1 Microstructural Targets for Ceramics

Roger Morrell

National Physical Laboratory, Teddington, Middlesex, U.K.

List of Symbols and Abbreviations 2

1.1 Introduction 3
1.2 Controlled Porosity 4
1.2.1 Macroporous Bodies 4
1.2.2 Microporous Bodies 5
1.2.3 Nanoporous Bodies 6
1.3 Mechanical Strength at Room Temperature 6
1.3.1 Young's Modulus 7
1.3.2 Size of Flaw 7
1.4 Fracture Energy 9
1.5 Resistance to High-Temperature Deformation 13
1.6 Resistance to Thermal Shock 14
1.7 Hardness and Wear Resistance 14
1.7.1 Hardness 15
1.7.2 Sliding Wear Resistance 15
1.8 Thermal Conductivity 16
1.8.1 Enhanced Thermal Conductivity 16
1.8.2 Minimised Thermal Conductivity 17
1.9 Thermal Expansion 17
1.10 Optical Functions 19
1.10.1 Transparency and Translucency 19
1.10.2 Colour 19
1.10.3 Emissivity 19
1.10.4 Special Optical Functions 19
1.11 Specific Electrical Functions 20
1.12 Magnetic Functions 22
1.13 Resistance to Corrosion 22
1.14 Joinability 22
1.15 Concluding Notes 23
1.16 References 23

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. Allrightsreserved.
2 1 Microstructural Targets for Ceramics

List of Symbols and Abbreviations

A constant
c size of crack
E Young's modulus

y{ toughness or work of fracture

af ultimate failure stress

YAG yttrium aluminium garnet

 1.1 Introduction

1.1 Introduction - control of thermal conductivity,

Commercial manufacture of ceramic - control of thermal expansion,
products is only undertaken against a - optical functions,
number of targets, including: - specific electrical functions,
- specific magnetic functions,
- properties of the material,
- resistance to corrosion,
- performance of the product,
- size and shape of component, and re- representing the principal technical perfor-
quired tolerances, mance criteria by which the suitability of
- cost of the product. ceramics is judged for particular applica-
The first two of these are determined tions. The assumption is made that we are
principally by the chemistry employed and concerned with polycrystalline ceramics
the microstructure achieved, while the lat- produced from powders or other appropri-
ter two are questions of choice of plant, ate fabrication routes, not single crystals.
reliability of manufacture, and the overall Clearly, chemistry is all-important in
cost-effectiveness of the process. Most defining the phases obtained within a
ceramics, with the principal exceptions of ceramic, and hence the properties dis-
tableware and decorative chinaware (al- played, but we will introduce chemistry
though some might dispute this exclusion), only by way of examples to illustrate par-
have technical functions for which particu- ticular types of product where optimisa-
lar sets of properties are required. To tion of microstructure has been achieved in
achieve these requires in turn, the selection technical products. In this way we can dis-
of chemistry, choice of raw materials, and cuss the principles uncluttered by excessive
optimisation of processing towards an ide- and diverting practical detail.
al target microstructure. This opening Before dealing with each one of the
chapter is devoted to a discussion of what above targets, it is necessary to look at the
the target microstructure should be. To basic characteristics of ceramic micro-
some extent, this is a very open-ended structures. When one takes an oxide or
issue, and is difficult to generalise, but there non-oxide powder or powders, forms a
are some guiding physical principles which shape from them and then consolidates the
are usefully described as a starting point shape by a high-temperature process, the
for the detailed formulation of materials. individual particles of the powder mass are
In order to do this, this chapter consid- encouraged thermodynamically (and, if
ers the targets from the point of view of the hot-pressing is applied, mechanically) to
following performance requirements or rearrange themselves and to join together
desirable properties: to form a solid body, usually by the move-
ment of ions, and with or without the pres-
- controlled porosity, ence of a liquid phase to assist the process.
- mechanical strength at ambient temper- The final result is seldom the classical reg-
ature, ular array of roughly equal-sized polyhe-
- toughness, dral grains (Fig. 1-1), but is controlled by
- resistance to deformation or creep at many factors, including:
elevated temperatures,
- thermal shock resistance, - the bulk chemical nature and overall
- hardness, sliding wear or abrasion resis- composition of the powder(s),
tance, - the surface chemistry of the powder(s),
 1 Microstructural Targets for Ceramics

results in the presence of residual porosity

in the microstructure. Its complete elimi-
nation is limited by the ineffectiveness of
sintering and grain growth processes,
which are trying to reduce the total surface
area of both exposed surface and grain
boundaries. While complete elimination of
pores is a desirable target for some me-
chanical properties, it is not for some oth-
ers, such as filters, membranes, catalyst
Figure 1-1. Typical single-phase, polycrystalline ce- supports, and thermal shock resistant and
ramic structure comprising regular shaped polyhe- low thermal conductivity materials for a
dral-shaped grains. variety of applications. However, the pow-
der process allows us to control porosity
by deliberately avoiding its removal in the
- the particle morphology (surface area, consolidation process.
particle size and shape),
- the shaping process and the pressure
used, 1.2.1 Macroporous Bodies
- the thermal cycle employed to consoli- Packing of irregular similar-size parti-
date the body, cles leaves a large volume fraction of pore
and may be described by the following mi- space with irregular-shaped channels for
crostructural features: permeation of fluids or for retention of air.
This principle has been used for many
- the chemistry of each phase, years for controlled filtration and for ther-
- the size and shape of each phase, mal insulation. All that needs to be done to
- the preferred orientation of each phase form a rigid ceramic filter body (e.g., Muil-
(texture), wijk and Tholen, 1989) is to bond the large
- the degree of elimination of porosity ini- particles together where they contact using
tially present as the unfilled gaps in the a ceramic glue, which can be amorphous
shaped body before firing (the "green" or crystalline in nature (Fig. 1-2 a). The
shape). amount of bonding phase controls the
This complex list of variables in real life strength of the material, but in small quan-
gives the ceramic manufacturer an almost tities does not affect the filter characteris-
inexhaustible supply of options for making tics.
products. It is the achievement of selected An interesting example is a ceramic dust
microstructural features that allows ceram- filter for power or incineration plant (Mor-
ics to find their applications, and thus to rell et al., 1990). A coarsely porous body is
create business. used as the filter wall to give a low pressure
drop across the thickness of material,
which obviously has to be adequately
1.2 Controlled Porosity strong to support itself mechanically under
the service conditions. So that fine parti-
The use of powder processing technolo- cles are trapped, a surface coating may be
gy for the fabrication of ceramics usually applied which has finer pores. This coating
 1.2 Controlled Porosity

additional porosity is included by adding

granular combustible or other fugitive or-
ganic material, such as sawdust, to the
green body.
Another example of a macroporous
body is a metallurgical filter (Minjolle,
1990). The main ceramic material is de-
signed to be dense, fine-grained and strong
after consolidation, but the controlled
pore or channel size is introduced typically
by using a plastic-based foaming technique
for green state shaping. The ceramic pre-
cursor powder is retained in the walls of
the foam, especially in the network of thick
ligaments between each void in the foam.
On removal of the plastic and consolida-
tion of the ceramic, a controlled channel
size remains (Fig. l-2b). Channel sizes
from a fraction of a millimetre, up to sever-
al millimetres, can be made in this way. By
appropriate choice of foaming agent, both
open cell and closed cell structures can be
developed, the latter having potential val-
Figure 1-2. Cross-sections of controlled open-porous
ue for low thermal conductivity.
ceramic structures for filters: (a) fabricated from ran-
dom packing of large grains bonded by a small pro-
portion of glass or ceramic bond (dark); (b) fabricated
by foaming a fine powder mix and sintering to give a 1.2.2 Microporous Bodies
dense skeletal structure.
In contrast, microporous bodies are
normally prepared simply by not allowing
traps the first layer of dust, which then acts the consolidation process to proceed to
as its own filter. greater than about 90% of the full theoret-
The easiest way of controlling the pore ical density. Such products retain open
size is by controlling the particle size of the pore channels through the structure, while
precursor powder. A narrow size distribu- at higher densities, residual pores are no
tion of large particles can be assembled longer interconnected. For electrical or
with some fine particles which on firing act mechanical applications this would be un-
as a bond to join the large particles togeth- desirable; penetration of water or dirt can
er. The channel size left between the large represent limitations in service; but if
particles is typically half the large particle porosity is controlled in size by the nature
size. In the dust filter example, the pressure of the precursor powder, the product can
drop across the filter body is controlled by be used to make particulate filters.
the channel size, which is controlled by the Sometimes, porosity of this type is ad-
starting large particle size of the ceramic ventitious, especially in conventional re-
body, not the bonding phase. In the cases fractory products (Chesters, 1973). Large
of insulating bricks and grinding wheels, grains and refractory bonding phases are
 1 Microstructural Targets for Ceramics

needed to ensure minimal dimensional 1.2.3 Nanoporous Bodies

change on consolidation, and dimensional
If controlled porosity at submicrometre
stability and retention of strength at very
levels is mandatory, a different strategy is
high temperatures. These large grains tend
required. Very small pores tend to be elim-
to restrict the elimination of the pore space
inated very rapidly in the powder consoli-
in the shaped green body because they are
dation process. It is then usually necessary
not as sinter-active as very fine particles
to consider specialized methods of obtain-
with large surface area, and as a result,
ing the required pore structure, and to
most bulk refractories have open porosity.
counter the thermodynamically driven ten-
However, this is of considerable advantage
dency for their removal. This usually calls
in controlling thermal stress damage resis-
for processing tricks to generate appropri-
tance. The presence of pores reduces the
ate structures at low temperatures where
elastic stiffness of the material, so less ener-
pore elimination rates are much reduced.
gy than in a fully dense body is available
For example, the process for forming sili-
from thermal strains to cause cracking
con nitride by the nitridation of silicon (the
damage. In addition, pores may tend to
reaction bonding process) can be used to
pin cracks, and such enhanced resistance
make very finely porous bodies by careful
to damage is much sought after by refrac-
control of the starting silicon powder size
tory engineers.
and homogeneity. Other routes can be
A useful attribute of many oxide-based used to make thin membranes with con-
ceramics is low thermal conductivity. trolled pore size, such as by applying a very
However, the thermal conductivity of air fine powder surface layer to a pre-sintered
in the absence of convection is even lower, coarsely porous body (Van Praag et al.,
and thus the trapping of air inside a low 1989), or by using chemical vapour deposi-
conductivity solid is the chief route for ob- tion partially to seal coarser open porosity,
taining improvements in thermal insula- or by anodising aluminium to form partic-
tion (Budnikov, 1964). However, there is a ular structures in aluminium oxide.
limit to how low the density of a solid body
On an even finer scale, members of the
formed from powder particles can be be-
zeolite family have crystal structures with
fore the strength is insufficient for practi-
defined "atomic" tunnels in them, which
cal use. Lower densities are better achieved
allow the passage of small gas molecules,
by using either hollow spheres (bubble alu-
but not large ones (Breck and Anderson,
mina, glassy cenospheres from fly ash) or,
1981). Zeolites can be formed into ceramic
much more commonly, fine fibres, which
bodies, or can be used as a coating on a
are now very widely used for ambient and
porous substrate. The chemical composi-
high-temperature thermal insulations. For
tion is used to control the structure type,
near-ambient temperature use, even better
and hence the tunnel size.
properties can be achieved from so-called
"aerogels" (Gowda and Harrison, 1987;
Fricke and Caps, 1988; Fricke and 1.3 Mechanical Strength
Emmerling, 1992), which are basically very at Room Temperature
low density skeletal forms of silica and
other oxides, which need encapsulating The mechanical strength of ceramic
within rigid boundaries in order to retain products is controlled by microstructural
them in place. defects which, if not already in the form of
 1.3 Mechanical Strength at Room Temperature

small cracks produced during microstruc- 1.3.2 Size of Flaw

tural development or by external damage,
form weak points or regions from which a Flaws or defects in ceramics take many
crack can propagate when subjected to a forms, from large scale voids or delamina-
sufficiently high stress level. The optimisa- tions several millimetres in length to
tion of strength in any microstructure thus micrometre sized intentional features of
requires the minimisation of the size and the microstructures. The larger the flaw (or
number of such defects or flaws by atten- its effective zone of influence) the lower
tion to processing conditions. This target will be the strength. Thus in generating a
applies irrespective of any additional ma- microstructure, strength can only be opti-
terial modifications designed to enhance mised by removing as many flaws as possi-
strength and/or toughness further. ble, but especially the largest ones. Very
The weakest link concept of brittle frac- small flaws can be considered as having an
ture in ceramics (Davidge, 1979) means insignificant role in determining strength
that the ultimate failure stress, c f , is deter- while large flaws remain. Table 1-1 shows
mined by the size of crack c and the tough- origins of and the typical scale of various
ness or work of fracture, yi5 of the material types of defects or flaws.
at the instant of failure: The size of defect that can be tolerated

"-m
where E is Young's modulus and A is a
constant determined by the geometry of
(1-1)
in a ceramic product depends on the pur-
pose of the product and the strength level
required for that purpose. For materials
where only intermediate to low strength is
required, such as in refractories, often
the crack in relation to the direction of strength tends to be controlled by the ef-
stressing. In order to maximise the fectiveness of bonding between large parti-
strength, it is desirable to maximise E and cles, with factors such as particle size and
yi9 whilst minimising c. Let us consider total porosity (0-40%) playing a lesser
each of these contributory factors. role. Unintentional defects, such as foreign
contaminants, which play a major role in
strength control in strong materials, can
often be safely ignored, especially in
1.3.1 Young's Modulus
coarse-grained products.
The Young's modulus of ceramics is de- Where medium levels of strength are re-
termined by two factors, the chemistry of quired (typically 100-400 MPa), it is es-
the crystalline and amorphous phases, and sential to remove the larger defects, but
the level of total amount of porosity. In small ones can usually be tolerated. For
order to maximise E, it is necessary to se- example, in high-alumina ceramics with
lect crystalline phases with inherently high flexural strengths of typically 300 MPa,
modulus and, particularly, to reduce strength and elastic modulus can be en-
porosity levels as far as possible. The for- hanced by removing as much of the total
mer of these options may clearly be re- porosity as possible, but usually a few per-
stricted by other performance attributes cent remains. Strength tends then to be
required of the material, but the latter is a controlled by the grain size rather than by
desirable target in the majority of circum- pores less than about 10 jim across, unless
stances. they form groups (e.g., resulting from a
8 1 Microstructural Targets for Ceramics

Table 1-1. Flaws - their size ranges and origins.

Type of flaw or defect Typical size range Origin

Delaminations 1-100 mm Problems in pressing or removal of binder;

drying shrinkage
Foreign organic matter 0.1-5 mm Contamination in powder handling
Foreign inorganic matter 0.01-1 mm Contamination in powder handling
Green machining surface faults 0.01-0.05 mm Binder not strong enough
to withstand machining
Aggregates causing voided regions 0.03-1 mm Powder binder mix too hard,
aggregates not crushed by pressing
Individual pores 0.001-1 mm Inhomogeneity of compacted powder;
lack of completeness of sintering
Large grains Microstructure Selection and processing of raw materials;
scale dependent chemical formulation; fired too hot
Machining damage 0.001 -0.5 mm Grinding with too coarse a diamond grit
or too great a depth of cut
Grain size 0.0005-1 mm Chemistry and physics of powder;
firing conditions

porous seam or aggregate). It may not be strength should not be ruined by inade-
worthwhile trying to remove a small quate attention to machining procedures
amount of fine-scale porosity. which generate surface flaws. In fact, the
For ultra-high strength (typically latter may be the fundamental limitation in
> 400 MPa), the defect or flaw size must any ceramic product, either in shaping to
be reduced as far as possible by the appro- the final size and tolerances, in handling,
priate attention to selection of raw materi- or in service.
als and their processing conditions. It is The mechanical integrity, or for that
imperative to minimise the risk of contam- matter many other functional attributes,
ination by adopting clean processing con- of advanced types of ceramics when pro-
ditions such that no foreign bodies enter cessed to avoid extraneous defects is thus
the batch. Even mill-ball debris has to be essentially determined by a combination of
avoided, and consideration should be giv- pore size and grain size. To obtain uni-
en to using the fired product as the wear- form, consistent performance, a uniform
resistant surfaces in processing equipment. grain size with a minimum of porosity
If binders are added, they should not form (preferably with pores rather smaller than
hard agglomerates which do not crush on the grains) is the practical target in many
shaping the green body, or which do not circumstances. Even then, the grains them-
shrink as much as surrounding material on selves will act as the fundamental limita-
firing. Chemistry must be controlled to tion. A grain boundary can act as a flaw,
achieve close-to-theoretical density with and the following are examples:
minimum residual porosity whilst main- 1. In some cases, the flaw will be present
taining control on the maximum grain size. as a crack in unstressed material, simply as
Finally, the achievement of high intrinsic a result of differences in thermal expansion
 1.4 Fracture Energy

coefficients between contacting grains 3. Surface machining by abrasives re-

which do not have the same crystallo- moves material by a ploughing and/or
graphic orientation, or between grains or chipping process that requires the forma-
grain boundary phases of different chemis- tion of microcracks. These can run into the
try or crystallography (Davidge, 1982) bulk of the material to a depth controlled
(Fig. 1-3). This can be shown to occur by the forces applied and by the toughness.
when the elastic energy associated with the It is thought that one factor limiting depth
thermal expansion mismatch exceeds that of damage may be the grain size, provided
required to fracture the boundary. Exam- that this is greater than about 5 jim. There
ples include coarse-grained alumina, alu- have been a number of demonstrations in
minium titanate, magnesium dititanate the literature on dense, single-phase ce-
(Kuszyk and Bradt, 1973), and beta-eu- ramics in which the flexural strength of
cryptite (Li 2 O A12O3 -2SiO2). The effect machined test-bars is proportional to c~1/2
can be minimised by reducing grain size, as predicted by the Griffith equation (1-1)
although this may not always be possible above (e.g., Davidge, 1979). This suggests
whilst retaining other desirable properties that damage is limited in depth by the
(e.g., low expansion coefficient in alumini- grain boundaries that the microcracks
um titanate). meet. Below about 5 jim grain size, dam-
2. Phase changes in the ceramic during age seems more likely to jump several
cooling after firing can in some circum- grains, and the c~1/2 relationship tends to
stances produce microcracks between disappear.
grains. A classic example is the quartz In summary, the first microstructural
transition seen in conventional siliceous target for strength should be the reduction
porcelains (Kingery et al., 1976), but it in size and amount of all types of defect,
also occurs with unstabilized zirconia used including pores, large grains, foreign bod-
as a reinforcing agent in many different ies and machining damage. The second
types of advanced technical ceramic. Opti- should then be the control of phase con-
misation of strength clearly requires such tent, grain shape, and the use of any rein-
phase changes to be controlled or sup- forcement that can conveniently and
pressed. simultaneously enhance the toughness.
This is discussed below.

1.4 Fracture Energy

The fracture energy of ceramics is close-

ly related to the phase composition, and to
some extent the grain structure (Davidge,
1973, Evans, 1988). Glasses, having fea-
tureless microstructures, show very little
resistance to the propagation of small
cracks, and thus possess low toughness.
Figure 1-3. As Fig. 1-1, but with microcracks on cer-
Substantially crystalline microstructures
tain boundaries caused by thermal expansion mis- which tend to fracture across the grains
match. (transgranularly) (Fig. 1-4 a) also show
10 1 Microstructural Targets for Ceramics

in its path. In ceramic/metal systems, such

as hardmetals or cermets, a metallic phase
bonding ceramic particles together is used
to ensure that fracture involves plastic de-
formation of the metal phase, which ab-
sorbs considerable amounts of energy. In
fact, ligaments of metal can often be seen
across the fracture (Fig. 1-5). However, in
an all-ceramic, i.e., all-brittle system, we
(a) cannot employ plastically deforming liga-
ments. Enhancement of toughness requires
absorption of energy by increasing the de-
viation of the crack from a planar condi-
tion. We already have some roughness if
the microstructure naturally cracks inter-
granularly, but we should seek to encour-
age either further deviations of the fracture
surface, or the generation of more damage
over a wider zone than just the immediate
crack plane, or we can make it more diffi-
cult for the crack to open by ensuring brit-
(b)
tle ligaments remain uncracked or wedged
Figure 1-4. Brittle fracture through the structure across the fracture (Fig. 1-6), or by grow-
shown in Fig. 1-1: (a) transgranular fracture (strong ing elongated grains. Reinforcement using
grain boundaries, weak grains); (b) intergranular frac- strong whiskers (Becher etal., 1986) is
ture (weak grain boundaries, strong grains). probably the most effective. This concept
is shown schematically in Fig. 1-7.
low toughness compared with those which There has been much research in recent
tend to fail along grain boundaries (inter- years into such toughening mechanisms,
granularly) (Fig. l-4b), because in the for- with a better understanding being devel-
mer case, grain boundaries offer little if
any resistance to the propagating crack
while in the latter case, the propagating
crack meets added resistance as segments
of it are diverted in fresh directions when
they reach the end of one grain facet. This
is a simple view of the typical situation in
a polycrystalline material. In practice, the
microstructure may be manipulated both
in chemistry and in physical arrangement
of phases to maximise the toughness
(Clarke, 1992), undoubtedly topics of con-
Figure 1-5. Toughening the microstructure shown in
tinuing research for many years to come.
Figure 1-1 by ductile metal particles which bridge the
To impede the development of a small crack faces. This effect results in rising crack resis-
flaw into a crack, obstacles must be placed tance with increasing crack length.
 1.4 Fracture Energy 11

Another method of increasing apparent

strength and toughness is by the use of an
expansile phase transformation. The now-
classic example (Green etal., 1989) is the
phase transformation in zirconia from the
tetragonal to the monoclinic form on cool-
ing compositions which have inadequate
stabilisers for full stability. If the particle
size is sufficiently small, the phase trans-
formation can be suppressed by the re-
Figure 1-6. As Fig. l-4b, but with incomplete separa- straining effect of the surrounding matrix.
tion of crack faces caused by residual bridging grains. Thus, in alumina with small dispersed zir-
This effect results in rising crack resistance with in-
conia grains, the alumina normally pre-
creasing crack length.
vents the transformation, but close to
cracks or a free surface, the transformation
is unimpeded (Fig. 1-8), and this has the
effect of placing such surfaces into com-
pression. Any attempt to propagate cracks
thus has to overcome the compressive
stress, and this tends to enhance the appar-
ent toughness. Strength is also improved
because the compressive zone tends to re-
strict the development of cracks from bulk
and surface defects. Some of the strongest
materials available (crf > 1 GPa) rely on
Figure 1-7. As Fig. l-4b, but with whisker reinforce- this principle. At the other end of the mi-
ment or acicular grains which bridge the crack and crostructural scale, the strength and frac-
resist propagation, giving a rising crack resistance ture resistance of coarse-grained refracto-
with increasing crack length. ries can also be improved by this principle.
Even if the transformation of the particles
has already occurred and the structure is
oped. However, the extent of reinforce-
ment by these methods is seldom more
than twice that the untoughened structure,
with the single exception of reinforcement
by long fibres (Marshall and Evans, 1986).
In addition, the introduction of micro-
structural features large enough to produce
an increase in apparent toughness often
results in "defective" microstructures with
larger intrinsic defects. This may result in
damage tolerance, but on balance, ulti-
mate strength is seldom improved, and of- Figure 1-8. As Fig. l-4b, but with zirconia particles
ten decreased compared with a fine- incorporated, which transform close to the crack sur-
grained unreinforced structure. faces, placing the crack tip into compression.
12 1 Microstructural Targets for Ceramics

microcracked, the apparent toughness can

still be improved because the moving crack
is diverted and pinned by the particles
(Fig. 1-9).
Recently, interest has developed in so-
called "nano-composite" ceramics in which
very fine scale particulate reinforcement of
the microstructure is developed (Nakahira
etal., 1989; Niihara, 1992). There are a
number of possible options for the mi-
crostructure (Fig. 1-10) but, typically, hard Figure 1-9. As Fig. 1-8, but with larger zirconia parti-
particles of size between 50 and 200 nm in cles which have already transformed as the ceramic
a concentration of typically 10 to 20% by was cooled from the firing temperature. As the crack
volume need to be distributed throughout approaches a particle, the matrix tends to microcrack
and absorb additional energy.
the matrix phase, both in the bulk and at
grain boundaries of the matrix phase. Ide-
ally, in order to remain effective the precip-
itate should not grow with the matrix
phase on firing. Limited demonstrations
have shown that very high strengths can be
achieved in materials such as SiC particu-
late reinforced alumina. While there has
been insufficient evidence yet published,
the concept is one of trying to limit the
growth of crack-like defects by fine-scale
obstacles which tend to pin them by virtue
of local compressive stress fields resulting
from thermal expansion mismatch. The
major problem is the achievement of the
required microstructure by processing
methods which avoid the agglomeration
typical of sub-micrometre powders. In the
example cited above, enormously extended
high-energy attrition milling has been re-
quired to generate the required dispersion
of the reinforcing phase in a powder of the
matrix phase, which then needs to be hot-
pressed to densify it whilst restricting grain
growth. The cost-effectiveness of such pro- (b)
duction methods is currently uncertain. Figure 1-10. Two examples of several concepts
Future developments need to concentrate providing nano-phase reinforcement: (a) with nano-
on methods of incorporating the reinforc- sized particles which have segregated to and grown at
grain boundaries, helping to resist grain boundary
ing particles directly into the matrix pow- fracture; (b) with nano-sized particles which have re-
der as it is made, rather than trying to mix mained widely dispersed throughout the grains, al-
them in later, so this requires an integrated lowing improved resistance to transgranular fracture.
 1.5 Resistance to High-Temperature Deformation 13

approach to powder production and mi- upper limits to the temperature that even
crostructural design. Ideal targets for the the most refractory system will tolerate,
microstructure are uncertain and probably posed by the natural tendency of ions to
phase composition dependent, but reten- diffuse within crystals provided that the
tion of a uniformly distributed nanometre- temperature is high enough.
scale reinforcing phase is clearly a major As a general rule, crystalline phase are
one. more resistant to permanent deformation
It is clear from the discussion so far that, than amorphous ones. So our first target
with the exception of long fibre reinforce- should be the elimination of amorphous
ment, improvements in toughness tend to phases (Lange, 1978). However, this is eas-
be rather limited. Toughness is determined ier said than done because many systems
principally by the spatial distribution of contain enough impurities to form a segre-
the phases of the material, and is thus de- gated secondary phase which resides at
termined by the system and the processing grain boundaries (Fig. 1-11). Since such
method employed. The target has to be amorphous phases usually wet the host
considered in terms of one or more of the crystalline phase well, the surface energy of
following: such a boundary is usually lower than that
of a crystal/crystal boundary, and so the
- deviation of the crack path,
amorphous phase becomes continuous
- pinning growing cracks in compressive
throughout the microstructure, even
stress fields,
though it may be present only in small
- crack bridging by uncracked ligaments, amounts. It then controls high-tempera-
- energy absorption by microcracking ture deformation by allowing sliding of the
around the main crack, grains. Very careful control of the purity of
- placing crack faces into compression, the system is needed to control the devel-
e.g., with phase transformations, opment of such phases. A classic example
without unduly compromising strength. is the progressive improvement over the
Most of the research work has been per- last thirty years in the refractoriness of
formed on materials for principally me- hot-pressed silicon nitride (Lange, 1983).
chanical use but there remain opportuni- Early versions were prepared with MgO as
ties to use such techniques on functional the hot-pressing aid, often in amounts as
electrical ceramics as well as to improve
mechanical properties and reliability.

1.5 Resistance to
High-Temperature Deformation
Resistance to high-temperature defor-
mation has to be seen in a rather different
light to that of strength at room tempera-
ture. Essentially, all microstructural fea-
tures of a contiguous nature need to be Figure 1-11. As Fig. 1-1, but typical microstructure
dimensionally stable and not deform by of ceramic containing continuous second phase which
any mechanism. Of course, there will be was liquid at the firing temperature.
14 1 Microstructural Targets for Ceramics

high as 5mass%, resulting in the forma- 1.6 Resistance to Thermal Shock

tion of an MgO-SiO2 glassy phase which
took up other impurities in the system, It is difficult to offer a unique set of
such as iron oxide. Bulk deformation was microstructural targets for thermal shock
obtained at temperatures as low as 900 °C. resistance, because much depends on the
Attention to the purity of the silicon ni- definition of "resistance" to be employed
tride, to the formulation of the secondary (Hasselman, 1970). In dense, strong ce-
phase, and to heat treatment to encourage ramics subjected to severe rates of temper-
it to crystallise, has resulted in the develop- ature change or severe temperature gradi-
ment of materials which are resistant to ents, resistance to the onset of cracking is
deformation at temperatures as high as determined principally by the ratio of
1400°C. strength to elastic modulus, and thermal
In cases where the presence of a sec- expansion coefficient. The latter is deter-
ondary phase is essential, if only to allow mined by chemistry with only a small mi-
processing at reasonably low tempera- crostructural influence in certain cases (see
tures, a secondary target to achieve refrac- Sec. 1.9, below), but the former ratio can
toriness is to try to develop continuous be maximised by ensuring that strength is
grain/grain bonding by control of compo- maximised relative to elastic modulus in
sition and firing conditions. Some high- accordance with the discussion in Sec. 1.3.
alumina ceramics achieve grain/grain bond- If the principal criterion is resistance to
ing in the presence of a CaO-Al2O3-SiO2 damage during thermal shock, crack prop-
amorphous phase, and are resistant to de- agation can be impeded by enhancing
formation at 1300 °C at which temperature toughness at the expense of strength (Has-
the amorphous phase is quite mobile selman, 1969), as outlined in Sec. 1.4. Al-
(Morrell, 1987). ternatively, in less strong refractory-type
Alternatively, the grains can be grown materials, resistance to thermal shock
unusually large by firing to temperatures tends to be controlled by porosity and its
higher than normal. Deformation rates of size distribution, as discussed in Sec. 1.2.
grains themselves, or even local accommo- Microstructural targets are therefore
dation rates at grain junctions, are general- rather varied, and are more likely to be
ly related to an inverse power of the grain fixed by other requirements of the product
size. Increasing the grain size thus reduces than thermal shock resistance.
the deformation rate. Coarse-grained alu-
mina is used for high-temperature work-
tubes and thermocouple insulators to tem-
peratures much in excess of the capability 1.7 Hardness and Wear Resistance
of fine-grained versions. Heavy refracto-
ries containing large grains of silicon car- Ceramic materials are generally harder
bide, magnesite, alumina, mullite, etc., are than most metal alloys, and thus find
much more stable dimensionally than fine- many applications in engineering where re-
grained equivalents, provided that they are sistance to erosion or sliding wear is of
well bonded by an equally refractory mate- prime concern. As with thermal shock re-
rial. sistance it is difficult to offer a unique set
of microstructural targets aside from the
choice of chemistry and crystallography.
 1.7 Hardness and Wear Resistance 15

1.7.1 Hardness thread guides. A running thread can be

extremely abrasive, because of the use of
Hardness is usually measured by some
nano-sized hard particle fillers; it slowly
form of indentation test which records a
removes material from the surface of the
penetration under a localised force. The
guide. When the contact conditions are es-
resistance to penetration by an indenter
sentially inert, the mechanism of removal
tends to be controlled more by the crystal-
is probably slow attrition at the atomic
lography of the material than the mi-
layer level. If the environment is such as to
crostructure itself, although it can be read-
react with the contact surface under the
ily demonstrated that elimination of
enhanced contact temperature and pres-
porosity is a desirable target in order to
sure conditions, removal rates can be
maximise hardness of a given material
much enhanced. Thus, silicon carbide
type. Thus, hot-pressed, fully-dense silicon
should be extremely resistant to this type
nitride is much harder than finely open-po-
of wear, but in practice, there is air present,
rous reaction-bonded silicon nitride of es-
and this produces a thin, probably
sentially the same composition and phases.
nanometre thick, layer of silica on the sur-
Also, glassy phases usually allow greater
face. This layer is much softer than the
deformation than crystalline ones, and
silicon carbide, and is much more readily
their elimination, or at least substantial re-
removed. Oxygen in the atmosphere then
duction in volume fraction, can also be
re-oxidises the silicon carbide, and a rela-
considered to be a target. An example is in
tively high wear rate results compared with
high-alumina ceramics; commercial mate-
expectations on the basis of hardness. It is
rials exist where the secondary, usually
for this reason that non-oxide ceramics
glassy, phase content can range between 0
have not found applications as thread
and 40 % by volume, with a corresponding
guides (Ramsey and Page, 1989). A further
wide range in hardness values (Morrell,
point related to thread-guides is the need
1987).
to have a macroscopically smooth surface
However, it must also be borne in mind consisting of protruding smoothly round-
that hardness, although often an advan- ed grains which minimises the contact area
tage for wear resistance, by itself does not with the thread. This surface microstruc-
control wear resistance, which is one of the tural target is usually achieved in oxide
most complex "properties" of ceramics. ceramics by shaping and smoothing to fi-
nal dimensions, and then re-firing to en-
1.7.2 Sliding Wear Resistance courage "thermal relief to develop. The
When two bodies slide over each other detailed bulk microstructure is not rele-
they contact at a small number of asperi- vant, but its ability to form this "smooth"
ties where contact pressures are very high, surface is critical.
and high temperatures can be generated The existence of a chemically modified
locally. Resistance to removal of material layer under wear conditions does not al-
from the two surfaces under such condi- ways produce an enhanced wear rate (Gee
tions depends on the mechanical and et al., 1989). In some cases of sliding wear,
chemical conditions experienced. Hard- the presence of a stable "tribochemical"
ness alone will not prevent wear, if this is film can help to lubricate the interface by
enhanced by some form of chemical reac- keeping the wearing bodies apart or by
tion. A simple example can be found in increasing the effective contact area, so re-
16 1 Microstructural Targets for Ceramics

during friction. The production of stable critical, and is discussed at greater length
films is clearly related to the microstruc- in Sec. 1.13.
ture, especially the distribution of species Under conditions of fixed or rolling
which produce the tribochemical film. abrasive wear or of erosion by impacting
Some examples include: hard particles, hardness and fracture tough-
ness tend to be the controlling parameters
1. Alumina ceramics in humid conditions. (Moore and King, 1980). High hardness
A film of hydrated alumina is produced resists the penetration of abrasive particles
at ambient temperature. into the material, or limits the ability to
2. Lubrication at elevated temperatures. plough grooves, while high toughness lim-
The presence of a glassy phase in suffi- its the extent of cracking that emanates
cient quantity produces a liquid film lu- from such contacts. In both cases, these
bricant, such as in some high-alumina properties limit the material removal rate.
ceramics. Alternatively, oxidation of a Aside from choice of crystalline phases,
non-oxide such as silicon nitride can microstructural targets are thus the re-
produce such a film (Gee et al., 1989). moval of porosity and soft secondary
3. Titania or chromia in alumina. Lubri- phases, and the optimisation of toughness
cious films of the respective oxides help in ways which do not compromise density
to reduce friction. The species can be and strength.
included in the bulk material or inserted
later by ion implantation.
4. Incorporation of obvious lubricating
media into the microstructure. Some
1.8 Thermal Conductivity
examples include graphite, molybdenum
disulphide, and boron nitride. Control of thermal conductivity very
definitely involves microstructural targets.
Thermal conductivity is a property deter-
In many industrial sliding wear condi-
mined by the crystallography of the phases
tions, such as in a chemical plant pump
present, and by their physical arrange-
shaft seal, contact loads are high. Seal
ment. The property targets may be very
counterfaces tend to be prepared by lap-
high or very low thermal conductivity, and
ping flat to a couple of wavebands of light,
each requires a different approach.
and are usually run against a carbon seal.
For seal materials, the critical microstruc-
tural targets are (1) the ability to lap or 1.8.1 Enhanced Thermal Conductivity
polish to a good-quality finish, and (2) the Apart from the choice of crystalline
choice of appropriate composition and phases with intrinsically high thermal con-
phases to withstand the chemical environ- ductivity, such as A1N, BeO, SiC, or vari-
ment under wear conditions. The former is ous borides (paying regard, of course, to
not difficult to achieve in most types of the strong temperature dependence of
ceramic, because a small percentage of thermal conductivity in many ceramic ma-
porosity does not impede the performance. terials), and the removal of most of the
Provided that the material has adequate porosity, attention has to be given to
strength and hardness there is no real means of fabricating such materials. Usu-
problem in achieving this target in many ally, some secondary chemical species need
material types. The latter target is more to be added to encourage sintering and to
 1.9 Thermal Expansion 17

control grain size. These species will tend low thermal conductivity. However, only
to have lower thermal conductivity, or limited porosity can be incorporated dur-
produce other phases with lower conduc- ing the flame spraying process, otherwise
tivity. Their minimisation or removal is a the coating will not have sufficient me-
principal microstructural target. chanical integrity to survive ablation by
Thus, silicate phases in BeO drastically hot gases. The development of some lentic-
reduce thermal conductivity, especially at ular pores during spraying is considered to
temperatures below 200 °C, and need to be be as far as it is possible to go in reducing
avoided when possible. In the case of alu- thermal conductivity.
minium nitride, there has been much re- As a general rule, fine-scale pores are
search in the last decade reliably to en- more effective at reducing thermal conduc-
hance thermal conductivity above that tivity than larger ones, since the latter may
achieved by BeO. The principal problems allow some convective effect by the con-
are that A1N is difficult to keep oxygen- tained gas. The more pores in the struc-
free, and that a sintering aid is needed for ture, the lower is the thermal conductivity.
processing to a dense body, both undesir- The maximum level of porosity achievable
able factors for optimisation of thermal in a handleable rigid body is probably
conductivity, with the microstructural about 80%, below which it becomes rather
targets being their removal (Riissel et al., friable unless in the form of fibres (Pratt,
1991). 1969). Fibre blankets, mats or vacuum-
In one particular case, reaction-bonded formed shapes can have very low density
silicon carbide using the silicon infiltration and very low thermal conductivity, bet-
process, the secondary phase is residual tered only by special forms of powders and
silicon which has a very high thermal by materials such as aerogels, highly
conductivity. Although undesirable for fragile dried aerated gel structures that
strength, toughness and resistance to ele- need to be contained within a rigid vessel.
vated temperatures, increasing the amount
of free silicon can greatly enhance the ther-
mal conductivity of this product. 1.9 Thermal Expansion
At first sight there might not seem to be
1.8.2 Minimised Thermal Conductivity
many opportunities to influence thermal
In this case one would choose materials expansion behaviour in ceramics through
of intrinsically low thermal conductivity, microstructural control, but there are a
and then consider the introduction of as number of possibilities that can be exploit-
much porosity as possible leaving a mini- ed for particular purposes:
mum of solid contact path through the ma- /. Surface compression. Strengthening of
terial. Naturally this has to be done as a ceramic components can be achieved by
compromise with other requirements that placing the surface into deliberate com-
are also controlled by the level of porosity, pression. In the case of partially stabilized
such as strength or hardness. An example zirconia, a phase change is used, but in
of this problem is found in thermal barrier others the development of different lower
coatings for jet engine parts. The first expansion phases at the surface can be of
choice of material sufficiently refractory advantage, such as cordierite or mullite on
for the task is zirconia with intrinsically the surface of alumina (Kirchner, 1979).
18 1 Microstructural Targets for Ceramics

The microstructural target here would be such cases, the ceramic can then be tailored
to control the local chemistry without in expansion coefficient to the metal.
affecting the bulk of the material. Glass-ceramic technology offers the widest
2. Ultra-low thermal expansion materi- range of possibilities, with expansion coef-
als. Many low thermal expansion products ficients ranging from — 2 x l O ~ 6 K ~ 1 to
are produced only by relying on the high 2 0 x l 0 ~ 6 K ~ 1 , by employing silicate
degree of crystalline anisotropy in ceramic phases, or even phase transitions of silica
phases causing microcracking on cooling itself, e.g., the a/p transition of cristobalite
from the firing temperature. Thus alumini- (McMillan, 1979). Provided that the grain
um titanate produced in very fine-grain size is very small, substantially sub-micro-
form would have a high expansion coeffi- metre, large mismatches between the indi-
cient if not allowed to microcrack. When vidual phases in the product can be tolerat-
microcracking takes place, the high expan- ed with no loss in mechanical properties. If
sion directions in each grain shrink away conventional ceramics are employed with
from grains in other orientations, and this rather large grain sizes, phase mixtures
reduces the effect such directions have on with widely mismatched expansion coeffi-
the net expansion coefficient. The result is cients should be avoided because mechani-
a material with poor strength (because of cal properties can be suspect.
the microcracks), but a very low expansion 4. Anisotropic thermal expansion. As in-
coefficient suitable for applications requir- timated above, a polycrystalline assem-
ing extremes of thermal shock resistance blage of grains with anisotropic expansion
(Buessem et al., 1952; Byrne et al., 1988). characteristics produces high levels of in-
A second example is in spodumene tergranular stress, which can lead to mi-
(Li2O • A12O3 • 4 SiO2) and eucryptite crocracking or worse. By the adoption of
(Li2O • A12O3 • 2 SiO2) ceramics. In this appropriate raw materials and processing
case the average expansion coefficient is techniques it is possible to orientate grains
low, but the anisotropy is so severe that in some materials into a common align-
useful products cannot be made by con- ment. In alumina ceramics, the use of platy
ventional ceramic processing (Gillery and alumina powder processed by extrusion
Bush, 1952). However, by using the glass- seems to align the grains, an alignment
ceramic process, the grain size of the crys- which is not lost during subsequent grain
tals of such phases grown from the precur- growth on firing. This would have the ef-
sor glass can be kept sub-micrometre, and fect of reducing local internal thermal
a reasonable strength can be retained stresses between grains, which has been
(McMillan, 1979). shown to correlate with increased mechan-
3. Matched thermal expansion. In many ical strength (Clinton etal., 1986). In an-
engineering applications it is necessary to other example from the literature on the
match thermal expansions of ceramics to thermal expansion characteristics of vehi-
other materials, particularly metals, and cle exhaust catalyst-support honeycomb
especially when vacuum-tight seals are to monoliths (Lachman etal., 1981), the
be produced. It is possible to control the choice of a particular platy clay in an ex-
expansion coefficient in metals as needed, trusion mix allowed the alignment of
especially in the nickel-iron-cobalt alloy cordierite grains developed subsequently
series, but often the properties of such on firing. The microstructure created has
metals are inadequate for the end use. In anisotropic thermal expansion which is
 1.10 Optical Functions 19

lower than the normal random polycrys- sons. Colours are typically generated by
talline value in the desired directions, i.e., impurity species present in the structure of
in the plane of the honeycomb walls. a material. Materials such as alumina and
In each of these cases, thermal expan- zirconia, normally white and transparent
sion characteristics, coupled with grain in single crystal form, are coloured by
size and orientation, have been exploited many transition metal or rare-earth species
through design of the appropriate mi- through the formation of colour centres. If
crostructure. If the target expansion char- colour needs to be controlled, the impuri-
acteristics are set, it is usually possible to ties in the product need to be controlled
select appropriate crystalline and/or glassy sufficiently to obtain reproducibility.
phases to meet the requirement by devel-
oping the appropriate microstructure. 1.10.3 Emissivity
The ability of a surface to absorb and
re-radiate energy is controlled by chemis-
1.10 Optical Functions
try and surface structure, and is usually a
function of both wavelength and tempera-
1.10.1 Transparency and Translucency
ture. Many ceramics show high emissivity
Many oxide ceramics and some non-ox- at low temperatures, decreasing with in-
ides are optically transparent, if not in the creasing temperature. If high emissivity is
visible range then in the infra-red range. needed at high temperatures, it is usually
Optical transmission characteristics are necessary to incorporate chemical species
controlled by the fundamental structure of into the microstructure which are thermal-
the phases of the product. For high levels ly stable and which confer the required
of transmission, radiation scattering is to properties. For example, significant in-
be avoided, and elimination of all porosity creases in high-temperature emissivity of
is essential. Thus, sodium vapour lamp en- white oxides can be achieved by additions
velopes are made from alumina of very of colouring oxides, such as oxides of tran-
high purity (to avoid optically mismatched sition metals. In addition, the surface tex-
secondary phases) and are fired in such a ture achieved, which is related to bulk mi-
way as to remove most of the porosity crostructure, can play a significant role. A
(Hing, 1976). Residual milkiness is a result surface which contains many accessible
of the optical anisotropy of each grain of pores has a higher emissivity than one
alumina. Crystallographically cubic mate- which is flat and featureless. Specific mi-
rials do not have such anisotropy, and are crostructural targets are difficult to define,
preferred for windows where focusing is and depend on other constraints, such as
required. Ideal materials are MgO, magne- refractoriness and operating atmosphere.
sium spinel and YAG (yttrium aluminium
garnet). 1.10.4 Special Optical Functions
Special functions, such as electro-optical
1.10.2 Colour
effects of change of optical polarisation
Colour, in many cases, is only a cosmet- under an applied voltage gradient, may
ic issue employed principally for product also involve microstructural targets, not
recognition. Only in a few cases is control only to achieve the appropriate optical
of colour actually needed for technical rea- transmission characteristics, but also the
20 1 Microstructural Targets for Ceramics

correct electro-optic function stemming trical ceramics. It is usually necessary to

from particular compositions (Haertling, add more of these species than is required
1988). Large grain size, zero porosity, and in the final product to compensate for the
avoiding segregation of dopants to the evaporation. If the evaporation is not ade-
grain boundaries are crucial. Control of quately controlled, too much or too little
overall composition and firing conditions may evaporate, and the microstructures
is especially needed for materials based on then contain respectively too little or too
lead oxide or barium oxide (see next sec- much of these species, changing the prop-
tion). erties. Ideal microstructures are difficult to
achieve, and there may be composition
gradients across sections more than a mil-
1.11 Specific Electrical Functions limetre or two thick.
The next important factor is the distri-
In order to produce a specific electrical bution of the phases and the requirement
function within a ceramic material the first for the presence or absence of small
priority has to be to develop the appropri- amounts of secondary phase. In low-loss
ate chemistry and crystallography that will dielectrics, control of purity is required to
produce the desired effect. Perhaps the on- minimise the amount of lossy phase, often
ly general microstructural target is minimi- a glassy grain boundary phase, which is
sation of porosity in order to achieve the allowed to develop. However, sometimes it
greatest mechanical reliability and the is better to let a glassy phase develop to act
greatest resistance to dielectric breakdown as a sink for impurities than to let them
(see below). All other targets are rather exist elsewhere in the microstructure. A
specific to the end function. useful example is high-alumina ceramics.
There are many instances where control Sodium as an impurity in the precursor
of properties is essential, and these are alumina powder tends to form the lossy
summarised in Table 1-2, together with an phase p-alumina. It is better to tie up the
indication of the microstructural means of sodium in a silicate glass with lower loss.
doing it. Mostly, the principal factor is the So the deliberate addition of components
chemical constitution of the primary phase, to make a glassy phase is often made. On
which may contain a variety of dopants to the other hand, if the sodium content is
achieve the desired function. Examples in- extremely low, this may not be necessary.
clude the use of dopants in barium titanate The target is controlled by economics.
dielectrics to control the permittivity, its Recent developments in piezoelectrics
temperature coefficient and the upper tem- have moved from all-ceramic bodies to ce-
perature limit of effective operation (Her- ramic/polymer composite bodies to im-
bert, 1982), and the use of calcium or stron- prove the electromechanical coupling coef-
tium dopants in lanthanum chromite to ficient (Shaulov et al., 1984; Clarke, 1992).
control electrical conductivity (St-Jacques In varistors (non-ohmic resistors) based
et al., 1974). In both of these examples an on zinc oxide, dopants are added to give
important factor to note in material for- the non-linear characteristics. The distri-
mulation is the risk of evaporation during bution of the dopants is important to the
firing. Species such as BaO, PbO and reliable functioning of the device, which
Cr 2 O 3 have significant vapour pressures at may be controlled by the degree of un-
the temperature needed for sintering elec- wanted segregation of species to grain
 1.11 Specific Electrical Functions 21

Table 1-2. Microstructural targets for electrical functions.

Function or property target Property range Microstructural targets

High relative permittivity >100 Control of composition,

control of domain size through grain size
Medium relative permittivity 6-100 No specific targets
Low relative permittivity <6 Selection of chemistry
Temperature coefficient - Control of phase composition employing
of permittivity a mix of negative and positive
(e.g., temperature compensated) temperature coefficients
Low dielectric loss tan<5< 0.001 Control of chemistry, especially impurities,
minimisation of glassy phases
High dielectric breakdown >10 Vmm" 1 Minimisation of porosity, minimisation
resistance of impurities, especially alkalis
High electrical resistance >1012Qmat25°C Minimisation of impurities, especially alkalis
Low electrical resistance < 10 Qm at 25°C Control of dopants and their distribution,
avoidance of glassy phases which
isolate grains
Non-linear resistivity Control of dopants and their distribution,
(varistors, thermistors, etc.) especially in grain boundary layers
Superconducting properties Zero resistance with Control of phase composition, impurities
high critical current at grain boundaries, and relative
orientation of grains
Ionic conduction Control of impurities, minimising segregation
(fuel cells, sensors, etc.) to grain boundaries while retaining dopants
to control defect concentration. Low porosity.
Control of grain size as compromise between
strength and resistance of grain boundaries
Piezoelectric properties Control of composition, fine grain size
(including pyroelectric and and minimised porosity for strength.
electrostrictive properties) Ceramic/polymer composites

boundary regions (Philipp and Levinson, shift and cause hysteresis in the character-
1983). istics and hence power loss.
Removal of porosity is a key target in In the particular cases of ionic and elec-
improved reliability in many electrical tronic conductivity, the optimisation of
functions by maximising mechanical and properties and improvement in reliability
dielectric reliability (Ward, 1989). This is can be complex, especially if the device is
particularly important for electronic sub- operating at elevated temperature. Factors
strates, electro-optic devices, ionic conduc- that need to be considered are impurities
tors, multilayer capacitors and piezoelec- required to give appropriate conduction
tric systems. Fine grain size is also an im- without their segregating to grain bound-
portant target in the control of dielectric aries where they may interrupt the conduc-
properties because it can control domain tion process (Seitz and Orlow, 1981). In-
size and the tendency of domain walls to creasing the grain size may reduce the
22 1 Microstructural Targets for Ceramics

number of boundaries involved in the con- growth. If this secondary phase is incor-
duction process, and thus reduce the rectly formulated, the corrosion resistance
boundary contribution to total resistance, may be completely inadequate because the
but this may result in mechanical weaken- corroding agent, such as a mineral acid, is
ing. Grain size therefore has to be chosen able to penetrate and thus weaken the
as a compromise. In addition, some pre- structure without necessarily attacking the
ferred orientation of grains may enhance alumina grains.
conductivity by lining up conducting This example illustrates the need to con-
planes across grain boundaries. In high- trol the composition and distribution of all
temperature superconductors, such align- phases in order to optimise corrosion resis-
ment is thought to be desirable for the tance. If a phase of poor corrosion resis-
achievement of high critical currents tance is necessarily present, a target could
(Clarke et al., 1989). be to ensure that the phase is isolated with-
in a corrosion resistant matrix of another
phase. The existence of continuous phases
1.12 Magnetic Functions with poor corrosion resistance, even as a
thin film along grain boundaries, can be
Microstructural targets for magnetic highly deleterious.
functions do not seem to be so critical as Control of porosity is also desirable, and
for electrical functions because less re- open or interconnected, porosity is espe-
liance is placed on control of grain bound- cially to be avoided since this allows fast
ary phases, and more is placed on the bulk penetration of corrodants which can then
properties of the major crystalline phase attack a large surface area. Some relatively
(Goldman, 1988). Control of composition, low-technology materials, such as chemi-
especially dopant levels, is the principal cal stoneware used as tower packing in
target, allowing control of permeability, chemical plants, may only be impervious
susceptibility, coercivity, remanence, heat by virtue of an impervious skin, sealing
dissipation in cycling, and magnetostric- interconnected pores in the bulk of the ma-
tive properties. terial.

1.13 Resistance to Corrosion 1.14 Joinability

Many ceramic phases have excellent re- A very important feature for many ap-
sistance to corrosion in water, mineral plications of ceramics is the ability to join
acids and gases at elevated temperature, it to other materials. If a rigid "chemical"
but such advantages are often lost when joint is to be produced, e.g., by metallising
they are incorporated into ceramic materi- and brazing, ceramic microstructural con-
als. The principal problem is that ceramics siderations may be involved because some
usually contain one or more interpenetrat- processes are highly dependent for their
ing phases (Lay, 1992). For example, a success on microstructural control. In the
high-alumina ceramic may have a network molybdenum/manganese process for
of alumina grains, but there is also a con- metallising, much research has been car-
tinuous grain boundary phase resulting ried out to establish the optimum mi-
from the need to control sintering or grain crostructure of alumina ceramics to obtain
 1.16 References 23

a well-bonded reliable metallised layer. chapter could not have been written with
The critical aspect appears to be the ability such an approach in mind. The essential
of a suitably fluid glassy phase to exude factors that comprise the target vary from
from the ceramic at high temperatures and application to application, or from func-
to enter the pore space in the sintering tion to function. It has generally been pos-
metallising, providing a strong key. The sible to indicate possible targets for opti-
channel width for the glass has to be large mising the performance of each function,
enough, and this is accomplished by allow- but it has to be borne in mind that only
ing adequate grain growth well beyond the very seldom does an application require
optimum size for high strength (Twenty- just a single function. Usually a combina-
man and Popper, 1975). On the other tion of material attributes is needed, and
hand, making the metallisation with a sometimes these conflict in terms of desir-
powdered glass incorporated has been able microstructures. It is always necessary
found to be less effective. Factors such as to consider compromising, e.g., between
these are peculiar to particular process strength and open porosity, or between
types, and thus there can be no generalised purity and cost. Of course, we also have to
microstructural target for joinability. consider how we will attain the target,
which may not be straightforward. So
while we might set ourselves some hypo-
1.15 Concluding Notes thetical microstructural targets, we must
be aware that real world will impose con-
This chapter has illustrated the strong siderable limitations. For example, we find
connection between microstructure and it hard to mimic Nature in its fabrication
properties or performance of ceramics, of tough sea-shells (Yasrebi et al., 1990).
whether coarse-grain refractories or the ul- However, the advent of chemical process-
timate advanced technical ceramics. Con- ing has opened up a wide range of new
trol of the microstructure through control possibilities, and in the future we will con-
of chemistry, form of the raw materials, tinue to see the diversity of ceramic micro-
processing of raw materials, shaping of structures increase and their quality im-
products, consolidation by firing, and prove.
modification by post-firing processes is the
key to manufacture of consistent products
with well-defined and reliable perfor-
mance. The first target must be to under- 1.16 References
stand the microstructure and the role it Becher, P. R, Tiegs, T. N., Ogle, I C , Warwick, W. H.
plays in determining performance, and (1986), in: Fracture Mechanics of Ceramics, Vol. 7:
thereby to try to set targets in all the pro- Bradt, R.C., Evans, A.G., Hasselman, D.P.H.,
Lange, E.F. (Eds.). New York: Plenum Press,
cessing factors which will result in the as- pp. 61-73.
sembly of a target improved microstruc- Breck, D.W., Anderson, R.A. (1981), in: Kirk-Oth-
ture. In some cases, microstructure plays mer Encyclopaedia of Chemical Technology, Vol. 15,
3rd. ed. New York: Wiley-Interscience, pp. 638-
only a minor role, but in others it is crucial 669. See also: Breck, D. W. (1974), Zeolite Molecu-
to the successful exploitation of the materi- lar Sieves, Structure, Chemistry and Use. New
al. York: John Wiley.
Budnikov, P.P. (1964), The Technology of Ceramics
It is not easy to set ideal microstructure and Refractories. London: Edward Arnold, Chap.
targets in a completely abstract way. This XIV.
24 1 Microstructural Targets for Ceramics

Buessem, W.R., Thielke, N.R., Sarahauskas, R.V. Lange, F. F. (1983), in: Progress in Nitrogen Ceram-
(1952), Ceram. Age, 60, 38. ics, Proc. NATO ASI Series E., Vol. 65: Riley, F. L.
Byrne, W. P., Morrell, R., Lawson J. (1988), Science of (Ed.). Leyden, Netherlands: Nordhoff, pp. 467-
Ceramics 14, 775. 490.
Chesters, J.H. (1973), Refractories: Production and Lay, L.A. (1992), Corrosion Resistance of Technical
Properties. London: The Iron and Steel Institute. Ceramics, 2nd ed. London: HMSO.
Clarke, D.R. (1992), J. Am. Ceram. Soc. 75, 739. Marshall, D. B., Evans, A. G. (1986), in: Fracture Me-
Clarke, D. R., Shaw, T.M., Dimos, D. (1989), J. Am. chanics of Ceramics Vol.7: Bradt, R . C , Evans,
Ceram. Soc. 72, 1103. A.G., Hasselman, D.P.H., Lange, F.F. (Eds.).
Clinton, D. X, Morrell, R., McNamee, M. (1986), Br. New York: Plenum Press, pp. 1-15.
Ceram. Trans. J. 85, 175. McMillan, P. W. (1979), Glass-Ceramics, 2nd ed. Lon-
Davidge, R. W. (1973), in: Fracture Mechanics of Ce- don: Academic Press.
ramics, Bradt, R.C., Hasselman, D.P.H., Lange, Minjolle, L. (1990), Ind. Ceram. (Paris) (846), 132.
F. F. (Eds.). New York: Plenum Press, pp. 447-468. Moore, M.A., King, F.S. (1980), Wear 60, 123.
Davidge, R.W. (1979), Mechanical Behaviour of Ce- Morrell, R. (1987), Handbook of Properties of Techni-
ramics. Cambridge, UK: Cambridge University cal and Engineering Ceramics: Part 2: Section 1:
Press. High-Alumina Ceramics. London: HMSO.
Davidge, R.W. (1982), Proc. Br. Ceram. Soc. 32, 199. Morrell, R., Butterfield, D.M., Clinton, D.J., Barratt,
Evans, A. G. (1988), Mater. Sci. Eng. A 105-106, 65. P.G., Oakey, J.E., Reed, G.P., Durst, M., Burnard,
Fricke, J. Caps, R. (1988), in: Ultrastructure Process- G. K. (1990), in: Ceramics in Energy Applications:
ing of Advanced Ceramics: MacKenzie, J.D., Ul- New Opportunities: Institute of Energy (Eds.). Bris-
rich, D.R. (eds.). New York: Wiley Interscience, tol, U.K.: Adam Hilger/IOP Publishing Ltd,
pp. 613-622. pp. 203-214.
Fricke, J., Emmerling, A. (1992), J. Am. Ceram. Soc. Muilwijk, F , Tholen, J. P. P. (1989), in: Euroceramics:
75, 2027. de With, G., Terpstra, R.A., Metselaar, R. (eds.).
Gee, M.G., Matharu, C.S., Almond, E.A., Eyre, London: Elsevier, pp. 3596-3599.
T.S. (1989), in: Proc. Conf Wear of Materials, Nakahira, A., Fukushima, Y, Niihara, K. (1989),
Vol. 1: Ludema, K.C. (Ed.). New York: American Funtani Oyobi Funmatsu Yakin 36, 746.
Society of Mechanical Engineers, pp. 387-397. Niihara, K. (1992), Mem. Inst. Sci. Ind. Res. Osaka
Gillery, F.H., Bush, E. A. (1952), J. Am. Ceram. Soc. Univ. 49, 21.
42, 175. Philipp, H. R., Levinson, L. M. (1983), in: Advances in
Goldman, A. (1988), in: Electronic Ceramics - Prop- Ceramics Vol. 7: Additives and Interfaces in Elec-
erties, Devices and Applications: Levinson, L. M. tronic Ceramics: Yan, M. F , Heuer, A. H. (Eds.).
(Ed.). London: Marcel Dekker, pp. 147-190. Columbus, Ohio: American Ceramic Society,
Gowda, G., Harrison, T. (1987), / Can. Ceram. Soc. pp. 1-21.
55, 68. Pratt, A.W. (1969), in: Thermal Conductivity, Vol. 1:
Green, D.J., Hannink, R.H.J., Swain, M.V. (1989), Tye, R. P. (Ed.). London: Academic Press, pp. 301-
Transformation Toughening of Ceramics. Boca Ra- 405.
ton, FL: CRC Press. Ramsey, P.M., Page, T. F. (1989), in: Proc. Conf.
Haertling, G.H. (1988), in: Electronic Ceramics - Wear of Materials 1989, Vol.2. Ludema, K.C.
Properties, Devices and Applications: Levinson, (ed.). New York: American Society of Mechanical
L. M. (Ed.). London: Marcel Dekker, pp. 371-492. Engineers, pp. 629-636.
Hasselman, D.P.H. (1969), /. Am. Ceram. Soc. 52, Russel, C , Hofmann, T., Limmer, G. (1991), CFI
600. Ber. Deut. Ker. Ges. 68, 22.
Hasselman, D. P. H. (1970), Bull. Am. Ceram. Soc. 49, Seitz, M. A., Orlow, W. (1981), in: Advances in Ceram-
1033. ics Vol. 1: Grain Boundary Phenomena in Electronic
Herbert, J. M. (1982), Ferroelectric Transducers and Ceramics: Levinson, L. M., Hill, D.C. (Eds.),
Sensors, New York: Gordon and Breach. Columbus, OH: American Ceramic Society,
Hing, P. (1976), Sci. Ceram. 8, 159. pp. 124-129.
Kingery, W. D., Bowen, H. K., Uhlmann, D. R. Shaulov, A. A., Smith, W.A., Singer, B.M. (1984),
(1976), Introduction to Ceramics, 2nd ed. New Proc. 1984 IEEE Ultrasonics Symp., pp. 545-548.
York: Wiley Interscience. St-Jacques, R. G., Moise, A., Yeroulchami, D. (1974),
Kirchner, H.P. (1979), Strengthening of Ceramics - /. Can. Ceram. Soc. 43, 23.
Treatments, Tests and Design Applications. New Twentyman, M. E., Popper, P. (1975), Special Ceram-
York: Marcel Dekker. ics, 6, 67.
Kuszyk, J. A., Bradt, R.C. (1973), J. Am. Ceram. Soc. Van Praag, W, Zaspalis, V.T., Keizer, K., Van Om-
56, 420. men, J. G., Ross, J. R. H., Burggraaf, A. J. (1989), in:
Lachman, I.M., Bagley, R.D., Lewis, R.M. (1981), Euroceramics, Vol. 3: de With, G., Terpstra, R. A.,
Bull. Am. Ceram. Soc. 60, 202. Metselaar, R. (Eds.). London: Elsevier, pp. 3605-
Lange, F. F. (1978), Mater. Sci. Res. 11, 597. 3609.
 1.16 References 25

Ward, C.P. (1989), Proc. Br. Ceram. Soc. 41, 85. Lay, L. (1991), Corrosion Resistance of Technical Ce-
Yasrebi, M., Kim, G. H., Gunnison, K.E., Milius, ramics. London: HMSO.
D.L., Sarikaya, M., Aksay, LA. (1990), Mater. Morrell, R. (1987), Handbook of Properties of Techni-
Res. Soc. Symp. Proc. 180, 625. cal and Engineering Ceramics: Part 2, Data Re-
views; Section I, High-Alumina Ceramics. London:
HMSO.
General Reading Moulson, A. I, Herbert, J.M. (1990), Electroceram-
ics: Materials, Properties, Applications. London:
Brook, R. I (Ed.) (1991), Concise Encyclopedia of Ad- Chapman and Hall.
vanced Ceramic Materials. Oxford: Pergamon Press. Ryshkewitch, E., Richerson, D.W. (1985), Oxide Ce-
Hoffmann, M. J., Petzow, G. (1994), Tailoring of Me- ramics. Haskell, NJ: General Ceramics Inc.
chanical Properties ofSi3N4 Ceramics, NATO Adv. Schwartz, M.M. (1992), Handbook of Structural Ce-
Sci. Inst. Ser., Ser E, Vol. 176. Dordrecht: Kluwer. ramics. London: McGraw-Hill.
Lee, W.E., Rainforth, W.M. (1994), Ceramic Micro- Somiya, S., Yamamoto, N., Hanagida, H. (Eds.)
structures - Property Control by Processing. Lon- (1988), Science and Technology of Zirconia III,
don: Chapman and Hall. Adv. Ceram. Vols. 24 A, B. Westerville, OH: Amer-
Clinton, D. J. (1987), A Guide to Polishing and Etching ican Ceramic Society.
of Technical and Engineering Ceramics. Stoke-on- Wachtman, J.B., Jr. (1989), Structural Ceramics.
Trent, U.K.: Institute of Ceramics. London, Academic Press.
2 Process Control in the Manufacture of Ceramics
Gijsbertus de With

Philips Research, Eindhoven, The Netherlands, and Eindhoven University of Technology,

Eindhoven, The Netherlands

List of Symbols and Abbreviations 29

2.1 Introduction 31
2.1.1 Process Control and the Need for Characterization 31
2.1.2 Complex Processes and the Need for Statistical Process Control 31
2.2 Overview of the Ceramic Process 32
2.3 Powders 35
2.3.1 Grain Size and Specific Surface 36
2.3.2 Chemical Composition and Moisture Content 38
2.3.2.1 Auxiliary Raw Materials 39
2.3.3 Processing and Functional Tests 39
2.3.4 Milling 41
2.3.4.1 Ball Milling 42
2.3.4.2 Attrition, Vibro-, and Jet Milling 42
2.3.4.3 Changing Characteristics 43
2.4 Consolidation 44
2.4.1 Compact Characterization 44
2.4.2 Pressing 46
2.4.2.1 Die Pressing 46
2.4.2.2 Isostatic Pressing 48
2.4.3 Injection Molding 49
2.4.3.1 Properties and Mixing 49
2.4.3.2 Molding 51
2.4.3.3 Binder Burn-Out 51
2.4.4 Extrusion 52
2.4.4.1 Process Control 53
2.4.5 Slip Casting 54
2.4.5.1 Normal Slip Casting 54
2.4.5.2 Other Casting Techniques 55
2.4.6 Tape Casting 56
2.4.6.1 Doctor Blade Processing 56
2.4.6.2 Solvent/Binder System 57
2.4.6.3 Casting/Drying Process 58
2.4.6.4 Other Tape Processes 58
2.4.7 Comparison of the Various Consolidation Methods 59

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
28 2 Process Control in the Manufacture of Ceramics

2.5 Sintering 59
2.5.1 Furnaces 59
2.5.2 Temperature 60
2.5.3 Atmosphere and Additives 60
2.5.4 Hot-Pressing 61
2.6 After-Sintering Control 62
2.6.1 Grinding 62
2.6.2 Visual and Geometrical Control 63
2.6.3 Microstructure 64
2.6.4 Functional Properties 65
2.7 Final Remarks 65
2.8 References 66
 List of Symbols and Abbreviations 29

List of Symbols and Abbreviations

c intrinsic viscosity
C capability index A/6s
d primary particle size
D agglomerate size; mill diameter
E Young's modulus
/ volume fraction
F cumulative size distribution
g acceleration due to gravity
h height between doctor blade and belt
k packing parameter in Dougherty-Krieger relation
K permeability
/ intercept length
L thickness
n sample size; number of data points
N number of particles
p normal stress
P pressure
q shear stress
r radius
R property range; fracture energy; gas constant; roughness
s sample standard deviation
S specific surface area
t time
T temperature; knife thickness
v settling velocity; velocity of doctor blade belt
V molar volume
x property (general); radius position; pore diameter; particle size
Y yield stress
z profile deviation

a width correction factor for doctor blade process

/? packing correction factor for doctor blade process
y surface energy
A upper specification limit —lower specification limit
rj viscosity
9 contact angle
Q density
co angular frequency; number of revolutions

AFM atomic force microscopy

BET Brunauer, Emmett, and Teller N 2 adsorption technique
CIP cold isostatic pressing
CLA center line average
30 2 Process Control in the Manufacture of Ceramics

CMC ceramic multilayer capacitor

DTA differential thermal analysis
EDX energy dispersive X-ray analysis
EPMA electron probe microanalysis
HIP hot isostatic pressing
LCL lower control limit
LSL lower specification limit
MIP mercury intrusion porosimetry
PTCR positive temperature coefficient of resistance
PVA poly (vinyl alcohol)
RMS root mean square
SEM scanning electron microscopy
STM scanning tunneling microscopy
TGA thermogravimetric analysis
UCL upper control limit
USL upper specification limit
XRD X-ray diffraction
 2.1 Introduction 31

2.1 Introduction Nowadays, process control generally

means active process control. This includes
Within the usual ceramic research labo- anticipating process deviations from mea-
ratory environment, many analysis and sured data and/or changing the raw mate-
characterization techniques are available. rial characteristics for process adjustment.
In a development laboratory the possibili- The control of a process therefore involves
ties are much more limited, since more characterization aspects as well as system
dedicated and pragmatic goals are usually aspects. For the latter a process specifica-
defined. Within the factor environment, tion, often represented by a flow chart, is
any analysis will be directly reflected in the quite useful. This flow chart indicates
price of the product and therefore only the where raw materials originate from, what
most necessary equipment is available. process conditions are applied, what mea-
Nevertheless, it is the task of the factory to surements should be performed at what
deliver products reliably and reproducibly stage, and what actions are to be taken
at the cheapest price. The aim of this chap- when the process runs outside the specifi-
ter is to summarize the control aspects of cations. This flow chart is often translated
the normal ceramic processing route in into a computer program which instructs
laboratory and factory practice, from operators where necessary, takes care of
powder to finished product. the data belonging to a semiproduct, and
Consequently, this is not a scientific controls the logistics. This part of process
Chapter but a pragmatic one, in which the control is rather more general than for ce-
relevant measurements that could be taken ramic processing only. Therefore in this
at each stage are indicated along with why article we primarily limit ourselves to the
they differ for different types of products characterization aspects.
and processes. Different analytical meth-
ods measure different characteristics.
2.1.2 Complex Processes and the Need
Therefore the combined use of numerous
for Statistical Process Control
characterization techniques offers consid-
erably more insight into the process. Many A complex process contains many
of the steps are illustrated by actual exam- parameters with many, usually nonlinear
ples of data from various processes. It is relations between the parameters (Juran
hoped that in this way a good grasp of the et al., 1974). Ceramic processes are consid-
actual situation in ceramic process control ered to be complex processes. In such a
can be obtained. complex process the change in settings is
usually chosen on the basis of designed
experiments, ranging from full factorial to
2.1.1 Process Control and the Need
Taguchi methods. Moreover, in the usual
for Characterization
ceramic factory a large number of prod-
The meaning of 'process control' is not ucts are made and this makes statistics in-
entirely unambiguous. In this case, a pro- dispensable. The process is monitored and
cess is defined as any combination of ma- if the process is misdirected or exceeds the
chines, tools, methods, materials, and per- statistical limits, corrective actions are tak-
sonnel to attain products with the desired en. Misdirection can be corrected by
quality, while control entails establishing changing the process settings. The reasons
and meeting standards. for running out of control may be twofold:
32 2 Process Control in the Manufacture of Ceramics

specific and common causes. Common Table 2-1. Types of processes and their capability in-
causes may be measurement and/or accu- dex C.
racy errors, variability of raw materials, C Type of process Type of control
and operator behavior. Specific causes
have to be identified, ranked in order of > 1.33 stable operator control
1.33-1.0 controlled factory engineer control
decreasing importance, and subsequently 1.0-0.67 not controlled control by development
removed. The 'Pareto principle' or the <0.67 not adequate improvement by
principle of the vital few and trivial many, development
tells us that only a few causes account for
most of the total effect. Typically, 20% of
the causes are responsible for 80 % of the property with time is a useful indicator for
effect (Juran etal., 1974). the process. When these values are plotted
In order to determine whether the pro- as a function of time together with the
cess is under control, a number of tools UCL and LCL, the resulting graphs are
have to be available. During processing usually denoted as the x and R control
samples of size n (subgroups) are taken. charts. Sometimes the USL and LSL are
For most properties a certain range of val- indicated as well. Process misdirection, in-
ues is obtained and allowed. On the one adequate measurement, process drift, and
hand, if the sample standard deviation of sudden changes can be spotted in this way.
the subgroup is denoted by s and the aver- If the process capability (6 s) is incompat-
age value by x, 99.7% of the individual ible with the product tolerances (A), either
measurements are within the range x-3s the process or the tolerances have to be
and x + 3s. For the subgroup average x changed, or else it is a case of 'suffer and
therefore, limits are set at x-3s/y/n and sort'.
x + 3s/^/n. These limits are usually denot- Many more and also more sophisticated
ed as the lower control limit (LCL) and the statistical techniques exist, see, e.g., Him-
upper control limit (UCL). On the other melblau (1970). In a number of examples
hand, the process specification is given by in this chapter, the previously mentioned
a lower specification limit (LSL) and an simple concepts will be used. These exam-
upper specification limit (USL). If the dif- ples should not be considered as represen-
ference between LSL and USL is denoted tative of the quality of the specific process.
by A, the so-called capability index, C, is Finally, it should be stated clearly right
given by C = A/6s. The process is deemed away that many aspects of ceramic pro-
to be under control (i.e., the product prop- cessing control are of a qualitative nature
erty is within the specifications) if C > 1. and therefore beyond the limits of statisti-
The actions to be taken for various values cal process control.
of C are indicated in Table 2-1 (Juran
et al., 1974). As a rule of thumb, the value
for C decreases by 40% when a process 2.2 Overview of the Ceramic
is transferred from development to the Process
factory.
A process may be under control but de- A close relation between processing and
viate from the standard. This calls for a the resulting properties exists for many
number of tools. Monitoring of the aver- materials, but this is particularly valid for
age value x and the range R of the relevant ceramic materials. The link between pro-
 2.2 Overview of the Ceramic Process 33

cessing and the properties of ceramic mate- product. Ceramic processing consists of
rials is represented by the microstructure four main stages: raw materials prepara-
of the material. According to Exner tion, consolidation to compacts, densifica-
(1983): 'the microstructure is defined by tion to dense ceramics, and machining to
the type, the structure and the number of achieve a specific surface finish and/or size
phases, by the number, the geometric ap- (see Table 2-2). Less emphasis is placed on
pearance (size, shape, etc.) and the topo- machining, since it is not always necessary.
logical arrangement of the individual Indeed, there is a trend to try to avoid
phase regions and their interfaces and by machining after densification and instead
the type, structure and geometry of lattice to try to make (near) net-shape compo-
defects'. The relation between processing, nents.
microstructure, and properties is given Raw materials can be divided into pow-
schematically in Fig. 2-1, adapted freely ders and fibers (including whiskers). Pow-
from Stuijts (1977). The essence of this re- ders can be characterized in several ways
lation is the difference between the 'com- (see Table 2-2). First of all chemically: the
pound' on the one hand and the 'material' stoichiometry and amount of impurities
on the other hand. The properties of a present. Secondly, crystallographically:
compound are intrinsic and can hardly be nonreacted phases and second phases (e.g.,
influenced when its composition is fixed. due to milling). Thirdly, morphologically:
They comprise properties like crystal the presence of agglomerates, the size (dis-
structure, thermal expansion coefficient, tribution) and shape, and the specific sur-
refractive index, and magnetic crystalline face area. Finally, the bulk properties, in-
anisotropy. The properties of a material cluding the packing and flow of the pow-
are to a large extent extrinsic and can be ders as well as their thermal response. For
drastically changed by altering the mi- fibers the diameter, the aspect ratio (the
crostructure through different processing ratio between length and diameter), and
routes. Typical examples are the mechani- mechanical properties like the Young's
cal properties like fracture strength and modulus and strength are also of major
fracture toughness, the permittivity for fer- importance. The art and science of milling
roelectric materials, and the permeability belongs to raw materials processing.
for ferromagnetic materials. Powder compacts can be realized in var-
Although in other chapters the various ious ways. The most commonly used meth-
steps of the ceramic processing route are od is still pressing, either using dies or iso-
described in detail, it is convenient to re- statically. Other important techniques are
peat here the main route from powder to slip casting, injection molding, and extru-

Figure 2-1. From compound to mate-

rial. While proper intrinsic compound
Processing properties are a prerequisite for ob-
taining good material properties, the
realization of the ultimate material
properties is achieved in the micro-
structure through proper processing.
34 2 Process Control in the Manufacture of Ceramics

Table 2-2. Ceramic processing. tion), the pore size (distribution), chemical
Stage Characterization homogeneity, and binder distribution
within the compact are the major aspects
Raw chemistry stoichiometry to consider (see Table 2-2).
materials - impurities Densification of the compacts can be
crystallography - nonreacted phases carried out in various ways. Almost all of
- second phases
them are variations on sintering. Sintering
morphology - agglomerates
- size distribution
means bringing the compacts to such a
- shape distribution high temperature that sufficient mobility is
- specific surface area present to release the excess surface energy
bulk packing of the powder, thereby joining the particles
- flowability together. If the diffusion takes place only
- thermal response in the solid state, the process is called solid-
Consoli- homogeneity - density state sintering. If enhanced mobility is re-
dation - binder distribution alized by a small amount of liquid phase,
- pore size
then the term liquid-phase sintering is
distribuition
- additive distribution used. It is also possible to use nonreacted
density - required level raw materials (where the final chemical
drying cracks compound is not yet formed at the start of
the sintering process) for so-called reaction
Densifi- grains - size distribution
cation
sintering. Finally, the application of an ex-
- growth
pores coordination
ternal pressure on the compact is called
number pressure sintering or hot-pressing. In this
- size distribution case the pressure can be applied (pseudo-)
2nd phases - overall distribution uniaxially or isostatically. Important as-
- size pects to consider are the grain size distri-
- grain boundary bution and growth, the densification rate,
pinning inhomogeneities in the densification, the
chemistry segregation
second phase (distribution), and the grain
- liquid phase
- furnace boundary chemistry (see Table 2-2).
contamination Due to the usually high hardness of the
- gas atmosphere sintered material, machining is conven-
shape - rate of densification tionally done by grinding using diamond
- inhomogeneous
grinding wheels. Nowadays, techniques
densification
- shrinkage like ultrasonic and laser machining are be-
coming more important. In the case of
Machining geometry size accuracy
- roughness
electrically conducting materials, spark
mechanics - introduction
erosion is another possibility. Tumbling is
of defects a frequently used operation for surface fin-
ishing. Apart from the demands for accu-
racy in dimensions, the surface roughness
sion. The importance of wet consolidation obtained is also important. An unwelcome
techniques like filtration, sedimentation, consequence of machining is the introduc-
and electrophoretic deposition is growing. tion of (mechanical) defects. These defects
For consolidation the density (distribu- usually have a negative influence on the
 2.3 Powders 35

mechanical behavior, in particular the agglomerate (10-100 pm)

strength of the material (Table 2-2).
This completes the overview of the vari- domain (0.1-1 jjm)
ous processing steps. More information
can be found in the various chapters of this
Volume and in McColm and Clark (1988),
Reed (1988) and less recently, Onoda and
Hench (1978) and Wang (1976). A lot of
information can also be found in the books \ J
on ultrastructure processing edited by particles (0.01-0.1 pm)

Hench and Ulrich (1984, 1986), McKenzie

and Ulrich (1988), and Uhlmann and Ul- Figure 2-2. Schematic view of an agglomerate in a
powder. Three substructures with decreasing size can
rich (1992) and the series Better Ceramics
be distinguished: agglomerate, domain, and primary
through Chemistry edited by Brinker et al. particle. Often the difference between domain and
(1984, 1986, 1988) and Zelinski etal. particle is not considered and the domain is called a
(1990). Attention will now be focused on particle.
the control aspects of the individual pro-
cess steps.
further attention is paid to the various
preparation routes.
Some of the characteristics of raw mate-
2.3 Powders
rials are indicated in Table 2-2. Only a few
It is convenient to define briefly what is of these are determined in factory practice;
meant by powder particles. A schematic in particular those parameters to which the
image of a powder is therefore presented in process control is sensitive. For example,
Fig. 2-2. Three size levels of powder parti-
cles are often distinguished: The largest
size, i.e., the agglomerate, consists of
smaller units, i.e., the domains, which in
g 40 -
turn consist of the smallest units, i.e., the
primary particles. Often just two levels are
used, neglecting the difference between do-
main and primary particle. We will do so in
this chapter. The preparation route of the
powders largely determines their charac-
teristics. Apart from the fact that different
0 5 10 50 100 500
preparation routes yield different powder compaction pressure (MPa)
characteristics, minor changes in one pro- Figure 2-3. Compaction curves of two types of zirco-
cess can also influence the characteristics nia powders. Both show two linear regions represent-
considerably. An example is given in ing rearrangement and fracture of the agglomerates,
Fig. 2-3, where the compaction behavior respectively. The transition pressure Pt at the intersec-
of powders similarly processed but with tion of the linear parts is an indication of the strength
of the agglomerates. One powder is washed with wa-
different washing agents is shown. Since ter (o) and the other with propanol-2 (•). The use of
powder preparation is becoming more and water clearly increases the strength of the agglomer-
more a separate branch of processing, no ates.
36 2 Process Control in the Manufacture of Ceramics

in the case of pressing this will be the The weight distribution can be calculated
pressed density, while for slip casting this from the latter if the density, Q, is known.
will be the rheology of the slurry to be Not realizing these differences may easi-
made. In most cases the powder size is also ly result in a difference in size of a factor of
characterized, since this is a fundamental about three compared with other workers.
parameter influencing nearly all further It should be noted that the ceramic pro-
steps. However, the chemical composition ducer often relies on the (contract) specifi-
is not usually checked since this is a costly cation of the powder producer instead of
operation. Let us now consider a number doing his own particle size measurements;
of the relevant parameters in turn. in particular in the case of comakership.
The size characteristics of powders can
be determined in various ways. A brief
enumeration of the more well-known tech-
2.3.1 Grain Size and Specific Surface
niques, together with their vices and
It should be apparent from the outset virtues, is given in Table 2-3. The most
that a clear statement of what is meant by frequently used ones are sedimentation
grain size should be made, since confusion analysis and specific surface measurement.
is possible. Any confusion can, however, Their use is discussed in some detail. The
be diminished by realizing the following other size characterization techniques are
points: less frequently used and are not discussed
• The definition of 'size' may differ, de- here. Although it may not be used for
pending on the technique used, e.g., circle quantitative purposes, observation of the
area (perimeter) equivalent diameter, powder by scanning electron microscopy
Stokes' diameter, and intercept. (SEM) is generally quite useful since exam-
• The method of measuring the size may ination of individual particles can reveal
differ: Whether mode, median, or mean features not spotted by techniques that
(or any other) size is used should be made measure collective properties.
clear. In sedimentation analysis the particle
• Different techniques measure different size is determined from Stokes' law, either
distributions; the most important ones be- using settling by normal gravity or settling
ing the number and volume distribution. reinforced by a centrifugal field. The di-

Table 2-3. Main characteristics of particle-size measurement methods.

Method Size Limiting Remarks Distribution

definition value typea

Sieving minimum > 1 - 5 um bias to smaller size V

Feret dependent on procedure
Microscopy nearly free choice > 1 nm bias to larger size N
Electrical pulse cross-section equivalent 0.5-100 um sphericity assumed N, V
Light scattering statistical <2 jim sphericity assumed V
Specific surface statistical <2 um primary particles V
Sedimentation Stokes 0.3-100 um agglomerate V
a
V: volume distribution, N: number distribution.
 2.3 Powders 37

ameter of a powder particle, D, settling a larger size distribution (due to aggrega-

under gravity and assumed to be a sphere, tion and/or flocculation). Techniques
is given by where a dry powder is used directly for
measurement without wet dispersing exist,
D = mr,v)l(Aeg)Y'2 (2-1) but are not generally used.
where rj denotes the viscosity, v the settling The specific surface is usually deter-
velocity, AQ the difference in density be- mined by the standard N 2 -adsorption
tween liquid and powder, and g the accel- technique, which is known as the BET
eration due to gravity. In the case of cen- technique after the inventors Brunauer,
trifuging objects, the corresponding equa- Emmett, and Teller (Gregg and Sing,
tion is 1982). The surface of the powder is first
degassed at an intermediate temperature
D = {[18 (2-2)
150-300°C). Then at 77 K, the boiling
where x and x0 denote the radius of the temperature of liquid nitrogen, N 2 gas is
particle's position at time t = t and t = 0, adsorbed and the absorption isotherm is
respectively, and co denotes the angular measured. Each N 2 molecule covers a
frequency. Recording the settled mass as a known surface area of about 0.162 nm2
function of time yields the cumulative size when it sticks to the surface. Hence the
distribution in both cases. The location of total surface area can be calculated from
the size distribution is conveniently de- the total amount of adsorbed N 2 in a
scribed by the median; here denoted by monolayer. Monolayer coverage is calcu-
D 50 . Commercial equipment is available lated according to the BET theory from
for both techniques (e.g., the Micromerit- the adsorption isotherm. For economy
ics Sedigraph 5000 D analyzer (Norcross, reasons, often only the low coverage range
GA) and the Shimadzu SA-CP4 centrifu- or just one or two points in the low cover-
gal analyzer (Kyoto), respectively). The age range are recorded instead of the
usual range of size measurements is 0.3- whole isotherm. The measuring range is
100 jim. Special and expensive equipment normally 3-300 m 2 /g. Incidentally, other
is required for much smaller sizes, while gases, notably Ar, are used for smaller ar-
sieving is useful for larger sizes. Because eas. The primary particle size can be calcu-
the powder agglomerates settle as a whole, lated from the specific surface via a shape
the sedimentation technique measures the assumption. The most frequently used as-
agglomerate size and not that of the prima- sumption is spherical primary particles
ry particles. Moreover, the pretreatment with a negligible contact area between
required for measuring may significantly them. This results in
alter the size distribution. This pretreat-
ment usually consists of dispersing about d=6/gS (2-3)
1 vol.% of the powder in a solution of where d denotes the primary particle size in
<0.01 wt.% Na 4 P 2 O 7 • 10 H 2 O in water. \xm, Q the density in g/cm3, and S the
At higher concentrations pronounced non- specific surface area in m 2 /g. From the
linearities appear. Nonaqueous media are combined use of the agglomerate size, D 5 0 ,
sometimes used as well as other salts. As a and the primary particle size, d, an esti-
consequence of this pretreatment, the mea- mate can be made for the number of parti-
surement can result in a smaller size distri- cles per agglomerate. Assuming spherical
bution (due to dispersal of the particles) or agglomerates, this number, N, is given ap-
38 2 Process Control in the Manufacture of Ceramics

proximately by 2.3.2 Chemical Composition

and Moisture Content
N~(D5O/d)3 (2-4)

Since frequently D50>d, this equation The most frequently used chemical anal-
indicates that agglomerates generally con- ysis methods are enumerated in Table 2-4.
sist of quite a large number of primary Each technique has its advantages and dis-
particles: however, well deagglomerated advantages, and some of these are indicat-
powders can be obtained. An example of ed. In many cases more than one method is
this is provided by an Y 3 A1 5 O 12 garnet necessary to characterize the material ful-
powder, which was prepared by calcining a ly. A particular problem in the chemical
spray-dried sulphate powder (de With, analysis of ceramics is the disclosure of the
1987). The BET surface was 5.0 m 2 /g, cor- elements, in particular impurities, due to
responding to a diameter d of 0.51 jim. the generally high inertness of ceramics.
Sedimentation analysis yielded a value of Additionally, and more frequently, stan-
D50 of 1.1 jim. Hence the median agglom- dard X-ray diffraction (XRD) is per-
erate contained about 10 primary parti- formed. Although it is a well-established
cles. Upon deagglomeration of the powder technique, it should be made clear that if
by ball milling with polymer balls, the val- all the X-ray reflections can be identified
ue of D50 decreased to 0.46 \xm, indicating as belonging to one phase, this does not
that 'on average' each agglomerate con- necessarily imply that the sample is single
sisted of one primary particle. With the phase, since the detection limit is typically
original powder nearly fully dense ceram- a few percent. Moreover, glassy inpurities
ics could be prepared, while with deag- do not show as sharp reflections.
glomeration fully dense and translucent Frequently, the chemical composition
ceramics could be obtained. This example and phases of a powder are checked only
thus illustrates the importance of deag- in the starting phase of using a powder for
glomeration. economy reasons. In later stages the con-

Table 2-4. Comparison of various chemical characterization methods.

Method Principle Remarks

Wet chemical analysis dissolve and use various inertness problem;

techniques (titration, spectroscopy) ppm level possible
X-ray fluorescence record characteristic X-ray lines nondestructive; calibration necessary
ppm level possible; surface sensitive
(urn depth)
Emission spectroscopy vaporize, record optical spectra ppm level possible
Neutron activation analysis irradiate, record radioactivity expensive
EDX a record energy X-ray lines qualitative; flat surface required;
light element problem
EPMA b record wavelength quantitative; flat surface required;
X-ray lines light element problem
XRDC record diffraction angles, rapid; qualitative;
compare with JCPDS files nondestructive
1 b
EDX: energy dispersive X-ray analysis; EPMA: electron probe microanalysis; c XRD: X-ray diffraction.
 2.3 Powders 39

tract specification is often relied on. There tent is frequently determined as the re-
are two exceptions to this. One is the mea- mainder after burning them in order to
surement of the water content of a powder, avoid undue contamination.
which is simply determined by weighing
before and after a specific drying process.
2.3.3 Processing and Functional Tests
A variation over various batches of about
0.2% is generally observed on an average Depending on the processing route, a
moisture content of 1 %. The moisture specific test is often performed in order to
content can greatly influence the process- establish the processability of the powder
ability of the powder. This is particularly (Table 2-2). In the case of pressing this may
true for the pressability, since moisture can include the tap density, the bulk density,
influence the glass transition temperature and the pressed density. The tap density is
of many binders, and the flowability, since the density of the powder after the powder
an increased moisture content increases container has been tapped a given number
the cohesion of the powder. The other ex- of times with a given force. The bulk densi-
ception is the determination of specific im- ty is the special case of tap density after
purities, which are particularly detrimental zero taps. The pressed density is the densi-
to the final product, e.g., the presence of Si ty of a compact after pressing and directly
in A12O3 powder for translucent alumina. determines the shrinkage of the compact
Although foreign elements can be detected during later processing. It is usually deter-
by chemical analysis, it is also frequently mined directly after pressing. Relaxation
useful to examine the powder by SEM can take place, however, and may lower
equipped with an energy or wavelength the density by up to a few percent.
dispersive analysis system (EDX or EP- Sometimes the flowability is also mea-
MA, respectively). Particles with a differ- sured, since the flow behavior of a powder
ent morphology are generally easily spot- is important in the filling of dies. To im-
ted and often correspond to impurities. prove the flow behavior powders are often
granulated (e.g., by spray-drying). The
goal is to make granules that are strong
2.3.2.1 Auxiliary Raw Materials
enough to do the (better filling) job but
Almost unavoidably, auxiliary raw ma- sufficiently weak so that they can still frac-
terials have to be used and they also form ture, or at least deform sufficiently, during
a source of impurities and/or contamina- compaction. Measurement of the flowabil-
tion. In many cases water has to be used ity has proved to be extremely difficult.
and the amount of foreign ions dissolved From the results of an extensive investiga-
may influence the final behavior of the ma- tion of (metal) powders, the ISO working
terial considerably. This is particularly group TC 119/SC2 concluded that proba-
true for those materials where the final bly no single method emulating die filling
properties are highly dependent on the de- will be suitable for a wide range of pow-
fect chemistry of the material. Instead of a ders. Consequently, various methods are
full chemical analysis, a conductivity mea- in use (Brown and Richards, 1970). Simple
surement of the water can be used for rou- methods record the angle of response, e.g.,
tine measurements. A similar problem may by pouring the powder onto a flat and
exist with binders and other additives. For measuring the angle between the flat and
these auxiliary raw materials, the ash con- the powder heap or by measuring the angle
40 2 Process Control in the Manufacture of Ceramics

at which a cylinder, filled about half-full

with powder, starts rolling from an in-
clined flat. More sophisticated methods
use the Jenike shear cell in one form or
another (Schwedes, 1984). In this test a
powder compact is loaded with a normal
force and the upper half is sheared with
respect to the lower half by a lateral force.
If the shear stress, q, is plotted against the
normal stress, p, a linear relationship is
often observed. The slope of this plot rep-
resents the (tangent of the) internal angle
of friction, while the intercept character-
10 15 20 25 30
izes the cohesion of the powder.
batch number
Another test, more directly related to
Figure 2-4. Batch-to-batch variation in yield value
the filling of dies, is the filling plate test for alumina extrusion paste. The horizontal lines de-
(Rice and Tengzelius, 1986). This test uses note the LSL and USL, respectively. Although the
a plate with a series of holes over which a yield value is (nearly) always within the specification
powder is moved by a filling shoe. The limits, it is clear that the process can easily run out of
relative density, that is the amount of pow- control. The average yield value is 24.7 kg/cm2 with a
sample standard deviation of 2.0 kg/cm2. The USL
der in each hole divided by the volume of and LSL are 29.8 and 21.9 kg/cm2, respectively. The
each hole, is plotted against the hole di- capability index is thus about (29.9-21.9)/12 = 0.67 in
ameter. From this plot a critical hole di- this case. Moreover, the average yield value deviates
ameter, which is the diameter where 98 % from the standard 25.9 kg/cm2. Therefore the process
of the density obtained at the largest hole is barely controlled and slightly misdirected.
is reached, can be determined. The critical
filling diameter is taken as a characteristic
of the powder. Factors that influence the shown in Fig. 2-5. Much better process
test result are the direction of motion of control is obtained in this case. The viscos-
the filling shoe, speed of motion of the ity and the yield point of a good, usable
filling shoe, number of passes of the shoe, slip are constant with time. An increasing
and vibration during filling. However, dif- viscosity and yield value with time make a
ferentiation between free-flowing and not- slip nonusable. The control of processabil-
free-flowing powders appears to be rather ity often also includes the determination of
difficult. shrinkage during sintering, either by
For extrusion the yield value in a partic- dilatometry or multiple temperature firing.
ular extrusion geometry is determined. An The thermal response of the powders is
example of batch-to-batch variability is important, particularly in the calcination
given in Fig. 2-4. In this particular case the of precursors, to ensure that various
process is not very well controlled. For slip batches of powder will exhibit similar sin-
and tape casting the solid-state content, tering behavior. Thermal gravimetric anal-
the viscosity, the yield value, and the ysis (TGA) or differential thermal analysis
ageing of these properties are often mea- (DTA) are normally used to monitor the
sured. The variability of the solid content process. An example of the thermal re-
of slip-cast foils over various batches is sponse of two MgAl 2 O 4 precursor pow-
 2.3 Powders 41
2.30
practice except for products produced by
2.28 reactive sintering. Since in this case the re-
USL
~T 2.26 actions and the sintering will probably in-
£ 2.24 terfere considerably with one another (see,
UCL e.g., Toolenaar and van Lierop-Verhees
r 2-22 (1988), a more optimized firing schedule
| 2.20 may be necessary and analysis as indicated
| 2.18 LCL above may provide the required informa-
1 2.16 tion.
Often a so-called functional test is also
2.14
LSL conducted. This involves the determina-
2.12 tion of a few or all of the functional prop-
2.10 erties for a small set of samples. The pre-
5 10 15
cise content depends considerably on the
batch number
material and application at hand; for ex-
Figure 2-5. Variability of solid content of BaTiO3
ample, in the case of (ferroelectric) BaTiO3
type-cast foil for various batches. The horizontal lines
denote the LSL and USL, respectively. In this case the for ceramic multilayer capacitors (CMCs)
processing is well within limits. The USL and LSL are the functional test generally includes the
2.26 mg/cm2 and 2.14 mg/cm2, respectively, while the determination of the dielectric permittivi-
average solid content is 2.2041 mg/cm2 with a sample ty, Curie temperature, and density of
standard deviation of 0.0068. The capability index C
pressed and sintered tablets, although the
is thus about (2.26-2.14)/0.0408 = 2.9. In addition, the
real average solid content is only slightly removed material in the actual application is usually
from the standard, set at 2.2 mg/cm2, so the process is processed differently.
well under control.

2.3.4 Milling
ders is shown in Fig. 2-6 (Reynen et al., Ideally, once prepared, powders have to
1983). The considerably lower calcination be consolidated. In practice, however, ad-
temperature of the more sophisticated pre- ditional processing is often still required.
cursor should be noted. Such an elaborate This additional processing of powders
analysis is practically never performed in includes milling, deagglomeration, and

Figure 2-6. Relative weight loss AG/G0

(%) of MgAl 2 O 4 precursors as a
function of temperature. A mixture
(A) of MgSO4 and A12O3 powders re-
sults in a two-step weight loss during
prefiring. A powder made with hot
kerosene drying (•) results in a much
faster reaction during prefiring,
20 - thereby resulting in a smaller particle
size. The weight loss of the MgSO4
(•) and A12(SO4)3 powders (•) alone
is also indicated.
500 600 700 800 900 1000 1100 1200
temperature (°C)
42 2 Process Control in the Manufacture of Ceramics

adding dopants. The adding of dopants is to be made between efficiency and impuri-
frequently done in the milling and/or deag- ty introduction. Higher density milling
glomeration step. We focus therefore on balls increase the milling efficiency. Very
milling. Many forms of milling equipment often agate (mineral SiO2) is used; this has
are known. Probably the most well-known good wear resistance. The density (Q) is,
and most used type of mill is the ball mill. however, rather low (2.2 g/cm3). Other
Other types of mills that are often used in materials that are often used include (den-
ceramic technology are the attritor, vibro- sity in brackets): porcelain (2.3), A12O3
mill, and jet mill. A discussion of modern (3.8), ZrO 2 (5.6), steel (7.7), or hard metal
milling theory is given by Austin et al. (WC-Co, 15.6). The latter two introduce
(1984) and Prasker (1987). large amounts of metal impurities into the
powder which can sometimes be removed
by washing with acids.
2.3.4.1 Ball Milling
The jars of the ball mills often have a
A ball mill is a cylindrical vessel, hori- so-called 'lining': a cover over the inner
zontally rotating along its axis. The length wall of the vessel, made of a wear-resistant
of the cylinder is usually more or less equal or a little contaminating material. Materi-
to the diameter. Above a critical number als used are vulcanized rubber, poly-
of revolutions, coc (rpm), centrifuging urethane, alumina, porcelain, hard metal,
takes place which renders highly inefficient or stainless steel. Contamination remains a
milling. The value of coc can be calculated serious problem, however. Less contami-
from nation is obtained by using 'autogenous
milling', that is, milling with balls and lin-
ings made of the same materials as that
where D is the diameter of the mill in m being milled. The large difference in size
and g is the acceleration of gravity. Below between the powder and parts chipped
a certain number of revolutions, depen- from the balls can present a problem
dent on the mill and the amount and na- though.
ture of the filling, cascading occurs which
should also be avoided. Usually the mill is
2.3.4.2 Attrition, Vibro-, and Jet Milling
operated at 0.7-0.8 coc in the case of dry
milling and at 0.5-0.65 coc for wet milling. An attritor is a ball mill in which the
The balls usually fill half of the mill. The balls are agitated by a series of stirring
remaining space between the balls is filled arms mounted on an axial shaft. Attrition
with the powder. In the case of dry milling is reported to be more efficient than ball
about 25 vol.% of powder is added, usual- milling for obtaining a small particle size
ly together with about 1 wt.% of lubricant and also results in less contamination. The
(e.g., stearic or oleic acid). In the case of starting powder must be not too coarse
wet milling about 30-40 vol.% of powder though.
is used together with typically 1 wt.% of In a vibro-mill energy is injected into the
dispersing agent (e.g., dispex) in the mill by means of vibrations. This can be
milling liquid (e.g., water, alcohol, or hex- done with two- or three-dimensional mo-
ane). Quite long milling times of up to, tion. In this way the balls in the mill are
e.g., 100 hours are sometimes used. Several agitated and reduce the particle size of the
milling media exist and a compromise has feed. Both dry and wet milling can be car-
 2.3 Powders 43

ried out. Moreover, batch as well as con- considerable reduction in size can be
tinuous operation is possible. The typical achieved, e.g., an increase in specific sur-
frequency used is about 1 kHz. A final par- face of about a factor of three. However, a
ticle size of less than 1 jim is possible. 'grind limit' is often observed. Below a cer-
In a jet mill a powder is transported by tain size grinding only introduces impuri-
means of a fluid (steam, N 2 , CO 2 , H 2 O) ties and no longer breaks particles. Various
and blown against another moving particle explanations exist. Firstly, the existence of
or against a plate. In both cases autoge- dynamic equilibria during grinding: not
nous grinding is quite possible. Also, no only are particles made smaller, but the
moving parts are present. Hence the broken particles also tend to agglomerate
amount of impurities picked up during again. Secondly, the so-called brittle-plas-
milling is usually much less than with other tic transition: below a certain particle size,
milling devices. In addition, the size distri- particles can be deformed only by plastic
bution of the jet-milled powder can be deformation since not enough elastic ener-
smaller than for the ball-milled powder gy can be stored in a single particle to
due to the inherent classification process of break it. This limiting size, dc, is estimated
the method. by Kendall (1978) for a splitting mecha-
A disadvantage is that inhomogeneities nism as
in the density, e.g., due to second phases,
uc — olKhj5 i \^~v)
are difficult to remove. The product of
particle size and particle density deter- where R is the fracture energy, E the
mines whether the particle leaves the mill Young's modulus, and Y the uniaxial yield
or not. The difference in density and size of stress of the material. The numerical factor
the second-phase particles therefore yields depends on the assumed shape, and thus
different classification rates for the major should not be taken too literally. For shear
constituent and the second phase. In the failure a similar expression, but with a dif-
case of lower density of the second-phase ferent numerical factor, will be obtained.
particles, this abnormal classification re- For glass the limiting size is estimated as
sults in relatively large particles with a dif- 0.5 jam, while for CaCO 3 about 4.5 jim is
ferent composition, which in their turn obtained.
promote an inhomogeneous microstruc- An important drawback of milling is the
ture. The results of jet milling are depen- introduction of impurities. In compounds
dent on a number of conditions. Among of more complex composition, autogenous
these are the feed, the feed rate, the design milling is often impossible and a change in
and alignment of the classification cham- composition or the introduction of impuri-
ber, and the pressure, temperature, and ties is unavoidable.
volume. Apart from decreasing the particle size
Although the above-mentioned tech- of and introducing impurities into the
niques can have distinct advantages, ball powder, milling is also capable of changing
milling remains the most frequently used other characteristics of the powder. First-
technique in practice. ly, the process can introduce considerable
lattice strain in the particles. In commer-
2.3.4.3 Changing Characteristics cially available alumina powders, derived
The aim of milling is obviously to obtain from the Bayer process and dry ball milled
a smaller particle size in the powder. A for 24 h to a median particle size of about
44 2 Process Control in the Manufacture of Ceramics

0.5 |im, it was shown (by the Warren-Aver- 2.4 Consolidation

bach method) that strain domains of about
39 nm were present and the average strain The primary goal of every consolidation
amounted to about 0.0027 (Page et al., technique is to obtain homogeneity in all
1978). The corresponding dislocation den- aspects (see Table 2-2), combined with the
sity is estimated to be 1.0-2.0 x 10 11 cm" 2 , secondary goal of sufficiently high density.
which is approaching the dislocation den- A variety of consolidation techniques are
sity in cold-worked metals. Secondly, used since the advantages and disadvan-
milling can change the crystallographical tages differ per technique. They are divid-
structure. For crystalline quartz it was ed into essentially dry consolidation (press-
shown that an amorphous layer appears ing), consolidation with doughs (injection
during the (dry) milling process, resulting molding and extrusion), and consolidation
in lower density particles (Somasundaran, with slips (slip and tape casting). The char-
1978). When milled in water this amor- acterization of consolidated powders is
phization is unnoticed, probably because briefly discussed first. The most important
the amorphous layer is continually dissolv- aspects of the consolidation techniques are
ing. Polymorphic transitions (the change discussed in the following sections.
of one crystallographic structure to anoth-
er) have been reported as well. Thirdly,
2.4.1 Compact Characterization
milling can also induce chemical reactions.
This kind of reaction belongs to a field The characterization of compacts is a
called tribochemistry (Heinicke, 1984). difficult but important task. This is true
When Fe 2 O 3 is milled, Fe 3 O 4 can be for the macroscale but even more so for the
formed, showing up as a distinct change in microscale. Small inhomogeneities can
color. Another, more interesting, example have a large influence on the resulting
is the reaction of PbS and CdSO 4 to properties, particularly the mechanical
PbS0 4 and CdS, which occurs gradually properties. The relative density of com-
during milling (Somasundaran, 1978). pacts is often measured and used as a feed-
More relevant examples are the decompo- back to the process. More sophisticated
sition of sulphate, chloride, and phosphate analysis can be done by using compaction
precursors. The oxidation of carbides and curves and porosimetry, but this is seldom
nitrides during milling provides yet an- done on a regular basis. Nevertheless, they
other example. The amount of oxygen in provide useful information and these tech-
these ceramic powders is quite important niques will therefore be discussed and illus-
for subsequent processing to form ceram- trated below.
ics. All these effects are most pronounced As mentioned above the compaction
when the atmosphere in the mill is dry. process can be followed by measuring
Sometimes pyrophoric powders are ob- compaction curves. In such a curve the rel-
tained, e.g., in the case of metals or sul- ative density is plotted against the pressure
phides. In other cases the final composi- applied. An example for two different ag-
tion of the powder may change, e.g., in the glomerated zirconia powders is shown in
case of BaTiO3 which, when milled in wa- Fig. 2-3 (van de Graaf etal., 1983). Two
ter, loses Ba by the formation and dissolu- regions can be distinguished. This is often
tion of Ba(OH)2 (Ba leaching). interpreted as follows: In the low pressure
region the granules of the powder rear-
 2.4 Consolidation 45

range, settle, and deform slightly. At a cer- filled and during desorption evaporation
tain pressure relatively few voids are left takes place. A variety of models is in use,
between the granules. Upon a further in- each yielding slightly different answers,
crease in the density, the granules split up but all based on the Kelvin equation de-
into separate particles. The densification scribing the pressure, P, above a surface
of these particles is then described by the with radius of curvature r
second, high-pressure part of the curve.
Sometimes such a correlation is well sup- RT\n(PIP0) = (2-8)
ported by microscopic examination, but The vapor pressure of the liquid, with
frequently not. surface tension y and molar volume V,
The pore-size distribution can be mea- above a flat surface, is given by P o . The
sured by mercury intrusion porosimetry temperature and the gas constant are de-
(MIP). Mercury does not wet most solids. noted by T and R, respectively. Pore sizes
Pressure is thus needed for the mercury to in the range of 1-500 nm can be deter-
intrude into the pores. It should be noted mined. Various pore size distributions,
that cylindrical channels are assumed, but which were determined for zirconia pow-
that the entrance diameter of the pore is der compacts with weak agglomerates (al-
the quantity actually measured. The di- so used in Fig. 2-3), are shown in Fig. 2-7
ameter of the pores that can be filled with (van de Graaf et al., 1983). For compacts
mercury is dependent on the pressure, and pressed at 4 MPa (4 MN m~ 2 ), a bimodal
is usually described by the Washburn pore size distribution exists. Using 8 MPa
equation (Gregg and Sing, 1982) removes the bimodality, although a long
tail remains in the pore size distribution.
x>(2y cos9)/P (2-7)
For pressures above the knee in the com-
where x is the pore diameter, y the surface
energy of the mercury (usually taken as
0.485 J/m2) 6 the contact angle (usually
14
taken as 140°), and P the applied pressure.
Measurement of the intruded cumulative
volume of mercury determines the cumu-
lative pore volume. The size of the pores
E
12
10
8
400 M P a

1
95 8 & 4

'I ' *
/ V' \-
--
6
1>x xv
/ \ / •'A\ x
x
4

UK '
that can be determined ranges typically I\ ' - / \
<>i
4 /
from a few nanometers to a few microme-
2 / 8 \ . •
ters. n
From the adsorption isotherm, as deter- 5 10 50 100 500
pore radius (nm)
mined for the BET specific surface analy-
sis, a pore size distribution can also be cal- Figure 2-7. Pore size distributions, as determined
with the N 2 adsorption/desorption technique for zir-
culated (Lecloux, 1981). For this purpose conia compacts pressed at 4, 8, 95, and 400 MPa, and
the complete adsorption and desorption corresponding to the arrows in Fig. 2-3 for the
isotherm has to be recorded. The method propanol-2 washed samples. Two peaks are only ob-
is based on the greater tendency of a liquid served for the compact pressed at 4 MPa, clearly indi-
to evaporate from a flat liquid than from a cating the presence of agglomerates. At 8 MPa, below
the knee in the corresponding compaction curve, the
curved surface in a capillary. Hence evapo- voids between the agglomerates have already largely
ration from larger pores proceeds more disappeared. Higher compaction pressures shift the
quickly. During adsorption the pores are intra-agglomerate porosity to smaller sizes.
46 2 Process Control in the Manufacture of Ceramics

paction curve (95 and 400 MPa), the pore methyl cellulose. A few percent of binder is
size and the amount of pores decrease with usually used. Lubricants such as paraffin
increasing pressure. This example shows oil or a stearine solution are sometimes
that the results of the BET analysis are used to improve the lubrication, typically a
more or less in agreement with the expecta- few tenths of a percent. A pressure of 2 0 -
tions based on the corresponding com- 100 MPa is usually applied. For single-ac-
paction curve. tion presses the pressing rate ranges from
0.01 to 5 s ~ \ while for high-speed rotary
2.4.2 Pressing presses rates of up to 100 s" 1 can be
reached. The tolerance in mass that can be
In pressing we can essentially distin-
achieved in dry pressing is typically about
guish two techniques. Firstly, die pressing,
1%. In size a tolerance of 0.02 mm in
in which a certain amount of powder is put
thickness and about twice that in plan-par-
into a die and shaped by the punches under
allelism can be achieved. In wet pressing
a load. Since in this method the load is
much more water is added to the powder,
applied uniaxially, the method is also
typically 10-15 vol.%, and the use of
called (pseudo-)uniaxial pressing. Second-
binders is quite normal. The weight toler-
ly, isostatic pressing, in which a powder
ance is somewhat larger than for dry press-
batch is consolidated by an isostatic pres-
ing, i.e., about 2%. Problems that can
sure applied by a fluid on a preshaped
arise with die pressing are wear of the dies
compact, provided with an impermeable
(contamination and loss of size tolerance),
cover. Many technical details of isostatic
cracks in the compact, and density varia-
pressing are given by James (1983), while a
tions.
review of the theory is given by Broese van
Groenou (1982). In die pressing density variations almost
always occur over the compact. These vari-
ations are due to inhomogeneous filling of
2.4.2.1 Die Pressing
the die and the pressing process itself.
Die pressing can be practiced using a dry More homogeneous filling can be achieved
or a wet powder. In dry powder pressing by using a powder with increased flowabil-
the powder is used either 'really' dry or ity. The flowability of a powder is often
with the addition of up to 5 vol.% water to increased by deliberate agglomeration,
improve the adhesion of the powder. If this e.g., by spray-drying. Further improve-
is not sufficient (which is often the case), ment may sometimes be obtained by so-
binders are used. Here a choice can be called tap-filling: tapping is applied after
made between 'hard' or 'soft' binders. The (or during) filling of the die, before the
first class yields rather hard agglomerates punches are lowered to consolidate the
in the powder which make the powder compact. A simple improvement in the
free-flowing but not self-lubricating. The pressing process itself, to avoid these den-
latter class yields soft agglomerates which sity variations to some extent, is the appli-
result in flow problems but less lubrication cation of the compaction force from two
problems. Examples of hard binders are sides. A relatively new improvement is the
dextrine or acrylate, while wax and Arab use of ultrasonic energy (Boch and
gum are examples of soft binders. A com- Rogeaux, 1986). By applying ultrasonic
promise is achieved with 'intermediate' waves (typically 20 kHz) during com-
binders like polyvinyl alcohol (PVA) or paction, particularly in the early stages, a
 2.4 Consolidation 47
5mm.
significant improvement in the homogene-
ity can be achieved. In particular, the
larger voids are removed. This improve-
ment in the homogeneity can result in an
increased value of the Weibull modulus for
the strength of the sintered compact (see
Vol. 11, Chap. 10, Sec. 10.5.4.1 of this Se-
ries). Although significant improvements
may be obtained in this way, the density
variations due to the pressing itself remain
present and the following section will deal
with some aspects of this phenomenon. To
show the basic effects clearly, we will limit
the discussion to cylindrical shapes (Broese
van Groenou and Knaapen, 1980). In
pressing cylinder-like tablets the density
variations increase upon increasing the
height of the tablet. In the lower corners of
a cross-section (pressing from above) low- Cone up
er densities are present, while in the upper Umm.
corners a somewhat higher density is usu-
ally present. Introducing a depression in
the center part of the upper punch results
in two effects. Firstly, below the center
part two lobes of higher density show up,
and secondly, the discontinuity at the top
does not affect the bottom corners. The
presence of increased density is indepen-
dent of the exact shape of the depression.
Below a 'positive' corner an increased den-
sity is always present. If we change the
shape to 'negative' corner, however, the
density in that corner will be much lower
than the average in the bulk of the speci-
men (Fig. 2-8). The density variations due
to these positive and negative corners also Cone down
occur in more complex shapes. Figure 2-8. Density distribution in a cross-section of
pressed ferrite compacts. High numbers correspond
Higher density powder compacts can be to low density, while low numbers correspond to high
obtained by increasing the compaction density (the numbers given are actually the optical
pressure. Upon increasing the density, oth- density of photographs of the X-ray transmission of
er properties of the compacts, e.g., hard- the compact). A lobe of high density is clearly distin-
guished below the cone down (positive corner), while
ness, Young's modulus, and strength, also a lobe of low density is observed for the cone up
increase. (negative corner). Note also the difference in the den-
As well as the powder characteristics, sity in the upper and lower corners due to wall fric-
the press and pressing cycle characteristics tion.
48 2 Process Control in the Manufacture of Ceramics

are also important. These include the de- ally easier to use displacement control,
sign of the die, the powder feeding and though in both cases the nonconstant be-
cycle control including load or displace- havior will result in increased control lim-
ment control, the dwell time, and the ejec- its.
tion mode. An abrasion-resistant steel is
normally used for soft powders. A good 2.4.2.2 Isostatic Pressing
general choice is AISIA2 steel hardened to
In cold isostatic pressing (CIP) two
Rockwell C 58-60, while for small quanti-
modes can again be distinguished: dry-bag
ties AISI D2 steel hardened to Rockwell C
tooling and wet-bag tooling. In the wet-
61-63 is appropriate. Hard metal dies con-
bag tooling technique a preshaped com-
taining 3-17 % Co are used for more abra-
pact is put in some kind of (disposable)
sive powders. The qualities obtained with a
envelope and isostatically pressed. When
lower Co content and fine grains are a
the pressing cycle is finished a 'wet' bag
more wear-resistant material which also
containing the product is returned. The ad-
chips more easily. Larger-grained material
vantages are the relatively low cost and
with a higher Co content has more
versatility of the technique. The pressure
strength but a faster wear rate. A first
used ranges up to 500 MPa. It can be done
choice would be material with about
on a laboratory or production scale, and
13-15% Co with a trend to higher Co
on shapes of intermediate complexity. A
contents for more irregular shapes in order
disadvantage is the relatively small num-
to minimize failure (Magdic, 1984). Die
ber of compacts that can be produced in a
lifetimes vary considerably, but several
given period of time. In the dry-bag tool-
tenths of thousands of cycles are common,
ing technique a batch of powder is put in a
while values of over 100000 cycles have
flexible, preshaped mold (which is used
been reported for particular cases. Some-
many times) and pressed isostatically.
times the die is coated with an abrasive
When the pressing cycle is finished only a
resistant layer, e.g., TiN or TiC. For soft
dry product is returned, since the bag re-
powders this may increase the lifetime con-
mains in the equipment. The name 'dry
siderably, but for the usual ceramic pow-
bag' is thus somewhat misleading. The
ders the cost of coating may be too high
tooling is durable, short cycle times are
compared with the lifetime improvement.
possible, and large numbers of specimens
For the control of the press cycle two
can be made. The mold can be made from,
options exist: load control or displacement
e.g., polyurethane, synthetic rubber, or sil-
control. If the behavior of the powder over
icone rubber. For the dry-bag technique a
various batches is constant, the choice is
lower pressure is generally used compared
relatively unimportant from the product
with the wet-bag technique, e.g., up to
point of view. Both methods have their
200 MPa (200 MN m~ 2 ). The most well-
technological advantages and disadvan-
known industrial example is the ceramic
tages. For nonconstant powder behavior,
part of a spark plug, which is typically
however, load control pressing will give
pressed for cycle times of 1-2 s. In both
green products with approximately con-
modes of isostatic pressing, several kinds
stant green density but with varying size,
of defects can be present. They are shown
while displacement control pressing will
in Fig. 2-9 (Morris, 1983).
give a constant green product size but with
different densities. In practice it is gener-
 2.4 Consolidation 49

Y~7 Advantages of the technique are: compli-

cated shapes are possible, and tight size
tolerances and homogeneous microstruc-
tures can be achieved. Disadvantages of
the technique are the cost (mixture, dies,
a b c d
binder burn-out), long product develop-
ment times, and inflexibility. Consequent-
ly, the technique is appropriate for rela-
tively expensive products which have to be

V
1 produced in large numbers. Three aspects
of the production process are considered in
more detail in the following sections: prop-
ZZ erties and mixing of the raw materials, the
e f g h molding process itself, and the removal of
Figure 2-9. Defects in isostatically pressed compacts:
the organic part after shaping. Many more
(a) necking due to underfill or uneven fill, possibly details can be found in German (1990).
originating from poorly flowing powder, (b) irregular
surface due to uneven powder fill or unsupported bag,
(c) 'elephant's foot' due to rigid closures in wet-bag 2.4.3.1 Properties and Mixing
tooling or to a highly compactable powder, (d) 'ba-
nana' due to unsupported bag in wet-bag tooling, (e) The most important properties of the
compression crack due to axial springback, typical for mixture are its processability and the con-
hard powders, (f) lamination due to compression tent of ceramic material. The latter deter-
cracking originating from unsuitable or too thick bag mines the shrinkage during sintering. The
material or weak compacts, (g) irregular surface due
to unsuitable or too thick bag material, weak com-
processability is dependent on the viscosity
pacts, or small corner radii, and (h) axial cracks due of the mixture as a function of shear rate
to insufficient elastic springback (in tubes). and temperature. The viscosity as a func-
tion of shear rate must show pseudo-plas-
tic or Bingham-plastic behavior. The vis-
cosity is also dependent on the volume
2.4.3 Injection Molding
fraction of solid material, as well as the size
The basic concept behind the injection (distribution) of the ceramic powder. An
molding of ceramic parts was to put the approximate relation between the relative
numerous possibilities available in plastics viscosity, rjr = f]/r]0 where Y\0 is the viscosity
technology to good use in ceramics as well. of the pure liquid, and the volume fraction
In ceramic processing a thermoplastic ma- of solid material, / , is the Dougherty-
terial is mixed with the ceramic powder. Krieger relation given by
The process consists essentially of three
rjr = (l-f/kpyck» (2-9)
steps: First, the filling of a relatively cold
mold with the hot melt of the feedstock, where kp is an empirical constant, which is
that is, a thermoplastic material mixed expected to be approximately equal to the
with the ceramic powder. Second, cooling maximum volume packing fraction of sol-
and solidification of the melt and, thirdly, id (kp~0.6-0.7). The parameter c in the
ejection of the formed product. After the exponent is the intrinsic viscosity defined
injection molding process the organic part by the limit of (nr-l)/f for /approaching
has to be removed, usually by burn-out. zero. For a dilute suspension c = 2.5. In
50 2 Process Control in the Manufacture of Ceramics

practice c has a slightly higher value (usu- The mixing of the powder and the ther-
ally 2.7), which is dependent on the size moplastic material is of the utmost impor-
distribution of the powder (Hunter, 1987). tance. In principle it consists, after cold
For agglomerated powders the factor f/kp blending of the ceramic powder and the
should read//(£ a £ p ), where ka is the pack- thermoplastic material, of hot kneading
ing density of the particles in the agglomer- (at about 200 °C) of the mixture. A good
ate, since agglomerates move as a whole. deagglomeration of the ceramic powder is
The viscosity increases rapidly with the essential. If the powder is not sufficiently
volume fraction. For multimodal size dis- deagglomerated, the processability of the
tributions much higher volume fractions molding mixture and the reproducibility of
of solid material can be reached than for a the molding process will be poor, and the
unimodal size distribution at the same vis- product quality will become insufficient
cosity (Farris, 1968). The demand for nar- (e.g., cracking during binder burn-out, in-
row size distributions from the sintering homogeneous microstructure). After de-
point of view, however, seriously compli- airing in the last stage of the kneading, the
cates the use of multimodal distributions. hot dough is cut into pieces and, after
In general, at a constant volume fraction cooling, it is milled to a granule size of
the viscosity increases with decreasing par- 1-4 mm.
ticle size. Volume fractions are usually be- The binder recipe is probably the most
tween 30 and 70 %, dependent on the pow- important factor for successful injection
der morphology and the nature of the ther- molding. Aspects of the binder recipe for
moplastic material. The influence of the injection molding are given in Table 2-5.
shear stress on the viscosity can be taken Obviously, the exact composition of the
into account as well (Hunter, 1987). binder recipe is often confidential.

Table 2-5. Aspects of the binder recipe for injection molding.

Components Comments

Powder oxide, nonoxide, metal

+ Thermoplast polymer
resulting in - thermoplastic material
- strong but brittle product
- low solid volume fraction
- macroscopic inhomogeneity
- microscopic inhomogeneity
Weakener oil-like components
resulting in - higher degree of filling
- more elastic product
- still inhomogeneous product
4- Lubricant wax
resulting in - higher degree of filling
- macroscopic homogeneity
+ Surface active components —> detergent
resulting in - microscopic homogeneity
 2.4 Consolidation 51
2.4.3.2 Molding The molding pressure during the injec-
The granules enter the injection molding tion molding cycle is generally closely
machine via the hopper. They are trans- monitored. For injection molding, full
ported by an auger (Archimedes screw) or closed-loop controlled equipment is avail-
plunger via a hot zone, where the granules able relatively cheaply at present.
melt, to the die. Here the material solidifies
because the temperature of the die is far 2.4.3.3 Binder Burn-Out
below that of the molten mixture. On com- After the molding process the organic
pletion of the solidification process, the part of the product has to be removed.
formed product is ejected. The distribution This can be done either by elutriation or by
channel for the distribution of the hot mix- combustion (binder burn-out). The latter
ture to several dies can be cold or heated. method is usually applied. A typical tem-
In the first case the material in this channel perature-time curves is given in Fig. 2-10.
is also ejected and (possible) re-used. In the During binder burn-out, carburization of
latter case the material in the channel does the organic compounds, which can easily
not solidify, and so higher demands are occur, should be avoided. In the initial
imposed on the mixture. Typical cycle stage of the burn-out process the injection-
times are 15 s in the case of cold distribu- molded compact becomes soft. This can
tion channels and 5 s in the case of heated lead to compact shape deformation due to
ones. The process parameters are the tem- its own weight (pyro-plasticity). Mechani-
perature of the mixture, material trans- cal cracking should also be avoided.
port, pressure in the die, and cooling rate. Cracks may be due to residual stresses
These parameters are dependent on ma- from the injection molding process (differ-
chine parameters, e.g., the injection tem- ential shrinkage, particularly around holes
perature, injection pressure, injection rate, in the product, and in thick parts; see
and design of the die. The last parameter in Fig. 2-11), or too rapid evaporation and
particular is much more important than in decomposition of the binder, possibly en-
the plastics industry because of the higher hanced by the exothermic nature of the
viscosities, higher wear rates, and higher burn-out process. Shape and thickness
thermal conductivity of ceramics com- play an important role. Thicker products
pared to pure thermoplastic materials. The require lower heating rates. Binder burn-
design of the die also determines whether
or not 'jetting' occurs. Jetting is the injec-
tion of the feedstock into the die in such a
way that progressive filling from the inlet
does not occur, and instead, when a jet of
the feedstock is sprayed into the die,
nonuniform filling of the die occurs, and
this inhomogeneous filling is reflected in
the microstructure of the final product.
Jetting can occur most easily if the inlet is Figure 2-10. Typical temperature-time curve used
positioned in front of a thick cross-section, for binder burn-out of injection-molded parts. At low
though it can be inhibited or reinforced by temperatures a low heating rate is applied, while at
the physical parameters of the feedstock higher temperatures a higher heating rate can be ap-
(Mangels and Trela, 1984). plied.
52 2 Process Control in the Manufacture of Ceramics

1 product. While for relatively inert materi-

V als like alumina temperature control of the

furnace used will suffice, more extensive
process control is required and performed
for more sensitive materials like BaTiO3.
Figure 2-11. Schematic diagram of typical defects in Monitoring of the oxygen content of the
injection-molded parts. Problems often occur at cor- exhaust gas is very helpful in the latter
ners and in the middle of thick sections. case. The process may be interrupted or
adjusted if the partial oxygen pressure ex-
out is often the limiting factor with respect ceeds certain specified limits.
to the degree of complexity and thickness
of the product. However, if properly exe- 2.4.4 Extrusion
cuted the injection molding process can While injection molding has its origin in
yield quite homogeneous compacts in plastics technology, extrusion originates
complicated shapes. An example of a zir- from metallurgy. In this process a piece of
conia razor knife is shown Fig. 2-12 (de metal is pressed through an orifice. In the
With and Witbreuk, 1993). case of ceramic extrusion, the piece of ma-
The control of the burn-out of the terial is a dough made from ceramic pow-
binder varies considerably from product to der and a binder. Lubricants and deter-
gents are added as well. Development of
the optimal recipe is quite analogous to the
injection molding recipe and thus also
largely empirical. After putting the feed in
the extrusion machine, the material is
transported to a de-airing chamber to re-
move air bubbles. The de-aired dough is
pressed into the extrusion chamber
through the extrusion nozzle. The neces-
sary force for extrusion through the nozzle
can be delivered by an auger (Archimedes
screw) or a simple plunger. While the for-
mer has the advantage of low forces, the
amount of impurities introduced is in prin-
ciple higher than by the use of a plunger
because of the high shear rates at the inter-
face with the screw. This disadvantage can
be circumvented by coating the screw with
a ceramic coating of the same composition
as the extrusion mass, although the limited
lifetime of this coating may provide anoth-
Figure 2-12. An example of a complicated injection- er problem.
molded shape. The part shown is a razor knife of
yttrium-doped tetragonal zirconia (Y-TZP). The ma-
terial has a relative density of 99% and a grain size of
about 0.3 urn.
 2.4 Consolidation 53

2.4.4.1 Process Control components is heavily dependent on the

component. Continuous microwave dry-
The extrudability of a mass depends on ing is a convenient way nowadays.
many parameters, related to the raw mate- Serious defects can be introduced into
rials or the equipment. The raw material the extruded mass if no optimum is found
properties involved are: among all the relevant parameters. The
1) The adhesion to the wall. This should most important ones are laminar defects,
be minimised in order to reduce the fric- surface and edge defects, and splitting. The
tion between the wall and the mass. latter is the occurrence of long axial cracks
2) The internal friction of the mass. through the entire product. Laminar de-
High internal friction results in a small, fects arise from the laminar flow in the
realizable turn-down ratio, i.e., the ratio extrusion process, possibly enhanced by
between the auger diameter and the die the particle shape characteristics. A better
diameter. design of the die and/or better kneading of
3) The cohesion of the mass. The lower the dough can remedy these defects. Sur-
the cohesion of the mass, the more likely face and edge tearing is mainly due to in-
the appearance of defects will be and the terfacial friction between the extrusion
less plastic strain the product can endure mass and the die, while splitting is due to
before breaking. an improper balance between friction and
4) The particle shape and size. Plate-like cohesion of the mass.
shapes will introduce a laminar structure For extrusion, as well as for injection
into the extruded product, so increasing molding, it is important to use a dough
the change of lamination defects. A which is as homogeneous as possible.
smaller particle size will improve the extru- Apart from implying good mixing of the
sion behavior. A multimodal particle size constituents, this also implies the absence
distribution also improves the extrusion of air bubbles, since their presence often
characteristics. results in a defect in the final product. Ex-
The suitability of the dough is often amples of these defects in extruded alumi-
characterized by its yield value. The yield na are shown in Fig. 2-13.
value is conveniently measured by pressing Because the extrusion process is largely
the dough through a specified orifice. A inflexible and a delicate balance between
typical example of the batch-to-batch many factors, it is primarily used for large-
variation for alumina pastes is given in scale operation. Apart from simple tubes
Fig. 2-4. From this figure it can be calcu- and rods, more difficult forms have been
lated that a sample standard deviation of made as well. The most appealing exam-
10% can be achieved. ples are probably the honeycomb struc-
The machine parameters are primarily tures made from low thermal expansion
related to the auger and die design. As- materials, e.g., Li-Al- and Mg-Al-silicate
pects of importance for the auger are the for heat exchangers and catalyst support
pitch, taper, number of wings, and speed. (Katz, 1980), or from modified BaTiO3
For the die the relevant aspects are the [positive temperature coefficient of resis-
taper, length, entrance angle, and the exact tance (PTCR) ceramic] for heater element
positioning of the die. Finally, the turn- applications (Wada, 1980).
down ratio and the de-airing procedure are
of importance. The method of drying the
54 2 Process Control in the Manufacture of Ceramics

a) other modifications are more frequently

used in factory practice. In all casting tech-
niques use is made of a slip that is poured
into a porous mold through which the
solvent is removed either by capillarity or
by external pressure. Fairly complicated
shapes can be produced (by using multi-
part molds) at relatively low cost, but the
production speed is low. The technique
can easily be applied in the laboratory and
also in industrial practice. For example,
sanitary porcelain is largely shaped by the
slip casting process.
b)

2.4.5.1 Normal Slip Casting

Slip casting is a rather simple but flexi-
ble technique. A slip can be defined as a
high volume fraction (45-60%) of solid
particles in a liquid (usually water), made
fluid by the addition of one or more defloc-
culants. Typical deflocculants are ammoni-
um polyacrylate, oleic acid, and carboxy-
methylcellulose. The slip is cast in a mold.
The walls of the mold extract the solvent
from the slip, thereby forming the compact
along the walls of the mold. After some
time, when the wall thickness of the com-
pact is large enough (or when the growth
of the wall thickness is largely stopped by
saturation of the mold), the remaining slip
is drained. The casting can then be re-
moved after partial drying. Hollow com-
pacts are normally produced. If a slip is
Figure 2-13. Defects in extruded alumina tubes for used which shows a gel-like behavior, it is
lamp application: (a) so-called 'white spot', which is
possible to leave the slip in the mold until
actually a pore due to an air bubble in the dough, (b)
so-called 'stripe', which is actually an array of pores it is (partially or wholly) gelled and to pro-
due to air entrapment, and (c) nonlinearity. duce solid compacts (gel casting). The rhe-
ological and casting properties of the slip
mainly depend on the type of solid used,
the particle size distribution and specific
2.4.5 Slip Casting surface of the powder, the pH of the slip,
For slip casting various modifications and the modifiers added.
exist. Normal slip casting is most frequent- The most common method of making a
ly used in laboratory practice, while the slip is wet ball milling. It is preferable for
 2.4 Consolidation 55

the slip to show thixotropy (shear thin- The most important measurable quanti-
ning) and have the maximum density pos- ties in casting are the viscosity and the
sible. This can most easily be obtained yield value. Consequently these parame-
with a wide particle size distribution. A ters are measured frequently, also as a
reasonably high density can be obtained function of time, since ageing may signifi-
by using the Andreasen cumulative size cantly alter the values. Usually this is due
distribution F(x) given by to flocculation and can only be solved by
using a more stable colloidal system.
= (x/X)m (2-10)
The mold is usually made of plaster of
where x is the particle size, X is the maxi- Paris (CaSO4 • Vi H 2 O). Upon the addi-
mum particle, size and the exponent m is tion of water the plaster hydrates, forming
between lA and lA (Gray, 1968; German, an interlocked mass of gypsum (CaS0 4 •
1990). However, sintering demands a nar- • 2H 2 O) needles. Although the theoretical
row size distribution, in which case the slip amount of water is only 0.185 kg water per
tends to become dilatant (shear thicken- kilogram of plaster, the practical range is
ing). During preparation of the slip the 0.4-1.2 kg water per kilogram of plaster.
viscosity may change from about 0.1 Pas The amount of water has a large influence
in the initial stages to 3 or 4 Pa s in the final on the absorption capacity of the mold,
stages of slip preparation. Ageing is often e.g., a mixture of 0.4 water: 1 plaster can
necessary in order to reach a constant vis- absorb an amount of water equal to 20 %
cosity with time. Although de-airing is eas- of its mass, while a mixture of 0.9 wa-
ier for slips than for doughs, it is an essen- ter : 1 plaster can absorb about 50 %. A
tial step for reaching homogeneity in the mold release agent is sometimes used. Typ-
cake. ical examples are graphite, talc, oil, and
The thickness of the deposited layer can starch. Although in principle the homoge-
be estimated quite accurately from theory neity of the resulting compact is high, in
(Tiller and Tsai, 1986). Taking into ac- some cases, if a sufficient wall thickness
count the limited permeability of the mold, cannot be reached by a single filling of the
Km, and the compressibility of the deposit- mold, the interface between the first and
ed layer, a parabolic dependence of the second filling of the mold can easily be
thickness of the layer, L, on time was spotted in the final product (Fig. 2-14). To
derived obtain consistent performance it is neces-
(25-11) sary to pay strict attention to the factors
2
L = that govern the physical and colloidal
properties of the slip. Moreover, the pro-
where Kc is the permeability of the cake,/ c cessing conditions (powder, additions,
the volume fraction of solid in the cake, / s mixing, casting rate, and temperature)
the volume fraction of solid in the slip, P should be kept as uniform as possible.
the pressure, and rj the viscosity of the slip.
The porosity of the cake appears to be
2.4.5.2 Other Casting Techniques
critical in terms of the deposition rate and
can be influenced through the degree of Modifications of the slip casting tech-
flocculation of the slip. For each type of nique are pressure casting, vacuum cast-
slip there is an optimum pore size for max- ing, centrifugal casting, and ultrasonic
imum cake build-up. casting. In the first modification the capil-
56 2 Process Control in the Manufacture of Ceramics

Figure 2-14. Cross-section of a slip-cast part of porous SiC, showing an inhomogeneous density distribution due
to repeated casting.

lary forces are replaced by a (physical) 2.4.6.1 Doctor Blade Processing

pressure on the slip. The slip is cast in a
In the case of the doctor blade process
mold provided with an air-tight closure
a slurry is made from the powder and a
and the air inside is pressurized. This
solvent. Some detergents are usually added
makes control of the thickness much eas-
to improve the dispersion of the powder.
ier. In vacuum casting the outside of the
After sufficient deagglomeration and dis-
mold is evacuated, also resulting in a phys-
persion, a binder is added. The slurry thus
ical pressure on the slip, necessarily limited
obtained is freed from air bubbles and put
to 1 bar (10~ 5 N m~ 2 ). Centrifugal casting
in a reservoir in front of a knife edge
can be used for complicated shapes. Spin-
(blade). At the bottom of the reservoir a
ning the mold ensures that the slip com-
carrier tape is pulled along the blade. After
pletely fills the mold, even in small chan-
the passing of the blade a film of slurry
nels and/or in protruding parts. In ultra-
adheres to the carrier film, which is subse-
sonic casting ultrasonic waves, usually in
quently dried by forced heating and rolled
the kilohertz range, are used to increase
upon a take-up reel. Nowadays, the carrier
the density, to improve the rheological be-
tape is usually absent and the slurry is di-
havior of thixotropic slips, to promote
rectly transported by an endless stainless
rapid air bubble removal, and to improve
steel belt. To avoid sticking after drying the
the uniformity. Electrophoretic deposition
film, the belt is cleaned and coated with an
is related to these techniques. In this tech-
antisticking agent (e.g., lecithine) before it
nique the driving force for consolidation is
enters the reservoir. Fairly thin films can
provided by an electric field. Homoge-
be made in this way. Thickness values can
neous products with a relatively thin wall
range from a few micrometers up to
thickness can be produced in this way.
1000 [im. The thicknesses are usually be-
tween 50 and 500 \im. Applications of tape
2.4.6 Tape Casting
casting are found mainly in electronics:
Ceramic tapes can be produced in sever- substrates for electronic circuitry, dielec-
al ways. The most well-known is the so- trics for capacitors, and piezo-electrics for
called doctor blade process. Less well- buzzers, igniters, and acutators. In these
known are the paper-casting and rolling applications binder burn-out and the
processes. stocking of foils is generally important
 2.4 Consolidation 57

(Fig. 2-15). The solvent/binder system and a)

the casting/drying process are discussed
below in some detail.

2.4.6.2 Solvent/Binder System

Several requirements exist for the
binder: It should form a flexible yet strong
enough tape when dried, leave little or no
impurities after firing (carbon!), be soluble
in a proper solvent, and preferably not too
expensive. Water or an organic liquid can
be used as the solvent. For environmental b)
reasons there is trend toward aqueous sys-
tems. Poly(vinyl alcohol) is the most com-
monly used binder for water-based sys-
tems. The choice of organic liquids may
depend on the thickness desired. For layers
up to 10 or 20 jim, methyl ethyl ketone can
be used as the solvent, with poly(vinyl ace-
tate) as the binder. For thicker layers
toluene or trichloroethylene can be used,
together with poly(vinyl butyral) as the
binder. However, other systems can also
be applied successfully. In many cases one
or more plasticizers are added in order to c)
obtain a greater flexibility for easy han-
dling after casting and drying. Examples
are poly(ethylene glycol) for toluene-based
systems and butyl benzyl phthalate for
methyl ethyl ketone systems. Again, many
other organic compounds may also suffice.
Finally, a deflocculant is often added in
order to keep the powder well dispersed.
Glyceryl esters of fatty acids are suitable
for this purpose, e.g., glyceryl oleate or
glyceryl stearate. The slip is usually pre-
Figure 2-15. Defects in ceramic multilayer capacitors
pared in two steps: First, milling of the
(CMCs), shown in cross-section. The electrode spac-
powder in the solvent without any binder ing is 20 urn. (a) Inner delamination due to uncon-
and plasticizer added and second, mixing trolled binder burn-out, (b) misalignment of electrodes
the slip obtained with the binder and plas- due to stacking error, and (c) large pore due to incom-
ticizer. These components are usually add- plete de-airing of the tape-cast foil.
ed after milling because the high viscosity
of the slip after addition of the binder pre-
vents proper milling.
58 2 Process Control in the Manufacture of Ceramics

2.4.6.3 Casting/Drying Process occurs in roughly two stages: In the first

stage the drying rate is approximately con-
After the slip is prepared, it is de-aired
stant. The slip is still fluid and the liquid
under a vacuum of a few millibars with
can be transported easily to the surface of
gentle stirring. De-airing is essential be-
the tape by liquid diffusion or capillary
cause the remaining bubbles can easily lead
action. Skin formation should be avoided.
to pinholes in the final foil (Fig. 2-15 c).
In the second stage the drying rate is grad-
Then the slip is sieved to remove large
ually decreasing. This step is limited by the
powder particles or undissolved binder.
inflow of heat into the tape.
The viscosity after these procedures is typ-
Apart from the viscosity, the solid con-
ically 3 Pas. The opening between the
tent in the as-cast type is of great impor-
transport belt and knife edge should be
tance for the control of the final thickness.
well controlled, e.g., by micrometer screws.
A typical example of the solid content over
Sometimes a dual blade configuration is
various batches of foils is given in Fig. 2-5.
used to improve the definition of the tape.
From this figure it is clear that a spread of
In this case, a temperature-controlled trans-
1 % is achievable. Since this parameter is
port belt is also often used. The thickness
the main one for thickness control after
of the tape, L depends on the viscosity of
sintering, narrow control limits are im-
the slurry, rj, the speed of the tape v, and
posed.
the opening between the knife edge and the
carrier tape, h. The speed of the tape is
usually from 0.2 cm/s to 5 cm/s. The thick- 2.4.6.4 Other Tape Processes
ness of the tape can be estimated from first
principles, as done by Chou et al. (1987). A modification of the tape-casting pro-
These authors assumed a Newtonian flow cess is the paper-casting process. A low-
for the slip and by applying the solutions ash paper is pulled through a reservoir
of the Stokes' equations for pressure and which again contains the slurry. After dry-
Couette flow, they obtained ing, the coated paper is rolled on a take-up
reel. Definition of the surface of the tape is
L = (apQh/2Qi)[l+(h2API6rivT)] (2-12) less than in the Doctor Blade process.
A dough is used in the roll process, in
where T is the thickness of the knife, AP contrast to the other tape processes, which
the pressure difference, Q the density of the makes it somewhere intermediate between
slip, and gt the density of the tape (after casting and extrusion. The dough is made
drying). The width correction factor a from the ceramic powder and an organic
takes into account that the as-deposited binder (and probably also some deter-
slip flows sidewards. It is expected to be gents). The dough is put in front of rollers,
close to one for wide knives. The packing which deform it into a band. This rolling is
correction factor /? corrects for the weight usually done more than once in order to
loss during drying and is expected to be obtain the correct thickness. On reaching
close to 0.6; the volume fraction for close the desired thickness, the tape is rolled on-
packing. to a reel. The thicknesses that can be ob-
The tape is dried directly after casting. tained with this technique are somewhat
Drying is performed by a flowing (filtered) larger than can be obtained with tape cast-
air current counter to the direction of the ing, i.e., a few hundred micrometers or
tape and by heating. The drying process larger.
 2.5 Sintering 59

2.4.7 Comparison of the Various Nearly all the techniques are thus used in
Consolidation Methods the laboratory and the factory. The sophis-
tication of the equipment used in the two
After the description of the most impor-
tant consolidation techniques, a critical situations is, however, quite different.
appraisal seems appropriate. Injection
molding, extrusion, and slip and tape cast-
ing all require a largely empirical, precon-
2.5 Sintering
solidation step. In this preparational step
Various control aspects can be distin-
of the feed dispersion, deflocculants, and
guished during sintering. These include the
wetting or other agents play an important
various types of furnaces in use as well as
role. The purpose of this step is to improve
their controlling parameters. The impor-
the rheological behavior of the feed in such
tance of these control aspects is governed
a way that optimum process control is re-
by the requirements put upon the micro-
alized. The result can be reasonable to
structure of the final material (Table 2-2).
good homogeneity of the consolidated
product. Powder to be used for pressing,
2.5.1 Furnaces
on the other hand, requires little prepara-
tion, but the homogeneity of the product is Sintering is done in various types of fur-
generally far from optimal. In a more de- naces. Tube and chamber furnaces are
tailed comparison, aspects to be consid- used up to about 1500°C. In pipe furnaces
ered are the feed, dimensional degrees of atmospheric control is relatively simple,
freedom, homogeneity of the product, but the sample size is limited. The reverse
speed, and cost. is true for chamber furnaces. For tempera-
A comparison of the consolidation tech- tures up to 1200°C the furnaces are fre-
niques is given in Table 2-6. From this quently equipped with Kanthal elements,
table it is clear that for laboratory experi- i.e., a metal alloy containing silicon, so
ments pressing, slip (and sometimes tape) that a protective SiO2 coating results dur-
casting, and extrusion are feasible tech- ing use. This protection mechanism is only
niques. From the production point of view useful in air. For temperatures above
injection molding, die pressing, isostatical 1200°C and up to 1600°C, SiC heating
pressing, tape casting, and extrusion are elements and so-called 'super Kanthal', ac-
the techniques to be considered primarily. tually MoSi 2 , elements can be used. Again,

Table 2-6. Comparison of consolidation techniques.

Technique Feed DF a Homogeneity Speed Cost c

Die-pressing powder 2.5 b bad moderate low

Iso-pressing powder 2.5 b moderate moderate moderate
Extrusion dough 1.5b moderate high moderate
Slip casting suspension 2.5 b reasonable low low
Tape casting suspension 1 good high moderate
Injection molding dough 3 good moderate high
a b
D F : dimensional degrees of freedom; shape limitations exist so that full 2D and 3D cannot be reached;
c
heavily dependent on exact shape.
60 2 Process Control in the Manufacture of Ceramics

in both cases a protective coating of SiO2 peratures the less sensitive B-couple (Pt/
is operative in use. Operational difficulties 6 % Rh-Pt/30 % Rh, up to 1700 °C) or even
with the furnaces increase rapidly above exotic types like B4C-C (up to 2200 °C) are
1500°C. A modern type of furnace is the applied. In pyrometry the use of two wave-
vacuum furnace, equipped with either a length pyrometers should be advocated.
carbon or a refractory metal (Mo, W, Ta) Single wavelength pyrometers rely on a
heating element. Carbon elements can re- clean window throughout the process.
sult in a fairly pure, but not so low, vacu- This requirement is not always satisfied.
um (about 10 ~2 mbar). The use of refrac- While the use of thermocouples and py-
tory metal elements, on the other hand, rometry is clear, the maximum tempera-
can result in a much better vacuum [about ture indicators may need some clarifica-
10~ 4 mbar (10" 2 N m~ 2 )], but vaporiza- tion. It has been shown to be possible to
tion of the element is easier than for car- make materials for which the shrinkage is
bon. In this type of furnace a temperature determined quite reproducibly by the
of 2500 °C, in the case of a tungsten ele- highest temperature encountered. One
ment, or 3000 °C, in the case of a carbon type of products, usually a ring or a cone,
element, can be reached without undue has a temperature measuring range of
complications. More exotic is the furnace about 150°C. In general, each degree cor-
equipped with a ZrO 2 element, which can responds to 0.01 mm shrinkage. The cones
be used in air up to about 2300 °C. Anoth- are known as Seeger cones. In rare cases
er type of laboratory furnace uses heating the controlled parameter is a relevant
via infrared lamps. Here a temperature up physical parameter rather than the temper-
to 1700°C can be reached with extremely ature, e.g., the shrinkage during sintering
high heating and cooling rates. Slightly or the weight loss during debinding. In this
lower rates can be achieved by induction case the processing is often denoted as
heating. However, the necessity of a sus- rate-controlled processing. For sintering
ceptor limits the heat input in the actual this type of control has been advocated
material to be sintered. Moreover, temper- notably by Palmour and Hare (1987).
ature control is considered to be difficult. The temperature distribution is of great
In industry (large) chamber furnaces are in importance. A homogeneous temperature
use, usually equipped with a feed-through distribution is necessary to obtain homo-
system. Heating elements of SiC are most- geneously densified compacts. The tem-
ly used, although other heating systems are perature homogeneity in a furnace is often
in use as well. overestimated. The temperature difference
at various places in the furnace can be as
high as 50 °C. It is rarely below 10-20 °C
2.5.2 Temperature
over the entire furnace. The homogeneity
For sintering the most important con- can be conveniently checked by the maxi-
trol parameter is obviously the tempera- mum temperature indicators.
ture. Virtually all 'normal' temperature
measurement methods are in use, i.e., the
2.5.3 Atmosphere and Additives
use of thermocouples, pyrometry, and
maximum temperature indicators. The S- With respect to the atmosphere, there
couple (Pt-Pt/10% Rh) is most frequently are considerations to be taken into account
used up to about 1500°C. At higher tem- with regard to both the product and the
 2.5 Sintering 61

environment. Control of the environment ties considerably. For example, the addi-
includes the water/oxygen content or, more tion of just 40 ppm CaO to translucent alu-
generally, the partial pressure. While, in mina, used as an envelope material for Na
general, a constant environment is the aim, discharge lamps (Fig. 2-13), results in a
the gas atmosphere is sometimes changed drastic decrease in the resistance towards
during heating and/or cooling in order to Na at elevated temperatures (de With et al,
control the defect chemistry of the materi- 1985).
al; this is known as stoichiometric process- Another possible method of controlling
ing. the grain size is the use of seeds. In this case
To avoid volatilization of certain com- minute particles of the parent compound
ponents, a powder bed or a capsule is used. (seeds) are added to the sintering powder.
In the former method a poorly sintering Although the mechanism differs depend-
powder is mixed with the product powder ing on the material, the effect is a more
or a powder containing the relevant com- homogeneous microstructure by con-
ponent. The green product is sintered trolled grain growth. The technique has
while embedded in this powder. Evapora- been successfully pursued for a number of
tion of the volatile component from the materials, e.g., BaTiO3 (Hennings et al.,
powder bed suppresses the volatilization 1987), ZnO (Hennings et al., 1990), A12O3
of that component from the product. A (Kumagai and Messing, 1985), and zirco-
well-known example is the sintering of nia toughened A12O3 (Messing and Kuma-
Pb(Ti, Zr)O 3 products in a bed of gar, 1989).
PbZrO 3 . In the capsule method the prod-
uct is simply put in a capsule of an inert
metal or glass so that a slight evaporation 2.5.4 Hot-Pressing
or decomposition suffices to provide a Hot-pressing or pressure sintering is
high enough partial pressure to stop the done either in dies or isostatically. The for-
evaporation or decomposition. mer is usually referred to as hot-pressing.
Additives are frequently used to en- This technique is used on a laboratory
hance the sintering and/or microstructure scale as well as on a factory scale. For the
control. The best-known example is prob- latter, BeO-doped SiC yielding an electri-
ably the addition of a few hundred ppm of cally isolating, though highly thermal con-
MgO to A12O3, in order to make the mate- ducting, ceramic is a recent example
rial translucent. The various opinions pre- (Takeda, 1988). If the mold is omitted and
sented in the literature, concerning this do- the material is pressed between the dies
pant behavior, seem to converge towards alone it is referred to as (press-)forging. In
an MgAl 2 O 4 second phase at the triple this case a significant texture is frequently
points at high dopant levels, and to slight developed. Separation of the product from
segregation at the grain boundaries, prob- the mold and/or dies may require special
ably as nonstoichiometric MgAl 2 O 4 at dis- precautions. A separator material, e.g., a
crete spots (Dorre and Hiibner, 1984). nonsintering powder, is often used. In the
Many ceramics additives are used and each case of hot isostatic pressing (HIP) an inert
seems to have its own specific action. It gas is used as the pressure medium. The
should be noted that in a number of cases compact must either be sintered to closed
the addition of minor amounts of additive porosity or it must be 'canned'. This can-
can change the microstructure and proper- ning can be made of stainless steel, a noble
62 2 Process Control in the Manufacture of Ceramics

metal, or a high viscosity glass. The equip- 2.6.1 Grinding

ment is usually pressurized to a limited ex-
tent at room temperature. An increase in As stated in Sec. 2.2, machining can be
pressure is achieved by the rise in tempera- done in various ways. Here we limit
ture. If a compact already sintered to high ourselves to abrasive processing. Detailed
density is isostatically hot-pressed to (al- information can be found in various pro-
most) 100% relative density, it is referred ceedings (Schneider and Rice, 1972, Hock-
to as HIP finishing. A not very frequently ey and Rice, 1979, Subramanian and Ko-
used variant of hot-pressing is the so-called manduri, 1985).
'continuous hot-pressing' technique (Stui- Grinding can be considered as repeated-
jts, 1973). In this case the lower die is ly grooving with single diamond grains,
lowered continuously, while the upper die bonded in a wheel. With increasing load
is intermittently lifted. While lifted, an during grooving, plastic deformation and
amount of powder is added. In this way, in median cracks appear, while during de-
principle, endless rods of various cross-sec- loading lateral cracks appear. The lateral
tions can be produced. Many materials cracks take care of the material removal,
have been densified by pressure sintering, while the median cracks act as mechanical
e.g., nonoxides include Si 3 N 4 (hot-press- flaws. There is a threshold load at which
ing and HIP) and SiC (hot-pressing and cracking occurs. Below this threshold only
HIP) (see, e.g., Wachtman, 1989, also for a plastic deformation takes place, while
general overview on these and other mate- above it cracking also occurs.
rials), and oxides include ferrites (continu- The control of grinding thus involves
ous hot-pressing and HIP) and titanates many variables dependent on the grinding
[(continuous) hot-pressing, forging, and wheel (wheel size and type of bonding:
HIP] (see, e.g., Hardtl, 1975, and Vol. 15, bronze or polymer, average diamond grain
Chap. 4, Sec. 4.6.2 of this Series). More size, diamond concentration), the grinding
complex, experimental materials, e.g., Al- machine (stiffness), and process parame-
con (A1 28 C 6 O 21 N 6 ) have also been densi- ters (rotation rate, single versus multiple
fied in this way (Groen et al., 1994). More- pass, depth of cut, perpendicular or paral-
over, the densification of integrated elec- lel grinding). The choice of parameters is
tronic components by hot-pressing has re- still often determined empirically, in spite
cently been described (Keizer and de Wild, of the fact that some guidance from theory
1994). can be obtained.
In normal grinding procedures a rela-
tively broad distribution of diamond sizes
is used in the grinding wheel. This results
2.6 After-Sintering Control in a wide range of loads, creating widely
different lateral crack sizes. Since widely
After sintering, the product processing different median crack sizes also result, the
may be finished but often further process- strength in principle degrades. This de-
ing steps, in particular machining, are re- crease in strength is, however, frequently
quired (Table 2-2). We will discuss here a compensated for by the compressive
number of possible checks that can be stresses introduced by the plastic deforma-
done, after a brief discussion on machin- tion. For zirconia ceramics the phase
ing. transformation from monoclinic to tetrag-
 2.6 After-Sintering Control 63

onal also plays an important role in the The erosion process and the resulting re-
balance between strength and grinding sidual stress have a significant influence on
(van den Berg and de With, 1993). More- the strength. Moreover, the results depend
over, since the grinding operation can be critically on the process parameters such as
considered as a closed-loop mechanical degree of filling, ratio of component/slur-
process, the stiffness of the grinding ma- ry, rate of rotation, and time (de With and
chine, which controls the amount of vibra- Sweegers, 1995).
tions, is also of importance. In view of the In a number of cases polishing is also
above-mentioned mechanisms, in a num- required, in particular for electronic appli-
ber of cases extremely stiff machines with cations. Even more so than for grinding,
carefully prepared and dressed grinding polishing (and lapping) is based on experi-
wheels are used (Shore, 1990). This results ence and different people obtain different
in so-called 'damage free' or 'ductile' results with the same material (Clinton,
grinding, which may increase the strength 1987). Very low roughness values can be
of the ground product considerably com- obtained, however, by using the correct
pared with normal grinding, but is, howev- procedures (Fig. 2-16).
er, accompanied by a considerable increase
in cost. An example of this is the three-
point bend strength of sialon, either 2.6.2 Visual and Geometrical Control
ground conventionally or damage free: for A visual check is frequently performed
the former type a strength of 482 MPa was after the sintering and/or grinding where,
reported while for the latter an increase to depending on the product, a full-scale in-
875 MPa was observed (Shore, 1990). spection or a more limited inspection is
Another finishing operation is 'tum- done. Visual checks may be for stains,
bling'. In this process sharp corners of cracks, nonlinearity, and nonbonded parts.
small products are rounded off by erosion Examples of defects found in extruded alu-
in a rotating container with an abrasive mina tubes are shown in Fig. 2-13. Visual
slurry. A smooth surface is also obtained. control also includes the internal examina-

Figure 2-16. Surface topog-

raphy of an extremely well
polished surface of hot-
pressed BaTiO3 with a
grain size of about 0.3 um
and 99.9% relative density,
as determined by STM.
800 The Ra value is 0.83 nm.
64 2 Process Control in the Manufacture of Ceramics

tion of a product, frequently by taking a Although the definitions are simple

micrograph. These micrographs can be enough, the tricky part is in the evaluation.
quite revealing about the processing. Some Here two factors play a role: First, various
examples of micrographs for CMCs are kinds of filters are used, either in the mea-
shown in Fig. 2-15. Defects are usually suring device or during data processing,
classified into a limited number of cate- which supposedly eliminate edge effects
gories according to which it is possible to due to the finite track length registered.
adjust the process or raw material. However, this filtering can have a large
Geometrical control includes verifica- influence on the values of Ra and Rq ob-
tion of the various dimensions, roughness, tained. The Ra and Rq values are generally
nonplanarity, eccentricity, etc. In many reduced by 10-50% when filtering is ap-
cases only a sample is tested and, when plied (de With and Corbijn, 1992). Second-
found to be satisfactory, the lot is ap- ly, the center line is frequently determined
proved. If the sample is not satisfactory, by applying such a filter repeatedly. This
frequently another, larger sample is taken, procedure can also cause a severe bias. A
from which a more reliable judgement can fairly simple but suitable approach seems
be made, before disposing of the whole lot to be to determine the center line by a
of products. The tolerance on as-fired di- least-squares fitting of the entire profile by
mensions can be quite small. For larger a suitable low-order polynomial, and to
dimensions (~100 mm) relative sample avoid further filtering as much as possible.
standard deviations of 0.1% can be ob- It should be noted that the surface mor-
tained, while for smaller dimensions phology is not adequately characterized by
(<10mm) the relative sample standard this type of measurement alone. Quite dif-
deviation generally increases, e.g., up to ferent morphologies can correspond to the
about 1%. same Ra or Rq values.
Dimensions, eccentricity, and nonpla- Although contactless measurements of
narity are easily determined, but rough- the surface roughness have been possible
ness is somewhat more tricky. Roughness for quite a longtime, they have become
measurements are usually done with a pro- much easier through the invention of the
filometer which registers the height of a scanning tunneling microscope (STM) and
component along a certain track by trac- the atomic force microscope (AFM). In
ing it with a stylus. The conventional pa- Fig. 2-16 the topography of a polished sur-
rameter for the roughness is the center line face of hot-pressed BaTiO3 with a grain
average (CLA) or Ra value defined by size of about 0.3 |im, as determined by
Ra = (L\Zi\)/n (2-13) STM, is displayed. This figure illustrates
the results that can be obtained by polish-
where n is the number of points on the ing and its representation by the STM.
center line at which a profile deviation z-x is
measured. The center line is defined as the
line which divides the profile in such a way 2.6.3 Microstructure
that the areas above and below that line For microstructure control the measure-
are equal. Another frequently used param- ments are usually limited to porosity and
eter is the root mean square deviation grain size. Depending on the product,
(RMS) or Rq value, defined by measurements are performed, ranging
f)/n]1'2 (2-14) from a simple manual count to fairly so-
 2.7 Final Remarks 65

phisticated, computer-assisted stereology ture superconductor YBa 2 Cu 3 0 :c . One

(see Vol. 2b, Chap. 15 of this Series). The simple way to assess the presence of micro-
grain size can be determined from the as- cracks is by BET and MIP analysis of the
fired surface, but in some materials the sintered material. For a 93% dense
surface has a different grain size from the YB 2 Cu 3 0 ;c ceramic (Severin and de With,
bulk of the material. A cross-section is 1988), processed in the conventional way,
therefore frequently prepared and the ma- these measurements indicate a specific sur-
terial is etched subsequently. Chemical face of 0.8 m 2 /g (MIP) to 1.5 m2/g (BET),
etching should be considered an art be- mainly divided over 'pores' with a size of
cause only a few general rules apply. Since 3-4 nm. Combined with a grain size of
thermal etching often provides a better 10-24 jim, this leads to the conclusion that
etch, many people use this method instead the material is heavily microcracked, as
(Clinton, 1987). The precise time, tempera- was also expected from mechanical consid-
ture, and atmosphere are quite important, erations.
but this is often not recognized. A recent
round robin on manual grain size determi- 2.6.4 Functional Properties
nation showed the effect of different an-
nealing cycles on the grain size found The specific properties determine to a
(Dortmans e t a l , 1992). The effects of large extent what functional tests are car-
sampling, interpretation, and micrograph ried out. Sometimes full testing is per-
preparation were also evaluated. Counting formed on a number of parameters, com-
at least a hundred grains, the scatter due to bined with a more limited test on other
sampling is about 10%, due to interpreta- parameters. For example, for CMCs the
tion another 10%, while preparation adds capacity, loss tangent, minimum break-
another 15%. From this round robin it through voltage, and electrical resistance
can be concluded that an accuracy of are usually tested on a complete scale. The
about 35% can be obtained with limited laydown of electrodes is checked by taking
effort. micrographs of a relatively large number
of products, while the Curie temperature is
A typical example of manual grain size
checked on only a few products. In other
control measurement can be found in the
cases, more limited testing is done, e.g., for
statistics of grain sizes for a particular
translucent alumina the transmittance is
batch of alumina: Taking five measure-
checked using a small sample from pro-
ments of the intercept, /, by counting the
duction. Whether full-scale or more limit-
number of grains intersecting a circle of
ed testing is done depends on both the pro-
specified diameter, resulted in an average
cess stability and customer requirements.
intercept length /=43.0 \im and a sample
standard deviation 5 = 4.5 j^m. Only the
USL was specified, in this case at 100 }im.
For this particular example the grain size is
2.7 Final Remarks
thus well within the specifications. Apart from taking all types of measure-
In some cases other microstructure ments, it is highly important, and in fact
characteristics are also determined. For prescribed by the ISO 9001 standard, to
highly anisotropic materials, microcrack- keep a record for the final product from
ing may occur. Typical examples are which the raw materials used, equipment
PbTiO 3 , Al 2 TiO 5 , and the high-tempera- used, data obtained, etc. can be traced. Al-
66 2 Process Control in the Manufacture of Ceramics

though it seems trivial, this bookkeeping Chou, Y.T., Ko, Y.T., Yan, M.F. (1987), J. Am. Cer-
am. Soc. 70, C280-C282.
was frequently not done in the past for Clinton, D. J. (1987), A Guide to Polishing and Etching
economy reasons. Ever-increasing quality of Technical and Engineering Ceramics. Stoke-on-
demands have changed this attitude. Trent, UK: The Institute of Ceramics.
de With, G. (1987), Philips J. Res. 42, 119-130.
Rather trivial but quite important, nev- de With, G., Corbijn, A. J. (1992), unpublished.
ertheless, is the packaging of the final de With, G., Sweegers, N. (1995), Wear, in press.
products. Apart from safe packaging, one de With, G., Witbreuk, P.N.M. (1993), J. Eur. Cer-
am. Soc. 12, 343-351.
of the most important aspects is the label- de With, G., Vrugt, P. J., van de Ven, A. J.C. (1985),
ing of the packed products. Considerable /. Mater. Sci. 20, 1215-1221.
costs can arise in the case of a delivery of Dorre, E., Hiibner, H. (1984), Alumina: Processing,
Properties and Applications, Berlin: Springer.
the wrong material. These costs not only Dortmans, L.J.M.G., Morrell, R., de With, G.
include the cost of replacement but also (1993), /. Eur. Ceram. Soc, 12, 205-213.
that of the delay in the follow-up process. Exner, H.E. (1983),in: Physical Metallurgy, 3rd ed.:
Cahn, R. W., Haasen, P. (Eds.) Amsterdam: North-
From the previous sections it is clear Holland, pp. 582-646.
that many aspects are involved in the con- Farris, R.J. (1968), Trans. Rheol Soc. 12, 281-301.
trol of ceramic processing. Measurements German, R. M. (1990), Powder Injection Molding.
Princeton, NJ: Metal Powder Industries Federa-
from such varying disciplines as rheology, tion.
powder mechanics, physical chemistry, Gray, W.A. (1968), The Packing of Solid Particles.
chemical analysis, etc. are required. A London: Chapman and Hall.
number of aspects can be quantified exact- Gregg, S.J., Sing, K.S.W. (1982), Adsorption, Surface
Area and Porosity, 2nd ed. London: Academic.
ly but the rest remain essentially qualita- Groen, W.A., Kraan, M.J., van Hal, P.F., Sweegers,
tive. Those responsible for a particular ce- N., de With, G. (1994), /. Mater. Sci., unpublished.
ramic process should be well aware of both Hardtl, K.H. (1975), Bull. Am. Ceram. Soc. 54, 210-
207.
varieties. Heinicke, G. (1984), Tribochemistry. Munich: C.
Hanser Verlag.
Hench, L.L., Ulrich, D.R. (Eds.) (1984), Ultrastruc-
ture Processing of Ceramics, Glasses and Com-
posites. New York: Wiley.
2.8 References Hench, L.L., Ulrich, D.R. (Eds.) (1986), Science of
Ceramic Chemical Processing, New York: Wiley.
Austin, L.G., Klimpel, R.R., Luckie, P.T. (1984), Hennings, D.F.K., Janssen, R., Reynen, P.J. L.
Process Engineering of Size Reduction: Ball (1987), J. Am. Ceram. Soc. 70, 23-27.
Milling. New York: AIME. Hennings, D.F.K., Hartung, R., Reynen, P.J.L.
Boch, P., Rogeaus, B. (1986), Brit. Ceram. Proc. 38, (1990), J. Am. Ceram. Soc. 73, 645-648.
91-101. Himmelblau, D.M. (1970), Process Analysis by
Brinker, C.J., Clark, D.E., Ulrich, D.R. (Eds.) Statistical Methods. New York: Wiley.
(1984), Better Ceramics through Chemistry, New Hockey, B.J., Rice, R.W. (1979), The Science of Ce-
York: North-Holland. ramic Machining and Surface Finishing II, NBS-SP
Brinker, C.J., Clark, D.E., Ulrich, D.R. (Eds.) 562, Washington, DC: Natl. Bureau Standards.
(1986), Better Ceramics through Chemistry II, Pitts- Hunter, R.J. (1987), Foundations of Colloid Science,
burgh, PA: Mater. Res. Soc. Vol. I, Oxford: Clarendon.
Brinker, C.J., Clark, D.E., Ulrich, D.R. (Eds.) James, P.J. (Ed.) (1983), Isostatic Pressing Technolo-
(1988), Better Ceramics through Chemistry III, gy, London: Appl. Sci.
Pittsburgh, PA: Mater. Res. Soc. Juran, J.M., Gryna, KM., Bingham, R.S. (Eds.)
Broese van Groenou, A. B. (1982), Ceramic Mono- (1974), Quality Control Handbook. New York: Mc-
graphs 1.4.5.1.1 - Handbook of Ceramics, Supple- Graw-Hill.
ment to Interceram 31/32, Freiburg i.Br.: Verlag Katz, R.N. (1980), in: Energy and Ceramics: Vincen-
Schmidt, pp. 1-10. zini, P. (Ed.). Amsterdam: Elsevier, pp. 449-467.
Broese van Groenou, A.B., Knaapen, A. C. (1980), Keizer, P. H. M., de Wild, W R. (1994), in: Electroce-
Sci. Ceram. 10, 93-99. ramics IV: Waser, R., Hoffmann, S., Bonnenberg,
Brown, R.L., Richards, J.C. (1970), Principles of D., Hoffmann, C. (Eds.). Aachen, Germany: Ver-
Powder Mechanics. Oxford: Pergamon. lag der Augustinus Buchhandlung; pp. 1045-1053.
 2.8 References 67

Kendall, K. (1978), Nature 272. 710-711. Stuijts, A. L. (1973), in: Annual Review Materials Sci-
Kumagai, M., Messing, G.L. (1985), J. Am. Ceram. ence Vol. 3: Huggins, R. A., Bube, R. H., Roberts,
Soc. 68, 500-505. R. W. (Eds.). Palo Alto, CA: Annu. Rev. Inc.,
Lecloux, A. J. (1981), in: Catalysis, Science and Tech- pp. 363-395.
nology, Vol. 2: Anderson, J. R., Boudart, M. (Eds.). Stuijts, A. L. (1977), in: Ceramic Microstructures '76:
Berlin: Springer, pp. 171-230. Fulrath, R.M., Pask, J.A. (Eds.). Boulder, CO:
Magdic, T.J. (1984), in: Forming of Ceramics, Adv. Westview Press, pp. 1-26.
Ceram. Vol.9: Mangels, J.A., Messing, G.L. Subramanian, K., Komanduri, R. (Eds.) (1985), Ma-
(Eds.). Columbus, OH: Am. Ceram. Soc, pp. 1-3. chining of Ceramics Materials and Components.
Mangels, J.A., Trela, W. (1984), in: Forming of Ce- New York: Am. Soc. Mech. Eng.
ramics, Adv. Ceram. Vol.9: Mangels, J.A., Mess- Takeda, Y. (1988), Bull Am. Ceram. Soc. 67, 1961-
ing, G. L. (Eds.). Columbus OH: Am. Ceram. Soc., 1963.
pp. 220-233. Tiller, F.M., Tsai, C.-D. (1986), J. Am. Ceram. Soc.
McColm, 1.1, Clark, N.J. (1988), Forming, Shaping 69, 882-887.
and Working of High Performance Ceramics. Glas- Toolenaar, F.J.C.M. van Lierop-Verhees, M.T.I
gow: Blackie and Sons. (1988), J. Mater. Sci. 23, 856-861.
McKenzie, J.D., Ulrich, D.R. (1988), Ultrastructure Uhlmann, D.R., Ulrich, D.R. (Eds.) (1992), Ultra-
Processing of Advanced Ceramics. New York: Wi- structure Processing of Advanced Materials. New
ley. York: Wiley.
Messing, G. L., Kumagai, M. (1988), J. Am. Ceram. van de Graaf, M.A.C.G., ter Maat, J.H.H.,
Soc. 72, 40-44. Burggraaf, A.I (1983), in: Ceramic Powders: Vin-
Morris, K. J. (1983), in: Isostatic Pressing Technology: cenzini, P. (Ed.). Amsterdam: Elsevier, pp. 783-
James, P. J. (Ed.). London: Appl. Sci., pp. 91-123. 794.
Onoda, G.Y., Hench, L.L. (Eds.) (1978), Ceramic van den Berg, P.H.I, de With, G. (1993), in: Science
Processing before Firing. New York: Wiley. and Technology of Zirconia V: Badwal, S.P. S.,
Page, I P . , Metzbower, E.A., Shanefield, D.J., Has- Bannister, M.I, Hannink, R.H.I (Eds.). Lancast-
selman, D. P. H. (1978), in: Ceramic Processing be- er, PA: Technomic, pp. 339-346.
fore Firing. Onoda, G. Y, Hench, L. L. (Eds.). New Wachtman, I B . (Ed.) (1989), Structural Ceramics,
York: Wiley, pp. 141-151. Treatise on Materials Science and Technology
Palmour III, H., Hare, T. M. (1987), in: Sintering '85. Vol. 29. Boston, MA: Academic Press.
Kuczynski, G.C., Uskokovic, D.P., Palmour, III, Wada, S. (1980), in: Energy and Ceramics: Vincenzini,
H., Ristic, M.M. (Eds.). New York: Plenum, P. (Ed.). Amsterdam: Elsevier, pp. 163-174.
pp. 17-34. Wang, F.F. Y (Ed.) (1976), Ceramic Fabrication Pro-
Prasker, C.L. (1987), Crushing and Grinding Process cesses, Treatise on Materials Science and Technol-
Handbook. Chichester, UK: Wiley. ogy Vol. 11. New York: Academic Press.
Reed, I S . (1988), Introduction to the Principles of Zellinski, B.I I , Brinker, C.I, Clark, D.E., Ulrich,
Ceramic Processing. New York: Wiley. D. R. (Eds.) (1990), Better Ceramics through Chem-
Reynen, P., Bastius, H., Fiedler, M. (1983), in: Ce- istry IV. Pittsburgh, PA: Mater. Res. Soc.
ramic Powders: Vincenzini, P. (Ed.). Amsterdam:
Elsevier, pp. 499-504.
Rice, E.R., Tengzelius, I (1986), Powder Metall. 29,
183-194.
Schneider, S.I, Rice, R.W (1972), The Science of General Reading
Ceramic Machining and Surface Finishing, NBS-SP
348. Washington, DC: Natl. Bureau Standards. McColm, I.I (1983), Ceramic Science for Material
Schwedes, I (1984), Powder Technology. Washington, Technologists. Glasgow: Leonard Hill.
DC: Hemisphere, pp. 89-98. McColm, L I , Clark, N.J. (1988), Forming Shaping
Severin, I, de With, G. (1988), Proc. Brit. Ceram. and Working of High Performance Ceramics. Glas-
Soc. 40, 249-256. gow: Blackie and Sons.
Shore, P. (1990), in: Brit. Ceram. Soc. Proc. 46 'Ad- Onoda, G.Y, Hench, L.L. (Eds.) (1978), Ceramic
vanced Engineering with Ceramics': Morrell, R. Processing before Firing. New York: Wiley.
(Ed.), Stoke-on-Trent, UK: The Institute of Ce- Reed, I S . (1988), Introduction to the Principles of
ramics, pp. 189-200. Ceramic Processing, New York: Wiley.
Somasundaran, P. (1978), in: Ceramic Processing be- Wang, F.F.Y (Ed.) (1976), Ceramic Fabrication Pro-
fore Firing: Onoda, G. Y, Hench, L. L. (Eds.). New cesses, Treatise on Materials Science and Technolo-
York: Wiley, pp. 105-123. gy Vol. 11. New York: Academic Press.
3 Chemical Preparation of Powders
David Segal

Materials Chemistry Department, AEA Technology, Harwell Laboratory,

Oxfordshire, U.K.

List of Symbols and Abbreviations 70

3.1 Introduction 71
3.2 Conventional Preparation of Powders 71
3.3 Requirements for Improved Powder Properties 72
3.4 Homogeneous Nucleation 72
3.4.1 Particle Growth in Solution 74
3.5 The Colloidal State 75
3.6 Coprecipitation 75
3.7 Sol-Gel Processing of Colloids 77
3.7.1 Hydrolysis of Cations 77
3.7.2 Sol-Gel in the Nuclear Industry 78
3.7.3 Ceramic Powders from Colloids 79
3.8 Sol-Gel Processing of Metalorganic Compounds 80
3.8.1 Preparation and Properties of Metal Alkoxides 80
3.8.2 Powders from Alkoxides 80
3.9 Alternative Gel Routes to Powders 82
3.10 Hydrothermal Synthesis 83
3.11 Forced Hydrolysis of Salt Solutions 85
3.12 Non-Aqueous Liquid Phase Reactions 86
3.13 Gas-Phase Reactions 88
3.14 Aerosol-Derived Powders 91
3.15 Polymer Pyrolysis 93
3.16 Emulsion Routes to Powders 94
3.17 Freeze- and Spray-Drying 94
3.18 Summary 95
3.19 Acknowledgements 95
3.20 References 95

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. Allrightsreserved.
70 3 Chemical Preparation of Powders

List of Symbols and Abbreviations

d droplet size
f ultrasonic frequency
AG free energy change
AG* free energy of activation
AGC critical free energy
AGS surface free energy per unit area
AGV bulk free energy per unit volume
h Planck's constant
I nucleation rate
k Boltzmann constant
n number of moles of bound water molecules
"i absolute hardness of an atom
N number of atoms per unit volume
P number of protons; vapour pressure of the droplet
Po saturation vapour pressure
P/Po supersaturation ratio
r radius
rc critical radius
T absolute temperature
vm molar volume
Z valency

y surface tension
8: partial charge
Q density
Xi electronegativity
Xi mean electronegativity
Xi electronegativity of the neutral atom

d.c. direct current

DLVO Deryagin, Landau, Verwey, Overbeek (theory)
RESA reactive electrode submerged arc
 3.2 Conventional Preparation of Powders 71

3.1 Introduction happens with electroceramic powders. In

addition, undesirable phases can form, for
Chemical routes for the preparation of example, BaTi 2 O 5 during synthesis of Ba-
powders on the laboratory and industrial TiO 3 .
scale are attracting increasing worldwide For non-oxide powders, conventional
attention in the early 1990s. Application syntheses include direct reaction of a metal
of chemistry to powder synthesis occupies with a gas. For example, titanium nitride is
an interface between conventional pursuits produced by nitridation of Ti metal in N 2
in chemistry and materials science. In this at 1773 K; a transmission electron mi-
chapter conventional methods for powder crograph for commercial TiN powder
synthesis are outlined followed by the ma- made by nitridation of metal is shown in
terials requirements for improved powder Fig. 3-1 (Blackburn etal., 1991). Carbo-
properties. Major chemical routes for syn- thermal reduction is another conventional
thesis of ceramic powders are then de- preparation of non-oxides. Thus titanium
scribed. While these routes differ from diboride is produced by carbothermal re-
each other in terms of their practical pro- duction of a mixture of TiO 2 and B 2 O 3 at
cedures they are often related through two 1273 K or by reduction of TiO 2 with bo-
areas of chemistry, namely colloid science ron carbide and carbon at 2273 K (Gal-
and homogeneous nucleation. Basic con- lagher etal., 1988) while in the Acheson
cepts on colloidal systems and homoge- process, industrial SiC powder is made by
neous nucleation are also described and carbothermal reduction of silica at temper-
related to the different chemical routes for atures in excess of 2073 K (Segal, 1989).
powder synthesis. The main disadvantage of these conven-

3.2 Conventional Preparation

of Powders

The main conventional synthesis for

multicomponent ceramic powders is a
solid-state reaction between oxide and/or
carbonate powder precursors. Thus for
barium titanate, BaCO 3 and TiO 2 pow-
ders are mixed, milled and calcined; re-
peated cycles of milling and calcination are
carried out to achieve the solid-state reac-
tion. High temperatures are required in
these solid-state reactions, thus for
BaFe 12 O 19 , between 1423 and 1523 K for
mixtures of a-Fe 2 O 3 and BaCO 3 (Barb
et al., 1986). Disadvantages of this method
are large grain sizes due to the high firing
Figure 3-1. Transmission electron micrograph of com-
temperatures and poor chemical homoge- mercial titanium nitride powder (Blackburn et al.,
neity especially when dopant oxides are in- 1991, courtesy of Tioxide Specialties Limited and the
troduced in small quantities as frequently Institute of Ceramics).
72 3 Chemical Preparation of Powders

tional syntheses for non-oxides is the re- 3.3 Requirements for Improved
quirement for extensive grinding of the re- Powder Properties
fractory materials for particle size reduc-
tion which introduces chemical impurities Advanced ceramic materials are used in
into the powders. the form of powders, coatings, fibres and
Precipitation from solution is a conven- monoliths. Conventional synthesis of ce-
tional preparation for one-component ox- ramics yield powders and these are not
ide powders. Thus in the Bayer process for particularly suited for fabrication of coat-
manufacture of a-Al 2 O 3 bauxite is hy- ings and fibres. Thus, one reason why
drothermally dissolved in sodium hydrox- chemical routes are attracting much atten-
ide to form sodium aluminate solution. An tion is that some of them allow direct fab-
aggregated gibbsite powder is produced by rication of coatings and fibres without
seeding this solution with gibbsite crystals powder intermediates. In addition chemi-
or by neutralisation with CO 2 gas which cal routes have the potential for achieving
results in precipitation of bayerite. The hy- improved chemical homogeneity on the
drous alumina is converted to a-Al 2 O 3 at molecular scale which is particularly im-
around 1873 K although addition of fluo- portant for electroceramic materials whose
rine compounds as mineralisers lowers the properties are often determined by small
conversion temperature to a-Al 2 O 3 and quantities of dopants. For structural ce-
produces platy crystals as illustrated in ramics improved mechanical properties
Fig. 3-2 (Southern, 1991). When applied to such as strength can be achieved by re-
multicomponent oxides (see Sec. 3.6) care- moval of aggregates in powder feedstocks.
ful control of solution conditions is re- Chemical routes are attractive methods for
quired in order to precipitate all cations synthesis of structural ceramics because
and thus maintain chemical homogeneity they allow production of submicrometre
on a molecular scale. powders in an unaggregated form. These
routes also use lower reaction tempera-
tures for producing the required crystalline
phases as components are mixed on the
colloidal or molecular scale so that diffu-
sion distances are smaller compared with
conventional preparations. Major chemi-
cal routes which are under intensive world-
wide investigation for powder preparation
are described below.

3.4 Homogeneous Nucleation

Random thermal fluctuations in a su-
persaturated vapour consisting of atoms
Figure 3-2. Scanning electron micrograph of fluo-
ride-mineralised Bayer alumina (Southern, 1991, or molecules give rise to local fluctuations
courtesy of Alcan Chemicals Limited and the Insti- in density and changes in the free energy of
tute of Ceramics). the system. Density fluctuations produce
 3.4 Homogeneous Nucleation 73

clusters of atoms or molecules known as where k is the Boltzmann constant and T

embryos which can grow by incorporation is the absolute temperature. The value for
of atoms or molecules from the vapour AGC can also be expressed by the equation
phase. A range of cluster sizes will be pres- (Christian, 1975)
ent in the vapour with vapour pressures
AGC = 16TT Vmy3/3k2T2ln(p/Po) (3-4)
determined by the Kelvin equation (Defay
etal., 1966); a consequence of the Kelvin where vm is the molar (or atomic) volume,
equation is that small clusters will evapo- y is the surface tension of the new phase,
rate back to the gas phase at the expense of and/?//?0 is the supersaturation ratio where
large ones. An analogy may be made with p0 is the saturation vapour pressure and p
the solubility of crystals whereby small is the vapour pressure of the droplet. As an
crystals are more soluble than larger ones. example, for a supersaturation ratio of 4
Embryos with a radius r, less than a critical there is about one critical nucleus per
value, rc, cannot grow to a new phase 1000 m 3 of air (Defay et al., 1966) whereas
whereas clusters with r>rc can. For nuclei at a ratio of 5 there are 6700 per m 3 . These
to form the embryos have to overcome an values indicate that nucleation of water va-
energy barrier and this may be illustrated pour occurs at a supersaturation ratio be-
by consideration of the free energy change, tween 4 and 5 as is observed in practice.
AG, on formation of a spherical embryo, For homogeneous nucleation in con-
radius r. It is assumed that the system is in densed systems such as viscous liquids the
thermodynamic equilibrium with all ener- activation energy for diffusion of molecules
gy states up to the energy state correspond- across a phase boundary can be a barrier
ing to the height of the free energy barrier. to nucleation (McMillan, 1964). Turnball
The magnitude of AG is given by the ex- and Fisher (1949) first obtained an expres-
pression (Stookey and Maurer, 1962) sion for the nucleation rate in condensed
systems by application of absolute reac-
AG = 4 7i r2 AG s -4/3 TT r3 AGV (3-1) tion rate theory. Thus,
where AGS is the surface free energy per exp (3-5)
unit area and AGV is the bulk free energy
per unit volume. The critical radius has a where N is the number of atoms per unit
value corresponding to d(AG)/dr = 0 and volume, h is Planck's constant, AG* is the
its critical free energy, AGC, that is the max- free energy of activation for diffusion of
imum value of free energy, can be written atoms from liquid to the new phase assum-
in the form ing only short range diffusion occurred,
and other terms have been defined previ-
= 16TTAG S 3 /3AG V (3-2) ously. Dunning (1961) modified the statis-
tical mechanical approach to nucleation
The nucleation rate, /, is the rate of for- from vapours to determine an expression
mation of critical nuclei as only these can for the rate of nucleation in a supersaturat-
grow to the new phase. Calculation of / ed solution.
has attracted much attention (Christian,
Homogeneous nucleation occurs
1975; McMillan, 1964) using the kinetic
throughout different chemical techniques
theory of gases and statistical mechanics.
for ceramic synthesis. Thus in gas phase
This rate can be written in the form
reactions (see Sec. 3.13) clusters are
exp [-AGJkT\ (3-3) formed from atoms or molecules and grow
74 3 Chemical Preparation of Powders

to the new phase. For formation of precip- particles. In polynuclear growth the parti-
itates (see Sec. 3.6) and their peptisation to cle size distribution decreases with growth
sols (see Sec. 3.7.3) polynuclear ions (see but to a smaller extent than for diffusion
Sec. 3.7) are the precursors for nucleation. controlled growth. Variation of the super-
In hydro thermal synthesis (see Sec. 3.10) saturation ratio (i.e., the ratio of solute
polynuclear ions are precursors for nucle- concentration to the concentration of a
ation as in powder preparation by forced saturated solution) with particle radius
hydrolysis (see Sec. 3.11). However in sol- (Nielsen, 1967) indicated boundary lines
gel processing of metal alkoxides (see separating different growth regions.
Sec. 3.8.1) polymers formed by alkoxide Polynuclear growth was characterised by
hydrolysis are the precursors for nucle- occupying a region bounded by supersatu-
ation from solution. ration ratios between 10 and 100 and parti-
cle radii in the range 0.3 nm to 1 jim. For
the mononuclear region, the particle
3.4.1 Particle Growth in Solution
growth rate is proportional to the surface
A nucleus that has formed from a super- area and surface reaction (i.e., desolvation,
saturated liquid phase can grow by trans- crystallisation, alignment) is considerably
port of solute species to the particle sur- slower than the two-dimensional growth
face, desolvation and alignment on the of surface nuclei. This results in particles
particle surface. Nielsen (1964, 1967) ap- growing by completion of layers before nu-
plied Fick's laws of diffusion to growing cleation of additional layers. The width of
nuclei and identified five possible crystal the particle size distribution increases with
growth mechanisms, two of which were growth so that this mechanism does not
formation of surface spiral steps leading to yield monodispersed particles. The signifi-
screw dislocations and enhanced diffusion- cance of these growth processes is that pre-
al rate control by convection around parti- cipitation of hydrous oxides is often a
cles. The other three mechanisms are rele- polynuclear growth process (Livage et al.,
vant to growth of ceramic particles in solu- 1990).
tion and are diffusion controlled growth, Generally, mononuclear growth is im-
polynuclear and mononuclear growth pro- portant at low values of particle radius and
cesses. In diffusion controlled growth, dif- concentration, and diffusion controlled
fusion of solute to the nucleus is the rate growth at high values. High supersatura-
determining step for particle growth which tion ratios during growth encourage diffu-
is inversely proportional to the particle ra- sion controlled or polynuclear layer mech-
dius. This inverse relationship leads to a anisms although, if the ratios are low, Ost-
narrowing of the particle size distribution wald ripening may occur in which smaller
with growth (Haruta and Delmon, 1986). particles dissolve and larger ones grow. If
Polynuclear and mononuclear growth supersaturation is kept too high during
mechanisms have rate controlling steps particle growth, further bursts of nucle-
that involve reactions at the particle sur- ation can occur.
face (Sugimoto, 1987). Surface nucleation LaMer and Dinegar (1950) were the first
is fast in polynuclear growth and nuclei to postulate how monodispersed powders
aggregate onto each other without forming were produced by homogeneous nucle-
complete monolayers; the growth rate is ation following a study of growth of sul-
independent of surface area of existing phur particles in acidified sodium thiosul-
 3.6 Coprecipitation 75

phate solutions. The concentration of Table 3-1. Classification of colloids.

molecular S produced by decomposition of System Disperse Dispersion
thiosulphate anions increased above a sat- Phase Medium
uration value until a critical supersatura-
tion concentration was reached. Homoge- Dispersion (sol) solid liquid
neous self-nucleation then occurred fol- Emulsion liquid liquid
lowed by particle growth involving diffu- Foam gas liquid
sion of molecular sulphur to the nuclei un- Fog, mist or aerosol liquid gas
(of liquid particles)
til the concentration of molecular S re-
Smoke or aerosol solid gas
turned to the saturation value. This idea of (of solid particles)
nucleation subsequent to growth-follow- Alloy, solid suspension solid solid
ing diffusion has been applied to alkoxide
hydrolysis (see Sec. 3.8.2) although it is im-
portant to note that particle growth can clear from Table 3-1 that colloids have an
result from homogeneous nucleation in so- important role in the preparation of ce-
lution, for example in alkoxides, succeeded ramic powders by chemical routes. Thus
by aggregation of small nuclei (Bogush sol-gel processing (see Sees. 3.7 and 3.8)
etal., 1988). Nucleation and growth can involves dispersions, gas-phase reactions
occur simultaneously (Haruta and Del- (see Sec. 3.13) utilise smokes while aero-
mon, 1986) but minimisation of the initial sols (see Sec. 3.14) are colloidal; emulsion
supersaturation reduces the period during routes to powders (see Sec. 3.16) also in-
which both nucleation and growth take volve colloids and hydrothermal tech-
place. When applied to powder synthesis niques (see Sec. 3.10) produce colloidal
supersaturation ratios are controlled by a dispersions.
control of solute release in order to limit
The stability of colloidal systems such as
local high supersaturation ratios which
dispersions is often explained in terms of
produce uncontrolled precipitation of
the DLVO theory named after the authors
solids. This is normally achieved by gener-
(Deryagin, Landau, Verwey and Over-
ating a "reservoir" of solute with a con-
beek) in which the potential energy be-
trolled release, and is particularly relevant
tween two particles is expressed as a sum of
to preparation of particles by forced hy-
two components. The first component is
drolysis for example for synthesis of sul-
an electrostatic double layer repulsion and
phides (see Sec. 3.11).
the second, the van der Waals attractive
energy between particles (Verwey and
Overbeek, 1948). A detailed account of
3.5 The Colloidal State colloid stability has been given by Lyklema
(1991) and by Goodwin (1982).
Colloids are systems that have, by defi-
nition, a disperse phase with at least one
dimension between 1 nm and 1 jum in a 3.6 Coprecipitation
dispersion medium; examples of colloidal
systems are shown in Table 3-1. The col- Preparation of a single component ce-
loidal dimension refers to particle diameter ramic such as alumina in the Bayer process
for sols but for macroscopic systems such was described in Sec. 3.2. The aim in co-
as foams it refers to film thickness. It is precipitation is to prepare multicompo-
76 3 Chemical Preparation of Powders

nent ceramic oxides through formation of of 0.99 + 0.01. The precipitating agent
intermediate precipitates, usually hydrous (COOH) 2 , can introduce carbon residues
oxides or oxalates so that an intimate mix- into the calcined product. Sherif (1989)
ture of components is formed during pre- showed that for synthesis of yttrium-based
cipitation, and chemical homogeneity is superconductors by the oxalate method,
maintained on calcination. Coprecipita- precipitation from mixed (Cu, Ba) acetate
tion is used to manufacture oxide powders. solutions containing yttrium chloride at
Thus Tosoh Corporation (Japan) produces pH 12 rather than under acidic conditions
unaggregated yttria-stabilised zirconia produced an intimate mixture of barium
powder with a particle size 0.3 jim by co- oxalate, yttrium hydroxide and copper ox-
precipitation of hydroxides from mixed yt- ide which limited the amount of oxalic acid
trium and zirconyl chloride solutions after required in the process, and thus carbon
which the metal hydroxide precipitates are residues.
solvent-dried, calcined and milled. The Synthesis of doped ZnO powder with a
Pechiney Group (France) manufactures composition 96.5 mol% ZnO, 0.5 mol%
BaTiO3 powder using an oxalate precipita- Bi 2 O 3 , 1.0mol% Sb 2 O 3 , 0.5mol% MnO,
tion process (Bind et al., 1987). Hence a 1.0mol% CoO, 0.5mol% Cr 2 O 3 for va-
barium titanyl oxalate precipitate is pro- ristor applications has been achieved
duced on addition of oxalic acid (COOH) 2 (Haile et al., 1989) by coprecipitation. A
to a mixed barium and titanyl chloride so- basic zinc salt was initially produced on
lution under controlled conditions of pH, addition of NH 4 OH to zinc sulfate solu-
temperature and reactant concentration. tion. Because this basic salt will not de-
compose to pure oxide it was dissolved in
BaCl2 + TiOCl2 + 2(COOH) 2 N H 4 0 H and the solution pH increased by
-> BaTiO(C 2 O 4 ) 2 • 4H 2 O + 4HC1 (1) evaporation of NH 3 (g). Homogeneous
Dopants such as lanthanides are intro- nucleation of zinc hydroxide occurred
duced by coprecipitation and the precipi- based on the nucleation and growth pro-
tate is calcined after collection by filtra- cess described in Sec. 3.4.1. Dopants were
tion, washing and drying. Decomposition coated onto Zn(OH) 2 particles by copre-
occurs through the following steps. cipitation from mixed (Bi, Sb, Mn, Co, Cr)
chloride solutions containing ammonium
BaTiO(C 2 O 4 ) 2 -4H 2 O (2) carbonate. Monodispersed, unaggregated
373-413K powders were produced in this coprecipita-
BaTiO(C 2 O 4 ) 2 + 4H 2 O tion process after calcination. Not all mix-
573-623K tures of cations can be easily precipitated
BaTiO(C 2 O 4 ) 2 > 0.5 BaTi,O
2^5- as hydrous oxides. However, chelating
+ 0.5BaCO agents have been used as precipitating
1.5CO2 (3)
agents (Kakegawa et al., 1984) during the
0.5BaTi 2 O 5 + (4) synthesis of Pb(Zr 0 3 Ti 0 7 )O 3 , as it is diffi-
873-973K cult to precipitate lead hydroxide. Thus Ti
+ 0.5BaCO. and Zr cations were precipitated as a cup-
ferron (C 6 H 5 N(NO)ONH 4 ) complex
The BaTiO3 powder has a nitrogen surface which, after drying, was fired at 1173 K to
area between 1 and 4 m 2 g" 1 , a particle a mixed oxide powder. The latter was
diameter 0.4-1 jim and a Ba/Ti mole ratio mixed and fired with PbO powder at
 3.7 Sol-Gel Processing of Colloids 11

1373 K to yield a homogeneous solid solu- where n is the number of moles of bound
tion of the lead zirconate titanate. Rajen- water molecules, p is the number of pro-
dran (1992) has extended the coprecipita- tons removed from the cation on hydroly-
tion technique to mullite, mullite-zirconia, sis and z is the valency of the cation. Fur-
alumina-zirconia and yttria-stabilised ther hydrolytic reactions can occur as rep-
ZrO2-alumina powders. resented by the equilibria

[M-OH2f [M-OH] ( *~ 1 ) + +H + T±
3.7 Sol-Gel Processing of Colloids +
(6)
Sol-gel processing of colloids is one of Thus three ligands result from hydrolysis,
two sol-gel techniques, the other one in- aquo species (H2O), hydroxyl, species
volving hydrolysis of metalorganic com- (OH) and oxo ligands (O). For example,
pounds (see Sec. 3.8). Since 1980 there has experimental studies (Baes and Mesmer,
been a vast increase in publications on 1976; Livage et al., 1988) indicate that
these two techniques and about 2000 scien- Cr(VI) precursors form two oxo-hydroxyl
tific articles a year are currently published species, CrO 2 (OH) 2 and [CrO 3 (OH)]" as
on sol-gel. Pouskouleli and Wheat (1990) well as an oxo species, [CrO 4 ] 2 ~, but no
showed from analysis of the bibliographic aquo complexes. However Cr(III) species
databases of the American Chemical Soci- form aquo complexes, [Cr(H 2 O] 6 ] 3 + ,
ety that about 9 % of all publications on three aquo-hydroxyl species, [Cr(O-
ceramics in 1988 were on sol-gel whereas H)(H 2 O) 5 ] 2 + , [Cr(OH) 2 (H 2 O) 4 ] + and
only 2% of publications in 1977 were on Cr(OH) 3 (H 2 O) 3 and one hydroxyl com-
this technique. However, it should be plex, [Cr(OH) 4 ] 2 ~ but no oxo complexes.
stressed that the usefulness of techniques Consideration of hydrolytic species
such as sol-gel for ceramic synthesis should formed in solution shows (Kepert, 1972)
not be judged by the number of publica- that ions with a valency less than 4 pro-
tions because factors such as availability of duce aquo-hydroxyl and/or hydroxyl spe-
precursors and the complexity of appara- cies over the whole pH range. Higher va-
tus can influence the interest in a synthetic lency species (>5) form oxo-hydroxyl
technique. and/or oxo complexes over the pH range
1-14. Tetravalent species have solution
3.7.1 Hydrolysis of Cations properties intermediate between the above
Metal ions, Mz + , are hydrated in solu- two regimes and form a series of solution
tion due to a high electronic charge or high complexes. Monomeric hydrolysis prod-
charge density. Hydration involves elec- ucts can condense to polyvalent metal or
tron transfer between coordinated H 2 O polynuclear ions which can be colloidal,
molecules and the central cation with a for example, [A1O4A112(OH)25(H2O)11]6 +
resultant weakening of the OH bond in the (Segal, 1989); polynuclear ions contain
water molecules. The result of cation hy- OH bridges, M-OH-M (olation) or oxygen
dration is that hydrolytic reactions can oc- bridges, M-O-M (oxolation).
cur in the following way, A quantitative approach for predicting
the products of cation hydrolysis known as
[M(H 2 O) n ] z (5) the partial charge model has recently been
[M(OH)p(H2O)n_p]< -*> +/>H +
2 +
developed by Livage and coworkers (Li-
78 3 Chemical Preparation of Powders

vage etal., 1990) and it may be sum- 3.7.2 Sol-Gel in the Nuclear Industry
marised as follows. When atoms combine, Sol-gel processing of colloids involves
the degree of electron transfer that takes (Segal, 1989) preparation of colloidal dis-
place depends on the electronegativity dif- persions of oxides or hydrous oxides, usu-
ference between them. Electron transfer ally in an aqueous medium followed by
stops when both atoms have the same elec- dehydration of the dispersions (i.e., sols) to
tronegativity. Thus the electronegativity of gels that are calcined to oxides. The origin
atom /, K{ varies with the partial charge on of this sol-gel process dates to the late
the atom, dt. In this thermodynamic model 1950s when it was used to prepare micro-
which is analogous to determining chemi- spherical particles of almost theoretically
cal equilibrium in reactions by equalisa- dense (U,Th)O2 for application as fuel in
tion of the chemical potential of a species high-temperature thermal nuclear reac-
in different phases, a balance of electronic tors. Initial work was carried out at Oak
charge enables the following expressions Ridge National Laboratory in the U.S.A.
to be derived: where thoria sols were prepared (Haas,
1989) by steam denitration of hydrated
(3-6) thorium nitrate after which the residue was
peptised in H N 0 3 to a sol. The latter was
doped with uranyl nitrate solution and the
where % is the mean electronegativity of an doped sol was dispersed to an emulsion in
atom, x? is the electro-negativity of the an immiscible solvent capable of extracting
neutral atom and pt the atomic stoichiome- water from the sol, for example, 2-ethyl
try of the atom is equivalent to the number hexanol; gelation occurred on transfer of
of protons removed from the cation on H 2 O from the aqueous to organic phase. A
hydrolysis (Eq. (5)). The partial charge dt photomicrograph of thoria spheres (380 jim
can be written in the form diameter) made by dewatering of aqueous
sol droplets is shown in Fig. 3-3.
^ = Cc-X?)/1.36 Vx? (3-7) Advantages of sol-gel processing for fuel
where %t is related to St in the following fabrication were as a dust-free route to
way,
Xt = X? + ni5i (3-8)
and nt is the absolute hardness of an atom
(Parr and Pearson, 1983). The partial
charge model can be applied for predicting
hydrolysis products not only of cations,
but also anions as well as alkoxides in solu-
tion. As the hydrolysis products are pre-
cursors for nucleation of ceramic powders
the model has considerable potential appli- Figure 3-3. Photomicrograph of sol-gel thoria
cation for gaining insight into the chemical spheres, diameter 380 um, made by dewatering of sol
processes involved in powder preparation. droplets (Haas, 1989, courtesy of Oak Ridge Na-
tional Laboratory managed by Martin Marietta, En-
ergy Systems Inc. for the U.S. Department of Energy
under Contract No. DE-AC05-840R21400).
 3.7 Sol-Gel Processing of Colloids 79

high-density spherical powders. The calci- trates the aqueous solution with the result-
nation temperature of 1423 K was consid- ing formation of polymeric colloidal spe-
erably lower than for conventional fabrica- cies (Segal and Woodhead, 1986). The
tion techniques of powder mixing (2000 K). third route involves dispersion of a flame-
Mixing components at the colloidal level hydrolysed oxide (see Sec. 3.13), for exam-
produced good chemical homogeneity as ple, fumed silica in water (Rabinovich,
well as the observed lower reaction tem- 1988). Other experimental methods for
peratures whereas the reversibility of the sols are thermal denitration which was
sol-gel transition allows recycling of mate- used for thoria (see Sec. 3.7.2) and electro-
rial that did not fulfil a particle size specifi- dialysis and these techniques have been de-
cation. Thus these processing advantages scribed by Dell (1972) who has also sum-
of the technique over conventional powder marised early activities on sols and their
fabrication by using solid-state reaction conversion to powders. Sols can also be
between oxide powders (see Sec. 3.2) are prepared in non-aqueous liquids although
also applicable to ceramic powders of in- these dispersions are not associated with
dustrial interest. sol-gel processing of colloids. Antimony
pentoxide cake made by addition of HC1
to sodium antimonate was dispersed in a
3.7.3 Ceramic Powders from Colloids
mixture of a surfactant, an alkylbenzene-
Aqueous oxide or hydrous oxide col- sulfonic acid in naphtha and peptised with
loidal dispersions which are used in sol- benzylamine, C 6 H 5 CH 2 NH 2 (Watanabe
gel processing are often synthesized by one etal., 1988).
of three general methods. The first in- Sol-gel processing of colloids and
volves peptisation of hydrous oxide pre- metalorganic compounds is particularly
cipitates. For example, ceria sols have been associated with fabrication of thin (ca.
made (Woodhead, 1974) by first adding 1 jam) oxide coatings. However for powder
NH 4 OH/H 2 O 2 to Ce(m)nitrate. After preparation dewatering of sol droplets,
careful washing of the Ce (iv)hydrate to spray-drying (see Sec. 3.17) and gelation of
remove entrained electrolyte the precipi- sol droplets by an organic amine or NH 3
tate was peptised with H N 0 3 to sols with (g) are the main methods used and they
a particle size of around 8 nm. Base-sta- have been well-described by Woodhead
bilised sols have also been made by pepti- (1984). These gelation methods have been
sation. Thus Lane and coworkers (Lane applied to multicomponent oxides includ-
et al., 1990) showed that metastannic acid, ing plasma-spray Cr 2 O 3 -ZrO 2 powders
H 2 SnO 3 , which was obtained on precipita- around 10 Jim diameter (Scott and Wood-
tion from a solution of tin metal in HNO 3 head, 1982), yttria-based superconductors
could be peptised to a sol with an organic (Arcangeli etal., 1988) as well as catalyst
amine, for example, butylamine, C 4 H 9 NH 2 . supports, electrically conducting ceramics
While peptisation encompasses breaking- such as ferrites, thus Ni 0 3 Zn 0 7 Fe 2 O 4 and
up coarse precipitates, the second synthetic 3%SnO 2 -In 2 O 3 and also to stabilised zir-
route to sols involves growing molecular conia (Segal, 1989).
species to colloidal units in the form of
polynuclear ions (see Sec. 3.7.1). For ex-
ample, treatment of Cr(m)nitrate solution
with a long-chain primary amine deni-
80 3 Chemical Preparation of Powders

3.8 Sol-Gel Processing temperature. Hydrolysis at low pH pro-

of Metalorganic Compounds duces a gel that can be calcined to oxide
whereas hydrolysis at high pH nucleates
oxide powder directly from solution. Sol-
3.8.1 Preparation and Properties gel processing shares the advantage of sol-
of Metal Alkoxides gel processing of colloids (see Sec. 3.7.2)
Metalorganic compounds may be de- but additionally enhanced chemical homo-
fined as molecules in which the organic geneity is considered to result due to mix-
groups are bound to a metal atom via oxy- ing of components at the molecular rather
gen. While this definition includes for- than at the colloidal level. Detailed ac-
mates, acetates and acetylacetonates the counts of sol-gel processing of alkoxides
main metalorganic compounds of interest are given in the books by Brinker and
in sol-gel processing are metal alkoxides. Scherer (1990) and by Klein (1988) while
The latter have the general formula comprehensive review articles have been
M (OR)Z, where z is the valency for a metal published (Colomban, 1989; Sakka and
M and R is an alkyl group. Choice of syn- Yoko, 1991).
thetic methods for alkoxides depends on
the electronegativity of the main element
3.8.2 Powders from Alkoxides
and include direct reaction between metal
and alcohol for NaOC 2 H 5 and reaction Sol-gel processing of metal alkoxides
between an alkali metal alkoxide with a has been extensively used for preparation
metal chloride for Ti(OC 2 H 5 ) 4 (Bradley, of submicrometre oxide powders. Hydrol-
1989; Bradley et al., 1978). Two properties ysis (Bernier, 1986) of a 0.2 mol dm" 3 so-
of alkoxides are important for sol-gel pro- lution of Ba and Ti ethoxides in C 2 H 5 OH
cessing. The first is their volatility which by addition of a 0.5 mol dm" 3 solution of
allows their preparation with high chemi- H 2 O in C 2 H 5 OH followed by drying and
cal purity by distillation. The second prop- pyrolysis yielded a BaTiO3 powder with a
erty is their ability to undergo hydrolysis mean particle size of 40 nm. Examples of
which forms the basis of this sol-gel pro- other oxide powders made from alkoxides
cessing method. The hydrolytic reaction of are PbTiO 3 after gel calcination at 873 K
an alcoholic alkoxide solution can be rep- (Ma et al., 1991), WO 3 from WO(OC 2 H 5 ) 4
resented by the overall equation and WO 2 (OC 2 H 5 ) 2 (Obvintseva et al.,
1988) and LiTaO3 (Jean, 1990). For the
M(OR) z + zH 2 O -> M(OH) z + zROH
preparation of the latter a hydrated lithi-
The M(OH) molecules are unstable and um acetate solution in methoxyethanol,
undergo condensation and polymerisation CH 3 OCH 2 CH 2 OH was dehydrated at
reactions in which the polymers are col- 398 K and the anhydrous salt reacted with
loidal. These polymerisation processes can tantalum ethoxide in CH 3 OCH 2 CH 2 OH
result in oxide particles, at 398 K to form a double alkoxide;
methoxyethylacetate was removed from
M (OH)Z -> MO z/2 + z/2 H 2 O (8)
the alkoxide solution by distillation.
In practice, the hydrolytic reactions are Amorphous gels produced on hydrolysis
complex and the formation, size and shape crystallised to LiTaO3 powders at 723 K.
of the solution polymers depend on many Sodium-doped y-Fe 2 O 3 powders were
factors, including water content, pH and made (Yamanobe et al., 1991) by calcina-
 3.8 Sol-Gel Processing of Metalorganic Compounds 81

tion of gels derived from Fe(m)nitrate,

sodium ethoxide and ethylene glycol,
HOCH 2 CH 2 OH at 973 K while KSbO-
SiO4 powder for non-linear optical applica-
tions was synthesized from reaction of tetra-
ethoxysilane, Si(OC 2 H 5 ) 4 and KSb(OH) 6
under conditions of acid or base hydrolysis
followed by calcination at 1373 K (Kanno
et al., 1991). Metal alkoxides have attract-
ed much attention as precursors for syn-
thesis of high-temperature superconduc-
tors. For example thallium-based super-
conducting powders with a particle size of
50 nm were obtained by reaction of Ba, Ca
and Th methoxyethoxides with Cu(n)ethox-
ide (Kordas and Teepe, 1990).
Alkoxide hydrolysis has been used to
prepare spherical, monodispersed sub-
micrometre oxide powders by the nucle-
ation and growth processes proposed by
LaMer and Dinegar (1950) and described
by Sec. 3.4.1. An example of the quality of Figure 3-4. Silica spheres made by hydrolysis of tetra-
spheres which can be obtained is shown, ethoxysilane (Milne, 1986, courtesy of the Institute of
for silica, in Fig. 3-4 (Milne, 1986). Fur- Ceramics).
ther examples of spherical powders made
by controlled hydrolysis are documented
by Brinker and Scherer (1990) and, for zir- the amorphous germanium sulphide crys-
conia, by Ogihara et al. (1987). Prepara- tallised on heating. However, a recent re-
tion of submicrometre spherical powders examination of this reaction (Seddon
from alkoxides has been extended to mate- etal., 1991) involving detailed characteri-
rials with surface functional groups (Riman sation of the products by X-ray diffraction
et al., 1989). For example, triisopropoxyti- and infrared spectroscopy showed that
tanium propanoate (C 3 H 7 O 3 )TiOOCC 2 H 5 GeS2 could be produced but was easily
was hydrolysed by addition of a solution contaminated with GeO 2 which had a sim-
of H 2 O in C 3 H 7 OH. The resulting uni- ilar X-ray diffraction pattern. Finally a
form spherical carboxy-hydrosols had par- complex submicrometre sulphide powder,
ticle diameters between 0.4 and 2.9 jum CaLa 4 S 4 , has been prepared (Wang et al.,
and contained surface carboxylic groups 1990) by reaction of CS 2 liquid with mixed
(— COOH) as only alkoxy groups in the Ti solutions of Ca and La methoxides fol-
alkoxide are removed on hydrolysis. lowed by sulphidization of the reaction
Sol-gel processing of metal alkoxides is product, an amorphous powder in H 2 S gas
not restricted to oxide systems and has at temperatures between 673 and 1023 K.
been applied to sulfide powders. Melling
(1984) showed that H 2 S gas reacted with
Ge(OC 2 H 5 ) 4 at ambient temperature and
82 3 Chemical Preparation of Powders

3.9 Alternative Gel Routes Table 3-2. Powders made by the Pechini method
(Eror and Anderson, 1986).
to Powders
Titanates: BaTiO 3 , SrTiO 3 , Pb (La, Zr, Ti) O 3
The phrase sol-gel processing has been Niobates: BaNb 2 O 6 , Pb 3 MgNb 2 O 9
used rather indiscriminately in the scientif- Zirconates: CaZrO 3
ic literature. While it refers strictly to Chromites: LaCrO 3 , MgCr 2 O 4
chemical preparations involving formation Ferrites: LiFeO 2 , CoFe 2 O 4
of gels from colloidal oxides or hydrous Manganites: LaMnO 3 , YMnO 3
oxide dispersions (see Sec. 3.7.3) as well as Aluminates: LaAlO 3 , MgAl 2 O 4
to hydrolytic reactions of metalorganic Cobaltites: LaCoO 3 , YCoO 3 , PrCoO 3
compounds such as alcoholic alkoxide so- Silicates: Zn 2 SiO 4
lutions (see Sec. 3.8.2), the term sol-gel has
frequently been used to describe ceramic
syntheses that involve formation of a rigid homogeneity through mixing at the molec-
gel or viscous resin intermediate from a ular level in solution, and control of the
liquid ceramic precursor. Preparative stoichiometry. Low firing temperatures are
methods which are described in this sec- required for conversion of resin to oxide.
tion have been particularly prone to this Thus, in the original description, BaTiO 3
incorrect nomenclature. powder was produced at 923 K compared
to 1273 K for solid-state reaction between
The Pechini Method BaCO3 and TiO 2 . In recent years the Pe-
chini method has been applied to many
In the Pechini method, named after complex compositions (Table 3-2; Eror
its inventor (Pechini, 1967) polybasic and Anderson, 1986) and it has become
chelates are formed between oc-hydro- more common to introduce metallic ions
xycarboxylic acids containing at least one from nitrate solutions; crystallite sizes of
hydroxyl group, for example citric acid, the oxide powders made by this method
HOC(CH 2 CO 2 H) 2 CO 2 H and glycolic are around 50 nm. The Pechini method has
acid, HOCH 2 CO 2 H with metallic ions. The attracted attention for preparation of
chelate underwent polyesterification on high-temperature ceramic superconduc-
heating with a polyfunctional alcohol, for tors, for example, Lax 85 Sr 0 15Cu04_CT
example ethylene glycol, HOCH 2 CH 2 OH. (Falter et al, 1989) and YBa 2 Cu 3 0 7 _ G
Further heating produced a viscous resin, (Kakihana etal., 1991) while substitution
then a rigid transparent, glassy gel and fi- of polyacrylic acid for citric acid has been
nally fine oxide powder. The original described by Lessing (1989). A potential
patent was limited to synthesis of lead and advantage of polyacrylic acid over citric
alkaline earth titanates, niobates and zir- acid is that its functionality, that is, the
conates whereby metallic ions were intro- number of reactive sites, is about 28 for a
duced from hydrous oxides, alkoxides or molecular weight of 2000 compared to 4
a-hydroxycarboxylates of Ti, Nb and Zr for citric acid: the higher functionality can
(e.g., zirconium lactate) and from oxides, aid the formation of the cross-linked poly-
hydrous oxides, carbonates or alkoxides mer resin.
for lead and alkaline earth metals. Advan-
tages of the Pechini method are the ability
to prepare complex compositions, good
 3.10 Hydrothermal Synthesis 83

The Citrate Gel Method (1953). Nowadays hydrothermal tech-

niques are widely used in industrial pro-
The citrate gel method was developed by
cesses for dissolution of bauxite prior to
Marcilly et al. (1970) and can be illustrated
precipitation of gibbsite in the Bayer pro-
by the preparation of the ceramic super-
cess and for preparation of aluminosilicate
conductor YBa2Cu3O7_CT (Blank etal.,
zeolites. However, only in recent years has
1988). Metal nitrate solutions of Y, Ba and
hydrothermal synthesis attracted increas-
Cu were added to citric acid solution and
ing attention for synthesis of advanced ce-
the pH raised to between 6.5-7.0 in order
ramic powders although interest in the
to dissolve insoluble barium citrate but not
method, as measured by the number of
to precipitate metal hydroxides. The solu-
publications, is considerably less than in
tion which contained polybasic chelates
sol-gel processing. This is surprising be-
was concentrated to a viscous resin and
cause hydrothermal synthesis offers a low
dried to a transparent gel that was py-
temperature, direct route to oxide pow-
rolysed to a fine powder. The citrate gel
ders, avoiding the calcination step required
and Pechini methods share the same ad-
in sol-gel processing. When applied to ce-
vantages with respect to chemical homoge-
ramic powders hydrothermal techniques
neity and compositional control.
often involve heating metal salts, oxides or
hydroxides as a solution or suspension in a
Acetate Precursors for Gels
liquid at elevated temperature and pres-
A resurgence of interest in the Pechini sure up to about 573 K and 100 MPa. Oth-
and citrate gel methods followed the dis- er routes to oxide powders (Somiya, 1991)
covery of high-temperature superconduc- include use of metal powders, hydrother-
tors (Bednorz and M tiller, 1986) which has mal anodic oxidation and reactive elec-
also highlighted another technique that in- trode submerged arc (RESA) processing
volves gel formation from metal acetate so- whereby reaction takes place in the arc re-
lutions. Acetate salts such as copper acetate gion between an electrode and liquid, kept
are weakly acidic in aqueous solution and under hydrothermal conditions.
act as buffer solutions. When copper ace- Barium titanate powders have been pre-
tate was added (Tarascon et al., 1988; Bar- pared by hydrothermal synthesis (Hen-
boux et al., 1988) to colloidal yttrium and nings etal., 1991). Thus a stable sol was
barium hydroxide dispersions, a blue glass- first prepared by mixing a Ti alkoxide (bu-
like amorphous gel was produced on drying toxide or ethoxide) with isopropanol and
which could not be obtained with nitrate acetic acid in the molar ratio, alkoxide: al-
or fluoride salts of copper. Heat treatment cohol: a c i d = l : 6:3. The sols were then
of the gel produced small (1 jam) particles mixed with barium acetate solution and
of superconducting YBa 2 Cu 3 O 7 _ a . acetic acid which resulted in formation of
a rigid gel. A suspension of the gel
(pH > 13) in tetramethylammonium hy-
3.10 Hydrothermal Synthesis droxide, N(CH 3 ) 4 OH was heated at 423 K
and 15 MPa for 10 hours. Dissolution-re-
Hydrothermal techniques are not new crystallisation processes occurred in asso-
and their application during the last centu- ciation with Ostwald ripening on this ther-
ry to the synthesis of minerals of geological mal treatment producing a monodis-
interest has been described by Morey persed, weakly agglomerated BaTiO3 dis-
84 3 Chemical Preparation of Powders

1423 and 1523 K in the conventional solid-

state synthesis by firing a mixture of oc-
Fe 2 O 3 and BaCO 3 .
The hydrolysis of metal ions in solution
was described in Sec. 3.7.1 and is shown
schematically in Eq. (5). The latter repre-
sents a forced hydrolysis reaction where
added base is not required to produce link-
ages between cations and hydroxyl groups.
Under hydrothermal conditions forced hy-
drolysis of cations in the absence of base is
more pronounced than at ambient temper-
ature and these hydrolytic reactions can
result in direct formation of oxide powders
from electrolyte solutions. In these reac-
tions polynuclear cations are the species
which result in supersaturated solutions
and are the precursors for nucleation from
solution. Thus, aggregated crystalline
Figure 3-5. Hydrothermally prepared barium titanate
powder (Hennings et al., 1991). TiO2 powder with a submicrometre prima-
ry particle size was produced by ageing an
aqueous Ti(iv)chloride solution between
persion, approximately spherical with par- 433 K and 503 K at 1.4-9.6 MPa for 1-2
ticle size 200-300 nm; an electron mi- hours (Kutty et al., 1988): the rate of pow-
crograph of the BaTiO 3 powder is shown der formation increased at higher tempera-
in Fig. 3-5. Another hydrothermal synthe- tures. Rutile was the crystalline phase al-
sis of spherical BaTiO3 powders, 0.1 jim in though presence of SOj" anions in the re-
diameter, was carried out (Fukai et al., actant solution yielded anatase. This ex-
1990) by treatment of a mixture of ample illustrates the high chemical purity
Ba(OH)2 and titanium(iv)hydroxide at that can be obtained by hydrothermal syn-
473 K for 5 hours, the latter made by ad- thesis. Hence hydrothermally derived TiO2
dition of aqueous ammonia solution to contained 32 ppm (parts per million) of Fe
Ti(iv)chloride solution. The ferroelectric compared with 2450 ppm in titania pow-
powder had a specific surface area of der made by calcination of the hydrous
10.6 m 2 g~ *; it was of high chemical purity titanium oxide obtained on addition of
and consisted of a single crystalline phase; aqueous ammonia solution to Ti(IV)chlo-
the synthesis was extended to single phase ride. Higher chemical purity arose because
BaTi0 8 Zr 0 2 O 3 . The advantage of hydro- iron compounds remained in solution after
thermal processing is illustrated by synthe- hydrothermal ageing of Ti(IV)chloride.
sis of BaFe 12 O 19 (Barb et al., 1986) from a The rutile powder could be converted to
suspension of barium hydroxide and oc- aggregated BaTiO3 powder with primary
FeOOH. Single-phase submicrometre fer- particle sizes between 0.3-0.6 jim by sus-
rite particles were obtained by dissolution- pension in Ba(OH)2 solution followed by a
recrystallisation of reactants at 598 K dissolution-recrystallisation process at
compared with a temperature between 463 K for 6 hours.
 3.11 Forced Hydrolysis of Salt Solutions 85

Another example of forced hydrolysis at

elevated temperatures and pressure is the
preparation of monoclinic ZrO 2 powder
from zirconyl nitrate, ZrO(NO 3 ) 2 (Denke-
wicz et al., 1990). Particles, 1-10 nm in di-
ameter, were produced by heating a 0.2 mol
dm~ 3 solution of zirconyl nitrate in Zr at
pH 0.5. A proposed mechanism for particle
formation involved association of a soluble
tetrameric species [Zr(OH) 2 -4H 2 O] 4 8+ to
form particles of critical size which nucle-
ated from solution and grew by aggrega-
tion.

3.11 Forced Hydrolysis of Salt

Solutions
While forced hydrolysis of electrolyte so-
lutions occurs at elevated temperature and
pressure (Sec. 3.10) it can also take place at
atmospheric pressure by use of elevated
temperatures. Thus, monodispersed amor-
phous chromium hydroxide (Matijevic,
1985,1988) was obtained from CrK(SO4)2 • Figure 3-6. Scanning electron micrographs of yttrium
oxide (a) prepared by forced hydrolysis and (b) com-
12H 2 O, chromic sulphate and nitrate
mercial grade powder (Sordelet and Akinc, 1988 a.
after ageing between 333-348 K when Reprinted by permission of the American Ceramic
the Cr(III) concentration was between Society).
2 x 10~ 4 -2 x 10~ 3 mol dm" 3 , the solution
had a pH<5.4 and contained SO^" ions;
particle diameter were betwen 293- N H 4 + and the cyanate ion, OCN (Shaw
490 nm. Another way of initiating forced and Bordeaux, 1955), at the acidic pH of
hydrolysis is by use of slow release of pre- the mixture and NCO3" ions then hydroly-
cipitating anions, for example, by decom- ses to CO 2 . Monodispersed YOHCO 3 pow-
position of urea, CO(NH 2 ) 2 in aqueous ders, 0.4 |Lim diameter, were obtained in a
solutions. Thus homogeneous precipitation reproducible manner and were decom-
of basic yttrium carbonate, YOHCO 3 was posed to oxide at 973 K. Scanning electron
achieved (Sordelet and Akinc, 1988 a, b) micrographs for yttrium oxide spheres
when mixed solutions of yttrium nitrate made by forced hydrolysis and for com-
(0.025 mol dm" 3 ) and CO(NH 2 ) 2 mercial grade yttria powder are shown in
(0.17 mol dm" 3 ) were kept at 373 K for Figs. 3-6 a and b. Evolution of CO 2 during
1 hour; low cation concentration is charac- decomposition of urea was essential for
teristic of reactions involving forced hy- particle formation as substitution of
drolysis in order to minimize bursts of nu- CO(NH 2 ) 2 by formamide which decom-
cleation. Urea decomposes initially to poses to formic acid and NH 3 , did not
86 3 Chemical Preparation of Powders

result in a precipitate. Monodispersed pyrolysis (see Sec. 3.15). In addition,

powders, about 0.3 |im in diameter have alkoxide-derived sulphide powders (see
also been prepared by homogeneous pre- Sec. 3.8.2) can also be considered to be the
cipitation of mixed solutions of yttrium, products of liquid-phase reactions and this
barium and copper nitrate in the presence class of reactions is associated with the
of urea at 373 K (Kayima and Qutubud- synthesis of non-oxide powders, particu-
din, 1989); carbonate powders could be de- larly silicon nitride.
composed on heating to oxide with a su- The reaction between silicon tetra-
perconducting composition. chloride and ammonia was investigated as
Non-oxide powders have been synthe- early as 1830 (Segal, 1989), although an
sized by forced hydrolysis (Matijevic, 1985). intermediate polymer product, silicon
Thus spherical CdS particles about 1.5 jim diimide, has been difficult to characterise.
in diameter were made by ageing a mixture Mazdiyasni and Cooke (1976) reacted
that was 0.0012 mol dm" 3 Cd(NO 3 ) 2 , SiCl4(l) with NH 3 (g) in dry hexane at
0.24 mol dm" 3 HNO 3 and 0.005 mol 273 K and concluded that polymeric inter-
dm" 3 in thioacetamide, CH 3 CSNH 2 , at mediates may have formed when initial re-
299 K for 14 hours. Slow hydrolysis of this action products underwent vacuum de-
amide results in a "reservoir" of precipitat- composition. The initial solid reaction
ing anions (S2~) which prevents local high product was amorphous to X-rays for de-
S2~ concentrations and thus non-uniform composition up to 1473 K but trans-
precipitation. Powders can also be made formed to oc-Si3N4 with a particle diameter
by slow release of cations through decom- 10-30 nm on prolonged heating between
position of a complexing agent rather 1473 and 1673 K. Very pure nitride pow-
than by release of precipitating anions. For ders were obtained with total impurities
example, copper(i)oxide in various mor- < 300 ppm while Fe and Ni, the main con-
phologies including cubic and octahedral taminants, were both present at less than
has been obtained by heating a solution 100 ppm. This liquid-phase reaction is not
of 2.7xlO~ 3 mol dm" 3 Cu(n)tartrate of obscure academic interest, as the inter-
(also known as Fehling's solution) and facial reaction between SiCl4(l) and
5.3xlO" 4 mol dm" 3 glucose at 368 K NH3(1) at 233 K has been scaled up for
(Matijevic, 1985). In this reaction, the production of commercial powders (Segal,
Cu(n)tartrate complex undergoes a con- 1989) that consist of equiaxed particles
trolled decomposition followed by reduc- about 0.2 jim in diameter. A variation of
tion of Cu(n) to Cu(i); the particle mor- the reaction between silicon tetrachloride
phology was determined, primarily, by the and ammonia has been made by Crosbie
concentration of reactants. et al. (1990) in which SiCl4(g) was reacted
with NH4(1) which yielded equiaxed Si 3 N 4
3.12 Non-Aqueous Liquid-Phase powders with particle diameters 0.2-
0.3 jim.
Reactions
In recent years liquid-phase reactions
Liquid-phase reactions take place in a have attracted attention for synthesis of
non-aqueous solvent which may be inert or aluminium nitride powders. Maya (1986)
one of the reactants. It can be difficult to developed synthetic pathways to A1(NH2)3
distinguish liquid-phase reactions from re- which was considered a useful intermedi-
actions which are associated with polymer ate compound for preparation of A1N
 3.12 Non-Aqueous Liquid-Phase Reactions 87

powders. In one synthesis, lithium alu- A solution of the alane in toluene was re-
minium hydride, LiAlH4, was reacted with acted with excess of NH 3 at 203 K. The
AICI3 at ambient temperature in diethyl product, a white precipitate was pyrolysed
ether, (C 2 H 5 ) 2 O, which yielded the adduct in N 2 at 1273 K to AIN powder with a
AIH3 • (C 2 H 5 ) 2 O. A solution of this adduct surface area of 118 m 2 g" 1 and a particle
was then reacted with excess NH 3 at size <0.1 jim. High-purity powders are
223 K. The polymeric reaction product, a characteristics of ceramics made from liq-
white precipitate with empirical formula uid-phase reactions and the oxygen con-
[Al(NH 2 ) 0i864 NH 1#069 ] B could be pyro- tent after heat treatment at 1473 K was
lysed to AIN powder, about 1 jim in size at 0.5wt.%. Another reaction pathway (In-
temperatures greater than 873 K. Reaction terrante et al., 1986) to AIN involves use of
between A1H3 and NH 3 in tetrahydro- trialkylaluminium compounds, for exam-
furan was studied by Ochi and coworkers ple, trimethylaluminium, (CH3)3A1 and
(1988). All of the hydrogen atoms in the triethylaluminium, (C2H5)3A1. The adduct
hydride were replaced by amino or imino (CH3)3A1 • NH 3 was produced by bubbling
groups during reaction at 243 K in the pres- NH 3 gas through (CH3)3A1 in a hydrocar-
ence of excess ammonia, an observation bon solvent at 195 K. Reaction at a higher
noted previously by Maya (1986). A white temperature, 203 K, yielded (CH3)2A1NH2
powder with the formula (Al(NH)NH2)n that could be collected as a solid with trimer
was obtained and pyrolysed to AIN under structure as shown by X-ray analysis. Reac-
vacuum at 1373 K. However, reaction with tion of (C2H3)3A1 with NH 3 produced an
a stoichiometric amount of NH 3 at 193 K adduct that on pyrolysis yielded high-puri-
gave a precipitate which dissolved on ty AIN with a surface area between 40 and
raising the temperature. Polymerisation 80m 2 g" 1 , oxygen content <0.3wt.%
occurred at ambient temperature with and carbon content of 0.06 wt.%; decom-
elimination of H 2 to give a solid gel con- position occurred through a series of inter-
taining hydroaluminium imide, (HA1NH). mediate alkyl aluminium amide and imide
Pieces of AIN with a crystallite size of species.
30 nm but contaminated with carbon were Synthesies of AIN described so far in
produced after pyrolysis of the gel at this section involve air-sensitive precur-
1373 K in a vacuum; carbon contamina- sors. A novel electrochemical route to alu-
tion was due to decomposition of tetrahy- minium nitride ceramics which avoids haz-
drofuran associated with the gel. ards associated with these precursors as
Dimethylaminoalane, H2A1N(CH3)2 has well as carbon contamination arising from
also been considered as a precursor for solvent decomposition has been developed
AIN powder (Einarsrud etal., 1989); its (Seibold and Riissel, 1988). Aluminium
solubility in solvents aids its removal and was anodically dissolved in a solvent of
thus, potentially reduces carbon contami- high polarity such as acetonitrile contain-
nation on pyrolysis. It was prepared by ing a primary amine and a tetraalkylam-
reaction of dimethylamine hydrochloride, monium salt, the latter for increasing the
HN(CH 3 ) 2 HC1 with LiAlH4 according conductivity of the solution. A liquid with
to the equation composition A1(NHR)3 could be obtained
which, after removal of excess amine and
LiAlH4 + HN(CH 3 ) 2 HCl -* (9) solvent, underwent polymerisation and set
-* H 2 AlN(CH 3 ) 2 + LiCl-h2H 2 to a gel with a final composition of
88 3 Chemical Preparation of Powders

A12(NR)3. An aluminium nitride powder

with a crystallite size of 25 nm was ob-
tained on pyrolysis of the gel above 1073 K
in NH 3 .
Liquid-phase metathetical reactions
have been developed for synthesis of tran-
sition metal dichalcogenides (Chianelli
and Dines, 1978). As an example reaction
between TiCl4(l) and H2S(g) is thermody-
namically unfavourable below 673 K.
However the tetrachloride will react quan-
titatively with a sulphide with greater ionic Figure 3-7. Scanning electron micrograph of tita-
character than H 2 S, for example, Li2S at nium diboride made by a non-aqueous liquid phase
298 K in a non-hydroxylic solvent such as reaction (Gallagher et al., 1988, by permission of the
tetrahydrofuran. Amorphous sulphide authors and the Ceramics Processing Research Labo-
crystallised between 673-873 K in an O 2 - ratory of MIT; ©1988 John Wiley & Sons, Inc.,
reprinted by permission).
free atmosphere to a submicrometre TiS 2
powder with surface area up to 96 m 2 g" 1 .
Reaction between TiCl4(l) and H 2 S could bubbling diborane, B 2 H 6 , through a solu-
take place at 298 K in tetrahydrofuran in tion of Ti(OC 4 H 9 ) 4 in tetrahydrofuran.
the presence of a long-chain amine, for ex- The solid borohydride decomposed at
ample trihexylamine (C 6 H 13 ) 3 N, accord- 413 K when heated in xylene,
ing to the reaction
2Ti(BH 4 ) 3 (12)
TiCl4 + 4(C 6 H 1 3 ) 3 N + 2H 2 S -> (10)
-> TiS 2 +4(C 6 H 1 3 ) 3 NHCl The resulting TiB2 powder was agglomer-
ated, of high chemical purity with a parti-
A rigid gel containing both amorphous cle size 0.1-0.2 jim; a scanning electron
sulphide and amine hydrochloride was micrograph for TiB2 powder is shown in
produced; crystalline TiS2 was obtained Fig. 3-7. The decomposition to a diboride
on gel calcination at 673 K. An aggregated at low temperatures illustrates an advan-
cubic (3-ZnS powder was obtained (John- tage of this liquid-phase reaction.
son et al., 1986) with primary particles
100 nm or smaller in size on passing H 2 S
gas through a diethyl zinc solution in hep- 3.13 Gas-Phase Reactions
tane at room temperature as indicated by
Reactions between gases have been used
Zn(C 2 H 5 ) 2 + H 2 S -» ZnS + 2C 2 H 6 (11)
to prepare both oxide and non-oxide pow-
A common feature of both this liquid- ders. Experimental work on these systems
phase reaction and sol-gel processing of is characterised by use of a variety of heat-
metalorganic compounds is that compo- ing techniques including furnace heating,
nents are mixed on the molecular level. lasers, gas plasmas and flame propagation
Liquid-phase reactions have been ex- while the underlying chemical principle
tended to metal diboride powders (Gal- that determines particle formation is ho-
lagher etal., 1988). Thus titanium(m)- mogeneous nucleation from supersaturat-
borohydride, Ti(BH 4 ) 3 can be prepared by ed vapours.
 3.13 Gas-Phase Reactions 89

Titanium dioxide powders have been Table 3-3. Oxides made by flame hydrolysis (Kriech-
synthesized (Suyama et al., 1975) in an baum and Kleinschmit, 1989).
electrically heated furnace by reaction of Oxide Raw Material Boiling Point (K)
TiCl4 and O 2 between 1173 and 1573 K.
Average particle sizes of 50-200 nm de- SiO2 SiCl4 330
creased with a reduction in TiCl4 concen- A12O3 AICI3 453 a
tration but increased on raising the reac- TiO 2 TiCl 4 410
tion temperature and oxygen concentra- ZrO 2 ZrCl 4 604a
tion; the rutile content increased from ZrO 2 /TiO 2 ZrCl 4 /TiCl 4 604 a/410
about 2 wt.% at 1173 K to nearly 40 wt.% Cr 2 O 3 CrO 2 Cl 2 390
at 1573 K. A supersaturated vapour of Fe 2 O 3 Fe(CO) 5 376
GeO 2 GeCl 4 357
gaseous TiO 2 was produced which resulted
NiO Ni(CO) 4 315
in formation of oxide clusters (see
SnO 2 SnCl4 387
Sec. 3.4). After the clusters reached a criti-
cal size, homogeneous nucleation occurred v2o5 VOCI3 400

and nuclei grew by adsorption of gaseous a

Sublimation temperature.
TiCl4 followed by decomposition of the
adsorbed tetrachloride to oxide. Critical
radii for rutile formation were smaller than In addition, an adduct between A1C13 and
for anatase. Thus for a 5 % conversion of NH 3 can be formed,
reactants at 1473 K the supersaturation AICI3+NH3 <=• AICI3NH3 +±
and critical radius were 4.5 and 0.33 nm
for anatase but 5.5 and 0.24 nm for rutile. *± AlN(s) + 3HCl(g) (16)
It was postulated that an anatase cluster A colourless A1N powder, amorphous to
formed initially due to its less dense pack- X-rays, was obtained in a 37 wt.% yield at
ing compared to rutile, then underwent 1473 K if NH 3 was introduced cold into
atomic rearrangement to rutile. This pro- the vertical chemical vapour deposition re-
cess became the dominant mechanism for actor before reaction with hot aluminium
rutile formation above 1173 K. Another chloride vapour. The powder was spheri-
synthesis involving furnace heating is the cal with a mean diameter of 77 nm, had a
preparation of A1N powder from mixtures nitrogen surface area of 23.5 m 2 g" 1 and
of A1C13 and NH 3 (Riedel and Gaudl, was free of residual chlorine; crystallisa-
1991). The following reactions can occur tion took place at 2173 K. All of this non-
between these reactants because alumini- oxide ceramic was present as a coating on
um chloride dimerises at its sublimation the reactor wall the gaseous reactants both
temperature (Table 3-3) and NH 3 can dis- being hot when mixed at 1273 K. Howev-
sociate into its elements above 400 K and er, when mixed first and then heated up to
0.1 MPa. 1123 K, the reactants yielded a solid prod-
(13) uct through formation of the adduct in
AlCl3(g) + NH 3 (g) *± AlN(s) + 3HCl(g) Eq. (16) which decomposed to yield A1N
coatings.
Al2Cl6(g) + 2NH 3 (g) (14) In flame hydrolysis (Kriechbaum and
2AlN(s) + 6HCl(g) Kleinschmit, 1989) volatile compounds
such as TiCl4 are passed through an oxy-
2NH, (15) gen-hydrogen stationary flame. Molten
90 3 Chemical Preparation of Powders

primary oxide particles formed by nucle- 2000 K. Powders were amorphous to X-

ation grow by coalescence to larger rays with a nitrogen surface area of 170 m 2
droplets. As particles solidify they stick to- g~ * and particle diameter in the range 10-
gether on collision forming solid aggre- 50 nm.
gates that then associate to loosely bound Use of high-powered lasers to heat gas-
agglomerates. Examples of flame-hy- eous reactants was pioneered by Haggerty
drolysed or fumed oxides which have been and coworkers (Cannon etal., 1982). Si-
made are shown in Table 3-3 and a trans- lane, SiH4, which had a strong absorbance
mission electron micrograph for a fumed at 10.6 |im, near the wavelength of CO 2
A12O3 is shown in Fig. 3-8. For mixed ox- lasers, was decomposed to a supersaturat-
ides such as ZrO 2 /TiO 2 chemical homoge- ed silicon vapour in a cross-flow cell and
neity can be maintained on the molecular the vapour reacted with NH 3 to silicon
scale while particle sizes are of the order of nitride powder. The reactions are charac-
50 nm. Whereas flame hydrolysis involves terised by fast heating and cooling rates,
a stationary flame, the propagation of a 106 K s~1 and 105 K s~ 19 respectively, and
flame through a gaseous mixture for main- reaction times around 10 ~3 s. Amorphous
taining a self-sustaining reaction has been powder had a particle size between 10-
investigated (Calcote et al., 1990) for pro- 25 nm and a nitrogen surface area of
duction of Si 3 N 4 powder. Stable flames 117 m 2 g" 1 . A recent laser driven reaction
could be maintained in mixtures of SiH 4 has used chlorinated silanes, for example,
and N 2 H 4 diluted in a carrier gas (N 2 ), at dichlorosilane, SiH2Cl2, (Bauer et al., 1991)
100 kPa for temperatures between 900 and as a reactant because they are cheaper and
safer to handle than silane. Dichlorosilane
reacts with NH 3 between 1100 and 1500 K
according to the equation

SiH2Cl2 + 4NH 3 Si(NH) 2 (s) + (17)

+ 2NH 4 C1 + 2H 2
where silicon diimide, Si(NH) 2 (s) decom-
poses to Si 3 N 4 , thus
3Si(NH) 2 (s) *± Si 3 N 4 (s) + 2NH 3 (18)
At temperatures between 1100 and 1200 K
a silicon nitride nucleus is formed that
grows by surface deposition,
3SiH 2 Cl 2 + 4NH 3 +± Si 3 N 4 (s)+ (19)
+ 6HC1 + 6H 2
but above the dissociation temperature of
Si 3 N 4 (2170 K), a Si nucleus formed which
1 jjtii reacted with ammonia,
Figure 3-8. Transmission electron micrograph of SiH2Cl2 +± Si(s,l) + 2HC1 (20)
flame-hydrolysed alumina C powder. (Courtesy of
Degussa AG, Germany). 3Si(s,l) + 4NH 3 *± Si 3 N 4 (s) + 6H 2 (21)
 3.14 Aerosol-Derived Powders 91

current (d.c.) argon plasma (Blackburn

etal., 1991). Titanium tetrachloride was
reacted with NH 3 at 1373 K in a reaction
zone heated by the argon plasma. Spheri-
cal TiN powders with diameters around
10 nm were produced and are shown in
the transmission electron micrograph in
Fig. 3-9. A comparison of the particle size
of this plasma-derived powder can be
made with commercial TiN prepared by
nitridation of titanium (see Sec. 3.2) in
Fig. 3-1 which shows that the plasma route
produces a much finer particle size. Titani-
um nitride powder made from these gas-
eous reactants could be pressureless sin-
tered at 1673 K to a ceramic with flexural
0.1 um
strength 500 MPa and fracture toughness
Figure J-y. Transmission electron micrograph ot 4 MPa mVi; the conventional powder il-
plasma-derived titanium nitride powder (Blackburn
et al., 1991, courtesy of Tioxide Specialties Limited
lustrated in Fig. 3-1 is difficult to sinter
and the Institute of Ceramics). and requires hot-pressing at temperatures
higher than 2073 K.

Loosely agglomerated Si 3 N 4 powder with

particle diameter between 25 and 100 nm 3,14 Aerosol-Derived Powders
were produced from the reactants. Laser
driven reactions (O'Neill et al., 1989) have Aerosols are colloidal systems and con-
also been carried out on chloromethylsi- sist of liquid or solid droplets in a gaseous
lanes such as (CH3)2SiCl2 as these reac- dispersion medium. Aerosol routes to
tants were cheaper and easier to handle powders fall into two categories. The first
than chlorosilanes or silane. Unlike chloro- involves generation of a supersaturated va-
silanes, the methylsilanes do not absorb pour from a reactant followed by homoge-
radiation from CO 2 lasers but have an ab- neous nucleation while the second involves
sorption band at 193 nm which corre- generation of liquid droplets, by a variety
sponds to the wavelength of radiation of techniques, which undergo a heat treat-
emitted by pulsed Ar-F excimer lasers. De- ment to solid particles. Thus, Okuyama
composition of reactants into fragments et al. (1986) produced a supersaturated va-
produced a supersaturated vapour of pour of TiO 2 by thermal decomposition of
atomic Si and powders formed by conden- Ti(OC 3 H 7 ) 4 using furnace temperatures
sation of decomposition products. Ag- up to 1273 K. Homogeneous nucleation
glomerated |}-SiC and amorphous Si pow- from the oxide vapour led initially to pri-
ders with low chlorine content and particle mary oxide particles which grew by coales-
sizes between 10 nm and 1.5 jim were pro- cence and by heterogeneous nucleation of
duced. vapour molecules onto clusters; particle
The final illustration of gas-phase reac- sizes for oxide powders made by this meth-
tions is the synthesis of TiN in a direct od were around 100 nm im diameter.
92 3 Chemical Preparation of Powders

Ultrasonic atomisation is a popular mean particle diameter of 0.24 jam (Ishiza-

technique for generation of liquid droplets wa et al., 1986); ultrasonic generation of
in the route to aerosol-derived powders. droplets was carried out from a mixed so-
The diameter of droplets generated by ul- lution of Zr(OC 4 H 9 ) 4 and Y(OC 3 H 7 ) 3 in
trasonic cavitation of a solution can be C 2 H 5 OH. Powders were calcined at
written in the form (Lang, 1962) 1173 K and, after compaction, sintered at
1723 K to a body with a grain size of
0.77 jum and fracture toughness of
where d is the calculated droplet size, y is ll.OMPa m 1/2 for a 2mol% yttria con-
the surface tension, Q is the density of the tent. This large fracture toughness arose
solution and / is the ultrasonic frequency due to the high quality of the powder with
employed for cavitation. In a recent study respect to its submicrometre size and unag-
(Ley etal., 1991) liquid droplets about glomerated nature.
0.4 jam in diameter were generated by ul- Electrostatic atomisation has also been
trasonic cavitation of a solution contain- used for droplet generation. Typically, a
ing Y, Ba and Cu nitrates. Droplets were high electrical potential (15 kV) is applied
transported to a furnace and mist pyrolysis to the liquid contained in a capillary tube.
took place at temperatures up to 1000 K. This potential causes the meniscus at the
A scanning electron micrograph for super- exit of the capillary to form a cone whose
conducting YBa 2 Cu 3 O 7 _ 5 powder made surface becomes mechanically unstable
by this method is shown in Fig. 3.10. Ad- due to charge accumulation. Liquid jets
vantages of this aerosol method are the are ejected from the cone and decompose
ready availability of reactant solutions and into droplets due to a Rayleigh instability
ability to produce unaggregated sub- (Rayleigh, 1878). This generation tech-
micrometre powder sizes. Decomposition nique (Slamovich and Lange, 1988) was
of aerosol droplets in a furnace is some- applied to droplets obtained from a zirco-
times referred to as spray-pyrolysis which nium acetate solution containing a poly-
was used to prepare dense unagglomerated ethylene oxide (molecular weight 200 000)
Y2O3-stabilised zirconia powder with a which increased the diameter from 100 nm
up to around 5 |im. Pyrolysis of droplets
yielded spherical oxide powders, either sin-
gle crystals, in the absence of Y 2 O 3 , or
polycrystalline submicrometre powders
when stabilised with yttria. Aerosol routes
are not restricted to oxide ceramics. Thus
micrometre-sized liquid droplets which
were generated from a solution of a poly-
(borazinylamine) in liquid ammonia could
be decomposed (Lindquist et al., 1991) in a
flame reactor at 1273 K to spherical pow-
ders, 0.5 jim in diameter that were amor-
phous to X-rays. Further heating of this
Figure 3-10. Yttrium-based superconducting powder
made by aerosol mist pyrolysis (Ley et al., 1991, cour-
powder at 1873 K under N 2 atmosphere
tesy of Professor B.C.H. Steele, Imperial College, yielded dense, crystalline BN with a crys-
London, UK, and the Institute of Ceramics). tallite size of 20 nm.
 3.15 Polymer Pyrolysis 93

Decomposition of aerosols has been in- and Trefonas, 1989). The synthesis of poly-
vestigated (Moser and Lennhoff, 1989) for carbosilanes has also been reviewed (Sey-
preparation of mixed metal oxides for ferth, 1988). However preceramic polymers
catalytic applications. Examples of com- have been pyrolysed (Burns and Chandra,
positions made for these applications are 1989) to silicon nitride powders. Thus a
CeNiO3, LaCoO 3 and SrFeO3. Spray- polycarbosilane, hydridopolysilazane, me-
roasting, also known as spray-calcination, thylchloropolysilane, and alkylsilsesquia-
involves converting a liquid feed to zane were first cross-linked in argon and
droplets which are fed directly to a fur- fired in an ammonia atmosphere at tem-
nace. It is related to other aerosol tech- peratures up to 1473 K. Amorphous pow-
niques but droplet sizes are larger, often ders were obtained after heart treatment at
greater than 1 ^im and thus beyond the col- this temperature and crystallised when
loidal range. Spray-roasting has been used heated to 1773 K in a nitrogen atmo-
to manufacture multicomponent ceramics, sphere.
for example, ferrites such as MnFe 2 O 4 Boron carbide powder has been made
(Ruthner, 1983). by using polymer pyrolysis (Seyferth et al.,
1989). Lewis bases (:L) react with de-
caborane, B 1 0 H 1 4 to yield L • B 1 0 H 1 2 • L
3.15 Polymer Pyrolysis with loss of one mole of hydrogen but
linear polymers are obtained when the
It was stated in Sec. 3.12 that it can be base has two electron pair donor sites
difficult to distinguish syntheses that in- (:L~L:). For example, polymeric solids
volve liquid-phase reactions from those in- were prepared by reaction of B 1 0 H 1 4
volving polymer pyrolysis. This is because initially with diphosphine bases. A sus-
polymer pyrolysis frequently involves syn- pension of B 1O H 12 -2((C 6 H 5 ) 2 PC1) in
thesis, in a non-aqueous solvent, of a poly- benzene was then reacted in the pres-
meric compound, sometimes referred to as ence of triethylamine to yield the poly-
a preceramic polymer, that is then py- mer, -[B 1 0 H 1 2 - (C 6 H 5 ) 2 POP(C 6 H 5 ) 2 ],T
rolysed to the ceramic. Polymer pyrolysis (molecular weight 27000). This precera-
is particularly associated with the synthesis mic polymer was pyrolysed at 1273 K in
of high-tensile strength |3-SiC fibre and at Ar to a powder containing boron carbide
the present time is not a major synthetic and excess carbon which, after doping with
pathway to powders. boron powder crystallised to B4C at
The pioneering work of Yajima and 1773 K. This reaction pathway to boron
coworkers (1976, 1978) showed how poly- carbide was extended (Seyferth and Rees,
carbosilanes could be derived from polysi- 1991) by reaction of B 1 0 H 1 4 with di-
lanes that were themselves prepared from amines, including H 2 NCH 2 CH 2 NH 2 and
chlorosilanes and then spun into fibres; af- (CH 3 ) 2 NCH 2 CH 2 N (CH 3 ) 2 in diethyl
ter cross-linking, the fibres were pyrolysed ether solution which yielded a solid prece-
to (3-SiC. Synthesis of polysilanes has been ramic polymer; the latter was pyrolysed in
reviewed by West and Maxka (1988) while an ammonia atmosphere at 1273 K to a
a detailed description of the synthesis of a boron nitride containing powder. Polymer
representative polysilane, namely poly- pyrolysis routes to non-oxide ceramics in-
methylphenylsilylene from methylphenyl- cluding powders have been reviewed by
dichlorosilane has been documented (West Pouskouleli (1989) and by Segal (1990).
94 3 Chemical Preparation of Powders

3.16 Emulsion Routes to Powders fected by bubbling NH 3 (g) through the

emulsion. After removal of water and
Emulsions are colloidal systems (see heptane by spray-drying unagglomerated
Sec. 3.5) when the liquid droplet diameter single phase p-alumina particles about
is less than 1 jim. Their use for preparation 100 nm in diameter were obtained on calci-
of ceramic powders is not new and they nation of the gel powder at 1273 K.
were employed for fabrication of micro- Powders have also been derived from
spherical (U,Th)O 2 nuclear fuels (see emulsions containing alkoxides. Thus an
Sec. 3.7.2) by dehydration of aqueous tho- emulsion of hydrolysed Ti(OC 2 H 5 ) 4 in
ria sol droplets in a water-immiscible sol- kerosene was rapidly dehydrated by mi-
vent. Gelation of water-in-oil emulsions crowave radiation (Komarneni and Roy,
derived from aqueous sols has been ex- 1985). After separation of gel spheres from
tended (Woodhead et al., 1984) to complex the organic phase an oxide powder with
ceramic compositions such as mixtures of particles about 30 Jim in diameter was ob-
hollandite, zirconolite and perovskite by tained on calcination. In another approach
use of dehydration techniques as well as by (Sherif, 1990) yttrium and zirconium alk-
use of organic amines and gaseous ammo- oxide solutions in toluene were vigorously
nia. The latter two reagents cause gelation stirred with water. Alkoxide hydrolysis
by removal of anions associated with sol produced stabilised zirconia spheres, after
droplets dispersed in a solvent; after gel collection and calcination of the gelled
calcination powders were around 10 jum in product, around 75 jim in diameter. Mi-
diameter. Electrolyte solutions can be con- croemulsions are thermodynamically sta-
verted to emulsions thus avoiding the need ble dispersions of liquid droplets less than
for sol precursors. As an example, 100 nm diameter in another liquid and
micrometre-sized liquid droplets contain- have been used for producing SiO2 pow-
ing yttrium, barium and copper nitrate so- ders from tetraethoxysilane (Messing and
lutions in heptane were gelled by addition Minehan, 1991). Thus, ammoniated water
of a primary amine (Cima et al., 1988) to was dispersed to a microemulsion in cyclo-
produce the superconducting compound, hexane containing a surfactant then
YBa 2 Cu 3 0 7 _ CT . In another gelation meth- Si(OC 2 H 5 ) 4 was dissolved in the oil phase
od (Akinc and Richardson, 1986), a water- diffused to the droplets where gelation
in-oil emulsion of yttrium nitrate solution produced monodispersed SiO2 particles
in a solvent, for example, toluene was with diameters 50 nm or less.
gelled by addition to a hot liquid which had
the same composition as the continuous
phase of the emulsion. Collection of gelled 3.17 Freeze- and Spray-Drying
particles by filtration yielded unaggregated
Y 2 O 3 powders about 1-2 jam in diameter Freeze-drying is a widely used industrial
at 1123 K. A further example of an emul- process for production of food and phar-
sion-derived powder from electrolyte solu- maceuticals (Boniface, 1989) and its appli-
tions is P-alumina, N a 2 O l l A12O3 (Sang cation to ceramic powders was first de-
and Ogilvie, 1991). An aqueous solution scribed by Schnettler and coworkers
containing aluminium and sodium nitrates (1968). The aim in freeze-drying is to ob-
was dispersed in heptane to droplets of tain chemically homogeneous powders.
about 100 nm diameter. Gelation was ef- This is done by freezing solutions to immo-
 3.20 References 95

bilize ions followed by sublimation of wa- tween powder mixtures and precipitation
ter. When applied to salt solutions, subli- for one-component systems, are described
mation yields an anhydrous salt. Mixed together with their limitations. Chemical
nitrate solutions were freeze-dried (Ander- routes to ceramic powders are discussed
ton and Sale, 1979), the products decom- with particular reference to two areas of
posed at 773 K and then calcined in air at chemistry, colloid science and homoge-
1333 K to strontium-doped lanthanum neous nucleation. These routes are copre-
cobaltite, La 0 3 Sr 0 7 CoO 3 ; the technique cipitation, sol-gel processing of colloids
has been applied to synthesis of supercon- and metalorganic compounds, use of cit-
ducting YBa 2 Cu 3 0 7 _ 6 powder (Medelius rate and acetate gels, the Pechini method,
and Rowcliffe, 1989). In a recent study, hydrothermal synthesis, forced hydrolysis,
alumina powders were derived from freeze- liquid- and gas-phase reactions, use of
dried, calcined aluminium sulfate solution aerosols, polymer pyrolysis and emulsions
at 1073 K (Wang and Lloyd, 1991). Sinter- as well as freeze- and spray-drying. The
ability of the alumina powders was vari- processing advantages of these methods,
able, possible due to the sensitivity of the for example, improved chemical homoge-
freeze-dried powder to calcination. neity and lower reaction temperatures over
Spray-drying is a popular industrial pro- conventional syntheses are illustrated.
cess for converting a liquid feed into a dry
powder by spraying the feed into a hot
drying atmosphere (Lukasiewicz, 1989). 3.19 Acknowledgements
Spray-drying of salt solutions is analogous
to aerosol techniques for powder produc- I thank my employer, the United King-
tion (see Sec. 3.14). The technique has been dom Atomic Energy Authority, for per-
used (Thomson, 1974) for preparation of mission to publish this chapter.
lanthanum-doped lead zirconate titanate
powder which, after hot-pressing, exhibited
improved optical homogeneity over hot- 3.20 References
pressed samples made from mixtures of
component oxides. Superconducting pow- Akinc, M., Richardson, K. (1986), Mat. Res. Soc.
ders with a composition YBa 2 Cu 3 O 7 _ § Symp. Proc. 73, 99.
Anderton, D.J., Sale, F.R. (1979), Powder Met. 1, 8.
have been prepared (Awano etal., 1988) Arcangeli, G., Fava, R., Masci, A., Nardi, A., Vat-
by spray-drying metal nitrate solutions teroni, R., Zondini, C. (1988), Int. J. Mod. Phys. B.
while alcoholic solutions of hydrolysed 2, 443.
Awano, M., Tanigawa, M., Takagi, H., Torii, Y,
alkoxides were spray-dried and calcined to Tsuzuki, A., Murayama, N., Ishii, E. (1988), /. Ce-
oxides (Marsh etal., 1987). Alkoxide- ram. Soc. Jpn. Int. Ed. 96, 417.
derived powders, for example, alumina Baes, Jr., C.F., Mesmer, R.E. (1976), The Hydrolysis
of Cations. New York: John Wiley & Sons.
had a high surface area, up to 900 m 2 g~1 Barb, D., Diamandescu, L,, Rusi, A., Tarabasanu-
with a particle size of several micrometres. Mihaila, D., Morariu, M., Teodorescu, V. (1986), J.
Mater. Sic. 21, 1118.
Barboux, P., Tarascon, J.M., Greene, L.H., Hull,
G. W., Bagley, B.G. (1988), / Appl. Phys. 63, 2725.
3.18 Summary Bauer, R. A., Becht, J. G. M., Kruis, F. E., Scarlett, B.,
Schoonman, J. (1991), J. Am. Ceram. Soc. 74, 2759.
Bednorz, J.G., Miiller, K.A. (1986), Z. Phys. B 64,
Conventional preparation of ceramic 189.
powders, namely solid-state reaction be- Bernier, J.C. (1986), Powd. Metall Int. 18, 164.
96 3 Chemical Preparation of Powders

Bind, J.M., Dupin, T,. Schafer, I, Titeux, M. (1987), ics: Mackenzie, J.D., Ulrich, D.R. (Eds.). New
J. Metals. 54, 60. York: Wiley, pp. 901-906.
Blackburn, S.R., Egerton, T.A., Jones, A.G. (1991), Goodwin, J.W (Ed.) (1982), Colloidal Dispersions.
Br. Ceram. Proc. 47, 87. London: Royal Society of Chemistry.
Blank, D.H.A., Kruidhof, H., Flokstra, J. (1988), /. Haas, P. A. (1989), Chem. Engin. Prog. 85, 44.
Phys. D: Appl. Phys. 226. Haile, S.M., Johnson, Jr. D.W, Wiseman, G.H.,
Bogush, G.H., Dickstein, G.L., Lee, K.C., Zukoski, Bowen, H.K. (1989), J. Am. Ceram. Soc. 72,
C.F. (1988), Mat. Res. Soc. Symp. Proc. 121, 57. 2004.
Boniface, A. (1989), Chem. Eng. 467, 21. Haruta, M., Delmon, B. (1986), J. Chim. Phys. 83,
Bradley, D.C. (1989), Chem. Rev. 89, 1317. 859.
Bradley, D . C , Mehrotra, R.C., Gaur, P.D. (1978), Hennings, D., Rosenstein, G., Schreinemacher, H.
Metal Alkoxides. London: Academic Press. (1991), J. Europ. Ceram. Soc. 8, 107.
Brinker, C.J., Scherer, G. W. (1990), Sol-Gel Science. Interrante, L.V., Carpenter II, L.E., Whitmarsh, C ,
The Physics and Chemistry of Sol-Gel Processing. Lee, W, Garbauskas, M., Slack, G.A. (1986),
New York: Academic Press. Mater. Res. Soc. Symp. Proc. 73, 359.
Burns, G.T., Chandra, G. (1989), /. Am. Ceram. Soc. Ishizawa, H., Sakurai, O., Mizutani, N., Kato, M.
72, 333. (1986), Bull. Am. Ceram. Soc. 65, 1399.
Calcote, H.F., Felder, W, Keil, D.G., Olson, D.B. Jean, J.H. (1990), /. Mater. Sci. 25, 859.
(1990), in: Proc. 23rd Int. Symp. on Combustion. Johnson, C.E., Hickey, D.K., Harris, D.C. (1986),
New Orleans, LA: The Combustion Institute, Mat. Res. Soc. Symp. Proc. 73, 785.
pp. 1739-1744. Kakegawa, K., Mohri, J.I., Shirasaki, S.I., Taka-
Cannon, W.R., Danforth, S.C., Flint, J.H., Hagger- hashi, K. (1984), /. Am. Ceram. Soc. 67, C2-C3.
ty, J. S., Marra, R. A. (1982), /. Am. Ceram. Soc. 65, Kakihana, M., Borjesson, L., Eriksson, S., Svedlindh,
324. P. (1991), J. Appl. Phys. 69, 867.
Chianelli, R. R., Dines, M. B. (1978), Inorg. Chem. 17, Kanno, Y, Pagnoux, C , Piffard, Y, Tournoux, M.
2758. (1991), J. Mater. Res. 6, 2499.
Christian, J. W. (1975), The Theory of Transformations Kayima, P.M., Qutubuddin, S. (1989), J. Mater. Sci.
in Metals and Alloys, 2nd ed. Oxford: Pergamon Lett. 8, 171.
Press. Kepert, D.L. (1972), The Early Transition Metals.
Cima, M.J., Chiu, R., Rhine, WE. (1988), Mat. Res. London: Academic Press.
Soc. Symp. Proc. 99, 241. Klein, L.C. (Ed.) (1988), Sol-Gel Technology for Thin
Colomban, Ph. (1989), Ceram. Inter. 15, 23. Films, Fibers, Preforms, Electronics and Specialty
Crosbie, G.M., Predmesky, R.L., Nicholson, J.M. Shapes. New Jersey: Noyes Publications.
(1990), Report ORNL/Sub/85-SB012ll, Oak Komarneni, S., Roy, R. (1985), Mater. Lett. 3, 165.
Ridge, Tennessee: Oak Ridge National Laborato- Kordas, G., Teepe, M. R. (1990), Appl. Phys. Lett. 57,
ry. 1461.
Defay, R., Prigogine, I., Bellemans, A., Everett, D. H. Kriechbaum, G. W, Kleinschmit, P. (1989), Adv. Ma-
(1966), Surface Tension and Adsorption, London: ter. 2, 330.
Longmans. Kutty, T. R.N., Vivekanadan, R., Murugaraj, P.
Dell, R.M. (1972), in: Proc. 7th Int. Symp. on the (1988), Mater. Chem. Phys. 19, 533.
Reactivity of Solids: Anderson, J. S. (Ed.). London: LaMer, V.K., Dinegar, R.H. (1950), / Am. Chem.
Chapman and Hall, pp. 553-566. Soc. 72, 4847.
Denkewicz, Jr., R. P., TenHuisen, K.S., Adair, J. H. Lane, E.S., Rush, D.F., Segal, D.L. (1990), U.S.
(1990), J. Mater. Res. 5, 2698. Patent 4946 820.
Dunning, W.J. (1961), in: Powders in Industry, SCI Lang, J. (1962), J. Acoust. Soc. Am. 34, 6.
Monograph No. 14. London: Society of Chemical Lessing, P. A. (1989), Bull. Am. Ceram. Soc. 68, 1002.
Industry, pp. 29-43. Ley, K., Kontoulis, I., Steele, B.C.H. (1991), Br.
Eror, N.G., Anderson, H.U. (1986), Mat. Res. Soc. Ceram. Proc. 47, 119.
Symp. Proc. 73, 571. Lindquist, D. A.,, Kodas, T. T, Smith, D. M., Xiu, X.,
Einarsrud, M. A., Rhine, W. E., Cima, M. J. (1989), in: Hietala, S.L., Paine, R.T. (1991), J. Am. Ceram.
Euroceramics, Vol. 1: de With G., Terpstra, R. A., Soc. 74, 3126.
Metselaar, R. (Eds.). London: Elsevier Applied Livage, I, Henry, M., Sanchez, C. (1988), Prog. Solid
Science, pp. 1.38-1.42. State Chem. 18, 259.
Falter, L.M., Payne, D.A., Friedmann, T. A., Livage, J., Henry, M., Jolivet, IP., Sanchez, C.
Wright, W.H., Ginsberg, D.M. (1989), Proc. Br. (1990), MRS Bulletin. 15, 18.
Ceram. Soc. 41, 261. Lukasiewicz, S. J. (1989), / Am. Ceram. Soc. 72, 617.
Fukai, K., Hidaka, K., Aoki, M., Abe, K. (1990), Lyklema, J. (1991), Fundamentals of Interface and
Ceram. Inter. 16, 285. Colloid Science, Vol. 1, London: Academic Press.
Gallagher, M.K., Rhine, WE., Bowen, H.K. (1988), Ma, J., Cheng, H., Zhao, Z., Qiang, D. (1991), Chem.
in: Ultrastructure Processing of Advanced Ceram- Mater. 3, 1006.
 3.20 References 97

Marcilly, C , Courty, P., Delmon, B. (1970), /. Am. Sang, XV., Ogilvie, C.X (1991), U.S. Patent 4 999 323.
Ceram. Soc. 53, 56. Schnettler, F.X, Monforte, F.R., Rhodes, W W (1968),
Marsh, G.B., Fanelli, A.J., Armor, J.N., Zambri, in: Science of Ceramics, Vol. 4: Stewart, G. H. (Ed.).
P.M. (1987), U.S. Patent 4649037. Stoke-on-Trent: British Ceramic Society, pp. 79-90.
Matijevic, E. (1985), Annu. Rev. Mater. Sci. 15, Scott, K.T., Woodhead, XL. (1982), Thin Solid Films,
483. 95, 219.
Matijevic, E. (1988), in: Ultrastructure Processing of Seddon, A. B., Hodgson, S. N. B., Scott, M. G. (1991),
Advanced Ceramics: Mackenzie, J. D., Ulrich, D. R. J. Mater. Sci. 26, 2599.
(Eds.). New York: Wiley, pp. 429-442. Segal, D. (1989), Chemical Synthesis of Advanced Ce-
Maya, L. (1986), Adv. Ceram. Mater. 1, 150. ramic Materials. Cambridge: Cambridge Universi-
Mazdiyasni, K.S., Cooke, C M . (1976), U.S. Patent ty Press.
3 959 446. Segal, D.L. (1990), in: The Physics and Chemistry of
McMillan, P.W. (1964), Glass Ceramics. London: Carbides, Nitrides and Borides: Freer, R. (Ed.). The
Academic Press. Hague: Kluwer Academic, pp. 3-11.
Medelius, H., Rowcliffe, D.J. (1989), Mater. Sci. Eng. Segal, D. L., Woodhead, X L. (1986), Proc. Br. Ceram.
A109, 289. Soc. 38, 245.
Melling, P. J. (1984), Bull. Am. Ceram. Soc. 63, 1427. Seibold, M., Riissel, C. (1988), Mater. Res. Soc.
Messing, G.L., Minehan, B. (1991), /. Ceram. Soc. Symp. Proc. 121, All.
Jpn., Int. Ed. 99, 1000. Seyferth, D. (1988), Am. Chem. Soc. Symp. Ser. 360,
Milne, S.J. (1986), Br. Ceram. Proc. 38, 81. 21.
Morey, G.W. (1953), /. Am. Ceram. Soc. 36, 279. Seyferth, D., Rees, Jr., WS. (1991), Chem. Mater. 3,
Moser, W.R., Lennhoff, I D . (1989), Chem. Eng. 1106.
Comm. 83, 241. Seyferth, D., Rees, Jr., W.S., Haggerty, J.S., Light-
Nielsen, A.E. (1964), Kinetics of Precipitation. Ox- foot, A. (1989), Chem. Mater. 1, 45.
ford: Pergamon Press. Shaw, W.H.R., Bordeaux, XX (1955), /. Am. Chem.
Nielsen, A. E. (1967), in: Crystal Growth: Peiser, H. S. Soc. 77, 4729.
(Ed.). Oxford: Pergamon Press, pp. 419-426. Sherif, KG. (1989), U.S. Patent 4 804649.
Obvintseva, I.E., Yanovskaya, M.I., Kucheiko, S.I., Sherif, F. G. (1990), European Patent Appl. 0 362 962.
Turova, N. Ya., Shifrina, R. R., Kukuev, V. I., Lub- Slamovich, E.B., Lange, F.F. (1988), Mater. Res.
nin, E. N., Venevtsvev, Yu. N. (1988), Inorg. Mater. Soc. Symp. Proc. 121, 257.
25, 661. Somiya, S. (1991), in: Concise Encyclopaedia of Ad-
Ochi, A., Bowen, H.K., Rhine, WE. (1988), Mater. vanced Ceramic Materials: Brook, R.X (Ed.). Ox-
Res. Soc. Symp. Proc. 121, 663. ford: Pergamon Press, pp. 375-377.
Ogihara, T., Mizutani, N., Kato, M. (1987), Ceram. Sordelet, D.X, Akinc, M. (1988 a), J. Am. Ceram. Soc.
Inter. 13, 35. 71, 1148.
Okuyama, K., Kousaka, Y, Tohge, N., Yamamoto, Sordelet, D.X, Akinc, M. (1988 b), /. Colloid Interface
S., Jwang Wu, X, Flagan, R.C., Seinfeld, J.H. Sci. 122, 47.
(1986), AIChE J. 32, 2010. Southern, XC. (1991), Br. Ceram. Proc. 47, 1.
O'Neill, J.A., Horsburgh, M., Tann, X, Grant, K.X, Stookey, S.D., Maurer, R.D. (1962), in: Progress in
Paul, G.L., Sinclair, W (1989), J. Am. Ceram. Soc. Ceramic Science, Vol. 2: Burke, X E. (Ed.). New
72, 1130. York: Pergamon Press, pp. 77-101.
Parr, R.G., Pearson, R.G. (1983), J. Am. Chem. Soc. Sugimoto, T. (1987), Adv. Coll. Inter. Sci. 28, 65.
105, 7512. Suyama, Y, Ho, K., Kato, A. (1975), J. Inorg. Nucl.
Pechini, M.P. (1967), U.S. Patent 3 330 697. Chem. 37, 1883.
Pouskouleli, G. (1989), Ceram. Int. 15, 213. Tarascon, J.M., Barboux, P.B., Bagley, B.G.,
Pouskouleli, G., Wheat, T.A. (1990), Can. Ceram. Greene, L.H., Hull, G.W. (1988), Mater. Sci. and
Quart. (May) 25. Engin. Bl 29.
Rabinovich, E.M. (1988), in: Sol-Gel Technology for Thomson, Jr., X (1974), Bull. Am. Ceram. Soc. 53,
Thin Films, Fibers, Preforms, Electronics and Spe- 421.
cialty Shapes: Klein, L.C. (Ed.). Park Ridge, NJ: Turnbull, D., Fisher, XC. (1949), /. Chem. Phys. 17,
Noyes Publ., pp. 260-294. 71.
Rajendran, S. (1992), /. Mater. Sci. 27, 433. Verwey, E.XW, Overbeek, J.Th.G. (1948), Theory of
Rayleigh, X WS. (1878), Proc. London Math. Soc. 10, the Stability of Lyphobic Colloids. Amsterdam: El-
1. sevier.
Riedel, R., Gaudl, K.U. (1991), J. Am. Ceram. Soc. Wang, L., Lloyd, I.K. (1991), J. Am. Ceram. Soc. 74,
74, 1331. 2934.
Riman, R.E., Landham, R.R., Bowen, H.K. (1989), Wang, L.H., Hon, M.H., Huang, W.L., Lin, W.Y
/. Am. Ceram. Soc. 72, 821. (1990), Mater. Sci. Eng. B7, 237.
Ruthner, M.X (1983), Mater. Sci. Monog. 16, 515. Watanabe, Y, Suzuki, K., Teranishi, M. (1988), U.S.
Sakka, S., Yoko, T. (1991), Ceram. Inter. 17, 217. Patent 4770 812.
98 3 Chemical Preparation of Powders

West, R., Maxka (1988), 1 Am. Chem., Soc. Symp. General Reading
Ser. 360, 6.
West, R., Trefonas, P. (1989), in: Inorganic Syntheses,
Vol. 25: Allcock, H. R. (Ed.). New York: Wiley In- Brinker, C.X, Scherer, G. W. (1990), Sol-Gel Science.
terscience, pp. 56-60. The Physics and Chemistry of Sol-Gel Processing.
Woodhead, XL. (1974), U.K. Patent 1 342 893. New York: Academic Press.
Woodhead, XL. (1984), /. Mat. Educ. 6, 887. Cheetham, A.K., Brinker, C.X, Mecartney, M.L.,
Woodhead, XL., Cole, K., Dalton, XT., Evans, J.P., Sanchez, C. (Eds.) (1994), Better Ceramics Through
Paige, E. L. (1984), in: Science of Ceramics, Vol. 12: Chemistry VI. Pittsburgh, PA: Materials Research
Vincenzini, P. (Ed.). Faenza: Ceramurgica, Society.
pp. 179-185. Christian, X W (1975), The Theory of Transformations
Yajima, S., Omori, M., Hayashi, X, Okamuya, K., in Metals and Alloys, 2nd Ed. Oxford: Pergamon.
Matsuzawa, T., Liaw, C. (1976), Chem. Lett. 551. Everett, D.H. (1988), Basic Principles of Colloid Sci-
Yajima, S., Hasegawa, Y, Hayashi, X, Imura, M. ence. London: Royal Society of Chemistry.
(1978), J. Mater. Sci. 13, 2569. Metlin, Y.G., Tretyakov, YD. (1994), /. Mater.
Yamanobe, Y, Yamaguchi, K., Matsumoto, K., Fu- Chem. 4, 1659.
jii, T. (1991), Jpn. J. Appl. Phys. 30, 478. Riman, R. E. (1992), in: Colloid and Surface Engineer-
ing: Applications in the Process Industries: Williams,
R.A. (Ed.). Oxford: Butterworth, pp. 140-167.
Segal, D. (1989), Chemical Synthesis of Advanced Ce-
ramic Materials. Cambridge: Cambridge Universi-
ty Press.
Uhlmann, R., Ulrich, D.R. (Eds.) (1992), Ultrastruc-
ture Processing of Advanced Materials. New York:
Wiley.
4 Characterization of Particles and Powders
Brian Scarlett

Delft University of Technology, Delft, The Netherlands

List of Symbols 100

4.1 Powder Characterization 101
4.2 Why Characterize and Why the Particles? 101
4.3 Particle Size Distribution - What is Measured 103
4.3.1 Particle Size 103
4.3.2 Particle Size Distribution 105
4.4 Particle Size Distribution Measurement 108
4.4.1 Electrical Sensing Zone (Coulter Counter) 108
4.4.2 Particle Sedimentation 109
4.4.3 Sieving Ill
4.4.4 Calibration Ill
4.5 On-Line Particle Measurement 112
4.5.1 Light Scattering Techniques 114
4.5.2 Acoustic Techniques 118
4.6 Statistical Diameters 118
4.6.1 Feret Diameters 120
4.6.2 Fourier Analysis 120
4.6.3 Fractal Dimension 120
4.6.4 Chord Size Distribution 121
4.6.5 Measuring and Modeling 122
4.7 Powder Properties 122
4.8 References 124

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. Allrightsreserved.
100 4 Characterization of Particles and Powders

List of Symbols
A area
C mean chord
D fractal dimension; drag
F Feret diameter
9 acceleration due to gravity
»||.*X intensity of light polarized parallel, perpendicular to incident plane
; intensity of scattered light
Jo intensity of incident light
/ length
m complex refractive index relative to medium
M moment
P perimeter
Px perimeter of polygon of step length X
1 distribution function
qo(x) fractional number density distribution
«3W fractional volume distribution
Q cumulative distribution function
Qo(x) cumulative fractional number undersize
Qs(x) cumulative fractional weight undersize
r radius
R distance from scatterer to receiver
Re Reynolds number
s standard deviation
S surface area
V velocity
V settling velocity; volume
X particle diameter
equivalent mesh diameter
equivalent surface diameter
equivalent volume diameter
equivalent settling diameter
Z distributed function
rj viscosity
e angle
X wavelength; step length
density, of fluid, of solid
sphericity
 4.2 Why Characterize and Why the Particles? 101

4.1 Powder Characterization their relationship to the microscopic prop-

erties of the powders, in particular their
The task of process engineers is to con- particle size distribution. The behavior of a
trol and manipulate three classes of vari- material is dependent upon the size and
ables: equipment, operating, and material shape characteristics of the particles which
variables. This chapter concerns that third constitute that material. Thus all the bulk
class: the material variables. There is no properties of the material, suspension, and
doubt that the properties of a ceramic compact reflect, in some way, an average
product depend upon those of the raw ma- of the properties of the particle population
terials from which they are formed, and which must be rigorously controlled.
which are usually in powder form. Thus it There is another reason why it is basical-
is obviously essential to control these ly more convenient to control the micro-
properties, but it is not quite so obvious scopic properties in the first instance. Mix-
what to control, and how and where to do ing of the ceramic materials is not really
it. These questions are open-ended and the the start of the process. These powders
role of the technologist is to provide and have already been prepared by operations
adapt answers to suit the circumstances. such as grinding, spraying, crystallization,
Thus the first question to be answered, is or vapor phase condensation. The act of
'Why characterize the powder?' The whole preparing the powders by these methods
chapter is based upon a sequence of deci- and then reconstituting them into a ceram-
sion making: Why? What? How? Where? ic product is all part of the total process in
which the basic material undergoes some
change at each stage. In the early stages of
4.2 Why Characterize and Why preparing the powders it is only the micro-
the Particles? scopic properties that can be related to the
process variables, for example, the grind-
In ceramic processing the starting mate- ing mechanisms to the particle size
rials are usually powders which are later achieved. Completion of the powder prep-
mixed with binder or are suspended in liq- aration is sometimes seen as a natural
uid to form a slip. The mixture is formed, break in the total process at which the
compacted, and sintered until the final sol- properties of the particles can be con-
id ceramic body is produced. In the first trolled. So long as ceramic processing is
instance, the chemical nature and purity of predominantly a batch operation there is
the materials must be correct. Thereafter, some logic in this, but the new nanotech-
numerous bulk properties of the mixture nologies may require continuous process-
are recognized as important during the ing and thus more monitoring throughout
process. The packing density of the pow- the whole sequence.
ders, their flow properties, conductivities, Meanwhile, control of the particulate
and compactability all affect the final properties has a twofold purpose. For the
strength and finish of the product. When preparation of the materials it is the speci-
one of the bulk properties of the powder is fication. Thus the growing, grinding, and
changed, then inevitably all the others classification of the materials must be con-
change. Although all these macroscopic trolled in such a manner as to produce the
properties must be measured, they can particle specification required. This speci-
only be manipulated by understanding fication, on the other hand, is related to the
102 4 Characterization of Particles and Powders

behavior of the materials in the subsequent which, whether empirical or theoretical,

processing. Powder-forming processes are quantitative or qualitative, relates the two.
controlled to produce a powder with a Since the bulk properties to be con-
specified particle distribution on the as- trolled are numerous and complex, it is
sumption that this powder will then lead to unlikely that a single, simple particle char-
the required product in the subsequent acterization technique will be appropriate
processing. In principle the sophistication for them all. On the other hand, once the
that could be applied to characterizing the specification is established, a single mea-
particles is almost unlimited, but the level surement of the particle size distribution
which is needed is defined entirely by the may be sufficient to ensure reproducibility.
final application. To use a very sophisticat- Similarly, in order to model the different
ed technique for a simple problem means particle preparation techniques, sophisti-
that much of the information will be dis- cated measurements may be required. Lat-
carded. On the other hand, to use a very er, routine control of the process may be
simple measurement to correlate complex achieved with quite simple measurements.
behavior ensures failure. The measure- Thus the particle size distribution is the
ment must be matched to the problem. In basic parameter that may be used for both
turn, the measuring technique must also quality control and process control. For
attain the reproducibility, accuracy, and more physically realistic models, other
sensitivity which is needed for the particu- parameters may be necessary. The ques-
lar problem. tion of what to measure and how to mea-
At the simplest level, particle characteri- sure it may be considered against these two
zation is used for quality control. Thus objectives, establishing a process and then
there is a specification, gained by experi- operating it routinely. A particle size dis-
ence, against which the powders must be tribution measurement may be sufficient
prepared and which must be adjusted be- to ensure the reproducibility of a powder
fore the ceramic processing continues. The preparation process that makes only one
main requirement of the particle character- product. As more versatility is required,
ization technique is that it be sensitive to then the characterization becomes more
any changes in the material which will extensive. In developing completely new
have a discernable effect on the final prod- products, a specification must be estab-
uct. The reproducibility of the technique lished which will predict the mixture of
must be an order of magnitude better than particle species, shapes, and sizes that will
the necessary sensitivity so that instrument give the required properties. Such a prob-
errors do not confuse the control. Abso- lem requires more sophisticated character-
lute accuracy is sometimes argued to be ization of both the particles and the pow-
less important. However, as scientific so- ders. Once established, a more restricted
phistication increases then accuracy be- measurement may be sufficient to ensure
comes more important, meaning that the reproducibility.
measurement reflects the physical reality.
There is then some mechanistic interpreta-
tion of the relationship of particle charac-
teristics and their bulk behavior. This
requires a correlation, theory, or model
 4.3 Particle Size Distribution - What is Measured 103

4.3 Particle Size Distribution - This classical approach to particle size

What is Measured measurement, the equivalent sphere ap-
proach, may be sufficient to control the
4.3.1 Particle Size manufacture and use of a known powder
once the process is established. Clearly
Given a spherical particle and asked to
much information about the size and
determine the size of that particle, it would
shape of each particle has already been dis-
seem quite natural to measure its diameter.
carded and so the characterization has in-
Confronted by a particle of any other
herent limitations. The first is that the size
shape, it is reasonable to rationalize the
of a particle is not unique, many different
dilemma by defining that particle to be
equivalents can be chosen, of which only a
equivalent to a spherical particle with re-
few are shown in Fig. 4-1. Secondly, each
spect to some property. The size of the
particle is characterized by only one scalar
particle is then the diameter of that sphere.
number, a diameter with the dimensions of
The most obvious equivalents are geomet-
length, and thus it can convey no informa-
rical, for example, the equivalent volume
tion about the shape of the particle. Fur-
diameter or the equivalent surface diame-
thermore it is clear that an equivalent di-
ter. Other equivalents may be behavioral,
ameter is not unique to a particular parti-
for example, a sphere of the same density
cle; many particles may have the same
which settles with the same velocity defines
equivalent diameter. For example, any
the equivalent settling diameter (see Fig.
particle may be deformed to produce a
4-1). Classically, particle size measurement
particle of different shape which has the
is based upon this approach: to define the
same equivalent volume diameter. Thus
size of the particle as the diameter of a
for all these reasons, although the equiva-
sphere which is equivalent in one aspect of
lent diameter of a particle is dependent
behavior.
upon its shape, a single particle size distri-
The ISO standard (ISO, 1995) recom- bution based upon one equivalent diame-
mends that the symbol x be used for parti- ter cannot convey information about the
cle size and that a subscript denote the particle shape. Establishing the process in
equivalent diameter used. For example: the first place may require more extensive
xy: equivalent volume diameter measurements.
xs: equivalent surface diameter This approach is, however, more elegant
xw: equivalent settling diameter than is apparent at first sight. It establishes
xn: equivalent mesh diameter a linear scale of measurement of particle

Figure 4-1. Some common examples of the equivalent sphere concept: 1, sphere of equivalent surface (xs);
2, sphere of equivalent volume (xv); 3, sphere of equivalent settling velocity, low Reynolds number (xw); 4, sphere
of equivalent settling velocity, high Reynolds number (xw); 5, sphere of equivalent sieve mesh (xn).
104 4 Characterization of Particles and Powders

\j/ = sphericity (4-1)

surface area of sphere of same volume
O surface area of particle
In fact this parameter is simply a ratio of
C) the equivalent volume and surface diame-
ters to the second power

nx (4-2)

a o
nx:
This ratio of equivalent diameters is di-
o o o
mensionless and is a shape factor which
has a value of less than one for any particle
equivalent equivalent equivalent that is not spherical. Two particles that
surface volume sieve mesh have the same shape but different size
(*») (*v) W would have the same Waddell sphericity
Figure 4-2. A linear scale of measurement. These factor, although this value might not be
equivalent spheres changes linearly with particle size completely unique to that particle shape.
for the same shape.
Of course the linearity of the size scale is
only assured so long as the law of equiva-
lence is valid. Thus, if an irregular particle
increases all of its dimensions by a factor
of two, then the equivalent surface and
size. When the linear dimensions of a par- volume diameters also change by a factor
ticle all change by some factor, without a of two. For these equivalents the relation-
change of shape, then all the equivalent ship between the size of a particle and its
diameters change by the same factor. For equivalent diameter is always linear. This
example, if the dimensions of a particle is not true of all the behavioral equiva-
double, then its surface is increased four- lents. The equivalent settling velocity of a
fold and its volume eightfold. However, small spherical particle of diameter x may
the equivalent surface diameter and equiv- be calculated by Stokes's law (Stokes,
alent volume diameter both double. 1845): V=[x2{Qs-Qi)g}l{\%n\ where Qs
This linear scale is illustrated in Fig. 4-2, and Q{ are the densities of solid and fluid,
showing that the ratio of the equivalent rj is the viscosity and Kis the settling veloc-
diameters is always the same for particles ity. As the particle becomes larger, or even
of the same shape. Thus, some measure of if a less viscous fluid is used with the same
particle shape can be achieved by deter- particle, the settling regime may change
mining the ratio of any two equivalent di- and Stokes's law is no longer applicable.
ameters. Such a parameter is dimension- Thus at low Reynolds number the settling
less and is dependent only on the shape of velocity is proportional to the square of
a particle, not upon its size. Although such the particle size and at high Reynolds
shape factors are not completely unique, number to the square root. (The Reynolds
they are useful. The most common exam- number Re = Qi Vx/rj characterizes the
ple is the Waddell sphericity factor (Wad- flow of the fluid around the settling parti-
dell, 1932, 1933). This is defined as cle.) The equivalent size of a particle is
 4.3 Particle Size Distribution - What is Measured 105

only linear with its dimensions so long as (a)

the law which defines the behavior of that
equivalent sphere is applicable.
4.3.2 Particle Size Distribution
When the size of an individual particle
has been defined and determined, then the
next problem is how to express the distri-
bution of particle sizes within a powder inx
sample. The distribution is usually deter-
mined either as a number distribution or as (b)
a distribution by weight, although this (i)
need not be exclusively so. The informa-
tion may be displayed graphically if so de-
sired, either as a histogram or as a contin-
uous distribution curve. A cumulative dis-
tribution curve is the most common be-
»-
cause the curves can be more easily inter-
In*
polated and normalized in this form. The
curves may be chosen to be either under- (ii)
Q~
size or oversize curves, and this is usually
dependent upon whether the maximum
and minimum sizes of the distribution are .A*--max. slope

known. The scale of the particle size pa- / \

In *
rameter is often chosen to be logarithmic
(iii)
in order to give appropriate significance to
the smaller particle sizes. These various \, 1 max. slope
graphical representations are illustrated in
Figs. 4-3 a and b.
The ISO standard (ISO, 1995) recom-
mends the symbol q for the distribution Inx
function and Q for the cumulative distri-
bution function. A subscript denotes the
(c)
power of particle size involved in the distri-
bution function. Thus
qo(x) = fractional number density
distribution
q3(x) = fractional volume distribution
[also fractional weight if density
is constant] In*

and Figure 4-3. Graphical representations of a particle

size distribution as (a) a histogram and (b) a continu-
Goto = cumulative fractional number ous distribution (i) or a cumulative distribution (ii,
undersize cumulative undersize; iii, cumulative oversize), (c)
Gato cumulative fractional volume
= Graphical representation of a bimodal particle size
undersize distribution.
106 4 Characterization of Particles and Powders

Almost all modern instruments present Any mean of the distribution can thus
the results of an analysis in graphical form be defined by two moments. For example
but it is, of course, also necessary to ex-
M 13
press the form of the curve quantitatively = average size of particles
so that the measurements can be used in a M 03 by volume (4-3)
theory or correlation. Some of the simplest
correlations reduce the distribution to a On the other hand
single parameter: a mean, median, or
modal size. These sizes all convey in one • M 30

single number the order of particle size in- = average volume of particles (4-4)
volved while giving no indication of the M oo
spread. All depend upon the distribution This elegant nomenclature is not often
function chosen: number, mass, or other. used in practice. It is most useful, and is
The mean size is the average of the distri- mentioned here, to illustrate that a dis-
bution. The median size is the midpoint, tribution can have an infinite number of
often denoted by x50. At this point half of 'average' sizes.
the particles are smaller and half are An alternative to single parameter mod-
larger. The modal size is the size which els is to fit the distribution curve with an
occurs most frequently. It is thus the loca- analytical function. Such functions are
tion of the maximum in the distribution usually two parameter models. These func-
function. None of these single value tions have the advantage that they can be
parameters are likely to be sufficient in ce- incorporated into the differential equa-
ramic processing. The packing and sinter- tions which sometimes describe a process.
ing behavior depends critically on the size The most commonly used functions are the
distribution. Often powders which have Gaudin-Schuman, the log-normal, and the
more than one modal value are utilized Rosin-Rammler distributions. In each case
(Fig. 4-3 c). the equation contains two constants, one
An extension of the use of a mean is to of which is a characteristic size and one of
use all the possible moments of the distri- which is dependent upon the width of the
bution. The German DIN standard (DIN, distribution. Although these functions are
1981) recommends that the moment of a sometimes convenient to use, they should
distribution be represented by the letter M not be extrapolated outside the range of
qualified by two subscripts. The first sub- the actual measurements:
script denotes the power of the moment, (i) Gaudin-Schuman distribution
the second the power of the distribution
function. For example: Q(x) = (4-5)
M13 = 1st moment by weight = § xq3dx
M23 = 2nd moment by weight = J x2 q3 dx This expression is included because it is
an example of a simple power law fit. Plot-
M33 = 3rd moment by weight = §x3q3dx
ted on a logarithmic scale, the distribution
M10 = 1st moment by number = §xqodx is a straight line. The characteristic size is
M20 = 2nd moment by number = j x2 q0 dx that of the largest particle, x max . A wider
M30 — 3rd moment by number = f x3 q0 dx distribution has a smaller slope, m.
 4.3 Particle Size Distribution - What is Measured 107

(ii) Rosin-Rammler distribution analytically differentiated and integrated

and fitting procedures can be used to
(4-6) smooth out the noise in the measurements.
This, however, becomes a disadvantage
This distribution is based on an expo- when the deviations from the function are
nential function and thus tends to be appli- not noise, but are real deviations. It is an
cable in a process where many random even bigger disadvantage when the devia-
events occur consecutively. It is often used tions are intentional. In modern processing
to describe the products of comminution the required particle mixtures may be quite
processes. The function can be plotted as a elaborate, far more complex than can be
straight line on a lnln vs. In scale. The char- described by a two-parameter model.
acteristic size, x\ is the size corresponding Modern computing techniques enable
to a cumulative percentage oversize of the numerical solution of equations and
0.368. A wider distribution has a smaller are easily able to accept the distribution
value of n. function as a series of numbers. Thus al-
(iii) Log-normal distribution: The nor- though a graph of the distribution curve is
mal distribution is described by the follow- convenient for visualization, its manipula-
ing equation: tion is nowadays best achieved numerical-
ly. There are two basic choices. The first is
q(z) = (4-7) to take fixed points on the distribution axis
and to read off the values of particle size at
those points. The median is one of these
This is the Gaussian normal distribution 'ogives' but often quartiles or 10% values
which describes a completely random dis- are also used. The second technique is to
tribution of sizes around some mean. This take fixed size classes and to determine the
curve can be useful for some narrow, natu- quantities in each size class. Thus, the form
rally occurring distributions such as those of a histogram is expressed as a column
relating to blood cells, pollen, or yeast vector, each term being the fraction of
cells. Usually, the size distributions of particles contained within a size range
powders are too wide for this function and (Fig. 4-4).
the log-normal distribution must be used. The size classes may be arbitrary and
The distributed function, z, is then given chosen to fit the problem in hand. The
by
In x — In x50
z = (4-8)
Ins
where s is the standard deviation of the
distribution.
This is the normally distributed form of
the ratio of the sizes. If this distribution is
applicable, a plot of probability versus In
size will yield a straight line.
The advantage of using these curve fits is
that the whole distribution is described by Figure 4-4. A particle size distribution represented as
only two parameters. The functions can be a vector, i.e., a column matrix.
108 4 Characterization of Particles and Powders

matrix may be normalized to express frac- In principle, any phenomenon which is

tions or percentages or may, alternatively, dependent on the particle size can be used
represent the actual quantities of powder to measure that size, and almost every-
at different stages of the process. The ad- thing has been tried at some time or other.
vantage of this representation is that in In practice, it is more convenient if the
each size class the terms in the matrix rep- relationship between the mechanism and
resent the actual particles present, whereas the particle size is known in the form of a
the values of all the ogives change when law, and if that relationship is monotonic.
one part of the distribution is changed. Thus there is currently a great diversity of
This method of representing particle size equipment commercially available, but
distributions is usually the most conven- this is based on relatively few physical
ient for use with numerical techniques, principles. These techniques are well sur-
that is, computer techniques. Further- veyed and discussed in a book by Allen
more, any part of the process which (Allen, 1990). Great reproducibility can be
changes the particle size distribution can obtained from any one technique if a stan-
be expressed as a second order matrix dard procedure is followed, for example as
described in a British Standard (British
Pll P12 Pl3 Pl4 Standard, 1995). The methods that are
Pll P22 P23 P24 42 most useful for characterizing ceramic
(4-9)
Pzi P32 P33 P34 powders at present are
?41 P42 P43 P44
- sieving,
or - sedimentation,
- electrical sensing zone,
Pq = r (4-10) - light scattering,
The process P changes the particle size dis- - ultrasonic attenuation.
tribution q into r. The first three are established tech-
niques which are well tried and developed
and from which good results can be ob-
4.4 Particle Size Distribution tained but they are less amenable to auto-
Measurement mation and on-line application. They will
be discussed briefly in the following three
In considering particle-measuring tech- sections; the other two techniques are de-
niques, the first major distinction to be ferred to later sections. The major purpose
made is between direct and indirect meth- of this discussion is to illustrate the fact
ods. If the particles are directly observed, that good particle characterization does
by microscope or electron microscope, not depend upon one technique.
then an indefinite amount of information
can be obtained from those pictures. The
only limitation is the number of particles 4.4.1 Electrical Sensing Zone
that can be analyzed in a reasonable time. (Coulter Counter)
These techniques are therefore usually The electrical sensing zone is an estab-
used for more sophisticated analyses. A lished and accurate instrument. In this
simple particle size distribution is more device the particles are suspended in an
often determined by an indirect method. electrolyte which is sucked through a small
 4.4 Particle Size Distribution Measurement 109

orifice of accurate dimensions. On either

side of the orifice is an electrode between
which a current is passed. Thus a large
electrical field is created in the orifice.
When a particle passes through the orifice
it distorts the electrical field and a voltage
pulse is produced (Fig. 4-5).
The magnitude of the pulse is propor-
tional to the volume of the particle if cer-
tain conditions are fulfilled: Figure 4-5. Schematic diagram of the Coulter coun-
ter.
(i) The concentration is low so that par-
ticles appear singly in the detection zone:
(ii) The particles pass through a region
'traceability' of an instrument is important
of homogeneous field; thus they pass
when it is used to measure standard refer-
through the center and are not bigger than
ence materials. An alternative calibration
half the diameter of the orifice.
technique is to pass a known mass of par-
(iii) The particles have a large effective
ticles through the instrument. If their den-
resistivity compared to the electrolyte. In
sity is known their total volume can be
practice this is usually the case, even for
compared with the integrated instrument
conducting particles, due to the surface po-
response. This total mass calibration is
tential of the particle.
traceable to the standard kilogram. This
The electrical sensing zone technique is instrument is invaluable when accurate
particularly useful if accurate reference re- and traceable results are required.
sults are required, for a number of reasons. The disadvantage of the instrument is
The instrument basically produces a num- that the size range of any one analysis is
ber distribution of the equivalent volume only of the order of 10:1 for accurate re-
diameter. If the pulses are integrated it sults. Of course, orifices of different sizes
then records a volume distribution of the are available, but a sample of wide distri-
volume diameter. It is one of the few in- bution cannot be accurately analyzed with
struments which measures the equivalent one orifice.
volume diameter directly, both as a num-
ber and a mass distribution.
The instrument can be calibrated in two 4.4.2 Particle Sedimentation
different ways: Monosized spherical parti- The equivalent settling diameter of a
cles can be used, whose size has been previ- particle is dependent upon the Reynolds
ously accurately determined by micro- number which appertains during the set-
scope techniques. By observing the re- tling. It is therefore customary to confine
sponse of the instrument to these particles, the analysis to the low Reynolds number
calibration can be achieved. If several regime (less than 0.2) where Stokes's law
monosized samples are used, the linearity applies (see Sec. 4.3.1). The settling veloc-
of the instrument can be established. Since ity of the particle is then given by
the size of the particles has been estab-
lished by direct observation, this calibra- (4-11)
tion is traceable to the standard meter. The v=
110 4 Characterization of Particles and Powders

Data reporting function

Relative concentration data

.2 o100 10 1.0 OX
Pore diameter

Elapsed time to Cell position Converter (X-ray

particle diameter translator intensity to
conversion vs. time concentration)

Outlet

X-ray tube s.n X-ray

detector

Slit

Density and
Control
viscosity
function
information
Dispersed
sample or
pure liquid

Cell position signal Figure 4-6. Schematic diagram of

the X-ray gravitational sedimen-
Elapsed time, densities and viscosity data
tometer.

where QS and gf are the densities of the solid For particles of modest aspect ratio, the
and the fluid, respectively, g is the acceler- Stokes diameter is only a little smaller than
ation due to gravity and rj is the viscosity the equivalent volume diameter. The ad-
of the fluid. vantage of the sedimentation method is
It is also customary to confine the con- that it can analyze samples with a wider
centration to minimize the interaction of range of particle sizes, of a ratio of at least
the particles with each other. The volume 100:1. The disadvantage is that the analy-
concentration should preferably be less sis may take a long time.
that 0.005 v/v, which corresponds to an For ceramic processing a particularly
average spacing of 10 particle diameters. useful instrument is a centrifugal sedimen-
The suspension is more stable when it is tation analyzer of which several versions
initially uniformly mixed, a so-called ho- are commercially available. The settling
mogeneous start. The particle size distribu- vessel, either a disk or a cuvette, is rotated
tion is deduced by recording the concen- and the consequential centrifugal force in-
tration as a function of time. One modern creases the rate of settling. Typical values
instrument scans an X-ray beam over the of acceleration of a few hundred times that
height of the settling vessel and then de- of gravity are employed which enable
convolutes a mass distribution of equiva- small particles, certainly as small as
lent Stokes diameter (Fig. 4-6). 0.1 jim, to be analyzed in a reasonable
 4.4 Particle Size Distribution Measurement 111

time. The concentration profile is also

recorded by a light beam or an X-ray
beam, although sometimes a 'line-start'
technique is used in order to improve the
resolution of multi-modal distributions.
With this technique a thin layer of the par-
ticles is created at the surface of the clear
fluid, which is already rotating at its equi-
librium speed. Thus the particles all start
to settle from the same point at the same
time and, in principle, only one size of par-
ticle can be present in the detection zone at
a subsequent time. If such a settling regime
can be stabilized it offers greater resolu-
tion.
This instrument covers a size range that
is inconvenient for many other instru-
ments but which is the range of many
ceramic powders, say lO^-lOjim. Even
finer particles can be analyzed with an-
other form of centrifugal device: sedimen-
tation field flow fractionation. With this
Figure 4-7. Traditional wire mesh test sieves.
device the operating range is further re-
duced by an order of magnitude (Gid-
dings, 1993).
calibration. Figure 4-7 shows a photo-
graph of typical test sieves.

4.4.3 Sieving 4.4.4 Calibration

Sieving is probably the oldest technique A standard particle characterization
and is ostensibly the simplest. The particles laboratory might be equipped with these
are separated into a number of size frac- three techniques: sieving, electrical sensing
tions which are subsequently weighed. The zone, and sedimentation, both gravitation-
result is a distribution by mass of the al and centrifugal. A microscope is also
equivalent mesh diameter. Since a large essential. Each technique has its strengths
sample can be used and the fractions accu- and weaknesses. The sieves can separate
rately weighed, determination of the distri- large samples into smaller fractions, while
bution function can be very accurate. The the Coulter counter can analyze the frac-
size range which can be handled is very tions particle by particle. The instruments
large; sieves with apertures as small as are sensitive over different size ranges, but
10 (im have been made and there is no up- by using them in conjunction, an accurate
per limit. The reproducibility of the tech- analysis can be achieved. The main prob-
nique depends upon the quality of the sep- lem is that they measure different equiva-
aration; it is never perfect. Even this im- lent diameters and so the results from dif-
perfection can be partially eliminated by ferent techniques are not easily combined.
112 4 Characterization of Particles and Powders

In order to combine the results of two dif- the holes in the sieve is the mesh size and
ferent techniques, care must be taken to the two will probably be different.
calibrate one to measure the same parame- By considering Fig. 4-9 it is easy to see
ter as the other. In fact, it is even better if the different checks and calibrations that
one parameter, i.e., the equivalent volume can be made on two particle size measure-
diameter, is considered to be the basic par- ments of the same material made by two
ticle size. The other techniques can be cal- different techniques. The figure compares
ibrated to measure this diameter as well as the plot of the distribution of equivalent
their natural parameter. volume diameter and of sieve diameter for
For example, consider the suggestion the same material. The two plots are, of
for calibrating the sieving technique, which course, different. Long, thin particles can
was made as early as 1927 by Andreason pass through a small sieve but have a large
(Andreason, 1927). In this technique a equivalent volume. Flaky particles, on the
sample of the actual material is sieved and other hand, are retained on the sieve but
the fraction which passes removed. The re- have a small equivalent volume. Spherical
mainder is than sieved further and only a particles should have the same size on both
few particles pass the mesh. These are the axes and so can be used to check the per-
'near mesh' particles (Fig. 4-8) and can be formance of the instrument. By using
considered to be a sample of all the shapes closely sized fractions of spherical parti-
which just pass the mesh. By dividing their cles, the linearity of the instruments can be
total weight by their number, the average checked. When the two instruments are to
volume diameter is determined. This is the be calibrated for irregular particles, several
effective cut size of the sieve, expressed as fractions must be created which are mono-
an equivalent volume diameter, specifi- sized on one axis. This is the basis of the
cally for that material. The actual size of Andreason technique for sieves, the parti-
cles in a near mesh fraction all have the
same sieve size but a distribution of equiv-
alent volume diameters. This basic idea
can be applied to other techniques. Some
particle-measuring techniques claim to be
absolute but, in fact, any technique bene-
fits from calibration.

4.5 On-Line Particle Measurement

The final question is where to measure
the particle characteristics. The traditional
off-line procedure is to take a sample of the
material back to the laboratory, to careful-
ly divide and prepare, it and then to ana-
lyze it (Fig. 4-10a). The maximum care
and expertise can be applied and good ac-
Figure 4-8. Particles which just pass the sieve - the curacy achieved. If the measurements are
"near-mesh" particles. being used to control the product of a pro-
 4.5 On-Line Particle Measurement 113

mass distribution

equivalent
^v volume diameter

J. increasing
aspect ratio Figure 4-9. Correlation of
the same particle size dis-
tribution measured by two
different methods.

cess, then the measurement is retrospec-

tive. If the measurement is being applied to
the starting materials of a process, then the
inherent implication is that the material
properties must be adjusted at the start of
the process and that no further monitoring
is possible.
(a)
In both cases the opportunity to use the
measurement to influence the process is
lost. The alternative is to make the mea-
surement in real time. Real time analysis
means that the measurement is available
(b) within a time scale that is short compared
to the process time. This usually implies an
automated measurement, either on-line or
in-line (see Fig. 4-10b, c). In principle, any
measurement which can be made in the
(c)
laboratory can be adapted to be made on-
line. The only difference is that the act of
sampling and preparation is automated.
There may be a loss of accuracy, but the
benefits are considerable. First, the num-
(d) ber of readings becomes statistically mean-
ingful and the course of the process can be
Figure 4-10. Modes of particle measurement - the
monitored. The measurements may be
transition from off-line to real time (top to bottom: passed to a control program and the
off-line, on-line, in-line, in situ). course of the process may be influenced.
114 4 Characterization of Particles and Powders

Such a control strategy, model-based con- A yet more modern development is to

trol, can be developed for both batch and use an array of detectors outside the pro-
continuous processes. However, besides cess to deduce the concentration of parti-
these obvious advantages the level of so- cles inside it (Beck et al., 1993), Fig. 4-10d.
phistication may be even higher. Thus the These tomographic techniques are current-
measurement system and the model can be ly being developed to also measure the par-
used conjunctively as a so-called 'observer' ticle velocity. In any case, the course of
system (Fig. 4-11). The model predicts the development is increasingly to techniques
measurement which is compared with the which can be automated and fast, and
actual measurement. If the two are wildly which can be adapted to real time analysis.
different then a fault alarm is raised. If In the past decade this has meant optical
they are slightly different, the model is ad- techniques.
justed slightly or the measurement decon-
volution is improved. Over short periods
any variation is probably due to small 4.5.1 Light Scattering Techniques
changes in the equipment caused by wear. An understanding of the phenomenon
Thus the model continuously updates it- of light scattering by small particles has
self. In the longer term, use of this model existed for more than one hundred years
improves the accuracy of the measure- (Rayleigh, 1871). The application of that
ment, because a measurement is always knowledge to particle size measurement
easier when made with prior knowledge. has been common practice for more than
An on-line measurement still requires a fifty years. However, its dominance in the
sample to be taken, which is probably still field of particle measurement has only oc-
the greatest source of variation and also curred in the last decade and is due to a
lack of accuracy. Some instruments are number of factors. The first is the ready
now being developed to operate in-line, availability of the laser, which provides an
that is, to intrude the particle detector into intense, monochromatic, and coherent
the sample. These techniques tend to re- source of light. The second is the availabil-
duce the sampling error but decrease the ity of cheap computers with which the nec-
accuracy. However, a suitable observer essary data recording, calculations, and
system can overcome this problem. deconvolutions can be rapidly made.
A general law of light scattering has
been known for more that 80 years (Mie,
u y __ 1908). For a spherical particle the intensity
Process Sensor
/, of the scattered light is given by
T 22
Process Sensor y (h+i±) (4-12)
FDI
model model
e
Error nx
feedback i,i±=/ (4-13)
Figure 4-11. An observer model that leads to im-
where R is the distance from the scatterer
provements in the measurement and the model, u,
Process input signal; y, sensor output; y, predicted to the detector, /(|, i± are the components
sensor output; e, error signal; FDI, fault detection polarized parallel and perpendicular to the
and isolation. incident plane, m is the complex refractive
 4.5 On-Line Particle Measurement 115

index of the particle relative to the medi- instructive to consider how these factors
um, x is the particle diameter, and X is the have combined to lead naturally to the de-
wavelength of the light. sign of some commercially available in-
This equation tells us a number of facts struments which are particularly useful for
about the scattering: preparing and characterizing ceramic pow-
ders. All need to be calibrated.
(a) For large particles, the scattering is
The light may be presented in the fol-
proportional to the square of the diameter,
lowing forms:
for small particles to the sixth power.
(b) The intensity of the scattered light is (i) The light beam may be small com-
proportional to the incident light and to pared to the particle. Thus the intention is
the square of the distance from the parti- that the beam will be intercepted for a time
cle. Thus the light from a single small par- that is dependent upon the size of the par-
ticle is very little indeed, and the instru- ticle. Since it is difficult to control the ve-
ment design must be very sophisticated. locity of the particles, the beam is usually
(c) The scattering pattern may be a com- scanned across the particle at a much
plex function of the refractive index and higher, but known, velocity. Since the par-
the particle size. An additional feature is ticles in a stream probably overlap, the in-
that the degree of polarization of the light strument works by detecting the backscat-
is changed by the scattering. tered signal.
The instrument shown in Fig. 4-13 is
This complexity of the signal gives an
one of the more successful for use in the
equal opportunity for error and for sophis-
in-line mode. It is useful for monitoring
tication in the measurements. In an optical
particles suspended in fluid and can work
instrument the light is passed through a
at higher concentrations. The parameter
suspension of the particles, either in liquid
measured is, in principle, a chord size,
or air, is scattered, and is received by some
which should be calibrated against particle
form of detector array (Fig. 4-12a).
size.
The light can be transmitted through
windows and lenses or, alternatively, opti- (ii) The beam may be comparable in size
cal fibers can be used to transmit the light. to the particle. The particle is immersed in
Three basic choices may be recognized in the beam for a time and scatters the light in
the design of an instrument: the way in every direction. This leads naturally to a
which the light is presented (Fig. 4-12 b), stream scanning counter; one in which the
the manner in which the particles are pre- particles are passed one by one through the
sented to the instrument (Fig. 4-12 c), and detection zone. Thus the instrument
the mode in which the scattered light is counts efficiently, but the detector must be
measured (Fig. 4-12 d). very sensitive to detect very small particles
When it is realized that these choices and ambiguity is possible about their size
can, in principle, also be combined into because the intensity distribution across a
one instrument, then it is clear that there laser beam is Gaussian.
are multifarious opportunities for instru- An example of this type of instrument is
ment designers. With the commercially shown in Fig. 4-14. It is particularly useful
available instruments, the greatest diver- if a number distribution of particle sizes is
sity is with the first two factors; most, but required. As with the Coulter counter, it is
not all, still only detect intensities. It is relatively easy to calibrate with spheres of
116 4 Characterization of Particles and Powders

(d)
Attenuation

Ratios
On/off

Scattering
(average
{fluctuation
Polarization
Phase

(b)
Imaged
-0- Small
Diffuse
Area
Line
Points
Convergent
Similar
Divergent

Collimated
Figure 4-12. Some possible arrangements of light
Q O scattering instruments, (a) General arrangement: the
0 Large
incident light is scattered and then detected, (b) Vari-
O n 0 ous modes of the incident light, (c) Various modes of
presentation of the sample, (d) Various modes of de-
White light
tection of the scattered light.
Multiple
Monochromatit

Coherent
Crossed
beams known size, but suffers the disadvantage
that the particles must pass through small
orifices or tubes.
(iii) The beam may be expanded to con-
(0 tain many, maybe several thousand, parti-
Presentation Motion cles. The scattered light is now much stron-
ger, but is due to the contribution of many
o o particles. Thus the signal must be deconvo-
Static
luted back into a particle size distribution.
There are two basic forms of this instru-
ment. Most instruments intended for
Random
0*"? I? motion larger particles use an array of detectors to
measure the intensity of the scattered light
Segregated
as a function of the angle. The deconvolu-
O tion then consists of using the known scat-
ter pattern of a number of size classes to
deduce the contribution made by each size
o f 1 ? ^ Constant
a f j 5 f speed
class to the total pattern.
Stream There are many commercial versions of
this forward scattering instrument (see
Accelerate or Fig. 4-15), and it is a versatile and power-
decelerate
ful technique for both off-line and on-line
 4.5 On-Line Particle Measurement 117

use. With calibration a wide particle size

range can be measured (0.1 jim-2000 jum).
/ probe head PARTICLE The only major limitation is that the parti-
I " FLOW cle concentration must be limited.
(iv) For smaller particles, a beam that is
large compared to the particles is also
used, but the light is received by a single
detector, often arrayed at right angles to
the incident light. The detector records not
the average intensity but the fluctuations
which arise mainly from the Brownian mo-
tion of the particles. This instrument is on-
ly possible because the laser provides a
source of coherent light. The size parame-
SCANNING FOCAL POINT
ter is now based not upon the light scatter-
Figure 4-13. Schematic diagram of a backscattering ing equations, but on the Stokes-Einstein
instrument (Lasentec). equation.

Figure 4-14. Schematic dia-

gram of a single-particle
counter (PMS).

HeNe loser beam particle collecting multi-element detector

exparider field lens in focal plane lens

P
.-.•.:.v.
r.»V
*7
^ 1
\
obscuration
J detector

[
f .«>=
r)arallel scattered 1
ight direct
beam
scattered
beam
Inonochromatic not collect
ight by tens
printer

computer electronics
Figure 4-15. Schematic dia-
video
gram of a forward laser
display 1 light scattering instrument
control keyboard
(Malvern).
118 4 Characterization of Particles and Powders

laser goniometer—-^ much easier to generate a wide range of

sample wavelengths. The more successful instru-
ments measure the attenuation of a wide
range of frequencies on passing through a
suspension of particles to one detector. As
with the forward scattering optical instru-
Figure 4-16. Typical layout for photon correlation ments, the signal must then be deconvolut-
spectroscopy. PMT: photomultiplier tube.
ed into a particle size distribution (Riebel
and Loffler, 1984). The principle of the
instrument is shown in Fig. 4-17. The com-
This instrument, as shown in Fig. 4-16,
mercially available versions are only able
is not used for particles larger than 1 jim,
to work with particles suspended in liquid,
but its lower limit is a few nanometers.
but can work at much higher particle con-
Thus it is a vital tool for characterizing
centrations than the forward scattering op-
nanoparticles. Deconvolution of particle
tical instrument.
size distributions is particularly difficult
with this instrument (Finsey, 1993).
(v) The beam may be split into two. The 4.6 Statistical Diameters
particles are projected with a known initial
velocity through the beams. The smaller To repeat, a particle size distribution is
particles have a larger deceleration and so often sufficiently unique to control the
take longer to traverse between the two process that makes the ceramic powder or
beams. This is the so-called 'time of flight' to control the reproducibility of the pow-
technique. It is only used for particles sus- ders as precursors. However, the particle
pended in air. size distribution ascribes one number only
(vi) The two laser beams may be crossed. to each particle and that may be too dras-
The light scattered from the point of inter- tic a reduction of the information neces-
section exhibits a Doppler shift between sary to establish the process in the first
the two beams. This may be used in several instance. The implication is then that each
ways to estimate the size of the particles. particle must be characterized by multiple
The most sophisticated uses two, or even measurements. This is usually achieved by
three, detectors to measure the phase dif-
ference of the light from the two beams.
This is the so-called phase Doppler tech-
nique. It is difficult to use this technique
other than with spherical particles, but it
has achieved considerable success in ana-
lyzing spray processes.

4.5.2 Acoustic Techniques

The move to make instruments more ro-
bust for on-line use has led to an increased
r—O~O—

Ultrasound
Measurement
section with
n
Ultrasound
transmitter sample receiver
interest in the use of an ultrasonic wave in in suspension
place of a light beam. It is more difficult to Figure 4-17. Schematic diagram of an ultrasound at-
make a complex detector array, but it is tenuation instrument (Sympatec).
 4.6 Statistical Diameters 119

direct particle characterization techniques,

observing a sample of the particles through
either a microscope or an electron micro-
scope. If such a picture is observed, a
wealth of detail about the particle shapes is
immediately obvious. The particles shown (a)
in Fig. 4-18 would probably yield similar
particle size measurements, but could be-
have very differently in the process.
Such visual information is often used
qualitatively, and adjectives describing
shape features are ascribed to the particles.
This information can be made quantitative
nowadays, relatively easily. A modern im- * .
age processor digitizes the whole picture.
Each pixel lies within a particle (black), in
the space (white), or on the boundary
(gray) (see Fig. 4-19). Thus the coordinates
of each pixel that lies on the boundary are
known, and the image processor is
equipped with edge enhancement pro-
grams to improve the definition.
Of course, there are some practical limi- Figure 4-18. Particles of the same equivalent volume
tations. The most important is that the im- but different shape. Three samples of copper powder:
(a) electrolytic, (b) reduced, (c) atomized.
age is a projected image; only the projected
outline of the particle is seen. If truly three-
dimensional characterization is required,
then polished sections of the particles must
be made and this can only be done with
large particles. A further limitation is that
the pixel must be small compared to the
features to be observed. This requirement
can imply an enormous magnification and
hence a number of pixels for one particle.
This implies, correspondingly, that a
smaller number of particles may be ana-
Figure 4-19. Digitizing the outline of a particle by a
lyzed; this number may become so small pixed array.
that it is no longer statistically representa-
tive of the powder. Thus the image proces-
sor is equipped with a selection of pro- they are still scalar parameters that de-
grams which reduce the array of contour scribe features of the particles but not of
coordinates to parameters of size and the array in which they lie. A few of the
shape. Most of these parameters assume more common parameters which are pro-
that the particles are lying in random ori- grammed into modern image processors
entation, at least in the field of view. Thus will now be described.
120 4 Characterization of Particles and Powders

4.6.1 Feret Diameters the perimeter of the particle, although it

may be a more convenient method of mak-
The simplest and earliest statistical
ing that measurement. Additional infor-
parameters are simply projections of the
mation about the particle shape can be ob-
particle outline. The Feret diameter (Feret,
tained if the relative angular position of
1931) is the distance between extreme tan-
some of the measurements can be re-
gents of a particle, as illustrated in Fig.
corded. Thus Heywood's elongation ratio
4-20, and is simply the projection of the
(Heywood, 1946) is formed by recording
perimeter of a particle in a fixed direction.
two particular measurements of the Feret
A single measurement has little signifi-
diameter which are mutually at right an-
cance but, if the Feret diameter is mea-
gles.
sured in all possible directions, then clearly
^each element of the perimeter is projected
at every orientation. 4.6.2 Fourier Analysis
For a completely convex particle, the av- If the profile of a particle is expressed in
erage Feret diameter is thus half the aver- polar coordinates, then its shape can be
,age projection, and the relationship be- considered to be a wave form having a
tween the Feret diameter (F) and the positive value of radius, r, for values of 6
! perimeter (P) is given by lying between 0 and 2n (Fig. 4-21).
This wave form can be expressed as a
•F-* (4-14) harmonic series, usually a Fourier series
n (Beddow etal. 1977). Thus
The equivalent three-dimensional pa- oo
rameter is the projected area, A, which is r(9) = a0 + X (ancosnO + bnsinn6)
similarly related to the particle surface, n=1
(4-16)
S by the Cauchy relationship (Cauchy,
1840). In principle the whole shape of the out-
line is contained in the coefficients of this
A =~ (4-15) series. The major practical limitation is
that it is difficult to know the point at
If the particle is not convex, there is which the series can be stopped. As the
overlap of portions of the projection, and particles become more angular and irregu-
there is no unique relationship between the lar, so higher order terms are involved.
average Feret diameter and the perimeter. Care must also be taken in the choice of
At the simplest, therefore, the Feret diame- origin; the actual value of the coefficients
ter can convey little more information than is dependent upon this choice although the
total sum is not.

4.6.3 Fractal Dimension

Mandelbrot has discussed the problem
of describing the highly rugged particle
boundaries which occur in nature (Man-
F F F delbrot, 1977). In particular, he showed
Figure 4-20. The Feret diameter of a particle depends that it is useful to consider the space-filling
on its orientation. nature of a rugged outline as being de-
 4.6 Statistical Diameters 121

a measure of the ruggedness of the surface.

The straight line power law fit implies geo-
metrical similarity on different scales of
scrutiny. However, the important point is
that the surface area of a particle is not an
absolute value; it is dependent upon the
size of the probe with which it is observed.
What is a smooth surface to the eye is a
mountain range to a nitrogen molecule.

100 200
I 4.6.4 Chord Size Distribution
300
One stereological approach is to mea-
Figure 4-21. Representation of a particle outline as a
waveform. sure the chord size distribution of the par-
ticle, as illustrated in Fig. 4-23.
The chord size can be defined by draw-
scribed by a mathematical dimension ing all the chords which intercept the parti-
which has fractal values lying between one cle in one direction. By rotating the parti-
and two. Mandelbrot called this dimension cle through all possible orientations, the
the fractal dimension. As applied to a par- total chord distribution is generated. This
ticle, the important fact to realize is that distribution has the advantage that it can
the perimeter of a particle outline is depen- be applied, in principle, to a three-dimen-
dent upon the scale with which it is ob- sional system. Thus if chord measurements
served (Kaye, 1994). If the perimeter is tra- are made on sections of the particle and all
versed with smaller and smaller steps, then the possible sections of the particle are
their total length increases. Thus a plot of taken, then the total chord distribution is
the total length of the inscribed polygon Pk related to the three-dimensional proper-
can be plotted against the step length X ties.
(Fig. 4-22).
Sometimes a power law fit is made
through these points and the correspond-
ing slope is the fractal dimension of the fractal

particle
non-fractal
Px = kX^~D) (4-17)
(k is a constant of proportionality). When non-fractal
this function is plotted as a straight line on
a log-log scale then D is a characteristic for
that particular particle. D is called the frac-
tal dimension of the particle. A larger val-
ue of D implies a more rugged profile.
Of course, the concept of fractal dimen-
sions is applicable in any dimension. For a
three-dimensional particle, the fractal di- Figure 4-22. Characterizing a particle outline by its
mension lies between two and three and is fractal dimension (PXi perimeter for a step length X).
122 4 Characterization of Particles and Powders

Figure 4-23. The chord size

distribution of a particle in
different orientations.

There are three approaches to the use of not to vector properties with direction. A
the chord size distribution. Some authors third use of the chord characterizations is
attempt to relate the chord size distribu- to regard the powder as consisting of a
tion back to the particle size distribution statistical array of chords rather than of
from which it has been derived. This can be particles (Scarlett and Todd, 1963).
done for particles of spherical shape using Chords have only length and so a com-
Wicksell's relationship (Wicksell, 1925, plex geometry may be more easily visual-
1926). In fact, in principle, the particle size ized as chords. For example, the pore
distribution of any set of particles, which space between an array of particles can be
have the same shape but varying particle better visualized as a statistical bundle of
size, can be derived from the chord size chords rather than as an enormously com-
distribution. However, a unique relation- plex pipe.
ship is not possible if both the size and
shape of the particles vary and, in any case, 4.6.5 Measuring and Modeling
there are easier methods of measuring the
These are only some of the parameters
particle size distribution. The chord distri-
which a modern image processor can de-
bution may, however, be a convenient way
termine, and they clearly contain far more
of recording some average of the size and
information about the particle size and
shape parameters. For example, the mean
shape than does the simple particle size
chord is related uniquely to the area-to-
distribution. What is less clear is how to
perimeter ratio, for any particle outline or
choose the characterization appropriate
array of outlines
for the purpose. This is the subject that will
A now be addressed.
(4-18)

The three-dimensional equivalent of this

equation relates the average chord to the 4.7 Powder Properties
volume-to-surface ratio
The production of a ceramic body is a
(4-19) long and complex process, starting with
some feedstock and eventually producing
This relationship applies to any array of a product. Powder technology plays an im-
particles with any shapes. It can, then, be portant role in the intermediate part of this
a convenient way of measuring parameters process. At a certain point the feedstock
such as the volume-to-surface ratio. The becomes powder, and at a later point the
limitation is that the chords must be mea- particles are fused again into a solid body.
sured in random directions and so can only In the interim there is a need to describe
be related to scalar averages of the array, how the millions of particles are interact-
 4.7 Powder Properties 123

ing with their environment and with each

Mechanical
other to achieve the next stage of the pro-
cess. At certain points, a characterization
must be made to answer the questions
posed at the beginning of this chapter, i.e.,
why, what, how, and where? In each case Chemical
the discussion has been intended to be suf-
ficient to make clear one point. That point
is that there are always wider possibilities
and thus there are always choices to be
made. The choices involve both scientific Cluster
and engineering considerations and there
COOH
is never only one correct answer. Con-
versely, there are many incorrect answers. Figure 4-24. The range of particle interactions, from
centimeters to nanometers.
It may be instructive to think of some of
the basic considerations which may lead to
a correct choice of a characterization tech- havior of the particles (Fig. 4-25). In these
nique for a particular circumstance. simpler systems it may well be that an ar-
In the first instance the order of magni- ray of spherical particles simulates the be-
tude of size of the particles dictates the havior of nonspherical particles sufficient-
forces and mechanisms which may pre- ly well. As the concentration becomes
dominate. Large particles are, in their own higher, the particles begin to interact but
right, macroscopic bodies, and it is the me- primarily through the continuum. Thus
chanical forces that dominate (Fig. 4-24). the properties of the interstitial gas or liq-
Particles larger than 100 jim fall into this uid play a more dominating role. When the
class. Their shape is important and they concentration becomes large enough, the
may easily fracture into smaller particles. particles touch their neighbors. The inter-
Particles which are of the order of 1 jum are actions are then more direct and the point
dominated by their surface. An adsorbed contact behavior can dominate.
layer or charge may be the dominating In ceramic processing all of these states
mechanism. Thus particle size measure- may occur, but the preoccupation is with
ments must be supplemented with mea-
surements of zeta potential or electrostatic
charge. When the particles reach the Dilute Concentrated Packed
nanometer range they contain a finite Systems systems systems
number of molecules and every molecule is ° °O
near to the surface. It is this factor, of
o o o
course, which gives nanoparticles their €)
special properties. Interpretation of their o u ©
O
behavior requires a knowledge of molecu- o O
lar bond theory. o
o o
The second consideration is the concen-
tration of the particles. In dilute systems
the particles behave individually and the Figure 4-25. The concentration of the particles leads
behavior of the system is the average be- to different regimes.
124 4 Characterization of Particles and Powders

the dense state. Thus the powders are com- that the characterization is the input pa-
pacted and sintered, and during this pro- rameter for the model and so must match
cess it may be necessary to model the dif- the sophistication of the model. The an-
ferent aspects of their macroscopic behav- swer to the first is that there is a wide
ior. It is in this packed state that the equiv- variety of choice in choosing the complex-
alent sphere approach is least applicable. It ity of the model. At one end of the spec-
is rare that an array of spherical particles trum a simple empirical equation which
mimics the behavior of other particles combines two or three variables into a
when they are closely packed. There re- power law fit may be used. At the other
mains a good chance that a simple particle extreme, the position coordinates of sever-
size analysis will be sufficient to control a al thousand particles may be fed to the
well-known process, but the development computer. The constitutive equations gov-
of a new process requires more informa- erning the flow of heat and fluid are added.
tion. The particle size and shape influence The particles are allowed to move or de-
the process in several different ways, as form by a small increment and the new
shown in Fig. 4-26. Some of the behavioral coordinates calculated. Such a model is re-
aspects depend primarily upon the pore ally a simulation, imitating in the comput-
structure and so the model must imitate er the actual process. Between these two
how the particles may pack together. extremes any intermediate level of model is
Other aspects are dominated by the point possible. The choice is whether, for a given
contacts and the forces and deformations problem, it is sufficient. Whatever that
that can occur there. The distribution and choice, the particle characterization tech-
disposition of the contacts is then the vital nique can be found.
geometrical factor. The most difficult
models are when the particles move rela-
tive to each other. A simple strain is impos-
sible in a powdered material when it shears 4.8 References
and compacts, and the model must simu-
Allen, T. (1990), Particle Size Measurement. London:
late the deformations which occur and an- Chapman & Hall.
isotropies which arise. Andreason, A. M. H. (1927), Sprechsaal 60, 515.
What sort of models can hope to de- Beck, M. S., Campogrande, E., Morris, M., Williams,
R. A., Waterfall, R. C. (Eds.) (1993), Tomographic
scribe such complex behavior and what Techniques for Process Design and Operation.
have they to do with particle characteriza- Southampton, U.K.: Computational Mechanics
tion? The answer to the second question is Publications.
Beddow, J. K., Philip, G. C , Vetter, A. F. (1977),
Powder Technol. 18, 19.
British Standard (1995), British Standard 3406, Part
Point Contact ' Pore Structure * Relative Movement 1-8.
Cauchy, A. (1840), C. R. Acad. Sci. 13, 1060.
DIN (1981), DIN 66142.
Feret, R. L. (1931), Assoc. Int. Essai Mater. 2,
Group D., Zurich.
Finsey, R. (1993), Part. Part. Syst. Charact. 10, 118.
Giddings, J. C. (1993), Science 260, 1456.
Heywood, H. (1946), Trans. - Inst. Min. Metall. 55,
391.
ISO (1995), ISO 9276-1.
Figure 4-26. The structure of packed beds: point con- Kaye, B. H. (1994), A Random Walk Through Fractal
tacts, pores and movement. Dimensions, 2nd ed. Weinheim: VCH.
 4.8 References 125

Mandelbrot, B. P. (1977), Fractals, Form, Chance and Bohren, C. R, Hoffman, D. R. (1993), Absorption and
Dimension. New York: W. H. Freeman. Scattering of Light by Small Particles. New York:
Mie, G. (1908), Ann. Phys. (Leipzig) 25, 377. Wiley.
Rayleigh, Lord (1871), Philos. Mag. 41, 107. Chu, B. (1991), Laser Light Scattering. San Diego,
Riebel, U., Loftier, F. (1984), Part. Part. Syst. Char- CA: Academic Press.
act. 6, 135. Friedlander, S. K. (1977), Smoke, Dust and Haze.
Scarlett, B., Todd, A. C. (1963), Trans. ASME 91, New York: Wiley.
478. Happel, I, Brenner, H. (1965), Low Reynolds Number
Stokes, G. G. (1845), Trans. Cambridge Philos. Soc. 8, Hydrodynamics. Englewood Cliffs, NJ: Prentice
281. Hall.
Waddell, H. (1932), /. Geol. 40, 443. Herdan, G. (1960), Small Particle Statistics. London:
Waddell, H. (1933), /. Geol. 41, 310. Butterworths.
Wicksell, S. D. (1925), Biometrika 17, 84. Hiemenz, P. C. (1986), Principles of Colloid and Sur-
Wicksell, S. D. (1926), Biometrika 18, 32. face Chemistry. New York: Marcel Dekker.
Iionoya, K., Gotoh, K., Higashitani, K. (Eds.)
(1991), Powder Technology Handbook. New York:
General Reading Marcel Dekker.
Nedderman, R. M. (1992), Statics and Kinetics of
Granular Materials. Cambridge: Cambridge Uni-
Allen, T. (1990), Practicle Size Measurement, 4th ed. versity Press.
London: Chapman & Hall. Rumpf, H. (1990), Particle Technology. London:
Beck, M.S., Campogrande, E., Morris, M., Williams, Chapman & Hall.
R. A., Waterfall, R. C. (Eds.) (1993), Tomographic Soo, S. L. (1990), Multiphase Fluid Dynamics. Beijing:
Techniques for Process Design and Operation. Science Press.
Southampton, U.K.: Computational Mechanics
Publications.
5 Die Pressing and Isostatic Pressing
Denis Bortzmeyer

Rhone-Poulenc Recherches, Aubervilliers, France

List of Symbols and Abbreviations 129

5.1 Introduction 130
5.2 Compaction Behavior 130
5.2.1 Density-Pressure Relationship 130
5.2.2 Density Variations 131
5.2.2.1 A Simple Model for Density Variations 131
5.2.2.2 Radial Pressure Coefficient 132
5.2.2.3 Wall Friction Coefficient 132
5.2.2.4 Continuum Mechanics 133
5.2.3 Microstructure 134
5.2.4 Green Sample Strength 134
5.2.5 Conclusion 135
5.3 Ungranulated (Unagglomerated) Powders 135
5.3.1 Mean Size 135
5.3.1.1 Density-Pressure Relationship 135
5.3.1.2 Radial Stress Coefficient 136
5.3.1.3 Wall Friction Coefficient 136
5.3.1.4 Green Fracture 137
5.3.2 Size Distribution 138
5.3.2.1 Compaction-Pressure Relationship 138
5.3.2.2 Other Parameters 138
5.3.3 Particle Morphology 139
5.3.3.1 Angular, Flat, or Acicular Particles 139
5.3.3.2 Rough Versus Smooth Particles 140
5.3.4 Particle Hardness 141
5.3.5 Aggregation 141
5.3.5.1 Density-Pressure Relationship 141
5.3.5.2 Microstructure 141
5.3.5.3 Other Parameters 141
5.3.6 Mixing Different Powders 142
5.3.7 Conclusions 142
5.4 Granulated (Agglomerated) Powders 143
5.4.1 Granule Strength 143
5.4.2 Influence of the Granule Parameters on the Compaction Behavior 145
5.4.2.1 Granule Size 145

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. Allrightsreserved.
128 5 Die Pressing and Isostatic Pressing

5.4.2.2 Granule Density 145

5.4.2.3 Granule Structure 146
5.4.3 Binder Influence 147
5.4.3.1 Powder + Binder Cohesion 148
5.4.3.2 Influence of Molecular Weight 148
5.4.3.3 Influence of the Glass Transition Temperature 149
5.4.3.4 Latexes 149
5.5 Conclusions 150
5.6 References 150
 List of Symbols and Abbreviations 129

List of Symbols and Abbreviations

A contact area
D diameter
dp particle diameter
%P mean particle diameter
F interparticle force; load
idyn friction force
h height
H total height
nP particle hardness
wall hardness
k radial pressure (stress) coefficient; stress ratio
molecular weight
N amount of powder
P pressure; load
r radius
R radius; gas constant
Rw wall roughness
T temperature
Tg glass transition temperature
V volume of pores
z height; coordination number
8 displacement
£ porosity
initial porosity
wall friction coefficient
G stress; pressure
°r radial pressure
ultimate tensile stress
wall stress
Gw mean wall stress
°z axial pressure
X friction force
AYP apparent yield point
PEG poly(ethylene glycol)
PVA poly(vinyl alcohol)
RH relative humidity
130 5 Die Pressing and Isostatic Pressing

5.1 Introduction this mean density is only a rough measure

of the sample structure. For a given mean
Forming by compaction is widely used density, a shaped mold may lead to local
for a variety of granular materials: ceramic density variations which depend on the
powders, pharmaceutical excipients, metal- powder. In addition, the microstructure of
lic powders, etc. Its advantages are high the green compact will be influenced by the
production rates and good dimensional powder morphology and size, etc. Last but
control. Ceramic powder compaction, for not least, green products must sustain ejec-
example, has been studied by many au- tion, handling, and machining without
thors. The aim of this paper is to give a fracture. The cohesion of the powder is
review of these studies, focused mainly on thus a very important parameter. The
the relationships between the characteris- measurement of these characteristics, as
tics of ceramic powders and their com- described in the literature, is reviewed be-
paction behavior. Section 5.2 deals with low.
the meaning of "compaction behavior";
the pressure/density relationship is often
5.2.1 Density-Pressure Relationship
used, but is obviously not sufficient. The
main measurements necessary for full If the green compact is assumed to be
characterization are described. homogeneous, what is the relationship be-
The powder behavior may be dramati- tween applied pressure and density?
cally changed by an agglomeration step, The pressure-density relationship in
such as occurs, for example, during spray- uniaxial compaction is easy to measure
drying. Thus it appeared to be preferable continuously. Of course precautions have
first to investigate the behavior of unag- to be taken in order to avoid erroneous
glomerated (ungranulated) powders in or- interpretation of the results: the sample
der to describe the influence of particle must be thin enough so that density varia-
size, morphology, etc. This is covered in tions are negligible; the compaction cell
Sect. 5.3. The problems caused by agglom- should preferably be shaped so that maxi-
eration (granulation) are addressed in mum symmetry is ensured; a blank curve
Sect. 5.4. Even without binder, spray-dry- should be recorded before the experiment
ing can create agglomerates, the structure in order to take into account the elastic
of which will have an influence on the com- stiffness of the load cell and punches. This
paction behavior. This relationship is stud- is quite easy to do.
ied first, then the problem of binder choice However, this kind of measurement is
is addressed. difficult in isostatic compaction since air
bubbles and liquid compressibility will in-
troduce a large amount of experimental
5.2 Compaction Behavior scatter (liquid compressibility may be of
the same order of magnitude as powder
What are the best measurements for compressibility).
complete characterization of the com- The result of the measurement in uniax-
paction behavior of a powder? ial compaction is usually slightly curved in
When a new powder is to be tested, the a semilogarithmic plot, as shown by Dynys
engineer first wants to know what its den- and Halloran (1983; Fig. 5-1). Several au-
sity is for a given compaction pressure. But thors stated that this relationship contains
 5.2 Compaction Behavior 131
50 -i 5.2.2 Density Variations
• 92% aggregated
2 45^ o 60% aggregated Here the green product is no longer as-
a 36% aggregated sumed to be homogeneous. The problem is
g 4<H • 6% aggregated
to predict and control its density varia-
tions, taking into account the powder and
powder/tool characteristics.
30

5.2.2.1 A Simple Model

25
for Density Variations
20
The simplest model for density varia-
Tapped density 11.8% tions in a cylindrical specimen assumes
0.01 OJ02 0.05 0.1 02 0.5 1 2 5 10 20 50 100 that the axial pressure is constant in a
Pressure(MPa) plane perpendicular to the compaction
Figure 5-1. Green density vs. compaction pressure axis. The force balance on such a plane
for aggregated powders (Dynys and Halloran, 1983). (Fig. 5-2) leads to
© The American Ceramic Society; reproduced by
permission. o{z + d z ) 7 i R 2 - G(z)nR2 =
(5-1)
Thus
two straight lines separated by a break-
do-
point (Niesz et al. 1972; Groot Zevert et al. (5-2)
1990), which is thought to be a measure of
the hardness of the granules (Lukasiewicz Two hypotheses are necessary to proceed
and Reed, 1978; Groot Zevert et al. 1990; further:
Harvey and Johnson, 1980). However, the
break-point is not always easy to measure - that the wall friction obeys Coulomb's
precisely; it seems that homogeneous and law: T(R) = juar(z), and
monodisperse aggregates (or granules) are - that the radial pressure (crr) is propor-
necessary for a clear determination. tional to the axial pressure (<7Z), i.e.,
crr(z) =koz(z).
Another advantage of continuous mea-
surement is that the springback which oc- Then an exponential decrease in the ax-
curs during ejection is easily measured. ial stress along the height is predicted:
This springback is often referred to in the
literature (either for ceramic or pharma- (5-3)
ceutical powders) as the "elastic spring-
back"; however, as usual in the mechanics where P is the pressure.
of granular media, every small deforma-
tion contains a substantial plastic contri- ff{z+dz)TiR
bution. This springback is often neglected,
although it is related to fracture upon
unloading (Thompson, 1981a, b). Hahn
(1986) has made an extensive study of the
influence of process conditions (punch
speed, etc.) on its magnitude. Figure 5-2. Force balance on a cylindrical specimen.
132 5 Die Pressing and Isostatic Pressing

This model is of course far too simple though this direct method is indeed prefer-
and shows that the pressure and density able, an indirect measurement of the prod-
variations along the mold height depend uct \xk is obtained through the determina-
mainly on the radial pressure coefficient k tion of the applied and transmitted pres-
and on the wall friction coefficient /i. THe sures, since the force balance on a cylindri-
following two sections describe how these cal sample leads approximately to
are measured.
p p _
applied transmitted
5.2.2.2 Radial Pressure Coefficient 9 (P J_ p ^
^ 7 V^ applied ~ ^ transmitted/
A device used for the measurement of jfv -2

the radial pressure coefficient (or "stress

To our knowledge, satisfactory theoreti-
ratio") k is described by Strijbos et al.
cal models for the pressure/density rela-
(1977; Fig. 5-3). Another example, using
tionship and the radial pressure coefficient
tool deformation, is described by Di Milia
do not exist; it is very difficult, if not im-
and Reed (1983).
possible, to determine the macroscopic be-
havior of the powder from its particle size,
5.2.2.3 Wall Friction Coefficient
morphology, and interaction laws. Wall
An example of an appropriate device for friction should be easier to model since it
the measurement of the Wall friction coef- involves the interaction between a single
ficient is given by Strijbos (1977). Al- particle and a surface. The two following
models, however, prove that this issue is
still controversial.
Kendall's model (Kendall, 1986) starts
with the contact law between a sphere and
a surface. Without the van der Waals' at-
traction, the elastic deformation of a
sphere obeys Hertz's law, but Johnson
etal. (1971) developed a model several
years ago that takes into account the inter-
action between the sphere deformation
and the van der Waals' attraction. Further
calculation using this model allowed
Kendall to state that the wall friction coef-
ficient is greater for smaller particles be-
cause of an increased particle/wall contact
area per macroscopic unit surface area.
Strijbos's model (Strijbos, 1977) is sim-
pler since it assumes that the friction is
greater with small particles because they
are able to enter the asperities of the wall
surface and form a coating on it. Thus the
Figure 5-3. Apparatus used to measure the distribu-
tion of the normal stress on the die wall and punches apparent friction coefficient is in fact the
during compaction (Strijbos et al., 1977). S: strain powder/powder friction, which is greater
gauge. than the powder/wall friction (Fig. 5-4).
 5.2 Compaction Behavior 133

Fine powder Coarse powder 1.5' -

°w /cf w

High friction Low friction

Independent of Dependent on
H9/H* and Hp/Hw and dp
2
Figure 5-4. Sliding friction between a stationary * w ( N / m 2 * 10 ) Hlmm)
0.5 -
powder compact and a moving wall (Strijbos, 1977). o 1.59 17.2
Note that small particles enter the wall asperities, so x 3.18 14.7
that the shear plane is located in the powder, thus 4- 4.78 13.0
6.37 11.9
increasing the apparent friction coefficient.

Strijbos's model has been proved experi-

mentally, as will be shown in the next sec- 05 1.0
tion. h/H-
Figure 5-5. The distribution of the wall normal stress
5.2.2.4 Continuum Mechanics for 8 g of BASF ferric oxide powder in a lubricated die
at four compacting pressures (Strijbos et aL 1977)
Figure 5-5 shows the variation of the
wall stress (aw), normalized by the mean
wall stress (oQ along the normalized tions the experimental determination of
height (h/H) of the compact. There is a the flow rule can be conducted on an
discrepancy with the theoretical prediction axisymmetric sample with an axisymmet-
given by Eq. (5-3) (exponential variation ric stress state. This is the purpose of the
of crw along the height). This is not surpris- triaxial pressure cell, which is nothing
ing since the hypothesis is far too simple more than an improved compaction device
(constant stress on a horizontal plane). in which the radial stress is controlled
Even if this hypothesis were true, the (Fig. 5-6).
simple calculation could not be used for a The question is, whether triaxial tests
shaped mold; a more complicated model is are absolutely necessary for such numeri-
needed for stress and density variations. cal calculations. A rigorous determination
Such a model involves the use of con- of the flow equation does indeed need ex-
tinuum mechanics, the basis of which is tensive tests on a triaxial cell. However, the
described in Gudehus (1977). Roughly simplest mechanical models may be fit us-
speaking, the mechanical behavior of the ing the simple tests described above (pres-
powder is no longer described by a simple sure/density relationship in uniaxial and
pressure/density relationship, but by a isostatic experiments, stress ratio, wall fric-
complete stress/strain flow rule. Stress and tion coefficient) and a few assumptions on
strain are 3 x 3 tensors instead of scalars, the yield surface and flow rule. Then a
but by making a few reasonable assump- complete determination of the stress and
134 5 Die Pressing and Isostatic Pressing

Upper punch

- High pressure cell

Sample
Measurement Polymer jacket preventing sample from fluid contact
of sample height
' ^ High-pressure fluid

Figure 5-6. Triaxial pressure cell.

density variations in a shaped sample can described in soil mechanics (Oda and
be carried out with the help of a computer Konishi, 1974). In ceramic technology, the
and a finite element method (for a few ref- green compact anisotropy is proven by the
erences, see Bortzmeyer, 1992 b). In a se- difference between the axial and radial
ries of very stimulating papers, Chandler sintering shrinkages of a compacted sam-
(1985, 1990) was able to determine analyt- ple obtained with an equiaxed powder
ically the powder flow equation using a (Fig. 5-8).
few hypotheses concerning the mecha- Pore morphology and size distribution
nisms involved in compaction (rearrange- are very important for the sintering behav-
ment and/or plastic deformation of gran- ior and tensile strength of the fired prod-
ules). However, the application of this the- uct. Examples are given in the following
ory to unagglomerated powders remains sections, showing that powder morpholo-
questionable. gy and the forming process can have a very
great influence on these factors.
5.2.3 Microstructure
5.2.4 Green Sample Strength
Two homogeneous samples with the
Sample fracture is a common problem
same green density may have unequal den-
encountered during powder processing. In
sities after sintering due to several factors,
order to avoid costly scrap, two factors
including anisotropy and pore size distri-
must be controlled:
bution. Anisotropy may involve either
particle anisotropy (flat particle powders, 1) It is necessary to predict the tensile
for example) or packing anisotropy stresses which will arise upon unloading
(Fig. 5-7; see Oda, 1979). Packing an- and ejection. Stress inhomogeneities dur-
isotropy arises because the particle/parti- ing compaction turn into tensile stresses
cle contacts tend to be oriented parallel to after ejection; this has been proven by nu-
the major stress axis. This has been widely merical simulations (Thompson, 1981a;

Particle anisotropy Packing anisotropy (contacts are

mainly oriented in a given direction)
Figure 5-7. Two kinds of anisotropy.
 5.3 Ungranulated (Unagglomerated) Powders 135

relationship; 2) the radial stress coefficient;

3) the wall friction coefficient; 4) the green
sample microstructure; and 5) the green
tensile strength. The influence of the pow-
der characteristics on these factors will
now be reviewed.

10
5.3 Ungranulated (Unagglomerated)
100 200 Powders
Pressure (MPa)

Figure 5-8. Axial and radial shrinkages for a com- Granulation (agglomeration) of pow-
pacted sample, as a function of compaction pressure ders, through spray-drying for example,
(equiaxed zirconia powder).
dramatically affects their behavior (green
microstructure and density). The influence
Brown and Weber, 1988). It is thus neces- of granulation can be as important as the
sary to be able to calculate the stress varia- influence of the raw powder characteris-
tions during compaction and ejection. tics. As a consequence, granulated pow-
2) The green cohesion must be adequate. ders will be studied in a separate section;
this section is only concerned with raw
While our knowledge of the tensile powders. The characteristics involved in-
strength of dense samples is very good, our clude mean size, size distribution, particle
understanding of the cohesion of particle morphology and roughness, surface chem-
packings is still quite poor. Earlier studies istry, and aggregates.
used tensile strength results to investigate
the relationships between packing density,
5.3.1 Mean Size
or binder content, and strength; these stud-
ies were more or less correlated with theo- 5.3.1.1 Density-Pressure Relationship
retical models (Schubert, 1975; Onoda,
For small particles ( < 10 |im), interpar-
1976). However, it is now recognized that,
ticle cohesion exceeds the influence of
as for any brittle material, the green prod-
weight. As a consequence, particles stick to
ucts must be studied with the help of frac-
each other so that the initial packing of the
ture mechanics. This has been addressed
powder is loose (relative density <0.6).
by several authors (Adams et al., 1989;
However, as soon as the compaction pres-
Kendall etal., 1986; Bortzmeyer, 1992a;
sure exceeds the tensile strength of the
Bortzmeyer et al., 1993), but their conclu-
packing (i.e., a few megapascals), the inter-
sions (interparticle forces, density depen-
particle forces created by this pressure are
dance) are still controversial, and further
greater than the cohesion forces. Thus the
studies are needed in this very interesting
cohesion/weight ratio is not able to ac-
field.
count for the porosity of the packing under
pressure. This porosity can only be ex-
5.2.5 Conclusion plained by a geometric effect; the packing
The powder compaction behavior can is organized in arches that are able to resist
be characterized quite well by the follow- the pressure even with small interparticle
ing measurements: 1) the pressure/density cohesion forces. These arches and macro-
136 5 Die Pressing and Isostatic Pressing

pores collapse through a buckling mecha- Table 5-1. Relationship between particle size and
nism (Kuhn etal., 1989; Bortzmeyer and stress ratio a .
Abouaf, 1989). Particle size 85/100 60/72 44/52 30/36 22/25
The bigger the particles, the smaller (mesh)
their cohesion/weight ratio. As a conse-
quence, the initial density of the packing is Stress ratio 0.31 0.30 0.29 0.28 0.23
expected to be higher. To our knowledge, a
Carless and Leigh (1974).
no model exists that is able to calculate the
slope of the pressure/density relationship
using the powder parameters. On the other 5.3.1.3 Wall Friction Coefficient
hand, if it is assumed that the buckling
As explained before, the influence of
mechanism described above is indepen-
particle size on the wall friction coefficient
dent of the particle size, then the slopes
depends on the roughness of both the wall
corresponding to different particle sizes
and the particles. If a perfectly smooth
must be equal. This is indeed observed in
wall is used with perfectly spherical parti-
ceramic technology (Fig. 5-9) and in phar-
cles, then it is expected from Kendall's the-
maceutical powder technology (Carless
oretical model (Kendall, 1986) that the
and Leigh, 1974).
friction coefficient will increase as the par-
ticle size decreases. If a "rough" wall (on a
5.3.1.2 Radial Stress Coefficient
particle scale) is used (and this is probably
The influence of particle size on the what usually happens), Strijbos's geomet-
stress ratio (see Sect. 5.2.2.2) is the same rical model may apply better. This model
for pharmaceutical powders (Carless and emphasizes the particle size/wall rough-
Leigh, 1974) and for ceramic powders ness ratio. If the particles are bigger than
(Bortzmeyer, 1990): the finer the particles, the wall asperities, the measured friction
the higher the stress ratio (Table 5-1). It is coefficient is the actual wall/powder coeffi-
sometimes stated that the higher stress ra- cient, but if they are smaller, they are able
tio of fine particles comes from the higher to enter these asperities so that the shear
density of particle/particle contacts in the plane is located in the powder, thus in-
sample. creasing the appearent friction coefficient
(Fig. 5-4). Figure 5-10, reprinted from
Strijbos's studies (Strijbos, 1977), supports
this model; the friction force is constant if
the ratio of the particle diameter/wall
roughness is less than one (see also Tan
and Newton, 1990).
Since the fik product is greater for small
0.50 pm
0.70jjm particles, their stress and density varia-
tions in a mold will be greater too. Indeed,
it is well known that small particles are
more difficult to process than large ones.
10 100 1000
Compaction pressure (MPa)

Figure 5-9. Influence of mean particle size on pres-

sure-density relationship (zirconia powder).
 5.3 Ungranulated (Unagglomerated) Powders 137
0.9 -|
Hp/HM*05 1) The air volume flow is calculated
*? 0.7 -
from the air pressure in the green compact
OO-o—
cu
and Darcy's law.
° 0.5- Op 2) The air volume and pressure in the
D
^0.04 green compact are related by the classic
0 3 - o0.1
\
| law P V=NR T, where V is the volume of
^4
u_ \ A
a25 pores, N the number of moles of gas, R the
0.1 -I
101 10° 101 10 2 10 3 gas constant, and T the temperature.
d/Rv
These assumptions result in a differen-
tial equation solved by numerical simula-
" I 0.7 - tion. The authors show good agreement
n ° n /H ^05
•—-
between calculated and experimental pres-
2 0.5- > >^3 sure profiles. They are able to determine
o

i 0.3-
an optimal pressure sequence in order to
B
minimize the entrapped air at the end of
£ 0.1 - i 1 the compaction cycle.
> 10 1 2
10 10 3
Thus fine particles are subjected to
higher tensile stresses than coarse particles.
Figure 5-10. Dynamic powder-wall friction between On the other hand, the influence of particle
ferric oxide powders and different walls. (A) Spark- size on the compact tensile strength has to
eroded and polished walls, (B) walls ground perpen-
dicular to the sliding direction (dp = particle diameter; be determined. The result may depend on
Rw = wall roughness;/dyn = friction force; from Strij- the size range and particle morphology,
bos, 1977). because both geometrical factors (number
of contacts per surface area) and mechani-
cal factors (interaction forces) will con-
tribute to the tensile strength.
The number of particle/particle contacts
5.3.1.4 Green Fracture
per surface area (for a given density) will
Green products obtained from small increase as \\d\ where dp is the diameter.
particles are more likely to crack upon un- Since the van der Waals force scales with
loading and ejection since the stress varia- dp, the compact strength is expected to in-
tions (turning into tensile stresses upon un- crease when the particle size decreases.
loading) will be greater. Moreover, fine This conclusion is also reached by Kendall
particles are involved in another cause of etal. (1986), although a different method
green fracture, that is, air entrapment. is used. Indeed, the cohesion of pharma-
Since the pore size is smaller with fine par- ceutical compacts is known to increase
ticles (for a given density), the entrapped when the particle size decreases (Carless
air is less likely to flow out of the green and Sheak, 1976; Krycer et al., 1982).
compact during compaction. Thus the However, if interparticle forces other
pore pressure and the tensile stresses after than the van der Waals force are involved
unloading will be greater. Note that an an- (including mechanical interlocking or hy-
alytical model has been derived by Al-Jew- drogen bonding), the result will depend on
aree and Chandler (1990) using two as- these mechanisms, and is difficult to fore-
sumptions: cast. From the author's experiments, the
138 5 Die Pressing and Isostatic Pressing

tensile strength of green compacts of zirco- For example, a mixture of two different
nia powders with different particle sizes sizes is always denser than a monosized
depends more on the compaction pressure population (Oger et al., 1986).
than on the green density (Bortzmeyer, The main problem is that these theoreti-
1992 a). This occurs because the tensile cal models are not usually able to take into
strength involves mechanical interlocking account the case of highly cohesive pow-
of the particles (Bortzmeyer et al., 1993). ders, and the problem of several particles
Thus this issue is still highly controver- falling one at a time. Some examples have
sial. However, the increase in the tensile been published, but they are not fully satis-
stress when the particle size is decreased is factory (cohesive powders: Suzuki and Os-
greater than the increase in the tensile hima, 1983; collective packing: Thomas
strength. Thus, the smaller the particles, etal., 1989; both: Yen and Chaki, 1992).
the more difficult they are to process. Moreover, they are not able to model the
evolution of the packing during com-
paction. Experiments are thus necessary.
5.3.2 Size Distribution
As an example, Fig. 5-11 shows that a
5.3.2.1 Compaction-Pressure Relationship polydisperse population (blend powder in
the figure) displays a steeper density/log
The choice of size distribution to be used
(pressure) relationship than a more nearly
in order to obtain the best green density
monodisperse one. However, to our
has often been dealt with in the literature
knowledge, the influence of size distribu-
(this problem is also encountered in soils
tion is often small compared to the influ-
mechanics, where it has been widely ad-
ence of mean size.
dressed). Hard sphere models have been
widely described (Yu and Standish, 1988;
5.3.2.2 Other Parameters
Ouchiyama and Tanaka, 1984); computer
modeling can also be used to predict the Unfortunately, to our knowledge there
density associated with any particular size is very little information about the influ-
distribution (Jullien and Meakin, 1990). ence of size distribution on the stress ratio

74
D A-152 SG Powder
0
Blend powder

68 -

.62-
Figure 5-11. Density versus
pressure for powders with
different maximum particle
56 - «S
packing densities (Zheng
and Reed, 1988). © The
American Ceramic Society;
50 i i i
reproduced by permission.
10u L J
* D ° ' ° 10 67
10'
Pressure (MPa) v
 5.3 Ungranulated (Unagglomerated) Powders 139

and tensile strength. If these factors de-

pend mainly on the number of contacts per
unit volume, then it can be expected that
they will not change very much with the
size distribution; Oger et al. (1986) have •BN

shown that the mean coordinate number

does not depend very much (for a given
density) on the size distribution.
100 1000
Pressure (MPa)
5.3.3 Particle Morphology
Figure 5-12. Density-pressure relationship for BN
Surprisingly, while it is possible to de- (plat particles) and zirconia (roughly spherical) pow-
scribe very accurately the particle mor- ders.
phology in terms of Fourier coefficients
(Beddow etal., 1977), or fractal dimen- ified by a cycling pressure. On the other
sions (Kaye, 1986) the literature does not hand, acicular or flat particles are able to
provide a complete and rigorous determi- provide very efficient packing (Fig. 5-7 a).
nation of its influence on the compaction The relative influence of these two fac-
step. tors (stable arched structures promoting
The morphology may be divided into porosity, and efficient packing promoting
two forms, which are very different in their density) depends on the particles' friction
theoretical and experimental conse- properties and aspect ratio. The result, as
quences: usual, is thus difficult to forecast. Flat BN
- "macroscopic" morphology, i.e., angu- particles are compared with equiaxed zir-
lar, flat, or acicular particles, versus conia particles in Fig. 5-12. The former is
spherical particles; and seen to be denser, owing to both morphol-
- "microscopic" morphology, i.e., surface ogy and friction properties.
roughness.
Influence on Particle Fracture
5.3.3.1 Angular, Flat, or Acicular Particles
Particle corners are expected to break
Influence on Density quite easily during compaction; this has
been observed for acicular particles by
Arches will be more easily constructed Leiser and Whittemore (1970), and for
with angular particles than with spherical sand during high-pressure compaction
particles. Thus the packing density of an- (Touati, 1982). However, the precise influ-
gular particles is expected to be lower than ence of this phenomenon on the mechani-
that of spherical particles. This has been cal behavior is unknown.
demonstrated with two-dimensional poly-
gons (Ammi et al., 1987): the greater the
Influence on Microstructure
size of these polygons, the greater the den-
sity. The arches are also stronger with an- The most important influence of particle
gular particles because the face-to-face morphology is probably its influence on
contacts are very stable. As a consequence, the microstructure. Flat particles are ex-
the slope of the pressure/density relation- pected to be oriented by the compaction
ship is lower, and the packing is not mod- pressure and this has several consequences:
140 5 Die Pressing and Isostatic Pressing

1) With flat particles, the pores are finer

and their size distribution is narrower than
with spherical particles (Yamagushi and
Kosha, 1981). Comparison between the in-
• BN
trusion and extrusion of mercury po-
rosimetry shows that the pores are "bottle-
like" for flat particles, while for acicular
particles they are cylindrical (Yamagushi 0 100 200
and Mian, 1991). These differences are im- Compaction pressure (MPa)
portant because it has been shown (Yama-
Figure 5-13. Radial springback against compaction
guchi and Kosha, 1981) that sintering is pressure for BN (flat particles) and zirconia (roughly
very much dependent on the pore mor- spherical) powders.
phology.
2) The radial/axial shrinkage ratio dur-
ing sintering is influenced by anisotropy.
Sintering of compacted samples of spheri- 5.3.3.2 Rough Versus Smooth Particles
cal powders leads to smaller shrinkage in
This point has not been addressed - to
the axial direction than in the radial direc-
our knowledge - experimentally, because
tion, as shown in Sect. 5.2.3. However,
it is difficult to synthesize two powders
with flat (i.e., platelet) particles (kaolin),
which are strictly identical except for their
the radial shrinkage is smaller (Table 5-2).
surface roughness. However, it is possible
This shows that the two contributions
to predict theoretically the influence of
due to packing anisotropy (namely, parti-
surface roughness on interaction forces
cle anisotropy and points of contact) have
(adhesion and friction). The most interest-
opposite influences on sintering.
ing articles are in our opinion those by
3) Flat powder particles slip easily over Adams et al. (1987) and Ross et al. (1991).
each other after compaction, resulting in a
significant springback upon ejection
Influence on Adhesion
(Fig. 5-13).
Rougher particles might be expected to
be less cohesive than smooth particles, be-
Table 5-2. Axial and radial shrinkages during sinter- cause the van der Waals force increases
ing of equiaxed and flat particle powders. Flat parti- with the radius of curvature of the surface,
cles: kaolin; green relative density 0.58; sintering tem- i.e., with the size of the asperities. A de-
perature 1300 °C, 2 h. Equiaxed powder: zirconia with
3% yttria; green relative density 51%; sintering scription of this problem is provided by
1500°C, 2h. Tabor (1987). However, rough particles
may increase their adhesion during com-
Axial Radial paction owing to the mechanical inter-
shrinkage shrinkage locking of asperities (Thompson, 1981b;
Bortzmeyer, 1992 a). The balance between
Equiaxed (zirconia) 16.6 19.5 these two phenomena is not easy to predict
and may depend on particle hardness
Flat particles (kaolin) 21.5 15.4
(Tabor, 1987).
 5.3 Ungranulated (Unagglomerated) Powders 141

Influence on the Friction Force 1988). This phenomenon is likely to de-

crease with increasing particle hardness.
Owing to their asperities, rough parti-
cles are expected to obey Coulomb's law
(i = /i(7), while the behavior of smooth 5.3.5 Aggregation
spherical particles is more complicated
The influence of aggregates (created for
(Adams et al., 1987). Smooth particles are
example during synthesis or calcination) is
expected to slip more easily on each other
very important in powder technology,
and on the mold walls, resulting in a
since they are detrimental to the sintered
greater stress ratio and a lower wall fric-
density and strength. As a consequence,
tion coefficient. Thus the resulting behav-
their influence on the pressure/density re-
ior of the fik product (which is important
lationship has been widely investigated.
for density variation prediction) is not easy
to predict; once again, experiments must
be carried out. Our own experiments 5.3.5.1 Density-Pressure Relationship
(Bortzmeyer, 1990) suggest that rough
particles are more easy to process than Since they are bigger and denser than
smooth ones, due to higher tensile strength the primary particles, aggregates are usual-
and smaller \ik product. This behavior al- ly more free-flowing and thus increase the
so depends on the particle hardness (Ross green density. The denser the aggregates,
e t a l , 1991). the denser the green sample, as shown in
Fig. 5-14 (Ciftcioglu et al., 1987).
5.3.4 Particle Hardness
5.3.5.2 Microstructure
Several studies describe the influence of
this parameter on different aspects of the An aggregated powder compact dis-
compaction behavior, mainly for pharma- plays a bimodal pore size distribution due
ceutical products, since particle deforma- to remnants of the original aggregates. The
tion makes an important contribution to inter-aggregate pores do not shrink easily
densification. The most important effect of during the sintering step, resulting in poor
particle hardness seems to be its influence sintered density (Fig. 5-14); thus the denser
on the stress ratio and the wall friction the aggregates, the less dense the sintered
coefficient. The softer the powder, the sample. Hence a high green density does
higher the stress ratio and the wall friction not necessarily result in a high sintered
coefficient (Carless and Leigh, 1974; Tan density, because it may be due to dense
and Newton, 1990; Strijbos, 1977). Softer aggregates. In the next section, dedicated
particles are expected to lead to steeper to the influence of spray-dried powders,
density/pressure relationships. However, this problem will be investigated further,
this is not so obvious since other factors including a measure of the aggregate's
influence this relationship (initial density, hardness.
etc.).
It has been shown that particles simulta-
5.3.5.3 Other Parameters
neously crush and reagglomerate during
compaction, leading to an increase then a To our knowledge, the influence of ag-
decrease in the specific surface area of a gregates on the stress ratio, friction behav-
green compact (Stanley-Wood and Sarrafi, ior, and tensile strength is unknown.
142 5 Die Pressing and Isostatic Pressing

Cf\
Figure 5-14. Pressure-den-
sity compaction curves for
powders A through F (i.e.,
£ 50- from loose and soft to hard
and tough aggregates).
Numbers in parentheses
.2 4 0 - are the sintered densities of
F{60.6) — pellets subsequently pressed
isostatically at 70 MPa and
30 air-sintered at 1400 °C for
s D(79.6)
2 h (Ciftcioglu et al. 1987).
BI85.0)
CI83.3) A(93.8)
© The American Ceramic
20 Society; reproduced by per-
10u 10' 10z
Pressure (MPa)

5.3.6 Mixing Different Powders low proportion of steel particles causes a

modification to the aluminum sphere
What happens when two different pow-
packing because of 'neighborhood' effects
ders are mixed together does not have a
near the steel spheres. In this case, the ma-
general answer. The following two exam-
trix density of the aluminum spheres is
ples shed a little light on the problem:
lower because of geometric effects (exclud-
1) Considering the green density/pres- ed volume around the steel spheres).
sure relationship, Gonthier (1984) showed
On the other hand, when a large number
that the density obtained for a given pres-
of steel spheres are mixed with aluminum
sure can be calculated as if the two pow-
spheres (i.e., above the percolation
ders were densified independently (Fig. 5-
threshold), the rigid network of these
15). This implies that each powder is inde-
spheres is able to resist the compaction
pendent of the other. This is true if the two
pressure. Thus for a given overall pressure,
mean diameters are not too different.
the matrix density of the aluminum
2) In the case of two metallic powders
spheres is lowered. This effect is purely me-
(aluminum and steel), Lange et al. (1991)
chanical, and during compaction the net-
showed that the inclusion of steel spheres
work of steel spheres bears a certain part
in a packing of aluminum spheres led to
of the load, thus reducing the effective
two different forms of behavior, depend-
stress on the aluminum particles.
ing on the proportion of steel spheres. A
5.3.7 Conclusions
This brief review shows that the influ-
ence of particle size on compaction behav-
ior is well-known. Obviously, the finer the
powder, the more difficult it is to process it
because of density effects (the initial densi-
ty is low), wall effects (higher friction and
Figure 5-15. The behavior of a mix of two powders A higher stress ratio), and tensile stresses
and B can (sometimes) be calculated as if the powders upon unloading (stress inhomogeneities,
were one on top of the other. entrapped air).
 5.4 Granulated (Agglomerated) Powders 143

The influence of other parameters (ag- methods have been developed in order to
gregates, morphology, roughness, hard- measure the granule strength. We will first
ness) on several areas of the mechanical briefly review these methods, then the in-
behavior has also been described. A com- fluence on powder processing will be ad-
plete description of their influence on the dressed. Note that the granule strength de-
overall mechanical behavior, is however, pends on three factors: the binder strength,
still lacking. the binder/particle interaction, and the
granule structure. The influence of the
granule structure (granule without binder)
5.4 Granulated (Agglomerated) on the compaction behavior will be ad-
dressed first.
Powders
Raw powders are not easy to process 5.4.1 Granule Strength
since, due to their high cohesion/weight
The granule strength is usually estimat-
ratio, they are not free-flowing and cannot
ed through the "break-point pressure", as
fill molds at a high rate. In order to over-
mentioned earlier (Sect. 5.21). But such a
come this problem, ceramic powders for
break-point is not always clearly distin-
compaction are usually spray-dried to
guishable, as shown by Brewer etal.
form granules whose size lies between 40
(1981), or by Dynys and Halloran (1983)
and 200 jLim.
in the case of aggregated powders.
This method has a severe drawback; if
The strength of granules or aggregates
the granules are too hard, the compaction
can also be measured by the evaluation of
pressure will not be able to destroy them.
the mean size in a slurry submitted to in-
The green microstructure will display the
creasing ultrasonic vibrations. Ciftcioglu
so-called "ghost" of the granules (i.e.,
etal. (1987) have shown that an equiva-
large intergranular pores), which is obvi-
lence can be found between the ultrasonic
ously seriously detrimental to the sintered
power and the compaction pressure. Thus
density (Fig. 5-16). This is why several
a correlation was found by these authors
between aggregate strength as measured
by the ultrasonic test, and the compaction
and sintering behavior. However, it is not
clear whether the compaction behavior is
always consistent with the results of ultra-
sonic measurement; the former involves
crushing the granules, while the latter in-
volves their comminution.
Both break-point pressure and ultrason-
ic resistance are indirect tests. A more
straightforward method is to measure di-
rectly one granule's crushing strength be-
tween two platens. It is necessary to apply
Figure 5-16. Fracture surface of a compact (10 MPa)
small loads (10 g), resulting in small dis-
of spray-dried zirconia powder (5% PVA, 2% PEG). placements (10 \xm); while difficult, this
The ghost of the granules is clearly seen. kind of measurement is not at all impossi-
144 5 Die Pressing and Isostatic Pressing

ble. Such devices have been described in a single constant: F=kA. This constant is
the literature (Kuno and Okada, 1982; determined by single granule testing.
Coupelle et al., 1991), and are even com- - However, in the packing there is a rela-
mercially available. The main problem tionship between the load (P) and the in-
comes from the handling of the granules; terparticle force (F) (such a law can be
while experiments on clay granules (1 mm derived theoretically, see Mehrabadi and
diameter) are quite easy, 50 jam granules Nemat-Nasser, 1982; Kanatani, 1981)
are far more difficult to handle. Figure 5-
D2P
17 shows such an apparatus. F= 71- (5-5)
The load/displacement curve of the Z8

granule can be interpreted within the where e is the porosity, D the diameter, and
elasto-plastic theory, which predicts a z the coordination number.
power-law curve (F= dn), where n lies be- - The relationships between porosity and
tween 1.5 (elastic behavior) and 1 (plastic coordination number, and between poro-
behavior). Many authors have described sity and contact area, are determined
such curves (Oberacker et al., 1988, for ex- through geometrical models
ample). Oberacker and co-workers were
able to detect the existence of a hard shell K D2(8-80)
(5-6)
around the larger granules, because the 12 (l-fi0)
strength/radius dependence changed from
where z = 12e, and s0 is the initial porosity.
an R2 law (volume-dependent strength) to
Combination of these equations with
a R1 law (surface-dependent strength).
F/A = k leads to
However, correlations between granule
tests and compaction behavior are not so (s-eo)s2
widely described. Baumard's model, relat- P=k (5-7)
(1 - e0)
ing granule strength (as measured by a
strength tester) and compaction behavior As a consequence, single granule experi-
is probably the most interesting model ments (giving k) are sufficient to predict
(Baumard et al., 1992): the compaction behavior. The model re-
- Owing to the pure plastic behavior of mains valid as long as the granules are not
these granules, the relationship between overly damaged, i.e., while the green densi-
force and contact area is characterized by ty is less than the initial granule density.

Differential
transformer
Counterbalance
Overflow
drain

Single granule

Figure 5-17. Diagram of an apparatus

to measure the strength and the lin-
Load transducer
ear shrinkage of a single granule
Recorder (from Kuno and Okada, 1982).
 5.4 Granulated (Agglomerated) Powders 145

(0%)

Figure 5-18. Pressure-den-

sity relationship, compari-
son between theory (lines:
Baumard's model) and ex-
periment (points); the num-
bers in parentheses are the
residual humidities. (Cour-
tesy of M. Baumard.)
10 15 20
Pressure (MPa)

Figure 5-18 shows that the theoretical pre- The size distribution has the same influ-
dictions are satisfactorily verified. ence as described in Sect. 5.3.2, which can
be calculated with an appropriate model
5.4.2 Influence of the Granule Parameters but is likely to be small compared with the
on the Compaction Behavior influence of granule density (see Sect.
5.4.2.2).
There are many results concerning the
However, the influence of granule size
influence of granule parameters (size, den-
on the microstructure is likely to be impor-
sity, microstructure) on the pressure/densi-
tant, since larger granules lead to larger
ty relationship, and moreover on the green
intergranular pores. These macropores
compact microstructure. This will be ex-
have an influence on the green strength
amined first. However, the influence of
since they act as strength-limiting flaws.
these parameters on the stress ratio, or on
Sintering is more or less able to correct the
the wall friction coefficient, is unknown. It
influence of granule size. For example, it
can be assumed, however, to be very small
has been reported that the strength and
compared to the binder influence, which
Weibull modulus (see Vol. 11, Chap. 10,
will be examined further.
Sect. 10.5.4.1 of this Series) of sintered
parts are not very dependent on the gran-
5.4.2.1 Granule Size
ule size (Fig. 5-19; Mosser etal., 1992).
Sufficiently large granules (about Moreover, Fig. 5-20 shows that the gran-
50 |im) are usually free-flowing and can be ule size does not affect the green density/
construed as cohesionless spheres. As a sintered density relationship.
consequence, the "classic" packing models
apply; the packing relative density of a
5.4.2.2 Granule Density
population of monosized free-flowing
granules should be about 0.60, whatever Since the relative density of the granule
their size, so that the influence of granule packing is nearly constant, the relative
size on green density is quite negligible. density of the green sample depends main-
146 5 Die Pressing and Isostatic Pressing

granules. The green compacts made from

dense granules are denser than those from
loose granules, but their density is always
smaller than the initial granule density.
This suggests that large intergranular
pores are probably not destroyed. Howev-
er, compacts made from loose granules are
denser than the initial granules, indicating
total destruction of the granules. As a con-
sequence, a high granule density is better
for the green density but not for the mi-
crostructure and the sintered density. This
result has already been mentioned in the
38-63 63-106 106-177
case of aggregates.
Pressed granule size
Figure 5-19. Weibull modulus of sintered compacts 5.4.2.3 Granule Structure
pressed from granules of differing size. The top of each
bar is the Weibull modulus of the green sample, while During the drying of a droplet in a
the two bottom sections (together) are the modulus of spray-drier, water migrates from the interi-
the fired (sintered) sample. Thus the Weibull modul of
the sintered samples are 17, 12, and 15 in increasing
or to the surface where it evaporates. Wa-
order of granule size. © The American Ceramic Soci- ter-soluble binder which is not adsorbed
ety; reproduced by permission. on the particles flows freely with the water,
and may form a hard shell on the surface
of the granule which is detrimental to the
ly on the granule density. Denser granules green microstructure and the green density
lead to denser packings, and usually to (Masters, 1979). As mentioned in Sect.
denser green compacts (Fig. 5-21). How- 5.4.1, this hard shell can be detected by a
ever, denser granules are also stronger and single granule strength measurement
do not deform easily. Figure 5-21 com- (Oberacker et al. 1988). It has been
pares the density/pressure relationships of claimed that this problem can be avoided
a powder granulated into dense and loose by using a polymer emulsion (i.e., latex)

<25fj.fr)
25/40
40/63
63/80
80/100
100/250 Figure 5-20. Relationship of
-250
green density to sintered den-
sity for granules of differing
5.6
2.9 3.0 3.1 3.2 3.3 3.4
Green density
 5.4 Granulated (Agglomerated) Powders 147

62
< 420 jum 46.3% Granule density
< 160 jum 46.3%
< 420 jum 61.5% Granule density
57- < 106 jum 61.5%

52-
* '
Figure 5-21. Density versus
pressure for a powder pre-
pared using two granule
47 -
densities (Zheng and Reed,
1988). © The American
Ceramic Society; repro-
duced by permission.
1OU
3 4 5 6' 2 3 4 5 6 7 1

Pressure (MPa)
V 10

instead of a water-soluble polymer (Ny- components are not used for their influ-
berg et al., 1988). ence on the compaction behavior, but for
process purposes (anti-foam).
The binder improves cohesion because
5.4.3 Binder Influence its large molecules are able to adsorb
The choice of binder depends on many simultaneously on different particles and
parameters, including compaction behav- to provide a network of entangled mole-
ior, pyrolysis behavior, and cost. This re- cules after drying (the rupture energy of
view will cover only the compaction be- the polymer is known to increase with the
havior. This influence is usually described molecular weight because of molecular en-
in terms of the pressure/density relation- tanglement). The plasticizer lowers the
ship, sometimes with the measurement of rigidity of the binder, because its molecules
the green sample strength and the radial prevent the functional groups of the binder
pressure coefficient. Care should be taken from becoming linked (H 2 O provides flex-
that the pressure/density relationship de- ibility between the OH groups of PVA).
pends not only on the binder characteris- The lubricant lowers the friction coeffi-
tics, but also on the granule density. This cient because of the low adhesion energy of
density will depend on the binder system its molecules (e.g., stearic acid bearing sev-
and also on the spray-dryer geometry; un- eral CH 2 groups).
fortunately this is difficult to allow for. In In such a slurry, it is very difficult to
fact, binder comparisons should be made assess exactly the influence of each compo-
with a constant granule density, which is nent. Moreover, the consequences of their
obviously difficult to realize. interaction are largely unknown. Accord-
A binder system usually contains several ingly, fundamental studies are mainly con-
components, each of them playing a (sup- cerned with simple systems involving only
posedly) precise role: the binder itself (e.g., one or two components, aiming to deter-
PVA), the plasticizer (e.g., PEG), and the mine the influence of a given parameter
lubricant (e.g., stearic acid). Some of these (r g , Mw) or of a given organic function.
148 5 Die Pressing and Isostatic Pressing

5.4.3.1 Powder + Binder Cohesion since bridging of the particles is promoted,

while in the other case the two polar
One of the most important characteris-
groups can be adsorbed on a single parti-
tic of a binder is its cohesion in a particle/
cle. This last example shows that several
binder system, because this will in turn
factors may influence the powder cohe-
affect the entire mechanical behavior.
sion. Careful examination is necessary in
Thus it is worth recalling a few results con-
each case.
cerning this characteristic.
The powder + binder cohesion depends
not only on the binder strength, but also
5.4.3.2 Influence of Molecular Weight
on the binder/particle adhesion. The for-
mer depends on several factors, including Influence on Green Strength
the molecular weight of the binder. The
In polymer technology, polymer strength
latter depends on the chemical nature of
is strongly related to molecular entangle-
the binder (the adhesion behavior of a par-
ment (see classical adhesion models), so
ticle/polymer system is governed roughly
that an increase in molecular weight (Mw)
by the same factors as the adsorption be-
increases the polymer strength. It also usu-
havior). These two parameters may vary
ally increases the particle/binder adhesion.
independently so that an increase in binder
Thus an increase in M w increases the
strength can be balanced by a decrease in
binder + powder cohesion.
particle/binder adhesion, resulting in a de-
crease in powder cohesion.
Influence on Pressure-Density Relation-
Moreover, other factors may have an
ship and Microstructure
influence on the powder cohesion. The fol-
lowing example has been described by This higher strength will of course pre-
Frisch and Thiele (1987; Fig. 5-22); the vent granule destruction. This has two
powder has been mixed with either suc- consequences. First, the green microstruc-
cinic acid or fumaric acid. These two or- ture will contain more intergranular
ganic acids (each with four carbons) have macropores. Second, the density-pressure
a carboxyl group at each end of the mole- relationship may be expected to display a
cule, but in one case the molecule is rigid higher break-point pressure, or, in other
since a double bond between the central words, to be shifted towards lower densi-
carbon atoms prevents free rotation. In ties. Both of these consequences are detri-
that case, the powder cohesion is better mental to the sintered density.

H2
/SB
CH2 \y CH 2 = C H 2 -C
V
CH 2 CH 2 HO OH

i Figure 5-22. Adsorption of

fumaric and succinic acid
on ceramic particles (Frisch
Succinic: low cohesion Fumaric: high cohesion and Thiele, 1987).
 5.4 Granulated (Agglomerated) Powders 149

Influence on Wall Friction Coefficient ule density is kept constant, then the hard-
and Stress Ratio er the binder, the lower the green density.
However, softer granules may give a lower
Briscoe and Evans (1991) showed that a
density if this softness comes from a low
higher molecular weight increases the wall
granule density. Thus several granule
friction. This is likely to be the result of
parameters (strength and density, at least)
greater adhesion between the wall and the
can have different influences on the com-
particles. These authors also showed that
pact density.
the higher M w , the smaller the stress ratio.
The following example shows the influ-
Once again, this is a result of the adhesion
ence of water on PVA/PEG binder. PEG
properties: the higher Mw, the greater the
or water is a good plasticizer for PVA
cohesion between particles. The particles
(polyvinyl alcohol); these systems have
do not easily flow and rearrange, thus low-
been widely described in the literature (the
ering the pressure transmitted to the walls.
higher the RH degree or the PEG content,
The important factor for density varia-
the lower the Tg: Brewer et al., 1981; Nies
tions is the fik product. Since \i increases
and Messing, 1984). As was stated above,
and k decreases with M w , the overall influ-
the lower Tg, the smaller the break-point
ence of the molecular weight on density
pressure and the higher the green density
variations is not self-evident. Low molecu-
(Fig. 5-23).
lar weight, is, however, probably prefer-
able since a higher stress ratio ensures bet- These authors also showed that the low-
ter filling of the mold corners. er the Tg, the lower the strength for a given
density. However, since the density is
These results explain why most industri-
higher for a given pressure, the strength-
al binder systems contain polymers of dif-
pressure relationship may be of greater im-
ferent molecular weights: small M w en-
portance for a lower Tg (Nies and Messing,
sures fluidity, high M w increases the green
1984). Due to the lower strength, it is likely
sample strength.
that the wall friction coefficient is lower,
and the stress ratio higher, if Tg is lower.
5.4.3.3 Influence of the Glass Transition
However, to our knowledge these results
Temperature
have not been reported.
Tg is the glass transition temperature, Our conclusion is the same as for Mw:
i.e., the temperature separating ductile and despite the decrease in green strength, a
brittle behavior of the binder. The influ- lower Tg is usually preferable for com-
ence of !Tg has been widely addressed be- paction shape forming.
cause the main effect of a plasticizer is to
5.4.3.4 Latexes
lower the Tg of the binder.
If the pressing temperature is below Tg9 Water-soluble binders such as PVA or
the binder is crystalline and brittle. On the PEG may be detrimental to the granule
other hand, if it is above Tg, the binder is structure, since they may flow with the
amorphous and soft. In the latter case, the evaporating water towards the outer sur-
granules are soft, and the particles are able face of the granules and form a shell of
to flow more freely. Thus the influence of hardened polymer. Latex (i.e., polymer
Tg is about the same as the influence of emulsion) is believed to avoid this problem
Mw: the higher it is, the harder the binder, because of the relatively large size of the
the lower the green density, etc. If the gran- particles.
150 5 Die Pressing and Isostatic Pressing

IUU

o92 % RH
600 - 0 75 % •
A
52 % • A-16-6 2.3%PVA in ,.
500 - • 35 %
D15 < r g ^
% • •
Figure 5-23. Compaction
ensit

r diagram for A-16-6 (alu-

400-
mina) samples. The figure
shows the evolution of the
itive

I
300 - break point with humidity
IU
"QJ
(Di Milia and Reed, 1983).
200 - © The American Ceramic
! AYP Society; reproduced by per-
100 - —i I I 1

10" 10D 10 10' 10 c

Applied pressure (Pa)

Experiments with latex binders are de- 5.6 References

scribed by Nyberg et al. (1988), and by
Gurak et al. (1987). Indeed, the granule Adams, M. J., Briscoe, B. X, Pope, L. (1987), in:
Tribology in Paniculate Technology: Briscoe, B. J.,
structure seems to be improved (no inter- Adams, M. J. (Eds.). Bristol: Adam Hilger, pp. 8-
granular holes). However, direct compari- 12.
son is difficult since the best polymer (for Adams, M. J., Williams, D., Williams, J. G. (1989) J.
Mater. Sci. 24, 1772-1776.
latex binders) is not necessarily the best in Al-Jewaree, H. A. M., Chandler, H. W. (1990), Br.
water-soluble binder applications. Ceram. Trans. J. 89, 207-210.
Ammi, M., Bideau, D., Troadec, J. P. (1987), J. Phys.
D: Appl. Phys. 20, 424-428.
Baumard, J. R, Coupelle, P., Destermes, J. (1992), in:
Journees Annuelles du GFC, Paris, 12-13 February,
1992. Abstracts: Paris: Groupe Francais de la Ce-
5.5 Conclusions ramique, pp. 408-412.
Beddow, J. K., Philip, G. C , Vetter, A. E, Nasta, M.
Ceramists have often focused their stud- D. (1977), Powder Technol. 18, 19-25.
Bortzmeyer, D. (1990), Dissertation, Ecole des Mines
ies on the pressure/density relationship. de Paris, France.
Thus the influence of most of the powder Bortzmeyer, D. (1992a), /. Mater. Sci. 27, 3305-
parameters on this relationship is known. 3308.
Bortzmeyer, D. (1992 b), Powder Technol. 70, 131-
On the other hand, among the powder 139.
parameters, the particle mean size is the Bortzmeyer, D., Abouaf, M. (1989), in: Proc. 1st Int.
easiest to change. Its influence on the stress Conf. on Micromechanics of Granular Media, Cler-
mont-Ferrand, France, 4-8 Sept., 1989: Biarez, I,
ratio and the wall friction coefficient is Gourves, R. (Eds.). Rotterdam: Balkema, pp. 279-
well-known. However, for other parame- 286.
ters, a complete knowledge of their influ- Bortzmeyer, D., Langguth, G., Orange, G. (1993), J.
ence on the whole mechanical behavior is Europ. Ceram. Soc. 11, 9-16.
Brewer, J. A., Moore, R. H., Reed, J. S. (1981), Am.
still lacking. Since this knowledge is neces- Ceram. Soc. Bull. 60, (2), 212.
sary for a good prediction of the com- Briscoe, B. I, Evans, P. D. (1991), Powder Technol.
paction behavior such studies will certainly 65, 7-20.
Brown, S. B., Weber, G. G. A. (1988), Mod. Devel.
be developed in the future. Powder Metall. 18, 465-476.
 5.6 References 151

Carless, J. E., Leigh, S. (1974), J. Pharm. Pharmacol. Niesz, D. R, Bennett, R. B., Snyder, M. J. (1972), Am.
26, 289-297. Ceram. Soc. Bull. 51, 677-680.
Carless, J. E., Sheak, A. (1976), /. Pharm. Pharmacol. Nyberg, B., Carlstrom, E., Persson, M., Carlsson, R.
28, 17-22. (1988), in: Proc. 2nd Int. Conf. on Ceramic Powder
Chandler, H. W. (1985), J. Mech. Phys. Solids 33, (3), Processing Science, Oct. 12-14, 1988, Berchtes-
215-226. gaden (Germany). Abstract.
Chandler, H. W. (1990), Int. J. Eng. Sci. 28, (8), 719- Oberacker, R., Ottenstein, A., Thummler, R (1988),
734. in: Proc. 2nd Int. Conf. on Ceramic Powder Process-
Ciftcioglu, M., Akinc, M., Burkhart, L. (1987), /. Am. ing Science, Oct. 12-14,1988, Berchtesgaden (Ger-
Ceram. Soc. 70, (11), C329-C334. many). Abstract.
Coupelle, P., Destermes, J., Miglioretti, R, Baumard, Oda, M. (1979), in: Proc. US/'Japan Seminar on Con-
J. F. (1991), Ind. Ceram. (Paris) 861, 408-412. tinuum Mechanics and Statistical Approaches in the
Di Milia, R. A., Reed, J. S. (1983), J. Am. Ceram. Soc. Mechanics of Granular Materials (Sendai, Japan,
66, (9), 667-672. 5-9 June, 1978): Cowin, S. C , Satake, M. (Eds.).
Dynys, R W, Halloran, J. W. (1983), J. Am. Ceram. Tokyo: Gakujutsu Bunken Pukyu-Kai, pp. 7-26.
Soc. 66, (9), 655-659. Oda, M., Konishi, X (1974), Soils Foundation 14, (4),
Frisch, B., Thiele, W. R. (1987), J. Adhesion 22, 81-95. 25-38.
Gonthier, Y. (1984), Dissertation, Universite Scienti- Oger, L., Troadec, J. P., Bideau, D., Dodds, X A.,
flque et Medicale de Grenoble, France. Powell, M. X (1986), Powder Technol. 46, 121-131.
Groot Zevert, W. R M., Winnubst, A. J. A., Theunis- Onoda, X (1976), J. Am. Ceram. Soc. 59, 236-239.
sen, G. S. A. M., Burggraaf, A. J. (1990), J. Mater. Ouchiyama, N., Tanaka, T. (1984), Ind. Eng. Chem.
Sci. 25, 3449-3455. Fundam. 23, 490-493.
Gudehus, G. (1977), Finite Elements in Geomechanics. Ross, X D. X, Pollock, H. M., Guo, Q. (1991), Powder
New York: Wiley. Technol. 65, 21-35.
Gurak, N. R., Josty, P. L., Thompson, R. J. (1987), Schubert, H. (1975), Powder Technol. 11, 107-119.
Am. Ceram. Soc. Bull. 66, 1495-1497. Stanley-Wood, N., Sarrafi, M. (1988), Part. Part.
Hahn, C. (1986), J. de Phys. Cl, 2, (47), C1.91-C1.96. Syst. Charact. 5, 186-192.
Harvey, J. W., Johnson, D. W. (1980), Am. Ceram. Strijbos, S. (1977), Powder Technol 18, 209-214.
Soc. Bull. 59, (6), 637-645. Strijbos, S., Rankin, P. X, Klein Wassink, R. X, Ban-
Johnson, K. L., Kendall, K., Roberts, A. D. (1971), nink, X, Oudemans, G. X (1977), Powder Technol.
Proc. R. Soc. London A 324, 301-313. 18, 187-200.
Jullien, R., Meakin, P. (1990), Nature 344, 425-427. Suzuki, M., Oshima, T. (1983), Powder Technol. 36,
Kanatani, K. I. (1981) Powder Technol. 28, 167-172. 181-188.
Kaye, B. H. (1986), Powder Technol. 46, 245-254. Tabor, D. (1987), in: Tribology in Paniculate Technol-
Kendall, K. (1986), Lett. Nature 319, 203-205. ogy: Briscoe, B. X, Adams, M. X (Eds.). Bristol:
Kendall, K., MacAlford, N., Birchall, J. D. (1986), Adam Hilger, pp. 206-219.
Spec. Ceram. 8, 255-265. Tan, S. B., Newton, X M. (1990), Int. J. Pharmaceutics
Krycer, I., Pope, D. G., Hersey, J. A. (1982), Powder 64, 227-234.
Technol. 33, 101-111. Thomas, G., Missiaen, X M., Rouille, L. (1989), in:
Kuhn, L., Mac Meeking, R. M., Lange, R F. (1989), Proc. 1st Int. Conf. on Micromechanics of Granular
in: Proc. 1st Int. Conf. on Micromechanics of Gran- Media, Clermont-Ferrand, France 4-8 Sept., 1989;
ular Media, Clermont-Ferrand, France, 4-8 Sept., Biarez, X, Gourves, R. (Eds.). Rotterdam: Balke-
1989: Biarez, X, Gourves, R. (Eds.). Rotterdam: ma, pp. 99-104.
Balkema, pp. 331-338. Thompson, R. A. (1981 a), Am. Ceram. Soc. Bull 60,
Kuno, H., Okada, J. (1982), Powder Technol. 33, 7 3 - 244-247.
79. Thompson, R. A. (1981 b), Am. Ceram. Soc. Bui. 60,
Lange, F. R, Atteraas, L., Zok, R (1991), Acta 248-251.
Metall. Mater. 39, (2), 209-219. Touati, A. (1982), Dissertation, Ecole des Ponts et
Leiser, D. B., Whittemore, O. J. (1970), Am. Ceram. Chaussees, Paris, France.
Soc. Bull. 49, 114-711. Yamaguchi, T, Kosha, H. (1981), /. Am. Ceram. Soc.
Lukasiewicz, S. J., Reed, J. S. (1978), Am. Ceram. Soc. 64, (5) C.84-C.85.
Bull. 57, 798-801. Yamagushi, T, Mian, G. (1991), J. Am. Ceram. Soc.
Masters, K. (1979), Spray Drying Handbook, 4th ed. 74, 1955-1958.
New York: Longman Scientific & Technical. Yen, K. Z. Y, Chaki, T. K. (1992), J. Appl Phys. 71,
Mehrabadi, M. M., Nemat-Nasser, S. (1982), Int. J. 3164-3173.
Num. Anal. Methods Geomechanics 6, 95-108. Yu, A. B., Standish, N. (1988), Powder Technol. 55,
Mosser, B. D., Reed, J. S., Varner, J. R. (1992), Am. 171-186.
Ceram. Soc. Bull. 71, (1), 105-109. Zheng, X, Reed, X S. (1988), / Am. Ceram. Soc. 71,
Nies, C. W., Messing, G. L. (1984), /. Am. Ceram. (11), C.456-C.458.
Soc. 67, 301-304.
152 5 Die Pressing and Isostatic Pressing

General Reading
deWith, G., Terpstra, R. A., Metselaar, R. (Eds.)
Capus, J. M., German, R. M. (Eds.) (1992), Advances (1989), Processing of Ceramics. London: Elsevier.
in Powder Metallurgy and Paniculate Materials. Thornton, C. (Ed.) (1993), Powders and Grains 93,
Princeton: APMI. Rotterdam: Balkema.
6 Slip-Casting and Filter-Pressing
Robert Fries and Brian Rand

School of Materials, The University of Leeds, Leeds, U.K.

List of Symbols and Abbreviations 155

6.1 Introduction 157
6.1.1 The Process 158
6.1.2 Requirements 159
6.2 Colloidal Stabilization Mechanisms for Ceramic Slip Systems 160
6.2.1 Electrostatic Stabilization 160
6.2.1.1 The Electrical Double Layer 160
6.2.1.2 The Zeta Potential 160
6.2.1.3 The Total Potential Energy Curve 161
6.2.1.4 Specific Adsorption 162
6.2.2 Polymeric Stabilization 163
6.2.3 Solvation Forces 164
6.3 Slip Structure and Rheology 165
6.3.1 General Background 165
6.3.1.1 Steady Shear Rheology 166
6.3.1.2 Viscoelastic Properties 166
6.3.2 Hard-Sphere Repulsion 168
6.3.3 Soft-Sphere Repulsion 169
6.3.4 Attractive Systems 170
6.3.4.1 Weakly Flocculated Systems 171
6.3.4.2 Strongly Flocculated Systems 171
6.4 Mechanism and Kinetics of Slip-Casting and Filtration 172
6.4.1 General Kinetics 172
6.4.2 Effects of Solids Concentration and Particle Size Distribution
on Cast Structure 176
6.5 Control of Interparticle Forces, Rheology and Cast Structure 177
6.5.1 Deflocculation 177
6.5.1.1 Advanced Ceramic Systems 177
6.5.1.2 Clay-Based Ceramic Systems 179
6.5.2 Interparticle Forces and the Control of Cast Structure 180
6.5.2.1 Clay-Based Ceramic Systems 181
6.5.2.2 Advanced Ceramic Systems 181
6.6 Defects and Microstructural Nonuniformities 183
6.6.1 Pinholes 183

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
154 6 Slip-Casting and Filter-Pressing

6.6.2 Preferred Orientation of Anisometric Particles 184

6.6.3 Segregation 184
6.6.4 Slip-Meets and Cast-Wreathing 184
6.7 Outlook 185
6.8 References 185
 List of Symbols and Abbreviations 155

List of Symbols and Abbreviations

A Hamaker constant
a particle radius
AF cross sectional area of the filter/cast
a max maximum particle radius in the size distribution
a min minimum particle radius in the size distribution
CFP floe structure indicator
D interparticle distance
F(a) cumulative volume fraction of particles smaller than radius a
FA van der Waals force
G storage modulus
G" loss modulus
G* complex shear modulus
j complex number: ^-1
k Boltzmann constant
K hydraulic tortuosity
Lc cast thickness
Lm wetted depth of the mold
m Weibull modulus
n0 counter ion concentration
AP pressure drop across the cast or filter cake
Po pressure at t h e interface between t h e d r y a n d wetted part of the mold
Pe Peclet n u m b e r
Pi hydraulic pressure at the interface between mold a n d consolidated layer
PT hydraulic pressure at the interface between slip a n d consolidated layer
RF resistance of the filter
SY surface area per unit volume
T absolute temperature
t time
V total potential energy
VA potential energy of attraction
Fmax energy barrier against flocculation
Vmin attractive primary potential energy minimum
VR potential energy of repulsion
F sec attractive secondary potential energy minimum
x distribution modulus
z counter ion valency
Z volume of filtrate removed from the slip in forming unit volume of cast
Z ave average volume of filtrate removed from the slip in forming unit volume of
compressible cast
occ specific cast resistance
a
c(ave) average specific cast resistance of compressible casts
am specific mold resistance
156 6 Slip-Casting and Filter-Pressing

y shear rate
S phase angle
s permittivity
sc cast porosity
em mold porosity
C zeta potential
r\ suspension viscosity
t]0 suspending m e d i u m viscosity
r]p plastic viscosity
rjr relative viscosity
[r]] intrinsic viscosity
x ~x Debye length
ji scaling constant
T shear stress
iy Bingham yield stress
0 volume fraction solids in the slip
cf)c volume fraction solids in the cast
^ccave) average volume fraction solids in compressible casts
4>crii critical volume fraction solids for network formation in a flocculated slip
(/>eff effective volume fraction of electrostatically stabilized soft spheres in the slip
(/>F volume fraction of floes in the slip
(j)m maximum packing volume fraction solids
Wo surface potential
*F6 Stern layer potential
DLVO Derjaguin, Landau, Verwey and Overbeek
IEP isoelectric point: pH of zero zeta potential
PAA poly(acrylic acid)
PVA poly(vinyl alcohol)
PZC point of zero charge: pH of zero net charge
 6.1 Introduction 157

6.1 Introduction nomic to bear the tooling costs of the pro-

duction of metal molds for pressing opera-
The fabrication of most ceramic articles tions. The use of colloidal systems, how-
utilizes the technique of sintering to den- ever, offers very significant additional ad-
sify a shaped porous compact of ceramic vantages in the processing of advanced ce-
particles. It is well established that the pro- ramics where homogeneity of structure in
gress of the sintering step is strongly influ- the green body is of paramount impor-
enced by the compositional and structural tance to eliminate defects in the final ce-
characteristics of the porous compact ramic that may limit its ultimate perfor-
(Lange, 1984, 1989; Lange et al., 1986; mance. For example, the strength limiting
Kendall, 1988). The compact, or green flaws in engineering ceramics may arise
body, should be as uniform in composition from nonuniform consolidation of the par-
and structure as is feasible within the con- ticles such that local density variations
straints of the industrial fabrication proce- lead to differential sintering densification
dure selected. This, of course, will be deter- which can lead to large voids in the final
mined by the nature of the ceramic to be ceramic (Lange, 1989). Figure 6-1 shows
fabricated, the economics of the produc- Weibull plots of the strength variation in
tion process and the engineering properties sintered Si 3 N 4 ceramics consolidated by a
required from the product. Slip-casting variety of techniques (Pasto etal., 1984).
and filter-pressing are related forming pro- Clearly, slip-casting gives a high value of
cesses based on colloidal systems, in which the Weibull modulus, i.e., a low spread in
dewatering is used to consolidate particles strength variation, indicating a narrow dis-
suspended in a slip. tribution of flaw sizes. This system also
Slip-casting is an old established tech- results in high mean strength, showing that
nique involving the use of shaped porous the flaws are generally small when this
molds, widely adopted in the whitewares
industry to produce hollowware, whilst fil-
ter-pressing has long been used as a tech-
nique for the preparation of plastic bodies, 0.95
also in the whitewares industry. However, 0.90
0.70
in recent years filter-pressing with con-
0.50
toured porous molds has become an estab-
lished shaping operation in its own right, 0.20
especially in the automated production of £ 0.10
clay-based ceramics. It offers the oppor-
tunity for achieving a higher green density
in the product than can be obtained by
slip-casting. 0.01
Slip-casting usually employs relatively 500 700 900 1100
cheap plaster of Paris molds and thus it is Flexural strength (MPa)

often selected as a forming method in the Figure 6-1. Flexural strength and Weibull moduli of
production of complicated shapes that Si 3 N 4 ceramics fabricated using different consolida-
tion techniques: (o) isopressed and sintered, (•) injec-
cannot be achieved by simple pressing or tion-molded and sintered, (•) slip-cast and sintered,
by injection-molding and for short runs of (•) injection-molded and HIP (after Pasto et al.,
components for which it would be uneco- 1984).
158 6 Slip-Casting and Filter-Pressing

consolidation technique is used; approach- Slip-casting is the most widespread of

ing the state obtained by hot isostatic these techniques for the final shaping oper-
pressing. Even higher densities are feasible ation. There are two modifications, drain-
from pressure filtration which make this casting and solid-casting, the former being
an important technique for future process- more common. In drain-casting, the mold
ing of high-performance materials. Cen- is filled with slurry, as depicted in Figure
trifugal-casting is another dewatering tech- 6-2. The thickness of cast increases pro-
nique that has potential for advanced ce- gressively with time as the fluid is with-
ramic fabrication. Many of the factors drawn, until the required thickness is at-
controlling the processes of slip-casting tained. Then the mold is inverted, drained
and pressure-filtration are relevant to this of slurry, and allowed to dry, partially. In
latter process also. the initial stages of drying, more fluid is
withdrawn into the mold and then evapo-
6.1.1 The Process ration from the exposed surface of the cast
In filter-pressing, pressure is applied to a leads to further consolidation which, ide-
slurry to force the fluid within which the ally, causes it to shrink away from the
particles are dispersed out through a po- mold and be released. The cast can then be
rous filter, so forming a filter cake which fully dried and fired. For complex shapes,
takes the shape of the filter surface. Slip- the mold usually comprises a number of
casting is a modification of this in which interlocking pieces which can be disman-
the dispersion fluid is drawn out from the tled to remove the product.
slurry into a shaped porous mold by capil- Solid-casting differs from drain-casting
lary action of the mold. A development is in that the slip is allowed to continually
that of pressure-slip-casting, in which pres- flow into the relevant part of the mold as
sure is applied to the slip in addition to the the dispersion fluid is withdrawn, until the
capillary suction from the mold itself to cast fills the cavity. The slip requirements
speed up the process. may be different in the two cases.

(a) (b)

(c)

Figure 6-2. Drain casting: (a) a porous

mold is (b) filled with a ceramic slip;
(c) excess slip is drained after a period
of casting leaving a plastic body (d) which
can be trimmed (after Reed, 1988).
 6.1 Introduction 159

6.1.2 Requirements posed perturbations, such as shear stress,

gravitational field, electrical field, hydro-
The following are ideal requirements of
dynamic effects, thermal gradients, and
the slip-casting and filter-pressing opera-
Brownian effects.
tions. The cast should have:
The size distribution and shapes of the
• homogeneous distribution of particles particles vary enormously in casting/filtra-
within the cast/cake, i.e., uniform distri- tion slips used in different branches of the
bution of chemical/mineralogical con- ceramic industry. In all casting slips it is
stituents, particle size and density on a essential to have a significant proportion
scale of size relevant to the product in of particles in the colloidal range of size to
question; control the rheology and sedimentation
• green density as high as possible; behavior. The powders used in advanced
• easy release from the mold; ceramics processing, however, are gener-
• high strength in the compacted state; ally all in this size range and tend to have
• good surface finish; a quite narrow size distribution to ensure
• good handling characteristics after de- rapid and uniform sintering leading to a
molding; uniform grain size in the final product. In
• uniform shrinkage on drying and firing the traditional area there is a mixture of
(preferably small). particle sizes ranging from the colloidal
clays, which are of plate-like character,
The slip is required to provide these often with large aspect ratios, to the silica
characteristics, but in addition should fillers and feldspathic fluxes which can be
have the following characteristics of its in the form of particles tens of microns in
own: size. There are also ceramic formulations
• ability to flow into and fill all the re- in which precalcined polycrystalline mate-
cesses in the mold; rial may be added to control the firing
• a casting rate as high as possible, to shrinkage. Such material may also be in
facilitate economic production. the tens of micron size range.
The particle size distribution exerts a
The attainment of these characteristics strong influence on the packing character-
in a controlled manner during a produc- istics during cast formation and thus will
tion process requires understanding and influence the properties of the green body.
control of the factors which determine the The other main factor is the nature of the
properties of slips and of the way such slips particle interactions. These control the
are consolidated during formation of the flocculation behavior of the particles in
cast/cake. It is the judicious choice of par- suspension which have a profound effect
ticle size distribution and control of parti- on the flow properties. The formation of
cle surface chemistry that leads to the suc- floes takes place when attractive forces be-
cessful use of these fabrication steps. tween particles are dominant. Floes link
Structure in both of these systems is up to form aggregated interconnected
characterized by the spatial distribution of structures which must be broken down to
particles and the nature of the interparticle allow the suspension to flow. A defloccu-
linkages which are controlled by particle lated slip is one in which repulsive forces
size distribution, solids volume fraction, between particles are the dominant inter-
particle shape, interparticle forces, and im- action. However, the relative magnitudes
160 6 Slip-Casting and Filter-Pressing

of attractive and repulsive forces can be particle surfaces. The subject has been
varied continuously by control of the extensively reviewed (e.g., Hunter, 1987;
chemical composition of the suspension, Kruyt, 1952). The surface electrical charge
allowing a wide range of behavior and density and the surface potential arise
casting characteristics. from the adsorption of 'potential deter-
Amongst the list of requirements given mining' ions, which are H + and OH~ for
above, some are favored by a state of floc- oxide-water systems (see Sec. 6.5.1.1).
culation, others require the slip to be in the
deflocculated state. Usually a balance has 6.2.1.1 The Electrical Double Layer
to be struck between opposing require-
ments and the casting slip is in a state of The electrical double layer consists of an
partial deflocculation. It is the structure in inner part, also called the Stern or
slip and cast that is paramount and so it is Helmholtz layer, the thickness of which is
essential to understand and control the determined by the effective radius of ad-
factors that determine this structure. sorbed ions, and an outer, diffuse part, in
which the ions are distributed according to
the balance of electrostatic forces and ther-
6.2 Colloidal mal motion (Fig. 6-3). The inverse of the
Stabilization Mechanisms Debye constant, x9 is a measure of the
for Ceramic Slip Systems thickness of the double layer,

8 7i
The most common colloidal stabiliza- x = (6-1)
tion mechanisms may be broadly catego- skT
rized into two main groups: in a polar, where n0 and z are, respectively, the coun-
e.g., aqueous, environment electrostatic ter ion concentration and valency, e the
phenomena predominate, whereas in non- electron charge, e the permittivity of the
polar media, such as most organic sol- double layer, k the Boltzmann constant
vents, often steric mechanisms are em- and Tthe absolute temperature. It follows
ployed. Electrostatic repulsion is devel- from this that the thickness of the double
oped by controlling electrical double layers layer decreases with increasing electrolyte
at the particle-solution interface, typi- concentration and valency, that is, it is
cally, achieved by adjusting the pH or by controlled entirely by the ionic strength.
adding specific ions which adsorb onto the
surfaces. Polymeric stabilization, in con-
6.2.1.2 The Zeta Potential
trast, originates from the interaction be-
tween polymeric additives adsorbed onto The electrostatic potential at the Stern
the particle surfaces which keep neighbor- layer cannot be readily obtained from
ing particles at a distance where attractive measurements, but the so-called zeta po-
van der Waals forces are ineffective. tential, C, can be determined by electroki-
netic measurements. C, also referred to as
the slipping plane potential, is assumed to
6.2.1 Electrostatic Stabilization
be located slightly further away from the
Electrostatic stabilization originates surface than the Stern potential (Fig. 6-3).
from the interaction between electrical A wide range of techniques is available
double layers that develop around charged (Hunter, 1981). The zeta potential is an
 6.2 Colloidal Stabilization Mechanisms for Ceramic Slip Systems 161

Figure 6-3. Schematic diagram showing

the variation in electrical potential with
distance from the particle surface: Wo is
the surface potential, *F5 the potential
at the Stern layer, f the zeta potential
and 1/x a measure of the electrical
double layer thickness.

important parameter in colloidal stability, two spheres as

since it reflects the variation in surface po-
tential for a specific material in a polar FA =- (6-2)
medium and the adsorption of ions into 12D2
the inner part of the double layer. £ de-
It can be seen that the magnitude of at-
creases with increasing ionic strength pro-
traction strongly depends on the interpar-
vided that the ionic species are of indiffer-
ticle distance, D, and also the Hamaker
ent character (i.e., adsorption determined
constant, A, and the particle radius, a
by electrostatic interaction).
(Horn, 1990). Both, the electrostatic and
van der Waals forces can be large com-
6.2.1.3 The Total Potential Energy Curve
pared to the diffusive force and can domi-
The behavior of slips is controlled by nate the behavior of the colloidal disper-
many-body interactions, that is, the forces sion. Thus, the interaction between parti-
acting between particles are determined by cles is essentially controlled by the range
a summation of all the interparticle forces and magnitude of the double layer repul-
between the particles in the slip. Theory sion, which can be manipulated for ceram-
does not, however, extend beyond the ics by varying, the ionic strength, composi-
summation of the forces acting between tion and pH of the solution. These control
pairs of particles, which only approxi- the magnitudes of the Debye length, 1/%,
mates trends in the behavior of concen- and the zeta potential. The van der Waals
trated suspensions. However, this limited forces, in contrast, are relatively insensitive
approach has provided a powerful insight to changes in the composition of the dis-
into the stabilization mechanisms of col- persion fluid. Figure 6-4 shows typical en-
loidal systems. The net-force acting be- ergy curves for two different electrolyte
tween two particles is determined by two concentrations at constant high surface
components, namely the electrical double potential. The total potential energy, K is
layer repulsion and the van der Waals at- obtained by the summation of the poten-
tractive force, FA, which acts between any tial energy of attraction, VA, and the poten-
kinds of particles and may be written for tial energy of double layer repulsion, VR.
162 6 Slip-Casting and Filter-Pressing

(a) (b)
Figure 6-4. (a) Typical potential
energy curve showing the pri-
mary minimum, Vmin, the energy
barrier against flocculation,
I Vmax and the secondary energy
minimum Vsec; (b) potential
-
eg
0 energy curves, V1 and V2, ob-
B
T Distance tained by the summation of the
o repulsive energy terms, VR1 and
VR2, and the attractive energy
VA showing the effect of indiffer-
ent electrolyte: At high elec-
trolyte concentration (curve V2)
Vmax becomes small, leading to
an unstable slip.

When the surface potentials are high and of interparticle distance and electrolyte
the ionic strength is low (high zeta poten- concentration. What is now called DLVO-
tial), repulsive forces predominate, and a theory has been reviewed extensively (e.g.,
distinctive maximum, Vmax, occurs in the Hunter, 1987; Russel etal., 1989).
potential energy curve. When this maxi- Especially at small ratios of particle ra-
mum is greater than approximately 10- dius-to-double layer thickness, xa, the
20 kT it is an effective energy barrier double layer repulsion is often referred to
against flocculation. However, with in- as a 'soft' repulsion, since its magnitude
creasing electrolyte concentration or de- varies with interparticle distance.
creasing surface potential, the energy bar-
rier becomes commensurate with the ther- 6.2.1.4 Specific Adsorption
mal energy of the particles which have a It is common practice to use so-called
greater probability of crossing the barrier deflocculating agents to control the proper-
into the flocculated state. At a threshold ties of ceramic slips. These are complex
value of ionic strength, the critical coagu- ionic species such as silicates, phosphates
lation concentration, the system 'floes' and and poly electrolytes. Such complex ions
a much higher energy, Vmin, is necessary to are specifically adsorbed in the inner part
separate the particles again. At intermedi- of the Stern layer. Specific adsorption de-
ate ionic strengths a shallow secondary scribes a special case where the adsorption
minimum, Vsec, exists at somewhat larger involves more than just electrostatic forces
interparticle distances. The depth of this and leads to unequal adsorption of posi-
minimum is usually only about 1-2 kT tive and negative ions. The potential in the
and so these systems form weak floes Stern layer shifts linearly towards higher
which can be readily separated by hydro- or lower values depending on whether
dynamic forces. Derjaguin and Landau counter- or co-ions have been preferen-
(1941) and Verwey and Overbeek (1948) tially adsorbed. Ultimately, this can lead
developed quantitative methods to de- to a reversal of charge within the Stern
scribe the stability of colloids as a function layer and/or enhancement of £ (Fig. 6-5).
 6.2 Colloidal Stabilization Mechanisms for Ceramic Slip Systems 163

(a) (b)

Figure 6-5. Potential-distance

curves showing the effect of
specific adsorption of (a) anionic
and (b) cationic species onto a
negatively charged surface: Wo is
the surface potential, f8 the
potential at the Stern layer and £
the zeta potential.

6.2.2 Polymeric Stabilization the local density of polymer increases,

which leads to a reduction in configura-
Water-sensitive ceramic powders are tional entropy. There is thus an increase in
commonly dispersed in organic solvents free energy which results in a steep rise in
using polymeric additives. Organic disper- the potential energy of repulsion (Fig. 6-7)
sion fluids are common practice in tape- (Everett, 1988).
casting operations (Chap. 7). The effects of The thickness of the adsorbed layer de-
polymer on colloid stability are complex pends on the concentration of polymer in
and not yet completely understood. For a the suspension, its molecular mass, the sol-
detailed exposition of the principles of ubility of the polymer in the dispersing
polymeric stabilization and flocculation of medium and the attractive forces between
colloids the reader is referred to Napper the polymer segments and the surface.
(1983), Hunter (1987) or Russel et al.
(1989). The major aspects of relevance to
slip-casting are steric stabilization and
bridging flocculation. The latter can arise
due to the presence of so-called 'binders' in
the slip.
Steric stabilization of ceramic suspen-
sions can occur when the adsorption of a
sufficiently thick layer of highly solvated
polymer molecules onto the particle sur-
face keeps neighboring particles at a dis-
tance where van der Waals forces are inef-
fective. The steric layer may consist of
Figure 6-6. Schematic representation of the adsorp-
loops, tails or trains of polymer segments tion of a steric stabilizing polymer onto a particle
(Fig. 6-6) (Russel e t a l , 1989). When the surface. The molecules can form loops, trains and
polymeric layers of two particles overlap, tails.
164 6 Slip-Casting and Filter-Pressing

systems, depending on the compressibility

of the adsorbed layers.
Polyacrylic acids (PAA) are electrosteric
stabilizers often used in ceramic systems.
Electrosteric stabilizers provide a mixture
of electrostatic and steric stabilization by
developing a charged polymer layer
around the particles. The mechanism can
be very effective as the stabilized system is
less susceptible to flocculation by in-
creased ionic strength.

Figure 6-7. Total potential energy curves for steri- 6.2.3 Solvation Forces
cally stabilized systems: (i) idealized hard surface,
(ii)-(iv) progressively decreasing density of adsorbed The DLVO treatment alone provides a
polymer layer at constant layer thickness (after satisfactory prediction of the pair interac-
Everett, 1988).
tion at relatively large distances of separa-
tion. However, it predicts that the depth of
the primary minimum is infinitely large.
Often, moderate attraction is preferred to This does not accord with experimental
very strong polymer adsorption since it en- observations on the redispersion of floccu-
ables the rearrangement of polymer seg- lated systems and the ease of break-up of
ments on the surface leading to a dense floes by relatively weak hydrodynamic
coverage (Horn, 1990). A low surface cov- forces. This was recognized by Frens and
erage by polymers of high molecular Overbeek (1972), who proposed that at
weight may lead to bridging effects due to short ranges the displacement of adsorbed
polymer segments adsorbing to more than liquid molecules would become a signifi-
one particle. Polymers adsorbing in a rela- cant factor limiting the depth of the pri-
tively weak manner may help to prevent mary minimum. The existence of solvated
such effects. layers at the surfaces of clay particles has
Another way to avoid bridging effects is long been invoked to account for the ready
the use of so-called block polymers. This dispersibility of such systems (Low, 1961;
type of molecule consists of sequences van Olphen, 1977).
within the polymer chains having different In recent years direct evidence has been
solubility characteristics. If block poly- obtained for the existence of this kind of
mers are used in a solvent in which one semi-incompressible surface layer on sur-
sequence has a higher solubility than the faces. Pashley (1981) and Pashley and
other, the low-solubility sequence will an- Israelachvili (1984) directly measured a
chor to the particle, allowing soluble parts short range (< 5 nm) repulsive force be-
to form loops and tails to provide the steric tween molecularly smooth mica surfaces in
repulsion. As Figure 6-7 shows, sterically aqueous solution at high salt concentra-
stabilized systems generally exhibit some- tions. It was suggested that in the solvation
what 'harder' particle interaction charac- layers at the solid-liquid interface the
teristics than electrostatically stabilized structure at the particle surface and in the
 6.3 Slip Structure and Rheology 165

liquid is modified. This structural pertur-

bation may extend over several molecular
layers into the liquid. Thus, the magnitude
VR (Hydration layer repulsion)
of solvation repulsion depends on the na-
ture of this perturbation and on how far it
extends into the liquid (Horn, 1990). At
present it is not possible to predict solva-
tion forces for a given system since no gen- Distance
eral theory has been brought forward.
Hydration repulsion, that is, repulsion
due to a solvation layer in water, has been
claimed to facilitate the processing of ox-
ide powders in concentrated aqueous elec-
trolyte solutions (Velamakanni et al.,
1990). At high electrolyte concentrations Figure 6-8. Total potential energy curve, V, obtained
( > 1 moldm" 3 ) hydrated cations such as by the summation of a short range hydration repul-
sion energy, VR, and the attractive van der Waals
Li + , Na + , K + , Mg 2 + will penetrate into energy, VA.
the Stern layer. Water molecules outside
the Stern layer may in turn adsorb or hy-
drogen bond onto the cationic surface giv-
ing rise to a structured solvation layer of However, these observations are rela-
several water molecules thickness extend- tively new and the effectiveness of solva-
ing into the liquid. In this case the resulting tion forces in processing remains to be es-
short range repulsion is believed to be as- tablished. The extent to which they are de-
sociated with the energy required to desorb pendent upon the ionic strength is crucial
the hydrated cations from the particle sur- as the processing of advanced ceramics
face. systems with high ionic strength raises
The classical DLVO theory predicts that problems in eliminating the electrolyte
the height of the energy barrier against prior to firing. Nevertheless, it is possible
flocculation decreases and ultimately dis- that the effects are of considerable signifi-
appears with increasing electrolyte concen- cance in controlling the behavior of con-
tration. Particles will flocculate and the en- solidated casts, filter cakes and plastic
ergy will drop into the deep primary mini- bodies where the particles have been force
mum of the total interaction curve (Fig. flocculated over any potential energy bar-
6-4). However, if solvation forces are sig- rier that may have existed when the parti-
nificant a different picture can emerge. The cles were well dispersed in the slip.
deep minimum due to van der Waals at-
traction is replaced by a moderate solva-
tion minimum at small interparticle dis-
tances around 5 nm, followed by strong
6.3 Slip Structure and Rheology
repulsion as particles approach each other
6.3.1 General Background
even closer (Fig. 6-8). The particle interac-
tion in a ceramic slip system stabilized uti- Rheological measurements are widely
lizing solvation forces would be character- used in the ceramics industry as a means of
ized by moderate attraction. determining appropriate dosages of addi-
166 6 Slip-Casting and Filter-Pressing

tives, especially dispersants, and in quality in such a way that they exhibit minimal
control in order to ensure good repeatabil- resistance against movement with respect
ity of different slip batches. Rheology also to each other. In densely packed suspen-
provides a powerful tool in monitoring the sions the viscosity may increase with in-
interparticle forces and structure in a sus- creasing shear rate. Such behavior is re-
pension which cannot be measured di- ferred to as shear-thickening or dilatant
rectly. Both, steady shear flow and visco- and may arise at shear rates above a criti-
elastic behavior at low strains are related cal value when shear planes are disrupted
to the interparticle spacing, which directly (Hoffmann, 1972).
depends on the solids loading, the inter- Sometimes the structural breakdown
particle forces and the particle packing and reformation of shear-thinning or
(Barnes et al., 1989). shear-thickening systems is not only de-
pendent on the forces applied, but also on
6.3.1.1 Steady Shear Rheology the time available for the system to reach
equilibrium. Time-dependent shear-thin-
The rheology of ceramic slips is com-
ning and time-dependent shear-thickening
monly characterized by complex non-
behavior are referred to, respectively, as
Newtonian behavior over a wide range of
thixotropy and rheopexy. The increasing
solids concentrations. Depending on the
and decreasing stress curves exhibit hys-
nature of the suspension structure the vis-
teresis loops. Figure 6-9 shows the typical
cosity can increase or decrease with shear
flow curves of time-dependent and time-
rate and time and often, especially in floc-
independent behavior, respectively.
culated systems, yield stresses appear
(Fig. 6-9).
6.3.1.2 Viscoelastic Properties
A Bingham system behaves similarly to a
Newtonian system in that once the shear Many non-Newtonian materials are vis-
stress exceeds a certain value, the yield coelastic in character. In contrast to New-
stress, the shear stress is proportional to tonian systems which dissipate all energy
the rate of shear. Pseudoplastic or shear- as heat and elastic Hookean solids which
thinning systems show a decrease in viscos- store all energy elastically, viscoelastic sys-
ity with increasing shear rate. Existing tems only store a fraction of the energy
structures break down and particles align elastically, whereas the remainder is dissi-

(a)

Shear - thinning

Figure 6-9. Typical flow curves

of suspensions: (a) time-inde-
pendent and (b) time-dependent.
Shear rate Shear rate
 6.3 Slip Structure and Rheology 167

Shear stress

Shear strain

Figure 6-10. Oscillatory stress-strain

behavior of (a) elastic solids and (b)
viscous liquids.

pated as heat to overcome the internal fric- strain curves are exactly 90° out of phase
tion in reaching a permanent deformation. (Fig. 6-10). Thus, the phase angle, 8, is a
However, to a great extent the measure- characteristic parameter of viscoelastic
ment time scale dictates whether a sub- materials and lies between 0 and 90°.
stance under stress behaves more like a Viscoelastic properties are usually ex-
Newtonian liquid or a Hookean solid. pressed in terms of complex functions. The
When stressed only for a very short time complex shear modulus, G*, can be deter-
most materials respond elastically, mined from oscillatory measurements and
whereas under permanent stress even is defined as
solids ordinarily thought of as elastic may
G* = G' +]G" (6-3)
show some viscous flow.
A number of excellent texts are available where j = yj — 1. The real part, G', is re-
giving guidance to the principles and tech- ferred to as the storage modulus, and the
niques of rheological measurement (Wal- imaginary part, G", is the loss modulus. The
ters, 1975, 1980; Collyer and Clegg, 1988). following relationship exists between the
Viscous systems can be adequately de- complex modulus and its components:
scribed by continuous shear measure- /
ments, but the most common way of char- \G* = V |G'| (6-4)
acterizing viscoelastic systems is to use os- and
cillatory measurements (Ferry, 1980; Col-
lyer and Clegg, 1988). The sample is ex- tan S = (6-5)
G7
posed to a sinusoidal shear stress of a low
magnitude to avoid disturbing the struc- This important rheological characteri-
ture of the sample. Elastic systems exhibit zation technique has been applied to ce-
stress curves which are in phase with the ramic systems only in recent years (Luther
resulting strain curve, whereas in Newto- etal., 1994).
nian systems the shear stress is propor- The rheological behavior of a slip is gov-
tional to the shear rate, and the stress and erned by its structure and how this re-
168 6 Slip-Casting and Filter-Pressing

sponds to the hydrodynamic forces. The

interparticle forces are responsible for de-
termining this behavior. Hence it is conve-
nient to discuss the different rheological
characteristics that may be displayed by
ceramic slips in relation to three specific,
idealized types of particle-particle interac-
tion.
10
Peclet number
6.3.2 Hard-Sphere Repulsion Figure 6-11. Relative steady shear viscosity of 'hard'
polystyrene spheres of various sizes at <>/ = 0.5, sus-
Hard-sphere repulsion is the simplest
pended in different media as a function of Peclet
case of particle-particle interaction in sus- number: (o) benzyl alcohol, (•) m-cresol, (-) water
pension. Particles can approach each other (after Krieger, 1972).
to separation distances slightly larger than
the hard-sphere diameter without any sig-
nificant repulsion. On closer approach, where \] is the suspension viscosity. The
however, they instantly become strongly normalized curves show well-defined
repulsive. In practice, true hard-sphere sys- plateau values for the low- and high-shear
tems are seldom found but the model is a viscosities. At low normalized shear rates
useful basis for the understanding of more (Pe«l) Brownian motion dominates the
complex systems. Both sterically and elec- flow behavior and shear forces do not sig-
trostatically stabilized systems can approx- nificantly perturb the equilibration struc-
imate to hard-sphere behavior when the ture. At intermediate normalized shear
ratio of particle size to stabilizing layer rates (Pe&l) the slips exhibit shear-thin-
thickness is large. ning behavior indicating the increasing in-
In hard-sphere systems only the relative fluence of viscous forces. Finally, at high
magnitudes of hydrodynamic forces and normalized shear rates (Pe » 1) the relative
Brownian motion characterized by the viscosity reaches a second plateau value
Peclet number, Pe, govern the behavior which is completely governed by hydrody-
under shear: namic viscous forces. Figure 6-12 shows
the Peclet number for spheres of different
Pe = (6-6) radii at shear rates typically encountered in
kT rheological measurements and casting pro-
where rj0 is the viscosity of the suspending cesses. On comparison with Fig. 6-11, the
medium and y is the rate of shear. It is Peclet numbers indicate that hard-sphere
possible to superimpose the flow curves of suspensions with a particle size between
slips having varying particle sizes and 0.01 and 0.1 |im should exhibit shear-thin-
medium viscosities at constant solids con- ning behavior at the given shear rates and
centration by plotting the relative viscos- solids volume fractions around 0.5. A sus-
ity, t]r, against the Peclet number (Fig. 6-11) pension of the smallest particle size should
(Krieger, 1972): also be able to restore an equilibrium state
at low shear rates. In a system containing
1 jam spheres, on the other hand, the flow
(6-7)
behavior is mainly governed by hydro-
 6.3 Slip Structure and Rheology 169

maximum attainable volume fraction of

solids. 4>m can be increased by broadening
the particle size distribution in the slip
(Reed, 1988; Smith and Haber, 1992) al-
though this potentially could cause prob-
lems of segregation through sedimentation
of the larger particles (see also Sec. 6.4.2).

"101
6.3.3 Soft-Sphere Repulsion
Shear rate (s1)
Systems with soft particle interactions
Figure 6-12. Peclet number as a function of shear exhibit a strong distance dependence of in-
rate for hard spheres of various radii (a).
terparticle repulsion. Soft interaction oc-
curs in both electrostatically and sterically
stabilized slip systems when the thickness
of the stabilizing layer is significant with
dynamic effects and it seems unlikely that respect to the particle size, characterized
an equilibrium can be attained at low shear by the ratio xa for the former type.
since Pe = 2.5 for 7 = 1 s" 1 . Thus, 1 jum Figure 6-13 shows the relative viscosity
hard spheres at 0.5 solids fraction should as a function of Peclet number for electro-
show Newtonian flow behavior. statically stabilized slips containing parti-
The low- and high-shear relative viscosi- cles of radius 110 nm at different ionic
ties of hard-sphere systems show an expo- strengths (Krieger and Eguiluz, 1976). Un-
tential increase with the solids volume like hard sphere systems, the curves no
fraction, </>, as described by the model of longer normalize with Pe, but show a
Krieger and Dougherty (1959): sharp increase in relative viscosity with de-
creasing electrolyte concentration. The
(6-8) trend is strongest at low shear rates where

where </>m is the so-called maximum pack-

ing fraction (^0.5-0.7) and [rj] the intrin-
sic viscosity which is 2.5 for spheres.
In viscoelastic measurements hard
sphere systems show a steep rise in storage
modulus and a corresponding drop in
phase angle once the solids volume frac-
tion approaches <pm which indicates a liq-
uid-solid phase transition. In the absence 10°
of soft repulsive layers this transition from Peclet number

viscous to elastic behavior occurs rapidly Figure 6-13. Relative steady shear viscosity of 'soft'
over a narrow range of volume fractions of polystyrene spheres with <2=110nm at <j) = 0A sus-
pended in aqueous media at various concentrations of
solids (see also the samples of ionic
HC1 as a function of Peclet number: (o) deionized,
strength above 1 mmol dm" 3 in Fig. 6-14). (•) 1.9xlO~ 4 moldm" 3 , (•) 1.9 x 10~3 mol dm~ 3 ,
4>m is a very important parameter in de- (•) 1.9 xlO" 2 moldm" 3 , (---) hard spheres (after
signing casting slips as it represents the Krieger and Eguiluz, 1976).
170 6 Slip-Casting and Filter-Pressing

Figure 6-14. Phase angle, S, as a function

of volume fraction solids for alumina powder
with a = 10 nm suspended in aqueous KC1
solutions at pH 3.6: (o) no KC1 added, (•)
10~ 3 moldm- 3 KC1, (•) 1(T2 mol dm" 3 KC1
(Fries, 1994).
0.06 0.08 0.10 0.12 0.14 0.16
Volume fraction solids

the samples with extended double layers of a thickness comparable to the particle
behave like solids, indicated by the pres- size which results in a large increase in </>eff.
ence of yield stresses that need to be over- Under these conditions, the maximum
come to enable flow. packing fraction of solids, (/>m, which is
Oscillatory measurements further illus- associated with a liquid-solid transition in
trate the effect of ionic strength: curves of the suspension, can be as low as 0.1. For
phase angle, S, against solids concentra- the majority of current ceramic powders
tion show that the liquid-solid transition, which are generally greater than 0.1 mm,
indicated by a marked drop in <5, is shifted xa only falls below a value of 10 for ionic
to higher solids concentrations as the elec- strengths below 10~ 3 mol dm" 3 . But this
trolyte concentration is decreased (Fig. problem becomes more important with the
6-14) (Fries, 1994). This may be explained use of finer particles to control grain size in
by an increase in effective solids concentra- sintering.
tion, (/>eff, with increasing double layer
thickness, x " 1 (Tadros, 1989): 6.3.4 Attractive Systems
Flocculated systems are difficult to
(6-9) characterize because of their time-depen-
In addition, the liquid-solid transition
occurs over a narrower range at higher
electrolyte concentrations. This indicates a
significant compression of the double layer
with increasing solids concentrations for
samples of low ionic strength before they
become 'solid-like', and an increasing
'hardness' of the interparticle repulsion as
the ionic strength is increased. Figure 6-15
shows the ratio of particle radius to electri-
10"1 10"4 10-3 10"2
cal double layer thickness, xa, for different Ionic strength (mol dm 3 )
particle radii and ionic strength values.
Figure 6-15. Ratio of particle radius to electrical
Suspensions of nanoparticles at low elec- double layer thickness, xa, as a function of ionic
trolyte concentrations have double layers strength for spheres of various radii.
 6.3 Slip Structure and Rheology 171

dent rheological behavior. The influence of 6.3.4.2 Strongly Flocculated Systems

Brownian motion is relatively weak com-
Strongly flocculated suspensions are
pared to the energy of attraction and when
characterized by the strong van der Waals
the slips are at rest nonequilibrium struc-
attraction in the primary DLVO minimum.
tures such as open particle-floc-aggregate
The systems cannot attain equilibrium in
networks, are encountered (Firth, 1976;
an experimentally realistic time scale as
Firth and Hunter, 1976a, b). Both flow
diffusion is negligible compared with the
and viscoelastic properties of a given sys-
attractive forces. The particles form strong
tem therefore may strongly depend on its
open floes which in turn grow into larger
shear history. Attractive systems are often
floe aggregates, leading to the formation
categorized as to whether they are 'weakly
of an attractive network throughout the
flocculated' (1 < Vmin/kT<20) or 'strong-
slip at a relatively low critical solids vol-
ly flocculated' (VmJkT> 20).
ume fraction, </)crit. The concepts of fractal
geometry have been successfully applied to
6.3.4.1 Weakly Flocculated Systems model the disordered structure of par-
Weak flocculation has been reported for ticle-floc aggregates as a function of shear
sterically stabilized systems which were rate at solids concentration below 0 c r i t
made attractive by the addition of free (Meakin, 1988). Percolation theory can be
non-adsorbing polymer (Prestidge and used to predict network formation at
Tadros, 1988) or by raising the tempera- higher solids concentrations (Safran et al.,
ture above a critical value (Woutersen and 1987; Seaton and Glandt, 1987).
de Kruif, 1991). Electrostatically stabilized The rheology of strongly attractive sys-
systems that reside in the secondary DLVO tems is characterized by the presence of
minimum are also weakly attractive (Bus- yield stresses and highly elastic behavior at
call et al., 1990). The steady shear behavior 0>(/> crit . The storage moduli of floccu-
of weakly flocculated systems is character- lated systems are reported to scale with cj)^
ized by strong shear-thinning character as where ft = 2 to 5 (Buscall et al., 1986, 1988;
a result of the breakdown of floes. Previ- Sonntag and Russel, 1987; Chen and Rus-
ously immobilized liquid inside the floes is sel, 1991). Firth and Hunter (1976 b) de-
liberated under shear leading to a decrease rived an 'elastic floe model' which corre-
in the effective solids concentration. Such lates floe structure with rheological prop-
systems may approximate to the Bingham erties. They measured the yield stresses of
system shown in Fig. 6-9 for which various systems and found a good agree-
ment between experimental and theoreti-
T-T=rjvy
(6-10) cal data. The yield stress scaled with </>2. At
where T is the shear stress, xy the Bingham low solids concentration the concept of a
yield stress and rjp a constant known as the floe structure indicator, C F P , due to
plastic viscosity. The Bingham yield stress Michaels and Bolger (1962), was found to
of weakly flocculated slips is related to the be useful:
energy required to break floes into single
particles or smaller flow units. CFP — (6-11)

where </>F is the volume fraction of floes.

Typical values of C F P are 5-13 for dilute
172 6 Slip-Casting and Filter-Pressing

coagulated ceramic suspensions (Michaels tance, Lc is the cast thickness at time, t,

and Bolger, 1962; Firth and Hunter, and Z is the volume of fluid filtrate re-
1976a, b; Bushnell-Watson, 1983). Firth moved from the slip in forming unit-vol-
(1976) and Firth and Hunter (1976 a, b) re- ume of cast:
lated CFP to the magnitudes of interparticle
forces and showed that CFP decreased lin- Z = (6-14)
early when plotted against C2, the square of
the zeta potential. This observation is con- where 4>c is the volume fraction of solids in
sistent with a wide range of experimental the cast. Where the resistance of the filter
studies which show that the rheological be- becomes significant, Equation (6-13) be-
havior changes systematically with ionic comes
concentration and pH for ceramic systems. 2RFAF 2AP t
For example, yield stresses increase with (6-15)
acf]0Z Lc
electrolyte concentration beyond the criti-
cal coagulation concentration required to in which AF is the cross-sectional area of
eliminate the energy barrier. This indicates the filter and RF is the resistance of the
that the depth of the primary minimum is filter, which is assumed constant. Thus a
also changing progressively. Thus there is plot of Lc vs. t/Lc should be linear in both
a complete spectrum of structures from the cases, provided that the cast layer is of
highly repulsive soft- and hard-sphere in- uniform structure such that occ and (/>c are
teractions through to the very strongly at- independent of L c . When the filter resis-
tractive. These differing particle-particle tance is significant it can be estimated from
interactions can result in widely different the intercept which appears in this case.
casting characteristics. The case of slip casting differs in two
ways. First, the pressure driving force is
the sum of the capillary pressure due to the
6.4 Mechanism and Kinetics suction pressure of the mold and any addi-
tional pressure applied to the slip or vac-
of Slip-Casting and Filtration uum applied to the mold. The capillary
pressure is determined by the pore size dis-
6.4.1 General Kinetics
tribution of the mold, the surface tension
The kinetics of filtration are well-estab- of the dispersion fluid and its contact angle
lished. The application of Darcy's law with the pore walls. It is reported that plas-
leads to the equations ter of Paris molds give suction pressures of
the order of 0.1-0.2MPa (Adcock and
AP McDowall, 1957; Dal and Berden, 1968).
(6-12)
dt acrj0ZLc Tiller and Tsai (1986) have estimated suc-
and tion pressures of molds comprising packed
2AP powder beds. Second, there is a resistance
t (6-13) due to the movement of the abstracted liq-
uid through the pore structure of the mold
for an incompressible filter cake, assuming as the cast builds up. Figure 6-16 shows a
a negligible resistance for the filter itself. schematic representation of the situation.
AP is the pressure drop across the filter In analyzing the kinetics of slip-casting,
cake or cast, occ is the specific cast resis- many authors have followed Adcock and
 6.4 Mechanism and Kinetics of Slip-Casting and Filtration 173

may still result. Thus, assuming that the

mmmmm —Slip
pores in the mold adjacent to the cast layer
are fully saturated with the abstracted fil-
trate,

-Consolidated Layer
dt Z dt f]0(xcLcZ
b -Mold (saturated)

(6-16)
1 I —Mold (dry)

where PT-Pi is the pressure drop across the

cast, P\-PQ that across the wetted part of
Figure 6-16. Schematic diagram of the slip cast sys- the mold and am the specific resistance of
tem.
the mold of porosity, em. The suction pres-
sure of the mold is PT-P0. Now
McDowall (1957) and ignored the effect of 2 £ m (P I -P 0 )
the mold, except in so far as it controls the (6-17)
suction pressure, assuming that the pres-
sure drop across the cast is equal to the When the hydraulic resistance of the
suction pressure. The parabolic nature of mold can be neglected, the Adcock and
the casting kinetics was established in McDowall expression (1957) ensues. This
many cases. However, Dal and Deen situation applies when the ratio
(1958) followed by Walker and Dinsdale
(1959), Dal and Berden (1968) and more ocmZ
(6-18)
recently Aksay and Schilling (1984) and
Tiller and Tsai (1986) have taken the pres- is very small, favored by a high solids con-
sure drop across the wetted part of the centration in the slip (near to </>c), a high
mold into account (Fig. 6-17). Since L m , mold porosity and a large differential be-
the wetted depth of the mold, and Lc are tween the mold and cast hydraulic resis-
both proportional to the amount of liquid tances.
abstracted by the mold, parabolic kinetics Viscous flow through porous media is
often modeled by means of the Kozeny-
Mold Consolidated layer
Carman model (Carman, 1956) which al-
lows the specific resistance to be related to
the porosity and surface area per unit vol-
ume, Sy of the porous body. Although the
assumptions implicit in the derivation are
Water
oversimplified and open to criticism, it
gives a mathematical representation of the
specific resistance that is useful here in elu-
cidating some of the effects of cast and
Distance- mold structure on the casting kinetics, i.e.,
Figure 6-17. Profile of hydraulic pressure across the
consolidated layer and mold (after Aksay and «c=^7^ (6-19)
Schilling, 1984).
174 6 Slip-Casting and Filter-Pressing

where K is a constant related to the tortu-

osity of the pore structure in the cast. It has
been shown that K&5 for many packed
powder beds (Carman, 1956). SY can be
related to the equivalent spherical radius,
B
a, of the particles in a monosize slip to
CO
CD
(SY = 3/a), and the specific resistance of the cr
cast expressed as
45 $1
OL = (6-20) 1 2 3
Particle diameter (\im)

Estimates of how the specific resistance Figure 6-18. Specific resistance of porous media
varies with particle size for different values formed by packing monosized spheres as a function
of cast solid volume fraction from Aksay of diameter for various cast porosities, ec (after Aksay
and Schilling, 1984).
and Schilling (1984) are shown in Fig.
6-18. Aksay and Schilling compared the
estimated values with a typical value for
the mold resistance and showed that the not present any experimental observations
mold resistance could become significant to validate their model.
when the diameter of the particles in the The predictions of specific resistance
slip is large and the solid volume fraction shown in Fig. 6-18 are useful in showing
in the slip and cast are low. A large particle the effect of changes in cast solid volume
size results in larger pore size and this com- fraction and particle size or the casting ki-
bined with a low solid volume fraction in netics. However, they should be viewed
the cast decreases its specific resistance, a c , with caution since they give values signifi-
whereas a decrease in </> increases the pa- cantly lower than those reported in the lit-
rameter Z. erature, which are usually in the range
Tiller and Tsai (1986) extended this from 1016 to 1018 m~ 2 , depending upon
treatment by applying the Kozeny-Car- the nature of the casting slip and the extent
man approach to the mold, relating the of deflocculation (Reed, 1988; Li and
pore size to an assumed particle size in the Rand, 1984; Hampton etal., 1992; Ad-
mold and estimating relationships between cock and McDowall, 1957). The Kozeny-
pore dimensions and the capillary suction Carman model is only approximate and
pressure. They showed that there should like most other models of flow through
be an optimum pore structure for each slip porous media it does not accurately repre-
to give the maximum pressure drop across sent the morphology of the flow channels.
the cast layer. They also went on to at- The subject is reviewed by Dullien (1992)
tempt to predict cast porosities for the var- and Scheidegger (1974). One specific prob-
ious slip-mold combinations using empir- lem is that the Kozeny-Carman model as-
ical relationships to predict local solid vol- sumes that all of the pore network is in-
ume fractions and permeabilities with volved in flow and the pore structure is
depth in the cast. This is an important ap- uniform throughout the sample. This is
proach since the relationship between cast frequently not the case, especially when
structure and casting conditions is not well flocculated slips are filtered or cast. The
understood. However, Tiller and Tsai did casts produced from such slips are usually
 6.4 Mechanism and Kinetics of Slip-Casting and Filtration 175

highly compressible, since the critical fac-

tor in determining the value of (j)c is the
extent to which the interconnected voids
between floes are eliminated during con-
solidation and the internal network within
the floes is rearranged. This will depend
upon the hydraulic pressure in the cast and
its compressibility. The compressibility is
controlled by particle size, shape and inter-
particle forces, the latter becoming more 0.4 0.6
significant the smaller the particle size. Fractional distance
Where strong repulsive forces prevail the Figure 6-19. Normalized hydraulic pressure in filter
particles can assemble themselves into cakes cast from slips of different degrees of floccula-
dense, relatively incompressible arrays. tion as a function of fractional distance from the filter
(after Tiller and Tsai, 1986).
When unconsolidated, a flocculated sys-
tem will exhibit a bimodal pore size distri-
bution and the larger interfloc voids may
be the principal channels for flow sure a short distance from the cake-filter
(Michaels et al., 1967). This can give a (mold) interface. In this latter case the
specific resistance much lower than that porosity, sc = 1 — <j>c, is lower near the filter
predicted from the Kozeny-Carman medium but increases rapidly, as the cake
model. On the other hand, Michaels et al. builds up, to an almost constant high
(1967) also showed that when filter cakes porosity region through the bulk of the
from flocculated clays are compressed to cake, which is uncompressed (Fig. 6-20).
low porosities the specific resistance be- Since the compressibility is controlled by
comes much larger than the Kozeny-Car- the strength of the particle attraction
man value, owing to particle orientation which controls the structure within floes,
effects which increase the hydraulic tortu- as well as by interfloc forces, it is related to
osity, K. There is a need for detailed char- the extent of deflocculation and thus also
acterization of the pore structure for pre-
cise modeling of the cast kinetics. Smith
et al. (1994) recently demonstrated the
need to use different techniques for charac-
terization of the pore structure of the cast.
Multi small angle neutron scattering was
able to characterize the bimodal pore size
distribution in an alumina cast from a par-
tially flocculated slip that was 'seen' as
monomodal by mercury porosimetry.
Tiller and Tsai (1986) showed how the
hydraulic pressure may vary with distance 0.2 0.4 0.6 0.8
Fractional distance
through a cast layer (Fig. 6-19). For an
Figure 6-20. Local porosity in filter cakes cast from
incompressible cake the variation is linear slips of different degrees of flocculation as a function
leading to a uniform cpc. Highly compress- of fractional distance from the filter (after Tiller and
ible cakes show a sharp increase in pres- Tsai, 1986).
176 6 Slip-Casting and Filter-Pressing

to the rheological behavior of the slip. It ties are achieved by fine particles filling the
was mentioned above that there is a con- interstices between the large. Monosized
tinuous spectrum of interparticle forces, slurries show no solids concentration de-
ranging from strongly attractive through pendency of green density due to this ef-
to strongly repulsive. Clearly this can re- fect.
sult in casts with continuously varying Hampton et al. (1988) studied the slip-
compressibilities. The above discussion casting-behavior of two alumina powders
shows that for compressible systems the having mean particle sizes of 4.2 and
parameters a c , (f)c and Z in the kinetic ex- 0.4 jim over a range of solids concentra-
pression will vary with cast depth. How- tions. They found that the green density of
ever, it is still possible to use the parabolic compacts formed using only the largest in-
expressions with average values, ac(ave), creased with solids concentration. This
0 c(ave) and Z ave (Tiller and Tsai, 1986). was attributed to the formation of close-
packed arrangements as indicated by the
increasing shear-thickening of the suspen-
6.4.2 Effects of Solids Concentration sions at high solids concentrations. On the
and Particle Size Distribution other hand, the densities of cakes consoli-
on Cast Structure dated from only the fines decreased
Although the interparticle forces in a ce- slightly over the same solids concentration
ramic slip have probably the strongest in- range possibly associated with a decrease
fluence on the resulting cast microstruc- in slip stability as indicated by increasing
ture, other factors also play a role in the shear-thinning at high solids concentra-
evolution of a successful casting route. In tions. A 50/50 mixture of the two powders
this section we will attempt to elucidate the yielded the highest green densities which
effects of solids concentration, particle size remained constant over the studied solids
and particle size distribution on casting be- concentration range. Velamakanni and
havior and cast structure. Lange (1991) examined the casting behav-
From Equations (6-12) to (6-14) it is ob- ior of mixtures of two alumina powders
vious that the concentration of solids in having mean sizes of 0.5 and 1.3 jim. The
the slip is directly proportional to the cast- highest green density (0.67) was achieved
ing rate. Its effect on the resulting green with a 40/60 mixture, whereas the powders
density, however, is somewhat more com- consolidated to a lower density (0.62)
plex. In general, a decrease in green density when individually cast. At solids concen-
at low solids concentration can be ex- trations below 0.5 the particles were found
pected if the sedimentation rate of col- to segregate.
loidal particles in a slip is significant with Maximum packing efficiency can be
respect to the filtration rate. Under these achieved with a polydispersion of spherical
conditions, segregation of particles of dif- particles that follows Andreasen and An-
ferent size can take place in dilute suspen- derson's (1930) particle packing equa-
sions; in concentrated slips, however, the tion
sedimentation velocity is diminished to
only a small percentage of that of an iso- F(a) = (6-21)
lated particle (Buscall etal.,1982). Of
course, this mechanism only applies for where F(a) is the cumulative volume frac-
size distributions where high green densi- tion of particles of diameter smaller than a,
 6.5 Control of Interparticle Forces, Rheology and Cast Structure 177

amax the maximum particle radius and x that the particles exhibit a high zeta poten-
the distribution modulus. More recently tial. This is achieved when the ionic
Dinger and co-workers (1982) modified strength is low (high Debye thickness, l/x)
the equation to allow for the more realistic and the surface potential is high. In oxide
case of a finite minimum particle size: systems the surface potential, !F0, is con-
trolled by the activities of potential deter-
F(a) = a —amin (6-22) mining ions (H + and O H ) in solution.
The adsorption of these species can be fol-
where amin is the minimum particle radius. lowed by simple titration procedures
A distribution modulus of 0.3 to 0.4 com- (Parks and de Bruyn, 1962) to establish the
monly is observed to yield the maximum surface charge density. At some pH-value
packing density for equiaxed particles. The there is equal adsorption of H + and OH~
pore size and hence the specific resistance leading to an electrically neutral surface.
of such a densely packed structure is deter- This is the point of zero charge, PZC,
mined by the size of the finest particles in which, in the absence of any specifically
the distribution. adsorbed anions or cations, is equal to the
If monodispersions of different particle pH of zero zeta potential, the isoelectric
size are consolidated to a given packing point, IEP. Site dissociation-site binding
arrangement, the porosity will be the same models have been proposed to account for
for each particle size. The pore size, how- the amphoteric nature of the surface (Yates
ever, will be related to the particle size, etal., 1974; Hunter, 1987; Pugh, 1994).
which has important implications for re- The fundamental factor in controlling the
sultant properties such as moisture stress, surface potential and hence the behavior of
casting rate, cake permeability and cake oxide dispersions is the pH-value of the
resistance. IEP. For oxide powders the values of IEP
are fairly reproducible and characteristic
of the oxide type. IEP values have been
6.5 Control of Interparticle Forces, reviewed by Parks (1965) and Yoon et al.
Rheology and Cast Structure (1979). Table 6-1 lists the IEPs for a range
of these materials. In general, the IEP is
6.5.1 Deflocculation low for acidic oxides, that is those having
tetravalent cations, and high for basic ox-
This section will focus on the practice of ides with cations of lower valency. For
stabilizing aqueous systems since in slip- complex oxides the IEP lies between the
casting most ceramic powders are stabi- values for the two oxides. For example Jo-
lized by electrostatic mechanisms or a mix- hansen and Buchanan (1957) synthesized a
ture of electrostatic and steric effects as range of alumina silicates in which the
exhibited by certain organic polyelec- A12O3 —SiO2 ratio was varied. The IEP
trolytes. Organic solvents are employed changed systematically from that of SiO2
when the powder properties deteriorate in to that of A12O3. This is significant for the
water. behavior of clay minerals (Sec. 6.5.1.2). In
practice it is the surface composition that
6.5.1.1 Advanced Ceramic Systems is important in controlling the IEP. Figure
As outlined in Section 6.2.1, electro- 6-21 shows the variation of zeta potential
static stabilization is attained by ensuring with pH for TiO 2 in the presence of differ-
178 6 Slip-Casting and Filter-Pressing

Table 6-1. Nominal isoelectric points of various ce- usually an optimum pH-region since mov-
ramic materials (after Reed, 1988) ing the pH to very low (pH<3) or very
Material IEP high ( p H > l l ) values increases the ionic
strength sufficiently to bring about floccu-
Silica 2 lation by compressing the electrical double
Silicon carbide 3
layer. Consequently oxides with low and
Soda lime silica glass 2-3
Potassium feldspar 3-5 high IEP-values show only one region of
Zirconia 4-6 stabilization.
Apatite 4-6 Nonoxide ceramics often have a surface
Tin oxide 4-5 layer of oxide which controls the colloid
Titania 4-6
chemical behavior. SiC, for example,
Silicon nitride 5-7
Kaolin (edges) 5-7 shows an IEP of pH 3 close to that of SiO2.
Mullite 6-8 However, the surfaces of nitrides are un-
Chromium oxide 6-7 stable in aqueous environments and show
Hematite 8-9 more complicated behavior (Pugh, 1994).
Zinc oxide 9
Alumina 8-9
Where the ionic species are specifically
Yttria 9 adsorbed into the Stern layer the IEP is
Calcium carbonate 9-10 shifted to low pH-values for anionic spe-
Magnesia 12 cies and to higher pH for cationic species
(Hunter, 1981). Such complex ionic species
often lead to very effective deflocculation.
ent concentrations of indifferent elec- Table 6-2 shows typical deflocculants for
trolyte (Wiese and Healy, 1974). ceramic systems. As outlined in Sec. 6.2.1,
It is evident that oxides should show specific adsorption may reverse the polar-
electrostatic stabilization at pH-values on ity near to the original IEP and increase
either side of the IEP provided the ionic the negative zeta potential of the particle at
strength is low enough. However, there is pH above the original IEP. This has a pro-
found beneficial effect on the electrostatic
stabilization.
A number of authors report the prepa-
ration of deflocculated alumina slips as in-
dicated by minima in the slip viscosity at
5 > p H > l l (e.g., Anderson and Murray,

Table 6-2. Common dellocculants for aqueous ce-

ramic systems (after Reed, 1988)

Inorganic Organic

Sodium carbonate Sodium polyacrylate

Sodium silicate Ammonium polyacrylate
Sodium borate Sodium citrate
Figure 6-21. Zeta potential as a function of pH for Tetrasodium pyrophosphate Sodium succinate
titania powder suspended in aqueous KNO 3 solu- Sodium tartrate
tions at various ionic strengths: (o) 10~ 4 moldm~ 3 Sodium polysulfonate
KNO 3 , (•) 1(T3 mol dm" 3 KNO 3 , (a) 10~2 mol Ammonium citrate
dm" 3 KNO3 (after Wiese and Healy, 1975).
 6.5 Control of Interparticle Forces, Rheology and Cast Structure 179

1959; Cooper and Miskin, 1965; Vela- of PAA required to deflocculate suspen-
makanni et al., 1990). This is consistent sions as a function of pH. Sumita et al.
with an IEP of pH 8-9. Nikumbh et al. (1991) studied the effects of various water
(1991) stabilized various zirconia powders soluble polymers and benzoic acid deriva-
in the absence of polyelectrolytes in the tives on the green densities of alumina
pH-range 2 - 5 (IEP^pH 4.5-6). A rever- compacts. The highest densities were
sal of charge was detected for these sys- achieved from slips deflocculated with
tems on the addition of a deflocculant con- — NH 2 or —OH derivatives of benzoic
taining alkali free carboxylic acid groups: acid, but several other deflocculants were
the pH-range of minimum viscosity shifted more effective than merely lowering the
to pH 8-12, consistent with a change in pH.
IEP to a low pH-value. Persson et al. Polymeric binders such as polyvinyl-
(1983) investigated £-SiC and Si 3 N 4 pow- alcohols (PVA), cellulose derivatives,
ders and established, respectively, a pH- starches, latex, polyethylenglycols etc., can
range from 8-12 and 5 > p H > 9 as suit- be added to the slip to increase the green
able for the preparation of casting slips. strength. Often the addition of binders re-
They also reported effective deflocculation sults in an increase in shear thinning and a
using ligno-sulphonates. The adsorption reduction in cast permeability and casting
of the latter deflocculant led to a shift in rate due to both an increase in fluid viscos-
the IEP from approximately pH 6.5 to ity, rj0, and the presence of binder in pores
pH 3. A commonly used polyelectrolyte which results in an increase in the specific
for ceramic systems is polyacrylic acid resistance of the cast.
(PAA), which has an electrosteric charac-
ter. It has been shown to strongly adsorb
onto various positively charged powders, 6.5.1.2 Clay-Based Ceramic Systems
e.g., alumina and titania, from low pH-val- The origin of charge on clay particles is
ues up to their respective IEP's (Gebhardt mainly the desorption of weakly bonded
and Fuerstenau, 1983; Cesareno III and alkali or alkaline earth cations which bal-
Aksay, 1988). Cesareno III and Aksay ance the negative charge in the crystal due
(1988) showed that the stabilizing action of to isomorphous substitution (van Olphen,
PAA is fairly independent of pH at low 1977). Thus, the basal faces of most clay
solids fractions, whereas for highly con- particles are negatively charged over a
centrated alumina slips the viscosity mini- wide pH-range. However, the fractured
mum was found to coincide with the origi- edges of the crystal are characterized by a
nal IEP of alumina. Below the original IEP pH-dependent surface potential and an
the viscosity was believed to increase be- IEP. This leads to the heteropolar nature
cause of a decrease in the negative charge of clay particles so that at a pH below the
characteristics of the adsorbed electrolyte edge IEP, the edges may be positively
and an effective decrease in the repulsive charged whilst the faces remain negative.
electrosteric barrier between the particles. The resulting flocculated 'house of cards'
At pH-values above the original IEP the structure exhibits strong mutual attraction
destabilization was believed to be associ- between edges and faces due to both, van
ated with the presence of excess polymer in der Waals and electrostatic forces. On the
solution. From their findings they devel- other hand, the clay particles become de-
oped a stability map showing the amount flocculated if the pH is raised to a value
180 6 Slip-Casting and Filter-Pressing

above the IEP of the edge surface, pro- provided the most effective deflocculation
vided the ionic strength is not too high combining electrostatic and steric ef-
(Flegmann et al., 1969). At high pH-values fects.
and high ionic strength, the double layer Diz etal. (1990) showed that organic
repulsion becomes too small to screen the matter adsorbed on the edge faces of ball
van der Waals attraction and the system clays can improve deflocculation and act
flocculates into a more dense attractive as a polyelectrolyte lowering the edge IEP.
face-to-face structure. In certain cases, however, when organic
Rand and Melton (1977) developed a matter was only present in low concentra-
simple rheological technique for determin- tions, deflocculation was not aided and the
ing the edge IEP for homoionic kaolinites. flocculating effect of cations at high pH
The edge IEP is a variable quantity de- was enhanced. Variations in clay compo-
pending on the SiO 2 -Al 2 O 3 ratio exposed sition and organic impurities, which can
at that surface which leads to defloccula- have both flocculating and deflocculating
tion at different pH-values. Natural sur- effects, can make the formulation of suit-
face active organic impurities adsorbed at able casting slips a difficult task. It is essen-
the edge faces of the particles may also tial to develop a formulation that balances
effect the IEP value (Rand et al., 1987; Diz the deflocculating and coagulating effects
etal., 1990). Na-montmorillonite suspen- of different electrolytes and also accom-
sions, in contrast, showed no tendency to modates slight batch-to-batch variations
coagulate over the pH-range 4-11 indicat- in the clay composition.
ing the absence of edge-to-face floccula- Sodium carbonate/sodium silicate mix-
tion in this range (Rand et al., 1980). tures have long been used as deflocculants
The anionic groups of poly electrolytes, for clay systems. Their behavior can be
e.g., Na 2 SiO 3 or Na 2 CO 3 for clay-based understood from the above discussion.
systems, adsorb specifically into the Stern Sodium carbonate acts to raise the pH
layer at the edge surfaces of clay particles, value above the edge IEP; the silicate is
the positive charge at the edge surface is also specifically adsorbed at the clay edges,
reversed within the Stern layer and the IEP lowering the IEP. Both of these promote
is shifted to a value below the suspension deflocculation. However, the use of
pH. What is more, the anions are able to sodium carbonate results in significant
form complexes with polyvalent ions in the sodium ion activities which will tend to
double layer and replace them with mono- compress electrical double layers ensuring
valent cations such as Na + . This leads to that the deflocculation is not too severe.
an effective decrease in ionic strength The deflocculation curve of a whiteware
which, in turn, increases the zeta potential slip is given in Figure 6-22 as a function of
and the repulsion between the particles. Na 2 SiO 3 dosage showing the casting
Diz and Rand (1990) investigated the range.
mechanisms of deflocculation of kaolinite
by sodium metasilicate, sodium pyrophos-
6.5.2 Interparticle Forces and the Control
phate and sodium acrylate via rheological
of Cast Structure
measurements. It was shown that the edge
IEP is progressively shifted to a lower pH- Although slip-casting is a major fabrica-
value as the extent of adsorption of the tion route for both traditional clay-based
polyanion is increased. The polyacrylate ceramics and high-performance ceramics,
 6.5 Control of Interparticle Forces, Rheology and Cast Structure 181

A typical casting slip for the traditional

ceramics industry is partially flocculated
(Fig. 6-22) and often contains a balance of
both dispersants and flocculants, as dis-
cussed in Sec. 6.5.1.2. The casting range of
the suspension does not coincide with the
minimum viscosity but is located at some-
what higher viscosity values. This is be-
cause of the need to provide a measure of
flocculation to prevent the sedimentation
of the filler and flux particles which are
Sodium silicate larger than the colloidal clay particles.
Figure 6-22. Deflocculation curve for a clay-based Also, by not fully deflocculating the slip, a
ceramic slip showing the optimum casting range as a higher casting rate can be attained and
function of sodium silicate dosage.
some plasticity is retained in the cast, as
described below. Due to the slight destabi-
the characteristics of good casting slips for
lization of the slip the resulting microstruc-
the two types of ceramics differ signifi-
ture of the cast is characterized by a higher
cantly in the degree of particle stabiliza-
porosity than in the deflocculated state
tion, and as a consequence there are differ-
with some pore dimensions being approxi-
ences in the casting and demolding behav-
mately equal to the size of the agglomer-
ior as well as in the resulting green struc-
ates, as discussed in Sec. 6.4. This control
tures. This section indicates how casting
of voidage in the slip allows control of the
and the cast structure are influenced by
specific resistance of the cast. The larger
those forces. First, traditional ceramics
pores also allow relatively rapid redistribu-
will be considered.
tion of moisture after demolding. The cast
is highly plastic and shows favorable trim-
6.5.2.1 Clay-Based Ceramic Systems ming and handling properties (Funk,
High green densities and controlled mi- 1984).
crostructures are the top priority in high- On the other hand, if a clay-based cast-
performance ceramics but are only of sec- ing slip is fully deflocculated, the casting
ondary importance in the traditional ce- rate is diminished by the high specific resis-
ramics industry which has to fabricate tance of the dense cast characterized by a
large numbers of products reliably in a small pore size. The cast is brittle and dila-
highly competitive environment. The re- tant, which makes trimming very difficult.
quirements for a good casting sMp and a Mold release is also impaired.
satisfactory cast were given in Section
6.1.2. For clay-based casts a relatively high
6.5.2.2 Advanced Ceramic Systems
moisture retention through the cast wall is
desirable to give a high plasticity in the High-performance ceramics are pro-
green body which facilitates handling and duced from slips that consolidate to high
trimming after demolding. What is more, a and uniform green density since this is an
'forgiving' slip is required to accommodate essential prerequisite for sintering to theo-
slight batch-to-batch variations in additive retical density in the absence of excessive
dosage and clay material characteristics. grain growth. This goal is commonly
182 6 Slip-Casting and Filter-Pressing

achieved by preparing casting slips at or solids concentrations due to the weakly

near to the viscosity minimum in the de- flocculated network and relatively high
flocculation curve. slip viscosity. This may be particularly use-
Fennelly and Reed (1972 a, b) systemati- ful for systems which contain powders of
cally studied the mechanics of filter-press- different densities, or when highly concen-
ing as a function of slip stability compar- trated slips prove difficult to formulate,
ing three aqueous alumina slips with dif- handle or store.
ferent polyelectrolyte additions. The cast- To our knowledge, there are no records
ing rate decreased for a given pressure with dealing with the conventional casting of
increasing extent of stabilization. In addi- advanced ceramics slurries at high salt
tion, filter cakes consolidated from the sta- concentration, but it seems likely that a
ble slips were incompressible at pressures pressure above the suction pressure of a
above 0.69 MPa indicated by their con- typical plaster mold is necessary to achieve
stant specific resistance. (For comparison: the high densities reported by Lange and
the suction pressure of a typical plaster co-workers. Also, the processing of slips
mold is approximately 0.1 to 0.2 MPa.) with such high levels of salt poses diffi-
The flocculated systems, in contrast, were culties owing to the effects of salt migra-
compressible and showed an increasing tion and crystallization in the surface re-
cast resistance and decreasing casting rate gions of the body during drying. Weakly
with increasing pressure. attracting systems can be produced by the
Lange and co-workers have recently addition of nonadsorbing polymer to a
postulated a new paradigm for efficient fil- stabilized dispersion and may be a more
ter-pressing, exploiting short-range repul- effective method of controlling the consol-
sive solvation forces. They stabilized idation process (Russel, 1987).
A12O3, 2-phase Al 2 O 3 -ZrO 2 (Chang Only a limited number of systematic
etal., 1991) and recently Si 3 N 4 powders casting studies using sterically stabilized
(Luther et al., 1994) in water at pH<IEP, suspensions exist. Kerkar etal. (1990a,b)
where the powder surfaces are positively stabilized silicon powders in benzene and
charged, and added large amounts of trichloroethylene using polystyrene (PS),
NH4C1 to the systems to achieve a short polymethyl methacrylate (PMMA) and
range repulsion as outlined in Section their co-polymers. Of the studied poly-
6.2.3. It is suggested that under these con- mers, PMMA and co-PS-PMMA pro-
ditions, ceramic suspensions flocculate vided adequate stabilization. The specific
and form a weakly attractive, nontouching cake resistance of pressure cast compacts
network allowing rearrangement of the initially increased with polymer concentra-
particles, presumably due to the depth of tion and then leveled off indicating com-
the minimum being less than about 10 kT. plete polymer surface coverage of the par-
Both filter-pressing and centrifuging ex- ticles. The green density of the compacts
periments using slips flocculated in this also increased with the polymer concentra-
manner yielded high green densities which tion and exhibited a distinct maximum
were insensitive to the applied pressure which was explained by the presence of the
and almost identical to those achieved polymer in the cast at high polymer con-
from fully deflocculated systems. In addi- centrations in the slip. The added polymers
tion, particle/mass segregation by sedi- had an additional beneficial effect by act-
mentation was eliminated at relatively low ing as a binder which was evident from an
 6.6 Defects and Microstructural Nonuniformities 183

increase in green strength with increasing laxation data as a function of time. Bodies
polymer concentration. formed from deflocculated slips showed an
Problems have been identified after de- irreproducible mixture of plastic and elas-
molding the highly dense samples pres- tic response. A possible explanation for
sure-cast from either flocculated or dis- this behavior would be mixed particle-
persed slurries of advanced ceramic pow- particle interactions: as particles in a com-
ders. As the last portion of slip is consoli- pressed network are exposed to a wide
dated in the filter press, the pressure gradi- range of forces, only a fraction of particles
ent through the cast becomes zero and the are believed to be pushed into the primary
applied stress is completely transferred to energy minimum whereas the remainder
the filter cake. On ejection from the filter stay on the repulsive side of the DLVO
press the filter cakes commonly exhibit curve. Second, compacts formed from
time-dependent strain relaxation mani- flocculated samples, prepared in the ab-
fested by an increase in compact dimen- sence of any significant surface charge,
sions. The release of elastically stored en- were able to withstand the highest stresses.
ergy requires fluid flow from the surface of The behavior was explained by the pres-
the cake to the interior leading to a gradi- ence of a strong cohesive network of at-
ent in fluid pressure governed by the per- tractive particles that can support high
meability of the cake. What is more, the stresses. Finally, weakly flocculated slips,
exterior of the sample will relax first, lead- prepared at positive surface charge with
ing to macroscopic stress gradients during salt addition, led to compacts that in-
relaxation. Lange and Miller (1987) found creased in initial stress with salt concentra-
that filter cakes cast from deflocculated tion, but behaved plastically, with most of
slurries resulted in a dilatant compact, the stress relaxing within a short period.
rigid and stiff at high shear rates but soft These results were suggested as being con-
and deformable at low ones. This imposes sistent with a short-range repulsive force
serious problems on the handling of sam- that lowers the attraction between parti-
ples where close dimensional tolerances cles.
are required. Drying prior to demolding is
possibly a solution to this problem. On the
other hand, bodies consolidated from floc-
culated slips showed slow time dependent
6.6 Defects and Microstructural
crack formation on strain release which Nonuniformities
could be counteracted by small binder ad-
ditions to the slip. According to Persson A number of specific defects can charac-
(1994), none of these problems are re- teristically arise in slip-cast products.
ported for pressure-cast clay-based prod- These are all due to nonuniformities in the
ucts, probably because of the fact that clay green body structure that arise during the
particles generate significant plasticity in casting operation and are related in some
the green body. Velamakanni et al. (1994) way to the slip structure and properties.
investigated the rheological behavior of fil-
ter cakes consolidated from deflocculated 6.6.1 Pinholes
and flocculated alumina slips. They com- Pinholes arise from the presence of air
pressed freshly cast compacts to 2 % axial bubbles in the slip which are incorporated
strain and recorded the resulting stress re- into the cast resulting in large voids. They
184 6 Slip-Casting and Filter-Pressing

can be avoided by deaeration and the

avoidance of turbulent flow in the subse-
quent operations carried out on the slip.
Removal of air bubbles is favored by a low
viscosity slip, which can be accomplished
by control of the solids concentration and
the extent of flocculation. Some advanced
ceramic powders such as Si or Si 3 N 4 can
be subject to hydrolysis in aqueous disper-
sion which releases hydrogen or ammonia.
Figure 6-23. Polarized light optical micrograph of
6.6.2 Preferred Orientation of Anisometric slip-cast alumina. The maltese cross shape of the ex-
Particles tinction pattern indicates circumferrential orientation
of the grains (Roberts, 1995).
Preferred orientation of anisometric par-
ticles is a common occurrence when such
entities are present in slips partially or fully Often it is of higher density after firing
deflocculated or of high solids content. than the surrounding body (Basnett et al.,
Plate-like particles will tend to form ori- 1961). When the slip is poured into the
ented domains and these may be preferen- mold the first layer casts rapidly; slip im-
tially aligned in certain directions, usually pinging on this layer is not subject to the
parallel to the mold surface. Figure 6-23 same suction pressure and flows over it.
shows an optical micrograph of circumfer- However, near the point of impingement it
ential preferred orientation in a cast solid experiences a rapid change in direction.
alumina rod arising from the plate-like The local shear forces impose the particle
character of the particles (Roberts, 1995). orientation. The effect is worse with slips
The drying and firing shrinkage of green of higher fluidity.
bodies with preferred orientation is aniso-
tropic and this differential shrinkage can 6.6.4 Slip-Meets and Cast-Wreathing
lead to stresses and cracking in the cast
ware. Slip-meets and cast-wreathing are two
additional effects that arise from a rapid
initial casting. When the mold is being
6.6.3 Segregation
filled, the initial layer forms almost instan-
Segregation can occur, usually by sedi- taneously. Slip flowing round the mold in
mentation of larger particles, and is con- opposite directions and then meeting up
trolled by adjusting the degree of floccula- does not homogenize near the mold sur-
tion in the slip. In the traditional area, a face because the cast has begun to form.
common defect known as casting spot There can be an inhomogeneity at this
arises from a combination of segregation point. Wreathing arises when there is dif-
and preferred orientation. Casting spot oc- ferential casting and manifests itself as a
curs as a thin layer just below the surface wavelike pattern on the drained surface
that was in contact with the mold. It arises (Reed, 1988).
from a thin oriented layer of clay and mica
particles from which the more equiaxed
flux and filler particles have been excluded.
 6.8 References 185

6.7 Outlook 6.8 References

Slip-casting has for a long time been an Adcock, D. S., McDowall, I. C. (1957), J. Am.
important shaping and consolidation pro- Ceram. Soc. 40, 355-362.
Aksay, I. A., Schilling, C. H. (1984), in: Advances in
cess for ceramics and this will continue Ceramics, Vol. 9: Forming of Ceramics. Columbus,
into the future. However, the drive to more OH: American Ceramic Society, pp. 85-93.
rapid and economic production with bet- Andreasen, A. H. M., Anderson, J. (1930), KolloidZ.
50, 217-228.
ter control of green body structure for Anderson, P. I, Murray, P. (1959), /. Am. Ceram.
both traditional and advanced ceramics Soc. 42, 70-74.
will maintain the trend towards pressure- Barnes, H. A., Hutton, J. R, Walters, K. (1989), An
Introduction to Rheology. Amsterdam: Elsevier.
filtration. This offers considerable advan- Basnett, D., Roberts, E. W, Ashley, M. (1961), Re-
tages where very fine powders are being search Paper No. 497. London: The British
processed as is the case in the advanced Ceramic Research Association.
Buscall, R., Goodwin, J. W, Ottewill, R. H., Tadros,
ceramics area. T. F. (1982), J. Colloid Interface Sci. 85, 78-86.
The processing of ceramic powders in Buscall, R., Mills, P. D. A., Yates, G. E. (1986), Col-
suspension is of considerable importance loids Surf. 18, 341-358.
Buscall, R., Mills, P. D. A., Goodwin, J. W., Lawson,
because it allows, through the control of D. W. (1988), /. Chem. Soc. Faraday Trans. I 84,
interparticle forces, a high degree of uni- 4249-4260.
formity to be obtained in the green body. Buscall, R., McGowan, 1.1, Mumme-Young, C. A.
(1990), Faraday Discuss. Chem. Soc. 90, 115-127.
Mixing is facilitated and the packing char- Bushnell-Watson, Y. M. (1983), The Sedimentation,
acteristics can be systematically changed. Rheological and Filtration Characteristics of Mag-
Understanding of colloidal phenomena nesium Hydroxide and Other Suspensions, Ph. D.
Thesis. Sheffield: Dept. of Engineering Materials,
has progressed significantly in recent years The University of Sheffield.
through precise studies of well-character- Carman, P. C. (1956), Flow of Gases Through Porous
ized model colloids. These concepts are be- Media. London: Butterworth.
Cesareno III, X, Aksay, I. A. (1988), J. Am. Ceram.
ginning to be applied to ceramic systems of Soc. 71, 1062-1967.
industrial interest, but there is still a long Chang, J. C , Velamakanni, B. V., Lange, F. R, Pear-
way to go. In the near future we shall see son, D. S. (1991), J. Am. Ceram. Soc. 74, 2201-
2204.
improvements in the control of colloidal Chen, M., Russel, W. B. (1991), J. Colloid Interface
processing probably through an improved Sci. 141, 564-577.
understanding of the use of polymeric sta- Collyer, A. A., Clegg, D. W. (Ed.) (1988), Rheological
Measurement. London: Elsevier.
bilizers in determining the interparticle Cooper, C. R, Miskin, S. F. A. (1965), Proc. Br. Ce-
forces and hence green body structure. The ram. Soc. 3, 135-154.
use of electrostatic stabilization will be- Dal, P. H., Berden, W. J. H. (1968), Sci. Ceram. 4,
113-131.
come less attractive because of the signifi- Dal, P. H., Deen, W. (1958), 6th International Ce-
cant influence of the electrical double layer ramic Congress. Wiesbaden: DKG, pp. 219-244.
in increasing the effective particle volume Derjaguin, B. V., Landau, L. D. (1941), Ada Phys-
iochim. URSS 14, 633-652.
fraction for very fine powders. Dinger, D. R., Funk Jr., J. E., Funk, J. E. (1982), 13th
In this chapter the most important con- Annual Meeting of the Fine Particle Society. Chi-
cago: The Fine Particle Society.
cepts in colloid chemistry have been out- Diz, H. M. M., Rand, B. (1990), Br. Ceram. Trans. J.
lined, some modern aspects of rheology of 89, 77-82.
suspensions have been presented and the Diz, H. M. M., Rand, B., Inwang, I. B. (1990), Br.
Ceram. Trans. J. 89, 124-129.
effect of these in the control and monitor- Dullien, R A. L. (1992), Porous Media: Fluid Trans-
ing of casting/filtration-slips discussed. port and Pore Structure. San Diego: Academic
Press.
186 6 Slip-Casting and Filter-Pressing

Everett, D. H. (1988), Basic Principles of Colloid Sci- Low, P. F. (1961), Adv. Agron. 13, 269-327.
ence. London: Royal Society of Chemistry. Luther, E. P., Kramer, T. M., Lange, F. F , Pearson,
Fennelly, T. X, Reed, J. S. (1972a), J. Am. Ceram. Soc. D. S. (1994), J. Am. Ceram. Soc. 77, 1047-1051.
55, 264-268. Meakin, P. (1988), Annu. Rev. Phys. Chem. 39, 237-
Fennelly, T. J., Reed, J. S. (1972b), J. Am. Ceram. Soc. 267.
55, 381-383. Michaels, A. S., Bolger, J. C. (1962), Ind. Eng. Chem.
Ferry, I D . (1980), Viscoelastic Properties of Poly- Fundam. 1, 153-162.
mers. New York: John Wiley. Michaels, A. S., Baker, W. E., Bixler, H. I , Vieth,
Firth, B. A. (1976), J. Colloid Interface Sci. 57, 257- W. R. (1967), Ind. Eng. Chem. Fundam. 6, 25-40.
265. Napper, D. H. (1983), Polymeric Stabilization of Col-
Firth, B. A., Hunter, R. J. (1976a), /. Colloid Inter- loidal Dispersions. New York: Academic Press.
face Sci. 57, 248-256. Nikumbh, A. K., Schmidt, H., Martin, K., Porz, F.
Firth, B. A., Hunter, R. J. (1976b), J. Colloid Inter- (1991), / Mater. Sci. 26, 3649-3656.
face Sci. 57, 266-215. Parks, G. A. (1965), Chem. Rev. 65, 177-198.
Flegmann, A. W, Goodwin, J. W, Ottewill, R. H. Parks, G. A., de Bruyn, P. L. (1962), /. Phys. Chem.
(1969), Proc. Br. Ceram. Soc. 13, 31-45. 66, 967-973.
Frens, G., Overbeek, J. Th. G. (1972), J. Colloid Inter- Pashley, R. M. (1981), /. Colloid Interface Sci. 83,
face Sci. 38, 376-387. 531-546.
Fries, R. (1994), Unpublished results. Leeds: School Pashley, R. M., Israelachvili, J. N. (1984), /. Colloid
of Materials, The University of Leeds. Interface Sci. 97, 446-455.
Funk, J. E. (1984), in: Advances in Ceramics Vol. 9: Pasto, A. E., Neil, J. T., Quackenbush, C. L. (1984),
Forming of Ceramics. Columbus, OH: American in: Ultrastructure Processing of Ceramics, Glasses
Ceramic Society, pp. 76-84. and Composites. New York: Wiley, pp. 476-489.
Gebhardt, J. E., Fuerstenau, D. W. (1983), Colloids Persson, M. (1994), in: Surfactant and Colloid Chem-
Surf. 7,221-231. istry in Ceramic Processing. New York: Marcel
Hampton, J. H. D., Savage, S. B., Drew, R. A. L. Dekker, pp. 279-307.
(1988), J. Am. Ceram. Soc. 71, 1040-1045. Persson, M., Hermansson, L., Carlsson, R. (1983),
Hampton, J. H. D., Savage, S. B., Drew, R. A. L. in: Ceramic Powders. Amsterdam: Elsevier, pp.
(1992), /. Am. Ceram. Soc. 75, 2726-2732. 735-742.
Hoffmann, R. L. (1972), Trans. Soc. Rheol. 16, 155- Prestidge, C , Tadros, T. F. (1988), Colloids Surf 31,
173. 325-346.
Horn, R. G. (1990), J. Am. Ceram. Soc. 73, 1117- Pugh, R. X (1994), in: Surfactant and Colloid Chemis-
1135. try in Ceramic Processing. New York: Marcel
Hunter, R. J. (1987), Foundations of Colloid Science, Dekker, pp. 127-193.
Vol. 1. Oxford: Oxford University Press. Rand, B., Melton, I. E. (1977), /. Colloid Interface
Hunter, R. J. (1981), Zeta Potential in Colloid Sci- Sci. 60, 308-320.
ence. New York: Academic Press. Rand, B., Diz, H. M. M., Li, X X, Inwang, I. B.
Johansen, P. G., Buchanan, A. S. (1957), Aust. J. (1987), 5c/. Ceram. 14, 231-236.
Chem. 10, 398-403. Rand, B., Pekenc E., Goodwin, X W, Smith, R. W.
Kendall, K. (1988), Mater. Forum 11, 61-70. (1980), J. Chem. Soc. Faraday Trans. 176, 225-235.
Kerkar, A. V., Henderson, R. J. M., Feke, D. L. Reed, X S. (1988), Introduction to the Principles of
(1990a), J. Am. Ceram. Soc. 73, 2879-2885. Ceramic Processing. New York: Wiley.
Kerkar, A. V., Henderson, R. J. M., Feke, D. L. Roberts. E. W. (1995), Private communication.
(1990b), /. Am. Ceram. Soc. 73, 2886-2891. Russel, W B. (1987), The Dynamics of Colloidal Sys-
Krieger, I. M. (1972), Adv. Colloid Interface Sci. 3, tems. Wisconsin: The University of Wisconsin
111-136. Press.
Krieger, I. M., Dougherty, T. J. (1959), Trans. Soc. Russel, W B., Saville, D. A., Schowalter, W. R.
Rheol. 3, 137-152. (1989), Colloidal Dispersions. Cambridge: Cam-
Krieger, I. M., Eguiluz, M. (1976), Trans. Soc. Rheol. bridge University Press.
20, 29-AS. Safran, S. A., Webman, I., Grest, G. S. (1987), Phys.
Kruyt, H. R. (1952), Colloid Science, Vol. 1. Amster- Rev. A 32, 506-511.
dam: Elsevier. Scheidegger, A. E. (1974), The Physics of Flow
Lange, F. F. (1984), J. Am. Ceram. Soc. 67, 83-89. Through Porous Media. Toronto: University of
Lange, F. F. (1989), J. Am. Ceram. Soc. 72, 3-15. Toronto Press.
Lange, F. F., Miller, K. T. (1987), Am. Ceram. Soc. Seaton, A. A., Glandt, E. D. (1987), /. Chem. Phys.
Bull. 66, 1498-1504. 86, 4668-4677.
Lange, F. F , Davis, B. A., Wright, E. (1986), / Am. Smith, P. A., Haber, R. A. (1992), J. Am. Ceram. Soc.
Ceram. Soc. 69, 66-69. 75, 290-294.
Li, I, Rand, B. (1984), Br. Ceram. Trans. J. 83, 186- Smith, P. A., Kerch, H., Krueger, S., Long, G. G.,
190. Keller, X, Haber, R. A. (1994), J. Am. Ceram. Soc.
77, 1777-1782.
 6.8 References 187

Sonntag, R. C , Russel, W. B. (1987), /. Colloid Inter- General Reading

face Set 116, 485-489.
Sumita, S., Rhine, W. E., Bowen, H. K. (1991), J. Am. Aksay, I. A., Schilling, C. H. (1984), in: Advances in
Ceram. Soc. 74, 2189-2196. Ceramics, Vol. 9: Forming of Ceramics. Columbus,
Tadros, Th. F. (1989), Prog. Colloid Polym. Sci. 79, OH: Am. Ceram. Soc, pp. 85-93.
120-127. Barnes, H. A., Hutton, J. R, Walters, K. (1989), An
Tiller, F. M., Tsai, C. D. (1986), J. Am. Ceram. Soc. Introduction to Rheology. Amsterdam: Elsevier.
69, 882-887. Everett, D. H. (1988), Basic Principles of Colloid Sci-
van Olphen, H. (1977), An Introduction to Clay Col- ence. London: Royal Society of Chemistry.
loid Chemistry. New York: Wiley. Funk, J. E. (1984), in: Advances in Ceramics, Vol. 9:
Velamakanni, B. V., Lange, F. F. (1991), J. Am. Ce- Forming of Ceramics. Columbus, OH: Am. Ceram.
ram. Soc. 74, 166-172. Soc, pp. 76-84.
Velamakanni, B. V., Chang, J. C , Lange, F. R, Pear- Goodwin, J. W. (1990), Am. Ceram. Soc. Bull. 69,
son, D. S (1990), Langmuir 6, 1323-1325. 1694-1699.
Velamakanni, B. V, Lange, F. R, Zok, F. W., Pear- Horn, R. G. (1990), J. Am. Ceram. Soc. 73, 1117 —
son, D. S. (1994), J. Am. Ceram. Soc. 77, 216-220. 1135.
Verwey, E. J. W, Overbeek, J. Th. G. (1948), The The- Hunter, R. J. (1987), Foundations of Colloid Science,
ory of Stability of Lycophobic Colloids. Amster- Vol.1. Oxford: Oxford University Press.
dam: Elsevier. Hunter, R. J. (1981), Zeta Potential in Colloid Sci-
Walker, E. G., Dinsdale, A. (1959), in: The A. T. Green ence. New York: Academic Press.
Book. Stoke on Trent, U.K.: British Ceramic Re- Kruyt, H. R. (1952), Colloid Science, Vol. 1. Amster-
search Association, pp. 142-150. dam: Elsevier.
Walters, K. (1975), Rheometry. London: Chap- Lange, R F. (1989), J. Am. Ceram. Soc. 72, 3-15.
man & Hall. Parks, G. A. (1965), Chem. Rev. 65, 177-198.
Walters, K. (Ed.) (1980), Rheometry: Industrial Appli- Pugh, R. I, Bergstrom, L. (Ed.) (1994), Surfactant
cations. New York: Wiley. and Colloid Chemistry in Ceramic Processing. New
Wiese, G. R., Healy, T. W. (1974), /. Colloid Interface York: Marcel Dekker.
Sci. 51, 427-433. Reed, J. S. (1988), Introduction to the Principles of
Woutersen, A. T. J. M., de Kruif, C. G. (1991), J. Ceramic Processing. New York: Wiley.
Chem. Phys. 94, 5739-5750. Russel, W. B. (1987), The Dynamics of Colloidal Sys-
Yates, D. E., Levine, S, Healy, T. W (1974), J. Chem. tems. Wisconsin: The University of Wisconsin
Soc. Faraday Trans. I 70, 1807-1818. Press.
Yoon, R. H., Salman, T., Donnay, G. (1979), /. Col- Russel, W.B., Saville, D. A., Schowalter, W. R.
loid Interface Sci. 70, 483-493. (1989), Colloidal Dispersions. Cambridge: Cam-
bridge University Press.
Tadros, Th. R (1985), Chem. Ind. 4, 210-218.
Tiller, F. M., Tsai, C. D. (1986), J. Am. Ceram. Soc.
69, 882-887.
Walters, K. (1975), Rheometry. London: Chap-
man & Hall.
7 Tape Casting
Hans Hellebrand

Siemens AG, Zentralabteilung Forschung und Entwicklung, Munich, Germany

List of Symbols and Abbreviations 190

7.1 Introduction 192
7.2 Applications and Their Demands 194
7.2.1 Ceramic Monolayers 194
7.2.1.1 Thin Sheet Capacitors 194
7.2.1.2 Piezoelectric and Electrostrictive Monolayers 195
7.2.1.3 Electronic Substrates 196
7.2.2 Multilayer Components 197
7.2.2.1 Lamination of Green Tapes 198
7.2.2.2 Heterogeneous Multilayer Composites 200
7.2.3 Special Components 204
7.3 Materials for Casting Ceramic Tapes 207
7.3.1 Ceramic Powder Materials 207
7.3.1.1 Powder Preparation 208
7.3.1.2 Powder Processing 212
7.3.1.3 Packing of Powder Particles 214
7.3.2 Slurries for Tape Casting 216
7.3.2.1 Polymer Binders 217
7.3.2.2 Plasticizers 221
7.3.2.3 Solvents 221
7.3.2.4 Functional Additives 226
7.3.2.5 Rheology of Tape Casting Slurries 235
7.3.2.6 Slurry Preparation 237
7.3.3 Assessment of Properties 238
7.4 The Tape Casting Process 242
7.4.1 The Practice of Tape Production 243
7.4.1.1 Tape Casting Methods 243
7.4.1.2 Tape Coating Methods 247
7.4.2 Drying of Ceramic Tapes 251
7.4.3 Characterization of Green Tapes 254
7.4.4 Binder Burn-Out 257
7.5 Conclusions and Outlook 258
7.6 Acknowledgements 260
7.7 References 260

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
190 7 Tape Casting

List of Symbols and Abbreviations

D thickness of a dried tape

d, dl9 d2 powder particle diameter, dx lowest value of d, d2 highest value of d
d50 median value of particle size distribution (50% size o n the cumulative
frequency curve)
£ T (30) empirical polarity parameter of solvents (in kJ/mol)
Gb weight loss of a sample tape of defined volume due to binder burn-out
Gp weight of a sample tape of defined volume after binder burn-out (ceramic
powder weight)
h gap height of the casting head
jc critical current density of a superconductor
L length of the casting g a p
pcap capillary pressure (positive a n d negative values)
Ap pressure difference in the casting head g a p
pHiep p H value at the isoelectric point (£ potential equals zero)
pHzpc p H value at the zero point of charge (no net charge)
r radius of a model capillary
s = f(d) distribution function of particle diameter d
Tc critical temperature of superconductivity
Tg glass transition temperature of polymers
V geometrical volume of a sample tape
Va potential energy of interaction caused by attractive van der Waals forces
Vr (electrostatic) repulsive electrostratic potential energy of interaction
Vr (steric) repulsive steric potential energy of interaction
v p , vh9 v% volume fractions (%) of powder, binder and gas, respectively, in a green
tape
v0 relative velocity of casting unit and tape substrate
di?/dx shear rate in a slurry
^s/v> ^S/L surface free energy of a solid particle surrounded by its own vapor or by
a liquid, respectively
w surface free energy per unit area of particle surface
a factor regarding tape shrinkage due to drying
y surface tension per unit length of solids and liquids
7s/v> ys/L> 7L/V surface tension per unit length related to solid/vapor, solid/liquid and
liquid/vapor contact, respectively
3 liquid/solid contact angle
C electric potential of charged particles with respect to the bulk of the
slurry (zeta potential)
Y\ slurry viscosity
rjo viscosity of the organic system only
rjTCl relative viscosity (rj/rj0)
Q specific density of ceramic bulk material
£p crystallographic specific density of ceramic powder particles
 List of Symbols and Abbreviations 191

Qh specific density of dried binder film

T shear stress in the slurry system
A" anionic electrolyte ion
BBP butyl benzyl phthalate
BET specific surface measurement method of solids by determination of gas
adsorption isotherms (theory by Brunauer, Emmett, Teller)
CLA center line average height (defined in Sec. 7.2.1.3)
c.m.c. critical micelle concentration
CVD chemical vapor deposition
C+ cationic electrolyte ion
DBP dibutyl phthalate
DLVO theory of electrostatic stabilization of dispersions
DOP dioctyl phthalate
ESA electrokinetic sonic analysis, measuring £ potential
IC integrated circuit
LSI large-scale-integrated
Me metal ion in the lattice
MEK methyl ethyl ketone
MW molecular weight
MCFC molten carbonate fuel cell
PCD particle charge detector
PEG polyethylene glycol)
PEMA polyethylmethacrylate
PVA poly(vinyl alcohol)
PVAc poly(vinyl acetate)
PVB poly(vinyl butyral)
PMAA polymethacrylic acid
PMMA polymethylmethacrylate
PTFE polytetrafluorethylene
PZT lead-zirconate-titanate ceramics
SEM scanning electron microscope
SOFC solid oxide fuel cell
TASA total apparent surface area calculated from size distributions
192 7 Tape Casting

7.1 Introduction a flat substrate, then evenly spread, and

the solvents subsequently evaporated. This
Ceramic technology enables us to ex- process can be illustrated by a flow chart
ploit the properties of crystalline materials according to Fig. 7-1. The result is a so-
by forming and shaping elements or struc- called 'green ceramic tape' containing
tures that are optimally adapted to the binders and ceramic particles, the shape
problem at hand. The flat, thin ceramic being a virtually two-dimensional, i.e., ex-
layers, for instance, required in the two tremely narrow gauge sheet or strip.
principal electroceramics markets, i.e., The slurries are closely related to pig-
electronic substrates and multilayer capac- mented paints and varnishes. Indeed, in
itors, are produced using various rival terms of ingredients, the main difference
technologies. Often the main criteria for between slurries for tape casting and paints
selection are production costs, a pre-exist- is that in slurries the powder content
ing familiarity with a given process, or should be as high as possible, in paints and
even the availability of suitable machines.
Table 7-1 lists the main processes com-
peting in the market for thin ceramic layer
fabrication as described in the literature,
together with some remarks characterizing
results and typical applications in flat sheet
production (Fukuura and Hirao, 1989;
Hyatt 1986; Prasad 1988; Mistier, 1991).
Other methods less frequently used for this
purpose, e.g., injection molding, spraying,
and isostatic pressing (Prasad, 1982) are
only mentioned here, as are combined
methods, e.g., extrusion + calendering
(Schat, 1970).
Compared with these mostly traditional
ceramic forming methods, tape casting is a
relatively new art, which was first used for tape casting
the production of ceramic sheet capacitors
in the 1940s (Howatt etal., 1947). In a
drying
1952 patent, it was clearly linked with the
well-known slip casting method, in which characterization
the solvent is removed by absorption in a
porous plaster of Paris mold (Howatt,
1952). There has been considerable pro-
gress in what is nowadays referred to as
"tape casting". Indeed, today it is now an
accepted high-precision method for high-
tech ceramics.
Tape casting may be defined as a process
in which a slurry of ceramic powder, Figure 7-1. Flowchart of the general production pro-
binder and slovents is poured or 'cast' onto cess of ceramic tape-based components.
 7.1 Introduction 193

Table 7-1. Processes producing thin ceramic layers.

Process Characterization Application

Dry pressing High packing factor, low binder content, For non-flexible sheets of small
low cost total area
Rough surface, density fluctuations, pore Thickness > 250 urn
size variations
Slip casting No binder content, large areas, curved and For non-flexible monolayered
structural surfaces possible, uniform structures
microstructure Thickness > 100 um
Stable dispersions necessary, batch process
only, no laminates possible, low casting
rates
Extrusion/extrusion plus No sedimentation/segregation, continuous For mass production of flexible
calendering process, self-supporting sheets, smooth sheets of limited width
surfaces Thickness (40) 100 um-1000 um
High binder content, length/lateral shrink-
age variations, warping, abrasion
Dry powder roll High packing factor, low binder content, For mass-production of flexible
compaction no sedimentation, self-supporting sheets sheets of standard thickness
Poor thickness uniformity, relatively high (thick film substrates)
process costs, difficulties with thickness Thickness 100 um-1500 um
variations
Screen printing Thin structured layers, relatively smooth Preferred thick film technology
surfaces for mono- or multilayer struc-
Generally discontinuous, small areas, tures
special ink systems, high binder content Thickness 3 um-30 (100) um
Electrophoretic deposition No binders necessary, structurable layers For coating of structured areas
Stable slurries of fine powders needed, (substrates)
discontinuous process, no self-supporting Thickness < 150 um
tapes
Tape casting Moderate to high packing factor, water- For flexible tapes in mono- and
and organic-based compositions, continu- multilayer technology
ous process possible, smooth surface, small For small and large devices
and large quantities, self-supporting tapes thickness 10 um-1000 um
Relatively high binder content, sophisti-
cated slurry compositions, drying and
debindering problems

varnishes the pigment content is as low as ing the use of solvent/binder systems and it
consistent with performance. may be set up as a continuous process to
Comparing the different process charac- manufacture large quantities of tapes or as
teristics from Table 7-1, it will be seen that, a small discontinuous laboratory route for
among all technologies described, tape cost-effective testing applications. The
casting offers the widest thickness range considerable literature published on this
and the thinnest self-supporting layers. It topic during the last decade clearly demon-
has no fundamental restrictions concern- strates the importance of the method, as
194 7 Tape Casting

well as the complexity of the associated from the peculiarities within their fields of
problems. Its versatile application in dif- employment. At the same time it also pre-
ferent fields of ceramic technology has sents a simple method for describing their
given rise to numerous specific demands, characteristics and quantifying structural
e.g., concerning powder and slurry prepa- differences.
ration, tape casting process and facilities,
and green tape properties.
7.2.1 Ceramic Monolayers
Starting with the description of the vari-
ous types of applications for thin ceramic This group includes all applications for
sheets, the following sections of this chap- which thin, flat, even, and unstructured el-
ter will deal with the different aspects of ements are produced by directly consoli-
sheet production using tape casting meth- dating the green cast tapes in a subsequent
ods. sintering route. The main task of the or-
ganic components in this monolayer tech-
nology is confined to guarantee optimum
7.2 Applications and Their casting process conditions and to provide
Demands homogeneous tapes with green density and
sintering activity as high as possible. Since
Listing only the most common ceramic little or no flexibility or plasticity of the
tape applications, and classifying them green tapes is needed for further process
according to fields of employment, it be- steps, binder content in this kind of green
comes clear that no single, uniform type of tape can be minimized and the type of
tape with standardized characteristics can binder system (water-based or organic-
meet all resulting requirements (Table 7-2). based) may be selected according to other
Consequently, tapes need to have a vari- conditions such as production costs or en-
able range of properties, enabling the vironmental considerations. Naturally, the
choice of a type ideally suited to the func- organic components should provide suffi-
tion at hand (Ettre and Castles, 1972). cient strength and dimensional stability to
This section describes the special prop- allow handling of the green tapes prior to
erties required in green tapes and resulting sintering. Beyond these general factors,
certain peculiarities relating to the many
special applications of such ceramic mono-
Table 7-2. Applications for ceramic tapes.
layers also have to be allowed for.
Field of Examples of application
employment
7.2.1.1 Thin Sheet Capacitors
Monolayers Capacitors, piezo-membranes,
substrates Because of their simple way of fabrica-
Multilayer Capacitors, piezo-actuators, tion (Goodmann, 1986), cast ceramic
laminates substrates sheets were first used as thin ceramic
Special porous structures for filters, monolayer capacitors (Howatt et al., 1947)
components catalysts and fuel cells and vacuum tube spacers (Thompson,
Gradient structures 1963). For elements of a given area, capac-
Tapes with preferred orientation ity is inversely proportional to the thick-
of particles for piezoelectrics and
superconductors
ness of the dielectric material. Producers
therefore seek to use the thinnest possible
 7.2 Applications and Their Demands 195

gauge sheet. There are, however, two seri- cutting lines

ous obstacles to thickness reduction.
First, such simple and extremely cheap
capacitors have to be produced as large
area sheets (e.g., 150 x 150 mm 2 ), the small
single capacitor units being separated only
after the entire sheet has been sintered and
electroded (Fig. 7-2). The brittleness of the
sintered ceramic tapes complicates the
handling of large thin areas, thus limiting
thickness reduction to about 100 \xm-
200 |im. fully electroded sintered sheets
Second, breakdown and electrical deg- Figure 7-2. Thin sheet capacitor production. The sin-
radation of the ceramic material is more gle capacitor units are produced by cutting them from
probable, if the electrical field strength ris- a stack of sintered and electroded thin ceramic sheets
es to values of about 3 kV/mm or more. of large area.
This effect is especially important, since in
an efficient production line these capaci-
small thickness tolerances within the entire
tors are normally electroded right up to the
batch.
edges, making edge shorts more likely to
happen.
7.2.1.2 Piezoelectric and Electrostrictive
For these reasons, thinner tapes for
Monolayers
higher capacity values are used predomi-
nantly in the multilayer technology, which Piezoelectric thin layers are used for ac-
will be described in Sec. 7.2.2. tuator applications like microphone mem-
In order to avoid warping of the sheets branes, buzzers, ultrasonic cleaners or
during sintering, some 30 to 50 green tapes bending elements. Such monolayered flat
are stacked with packing powder in be- structures are manufactured at a great
tween, e.g., presintered, milled powder of variety of geometrical shapes (thick, thin,
the same material or another inert powder. round, square or ring-shaped). If electri-
This packing powder prevents reactions cally stimulated, they produce small and
between the sheets and allows the separa- very fast displacements or, inversely, an
tion of the tapes after firing. Subsequently, electric field signal corresponding to a me-
the sheets are electroded by dipping, spray- chanical stress.
ing, or evaporating and reassembled on a The direction of the displacement or the
waxed block for cutting. Sawing with a actuated electric signal are related to a re-
diamond plated circular saw has proved an sulting remanent polarization within the
effective way of separating capacitors with ceramic element. This is established by a so
well-defined capacity values. called 'poling process', where randomly
From these remarks, certain require- distributed ferroelectric domains are sta-
ments regarding tape properties for mono- tistically aligned to the applied electric
layer capacitors may be derived. They field. This poling itself is always associated
should be free of cracks and flaws, and with some geometrical deformation of the
show uniform high density over the whole membranes, both in direction of the poling
area, little, if any, warping or bending, and field (elongation) and perpendicular to it
196 7 Tape Casting

(contraction), which must not result in Sintering, therefore, is normally carried

warping or bending of the thin sheets. out in a stacked assembly and under a
Perfectly flat, homogeneous, dense, and moderate weight load using very fine-sized
untextured sintered layers are therefore re- packing powder as a separator (Brewer,
quired for this application - a criterion af- 1990), as described in the section on mono-
fecting likewise the quality of the green layer capacitors.
tapes used. These properties may be
achieved either by using a high standard
7.2.1.3 Electronic Substrates
tape casting process or by carefully press-
ing the green tapes prior to sintering A great deal of progress in tape casting
(Prasad and Panchapakesan, 1982). This over the last decades has been stimulated
kind of "postcasting pressing" was pro- and promoted by the application of this
posed in 1978 by Biggers et al. (1979) and method to the production of substrates for
may be compared in its effect with what thick- and thin-film technology (Shane-
happens during lamination in multilayer field and Mistier, 1971). Fabrication of
technology (see Sec. 7.2.2). substrates by tape casting - mainly of
Frequently such a piezo-active mem- A12O3 - in the thickness range of 0.2 to
brane is stuck to a layer of inactive materi- 1 mm is now a standard mass production
al (ceramics, metal, etc.) to form a so- process of outstanding importance. Al-
called bending unit reproduced schemati- though significantly more expensive, ce-
cally in Fig. 7-3. If the active part is electri- ramic substrates have prevailed over poly-
cally stimulated, the very small piezoelec- mers and glasses, because of the demand in
tric displacements in the micrometer range the rapidly developing thick-film technolo-
may be magnified between 20 and 40 times gy for substrates with higher temperature
(usable, for instance, in a piezoelectric resistance and with chemical inertness.
switching element). This principle of dis- Meanwhile, the quality of electronic sub-
placement increase is very common with strates has increased extraordinarily, re-
the temperature controlled bi-metal sulting today even in their joint use for
switches. Evidently, flaw-free forming of thick- and thin-film applications upon the
such compounds of about 100 to 200 jim same substrate to form hybrid integrated
thickness, often with large lateral dimen- circuits (Belosi etal., 1980; Ziegler, 1986).
sions in the range of several centimeters, Many of the imperative requirements re-
requires the production of absolutely flat, sult from the various process steps in-
untextured membranes with very even sur- volved in these technologies. For screen
faces. printing, for instance, sufficiently smooth
uncambered substrates with CLA value
^0.5 |im (center line average height =
height of a rectangle with an area equiva-
electrodes piezoelectric layer lent to the area of the profile peaks, count-
ed from the average line of the substrate
profile) and with adequate mechanical
strength ( « 500 MPa) are essential, while
adhesive firing processes, bonding, or brazing re-
contraction
quire resistance to aggressive chemicals
Figure 7-3. Piezoelectric bending element. and high temperature (^1000°C). To en-
 7.2 Applications and Their Demands 197

sure exact punching of positioning holes the case of ceramic sheet structures is fre-
and of vias, unfired green tapes should dis- quently closely related to a decrease in lay-
play good dimensional stability. At the er thickness.
same time deviations in shrinking rates Normally this tendency is limited by the
should be minimal (0.1-0.5% over the brittleness and the low strength of such
whole area of one piece as well as from individual self-supporting elements. Espe-
piece to piece) to facilitate the use of auto- cially in the two main segments, i.e., capac-
matically controlled tools. itors and substrates, where higher degrees
In thin film technology, e.g., with tanta- of integration demand smaller areas for
lum and tantalum nitride circuits, extreme- unit capacity or advanced space-saving
ly smooth surfaces are needed. With fine- cross over techniques of leads, the so-
grained alumina substrates (grain size called multilayer technology has achieved
« 1 jum) of very low porosity ( < 0.5 vol. %) a high standard. It was first proposed by
CLA-values of less than 0.1 j^irn have been Howatt et al. (1947) with the aim of build-
achieved (Mistier et al., 1974) and 0.02 jim ing up low-weight, small-volume ceramic
is possible, thus permitting the use of sub- capacitors for high temperature applica-
strates in the 'as fired' state (Mistier et al., tions and has been subsequently developed
1974; Cooper et al., 1987). Conductor and to an industrial production method by
resistor materials are evaporated or sput- Stetson and Schwartz (1961).
tered directly on the substrate surface Meanwhile, this multilayer technology
and structured by photolithographic tech- is used in nearly all other fields of ceramic
niques, thus avoiding high-cost polishing tape application (Prasad, 1988; Boch and
steps (Feil, 1986). Chartier, 1988). Piezoelectric actuators for
Finally, higher frequency applications small driving voltages are produced (Wers-
necessitate lower permittivities and losses, ing et al., 1986b) as well as high tempera-
while the heat produced in large-scale-in- ture ceramic fuel cells and multilayer chip
tergrated (LSI) circuits needs to be trans- varistors (Utsumi, 1991). Other heteroge-
ferred to the environment by materials neous, composite structures, such as in-
with high thermal conductivity (Schwartz, ductors (Takaya etal., 1990), heat ex-
1988). Consequently, materials like BeO changers (Heinrich etal., 1987a, b), pho-
(Lynch et al., 1989), A1N (Brunner, 1988; tovoltaic solar cells (Fiori and De Portu,
Descamps etal., 1994), or mullite (Char- 1986), or ZrO 2 -Al 2 O 3 laminated com-
tier and Boch, 1988; Fiori and De Portu, pounds (Besson etal., 1987; Boch etal.,
1986) have been introduced which has also 1986) for substrates with increased
had effects on the process of substrate pro- strength and toughness are further exam-
duction, resulting mainly from the special ples for this technology.
chemistry of these powders (see Sec. 7.3.1). The production of multilayers involves
laminating stacked green ceramic layers by
heat compression, with the aim of forming
7.2.2 Multilayer Components a monolithic composite featuring specified
structural properties. It is essential, there-
The main driving force in the develop- fore, for the individual tapes to have ade-
ment of ceramic multilayer technology was quate sealing properties, in addition to the
the continuous demand for miniaturiza- multiple demands described in the preced-
tion of electronic components, which in ing sections (Prasad, 1982).
198 7 Tape Casting

Multilayer technology is predominantly

used to build up composites with a hetero-
geneous structure, consisting of ceramics tape 1
and electrode metals. To determine the
properties required in ceramic tapes suit-
able for this technology, the basic process
of laminating two unstructured green ce-
ramic layers in a die by application of a
resonable amount of pressure and heat will tape 2
be considered first. Subsequently, the spe-
cial problems arising with heterogeneously
composed multilayers will be dealt with.
a) 100 pill

7.2.2.1 Lamination of Green Tapes

The structure of a green tape can be de-
fined in terms of the specific volume frac-
tions of three different components
(Chartier and Bruneau, 1993): the ceramic
tape 1
powder particles (vp in vol.%), the inor-
ganic film forming polymer binders sur-
rounding the particles (vh in vol.%) and
the voids, filled with gaseous residues of
the solvents and air (rg in vol.%) with
vp-\-vh + vg = 100%. For ideal lamination, tape 2
the boundary between two adjacent tapes
should be undetectable after compression,
i.e., in a cross sectional view all compo-
nents should be uniformly distributed in 100 pin
the transitional region of the two surfaces
Figure 7-4. Cross-sectional view of the interface be-
and in the interior of the tape (Fig. 7-4). tween two laminated tapes. After lamination the in-
This can only be achieved if connection terface should no longer be detectable (a). Insufficient
of the tapes is not implemented by sticking lamination is indicated by visible line defects (b).
or gluing, but by a process comparable in
its effect to joining or welding. Individual
particles at the surfaces of the tapes in con- required amount of densification by parti-
tact have to move and interpenetrate cle rearrangement (Utsumi, 1991; Gardner
within a thin, surface-near region, thus and Nufer, 1974). In this latter case, a neg-
smoothing the micro-roughness of the ligible initial phase of densification would
tapes and forming a homogeneous com- be followed by the onset of an overall
pound. The mobility is guaranteed by us- undirected hydrostatic pressure with no re-
ing a sufficient amount of a thermoplastic sultant driving forces for further reorienta-
binder (vh) to achieve the desired bonding tional movement.
effect, but not so much that the remaining In Fig. 7-5 the effect of these two limit-
gas volume (vg) is insufficient to allow the ing conditions for binder content and vol-
 7.2 Applications and Their Demands 199
before pressing after pressing Figure 7-5. Schematic view
of the interface of two
green ceramic tapes during
poor gliding of the particles
poor densification lamination. Three tapes
only tacking at the with different binder vol-
interface of the tapes ume fractions (vh) and,
hence, gas volume fractions
(vg) are shown while the
a) very low binder content Vb and very high gas volume vg packing of the powder par-
ticles is kept approximately
constant. Tapes with ex-
tape 1 poor dislocation of the particles tremely low (a) and ex-
poor densification tremely high (b) binder
only tacking at the contents result in laminates
tape 2 interface of the tapes with insufficient interpene-
tration of powder particles
at their interface. Optimum
b) very high binder content vb and very low gas volume vg
conditions (c) are charac-
terized by a sufficiently
high compressibility and,
tape 1 good gliding and dislocation simultaneously, by a high
of the particles mobility of the particles
interpenetration of particles
during lamination to guar-
tape 2 at the interface of the tapes
antee their reorientation
and interpenetration at the
c) optimum binder content Vb and optimum gas volume vg seams of the tapes.

ume fraction of gas on the processes taking tion parameters, they have to be defined
place in or near the tape seams are repre- empirically from case to case and are
sented schematically, and are contrasted therefore often a proprietary secret. As a
with the case of optimized green tape com- rule of thumb, characteristic values for
position. It is evident, that only in this op- green tape compositions best suited for
timized case will the interface between the producing multilayer laminates fall ap-
two tapes be undetectable after burn-out proximately within the following ranges
of the binder. In both other cases, a (see also Fig. 7-10 in Sec. 7.2.2.2):
boundary layer with a more or less en-
larged pore fraction occurs and is unlikely 40vol.%<v p <55vol.%
20 vol. % < vh < 40 vol. %
to readily disappear during sintering.
10vol.%<t; g <35vol.%
Given these restrictive requirements on
with vp + vh + vg = 100 vol. %
green tape characteristics, together with
other factors, such as flexibility, dimen- Pure laminates of several ceramic layers,
sional stability, sinterability, or binder however, represent only a very small por-
burn-out behavior, not to mention the tion of multilayer products. They are used,
constraints resulting from process parame- for instance, as devices built up by lami-
ters involving pressure and temperature nating thinner layers with preferred orien-
limits, compromise on tape specifications tation of the particles (see Sec. 7.3.2)
is inevitable. Since these will in turn de- (Watanabe etal., 1991), as tough sub-
pend very much on materials and produc- strates of alternately stacked A12O3 and
200 7 Tape Casting

ZrO 2 layers (Chartier etal., 1988), or as ceramic

interlayers between tapes of varying com-
position for the matching of thermal ex-
pansion coefficients in fuel cell technology
(SchieBl etal., 1991). Nevertheless, this
basic process step in multilayer production
should be performed very carefully, since it
might be the origin of troublesome failures
in these and all other ceramic laminates metal layers termination
(Pepinetal., 1989). Figure 7-6. Schematic diagram of a ceramic multi-
layer capacitor.
7.2.2.2 Heterogeneous Multilayer
Composites
pacitor (Fig. 7-6) reveals its principal con-
The most typical applications for cast structive feature: the total encapsulation of
ceramic tapes are multilayer capacitors all internal capacity-determining elec-
and electronic substrates with integrated trodes within the ceramic material and
passive functional components and three- their protection from detrimental environ-
dimensional crossover of leads. ment influences.
The ceramic multilayer capacitor offers Driving field strength can thus be in-
the possibility of combining a whole spec- creased to a multiple of that of mono-
trum of adjustable ceramic material prop- layers. At a given driving voltage, this is
erties with the advantages of a monolithic, equivalent of achieving an extreme thick-
low failure rate, and automatically posi- ness reduction of the ceramic sheet and,
tionable element with high specific volume hence, an increase of volume capacity. A
capacity. dielectric layer thickness of about 20 jam is
Unsintered green ceramic tapes are pro- technical standard today with the potential
vided with electrode patterns by screen of about 10-15 jim likely to the near fu-
printing or spraying a metal paste, then ture.
stacked and pressure-laminated into a Because of the total encapsulation of the
monolithic structure and cofired at an ap- electrodes, however, it will no longer be
propriate temperature (Hagemann et al., possible to adjust capacity values after sin-
1983/84; Kahn et al., 1988). The thickness tering. On commercial production lines,
of the printed patterns can cause problems therefore, narrow tolerances for layer
during lamination if a complete flaw-free thickness, for the positioning of electrode
embedding in the ceramic layer material patterns, and for compositional fluctua-
has to take place. It is not uncommon for tions have to be guaranteed, since these
this to result in density fluctuations in the factors directly affect process output.
compressed green compounds, which may Thickness tolerances and criteria deter-
give rise to cracks during sintering. mining the homogeneity of green tape
After sintering, the layers are electrically structure (vp9 vh, vg) will become corre-
interconnected in parallel, while mechani- spondingly more stringent.
cally the strength of a thin single layer is The multilayer technology was very
multiplied to that of the thicker com- soon applied to alumina substrates. Lay-
pound. A scheme of such a multilayer ca- ered structures were produced to allow
 7.2 Applications and Their Demands 201

space-saving crossover between conduc- ter to guarantee matching of vias and elec-
tors in different planes (Stetson, 1965; trode patterns in different planes (Young,
Schwartz and Wilcox, 1967; Blodgett, 1986), Plasticity, therefore, must be com-
1980). These efforts have now culminated paratively low and the lamination process
in a large-scale hybrid packaging technolo- should be carried out in a die with high
gy, where hundreds of active (transistors, stability and precise geometry.
integrated circuits) and passive (resistors, Recently, piezoelectric and electrostric-
conductors, capacitors) functions are inte- tive multilayer structures have been gaining
grated within and upon one single sub- in importance, since they offer the possibil-
strate. ity of cumulating the very small piezoelec-
This technology comprises the applica- tric movements of a great number of single
tion of conductors on the green ceramic thin layers without increasing their driving
tapes, punching and filling vias with metal voltage. Both the high forces generated by
inks for the interconnection of leads in the the single units, and their very short delay
different planes and finally the laminating time between electrical signal and mechan-
and sintering of these tapes to form a ical response (measured in microseconds)
composite of 10-30 layers (Fig. 7-7). The are more or less conserved. Actuators, sen-
upper surface of such a substrate is then sors, transducers or sonar devices are be-
equipped with thick or thin film circuits, ing produced with up to several hundred
discrete capacitors, inductors, IC-chips, layers of 20-200 |im thickness and with
and heat sinks, while the bottom surface cofired electrodes between them as de-
is provided with a very high number of scribed in the section on multilayer capaci-
pins for the purpose of interconnection tors (Dayton etal., 1984; Wersing e t a l ,
(Schwartz, 1988). During handling and 1988; Takahashi, 1986; Ohde et al., 1988;
pressing the structures must not shift later- Lubitz and Hellebrand, 1990; Lubitz et al.,
ally within the range of tenths of a millime- 1991; Lubitz, 1995).
Since the elongation of the units is pro-
via hole external resistor portional to their length while the forces
IC-chip ' IC-chip
are proportional to the cross section area
perpendicular to it, such elements often
have very large volumes of several cubic
•± centimeters, which gives rise to problems
during the burn-out of binders prior to
sintering (Hellebrand etal., 1994). Hence,
binder type and amount of ingredients
have to be selected and adjusted to ensure
uncritical burn-out behavior and excellent
laminating properties. The latter is impor-
tant because piezoelectric components are
stressed both electrically and mechanical-
ly, generating considerable strains and
connecting pins stresses within the element during opera-
ceramic layers internal conductors tion (acceleration forces during fast move-
Figure 7-7. Schematic cross-sectional view of a ceramic ments, fatigue problems with bending
multilayer substrate. devices, etc.).
202 7 Tape Casting

Reviewing the above requirements for unelcctroded

ceramic layers
tapes used in heterogeneous multilayer
every second
composites, at least two central problems laver elect roded
will be evident, namely compatibility of
the different binders for the tape and that every ceramic
laver elect roded
for the printing ink, and the embedding of
the structure elements in the green tape
material.
Typical problems arising from the print-
ing of green tapes with electrode inks are
warping or even dissolving of the sheets. Figure 7-8. Step formation with isostatically lami-
To avoid these difficulties, binders and sol- nated multilayer units containing printed patterns of
internal electrodes.
vents of both systems should be carefully
correlated with each other, especially if the
tapes are very thin. Although it may seem
would be created, which in most cases is
advantageous to select two binder systems
not acceptable (Fig. 7-8).
which are exclusively soluble in different
Alternatively, the structured patterns
solvents (e.g., water-based tape binders
can simply be pressed into the thermoplas-
and organic-based inks) this might not al-
tic ceramic layers by laminating the stack
ways be possible or even desirable, since
at a sufficiently high pressure between flat
some interpenetration of the ink solvents
blocks of metal. For most applications this
into the tapes could enhance anchoring of
is the preferred and often the only feasible
the printed patterns. Thus, printing sys-
method. But there are some serious prob-
tems for green tapes are often purpose-
lems associated with it, especially if very
developed and proprietary.
thin ceramic tapes are concerned. Looking
The second very important general more closely at this process, three stages of
problem in multilayer fabrication arises compaction can be distinguished.
from density gradients generated during During the first stage the pressure is
pressure lamination of the electroded concentrated exclusively at or near the
green tapes. It is evident that any screen electrodes, causing a densification of the
printed elecdrode or conductor pattern has green ceramic tape under the pattern until
a certain thickness (normally 3-5 jim), the whole electrode structure has been
making the even surface of a green ceramic pressed into the layer. The neighboring,
tape structured. By stacking these partly unelectroded areas have not yet received
metallized sheets to form a multilayer any appreciable amount of pressure, thus
compound of up to several hundred layers, giving rise to large gradients of ceramic
the amount of material can vary consider- material concentration.
ably over the whole area of the stack. Lam- If the pressure is increased, the whole
inating such a stack, by pressing it isostat- stack is densified in the thickness direction,
ically or quasi-isostatically (e.g., between accompanied by a break-up of soft ag-
rubber plates in a die) in order to avoid glomerates and by the reorientation of
density fluctuations, would result in com- particles, together with a strong tendency
pounds with uneven surfaces. Steps of up of the above-mentioned density gradients
to several tenths of a millimeter in height to even up. This second stage of pressure
 7.2 Applications and Their Demands 203

lamination lasts until in some areas of the

laminate (predominantly in the region of
the electrodes) maximum particle packing
is achieved. A further increase of pressure
(stage 3) results in conditions which may
be characterized as approximately hydro-
static, and which may fully remove any
small residual density gradients present
(see Fig. 7-9).
From such considerations it is possible
to derive some general remarks on green
tape characteristics suitable for multilayer
composite production:
1. Tapes with a very low gas volume, i.e.,
high binder content (i?b) and high parti-
cle packing (t?p) reach stage 3 very quick-
ly, i.e., even at low pressure and with
practically no densification but with a
very pronounced removal of internal
stresses. Unfortunately, such tapes not
only have unsatisfactory laminating
properties (see Sec. 7.2.2.1) but the high
amount of organic components in them stage 3
also displays unattractive burn-out
properties.
2. Reducing binder volume fraction vb,
while retaining high particle packing Figure 7-9. Embedding of an electrode pattern in a
green ceramic tape by pressure lamination (schemati-
fraction vp, results in a low degree of cally). Stage 1: Electrodes are impressed in the ce-
particle mobility, and through that in ramic layer at the expense of parts of the gas volume.
unsatisfactory lamination and an insuf- The green ceramic is densified mainly in the region
ficient diminution of density gradients. below these electrodes, generating large density gradi-
Admittedly, this texture problem can be ents. Stage 2: The whole stack is compressed uniax-
ially, accompanied by particle rearrangement. Pack-
reduced by the use of higher pressures, ing density increases due to agglomerate destruction
but the resultant lateral particle move- and reorientation of particles. Gas volume fraction
ments might then lead to an undesirable decreases and density gradients are strongly reduced.
lateral shift of the structures. Stage 3: In some parts of the stack maximum packing
density is achieved. Residual density gradients are
3. Only if the gas volume fraction is in- eliminated due to the onset of quasi-isostatic pressure
creased at the expense of the powder conditions.
packing density of the green tape while
the binder content is maintained at a
medium level will compressibility be suf- After pressing, the laminate should dis-
ficient to ensure that all three stages of play a high packing of ceramic powder
compaction take place at meaningful (55-65%) to guarantee optimum sintering
moderate pressures (Biggers et al., 1979; properties. For reasons of better binder
Gardner and Nufer, 1974). burn-out, a certain amount of gas volume
204 7 Tape Casting

(about 5 %) should be left over in the com- In Fig. 7-10 these results are depicted
pound (Kahn et al., 1988), while the rest of schematically in a diagram, which enables
the volume is occupied by organic binder a comparison with products made by using
components and the embedded structures. other technologies such as dry pressing,
The volume of the compressed compound slip casting, isostatic pressing, or extru-
is typically about 10% smaller compared sion, which are described in detail in
with the volume of the unpressed printed Chaps. 5 and 6 of this volume. Naturally,
tapes, but this value can vary from 5-15 % green products manufactured by these
depending on the volume proportion of technologies would normally have wider,
ceramics and electrodes. overlapping ranges of values in this graph,
Remembering the considerations on the but for greater clarity, they are referred to
lamination process in Sec. 7.2.2.1 and al- as typical ranges.
lowing for a certain need of additional gas
volume for the embedding of the electrode 7.2.3 Special Components
structures and the required reduction of
So far, only the most common and nu-
density gradients, the optimum range of
merically largest groups of tape cast ce-
green tape volume fractions can thus be re-
ramic sheet applications have been dis-
calculated:
cussed. Moreover, numerous specialized
40 vol. % < vp < 50 vol. % applications are known from the litera-
20vol.%<t? b <30vol.% ture. These have to be dealt with separately
25vol.%<i; g <35vol.% since tapes used for these have very special

100% vB

50% 50%

highly packed
tapes from sped
powder fractions

100% vn 100% vb
50%
Figure 7-10. Ranges of green ceramic tape volume fractions (vp, vb, vs) suitable for different applications con-
trasted with some well-known related products.
 7.2 Applications and Their Demands 205

functional properties. Although only some • powders with bimodal distribution

examples can be referred to here, they will (Costa et al., 1993)
at least give an impression of the large va- • powders with controlled state of ag-
riety of possible tape applications. glomeration (Bonekamp etal., 1989)
During sintering, ceramic components • powders from dried sol gel systems
always pass through stages where their (Keizer and Burggraaf, 1988)
structure is more or less porous, while al- • tapes containing organic inclusions
ready displaying sufficient mechanical (spheres, fibers, etc.) (Zhao and Harper,
strength to be used for some special task. 1988; Wersing etal., 1986a)
In the early stages of sintering, porosity is • uniaxial sintering of layers joined to
normally open, i.e., the ceramic substance rigid substrates (Garino and Bowen, 1987;
is accessible to liquid or gaseous media Scherer and Garino, 1985).
from the environment. Components with
this kind of structure can be used as filters High-temperature fuel cells are current-
and as membranes for the separation of ly developed by bonding porous anodes
gases or liquids with the demand for a nar- and cathodes to a ceramic electrolyte layer
rowly defined pore size distribution (Keiz- (Fig. 7-11). The electrolyte layer consists
er and Burggraaf, 1988). For other impor- either of a dense sheet of Y-stabilized ZrO 2
tant applications such as sensors and cata- (solid oxide fuel cell, SOFC) (Nguyen
lysts, the total amount of accessible inter- Quang Minh, 1991) or of a porous tape of
nal ceramic surfaces of the porous unit LiAlO 2 , which has to be filled subsequent-
must be as large as possible. ly with molten carbonates of K and Li
If sintering is continued, the porosity (molten carbonate fuel cell, MCFC) (Na-
within the ceramic body may become mensma and Van der Molen, 1989). Poros-
"closed". Substrates with such a porous ity of anodes and cathodes has to be open
structure will have a lower permittivity after sintering, and especially in the
compared with a dense ceramic substrate, MCFC, the pore sizes of the different lay-
allowing the operation of integrated cir- ers must be adjusted relative to each other,
cuits at higher signal speeds (Yamamoto in order to insure secure retention of the
et al., 1989). With hydrostatic piezoelectric liquid carbonates in the LiAlO 2 .
sensor and actuator applications, more Additional layers are needed to separate
compliant elements are required. This can the single cell units and also to provide the
be achieved by embedding arrays of isolat- interior of the cell with sufficient gaseous
ed voids, building up a 3-0 connectivity reactants. For this reason, they are struc-
with the ceramic matrix, i.e., the ceramic tured in the green state, either by corrugat-
phase is three-dimensionally interconnect- ing them by hot forming (Sawicka et al.,
ed while the phase of the voids has zero 1987; McPheeters et al., 1988) or by em-
connectivity (Newnham et al., 1978; Kahn bossing edge-shaped channels on the flexi-
and Chase, 1992). ble thermoplastic sheets (Tai and Lessing,
There are several methods for the form- 1991). To ensure sufficient strength or, in
ing of porous ceramic structures for appli- some cases, effective separation of the dif-
cations determined by the required ferent gases, these layers must be absolute-
amount, size and distribution of the pores. ly free of cracks. The forming and sintering
Appropriate green tape configurations are of such complex green composites is a very
available by the use of: sophisticated process.
206 7 Tape Casting

anode

electrolyte

Figure 7-11. Cross-sectional view of a

solid oxide fuel cell, consisting of a dense
electrolyte layer of ZrO 2 and two porous
cathode layers working as anode and cathode, re-
spectively (courtesy of G. Preu, Siemens
AG).

100 pm

Mismatch of thermal expansion coeffi- in the substrate plane and parallel to the
cients may cause crack formation. This can casting direction.
be avoided by using electrodes consisting For all powders with pronounced aniso-
of a number of layers with slightly differ- tropic properties this effect may be of par-
ent expansion coefficients, thus bridging ticular practical significance. An example
the troublesome difference between the which has attracted some interest is repre-
two end components (SchieBl et al., 1991). sented by the group of the so called 'bis-
Such gradient composites may not only be muth-layer-structured ferroelectics' which
used for SOFC-structures. They are a gen- crystallize in a pronounced platelike struc-
eral, effective way of adapting non-con- ture (Holmes et al., 1979; Lin et al., 1983).
ciliatory components by the use of inter- In spite of their relatively weak piezo-
layers with slightly changing properties, electric effects, materials like Bi2WO6
e.g., composition, microstructure, or den- (Kimura et al., 1982), Bi 4 Ti 3 0 1 2 (Holmes
sity. An indispensable prerequisite for this etal., 1979; Watanabe et al., 1989) and
method is the ability to make green tapes PbBi 2 Nb 2 0 9 (Lin et al., 1983; Hofer, 1990)
with accordingly narrow tolerances for the may prove advantageous because of their
characteristics in question. high Curie temperatures (450°C-950°C).
During the last two decades, tapes with Watanabe etal. (1989, 1991) have
preferred orientation of particles have at- demonstrated that interactions between
tracted a great deal of interest. The phe- powder particles during the flow of the
nomenon was first reported by Di Mar- slurry beneath the casting blade give rise to
cello et al. (1972) who found that morpho- the mutual parallelism of the particles in
logically anisotropic particles of A12O3 are the tape. Thus, for the casting of highly
slightly aligned during tape casting. This oriented green sheets, slurries with a high
small orientational effect increases in the powder content are required, while rheo-
course of subsequent sintering, resulting in logical values have no appreciable influ-
a fibrous texture, with the long axis lying ence. The green density of the tapes should
 7.3 Materials for Casting Ceramic Tapes 207

be as high as possible in order to promote ly temporary, components necessary to

alignment during sintering and densifica- form the green body.
tion. Looking at the general flow chart in
Another very important group of mate- Fig. 7-1 for casting ceramic tapes, the cen-
rials that should be mentioned in this con- tral significance of these process steps be-
text are ceramic superconductors with comes evident. Adequate powder prepara-
high critical temperature Tc. It has been tion techniques on the one hand and ap-
found that the critical current density jc in propriate slurry formation on the other
YBa 2 Cu 3 O 7 _x, at present the most impor- hand are the key factors in ceramic tape
tant superconducting ceramic material, is production.
highest in a plane parallel to the copper- In the first part of this section, emphasis
oxygen layers (Suzuki etal., 1990; Hum- will be laid on the description of available
phreys, 1991). It is therefore important for techniques of preparing ceramic powders
crystallites in wires or layers to have a and their suitability for the formation of
preferred orientation. Furthermore, the specific green structures by tape casting.
strength and life time of these materials The second part will deal with the role and
depend very much on the density of the the tasks of the organic components dur-
sintered components. ing this process and their influence on slur-
As a result of these requirements, green ry properties and tape characteristics. Fi-
tapes should be textured and have a high nally, some examples of practically proven
green density, and since they are often used slurry compositions for the tape produc-
to form curled-up cylindrical devices, the tion will be cited.
tapes must also be sufficiently flexible
(Matsubara etal., 1989; Fisher and
7.3.1 Ceramic Powder Materials
Schober, 1987).
Ceramic powders are nearly always pro-
duced synthetically, since modern ceramic
7.3 Materials for Casting Ceramic products demand strict control of chemical
Tapes composition and micro structure. Today, a
great number of powder preparation tech-
Ceramic products are normally manu- niques are available, differing widely in the
factured from fine, crystalline ceramic characteristics of the produced powders,
powders by forming a so-called 'green' as well as in parameters such as process
body and consolidating it by sintering. The yield and costs of production. The various
properties of the sintered product (me- aspects of these topics will not be treated in
chanical, electrical, magnetic, or thermal) this discourse. For a detailed description
depend not only on the chemical composi- of these matters the reader is referred to
tion of the starting powder particles. They Chap. 3 of this volume as well as to the
are also strongly related to characteristics numerous special publications on this sub-
such as powder morphology and constitu- ject cited, for instance, in relevant review
tion of the particle surfaces. These charac- articles (Johnson and Gallagher, 1978;
teristics, however, are to a great extent de- Abicht etal., 1990; Gallagher, 1991).
termined by the choice of the special pow- In the context of tape casting, it may be
der preparation process and the interac- useful to briefly mention the most fre-
tion of the particles with the various, most- quently applied methods of fabrication
208 7 Tape Casting

along with some powder processing tech- are normally highly aggregated, and have
niques, and to point out some characteris- to be milled down prior to use for thin
tics associated with them. Considering the layer production. Particle size distribution,
broad range of applications described in therefore, is rather broad, and chemical
the previous section, it is not possible to purity and phase homogeneity often are
define a special set of powder characteris- not very high. In some cases, stoichiometry
tics suitable for tape casting in general. is impaired by the partial evaporation of
The choice has to be made according to the certain components at the temperature of
demands of the present application and calcination (PbO, Bi2O3).
with respect to criteria such as: In spite of these restrictions, mixed ox-
ide preparation remains by far the most
chemical purity
frequently used process for manufacturing
particle size and size distribution
complex powders for capacitors, piezo-
particle morphology
electric and electrostrictive components,
degree of aggregation or agglomeration
ferrites, or high-temperature superconduc-
compositional homogeneity
tors, since this method provides large pow-
sintering activity
der quantities at relatively low costs. Non-
ability to be mass-produced
oxide powders such as A1N and Si 3 N 4 ,
costs of production
used for electronic substrates with high
thermal conductivity, can be produced by
a similar process but in a nitrogen environ-
7.3.1.1 Powder Preparation ment instead of oxygen or air (Segal, 1994;
Ceramic powders either consist of single Nagai and Kimura, 1989).
oxide or non-oxide components (A12O3, Another very important process is char-
SiC), a prereacted complex combination acterized by precipitating or coprecipitating
with a well-defined chemical formula hydroxides or oxalates from aqueous solu-
(BaTiO3, PbNd(Ti,Zr)O 3 ), or a mixture tions of the components (Abicht et al.,
of two or more of these substances 1990; Matijevic, 1989 a). The so-called
(Al 2 O 3 + ZrO 2 , dielectric materials with Bayer process (Flock, 1978) for manufac-
added sintering aids). turing alumina from bauxite is a well-
If powder mixtures must be processed known example. It provides powders with
with components having greatly varying a broad variety of particle properties
particle sizes and specific weights, prob- concerning characteristics such as size,
lems can arise concerning homogeneity morphology, specific surface area, or
and segregation during slurry formation. chemical purity (Fig. 7-12). Soluble sodi-
Special precautions have to be made in or- um aluminate is formed by dissolving
der to avoid these effects (see Sec. 7.3.2). bauxite in hot sodium hydroxide and then
A very common method of forming ce- precipitating alumina trihydrate from that
ramic combinations is the so-called mixed solution. This intermediate product has to
oxide route, a solid phase reaction method. be washed and calcined for dehydration.
Here ceramic synthesis is achieved by cal- However, the remaining soda content of
cining premixed and activated oxides, hy- about 0.7% would still be too high for
droxides, or carbonates of the elements in most electronic substrate applications, be-
question at temperatures of approximately cause of its deleterious effect on dielectric
500 °C to 1100 °C. The resulting crystallites losses. A further reduction of soda content
 7.3 Materials for Casting Ceramic Tapes 209

v;** face 1.4 m2/g b) spec, surface 84

Figure 7-12. Electron micrographs of

several Bayer-process A12O3 powders,
showing the variability of particle char-
acteristics. The specific surface of the
powders normally increases with decreas-
ing particle size (a-c). In (d), however,
the larger particles consist of porous ag-
gregates of extremely fine primary parti-
cles, giving rise to a very high specific
surface.
c) spec, surface 13.2 mVg d) spec, surface 15.8 mVg

10 um

( < 0.01 %) can be achieved by calcining at which can be divided into three groups:
high temperatures in the presence of gas- solvent vaporization, solution combus-
eous chlorine (Prasad, 1982). Repeated tion, and precipitation filtration (Johnson
purification and milling may thus lead and Gallagher, 1978).
to a reactive, high purity, a-Al 2 O 3 with The first group comprises processes
small crystallite size (d5O&0.5 |im), about such as spray drying, freeze drying, or
10m 2 /g specific surface area, and 99.9% emulsion drying. The resultant powders
alumina content (Cooper et al., 1987; normally are amorphous hollow spheres
Nagai and Kimura, 1989; Shanefield and which, prior to use for tape casting, have
Mistier, 1974). to be broken up and calcined. Another
The Bayer process belongs to the large method with solvent vaporization is the
group of liquid-phase powder processes. so-called sol-gel technique. It uses colloidal
In general, producing powders from a mix- suspensions of precursor materials, for in-
ture of the components in liquid media stance metal-organic compounds such as
(e.g., solutions or sols) is promising be- alkoxides or alkanes. If partially hydro-
cause of its homogeneity at an atomic lyzed, these compounds will build up a gel
scale. The main problem which arises is to and link together to form macromolecules.
provide a method to remove the liquid Such gels collapse if dried by normal evap-
phase, while preserving the homogeneous oration, forming very hard and porous ag-
distribution of the components. A great gregates (xerogels) suitable for powder
number of techniques has been tried out production only after intense comminu-
for this purpose of solid/liquid separation, tion.
210 7 Tape Casting

Drying the gels hypercritically in an jevic, 1989b; Kulig etal., 1995). Another
autoclave, however, will result in very wet synthesis route providing powders
loosely-packed skeletons (aerogels), which with similar characteristics is called emul-
can be ground down before calcining sion precipitation (Gassner etal., 1991;
(Brinker and Scherer, 1990 a). Because of Ponton, 1993). In this technique, precipita-
its relatively high costs and low material tion of precurser materials takes place in
yields, this process is mostly used for spe- the submicrometer water droplets of a
cial applications at a laboratory scale. water-in-oil emulsion which permits the
It is preferable, however, to avoid full production of powders with selectable
gelation of the whole batch, by using the sizes and size distributions. However, yield
polymeric metal-organic intermediates of material is small compared to hydro-
and decomposing them pyrolytically. With thermal processes.
this solution combustion technique, mono- A special technique, the molten salt pro-
phase (ZrO 2 or A12O3) and multiphase cess, was mentioned above with regard
(spinels) oxide powders with submicrome- to platelet bismuth layer structures (see
ter particle sizes and very high purity are Sec. 7.2.3). In a flux of KC1 and NaCl, for
formed as well as non-oxides such as SiC instance, oxides are molten, reacted, and
and Si 3 N 4 (Riedel, 1991). crystallized, preferably in their idio-
Precipitation techniques with subsequent morphic shape. After cooling, the water-
filtration, such as the described Bayer pro- soluble salts are washed out carefully and
cess, are most widely used in preparing the morphologically anisotropic crystals
high purity ceramics. Precipitates or copre- can be dried and used for textured ceram-
cipitates of hydroxides and oxalates are ics (Arendt and Rosolowski, 1979).
affected by parameters such as tempera- Small quantities of uniform, nearly
ture, concentration, pH values, reacting monodisperse, very high purity powders of
agents or equilibrium factor which in turn submicrometer particle size can be gener-
will influence particle stoichiometry, size, ated using vapor phase reaction processes.
and degree of agglomeration. In most With chemical vapor deposition (CVD),
multicomponent systems, coprecipitates for instance, small quantities of A12O3,
will not be a chemically uniform com- SiC, TiB2 Si 3 N 2 , or A1N with special mor-
pound but a mixture of hydroxides or phological characteristics are produced.
oxalates which have to be reacted subse- On a much larger scale, similar particles
quently. (often single-phased monoxides) can be
An advanced modern method for pro- formed by coalescing reacted gases at high
ducing ultra-fine (<100nm), approxi- temperatures or by laser radiation (Can-
mately monodisperse, highly reactive and non et al., 1982; Haggerty, 1991). In Table
homogeneous single crystals of mono- or 7-3, some frequently used powder materi-
multicomponent systems (such as ZrO 2 , als for tape casting are listed together with
BaTiO 3 , PZT, or ferrites) is the so-called some remarks on their application, their
hydrothermal technique. Oxidation, precip- properties, and the way they are generally
itation, hydrolysis, or crystallization is per- produced (Brown et al., 1991).
formed under 'hydrothermal' conditions,
i.e., high temperature (>200°C) and high
pressure (^100 MPa) in an autoclave
(Segal, 1994; Somiya etal., 1984; Mati-
 7.3 Materials for Casting Ceramic Tapes 211

Table 7-3. Ceramic powder materials for tape production.

Powder material Generally applied Application of tapes Properties, requirements,

process of fabrication figures of merit

a-Al2O3 Bayer process electronic substrates, 97-99.9% A12O3 content;

insulators rel. permittivity ^10;
thermal exp. coefficient
(6-9)xlO" 6 K" 1 ;
thermal conductivity
10- ^ 1
Al2O3/SiO2 (mullite) natural sources; mixed electronic substrates, rel. permittivity^7-8;
oxide; hydrothermal photovoltaic solar thermal exp. coefficient
precipitation cells (4-6)xlO~ 6 K- 1 ;
matchable to silica
(»5xlO-6K~1);
thermal conductivity
4-6Wm-1K~1
A1N carbothermal reduc- electronic substrates rel. permittivity « 8;
tion of A12O3; thermal exp. coefficient
nitridation of Al (4-5)xlO~ 6 K- 1 ;
thermal conductivity
100-200 ^ 1
BeO thermal decomposition electronic substrates rel. permittivity = 6.7;
of Be(OH)2 from thermal exp. coefficient
minerals (6-9)xlO" 6 K~ 1 ;
thermal conductivity

very low loss tangent; toxic

BaTiO3 + dopants mixed oxide; hydro- multilayer capacitors rel. permitivity « 2000-6000
thermal precipitation with tailorable temp, depen-
dence
ZnO from ore or metal varistors nonohmic current/
oxidation voltage response (/ oc Vn)
with n&20 to 50
Pb(Zr,Ti)O 3 +dopants mixed oxide; hydro- piezoelectric compo- high displacement per unit
thermal precipitation nents voltage ( « 1000 pm/V)
YBa 2 Cu 3 O 7 _ x mixed oxide high temperature high critical temperature
superconductors ( « 70-90 K);
high current densities;
grain oriented ceramics
molten salt synthesis high temperature Curie temperatures
Bi 4 Ti 3 O 12 piezoelectric compo- 450-950 °C; enhancement of
Bi2WO6 nents piezoelectric effect by grain
orientation
ZrO 2 (Y-stab.) melting of ZrO 2 sand; oxygen sensors; ionic conductivity
hydrothermal precipi- SOFC electrolyte ^0.09 Q" 1 cm" 1 at 1000°C;
tation thin, dense, flaw-free layers
Sr-LaMnOq mixed oxide SOFC cathode thermal exp. coefficient close
to ZrO 2 ; porous structure;
conductivity 200-400 Q " x cm " J
LanSr 1 _ nMn 1 _ ^Co^O3 _ 5 mixed oxide SOFC cathode porous structure; conductivity
200-1000 Q 1 cm" 1
212 7 Tape Casting

7.3.1.2 Powder Processing Since such agglomerates behave like sta-

ble, irregularly-shaped large particles, they
Most ceramic powders, synthesized from
will not only complicate formation of sta-
the above described production lines, are
ble dispersions, but will also have a partic-
aggregated, i.e., they consist of more or
ularly deleterious effect on the packing
less porous structures of intensely inter-
density and homogeneity of the green tape.
connected primary particles (Fig. 7-13)
Small /ft/raagglomerate pores and large
(Niesz et al., 1972; Schubert and Petzow,
mteragglomerate voids due to bridging ef-
1988; Halloran, 1991). These hard agglom-
fects (Kingery, 1978; Ueyama et al., 1988)
erates can have considerable strength and
will be produced, which cannot be fully
cannot be readily destroyed by normal dis-
eliminated during sintering (Rhodes, 1981;
persion techniques such as ultrasonic vi-
Halloran, 1984; Lange 1984) because of
bration and slurry milling or by pressure
preferential densification of the fine prima-
compaction of the green tapes (Ciftcioglu
ry particles. Disregarding those tape appli-
etal., 1987; Chane-Ching etal., 1989).
cations which require open, porous struc-
tures, existence of large pores within thin
ceramic layers is not tolerable in the great
majority of applications. Therefore, in
general, it is imperative that powders be
reduced to small primary particles prior to
use for tape production.
A great variety of methods is available
for this purpose, all of which have their
special strengths and weaknesses. For fur-
ther discussion of this matter reference
should be made to Chaps. 1-3 of this vol-
ume. Only the most frequently used proce-
10 jim dures will be mentioned briefly in this con-
text (Malghan, 1991).
Ball milling, as either dry or wet milling,
is presumably the most widely used tech-
nique. Powder particles are predominantly
broken between the spheres or cylinders of
the grinding media. The energy results
from collisions between the tumbling and
rolling balls. Since the source of energy is
gravity, allowance must be made for a cer-
tain buoyancy effect of the spherical or
cylindrical grinding media in the concen-
trated slips of sometimes high specific
2.5 pm
gravity. A12O3 or SiO2 grinding media, for
Figure 7-13. Electron micrographs of an aggregated instance, can be suitable for very diluted
ZrO 2 powder particle. The high-resolution figure re-
veals the fact that the aggregate consists of small pri- slips or slips from specifically light materi-
mary particles baked together strongly, predomi- als, containing mainly Al, Be, Mg, or Si.
nantly by a close contact of their surfaces. Such grinding media would be nearly inef-
 7.3 Materials for Casting Ceramic Tapes 213

fective, however, with concentrated slips

made from Pb or Fe combinations, where
zirconia or iron media are considerably
more effective.
Vibratory milling, either wet or dry,
makes use of the local impacts and shear
forces from collisions controlled by inertial
movements of the media balls. It is usually
more efficient and yields a finer powder
than ball milling, if parameters such as fill-
ing factor of the mill, ratio of powder to
1 10 100
media, viscosity of the slip, or size and
particle diameter [\im]
specific gravity of the media balls are opti-
mized (Choi, 1971; Sliva et al., 1989). Figure 7-14. Cumulative particle size distributions at
different stages of attritor milling. Starting from a
Attrition milling is a still more effective coarse, agglomerated or aggregated powder with a
wet grinding process. A large number of broad size distribution, milling results in a successive
very small grinding balls are greatly agitat- decomposition of these large compounds to small pri-
ed by specifically formed arms rotating in mary particles. Unmilled powders and also intermedi-
ate stages with incomplete milling are often indicated
a cylindrical chamber at a high angular by the occurrence of a shoulder in the curves due to
speed (Schmidt, 1989). The slips can be the presence of a bimodal distribution of primary
pumped through the mill several times, in particles and agglomerates. Milling times: (a) 0 min,
order to yield a fine particle size and a (b) 10 min, (c) 20 min, (d) 40 min.
narrow size distribution within short time
spans. As an example, Fig. 7-14 shows cu-
mulative particle size distributions of an size distribution than does attrition mill-
attrition-milled PZT ceramic at different ing, where particles near the axis move at
stages of the milling procedure. The disap- a lower velocity than those in a more
pearance of the bimodal distribution indi- peripheral position (Leschonski, 1989).
cates destruction of the aggregates to pri- Wear of the grinding media as well as of
mary particles of 0.6 ^m median size with- the walls of the chamber is unavoidable
in less than one hour of milling time. with the comminution techniques de-
Another effective variation of this mill- scribed so far. In order to minimize in-
ing system, the so-called "ring gap mill", crease of impurities in the powder batch,
makes use of shear and compressive forces the milling procedure has to be optimized
exerted in a narrow slit of coaxially ar- by strictly controlling parameters such as
ranged cylinders. One cylinder rotates at a weight fractions of powder, media, and
high speed while the other one is fixed. The fluid, grinding time, initial particle size and
forces are exerted by tiny grinding balls of selection of adequate media material akin
less than 1 mm diameter. The outlet con- to powder composition.
sists of an adjustable small ring gap, which Alternatively, and with appreciably
has to retain the balls of the milling media. greater success, contamination can be min-
In this system, all particles move at nearly imized by using the very effective fluid en-
equal velocity and are therefore subject to ergy technique known as "jet milling". A
the same comminutive forces. This tech- beam of gas or liquid containing the pow-
nique produces powders with a narrower der is accelerated to near sonic velocity
214 7 Tape Casting

and directed against a fixed wall of non- be present immediately after casting and
abrasive material, or two beams are direct- drying. The characteristics of suitable slur-
ed against each other. The energy for com- ry compositions will be featured in the fol-
minution results from particle impacts, lowing Sec. 7.3.2. In the case of multilayer
and in the latter case evidently no contam- components, these properties are only re-
ination takes place. However, the separa- quired after lamination, i.e., after a process
tion of large quantities of gases from the of compression.
powders is not an easy task, especially with In the following, some general aspects
particles in the size range of micrometers concerning packing of particles shall be
and even less (Richerson, 1982). dealt with in order to define factors gov-
To conclude this section on ceramic erning these properties and feasibility lim-
powder processing related to the special its in real tape production processes.
conditions of tape casting, the problem of Packing of powder particles has been in-
possible undesirable interaction of pow- vestigated both theoretically (McGeary,
ders with processing materials or the en- 1961), and by experiment (Patankar and
vironment has to be considered. Some Mandal, 1980; Barringer and Bowen,
very interesting ceramic powders such as 1982). From a theoretical point of view,
A1N, BeO, or the superconductor material ideal monosized spheres with no (or negli-
YBa 2 Cu 3 O 7 _ :c are not water-resistant. Of gible) particle interaction due to van der
course, water-soluble binder systems are Waals forces, should achieve a maximum
prohibited with them, and milling pro- packing of 74 vol. % in an octahedral or
cesses have to be implemented in water- tetrahedral arrangement with 12 closest
free organic media. Moreover, it should be neighbors (coordination number 12). Other
pointed out that very fine-sized powders of packing arrangements of such spheres
non-oxide materials have often a pro- with coordination numbers 10, 8, or 6
nounced tendency for spontaneous exother- would have theoretical packing densities
mic oxidation, which complicates the use of 69.8 vol.%, 60.5 vol.% and 52.4 vol.%,
of dry-milling processes. respectively (Fig. 7-15).
Experiments with tapping, tamping or
7.3.1.3 Packing of Powder Particles vibrating monosized spheres (50 |im to
The optimum configuration of a green 2 mm) resulted in dry powder packing den-
ceramic body prior to sintering is generally sities of about 58 to 63 vol.%, indicating
characterized by a high powder packing that normally a combination of these
density of submicrometer, homogeneously packing arrangements is present simulta-
arranged particles of nearly spherical neously, with a mean coordination num-
shape and with narrow size distribution. In ber of just under 8 (Ueyama and Kaneko,
green structures of this type, the interstices 1987). If such dry state experiments are
between the powder particles are likewise made with fine-sized powders in the range
small and homogeneously distributed, and of micrometers, submicrometers, or nano-
the number of particle contacts in the com- meters, achievable packing densities will
pound after binder burn-out is high, giving be appreciably lower (^45 vol.% with
rise to a pronounced sintering activity and jim-sized particles; ^ 3 0 % with nanosized
a low amount of shrinkage (Rhines, 1978). powders) since particle arrangement is
If individual tapes (monolayers) are then predominantly determined by the for-
concerned, the described properties should mation of agglomerates due to attractive
 7.3 Materials for Casting Ceramic Tapes 215

mmmm
coordination no. 6 coordination no. 8 coordination no. 10
packing dens. 52.4 vol% packing dens. 60.5 vol% packing dens. 69.8 vol%

Figure 7-15. Packing arrangements of

spheres with coordination numbers from
coordination no. 12 coordination no. 12 6 to 12 (cidapted from Ueyama and
packing dens. 74.1 vol% packing dens. 74.1 vol% Kaneko, 1987).

van der Waals forces (Paulus, 1984; Cutler, dered regions ('domains') which after sin-
1978; Kitoako and Seki, 1988). • tering might result in large pores, flaws, or
These considerations are transferable to cracks (Liniger and Raj, 1987, 1988).
powder packing in fluid media, i.e., for Most powder production routes, how-
green tape consolidation from slurries. ever, produce particles with a more or less
Without any precautions (e.g., the use of wide size distribution. Theoretically, pow-
suitable solvents, binders, or special addi- ders with a bimodal or multimodal size
tives) to prevent clustering, packing densi- distribution (with the smaller particle frac-
ties would not exceed values of v p ^ 4 0 - tion filling the respective interstices be-
45 vol.% for powder particles with about tween the larger ones) would allow very
1 jim diameter. high values of packing density of 90 vol.%
So far, only monosized powders have and more. But the necessary large ratios of
been considered. Even if some of the pow- particle diameters (1:7:38:. . .) (see Fig.
der preparation techniques produce nearly 7-16) would bring about a pronounced
monosized particles, and even if similar segregation within the slurry, which would
powders could be fractionated from mate- be counterproductive (McGeary, 1961).
rials with wide size distribution (Richer- A far better approach is to use disor-
son, 1982; Pober et al., 1984; Mizuta et al., dered, randomly packed, but uniform
1984), their use for a broad range of appli- green structures which exhibit smaller de-
cations seems to be limited. Disregarding fect sizes and may be sintered without
the economic restraints, truly monosized flaws (Brook, 1989). Such structures can
particle arrangements have separated or- be manufactured by using powders with
216 7 Tape Casting

size distribution. After compression, the

larger pore size fraction should disappear
due to agglomerate distruction. For this
effect to happen, it is, however, important
that these agglomerates are soft enough,
not excessively large, and that they are ho-
mogeneously distributed in the whole tape
(Niesz et al., 1972; Ciftcioglu et al., 1987).
Figure 7-16. High particle packing, achievable theo-
retically by the use of multidisperse particle fractions 7.3.2 Slurries for Tape Casting
of spheres with special size ratios (1:7:38...).
Following the flow chart of Fig. 7-1, the
powders have to be suspended homoge-
narrow particle size distribution (Mizuta neously in a suitable liquid system and a
et al., 1984), the largest grains preferably highly-concentrated castable slurry has to
being smaller than twice the mean particle be formed. Its merely temporary, but abso-
size (Hillert criterion), in order to avoid lutely decisive task is to provide a stable
abnormal grain growth (Brook, 1976; Gat- dispersion of adequate viscosity, while pre-
tuso and Bowen, 1985). More recent re- serving the tailored morphological charac-
sults concerning sintering of alumina per- teristics of the powders, and transferring
mit even the use of larger size ratios, but them to the consolidated state of the dried
acknowledging at the same time the detri- tape. This tape should exhibit sufficiently
mental effect of exaggerated grain growth high strength and flexibility for further
due to excessively broad size distributions procedural handling as well as a homoge-
(Yeh and Sacks, 1988). neous, flawless microstructure (Bohnlein-
Adequate powders with suitable, nar- MauB etal., 1992), since defects prior to
row size distribution are provided by most sintering are generally known to be en-
of the described preparation routes, espe- hanced during firing (Kingery, 1978).
cially if they are combined with subsequent From the beginning of the art of tape
high energy milling and/or sizing by parti- casting, this has been done by using organ-
cle separation with cyclones or dispersion ic binder polymers dissolved in suitable
by sedimentation. solvents like in the paint industry (Mistier,
It was demonstrated in Sec. 7-2 of this 1991). While the powder after firing is re-
chapter that for the very important branch sponsible for the electrical and mechanical
of multilayer technology, an extremely properties of the ceramic body, the pur-
high powder packing density in the green pose of the organic components is merely
tapes would not be desirable (Utsumi, transient. They have to be evaporable,
1991). Before lamination, tapes should therefore, with no or only negligible resi-
rather consists of fine-sized primary parti- dues and at moderate temperatures far be-
cles, bound in soft agglomerates, which low the sintering temperature.
can easily be destroyed by the compressive This basic approach has prevailed until
forces at work. now, although some decades of intense re-
On account of bridging effects, agglom- search on this matter have brought about
erates settle in a less densely packed ar- profound insight into the physical and
rangement with a marked bimodal pore chemical mechanisms which govern the
 7.3 Materials for Casting Ceramic Tapes 217

Table 7-4. Special functions and characteristic parameters of slurry components.

Component Functions Characteristic parameters

Powders provide desired ceramic properties composition - morphology - size disbribu-

tion - state of agglomeration
Solvents solvate polymers, plasticizers and additives - efficiency - rate and heat of evaporation -
disperse powder particles - determine slurry polarity - environmental acceptability -
viscosity costs
Binders interconnect powder particles - provide solubility - film formation - tensile strength
green tape strength - guarantee laminate of film - compatibility with other compo-
formation - control rheological behavior nents - thermoplasticity - burn-out behavior
Plasticizers flexibilize polymer films - dissolve organic effectivity - volatility - miscibility with
compounds solvents - solvating properties
Functional disperse powder particles - provide wetting adsorption on particle surface - molecular
additives of powders and substrates structure (ionic, non-ionic, polymer chains,
special agents: flocculate, defoam, homoge- charge) - solubility - surface activity
nize slurries - control viscosity and drying
properties

specific functions of the different slurry sion in liquid media. Additionally, some
components and which affect the proper- slurry formulations will be cited quantita-
ties of the slurry as a whole. According to tively.
their specific function in the suspension,
these organic components can be classified
7.3.2.1 Polymer Binders
in mainly four groups of materials, listed
in Table 7-4 together with some of the When formulating a special slurry com-
properties generally associated with them. position, one will generally begin with the
Theoretically, in each of these groups a selection of a suitable polymer binder sys-
very large number of organic materials can tem because of its central role in determin-
provide the demanded properties. But in ing the green tape properties. Its most im-
practice, only a few of them have proven portant task is to provide a certain amount
the test of time and are now in common of strength and toughness to the thin
use on an industrial scale. sheets by surrounding the powder parti-
Concerning the functional additives, cles, anchoring itself to their surfaces, and
however, there is a permanently growing creating a strong 3-dimensionally inter-
number of specifically effective substances connected skeleton of resin (Mistier, 1991;
(dispersants, wetting, agents, defoamers, Bohnlein-MauB etal., 1992; Moreno,
etc.), which are used to optimize the prop- 1992a).
erties of the slurries and, hence, of the For this reason, only long-chained high
green tapes. molecular polymers (MW 30000-80000)
In this section, the special tasks of the can be used. After drying, they are able to
organic components will be outlined, with form films of adequate strength. In princi-
emphasis on their interaction with the ple, such films can either be made from
powder particles and the mechanisms molecular solutions of the binder in a true
which govern the effect of powder disper- solvent or from macromolecular disper-
218 7 Tape Casting

sions of the polymer in non-solvating

liquids (Nahass etal., 1990; Gurak et al., drying of the
polymer dispersion
1987).
Molecular solutions are created by sol- GO

mm,
vating polymer molecules. Solvent mole-
cules attach themselves to 'active centers'
of the polymer chains, points of inter- forming of
capillaries
molecular Van der Waals bonds, disaggre-
gating them and forming a temperature-
and concentration-dependent solvation/
desolvation equilibrium. During drying film formation
(solvent evaporation) this equilibrium is by particle
shifted towards polymer aggregation coalescence

which is accompanied by an increase of

viscosity. With further loss of solvent, this Figure 7-17. Polymer film formation from disper-
causes full reaggregation of polymer bonds sions by particle coalescence (after Talen, 1962).
(Talen, 1962).
In contrast to normal solutions of small
molecules, solvation of large polymer flow under the conditions of drying. Their
molecules provides the whole range of vis- 'glass transition temperature' should be
cosity values between those of the con- low enough (near room temperature),
densed plastic and the pure solvents. This which in most cases can only be achieved
is an important fact as regards the possibil- by using plasticizers.
ity of adjusting an adequately broad range Generally, flexible long-chained poly-
of slurry viscosity by gradually controlling mers are used instead of polymers with
the content of solvents. three-dimensionally interlinked chains
Macromolecular dispersions, referred to (Mistier etal., 1978), which have higher
as latexes (Gurak et al., 1987), are formed glass transition temperatures and form
by small submicrometer polymer particles rather rigid films. Since tapes to be used
suspended in a non-solvent, mostly water, for the production of laminates have to
by the use of stabilizing agents. Evapora- provide high particle mobility at moder-
tion of water results in small water residues ately high temperatures, binders should
in the interstices of the polymer spheres, exhibit thermoplastic behavior, i.e., they
which latter are provided with the stabiliz- should become more flexible when the
ing hydrophilic surface molecules. This temperature is increased. Moreover, the
causes negatively-curved capillary water problems concerning subsequent binder
surfaces there with large compressive cap- burn-out have to be considered. The
illary forces. amount of polymers should be kept to a
Coalescence is achieved by plastic defor- necessary minimum, and also binders
mation of the polymer spheres into poly- should be preferred which evaporate in in-
hedra under these forces (Fig. 7-17) and a ert or reducing atmospheres. This can be
water-insoluble polymer film results (Hen- accomplished by the unzipping of the
son etal., 1953; Brown, 1956). Thus, film molecular chains rather than by decom-
forming is enhanced if the polymer parti- posing them to carbon residues (Fiori and
cles exhibit a certain amount of plastic DePortu, 1986; Roosen 1988).
 7.3 Materials for Casting Ceramic Tapes 219

Since the early days of tape casting, a This is the basic cellulose unit for the
great number of binder systems have been formation of a great number of ether
tried out, even including natural polymers derivatives, with various radicals such as
such as gums, lignins, alginates or saccha- methyl, ethyl, hydroxyethyl, or carboxy-
rides (Moreno, 1992 a). In spite of their methyl groups replacing some of the OH
economic attractiveness they remain in use groups. Some of them, such as ethyl cellu-
only as binders for tapes of low technical lose, are soluble only in non-polar liquids.
standard because of their undefined chem- But most of them are water-soluble with a
ical composition and poor reproducibility pronounced tendency to foam (Bast,
of properties. Tapes for modern advanced 1990), and they need long drying times re-
ceramics, however, need well-defined and sulting in relatively brittle tapes.
often tailorable binder systems with con-
trolled chemical functions. This can only Poly (vinyl acetatejs
be expected from synthetic combinations. (2)
Basically, polymer binders consist of basic group of H H
small subunits (monomers) which deter- poly(vinyl acetate) I
mine the general character of the macro- PVAc, n x C 4 H 5 O 2 C— C-
I I
molecules. Introducing special short side O H
groups at suitable sites of the chain links I
establishes the great variety of polymer O=C-CH 3
properties concerning solubility, polarity,
glass transition temperature, or anchoring PVAc binders are either used directly,
to powder particles. dissolved in non-polar liquids, or as copoly-
Only a few of these systems span the mers in conjunction with PVC (Bohnlein-
major portion of all tape cast activities, MauB etal., 1992; Rabin, 1990) and are
since they proved to be a good compro- especially suitable for forming thin tapes.
mise with respect to most of the various Difficulties can arise from thermal decom-
demanded characteristics. They will be position due to chlorinated residuals.
briefly cited here and their general chemi-
cal structure as well as some of their rele- Poly (vinyl alcohol) s
vant properties will be described (Onoda,
1978). H H
1 1
H H
Cellulose ethers 1 1 1 1 (3)
O H
1 Q1 1
OH H X
O=C-CH 3

basic groups of poly(vinyl alcohol) (PVA)

A monomer vinyl alcohol does not exist.

PVA, therefore, is synthesized by saponifi-
cation of poly (vinyl acetate) (PVAc). De-
pending upon the degree of saponification,
a variety of PVA types with various ratios
basic group of cellulose, n x C 6 H 1 0 O 5 of basic OH groups to acid O = C-CH 3
220 7 Tape Casting

groups are available. Most poly(vinyl alcohol)s are water soluble and exhibit a pro-
nounced tendency to foam.

Poly (vinyl butyral)

basic groups of H H H H H H H H
poly(vinyl butyral) I I I I I I I I
(PVB) -c-c-c-c- -C- -C- -C-
I I I I I
H H H H (4)
O O O OH

C 3 H 7 -C-H CH3-00

Polybutyral homopolymer does not ex-

ist. PVB is produced by saponification of From polymethacrylic acid, a very large
PVAc and subsequent addition of butyral- number of esters can be formed, by intro-
dehyde to the generated PVA (Bohnlein- ducing one or more small-chained alcohol
MauB etal., 1992). Thus, in commercial groups (methyl, propyl, butyl groups or
products, all three components are nor- combinations of them). They all contain a
mally present with the approximate pro- more or less large number of methacrylic
portions of x& 0.70-0.90; y& 0-0.05; acid groups which are responsible for
z^0.10-0.30 (Butvar B76, B98, etc., Mon- strong anchoring to an oxide powder sur-
santo) (Roosen et al., 1990). Because of its face (Boch and Chartier, 1988). Methyl-
only moderate thermoplasticity but excel- esters (PMMA) and ethylesters (PEMA)
lent dimensional stability and high tensile are by far the most frequently used combi-
strength, PVB is extensively used with nations.
tapes for A12O3 substrates. Anchoring to They have excellent thermoplastic prop-
the alumina powder particles occurs via erties and degrade in inert atmospheres by
hydrogen bonding of the OH groups on depolimerisation, with low residual ash
the powder surfaces and those within the content. Especially thin tapes for multi-
PVB molecule. Alternatively, acid-base re- layer components are produced on a very
actions are a possible mechanism (Howard large scale. Acryloid B7 (Rohm and Haas,
et al., 1990). PVB is soluble in non-polar or U.S.A.) is a very common binder system
weakly polar organic solvents. During which is composed of PEMA (62.3%),
burn-out oxygen has to be present, since PMMA (37.1 %) and polymethacrylic acid
otherwise carbonization takes place. (PMAA) (0.6%). Its dispersing properties

Polymethacrylates

basic groups of CH3 H CH3 H CH3 H

a PMMA/PEMA I
copolymer -C—\-
I 1 (5)
H H H
C
o7 V O-CH 3 O-Co <TOH
 7.3 Materials for Casting Ceramic Tapes 221

on oxide powders are only moderate com- Table 7-5. Selection of binder/plasticizer systems for
pared with other acrylic compositions tape casting.
(PM 685, Rohm, Germany). This is ad- Binder Plasticizers
vantageous for laminate production (see
Sec. 7.2.2). Ethyl cellulose diethyl oxalate
PVA glycerine
7.3.2.2 Plasticizers triethylene glycol
Owing to their interlinked chains, most PVAc + PVC buthyl benzyl phthalate
of the described polymer binders have dibutyl phthalate (DBP)
polyethylene glycol) (PEG)
glass transition temperatures Tg well above
room temperature. Thin tapes produced PVB triethylene glycol hexoate
dibutyl phthalate
with these binders would not be flexible
dioctyl phthalate (DOP)
enough for secure handling even if their poly(ethylene glycol)
tensile strength were sufficiently great.
PMMA, PEMA polyethylene glycol)
Plasticizers are able to shift Tg to a desir- butyl benzyl phthalate
able temperature value, near or below dibutyl phthalate
room temperature, but unfortunately at dioctyl phthalate
the expense of the strength of the tapes Acrylic co-polymer butyl benzyl phthalate
(Bohnlein-MauB etal., 1992; Moreno, latex dibutyl phthalate
1992a; Roosen 1988; McHale, 1991; Kit- poly(ethylene glycol)
tel, 1976a). glycerine

The mechanism of this polymer soften-

ing can be described as a disruption of in-
termolecular bonds when the plasticizer In Table 7-5 suitable, most broadly used
molecules attach themselves to the active binder/plasticizer systems are presented.
centers responsible for chain linkage
(Bohnlein-MauB etal., 1992). This is a 7.3.2.3 Solvents
process comparable to polymer solvation.
Even though the role of the liquid media
Indeed, plasticizers can be defined as hard-
is certainly not restricted to solvating, the
ly evaporable "heavy" solvents (Boch and
name 'solvent' for the temporary liquid ve-
Chartier, 1988). Preferably, their molecu-
hicles in tape casting has become common.
lar weight should be in the range of 300 to
The real significance of the liquid media
400, and their boiling temperature should
becomes apparent when considering the
at least be higher than 200 °C.
main tasks they must perform simulta-
The amount of plasticizers necessary
neously:
to form suitable 'leather hard' tapes
(Williams, 1976) often exceeds the total - dissolving binders, plasticizers, and ad-
weight of the binder. However, with re- ditives
spect to their effect on green tape strength, - dispersing ceramic powder particles
their content should be kept as low as pos- - providing suitable viscosity for the
sible, i.e., plasticizers should be highly effi- slurry
cient. Additionally, they should evaporate - evaporating substantially at moderate
without leaving residues, and should be temperatures
chemically and physically stable, inexpen- - guaranteeing flawless consolidation of
sive and non-toxic. tapes
222 7 Tape Casting

Additionally, they should be pose undue a criterion for a liquid to be a suitable

hazards for health and environment and solvent for a certain binder may be de-
should be inexpensive since they are used fined: the chemical structures of the sol-
in great quantities and in most cases can- vent and the binder should resemble each
not easily be recycled. other in the character of their functional
In Sec. 7.3.2.1 the mechanisms govern- groups (i.e., mainly in their polarity) and/
ing polymer solvation were indicated in or should consist of corresponding groups
general terms. In practice, technicians are for hydrogen bonding.
frequently confronted with the problem of This remains approximately valid for
selecting suitable solvents for a binder sys- the solution of plasticizers and additives
tem with a certain chemical composition. and even for the miscibility of solvents. In
In order to be able to give some general Table 7-6 a selection of frequently used
clues with regard to this decision, it is nec- solvents for tape casting is presented.
essary to take a closer look at some fea- Among other relevant characteristics, such
tures which govern this process. as boiling point, heat of vaporization,
The nature of intermolecular interaction evaporation rate, and viscosity, empirical
forces between solvent molecules and polarity values of the solvents are included
solids can be classified according to their to allow a rough estimate of solubility.
atomistic origin: Such a value, £ T (30), was introduced in
the 1960s by Dimroth and Reichardt [cited
- ionic forces and defined in Reichardt (1979) and in
- dipole/dipole forces Langhals (1982)]. It represents all contri-
- hydrogen bonding forces butions to the solvent polarity by measur-
- van der Waals forces ing the change of molar electron excitation
If polymers are concerned, ionic forces energy of a standard substance (in kJ/mol)
are of minor interest. However, dipole/ dissolved in the solvent in question com-
dipole interations occur if either both pared to its solution in the totally nonpolar
solvent, and polymer have permanent hexane.
polar groups, such as - C = O, - C - O - , It is very common in practice to use mix-
- C - N = , or - C - N O 2 , or if one of them tures of solvents. This is advantageous not
has a permanent polar group and is able to only with respect to the greater adaptabili-
induce a polar group in the molecule of the ty to the various organic ingredients, but
other. Hydrogen bonding is related to the also because polymers are known to be
tendency of hydrogen atoms in (-OH) or more soluble in an optimized mixture of
(-NH 2 ) groups (water, alcohols, glycols, solvents than in any of the individual
amines) to be attracted to acceptor groups liquids. Mixtures of so-called 'kinetic sol-
such as - C - O - , - C = O, - N H 2 (esters, vents' (small molecules such as ethanol)
ketones). Van der Waals forces will always and 'thermodynamic solvents' (esters, ke-
be present. They are generated if the dis- tones such as ethylmethylketone) are most
tance between atoms of the molecules be- effective and thus can be used to minimize
comes small enough for dipoles to be in- the necessary solvent quantities (Moreno,
duced in both of them by electron/nucleus 1992 b; Morris and Cannon, 1986).
displacement (Kittel, 1976 b). Often azeotropic mixtures of solvents
As a general, only qualitatively valid are proposed, especially with organic-
rule which has some important exceptions, based binder systems (Boch and Chartier,
 7.3 Materials for Casting Ceramic Tapes 223

Table 7-6. Characteristic parameters of some selected solvents.

Solvent Boiling Evaporation Heat of Viscosity Polarity Relative Surface

point rate vaporization at25°C £T(30) permit- tension
(°Q (w-butyl (J/g) (mPa s) (kJ/mol) tivity (mN/m)
acet. = l)

Water 100 0.16 2260 1 264 80 73

Methanol 65 3.70 1100 0.6 232 33 23
Ethanol 78 2.65 860 1.2 217 24 23
Butanol 118 0.44 578 2.9 210 18 22
Benzyl alcohol 205 <0.01 415 5.8 213 13 35
Isopropyl alcohol 82 2.08 578 2.4 204 18 22
Ethylene glycol 198 <0.01 800 20.0 236 37 48
Ethyl acetate 77 4.95 360 0.4 160 6 23
Butyl acetate 127 1.00 310 0.5 155 5 26
Methyl ethyl ketone 80 4.59 444 0.4 173 18 25
Acetone 56 7.70 524 0.3 177 21 25
Trichloroethylene 86 4.90 243 0.4 150 3 25
Toluene 110 1.96 352 0.6 142 2 29
/?-Xylene/oxylene 140 0.55 327 0.7 141 2 28
Cyclohexanone 155 0.24 427 0.8 171 18 35

1988). They provide good solubility char- or air (Ws/y) with the energy of the same
acteristics and allow solvent evaporation particle surrounded by the liquid (Ws/L)
at a constant composition. In Table 7-7, (Parfitt, 1986 a). Since the surface free en-
therefore, some binary and ternary azeo- ergy of unit area (w) is equivalent to the
tropic systems are listed. surface tension per unit length y, the unit
On the other hand, there are also good area energy difference can be calculated
arguments for the use of non-azeotropic
Aw = ys/L - y s/v (7-1)
solvent mixtures. Their successive volatil-
ization according to their different evapo- Combining this with the well-known
ration rates and the diversity of their boil- Young's equation for the contact angle 5 of
ing points can yield a higher variability of a drop of liquid in contact with an even,
the drying conditions. solid surface (see Fig. 7-18)
The second very important task of sol-
yS/v = ys/L + 7L/vCos<5 (7-2)
vents in tape casting slurries is to provide
a liquid medium for the dispersion of the an expression for the total work of disper-
ceramic powders. With respect to this task sion for unit powder surface can be derived
the surface properties of the materials are
Aw = yh/y cos 5 (7-3)
significant, i.e., physical and chemical phe-
nomena at the solid/liquid interface have If Aw is positive, i.e., for 5>90°, energy
to be considered. has to be added to the system (by stirring
Phenomenologically, immersion of solid or milling) for the powder to be dispersed,
particles in liquids may be described as the and the liquid is said not to wet the pow-
exchange of surface free energy of a pow- der. But if 0°<<5<90°the process of dis-
der particle surrounded by its own vapor persion produces energy, and the solid is
224 7 Tape Casting

Table 7-7. Azeotropic systems.

(a) Binary systems

Component 1 wt% Component 2 wt% Boiling point

Methyl ethyl ketone 88.6 water 11.4 73.6

Methyl ethyl ketone 40 ethyl alcohol 60 74.8
Methyl ethyl ketone 30 isopropyl alcohol 70 77.3
Methyl ethyl ketone 86 methyl alcohol 14 55.9
Ethyl alcohol 95.6 water 4.4 78.2
Ethyl alcohol 27 trichloroethylene 73 70.9
Ethyl alcohol 68 toluene 32 76.7
Isopropyl alcohol 87.7 water 12.3 80.4
Isopropyl alcohol 30 trichloroethylene 70 75.5
Isopropyl alcohol 21 ethyl acetate 79 75.3
Isopropyl alcohol 52 butyl acetate 48 80.1
Isopropyl alcohol 48 butyl alcohol 52 82.3
w-Butyl acetate 71.3 water 28.7 90.2
n-Butyl acetate 53 w-butyl acetate 47 113.5
m-Xylene 20 n-butyl alcohol 80 116.0
p-Xylene 17 isobutyl alcohol 83 107.5
Acetone 88 methyl alcohol 12 55.9

(b) Ternary systems

Component 1 wt% Component 2 wt% Component 3 wt% Boiling point

Methyl ethyl ketone 81 water 8 ethyl alcohol 11 73.5

Trichloroethylene 69 ethyl alcohol 26 water 5 67.3
w-Butyl acetate 35.3 w-butyl alcohol 27.4 water 37.3 89.4
Ethyl acetate 83.2 ethyl alcohol 9.0 water 7.8 70.3
Isopropyl acetate 63.7 isopropyl alcohol 26.2 water 10.1 76.2

5=0

X//7ZZZA solid plate

Figure 7-18. Different wet-

ting conditions of a liquid
on a plane solid surface.
liquid
Wetting occurs if 0° < 5 < 90°.
5 = wetting angle If S = 0, the liquid spreads
over the solid. If 5>90°,
the liquid is called non-
wetting.
 7.3 Materials for Casting Ceramic Tapes 225

wet by the solvent. It can be shown, how-

ever, that only in the case of S = 0° does
spontaneous spreading wetting occur, i.e.,
the powder submerges instantly without
any intermediate energy consumption. In
all other cases where 0<5<90°, the pow-
der floats on the surface of the liquid and
does not submerge unless some small
amount of energy is provided (by gravity
or stirring), even if the whole process pro-
duces a certain amount of energy surplus.
Dry ceramic powders are always ag-
glomerated to a certain extent, forming
weakly bound structures with channels
and caverns of variable depth and width. b)
Modeling these caverns as capillary tube of liquid medium
radius r, and with one side closed, such as
is shown in Fig. 7-19, the pressure required
to force the liquid into these tubes can be
expressed as

(7-4)

Negative values for this pressure (i.e.,

0° < d < 90°) indicate spontaneous penetra-
tion. In this case the air in the caverns is
compressed and the absolute value of /?cap
may favorably become high enough for
the destruction of the weak agglomerate
bonds, if d is nearly 0° while yL/v is large. Figure 7-19. Modeling agglomerate pores by capil-
laries, (a) Schematic view of a powder agglomerate of
It is possible, indeed, to shift the wetting primary particles which contains caverns closed on
angle to a value close to zero by the use of one side, (b) The caverns can be approximately mod-
wetting agents. However, this is normally eled as capillaries with different diameters. Brought in
accompanied by a decrease in the surface contact with a wetting liquid, the air within the pores
tension yL/v of the liquid and, hence, must becomes compressed to some extent (pcap), which can
be helpful for the destruction of weak agglomerates.
be optimized with respect to both values
(McHale, 1991). It should be pointed out
that an agglomerate destruction by this
kind of effect does not strictly demand the additional mechanical energy from milling
existence of one-side closed pores. Even in or agitation.
more "open" agglomerate structures air The influence of solvents on the creation
bubbles may be entrapped and compressed of suitable slurry viscosities will be de-
by capillary forces, resulting in a similar scribed later in Sec. 7.3.2.5. Likewise, it
effect of decomposition. In principle, how- will be more appropriate to deal with sol-
ever, de-agglomeration necessitates some vent evaporation problems when present-
226 7 Tape Casting

ing the facts governing the process of tape Vt=Va (van der Waals) + Vx (electrostatic)
drying (see Sec. 7.4.2). + Vr (steric)
Attractive van der Waals forces are al-
7.3.2.4 Functional Additives ways present, since they originate in the
Ceramic powder dispersions in liquids interactions of atoms with either perma-
are normally very unstable and inhomoge- nent or induced electron/nucleus dipoles
neous, because the small powder particles on the surface of the particles.
have a pronounced tendency to arrange Repulsive electrostatic forces arise from
themselves in large agglomerates. This the interaction of particles which carry
gives rise to undesirable separation effects electric charges of the same sign. They pre-
caused by fast sedimentation of these clus- vail in polar solvents of high dielectric con-
ters. This effect is the more pronounced stant (especially water) and are of minor
the smaller the primary particles are and significance in most organic-based slurry
the more polar the suspending liquid is, systems.
i.e., in water-based systems. A certain Repulsive steric forces, on the contrary,
amount of agglomeration can be advan- are predominant in non-polar organic sol-
tageous such as for tapes for multilayer vents, where electrostatic forces are more
applications. But even then uncontrolled likely to be small. They are due to the in-
agglomeration in the slurry would have teraction of long chained macromolecules
undesirable consequences for the further adsorbed at the surfaces of the particles.
processing and the properties of the green In real tape casting suspensions, steric
tapes. and electrostatic repulsive forces are gener-
Most tape casting slurries therefore con- ally present simultaneously. For reasons of
tain special dispersing agents, temporary lucidity, however, these effects can be dis-
additives which control the degree of parti- cussed separately, if one keeps in mind that
cle agglomeration as well as the strength of their pure manifestation has to be consid-
agglomerates. Hence, it might be appropri- ered only as border-line cases of a more
ate and helpful at this point to deal with general situation.
some fundamental aspects governing the
Electrostatic Stabilization
dispersing process although only a very
brief and consequently simplified sum- On the basis of these considerations, the
mary can be offered. Reference should be case of exclusively electrostatic stabiliza-
made to the very comprehensive literature tion of powder dispersions in dissociated,
on this subject (Shaw, 1975 a; Temperley highly polar media (e.g., water) will be de-
and Trevena, 1978; Parfitt and Rochester, scribed in some detail, following the theo-
1983; Tadros, 1984a; Parfitt, 1986b). retical approach given by the so-called
Agglomeration effects of particles in liq- DLVO theory of Deryaguin, Landau,
uid media can be described in terms of the Verwey, and Overbeek (Deryaguin and
attractive and repulsive forces between the Landau, 1941; Verwey and Overbeek,
particles, or rather in terms of their poten- 1948).
tial energies Va and Vr. The total energy of This model of particle-particle interac-
interaction between two idealized parti- tion produces results the details of which
cles, Vt9 can be determined by the superpo- are often controversial. Many corrections
sition of three main components: and improvements have therefore been
 7.3 Materials for Casting Ceramic Tapes 227

necessary. Nonetheless, the validity of its

basic concepts remains undisputed. De- repulsive energy
scribing the fundamental principles of this
theory in context with the topics at hand,
offers the opportunity to introduce the total energy
special terminology used in this physico-
chemical field of interface science and to
focus attention on the specific methods ap- distance
plied to describe the relevant phenomena.
The DLVO theory starts with an expres- secondary
minimum
sion for the van der Waals attractive ener-
gy of two particles as a function of their / attractive energy
distance. The total interaction energy is
calculated by superposing a corresponding primary minimum
term for the repulsive coulombic energy of
equally-charged particles. This produces
Figure 7-20. Potential energy curve for two charged
total potential energy graphs of the type
particles as a result of superposition of attractive and
presented in Fig. 7-20. repulsive forces. A close approach of the particles and
At very small particle distances, a deep a strong attraction in a deep primary energy mini-
energy minimum indicates the relatively mum is prevented by a potential barrier. At a larger
high strength of particle agglomerates. At distance a flat secondary minimum may occur, im-
mobilizing the particles at this distance (after Tadros,
moderate distances, and depending on the 1984 b).
characteristics of the particle surface and
the liquid medium (such as particle surface
potential, permittivity of the liquid, con-
centration and valency of electrolyte ions)
a more or less pronounced positive energy forces occurring during tape casting (Par-
barrier can be built up, while at larger dis- fitt, 1986 a).
tances the interaction energy becomes The origin, character and local distribu-
zero. If the energy barrier is high com- tion of the electric charges are of special
pared with the thermal kinetic energy of interest. Ionization by dissociation or pre-
the particles, agglomeration is prevented ferred dissolution of particle ions are pos-
and the particles are kept apart from each sible mechanisms, but they are rarely pres-
other at some distance. ent in ceramic dispersions. The most fre-
At specific conditions, e.g., fairly large quent source of charge in this field is the
particle diameters and moderate elec- adsorption of ions or dipoles at the particle
trolyte concentrations, a flat secondary en- surface, stemming either from the suspen-
ergy minimum may occur. This can cause sion medium (e.g., H + and OH~ water
the particles to be immobilized at a rela- ions) or from small cationic or anionic
tively large distance - the paste-like system electrolytes in the slurry (NH^, SO^~).
is said to be stabilized by flocculation. Neutral oxide particle surfaces and even
Since the energy levels are relatively low, partially oxidized surfaces of nitrides and
this flocculated state is reversible, i.e., par- carbides, for example, form hydroxylated
ticles can easily be redispersed, for in- amphoteric groups in the presence of water
stance, by stirring or by the action of shear molecules (Me stands for a metal ion)
228 7 Tape Casting

(Parish et al., 1985): Particle surface

(6) Surface of shear

I I I I
-Me-O -Me-0—H
I I +H,0 o I I
-O—Me -O—Me-OH
I I I I
Depending on the pH of the ionic envi-
ronment, they will either dissociate form-
ing negatively-charged surface ions
(7)
I I
-Me-O—H - M e - O - + H(+)
( }
Diffuse layer
I I I I Stern layer
-0—Me-OH -O—Me—OH
I I I I

or they will be positively charged by bind-

ing H + -ions from the medium to the very
polar OH-groups
(8)
II II
-Me-O—H ,( +, ) -Me-O—H Distance
I I +H o I I
-O—Me-OH -O—Me-OH Figure 7-21. The structure of the electric double layer
11
V> according to Stern's theory. The potential at the
boundary between the immobilized charged mole-
cules in the Stern layer and the mobile charged
At high pH values, negatively charged molecules in the diffuse layer is defined as the £ poten-
surface sites are predominant (Morrison, tial (after Shaw, 1975 a).
1985), while at low pHs (i.e., high H + con-
centrations), the total net charge will be
positive. Consequently at a certain value Stern layer has to be regarded as an inte-
pH zpc , the so-called zero point of charge, gral part of the particle and the apparent
no net surface charge will be present. surface charge (responsible, for instance,
An important fact is that charged parti- for the repulsive forces of interacting parti-
cle surface sites attract corresponding cles) will be the true surface charge reduced
counter-ions from the solution. They will by the usually smaller counter-charges of
surround the pigments, a certain portion the Stern layer.
of them being bonded to the surface. Ac- From electrokinetic phenomena (e.g.,
cording to the so-called 'double layer theo- from measurements of particle movements
ry' of Stern (Jaycock, 1986), they will cre- due to electric fields), an electric potential
ate a thin layer of strongly-bound, immo- difference between the Stern layer and the
bile, charged molecules (Stern layer) while bulk of the liquid can be deduced. This
the larger portion of the counterions will so-called £ potential (especially its depen-
form an adjacent layer of loosely-bound dence on parameters such as pH value and
molecules extending much deeper into the permittivity of the liquid, or type and con-
medium (diffuse layer) (Fig. 7-21). The centration of particles and electrolytes in
 7.3 Materials for Casting Ceramic Tapes 229

the slurry) is of utmost importance for the

stabilization and characterization of ionic
dispersions. For oxide particles, the £ po-
tential is normally positive at low pH val-
ues and negative at high pH values, pass-
pH ie pH
ing the zero voltage line at the so-called
isoelectric point, pH iep (Fig. 7-22).
Ionic dispersions will be stable only if
they have a high absolute value of £ poten-
tial, i.e., if their pH is adjusted to a value Figure 7-22. Typical potential/pH dependence of an
sufficiently distant from the isoelectric oxide ceramic particle dispersion. The £ potential be-
point. As a working rule, a ( potential of at comes zero at the isoelectric point (pHiep).
least ± 30 mV is necessary for long-term
dispersion stability (Smith, 1986). Meth- crease the wetting properties of the liquid
ods for measuring this potential will be media, and to stabilize dispersions and
described in Sec. 7.3.3. emulsions.
If the solution contains additional elec- Non-ionic surfactants have OH groups
trolytic additives, they will compete with with the possibility of hydrogen bonding
the solvent ions for the charged particle to the surface. They may be used simulta-
surface sites. In the case of multivalent neously with ionic agents in the same sus-
electrolyte ions, their selective adsorption pension. Cationic surfactants are generally
can even result in a charge sign reversal of basic. They are not commonly used and
the Stern layer and, hence, of the £ poten- are often toxic. Zwitter-ionic agents like
tial according to the chemical equations of the well-known triglycerides contain an-
equilibrium: ionic and cationic groups and are able to
(9) bind in different types, according to the
() (2+)
ambient conditions upon the powder sur-
-Me-0 +K <^ - M e - 0 K(+) face and in the slurry. They are not gener-
II II ally water-soluble.
() Anionic and non-ionic surfactants, even
Me-OH 2 (+) -Me-OH2-A mixtures of them, are very popular disper-
sion stabilizers. Preferably, the hydropho-
C 2 + and A 2 ~ stand for divalent cationic bic chains consist of 10-20 carbon atoms
and anionic electrolyte ions, respectively. and can be interrupted by oxygen atoms,
Very effective in this context are surface double bonds, amide groups or other func-
active agents with molecules consisting of tional groups. Anionic surfactants are
a polar or ionic hydrophilic 'head' and a acidic in nature and may interfere with
non-polar hydrophobic 'tail' of linear or binder components. The most frequent an-
branched hydrocarbons. These so-called ions for hydrophilic head groups are car-
surfactants (Tormey, 1987; Mikeska and boxylates, sulfonates, sulphates, and phos-
Cannon, 1988; Ottewill, 1984), classified as phates. Ethylene oxide chains and hydrox-
anionic, cationic, zwitter-ionic, or non- yl groups are the preferred head groups for
ionic agents, are able to reduce surface and non-ionic surfactants.
interface tensions of liquids (especially Adsorption to uncharged surfaces (e.g.,
water). They are most broadly used to in- colloidal polymer particles) in highly polar
230 7 Tape Casting

media (water) normally occurs due to (Mikeska and Cannon, 1984). In other cas-
strong repulsive forces generated by the es, the hydrophobic tails may be squeezed
polar molecules. They act against the hy- out of the medium and then molecules will
drocarbon tail while the ionic part is dis- be oriented parallel to the surface, with
solved in the medium. The surfactant some resulting steric repulsion.
molecules are thus oriented vertical to the A characteristic feature of surfactants is
interface with the tails towards the surface that at a certain concentration these
of the particle. molecules associate to form larger units,
In the case of charged particles, things called micelles. The form of these associ-
generally become much more intricate. ates is rather variable (spheres, tubes,
There may be a repulsion between the plates, etc.), but their general constructive
charge of the particle and the ionic head or characteristic features the existence of re-
an attraction, depending on the signs of pulsive forces, exerted by the water mole-
the charges. In this latter case, the tails cules on the hydrocarbon tails of the sur-
may then extend into the more or less polar factants. The hydrophobic chains form an
medium. If water is concerned, their very inner core while the hydrophilic 'heads'
hydrophobic tails will attract tails of other generate the outer surface (Fig. 7-24)
ions, forming so-called hemimicelles, with (Cooper, 1984). The concentration at
the ionic heads of the surfactants oriented which this association phenomenon oc-
towards the medium. An overall charge curs, the so-called critical micelle concen-
sign reversal will be the consequence tration (c.m.c), depends on the type of the
(Fig. 7-23). surfactant and on temperature. At the
This sign reversal will not occur in less c.m.c. the solubility of the surfactant in-
polar liquids such as alcohols, where the creases strongly, while its activity remains
hydrophobic behavior of the tails is of mi- approximately constant.
nor relevance. The Stern layer may then Aqueous polymer dispersions are
keep its charge, i.e., remain positive in the known to be appropriate binders for the
case of anionic agents, for instance. But, tape casting process (see Sec. 7.3.2.1). The
naturally, a strong shift of the ( potential colloidal polymer particles may successful-
slope to smaller pH values will occur ly be stabilized in the water medium, for
instance, electrostatically by the use of ion-
ic surfactants. The tails of these molecules
are anchored to the surface of the polymer
particles by hydrogen bonding, while the
ionic heads are directed into the water
head medium. It goes without saying that sim-
tail head ultaneous electrostatic stabilization of
tail binder and powder particles in one singel
surfactant molecule dispersion may be difficult, since the oper-
ating window will become rather narrow.
In cases like this, it will be more appropri-
ate to use, at least partly, other methods
Figure 7-23. Charge sign reversal of particles due to such as steric stabilization.
hemimicelle formation of surfactant molecules (after Another interesting group of electrostat-
Mikeska and Cannon, 1984). ically stabilizing agents, very widely used
 7.3 Materials for Casting Ceramic Tapes 231

At pH values larger than approximately

4, the COONa groups begin to dissociate,
and at a pH value of about 9, full dissoci-
ation provides about 150 negatively
charged acid sites per molecule (Cesarano
et al., 1988). As anionic counter-ions, these
molecules are bound by the positive sur-
face sites. There, their large surplus of neg-
ative charges will cause a sign reversal of
the C potential even at very low concentra-
tions of the polymer (Fig. 7-25).
This assumes the pH value of the disper-
sion is larger than 4 (for reasons of suffi-
cient dissociation) and smaller than pH iep ,
where normally no adsorption takes place.
For maximum stability of the dispersion,
the quantity of adsorbent per square meter

• particle surface

Stern plane

Figure 7-24. Diagrammatic view of some micelle

structures.

in modern slurry technology, are the so-

called polyelectrolytes. These are salts of
polycarboxylic oxides with a very high £ potential
molecular weight (up to 20 000). They con-
tain a large number of acid sites which Figure 7-25. Sign reversal of the £ potential caused by
become progressively dissociated with in- the adsorption of multiply charged counter-ions. If
negatively charged polyelectrolytes are bonded to a
creasing pH values of the solution. As an positively charged particle surface, the £ potential (i.e.,
example, the structure of a Na salt of poly - the potential at the Stern plane) changes to negative
methacrylic acid (PMAA) is given: values (after Shaw, 1975 a).

Na salt of PMAA CH, CH,

with rc~50 H H H
I I I
—C- -C- -C- -C—h (10)
I I I
H H H
COONa COONa COONa
232 7 Tape Casting

powder surface has to be matched against to the sizes of the particles, thus limiting
the degree of its dissociation. This may re- their close approach in the suspension.
sult in narrow windows of operation for With such powders, highly concentrated
powders with low values of pH iep (Mor- slurries and densely packed green tapes,
rison, 1985). therefore, are not achievable.
Moreover, the PMAA ions can be multi-
ply-bonded to the surface sites, forming Steric Stabilization
loops and tails. They thus cover the parti-
In non-polar organic systems of low per-
cle surface even at a low polymer content
mittivity (aromatics, aliphatics) electro-
and establish an additional repulsion, re-
static repulsion will be only of minor im-
ferred to as 'electrosteric' stabilization
portance, while stabilization by adsorption
(Fig. 7-26). Such dispersants are very ef-
of long-chained polymer molecules plays
fective and are well able to stabilize sub-
the main role. Again simplifying the mech-
micrometer, highly concentrated aqueous
anism, macromolecules consisting of hy-
suspensions for tape casting. They can
drocarbons with acid or basic head groups
work even in slurries with particle mix-
or of amphipatic copolymers (i.e., hydro-
tures. With nanosized particles, the dimen-
philic and lipophilic groups in the same
sions of the organic hulls are comparable
molecule such as in phosphate esters (Bast,
1990), attach themselves to the particle
surface by means of suitable anchoring
groups, while the rest of the molecule ex-
tends into the liquid medium. These hy-
drocarbon chains are responsible for a
steric hindrance effect in that repulsive
forces from chain interactions prevent the
particles from approaching closely (Fig. 7-
loops
27). A very effective dispersing agent of
this group is menhaden fish oil. It attaches
to the particle surface by acid carboxylic
groups (Calvert et al., 1986 a, b).
The stabilizing effect is pronounced if
surfactant (1) good anchoring of the polymer on the
molecule
particle is achieved, (2) the loops and tails
of the polymer are compatible with the liq-

polymer chain

Figure 7-26. The principle of electrosteric stabiliza- particle

tion. Multiply bonded surfactant molecules have a
high number of charged sites. The free charges cause
a sign reversal of the apparent particle charge, provid-
ing repulsive electric forces between the particles. Ad- Figure 7-27. Schematic view of the stabilizing effect
ditionally steric repulsive forces are built up by the of long hydrocarbon polymer chains caused by steric
"loops" and "tails" of the polymer chains. hindrance (after Bohnlein-MauB et al., 1992).
 7.3 Materials for Casting Ceramic Tapes 233

polymer cha

Figure 7-28. Very effective steric stabili-

zation of particles by polymer molecules
multiply anchored via hydrogen bonds.

uid medium, (3) the polymer covers a great stabilization if their molecules had incor-
part of the particle surface and the layer porated specialized parts both for an-
cover is thick enough (Sacks and Scheif- choring (A) and for stabilizing tasks (B).
fele, 1986; Tadros, 1982). This can be accomplished by forming ap-
Particle anchoring can be achieved, for propriate AB block-copolymers, tailored
instance, by hydrogen bonding of hydrox- and optimized with regard to the A and B
yl groups in the polymer to the oxide sur- composition as well as to their sequence
face, or, if shorter surfactant molecules are and their position in the molecule, for
involved, by acid-base reactions. Multiple optimum adaptation to the current prob-
anchoring of one individual polymer mole- lem (B6hnlein-Mau6 et al., 1992; Dawkins
cule by means of suitable side groups is etal., 1982).
very efficient, since the molecule forms In terms of the potentials of the steric
loops and tails, thus covering a larger part repulsive forces (Vs) and the attractive van
of the surface (Fig. 7-28). These parts of der Waals forces (JQ, superposition will
the molecules should be compatible with have the general form given in Fig. 7-29.
the medium, i.e., they should be well sol- The repulsive potential energy (Vs) results
vated. If not, the theory predicts an attrac- from the interaction of particles that carry
tion between the polymer chains rather adsorbed polymer layers. There are two
than a stabilizing steric repulsion. main contributions to this steric repulsion
Long-chained polymers with molecular potential. The first results from an increase
weights of more than 10000 are more ef- in free energy and the second is due to a
fective in steric stabilization than shorter reduction of entropy of the interfering
molecules, since they form thicker adsorp- polymer chains (Tadros, 1984 b). No short
tion layers, and the sensitivity of the sys- distance primary minimum will be found,
tem to fluctuations of process parameters since the polymer cover of the particles will
is less pronounced. prevent a close approach.
From these statements it can be implied The secondary minimum is relatively
that polymers would be highly suitable for deep for short polymer chains and shallow
234 7 Tape Casting

special slurry properties or to establish de-

total energy sired characteristics in the dried tape.
Remembering the described difficulties
repulsive energy which may arise in multicomponent sus-
pensions from undesired interactions of
the various components, it is consistent to
say that the number of additives should be
kept to an absolute minimum. Therefore,
the large majority of slurry compositions
described in related publications do not
contain any of these special functional
agents. Nevertheless, it may be useful to
enumerate them here and to make some
brief remarks on how they are used
(Moreno, 1992 a; Roosen, 1988; McHale,
energy minimum
1991; Morris and Cannon, 1986; Karas
etal., 1988).
Wetting agents are surface-active agents
primarily soluble in the liquid phase which
Figure 7-29. Total potential energy of interaction (Vt) are required to reduce the surface tension
for steric stabilization. The high potential barrier at of the liquids (especially of water), and en-
small distances prevents a close approach of the par- hance their wetting properties for powders
ticles (no primary minimum) (after Tadros, 1984 b).
and substrates. For this reason, these sur-
factants are also employed as dispersants,
as mentioned earlier in this section.
for larger molecules, and depends on the Defoamers are helpful mostly in aque-
polymer solvency in the system. This sec- ous media, where the build-up of deleteri-
ondary minimum is responsible for a re- ous stable foams from polymer solutions
versible flocculation which may or may (PVA) or polymer dispersions (PMMA) is
not be desired. If sufficiently pronounced, likely to occur, especially during milling. It
it may even result in a gelation of the sus- is much more effective to prevent foam
pension. Furthermore, in contrast to elec- formation (with agents such as special
trostatically stabilized suspensions, most waxes, or with vacuum milling) than to
steric dispersions are extremely sensitive to attempt to destroy the foams.
temperature variations, flocculating either Homogenizers such as cyclohexanone
on cooling or on heating. This effect is are used to increase the mutual solubility
reversible and can be attributed to the spe- of the components, thus preventing skin
cial temperature dependence of the poly- formation during drying. They will also
mer solvency. increase the density and tensile strength of
In Table 7-8 a small selection of dispers- the tapes.
ing agents, frequently used in tape casting Preservatives, which are generally very
systems, is presented. toxic, can be added to suppress bacterial or
Concluding this section on slurry addi- fungal attacks which occur frequently in
tives some functional ingredients should be aqueous binder systems and in tapes stem-
mentioned which may be used to provide ming from them.
 7.3 Materials for Casting Ceramic Tapes 235

Table 7-8. Selected dispersants for tape casting slurries.

Dispersants for Electrostatic Stabilization

Aqueous systems Non-aqueous systems
Sodium silicates chloroform
Sodium carbonates methylene chloride
Sodium polyphosphates ketones, ethers
Ammonium polyphosphates acetonitrile
methyl ethyl ketone + methyl alcohol
Surface-Active Agents
Non-ionic surfactants Cationic surfactants
Ethoxylate of caster oil amines from fatty acids
Diethanolamine condensates amine oxides
Glycerol esters substituted imidazolines
Anionic surfactants Zwitterionic surfactants
Phosphate mono- and diesters triglycerides
Sulfonates, sulfates of oils and fats methylamine salts of polycarbonic acids
Carboxylic acid esters, glycerol esters
Sulfosuccinates

Amphiphilic Copolymers for Steric Stabilization

(A, anchor group; B, stabilizing group)
Aqueous systems Non-aqueous systems
Poly(vinyl chloride) (A) polyethylene (A)
Polyethylene (A) polyoxyethylene (A)
Polymethylmethacrylates (A) poly(vinyl chloride) (A)
Polyoxyethylene (A) polymethylmethacrylate (A)
Poly(vinyl alcohol) (B) polystyrene (B)
Poly(acrylic acid) (B) polymethylmethacrylate (B)
Poly(methacrylic acid) (B) polyisobutylene (B)
Polyoxyethylene (B) poly(vinyl acetate) (B)
oxidized menhaden fish oil (A and B)

Flow control agents such as liquid poly- liquid (Shaw, 1975 b). To produce a certain
ethylene are sometimes added in small shear rate dvl&x (i.e., a local gradient of
quantities to prevent the surface of the velocity within a fluid system), an ade-
tape from drying too rapidly, which might quate shear stress x has to be provided for
provoke cracks. the molecules or particles to be moved rel-
Flocculants are agents which prevent ative to each other. If there is a propor-
dispersions from forming extremely high tional response between this shear stress
density sediments. They are counteractive and the shear rate over a large range of the
to dispersing agents by shifting the pH of parameters (e.g., in pure liquids), the rheo-
the suspension to values close to pH iep . logical behavior is said to be Newtonian
and the factor of proportionality rj is de-
7.3.2.5 Rheology of Tape Casting Slurries fined as viscosity (T = rjdv/dx).
Flowing liquids or objects moved rela- In actual dispersions, however, and es-
tive to a liquid phase generate shearing ef- pecially in concentrated dispersions, such
fects which can be used to characterize the behavior is rather unlikely to occur be-
236 7 Tape Casting

cause of the multiple interactions between high enough for them to be destroyed
the components (Goodwin, 1990). A non- (McKinnon and Blum, 1984). This shear
Newtonian flow characteristic can then be thinning behavior is advantageous with
established, with the viscosity rj itself de- tape casting slurries. The high viscosity of
pending intricately on the shear rate (Rus- the undisturbed dispersion in the reservoir
sel, 1987). contributes to the resistance of the slurry
With tape casting slurries, the most fre- to sedimentation, while during tape cast-
quent forms of shear rate/shear stress ing the relatively high operating shear rates
curves are of plastic, pseudoplastic, dila- of 100-1000 s" 1 cause a more suitable,
tant, and of thixotropic type (Fig. 7-30). lower viscosity (Chartier et al., 1988; Day-
Plastic behavior is characterized by the ton et al., 1984).
existence of a more or less pronounced Dilatancy, i.e., shear thickening, is in
yield value of shear stress below which no contrast very deleterious since it can pro-
flow occurs. It can frequently be found in voke insufficient flow during slurry pro-
concentrated dispersions, which generate a cessing and, above all, during tape casting.
structural network. If the yield value is Its manifestation indicates the existence of
very small, plasticity changes to pseudo- an extremely concentrated, stabilized dis-
plasticity, which is considered to be the persion with insufficient compatibility be-
most suitable rheological behavior for tape tween medium and stabilizing molecules.
casting. It is associated with a shear thin- By exertion of shear stress, the liquid is
ning effect, i.e., a decrease of viscosity val- pressed out of the interstices between the
ues at increasing shear rates. particles, thus concentrating the dispersion
Pseudoplasticity is the characteristic vis- in this region over a tolerable upper limit
cosity behavior of weakly flocculated sus- for any flow (Hampton et al., 1988).
pensions, the particles being held for in- Thixotropy is a special time-dependent
stance in a shallow secondary energy mini- type of flow behavior. If, especially with
mum. The mechanism governing this effect sterically-stabilized dispersions, a slurry is
can be described in terms of the release of left undisturbed for a certain period of
liquid medium entrapped within agglom- time, a skeleton of very weak interlinked
erated floes, if the shear stresses become bonds may be generated, leading to a mea-
sure of rigidity which can be readily de-
stroyed even by the small shear stresses
thixotropic built up during measurement. If the effect
of thixotropy is not very pronounced, i.e.,
dilatant if the times for destruction and recovery
pseudoplastic are short, this behavior is considered to be
plastic advantageous in some special cases of tape
yield value fabrication.
Slurry viscosity is, moreover, an appro-
shear stress priate means for characterizing the effi-
Figure 7-30. Non-Newtonian flow characteristics of ciency of dispersing agents. When particles
tape casting slurries. With increasing shear-rates, are dispersed in a liquid medium, the vis-
shear stresses increase sub-linearly with plastic, pseu-
doplastic and thixotropic slurries (shear thinning be- cosity of the system is higher than that of
havior) and super-linearly with dilatant slurries (shear the pure liquid. The magnitude of this in-
thickening behavior) (after Shaw, 1975 b). crease provides some information on
 7.3 Materials for Casting Ceramic Tapes 237

70
parameters such as degree of dispersion, shear rate 9 s"1
interaction of slurry components, or effi- ^ 60
ciency of dispersing agents. in
& 50
Defining rj0 to be the viscosity of the
24o
liquid medium and r\ the viscosity of the
dispersion, the so-called relative viscosity,
rjTel, can be calculated from their ratio 20
(Morris and Cannon, 1986; Braun etal., shear rate 90 s
10
1985)
rjrel = t]/r]0 (7-5) 10 12
pH
In the case of concentrated dispersions, Figure 7-31. Viscosity of an electrostatically stabi-
the values of rjrcl are always appreciably lized aqueous powder suspension of a TiO 2 /Al 2 O 3
higher than unity. Plotting the relative vis- (1:2 wt%) mixture as a function of pH. At low shear
cosity at a constant shear rate against the rates the formation of pronounced viscosity maxima
at the isoelectric points for TiO 2 (pHiep = 5.6) and
quantity of the added electrostatically dis- A12O3 (pHiep = 9.0) becomes visible [after Rao (1987),
persing agent will show a pronounced de- with kind permission from Elsevier Science Ltd. The
crease of viscosity followed by a normally Boulevard, Langford Lane, Kidlington OX5 1GB,
very flat viscosity minimum. The reason UK].
for this effect is again the dstruction of
agglomerates in the course of powder dis-
persion, which releases the liquid medium methacrylates (Sacks and Scheiffele, 1986;
which was entrapped and immobilized Braun etal., 1985, Cannon etal., 1986).
within the pores of these agglomerates Slurry compositions are therefore often
(Mikeska and Cannon, 1984). The subse- formulated without additional disper-
quent increase of rjrel is due to the begin- sants, which greatly simplifies the set-up of
ning of a sign reversal of the £ potential. the suspension. Thus, the process of com-
Moreover, electrostatically stabilized peting adsorption of these dispersant
suspensions are deflocculated below and molecules on the one hand and of solvent,
above their isoelectric point. Viscosity/pH plasticizer, and binder molecules on the
dependences at low shear rates, therefore, other is avoidable and the result will be less
are characterized by a pronounced maxi- equivocal.
mum at this isoelectric point. Rao (1987)
has reported the appearance of even two
7.3.2.6 Slurry Preparation
distinct viscosity maxima as a function of
pH value, if mixed powder dispersions of Preparing a tape casting slurry will begin
2 part alumina and 1 part titania are with a procedure that can be described as
measured at a small shear rate of 9 s" 1 milling or intense mixing of the ceramic
(Fig. 7-31). powder with some liquid medium, for in-
Careful measurement of relative viscosi- stance by using an ultrasonic probe (Blum
ty values of binder solutions before and and Cannon, 1985).
after pigmentation with powders has re- Ball milling is appropriate if wear and
vealed a very pronounced dispersing ef- solvent resistant materials are used (e.g.,
fect, i.e., a viscosity decrease developed by polyethylene). In some special cases, the
polymer binders such as butyrals and poly- milling vessel has to allow the use of inert
238 7 Tape Casting

atmospheres or even vacuum conditions. If casting facilities demand higher slurry

Mixing and milling will be relatively inef- viscosity, the binder must be added sepa-
fective if the viscosity of the slurry is high, rately using, for instance, a stirring ma-
which unfortunately is normally the case chine. Another, more practicable and effi-
with suspensions ready for casting. More- cient expedient would be to add a certain
over, intense milling in the presence of surplus of solvents to the slurry, in order to
binders could even cause damage to the reduce the viscosity, and to evaporate
long molecular chains of the polymer. It is these solvents at the end of the mixing pro-
favorable, therefore, to observe a certain cess. It should be pointed out, however,
sequence when adding the slurry compo- that if non-azeotropic solvent mixtures are
nents and to divide the process into two applied, their composition in the slurry
parts (Boch and Chartier, 1988; Morris will be changed during this procedure ac-
and Cannon, 1986; Cannon et al., 1986). cording to their different volatilities.
During the first period the milling effect After milling and mixing, slurries nor-
should be predominant, with the aim of mally contain a certain amount of air
breaking up particle agglomerates and which has to be removed by evacuation
wetting the powder. Hence, the slurry prior to casting. An associated increase of
should contain only the powder, the sol- the slurry viscosity due to the loss of sol-
vents, and the dispersing agents (Cannon vents must therefore be taken into consid-
et al., 1989). During this part of the milling eration at the set-up of the slurry formula-
process, the dispersing agents will have tion. Deairing should be continued until
enough time to occupy most of the corre- the desired value of slurry viscosity is
sponding sites on the powder surfaces, reached.
which is a necessary prerequisite for their Finally, immediately before casting, or-
optimum efficacy. ganic or inorganic residues such as binder
The second part of the process should be lumps or small debris of grinding medium
dedicated to mixing this slurry with the must be removed from the slurry by pass-
more viscous plasticizers, the highly vis- ing it through a fine-meshed sieve (5-30 Jim
cous polymer solutions (solvated in a small openings) (Williams, 1976; Shanefield and
portion of the solvents), and any addition- Mistier, 1976).
al functional additives. If no special disper- In Table 7-9 some typical slurry formu-
sants are applied (i.e., if only the dispersing lations used for tape casting are compiled.
properties of the binder polymer are used),
a small amount of this polymer binder has
to be included in the slurry during the first 7.3.3 Assessment of Properties
milling stage. The main part of the binder
Powder Properties
has to be added subsequently, as described
above. Characterization of powders consists
Times for milling and mixing should be mainly of determining their size, their
long enough to achieve stable conditions shape, their surface condition, and their
and a high homogeneity. A milling time of state of agglomeration. For this purpose a
24 h plus a mixing time of the same range number of effective methods are available
have been proven to be adequate, provided some of which will now be described
the viscosity is at all low enough for ball briefly (Malghan and Dragoo, 1991;
milling (rj < 1000 mPa s). Wood, 1991).
 7.3 Materials for Casting Ceramic Tapes 239

C ON

II 1
O O

§1

6 jam

8
a
I
hai

•8 o
nsi

o
(-1
1

oo © ON
(N <N ©
ylph

o PH

i o
0
w
PH
PH
PQ W
PQ PH
0 PH
PQ
Q
0
P-i
PH
PQ
PQ w
PH
'ao PH
W o
b)
2|im
Figure 7-32. Electron micrographs of two A12O3
powders differing in particle size and shape and in the
state of agglomeration.

PQ PQ
>
PH PH PQ S u
Electron microscopy is a very useful and
© effective tool for this task. Most of the
©
delineated characteristic features may at
least qualitatively be obtained from one
II single image (Fig. 7-32). Problems may
O
CJ
a>
O
o
T! P
<i>
arise from the normally non-representative
13
small section being examined and the limi-
i S 11 tation to dry state powder conditions.
Grain size and size distributions in very
© r^
dilute suspensions can be determined by
oo oo o laser light methods. Using laser scattering
or laser diffraction methods, particles sized
240 7 Tape Casting

from 3 nm to 1000 jim are detectable, with tribution of the primary particle surfaces
the resulting diameters being equal to in the interior of the agglomerates and,
those of spheres which would produce hence, represents the degree of agglomera-
equiareal scattering or diffraction pat- tion.
terns.
Classical methods, such as sedimenta-
Slurry Properties
tion of particles in gravitational or cen-
trifugal force fields, have simultaneously Measurements concerning the charac-
been developed to a high standard (Bern- teristics of the suspending media, such as
hardt, 1990). These relatively fast measure- their wetting behavior and their surface
ments may result in particle diameters tension, are performed by the use of ten-
equivalent to spheres with the same set- siometers and optical systems which visu-
tling velocity and ranging from 10 nm to alize the angle of contact at the solid/liquid
60 jim. interface. Tensiometers measure the forces
There seems to be no simple, commer- necessary to enlarge the free surface of a
cially available, practicable method for de- liquid by a unit area. A low surface tension
tecting actual particle sizes, i.e., the status is not only important for the immersion of
of agglomeration, in highly-concentrated powders into the liquid medium but also
dispersions such as tape casting slurries. for their dispersion within this liquid. Sim-
The total specific particle surface, in- ilar information can be obtained measur-
cluding all open accessible pore surfaces, ing wetting angles. The silhouette of a drop
can be detected by well-established meth- of the liquid on a polished, plane surface of
ods such as N 2 adsorption (BET) (Allan, the solid material is optically reproduced
1981). Particle size measurements and sur- in the ocular of a microscope where the
face measurements can be combined to ob- wetting angle yL/v can be registered. Both
tain some information on the state of measurements are useful for assessing the
agglomeration. From the grain size distri- effectivity of additives such as dispersants
bution function s=f(d), where s is the rel- or wetting agents.
ative frequency with which a certain di- Values of the f potential of dilute aque-
ameter d occurs in the distribution (sAd ous suspensions are obtainable from elec-
being the mass fraction of the powder with trophoretic mobility data of the charged
particles diameters of d± Ad/2), a theoret- particles in an electric field. If measure-
ical value for the total apparent surface ments are performed over the whole range
area (TAS A) can be calculated by integrat- of pH values, a potential/pH diagram is
ing these differential values for the whole produced, a key factor for controlling col-
range of measured particle sizes (Davis loidal particle behavior in slurries. From
etal., 1971): these plots, the isoelectric point (pHiep) is
readily obtained as the point where the
TASA = 6 t ^t
d\ Qd
d(d) (7-6) £ potential crosses the zero voltage line.
Technicians are often exclusively inter-
ested in the determination of this special
The difference between the calculated pH iep value or, for reasons of process con-
TASA value of a certain powder and its trol, in the actual sum of all particle
actually accessible surface from adsorp- charges. For this purpose, simple inexpen-
tion measurement is mainly due to the con- sive measuring devices are available, e.g.,
 7.3 Materials for Casting Ceramic Tapes 241

the particle charge detector (PCD) from An appropriate low-cost method for
Malvern Instruments, U.K. Its operating evaluating the degree of dispersion in tape
principle is based on the build-up of an casting slurries, either electrically or steri-
electric potential between an electrode, cally stabilized, is based on sedimentation
covered with an adsorbed layer of charged measurements. Flocculated suspensions
particles and the bulk of the slurry. Their will settle faster and, due to bridging ef-
relative motion separates the fixed particle fects, will result in rather loosely-packed
charges from the diffuse part of the double sediments. Highly stabilized suspensions
layer. By titration with a suitable elec- take much longer to settle, and produce
trolyte, the pH iep can be determined as the very dense 'cakes' of powder particles
pH value at which the signal becomes zero. which can hardly be redispersed by stir-
The original mean charge of the Stern lay- ring. The high density of these sediments is
er of the particles can be calculated from caused by the fact that the small primary
the quantity of consumed electrolyte. particles of the stabilized suspensions are
A more recently developed method for able to reorient themselves after settling,
measuring f potentials is the so-called elec- ideally forming a structure of theoretically
trokinetic sonic analysis (ESA) (Lehmann ultradense particle packing. This is very
etal., 1993; Graule and Gauckler, 1993). close to the situation present in dried tapes
With a pair of electrodes, a high-frequency cast from well-dispersed slurries, even if
electric field is applied to the suspension, the process of tape drying cannot be de-
generating a corresponding high-frequen- scribed in terms of particle sedimentation
cy oscillation of the charged particles. (see Sec. 7.4.2).
Ultrasonic waves propagate through the Several commercially available instru-
medium and are detected, for instance, by ments are widely used for viscosity mea-
piezoelectric sensors. The received signal surements. In the simplest method, the
amplitude is proportional to the £ poten- time required for a certain quantity of slur-
tial and to factors such as particle concen- ry to pass through an adequately narrow
tration, amplitude of the electric field, or capillary tube is measured. Since no varia-
particle mobility. tion of the shear rate is possible, this so-
Measurements based on this method are called Ostwald viscometer is suitable for a
very accurate and available within short fast viscosity check during a running tape
times. Their most important feature, how- casting process or for a rough viscosity
ever, is that they are obtainable from actu- comparison between different slurry
al, concentrated slurries ready for tape batches.
casting. The measurement samples need Modern, mostly computer-controlled
not to be greatly diluted, as is the case with viscometers, suitable for the measurement
methods based on mobility measurements. of complex rheological behavior of slur-
This is an important advantage, since in ries, are based on the relative movement of
most cases the £ potential of a dispersion rotating concentric cylinders placed at
changes by extreme dilution. Additionally, very small distances from each other. The
the C potential of non-aqueous suspen- liquid within this narrow gap attaches to
sions, which is not equal to zero for many the cylinder walls and is sheared if one of
organic based slurries, can be easily ob- them is rotated while the other is kept
tained by this method, without special fixed. The shear rate is determined by the
titrating agents. surface velocity of the rotating part and
242 7 Tape Casting

the gap width between the cylinders and ing smaller units for tape casting on a lab-
should be adjustable within a broad range. oratory scale or close to that. Diversified
Shear stress is obtainable by measuring the ceramic tape applications, as well as prop-
restoring torque exerted upon the rotating er compositions and processes, demand
cylinder. Shear rate- as well as shear stress- rather highly adapted and specialized pro-
controlled units are available. Efficient duction lines. For this reason, these pro-
cooling of the system is required, especially duction lines are often in-house develop-
at high rotational speeds, since the vis- ments.
cosity of binder-containing suspensions is Two types of production methods are
extremely sensitive to temperature varia- distinguishable as illustrated schematically
tions. in Fig. 7-33. They mainly differ in the way
Optimum viscosity values for tape cast- by which the slurry is applied to the sup-
ing slurries vary in a relatively broad range porting substrate. It makes sense to refer
of about 500 to 25 000 mPa s, depending to the technique shown in Fig. 7-33 a strict-
upon the working principle of the casting ly speaking as a 'casting' process, while for
device and the desired thickness of the tape the process of Fig. 7-33 b 'coating' seems
(see Sec. 7.4.1). to be a more appropriate name.
With casting methods the slurry is 'doc-
tor bladed' onto the substrate by forcing it
7.4 The Tape Casting Process to pass through a gap of well-defined
width, which is the determining factor for
Detailed technical descriptions of tape the thickness of the ceramic tape. This can
casting facilities are rather rare in the pub- either be done by moving the casting head
lished literature, and only a few commer- over a fixed substrate (non-continuous
cial suppliers are known, mainly produc- working) or, vice versa, by pulling the sub-

a)
fixed casting head

slurry
plastic substrate
or steel belt

-_-_-_-_-_-__-_frC-__~-~__~~. wet tape

j direction of
movement

b)
plastic substrate

direction of double coated Figure 7-33. Diagrammatic view of two

movement tape substrate different types of tape production, (a) A
slurry layer is doctor bladed onto a sub-
strate that is moving relative to the cast-
fixed casting head
ing head (casting method), (b) A flexible
substrate is passed through a suspension
slurry
and is covered with a layer of slurry
(coating method).
 7.4 The Tape Casting Process 243

strate under the fixed reservoir (continu-

ous working).
Coating methods are based on the fact
that when a flexible substrate is passed
through a slurry, a certain quantity of sus-
pension will adhere to the substrate and
after it has left the slurry surface, the sub-
strate will be covered by a thin layer of
suspension. In this case, the most impor-
tant thickness-determining factor is slurry
Figure 7-34. Non-continuously-working laboratory
viscosity. Other significant factors are the casting units. A casting head with a gap that can be
velocity of substrate movement, the angle adjusted or chosen is drawn over a glass plate (pro-
of declination at which the substrate leaves vided with a plastic tape support) either by means of
the suspension level and the specific weight a driving motor (a) or simply by hand (b). In the
illustrated case, a movable infrared lamp is used for
of the slurry. It is self-evident that very
drying (c).
good wetting of the substrate by the sus-
pension is a necessary prerequisite for this
method. Generally, coating units are con-
tinuously working plants and hence, suit-
able for the production of somewhat larger
quantities of tapes.

7.4.1 The Practice of Tape Production

7.4.1.1 Tape Casting Methods
Laboratory tape casting facilities for in- //* ?
vestigation purposes are generally small,
non-continuously working units of the Figure 7-35. Continuously working industrial cast-
doctor blade type. Since the quantity of ing unit showing the long drying tract and a closed
required tapes is usually small, batch-wise drying tunnel (courtesy of C. v. Stein, Siemens AG).
casting is appropriate, with the unit mov-
ing over a fixed substrate. Drying can then
be done in separate drying units of ade- ing of a covered slurry reservoir to prevent
quate size. As an example, a photograph solvent evaporation, and an adjustable
showing two such laboratory casting facil- blade for determining the tape thickness.
ities is reproduced in Fig. 7-34. Larger in- The gap of the blade was monitored ac-
dustrial units of this type, such as that cording to an X-ray transmission signal,
shown in Fig. 7-35, normally have to work the magnitude of which depends on the
continuously, the largest part of the ma- deposited slurry layer thickness. The tape
chine being dedicated to the drying of the was cast on a smooth aluminum plate cov-
tape. ered with a thin band of cellulose acetate.
Frequently, publications refer to a plant This substrate tape was drawn under the
first mentioned by Shanefield and Mistier casting head at a velocity of approximately
(1971). They described a casting apparatus 18 cm/min with a gap of 1.5 mm, produc-
for monolayer alumina substrates, consist- ing a dried tape of about 650 |im thickness.
244 7 Tape Casting

Because of the low casting velocity, the substrate is governed almost entirely by
drying zone could be limited to only 7 m. the height of the blade gap. Further im-
This machine, itself based on a similar provements concerned the stability of the
earlier apparatus introduced by Howatt table base, the thickness tolerances of the
et al. (1947), has been improved by Runk carrier and the smoothness of the tape
and Andrejco (1975) for the production movement. Moreover, detrimental tape vi-
of 25-250 nm PZT tapes. They found a brations were avoided by drawing the sub-
double-blade construction to be able to strate belt over a plate glass bed, bent into
provide narrower thickness tolerances an arc with a radius as large as 130 m
( + 4%) over large tape areas. The first (Fig. 7-37). This enabled close contact be-
blade reduces the hydrostatic pressure in tween the carrier and the bed.
the region of the second blade (Fig. 7-36). These early casting units have been the
Thus the rate of slurry deposition on the basis for many later industrial and labora-

slurry reservoir
casting head

two doctor blades

polymer substrate
glass bench

direction of * l
movement T
cast tape
Figure 7-36. Double-blade construction for the production of tapes with narrow thickness tolerances. A first
blade reduces the hydrostatic pressure in the slurry reservoir to a lower value in the casting chamber between
the two blades. During casting, this pressure in front of the second blade, which determines the tape thickness,
remains approximately constant (after Runk and Andrejco, 1975).

drying ceramic tape

casting head
polymer substrate

coiling drum

curved glass bench

Figure 7-37. Tape casting unit with a flexible substrate drawn over a lightly arched glass bed in order to avoid
tape vibrations (after Runk and Andrejco, 1975).
 7.4 The Tape Casting Process 245

tory plants, with improvements mainly might be even impossible for tapes thicker
dealing with thickness control and casting than approximately 150 \im. A small ra-
velocity. Modern casting units normally dius of curvature would cause a non-elastic
work at velocities of several meters per deleterious stretching of the outward sur-
minute (Mistier, 1991; Fiori and DePortu, face of the tapes. Thick tapes, used pre-
1986; Boch and Chartier, 1988; Roosen, dominantly for monolayer structures,
1988). therefore have to be kept plane by cutting
With continuously working machines, them into smaller pieces and storing them
maximum casting velocity is closely related as stacks of plates or stripes.
to the length of the drying tract, which Especially with the production of thick
corresponds to the available time for the tapes, a special problem has to be consid-
evaporation of solvents. The required ered, caused by the tendency for a certain
length of the drying bench depends on the side flow of slurry to occur at the edges of
thickness of the cast tape and the evapora- the tape immediately after the cast layer
tion rate of the solvents. For tapes with leaves the blade gap. This effect is gov-
thicknesses close to 1 mm, casting ma- erned by the low shear rate viscosity. This
chines up to 35 m in length have been built. viscosity should be adjusted to relatively
On the other hand, thin tapes with thick- high values (3000-10000 mPas) even if a
ness ranging from 25 \im to 50 jim can be paste-like consistency might bring about
cast at velocities of up to 15 m/min, requir- problems with slurry preparation (see
ing a drying distance of only several me- Sec. 7.3.2.6).
ters. These problems, however, are less signif-
At the end of this distance, the tape icant and dimensional stability is even im-
should be dry enough to be peeled off from proved, if slurries with strongly pseudo-
the substrate if the latter consists, for in- plastic behavior are used. During passage
stance, of an endless steel belt. Or it should through the blade gap, the viscosity of the
be rolled up together with the substrate if slurry is low due to the high shear rates in
the tape is thin and the substrate likewise this region. Behind the gap the shear rate
consists of a thin, flexible band of plastic. drops to values near zero, with a pro-
Cellulose triacetate, Mylar, Teflon, coated nounced increase of viscosity. Thixotropic
paper, polyethylene (PE), polytetrafluor- behavior, evidently, is much less suitable
ethylene (PTFE) and some other materials for this purpose, because the viscosity in-
have been reported to be suitable for this crease behind the gap requires a certain
latter purpose (Mistier, 1991, 1990). Con- recovery time.
trary to frequent reports, these substrate The described casting method can like-
belts should not be used repeatedly. The wise be used to produce tapes as thin as
plastic substrate could be chemically agi- 20 |im. Generally, casting units will be-
tated by the slurry solvents or stretched by come more compact with thin layers, since
drawing it over the supporting bed. One extended drying distances are not re-
important criterion for substrate selection, quired, and viscosity can then be reduced
therefore, should be besides its costs, the drastically to values less than 500 mPa s.
recyclability of the material, or at least, Furthermore, the use of space-saving heat-
easy disposability of residuals. ed wheels becomes possible. In this case, a
Space-saving storing of the tapes as first preconsolidation of the wet layer in a
coiled-up spools is not always adequate or small horizontal passage is required before
246 7 Tape Casting

the tape comes in contact with the wheel,

where it is inclined due to its curvature.
Consequently, this method is only suitable
for thin organic-based tapes with fast rates
of evaporation.
A similar drying principle is used in a
compact commercial tape casting unit pro-
duced by Cladan, Inc., U.S.A. With this
system, suitable for organic solvent-based
tapes of approximately 100 j^m maximum
thickness the tape is dried in an enclosed,
ventilated box by moving it after a first Figure 7-38. Continuously working casting unit with
horizontal distance over a series of rollers tangential deposition of slurry on a plastic belt and
with a totally enclosed slurry reservoir (Cladan Inc.,
in a spiral rate. The slurry is deposited U.S.A.). The casting head (a) is provided with an ad-
through an adjustable blade clearance tan- justable blade clearance. A uniformly turning wheel
gential to a driven cylinder in contact with with a very smooth surface (b) moves the plastic belt,
the plastic belt support (Fig. 7-38). The which is covered on one side with a thin layer of
thickness tolerances over the whole tape slurry. In a ventilated drying chamber (c) the tape is
drawn over a series of rollers at a spiral rate.
are reported to be about 2.5 jum.
In order to estimate the influence of the
various casting parameters on the thick-
ness of the deposited layer, a theoretical / wet tape
fluid flow model of the casting process has zdry tape
been developed by Otsuka et al. (1986 a, b)
ID
and likewise by Chou et al. (1987). For a
slurry with Newtonian rheological behav- substrate * L drying
distance
ior and a laminar flow in the gap of a
simple casting unit (Fig. 7-39), they found Figure 7-39. Model casting unit for tape thickness
calculation. D, thickness of the dried tape; h, gap
an expression for the thickness D of the height; L, gap length; v0, casting velocity.
dried tape

(7-7)
6rj Lv0 tional to the gap height and nearly inde-
a signifies a factor which reflects the thick- pendent of all other casting parameters if
ness shrinkage of the wet layer during dry- the second term within the parenthesis
ing, h and L are the height and the length could be reduced to a value of far below
of the blade gap, respectively, r\ is the vis- unity.
cosity of the slurry, A/? (normally deter- For narrow gaps of less than approxi-
mined by the height of the slurry level) is mately 200 jam, which means for tapes
the pressure difference between entry and thinner than about 100 jim, this can be
exit of the gap, and v0 stands for the rela- achieved if the parameters rj, v0, L and Ap
tive velocity of the casting unit and the are kept within certain ranges. This will
substrate. become more obvious if a dimensionless,
From this equation, it follows that the normalized value for the tape thickness
tape thickness would simply be propor- (the term of relation in parentheses) is cal-
 7.4 The Tape Casting Process 247

Table 7-10. Values of the calculated normalized tape thickness D [2/(a h)] for different sets of casting parameters.

Parameter PZT slurry PZT slurry A12O3 slurry A12O3 slurry

Q slurry (g/cm3) 3.9 3.9 2.0 2.0

Slurry level (cm) 4 4 4 4
Mem) 0.065 0.020 0.065 0.020
rj (mPa s) 1000 1000 1000 1000
v o (cm/s) 1 5 5 5
L (cm) 0.1 1 0.1 1
Ap (Pa) 1500 1500 800 800
/ 2\ / h2 Ap \ 11.6 1.02 2.13 1.01
\ahj \ 6r]Lv0J

culated and compared for different sets of is a much more convenient solution, since
casting parameters, as compiled in Table the system becomes less sensitive to fluctu-
7-10. ations of the slurry level in the reservoir.
Large gaps can give rise to appreciable With pseudoplastic fluids the tape thick-
deviations from this simple linear thick- ness depends even more strongly on the
ness dependence on the gap height. The gap width and the casting velocity. These
effect is more pronounced with small cast- parameters determine the viscosity in the
ing velocities v0, low slurry viscosities, and casting head and velocity profiles in the
small gap lengths L. In particular, a knife gap of the casting head are therefore much
edge-shaped gap clearance (i.e., a very low more difficult to analyze (Ring, 1989).
value of L) results to be very unsuitable for
7.4.1.2 Tape Coating Methods
tape casting. In Fig. 7-40, for such a blade
with L equal to 1 mm, the dependence of The principle of this method defined in
the parenthetical term of Eq. (7-7) is plot- Fig. 7-33 b earlier in this section, is charac-
ted against the gap height, with the casting
velocity as a parameter.
From this diagram it can be deduced vo = lem/s /
that the value for the term is very close to
1, and hence, will have only a small effect
on the tape thickness Z), for gap heights / vo = 5 c m / s X
/
below approximately 200 jum and casting
velocities higher than about 5 cm/s. But - 10 cm/s

for larger gap clearances h, and for small

- ^
casting velocities v0, the value of the term
will become noticeable or even decisive, i

and the height of the slurry level (which is 0.0 0.5 1.0
h [mm]
proportional to Ap) will exert a strong in-
fluence on the tape thickness. This level Figure 7-40. Dependence of normalized tape thick-
ness [D = 2/(och)] on gap height and velocity (A12O3
should then be kept constant during the
slurry, ^slurry = 2g/cm 3 , ^slurry = 1000mPa s) for a
whole casting cycle if a constant tape casting head with a knife-edge-shaped gap clearance
thickness is desired. Otsuka et al. (1986 b) (L = 1 mm, Ap = 800 Pa, corresponding to about 4 cm
have shown that a dual blade casting head height of the slurry level).
248 7 Tape Casting

terized by the adherence of a certain than about 80 jam, while the yielded tape
amount of slurry to a carrier tape while it area per unit time is large due to relative
passes through the slurry reservoir. If no high casting velocities and the double-
special precautions are taken, the plastic sided covering of the support. Such facili-
tape support will be coated on both sides ties, therefore, are predominantly used for
after it leaves the suspension surface. Un- large production lines, for instance in
der stationary conditions, the thickness of multilayer capacitor technology, where ex-
the wet ceramic layer is proportional to the tremely thin tapes in the range of 20 }im
volumetric rate of slurry material dragged are typical.
along. Qualitatively, the influence of the most
This slurry volume per unit time is, how- significant casting parameters on the wet
ever, not readily calculable from the hy- layer thickness can be stated. Tape thick-
drodynamic equations. Viscous flow forces ness increases with
which generate the flux of material are
• increasing slurry viscosity
counteracted by gravity forces causing a
• increasing velocity of the support
certain reflow of material and resulting in
• decreasing specific weight of the slurry
undesirable conditions such as drop forma-
• decreasing wetting angle between slurry
tion. A third force, which determines the
and support
quantity of transported material, is due to
the surface tension of the slurry. The latter Most of the above delineated parame-
also contributes to a reflow of material. ters are inaccessible for use as thickness
The greatest difficulty for a theoretical control during a running casting process
treatment, however, arises from the fact since they depend on slurry properties. As
that the slurry parameters (such as viscosi- a means of an on-line thickness variation,
ty, density, or surface tension) will become only the velocity of the tape support can be
intricate unknown functions of the dis- used within certain limits.
tance from the reservoir surface as a conse- In general, a certain desired tape thick-
quence of progressive solvent evaporation. ness with small tolerances is adjusted in
In order to avoid an excessive reflow of two steps. A first rough determination of
material and drop formation, a fast in- the thickness range is made by selecting
crease of viscosity of the deposited slurry is slurry parameters such as viscosity and sol-
necessary. Again a pronounced pseudo- id content. For a fine-scale thickness ad-
plasticity is the adequate type of rheologi- justment, the casting velocity is disposable
cal behavior in this context, but at the and can even be used in a feedback control
same time a rapid expulsion of solvents system.
should likewise contribute to a rapid rate A special coating method (which can be
of tape consolidation. This latter effect, regarded as a variety of the double-side
however, must not be overrated since skin coating principle decribed above) provides
formation as a possible consequence an additional thickness-determining pa-
would be very deleterious with respect to rameter. If the supporting tape is not
the rate of tape drying and the formation drawn vertically but at a certain inclina-
of cracks in the course of this process (see tion with respect to the slurry surface, the
Sec. 7.4.2). upper side of the coated tape will be less
Hence, with this type of production, subjected to slurry reflow and, hence, will
tape thickness is limited to values smaller be thicker than layers of substrates pulled
 7.4 The Tape Casting Process 249

vertically. The effect is the more pro-

nounced the larger the deviation from the
vertical direction, i.e., the smaller the angle
between substrate and slurry surface.
Thus, this angle of inclination can be
used as a further parameter for the adjust-
ment of tape thickness. The coating on the
reverse side of the tape is exposed to un-
controlled slurry reflow accompanied by a
pronounced drop formation. The liquid r:
slurry is therefore stripped off from this
side by moving the tape across a high pre-
cision knife blade, immediately after it has Figure 7-42. Continuously working type-coating ma-
chine, a, slurry reservoir; b, knife blade; c, one-side-
left the slurry surface (Fig. 7-41). coated carrier tape; d, heated drying wheel; e, coiling-
A coating unit of this type, including a up spool for the dried tape.
drying wheel, is shown in Fig. 7-42. The
system is employable for tapes with thick-
ness ranging from about 30 \xm to 150 jam. Vibrations of the supporting tape can pro-
Typical thickness fluctuations or flaws duce similar patterns, but the intervals be-
can occur in the course of this coating pro- tween the stripes would be much smaller.
cess. The cause can often be deduced from A special type of thickness fluctuation
the visible traces and patterns in the tape, can be still more deleterious. Solid com-
at least when thin tapes are concerned. In- pacts, comparable in size with the tape
cident or transmitted light views or prints thickness dimensions, can be entrapped in
will clearly reveal these irregularities as the roll of slurry which develops in the
shown in Fig. 7-43. interfacial angle between the slurry surface
Stripe patterns perpendicular to the di- and substrate tape. By hindering the trans-
rection of the tape movement are normally port of slurry in the meniscus of the wet
caused by irregularities of the tape veloci- layer surface, these particles will cause
ty, especially with tape coating machines. stripe-shaped structures in the direction of
the tape's movement. These flaws may
have the form of periodically recurring
thin, short lines if the particle remains in
the slurry roll for a certain period, or they
may manifest themselves as a single line if
blade the particle is dragged along by the slurry.
In this latter case, the line shape is an out-
angle of come of disturbances in slurry material re-
inclination flow, as discussed earlier in this section.
slurry
With both coating and casting units a
continuous line pattern occurs if the par-
ticle is too large to be dragged along, or if
Figure 7-41. Principle of one-sided tape coating unit.
The slurry is stripped off from one side of the support it is kept in a fixed position, for instance, at
by a knife-edge-shaped blade. The tape thickness de- the knife edge of a blade or at the edge of
creases with increasing angle of inclination. a gap clearance.
250 7 Tape Casting

pin

cm

Figure 7-43. Typical incident (a-c) and transmitted (d-f) light views of green tapes showing specific irregularities
and flaws (the arrows indicate the direction of the tape movement), (a) Pattern caused by enhanced material
reflow in coating machines (very thick tapes), (b) Surface structure of a tape containing incompatible organic
components. Similar structures can occur with tapes from extremely diluted slurries, (c) Tape surface showing
large agglomerates in a matrix of small-sized powder particles [see also (a)], (d) Stripe patterns caused by tape
 7.4 The Tape Casting Process 251

In order to avoid such flaws, fine- liquid medium, is an indispensable process

meshed sieving of the slurry (10-30 jim) is step prior to final consolidation by firing.
indispensable prior to casting. Moreover, It is not astonishing, therefore, that in the
the slurry surface has to be protected from course of the very long history of ceramics
drying up by encapsulating of the reser- there has always been a considerable inter-
voir, and if necessary, by depositing an est in this important production step, as
additional source of solvent vapor in the regards both the physics of the process and
reservoir capsule. its practical performance as part of an ac-
Thickness fluctuations of a more "long tual production line.
wave" type, i.e., slow changes of the mean This section will only deal with some
tape thickness in the course of the casting selected problems in the context of ceramic
or coating process, are generally due to tapes, and will briefly present the basic
changes in slurry viscosity caused either by phenomena relevant to the process of dry-
solvent evaporation or by variations in ing in general. From the large number of
temperature. Especially with organic- publications dealing with this topic, only a
based slurry compositions, a thermostati- few books, review articles, and special
cally stabilized, covered casting head publications shall be cited here (Descamps
would be highly suitable for the produc- etal., 1994; Brinker and Scherer, 1990b;
tion of low tolerance tapes. Mujumdar, 1980,1983; Scherer, 1990a, b).
With an optimum selection of parame- During the process of tape drying, two
ters concerning slurry formulation, the successive stages can be distinguished, dif-
tape casting process, and the drying proce- fering in the rate of solvent evaporation. A
dure, the described sources of thickness first stage with constant rate is followed by
fluctuations can be avoided and overall a second stage with falling evaporation
tape thickness tolerances in the range of rate. In Fig. 7-44, this behavior is depicted
1-2 jim are possible. This can only be true, schematically. The weight of a certain wet
however, for tapes made from fine sized tape volume is plotted versus the drying
submicrometer ceramic powders with very time. The beginning of the second period is
smooth surfaces, since the thickness of a indicated by the deviation of the weight
single surface layer of particles theoretical- dependence from proportionally at the so-
ly represents the minimum value of tape called critical point.
thickness tolerance. Immediately after a wet ceramic tape
leaves the gap of the casting head, respec-
tively the slurry level in the casting reser-
7.4.2 Drying of Ceramic Tapes
voir, solvents begin to evaporate from its
With all ceramic wet forming technolo- exposed surface. This effect is accompa-
gies, drying, i.e., the evaporation of the nied by a shrinkage of the slurry layer in

vibrations (top) and by irregularities of the casting velocity (bottom). In the lower part some "pinholes", arising
from small air bubbles, are also visible, (e) Top: Track of a short-range flaw caused by a temporarily effec-
tive foreign body in the gap clearance of a casting head. Bottom: Continuous track of a long-distance flaw
stemming from a foreign body entrapped and stabilized in a fixed position of the gap clearance, (f) Short-range
flows of the type shown in the upper part occur if larger air bubbles are temporarily entrapped in the slurry roll
of a coating machine. Pinholes from very small bubbles are likewise visible. Bottom: The surface of a very homo-
geneous, flaw-free tape.
252 7 Tape Casting

stance, evaporation of liquid at the layer

constant
surface is increased excessively by a strong
drying rate stream of air and/or by overheating it, the
flux of liquids from the interior to the sur-
face may become insufficient and the liq-
uid/vapor interface will recede into the
tape's body.
Because of the plane geometry of a tape
critical point time
this effect can readily occur. It is character-
Figure 7-44. Time dependence of the weight of a dry-
ized by a change of the surface from shiny
ing tape. Initially, the weight loss, i.e., the rate of
evaporation remains constant. After the "critical and reflecting to a tarnished appearance
point" is achieved, a second drying period follows, due to the formation of an undesirable, dry
characterized by a decreasing rate of evaporation. skin (Shanefield, 1986) and a decrease of
the evaporation rate due to this barrier.
The consequences are more detrimental
thickness direction according to the loss of with thick tapes, where the total dwell time
liquid volume. This is associated with a in this first drying stage is much larger and
corresponding approach of the ceramic gradients of tape properties are more pro-
particles in the layer. The first stage of dry- nounced. The use of small quantities of
ing continues until contact between the non-volatile flow control agents such as
solid particles makes their further ap- liquid polyethylene can be appropriate to
proach more difficult. Three consecutive reduce the risk of skin formation (Brewer,
physical mechanisms govern this process, 1990).
the slowest of which determines the rate of In order to provide optimum drying
drying (Mistier et al., 1978): conditions, tape casting units are often
supplied with a closed drying tunnel,
1. The flow of liquid from the inner part of
where the vapor is swept away from the
the slurry to the surface.
tape surface by a moderate counterflow of
2. The evaporation of the solvent at the
slightly warmed air. The air enters the tun-
surface.
nel at a distance from the casting head
3. The transportation of vapor from the
where the tape has just reached a leather-
region near the surface.
hard consistency. During its passage oppo-
During the whole first stage of drying, site to the direction of the tape movement,
the so-called constant rate period, the sur- the air is consecutively enriched with the
face of the wet layer remains approximate- solvent vapor. Near the casting head,
ly in the same condition, i.e., the surface is where the flowing air leaves the drying tun-
always covered with a liquid layer fed by nel, it is nearly saturated with the solvent,
solvent flow from the interior of the tape. which retards its evaporation and prevents
Thus, the rate of evaporation remains con- skin formation. Simultaneously, the tem-
stant. The liquid/vapor interface coincides perature decrease due to solvent evapora-
with the surface of the layer, and the rate tion at the tape surface is compensated by
of evaporation is close to that of a plane, the application of moderately heated air.
exposed surface of bulk liquid. A second stage with a falling rate of sol-
This is only true, however, as long as a vent evaporation follows the initial con-
stationary state is maintained. If, for in- stant rate period. As the meniscus of the
 7.4 The Tape Casting Process 253

solvents recedes into the tape body, the that such tapes, having already endured
process of evaporation is displaced more extremely high compressive forces during
and more from the surface into the interior their fabrication, can hardly be com-
of the tape. Since now the slowest process pressed further during lamination.
steps are the diffusion of solvent vapor On the other hand, if the tapes were
from the receding front of liquid to the dried from less stabilized slurries contain-
surface, and the transportation of heat to ing soft agglomerates, the pore radii will be
this front (which is necessary to provide much larger, the capillary tensions will be
the heat of vaporization), the drying rate much less, and the more porous green
drops down gradually due to the increas- tapes will be more compressible during the
ing distance between the front and the sur- conditions of laminate formation.
face (Ford, 1986). At this point in the discussion, it should
At this critical point of drying, where the be pointed out that the tape drying process
vapor/liquid interface moves into the tape, normally is not governed by the laws of
a tendency for crack formation can be not- particle sedimentation, with larger parti-
ed. This is most likely to occur in thick cles settling first and smaller particles last.
tapes and at high drying rates. The effect This condition should even be avoided
can be attributed to local stresses caused carefully lest the tapes form inhomoge-
by differences of capillary pressures be- neous, textured layers. Sedimentation ef-
tween larger and smaller pore channels fects during drying are promoted by the
built up by the randomly packed ceramic following tape casting conditions:
particles (Scherer, 1990 b).
non-stabilized slurries
If again the pores are idealized as cylin-
large particle size distributions
drical tubes with a radius r, a capillary
low slurry viscosity
pressure, /? cap , is produced according to re-
thick tapes
lation (7-4) in Sec. 7.3.2.3. For contact an-
slow solvent evaporation
gles d near zero (i.e., for excellent wetting
conditions) capillary pressures are com- The last two conditions may be present,
pressive, and for small capillary pore sizes for instance, if thick tapes are produced
with r in the range of the particle diameters from water-based slurries which have a
(i.e., some tenths of a micrometer), they rather high latent heat of evaporation. The
can become rather high (200 MPa and rate of evaporation can of course be in-
even more). With these high compressive creased by the use of a strong, heated
forces, the leather-hard tape shrinks fur- stream of air. But the applicability of this
ther owing to displacements and reorienta- measure is limited by the risk of skin for-
tion of particles and with the aid of binder mation described above.
plasticity. Detrimental sedimentation effects (espe-
If the drying process is continued, such cially in thick tapes) can only be excluded
tapes will result in a green body structure, by using well stabilized slurries and partic-
with the ceramic forming a matrix of ularly by using pseudoplastic slurries with
densely-packed incompressible single par- high viscosity (rj«10 000 mPa s). With thin
ticles. Each such particle is covered by a tapes, especially if they are based on or-
thin plastic binder layer, and the remaining ganic solvents, these problems are not like-
interstices are filled with some residual ly to occur, since the overall drying times
portions of solvent vapors. It is evident are incomparably shorter.
254 7 Tape Casting

7.4.3 Characterization of Green Tapes Visual or light microscopical inspection

will be adequate for the detection of the
The characterization of green ceramic described flaws as well as of line flaws and
bodies in general is treated in Chap. 10 of thickness fluctuations as discussed in
this volume, to which reference shall be Sec. 7.4.1.2 of this chapter (visible as fluctu-
made at this point. With green ceramic ations of translucency). Electron microsco-
tapes, however, some special characteris- py, however, allows a much more informa-
tics are of interest, and their determination tive look at structural characteristics such
may involve particular problems which as the size and arrangement of particles,
will be briefly discussed subsequently. binder distribution, degree of agglomera-
Tape properties can be assessed both tion, and pore sizes. Exemplary SEM prints
qualitatively by visual inspection and of some green ceramic tape surfaces are re-
quantitatively by mechanical, optical, and produced in Fig. 7-45, with the composi-
thermal measurements. Some of the most tional differences explained in the caption.
important characteristics concern Tensile strength and flexibility of the
tapes are properties of only temporary in-
surface conditions
terest in the course of green tape process-
tensile strength
ing. They permit a destruction-free hand-
flexibility
ling especially of the thin tapes employed
thermal plasticity
in multilayer technology. Their measure-
structural characteristics (vp, vh, vg)
ment follows well-established methods
dimensional stability
and, hence, should not be treated here.
binder burn-out behavior
A property of specific interest for multi-
An initial visual inspection of the cast layer technology is a tape's ability to form
tape is informative and necessary, since laminates under the influence of pressure
compositional and operational shortcom- and temperature. This process was de-
ings will often become obvious because of scribed in Sec. 7.2.2. Here it should only be
their typical effects on the tape appear- pointed out that testing the suitability of a
ance. A rough tape surface indicates a certain tape for multilayer technology is a
highly agglomerated state of the powder in useful and necessary step of characteriza-
the slurry. Typical circular so-called pin- tion.
holes may be caused by the water content It can be carried out by forming small
in hydrophobic dispersions or insufficient test stacks at various temperature and
deairing before casting. pressure conditions. The success can then
The occurrence of cracks in thick tapes be evaluated by looking at the borders be-
or bubbles in thin tapes is frequently the tween the individual tapes in a cross-sec-
consequence of too rigorous drying condi- tional microscopic view. A well laminated
tions. Dust particles could be entrapped by stack is characterized by the absence of
the wet tape surface, causing lumps, since any distinguishable interface between the
the liquid generally endeavors to wet them individual tapes. As has been pointed out
and the slurry is therefore pulled up. previously, an excellent thermoplasticity
Hence, clean room conditions would be of the binder and a tape microstructure,
very appropriate for tape casting lines. As which allows a certain amount of compres-
a minimal precaution, filtered air should sion during lamination are indispensable
be used during tape drying. properties for this purpose.
 7.4 The Tape Casting Process 255

Figure 7-45. SEM micrographs of green

ceramic PZT tapes with different produc-
tion parameters, showing their influence
on the tape structure, (a) Binder B7, 3 h
ball mill mixing, t>p = 43%, t?b = 39%,
ug = 18%. Large particle agglomerates
form an inhomogeneous structure with
large voids and low packing of particles,
(b) Binder B7, high energy milled,
t>p = 51%, ^, = 35%, ug = 14%. The ag-
glomerates are largely destroyed, the
structure becomes more homogeneous
with smaller voids and higher packing of
particles, (c) Acrylic binder dispersion,
high energy milled powder, t?p = 46%,
yb = 31°/0) yg = 23%. Homogeneous struc-
ture, lightly agglomerates particles,
(d) Binder Butvar B76, high energy
milled powder, t?p = 44%, i?b = 30%,
ug = 26%. Homogeneous structure, lightly
agglomerated particles, (e) Binder PM685,
high energy milled powder, t;p = 58%,
ub = 27%, i;g = 15%. Very homogeneous
structure, high packing of powder parti-
cles, small voids, (f) Green tapes of (e)
densified by pressure and temperature,
vp = 63%, Db = 29%, t?g = 8%. Extremely
homogeneous structure with low gas vol-
ume and very high packing of powder
particles.
e)
10 pni 10 p.m

The tape volume fractions of powder corresponding values of their specific den-
(vp), binder (vh), and gas (vg) are a basic sities, £p and £b. The gas volume is the
structural characteristic property in this difference between the original tape vol-
context. The powder and binder portions ume and the sum of binder and powder
in a green sheet with a well-defined tape volumes:
volume V, (e.g., available from measure-
ment of thickness and area), are deter-
mined by heating the tape to approxi- vh=100Gh/(QhV)% (7-8)
mately 500 to 700 °C in oxygen or air, and vo = 100-vn-vh%
measuring its mass after binder burn-out,
Gp, and the associated mass loss, Gh. The The density of the ceramic powder crys-
percentage by volume of ceramic material, tals normally equals the X-ray density of
vp, and dried binder film, vb9 can then eas- the special composition, while the density
ily be calculated from these masses and the of the binder film containing all non-vola-
256 7 Tape Casting

tile organic components has to be deter- stabilized or not. If, with the same powder
mined separately from a dried piece of this and the same binder composition, the
binder film. For most of the generally ap- binder content is increased successively, a
plied tape casting binder systems, the more or less pronounced plateau of con-
binder film density, gh has a value ranging stant packing fraction occurs. Its level de-
from 1.1 to 1.2g/cm3. pends on the actual binder composition
As long as thick tapes are characterized and, of course, on the degree of particle
(100 jLim and more), this method provides dispersion. The influence of the binder
values for the volume fractions with toler- type is a result of the fact that binders may
ances of about 1-2%. With thin tapes, act as dispersing agents (see Sec. 7.3.2).
however, their uncertainty becomes appre- At a certain binder content all voids be-
ciably higher. This is mainly due to the tween the particles are filled up (vg be-
increasing tolerances associated with tape comes zero). Any further addition of non-
volume values, or more precisely, with the volatile organic materials will result in an
values for the tape thickness. This latter increase of particle distances in the green
property can only be determined with an tape and, hence, in a reduction of the
uncertainty in the range of surface rough- powder packing fraction vp (Gardner and
ness, about 1 jim at best. For a 20 (im thick Nufer, 1974). In Fig. 7-46, these relations
tape this would just mean a tape volume are illustrated schematically and qualita-
tolerance of 5%. tively, using arbitrary units for the amount
In some rare, special cases, a direct vol- of binder and for vp.
ume measurement using the Archimedean Dimensional stability of the green tapes
buoyancy principle can produce more pre- means not only the tapes do not shrink in
cise results. With this method, however, it thickness and laterally over the long term.
is a necessary prerequisite that the used The effect of such shrinkage on further
liquid medium does not dissolve any of the processing of the tapes could be minimized
organic tape components and does not in most cases by freeing them from their
penetrate into the capillary tape pores. supports and storing them under suitable
This is only possible if the the pores are conditions for a sufficiently long period of
very small and the liquid exhibits pro- time. Much more deleterious would be the
nounced non-wetting behavior. For exam- occurrence of effects due to non-isotropic
ple, the use of higher aliphatic hydrocar- distribution of tape properties. If the dry-
bons such as hexane or octane would be ing rate of the wet tapes is made excessively
appropriate for most organic-based and high, the flow of the solvent from the inte-
water-based binder systems. Benzene de- rior to the surface of the tape and its re-
rivatives and waxes could likewise be ap- distribution may become unsatisfactory or
plicable. even interrupted. The higher capillary ten-
The ceramic powder packing fraction vp, sions in the drier tape regions near the sur-
qualitatively displays a typical dependence face may then cause thin tapes to curl up or
on the content of binder in the tape. At thicker ones to crack.
zero, or very low binder content (e.g., with Moreover, every process step which
dried powder layers from nearly binder- could result in a gradient of tape composi-
free slurries), the resulting powder density tion (such as concentration and size distri-
may either be very small or relatively high, bution of particles and pores or distribu-
depending on whether the slurries were tion of binder) may potentially cause these
 7.4 The Tape Casting Process 257

composition products, their diffusion to

the surface, and their evaporation from
'b C££S> it (Cima et al., 1989 b; Calvert and Cima,
1990). The whole process, however, takes
place at much higher temperatures (about
250 to 600 °C), and since organic materials
are concerned, the surrounding atmo-
sphere becomes a very important factor to
be taken into consideration.
The character of the ceramic powder de-
termines whether an oxidizing or a reduc-
ing environment is admissible or, more
0 volume fraction vb precisely, what oxygen partial pressure
Figure 7-46. Particle packing fraction vp of a dry, would optimize the current burn-out prob-
green ceramic tape as a function of its binder content lem (Kahn, 1986). With oxide ceramics,
vh. In layers from powder/solvent suspensions with- for instance, an oxidizing atmosphere
out any binder content, particle packing may be low would be appropriate, but precautions
(a) or relatively high (b), depending on the wetting have to be taken to prevent any uncon-
behavior of the liquid. If small quantities of binder are
present, the particles are bonded together, which is trolled exothermic reactions or even in-
sometimes accompanied by a densiflcation of the flammation of the organic exhausts. Espe-
layer through capillary forces. With increasing binder cially with large, heterogeneous compo-
content the interstitials between the particles are suc- nents such as multilayer laminates, this
cessively filled up, while the distance between the par-
would cause local stresses and, as a conse-
ticles (and correspondingly vp) remains approxi-
mately constant. After the voids have been filled up, quence, the destruction of the part. The
the particle distances increase with further binder ad- quantity of oxygen has to be restricted to
dition, and hence the powder packing density of the concentrations low enough to prevent
tape decreases (adapted from Gardner and Nufer, exothermic chain reactions and sometimes
1974).
even reducing conditions have to be ad-
mitted.
In some other cases, e.g., with lead-con-
effects (Karneko et al., 1988). The selective
taining ceramics, reducing conditions have
sedimentation of large and fine powder
to be avoided especially at higher tempera-
particles from very dilute, slowly drying
tures beyond about 500 °C. The burn-out
slurries, as mentioned above, is an example
procedure, therefore, may become an intri-
of the kinds of conditions that cause pore
cate process, necessitating simultaneous
size gradients in a vertical direction. Pro-
control of the gas flow rate, the composi-
nounced lateral gradients, such as thick-
tion of the gas stream and the rate of tem-
ness fluctuations, could cause warping of
perature increase. Unfavorably large ele-
the dried tapes due to non-isotropic drying
ments with long distances from the interior
conditions.
to the surface necessitate very slow slope
burn-out routes of two and more days'
7.4.4 Binder Burn-Out
duration.
This process is governed by kinetic laws With non-oxide ceramics such as A1N or
at least in part similar to those involved in Si 3 N 4 , binder burn-out has to be carried
drying ceramic slurries: generation of de- out in non-oxidizing atmospheres. The
258 7 Tape Casting

generation of carbon residues as a product duction of large quantities of flat, thin and
of binder decomposition, however, must very homogeneous ceramic sheets at a
be carefully avoided. This is only possible moderate cost as a starting product for a
if the binder decomposes to low molecular great variety of ceramic components. With
weight, evaporable subunits. Such 'unzip- respect to special requirements, tape char-
ping' degradation is known to take place, acteristics are tailorable by applying mate-
for instance, with polymethylmethacrylate rials with well-defined and controlled func-
binders which depolymerize, producing tional properties as well as highly adapted
gaseous methylmethacrylate monomers. processing routes.
Their rate of production and their diffu- In this chapter on tape casting, an at-
sion to the surface through residual poly- tempt was made to emphasize that the gen-
mer binder or green body porosity are tem- erally pursued goal of high ceramic pow-
perature-controlled effects. Thus, a rapid der packing in the green tape can only be
increase in temperature could produce recommended for monolayer applications.
large quantities of gaseous components Multilayer manufacturing with its need for
which could not be readily transported to a lamination process requires tapes of low-
the surface, especially if the body has not er green density, without however allowing
yet achieved a sufficiently high permeabili- inhomogeneities or hard agglomerates.
ty. Vapor bubbles are likely to occur with The increasing interest in low-cost ce-
such conditions, i.e., if the partial pressure ramic tapes with narrowly defined proper-
of the gases in some local internal regions ties has brought about intensified activities
rises above the pressure of the environ- in theoretical and experimental research
mental atmosphere. This condition for an and development. Their outcome is a
unsatisfactory burn-out process has to be steadily growing understanding of the
avoided by using very slow heating rates chemical and physical mechanisms gov-
with overall process times of several days erning the art of tape casting in general,
(Calvert and Cima, 1990). and the special effects, tasks, and interac-
A number of published theoretical (Ger- tions of the various participating compo-
man, 1987; Cima etal., 1989; Sohn and nents in particular.
Wall, 1990) and experimental (Dong and As a consequence, the present situation
Bowen, 1989; Masia et al., 1989; Verweij in tape casting is characterized by perma-
and Bruggink, 1990) treatments have dealt nent improvements in material properties
with the problems related to binder burn- and process technology. Simultaneously
out with ceramics in general (Ferrato some tendencies and trends for future ac-
et al., 1994) and with ceramic tapes in par- tivities come into view (Tormey et al.,
ticular (Kahn and Chase, 1992; McAn- 1984). Anticipating certain foreseeable in-
drew, 1992). Only a few of them could be novations, some of which are in fact al-
cited here. ready practicable, the following examples
exemplify a far greater number of imagin-
able developments.
7,5 Conclusions and Outlook Chemical routes for ceramic powder
synthesis, including hydrothermal tech-
Tape casting as a wet forming process niques, sol-gel processing of metal organic
for ceramics reveals its outstanding merits compounds, or liquid phase reactions will
whenever applications demand the pro- be able to provide pure, fine-sized powders
 7.5 Conclusions and Outlook 259

with adjustable size distributions at rela- and the surrounding medium. Concentrat-
tively low costs (Segal, 1994). The fabrica- ed and yet sufficiently fluid suspensions of
tion of devices with highly reliable and re- nanometer-sized powder particles should
producible properties will greatly benefit then be achievable (Aksay et al., 1989).
from these powders. In the context of gradient compounds,
Ceramic layers with embedded, pre- the technique of multiple tape casting may
fabricated ceramic fibers can be expected become an attractive variant of produc-
to provide non-isotropic characteristics tion. Mistier (1973) has proposed a triple-
and improved mechanical or electrical layer casting unit for this purpose which
properties parallel to the ceramic sheet can easily be extended to higher numbers
plane (Kelley and Amateau, 1990). Like- of layers. Its main advantage is due to the
wise, non-isotropic grain-oriented ceram- fact that no special lamination process is
ics could be introduced to a number of required for the formation of the multilay-
tape applications. The incorporation of er compound. Moreover, the thickness of
specially shaped and positioned voids in a the individual layers is not limited by the
ceramic matrix offers some potential for requirements related to their separate han-
control of mechanical and electrical prop- dling, and binder content can be mini-
erties such as non-isotropic compliance, mized for this same reason.
resonance frequencies, or coupling charac- Thermal decomposition of tape binders
teristics. involves intricate problems, which have
Future binder materials might consist of been described above. The binder burn-
liquid organic monomers, which may be out process and the problem of environ-
polymerized after casting by chemical or mental contamination by the temporary
physical reactions (Tormey etal., 1984; organic components, in general, will have
Landham etal. 1987; Yoshikawa etal., a decisive influence on future develop-
1990). Thus, special evaporable solvents ments in the art of tape casting. In spite of
could be omitted and no drying time the described restrictions with water-based
would have to be provided. Only marginal binder systems, clean-air regulations will
shrinkage during curing of the polymer exert increasing pressure to replace organic
phase would occur, which could make it solvents by water, both in the form of solu-
possible to attain a desired tape thickness tions and, above all, as water-based binder
with narrow tolerances. Furthermore, dispersions. It should be possible, more-
structuring of dried green tapes by using over, to reduce the overall content of organ-
photosensitive polymer binders might be ic components drastically, to approximately
attractive. 10% of the tape volume and even less.
Concerning dispersants, the use of syn- A very promising method for polymer
thetic additives instead of natural products expulsion could be reactive chemical pro-
has just proved their superiority. For one cessing instead of pyrolysis of the organic
thing they are not only far better defined components (Boch and Chartier, 1988;
compared, for instance, with the broadly Tormey et al., 1984). This surface-con-
used Menhaden fish oil, which consists of trolled reaction could bring about a better
up to 44 different fatty acids (Bohnlein- management of the organic residuals and
MauB et al., 1992). They are also easy to fewer problems with destruction of com-
reproduce and are suitable for adaptation ponents due to stresses during binder
to both the surface of the ceramic powder burn-out.
260 7 Tape Casting

7.6 Acknowledgements Blodgett, A. J., Jr. (1980), IEEE Transactions on Com-

ponents, Hybrids, and Manufacturing Technology,
CHMT 3, 634.
I wish to express my gratitude to the Blum, J. B., Cannon W. R. (1985), Mater. Res. Soc.
editor of this volume, Prof. R.J. Brook, Symp. Proc. 40, 11.
Boch, P., Chartier, T. (1988), in: Ceramic Develop-
and one of the editors-in-chief, Prof. R.W. ments, Vol. 34-38: Sorrel, C. C , Ben-Nissan, B.
Cahn, for their constant encouragement (Eds.), Aedermannsdorf, Switzerland: Trans Tech,
and extraordinary patience during the Materials Science Forum, pp. 813-819
Boch, P., Chartier, T, Huttepain, M. (1986), /. Am.
preparation of this chapter. Ceram. Soc. 69 (8), C191.
Acknowledgement is further made to Bohnlein-MauB, I, Sigmund, W., Wegner, G., Meyer,
the Siemens AG for giving me the opportu- W. H., HeBel, K, Seitz, K., Roosen, A. (1992), Adv.
Mater 4, 73.
nity to write this treatment. I would espe- Bonekamp, B. C , Van't Veen, W. H., Schoute, M. J.,
cially like to thank my superiors as well as Veringa, H. J. (1989), in: Euro-Ceramics I, Vol. 1:
all my colleagues for their continuous ap- De Wirth, G., Terpstra, R. A., Metselaar, R. (Eds.).
London: Elsevier Applied Science, pp. 1145-1149.
preciation and sympathy during the whole Braun, L. Morris, J. R., Jr., Cannon, W. R. (1985),
period of writing this article. Am. Ceram. Soc. Bull. 64, 727.
I am deeply indebted to my fellow devel- Brewer, J. A. (1990), Ceram. Trans. 15, 651.
Brinker, C. J., Scherer, G. W. (1990a), Sol-Gel Sci-
opers, Mrs. G. Preu for producing most of ence, Boston: Academic Press, Chap. 8, pp. 453-
the SEM micrographs, and Mr. A. Wolff 513.
for his aid and advice concerning the pro- Brinker, C. J., Scherer, G. W. (1990 b), Sol-Gel Sci-
ence. London: Academic Press.
duction of the figures and graphs. Brook, R. J. (1976), in: Treatise on Material Science
Finally I am very grateful to Mrs. B. and Technology, Vol. 9: Wang, F. F. Y. (Ed.). New
Kirsch, for her patience and her skillful York: Academic Press, pp. 331-364.
Brook, R.J. (1989), in: Ceramic Powder Processing
typing of the manuscript from my not Science: Hausner, H., Messing, G. L., Hirano, S.
easy-to-read handwritten draft. (Eds.). Koln, FRG: Deutsche Keramische Ge-
sellschaft, pp. 667-672.
Brown, G. L. (1956), /. Polym. Sci. 12, 423.
7.7 References Brown, A. J., Bultitude, I, Lawson, J. M., Winbow,
H. D., Witek, S. (1991), in: Engineered Materials
Abicht, H. P., Voltzke, D., Miiller, T. (1990), Z. Handbook, Vol. 4, Ceramics and Glasses: Schneider,
Chem. 30 (11), 385. S.J., Jr. (Ed.). Materials Park, OH: ASM Int.,
Aksay, I. A., Stangle, G. C , Sarikaya, M. (1989), in: pp. 43-51.
Ceramic Powder Processing Science: Hausner, H., Brunner, D. G. (1988), Sprechsaal 121(3), 181.
Messing, G. L., Hirano, S. (Eds.). Koln, FRG: Calvert, P., Cima, M. J. (1990), J. Am. Ceram. Soc. 73,
Deutsche Keramische Gesellschaft, pp. 463-470. 575.
Allan, T. (1981), Particle Size Measurement, 3rd ed. Calvert, P. D., Tormey, E. S., Pober, R. L. (1986 a),
London: Chapman and Hall,. Am. Ceram. Soc. Bull. 65, 669.
Arendt, R. H., Rosolowski, J. H. (1979), U. S. Patent Calvert, P. D., Lalanandham, R. R., Parish, M. V,
4 152 280. Fox, J., Lee, H., Pober, R. L., Tormey, E. S.,
Barringer, E. A., Bowen, H. K. (1982), /. Am. Ceram. Bowen, H. K. (1986 b), Mater. Res. Soc. Symp.
Soc. 65, C199. Proc. 73, 579.
Bast, S. (1990), Keram. Z. 42, 650. Cannon, W. R., Becker, R., Mikeska, K. R. (1989),
Belosi, A., Fiori, C , Vincenzini, P. (1980), La Chim. in: Advances in Ceramics, Vol. 26: Yan, M. F ,
Ind. (Milan) 62, 913. Niwa, K., O'Bryan, H. M., Young, W. S. (Eds.).
Bernhardt, C. (1990), Granulometrie, Leipzig: Westerville, OH: Am. Ceram. Soc, pp. 525-541.
Deutscher Verlag fur Grundstofflndustrie. Cannon, W. R., Morris, J. R., Mikeska, K. R. (1986),
Besson, J. L., Boch, P., Chartier, T. (1987), in: High in: Advances in Ceramics, Vol. 19: Blum, J. B.,
Tech Ceramics, Materials Science Monographs, Cannon, W. R. (Eds.). Westerville, OH: Am. Ce-
Vol. 38. Vincenzini, P. (Ed.). Amsterdam: Elsevier, ram. Soc, pp. 161-174.
pp. 633-642. Cannon, W. R., Danforth, S. C , Flint, J. H., Hag-
Biggers, J. V, Shrout, T. R., Schulze, W. A. (1979), gerty, J. S., Marra, R. A. (1982), J. Am. Ceram.
Am. Ceram. Soc. Bull. 58, 516. Soc. 65, 324.
 7.7 References 261

Cesarano, J., Ill, Aksay, I. A., Bleier, A. (1988), J. Feil, M. (1986), in: Hybridintegration: Reichl, H.
Am. Ceram. Soc. 71, 250. (Ed.). Heidelberg: Hiithig, pp. 297-313.
Chane-Ching, J. Y, Paraud, N., Bortzmeyer, D., Ferrato, M., Chartier, T, Baumard, J. F., Coudamy,
Abouaf, M. (1989), in: Ceramic Powder Processing G. (1994), CFI, Ceram. Forum Int. 71 (1/2), 8.
Science: Hausner, H., Messing, G. L., Hirano, S. Fiori, C , De Portu, G. (1986), Brit. Ceram. Proc 38,
(Eds.). Koln, FRG: Deutsche Keramische Ge- 213.
sellschaft, pp. 547-553. Fisher, G., Schober, M. (1987), Am. Ceram. Soc. Bull.
Chartier, T., Boch, P. (1988), in: Science of Ceramics, 66, 1087.
Vol. 14: Taylor, D. (Ed.). Shelton, U. K.: The In- Flock, W M. (1978), in: Ceramic Processing Before
stitute of Ceramics, pp. 169-174. Firing: Onoda, G. Y, Jr., Hench, L. L. (Eds.). New
Chartier, T., Bruneau, A. (1993), /. Eur. Ceram. Soc. York: Wiley, Chap. 9, pp. 85-100.
12, 243. Ford, R. W. (1986), Ceramics Drying, New York:
Chartier, T., Streicher, E., Boch, P. (1987), Am. Ce- Pergamon Press.
ram. Soc. Bull. 66, 1653. Fukuura, I., Hirao, T. (1989), in: Advanced Technical
Chartier, X, Besson, J. L., Boch, P. (1988), Adv. Ce- Ceramics: Somiya, S. (ed.). Tokyo: (Transl.) Aca-
ram. 24B. Westerville, OH: Am. Ceram. Soc, pp. demic Press, pp. 65-81.
1131-1138. Gallagher P. K. (1991), in: Engineered Materials
Choi, G. R. (1971), /. Am. Ceram. Soc. 54, 34. Handbook, Vol. 4, Ceramics and Glasses: Schneider,
Chou, Y. T., Ko, Y T, Yan, M. F. (1987), J. Am. Ce- S. J., Jr. (Ed.). Materials Park, OH: ASM Int., pp.
ram. Soc. 70, C280. 52-64.
Ciftcioglu, M., Akinc, M., Burkhart, L. (1987), J. Am. Gardner, R. A., Nufer, R. W. (1974), Solid State
Ceram. Soc. 70, C329. Technol. 17 (May), 38.
Cima, M. J., Dudziak, M., Lewis, J. A. (1989a), J. Garino, T. I, Bowen, H. K. (1987), /. Am. Ceram.
Am. Ceram. Soc. 72, 1087. Soc. 70, C315.
Cima, M. X, Lewis, J. A., Devoe, A. D., (1989b), J. Gassner, W, Rossner, W, Tomandl, G. (1991), in:
Am. Ceram. Soc. 72, 1192. Ceramics Today - Tomorrow's Ceramics: Vin-
Cooper, A. (1984), in: Surfactants, Chap. 2: Tadros, cenzini, P. (Ed.). Amsterdam: Elsevier, pp. 951 —
T. F. (Ed.). London: Academic Press, pp. 19- 958.
52. Gattuso, T. R., Bowen, H. K. (1985), in: Advances in
Cooper, D., Newland, B. G., Shepley, F. W. (1987), Ceramics, Vol. 10: Kingery, W. D. (Ed.). Wester-
in: High Tech Ceramics: Vincenzini, P. (Ed.). Am- ville, OH: Am. Ceram. Soc, 644-655.
sterdam: Elsevier, pp. 1549-1558. German, R. M. (1987), Int. J. Powder Metall. 23, 237.
Costa, A. D. S., Traqueia, L. S. M., Labrincha, J. A. Goodman, G. (1986), in: Ceramic Materials for Elec-
(1993), in: Third Euro-Ceramics, Vol 1: Duran, P., tronics: Buchanan, R. C. (Ed.). New York: Marcel
Fernandez, J. F. (Eds.). San Vicente, Spain: Faenza Dekker, pp. 79-138.
Editrice Iberica, pp. 573-578. Goodwin, J. W (1990), Am. Ceram. Soc. Bull. 69,
Cutler, I. B. (1978), in: Ceramic Processing Before 1694.
Firing: Onoda, G. Y, Jr., Hench, L. L. (Eds.). New Graule, T, Gauckler, L. X (1993), in: Third Euro-
York: Wiley, Chap. 3, pp. 21-29. Ceramics, Vol. 1: Duran, P., Fernandez, J. F.
Davis, J. E., Carithers, V. G., Watson, D. R. (1971), (Eds.). San Vicente, Spain: Faenza Editrice Iber-
Am. Ceram. Soc. Bull. 50, 906. ica, pp. 491-500.
Dawkins, J. V, Guest, M. J., Taylor, G. (1982), in: Gurak, N. R., Josty, P. L., Thompson, R. J. (1987),
The Effect of Polymers on Dispersion Properties: Am. Ceram. Soc. Bull. 66, 1495.
Tadros, T. F. (Ed.). London: Academic Press, Hagemann, H. J., Henning, D., Wernicke, R. (1983/
Chap. 2, pp. 39-58. 84), Philips Tech. Rev. 41(3), 89.
Dayton, G. O., Schulze, W A., Shrout, T. R. Swartz, Haggerty, J. S. (1991), in: Concise Encyclopedia of
S., Biggers, J. V. (1984), in: Advances in Ceramics, Advanced Ceramic Materials: Brook, R. J. (Ed.).
Vol. 9. Mangels, J. A., Messing, G. L. (Eds.). Oxford: Pergamon Press, pp. 165-166.
Columbus, OH: Am. Ceram. Soc, pp. 115-139. Halloran, J. W. (1984), in: Advances in Ceramics,
Deryaguin, B. V, Landau, L. D. (1941), Ada Physi- Vol.9: Mangels, J. A., Messing, G. L. (Eds.).
cochim. 14, 633. Columbus, OH: Am. Ceram. Soc, pp. 67-75.
Descamps, M., Moreau, G., Mascart, M., Thierry, B. Halloran, J. W (1991), in: Engineered Materials
(1994), J. Eur. Ceram. Soc. 13, 221. Handbook, Vol. 4, Ceramics and Glasses: Schneider,
Di Marcello, F. W, Key, P. L., Williams, J. C. (1972), S. I , Jr. (Ed.), Materials Park, OH: ASM Int., pp.
/. Am. Ceram. Soc. 55, 509. 109-114.
Dong, C , Bowen, H. K. (1989), J. Am. Ceram. Soc. Hampton, J. H. D., Savage, S. B., Drew, R. A. L.
72, 1082. (1988), J. Am. Ceram. Soc. 71, 1040.
Ettre, K., Castles, G. R. (1972), Am. Ceram. Soc. Heinrich, J., Huber, X, Forster, S., Quell, P. (1987a),
Bull. 51, 483. Ind. Ceram. 7, 34.
262 7 Tape Casting

Heinrich, X, Huber, X, Schelter, H., Ganz, R., Golly, Kulig, M., Preu, G., Cramer, D., Lubitz, K. (1995),
R., Forster, S., Quell, P. (1987b), Brit. Ceram. Proc. 8th CIMTEC, Florence, in press.
Trans. J. 86 (6), 178. Landham, R. R., Nahass, P., Leung, D. K., Un-
Hellebrand, H., Cramer, D., Probst, L, Wolff, A., Lu- gureit, M., Rhine, W. E., Bowen, H. K., Calvert,
bitz, K. (1994), in: Proc. 4th Int. Conf. on New P. D. (1987), Am. Ceram. Soc. Bull. 66, 1513.
Actuators'. Borgmann, H., Lenz, K. (Eds.). Bremen, Lange, F. F. (1984), J. Am. Ceram. Soc. 67, 83.
FRG: AXON Technology Consult, pp. 119-122. Langhals, H. (1982), Angew. Chem. Int. Ed. Engl. 21,
Henson, W. A., Taber, D. A., Bradford, E. B. (1953), 724 [Angew. Chem. 94, 739].
Ind. Eng. Chem. Res. 45, 735. Lehmann, X, RieB, W, Ziegler, G. (1993), in: Third
Hofer, M. (1990), Dissertation, Technische Univer- Euro-Ceramics, Vol. 1: Duran, P., Fernandez, X F.
sitat Miinchen. (Eds.). San Vicente, Spain: Faenza Editrice Iber-
Holmes, M. H., Newnham, R. E., Cross, L. E. ica, pp. 485-489.
(1979), Am. Ceram. Soc. Bull. 58, 872. Leschonski, K. (1989), in: Ceramic Powder Process-
Howard, K. E., Lakeman, C. D. E., Payne, D. A. ing Science: Hausner, H., Messing, G. L., Hirano,
(1990), J. Am. Ceram. Soc. 73, 2543. S. (Eds.). Koln, FRG: Deutsche Keramische
Howatt, G. N. (1952), U. S. Patent 2 582 993. Gesellschaft, pp. 521-534.
Howatt, G. N., Breckenridge, R. G., Brownlow, Lin, S.-H., Swartz, S. L., Schulze, W. A., Biggers, J. V
X M. (1947), /. Am. Ceram. Soc. 30, 237. (1983), /. Am. Ceram. Soc. 66, 881.
Humphreys, C. X (1991), in: Concise Encyclopedia of Liniger, E., Raj, R. (1987), /. Am. Ceram. Soc. 70,
Advanced Ceramic Materials: Brook, R. X (Ed.). 843.
Oxford: Pergamon Press, pp. 67-73. Liniger, E., Raj, R. (1988), J. Am. Ceram. Soc. 71,
Hyatt, E. P. (1986), Am. Ceram. Soc. Bull. 65, 637. C408.
Jaycock, M. X (1986), in: Dispersion of Powders in Lubitz, K. (1995), in: Piezokeramik: Ruschmeyer, K.
Liquids, 3rd ed.: Parfitt, G. D. (Ed.). London: (Ed.). Reiningen-Malmsheim, FRG: Expert Ver-
Elsevier Applied Science, Chap. 2, pp. 51-98. lag, pp. 59-80.
Johnson, D. W, Jr., Gallagher, P. K. (1978), in: Ce- Lubitz, K., Hellebrand, H. (1990), in: Proc 7th Int.
ramic Processing Before Firing: Onoda, G. Y., Jr., Symp. on Application of Ferroelectrics, ISAF 90:
Hench, L. L. (Eds.). New York: Wiley, Chap. 12, Krupanidhi, S. B., Kurtz, S. K. (Eds.). Piscataway,
pp. 125-139. NJ: IEEE, pp. 509-912.
Kahn, M. (1986), in: Advances in Ceramics, Vol. 19: Lubitz, K., Hellebrand, H., Cramer, D., Probst, I.
Blum, J. B., Cannon, W. R. (Eds.). Westerville, (1991), in: Euro-Ceramics II, Vol. 3: Ziegler, G.,
OH: Am. Ceram. Soc, pp. 185-188. Hausner, H. (Eds.). Koln, FRG: Deutsche
Kahn, M., Chase, M. (1992), J. Am. Ceram. Soc. 75,649. Keramische Gesellschaft, pp. 1955-1972.
Kahn, M., Burks, D. P., Burn, I., Schulze, W. A. Lynch, J. F., Gaibrois, A. M., Lands, W. F., Vernon,
(1988), in: Electronic Ceramics: Levinson, L. M. P. M., Jr. (1989), in: Advances in Ceramics, Vol.
(Ed.). New York: Marcel Dekker, pp. 191-274. 26: Yan, M. F., Niwa, K., O'Bryan, H. M.,Young,
Karas, A., Kumagai, T., Cannon, W. R. (1988), Adv. W. B. (Eds). Westerville, OH: Am. Ceram. Soc,
Ceram. Mater. 3, 374-377. pp. 577-583.
Karneko, N., Rhine, W. E., Bowen, H. K. (1988), Ce- Malghan, S. G. (1991), in: Engineered Materials
ram. Trans. 1A, 410. Handbook, Vol. 4, Ceramics and Glasses: Schneider,
Keizer, K., Burggraaf, A. J. (1988), in: Science of Ce- S. X, Jr., (Ed.). Materials Park, OH: ASM Int., pp.
ramics, Vol. 14: Taylor, D. (Ed.). Shelton, U. K.: 75-82.
Inst. of Ceramics, pp. 83-93. Malghan, S. G., Dragoo, A. L. (1991), in: Engineered
Kelley, M. A., Amateau, M. F. (1990), in: Metal and Materials Handbook, Vol. 4, Ceramics and Glasses:
Ceramic Matrix Composites, Baghat, R. B., Clauer, Schneider, S. X, Jr. (Ed.). Materials Park, OH:
A. H., Kumar, P., Ritter, A. M. (Eds.). Warrendale, ASM Int., pp. 65-74.
PA: Minerals, Metals & Materials Soc, pp. 23-30. Masia, S., Calvert, P. D., Rhine, W. E., Bowen, H. K.
Kimura, T., Holmes, M. H., Newnham, R. E. (1982), (1989), J. Mater. Sci. 24, 1907.
J. Am. Ceram. Soc. 65, 223. Matijevic, E. (1989a), in: Ceramic Powder Processing
Kingery, W. D. (1978), in: Ceramic Processing Before Science: Hausner, H., Messing, G. L., Hirano, S.
Firing: Onoda, G. Y, Jr., Hench, L. L. (Eds.). New (Eds.). Koln, FRG: Deutsche Keramische Gesell-
York: Wiley, Chap. 23, pp. 291-305. schaft, pp. 25-34.
Kitoako, T., Seki, Y (1988), /. Ceram. Soc. Jpn., Int. Matijevic, E. (1989b), in: Science of Sintering:
Ed. 96, 572. Uskokovic, D. P., Palmour, H., Ill, Spriggs, R. M.
Kittel, H. (1976a), Lehrbuch der Lacke und Beschich- (Eds.). New York: Plenum Press, pp. 101-115.
tungen, Bd. Ill, Berlin: Verlag W. A. Colomb, pp. Matsubara, T., Yamada, Y, Kase, X, Irisawa, N., Shi-
137-236. moyama, X, Mitsuhashi, M., Sasaki, M., Abe, H.,
Kittel, H. (1976b), Lehrbuch der Lacke und Beschich- Morimoto, T. (1989), in: Advanced Superconductors,
tungen, Bd. Ill, Berlin: Verlag W. A. Colomb, pp. Proc. 1st Int. Symp. Supercond.: Kitazawa, K.,
17-136. Ishiguro, T. (Eds.). Tokyo: Springer, pp. 313-317.
 7.7 References 263

McAndrew, T. P. (1992), U. S. Patent 5 089 070. Onoda, G. J., Jr. (1978), in: Ceramic Processing Be-
McGeary, R. K. (1961), J. Am. Ceram. Soc. 44, 513. fore Firing: Onoda, G. Y, Jr., Hench, L. L. (Eds.).
McHale, A. E. (1991), in: Engineered Materials New York: Wiley, Chap. 19, pp. 239-251.
Handbook, Vol. 4, Ceramics and Glasses: Schneider, Otsuka, K., Ohsawa, Y, Yamada, K. (1986a), Yogyo
S. J., Jr. (Ed.). Materials Park, OH: ASM Int., pp. Kyokai Shi 94, 351.
115-121. Otsuka, K., Kitamura, W, Ohsawa, Y, Sekibata, M.
McKinnon, R. I, Blum, J. B. (1984), in: Advances in (1986b), Yogyo Kyokai Shi 94, 1136.
Ceramics, Vol. 9: Mangels, J. A., Messing, G. L. Ottewill, R. H. (1984), in: Surfactants: Tadros, T. F.
(Eds.). Columbus, OH: Am. Ceram. Soc, pp. (Ed.). London: Academic Press, Chap. 1, pp. 1 -
158-163. 18.
McPheeters, C. C , Dees, D. W, Dorris, S. E., Picci- Parfitt, G. D. (1986a), in: Dispersion of Powders in
olo, J. J. (1988), Fuel Cell Seminar, Washington, Liquids, 3rd ed.: Parfitt, G. D. (Ed.). London:
DC: Courtesy Associates, pp. 29-32. Elsevier Appl. Sci. Publ., Chap. 1, pp. 1-50.
Mikeska, K. R., Cannon, W. R. (1984), in: Advances Parfitt, G. D. (Ed.) (1986b). Dispersion of Powders
in Ceramics, Vol. 9: Mangels, J. A., Messing, G. L. in Liquids, 3rd ed. London: Elsevier Applied Sci-
(Eds.). Columbus, OH: Am. Ceram. Soc, pp. ence.
164-183. Parfitt, G. D., Rochester, C. H. (Eds.) (1983), Ad-
Mikeska, K. R., Cannon, W. R. (1988), Colloids Surf. sorption from Solution at the Solid/Liquid Interface.
29, 305. London: Academic Press.
Mistier, R. E. (1973), Am. Ceram. Soc. Bull. 52, 850. Parish, M. V, Garcia, R. R., Bowen, H. K. (1985), /.
Mistier, R. E. (1990), Am. Ceram. Soc. Bull. 69,1022. Mater. Sci. 20, 996.
Mistier, R. E. (1991), in: Engineered Materials Hand- Patankar, A. N., Mandal, G. (1980), Trans. J. Br.
book, Vol. 4, Ceramics and Glasses: Schneider, S. J., Ceram. Soc. 79, 59.
Jr. (Ed.). Materials Park, OH: ASM Int., pp. 159- Paulus, M. (1984), in: Materials Science Research,
165. Vol. 17: Davis, R. R, Palmour, H., Ill, Porter,
Mistier, R. E., Morzenti, P. T., Shanefield, D. J. R. L. (Eds.). New York: Plenum Press, pp. 177-
(1974), Am. Ceram. Soc. Bull 53, 564. 191.
Mistier, R. E., Shanefield, D. I, Runk, R. B. (1978), Pepin, J. G., Borland, W., O'Callaghan, P., Young,
in: Ceramic Processing Before Firing, Chap. 30: R. J. S. (1989), J. Am. Ceram. Soc. 72, 2287.
Onoda, G. Y, Jr., Hench, L. L. (Eds.). New York: Pober, R. L., Barringer, E. A., Parish, M. V, Levoy,
Wiley, pp. 411-447. N., Bowen, H. K. (1984), in: Materials Science Re-
Mizuta,S., Parish, M. V, Bowen, H. K. (1984), Ce- search, Vol. 17: Davis, R. R, Palmour, H., Ill,
ram. Int. 10, 83. Porter, R. L. (Eds.). New York: Plenum Press, pp.
Moreno, R. (1992a), Am. Ceram. Soc. Bull. 71, 1647. 193-206.
Moreno, R. (1992 b), Am. Ceram. Soc. Bull. 71, 1521. Ponton, C. B. (1993), in: Third Euro-Ceramics, Vol. 1:
Morris, J. R., Cannon,W. R. (1986), Mater. Res. Soc. Duran, P., Fernandez, J. F. (Eds.). San Vicente,
Symp. Proc. 60, 135. Spain: Faenza Editrice Iberica, pp. 189-194.
Morrison, W. H., Jr. (1985), /. Coat. Technol. 57, 55. Prasad, V. C. S. (1982), Trans. Ind. Ceram. Soc. 41,
Mujumdar, A. S. (Ed.) (1980), Advances in Drying, 85.
Vol. 1. New York: Hemisphere Publishing. Prasad, V. C. S. (1988), Trans. Ind. Ceram. Soc 47, 25.
Mujumdar, A. S. (Ed.) (1983), Advances in Drying, Prasad, V C. S., Panchapakesan, K. (1982), Am. Ce-
Vol. 2. New York: Hemisphere Publishing. ram. Soc. Bull. 61, 1234.
Nagai, A., Kimura, Y (1989), in: Advanced Technical Rabin, B. H. (1990), J. Am. Ceram. Soc. 73, 2757.
Ceramics: Somiya, S. (Ed.). Tokyo: (Transl.) Aca- Rao, A. S. (1987), Ceram. Int. 13, 233.
demic Press, Chap. 3, pp. 27-64. Reichardt, C. (1979), Angew. Chem. Int. Ed. Engl. 18,
Nahass, P., Rhine, W. E., Pober, R. L., Bowen, H. K. 98 [Angew. Chem. 91, 119].
(1990), Ceram. Trans 15, 355. Rhines, R N. (1978), in: Ceramic Processing Before
Namensma, P. A., Van der Molen, S. B. (1989), Euro- Firing: Onoda, G. Y, Jr., Hench, L. L. (Eds.). New
Ceramics I, Vol. 2: De Wirth, G., Terpstra, R. A., York: Wiley, Chap. 25, pp. 321-341.
Metselaar, R. (Eds.). London: Elsevier Applied Rhodes, W. H. (1981), J. Am. Ceram. Soc. 64, 19.
Science, pp. 2131-2135. Richerson, D. W. (1982), Modern Ceramic Engineer-
Newnham, R. E., Skinner, D. P., Cross, L. E. (1978), ing, New York: Marcel Dekker, Vol. II, Chap. 5,
Mater. Res. Bull. 13, 525. pp. 149-177.
Nguyen Quang Minh (1991), CHEMTECH, Feb., Riedel, R. (1991), in: Concise Encyclopedia of Ad-
120. vanced Ceramic Materials: Brook, R. J. (Ed.).
Niesz, D. E., Bennett, R. B., Snyder, M. J. (1972), Oxford: Pergamon Press, pp. 299-303.
Am. Ceram Soc. Bull. 51, 611. Ring, T. A. (1989), in: Advances in Ceramics, Vol. 26:
Ohde, N., Utsumi, K., Ochi, A., Takahashi, S. (1988), Yan, M. R, Niwa, K., O'Bryan, H. M., Young,
Inst. Electron., Inf. Commun. Eng. Jpn. Tech. Rep. W. S. (Eds.). Westerville, OH: Am. Ceram. Soc,
52, 220. pp. 569-576.
264 7 Tape Casting

Roosen, A. (1988), Ceram. Trans. IB, 675. Somiya, S., Yoshimura, M., Kikugawa, S. (1984), in:
Roosen, A., Hessel, R, Fischer, H., Aldinger, R Materials Science Research, Vol. 17: Davis, R. R,
(1990), Ceram. Trans. 12, 451. Palmer, H., Ill, Porter, R. L. (Eds.). New York:
Runk, R. B., Andrejco, M. X (1975), Am. Ceram. Soc. Plenum Press, pp. 155-166.
Bull. 54, 199. Stetson, H. W. (1965), U. S. Patent 3 189 978.
Russel, W. B. (1987), Powder Technol. 51, 15. Stetson, H. W, Schwartz, B. (1961), Am. Ceram. Soc.
Sacks, M. D., Scheiffele, G. W. (1986), in: Advances Bull. 40, 584.
in Ceramics, Vol. 19: Blum, J. B., Cannon, W. R. Suzuki, M., Kondoh, S., Yanagisawa, E., Shimo-
; (Eds.). Westerville, OH: Am. Ceram. Soc, pp. yama, I , Irisawa, N., Morimoto, T. (1990), Jpn. J.
175-184. Appl. Phys. 29, L61.
Sawicka, B. D., Ellingson, W. A., McPheeters, C C. Tadros, T. F. (1982), in: The Effect of Polymers on
(1987), Ceram. Eng. Set Proc. 8, 525. Dispersion Properties: Tadros, T. F. (Ed.). London:
Schat, B. R. (1970), Proc. Br. Ceram. Soc. 18, 281. Academic Press, Chap. 1, pp. 1-38.
Scherer, G. W. (1990a), J. Am. Ceram. Soc. 73, 3. Tadros, T. R, (Ed.) (1984a), Surfactants. London:
Scherer, G. W. (1990 b), Ceram. Trans. 12, 561. Academic Press.
Scherer, G. W., Garino, T. J. (1985), J. Am. Ceram. Tadros, T. F. (1984b), in: Surfactants: Tadros, T. R
Soc. 68, 216. (Ed.). London: Academic Press, Chap. 9, pp. 197 —
SchieBl, M., Ivers-Tiffee, E., Wersing, W. (1991), in: 220.
Ceramics Today - Tomorrow's Ceramics, Part D: Tai, L.-W, Lessing, P. A. (1991), J. Am. Ceram. Soc.
Vincenzini, P. (Ed.). Amsterdam: Elsevier, pp. 74, 155.
2607-2614. Takahashi, S. (1986), Am. Ceram. Soc. Bull. 65,1156.
Schmidt, W. G. (1989), in: Ceramic Powder Process- Takaya, M., Mochizuki, Y, Yazuda, K. (1990), Euro-
ing Science: Hausner, H., Messing, G. L., Hirano, pean Patent EP0433176 A2.
S. (Eds.). Koln, FRG: Deutsche Keramische Talen, H. W. (1962), FATIPEC Congressbook V, Mi-
Gesellschaft, pp. 535-546. lano, pp. 137-161.
Schubert, H., Petzow, G. (1988), in: Science of Ce- Temperley, H. N. V, Trevena, D. H. (1978), Liquids
ramics, Vol. 14: Taylor, D. (Ed.). Shelton, UK: and Their Properties. Chichester: Ellis Horwood.
Institute of Ceramics, pp. 107-112. Thompson, J. J. (1963), Am. Ceram. Soc. Bull. 42,
Schwartz, B. (1988), in: Electronic Ceramics. Levinson, 480.
L. M. (Ed.). New York: Marcel Dekker, pp. 1 -44. Tormey, E. S. (1987), Surfactant Sci. Ser., Surfactants
Schwartz, B., Wilcox, D. L. (1967), Ceram. Age 83, Emerging Technol. 26, 85.
40. Tormey, E. S., Pober, R. L., Bowen, H. K., Calvert,
Segal, D. (1994), in: Ceramic Technology Interna- P. D. (1984), in: Advances in Ceramics, Vol.9:
tional 1994: Birkby, I. (Ed.). London: Sterling Mangels, J. A., Messing, G. L. (Eds.). Westerville,
Publications, pp. 34-36. OH: Am. Ceram. Soc. pp. 140-149.
Shanefield, D. J. (1986), in: Advances in Ceramics, Ueyama, T., Kaneko, N. (1987), in: High Tech
Vol. 19: Blum, J. B., Cannon, W. R. (Eds.). Wester- Ceramics: Vincenzini, P. (Ed.). Amsterdam: Else-
ville, OH: Am. Ceram. Soc, pp. 155-160. vier, pp. 1451-1458.
Shanefield, D. J., Mistier, R. E. (1971), West. Electr. Ueyama, T, Wada, H., Kaneko, N. (1988), J. Am.
Eng. 15, 26. Ceram. Soc. 71, CIA.
Shanefield, D. J., Mistier, R. E. (1974), Am. Ceram Utsumi, K. (1991), Am. Ceram. Soc. Bull. 70, 1050.
Soc. Bull. 53, 416. Verweij, H., Bruggink, W. H. M. (1990), J. Am.
Shanefield, D. J., Mistier, R. E. (1976), Am. Ceram. Ceram. Soc. 73, 226.
Soc. Bull. 55, 213. Verwey, E. J. W, Overbeek, J. T. G. (1948), Theory of
Shaw, D. J. (1975 a), Introduction to Colloid and Sur- Stability of Lyophobic Colloids. Amsterdam: Else-
face Chemistry, London: Butterworths. vier.
Shaw, D. J. (1975 b), Introduction to Colloid and Sur- Watanabe, H., Kimura, T. Yamaguchi, T. (1989),
face Chemistry: London: Butterworths, Chap. 9, /. Am. Ceram. Soc. 72, 289.
pp. 187-205. Watanabe, H., Kimura, T., Yamaguchi, T. (1991),
Sliva, P., Dayton, G. O., Cross, L. E., Scheetz, B. E. J. Am. Ceram. Soc. 74, 139.
(1989), in: Advances in Ceramics, Vol. 26: Yan, Wersing, W, Lubitz, K., Mohaupt, J. (1986a), Ferro-
M. P., Niwa, K., O'Bryan, H. M., Young, W. S. electries 68, 77.
(Eds.). Westerville, OH: Am. Ceram. Soc, pp. Wersing, W, Schnoller, M., Wahl, H. (1986b), Ferro-
583-596. electrics 68, 145.
Smith, A. L. (1986), in: Dispersion of Powders in Wersing, W, Wahl, H., Schnoller, M. (1988), Ferro-
Liquid, 3rd ed.: Parfitt, G. D. (Ed.). London: El- electrics 87, 271.
sevier Applied Science, Chap. 3, pp. 99-148. Williams, J. C. (1976), in: Treatise on Material Science
Sohn, H. Y, Wall, D. R. (1990), J. Am. Ceram. Soc. and Technology, Vol. 9: Wang, F. F. Y. (Ed.). New
73, 2953. York: Academic Press, pp. 331-364.
 7.7 References 265

Wood, B. C. (1991), in: Engineered Materials Hand- Buchanan, R. C. (Ed.) (1986), Ceramic Materials for
book, Vol. 4, Ceramics and Glasses: Schneider, S. X, Electronics. New York: Marcel Dekker.
Jr. (Ed.). Materials Park, OH: ASM Int., pp. 8 3 - Kingery, W. D., Bowen, H. K., Uhlmann, D. R.
89. (1976), Introduction to Ceramics, 2nd ed. New York:
Yamamoto, J. K., Kata, K., Shimada, Y. (1989), Wiley.
Mater. Lett. 8, 278. Onoda, G. Y Jr., Hench, L. L. (Eds.) (1978), Ceramic
Yeh, T.-S., Sacks, M.D. (1988), J. Am. Ceram. Soc. Processing Before Firing. New York: Wiley.
71, C484. Parfitt, G. D. (Ed.) (1986), Dispersion of Powders in
Yoshikawa, S., Haertling, C , Smith, D., Newnham, Liquids, 3rd ed. London: Elsevier Applied Science.
R. (1990), Ceram. Trans. 12, 553. Parfitt, G. D., Rochester, C. H. (Eds.) (1983), Adsorp-
Young, W. S. (1986), in: Ceramic Materials for Elec- tion from Solution at the Solid/Liquid Interface.
tronics: Buchanan, R. C. (Ed.). New York: Marcel London: Academic Pre^s.
Dekker, pp. 403-424. Shaw, D. J. (1975), Introduction to Colloid and Surface
Zhao, J., Harper, M. P. (1988), J. Am. Ceram. Soc. 71, Chemistry. London: Butterworths.
530. Tadros, T. F. (Ed.) (1982), The Effect of Polymers on
Ziegler, E. (1986), in: Hybridintegration: Reichl, H. Dispersion Properties. London: Academic Press.
(Ed.). Heidelberg: Hiithig. pp. 13-27. Tadros, T. F. (Ed.) (1986), Surfactants. London: Aca-
demic Press.
Taylor, G. W. (Ed.) (1986), Special Issue on Multilayer
General Reading and Other Ferroelectric Ceramic Composites. Ferro-
electrics, Vol.68 (1-4). New York: Gordon and
Allan, T. (1981), Particle Size Measurement, 3rd ed. Breach.
London: Chapman and Hall. Temperley, H. N. V., Trevena, D. H. (1978), Liquids
Brinker, C. J., Scherer, G. W. (1990), Sol-Gel Science. and Their Properties. Chichester: Ellis Horwood.
London: Academic Press.
8 Injection Moulding
Julian R. G. Evans

Department of Materials Technology, Brunei University, Uxbridge, Middlesex, U.K.

List of Symbols and Abbreviations 268

8.1 Introduction 270
8.2 Powder Characteristics 274
8.2.1 Agglomeration 274
8.2.2 Specific Surface Area 274
8.2.3 Particle Size 276
8.2.4 Particle Size Distribution 276
8.2.5 Anisotropy of Particle Shape 277
8.3 Mixing Operations 277
8.4 Characterization of Dispersion 279
8.5 Flow Properties of Injection Moulding Suspensions 280
8.6 Physical Properties of Suspensions 285
8.7 Solidification in the Cavity 287
8.7.1 Void Formation 289
8.7.2 Stresses in Mouldings 292
8.7.3 Measurement of Residual Stress 295
8.7.4 Methods of Prolonging Sprue Solidification Time 296
8.8 Removal of Organic Vehicle 298
8.8.1 The Effect of Reheating Mouldings 298
8.8.2 Extraction by Capillary Flow 298
8.8.3 Solvent Extraction 299
8.8.4 Pyrolytic Extraction 299
8.8.5 Shrinkage During Pyrolysis 303
8.9 E x a m p l e s o f O r g a n i c M a t e r i a l s U s e d in C e r a m i c Injection M o u l d i n g . . . . 304
8.10 State-of-the-Art Summary 304
8.11 Acknowledgements 306
8.12 References 306

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
268 8 Injection Moulding

List of Symbols and Abbreviations

a activity (dimensionless)
a0 particle separation distance [m]
am adsorption area per molecule [m2]
A Hamaker constant [J]
b constant [K" 1 ]
c critical defect size [m]
c1? c 2 WLF constants [dimensionless, K]
C constant in viscosity equation (dimensionless); specific heat [J kg" 1 K" 1 ]
d0 particle diameter [m]
D diffusion coefficient in the composite [m2 s" 1 ]
D2 diffusion coefficient in the continuous phase [m2 s" 1 ]
E elastic modulus [N m~ 2 ]; activation energy [J mol" 1 ]
/ shape factor (dimensionless)
F van der Waals force [N]
Fg gravitational force [N]
h remainder weight fraction ( 0 < / i < l ) ; chain end-to-end distance [m];
surface heat transfer coefficient [Wm~ 2 K" 1 ]
Ho distance between mill rolls [m]
AJJvap enthalpy of vaporization [J mol" x ]
i Clausius-Clapeyron constant [Pa]
k fraction of adsorbed layer thickness (0</c<l)
K sorption constant [m s" 1/2 ]
Ko first order rate constant [s - 1 ]
Kp permeability [m2]
/ bond length [m]; thickness of disc [m]
L capillary length [m]
m mass fraction (0 < m < 1)
M molecular weight [g mol" 1 ]
n number of links in polymer chain (dimensionless);
flow behaviour index (dimensionless)
N Avogadro's number (dimensionless)
P pressure
q specific rate of decomposition [s" 1 ]
Q rate of monomer generation [kg m " 3 s" 1 ]; volume flow rate [m3 s" 1 ]
r radius [m]
JR gas constant [J mol" 1 K'1]
S volumetric specific surface area [m" 1 ]
t time [s]
T temperature [K]
Tg glass transition temperature [K]
Tm melting temperature [K]
vl9v2,v surface velocities of rolls 1 and 2 and average velocity [m s" 1 ]
 List of Symbols and Abbreviations 269

V ceramic volume fraction (0< V< 1); packing fraction (0< V< 1);
absolute volume [m3]
Ve effective ceramic volume fraction (0 < Ve < 1)
Fmax ceramic volume fraction a t which relative viscosity a p p r o a c h e s infinity
(0<F m a x <l)
Vmax ceramic volume fraction a t which shrinkage is arrested (V£ax< F m a x )
Fv void volume fraction (0 < Vy < 1)
w weight of sample [g]
x reaction depth [m]
X weight fraction ( 0 < X < 1) reaction boundary
Y plate half thickness
Z heating rate [Ks" 1 ]

a thermal diffusivity [m2 s" 1 ]

P die swell ratio (dimensionless); volume thermal expansion [K" 1 ]
y shear rate [s" 1 ]
Fc specific interfacial energy between particles [J m " 2 ]
F measured fracture energy [ J m ~ 2 ]
rj viscosity of suspension [Pa s]
f/A apparent viscosity [Pa s]
rj0 viscosity of vehicle [Pa s]
rjr relative viscosity (dimensionless)
8 polymer a n d m o n o m e r volume fractions ( O < 0 < 1 )
x isothermal compressibility [ P a " 1 ]
X constant (dimensionless); thermal conductivity [ W m " 1 K " 1 ]
n internal pressure [Pa]
Q density [ k g m ~ 3 ]
G stress [Pa]
of shrinking core reaction modulus (dimensionless)
i characteristic time [s]; shear stress [Pa]
X polymer-solvent interaction parameter (dimensionless)
co volume constant in van der Waals equation [m 3 ]

CPVC critical powder volume concentration

270 8 Injection Moulding

8.1 Introduction moulding machines of the late nineteenth

century were based on die casting methods
Ceramic injection moulding is the most (Hyatt and Hyatt, 1872). One of the earli-
versatile member of a set of plastic forming est reports of ceramic injection moulding
operations in which ceramic particles are was published in 1949 claiming that it had
carried into their prefiring positions in an been in use since 1937, using polymer
organic vehicle. It compares with slip cast- moulding machines (Schwartzwalder,
ing and the related pressure-casting pro- 1949). Ceramic injection moulding has in-
cesses as a method of producing complexi- herited some of its problems from these
ty of form in ceramic manufacturing. It antecedents.
differs from these processes in that the ve- In all these solidification processes, fluid
hicle is not extracted through the walls of is introduced through an orifice into a cav-
the cavity but remains in interparticle ity which generally exceeds the section di-
space to be removed at a later stage. The mensions of the entry channels. A change
main advantages of injection moulding of state then occurs progressively from the
are: walls under unsteady state heat transfer
conditions. In ceramic and metal powder
1. Rapid automated mass production with injection moulding this change of state oc-
fine process control. curs by the solidification of the organic
2. Manufacture of complex shaped com- vehicle either because it passes its glass
ponents with dimensional accuracy. transition temperature during cooling or
Set against these advantages are: because it is a semicrystalline wax or poly-
mer which passes its melting point. It may
1. A high capital and tooling cost justified also be brought about by the crosslinking
only by large production runs. or polymerization of a low molecular
2. A limitation on component section size weight resin (reaction injection moulding)
controlled by the problems of non-uni- or by the thermo-gelling of a water soluble
formity of solidification and removal of polymer (Rivers, 1978). In contrast, the so-
organic matter before sintering. lidification in slip or pressure-casting oc-
Table 8-1 shows that injection moulding curs by raising the ceramic volume fraction
has its origin in the foundry (Tylecote, by extraction of vehicle until fluid proper-
1962). In the mid-nineteenth century the ties are lost. Figure 8-1 shows a range of
invention of die casting offered speed and sintered ceramic components, manufac-
scope for automation to the foundryman tured by injection moulding, and illus-
(Sturges, 1949). The first polymer injection trates the complexity of shape attainable.
Ceramic injection moulding should
therefore be set in the context of other
Table 8-1. Historical development of ceramic injec- manufacturing techniques which have
tion moulding. been drawn from the polymer processing
arena (Evans, 1993). In each case, a well-
Process Date
established method has been applied, mu-
Metal casting 3000 B.C. tatis mutandis, to ceramic processing.
Die casting 1849 A.D. Table 8-2 gives a summary of these tech-
Polymer injection moulding 1872 A.D.
Ceramic injection moulding
niques and it can be seen that the ceramic
1937 A.D.
designer now has a large armory of manu-
 8.1 Introduction 271

a. b.
Figure 8-1. Ceramic com-
ponents manufactured by
injection moulding, a) SiC
seal rings (up to 70 mm di-
ameter) b) Si 3 N 4 bucket
tappets; unsintered and sin-
tered (30 mm outer diame-
ter) c) SiC turbocharger
rotor on shaft (55 mm
diameter) d) marine turbo-
charger plain bearing
(125 mm diameter). Cour-
tesy of John Woodthorpe,
Principal Scientist, T&N
Technology, Rugby, U.K.
c.

facturing methods at his disposal (Edirs- shear or extensional flow. Figure 8-2 illus-
inghe and Evans, 1986; Cass, 1991; Shep- trates the occupation of space. Here Vmax is
pard, 1991; Alford et al., 1990; Wright the ceramic volume fraction at which the
et al., 1990a; Kobayashi et al., 1981; Ham- viscosity of the suspension approaches in-
mond and Evans, 1991; Tummala, 1988; finity as particles come into contact. It fol-
Haunton et al., 1990; Rashid et al., 1991 a, lows that the working range of ceramic
b; Rashid and Evans, 1991; Greener and volume fraction V must be less than VmaLX.
Evans, 1993). These processes all make use This gives rise to a free volume fraction
of the various features of macromolecules (Vmax~ V) which represents the fraction of
such as high adhesive strength, pseudo- space occupied by organic vehicle over and
plastic flow properties and extensional above that needed to fill the space between
melt strength. contacting particles at Vmax. This free vol-
In each case, organic matter fills the ume concept is extremely useful and it can
space between particles, ideally to the ex- be shown that it controls viscosity, shrink-
clusion of air, and is present in sufficient age during binder removal and the change
excess to allow the particles to rotate in of state from liquid to quasi-solid as or-
272 8 Injection Moulding

Table 8-2. Polymer processing techniques used for ery stage in the process, including isother-
ceramics manufacture. mal storage after moulding and after
Process References debinding. Strain in the particle-filled
composite implies particle displacement.
Injection moulding Edirisinghe and Evans (1986) Although the displacements may be small,
Extrusion Benbow and Bridgwater (1993) they may act as precursors for the subse-
Solvent casting this work Chap. 7 quent formation of defects. Particles
(tape casting)
should be regarded as being in permanent
Melt spinning Cass (1991); Sheppard (1991) motion, albeit with small displacements,
Winding of coils Alford et al. (1990);
Wright etal. (1990 a)
from the beginning of the process to the
Blow moulding Kobayashi et al. (1981); end. Although alarming, this is a useful
Hammond and Evans (1991) concept and Table 8-3 summarizes some of
Thermolamination Tummala (1988) the movements which are dealt with in
Vacuum forming Haunton etal. (1990) greater detail below.
Solvent welding Rashid etal. (1991b) The formal simplicity of injection
Hot plate welding Rasmid and Evans (1991) moulding machines has remained un-
Ultrasonic welding Rashid etal. (1991a) changed over the years, while process con-
Film blowing Greener and Evans (1993) trol has advanced so that speeds, delays
and pressure profiles can be finely adjusted
and reproduced. Mutsuddy (1989) reviews
mixing and moulding equipment available
for ceramics. The machine (Fig. 8-3) con-
1-V. tains a clamp which houses the mould tool
}Vmax-V
and whose purpose is to open, eject the
moulding and close again. The clamp may
open in the axial or vertical direction with
respect to the injection axis. In either case,
the clamp force is dictated by the maxi-
Figure 8-2. The occupation of space by ceramic of
volume fraction V and organic vehicle of volume frac-
mum injection pressure multiplied by the
tion (1 - V). (Vmax from viscosity by extrapolation; V£ax projected area of the moulding in the
from shrinkage by measurement, F ^ opening direction.
The other end of the machine supports a
moving carriage which conveys the barrel.
ganic vehicle is removed, i.e., as Vincreas-
es (Wright et al., 1990 b). V£ax on the other
hand represents the maximum volume sprue
fraction of ceramic after binder removal bush nozzle heaters hopper
and can be deduced from linear shrinkage
measurements. It is always less than Vmax. clamp
It would be oversimplification to the
point of error to suppose that the presin-
tering positions of particles are taken up
immediately after mould filling. Relative carnage
and often heterogeneous motion of parti- Figure 8-3. Schematic diagram of a reciprocating
cles can and usually does take place at ev- screw injection moulding machine.
 8.1 Introduction 273

Table 8-3. The causes of particle motion throughout the injection moulding process.

Stage Source of particle relative motion

Dispersive mixing breaking down of agglomerates.

Distributive mixing rearrangement of ultimate particles.
Moulding filling rotation in shear flow.
Solidification non-uniform thermal contraction in the liquid state;
non-uniform thermal contraction in the solid state;
strains associated with the partial release of orientation and residual stress
during cooling.
Storage after time dependent failure (static fatigue) under residual stresses;
moulding strain associated with stress relaxation and recovery of orientation.
Pyrolysis uniform thermal expansion under slow (steady state) heating;
release of residual stresses;
recovery of orientationi in macromolecules;
flocculation;
non-uniform change in ceramic volume fraction caused by non-uniform binder
removal;
overall shrinkage due to loss of binder from particle junctions;
uniform thermal contraction under slow (steady state) cooling.
Storage after expansion of dry ultrafine particles as they are non-uniformly hydrated
debinding by ambient air.
Sintering approach of particle centres, possibly non-uniformly.

Two types of barrel are available - recipro- that the material is uniformly heated in
cating screw and plunger. The latter suffers the barrel.
from the inherent disadvantage of inject- b)The load on the carriage is released to
ing non-uniformly heated material, al- avoid deformation of the stationary
though this disadvantage is overcome if mould half when the clamp is opened,
the plunger barrel is itself fed from a sepa- c) The clamp is opened and the part is
rate plasticizing device. ejected.
Starting from the position with the d)The clamp closes and the carriage moves
mould cavity full - a shot has just been forward, loaded to ensure a seal between
made - the sequence is as follows: nozzle and sprue bush.
e) The screw moves forward without rota-
a) The helical screw rotates, conveying ma-
tion, at controlled speeds to displace
terial forward and itself floating back
material from the barrel to the cavity.
against a plasticization counter pres-
The contents of the barrel are always set
sure. The nozzle is closed by the material
to exceed the shot size so that a 'cushion'
in the cavity or by a spring-operated
of material is available to fill the cavity
valve. The forward flow of material is
as shrinkage occurs for as long as the
dependent on the adhesion to the barrel
sprue remains molten.
wall because the single screw offers no
f) After sufficient time has elapsed for so-
positive displacement pumping action.
lidification and cooling, stage a) is re-
Rotation of material within and around
peated.
the helical grooves of the screw ensures
274 8 Injection Moulding

8.2 Powder Characteristics which shows a much greater dependence

on packing fraction and introduces a de-
8.2.1 Agglomeration pendence on critical flaw size c, within an
agglomerate.
It is well recognized that sintering pro- While the mixing stage must be able to
cesses require powders with a fine, narrow break down agglomerates, a burden is also
size distribution. Fineness confers a high placed on the powder manufacturer to
surface excess free energy for sintering and control agglomerate strength. For exam-
mass transport is enhanced by low diffu- ple, the manufacturers of pigments for
sion distances. Uniform particle size re- paint or printing ink applications are able
duces the driving force for grain growth. to produce easily dispersed fine powders in
However, the fabrication method also the 0.2 \xm ultimate particle size range
places restrictions on the powder and these (Kendall, 1988; Zhang etal., 1989 b).
frequently conflict with the demands of the Residual agglomerates are thought to
sintering stage. control defect size in the fired ceramic
Fine ceramic powders agglomerate nat- body either because of differential sinter-
urally under the influence of London dis- ing (Lange, 1983) or because pores with
persion forces and the force between high grain coordination number reside at
neighbouring particles of equal diameter agglomerate junctions and achieve stabili-
dQ is given by ty (Lange, 1984). Thus residual agglomer-
A d0 ate dimensions can give an approximate
F= (8-1) indication of critical flaw size. It follows
24a20
from the statistical nature of brittle failure
where a0 is the separation distance, taken that uniformity of agglomerate strength in
as about 2 lattice spacings if the particles a powder will influence uniformity of
are in contact. The Hamaker constant A, strength in the ceramic product, a small
can be calculated for a range of ceramics in proportion of strong agglomerates having
different media and is usually close to a disproportionate influence on strength.
10" 19 J (Israelachvili, 1991 b). This attrac-
tion accounts for the poor flow properties
of fine powders and the tendency to ag- 8.2.2 Specific Surface Area
glomerate and to flocculate in suspension
A ceramic fabrication process should be
(Song and Evans, 1994). Rumpf (1962) ex- capable of serving a wide range of powders
presses the tensile strength of a particle but significant problems emerge when at-
assembly of packing fraction V as tempts are made to injection mould ultra-
1.1 VA fine powders with volume specific surface
(82) areas in the 1 -2 x 108 m~ * range. A gener-
al finding is that maximum loadings are
so that for the conditions discussed above,
reduced. In a particle-filled polymer or
the agglomerate strength for 100 nm and
wax, the inorganic surface controls the
1 |im particles are 0.6 MPa and 0.06 MPa,
conformation and mobility of the adjacent
respectively, for volume fraction of 0.36.
organic molecules through adsorption
Kendall (1988) also gives an expression
(Stromberg, 1967). Figure 8-4 shows the
for the strength of agglomerates:
schematic structure of adsorbed polymer
J .A-1/2 / o ->\ layers on inorganic surfaces. Molecules
20 C) \°-3)
 8.2 Powder Characteristics 275

tail

train

Figure 8-4. Schematic arrangement

of adsorbed polymer molecules on a
crystalline lattice showing 'trains',
'loops' and 'tails'.

protrude from the surface as loops or bris- lar weight of polymers for plastic forming
tles with only a small fraction of segments must be judiciously selected in the context
in the adsorbed state (Patat etal., 1966). of particle size (Evans, 1990). This is re-
Similarly, only a fraction of adsorption flected in recent work which shows that,
sites are used (Taylor and Rutzler, 1958). with appropriate dispersants, it is possible
This means that adsorbed layer thickness to injection mould an ultrafine zirconia
considerably exceeds the dimensions of a powder (average particle size 70 nm) at
flat monomolecular layer. Studies of poly- over 60 vol. % in a wax-based system
mer adsorption are generally from solu- (Song and Evans, 1995).
tion but the situation in the melt is believed The end-to-end distance of a polymer
to be similar (Israelachvili, 1991a). chain h is calculated from the dimensions
Such adsorbed layers confer advantages of a random coil (Tanford, 1967)
and disadvantages in ceramic processing.
h« 3 (8-4)
The main advantage is that overlap results
in a repulsion between particles leading to where n is the number of links and / is the
stabilization of the dispersion known as bond length. The number of links is depen-
elastic stabilization. The conformational dent on the degree of polymerization and
freedom of the loops and tails is reduced the number of links contributed per
during compressive overlap, leading to a monomer unit. Thus h oc ^ M .
reduction in entropy and hence an increase Probably the most deleterious effects of
in free energy. As a result, a force of repul- fine powders emerge during the process of
sion develops between particles (Gregory, debinding. This is illustrated in a compari-
1978). In low molecular weight polymers son of a coarse alumina (S= 1.2 x 106 m~ *)
such as waxes, steric stabilization of sus- and a fine alumina of (5 = 37 x 106 m" 1 )
pensions using amphipathic molecules can (Evans and Edirisinghe, 1991). Cylinders
be achieved (Schofield, 1990). of the coarse powder could be pyrolysed
A number of studies of adsorption indi- successfully up to 25 mm diameter while
cate that the adsorbed layer thickness of the maximum diameter for the fine powder
polymers is comparable to the average was 8 mm.
end-to-end distance of the molecule Adsorption can be expected to reduce
(Stromberg etal., 1965; Priel and Silber- mass transport both as diffusion of poly-
berg, 1978). If this situation prevails in the mer degradation fragments in the melt and
melt, then 100 nm particles are compara- as fluid flow in the pore structure. Thus, in
ble in size to adsorbed layers and molecu- the case of diffusion, an effective volume
276 8 Injection Moulding

fraction of ceramic should be defined as (1989) observed a strong catalytic effect of

"inert" ceramic oxides on the decomposi-
+kSh) (8-5) tion of polyvinyl butyral in both oxidizing
where k is a constant which indicates the and non-oxidizing atmospheres.
fraction of adsorbed layer for which diffu-
sion coefficient is effectively zero. Thus 8.2.3 Particle Size
for the fine alumina (3.7 x 107 m" 1 ) at
For a given particle size distribution, the
61 vol.% in polypropylene, the effective
effect of particle size does not appear in
volume fraction V was 81 vol.%, taking
equations for relative viscosity in terms of
£ = 0.5 (Evans and Edirisinghe, 1991).
volume fraction of dispersed phase (Ediri-
A further disadvantage of high surface singhe and Evans, 1986). Nevertheless,
area powders is that they can hold consid- several studies of viscosity of dispersions
erable amounts of adsorbed water. The ad- reveal an inverse dependence on particle
sorption area per molecule of water is giv- size (Sherman, 1970; Parkinson etal.,
en by (Gregg and Sing, 1982) 1970; Sweeney and Geckler, 1954) which
can be explained in terms of the volume
(8-6) occupied by adsorbed layers. Landel et al.
\QLN
(1965) found that for a range of powders
where N is Avogadro's number, a n d / i s a curves of log rj/rj0 versus V/Vmax were coin-
shape factor (1.019) and M is molecular cident suggesting that particle size princi-
weight. For water, am = 10.6 x 10~ 20 m2 pally influenced Vmax. This is compatible
molecule"1. The weight fraction of water with the effect of adsorption which in-
adsorbed on a powder is then creases the effective volume fraction of ce-
ramic. German (1990 a) interprets the ef-
nMS fect in terms of interparticle friction result-
A= (8-7)
amN Q ing from the increased surface area.
Nogueira et al. (1992) prefer a particle size
where S is the volumetric specific surface, of just under 1 |im, near spherical shape
area, Q is the powder density and n is the and a surface area of 12 x 106 m~* for in-
number of monolayers. Thus a 35 m2/g jection moulding.
powder carries 1 wt.% water per monolay-
er. During pyrolysis, this is likely to
8.2.4 Particle Size Distribution
emerge in the 100-200 °C region while all
pores are blocked and must diffuse It is well-known that the intimate mixing
through the organic vehicle; a transport of particles which are multimodal in size
path which is partly dependent on polarity confers a packing advantage over mono-
of the organic vehicle. modal arrangements (Westman and
Surface area may also influence the ex- Hugill, 1930; Lewis and Goldman, 1966).
tent of catalysis of degradation of the or- A similar situation applies to the packing
ganic vehicle by the inorganic surface and of particles in a fluid. Thus for the same
hence influence the rate and reaction route viscosity, a higher volume fraction of ce-
for decomposition during pyrolysis. It is ramic can be accommodated if the size dis-
well-known that transition metal cations tribution is multimodal (Farris, 1968; Lee,
influence the oxidative degradation of 1969). Unfortunately grain growth and
polymers (Hansen, 1970) but Masia etal. sintering are necessarily concurrent in
 8.3 Mixing Operations 277

crystalline materials and the use of multi- and is restricted in the directions parallel to
modal powders enhances the driving force the platelets.
for grain growth. Thus, where final grain Considerable control of orientation in
size is critical for high strength, metastable moulding has been achieved for short-sta-
phase retention or ferroelectric properties, ple reinforced polymer (Allan and Bevis,
the strategy of attaining efficient multimo- 1987) and metal (Pinwill et al., 1992) com-
dal packing is to be treated with caution. posites. Shear-induced orientation is set up
during solidification in a multi-gated
mould cavity as a result of forced convec-
8.2.5 Anisotropy of Particle Shape
tive oscillatory flow in the mould during
The pattern of flow into the simplest test cooling (Fig. 8-5). For a fully aligned ar-
bar cavity is complex. Since the flow pat- rangement, the resulting shrinkage an-
tern influences the final orientation of any isotropy is more controllable.
anisotropic particles conveyed by the vehi-
cle, the orientation distribution is hetero-
geneous throughout the cavity. The conse-
quences of this may only appear at the 8.3 Mixing Operations
sintering stage: non-uniform shrinkage
and loss of shape. There is an abundant Two types of mixing are generally distin-
supply of powders with reasonably guished. Dispersive mixing involves the
equiaxed particles but the popular interest separation of particles and hence the de-
in the manufacture of fibre, whisker or struction of agglomerates either by erosion
platelet reinforced metals or ceramics in- or successive fracture. It is dependent on
troduces a major difficulty for injection the imposition of high shear stresses
moulding. (Edirisinghe and Evans, 1986). Distribu-
Such second phase reinforcements may tive mixing describes the spatial rearrange-
confer additional strength or toughness ment of particles by forced convection.
(Wei and Becker, 1986; Lange, 1973) and In the dispersion of powders in liquids
attempts have been made to injection three stages are recognized (Bell and
mould metal (Pinwill et al., 1992) and ce- Crowl, 1973). Wetting describes the re-
ramic (Kandori et al., 1987; Neil and placement of solid-vapour interfaces with
Noris, 1988) matrix composites. The diffi- solid-liquid interfaces and may affect ag-
culties are illustrated by the gross deforma- glomerate strength. Mechanical disruption
tion of test bars and turbine blades injec- involves the fracture of junctions until the
tion moulded from SiC-whisker-reinforced required degree of dispersion is obtained
Si 3 N 4 by conventional moulding tech- and stabilization indicates that conditions
niques (Neil and Noris, 1988). have been arranged to prevent floccula-
The anisotropic sintering shrinkage in a tion. It is questionable whether wetting
silicon carbide-silicon nitride composite precedes dispersion in polymer melts. The
has been detected during sintering by time required for intrusion of the medium
dilatometry. Non-uniform shrinkage be- may exceed the mixing period and agglom-
gins at an early stage and produces a dis- erates can be detected in which polymer is
torted shape (Stedman et al., 1993). The largely absent (Song and Evans, 1993 a)
rate of sintering is highest in the direction with important consequences for the cal-
perpendicular to the plane of the platelets culation of true volume fractions.
278 8 Injection Moulding

MOULD CAVITY

INJECTION MOULDER BARREL

Figure 8-5. Schematic diagram of

SHEARING OF DOUBLE LIVE-FEED
THE MELT
a double-gated multi-live-feed
DEVICE
device for control of fibre orienta-
tion in mouldings.

There is now general recognition that where Ho is measured with slip gauges and
high shear mixers are preferred for ceramic the surface speeds are recorded with a
injection moulding suspensions. There are tachometer, the shear rates experienced by
three main devices, each incorporating two material in the nip can be found and the
principal moving parts; the two-roll mill, shear stresses obtained from a knowledge
the twin-screw extruder and various dou- of viscosity as a function of temperature
ble axis lobed mixers (Fig. 8-6). Mixers and shear rate. This has been done for a
and their operation are thoroughly de- range of powders and the resulting disper-
scribed in a number of good reviews (Tad- sion has been assessed (Song and Evans,
mor and Gogos, 1979; Weidenbaum, 1958; 1993 a). In general, applied shear stresses
Irving and Saxton, 1967; Middleman, must be 5-10 times greater than the ag-
1977). glomerate strength. This can be under-
Often, the geometry of flow and leakage stood in terms of Kendall's fracture me-
paths prevents calculation of shear rates chanics theory of agglomerate strength
experienced by the material. This is not so (Kendall, 1988). Dispersion requires suc-
for the twin-roll mill. For this simple ge- cessive fracture at ascending stress levels,
ometry (Fig. 8-6 a) the maximum shear the first fracture being propagated by the
rate in the nip is given by (Cogswell, 1975) largest defect at the lowest stress.
Dispersion in shear flow can also be re-
(8-8) garded as an erosion, rather than a frac-
2H,
,T 3 2
ture proces but observation of carbon
1 or black agglomerates suggested that erosion
Where
^=^-4(1+P) 7i ="
Hr was a slow process and that agglomerate
2
whichever is larger, and A = /3— 1 where size remains largely unchanged until a cer-
/? is the die swell ratio. The parameter v is tain shear stress is reached (Rwei et al.,
the average surface speed. This means that 1990).
 8.4 Characterization of Dispersion 279

2 Ho
extensive work on dispersion of inorganic
solids in organic media developed by the
paint, polymer and printing-ink industries.
Optical and electron microscopical
methods are used for the characterization
of dispersion in all these industries but
they tend to require laborious preparation.
Attempts to remove the organic phase by
pyrolysis or solvent extraction are prob-
lematic (Ess et al., 1984). Plasma etching in
an oxygen environment is claimed to re-
move the continuous phase at the surface
and is well suited for paint films (Prosser,
1985). Microtomy or cryomicrotomy are
successful in producing film in the 2-5 jim
region for optical transmission for rubbers
(Leigh-Dugmore, 1956) and for low filler
loadings ( < 35 vol.%) in polymer but are
of limited use for loadings encountered in
ceramic processing. Reflected light micros-
copy of polished sections is also successful
for detection of agglomerates (Ess et al.,
1984).
Optical microscopy is acceptable for dis-
persion assessment of particles above
0.5 j^m or 1 jum (Vondracek and Vys, 1985)
and standard procedures have been devel-
oped for polymer fillers (ASTM, 1975). An
c. extensive review of the practical aspects
Figure 8-6. Schematic diagram of the A) twin roll
mill, B) twin screw extruder and C) double axis lobed has been published by Hess (1991). Dark
field methods in which surface steps or ir-
regularities act as diffracting centres in re-
flected light microscopy have been devel-
oped for rubber- (Ebell and Hemsley,
8.4 Characterization of Dispersion 1981) and polymer- (Ogbobe and Hems-
ley, 1989) filled materials and these tech-
In a manufacturing operation with as niques have been automated. For ultrafine
many steps as injection moulding, the me- particles, scanning electron microscopy
chanical strength of the final ceramic body (Shimizu et al., 1986; Leriche and Deletter,
does not disclose the stage at which defects 1986) and transmission electron microsco-
originated. The systematic and preferably py are used (Hess etal., 1969; Ribeiro
quantitative assessment of dispersion at etal., 1981).
the mixing stage is desirable but notorious- Various methods are employed to quan-
ly difficult for fine powders. Fortunately, tify dispersion (Shimizu etal., 1986;
the ceramics community can draw on the Leriche and Deletter, 1986) using quanti-
280 8 Injection Moulding

tative microscopy or other image analysis been used as an indication of dispersion

techniques (Hess and Garret, 1971; using a surface roughness analyser (Veg-
Suetsuga, 1990). Ceramic injection mould- varietal., 1978).
ing mixtures generally may be diluted in
the mixing machinery without affecting
agglomerate size distribution because low-
er shear stress prevail at low volume frac-
8.5 Flow Properties of Injection
tion of powder. This facilitates microscopy Moulding Suspensions
and renders other techniques usable. Song
and Evans (1993 a) used suspensions at Unlike other ceramic forming methods,
10 vol. % to assess deagglomeration in in the plastic processes particles are se-
< 100 nm ZrO 2 powders and 250 nm TiO 2 parated during mixing and remain separat-
pigments by collecting the backscattered ed after moulding. Figure 8-2 shows the
electron image from fracture surfaces. free volume fraction of organic vehicle
(Knax ~ V) without which there would be
Suetsugu etal. (1990) used light scat-
no fluidity. This excess remains until it is
tering for particles down to 70 nm at
extracted after moulding. The free volume
5-30 vol.% while the back scattering of
fraction has a commanding influence on
infra-red radiation is used for pigment ag-
mass and heat transfer and on the origin of
glomeration in paint films (Rutherford
defects in injection moulding.
and Simpson, 1985). Small angle X-ray
scattering is also applied to dispersion as- There is a large number of expressions
sessment in polymers for particles up to for relative viscosity, rjr= rj/rj0 as a func-
500 nm (Wilson, 1990) and neutron scat- tion of ceramic volume fraction V. It di-
tering has been used of the special case of vides into two subsets. First, those which
particle size measurement of catalysts on focus attention on relative viscosity at low
ceramic carriers (Baston etal., 1981). volume fraction and are developments of
The state of dispersion of a suspension Einstein's equation (Einstein, 1956):
also has a pronounced effect on some rhe-
i//»jo = 1 + 2 . 5 V. (8-9)
ological characteristics, particularly those
measured in oscillatory mode. In filled These rely on the addition of extra terms
polymers, dynamic and viscous moduli de- such as (Vand, 1948)
crease as particles are dispersed (Payne,
1965; Bohlin, 1990) but capillary viscome- 0 = 1 + 2.5 V+ 7.349 V2 (8-10)
try is insensitive to the changes (Bohlin,
1990). The same observations are made for Secondly there are those that focus at-
ceramic moulding suspensions (Dow et al., tention on the concentrated end of the
1988). range and therefore incorporate a term
Other methods of dispersion measure- Vmax (see Fig. 8-2) which represents the
ment include the Coulter Counter after maximum packing fraction of ceramic par-
dissolution of the continuous phase in an ticles at which viscosity approaches infini-
electrolyte (James, 1968) and contact mi- ty. This includes equations due to Eilers
cro-radiography followed by enhancement (1941):
of the image in a projection microscope
(Waterfield and Peacock, 1973). The sur- (8-11)
face roughness of filled material has also
 8.5 Flow Properties of Injection Moulding Suspensions 281

110
100 »« •
90
80
§ 70
8 60
o 50 Figure 8-7. Viscosity-volume
40 fraction curves for composites in
CE the 0 to 30 vol.% whisker content
30
range based on Chong's equation
20 Kmax=(n) 0.750; (•) 0.635;
10 (•) 0.620; (o) 0.600 and (A) 0.584
(Stedman et al., 1990 a).
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Ceramic Volume Fraction

to Mooney (1951): of whisker fraction extrapolated to

Vmax = 0.35 at a whisker fraction = 1, corre-
/2.5 V
17/1/0 = exp (8-12) sponding to model predictions of Milewski
V —V
Y
(1986).
'max
to Maron and Pierce (1956) as develop- Some powders do not fit Chong's equa-
ment by Kitano et al. (1981): tion perfectly. Vmax is a variable deduced
from the experimental data which defines
(8-13) the assymptote in Eq. (8-14). The constant
V -V 0.25 influences the curvature. Replacing
r r
max >
0.25 with a variable C therefore gives
and to Chong, Christiansen and Baer
(1971): V —
(8-15)
- 0.25 V\2
(8-14)
v -v v
'max
The latter is remarkably useful in fitting
apparent viscosity data of ceramic injec-
tion moulding suspensions of fine zirconia
(Hunt etal., 1988 a), alumina (Wright
etal., 1990 b) silicon nitride and silicon
carbide reinforced silicon nitride (Stedman
etal., 1990a). In the composite case,
Chong's equation fits the suspensions of
unreinforced silicon nitride and suspen-
sions with 0-30% silicon carbide whiskers
(Fig. 8-7) so that Vmax can be deduced as a
function of whisker loading (Fig. 8-8). 0.1 0.2 0.3 0.4
This means the entire composition range Whisker Fraction of Ceramic
0-Vmax and 0-30% reinforcement is Figure 8-8. Vmax from Chong's equation plotted against
mapped. Furthermore, Vmax as a function whisker content (Stedman et al., 1990 a).
282 8 Injection Moulding

which can be rewritten as (Zhang and tion is given by

Evans, 1989)
drjr 1.5 F m a x (F m a x - 0.25 V)
1 Vm 1 1 (8 17)
(8-16) w = Jy^vf "
1-C so that for Vmax = 0.75 and K=0.6, a
which is a straight line with slope Vm/ 1 vol.% change in ceramic volume fraction
( 1 - C ) and intercept (1/(1 - C ) . Thus for a produces a change of 2 in rjr increasing it
well-characterized organic vehicle, viscosi- from 16 to 18. If a suspension is too crowd-
ty measurements on only two suspensions ed it will cause injection moulding machine
of different volume fraction are sufficient seizure. The author's experience of remov-
to characterize the entire range. With C ing a seized helical screw with hydraulic
and Vmax as empirical constants the equa- jacks counsels practitioners to screen all
tion is widely applicable as a way of pre- suspensions by apparent viscosity mea-
dicting viscosity with minimal experimen- surement.
tation. A similar procedure can be applied There are many methods of characteriz-
to the other equations (Edirisinghe et al., ing the flow properties of ceramic suspen-
1992 a; Wildemuth and Williams, 1984). sions, each having its own advantages and
Since mixtures which approach Vmax disadvantages. The measurement of ap-
cannot be made without air entrapment, parent viscosity in a capillary rheometer is
this fraction cannot be directly measured a quick procedure giving f/ A =/(y, T, V).
and its physical significance is elusive. The shear rate dependence is given by
Thus it has been argued that relative vis- rj = Kf~^ where n is the flow behaviour
cosity should be calculated from constant index. This describes the fluid as Newtoni-
shear stress, not constant shear rate (Ki- an (n = 1), pseudoplastic (n < 1) or dilatant
tano et al., 1981) and this will give a slight- {n>\). For ceramic injection moulding it
ly different volume of Vmax (Zhang and should be pseudoplastic in the shear rate
Evans, 1989). Furthermore at low shear range experienced in the machine and cav-
stress, Vmax is itself dependent on shear ity. Coarse particles in low molecular
stress (Wildemuth and Williams, 1984), a weight liquids sometimes undergo a dila-
phenomena related to the yield stress of tant transition at high shear rates (Issitt
suspensions. and James, 1986). A three-dimensional
The constants Vmax and C tend to vary particle network forms from which the ve-
widely for different powders and slightly hicle can sometimes separate. Shear thick-
between batches of the same powder. It ening (dilatancy) is well reviewed by
would be ideal if Vmax could be deduced Barnes (1989).
from independent powder data such as The shear rate of interest is 100 s" 1 to
particle size distribution, shape factors or 1000 s" 1 approximately and this corre-
volume specific surface area. A number of sponds to the region where n is often con-
calculations of packing efficiency have stant so that logarithmic plots of rj against
been made (Bierwagen and Saunders, y give straight lines. At lower shear rates,
1974) but the problem of applying them to unfilled polymers tend to show near-New-
viscosity prediction is that accuracy to « tonian behaviour while filled systems show
1 vol.% is needed. the effects of yield stress. A general guide-
The local error in relative viscosity in- line for the fluidity of suspensions is that
curred by an error in ceramic volume frac- viscosity at the lower shear rate end
 8.5 Flow Properties of Injection Moulding Suspensions 283

(100 s" 1 ) should not exceed 1000 Pas 50% of the flow (Tsao and Danforth,
(Mutsuddy, 1983]. This has been success- 1993).
fully used as a guide for the development There is evidence that with coarse parti-
of compositions (Mutsuddy, 1983 a; cles, radial separation occurs in capillary
Edirisinghe and Evans, 1987 a, b). It flow such that the fluid near the wall is
should not be interpreted too strictly; sus- depleted in dispersed phase (Gauthier
pensions with viscosity of 1500 Pas and etal., 1971; Kubat and Szalanczi, 1974).
occasionally higher can sometimes be These effects make the true viscosity an
moulded. The likelihood of a suspension elusive quantity. One of the main justifica-
causing machine seizure is dependent on a tions for the use of the capillary rheometer
combination of factors among which fluid- is that it is a test whose configuration sim-
ity is important but not exclusive. ulates the flow of suspensions in the ma-
Probably the main criticism that can be chine and cavity.
levelled at the relevance of capillary The spiral flow mould is also popular in
rheometry is that the parabolic velocity ceramic injection moulding (Willermett
profile across the capillary does not prevail etal., 1978; Mangels, 1978). The distance
with highly filled suspensions. Thus the ex- along a spiral channel reached by the melt
pressions for shear rate and shear stress at before solidification occurs is taken as a
the wall (Cogswell, 1981a) measure of fluidity. In fact it has a complex
dependence on fluidity, temperature de-
. 4Q APr pendence of fluidity, thermal diffusivity
(8-18)
y = nr
—~ and T = and injection temperature. Nevertheless,
are not strictly valid. There is likely to be Skinner and Taylor (1960) were able to
a much steeper velocity profile at the wall correlate spiral flow lengths with capillary
and a flatter profile near the centre. Such a viscosity data for unfilled polystyrene.
profile is inherent in pseudoplastic materi- Their view is that a single capillary flow
als and is accommodated by the Rabinow- viscosity measurement gives a good mea-
itsch correction which adjusts wall shear sure of mouldability. Perhaps one of the
rate (Cogswell, 1981b): disadvantages of the spiral flow test is that
the suitability of the material for the ma-
(8-19) chine is only discovered post facto!
y=
nr Since the injection moulding process in-
However, this does not correct for the volves melt flow into a series of colder
effect of 'slip' at the die wall. This can be channels, the temperature dependence of
done by using dies of fixed L/r to give the viscosity also controls mould filling (Weir,
same pressure gradient but different r. The 1963) and is reflected, inter alia, in the spi-
apparent wall shear rate is then ral flow test (Weir et al., 1963). When the
formation of defects in large sections is
4V
(8-20) examined it again becomes clear that a low
temperature dependence of viscosity is de-
where the true wall shear rate is yt and V is sirable (Hunt et al., 1991 c). The final stage
the slip velocity obtained from the slope of of packing involves a competition between
y against l/r (Cogswell, 1981b). In com- the overall shrinkage of the material in the
posite ceramic suspensions at shear rates cavity and the diminishing flow of liquid
above 1000 s" 1 wall slip can account for along the core of a solidifying sprue. The
284 8 Injection Moulding

decay of cavity pressure is consequent where a is a constant viscosity and b is a

upon the inequality constant with units K~l. Graphs of \nr\
against T show negative slope but also
dF
with limited linearity. Thus, a low value of
shrinkage flow 'activation energy' for viscous flow or a
low value of d(lnrj)/dT is a desirable fea-
being met. Thus, Hunt (1991) showed that
ture for the filling of thin sections without
with an amorphous polymer as the organic
short shots and for the final stage packing
vehicle, where viscosity decreases steeply
of large sections.
with temperature in the region above Tg,
the calculated solidification time for the The viscosity of polymer melts can also
sprue based on Tg is significantly preceded be expressed as a function of the difference
by cavity pressure decay. Conversely for between the melt temperature and the glass
semicrystalline thermoplastic polymers, transition temperature (T—Tg) by the
which are processed well above Tg, the Williams-Landel-Ferry equation (Ferry,
temperature dependence of viscosity is low 1970) which arises from free volume theory
and solidification provides a sharp discon- and is general:
tinuity so that calculated sprue solidifica-
tion time matches the time at which cavity log 10 (8-23)
pressure falls quite well (Zhang and Evans,
1993 a). It may be thought, ipso facto, that Differentiating Eq. (8-23) gives d(log10 rj)/
semicrystalline polymers are preferable. d l a s a negative slope which is inversely
However, they also tend to present a sig- proportional to (T-Tg) (Brydson, 1981).
nificant reduction in specific volume on Thus, as discussed above, polymers which
crystallization which enhances shrinkage. are processed well above Tg show a low
The temperature dependence of viscosi- temperature dependence of viscosity.
ty can be expressed in several ways (Bryd- The constants c1 and c2 are characteris-
son, 1981). The well-known Arrhenius re- tic of the polymer. Ferry (1970) gives ex-
lationship gives amples. Interestingly, these constants
along with Tg appear again in Sec. 8.8.4
= AeE'RT (8-21) where they control the diffusion coefficient
for degradation products in solution in the
and plots of In rj versus l/T for polymeff
parent polymer during binder removal
give positive slopes which are linear over
through free volume theory.
ranges of about 60 K. Ideally this should
A third type of test is the dynamic cone
be plotted at limiting zero shear rate but
for characterization of ceramic suspen- and plate viscometer in oscillatory mode
sions a shear rate of 100 s~* has been used with the ability to control the frequency
(Edirisinghe and Evans, 1987 a, b). The and the amplitude. It is argued that this
constant E is often called the activation overcomes the problems at the surface of
energy for viscous flow but care should be the die in capillary flow (Mutsuddy,
exercised in its interpretation (Hildebrand, 1983 b). For unfilled polymers it is an em-
1977). pirical observation that the angular fre-
quency corresponds to shear rate in the
An alternative relationship is (Brydson,
capillary rheometer in a restricted shear
1981)
rate range 10" 2 s" 1 to 10 s" 1 (Cox and
= ae~bT (8-22) Merz, 1958).
 8.6 Physical Properties of Suspensions 285

A different approach to fluidity assess- the ability to deduce the properties of poly-
ment is provided by the critical powder mer blends from their compositions.
volume concentration (CPVC). A crude es- In order to model the solidification stage
timate is obtained from the volume frac- of ceramic mouldings to predict the inci-
tion of powder which makes a non-crum- dence of voids (Hunt et al., 1991 b), cracks
bling paste with linseed oil (ANSI ASTM, (Hunt etal., 1991 d) and sprue solidifica-
1974). A better method uses a torque tion time (Hunt et al., 1991 c), Hunt devel-
rheometer to measure the maximum oped techniques to obtain thermal expan-
torque during the progressive addition of sion coefficient of the solid, equation of
an oil or carboxylic acid to the powder state, specific heat, thermal diffusivity and
(Markoff etal., 1984). The volume frac- volume thermal expansion coefficient in
tion of powder at this point is called the the liquid state (Hunt et al., 1991 a).
CPVC. The CPVC can also be correlated If the matrix bulk modulus is low com-
with a temperature rise during mixing (Pu- pared with the bulk modulus of the filler
jari, 1988, 1989). The maximum density of then the specific heat of a composite is
ceramic suspensions is recorded at the given by the law of mixtures (Hunt et al.,
CPVC, showing that above this volume 1991a; Christiansen, 1979; Hale, 1976;
fraction air entrapment occurs (Pujari, Taylor, 1991). Deviations of up to 9%
1989). from the law of mixtures have been found
(Zhang etal., 1989d) in the mid-volume
fraction range. It is possible to estimate the
specific heat of polymers and polymer
8.6 Physical Properties blends using group contributions (Zhang
of Suspensions etal., 1989d) by using data and proce-
dures from the excellent review by van
The injection moulding process involves Krevelen (1972a).
a sequence of operations during which in- Hunt etal. (1991a) measured the ther-
ternal defects can accumulate, presenting mal diffusivity of a ceramic suspension in
themselves for inspection in the final prod- the solid and liquid states over a range of
uct but being reluctant to disclose the stage temperature using a modification of the
of their origins. It is often held that binder method of Hands and Horsfall (1977). In
removal is the main source of defects but this method two discs of the material are
evidence sometimes suggests that these de- sandwiched between bronze plates with
fects have their origin during solidification thermocouples at the surfaces and at the
in the cavity and are enhanced during centre. The assembly is then heated at a
binder removal (Thomas and Evans, controlled linear rate and the thermal dif-
1988). To better understand the solidifica- fusivity at any temperature may be found
tion process, the thermal and mechanical from
properties of suspensions are needed and
preferably these properties should be capa- a = -12(7; - r ) (8-24)
ble of estimation without the laborious ex- 0

perimental work needed to screen a large where / is the thickness of each disc, T1 the
number of formulations. This implies a average of the two surface temperatures,
knowledge of the dependence of physical To the centre temperature and the prime
properties on ceramic volume fraction and indicates the rate of temperature rise.
286 8 Injection Moulding

III. IV.

Figure 8-9. Thermal con-

ductivity as a function of
ceramic volume fraction for
alumina -polypropylene
suspensions. Superimposed
are model curves for i)
Hashin and Shtrikman up-
per bound, ii) Landauer,
iii) and iv) Nielsen,
v) Eq. (8-26), vi) Maxwell
and Hashin-Shtrikman
lower bound (Zhang et al.,
1989 a).
10 20 30 40 50 60 70
volume percent ceramic

Zhang etal. (1989 d) used the same most metals and ceramics in continuous
method to measure the thermal conductiv- polymer media.
ity as a function of ceramic volume frac- The method measures thermal diffusivi-
tion for alumina suspensions up to ty directly in the solid and liquid states.
62 vol.%. A large number of expressions Thermal conductivity is obtained from the
governing mass, electrical, magnetic and diffusivity as X = a Q C, where C is the
thermal transport phenomena in two- specific heat. Van Krevelen (1972 b) also
phase composite materials have been discusses ways of estimating thermal con-
derived and are well reviewed (Hale, 1976; ductivity of polymers from group contri-
Barrer, 1968; Mohram and Taylor, 1991; butions. Linear thermal expansion of ce-
Crane et al., 1977). Of a number of such ramic suspensions can be measured in the
expressions, that of Maxwell (1892) pro- solid state by dilatometry (Zhang and
vided the best fit to the data (Fig. 8-9): Evans, 1990 a) and the bulk expansion co-
efficient can be measured in the liquid and
(1 +2F 1 )2 1 + 2(1-K 1 )A 2 solid states on degassed samples by mer-
1 kl (8 25)
~ d W +e + w " cury immersion dilatometry (Hunt et al.,
1991a; Zhang and Evans, 1990a; ASTM,
where Vx is the ceramic volume fraction. It 1978). This method provides the dilation
can be seen that if A* g> k2 > thermal con- associated with the melting of the crys-
ductivity is controlled almost entirely by talline phase.
the continuous phase and
Certain precautions are needed when
dilatometry is applied to heavily filled in-
^ ~ ^2 ( ? / 2 T ^ ) (8"26) jection moulded polymers. The thermal ex-
pansion in the solid state is perturbed by
Interestingly, this simple approximation the release of residual stresses or the relax-
provides a slightly better fit (Zhang et al., ation of orientation due to moulding
1989d). The simplification is valid for (Zhang and Evans, 1990b, 1992 a). The
 8.7 Solidification in the Cavity 287

anomalous expansion or contraction varies T). The specific volume of the ceramic is
throughout the moulding with direction, thus given by
giving sometimes negative apparent ex-
r 1 = K 1 ( 2 9 8 ) [l+/?(7'-298)] (8-31)
pansion coefficients (Zhang and Evans,
1992 a). Thus extensive annealing is neces- and V2 is found from Eq. (8-27) so that the
sary before true expansion coefficient can specific volume of the suspension is given
be recorded. by
Taylor (1991) and Hale (1976) review
expressions for the thermal expansion co-
V*=V1X1 (8-32)
efficients of composites. The expansion co- An example (Zhang and Evans, 1990 a)
efficient of polymers can also be deduced of the specific volume of a 56 vol.% alumi-
from group contribution theory with rea- na suspension in polypropylene is given in
sonable accuracy (van Krevelen, 1972 c). Fig. 8-10.
Hunt et al. (1991 a) derived the equation The analysis of the development of
of state for a ceramic suspension using the stresses in mouldings requires a knowledge
Spencer and Gilmore (1949, 1950) version of mechanical properties such as Poissons
of the van der Waals equation as applied to ratio, elastic modulus, fracture stress and
polymer systems: ftie temperature and ceramic volume frac-
tion dependence of these. Some generaliza-
(8-27) tions for these properties have been de-
duced (Hunt etal., 1991a; Zhang and
where M is the molecular weight of the Evans, 1991a).
repeat unit and n and co are constants to be
determined by experiment. The parameter
co can be found by extrapolation of ther- 8.7 Solidification in the Cavity
mal expansion data to 0 K whereupon
V(0) = co. The slope of the volume expan- Remembering that injection moulding
sion curve dV/dT at nominally zero pres- has its origin in the foundry, it is not sur-
sure then gives a value for n prising that it has inherited some of the
problems of casting. In his paper on solid-
(8-28) ification, Davies (1973) describes the for-
mation of shrinkage voids in a large metal
casting. The premature solidification of
The volume thermal expansivity is given
the feeder, coupled with the volumetric
by
change on solidification and the high ex-
pansion coefficient of liquids compared to
'•ME (8-29) their solids means that voids form in a
sealed pocket of fluid. In the foundry, a
and the isothermal compressibility by range of techniques is used to prevent
1 fhV shrinkage voids (Francis and Pardoe,
(8-30) 1970). Temperature gradients in the cast-
ing are controlled with chills, feeders are
The ceramic phase can be treated as in- insulated and sometimes are made from
compressible compared to the continuous materials which react with the molten met-
organic phase so that Vt =f(T), V2 =f(P, al (exothermic feeders). The machining of
288 8 Injection Moulding

4.10-

4.05-

4.00H

8. 3.90-

3.85

0 50 100 150
pressure / MPa

Figure8-10. P-V-T behaviour of a liquid suspension of 56vol.% alumina based on polypropylene in the
temperature range relevant to injection moulding (Zhang and Evans, 1990 a).

a casting sometimes results in significant Thomas and Evans (1988) show how the
deformation because of the unbalancing of incidence of voids is dependent on section
residual stresses which were set up during size and hold pressure (Fig. 8-11). As the
solidification. Castings are often therefore hold pressure is increased, the voidage re-
given a stress-relieving heat treatment. cedes toward the centre and then disap-
The use of a step wedge cavity allows pears. Unfortunately, the use of high pres-
solidification defects in ceramic moulding sure has the effect of changing the stress
to be studied as a function of mould thick- distribution in the moulding so that cracks
ness. Silicon powder suspensions in appear.
polypropylene-based binders produced Thomas and Evans (1988) showed that
shrinkage defects which appeared in thick- the solidification stage also influenced the
er sections (10-20 mm) and could be relat- morphology of defects which appeared af-
ed to the total crystalline volumetric ter binder removal. Cracks which could
shrinkage of the organic vehicle and to the just be detected after reheating mouldings
temperature dependence of viscosity to the softening point were extended once
(Edirisinghe and Evans, 1987c). Low val- the wax had been removed (Fig. 8-11). De-
ues of both parameters are preferred. fects sometimes associated with the binder
Once material has filled the cavity, the removal step therefore have their origin in
machine continues to apply pressure which the solidification process. In one example
is transmitted to the solidifying moulding (Katayama, 1986) the problems of injec-
for as long as the sprue allows flow. This tion moulding a thick section have been
sets the level of pressure in the core of the assessed to be so severe that a two-stage
moulding at the point of sprue closure. manufacturing process has been adopted.
 8.7 Solidification in the Cavity 289

A: as moulded 8.7.1 Void Formation

Hunt et al. (1991 b) compared predicted
and experimental voiding for a zirconia-
polystyrene suspension in the form of long
rectangular bars. Residual pressure was
plotted as a function of injection pressure
and was seen to be very sensitive to the
sprue solidification time ts. For an amor-
91MPa 104MPa 117MPa 129MPa phous polymer 'solidifying' through Tg9
this time is not specific (Hunt et al.,
1991c), and calculated values of /s differ
from the cavity pressure decay time
(Fig. 8-12). This means that upper and
lower bounds for the minimum hold pres-
sure to cause voiding can be found from
the experimental and calculated ts, respec-
tively. The minimum hold pressure is that
needed to prevent the pressure in the last
B. C. D. E. pocket of fluid from falling to zero.
Figure 8-11. Contact radiographs of alumina-wax
Zhang and Evans (1993 a) approached
step-wedge bars: A) as moulded at various pressures the same problem for a short cylindrical
showing the dependence of voidage on hold pressure; moulding but in a slightly different way.
B) a high pressure moulding reheated to the softening An experimental study of the effect of in-
point of 80°C (note the crack near the bottom); C) jection pressure shows the decrease in
after binder removal: cracks focus around the gate;
D) after sintering (plan); E) after sintering (elevation)
voiding, and its recession to the centre of
(Thomas and Evans, 1988). the moulding as hold pressure is raised.

60-

Thus a sintered silicon nitride rotor is

made by joining an isostatically pressed
hub to an injection moulded blade ring.
The cracking found in large mouldings
made at high hold pressures does not nec-
essarily occur in the cavity. It can be de- 20
tected audibly a few minutes after ejection
from the die or several weeks later during
storage (Zhang et al., 1989 a) or during re-
0 20 40 60
heating.
Time /s
The aim then is to understand the origin
Figure 8-12. Cavity pressure as a function of time for
of voiding and of stress development so
polystyrene-zirconia rectangular bars moulded at
that interventions in materials selection 493 K with a sprue radius of 5 mm (-) and 3.5 mm
and machine techniques can be implement- ( ). The calculated times for sprue solidification
ed. are marked (Hunt et al., 1991c).
290 8 Injection Moulding

Figure 8-13. Sections of

40 mm diameter cylindrical
mouldings of 56 vol. % alu-
mina-polypropylene made
at pressures (top left to
bottom right) of 65,87,108,
119,130,141,152,152 MPa
(Zhang and Evans, 1993 a).

When voiding ceases, higher hold pres- transfer coefficient h to be calculated dur-
sures cause cracking (Fig. 8-13). ing the course of solidification. At high
These calculations require a knowledge injection and hold pressures, the value of h
of the temperature at each point in a was about 2000 W m " 2 K " 1 but when low
moulding during solidification and cooling hold pressures were employed, the surface
which is obtained by finite difference temperature of the moulding actually rose
methods. However, this information can during solidification corresponding to a
be deduced from charts for symmetrical loss of contact at the mould wall due to
shapes and readily used to estimate solidi- shrinkage. This produced to a transition to
fication times. / ? ^ 5 0 0 W m " 2 K " 1 . This also explains the
In Fig. 8-14 graphical reference plots of phenomenon of sinking deformation seen
normalized surface and centre tempera- in some mouldings (Hunt etal., 1991c).
tures for infinite flat plates and cylinders Sinking is only possible when the wall of
are given for values of Biot's modulus rele- the moulding is soft, but instability is in-
vant to ceramic and metal injection mould- troduced because separation from the wall
ing. Inspection of these general charts al- causes a rise in wall temperature. Such de-
lows rapid estimation of the sprue and formation is characteristic of mouldings
mould solidification times. Values of ther- with thick sections with low ceramic vol-
mal diffusivity can be found using the ume fractions, made with low hold pres-
methods described in Zhang et al. (1989). sure.
Zhang and Evans (1992 b) measured the While Hunt's model for void formation
surface temperature during injection (Hunt etal., 1991c) referred to an amor-
moulding by trapping a fine thermocouple phous polymer, Zhang and Evans (1993 b)
in the clamp. This allowed the surface heat dealt with the situation for a semicrys-
 8.7 Solidification in the Cavity 291

talline polymer (a mixture of atactic and avoids the very considerable complexity
isotactic polypropylenes with 30 vol.% introduced by considering the distribution
crystallinity). The residual amorphous of pressure in the semicrystalline solid and
fraction is well above its Tg of-14°C and hence the variable volume strain through-
the equation of state now includes a out this region.
specific volume change as crystallization When the sprue solidifies, WQ can be
occurs. found from the hold pressure condition
In the first stage of solidification the and from the temperature distributions.
sprue is molten and flow into the cavity After sprue solidification, the total weight
compensates for volume shrinkage so that needed to fill the cavity, solid and liquid,
the mass in the cavity increases. The pres- can be calculated at time intervals for var-
sure in the cavity at this stage is taken as ious cavity pressures. The pressure at
equal to the pressure on material set by the which this weight is equal to Wo can then
machine because the pressure defect in the be found from a computer generated file.
flow channels is low. This is the pressure in the cavity at that
The specific volumes of each element time. Thus the time required for ^ = 0 is
(i, j) are given by read from the file for the time at which the
weight at P = 0 corresponds to J¥o. The
pressure at any time can be found in a
similar way.
(8-33)
In fact, results for the model that as-
when JTJJ is higher than Ts, A, n and w are sumes incompressibility in the solid state
constants in the equation of state, X repre- works well (Table 8-4). These models pre-
sents weight fraction and subscripts v and dict that the gap between the injection and
c refer to organic vehicle and ceramic pow- solidification temperatures should be as
der. low as possible commensurate with mould
Two situations were explored below Tg, filling requirements, that the volume ex-
one where the solid was treated as incom-
pressible and
TUi<Ts (8-34) Table 8-4 Comparison of predicted and observed in-
cidence of void formation (Zhang and Evans, 1993 b).
where /Js is the volume expansion coeffi-
cient of the suspension. In the lower bound Cylinder Mould Predicted Observed
calculation an equation of the type shown diameter temp minimum minimum
(mm) (°C) hold pressure hold pressure3
by Eq. (8-33) was used with the coefficients (MPa) (MPa)
modified to fit experimental data below Ts.
In the second stage of solidification, the upper lower
sprue has solidified and no more material
20 20 115 68 119-130
can enter the cavity. The weight of the 20 80 110 63 108-119
sample Wo is therefore constant. The crite- 40 20 174 115 119-130
rion for a void-free moulding is that when 40 80 142 88 108-130
the centre temperature has reached the so- a
The range of pressure given arises from the pres-
lidification temperature, the pressure sures selected in the experiments. Thus at the lower
should still be greater than zero. The use of pressure voids were present and at the higher values
an upper and lower bound procedure they were not.
292 8 Injection Moulding

0.2 0.3
crt/R2

Figure 8-14. Graphical ref-

erence plots of normalized
surface and centre tempera-
ture for infinite flat plates
and cylinders for values of
Biot's modulus P = hR/A
relevant to ceramic and
metal injection moulding
where JR is the plate half-
thickness.
02 03 0.4 0.5 0.6 ..0.7 0.8 0.9

pansion coefficient in the bulk should be shrinkage on crystallization it is a some-

low and that the time for sprue solidifica- what abstract exercise.
tion should be long.
Methods for extending the sprue solidi-
8.7.2 Stresses in Mouldings
fication time are discussed below. Zhang
and Evans (1993 b) also calculated the Cracks which present themselves after
pressure decay curve when the volumetric moulding or during storage are prevalent
shrinkage associated with crystallization when large sections are moulded or when
was notionally eliminated. This gave a pro- the cavity design restricts shrinkage. In the
longed cavity pressure decay curve but moulded state, ceramic suspensions have
since the majority of polymers undergo a low strengths 5-18 MPa (Hunt etal.,
 8.7 Solidification in the Cavity 293

Figure 8-14. Continued

1991a; Zhang and Evans, 1991) and this ly must accommodate the following ef-
strength decreases rapidly with tempera- fects.
ture. Cracking is often noted when high
hold pressures are used (Zhang and Evans, 1. The solidification and thermal contrac-
1993 a; Thomas and Evans, 1988; Hunt tion of successive layers in a material
et a l , 1991 d). A full analysis of the devel- where elastic modulus increases sub-
opment of stresses in polymer mouldings stantially as temperature falls results in
has yet to be made but specific aspects of each layer placing its predecessor in
the problem have been addresed (Hunt compression and itself in tension. This
etal., 1991 d). A full analysis, which can effect is similar to the quench-toughen-
only be achieved numerically and iterative- ing of glass and it produces the same
294 8 Injection Moulding

surface-compressive, centre-tensile stress 5. Stress relaxation occurs in solid layers

distribution. Mills (1982) has used a sim- during cooling. Quasi-static measure-
ilar analysis for the stress build up in ments of modulus as a function of tem-
glass (Aggarwala and Sabiel, 1961) to perature which are needed for numerical
predict the stress distribution in polymer stress analysis do not accommodate the
extrudates. relaxation of stress as a function of tem-
2. The change in pressure in the liquid core perature in the recently solidified layers.
effects stresses in already-solid layers in This is particularly relevant for semi-
a complex way. Mills [1983] analysed the crystalline organic materials where the
stress development in an injection amorphous fraction continues to relax
moulded polymer assuming that only well below r m .
the change in pressure influenced the
stress distribution. The decay of pres- It can be seen that the combination of
sure then causes a hydrostatic dilation in these effects presents a formidable prob-
the solid layer. lem and it is understandable that it has
3. A restriction on shrinkage is caused by been addressed only in part even for poly-
the cavity design. Hunt et al. (1991 d) mer mouldings. Some deductions concern-
compared the predicted stresses in re- ing the influence of material properties can
strained mouldings with the incidence of be made from existing models. Low ther-
cracking using a circular disc with a cir- mal expansion coefficient and a narrow
cumferential lip. The stresses that devel- band between softening point and mould
op in the mould cavity during cooling temperature are desirable.
are greater than those in the residual Hunt (1991 d) analysed the stress build
stress distribution out of cavity. For a up in restrained and unrestrained centre-
crack/no crack criterion it is the in-cavi- gated circular discs by incorporating ef-
ty stresses that are needed. These are fects 1, 2 and 3. This model, like existing
very sensitive to the temperature inter- solutions, presents a parabolic stress pro-
val between the solidification point and file with the tensile stress at the centre with
the mould temperature. variations produced by the pressure re-
4. The pressure in the liquid core restricts sponse with time and by the rate of cool-
shrinkage by acting on its projected ar- ing. The effect of pressure decay is itself
eas in the solid walls. Few analysts complicated because the reactions in the
(Titomanlio et al., 1987) have attempted mould wall at the edges are unknown.
to incorporate this effect because it im- Thus, for recently solidified layers, the
plies that the bending moment in the strain consequent upon pressure decay is
solid at the corners must also be consid- an hydrostatic dilation as analysed by
ered. However, it is this effect which Mills (1983) but for layers which have been
causes the overall shrinkage of unre- solid for some time and have undergone
strained mouldings to be substantially substantial thermal contraction, the decay
less than the theoretical shrinkage in the of pressure causes a biaxial extension as
cavity. This effect also appears to be re- discussed by Hunt et al. (1991 d).
sponsible for the cracking seen in some Cracking appeared when high hold pres-
heated-sprue mouldings which are dis- sures were applied and this could be allevi-
cussed below. ated by high mould temperature. The so-
lidification time for the sprue is also an
 8.7 Solidification in the Cavity 295

influential parameter for the control of obtained. Treuting and Read (1951) devel-
stresses. Thus, if the sprue were kept oped a method of obtaining the stress dis-
molten for the duration of solidification of tribution from the curvature plot.
the moulding, the moulding would experi- The curvature of the bar can be found
ence no cavity pressure decay. Mills (1983) optically (Coxon and White, 1980) or by
states "The level of residual stress can be the use of strain gauges (Tandon and
reduced by any method that minimizes the Green 1990) provided precautions are tak-
variation in the cavity pressure during so- en to avoid strain gauge heating (Zhang
lidification". Allowing for the difficulties et al., 1992). One of the other weaknesses
in knowing the sprue solidification time, of the method as applied to semicrystalline
Hunt's calculation gave a reasonable organic systems is that stress relaxation
agreement with his crack/no crack criteri- continues at such a rate after removal from
on. Since the tensile strength of the mould- the cavity that the long time required for
ed material was 18 MPa at room tempera- the experimental method precludes the ac-
ture, cracking was predicted if the centre curate measure of maximum stress distri-
tensile stress exceeds this value. bution. Thus, Kostic et al. (1992) mea-
sured residual stress distribution in injec-
tion moulded alumina-polypropylene sus-
8.7.3 Measurement of Residual Stress
pensions 600 ks after ejection. They also
The methods that have been applied to assessed the errors inherent in the process.
the measurement of residual stresses in Stress distributions were found to be cen-
polymer mouldings include layer removal tre-tensile surface-compressive as predict-
(Coxon and White, 1980), stress relaxation ed from theory (Mills, 1982, 1983; Aggar-
(Kubat et al., 1975) and hardness indenta- wala and Saibel, 1961) and the maximum
tion (Rache and Felt, 1971). The layer re- tensile stress was sensitive to hold pressure
moval method is based on unbalancing the during moulding as shown in Fig. 8-15. In-
internal stresses over the cross section by terestingly, heated sprue mouldings showed
removing layers from one surface and extremely low residual stresses, in agree-
recording the resulting deformation. This ment with Mill's assertion quoted above.
is repeated until the centre line is reached A further observation was that annealing
and a plot of curvature against thickness is just above the softening point of the sus-

Figure 8-15. Stress distributions mea-

sured by the layer removal method in
injection moulded plaques of 10 mm
thickness at hold pressure of A)
22 MPa, B) 43 MPa, C) 108 MPa.
(Measurements were made 600 ks
after moulding) (Kostic et al., 1992).
1 2 3
Distance from Centre /mm
296 8 Injection Moulding

pension failed to remove the residual appear in large sections are internal
stresses completely, implying a restriction (Zhang and Evans, 1993 a; Thomas and
on the relaxation of polymer molecules Evans, 1988; Edirisinghe and Evans,
perhaps associated with adsorption on 1987c; Zhang et al., 1989a) corresponding
high energy surfaces. to the centre-tensile stress distribution. In
heated sprue mouldings, the cracks ap-
8.7.4 Methods of Prolonging Sprue peared to initiate from the surface. It is
Solidification Time likely therefore that they originate from
the overall restriction on shrinkage of the
The control of sprue solidification time solidified skin of the moulding caused by
is one of the most important machine in- internal pressure, the fourth contribution
terventions in ceramic injection moulding, to residual stress discussed in Sec. 8.7.2.
particularly for large sections. There are
An inherent disadvantage of both these
three methods of control:
techniques is that the sprue, enlarged or
1. The use of a large sprue runner and gate. heated, must enter the thickest section di-
2. The insulation and heating of the sprue. rectly. The modulated hold pressure tech-
3. The modulation of hold pressure to nique does not suffer from this disadvan-
cause oscillatory flow in the feed system. tage. Originally used for polymers
The first method is the simplest but sad- (Menges et al., 1980) and filled polymers
ly very large sprues are needed, compara- (Allan and Bevis, 1983), it allows large ce-
ble, in fact, to the largest section size of the ramic sections to be made without voids or
"moulding and they must enter that section cracks (Zhang etal., 1989 a; Allan et al.,
directly. Zhang and Evans (1993 a), for ex- 1987; Edirisinghe and Evans, 1987d,
ample, show how solidification time de- 1988). Once the cavity is full, the pressure
pends on sprue diameter. on the screw is reduced to a low level suffi-
Hot runners are widely used in thermo- cient to refill the oscillating cylinder
plastic polymer moulding. Hunt and (Fig. 8-16). As the oscillating pressure am-
Evans (1991) have shown how mouldings plitude is increased a region is found for
can be made at very low pressure using a each material and cavity where macro-de-
heated sprue. Because the pressure is main- fect-free mouldings are produced (Zhang
tained throughout solidification, shrink- et al., 1989a, e; Zhang and Evans, 1993 d).
age voids are not created and because the As the amplitude is raised, the mouldings
pressure is constant throughout solidifica- again show cracking, either on ejection or
tion the residual stress distribution is near- during storage. The cavity pressure traces
ly flat. The method bears a formal similar- show that sprue solification time increases
ity to the exothermic feeder used in the monotonically with pressure amplitude.
foundry (Francis and Pardoe, 1970). The magnitude of displacement in single-
Zhang and Evans (1993 c) have applied gated modulated pressure relies on the
the same device to the moulding of large compression of fluid in the cavity so the
(40 mm diameter, 60 mm length) cylinders technique is not suitable for small mould-
confirming that large mouldings can be ings (Zhang and Evans, 1993 e). However,
made free from voids at very low pressure for large mouldings, the solidification time
(1 MPa). In such large mouldings, howev- is prolonged beyond the calculated time
er, a new type of cracking was found. In for the centre of the moulding itself to so-
conventional mouldings, the cracks that lidify.
 8.7 Solidification in the Cavity 297

PISTON
ROD UPPER PLATE

HEATER

4 BOLTS
HEATER
6 BOLTS
(equally
NOZZLE,
spaced)
SEAT
NOZZLE

Figure 8-16. The oscillating

pressure valve which is in-
4 BOLTSy
(equally \ serted between nozzle and
A INNER LINER
spaced) sprue and is externally
B INNER LINER powered to modulate the
0 PISTON pressure in the runner and
C CYLINDER mould cavity.
20mm

Reciprocating flow can continue for magnitude of reciprocating volume as a

700 s when the calculated solidification function of time during solidification
time is 20 s (Fig. 8-17), an effect even more (Zhang and Evans, 1993 d). The limitation
pronounced in polymer mouldings where on single-gated modulated pressure
mouldings can be kept "live" for « 1 hour mouldings is of course overcome for multi-
(Alan and Bevis, 1983). gated devices (Allan and Bevis, 1987).
It is extremely difficult to estimate the Nevertheless, for mouldings with large sec-
heat fluxes in this process but initial at- tions, central moulding cracks and voids
tempts have been made by using the equa- are overcome by single-gated modulated
tion of state, together with a finite differ- pressure moulding as shown in Fig. 8-18.
ence temperature calculation to find the

700-

600-

500-

^ 400
CD
Figure 8-17. The effect of
increased modulated pres-
.§ 300 sure amplitude on sprue
solidification time for cylin-
200 drical mouldings of 20 mm
diameter (•) and 40 mm
100 diameter (o) (Zhang and
0-W- Evans, 1993 d).
100 120 140
Pressure /MPa
298 8 Injection Moulding

tent. During loss of organic matter the

residue may remain immobile or redis-
tribute itself under capillary forces. As
matter is removed, V approaches Vmax and
there is a consequent shrinkage and an ef-
fective change of state as fluid properties
are lost (Wright, 1990 b). The movement of
particles accompanies each stage.

8.8.1 The Effect of Reheating Mouldings

Most of the methods of binder removal
described above involve the reheating of
mouldings until the binder is again fluid.
20mm The exceptions are freeze drying and cata-
lytic degradation of polymers in the solid
Figure 8-18. Polished sections of moulded alumina
polypropylene cylinders (40 mm diameter) made with state.
112 MPa and 126 MPa modulated pressure ampli- The anomaly in thermal expansion mea-
tude. surement discussed above (Zhang and
Evans, 1990 b) is indicative of relaxation
processes associated with residual stresses
8.8 Removal of Organic Vehicle or orientation of macromolecules. This in-
volves substantial strains which can pro-
Regarded as the most problematic state
duce negative apparent expansion (Zhang
in plastic forming operations for ceramics
and Evans, 1992 a, 1993 f). Particle move-
the removal of organic vehicle is also the
ment accompanies these strains and there-
least studied and the least understood. The
fore the rearrangement of particles before
binder can be removed by capillary flow
firing can be substantial.
into a porous body (Wiech, 1981 a), by sol-
When the softening point of the organic
vent extraction (Wiech, 1980), by freeze
vehicle is reached, particle motion is again
drying of low molecular weight species
such as water (Hausner, 1993), by catalytic possible. On a macroscopic scale, slump-
degradation in the solid state (Ter Maat ing can change the shape of the body under
et al., 1991), or by pyrolysis in either oxi- gravitational forces (Zhang etal., 1989c)
dizing or inert atmospheres (Edirisinghe and mouldings are often supported on a
and Evans, 1986). powder bed for this reason. On a micro-
scopic scale, flocculation can occur nota-
The latter method is most popular. Dur-
bly in low viscosity binders. Thus, Song
ing the early stages of reheating a series of
and Evans (1993 b) report the observation
physical changes may take place before the
of cracks at the softening point of wax
softening point is reached. Once it is
before any matter has been lost in poorly
reached, particles are free to move at a
stabilized suspensions.
macroscopic scale causing slumping or at a
microscopic scale causing flocculation. Py-
rolytic processes may result in evaporation 8.8.2 Extraction by Capillary Flow
or oxidative degradation but thermal deg- Wiech (1981 a) was able to remove large
radation will accompany both to some ex- fractions of wax-based binder by placing
 8.8 Removal of Organic Vehicle 299

the moulding in contact with a powder or The sorption constant was very low
a porous tile. Contact over a small region compared to the corresponding value for
was sufficient. German (1987) has pio- an unlimited supply. Saturation of powder
neered the quantitative study of extraction beds were much lower indicating that only
by capillary flow. Wright and Evans smaller pores were capable of opposing the
(1991 a) studied flow of wax from a mould- capillary forces in the moulding. A full
ed body into a powder and found that the analysis of the problem therefore requires
distribution of wax remained uniform and permeability and capillary pressure for
saturation decreased as extraction pro- each given powder as a function of satura-
ceeded. It was generally not feasible to tion. Intermittent loss on ignition surveys
treat the permeability of the powder bed as for sections of ceramic body and powder
infinite by comparison with that of the ce- bed confirmed that the saturation of the
ramic body. These observations make the body was uniform throughout the mould-
mathematical modelling more complex. ing at each stage. The strong dependence
Bao studied the sorption of wax by pow- of zero shear rate viscosity on molecular
der beds from an unlimited supply and weight means that capillary flow is gener-
then the corresponding one-dimensional ally useful for waxes but not high poly-
sorption from a ceramic body (Bao and mers.
Evans, 1991). No single expression for per-
meability in terms of powder characteris- 8.8.3 Solvent Extraction
tics was valid for all the powders studied.
The extraction of organic matter in a
Deviations were frequently greater than
solvent has also been developed by Wiech
two orders of magnitude. Sorption from
(1980) and the advantages of using an
an unlimited supply followed a parabolic
over-pressure during solvent extraction
law as expected:
have been demonstrated (Wiech, 1981b).
= Kt1/2 (8-35) Binders for solvent extraction are often
duplex, one species being soluble, the other
where K is the sorption constant given by remaining to restrict particle mobility
(German, 1990b). Although the literature
F) K 1/2
is not explicit, it appears that since the in-
K= V P (8-36)
teraction between solvent and polymer can
n involve both swelling and dissolution, the
with units [m s 1 / 2 ]. Interestingly, even for solvent should be a 'good' solvent for the
sorption from an unlimited supply, the polymer or wax and extraction should be
void space in the powder bed was not fully carried out above the theta temperature to
saturated; saturation was as low as 66% minimize the swelling effect.
for some powders. This means that general
expressions for capillary removal of binder 8.8.4 Pyrolytic Extraction
for diverse systems promise to be elusive.
When the wax was removed from a ceram- A wealth of literature, previously re-
ic body, there was an initial high flux in the viewed (Edirisinghe and Evans, 1986; Ger-
first few seconds and thereafter a steady man, 1990 a), describes the use of con-
parabolic loss curve given by trolled heating in various atmospheres and
pressures to remove organic matter from
(8-37) ceramic mouldings. The mechanism of
300 8 Injection Moulding

weight loss can be i) evaporation, ii) oxida- rate the mass transport kinetics within the
tive chain scission followed by evaporation ceramic moulding.
or iii) thermal degradation followed by Wright and Evans (1991 b) analysed the
evaporation. Wright et al. (1989) studied a oxidative pyrolysis of mouldings by apply-
range of polyolefins and their blends of ing shrinking core reaction kinetics. The
different molecular weight in the form of weight loss from small mouldings is known
small mouldings and fines by thermo- to be sensitive to sample size (Wright et al.,
gravimetry. Low molecular weight waxes 1989) implying some diffusion control.
show activation energies corresponding to The extent of diffusion control and esti-
the enthalpy of vaporization of long chain mates of effective diffusion coefficient can
olefms and the process is zero order. For be obtained for an infinite flat plate
higher molecular weight waxes, a depen- (Szekely et al., 1976) for isothermal condi-
dence on atmosphere presents itself in the tions from
thermograms showing that chain scission
t = TX+T(T*X2 (8-38)
precedes evaporation and this effect is
most pronounced for high molecular The extent of the reaction boundary Xis
weight polymers where the expected first given by x\ Y where x is the actual depth
order activation energies for oxidative and calculated from weight loss and Y is the
thermal degradation are recorded for fine- plate half thickness for reaction at both
ly divided material heated in air and nitro- faces.
gen, respectively. The value of the shrinking core reaction
From one point of view, materials selec- modulus of indicates the extent of diffu-
tion should aim to give a blend of materials sion or reaction rate control. For of < 0.1
with a small negative slope in the ther- the weight loss is reaction controlled. For
mogram. This approach was used in form- of > 10 the reaction is entirely diffusion
ing wide molecular weight distribution controlled. The characteristic time T gives
blends (Saito et al., 1976) and Wright et al. the reaction rate constant. From this and
(1989) show that the thermogram of as the effective diffusion constant can be
blends can be predicted with some accura- deduced.
cy from those of the constituents. Stedman Experiments on slabs of thickness
et al. (1990 b) show how blends can be se- 0.5-2 mm of alumina-polyethylene sus-
lected by a computer program which cal- pensions showed mainly diffusion control.
culates the thermograms of blends assum- For isothermal experiments at 180°C the
ing each component exerts its own decom- average effective diffusion coefficient was
position kinetics in the blend and then cal- 7 x l O ~ 1 1 m 2 s " 1 . This suggests that rate is
culates E(dw/dr) 2 for each combination controlled partly by the diffusion of reac-
of components. The compositions with the tion products out of the reaction layer be-
lowest values of this sum of squares are cause the value of the effective diffusion
selected. constant is too low for oxygen transport.
A similar effect is achieved by a process Activation energies for rate constant and
control loop which senses weight loss and diffusion coefficient would allow reaction
adjusts temperature to give a linear ther- rates to be calculated numerically for as-
mogram (Johnsson et al., 1983). These ap- cending temperature ramps and holds.
proaches, though effective, do not address Barone and Ulicny (1990) consider the
the full problem which needs to incorpo- ceramic particles to form a rigid network
 8.8 Removal of Organic Vehicle 301

within which the binder generates an hy- throughout the ceramic body producing
drostatic pressure because of its higher low molecular weight degradation prod-
thermal expansion. This causes liquid to be ucts throughout.
expelled from the body and when satura- The models described above are relevant
tion has decreased, evaporation prevails. when initial porosity exists but do not ad-
The rigid skeleton is difficult to reconcile dress the critical initial situation of a poly-
with a dispersed system with a free volume mer undergoing thermal degradation in
fraction of organic vehicle which implies the interior of a moulding which does not
the spatial separation of particles and con- contain continuous porosity. Under these
fers the ability to flow into the mould cav- conditions, the degradation products first
ity. The skeletal model is more appropriate dissolve in the 'parent' polymer. Mass
for pressed bodies or for the stage where transport then occurs under the concentra-
volume has collapsed through loss of tion gradient between centre and surface
binder so that particles are in contact. This by diffusion in the parent phase. This
has been analysed by Tsai (1991) for a transport path through the bulk provides
system containing 60 vol. % alumina and the limiting step until continuous porosity
17 vol.% binder. The pressure in pore space develops. Models based on this analysis
caused by the evolution of gas by pyrolysis (Calvert and Cima, 1990; Evans et al.,
and its thermal expansion is found and 1991) accurately predict the formation of
used to calculate the stresses on the skele- defects by locating the heating rate which
ton. Clearly a knowledge of mechanical causes the vapour pressure of diffusant
properties of the assembly as a function of over solution to reach ambient pressure,
temperature then offers prediction of fail- the condition for boiling. The unsteady
ure. Because the pore size may be com- state model and experiments are per-
parable to the mean free path of migrating formed for a polymer which decomposes
gases, Tsai considered both laminar and exclusively to monomer (Matar et al.,
slip flow gas dynamics. 1995). At any temperature T, the rate of
Stangle and Aksay (1990) provide a generation of monomer which is uniform
model for mass transport which incorpo- throughout the moulding is given by
rates unsteady state heating and momen-
q — Ko exp( — E/R T)exp x
tum effects. The description of mass trans-
port includes liquid flow in the pore struc- K0RT2Qxp(E/RT)
X A — X
ture, the evolution of gas by evaporation ZE
or by chemical reactions of the liquid and
2RT 6{RT}2
the resulting gas transport by diffusion or x l — (8-39)
convection in the pore structure.
In a ceramic moulding, the initial condi- For a infinite cylindrical moulding, the
tion is that all pores are filled with organic concentration gradient at any time is de-
matter which exists in excess (Fig. 8-2) so fined by
that particles are not in contact. During 8
heating, matter may be lost initially from (8-40)
free surfaces by evaporation or oxidation.
However, thermal degradation always ac- Time t, and temperature are related for a
companies these processes to some extent constant linear heating rate Z by Z = AT/
(Wright and Evans, 1991) and takes place At. This allows the concentration of
302 8 Injection Moulding

monomer at the centre of the cylinder to be and Bruggeman:

found at any time and hence its activity
D/D2 = (1 - V)3'2 (8-45)
from the Flory-Huggins equation:
Due to adsorption however, polymer
a = 01exp{62 + x622} (8-41) layers which are likely to correspond in
where 91} 62 are the volume fractions of thickness to the chain end-to-end distance
monomer and polymer respectively and x extend into the continuous phase and the
is the interaction parameter. The vapour mobility of chain segments in that region
pressure at a given temperature can be are limited. Ve should be increased to in-
derived from the Clausius-Clapeyron re- clude the adsorbed immobile layers as ob-
lation if the enthalpy of vaporiziation is structions discussed in Sec. 8.2.2.
known: In its initial form this model ignores the
effect of loss of matter on transport: its
In P° = - AHWJR T + / (8-42) objective being to find the critical heating
where / is a constant. Thus if Px is the rate Z c at which the rate of diffusion of
vapour pressure of monomer over the so- monomer to the surface just exceeds the
lution and /\° is the vapour pressure over rate of generation. It therefore treats the
the pure liquid, critical initial stage when all pores are
blocked. It is developed to include the for-
P1=P°91 exp (02 + x Bl) (8-43) mation of porosity in two configurations
When the vapour pressure of monomer (Matar et al., 1993). In one, the polymer
over a solution of concentration Cmax at recedes to the centre and a porous outer
the centre of the cylinder reaches ambient layer is formed. The effective radius then
pressure, the liquid becomes unstable with reduces with time as
respect to its vapour and provided condi-
r1 = r0h1/2 (8-46)
tions for nucleation are available, a bubble
will form in the ceramic suspension. where h is the fraction of initial polymer
Since the intermolecular forces in a ther- remaining. Gaseous transport through the
moplastic polymer are the relatively weak outer layer experiences a much lower resis-
van der Waals forces, the diffusion coeffi- tance than that in the polymer and can
cient for small molecules is highly concen- therefore be ignored except for the case of
tration dependent as well as showing tem- ultrafine powders. In the second case, a
perature dependence. It is calculated from uniform distribution of porosity develops
the free volume equation of Duda et al. throughout the body and mass transport
(1982) at each step during numerical solu- across a pore is regarded as fast compared
tion. This gives the coefficient for the con- to diffusion in the polymer. The diffusion
tinuous phase D2. The effective coefficient coefficient in the three-phase composite
for the composite D is found from expres- with Dc = 0 is given by Kerner (1956) as
sions for transport in two-phase media
with Dx = 0 for the dispersed phase. These p p
ID +D
models are reviewed by Barrer (1968) and
two frequently used (Evans and Ediri- 3 3VD
Vp + +- v p
singhe, 1991) are due to Maxwell: 2

3V where subscripts c, p, v refer to ceramic,

D/D2 = 1 - (8-44) polymer and void, respectively. The three
V +2
 8.8 Removal of Organic Vehicle 303

• constant radius
O shrinking core
o distributed porosity

Figure 8-19. Calculated critical heating

rates for cylinders of the model system
alumina-poly(a-methylstyrene) as a func-
_- . . . . , ^§&§=e i—i tion of radius (Matar et al., 1993).
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
Radius (mm)

situations are shown in Fig. 8-19 for the predicts the effects of increased ambient
poly(oc-methylstyrene) - 50 vol. % alumina pressures which suppress boiling (Ham-
system. The strong dependence of Z c on mond and Evans, 1995). This is a tech-
radius is clearly shown. The low heating nique for which commercial overpressure
rates are obtained because the binder is furnaces are available (Katagiri, 1990).
single phase and presents a steep ther-
mogravimetric loss. The addition of dilu- 8.8.5 Shrinkage During Pyrolysis
ents and plasticizers generally increases
diffusion coefficient by several orders of As organic vehicle is lost, the free vol-
magnitude as well as levelling the ther- ume fraction (Fmax— V) decreases (Fig. 8-2)
mogram. The development of porosity in and viscosity increases as particles ap-
real systems lies somewhere between the proach. The maximum volume fraction of
two extremes depending upon the mobility ceramic in the binary system V£ax can
of the organic phases in the pore structure. therefore be deduced from linear shrinkage
The models have been developed for the measurements before and after pyrolysis
geometry of the sphere and infinite flat assuming isotropic shrinkage:
plate (Matar etal., 1995) and have been 3 AL /AL AL
used to predict the influence of monomer 1 -
y*
and polymer properties on critical heating r
max
rate. Perhaps the most interesting aspect of (8-48)
these models is their prediction of the ef- where Lo is the initial length and AL the
fects of individual polymer and monomer change in length. Shrinkage measurements
properties. They give guidance for the of A16 alumina mouldings with different
specific design of organic species for plas- initial ceramic volume fraction V are
tic forming of ceramics. The model also shown in Fig. 8-20. V£ax for this system
304 8 Injection Moulding

always possible to explain the reasons for

their success. It should be noted that al-
5- most any organic blend will allow small
objects to be manufactured. The criterion
4-
of section size is what classifies the quality
of such blends. For every blend there will
be a limiting section for a given powder
3-
above which defects are produced. This
table reflects the diverse approaches to ma-
2- terials selection. Part a) reflects the work
of those who have sought consistency be-
1- tween materials and process; injection
moulding is a process developed for high
polymers. Part b) reflects concern about
48 50 52 54 56 58 60 62 64 the use of high molecular weight materials.
ceramic vol %
Part c) allows the use of thermosets be-
Figure 8-20. Shrinkage of alumina-polypropylene
mouldings as a function of initial ceramic volume
cause SiC requires a carbon residue as sin-
fraction during pyrolysis in air or nitrogen (Wright tering aid. Part d) reflects a strong bias to
etal., 1990 b). keep the water-based tradition in ceramics
manufacture. Part e) represents a logical
desire to make use of the organic vehicle as
was 65 vol.% which corresponds with the a ceramic precursor.
CPVC for the same powder grade obtained
by Markhoft etal. (1984). The restriction
on shrinkage of the samples pyrolysed in
air is attributed to the non-uniformity of 8.10 State-of-the-Art Summary
the reaction, the loss of binder from outer
layers restricting the subsequent shrinkage Processing science and technology are
of the body. The loss of binder also effects dynamic and a snapshot in time presents
a change of state from fluid, in which de- us at once with problems that are solved,
fects may appear as bubbles, to a quasi- problems that are understood but not
solid in which cracking predominates. This solved and problems that are neither un-
transition can be detected by thermome- derstood nor solved.
chanical measurements (Wright et al., Nevertheless, certain themes emerge.
1990 b). Principal is the problem of particle contact
before firing. It is inherent in the ceramic
moulding process that particles are held
8.9 Examples of Organic apart throughout the process until the
Materials Used in Ceramic binder removal stage. It has also been
Injection Moulding made clear that inter-particle distances are
permanently changing from the beginning
Table 8-5 presents a wide range of or- to the end of the process: particles are con-
ganic blends which are openly reported. stantly undergoing small relative displace-
These formulations have been developed ments even during storage mainly as a re-
by extensive trial and error and it is not sult of changes in the organic vehicle.
 8.10 State-of-the-Art Summary 305

Table 8-5. Examples of organic blends used for injection moulding.

Constituent Source

a) Compositions based on thermoplastic polymers:

Styrene and its copolymers Shung (1961); Pett et al. (1981); Muller-Zell et al. (1983); Sugano (1983);
Inoue et al. (1989); Maeda (1989 a); Sanpei et al. (1988)
Polyethylene Johnsson et al. (1983); Inoue et al. (1989); Bendix (1954);
Linguist et al. (1989); Arakarva et al. (1988); Seno et al. (1985);
Matkin et al. (1973); Inoue et al. (1987)
Polypropylene and atactic Hunt et al. (1988 a); Pujari (1989); Edirisinghe and Evans (1987 c);
polypropylene Saito et al. (1976); Inoue et al. (1989); Wiech (1986); Hunt et al. (1986);
Lee et al. (1989)
Polyethylene glycol Bailey (1966); Nagai et al. (1990)
Ethylene-vinyl acetate Thomas and Evans (1988); Inoue (1989); Mutsuddy (1990); Renlund and
copolymers Curtis (1986); Kato and Inoue (1990); Kumagai and Aoba (1988)
Polyoxymethylene and Ter Maat et al. (1991); Farrow and Conciatori (1984)
copolymers
Polyamides Maeda (1987); Sakai et al. (1986)

b) Compositions based on waxes or volatilizable organic compounds:

Paraffin wax Pujari (1989); Wiech (1981, 1983); Lee et al. (1989); Seno et al. (1985);
Bhadori et al. (1990); Bandyopadhyay and French (1986)
Microcrystalline wax Edirisinghe and Evans (1987 c); Kato and Inoue (1990);
Bhadori et al. (1990); Bandyopadhyay and French (1986); Mutsuddy (1987 a)
Montan wax Nagai et al. (1990)
Carnauba wax Wiech (1981 a, 1986); Lee et al. (1989); Inoue et al. (1989)
Candelilla wax Wiech (1983)
Beeswax Bendix (1954); Inoue et al. (1989); Yamazaki et al. (1985)
Miscellaneous waxes McLean and Fisher (1981)
Polypropylene wax Seno et al. (1985)
Napthalene or camphor Hermann (1966)

c) Compositions incorporating thermosetting resins:

Epoxy resin Strivens (1952)
Phenolformaldehyde Ohnsorg (1979)
Silane crosslinked polyethylene, Huether (1986)
phenolic resin

d) Compositions based on water or alcohol:

Cellulose ethers Rivers (1978); Wada et al. (1986); Maeda (1989 b); Chiba et al. (1987);
Kuwabara and Inoue (1988)
Agar Fanellietal. (1989)
2-Methylpropan-2-ol Edirisinghe et al. (1992 b)

e) Compositions wherein pyrolysis leaves a ceramic residue:

Silicone oil Burroughs and Thornton (1966); Newfield and Gac (1978)
Polycarbosilane Mutsuddy (1987b); Zhang and Evans (1991 b); Sugurawa (1986);
Okumura (1984)
Polyphenylsilsesquioxane Mine and Komazaki (1985)
306 8 Injection Moulding

Second is the magnitude of the empirical 8.12 References

problem. Stedman et al. (1990 b) calculates Aggarwala, B. D., Saibel, E. (1961), Phys. Chem.
that if an organic vehicle is composed of 9 Glasses 2, 137.
components drawn from a mere 19 in mul- Alford, N., McN., Button, T. W., Birchall, X D.
(1990), Supercond. Sci. Technol. 3, 1.
tiples of 10wt.% (10wt.% stearic acid Allan, P. S., Bevis, M. J. (1983), Plast. Rubber Proc.
fixed) there are 4496901 compositions to Appl. 3, 85.
be made up and tested! While some suc- Allan, P. S., Bevis, M. J. (1987), Plast. Rubber Proc.
cesses can be achieved empirically we are Appl 7, 3.
Allan, P. S., Bevis, M. J., Edirisinghe, M. X, Evans, X
getting much closer to the stage where the R. G., Hornsby, P. R. (1987), J. Mater. Sci. Lett. 6,
critical mass transport paths and kinetics 165.
of debinding can be quantitatively under- ANSI ASTM D281-131 (1974), Book of ASTM Stan-
dards. Philadelphia, PA: ASTM.
stood and predictions made. This will al- Arakarva, M., Hirota, K., Yamono, M. (1988), Nip-
low polymers to be deliberately synthe- pon Seramikkusa Kyoka Gakujutsu Ronbunshi 96,
sised to meet ceramic processing require- 942. CA109(20):175144w.
ASTM D-2663 (1975), Rubber Compounds - Disper-
ments. Thirdly, the restriction placed on sion of Carbon Black. Book of ASTM Standards,
polymer conformation by high energy sur- Part 37. Philadelphia, PA: ASTM.
faces means that there is now a recognition ASTM D864-52 (1978), Philadelphia, PA: ASTM.
Bailey, D. F. (1966), U.S. Patent 3 285 875.
of the problems of using high molecular Bandyopadhyay, G., French, K. W. (1986), ASME
weight organic materials in conjunction Tech. Paper 86-GT-ll.
with fine powders. The desire to use low Bao, Y, Evans, X R. G. (1991), /. Euro. Ceram. Soc.
8,95.
molecular weight species to simplify binder Barnes, H. A. (1989), J. Rheol. 33, 329.
removal, however, seems often to result in Barone, M. R., Ulicny, X C. (1990), J. Am. Ceram.
poor mixing and poor flow properties. For Soc. 73, 3323.
Barrer, R. M. (1968), in: Diffusion in Polymers:
this reason there is growing interest in the Crank, X, Park, G. S. (Eds.). London: Academic
processes of dispersion and the deploy- Press, pp. 165-216.
ment of dispersants for which much rele- Baston, A. H., Potton, X A., Twigg, M. V., Wright, C.
X (1981), /. Catalysis 71, 426.
vant knowledge can be transferred from Bell, S. H., Crowl, V. T. (1973), in: Dispersion of Pow-
the paint and printing ink industries. ders in Liquids, 2nd ed.: Parfitt, G. D. (Ed.). Lon-
don: Appl. Sci. Publ., pp. 267-273.
Benbow, X X, Bridgwater, X (1993), Paste Flow and
Extrusion. Oxford: Oxford University Press.
8.11 Acknowledgements Bendix Aviation Corporation (1954), U.K. Patent
706728.
Bhadori, S. B., Chakrabortz, A., Janardhana Reddy,
The author thanks his colleagues, past X (1990), /. Mater. Sci. Lett. 9, 209.
and present, in the Ceramics Fabrication Bierwagen, G. P., Saunders, T. E. (1974), Powder
Group at Brunei University for their con- Technol. 10, 111.
Bohlin, L. (1990), Plastics Compounding, Jan/Feb,
tributions to this subject. The continued pp. 32-37.
enthusiastic support of the U.K. Engineer- Brydson, X A. (1981), Flow Properties of Polymer
ing and Physical Sciences Research Coun- Melts, 2nd ed. London: Geo Goodwin, pp. 54-57.
Burroughs, X E., Thornton, H. R. (1966), Am. Cer-
cil is gratefully acknowledged. The author am. Soc. Bull. 45, 187.
thanks Mrs K. Goddard for typing the Calvert, P. D., Cima, M. X (1990), /. Am. Ceram. Soc.
manuscript. 73, 575.
Cass, R. B. (1991), Am. Ceram. Soc. Bull 70, 424.
Chiba, T, Nakamura, S., Hayakawa, K. (1987), Jpn.
Kokai Tokkyo Koho 62105948. CA107(10): 82694a.
Chong, X S., Christiansen, E. B., Baer, A. D. (1971),
/. Appl Polym. Sci. 15, 2007.
 8.12 References 307

Christiansen, R. M. (1979), Mechanics of Composite Fanelli, A. X, Silvers, R. D., Frei, W. S., Burlew, X V.,
Materials. New York: Wiley, p. 325. Marsh, G. B. (1989), J. Am. Ceram. Soc. 72, 1833.
Cogswell, F. N. (1981a), Polymer Melt Rheology. Farris, R. X 81968), Trans. Soc. Rheol. 12, 281.
London: Geo Goodwin, pp. 15-38. Farrow, G., Conciatori, A. B. (1984), European
Cogswell, F. N. (1981b), Polymer Melt Rheology. Patent 0114 746.
London: Geo Goodwin, pp. 135-136. Ferry, X D. (1970), Viscoelastic Properties of Poly-
Cogswell, P. N. (1975), Journee Calendrage Papier mers, 2nd ed. New York: Wiley, pp. 313-318.
d'une Reunion Group Frangaise d'Etudes et d'Appli- Francis, X L., Pardoe, P. G. A. (1970), in: Applied
cation des Polymeres, Section Lyonaise: Niviere, J. Science in the Casting of Metals: Strauss, K. (Ed.).
(Ed.), p. 6. Oxford, U.K.: Pergamon, pp. 467-486.
Cox, W. P., Merz, E. H. (1958), J. Polym. Sci. 28, 619. Gauthier, F , Goldsmith, H. L., Mason, S. G. (1971),
Coxon, L. D., White, J. R. (1980), Polym. Eng. Sci. Trans. Soc. Rheol. 15, 297.
20, 230. German, R. M. (1987), Powder Metall. 23, 237.
Crane, R. A., Vachon, R. I., Khader, M. S. (1977), German, R. M. (1990 a), Powder Injection Moulding.
Proc. 1st Symp. Thermophys. Prop., Geithersburg, Princeton, N.X: Metal Powder Ind. Fed., p. 163.
MA: ASME, pp. 109-123. German, R. M. (1990 b), Powder Injection Moulding.
Davies, G. J. (1973), Solidification and Casting. Lon- Princeton, N.X: Metal Powder Ind. Fed., pp. 311-
don: Appl. Sci. Publ., pp. 180-185. 313.
Dow, J. H., Sacks, M. D., Shenoy,A. V. (1988), Cer- Greener, X, Evans, X R. G. (1993), M. Mater. Sci. 28,
am. Trans., Vol. 1, Ceram. Powder Sci. 2, Part A, 6190
380. Gregg, S. X, Sing, K. S. W. (1982), Adsorption, Sur-
Duda, J. L., Vrentas, X S., Ju, S. T., Lin, H. T. (1982), face Area and Porosity. London: Academic Press,
AIChE J. 28, 279. p. 62 and p. 238.
Ebell, P. C , Hemsley, D. A. (1981), Rubber Chem. Gregory, X (1978), in: The Scientific Basis ofFloccula-
Technol. 54, 698. tion: Ives, K. X (Ed.). Alphen aan den Rijn, Nether-
Edirisinghe, M. J., Evans, J. R. G. (1986), Int. J. High lands: Sijthoff and Hoordhoff, pp. 101-130.
Technol. Ceram. 2, 1 and 249. Hale, D. K. (1976), J. Mater. Sci. 11, 2105.
Edirisinghe, M. J., Evans, J. R. G. (1987a), J. Mater. Hammond, P., Evans, X R. G. (1991), /. Mater. Sci.
Sci. 22, 269. Lett. 10, 294.
Edirisinghe, M. X, Evans, I R. G. (1987b), Br. Cer- Hammond, P. D., Evans, X R. G. (1995), J. Euro.
am. Trans. J. 86, 81. Ceram. Soc. 15, 117.
Edirisinghe, M. J., Evans, J. R. G. (1987 c), J. Mater. Hands, D., Horsfall, F. (1977), Rubber Chem. Tech-
Sci. 22, 2267. nol. 50, 253.
Edirisinghe, M. X, Evans, J. R. G. (1987d), Materials Hansen, R. H. (1970), in: Thermal Stability of Poly-
and Design 8, 284. mers, Vol. 1: Conley, R. T. (Ed.). New York:
Edirisinghe, M. J., Evans, J. R. G. (1988), Materials Marcel Dekker, pp. 153-187.
and Design 9, 85. Haunton, K. M., Wright, X K., Evans, X R. G. (1990),
Edirisinghe, M. J., Shaw, H. M., Tomkins, K. L. Br. Ceram. Trans. J. 89, 53.
(1992a), Ceram. Int. 18, 193. Hausner, H. (1993), Proc. 4th Int. Ceramitec Symp.
Edirisinghe, M. J., Tomkins, K. L., Patching, M. Dtsch. Keram. Ges., Wiesbaden, pp. 107-121.
(1992b), Ceram. Trans. 26, Forming Sci. Technol. Herrmann, E. R. (1966), U.S. Patent 3 234 308.
Ceram.: Cima, M. J. (Ed.). Westville, OH: Am. Hess, W. M. (1991), Rubber Chem. Technol. 64, 386.
Ceram. Soc, pp. 165-172. Hess, W. M., Garret, M. D. (1971), J. Oil Col. Chem.
Eilers, H. (1941), Kolloid Z. 97, 313. Assoc. 54, 24.
Einstein, A. (1956), Investigations on the Theory of Hess, W. M., Ban, L. L., McDonald, G. C. (1969),
Brownian Movement: Further, R. (Ed.). New York: Rubber Chem. Technol. 42, 1209.
Dover Publ., p. 55. Hildebrand, X H. (1977), in: Structure-Solubility Re-
Ess, J. W, Hornsby, P. R., Lin, S. Y, Bevis, M. J. lationships in Polymers: Harris, F. W, Seymour, R.
(1984), Plast. Rubber Proc. Appl. 4, 1. B. (Eds.). New York: Academic Press, pp. 1-10.
Evans, J. R. G. (1990), in: New Materials and their Huether, W (1986) WO Patient 8 605424.
Application: Holland, D. (Ed.). Bristol, U. K.; Inst. CA106(22) :177671c.
of Physics, pp. 25-32. Hunt, K. N., Evans, X R. G. (1991), J. Mater. Sci.
Evans, J. R. G. (1993), Proc. Int. 4th Ceramitec Lett. 10, 730.
Symp., Dtsch. Keram. Ges., Wiesbaden, pp. 8 1 - Hunt, K. N., Evans, X R. G., Woodthorpe, X (1988a),
106. Br. Ceram. Trans. J. 87, 17.
Evans, J. R. G., Edirisinghe, M. J. (1991), J. Mater. Hunt, K. N., Evans, X R. G., Woodthorpe, X (1988 b),
Sci. 26, 2081. Polym. Eng. Sci. 28, 1572.
Evans, J. R. G., Edirisinghe, M. X, Wright, X K., Hunt, K. N., Evans, X R. G., Woodthorpe, X (1991 a),
Crank, X (1991), Proc. Roy. Soc. A 432, 321. J. Mater. Sci. 26, 285.
308 8 Injection Moulding

Hunt, K. N., Evans, I R. G., Woodthorpe, J. (1991 b), Leigh-Dugmore, C. H. (1956), Rubber Chem. Tech-
/. Mater. ScL 26, 292. nol. 29, 1303.
Hunt, K. N., Evans, J. R. G., Woodthorpe, J. (1991 c), Leriche, A., Deletter, M. (1986), Proc. Adv. Ceram.,
/. Mater. ScL 26, 2143. Proc. Workshop Mtg.: Morgan, J. S. (Ed.). Madrid,
Hunt, K. N., Evans, I R. G., Woodthorpe, J. (1991 d), Spain: Soc. Esp. Ceram. Vidrio, pp. 157-167.
/. Mater. Sci. 26, 5229. Lewis, H. D., Goldman A. (1966), /. Am. Ceram. Soc.
Hyatt, I. S., Hyatt, J. W. (1872), U.S. Patent 133 229. 49, 323.
Inoue, M., Sakai, T. (1987), Jpn. Kokai Tokkyo, Ko- Lindquist, K., Carlstroem, E., Persson, M., Carlsson,
ho 62270458. CA108(12): 99977g. R. J. (1989), J. Am. Ceram. Soc. 72, 99.
Inoue, M., Kihora, Y, Arakida, Y (1989), Intercer- Maeda, K. (1987), Jpn. Kokai Tokkyo Koho
am. 38 (2), 53. 61006172. CA 104(18): 154481k.
Irving, H. K, Saxton, R. L. (1967), in: Theory and Maeda, K. (1989 a), Jpn. Kokai, Tokkyo Koho
Practice, Vol. 1: Uhl, V. W, Gray, J. B. (Eds.), New 01308862 (192). CA113(14): 119834s.
York: Academic Press, pp. 169-224. Maeda, K. (1989 b), Jpn. Kokai Tokkyo Koho
Israelachvili, J. N. (1991a), Intermolecular and Sur- 01249650 (A2). CA113(4):28211v.
face Forces, 2nd ed. London: Academic Press, Mangels, J. A. (1978), in: Ceramics in High Perfor-
p. 298. mance Applications II: Burke, J. X, Lenoe, E. N.,
Israelachvili, J. N. (1991b), Intermolecular and Sur- Katz, R. N. (Eds.). Massachusetts: Brook Hill
face Forces, 2nd ed. London: Academic Press, Publ., pp. 113-130.
p. 178. Markhoff, C. J, Mutsuddy, B. C , Lennon, X W.
Issitt, D. A., James, P. J. (1986), Powder Metall. 29, (1984), Advances in Ceramics, Vol. 9, Forming of
259. Ceramics: Mangels, X, Messing, G. L. (Eds.).
James, D. E. (1968), Polym. Eng. Sci. 8, 241. Columbus, OH: ACS, pp. 246-250.
Johnsson, A., Carlstrom, E., Hermansson, L., Maron, S. H., Pierce, P. E. (1956), J. Colloid Sci. 11,
Carlsson, R. (1983), Proc. Br. Ceram. Soc. 33, 139. 80.
Kandori, T, Kobugashi, S., Waila, S., Kamizaito, O. Masia, S., Calvert, P. D., Rhine, W. E., Bowen, H. K.
(1987), J. Mater. Sci. Lett. 6, 1356. (1989), / Mater. Sci. 24, 1907.
Katagiri, T. (1990), Powder Metall. Int. 22, 37. Matar, X A., Edirisinghe, M. X, Evans, X R. G., Twiz-
Katayama, K., Watanabe, T., Matoba, K., Katoh, N. ell, E. H. (1993), /. Mater. Res. 8, 617.
(1986), Society of Automotive Engineers Tech. Pa- Matar, S. A., Edirisinghe, M. X, Evans, X R. G.,
per 861 128. Twizell, E. H., (1995), /. Am. Ceram. Soc, in press.
Kato, S., Inoue, K. (1990), German Patent 3 942 745 Matkin, D. I., Denton, I. E., Valentine, T. M., War-
(Al). CA113(14): 119886k. rington, P. (1973), Proc. Br. Ceram. Soc. 22, 291.
Kendall, K. (1988), Powder Metall. 31, 28. Maxwell, X C. (1892), A Treatise on Electricity and
Kerner, E. H. (1956), Proc. Phys. Soc. B69, 802. Magnetism, Vol. 1, 3rd ed. Oxford, U.K.: Claren-
Kitano, T., Kataoka, T., Shirota, T. (1981), Rheol. don Press, p. 440.
Acta 20, 207. McLean, A. F , Fisher, E. A. (1981), Final Report
Kobayashi, K., Furuta, M., Maeno, Y. European AMMRC-TR-81-14. Watertown, MA: Army Ma-
Patent 0034056. terials and Mechanics Research Center.
Kostic, B., Zhang, T., Evans, J. R. G. (1992), /. Am. Menges, G., Koenig, D., Luettgens, R., Sarholz, R.,
Ceram. Soc. 75, 2773. Schuermann, E. (1980), Plastikverarbeiter 31, 185.
Kubat, X, Szalanczi, A. (1974), Polym. Eng. Sci. 14, Middleman, S. (1977), Fundamentals of Polymer Pro-
873. cessing. New York: McGraw-Hill, pp. 295-353.
Kubat, J., Petterman, J., Rigdhal, M. (1975), Mater. Milewski, X V. (1986), Adv. Ceram. Mat. 1, 36.
Sci. Eng. 19, 185. Mills, N. X (1982), /. Mater. Sci. 17, 558.
Kumagai, M., Aoba, T. (1988), Jpn. Kokai Tokkyo Mills, N. X (1983), Plast. Rubber Proc. Appl. 3, 181.
Koho 63239143. CAllO(lO): 81347h. Mine, T, Komazaki, S. (1985), Jpn. Kokai Tokkyo
Kuwabara, M., Inoue, M. (1988), Jpn. Kokai Tokkyo Koho 60210570. CA104(18): 154450z.
Koho 63265849 (A2). CA110(20): 178332n. Mohram, X T, Taylor, R. (1991), in: International
Landel, R. F., Moser, B. G., Bauman, A. J. (1965), Encyclopaedia of Composites, Vol. 5: Lee, S. M.
Proc. 4th Int. Conf. Rheol., Part, 2. (Ed.). New York: VCH, pp. 476-496.
Lee, E. H. (Ed.). New York: Interscience, pp. 663- Mooney, M. (1951), J. Colloid Sci. 6, 162.
692. Muller-Zell, A., Schwarzmeier, R. (1983), Keram. Z.
Lange, F. F. (1973), /. Am. Ceram. Soc. 56, 445. 55, 524.
Lange, F. F. (1984), /. Am. Ceram. Soc. 67, 83. Mutsuddy, B. C. (1983 a), J. Ind. Res. Dev. 25, 76.
Lange, F. F , Metcalf, M. (1983), /. Am. Ceram. Soc. Mutsuddy, B. C. (1983 b), Proc. Br. Ceram. Soc. 33,
66, 398. 117.
Lee, D. I. (1969), Trans. Soc. Rheol. 13, 273. Mutsuddy B. C. (1987 a), Adv. Ceram. Mater. 2,
Lee, D., Hens, K. F., Najmi, L., Fox, R. T. (1989), 213.
Adv. Powder Metall. 3, 1. Mutsuddy, B. C. (1987 b), Ceram. Int. 13, 41.
 8.12 References 309

Mutsuddy, B. C. (1989), Am. Ceram. Soc. Bull 68, Schofield, X D. (1990), in: Recent Developments in the
1796. Technology of Surfactants, Critical Reports on Ap-
Mutsuddy, B. C. (1990), Langmuir 6, 24. plied Chemistry, Vol. 30: Porter, M. R. (Ed.). Am-
Nagai, T., Yamanashi, H., Hachimori, H. (1990), sterdam: Elsevier, pp. 35-63.
U.S. Patent 4898 902. Schwartzwalder, K. (1949), Am. Ceram. Soc. Bull. 28,
Neil, J. T., Norris, D. A. (1988), ASME Tech. Paper 459.
88-GT-193, pp. 1-6. Seno, Y, Fujii, H., Narase, T. (1985), Aichi-ken Ko-
Newfield, S. E., Gac, F. D. (1978), Injection Mould- gyo, Gijutsi Sentra Hokuka 21, 25. CA105(20):
able Ceramics. Los Alamos Scientific Laboratory 177241s.
Report LA6960 UC25. Sheppard, L. M. (1991), Am. Ceram. Soc. Bull. 70,
Nogueira, R. E. F. Q., Edirisinghe, M. X, Gawne, D. 692.
T. (1992), J. Mater. Sci. 27, 6525. Sherman, P. (1970), Proc. 5th Int. Conf RheoL:
Ogbobe, O., Hemsley, D. A. (1989), Ind. J. TechnoL Onogi, S. (Ed.). Baltimore: University Park Press,
27, 409. pp. 327-338.
Ohnsorg, R. W. (1979), U.S. Patent 4144 207. Shimizu, Y, Tanemura, M., Sumita, M., Miyadea,
Okumura, M. (1984), German Patent 3 409385. N., Miyasaka, K. (1986), SEN-I Gakkaishi 42,
CA102(4): 30765v. T261.
Parkinson, C , Matsumoto, S., Sherman, P. (1970), /. Shung, M. C , Hitachi Corp. (1961), Japanese Patent
Coll. Interf. Sci. 33, 150. 7 883.
Patat, R, Killmann, E., Schliebener, C. (1966), Rub- Skinner, S. X, Taylor, W. (1960), Trans. J. Plast. Inst.
ber Chem. TechnoL 39, 36. 28, 237.
Payne, A. R. (1965), /. Appl. Polym. Sci. 9, 2273. Song, X, Evans, X R. G. (1993 a), /. Euro. Ceram. Soc.
Pett, R. A., Rao, V. D. N., Qaderi, S. B. A. (1981), 12, 467.
U.S. Patent 4265 794 Song, X, Evans, X R. G. (1993b), Process. Adv. Ma-
Pinwill, I. E., Ahmad, F , Allan, P. S., Bevis, M. J. ter. 3, 193.
(1992), Powder Metall. 35, 107. Song, X H., Evans, X R. G. (1994), J. Mater. Res. 9,
Priel, Z., Silberberg, A. (1978), J. Polym. Phys. 16, 2386.
917. Song, X H., Evans, X R. G. (1995), Ceram. Int., in
Prosser, J. (1985), Polym. Paint Col. J. 175, 390. press.
Pujari, V. K. (1988), Ceram. Trans. 1 (Ceram. Powder Spencer, R. S., Gilmore, G. D. (1949), J. Appl. Phys.
Sci. 2, Part B), 635. 20, 502.
Pujari, V. K. (1989), 1 Am. Ceram. Soc. 72, 1981. Spencer, R. S., Gilmore, G. D. (1950), J. Appl. Phys.
Rache, H. H., Felt, T. (1971), Dtsch. Verb. Material- 21, 523-526.
prufung 13, 37. Stangle, G. C , Aksay, I. A. (1990), Chem. Eng. Sci.
Rashid, H., Evans, J. R. G. (1991), Ceram. Int. 17, 45, 1719.
259. Stedman, S. X, Evans, X R. G., Woodthorpe, X
Rashid, H., Hunt, K. N., Evans, J. R. G. (1991 a), /. (1990a), J. Mater. Sci. 25, 1833.
Euro. Ceram. Soc. 8, 329. Stedman, S. X, Evans, X R. G., Woodthorpe, X
Rashid, H., Lindsay, K., Evans, X R. G. (1991b), J. (1990 b), Ceram. Int. 16, 107.
Euro. Ceram. Soc. 7, 165. Stedman, S. X, Evans, X R. G., Brook, R. X, Hoff-
Renlund, G. M., Curtis, A. (1986), U.S. Patent mann, M. X (1993), J. Euro. Ceram. Soc. 11, 523.
4 571414. Strivens, M. A. (1952), U.K. Patent 779242.
Ribeiro, C. A., Vargas, H., Galembeck, X (1981), Stromberg, R. R. (1967), in: Treatise on Adhesion and
Electron Microscopy 30, 148. Adhesives, Vol. 1: Patrick, R. L. (Ed.). New York:
Rivers, R. D. (1978), U.S. Patent 4113 480. Marcel Dekker, pp. 69-118.
Rumpf, H. (1962), Proc. Int. Symp. Agglomeration: Stromberg, R. R., Tutas, D. X, Passaglia, E. (1965), J.
Knepper, W. A. (Ed.). London: Interscience, Phys. Chem. 69, 3955.
p. 379. Sturges, R. F (1949), U.K. Patent 12480.
Rutherford, D. X, Simpson, L. A. (1985), /. Coatings Suetsuga, Y (1990), Int. Polym. Proc. V, 184.
Techol. 57, 75. Suetsuga, Y, Kikutani, T, Kya, T, White, X L.
Rwei, S. P., Manas-Zloczower, L, Feke, D. L. (1990), (1990), Colloid Polym. Sci. 268, 118.
Polym. Eng. Sci. 30, 701. Sugano, T. (1983), Proc. 1st Symp. Res. Dev. Basic
Saito, K., Tanaka, T., Hibino, T. (1976), U. K. Patent TechnoL Future Ind. Tokyo: Japan TechnoL Assoc,
1426 317. pp. 67-84.
Sakai, T., Kawabata, Y, Inoue, M., Kihara, Z. Suguwara, X (1986), Jpn. Kokai Tokkyo Koho
(1986), Jpn. Kokai Tokkyo, Koho 61006172. 61053149. CA104(26): 229430d.
CA104(18): 154481k. Sweeney, K. H., Geckler, R. D. (1954), /. Appl. Phys.
Sanpei, H., Aoba, T., Kumagi, M., Kohn, T. (1988), 25, 1135.
Jpn. Kokai Tokkyo Koho 53233038, CA110(10): Szekely, X, Evans, X W, Sohn, H. Y (1976), Gas-Solid
81335c. Reactions. New York: Academic Press, pp. 65-107.
310 8 Injection Moulding

Tadmor, Z., Gogos, C. G. (1979), Principles of Poly- Wildemuth, C. R., Williams, M. C. (1984), Rheol.
mer Processing. New York: Wiley, pp. 404-463. Acta 23, 627.
Tandon, R., Green, D. J. (1990), /. Am. Ceram. Soc. Willermett, P. A., Pett, R. A., Whalen, T. J. (1978),
73, 2628. Am. Ceram. Soc. Bull. 57, 744.
Tanford, C. (1967), Physical Chemistrys of Macro- Wilson, F. C. (1990), in: Proc. Poly. Based Mol. Com-
molecules. New York: Wiley, p. 150. pos. Symp., Vol. 171: Schaefer, D., Mark, J. E.
Taylor, D., Rutzler, J. E. (1958), Ind. Eng. Chem. 50, (Eds.). Pittsburgh, PA: Mater. Res. Soc, pp. 413-
928. 418.
Taylor, R. (1991), in: International Encyclopaedia of Wright, J. K., Evans, J. R. G. (1991 a), Ceram. Int. 17,
Composites, Vol. 5: Lee, S. M. (Ed.). New York: 79.
VCH, pp. 530-548. Wright, J. K., Evans, J. R. G. (1991b), /. Mater. 26,
Ter Maat, J. H. H., Ebenhoch, X, Sterzel, H. J. (1991), 4897.
4th Int. Symp. Ceram. Mater. Comp. for Engines, Wright, J. K., Evans, J. R. G., Edirisinghe, M. J.
Gotenberg, Sweden. (1989), /. Am. Ceram. Soc. 72, 1822.
Thomas, M. S., Evans, J. R. G. (1988), Br. Ceram. Wright, J. K., Thomson, R. M., Evans, J. R. G.
Trans. J. 87, 22. (1990 a), /. Mater. Sci. 25, 149.
Titomanlio, G., Drucato, V., Kamal, M. R. (1987), Wright, J. K., Edirisinghe, M. X, Zhang, X G., Evans,
Int. Polym. Proc. 1, 55, J. R. G. (1990b), J. Am. Ceram. Soc. 73, 2653.
Treuting, R. G., Read, W T. (1951), J. Appl. Phys. 22, Yamazaki, T, Hoshi, K., Torii, T. (1985), Aichi-ken
230. Tokoname Yogyo Gijutsu Senta Hokaku 12, 21.
Tsai, D. S. (1991), AIChE J. 37, 547. CA104(18): 154216c.
Tsao, I., Danforth, S. C. (1993), J. Am. Ceram. Soc. Zhang, T, Evans, J. R. G. (1989), /. Euro. Ceram. Soc.
76, 2977. 5, 165.
Tummala, R. R. (1988), Am. Ceram. Soc. Bull. 67, Zhang, T, Evans, I R. G. (1990a), J. Euro. Ceram.
752. Soc. 6, 15.
Tylecote, R. F. (1962), Metallurgy in Archaeology. Zhang, T., Evans, J. R. G. (1990 b), J. Mater. Sci. Lett.
London: Arnold, p. 107. 9, 673.
van Krevelen, D. W (1972 a), Properties of Polymers. Zhang, T, Evans, X R. G. (1991a), J. Euro. Ceram.
Amsterdam: Elsevier, pp. 67-79. Soc. 7, 155.
van Krevelen, D. W. (1972 b), Properties of Polymers. Zhang, T., Evans, X R. G. (1991b), /. Euro. Ceram.
Amsterdam: Elsevier, pp. 223-229. Soc. 7, 405.
van Krevelen, D. W. (1972 c), Properties of Polymers. Zhang, T., Evans, X R. G. (1992a), in: Processing
Amsterdam: Elsevier, pp. 55-65. Properties and Applications of Metallic and Ceram-
Vand, V. (1948), Phys. Chem. 52, 277. ic Materials: Loretto, M. H., Beevers, C. X (Eds.).
Vegvari, P. C , Hess, W M., Chirico, V. E. (1978), Birmingham, U.K.: MCE Publ., pp. 79-84.
Rubber Chem. Technol. 51, 817. Zhang, T, Evans, X R. G. (1992 b), J. Am. Ceram.
Vondracek, P., Vys, S. B. (1985), Sk. Chem.-Technol. Soc. 75, 2260.
Praze, Prague Poly. Chem. Wastnosti. Z. Pracov. Zhang, T, Evans, X R. G. (1993 a), J. Mater. Res. 8,
512,41. 187.
Wada, S., Masaki, H., Kamitori, T. (1986), Jpn. Zhang, T, Evans, X R. G. (1993 b), /. Am. Ceram.
Kokai Tokkyo Koho 61101447. CA105(12): Soc. 76, 481.
101643d. Zhang, T., Evans, X R. G. (1993 c), Br. Ceram. Trans.
Waterfield, C. G., Peacock, X (1973), Soc. Plast. Eng. J. 92, 146.
Tech. Paper 19, 393. Zhang, T, Evans, X R. G. (1993 d), M. Mater. Res. 8,
Wei, G. C , Becker, P. F. (1985), Bull. Am. Ceram. 345.
Soc. 64, 298. Zhang, T, Evans, X R. G. (1993 e), Ceram. Int. 19,
Weidenbaum, S. S. (1958), in: Advances in Chemical 367.
Engineering, Vol. 2: Drew, T. B., Hoopes, I W. Zhang, T., Evans, X R. G. (1993 f), J. Euro. Ceram.
(Eds.). New York: Academic Press, pp. 209-234. Soc. 12, 51.
Weir, F. E. (1963), SPE Trans. 3, 32. Zhang, X G., Edirisinghe, M. X, Evans, X R. G.
Weir, F. E., Doyle, M. E., Norton, G. D. (1963), SPE (1989a), J. Euro. Ceram. Soc. 5, 63.
Trans. 3, 37. Zhang, X G., Edirisinghe, M. X, Evans, X R. G.
Westman, A. E. R., Hugill, H. R. (1930), J. Am. Cer- (1989 b), Proc. Br. Ceram. Soc. 42, 91.
am. Soc. 13, 161. Zhang, X G., Edirisinghe, M. X, Evans, X R. G.
Wiech, R. E. (1980), U.S. Patent 4197 118. (1989 c), Ind. Ceram. 9, 72.
Wiech, R. E. (1981a), European Patent 0 032403. Zhang, X, Evans, X R. G., Dutta, K. K. (1989d), J.
Wiech, R. E. (1981b), European Patent 0032404. Euro. Ceram. Soc. 5, 303.
Wiech, R. E. (1983), U.S. Patent 4415 528. Zhang, T., Evans, X R. G., Rees, D. W. A. (1992),
Wiech, R. E. (1986), U.S. Patent 4 602953. Strain 28, 107.
 8.12 References 311

General Reading
Edirisinghe, M. J., Evans, J. R. G. (1986), "Fabrica- Mutsuddy, B. C , Ford, R. G. (1995), Ceramic Injec-
tion of Engineering Ceramics by Injection Mould- tion Moulding. London: Chapman & Hall.
ing"; Int. J. High Technol. Ceram. 2, 1-31, 249- Napper, D. H. (1983), Polymeric Stabilization of Col-
278. loidal Dispersions. London: Academic Press.
German, R. M. (1990), Powder Injection Moulding. Richerson, D. W. (1992), Modern Ceramic Engineer-
Princeton, NJ: Metal Powder Industries Federa- ing, 2nd ed. New York: Marcel Dekker.
tion. Rubin, I. I. (1973), Injection Moulding Theory and
Israelachvili, J. N. (1991), Intermolecular and Surface Practice. New York: Wiley.
Forces. London: Academic Press.
9 Single Crystals
Anthony L. Gentile
American Association for Crystal Growth, Thousand Oaks, CA, U.S.A.

Frank W. Ainger
Materials Research Laboratory, The Pennsylvania State University, University Park,
PA, U.S.A.

List of Symbols and Abbreviations 314

9.1 Introduction 315
9.2 Crystals and Crystallography 316
9.2.1 Ferroelectricity in Crystals 317
9.3 Growth of Single Crystals 319
9.4 Classification of Crystal Growth Techniques 320
9.5 Brief Fundamental Aspects for Selection of Crystal Growth Parameters . . . 321
9.6 Phase Equilibria and the Gibbs Phase Rule 322
9.6.1 Solid Solution Series 324
9.7 Crystal Growth Techniques 326
9.7.1 Growth from a Melt 326
9.7.1.1 Directional Solidification 326
9.7.1.2 Bridgman-Stockbarger Technique 326
9.7.1.3 Verneuil Flame Fusion Technique 327
9.7.1.4 Czochralski Technique 328
9.7.1.5 Skull Melting 329
9.7.1.6 Zoning Techniques 330
9.7.2 Indirect Crystal Growth - Growth from Solution 331
9.7.2.1 Top-Seeded Solution Growth 331
9.7.2.2 Hydrothermal Growth 333
9.7.2.3 The Sol-Gel Process and Its Derivatives 333
9.7.3 Crystal Growth from the Vapor Phase 334
9.7.3.1 Physical Vapor Deposition 334
9.7.3.2 Vapor Deposition of Single-Crystal Thin Films 334
9.7.3.3 Metal-Organic Chemical Vapor Deposition 334
9.7.3.4 Atomic Layer Epitaxy 336
9.7.4 Solid-to-Solid Crystal Growth 336
9.8 Distribution Coefficient and Mass Transport 337
9.9 Heat Transfer 339
9.10 Crystal Growth Theory Versus Experiment 340
9.11 References 340

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
314 9 Single Crystals

List of Symbols and Abbreviations

A, B component of a mixture
C number of compounds
E eutectic point
F degrees of freedom
k distribution or segregation coefficient
P number of phases present at equilibrium pressure
T temperature
Tc Curie temperature
AT change in temperature
AT/Ax thermal gradient
Ax distance in furnace
X composition

ACRT accelerated crucible rotation technique

ALE atomic layer epitaxy
BST barium strontium titanate
CVD chemical vapor deposition
EO electro-optic
KTN potassium (K) tantalate niobate
LEC liquid encapsulated Czochralski
LHPG laser heated pedestal growth
MBE molecular beam epitaxy
MOCVD metal-organic chemical vapor deposition
MOMBE metal-organic molecular beam epitaxy
OMVPE organo-metallic vapor phase epitaxy
PC polycrystal
PVD physical vapor deposition
PZT lead (Pb) zirconate titanate
rf radio frequency
TSSG top-seeded solution growth
YBCO yttrium barium copper oxide (superconductor)
 9.1 Introduction 315

9.1 Introduction These materials include simple oxides of

the type A / ) , , e.g., SiO 2 , A12O3, MgO,
A ceramic has been defined (NAS, 1968) TiO 2 , ZrO 2 , and complex structures of the
as an inorganic non-metallic material or formula type ABO 3 where B may be Nb,
article. Ceramics may be polycrystals, Ta, Ti, or Zr; and A may be Na, K, Ca, Ba,
glasses, or combinations thereof, or single Sr, Pb. Other refractory materials includ-
crystals. A single crystal is defined as the ing spinel structures such as MgAl 2 O 4 ,
macroscopic extension of a regular re- garnets (including aluminosilicates and
peated geometric network of atoms (called rare earth garnets), and other aluminates,
the crystal lattice), consisting of one or borates, silicates, and aluminosilicates as
more elements, from a microscopic scale to well as carbides, borides, and nitrides
a unit ingot (or film). Therefore, the bulk would fit into this chapter but would re-
single crystal (ingot) as well as the single quire extensive discussion to cover. Many
crystal thin film is expected to contain no of the growth techniques discussed are ap-
crystal grain boundaries and to have a con- plicable to these compounds.
tinuous, repeated symmetry of its constitu- In these materials, bonding occurs pri-
ent atoms. In contrast, a ceramic body marily between charged atoms or groups
may be seen as a consolidation of many of atoms, i.e., ionic bonding. Data on ionic
small single crystal grains, a polycrys- radii and coordination numbers are avail-
talline form, with the grains having ran- able. The coordination number represents
dom crystallographic orientation. Ceramic the number of surrounding cations around
Processing is defined (NAS, 1968) as a a positive metal ion or anion. It is believed
combination of science and engineering to be a major factor in the determination
that is directed initially toward developing of certain properties, e.g., ferroelectric and
-and ultimately of reliably manufacturing- nonlinear properties.
a ceramic product with specific desirable Polar materials are of great interest be-
properties tailored to its application. The cause of their nonlinear optical properties
processing equivalent for single crystals is (large nonlinear coefficients) leading to
the synthesis or growth procedure known efficient electro-optic modulation, sec-
as crystal growth. ond (or higher) harmonic generation, fre-
Many materials which have been quency conversion, and optical parametric
thought of as traditional ceramic materials oscillation, operating in wavelengths from
have extensive applications in single crys- the visible to near infrared. In these pro-
tal form. Some devices, e.g., lasers, pyro- cesses, energy exchange between a number
electric detectors, have been demonstrated of optical fields of different frequencies is
using both single crystal and ceramic brought about by a field-dependent dielec-
forms. However, the efficiency of the tric constant. Electro-optic modulation is
devices in most cases has been lower for based on manipulation of the optical prop-
the ceramic body than for the single crys- erties of crystals by means of external elec-
tal, lower even than predicted by calcula- tric fields. In optical parametric oscilla-
tions taking into account differences based tion, a fixed frequency pump wave is con-
on anisotropic properties. verted to a higher frequency signal and a
The materials which we will stress in this lower frequency idler wave. Phase match-
chapter are the refractory (high melting ing conditions are altered by changes in
point) oxides, both simple and complex. temperature or orientation (angular posi-
316 9 Single Crystals

tion) of the crystal which allows tuning of

the signal or idler wavelengths. It is not
surprising then that the demand for optical
quality (striae free) crystals applicable to
electro-optic devices has also advanced the
techniques for the growth of oxide single
crystals.
Some aspects of crystal growth are also
treated in Vol. 16 (Chap. 2, Sec. 2.7) of this
Series.

9.2 Crystals and Crystallography

From a crystallographic point of view,
the spatial arrangement of atoms in a crys- A CATION
o
OXYGEN
o
B CATION
tal forms a unit cell which is defined as the Figure 9-1. Perovskite (ABO3) structure.
smallest crystallographic "repeat" unit,
i.e., a sufficient number of atoms are re-
peated to define the 3-dimensional rela- An extended discussion of crystallography
tionship of the atoms throughout the crys- appears in Chap. 1, Vol. 1, of this Series.
tal when the unit cell is reproduced by sim- For many electronic and optical appli-
ple translations in the same orientation. cations, one can select useful crystals based
The unit cell for perovskites which repre- on their chemical and structural makeup.
sent an important family of ferroelectric The linear electro-optic (EO) effect is ob-
structures is shown in Fig. 9-1. The name served in crystals which lack inversion
is derived from the mineral perovskite symmetry (noncentrosymmetric); in crys-
CaTiO 3 which is isostructural with other tals which possess inversion symmetry, the
ABO3 crystals where A = Ca, Ba, Sr, K, linear electro-optic effect vanishes (Yariv,
Na, Li, and B = Nb, Ta, Ti, Zr. The unit 1967) and the dominant term describing
cell of a cubic perovskite shows a metal ion the dependence of the dielectric tensor on
B (= Ca) at the center of the basic cube, the electric field is a quadratic term. Some
and an anion X ( = O) at the center of the of the best 'quadratic' EO materials are
cube faces, yielding an octahedron with the ferroelectric when operated above their
metal ion (B) in its center, the BX6 octahe- Curie temperature, i.e., in a paraelectric
dra. phase. This indicates that changes in the
The unit cell usually is not the simple optical index are linked more fundamen-
chemical formula which designates the tally to the induced polarization than to
overall ratios of its constituent elements. the applied electric field. In this region,
The 3-dimensional extension of a unit cell susceptibility is temperature dependent.
in an infinite pattern produces a single As illustrated in Fig. 9-2 for the longitudi-
crystal. The unit cell defines the basic sym- nal electro-optic effect, specific point
metry of the crystal which classifies the groups will contain crystal compounds
crystal into one of seven crystal systems. having longitudinal (modulator field and
 9.2 Crystals and Crystallography 317

CRYSTAL SYMMETRY
SYSTEM POINT GROUP

CUBIC 23, M3, 432, 43M, M3M

HEXAGONAL 6. 1 6, 6/M, 1622J 6MM, 6M2, 6/MMM

TRIGONAL
0 3, ( 32, ) 3M, 3M

TETRAGONAL 4, ) ( 4, ) 4/M, (422,) 4MM, 142M,) 4/MMM

ORTHORHOMBIC (222,1 MM2, MMM

MONOCLINIC 2, ) M, 2/M

TRICLINIC
0
Figure 9-2. Point groups of longitudinal electro-optic crystals.

light propagation in same direction) or ences in properties compared to its chemi-

transverse (field normal to direction of cally-similar ceramic body, especially re-
light propagation) electro-optical effects. garding optical, optoelectronic, and ferro-
Although we can select EO crystals from electric properties.
these groups, rarely can we predict the
magnitude of the effect based on present-
9.2.1 Ferroelectricity in Crystals
day knowledge. However, in the case of
EO coefficients, theoretical calculations A crystal is ferroelectric if it demon-
based on structural properties (especially strates reversible spontaneous polarization
bond length and angle) of the electro-optic (Megaw, 1957). A ferroelectric is a py-
coefficient made by Shih and Yariv (1980, roelectric with a reversible spontaneous
1982) for binary compounds including polarization which goes to zero at temper-
GaAs and CdTe have compared favorably atures higher than Tc, the Curie tempera-
with measured results. The calculations for ture. At Tc the polar phase becomes non-
compounds of the type ABO 3 have not polar, i.e. centrosymmetric, and is repre-
been as accurate. Additional theoretical sented by a dielectric anomaly whereby the
and experimental work is required for ulti- dielectric constant reaches a maximum
mate device advancement in many fields value and then decreases above Tc accord-
including the synthesis and characteriza- ing to the Curie-Weiss law. When a ferro-
tion of single crystal materials. The conti- electric crystal is cooled through Tc, it
nuity of the internal structure of a single spontaneously polarizes into domains, the
crystal explains the many observed differ- directions of which are dictated by the
318 9 Single Crystals

space group and crystal structure. In order sitions as well as nonlinear properties. The
to minimize the internal strain, the do- movement of the B ion in its oxygen cage
mains adopt those polar crystallographic constitutes a simple microscopic descrip-
structures which lead to this condition. tion of ferroelectricity in perovskite com-
Therefore in order to obtain a single do- pounds.
main crystal either an electrical or mechan- The sequence of ferroelectric transi-
ical stress is applied in the required polar tions:
direction on cooling through Tc.
cubic -• tetragonal -* orthorhombic
There are related materials which con-
-> rhombohedral
tain aligned dipoles on the lattice dimen-
sion but are non-polar owing to their com- such as observed in BaTiO3 and KNbO 3 ,
pensating antiparallel arrangement. These is readily understood if one pictures the B
are known as antiferroelectrics and fre- ion (Nb 5 + , Ti 4 + ) being successively dis-
quently have free energies near to those of placed along (100), then (110), and finally
a polar form. Some antiferroelectrics may (111). This mechanism is typical of a dis-
be transposed into a polar form by stress, placive ferroelectric transition. A distinc-
electric field or chemical modification- tion between displacive and order-disorder
enforced ferroelectricity; lead zirconate ferroelectrics can be made. The order-dis-
PbZrO 3 is a prime example. order type is characterized by the existence
In ABO 3 compounds, where B = Ti, Nb, of permanent dipoles primarily in the fer-
six-fold coordination is typical, consisting roelectric phase (Rytz, 1983). Thermal agi-
of an octahedral arrangement of oxygen tation competes with the dipole-dipole in-
atoms around niobium or titanium atom teraction and, at the Curie temperature,
as shown in Fig. 9-3. The shape or distor- the alignment of the dipoles (perfect at ab-
tion of this atomic arrangement relative to solute zero) disappears. In the paraelectric
the position of the B atom, is considered to phase, dipole orientation is disordered. In
have an essential role in ferroelectric tran- displacive ferroelectrics, dipoles induced in
the ferroelectric phase disappear in the
paraelectric phase via a displacement of
LiNbO
atoms towards a position of higher sym-
metry.
Recent work (Muller, 1981; Miiller
Nb etal., 1982; Burns and Dacol, 1982) indi-
cates that a ferroelectric transition demon-
strates simultaneously both displacive and
order-disorder characteristics, with one
dominating but not excluding the other.
This is why BaTiO3 is not a simple exam-
ple and points to the need for further in-
vestigations of polycrystalline and single
crystal ferroelectrics.
Many oxide ferroelectrics exhibit the
perovskite structure and possess high
spontaneous polarization and dielectric
Figure 9-3. Niobium (Nb) octahedron. constants which determine their pyroelec-
 9.3 Growth of Single Crystals 319

trie, piezoelectric and electro-optic proper- linked with chemical synthesis; numerous
ties. Although the polycrystalline ceramic methods are available for the formation of
form has been successfully commercialized single crystals.
for components such as capacitors, ther- Crystal growth occurs through a con-
mal detectors, and piezoelectric elements trolled accumulation of atoms or ions or
because of its low-cost, high-volume pro- molecules around a unique nucleus. The
duction technology, there exist important nucleus or nucleation site may be self-in-
applications for ferroelectric single-do- duced or introduced as a seed. The meth-
main crystals where high coefficients and/ ods of single crystal growth involve a con-
or optical quality are deemed necessary. trolled change of state, or phase change, to
Such crystals may be either in bulk or thin the solid (condensed) state. This transition
film form according to the application. may occur from the vapor, liquid, or, in
certain cases, within the solid state itself.
The growth of crystals occurs through the
9.3 Growth of Single Crystals application of many disciplines within
chemistry and physics including thermo-
The process of forming a single crystal is dynamics, kinetics, and fluid dynamics.
referred to as crystal growth. The process- Essentially, the crystal grower uses the in-
ing of ceramic oxides is done by chemical formation available in these fields to de-
synthesis and two examples of ways in velop techniques to react components to
which polycrystalline barium titanate is form a single compound, to create a
made are given below: physico-chemical environment for con-
trolled nucleation, and to apply (or re-
Solid state reaction: move) heat to transform the compound to
a condition wherein programmed heat
transfer will yield a condensed state accru-
Chemical precipitation from solution: ing on the nucleation site to form a single
Ba(NO 3 ) 2 + H 2 [TiO (gg)] • «H 2 O -> (9-2) crystal.
The emphasis in this chapter is primarily
in solution on the practical aspects of single crystal
growth as related to the synthesis of refrac-
tory oxide crystals and thin films. Crystal
2HNO 3 (9-3) growth theory, although rapidly develop-
(Precipitate) ing today, is also changing, and is some-
BaTiO(ggg)2 ^ what limited in its applicability to the prac-
tice of crystal growth. Most theoretical
+ H2T (9-4)
models are successful in two dimensions
The solid-state reaction [Eq. (9-1)] is com- but do not carry over as well to the third
monly used for commercial ceramics be- dimension.
cause it is inexpensive and facilitates the Before we get into a detailed discussion
inclusion of numerous other oxides by sub- of crystal growth, we present a classifica-
stitution or doping, whilst the chemical tion of crystal growth techniques with ex-
precipitation [Eq. (9-2)to (9-4)] provides a amples from a variety of materials in order
high-purity form which can be used in the to give the reader an overall view of crystal
growth of single crystals. Crystal growth is growth. Following that, we discuss growth
320 9 Single Crystals

parameters, phase equilibria, and then illustrated for specific materials often rep-
present specific examples of some of the resentative of a group chosen from the
common growth techniques in use today to most popular among those in practice. The
produce single crystals of refractory oxides selection of a growth technique is dis-
and mixed crystal species which have inter- cussed in terms of phase equilibria. The
esting properties applicable to devices in description of crystal growth techniques is
use in many modern technologies. followed by a brief introduction to the
roles of energy and mass transport.
The various chemical thermodynamic
9.4 Classification of Crystal and kinetic influences are pointed out as
Growth Techniques we develop the determination of the
parameters of crystal growth. A short sec-
The classification of crystal growth tech- tion points out some recent work which
niques presented here (Table 9-1) is based has shown remarkable correlation between
on that developed by Laudise (1967,1970). crystal growth theory and experiment. Sin-
The various crystal growth approaches are gle crystals may be grown by many tech-

Table 9-1. Classification of crystal growth techniques (Gentile, 1983).

Phase Direct Indirect

Source Growth technique Source Growth technique

Liquid Melt3 Verneuil Flux-solution Slow cooling

Pedestal (solid solution) Temp, differential
Directional solidification Solvent evaporation
Bridgman Top seeded (isothermal)
Stockberger Hydrothermal
Cooled seed Any solvent: aqueous,
Kyropoulos molten flux, etc.
Pulling
Czochralski
Zoning techniques Reaction Chemical, electrochemical
Float zone
LHPG
Skull melting
Vapor Constituent gas Sublimation - condensation Compound gas Reaction - condensation
(Epitaxy) PVD b , MBE, ALE CVD, MOCVD,
(MOVPE), etc.
Solid Solid Recrystallization Solid solution Exsolution
Strain Spinodal decomposition
Polycrystal
a
Melt may be off stoichiometry but direct if composition is same as growing crystal; b PVD, physical vapor
deposition; MBE, molecular beam epitaxy; ALE, atomic layer epitaxy; CVD, chemical vapor deposition;
MOCVD, metal-organic chemical vapor deposition; MOVPE, metal-organic vapor phase epitaxy.
Reproduced from Tutorial Lectures in Electrochemical Engineering and Technology II, AIChE Symposium Series,
Vol. 19, No. 229 (1983) page 144 by permission of the American Institute of Chemical Engineers, © 1983, AIChE,
all rights reserved.
 9.5 Brief Fundamental Aspects for Selection of Crystal Growth Parameters 321

niques, some of which will be discussed Indirect techniques, like Thurmond's

comprehensively in this article. "nonconservative" system, involve the use
Crystal growth techniques are generally of either additional foreign components or
named after the first person to describe excesses of constituent components which
them (e.g., Czochralski, Bridgman) or a act as mineralizers, fluxes, solvents, or car-
descriptive phrase of the process, e.g., riers of the constituents. These indirect
physical vapor deposition (PVD), chemi- techniques include ways of growing incon-
cal vapor deposition (CVD). The classifi- gruently melting crystals which decompose
cation shown in Table 9-1 (Gentile, 1983) upon melting to form a liquid and another
is a slightly modified version of the classifi- solid of different compositions. Indirect
cation first presented by Laudise (1967, techniques using solvents (fluxes or miner-
1970) with some of the ideas of Thurmond alizers) may be employed to grow crystals
(1959). The principal difference is that which can also be grown directly. Solvent
crystal growth techniques have been sepa- or fluxes are frequently utilized to lower
rated initially into two categories: direct the growth temperature to more practical
and indirect. However, the techniques working levels which, in addition, may
have not been separated into single and avoid subsolidus (solid-solid) phase tran-
multicomponent types. The direct tech- sitions. The phases or states of matter (liq-
niques are those which do not contain any uid, vapor, and solid) which are changed in
components other than those of the de- order to obtain single crystals represent the
sired crystal product. This accords with phases present with the solid single crystal
Thurmond's "conservative" system which at the interface in both direct and indirect
is similarly defined "when nothing is categories. The materials from which these
added to or taken from the melt except by phases are generated are referred to as
the freezing process". The phase from source materials.
which the crystal is growing, whether it is Off-stoichiometric melts, i.e., melts
vapor, liquid, or solid, has the same overall whose composition vary slightly from the
constitution (excluding trace impurities precise chemical formula of the com-
whether added for doping purposes or in- pound, are included here as direct tech-
advertently present) as the growing crystal. niques. Crystals such as LiNbO 3 and
A common and simple example is the melt LiTaO3 are grown from melt compositions
growth of single crystals of sapphire which vary from formula but yield solid
(a-Al2O3) from a molten pool of A12O3 crystals of the same composition as the
which may (or may not) contain trace melt. Frequently, these compounds will
amounts of a "dopant" metal added in or- demonstrate off-stoichiometric maximum
der to modify electronic and/or optical melting compositions as well as an "exis-
properties, such as the addition of Cr 3 + tence region" wherein the solid compound
to sapphire for the fabrication of ruby la- exists with variations in stoichiometry.
ser crystals. The A12O3 system is consid-
ered simple because it contains one metal- 9.5 Brief Fundamental Aspects
lic element, melts congruently to form a
liquid melt of the same composition and
for Selection of Crystal Growth
has no solid-solid phase transitions from Parameters
room temperature to its melting point at The selection of a specific technique for
2050 °C. crystal growth and the source materials to
322 9 Single Crystals

be used is based on the physical and chem- pound may also be grown indirectly within
ical properties of the material to be grown, the phase field where it is in equilibrium (as
including melting point, component vapor a solid) with a liquid in a temperature
pressures, and constituent reactivity. The range lying below its melting point, but
final equilibrium state is that having the above TE. The selection of this range de-
lowest free energy. Therefore, the thermo- pends upon the selection of the starting
dynamics of the reaction(s) must be taken composition. However, below TE, a solid
into consideration and frequently may be mixture of A and B coexist. In order to
used to calculate the first usable form of grow a single crystal of one compound (A)
pertinent data to the crystal grower: the from solution, the crystal must be sepa-
phase equilibrium diagram. Although rated from the remaining melt before TE is
crystal growth is not an equilibrium but reached. A pertinent method to grow sin-
rather a steady state process, the informa- gle crystal A could be the Czochralski
tion required to engineer a crystal growth technique. Selection of crystal growth
process appears in the phase equilibrium parameters and the allowed variations of
diagram. A simple binary phase diagram parameters (degrees of freedom) are deter-
(Fig. 9-4) illustrates the temperature-com- mined from the phase equilibrium diagram
position (T-X) relationships for two com- and governed by the Gibbs phase rule (see
ponents which may be compounds A and also Vol. 5, Chap. 1, Sec. 1.5.12, of this
B. From the diagram, it appears that both Series).
compounds melt without decomposition
(congruently) at temperatures TA and TB,
respectively, and thus could be grown by 9.6 Phase Equilibria and the
direct techniques from the melt. In addi- Gibbs Phase Rule
tion, this diagram illustrates a simple
eutectic point, E\ i.e., a minimum melting The Gibbs phase rule can be written as
mixture of the constituents at a tempera-
ture TE<TA, TB. Therefore, either com- P+F=C+2 (9-5)
where P = number of phases present at
equilibrium; F= degrees of freedom of sys-
tem (temperature, pressure, composition);
and C=number of components of system.
LIQUID
The rule can be applied to all phase equi-
librium diagrams to ascertain information
concerning the phase relations and the re-
B
• LIQUID maining degrees of freedom which repre-
sent the control parameters of crystal
EUTECTIC growth. In practice, the Gibbs phase rule is
SOLID A + B frequently used in its reduced form derived
by the removal of pressure as a variable. It
can then be written as
P + F=c+1. (9-6)
A X • B Applying the reduced version of the
Figure 9-4. Phase diagram for binary systems. phase rule, the eutectic point E in a binary
 9.6 Phase Equilibria and the Gibbs Phase Rule 323

(two-component) phase equilibrium dia- growth apparatus leading to an explosion;

gram (Fig. 9-4) is determined to be an in- this is not usually a problem in refractory
variant point, i.e., there are no degrees of materials. In other cases, control of con-
freedom since three phases are present: stituent overpressure may be required to
two solid phases (A and B) and a liquid achieve the desired phase, compound or
phase: stoichiometry.
Many of the systems of interest to the
P = 3 (solid A, solid B, liquid);
crystal grower today are indirect ones
C = 2(A,B). which include crystals which melt incon-
gruently. A phase equilibrium diagram of
Thus, a system containing an incongruently
F=0. melting compound is illustrated in Fig. 9-5
(Reisman, 1970). The compound A^B^ in
This is obvious from the diagram itself, the diagram melts at temperature Tt to
since at the point E both the composition form a solid of composition A in equilib-
and temperature (the degrees of freedom in rium with a liquid (melt) of composition i.
a binary system) are fixed (invariant). In order to grow compound A^B^ from the
However, along the curve E-TB there exist melt, an indirect approach is required uti-
only 2 phases (solid B plus liquid), which lizing a melt in the phase field "A^B^
leaves one degree of freedom. The selec- (solid) 4- liquid" under "i-e," liquidus line
tion of either a temperature within the section.
range from TB to TE or a composition A complex phase diagram such as the
within the range from E to B fixes a point system BaO-TiO 2 (Fig. 9-6) (Kirby and
on the curve. In the phase field "B + liq- Wechsler, 1991; Wechsler and Kirby, 1992)
uid", the crystal grower may select a tem-
perature and composition with some vari-
ability to optimize, for example, the indi-
rect growth of solid B. The T-X diagram 'A

is a result of the system thermodynamics in Liquid

that it indicates the result of the equilib- A + Lick

rium reaction for a given temperature and

a specified mixture (composition) of the
constituents. The phase diagram, however, AxBy + L i q \ / B + Liq
tells us nothing about the kinetics of the
reaction which includes the rates of the
A + AxBy
reaction, or the formation of metastable
(intermediate) phases. In addition, the typ- AxBy + B
ical binary phase diagram represents an
isobaric plane which may not be a true
representation of reality for systems con-
taining components with significant vapor B
pressures. In many crystal growth situa-
Figure 9-5. Phase diagram for incongruently melting
tions, the pressures generated at the indi-
compound. Reprinted with permission from Reisman
cated temperature and composition may (1970). © 1970, Academic Press, Orlando, FL; ©
be beyond the capability of the crystal 1990, MCNC, Research Triangle Park, NC.
324 9 S i n g l e Crystals

1660
LIQUID
1620
o ALL LIQUID HEXAGONAL BaTiO 3 S.S.
gi580 + LIQUID
• SOLID + LIQUID PHASES LU
HEX
• TWO SOLID PHASES I BaTiCs
§1540
a ONE SOLID PHASE S.S. HEX + CUBIC S.S.
£ 1500
3 1460 CUBIC BaTiO3 S. S.
• CUBIC
+ LIQUID
1420 BaTiO3 S.S.
I
1380 53

Figure 9-6. Phase diagram for the system BaTiO 3 -TiO 2 . From Kirby and Wechsler (1991) < 1991. Reprinted
by permission of the American Ceramic Society.

shows the formation of many stoichiomet- ist, from one pure end member to the
ric compounds and illustrates many fea- other, which is called a solid solution
tures including incongruent melting. The series. A series of the general formula
diagram is used here to delimit the phase KTa:cNb1_JCO3 where 0 < x < l exists for
field wherein conditions exist for the all values of x in the designated range. This
growth of cubic BaTiO3 from an off-stoi- relationship can be shown as a phase equi-
chiometric melt, i.e., one rich in TiO 2 . librium diagram (Fig. 9-7): a plot of tem-
perature vs. mol% KTaO 3 in the liquid-
solid thermal range (Fig. 9-7 a), and the
9.6.1 Solid Solution Series subsolidus phase equilibrium (Fig. 9-7 b).
In isostructural compounds such as The essential difference (see Fig. 9-7 a)
BaTiO 3 -SrTiO 3 , KTaO 3 -KNbO 3 , from most phase equilibrium diagrams is
PbZrO 3 -PbTiO 3 , a complete continuous the constantly varying compositions of the
series of quaternary compounds may ex- liquid (shown by the liquidus curve) and
 9.6 Phase Equilibria and the Gibbs Phase Rule 325

IHVJU
I I I I I I I I the solid (solidus) with changes in temper-
1350 - ature. The region between the curves rep-
LIQUID
1300 resents solid in equilibrium with liquid.
1250 The area above the liquidus is entirely liq-
1200 uid; and that below the solidus is all solid.
S^ SSAND
S^ LIQUID In Fig. 9-7 b, we observe that the first-to-
1150
SOLID form solid is cubic. The transition
1100 SOLUTION _
boundary from cubic to tetragonal repre-
1050( - sents the Curie temperature which varies
1000 - a _ with composition (value of x). In this sys-
500 - b - tem, this is a nondestructive transition into
400
the ferroelectric phase.
^SA CUBIC
In the 'KTN' system shown in Fig. 9-8,
300
~~ \ /-TETRAGONAL ~
a melt composition at A (approximate-
200
^Sw O \ CURIE TEMPERATURE ~
c
ly KTa0 3 7 Nb 0 63 O 3 ) is entirely liquid
_
100 above 1200°C. As the temperature is
ORTHORHOMBIC^p^.^^
0 lowered to cross the liquidus curve, the
-100 -RHOMBOHEDRAL^ ^' ^ ^ ^ first-to-freeze composition (approximately
-200 I I I I I I I I KTa0 6 6 Nb 0 34 O 3 ) is designated by a' on
0 10 20 30 40 50 60 70 80 90 100
the solidus curve; the tie line a-a' shows
KNbO3 Mol % KTaOg KTaO3
the liquid-solid equilibrium at tempera-
Figure 9-7. KTN phase diagrams, (a) Liquidus-soli- ture r=1200°C. As the temperature is
dus relations; (b) subsolidus relations. lowered, the composition of the liquid
varies along the liquidus curve, and that of
the solid along the solidus curve. For a
starting composition A, solidification is
complete when the vertical extension of A

Figure 9-8. Phase diagram for the system

1000 KNbO 3 -KTaO 3 .
10 20 30
COMPOSITION MOLE % KTaO Q
326 9 Single Crystals

hits the solidus (x'); theoretically, the last The selected growth technique may re-
drop of liquid has the composition x. The quire an ampoule or container, or one may
solid obtained by a slow cooling procedure choose a "containerless" technique de-
varies continuously in composition. fined here as a method wherein the source
In order to obtain a crystal or solid of material does not come into contact with a
composition a' at a working temperature container in the region of the growth inter-
of 1200 °C, the selected melt composition face. Container selection is influenced by
must lie to the right of A, i.e., must be temperature and reactivity of the constitu-
richer in Ta. This is shown for a composi- ent materials. In many cases, the furnace
tion at B (KTa0 4 3 Nb 0 57 O 3 ). The intersec- may be used in either a vertical or horizon-
tion of B with the isothermal temperature tal position so long as the growth parame-
(1200 °C) line extension designates the pro- ters may be stabilized and controlled.
portional amounts of solid (S) and liquid
(L) at that temperature. Starting with com- 9.7.1.1 Directional Solidification
position B, precipitation of solid material
Directional solidification involves cool-
commences as the temperature is lowered
ing through the melting point of a liquid
just below the liquidus curve, approxi-
melt in an ampoule. The ampoule is fre-
mately 1220 °C. As discussed above for A,
quently conically shaped at the bottom tip
the first solid to crystallize lies at the inter-
to achieve nucleation of a single crystal.
section of a horizontal line (the 1220 °C
This method is often used where no seed of
isotherm) with the solidus curve. As cool-
the desired single crystal is available.
ing continues, the compositions of both
When a seed is available, it may be encap-
the liquid and solid continuously change
sulated into the base of the ampoule for
along their respective curves. This relation-
"seeded" growth. Solidification continues
ship poses a significant problem for the
from the tip to the top of the ampoule by
crystal grower seeking a uniform single
cooling and directing heat transfer or re-
crystal and requires a technique which
moval of heat through the growing crystal.
varies from the standard cooldown proce-
Bridgman (1925) enhanced the direct so-
dure. Techniques have been developed to
lidification technique by invoking a ther-
overcome this problem and are discussed
mal gradient profile initially accomplished
in Sec. 9.7.2.1.
by having the center of the furnace at max-
imum temperature and then allowing the
9.7 Crystal Growth Techniques natural cooling toward the ends to create
the gradient. Thermal techniques have
9.7.1 Growth from a Melt been applied to control growth gradients
in both contained and containerless tech-
Direct crystal growth from a liquid melt
niques.
to a solid involves the solidification by a
controlled pass through the melting point
9.7.1.2 Bridgman-Stockbarger Technique
of the compound. Typically, the source
material is heated to melting and then In the Bridgman technique, the furnace
cooled slowly by either lowering an am- temperature is kept constant and the am-
poule through a suitable thermal gradient poule is lowered at a controlled rate
or lowering furnace temperature (ramp- through a preset thermal profile. Stock-
ing). barger's (1936) interest in growing large
 9.7 Crystal Growth Techniques 327

crystals (several inches in diameter) of al- CONICALLY

TIPPED HEATER
kali halides led him to take more drastic CRUCIBLE
steps to remove heat and create a steeper
thermal gradient (larger AT/Ax, where AT
is the change of temperature with distance
in the furnace, Ax). He added baffles inside
the furnace to assist in cooling the desired
BAFFLE —
region. Today many variations of these
methods are in practice and different ways
are used to vary thermal gradients. A typ-
ical vertical crystal growth method using
an ampoule and cooling in a modified-
steep thermal gradient is commonly called
Bridgman-Stockbarger (Fig. 9-9). Natu- CRUCIBLE
ral steep thermal gradients are typical of SUPPORT
some of the containerless techniques be- LOWERING
MECHANISM
cause of their geometric configuration.
Figure 9-9. Bridgman-Stockbarger growth appa-
ratus. Reprinted with permission from Academic
9.7.1.3 Verneuil Flame Fusion Technique Press, Inc., Orlando, FL, from Gentile (1987, 1992),
© 1987, page 4; © 1992, page 712.
The growth of crystals from the melt is
most easily achieved by the Bridgman-
Stockbarger technique. However, the
properties of refractory oxide materials
rule out the use of conventional crucible TAPPER
POWDER
(container) materials, e.g., fused silica, be-
cause of their high melting points and sub-
sequent reactivity. Several approaches can
be used to bypass these problems essen-
tially by employing "containerless" tech-
niques. The "flame fusion" or Verneuil
technique (Verneuil, 1902) has been used
extensively for the synthesis of sapphire,
ruby, rutile (TiO2), and spinel (MgAl2O4).
The Verneuil technique represents one FURNACE

of the early direct crystal growth tech-

niques from the melt. Growth takes place CRYSTAL _
on a pedestal as shown schematically in
Fig. 9-10. Verneuil used a hopper to hold
and feed the Cr-doped aluminum oxide
powder above a pedestal. Heat was pro-
vided by an oxy-hydrogen torch to initiate
rapid melting of the powder at 2050 °C. As Figure 9-10. Verneuil flame fusion apparatus. Re-
the molten powder is distributed around printed with permission from Hurle (1979), page 110,
the pedestal, it loses heat and solidifies. © 1979, Elsevier Science Publishers BV, Amsterdam.
328 9 Single Crystals

causes the crystals to grow very highly

strained. Crystals often split in half in stor-
age to relieve the built-in strain. The
Verneuil technique has been replaced
(Charvat et al., 1967) for the most part by
the Czochralski technique.

9.7.1.4 Czochralski Technique

The Czochralski technique (1918) (Fig.
9-12) employs the use of a crucible con-
taining the melt from which a solid single
crystal is pulled by controlled lifting after
Figure 9-11. Verneuil-grown sapphire boules.
dipping of a seed (or a noble metal rod, e.g.
Pt, when no seed is available) to induce
nucleation. If multiple crystals nucleate, a
The pedestal for single crystal growth held "necking down" procedure is applied to
a single crystal rod of sapphire or ruby to restrict the growth to one single crystal.
nucleate and promote continuous single Heating of the crucible may be accom-
crystal growth, yielding boules and rods of plished by use of a resistance-heated fur-
various sizes. Modification of the standard nace or radio frequency (rf) induction in
pedestal approach can yield disk-shaped which case the crucible may act as an rf
sapphire boules as shown in Fig. 9-11. susceptor. Heat transfer through the grow-
However, the fast cooling of the material ing crystal and holder rod is enhanced by
in this zone of steep thermal gradients the steep thermal gradient and causes so-

MECHANISM FOR RAISING AND

ROTATING CRYSTAL

GAS
OUTLET

VIEWPORT

CRUCIBLE

HEATING ELEMENT
AND SHIELD

MELT THERMOCOUPLE Figure 9-12. Czochralski crystal growth system.

Reprinted with permission from Academic Press,
Inc., Orlando, FL, from Gentile (1987, 1992),
GAS INLET © 1987, page 5; © 1992, page 713.
 9.7 Crystal Growth Techniques 329

lidification to occur. Constant tempera-

ture control (frequently automated) is nec-
essary to continue growth as well as to
maintain a uniform diameter of the grow-
ing ingot. This is the standard technique
for commercial growth of ruby and sap- VEW1NG STAINLESS
PORT STEEL WALLS
phire from a melt contained in an iridium
crucible.
The Czochralski technique was sub-
sequently developed for commercial elec-
tronic-grade silicon pulled from fused
silica crucibles. However, requirements for N 8 [P > 3 0 ATM.)
high-purity, "zero-defect" silicon free of
oxygen impurity caused much of the pro-
duction of silicon wafers to utilize the float
GaP MELT*
zone technique, as described in Sect.
9.7.1.6, below.
A special variant of the Czochralski
method, the Kyropoulos technique, utilizes
a cooled seed to initiate single crystal
growth within the melt-containing
crucible. However, heat removal continues
by controlled lowering of the furnace tem- Figure 9-13. Liquid encapsulated Czochralski (LEC)
apparatus. Reprinted with permission from Hurle
perature (ramping) to grow the crystal (1979), page 99, © 1979, Elsevier Science Publishers
within the crucible, without any pulling BV, Amsterdam.
and rotation.
Other variations of both these tech-
niques have been developed, one of which,
though some problems of constituent loss
liquid encapsulated Czochralski (LEC)
occur in certain systems which contain ox-
growth, is especially useful for compound
ides which are volatile at growth tempera-
semiconductors which are volatile at their
tures. A major problem for the application
melting points, e.g., GaAs, GaP, InP,
of LEC is the reactivity of most refractory
PbTe, Bi2Te3, etc. Boron oxide, which
oxides with boric oxide as well as many
melts as a glass at 450 °C, remains viscous
other possible encapsulants (see also
to high temperatures and does not react
Vol. 16, Ch. 2, Sec. 2.5.4).
with the above compounds, is most com-
monly used as a molten layer around the
9.7.1.5 Skull Melting
melt. Figure 9-13 shows the experimental
setup for the growth of gallium phosphide A more controlled containerless (or ac-
in which an overpressure (P>30 atmo- tually "self-container") technique than the
spheres) of inert gas is maintained in the Verneuil method discussed above is skull
closed system along with the B 2 O 3 layer, melting (Aleksandrov et al., 1973). Skull
to prevent loss of phosphorus. LEC is not melting has been demonstrated (Nassau,
usually required for the less volatile com- 1981; Wenckus etal., 1977) to produce
pounds such as the refractory oxides al- very large crystals of such materials as cu-
330 9 Single Crystals

bic zirconia (ZrO 2 stabilized in the cubic (just prior to crystal nucleation) com-
form by the addition of Y 2 O 3 , CaO, or pletely in the liquid state and then pro-
MgO). This method utilizes a cold crucible ceeds to solidification. A different ap-
or skull to contain the melt inside a crust of proach is involved in "zoning" techniques
its own powder. The skull is designed for where only a narrow zone of liquid is made
the use of radio frequency (rf) energy to to travel through a polycrystalline ingot
heat and melt the material inside, e.g., cu- for transformation into a single crystal.
bic zirconia. The skull consists of a split The width of the molten zone is dependent
cup (usually split in half), closed at the on the viscosity/surface tension of the
bottom and open at the top, made up of melt. The float zone technique shown in
numerous water-cooled copper tube fin- Fig. 9-14 employs rf heating of a polycrys-
gers. The skull is placed inside a copper talline ingot in contact at one end with a
coil which is energized with an rf genera- single crystal. Melting is accomplished at
tor. Radio frequency penetrates into the the interface and the zone is moved, by the
skull filled with the zirconia powder con- motion of either the ingots or the furnace,
taining a stabilizer and powdered zirco- to continue single crystal growth. Al-
nium metal. The metal is required to act as though performed vertically and contain-
the initial susceptor because the zirconia erless for silicon, horizontal zoning tech-
powder is an insulator at room tempera- niques similar to the zone-refining process
ture. As the zirconia gets hot, it becomes
conducting (similar to alumina) and melts
in the rf field. Eventually, the zirconium
metal reacts with oxygen from the sur-
rounding air to form additional zirconia.
A thin solid skin (less than 1 mm thick) POLYCRYSTAL
ROD
remains next to the walls of the cup be-
cause it is cooled by contact with the
water-cooled copper fingers. This skin,
acting as a container, prevents contamina-
tion as well as reaction between the melt
and the skull. The melt is maintained for
several hours to ensure uniformity; then, it
is slowly cooled. Self nucleation has been
MOLTEN ZONE
observed to start at the bottom of the cup;
crystal growth proceeds from the bottom
to the top until the melt is consumed.
Columns of single crystals as large as 2 cm SINGLE CRYSTAL

across and 1.5 cm long have been grown by

this technique. This process has been com- SEED
mercialized for the growth of large cubic
zirconia crystals (Wenckus, 1993).
Figure 9-14. Float zone technique. Reprinted with
9.1 A.6 Zoning Techniques permission from Encyclopedia of Physical Science and
Technology Vol. 4, Academic Press, Inc., Orlando,
In the liquid-to-solid techniques dis- FL, from Gentile (1987, 1992), © 1987, page 6;
cussed above, the source material exists © 1992, page 714.
 9.7 Crystal Growth Techniques 331

developed by Pfann and Olson (1953), pri- crystal (an excess of one of its constituents)
marily for purification, have been used for to a totally foreign material which dis-
single crystal synthesis. The high melting solves the desired compound under certain
points, reactivity, and relatively low vis- conditions such as heating, but will allow it
cosity of molten refractory oxides limit the to solidify intact upon cooling.
application of zone melting techniques. Al- An ideal solvent should meet the follow-
ternative methods of heating have been ing requirements (Elwell & Scheel, 1975):
used, including lamps and lasers. Ainger
high solubility for crystal constituent;
et al. (1970) successfully grew ferroelectric
crystal phase is only stable phase at growth
tungsten bronze oxide crystals using an
temperature;
arc-image furnace.
appreciable change of solubility with tem-
A recent development (Feigelson, 1985)
perature ;
is laser heated pedestal growth (LHPG)
low viscosity at the applied temperature;
which is a float zone technique for the
low melting point;
growth of small diameter or fiber single
low volatility;
crystals where the material is self con-
nonreactive with crucible;
tained by the surface tension of the melt. A
absence of elements which are incorpo-
laser, usually carbon dioxide, is used for
rated into the melt;
oxide melts since the 10.6 jam photon en-
suitable density;
ergy is readily absorbed, making it rela-
ease of separation from crystal;
tively easy to grow crystals from ceramic
low toxicity.
feed rods and attain temperatures of more
than 3000 °C. LHPG is a versatile tech-
9.7.2.1 Top-Seeded Solution Growth
nique which enables the study of the melt-
ing behavior and crystallization of some of Top-seeded solution growth (TSSG) was
the more refractory oxide compounds and first introduced by Linz et al. (1965) and
to access either those for which there is no described in detail by Belruss et al. (1971),
suitable crucible, or those which do not and has been used successfully for many
lend themselves to skull-melting and where ABO 3 compounds (Rytz et al., 1990) in-
also small quantities of materials are avail- cluding BaTiO 3 , KNbO 3 , SrTiO 3 , and
able. solid solution series which include
Ba^Si^TiOa, BST (Rytz et al., 1985),
and KTa1_JCNb:cO3, KTN (Rytz, 1983;
9.7.2 Indirect Crystal Growth -
Gentile and Andres 1967). The substitu-
Growth from Solution
tion of Sr for Ba in BaTiO3 stabilizes the
Indirect crystal growth is dominated by cubic phase and allows solidification from
solution growth where a solid crystal (es- a melt while avoiding excessive strain or
sentially, the solute) can be retrieved from cracking in the crystal. KTN grows in a
a liquid solution by either cooling slowly cubic phase and the ferroelectric transi-
and thus changing the solubility relations, tions are nondestructive.
or by evaporation of a solvent which is Top-seeded solution growth utilizes a
volatile under the conditions of crystal setup similar to that used for Czochralski
growth. The solvent can range in composi- growth except for a few significant differ-
tion from one that varies slightly from the ences. As illustrated in Fig. 9-15, the
stoichiometric composition of the desired crucible is on an insulated stand within a
332 9 Single Crystals

Figure 9-15. Apparatus for

top-seeded solution growth.

resistance-heated furnace. Crucibles used 1400 °C) but above the eutectic tempera-
for these materials can be platinum or plat- ture (1332 °C).
inum/rhodium alloys. The seed is sus- Solid-solution growth of either BST or
pended from a platinum rod which is not KTN of compositions away from the end
reactive with the melt and acts as a heat members is somewhat more complex. As
sink. The melt consists of the component described in Section 9.6 on phase equi-
oxides with an excess of one. A look at the libria, KTN solid solutions are grown
phase diagram (Fig. 9-16) (Kirby and from a melt whose composition is widely
Wechsler, 1991) indicates that a solution different from the solid material in equilib-
of approximately 35 mol% BaO and rium with it. In addition, lowering the tem-
65 mol% TiO2 yields the cubic phase of perature causes a large, continuous change
BaTiO3 (Rytz et al., 1990) when cooled in crystal composition. This would mean
below the liquidus temperature (near large variations in Curie temperature and

cm Figure 9-16. KTN single

crystal.
 9.7 Crystal Growth Techniques 333

other ferroelectric and optical inhomo- niques. A knowledge of the phase equilib-
geneities within the crystal. In order to rium conditions for the system involved is
maintain uniform composition, KTN is essential under critical and supercritical
grown isothermally (see Fig. 9-8) from a conditions. Autoclave containers for hy-
melt rich in K 2 O using solvent evaporation drothermal growth must be able to with-
of the volatile K 2 O (Gentile and Andres, stand high pressures at elevated tempera-
1967). The use of Pt/Rh crucibles holding tures while resisting attack from either
as much as 800 g of melt yielded large acidic, basic or oxidizing hydrothermal so-
KTN crystals as shown in Fig. 9-16. Be- lutions. Special stainless steel formulations
cause of the nature of the growth and a have proven successful; however, for many
viscous melt, stirring is necessary to main- materials, the container must be lined with
tain a uniform melt. In addition to seed a noble metal to prevent reaction.
rotation at several rpm, ACRT (acceler-
ated crucible rotation technique) devel- 9.7.2.3 The Sol-Gel Process and Its
oped by Scheel and Schulz-DuBois (1971) Derivatives
has been used with success (Scheel, 1972). The sol-gel process is popular today for
Without effective stirring, depletion of re- the growth of polycrystalline thick films of
quired constituents can occur at the solid- such materials as PZT, BST, and KTN for
melt interface which can slow down or applications including non-volatile fer-
stop continued crystal growth or cause roelectric memory devices, pyroelectric de-
constitutional supercooling, resulting in tectors, and capacitors. The approach uti-
spurious deposition of undesired material lizes the chemistry of viscous solutions
on the growing crystal. containing an intimate mixture of the con-
stituent elements of interest (often as
organo-metallic compounds) which are
9.7.2.2 Hydrothermal Growth
uniformly distributed so that upon calci-
Hydrothermal crystal growth is a special nation, the desired stoichiometric com-
case of solution growth utilizing the in- pound forms with little or no segregation.
creased solubility of many compounds in Although the sol-gel technique is not
water under high pressures and tempera- specifically aimed at the growth of thin
tures. The technique is well-known for the single crystal films, many approaches to-
synthesis of large quartz (SiO2) crystals wards growing ferroelectric thin films are
(Laudise, 1987; Laudise and Barns, 1988) based on a solution process with con-
weighing in the range of 5 to 8 kg (Laudise, trolled viscosity - however, not always a
1994) for the electronic communications gel. The viscosity controls the film thick-
industry. Other oxide crystals including ness distributed on a substrate via a spin-
beryl (Be3Al2Si6O18) and ruby (chromi- ning technique as developed in the semi-
um-doped sapphire, a-Al 2 O 3 :Cr), have conductor industry. In this case, the solu-
also been grown hydro thermally. An im- tion is placed in the center of a wafer which
portant advantage of the hydrothermal is rotated at an appropriate velocity to
method is the use of pressure as a variable spread the liquid uniformly across the sur-
which adds an important dimension to face. Highly preferred orientations of the
process control and, in addition, may al- films can be obtained by using a single
low variations in product characteristics crystal substrate with or without buffer
which are not attainable with other tech- layers.
334 9 Single Crystals

9.7.3 Crystal Growth from the Vapor Phase 9.7.3.2 Vapor Deposition of Single-Crystal
Thin Films
9.7.3.1 Physical Vapor Deposition
When the volatility of the constituents is
Crystal growth from a vapor phase has insufficient to cause sublimation, as is the
many complexities, some of which will be case for most refractory oxides, other tech-
covered herein. The simplest case is the so- niques may be employed. The use of high
lidification or condensation of a constituent vacuum frequently creates phase conditions
gas or gases; essentially the opposite of sub- where the constituents can be heated to va-
limation. Under certain conditions of tem- porization. Such is the case in molecular
perature and pressure, for example, water beam epitaxy (MBE) where effusion cells
vapor will form ice. Similarly, II-VI com- containing elemental sources are heated to
pounds, e.g., CdS, ZnS, will grow as single form directional beams of constituent
crystals from their constituent gases: Cd or atoms. A schematic diagram for a MBE
Zn and S 2 . The technique utilized is called system is shown in Fig. 9-17 (Panish, 1986)
Physical Vapor Deposition (PVD) and does for GaAs/GaAlAs. Recently, investigators
not involve any extraneous compound for- have used MBE for the synthesis of new
mation or reactions. In the case of these high temperature superconductors includ-
II-VI compounds, the source material can ing yttrium barium copper oxide (YBCO,
also be the solid compound itself which YBa 2 Cu 3 0 7 _ :c ) and similar systems. Stoi-
sublimes upon heating to form the two gas- chiometry is controlled both by tempera-
eous species. Under the proper conditions ture of the substrate (where solidification is
of pressure and temperature, the gases will occurring) and concentration of the con-
combine upon condensation in stoichio- stituents in the vicinity of the substrate. In
metric proportion to form single crystals of many experiments, the unsuitable proper-
the compounds, or even mixtures (mixed ties of certain elements have been overcome
crystals) of two or more compounds. This is by employing metal-organic (MO) sources
a unique property of II-VI compounds. creating a technique referred to as
PVD as well as other vapor deposition tech- MOMBE. Under these conditions, MBE
niques are not an applicable technique for may be considered an indirect technique.
the synthesis of single crystals of refractory
oxide compounds since the vapor pressures 9.7.3.3 Metal-Organic Chemical Vapor
of most constituents is extremely low even Deposition
at elevated temperatures. However, many
investigators are currently pursuing vapor Approaches to the growth of single crys-
deposition of thin single-crystal films of fer- tal films where the constituent elements are
roelectric, nonlinear, and superconducting introduced as metal-organic compounds
oxides using high-vacuum molecular beam have been designated metal organic chemi-
epitaxy (MBE), metal-organic molecular cal vapor deposition (MOCVD) or organo-
beam epitaxy (MOMBE), and chemical va- metallic vapor phase epitaxy (OMVPE).
por deposition (CVD) (MOCVD or Gaseous compounds of the constituent ele-
OMVPE) techniques. These are covered ments are introduced in proportions cali-
briefly in the discussion that follows. brated to yield the desired stoichiometry
under the growth conditions. A simplified
schematic diagram for MOCVD utilizing
three constituents is shown in Figure 9-18.
 9.7 Crystal Growth Techniques 335

ULTRAHIGH-VACUUM
CHAMBER

RHEED
SCREEN

LIQUID
NITROGEN-COOLED
CRYOPANEL

HEATING
COIL

Figure 9-17. Molecular beam epitaxy (MBE) schematic. Reprinted from Panish (1986), © 1986, with permission
from Pergamon Press Ltd., Oxford.

^ EXHAUST

> Dc n
SUBSTRATE-^!
REACTOR^J Rp
- O - - ^ BYPASS ^ I SOURCE

BYPASS

V E NT VENT

(R) REGULATOR

® VALVE

nTT^l MASS FLOW

CONTROLLER

Figure 9-18. Simplified MOCVD apparatus.

336 9 Single Crystals

In "cold wall" CVD techniques, the source Zn+1/2S 2 — ZnS ZnC!2 + H2S-~ ZnS + 2HCII
gases are selected so that they decompose/
\\\\\\\\\\\\\\\\\\\
react only when they come into contact
with the hot substrate - usually a single
crystal to nucleate epitaxial growth. Epitax-
ial growth occurs when the deposited layer \\\\\\\yv\\\\\\\\\\\
« • « ^» • • • • •
follows the lattice network of the substrate
whether the substrate is the same material
(homo-epitaxy) or a different crystal sub-
stance (hetero-epitaxy). The major factors o o o o oo
for substrate matching in order to obtain o o o° o
o oo
optimum layer quality are lattice parame-
ters and coefficients of thermal expansion.
In cold wall reactors, the substrate is heated VoWcfoW >o*o*o*oV
usually by placing it on a susceptor, e.g., a
carbon block, and using rf induction heat-
ing. In some cases, heating is accomplished fcfo
or assisted by internal or external radiant
lamps. Following the reaction and deposi-
tion of the desired layer, residual gaseous Figure 9-19. Atomic layer epitaxy (ALE). After Sun-
tola and Hyvarinen (1985). Reproduced with permis-
products are removed from the reactor by sion from Annual Reviews Inc., © 1985.
pumping or flowthrough of inert gases. Re-
actor shapes (geometries) may vary to ac-
commodate the reaction relative to the sur- ZnS is shown on the left by direct reaction
face of the substrate; they may be vertical of the elemental constituents, zinc and sul-
or horizontal. Similar geometries may be fur; the right side illustrates the layer-by-
used in "hot wall" reactors which are con- layer reaction of ZnCl2 and H 2 S. Condi-
tained in furnaces. The entire reactor is tions are established to remove excess
heated to maintain constituent elements (or atoms other than the desired monolayers.
compounds) in the vapor state until they ALE has been extremely successful for syn-
react and deposit on the substrate. This thesis of II-VI compound layers because of
may result in deposition which is not re- the properties of II-VI materials. The
stricted to the substrate. II-VI bond, e.g., Zn-S, is much stronger
than either the Zn-Zn or the S-S bonds.
The method could be extended to the ALE
9.73.4 Atomic Layer Epitaxy
growth of oxides, for instance titanium
Atomic Layer Epitaxy (ALE) (Suntola tetrachloride could be reacted with wet air
and Hyvarinen, 1985) is a vapor deposition and/or oxygen to give titanium dioxide.
technique applicable to MBE and/or
MOCYD. A monolayer of each constituent
9.7.4 Solid-to-Solid Crystal Growth
is deposited either individually from an ele-
mental source or as a reactant species to Solid-to-solid crystal growth is domi-
form the desired compound by reaction nated by recrystallization which is typically
with a subsequent layer as illustrated in invoked by sintering to cause grain growth
Fig. 9-19. In the figure, the reaction to form and may ultimately yield single crystal ma-
 9.8 Distribution Coefficient and Mass Transport 337

terial. Direct recrystallization begins and 9.8 Distribution Coefficient and

ends with the same compound but yields Mass Transport
larger crystal grains. Grain growth occurs
to reach a state of lowest free energy. Built- The selective incorporation (or rejec-
in strain in a polycrystalline mass activates tion) of components at a fluid-solid inter-
the process of recrystallization. Gilman face causes the interface of a growing crys-
(1963) describes exaggerated grain growth: tal to represent not only a sink for matter,
"... in certain cases, after prolonged anneal- but also a source of both heat and matter.
ing of a fine-grained matrix, only one (or Latent heat that is released during the
a few) large grain grows at the expense of growth process (when atoms/ions are at-
the whole specimen." (See also Vol. 15, taching to the solid) at the interface, to-
Chap. 9, Sec. 9.6.3, of this Series.) Although gether with matter that is not incorporated
first observed in metals where reorientation into the solid, must be transported away
across grain boundaries leading to grains from the advancing interface. In this man-
disappearing is more frequently observed, ner, heat transfer proceeds through both
this process has been observed in semicon- the fluid and the solid; the rejected con-
ductors, organic materials, ice, and in some stituents are transported through the fluid
minerals when compressed and heated only. The rate with which constituent
(McCrone, 1949). The latter include anhy- atoms (ions) are incorporated into the
drite (CaSO4), fluorite (CaF 2 ), periclase solid is governed by the difference in the
(MgO), and corundum (a-Al2O3). In addi- chemical potentials of the two phases in
tion, as a result of heat and compression of contact. In turn, the chemical potentials
powders (Buckley, 1961), significant grain depend on the concentration of all species
growth has been observed in rock salt present, as well as on the local temperature
(NaCl) and sylvite (KC1). and pressure. Thus, the mass and heat
One of the most successful applications transport problems are coupled. In order
of solid-to-solid crystal growth has been in to determine the time-dependent position
hexagonal ferrites. Single crystals of ap- and shape of an interface, one must simul-
proximately 10 mm diameter by several taneously consider the conservation of
centimeters in length have been reported mass, momentum, and energy for the en-
(Lacour and Paulmus, 1968) for barium tire system. Macroscopic mass and heat
hexaferrite, BaFe 12 O 19 . transport play a central role in crystal
Indirect solid-to-solid crystal growth in- growth. Molecules or atoms must be trans-
cludes the phenomena of spinodal decom- ported in the fluid over macroscopic dis-
position where a "mixed crystal" (a single tances toward the interface where they at-
crystal composed of the constituents of two tach to the crystal. This transport can be
isomorphous crystals) breaks down into its fast or slow compared to the growth (at-
constituent compounds or different mix- tachment) kinetics. The rate at which a
tures because the compounds formed are crystal grows can be limited by interfacial
the equilibrium phases (lowest free energy kinetics or by the transport process (diffu-
states) under the conditions of temperature sion limited). These are the factors which
and pressure (see Vol. 5, Chap. 7 of this Se- must be established by the crystal grower
ries). in terms of control of overall constituent
concentration (source material), tempera-
ture, temperature gradient, pressure, and
338 9 Single Crystals

i
growth rate. When growth is by an indirect i
method, there will be a buildup of those A 1
A' *
components that tend to be rejected from \ * 1
\ \
the solid adjacent to the growing interface. t \ \
I
If the distribution constant (or segrega- X
\ 1

POI
tion coefficient) k, defined as the concen- s ^ \ 1
tration of a constituent, A, in the solid over *Sv \ i

;
KJ
N\\ 1
the concentration of that constituent in the 1
liquid XI
T
si B
lAsolidJ
k= (9-7) 1
CRYSTAL
lAliquidJ 1
1
of a given component is less than unity, DISTANCE -
k< 1, then the component will be rejected Figure 9-20. Composition melting points as a func-
from the solid and will tend to concentrate tion of distance near the growing interface.
close to the growing interface. Conversely,
if k > 1, there will be a depletion of that
component close to the interface as com- point and, therefore, tends to freeze out
pared to the bulk composition. The effec- rapidly (Hurle, 1962). Such a case is illus-
tive segregation coefficient keff in the real trated in Fig. 9-20 which shows the melting
growth regime is different from the equi- point relationships of the compositions
librium k derived from the phase equilib- that exist near the interface of the growing
rium diagram. Thus, k&ff can indicate the crystal as a function of distance. The melt-
impact of crystallization rate, diffusion, ing point of the solution (TS) decreases as
and hydrodynamic processes at the crystal the interface is approached. Lines AB and
growth interface (Burton et al., 1953). Dif- A'B represent two different possible tem-
fusion processes, activated by concentra- perature gradients in the solution. For the
tion and thermal gradients, are required to larger (steeper) gradient, AB, there is no
homogenize the composition at the grow- supercooling at all positions in front of the
ing interface. If crystal growth is so slow as interface. For the smaller gradient, A'B,
to be almost at equilibrium, diffusion can the region CB is supercooled and crystal-
readily counteract this effect. In most lization will tend to take place in front of
cases, however, severe build-up can occur. the solid interface at a rapid rate. Large
Build-up, in turn, leads to a morphological temperature gradients as well as slower
stability problem. This problem is com- growth rates can be used to lessen the ten-
monly encountered in indirect crystal dency toward constitutional supercooling.
growth and can result in constitutional su- If the growing interface is not an "equilib-
percooling. rium form" of the crystal, it will usually
When the rate of growth is too rapid for change to a surface composed of facets
the diffusion of the constituents of the whose faces are equilibrium faces, or at
growing crystal (in their proper propor- least faces of lower interfacial free energy.
tions) for balance at the growth interface, There exists a tendency toward runaway
the composition of the liquid becomes de- dendritic growth resulting in a mosaic
pleted of certain constituents. This leaves a structure. In addition to the thermal gradi-
composition which has a lower freezing ent (which may be furnace limited), the
 9.9 Heat Transfer 339

control of crystal growth at this point de- port properties such as diffusivities, vis-
pends on the growth rate and use of meth- cosities, emissivities, and densities. This
ods such as accelerated crucible rotation does not imply that these parameters can
technique (ACRT) to enhance diffusion be controlled independently; indeed, they
and mixing of the constituents. In essence, cannot. For this reason, modern crystal
we must overcome the constitutional growth still depends largely on data
(composition) differences at the liquid- derived from empirical investigations. Al-
solid or liquid-vapor interface referred to though there has been significant develop-
as the "boundary layer." ment in crystal growth theory, and many
Crystal growth literature contains nu- observed phenomena have been explained
merous mathematical treatments of the and even modelled in mathematical terms
boundary layer. A good comprehensive (Rosenberger, 1979), there still exists a
treatment of the subject appears in the large separation between experiment and
book by Rosenberger (1979). The relation theory. The gaps are being narrowed by
of the boundary layer concept to real crys- many investigators (Koai et al., 1994a). In
tal growth situations is very complicated. crystals grown from the melt utilizing such
In fact, crystal growth situations may or techniques as Bridgman, Czochralski
may not have boundary layers. If crystal (Derby, 1988; Brown, 1988) and float-
growth rates are sufficiently slow, diffusive zone, the uniformity of composition is
flow may spread over the entire region. strongly dependent on the pattern and in-
The important and related roles of macro- tensity of flow in the melt (Oshima et al.,
scopic mass and heat transport have been 1994) as well as the shape of the solid-liq-
noted above for their central role in crystal uid interface (Koai et al., 1994b). The un-
growth processes. Molecules must be equal distribution of a dopant between
transported in the fluid over macroscopic crystal and melt causes a concentration
distances to the crystal-fluid interface. gradient normal to the interface which is
From this position they take their (rela- influenced along the interface by convec-
tively) permanent position on the crystal tion in the melt and the interface shape.
surface. The control factor at this critical This concentration gradient represents the
point is the boundary layer. boundary layer discussed above. The
boundary layer (or diffusion layer) decays
exponentially with distance from the inter-
9.9 Heat Transfer face into the melt when the interface is
planar. The melt is quiescent except for the
The transfer of thermal energy within motion caused by growth of the crystal.
materials and experimental setups is as im- Analysis of the growth of a large diameter
portant for the understanding of crystal crystal with similar velocity field along the
growth as the transport of matter. Specific interface indicates that convection only al-
temperature ranges and thermal profiles ters the concentration field perpendicular
(gradients) are selected. This is done to to the interface. The idea of diffusion-con-
control interphase mass transport rates trolled mass transfer was followed up with
through the temperature dependence of expressions to calculate the thickness of
the chemical potentials; to dissipate or the axial boundary layer without consider-
supply latent heat that is generated or con- ation for the fluid motion in the melt. In-
sumed at interfaces; and to control trans- vestigators typically applied the concept of
340 9 Single Crystals

a stagnant-film layer which masked a Gilmer and Bakker, 1991) to vapor deposi-
growing crystal from a well-mixed melt. tion using MBE in a study of effects of
Recently, some investigators (Hayakawa deposition on misfit surfaces. Simulations
et al, 1993) have approached the problem of MBE deposition of strained Si films
by a consistent analysis between mass provide information on the mobility of
transfer and flow in the fluid. atoms at the surface and give insight into
mechanisms by which strain relief can
occur. In an investigation of columnar
growth morphology in MBE-grown films
9.10 Crystal Growth Theory (which is farthest from the equilibrium
Versus Experiment state of a film), molecular dynamics simu-
lations show that the angle of the beam,
Recent investigations utilizing comput- interatomic forces, substrate temperature,
er-aided analysis of the interactions of nat- and deposition rate all have a strong effect
ural convection and the shape of the crys- on the film structure.
tal/melt interface have established the in- Although many limitations remain in
fluence of these parameters in crystal the experimental arena which require fur-
growth from the melt particularly in Bridg- ther engineering and control over interre-
man and Czochralski systems (Brown, lated parameters, some fall into an area
1988; Derby, 1988; Hurle, 1983). New al- over which we have some control, but re-
gorithms are currently being used which quire still more for the perfection of crys-
pertain to the mathematical free-boundary tals. An insight into the fine tuning of these
problem and simultaneously compute the parameters is being attained by the theo-
interface shape, velocity and pressure retical work in progress.
fields in the melt, and the temperature dis-
tribution in both the crystal and the melt.
Investigators using these algorithms have
concluded that for moderate convection 9.11 References
levels the boundary-layer model is an over- Ainger, F. W., Bickley, W. P., Smith, G. V. (1970),
simplification of the interactions between Proc. Br. Ceram. Soc. 18, 221-237.
complex flow patterns and the dopant Aleksandrov, V. I., Osiko, V. V., Prokhorov, A. M.,
Tatarintsev, V. M. (1973), Vestn. Akad. Nauk SSSR
field. An important difference revealed by 12, 29.
this approach is that the concentration Belruss, V., Kalnajas, J., Linz, A., Fotweiler, R. C.
gradient next to the crystal is far from radi- (1971), Mater. Res. Bull. 6, 899.
Bridgman, P. (1925), Proc. Am. Acad. Arts Sci. 60,
ally uniform; in fact, as much as 60% vari- 305.
ation in radial segregation is calculated. Brown, R. A. (1988), AIChE J. 34, 881.
For the most part, accurate comparisons Buckley, H. E. (1961), Crystal Growth, 5th ed. New
York: Wiley, p. 294.
with experimental values require the Burns, G., Dacol, F. H. (1982), Solid State Commun.
determination of the variation of thermo- 42,9.
physical properties with temperature Burton, J. A., Prim, R. C , Slichter, J. (1953), /. Chem.
Phys. 21, 1987.
which are unknown to date. However, ex- Charvat, F. R., Smith, J. C , Nestor, O. H. (1967),
cellent agreement has been shown where Proc. Int. Conf. on Crystal Growth, Boston 1966;
experimental and property data exist. Suppl. J. Phys. Chem. Solids. Oxford: Pergamon
Press.
Molecular dynamics simulations have Czochralski, J. (1918), /. Phys. Chem. 92, 219.
been applied (Grabow and Gilmer, 1987; Derby, J. J. (1988), MRS Bull. 13 (10), 29.
 9.11 References 341

Elwell, D., Scheel, H. J. (1975), Crystal Growth from McCrone, W. C. (1949), in: Crystal Growth, Discus-
High-Temperature Solutions. London: Academic sion of the Faraday Society, Vol. 5. London: Butter-
Press. worths, pp. 158-166.
Feigelson, R. S. (1985), in: Crystal Growth of Elec- Muller, K. A. (1981), in: Nonlinear Phenomena at
tronic Materials: Kaldis, E. (Ed.)- New York: El- Phase Transitions and Instabilities: Riste, T. (Ed.).
sevier, Chap. 11. New York: Plenum Press.
Gentile, A. L. (1983), AIChE Symp. Series, Tutorial Muller, K. A., Luspin, Y, Servoin, J. L., Gervais, F.
Led. Electrochem. Eng. Technol.-II 79, 144. (1982), J. Phys. Lett. (Paris) 43, L-542.
Gentile, A. L. (1987), in: Encyclopedia of Physical NAS (National Academy of Sciences, USA), (1968),
Science and Technology, Vol. 4: Crystal Growth. Ceramic Processing, Publication 1576. Washington,
Orlando, FL: Academic Press, pp. 1-14. D.C.: National Academy of Sciences.
Gentile, A. L. (1992), in: Encyclopedia of Physical Nassau, K. (1981), The Lapidary Journal 35, 1194-
Science and Technology, Vol. 4: Crystal Growth. Or- 1200, 1210-1214.
lando, FL: Academic Press, pp. 709-725. Oshima, M., Taniguchi, N., Kobayashi, T. (1994), /.
Gentile, A. L., Andres, F. H. (1967), Mater. Res. Bull Cryst. Growth 137, 48-53.
2, 853. Panish, M. B. (1986), Prog. Cryst. Growth Charact.
Gilmer, J. J. (Ed.) (1963), Art and Science of Growing 12, 1.
Crystals. New York: Wiley, p. 434. Pfann, W G., Olsen, K. M. (1953), Phys. Rev. 89,
Gilmer, G. H., Bakker, A. F. (1991), in: Computer 322.
Aided Innovation of New Materials: Doyama, M., Reisman, A. (1970), Phase Equilibria - Basic Princi-
Suzuki, T., Kihara, I, Yamamoto, R. (Eds.). Am- ples, Applications, Experimental Techniques.
sterdam: North-Holland, p. 687. New York: Academic Press.
Grabow, M. H., Gilmer, G. H. (1987), in: MRS Rosenberger, F. (1979), Fundamentals of Crystal
Symp. Proc. 94, 15. Growth I. Berlin: Springer.
Hayakawa, Y, Asakawa, K., Torimoto, Y, Yama- Rytz, D. (1983), Ferroelectricite Quantique dans
shita, K., Nakayama, A. (1993), J. Cryst. Growth KTairxNbxO3, These N° 475, Departement de
128, 159-162. Physique, Ecole Polytechnique Federate de
Hurle, D. T. J. (1962), Prog. Mater. Sci. 10, 79. Lausanne.
Hurle, D. T. J. (1979), Crystal Growth: A Tutorial Ap- Rytz, D., Wechsler, B. A., Kirby, K. W, Nelson,
proach, North-Holland, Series in Crystal Growth, C. C. (1985), Jpn. J Appl. Phys. 24, 622-624, Proc.
Vol. 2, W. Bardsley, D.T.J. Hurle, J. B. Muller 6th Int. Mtg. on Ferroelectricity, Kobe, Japan
(Eds.). Amsterdam: Elsevier Science Publishers, 1985.
p. 110. Rytz, D., Wechsler, B. A., Nelson, C. C , Kirby,
Hurle, D. T. J. (1983), J. Cryst. Growth 65, 124. K. W. (1990), /. Cryst. Growth 99, 864-868.
Kirby, K., Wechsler, B. A. (1991), /. Am. Ceram. Soc. Scheel, H. J., Schulz-DuBois, E. O. (1971), J. Cryst.
74, 1841-1847. Growth 8, 304.
Koai, K., Seidl, A., Leister, H.-J., Muller, G., Kohler, Scheel, H. J. (1972), J. Cryst. Growth 13/14, 560.
A. (1994a), /. Crvst. Growth 137, 41-47. Shih, C.-C, Yariv, A. (1980), Phys. Rev. Lett. 4, 281 -
Koai, K., Sonnenberg, K., Wenzl, H. (1994b), J 284.
Cryst. Growth 137, 59-63. Shih, C.-C, Yariv, A. (1982), J. Phys. C: Solid State
Lacour, C , Paulus, M. (1968), /. Cryst. Growth 3-4, Phys. 15, 825-846.
814-817. Stockbarger, D. C. (1936). Rev. Sci. Instrum. 7, 133-
Laudise, R. A. (1967), Proc. Int. Conf on Crystal 136.
Growth, Boston 1966; Suppl to J. Phys. Chem. Stringfellow, G. B. (1984), J. Cryst. Growth 68, 111 —
Solids. Oxford: Pergamon Press, 3-16. 122.
Laudise, R. A. (1970), The Growth of Single Crystals. Stringfellow, G. B. (1993), /. Cryst. Growth 128, 503-
Englewood Cliffs, NJ: Prentice-Hall. 510.
Laudise, R. A. (1987), Chem. Eng. News 65 (39), Suntola, T., Hyvarinen, J. (1985), Annu. Rev. Mater.
30-43. Sci. 15, 111.
Laudise, R. A. (1994), Private communication. Thurmond, C. D. (1959), in: Semiconductors: Han-
Laudise, R. A., Barns, R. L. (1988), IEEE Trans. nay, N. B. (Ed.). New York: Reinhold.
Ultrasonics, Ferroelectrics, and Frequency Control Verneuil, A. (1902), C. R. Acad. Sci. (Paris) 135,
35 (3), 277-287. 791-794.
Linz, A., Belruss V., Naiman, C. S. (1965), Meeting of Wechsler, B. A., Kirby, K. W (1992), J. Am. Ceram.
Electrochem. Soc. San Francisco. Extended Ab- Soc. 75, 981-984.
stracts 2, 87. Wells, A. F. (1984), in: Structural Inorganic Chemis-
Megaw, H. D. (1957), Ferroelectricity in Crystals. try, 5th ed. Oxford: Clarendon Press, Chap. 2.
London: Methuen. Wenckus, J. R, Menashi, W. P., Castonguay, R. A.
(1977), U.S. Patent 4049384.
342 9 Single Crystals

Wenckus, J. (1993), in: Proc. 10th Int. Conf. on Crystal Laudise, R. A. (1970), The Growth of Single Crystals.
Growth, San Diego, CA, 1992, J. Cryst. Growth 128, Englewood Cliffs, NJ: Prentice-Hall.
13-14. Lines, M. E., Glass, A. M. (1977), Principles and Ap-
Yariv, A. (1967), Quantum Electronics. New York: plications of Ferroelectrics and Related Materials.
Wiley, Chap. 18. Oxford: Clarendon Press.
Megaw, Helen D. (1957), Ferroelectricity in Crystals.
London: Methuen.
Nye, J. F. (1969), Physical Properties of Crystals. Ox-
ford: Clarendon Press.
General Reading Pfann, W G. (1966), Zone Melting, 2nd ed. New
York: Wiley.
Buckley, H. E. (1961), Crystal Growth, 5th ed. New Rosenberger, F. (1979), Fundamentals of Crystal
York: Wiley. Growth I, Springer Series in Solid-State Sciences,
Elwell, D., Scheel, H. J. (1975), Crystal Growth from Vol. 5. Berlin: Springer.
High-Temperature Solutions. London: Academic Roy, R. (Ed.) (1994), Crystal Chemistry of Non-
Press. Metallic Materials, Vol. 4. Berlin: Springer.
Gilman, J. J. (Ed.) (1963), Art and Science of Growing Wells, A. F. (1984), Structural Inorganic Chemistry,
Crystals. New York: Wiley. 5th ed. Oxford: Clarendon Press.
Hannay, N. B. (1967), Solid State Chemistry. Engle- Wilke, K.-T. (1963), Methoden der Kristallziichtung.
wood Cliffs, NJ: Prentice-Hall. Berlin: VEB Deutscher Verlag der Wissenschaften.
Herbert, J. M. (1982), Ferroelectric Transducers and Wilke, K.-T. (1973), Kristallziichtung. Berlin: VEB
Sensors. London: Gordon and Breach. Deutscher Verlag der Wissenschaften, Berlin.
Hurle, D. T. J. (Ed.) (1992), A Perspective on Crystal Crystals: Growth, Properties, and Applications, Vol-
Growth. Amsterdam: Elsevier. umes 1-13. Berlin: Springer.
Jona, R, Shirane, G. (1962), Ferroelectric Crystals. Crystal Growth, Discussions of the Faraday Society,
Mineola, NY: Dover Press. No. 5, 1949. London: Butterworths.
10 Green Microstructures and Their Characterization
Ben C. Bonekamp and Hubert J. Veringa

Netherlands Energy Research Foundation ECN, Petten, The Netherlands

List of Symbols and Abbreviations 344

10.1 Introduction 346
10.2 Structure of Green Bodies 347
10.2.1 Definition of Green Bodies 347
10.2.2 Green Bodies in Ceramic Processing 348
10.2.3 Macrostructure, Microstructure and Texture 348
10.2.4 What is a Homogeneous Green Material? 349
10.2.5 Ceramic Membranes, an Example 350
10.3 The Structure of Particle Packings 353
10.3.1 Introduction 353
10.3.2 Packing of Spherical Particles of Uniform Sizes 354
10.3.3 Bimodal Stackings of Spherical Particles 358
10.3.4 Sol-Gel Structures 359
10.3.5 Hierarchical Cluster Packing 362
10.3.6 Measurable Quantities 363
10.3.7 Processing Technology in Relation to Green Structures 364
10.3.7.1 Wet Processing 364
10.3.7.2 Dry Processing 368
10.3.7.3 Drying and Sintering 368
10.4 Characterization Methods 371
10.4.1 Types of Green Bodies and Usability of Characterization Techniques . . . 371
10.4.2 Macrostructure and Texture 373
10.4.3 Green Microstructures, Current Practice 374
10.4.4 Imaging Techniques 377
10.4.5 Capillary and Fluid Flow Techniques 378
10.4.6 Mercury Porosimetry 380
10.4.7 NMR and Small Angle Scattering Techniques 382
10.4.8 Rheological Measurements 383
10.5 Appendix 383
10.5.1 Consolidation Effect During Wet Processing 383
10.5.2 Brownian and Colloidal Effect During Sedimentation 384
10.6 References 385

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
344 10 Green Microstructures and Their Characterization

List of Symbols and Abbreviations

a particle radius
A area
C(r) density-density correlation function
d distance
D fractional dimensionality
F force
G average coordination
k Boltzmann constant
M mass
n number of particles
AT number of contacts
p material property; hydraulic pressure
P porosity; pressure
r particle radius; length scale
R radius of a fractal cluster; particle radius
S correlation function
t time
T temperature
U free energy
v velocity
V volume
x spatial coordinate
zg densification parameter

a proportion of hexagonal stacked material; fraction of coarse particles

y water surface tension
e coordination probability, stacking density; porosity
rj viscosity
Q stacking density; average density
a solid density
q> particle density
<P volume fraction
Q molecular volume

AFM atomic force microscopy

BSD broad (particle) size distribution
CAT computerized axial tomography
CPVC critical particle volume concentration
CSLM confocal scanning laser microscopy
DLA diffusion-limited agglomeration
DLCA diffusion-limited cluster-cluster agglomeration
DLVO Derjaquin, Landau, Verwey, Overbeek (theory)
MRI magnetic resonance imaging
 List of Symbols and Abbreviations 345

NMR nuclear magnetic resonance

NSD narrow (particle) size distribution
NSOM near-field scanning optical microscopy
REV representative elementary volume
RLA reaction-limited agglomeration
RLCA reaction-limited cluster-cluster agglomeration
SANS small angle neutron scattering
SAXS small angle X-ray scattering
SEM scanning electron microscopy
TEM transmission electron microscopy
346 10 Green Microstructures and Their Characterization

10.1 Introduction ramie product, but also of the intermediate

products between the several unit opera-
Progress in technology depends on tech- tions used. For the discussion in this chap-
nical and scientific advancements and the ter it is important to recognize that the
boosting effect that emerges from the mu- three main classes of forming processes:
tual interaction between the two. This is pressing, plastic forming and casting, also
also the case for engineering and engineer- give rise to green bodies which differ in
ing science in the ceramic processing field. composition, structure and texture, even
Although in the past the development of when the same starting powder is used.
processing routes for ceramic materials These differences are due to differences in
depended heavily on empirical correla- additives such as polymers and surfactants
tions being established between final prop- used, but also to differences in structure
erties and processing variables (Fig. 10-1), imposed by the forming process itself.
nowadays, little progress can be expected Occasionally we refer in this chapter to
in the development of advanced ceramic processing-microstructure relations but
materials and their processing routes on an in depth discussion of this subject is
the basis of empirical correlations alone beyond the aim of this chapter.
(Reed, 1988). This motivates the develop- In order to be able to characterize mi-
ment of the science of ceramic processing. crostructural properties of a green body it
The objectives of this science, as defined by is first necessary to have information
Reed (1988), are to identify the important about the composition of the body. Here it
characteristics of the system and to under- is meant that one should know the inor-
stand the effects of processing variables on ganic solid content, the liquid content and
the evolution of these characteristics the amount and types of organic matter
(Fig. 10-2). The objectives in ceramic pro- present. The latter should be known from
cess engineering should be to change these the granule formulation in the case of dry
characteristics purposefully to improve pressing or the dispersion formulation in
product quality (Reed, 1988 a). the case of casting and plastic forming
An important part of ceramic process- techniques. Such information is necessary
ing science is the characterization of not for a correct pretreatment of specimens for
only the initial materials and the final ce- characterization but also for the interpre-
tation of results from characterization
methods as porosimetry, gasadsorption
and (electron) microscopy.
Empirical
Our goal is to give insight into the pres-
Correlation ent day knowledge of the microstructural
properties of particulate materials, which
Processing Variables Properties are of importance in ceramic processing
Batch composition Modulus of rupture
science and technology.
Milling time Dielectric constant Our treatment is necessarily biased by
Pressing pressure Thermal conductivity our own research experience in the area of
ceramic membrane development.
Figure 10-1. Correlation between the final properties
and processing variables may identify the more sensi- In Section 10.2 some general aspects
tive processing parameters but is empirical (from concerning green bodies, their structure,
Reed, 1988). and the role they play in ceramic process-
 10.2 Structure of Green Bodies 347

Ceramic Processing

Batching Forming Drying Finishing

^^ w

Beneficiation Firing

Characteristics Characteristics Characteristics Characteristics Characteristics

Initial Materials Process System Body Unfired Product Final Product

t
Physical
(Chemical)
Properties

Figure 10-2. Processing develops the characteristics of the system (from Reed, 1988).

ing are discussed. Section 10.3 deals with body is the final fired product. We remark
theoretical aspects of the structure of parti- that often only the particle compact ob-
cle packings and in Section 10.4 experi- tained after forming and drying is consid-
mental methods for the characterization of ered as the green body. In this chapter we
green bodies and green microstructures in mainly discuss microstructures of green
current use will be discussed as well as new bodies obtained by packing of isometric
possibilities for characterization not yet particles with sizes in the colloidal regime.
commonly used in ceramics and still in de- Hence we restrict ourselves to the discus-
velopment. The main emphasis from the sion of the structure of green bodies ob-
experimental side is, as said before, on the tained by the processing of submicron
characterization of the pore structure. We powders and sol-gel processing with par-
take porous ceramic membranes as a spe- ticulate sols.
cial example because here submicrometer To begin with, it should be mentioned
ceramics and sol-gel ceramics both play a that green bodies obtained from process-
role and our own R&D experience is in this ing of submicron powders and sols consist
field. of a porous three dimensional structure of
particles. The space between the particles
consists of at least one phase, which can be
10.2 Structure of Green Bodies fluid or solid or both. When the void space
is completely gaseous the body is called
10.2.1 Definition of Green Bodies
completely unsaturated. When the void
In our treatment we consider all prefired space is filled with liquid or (organic) solid
bodies prepared by some forming tech- the body is called saturated.
nique as green bodies. Therefore, calcined Injection moulding and extrusion of
bodies are also considered as such and in polymer-filled melts result in green bodies
some cases we consider even a "bisque which are nearly completely saturated with
body", where still no overall densification solid polymer. Pressing of pure powders
has taken place, as a green body. However, results in completely unsaturated green
for several types of porous ceramics, for bodies, unless moisture sorption occurs to
example ceramic membranes, the bisque an appreciable extent. Undried green bod-
348 10 Green Microstructures and Their Characterization

ies from casting techniques such as pres- in sharp contrast to the, often very de-
sure filtration, slip- and tape casting are tailed, studies concerning the mechanical
completely saturated with liquid. Upon and structural properties of the fired prod-
drying, a green body forms, which, de- uct. More detailed characterization of the
pending on the degree of shrinkage and green structure is however needed in order
organic solid content, is saturated or un- to be able to understand the relation be-
saturated. tween powder forming and compact struc-
The kind and distribution of phases in ture.
the void space of the inorganic particle net- A quotation from Reed's (1988 b) article
work of a green body are important mi- entitled: "Critical Issues and Future Direc-
crostructural characteristics. They are im- tions in Powder Forming Processes" reads
portant for the (mechanical) properties of as follows: "Ceramic forming is one of sev-
the green body and for the development of eral critical steps in processing powders in-
the particle compact during further pro- to green products, but at present this is the
cessing operations as drying, calcining processing step that is most poorly under-
(binder burn out) and firing. Unfortunate- stood in a scientific context".
ly, information on these aspects of green Although progress has been made since
microstructure is very scarce and we will then, Reed's statement is still valid.
discuss it further only briefly.
Characterization of green bodies is
10.2.3 Macrostructure, Microstructure
many sided. On the one hand the investiga-
and Texture
tion of overall properties such as type and
quantity of organic matter, inorganc solid, Before proceeding further we now want
and liquid content are part of it; on the to define more precisely the notion of mi-
other hand, there is the determination of crostructure of a particulate material.
microstructural and textural properties of It is not possible to define the mi-
the body. We discuss mostly the last two crostructure of a material as the structure
and to be more specific the pore structure at a certain length scale between atomic
of a green compact is our main concern. and macroscopic dimensions. Consider for
example a xerogel built from nanometer-
sized colloidal particles which has only
10.2.2 Green Bodies in Ceramic Processing
pores smaller than 5 nm, and a slip-cast
Green bodies are, in fact, a subclass of body with pores larger than 1000 nm. With
the so-called particulate materials. Soils, a yardstick of 1 jum, the xerogel is struc-
pastes and paints are other examples. The tureless but the slip-cast body is not.
study of particulate materials is rather ad- We define the microstructure of a partic-
vanced in colloid science, soil mechanics ulate body, consisting of one solid phase
and coating technology. However, the and a fluid, as the three-dimensional ar-
study of green bodies and their forming is rangement of the solid particles, which
only in its infancy in the science of ceramic were the kinetic units during the forming
forming. operation. These solid particles are called
Usually the ceramist is only interested in the structural units of the particulate mate-
the green density and green strength, not in rial (Feda, 1982). Hence these structural
detailed compositional and structural in- units can be primary particles (crystallites)
formation about a green compact. This is as usually defined in ceramics or aggre-
 10.2 Structure of Green Bodies 349

gates (hard agglomerates). The three di- Bachmat (1991) (their Chap. 1). A few
mensional arrangement of voids between powder particles packed together consti-
the inorganic structural units is the "pore" tute a microscopic rather than a macro-
microstructure. When the structural units scopic system. The properties of a micro-
itself have no open porosity the "particle" scopic sample cannot be expected to be
microstructure and "pore" microstructure representative of those of the macroscopic
are complementary at least when the pore green body from which it has been re-
fluid is considered as structureless. When moved.
the pore space is filled with solid polymer Let us suppose that a macroscopic struc-
having amorphous and crystalline regions ture parameter, such as porosity or perme-
the distribution of these regions in the pore ability, is determined in a series of samples
space are microstructural features super- of increasing size, taken from a large po-
imposed on the pore microstructure. The rous compact, and the results are plotted
pore microstructure is also more compli- against the sample volume (Fig. 10-3). The
cated for partially saturated green materi- results will vary with the sample size as
als. The distribution of solid polymer or indicated. With increasing sample size,
fluid over the pore space is then an addi- however, the amplitude of the random
tional microstructural level, which is im- fluctuations gradually decreases, until fi-
portant for the burn-out and drying kinet- nally a smooth line is obtained. By defini-
ics of the material. tion, the sample is said to be macroscopic
Spatial heterogeneity in microstructure, whenever the macroscopic pore structure
i.e., alternation of layers of smaller and parameter studied is not fluctuating any-
larger structural units, of structural units more when including more material
of different shapes, orientation, composi- around the initial sample point, but its
tions or of regions of different densities is variation can be represented by a smooth
called texture (Feda, 1982). Sometimes the line. When this line runs parallel to the
term macrostructure is used instead (Feda, abscissa, the medium is said to be macro-
1982) but we reserve this term for hetero- scopically, or statistically, homogeneous.
geneities in the microstructure on an ap- This definition of homogeneity is
preciably larger scale than the size of a worked out mathematically, amongst oth-
structural unit other than those mentioned ers, by Bear and Bachmat (1991). They
above, for example cracks, large voids define a so-called representative elemen-
caused by air bubbles, inclusions such as tary volume (REV) of a porous medium,
dust particles or large aggregates. which allows for the transition from a mi-
Hornbogen (1986) gives more general croscopic to a macroscopic continuum de-
and elaborate definitions of microstruc- scription of a porous medium. The size of
ture. He characterizes microstructures as a the REV is such that parameters that rep-
nonequilibrium phenomenon. resent the distributions of the void space
and of the solid matrix within it are statis-
tically meaningful. The REV is between
10.2.4 What is a Homogeneous
C/min and C/max in Fig. 10-3.
Green Material?
When a macroscopic parameter of a
In this section we quote from and follow green body first remains constant and then
closely Dullien (1979) (his Sec. 3.1.6) and suddenly changes to a different but con-
some arguments discussed by Bear and stant value the compact is called macro-
350 10 Green Microstructures and Their Characterization

Domain of
microscopic Domain of Solid particles

1.0 . inhomogeneity porous medium Domain of (possible)

macroscopic
heterogeneity

^ ^ Heterogeneous
Figure 10-3. Variation of porosity,
5 I ^v**i*^
! \i
medium
n, in the neighborhood of a point
I | Homogeneous
Range for J medium as a function of the extended
volume (from Bear and Bachmat,
1991)
Volume, U

scopically heterogeneous in the discontin- strongly dependent on the pore structure

uous sense. Proceeding in any chosen di- or microstructure of the separating layer,
rection in the medium you may find that which is the top layer with the finest pore
the macroscopic microstructure parameter structure on top of several other support-
measured is a continuous, nonconstant ing layers, shown in Figs. 10-4 and 10-5.
function of position. Then the medium is The tubular support structure is made out
called macroscopically heterogeneous in the of material with particles in the range of
continuous sense. some tens of microns and particles in the
Although the preparation of homoge- submicron range. The first is to establish
neous green bodies is in many cases very an open structure and the second have
important for the sintering step and the some sintering at moderate temperatures.
properties of the final product, quantita- The second intermediate layer, which can
tive verification of the homogeneity of already be considered as a microfiltration
green bodies is, as far as we know, not layer, is made out of particles in the submi-
being practiced yet. cron range.
The pore size distribution of the func-
tional top layer should be as narrow as
10.2.5 Ceramic Membranes, an Example
possible in order to increase the selectivity
A ceramic membrane is a porous lami- of the membrane. The kind of species that
nate material in the form of tubes or plates can be separated by the membrane de-
which serves as a filter for the filtration of pends on the pore size and its distribution
liquids in order to purify or to concentrate in the membrane as well as the interaction
a specific component of a liquid or for the of the species with the pore surface. A
separation of mixtures of gases. Hence ce- membrane has gas separation properties as
ramic membranes are functional ceramics the pore size is smaller than 2 nm.
just as ceramic superconductors are. The The membrane as a whole is an asym-
functioning of the ceramic membrane is metric laminar composite, which is on pur-
 10.2 Structure of Green Bodies 351

Separation Toplayer
3 - 100 nm pores

Intermediate Layer
100 - 1500 nm pores

Porous Support
1000 - 15000 nm pores

Figure 10-4. Schematic representation of an asym-

metric ceramic membrane system.

pose macroscopically heterogeneous in the

discontinuous sense. For this chapter we
are mainly interested in the microstructure
of the intermediate and toplayers. The in-
termediate layers have a mean pore diame-
ter in the range 4-1500 nm. They are
formed by coating of a support having
pores of 1-15 jum with a (colloidal) sus-
pension of inorganic particles using the
technique of slipcasting and/or filmcoating
(Bhave, 1991; Terpstra et al., 1988), fol-
lowed by drying and sintering. A top layer
is formed using sol-gel processing, e.g.,
colloidal filtration of a sol, with the inter-
mediate layer as a substrate or mould. The
layers should have a pemeability as high as
possible. The thickness of especially the
toplayer system should be as thin as possi-
ble in order to optimize the flux through
the membrane.
It is the interplay of the structural stabil-
ity of the supporting structures and the
particle stacking of the green structure of
the subsequent layers that makes that the
green structure, also during sintering, is
largely being preserved, even when the ma- (b)
terial itself would sinter to its full density Figure 10-5. (a) SEM of a fracture surface of a four
under unconstrained conditions. There- layer alumina ceramic membrane developed at the
Netherlands Energy Research Foundation, ECN.
fore the microstructure of the dried and Two intermediate layers and one top layer are visible,
fired layers deviates appreciably from that (b) Detail of the membrane system shown in Fig. 10-5 a.
of bulk material prepared from the same The second alpha alumina intermediate layer and the
dispersions and treated similarly, as shown gamma alumina separation top layer are shown.
352 10 Green Microstructures and Their Characterization

(a)

(b) (b)
Figure 10-6. SEM of fracture surfaces: (a) An a-alu- Figure 10-7. Same as Fig. 10-6 but now after sinter-
mina plate (bulk material) prepared by casting a con- ing for 4h at 1500°C. (a) Bulk alumina membrane
centrated suspension of a submicrometer alumina layer; (b) intermediate alumina layer. Note that the
powder followed by drying and sintering, (b) Interme- bulk material is sintered to full density as expected
diate ceramic thin layer on a porous support prepared but that the thin supported coating is still porous and
by film coating of the same suspension as (a). Sinter- has a granular appearance similar to that after sinter-
ing conditions are 2 h at 1200°C in both cases. ing at 1200°C.

in Figs. 10-6 and 10-7. This is caused by bad situation because important experi-
the constrained microstructural develop- mental techniques as gas adsorption and
ment of the coatings during firing due to Hg porosimetry cannot be used for the
the adherence to a substrate. We will come characterization of thin layers. On the oth-
back to this point in Sec. 10.3.7.3. er hand, the practice of considering a ma-
The microstructural characterization of terial, after the first stage of sintering, as
unsupported membrane layers or bulk similar to a dried green body is better for
specimens in stead of the supported mem- ceramic membranes and other thin layers
branes should therefore be looked at with than for bulk materials.
caution. On the one hand, this is a rather
 10.3 The Structure of Particle Packings 353

10.3 The Structure of Particle • The dilatancy, which is the tendency of

Packings a stacking to deform when a shear stress
is applied during a relatively long period
10.3.1 Introduction of time. In such a case the particles in the
stacking should slide past one another
A particle packing, in fact, can be under- leading to an expansion of the structure.
stood as a collection of particles, which, as Another consequence is that the re-
a whole, has not undergone any thermal sponse of the stacking under a sudden
treatment, other than drying, but is able to application of the load is absent or very
withstand mechanical handling without limited. A more gradual application of a
deterioration. Such a packing should be shear stress and for a long period would
considered as a collection of particles, nor- lead to quite extensive deformations.
mally of sizes less than 10 jim and loosely
held together by dispersion or other col- Looking into more detail, the individual
loidal forces. The particles may have particles and the mutual interaction as well
shapes varying from spherical or platelike as the interaction with the surrounding will
to fibrous, while the structure itself may be manifest themselves, leading to parameters
made up of a combination of these de- determining the statistical nature of the
pending on the required final properties of packing.
the body to be made. • Particle size and size distribution. Cur-
A number of different parameters are rent practice is that for making high
used to characterize such a green mi- quality ceramic materials the size distri-
crostructure, although completeness is bution of the particles is as narrow as
very hard to obtain since two microstruc- possible. This means that normally uni-
tures with the same characteristics may modal or bimodal particle arrangements
show different properties. are considered.
We may distinguish between bulk or av- • Particle shape: spherical, platelike etc.,
erage properties, properties revealing the often expressed as terms of aspect ratios.
statistical nature of such a packing and the • Coordination defined as the number of
characteristics of the individual particles. contacts between neighboring particles
Bulk properties are: and counted per particle. In this respect
• The overall density, which can be either the occurrence of a real physical contact
the weight of the stacking per unit of between adjacent particles should be
volume or the total volume of the mate- considered rather loosely if we want to
rial per unit of bulk volume. relate the sintering behavior to the parti-
• The porosity, which is the pore volume cle stacking characteristics. During sin-
per unit of stacking volume. tering the particles will, to some extent,
• The green strength, defined as the stress move into the direction of one another
to failure of the stacking. leading to a rapidly increasing coordina-
tion in the very first beginning of the
Further we may also introduce some sintering.
rheological characteristics when it is ex-
pected that the stacking will show a dy- At this moment the actual difficulty of
namic response upon the application of a characterizing a structure is showing up.
shear stress: We may speak of a statistical distribution
354 10 Green Microstructures and Their Characterization

of locations of particles in a structure and, ing each particle move to their nearest
connected with that, a pore size and shape neighbor as long as no new overlappings
distribution, but it is impossible to connect are created. Such a stacking will be termed
numerical values to the structural peculiar- a loose random packing. This structure is
ities. For instance, the pores are intercon- certainly not in its densest random state
nected such that small and large pores will and we can allow for further movement of
be mixed up to a high degree of random- the particles to give rise to some additional
ness. Unfortunately the densification of densification. The final structure is called a
the local structure is strongly depending on dense random packing. In the practical situ-
the local pore size and shape and their ation we can make a loose random packing
statistics which means that a description of by filling up a large container with small,
the local densification effect in relation equally sized, spherical particles, one by
with the local stacking structure as well as one. It will be observed in such a case that
its influence on the microstructure devel- a further densification is obtained when we
opment should be the ultimate goal of the shake this packing for some period of time.
present considerations. In the real situation it is found that by
In this section emphasis will be given to extrapolating to infinite random packings,
a method to describe both quantitatively where wall effects are excluded, the density
and qualitatively a green microstructure has an upper limit of 0.6366 + 0.0004.
such that measurable parameters can be Most surprisingly this value is very close to
used to describe the microstructural devel- (2/TC) = 0.6366197, for which no theoretical
opment of a sintering or at least densifying explanation has been found so far.
ceramic structure. The loose random packing is not easily
identified with an actual stacking situa-
tion, but it can be a reasonable approxima-
10.3.2 Packing of Spherical Particles
tion of a stacking of particles which have
of Uniform Sizes
undergone a limited amount of floccula-
Most generally, a packing that is to serve tion before stacking, of which the floes are
as a precursor for a ceramic microstructure in a more or less dense random order. Ob-
should be considered as random. The term viously, to attain a dense random material
randomness is however difficult to define after sintering, such a structure is of great
formally, but there are some ways to make importance, but also the effect of modality
a packing of particles where the method should be considered into more detail.
itself uses procedures giving a high degree It has already been mentioned that the
of randomness. coordination number is the number of
One possibility is to consider a container contacts of a given particle with its adja-
into which we accommodate particles hav- cent particles. Particularly for ceramic
ing dimensions smaller than the actual size technology this coordination is very im-
of the container. We may put the particles portant since material transport during
in the container such that their centers of sintering will necessarily take place where
gravity are at random positions as long as mutual contacts between particles are
no overlappings occur. We can proceed present. Of greatest importance for a par-
with this filling up until it is no longer ticular packing, therefore, is the average
possible to accommodate new particles, af- coordination and the stacking density as
ter which the structure is densified by mak- well as the distributions of the coordina-
 10.3 The Structure of Particle Packings 355

tion over the entire ensemble of particles. Eliminating the parameter a and substitut-
In this respect it should be mentioned that ing the result into the expression for the
a random packing with a homogeneous porosity we obtain
density will have a coordination distribu-
G = 26.493 - 10.737/e (10-5)
tion, together with additional variations
due to global fluctuations as a conse- where the density, g, is equal to \—P.
quence of peculiarities of the processing The result is a very smooth and simple
technology. A simple relation between the connection between the average density
coordination number and the stacking and the average coordination. It is graphi-
density is obtained by the following argu- cally displayed in Fig. 10-8. The same ar-
ment (Hudson, 1947). gument can be carried some steps further
Consider a random dense packing as a by taking into account the local coordina-
system of intervening stackings of both tion of both cubic, orthorhombic, tetrago-
hexagonal and cubic structures which have nal and rhombohedral packings where odd
porosities of 0.2595 and 0.4764 respective- coordinations are necessarily considered
ly. If the proportion of hexagonal stacked as averages of structures with different
material is given by a, the relative amount even coordinations (Hudson, 1947). The
of cubically stacked material will be given result of such an analysis is given in the
by 1 — a so that the average porosity, P, is same figure representing the result of the
equal to former analysis. It shows that the value for
the coordination as a function of stacking
P = 0.2595a + 0.4764(1 - a ) (10-1) density is slightly reduced.
Further if we consider a specific volume Another analysis based upon geometric
of the cubic structure equal to (2r)3 with r arguments is given by Ouchiyama et al.
the radius of the spherical particle, we can (1980), leading to a similar expression as
place one particle of a cubic stacking order the former result. The actual expression
in such a volume and y/2 particles of a giving this relation is
hexagonal stacking order. The cubic struc-
ture has coordination 6 and the hexagonal (10-6)
structure coordination 12 so that a volume
V will have a total number of contacts A difference with the former interpreta-
equal to TV given by tion is that this latter relation is based
upon the calculation of the local void vol-
N= 6(l-a)]/(2r) 3 (10-2) ume of a particle surrounded by particles
in its vicinity. As a consequence this latter
This volume contains a number of parti- relation can be formally applied to situa-
cles, ft, equal to tions where information on the local coor-
dination in dependence with stacking den-
13 (10-3)
sity is needed.
so that the average coordination, G, is Model experiments as have been de-
given by scribed earlier have given information on
both average stacking density and the co-
12a v / 2 + 6 ( 1 - a ) , 1 +1.828a ordination distribution of a dense random
G—
= = 6 packing. This distribution is shown in
1+0.414 a Fig. 10-9 (Wadsworth, 1960; Belik, 1989).
(10-4)
356 10 Green Microstructures and Their Characterization

Rhorabohedral

2!
i Tetragonal
sphenoktal v ^ \

e + Figure 10-8. The average number of

A
mutual contacts per spherical particle
Orthorhotnbic as a function of the overall porosity
0 of a number of specific packing struc-
Bod^ tures and the calculated values on the
centred e
# basis of the intermixing of different
+ 0^v Simple
specific packings (from Hudson,
N^ubic
1947 b).
*> 0-25 0-30 035 (H5 0-50
Porosity. P

According to Smith et al
e More probable experimental values (Wadsworth)
Regular packings
Hudson's artifice values

It is seen that the average coordination is

equal to about 8 leading to a stacking den-
sity of 0.6 being very close to the density
belonging to the dense random packing.
It should be mentioned that in green
structures a dense homogeneous packing
will generally not occur. This cannot be the
case since an actual ceramic structure is
6 8 10 12 obtained by stacking particles under the
Contacts per sphere n (a) action of strong interparticle interactions.
This will obviously lead to voids in the
structure of sizes larger than the particle
dimensions. They may continue to exist
also if the structure is mechanically agitat-
ed. For this reason we should more or less
consider the dense random packing as a
theoretical maximum and therefore as a
structure which should be reached as close-
ly as possible during ceramic processing if
it is intended to produce strong and reli-
0 2 10 12 n (b) able ceramics.
Figure 10-9. (a) The coordination density distribu- To arrive at a statistical treatment of the
tion as experimentally determined by Wadsworth coordination problem we now should try
(1960); (b) same as (a) as calculated by Belik (1989). to build up a stacking taking into account
 10.3 The Structure of Particle Packings 357

the statistical nature of the system. A sim- The expression for the relation between G
ple analysis, which also allows to some ex- and e' is found with the help of Eq. (10-8)
tent for stacking peculiarities related to and reads
prior coagulation of particles which are
= 14.4 s ' - 2 . 4 (10-10)
not easily removed by shaking, is derived
in the following way. First we define a Experiments done to find realistic values
packing lattice with a uniform coordina- for G have shown that coordinations as
tion Go which can be as high as the theoret- high as 12 are very unlikely whereas values
ical maximum. This lattice is subsequently 11 or 10 are more realistic. Further an av-
filled up with both particles and voids. If s erage stacking density of between 0.8 and
is the stacking density, there is a probabil- 0.65 is often encountered, depending on
ity that at a specific lattice point next to a the way of processing the particles in nor-
particular particle which is equal to s. If we mal ceramic practice. In Fig. 10-10 two
now fill up the surrounding of the particle distributions for the coordination, one for
under consideration with G neighbors and a dense packing and one for a loose pack-
Go-G voids, we have G0\/[G\(G0-G)\] ing are given. In the equation for the coor-
possibilities to do so, all with equal proba- dination distribution, Eq. (10-7), the actual
bility. coordination and the stacking density
There is also a probability that there are should be treated as independent parame-
G neighbors and Go — G voids given by ters. We can, however, take into consider-
ation the variations in the stacking density
The proability that a particle somewhere in dependence of the variability introduced
in the stacking, having a density £, has a by the processing technology if we allow
coordination G now is given by the proba- for a stacking density distribution cp' (e).
bility density function <p (e, G) If we integrate over all possible values of
the stacking density we arrive at the real
<p(G,e) = (10-7)
(G0-G)!G!

In this equation, G can be chosen freely but

it should not exceed 12.
The parameter s has a direct relation
with the density as

(10-8)

If we further interpret e' as the local val-

ue for the stacking density we may easily
find the distribution of the local values for
the stacking density distribution, q> (e'), if
we know the relation between coordina-
10 12
tion and density
Figure 10-10. The coordination density distribution
(10-9) calculated following the simple model presented in
this chapter.
358 10 Green Microstructures and Their Characterization

distribution of the coordination as a func- Wertman et al., 1930; Furnas, 1928/1929;

tion of the average stacking density, 8, in Messing etal., 1978), but it is in this re-
which also the variation in the structural spect of interest to indicate at least qualita-
density is taken into account: tively what kind of benefit may be attained
when dealing with multimodal stackings.
8
Suppose we have two systems of parti-
Summing up over all values of the coor- cles, each of which can be stacked up to
dination it will be concluded that normal- their random close packing individually.
ization to unity is guaranteed: The system of particles with a volume frac-
tion of coarse particles equal to a has a
solid volume equal to V01. We then have a
G 8
volume fraction of a — 1 fine particles with
Go! a solid volume of V02. Further Fis the bulk
v
(G0-G)\G\ volume of the stacking and Vo is the solid
volume of these particles.
= l(s + (l-s)f°(p'(e)de = I (f)'(e)de = 1 If the total mass of the system is M and
8 8
there is a solid density of o, it follows that
For the calculated coordination density
distribution, shown in Fig. 10-10 a varia- V01 = Moc/a; V02 = M(l-a)/a (10-12)
tion in the stacking density with a standard If Vl and V2 are the bulk volumes of the
deviation of 5% and the same average coarse and fine stackings respectively, we
stacking density is assumed. It may be con- may define the fractional solids volume
cluded from this analysis that, for statisti- v1 = V01/V1 and v2=V02/V2 so that
cal reasons, a relatively broad spectrum of V1 = Ma/(iv1 and V2 = M(l—a)/av2.
coordinations are obtained and also that a The open space left over by the coarse
slight scatter in the local average of the material is V1 — V01 = V2 and we have
stacking density has a profound effect on
this variation. It should further be empha- a)/<7 v2 = a (1 — (10-13)
sized that, in normal ceramic practice, This equation is based upon the idea
quite considerable variations in stacking that we have filled the space first with large
density are easily obtained. This variation particles and then filled the rest up with the
will have a serious influence on the stabili- fine material. Therefore we started with
ty of the structure during sintering and eas- a = 1 which by introducing small material
ily may induce zones with reduced and in- is reduced down to the point where no
creased sintering rates. Ultimately it may space is left for this material. Hence the
lead to defects and distortions of the final solution of Eq. (10-13) gives the minimum
product. value of a. When we take the values for the
random close packing we find about 0.73.
When we define the maximum solids
10.3.3 Bimodal Stackings of Spherical
content for the stacking, v, we find
Particles
To treat the problem of bimodality, we
= VJOL = v1 (10-14)
should consider the stacking of particles of
different sizes. A thorough treatment is As a consequence a theoretical maxi-
given by many authors (McGeary, 1961; mum of 0.86 is obtainable for a = 0.73. In
 10.3 The Structure of Particle Packings 359

the previous analysis we started from a 0.90 r

dense random stacking of large particles
where we gradually filled up the voids with
small (in fact infmitesimally small) parti-
cles. This latter was also stacked in a local
dense random packing order. A similar ar-
gument can be made starting from the lim-
it of stacking of small particles where we
gradually grow large particles by making
the small ones to merging together.
If 1 — a is, as before, the fraction of small
particles in the mixture, their volume will
be given by
V2 = M(l-a)/av2 (10-15)
Volume fraction of coarse spheres in mixture
The volume for the mixture is derived
along the above given argument and is Figure 10-11. The calculated fractional solids content
equal to of a random mixture of spherical coarse and fine par-
ticles.
V=Ma/(7 + M(l -a)/ov2 (10-16)
The solid volume of the particles now is
equal to ajM and the fractional solids con-
10.3.4 Sol-Gel Structures
tent, defined as G/M V is given by
A sol is a fluid colloidal system consist-
a [OL/G + (1 - GC)/V2G] ~i = v2/[l - a (1 - v2)]
ing of two or more components. A disper-
(10-17) sion of submicrometer powders in a liquid
If we now join the two analyses together is a sol but also a solution of inorganic
we conclude that the fractional solids con- polymers.
tent is given by two equations in their par- Colloidal ceramic processing is the prep-
ticular intervals from which they have been aration of sols with controlled properties
derived: and processing the sols further into ceram-
v = vja for <x(v = vmax) < a < 1; ic products with a controlled microstruc-
ture by using the possibilities of controlling
i> = i> 2 /[l-a(l-f; 2 )]
the interactions between the particles and
f o r O < a < a = a(i;max) (10-18) hence of controlling the green microstruc-
Clearly when equating both expressions ture.
we find that if we eliminate a again For the purpose of colloidal ceramic
v = V)^^-X)2 — v2v^ as it should be. processing we define a gel as a solid-liquid
Figure 10-11 gives a graphical represen- colloidal system which shows elastico-vis-
tation of this model for the system consid- cous behavior and a usually small yield
ered. Model experiments, where obviously stress at the time scale of interest for pro-
the size relation is non ideal, have revealed cessing operations. More general and elab-
indeed a high packing efficiency although orate discussions of the gel state of matter
the maximum rapidly reduces to lower val- are given by Brinker and Scherer (1990).
ues when the size ratio approaches 1. Depending on the properties of the parti-
360 10 Green Microstructures and Their Characterization

cles in a sol, their interaction and their rogels result with a relatively narrow pore
volume fraction, a sol can behave as a gel size distribution. Therefore such sol-gels
instead of a viscous liquid. are, in the form of a supported calcined
A xerogel is an air-dried gel. An aerogel thin layer, especially suitable as a ceramic
is a supercritically dried gel. Dried gels are membrane.
actually not gels according to the above In sols where there is a certain sticking
definition. Sol-gel processing concerns the probability between the particles due to
preparation of ceramic materials by first, the presence of an attraction between the
preparing a concentrated sol, second, cast- particles and an activation energy for ag-
ing of the sol in a mould for the prepara- glomeration, particle clusters form and
tion of monoliths or coating a substrate for cluster-cluster aggregation occurs. The
preparing dense thin films or porous thin appearance of a volume filling (percolating
films, third, gelation of the sol, fourth, dry- cluster) is the transition point from sol to
ing of the gel, and fifth, sintering of the a gel. When the sticking probability is high
xerogel or aerogel. (low activation energy) a gel is formed at
Sol-gel processing and the processing of an high enough particle volume fraction by
submicrometer powders have much in diffusion limited agglomeration (DLA)
common. In both cases the methods of col- and diffusion limited cluster-cluster ag-
loidal processing are applied. The differ- glomeration (DLCA). When the activation
ences are only quantitative and not quali- energy is high, more dense gels are formed
tative in many cases. However, usually the by reaction limited agglomeration (RLA)
term sol-gel process is reserved in ceramics and reaction limited cluster-cluster ag-
for a process where the sol is synthesized glomeration (RLCA) (see for example:
by hydrolysis and/or condensation of a Jullien 1988; Meakin, 1988).
metal(ion) surrounded by various ligands,
i.e. inorganic salts or metal-organic com- An agglomerated particulate gel usually
pounds such as A1(OC 4 H 9 ) 3 . Sols of sub- has a much more open pore structure due
micrometer powders are prepared by some to a reduction in the coordination number,
mixing and milling operation. from about 8 corresponding to close pack-
We use the term sol-gel for the colloidal ing, to about 3, which is an average equiv-
processing of sols with particles or poly- alent to trihedral packing (Brinker and
mers in the lower range of colloidal size, Scherer, 1990). Upon drying an open gel
i.e. < 100 nm. Now we will focus on some shrinks appreciably. The microstructure of
theoretical aspects of the structure of par- the resulting xerogel will be a contracted
ticulate gels and their xerogels. Experimen- and distorted version of the structure of
tal characterization of gels and xerogels the original gel and the coordination num-
are treated in Sec. 10.4. ber increases (Brinker and Scherer, 1990).
The microstructure of a gel is strongly Due to the often large shrinkage there is a
connected with the properties of the sol substantial risk that macrostructural de-
from which it is formed. Is the sol a con- fects such as cracks develop (Zarzycki
centrated dispersion where the particles re- etal., 1982).
pel each other strongly, the sol can behave Insight in the forming and structure of
as a gel due to electro-viscous effects. In gels can be obtained by computer simula-
these gels there are actually no solid-solid tions of their forming process by DLCA or
contacts. Upon drying densely packed xe- RLCA. The simulated structures can then
 10.3 The Structure of Particle Packings 361

be studied by for example the methods of R is given by

geometrical statistics.
The microstructure of particulate gels (10-21)
formed by DLCA or RLCA cannot be de-
scribed by the random particle packing where Rt is a cluster of size i.
models discussed before. However these With the help of the density-density
structures can be described successfully by correlation function C{r) of a gel a quanti-
the concept of fractal geometry as has been tative description of the density fluctua-
shown by Bremer et al. (1989). tions in a gel can be given. C(r) is defined
This means that between certain limits as
the structure of the gel is scale invariant.
This property of particle networks origi- C(r) = (10-22)
nates from the fact that the particle clus- Q2
ters from which the gel is formed have also The brackets denote an averaging over all
fractal properties (Jullien, 1988; Meakin, orientations and positions, x, in the gel
1988). However on a macroscopic scale and Q is the average density. The density-
particulate gels are homogeneous. density correlation function is a two point
The total number of particles TV in a correlation function (see also Sec. 10.3.6).
fractal cluster is given by On length scales greater than the particle
radius a, C(r) is the same as the radial
(10-19) distribution function.
Integration of the density-density cor-
relation function leads to the mass in a gel
where R is the radius of the cluster and a sphere with radius R:
the radius of the particle. D is the fractal
dimensionality of the cluster. D is always Mp = g J 4TTr 2 C(r)dr (10-23)
smaller than the euclidean dimension d. o
Hence the density of a cluster decreases
A fractal cluster of radius R has a mass
when the size increases. The value of D for
given by
a cluster in a dilute system or in a gel char-
acterizes its microstructure. Computer 'R\D 4
(10-24)
simulations of rigid DLCA and RLCA
clusters showed that D = 1.78 and 2.04 re-
spectively (Botet and Jullien, 1988). The The last two equations lead to a relation
clusters become more dense (higher D) between the correlation function C (r) and
when restructuring after sticking is al- the length scale r(r>a) (Bremer, 1991):
lowed or when the agglomeration is re- D D-3
versible (Jullien, 1988). C(r) = (10-25)
Bremer (1991) derived the following re-
lation between the volume fraction of par- We now see that the fractal dimensional-
ticles in a gel and the average radius of the ity can be determined by plotting C (r) ver-
clusters jRg in a gel: sus r on a log-log scale in the appropriate
r region. At large r, C (r) becomes indepen-
D-3
(10-20) dent of r for a homogeneous gel. More
information about the use of fractal ideas
362 10 Green Micrestructures and Their Characterization

in ceramics can be found in Schaefer descent properties (Pieranski, 1983). With

(1989) and Schaefer and Keefer (1986). A increasing interparticle binding energies,
lucid treatment of the gelation and gel domain interlocking decreases due to the
structure of silicate systems derived from rigidness of the clusters and the clusters
metal alkoxides is given by Brinker and themselves will stack as entities. The com-
Scherer (1985). bined effect of the decreasing domain size
and the increasing interdomain void
spaces results in the loss of iridescence.
10.3.5 Hierarchical Cluster Packing
An additional contribution to the low
The concept of hierarchical cluster packing efficiency may be the formation of
packing is strongly related to the idea that a third generation clusters. The associated
particles may cluster together depending void space is the grouping of the second
on their mutual interaction. This interac- generation of clusters.
tion can be repulsive or attractive. In elec- The most important aspect of this is that
trocratic systems the interaction between a hierarchy of microstructures can be ob-
two colloidal particles is described by the tained by varying the degree of interaction
DLVO theory. The combination of the re- between the individual particles as de-
pulsive energy and the van der Waals scribed by the zeta potential.
forces results in a pairwise potential that, A further point worth mentioning is the
in general, has a maximum either in the description of the hierarchical particle
repulsive or the attractive range. It sepa- packing from the void's point of view.
rates two minima which are always attrac- Although a void space is interconnect-
tive. ing, one can visualize this space as divided
Aksay et al. (1986) have characterized into separate pores, each bounded by a
the different states in which a colloidal sys- surface of touching spheres. Each pore has
tem can occur in terms of the effective zeta a volume, a shape, and a coordination
potential. They deviate between solid-like, number. The coordination number of a
liquid-like and gas-like colloidal structures. pore is the number of particles forming its
In such a structure a number of genera- surface boundary. It may be easily seen
tions of particle clusters are seen to be that also the hierarchical packing structure
present. Particle clusters are formed during is reflected in the pore coordination. Par-
colloidal solidification. ticularly this interpretation is important in
The first generation of clusters begins as that the reduction of voids in the structure
domains, formed by close packing of pri- during sintering is only possible when the
mary particles. The domain size decreases pore coordination is lower than some crit-
with increasing interparticle binding ener- ical value, depending on the dihedral angle
gy. The collection of a number of these defined by the free surface bounding the
clusters results in the formation of a sec- pore and the grain boundary (Lange,
ond generation of clusters. This second 1984).
generation of clusters normally shows con- It should be mentioned further that this
tinuous variations in the interdomain void interpretation in terms of hierarchical
space. When the interparticle binding ener- packing structure also has relevance for
gy is low, tight interlocking results in poly- the description of slipcasting and the sin-
domain structures which resemble poly- tering process. We will come back to this
crystalline atomic structures and have iri- point later in this chapter.
 10.3 The Structure of Particle Packings 363

10.3.6 Measurable Quantities £ = 1—0. This can be easily seen by consid-

ering that Sx is in fact the probability that
Consider again a simple particulate
a randomly chosen point is in the solid
green body composed of solid particles
phase (first order statistics). For isotropic
and gas only (bicontinuous two-phase sys-
materials Sx can be measured from images
tem).
of 2D or even ID cuts through the materi-
We are interested in the characteristics
al.
of the material which have a predictive val-
The 2-point correlation function reads
ue for the further processing and final
properties of the ceramic and which can be f x)> (10-28)
defined in a statistically geometrical way.
for an isotropic material JC = | jr | and
We already saw in Sec. 10.3.4 that knowl-
S2 (x) = S2 (x). S2 (x) is the probability that
edge of spatial correlation functions means
two points a distance x apart are both in
quantitative knowledge about the mi-
material 1. Because f2 (r) =/(r) it follows
crostructure. Let us therefore first have a
that S-y (0) = S,=d> and
closer look at the definition and properties
of these functions. lim [S2 (x)] = (10-29)
Following Berryman (1985) we can de-
when no long range order is present in the
fine a function p (r), which gives the value
material.
of some material property p at a spatial
Debye (1957) was the first to show that
position given by the vector r. For the ran-
the specific surface as of the material can
dom materials we consider p can have only
be expressed as
two values p0, for a position in the pore
space or px for a position in the solid par-
ticles. We can then define the stochastic (10-30)
function
where As is the total interfacial surface and
V the total volume of the material.
(10-26) Correlation functions can be measured
PI -Po
by small-angle scattering techniques (e.g.,
which is defined as 0 for r in material 0 Debye, 1957) and/or image processing
(void) and 1 for r in material 1 (solid). This techniques (e.g., Berryman, 1985). Which
function defines the microstructure of the technique can be used best depends,
porous material completely. among other factors, on the sizes of the
However, in practice only a few of its structural units in the material.
statistical properties can be measured, as Measuring S2 (x) is in fact equivalent to
we will see. We mention that a slightly dif- applying dipole statistics (second order
ferent random function was defined earlier statistics), i.e., placing a stick with length x
by Fara and Scheidegger (1961) leading, randomly on a plane section n times and
however, to similar results. calculating from these data the probability
The 1-point correlation function is given that both ends are in the solid or in a pore.
by Many other quantitative stereological
S1 = (f\r\) = ct) (10-27) and image analysis methods exist to mea-
sure morphological parameters from 2D
cj) is the volume fraction of component 1 images (Fishmeister, 1974; Serra, 1982;
(solid) and the porosity of the material Russ, 1990). These methods can be used
364 10 Green Microstructures and Their Characterization

with relative ease due to the present possi- strictions it is often useful to discriminate
bilities of computer-assisted microscopy. between pore necks or throats and pore
The porosity and specific surface of a voids and their respective distributions.
green body can also be obtained by other Porosity, specific surface, and pore size
methods. For example the porosity of a (neck and void) distribution are important
green body can be determined by measur- descriptors of the microstructure of a
ing the total volume VT by Archimedes green body with respect to drying, binder
method by immersing the body in a non- burnout and sintering behavior. However,
wetting liquid and measuring the volume for a prediction of the sintering behavior
of solid Vs with He pycnometry. The also other structure parameters such as the
porosity e is then e = (VT— VS)/VT. particle coordination and spatial fluctua-
Gas adsorption is widely accepted for tions therein should be known. The mean
the determination of the specific surface coordination can be determined using suf-
area. ficiently resolved micrographs of 2D cuts
When we express the specific surface of monosized sphere packings.
area as surface/pore volume ( = A/Vp), we
see that VJA has units of length. Twice this
10.3.7 Processing Technology
length is defined as the "hydraulic radius"
in Relation to Green Structures
of the porous material (for cylinders this
value is equal to the cylinder radius). Within the scope of the present treat-
Problems in the definition and opera- ment it is impossible to give a comprehen-
tional determination of porosity and sive description of the relation between
specific surface area arise when the pores green structures and the preceding pro-
become very small (a few times molecular cessing route, although a number of pecu-
dimensions) or when the pore surface be- liarities can be considered to give at least
comes fractally rough (Everett, 1988). some insight into the most important fac-
The term pore size or pore size distribu- tors controlling the microstructures in-
tion is usually used rather loosely. In fact volved.
a bicontinuous particulate material has For this purpose, we shall firstly elabo-
one pore with a very complicated geome- rate on the technique of sedimentation and
try. Defining a pore size (distribution) slip casting as a means to obtain a green
means a discretization of the pore space. structure which can be either a precursor
This can only be done in a more or less for a dense-sintered material or to yield a
arbitrary way (Dullien, 1979; Everett, material with a carefully controlled poros-
1988). The volume that is assigned to a ity. Secondly also an example of a dry pro-
particular pore size depends on the pore cessing route will be considered.
structure model and the experimental
method used (Sec. 10.4). As we will see this
10.3.7.1 Wet Processing
has severe consequences for the extraction
of pore "size" data bearing a relation to To arrive at an understanding of the re-
the real pore structure from for example lation between the green structure and the
Hg porosimetry or gas desorption mea- processing parameters we shall in the pres-
surements. ent treatment focus our attention on the
Because a pore space usually has con- interface where during consolidation the
strictions and wider regions between con- green wet structure is formed.
 10.3 The Structure of Particle Packings 365

The discussion of the particle interac-

tion will consider systems where the parti-
cle interaction is a result of van der Waals
attraction, electrostatic repulsion and, per- particle trajectory
haps, steric repulsion. Depending on the JC
_CD
relative strength of these forces and the O

interparticle distance, the resultant force is

attractive or repulsive. Generally the elec-
a
bcpv
trostatic interaction may dominate at large consolidation front
bcp
distances over the van der Waals force
time •
while steric repulsion is active only at short
interparticle distances. In the case that the Figure 10-12. Schematic representation of the con-
solidation of a particle dispersion under the action of
repulsion dominates at a considerable dis- a downward force leading to the formation of a sedi-
tance, we may conclude that the particles ment and a consolidation front.
in a suspension are not flocculated. Ther-
mal movement of the particles can over-
come the repulsive energy barrier, which We can easily show this result as in
makes the particles in the suspension floc- Fig. 10-12, where we consider three lines.
culate and large agglomerates grow in the The upper line at the left represents the
suspension. It is a well established fact that location of a specific particle as a function
a sediment formed out of a stable (not floc- of time. As soon as this particle reaches the
culated) suspension has a very dense struc- consolidation surface, it becomes immobi-
ture whereas a sediment formed out of a lized and part of the green cake structure.
flocculated suspension has a high and open The lower line represents the progression
porosity. of the consolidation surface as determined
In this former discussion some conclu- by Eq. (10-41). This latter line also sepa-
sions are drawn from the state in which the rates two areas, one with the particle densi-
suspension occurs before sedimentation ty of the suspension and one with the den-
has taken place. It is, however, also inter- sity of the cake. It is evident in this case
esting to see what is actually occurring at that the density has a discontinuity at the
the interface between the suspension and consolidation front.
the wet structure being consolidated. The Obviously the case considered above is a
actual behavior at the interface is deter- theoretical extremum. If the suspension is
mined by the balance between the sedi- not flocculated, there will be a repulsive
mentation force (gravitational or centrifu- force between the particles which have to
gal and drag force as in the case of slip be overcome to be able to densify. This
casting) and the osmotic pressure which effect may be thermally activated and
appears when large particle density gradi- therefore will require some time for the
ents are existing. particles to be forced into one anothers
It is shown in the appendix (Sec. 10.5.1) vicinity such that van der Waals forces be-
that the rate of growth, vf, of the consoli- come the dominant interaction. In any of
dated particle stacking is given by these cases particle flow and densification
are opposed and discontinuities may be
= (d<pv\ levelled out. Actually there occurs a con-
vf = (10-31)
dcp ) t tinuous relation between particle flow and
366 10 Green Microstructures and Their Characterization

density. It should be remembered, how- In the appendix the case of hindered set-
ever, that a consolidation front exists only tling due to hydrodynamic forces is
when dcp v/dcp is negative. worked out as an example.
Now we should try to find out what is In the treatment on green structure de-
actually happening at the place where ma- velopment given until now it has been
jor changes in the density are taking place. shown that particle density discontinuities
As long as the particle flow decreases may arise during settling from a suspen-
with increasing particle density a consoli- sion. This discontinuity can occur during
dation front can exist. At relatively low forced or gravitational settling as well as
densities, far away from the cake surface, under conditions prevailing when normal
this speed is low, whereas it increases the slip casting is chosen as the ceramic pro-
more we reach the cake surface. The conse- cessing route.
quence will be that under stationary condi- There is one more effect which may re-
tions, also in the case of hindered settling, duce at least to some extent the develop-
a particle density gradient will occur. Fig- ment of density discontinuities, particular-
ure 10-13 shows how in general the particle ly when the particle sizes are small. This
density discontinuity may be found when effect is directly related to the osmotic
the relation between particle settling flux pressure, and is strongest when steep gra-
and density is known. dients in this pressure just in the vicinity of
When the behavior of the particle flux is the consolidation front are present. At this
more gradual than in the example given, moment it is however more convenient to
the discontinuity is still present, but the consider this effect from the Brownian
structure will be densifying also at higher movement point of view, where also a di-
values of cp than at the discontinuity. Obvi- rect interaction between the particles other
ously this means that a green structure than an entropic effect is taken into con-
which is formed under such situations will sideration. In the appendix some basic
have a high risk for uncontrolled porosity equations which are important to describe
since the particles entering the consolida- the process are given.
tion zone may become immobilized and It is shown that in order to have stable
cannot settle to achieve a stable and dense formation of a wet green structure, the
environment. particle flow due to gravity, forced sedi-
mentation or a mechanical drag is higher
than the osmotic pressure exerted by the
particles.
When the purpose is to achieve at a ho-
density gap
mogeneous well-organized green structure,
it is favorable to work with a system which
is slightly above the limiting case of hin-
dered settling by the osmotic effect. Under
such circumstances the stacking of the par-
ticles yielding the green structure is a rela-
tively slow process where each particle is
Figure 10-13. The sedimenting particle flux as a func-
tion of the suspended particle density leading to a gap given some time to migrate into a location
in the settling density under conditions of hindered of lowest energy to give a dense green
settling. structure.
 10.3 The Structure of Particle Packings 367

Normal slip casting, using a mould of a of the flow of the liquid in which these
very fine nonleaching porous material, has particles are suspended. It has been shown
initially a very high rate of cake formation that dense and homogeneous structures
which gradually reduces. Ultimately a situ- will occur when slow settling is allowed
ation occurs where the drag balances the whereas a more open structure results
osmotic pressure. It is found that under when a cake is formed under forced condi-
such circumstances the green structure at tions. It is well-known that particles will
the mould side normally has a slightly low- behave individually when the interparticle
er density than the area directly at the sus- forces are repulsive, which means that gen-
pension side. This effect is reflected also in erally electrostatic forces or dispersion
the sintering behavior where the high den- forces due to steric interaction are suffi-
sity side sinters at a higher rate than the cient to overcome the van der Waals at-
low density site. Complicated geometries traction. The interplay between the elec-
will normally generate internal stresses and trostatic repulsion and the van der Waals
may show failure or the formation of force can result in a long-range (sec-
flaws, whereas simple initially flat green ondary) attractive minimum a few kT
bodies show bending with the convex sur- deep. A similar minimum can also result
face directed to the mould side. from the interplay between steric repulsion
As a conclusion of the foregoing de- and the van der Waals force.
scription of the consolidation process we When such a situation exists, the parti-
can state that the wet green structure is, cles will, to some extent, organize into rel-
apart from its particle density, similar to atively dense agglomerates prior to set-
the structure of the suspension. This state- tling. When a green structure is formed out
ment holds at least for dense structures, of a suspension which in this way has lost
but a porous particle stacking is also be- his stability, we will observe that the struc-
lieved to represent the state of the floccu- ture is made up of densely packed domains
lated suspension. This latter suspension which are further interlocked into one an-
contains, depending on the state of desta- other with a high degree of packing effi-
bilization, a large number of clusters with ciency. The overall structure is character-
a random but open structure. During sedi- ized by a bimodal pore size distribution
mentation these clusters may organize into where the small pores are due to the dense
a dense random packing such that a pore packing in the agglomerates and the large
structure, characterized by a high degree of pores represent the interdomain pores. The
multimodality, results. Also slipcasting of conditions responsible for stacking of the
such a structure is a rapid process due to particles individually or as agglomerates
the fact that liquid transport to the mould have been studied extensively by Aksay
is taking place rapidly. We generally ob- et al. (1986). It was found that settling out
serve that no great differences in particle of a pre-agglomerated suspension rapidly
stacking structure exist between the upper forms a thick cake with an open structure
and lower side of the cake. while settling from a stable suspension oc-
So far we have assumed in our consider- curs slowly and results into a very dense
ations that particles in a suspension may green structure. The same is found when
react independently of the particles in their applying normal or pressurized slipcast-
direct vicinity. Only mutual interactions ing. When the intention is to make a dense
are dealt with due to the specific behavior and thin cake, a stable suspension is nor-
368 10 Green Microstructures and Their Characterization

mally used. An unstable suspension gives a body may be found due to macroscopic
thick and more porous sediment. It is im- forces introduced by friction effects of the
portant to note in this respect that the pressing dies or nonproper dimensions of
agglomerates generally are not strong the green body inhibiting a well developed
enough to support the drying forces so that flow of the powder compact. This effect
extensive reorganisation of the green struc- can, however, not be dealt with from a
ture normally is very likely to occur. This green structure point of view but should be
effect greatly obscures the peculiarities of solved by a proper choice of the conditions
the green structures so that it is very hard at which the shaping technology is being
to draw any conclusion from the structural done.
characteristics determined in the dried Hot isostatic pressing is important from
cake. the sintering point of view. It may be
shown (Veringa, 1993) that structural reor-
ganization takes place in such a way that
10.3.7.2 Dry Processing
any variation in the green packing density,
It is very difficult to describe the struc- which actually is reflected in a variation of
tural characteristics obtained by dry press- the coordination number, is amplified by
ing. Normally the structure is a densified the same effect responsible for the overall
representation of the powder structure. It densification due to sintering. However,
is common practice to organize the powder the accelerated densification due to hot
particle into weak agglomerates to ensure pressing does not influence the reorganiza-
good flowing properties when the pressing tion rate, so that overall sintering can be
dies have to be completely filled with the speeded up to such a level that any struc-
powder precursor. Generally also a lubri- tural reorganization has unsufficient time
cant is added to allow to some extent the to develop properly. For this reason a hot-
particles to slide along one another during pressed material has a high density, a low
pressing. The application of a high me- variation of grain size on both microscopic
chanical load to the particles will destroy and macroscopic scale.
the large agglomerates and therewith make
the large pores disappear. Studies have re-
10.3.7.3 Drying and Sintering
vealed that the green structure of such a
pressed compact does not show any signif- Although a number of mechanisms have
icant structural detail which is a conse- been presented to gain some insight into
quence of the preagglomeration. It should the effects which are operative during
be emphasised, in this respect, that the rel- forming of the green structure, we will ob-
ative distances which are travelled by the serve in many cases that the drying and
particles during pressing are very limited. sintering completely reorganizes the struc-
As a consequence it should be concluded ture such that only a few "footprints" are
that structural peculiarities, when present, left behind to reveal some peculiarities of
are to be found at a scale of some times the the green structure.
particle size, which is also the scale at During drying normally high internal
which variations do occur due to the statis- capillary forces are developed which will
tical nature of the packing process. It is make a dense green structure shrink and in
further worth mentioning that structural the worst case, where drying occurs with
variations at a scale of the size of the green great inhomogeneity, extensive flaws will
 10.3 The Structure of Particle Packings 369

develop. These flaws can easily attain a

size which is a large fraction of the entire
ceramic body, so that we must conclude
that the capillary forces are quite consider-
able and at least strong enough to break up
or shrink particle agglomerates. Due to the
fact that capillary forces are higher in areas
with small pore sizes and also because wa-
ter is more difficult to remove from low
porosity volumes, we normally expect a
high contraction of high density zones and Figure 10-14. Schematic representation of a water
bridge between two adjacent spherical particles lead-
relatively low shrinkage of low density ing to interactive drying forces.
areas. In the next geometric analysis we
will investigate the relevant parameters
which are of importance for establishing
the dry green structure and the possibility of the force on the relative interparticle
of instabilities to occur. distance which is shown in Fig. 10-15.
The pressure p is the pressure which ex-
We start from an array of monosize
ists under the surface of the liquid and is
spherical particles and assume complete
directly related to the external water vapor
wetting. Obviously this approach is an
pressure, P by
oversimplification of the real situation, but
it can be shown that assuming a three di- = Qy(l/r1-l/r2) =
mensional stacking with particles differing (10-34)
in size and also nonperfect wetting will re-
where P^ is the vapor pressure above a flat
sult in a more complex behavior while the
surface and Q the molecular volume. We
main conclusions are essentially the same
as in the present case. The particle array
and the relevant geometric variables are
shown in Fig. 10-14. The equations de-
scribing the capillary interaction are

F= -pnx2 + 2ny x2/R (10-32)

where p is the water pressure, y the water
surface tension, d the interparticle distance
and F the force between two adjacent par-
ticles which are connected by a water
bridge. Further we can write for x • - d/R
Figure 10-15. The dimensionless interactive force be-
(10-33) tween two adjacent spherical particles as a function of
the dimensionless interparticle distance for different
Solving x and r2 and substituting the values of the normalized hydrodynamic pressure in-
result into the equation for the interparti- side the water bridge. The values are calculated on the
cle force, we find a two-valued dependence basis of a simple model presented in this chapter.
370 10 Green Microstructures and Their Characterization

can interprete the behavior of the particles sult into an enhanced restacking effect of
in the stacking in the following way. At a the structure during drying.
certain interparticle distance less than The last step in the processing of a ce-
1.36 x R and a value of the partial vapor ramic body is the sintering to obtain the
pressure less than the saturation pressure required solid structure and also this sin-
we see that two values for the force result. tering obscures the peculiarities of the par-
The highest force corresponds to a large ticle stacking very severely. In very much
water bridge between the two particles, the same way as during drying, reorganiza-
which we call the wet state and the low tion of the structure will take place. This is
value represents the small water bridge or most easily understood by considering the
the dry state. Most important is, however, free energy of the structure. It is shown by
that both conditions occur at one and the Veringa (1993) that the free energy, U, of a
same partial vapor pressure. Obviously particle in a stacking, which is locally or
when drying, the system must move from globally densifying, can be represented by
the wet to the dry situation, so that after
U(r)=U0-ynR2G(i-l/z2(r)) (10-35)
this shift no restacking is likely to occur.
The wet state therefore is most important where the second order contribution due
and it is seen that interparticle forces in- to the increasing coordination during den-
crease when both the particles and their sification is neglected. G (r) is the value for
distances get smaller. We may therefore the local coordination and z expresses the
conclude that high density clusters have a relative amount of interpenetration of the
tendency to shrink more than average in a particles during sintering as explained in
stacking, on the one hand because of the Fig. 10-16. If we introduce <ps(r) for the
small interparticle distance and on the oth- local particle stacking density, which is at
er hand due to the fact that evaporation of least a function of the position in the mate-
water is slower than average so that drying rial under consideration, we may integrate
forces relax at a later moment than the over the entire volume and also assume
more open structure. Further it is seen that that there exists a variation in the local
for interparticle distances higher than densification expressed by: z(r) = z
1.36 x R, no solution exists other than the and find as a first approximation
complete wet state and no interparticle
forces will develop. Most important is the
fact that the wet structure does show some
ability to deform in order to comply with (10-36)
the structural changes in the transition
zone.
Another aspect which can be worked
out on the basis of this, oversimplified,
model is the behavior of the drying front at
the very moment that the local structure
moves from the wet to the dry state and
therewith the development of variations in
the stacking pressures when the drying
front tends to bend due to global varia- Figure 10-16. Representation of two interpenetrating
tions in the stacking density. This can re- spheres during sintering of a particle stacking.
 10.4 Characterization Methods 371

So that as a first approximation it may be former analysis a more gradual variation is

written that assumed and overall densification is not a
priori necessary. It should be mentioned,
= $U0(ps(r)-ynR2$G(r)(psdr however, that the restacking process is
speeded up when the structure as a whole
00-37) is allowed to densify.
This latter effect is shown most dramat-
With the auxiliary condition that ically in the case of a membrane as shown
in the Figs. 10-6 and 10-7.
Here a suspension is filmcoated on a
nonsucking, but porous, presintered ce-
we can conclude that the structure can, ramic substrate and also transformed into
also when no global densification is al- a solid body under exactly the same condi-
lowed (z0 is constant), reduce the free ener- tions. These two dispersions are dried and
gy when it densities locally where G (r) and sintered together but at a temperature low-
cps(r) are highest and compensates by an er than the presintering temperature of the
expansion of the structure in areas where substrate. What we finally see is that the
the coordination and stacking density is mechanical constraint of the substrate im-
lower than average. posed on the thin filmcoated layer com-
The final conclusion is that structural pletey suppresses any dimensional changes
changes during the initial phase of sinter- and also the extensive restacking of the
ing will occur in such a way that the high particles, while the material itself does
coordination and/or density areas shrink show sufficient neck-formation to become
faster than areas with low coordination a strong structure. The material originat-
and/or density. Since the density is closely ing from the thick body, on the other
connected with coordination, we will see hand, has undergone extensive densifica-
that the densest areas densify fastest dur- tion and particle growth.
ing the initial phase of sintering.
The same sort of analysis is given by
Lange (1984) who treats the problem of
restacking from the void's point of view. It
10.4 Characterization Methods
has already been shown that the pore
Most techniques used for the characteri-
structure, as described in term of coordina-
zation of green bodies are also used for the
tion, is dictated by the hierarchical particle
characterization of fired bodies and are
packing. It further is found that all pores
well-known in other areas of materials sci-
will initially undergo some shrinkage after
ence. It is assumed that some basic knowl-
necking during sintering, but only those
edge of the methods to be discussed below
which have a coordination lower than
is already available.
some critical value will have the thermody-
namic potential to disappear. Pores with
10.4.1 Types of Green Bodies and Usability
coordination larger than this critical value
of Characterization Techniques
will shrink to an equilibrium size. Obvi-
ously this analysis starts from the assump- Green bodies have already been defined
tion of a very high variation of coordina- and a distinction is made between saturat-
tion over the entire structure, where in the ed and unsaturated bodies. Now we will
372 10 Green Microstructures and Their Characterization

elaborate this somewhat further in order to When the interparticle space is filled with
be able to classify green bodies and to dis- polymer, intrusive techniques cannot be
cuss the suitability of well-known experi- used. A relatively new technique that may
mental techniques for each of these classes. be used at all PVC is atomic force micros-
The concept of a critical particle volume copy (AFM). However this technique is
concentration (CPVC) has been intro- limited to the (fracture) surface of the spec-
duced in accordance with paint science imens.
(Bierwagen, 1992). The CPVC is the parti- Above the CPVC Hg intrusion and gas
cle volume fraction in a compact where adsorption can be used when solid poly-
there is just enough liquid or polymer ma- mer fills the voids partially. Compression
trix to wet and completely fill the voids of the polymer and distortion of the parti-
between the particles. Below the CVPC the cle packing due to high Hg pressures may
fluid or polymer phase is continuous and complicate the interpretations of the mea-
the particles are randomly dispersed in this surements.
matrix. Above the CPVC there are void Xerogels and dry compacts obtained by
structures filled with gas or vapor due to colloidal filtration passed the CPVC dur-
insufficient liquid or polymer. The parti- ing the transition from the constant rate to
cles are still continuously connected. the decreasing rate drying period. At this
At the CPVC mechanical, transport, point air intrudes the body and the PVC
and optical properties of the green com- stays constant with increasing unsatura-
posit material change drastically. tion. These materials can be characterized
For a completely dispersed powder, with all above mentioned techniques.
where all the particles can act individually, However, Hg porosimetry may suffer from
the densest random packing of the particu- the same drawbacks as mentioned above
lar powder is the CPVC. In practice, for a and microscopic techniques demand spe-
particular system the CVPC will appear to cial sample preparation techniques which
be lower due to flocculation, aggregation, may introduce artefacts as well.
phase separation etc. In order to classify porous bodies ac-
In most ceramic systems the ceramic en- cording to their pore size range we follow
gineer wants an as high as possible CPVC. the generally accepted classification ac-
The CPVC concept was used in ceramics cording to IUPAC (1972):
by Pujari (1988) for injection-moulded
a. macropores: d > 50 nm
green materials and by Nahass et al. (1992)
b.mesopores: 2 nm < d < 50 nm
for green tapes. These materials are slight-
c. micropores: d < 2 nm
ly above CPVC.
Investigation of the interparticle space This classification is mainly based on ex-
of green bodies at or below the CPVC is perience with gas adsorption and mercury
not possible with intrusive techniques as intrusion measurements. Notice that a
mercury porosimetry or gas adsorption. green body can have macropores due to a
When the interparticle space is filled with particular microstructure! With the for-
liquid the body can be investigated by per- mer, due to capillary hysteresis, pores in
meability measurements, small angle scat- the mesopore range can be measured. Mer-
tering techniques and NMR. Optical and cury intrusion techniques are useful in the
EM techniques can be used after preparing macropore range. In our discussion of sub-
specimens by fast freezing techniques. micrometer and sol-gel green bodies the
 10.4 Characterization Methods 373

mesopore region is especially important alumina suspensions had still an apprecia-

for xerogels and the low macropore region, ble closed void density which shows up in
i.e., 50 nm < d < 300 nm is important for optical micrographs of polished sections of
submicrometer green bodies. With sol-gel material sintered for 1 h at 1500°C and at
techniques also microporous supported a magnification of 50. Similar results with
membranes for gas separation can be pre- green bodies were obtained by Uematsu
pared. In such membranes mesopores and et al. (1991). These authors describe a fast
macropores are defects and should be ab- method for the evaluation of macrostruc-
sent to guarantee proper working of the tural entities in green ceramics. They used
membrane. A discussion of the characteri- refractive index matching between alumi-
zation of microporous materials is, howev- na and organic liquid imbibed in the pores
er beyond the scope of this chapter. Much of the green body by immersion of thin
information can be found in the proceed- specimens (0.2 mm). Structural features
ings of the IUPAC symposia: Characteri- like defects of the order of 5 |im could eas-
zation of Porous Solids I (Unger et al., ily be detected. Hence, the resolution is
1988) and II (Rodriguez-Reinoso etal., better than that of more sophisticated
1991). methods as X-ray tomography and ultra-
sonics (Friedman, 1987). We think that the
method of Uematsu etal. (1991) can be
10.4.2 Macrostructure and Texture
improved appreciably when a confocal
The microstructure of a green body is scanning laser microscope (CSLM) is used.
important for the microstructural develop- Acoustic microscopy at ultrasonic fre-
ment during firing and hence for the mi- quencies in the range 10-1000 MHz can be
crostructure of the final ceramic. The used to detect macrostructural features in
transport properties of a porous ceramic optically dense materials. Slabs up to
and the mechanical properties of a dense 2 mm thickness can be investigated with a
ceramic material depend on the mi- resolution of typically 50 um, although of-
crostructure. However, when the material ten features smaller than this size can be
is textured, or shows macrostructural enti- detected (Briggs etal., 1982). We do not
ties as large aggregates or voids, these may know application of this technique to
dominate the mechanical properties and green ceramics.
transport properties. Also, since many Other techniques such as sound velocity
techniques for determining microstruc- and attenuation measurements can be used
tural properties determine an average over to detect spatial variations in density (vol-
the whole specimen, one should know ume fraction) in concentrated suspensions
whether a sample has texture or a and sediments like wet green compacts
macrostructure. (McClements, 1991).
Macrostructural features are usually un- With imaging techniques as computer-
wanted and often caused by improper pro- ized axial tomography (CAT) scan and
cessing. Such structural defects are easily magnetic resonance imaging (MRI) real
overlooked when looking only at the scale time imaging of the green body forming as
of microstructural details with for example well as the evaluation of the green body
SEM. This was shown clearly by Correia after compaction is possible. Mapping of
et al. (1989) who found that ceramic parts porosity and binder distribution in green
prepared by slip casting of submicrometer bodies with a resolution of several hun-
374 10 Green Microstructures and Their Characterization

dreds of jjm can be done. Work is in pro- values of the calculated pore structure
gress to improve upon this (Bridger and parameters and the real pore structure is
Massuda, 1990; Ellington etal., 1987). not too strong. The incorporation of per-
Texture, such as density gradients due to colation ideas, fractality, and simulation
gradients in packing density of the struc- of pore networks in models underlying the
tural units in bodies prepared by, for ex- characterization methods may improve
ample, colloidal filtration, can be traced by upon this.
y-ray attenuation as shown by, among oth- It is illustrative to give a brief descrip-
ers, Schilling and Aksay (1988). The reso- tion of current practice in green mi-
lution depends on the beam diameter, crostructure characterization as it ap-
which was 3.2 mm in the work of Schilling peared in recent ceramic literature.
and Aksay. Rhodes (1981) studied agglomerate and
particle size effects on the sintering of
yttria-stabilized zirconia. They determined
10.4.3 Green Microstructures,
only green densities.
Current Practice
An interesting study on the packing of
In ceramic processing science the study monosized silica spheres by gravity sedi-
of green microstructures is seldom a sub- mentation was that of Sacks and Tseng
ject in its own right. Usually powder pro- (1984). Green compacts obtained from sta-
cessing by some compacting technique and ble and coagulated sols were investigated
structure-function relations are the main with SEM, mercury porosimetry and bulk
subjects. Experimental methods used in ce- density measurements. Green body sur-
ramic science and technology on a more or faces were coated with a thin gold-palla-
less routine basis for the characterization dium layer for SEM analysis. The bulk-
of green microstructure are: density was determined by measuring sam-
ple weight and geometric dimensions. The
• Density determination by Archimedes
specific volume distribution of pore radii
method.
was determined by applying the Washburn
• Determination of open porosity by Hg
equation fdr liquid penetration into a
porosimetry and the apparent pore size
cylindrical capillary. Standard values were
distribution presupposing cylindrical
used for the mercury surface tension. SEM
pores.
results showed that the green compacts
• The use of EM on a qualitative basis. from stable sols were polycrystalline col-
• Mesopore characterization by gas ad- loid crystals with many crystal and grain
sorption hysteresis. boundary defects and that no ordering
The use of simple pore models, such as other than at a short range occurred with
cylinders or slits, and some assumptions the coagulated sols. The pore size distribu-
about contact angle and wetting, in the tion of the compact from the coagulated
interpretation of mercury intrusion or gas dispersion was highly bimodal. From these
desorption measurements, is current prac- results and the SEM pictures the authors
tice. This practice is valuable in ceramic conclude that the bimodality is caused by
processing for the comparison of similar inter and intra agglomerate pores in the
green bodies and to evaluate the process- case of the "coagulated" compact and by
ing performance. However, one should three-particle and four-particle pore chan-
realize that the connection between the nels in case of the "stable" compacts.
 10.4 Characterization Methods 375

Yeh and Sacks (1988 b) investigated the dried at 120 °C at a heating rate of 5°C/h.
effect of the particle size of sub-microme- Parts of the green compacts were used for
ter alumina powders on the sintering. SEM investigation of the packing struc-
Green bodies were obtained by slip casting ture. Hg porosimetry was used for green
well-dispersed suspensions of powders density determinations and pore size distri-
with narrow (NSD) and broad (BSD) par- bution. The Hg porosimetry procedure is
ticle size distributions, both with a median described in more detail than usual:
Stokes diameter of 400 nm.
- 2 g compact divided into 10 pieces was
With the Archimedes method using wa-
used.
ter as the liquid medium the density of the
- The pressure was increased in 220 steps
compacts was determined to be 65 % and
to 414 MPa (3-4 nm pores).
73% of the theoretical density for the
- Constant pressure was attained at each
NSD and BSD respectively. The median
step.
pore diameter of the green NSD compact
- Surface tension Hg 485 mN/m, contact
was found to be ~100nm and the BSD
angle 130°.
compacts ~58nm. Yeh and Sacks used
- Exactly the same procedure was used for
mercury porosimetry for this without men-
each measurement.
tioning how they analyzed their intrusion
data. This practice shows that Hg This description is important in the dis-
porosimetry is considered as an established cussion of small differences in pore size
technique needing no further discussion. distribution found for the various consoli-
Yeh and Sacks mention further that their dation techniques. It is remarkable again
green bodies still contain local packing that they do not mention assumptions
density gradients and packing "defects", about pore geometry made in the analysis.
which will influence sintering behavior and Due to the careful procedure followed dif-
microstructure evolution. No further evi- ferences in pore size distribution found
dence was given for this except for a re- could be ascribed to differences in particle
mark about SEM observations (Yeh and packing in the green compacts.
Sacks, 1989). In a subsequent study Yeh Bellosi et al. (1990) studied the forming
and Sacks (1988 a) used the same methods of alumina-zirconia composites by slip-
to investigate the processing performance casting. Stable suspensions, as evaluated
of a fractionated submicrometer alumina by rheological measurements, were pre-
powder. pared by ball milling and using a ligninsul-
In an excellent systematic study Roosen fonate as a dispersant. Wet green bodies
and Bowen (1988) investigated the effect were in situ characterized by measurement
colloidal compaction techniques such as of the casting rate constant defined as
colloidal pressing, vacuum filter casting, V= Vo + Kt1/2 where V is the volume water
and centrifugal casting and isostatic dry absorbed by the mould at time t. Vo is the
pressing have on the green microstruc- initial value of V related to the air gap
ture and sintering of particle compacts above the suspension and below the mould
obtained from unclassified powder (d= and ^ i s a casting rate constant normalized
610 nm; as = 7.9 m2/g) containing agglom- to the unit area of the mould-slip inter-
erates and a classified powder (d= 380 nm; face. For fixed values of the solid phase K
tfs = 11.4m2/g). The green compact mi- depends only on the permeability of the
crostructure was studied on specimens cast. Lower values of K denote higher
376 10 Green Microstructures and Their Characterization

packing density. Dry green bodies were isostatically pressed compacts of a binary
evaluated for pore structure and particle mixtures of a spherical alumina powder c
packing with classical Hg porosimetry and (6 jim < d< 7 jim; D50 = 6.3 jim) and a sub-
SEM. micrometer alumina powder / (d < 1 |im;
Cao etal. (1988) investigated the pore Z)5O = 520nm) was made by Taruta etal.
structure of silicate xerogels with Hg (1990). Ratios cjf'm the range 1-10 were
porosimetry using the Washburn equation investigated. They found their results were
and N 2 adsorption using a modified Kel- in reasonable agreement with the Furnas
vin equation. particle packing model. They concluded
Also Kunze and Segal (1991) used stan- further from the bimodal pore size distri-
dard Hg porosimetry and gas adsorption butions, known particle sizes, and SEM
to investigate the modification of the pore observations that the packing structure of
structure of sol-gel derived ceramic oxide the compact of c\f—10 is an intermediate
powders by water-soluble organic addi- structure between rhombic and cubic
tives. Only calcined (973 K) gel powders packing. The green compacts with c\f be-
were studied. Xerogel powders which are tween 9 to 5 had a wide pore size distribu-
partially saturated with watersoluble or- tion with dense and loose packed regions.
ganics can probably not be studied easily The number of large pores decreased with
because of outgassing problems due to the an increasing amount of fine particles. The
additives and also because interpretation compacts of c/f between 4 and 0 were
of results would be difficult. found to have a narrow pore size distribu-
Gauthier and Danforth (1988) studied tion, in which large pores were not found.
the packing of particles from monomodal The agreement found between mercury
and bimodal silica sphere dispersions by pore sizes and particle packing consider-
colloidal filtration. They aimed at deter- ations may be qualitatively correct but
mining the effect of the size and size distri- may change somewhat when a more so-
bution of model ceramic powders on the phisticated analysis of mercury intrusion
packing behavior of the resulting green data would be used. We mention that also
bodies. The green cakes were characterized Onoda (1976) analyzed the behavior of
with gas adsorption-desorption from green binary particle packings of course
which the surface area and pore size distri- and fine powders and their sintering be-
butions were determined in the standard havior with respect to packing, pore struc-
way. SEM was used to observe fracture ture and sintering.
surfaces of green compacts. Wet cakes Recently several researchers (Young
were characterized by measuring the cake etal., 1991; Katsuki etal., 1992) reported
forming constant in a similar way as Bel- on the preparation of alumina ceramics by
losi's (1990). They observed that the per- gelcasting. With this method an organic
meability (cake forming rate constant) and polymer gel filled with alumina particles is
pore size distribution decreased with in- formed usually in a mold. After removing
creasing volume fraction. Their results of the mold the green body obtained is
agreed with Furnas model of random dried, calcined and sintered. Young et al.,
dense packing of binary mixtures. interested in preparing dense ceramics,
An interesting attempt to apply particle looked at some macro structural aspects as
packing principles, density measurements, knit lines and air bubbles in their green
Hg porosimetry (classical), and SEM to bodies. These must be absent before fur-
 10.4 Characterization Methods 377

ther microstructure studies become useful. The wet compact can only be studied by
Katsuki et al., interested in the prepara- using cryo-techniques. Very fast freezing of
tion of porous ceramic spheres (1-3 mm), the body is necessary in order to avoid
investigated the pore size and porosity of artefacts due to water crystallization
sintered gels with classical Hg porosime- (Menold etal., 1976; Luckham etal.,
try. Microstructural aspects of the wet and 1983). After freezing, fracture surfaces can
dry gel were not looked at. be investigated in a cryo-SEM or after
slight sublimation of water from the (frac-
10.4.4 Imaging Techniques ture) surface a replica can be obtained and
Direct observation of green bodies with observed with EM (Stewart and Sutton,
electron microscopy is a powerful and fast 1984).
tool to evaluate microstructural proper- Fracture surfaces of dried, calcined, or 1
ties. Details of about 5 nm can be resolved stage sintered particle compacts can be
with SEM and even better resolution can studied after depositing a gold/palladium
be obtained when field emission SEM is thin layer in the same way as for non-
used (Kumar et al., 1992). The depth of porous materials. The preparation of pol-
field is rather high (150 |im) with SEM, ished specimens or microtome slices for 2D
which makes this technique very suitable quantitative microscopy is much more dif-
for the observation of rough surfaces. ficult. Most of the techniques published
Studying 2D cuts on polished specimen (Takasu et al., 1990; Weeks and Laughner,
such a depth of field is not desirable, espe- 1987; Spurr, 1969; Pickles and Lilley,
cially when quantitative image analysis 1985) rely on the impregnation of the po-
methods are applied. The resolving power rous compact with a low viscosity resin.
of TEM is generally higher than SEM This impregnation is done under vacuum
(0.3 nm) but the depth of field is much to fully infiltrate the material with the
lower, about 2 jam, depending on the mag- liquid when gradually the pressure is ap-
nification. Microstructural analysis of plied and the whole compact is cured at
green bodies by TEM investigation of (se- elevated temperature. Microtomic sections
rial) ultramicrotome sections is possible can be made for detailed study involving
but very tedious (Kerch and Gerhardt, SEM or TEM. These slices are analyzed to
1989; Pickles and Lilley, 1985). Very re- evaluate the stacking features. Where silica
cently Kerch et al. (1993) showed that us- stackings are to be investigated, also the
ing the technique of defocus contrast TEM ceramic can be dissolved with an aqueous
it is possible to measure size and distribu- solution of HF (Takasu etal., 1990). The
tion of mesopores in xerogels. resolution of this method is, however, lim-
Imaging of green ceramic structures ited to structural features larger than
with SEM or TEM is more delicate and about 50 nm due to restrictions imposed
difficult than that of their dense ceramic by the deposition of the current conduct-
counterparts. This is due to the fragility of ing phase on the polymer surface necessary
most green bodies necessitating special for electron microscopic studies.
preparation methods. The preparation of Polished fracture surfaces of impregnat-
green body specimens for SEM becomes ed and cured samples are more difficult to
more difficult in the order: 1. porous prepare for "dried only" compacts than
bisque body; 2. calcined body; 3. dried for slightly sintered compacts. This is due
green body, and 4. wet compact. to the hard structural units in ceramic
378 10 Green Microstructures and Their Characterization

green bodies, which easily break out dur- AFM has already proven to be useful for
ing polishing and damage the structure. optimization studies of highly selective gas
This is the reason that usually only partial- separation membranes (Fritzsche et al.,
ly sintered samples are investigated by ce- 1992; Dietz et al., 1992). Additional inde-
ramic researchers. Although the heat treat- pendent measurement of lateral friction
ment causes reorganization of the particle forces allows for chemical differentiation
packing, the original packing structure can of the topological information obtained by
be traced back at least qualitatively. It re- the normal force measurements (Meyer
mains, however, very hard to draw conclu- and Amer, 1990; Overney et al., 1992). An
sions on the actual packing mechanisms additional advantage of AFM above SEM
involved to generate the green structure. is that direct measurement of colloidal
Determination of statistical microstruc- forces appears to be possible (Ducker
tural quantities from images is possible in etal., 1991; Meagher, 1992). Preliminary
principle. This is however not practiced yet investigation of our own ceramic mem-
in ceramics. An advantage of quantitative branes with AFM showed promising
computer-assisted microscopy is that the structural detail and we think that AFM
comparison of micrographs may become will be a valuable tool for the investigation
less subjective. However the problem of of microstructural evolution and interac-
obtaining objective binary images (e.g., tion forces during processing of sub-
/ = 0 : p o r e ; / = l : solid) from digital grey micrometer and sol-gel ceramics in the
value images is not resolved satisfactorily near future.
(Vivier etal., 1989). Better sample prepa- To conclude this section we mention
ration methods where contrast variation that recently Betzig and Trautman (1992)
artefacts are avoided are needed. described a scanning probe method called
Using image analysis differences in first near-field-scanning optical-microscopy
and second order statistics, which can also (NSOM). The near field optical interaction
be percepted visually, can be expressed in between a sharp probe and a sample of
numbers and therewith benefit for exam- interest is used to image or spectroscopi-
ple quality control. Variations in higher cally probe the interface at a resolution
order statistics cannot be percepted visual- down to 12 nm.
ly but nevertheless determine the mi-
crostructure and hence transport and sin-
10.4.5 Capillary and Fluid Flow Techniques
tering properties of the material. Quantita-
tive image analysis will be used in the near Capillary phenomena in porous materi-
future to elucidate also these aspects of als are pore-size-related as expressed by
microstructure. the Kelvin relation in the case of capillary
No sample preparation is required for condensation and via the Laplace equation
atomic force microscopy (AFM). Images in the case of fluid displacement processes.
of (fracture) surfaces from atomic scale up For several decades capillary pressure
to the macroscopic level can be obtained curves and gas adsorption isotherms have
(Binnig etal., 1987; Radmacher etal., been used to obtain information about the
1992; Fritzsche et al., 1992). Besides AFM pore size distribution of porous media in
can be used for structural investigation of the macropore and mesopore regions, re-
all kinds of green bodies, including wet spectively. In the traditional interpretation
ones. In the case of polymeric membranes the pores are considered as independent
 10.4 Characterization Methods 379

and capillary condensation isotherms and be used for pore size determinations is an-
capillary pressure curves are directly con- other matter of endless dispute (Ro-
nected to a pore size via the Kelvin equa- driguez-Reinoso etal., 1991). Traditional-
tion using the desorption isotherm and ly the desorption branch was used because
corrections for multilayer adsorption or it corresponds to the adsorbate condition
the Laplace equation using the Hg intru- with the lowest free energy (Lowell and
sion curve. Further some assumptions Shields, 1984). However Everett (1988)
about pore shape, e.g., cylindrical, slit- pointed out that network effects play no
shaped etc., are made (e.g., Adamson, role for the adsorption branch and hence
1991). In the case of gas adsorption/des- this branch should be considered for pore
orption hysteresis loops (de Boer, 1958) size calculations. In fact information of
discriminated between several characteris- both branches is needed as becomes clear
tic shapes and associated them with the from percolation models. It is clear that
individual pore morphology. The "de Boer using the desorption branch and the tradi-
classification" is outdated now and re- tional interpretation the most frequent
placed by the IUPAC classification into pore size is essentially correct but the
types H1-H4 (Gregg, 1986). broadness of the distribution is not.
For 35 years now it has been recognized A relatively new technique related to the
that pore blocking effects due to the fact capillary condensation method is perm-
the pores in most systems form a void/neck porometry (Cuperus, 1990). In this tech-
network are very important in determining nique the gas flux of a gas-vapor mixture
the shape and hysteresis in the isotherms is measured at a constant small pressure
and intrusion/extrusion curves (Everett, difference across a mesoporous sample,
1958; Barker, 1958). Hence, realistic pore while varying the relative vapor pressure
size information can be obtained only by from 1 to 0. At P/P0 = l all pores are
taking these effects into account. But it is blocked due to capillary condensation of
only relatively recently that the gas desorp- the vapor. On lowering the vapor pressure
tion and mercury intrusion processes have there is a P/Po where a percolating path
been recognized as a percolation problem for gas transport develops. The size distri-
and that network/percolation models have bution is calculated similarly to the tradi-
been developed (Mason, 1988a,b; Parlar tional interpretation of capillary desorp-
and Yortsos, 1988, 1989; Zhadanov et al., tion and is therefore a void-weighted nar-
1987). rowest constriction distribution. The tech-
Despite these new developments the tra- nique is especially useful for characteriza-
ditional interpretation is still general prac- tion of ceramic membranes. The same is
tice in ceramics, it stays useful when only true for gas imbibition porometry. With
relative comparisons are the goal, al- this method the gas flow through a sample,
though the "pore sizes" obtained are at initially completely saturated with a wet-
best a crude estimation of the real pore ting liquid, is measured as a function of the
structure in case of void/neck pore net- pressure across the sample. The flow-pres-
works. The status of the Kelvin equation sure curve is recalculated as a flow-pore
in the interpretation of gas adsorption- size curve in much the same way as in tra-
desorption data is still an open question ditional mercury porosimetry. Percolation
(Sing, 1989) and whether the adsorption or theory is needed for a more realistic inter-
desorption branch of the isotherm should pretation in most cases.
380 10 Green Microstructures and Their Characterization

Phase transition porometry is based on 10.4.6 Mercury Porosimetry

the microcalorimetric (Brun et al., 1976;
Quinson and Brun, 1988; Eyraud et al., The easiest tool for determining the pore
1988) or dilatometric (Enustun et al., 1985, structure, but which is however afflicted
1990) analysis of solid-liquid transforma- with a number of problems in interpreting
tions in porous materials. The method is the results, is the mercury porosimetry
based on the freezing (or melting) point (Lowell, 1979 a, b; Drake, 1945; Mishra,
depression of a liquid in small pores (1.5- 1988; Park and Ihm, 1990; Tsakiroglou,
150 nm). The correctness of the models 1990, 1991). Since mercury does not wet
used is still a matter of debate, but the the pore surface, an external pressure must
method is useful in the same way as tradi- be applied, while at the same time the in-
tional gas adsorption and Hg intrusion trusion of mercury into the material is
measurements are. An additional advan- recorded. By assuming that the pores are
tage is that wet systems can be studied, cylindrical, the pore radius is related to the
although the pore fluid must be very pure. applied pressure by the Washburn equa-
Because Hg porosimetry takes such a tion (Lowell, 1979).
prominent place in pore size characteriza- The surface tension is usually taken as
tion in prefired ceramics, we discuss below 0.484 N/m and the mercury-solid contact
recent developments in more detail. angle ranges between 130° and 150°. De-
Permeation of fluids through a green spite the widespread application of this
body is directly related to the structure of method, it suffers from several inherent
the body. The permeability of the material disadvantages. A number of these are:
as defined by Darcy's law contains mi-
crostructural information in a strongly • The pore size measured is that of the
convoluted form. Many attemps have been smallest constriction in a pore and cer-
made to relate the permeability to pore tainly not the average radius. Therefore,
properties as porosity and pore size. Well- application of the correct equations
known and successful in this respect from without taking into account the possible
a practical point of view is the Kozeny Car- variation of one single pore shape will
man equation, which works quite well for lead to incorrect analysis of the pres-
random particle packings. Philipse et al. sure-volume data.
(1990) used casting rate constants as a rel- • A high pressure exerted on some sam-
ative measure of the wet cake permeability ples results in significant sample com-
to show that the microstructure of wet pression (Johnston, 1990).
green compacts of monodisperse silica • Mercury porosimetry is a destructive
spheres was the same for coagulated and testing technique and has limitations
stable sols under the circumstances they concerning the sample sizes which can
studied, but more open for coagulated be determined.
than for dense aqueous alumina suspen- • Porosimetry cannot give any informa-
sions. This led them to the conclusion that tion on the connectivity of the pore
agglomerate strength is an important fac- structure. Actually only an interpreta-
tor determining the green compact proper- tion of the numerical data will rely on a
ties. specific assumption that the pore con-
nectivity does not influence the mercury
intrusion or the extrusion behavior.
 10.4 Characterization Methods 381

• The contact angle in a specific material is considered to be more realistic than the
may vary from place to place, within the parallel bundle of pore and represents an
interval given. extension of the more familiar "ink bottle"
• Limited accessibility of the interior of a pore concept suggested by Drake (1945). It
sample material or even the portion of a is seen that any pore constriction located
pore which is located underneath the between a mercury interface advancing to
surface of the material. fill still open pores and a wider pore seg-
ment would temporarily hinder the filling
A study by Lee (1990) involved mercury of this latter pore and therefore induces a
porosimetry of special sample material large amount of hysteresis. Similar obser-
made to find the limits of applicability of vations can be made for the retraction of
the method. They prepared samples with mercury as the external pressure is re-
interior voids which were larger than the duced. In this case mercury is pushed
exterior pores by slipcasting and showed preferably out of the finest pore segments
that the pore size distribution resembled adjacent to the surface. Wide segments de-
that of the material located on the outside lay the outgoing flow and contribute to
of the sample. Also when the internal voids hysteresis as well. If retraction rates are
were, to a high degree, interconnected, no different in one and the same pore, mer-
significant changes in the measuring data cury separation from the bulk may occur
were found. As soon as interconnected and further complicates a proper analysis
voids in contact with the outer surface of the experimental data.
were present in the material, significant in- It is found in this study that the greater
trusion and extrusion at much lower pres- the entrapment and hysteresis, the greater
sures occurred. There remains, however, a the deviation of the intrinsic pore size dis-
high degree of hysteresis, which indicates tribution from the relevant mercury pene-
that the extrusion spectra shows a consid- tration data. Particularly the pore length is
erable shift to larger pores. It should there- of great importance for the extent of the
fore be concluded that extrusion data can- hysteresis effect. This conclusion is also
not reveal the actual spectra of the pores. derived from the study of Lee (1990) and
Since it was clearly established in this Larson and Morrow (1981).
study that mercury intrusion into the inter- Also as in the former case a mercury
nal voids is regulated by the size of the porosimetry experiment gives the best in-
particle packing channels, it is concluded formation on the pore distribution in the
that mercury porosimetry is useful in char- direct vicinity of the outer surface of the
acterizing the particle packing in the direct material.
vicinity of the outer surface of the sample A bimodal pore spectrum could in such
material. a case be beneficial if most interest is to be
Another study done by Tsetsekou (1991) given to the pores with the smallest sizes.
leads to a model to analyze the data based The large pores are in such a case to give
upon the assumption of a random corru- access to the mercury to the inner structure
gated pore structure. Here the pore model of the material. The intrusion data will
is assumed to comprise non-intersecting however not readily reveal information on
pores wherein individual pore segments the bimodality so that other characteriza-
are made up of cylindrical segments of dis- tion methods will have to be applied beside
tributed sizes. This corrugated pore model the mercury porosimetry to give full ac-
382 10 Green Microstructures and Their Characterization

count of the pore structure of a ceramic for this so-called Guinier region approxi-
material. mation). For highly sintered almost dense
compacts these requirements are fulfilled
and rest porosity can be characterized in
10.4.7 NMR and Small Angle Scattering
this way (see Page, 1988). Green bodies
Techniques
have a much higher "pore concentration"
Pore structure analysis of green bodies and analysis of scattering data is much
with nuclear magnetic resonance or small more difficult because now multiple scat-
angle scattering is not complicated by per- tering can play a dominant role.
colation effects occurring, as we saw was In a comparison of nitrogen adsorption,
the case with capillary methods. This SANS and TEM techniques for the char-
means that pore volume-pore size infor- acterization of mesoporous oxides, Stacey
mation concerning both the voids and (1988) analyzed SANS data obtained from
necks can be obtained more directly. Other mesoporous alumina and zirconia fibres in
advantages of these techniques are that the low Q area for the mean pore gyration
they are nondestructive and that in situ radius and pore distribution using the
measurements of the transition wet-dry Guinier approximation and an assumption
compact are possible. for the pore geometry (spheres or cylin-
Small angle X-ray scattering (SAXS) ders). It appeared that reasonable pore dis-
and small angle neutron scattering (SANS) tributions can be obtained by fitting the
in principle provide similar information SANS data despite the fact that the pore
about a green compact. Structures from system is actually concentrated. The SANS
1-100 nm can be probed by SAXS and a pore distribution appears to be in between
larger range by SANS 1-1000 nm. A dis- the distribution obtained from nitrogen
advantage of SANS is its low availability. adsorption and nitrogen desorption data
A major advantage of neutron scattering as expected.
over X-ray scattering is the possibility of Better methods for analyzing SANS
contrast variation, which for example can data of highly porous solids for pore size
be used for studying the drying (i.e., water distribution are developed by Hardman-
sorption) process in mesoporous green gels Rhyne and co-workers (Frase and Hard-
and green compacts (Ramsay and Wing, man-Rhyne 1988; Hardman-Rhyne et al.,
1991). An other advantage is the possibili- 1986; Hardman-Rhyne, 1987; Page, 1988)
ty of measuring the energy exchange oc- and more recently by Long et al. (1990).
curring in the scattering process. Spectros- Low-field NMR can be used as a pore
copic detection of this exchange processes structure tool with the advantage that a
allows for the study of dynamic processes pore shape assumption is only needed for
with time scales ranging from about pores smaller than a few nm. This method
1 0 " 1 4 s t o 10" 6 s. is still, just as SANS pore characterization,
When a compact can be considered as a in a stage of development but the promise
two-phase system with the pores as the di- is there that it can possibly determine
luted dispersed phase and the solid as the pores with dimensions smaller than the
continuous phase the pore size distribution usual porosimetry. The method relies on
can be obtained from scattering intensity the fact that the spin-lattice relaxation de-
data in the limit of low wave vector Q (see cay time of a volume surrounded by a solid
e.g., Ramsay, 1988; Dore and North, 1991 surface is smaller than that of a bulk fluid
 10.5 Appendix 383

(Smith et al., 1987). Therefore by relating In ceramics, rheology is mainly used to

the measured decay time with the decay characterize properties of preforming ma-
time of the bulk, information on the terials as concentrated sols and suspen-
amount of surface area surrounding the sions (Goodwin, 1990 b; Sacks, 1986). In
liquid, usually water, is obtained. So the such studies researchers try to connect rhe-
volume/surface area diameter can be ob- ological properties of dispersions to the
tained directly. forming performance and green material
The pore size distribution can be ob- properties (Bellosi et al., 1990). In situ
tained by deconvolution of the measured measurement of the rheology, i.e., vis-
magnetization relaxation data (see e.g., coelastic properties, of wet and dry green
Smith and Davis, 1991). The pore size dis- materials could give valuable additional
tribution obtained by NMR is mainly a information. To end this section we men-
void size distribution as can be clearly seen tion that Sacks and Sheu (1987) discuss the
by comparing the distribution obtained rheology of silica sol-gel materials and that
with those from gas adsorption/desorption a good introduction to the rheology of sus-
measurements. pensions and the relation with particle in-
An advantage of the NMR method is teractions and particle packing can be
further that also pores in thin (green) lay- found in Barnes et al. (1989).
ers (membranes) on a porous support can
be measured, as was shown by Glaves et al.
(1989), by measuring at relative vapor
pressures where only the mesoporous coat-
ing is filled with liquid.
10.5 Appendix
10.4.8 Rheological Measurements
Rheological properties of green bodies 10.5.1 Consolidation Effect During
are connected to the interaction forces be- Wet Processing
tween the particles and to the micro struc- We may formally describe the behavior
ture of the material. This makes the inter- of the consolidation surface by the conti-
pretation of rheological quantities in terms nuity equation as follows:
of microstructure more difficult than for
measurement techniques where the mea- 8<p 8
(10-38)
sured properties depend only on the geom- dt dx
etry of the structure. On the other hand the where (p is a particle density to be specified
additional information about interaction for either the suspension or the cake and v
forces that can be obtained when structur- is the settling velocity of the individual
al information is available by other means particles. We will assume that the particle
may help in the explanation of microstruc- density depends on the location in the sus-
ture development during processing. Un- pension and the time as independent vari-
fortunately detailed quantitative interpre- ables: q> (x, t). Under this assumption it is
tation in terms of microstructure is only necessary to indicate exactly which vari-
possible for model systems of mono- able is kept constant:
disperse particles where the interaction
energies are known precisely (Goodwin,
1990a).
384 10 Green Microstructures and Their Characterization

so that we obtain directly We may now apply this analysis to a

simple case of sedimentation where a hy-
dcp\ f^x\ ft drodynamic drag is considered.
If particles are flowing through the sus-
pension, a liquid counter flow reduces its
and after substitution of the continuity
speed by a factor 1 — a cp, where a is a pro-
equation
portionality constant. If £0 is the liquid
dx\ f^x\ {d<Pv\ fdcpv\ density and Q the solid density, the buoy-
dt/cp \dcpjt \ dx )t \ dcp )t s ancy force determined by the overall
(10-39) suspension specific weight is given by
In this result vs is the speed at which the Therefore the force balance is as follows:
location characterized by a constant densi-
ty is moving. If we therefore can relate a
6nrjR
particular particle density to a consolida- 1 — acp
tion front, we may identify vs with the 4
growth rate of the cake. For the present = -TZR3[Q-(1 -cc(p)Q0-Q(pa]
treatment, where green structures are of
interest, we could identify the consolida- so that
tion with the density where particles are v = vo(l -acp)2 (10-41)
completely immobilized by their surround-
which shows directly that there is a gradual
ing, although, as it will be shown, this will
dependence of the particle flow rate on the
not be necessarily an obvious choice.
density. A consolidation front will occur
Before we go into more detail concern-
when
ing the result of the analysis, it is illustra-
tive to define a consolidation boundary at dcpv
x0 such that the particle density and the <0
dcp
particle flow exhibit a restricted disconti-
nuity, for example, or alternatively

dcpv 3acp > 1

= A (cp v) 5 (x — x0) with
~8x~
A {(p v) = js - j c = vscps- vc cpc 10.5.2 Brownian and Colloidal Effect
During Sedimentation
dcp
^-=A(cp)d(x-x0) with The expression for the diffusivity, taking
into account the osmotic effect and the in-
terparticle interaction, can be shown to be
The indices s and c refer to the suspension equal to
(s) and the cake (c).
With the above equation we find 1
D=
6nRr]\_ dcp dcp J
(10-40) where U(d) is the interaction potential be-
A(cp)
tween two colloidal particles at a separa-
This result is not afflicted with the fact that tion d. This separation in turn is a function
a discontinuity occurs at the interface. of the particle concentration. Knowing
 10.6 References 385

this diffusivity we can again set up the Barker, J. A (1958), Discussion Colston Papers X.
London: Butterworth, pp. 125-128.
transport equation, which is similar to the Barnes, H. A., Hutton, F. K, Walters, K. (1989), An
continuity equation but with the diffusivi- Introduction to Rheology, Rheology Series, Vol. 3,
ty term added: Amsterdam: Elsevier.
Bear, J., Bachmat, Y. (1991), Theory and Applications
dcp 9 9 do of Transport in Porous Media, Vol. 4: Bear, J. (Ed.).
(10-43) Dordrecht: Kluwer Acad. Publ.
a7 dx <pv =dx dx*
— D- Belik, V. D. (1989), Poroshkovaya Metallurgiya
(Ukrainian Acad. Set). (8), 320, 18-20. Aug.
A complication is the fact that the diffu- Bellosi, A., Gallassi, C , Guicciardi, S. (1990), /. Ma-
sion term is a function of the particle den- ter. Sci. 25, 4331-4340.
sity. Further we should, to describe also Berryman, J. G. (1985), /. Appl. Phys. 57, 2374-2384.
Betzig, E., Trautman, J. K. (1992), Science 257, 189-
structure development during slip casting, 195.
give an extended expression for the migra- Bhave, R. R. (1991), Inorganic Membranes, Synthesis,
tion velocity of the particles through the Characteristics, and Applications. New York: van
Nostrand Reinhold.
liquid taking into account the forced flow Bierwagen, G. P. (1992), J. Coatings Technol. 64, 71-75.
of the liquid through the cake and there- Binnig, G., Gerber, Ch., Stoll, E., Albrecht, T. R.,
with increasing the rate of consolidation: Quate, C. F. (1987), Europhys. Lett. 3, 1281-1286.
Botet, R., Jullien, R. (1988), Ann. Phys. Fr. 13, 153-
v = vm + (2/9) • {R2ln) • (Ag g) where vm is the 221.
settling speed generated by the mold and Bremer, L. G. B. (1991), "Fractal Aggregation in Re-
AQ the density difference between liquid lation to Formation and Properties of Particle Gels"
PhD thesis. Agricultural University Wageningen,
and solid. Without knowing the exact solu- The Netherlands.
tion, it is seen directly that a cake is being Bremer, L. G. B., van Vliet, T., Walstra, P. (1989),
formed when at least somewhere the parti- /. Chem. Soc. Faraday Trans. 1, 85, 3359.
Bridger, K., Massuda, M. (1990), Ceramic Powder
cle concentration is increasing. This means Sci. Ill Ceramic Transactions 12, 507-519.
that Briggs, G. A. D., Somekh, M. G., Ilett, C. (1982),
Acoustic Microscopy in Materials Science, SPIE
dcp 368, Microscopy, Techniques and Capabilities.
>0 Bellingham, WA: SPIE, pp. 74-79.
97 Brinker, C. X, Scherer, G. W. (1985), / Non-Cryst.
Since the coordinate x is taken positive Solids 70, 301-322.
Brinker, C. X, Scherer, G. W. (1990), Sol-Gel Science,
everywhere in the system, we must have The Physics and Chemistry of Sol-Gel Processing.
after integration of the equation Boston: Academic Press.
Brun, M., Lallemand, A., Quinson, J.-F., Eyraud, C.
Ddo (1976), Thermochimica Acta 21, 59-88.
o = = const > 0 Cao, W, Gerhardt, R., Wachtman, X B. Jr. (1988), J.
v ox Am. Ceram. Soc. 71, 1108-1113.
Correia, L. A., Pex, P. P. A. C , Veringa, H. X (1989),
So that after some rearranging we get Euro-Ceramics 3, Proc. 1st ECerS Conf.
Maastricht, 1989: de With, G., Terpstra, R.A.,
6ntiRvo > kT— (10-44) Metselaar, R. (Eds.). New York: Elsevier.
9x Cuperus, F. P. (1990), Characterisation of Ultrafiltra-
tion Membranes, Pore Structure and Top-Layer
Thickness, Thesis, University of Twente, Twente,
The Netherlands,
de Boer, X H. (1958), The Structure and Properties of
10.6 References Porous Materials, Colston Papers X. London: But-
terworth, pp. 68-94.
Adamson, A. W. (1990), Physical Chemistry of Sur- Debye, P., Anderson, H.R., Brumberger, H. (1957), /.
faces, 5th ed. New York: Wiley. Applied Phys. 28, 679-683.
Aksay, A. I., Kikuchi, R. (1986), Science of Ceramic Dietz, P., Hansma, P. K., Herrmann, K.-H., Inacker,
Chemical Processing. New York: Wiley, pp. 513— O., Lehmann, H.-D. (1992), Ultramicroscopy 35,
521. 155-159.
386 10 Green Microstructures and Their Characterization

Dore, J. C , North. A. N. (1991), in: Characterisation NATO ASISeries C 324: Bloor, D.M., Wyn-Jones,
of Porous Solids II: Rodriguez-Reinoso, R, Rou- E. (Eds.). Dordrecht, The Netherlands: Kluwer,
querol, J., Sing, K.S.W., Unger, K.K. (Eds.). Am- pp. 659-679.
sterdam: Elsevier, pp. 245-255. Gregg, S. J. (1986), Colloids and Surfaces 21, 109-
Drake, L. C. (1945). Ind. Eng. Chem. Anal Ed. 17, 124.
787. Hardman-Rhyne, K. A., Frase, K. G., Berk, N. F.
Ducker, W. A., Senden, T. I , Pashly, R. M. (1991), (1986), Physica 136B, 223-225.
Nature 353, 239-241. Hardman-Rhyne, K. A. (1987), Adv. Ceramics 21,
Dullien, F. A. L. (1979), Porous Media, Fluid Trans- 767-778.
port and Pore Structure. New York: Academic Hornbogen, E. (1986), /. Mater. Sci. 21, 3737-3747.
Press. Hudson, D. R. (1947a), Machinery, 70, 681-683.
Ellington, W. A., Ackerman, J. L., Garrido, L., Hudson, D. R. (1947b), Machinery, 70, 617-622.
Weyand, X D., Dimilia, R. A. (1987), /. Ceram. IUPAC (1972), Manual of Symbols and Terminology
Eng. Sci. Proc. 8, 503-512. for Physico-Chemical Quantities and Units, Part I.
Enustun, B. V., Gunnink, B. W, Demirel, T. (1990), /. London: Butterworth.
Colloid Interface Sci. 134, 264-275. Johnston, G. P. (1990), Powder Technol. 61, 289-294.
Enustun, B. V., Eckrich, X, Demirel, T. (1985), Phase Jullien, R. (1988), Contemp. Phys. 28, 477-493.
Transition Porosimetry, Proc. Int. Symp. Panicu- Katsuki, H., Kawahara, A., Ichinose, H. (1992), /.
late and Multi-Phase Processes, 16th Annual Meet- Mater. Sci. 27, 6067-6070.
ing Fine Particle Society, April 1985, Florida 68-1- Kerch, H. M., Cosandrey, F, Gerhardt, R. A. (1993),
68-6. /. Non-Cryst. Solids 152, 18-31.
Everett, D. H. (1958), Some Problems in the Investiga- Kerch, H. M., Gerhardt, R. A. (1989), Ceram. Trans.
tion of Porosity by Adsorption Methods, Colston 5, 360-367.
Papers X. London: Butterworth, pp. 95-120. Kumar, K.-N., Keizer, P., Burggraaf, A. X, Okubo,
Everett, D. H. (1988), in: Studies in Surface Science X, Nagamoto, H., Morooka, S. (1992), Nature 58,
and Catalysis, Vol. 39: Characterisation of Porous 48-50.
Solids I: Unger, K. K., Rouquerol, J., Sing, K. S. Kunze, K., Segal, D. (1991), Colloids Surf 58, 327-
W, Krai, H. (Eds.). Amsterdam: Elsevier, pp. 1-21. 337.
Eyraud, C , Quinson, J. K, Brun, M. (1988), in: Stud- Lange, F. F. (1984), /. Am. Ceram. Soc. 2, 83-89.
ies in Surface Science and Catalysis Vol. 39: Char- Larson, R. G., Morrow, N. R. (1981), Powder Tech-
acterisation of Porous Solids I: Unger, K.K., Rou- nology, 30, 123-138.
querol, I, Sing, K. S. W, Krai, H. (Eds.). Amster- Lee, H. H. D. (1990), J. Am. Ceram. Soc. 73, 2309-
dam: Elsevier, pp. 295-305. 2315.
Fara, H. D., Scheidegger, A. E. (1961), /. Geophys. Long, G. G., Krueger, S., Jemian, P. R., Black, D. R.,
Res. 66, 3279-3284. Burdette, H. E., Cline, X P., Gerhardt, R. A.
Feda, J. (1982), Dev. Geotech. Engin. Vol. 30: Me- (1990), J. Appl. Cryst. 23, 535-544.
chanics of Paniculate Materials, the Principles. Lowell, C. (1979), Introduction to Powder Surface
Amsterdam: Elsevier. Area. New York: Wiley.
Fishmeister, H. F. (1974), in: Pore Structure and Lowell, S., Shields, J.E. (1984), Powder Technology
Properties of Materials Proc. Int. Symp. RILEM/ Series: Powder Surface Area and Porosity, 2nd ed.:
IUPAC Prague 1973. Scarlett, B. (Ed.). London: Chapman and Hall.
Frase, K. G., Hardman-Rhyne, K. (1988), /. Am. Luckham, P. F , Vincent, B., McMahon, X, Tadros,
Ceram. Soc. 71, 1-6. Th. F. (1983), Colloids and Surfaces 6, 83-95.
Friedman, W. D. (1987), Characterisation of Green Mason, G. (1988 b), Proc. R. Soc. Lond. A 415, 453-
Ceramics with X-ray Tomography and Ultrasonics, 486.
pp. 28-30. Mason, G. (1988 a), Catalysis 39, 323-332.
Fritzsche, A. K., Arevalo, A. R., Moore, M. D., We- McGeary, R. K. (1961), J. Am. Ceram. Soc. 44, 513-
ber, C. J., Elings, V. B., Kjoller, K., Wu, C. M. 522.
(1992), /. Appl. Pol. Sci., 46, 167-178. McClements, D. X (1991), Adv. Colloid Interface Sci.
Furnas, C. C. (1928), U.S. Bur. Mines Rep. Invest., 37, 33-72.
2894. Meagher, L. X (1992), Colloid Interface Sci. 152, 293-
Furnas, C. C. (1929), U.S. Bur. Mines Bull., 307. 295.
Gauthier, F.G.R., Danforth, S.C. (1988), Ceramic Meakin, P. (1988), Adv. Colloid Interface Sci. 28,
Trans. 1, 709-715. 249-331.
Glaves, C. L., Davis, P. J., Moore, K. A., Smith, D. Meakin, P. (1988), Ann. Rev. Phys. Chem. 39, 237-
M., Hsieh, P. (1989), J. Colloid Interface Sci., 133, 267.
377-389. Menold, R., Luutge, B., Kaiser, W (1976) Adv. Col-
Goodwin, J. W. (1990a), Ceram. Bull. 69, 1694-1698. loid Interface Sci. 5, 281-335.
Goodwin, J. W. (1990 b), in: The Structure, Dynamics Messing, G. L., Onoda, Jr., G. Y (1978), /. Am. Cer-
and Equilibrium Properties of Colloidal Systems. am. Soc. 61, 1-5
 10.6 References 387

Meyer, G., Amer, N. (1990), J. Appl. Phys. Lett. 57, Schaefer, D. W (1989), Science 243, 1023-1027.
2089-2091. Schaefer, D. W, Keefer, K. D. (1986), Mat. Res. Soc.
Mishra, B. K. (1988), AIChE J. 34, 684. Symp. Proc. 73, 277-288.
Nahass, P., Pober, R. L., Rhine, W. E., Robbins, L., Schilling, C. H., Aksay, I. A. (1988), Ceramic Trans.
Bowen, H. K. (1992), J. Am. Ceram. Soc. 75, 2373- 1, 800-808.
2379. Serra, J. (1982), Image Analysis and Mathematical
Onoda, G. Y. (1977), in: Ceramic Microstructures 76: Morphology. New York: Academic Press.
Fulrath, R. A., Pask, J. A. (Eds.). Boulder, CO: Sing, K. S. W. (1989), Colloids and Surfaces 38, 113-
Westview Press, 163-181. 124.
Ouchiyama, N. S., Tanaka, T. (1980), Ind. Eng. Chem. Smith, D. M., Davis, P. J. (1991), in: Characterisation
Fundam. 19, 338-340. of Porous Solids II: Rodriguez-Reinoso F , Rou-
Overney, R. M., Meyer, E., Frommer, X, Brodbeck, querol, X, Sing, K. S. W, Unger, K. K. (Eds.).
D., Luthi, R., Howald, L., Guntherodt, H.-J., Fuji- Amsterdam: Elsevier, pp. 301-310.
hara, M., Takano, H., Gotoh, Y. (1992), Nature Smith, D. M. (1987), Adv. Ceram. 21, 779.
359, 133-134. Spurr, A. R. (1969), /. Ultrastructural Res. 26, 31-43.
Page, R. A. (1988), /. Appl. Cryst. 21, 795-804. Stacey, M. H. (1988), in: Characterisation of Porous
Park, C.-Y, Ihm, S.-K. (1990), AIChE J. 36, 1641. Solids: Unger, F , Rouquerol, X, Sing, K.S.W.,
Parlar, M., Yortsos, Y C. (1989), J. Colloid Interface Unger, K. K. (Eds). Amsterdam: Elsevier.
Sci. 132, 425-443. Stewart, R. F , Sutton, D. (1984), in: Solid-Liquid
Parlar, M., Yortsos, Y C. (1988), J. Colloid Interface Separation. Gregory, X (Ed.). Edis Horwood:
Sci. 124, 162-176. Chichester.
Philipse, A. P., Bonekamp, B. C , Veringa, H. J. Takasu, Y, Suzawa, K., Ueno, M., Yahikozawa, K.
(1990), J. Am. Ceram. Soc. 73, 2720-2726. (1990), J. Catalysis 123, 279-281.
Pickles, D. G., Lilley, E. (1985), /. Am. Ceram. Soc. Taruta, S., Okada, K., Otsuka, N. (1990), J. Ceram.
68, C222-C223. Soc. Jpn., Int. Ed. 98, 30-36.
Pieranski, P. (1983), Contemp. Phys. 24, 25. Terpstra, R.A., Bonekamp, B. C , Veringa, H. X
Pujari, V. K. (1988), Ceram. Trans. 1, 635-644. (1988), Desalination 70, 395-404.
Quinson, J. F, Brun, M. (1988), in: Studies in Surface Tsakiroglou, C. D., Payatakas, A. C. (1991), J. Col-
Science and Catalysis Vol. 39: Characterisation of loid Interface Sci. 146, 479.
Porous Solids I: Unger, K. K., Rouquerol, X, Sing, Tsakiroglou, C. D., Payatakas, A. C. (1990), J. Col-
K. S. W., Krai, H. (Eds.). Amsterdam: Elsevier, loid Interface Sci. 137, 315.
pp. 307-315. Tsetsekou, A. (1991), Chem. Eng. Comm. 110, 1-29.
Radmacher, M., Tillmann, R. W, Fritz, M., Gaub, Uematsu, K., Miyashita, M., Kim, X X, Kato, Z.,
H. E. (1992), Science 257, 1900-1905. Uchida, H. (1991), /. Am. Ceram. Soc. 74, 2170-
Ramsay, J. D., Wing, G. (1991), J. Colloid and Inter- 2174.
face Sci. 141, 475-485. Unger, K. K., Rouquerol, X, Sing, K. S. W, Krai, H.
Ramsay, J. D. F. (1988), in: Studies in Surface Science (Eds.) (1988), Characterisation of Porous Solids I,
and Catalysis Vol. 39: Characterisation of Porous Proc. IUPAC Symp. (COPS I), Bad Soden, 1987.
Solids I: Unger, K. K., Rouquerol, I, Sing, K. S. Amsterdam: Elsevier.
W., Keral, H. (Eds.). Amsterdam: Elsevier. Veringa, H. X (1993), J. Mater. Sci. 28, 2757-2762.
Reed, J. S. (1988 a), Introduction to the Principles of Vivier, H., Noelle Pons, M., Portala, X-F (1989), /.
Ceramic Processing, New York: Wiley. Membr. Sci., 46, 81-91.
Reed, J. S. (1988b), Ceramic Trans. 1, 601-611. Wadsworth, X (1960), Nat. Res. Counc. Canada,
Rhodes, W. H. (1981), J. Am. Ceram. Soc. 64, 19-22. Mech. Eng. Rep. MT-41, Feb. i960, NRC 5895.
Rodriguez-Reinoso, F , Rouquerol, J., Sing, K. S. W, Weeks, M.D., Laughner, X W (1987), Adv. Ceram. 21,
Unger, K. K. (Eds.) (1991), Characterisation of Po- 793-800.
rous Solids II, Proc. IUPAC Symp. (COPSII) Ali- Wertman, A. E. R., Hugill, H. R. (1930), J. Am.
cante, 1990, Amsterdam: Elsevier. Ceram. Soc. 13, 767-779.
Roosen, A., Bowen, H. K. (1988), J. Am. Ceram. Soc. Yeh, T.-S., Sacks, M. D. (1988 a), X Am. Ceram. Soc.
71, 9970-9977. 71, 841-844.
Russ, J. C. (1990), Computer-Assisted Microscopy. Yeh, T.-S., Sacks, M. D. (1988 b), J. Am. Ceram. Soc.
New York: Plenum Press. 71, C-484-C-487.
Sacks, M. D. (1986), in: Science of Ceramic Process- Young, A. C , Omatete, O. O., Janney, M. A.,
ing: Hench, L. L., Ulrich, D. R. (Eds.). New York: Menchhofer, P. A. (1991), J. Am. Ceram. Soc. 74,
Wiley, pp. 522-538. 612-618.
Sacks, M. D., Tseng, T.-Y (1984), /. Am. Ceram. Soc. Zarzycki, X, Prassas, M., Phalippou, X (1982), /. Ma-
67, 526-532. ter. Sci. 17, 3371-3379.
Sacks, M. D., Sheu, R. S. (1987), J. Non-Cryst. Solids Zhadanov, V. P., Fenelonov, V. B., Efremov, D. K.
92, 383-396. (1997), J. Colloid Interface Sci. 120, 218-223.
11 Advanced Ceramics from Inorganic Polymers
Ralf Riedel

Fachbereich Materialwissenschaft, Technische Hochschule Darmstadt, Darmstadt,

Federal Republic of Germany

List of Symbols and Abbreviations 2

11.1 Introduction 4
11.2 Polymer Pyrolysis: The Process 5
11.2.1 A Process Description 5
11.2.2 Properties of Organoelement Polymers 9
11.3 SiC Ceramics from Poly(organo)silanes 10
11.3.1 Poly(diorgano)silanes - Properties and Synthesis 11
11.3.2 Synthesis of Polysilanes and Polycarbosilanes from Dichlorodimethyl-
and Dichloromethylphenylsilane 12
11.3.2.1 Thermogravimetric Analysis 14
11.3.2.2 Infrared and NMR Spectroscopic Studies of Poly(methylphenyl)silanes
and Carbosilanes 17
11.3.2.3 Molecular Weight Distribution Analysis of the Poly(methylphenyl)silanes 19
11.3.3 Synthesis of Boron-Containing Poly(organo)silanes 20
11.3.3.1 Hydroboration of Poly(methylvinyl)silane 20
11.3.3.2 Polycondensation of Tris[(dichloromethylsilyl)ethyl]borane to Form
Boron-Containing Si Polymers 21
11.3.4 X-Ray Studies of Ceramic Materials Produced from
Poly(methylphenyl)silanes 22
11.3.5 In Situ Generation of SiC Dispersions in Si 3 N 4 and B4C Composites . . . . 25
11.4 Si-N Ceramics Based on Poly(organo)silazanes 25
11.4.1 Poly(organo)silazanes - Synthesis and Properties 26
11.4.2 Synthesis and Characterization of S i - C - N Ceramics
from Polysilazanes 28
11.4.2.1 The Poly(organo)silazanes Used 28
11.4.2.2 Characterization and Pyrolysis of Poly(hydridomethyl)silazane 28
11.4.2.3 Characterization of Silicon Carbonitride 30
11.5 Production of Non-Oxide Si-Based Ceramic Parts 34
11.5.1 Conventional Production of SiC and Si 3 N 4 Parts 34
11.5.2 Production of Si-Based Ceramic Parts from Polymeric Compacts 34
11.5.2.1 Crack-Free, Dense Ceramic Materials from Organoelement Polymers . . . . 35
11.5.2.2 Si-Based Ceramic Parts from Poly(organo)silazanes 36
11.5.2.3 Si-Based Ceramic Parts from Polycarbosilanes 43
11.5.2.4 Densification 44
11.6 Summary and Outlook 46
11.7 Acknowledgements 47
11.8 References 47

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
2 11 Advanced Ceramics from Inorganic Polymers

List of Symbols and Abbreviations

c lattice constant
d particle diameter
Dc density of ceramic
Dp density of polymer
Dx_y bond energy between x and y
e charge of an ion
h crystallite size
m mass of an ion
Mc mass of the obtained ceramic pyrolysis product
Mp mass of the starting polymer
P sintering pressure; porosity
r particle radius
T temperature
Tg glass transition temperature
V volume

a ceramic yield
P integral width of an X-ray reflection
y interface energy
(5 chemical shift
r\ viscosity
9 diffraction angle
9 relative density
X wavelength
Q density
ah hydrostatic stress
crr radial stress
crt hoop stress

AFCOP active filler controlled pyrolysis

CVD chemical vapor deposition
DEPT distortionless enhancement by polarization transfer
DTA differential thermal analysis
EI-MS electron ionization mass spectrometry
ESCA electron spectroscopy for chemical analysis
ESR electron spin resonance
FT-IR Fourier transform infrared
GPC gel permeation chromatography
HPZ hydridopolysilazane
HV Vickers hardness
IR infrared
MOCVD metal-organic chemical vapor deposition
MS mass spectrometry
 List of Symbols and Abbreviations

NMR nuclear magnetic resonance

PAN polyacrylonitrile
PCS polycarbosilane
PDSC polymer derived silicon carbide
PDSCN polymer derived silicon carbonitride
PMPS poly(methylphenyl)silane
PVS polyvinylsilane
RBSN reaction bonded silicon nitride
SEM scanning electron microscope
SiAlONs S i - A l - O - N based ceramics
TEM transmission electron microscope
TGA thermogravimetric analysis
TH F tetrahy drofuran
TM S te t ramethylsilane
WDX wavelength-dispersive X-ray
XPS X-ray photoelectron spectroscopy
 11 Advanced Ceramics from Inorganic Polymers

11.1 Introduction large technical scale (700000 tons annual-

ly) using the Rochow synthesis (Rochow,
In recent times, methods such as chemi- 1952) and are used as starting materials for
cal vapor deposition (CVD), sol-gel tech- silicones. The thermal decomposition of
nology and polymer pyrolysis have been the polymers, for example polysiloxanes,
used for the production of high-purity ce- yields ceramic materials in the Si/C/O sys-
ramic starting materials alongside the clas- tem, so-called silicon oxycarbides. Polysi-
sical powder metallurgical processes. Due lanes and polycarbosilanes yield silicon-
to the use of these methods, new applica- carbide-containing materials SiC 1 + x with
tions of these materials have been made a carbon content usually higher than in
possible. What the three methods have in pure SiC, and polysilazanes result in Si/C/
common is that the ceramic is produced N-based materials which can be described
from low molecular weight, inorganic or as silicon carbonitrides.
organoelement precursors. The latter are Ceramic materials resulting from the py-
molecular species containing direct, polar rolysis of high-molecular-weight materials
E<5 + _c<5- b o n c j s between the element E at around 1000 °C exhibit greatly different
and carbon. E can be a metal or a non- composition, purity, and crystallinity to
metal, giving these methods a great poten- conventionally produced materials. The
tial in ceramic production and processing. variation of the polymer structure in prin-
In this chapter, polymer pyrolysis, that is ciple allows the production of new materi-
the production of ceramic materials from als (e.g. metastable phases in the B-C-N or
inorganic polymers, will be described. Si-C-N systems) with the stoichiometry
In contrast to the sol-gel process, which controlled at the molecular level. The
is ideally suited to the production of oxide properties of ceramic materials produced
glasses and ceramics, the solid-state pyrol- by polymer pyrolysis depend not only on
ysis of inorganic polymers is best suited to the chemical content and bonding in the
the synthesis of non-oxide, nonmetallic solid ceramic but also on the structure and
materials. In the lost 25 years a number of bonding in the precursor polymer.
inorganic polymers have been developed. This chapter will concentrate on the syn-
The following types of silicon-based poly- thesis of inorganic polymers and the pro-
mers are known: cessing of these species to form nonmetal-
(a) polysiloxanes: [-R 2 Si-O-] n lic, inorganic, silicon-carbide- and silicon-
(b) polysilanes: [-R2Si-]n nitride-based materials. The chemical and
(c) polycarbosilanes: [-R 2 Si-CH 2 -] n structural investigation of the polymers as
(d) polysilazane: [-R 2 Si-NH-] w well as their pyrolysis behavior and the
A number of copolymers with mixed characterization of the ceramic products
polymer units are also known. The chemi- will be described, as will the microstruc-
cal behavior of the polymer depends both tural analysis (spectroscopic and electron
on the type of polymer and on the nature microscopic) of these metastable materi-
of the R groups on the silicon atoms, als. The crystallization and sintering be-
which can be hydrogen or alkyl or aryl havior of the ceramics and in particular the
groups. The production of polymer types phase transformations which occur both
(a)-(d) starts with abundantly available during the transformation of the polymers
chlorosilanes. Chlorosilanes containing to the inorganic solids and at high temper-
organic substituents are produced on a ature (over 1200°C) will also be discussed.
 11.2 Polymer Pyrolysis: The Process

The production of thin films and fibers (2) Thermal decomposition (pyrolysis) of
that are very small at least in one dimen- the high-molecular-weight compounds to
sion has been demonstrated (Schulenburg, form nonmetallic, inorganic solids.
1991). It is also possible now to produce This process is therefore analogous to the
monolithic materials (as opposed to pow- production of pyrolytic carbon where, e.g.,
ders) using the pyrolysis method. This polyacrylonitrile (PAN) filaments are py-
method also has the large advantage that rolyzed in several steps to form carbon
the reaction temperature is low at 800- fibers (Fitzer, 1985; Boder et al., 1980).
1500°C compared with traditional meth- The thermal decomposition of tetra-
ods, resulting in considerable energy sav- methylsilane (TMS) and the characteriza-
ings, and, due to the complex nature and tion of the reaction products has been the
amorphous structure of the polymer- subject of intense investigation since 1937
derived materials, totally new properties (Helm and Mark, 1937; Fritz, 1987). How-
can be expected. ever, it has been known only since the
middle of the 1970s that ceramic materials
can be produced from organoelement
11.2 Polymer Pyrolysis: compounds (Verbeek, 1973; Verbeek and
The Process Winter, 1974; Yajima etal., 1976), the
initial work concerning the thermolysis
11.2.1 A Process Description of poly(alkyl)- or poly(aryl)carbosilanes
The conventional production of ceramic ([RSiH-CH 2 ]J to form SiC-containing
materials is based on inorganic high-tem- material. Since then much work on the syn-
perature synthesis. Oxide ceramics are ob- thesis of silicon-containing polymers -
tained from minerals, for example A12O3 is mainly poly(organyl)silanes, polycarbosi-
extracted from bauxite using the Bayer lanes and polysilazanes - for the production
process. Non-oxide ceramics are synthe- of SiC and Si 3 N 4 has been published.
sized either through reaction of the constit- The process of converting chlorosilanes
uent elements or through carbothermal re- R 4 _ x SiCl x to SiC and Si 3 N 4 ceramics can
duction of the corresponding oxides. be described as follows: From Fig. 11-1 it
Examples of the latter are the following can be seen that the first step is the conver-
Si 3 N 4 and SiC synthesis: sion of organoelement compounds into
oligomeric and polymeric organoelement
3Si + 2 N 2 - ^ ^ S i 3 N 4 (11-1) intermediates. Where x is two or more, in
SiO2 + 3C 2200°c>SiC + 2CO analogy to the Wurtz-Fittig synthesis, the
(Acheson process) (11-2) treatment with alkali metals results in
CO polysilanes (Burkhard, 1949); with dilithi-
um acetylide, polycarbosilanes (Boury
(11-3) etal., 1990) are formed; with ammonia,
On the other hand, the solid-state pyroly- polysilazanes result (Seyferth and Wise-
sis of inorganic precursors to ceramic mate- man, 1984) and with water, polysiloxanes
rials is a low-temperature process which in (silicones) (Noll, 1968).
principle involves two steps: The second step is the thermal conver-
(1) Synthesis of inorganic oligomers or sion of the polymers to ceramic materials.
polymers from low-molecular-weight com- This process results not in the thermody-
pounds. namically stable oc- or (3-SiC and a- and
 11 Advanced Ceramics from Inorganic Polymers

loocrc |SiC|
[RSiH-CH 2 ] n
Polycarbosilane
nooox
v400°C
Si + CH 2 -CH 2 -Si + + CH-Si + C 2 H 4 -SiH 3
[R,R 2 SiJ n u u | m | \ n
H H
Polysilane - H CH 3 H

R,=R 2 = R 1= H;R 2 =HC=CH 2

1)NH3
2 ) K H
'R 2 -CH 3

[R 1 R 2 SiNH] m [R 2 SiN] n [RiR2SiO]n

Figure 11-1.
Polysilazane
Polysiloxane Organochlorosilanes used
J000°C as starting materials for the
1000°C
production of silicon-con-
NHq
taining polymers and ce-
|SiC/SiO 2 l ramics.

(3-Si3N4 phases but rather metastable, Table 11-1 contains details of analogous
amorphous solids where the kinetics of processes used to produce other ceramic
crystallization depend strongly on the stoi- materials including A1N, BN, and B4C.
chiometry of the amorphous product. Also, Figure 11-2 shows the basic reaction se-
in many cases, the crystallization of such quence with respect to the volume and den-
materials only starts well above 1000 °C. sity changes associated with the structural
conversion of polymers to inorganic solids
at different temperatures. Between 100 and
400 °C the major processes are the evapora-
tion of low-molecular-weight compounds
and condensation, polyaddition, and poly-
merization reactions which result in further
cross-linking and therefore an increase in
the molecular weight. Above 400 °C the
•i Crystallisation
Ceramisation thermal decomposition of the organoele-
Sintering
ment polymers begins, a process which is
500 1000 normally completed between 800 and
Temperature [°C]
1000 °C. At these temperatures hydrocar-
Figure 11-2. Volume and density changes during the bons and hydrogen are released. Chlorine-
thermal decomposition of Si-containing, polymeric
containing precursors in many cases exhibit
starting materials such as polysilanes or polysilazanes
to dense Si based bulk ceramics. Crystallization to the significant weight loss up to 1200 °C due to
a or f$ forms or mixtures thereof follows the ceramiza- the elimination of chlorine or HC1 (Riedel
tion. etal., 1990a; Riedel etal., 1991). Crystal-
 11.2 Polymer Pyrolysis: The Process

Table 11-1. Inorganic ceramic precursors for the production of non-oxide ceramics.

Inorganic precursor Ceramic References

[RAlNH]n A1N Interrante et al. (1986)

[Cl2Al-N(H)Si(CH3)3]2 A1N Riedel et al. (1990 a)
BH3 C 5 H 5 N B,CyNz Riedeletal. (1989 b)
[H 2 NBNC 6 H 5 ] 3 BN Tanigushi et al. (1976)
(CH3)4Si SiC Helm and Mark (1937)
[(CH 3 ) 2 SiO] m [CH 3 SiO 15 ] B Si,CyO2 Remlund et al. (1991 a, b)
[CH 3 SiH-CH 2 ] n SiC/C Yajima et al. (1976)
[((CH3)2Si)JCH3SiC6Hs)J. SiC/C Mazdiyasni et al. (1978)
[CH3SiHNH]ra • [CH3SiN]n Seyferth and Wiseman (1984)
[B 10 H 12 diamine] n B4C/BN Rees and Seyferth (1988)
[(C4H9N)2Ti]n TiN Seyferth and Mignani (1988)
Zr[BHJ4 ZrB2 Rice and Woodin (1988)

lization starts above 1000 °C and is usually - The infiltration of porous substrates, for
accompanied by a further increase in densi- example in the production of fiber-rein-
ty. forced composites
Polymer pyrolysis has the following ad- - The production of inorganic membranes
vantages over conventional methods: with well-defined porosity in the nanome-
• Monomeric and polymeric starting ma- ter range (Burggraaf, 1991)
terials can be synthesized in a very pure The commercial production of SiC
form which results, in turn, in purer ceram- fibers (Nicalon or Tyranno fibers) has been
ics than those produced in the traditional carried out for several years using the py-
manner from natural raw materials rolysis of carbosilane (Yajima et al., 1976):
(Pouskouleli, 1989). Owing to the sensitiv- ra(CH3)2SiCl2 —
ity towards moisture nature of the non-
oxide pyrolysis products, which results + 2mNaCl (11-4)
from the large surface area of the powders [Si(CH3)2]M 450 °C
[CH 3 SiH-CH 2 ] w (11-5)
(up to 150-200 m 2 /g), and the amorphous
state of the materials, and in order to pre- The dechlorination of dichlorodimethyl-
vent contamination with oxygen, the mate- silane with sodium results in poly(dimeth-
rials are handled under inert gas. The sus- yl)silane. Poly(diorganyl)silanes are often
ceptibility to hydrolysis generally decreas- referred to as poly(diorganylsilylenes) or
es with increasing carbon content of the perorganopolysilanes. The polydimethyl-
materials. silane is converted to polycarbosilane
• The pyrolysis process opens up new ap- [CH 3 SiH-CH 2 ] n in an autoclave at 450°C
plications for ceramic materials, and al- under 10 MPa argon.
lows the development of new production Polycarbosilanes are organosilicon poly-
technology such as: mers in which the silicon atoms are bridged
- The fabrication of ceramic fibers (Yaji- by bifunctional organic groups [e.g.
ma et al., 1976; Peuckert et al., 1990) [R 1 R 2 Si-(CR 1 R 2 )] n , where Ri = R 2 = al-
- The coating of substrates with ceramics kyl, phenyl, vinyl, etc. (Seyferth, 1988)].
(surface modification, Peuckert et al., The resulting polycarbosilane can be spun
1990) to form fibers and, in a further step at
8 11 Advanced Ceramics from Inorganic Polymers

200 °C in air, annealed to form cross-linked the powder and reduced particle growth at
Si-O-Si bridges, thus increasing the rigid- high temperatures (Riedel etal., 1989b;
ity of the polymer network. The next stage Passing etal., 1991). The control of the
is pyrolysis to form SiC fibers which, apart temperature of the pyrolysis process al-
from an excess of carbon, contain around lows the production of amorphous or crys-
28 wt.% SiO2 due to heating in air (Pysher talline ceramic powders.
etal., 1988): The crystallinity, the morphology and
[(CH 3 ) 2 Si] 1 [CH 3 SiH-CH 2 ] m the homogeneity of the ceramic starting
powder have great influence on the sinter
[O-(CH 3 ) 2 Si-CH 2 ] w 1200C/Ar>
activity, particle growth, and microstruc-
(11-6) ture of the densified material (Riedel et al.,
In a modified form, this process also 1989 a; Sawhill and Haggerty, 1982). The
makes possible the production of other fi- production and processing temperatures
ber materials based on BN, Si 3 N 4 and used for polymer-derived ceramic powders
SixCyNz (Paine and Narula, 1990; Yokoy- are, in many cases, lower than those used
oma etal., 1991; Peuckert etal., 1990). for conventional raw materials (Barringer
Table 11-2 contains details of several com- and Bowen, 1982; Matijevic, 1989; Pro-
mercially available ceramic fibers based on chazka and Klug, 1983; Riedel etal.,
inorganic polymers. 1989 a; Riedel et al., 1990 b, c). Apart from
Organoelement compounds have also this, Fig. 11-3 shows the narrow particle-
been successfully used on the laboratory size distribution (size in the nanometer
scale for the production of ceramic com- range) of such materials. These properties
posite material (Seyferth and Wiseman, point towards higher strength and reliabil-
1984; Riedel etal., 1989a, Toreki etal., ity for these materials, providing that the
1990), multicomponent glasses (Dislich, other factors affecting strength, e.g., cracks,
1971), or the liquid-phase coating of ce- pores, impurities, agglomerates, or surface
ramic powder surfaces with additives defects are controlled during processing.
(Roosen and Bowen, 1989; Jenett etal., The in situ crystallization of the ceramic
1990; Passing etal., 1991; Riedel etal., phases during the sintering of amorphous,
1988). The latter method results in a better pyrolytic starting powders could be ex-
distribution of the additives which can re- ploited for the production of polycrys-
sult in an increased sintering activity for talline materials with tailor-made grain

Table 11-2. Producers and properties of selected commercially available non-oxide Si-based ceramic fibers
synthesized from inorganic polymers.

Fiber type Producer Density E-modulus

(g/cm3) (GPa) (GPa) (um)

fi-SiC "Nicalon" Nippon Carbon Co., Tokyo, Japan 2.6 2.6 193 100
Si-Ti-C-O "Tyranno" UBE Industries, Tokyo, Japan 2.5 2.8 193 10
SiC Berghof, Tubingen, F.R.G. 3.4 3.5 410 100
Si3N4 "TNSN" Toa Nenryo Kokyo K. K. (Tonen), 2.5 2.5 250 10
Tokyo, Japan
b
Strength; diameter.
 11.2 Polymer Pyrolysis: The Process 9

partial pressure of ammonia the carbon

content of the Si-C-N powder can be re-
duced. This powder is then conventionally
processed. After densification (liquid
phase sintering under hot pressing condi-
tions the material contains both p-Si 3 N 4
and p-SiC particles with sizes in the 200-
500 nm range. The superplasticity of this
covalent composite material is the result of
the extremely small particle size and this
effect was previously only known for
metallic and occasionally for ionic poly-
crystalline (ZrO 2 /Y 2 O 3 ) materials.

11.2.2 Properties of Organoelement

Polymers
The chemical and physical properties of
organoelement polymers, such as solubili-
Figure 11-3. SEM images of (a) £-SiC (Riedel et al.,
ty in organic solvents, fusibility, hydrolytic
1989 c) produced from polysilane and (b) a /?-Si3N4/ nature, viscosity or volatility, depend not
/?-SiC composite produced from polysilazane. The only on the type of polymer but also to a
particle size is extremely low (100-200 nm) and the great extent on the molecular weight distri-
particle size distribution is very narrow. bution and the degree of cross-linking. The
properties of the polymer have an impor-
sizes, and for the synthesis of nanocrys- tant influence on the ceramic production
talline materials. Nanocrystalline materi- and therefore have to be studied carefully.
als (Birringer and Gleiter, 1988; Burg- An increase in the molecular weight or
graaf, 1991; Karch and Birringer, 1990) the degree of cross-linking usually results
can exhibit interesting properties, such as in decreased solubility, reduced reactivity
superplasticity, and have particles in the with air and water, a lower volatility, and
range 10-100 nm. This means that the an increase in the softening temperature of
physical and mechanical properties are the polymer. Ceramic coatings can be pro-
mainly influenced by surface properties, duced either through the thermal decom-
due to the high surface-to-volume ratio position of low molecular weight, volatile
(Burggraaf, 1991). oligomers in the gas phase [metalorganic
Just recently the superplastic behavior chemical vapor deposition, MOCVD (Rie-
of a polycrystalline, covalent composite ce- del et al., 1990d; Interrante et al., 1986), or
ramic (Si3N4/20 wt.% SiC) was reported through the pyrolysis of higher molecular
(Wakai etal., 1990). The synthesis of this weight, insoluble polymer films on the sub-
material was based on the thermolysis of strate (Strecker, 1990). Soluble or liquid
hexamethyldisilazane ([(CH3)3Si]2NH) at polymers with low volatility are suited for
1000 °C in the gas phase under N 2 . This "dip" or "spin on" coating processes
results in an Si-C-N powder and if the gas- (Schmidt etal., 1988; Cranmer, 1988). In
phase decomposition is carried out in a the latter processes a polymer film is ap-
10 11 Advanced Ceramics from Inorganic Polymers

plied to a substrate and then chemically 11.3 SiC Ceramics from

(e.g. by hydrolysis) and/or thermally con- Poly(organo)silanes
verted to a ceramic film (Cranmer, 1988).
For the production of homogeneous and
dense coatings the quality of the wetting of Because of its hardness (9.5 on the Mohs
the polymer to the surface of the substrate scale) and its good oxidation and corro-
is important. sion resistance, silicon carbide (SiC) has
The molecular structure and the chemi- potential in the construction of motors and
cal composition of the organoelement turbines. The high hardness also results in
starting materials strongly influence the a high resistance to wear. Due also to its
constitution of the ceramic product. For good thermal conductivity and its semi-
the production of SiC, polymers with alter- conducting properties SiC is used as a
nating Si-C-Si structures are best suited grinding material, for protection against
and for Si 3 N 4 , those with Si-N-Si struc- abrasion, in high-temperature and chemi-
tures. From materials containing molecu- cal engineering, as well as in the produc-
lar Si-C-Si and Si-N-Si units, amor- tion of heat exchangers and heating ele-
phous silicon carbonitride (Si^C^NJ can ments.
be produced and heterogeneous SiC/Si 3 N 4 The bonding in SiC is primarily covalent
materials can be produced at higher tem- in nature, the ionic contribution, accord-
peratures. The molecular structure also af- ing to Pauling being only 12%. This is the
fects the pyrolysis behavior of the polymer. reason for the above mentioned proper-
Polymers with strongly branching net- ties. There are two modifications of SiC,
works or with ring-type structures result in a cubic, low-temperature modification
a high ceramic yield, whereas linear com- (P-SiC) and a hexagonal, high-tempera-
pounds on thermolysis generate only low ture modification (a-SiC). The phase
ceramic yields. The reason for this is the transformation p -> a occurs at around
easy thermal depolymerization of linear 2100 °C and the back reaction from a -* p
polymers through cyclization and the loss does not take place below this tempera-
of low-molecular-weight fragments (Wyn- ture, meaning that commercially available
ne and Rice, 1984). An example based on SiC mostly contains a-SiC. Another fea-
linear polysilane is ture of SiC are its extensive poly types.
Around 150 polytypes are known which
(CH3)3Si-[(CH3)2Si]n-Si(CH3)3 differ in their stacking modes. The most
^(CH3)3Si-[(CH3)2Si]n_6- common types are 4H, 6H, and 15 R (see
Si(CH 3 ) 3 + [(CH3)2Si]6 (11-7) Chapters 4 and 10 in Volume 11 of this Se-
ries). 4 H and 6 H are based on the hexago-
Due to this process linear polymers can nal structure, 4H, for example, having the
sometimes be evaporated without leaving stacking order ABAC ABAC . . ., whereas
a deposit and are therefore only of interest 15 R exhibits a rhombohedral structure
for gas-phase pyrolysis. Further cross-link-
based on the cubic modification. General-
ing of the linear polymers either chemically
ly, a-SiC contains a mixture of the various
(e.g. with catalysts) or physically (thermal-
polytypes. The binary system Si-C con-
ly or with UV light) can be used to increase
tains only one thermodynamically stable,
the ceramic yield of these materials.
stoichiometric compound, SiC (Scace and
Slack, 1960).
 11.3 SiC Ceramics from Poly(organo)silanes 11

The production of technical-grade SiC pler et al., 1987; Stolka et aL, 1987), and as
powder is carried out using the Acheson materials for microlithographic applica-
process, which involves the reaction of tions (Miller, 1989). Polysilanes have a
quartz with coke at temperatures above smaller bandgap than saturated carbon-
2000 °C: based polymers (4 eV against 8 eV). There-
fore, the doping of polysilane films with,
SiO2 3C->SiC + 3CO (11-8)
for example, AsF 5 , results in electrically
The gas-phase pyrolytic decomposition conducting films (West et al., 1981).
of silicon tetrachloride with hydrocarbons, Poly(diorgano)silanes are thermally sta-
or of tetramethylsilane or methylchlorosi- ble in air, but photochemically labile, they
lane between 1000 and 1400 °C is used to absorb light over a broad section of the
produce SiC coatings. SiC fibers are pro- spectrum, they are incompatible with most
duced using the polymer pyrolysis process organic polymers, they are relatively stable
described in detail in Sec. 11.2. The pro- against etching processes in oxygen plas-
duction of ceramic bulk materials directly mas, and can be sensitized using X-rays,
from organoelement polymer compacts is gamma rays, and electron beams. These
a completely new process which will be properties make polysilanes interesting for
described in detail in Sec. 11.5. A further lithography. Recent studies demonstrate
applicational possibility for the use of the nonlinear optical activity of the polysi-
solid-state pyrolysis is in the production of lanes, which opens up possible applica-
composite materials through the impreg- tions in the telecommunications industry
nation of porous substrates with polymeric (Miller, 1989).
ceramic precursors followed by thermal The a-conjugated backbone of the
decomposition, or the pyrolysis of parts polysilanes is highly delocalized, resulting
made of polymer-ceramic powder mix- in high absorption in the UV range. In this
tures. respect the electronic properties of the
In the following section an introduction polysilanes [e.g. poly(di-w-hexyl)silane
to the chemistry of the polysilanes will be (Michl et al., 1988)] are similar to those of
given before the synthesis of polysilanes, 7i-conjugated systems, such as poly-
and carbosilanes, and their pyrolysis to acetylene (Kuzmany et al., 1985).
form silicon carbide based materials is dis- Poly(diphenyl)silane was first described
cussed. in the 1920s (Kipping and Sands 1921)
but it was only many years later
that poly(dimethyl)silane was synthesized
11.3.1 Poly(diorgano)silanes -
(Burkhard, 1949). According to the
Properties and Synthesis
IUPAC regulations poly(dimethyl)silane
Poly(diorgano)silanes exhibit a number should be denoted as catena-po\y[di-
of interesting chemical and physical prop- methylsilicon], (Donaruma etal., 1981).
erties which make them technologically Poly(dimethyl)silane is a colorless powder,
important. Apart from their use as precur- insoluble in organic solvents, and decom-
sors for SiC ceramics, polysilanes are em- poses without melting at temperatures
ployed as photoinitiators for radical reac- above 250 °C. It was only with the discov-
tions, for example vinyl polymerization ery that poly(dimethyl)silane could be
(Wolff and West, 1987), they can be used as used to produce SiC fibers (Verbeek, 1973;
photoelectric conducting materials (Ke- Verbeek and Winter, 1974; Yajima et al.,
12 11 Advanced Ceramics from Inorganic Polymers

1976) that serious interest was shown in potassium alloy in boiling toluene results
the material, an interest which continues in the copolymer with the IUPAC (Donar-
today. A comprehensive review of the syn- uma etal., 1981) name catena-poly[(di-
thesis, characterization, and properties of methylsilicon)(methylphenylsilicon)], with
poly(diorgano)silanes has been provided the idealized formula
by West (1989). xme 2 SiCl 2 +jme(ph)SiCl 2
The production of poly(diorganyl)si-
^^>H( m e 2Si)ime(ph)Si),] n
lanes is generally done using a process
analogous to the Wurtz-Fittig synthesis, + 2(x+j)NaCl/KCl (11-10)
which involves the dehalogenation of where me stands for methyl, and ph,
dichlorodiorganosilanes with sodium, phenyl (Mazdiyasni et al., 1978).
potassium, or a sodium-potassium alloy. With a 5:1 ratio of me2SiCl2 to me(ph)-
The mechanism of the reaction is still not SiCl2 there are two important advantages
clear, both silyl radicals and silyl anions in the pyrolytic synthesis of SiC from the
being suggested as intermediates in the copolymer. Firstly, Mazdiyasni et al. (1978)
chain-growth process (Worsfold, 1988). have shown that conversion of the copoly-
The reaction results in a mixture of linear mer into SiC without the need for interme-
and cyclic oligomers as well as a high- diate tempering in an autoclave is possible.
molecular-weight fraction, the relative This is in contrast to pure poly(dimethyl)-
proportions of which are determined by silane which has to be converted first into
variation of the solvent, the type of sodium polycarbosilane at 450 °C and a pressure of
dispersion, and the order in which the 10 MPa before the pyrolysis to form SiC
reagents are added to the reaction mixture. (Yajima etal., 1978a). Secondly, the 5:1
Processes for the synthesis of polysilanes ratio results in a relatively low phenyl con-
that obviate the need for the use of sodium tent; high phenyl contents lead to high car-
or potassium are also known. Diorganosi- bon excesses in the SiC.
lanes can be produced through the anionic The chemical and physical properties of
polymerization of l-phenyl-7,8-disilabicy- the polymeric products depend on the
clo[2.2.2]octa-2,5-diene with alkyllithium length of reaction and the subsequent heat
(Sakamoto et al., 1989). The transition- treatment (Fig. 11-4). After 36 h reaction
metal catalyzed condensation of mono- an amber colored oil [poly(methylphenyl)-
organosilanes silane (PMPS A)] and a large proportion
CH SiH °-2 m o l % Cp2Ti(CH3)2/toluene^ of insoluble and infusible material were
obtained. The insoluble poly(methyl-
H-[CH 3 SiH] x -H + H 2 (11-9) phenyl)silane, which is produced in yields
results in poly(organo)silanes, and in the of up to 30%, can be annealed at 450 °C in
case of methyl substituents these materials argon to form soluble polycarbosilane, a
yield almost stoichiometric SiC (Zhang process which increases the yield of soluble
etal., 1991). materials to 88 %. The oil can be pyrolyzed
at 1000 °C with a ceramic yield of 44%.
11.3.2 Synthesis of Polysilanes and In contrast, a reaction time of only 24 h
Polycarbosilanes from Dichlorodimethyl- results in a colorless, opaque product
and Dichloromethylphenylsilane (PMPS B), which leaves no residue on
The dechlorination of a mixture of heating to 1000 °C. If PMPS B is subse-
dichloromethylphenylsilane with sodium/ quently annealed for 7 h at 500 °C in argon
 11.3 SiC Ceramics from Poly(organo)silanes 13

5Me2SiCl2

rt# JPhMeSiCl 2 |^ jV)

POLYSILANE A POLYSILANE B
(amber oil) (opaque oil)

Ar/500°C/7h

Ar/l000°C POLYSILANE C
no SiC
(waxy solid or amber oil)

Ar/l000°C

44 wt/o ceramic yield

51 - 68 wt/o ceramic yi
75 wt/o SiC + 25 wt/o C

Figure 11-4. Dependence of the properties of the product on the reaction time and the heat treatment during the
production of c<3^wa-poly[(dimethylsiliconXmethylphenylsilicon)] (Riedel et al., 1989 c).

at 0.1 MPa, here also an amber, highly vis- 25 wt.% carbon. This content was calcu-
cous product (PMPS C) is obtained which lated based on the assumption that all Si
on pyrolysis at 1000 °C in argon has a ce- atoms are bound to carbon to form SiC,
ramic yield of 51 %. The distillation of the and since SiC dissolves almost no carbon,
volatile fraction under vacuum (200 °C, excess carbon must be present as elemental
10~ 3 Torr) results in a wax-like material carbon. Table 11-3 contains the results of
(PMPS C*) which on pyrolysis at 1000 °C the elemental analysis and ceramic yields
in argon shows a ceramic yield of 68 %. of the various products PMPS B, C, and
The annealed reaction product is fusible C*. The analytical data for the untreated
(softening point = 120°C), and soluble in polysilane product PMPS B are in good
organic solvents, which means it can be agreement with the theoretical values cal-
used for coating substrates and for the in- culated on the basis of Eq. (11-10).
filtration of porous materials. The infiltra- The electron ionization (El) mass spec-
tion of organoelement ceramic precursors trum of PMPS B indicates the presence
in porous substrates is of particular impor- of volatile, cyclic oligomers. Fragments
tance in the development of carbon- and from methylphenylsilane, for example
ceramic-fiber reinforced composites. trimethylsilyl, as well as [Si(CH3)2]6 (m/e
The ceramic material obtained from 348), [Si(CH3)2]5[C6H5SiCH3] (m/e 410),
PMPS C* contains about 75 wt.% SiC and [Si(CH 3 ) 2 ] 4 [C 6 H 5 SiCH 3 ] 2 (m/e 472), and
14 11 Advanced Ceramics from Inorganic Polymers

Table 11-3. The dependence of the chemical analysis and ceramic yield of the copolymerisates PMPS B, PMPS C
and PMPS C* on the heat treatment employed (Riedel et al., 1989 c).

Copolymerc Heat treatment3 Composition Ceramic yield b

T (°C)/time (h) (wt.%) (wt.%)

C H Si

PMPSB d 49.6 9.1 39.0 <1

PMPSB 350/24 48.4 9.2 40.0
PMPSB 450/24 47.9 8.9 44.0 44
PMPSC 500/7 42.9 8.4 48.0 51
PMPSC* e 500/7 37.2 7.2 55.7 68
a
The heat treatment takes place at 0.1 MPa under flowing argon; b after pyrolysis up to 1500°C in flowing
argon; c all polymers are soluble in organic solvents; d reaction product without subsequent heat treatment;
e
PMPS C where the low-molecular-weight components have been distilled off.

[Si(CH3)2]3[C6H5SiCH3]3 (m/e 534) are nealed polymer PMPS C*. The theoretical
seen. From this it can be concluded that ceramic (SiC) yield from PMPS C* is 81 %,
the ratio of dimethylsilylene to methyl- which can be compared to the experimen-
phenylsilyene groups varies greatly along tally found 68 % for a heating rate of 5 °C/
the polymer chain resulting in regions of min under a continuous argon flow (Riedel
high and low carbon content. etal., 1989c).
On subsequent heat treatment at 500 °C
the carbon content decreases and the Si
content increases (Table 11-3) which indi-
11.3.2.1 Thermogravimetric Analysis
cates that in the temperature range be-
tween 350 °C and 500 °C mostly carbon- The ceramic yield and the behavior of
containing products are expelled. The loss the polymers during pyrolysis can be stud-
of methane and hydrogen can be observed ied using thermogravimetric analysis
above 300 °C using mass spectrometry. (TGA; see the chapter by Gallagher in Vol-
The weight loss during heat treatment for ume 2 A of this Series). A typical TGA
7 h at 500 °C under argon is between 45 curve of poly(methylphenyl)silane PMPS
and 50 % and is partially the result of de- C* is shown in Fig. 11-5. Three weight-loss
polymerization reactions (Wynne and regions can be identified. The first stage is
Rice, 1984). The annealed polymers are the evaporation of low-molecular-weight
soluble in organic solvents such as THF silanes, the second is the start of the
but the solubility decreases quickly if the thermal decomposition accompanied by a
annealing time at 500 °C is greater than change in the molecular structure (poly-
7h. silane -> polycarbosilane), and the third
The ceramic yield depends to a great ex- weight-loss stage is the loss of hydrogen
tent on the heat treatment of the copoly- and methane. This interpretation of the
mers. A heat treatment at around 500 °C TGA curve is supported by the mass spec-
after the synthesis results in higher ceramic tral analysis of the gaseous reaction prod-
yields. The composition changes from ucts up to 1500°C (Riedel etal., 1989c).
C 3 H 6 5Si in the polymeric reaction prod- Table 11-4 contains the details of these
uct (PMPS B) to C 1 5 H 3 5Si in the an- mass spectrometric studies.
 11.3 SiC Ceramics from Poly(organo)silanes 15

Figure 11-5. TGA and DTA

curves of the copolymer
product PMPS C* under
an argon flow. Heating rate
5°C/min. (1) Evaporation
of volatile components. (2)
Polycondensation and de-
composition. (3) Loss of H 2
and CH 4 (Riedel et al,
1989 c).
200 400 600 800 1000 1200 1400
Temperature [°C]

Table 11-4. Reactions of PMPS C* which occur during pyrolysis between room temperature and 1500 °C (Riedel
etal., 1989 c).

Temperature Reactions Gaseous species and fragments3

interval

U] < 300 °C Evaporation of volatile compounds [(CH3)2Si]6 (348), [[(CH3)2Si]5 • [phSiCH3]] (410)
\J\ 300-450 °C Polycondensation, initial decomposition, H 2 (2), CH 4 (16), (CH3)2SiH (59), CH3Si (43)
conversion to polycarbosilane C 3 H 8 (44), ph(CH3)Si (135), (CH3)3Si (73),
C 6 H 6 (78)
450-900°C Main decomposition H 2 (2), CH 4 (16), ph(CH3)2Si (135), C 6 H 6 (78)
> 1200 °C Reaction of Si-O groups with CO (28), SiO (44)
elemental carbon

Using electron ionization mass spectrometry (EI-MS). The m/e values are given in brackets. ph = phenyl.

The mass spectra can only be measured 600 °C, and of H 2 at 680 °C. The hydrogen
in high vacuum conditions resulting in the ion flux exhibits another broad maximum
weight-loss steps 1, 2, and 3 in Fig. 11-5 at 1150°C but this has a lower relative in-
being shifted to lower temperatures. Up to tensity (7%). In the temperature range up
300 °C the cyclic oligomers (see above) and to 700 °C the loss of Si-containing com-
their fragmentation products observed in pounds can also be seen. The (CH3)3Si
the mass spectrum (e.g. (CH3)3Si (m/e 73) fragment appears between 100 and 450 °C,
or C 6 H 5 Si(CH 3 ) 2 (m/e 135). Figure 11-6 and the ion flux of dimethylphenylsilane
shows the change in the H 2 (m/e 2) and (m/e 135) fragment has maxima at 380 and
CH 4 (m/e 16) peaks between room tem- 650 °C. The fragmentation of oligosilanes
perature and 1500°C. H 2 and CH 4 loss demonstrates the existence of very stable
begins at 300 °C whereby the maximum R 1 R 2 R 3 Si + ions (Saalfeld and Svec, 1963,
loss (100% relative ion flux) of CH 4 is at 1964).
16 11 Advanced Ceramics from Inorganic Polymers

< 3 .

o 2
Figure 11-6. Mass spectral ion
current of H 2 (m/e = 2) and CH 4
(m/e = 16) during the pyrolysis of
PMPS C* under high vacuum be-
tween 25 and 1500 °C.
200 A00 600 800 1000 1200 1400

Temperature [°C]

The variation of the total gas pressure is drogen and methane. Between 300 and
mainly caused by the generation of H 2 and 900 °C H 2 and CH 4 are split off, which
CH 4 due to the thermal decomposition of results in a large increase in gas pressure
the polymers (Fig. 11-7). Despite there be- but only a small weight loss. Between 1200
ing only a small increase in the total gas and 1500°C another large increase in the
pressure between 25 and 400 °C there is a total gas pressure is observed (Fig. 11-7)
weight loss of 55 wt.% (this weight loss is which is due to the reaction of Si-O units
higher than that measured for pyrolysis with elemental carbon and the generation
in argon (32%) as it is measured in high of CO [Nickel et al. (1988)]
vacuum, see Fig. 11-5). The greater weight
loss at almost constant gas pressure is due SiO2 + 3 C ^ S i C + 2CO (11-11)
to the evaporation of cyclic silanes which
have a higher molecular weight than hy- SiO2 + C-+SiO + CO (11-12)

3
2 Figure 11-7. Weight loss
5 and total gas pressure as
a function of temperature
during the pyrolysis of
PMPS C* under high vac-
uum between 25 and
1500°C(RiedeletaL,
1989 c).
200 £00 600 800 1000 1200 U00
Temperature [°C]
 11.3 SiC Ceramics from Poly(organo)silanes 17

Mass spectrometry indicates that in

temperature range 4 (Table 11-4) the gas
present is mainly made up of CO (m/e 28)
and SiO (m/e 44). These gases are generat-
ed as a result of the oxygen content of the
polymers, which due to the aqueous work-
up is around 2.6 wt.%. The linear polysi-
lane chains contain Si-Cl end groups that
form silanol groups on hydrolysis, which
in turn lose water to form siloxanes.
Through the nonaqueous work-up of the
reaction mixture the oxygen content can be
held below 0.3 wt.%.
C-H
11.3.2.2 Infrared and NMR Spectroscopic
Studies of Poly(methylphenyl)silanes
and Carbosilanes
(CH 3 ) 2 Si
~T 1 1 1—1—1—1 p^ r
Infrared spectroscopy can be used to
follow the structural changes taking place £000 3000 2000 U00 1000 600
[cm* 1 ]

Tabelle 11-5. Infrared absorption frequencies of Figure 11-8. Infrared spectrum of poly(methyl-
poly(methylphenyl)silane PMPS B annealed at 350 °C phenyl)silane PMPS B after synthesis (a), after the
for one day (Riedel et al., 1989 c). heat treatment at 350 °C/1 d under argon (b), and at
500°C/7h(c).
Wave- Rel. Assignment Vibra-
number intensity8 tional
(cm- 1 ) mode b during the isothermal annealing process.
The occurrence of characteristic vibra-
3080 m Ph H st tional bands in the IR spectrum due to the
3060 m Ph-H st
C-H
generation of SiH and Si-CH 2 -Si groups
2965 vs st
2900 s C H st in annealed poly(methylphenyl)silane
2110 m Si-H c st PMPS B is observed. Table 11-5 contains
1600 w C=C st IR data for PMPS B annealed at 350 °C for
1492 w C=C st one day. Stretching and bending vibra-
1438 s CH 3 -Si 3
tions which can be assigned to Si-H and
1410 s CH 3 -Si 3
1270 vs CH 3 -SI Si-CH 2 -Si groups are not present in the
y
1255 vs CH 3 -Si y spectra of PMPS A and B. Figure 11-8
1085 vs Si-O-Si st shows the IR spectra of both PMPS B and
1028 vs Si-CH 2 -Si c — the polymers which are the result of an-
890 s Si-H c 3
nealing at 350 and 500 °C. Comparison of
800 vs (CH3)2Si Q
735 s CH 3 -Si the IR spectra of annealed and unannealed
Q
700 s Si-C st polymer shows that heat treatment at
652 s Si-C st 350 °C results in the conversion of the
a
polysilane structure (with Si-Si-Si units)
w = weak, m = medium, s = strong, vs = very strong;
b
st = stretching, 3 = bending, y = wagging, Q = rocking; to the polycarbosilane structure (with
c
absorptions not observed in unannealed PMPS B. Si-C-Si units).
18 11 Advanced Ceramics from Inorganic Polymers

The polysilane-polycarbosilane conver- tion of silyl radicals. In poly(diorganyl)si-

sion can also be followed using 13 C NMR. lanes the Si-Si bond, which has an average
In the DEPT spectrum (DEPT stands for bond energy of 210-250 kJmol" 1 , is
distortionless enhancement by polariza- weaker than the Si-C bond (250-
tion transfer) of PMPS C* the carbon 335kJmol" 1 ) and is therefore preferen-
atoms characteristics for the Si-CH 2 -Si tially broken. It is assumed that the silyl
of carbosilane are observed with chemical radicals subsequently rearrange to form
shifts of 5 = 23 (-CH 3 SiH-CH 2 -) and methylene radicals = Si-CH 2 -, which
33 (-phSiH-CH 2 -). The methyl groups eventually recombine to form polycarbosi-
CH 3 Si= and(CH 3 ) 2 Si = exhibit chemical lanes based on the Si-CH 2 -Si structural
shifts of-6 to + 5 ppm. The phenyl region unit (Yajima et al., 1978b; Yajima, 1985).
(128-136 ppm) of the annealed material
exhibits no changes relative to the unan- -(CH 3 ) 2 Si-(CH 3 ) 2 Si-(CH 3 ) 2 Si-
nealed starting polymer. This indicates ( C H 3 ) 2 S i - ^ - ( C H 3 ) 2 S i - C H 3 S i ( C H 2 •)
that the silicon-bound phenyl groups are + H-Si(CH 3 ) 2 -Si(CH 3 ) 2 ->-(CH 3 ) 2 Si-
not involved in the structure conversion CH 3 SiH-CH 2 -(CH 3 ) 2 Si-(CH 3 ) 2 Si-
during the annealing process at 500 °C.
The 29Si spectrum of PMPS C* exhibits The formation of the Si-H groups can
three singlets at 5 = 38.0 ppm (relative in- also be seen as the insertion of a CH 2
tensity 7%), -38.4 ppm (100%) and group into an Si-Si bond and it was in this
-38.6 ppm (7%) (Strecker, 1990). way that it was described for the pyrolysis
of hexamethyldisilane and trisilacyclopen-
The Mechanism of Formation of tadiene (Davidson and Eaborn, 1974; Fritz
Polycarbosilanes from Polysilanes and Grunert, 1976; Davidson et al., 1982).
The loss of H 2 , CH 4 , and higher hydrocar-
This reaction involves the well-known
bons can also be explained on the basis of
insertion of methylene into an Si-Si bond
a radical mechanism. It has also been
(Fritz, 1987), see Formula 1. The rear-
shown that the linear polysilane chains can
rangement to form polycarbosilanes has
be cross-linked (Okamura, 1987). The
been explained in terms of a radical mech-
higher SiC yields from the decomposition
anism (Yajima et al., 1978 b; Yajima,
of phenyl-containing polysilanes could be
1985). In order to examine the mechanism
explained by the relatively easy loss of
the products of the thermal decomposition
phenyl radicals followed by hydrogen ab-
of hexamethyldisilane have been charac-
straction and recombination to form poly-
terized in detail (Sakurai et al., 1968; Shi-
carbosilane (Carlsson et al., 1990), see
ma and Kumada, 1958; Davidson and
Formula 2.
Stephanson, 1968). The first stage is the
However, taking into account the rele-
breaking of an Si-Si bond and the forma-
vant bond energies it is less likely that the
reaction proceeds through the formation
H3C CH3 H3C H of methylene radicals [ = Si-CH 2 -, (Cot-
I I 35O°C/ld ,
ton and Wilkinson, 1980)].
-Si-Si- -Si-CH?-Si-
I I I = 210-250 kJmol - I
H3C CH3 CH3

Formula 1
 11.3 SiC Ceramics from Poly(organo)silanes 19

-CH 3 Siph-CH 3 Siph- -> -CH 3 Siph-CH 3 Si-+ ph-

ph CH3 ph ph
Recombinatlon
-(CH 2 )Siph-CH 3 Siph-+C 6 H 6 > -Si-Si-CH2-Si-Si-
Formula 2 CH3 CH3

Another possibility is therefore the elim- weight seen are in the region of 65000
ination of dimethylsilylene [(CH3)2Si:] or g/mol.
methylphenylsilylene [CH 3PhSi:] from (2) After annealing the PMPS B at 350°C
PMPS according to for one day a decrease in the high molecu-
lar weight fraction was observed, due to
[(CH 3 ) 2 Si] n ^>[(CH 3 ) 2 Si] ri _ 1
depolymerization, so that the maximum at
+ (CH 3 ) 2 Si: (11-13) 350 g/mol is now only seen as a shoulder in
followed by repolymerization and inser- the high molecular weight region.
tion to form polycarbosilane (Atwell and (3) Polysilane which has been annealed at
Weyenberg, 1969). The silylene mechanism 450 °C for one day exhibits a similar
is also supported by Sakurai et al. (1969), molecular weight distribution to that pro-
who observed the formation of higher duced at 350 °C.
poly(methyl)silanes from the pyrolysis of
pentamethyldisilane. Although the struc-
tures of many low-molecular-weight car-
bosilanes resulting from pyrolysis reac-
tions have been elucidated, in particular by
Fritz (1987), the structures of the higher
molecular weight carbosilanes are still al-
most completely unknown.
PMPS C*
11.3.2.3 Molecular Weight Distribution
Analysis of the Poly(methylphenyl)silanes PMPSC
500°C/7h
A shift in the molecular weight distribu-
tion results as the amount of cross-linking
increases with increasing temperature. Fig-
ure 11-9 shows the results of gel perme- PMPS B
ation chromatography (GPC) of materials 350°C/1d
annealed at different temperatures. The
following differences to unannealed sam- PMPS B
ples are seen.
215 68 23 7.4 0.6
(1) PMPS B exhibits a bimodal distribu-
tion. The sharp peak at the relative molar
MW x 10" 3
mass 350 g/mol is assigned to the cyclic do-
Figure 11-9. Molecular-weight distribution analysis
decamethylcyclohexasilane ([Si(CH3)2]6). (GPC) of the poly(methylphenyl)silanes PMPS B,
Next to this, a broad maximum at 3500 PMPS C and PMPS C*. The molecular weights are
g/mol is observed. The highest molecular given relative to polystyrene.
20 11 Advanced Ceramics from Inorganic Polymers

(4) On annealing at 500 °C for 7 h, a strong suitable monomers to form boron-con-

increase in molecular weights is observed, taining Si-polymers. There are no litera-
resulting again in a bimodal distribution, ture reports of such work, a particular aim
this time, however, with a higher propor- of which was to discover whether the hy-
tion of high molecular weights. The droboration of Si-polymers would work
highest molecular weight is ca. 50000 g/ and whether boron-containing precursors
mol. could be pyrolyzed to form boron-contain-
(5) Vacuum distillation at 10" 3 Torr can be ing ceramics.
used to further increase the high molecular
weight portion, at the same time reducing
11.3.3.1 Hydroboration of
the intensity of the peak at 350 g/mol.
Poly(methylvinyl)silane
Boron-containing polysilanes could, in
11.3.3 Synthesis of Boron-Containing
the past, only be synthesized through the
Poly(organo)silanes
reaction of dichloroorganylsilanes and
The synthesis of new, boron-containing dichloroorganylboranes with alkali metals
polysilanes is interesting for the following in aprotic solvents (Riccitiello et al., 1987).
reasons: This method results in the formation of
• For the production of sinter-active, polymers containing Si-B-Si chains. Par-
boron-containing SiC powder. Because of ticularly with the aim of being able to con-
the distribution of the boron at the molec- trol the polymer properties, a search for
ular level, boron-containing polysilanes new synthetic methods was undertaken.
are SiC precursors which on pyrolysis yield Soluble polysilanes should be the starting
nanodisperse boron carbide: boron car- materials which can be chemically modi-
bide (B4C) functions as a sintering aid in fied at their reactive side-group. The vinyl
the compaction of SiC powders (Schwetz group, which, for example, can be hydrob-
and Lipp, 1980). Diffusion processes im- orated with borane, proved to be a suitable
portant in sintering occur faster in this ma- reactive side group.
terial as the diffusion distances are smaller. The hydrosilylation of dichlorovinylsi-
• For the homogeneous coating of con- lane followed by reduction of the Cl
ventional SiC powder with B4C precur- groups with LiAlH 4 has been shown to
sors. Boron-containing polysilanes can al- yield ethylene-bridged poly(organyl)si-
so be used as binders in the injection mold- lanes (Boury et al., 1990). The requirement
ing of SiC powders. The polysilane on the for this is the synthesis of polysilanes with
powder surface flections as binder but al- intact vinyl groups. This has been achieved
so as sintering aid as on pyrolysis it decom- by the reaction of dichlorodimethylsilane
poses to homogeneously distributed SiC, and dichloromethylvinylsilane with sodi-
C, and B 4 C. um (Schilling, 1986). By using Na/K alloy,
• For the targeted influence of polymer however, apart from the dechlorination
properties such as Tg (glass or softening the vinyl groups are also attacked. As a
transition temperature), viscosity or ce- result, trimethylvinylsilane in the presence
ramic yield. of chlorotrimethylsilane and potassium in
Therefore, we have studied the hydro- toluene reacts to form tris(trimethylsi-
boration of polysilanes containing vinyl lyl)ethane while with sodium no analogous
groups as well as the polycondensation of reaction is observed (Schilling, 1986).
 11.3 SiC Ceramics from Poly(organo)silanes 21

Therefore, it is clear that with potassium Table 11-6. Change in the molecular weight distribu-
silylation of the vinyl group dominates tion of poly(methylvinyl)silane (PVS) on hydrobora-
tion.
the dechlorination. The reaction of di-
chloromethylvinyl- and dichlorodimethyl- Molecular weight PVS Hydro-
silane in a 1:1 molar ratio, with sodium borated
and in the presence of chlorotrimethylsi- PVS
lane in a solvent mixture of toluene and Mn
THF at 100 °C leads to the formation of No. average of the molecular
the soluble ctfte«a-poly[(dimethylsilicon)- weight (g/mol) 988 1464
(methylvinylsilicon)] (PVS) in 84% yield.
A solution of PVS in THF is reacted Weight average of the
with THF-borane (see Formula 3) or molecular weight (g/mol) 5 538 23 664
Mcf
Centrifuge average of the
I molecular weight (g/mol) 50 371 574181
RSi-CH=CH 2
Highest molecular weight
(g/mol) 314420 5 060468

THF • BH3
I /
RSi-CH 2 -CH 2 -B
act as a sintering aid at temperatures above
Formula 3 1000 °C and lead to a reduction of the
porosity and therefore to an increase in the
with the dimethylsulfide-borane complex sintered density.
(CH 3 ) 2 SBH 3 at room temperature to
produce boron-containing polysilanes. In- 11.3.3.2 Poly condensation of
frared and 13 C-NMR spectroscopy detect Tris[(dichloromethylsilyl)ethyl]borane
a significant decrease in the relative inten- to Form Boron-Containing Si Polymers
sities of the vinyl C - H vibrations at
A further possibility for the production
3050 cm" 1 and <5 = 131 ppm respectively.
of boron-containing Si Polymers is the use
The resin-like hydroboration product has
of tris[(dichloromethylsilyl)ethyl]borane
a higher viscosity than the oil-like PVS and
as a monomeric starting material. This ma-
can as a result be drawn into fibers. The
terial was first synthesized by Mikhailev
cross-linking of PVS over the vinyl groups
and Aronovich (1960) and used by Jones
also results in a shift in the molecular-
and Lim (1976) as an intermediate in the
weight distribution, from 988 g/mol in
production of alcohols through oxidative
PVS to 1464 g/mol in hydroborated PVS
protolysis.
(see the GPC data in Table 11-6).
The reaction of dichloromethylvinylsi-
By controlling the proportion of vinyl lane with THF-borane [Eq. (11-14)] yields
groups in the PVS the amount of cross- a mixture of stereoisomers.
linking on hydroboration and therefore al-
so the properties of the polymer can be 3 H 2 C = CH-SiCl 2 -CH 3 + THF • BH 3
influenced. The production of highly ->B[CH(CH 3 )-SiCl 2 -CH 3 ] 3 _ :c
cross-linked, infusible, boron-containing [CH 2 -CH 2 -SiCl 2 -CH 3 L (11-14)
polysilanes is also of interest in the solid-
state pyrolysis of compacted polymer pow- Apart from the addition of BH 3 in
ders (see Sec. 11.3). In situ formed B4C can the P position [anti-Markovnikov addi-
22 11 Advanced Ceramics from Inorganic Polymers

tion, Eq. (11-16)] the hydroboration in the silyl)ethyl]borane. A wavelength-disper-

a position producing a chiral center is fa- sive X-ray analysis (WDX) in the SEM
vored [Eq. (11-15)]: reveals that the ceramic material is homo-
geneous. Figure 11-11 shows the various
CH 3 ~Cl 2 Si-CH(CH 3 )-B< (a-product)
reactions possible with the monomeric bo-
(11-15)
rane compounds for the production of new
C H 3 - C l 2 S i - C H 2 - C H 2 - B < (p-product) organoelement polymers.
(11-16)
Tris[(dichloromethylsilyl)ethyl]borane 11.3.4 X-Ray Studies of Ceramic
can be converted into new organoelement Materials Produced from
polymers in a number of ways. Reaction Poly(methylphenyl)silanes
with water leads to boron-containing The pyrolysis of the pure poly(methyl-
polysiloxanes, reaction with ammonia phenyl)silane PMPS C* usually gives
leads to the respective polysilazanes, and foam-like SiC2 ceramic products due to
with Na/K alloy leads to boron-containing intermediate melting. On part of the sur-
polysilanes. Pyrolysis of these polymeric face of the pyrolyzed product honeycomb-
materials at 1100°C under argon results in like structures are formed which penetrate
ceramic materials in 62-70 wt.% yield. El- the bulk forming a porous network of
emental analysis of these X-ray amor- channels which carry away reaction gases
phous ceramic products gives the follow- during the thermal decomposition. Such
ing empirical formulas: highly porous materials could find appli-
• The cross-linking of B[CH(CH 3 )- cation as chemically resistant filters or cat-
Si(Cl 2 )-CH 3 ] 3 _JCH 2 -CH 2 -Si(Cl 2 )- alysts. Only above 1200 °C, that is above
CH3]X with H 2 O in THF followed by the pyrolysis temperature, is crystalliza-
pyrolysis at 1100°C in Ar gives tion detectable by X-ray diffraction
Si2.sC5.1O4 8 B 1 0 . >1200°C
P-SiC + xC (11-17)
• Cross-linking from B[CH(CH 3 )-
Si(Cl 2 )-CH 3 ] 3 _ JCH 2 -CH 2 -Si(Cl 2 - Transmission electron microscopy
CH3]X with NH 3 in THF followed by (TEM) reveals, however, individual crys-
pyrolysis at 1100°C in Ar gives
^ J 3.0^4.3^2.0^1. OOQ. 4*

The elemental analysis shows that the

Si/B atomic ratio in the starting materials
is maintained in the reaction products. No
reversal of the hydroboration with loss of
borane is observed in this case although it
has in others (Elschenbroich and Salzer,
1989). The water and ammonia cross-
linked polymer materials can be drawn in-
to fibers which on pyrolysis yield ceramic
fibers with the above composition.
Figure 11-10. Ceramic S i - C - N - B fibers produced
Figure 11-10 shows an SEM image of an by the pyrolysis (at 1100°C under argon) of polymer
Si-B-C-N ceramic fiber produced from fibers obtained by the ammonolysis product of
ammonia cross-linked trisfdichloromethyl- tris[(dichloromethylsilyl)ethyl]borane.
 11.3 SiC Ceramics from Poly(organo)silanes 23

CH
\ ^ n 33 Cv-»Hn 33
+ NH 3
•N-
H
B
\ \

B-CH-CH 3
C 2 H 4 (OH) 2 r ?H3 ?H3 i
-Si-O-CH 2 -CH 2 -O-Si-O- L
L i i -I
SiCI 2 (CH 3 )
H 3 C-CH CH-CH 3
CH,
B B
/ \

+ R2SiCI2+Na CH3

[-R 2 Si-Si-R 2 Si-] n

R= -CH 3 ;Phenyl; Vinyl etc. H3C-CH
B
/ \
Figure 11-11. Possibilities for the synthesis of Si-based polymers from tris[(dichloromethylsilyl)ethyl]borane.

tallites in the amorphous matrix of the py- ygen) are also observed. Figure 11-12
rolysis product produced at 1000 °C which shows X-ray powder diffraction results
have been identified as (3-SiC on the basis taken on PMPS C* which has been py-
of electron diffraction. (3-Cristobalite crys- rolyzed at 1000 °C in Ar and annealed at
tals, formed from oxygen-containing im- various temperatures. The pyrolysis prod-
purities (PMPS C* contains 2-3 wt.% ox- uct subsequently heat-treated at 1250°C

JS
^
i/V 1750
2000

Figure 11-12. X-ray diffrac-

1500 tion patterns of PMPS C*
which has been pyrolyzed
1250 at 1000°Cfor 2 h and an-
/ nealed for 1 h under argon
1000 at different temperatures.
20 40 60 80
All reflections can be as-
20 CuK,, signed to j#-SiC
24 11 Advanced Ceramics from Inorganic Polymers

exhibits every broad X-ray reflections Table 11-7. Chemical analysis and SiC content of ce-
which sharpen on annealing at higher tem- ramic materials produced from PMPS C* at various
pyrolysis temperatures under argon (Riedel et al.,
peratures and which can be assigned to the 1989 c).
cubic SiC phase (P-SiC). The sizes of the
crystallites produced at different tempera- Temper- Composition (wt.%) SiC content (wt.%)
tures can be estimated on the basis of the ature (Molar
(°Q Si C composition)
breadth of the (100) reflections of (3-SiC
(20 = 35.61°), using the Scherrer equation 1000 52.4 45.3 75 (SiC2 0)
(Cullity, 1956) 1250 49.9 49.7 71 (SiC2.3)
1500 49.7 49.9 71(SiC 24 )
KX 2000 49.1 49.5 70(SiC 24 )
P = /zcosO (11-18)

where ^ i s a constant with a value of 1.08

for cubic materials, X is the wavelength of the material up to temperatures of 2200 °C.
the X-rays (XCuKa = 154.18 pm), h is the Surprisingly, the conversion of (3-SiC to
crystallite size, 6 the diffraction angle and oc-SiC does not occur up to 2200 °C, a pro-
p the integrated breadth of the reflection. cess normally observed around 2100 °C
The dependence of the particle size on (Williams et al., 1985).
the annealing temperature after a 1 h hold- SiC2 materials crystallized at 2000 °C ex-
ing period is shown in Fig. 11-13. The av- hibit an ultra fine grain size and also an
erage value of the particle size increases extremely narrow grain size distribution
from <10nm at 1250 °C to 45 nm at (Fig. 11-3). This shows that the production
2000 °C. After annealing at 1750°C aver- of nanocrystalline SiC ceramics using
age crystallite sizes of only 15 nm are still solid-state pyrolysis is in principle possi-
the norm. Excess elemental carbon ble. In contrast, conventional powder
(Table 11-7) cannot be detected using X- metallurgy cannot produce comparable
ray techniques even after heat-treatment of SiC materials as the starting particles have
diameters d50 of 0.3-0.4 jim and higher
sintering temperatures would be necessary.
Elemental analysis of the pyrolysis
products shows large excesses of carbon
(25-30 wt.%) in SiC (Table 11-7). This ex-
cess increases on annealing at high temper-
atures as part of the oxygen-bound silicon
is lost in the form of SiO, a process which
can also be followed mass spectrometrical-
ly (Table 11-4).
The pyrolysis product obtained from
PMPS C* at 1000 °C can be hot-pressed
without additives at 2200 °C and 62 MPa
1000 1200 1400 1600 18002000 to make SiC-based compacts with a rela-
Temperature [°C] tive density of around 82 %. The bulk den-
Figure 11-13. Dependence of the average crystallite sity was calculated (75 wt.% SiC, 25 wt.%
size of pyrolyzed PMPS C* on the annealing temper- C) to be 2.87 g/cm3. The X-ray powder
ature in argon after an isothermal holding time of 1 h. diffraction pattern indicates only the pres-
 11.4 Si-N Ceramics Based on Poly(organo)silazanes 25

ence of (3-SiC; the expected phase transfor- SiC particles are put under stress which
mation to the high-temperature modifica- can lead to the formation of microcracks
tion, a-SiC, also did not take place under (Raj and Bordia, 1984; Bordia and Raj,
these conditions. The question of the ex- 1988).
tent to which the carbon excess is responsi- At the interface between the nonsinter-
ble for the stabilization of the (3-SiC phase ing SiC and the matrix, radial stresses a r
is the subject of continuing research. and hoop stresses a t occur (Timoshenko
and Goodier, 1970; Bordia and Raj, 1988).
The tangential stress a t generated on the
11.3.5 In Situ Generation of SiC
secondary phase in the matrix can lead to
Dispersions in Si 3 N 4 and B 4 C Composites
the formation of cracks. Low sinter densi-
The production of new materials with ties and poor mechanical properties are the
defined mechanical and physical proper- result.
ties is a great challenge to the materials The reproducible production and reli-
scientist. One approach is to combine the ability of composite materials with well-
properties of several materials in one struc- defined mechanical and physical proper-
ture. Silicon nitride (Si3N4) exhibits high ties depends to a great extent on the homo-
strength and good thermal shock proper- geneous distribution of the dispersed
ties while silicon carbide (SiC) exhibits phase. Composite materials produced us-
greater hardness and better oxidation and ing powder metallurgical methods often
creep resistance. Therefore, it could be ex- exhibit agglomerations of the inclusions in
pected that a combination of these materi- the matrix. For this reason we have studied
als (a Si3N4/SiC composite) should show the in situ generation of SiC particles
improved characteristics over the individu- through the pyrolysis of matrix powder/
al materials. polysilane mixtures with the aim of im-
It is known that Si 3 N 4 particle growth is proving the homogeneity of the dispersed
reduced during sintering in Si3N4/SiC SiC phase in Si 3 N 4 and B4C bulk materials
composites which have been produced us- (Riedel et al., 1989c; Riedel, 1993). The py-
ing conventional powder metallurgical rolysis of alkyl/aryl substituted polysilanes
methods (Greil et al., 1987). These com- or polycarbosilanes leads, depending on
posites also exhibit higher fracture the length of the heat treatment and the
strength (Greil et al., 1987; Sawaguchi final temperature, to very fine crystallites
et al., 1991). Lange (1973) reported an in- of P-SiC with particle sizes in the range of
crease in the fracture toughness with in- 10-100 nm. The homogeneous distribu-
creasing SiC content. tion of the nanocrystalline SiC phase
Sufficient densification of Si 3 N 4 and should make homogeneous densification
SiC powders was previously only achieved and high sinter activity possible.
through hot pressing (Lange, 1973). Pres-
sureless sintering of Si 3 N 4 materials with a
high SiC content is difficult because the 11.4 Si-N Ceramics Based on
Si 3 N 4 and SiC particles exhibit widely dif- Poly(organo)silazanes
ferent sintering behavior at the same tem-
perature (Raj etal., 1984; Bordia etal., (Si6N2)n and Si 3 N 4 are the only binary
1988). Due to the much better densifica- nitrogen compounds known. In contrast
tion of the Si 3 N 4 matrix the nonsintering to Si 3 N 4 , with silicon in the oxidation state
26 11 Advanced Ceramics from Inorganic Polymers

4-4, the structure of the silicon nitride acids which block the free electron pair on
(Si6N2)n exhibits Si-Si bonds resulting in a the nitrogen. Adducts of Lewis acids and
lower oxidation state (+1) for silicon Si-N compounds dissociate in many cases
(Hengge, 1962). Silicon nitride (Si3N4) has below room temperature and are extreme-
been investigated intensely in recent years ly sensitive to hydrolysis because with the
with respect to its application as a synthet- dn-pn interactions now blocked the silicon
ic ceramic material (Boberski et al., 1989). atom is now prone to nucleophilic attack.
Si 3 N 4 exhibits excellent mechanical prop- A comprehensive description of the syn-
erties (high strength even at high tempera- thesis and properties of molecular Si-N
tures, good thermal shock behavior with a compounds is to be found in the literature
low thermal expansion coefficient of 2.9- (Wannagat, 1964).
3.6xlO~ 6 o C, good oxidation resistance The first work describing the synthesis
due to passivation by an SiO2 layer, and its of Si-N polymers was published by
low density (£Si3N4 = 3.2 g/cm3) and as such Chantrell and Popper (1965) and Cheronis
has great potential, especially in motor and (1951). Verbeek (1974) described the syn-
turbine construction. thesis of polysilazanes through the reac-
The structure, technological production, tion of alkylchlorosilanes (CH3)JCSiCl4_JC
and processing of Si 3 N 4 to dense materials with alkylamines such as CH 3 NH 2 . The
are comprehensively reviewed in Vol. 11 reaction mixture containing di- and
(Sec. 3.3) of this Series. trichlorosilanes was subsequently ther-
An alternative method for the fabrica- molyzed at 520 or 650 °C to form resin-like
tion of Si3N4-based ceramics is the pyroly- polysilazanes the exact composition of
sis of polysilazanes. After a short summary which is still a topic of research today.
of the synthetic methods used for the pro- In a process developed by Dow Corning,
duction of polysilazanes, the pyrolysis of hexamethyldisilazane is reacted with
these materials to form Si-N-based ceram- trichlorosilane to produce a highly cross-
ics and the characterization of the materi- linked, chlorine-containing hydridopolysi-
als will be discussed. lazane (HPZ) (Legrow et al., 1987; Lipow-
itz etal., 1986).
On reaction of alkyldichloro- and di-
11.4.1 Poly(organo)silazanes -
alkyldichlorosilanes with ammonia in so-
Synthesis and Properties
lution the main products are found to be
Si-N compounds are usually produced cyclic oligosilazanes with the general for-
through the reaction of =Si-Cl and mula [R 1 R 2 SiNH] n (« = 3, 4). These
= N - H groups. The d n -p n interactions in oligomeric materials cannot however be
Si-N compounds result in short bond pyrolyzed because, due to their low molec-
lengths and higher bond orders as might be ular weight, they evaporate on heating.
expected for single Si-N bonds. The Si The oligosilazanes must therefore be react-
atoms are protected against nucleophilic ed to form products with higher molecular
attack. As a result, these compounds are weights. In the case of the oligomers ob-
relatively stable with respect to hydrolysis. tained on reaction of methyldichlorosilane
Tris(silyl)amines, for example, can be heat- with ammonia [CH3SiHNH]n this is possi-
ed in strong bases for long periods without ble using strong bases such as KH (Sey-
decomposing (Wannagat, 1964). Si-N ferth and Wiseman, 1984), see Formula 4.
compounds react preferentially with Lewis The KH catalyzes the loss of H 2 . It is not
 11.4 Si-N Ceramics Based on Poly(organo)silazanes 27

RH
\/
Si-NH H
NH,
4 RSiHCl, HN Si. /R=CH 3
H I I R
>Si NH
R \ /
HN-Si
/\
Formula 4 RH RH

RR RR RR RR
\/
Si
HN NH l)2n-BuLi / \ I / \ I / \
2)2R 2 SiF 2 R3Si-N N-Si-N N-Si-N N-SiR3 /R=CH 3
.R \ / I \ / | \ /
Si R Si R Si
\N R /\ /\ /\
I R R RR R R
SiR3
Formula 5

known to what extent the oligosilazanes through pyrolysis in ammonia in both the
form Si = N double bonds or whether solid state and the gas phase (Werner et al.,
cross-linking due to cyclopolymerization 1991).
occurs. The company Hoechst has developed
Transition metal catalysts can be used to several methods for the synthesis of new
lead to the ring-opening polymerization of polysilazanes which are mainly based on
cyclic oligosilazanes (Blum et a l , 1989): the reaction of alkyl-substituted chlorosi-
lanes with ammonia followed by cross-
•x[R2SiNH]n RU3(CO)Y [R 2 SiNH], (11-19) linking of the cyclic oligo(alkylhydrido)si-
These catalysts can also be used for lazane with alkyldichlorohydridosilanes
the dehydrocoupling of organosilanes (Gerdau et al., 1989), see Formula 6. Side
(R'R"SiH 2 ) with ammonia (Blum etal., products include alkylchlorosilanes such
1989): as R x SiHCl 3 _ x and R^SiCV,, HC1, H 2
and NH 4 C1, which evaporate or sublime
R 2 SiH 2 (11-20)

30-300°C
[CH3SiHNH]n + CH3SiHCl2
Clegg et al. (1980) report that the reac-
tion of N-trimethylsilylhexamethyltrisila- CH3 CH3 CH,
zane with butyllithium, followed by reac- I
tion with difluorodimethylsilane (Formu- -Si-N- -Si-N- -Si-N-
I I I I
la 5), which results in the loss of LiF, yields H N- Cl
a chain-like molecule made up of three I
four-membered rings joined by silylene -Si-
bridges. These Si-N compounds can be I
converted into Si-N ceramic powders Formula 6
28 11 Advanced Ceramics from Inorganic Polymers

out of the reaction mixture. The investiga- and c also increase. The preferred values
tions on the conversion of polysilazanes to for the mole fraction c are between 0.3 and
Si-containing ceramics discussed in the fol- 0.6 (Gerdau et al., 1989).
lowing section were carried out on labora-
tory products from Hoechst.
11.4.2.2 Characterization and Pyrolysis
of Poly(hydridomethyl)silazane
11.4.2 Synthesis and Characterization
of Si-C-N Ceramics from Polysilazanes Characterization and studies of the py-
rolysis behavior of poly(hydridomethyl)si-
X-ray amorphous Si, C, and N-contain-
lazane [CH 3 SiHNH] 04 [CH 3 SiN] 06 have
ing ceramic powders (silicon carbonitride)
been carried out by mass spectrometry and
can be obtained by the pyrolysis of polysi-
by thermal gravimetric analysis. Fig-
lazanes in nitrogen and argon atmo-
ure 11-14 shows the El mass spectrum of
spheres. Thermal decomposition in ammo-
the polymer compound. Symmetrical sig-
nia on the other hand yields pure Si 3 N 4 .
nal patterns with maxima differing in m/e
Si3N4/SiC composite materials can then
by 57, 58, and 59 can be observed. These
be produced from the carbonitride starting
differences are assigned to the monomeric
powder by liquid phase sintering with
building blocks of the polymer such
simultaneous crystallization (Riedel et al.,
as [CH3SiHNH] with ra/e = 59, and
1989a).
[CH3SiN] with m/e = 51.
The highest detected mass is at around
11.4.2.1 The Poly(organo)silazanes Used
m/e = 1000, and the peak with 100% rela-
The following polymeric starting mate- tive intensity, which is assigned to the
rials are used in the production of Si ce- HSiNH + fragment, is at m/e = 44. The
ramics. peaks with m/e = 73, 499, and 463 can
• Poly(hydridomethyl)silazane: be characterized using high resolution.
[CH 3 SiHNH] 0 . 4 [CH 3 SiN] 0 . 6 m/e = 73 has been assigned to (CH3)3Si + .
• Poly(hydridochloro)silazanes of the The exact masses of the m/e = 449 and 463,
form shown in Formula 7. with relative intensities of 77 and 25 % re-
The substituent R is usually methyl; the spectively, are 449.0672 (calc: 449.06693)
sum of the mole fractions a + b + c is 1. and 463.0415 (calc: 463.04079), and point
The proportions of the mole fractions a, b, toward the presence of C 7 H 29 Si 8 N 8 ,
and c are determined using elemental which is [CH 3 SiHNH] 4 [CH 3 SiN] 4 -15] + ,
analysis and 1 H-NMR spectroscopy. As and C 7 H 29 Si 9 N 7 . On the basis of this
the proportion of alkyldichlorosilane to one can assume that building blocks such
oligosilazane is increased the values of b as that shown in Formula 8 must be

R R' R" RH \
R
R W T-»
\/
\ / «•

I I I Si-NH Si-NH /
-Si-N- -Si-N- -Si-N- / / \/
I I I I I I -N Si-N Si- /R=CH 3 ;
H N- Cl 1 1 1 1 m/e=449
I -Si N-Si N-
-Si- / HN-Si / HN-Si
I
R /\ R /\
Formula 7 RH RH Formula £
 11.4 Si-N Ceramics Based on Poly(organo)silazanes 29

44 450
R H H
?i-*R > N R C 7 H M Si 8 N 8

m/e = 449

CH3SiHNH (59m.u.)
CH3SiN (57 m.u.)

Figure 11-14. Electron ion-

ization mass spectrum of
[CH3SiHNH]0.4[CH3SiN]0.6
at 70 eV; sample tempera-
tures 250-300 °C.
300 500 700
m/e

present in poly(hydridomethyl)silazane, mum loss at 525 °C) but H 2 loss occurs

[CH 3 SiHNH] 0 . 4 [CH 3 SiN] 0 . 6 . over a broad range (maxima at 500 and
Thermogravimetric studies show that 700 °C). Apart from the molecular ion of
the thermal decomposition of the poly(hy- methane at m/e = 16 fragments of this ion
dridomethyl)silazane is fastest at 600 °C in at m/e =15 and 14 are observed with the
a stationary N 2 atmosphere or at 560 °C expected intensities (Kienitz, 1968).
under high vacuum. Between room tem- The presence of the CH^~ ion is also in-
perature and 400 °C two other maxima in dicated in the same temperature range by
weight loss are observed, at 100 and the observation of a peak with m/e = 17
350 °C, which are due to the evaporation (Fig. 11-16). The presence of CH^ ions in
of low molecular weight fractions. Above CH 4 /H 2 mixtures has been confirmed
850 °C very little weight loss is observed. mass spectrometrically by Sefcik et al.
The ceramic yield of pyrolysis at 1000°C
under vacuum is 81.5% and 90% under
N 2 . The mass-loss curve of the sample py- 100 0.2
rolyzed under N 2 is shown in Fig. 11-15. 0.0 CD
o
98 - \ / \ \
The assignment of the masses observed -0.2 -

in the various temperature ranges are col- 1 96 1\ / -0.4 |

lected in Table 11-8. Thermogravimetric 1

S. 94
lCH SiN] [CH SiHNH]
3 m 3 n \ -0.6 g
_
analysis coupled with mass spectrometry -0.8
92
1 A -1.0
allows the qualitative analysis of tempera- v SixCyNz ~

ture dependent emission of pyrolysis gases. i i i i i i i i i i i V i i M">4 i i i r -1.2

Figure 11-16 shows the temperature de- 100 300 500 700 900 1100
Temperature [°C]
pendence of the ion fluxes of various mass- Figure 11-15. TGA of poly(hydridomethyl)silazane
es between m/e = 2 and m/e = 28. Between [CH3SiHNH]0 4[CH3SiN]0 6 under stationary nitro-
400 and 700 °C mostly CH 4 is lost (maxi- gen.
30 11 Advanced Ceramics from Inorganic Polymers

16 / \

8 -
P 2 H2
o

I6-
U-17 Cf-U
17 NH3 Figure 11-16. Temperature
28 N2iC0 dependence of the ion

f/^\ \
CD

St.- current of ions with

c m/e = 2, 14, 15, 16, 17, and
o
/ 28 produced during the
2- /
pyrolysis of poly(hydri-
A__28 ^S I domethyl)silazane
[CH3SiHNH]04[CH3SiNH]0 6
under vacuum at 70 eV.
100 200 300 400 500 600 700 800
Temperature [°CJ

(1974) and Tal'rose and Lyubimova pie form as

(1952). The mje = \l peak can also be as-
[CH 3 SiHNH] 0 . 4 [CH 3 SiN] 0 . 6 (11-21)
signed to 1 3 CH 4 and NH 3 . Methane and 1000°C/N2
hydrogen exhibit the most intense ion flux-
es. All other ions (Table 11-8) have signifi-
cantly lower intensities. The thermal de- 11.4.2.3 Characterization of Silicon
composition of poly(hydridomethyl)sila- Carbonitride
zane can therefore be represented in a sim- The black, X-ray amorphous pyrolysis
product obtained at 1000 °C under N 2 or
Ar is denoted Si^Cy^ (silicon carboni-
Table 11-8. Section of the El mass spectrum (70 eV) tride) in Eq. (11-21). Elemental analysis re-
of poly(hydridomethyl)silazane [CH 3 SiHNH] 04 veals that silicon carbonitride produced
[CH3SiN]0 6 in various temperature ranges. by pyrolyzing poly(hydridomethyl)sila-
Temperature m/e Assignment zane contains 58.9 wt.% Si, 12.6 wt.% C,
range (°C) and 26.8 wt.% N, giving a formula of
Si 2 . 0 N 1 8Ct 0. Oxygen contamination is at
25-200 16-18 H 2 O, NH 3 a level of 0.5 wt.%. If one assumes that all
56 [CH 3 SiN]-H
73 THF + H (solvent) or
nitrogen is bound to silicon in the form of
CH 3 SiH-NHCH 3 Si 3 N 4 and the rest of the silicon is in the
250-325 14-17 NH 3 form of SiC then the formal composition
31 CH 3 NH 2 , CH 2 OH of silicon carbonitride is 67 wt.% Si 3 N 4 ,
(from THF) 27 wt.% SiC, and 4 wt.% C. S i / ^ N ,
270-450 56, 73 [CH3SiN]-H,
CH 3 SiH-NHCH 3 produced from poly(hydridochochloro)si-
350-525 59, 73 CH3SiHNH, lazane has the following composition:
CH 3 SiH-NHCH 3 57.4 wt.% Si, 14.2 wt.% C, and 25.2 wt.%
375 750 12-17 CH 4 , NH 3 N, representing a formula of Si xi7 N t>5 C 1 0 .
200-900 2 H2
Oxygen and chlorine impurities were found
700-900 56 [CH 3 SiN]-H
to be 2.4 and 0.6 wt.% respectively.
 11.4 Si-N Ceramics Based on Poly(organo)silazanes 31

The infrared spectrum of thermally ther amorphous or crystalline inhomoge-

treated (1150°C) poly(hydridomethyl)sila- neous regions could be detected in the
zane exhibits strong vibrational bands as- range of 20 A. The C and N distributions
signed to Si-N at 1250-720 cm" 1 and shown in Fig. 11-17 for an amorphous Si-
Si-C at 600-415 c m ' 1 as well as bands C-N particle show no evidence of C or N
for Si-H (2164 cm" 1 ) and C - H groups enrichment.
(2950-2850 cm" 1 ). Many of the absorp- High-resolution Si(2p) ESCA measure-
tion bands are broadened due to the disor- ments on silicon carbonitride obtained at
dered structure of the pyrolysis product. 850-1850 °C under Ar or N 2 from poly(-
The 29Si solid-state NMR spectrum of hydridomethyl)silazane allow the detec-
poly(hydridomethyl)silazane exhibits a tion of both Si-C and Si-N bonding con-
broad singlet at 8 = — 25 ppm. Annealing tributions. To prevent electrical charging
at 550 °C causes the signal to broaden and the pyrolysis products were milled together
weaken, although the chemical shift of the with copper powder and pressed into pel-
maximum remains constant, and at 850 °C lets prior to the ESCA measurements. The
no further signal can be detected. This be- bond energies were referenced to the ener-
havior has been found on the basis of elec- gy of the 0(1 s) electron in silicon carboni-
tron spin resonance measurements on the tride (531.6 eV). The presence of oxygen
pyrolysis product to be due to paramag- was due to impurities in the polysilazane
netism. Above 1200°C signals reappear and to the handling of the pyrolysis prod-
whose half width decreases with increasing uct in air and the oxygen was present in the
crystallization, and whose 29 Si-NMR form of SiO2 or SiOJCC};Nz (silicon oxy-
chemical shifts are in the range of SiC carbonitride).
[((3)=-16 ppm, (a) = - 2 4 ppm (15R), The O(ls) bond energies of all the pyrol-
— 25 ppm (6H), Dando and Tadayyoni ysis products annealed between 850 and
(1990)], and Si 3 N 4 [(oc)= -46.8 ppm and 1850°C are in the range 531.4 to 531.8 eV
-48.9 ppm, ( p ) = - 4 8 . 7 ppm, Carduner (median, n = 7,is 531.6 eV). The bond en-
etal. (1990)]. ergy of the O(ls) electron in SiO2 varied
Analysis of the pyrolysis product ob- between 532.3 (cristobalite) and 532.7 eV
tained at 1000 °C with an electron micro- (SiO2 gel) (Peuckert and Greil, 1987). The
scope indicates the presence of a homoge- experimental curves could only be simulat-
neous, amorphous solid (Fig. 11-17). Nei- ed by overlapping at least four Gaussian

Figure 11-17. TEM bright-field im-

age of Si 2>0 N 1-8 C li0 obtained from
poly(hydridomethyl)silazane at
1000 °C (left) with elemental distri-
bution analysis for C (center) and
N (right).
32 11 Advanced Ceramics from Inorganic Polymers

1U — i

9
, , , , —-H 1 1 1

N
1 1 1 1—

A>
/
:
8 ioo. A 9 :
N-Si-N /
7

u, 6 :
N /
~5- /101.6/
/I >*—~^ / \ \ '•

z 4-

103.2
//h
/ / /
3-
\ \V •
2
1-
\ \\ •
0^
107.0 106.1 105.2 104.3 103.4 102.5 101.6 100.7 99.8 98.9 98.0
Binding Energy [ e V ]

Figure 11-18. XPS binding energies of the Si(2p) electrons of silicon carbonitride S i ^ N ^ C ^ o annealed at
1850 °C under N 2 . The experimental curve can be modeled by overlaying four Gaussian curves. The binding
energy at 101.6 eV is assigned to Si surrounded in a tetrahedral bonding environment by N atoms, while those
in the 99.5 eV region result from the tetrahedral coordination of Si with the more electropositive C atoms.

curves. Figure 11-18 shows as an example

Si 2p
the Si(2p)-spectrum of Si le7 N 1>5 C 1>0 . The
calculated maxima are at 103.2, 101.6,
100.4, and 99.5 eV. The maximum at
103.2 eV can be assigned to the presence of
SiO4 tetrahedra and therefore to oxygen
impurities. The bond energy at 101.6 eV
indicates the presence of SiN4 or
SiN 4 _ x O :c tetrahedra, that at 100.4 eV to
the presence of SiN4_JCCJC tetrahedra, and
that at 99.5 eV the presence of SiC4 tetra-
hedra. The Si(2p) bond energies of pure
Si 3 N 4 and SiC are between 101.1 and
101.7 eV and at 99.4 eV, respectively. The
mixed valence SiN 4 _ x C x exhibited the
highest intensity over ca. 50 % of the peak
surface, an observation which agrees with
the X-ray amorphous structure of the ma-
terial.
The high-resolution C(ls) spectrum of
samples pyrolyzed at over 1000 °C shows 112.0
110.5
109.0
107.5
106.0
104.5
103.0
101.5
100.0
98.5
97 0

the presence of C-C (284.2 eV), C - N Binding Energy [ e V ]

(286.5 eV), and C-Si bonds (282.7 eV). The
high-resolution N(ls) spectrum reveals Figure 11-19. XPS binding energies of the Si(2p) elec-
N-Si (397.4 eV) and N - C (398.6 eV) trons (ESCA spectrum) of Si1#7N1>5C1<0 synthesized
from poly(hydridochloro)silazane as a function of the
bonding contributions. The values were pyrolysis temperature and atmosphere. The vertical
referenced to Si-O in silicon carbonitride lines indicate the chemical shifts of the Si(2p) elec-
and are collected in Table 11-9. trons in pure Si 3 N 4 and SiC.
 11.4 Si-N Ceramics Based on Poly(organo)silazanes 33

Table 11-9. XPS bond energies of the Si(2p), N(l s) and C(l s) electrons of the pyrolysis product (Si 1/7 N 15 C 1 0)
obtained at various temperatures from poly(hydridochloro)silazane. The values given correspond to the peak
maxima. The literature values for the XPS bond energies in Si3N4 and SiC are included for comparison.

Compound T(°C) XPS binding energy3 (eV)

Si(2p) N(ls) C(ls)

Si3N4 (amorph.)b 1300 101.8 397.6

Si3N4(cryst.)c 101.1-101 .7 397.1-397.5
SiCd 99.4 282.5

Si3N4 (amorph.)e 1000 102.5 398.4 284.6

850 101.3 284.4
1000 101.8 397.8 284.2
1150 101.3 397.5 284.1
Si,CyNz 1850 100.1 396.0 283.3
a
After 15-30 min sputtering of the sample surface with Ar+ ions. b Goto and Hirai (1988). c Homeny et al.
(1990); Peuckert and Greil (1987). d A10 powder from H. C. Starck, copper compact. e The pyrolysis of
poly(hydridochloro)silazane under NH 3 at 1000 °C yields amorphous Si 3 N 4 . Experimental values for this sample
are referenced to the C(l s) electron binding energy (284.6 eV). The value for O(l s) is 532.5 eV and agrees well
with literature values for SiO2 or Si 2 N 2 O (Peuckert and Greil, 1987).

Figure 11-19 shows the change in XPS placed by N-Si bonds, and C-H bonds are
bond energies with temperature and pyrol- replaced by C-Si bonds.
ysis atmosphere. In the temperature range
Spectroscopic studies, electron micros-
1000-1850 °C the bond-energy maximum
copy, and elemental analysis of silicon car-
for Si(2p) electrons moves from 101.8 to
bonitride synthesized from poly(hydri-
101.1 eV, that of the N(ls) from 397.8 to
domethyl)- and poly(hydridochloro)sila-
396.0 eV, and that of the C(ls) electron
zanes have shown that these are new
from 284.2 to 283.3 eV. Despite the high
metastable materials in which the silicon
annealing temperature (1850°C) no crys-
atoms are tetrahedrally coordinated by
tallization can be observed within one
both carbon and nitrogen. This is also
hour, the material remaining X-ray amor-
demonstrated by the presence of mixed va-
phous. The low bonding energy (283.3 eV)
lence components such as NSi 3 _ x C :c and
of the C(ls) electron in the pyrolysis prod-
uct annealed at 1850 °C hints at a high con-
tent of Si-C bonds. The decrease of the The silicon carbonitrides produced are
Si(2p), C(ls) and N(ls) bonding energies non-oxide glasses which can be represent-
can only be explained by the presence of ed by the general formula 813+^4(^+3,.
electropositive bonding partners in the ce- From this the formal stoichiometry which
ramic solid. Si-H and Si-N bonds in the is expected on crystallization of the materi-
polymer starting material and in the pyrol- al Si 3 N 4 xSiC-jC can be deduced. Ac-
ysis product respectively are increasingly cordingly, the silicon carbonitrides derived
replaced by Si-C bonds as the temperature from different precursors given below are
and the conversion to ceramic material in- represented by the following molar com-
creases. N - H and N - C bonds are re- positions:
34 11 Advanced Ceramics from Inorganic Polymers

• From poly(hydridomethyl)silazane powders cannot be sintered to high densi-

[CH 3 SiHNH] 0 . 4 [CH 3 SiN] 0 . 6 : ties ( > 9 0 % relative density) without the
use of additives. In the case of SiC the
= Si 3 N 4 +1.40 SiC + 0.80 C addition of elemental boron or aluminum
• From poly(hydridomethyl)silazane or their compounds, such as BN, B4C,
[CH3SiHNH]m[(CH3)2SiNH]n: A1N, or A14C3, leads to a reduction of the
grain boundary energy and therefore a
- S i 3 N 4-h 1.40SiC + 1.2C change in the ratio of the grain boundary
From poly(hydridochloro)silazanes: to surface energy (Prochazka, 1975). For
further details of the conventional fabrica-
= Si 3 N 4 + 1.50SiC + l . l C tion of SiC and Si 3 N 4 parts see Vol. 11 of
• From poly(methylvinyl)silylhydrazine this Series (Chaps. 3, 4, and 10).
[-(CH2 = CH)Si(CH3)-NH-NH-]n: A homogeneous distribution of addi-
Sii.00N1.03C1.36 = Si 3 . 9 N 4 C 5 3
tives is essential in order to sinter the pow-
= Si 3 N 4 + 0.9SiC + 4.4C. ders to high-density bulk materials and can
be achieved using milling or precipitation
processes. Oxide-based sintering additives
11.5 Production of Non-Oxide form glass-like secondary phases, which,
Si-Based Ceramic Parts due to their relatively low softening tem-
peratures, can be detrimental to the high-
11.5.1 Conventional Production of SiC temperature properties of the material.
and Si 3 N 4 Parts Therefore, work aimed at the production
of non-oxide Si-ceramic materials without
SiC and Si 3 N 4 powders are usually den- the addition of sintering aids is particularly
sified to form ceramic parts at very high important.
temperatures and both with and without
the use of pressure (hot pressing and hot
11.5.2 Production of Si-Based Ceramic
isostatic pressing). The sintering tempera-
Parts from Polymeric Compacts
tures for SiC are in the range 2000-
2200 °C (Hunold, 1989) and for Si 3 N 4 The direct pyrolysis of large-volume
1600-1950°C (Ziegler et al., 1987). Due to polymeric green bodies to ceramic parts is
their extremely small self-diffusion con- very difficult. The reason is the release of
stants (see Table 11-10) SiC and Si 3 N 4 gaseous low-molecular-weight reaction

Table 11-10. Self-diffusion constants for C, N, and Si in SiC and Si 3 N 4 .

Element Material Temperature Diffusion constant References

(°Q (cn^s" 1 )

C fi-SiC (polycrystalline) 2000 10" 11 Hon and Davis (1979)

Si ^-SiC (polycrystalline) 2000 10~ 13 Hon and Davis (1980)
C a-SiC (single crystal) 2000 3X10" 1 1 Hong and Davis (1980)
Si a-SiC (single crystal) 2000 5xlO~ 1 4 Hongetal. (1981)
10 -i9
N a-Si3N4 1400 Ziegler et al. (1987)
N £-Si3N4 1400 8xlO-18 Ziegler et al. (1987)
Si Si 3 N 4 1400 10" 13 Ziegler et al. (1987)
 11.5 Production of Non-Oxide Si-Based Ceramic Parts 35

products which are formed during the 11.5.2.1 Crack-Free, Dense Ceramic
thermal decomposition and which lead in Materials from Organoelement Polymers
many cases to foaming and the occurrence
of cracks. The following requirements have to be
The AFCOP process (active filler con- met in order to be able to convert orga-
trolled pyrolysis) developed by Greil et al. noelement polymers into crack-free ceram-
(Erny et al., 1993), which avoids this prob- ic materials:
lem, involves the processing of mixtures • To prevent foaming and cracking the
of active metals (Ti, CrSi, etc.) and materials should not melt during the ther-
polysiloxanes (poly(organyl)silsesquioxane mal decomposition
[RSiOi 5]n with R = alkyl and/or vinyl). • The polymer green bodies should have
The polymer is cross-linked in a thermal or an open, porous structure to allow the re-
radical process. The use of polymers with moval of the gaseous reaction products
intact vinyl groups (CH = CH 2 ) leads to • The maximum pore size should be small
cross-linking through polyaddition with- enough to allow the pores to be closed at
out gas evolution. In contrast, the cross- the end of the pyrolysis, i.e. during the
linking of polymers through polyconden- subsequent sintering process at tempera-
sation leads to the formation of low- tures above that used for the decomposi-
molecular-weight products such as H 2 O tion.
and NH 3 which lead to the foaming and Basically, three methods can be consid-
cracking mentioned above. ered to produce open, porous structures:
Subsequent pyrolysis of the polymer/ (a) The generation of porosity through the
metal mixture results in crack-free dense cold or hot pressing of infusible polymer
materials (Erny et al., 1993). Gaseous py- powder. In this case the size of the polymer
rolysis products such as hydrocarbons, particles, the pressure, and the formability
and condensed carbon react directly with of the polymer determine the pore size and
the metal, forming carbides and silicides, the pore distribution in the polymer green
thus avoiding the formation of large body.
gas volumes. Due to the polysiloxane con- (b) Generation of the porosity via a gel
tent, large amounts of oxygen-containing process. Three-dimensional networks ex-
phases, such as SiO2 and metal oxides, are hibiting microporosity can be formed as a
formed. The hydrogen formed during the gel from solution. The porosity can be in-
thermolysis must be able to diffuse out of fluenced by additives, polymerization time
the part. or by the solvent. This method is anal-
The formation of gas-phase reaction ogous to the sol-gel process used for the
products is not so damaging in fiber pro- production of monolithic oxide glasses
duction from organic polymers as the low and ceramics.
fiber diameter (10-100 jum) allows the dif- (c) Generation of the porosity in situ dur-
fusion of the gases through the solid there- ing the pyrolysis. Gases generated during
fore avoiding the formation of cracks and the pyrolysis, preferentially of liquid poly-
bubbles on thermolysis. mers, can be used to produce the porosity
during the early stages of conversion to the
ceramic solid. Here, an extremely slow and
controllable gas emission is required in or-
der to avoid the formation of cracks.
36 11 Advanced Ceramics from Inorganic Polymers

All three processes benefit from the gas

[CH3SiHNH]m[(CH3)2SiNH]n
evolution occurring over a broad tempera-
ture range. However, the route considered POLYSILAZANE
in (a) is the most successful. It has been
demonstrated recently (Riedel et al., 1992)
Ar, 350°C Crosslinking
and will be discussed in more detail in the
following sections. Accordingly, using
method (a) polymer green bodies have
been pyrolyzed to dense crack-free ceram- INFUSIBLE POLYSILAZANE
ics. The process has the following special
features:
• Avoidance of the high-temperature syn- Milling Shaping
thesis of ceramic powders
• Low temperatures (800-1200 °C) are
sufficient for the production of dense ce-
ramic parts POLYSILAZANE

• Sintering additives are not necessary for GREEN COMPACT

densification
• The production of amorphous and crys-
talline materials is possible. The relatively Ar, 1000°C Pyrolysis

low process temperature (ca. 1000 °C) re-

sults initially in amorphous materials
which can be subsequently crystallized by
CRACK-FREE MONOLITHIC
annealing. This method also can be used to
synthesize nanocrystalline materials, SILICON CARBONITRIDE
which are, due to their unusual properties
such as superplasticity, of great interest
but difficult to produce by conventional Figure 11-20. Process used for the production of
processes. monolithic silicon carbonitride ceramics from
A basic disadvantage of this process is poly(hydridomethyl)silazane (Riedel et al., 1992).
the high degree of shrinkage (in silazanes a
28 % linear or a 60 % by volume shrinkage
has been observed) (Riedel et al., 1992).
of a starting material, which according to
IUPAC rules should be known as catena-
11.5.2.2 Si-Based Ceramic Parts
poly[methylsilicon-(i-amino) (dimethylsili-
from Poly(organo)silazanes
con-|i-amino)] (Donaruma etal., 1981).
The flow diagram shown in Fig. 11-20 The structural formula is purely empirical
is a schematic representation of the pro- and reflects the molar composition deter-
cess used for the production of near mined using spectroscopic and analytical
dense monolithic silicon carbonitride techniques.
(SixCyNz) parts from highly cross-linked Poly(hydridomethyl)silazane is synthesized
poly(organo)silazanes. [CH3SiHNH]m- by the reaction of dichloromethyl- and
[(CH3)2SiNH]n, a commerically available dichlorodimethylsilane with ammonia fol-
poly(hydridomethyl)silazane, is an example lowed by cross-linking using a base (e.g.
 11.5 Production of Non-Oxide Si-Based Ceramic Parts 37

KH, Seyferth and Wiseman, 1984), accord-

ing to
CH3SiHCl2 + (CH3)2SiCl2 (1)NH3'(2)KH/CH3l>
[CH 3 SiHNH]J(CH 3 ) 2 SiNH] n (11-22)
Table 11-11 contains the properties of
the poly(hydridomethyl)silazane used. In
the following section the production of Si-
based ceramic parts from polymeric start-
ing materials is examined in more detail.
Figure 11-21. SEM image of poly(hydridomethyl)sila-
Poly ( hydridomethyl) silazane zane [CH3SiHNH]J(CH3)2SiNH]n annealed at
350 °C for 3 h under argon.
For the production of dense monolithic
Si-C-N ceramics the polysilazane shown in
During annealing, low-molecular-weight
Fig. 11-20 is first annealed at 350 °C. Poly-
components distill off, increasing the total
condensation occurs, resulting in the loss
weight loss to 12%. The scanning electron
of H 2 and NH 3 (which can be detected
microscopy images in Fig. 11-21 show that
using mass spectrometry) and in the for-
the cross-linked material produced in this
mation of N(SiR 2 -) 3 units
way is slightly foamed. Cross-linking re-
350 °C
[RSiHNH]n [RSiHNH]m_n[RSiN]M sults in a decrease in the solubility of the Si
(11-23) polymer in organic solvents and in infus-
ibility. Infusibility is an important require-
The viscosity of the polymer increases ment in order that the green parts retain
significantly above 250 °C due to increas- their shape during pyrolysis. As a result,
ing cross-linking. When the final tempera- organoelement polymers, which are easily
ture is reached (350 °C) at a heating rate of thermally or chemically cross-linked, are
5°C/min all of the liquid has solidified. of great interest.
The annealed polymer is subsequently
Table 11-11. Selected physical properties of a com- ground to a powder which is then uniaxial-
mercial poly(hydridomethyl)silazane. ly or cold-isostatically pressed into poly-
meric green bodies. The particle size of a
Property Poly(hydridomethyl)silazanea
polymer powder ground for 1 h in a ball
Composition [ - RSiH - NH - ]m mill with ZrO 2 milling balls is shown in
•[-R2Si-NH-]n; Fig. 11-22. The distribution is bimodal and
R = CH 3 shows particle sizes in the range < 0 . 1 -
Consistency colorless powder 60 \xm with maxima at 0.35 and 10.5 jim.
Softening temperature ca. 120 °C
Average molecular
After cold isostatic pressing at 640 MPa
weight 1100-1300 the relative density of the densified poly-
Solvent toluene, THF mer is calculated to be 84-89% based on
Oxygen content (wt.%) 0.2 (1100°C, N2) a solid density of 1.22 g/cm3. The high
1.1 (1100°C, Ar) green density of the cross-linked and com-
Ceramic yield (wt.%) 69 (1000 °C, N2)
74(1000 °C, Ar) pacted polymer can be explained on the
basis of plastic deformation of the polysi-
a
Polysilazane NCP 200, Chisso Corp., Tokyo, Japan. lazane during pressing. Figure 11-23 shows
38 11 Advanced Ceramics from Inorganic Polymers

further weight loss being detected in argon

up to 1000 °C. The black pyrolysis product
contains 57.8 wt.% Si, 14.3 wt.% C and
26.0 wt.% N, which indicates a molar
composition of Si2 7N1 6CX 0 or Si 3 N 4 -1.4
SiCl.lC.
If the poly(hydridomethyl)silazane is
pyrolyzed under nitrogen instead of argon
the silicon carbonitride has a significantly
I higher C content as well as 7.4 wt.% hy-
100 200
Particle Diameter [ (I m ]
drogen, which results in a formula of
Sii.oNi.iCi. 4 H 4 7. The carbon content of
Figure 11-22. Laser-light scattering histogram of the
particle size distribution of poly(hydridomethyl)sila- the carbonitride can be widely varied by
zane powder which has been thermally cross-linked adjusting the pyrolysis atmosphere. On the
and milled in a ball mill with ZrO 2 milling balls. basis of the rule of mixtures, densities of
the pyrolysis products obtained in argon
were found to be 2.66 g/cm3 for the purely
amorphous phases (Lipowitz etal., 1987)
and 2.90 g/cm3 for the crystalline phases
(Lipowitz et al., 1987). The absolute solid
phase density of the ceramic product
synthesized at 1000 °C under argon is
2.3 g/cm3, a much lower value. This dis-
crepancy is due to the residual hydrogen
content of the pyrolysis product (even af-
ter processing at 1150°C) which can be
detected in the form of C-H and Si-H vi-
Figure 11-23. SEM image of a fracture surface of
poly(hydridomethyl)silazane powder which has been
brational bands in the FT-IR spectrum at
cold isostatically pressed at 640 MPa (Riedel et al., 2850-2950 cm" 1 and 2100-2200 cm" 1 ,
1992). respectively, and which results in the ex-
tension of the amorphous network. Ele-
mental analysis shows the hydrogen con-
a scanning electron microscopy image of a tent of the pyrolysis product to be
fracture surface of a polymer green body. 0.2 wt.%.
The relatively dense packing of the parti- Microstructural analysis using pho-
cles and the extremely flat particle contacts toelectron spectroscopy (ESCA) and ana-
can be seen. lytical transmission electron microscopy
In the last step of the process the ther- (TEM) have shown (Riedel etal., 1995)
mal decomposition of the polymer com- that the X-ray amorphous Si-C-N materials
pact takes place at temperatures above obtained from poly(hydridomethyl)silazane
500 °C. Pyrolysis under argon results in are single-phase silicon carbonitride in
gaseous products such as hydrogen and which Si is statistically surrounded by C
the hydrocarbons CH 4 , C 2 H 6 , C 2 H 4 , and and N (see also Sec. 11.4.2.3).
C 2 H 2 . Thermal analysis indicates that the Due to the open porosity of the polymer
decomposition is complete at 900 °C, no green body (11-16%) gaseous reaction
 11.5 Production of Non-Oxide Si-Based Ceramic Parts 39

products can be removed easily. The poly-

mer particles are converted to a nonmetal-
lic, inorganic solid on pyrolysis. The green
body shrinks linearly by 20-28% during
this process, depending on the initial resid-
ual porosity. Ceramic parts produced in
this way are crack free and exhibit a rela-
tive density of 92-94%. The pyrolytically
formed materials are wear resistant and of
such hardness that they cannot be
Figure 11-24. SEM image of the microstructure of
scratched with steel. The good mechanical Sii.7N1-6C1-0 (relative density £ = 93%) produced
strength results from the formation of new from [CH3SiHNH]IB[(CH3)2SiNH]ll at 1000°C.
covalent bonds which are generated during
surface reactions between the individual
polymer particles during pyrolysis. Fig-
ure 11-24 shows an SEM image of the con-
tinuous three-dimensional SixCyNz net-
work obtained. Dense regions broken up
by pores with a maximum diameter of
10 |im can be seen and, surprisingly, no
cracks can be found. Higher magnification
in the SEM allows the identification of
pore-free dense particles which are bound
to each other in the absence of secondary
Figure 11-25. As Figure 11-24 but with a higher mag-
phases. Evidence for the intermediate
nification.
melting of the polymer particles during the
heating process cannot be found (Fig. 11-
25). /
a) -N b) -N
Figure 11-26 shows the effect of heat \ ^ \
-Si-CH 3
treatment during cross-linking on the K /
/ H -N
bonding characteristics of the surface of N-H \
the polymer particles. After a short anneal- Si -Si-H
/< H I
ing period, reactive Si-H and N - H groups -N -N
are still present which can take part in in- \ \
ter- and intraparticle cross-linking during Figure 11-26. Possible surface structures of poly(hy-
subsequent pyrolysis (Fig. 11-26 a). With dridomethyl)silazane particles after different anneal-
longer periods of annealing the reactive ing times in the temperature range 100-400°C. (a)
After a short annealing time, (b) After a longer anneal-
groups have partially taken part in intra- ing time.
particle polycondensation reactions
(Fig. 11-26b). Accordingly, the surface re-
activity of the polymer particles decreases > RSi-HNH + HNH-SiR <
with increasing time of thermal cross-link-
>RSi-NH-SiR<+NH3 (11-24)
ing mainly via condensation reactions in-
volving the loss of H 2 [Eq. (11-23)] or NH 3 The reaction of low-molecular-weight
[Eq. (11-24)]. species with the polymer particles can also
40 11 Advanced Ceramics from Inorganic Polymers

be considered to lead to the bonding of the ics the maximum pore radius observed was
particles. On heating of the polymer green 110 nm, the larger pores seen in Fig. 11-24
body, low-molecular-weight monomeric not being detected by the mercury-pres-
or oligomeric silazanes ([CH3Si(H)NH]n) sure porosimetry. From this it can be con-
can form, which condense on the surface cluded that, due to flaws generated during
of the polymer particles and react there pressing of the polymer powder, the matrix
with Si-H and N - H groups. Low-molecu- contains isolated pores which are joined
lar-weight components are formed in de- only by very narrow channels or are not
polymerization processes during heating. joined with each other at all. Figure 11-27
In the contact regions between the parti- also shows that the pore-size distribution
cles covalent bonds can be formed. of the polysilazane green bodies is broader
Another explanation for the observed than that observed for the silicon carboni-
densification during the thermal decompo- tride samples. All pores with a radius
sition of polymer powder compacts is the < 15 nm were closed during pyrolysis. The
formation of intermediate liquid phases, cumulative and mass-unit related pore vol-
which on decomposition can also lead to a ume is smaller, indicating that during the
coalescence of the polymer particles. thermal decomposition volume shrinkage
Using the above technique polysilazane takes place, due to the loss of gaseous
green bodies can be pyrolyzed to crack- products, in addition to densification.
free S i 1 7 N 1 6 C 1 0 ceramic parts. Mercury- Some mechanical properties of the dense
pressure porosimetry of silicon carboni- Si-C-N materials produced under argon at
tride bulk materials synthesized at 1000°C 1100°C are summarized in Table 11-12.
reveals an absolute density of 2.15 g/cm3, The fracture surface of a dense silicon
or 93% of the solid-phase density of 2.3 carbonitride sample is shown in the SEM
g/cm3. image of Fig. 11-28. The relatively low
The pore-size distribution is shown in porosity and the flat fracture surface typi-
Fig. 11-27. In the pyrolytic Si-C-N ceram- cal of amorphous materials can clearly be

200

100 Figure 11-27. Pore-size dis-

tribution of infusible, pow-
dered poly(hydridomethyl)-
silazane: (a) isostatically
pressed at 640 MPa and (b)

n~hhLjP subsequently pyrolyzed at

1000°C under argon
h H ^ (hatched sections, Riedel
et al, 1992).
r 10 LLJI 100 1000
Pore Radius [nm]
 11.5 Production of Non-Oxide Si-Based Ceramic Parts 41

1000 °C under flowing ammonia then com-

pletely colorless Si 3 N 4 is obtained. The mi-
crostructure exhibits dense regions up to
80 jum in size. Thermolysis results also in
this case in the reaction of the original
polymer particles and the formation of a
stable amorphous Si 3 N 4 matrix.
The crystallization of amorphous Si 3 N 4
to oc-Si3N4 starts above 1000 °C. In the
presence of carbon (i.e. in silicon carboni-
Figure 11-28. SEM image of a fracture surface of tride) the onset of crystallization is shifted
Si 1-7 N 1#6 C lt0 . The sample was produced by the py- to higher temperatures. In this way
rolysis of poly(hydridomethyl)silazane compacts at
1000°C under argon.
Sii.vNi 6 C 1 - 0 produced from poly(hydri-
domethyl)silazane forms a-SiC and oc-
Si 3 N 4 at T> 1400 °C under nitrogen or un-
Table 11-12. Average values of the four-point bending der argon.
fracture strength oB at room temperature, the Vickers The high thermal stability of the Si-C-N
hardness, and the elasticity modulus of silicon car- bulk material is documented by the low
bonitride compacts produced from poly(hydri- relative mass changes and the low linear
domethyl)silazane[CH3SiHNH]m[(CH3)2SiNH]n. shrinkages of the samples in the tempera-
Heat Open Vickers E-modulus ture range between 1000 and 1400 °C and
treatment porosity hardness is shown in Fig. 11-29. The higher weight
(%) (MPa) (GPa) (GPa) loss observed at r>1400°C for samples
pyrolyzed in argon can be explained on a
1100°C
in Ar 6 172a 9.8 108 thermodynamic basis in analogy to the be-
1400°C havior of Si 3 N 4 , which reacts in the pres-
inN 2 7 176 12 122 ence of free carbon above 1440 °C at
185O°C 0.1 MPa N 2 partial pressure to form SiC
_ b _ b
inN 2 24 6 and elemental N 2 (Nickel et al., 1988)
a
Average over 5 samples, ignoring the maximum •3SiC + 2N 2 (11-25)
value of 375 MPa; b not measured. Si3N4 + 3C-
The driving force behind this reaction is
the gain in entropy and the high formation
seen. The Vickers hardness HV is 9.8 GPa, energy of elemental nitrogen.
which is a low value compared to that of Silicon carbonitride reacts in an entirely
pure, crystalline SiC (27-30 GPa) and analogous fashion to form SiC with loss of
Si 3 N 4 (16-22 GPa) although it is com- N2:
parable to that of reaction-bonded Si 3 N 4
with a similar porosity (VH = 9 GPa, Rit- ^ 1.73^1.0^ 1.56 >

ter etal., 1988). Table 11-12 also shows 0.24Si 3 N 4 +lSiC + 0.3N 2 (11-26)
that the subsequent annealing of the sam-
ples at 1400 °C under nitrogen results in an In support of this interpretation is the
increase of these values. observation that a-SiC crystallizes out un-
If the compacted polymer powder com- der argon at 1600°C. Under an N 2 atmo-
posed of polysilazane is pyrolyzed at sphere, in contrast, the reaction shown in
42 11 Advanced Ceramics from Inorganic Polymers

The oc-Si3N4- and a-SiC-containing ma-

terial, crystallized from silicon carboni-
tride at 1850°C under 0.1 MPa, N 2 , ex-
hibits, despite its high porosity, excellent
mechanical stability and hardness. This
composite cannot be scratched by glass.
This process is all the more remarkable as
it makes possible the production of well-
defined Si3N4/SiC composite materials
with pore sizes in the nanometer range.
Such materials have potential applications
1700 as microfiltration membranes. Average
crystallite sizes in the range of 10 nm are
derived from the X-ray diffraction pattern
using the Scherrer equation [Eq. (11-18)].
Silicon carbonitride bulk materials are
highly oxidation resistant. Annealing the
material (6% open porosity) for 24 h at
1400 and 1500°C in air results in weight
increases of 0.3 and 0.6%, respectively.
The majority of the weight increase occurs
during the first hour and at times longer
than 1 h the oxidation rate falls to less than
0.1 % h" *. Due to the open porosity of the
samples no time constants for the oxida-
900 1700 tion reaction can be calculated as the inner
surface of the samples is not known and
Figure 11-29. Relative change in weight (m/m0) and even changes during oxidation.
length {L/Lo) of silicon carbonitride parts in the tem- The crystallization behavior of silicon
perature range 1000-1600 °C (Riedel et al., 1992). carbonitride annealed in air is quite differ-
ent to that found under nitrogen or argon.
Eq. (11-26) is shifted to higher tempera- Even after 70 h at 1500 °C in air no crystal-
tures and is slower than that under argon lization takes place, the process starting
(Fig. 11-29). The samples change color on only above 1600 °C. After 60 h oxidation
annealing from black to gray. at 1600°C a weight increase of 0.4%, due
The weight loss measured (ca. 8 %) after to the surface passivation with SiO2, can
completion of the carburization reaction be detected and X-ray diffraction shows
agrees well with the calculated value the formation of oc-SiC and oc-Si3N4. As no
(10%). The amorphous-crystalline con- weight loss can be detected, crystallization
version above 1440 °C takes place without of the silicon carbonitride matrix must
damage to the densified bulk material. The proceed with the segregation of elemental
evolution of nitrogen is, however, associat- carbon
ed with a large increase in the porosity
Si 1 . 7 N 1 . 6 C 1 . 0 ^Si 4 . 4 3 N 4 C 2 . 5 6 (H-27)
(Table 11-12). At the same time the aver-
r 1600OC/air
age pore size increases from 75 to 100 nm. - )0t-Si3N4 +1.43 oc-SiC + 1.13 C
 11.5 Production of Non-Oxide Si-Based Ceramic Parts 43

Crystallite sizes of 50 and 25 nm can be

determined from the X-ray diffraction
pattern using Eq. (11-18) for a-Si 3 N 4 and
oc-SiC, respectively.

11.5.2.3 Si-Based Ceramic Parts from

Polycarbosilanes
For the production of Si-based ceramic
components from polycarbosilanes, spe-
cial infusible Si-C polymers can be synthe- Figure 11-30. SEM image of polycarbosilane (PCS)
sized which can be processed in an powder particles formed by the reaction of dichloro-
analogous way to the Si^C^N^ bulk ceram- methylvinyl- and dichlorodimethylsilane with Na/K
ics. A typical synthetic route, which in this alloy in boiling THF.
case leads to Si polymers used in the pro-
duction of monolithic SiC-based materials ed in the physical properties of the poly-
will be detailed here. mer, which is insoluble in organic solvents
Dichloromethylvinyl- and dichlorodi- and does not melt on heating.
methylsilane are reacted with Na/K al- PCS is obtained in 94 % yield as a color-
loy in boiling tetrahydrofuran (THF) to less, fine powder, obviating the need for
form a colorless poly(organyl)silane. This milling before further processing. The PCS
polymer contains [CH2 = CHSiCH3] and powder is composed of particles with a di-
[(CH3)2Si] units which are also cross- ameter of < 1 jim-80 jam (Fig. 11-30)
linked via the vinyl group. As a result, a which can be uniaxially or isostatically
polymer with an Si-Si and Si-C-C-Si pressed to form polymer green samples.
framework is formed, which in the follow- The infusibility of the material guarantees
ing will be called polycarbosilane (PCS). that the compact retains its form during
Through the addition of chlorotrimethylsi- thermal decomposition. The pyrolysis in-
lane to the reaction mixture the ends of volves heating at 1 °C/min up to 1100 °C in
polymer chains can be substituted with argon. A black, crack-free and scratch-re-
trimethylsilyl groups (Formula 9). sistant SiC part is obtained which exhibits
Under these reaction conditions (which a greater hardness than glass. Figure 11-31
are analogous to the Wurtz reaction) the shows that polymeric green bodies with in-
formation of Si-Si bonds is accompanied tricate or complex structures such as cylin-
by reaction of the vinyl group. The result- ders or threads can be produced by this
ing large amount of cross-linking is reflect- process.

Cl Cl CH3
I I Na/K
CH 2 =CH-Si-CH 3 -Si- 2CH3-Si-CH3
I I
Cl Cl Cl

CH, CH3
I I
H3C-Si- -CH2-CH-Si- -Si- -Si-CH3
I
Formula 9 CH3 H3C CH 3 J CH3
44 11 Advanced Ceramics from Inorganic Polymers

11.5.2.4 Densification
The sintering mechanism by which com-
pacted polymer powders give dense ceram-
ics has yet to be fully explained. Several
points are important for the discussion of
the mechanism, which can be seen in three
variations:
• Shrinkage of the polymer particles with-
Figure 11-31. SiC1-8 parts produced by the thermal
decomposition of compacted PCS powders at 1100°C out change in porosity or with an increase
under argon. of the porosity and without the formation
of interparticle contacts.
• Shrinkage of the polymer particles with-
out change in porosity or with an increase
of the porosity and with the formation of
The pyrolytic material is X-ray amor-
interparticle contacts.
phous and exhibits a Vickers hardness of
• Shrinkage of the polymer particles with
18.5 GPa. For comparison the Vickers
a reduction in porosity and with the forma-
hardness of pure, polycrystalline sintered
tion of interparticle contacts.
SiC is 25-30 GPa. The lower value of
18.5 GPa results from the relatively high Apart from the volume decrease caused
porosity (30%) of the pyrolysis product by the evaporation of low-molecular-
and that its stoichiometry is SiCx 8 , not weight products and the phase transforma-
SiC. Recent studies by Soraru et al. (1990) tion (to ceramic), the third option involves
have shown that an Si ceramic produced an additional densification. An example of
from a commercially available polycar- this is the Si-C-N bulk materials produced
bosilane is not a mixture of silicon carbide from polysilazane.
and carbon, but must be seen as SiC1>4_ 1>6 The shrinkage during the thermal de-
Solid-state NMR measurements reveal composition of polymer (P) to ceramic (C)
the presence of SiC4 tetrahedra and struc- and to gaseous reaction products (R) as in
tural elements such as Si-C-C = C-Si and
C-H in the X-ray amorphous material ob- R (g) (11-29)
tained at 840 °C, and the presence of Si-Si
is dependent on the density and mass
groups can be excluded. The strongly dis-
changes and the ceramic yield (a). The in-
ordered amorphous structure, which is al-
dices s, 1, and g represent the aggregation
so observed in chemical-vapor deposited
state, solid, liquid, and gas. The ceramic
SiC1>5 (Liedike, 1987), explains the lower yield is the relationship between the mass
hardness value of SiC1>8 compared to pure of the ceramic pyrolysis product (Mc) to
SiC. Equation (11-28) also shows that the the mass of the polymer starting material
hardness decreases with increasing porosi-
ty P (Kollenberg, 1991). Figure 11-32 pre-
sents a summary of the method for the <ii 3o)
production of "SiC" bulk materials from '-if. -
PCS.
The change in volume (AF) can be ex-
HV = (11-28) pressed in terms of the mass (M) and the
 11.5 Production of Non-Oxide Si-Based Ceramic Parts 45

x CH2=CH-Si-CH3 + yCH 3 -Si-CH 3 + 2 CH 3 -Si-CH 3

Na/K, THF, 68°C Wurtz-Fittig analogous reaction

H 3 C-Si- [-CH 2 -CH-Si- }x • [-Si- ] y -Si-CH 3

Shaping

POLYCARBOSILANE
GREEN COMPACT

Ar, 1100°C Pyrolysis

Figure 11-32. Schematic representation of

CRACK-FREE the polymer process for the production
MONOLITHIC of monolithic SiC1 + x bulk materials
through the direct pyrolysis of infusible
polycarbosilane green bodies (PCS: poly-
carbosilane).

density (D) of the phases involved: or in percentages

100 (11-34)

For the thermal decomposition of

(11-31) the silazane to form Si1>7N1>6C1>0,
Dp = 1.22 g/cm3, Dc = 2.28 g/cm3 and
AV=Vp-Vc a = 77 %, a shrinkage in volume by 58.8 %,
which with isotropic volume change repre-
,-K 100 (11-32) sents a linear shrinkage of 25.4%. This
value represents the volume change of the
polymer particles during pyrolysis without
taking into account the formation of pores
or shrinkage processes. In the case of the
(11-33) decomposition of the polysilazane bulk
materials, linear shrinkage of up to 28 %
46 11 Advanced Ceramics from Inorganic Polymers

has been measured. The 2-3 % difference ity must therefore be retained in reactive
can be attributed to sintering processes sintering in order that the gases produced
and represents a further 5 % shrinkage in can diffuse in or out of the bulk. Gas evo-
volume. These calculations agree well with lution also results in weight loss during the
the observed relative densities Z>p = 84 to pyrolysis of the polymer.
89 % in the polymer green bodies and up to
2)c = 94% in the pyrolysis product
(1000 °C).
On the basis of these considerations it 11.6 Summary and Outlook
must be assumed that densification can in
principle occur during the pyrolysis of po- The pyrolysis of polymers provides an
rous, polymeric compacts. This is remark- attractive method for the production of
able as diffusion-controlled sinter process- ceramic parts without the need to start
es in covalent, non-oxide materials are with ceramic powders. Si ceramic materi-
generally only activated at temperatures als with a residual open porosity of 6%
far above 1000 °C. Thus it is possible that can be prepared at 1000 °C. Ceramics of a
the mechanism of this densification in- covalent nature, such as carbides and ni-
volves reactive sintering and viscous flow. trides, can be produced at relatively low
temperatures and in the form of near dense
materials without the need for the addition
Viscous Flow
of sintering aids. The acronyms PDSCN
The densification rate in sintering by vis- (polymer derived silicon carbonitride) and
cous flow is determined by the interface PDSC (polymer derived silicon carbide)
energy y, the viscosity i\, and the particle have been proposed for these materials.
radius r. Mackenzie and Shuttleworth The X-ray amorphous silicon carboni-
(1949) formulated the relationship trides are stable up to 1400 °C and crystal-
lize only above this temperature with the
dS 3y
(11-35) formation of thermodynamically stable
— = —(1-9)
phases. They also have a lower density
At 2r]r
where S is the relative density Q/Q0. AS the than the crystalline phases and, due to the
interface energy and the viscosity are mate- low process temperature and the stability
rial dependent and are therefore difficult of the complex amorphous phases, materi-
to influence, greater densification is best als with novel property profiles can be pro-
achieved through the use of smaller parti- duced.
cles. Just such a particle-size dependence The ceramic matrix is free of condensed
has been observed in the densification of secondary phases which would tend to be
annealed polysilazane. detrimental to the high-temperature prop-
erties of the material. The thermal stability
of the amorphous silicon carbonitride al-
Reactive Sintering
lows its use in inert gas atmospheres up to
In this process the porosity is only 1440 °C and even above this temperature
slightly reduced, resulting in residual under oxidizing conditions in air.
porosities of between 12 and 30%. The A further potential of the polymer py-
process follows a diffusion mechanism rolysis process is the possibility of produc-
(Salmang and Scholze, 1982). Open poros- ing multi-component materials such as
 11.8 References 47

complex nitrides and carbides which are Burggraaf, A. I (1991), in: Concise Encyclopedia of
very difficult to synthesize using conven- Advanced Ceramic Materials: Brook, R. J. (Ed.).
Oxford: Pergamon Press, p. 62.
tional methods. Burkhard, C. A. (1949), /. Am. Chem. Soc. 71, 683.
Carduner, K. R., Blackwell, C. S., Hammond, W. B.,
Reidinger, R, Hatfield, G. R. (1990), J. Am. Chem.
Soc. 112, 4676.
11.7 Acknowledgements Carlsson, D. J., Cooney, J. D., Gauthier, S., Wors-
fold, D. J. (1990), J. Am. Ceram. Soc. 73, 237.
Chantrell, P. G., Popper, P. (1965), in: Special Ce-
I gratefully acknowledge the contribu- ramics 1964: Popper, P. (Ed.). London: Academic
tions of my co-workers Dr. A. Kienzle, Dr. Press, p. 87.
Cheronis, N. D. (1951), U.S. Patent 2579416.
G. Passing, Dr. H. Schonfelder, Dr. M. Clegg, W, Hesse, M., Klingebiel, U., Sheldrick, G.
Seher, and Dr. K. Strecker. I would also M., Skoda, L. (1980), Z. Naturforsch. 35b, 1359.
like to thank Dr. J. Mayer, Max Planck Cotton, R A., Wilkinson, G. (1980), in: Anorganische
Chemie, 3rd Ed. Weinheim: VCH, p. 320.
Institute for Metals Research, Stuttgart, Cranmer, D. C. (1988), Ceram. Eng. Sci. Proc. 9,
for the TEM results on the silicon carboni- 1121.
tride material, and Hoechst, Frankfurt, for Cullity, B. D. (1956), in: Elements of X-Ray Diffrac-
tion. Reading, MA: Addison Wesley, pp. 50 and
providing the polysilazanes. The generous 389.
financial support by the European Com- Dando, N. R., Tadayyoni, M. A. (1990), /. Am. Cer-
munity, the Deutsche Forschungsgemein- am. Soc. 73, 2242.
Davidson, I. M. X, Eaborn, C. (1974), J. Chem. Soc,
schaft, the Keramikverband Karlsruhe- Faraday Trans. 1, 70, 249.
Stuttgart, the KSB Foundation, and the Davidson, I. M. T., Stephanson, J. L. (1968), /. Chem.
Fonds der Chemischen Industrie is also Soc. A 282.
Davidson, I. M. T., Lawrence, R T., Fritz, G.,
gratefully acknowledged. Thanks are also Matern, E. (1982), Organometallics 1, 1453.
due to P. Fischer and P. Gartner for their Dislich, H. (1971), Angew. Chem. Int. Ed. Engl. 10,
assistance in the preparation of this 363; Angew. Chem. 83, 428.
Donaruma, L. G., Block, B. P., Loening, K. L., Plate,
manuscript. N., Tsuruta, T, Buschbeck, K. C , Powell, W. H.,
Reedijk, J. (1981), Pure Appl. Chem. 53, 2283.
Elschenbroich, C , Salzer, A. (1989), in: Organometal-
lics, A Concise Introduction. Weinheim: VCH,
p. 60.
11.8 References Erny, T., Seibold, M., Jarchow, O., Greil, P. (1993),
/. Am. Ceram. Soc. 76, 207.
Atwell, W. H., Weyenberg, D. R. (1969), Angew. Fitzer, E. (1985), in: Carbon Fibres and Their Com-
Chem. Int. Ed. Engl. 8, 469. posites. Berlin: Springer.
Barringer, E. A., Bowen, H. K. (1982), J. Am. Ceram. Fritz, G. (1987), Angew. Chem. Int. Ed. Engl. 26,
Soc. 65, C-199. 1111; Angew. Chem. 99, 1150.
Birringer, R., Gleiter, H. (1988), in: Advances in Ma- Fritz, G., Grunert, B. (1976), Z. Anorg. Allg. Chem.
terials Science and Engineering: Cahn, R. W. (Ed.). 419, 249.
Oxford: Pergamon Press, p. 339. Gerdau, T., Kleiner, H.-J., Peuckert, M., Briick, M.,
Blum, Y. D., Schwartz, K. B., Laine, R. M. (1989), J. Aldinger, R (1989), German Patent DE 3733727
Mater. Sci. 24, 1707. Al.
Boberski, C , Hamminger, R., Peuckert, M., Goto, T., Hirai, T. (1988), J. Mater. Sci. Lett. 7, 548.
Aldinger, R, Dillinger, R., Heinrich, J., Huber, J. Greil, P., Petzow, G., Tanaka, H. (1987), Ceram. Int.
(1989), Adv. Mater. 1, 378; Angew. Chem. Adv. 13, 19.
Mater. 101, 1592. Helm, D. R, Mark, E. (1937), J. Am. Chem. Soc. 59,
Boder, H., Golden, D., Rose, P., Wiirmseher, H. 60.
(1980), Z. Werkstofftech. 11, 275. Hengge, E. (1962), Z. Anorg. Allg. Chem. 315, 298.
Bordia, R., Raj, R. (1988), Adv. Ceram. Mater. 3(2), Homeny, I , Neergaard, L. I , Karasek, K. R., Don-
122. ner, J. T., Bradley, S. A. (1990), /. Am. Ceram. Soc.
Boury, B., Carpenter, L., Corriu, R. J. P. (1990), 73, 102.
Angew. Chem. Int. Ed. Engl. 29, 785; Angew. Chem. Hon, M. H., Davis, R. R (1979), /. Mater. Sci. 14,
102, 818. 2411.
48 11 Advanced Ceramics from Inorganic Polymers

Hon, M. H., Davis, R. F. (1980), /. Mater. Sci. 15, Okamura, K. (1987), Composites 18, 107.
2073. Paine, R. T., Narula, C. K. (1990), Chem. Rev. 90, 73.
Hong, J. D., Davis, R. F. (1980), J. Am. Ceram. Soc. Passing, G., Riedel, R., Petzow, G. (1991), </. Am.
63, 546. Ceram. Soc. 74, 642.
Hong, J. D., Davis, R. K, Newbury, D. E. (1981), /. Peuckert, M., Greil, P. (1987), J. Mater. Sci. 22, 3111.
Mater. Sci. 16, 2485. Peuckert, M., Vaahs, T., Briick, M. (1990), Adv. Ma-
Hunold, K. (1989), Powder Metall. Int. 21(3), 22. ter. 2, 398.
Interrante, L. V., Carpenter, L. E., Whitmarsh, C , Pouskouleli, G. (1989), Ceram. Int. 15, 213.
Lee, W, Garbauskas, M., Slack, G. A. (1986), in: Prochazka, S. (1975), Mater. Sci. Res. 9, 421.
Better Ceramics through Chemistry II: Brinker C. Prochazka, S., Klug, F. J. (1983), /. Am. Ceram. Soc.
I, Clark, D. E., Ulrich, D. R. (Eds.). Mater. Res. 66, 874.
Soc. Symp. Proc, Vo.. 73. Pittsburgh, PA: MRS, Pysher, D. J., Goretta, K. C , Hodder, R. S., Tressler,
p. 359. R. E. (1988), /. Am. Ceram. Soc. 72, 284.
Jenett, H., Bubert, H., Riedel, R., Stadelmann, H. Raj, R., Bordia, K. (1984), Acta Metall. 32, 1003.
(1990), Mikrochim. Acta 207. Rees, W. S., Seyferth, D. (1988), /. Am. Ceram. Soc.
Jones, P. R., Lim, T. F. O. (1976), J. Organomet. 88, C-194.
Chem. 38, 27. Remlund, G. M., Prochazka, S., Doremus, R. H.
Karch, J., Birringer, R. (1990), Ceram. Int. 16, 291. (1991a), J. Mater. Res. Soc. 6, 2716.
Kepler, R. G., Zeigler, J. M., Harrah, L. A., Kurtz, S. Remlund, R. M., Prochazka, S., Doremus, R. H.
R. (1987), Phys. Rev. B 35, 2818. (1991b), /. Mater. Res. Soc. 6, 2723.
Kienitz, H. (1968), Massenspektrometrie. Weinheim: Riccitiello, S. R., Hsu, M.-T. S., Chen, T. S.-S. (1987),
VCH. European Patent Application 0254809.
Kipping, F. S., Sands, J. E. (1921), J. Chem. Soc. 119, Rice, G. W, Woodin, R. L. (1988), J. Am. Ceram. Soc.
830. 71, C-181.
Kollenberg, W. (1991), in: Concise Encyclopedia of Riedel, R. (1993), in: Nicht-oxidische Keramiken aus
Advanced Ceramic Materials: Brook, R. J. (Ed.). anorganischen Vorstufen, Materialkundlich-Tech-
Oxford: Pergamon Press. nische Reihe 12. Stuttgart: Gebriider Borntraeger.
Kuzmany, H., Mehring, M., Roth, S. (1985), in: Elec- Riedel, R., Passing, G., Petzow, G., Krumbe, W,
tronic Properties of Polymers and Related Com- Franz, G. (1988), German Patent DE 3834325 Al.
pounds, Springer Ser. in Solid-State Sci., Vol. 63. Riedel, R., Seher, M., Becker, G. (1989a), /. Eur.
Berlin: Springer. Ceram. Soc. 5, 113.
Lange, F. F. (1973), /. Am. Ceram. Soc. 56, 445. Riedel, R., Passing, G., Petzow, G. (1989b), in: Proc.
Legrow, G. E., Lim, T. F , Lipowitz, J., Reaoch, R. S. 2nd Conf. Ceram. Powder Processing Sci: Hausner,
(1987), Am. Ceram. Soc. Bull. 66, 363. H. (Ed.). Bad Honnef: Deutsche Keramische
Liedike, S. (1987), J. Non-Cryst. Solids 97/98, 1083. Gesellschaft, p. 843.
Lipowitz, J., Freeman, H. A., Goldberg, H. A., Chen, Riedel, R., Strecker, K., Petzow, G. (1989c), /. Am.
R. T., Prack, E. R. (1986), Mater. Res. Soc. Symp. Ceram. Soc. 72, 2071.
Proc. 73, 489. Riedel, R., Petzow, G., Klingebiel, U. (1990 a), J.
Lipowitz, I, Freeman, H. A., Chen, R. T., Prack, E. Mater. Sci. Lett. 9, 222.
R. (1987), Adv. Ceram. Mater. 2, 121. Riedel, R., Seher, M., Petzow, G., Gerdau, T., Vaahs,
Mackenzie, J. K., Shuttleworth, R. (1949), Proc. T., Briick, M., Kleiner, H.-J., Peuckert, M.,
Phys. Soc. London B62, 833. Aldinger, F. (1990 b), German Patent DE 3840773
Matijevic, E. (1989), in: Proc. 2nd Int. Conf. Ceram. Al.
Powder Processing Sci.: Hausner, H. (Ed.). Bad Riedel, R., Seher, M., Petzow, G., Vaahs, T., Briick,
Honnef: Deutsche Keramische Gesellschaft. M., Kleiner, H.-J., Peuckert, M. (1990c), German
Mazdiyasni, K. S., West, R., David, L. D. (1978), /. Patent DE 3909078 Al.
Am. Ceram. Soc. 61, 504. Riedel, R., Kunesch, X, Passing, G., Kaysser, W A.,
Michl, J., Downing, J. W., Karatsu, T., Klingensmith, Petzow, G. (1990d), in: Proc. Int. Cont. on Metal
K. A., Wallraff, G. M., Miller, R. D. (1988), in: and Ceramic Matrix Composites: Processing, Mod-
Inorganic and Organometallic Polymers: Zeldin, elling and Mechanical Behaviour: Bhagat, R. B.,
M., Wynne, K. J., Allcock, H. R. (Eds.). Washing- Clauer, A. H., Kumar, P., Ritter, A. M. (Eds.).
ton, DC: American Chemical Society. Warrendale, PA: TMS, p. 579.
Mikhailev, B. M., Aronovich, P. M. (1960), Zh. Ob- Riedel, R., Schaible, S., Klingebiel, U., Noltemayer,
shch. Khim. 30, 3615. M., Werner, E. (1991), Z. Anorg. Allg. Chem. 603,
Miller, R. D. (1989), Adv. Mater. 1, 433; Angew. 119.
Chem. Adv. Mater. 101, 1713. Riedel, R., Passing, G., Schonfelder, H., Brook, R. J.
Nickel, K. G., Hoffmann, M., Greil, P., Petzow, P. (1992), Nature 355, 714.
(1988), Adv. Ceram. Mater. 3, 557. Riedel, R., Seher, M., Mayer, X, Szabo, V. (1995),
Noll, W. (1968), in: Chemie und Technologie der Sili- J. Eur. Ceram. Soc, in press.
cone, 2nd Ed. Weinheim: YCH.
 11.8 References 49

Ritter, X, Nair, S., Gennon, P., Dunlay, W. (1988), Timoshenko, S. P., Goodier, X N. (1970), Theory of
Adv. Ceram. Mater. 3, 415. Elasticity, 3rd Ed. New York: McGraw-Hill,
Rochow, E. G. (1952), in: Introduction to the Chemis- p. 395.
try of Silk ones, Weinheim: Verlag-Chemie. Toreki, W, Batich, C. D., Sacks, M. D., Morrone, A.
Roosen, A., Bowen, H. K. (1989), in: Proc. 2nd. Int. A. (1990), Ceram. Eng. Sci. Proc. 11, 1371.
Conf. Ceram. Powder Processing Sci.: Hausner, H. Verbeek, W (1973), German Patent 2218960.
(Ed.). Bad Honnef: Deutsche Keramische Verbeek, W (1974), U. S. Patent 3853 567.
Gesellschaft. Verbeek, W, Winter, G. (1974), German Patent
Saalfeld, F. E., Svec, H. J. (1963), J. Inorg. Chem. 2, 2236078.
50. Wakai, F , Kodoma, Y, Sakaguchi, S., Murayama,
Saalfeld, F. E., Svec, H. J. (1964), J. Inorg. Chem. 3, N., Izaki, K., Niihara, K. (1990), Nature 344, 421.
1442. Wannagat, U. (1964), Adv. Inorg. Chem. Radiochem.
Sakamoto, K., Obata, K., Hirata, H., Nakajima, M., 6, 225.
Sakurai, H. (1989), J. Am. Chem. Soc. 1111, 7641. Werner, E., Klingebiel, U., Pauer, F., Stalke, D.,
Sakurai, H., Hosomi, A., Kumada, M. (1968), /. Riedel, R., Schaible, S. (1991), Z. Anorg. Allg.
Chem. Soc. D 930. Chem. 596, 35.
Sakurati, H., Hosomi, A., Kumada, M. (1969), /. West, R. (1989), in: The Chemistry of Organic Silicon
Chem. Soc. D 1969. Compounds: Patai, S., Rappoport, Z. (Eds.). New
Salmang, H., Scholze, H. (1982), in: Keramik, Teil 1: York: Wiley.
Allgemeine Grundlagen und wichtige Eigenschaften. West, R., David, L. D., Djurovich, P. I., Stearley, K.
Berlin: Springer, p. 185. L., Srinivasan, K. S. V., Yu, H. (1981), J. Am.
Sawaguchi, A., Toda, K., Niihara, K. (1991), /. Am. Chem. Soc. 103, 7352.
Ceram. Soc. 74, C1142. Williams, R. W, Juterbock, B. N., Shinozaki, S. S.,
Sawhill, H. T, Haggerty, J. (1982), / Am. Ceram. Soc. Peters, C. R., Whalen, T. X (1985), Am. Ceram. Soc.
60, C131. Bull. 64, 1385.
Scace, R. I., Slack, G. A. (1960), in: Silicon Carbide: Wolff, A., West, R. (1987), Appl. Organomet. Chem.
Conner, J. R., Smiltens, J. (Eds.). Oxford: Perga- 1, 7.
mon, p. 24. Worsfold, D. X (1988), in: Inorganic and
Schilling, C. L. (1986), Br. Polym. J. 18, 355. Organometallic Polymers: Zeldin, M., Wynne, K.
Schmidt, H., Seiferling, B., Philipp, G., Deichmann, X, Allcock, H. R. (Eds.). Washington, DC: Ameri-
K. (1988), in: infrastructure Processing of Ad- can Chemical Society.
vanced Ceramics: Mackenzie, J. D., Ulrich, D. Wynne, K. X, Rice, R. W. (1984), Annu. Rev. Mater.
(Eds.). New York: Wiley, p. 651. Sci. 14, 297.
Schulenburg, S. (1991), Hoechst High Chem. Magazin Yamjima, S. (1985), in: Handbook of Composites,
11, 49. Vol. 1: Watt, W, Perov, B. V. (Eds.). New York:
Schwetz, K. A., Lipp, A. (1980), in: Science of Ceram- Elsevier.
ics, Vol. 10: Hausner, H. (Ed.). Bad Honnef: Yajima, S., Hayashi, X, Omori, M., Okamura, K.
Deutsche Keramische Gesellschaft, p. 149. (1976), Nature 261, 683.
Sefcik, M. D., Henis, J. M. S., Gaspar, P. P. (1974), Yajima, S., Hasegawa, Y, Hayashi, X, Imura, M.
/. Chem. Phys. 61, 4321. (1978a), /. Mater. Sci. 13, 2569.
Seyferth, D. (1988), in: Inorganic and Organometallic Yajima, S., Hasegawa, Y, Okamura, K., Matsuzawa,
Polymers: Zeldin, M., Wynne, K. X, Allcock, H. R. T. (1978 b), Nature 273, 525.
(Eds.). Washington, DC: American Chemical Soci- Yokoyoma, Y, Nanba, T., Yasui, I., Kaya, H.,
ety, p. 21. Maeshima, T., Isoda, T. (1991), /. Am. Ceram. Soc.
Seyferth, D., Mignani, G. (1988), J. Mater. Sci. Lett. 74, 654.
7, 487. Zhang, Z.-F., Babonneau, F, Laine, R. M., Mu, Y,
Seyferth, D., Wiseman, G. (1984), /. Am. Ceram. Soc. Harrod, X F., Rahn, X A. (1991), J. Am. Ceram.
67, C-132. Soc. 74, 670.
Shiina, K., Kumada, M. (1958), J. Org. Chem. 23, Ziegler, G., Heinrich, X, Wotting, G. (1987), /. Mater.
139. Sci. 22, 3041.
Soraru, G. D., Babonneau, F., Mackenzie, J. D.
(1990), /. Mater. Sci. 25, 3886.
Stolka, M., Yuh, H.-X, McGrane, K., Pai, D. M. General Reading
(1987), J. Polym. Sci,, Polym. Chem. Ed. 25, 823.
Strecker, K. (1990), Dissertation, University of Stutt- Auner, N., Weis, X (Eds.) (1994), Organosilicon
gart. Chemistry; From Molecules to Materials. Wein-
Tal'rose, V. L., Lyubimova, A. K. (1952), Dokl. Akad. heim: VCH.
Nauk SSSR 86, 909. Bassindale, A. R., Gaspar, P. P. (Eds.) (1991), Fron-
Tanigushi, L, Harada, K., Maeda, T. (1976), Japanese tiers of Organosilicon Chemistry. Cambridge: Roy-
Patent 76 530000. al Society of Chemistry.
50 11 Advanced Ceramics from Inorganic Polymers

Hench, L. L., Ulrich, D. R. (Eds.) (1986), Science of Riedel, R. (1993), Nicht-oxidische Keramiken aus
Ceramic Chemical Processing. New York: Wiley. Anorganischen Vorstufen, Materialkundlich-Tech-
Herrmann, W. A. (1990), "Chemie der metallorganis- nische Reihe 12. Stuttgart: Gebriider Borntraeger.
chen Verbindungen". Erzmetall 43, 69. Segal, D. (1989), Chemical Synthesis of Advanced Ce-
Laine, R. M. (Ed.) (1988), Transformation of ramic Materials, Chemistry of Solid State Materi-
Organometallics into Common and Exotic Materi- als 1: West, A. R., Baxter, H. (Eds.). Cambridge:
als: Design and Activation, NATO ASI Series. Cambridge University Press.
No. 141. Dordrecht: Martinus Nijhoff. Uhlmann, D. R., Ulrich, D. R. (Eds.) (1992), Ultra-
Peuckert, M., Vaahs, T., Briick, M. (1990), "Ceramics structure Processing of Advanced Materials. New
from Organometallic Polymers". Adv. Mater. 2, York: Wiley.
398. Zeldin, M., Wynne, K. J., Allcock, H. R. (Eds.)
Pouskouleli, G. (1989), "Metallorganic Compounds (1988), Inorganic and Organometallic Polymers.
as Preceramic Materials I, Non-Oxide Ceramics". Washington, DC: American Chemical Society.
Ceram. Int. 15, 213.
12 Biomimetic Processing
Paul Calvert

Department of Materials Science and Engineering, University of Arizona,

Tucson, AZ, U.S.A.

List of Symbols and Abbreviations 52

Glossary 52
12.1 The Meaning of "Biomimetic" 53
12.2 Structure and Properties of Natural Ceramics 53
12.2.1 Shell 54
12.2.2 Tooth 56
12.2.3 Bone 56
12.2.4 Sponge Spicules 58
12.2.5 Jade 58
12.2.6 Eggshell 59
12.2.7 Diatoms and Coccoliths 60
12.2.8 Potential for New Materials 60
12.3 Processing Methods in Biology 60
12.3.1 Mineralization in Vesicles 61
12.3.2 Extracellular Mineralization 62
12.3.2.1 Shell Growth 63
12.3.2.2 Bone Mineralization 64
12.3.2.3 Mineralization Methods Summarized 66
12.4 Biomimetic Processing 66
12.4.1 Powder Formation 67
12.4.1.1 Particle Growth in Liposomes 67
12.4.1.2 Other Synthetic Routes to Powders 69
12.4.1.3 Bacterial Fermentation 70
12.4.2 Bulk Parts 70
12.4.2.1 Fibrous and Laminated Ceramics 70
12.4.2.2 Polymer-Ceramic Composites 72
12.4.2.3 In Situ Mineralization of Polymers 74
12.4.2.4 Controlled Mineralization 76
12.5 Applications 77
12.6 Advanced Concepts 77
12.7 Conclusions 78
12.8 Acknowledgements 79
12.9 References 79

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
52 12 Biomimetic Processing

List of Symbols and Abbreviations

Gc critical strain energy release rate for fracture (toughness)
Klc critical stress intensity factor for fracture
Tg glass transition temperature

ABS acrylonitrile-butadiene-styrene
CAD computer-aided design
PDMS poly(dimethyl siloxane)
PEEK/AS4 poly(ether ether ketone)
PMMA poly(methylmethacrylate)
PTMO poly(tetramethylene oxide)
SFF solid freeform fabrication
TEOS tetraethoxy silane

Glossary
Apoferritin Ferritin protein without the iron oxide core
Chordate An animal with a backbone, includes fish, mammals, birds, reptiles
Coccolithophores Bacteria with a shell of calcite
Ferritin Protein shell enclosing a cluster of iron oxide
Frustule The two-part silica shell of a diatom
Liposomes Lipid bilayer enclosing a small volume of solution
Magnetotactic Bacteria that respond to a magnetic field
bacteria
Vesicle A liquid-filled volume surrounded by a membrane
Ormocers Organic-ceramic hybrids, also called ormosils, polycerams,
or ceramers
Spicule An elongated silica particle - these connect to form the skeletons
of sponges
Vacuole A small cavity containing air or liquid
 12.2 Structure and Properties of Natural Ceramics 53

12.1 The Meaning Structural biological materials range

of "Biomimetic" from very soft gels to hard ceramics. In the
context of ceramics processing, the wholly
In modern biology, great emphasis has organic structures are not relevant. Mate-
been placed on the study of cellular metab- rials like tooth and shell, which have more
olism and genetics, with a view to under- than 90 vol% of a mineral phase and a few
standing the chemical relationships under- percent of organic polymers, are certainly
lying living processes. Organisms are also pertinent. There is also a range of materi-
structurally interesting in that there are en- als with increasing organic content, which
lightening solutions to many of the me- are really composites but should be consid-
chanical problems that affect the design of ered because they form a bridge between
machines and structures. ceramics and filled polymers. In synthetic
Since ancient history there has been a materials, these two groups are regarded as
tradition of machines that are built to re- unrelated classes. The current interest in
semble animals in their operation, early ceramic-polymer and glass-polymer blends,
aeronautical engineering has many exam- made under the names of polycerams or
ples, most of which were disastrous. More ceramers, promises to bridge this gap.
recently, the term "biomimetic" has been Table 12-1 summarizes the mechanical
applied to studies of synthetic polymers properties of some biological materials.
that exhibit the catalytic properties of en- This article will briefly survey the struc-
zymes or the information storage proper- ture and properties of a number of biolog-
ties of nucleic acids (Gebelein, 1990). In ical composites and ceramics. The rela-
the last few years materials scientists have tionship between structure and properties
also come to realize that natural materials will provide a basis on which to argue that
can be informative models. improved properties could be obtained if
Structural biological materials, such as similar morphologies were produced in
human tendon or crab cuticle, are com- synthetic ceramics and composites. Fol-
posites where matrix phases, reinforcing lowing this there is a review of the methods
fibers and filler particles can be distin- by which biological ceramics and com-
guished. They differ notably from synthet- posites are formed. We then discuss pro-
ic composites in the complexity of the mor- cessing methods for reproducing these
phology. There is usually a hierarchy of morphologies.
structures and there are changes in this
pattern from site to site within the whole
component. We could learn from this hier- 12.2 Structure and Properties
archy and produce better composites by of Natural Ceramics
copying biological design. We may also try
to copy the methods by which the biologi- The immediate argument for biomimet-
cal materials are made. As with the poly- ic materials derives from a belief that the
mers, the adjective "biomimetic" in mate- mechanical properties of biological mate-
rials is taken to mean anything from "is a rials are superior to the equivalent synthet-
close reproduction of a biological materi- ic materials. This belief is partly based on
al" to "has some concepts in common with admiration for the elegant microstructures
our understanding of biological materi- of bone or shell but it can be supported by
als." measured mechanical properties. In iso-
54 12 Biomimetic Processing

Table 12-1. Properties of biological ceramics and composites.

Inorganic Modulus Strength Work

volume fraction of fracture
(GPa) (MPa)

Nacre, wet 95 64 130 1240

Nacre, dry 95 73 167 464
Macadamia nut shell 0 25-80 100-1000
Insect cuticle 0 6-10 80
Enamel 92 45 76 200
Whale bulla 66 30 33 200
Dentine 48 12 250 550
Bone 41 16 270 1700
Antler 31 7.7 179 6200

Table 12-2. Properties of natural and synthetic ceramics.

Modulus (GPa) Strength (MPa) K l c (MPam 1 / 2 ) Toughness

G c (Jm- 2 )

Alumina 350 100-1000 3 7

Fused silica 72 9
Nacre 64 130 600-1240
Enamel 45 76 13-200
Nephrite 222 3.6 562
5.5 619
2.9 269
Hornblende 151 7-16 49
Jadeite 205 205 94-120

tropic materials one expec ts microstruc- 12.2.1 Shell

ture to have a small effect on elastic modu-
lus but a much larger effect on fracture Mollusc shell has attracted the attention
behavior. We thus hope that biological of ceramists because it combines strength
materials will show superior fracture with toughness. The task of the shell is
strength or toughness, especially for the rather like that of tank armor. It protects
brittle ceramics. The comparison of a the soft contents from damage by impact,
range of hard tissues also shows that there crushing or boring. Different species of
is no single best structure. Fracture prop- mollusc have many differing shell mor-
erties are very sensitive to the type of frac- phologies. Laria and Heuer (1990) have
ture test and the different lifestyles of ani- studied the microstructure and mechanical
mals from the same evolutionary family properties of a number of mollusc shells,
can impose different critical loadings and including nacreous, foliated and crossed-
favor different micro structures. This is lamellar morphologies.
particularly apparent in shells. Table 12-2 The crossed-lamellar structure of the
compares the properties of a variety of nat- conch, Strombus gigas, is constructed of
ural ceramics with those of alumina and sheets of aragonite, 25-50 jim thick, ori-
silica. ented perpendicular to the surface of the
 12.2 Structure and Properties of Natural Ceramics 55

shell. The sheets are composed of laths,

which run at an angle to the surface of the
shell and are rotated by approximately 90°
in successive lamellae. These laths are
polycrystalline aggregates of aragonite
crystals, which are less than 50 nm thick,
100-500 nm wide and can be 10 |im long.
The grains are densely twinned on (llO)
planes which are parallel to the long edge
of the laths. The flexural strength of this
structure is about 100 MPa; "graceful",
non-catastrophic, failure occurs if the shell
is bent so that the shell interior is in ten-
sion. Currey and Kohn (1976) made simi-
lar measurements on a Conus shell and
found flexural strengths of 70 MPa for
bending designed to separate the large
sheets and 200 MPa for bending designed
to tear the sheets.
The pacific scallop, Pecten caurinus, has
a calcite structure made of laths assembled
into sheets which are parallel to the shell
surface. The flexural strength is again
about 100 MPa. The Knoop microhard-
ness of this shell is 1.5-2 times that of a
Figure 12-1. Structure of a mollusc shell, a) Outer
calcite single crystal. Scratch tests confirm
columnar structure, b) Inner layered nacreous struc-
the higher hardness of the shell. This has ture.
been attributed to microstructural effects
rather than to twins, entrained organic
material or other intrinsic crystal proper- wall. The organic thick layer is 20-50 nm
ties. thick and comprises five layers, see
The shell of the pacific mussel, Mytilus Sec. 12.3.2.1. During fracture, fibrils of the
edulis, is typical of many bivalves and has polymer are seen bridging cracks. This
a nacreous inner shell of aragonite with a mechanism may contribute significantly to
prismatic outer layer of calcite (Fig. 12-1) the toughness of this structure.
(Lowenstam and Weiner, 1989). The red Jackson etal. (1988, 1989) have mea-
abalone, Haliotis rufescens, has a similar sured the mechanical properties of Pinctada
structure (Sarikaya etal., 1990). The (a pearl oyster, which is not a true oyster)
whole shell is reported to have a strength nacre under wet and dry conditions and
of 180 MPa and a toughness Klc of have developed models. Drying the shell
8 ± 3 M P a m 1 / 2 . The nacreous layer has increases the modulus (60 GPa to 70 GPa)
thin, roughly hexagonal, plates of arago- and increases the strength (140 MPa to
nite separated by an organic layer. The 170 MPa) but reduces the work of fracture
plates are 0.25-0.5 jim by several |im wide (1240 J m~ 2 to 350 J m~ 2 ). Since water
and are usually staggered, like bricks in a plasticizes the protein binding layer, this
56 12 Biomimetic Processing

suggests that much of the fracture tough-

ness is due to plastic deformation of the
organic matrix, which pulls out into long
fibrils. This group shows that the modulus
can be fitted to a shear lag model and that
the strength and work of fracture can be
interpreted in terms of the pull-out stress
and energy for the platelets. The behavior
of nacre is compared with that of a model
comprising glass microscope slides glued
together with epoxy resin.
Jackson et al. (1988) also discuss the rel-
ative importance of fracture toughness Figure 12-2. Enamel of a rat tooth showing crossed
when compared to the greater crack-stop- rods of hydroxyapatite.
ping ability of the crossed-lamellar struc-
ture. They argue that avoidance of cata-
strophic failure is often more important A number of other tooth structures have
than strength and hence the crossed lamel- been described which seem to offer lessons
lar structure should be favored, especially for the development of tough ceramic struc-
by larger molluscs. It does seem that shell tures. The teeth of the limpet Patella vul-
microstructure is not conserved within gata (Runham et al., 1969) contain 78 wt%
families but is adapted to the requirements goethite (Fe 2 O 3 H 2 O) and 13wt% silica
of the animal. This suggests that different at the cutting tip while the back of the tip
types of toughness will also be important has 9 wt% goethite and 75 wt% silica. The
for different artificial applications of ce- whole tooth shows an increasing hardness
ramics. and inorganic content towards the tip.
Similarly the teeth of chitons, gastropods
12.2.2 Tooth resembling limpets, contain magnetite at
Tooth is a material which is designed for the tip and carbonated hydroxyapatite or
wear and chemical resistance. Mammalian amorphous ferric phosphate in the interi-
tooth enamel has more than 95 vol% hy- or. The organic matrix contains chitin,
droxyapatite as rods. In the rat tooth which is the predominant arthropod struc-
(Fig. 12-2) the rods are arranged in layers tural polymer (Lowenstam and Weiner,
with alternate layers at 90°. Many other 1989).
arrangements are also possible. The rods
are about 2 jam in diameter and are com-
12.2.3 Bone
posed of long crystals of about 50 nm
width. There has been little work on the A number of skeletal materials resemble
fracture properties of enamel. The material composites or particle reinforced polymers
is less tough and strong then nacre (Table in that the mineral phase occupies between
12-1) and there has been no evidence that 20% and 70% of the volume (Table 12-1).
the organic component has any influence The modulus is considerably less than that
on the fracture toughness. It is believed of a ceramic but the strength and tough-
that the protein component controls the ness are high. Bone has a matrix of colla-
morphological development. gen reinforced with hydroxyapatite plates
 12.2 Structure and Properties of Natural Ceramics 57

at volume fractions from 20 % in deer ant- filled synthetic polymers (Table 12-3). Bigg
ler to 50 % in a penguin humerus (Currey, (1987) has described the properties of No-
1984). Over this range the Young's modulus ryl, a tough blend of polyphenylene oxide
goes from 4 GPa to 28 GPa and the bend- and polystyrene, filled with talc, a platy
ing strength from 30 MPa to 300 MPa. At filler. At 30 vol% talc, the modulus in-
the higher volume fractions the fracture creases from 2.2 GPa for the polymer to
toughness decreases with increasing min- 6 GPa and the strength goes from 48 MPa
eral content from a high value around to 70 MPa, the impact strength decreases
6 kJ m ~ 2 to 2 kJ m ~ 2 in the penguin bone. by a factor of eight. Higher filler loadings
Ku values for bovine bone are around are not practical because of loss of tough-
5 MPa m 1/2 , but show considerable varia- ness. These properties are for a well-
tion between samples (Behri and Bonfield, bonded system, ABS-talc - where ABS is
1980). acrylonitrile-butadiene-styrene, a tough-
These properties are quite anisotropic. ened polystyrene - has poorer interfacial
Lewis (1990) has collected data on elastic bonding and shows a decrease in strength
constants which show ratios of 1.5-3 be- with added filler.
tween the longitudinal and radial Young's As shown in Table 12-4, there is a prob-
moduli. There is a similar anisotropy in lem in finding prosthetic materials which
bend strength. The tympanic bulla (ear provide a good match to the mechanical
bone) of a whale is interesting in that it has properties of bone. Bonfield (Bonfield
a very high mineral content, 66 vol%, and etal., 1981; Doyle etal., 1991) has devel-
high modulus but very poor strength and oped polyethylene/hydroxyapatite com-
toughness. The material is isotropic and is posites as potential bone substitutes. The
presumably optimized for its elastic prop- elastic moduli rise with volume fraction of
erties rather than for load bearing. mineral and reach 9 GPa at 48 vol%. The
This behavior of biological composites strength is about 25 MPa and ATlc is 3 MPa
should be compared with that of particle- m 1/2 . At higher volume fractions the corn-

Table 12-3. Properties of synthetic and biological composites.

Volume Modulus Strength Strain Work of

fraction to break fracture
(%) (GPa) (MPa) (%) (Jm~ 2 )

Continuous fiber PEEK/AS4a 61 140 2200 1900

Sheet molding compound 10 130
Poly(butyleneterephthalate)/glass beads 25 4.9 95
Poly(ethyleneterephthalate) + short 35 20 165 1 3200
glass fiber
Bone (bovine femur) 41 20 220 10 1700
Collagen (tendon) 0 3 100
Hydroxyapatite 100 110 100
Poly(ethyleneterephthalate) 0 3.3 60 275 7300
E glass 100 70 3000
a
PEEK: poly(ether ether ketone).
58 12 Biomimetic Processing

Table 12-4. Mechanical properties of bone and replacement materials for bone.

Modulus Strength Toughness

(GPa) (MPa) (MPam 1/2 ) G c (Jm~ 2 )

Cortical bone 7-30 50-150 2-12 600-5000

Alumina 365 3 40
Ti Alloy 106 900 80 104
PMMA 3.5 70 1.5 400
Polyethylene 1 30 8000
Polyethylene-Hydroxyapatite, 40% 5 22 3 1800
Polyethylene-Hydroxyapatite, 50% 9 26 3 1000

posite becomes unmanageably brittle. silica is deposited as concentric cylinders

Thus it is only possible to achieve moduli on an axial organic filament. The water
in the lower edge of the bone range and the content of the silica is 9%. The Young's
strength is much poorer. The lesson is that modulus is 36 GPa, about half that of sili-
not only the composition of the composite ca, and the bending strength is 600 MPa.
is important but also the morphology and, The layered structure is believed to induce
probably the interfacial bonding. Bonfield a higher strength by stopping cracks and
avoided using coupling agents because of minimizing the initial flaw size.
concerns about biocompatibility of the
bone substitutes. 12.2.5 Jade
Crustacean cuticle (e.g. crabshell) is sim-
ilar to bone. The matrix is principally chit- Jade is a group of tough minerals that
in as it is insect cuticle, but it is mineralized are not biological materials but share a
with calcium carbonate, either as calcite or fibrous structure rather like that of biolog-
as amorphous calcium carbonate (Lowen- ical ceramics. The fibrous morphology
stam and Weiner, 1989). The calcite can arises from the chain silicate structure
form elongated crystals which follow the found in jadeite and nephrite. Rowcliffe
direction of the chitin fibrils. and Fruhauf (1977) report fracture tough-
ness values Klc from 2 to 5.5 MPa m 1/2 ,
strength from 22 to 74 MPa, work of frac-
12.2.4 Sponge Spicules
ture Gc from 270 to 990 J m " 2 . Pre-
The skeletons of sponges (Porifera) cracked samples yielded Klc of 3.7 MPa
comprise silica rods which are formed in m 1/2 and work of fracture of 1840 J m~ 2 .
vesicles then joined at the ends to form The authors point out that jade stands out
scaffolding. The rods can have complex from ceramics in terms of the high value of
shapes and are typically 100 jim long by Gc whereas Klc is comparable. Measure-
10 jim in diameter. ments by Bradt (1973) show that jadeite is
There is one example of a macroscopic less tough than nephrite and has a less
silica deposit, namely the megasclere, a rod fibrous texture. Wu et al. (1990) have
which is used by some sponges to anchor shown that a highly oriented fibrous horn-
them to the sea bed (Levi et al., 1989). The blende was strong when the fracture propa-
sponge Monoraphis, which is found at gated across the fibers but apparently
depths of 1 km, forms a silica rod of a few tougher when the fracture was parallel to
millimeters in diameter and 1 m long. The the fibers. However, the high toughness
 12.2 Structure and Properties of Natural Ceramics 59

(X lc = 16 MPa m 1/2 ) was coupled with cat-

astrophic failure. Increasing fiber length
and smaller fiber diameter led to greater
toughness.
From these studies on jade we learn that
fibrous structures can show high values of
fracture energy and of Klc. Fiber size may
be important for strength in limiting the
maximum flaw size to fiber diameter. Ax-
ial ratio is important for developing tough-
ness. The relationship is unclear between
Klc, Gc, test method and the appearance of
catastrophic or non-catastrophic failure.
Jade has a reputation for toughness that
comes from its ability to be machined. In
mollusc shells, there seems to be a similar
dependence of the optimum microstruc-
ture on the expected loading mechanism.

12.2.6 Eggshell
A recent study has concentrated on the
structure and formation of eggshell (Fink
et al., 1992). This represents a particularly
interesting biological structure as it forms
in the oviduct by deposition of calcite on
an organic membrane, which surrounds
the contents of the egg. Once the mem-
brane is made there is apparently no direct
cellular involvement in the process, but
simply controlled secretion of the precipi- Figure 12-3. Structure of avian eggshell, a) Micro-
tants. This suggests that we could hope to graph of a fracture surface, regions of the eggshell:
reproduce such a process synthetically SM, shell membranes; M, mammillae; P, palisades; C,
without needing to duplicate the great cutile. b) Schematic representation of the zones of
mineralization. (A. H. Heuer, L. Kuhn and M. Agar-
complexity of a cellular matrix. wal, Case Western Reserve University.)
The bulk of the thickness of a shell con-
sists of palisades, columns of calcite which
are rounded on the inner side at a point of zones where the palisades adjoin may be
origin on the membrane. The structure rich in matrix polymers, principally pro-
very much resembles the columnar zone of teoglycans. A disordered region of calcite
a cast metal. The content of organic mate- crystals at the base of the palisades acts as
rial is 2-4%. Within the palisade are cal- a calcium reservoir for the growing chick.
cite crystals 1 (am wide by 0.3 |im thick in Dissolution in this zone may also aid frac-
a brick-like pattern with the long axis par- ture from the inside as the chick pecks its
allel to the shell surface (Fig. 12-3). The way out of the shell.
60 12 Biomimetic Processing

Some mechanical measurements on b) high-temperature ceramics with a fi-

shell have been reported (Vincent, 1990) brous microstructure, which are isotrop-
but the porosity and asymmetry of the ic on the scale of several fiber lengths;
structure make strength measurements al- c) highly fiber-filled composites with bone-
most impossible. like properties.
The key questions which we may hope
In principle the above materials should
to see answered by studies of shell forma-
be formable from carbide, nitride or oxide
tion concern the mechanisms by which the
ceramics. In the longer term, more com-
sequence of structures shown in Fig. 12-3
plex composites could be constructed,
are formed by sequential secretion of nu-
which will be discussed at the end of this
cleating solid polymers, calcium and car-
chapter.
bonate ions, growth inhibitors and soluble
The examples of particle formation in
matrix polymers.
vesicles imply that similar methods can be
used to make ceramic powders with parti-
12.2.7 Diatoms and Coccoliths cles on complex shapes and that very pre-
Siliceous diatoms are single-celled ani- cisely patterned structures can be grown by
mals which form an outer shell of silica solution processing, as opposed, to the
with a well defined pattern of pores. The vapor-phase processing used for integrated
frustule completely encloses the diatom. circuits.
The frustule forms in two halves, each of
which is enclosed by a membrane during
growth. The interest in these structures is 12.3 Processing Methods
that they are patterned on a scale of in Biology
100 nm suggesting that biomimetic pro-
cesses could be used to make small scale Given that there are biological materials
devices (Li and Volcani, 1984). for which we wish to mimic the composite
Unicellular photosynthetic algae (coc- morphology, several approaches may be
colithophores) are enclosed within a com- taken. It may be possible in some cases to
plex calcite shell, the coccolith. The shell is use organisms directly to produce the de-
formed by the assembly of calcite single sired product. This is certainly feasible for
crystals which are shaped rather like wish- the preparation of inorganic powders by
bones with an overall length of about microorganisms, but it is unlikely that
0.5 jam. The development of these peculiar large components can be made by collec-
crystals is again controlled by a surround- tion of animal cells in the foreseeable fu-
ing membrane (Young et al., 1992). ture.
A second approach is to use wholly syn-
12.2.8 Potential for New Materials thetic methods, such as tape-casting or
spin-casting to make the required struc-
The examples given above suggest a
tures. This approach will be discussed be-
number of immediate goals for biomimetic
low under "fibrous and laminated ceram-
structural materials. These include
ics". Alternatively we can explore how bio-
a) tough ceramics for low-temperature ap- logical materials are produced and plan to
plications, with elongated grains bound use some modification of the biological
by a thin layer of polymer; processing technique to make the material.
 12.3 Processing Methods in Biology 61

Biological methods of mineralization are compartment. In addition, shape control is

very varied. At the extreme of low levels of exerted by some combination of the shape
control, the dental caries-forming organ- of the membrane and a nucleating surface.
isms Streptococcus sanguis and Corynebac- In the second category, the mineral is
terium matruchotii can nucleate hydroxy- formed outside the cell and possibly out-
apatite crystallization at low supersatura- side the organism. The process is then a
tions, close to those at which hydroxyapa- simple precipitation, primarily controlled
tite forms on seed crystals. S. mutans and by local concentrations of solutes but
S. sanguis excrete polysaccharides, dextran modulated by growth inhibitors, polymer-
and glucan, which act as a binding matrix ic and non-polymeric, and by solid poly-
for apatite growth. The result is a tough, mers which may provide nucleation sites
adherent composite (Boyan etal., 1992; or impose preferred orientations.
Brooker, 1979).
In coral, the mineralization happens
12.3.1 Mineralization in Vesicles
also completely outside the organism but
there may be manipulation of the solution The sponge spicule is an example of vesi-
concentrations close to the cell surface as cle mineralization that has been studied in
well as the provision of a nucleating site. In some detail (Garrone etal., 1981). The
bone, mineralization takes place outside composition and properties of the spicule
the cells but within a matrix whose proper- have been described above. Silica is appar-
ties and solute concentrations are under ently transported in the cell as silicic acid
cellular control. Sponges and many single- (Si (OH)4) which may be complexed by cat-
celled organisms form minerals within echols or other hydroxylated compounds
vesicles, surrounded by a membrane and to raise the solubility (Sullivan, 1986).
wholly within the cell. There is, however, no known enzyme sys-
Mann has offered a classification of bio- tem which is responsible for metabolizing
logical mineralization processes which is or transporting silica. The silicic acid is
focused on the role of the biological inter- carried through the membrane into the
face in nucleating the particle (Calvert and vacuole. Super saturation may be achieved
Mann, 1988). Four types are distinguished inside the vesicle by manipulation of pH.
according to whether the matrix is (I) a Sea water is roughly saturated for amor-
passive support, (II) provides specific nu- phous silica and supersaturated for crys-
cleation sites, (III) limits growth by a talline quartz. All biogenic silica precipi-
membrane, or (IV) controls both the site tates are amorphous. Thus, the growing
and orientation of nucleation. Biological precipitate would be unstable in the envi-
processes are so diverse that any such divi- ronment outside the cell.
sions must be artificial and only partly sat- Shape control is exerted by two meth-
isfactory. ods. Spicules have an internal polymer
For the purposes of biomimetic materi- thread of about 0.1 |im diameter which is
als formation, it may be useful to stress probably protein. On this thread the silica
another categorization. We can identify initially precipitates. In siliceous diatoms,
two processing systems. In one case, the complicated sheet structures are formed
mineral is formed in a vesicle within the which become the shell of the diatom. The
cell. The surrounding membrane controls shape of these is the product of action by
the concentrations of reagents inside the the external membrane whose shape is in
62 12 Biomimetic Processing

turn presumably controlled by the cellular cium oxalate in plants, including prickly
cytoskeleton. pear (Rivera and Smith, 1979).
This intravesicular mineralization route The growth of magnetite single crystals
also applies to silica in diatoms (Sullivan, in Aquaspirillum apparently proceeds via
1986). Magnetite formation in the magne- deposition of a ferric oxide gel, followed
totactic bacterium Aquaspirillum magneto- by delivery through the membrane of Iron
tacticum is similar (Fig. 12-4). Also coccol- (II) which reacts at the surface of the grow-
iths form complex shapes in calcite by this ing crystal (Frankel and Blakemore, 1984;
method (Fig. 12-5). It applies to a number Mann et al., 1984). How the crystal is nu-
of precipitates which may have the func- cleated is not clear but there may be a
tion of either binding a toxic element, such specific region on the vesicle membrane
as the precipitation of cadmium sulfide by provided for this purpose.
yeast (Dameron et al., 1989), or of render-
ing the organism toxic to whatever might 12.3.2 Extracellular Mineralization
otherwise eat it, such as the storage of cal-
Extra-cellular mineralization has been
studied in coral but the mechanism is still
not resolved. Constanz (1986) has shown
that aragonite grows on nuclei produced
by the coral. These nucleating bodies do
contain calcite crystals on which the arago-
nite seems to form. Constanz suggests that
this is the main control mechanism, with
no controlled modulation of local solute
concentrations or secretion of growth-
modifying polymers. However, the process
of crystallization is faster in the light as
photosynthesis by algae removes carbon
Figure 12-4 Bacterial magnetite particles. (S. Seraphin dioxide from the sea water and shifts
and K. Law, University of Arizona.) the bicarbonate/carbonate equilibrium.

Figure 12-5. Growth se-

quence of Emiliania huxleyi
coccoliths. (Young et al.,
1992.)
 12.3 Processing Methods in Biology 63

Aragonite

Acid Phosphorprotein t

Beta - sheet protein

Chitin Figure 12-6. Scheme of five

polymer sheets between layers
Beta - sheet protein of aragonite crystals in
nacreous shell. (After Lowen-
Acid Phosphorprotein t stam and Weiner, 1989.)

Growth rates of 22 \im per day have been produces hydrogen ions which must some-
measured, which dropped to zero in the how be removed from the growth front
east Pacific when El Nino raised the water (Wheeler, 1992).
temperature (Risk and Pearce, 1992). Electron microscopy of nacreous shell
The calcite crystals (otoconia) found in edges shows empty "mortar" compart-
the balance organs of mammals are a sim- ments of the brick-and-mortar structure.
ilar case of controlled extracellular crystal- This suggests that the organic compart-
lization (Mann et al., 1983). What is ap- ments form first and are filled afterwards.
parently a single crystal several microme- However, such a structure would be ex-
ters long, actually has an organic core cov- pected if there were any dissolution during
ered with nucleation sites which is extrud- sample preparation. Hence it is also possi-
ed by a cell. Oriented nucleation on the ble that crystals grow on an exposed poly-
core gives rise to radially growing crystals mer layer and are then overcoated with a
which then merge to form a single unit. new organic layer (Lowenstam and Weiner,
1989). Either interpretation leaves many
12.3.2.1 Shell Growth
puzzling questions.
Mollusc shell is in contact with a pool of The organic mortar comprises up to five
fluid which is trapped between the tissue of layers (Fig. 12-6) (Lowenstam and Weiner,
the mantle and the shell. This extrapallial 1989). The outer surfaces are acidic
fluid has a supersaturation with respect to proteins, rich in aspartic acid, with at-
calcium carbonate of several times, which tached sulphated sugar units. This struc-
is about the same as sea water. The con- ture would be expected to interact with
centration of ions in the immediate region ions in solution or on crystal surfaces. In-
of the growing front may be higher and has ner layers on either face are hydrophobic
not been determined. The enzyme, carbon- silk-like proteins and the central region is
ic anhydrase, is associated with shell for- the crystalline polysaccharide chitin. This
mation in mollusc shell and eggshell. It structure thus has a central structural chit-
converts carbon dioxide and water to bi- in layer, the silk layer, which is tough and
carbonate ions - normally quite a slow extensible, and may be responsible for in-
equilibration. This could supply carbonate creasing the work of fracture while the out-
to the shell at a higher rate, but it also er layers control crystal growth.
64 12 Biomimetic Processing

More recently, Keith et al. (1993) have ation or to adsorption of nuclei on the
analyzed macromolecules from mussel, surface once they have formed. Crenshaw
nautilus and abalone. 7V-Acetyl glucosa- et al. (1987) noted that shell proteins were
mine, characteristic of chitin, was found in also effective in forming hydroxyapatite
nautilus but not mussel or abalone. The crystals. Thus the effect is neither very
implication is that the five-layer model is strong nor very specific.
not always correct.
A number of groups have studied the
12.3.2.2 Bone Mineralization
inhibitory effect of the soluble shell protein
on crystal growth of calcium salts. Addadi Bone mineralization may be seen as a
and Weiner (1985) showed that the proteins more advanced form of coral formation.
affect crystal morphology by binding to The precipitation is also extracellular but
specific crystal faces. Sikes et al. (1991) occurs here in a matrix which has a struc-
showed that polyaspartic acid slowed the ture that can guide the mineralization. The
growth of calcium carbonate with the ef- subject has recently been reviewed by
fect increasing with chain length up to 15 Lowenstam and Weiner (1989). Bone for-
aspartate units. Mollusc proteins can be- mation is very complex and not well un-
come incorporated into the growing car- derstood despite an enormous amount of
bonate crystals and modify the fracture research. Dentine, calcified tendon and
surface from faceted to conchoidal, but it calcified cartilage have much in common
is not certain whether there is an actual with bone.
change in strength (Berman et al., 1990). The bone mineral is hydroxyapatite, a
While it is clear that polymers can be basic calcium phosphate Ca 10 (PO 4 ) 6 (OH) 2
very effective in inhibiting crystal growth, with 4 - 6 % carbonate substitution and
less is known about how they induce nucle- some monohydrogen phosphate. Much of
ation. It has been suggested that the same the non-stoichiometry may be accounted
polymers which inhibit growth when in so- for by surface sites. The crystals have the
lution become nucleating agents when at- form of thin plates about 50 nm x 25 nm
tached to a surface (Crenshaw et al., 1987). x 3 nm. The crystals show central dark
Sheets of matrix polymer from mollusc lines in electron micrographs, which may
shell would induce growth of calcium car- be a layer of octacalcium phosphate, which
bonate if the anionic polymer was present. in turn may be the first precipitate.
Addadi and Weiner (1985) have shown The first evidence of bone crystals in
that the anionic proteins act similarly many cases is associated with matrix vesi-
when adsorbed to glass surfaces. Rieke cles. These vesicles are buds that develop
(1987) and Addadi et al. (1987) have shown from cell membranes. They are rich in
that carboxylate groups on polyethylene phosphatase enzymes and the hydroxyap-
or sulphate groups on polystyrene result in atite crystals first form in the vesicle mem-
attached carbonate crystals with a prefer- brane. When the bulk of the mineraliza-
ence for the (001) plane being parallel to tion takes place these first randomly-ar-
the surface. However, it is noticeable that ranged crystals seem to disappear and their
this effect is not great, the density of at- significance is unclear (Sela et al., 1992).
tached crystals is quite small, several hun- Subsequent crystallization takes place in
dred per cm2. Wheeler (1991) has pointed association with collagen fibrils. It is an
out that this effect may be due to nucle- open question whether amorphous calci-
 12.3 Processing Methods in Biology 65

urn phosphate is generated before hy- droxyapatite but mineralization is inhib-

droxyapatite formation and also whether ited. Pyrophosphate (P2C>7~) may be re-
octacalcium phosphate is the initial precip- sponsible for growth inhibition; it turns
itate as mentioned above. It has long been over very rapidly in the body because of
believed that collagen is responsible for enzymatic action and is a strong inhibitor
nucleating hydroxyapatite. Collagen fibrils of crystallization. The presence of phos-
are composed of rod-like triple helices phatases in matrix vesicles may be due to
grouped into bundles and laid end-to-end their role in removing pyrophosphate in
with a 36 nm gap between ends. The triple the mineralization zones (Eanes, 1992).
helix of collagen is 300 nm long and adja- Until recently it was not possible to
cent fibrils are arranged with a % stagger. grow bone from cell cultures in the labora-
As a result the fibrils have gap regions ev- tory. Bone cells, osteocytes, can now be
ery 68 nm where a group of helices ends. taken from young tissue and grown on
This 64 nm spacing becomes very marked mats of collagen that become mineralized
in the early stages of mineralization, sug- (Heuer etal., 1992; Robey, 1992).
gesting that collections of acidic groups As in shell, there has been much interest
near the ends of the helices may form a in identifying the roles of collagen and sol-
nucleating site. The hydroxyapatite plates uble proteins in the control of bone miner-
ultimately form a parallel stack within a alization. Collagen has been shown to act
collagen fibril. as a nucleating agent for hydroxyapatite in
At a more macroscopic level, the long vitro but it is a disappointingly ineffective
bones of the newborn are cartilage which one (Koutsoukos and Nancollas, 1987). It
contains some mineral. This is invaded by has been shown that various insoluble
bone-forming cells which propagate min- polyamides with attached organophosphi-
eralization towards the bone ends. A resid- nates [(CH3O)2P = O]-polymer can act as
ual plane of cartilage is left near the bone effective nucleating agents for hydroxyap-
ends. In this zone new matrix continues to atite but with a rate almost three orders of
form and calcify as the youth matures and magnitude slower than needed for crystal-
the bones lengthen. The bones thicken by lization on hydroxyapatite seeds and 20
deposition at a membrane surrounding the times less than for a similar nominal sur-
outer surface (Engfeldt et al., 1992). face area of collagen (Dalas et al., 1991).
Bone is not static but is continually re- To summarize the "processing" of bone
sorbed and redeposited. This allows re- we can derive the following points:
modelling to account for changing stress
distributions. The mechanism of stress • organic matrix is first laid down then
sensing which permits the appropriate mineralized;
growth patterns is not known but piezoelec- • long bones extend by extrusion from a
tric effects have been suggested. growth plate;
Pathological mineralization can occur in • the details of nucleation and growth are
damaged soft tissues (Daculsi et al., 1992). not understood;
Mineralization of tendons and in joints • growth inhibitors are important for pre-
can happen after damage and leads to in- venting unwanted mineralization while
flammation and pain (Calvert and Dieppe, allowing the body always to be supersat-
1982). It is apparent that normal body flu- urated;
ids and tissue are supersaturated for hy- • very small crystals are formed with pre-
66 12 Biomimetic Processing

cise shape as a result of the composition have no reason to believe that intrinsic me-
of inorganic salts present; chanical performance was an overriding
• these crystals are laid down parallel un- basis for choice of mineral in living crea-
der the influence of the collagen fibrillar tures.
structure; Vesicle growth applies to small complex
• the body has many types of collagen but particles that may then be assembled. Ex-
most of them do not mineralize. tra-cellular growth allows formation of
bulky solid parts. In each system of extra-
cellular mineralization, there is a growth
12.3.2.3 Mineralization Methods
front at which mineralization occurs. Cells
Summarized at the leading edge of this front control the
Based on the foregoing examples we can nucleation process, with continuing growth
attempt some generalizations, while keep- at greater depths until full density is
ing in mind that biology favors variety reached, over a distance which is typically
over consistency. The main structural bio- 100 \xm from the growth front. Structures
logical minerals are calcium carbonate, hy- at the scale of 1 jim or less must be con-
droxyapatite and silica. Both carbonate trolled by the kinetics and thermodynam-
and silica are grown in vesicles in circum- ics of the process. At longer scales the cells
stances where complex shapes are formed. can directly control the morphology.
All three minerals can also form under cel- Biological growth rates are generally
lular control in an extracellular matrix, in thought of as being slow. Maximum
the case of silica this occurs in some mol- growth rates for bone and antler reach cen-
lusc teeth (Runham et al., 1969). Both car- timeters per day. Two separate rates are of
bonate and hydroxyapatite can be exploit- interest here. The first is the growth rate of
ed by controlled mineralization from su- an individual crystal or particle and we
persaturated solutions. Carbonate exam- would expect it to be comparable to that
ples include eggshell and coral, for hy- for a similar crystal at similar supersatura-
droxyapatite the supersaturation occurs in tions measured in vitro. The second rate is
saliva. Thus, there is no intrinsic difference that for the overall advance of the mineral-
between the handling of crystalline and ized interface. This may be due in part to
amorphous precipitates. the rate at which solutes can be supplied
The restriction of silica mineralization but will also depend on the rate of forma-
to plants and marine animals is probably tion of the associated organic structures.
related to its low solubility and the conse- These rates will be of importance if we
quent difficulty of obtaining large amounts contemplate the use of organisms to grow
without a large throughput of water. Ma- synthetic parts. They will also limit the rate
rine animals do apparently use hydroxyap- at which bacterial fermentation methods
atite rather than carbonate in circum- can be used to produce ceramic particles.
stances where better mechanical properties
are desired. Otherwise the dominance of
hydroxyapatite in chordates (animals with 12.4 Biomimetic Processing
backbones) may reflect the need to main-
tain a large phosphate reserve in view of With what we have learnt from biology
the relative rarity of phosphorus in the en- we can outline a group of processing
vironment. With the exception of teeth, we strategies for ceramics and composites that
 12.4 Biomimetic Processing 67

could be called biomimetic. Particle particles from Aquaspirillum magnetotacti-

growth within a membrane should allow cum. The particle is a submicrometer, sin-
the formation of particles of controlled gle crystal and kept dispersed by the sur-
size and shape with agglomeration being rounding membrane. Highly uniform par-
prevented by the surrounding organic lay- ticles can be formed by controlled precipi-
er. Bulk ceramics and composites could be tation from solution (Matijevic, 1981).
made by in situ formation of inorganic These precipitation methods tend to be
particles in a polymer or gel matrix. In rather slow and take place in dilute solu-
addition more direct methods could be tion and so far have not yet been practica-
used to reproduce the same types of lay- ble for commercial ceramics.
ered and fibrous composites that are found
in biology. 12.4.1.1 Particle Growth in Liposomes
The clearest synthetic analogue for par-
12.4.1 Powder Formation
ticle formation in a vesicle within a cell is
Most oxide ceramic powders are formed particle formation within liposomes. These
by a solution precipitation route, followed closed spherical shells of lipid can be
by a calcination step to remove water and formed by ultrasonic treatment of a small
then extensive milling to reduce the aver- amount of a natural phospholipid such as
age particle size into the micrometer re- phosphatidylcholine in a solution of salts
gion. Nanoparticles can be formed by pre- (Bhandarkar and Bose, 1990; Bhandarkar
cipitation from dilute solution or gas reac- et al., 1990). The two hydrocarbon tails on
tions followed by collection at low temper- the lipid favor formation of a bilayer
atures to prevent agglomeration. rather than a micelle (Fig. 12-7). As the
Fine scale carbide and nitride particles shells form, some solution is trapped in-
can be made by gas-phase reactions, where- side. The liposomes can be separated by
as melt processes give much larger particle centrifugation from the solution and trans-
sizes. ferred to a new aqueous medium. The lipid
The attraction of biomimetic powders is layer is somewhat permeable to hydroxyl
summarized by the picture of magnetite ions, which increase the internal pH and

(a)

Figure 12-7. Schematic arrangement of

phospholipid (a) into a bilayer vesicle
(b) with the anionic head groups facing
towards the inner and outer aqueous
phase.
68 12 Biomimetic Processing

induce precipitation of oxides inside the Liposomes have been made containing
liposome. Bose and co-workers (Bhandar- up to 0.5 M solutions of A1C13 and higher
kar and Bose, 1990; Bhandarkar etal., concentrations are certainly possibly. It is
1990) produced cobalt ferrite and alumina. not clear how concentrated a solution can
Mann and Hannington (1988) made silver be maintained inside a vesicle, how con-
oxide, goethite and magnetite. Particle centrated a solution can be used to sur-
sizes are in the range of 10-50 nm. Oxide round vesicles or how stable vesicles will be
particles form in the aqueous phase inside in high-number densities in suspension.
the vesicle, whereas hydrophobic cadmium These factors will all have a significant im-
and zinc sulfides form within the lipid pact on the viability of this process for
membrane (Heywood et al., 1990). powder production.
To view this as a process for making A 0.5 M solution of aluminum chloride
particles we need to attain a number of represents a final solid volume fraction of
goals. The liposomes must be inexpensive alumina within the vesicle of 0.7% assum-
to make, they must reach a high volume ing that the particle has full density. One
fraction of solids, they must be stable at way to reach higher densities is to come
high-number densities, the particles must closer to the biological system and develop
be dense and preferably crystalline, and a method for continuous uptake of metal
the process must be reasonably fast. Bose by the liposome. The requirement would
and co-workers (Bhandarkar and Bose, be fulfilled by a metal complex that was
1990; Bhandarkar et al., 1990), stress that stable (or metastable) and soluble in the
one advantage of this method is that co- mother liquor, that had sufficient lipid sol-
precipitation is limited to a very fine scale ubility to be transported through the bi-
by the size of the individual vesicles. Hence layer and that was destabilized by condi-
mixed oxides, such as barium titanate, can tions inside the liposome. This is a viable
be made with a high level of homogeneity goal but it is not clear exactly how it will be
instead of segregation when one metal achieved.
compound hydrolyzes before the other. The particles which form within the vesi-
Natural phospholipids are relatively ex- cles are often single crystals but may be
pensive because they are extracted from quite disordered; the latter have not been
egg yolk and extensive purification is fully characterized. More control of the
needed. For medical uses of liposomes, process should yield more dense and con-
these high costs are not prohibitive. The sistent particles, especially if temperatures
availability of cosmetics based on lipo- closer to 100 °C can be used so that the
somes has promise for their use in materi- conditions become more similar to those
als synthesis. Synthetic vesicle-forming used by Matijevic to form dense particles
compounds are available, with two long from dilute solutions. The slow kinetics of
hydrocarbon tails attached to a sulphonate ion transport through the lipid membrane
or phosphate ion (Fendler, 1984). Innova- may lead to more stable products than
tive work has shown that mixtures of those produced from solution as the pre-
cationic and anionic synthetic detergents cipitation process occurs much closer to
can spontaneously form liposomes (Kaler equilibrium as a result.
et al., 1989). This promises to reduce the
cost considerably.
 12.4 Biomimetic Processing 69

12.4.1.2 Other Synthetic Routes under slightly basic conditions. It is be-

to Powders lieved that groups inside the protein cata-
lyze the oxidation and precipitation.
A number of related approaches are also A further alternative is to create a poly-
being investigated. Alkoxides may be hy- meric or gel mold to grow the particle in.
drolyzed in oil-in-water or water-in-oil Cross-linked dextran or ionic polystyrene
(reversed) emulsions (Hardy etal., 1988). beads have been swollen with water and
These systems lack the great stability of then reacted with silicon alkoxides to pro-
vesicles and thus tend to form agglomerat- duce ceramic particles, although the scale
ed powders. Microemulsions are stable is large, 10 jim-1 mm (Hardy et al., 1990).
phases which form at high surfactant con- The process may be adaptable to smaller
centrations. Lamellar and rod phases exist particles, such as emulsions of polymer
at suitable water: oil: surfactant ratios. elastomers.
The latter have been used to form oxide In a related process we have developed a
particles (Inouye et al., 1982). As with con- method for forming a polymeric "mould"
ventional emulsions, particle agglomera- which can be used to grow a shaped ceram-
tion may occur if hydrolysis takes place in ic particle (Burdon and Calvert, 1992). A
the oil phase, with the water moving to the two-phase polymer film is cast and drawn
alkoxide rather than vice versa. Also, these in a way that the discontinuous phase is in
systems are more dynamic than vesicles the form of elongated rods. This film can
and particles in separate aqueous compart- be swollen with titanium alkoxide which
ments can still agglomerate. Particles may preferentially dissolves into the elongated
be stabilized after growth and before sepa- polymer particles. On subsequent hydrolyt-
ration by a surface treatment as has been ic treatment of the film, elongated particles
done for cadmium selenide (Steigerwald of titanium dioxide are formed (Fig. 12-8).
etal., 1988). The shape of these particles can be mod-
One goal of this work is to develop a ified by different drawing treatments on
method that yields a fluid containing 30- the two-phase polymer. A similar process
50 vol% of uniform submicrometer parti- has been used by Mark and co-workers
cles which are stable against agglomera- (Wang and Mark, 1990) to elongate parti-
tion. Such a suspension could be used di-
rectly for casting ceramic green bodies.
These methods are promising but still have
some way to go.
Ferritin is a protein comprising 24 sub-
units which form a cage around a cluster of
iron oxide with a diameter of 8-9 nm, con-
taining 2500 iron atoms. The iron can be
removed to yield the empty protein shell,
apoferritin. Meldrum etal. (1991) have
used the apoferritin shell to synthesize par-
ticles of iron sulphide, manganese oxide
and uranyl oxyhydroxide. Iron oxide and
manganese dioxide can be formed in ferrit- Figure 12-8. Elongated titania particles grown in an
in by oxidation of Fe11 or Mn11 solutions oriented polymer matrix which has been burnt off.
70 12 Biomimetic Processing

cles of one polymer within another. If the 12.4.2 Bulk Parts

polymer is burnt out and the particles
calcined, they form rods, each of which 12.4.2.1 Fibrous and Laminated Ceramics
is made of fine-grained titania. As they
have been produced so far, the rods are In the first part of this review we extract-
internally porous since they arise from a ed some lessons from biological materials
T i O 2 x H 2 O gel. The requirements for regarding the toughness of ceramic lami-
production of ceramic rods by this ap- nates and fibrous ceramics with small re-
proach would be fulfilled by a system sidual polymer contents. Several groups
which swells to a higher volume fraction of are studying methods for reproducing such
alkoxide, and a set of conditions where the structures though not all of them would
initial inorganic precipitate has a higher necessarily describe their work as bio-
density. These criteria should be achiev- mimetic.
able by adjusting the polymer-alkoxide Among the attributes of shell or tooth
compatibility and the treatment condi- that we would want to duplicate are a
tions. small volume fraction, but continuous,
polymeric phase, a submicrometer grain
12.4.1.3 Bacterial Fermentation
size, elongated particles, and orientation
Bacteria have long been fermented to parallel to the exposed surface.
produce water-soluble polymers. In recent Aksay and co-workers have described a
years, polyhydroxybutyrate has been pro- method for making laminates of boron
duced on a small commercial scale by ICI. carbide and aluminum in which B4C is tape
This fiber-forming plastic has properties cast, partly sintered and the pores back-
similar to polypropylene and is biodegrad- filled with aluminum (Yasrebi et al., 1990).
able. Many families of bacteria form the Very high toughness (Klc = 14 MPa m 1/2 )
polymer as intracellular granules for car- is claimed for the resulting laminates, with
bon storage, as root vegetables store a strength up to 950 MPa at 30 vol% alu-
starch. Under suitable conditions 80% of minum.
the dry weight of the cell can be polymer. The apparent problem with an infiltra-
The formation of a true polymeric material tion approach is that there must be a con-
by fermentation implies that inorganic tinuous inorganic phase before infiltration
particles might also be manufactured com- and this may provide an easy crack path
mercially. which requires little deformation of the
In the absence of any detailed knowl- tough matrix. Infiltration is also likely to
edge of particular growth cycles and their be least successful in the best structures
dependence on conditions, it seems reason- where the ceramic particles are tightly
able to assume that cycle times could be of packed.
the order of a few hours to a day, that cell Clegg etal. (1990) made laminated
densities will be about 1012 per liter in or- structures from 200 Jim thick sheets of sili-
der to allow a good supply of nutrients, con carbide separated by graphite. The fi-
and that ultimate production costs might nal sintered structure bears a close resem-
parallel those of polyhydroxybutyrate, blance to nacreous shell, although the scale
which currently sells for about $ 15/lb but is 400 times larger. The three-point-bend
might be reduced to $ 3/lb with large-scale strength of the composite was 633 MPa
production. and the fracture toughness Klc was 15 MPa
 12.4 Biomimetic Processing 71

m 1/2 . The average work of fracture was Folsom et al. (1992) have made com-
4625 J m ~ 2 compared to 6 2 J m ~ 2 for posites of dense alumina sheets, 630 |im
monolithic silicon carbide. Further work thick, with layers of carbon fiber-epoxy of
by Clegg (1992) shows that the properties about 100 |im thick. Fracture in three-
of the laminate can be treated as the se- point-bending was catastrophic, with little
quential failure of a series of unnotched delamination, when the crack length was
beams of silicon carbide as each lamina parallel to the carbon fibers. When the
fractures independently. The thickness of crack was across the fiber axis, a sequence
the graphite interface has little effect above of lamina fracture, crack arrest, delamina-
3 jLim. With thinner graphite layers the tion, and further lamina fracture was seen,
crack propagates across the interface from similar to that found by Clegg et al. (1990).
the tip of the delamination crack, possibly The initial fracture stress was limited by
due to bridging of the silicon carbide that of the alumina.
through the graphite layer. Oxidation lim- Clegg (1992) finds a ratio of the fracture
its the temperature resistance of this struc- energy of graphite-silicon carbide interface
ture to 600 °C in air. Similar structures to that of silicon carbide as 0.18. This com-
have been made from extruded 150 jim zir- pares with calculations by Kendall (1975)
conia fibers and have a work of fracture 10 that suggest a ratio of 0.1 or less is needed
times that of monolithic zirconia. for crack deflection. Evans (1988) derives a
Halloran and co-workers (Baskaran value of 0.25. Also the friction coefficient
etal., 1993) have formed green ceramic should be less than 0.1. For fibers with a
fibers which are then coated with graphite reasonably sharp strength distribution, the
and dry-pressed. The resulting green part composite toughness is expected to in-
is burnt-out and sintered under reducing crease with increasing fiber diameter. The
conditions. The polymeric binder is lost same would be true for short fibers or
from the green fibers which then sinter to plates of constant aspect ratio and volume
polycrystalline rods surrounded by a weak fraction, the composite becomes tougher
graphite interface. High fracture tough- as the fiber size increases. This is contrary
nesses are observed. The advantage of this to a natural prejudice to assume that finer
green fiber approach is that the fibers de- microstructures will be better and it has
form during dry-pressing to leave a very not been tested for fibers in the micrometer
thin graphite layer, A dense fiber would and submicrometer range. Not taken into
leave relatively large zones of weak matrix account is the increase in fiber strength
surrounding each fiber. that is expected to come from the decrease
A fibrous composite has been derived in intrinsic flaw size with decreasing di-
from polytitanocarbosilane precursor fibers ameter.
(Yamamura et al., 1989). Woven fabrics of Biomimetic structures have been pre-
green fiber were hot-pressed, resulting in a pared by starting from coral which is an
dense ceramic of close packed hexagonal interpenetrating network of holes and cal-
fibers. A tensile strength of 400 MPa was cite (White etal., 1972; Skinner etal.,
reported for a density of 90 % of the theo- 1978). The organic residue is removed and
retical value. Stress-strain curves showed the holes filled with resin, then the carbon-
complex failure in bending with a 2 % ate can be dissolved out and replaced with
strain to final failure and presumably a metal or ceramic. The pores in coral are in
high work of fracture. the range of 20-200 |im. The predominant
72 12 Biomimetic Processing

application for this type of material lies in parency may also be promoted by the close
the area of bone replacement since the refractive index match between silica and
large porosity allows cells to invade the many polymers. The morphology of these
structure. materials is unclear.
In summary, composites with structures Wilkes and co-workers (Huang et al.,
analogous to nacre have been synthesized 1992) have prepared a series of combina-
and have been shown to exhibit improved tions of silica and titania with polytetra-
toughness. The morphology is on the scale methylene oxide (PTMO) and with poly-
of hundreds of micrometers rather than dimethyl siloxane (PDMS). These polymer
0.5 micrometer but it is not certain whether systems are of especial interest since they
this will make a difference. So far these have glass transition temperatures (Tg)
materials are all highly anisotropic. How- well below room temperature. The result-
ever, isotropic materials formed from bun- ing high mobility at room temperature
dles of plates or fibers should be possible. may be important for retaining toughness
at room temperature in the composites. In
each case the polymer has a relatively low
12.4.2.2 Polymer-Ceramic Composites molecular weight and is end-capped with
Polymer-ceramic composites can be alkoxide groups. During the hydrolysis of
made by a variety of direct-reaction pro- the alkoxide, the polymer becomes cou-
cesses. Existing ceramic particles can be pled into the inorganic network. These
dispersed into resin, particles can be dis- materials tend to be transparent, implying
persed into monomer which is polymer- that phase separation occurs on a very fine
ized, ceramic precursors can be dispersed scale. Fractal analysis has been used with
into resin and reacted, or precipitation and small angle X-ray scattering data (Rodri-
polymerization can occur simultaneously. gues et al., 1992). The interparticle scatter-
Polymers have long been filled with ceram- ing peak (at about 10 nm) moves to lower
ic particles to increase the modulus and angles and becomes more pronounced
yield strength and to decrease creep at ele- with increasing molecular weight of the
vated temperatures (see Sec. 9.4 in Vol- organic component.
ume 13 of this Series). Generally these im- Dynamic mechanical measurements of
provements are accompanied by a reduc- modulus for the PTMO ceramers show a
tion in toughness which becomes close to glass transition temperature at -70 °C
zero at about 50 vol% particles. In this which does not change much as silica is
regime of low volume fractions of filler, we incorporated (Brennan and Wilkes, 1991).
are clearly dealing with hard plastics rather However, as the silica content increases be-
than with tough ceramics. yond 20% (60-70 wt% tetraethoxy silane
Combinations of polymers with sol-gel in the starting resin), the modulus decrease
glasses (variously named polycerams, cer- at the glass transition is lost and the mod-
amers, ormosils and ormocers) have been ulus at room temperature becomes about
prepared by several groups. The polymer 1 GPa. The strain-to-failure decreases from
allows the system to fully densify at room about 70% in unfilled polymer to 20% in
temperature. Transparency is achieved by the ceramers. The properties are quite sen-
incorporating silicon alkoxide groups onto sitive to the extent of alkoxide functionali-
the polymer so that the two phases remain ty of the polymer, to the molecular weight
intimately mixed. In silica systems, trans- of the polymer, and to the curing condi-
 12.4 Biomimetic Processing 73

tions. Thus the properties of these materi- Similar methods have been used with a
als are very sensitive not only to composi- wide range of polymers and a range of
tion, but also to morphology and degree of inorganic precipitates. Mark (Zhao et al.,
connectivity between the phases. 1993) has swollen poly dimethyl siloxane
Fitzgerald et al. (1992) have incorporat- (silicone rubber) with TEOS vapor and has
ed silica into a series of acrylic polymers found a progressive modulus increase to
and polyvinylacetate with similar results. 24 MPa at 33 vol% silica. The strain to
Instead of dropping rapidly with tempera- failure reduces from above 40% to 8%.
ture, the modulus above Tg levels off at Polyphosphazenes with acid catalyzed
10-100 MPa at silica contents of 15 wt%. TEOS have been converted from fluids to
The behavior is complicated by further tough plastics with a modulus of 1 GPa
curing of the gel and changing bonding to (Coltrain et al., 1992).
the polymer at temperatures above 100 °C. The scale of phase separation in polyce-
At still higher levels, above 27 vol% silica, rams must depend on the compatibility of
the material becomes unworkably brittle the polymer with partly polymerized
(Landry et al., 1992). Acid catalysis of the TEOS. Thus, David and Scherer (1991)
hydrolysis of tetraethoxysilane (TEOS) have shown that transparent composites
generally resulted in transparent materials can be made if a compatible polymer is
while base catalysis gave white composites. chosen. In addition the polymer may be
This reflects the fact that base catalysis of modified to carry sufficient silicon alkox-
TEOS hydrolysis gives large particles ide groups so that intimate mixing is fa-
while acid hydrolysis gives a transparent vored. Novak and Ellsworth (1991) have
gel with particles of a few nanometers in produced transparent composites by ar-
diameter. The opaque base-catalyzed sam- ranging that the organic polymerization
ples did not show a high modulus above and silica gelation occur simultaneously
Tg, showing that fine scale phase separa- and rapidly. Furthermore, the shrinkage
tion is required to restrict polymer motion. of these materials can be greatly reduced
Pope et al. (1989) made a similar com- by use of the silicon alkoxide of a polymer-
posite by impregnating porous sol-gel sili- izable alcohol, such as hydroxyethyl-
ca with methylmethacrylate (MMA) and methacrylate, so that the alcohol becomes
polymerizing it. In this case, they found a part of the polymer network (Novak and
simple rule of mixtures relationship be- Davies, 1991).
tween volume fraction of polymer and Schmidt etal. (1988, 1992) have devel-
both compressive strength and 4-point oped hard coatings for polymers based on
bend strength. The elastic modulus fell be- polymer-ceramic hybrids which cure at
tween the Hashin-Shtrikman bounds and low temperature.
was about 30 GPa at 50 vol% PMMA (see Having established that polymer-silica
Sec. 9.5.1.1 in Volume 13 of this Series). A hybrids can be made, questions arise about
similar system prepared by Abramoff and whether we can extend the chemistry to
Klein (1992) gave a modulus of 14 GPa other oxides and non-oxide systems and
and strength of 132 MPa at about 45 vol% whether more complex morphologies can
silica. These composites differ from those be synthesized.
of Landry et al. (1992) in that there should Much work has been done on formation
be two interpenetrating phases rather than of sulfides in various matrices. Spanhel
a continuous polymer phase. et al. (1992) have grown CdS and CdS-PbS
74 12 Biomjmetic Processing

in ormocers. Bianconi et al. (1991) have be introduced. It is believed that mica lay-
grown cadmium sulfide in polyethylene ers are separated by a single layer of poly-
oxide films and have shown how the parti- mer chains. To exploit this type of interca-
cle morphology is controlled by the poly- lation chemistry for ceramics, one would
mer matrix. need to be able to introduce a wide range
Magnetic iron oxides have been formed of polymers between silicate layers of vary-
in polyimides (Bergmeister and Taylor, ing thickness. It is not clear that this type
1992; Madeleine et al., 1988). Cross-linked of control is available to allow general ap-
sulfonated polystyrene beads have been plication, though the materials may have
impregnated with y-iron oxide by oxida- interesting properties as sensors. A related
tion of absorbed Fe(n) (Ziolo et al., 1992). approach is to grow a layered structure of
This precipitate is of interest because the calcium aluminate in the presence of poly-
small particle size leads to increased trans- vinyl alcohol (Messersmith and Stupp,
parency of the iron oxide. Similarly, cellu- 1991).
lose and paper can be impregnated with Okada et al. (1987, 1991) have polymer-
iron salts which are hydrolyzed to the ized caprolactam, the monomer of nylon 6,
oxide (Marchessault et al., 1992). Nandi in the presence of montmorillonite clay.
et al. (1990) have made chromium oxide The lactam intercalates and separates the
and ferrihydrite in polyimide films by ther- clay sheets, giving 1 nm thick clay sheets
mal decomposition of chromium or iron dispersed in a nylon matrix. With 5 wt%
carbonyl. clay the polymer showed an increase in
Metals can be precipitated into polymer modulus from 1 - 2 GPa and in strength
films to make composites by reduction from 70 to 110 MPa with no decrease in
(Mazur etal., 1989; Calvert and Broad, impact strength. This would be a very in-
1989). teresting approach, if these improvements
The studies reviewed above all involve continue to higher volume fractions of
homogeneous amorphous polymers that mineral. Yano et al. (1991) have swollen
impose little constraint on the sol-gel reac- the layer structure of montmorillonite with
tion. Mauritz and Warren (1989) swelled precursors of polyimides to produce com-
Nafion membranes with TEOS and hydro- posites which again show significant prop-
lyzed it. The two-phase membrane might erty changes with small amounts of clay.
be expected to incorporate silica into the
hydrophilic phase between the crystalline
12.4.2.3 In Situ Mineralization
fluoropolymer regions. Studies of dielec-
of Polymers
tric properties and small angle X-ray dif-
fraction have not yet produced a clear To translate the biological processes into
model for the morphology (Mauritz and physical terms, composites can be formed
Stefanithis, 1990). by the precipitation of reinforcing particles
Another approach to polymer-ceramic into a pre-existing matrix. By controlling
composites has been developed by Gianel- the production of the precipitants and the
lis (1992) and Gianellis etal. (1990). Lay- properties of the matrix we should be able
ered silicates such as talc or mica can be to form composites with shaped and ori-
intercalated with aniline which is then con- ented particles, packed to high enough vol-
verted to poly aniline. Alternatively, polar ume fractions to be hard composites or
polymers, such as polyethylene oxide can ceramics. In the preceding discussion of
 12.4 Biomimetic Processing 75

polycerams little attention has been given cipitated into a polymeric matrix if it is
to using the polymer to really control the formable at temperatures below about
form of the ceramic phase. 300 °C. However, there are frequently
We have approached this goal by start- compatibility problems that must be re-
ing from simple precipitations in organic solved for any particular combination of
polymer matrices and have explored meth- polymer and particle.
ods of exerting increasing amounts of con- Having explored the types of precipi-
trol over the resulting morphology. tates that can be formed, our recent work
One system of interest is the precipita- has looked at control of particle size and
tion of organic solutes from supersaturat- shape in the precipitation of titania, silica
ed solution in a polymer. This has long and zirconia from the alkoxides. For the
been a concern in the form of "blooming" precipitation of an amorphous particle,
- surface crystallization - of antioxidants the particle size is controlled by a combina-
from polymers. We incorporated nitroani- tion of polymer mobility and solubility of
lines with high optical second-harmonic the precipitating species. If phase separa-
coefficients into various polymers with a tion of a liquid precedes formation of the
view to producing transparent composites inorganic particle, the scale of the precipi-
that would combine high optical activity tate is typically a few micrometers. If the
with polymeric toughness. Rather than precursor remains soluble until reaction
aiming for a homogeneous solution we felt occurs, then we find a submicrometer pre-
that a combination of very fine scale crys- cipitate and a transparent composite re-
tallization with a highly oriented structure sults.
should give good transparency. Very high- Highly elongated titania particles can be
ly-oriented organic crystals can be grown formed by drawing of the polymer matrix
in polymers either by melting a composite during the precipitation process (Fig. 12-9).
film and cooling it by slowly pulling We believe that we are then elongating
through a temperature gradient, or by pre- partly hydrolyzed gel particles. We have
drawing the polymer before allowing the been seeking for a demonstration of elon-
solute to crystallize. Drawing a polymer gated particle formation in a drawn poly-
that contained nitroaniline crystals did not
result in orientation of the precipitate, ap-
parently the crystals just broke up (Azoz
et al., 1990; Calvert and Moyle, 1988).
We have since explored methods to form
a range of inorganic precipitates in poly-
mers, including metal oxides by hydrolysis
of alkoxides, iron oxides from iron chlo-
ride and metals by reduction (Calvert and
Broad, 1989). Other groups have devel-
oped methods for the formation of sul-
fides, for depositing metal films internally
in polymers and for laser "writing" of
metal lines into polymers (Bianconi et al.,
1991; Madeleine et al., 1988; Mazur et al., Figure 12-9. Elongated titania particles in a matrix of
1989). In essence, any material can be pre- polyvinyl chloride.
76 12 Biomimetic Processing

mer matrix to show that anisotropy of the

matrix can impose anisotropy on the parti-
cles forming within it. This has not yet
been successful. We have used drawn poly-
mer blends, of polymethylmethacrylate and
polyvinylidene fluoride, to induce elongat-
ed particle formation. Here the oxide pre-
cipitate forms preferentially within the
acrylate-rich regions and the elongated
shape of these results in elongated particles
(Burdon and Calvert, 1992).
A further goal, which we are working
towards, is to demonstrate that a polymer
Figure 12-10. Structure of siliceous diatoms: (A) and
matrix can catalyze precipitation of an
(C) whole shells, (B) and (D) fine structure of pores in
oxide. For instance, the polymer could shells. Scale bars: A: 1.7 urn, B: 105 nm, C: 2.8 urn, D:
provide a localized highly basic environ- 110 nm. Note the pore size of about 0.1 um. (From
ment to promote the condensation reac- Mann and Perry, (1986).)
tion of silicic acid to silica (Burdon and
Calvert, 1991).
In this work we are hoping to show that surfaces which have been functionalized
a totally non-biological system can be de- with organic groups. Dense thin films were
veloped to form composites with an equiv- formed. By using lithographic techniques
alent degree of structural control to that to modify only some parts of the surface,
seen in mineralized tissues. patterns can be produced. The general re-
quirements for this type of layer growth
are not yet understood. The solutions used
12.4.2.4 Controlled Mineralization
in these treatments generate the precipitat-
By use of photosensitive salts it is possi- ing species by slow reaction, either hydrol-
ble to precipitate patterns of metals into ysis of iron salts or chemical formation of
polymer films. Decomposition of the salt is sulphide. The treated surface may catalyze
induced by a laser (Auerbach, 1985). It these reactions and produce locally high
would also be desirable to grow multilayer concentrations. Alternatively the surfaces
patterns of ceramic on polymer or other may act as nucleating agents or adsorb
substrates. This could be used to make nuclei which form in solution. Presumably
devices or to deposit protective coatings. the process is generizable to a wide range
The biomimetic approach is to use solu- of surfaces and precipitates.
tion deposition and activated regions on A similar and intriguing effect has been
the substrate rather than vapor phase described by Kokubo and co-workers (Abe
deposition. Silica structures in diatoms, et al., 1990). Apatite coatings are grown
grown from solution, show levels of reso- from solutions containing calcium and
lution comparable to those currently at- phosphate in the presence of a plate of
tainable by photolithography (Fig. 12-10). glass ceramic comprising calcium and sili-
Rieke and co-workers (1992) have de- cate. Apatite layers form on any substrate
posited iron oxide and cadmium sulphide which is placed parallel to the glass ceram-
on treated polymer surfaces and on silicon ic, with a gap of less than 1 mm. It is sug-
 12.6 Advanced Concepts 77

gested that silica from the glass ceramic cation for biomimetic coatings. The same
adsorbs to the substrate and induces nucle- arguments apply to barrier coatings. There
ation of the apatite. Once the apatite layer are no good polymer barriers to oxygen,
is growing, the glass ceramic plate can be liquid hydrocarbons or many other sol-
removed. It is known that mineralizing vents. A coherent layer of silica or another
bone contains high levels of silicon but no oxide would be very desirable, especially if
reason for this has yet been found. it could be deposited within the polymer.
As was brought out in the discussion of
the properties of bone, there is currently a
12.5 Applications gap between the mechanical properties of
moldable filled-polymer composites and
Some materials now being used as med- those of ceramics. Biological materials,
ical prosthetics could be justifiably called from antler to enamel, bridge this gap and
biomimetic. Otherwise there is no current so one can expect synthetic materials to be
commercial ceramic product that could re- developed. Stiff, tough and formable plas-
ally be so called. A number of applications tic panels with properties comparable to
should be viable with the knowledge and bone would be expected to find many ap-
techniques that we have now. Biomimetic plications where sheet metal is now used.
materials, with structures resembling those Continuous fiber composites do have
of biological materials, could be made by many of the right properties but are too
conventional processing methods. expensive to process.
Thin films of ceramics can be made by in Shell, and possibly tooth enamel, rely on
situ precipitation of particles into poly- the polymer content for their toughness. In
mers followed by firing (Calvert and synthetic applications this would limit the
Broad, 1990). Such processing methods temperature range of biomimetic ceramics
could, in principle, be applied to ceramic to below 300 °C. Many applications of ce-
multilayers such as capacitors and elec- ramics for wear-resistance would be suit-
tronic packaging. The main advantages able. A polymer layer is only tough if it is
would stem from the fine particle sizes, above, or close to, its glass transition and
absence of agglomerates and purity of liq- is below the decomposition temperature. It
uid starting materials. The large disadvan- is not clear whether it would be possible to
tage is that the laboratory method would devise bonding materials with windows of
require considerable development to be- toughness in higher temperature ranges.
come a commercial process. These com- The example of jade also shows that a
ments apply to most applications where good morphology can give enhanced
biomimetic methods may be better but ex- toughness even without a polymer binder.
isting methods do work.
One area where existing methods are
very limited is in barrier coating of plastic 12,6 Advanced Concepts
film and sheet. Wear resistant and scratch
resistant coating for plastics (e.g., Schmidt, At the more speculative level, we can
1992) exist but are not adequate. High expect to be able to build components with
temperature routes destroy the polymer; complex hierarchies of structure akin to
evaporated coatings are either very thin or those of bone or cartilage (Baer et al.,
not coherent. This should be a good appli- 1991). Thus, a composite component could
78 12 Biomimetic Processing

have regions of pure ceramic and others of cellent. Composites of polymer reinforced
pure polymer with gradients of properties with silica or carbon fiber have also been
in between and no interfaces. Orientation made, including silica formed by an in situ
of reinforcing fibers and controlled porosi- reaction (Stuffle etal., 1994). This and
ty could be used to modulate the proper- other SFF methods will provide a new
ties. Such a material would have to be route to making tough structures resem-
made by biomimetic processing, using bling those of tooth and shell.
chemical methods for forming compo- In addition to pursuing the desirable
nents in situ and growing the structure mechanical properties of biological mate-
rather than making it in a single action. rials, there is much interest in the possibil-
The obvious way to construct such a ity of incorporating intelligence in the
part is to build it up layer by layer, either form of sensors and actuators and the
by the methods now used for multilayer means to repair damage or to remodel.
ceramics or by using techniques akin to These goals require the ability to deliver
stereolithography (Lauder et al., 1992). signals or reagents to any point within the
A number of methods for solid freeform material and so presuppose some equiva-
fabrication (SFF) have been developed re- lent of a nervous system or a blood supply.
cently. Each is based on the delivery of
energy and/or material to a point under
the control of a three-dimensional CAD 12.7 Conclusions
design. The point sweeps out a layer and a
series of layers build up a solid. The mate- Biological materials have mechanical
rial starts as a mobile fluid or powder and properties which are superior to those of
is converted point-by-point to a solid. Ide- their synthetic equivalents, at least when
ally this cure time should be less than a compared on some basis that allows for
minute. At one minute per layer, a part is the intrinsic weakness of the chemical
produced in a few hours. While this sounds structures of calcium carbonate and phos-
like an impracticably long time, some ma- phate. Structural tissues achieve good frac-
chines will fit on a desk top and thus lend ture toughness in combination with high
themselves to distributed and customized levels of stiffness. This combination of
manufacture. Although SFF methods properties has its origin in the morphology
have so far found relatively minor applica- of the inorganic and organic phases.
tions in manufacturing, they have the po- There exist methods which should allow
tential to change the way in which we re- us to produce similar microstructures in
gard the relationship between the chemis- synthetic ceramics and composites. We can
try and processing of materials, especially also develop processing methods which are
in regard to composites. analogous to the processes of biological
Over the last two years we have devel- growth and mineralization. In particular
oped an SFF system based on the extru- these methods will depend on low temper-
sion of material from a syringe fitted with ature chemical processing rather than con-
a fine (0.3 mm) needle. Materials produced ventional thermal processing. These meth-
include alumina from a slurry of powder in ods should also allow structure and prop-
acrylic monomer which is thermally poly- erty gradients to be built into components
merized as it is deposited (Stuffle et al., so that single materials can replace jointed
1993). The mechanical properties are ex- assemblies.
 12.9 References 79

Beyond this first set of biomimetic mate- Bigg, D.M. (1987), Polym. Compos. 8, 115-122.
rials, we can work to develop materials Bonfield, W, Grynpas, M.D., Tully, A. E., Bowman,
X, Abram, J. (1981), Biomaterials 2, 185-191.
with hierarchical composite structures that Boyan, B. D., Swain, L. D., Everett, M. M., Schwartz,
incorporate aspects of intelligence and self- Z. (1992), in: Calcification in Biological Systems:
repair. Bonucci, E. (Ed.). Boca Raton: CRC Press,
pp. 129-156.
Bradt, R.C., Newnham, R.E., Biggers, J.V. (1973),
Am. Mineral 58, 727-732.
12.8 Acknowledgements Brennan, A.B., Wilkes, G.L. (1991), Polymer 32,
111-178.
Brooker, B.E. (1979), in: Microbial Polysaccharides
I would like to thank the following orga- and Polysaccharases: Berkeley, R.C.W., Gooday,
nizations for support of our work on G. W, Ellwood, D.C. (Eds.). New York: Academic
biomimetic materials: Courtaulds Perfor- Press, pp. 85-116.
Burdon, J.W, Calvert, P. (1991), Mater. Res. Soc.
mance Films, Elf Atochem, US Air Force Symp. Proc. 218, 203-212.
Office of Scientific Research, US Army Burdon, J.W, Calvert, P. (1992), Hierarchically
Research Office. Funding from the BP Structured Materials, Mater. Res. Soc. Symp. 255:
Aksay, I., Baer, E., Sarikaya, M., Tirrell, D.A.
Venture Research Unit in 1985 enabled us (Eds.), pp. 375-383.
to start this research program. Calvert, P., Broad, R. A. (1990), Materials Synthesis
Utilizing Biological Processes, Mater. Res. Soc.
Symp. 174: Rieke, P.C., Calvert, P.D., Alper, M.
(Eds.) pp. 61-67.
12.9 References Calvert, P., Broad, R.A. (1989), in: Contemporary
Topics in Polymer Science, Vol. 6: Culbertson,
Abe, Y, Kokubo, T., Yamamuro, T. (1990), J. Mater. W.M. (Ed.). New York: Plenum.
Sci.: Mater. Med. 1, 233-238. Calvert, P.D., Mann, S. (1988), J. Mater. Sci. 23,
Abramoff, B., Klein, L.C. (1992), in: Ultrastructure 3801-3805.
Processing of Advanced Materials: Uhlmann, Calvert, P.D., Moyle, B.D. (1988), Mater. Res. Soc.
D.R., Ulrich, D.R. (Eds.). New York: Wiley, Symp. Proc. 109, 357-362.
pp. 401-407. Calvert, P., Dieppe, P. (1982), Crystals and Joint Dis-
Addadi, L., Weiner, S. (1985), Proc. Nat. Acad. Sci. ease. London: Chapman and Hall.
USA 82, 4110-4114. Clegg, WJ. (1992), Ada Metall. Mater. 40, 3085-
Addadi, L., Moradian, J., Shay, E., Maroudas, N. G., 3093.
Weiner, S. (1987), Proc. Nat. Acad. Sci. USA 84, Clegg, W.J., Kendall, K., Alford, N. McN., Button,
2732-2736. T. W, Birchall, J.D. (1990), Nature 347, 455-457.
Auerbach, A. (1985), J. Electrochem. Soc. 132, 1437- Coltrain, B.K., Ferrar, W.T., Landry, C.J.T., Mo-
1440. laire, T.R., Zumbulyadis, N. (1992), Chem. Mater.
Azoz, N., Calvert, P. D., Kadim, M., McCaffery, A. J. 4, 358-364.
Seddon K. J. (1990), Nature 344, 49. Constanz, B. (1986), Paliaos 1, 152-157.
Baer, E., Cassidy, J. J., Hiltner, A. (1991), Pure Appl. Crenshaw, M.A., Linde, A., Lussi, A. (1987), in:
Chem. 65,961-973. Atomic and Molecular Processing of Electronic and
Baskaran, S., Nunn, S., Popovic, D., Halloran, J.W. Ceramic Materials: Aksay, LA., McVay, G.L.,
(1993), J. Am. Ceram. Soc. 76, 2209-2216, 2217- Stoebe, T. G., Wager, J. F. (Eds.). Pittsburgh: Mate-
2224. rials Research Society, pp. 99-107.
Behri, J.C., Bonfield, W. (1980), /. Mater. Sci. 15, Currey, I D . , Kohn, A.J. (1976), J. Mater. Sci. 11,
1841-1849. 1615-1623.
Bergmeister, J.J., Taylor, L.T. (1992), Chem. Mater. Currey, J.D. (1984), Phil. Trans. R. Soc. London B
4, 729-737. 304, 509-518.
Berman, A., Addadi, L., Kvick, A., Lieserowitz, L., Daculsi, G., Pouezat, I, Peru, L., Maugars, Y,
Nelson, M., Weiner, S. (1990), Science 250, 664- LeGeros, R.Z. (1992), in: Calcification in Biologi-
667. cal Systems: Bonucci, E. (Ed.). Boca Raton: CRC
Bhandarkar, S., Bose, A. (1990), J. Colloid Interface Press, pp. 365-397.
Sci. 139, 541-550. Dalas, E., Kallitsis, J.K., Koutsoukos, P.G. (1991),
Bhandarkar, S., Yacob, I., Bose, A. (1990), Mater. Langmuir 7, 1822-1826.
Res. Soc. Symp. 180, 637-641. Dameron, C.T., Reese, R.N., Mehra, R.K., Kortan,
Bianconi, P. A., Lin, J., Strzelecki, A. (1991), Nature A. R., Carroll, P. X, Steigerwald, M. L., Brus, L. E.,
349, 315-317. Winge, D.R. (1989), Nature 338, 596-597.
80 12 Biomimetic Processing

David, I. A., Scherer, G.W. (1991), Polym. Prepr. 32, Kaler, E.W, Murthy, A.K., Rodriguez, B.E.,
530-531. Zadasinski, J.A.N. (1989), Science 245, 1371-
Doyle, C , Tanner, E.X, Bonfield, W. (1991), Bioma- 1374.
terials 12, 841-847. Keith, X, Stockwell, S., Ball, D., Remillard, K.,
Eanes, E.D. (1992), in: Calcification in Biological Kaplan, D., Thannhauser, T, Sherwood, R. (1993),
Systems: Bonucci, E. (Ed.). Boca Raton: CRC Comp. Biochem. Physio 1. B: Cornp. Biochem. 105,
Press, pp. 1-18. 487-496.
Ellsworth, M.W., Novak, B.M. (1991), J. Am. Chem. Kendall, K. (1975), Proc. R. Soc. London A 344, 287-
Soc. 113, 2756-2758. 302.
Engfeldt, B., Reinholt, F. P. (1992), in: Calcification in Koutsoukos, P.G., Nancollas, G.H. (1987), Colloids
Biological Systems: Bonucci, E. (Ed.). Boca Raton: Surf. 28, 95-108.
CRC Press, pp. 217-242. Landry, C.J.T., Coltrain, B.K., Brady, B.K. (1992),
Evans, A. G. (1988), in: Metallic and Ceramic Com- Polymer 33, 1486-1495.
posites, Proc. 9th Rise Int. Symp. Roskilde: Ris0 Laria, V X, Heuer, A. H. (1990), in: Structural Ceram-
National Lab. pp. 13-34. ics, Processing, Microstructure and Properties,
Fendler, J.H. (1984), Science 223, 888-894. Proc. 11th Riso Int. Symp. Roskilde: Riso National
Fink, D.J., Caplan, A. I., Heuer, A.H. (1992), MRS Lab., pp. 79-96.
Bull, October, 27-31. Lauder, A., Cima, M.X, Sachs, E., Fan, T. (1992),
Fitzgerald, J.J., Landry, C.J.T., Pochan, J.M. (1992), Mater. Res. Soc. Symp. Proc. 249, 331-336.
Macromolecules 25, 3715-3722. Levi, C , Barton, XL., Guillemet, C , LeBras, E.,
Folsom, C.A., Zok, F.W., Lange, F. F, Marshall, Lehuede, P. (1989), J. Mater. Sci. Lett. 8, 337-339.
D.B. (1992), /. Am. Ceram. Soc. 75, 2969-2975. Lewis, G. (1990), /. Mater. Educ. 12, 197-222.
Frankel, R.B., Blakemore, R.P. (1984), Phil Trans. Li, C.-W, Volcani, B.E. (1984), Phil. Trans. R. Soc.
R. Soc. London B 304, 561-51 A. London B 304, 519-528.
Garrone, R., Simpson, T.L., Pottu-Boumendil, J. Lowenstam, H. A., Weiner, S. (1989), On Biomineral-
(1981), in: Silicon and Siliceous Structures in Bio- ization. Oxford: Oxford University Press.
logical Systems: Simpson, T.L., Volcani, B.E. Madeleine, D.G., Ward, T.C., Taylor, L.T. (1988), /.
(Eds.). Berlin: Springer, pp. 495-525. Polym. Sci. B: Polym. Phys. 26, 1641-1655.
Gebelein, C.G. (Ed.) (1990), Biomimetic Polymers. Mann, S., Hannington, XP. (1988), J. Colloid Inter-
New York: Plenum Press. face Sci. 122, 326-335.
Gianellis, E.P. (1992), JOM 44, 28-30. Mann, S., Perry, C.C. (1986), in: Silicon Biochem-
Gianellis, E.P., Mehrotra, V., Russell, M.W. (1990), istry: Ciba Foundation Symp. Vol. 121: Evered, D.,
Mater. Res. Soc. Symp. 180, 685-696. O'Connor, M. (Eds.). Chichester, UK: Wiley,
Hardy, A.B., Gowda, G., McMahon, T.X, Riman, pp. 40-58.
R.E., Rhine, WE., Bowen, H.K. (1988), in: Ultra- Mann, S., Frankel, R.B., Blakemore, R.P. (1984),
structure Processing of Advanced Ceramics: Nature 310, 405-407.
Mackenzie, J.D., Ulrich, D.R. (Eds.). New York: Mann, S., Parker, S.B., Ross, M.B., Skarnulis, A.X,
Wiley. Williams, R.J.P. (1983), Proc. R. Soc. London B
Hardy, A. B., Rhine, W. E., Bowen, H. K. (1990), Ma- 218, 415-424.
ter. Res. Soc. Symp. Proc. 180, 1009-1014. Marchessault, R. H., Riuoux, P., Raymond, L.
Heuer, A.H., Fink, D.J., Laria, V.J., Arias, XL., (1992), Polymer 33, 4024-4028.
Calvert, P.D., Kendall, K., Messing, G.L., Black- Matijevic, E. (1981), Ace. Chem. Res. 14, 22-29.
well, X, Rieke, P.C., Thompson, D.H., Wheeler, Mauritz, K.A., Stefanithis, I.D. (1990), Macro-
A.P., Veis, A., Caplan, A.I. (1992), Science 255, molecules 23, 1380-1388.
1098-1105. Mauritz, K.A., Warren, R.M. (1989), Macro-
Heywood, B.R., Fendler, J.H., Mann, S. (1990), J. molecules 22, 1730-1734.
Colloid Interface Sci. 138, 295-298. Mazur, S., Manring, L. E., Levy, M., Dee, G.T.,
Huang, H., Glaser, R.H., Brennan, A. B., Rodrigues, Reich, S., Jackson, C.E. (1989), in: Metallized
D., Wilkes, G. L. (1992), in: Ultrastructure Process- Plastics: Mittal, K. L. (Ed.). New York: Plenum,
ing of Advanced Materials: Uhlmann, D.R., Ul- pp. 115-134.
rich, D.R. (Eds.). New York: Wiley, pp. 425-438. Meldrum, F.C., Wade, V.X, Nimmo, D.L., Hey-
Inouye, K., Endo, R., Otsuka, Y, Miyashiro, K., wood, B.R., Mann, S. (1991), Nature 349, 684-
Kaneko, K., Ishikawa, T. (1982), /. Phys. Chem. 86, 686.
1465-1469. Messersmith, P.B., Stupp, S.I. (1991), Polym. Prepr.
Jackson, A. P., Vincent, XF.V., Turner, R.M. (1988), 32, 536-537.
Proc. R. Soc. London B 234, 415-440. Nandi, M., Conklin, X A., Salvati, L., Sen, A. (1990),
Jackson, A.P., Vincent, XF.V, Turner, R.M. (1989), Chem. Mater. 2, 772-776.
Compos. Sci. Technol. 36, 255-266. Novak, B.M., Davies, C. (1991), Polym. Prepr. 32,
512-513.
 12.9 References 81

Okada, A., Kawasumi, M., Usuki, A., Kojima, Y, Stuffle, K., Mulligan, A., Lombardi, I , Calvert, P.,
Kurauchi, T., Kamigaito, O. (1987), Polym. Prepr. Fabes, B. (1994), Mater. Res. Soc. Symp. Proc. 346,
28, 447-448. 1027.
Okada, A., Fukumori, K., Usuki, A., Kojima, Y, Sullivan, C.W. (1986), in: Silicon Biochemistry Ciba
Sato, N., Kurauchi, T., Kamigaito, O. (1991), Symp. 121: Evered, D., O'Connor, M. (Eds). Chi-
Polym. Prepr. 32, 540-541. chester, UK: Wiley, pp. 59-89.
Pope, E. J.A., Asami, M., Mackenzie, I D . (1989), J. Vincent, J.F.V. (1990), Structural Biomaterials, 2nd
Mater. Res. 4, 1018-1026. ed. Princeton, NJ: Princeton University Press.
Rieke, P.C., Tarasevich, B.J., Bentjen, S.B., Fryxell, Wang, S., Mark, I E . (1990), Macromolecules 23,
G.E., Campbell, A. A. (1992), in: Supramolecular 4288-4291.
Architecture, ACS Symp. 499: Bein, T. (Ed.). Wash- Wheeler, A. P. (1992), in: Calcification in Biological
ington, DC: American Chemical Society, pp. 6 1 - Systems: Bonucci, E. (Ed.). Boca Raton: CRC
75. Press, pp. 180 216.
Rieke, P. (1987), in: Atomic and Molecular Processing White, R.A., Weber, I N . , White, E.W. (1972), Sci-
of Electronic and Ceramic Materials: Aksay, LA., ence 176, 922-923.
McVay, G.L., Stoebe, T.G., Wager, I F . (Eds.). Wu, C.Cm., McKinney, K.R., Rice, R.W. (1990), /.
Pittsburgh: Materials Research Society, pp. 109- Mater. Sci. 25, 2170-2174.
114. Yamamura, T., Ishikawa, T., Shibuya, M., Tamura,
Risk, M.I, Pearce, T.H. (1992), Nature358, 572-573. M., Nagasawa, T., Okamura, K. (1989), Ceram.
Rivera, E.R., Smith, B.N. (1979), Plant Physiol. 64, Eng. Sci. Proc. 10, 736-747.
966-970. Yano, K., Usuki, A., Okada, A., Kurauchi, T., Kami-
Robey, P.G. (1992), in: Calcification in Biological gaito, O. (1991), Polym. Prepr. 32, 65-66.
Systems: Bonucci, E. (Ed.). Boca Raton: CRC Yasrebi, M., Kim, G.H., Gunnison, K.E., Milius,
Press, pp. 107-127. D.L., Sarikaya, M., Aksay, LA. (1990), Mater.
Rodrigues, D. E., Brennan, A. B., Betrabet, C , Wang, Res. Soc. Symp. Proc. 180, 625-633.
B., Wilkes, G.L. (1992), Chem. Mater. 4, 1437- Young, J.R., Didymus, J.M., Bown, P.R., Prins, B.,
1446. Mann, S. (1992), Nature 356, 516-518.
Rowcliffe, D. J., Fruhauf, V. (1977), J. Mater. Sci. 12, Zhao, M.X., Ning, Y.P., Mark, I E . (1993), Proc. of
35-42. Symp. on Composites: Processing, Microstructure
Runham, N.W., Thornton, P.R., Shaw, D.A., and Properties: Sacks, M.D. (Ed.). Westerville,
Wayte, R.C. (1969), Z. Zellforsch. 99, 608-626. OH: American Ceramic Society.
Sarikaya, M., Gunnison, K.E., Yasrebi, M., Aksay, Ziolo, R.F., Giannelis, E.P., Weinstein, B.A.,
LA. (1990), in: Mater. Res. Soc. Symp. Proc. 174, O'Horo, M.P., Ganguly, B.N., Mehrotra, V., Rus-
109-116. sell, M.W., Huffman, D.R. (1992), Science 257,
Schmidt, H. (1992), in: Ultrastructure Processing of 219-223.
Advanced Materials: Uhlmann, D.R., Ulrich,
D.R. (Eds.). New York: Wiley, pp. 409-423.
Schmidt, H., Seiferling, B., Philipp, G., Deichmann,
K. (1988), in: Ultrastructure Processing of Ad-
vanced Ceramics: Mackenzie, I D . , Ulrich, D.R. General Reading
(Eds.). New York: Wiley, pp. 651-660.
Sela, I, Schwartz, Z., Swain, L.D., Boyan, B.D. Aksay, I., Baer, E., Sarikaya, M., Tirrell, D. A. (Eds.)
(1992), in: Calcification in Biological Systems: (1992), Hierarchically Structured Materials: Mater.
Bonucci, E. (Ed.). Boca Raton: CRC Press, Res. Soc. Symp. 255. Pittsburgh, PA: MRS.
pp. 73-105. Alper, M., Calvert, P., Frankel, R., Rieke, P., Tirrell,
Sikes, C.S., Yeung, M.L., Wheeler, A. P. (1991), in: D. (Eds.) (1991), Materials Synthesis Based on Bio-
Surface Reactive Pep tides and Polymers: Sikes, logical Processes: Mater. Res. Soc. Symp. 218.
C.S., Wheeler, A. P. (Eds.). ACS Symp. Ser. 444, Pittsburgh, PA: MRS.
pp. 50-71. Alper, M., Bayley, H., Kaplan, D., Navia, M. (Eds.)
Skinner, D.P., Newnham, R.E., Cross, L.E. (1978), (1994), Biomolecular Materials by Design: Mater.
Mater. Res. Bull. 13, 599-607. Res. Soc. Symp. 330. Pittsburgh, PA: MRS.
Spanhel, L., Arpac, E., Schmidt, H. (1992), J. Non- Gonsalves, K.E., Chow, G.-M., Xiao, T.D., Cam-
Cryst. Solids 147, 657-662. marata, R.C. (Eds.) (1994), Molecularly Designed
Steigerwald, M.L., Alivisatos, A. P., Gibson, J.M., Ultrafine/Nanostructured Materials: Mater. Res.
Harris, T.D., Kortan, R., Muller, A.I, Thayer, Soc. Symp. 351. Pittsburgh, PA: MRS.
A.M., Duncan, T.M., Douglass, D.C., Brus, L.E. Komareni, S., Parker, J.C., Thomas, G.I (Eds.)
(1988), / Am. Chem. Soc. 110, 3046-3050. (1993), Nanophase and Nanocomposite Materials:
Stuffle, K., Mulligan, A., Calvert, P., Lombardi, I Mater. Res. Soc. Symp. 286. Pittsburgh, PA: MRS.
(1993), in: Solid Freeform Fabrication Symp. Proc. Lowenstam, H. A., Weiner, S. (1989), On Biomineral-
Austin, TX: University of Texas, p. 60. ization. Oxford: Oxford University Press.
82 12 Biomimetic Processing

Mann, S., Webb, J., Williams, R.J.P. (Eds.) (1989), Biology as a Guide for New Materials Technology.
Biomineralization: Chemical and Biochemical Per- Washington, DC: National Academy Press.
spectives. Weinheim: VCH. Vincent, J. (1990), Structural Biomaterials. Princeton,
Rieke, P.C., Calvert, P.D., Alper, M. (Eds.) (1990) NJ: Princeton University Press.
Materials Synthesis Using Biological Processes: Wainwright, S.A., Biggs, W.D., Currey, J.D., Gos-
Mater. Res. Soc. Symp. 174. Pittsburgh, PA: MRS. line, J.M. (1982), Mechanical Design in Organisms.
Tirrell, D. A. et al. (1994), Hierarchical Structures in Princeton, NJ: Princeton University Press.
13 Sintering and Hot-Pressing
Ulrich Eisele

Robert Bosch GmbH, Stuttgart, Germany

List of Symbols 84
13.1 Introduction 85
13.2 Definition and Stages 85
13.2.1 Initial Stage 85
13.2.2 Intermediate Stage 86
13.2.3 Final Stage 87
13.3 Driving Force 88
13.4 Kinetics 90
13.5 Coarsening 93
13.6 Inhomogeneities 96
13.7 References 97

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
84 13 Sintering and Hot-Pressing

List of Symbols
Ah total grain-boundary area
Ap projected grain-boundary area occupied by pores
As total pore surface area
B constant
C size of a grain cell
D b , Dx grain-boundary and lattice diffusion coefficients
9 9 2/9 s
9IIQ?> geometrical factors relating the size of a cell to its surface area
G grain size
G grain-growth rate
GF Gibbs free energy
J, J b , J, volume flux at the sintering neck, any path, through grain boundary, lattice
k Boltzmann constant
Lvj/ total length of p o r e - b o u n d a r y intersection line
m grain growth exponent
Mb, Mp grain-boundary and pore mobilities
p externally applied pressure
Q activation energy for diffusion
r polar coordinate; equivalent spherical pore radius
R neck radius
T absolute temperature
V volume
Z coordination number
a b , ax fraction of atomic volume transported by one species in grain-boundary and
lattice diffusion
y b , ys grain-boundary and specific surface energy
3 thickness
C number of atoms transferred
x average curvature of all the boundaries
xs curvature of the pore surface
/i chemical potential
Apt sintering potential
A/i GG driving force for grain growth
Q, Q relative density; densification rate
as sintering stress
cp <p = I (71 — i//) (see Fig. 13-8); polar coordinate
<P efficiency factor
W dihedral angle
coh grain-boundary width
Q atomic or molecular volume
 13.2 Definition and Stages 85

13.1 Introduction
On sintering, which is the proof test for
ceramic processing (Kingery, 1978), the
ceramic part obtains its final shape and
properties. Although the main objective is
to densify the powder compact, control of
the final microstructure is equally impor-
tant. Microstructure-related properties
such as strength, toughness, and dielectric
Figure 13-1. The shrinkage of a powder compact im-
properties are therefore strongly influ- plies the shape change of individual grains. This is
enced by the sintering schedule. illustrated in 2 D by the change from circles to hexa-
In this chapter, a treatment of the driv- gons (left). In 3D, ideal packing is achieved by a stack
ing forces and kinetics of solid state sinter- of tetrakaidecahedra (right).
ing is presented for pressureless sintering,
and for sintering with an external force ap-
plied. We will only consider solid-state sin- In a crystalline material, atoms can only
tering; the special aspects of sintering in the be removed and added at interfaces (with
presence of a liquid phase are covered in the exception of dislocation climb). The
Chapter 14 of this Volume. grain center approach is achieved by re-
moving atoms from the grain boundaries
and adding them at the pore surfaces,
transported by thermally activated diffu-
13.2 Definition and Stages sion. The idea that the grain boundaries
act as a source for atoms was first pub-
Sintering can be regarded as the coordi- lished by Kuczynski (1949) and Herring
nated shape change of all grains in a pow- (1951), but it was first advanced by Pines
der compact to allow them arrange them- (Schatt, 1989). In contrast, in amorphous
selves in a space filling manner. This im- materials the grain shape change can be
plies that the grain centers move towards brought about by viscous flow, and in ma-
each other, thereby reducing the size of the terials with glide systems with sufficiently
compact and eliminating the pores. Fig- low shear stress [most metals, and possibly
ure 13-1 illustrates this principle: in two MgO (Vieira and Brook, 1984)] it can
dimensions, initially circular grains would occur by dislocation glide.
become hexagons to give full packing; in The sintering process is usually divided
three dimensions, spheres would transform into the following three stages.
into tetrakaidecahedra (Coble, 1961). Both
final arrangements give a space filling
13.2.1 Initial Stage
packing with minimum possible specific
interface area (Kelvin, 1887) in their re- After forming, drying, and debinding
spective dimension. The reduction of sur- (see Chapters 5-9 of this Volume), the
face and interface area is the driving force powder compact has 40-70% of the theo-
for the process. In the case of hot-pressing, retical density, (see Chapter 10 of this Vol-
an additional driving force is the reduction ume). The first step to achieving full density
of the compact volume under stress. is to maximize the contact points per grain
86 13 Sintering and Hot-Pressing

(coordination number) by reordering the A coordination number of 12.5-14.5 is

grains through translational and rotational found in dense structures (Smith, 1964).
movement. Secondly, as soon as the tem- The initial stage has ended when this final
perature is high enough to allow diffusion, value is reached and no further rearrange-
the network of grain boundaries and pore ment of the grains is possible. Irregular
surfaces has to reach a status of local force packing of monosized spheres can be
equilibrium. This implies that the angle at viewed as a mixture of simple cubic and
the junction between the surface and the hexagonal close-packing (Smith et al.,
grain boundary (the dihedral angle, W) is 1929). The resulting relation between den-
fixed and given by the ratio of the specific sity and coordination number is plotted in
surface to grain boundary energies (ys and Fig. 13-4. From this the initial stage can be
7b) estimated as ending at 75% relative den-
sity. For coarse-grained materials with
7b long diffusion paths (e.g., sinter metals),
COSy = (13-1)
densification will stop at this stage.
As a consequence the grain contact areas
have to widen until the angle *F is reached
13.2.2 Intermediate Stage
(Fig. 13-2). Together with the increase in
the coordination number, this leads to a All grains are now in contact with their
change in curvature of the free grain sur- nearest neighbors, so that the movement of
faces from convex to concave. Figure 13-3 grains as a whole has stopped. Shrinkage
shows a real powder compact somewhere can only proceed by transfer of material
during this transition. The surface trans- from between the grains to the neck surface
port processes also lead to smoothing of brought about by lattice or grain-
the grain surfaces, leading to a reduction in boundary diffusion (Fig. 13-5). The pores
the specific surface area as measured by form a continuous network consisting of
nitrogen adsorption (Burke et al., 1980). cylindrical channels around the necks

initial

intermediate

final

Figure 13-2. Evolution of

the grain arrangement with
the corresponding sintering
stages.
 13.2 Definition and Stages 87

Figure 13-3. Grain structure during the initial stage (A12O3 hot-pressed using 22 MPa for 5 s at 1450 °C; relative
density: 67%). This and all other micrographs were taken by the author.

(Beere, 1975 a; Svoboda et al, 1994). This 13.2.3 Final Stage

stage ends at around 93 % relative density
when the pore channels become too nar- The pores are now closed and situated
row to be stable and decompose into mainly at four grain junctions (Fig. 13-6).
closed pores. Major grain growth can now occur. If the

0.52 0.54 0.56 0.58 0.6 0.62 0.64 0.66 0.68 0.7 0.72 0.74 Figure 13-5. Diffusion paths at a sintering neck: a
Relative density slice of material is removed from the grain boundary
Figure 13-4. Relation between relative density and by (1) lattice or (2) grain boundary diffusion and is
average coordination number for irregularly packed distributed on the pore surface, mainly by (3) surface
monosized spheres (Smith et al., 1929). diffusion.
88 13 Sintering and Hot-Pressing

pact leaving the number of entities con-

stant
dGF = ysdAs + y b d,4 b + pdV (13-2)
where Ah is the total grain-boundary area,
As the total pore surface area, and p an
externally applied pressure.
The last term represents the work done
against an external load in the case of hot-
pressing. The transfer of atoms is caused by
a chemical potential difference between the
source and the sink. This difference is not
the same for all atoms, but its average
value (averaged over all atoms transferred
during a period of time) may be termed the
sintering potential A/L This is given by

(13-3)

where ( denotes the number of atoms

transferred. If one atom is transferred from
Figure 13-6. Grain structure at the beginning of the
final stage (A12O3 hot-pressed using 10 MPa for
a grain boundary into a pore, then the pore
40 min at 1450 °C; relative density: 94%). shrinks because the atom now occupies
part of it, and because the surrounding
structure can move towards the pore ow-
ing to the removed atom. The ratio be-
pores contain gases which are insoluble in tween the shrinkage of the compact as a
the solid, then on shrinking the internal gas result of the transfer of one atom and the
pressure will rise and eventually stop atomic volume, Q, has been labeled the
shrinkage. "efficiency factor", <£ (DeHoff, 1984)
dV
(13-4)
13.3 Driving Force
In crystalline materials, when a layer of
The driving force for sintering is the de- thickness S is removed from grain bound-
crease in the Gibbs free energy, GF, of a aries of area Ah (volume 5Ah\ then the
powder compact brought about by the compact shrinks by a volume S(Ab + Av)9
transfer of material, such that the total where Ap is the grain-boundary area occu-
pore volume and therefore the volume of pied by pores (see Fig. 13-7),
the powder compact, V, are reduced. The
process is driven by replacing high-energy Icrystalline (13-5)
grain surfaces by lower-energy grain
boundaries, and by a reduction in the total In contrast, in an amorphous solid the
interface area due to shrinkage of the com- atom source is not confined to a specific
 13.3 Driving Force 89

To progress any further, we have to rely on

geometrical models for the compact. A vast

O o number of such models have been pro-

posed, but we shall develop a fairly general
type of expression here in order to demon-
strate the kind of concept that h$s been
used to develop sintering rate equations for
the intermediate and final stages.
We assume local equilibrium, i.e., that
o the pore surfaces are spherical with discon-
tinuities in the curvature only at the pore/
o grain-boundary junctions. This implies
that the process which distributes the ma-
terial evenly over the pore surface (surface
or vapor diffusion) is fast compared to the
rate-determining grain-boundary and vol-
ume diffusion.
The change in pore surface area dAs has
two contributions: at the spherical por-
tions it is the inward movement of the pore
surfaces when atoms are added
dAs\s = -xsQdC (13-8)
where xs is the pore curvature and ^4S|S is
the change in pore surface area at spherical
Figure 13-7. Illustration of the efficiency factor 0 for
portions. At the pore-boundary junctions
an amorphous (top) and a crystalline solid (bottom).
The thick "boundary" (top) represents a notional (of total length L^), pore surface disappears
slice, not a physical boundary. in favor of additional grain-boundary area.
We find from Fig. 13-8
1 dAs l j
location. A randomly placed slice will con- cos— =sin(p = - L (13-9)
tain a fraction Q of solid and the remainder
as pore space (Underwood, 1970), so that where As\} is the change in pore surface
in this case the efficiency factor would be area at pore-boundary junctions and cp is
given by (Mackenzie and Shuttleworth, defined in Fig. 13-8. From the definition
1949) of the dihedral angle V [Eqs. (13-1) and
(13-9)]
^amorphous = ~ (13-6)
dAB\t = dAp?± (13-10)
Ys
So Eq. (13-3) can be rewritten as
Combined with Eq. (13-4) and (13-8) the
Aju = (13-7) total change in pore surface area
dV
^ij
So far the treatment has been very gen- dV
eral and valid for all three sintering stages. (13-11)
90 13 Sintering and Hot-Pressing

a tendency to reduce the compact size, and

the externally applied pressure. In the liter-
ature the 'sintering stress' is often used
rather than 'sintering potential'. The sin-
tering stress aT is the magnitude of an ex-
pore
ternally applied stress needed to produce
the same driving force as the surface ten-
sions in the compact (Mackenzie and Shut-
tleworth, 1949; De Jonghe and Rahaman,
1988). It follows that

9
(13-16)

Alternatively, an applied hydrostatic ten-

sile stress of magnitude — oz can be consid-
Figure 13-8. Pore-boundary intersection geometry ered to stop shrinkage (Gregg and Rhines,
for infinitesimal shrinkage (Eisele, 1989).
1973). During the final stage, shrinkage will
stop as soon as the gas pressure inside the
pores reaches — (a s + p).
We define a grain cell as consisting of a For a tetrakaidecahedron g2 = 3.45,
grain with its share of adjacent pores ^3 = 0.523, and 0 = 6.60 (Stevens, 1971). In-
C3Q = G3 (13-12) serting reasonable values into Eq. (13-16),
we find a sintering stress of the order of
where C is the size of a grain cell, Q is the 1 MPa. Considering the applied stresses in
relative density, and G is the grain size. axial hot-pressing (< 30 MPa) and for hot
Furthermore, the geometrical factors g2 isostatic pressing (< 200 MPa), the sinter-
and #3 are introduced such that g2 C2 is the ing potential is then simply given by
surface area and g3C3 is the volume per 5/1« —<PQp. Both terms on the right side
cell. To calculate the grain-boundary area, of Eq. (13-16) are inversely proportional to
we first state that the grain size.
i<? 2 c 2 (13-13)
V

or in differential form with g = g2/g3

13.4 Kinetics
+ Ap)=i-^dV (13-14) The chemical potential difference A/x
translates into shrinkage by driving atoms
Combining Eqs. (13-7), (13-12) and (13- from the boundary to the neck surface.
14) gives This removal of matter must take place at
(13-15) a constant rate everywhere on the neck
boundary, otherwise voids will be created.
The rate of removal must also be propor-
The three terms in the parentheses are the
tional to the sintering potential, so that
capillary force of the curved pore surfaces,
the surface tension of the boundaries with (13-17)
 13.4 Kinetics 91

on the grain boundary, where pi is the and the volume flux at each neck through
chemical potential and K a constant. Since the grain boundary, J b , is
no material flux can be created inside a
crystalline grain (we do not consider dis-
location climb), we have kT
= 0 (13-18) (13-24)
^inside grain kT
In the following, we have to make some
geometrical assumptions to set up the where ab represents the fraction of the
boundary conditions for solving Eqs. (13- atomic volume transported by the species
17) and (13-18). Due to the circular symme- to which the grain-boundary diffusion co-
try of the neck, the solution is most easily efficient relates, e.g., in A12O3 if Dh = D£l
demonstrated for the intermediate stage. then ab = £ (Readey, 1966), coh is the grain-
We introduce the polar coordinates, r and boundary width, k the Boltzmann con-
cp, and the neck radius R. Considering stant, and T the absolute temperature.
boundary diffusion only at first, so that The volume diffusion flux at the neck
V\i = d/i/dr, we find through the lattice, J{, is not calculated so
easily. A good estimate is (Eadie and
1 Weatherly, 1975)
A (13-19)
r dr\ dr
(13-25)
By definition, the average potential differ- '4co b D b a b
ence for each atom from source to sink is
where Dl is the lattice diffusion coefficient
\ R 2n and a, has the corresponding meaning to
/*(*)- (13-20) ab for lattice diffusion. R can be substituted
R^ii I using
All fluxes created in the boundary must
arrive at the neck; as that is the Gauss ZR2n Au 1
2
(13-26)
divergence theorem for two dimensions g2c
R 2n 2n where Z is the coordination number, to
J J V2jurd(pdr= j RVfidcp (13-21)
give
0 0 0

(13-27)
Integrating Eq. (13-20) twice by parts kT
and using Eq. (13-19) finally yields
There are Z necks per grain cell, each
K= (13-22) shared between two cells, so that the vol-
R2 ume change rate per cell is ^ZJ<P, with
From Eq. (13-21) we find the chemical po- J = Jj + J b . When related to the cell vol-
ume, this gives the relative densification
tential gradient at the pore boundary inter-
rate of the whole compact as
section
Q O.5ZJ<2>
(13-23) (13-28)
R Q
92 13 Sintering and Hot-Pressing

and finally rate. Setting, in the case of boundary dif-

fusion control, ^ocexp[-Q/(/cT)]G~ 4 -
(13-29) const, it is found that
Q (13-30)
driving force lnG = const —
AkT
where Q is the activation energy for diffu-
sion.
boundary diffusion lattice diffusion This relation can be used to rationalize
the results of different sintering experi-
kinetic factor
ments (Fig. (13-9), and also to predict the
In the literature there are alternative rate required powder grain size to allow densifi-
equations (Lifshits and Shikin, 1964; John- cation at a desired temperature. Equation
son, 1969; Coble, 1961,1970; Beere, 1975 b; (13-30) is only valid if the compact topol-
Eadie and Weatherly, 1978; Swinkels and ogy and homogeneity are independent of
Ashby, 1981; Hsueh et al., 1986a) which the grain size. A requirement which, espe-
differ from Eq. (13-29) in the choice of their cially for very fine powders, is difficult to
specific geometrical assumptions. How- meet.
ever, all rate equations have the depen- In ionic materials, both cations and an-
dence of the densification rate on the grain ions must diffuse simultaneously. Different
size as the inverse third (lattice diffusion ions can travel on different paths, but at
control) or fourth (boundary diffusion con- the same rate. The slowest ion on its fastest
trol) power in common. For hot-pressing it path determines the rate (Cannon and
is the inverse second and third power since Coble, 1975). For high driving forces (e.g.,
the driving force then is grain-size indepen- hot-pressing) and fast kinetics (fine-grained
dent. These scaling laws (Herring, 1950) material, increased diffusion coefficients
can be used to estimate the influence of the due to dopants), the step to remove or add
grain size on the necessary sintering tem- atoms at the interfaces can become rate-
perature to yield the same densification determining (Wills and McCoy, 1985). In

1000
f1600 +1500 +1400 +1300 +1200 +1100 Figure 13-9. Relation be-
tween sintering tempera-
—Experimental ture and powder grain size
resulting in the same
shrinkage rate; compilation
of some experimental data
-100-
for A12O3 (Harbach et al.,
1990; Xue and Chen, 1990;
Yeh and Sacks, 1988) and
Activation energy for Al 3+ grain - how they can be correlated
boundary diffusion: 418 kJ/mol assuming rate control by
cation grain-boundary dif-
fusion (Cannon and Coble,
104— 1975).
0.5 0.55 0.6 0.65 0.7 0.75
1000 K/T
 13.5 Coarsening 93

this case (called interface control), the ki- ing force changes with 1/G, the grain
netic factor in Eq. (13-29) scales with the growth rate G also changes with 1/G pro-
inverse of the grain size. vided that the boundary mobility M b is
constant. This leads to a parabolic growth
law

13.5 Coarsening G
(13-32)
G G2
Sintering is always accompanied by as found in various theories (Feltham,
coarsening of the microstructure. As can be 1957; Hillert, 1965; Kurtz and Carpay,
seen fom Eq. (13-29), grain coarsening re- 1980), which also all predict an invariant
duces the sintering rate by decreasing the normalized grain-size distribution. The
driving force and by making diffusion same result is obtained when grain growth
paths longer and therefore slowing down is regarded as a result of random size fluc-
the kinetics. tuations of the grain (Louat, 1974), just as
Analogous to sintering, the driving force diffusion down a concentration gradient
for grain growth AfiGG is the average free results from random movement of the
energy decrease per one atom transfer atoms. Both approaches are equivalent
across a grain boundary (Chen, 1987) -just as diffusion can be de-
(13-31) scribed atomostically by random walks or
thermodynamically from a concentration
where x is the average curvature of all the gradient - provided the boundary mobility
boundaries, which is normally taken as in- is constant.
versely proportional to the grain size. Most experimentally observed grain-
Grain boundaries are curved because growth exponents are larger than two. A
space needs to be filled with solids where review of the exponents measured for zone-
the boundaries at three grain junctions refined metals and nominally pure ceram-
meet at 120°, to maintain local force equi- ics (Anderson et al, 1984) showed them to
librium. In principle, this can be achieved lie between 2 and 4 with an average around
by a stack of equally sized tetrakaideca- 2.6. This apparent discrepancy from theory
hedra with double curved hexagonal faces may arise from a driving force that is not
(Kelvin, 1887; Smith, 1948). However, this proportional to 1/G or from a boundary
arrangement is extremely unlikely and mobility that changes with grain size (or
only metastable (a finite disturbance sets time). The latter may be caused by the ac-
off grain growth ad infinitum), so it has cumulation of impurities at the boundaries
never been observed in nature. (Cahn, 1962; Brook, 1968), or by a change
A grain-growth law may be established in the concentration of (mobility con-
from simple scaling arguments (Weaire and trolling) boundary ledges as grain growth
Kermode, 1984). Suppose the topology of proceeds (Chen, 1987). However, it is gen-
the microstructure, i.e., the grain shape and erally accepted (Atkinson, 1988) that grain
the normalised grain-size distribution, is growth in dense materials follows a power
constant with time. Then the average cur- law of the form
vature changes with 1/G, where G is the
mean grain size or any other linear mea- G B
(13-33)
sure of the grain structure. When the driv- G G™
94 13 Sintering and Hot-Pressing

where B is a constant which includes yb

and M b , and m is the grain-growth expo-
nent.
Equation (13-33) remains true when the
grain growth is controlled by a continuous
second phase at the boundaries with the
boundary mobility then being inversely
proportional to the layer thickness (Lay,
1968). Thus, formally, Bocl/G in Eq. (13-
33), and it is found that

mi m p 'lire ^ p u r e •" (13-34)

The same relationship [Eq. (13-34)] is
found if impurities are segregated near the
boundary due to electrostatic (Kliewer and
Koehler, 1965) or size (Johnson, 1977). In-
compatibilities (no second phase), and if
grain growth is controlled by impurity
drag (Brook, 1968).
If the average size of the grains increases,
then, because of mass conservation, the
total number of grains must decrease. A Figure 13-10. The disappearance of small grains dur-
grain shrinks if its boundaries are convex, ing grain coarsening leads to pore coarsening.
i.e., when it has less than the average num-
ber of sides. In 3D a grain can have a
minimum of four sides. Grain growth can grain growth, pores will move with them
thus be regarded as the disappearance of unless it is more economical for the system
four-sided grains (and grains with more to leave the pores behind. The resistance of
than four sides becoming four-sided a pore-laden boundary to movement is the
through the elimination of grain facets). If sum of the resistance resulting from the
pores are situated on the grain corners, pores and from the boundary itself
then these pores will coalesce on disap- A _L A A A
pearance of the grain (Fig. 13-10). There- (13-35)
fore, grain coarsening is always accompa- Mt,:otal Mn
nied by pore coarsening (Kingery and where M p and M b are the mobilities of
Francis, 1965). This is true for any second pore and boundary, defined as velocity per
phase inclusions, as long as they remain unit stress.
attached to the grain boundaries and diffu- The pore mobility depends upon the
sion of second phase atoms is fast enough pore size and the transport path to move
(Green, 1982). material from the leading to the trailing
A pore (or any inclusion) situated at a side (Shewman, 1964; Yan et al., 1983)
grain boundary has a tendency to maintain
M p ocr" (13-36)
its attachment, which originates from the
saving in boundary energy made by its where r is the equivalent spherical pore
presence. When boundaries move during radius, and v = 2 for surface diffusion v = 1
 13.5 Coarsening 95

for lattice and gas diffusion, and v = 0 for

evaporation/condensation and gas diffu-
sion if the total gas pressure is in equilib-
rium with the pore surface tension, i.e.,
pgas = 2yjr. The proportionality constant
contains the appropriate diffusion coeffi-
cient and the dihedral angle (Hsueh et al.,
1982; Svoboda and Riedel, 1992).
Equation (13-35) remains true as long as
the pores remain attached to the bound-
ary. However, if A/nGG is large (small grain
size), breakaway may occur. This is likely
to happen when the pores are large and
widely dispersed. Coarsening is controlled
by the intrinsic boundary mobility when
the pores are small and widely dispersed
(AJMh »y4p/Mp), and by the pores when
they are large and close together (Ah/Mh«
AJMV) (Brook, 1969; Rodel and Glaeser,
1990). A pore left behind by a moving
boundary finds itself trapped inside a
grain. Here the diffusion paths to the next Figure 13-11. Grain structure during the intermedi-
ate stage where some regions are already nearly den-
boundary are long and all species have to sified (A12O3 hot-pressed using 22 MPa for 50 s at
diffuse through the lattice. The shrinkage 1450°C; relative density: 80%, 19.5% open poros-
of such a pore is slow, and it will probably ity).
survive the sintering schedule.
A time only dependent power law with
an exponent close to three has been ob- clear how coarsening can occur in nearly
served for various porous materials during densified regions when the porosity is un-
sintering (Coble, 1961; Coble and Gupta, evenly distributed.
1967; Rhaman et al, 1986). This is not sig- Since the densification rate depends on
nificantly different from what has been ob- the actual grain size, and the grain-growth
served in dense materials. It must be con- rate is a function of the sintering progress,
cluded that grain growth is not primarily both processes are interrelated. Some of
controlled by pore drag alone. Indeed, it the equations developed so far have been
has been observed (Shaw and Brook, 1986) used to calculate density-grain size rela-
that most of the grain growth during the tionships (Harmer, 1984). Inserting dif-
intermediate stage occurs in locally densi- ferent values for diffusion coefficients,
fied areas and is controlled by either intrin- boundary mobility, driving force, etc., it is
sic or impurity drag. The cylindrical pores possible (at least theoretically) to study the
can barely detach from migrating bound- influence of process parameters and do-
aries (Riegel and Svoboda, 1993), so coars- pants on microstructural evolution. In
ening of the channel network structure is principle any measure that favors densifi-
unlikely. However, when considering a cation compared to coarsening will lead to
structure as shown in Fig. 13-11 it becomes faster sintering rates and a finer final mi-
96 13 Sintering and Hot-Pressing

crostructure. The most obvious of such

measures is to increase the driving force for
densification by hot-pressing (see Chap-
ter 16 of this Volume).

13.6 Inhomogeneities
In reality sintering compacts are not as
homogeneous as assumed in most of the
foregoing text. There are three main types
of heterogeneities:
(a) Evenly dispersed inclusions which are
either rigid or at least show a far slower
shrinkage rate than the surrounding
matrix; these can be agglomerates
(Fig. 13-12), an inhomogeneously dis-
tributed second phase, or deliberately
introduced dense particles such as fi-
bres and whiskers.
(b) Macroscopic gradients in the green Figure 13-12. Low density region caused by an ag-
density or composition. glomerate in which the grains could only obtain a low
coordination. Consequently, there is little shrinkage
(c) Macropores (Lange, 1984; Hirata et al., of the agglomerate and grain growth is hardly pos-
1990) created by either poor powder sible (same sample as in Fig. 13-11).
processing or burning out of organic
'dirt'.
Rigid inclusions do not follow the densi- are manufactured by dry axial pressing.
fication of the surrounding porous com- However, there have been successful at-
pact. As a result, a tensile hydrostatic stress tempts to compensate for the differential
is created in the matrix (Hsueh et al., shrinkage by forming a slightly distorted
1986a; Bordia and Scherer, 1988) which is green body (Riedel and Sun, 1992).
opposite to the sintering stress. This stress The surface curvature of large pores
can be relieved with time by creep of the (larger than the sourrounding grains) is
matrix around the inclusions. However, it concave (as seen from the pore). Therefore
has been shown that as little as 1 % volume the left term on the right side of Eq. (13-16)
fraction of evenly distributed rigid hetero- has the wrong sign and the driving force
geneities can significantly reduce the densi- for such macropores to shrink is reduced.
fication rate (Hsueh et al., 1986 b). The pore and its surrounding grains then
Macroscopic density gradients will nor- behave in a similar manner to a rigid inclu-
mally lead to a gradient in the densification sion.
rate (causing stresses during sintering) and The lesson is that sintering cannot nor-
to a gradient in the final shrinkage (causing mally heal defects that have been intro-
shape distortions). Density gradients are duced by prior processing. However, inho-
often unavoidable when complicated parts mogeneities are always there and normal.
 13.7 References 97

If appropriate, the sintering process must Harmer, M. P. (1984), in: Structure and Properties of
be designed to overcome the detrimental MgO and A12O3 Ceramics: Kingery, W. D. (Ed.).
Westerville, O.H.: American Ceramic Society.
effects of heterogeneities. Grain-growth in- Herring, C. (1950), /. Appl Phys. 21, 301.
hibitors can reduce the coarsening in re- Herring, C. (1951), in: The Physical of Powder Metal-
gions that are prematurely densified, and a lurgie: Kingston, W. E. (Ed.). New York: McGraw-
Hill, p. 143.
higher sintering temperature or a liquid Hillert, M. (1965), Acta MetalL 13, 227.
phase can help the system to relax stresses Hirata, Y, Aksay, I. A., Kikuchi, R. (1990), Nippon
caused by heterogeneities. Sermakkusu Kyokai Gakujutsu Ronbunshi 98 (2),
126.
Hsueh, C. H., Evans, A. G., Coble R. L. (1982), Acta
MetalL 30, 1269.
13,7 References Hsueh, C. H., Evans, A. G., Cannon, R. M., Brook,
R. J. (1986 a), Acta MetalL 34, 927.
Anderson, M. P., Srolovitz, D. J., Grest, G. S., Sahni, Hsueh, C.H., Evans, A. G., McMeeking, R. M.
P. S. (1984), Acta MetalL 32, 783. (1986 b), /. Am. Ceram. Soc. 69 (4), C-64.
Atkinson, A. V. (1988), Ada MetalL 36, 469. Johnson, D. L. (1969), J. Appl Phys. 40, 192.
Beere, W. (1975 a), Ada MetalL 23, 131. Johnson, W. C. (1977), Met. Trans. A8, 1413.
Beere, W. (1975 b), Acta MetalL 23, 139. Kelvin, L., Thomson, W. (1987), Phil. Mag. 24, 503.
Bordia, R. K., Scherer, G. W. (1988), Ada MetalL 36, Kingery, W. D. (1978), in: Ceramic Processing Before
2411. Firing: Onoda, G. Y, Hench, L. L. (Eds.). New
Brook, R. J. (1968), Scr. MetalL 2, 375. York: Wiley, pp. 291-305.
Brook, R. J. (1969), /. Am. Ceram. Soc. 52, 56. Kingery, W. D., Francois, B. (1965), /. Am. Ceram.
Burke, J. E., Lay, K. W., Prochazka, S. (1980), in: Soc. 48 (10). 546.
Sintering Processes: Kuczynski, G. C. (Ed.). New Kliewer, K. L., Koehler, J. S. (1965), Phys. Rev. A140,
York: Plenum, p. 417. 1226.
Cahn, J. W. (1962), Acta MetalL 10, 789. Kuczynski, G. C. (1949), Trans. AIME 185, 169.
Cannon, R. M., Coble, R. L., (1975), in: Deformation Kurtz, S. K., Carpay, F. M. A. (1980), /. Appl Phys.
of Ceramic Materials: Bradt, R. C , Tressler, R. E. 51, 5725.
(Eds.). New York: Plenum, pp. 61-100. Lange, F. F. (1984), J. Am. Ceram. Soc. 67 (2), 83.
Chen, I.-W. (1987), Acta MetalL 35, 1723. Lay, K. W. (1968), /. Am. Ceram. Soc. 51, 373.
Coble, R. L. (1961), J. Appl Phys. 32, 787 [Erratum: Lifshits, I. M., Shikin, V. B. (1964), Sov. Phys. - Solid
(1965), J. Appl Phys. 36, 2327]. State 6, 1362.
Coble, R. L. (1970), J. Appl Phys. 41, 4798. Louat, N. P. (1974), Acta MetalL 22, 721.
Coble, R. L., Gupta, T. K. (1967), in: Sintering and Mackenzie, J. K., Shuttleworth, R. (1949), Proc.
Related Phenomena: Kuczynski, G. C , Hooton, Phys. Soc. (London) B62, 833.
N. A., Gibbon, C. F. (Eds.). London: Gordon and Rahaman, M. N., DeJonghe, L. C , Brook, R. J.
Breach, p. 423. (1986), J. Am. Ceram. Soc. 69, 53.
DeHoff, R. T. (1984), in: Sintering and Heterogeneous Readey, D. W (1966), J. Appl Phys. 37, 2309.
Catalysis: Kuczynski, G. C , Miller, A. E., Sargent, Riedel, H., Sun, D.-Z. (1992), in: Numerical Methods
G. A. (Eds.). New York: Plenum, p. 23. in Industrial Forming Processes: Chenot, J.-L.,
DeJonghe, L. C , Rahaman, M. N. (1988), Acta Wood, R. D., Zienkiewicz, O. C. (Eds.). Rotter-
MetalL 3,6, 223. dam: Baldema, pp. 883-886.
Eadie, R. L., Weatherly, G. C. (1975), Scri. MetalL 9 Riedel, H., Svoboda, J. (1993), Acta MetalL 41, 1929.
(3), 285. Rodel, J., Glaeser, A. M. (1990), J. Am. Ceram. Soc.
Eadie, R. L., Weatherly, G. C , Aust, K. T. (1978), 73, 3302.
Acta MetalL 26, 759. Schatt, W. (1989), personal communication.
Eisele, U. (1989), "The Value of Kinetic Studies for Shaw, N. X, Brook, R. J. (1986), J. Am. Ceram. Soc.
the Determination of Sintering Mechanisms". 69, 107.
Ph.D. Thesis, University of Leeds. Shewmon, P. G. (1964), Trans. AIME 230, 1134.
Feltham, P. (1957), Acta MetalL 5, 97. Smith, C. S. (1948), Trans. AIME 175, 15.
Green, D. J. (1982), /. Am. Ceram. Soc. 65, 610. Smith, C. S. (1964), MetalL Rev. 9, 1.
Gregg, R. A., Rhines, F. N. (1973), Met. Trans. 4, Smith, W. O., Foote, P. D., Busang, P. F. (1929),
1365. Phys. Rev. 34, 1271.
Harbach, F, Neeff, R., Nienburg, H., Weiler, L. Stevens, R. N. (1971), Phil. Mag. 23, 265.
(1990), in: Ceramic Powder Processing Science: Svoboda, J., Riedel, H. (1992), Acta MetalL 40, 2829.
Hausner, H., Messing, O. L., Hirano, S. (Eds.). Svoboda, J., Riedel, H., Zipse, H. (1994), Acta
Cologne: Deutsche Keramische Gesellschaft, p. 609. MetalL 42, 435.
98 13 Sintering and Hot-Pressing

Swinkels, F. B., Ashby, M. F. (1981), Acta Metall. 29, Beere, W. (1975), Acta Metall. 23,131 and 139.
259. Brinker, C. J., Scherer, G. W. (1990), Sol-Gel Science.
Underwood, E. E. (1970), Quantitative Stereology. Orlando, FL: Academic Press, Chap. 11.
Reading, MA: Addison-Wesley. Coble, R. L., Burke, J. E. (1963), in: Progress in Ce-
Vieira, J. M., Brook, R. I (1984), /. Am. Ceram. Soc. ramic Science, Vol. 3: Burke, J. E. (Ed.). New York:
67, 450. Pergamon, pp. 197-251.
Weaire, D., Kermode, J. P. (1984), Phil Mag. B50, Handwerker, C. A., Blendell, J. E., Kaysser, W. (Eds.)
379. (1990), Sintering of Advanced Ceramics, Ceramic
Wills, R. R., McCoy, K. (1985), /. Am. Ceram. Soc. Transactions, Vol. 7. Westerville, OH: American
68, C-95. Ceramic Society.
Xue, L. A., Chen, I. W. (1990), J. Am. Ceram. Soc. 73, Hsueh, C. H., Evans, A. G., Cannon, R. M., Brook,
3518. R. J. (1986), Acta Metall. 34, 927.
Yan, M. R, Cannon, R. M., Bowen, H. K., Chowdhry, Hsueh, C. H., Evans, A. G., McMeeking, R. M.
U. (1983), Mater. Sci. Eng. 60, 275. (1986), /. Am. Ceram. Soc. 69(4), C-64.
Yeh, T. S., Sacks, M. D. (1988), J. Am. Ceram. Soc. Kuczynski, G. C. (Ed.) (1980), Sintering Processes,
71, 841. Materials Science Research, Vol.13. New York:
Plenum Press.
Kuczynski, G. C , Miller, A. E., Sargent, G. A. (Eds.)
(1984), Sintering and Heterogeneous Catalysis,
General Reading Materials Science Research, Vol.16. New York:
Plenum Press.
Ashby, M. F. (1974), Acta Metall. 22, 275. Riedel, H., Zipse, H., Svoboda, J. (1994), Acta
Atkinson, A. V. (1988), Acta Metall. 36, 469. Metall. Mater. 42, 445.
14 Liquid-Phase Sintering
Oh-Hun Kwon

Northboro R&D Center, Saint-Gobain/Norton Industrial Ceramics Corporation,

Northboro, MA, U.S.A.

List of Symbols and Abbreviations 100

14.1 Introduction 102
14.2 Stages of Liquid-Phase Sintering 103
14.3 Wetting by Liquid 105
14.4 Driving Force for Sintering 106
14.5 Elementary Densification Mechanisms 107
14.5.1 Rearrangement 107
14.5.2 Solution-Precipitation 108
14.5.3 Pore Removal 110
14.6 Grain Boundaries Ill
14.7 Grain Growth and Coalescence 113
14.8 Use of Phase Diagrams 113
14.9 Reactive Liquid-Phase Sintering 116
14.10 Transient Liquid-Phase Sintering 117
14.11 Real Powder Compacts 117
14.12 Outlook for the Future 119
14.13 References 120

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
100 14 Liquid-Phase Sintering

List of Symbols and Abbreviations

Asy9 A s s , A s l 9 AlY interfacial areas between solid (s), liquid (1), and vapor (v) in a particle
compact
A (g), B(g), C(g) geometrical constants, which are dependent on relative density, liquid
volume, and geometry at particle contacts
cx solubility of solute in a liquid
Dh grain boundary diffusion coefficient
E activation energy
F force
AG change in free energy of a system
h thickness of liquid film
K interface reaction constant
k rate constant for grain growth
kB Boltzmann constant
n radius (or grain size) exponent
Rx, R2 principal radii of the liquid meniscus
rc contact radius
rY radius of liquid lens at a particle contact
rp radius of pore
rs radius of solid particle
SD driving stress for densification
t time
T absolute temperature
Vx, K p , Vs volume fraction of liquid, pores, and solid, respectively
ysv, yss, y sl , ylv solid-vapor, solid-solid, solid-liquid, and liquid-vapor
(surface tension) interfacial energies, respectively
3 thickness of liquid film in two-grain boundary
rj viscosity of liquid
9 contact or wetting angle
[i chemical potential
Q relative density
£0 initial density
A@ density difference
@r, @s, ^ p densification rates by rearrangement, solution-precipitation,
and pore removal, respectively
G stress (or pressure)
cra applied stress
<7 p vapor pressure inside pore
Q molecular volume of solid
D.F. driving force for densification
EDS energy dispersive spectroscopy
HIPing hot isostatic pressing
HRTEM high resolution transmission electron microscope
 List of Symbols and Abbreviations 101

LFM liquid film migration

LPS liquid-phase sintering
MAS magnesium-alumino-silicate (MgO-Al2O3-SiO2)
PLZT (Pb,La)(Zr,Ti)-O 3
P.S. particle size
PZT Pb(Zr,Ti)O 3
SEM scanning electron microscope
SSS solid-state sintering
TEM transmission electron microscope
TKD tetrakaidecahedron
VCS viscous composite sintering
VGS viscous glass sintering
YAG yttrium-aluminum garnet
102 14 Liquid-Phase Sintering

14.1 Introduction Table 14-1. Various kinetic processes that occur dur-
ing liquid-phase sintering.
Liquid-phase sintering (LPS) is an im- Process Definition
portant consolidation process for manu-
facturing a variety of dense ceramic com- Melting initial liquid formation
ponents from porous, powder compacts. Wetting:
spreading wetting by liquid on free
The first industrial use for LPS was for solid surface
powder metallurgy products (Price et al., penetration wetting by liquid between
1938) with a sintering experiment on an 80 solid surfaces
Fe-20 Cu compact. Most ancient ceramics, Dissolution of solid dissolution of solid in a liquid
e.g., pottery and porcelains, though, were Diffusion of liquid diffusion of liquid
also produced by complex LPS processes into solid component(s) into solid
which were practiced as a craft. At present Chemical reaction reactions between solid, liquid
technical ceramics manufactured by LPS and vapor, e.g., formation
of reaction product(s) by
include alumina and A1N substrates for
incongruent melting
electronics, alumina and SiC mechanical
Rearrangement capillary induced, lubricated
seals, alumina and Si 3 N 4 glow plugs, sili- particle sliding (movement)
con nitride/sialon structural parts, ZnO toward higher compact density
varistors, BaTiO3 capacitors, PLZT Solution- dissolution of solid and re-
[(Pb,La)(Zr,Ti)-O3] piezoelectric compo- precipitation precipitation of solute resulting
nents, and various composites. The pri- in mass transfer
mary advantage of LPS as a densification Pore isolation isolation of continuous pore
process is the enhanced kinetics of sinter- (closure) channels
ing. Solid powders which are hard to sinter Pore removal gas and vacancy diffusion
from internal pore to surface
by solid-state sintering (SSS), can in many of compact
cases be easily sintered by LPS at lower Grain growth and growth in pore size and
temperatures than those by SSS. Another coalescence (Ost- decrease in the number
major benefit of LPS is that it is an impor- wald ripening) of grains
tant method for manufacturing ceramic Neck growth growth in solid-solid contact
alloys with tailored microstructures and area
optimized properties, as recently demon- Pore growth growth in pore size and de-
strated by a number of silicon nitride and coalescence crease in number of pores
alloys with significantly improved fracture Grain/liquid flow simultaneous flow of grain
and liquid into macropores
toughness.
Bloating gas pressure-induced localized
Overall the processing steps of LPS are swelling in a compact
quite similar to those-of SSS. First, two or Solidification solidification of liquid phase
more solid powders are homogeneously during cooling
mixed by dry or wet mixing techniques. Crystallization crystallization of liquid during
The mixed powder is formed into a green cooling
shape of 50-65 % relative density by vari-
ous forming methods, e.g., uniaxial die
pressing, cold isostatic pressing, slip cast- phase, typically 1-20 vol% of the com-
ing, injection molding, etc. The structure pact, is formed upon heating the powder
of a ceramic green compact can be a ran- compact, either by melting of one or more
dom loose or dense packing. A liquid of the constituents or by reaction between
 14.2 Stages of Liquid-Phase Sintering 103

Characteristic
viscosities of Melting Softening Annealing Figure 14-1. Viscosity
glass Working regimes of liquids for vari-
ous densification tech-
I I I I I I I niques compared to char-
Viscosity (Pa s) acteristic viscosities of
2 2 4 6 8 10
10- 10° 10 10 10 10 10 10 12 10 14 glass; LPS: liquid-phase
Sintering sintering, VCS: viscous
.—> <
LPS VCS VGS
composite sintering,
<—> <—> VGS: viscous glass sinter-
Metals Ceramics Hot pressing of glass ing.

the constituents. While the mixture of solid viscosity regimes of liquids for different
particles and liquid sinters together, the densification techniques. It is noted that
porosity of the powder compact gradually the level of viscosity of a silicate grain
diminishes to form a dense ceramic part, boundary phase (see, for example, Riebling,
resulting in a useful engineering compo- 1964), which is most common in ceramics,
nent. A number of physical and chemical is higher than that of metals by about three
processes occur either simultaneously or orders of magnitude. However, the viscosi-
successively during LPS, as summarized in ties of nonglass forming liquids could be as
Table 14-1 along with their definitions. low as those of metals. Consequently, the
While some processes promote densifica- LPS behavior of ceramics is appreciably
tion, other processes listed in Table 14-1 different from that of metals.
are deleterious to densification. Hence the
material system and processing step of
LPS should be designed to promote favor- 14.2 Stages of Liquid-Phase
able processes and to minimize unfavor- Sintering
able processes during firing, on the basis of
the fundamental physical and chemical un- Densification processes by LPS are,
derstanding of processes. classically divided into three distinctive
There are three general requirements for stages defined by three different rate con-
LPS (Kingery, 1950; Kingery and Nara- trolling mechanisms (Kingery, 1950), which
simhan, 1950): (i) a liquid must be present are schematically presented as Stages I, II,
at the sintering temperature, (ii) there and III in Fig. 14-2. However, prior to ap-
should be good wetting of the solid by the preciable densification, a few important
liquid (i.e., a low contact angle), and (iii) physical and chemical processes take place
the solid must be reasonable soluble in the such as melting, wetting (or liquid flow),
liquid. LPS of metals has been studied ex- and reaction(s) between solid and liquid;
tensively, and there are some excellent Stage 0 in Fig. 14-2. Stage 0 is transient
monographs on the subject (Eremenko et and results in negligible densification.
al., 1970; German, 1985). In contrast to With increasing density, the controlling
metal systems, ceramic systems are charac- densification mechanism progressively
terized by a viscous grain boundary phase, changes from rearrangement (Stage I) to
limited mutual solubility, complex crystal solution-precipitation (Stage II), and to
systems, and slow reactions between con- final pore (or vapor phase) removal (Stage
stituents. For example, Fig. 14-1 shows the III). Table 14-2 summarizes the competing
104 14 Liquid-Phase Sintering

(a) loose packed] close packed

open pores closed pores

(b)
I.0 Figure 14-2. (a) Schematic
diagrams of the stages of
0.9 liquid phase sintering
(0: melting; I: rearrange-
ment; II: solution-precipi-
0.8- = 7.5 vol % tation, and III: pore re-
RS. = 3.6/i. m moval), (b) Stages of LPS
0.7 a ,I625°C with an example of actual
o,|600°C densification as a function
A,I575°C of sintering time in an alu-
0.6 — 0, melting ©,I55O°C mina-glass system at dif-
I I I I II I I I I II Mil ferent temperatures.
0.5 10 100 1000
Sintering time (min)

Table 14-2. Stages of liquid-phase sintering and cor- densification mechanisms and presents
responding densification rates a . typical densification (volume shrinkage)
SPS stages Densification rate a rates. However, significant overlapping
exists between the connecting stages in ac-
Competing Typical tual powder compacts, as illustrated in
densifica- densifica-
tion rates b tion rate b
Fig. 14-2 b. In general, the densification
rate decreases significantly as the sintering
Stage 0: melting and progresses, typically from 10~3/s to 10~6/s.
wetting An appreciable amount of dedensification
Stage 1: rearrangement (or desintering) may occur at the end of
Stage 1/2: transition LPS, as shown in Fig. 14-2 b with an ex-
Stage 2: solution-pre- tended sintering time.
cipitation To quantitatively describe the three
Stage 2/3: transition regimes of sintering, Kwon and Messing
Stage 3: pore removal (1991) have developed a ternary liquid-
a
Estimated from LPS kinetic data for an alumina-
phase sintering diagram (Fig. 14-3). The
glass system; b QX, QS, QP are densification rates for stages of LPS and dominant sintering
rearrangement, solution-precipitation, and pore re- mechanisms are mapped as a function of
moval, respectively. the relative volume fractions of solid (Vs),
 14.3 Wetting by Liquid 105

liquid (V^) and pores (Fp). Changes in the The final sintering stage (pore removal)
relative volume fractions during densifica- may start immediately after pore closure,
tion for solid-state sintering (SSS), LPS, i.e., when Q (= Vs + JQ > 0.92, at the end of
viscous composite sintering (VCS), and solution-precipitation. The sintering be-
viscous glass sintering (VGS) are shown as havior of a powder compact with an exces-
densification loci. For LPS, a porous pow- sive amount of liquid (20-100 vol%; VCS
der compact at O is densified by traversing and VGS in Fig. 14-3) is quite different
three regions of successive mechanisms, (I, from that of LPS. VCS and VGS do not
II, and III in Fig. 14-3), along the arrow to necessarily require solution-precipitation
a dense compact at Q. The two boundaries as a densification mechanism. Therefore
between the three regions can be deter- these two processes should not be confused
mined by geometrical analyses of compact with LPS.
structures. Assuming monosized spherical
particles, the rearrangement of particles
will cease at ^ = 0.74 on achieving a close- 14.3 Wetting by Liquid
packed structure. The boundaries for den-
sification by solution-precipitation are Good wetting of a solid by a liquid is a
conservatively determined as a triangular fundamental requirement for LPS. Sup-
region DEF in Fig. 14-3 pose a drop of liquid is placed on a per-
fectly smooth surface and that these phases
0.74 < ^ < 0.92
are in equilibrium with the surrounding
0 <^<0.20 (14-1) vapor phase. Young's equation describes
0.08 <Vp< 0.26 the equilibrium force balance (see, e.g.,
Hiemenz, 1977)
ylvcos0 = y sv -y sl (14-2)
where ylv is the liquid-vapor interfacial
energy (surface tension), 9 is the contact
V s =0.74 (or wetting) angle, and ysv and ysl are the
solid-vapor and solid-liquid interfacial
energies. For efficient LPS, the contact an-
gle should be small enough (e.g., <45°) to
achieve complete wetting of the solid parti-
cles.
The wetting of a solid by a liquid
strongly depends on the temperature of the
Liquid VGS Pore system and chemical reaction. A small
Figure 14-3. Ternary diagram showing volumetric
amount of additive can also significantly
phase relationships during densification by SSS, LPS, alter the wettability of a liquid. A liquid
viscous composite sintering (VCS), and viscous glass composition that has a lower contact angle
sintering (VGS). The arrows represent loci of the vol- can be designed by choosing a more effec-
umetric changes of the phases with an initial compact tive additive(s). The contact angle also
density (Q) of 60%. In the LPS region, ABCS, subdi-
visions for dominant mechanisms are also displayed;
varies as a function of time in most practi-
I: rearrangement, II: solution-precipitation, III: pore cal measurements (Towers, 1954). As a re-
removal (Kwon and Messing, 1991). sult of reactions and solubilities between
106 14 Liquid-Phase Sintering

Ylv the interfacial free energy of the system. In

general, the change in free energy on going
from one configuration to another (see
Fig. 14-2 a) in a solid-liquid-vapor sys-
Solid tem is

Figure 14-4. Schematic diagram of a sessile drop on

a substrate when the liquid dissolves the crystalline ?iv) (14-3)
phase (Kwon, 1986).
where A^4SV, A^4SS, AAsl, and AAU are
changes in the various interfacial areas and
liquid and solid in LPS systems, the wet- ?Sv> yss> Vs\> a n d 7iv a r e t h e i r corresponding
ting angle can be determined by a method interfacial energies (subscripts s, 1 and v
shown schematically in Fig. 14-4. For represent solid, liquid, and vapor), respec-
more precise determination of the contact tively. If good wetting of the solid by a
angle, a liquid of equilibrium composition liquid is assumed, AAsy and AASS are unim-
at the temperature of interest (e.g., the sin- portant. Also, when there is no grain
tering temperature) can be used. growth, AAsl is negligible. Therefore, AAU
Kwon (1986) reported wetting angles of is the primary and most important variable
a MAS (magnesium -alumino- silicate) in determining the driving force for LPS.
glass on the (0001) plane of a sapphire sub- Assuming a liquid lens at a contact be-
strate. Alumina was dissolved in a eutectic tween two spheres, as shown in Fig. 14-5,
MAS glass to form an equilibrium glass
composition upon heating for 1 hour, fol-
lowed by quenching in air. The measured
contact angles varied from 45 to 22° at
temperatures from 1400 to 1600°C for this
system, as determined by the method illus-
trated in Fig. 14-4. Note that the LPS of
the alumina/glass system is usually per-
formed at 1500-1600°C. The sintering
rate at 1400 °C is too slow, and is therefore
not used in practice. The technique in
Fig. 14-4 is also useful for determining any
interface reaction and crystallization of the
grain boundary phase as a function of heat
treatment.

14.4 Driving Force for Sintering

(0,0)

Figure 14-5 illustrates a simplified two-

grain contact (two-sphere model) depict- Figure 14-5. Classical two-sphere model for liquid-
phase sintering. The Laplacian force develops at a
ing a solid-liquid-vapor assemblage. grain contact due to the concave liquid lens, resulting
Densification during LPS is driven by the in a compressive normal force at the contact (Heady
thermodynamic driving force to minimize and Cahn, 1970).
 14.5 Elementary Densification Mechanisms 107

the force per contact is centration gradient, A(^, and consequently

(14-4)
a driving force for mass transfer away
1 from the contact.
F= (2 n rx ylv cos cp) — nrfy

where rx is the radius of the lens, (p is de-

fined in Fig. 14-5, and Rx and R2 are the 14,5 Elementary Densification
principal radii of the liquid meniscus. The Mechanisms
first term in Eq. (14-4) is due to the surface 14.5.1 Rearrangement
tension of the liquid. The second term re-
sults from the curved liquid-vapor inter- During the initial LPS stage, a number
face and is known as the Laplacian force. of consecutive and simultaneous processes
However, for good wetting conditions in may occur, including melting, wetting,
which the contact angle #<30°, the angle spreading, and redistribution of liquid.
(p becomes large and the first term is negli- Both solid and liquid are subject to appre-
gible. Thus the driving force (D.F.) for ciable rearrangement due to local capillary
densification during the intermediate stage forces in random directions around the
of sintering with good wetting conditions solid particles. The local rearrangement is
can be approximated by the Laplacian dictated by particle contact and liquid
force arising from the liquid meniscus at meniscus geometries, resulting in shearing
the particle contact. and rotational movements of particles.
In Fig. 14-5, the capillary pressure is The liquid films between the particles act
balanced by substantial compressive forces as a lubricant during LPS. The rearrange-
in the contact area. This pressure results in ment of particles proceeds in the direction
an increase in the chemical potential, /x. of reducing porosity [to reduce AAU ylv in
Assuming the strain energy contribution is Eq. (14-3)] with a simultaneous reduction
negligible, the relation between the stress in the surface free energy of the system. As
and the chemical potential difference is the density of the powder compact in-
creases, the particles experience increasing
Afi = fi/-fi° = aQ (14-5) resistance to further rearrangement due to
where A/i is the chemical potential differ- impingement by neighboring particles, un-
ence induced by a normal stress, c, on a til a close packed structure is formed.
liquid-solid interface, \JC is the chemical Earlier models based on axial symmetry
potential at a normal stress, /i° is the chem- (e.g., Fig. 14-5) do not adequately explain
ical potential at a reference state, and Q is the driving force and corresponding mov-
the molecular volume of solid. Assuming a ing direction for rearrangement. The driv-
regular solution, the corresponding solu- ing force for rearrangement arises from an
bility due to normal stress is imbalance in the capillary pressure which
results from distribution of the particles
ln|^] = ^4 (14-6) and their particle sizes, their irregular
v? — shape, the distribution of liquid, the local
where, c[ is the solubility at a normal stress, density fluctuation in the powder compact,
Cy is the solubility at a reference state, kB is and the anisotropic material properties. If
the Boltzmann constant, and Tis the abso- the geometry of the particle contact is
lute temperature. Thus the solubility in- known, the driving force for rearrange-
crease at a contact results in a solute con- ment can be calculated for various particle
108 14 Liquid-Phase Sintering

shapes and contact geometries (Heady and tion with assumed friction coefficients and
Cahn, 1970; Cahn and Heady, 1970). The monosized spheres has also been devel-
random nature of particle packing results oped and has successfully described the re-
in local movements of particles: push-pull, arrangement of particle/liquid.
sliding, and rotation.
Modeling shows that the viscous flow of
14.5.2 Solution-Precipitation
a liquid sandwiched between solid particles
limits the rearrangement process (Kwon, When further rearrangement becomes
1986). Assuming there is a Newtonian liq- negligible, an additional densification
uid between two particles, the deformation mechanism must be operative to attain fur-
rate is proportional to the shearing stress ther densification. At the end of the rear-
exerted on the particles. Accordingly, the rangement stages, the densification rate
resulting densification rate is given by due to solution-precipitation becomes sig-
nificant compared to that of rearrange-
= Al Tiv (14-7) ment, as shown in Table 14-2. The solubil-
d/
ity increase at a grain contact, Acl9 is
where Q is the relative density, QO is the proportional to the normal traction result-
initial compact density, AQ is the density ing from the capillary force (Laplacian
difference, t is the time, A (g) is the geomet- force) compressing the solid particles to-
rical constant, which is a function of Vl9 Q gether, i.e., Eq. (14-6). The volume shrink-
and the contact geometry, rj is the viscosity age at this stage is primarily obtained from
of the liquid, and rs is the radius of the the center-to-center distance between
solid particle. A (g) increases with increas- neighboring particles as a result of solu-
ing volume fraction of solid and liquid, tion-precipitation at particle contacts.
and decreases with increasing relative den- The high concentration of solute at
sity. In actual particle compacts, full densi- compressed particle contacts transfers to
fication can be achieved by rearrangement the uncompressed part of the grain struc-
alone at approximately 30-35 vol% liq- ture by diffusion through the liquid phase,
uid. (Kingery, 1950; Kingery and Nara- followed by reprecipitation of the solute
simhan, 1950). on an uncompressed (free) solid surface, as
The rearrangement behavior of powder shown schematically in Fig. 14-6 for a
compacts with an excessive amount of liq- multicomponent system. This mass trans-
uid (Ewsuk and Harrison, 1990) is quite fer results in contact-point flattening and a
different from that in LPS. In contrast, in corresponding linear shrinkage in the pow-
solid-state sintering (SSS), particle rear- der compact. The dissolution rate of the
rangement is not a significant mechanism solid decreases as the contact area in-
for the densification due to the absence of creases due to a simultaneous reduction in
liquid capillarity and lubricating films be- the effective stress in the contact area.
tween particles. Direct observation of the Accordingly, the densification (volume
rearrangement by Huppmann et al. (1979) shrinkage) rate decreases as the density of
using a hot-stage scanning electron micro- the powder compact increases. At the later
scope (SEM) showed that the process is stage of solution-precipitation, the inter-
discontinuous, with discrete motion of in- connected pore structure pinches off to
dividual particles or small groups of parti- form isolated (closed) pores (Budworth,
cles. Two-dimensional computer simula- 1970).
 14.5 Elementary Densification Mechanisms 109

rate is (Kwon and Messing, 1991)

where B{g) is the geometrical constant

which depends on V39 Vu Q and the appar-
ent dihedral angle, 6 is the thickness of the
liquid boundary (typically 1-3 nm), Dh is
the grain boundary diffusion constant of
the solute, and cx is the solubility of the
solute. If material transport is controlled
by the interface reaction, the densification
(b) rate is

*-2
(14-9)
Concentration
dt kBT
where C(g) is the geometrical constant and
^Tis the interface-reaction constant. Again,
the magnitude of the geometrical constant,
C(g), is determined by the relative density,
the liquid content, and the contact geo-
Figure 14-6. (a) Schematic diagram of a two-grain metry. For example, if the liquid content
contact during LPS by solution-precipitation, show- (Vx) is greater at a constant density, then
ing the three paths for mass transfer: 1: out-diffusion C(g) will be larger. It should be noted that
of solute (•), 2 and 4: influx of solution components both Eqs. (14-8) and (14-9) indicate that
(o and A) into the grain contact region, and 3: disso-
lution-reprecipitation of solute within the contact
the densification strongly depends on rs
region, (b) The corresponding concentration gradi- with exponents of 4 and 2 for diffusion and
ents of the three component liquids as a function of r, interface reaction control, respectively. Ac-
where rc is the contact radius and h is the thickness of cordingly, measurement of the grain size
liquid film (Kwon and Messing, 1991). exponent could be a simple method to
determine the controlling densification
mechanism for a system of interest. Fur-
Using appropriate geometrical models ther analysis (Kwon and Messing, 1991)
for the grain, liquid and pore structures, predicts that interface-reaction control is
parametric relationships for the densifica- more likely with small particles, which is
tion rate can be derived. Assuming the consistent with a simple geometrical analy-
pores are located at the edges or corners sis, in that a larger grain requires a longer
of a tetrakaidecahedron (TKD) grain diffusion distance from the grain contact
(Budworth, 1970; Wray, 1976), the driving to pore sites for densification. If grain
force can be determined from geometries growth occurs rapidly during this stage of
of a solid-liquid-vapor assemblage. LPS, the rate controlling mechanism could
In general, there are two rate limiting shift from interface reaction to diffusion
processes for solution-precipitation. When controlled.
material transport is limited by diffusion There are few critical analyses of the
through the liquid phase, the densification controlling densification mechanism in the
110 14 Liquid-Phase Sintering

solution-precipitation stage of LPS pri-

marily due to previous, oversimplified
models and difficulties in executing rigor-
ous experiments with respect to ideal mod-
els. Kwon and Messing (1989) analyzed
the solution-precipitation kinetics of an
alumina-glass system densified by iso-
thermal LPS using air-classified powders
of different particle sizes. The controlling
mechanism was determined by evaluating
the particle size dependency of the densifi-
cation rate based on the above models and
by determining the activation energy of the
process. Microstructural observation is
another method of confirming the occur-
rence of solution-precipitation. In order
to examine microstructural changes during
solution-precipitation, polished alumina-
MAS glass sections were heavily etched to
reveal the grain structures, as shown in
Fig. 14-7. In the initial stage of solution-
precipitation, Fig. 14-7 A, the particle con-
tacts are relatively narrow. However, in
the later stage of solution-precipitation,
Fig. 14-7 B, the particle contacts are sub-
stantially flattened, indicating that exten-
sive solution-precipitation was operative.

Figure 14-7. Grain structure of heavily-etched alu-

14.5.3 Pore Removal mina-MAS glass specimens showing (A) the initial
and (B) the later stage of solution-precipitation.
During the intermediate stage of sinter- Note that the grain contacts are significantly flat-
ing, interconnected pore channels pinch tened with densification (Kwon and Messing, 1990).
off to form closed pores in the density
range from 0.9 to 0.95, depending on the
material. In practice, this pore closure can
occur at a lower density by LPS than by where ap is the vapor pressure inside the
SSS. The final stage of LPS starts immedi- pores and rp is the radius of the pores. If rp
ately after pore closure. The closed pores and ap remain small (i.e., SD > 0), then den-
usually contain gaseous species from the sification will proceed. As the contacts be-
sintering atmosphere and vapor(s) from tween the solid particles flatten, the densi-
the liquid. After pore closure, the driving fication rate by solution-precipitation de-
stress for densification is creases. However, if rs increases due to
growth and/or coalescence of pores, and
27l, ap increases due to any internal reactions
(14-10)
causing gas evolution (for example, reduc-
 14.6 Grain Boundaries 111

tion of metal oxide and oxidation of resid- can increase as a function of pressure.
ual carbon), the driving force for densi- Consequently, if dense bodies, which are
fication could be negative, resulting in prepared by pressure-assisted densifica-
dedensification in some cases. tion, are heated to an elevated temperature
Several processes can simultaneously under atmospheric pressure, bloating and
occur during the final stage of LPS, includ- swelling of bodies can result as a function
ing growth and coalescence of grains and of heating (Kwon and Messing, 1989).
pores, diffusion of the liquid component(s)
into the solid, phase transformations, and
the formation of reaction products be- 14.6 Grain Boundaries
tween the solid, liquid, and gas. The lack of
critical experiments and models for these There are outstanding issues concerning
concurrent processes adversely affects the the structure of the thin liquid layers at the
predictability of the final stage densifica- grain boundaries in LPS systems and its
tion in LPS, i.e., the final density and mi- load-bearing capacity during LPS and
crostructure. high temperature deformation processes.
Pressure-assisted sintering techniques, Island structure (Raj and Chyung, 1981),
e.g., hot pressing and hot isostatic pressing semicrystalline interface (Marion et al.,
(HIPing), can be used to lower the sinter- 1987; Clarke, 1987), and hydrodynamic
ing temperature, to achieve a higher final squeeze film (Lange, 1982; Kwon and
density, and to produce a more homoge- Messing, 1991) models have been pro-
neous microstructure. Pressure-assisted sin- posed to date. The existence of a continu-
tering is often employed to prepare high ous liquid layer at the grain boundary, typ-
performance or optical quality compo- ically ~ 2 nm thick, has been reported in
nents. The applied stress (or pressure) en- various liquid-phase sintered ceramics on
hances the driving force for densification the basis of observations using the trans-
in all three LPS stages. During the final mission electron microscope (TEM).
stage of pressure assisted LPS, the driving As mentioned in the introduction
stress while under an applied stress is given (Sec. 1), one important premise for liquid-
by phase sintering is the existence of a liquid
at the grain boundaries, especially at
, — <?* (14-11) two-grain junctions. At some low-energy
boundaries, wetting (or penetration) by
where aa is the applied stress. The cra can be liquid is not energetically favorable, and so
as high as 400 MPa using hot isostatic solid-solid boundaries are retained. In
pressing (HIPing). Consequently, the up- other cases, the liquid film can disappear
per boundary of solution-precipitation from the boundary by vaporization or by
could move to a higher density as a func- diffusion into the grains forming solid so-
tion of the applied stress. The iso-density lution. However, a thin liquid film exists at
line EF in Fig. 14-3 can move toward the most two-grain boundaries in ceramics
line SA (Q = 1). In general, gases in the with a silicate grain boundary phase, as
pores can be dissolved in the liquid as a observed using the high resolution trans-
function of the gas pressure and tempera- mission electron microscope (HRTEM).
ture. While the pore size shrinks signifi- The thickness of the liquid film varies from
cantly, the gas pressure inside the pores 1 to 3 nm (10 to 30 A) in most alumina and
112 14 Liquid-Phase Sintering

silicon nitride ceramics prepared by LPS. oped a theoretical basis for the existence of
The nature of the liquid film at the two- an equilibrium liquid film thickness in
grain junction has been an interesting topic some ceramic systems. He analyzed a force
in recent publications along with improved normal to the boundary to explain the
observations using various transmission equilibrium liquid film thickness and mod-
electron microscopic techniques. For ex- eled it as a balance between an attractive
ample, Cinibulk et al. (1993) recently com- van der Waals dispersion force and a repul-
pared techniques using the TEM for deter- sive disjoining force due to distortions of
mining the thickness of the amorphous, SiO4-tetrahedra. Clarke (1989) further re-
intergranular film in Si 3 N 4 ; diffuse dark- ported that the film thickness for rapidly-
field imaging and defocus Fresnel fringe cooled Si 3 N 4 is greater than that for slowly
imaging vs. HRTEM. cooled, typically 2-10 nm compared to
By applying Reynolds' (1886) theory on ~ 1 nm for normally cooled. (Note that
lubrication, Lange (1982) proposed that a most reported film thicknesses were mea-
finite thickness of liquid film always re- sured for slowly cooled samples.) The film
mains between the particles, although the thickness at an elevated temperature is also
thickness decreases with time. Kwon and an important factor in relation to the high
Messing (1991) further developed the hy- temperature creep behavior of liquid-
drostatic squeeze film analysis. Equations phase sintered materials. The molecular
for the equilibrium thickness were derived structure of the intergranular liquid phase
for diffusion and interface reaction con- could be nonrandom. Its viscosity and dif-
trolled densification. In both cases, the fusivity might be markedly different from
equilibrium thickness is proportional to those properties measured on a bulk sam-
the solubility of solid in liquid. The analy- ple of the same liquid. Further understand-
sis further indicates that the film thickness ing of the factors affecting the film thick-
can vary as a function of the cooling rate ness and resulting creep behavior would
after sintering due to a strong temperature eventually make it possible to design and
dependency of the solubility. Local repre- manufacture an optimized material for
cipitation at solid-liquid-solid boundaries specific applications of interest.
is another factor that would reduce the In recent investigations, it has been ob-
observed film thickness. Estimation of the served that the liquid film can migrate by
equilibrium film thickness at a silicate changing the chemical composition of ce-
grain boundary, based on the derived ramic alloys. In the 8% Y 2 O 3 -ZrO 2 sys-
equations indicated that the thickness may tem with a small amount of silica as an
vary from 1 to 2 nm in an alumina-glass impurity, annealing at 700-1400 °C re-
system, depending on the densification sulted in enrichment in Y 2 O 3 content of
mechanism responsible. the cubic ZrO 2 grains, which was assisted
On the basis of observations of liquid- by liquid film migration (LFM) involving
phase sintered alumina, using the TEM, the ubiquitous silicate grain boundary
Marion etal. (1987) proposed that the phase (Chaim etal., 1986). The resulting
stress distribution needed to maintain the microstructure indicated a significant al-
chemical potential gradient in a thin liquid teration of the properties.
layer during LPS is associated with atomic
structuring of the liquid at the grain
boundaries. Clarke (1987) further devel-
 14.8 Use of Phase Diagrams 113

14/7 Grain Growth grain growth, a small concentration of ad-

and Coalescence ditives in the liquid can strongly influence
the kinetics and morphology of the grain
Grain growth in LPS systems is quite growth. For example, in the sintering of
different from that in SSS system in its alumina/glass, when CaO is added as a
controlling mechanisms. Mass transfer be- sintering additive with SiO2 in alumina,
tween grains occurs only through a liquid faster grain growth and more faceting in
phase, assuming good wetting of the solid the alumina result compared to when MgO
by the liquid. The liquid phase can either is added (Kaysser et al., 1987).
enhance or inhibit grain growth, depend- Extensive studies have recently been car-
ing on the system. In some cases, the grain ried out on anisotropic grain growth in
growth rate by LPS is much faster than liquid-phase sintered ceramics, e.g., alu-
that by SSS for a given solid due to a mina and silicon nitride. Optimization of
higher mass transfer rate through the liq- the grain growth can result in toughening
uid. In other cases, though, the liquid and R-curve behavior (see Chap. 12 in
phase can act as a grain growth inhibitor Volume 6 of this series). Nagaoka et al.,
(Bennison and Harmer, 1985; Kaysser (1992) recently determined the two and
etal., 1987). three-dimensional distributions of grain
In general, the grain growth of spherical size and the aspect ratio of silicon nitride
particles in a large amount of liquid is sintered at 1900°C for 2-24 h. The im-
given by (Greenwood, 1956; Lifshitz and provement in fracture toughness is attrib-
Slyozov, 1961; Wagner, 1961) uted to the increase in the volume and
number fractions of a particular elongated
(rjr-(r?y = kt (14-12) grain group in the 3-D distribution. The
where rs is the average radius of the grains increase in the fracture toughness is also
at time t, r° is the average radius of the explained by the increase in the area and
grains at time 0, and k is the rate constant number fractions of the grains in a particu-
for grain growth. The radius (or grain size) lar group in the 2-D distribution. Lai and
exponent, n, depends on the grain growth Tien (1993) measured the kinetics of an-
mechanism; n = 3 and n = 2 for diffusion isotropic j8-Si3N4 grain growth. It was
and interface reaction controlled, respec- demonstrated that the grain growth behav-
tively. Later, it was demonstrated that the ior of /?-Si3N4 grains follows the general
relationship is valid for a high volume frac- grain growth equation [Eq. (14-12)], with
tion of solid (Sarian and Weart, 1972). the exponents of 3 and 5 for the length
While the dissolution of solid in a liquid [001] and width [210] directions. These dif-
contributes to densification, the differen- ferences in the growth rate constants and
tial dissolution between different shapes exponents for the length and width direc-
and sizes of particles results in grain tions are attributed to anisotropy in the
growth by Ostwald ripening (Wagner, /J-Si3N4 growth during isothermal grain
1961; Buist et al., 1965). The dissolved sol- growth.
ute from small or sharp corners of particles
tends to reprecipitate on coarser particles. 14.8 Use of Phase Diagrams
Accordingly, coarse particles grow as fine
particles are eliminated. While the amount Information gained from phase equi-
of liquid can be a dominant variable for libria is critically important for under-
114 14 Liquid-Phase Sintering

standing LPS and for designing material combination of a base B (the major phase)
systems for LPS, because the amount of and an additive A (the minor phase).
liquid increases as the temperature in- Transport of B through liquid rich in A
creases above the eutectic temperature. In requires high solubility of the base mate-
addition, the viscosity of liquids is expo- rial in the liquid. A high solubility en-
nentially related to temperature, and diffu- hances the wetting of B by the liquid. In
sion in a liquid is much faster than it is in addition, the additive should have a low
a solid. As a result, LPS takes place rapidly solubility in the base. If the solubility of A
because both the amount of liquid and dif- in B is large, then depletion of the solid
fusion increase (or the viscosity decreases) and/or swelling may occur during densifi-
with increasing temperature (Johnson and cation. The decreasing liquidus and solidus
Cutler, 1970). lines indicate segregation of the liquid to
The phase diagram is also an important the interparticle regions. Segregation is
means of predicting systems with favor- also aided by selecting an additive A with
able characteristic for LPS (German, et al., an atomic (or ionic) size smaller than that
1988). Phase diagrams constructed using of B. A large melting difference between
computer simulations could provide a link the liquid and the base is essential for en-
between the thermodynamics and the sin- hanced diffusive transport of the base.
tering behavior. Although LPS is not an Thus the optimum phase diagrams for
equilibrium process, phase diagrams do LPS will have deep eutectics, and no
provide useful information with respect to high melting, intermediate components. It
the behavior of mixed components during should be noted that these characteristics
sintering. Obviously, the melting tempera- are identical to those desired for activated
ture influences the selection of a sintering sintering, except for a lower process tem-
temperature. Figure 14-8 shows an ideal perature. Compositions for LPS should be
system for LPS. The desired characteristics chosen away from eutectics such that the
are related to the diffusivity, solubility, volume fraction of liquid increases slowly
and segregation tendencies, as noted in the with increasing temperature. This will min-
phase diagram. The system is a binary imize the warpage and migration of the
liquid phase associated with temperature
gradients during heating.
temperature decreasing Figure 14-9 shows the ternary phase dia-
liquidus & gram for MgO-Al2O3-SiO2 (Osborn and
solidus
Muan, 1964) and displays many techno-
liquid logically important ceramics. Alumina
phase substrates for electronics and wear resis-
high tant components are manufactured by
solubility activated
low mixing a glass of composition G1 and alu-
solubility mina. Minerals in this phase diagram, e.g.,
kaolinite and talc, can also be used as raw
materials for the glass composition. Upon
Figure 14-8. The ideal binary phase diagram for LPS heating, the alumina content of the glass
showing the critical elements of solubility, diffusivity,
and segregation. LPS occurs in the two-phase field, at
will be increased and its composition will
a temperature slightly above that for activated sinter- change from G^to G2 at 1600 °C, if equilib-
ing (German et al., 1988). rium is reached. The diagram indicates
 14.8 Use of Phase Diagrams 115

SiO2

Two
Liquids

2MgOSiO2

MgO
Figure 14-9. Ternary phase diagram of the MgO-Al2O3-SiO2 system. The dashed lines show isotherms in
degrees centigrade at which liquid phase is present.

that corundum (a-Al2O3) is the only solid inhibit the crystallization of grain boundary
phase in equilibrium with the liquid at the phases. For example, mullite, spinel, sap-
composition G 2 . The optimum sintering phrine, and cordierite phases can be pre-
temperature can be estimated using phase cipitated in the grain boundaries of an alu-
equilibria and viscosity data. The amount mina-MAS glass system during controlled
of liquid in the final microstructure can cooling or by an appropriate post-sinter-
also be estimated from solubility data and ing heat treatment (Zdaniewski and Kirch-
the density of the liquid. ner, 1986). Powel-Dogan and Heuer (1990)
The phase diagram information can also reported extensive observations using the
be used to optimize cooling or heat-treat- TEM of 96 % aluminas after various heat
ment cycles after sintering to promote or treatments on the basis of phase diagram
116 14 Liquid-Phase Sintering

information. It has also been realized that It has been realized that the chemical
thermodynamic equilibrium is barely driving force due to various reactions can
reached during most experimental time be much larger than the driving force from
frames of sintering and heat treatments. interfacial energies in some multiphase
For silicon nitride ceramics, a number of LPS systems (Petzow and Kaysser, 1980).
efforts have been devoted to improving the The chemical reactions which occur in
mechanical properties by post-sintering most systems with a liquid phase provide a
heat treatments (Clarke et al., 1982; Bon- decrease in the free energy of the system of
nel etal., 1987). Complete crystallization typically between 100 and lOOOJmol" 1 .
of the two-grain boundaries has not been Since during the sintering of particles of
observed to date in both alumina and about 10 |im diameter a free energy reduc-
silicon nitride with viscous silicate grain tion of only 1 -10 J mol" 1 occurs due to a
boundaries. decrease in the interfacial energy, chemical
reactions are expected to have a pro-
nounced effect on the initiation and accel-
eration of mass transport processes. In
14.9 Reactive Liquid-Phase spite of the magnitude of the free energy
Sintering decrease caused by chemical reactions, it
must be noted that only a decrease in the
There are many ceramics in which a interfacial energy results in the driving
chemical reaction takes place during sin- force necessary for densification.
tering. These are comprised of mixtures of LPS is the only practical means to den-
powders of different phases, or coated par- sify Si 3 N 4 owing to its refractoriness and
ticles, or of a phase that undergoes a phase susceptibility to decomposition. A small
transformation during heat treatment. amount of liquid-forming additives such
One of the main interests in reactive sys- as MgO, Y 2 O 3 , A12O3 + Y 2 O 3 , BeSiN2,
tems is that the densification behavior be and other rare earth oxides is known to
"activated". Several types of reactive LPS be effective as a sintering aid. During
may occur, as listed in Table 14-3 (Coble, sintering, a silicate liquid phase, which
1982). Firing of whiteware bodies (e.g., promotes the a- to jS-Si3N4 transformation
porcelain) is a classical example of reactive at ~1800°C, provides the rapid mass
LPS. Densification typically occurs simul- transfer path (Bowen et al., 1976). In order
taneously with the reaction and dissolving to achieve full density, the temperature-
of raw materials that produce new glass time cycle must be optimized for the com-
and crystalline phases (Reed, 1988). ponent size to allow for gas transport from

Table 14-3. Types of reactive liquid-phase sintering

Reaction Example

(a) A(s) + B(s) -• A + liquid Al 2 O 3 + anorthite, Si 3 N 4 + MgO

(b) A(s) + B(s) + C(s) - » A + £ + liq. SiC + Al 4 C 3 + Al 2 O 3
(c) A(s) + B(s) -> A'+ liquid a-Si3N4 to /?-Si3N4
(d) A(s) + B(liq.) -> A' amalgam (transient liquid)
(e) A(s) + B(s) ^ A ^ B ^ liquid BaTiO,, PZT
 14.11 Real Powder Compacts 117

the pores (Greskovich, 1981). Complex the theoretical density of MgO, had a lat-
chemical reactions can also occur between tice parameter equal to that of MgO, and
the sintering atmosphere and the powder became colorless and transparent. The de-
compact at high temperatures, and are velopment of this transparency started in
partially responsible for incomplete densi- the center of the specimens as the transient
fication during LPS. Researchers have not liquid was evaporated from the center to
yet been successful in developing a robust- the surface. Hence, transient LPS is an al-
enough process for silicon nitride, after ternative route for fabricating fully dense,
25 years intensive developmental efforts, polycrystalline ceramics if direct solid-
primarily owing to its complicated reactive state sintering is difficult.
sintering coupled with insufficient under- Silicon carbide ceramics can be rapidly
standing of related phenomena. densified above approximately 1850°C
Sigle and Kleebe (1993) demonstrated due to a transient liquid phase resulting
that a core/rim structure can be developed from the reaction between alumina and
during the sintering of a-SiC with yttrium- aluminum oxycarbides. The resulting ce-
aluminum garnet (YAG). Chemical analy- ramics are fine grained, dense, and exhibit
sis by EDS/transmission electron micros- high strength at room temperature (Huang,
copy revealed that the rim has excess et al., 1986). The high temperature creep of
yttrium, aluminum, and oxygen, while this transient liquid-phase sintered a-SiC
these elements are missing in the core. The showed similar behavior to that of solid-
core/rim interface was found to be coher- state sintered a-SiC due to the absence of a
ent, and both the core and rim are com- grain boundary liquid phase at the two
posed of the same polytype. These results grain junctions (Jou et al., 1991). It is
suggest that the nonequilibrium phase in noted that, in this case, the grain boundary
the liquid boundary has precipitated onto liquid is not a silicate glass.
the undissolved, nonequilibrium particles
which constitute the cores. Consequently,
it is proposed that Ostwald ripening by 14.11 Real Powder Compacts
solution-precipitation controls the sinter-
ing mechanism in this system. In real powder compacts with different
particle shapes, sizes, and phase distribu-
tions, a number of complex processes may
14.10 Transient Liquid-Phase occur in addition to the foregoing idealized
Sintering processes. Shaw (1986,1993) reported, as a
result of 2-D and 3-D analyses, that the
A transient liquid phase has been used liquid can rapidly redistribute to a low en-
to take advantage of the enhanced densifi- ergy configuration in real powder com-
cation kinetics in some special systems. pacts. The analyses indicate that the distri-
The transient liquid may be removed by bution of the liquid phase is not homoge-
evaporation or remain in the compact by neous in real systems. The pores will try to
forming solid solution or crystalline reac- fill sequentially in order of increasing size
tion products with the solids. For example, as shrinkage occurs. It is also predicted
MgO specimens hot pressed with a small that, when the pore size distribution is
amount of LiF (Hart et al., 1970) and an- broad, a significant drop in capillary pres-
nealed at 1300°C for 3 h in air approached sure can occur during sintering due to an
118 14 Liquid-Phase Sintering

abrupt change in the pore size from small 1987). Additional rearrangement may also
to big pores. In practice, the development occur in liquid-rich regions with a loose
of a short sintering cycle would be pre- compact structure, as observed by the
ferred, e.g., several hours rather than a few particle-liquid mixture flow in different
days. Consequently, in most cases of LPS, material systems (Kim et al., 1987; Kang
thermodynamic equilibrium cannot be ob- et al., 1989). Figure 14-10 illustrates a typ-
tained. If the liquid becomes nonwetting or ical microstructural heterogeneity that can
coalesces during sintering, a heteroge- be created during LPS. Large pores may
neous microstructure may be produced. result from the burn-out of organic inclu-
The sintering atmosphere may influence sions or the melting of a liquid forming
the final stage densification. If gaseous particle, as shown in Fig. 14-10 A, and act
species are entrapped in closed pores, the as critical flaws for structural reliability.
diffusion of gases out of the pores to the The coarse pores may be filled with liquid
surface of the powder compact can control or a particle-liquid mixture, Fig. 14-10 B,
the densification as well as the size and as a result of extended sintering or pres-
distribution of the pores. It has been
demonstrated that changing of sintering
atmosphere for the final stage is effective
for achieving a higher final density, due to
the increased driving force for the transfer
of entrapped gases from the pores (Kim et
al., 1987; Kang et al., 1989; Burneburg,
1991). The sintering atmosphere can not
only alter the diffusivity of the controlling
atmosphere, but it can promote or inhibit
certain gas-forming reactions. If a gas-
forming reaction is operative during the
later stage of sintering, the bloating and
swelling of bodies is frequently observed
(Kwon and Messing, 1989). The liquid
grain boundary phase has a high solubility
for gases as well as providing a highly dif-
fusive path for gas transport. It has been
reported that 5% glass-alumina bodies
can be dedensified during containerless
HIPing in argon by gas diffusion and/or
gas forming reactions inside (Kwon, 1986).
In some cases, liquid penetration into
aggregates and agglomerates of polycrys-
talline particles can limit densification
(Petzow and Kaysser, 1980). Liquid pools
may appear in a sintered microstructure
Figure 14-10. (A) Macropores can be formed by the
due to heterogeneous liquid redistribution burning-out of an organic inclusion, and (B) liquid
and liquid pore-filling processes (Kwon may flow into a macropore, resulting in a liquid pool
and Yoon, 1980; Kwon and Messing, (Kwon, 1986).
 14.12 Outlook for the Future 119

sure-assisted densification (Kwon and A few shortcomings of LPS as a manu-

Messing, 1984; Kang and Yoon, 1989). facturing process are that ceramics densi-
The extent of overlapping of the densifi- fied by LPS are susceptible to shape distor-
cation mechanisms between rearrange- tion and sticking on setters. Therefore care
ment and solution-precipitation was de- must be taken to design an optimum com-
termined by measuring the activation position and a sensible firing cycle for
energies for densification as a function of LPS. For example, the viscosity of the liq-
density in an alumina-glass system (Kwon uid does not have to be high to resist
and Messing, 1990). Figure 14-11 shows slumping during sintering. In addition, the
the measured apparent activation energies reactions between the liquid, vapor, and
as a function of density. The two plateau atmosphere are vigorous due to the pres-
regions are equivalent to the activation en- ence of the liquid, often resulting in a sur-
ergies for rearrangement and solution- face layer with an appreciably different
precipitation, respectively. A broad over- composition and properties compared to
lapping zone for the controlling mechanism those of the bulk.
is noticed in this system between the two
plateaus due to deviations from the ideal
powder compact. Although an air-classi-
fied powder of narrow particle size distri- 14.12 Outlook for the Future
bution was used to minimize the deviation
from the ideality of the models, it appears Unlike well-established processing tech-
that the two mechanisms occur concur- niques for metal alloys, ceramic processing
rently over most density regimes of LPS. is being developed for extended future use.
As an important ceramic processing tech-
nique, the present understanding of LPS is
not sufficient for its broader use for ad-
vanced ceramics. Areas of future interest
\ I include: (1) a fundamental understanding
600 -
AI 2 0 3 -MAS glass of the liquid grain boundary on an atom-
P.S.=3.6jjm istic scale, (2) establishment of the critical
500
processing factors to tailor desired micro-
400 structures and optimized properties, and
(3) a further understanding of the final
300
(mAEr+nAEs) stage of LPS in order to prepare highly
< 200 reliable ceramic components. It is believed
that LPS is an important means of devel-
e = 0,075,0.10
100 oping a variety of ceramic alloys for engi-
I I
neering applications in the future. Tailored
0
0.6 0.7 0.8 0.9 1.0 microstructures and various properties can
Relative density be developed analogous to metal alloys.
Figure 14-11. Apparent activation energies as a func- The principles of LPS can also be applied
tion of relative density for samples of Vx — l.l and to exploit liquid-phase enhanced super-
10%, where, AE is the apparent activation energy, plastic forming (Wang and Raj, 1984; Wu
AEr and AES are the activation energies for rearrange-
ment and solution-precipitation, respectively, and m and Chen, 1992) as a future manufacturing
and n are fractions (m + n=l). technology.
120 14 Liquid-Phase Sintering

14.13 References Kang, S.-X, Yoon, K.-X (1989), J. Eur. Ceram. Soc. 5,
135-139.
Kang, S.-X, Greil, P., Mitomo, M., Moon, X H.
Bennison, S. J., Harmer, M. P. (1985), J. Am. Ceram. (1989), J. Am. Ceram. Soc. 72, 1166-1169.
Soc. 68, c22-24. Kaysser, W. A., Sprissler, M., Handwerker, C. A.,
Bonnel, D. A., Tien, T.-Y, Ruhle, M. (1987), /. Am. Blendell, X E. (1987), J. Am. Ceram. Soc. 70, 339-
Ceram. Soc. 70, 460-465. 343.
Bowen, L. J., Weston, T. X, Carruthers, T. G., Brook, Kim, J.-X, Kim, B.-K., Song, B.-M., Kim, D. Y,
R. J. (1976), Ceram. Int. 2, 173-176. Yoon, D. N. (1987), / Am. Ceram. Soc. 70, 734-
Budworth, D. W. (1970), Trans. Brit. Ceram. Soc. 69, 737.
29-31. Kingery, W. D. (1950), /. Appl. Phys. 30, 301-306.
Buist, D. S., Jackson, B., Stephenson, I. M., Ford, Kingery, W. D., Narasimhan, M. D. (1950), /. Appl.
W. R, White, J. (1965), Trans. Br. Ceram. Soc. 64, Phys. 30, 307-310.
173-209. Kwon, O.-H.(1986), Ph. D. Thesis, Pennsylvania
Burneburg, P. L. (1991), U. S. Patent 5 053 370. State University, University Park, PA.
Cahn, J. W, Heady, R. B. (1970), /. Am. Ceram. Soc. Kwon, O.-H., Messing, G. L. (1984), /. Am. Ceram.
7, 406-409. Soc. 68, C43-45.
Chaim, R. Heuer, A. H., Brandon, D. G. (1986), Kwon, O.-H., Messing, G. L. (1987), in: Sintering
/. Am. Ceram. Soc. 65, 243-248. '85: Kuczynski, G. C , Uskokovic, D. P., Palmer
Cinibulk, M. K., Kleebe, H.-J., Ruhle, M. (1993), III, H., Ristic, M. M. (Eds.). New York: Plenum,
/. Am. Ceram. Soc. 76, 426-432. pp. 203-218.
Clarke, D. R. (1987), J. Am. Ceram. Soc. 70, 15-22. Kwon, O.-H., Messing, G. L. (1989), /. Am. Ceram.
Clarke, D. R. (1989), J. Am. Ceram. Soc. 72, 1604- Soc. 72, 1011-1015.
1608. Kwon, O.-H., Messing, G. L. (1990), /. Am. Ceram.
Clarke, D. R., Lange, F. R, Schnittgrund, G. D. Soc. 73, 275-281.
(1982), J. Am. Ceram. Soc. 65, c51-52. Kwon, O.-H., Messing, G. L. (1991), Acta Metall. 39,
Coble, R. L. (1982), in: Sintering- Theory and Prac- 2059-2068.
tice: Kolar, D. (Ed.). New York: Elsevier Scientific, Kwon, O.-X, Yoon, D. N. (1980), in: Sintering Pro-
pp. 145-151. cesses: Kuczynski, G. C. (Ed.). New York: Plenum,
Eremenko, V. N., Naidich, Y. V., Lavrinko, I. A. pp. 165-171.
(1970), Liquid Phase Sintering. New York: Consul- Lai, K.-R., Tien, T.-Y (1993), /. Am. Ceram. Soc. 76,
tant Bureau. 91-96.
Ewsuk, K. G., Harrison, L. W. (1990), in: Sintering of Lange, F. F. (1982), /. Am. Ceram. Soc. 65, c-23.
Advanced Ceramics: Handwerker, C. A., Blendell, Lifshitz, I. M., Slyozov, V. V. (1961), Phys. Chem.
J. E., Kaysser, W. A. (Eds.), Westerville, OH: The Solids 19, 35-50.
American Ceramic Society, pp.436-451. Marion, J. E., Hsueh, C. H., Evans, A. G. (1987), /
German, R. M. (1985), Liquid Phase Sintering. New Am. Ceram. Soc. 70, 708-711.
York: Plenum. Nagaoka, T, Watari, K., Yasuoka, M., Hirao, K.,
German, R. M. Farooq, S., Kipphut, C. M. (1988), Kanzaki, S. (1992), J. Ceram. Soc. Jpn. 100, 1256-
Mater. Sci. Eng. A1051106, 215-224. 1260.
Greenwood, G. W. (1956), Acta Metall. 4, 243-248. Osborn, E. F , Muan, A. (1964), in: Phase Diagrams
Greskovich, C (1981), J. Am. Ceram. Soc. 64, 725- for Ceramists: Levin, E. M., Robins, C. R., Mc-
730. Murdie, H. F. (Eds.). Westerville, OH: The Ameri-
Hart, P. E., Atkin, R. B., Pask, J. A. (1970), /. Am. can Ceramic Society, p. 246.
Ceram. Soc. 53, 83-86. Petzow, G., Kaysser, W A. (1980), in: Science of
Heady, R. B., Cahn, J. W. (1970), Metall. Trans. 1, Ceramics, Vol. 10. Deut. Keram. Gesellschaft, pp.
185-189. 269-278.
Hiemenz, P. C. (1977), Principles of Colloid and Sur- Powell-Dogan, C. A., Heuer, A. H. (1990), J. Am.
face Chemistry, New York, Marcel Dekker, pp. Ceram. Soc. 73, 3670, 3677, and 3684.
235-238. Price, G. H. S., Smithelles, C. X, Williams, S. V.
Huang, J.-L., Hurford, A. C , Cutler, R. A., Virkar, (1938), J. Inst. Met. 62, 239-264.
A. V. (1986), J. Mater. Sci. 21, 1448-1456. Raj, R., Chyung, C. K. (1981), Acta Metall. 29, 159-
Huppmann, W. J., Riegger, H., Kaysser, W. A. 166.
Smolej, V., Pejovnik, S. (1979), Z. Metallkd. 70, Reed, X S. (1988), Principles of Ceramic Processing.
703-713. New York: Wiley, p. 464.
Johnson, D. L., Cutler, I. B. (1970), in: Phase Dia- Reynolds, O. (1886), Trans. R. Soc. London 177,190-
grams: Materials Science and Technology: Alper, 234.
A. M. (Ed.). New York: Academic, pp. 265-291. Riebling, E. F. (1964), Can. J. Chem. 42, 2811-2821.
Jou, Z. C , Virkar, A. V., Cutler, R. A. (1991), /. Ma- Sarian, S., Weart, H. W. (1966), J. Appl. Phys. 37,
ter. Sci. 6, 1945-1949. 1675-1681.
 14.13 References 121

Shaw, T. M. (1986), J. Am. Ceram. Soc. 69, 27-34. Finney, J. L. (1970), Proc. R. Soc. London A 319,
Shaw, T. M. (1993), J. Am. Ceram. Soc. 76, 664-670. 479-507.
Sigl, L. S., Kleebe, H.-J. (1993), /. Am. Ceram. Soc. Froschauer, L., Fularath, R. M. (1976), /. Mater. Sci.
76, 773-776. 11, 142-149.
Towers, H. (1954), Trans. Brit. Ceram. Soc. 53, 180- Hildebrand, J. H., Scott, R. L. (1950), The Solubility
202. of Nonelectrolytes, 3rd ed. New York: Reinhold,
Wagner, C. (1961), Z. Elektrochem. 65, 581-591. pp. 270-299.
Wang, J.-G., Raj, R. (1984), /. Am. Ceram. Soc. 67, Huppmann, W. I , Riegger, H. (1975), Acta Metall.
399-409. 23, 965-971.
Wray, P. J. (1976), Acta Metall. 24, 125-136. Kingery, W. D., Niki, E., Narasimhan, M. D. (1961),
Wu, X., Chen, L-W. (1992), J. Am. Ceram. Soc. 75, /. Am. Ceram. Soc. 44, 29-35.
2733-2741. Prill, A. L., Hayden, H. W, Brophy, J. H. (1965),
Zdaniewski, W. A., Kirchner, H. P. (1986), Adv. Trans. Metall. Soc. AIME 233, 960-964.
Ceram. Mater. 1, 99-103. Rhines, F. N. (1978), in: Ceramic Processing Before
Firing: Onoda, G. Y, Hench, L. L. (Eds.). New
York: Wiley, pp. 321-347.
Stephenson, I. M., White, J. (1967), Trans. Br. Ceram.
General Reading Soc. 66, 443-483.
Warren, R., Waldron, M. B. (1972), Powder Metall.
Eremenko, V. N., Naidich, Yu. V., Lavrinko, I. A. 15, 166-201.
(1970), Liquid Phase Sintering. New York: Consul- Whalen, T. J., Humenik, Jr., M. (1967), in: Sintering
tants Bureau. and Related Phenomena: Kuczynski, G. C, Hooton,
German, R. M. (1985), Liquid Phase Sintering. N. A., Gibbon, C. F. (Eds.). New York: Gordon
New York: Plenum Press. and Breach, pp. 715-746.
15 Vitrification
Francis Cambier

Centre de Recherches de l'lndustrie Beige de la Ceramique, Mons, Belgium

Anne Leriche

Laboratoire des Materiaux/Avances Ceramiques, Universite de Valenciennes et du

Hainaut-Cambresis, Maubeuge, France

List of Symbols and Abbreviations 124

15.1 Introduction and Definition 126
15.2 General Considerations 126
15.2.1 Ceramic Materials Which Undergo Vitrification 126
15.2.2 Phenomena Occurring During the Vitrification of Classical Ceramics . . . . 127
15.3 Role of Viscosity and Surface Tension in Vitrification Processes 127
15.3.1 Surface Tension 128
15.3.2 Viscosity 129
15.3.3 Vitrification Paths 130
15.4 Description of Theoretical Models for Vitrification 131
15.4.1 Models for Determining the Densification Mechanism 132
15.4.1.1 Kuczynski's Method 132
15.4.1.2 Herring's Method 132
15.4.2 Models Describing Vitrification Kinetics from Geometrical Assumptions 133
15.4.2.1 Frenkel's Model 133
15.4.2.2 Frenkel's Model Modified by Clasen 134
15.4.2.3 Mackenzie and Shuttleworth's Model 136
15.4.2.4 Scherer's Model 137
15.4.2.5 Conclusions Related to the Geometrical Models 139
15.4.3 Models Describing Vitrification Kinetics from Phenomenological
Assumptions 139
15.4.3.1 Ivensen's Model 139
15.4.3.2 Anseau, Cambier, and Deletter's Model 141
15.4.3.3 Conclusions Relating to the Phenomenological Models 142
15.4.4 Comparison of the Models 142
15.5 References 144

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
124 15 Vitrification

List of Symbols and Abbreviations

a cylinder radius
al9a2,.. ,,at area of contact
A see Eq. (15-4)
B seeEq. (15-5)
c r sinjS
dm measured diameter
e seeEq. (15-23)
E{ rate of energy dissipation in viscous flow
Em molar energy gap
Es energy supplied by the reduction in surface area
F3 cross-sectional area
h height of cylinder
partial molar enthalpy
i see Eq. (15-26)
j r cos/?
k,k' constants
K see Eq. (15-37)
l,l0 edge length of unit cell, initial length
L,L0 length of sample, initial length
m constant determining sintering mechanism; constant
mt molar percentage of phase /
Mu molar mass of structural unit
n coordination number
n
P
number of pores
Nh number of holes
P pressure
Pu pressure tensor element
d starting porosity, after time t, after time td
P* jump probability
P, vacancy probability
q e + xs; constant [(l/F p0 ) dKp0/d/]
r neck radius
R gas constant
r,R radius of spheres or cylinders after time t
initial radius of spheres or cylinders
s strain
S surface
So calculated surface
s1 surface of sphere segment
S3 half surface of the neck
sm partial molar entropy
total surface
Mo time of heat-treatment, initial time
 List of Symbols and Abbreviations 125

starting time of vitrification step

T temperature
V bulk volume at time t
AV/V0 volume shrinkage
v, displaced volume of spherical particle
v2 half-volume of neck
y3 half-volume of torus segment
vo,vd initial apparent volume, such volume at time td
initial pore volume, pore volume
pore volume at time t
K volume of solid phase
K molar volume of structural unit
X neck radius
Xl all
xF radius of contact area in Frenkel model
xs radius of gravity center of torus segment
radius of contact without neck
y half the shrinkage of the compact
z r cos/?
a constant
/} complementary angle to y
y angle between symmetry axis and end-point of neck
r , Ft surface energy or tension, for oxide /
f/, rj0 viscosity, initial viscosity
6 angle, see Fig. 15-10
X constant
£ constant determining sintering mechanism
Q true density
Q' relative density
£ 0 , qt bulk density (initial and after time t)
QS theoretical density
126 15 Vitrification

15.1 Introduction and Definition occurs in certain reaction sintering pro-

cesses using zircon (ZrSiO4) as a raw mate-
The most common process used to fabri- rial. At high temperatures (>1500°C, de-
cate ceramic materials involves the form- pending on the impurities), zircon dissoci-
ing of suitable powders into a desired ates to give amorphous silica which can
shape followed by heat-treatment of the react with other oxides, leading to the for-
shaped powder at a temperature below the mation of a viscous liquid silicate.
melting point of the principal constituents. Sintered glasses are fabricated by vitrifi-
The result of this heat-treatment is densi- cation. Indeed, glasses have finite viscosity
fication of the powder shape leading to a above the glass-transition temperature,
reduction in the porosity, to shrinkage, and and are then capable of viscous flow. In
to an improvement in the mechanical resis- materials of this type, vitrification is com-
tance. monly not pushed to its limit so that a
During the heat-treatment, a series of desired level of porosity remains. Typical
physical changes and chemical reactions examples are the soda-lime glasses, the
occur in the compacted powder. These of- borosilicate glasses, fluorine containing
ten lead to the production of a liquid phase glasses, and bioglasses. Certain glass ce-
which plays a major role in the achieve- ramics also fall into this group: for these,
ment of densification. Where the volume of two processes occur simultaneously during
liquid is sufficient to fill all the pores in the processing, namely densification by vis-
compacted powder, complete elimination cous flow and devitrification of the glass.
of the porosity can occur by flow of the The classical ceramics constitute the
liquid into the pores under the influence of main group of materials which undergo
surface tension forces, resulting in a dense, vitrification. They include earthenware,
pore-free, solid product. vitrified tiles, sanitary ware, domestic and
Although strictly applicable in the case industrial porcelains, cordierite-based
of glass formation, the term vitrification is products, and some traditional refracto-
currently applied to all densification pro- ries. The main systems involved are SiO 2 -
cesses where liquid flow at high tempera- A12O3-XO, where XO represents an oxide
tures is sufficient to bring about complete or mixture of oxides such as Na 2 O, K 2 O,
densification. It is also usual to call this MgO, CaO, Fe 2 O 3 , and TiO 2 . The quan-
phenomenon "viscous flow sintering". tity of SiO2 is relatively large, the silica
generally being present both in the free
state (as quartz brought in the form of
sands) and in a combined form in the clays.
15.2 General Considerations The presence of different oxides (XO) is due
both to impurities in the raw materials and
15.2.1 Ceramic Materials Which Undergo
to the deliberate addition of fluxing agents.
Vitrification
Classical fluxing agents are feldspars of
Vitrification is the principal mechanism sodium (albite and nepheline), or potas-
of densification in most traditional ceram- sium (orthoclase).
ics and in sintered glasses. It also plays a
role in some technical products in which a
large amount of transitory liquid phase is
produced during heating. For instance, it
 15.3 Role of Viscosity and Surface Tension in Vitrification Processes 127

1 2 3
Figure 15-1. Schematic representation of the phe- (3) amorphous phase of clay composition, (4) dis-
nomena occurring during vitrification of a classical solution of part of the quartz into the amorphous
ceramic body: (1) quartz (hatched) and clay partic- phase, and (5) crystallization during cooling (dotted).
les, (2) dehydration and collapse of clay lattice,

15.2.2 Phenomena Occurring During with the amorphous phase and partly dis-
the Vitrification of Classical Ceramics solve in it, so modifying its composition
and its flow characteristics and leading
As a consequence of the nature of the
sometimes to the crystallization of new
constituents in classical ceramic bodies, a
phases.
range of phenomena, generally sequential
5) Finally, during cooling further crystal-
but sometimes simultaneous (schemati-
lization may occur.
cally represented in Fig. 15-1), occur dur-
ing heat treatment:
1) Changes occur in the constituent miner- 15.3 Role of Viscosity and Surface
als, independent of the presence of the Tension in Vitrification Processes
other constituents in the body; these in-
clude phase transformations (i.e., OL+±P Vitrification is driven by the energy re-
quartz, dehydration of clays, or collapse of duction brought about by the reduction in
the clay lattice). surface area of the porous body. Various
2) A liquid phase is formed when the kiln attempts have been made to describe the
temperature exceeds (a) the melting tem- phenomenon in terms of energy balance
perature of one particular constituent (i.e., and deformation. These models (details are
feldspars), (b) the eutectic temperature of a given in Sect. 15.4) show the particular im-
mixture corresponding to grains in con- portance of three variables, i.e., a geometri-
tact, or (c) the glass-transition temperature cal factor: the particle size, a kinetic factor:
of a pre-existing amorphous phase.
3) At a sufficiently high temperature, the
amorphous phase has a reduced viscosity
which allows amorphous particles to flow.
Following the process in Fig. 15-2, this re-
sults in the formation of necks between
joined particles, and an increased contact
area between them, thus reducing the
porosity and therefore the air amorphous
phase surface area of the compact. Vitrification time
4) As a function of time at a sufficiently Figure 15-2. Viscous flow of two joined amorphous
high temperature, solid particles can react particles as a function of vitrification time.
128 15 Vitrification

the viscosity, and a thermodynamic factor: and experimental data for some binary
the surface tension. SiO2-XO glasses.
From the same table, it is possible to
approximate the surface tension versus
15.3.1 Surface Tension
temperature variation by subtracting or
The surface tension is the force required adding 0.004 N/m when the temperature
per unit length to extend a liquid surface; it decreases or increases by 100 °C, respec-
is equivalent to the work required to form tively. Again, good correlation is obtained,
a new surface of unit area, i.e., the surface as can be seen in Fig. 15-4, for silica glass.
energy. Measurement of the surface ten-
sion at high temperatures is not easy; vari-
15.3.2 Viscosity
ous methods have been proposed based on
the measurement of the tearing forces Table 15-1 and Figs. 15-3 and 15-4 show
when a metallic wire, ring, or plate is pulled that for silicate ceramics the surface ten-
from the liquid surface. sion is weakly dependent on both the corn-
An alternative method is to approxi-
mately calculate the surface tension of an
amorphous phase starting from its chemi- Table 15-1. Tension coefficients of some oxides at
cal composition. The surface tension of a 1300°C.
glass is high if the attraction forces between Oxide
its constituents are high. For example, pure
silica glass made up of strongly bonded 0.01
SiO 4 tetrahedra shows a high surface ten- PbO 0.12
TiO 2 0.25
sion. When increasing amounts of Na 2 O 0.29
SiO2
are added within the SiO 2 glass, the surface 0.295
tension stays almost constant, whereas 0.45
K 2 O leads to a sharp decrease and Li 2 O to CaO 0.51
a dramatic increase. This can be related to MgO 0.52
A12O3 0.58
a decrease in the binding force due to the
higher polarization of K + , whereas Li + is
less polarized and acts in an opposite way.
0.6
It has been shown by Tillotson and
Oppen (Jouenne, 1990) that the surface _ 0.5 -
SiO 2 - C a O
tension (F) can be expressed as a linear
^0.4
function of the composition
SiO2 - N a 2 O^^^f
I 0.3
SiO 2 - P b O
(15-1) 45 0.2
""SiOa-KaO"^
100
0.1 -

where T{ is the tension coefficient corre- i i 1 1

sponding to each oxide i and m{ is its molar 10 30 50 70 90

SiO 2 (mol %)
percentage. Table 15-1 gives the tension
Figure 15-3. Surface tension of binary XO-SiO2 mix-
coefficient values for some oxides present tures: full lines correspond to experimental data and
in classical ceramics, whereas Fig. 15-3 dashed lines to calculations (adapted from Kingery,
provides a comparison between calculated 1960).
 15.3 Role of Viscosity and Surface Tension in Vitrification Processes 129

be fulfilled: an energetic condition (the

availability of sufficient energy to jump
into a free site, P,) and a geometric condi-
tion (free volume, i.e., the presence of a hole
to receive the structural element, Pv).
Therefore the viscosity (rj) is expressed as
1
rjcc (15-2)
1000 1400 1800 2200 P P
J v
Temperature (°C)
Figure 15-4. Silica surface tension vs. temperature: Using the statistical approach of Weymann
circles are experimental data, line is calculation (1962), a viscosity-temperature (T) relation
(adapted from Kingery, 1960). can be developed
n = ATe(1000BIT) (15-3)
position and the temperature. In contrast,
the viscosity is much more sensitive to where A and B have a physical significance
these parameters. Most laboratory equip-
ment for the measurement of viscosity con-
sists of coaxial cylinders between which the
fluid is placed (Coulette's device). By sim- and
ply computing the applied torque, the rota-
tion speed, and the dimensions of the cylin- 1000B = (15-5)
R
ders, it is possible to calculate the viscosity.
However, at high temperatures, accurate Here R is the gas constant, M u and Vu are
measurements are difficult, time consum- the molar mass and volume, respectively,
ing, and expensive. Therefore several em- of the structural unit, Em is the molar en-
pirical methods have been proposed for the ergy gap to release this structural unit, Hm
estimation of viscosity. These include the and Sm are the partial molar enthalpy and
linear additive method (Bottinga, 1972) entropy, respectively, associated with hole
covering a restricted range of composi- formation, and Nh is the number of holes.
tions, and a more accurate model proposed A relation exists between A and B and
by Urbain et al. (1981) based on a general can be written as
physical background. -\nA = kB + kf (15-6)
The main feature of Urbain's approach
is to describe high-temperature liquids in where k and k! are typical of the family of
terms of structural elements, i.e., polyan- liquids considered, these being defined as
ions, the size of which is based on the rela- having the same type of bonds between
tive proportions of glass former (Si4 + , their structural elements. Figure 15-5
Ge 4 + , p 5 + , etc.), modifier (Na + , K + , Ca 2 + , shows the linear relation obtained for
etc.), and amphoteric ions (Al3 + , Fe 3 + ), liquid silicates where k = 0.253 and
which are either glass formers or modifiers k' = 12.263.
depending on their relative proportions. Equation (15-6) makes it possible to
As viscous flow involves relative move- completely calculate the viscosity-temper-
ment of the structural elements in the liq- ature dependence of a silicate liquid com-
uid, two simultaneous conditions have to position if either A or B are known. Since
 130 15 Vitrification

27.5 15.3.3 Vitrification Paths

22.5 ; A good example of vitrification behavior
is that of kaolin, a mineral composed pre-
17.5
* * *
dominantly of kaolinite clay, quartz sand,
^ 12.5 and certain impurities (K 2 O, Na 2 O). Dur-
ing heat-treatment, dehydration of the
7.5 -
kaolinite occurs progressively at tempera-
2.5 _ tures of up to about 700 °C, leading to the
t i l l i i i i i i i
5 10 15 20 25 30 35 40 45 50 55 formation of an amorphous phase, the
B composition of which corresponds to the
Figure 15-5. Plot of -In A versus B for 57 different kaolinite SiO 2 /Al 2 O 3 ratio of 2/1. Above
liquid silicates. 1000°C, mullite (SiO 2 /Al 2 O 3 = 2/3) is the
first phase to crystallize, which increases
the silica content in the amorphous phase.
It has been observed that where the fluxing
better accuracy can be obtained for B, iso- oxide content is low (typically Na 2 O and
B diagrams have been proposed by Urbain K 2 O), secondary crystallization of cristo-
etal. (1981) and Deletter et al. (1984). A balite occurs, dramatically reducing the sil-
typical example is given for the ternary ica content of the amorphous phase; on the
CaO-SiO 2 -Al 2 O 3 system in Fig. 15-6. Ap- other hand, where the flux content is suffi-
proximations for more complicated com- ciently high, secondary crystallization does
positions can be otained by computer cal- not occur. This phenomenon leads to dif-
culation. ferent vitrification paths within the ternary

SiO2 SiO 2

MO [0.013] AI 2 O 3 MO [0.026 ] AI2O

A B

/1Q12-.14-!
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 MO [0.063 ] AI 2 O 3
CaO — A/(CaO) AI 2 O 3 C

Figure 15-6. Iso-B curves for the SiO 2 -Al 2 O 3 -CaO Figure 15-7. Vitrification paths determined at
system. N is the molar ratio (courtesy of Urbain 1150°C for various atomic contents of Na([]) for
etal., 1981). kaolin (Cambier et al., 1984).
 15.4 Description of Theoretical Models for Vitrification 131

6
Onion DDDDC

5 ^ °°DDD.

a? 4 \

| 3 o = Na 0.013
\
O
D = Na 0.026
§ 2-
• = Na 0.063 Figure 15-8. Viscosity (log scale) versus
1- time for different fluxing oxide content,
as determined for a typical kaolin vitrifi-
I i i i cation (1 poise = 0.1 Pa s).
12 16
Time (h)

system SiO2-Al2O3-XO (XO being the semblies of spherical, viscous particles

fluxing oxides). whose dimensions and physical proper-
A typical evolution is given in Fig. 15-7 ties (surface energy and viscosity) remain
for increasing amounts of Na 2 O from A constant during processing. A schematic
to C. diagram of the sintering of two spheres is
As a consequence of the existence of dif- presented in Fig. 15-10.
ferent vitrification paths, the variation of - the other based on a phenomenological
viscosity with time can be dramatically dif- description for the full span of the densi-
ferent even for the same materials, as can fication process.
be seen in Fig. 15-8. It should be noted that
In addition to these models, various
the total volume of amorphous phase is
methods were proposed some time ago by
also quite different in the three examples
Kuczynski (1949) and Herring (1950)
considered.
which make it possible to distinguish on
the basis of the densification kinetics
whether the main active mechanism is vit-
15.4 Description of Theoretical rification.
Models for Vitrification
The kinetics of densification or porosity
removal in a ceramic which is undergoing
vitrification are described by the general
form of curve shown in Fig. 15-9. The
porosity P falls and the extent of shrinkage
AV/V0 rises as a function of time.
Among the large number of models in
the literature, it is possible to distinguish
between two different approaches:
Time
- one based on geometrical assumptions Figure 15-9. General form of the shrinkage curve en-
that relate to idealized systems, i.e., as- countered during vitrification.
132 15 Vitrification

the same composition, the results of which

are presented in Fig. 15-11, have shown
that viscous flow controls glass sintering.

15.4.1.2 Herring's Method

Herring (1950) considers two partially
sintered clusters of spherical, amorphous
grains which are geometrically similar, the
linear dimensions of the one being X times
those of the other. Figure 15-12 shows a
schematic representation of two clusters.
Each cluster consists of spherical grains of
Figure 15-10. Schematic diagram of the sintering of
radius rt sintered until the radius of the
two spheres, where r0 is the initial radius of the area of contact between the adjacent grains
sphere, r the radius after time /, x = (r2) — r2)112, and is at.
y = ro-r. The surface energy (y) of a spherical drop
of liquid of radius .R produces a pressure
(p) increase in the interior of the drop given
15.4.1 Models for Determining the by
Densification Mechanism 2nRy = nR2p (15-8)
15.4.1.1 Kuczynski's Method From this relation, the pressure can be cal-
culated as p = 2y/R. This equation can be
From kinetic considerations, Kuczynski extrapolated to a curved surface of any
(1949) proposes a method allowing distinc- shape using pressure tensors. The stress
tion between the main sintering mecha- tensors related to the two-grain clusters
nisms. Matter is transported in four ways: presented in Fig. 15-12 are, respectively,
a) evaporation-condensation, b) volume 2y
diffusion, c) surface diffusion, and d) vis- (15-9)
cous flow (vitrification).
The kinetic contributions of these phe-
nomena can be described by a common
relation linking the neck radius (x) to the 715°C/
sintering time (t) •
0.2
•
(15-7) 5 675°C #.

The value of m depends on the main 0.1 - /

• •—'
mechanism responsible for sintering: ra = 2
for viscous flow (vitrification), m = 3 for
evaporation-condensation, m = 5 for lat-
tice diffusion, and m = l for surface diffu-
L i
50
Time (min)
i

100
i

150

sion. 2
Figure 15-11. (x/r) versus time plots for glass
Sintering experiments with glass grains spheres bonded to glass plates at various tempera-
placed on a glass plate of approximately tures.
 15.4 Description of Theoretical Models for Vitrification 133

R>N fa, !

Figure 15-12. Schematic representation

of grain clusters defined by Herring
System 1 System 2 (1950).

and 15.4.2 Models Describing Vitrification

Kinetics from Geometrical Assumptions
(15-10)
15.4.2.1 Frenkel's Model
and can be correlated by
Frenkel (1945) was the first to propose a
(15-11) model describing the sintering of viscous
Py(2) =
X materials. By equating the energy variation
due to surface decrease to the energy dissi-
Thus, assuming viscous behavior, the
pated by viscous flow, Frenkel provides an
strain rates dstj/dt at corresponding points
equation which describes the first stage of
will satisfy
vitrification
(15-12) AV 9y
dt k dt (15-15)
4rir0
where k is a constant.
where Vo is the initial volume, y the surface
So, by defining Atx the time necessary
tension, r\ the viscosity, r0 the initial parti-
to realize a given fractional dimensional
cle radius, and t the time of heat-treatment.
change in the first system, and At2 the time
According to this model, the volume
to realize the same change in the second
shrinkage is proportional to the duration
system, At2 is linked to Atx by
of the thermal treatment. The kinetic con-
At2 = kAt1 (15-13) stant depends on the surface tension, glass
viscosity, and powder grain size.
By applying this analysis to other mecha- Most experimental results and notably
nisms, Herring (1950) proposes the general those on clays suggest that the range of
relation application of this equation is very limited.
At2 = X*At± (15-14) This led Lemaitre and Bulens (1976) to
modify the Frenkel equation (Frenkel,
with £ as a constant depending on the sin- 1945) on the basis of the hypothesis that
tering mechanism: £ = 1 for viscous flow the viscosity of the material increases lin-
(vitrification), £ = 2 for evaporation-con- early with time during processing as fol-
densation, £ = 3 for volume diffusion, and lows rj = rjo(l+at% where a is a constant.
£ = 4 for surface diffusion. On replacing rj in Frenkel's relation
134 15 Vitrification

[Eq. (15-15)], forfc= 9y/4r 0

(15-16)
AV/V0 k k
It has to be noted that this linear expres-
sion cannot properly describe the final
stage of shrinkage.

15.4.2.2 Frenkel's Model Modified

by Clasen
In the work by Clasen (1989), the viscous Figure 15-13. Definition of the parameters for calcu-
flow of single particles is represented by the lation of the interaction between spherical particles
decrease in the surface area of the compact during sintering, after the described modified Frenkel
model. y = overlap along the symmetry axis with re-
during sintering. The calculations were
spect to the symmetry plane (half the shrinkage of the
performed based on an improved Frenkel compact); R = radius of the sperical particle; r = radi-
model. This new model is based on the us of the neck between particles in contact; ^ = surface
following assumptions: of the displaced segment of the particle; S3 = half-
surface of the neck formed between two particles;
V1 = displaced volume of the spherical particle as a
i) The glass particles are incompressible result of approach of a second particle; V2 = half-
and the volume displaced at the approach volume of the neck formed between two particles;
of two particles contributes exclusively to V3 = half-volume of the torus segment; xF = radius of
the formation of the neck, the contact area in the Frenkel model; xy = radius of
ii) The particle radius R remains un- the contact between particles without neck; xs =
radius of the gravity center of the torus segment;
changed. 7 = angle between the symmetry axis and the end-
iii) The neck surface is a curved cylinder point of the neck; p = complementary angle to y.
and can be described by the radii r, xF. Quantities defined for calculations: c = rsin/?;
iv) The number of contact points to neigh- q = e + xs; e is defined in Eq. (15-23); z = R cosfi;
boring particles (the coordination number) j=r cosfi (courtesy of Clasen, 1989).
depends on the green density.
with
The parameters needed for the calcula-
c = r sin/?,
tions are defined in Fig. 15-13.
The volume V± (segment of a sphere) is z = R cos/?,
easily calculated by (15-19)
and
(15-17)

The volume V2 is composed of a cylinder

which gives
from which a torus segment V3 has to be
subtracted V2 = nr R2 sinj?(l -sin 2 j5) - \ r sin£
6
n
, = C71Z*- I -
(15-18) -V3 (15-20)
 15.4 Description of Theoretical Models for Vitrification 135

Calculation of V3 is done according to

Guldin's rule (Smirnov, 1975), which states
that volume results from the production of
the cross-sectional area F 3 and the distance
resulting from rotation of the center of
gravity around the axis of rotation
2 4 6 8 10 12 14 16
2V,= x2K (15-21)
Neck radius r (nm) •
where xs is the position of the center of Figure 15-14. Calculated correlation between the
gravity, and can be calculated from overlap y of the particles and the neck radius r (cour-
tesy of Clasen, 1989).
x s = z + rcos/J — e (15-22)
with
2c{r2-lAc2)
e = (15-23)

By inserting the center of gravity value, V3

can be expressed as
20 40 60 80
Solid matter content (vol. %) —
Figure 15-15. Correlation between the coordination
2pn number of spherical particles and the solid matter
(15-24)
T80" content in the sample (courtesy of Clasen, 1989).

Therefore V2 can now be calculated from

Eq. (15-20). As Vx = V2, two equations with St of the sphere segment
two unknown variables exist which can be
solved by computer iteration. ST = S + nS3 -nS1 (15-25)
In Fig. 15-14, the results are plotted for with n denoting the coordination number.
particles with an average radius of 40 nm. The surface of the neck is obtained from
It can be seen that Frenkel's approxima- Guldin's rule by
tion y = r is only valid for a small overlap
S3 = n2rq — 2nrqi — 2%r2 cosz (15-26)
of the particles.
We also have to take into account the where i is arccos sin /?. As the relations be-
number of contact points (the coordina- tween y, r, and y are already known, it is
tion number) which depends on the green easy to calculate the ratio of the initial to
density of the compact, as shown in the instantaneous surface of the compact
Fig. 15-15. as a function of linear shrinkage. The re-
With the coordination number as a pa- sults are independent of the particle size
rameter, the surface area of an arrange- but the coordination number plays a sig-
ment of spheres can be calculated. Basi- nificant role.
cally, the total surface ST contributed by Figure 15-16 compares the measured
one particle in the compact is given by the and calculated surfaces S/So for different
surface S of the particle and the half-sur- coordination numbers for monodispersed
face S3 of the neck reduced by the surface silica particles and polydispersed particles.
136 15 Vitrification

their higher density, these agglomerates

sinter more rapidly, so reducing the parti-
cle surface.
In conclusion, Clasen's modified model
(1989) gives acceptable agreement between
the calculated and measured values only
for monosized particles. The deviations for
compacts made from polydispersed parti-
cles are best interpreted by the existence of
small agglomerates.
Figure 15-16. Comparison of the measured and cal-
15.4.2.3 Mackenzie and Shuttle worth's
culated surfaces S/So (thin lines) for different coordi-
nation numbers (S = calculated sample surface, Model
So = calculated standard surface). The compacts were
Mackenzie and Shuttleworth (1949)
made from monodispersed silica particles (•) and
from Degussa Aerosil OX50 with Q' = 40% (•) and have provided a theory for the closing
0'=45% (A) (courtesy of Clasen, 1989). stages of densification when the pores can
be considered as isolated spheres set in a
matrix of the viscous phase. Following
From these results, it can be noted that their assumptions, the solid material in-
while the deviation for the monodispersed cludes identical spherical pores as shown
particle compact is relatively small, a con- in Fig. 15-18.
siderable discrepancy is found for both of They consider a pore of radius r1 sur-
the polydispersed compacts with increased rounded by a spherical shell of incompres-
sintering. Consequently, during sintering sible material of radius (r2 — r1). The exter-
the number of contact points of a particle nal medium's properties are assumed to be
increases, which could be due to an in- identical to those of the "pore-shell" sys-
crease in the particle size [which is in con- tem. The system is schematically presented
tradiction to assumption (ii) and the exper- in Fig. 15-19.
imental results presented in Fig. 15-17], or Such a model is only acceptable if the
to the existence of small agglomerates r1/r2 ratio is much less than unity, i.e., if no
(which seems to be more probable). Due to interaction occurs between the pores. The

0.05 0.10 0.15 0.20

AZ.//-0 -
Figure 15-17. Measured diameter (dm) of monodis- Figure 15-18. Spherical pores of radius rx embedded
persed particles at different shrinkages AL/L0 during in a viscous matrix.
sintering (L = length of the sample, Lo = original
length of the sample) (courtesy of Clasen, 1989).
 15.4 Description of Theoretical Models for Vitrification 137

\VEquivalent homogeneous continuum

where P is the volume porosity.

In accordance with the hypothesis, this
model is only applicable to the removal of
closed porosity, that is to say, to the final
densification step. However, good agree-
ment has been found between theory and
experiment for densification at relative
densities of between 0.7 and 1.0, i.e., includ-
ing a span in which the pores are open.
An example of the results is given in
Fig. 15-20.

15.4.2.4 Scherer's Mode!

Figure 15-19. Representation of the "pore-shell" sys-
tem. In the previous models, the correlation
between the shrinkage and the microscopic
surface energy effect is similar to an inter- interactions can only be obtained for the
nal pressure inside the pores of — 2y/ru total sintering regime with assumptions
and sintering is described by the decrease which simplify the real particle arrange-
in the pore radius. The rate of closing of the ment. Scherer (1977, 1984, 1986, 1987) re-
pores is calculated by equating the energy placed the model of individual spherical
dissipated by the flow of the material in the particles by chains of particles which were
shell to the work done by surface tension. approximated by cylinders forming an ar-
Considering the solid to have Newton- ray of cubic cells. Figure 15-21 shows the
ian viscosity, we have cubic array of a unit cell.

drx 1 (15-27)
2flQ'
where Q' is the relative density. The pore
volume for unit volume is expressed by
4
(15-28)
Q '3

where np is the number of pores. From this

expression, the T1(Q') function can be de-
fined and derived. The derived function
can be combined with Eq. (15-27), leading
to the Mackenzie and Shuttleworth rela- Figure 15-20. Densification of a soda-lime silica
tions glass. The solid lines are calculated from Eq. (15-29).
The initial rates of sintering indicated by the dashed
(15-29) curves are calculated from Eq. (15-15) (courtesy of
Kingery, 1960).
138 15 Vitrification

Figure 15-21. Microstructural model

consisting of cylinders in a cubic array:
(a) Unit cell showing edge length /;
(b) model of low-density microstructure
(relative density ^0.05); (c) model of
microstructure with relative density of
0.50 (courtesy of Scherer, 1984).

The rate of sintering of the model struc- ergy supplied by the reduction in surface
ture is calculated following Frenkel's meth- area is given by
ods (Frenkel, 1945) by equating the rate of
energy gain from the reduction in the sur- dS
(15-32)
face area to the energy dissipation in vis- dt
cous flow. In this way, the complex calcula-
where y is the surface energy. Assuming
tions of the stress tensors can be carried
dEf/dt= -dEs/dt leads to
out giving a relation between the dimen-
sions of the cylinders in the cubic cell and
(15-33)
the surface tension, the viscosity of the
glass, the increase in the green density, and
the sintering time, respectively. where x , « a/I.
The derivation of this last relation can The density, Q, of the cell is given by
be summarized as follows. In the model,
the rate of energy dissipation in viscous (15-34)
flow, Ef, as a cylinder decreases in height is
given by with QS and Vs the theoretical density and
the volume of the solid phase. Recognizing
d£ f dh
(15-31) that Vs is constant and using the density
h ~dt
expression leads to
where r and h are the radius and height of
the cylinder, and r\ is the viscosity. For the
cell, r = a and h = l-[Sy/2/(3n)]a. The en-
w= (15-35)
 15.4 Description of Theoretical Models for Vitrification 139

where Q0 is the initial density and /0 is the

1
initial value of /. Substituting Eq. (15-35) Symbol Temp. (°C
into Eq. (15-33) and integrating leads to 1115 /
0.8
• 1165 /**
0.6 O 1296 o
dx (15-36)
o 1327 /
0.4
where
0.2
K= (15-37)
til
This equation determines xt&a/l as a -tn)
function of time; since Q/QS is a function of Figure 15-22. Relative density versus reduced time
only xt(t\ the density of the cell is deter- for SiO2 soot preform fired in ambient air (courtesy
mined as a function of time. of Scherer, 1977).
After integration, a theoretical curve is
obtained; this is compared with experimen-
tal data for silica in Fig. 15-22. It has to be der compacts. This variation can be ex-
noted that the reduced-time expression pressed by the Vpt/Vp0 ratio (pore volume at
used in the Mackenzie-Shuttleworth anal- time t/initial pore volume), which can be
ysis (Mackenzie and Shuttleworth, 1949), is found from the bulk densities (Q0, @t) and
equivalent to that of the Scherer (1977) the true density (Q) as follows:
analysis when the unit cell contains a single
£o x (Q-Qt)
pore. This model has also been extended to (15-38)
Qt (Q-Qo)
the special cases of a porous glass layer on
a rigid substrate and of composites con- Ivensen derived the following empirical
taining rigid inclusions. equation [Eq. (15-39)], which has been veri-
fied in the sintering of metal and carbide
15.4.2.5 Conclusions Related to the powders,
Geometrical Models Kn
(15-39)
All the different approaches focus on a
specific, idealized geometry for only one of where q and m are constants; q is the re-
the sintering stages, which strongly limits lative rate of pore volume reduction [(1/
the applicability of such models in actual Vp0)dVp0/dt] and m is a dimensionless con-
situations. It therefore remains desirable to stant characterizing the intensity of the de-
derive models describing the entire sinter- crease in the densification rate.
ing process. Equation (15-39) reflects the combined
effect of the "geometric factor" (change in
15.4.3 Models Describing Vitrification pore geometry and volume) and the "sub-
Kinetics from Phenomenological structural factor" (change in defect concen-
Assumptions tration). Ivensen (1970) assumes that the
only difference between the densification of
15.4.3.1 Ivensen's Model
crystalline and amorphous bodies is that
Ivensen (1970) investigated the change in with crystalline powders the densification
pore volume during the sintering of pow- is simultaneously influenced by both fac-
140 15 Vitrification

tors, whereas the densification of amor-

phous powders reflects the effect of the geo-
metrical factor alone.
Investigations into the change in pore
volume of glass powder compacts hjas
demonstrated that the rate of the process is
directly related to the pore volume
Al/
-kVv (15-40)

After integration 10 15
Time (h)
In Kp_. =
_
-kt (15-41) Figure 15-24. Plots of - l n ( j y Fp0) versus time for
"pO glass powders: 61% SiO2; 15% A12O3; 24% CaO
(mol%) sintered at 900 °C (Leriche et al, 1983).
where k is a constant determined by both
the viscosity of the substance at the investi-
gation temperature and the geometric ried out by Leriche et al. (1983) has shown
characteristics of the initial pores. that the Ivensen relation (Ivensen, 1970)
When curves of log Vp/Vp0 were plotted successfully describes all the sintering
against time, as shown in Fig. 15-23, the steps, as shown in Fig. 15-24.
plots were found to be approximately lin- Zagar (1975) has proposed a derivation
ear for all sintering temperatures (with the of Ivensen's relation from the usual model
exception of an initial period of sintering of two spheres (Ivensen, 1970) by consider-
lasting from 1 to 2 h). ing the motion of space between them in-
A comparison of the applicability of the stead of the motion of material during the
different equations describing vitrification sintering. Figure 15-25 represents the
to ternary glasses (CaO-SiO2-Al2O3) car- model of two grains at the sintering time
f = 0 (Fig. 15-25 a) and at the sintering time
t = t (Fig. 15-25 b).
The pore volume at the initial time can
be derived by subtracting the volume of the
solid material (two hemispheres of radius
r0) from the total volume of the system
ABCD, as shown in Fig. 15-25. The pore
volume is then

Vp0 = 2nrl-^nrl=~nrl (15-42)

After a sintering time t, the system changes

0 1 2 3 4 5 6 to A'B'C'D' and the pore volume now
Time (h) equals
Figure 15-23. Plots of log(Fp/Fp0) as a function of Vpt = (2nrf-2nrfy) (15-43)
time for glass powder compacts at different tem-
peratures: (I) 580, (II) 590, (III) 600, (IV) 610, and
(V) 620 °C (courtesy of Ivensen, 1970).
 15.4 Description of Theoretical Models for Vitrification 141

B'

Figure 15-25. Zagar's sintering model (Zagar, 1977).

It should be noted that Zagar (1977) as- (1977) takes the right-hand term of the
sumes that: i) the sphere volume decreases equation as an expansion of an exponen-
with time, and ii) the sphere centers ap- tial function giving the Ivensen equation
proach each other. However, he does not (Ivensen, 1970). Let us remark that such an
take into account the solid volumes in- approximation induces higher inaccuracy
volved in these two phenomena which con- as the Vpt/Vp0 ratio becomes small.
tribute to the pore volume decrease.
Neglecting the term with y 3 and assum- 15.43.2 Anseau, Cambier, and
ing that y is much smaller than rr, Zagar Deletter's Model
(1977) obtains
Anseau et al. (1981) developed a model
based on a hypothesis linking densification
- nrf — nrtx (15-44) to the fluidity of the viscous phase resulting
in the expression
Assuming, as did Kuczynski (1949), that 1 dV _ 1
during the period t the original radius r0 of (15-47)
the particle does not change appreciably
V~dt ~ ~ w
(rt = r0), it may be deduced that where V is the bulk volume and k is a con-
stant. Integration of this equation gives
2 (15-45) , i i
VPo z1 r0 In dt (15-48)
According to Frenkel (1945) for viscous
with AV= V- Vd9 where Vd is the bulk vol-
flow, x2 equals 3ytro/(2r]). Thus we obtain
ume at the time td from which viscous flow
completely controls densification. Finally,
(15-46)
4 t]r0 1
in''100 — dt (15-49)
In this relation, y, rj, and r0 can be consid- loo- pj in(t)
ered as constants for glass particles. Zagar where Pd is the volume porosity at time td.
142 15 Vitrification

In the case of clays, the viscosity of the — •' # #

*1150°C
liquid phase changes with time and the . _# •-
authors calculate it from the instantaneous -0.3 - 1200°C
composition [Urbain's method (Urbain
etal., 1981)] at each stage during process- <k -0.2 - 1250°C
r
ing. The models of Lemaitre and Bulens
(1976) and Anseau et al. (1981) are alone in
-0.1
considering this change. Equations (15-48)
and (15-49) satisfy the sintering data for
i
kaolin or illite-kaolinite mixtures (Cam- 5x103 10*
bier etal., 1984). Figure 15-26 shows the A 6t
application of Eq. (15-48) to sintering data
for kaolin at different temperatures. Figure 15-26. Equation (15-48) applied to the sinter-
The application of Eq. (15-47) to glass ing data of kaolin (Cambier et al., 1984).
sintering implies dF= dF p because of a con-
stant matter volume and no viscosity
change. As shown in Fig. 15-27, Leriche
etal. (1983) also obtained a good fit with
experimental data.

15.4.3.3 Conclusions Relating to the

Phenomenological Models
Due to the absence of restrictive as-
sumptions, the phenomenological models
have shown themselves to be capable of
describing all the sintering stages.

15.4.4 Comparison of the Models

The first two models presented, from
Figure 15-27. Plots of ln(^/F 0 ) versus time for glass
Kuczynski (1949) and Herring (1950),
powders: 77% SiO 2 ,12% A1 2 O 3 ,11% CaO (mol%),
make it possible to distinguish the densifi- sintered at 1000°C (Leriche etal., 1983).
cation kinetics of the main active mecha-
nism. When viscous flow controls the den-
sification kinetics, the other models de- sion decrease slow the densification rate.
scribing vitrification using various as- Some models, called geometrical ones, in-
sumptions can be applied. All these last clude these parameters in the kinetic equa-
models agree about taking into account tion; others, called phenomenological
similar parameters such as the particle ra- ones, include them inside the kinetic con-
dius, the liquid phase viscosity, and the stant and do not impose restrictive kinetic
surface tension. Moreover, these physical hypothesis. We note that all the kinetic
parameters act in the same way for all the equations issued from these models can be
proposed theories: the viscosity and parti- applied to describe the densification of
cle size increases, as well as the surface ten- glass powders but in different, more or less
 15.4 Description of Theoretical Models for Vitrification 143

Table 15-2. Practical forms of the different equations: Y= Kt.

Y K Model Shrinkage range applicability

following VJVp0
(tested on pure glasses)

Frenkel 1.0-• 0.7 initial stage

(1945)
AV Lemaitre and - initial and intermediate
Bulens stages
(1976)
Computer iteration Clasen (1989) initial stage?

Mackenzie and 0.6 -» 0.0 final stage

Shuttleworth
(1949)
Scherer — intermediate stage?
(1977)
Qo(Q-Qt) Ivensen 1.0-• 0.13 all stages
In
1
Qt(Q-Qo) (1970)
Zagar 1.0 -> 0.35 initial and intermediate
(1977) stages
Anseau et al. 1.0-• 0.16 all stages
(1981)

restrictive densification ranges. The practi- - Mackenzie and Shuttleworth (1949)

cal forms of these different kinetic equa- consider spherical pores isolated in a
tions are presented in Table 15-2 with the matrix of the viscous phase.
corresponding application ranges.
The geometrical models are derived However, the actual system geometry is
from Frenkel's assumption (Frenkel, 1945) not as simple because of continuous
that the rate of strain (densification) can be changes in the neck shape during sintering.
deduced by equating the rate of change in The mathematical description of neck
the surface energy to the rate of energy shape is then difficult and necessitates
dissipation. The models differ only in the severe simplifications valid only for limited
assumed geometry of the particles: stages of sintering, so:

- Frenkel based models (Frenkel, 1945) - Frenkel's expression (Frenkel, 1945)

consider spherical particles in contact closely describes the initial stage of sin-
and assume that the change in distance tering during which there is considerable
between the centers of two spheres neck growth but little densification.
equals the linear shrinkage of a powder - Clasen's modified model (Clasen, 1989)
compact. gives a large discrepancy for polydis-
- Scherer's model (Scherer, 1977) is based persed particles.
on cylinders forming an array of cubic - Scherer's model (Scherer, 1977) de-
cells. scribes the intermediate stage of sinter-
144 15 Vitrification

ing during which the pores are intercon- Cambier, R, Ilunga N'Dala, Deletter, M., Anseau,
M. R. (1984), Silk. Ind. 49, 57.
nected. Clasen, R. (1989), Glastech. Ber. 62 (7), 234.
- Mackenzie and Shuttleworth's model Deletter, M., Cambier, R, Ilunga N'Dala, Urbain, G.
(Mackenzie and Shuttleworth, 1949) is (1984), Br. Ceram. Trans. J. 83, 108.
Frenkel, J. (1945), /. Tech. Phys. Leningrad 9, 385.
limited to the final stage of sintering dur- Herring, C. (1950), /. Appl. Phys. 21, 301.
ing which the pores are isolated. Ivensen, V. A. (1970), Powder Metall. (USSR), 4, 20.
Jouenne, C. A. (Ed.) (1990), Traite de Ceramiques et
In contrast, phenomenological models Materiaux Mineraux, 5th ed. Paris: Septima,
such as those of Ivensen (1970), and p. 568.
Kingery, W. D. (1960), Introduction to Ceramics.
Anseau-Cambier-Deletter (Anseau et al, New York: Wiley.
1981) can describe all the sintering stages Kuczynski, G. C. (1949), J. Appl. Phys. 20, 1160.
within one equation. These models do not Lemaitre, J., Bulens, M. (1976), Clay Miner. 11, 313.
Leriche, A., Pilate, P., Anseau, M. R., Leblud, C ,
completely express the kinetic constant Cambier, F. (1983), Rev. Int. Hautes Temp. Refract.
and limit restrictive hypothesis. Zagar's 20, 25.
model (Zagar, 1975, 1977) proposes an Mackenzie, J. K., Shuttleworth, R. (1949), Proc.
expression of Ivensen's kinetic constant Phys. Soc. B. 62, 833.
Scherer, G. W. (1977), /. Am. Ceram. Soc. 60, 236.
(Ivensen, 1970) deduced from a geometrical Scherer, G. W. (1984), /. Am. Ceram. Soc. 67, 709.
model of two spheres. It should also be Scherer, G. W. (1986), /. Am. Ceram. Soc. 69, C206.
noted that the Anseau et al. (1981) model Scherer, G. W. (1987), J. Am. Ceram. Soc. 70, 719.
Scherer, G. W, Bachman, D. L. (1977), /. Am. Ceram.
takes into account the glass viscosity dur- Soc. 60, 239.
ing vitrification as does the model of Scherer, G. W, Garino, T. (1985), /. Am. Ceram. Soc.
Lemaitre and Bulens (1976). Hence only 68, 216.
Smirnov, W. I. (1975), Lehrgang der hoheren Mathe-
these two models can satisfactorily de- matik, Vol. 1. Berlin: Verlag Dt. Wissenschaften,
scribe the sintering of the clays, during p. 284.
which the viscosity of the glassy phase Urbain, G., Cambier, R, Deletter, M., Anseau, M. R.
(1981), Trans. J. Br. Ceram. Soc. 80, 139.
changes. Weymann, H. D. (1962), KolloidZ. Polym. 181, 131.
The application of these various equa- Zagar, L. (1975), Sci. Sintering 7 (1), 35-43.
Zagar, L. (1977), in: 4th Int. Round Table Conf on
tions to the densification data of glasses Sintering, Dubrovnik, Yugoslavia, pp. 57-64.
and clays makes it possible for the ceramist
to predict sintering conditions and under-
stand the behavior of clay products during
firing. General Reading
Bever, M. B. (Ed.) (1986), Encyclopedia of Materials
Science and Engineering. Oxford: Pergamon.
15.5 References Jouenne, C. A. (Ed.) (1990), Traite de Ceramiques et
Materiaux Mineraux, 5th ed. Paris: Septima.
Anseau, M. R., Deletter, M., Cambier, F. (1981), Kingery, W. D., Bowen, H. K., Uhlmann, D. R.
Trans. J. Br. Ceram. Soc. 80, 142. (Eds.) (1976), An Introduction to Ceramics, 2nd ed.
Bottinga, Y (1972) Am. J. Sci. 272, 438. New York: Wiley.
Cambier, K, Deletter, M., Anseau M. R. (1981), Rev. Urbain, G. (1985), "Viscosity of Silicate Melts. Mea-
Int. Hautes Temp. Refract. 18, 57. sure and Estimation", /. Mater. Educ. 7, 1007.
16 Hot Isostatic Pressing
Hans T. Larker and Richard Larker

Division of Engineering Materials, Lulea University of Technology, Lulea, Sweden

List of Abbreviations 146

16.1 Introduction 147
16.2 Equipment for Hot Isostatic Pressing 148
16.3 HIP Technologies for Ceramics 150
16.3.1 Direct HIP Methods 150
16.3.1.1 Direct HIP in a Pre-Shaped Container 151
16.3.1.2 Direct HIP with Shape Determined by Green Body 152
16.3.2 Post-HIP of Sintered Green Bodies 157
16.3.3 Quasi-Isostatic HIP Methods 158
16.4 Influence of Gaseous Species 159
16.4.1 Cladless HIP 159
16.4.2 Direct HIP 160
16.4.3 Special Processing Options 160
16.5 Theoretical Tools and Quality Assurance 161
16.5.1 Developments in HIP Theory 161
16.5.1.1 Modeling of HIP Mechanisms 161
16.5.1.2 HIP Phase Diagrams 162
16.5.2 Quality Assurance 163
16.6 Some Applications and Their Characteristics 163
16.7 Diffusion Bonding 164
16.7.1 Modeling of Diffusion Bonding 165
16.7.2 Parameters Influencing Diffusion Bonding 166
16.7.3 Diffusion Bonding by HIP 166
16.7.4 Diffusion Bonding of Ceramics to Metals by HIP 167
16.7.4.1 Importance of Ceramic/Metal Joining 167
16.7.4.2 Difficulties Associated with Ceramic/Metal Joining 168
16.7.4.3 Origin of Residual Stresses in Ceramic/Metal Joints 168
16.1 AA Concepts Suggested for the Reduction of Residual Stresses 168
16.7.4.5 Formation of Reaction Layers During Joining 169
16.7.4.6 Reaction Layers Formed During Diffusion Bonding by HIP 169
16.7.4.7 Effects on Microstructure from Joining Heat Treatments 172
16.7.5 Diffusion Bonding of Ceramics to Ceramics by HIP 173
16.7.5.1 Joining of Ceramic Green Bodies by HIP 173
16.7.5.2 Joining of Solid Ceramic Bodies by HIP 173
16.8 References 174

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
146 16 Hot Isostatic Pressing

List of Abbreviations
AEM analytical electron microscopy
APU auxiliary power units
CIP cold isostatic pressing
CTE coefficient of thermal expansion
CVI chemical vapor infiltration
EDS energy dispersive X-ray spectroscopy
f.c.c. face-centered cubic
FEM finite element method
FGM functionally graded materials
HIP hot isostatic pressing
HPCS high pressure self-combustion sintering
m Weibull modulus
MMC metal matrix composites
NNS near net shape
ROC rapid omnidirectional consolidation
SEM scanning electron microscopy
T.D. theoretical density
 16.1 Introduction 147

16.1 Introduction can be efficiently reduced in size and/or

appearance in most materials. Large pro-
Hot isostatic pressing (HIP) is an en- cessing pores at three granule junctions,
abling technology providing a most effi- which do not shrink in normal sintering,
cient method for the densification of ce- disappear in the early stages of encapsulat-
ramic powders, also allowing production ed HIP (J.-Y. Kim etal., 1992). Ceramics
of net-shape ceramics with superior and can therefore, with the use of encapsula-
consistent properties. A further option is tion, be densified at relatively low temper-
solid-solid, solid-powder or powder- atures. The reduced sintering temperature
powder joining of ceramics to ceramics or enables control or even avoidance of grain
metals. growth and undesirable reactions. A very
Engineered ceramic materials are char- high uniformity of properties as well as
acterized by very strong interatomic freedom from directionality can, if desired,
bonds. These bonds provide ceramics with be obtained.
desirable properties such as high hardness, The absence of shear forces in the body
stiffness, and strength in applications re- (on a scale substantially larger than parti-
quiring elevated temperatures for pro- cle size) and of any die friction, combined
longed periods of time. However, the few with a pressure level 5-10 times higher
slip systems available contribute to make than in uniaxial hot pressing, makes HIP
the material brittle. A consequence of the an even more powerful densification meth-
strong bond is that ceramic components od for ceramics than hot pressing. Even
are often very difficult to fully densify and ceramics extremely difficult to sinter can
to machine to the required final shape. be fully densified. High purity silicon ni-
In hot isostatic pressing, no rigid tools tride powders without any sintering addi-
with limited strength (such as graphite tives at all are one example, ceramic/ce-
tools in uniaxial hot pressing) are needed ramic composites with high loading of par-
to transmit the pressure to the body, thus ticles, whiskers or long fibers are another.
allowing high pressure levels to be chosen, Actually the ability to eliminate large
typically 100-320 MPa (15-50 ksi), even defects, which have a particularly detri-
above 2000 °C and in large size industrial mental influence on properties, was used
equipment. for the first widespread commercial appli-
The gas pressure is applied perpendicu- cation of HIP in the early 1970s. Large
larly to - and with the same magnitude on rolls of cemented carbide (a cermet) were
- all accessible surfaces of the body. It pro- post-densified (Nilsson, 1974). As the
vides a multiplication of the driving force parts had already reached over 99% of the
for densification, relative to the forces de- theoretical density (T.D.) after pressureless
veloped by the common sintering mecha- sintering, the remaining porosity was well
nisms related to surface tension on pore separated from the surface. The high pres-
surfaces. In contrast to these, the contribu- sure gas (argon) could therefore be allowed
tion of the gas pressure to "effective pres- to act directly on the surface of the part,
sure in particle contacts" (see Fig. 16-3 and a very good effect on elimination of
and Sec. 16.3.1.2 below) is not influenced internal pores was found.
by particle size. Similarly processed as in this example,
In HIP generally all the voids, particu- "post-HIP" of ceramics has widespread
larly large ones (Evans and Hsueh, 1986), use, not least for oxide-based ceramics.
148 16 Hot Isostatic Pressing

The material can in this case be processed Joining of ceramics to metals by diffu-
without the need for separate encapsula- sion bonding can also be improved by
tion. Several additional effects of pressure HIP. The high isostatic pressure forces the
can, however, be taken advantage of if metal to plasticize locally at the joint inter-
pressure is used throughout the shrinkage face at lower temperatures than are needed
and densification stages - from the green using traditional uniaxial pressure in vacu-
powder body to the fully dense part - the um. Reduction of bonding temperature
"direct HIP" methods. can be advantageous in several aspects,
Hot isostatic pressing is thus a very ver- such as reduced level of residual stress dur-
satile process with many advantages. ing cooling (originating from large ceram-
Commercially most important, however, ic/metal mismatch in thermal expansion),
may be the ability to control the size and reduced formation of reaction layers (also
shape of the product to a very high preci- favored by the closed system inside the
sion without costly diamond machining capsule), and retained microstructure in
operations. Under ideal conditions no the metal part. Furthermore the joint de-
change of shape (just a change of scale) of sign is not restricted to the planar butt
the body occurs. HIP has, as a matter of geometry.
fact, an inherent ability to produce parts
with exceptionally accurate shape. There is
also virtually no dimensional or shape lim- 16.2 Equipment for Hot Isostatic
itation. Pressing
The precision in shape and size, of
course, depends on consistency in process- Within the powder body to be pro-
es before sintering such as powder prepa- cessed, equipment for hot isostatic press-
ration and green body forming (injection ing must be able to simultaneously main-
molding, slip casting, etc.). For HIP, a tain a high, isostatically acting pressure
variation in green density between differ- and a high temperature for a desired length
ent parts of a green body - or between of time - in a well controlled way. Today
individual green bodies - generally does practically all equipment is of the cold
not result in a product with approximately pressure vessel wall, internal furnace type
the intended shape but varying residual (Fig. 16-1). It resembles other much more
porosity, but rather in a fully dense prod- common equipment: the cold wall, high
uct with an altered shape and size. A con- temperature, high vacuum furnace. In
sistent density gradient in the green bodies both cases the pressure difference to ambi-
can, however, be compensated for by mod- ent is taken by the cooled wall of the equip-
ifying the shape and size of the green body ment. The temperature in the workspace is
correspondingly. Flexible glass powder produced by electrical furnaces, usually
based encapsulation applied on injection with resistance heating and with a thermal
molded, highly reproducible powder green insulation tailored to the requirements.
bodies is now used in high series produc- At the low pressure in high vacuum fur-
tion of precision parts, e.g., for textile ma- naces, which is typically ten orders of mag-
chinery. Contrary to the general concep- nitude below ambient, only heat transport
tion, it is for some applications the low by radiation needs to be considered. How-
cost alternative! (H. T. Larker et al., 1993). ever, for pressures common in HIP equip-
ment, typically 3 to 3 V2 orders of magni-
 16.2 Equipment for Hot Isostatic Pressing 149

Heating elements and gas-tight sheats

Insulation cover for convection control in the heat insula-
Insulation mantle tion are typically made of refractory
metals such as molybdenum for uses up to
Heater
1400 °C, and of graphite, graphite paper
Wire-wound and graphite/graphite composites at
vessel
higher temperatures. Platinum is used in
|-Work piece
some research units operating with partial
oxygen atmospheres, e.g., oxygen/argon
(20%/80%).
Temperature measurement is usually by
means of platinum/rhodium thermocou-
Closure ples (Pt-6 Rh/Pt-30 Th) up to 1750°C and
Figure 16-1. Typical construction of a cold wall, in- tungsten/rhenium up to 2000 °C. The ther-
ternal-type HIP furnace. Electrical feedthrough for mocouple life is, however, short at higher
power and temperature control, as well as ducts for
the high-pressure gas, are provided through the end
temperatures (>1600°C and >1800°C,
closures (courtesy of Asea Brown Boveri). respectively), and the read-out may be un-
certain, especially in nitrogen gas. Particu-
larly important for thermocouple life is the
hide over ambient (100-320 MPa), the quality of thermocouple insulators.
gases used to transfer the pressure (usually HIPed, high purity boron nitride, which
argon but sometimes nitrogen) become can be made with very low boron oxide
very dense. In areas near ambient tempera- content, appears to be a good choice. B4C/
ture, e.g., just inside the pressure vessel C thermocouples can be used up to 2400 °C
wall, the density of the gas may actually (longer term up to 2200 °C) (Hunold,
exceed the density of water, however, still 1986), and an all-graphite system based on
not be a liquid. The combination of high mechanical principles can be used up to
gas density gradients in combination with 3000 °C (longer term 2700 °C) (Traeff,
low gas viscosity at high temperatures (less 1990).
than a factor of two higher at 200 MPa The furnace system with temperature
than at atmospheric pressure) makes the control devices is probably the technically
tendency to convection extremely strong. most critical part of HIP equipment. Most
Control of gas convection is conse- important for safety reasons are, however,
quently crucial, but based on principles the pressure vessel and high pressure gas
proposed by H. T. Larker (1966) and Boy- installation.
er and Orcutt (1967), heat losses by con- The safety codes as prescribed by au-
vection can be mastered even in large in- thorities in different countries must of
dustrial units. Already a relatively thin course be strictly followed, as the stored
heat insulation wall, leaving a large avail- energy in HIP vessels under pressure is
able workspace within a given pressure very high. The pressure vessel may be
vessel size, can be effective in controlling made with thread-type closures for small
convection heat losses. Heat conduction size or lower pressure applications. For
can then become dominant and determine large volumes or high pressure applications,
the minimum thickness of the heat insula- yoke closures are preferred (Fig. 16-2), and
tion wall. vessels with wire-wound cylinders and
150 16 Hot Isostatic Pressing

yokes can be made in "leak before break"

designs (Haerkegaard et al., 1984).
Typical tooling for the supporting of ce-
ramic parts to be processed by HIP is made
of graphite. Often cylindrical trays are
used, for small furnaces to the full
workspace diameter of the unit and for
larger diameter units seven trays in each
plane. On surfaces which may come in
contact with the ceramic, the graphite is
covered by a boron nitride wash, some-
times in combination with molybdenum
foils.

16.3 HIP Technologies for

Ceramics
HIP has in many cases been found to
provide unique solutions for problems re-
lated to densification and shaping of ce-
ramics. The process utilizes uniform and
omnidirectional pressure at elevated tem-
Figure 16-2. HIP furnace in a pressure vessel incor-
perature to enhance densification and in-
porating a yoke closure. The wire-wound QUINTUS
pressure vessel is of leak-before-break design (cour- ter-member bonding, usually of particles,
tesy of Asea Brown Boveri). as schematically illustrated in Fig. 16-3.
An inert gas (usually argon for cost and
availability reasons) is used to transmit
pressure. The gas must be prevented from
penetrating into any voids, which are to be
eliminated, of the processed body.
Effective Pressure in
Particle Contacts
There are several different ways to pro-
duce shaped parts by HIP, but based on
Force between Particles how the gas is prevented from entering the
Contact Area voids of the powder body, they can be
grouped into two categories, "direct HIP"
and "post-HIP", with characteristics as
shown in Fig. 16-4.
Figure 16-3. Key components of the HIP process. A
porous body (e.g., of powder particles) is subjected to
an isostatic pressure /?, acting on an impervious gas 16.3.1 Direct HIP Methods
barrier at its boundary at a temperature 7" for a time t.
As a consequence the "effective pressure in particle
"Encapsulated HIP", "glass encapsulat-
contacts" increases with/?, which at typical pressures ed HIP", "glass HIP", "capsule HIP", or
greatly enhances densification and form-stability of "container HIP" are other terms used for
the processed part. direct HIP processing of ceramics. The en-
 16.3 HIP Technologies for Ceramics 151

[ Direct HIP ] f Post-HIP

Establishing continuous Powder brought to a state with

Gas Impermeable Barrier No Surface Connected

Porosity Figure 16-4. Main princi-
around porous body ples for HIP of shaped
before pressurization green powder bodies to
before sintering starts
high density parts.

capsulation, impervious to the gas used for ties sensitive to surface defects, like the
the HIP process, must be evacuated and strength of ceramics - and machining be-
sealed before the gas is pressurized. ing generally very costly - the desire to
Metals, ceramics or, most commonly, make net shape products is very pro-
glasses may be used for such an encapsula- nounced for ceramic materials.
tion. The HIP method to control the shape of
the product by the HIP container was used
for large oc-alumina canisters (H. T. Lark-
16.3.1.1 Direct HIP in a Pre-Shaped
er, 1980) in a technical feasibility study of
Container
making long term resistant containment
This is the most common method for for spent nuclear fuel. About 1600 kg of
near net shape manufacture of powder 99.8% pure aluminum oxide powder was
metallurgy parts. Sheet metal or glass con- carefully packed in a low carbon steel con-
tainers are used. They have a similar shape tainer with 3 mm wall thickness, 3000 mm
(facing the powder) as the product to be length and an outer/inner diameter space
made, but are enlarged to compensate for for the alumina powder of 600/350 mm
the ratio between fill density and final den- (Fig. 16-5). At 1350°C and 150 MPa
sity of the powder. Much progress is being for 4 h, a fully dense alumina canister hav-
made in the development of computer-aid- ing 500 mm as the largest diameter with
ed methods to adjust the container shape a tolerance of 1-2 mm was produced
to anomalies in container deformation and (Fig. 16-6). Such a tolerance was accept-
powder densification during HIP (Seliver- able in this case, but could only be reached
stovetal., 1994); see also Sec. 16.5.1.1. For by the use of an efficient system to uni-
metallic materials with metal containers, formly pack the alumina powder into the
removal of the encapsulation after HIP annular steel container, and because of the
can often be combined with finish machin- relatively simple geometry.
ing, and near net shape (NNS) production For ceramics requiring higher process-
is good enough. A combination of finish ing temperatures, it is difficult to find suit-
machining with encapsulation removal is, able container materials, to control and
however, seldom possible for ceramic ma- limit reactions with the processed product,
terials because of the difference in material and to meet close shape tolerances for such
properties between container and product. containers at a reasonable cost. Further-
As any machining can deteriorate proper- more, the desired consistency in powder
152 16 Hot Isostatic Pressing

Figure 16-5. Principle for

HIP with the shape con-
trolled by the container.
Left: powder filling; center:
sealing of container; right:
HIP.

particle distribution is much more difficult

to attain by filling ceramic powders of fine
particle size into containers than by some
methods utilized for green body forming
with temporary binder systems, e.g., injec-
tion molding or pressure slip casting.
These findings have all contributed to
make the interest in this HIP method very
limited for products of high performance
ceramics with net or near net shape re-
quirements. Tantalum cans have, however,
been used for making material samples of
e.g., silicon nitride, with good results
(Richerson and Wimmer, 1983) and boron
(Hoenig et al., 1991).

16.3.1.2 Direct HIP with Shape

Determined by Green Body
HIP is, as discussed above, a very pow-
erful method to densify and sinter materi-
Figure 16-6. Large high-purity a-alumina canister als. However, commercially even more
produced by the principle of Fig. 16-5 and using a important is the inherent ability of the
shaped mild-steel container. The 500 mm diameter,
canister with a 100 mm wall thickness is being assem- HIP process to produce components with
bled to join a similarly made lid using HIP bonding complex and well-defined shape. The char-
(see Sec. 16.7.5.2) (courtesy of AC Cerama). acteristics of the green powder body
 16.3 HIP Technologies for Ceramics 153

(Fig. 16-7) will have a decisive influence on

SHAPE CONTROL
the extent this inherent ability can be uti-
Good finished product shape lized to consistently make high precision
can be achieved with products.
any green body forming method giving
consistency in In an ideal HIP process, a pure isostatic
pressure should only be exerted on the
• particle distribution and orientation body boundary, and the stress state in the
and body should only emanate from that and
• shape of body boundary
from sintering mechanisms. Weightless
Examples of conditions and an impervious pressure gas
green powder body shaping methods
(more or less fulfilling the above) seal at the body boundary, creating no
are shear stresses, would satisfy these require-
injection moulding, ments.
slip casting, The main difference between the pro-
pressure slip casting,
CIP + green machining, cessing routes to the left and to the right in
die pressing Fig. 16-8 concerns how early the process-
ing of the body can take advantage of the
Figure 16-7. Requirements on green body forming
processes enabling high-precision parts to be pro-
high isostatic pressure. Only the direct HIP
duced repeatedly. alternatives can utilize the effect of pres-
sure throughout the shrinkage and densifi-
cation phase, while the post-HIP route
cannot do so until the body has been
brought to a density of 91-97% T.D. (that

Product shape
determined by
Green Powder Body Shape

/ \
Establishing a Sintering

Shape Conforming with gas communication;

Envelope inside to outside
the powder body
for the green body
E.g. vacuum, pressure-less,
E.g. giass ampoule over-pressure sintering (GPS)....
or particle type encapsulation until

Sealing Disconnection Figure 16-8. HIP alterna-

of envelope of inner pores from surface tives for making shaped
parts based on the green
body form. To the left,
\ f
"direct HIP" and to the
H 1p iI ip right, "post-HIP" alterna-
tives.
154 16 Hot Isostatic Pressing

is, when all internal pores are isolated from , (MPa)

\
the body surface). Obviously direct HIP \
\
\
also gives added sintering control up to \
- 1000
such a density level, but less obvious are •V.
f Pgas = 200 MPa
"*" [ (Independent of R)
the advantages offered for controlling the
shape. -100
In direct HIP, the high gas pressure
when transmitted to the porous body via
the encapsulation, makes the body ex- -10
tremely rigid and insensitive to external in- Uz 1 Jm-2(Al203,SiC)
/ R = 1[im
fluence, which is particularly important in /
-1 / / / /
the initial stages of densification when /
bonds between particles are still very weak. /
/
An explanation of this exceptional dimen- - 0.1 £ hz 0.3 Jm" 2 (Glass)
sional consistency was presented by H. T. \R = 10|im
Larker (1985); see production examples in 0.6 0.7 0.8 0.9 1.0
i i i
Sec. 16.6. The gas pressure which acts iso- l

statically via the glass encapsulation on the RELATIVE DENSITY D

boundary of the porous powder body (see Figure 16-9. Effective pressure in particle contacts
Fig. 16-3) has a very large influence on the caused by HIP gas pressure (broken curve) and sin-
effective pressure in particle contacts tering mechanisms (solid curve). An example of a
tensile stress component in a horizontal beam with
(Fig. 16-11). As a result, gravity-induced length-to-height ratio of 10 (dotted curve) is also
tensile stress components, which could shown.
cause shape deformation, will become vir-
tually negligible (dotted line in Fig. 16-9).
In typical glass powder encapsulated and densification process has to occur be-
HIP the powder body is kept in a very firm fore the restraint by gas pressure comes
hold by the gas pressure acting on its into action. It should be noted that there is
boundary throughout its entire shrinkage no longer any such effect in encapsulated
and sintering process. The rigidity of the HIP when a part is fully densified and a
porous powder body may be visualized by uniform pressure throughout the part is
thinking of the common vacuum-packed already established.
packet of ground coffee. The ceramic pow- This inherent characteristic of the en-
der is, however, much harder and the pres- capsulated HIP process may actually re-
sure difference over its boundary more sult in a better shape precision for dense
than a thousand times higher, (similar to ceramics of complex shape from a uniform
the isostatic pressure at the bottom of the green powder body than any other sinter-
Mariana Trench!). ing process. It is important that the green
In contrast, when sintering is carried out powder body is encapsulated before any
without encapsulation, and consequently significant shrinkage of the body occurs. A
also in sinter-HIP or gas pressure sintering further requirement is, however, that the
(a process which may be seen as sintering encapsulation is "soft" and virtually does
and low pressure HIP in one sequence), not distort the shape of the powder body.
such an effect cannot be taken advantage In the first published realization of this
of, because the major part of the shrinkage principle for silicon nitride, a conformable
 16.3 HIP Technologies for Ceramics 155

high silica glass container was used

(Fig. 16-8, left), reported by H. T. Larker
et al. (1977) [the first patents were, howev-
er, applied for several years earlier; see e.g.,
Adlerborn and Larker (1974)]. That meth-
od is still well-suited for material samples.
With Vycor glass, which due to its soften-
ing behavior is considerably easier to work
with than quartz glass, the experimental
conditions can be particularly well-de-
fined. Hot evacuation can, for example, be
made at a relatively high temperature Figure 16-10. Steps of the encapsulated HIP process
using a conformable high silica glass container (am-
(about 1000 °C) and if desired, the contain- poule). Rubber bag for cold isostatic pressing (CIP)
er can be back-filled with a defined of silicon nitride powder with 1 wt.% yttria additive
amount of, for example, a gas before being to a cylinder, which is green-machined (left). Evacu-
sealed (Fig. 16-10). However, this type of ated and sealed container ready for HIP (center left).
glass encapsulation is not suited for sam- Container halfway through the HIP cycle after heat-
ing the glass to softening (at about 1300°C) and pres-
ples with weak, protruding parts, as the surizing the glass to conform to the shape of the green
viscous glass container may break these off body (center right). Fully dense part after HIP pro-
the green body when conforming to its cessing at 1750°C and 200 MPa for 1 h and finishing
shape. by sand blasting (right).
Shapes such as turbine blades have been
processed using a boron nitride powder
bed to fill the space between the green
powder body and the glass container
(Boehmer and Heinrich, 1980). The pow-
der bed should preferably have about the
same relative density as the green powder
body in order to reduce shape distortion.
A further function of the boron nitride is
also to provide a chemical separation from
the glass.
Containers of glasses with a lower soft-
ening point, e.g. Pyrex, have been used,
surrounding the sample packed in a boron
nitride bed and wrapped with molybde-
num foil, for example in silicon nitride/sil-
icon nitride joining experiments (Shimada
etal., 1984).
Encapsulation techniques using glass Figure 16-11. Axial turbine wheel of silicon nitride
particles (see Fig. 16-8, left), have achieved produced by injection molding the green body (left),
glass particle encapsulate, HIP at 1750°C and
the widest industrial see (H. T. Larker,
200 MPa for 1 h and finishing with sand blasting
1979 b). Such encapsulation was used for (right). The trailing edges of the 37 blades are about
HIPing of the complex axial turbine wheel 0.3 mm thick (courtesy of United Turbine and AC
in Fig. 16-11. A flow chart of the process is Cerama).
156 16 Hot Isostatic Pressing

shown in Fig. 16-12. Different realizations Powder Body of Consistent

of particle-type encapsulation have been Shape and Material Distribution

published in the patent literature. There

are single- or multiple-layer techniques Removal of forming
vehicles, e.g. binders
(Fig. 16-13, top), reported by Isaksson and
Larker (1971), Adlerborn and Larker
(1976), and Heinrich and Boehmer (1989). Application of glass
particle envelope
Pure and highly doped silicon nitride pow-
der layers have found use as inner/outer
Hot evacuation through
particle layers, respectively, on reaction- permeable glass envelope
bonded silicon nitride bodies (Heinrich
and Boehmer, 1989), and high- and low-
Sealing of glass envelope
melting glasses on green silicon nitride
powder bodies have fulfilled similar func-
tions (Adlerborn and Larker, 1976).
There are also technique using an open Pressurization and HIP
vessel to contain the glass (Fig. 16-13, bot- at selected parameters
tom), as reported by Adlerborn et al.
(1978). In this case glasses of lower viscos- Depressurization and cooling
ity can be used. Separation layers, which
usually contain boron nitride powder to
some extent, may often be a part of the Glass removal
encapsulation system in order to control
possible reaction and penetration phenom- Figure 16-12. Flow chart showing processing se-
ena. The glass encapsulation may be me- quence of the glass particle encapsulation and HIP
chanically removed after HIP, e.g., by process.
sand blasting, a method reported also for
the ceramic/ceramic multiple layers (Hein-
rich and Boehmer, 1989). The sand blast-
ing must, however, be gentle when in con-
tact with the ceramic, otherwise the surface
strength may be drastically reduced.
Chemical methods can be used in combi-
nation with specially developed glasses.
Glass removal by thermal chocks, as some-
times mentioned in the literature, is not
recommended as a crack in the glass very
often continues into the ceramic.
Encapsulated HIP allows a great free-
dom of composition choice, which has
been taken advantage of in particular in W&txM&M^^^
materials developed for good properties at Figure 16-13. Two principles for particle-type encap-
high temperatures (Tanaka e t a l , 1992) sulation: single/multiple layer technique (top) and
and in composites. It has found extensive technique using an open vessel to contain the glass
use for prototype manufacture of silicon (bottom).
 16.3 HIP Technologies for Ceramics 157

nitride components for heat engines like 16.3.2 Post-HIP of sintered green bodies
turbine rotors or stators. Important silicon
nitride materials in the current U.S. heat The alternative to the right in Fig. 16-8,
engine programs such as NT-154, NT-164 namely post-HIP, also called "cladless
and NCX-5102 (Norton/TRW), GN10 HIP", "sinter-HIP" or "sinter plus HIP",
(Garrett Ceramic Components) and PY6 was used for the first widespread commer-
(GTE Labs) are all processed by glass en- cial application of HIP in the early 1970s.
capsulated HIP. Large cemented carbide products such as
Exceptional properties were reported by rolls for rolling mills were post-densified.
Pujari and Tracey (1993), for example, for As the parts after liquid phase sintering
NCX-5102. By selecting starting powder, had already reached > 9 9 % T.D., the re-
carefully optimizing green body processing maining porosity was well separated from
and sintering, remarkable results were the surface. The high pressure gas (argon)
achieved. The material designated has a could therefore be allowed to act directly
Si 3 N 4 -4% Y 2 O 3 composition and was on the surface of the part, and good effect
sintered using glass encapsulation HIP. On on closure of internal pores was found.
320 tensile test rods with gauge diameter of The frequency of appearance of a certain
6 mm a mean tensile strength value of size of pores was reduced by two orders of
997 MPa at room temperature was ob- magnitude after HIP processing.
tained. A set out of these (n = 170) repre- Cladless HIP is a widely practiced way
senting strength of test rods with intrinsic to HIP ceramics. It works particularly well
defects did not fit the common two-param- with oxide ceramics and is successfully
eter Weibull model. However, the strength used for production of tool bits of oxide
data were well represented by a three-pa- ceramics, e.g., on alumina base (i.e. an
rameter Weibull distribution with a oxide alloy with alumina as base materi-
threshold stress of 665 MPa (Fig. 16-14). al, not an oxide coating on alumina). In

99.94

63.21-

0.03-
602 735 898 1097 1339
Strength (MPa)
Figure 16-14. Weibull plot of 170 intrinsic tensile strength data on 6 mm diameter test bars of Norton NCX-5102
silicon nitride. The material was made by colloidal processing and glass encapsulated HIP. The line marked
"TARGET" represents a mean tensile strength of 900 MPa with a Weibull modulus of 20 in two-parameter
representation (from Pujari and Tracey, 1993, Fig. 9).
158 16 Hot Isostatic Pressing

Mn-Zn and Ni-Zn ferrites, freedom from least 91 % T.D. Improved results were ob-
pores results in high permeability, high sat- tained if the pressurization for HIP started
uration induction and improved wear re- just after the stage of pore closure, accord-
sistance. The fine and uniform grain size ing to Ritzhaupt-Kleissl etal. (1992).
that can be obtained gives better high fre- For magnesia-doped silicon nitride, the
quency characteristics and low ferrite densification is often more effective. Gen-
noise. In piezoelectric ceramics, which are erally densification is positively affected by
used in oscillators, for example, the fine long processing times, 6 h or more. If there
and uniform grain size facilitates process- is a sufficiently high volume fraction of
ing to thin (50 jim) discs. The absence of glassy boundary phase present, the densifi-
pores is important in order to improve cation is often improved for both alloying
strength and obtain a surface without de- systems. Sometimes pools of boundary
fects. phase may be squeezed into internal voids,
A necessary prerequisite operation prior which is hardly desirable in most cases.
to the HIP stage is to sinter the ceramic Lowering the pressure may reduce this ef-
powder components to such a density, typ- fect, as well as increasing the temperature
ically 91-97% T.D., so that the pores at pressurization, as reported by Kito et al.
cease to be interconnected to the surface of (1991) (for the system tungsten carbide/
the part. HIP is then carried out with gas cobalt, however, a reduction of tempera-
pressure acting directly on the surface of ture to just below the liquidus gave good
the part. The gas composition is important pore closure without squeezing cobalt into
for many materials, which will be dis- the large pores).
cussed in Sec. 16.4. Also in silicon carbide materials, the
The necessary pre-sintering to high den- grains formed during the pre-sintering
sity limits the composition to those materi- stage often resist further densification.
als that can be sintered to the required den- However, if grain growth in the pre-sinter-
sity without unacceptable deterioration of ing stage can be restricted and a material
microstructure, phase composition, shape, with fine grain size is used, good densifica-
etc. For many ceramic materials with high tion during post-HIP can be obtained
strength at high temperatures other limita- (Hunold, 1985).
tions also occur. In pre-sintering of silicon
nitride, for example, elongated (3-Si3N4
grains are formed. These new grains often 16.3.3 Quasi-Isostatic HIP Methods
form a very rigid interlocking structure. There are several reports about methods
With the yttria-doped systems, several to heat a green powder compact surround-
investigators (Ziegler and Woetting, 1985; ed by a ceramic powder and/or glass mix-
S. S. Kim and Baik, 1992) have shown that ture which is used to transmit pressure.
typically only a few percent of density in- After heating outside or inside a forging
crease is obtained by subsequent post- die, pressure is applied at normal or slow
HIP, regardless of the pre-sintered density forging speeds. Near isostatic conditions
of the ceramic body. This may not be sur- are claimed and in laboratory such meth-
prising in consideration of the high ods have proven fast and effective (World
strength and low creep rates of pure, co- Report, 1989). A process called ROC
lumnar P-Si3N4 grains, which generally (rapid omnidirectional consolidation), us-
are very well developed after sintering to at ing heating outside the die and thus using
 16.4 Influence of Gaseous Species 159

a minimal amount of heat, has been used materials that tend to decompose or disso-
for the production of ceramic tool bits. In ciate at sintering temperature, an increased
the first stage, what is called a fluid die partial pressure of a gaseous decomposi-
containing a sample (a formed green body) tion product will reduce or eliminate the
is heated to the required temperature in an decomposition. Nitrogen gas is therefore
argon-purged furnace and allowed to re- used for cladless HIP of nitride ceramics,
main there for a "soaking time" to obtain usually as pure nitrogen gas. Better densifi-
a uniform temperature. The glass fluid die cation and reduced bloating upon pres-
is removed from the furnace and placed in sureless reheating to high temperature
a pot die in a press. Pressure is applied by was, however, observed with a gas mixture
a ram for typically 3-5 s at a pressure level of argon with 7vol.% of nitrogen, thus
of 830 MPa. The fluid die exhibits plastic preventing excessive dissolution of nitro-
flow and transmits pressure in a quasi-iso- gen in the glassy phases (S.S. Kim and
static manner to the sample, which densi- Baik, 1992).
fies, mainly by plastic deformation. Once The powder embedding method is quite
the pressure has been released, the die is common in addition to the two above-
removed from the press and cooled down mentioned methods (using a chemically ac-
(Pyzik and Pechenik, 1988). tive gas to full pressure or to a partial pres-
sure). It is widely used in the HIP of oxide
functional ceramics such as ferrites and
16.4 Influence of Gaseous Species lead titanates. A powder bed of composi-
tion similar to the processed body must be
The influence of gases in chemical con-
densely packed, but can then provide just
tact with the ceramic material plays an im-
the gas species desired. The rather stiff
portant role in all the different HIP pro-
powder bed creates no shape problems in
cessing alternatives. Many observed phe-
these applications, as material blocks for
nomena can be better understood if the
cutting into the final shape are produced.
thermodynamics of gases are more fre-
Engineering ceramic components with a
quently considered (Ishizaki, 1991), see
more complex shape, with causes partial
Sec. 16.5.1.2. The presence (or lack) of re-
engulfing of the powder bed (e.g., turbine
actant gas partial pressures as well as total
wheels) would, however, be distorted as
gas pressure are important variables.
the bed interferes with part shrinkage (H.
T. Larker, 1984).
16.4.1 Cladless HIP
In all cases where such gases are used
Often an inert gas, usually argon, is used which are not fully inert, durable furnace
for cladless HIP. The presence of a chemi- materials must be chosen. Even so there is
cally active pressure gas is, however, to be often a reduction of service life for furnace
preferred in several cases. Some oxide ma- components like thermocouples. An im-
terials tend to be slightly reduced in an portant point to be aware of in cladless
inert atmosphere, causing discoloration HIP with any type of gas is that small
and deterioration of properties. An oxygen amounts of impurities in the pressurized
partial pressure can be used for "white gas become chemically more active in pro-
HIP", e.g., of zirconia or alumina (Man- portion to the gas pressure, and may react
abe etal., 1991). Oxide superconductors with the furnace or with the material pro-
are HIPed under similar conditions. For cessed.
160 16 Hot Isostatic Pressing

16.4.2 Direct HIP stance, which may still adhere to powder

surfaces at temperatures typically used to
The impervious gas barrier is typically seal the encapsulation (Hermansson et al.,
sealed at <1000°C for metal encapsula- 1988). Humidity, which is always present
tion and usually at 1000-1300 °C for glass adhering to powder surfaces will also, if
encapsulation. It is essential to understand not removed, cause reduction of high tem-
that this isolation barrier is active in both perature properties for materials such as
directions! silicon nitride. Exaggerated grain growth
Firstly, it provides chemical isolation of occurred in the work on the alumina canis-
the material from the processing gas. The ters shown in Fig. 16-6, unless the ad-
chemical activity of trace substances (de- sorbed water content was reduced to
scribed in the previous section) in the pres- <200 ppm (H. T. Larker, 1980). An other-
surized gas are prevented from attacking wise uniform micro structure with a grain
and deteriorating the properties of the size of a few micrometers exhibited singu-
powder body. On the other hand, any gas- lar millimeter-long columnar oc-alumina
eous species formed in the powder body crystals grown during HIP at 1350°C at
cannot escape. That is a great advantage higher adsorbed water content, and the
for materials like nitrides, which tend to bend strength was lowered by almost one
decompose at the high temperatures neces- order of magnitude.
sary for densification. The gaseous decom-
position products are contained within the
16.4.3 Special Processing Options
encapsulation, which is supported by the
high-pressure gas. As soon as equilibrium Seino et al. (1989) pointed out an inter-
pressure is reached inside the capsule, de- esting way to control the oxygen chemistry
composition ceases. This is also beneficial during glass encapsulated HIP of high-Tc
for preventing the formation of interfacial ceramic B i - ( P b ) - S r - C a - C u - 0 super-
gas-filled voids during diffusion bonding, conductors. By using the PbO/PbO 2 sys-
e.g., for silicon nitride to metals that tem, they succeeded in producing highly
preferably form silicides. dense material with the desired oxygen bal-
For materials that require removal of ance, giving very good superconducting
gaseous reaction products before final properties without the additional treat-
densification, adequate pre-processing ment in oxidizing atmospheres that was
must be adopted. For example, the re- previously considered necessary. Richards
moval of carbon oxides from ceramics and Benfer (1991) used borosilicate-glass
containing oxides and carbonaceous mate- capsules and BaO2 as oxygen donor in
rials (as required by many silicon carbide HIP of high-Tc YBa 2 Cu 3 0 7 _ x ceramics.
powders) must be carried out before the They achieved a high superconducting
encapsulation is sealed. transition temperature in the as-HIPed
Adequate selection and handling of the condition also in this material.
powder and the powder body is therefore Koizumi (1988) reported on high pres-
needed, in order to avoid deleterious sub- sure self-combustion sintering (HPCS) of
stances becoming trapped in the material. TiB 2 . A powder mixture of Ti metal and
Fluorine (from HF acid treatments of amorphous boron was formed into a rod
powder in order to remove iron impurities) by cold isostatic pressing. It was sealed to-
is one example of a very tenacious sub- gether with a tungsten heating wire under
 16.5 Theoretical Tools and Quality Assurance 161

vacuum in a glass capsule (of the type de- predicting, e.g., the final shape and density
scribed in Shimada et al. (1984)). The glass of the HIPed products, than by empirical
was heated to softening, then 100 MPa ar- results only.
gon gas pressure was applied to the assem- A further theoretical approach of major
blage and the mixture electrically ignited importance is consideration and analysis
by the tungsten wire. The best densifica- of thermodynamic conditions related to
tion was obtained with an excess of Ti met- the high total pressure and/or partial pres-
al. sures of different gas species. These effect
Synthesis with as much as 50 % of the not only densification but also phase
volume liquid at processing temperatures transformation and chemical reactions.
in the range 1900-1950 °C (some 300 °C Modified Ellingham diagrams can be con-
above the eutectic temperature) using structed and be used as a very powerful
glass-encapsulated HIP resulted in good tool. They may be called HIP phase dia-
part integrity, as reported by R. Larker grams (Ishizaki, 1990).
(1992 a). A block formed by cold isostatic Regarding control of final shape, the
pressing of an equimolar mixture of silicon possibility of moving towards ideal condi-
nitride and silica powders without other tions, both regarding the temperature dis-
oxides was more than 90% converted to tribution in the body being densified and
silicon oxynitride, a material that dissoci- the properties of the gas-impermeable en-
ates at atmospheric pressure above capsulation, should not be overlooked (see
1700°C, by HIP at 200 MPa and 1900 °C the second paragraph of Sec. 16.3.1.2).
for 4 h or 1950°Cfor2h. Under these conditions both the highest
precision and the lowest cost may be the
result.
16.5 Theoretical Tools and Quality
Assurance 16.5.1.1 Modeling of HIP Mechanisms
In the micromechanical approach the
16.5.1 Developments in HIP Theory
rate equations for the behavior of the pow-
The rapid development of HIP technol- ders under densification are derived con-
ogy in recent years has been accompanied sidering physical aspects such as creep and
by major advances in the basic under- diffusion from the view of individual
standing of the HIP process, e.g., by using grains, particles and their surroundings.
"HIP maps" and/or "HIP phase dia- Helle et al. (1985) proposed densification
grams". Mastering of such techniques rate equations and constructed HIP mech-
would make it possible - in a systematic anism diagrams, popularly called "HIP
way - to preselect the process variables maps". The equations proposed are differ-
(pressures, temperature, time), and how ent for relative densities below and above
pressure and temperature are to be varied 0.9, respectively. Such mechanisms as
with time to give optimum results, both for grain boundary diffusion, lattice diffusion,
material properties and for product shape power-law creep, and in the latter case also
and precision. Micromechanical modeling boundary diffusion, are considered. Li
and macromechanical modeling are two et al. (1987) investigated the phenomena of
approaches to studying the complex pro- shape changes during HIP. Densification
cess of hot isostatic pressing and better during HIP, at constant temperature and
162 16 Hot Isostatic Pressing

pressure, can be modeled satisfactorily. tween dilatation and distortion during

But in practice the heat flow through the compaction (Svoboda, 1994). Good agree-
ceramic powder being heated up to the se- ment between computed and experimen-
lected HIP temperature is slow, and, under tally obtained shapes has been reported for
normal HIP conditions, temperature gra- axisymmetric parts. Computing time is still
dients are inevitable. If a density front in fairly long (Svoboda mentions 10 h for a
the powder body is formed by too rapid cylindrical-conical part).
heating of the part while under pressure,
large changes of shape may develop. The
16.5.1.2 HIP Phase Diagrams
analysis in this paper proposes that this
may happen when the densification rate is The high isostatic pressure, and the form
greater than a characteristic heat flow rate. in which it is applied in HIP, influences not
The behavior of a densification front was only densification and shape retention, but
modeled for a one-dimensional heat flow also phase transformations and chemical/
field. Limited experiments confirmed the physical reactions. Sintering conditions
trends. during HIP, such as high total pressure,
In the macromechanical approach the and/or partial pressures of gas species in
powder is considered as a continuum. The contact with the processed powder body,
behavior of the powder under densifica- create peculiar thermodynamic conditions.
tion is modeled by constitutive equations These have to be considered in order to
based on a modification of the theory of better understand and to be able to take
plasticity for porous materials. Porosity or full advantage of the potential of HIP pro-
relative density is usually used as an inter- cessing.
nal variable. Temperature calculations Ellingham diagrams, in their usual
taking into account variations of specific form, indicate the oxygen partial pressure
heat and thermal conductivity as function in which a metal and its oxide are in
of relative density and temperature may be equibrilium at any given temperature (see,
considered using a finite element heat e.g., Gaskel, 1973). They have to be re-
transfer code. drawn because of the extremely high total
In order to show the correct behavior pressure involved. Such modified dia-
the constitutive model has to be based on grams are referred to as "HIP phase dia-
proper experimental results. Pure dilata- grams" and Ishizaki (1990) describes a
tion measurements can be made using way to construct these. The change in
dilatometers suitable for operation inside a Gibbs' free energy of a reaction may be
HIP furnace, so-called "HIP dilatometers" plotted at any given pressure, as a function
(Traeff, 1990). The powder body to be of temperature, considering only the vol-
studied is usually enclosed in a cylindrical ume difference between the gaseous reac-
HIP container, whose geometry for such tants and the reaction products and as-
tests is crucial (Dietze, 1991), because the suming an ideal gas behavior for each gas.
influence of container deformation can Because of the importance (and too of-
easily give incorrect results. Collecting ten seemingly ignorance) of the influence
data for the other direct case, distortion, of gases in respects other than as the trans-
using torsion tests, is problematical for ce- mitter of the isostatic pressure to the work-
ramic materials, but free compression tests piece, the preceding section (Sec. 16.4) has
may be used for assessing the ratio be- been devoted to practical examples of this
 16.6 Some Applications and Their Characteristics 163

topic. One further illustration of the use-

fulness of HIP phase diagrams is ceramic/
ceramic composites with carbon fibers in a
silicon nitride matrix. Such diagrams
(Celis and Ishizaki, 1991) show that in en-
capsulated HIP silicon carbide would not
be formed at temperatures lower than
2200 K (but would be formed in normal
sintering or hot pressing).

16.5.2 Quality Assurance Figure 16-15. Glass encapsulated and HIPed silicon
nitride ball blanks to net shape (right); to the left
Controlling the product quality of ce- starting materials, and in the middle pre-shaped green
ramic parts produced by HIP must not only powder bodies (courtesy of SKF).
involve a before-delivery-check-control
but quality assurance built into all the pro-
cessing steps. Process optimization tech-
only HIP processing gives optimum prop-
niques with the goal to make the process
erties, is silicon nitride balls for bearings
more efficient, controlled and reliable
(Fig. 16-15) (Cundill, 1993). Silicon ni-
should be applied. Knowledge of the pro-
trides processed by glass encapsulated HIP
cess technology is the key to controlling
and with a reduced volume of intergranu-
and thus assuring product quality. When-
lar phase provided superior fatigue and
ever possible the processing parameters
rolling contact wear performance, com-
should be set in the area which in the liter-
pared to materials with high additive con-
ature is called the "processing map" or
tent (Lucek, 1990).
"response surface", where the influence of
parameter variations or uncontrolled vari- Another emerging industrial application
ations in the process influences the product area is small high precision parts, e.g., for
properties the least. textile machinery (Fig. 16-16). In some ap-
plications, the reason for competitiveness
The basic rule in quality control, i.e. to
for components made by HIP of silicon
check instruments and sensors against cal-
nitride base ceramics is mainly the very
ibrated masters at predetermined intervals,
high precision compared to parts of sin-
must of course be adhered to. Particularly
tered alumina, that can be reached without
important are the high temperature mea-
any post-machining. With injection mold-
surement sensors in the HIP equipment.
ed green bodies (which can reach very high
uniformity and precision) and with glass
powder encapsulation (which is flexible
16.6 Some Applications and enough not to restrain the shrinkage of the
Their Characteristics ceramic powder body), it appears that al-
most ideal conditions can be achieved (H.
Several application examples, mainly T. Larker et al, 1993). On a 59.7 mm long,
for post-HIP techniques, have been men- curved thread guide of silicon nitride (in
tioned above, for example cutting tools the upper part of Fig. 16-16), of which
and different electronic ceramics. A major over 100000 parts have now been deliv-
emerging industrial application, for which ered, the standard deviation of the length
164 16 Hot Isostatic Pressing

Figure 16-17. Stator vane segments for an aircraft

auxiliary power unit gas turbine. The parts are slip-
cast GN-10 silicon nitride densified by glass encapsu-
lated HIP (courtesy of Allied Signal Aerospace).

A comprehensive testing and quality as-

surance program is required for compo-
nents of the hot gas path in gas turbines for
use aboard aircrafts. Silicon nitride vanes
(Fig. 16-17) have successfully passed test-
ing in APUs (auxiliary power units) for
Figure 16-16. Series-produced high-precision parts aircraft (the gas turbine powered electric
of silicon nitride for advanced textile machinery man- generator set). One of their advantages
ufactured to net shape by injection molding, glass over their superalloy counterparts is their
encapsulated HIP and with vibratory tumbling as the erosion resistance.
finishing operation (courtesy of AC Cerama).

16.7 Diffusion Bonding

was found to be only 42 jum or 0.07 % in a
production batch of 768 parts. The first development of HIP technolo-
A standard deviation limited to only a gy in the 1950s originated from the need to
few tenths of one percent in green body clad uranium oxide fuel elements with zir-
density would in fact leave room for no calloy for use in nuclear reactors. The use
other shape distorting effect whatsoever of HIP for joining purposes has since then
during sintering, if such a low dimensional mainly concerned diffusion bonding of dif-
variation as 0.07% is to be attained. This ferent metal alloys, such as austenitic steel
is obvious from the elementary densifica- to ferritic steel.
tion equation, knowing that the standard The HIP process may be used for the
deviation in final density after HIP is typi- joining of ceramics to ceramics or for ce-
cally only some hundredths of one percent. ramics to metals (or, of course, also for
An explanation for this exceptional dimen- metals to metals, but that is outside the
sional consistency can be found in the con- scope of this volume); one or all of the
ditions during encapsulated HIP; see materials may either be in fully dense, in-
Figs. 16-3 and 16-9 and Sec. 16.3.1.2. termediate or green body states. The use of
 16.7 Diffusion Bonding 165

large HIP facilities could make diffusion portant factors are the reactivity between
bonding more accessible as a mass produc- the materials, the enrichment of impurities
tion process and increase the range of ge- and/or ceramic sintering additives at the
ometries that can be bonded, according to interface, and possible reactions with the
Nicholas (1991). In this section, ceramic/ surrounding atmosphere.
metal joining will be emphasized, but ex- The most accurate modeling of diffusion
amples will also be given of ceramic/ce- bonding published so far was made by Hill
ramic joining, with both fully dense and and Wallach (1989). Their model was
green body components. claimed to be applicable to any single-
For applications with the highest re- phase, similar-to-similar metal bond. They
quirements on joint durability, joining in proposed an elliptical void shape, instead
the solid state by diffusion bonding is the of the earlier adopted cylindrical geome-
prime choice. This is due to the possibili- try. This was supported by micrographs
ties of minimizing the formation of reac- indicating a relatively low aspect ratio
tion layers having inferior mechanical and (height/length) of the voids. The model
chemical properties. A second area of gen- considers the contributions from seven
eral importance for durable joints con- mechanisms (also operating in pressure
cerns the necessity to retain the desired mi- sintering) to void elimination: plastic yield-
crostructures of the joined materials after ing deforming an original contacting as-
heat treatments experienced during joining perity, power-law creep, volume/grain-
(or otherwise possibly to restore them by boundary diffusion from an interfacial
subsequent heat treatments). For rigid source to a neck, evaporation from a sur-
joints between dissimilar materials having face source to condensation at a neck, or
large differences in stiffness and thermal volume/surface diffusion from a surface
expansion behavior, a third area of major source to a neck.
importance concerns the reduction of ex- Initially, plastic deformation rapidly in-
treme residual stresses, caused by thermal- creases the contact area until the applied
ly induced strains either on cooling from load can be supported, i.e., the local stress
the joining temperature or during thermal falls below the yield strength at the joining
cycling in use. temperature. The contributions from the
remaining six time-dependent mechanisms
are then considered to add together, giving
16.7.1 Modeling of Diffusion Bonding
the overall amount of bonding. For single-
In diffusion bonding, the most crucial phase, similar-to-similar metal bonds such
aspect is the degree of contact between the as for copper, oc-iron or y-iron, it was
surfaces to be joined. Since no surface is found that the main mechanisms after fin-
perfectly flat on the atomic level, the sur- ished yielding were either power-law creep,
faces must adapt to each other by yielding grain-boundary diffusion or surface diffu-
followed by power-law creep and diffusion sion. Their model considers the influence
processes. The degree of contact is gov- from the amount of grain boundaries
erned by three factors: the surface rough- (which depends on the grain size) and also
ness of the mating surfaces, the yield/creep the variation in chemical potential due to
behavior of the materials at the joining the angle between applied (uniaxial) pres-
temperature, and the pressure level applied sure and a particular grain boundary. It
to bring the surfaces together. Other im- does not, however, consider the influences
166 16 Hot Isostatic Pressing

from surface oxides or from the formation tion, while hot pressing at 27 MPa de-
of intermetallic phases, obviously relevant formed the molybdenum and did not result
in joints between metals and nonoxide ce- in bonding.
ramics. Observe that a majority (the first four
out of seven) of the mechanisms in the Hill
and Wallach model, discussed in the previ-
16.7.2 Parameters Influencing
ous section, are directly or indirectly de-
Diffusion Bonding
pendent on the level of the applied pres-
Among the three main process parame- sure.
ters in diffusion bonding (temperature,
time and pressure), temperature is usually
16.7.3 Diffusion Bonding by HIP
considered to be the most important, since
it has a great influence both on yield In the case of ceramic/metal joining, the
strength and on kinetics for creep and dif- high isostatic pressure (usually 100-
fusion. As will be discussed later, most 200 MPa) acting on the encapsulation of
work in the field of ceramic/metal joining the components to be joined (which is
has been concerned with methods to over- evacuated before sealing and application
come the obvious restriction from cracking of pressure), forces the metal at the joint
mainly caused by large mismatch in the interface to plasticize and accommodate to
coefficient of thermal expansion (CTE). In the surface of the ceramic. At the joining
a review on diffusion bonding of ceramics temperature, the yield strength of the
by Akselsen (1992), it was stated that the metal alloy is usually low compared to the
optimum temperature for joining occurs at applied HIP pressure, since, even for most
a point where strength reduction due to superalloys, the major strengthening pre-
residual stresses (present after cooling) cipitates are dissolved at these tempera-
starts to balance the strength enhancement tures. The major part of the void elimina-
due to void elimination. The bonding time tion can therefore be performed by yield-
mainly influences the reaction layer thick- ing, mainly followed by power-law creep.
ness, usually following a parabolic growth The ability of two superalloys, Hastelloy
with bonding time. X and Incoloy 909, to adapt to the very
The influence of the level and direction rough surface of a highly porous (10%)
of applied pressure are often neglected. SiC/SiC continuous fiber composite dur-
One plausible reason for this is that usually ing diffusion bonding by HIP at relatively
diffusion bonding is carried out under uni- low temperatures (900 °C or 1000 °C,
axial pressure inside a vacuum furnace, 200 MPa, 1 h) was demonstrated by R.
and therefore pressure is limited to avoid Larker et al. (1992b). However, an initial
macroscopic distortion at joining tempera- run at 800 °C proved to be too low for the
tures. This quite severe limitation does not, superalloys (especially Hastelloy X) to
however, apply for diffusion bonding by plasticize and accommodate to the rough
HIP, thus allowing pressures that exceed ceramic surface. The composite structure,
the yield/creep strength of the materials at made of stacked two-dimensional weaves
joining temperatures. According to Sug- of continuous SiC fibers with a silicon car-
anuma (1990), Si 3 N 4 /Mo joints (10 x bide matrix deposited by chemical vapor
10 mm area) could be bonded by HIP at infiltration (CVI), survived the diffusion
100 MPa without macroscopic deforma- bonding procedure without damage, in
 16.7 Diffusion Bonding 167

spite of the customary initial pressure in- peratures, resulting both in lower residual
crease in the early stages of the HIP cycle stresses when cooled to ambient tempera-
before the temperature was raised high ture, and in a very thin interface, if chemi-
enough to reduce the yield strength of the cal stability between the ceramic and the
superalloy sufficiently. The rough surface superalloy can be obtained. Moreover, the
of the composite, combined with proper de- possibility of retaining the optimum mi-
sign, might be useful for mechanical inter- crostructure in the metal alloy is increased
locking by plastic deformation of the super- with lower joining temperatures.
alloy during diffusion bonding by HIP.
For ceramic/ceramic joining of solid 16.7.4 Diffusion Bonding of Ceramics
materials, creep in grain-boundary glassy to Metals by HIP
phases may be responsible for the major
16.7.4.1 Importance of Ceramic/Metal
part of the void elimination, followed by
Joining
diffusion processes. For joining with green
body components, obviously densification The evolution during recent decades of
and joining processes take place simulta- structural ceramics such as silicon nitride
neously. (Si3N4) and silicon carbide (SiC) for use at
Components to be joined are not limited high temperatures has spurred develop-
to planar geometries with small areas pro- ment of technologies for joining these ma-
vided that an encapsulation can be ap- terials to metallic superalloys. Compo-
plied. Suitable capsule materials must have nents intended to sustain a significant
a low yield strength at HIP temperature, a stress level at elevated temperatures
low reactivity with the joined materials (> 800 °C) under simultaneous exposure to
and permit easy and reliable sealing. For corrosive environments such as combus-
joining around 1000°C, copper or mild tion gases in heat engines can be manufac-
steel are often used, while for higher tem- tured from these two classes of materials.
peratures glass or tantalum are possible Structural ceramics should, owing to
alternatives. their volume-dependent strength, manu-
In contrast to other joining methods, facturing limitations and associated costs,
such as active metal brazing or conven- be applied only in parts where their prop-
tional diffusion bonding using uniaxial erties can be utilized efficiently, mainly in
pressure under vacuum, the interfacial components facing very high temperatures
joining processes during HIP are taking and/or aggressive environments at certain
place in a closed system inside the capsule; levels of mechanical or thermomechanical
(see also Sec. 16.4.2). The pressurized en- stress. Superalloys are preferred in all parts
capsulation prevents both unwanted reac- where their properties are sufficient. The
tions with the furnace atmosphere and utilization of Si 3 N 4 and SiC in hot applica-
possible degradation by the formation of tions such as heat engines is thus presently
voids due to released gases, e.g., nitrogen restricted by the lack of efficient methods
during joining of silicon nitride to metals for joining them to superalloys.
such as Ni that preferably form silicides, Joints intended for use at elevated tem-
see Brito et al. (1989), Heikinheimo et al. peratures can be made by mechanical at-
(1992). tachment (shrink fit), diffusion bonding or
These advantages for diffusion bonding active metal brazing. Since an efficient use
by HIP permit reduction of bonding tem- of these materials requires the joint to be
168 16 Hot Isostatic Pressing

positioned as far into the hot zone as pos- joining temperature or later during ther-
sible, the joined area must be able to sus- mal cycling in use. The CTE mismatch is
tain thermal cycling to high temperatures considerably larger for joints between
(500-700 °C) during use, and, for the two Si 3 N 4 or SiC (where their strong, mainly
latter methods, to sustain even higher tem- covalent bonding results in lower CTE lev-
peratures during the formation of the els than for ZrO 2 or A12O3), and superal-
joint. A design using an efficient materials loys (where the creep resistant f.c.c. struc-
combination is strongly related to the ob- ture results in an even higher CTE than for
tained durability of the joint. ferritic steels). When rigidly joined, ther-
mally induced strains result in extreme re-
16.7.4.2 Difficulties Associated sidual stresses that frequently cause frac-
with Ceramic/Metal Joining ture in the joint or extending some hun-
dred micrometers into the ceramic; (see
The fabrication of a joint between these
Yamada et al., 1987 and R. Larker et al.,
two classes of materials is, however, com-
1989).
plicated by the extensive structural and
The use of a mechanical attachment
chemical differences between ceramics and
(such as a shrink fit consisting of a sleeve
metals. Methods for ceramic/metal joints
of the low-expansion superalloy Incoloy
in structures used at elevated temperatures
909 around a Si 3 N 4 turbine wheel hub) are
must deal with two major restrictions,
currently limited in temperature to ap-
namely high residual stresses and excessive
proximately 500 °C, due to either loosen-
reaction zones that occur in the bonded
ing of the firm grip at higher temperatures
region during joining and use; see reviews
or exceeding the yield strength at the low-
by Elssner and Petzow (1990), Pejryd
est temperature.
(1992), Nicholas (1991), or Suganuma
A proper design based on a shrink fit
(1990). A third restriction, which accord-
combined with either diffusion bonding or
ing to Pejryd (1992) and R. Larker (1992 b)
active metal brazing might allow higher
has received very limited attention, is the
temperatures for the joint in use.
necessity to retain the optimum mi-
crostructure of the metal alloy after the
heat treatments experienced during join- 16.7.4.4 Concepts Suggested for the
ing. These three restrictions are, more or Reduction of Residual Stresses
less, influenced by both the selection of the
joining method itself and of its associated Considerable efforts have been made to
parameters (such as temperature, time, reduce the residual stresses by applying
pressure and chemical environment). metallic interlayers, especially in diffusion
bonded or brazed joints. The suggested in-
terlayers are of three kinds: ductile metals,
16.7.4.3 Origin of Residual Stresses
refractory metals (with low CTE) and low-
in Ceramic/Metal Joints
expansion alloys. In the literature, these
The thermal expansion mismatch for are often combined to reduce the stresses
materials with only slightly different CTE in the ceramic down to a safe level.
levels, such as between ferritic and The often proposed solutions of chang-
austenitic steels or between ZrO 2 and fer- ing the composition from the metal side
ritic steel can already cause high stresses in through ductile metal interlayers such as
rigid joints, either during cooling from the nickel or copper (see Yamada et al., 1987;
 16.7 Diffusion Bonding 169

Brito et al., 1989; Frisch et al., 1991), and/ rials (FGMs) (Okamura, 1991) for inter-
or refractory metals such as tungsten, mediate joint pieces, which have a grada-
molybdenum, tantalum, niobium or hafni- tion in composition and consequently also
um (see Suganuma etal., 1986; Yamada in important physical properties such as
et al., 1987; Brito et al., 1989; Frisch et al., CTE and Young's modulus, applied be-
1990 are, unfortunately, difficult to protect tween the materials to be joined. Due to
against fatigue and oxidation, respectively the reactivity between Si 3 N 4 and the rele-
(Pejryd, 1992). The low-expansion charac- vant metals at joining temperatures, a
teristics of Invar and Kovar are restricted third phase is needed to avoid the forma-
to relatively low temperatures (below tion of reaction layers with inferior proper-
200 °C and 300 °C, respectively). In a study ties. Titanium nitride (TiN) has shown to
on Incoloy 909, R. Larker etal. (1992a) possess suitable physical and chemical
found that the benefits usually claimed for properties for the third phase against both
the use of this superalloy in joints, a rela- Si 3 N 4 and pure Ni metal or Ni-based al-
tively low CTE up to 400 °C, is reduced by loys (R. Larker and Beckman, 1994).
a considerably higher expansion above
400 °C up to possible joining temperatures. 16.7.4.5 Formation of Reaction Layers
For interlayers intended to reduce During Joining
stresses in the ceramic without plastic de-
For the two joining methods involving
formation, such as refractory metals (W,
high temperatures during joining, namely
Mo), the interlayer must generally be
diffusion bonding and active metal braz-
rather thick, even when combined with a
ing, chemical reactions can be anticipated
ductile layer (see Suganuma etal., 1986;
both between the materials to be joined
Yamada etal., 1987). In the latter work,
and, when applicable, also with brazes and
calculations were made by the finite ele-
interlayers. The formation of reaction lay-
ment method (FEM) for several insert
ers are, in general, governed by thermody-
metal combinations of tungsten, nickel
namic and kinetic data for possible reac-
and copper, positioned between silicon ni-
tions, in conjunction with parameters as-
tride and Nimonic 80 A. They found that
sociated with the joining method (such as
residual stresses were reduced with increas-
temperature, time, pressure, and chemical
ing thickness of the tungsten interlayer up
environment). The reactivities with metals
to 5 mm, but due to the difficulties in pro-
at joining temperatures are, due to their
tecting tungsten against oxidation at high
proneness to form silicides, nitrides or car-
temperatures, they used three layers con-
bides, considerably higher for silicon ni-
sisting of 500 jum Ni + 800 jum W + 500 jim
tride and silicon carbide compared to ox-
Cu in their diffusion bonding experiments.
ide ceramics such as alumina. Diffusion
In a review, R. Larker's (1992 b) main bonding of these two nonoxide ceramics to
conclusion concerning residual stresses metals is therefore discussed in detail in the
was that for Si3N4/metal joints intended next section.
for high service temperatures (500-
700 °C), the CTE mismatch could not be
16.7.4.6 Reaction Layers Formed During
sufficiently reduced by modifying only the
Diffusion Bonding by HIP
metallic part of the joint. A fourth concept
for reducing the residual stresses could Diffusion bonding by HIP between sili-
then be to develop functional graded mate- con nitride and metals have been performed
170 16 Hot Isostatic Pressing

with joining parameters in the range ing between consecutive points. Elemental
900-1400°C and 50-200 MPa for 0.5-4 h concentrations were calculated using the
(see Suganuma et al., 1986; R. Larker thin foil approximation with corrections
etal., 1989; Frisch et al., 1991, 1992; for absorption and normalized to
R. Larker etal., 1992c). Between silicon 100 at. % (nitrogen, carbon and oxygen
carbide and metals, diffusion bonding by could not be detected). Since the electron
HIP has been performed at 1160°C and beam of about 6 nm diameter while pene-
103 MPa for 3 h (Moseley et al., 1991) or trating the thin foil (150-200 nm) broad-
at 900-1000°C and 200 MPa for l h ens up to about 25-30 nm diameter, there
(R. Larker etal., 1992b). Most work on was still only a slight overlap between con-
diffusion bonding of Si 3 N 4 or SiC to secutive points. The resulting spatial reso-
metals have, however, been conducted by lution was nearly two orders of magnitude
uniaxial pressing at lower pressure levels higher than for quantitative point analyses
(7-60 MPa) in vacuum. earlier conducted on bulk diffusion cou-
The joining temperatures selected in dif- ples in SEM (R. Larker etal., 1989). A
ferent papers vary over a rather large typical compositional profile for a wide
range. The resulting formation of reaction grain of this semi-continuous phase, deter-
layers and microstructural changes of the mined to be a G-phase silicide
metal alloy therefore differ considerably. (Ni 16 Nb 6 Si 7 ) with some substitution of
Thin reaction layers formed during dif- cobalt and iron for nickel and titanium for
fusion bonding by HIP between Incoloy niobium, is shown in Fig. 16-18.
909 and Si 3 N 4 or Si 3 N 4 /60 vol.% TiN Considerably thicker reaction layers
composites (later developed for graded were formed during diffusion bonding by
joints) were studied by R. Larker et al. HIP between Incoloy 909 or Hastelloy X
(1994). Joining was performed at 1200 K and a highly porous (10%) SiC/SiC con-
(927 °C) and 200 MPa for 4 h. The maxi- tinuous fiber composite in a study (also
mum total thickness of the reaction layers cited in Sec. 16.8.3) by R. Larker etal.
formed was < 1 Jim, being below the spa- (1992b). Joining was performed at 900-
tial resolution limit for scanning electron 1000 °C and 200 MPa for 1 h.
microscopy (SEM). Therefore analytical
electron microscopy (AEM) was adopted
to study the two reaction phases observed Relative concentration [at%]
in all specimens examined; the first phase 90 - ID Ni
was a continuous, about 100 nm thick 80 - • Co
bright layer consisting of numerous small 70 -
TiN crystals along the ceramic/metal inter- 60- I ,.,..„,. B Fe
face, while the second was a semi-continu-
50 - .'*•••"•! r i -ir--'l:'-!ff!*ffi
. ^ x x x x x x x X X Xm
11 Ti
40 -
ous dark layer of larger crystals formed 30 - f X X X X X X X X x S xV\ ^ • Nb
^^N<-X% —~
with varying thickness ( « 100-500 nm) 20 -_ • Si
into the superalloy. 10 -
Compositional profiles were determined o:
0 100 200 300 400 500 600 700 [nm]
along straight lines across these grains per-
Figure 16-18. Compositional profile perpendicular to
pendicular to the interface by quantitative the interface through a semi-continuous phase grain,
EDS (energy dispersive X-ray spectrosco- determined by quantitative EDS point (25 nm spac-
py) point microanalysis with 25 nm spac- ing) microanalysis (N, C and O omitted) in AEM.
 16.7 Diffusion Bonding 171
Relative concentration [at%]
100

0
0 10 20 30 40 50 60 70 80 90 100 110 120
Distance [jim]
[D Ni • Co m Fe M Ti • Nb • Si
Figure 16-19. Compositional profile (C omitted) from the interface SiC/SiC-Incoloy 909 HIPed at 1000 °C,
determined by quantitative EDS point (1 um spacing) microanalysis in SEM.

The reaction layer formed at 1000°C ing of both (Cr, Mo)xCy and (Ni, Fe)5Si2;
with Incoloy 909 consisted of a high num- and a 15 jim zone with linearly decreasing
ber of clearly distinguishable zones. EDS Si and simultaneously increasing Ni level,
quantitative point measurements were until the composition of Hastelloy X was
conducted in SEM using a line consisting reached.
of 121 points with 1 jim spacing and locat- The explanation for the disparity in lay-
ed perpendicular to the reaction zone. The er thickness is most likely the different
result is shown in Fig. 16-19. amounts of carbide formers present in
The total reaction layer (about 100 [im these alloys, as carbides can form an effec-
wide) can be divided into four zones: a tive barrier against the diffusion of mainly
78 \im wide zone probably consisting of Ni and Fe. Similar results were found by
alternating layers of (Ni, Fe)5Si2 and (Ni, Moseley et al. (1991). They found large
Fe)5Si2 + C; a 6 |im zone consisting either differences between Incoloy 909 and In-
of oc-Fe(Si) or ordered cubic a-Fe3Si, both conel 718 concerning the reactivity with
with graphite precipitates; a 4 ^im zone Si/SiC. The compositions of both Inconel
likely to almost entirely contain NbC and 718 and Hastelloy X contain approximate-
TiC; and a 14 jim zone with slowly de- ly five times the amount of carbide formers
creasing Si level from 21 to 15 at.%, fol- compared with Incoloy 909. Their investi-
lowed by a steep decrease down below gation did however involve reactions in the
1 at.%. On the other side of a crack paral- liquid state (with a HIP temperature of
lel to the joint the composition of Incoloy 1160°C!) and might therefore not be total-
909 was almost immediately reached. ly comparable.
The reaction layer formed with Hastel- The SiC/SiC composite was found to be
loy X was narrower (about 40 ^im) and considerably more prone to reactions with
consisted of only three zones: a 20 jim superalloys compared to the behavior of
wide zone probably consisting of (Ni, Si 3 N 4 with superalloys under similar con-
Fe)5Si2 -f C; a 8 |Lim zone probably consist- ditions. This is not surprising since, ac-
172 16 Hot Isostatic Pressing

cording to Pejryd (1992), Si 3 N 4 is thermo- It was shown that heat treatments per-
dynamically stable with Ni and Fe at the formed at 1060 °C or 1150°C caused disso-
relevant temperatures (for normal and lution of the grain-boundary Laves phase.
higher pressures), while SiC is not. At Due to the lack of grain-boundary pinning
higher temperatures (1100-1300 °C), how- carbides, excessive grain growth occurred
ever, also Si 3 N 4 forms thick reaction lay- from the initial grain size of 5-16 |im up to
ers with silicide formers (Frisch et al., an abnormal 200-440 (im, which is clearly
1992); this is further promoted by low ni- detrimental for tensile strength and ductil-
trogen partial pressures (Heikinheimo ity.
etal., 1992), as observed by Brito et al. The heat treatments at 800 °C in the
(1989) after diffusion bonding under uni- study caused an extensive precipitation of
axial pressure in vacuum. intragranular platelets of e" and 8 phases,
which may cause an undesired embrittling
effect. The formation of platelets at 800 °C
was more pronounced in the HIP samples
16.7.4.7 Effects on Microstructure from
than in vacuum furnace heat-treated sam-
Joining Heat Treatments
ples, probably caused by aged thermocou-
During active metal brazing or diffusion ples in the HIP furnace, resulting in a
bonding, the metal alloy to be joined to slightly higher (^10°C) temperature dur-
ceramics will be subject to a heat treat- ing HIP treatment.
ment. The desired structure of the metal The conclusions in this study were that
alloy must either be retained during join- Incoloy 909 must not be joined/densified
ing, or possibly be restored by other heat above the annealing temperature region
treatments afterwards. This influence from (930-1040 °C) specified by the manufactur-
joining treatment on the resulting mi- er, due to the exaggerated grain growth oc-
crostructure of the metal has, however, curring when grain-boundary Laves phase
been almost ignored in the literature on is dissolved. Furthermore, joining between
ceramic-metal joining. the recommended range for the higher
Due to the interest of using Incoloy 909 ageing temperature (720-775 °C) and ap-
in two areas, namely in joints to Si 3 N 4 or proximately 900 °C should also be avoided,
SiC, and as a matrix with W or SiC fibers since prolonged exposure can cause over-
in metal matrix composites (MMC), a aging by excessive formation of 8 phases.
study by R. Larker et al. (1992 a) has ad- In general it would be useful to combine
dressed this topic. Samples of "as-re- the joining procedure with heat treatment
ceived" Incoloy 909 were heat-treated at of the superalloy. This might be possible in
five selected temperatures: 800 °C, 900 °C the case of Incoloy 909, where diffusion
or 1000 °C for joining, and 1060 °C or bonding to Si 3 N 4 via graded joint pieces of
1150°C relevant for MMC purposes. The Si 3 N 4 /TiN and TiN/Ni composites could
heat treatments were conducted either in a be combined with annealing, and the fol-
HIP equipment at 200 MPa argon pressure lowing ageing treatments could optionally
(due to the possibility to join or densify at be combined with partial relaxation of re-
as low a temperature as possible) or in a sidual stresses by creep in the nickel phase
vacuum furnace. The heating rate was in at lower temperatures under remaining
both cases 25°C/min, the dwell time 4 h HIP pressure (> 100 MPa), as suggested
and the cooling rate 15°C/min. by R. Larker and Beckman (1994).
 16.7 Diffusion Bonding 173

Concerning Hastelloy X, the situation is

complicated by the fact that the solution
treatment recommended at 1175°C to
provide optimum mechanical properties
and corrosion resistance must be followed
by rapid cooling to 540 °C to prevent car-
bide precipitation. The solution treatment
temperature is clearly much too high for
diffusion bonding to SiC, while sufficient
topographical accommodation of Hastel-
loy X to the surface of the ceramic com-
posite during joining - even by HIP - Figure 16-20. Cold iso-pressed (CIPed), green-ma-
chined hub and injection molded "dummy wheel"
would require at least 900-950 °C, accord- blade-ring (right). Successfully HIPed and joined
ing to R. Larker et al. (1992b). wheel to the left.

16.7.5 Diffusion Bonding of Ceramics

pleted wheel - has been developed as well
to Ceramics by HIP
as slips for initial sealing of the joint. The
assembly is then glass encapsulated and
16.7.5.1 Joining of Ceramic Green Bodies
HIPed. Four-point flexure tests on thest
by HIP
bars cut perpendicularly to the joint and
Injection moulding allows products with stressed at the joint gave good results. The
very high accuracy of as-processed shape strength was 913 + 49 MPa (m = 23) at
to be made. However, a major drawback is room temperature.
the difficulty in removing the organic
binder system, necessary for forming, from
thick material sections. H. T. Larker et al. 16.7.5.2 Joining of Solid Ceramic Bodies
(1994) report a solution for a radial turbine by HIP
wheel in an automotive gas turbine, by us- In the late 1970s large area solid-solid
ing different green forming techniques for ceramic-ceramic diffusion bonding was
two parts which are then green joined and applied to large oc-alumina canisters for the
HIPed, using a modified glass encapsula- containment of spent nuclear fuel (H. T.
tion. Not the entire radial turbine wheel of Larker, 1979 a). The canister was made of
silicon nitride but only the blade ring, with 99.8% pure Al2e>3 HIPed to shape (see
a relatively thin maximum cross section, is Sec. 16.3.1.1) with a density exceeding
made by injection molding. The thick hub 99.5% T.D. and had an outside diameter
of the wheel is made by CIP (cold isostatic of 500 mm with 100 mm wall thickness. It
pressing). The injection molded blade ring was joined (see Fig. 16-6) to a hemispheri-
has a conical inner surface, and the outer cally shaped lid of the same material with
surface of the hub is green-machined to the same dimensions of the mating surface,
corresponding shape. After binder re- which on both components was machined
moval the hub and the blade ring are flat by diamond grinding (within a few
green-joined (Fig. 16-20). A design of the micrometers flatness over the entire sur-
joint area with lips - which can easily be face). Because of the requirement for the
removed on finish machining of the com- same long-term corrosion resistance of the
174 16 Hot Isostatic Pressing

joint as for the 100 mm thick container Frisch, A., Kaysser, W. A., Zhang, W, Petzow, G.
(1991), in: Hot Isostatic Pressing - Theory and Ap-
wall, no additives (e.g. glass formers) facil- plications: Koizumi, M. (Ed.). Amsterdam: El-
itating the joining could be used. sevier, pp. 319-325.
After the two components to be joined Frisch, A., Kaysser, W. A., Zhang, W, Petzow, G.
(1992), Ada Metall. Mater. 40, S. 361.
had been positioned, the gas pressure in- Gaskel, D. R. (1973), Introduction to Metallurgical
side the canister was reduced and on outer Thermodynamics. New York: McGraw-Hill,
sheet metal container made of low carbon pp. 253-268.
Haerkegaard, G., Liljeblad, J., Ohlsson, L. (1984),
steel was sealed gas-tight. The HIP joining ASME Paper 84-PVP-116.
was successfully carried out at a HIP pres- Heikinheimo, E., Kodentsov, A., Van Beck, J. A.,
sure of 100 MPa with a dwell time of 2-5 h Klomp, J. T., Van Loo, F. J. J. (1992), Acta Metall.
Mater. 40, SAIL
at 1350 °C at the joint. It is interesting to Heinrich, I, Boehmer, M. (1989), U.S. Patent
note that the joined area was approx. 4812272.
0.13 m 3 and the total force across the Helle, A. S., Easterling, K. E., Ashby, M. F. (1985),
Acta Metall. 33, 2163.
joined area approx. 200 MN (comparable Hermansson, L. A. G., Burstroem, M., Johansson,
to the weight of 2000000 kg)! Test bars T., Hatcher, M. (1988), Commun. Am. Ceram. Soc.
cut out by diamond saw across the joint 67, C183.
Hill, A., Wallach, E. R. (1989), Acta Metall. 37, 2425.
of half-scale canisters showed that the Hoenig, C , Otto, R., Stutler, W. (1991), in: Proc. 7th
mechanical strength of the joint (373 + CIMTEC: Vincenzini, P. (Ed.). Amsterdam: El-
55 MPa) was comparable to the strength sevier.
of the lid (370 + 45 MPa) and of the con- Hunold, K. (1985), Interceram 85, 38.
Hunold, K. (1986), Adv. Mater. Proc. 9, 5.
tainer wall (389 + 22 MPa). The tests were Isaksson, S.-E., Larker, H. T. (1971), U.S. Patent
made three-point bending tests with a 4 339 271.
20 mm span on 3 x 3 x 48 mm test bars. Ishizaki, K. (1990), Acta Metall. Mater. 35, 2059.
Ishizaki, K. (1991), in: Hot Isostatic Pressing - Theory
and Applications: Schaefer, R. J., Linzer, M. (Eds.).
Materials Park, OH: ASM Int., pp. 129-138.
Kim, J.-Y, Uchida, N., Kato, Z., Miyamoto, A., Ue-
matsu, K. (1992), in: Hot Isostatic Pressing - Theo-
ry and Applications: Koizumi, M. (Ed.). Amster-
dam: Elsevier, pp. 129-134.
16.8 References Kim, S. S., Baik, S. (1992), in: Hot Isostatic Pressing
- Theory and Applications: Koizumi, M. (Ed.). Am-
Adlerborn, I, Larker, H. T. (1974), U.S. Patent sterdam: Elsevier, pp. 67-72.
4455 275. Kito, T., Yabuta, K., Watanabe, M., Matsou, Y.
Adlerborn, I, Larker, H. T. (1976), U.S. Patent (1991), in: Hot Isostatic Pressing - Theory and Ap-
4112143. plications: Schaefer, R. J., Linzer, M. (Eds.). Mate-
Adlerborn, X, Larker, H. T., Mattsson, B., Nilsson, J. rials Park, OH: ASM Int., pp. 155-158.
(1978), German Patent 2950158 and U.S. Patent Koizumi, M. (1988), in: Hot Isostatic Pressing - The-
4478 789. ory and Applications: Garvare, T. (Ed.). Lulea,
Akselsen, O. M. (1992), J. Mater. Set 27, 569. Sweden: Centek Publishers, pp. 287-296.
Boehmer, M., Heinrich, J. (1980), German Patent Larker, H. T. (1966), U.S. Patent 3 470 303.
3 037 237. Larker, H. T. (1979a), in: Ceramics in Nuclear Waste
Boyer, C. B., Orcutt, F. D. (1967), U.S. Patent Management: Chikalla, T. D., Mendel, J. E. (Eds.).
3467 011. Springfield, VA, USA: NTIS.
Brito, M. E., Yokohama, H., Hirotsu, Y., Mutoh, Y. Larker, H. T. (1979b), in: AGARD: CP-276, 18/1-4.
(1989), ISIJ Int. 30, 1071. Larker, H. T. (1980), in: High Pressure Science and
Celis, P. B., Ishizaki, K. (1991), Materials at High Technology: Vodar, B., Marteau, P. (Eds.). Oxford:
Temperatures 9, 80. Pergamon Press, pp. 571-582.
Cundill, R. (1993), Ball Bearing J. 241, SKF, 26. Larker, H. T. (1984), in: Ceramic Components for
Dietze, M. (1991), Doctoral Thesis, D82 JUL-2521. Engines: Somiya, S., Kanai, E., Ando, K. (Eds.).
Elssner, G., Petzow, G. (1990), ISIJ Int. 30, 1011. Tokyo, KTK Scientific Publishers/D. Reidel,
Evans, A. G., Hsueh, C. H. (1986), /. Am. Ceram. pp. 304-310.
Soc. 69, AAA. Larker, H. T. (1985), Mater. Sci. Eng. 71, 329.
 16.8 References 175

Larker, H. T., Adlerborn, J., Bohman, H. (1977), SAE Richards, K. X, Benfer, R. H. (1991), /. Am. Ceram.
Technical Paper No. 770335. Soc. 74, 2014.
Larker, H. X, Adlerborn, J. E., Karlsson, E. (1993), Richerson, D. W, Wimmer, X M. (1983), Commun.
Ceram. Eng. Sci. Proc. 14, 274. Am. Ceram. Soc. 62, C173.
Larker, H. T, Adlerborn, J. E., Lundberg, R. (1995), Ritzhaupt-Kleissl, H.-X, Kiihne, A., Oberacker, R.
in: Ceramic Materials and Components for Engines; (1992), in: Hot Isostatic Pressing - Theory and Ap-
Yan, D. S., Fu, X. R., Shi, S. X. (Eds.). Singapore: plications: Koizumi, M. (Ed.). Amsterdam: El-
World Scientific, pp. 741-744. sevier, pp. 165-170.
Larker, R. (1992 a), J. Am. Ceram. Soc. 75, 62. Seino, H., Ishizaki, K., Takata, M. (1989), Jpn. J.
Larker, R. (1992 b), Doctoral Thesis 1992:102D, Appl. Phys. 28, L78.
(ISSN 0348-8373), Lulea University of Technolo- Seliverstov, D. G., Samarov, V., Goloveshkin, V. A.,
gy, Sweden. Alexandrov, S. E., Ekstrom, P. (1994), in: Hot Iso-
Larker, R., Beckman, T. (1995), in: Proc. 3rd Int. static Pressing '93: Delaey, L., Tas, H. (Eds.). Am-
Symp. on Structural and Functional Gradient Mate- sterdam: Elsevier, pp. 555-560.
rials: Ilschner, B., Cherradi, N. (Eds.). Lausanne, Shimada, M., Tanihata, K., Kaba, X, Koizumi, M.
Switzerland: Presses poly techniques et universi- (1984), in: Emergent Process Methods for High-
taires romandes, pp. 495-501. Technology Ceramics: Davis, R. F, Palmour III,
Larker, R., Loberg, B., Johansson, T. (1989), in: H., Porter, R. L. (Eds.). New York: Plenum,
Proc. 3rd Int. Symp. on Ceramic Materials and pp. 591-596.
Components for Engines: Tennery, V. J. (Ed.). Las Svoboda, A. (1994), Licentiate Thesis 1994: 29 L,
Vegas, NV: American Ceramic Society, pp. 503- (ISSN 0280-8242). Lulea University of Technolo-
512. gy, Sweden.
Larker, R., Anevik, K., Kristiansson, S., Loberg, B. Suganuma, K. (1990), ISIJ Int. 30, 1046.
(1992 a), Mater. Des. 13, 11. Suganuma, K., Okamoto, X, Miyamoto, Y, Shi-
Larker, R., Nissen, A., Pejryd, L., Loberg, B. mada, M., Koizumi, M. (1986), Mater. Sci. Tech-
(1992b), Ada Metall. Mater. 40, 3129. nol. 2, 1156.
Larker, R., Wei, L.-Y, Olsson, M., Loberg, B. Tanaka, I., Pezzotti, G., Okamoto, X, Miyamoto, Y,
(1992c), in: Proc. 4th Int. Symp. on Ceramic Mate- Niihara, K. (1992), in: Hot Isostatic Pressing - The-
rials and Components for Engines: Carlsson, R., ory and Applications: Koizumi, M. (Ed.). Amster-
Johansson, X, Kahlman, L. (Eds.). Gothenburg, dam: Elsevier, pp. 73-78.
Sweden: Elsevier, pp. 340-347. Traeff, A. (1990), Met. Powder Rep. 45 (4), 279.
Larker, R., Wei, L.-Y, Loberg, B., Olsson, M., Jo- World Report on Advanced Ceramics (1989), Engle-
hansson, S. (1994), /. Mater. Sci. 29, 4404. wood, NY, USA: Technical Insights Inc., pp. 1,
Li, W.-B., Ashby, M. R, Easterling, K. E. (1987), 2,5.
Acta Metall. 35, 2831. Yamada, X, Sekiguchi, H., Okamoto, H., Azuma, S.,
Lucek, J. W. (1990), ASME Paper 90-GT-165/1-7. Kitamura, A., Fukaya, K. (1987), High Temp.
Manabe, Y, Fujikawa, T, Narukawa, Y. (1991), in: Technol. 5, 193.
Hot Isostatic Pressing - Theory and Applications: Ziegler, G., Woetting, G. (1985), Int. J. High Tech.
Schaefer, R. X, Linzer, M. (Eds.). Materials Park, Ceram. 1, 31.
OH: ASM Int., pp. 139-144.
Moseley, S. G., Blackford, I, Jones, H., Greenwood,
G. W, Walker, R. A. (1991), in: Diffusion Bonding 2:
Stephenson, D. J. (Ed.). Amsterdam: Elsevier,
pp. 183-199. General Reading
Nicholas, M. G. (1991), in: Proc. Int. Institute of
Welding Congress on Joining Research: North, T. Delaey, L., Tas, H. (Eds.) (1994), Hot Isostatic Press-
H. (Ed.). London: Chapman and Hall, pp. 160- ing '93: Amsterdam: Elsevier.
171. Garvare, T. (Ed.) (1988), Hot Isostatic Pressing - The-
Nilsson, M. (1974), Interceram 23, 55. ory and Applications: Lulea, Sweden: Centek Pub-
Okamura, H. (1991), Mater. Sci. Eng. A143, 3. lishers.
Pejryd, L. (1992), in: Proc. 4th Int. Symp. on Ceramic Koizumi, M. (Ed.) (1992), Hot Isostatic Pressing -
Materials and Components for Engines: Carlsson, Theory and Applications: Amsterdam: Elsevier.
R., Johansson, X, Kahlman, L. (Eds.). Gothen- Larker, H. T. (1991), Hot Isostatic Pressing in: Engi-
burg, Sweden: Elsevier, pp. 50-66. neered Materials Handbook, Vol. 4: Schneider, S. X
Pujari, V. K., Tracey, D. M. (1993), ASME Technical (Ed.). Materials Park, OH: ASM Int., pp. 194-201.
Paper 93-GT-319. Schaefer, R. X, Linzer, M. (Eds.) (1991), Hot Isostatic
Pyzik, A. X, Pechenik, A. (1988), Ceram. Eng. Sci. Pressing - Theory and Applications. Materials
Proc. 9, 965. Park, OH: ASM Int.
17 Fired Microstructures and Their Characterization
Helen M. Chan and Martin P. Harmer

Materials Research Center, Lehigh University, Bethlehem, PA, U.S.A.

List of Symbols and Abbreviations 178

17.1 Characterization Techniques 179
17.1.1 Compositional Mapping Techniques 179
17.1.1.1 Scanning Auger Microscopy 179
17.1.1.2 Scanning Ion Microscopy 179
17.1.2 Techniques Which Reveal Topographical Contrast 180
17.1.2.1 Nomarski Interferometry 180
17.1.2.2 Environmental Scanning Electron Microscopy 181
17.1.2.3 Low Voltage Scanning Electron Microscopy 182
17.1.2.4 Differential Etching of Ferroelectric Domains 183
17.1.3 Piezospectroscopy 184
17.2 Defect-Containing Microstructures 184
17.2.1 Processing Defects 185
17.2.2 High-Temperature Defects 188
17.3 Tough Ceramic Microstructures 191
17.3.1 Process Zone Toughening Mechanisms 192
17.3.2 Bridging Zone Toughening Mechanisms 195
17.4 Novel Microstructures and Processing Methods 196
17.4.1 Fibrous Monolithic Ceramics 196
17.4.2 Duplex Bimodal Structures 197
17.4.3 Processing Techniques Involving Metallic Precursors 198
17.4.3.1 Reaction Bonding 198
17.4.3.2 Directed Metal Oxidation 199
17.4.3.3 Co-Continuous Ceramic Composites 199
17.4.4 Microstructures Formed by Controlled Nucleation 200
17.4.4.1 Seeding of YBa 2 Cu 3 O 6 + x High Tc Superconductors 201
17.4.4.2 Seeding of Boehmite Sol-Gels 201
17.4.5 Model Ceramic Microstructures 202
17.5 Electronic and Optical Ceramic Microstructures 205
17.6 Acknowledgements 210
17.7 References 210

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
178 17 Fired Microstructures and Their Characterization

List of Symbols and Abbreviations

E electron energy
Jc critical current density
incurve resistance curve
Tc superconducting critical temperature
C4 co-continuous ceramic composites
DIMOX direct metal oxidation
ESEM environmental scanning electron microscopy
HIP hot isostatic pressing
HPSN hot-pressed silicon nitride
LAS lithium alumino-silicate
MLC multilayer ceramic capacitor
PLZT lead lanthanum zirconium titanate
PMN Pb(Mg 1 / 3 Nb 2 / 3)O 3
PST Pb(Sc 1 / 2 Ta 1 / 2 )O 3
PSZ partially stabilized zirconia
RBAO reaction bonding of aluminum oxide
SAM scanning Auger microscopy
SEM scanning electron microscopy
SIM scanning ion microscopy
TEM transmission electron microscopy
TFR thick film resistor
TZP tetragonal zirconia polycrystals
YAG yttrium aluminum garnet
YTZP yttria tetragonal zirconia polycrystals
ZTA zirconia toughened alumina
 17.1 Characterization Techniques 179

17.1 Characterization Techniques due to the limited escape depth of the low
energy Auger electrons. The lateral resolu-
In any discussion of fired ceramic mi- tion is « 50 nm, and relative to conven-
cro structures, it is perhaps appropriate to tional X-ray mapping, SAM is more effi-
give a brief overview of the most important cient for low atomic number elements.
techniques which are employed to charac-
terize such structures. In this review, we
17.1.1.2 Scanning Ion Microscopy
will describe techniques for compositional
mapping, techniques which give topo- In scanning ion microscopy (SIM) (Soni
graphical contrast and a new technique for et al., 1994; Williams et al., 1991, 1993), a
characterizing the stress state in crystalline liquid metal ion source (usually Ga) is used
materials (piezospectroscopy). Note that to produce a focused beam of high energy
this review was not intended to be exhaus- (20-60 keV) ions which is rastered across a
tive; in particular such extensively used selected area of the specimen. This results
methods as conventional scanning and in the emission of secondary ions, neutral
transmission electron microscopy will not atoms and secondary electrons. Of these
be covered, and the reader is referred to signals, the secondary ions are of the great-
several excellent texts on the subject est importance for imaging in SIM. Dis-
(Goldstein et al., 1992; Williams, 1984; persion of the secondary ions with regard
Thomas and Goringe, 1979; Wachtman, to their mass/charge ratio can be achieved
1993). by mass spectroscopy. By mapping the in-
tensity of a certain species of ion as a func-
17.1.1 Compositional Mapping Techniques tion of the position of the ion probe, chem-
ical mapping of the surface can be
17.1.1.1 Scanning Auger Microscopy
achieved, in a manner highly analogous to
Irradiation by high energy electrons re- X-ray mapping in the scanning electron
sults in inner shell ionization events, such microprobe. Alternately, the total inten-
that the atom is left as an ion in an excited, sity of emitted secondary ions may be used
energetic state. De-excitation can give rise to modulate the image, in which case topo-
to either the emission of a characteristic graphic contrast may be achieved. Since
X-ray, or the ejection of an outer-shell the surface of the specimen is continually
electron (Goldstein etal., 1992; Williams, being eroded by the impingement of the
1984). These so-called Auger electrons (see primary ion beam, this technique can en-
Vol. 2B, Chap. 13, Sec. 13.2 of this series) able a three-dimensional construction of
have energies which correspond to transi- the distribution of a chemical species
tions between well-defined energy levels, within the specimen.
and hence are characteristic of the parent Compared to the electron microprobe,
atom. In scanning Auger microscopy SIM has the following advantages. Be-
(SAM) (Wachtman, 1993), compositional cause of the much lower penetration dis-
imaging is achieved by scanning the speci- tance of the primary ions, SIM has greater
men with a focused probe of high energy surface sensitivity. Further, SIM enjoys a
electrons, and using the intensity of emit- factor of 10 improvement in lateral resolu-
ted Auger electrons to modulate the tion over scanning electron microscopy
brightness of the image. This technique is (SEM). Excellent examples of the power of
highly surface specific (within 0.3-3 nm) this technique are shown in Figures 17-1
180 17 Fired Microstructures and Their Characterization

Figure 17-1. SIM Ca (left) and Mg (right) maps of polished A12O3 specimen doped with 250 ppm of MgO. Note
segregation of Ca to pores and grain boundaries (arrowed) and Mg segregation to pores only. Courtesy K. K.
Soni and A. M. Thompson.

and 17-2, which depict the preferential Clearly the technique has the potential to
grain boundary segregation of lanthanum reveal segregation at particular mi-
in an alumina sample doped with crostructural sites, e.g., special grain
1000 ppm La 2 O 3 , and the segregation of boundaries, pores etc., in a manner which
magnesium to pore surfaces in A12O3. is considerably less time consuming than
say analytical electron microscopy.
Finally, SIM images can distinguish be-
tween different isotopes of the same ele-
ment, and can provide chemical, as well as
elemental mapping, since the signal can
consist of clusters of atoms/ions or discrete
molecules. Unfortunately, relative to
SEM, the extraction of quantitative com-
positional information from SIM data is
relatively difficult.

17.1.2 Techniques Which Reveal

»'•*•/' • / " ' •
Topographical Contrast
17.1.2.1 Nomarski Interferometry
Nomarski interferometry is an optical
technique which accentuates any surface
Figure 17-2. SIM map showing preferential grain
boundary segregation of lanthanum in an alumina relief in the specimen (see Vol. 2 A,
sample doped with 1000 ppm La 2 O 3 . Courtesy K. K. Chap. 5, Sec. 5.3.1.2 of this series). A shad-
Soni and A. M. Thompson. owing effect is produced which has an ap-
 17.1 Characterization Techniques 181

pearance very much like oblique illumina-

tion. For details, the reader is referred to
the excellent text by Gifkins (1970). Essen-
tially however, in Nomarski interferome-
try, plane polarised illumination. passes
through one half of a Wollaston prism be-
fore impinging on the sample, and subse-
quently travels through the other half of f^yf*^ '-c*k'S'CVv
the prism before reaching the analyzer.
The prism is adjusted so that if the surface
were completely flat, the path lengths
through the prism would be completely
matched so that extinction would be ob-
tained at the analyser. Any surface relief
causes a change in the path difference and
hence the intensity of light. An example of
Nomarski contrast is shown in Figure 17-3,
which depicts surface damage in a coarse
grained alumina resulting from repeated
indentation from a spherical indenter
(Guiberteau et al., 1993). Uplifting of indi-
vidual grains, together with deformation
twinning are clearly delineated.

17.1.2.2 Environmental Scanning Electron

Microscopy
In the environmental scanning electron
microscope (ESEM), a specially designed
differential pumping system allows an at-
mosphere of up to 50 torr at the specimen, Figure 17-3. Optical micrographs of surface damage
whilst maintaining the required high vac- in coarse grained alumina resulting from cyclic inden-
uum in the remainder of the column, and tation (load = 1000 N, « = 10000 cycles, frequency =
at the gun. The two chambers are sepa- 1 Hz) by WC sphere (r = 3.8 mm). Tests in (a) air, and
(b) water. Marker denotes contact diameter (454 urn).
rated by a series of fine, pressure limiting
Viewed in Nomarksi interference illumination. Cour-
apertures. Any gas leaking out of the spec- tesy F. Guiberteau, N. Padture, H. Cai and B. R.
imen chamber through the apertures is Lawn.
successively extracted via a system of tubes
and pumps. The major concern with oper-
ation of the ESEM (as compared to a con- diameter is relatively insensitive to the gas
ventional SEM) is the degradation of per- pressure, there is a loss of beam intensity
formance due to scattering of the electron (Danilatos, 1988). Accordingly, to achieve
beam by gas molecules in the specimen a comparable signal to noise ratio, it is
chamber. Theoretical and experimental necessary to utilize a higher beam current,
studies have shown that whereas the beam and hence greater spot size. One advantage
182 17 Fired Microstructures and Their Characterization

of the ESEM (which is particularly conve- 17.1.2.3 Low Voltage Scanning Electron
nient for the study of ceramic specimens), Microscopy
is that the ionized gas molecules can dis-
charge the specimen, hence rendering coat- Since the majority of ceramics are poor
ing unnecessary. Details of the instrumen- electrical conductors, for conventional
tation and the development of the ESEM SEM imaging, the application of a thin
have been reviewed by Danilatos (1988, conducting coating (usually carbon) is nec-
1991). essary to prevent charging. This require-
In conjunction with a heating stage, the ment can be disadvantageous, however,
ESEM is particularly useful for in-situ because very fine surface relief may be ob-
study of reactions involving solids and scured by the coating. It has long been
gases (Brown etal., 1993; D. A. Lange recognized that this problem could be
etal., 1991; Rodriguez etal., 1992), melt- overcome by operating at very low acceler-
ing reactions (McKernan, 1993; Bergstrom ating voltages (1-3 keV), however, chro-
and Jennings, 1992), and also for specimens matic aberration proved to be the limiting
which are very environmentally sensitive. factor, since for thermionic electron
For an extensive bibliography of such sources the electron energy spread (A£) of
studies, see reference (Danilatos, 1993). An ^ 2 - 3 eV is an unacceptably large fraction
example of the use of ESEM is shown in of the total energy. Because of their much
Figure 17-4, which depicts the solidifi- narrower energy range (A£^0.3 eV), how-
cation of calcia-rich Bi 2 Sr 2 CaCu 2 O 8 in ever, the field emission sources in new gen-
3 torr of oxygen. The crystallization of eration SEM's can achieve low keV images
platelets of the superconducting phase of uncoated nonconducting specimens.
Bi 2 Sr 2 CaCu 2 0 8 is clearly visible, together For example, Figure 17-5 shows a pair of
with their subsequent growth at the ex- scanning electron micrographs which de-
pense of the calcia particles. pict crack healing in an A12O3 "nanocom-

Figure 17-4. In situ solidification of calcia-rich Bi2Sr2CaCu2O8 in 3 torr of oxygen. The crystallization of
platelets of the superconducting phase. Bi2Sr2CaCu2O8 is clearly visible, together with their subsequent growth
at the expense of the calcia particles (small dark features). Courtesy S. McKernan.
 17.1 Characterization Techniques 183

Figure 17-5. Low voltage (3.1 kV) SEM

images of an uncoated alumina-SiC
b) "nanocomposite" (Al 2 O 3 -5 vol.% SiC)
(a) prior to, and (b) after annealing for
2 h at 1300°C in argon. Note that heal-
ing of the indentation-induced crack has
taken place. Courtesy A. M. Thompson.

posite" containing 5 vol.% SiC particles.

The micrographs were recorded at an ac-
celerating voltage of 3 keV, and the speci-
mens were uncoated. Other advantages of
low voltage SEM include significantly re-
duced beam damage, and greater surface
sensitivity.

17.1.2.4 Differential Etching

of Ferroelectric Domains
Between 1460°C and 130°C, BaTiO3
adopts a cubic perovskite structure and is
paraelectric. On cooling below its Curie
temperature (130°C), BaTiO3 undergoes a
phase transformation from cubic to tetrag- (b)
onal (c/a^l.01). The tetragonal phase is
ferroelectric, the direction of spontaneous
polarization being parallel to the oaxis,
which lies parallel to one of the three orig-
inal <100> cube axes. Hooton and Merz
(1955) discovered that {100} surfaces of
ferroelectric BaTiO3 single crystals show a
differential etch rate in hydrochloric acid
(see Fig. 17-6). Namely, domains which
are oriented so that the polarization vector
is perpendicular to the surface and directed
outwards, etch at the greatest rate. In con-
strast, domains where the polarization vec-
tor is perpendicular to the surface but ori-
ented inwards, etch at the slowest rate. Figure 17-6. (a) SEM micrograph of an etched sur-
face of Ca-doped BaTiO3 (Ba0>95Ca 05 TiO 3 ). The
Ferroelectric domains in which the polar- encircled region shows an area of the specimen where
ization vector lies parallel to an in-plane the surface is parallel to (100). (b) Domain configura-
<100> direction etch at an intermediate tion consistent with (a). Courtesy Y.-H. Hu.
184 17 Fired Microstructures and Their Characterization

rate. As depicted in Figure 17-6, this differ- chromium ions in A12O3 when a single
ence in etch rates gives rise to topographi- crystal sapphire was compressed along its
cal contrast which can be used to identify c-axis.
the domain configurations (Hu et al., While the physical processes for the two
1986). phenomena differ significantly, the tech-
niques employing them for stress measure-
17.1.3 Piezospectroscopy ment are virtually the same (Adar and
Clarke, 1982; Ma and Clarke, 1993 b). The
Both Raman and optical fluorescence Raman or fluorescence signals induced by
spectra exhibit piezospectroscopic phe- a laser beam are analyzed by a monochro-
nomena and have been used to measure mator with a set of gratings to achieve ex-
both residual and applied stresses in crys- treme precision of frequency. The resulting
talline materials (Anastassakis et al., 1970; position of a characteristic spectral line
Ma and Clarke, 1993 a; Ma et al., 1994). In from a stressed sample is compared to that
the Raman effect, photons are scattered at the stress-free state to obtain the fre-
inelastically by a crystal, with creation or quency shift, from which the stress level is
annihilation of photons. The energies of determined using empirical piezospectro-
these normal vibration modes change sys- scopic coefficients. When an optical micro-
tematically with the applied stress, giving scope is used as part of the probe, high
rise to the frequency shifts of the corre- spatial resolution ( « 1 jim) can be achieved.
sponding Raman lines. The optical fluo- In addition, due to the imaging capabilities
rescence of a crystal is generated by impu- of the microscope, particular microscopic
rity ions, whose outer shell electrons, upon features can be easily located and probed
absorbing energy, can produce sharp char- selectively.
acteristic fluorescence lines during radia-
tive transition from an excited state to an
intermediate or ground state. The energies 17.2 Defect-Containing
of these outer shell electronic states are Microstructures
influenced by the surrounding crystal field
and are therefore sensitive to applied A key to the successful processing of ce-
stresses. Figure 17-7 illustrates the shifts of ramics, especially monolithic structural ce-
the characteristic K1 and R2 peaks of ramics where high fracture strength is de-

10000
Stress Free
3.2 GPa Compression J Neon
J? 7500
O
O
5000
c/)
c Figure 17-7. Optical fluorescence spectra
0

2500 of chromium ions in sapphire showing

the shifts of the characteristics peaks
(Rx and R 2 ) due to compression along
the c-axis. Courtesy Q. Ma.
14380 14400 14420 14440
Frequency (cm"1)
 17.2 Defect-Containing Microstructures 185

sirable, is the avoidance of microstructural

defects. In this section we provide exam-
ples of common types of microstructural
defects found in structural ceramics, pro-
duced both during firing and during test-
ing in a service environment. First, we will
show examples of processing-related mi-
crostructural defects of the type which lead
to mechanical failures in structural ceram-
ics at room temperature. Then we will give
examples of several types of microstruc-
tural defects corresponding to mechanical
failures at high temperatures.

17.2.1 Processing Defects

Large pores are a prevalent source of
mechanical failure in structural ceramics (as Figure 17-8. Large interagglomerate pore in an alu-
well as electrical failure in electronic ceram- mina ceramic (SEM, polished section). Courtesy S. J.
Bennison.
ics). An example of a large pore in an alu-
mina ceramic is illustrated in Figure 17-8.
This pore most likely originated from poor
powder packing in the starting powder. A
different type of large pore in a sample of
alumina-zirconia is shown in Figure 17-9.
The elongated shape and size of this pore
suggests that it was the remnant of a strand
of hair or an organic fiber that was acci-
dentally introduced into the starting pow-
der. Such observations highlight the need
to use clean handling procedures during
ceramic processing. Commercial manufac-
turers of advanced ceramic products are
increasingly resorting to the use of dedicat-
ed clean rooms for this purpose.
Abnormally large grains are also a com-
mon source of mechanical failure in struc-
tural ceramics. Figure 17-10 shows one
such abnormal grain in an alumina ceram-
ic. Note the transgranular fracture of the
Figure 17-9. Elongated pore due to contamination
abnormal grain compared to the inter-
by an organic fiber or a hair strand in an alumina-zir-
granular fracture of the matrix grains. Ab- conia ceramic (SEM, fracture surface). Courtesy F. F.
normal grain growth in this sample was Lange.
caused by inhomogeneous densification
during sintering. Inhomogeneous densifi-
186 17 Fired Microstructures and Their Characterization

stoichiometry and poorly mixed resulting

in the formation of a non-uniform distri-
bution of second-phase precipitates. Ab-
normal grain growth in this material was
believed to have resulted from non-uni-
form pinning of the grain boundaries by
the second phase particles. Abnormal
grain growth can also be promoted by a
small quantity of liquid phase at the grain
boundaries (Stuijts, 1977). Figure 17-12
shows abnormal grains in an alumina ce-
ramic containing a small amount (0.2%)
Figure 17-10. Large abnormal grain which acted as a of glassy second phase at the grain
fracture origin in a high purity alumina ceramic boundaries. Note the elongated grain
(SEM, fracture surface). Courtesy S. J. Bennison! structure which is a classic indication of
the presence of a liquid phase in alumina
(Rossi and Burke, 1973; Bennison and
cation caused local regions of the mi- Harmer, 1983, 1985). Abnormal grain
crostructure to densify faster than the sur- growth in alumina can be avoided either
rounding matrix. These locally dense by using ultra-high purity starting powders
patches then provided the nuclei for ab- and clean processing procedures to guard
normal grain growth (Harmer et al., 1983; against the formation of liquid-phase
Shaw and Brook, 1986). An extreme exam- forming contaminants, or by doping with
ple of abnormal grain growth in a sample MgO (Bennison and Harmer, 1983, 1985).
of Y3A15O12 (YAG) is illustrated in Figure Agglomerates and second phase inclu-
17-11. This material was prepared by hot- sions are a serious and common cause of
pressing a powder mixture of A12O3 and microstructural defects in fired ceramics.
Y 2 O 3 . The powder composition was off Examples of microstructural defects asso-
ciated with both hard and soft agglomer-
ates in alumina-zirconia ceramics are
shown in Figures 17-13 and 17-14 respec-
tively. The results of poor mixing leading
to phase separation and agglomeration in
alumina-zirconia is shown in Figure 17-15.
Phase segregation and agglomeration in
powder suspensions can be controlled by
adjusting the suspension chemistry. For
example, in the alumina-SiC system, meth-
anol produces a good dispersion of SiC,
whereas hexane causes the SiC particles to
flocculate (Stearns et al. 1992). Figures
17-16 and 17-17 compare the fired mi-
Figure 17-11. Abnormal grains in a sample of hot- crostructures of hot pressed composites of
pressed Y3A15O12 (YAG). The small dark specks are
second phase particles (optical, polished section). A12O3 containing 5vol.% of 0.15 micron
Courtesy J. D. French. SiC particles prepared using methanol and
 17.2 Defect-Containing Microstructures 187

Figure 17-12. Abnormal

grain growth and pore
entrapment in an undoped
(99.98% pure) alumina.
Note the elongated grain
morphology due to the
presence of a glassy phase
at the grain boundaries.
(SEM, polished section).
Courtesv S. J. Bennison.

Figure 17-13. Crack associated with a hard agglom- Figure 17-14. Circumferential void associated with
erate in an alumina-zirconia ceramic. (SEM, fracture an agglomerate of zirconia (light phase) in an alumi-
surface). Courtesy F. F. Lange. na-zirconia ceramic. (SEM, polished section). Cour-
tesy F. F. Lange.

hexane. The hexane-prepared samples pared samples the SiC particles are well
contain large voids due to agglomeration distributed throughout the A12O3 matrix,
of the SiC particles during mixing. The fi- and the final density reached 99%. Figure
nal density of these samples only reached 17-18 shows a fracture surface of a sample
92%. By comparison, in the methanol-pre- of Si 3 N 4 where the fracture origin can be
188 17 Fired Microstructures and Their Characterization

Figure 17-16. Nanocomposite of A12O3 containing

5 vol.% of 0.15 urn SiC particles prepared using
methanol, hot pressed at 1450 °C. Note the uniform
distribution of SiC particles. (SEM, polished section).
Courtesy L. C. Stearns.

Figure 17-15. Zirconia agglomerate due to poor mix-

ing in an alumina-zirconia ceramic, (a) Secondary
electron image, (b) Backscattered image showing that Figure 17-17. Nanocomposite of A12O3 containing
the agglomerate is rich in zirconia (lighter phase). 5 vol.% of 0.15 pirn SiC particles prepared using hex-
(SEM, free surface). Courtesy F. F. Lange. ane, hot pressed at 1450 °C. Note the agglomerated
SiC particles. (SEM, polished section). Courtesy L.
C. Stearns.

traced to a large second phase inclusion of 17.2.2 High-Temperature Defects

SiC. Significant improvements in the
strength and reliability of structural ce- The strength of a ceramic usually de-
ramics have been achieved through pro- creases with increasing temperature. The
cessing activities directed toward the sys- predominant high-temperature flaws are
tematic removal of strength degrading de- generally different from the flaws that con-
fects (E E Lange, 1989). trol strength at ambient temperature
 17.2 Defect-Containing Microstructures 189

a) b)
Figure 17-18. A large SiC inclusion in Si 3 N 4 which acted as a source of mechanical failure.
(SEM, fracture surface). Courtesy F. F. Lange.

Figure 17-19. Typical high temperature flaws in structural ceramics. (SEM, polished sections).
Courtesy B. J. Dalgleish.
190 17 Fired Microstructures and Their Characterization

(Fig. 17-19). High-temperature failure is a

gradual process that results from the wide-
spread accumulation of microstructural
damage, whereas failure of brittle materi-
als at ambient temperature occurs
catastrophically from preexisting flaws.
Damage formation at high temperatures
results from the ability of a material to
undergo creep and creep rupture (Evans
and Dalgleish, 1986). An important factor
with regard to high-temperature mechani-
cal behavior is the existence of a grain
boundary phase (Fig. 17-20). Hot-pressed Figure 17-21. High resolution lattice fringe image of
silicon nitride (HPSN) is a classic example grain boundary phase in hot pressed Si 3 N 4 . (TEM).
Courtesy D. R. Clarke.
of a material whose high-temperature
strength is compromised by the presence of
a grain boundary phase. HPSN is fabricat-
ed with the aid of sintering additives such al is reheated and stressed at temperatures
as MgO and Y 2 O 3 which react with SiO2 above the glass transition temperature it
impurity in the Si 3 N 4 to form a silicate fails by creep crack growth. The glassy
liquid phase. Upon cooling the liquid phase provides a region of rapid mass
phase solidifies to form a glassy phase situ- transport for the creep cavitation to occur
ated both at the triple points and as a thin rapidly. A diffuse network of cracks is
( « 1 nm) amorphous layer along the grain formed which ultimately interconnect
boundaries (Fig. 17-21). When the materi- leading to the total failure of the material.
Under some conditions of stress and strain
rate, ligaments of glassy phase can remain
intact behind the crack tip and act to di-
minish the creep crack growth rate. Figure
17-22 shows glass ligaments and diffuse
cracking in a crept sample of HPSN. The
cavities and cracks that form by high-tem-
perature deformation align preferentially
with respect to the direction of applied
stress. Figure 17-23 shows the alignment of
creep cavities in a sample of siliconized
SiC.
Cavity formation at high temperatures
can also occur in the absence of an applied
stress. Figure 17-24 shows the microstruc-
ture of a sample of hot-pressed alumina
after it was heated in air for 36 h at
Figure 17-20. Creep cavity penetrating an intergran- 1600°C. The sample started out fully
ular glassy phase along a grain boundary in alumina. dense, but bloated during the oxidizing an-
(TEM). Courtesy D. R. Clarke. nealing treatment due to the formation of
 17.3 Tough Ceramic Microstructures 191

Figure 17-22, Diffuse cracking in hot-pressed Si 3 N 4

due to creep rupture at high temperatures. Note the
presence of intact ligaments of glass bridging the
crack faces. (SEM, polished section). Courtesy N.
Tighe.
Figure 17-24. Hot-pressed MgO-doped A12O3 after
annealing in air at 1600°C for 36 h. Note the exten-
sive bloating and cavity formation due to oxidation
of carbon impurities forming CO gas bubbles. (SEM,
fracture surface). Courtesy of S. J. Bennison.

17.3 Tough Ceramic

Microstructures

In this section we will illustrate ceramic

microstructures designed to have high val-
ues of fracture toughness (10-30 MPam 1/2 ).
In all of the cases to be discussed toughen-
Figure 17-23. Alignment of creep cavities with re- ing results in a resistance curve (incurve),
spect to the direction of applied stress in siliconized in which the fracture resistance increases
SiC. (SEM, polished section). Courtesy of S. M. with crack extension (see Vol. 6, Chap. 12,
Wiederhorn.
Sec. 12.6.6.1 of this series). The slope of
the resistance curve (the tearing modulus)
determines the effect of the flaw size on the
grain boundary cavities (Bennison and material's strength. A microstructure with
Harmer, 1985). The cavitation is due to a a high tearing modulus will have a strength
grain boundary reaction between in-dif- that is less sensitive to the the size of the
fused oxygen and trapped carbon impuri- flaws. The material's strength then de-
ties at the grain boundaries which pro- pends on the form of the incurve and the
duces CO gas. Similar bloating reactions initial crack length (Evans, 1990).
have been observed in hot-pressed zirconia Numerous approaches have been devel-
and magnesia. oped to promote toughening in ceramics
192 17 Fired Microstructures and Their Characterization

(Evans, 1990; Harmer et al., 1992). Tough-

ening mechanisms can be conveniently di-
vided into two types: process zone and
bridging zone mechanisms.

17.3.1 Process Zone Toughening

Mechanisms
In process zone toughening the applied
stress activates a microstructural change
within a frontal "process zone" ahead of
the crack. Then, as the crack grows, it de-
velops an extended process zone in the Figure 17-25. Low magnification image of sintered
wake of the crack. If the microstructural Mg-PSZ transformation toughened ceramic. The
large grains are cubic zirconia ( « 50 urn) containing a
change within the process zone involves an
fine dispersion of tetragonal precipitates (^300 nm).
irreversible volume expansion, the crack is (SEM, polished section). Courtesy D. B. Marshall.
placed under compression and toughening
results. The extension of the process zone
with crack extension automatically leads
to incurve behavior. The toughening
mechanisms in this class include transfor-
mation toughening and microcrack tough-
ening.
Transformation toughening is mainly
confined to zirconia-containing ceramics
where the volume expansion arises from a
martensitic phase transformation from te-
tragonal to monoclinic zirconia. There are
three important classes of transformation
toughened ceramics, namely: partially sta-
bilized zirconia (PSZ), tetragonal zirconia Figure 17-26. Higher magnification image of Figure
polycrystals (TZP) and zirconia toughened 17-25, chemically etched to reveal the tetragonal pre-
alumina (ZTA). Figure 17-25 shows a low cipitates. (SEM, polished and chemically etched).
magnification view of a partially stabilized Courtesy D. B. Marshall.
zirconia containing 9 mol% MgO (i.e.,
Mg-PSZ). The material was sintered in the
cubic phase field and then heat treated in to reveal the microstructure. This material
the cubic plus tetragonal phase field. The has a steady state toughness (i.e. at large
resultant microstructure consists of cubic crack extensions) of «12-14 MPa m 1/2
ZrO 2 grains 50 microns in diameter, con- and a flexural strength of «600 MPa
taining lens-shaped tetragonal precipitates which does not vary significantly with flaw
(35-40 vol.%) with largest dimension size (Marshall, 1986). Figure 17-27 shows
300 nm. The tetragonal precipitates can be the microstructure of a yttria tetragonal
seen in the higher magnification image of zirconia polycrystal containing 3 mol%
Figure 17-26 which was chemically etched Y 2 O 3 (Y-TZP). The material was sintered
 17.3 Tough Ceramic Microstructures 193

Figure 17-27. Microstructure of yttria zirconia poly- Figure 17-28. Microstructure of zirconia-toughened
crystal (Y-TZP) containing 3 mol% Y 2 O 3 . (SEM, alumina (ZTA) containing 15mol% of 3Y-ZrO 2 .
polished section). Courtesy R. Hannink. (SEM, polished section). The light phase is tetragonal
zirconia. Courtesy D. R. Clarke.

in the all tetragonal phase field to produce within regions of local residual tension
a sub-micron grain microstructure of all caused by thermal expansion mismatch. If
tetragonal grains. Y-TZP's are usually the second phase particles have a higher
much stronger but less tough than thermal expansion coefficient than the ma-
Mg-PSZ's. Typical strength and toughness trix, the particles will be in hydrostatic ten-
values for Y-TZP fall in the range of sion and the stress field of the applied load
0.5-1.5 GPa. and 6-8 MPa m 1/2 respec- will cause the particles to microcrack.
tively. Figure 17-28 shows the microstruc- Conversely, if the matrix has a higher ther-
ture of a zirconia toughened alumina ce- mal expansion coefficient than the second
ramic (ZTA) containing 15mol% of te- phase particles, the matrix will microcrack.
tragonal zirconia particles. ZTA's have In either case a volume expansion occurs
been produced with toughnesses in the governed by the volume displaced by the
range of ^ 4 - 8 MPa m 1/2 . microcrack, producing toughening in an
Microcrack toughening has been identi- analogous way to transformation tough-
fied in various two-phase ceramics includ- ening. Several techniques have been ap-
ing A12O3 toughened with monoclinic plied to characterize microcrack process
ZrO 2 , and SiC toughened with TiB2 zones. The scanning acoustic microscope
(Evans, 1990; Riihle et al., 1987). In micro- (SAM), which allows measurement of the
crack toughening, the stress field of the elastic modulus with high spatial resolu-
applied load causes microcracks to occur tion (Quinten and Arnold, 1989) has been
194 17 Fired Microstructures and Their Characterization

a)

b)

C)

Figure 17-29. Bright field images of microcracks due

to phase transformation from tetragonal to mono-
clinic in zirconia. The contrast of the microcrack re- Figure 17-30. Microcracks associated with A12O3 ag-
verses in going from the under-focus condition (a), to glomerates in an alumina-mullite matrix. (Low-
the overfocus condition (b). (TEM). Courtesy D. R. voltage SEM, polished section, uncoated). Courtesy
Clarke. A. Khan.

successful in recording the process zones in radial microcracks depends on their incli-
polycrystalline alumina and a glass ceram- nation with respect to the electron beam.
ic material (Evans 1990). Transmission Tilting in the TEM is limited to + or - 45
electron microscopy (TEM) has been used degrees in all directions, therefore the frac-
to characterize microcracks in A12O3/ tion of detectable microcracks is limited to
ZrO 2 and SiC/TiB2 (Evans, 1990). A TEM 30% (Evans, 1990). Low-voltage high-
image of microcracks in Al 2 O 3 /ZrO 2 is resolution SEM on uncoated specimens
shown in Figure 17-29. The detectability of was able to resolve microcracks in the
 17.3 Tough Ceramic Microstructures 195

Al2O3-mullite system as illustrated in Fig-

ure 17-30. Staining followed by optical ex-
amination has been successful in revealing
the microcrack process zone in the A12O3-
ZrO 2 system as shown in Figure 17-31.

17.3.2 Bridging Zone Toughening

Mechanisms
Toughening by bridging occurs when re-
inforcing elements (such as fibers,
whiskers, ductile phases and large grains)
remain intact behind the crack tip, shield- Figure 17-31. Microcrack process zone in alumina-
ing the crack tip from the full effect of the zirconia revealed by staining. (Optical, polished sec-
applied load (Evans, 1990). Ductile rein- tion). Courtesy E. H. Lutz and N. Claussen.
forcements can significantly increase the
toughness by bridging, the classic example
being WC/Co which has a fracture tough-
ness of ^ 1 6 M P a m 1/2 . A12O3/A1 pro-
duced by the directed metal oxidation
(DIMOX) process is another example of a
ductile reinforcement toughened material.
In brittle systems, by far the greatest
toughening occurs in ceramic matrices
reinforced with continuous fibers. Figure
17-32 shows a fracture surface of a lithium
alumino-silicate (LAS) glass reinforced
with continuous SiC fibers. Toughening in
these systems is dominated by debonding
and sliding resistance at the fiber-matrix
interface. Fibers are often coated in .order
to control the interfacial bond strength
and shear resistance (Figure 17-33). A Figure 17-32. Fracture surface of lithium alumino-
technique for measuring the frictional silicate glass (LAS) reinforced with continuous SiC
stresses between the matrix and individual fibers showing fiber debonding and pull-out. (SEM).
Courtesy D. R. Clarke.
fibers has been developed by Marshall and
Oliver (1987). A standard microhardness
indenter is used to apply a force to the end
of the fiber and depress it below the matrix ness by grain-bridging. Figure 17-35 illus-
surface as shown in Figure 17-34. The fric- trates a crack interacting with elongated
tional stress is calculated from measure- grains in a silicon nitride ceramic to pro-
ment of the applied force and the amount duce toughening by grain bridging.
of slipping between the fiber and matrix.
In monolithic ceramics, such as Si 3 N 4 ,
elongated grains can lead to higher tough-
196 17 Fired Microstructures and Their Characterization

Coblenz (1988), Halloran and co-workers

have developed a variety of so-called fi-
brous monolithic ceramics of varying
phase constituents (Baskaran etal., 1993;
Baskaran and Halloran, 1993, 1994). As
depicted in Figure 17-36, the above ceram-
ics consist of elongated cells of one phase
separated by thin cell boundaries of a sec-
ond phase. If the cell boundary material is
deliberately chosen to be mechanically
weak, e.g., graphite, then delamination of
the cells occurs during fracture, leading to
noncatastrophic or graceful failure (Basker-
an and Halloran, 1993). Clearly by suitable
tailoring of the phases and morphology,
the potential exists to design microstruc-
tures with greatly increased work of frac-
tures.
Details of the processing procedure are
Figure 17-33. Sapphire fiber coated with copper by described by Baskaran et al. (1993). Essen-
electron beam evaporation. (SEM). Courtesy C. Levi. tially, the process consists of producing
green fibers of the primary phase, dip coat-
ing the green fibers in a suspension of the
second phase powder, consolidating the
17.4 Novel Microstructures and coated fibers in a die, and finally sintering/
hot pressing. The technique is very ver-
Processing Methods satile, and readily applied to a variety of
ceramic compositions. To date, fibrous
17.4.1 Fibrous Monolithic Ceramics
monolithic microstructures have been
Recently, a class of ceramics has achieved for the following cell/cell wall
emerged possessing somewhat unique mi- combinations: SiC/graphite, Si 3 N 4 /BN,
crostructural characteristics. Based on a A12O3/A12O3: ZrO 2 , Al 2 O 3 /Al 2 TiO 5 ,
processing concept first proposed by Al 2 O 3 /Ni and Ce-TZP/Ce-TZP: A12O3.

Figure 17-34. Fiber in SiC-

lithium alumino-silicate
composite after indentation
with a triangular pyramid
indentor. (SEM, polished
section). Courtesy D. B.
Marshall.
 17.4 Novel Microstructures and Processing Methods 197

Figure 17-35. Cracks interacting with elongated grains

in a high toughness Si3N4. The fracture toughness of
the material is ^ 9 M P a m 1/2 at a crack size of
^800 jim. (SEM, polished section). Courtesy C. W.
Li.

17.4.2 Duplex Bimodal Structures

If ceramic powder is mixed with spray
dried agglomerates of a different powder
composition, it is possible to achieve novel
ceramic composites in which approximate-
ly spherical zones of one type of structure
are dispersed in a matrix of a second struc-
ture. Claussen and co-workers (Claussen,
1984; Lutz et al., 1991; Lutz and Claussen,
1991 a, b) have designated such structures
"duplex", and have used this approach in
compositions based on partially stabilized
zirconias. Specifically, Lutz et al. (1991)
reported that if the agglomerates are
higher in m-ZrO 2 content relative to the
matrix, on transformation, pressure zones
are created which give rise to radial com-
pressive stresses in the matrix. These in Figure 17-36. SEM micrographs of polished surfaces
turn lead to enhanced crack branching and of a SiC/BN fibrous monolith prepared with finely
incurve behavior. chopped, randomly oriented fibers (SiC-cell, BN-cell
wall): (A) Section perpendicular to hot-pressing di-
In the case where single phase agglomer- rection, (B) section parallel to hot-pressing direction.
ates are dispersed in a two-phase powder, Courtesy S. Baskaran and J. W. Halloran.
the more rapid grain growth in the single
phase material results in a microstructure
with a bimodal grain size (Harmer et al.,
1992). Such a structure is depicted in Fig-
ure 17-37, which shows spherical regions
198 17 Fired Microstructures and Their Characterization

ing. Note that although the reaction bond-

ing (RBAO) process can be taken to com-
pletion to produce 100% ceramic samples,
both the directed metal oxidation and C 4
materials (see Sec. 17.4.3.3) contain residu-
al metal.

17.4.3.1 Reaction Bonding

Recently Claussen and co-workers have
developed a novel method of processing
alumina based ceramics which they have
Figure 17-37. Optical micrograph showing duplex bi- designated RBAO (reaction bonding of
modal sample consisting of 30 vol. % alumina ag- aluminum oxide) (Claussen etal., 1989,
glomerates dispersed in a matrix of 50:50 vol.% alu-
mina :mulite. Courtesy A. Khan.
1990; Wu et al., 1991). In the RBAO pro-
cess, a mixture of alumina and aluminum
powder (30-60 vol.%) is intensively attri-
tion milled, and then die or isostatically
of coarse-grained alumina in a matrix of
pressed at pressures ranging from 300-
fine grained alumina-mullite. These so-
900 MPa. Next, the compacted green bod-
called duplex bimodal structures are
ies are subjected to an oxidation/sintering
promising in that they possess superior
heat-treatment. Under suitable processing
flaw tolerance to the monolithic matrix
conditions, 100% ceramic bodies can be
materials, yet exhibit comparable resis-
achieved, with theoretical densities of the
tance to microstructural coarsening. This
order of 96-98 %. Advantages of this pro-
phenomenon has been successfully demon-
cess include: a) improved strength and
strated in several duplex bimodal systems,
machinability of the green bodies resulting
including Al 2 O 3 /Al 2 TiO 5 (Padture et al.,
from the high metal content, b) the ability
1991, 1993) and Al 2 O 3 /c-ZrO 2 (French,
to use relatively inexpensive, low purity
1990).
starting materials, and c) near net shape
forming, since the volume expansion on
17.4.3 Processing Techniques Involving
oxidation of the aluminum powder can
Metallic Precursors
compensate for the shrinkage on sintering.
In recent years, there has been renewed As yet, the reaction mechanisms of the
interest in the processing of ceramic mate- RBAO process are not fully understood,
rials by reaction of metallic precursors, although it has been determined that a sig-
particularly with regard to high Tc super- nificant fraction of the oxidation reaction
conductors (Yurek etal., 1987; Haldar takes place while the aluminum is still in
etal., 1987; Yurek etal., 1988; Sandhage the solid state. Further, the presence of
et al., 1991). In addition, several novel pro- ZrO 2 has been found to be beneficial to the
cessing techniques (all involving the oxida- process, the postulate being that it en-
tion of aluminum) have been developed, hances the oxygen transport kinetics. Per-
which offer inherent advantages over con- haps not surprisingly, given the complexity
ventional processing. These methods will of the reaction sequence (Wu et al., 1993),
be described in greater detail in the follow- the final microstructures are highly sensi-
 17.4 Novel Microstructures and Processing Methods 199

tive to processing parameters such as the

milling conditions, pressing pressure, and
heating rate. If these parameters are not
optimised, bloating or cracking of the sam-
ples may occur, and/or there may be in-
complete reaction at the specimen core
(Holz etal., 1994). Although originally
used in the processing of alumina, the
RBAO technology has also been success-
fully applied to other ceramic systems, e.g.
mullite (Wu and Claussen, 1991; Claussen Figure 17-38. SEM micrograph depicting the struc-
and Wu 1991), Al 2 O 3 -ZrO 2 (Wu etal., ture of a mullite sample processed by the reaction
1993; Wu and Claussen, 1992), and bonding technique. The starting powder composition
Al 2 O 3 -SiC (Gesing et al., 1990). For ex- was 34 vol.% A12O3, 26 vol.% SiC and 40 vol.% AL
ample, Figure 17-38 is a SEM micrograph The bright grains are t-ZrO2 introducted during
milling. Courtesy S. Wu and N. Claussen.
depicting the structure of a mullite sample
processed by the reaction bonding tech-
nique. Notice the high density and lack of
residual aluminum. The bright grains are The chief drawback of DIMOX, however,
t-ZrO 2 introduced during the milling pro- is the relatively slow rate of reaction, 0.2-
cess. 0.3 mm/h.
Current understanding of the A12O3
growth mechanism during directed metal
17.4.3.2 Directed Metal Oxidation
oxidation has been reviewed by Brandon
In the directed metal oxidation (1994). It seems that the presence of magne-
(DIMOX) process pioneered by Lanxide sium as an alloying element (> 0.2 wt.%)
Corporation (Newkirk etal., 1986, 1987; is vital to the DIMOX process, since oxi-
Claussen and Urquart, 1990; Brandon, dation proceeds via several stages, involv-
1994), a planar oxidation front grows out- ing both the formation of a spinel layer,
wards from a reservoir of molten alu- and then a thin metastable layer of magne-
minum, permeating through a porous pre- sium oxide. It is the reduction of this
form. The resultant structure consists of metastable magnesium oxide which sup-
coarse, columnar grains of A12O3, with in- plies the oxygen to the aluminum melt
terconnecting channels of aluminum met- for subsequent growth of a-Al 2 O 3 . Figure
al. The process enjoys several advantages 17-39 illustrates the microstructure of a
over conventional ceramic processing SiC particle reinforced A12O3 matrix com-
techniques. Perhaps the most important of posite which was made by the DIMOX
these is that since the oxidation front can process.
be readily contained within a preform or
mold, near net shape forming of large
17.4.3.3 Co-Continuous Ceramic
components can be achieved. In addition,
Composites
the precursor materials are relatively inex-
pensive, and the processing temperatures A new technique for producing co-con-
(950-1250 °C) are significantly lower than tinuous composites (C4) of A12O3/A1 com-
those required for conventional sintering. posites has been discovered by Breslin
200 17 Fired Microstructures and Their Characterization

A unique aspect of this process is that

the transformed material closely replicates
that of the glassy preform, thus offering
the potential for near netshape forming of
complex components! Another advantage
of this processing method is that the
reported reaction rates (> 1 mm/h at
1050 °C), are rapid enough to be economi-
cally viable. Although details of the reac-
tion mechanism are not yet fully under-
stood, Breslin et al. (1993, 1994) have pos-
tulated a model whereby reduction of the
Figure 17-39. Optical micrograph showing com- silica occurs initially at the surface of the
posite consisting of SiC particles in a matrix of preform, resulting in a thin layer of alumi-
A12O3/A1 produced by the DIMOX process. Cour- na. It is proposed that subsequent penetra-
tesy P. Niskanen.
tion of the molten aluminum occurs
through cracks in the alumina layer, which
are induced by the tensile stresses resulting
(1993). The method is experimentally
from the volume decrease on transforma-
straight-forward, and simply involves im-
tion. C 4 materials have attractive proper-
mersing a sacrificial silica preform in
ties for automotive applications, these in-
molten aluminum. Reduction of the pre-
clude high thermal conductivity, good
form takes place such that the silica is com-
strength (^700 MPa) and toughness, and
pletely replaced by a fine interpenetrating
improved stiffness and wear resistance rel-
network of aluminum and alumina. A mi-
ative to 100% metallic components.
crograph depicting this unusual mi-
crostructure is shown in Figure 17-40.
17.4.4 Microstructures Formed by
Controlled Nucleation
One approach which has proven to be
very successful in influencing the final mi-
crostructure of a ceramic body is that of
controlled nucleation. In this procedure,
"seeds" of the desired orientation/phase
are incorporated into the green body, and
these can subsequently behave as preferred
nucleation sites for transformations occur-
ing during the sintering (or other consoli-
dation) process. Specific examples of the
application of controlled nucleation are
Figure 17-40. SEM micrograph of a typical mi-
crostructure of a co-continuous ceramic composite described in the following.
(C4) consisting of a-alumina (light phase) and alu-
minum (dark phase). The sample was prepared by
immersing a high purity silica glass preform in com-
mercially pure aluminum for 4 h at 1423 K. Courtesy
M. C. Breslin.
 17.4 Novel Microstructures and Processing Methods 201

17.4.4.1 Seeding of YBa 2 Cu 3 O 6+JC High

Tc Superconductors
Due to the highly anisotropic nature
of the transport properties of the high
temperature ceramic superconductor
YBa 2 Cu 3 0 6 + : c (123), a great deal of effort
has been directed at controlling the crystal-
lographic texture in these materials. To
date, the processing method which has
yielded the highest values of Jc (critical
current density) is that of melt-texturing
(Jin et al., 1988; McGinn et al., 1990; Mor-
ita et al., 1990). Although there are many
variations of this technique, essentially it
involves heating the green body above its
incongruent melting temperature, so that
the 123 dissociates into Y 2 BaCuO 5 (211)
and a liquid phase. On subsequent cooling,
the reverse reaction occurs, resulting in a
microstructure consisting of domains of Figure 17-41. Optical micrograph showing preferen-
tial nucleation of YBa 2 Cu 3 O 6+x grains at the inter-
highly aligned, plate-like grains of 123.
face between the YBa 2 Cu 3 O 6+x melt and the reaction
The degree of alignment can be significant- layer (R) formed at the A12O3 seed particle (A). The
ly enhanced by the application of a tem- sample was held at 0.5 h at 1050 °C, 0.5 h at 970 °C,
perature gradient during the cooling pro- and then quenched; the incongruent melting tempera-
cess. Furthermore, several groups have ture of YBa 2 Cu 3 O 6 + x is ^1005°C. Courtesy Y. L.
Chen.
achieved almost single domain bulk speci-
mens (of the order of 100 cm3) by seeding
the green body with a single crystal of
REBa 2 Cu 3 O 6 + ;c, where RE is a rare earth 17.4.4.2 Seeding of Boehmite Sol-Gels
element, e.g., Sm, Eu, Gd etc. (Murakami, When boehmite (y-AlOOH) is heated, it
1992; Morita etal. 1991). The advantage undergoes the following sequence of reac-
of using the rare earth substituted compo- tions:
sition, is that the seeds exhibit a higher
melting point than the 123 powder, and boehmite —>y—> 5 — • 6 - • OC-A1 2 0 3
hence remain completely solid during the Messing and co-workers (Kumagai and
melt-texturing process. Messing, 1984, 1985) have demonstrated
Preferred nucleation of 123 during melt- that by seeding the boehmite with a-Al 2 O 3
texturing has also been observed at other particles, the kinetics of the 9 - K X - A 1 2 O 3
types of heterogeneous sites. For example, transformation are increased, and there is
Chen et al. (1993) have shown that when a lowering of the transformation tempera-
sintered A12O3 particles (500-1000 jim) ture. Further, by modifying the intermedi-
are introduced into pressed powder pellets ate pore structures during sintering, for a
of 123,123 domains nucleate preferentially given heat-treatment seeding results in a
at the reaction layer between the particles denser, more uniform final microstructure
and matrix (see Fig. 17-41). (see Fig. 17-42). The effectiveness of the
202 17 Fired Microstructures and Their Characterization

a) most elegant method developed by Glaeser

and Rodel involves the use of photolithog-
raphy to produce arrays of pores and pore
channels of controlled size and shape in
sapphire. The patterned sapphire surfaces
are bonded to dense polycrystalline alumi-
na to generate interfaces with highly con-
trolled pore structures. These tailored mi-
crostructures have been utilized to study
pore-drag (see Fig. 17-43) and crack heal-
ing in alumina (Rodel and Glaeser, 1987,
b) 1990).
Model microstructures for studying fi-
nal-stage sintering have been developed
using a latex sphere impregnation-burnout
technique (Zhao and Harmer, 1992). In
this procedure, ceramic powder is mixed
with latex spheres of controlled size which
upon calcination, burn out to create artifi-
cial pores of similar size. During subse-
quent sintering and grain growth the struc-
ture evolves into an ideal final-stage mi-
Figure 17-42. SEM micrographs showing difference
in sintered structures for unseeded and seeded boeh- crostructure as shown in Figure 17-44.
mite gels: (a) Unseeded boehmite gel, heated in air at Such model microstructures have been
1200 °C for 100 min. (b) Boehmite gel seeded with used to study the effect of pore distribution
5 x 1013 a-Al2O3 seed particles/cm3, sintered in air at on microstructure development during fi-
1185°C for 100 min. Courtesy G. Messing.
nal-stage sintering (Zhao and Harmer,
1988, 1992). Figure 17-45 demonstrates
seeding procedure was found to depend on
the size of the a-Al 2 O 3 seed particles, thus
0.1 jim seeds were found to more beneficial
in promoting high final densities as com-
pared to 0.4 |im seeds (Kumagai and
Messing, 1985). In addition, Kumagai and
Messing (1985) reported an optimum seed
volume fraction of ^ 5 x l O 1 3 seeds/cm3
(1.5 wt.% of 0.1 |im seeds).

17.4.5 Model Ceramic Microstructures

In this section we provide examples of
several types of model microstructures
Figure 17-43. Periodic array of pores being dragged
used in ceramic processing research. Vari- by a migrating interface between sapphire and poly-
ous methods have been developed to intro- crystalline alumina. (SEM, polished section). Cour-
duce controlled porosity into ceramics. A tesy J. Rodel.
 17.4 Novel Microstructures and Processing Methods 203

Figure 17-44. Model fmal-

stage microstructure in alu-
mina produced by the latex
sphere impregnation and
burnout technique.
Left: Fracture surface.
Right: Polished section.
(SEM, polished section).
Courtesy J. Zhao.

the effect of grain growth and grain coor-

dination number on the closure of large
pores in alumina. At small grain sizes (less
than a critical value) the large pore is un-
able to close during sintering for thermo-
dynamic reasons. Below a critical grain co-
ordination number, the pore is thermody-
namically stable due to the concave pore
curvature (with respect to the solid) set by
the high coordination of grains surround-
ing the pore. Above the critical grain coor-
dination number, the curvature of the pore
becomes favorable for it to shrink, howev-
er, the kinetics of the process is so slow
that no significant shrinkage actually oc-
curs (Zhao and Harmer, 1988).
Readey has developed a process for pro- b)
ducing controlled porous microstructures
in ceramic powder compacts by sintering
in a reactive gaseous environment to pro-
mote coarsening by enhanced vapor trans-

Figure 17-45. Microstructure of undoped A12O3 im-

pregnated with large ( « 5 urn) model spherical pores
after firing for several minutes at 1620°C (a) and 5 h
at 1800°C (b). Note the change in pore curvature
from concave in (a) to convex in (b) due to a change
in the grain coordination number. Despite the favor-
able curvature in (b), the pore persists due to slow
densification kinetics. (SEM, polished section). Cour-
tesy J. Zhao.
204 17 Fired Micrestructures and Their Characterization

port (Readey etal., 1987). Figure 17-46

shows the final microstructure of a com-
pact of Fe 2 O 3 that was subjected to en-
hanced vapor transport by sintering in an
atmosphere containing HC1 gas. The final
grain size is ^25jum and the sample is
only 55 % dense. The starting particle size
was only 0.2 |im. The method has been ap-
plied to a variety of other oxides including
ZnO, MgO and A12O3.
A technique for introducing artificially
Figure 17-46. Coarsened micro structure of Fe 2 O 3 induced internal cracks into single crystal
processed by enhanced vapor transport. The sample ceramics and glasses is laser-induced
was sintered in 50% HC1 + 50% He at 1300°C for cracking (Wang etal., 1988). By focusing
30 minutes. The structure is ~ 5 5 % dense and the the radiation of a laser beam into the inte-
starting particle size was 0.2 urn. (SEM, fracture sur-
face). Courtesy D. W. Readey. rior of a transparent solid, well defined
internal cracks can be generated. Figure
17-47 shows a laser-generated internal
crack in LiF. The damage pattern in LiF
caused by the laser consists of penny-like
radial cracks generated along {100}
planes. The stress birefringence reveals a
strong residual stress field around the
damage zone strikingly similar to that cre-
ated by a Vickers indentation. Laser-in-
duced cracks have been used to study
crack healing and fracture phenomena in
various single crystal ceramics and glasses
(Wang et al., 1992). A key feature of this
method is that it produces internal surfaces
free of any environmental contamination.
Monahan and Halloran (1979) and
more recently Kaysser et al. (1987) have
developed model microstructures for in-
vestigating abnormal grain growth in alu-
mina. Large single-crystal spheres of sap-
phire ( « 500 jim diameter) were embedded
into A12O3 powders of different chemical
composition, densified by hot-pressing,
and then annealed to study the grain
Figure 17-47. Laser induced crack in LiF crystal. Pri-
mary cracks are formed along {100} planes. Stress
growth behavior. Figure 17-48 illustrates
birefringence reveals a strong residual stress field the microstructure of a single-crystal seed
around the damage zone. (Optical, transmitted light, grown into a matrix of MgO-doped alumi-
crossed polars). Courtesy Z. Y Wang. na.
 17.5 Electronic and Optical Ceramic Microstructures 205

Figure 17-48. Sapphire single crystal seed growing

into a matrix of A12O3 + 0.1 wt.% MgO. The sample
was hot pressed (35 MPa) at 1850 °C for 1 h. (Optical,
polished section). Courtesy J. Blendell.

10 urn
Figure 17-49. Microstructure of a commercial alu-
mina substrate used in the manufacture of hybrid
17.5 Electronic and Optical circuits. (SEM, polished section). Courtesy A. DiGio-
Ceramic Microstructures
Ceramics are used in a wide variety of sists of convoluted chains of highly ag-
electronic and optical applications includ- glomerated conductive particles of
ing electronic packaging, hybrid circuits, Pb 2 Ru 2 O 6 embedded in a lead silicate
capacitors, transducers, actuators, sensors, glass. The individual conductive oxide par-
electro-optics, infrared windows and lamp ticles within the agglomerates are separat-
envelopes (Moulson and Herbert, 1990). ed by a thin (1-5 nm) amorphous layer
Due to space limitations we will limit our (Fig. 17-50). The resistance of a TFR is
discussion here to just a few important determined by several factors including the
types of electronic and optical ceramics. volume fraction and connectivity of the
Electronic packaging is the single largest conductive particles, and the solubility of
application of electronic ceramics. Alumi- the conductive oxide in the glass. The con-
na continues to be the material of choice nectivity of the conductor particles is con-
for most ceramic substrate and packaging trolled by adjusting the particle size ratio
applications. Figure 17-49 shows the mi- of glass frit and conductive particles. The
crostructure of a commercial alumina sub- small Pb 2 Ru 2 0 6 particle agglomerates
strate material used in the manufacture of ( « 300 nm) are located around the larger
hybrid circuits. This material contains a ( « 2 jam) glass particles prior to the melt-
glassy phase at the grain boundaries which ing of the glass phase. Following melting
promotes diffusional bonding between the of the glass, the Pb 2 Ru 2 O 6 agglomerates
substrate and the materials which make up form convoluted chains (bright areas in
the circuit overlayers (thick film resistors, Fig. 17-50) around the sites of the original
conductors etc.). glass particles. Limited solubility of the
The microstructure of a thick film resis- conductive oxide in the glass leads to a
tor (TFR) is shown in Figure 17-50. It con- compositional gradient of ruthenium ions
206 17 Fired Microstructures and Their Characterization

a) ture. The liquid phase crystallizes on cool-

ing to form a solid second phase at the
grain boundaries. This grain boundary
phase lowers the thermal conductivity of
the material. The thermal conductivity can
be increased by a post-sintering heat treat-
ment designed either to remove the grain
boundary phase by evaporation, or to pro-
mote dewetting of the second phase to in-
crease the amount of A1N grain-grain con-
tact. A1N ceramics also contain planar de-
b) fects within the grain interiors which affect
the thermal conductivity (Fig. 17-52). De-
tailed microstructural and microchemical
characterization has revealed that these
defects are inversion domain boundaries
rich in oxygen (Westwood and Notis,
1991).
Multilayer ceramic capacitors (MLCs)
constitute the second largest application of
Figure 17-50. Microstructure of a thick film resistor electronic ceramics, with over 1O10 units
containing Pb 2 Ru 2 0 6 conductive particle agglomer- being manufactured each year. Figure
ates in a lead silicate glass, (a) Low magnification
SEM image showing ring-like structure of conductive
particles (light phase) and (b) high magnification
TEM image showing presence of intergranular glass
phase within the agglomerates. Courtesy Y. L. Chen.

extending w 1 |im away from the particles

into the glass phase. This solubility has a
significant effect on the electrical conduc-
tion.
Aluminum nitride substrates are being
used in specialized applications for their
high heat conduction. A1N ceramics with
thermal conductivities as high as 260
W M " 1 K~ 1 (i.e., approaching the thermal
conductivity of copper!) are now commer-
cially available. The microstructure of a
commercial A1N ceramic is shown in Fig-
ure 17-51. A1N is sintered with the aid of
Figure 17-51. Microstructure of a commercial A1N
additives such as Y 2 O 3 and CaO which substrate material containing a yttrium aluminate
react with A12O3 in the A1N powder to grain boundary phase. (TEM). Courtesy M. R. Notis
form a liquid phase at the firing tempera- and A. Westwood.
 17.5 Electronic and Optical Ceramic Micrestructures 207

17-53 shows the configuration of a typical

multilayer ceramic capacitor device. It
consists of alternating layers of a fine
grained (1-2 jim) ceramic dielectric
(^25 jiim thick) and conducting electrode
(Ag/Pd). The most widely used material
for the ceramic dielectric is barium titanate
modified with various dopants to control
the grain size and dielectric properties. A
small grain size of the ceramic is preferred
since the dielectric constant of BaTiO3 in-
creases with decreasing grain size (Arlt
Figure 17-52. Inversion boundary faults in A1N. An- etal., 1985). The temperature dependence
alytical electron microscopy revealed a high concen- of the dielectric constant is smoothed out
trations of oxygen at the fault. (TEM). Courtesy M.
R. Notis and A. Westwood.
by controlling the dopant distribution
within the grains. Incomplete interdiffu-
sion of the constituent oxides during sin-
tering and grain growth gives rise to a gra-
dation in composition within the grains,
producing a so-called "core-shell" mi-
crostructure (Fig. 17-54). The resulting in-

1
, .f *<.

"f -..

m
• i , >

1 jam

Figure 17-53. Cross section of a multilayer ceramic Figure 17-54. Doped barium titanate capacitor dielec-
capacitor (MLC) showing conducting electrodes (Ag/ tric with a "core-shell" microstructure. The center
Pd) separating dielectric layers of barium titanate. region of each grain exhibits a ferroelectric domain
(SEM, polished section). Courtesy J. Chen. structure which analysis shows to be low in substitu-
ent ions. (TEM). Courtesy M. Mecartney.
208 17 Fired Microstructures and Their Characterization

homogeneity comprises regions with dif- unity and replacing some of the Pb 2 + with
ferent Curie points, and the net effect is a La 3 + to maintain charge neutrality (see
flattened dielectric constant versus temper- Fig. 17-56). A practical problem encoun-
ature characteristic. tered in the processing of relaxor ferroelec-
Relaxor ferroelectrics are emerging as a trics is the formation of unwanted pyro-
technologically important class of materi- chlore phases. Figure 17-57 illustrates the
als for use in a wide variety of electronic microstructure of a PMN ceramic contain-
applications including multilayer capaci- ing a pyrochlore phase with a chemical
tors, piezoelectric transducers, electrostric- composition of Pb 2 (Mg 0 2 5 Nb l i 7 5 )O 6 . 6 2 5.
tive actuators and sensors (Cross, 1987). The pyrochlore grains are easily distin-
A characteristic feature of these materials guished in this case due to their distinct
is that they contain a compensated mixture octahedral morphology (Chen and
of higher ( > + 4) and lower ( < + 4) va- Harmer, 1990). Processing routes have
lence cations on the ocahedral B-site sub- been developed to minimize pyrochlore
lattice of the perovskite-related crystal formation, including precalcining to form
structure (Cross, 1987). Examples of relax- MgNb 2 O 6 and additions of excess MgO or
or ferroelectrics include Pb(Sc 1/2 Ta 1/2 )O 3 excess PbO (Swartz and Shrout, 1982).
(PST), Pb(Mg 1/3 Nb 2/3 )O 3 (PMN) and Finally, we will illustrate the microstruc-
Pb(Fe 2/3 W 1/3 )O 3 (PFW). The electrical tures of several transparent polycrystalline
properties of relaxor ferroelectrics are in- oxides that have been developed for vari-
fluenced to a large degree by the manner in ous optical, infra-red and electro-optic ap-
which the B-site ions are ordered on the plications. The largest application for
B-site sublattice. Depending on the system, transparent A12O3 is high-pressure sodium
ordering can be controlled by heat treat- discharge lamp envelopes, with an estimat-
ment or chemical composition (Harmer ed 3 x 107 units manufactured annually. A
et al., 1989). PST is a classic example of a new application for translucent A12O3 is
system whose ordering can be controlled dental brackets which are more visually
reversibly by heat treatment (Setter and appealing than metal braces. Alumina can
Cross, 1980). PMN, on the other hand, is be sintered to translucency with the aid of
a system in which the ordering lends itself MgO as a solid-solution sintering additive
to control by compositional variation (Bennison and Harmer, 1990). Figure 17-58
(Harmer et al. 1989; Chen et al., 1989). illustrates the fired microstructure of
TEM studies of pure PMN have shown MgO-doped alumina. This material has a
that it contains discrete ordered micro- uniform, single phase, low porosity, high-
chemical domains ^ 2 - 4 n m in size in a transmittance microstructure, in contrast
disordered matrix (see Fig. 17-55), and to the microstructure of undoped A12O3
that the domain size cannot be changed by (Fig. 17-12) which contains abnormal
annealing (Chen et al., 1989). The resis- grains and trapped pores which result in
tance to domain growth has been taken as the A12O3 being opaque. The transparency
evidence for a compositional partitioning generally increases with increasing grain
between the ordered and disordered re- size due to a decrease in light scattering by
gions such that the Mg:Nb ratio in the grain boundaries. Transparent Y 2 O 3 -
ordered region is closer to 1:1. Ordering La 2 O 3 ceramics have been developed for
can be chemically induced in PMN by infrared window applications by Rhodes
adjusting the Mg:Nb ratio closer to using the technique of transient second
 17.5 Electronic and Optical Ceramic Microstructures 209

Figure 17-55. Dark field

micrograph of undoped
Pb(Mg 1/3 Nb 2/3 )O 3 showing
ordered (bright) domains
in a disordered matrix. The
electron diffraction pattern
shows superlattice spots
that correspond to 1:1 or-
dering in the [111] direc-
tion. (TEM). Courtesy J.
Chen.

Figure 17-56. Dark field

micrograph of PbxLax _x
(Mg (1+x)/3 Nb (2 _ x)/3 )O 3 (A)
x = 0.05 (B) x = 0A0 and
(C) x = 0.20 showing exten-
sive ordering and antiphase
boundaries. The superlat-
tice spots indicate a larger
amount of ordering than
for the undoped material.
(TEM). Courtesy J. Chen.

phase sintering (Rhodes, 1981). In this sys- sintering. Transparent lead-lanthanum-

tem, Rhodes used a two phase field to con- zirconium-titanate (PLZT), an important
trol grain growth during the pore-removal electrooptic ceramic, was first fabricated
period of sintering, and then shifted to a by hot pressing by Heartling (1970). It was
single phase field to anneal to transparen- later fabricated by Snow (1973) using an
cy. Figure 17-59 illustrates the pore-free atmospheric sintering technique and by
microstructure of a 0.09 La 2 O 3 0.91 Y 2 O 3 Hardtl (1975) by sinter-HIPing. Excess
ceramic sintered by transient second phase PbO is required in all cases to form a liquid

Figure 17-57. Opposing fracture surfaces

of PMN showing a large pyrochlore grain
with octahedral morphology. (SEM).
Courtesy J. Chen.
210 17 Fired Microstructures and Their Characterization

Figure 17-58. High density microstructure of translu-

cent alumina achieved by doping with MgO and sin-
Figure 17-60. Microstructure of transparent hot-
tering in H 2 . (SEM, polished section). Courtesy W. H.
pressed PLZT ceramic (optical). Courtesy G. H.
Rhodes and G. C. Wei.
Haertling.

0.9 Y 2 O 3
PLZT, which Rhodes (1995) suggests is
due to an increase in scattering from the
ferroelectric domain boundaries. Figure
17-60 illustrates the microstructure of a
transparent PLZT ceramic. For more in-
formation on the development of transpar-
ent polycrystalline oxides the reader is re-
ferred to a comprehensive review article on
this subject by Rhodes (1995).

17.6 Acknowledgements
The authors are grateful for financial
support provided by the Ford Motor Co.,
100
the National Science Foundation (NSF),
DARPA, the Office of Naval Research
Figure 17-59. Pore-free microstructure of transpar-
ent 0.09 La 2 O 3 0.91 Y 2 O 3 achieved by sintering in (ONR), and the Air Force Office of Scien-
the two-phase (2150°C) and annealing in the single tific Research (AFOSR).
phase (1900°C) region of the phase diagram. Cour-
tesy W. H. Rhodes.
17.7 References
phase at the firing temperature (1200- Adar, R, Clarke, D. R. (1982), in: Microbeam Analy-
sis: Heinrich, K. R J. (Ed.). San Francisco: San
1350°C). Extended processing times of Francisco Press, pp. 307-310.
18-60 h were used to volatilize the liquid Anastassakis, E., Pinczuk, A., Burstein, E., Pollak, F.
phase after densification to improve trans- H., Cardona, M. (1970), Solid State Commun. 8,
133.
parency. It has been shown that transpar- Arlt, G., Hennings, D., de With, G. (1985), J. Appl.
ency decreases with increased grain size in Phys. 58, 1619.
 17.7 References 211

Baskaran, S., Halloran, J. W. (1993), J. Am. Ceram. Evans, A. G., Dalgleish, B. X (1986), Ceram. Eng. Sci.
Soc. 76, 2217. Proc. 7, 1073.
Baskaran, S., Halloran, J. W. (1994), /. Am. Ceram. French, X D. (1990), Thesis, Lehigh University, Beth-
Soc. 77, 1249. lehem, PA.
Baskaran, S., Nunn, S. D., Popovic, D., Halloran, J. Gesing, A. X, Burger, G., Luce, E., Claussen, N., Wu,
W. (1993), J. Am. Ceram. Soc. 76, 2209. S., Travitzky, N. A. (1990), Ceram. Eng. Sci. Proc.
Bennison, S. X, Harmer, M. P., (1983), /. Am. Ceram. 11, 821.
Soc. 66, C90. Gifkins, C. (1970), Optical Microscopy of Metals.
Bennison, S. X, Harmer, M. P. (1985 a), J. Am. Cer- New York: Elsevier.
am. Soc. 68, C22. Goldstein, X I., Newbury, D. E., Echlin, P., Joy, D.
Bennison, S. J., Harmer, M. P. (1985b), /. Am. Cer- C , Roming Jr., A. D., Lyman, C. E., Fiori, C ,
am. Soc. 68, 591. Lifshin, E. (1992), Scanning Electron Microscopy
Bennison, S. X, Harmer, M. P. (1990), Ceram. Trans. and X-Ray Microanalysis, 2nd ed. New York:
Am. Ceram. Soc. 7, 13. Plenum Press.
Bergstrom, T. B., Jennings, H. M. (1992), J. Mater. Guiberteau, K, Padture, N. P., Cai, H., Lawn, B. R.
Sci. Lett. 11, 1620-1622. (1993), Phil. Mag. A68, 1003.
Brandon, D. G. (1994), in: Encyclopedia of Advanced Haertling, G. H. (1970), Bull. Am. Ceram. Soc. 49,
Materials: Bloor, D., Brook, R. X, Flemings, 564.
M. C , Mahajan, S. (Eds.). Oxford: Pergamon Haldar, R., Lu, Y. Z., Giessen, B. C. (1987), Appl.
Press. Phys. Lett. 51, 538.
Breslin, M. C. (1993), U.S. Patent 5214011. Hardtl, K. H. (1975), Bull. Am. Ceram. Soc. 54, 201.
Breslin, M. C , Xu, L., Daehn, G. S., Fraser, H. L. Harmer, M. P., Bennison, S. X, Narayen, C. (1983),
(1993), Applications of Continuous Ceramic Com- Mater. Sci. Res. 15, 309.
posite Materials for Automotive Components, SAE Harmer, M. P., Chen, X, Peng, P., Chan, H. M.,
Technical Paper Series No. 930184. Warrendale, Smyth, D. M. (1989), Ferroelectrics 97, 263.
PA: Society of Automotive Engineers. Harmer, M. P., Chan, H. M., Miller, G. A. (1992),
Breslin, M. C , Daehn, G. S., Fraser, H. L. (1994), in: /. Am. Ceram. Soc. 75, 1715.
Low Thermal Expansion Alloys and Composites: Holz, D., Wu, S., Scheppokat, S., Claussen, N.
Stephens, X X, Frear D. R. (Eds.). Warrendale, PA: (1994), J. Am. Ceram. Soc. 77, 2898.
TMS, pp. 185-194. Hooton, X A., Merz, W. X (1955), Phys. Rev. 98, 409.
Brown, P. W, Hellmann, X R., Klimkiewicz, M. Hu, Y.-H., Chan, H. M., Wen, Z. X., Harmer, M. P.
(1993), Microsc. Res. Tech. 25, 474. (1986), J. Am. Ceram. Soc. 69, 594.
Chen, X, Harmer, M. P. (1990), J. Am. Ceram. Soc. Jin, S., Tiefel, T. H., Sherwood, R. C , Davis, M. E.,
73, 68. van Dover, R. B., Kammlott, G. W, Fastnacht, R.
Chen, X, Chan, H. M., Harmer, M. P. (1989), /. Am. A., Keith, H. D. (1988), Appl. Phys. Lett. 52, 74.
Ceram. Soc. 72, 593. Kaysser, W. A., Sprissler, M., Handwerker, C. A.,
Chen, Y. L., Zhang, L., Chan, H. M., Harmer, M. P. Blendell, X E. (1987), /. Am. Ceram. Soc. 70, 339.
(1993), /. Mater. Res. 8, 2128. Kumagai, M., Messing, G. L. (1984), /. Am. Ceram.
Claussen, N. (1984), in: Advances in Ceramics, Soc. 67, C230.
Vol. 12: Science and Technology of Zirconia II: Kumagai, M., Messing, G. L. (1985), /. Am. Ceram.
Claussen, N., Riihle, M., Heuer, A. H. (Eds.). Soc. 68, 500.
Columbus, OH: American Ceramic Society, Lange, D. A., Sujata, K., Jennings, H. M. (1991),
pp. 325-351. Ultramicrosc. 37, 234.
Claussen, N., Urquart, A. W. (1990), in: Encyclopedia Lange, F. F. (1989), /. Am. Ceram. Soc. 72, 3.
of Materials and Engineering, Suppl. Vol. 2: Cahn, Lutz, H. E., Claussen, N. (1991 a), / Eur. Ceram. Soc.
R. W. (Ed.). Oxford: Pergamon Press, pp. 1111- 7, 209.
1115. Lutz, H. E., Claussen, N. (1991 b), /. Eur. Ceram. Soc.
Claussen, N., Wu, S. (1991), Ceram. Trans. 22, 631. 7, 219.
Claussen, N., Le, T., Wu, S. (1989), /. Eur. Ceram. Lutz, H. E., Claussen, N., Swain, M. V. (1991), J. Am.
Soc. 5, 29. Ceram. Soc. 74, 11.
Claussen, N., Travitzky, N. A., Wu, S. (1990), Ceram. Ma, Q., Clarke, D. R. (1993 a), J. Am. Ceram. Soc. 76,
Eng. Sci. Proc. 11, 806. 1433.
Coblenz, W S. (1988), U.S. Patent 4 772 524. Ma, Q., Clarke, D. R. (1993 b), Experiments in Smart
Cross, L. E. (1987), Ferroelectrics 76, 241. Materials and Structures, AMD Vol. 181, ASM
Danilatos, G. D. (1988), Adv. Electron. Electron Symp. Proc. Materials Park, OH: ASM.
Phys. 71, 109. Ma, Q., Pompe, W, French, X D., Clarke, D. R.
Danilatos, G. D. (1991), J. Microsc. 162, 391. (1994), Acta Metall. Mater. 42, 1673.
Danilatos, G. D. (1993), Microsc. Res. Tech. 25, 529. Marshall, D. B. (1986), J. Am. Ceram. Soc. 69, 173.
Evans, A. G. (1990), J. Am. Ceram. Soc. 73, 187. Marshall, D. B., Oliver, W C. (1987), J. Am. Ceram.
Soc. 70, 542.
212 17 Fired Microstructures and Their Characterization

McGinn, P., Chen, W, Zhu, N., Lanagan, M., Bal- Setter, N., Cross, L. E. (1980), J. Appl. Phys. 51,4256.
achandran, U. (1990), Appl. Phys. Lett. 57, 1455. Shaw, N. X, Brook, R. X (1986), J. Am. Ceram. Soc.
McKernan, S. (1993), Proc. 51st Annu. Meeting 69, 107.
MSA: Bailey, G. W, Rieder, C. L. (Eds.). San Snow, G. S. (1973), J. Am. Ceram. Soc. 56, 91.
Francisco: San Francisco Press, pp. 910-911. Soni, K. K., Chabala, X M., Mogilevsky, R., Levi-
Monahan, R. D., Halloran, X W. (1979), J. Am. Cer- Setti, R, Zhang, K., Wolbach, W. S., Bryan, S. R.
am. Soc. 62, 564. (1994), Surf. Interface Anal. 21,117.
Morita, M., Miyamoto, K., Doi, K., Murakami, M., Stearns, L. C , Zhao, X, Harmer, M. P. (1992), J. Eur.
Sawano, K., Matsuda, S. (1990), Physica C 172, Ceram. Soc. 10, 473.
383. Stujits, A. L. (1977), in: Ceramic Microstructures '76:
Morita, M., Takebayashi, S., Tanaka, M., Kimura, Fulrath, R. M., Pask, X A., (Eds.). Boulder, CO:
K., Miyamoto, K., Sawano, K. (1991), in: Proc. Westwood Press.
ISS '90: Advances in Superconductivity III: Kajimu- Swartz, S. L., Shrout, T. R. (1982), Mater. Res. Bull.
ra, K., Hayakawa, H. (Eds.). Berlin: Springer. 17, 1245.
Moulson, A. X, Herbert, X M. (1990), Electroceram- Thomas, G., Goringe, M. X (1979), Transmission
ics. London: Chapman and Hall. Electron Microscopy of Materials. New York: Wi-
Murakami, M. (Ed.) (1992), Melt Processed High ley.
Temperature Superconductors. Singapore: World Wachtman, X B. (1993), Characterization of Materi-
Scientific. als. Stoneham, MA: Butterworth-Heinemann.
Newkirk, M. S., Urquart, A. W., Zwicker, H. R. Wang, Z., Harmer, M. P., Chou, Y. T. (1988), Mater.
(1986),/. Mater. Res. 1, 81. Lett. 7, 224.
Newkirk, M. S., Lesher, H. D., White, D. R., Wang, Z., Harmer, M. P., Chou, Y T. (1992), /. Am.
Kennedy, C. R., Urquart, A. W, Claar, T. D. Ceram. Soc. 75, 1596.
(1987), Ceram. Eng. Sci. Proc. 8, 879. Westwood, A., Notis, M. R. (1991), /. Am. Ceram.
Padture, N., Bennison, S. X, Runyan, X, Rodel, X, Soc. 74, 1226.
Chan, H. M., Lawn, B. R. (1991), Ceram. Trans., Williams, D. B. (1984), Practical Analytical Electron
Vol. 19: Advanced Composite Materials: Nair, K. Microscopy in Material Science. Mahwah, NJ:
M. (Ed.). Westerville, OH: American Ceramic So- Philips Electronic Instruments, Electron Optics
ciety, pp. 715-721. Publishing Group.
Padture, N. P., Bennison, S. X, Chan, H. M. (1993), Williams, D. B., Pelton, A. R., Gronsky, R. (Eds.)
J. Am. Ceram. Soc. 76, 2312. (1991), Images of Materials. New York: Oxford
Quinten, A., Arnold, C. (1989), Mater. Sci. Eng. 122, University Press.
15. Williams, D. B., Soni, K. K., Tseng, M. W, Chabala,
Readey, D. W, Quadir, T., Lee, X (1987), Mat. Sci. X M., Levi-Setti, R. (1993), J. Microsc. 169, 163.
Res. 21, 485. Wu, S., Claussen, N. (1991), J. Am. Ceram. Soc. 74,
Rhodes, W. H. (1981), J. Am. Ceram. Soc. 64, 13. 2460.
Rhodes, W. H. (1995), in: Phase Diagrams in Ad- Wu, S., Claussen, N. (1992), in: Solid State Phenome-
vanced Ceramics: Alper, A. M. (Ed.). San Diego, na, Vol. 25/26: Sintering '91: Chaklader, A. C. D.,
CA: Academic Press, pp. 1-41. Lund, X A. (Eds.). Vancouver, Canada: Sci-Tech
Rodel, X, Glaeser, A. M. (1987), J. Am. Ceram. Soc. Publications, pp. 293-300.
70, C 172. Wu, S., Gesing, A. X, Travitzky, N. A., Claussen, N.
Rodel, X, Glaeser, A. M. (1990), J. Am. Ceram. Soc. (1991), J. Eur. Ceram. Soc. 7, 277.
73, 592. Wu, S., Holz, D., Claussen, N. (1993), /. Am. Ceram.
Rodriguez, M. A., Chen, B. X, Snyder, R. L. (1992), Soc. 76, 970.
Physica C 195, 185. Yurek, G. X, Van der Sande, X B., Wang, W X.,
Rossi, G., Burke, X E. (1973), J. Am. Ceram. Soc. 56, Rudman, D. A., Zhang, Y, Matthiesen, M. M.
654. (1987), Met. Trans. 18A, 1813.
Riihle, M., Evans, A. G., McMeeking, R. M., Yurek, G. X, Van der Sande, X B., Rudman, D. A.,
Charalambides, P. G., Hutchinson, X W. (1987), Chiang, Y.-M. (1988), J. Metals 40, 16.
ActaMetall. 55,2701. Zhao, X, Harmer, M. P. (1988), J. Am. Ceram. Soc.
Sandhage, K. H., Masur, L., Smith, G., Poole, X, 71, 530.
McKimpson, M. (1991), in: High Temperature Su- Zhao, X, Harmer, M. P. (1992), J. Am. Ceram. Soc.
perconducting Compounds HI: Whang, S. H., Das- 75, 830.
Gupta, A., Collings, E. (Eds.). Warrendale, PA:
Minerals, Metals, and Materials Society.
 17.7 References 213

General Reading
Brook, R. J. (Ed.) (1991), Concise Encyclopedia of Lee, W. E., Rainforth, W. M. (1994) Ceramic Mi-
Advanced Ceramic Materials. Oxford: Pergamon crostructures, Property Control by Processing. Lon-
Press. don: Chapman & Hall.
Fulrath, R. M., Pask, J. A. (Eds.) (1977), Ceramic Pask, J. A., Evans, A. G. (Eds.) (1987), Ceramic Mi-
Microstructures 76'. Boulder, CO: Westview Press. crostructures '86\ Materials Science Research,
Kingery, W. D., Bowen, H. K., Uhlmann D. R. (1976), Vol. 21. New York: Plenum Press.
Introduction to Ceramics, 2nd ed. New York: Wi- Rhodes, W. H. (1995), Phase Diagrams in Advanced
ley. Ceramics. San Diego, CA: Academic Press.
18 Finishing

Krishnamoorthy Subramanian

World Grinding Technology Center, Norton Company, Worcester, MA, U.S.A.

List of Symbols and Abbreviations 217

18.1 Introduction 218
18.2 Requirements for Finishing Ceramics 218
18.2.1 Ceramic Grinding Applications 218
18.2.1.1 Electronic Ceramics 218
18.2.1.2 Technical Ceramics 219
18.2.1.3 Traditional Ceramics 219
18.2.1.4 Advanced Ceramics 220
18.2.2 Ceramic Finishing Methods 221
18.2.3 Ceramics Versus Metals 221
18.3 Principles of Finishing Ceramics 224
18.3.1 Interactions at the Grinding Zone - Where Abrasive Product
and Work Materials Interact 224
18.3.2 Effect of Grinding Direction on Ceramic Strength 224
18.3.3 Surface Finish and the Retained Strength 225
18.3.4 Effect of Abrasive Grain Size on Retained Strength 226
18.3.5 Mechanism of Material Removal in the Grinding of Ceramics 226
18.3.5.1 Indentation Fracture Mechanism 228
18.3.5.2 Ductile Regime Grinding Model 228
18.3.5.3 Chip Formation Model for Precision Grinding of Ceramics 229
18.3.6 Systems Approach for Grinding of Ceramics 234
18.4 Practical Aspects of Finishing Ceramics 239
18.4.1 Machine Tool Parameters 239
18.4.1.1 Rigidity/Stiffness 240
18.4.1.2 Vibration Level 240
18.4.1.3 Coolant Systems 240
18.4.1.4 Creep-Feed Grinding 240
18.4.1.5 Precision Movements and Positioning 241
18.4.1.6 On-Machine Dynamic Balancing 241
18.4.1.7 Truing and Dressing Systems 241
18.4.1.8 Multiaxis CNC Capability 243
18.4.1.9 Materials Handling Systems 243
18.4.1.10 New Machine Tools Versus Retrofitting of Current Machines 243
18.4.2 Abrasive Product Selection 243
18.4.2.1 Abrasives Used in the Finishing of Ceramics 243

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. Allrightsreserved.
216 18 Finishing

18.4.2.2 Diamond Grinding Wheels 245

18.4.3 Effect of Workpiece Material Properties 246
18.4.4 Operational Factors 248
18.4.4.1 Chip Thickness 249
18.4.4.2 Specific Energy 249
18.4.4.3 Grindability 250
18.4.4.4 Chippage 251
18.4.4.5 Chatter 251
18.4.4.6 Cracking of Ceramics 251
18.4.5 Output of the Finishing Process 251
18.4.5.1 Production Viable Grinding 251
18.4.5.2 Mirror-Finish Grinding 252
18.4.5.3 Grinding from Simple Solid Shape 252
18.4.5.4 Microgrinding of Ceramics 254
18.5 Economic Aspects of Finishing 254
18.6 Acknowledgements 257
18.7 References 257
 List of Symbols and Abbreviations 217

List of Symbols and Abbreviations

area of new surface

c number of grains per unit area
d depth of cut
equivalent diameter
d% size of abrasive grain
£f energy for fracture
^P energy for plastic deformation
F. force per abrasive grain
Fn normal load
G Griffith crack propagation parameter
h chip thickness
K constant
N number of parts
P power
K surface finish
T time
U specific energy
volume of material under deformation
K wheel speed
work speed
yield stress
CBN cubic boron nitride
CNC computer numerical control
HIP hot isostatic pressing
HPSN hot pressed silicon nitride
L.C labor cost
MOR modulus of resilience
MRR material removal rate
M.T.C machine tool cost
OHC overhead cost
218 18 Finishing

18.1 Introduction powders and lapping slurries reserved for

special applications. Also, green or white-
This chapter covers the final step in the state components are occasionally ma-
production of ceramic parts before they chined using conventional abrasives, car-
are put to use. Ceramics are finished be- bide saws, or single-point cutting tools
fore final use to meet shape, size, finish or made of carbide, ceramic, or diamond.
surface, and quality requirements. The ex- Non-abrasive machining methods uti-
tent of finishing depends upon the applica- lize a variety of means to transfer energy to
tion. the ceramic material surface. These include
Machining is the most common method laser beam machining, electrical discharge
of finishing. It is the application of energy machining, ion-beam machining, electron-
to the part to remove stock and create new beam machining, chemical machining, and
surfaces which define a shape. Such energy electrochemical machining.
transfer is usually accomplished through Abrasive machining methods are covered
the use of abrasives. Machining processes in this section. These are the most common
utilizing abrasives are referred to as abra- processes, used for finishing ceramics. De-
sive machining methods. Methods which tails on non-abrasive finishing methods
result in such energy transfer without the may be obtained from the references, e.g.
use of abrasives are called non-abrasive Engineered Materials Handbook (1991).
machining methods. A word about terminology. Ceramists
Abrasive machining is performed on often use "grinding" to imply comminu-
green (unfired), white (partially fired), and tion or reduction in size. Other terms used
hard (fully dense) ceramic components. to denote this mass finishing process are
Abrasive machining and surface finishing "milling", "crushing", etc. This is not the
methods include grinding, lapping, hon- meaning of the word "grinding" used in
ing, polishing, abrasive fluid jet cutting, this chapter. Instead, grinding here refers
and ultrasonic machining. to a finishing method, using abrasive prod-
Grinding methods used on ceramic com- ucts and a variety of grinding machines
ponents are the most versatile. They em- and processes to generate surfaces of re-
ploy diamond abrasives held fixed in a quired geometry and characteristics on
grinding wheel and applied against the discrete components.
work surface in a variety of configura-
tions. Sometimes these operations are also
termed honing, superfinishing, etc. 18.2 Requirements for Finishing
Abrasives can also be applied to the Ceramics
work surface without being rigidly held in
a grinding wheel. These methods are called 18.2.1 Ceramic Grinding Applications
free-abrasive machining. Some of these
18.2.1.1 Electronic Ceramics
methods are: Lapping, polishing, ultrason-
ic machining, water jet, or abrasive jet fin- Electronic ceramic components are
ishing, etc. computer parts such as microchips, mag-
Diamond is the abrasive most widely netic heads and substrates. They require
used for finishing ceramics. Conventional extensive finishing operations. These oper-
abrasives, however, are generally selected ations are usually carried out to create in-
for lapping operations, with diamond dividual components such as microchips
 18.2 Requirements for Finishing Ceramics 219

from a large array of components made on semiconductor component casings, or-

single wafers. They are also carried out to thodontics, nuclear, biomedical applica-
impart surface features that render the tions, fiber optics, etc. Technical ceramics
components of unique function. For ex- grinding often involves cutting a large part
ample, the reliability and consistency of or tube into subcomponents having mini-
the gap height between the magnetic head mal chippage. Frequently, finishing of flat
and hard disk are achieved by precision surfaces which must be scratch-free or a
profiling of the surfaces of the magnetic mirror finish is also required. Figure 18-1
head. In general, electronic ceramics fin- shows typical arrangements employed in
ishing requires extremely close tolerances the cutting of parts using both single
on part geometry. In addition, residual wheels and gang wheels. A wide range of
stresses in the finished surfaces are critical bond types and grit sizes are used in these
as they affect the electromagnetic proper- applications. The typical products used for
ties of ceramics. Usually, electronic ceram- grinding technical ceramics are cut-off
ics are finished using wheels of very small blades and Blanchard or disk wheels.
thicknesses, in order to reduce the kerf Grinding process optimization, the toler-
width or the material loss during the grind- ances of grinding wheels, and the work
ing. Minimizing the kerf loss helps to in- material features, are among the factors
crease the number of components made that heavily influence the success of these
per wafer. Micrometer-sized diamond grinding operations. Both resinoid and
grains are extensively used in these opera- metal bonded grinding wheels are in use
tions. The grain size uniformity, quality, for these applications. Loose abrasives
and distribution heavily impact the grind- (both conventional and diamond) are also
ing performance. In addition, the consis- extensively used in the finishing of techni-
tency, tolerances, homogeneity, and bal- cal ceramics for lapping and polishing.
ance limits of the grinding wheels also in-
fluence the grinding results. Several spe-
cially built grinding machines are used for 18.2.1.3 Traditional Ceramics
various operations in electronic ceramic Traditional ceramics are low density,
grinding. A variety of conventional abra- porous ceramics used for many applica-
sives, as well as diamond abrasives, are tions where the thermal, electrical or
used in the form of powder, slurry, or com- chemical properties of ceramics are cru-
pounds for the lapping and polishing of cial, often in large or bulk shapes. Typical
electronic ceramic components. Occasion- examples of traditional ceramics are: re-
ally very small surface modifications are fractories, furnace liners, electrical parts,
carried out using electron beam machin- coated ceramics, etc. The grinding opera-
ing, under special circumstances. tions for traditional ceramics vary widely
in shape and configuration. Metal bonded
grinding wheels are mostly employed for
18.2.1.2 Technical Ceramics these operations to cut off parts or to gen-
Technical ceramics are components used erate large flat surfaces. Metal bonded and
to exploit a variety of ceramic properties electroplated diamond wheels are exten-
such as electrical, thermal properties, cor- sively used for these applications to
rosion resistance, etc. These applications achieve long durability for wheels. The
include quartz tubing used for lighting, chippage criteria and tolerances for these
220 18 Finishing

Sliding table, angle, or nested support Swing arm Vertical cut

(a) (b)

Single or multiple parts

n n
Clamped on both
sides of cut

(c) (d)

(e) Parts totally fixed

Figure 18-1. Schematic representation of single and multiple or gang wheel cutting of ceramic parts (Dudley,
1990): (a) single wheel cutting (moving parts); (b) single wheel (fixed parts); (c) single wheel (fixed parts); (d) multiple
wheels/reciprocating table; (e) gang wheels/multiply fixtured.

applications are generally rather broad rel- 18.2.1.4 Advanced Ceramics

ative to other applications. Hence, they
can tolerate the higher grinding power and Advanced ceramics are high-strength,
forces generated by the metal bonded high-density (low-porosity) ceramics for
wheels or the poor finish and damage pro- mechanical or structural applications.
duced by the electroplated wheels. Howev- These materials are the most difficult to
er, for applications requiring tighter toler- grind owing to their high hardness and
ances, the choice of metal bond type and strength. Figure 18-2 shows a range of ad-
its properties, as well as the dressing prac- vanced ceramic parts after grinding. Typi-
tices used, will require careful consider- cal requirements for these components are
ation. high retained strength after grinding and
 18.2 Requirements for Finishing Ceramics 221

Figure 18-2. Selected advanced ceramic

workpieces finished by production grind-
ing techniques.

also production-viable grinding methods to closer flatness and parallelism than that
(that is, short cycle time, an economic possible for metals. Ceramics are generally
grinding process, as well consistent part more chemically stable than metals. Hence
quality). the burn sometimes observed on metals
during grinding is rarely observed on ce-
18.2.2 Ceramic Finishing Methods ramic materials. Any large scale thermal
softening that aids in the grinding of
The range of abrasive finishing opera-
metals can rarely be counted on for the
tions required for advanced ceramics par-
grinding of ceramics. Thermal conductivi-
allel the finishing operations used for met-
ty of ceramic materials varies widely. It is
al and carbide components (Subramanian,
more difficult to grind poorly conducting
1987 a). Figure 18-3 a shows the typical
ceramics and the problem becomes worse
range of diamond grinding processes avail-
when they are also poor in thermal shock
able for advanced ceramic component
resistance.
grinding. Figure 18-3b presents the range
One characteristic that significantly dis-
of ultraprecision finishing processes used.
tinguishes ceramics from metals is their
low fracture toughness. While recognizing
18.2.3 Ceramics Versus Metals
the role of all the above properties it would
Ceramics include a wide range of mate- appear that it is possible to achieve suc-
rials, with thermophysical properties de- cessful grinding of ceramics if the genera-
pendent on the type, composition, mi- tion and propagation of cracks during the
crostructure, and the processing methods. grinding process can be minimized.
The strength of ceramic material varies Methodology and principles pertinent to
widely, depending on the material chosen. such successful grinding of ceramics are
Even for a given material like silicon ni- outlined in later sections. In the absence of
tride for example, the strength depends on such a systematic approach, severe dam-
the sintering aids used and the sintering age can be caused during the finishing pro-
methods applied (pressureless sintering, cess. Alternatively, the finishing process
hot pressing, hot isostatic pressing (HIP), may be carried out at low production effi-
etc.). In general, ceramic materials have ciency (long cycle time and poor yield) be-
higher stiffnesses (Young's modulus) than cause of the potential for damaging the
metals. As a result, they can be machined ceramics. On the other hand, through the
 222 18 Finishing

Diamond

7 Workplace
(a) Horizontal-spindle (b) Vertical-spindle (c) Creep-feed grinding
surface grinding surface grinding

Workpiece Diamond wheel Workpiece

Diamond wheel

Regulating
Workpiece Guide blade wheel

(e) Internal grinding (f) Centeriess grinding

(d) Outside diameter with a chucking machine
cylindrical grinding

wheel
-a-y--•-£
Diamond . _ Diamond
wheel
(left) (right)

Work-
(g) Inside diameter
piece
form grinding (h) Jig grinding
(I) Double-disk grinding

Diamond .
wheel ttrV Workpiece

- • ~ ^

(j) Thread grinding (k) Outside diameter (I) Slot grinding

form grinding

Single or multiple parts

(m) Multiple wheels/reciprocating table (O) Slicing

(A) (p) Flatting (q) Cropping (r) Dicing

 18.2 Requirements for Finishing Ceramics 223

(c)
(a) HONING
FORCE OR PRESSURE ON EXPANDING
SINGLE SIDED LAPPING
ARBOR OR MANDREL

STROKE
LENGTH
OF HONING
TOOL
HONING STONE
OR STICK

(b)
MICROFINISHING
SUPERFINISHING STONE
OR STICK
WORK PIECE ROTATION
LAPPING SLURRY

(d)
STROKE LENGTH^
OF SUPERFINISHING TOOL
(e)
CYLINDRICAL PARTS LAPPING

(g)
ABRASIVE POLISHING
ABRASIVE PAD OR
FILM ADHERED TO
W O R K >. PRESSURE THE POLISHING WHEEL

Figure 18-3. (A) Typical selection of precision grinding processes for finishing of ceramics. (B) Schematic repre-
sentation of ultraprecision finishing processes, (a) Honing, (b) microfinishing, (c) single sided lapping, (d) double
sided lapping, (e) cylindrical parts lapping, (f) flat honing and (g) polishing.
224 18 Finishing

use of suitable grinding principles, meth- material. The influence of coolants should
ods and systems, ceramics can at times be also be included in these interactions. The
finished to closer tolerances, better surface grinding process differs from the cutting
finish and more complex geometry than process in two key areas: Chip-bond inter-
their metallic counterparts, while meeting action and the bond-work interaction are
the necessary economic considerations. absent in single or multiple-point cutting
processes such as turning or milling; the
cutting geometry of the abrasive constant-
18.3 Principles of Finishing ly changes in the grinding processes. An
efficient grinding process attempts to max-
Ceramics imize the abrasive/work interaction, and
to minimize the other three frictional inter-
18.3.1 Interactions at the Grinding Zone -
actions (chip/bond, chip/work, and bond/
Where Abrasive Product
work interfaces). The ability to preferen-
and Work Materials Interact
tially control the abrasive/work interac-
The interactions at the grinding zone tion depends to a large degree on our un-
may be characterized as shown in Fig. 18-4. derstanding of the principles of grinding
As the grinding wheel is applied to the ceramics.
work material at a given wheel speed, work In metallic materials this balance be-
speed, and depth of cut, there are generally tween the "cutting action or chip forma-
four types of surface interactions taking tion" and the "sliding interactions" is pri-
place at the grinding zone. The abrasive marily to achieve desired surface quality,
grain interferes with the work material un- while maintaining the desired production
der stress, strain and strain rate conditions rate. However, in the case of ceramics,
dependent on the grinding process vari- these interactions have to be balanced first
ables, abrasive geometry, and work mate- in terms of minimizing residual damage
rial properties. Simultaneously, there are such as surface or subsurface cracks, which
three frictional interactions due to the rub- lead to chippage, gross fracture or degra-
bing of chips produced, against the bond dation in strength. Once this is accom-
matrix of the wheel, bond matrix in the plished, the balancing act between "chip
wheel against the work material and chips formation" and "friction" can be pursued
(entrained in the wheel) against the work in a manner very similar to the principles
of grinding metallic components.

18.3.2 Effect of Grinding Direction

on Ceramic Strength
Figure 18-5 shows the strength (modu-
lus of resilience, MOR) of a hot pressed
silicon nitride (HPSN) after it has been
precision ground. It is observed that the
grinding direction has a significant influ-
Figure 18-4. Interactions in the grinding zone: (1) ab-
rasive/workpiece interface; (2) chip/bond interface; ence on the strength of the material (Ohta
(3) chip/workpiece interface; (4) bond/workpiece in- and Miyahara, 1990). As the abrasive
terface. grain size decreases, this anisotropy ob-
 18.3 Principles of Finishing Ceramics 225

Grain size, in.

0.001 0.002 0.003 0.004 0.005 0.006
800 116

700 102

>- -
*-- —.
600 87
" f' (/
M L.ongitudinal

i 50°
No 6(30 -
73 B
\
No.400 \
\
400 58
*
^ ^ - . __
< y///////////)
• r i/
No. 200
Transverse
300 44
No 120

200 29
25 50 75 100 125 150
Grain size, jim

Figure 18-5. Influence of grinding direction and abrasive grit size on the strength of ground ceramic material
(Ohtaetal., 1987).

served in the strength gradually diminish- Miyahara, 1990). However, improvement

es. Thus, it appears beneficial to grind with in surface finish alone is not adequate to
finer abrasive particles in order to obtain improve the strength. There appears to be
high retained strength as well as strength no improvement in strength withfinersur-
which is independent of the grinding direc- face finish if coarser abrasive grits are
tion. Later studies have shown that this used. This would suggest that the damage
beneficial effect of abrasive grain size is caused in ceramic materials may not be
related to the size of the "chip thickness" removed simply by further burnishing of
(Subramanian and Ramanath, 1992; the work surface to achieve better surface
Mayer and Fang, 1994). This will be dis- finish. The finer abrasive grits, on the oth-
cussed further in a later section. er hand, may not cause the damage to be-
gin with, thus assuring high strength. The
18.3.3 Surface Finish and the Retained damage referred to here is predominantly
Strength crack generation and propagation through
Figure 18-6 shows that the strength im- brittle fracture. Other references (Ander-
provement with fine abrasive particles is son and Bratton, 1979) also refer to the
also associated with an improvement in relationship between surface finish and re-
surface finish (Ohta et al., 1987; Ohta and tained strength of ceramics.
226 18 Finishing

Maximum peak-to-valley roughness height (Rmax), \x\n.

40 80 120 160 200 240
700 102

Grain Lay
A
size symbol
600 87
O 400 /
• 200 J-
A 800 1
oo
500
o &*
o

ex o
400 58
t
\ \
\
300 44

200 29
0 1 2 3 4 5 6
Maximum peak-to-valley roughness height (Rmax), |im

Figure 18-6. Strength versus surface finish of ground ceramic material (Ohta et al., 1987).

18.3.4 Effect of Abrasive Grain Size grit size decreases, the specific energy in-
on Retained Strength creases rather dramatically.
Figures 18-5 and 18-6 indicate that the
18.3.5 Mechanism of Material Removal
finer the abrasive grain size, the better the
in the Grinding of Ceramics
strength of the ground ceramic and the
better the surface finish. Table 18-1 shows Recognizing the low fracture toughness
the variation of surface finish as a function of ceramics, a variety of models have been
of grit size and other grinding parameters. proposed which describe the initiation and
Tests have shown that surface finish for propagation of microcracks during the
alumina ceramic is generally unchanged by grinding process. These models are gener-
the depth of cut or the table speed when ally described as "indentation fracture
coarse abrasive grain is used. However, the model". A second variety of model has
finish improves gradually as finer grit size also been proposed based on the occurence
is used. Figure 18-7 shows the specific en- of ductile deformation under extremely lo-
ergy (energy required per unit volume of calized conditions. This second type of
material removed) required in the grinding model is described as "ductile regime
of Sialon material as a function of the grinding" and assumes occurrence of plas-
abrasive grit size. It is observed that as the tic deformation only and seeks to achieve
 18.3 Principles of Finishing Ceramics 227

Grit size, in.

0.0008 0.0016 0.0024 0.0032 0.0040 0.0048
100

90 Depth of cut Unit-width material removal rate

|im fiin. mm 3 /s, mm in. 3 /min, in.
80 O 2.5 100 0.017 0.0016
• 5.0 200 0.034 0.0032
A 10.0 400 0.068 0.0064
15 r
70 30 0 ' ^* fi one n n-iao

60

50
{
40

30
\
20

10

0 10
!P
t—.

20 30
F ^ ^

40
——,_«
50
^

60 70
t
80 90
<

100 110 120

Grit size, |im

Figure 18-7. Specific energy versus abrasive grit size of ground SiAlON ceramic as a function of depth of cut and
unit width material removal rate (computed values adapted from Ichida, 1986).

Table 18-1. Surface finish (roughness average, Ra) as a function of grinding parameters.

Coolant Wheel Table Surface finish, Ra, at selected downfeeds

speed
0.0025 mm 0.0050 mm 0.013 mm 0.025 mm 0.050 mm
downfeed downfeed downfeed downfeed downfeed
(mm/min) (jun) (jam) (um) (urn) (um)

100% Oil D150-N100 355 _ 1.3 1.08 1.00 -

710 - 0.95 1.03 1.08 0.95
1270 - 1.3 1.03 1.03 1.00
2.5% Oil in SD320-R100B95 1145 - 0.86 - - -
water SD320-R100B95
(after spark out) 1145 - 0.81 - - -
SD20/30JU-R100B69 1145 - 0.45 - - -
SD20/30/i-R100B69
(after spark out) 1145 - 0.40 - - -
SD4/8AI-R100B69 1145 0.24 - -
SD4/8/I-R100B69
(after spark out) 1145 0.18 - - — —
228 18 Finishing

this condition, usually at extremely low material. Figure 18-9 shows the normal
material removal rates. We shall describe forces per abrasive grain as a function of
these two models in the following and then abrasive grain size. It is observed that the
we will utilize other evidence which sug- force per grain decreases significantly with
gests that both these models may operate grit size. This could account for the de-
simultaneously. This leads to a unified crease in the size of the median cracks, thus
"chip formation model" for grinding of leading to the higher strength of the ceram-
ceramics. ic material. This model seems to simplify a
rather complicated problem.
18.3.5.1 Indentation Fracture Mechanism However, there are several limitations to
this model (Conway and Kirchner, 1988;
In this model, it is assumed that the
Malkin and Ritter, 1988). For instance,
abrasive grain acts like an indenter, which
while the force per grain decreases with
under a normal load (Fn) initiates a large
abrasive grain size, when this force is nor-
median crack. Figure 18-8 depicts lateral
malized with respect to the volume of ma-
or vent cracks, which when propagated
terial removed per grain (specific force),
back to the surface, remove or lift a piece
there appears to be a disproportionate in-
of material off the work surface. During
crease. This is not consistent with a mech-
such a brittle fracture-dominated process,
anism purely governed by brittle fracture.
the surface finish obtained may be per-
Similarly, the grinding energy (specific en-
ceived to be independent of grinding pro-
ergy) required per unit volume of material
cess parameters (Lawn etal., 1980; Con-
removed also appears to increase signifi-
way, Jr. and Kirchner, 1980; Rice and Me-
cantly as the abrasive particle size used or
cholsky, 1976; Inasaki, 1987; Spur et al.,
chip volume removed per abrasive grain
1985).
decreases as noted earlier in Fig. 18-7.
This model serves to explain the benefi- These increases in specific force and
cial effect of fine abrasive particles to en- specific energy with the decrease in grit
hance the strength of the ground ceramic (grain) size, depth of cut, and material re-
moval rate are remarkable and more than
required for a purely brittle fracture pro-
cess (that is, the force is higher than that
which would cause lateral cracks).

Grinding groove 18.3.5.2 Ductile Regime Grinding Model

In an effort to explain the higher grind-
Radial crack ing specific forces and specific energy asso-
ciated with a smaller depth of cut
(Miyashita, 1985), the mechanism of mate-
rial removal has been designated as the
/ Lateral crack "ductile regime grinding". In this model if
the abrasive grain depth of cut is precisely
Median vent crack controlled, it is believed that under certain
Figure 18-8. Schematic representation of indentation "critical depth of cut", the material re-
fracture model of ceramic materials. moval mechanism is purely due to ductile
 18.3 Principles of Finishing Ceramics 229

Average grit size, in.

0.0008 0.0016 0.0024 0.0032 0.0040
0.045 0.010

0.04
/ - 0.009
/
/ - 0.008
0.035
/
- 0.007
0.03 /

-- 0.006
#t? 0.025
5 - 0.005 (5
/
0.02

/ -- 0.004
o
0.015 /
/ -— 0.003

0.01
/
/ - 0.002

c\ /
0.005 - 0.001

/
0 10 20 30 40 50 60 70 80 90 100 110
Average grit size, pm

Figure 18-9. Force/grit versus average abrasive grit size, while grinding SiAlON ceramic material; wheel: resin
bonded grinding wheel, 150 concentration.

deformation. This is a good approach be- 18.3.5.3 Chip Formation Model

cause it recognizes the possibility of ductile for Precision Grinding of Ceramics
deformation in the grinding of ceramic
materials (Bifano, 1991). It is possible to treat hard ceramic mate-
The limitation of this approach is the rials like any other material subjected to
primary dependence on the depth of cut as machining or grinding processes. Under
the controlling factor for plastic deforma- such processes, shear deformation at the
tion during grinding. Also a discrete cutting or grinding zone produces a chip
change between ductile deformation and (Fig. 18-10). A sequential removal of chips
brittle fracture may not be possible in a leads to the generation of the machined
multi-edge cutting process such as grind- surface. The nature of the chips produced
ing. Even when the depth of cut is precisely (that is, continuous or discontinuous) with
controlled coarse abrasive grains could varying degrees of plastic deformation will
leave a pattern of microcracks on an other- depend on the cutting tool geometry,
wise plastically deformed surface due to depth of cut, cutting velocity, work materi-
their dull cutting action and excessive fric- al properties, and so on. This is the classi-
tion during grinding. cal cutting model (Ernst and Merchant,
230 18 Finishing

creased probability for the presence of a

large number of defects or brittle fracture
sites in the volume removed. Conversely,
dense ceramics with low flaw population
are likely to resist brittle fracture, particu-
larly as the size of the chip decreases. In the
absence of brittle fracture, these hard ce-
ramics are likely to offer large resistance to
deformation, possibly accounting for the
higher specific grinding forces and the
specific energy required as the grit size or
(a) depth of cut decreases.
Superabrasive In addition to the material properties of
the workpiece, the degree of plastic defor-
mation in a cutting or chip formation
model will also depend on the geometry of
the cutting tool, abrasive grain size, and
the depth of cut per grain or chip thick-
ness. There are several literature references
that discuss potential localized deforma-
tion in small sections of brittle materials
caused by small indenter forces and depen-
(b)
dent on the indenter geometry. It has been
Figure 18-10. Scheme of the cutting or chip forma- suggested that it is impossible to fracture
tion model chip generated by (a) single point cutting brittle materials at small sizes because they
tool; (b) abrasive grit; /?, clearance angle; a, rank angle.
cannot store enough elastic energy to pro-
pel a crack through the particle before
1941), and has been successfully used in plastic flow takes place (Feng and Field,
metal grinding (Lindsay, 1989; Hahn and 1989). The existence of a critical depth and
King, 1986; Subramanian and Lindsay, a critical load per grain, below which plas-
1989; Malkin, 1984). tic deformation occurs in preference to
Such a model would suggest that materi- elastic loading and brittle fracture, has also
als of high strength and fracture tough- been observed (Nakajima et al., 1989).
ness, such as ZrO 2 and hot pressed silicon The models of localized deformations
nitride, would exhibit greater plastic defor- due to indentation are shown in Fig. 18-11.
mation during grinding. On the other Under lightly loaded conditions and a
hand, materials of low strength and low small radius of the indenter, plastic defor-
fracture toughness (for example, ferrite), mation would be promoted in the brittle
could produce a large degree of discontin- material subjected to the indentation force.
uous brittle fractured chips. It would also Under a large radius of the indenter, the
be reasonable to state that grinding condi- so-called cone crack is initiated. Thus,
tions, removing a large volume of material even in situations where ductile regime
per grain (such as with a large depth of cut grinding (see the section on "Ductile
and coarse abrasive grains), are likely to Regime Grinding Model" in this chapter)
result in brittle fracture due to the in- may be promoted by a precisely trued
 18.3 Principles of Finishing Ceramics 231

Brittle materials Ductile materials

Large
y
radius
of tip

Cone crack Flaws Plastic deformation

Low pressure Low pressure

P
I /

iiiiiiiWiniiii - # -
Plastic deformation Plastic deformation
Small
radius
of tip High High pressure
1P pressure

11 mj2JjUaj.mil i
Median ' Lateral Plastic deformation
crack crack

Figure 18-11. Localized deformation and fracture generated by an indenter in both brittle and ductile materials
under varying conditions of indenter radii, r, and pressure, P (Nakajima et al., 1989).

wheel, using an ultrasmall depth of grind- Fig. 18-12. The resultant surface on the ce-
ing, it is possible that such deformation ramic material also shows evidence of duc-
may be accompanied by surface cracks tile deformation, which improves with the
caused by the large apparent radius of the fracture toughness of the material being
indenter (abrasive grain). However, it may worked as well as with a decrease in the
be possible to obtain plastically deformed abrasive grain used (Fig. 18-13).
surfaces without initiation of surface The energy required to plastically de-
cracks in the high-density high-strength ce- form a specified volume (F p ), may be writ-
ramics if a small indenter radius and a ten as
small force per grit particle are applied
Ep = oyVp (18-1)
during grinding.
Evidence of chips generated during where ay is the yield stress.
grinding of ceramics with morphology The material property characterizing re-
similar to ductile deformed chips in sistance to fracture is the Griffith crack
addition to the brittle fractured chips is propagation parameter G. The energy re-
reported in the literature (Subramanian quired for fracture is a function of the area
and Ramanath, 1992), and are shown in (A{) of new surface generated by crack
232 18 Finishing

(B)

(E)

Figure 18-12. Morphology of chips generated in grinding of ceramics (A) and (B): ferrite; (C), (D) and (E): alumina
ceramic; (F): zirconia. All these chips were observed during surface grinding using diamond abrasive wheels, with
average grain size of 91 um.
 18.3 Principles of Finishing Ceramics 233

(a)

(b)

Figure 18-13. Surfaces generated by grinding four different ceramics: zirconia, silicon nitride, alumina, and ferrite,
with decreasing fracture toughness, (a) Surface generated using coarse abrasive grains (91 um); (b) surfaces
generated using fine abrasive grains (6 um).
234 18 Finishing

propagation. Thus be considered to be influenced by two

probabilistic events (plastic deformation
Ei = G Af (18-2)
and brittle fracture). The end result may be
In his analysis Bifano (1991) has ap- a cumulative result of these two probabilis-
proximated both Vp and Af as functions of tic events and their relative frequency of
depth of cut d. However, in grinding pro- occurrence. This model is schematically
cesses, while d represents the total depth of shown in Figs. 18-14a, b.
cut, h represents the chip thickness or It is evident that while we refer to aver-
depth of cut per grain, as shown in Fig. 18- age values of the three process parameters
10. In addition, as described in Fig. 18-11, (force per grain, radius of the grain, and
the volume deformed under plastic defor- chip thickness), they should also be main-
mation is influenced by the radius of the tained over a rather small range. Otherwise
indenter (approximately equal to half the the output of the grinding process will be
size of the abrasive grain dg), and also the influenced by extreme conditions.
force applied per abrasive grain, Fg.
The chip thickness h in grinding may be
18.3.6 Systems Approach for Grinding
expressed as
of Ceramics
K\li2
1/2
h= (18-3) From the details observed thus far, it is
Dj \kc(dg)J clear that abrasive machining processes
where V^ work speed, Vs wheel speed, d will be influenced by the abrasive product
depth of cut, De equivalent diameter, K a used, machine tool involved, work materi-
constant, generally a function of abrasive al and/or operational variables. All these
type and its geometry, and c number of four input categories interact with each
grains in the grinding zone per unit area. other, which culminates in the output of
c =f(dg), where dg is the abrasive grain size. abrasive machining process results. This is
It is also a function of the abrasive volume described in Fig. 18-15 a (Subramanian,
contained in the grinding wheel. Hence, 1987 b). However, each of these input
groupings consists of a multitude of vari-
ig,Fg) ( 18 " 4 ) ables, as described in Fig. 18-16. When
dg,Fg) (18-5) perceived in this manner, the abrasive ma-
chining processes appear to be complex
The ratio of material removal energies is and managed only as a black art. Howev-
then given by er, this is not the case.
plastic flow energy / (/i3, dg, Fg) Irrespective of the choice of variables in
fracture energy ~ / (h2, dg, Fg) the four input categories, for every abra-
sive machining process it is possible to vi-
g,Fg) (18-6)
sualize the four interactions described ear-
Hence each incidence of abrasive/work lier (Fig. 18-4) between the abrasive prod-
interaction may be energetically favored uct and the work material. Of these, the
for plastic deformation or for brittle frac- abrasive/work interaction is the most criti-
ture, dependent on the chip thickness, cal, which in many respects is analogous to
force per grain and radius of the abrasive machining processes with cutting tools.
grain pertinent to that incidence. Hence Therefore the interactions between the
grinding of ceramics on the average may abrasive product and the work material
 18.3 Principles of Finishing Ceramics 235

(a)

INDENTER RADIUS

CHIP THICKNESS

FORCE PER GRAIN

GRINDING OF CERAMICS
(b) (GOVERNING MECHANISMS)

Figure 18-14. (a) Schematic

25 representation of chip for-
DC <

UJ LU
mation model combined
with the treatment abrasive
grit as an indenter.
(b) Governing mechanisms
and their key controlling
CHIP THICKNESS = V/((Vw/Vs) (d.o.c./ De) X 1/(KxC) variables in the grinding of
ceramics (combining chip
formation model with the
treatment of abrasive grain
FORCE PER GRAIN
in an indenter).
ABRASIVE GRAIN RADIUS

may be grouped into cutting (material re- measured or monitored using macroscopic
moval process), plowing (material dis- process variables such as forces, power,
placement process) and sliding (surface temperature, etc. These in turn result in
modification process). Hence every abra- certain technical outputs, which, when
sive machining process is an effort to bal- viewed on the basis of the rules of manu-
ance between cutting (surface generation) facturing economics, result in economic or
and plowing/sliding (which controls the system output. This input/output repre-
characteristics of the generated surface). sentation - called systems approach - is
Thus every abrasive machining process shown in Fig. 18-15 b (Subramanian et al.,
may be thought of as an input/output pro- 1994).
cess with defined microscopic interactions The above system description of the
of cutting and tribological aspects of plow- abrasive machining processes greatly sim-
ing and sliding. These interactions can be plifies our understanding and use of the
 18 Finishing

OPERATIONAL
FACTORS

GRINDING RESULTS
SURFACE QUALITY

TOLERANCES/FINISH

PRODUCTION RATE

PRODUCTION ECONOMICS

NEW PROCESS/PRODUCT

WHEEL SELECTION WORK MATERIAL

(b)
Machine Tool
Abrasive Product
Input Work Material
Operational Factors

Grinding Forces Cutting

Process and ^ — Plowing
Energy Rubbing/Friction
I

I (Macroscopic) (Microscopic)

Surface Generation Part Quality Figure 18-15. (a) Schematic

Wheel Wear Production Economics representation of the pro-
Output Chips ~*~ New Products/Processes
Coolant Interactions Surface Integrity/performanc
duction grinding system,
Residual Stress (b) A systems approach for
(Technical) (Economic) abrasive finishing processes.
 18.3 Principles of Finishing Ceramics 237

MACHINE TOOL FACTORS WORK MATERIAL FACTORS

DESIGN PROPERTIES
-RIGIDITY -MECHANICAL
-PRECISION -THERMAL
-DYNAMIC STABILITY -CHEMICAL
FEATURES -ABRASION RESISTANCE
-MICROSTRUCTURE
-CONTROLS
-HORSE POWER
GEOMETRY
-SPINDLE SPEED
-SLIDE MOVEMENTS/AXES -WHEEL - PART CONFORMITY
-POSITIONING ACCURACY -ACCESS FOR COOLANT
-REPEATABILITY -SHAPE/FORM REQUIRED
-THERMAL STABILITY
-TRUING AND DRESSING PART QUALITY
EQUIPMENT -GEOMETRY
-TOLERANCES
COOLANT SYSTEM -CONSISTENCY
-TYPE, PRESSURE, FLOW -SURFACE CHARACTERISTICS
-NOZZLE ARRANGEMENTS
-FILTRATION SYSTEM

GRINDING RESULTS
-SURFACE QUALITY
-RETAINED STRENGTH
-PRODUCTION RATE
J
ABRASIVE
rl
PRODUCT
-COST PER PART
-COMPONENT PERFORMANCE

OPERATIONAL FACTORS

ABRASIVE
FIXTURES AND WORK
-TYPE
HOLDING
-PROPERTIES
=PARTICLE SIZE AND SHAPE WHEEL BALANCING
-SIZE DISTRIBUTION TRUING, DRESSING
-CONTENT/CONCENTRATION AND CONDITIONING
BOND -DEVICES
-TECHNIQUES
-TYPE
-PARAMETERS
-HARDNESS/GRADE
-STIFFNESS GRINDING CYCLE DESIGN
-POROSITY WHEEL/WORK CONFORMITY
-THERMAL CONDUCTION
CHIP THICKNESS
WHEEL DESIGN INSPECTION METHODS
-SIZE/SHAPE
-CORE MATERIAL
-FORM OR PROFILE

Figure 18-16. Selected variables influencing the abrasive machining system (selected examples only).

principles of machining and tribology to governing mechanisms and key control

manage and/or improve abrasive machin- variables in the finishing of ceramics as
ing processes. We can now incorporate our shown in Fig. 18-17 (Subramanian et al.,
earlier observations via the mechanisms of 1994). This figure represents key micro-
material removal in precision finishing of scopic interactions (brittle versus ductile
ceramics into the system approach. This material removal processes) and key tech-
results in a schematic representation of nical inputs governing them, based only on
238 18 Finishing

PLASTIC
DEFORMATION BFKITTLE ^ /
MATERIAL FFWCTURE /
FINE OR DUCTILE
S^/ REGIME GRINDING

o
^ 1
ac z
(X
«_->
<
LJ
. . — o
o 2:
LU _»J
AIIV
OVA

LU LU
(X. CC
^" / *~
\ ' COARSE
GRINDING
FIINISH GRINDING (PRESENT)
GRINDING
(PROPOSED)

INPUT
M/C TOOL
LOW VIBRATION LARGE
PROCESSING TOOL
SMALL GRAIN SIZE LARGE
SMALL GRAIN TIP RADIUS LARGE
WORK MATERIAL

SMALL GRAIN SIZE LARGE

HIGH FRACTURE LOW
TOUGHNESS
OPERATIONAL FACTORS
LARGE CUTTING VELOCITY SMALL
SMALL DEPTH OF CUT LARGE
SMALL CHIP THICKNESS LARGE

PROCESS VARIABLES

MICROSCOPIC

HIGH "CUTTING" PROCESS THROUGH LOW

PLASTIC DEFORMATION

MACROSCOPIC

SMALL FORCE PER GRAIN LARGE

LARGE SPECIFIC ENERGY SMALL
LARGE SPECIFIC GRINDING FORCE SMALL

OUTPUT
HIGH PART STRENGTH LOW

FINE SURFACE FINISH POOR

GOOD EDGE CHIPPAGE POOR

LOW MATERIAL REMOVAL HIGH

RATE

Figure 18-17. Schematic representation of abrasive/workpiece interactions in the ceramic grinding system and
their governing mechanisms.
 18.4 Practical Aspects of Finishing Ceramics 239

(INPUT)
MACHINE TOOL
MULTI AXIS CNC GRINDER WITH
HIGH RIGIDITY AND HP
PROCESSING TOOL
FINE GRIT DIAMOND GRINDING WHEEL

WORK MATERIAL
SIMPLE SHAPED CERAMIC BLANK
GOOD HOMOGENEITY AND MINIMUM
OF DISTORTION
OPERATIONAL FACTORS
GRINDING CYCLE DESIGN
CHIP THICKNESS
PART HOLDING/FIXTURING
CREEP FEED GRINDING
ON-MACHINE TRUING, DRESSING
AND BALANCING
COOLANT SYSTEM AND FILTRATION

(MACROSCOPIC)
DECREASE THE FORCE (MICROSCOPIC)
PER GRAIN MAXIMISE THE "CUTTING" PROCESS
DECREASE SPECIFIC ENERGY (THROUGH PLASTIC DEFORMATION
AND SPECIFIC FORCE VS. BRITTLE FRACTURE)
DECREASE FORCE VARIATIONS MINIMISE THE FRICTIONAL
DECREASE TOTAL FORCE INTERACTIONS
AND POWER

(PROCESS OUTPUT) (SYSTEM OUTPUT)

HIGH MATERIAL REMOVAL RATE COMPONENT RELIABILITY
COMPLEX PART GEOMETRY » HIGH YIELD Figure 18-18. Systems de-
GOOD SURFACE FINISH LOW TOTAL COST PER PART scription for the produc-
MAXIMUM RETAINED STRENGTH EARLY COMMERCIALISATION
tion grinding of ceramics.

abrasive/workpiece interactions. In a total cussed in a later section. This will require

system consideration for abrasive finishing additional considerations in each of the in-
processes, we must accomplish the desired put groupings. When all of these are taken
abrasive/workpiece interaction with the into account, a system description of the
minimum of tribological interactions, i.e., production grinding of ceramics, may be
bond/workpiece, chip/bond, chip/work- represented as shown in Fig. 18-18 (Subra-
piece interactions. In addition, any point manian etal., 1994). We shall discuss
on the X axis of Fig. 18-17 represents an specific aspects of this system in further
average of abrasive/workpiece interac- detail in the next section.
tions, from a range of results based on a
range of probabilistic events occurring
over a period of time and throughout the 18.4 Practical Aspects of Finishing
grinding zone. In practice, however, this
range should be held to a minimum, which
Ceramics
is influenced by uniformity of abrasive
18.4.1 Machine Tool Parameters
grain size, shape and their distribution in
the abrasive product, for example. Fur- The following machine tool parameters
ther, the technical output of the system has and operational factors predominantly de-
to become economically viable, as dis- termine the product quality of precision
240 18 Finishing

ground ceramic components (Subramani- monly held views, oil in some instances
an and Ramanath, 1989). may be preferable to water combined with
- Rigidity/stiffness rust inhibitor as a coolant such as is nor-
- Vibration level mally used for ceramic grinding. This se-
- Coolant systems lection will depend heavily on the material
- Creep-feed grinding properties of the workpiece as well as the
- Precision movements and positioning grinding process conditions. The state-of-
- On-machine dynamic balancing the-art coolant systems employed for pro-
- Truing and dressing systems duction grinding of conventional materials
- Multiaxis CNC capability will be equally valuable for production
- Materials handling systems grinding of ceramics. The size of the chips
produced in the grinding of dense, high
18.4.1.1 Rigidity/Stiffness strength ceramics are of the order of 1 to
10 |im. These are substantially smaller
In general, grinding forces for dense ce- compared to typical sizes of chips pro-
ramics are higher than those encountered duced in metal grinding (of the order of
in the grinding of carbides. This implies 100 to 1000 jam). In addition, the lower
that in order for the grinding wheel to pro- density and non-magnetic nature of the ce-
duce the necessary straightness, flatness or ramic chips also pose special filtration
similar surface requirements, we need a problems. In general finer filters of higher
spindle assembly of greater resistance to quality are required for use with advanced
deflection. This higher stiffness is required ceramics grinding. Novel methods of fil-
in the entire spindle/wheel/work fixture/ tration such as centrifuging also require
table assembly. In addition, this stiffness consideration on occasion.
property is required at operating speeds
Because of their fine size and light
(that is, dynamic stiffness rather than stat-
weight, ceramic chips may float and get
ic stiffness alone).
carried into the ways and guides of the
machine tool more easily than steel chips.
18.4.1.2 Vibration Level Such entrainment could accelerate the
In some respects, vibration level is a wear of machine tool parts that are not
measure of stiffness. In addition, it is a adequately sealed.
measure of the damping characteristics in
the spindle itself and the machine tool as- 18.4.1.4 Creep-Feed Grinding
sembly as a unit. Low levels of vibration
are critical to minimizing chippage for As discussed earlier ceramic grinding
grinding of ceramics with thin sections or processes will be driven towards lower
precision forms, especially for low strength grinding forces per abrasive grain to ob-
ceramics such as ferrite and/or low frac- tain maximum retained strength. A com-
ture toughness materials. parison between creep-feed and surface
grinding of hot pressed silicon nitride, for
example, reveals that while the creep-feed
18.4.1.3 Coolant Systems
process requires high total forces and pow-
The direction, pressure and flow of er, the intensities of contact stress and
coolant applied are very critical in the power at the grinding zone are significant-
grinding of ceramics. Contrary to com- ly lower (Pukaite and Subramanian, 1987).
 18.4 Practical Aspects of Finishing Ceramics 241

Creep-feed grinding processes generally available grinding equipment used for elec-
utilize large depths of cut and very low tronic ceramics finishing operations.
work speeds. Machines capable of such
creep-feed grinding in a variety of situa- 18.4.1.7 Truing and Dressing Systems
tions would find preferential use (if ma-
Truing is the process of generating a
chining allowance is large enough) in the
concentric wheel face with accurate form
production grinding of ceramics.
or straightness as required (Fig. 18-19).
Dressing is the process of exposing the dia-
18.4.1.5 Precision Movements
mond abrasives above the bond matrix for
and Positioning
efficient grinding operation (Fig. 18-20).
Ceramic materials, because of their su- Truing and dressing methods applied to
perior hardness and thermal stability, find diamond grinding wheels differ signifi-
applications where tolerances are much cantly from the methods used for conven-
tighter than for metal parts. The surface tional abrasive wheels. There are at least
finishes required in such ceramic parts will six important reasons for this difference:
also be smoother compared to their metal-
lic counterparts. Such precision tolerances • Diamond abrasives (being the hardest
and finish, in turn, will require machine material known on earth) are difficult
tools capable of precision movements, to cut or shape as required by the truing
with a high degree of repeatability, stabili- process.
ty and positioning accuracy (Yoshioka • The tools used for truing diamond
et al, 1985; Bifano et al. 1985). These fea- wheels undergo rapid wear unless the
tures of the machine tools will be enhanced truing system is properly designed and
by the proper selection and application of implemented. This rapid wear of the
diamond wheels. truing tool has serious implications in
setting up automated production grind-
ing cycles.
18.4.1.6 On-Machine Dynamic Balancing
• The amount of diamond wheel removed
The grinding wheel is the final dynamic during the truing process should be
element of the machine tool system that minimized to make the production pro-
contacts the work material. Hence, it is cess cost-effective. Conventional truing
critical that grinding wheels operate at low methods with silicon carbide wheels
levels of vibration. In many instances, this erode the bond matrix of the diamond
will require on-machine balancing systems wheel, thereby removing large wheel
that correct for any degree of imbalance volume.
while the grinding wheel is rotating at op- • Diamond abrasives can be damaged if
erating speeds (Layne and Heck, 1989; the truing process is harsh and carried
Yoshioka et al., 1987). out with high forces. Similarly, the resin
This on-machine balancing operation bond matrix can be thermally damaged
should be applied as part of the truing pro- if the truing and dressing processes are
cess prior to the use of the grinding wheel. not controlled.
In addition, such on-machine balancing is • The precision required in truing dia-
required periodically during the use of the mond wheels for ceramic grinding will
grinding wheel in the grinding operation. be significantly tighter than the preci-
Several such systems are in commercially sion required in current practice for job
242 18 Finishing

Before truing After truing Before truing After truing

Type 1 wheels

I i I
I ! ! I t

Type 1 B1 Typei F1V

sis
Type 2 or 6 wheels

Figure 18-19. Typical examples of conditions that require truing.

shop uses of diamond wheels or for ap-

plications such as production grinding
of glass or carbide parts.
• There will be situations where diamond
wheels of resin or metal bond are likely
to be used in addition to vitrified bond-
ed wheels for production grinding of ce-
Tail (bond
ramics. Such situations require dressing
Abrasive Grain supporting grit) methods and equipment distinctly dif-
ferent from the truing equipment. Au-
tomation and consistency of such a
dressing process will be key require-
ments.
Figure 18-20. Scheme of a wheel that has been trued Truing and dressing equipment or
and dressed; (1) after truing, the wheel face is smooth devices that meet the above requirements
and closed; (2) after dressing, the wheel face is open
will be a key requirement for the precision
with grits exposed and ready for efficient grinding;
(3) after dressing, bond supports the grit; (4) after production grinding of ceramics. Integra-
dressing. Note: path connecting the tails for efficient tion of such equipment with the machine
cooling and chip flow. tool will be necessary for production
 18.4 Practical Aspects of Finishing Ceramics 243

grinding of ceramics {Engineered Materials great deal of effort to adapt existing equip-
Handbook, 1989). ment. However, the level of machine tool
developments or modifications required
are rather involved. Hence, machine tools
18.4.1.8 Multiaxis CNC Capability
developed specifically for ceramics grind-
One of the key requirements of produc- ing may find greater success and accep-
tion grinding economics is the decrease in tance for precision production grinding of
the total cost of grinding. Set-up time, ma- ceramics.
chine-to-machine movement time, and in-
process inventory costs contribute heavily
18.4.2 Abrasive Product Selection
to the total cost of fabrication. These costs
can be significantly decreased if the part All abrasive products used for finishing
can be fabricated on one machine using a of ceramics may be classified on the basis
single set up. This approach is being used of the abrasive used (type, shape, strength
successfully via the application of multi- and other characteristics), abrasive grit
axis CNC grinding systems specifically de- size, bond type (matrix used to hold the
veloped for use with CBN wheels to grind abrasives together and its wear behavior),
steel parts. Similar concepts and cycle de- abrasive content and the abrasive product
sign strategies are likely to find many uses shape and configuration. The details of
in the precision production grinding of ce- abrasive product selection can be obtained
ramics. from a number of references {Engineered
Materials Handbook, 1989, 1994). Abra-
sive product manufacturers should be con-
18.4.1.9 Materials Handling Systems
sulted on the specifics. However, there are
While ceramic materials can meet a wide also general guidelines on this matter
range of applications, it is imperative that which are discussed below.
their potential for chippage be recognized
throughout the production process until
their eventual installation into a finished 18.4.2.1 Abrasives Used in the Finishing
assembly. This necessitates a sequence of of Ceramics
materials handling systems compatible Diamond is the preferred abrasive for
with the ceramic material properties. In finishing of ceramis, the reason being
many respects, these operations are similar highest hardness as a material and relative-
to the materials handing systems devel- ly larger wear resistance compared to ce-
oped for the processing of electronic ce- ramic workpiece materials. Figure 18-21
ramic parts. shows the comparison of hardness of dia-
mond with other workpiece materials and
abrasives.
18.4.1.10 New Machine Tools Versus
Diamond abrasives are used in a wide
Retrofitting of Current Machines
range of sizes, as shown in Fig. 18-22,
A frequently asked question is: Can the based on the application. In general, finer
machine tools used for ceramic grinding be abrasive sizes are preferred for finishing
adapted from existing machines or will a dense and high strength ceramics, while
new generation of machine tools be re- coarse and blockier (higher fracture
quired? Initially there will probably be a strength) abrasives are preferred for mass
244 18 Finishing

Diamond
(superabrasive for ceramics,
glass, carbide, and stone) ~"
CBN
(superabrasive for steel)

Boron carbide

Vanadium carbide
Silicon carbide
(conventional abrasive for steel)
Aluminum oxide
(conventional abrasive for steel)

Tungsten carbide

Silicon nitride

Zirconium oxide

Garnet

Silicon
Hardened steel
(65 HRC)

Quartz

Ferrite

Glass

Magnesia

Soft steel
(85 HftB)

Calcite

2000 4000 6000 8000

Hardness, HK
Figure 18-21. Comparison of hardness of work materials and the abrasives used to machine them.

[Rough (300-6500)

| Prec (50-500)

Ultra precision (0.1-80)

Figure 18-22. Abrasive grain size versus
Illlli 1 1 I Mill i 1 1 1 111 1 application type.
0.1 1 10 100 1000 10000
Average abrasive grit size
 18.4 Practical Aspects of Finishing Ceramics 245

Mesh size
230/270 100/120 60/80 50/60 40/50 30/40
100
I I
A, high toughness diamond
used in hard metal bonds
for low MOR workpieces

75
\

CD
C
sz B, medium toughness
50
CD
diamond used in medium
to soft metal bonds for
medium MOR workpieces
.a
to \
CBN

25

C, low toughness diamond

used in resin and vitrified for
high MOR workpieces

0(0) 100(4) 200 (8) 300(12) 400(16) 500 (20) 600 (24)
Mean grain diameter, um (mil)
Figure 18-23. Toughness of superabrasives shown as a function of abrasive grain size. In addition, the most
effective bond system for each of the three ranges of diamond toughness - A (high toughness), B (medium
toughness), and C (low toughness) - is described in terms of the MOR value (Ratterman and Cassidy, 1991).

finishing of low strength or porous ceram- toughness and work material strength is
ics. These choices are based on both avoid- shown in Fig. 18-23.
ance of surface damage when necessary
and process economics. There is a wide
18.4.2.2 Diamond Grinding Wheels
variety of diamond abrasives available in
terms of their impact strength, shape, size Diamond grinding wheels used for ce-
and the coatings on abrasives. Illustrations ramic grinding are generally of four bond
of abrasive selection based on abrasive types: resin, metal, vitrified, and single-
246 18 Finishing

Table 18-2. Advantages and limitations of diamond abrasive bond types.

Resin bond Vitrified bond Metal bond Layered products

Readily available Free cutting Very durable Single abrasive layer

Easy to true and dress Easy to true Excellent for thin slot, plated on a pre-
Moderate freeness of cut Does not need dressing groove, cutoff, simple machined steel preform
Applicable for a wide (if selected and trued form, or slot grinding Extremely free cutting
range of operations properly) High stiffness High unit-width material
First selection for deter- Controlled porosity to Good form holding removal rates
mining the optimum enable coolant flow to Good thermal conduc- Form wheels, easily pro-
use of diamond the grinding zone and tivity duced
wheels chip removal Potential for high-speed Form accuracy dependent
Intricate forms can be operation on preform and plating
crush formed on the Generally requires high accuracy
wheels grinding forces and High abrasive density
Suitable for creep-feed power Generally not truable
or deep grinding. Difficult to true and dress Generally poorer surface
ID grinding, or high- finish than bonded
conformity grinding abrasive wheels
Potential for longer wheel
life than resin bond
Excellent with oil as
coolant

layer achieved by electroplating or brazing power are the most frequent limiting fac-
the diamond onto a steel preform. The ad- tors in the use of metal bonded wheels.
vantages and limitations of each bond type Many traditional ceramics of low strength
are listed in Table 18-2. and large porosity are cut, finished and
With the wide range of operations and sawed using metal bonded diamond
grinding configurations, it is difficult to set wheels under dry grinding conditions. Fig-
guidelines for the selection of bond type. ure 18-24 Shows a variety of diamond
However, flexibility, ease of use, and re- abrasive products used in the finishing of
silience are the most common factors in ceramics. Table 18-3 presents the parame-
favor of resin bonded diamond wheels. ters influencing diamond wheel selection
Vitrified bonded diamond wheels have for finishing of ceramics.
several advantages for production grind-
ing, including: form holding, higher stiff-
ness, tighter tolerances and light weight. 18.4.3 Effect of Workpiece Material
Metal bonded diamond wheels are nor- Properties
mally chosen when durability or long life is Grinding force requirements vary de-
the primary objective (for example, large pending on the ceramic material chosen
contact area grinding, slot grinding). In (Fig. 18-25) (Subramanian and Keat,
some instances, metal bonded grinding 1985). Grinding power requirements also
wheels have been used in a machining cen- vary with the workpiece material (Fig. 18-
ter to grind complex profiles using small 26). These variations would appear to be
diamond wheels (Nakagawa et al., 1985). dependent on the work material proper-
In general, higher grinding forces and ties. Creep-feed grinding of hot pressed sil-
 18.4 Practical Aspects of Finishing Ceramics 247

Figure 18-24. Typical diamond abrasive

products required for grinding advanced
ceramics.

icon nitride and tungsten carbide indicates ceramic workpiece materials by proper se-
that the HPSN material with its higher lection of grinding parameters (for exam-
hardness (and hence resistance to penetra- ple, unit-width material removal rate, grit
tion) requires higher normal grinding size and depth of cut). In case the ceramic
forces than the tungsten carbide material. grinding is primarily dominated by brittle
However, the tungsten carbide material fracture, there is generally poor control of
(with higher strength) requires higher the surface finish. However, by proper se-
grinding power than the HPSN material. lection of the abrasive grains, a rigid grind-
Porosity, grain size and microstructure ing system, proper truing, dressing and
could have a major effect on surface finish balancing, extremely fine finishes in the or-
and surface quality (Roth and Tonshoff, der of < 0.025 pirn (< 1 juin) can be readily
1993). Surface finish can be controlled in obtained. With mirror finish grinding tech-

30
Unit-width
metal CD
removal rate 25
O
Q.
CD Low
Intermediate £ 20
High c
15
K
I 10
l

<D
1
HPSN ZrO2 Ferrite AI2O3-TiC WC HPSN ZrO2 Ferrite AI2O3-TiC WC
Figure 18-25. Relative unit-width normal force re- Figure 18-26. Relative unit-width grinding power re-
quired to machine various structural and electronic quired to machine various structural and electronic
ceramics. Unit-width material removal rates classified ceramics.
as low (2 mm3 s'1 mm"1), medium (5 mm3 s" 1 mm" 1 ),
and high (10 mm3 s"^1 mm" 1 ) (Subramanian and
Keat, 1985).
248 18 Finishing

Table 18-3. Parameters affecting diamond abrasive and wheel bond type selection for machining ceramics.

Parameter Diamond abrasive Effect on grind process

designation

Type natural or synthetic abrasive strength, consistency, and sharpness of abrasive

grit
monocrystalline self-sharpening characteristics
or polycrystalline
coated or uncoated grit retention strength
Particle size coarse high MRR, poor finish, poor strength
fine (including low MRR, improved finish and good strength
micron powders)
Particle size distribution inconsistent Nonuniform grinding results
uniform/consistent Reliable grinding at low chippage
Content concentration low free-cutting action, low life
high long life, higher grinding forces, and power consumption
Bond properties hardness/grade freeness of cut
stiff/resilient dampening characteristics, deflection and chippage,
slot or form accuracy
porosity freeness of cut, coolant entrainment
thermal conductivity heat removal from grinding zone, thermal gradients

nology, finishes in the order of 1 nm attention in the grinding of ceramics (Sa-

(0.04 juin) have been produced (Subrama- muel etal., 1989; Chandrasekar etal.,
nian and Redington, 1995). 1985). While research in this area will con-
The surface finish obtained in workpiece tinue for years to come, proper attention
materials is also a function of the type of to optimizing the grinding process (that is,
workpiece material. For low-fracture to minimize grinding forces, power, con-
toughness materials such as alumina, the tact stress and heat input to the workpiece
surface finish obtained is poor when com- material) will result in significant near-
pared to HPSN or ZrO 2 materials with the term and practical end-results. Some ef-
same grinding conditions. Surface finish is forts to measure the temperature at the
also related to the thermal shock resistance grinding interface is reported in the litera-
of the workpiece material and its effect via ture (Hebbar et al., 1992). Temperature of
the coolant application. Poor thermal the order of 1600°C is measured in single
shock resisting materials such as alumina abrasive incidence against zirconia ceram-
are prone to thermal cracking during the ic. However, the average temperature mea-
grinding process. Pullout of grains can oc- sured is much smaller when grinding with
cur, particularly in poorly bonded work- an abrasive grinding wheel.
piece materials with coarse grain structure,
especially when fine abrasive grained
18.4.4 Operational Factors
wheels are used to obtain ultrasmooth sur-
faces. While the machine tool, workpiece ma-
Residual stress and surface damage are terial, and the grinding wheel are the dis-
two factors that have attracted significant crete inputs into the grinding system, the
 18.4 Practical Aspects of Finishing Ceramics 249

Table 18-4. Typical operational factors in finishing of ceramics

Relative motions: wheel speed, work speed, feed rate.

Factors based on kinematics: arc length of wheel/workpiece contact, area of contact, depth of cut,
equivalent diameter,
Factors based on time: material removal rate, grinding cycle.
Factors based on geometry: wheel profile, run out, balance limits, wheel face morphology,
Factors based on assembly: stiffness, squareness wheel deflection and side run-out,
Resultant operational factors: chip thickness, specific energy, chatter or vibrations, thermal effects,
frictional interactions, lubrication and cooling conditions.

actual grinding process occurs when these tions (as discussed later) and perhaps less
three inputs interact through the opera- desirable. Decrease in both Vw and d, has,
tional factors such as grinding parameters, the negative consequence of lower material
type of grinding, coolant interactions, tru- removal rate, which in turn influences pro-
ing and dressing, and so on. Table 18-4 ductivity and economics of the process.
indicates some of the typical operational
factors. We shall consider a specific few of 18.4.4.2 Specific Energy
these factors in the following section and
This is the energy required to remove
their influence on grinding results.
unit volume of work material. This is often
obtained as follows
18.4.4.1 Chip Thickness
specific energy (£/) (18-7)
This is the hypothetical quantity of the
= power (immaterial removal rate
depth of the material removed per abra-
sive/workpiece interaction. The average However,
chip thickness may be defined as shown
t/=*y c h +t/ f r i c t i o n (18-8)
earlier in Eq. 18-3. As previously dis-
cussed, there is evidence that suggests that where Uch is the specific energy associated
as the chip thickness is reduced, the abra- with the chip formation process, while
sive/workpiece interaction in the grinding Ufriction
t is the specific energy associated
of ceramics is probabilistically more fa- with all the frictional interactions.
vored to occur under plastic deformation The objective of any abrasive finishing
conditions, which in turn results in better process is to achieve the desired geometry
retained strength, improved finish and bet- and productivity with the minimum of
ter edge quality. specific energy. Uch can be the result of
Reducing chip thickness would imply brittle fracture, in which case it is generally
decrease in work speed (Vw) and depth of small; it may also be due to the result of
cut (d) and increase in wheel speed (Vs), ductile deformation, when the specific en-
equivalent diameter (De) and number of ergy is large. It is absolutely critical to rec-
grains per unit area (c). Of these options, ognize this dichotomy of the need to lower
increasing c by decreasing the abrasive the overall specific energy and yet main-
grain size is the first and most practical tain its component for chip formation
option. Then increasing Vs is the second large, when ductile deformation is the pre-
more desirable option. Increase in De is ferred mode of chip formation mechanism!
often associated with frictional interac- After allowing for this dichotomy, the pro-
250 18 Finishing

1.ABRASIVE/WORK

CUTTING DUCTILE
DEFORMATION
(MATERIAL
REMOVAL
PROCESS) BRITTLE
FRACTURE
_h u chip

PLOWING
(MATERIAL
DISPLACEMENT U
PROCESS)

SLIDING
(SURFACE
MODIFICATION friction
PROCESS)

SLIDING

ADHESION
ABRASION
CORROSION
3.CHIP/WORK EROSION
STRESS CORROSION
FATIGUE
SLIDING THERMAL
MECHANICAL
CREEP
FRETTING
(STICK/SLIP)

4,BOND/WORK

SLIDING
Figure 18-27. Interactions in the
grinding zone and their relationship
to specific energy.

cedure is simply to concentrate on mini- {Engineered Materials Handbook, 1994). It

mizing the C/friction component. This is il- is the ratio of the life or durability of abra-
lustrated in Fig. 18-27. sive product (G-ratio) and the specific en-
ergy consumed during the grinding pro-
cess. G-ratio is the ratio of workpiece vol-
18.4.4.3 Grindability
ume removed to the ratio of abrasive prod-
Sometimes it is useful to compare the uct volume consumed. In the "grindabili-
grinding results in terms of "grindability" ty" of ceramics, this term should be modi-
 18.4 Practical Aspects of Finishing Ceramics 251

fied to take into account the retained ness. When machining ceramic materials,
strength after grinding. chatter usually translates into brittle frac-
There are a number of other operational ture leading to very poor surface finish,
factors which help to minimize the brittle lower strength or chippage.
fracture during grinding as well as to re-
duce the frictional interactions. Instead of 18.4.4.6 Cracking of Ceramics
pursuing one at a time, a more desirable
Mechanically or thermally induced
approach is to minimize these two results
cracks limit the grinding of ceramics. Con-
in every aspect of the grinding system.
ditions that produce cracking in ceramics
include high grinding forces, thermal
18.4.4.4 Chippage
shock, poor removal of heat from the
Cippage is one of the predominant limi- grinding zone and large contact stresses at
tations in the grinding of ceramics. It is the grinding zone.
accelerated when the ceramic has large
porosity or poor strength and when the 18.4.5 Output of the Finishing Process
grinding forces are large or widely variable
The end-product of all the finishing pro-
in magnitude. Chippage is invariably due
cesses is the generation of surfaces of re-
to a brittle fracture process. Any effort to
quired characteristics at acceptable pro-
minimize brittle fracture will also minimize
duction economics. The results pertaining
chippage. Great care is taken when grind-
to surface generation and the associated
ing electronic ceramics to ensure uniform
emerging technologies are described here.
and vibration-free conditions that mini-
mize chippage.
18.4.5.1 Production Viable Grinding
Vibrations induced by machine reversals
are of low frequency and can produce large Table 18-5 describes the results towards
chips. The machine spindle/workpiece/ production viable grinding of ceramics.
holder assembly constitutes a dynamic sys- Material removal rates (MRRs) of 1 -
tem with its own natural frequency of vi- 30 mm 3 /mm s (0.1-3 in3/min in) are now
bration. Resonance at the natural frequen- routinely achieved; the higher MRR are
cy of vibration of the system can lead to used for rough grinding processes while
chippage. Wheel runout is a key factor for the lower values are used for finish grind-
uneven grinding that leads to chippage. Vi-
brations induced by the hydraulic pump
Table 18-5. Production viable ceramics grinding.
impeller can induce a high frequency vi-
bration leading to fine chippage. Some- Past Now
times chippage in a given operation may be
Material removal rate
the result of exposing brittle cracks gener-
(mm3/(sec/mm)) 0.1-1.0 1-30
ated in the previous finishing operation. (in3/(min/in)) 0.01-0.1 0.1-3
Surface finish
18.4.4.5 Chatter (urn) 0.750-1.27 0.0075-0.125
(microinch) 30-50 0.3-5
Chatter are the sustained long-term vi-
brations in the grinding zone. In general, Cycle time: lh
hours -»• 35 min
chatter in the machining of metallic mate- 48 min -> less than 10 min
rials leads to poor surface finish or wavi-
252 18 Finishing

ing. Such MRR, if they are achieved using

coarse abrasive particles (100 jim or
larger), predominantly result in brittle
fracture with poor surface quality (low
strength and poor surface finish). Finer
abrasive grains can be successfully used to
achieve the high MRRs mentioned earlier,
if the machine tool has rigidity and power
to transmit the forces and energy required.
This may, on occasion, also require unique
wheel design to overcome the limitations
of part geometry and fixturing.

18.4.5.2 Mirror-Finish Grinding

Plastic deformation-dominated grinding
of ceramics using fine abrasive particles
often permits controlled finishing of ce-
ramic surfaces. As a result, extremely
smooth surfaces have been produced in (b)
various grinding modes, e.g., surface
Figure 18-28. Mirror-finish grinding of ceramics, (a)
grinding, cylindrical grinding, creep-feed Surface grinding of a silicon carbide ceramic; (b) cylin-
grinding. For example, mirror-finish drical grinding of a hot pressed silicon nitride ce-
grinding of ceramics in external cylindrical ramic.
grinding has resulted in finish of the order
of 7 nm in hot pressed silicon nitride. Such
superior finish is often dependent on the Figure 18-30 a shows four fluted end
work material grain size, uniformity and mill ground from a solid cylindrical rod of
strength. Figure 18-28 shows a sample 16 mm diameter. The entire end mill was
with such fine surface finish generated in finished in 35 min, which compares well
cylindrical grinding. Figure 18-29 shows with 5-15 min required for grinding steel
the scanning electron microscope pictures end mills from cylindrical blanks. The fig-
of this cylindrical surface. ure also depicts a tensile test rod also
gound from simple solid blanks. Figure 18-
30 b shows a number of cylindrical compo-
18.4.5.3 Grinding from Simple Solid Shape
nents and thin rods generated by plunge
When the grinding system is integrated cylindrical grinding of ceramic blanks or
as described earlier, with attention to pro- rods. It presents a rod plunge ground from
cess interactions that minimize the force a nominal size of 10-15 mm to less than
per grain, high MRR grinding conditions 1 mm diameter. This was carried out
can be used to grind components of com- through a plunge cylindrical grinding pro-
plex geometry from simple solid shapes. cess using a vitrified bonded diamond
Figure 18-30 shows several components wheel. Such thin sections are difficult to
that were ground from simple shaped raw produce in the grinding of metals. Howev-
material blanks. er, ceramic materials due to their higher
 18.4 Practical Aspects of Finishing Ceramics 253

>f
i •**

(b)

Figure 18-29. Hot pressed silicon nitride; process: fine grinding using diamond abrasives; Ka = 0.3uin
( = 0.0075 um). (a) Photon tunneling microscope pictures (Guerra et al., 1993 for a description of the photon
tunneling microscope), (b) SEM micrographs.

(a)

Figure 18-30. (a) Ceramic end mills and tensile rod

finished by grinding of solid cylindrical blanks. Work-
piece material: hipped silicon nitride, (b) Thin diame-
ter ceramic rods and other cylindrical components
generated from standard cylindrical rods by plunge
grinding. Workpiece material: zirconia, HPSN,
SiAlON. (c) A selection of complex geometry ceramic
parts with surfaces generated by diamond grinding
processes, from simple solid blanks. (c)
 254 18 Finishing

stiffness do not deflect as much and hence, achieved in microgrinding of ceramics on a

can be machined to achieve extremely thin variety of work materials. In addition to
sections or small cross sections (if brittle the obvious productivity and quality ad-
fracture during grinding is minimized) vantages, this process is environmentally
through the systems approach. Figure 18- benign, as simple coolants can be used
30 c shows a range of parts, most of which without the need for lapping fluids. Also
are produced from simple shaped unfin- oil-bearing abrasive slurry need not be dis-
ished blanks. posed of in microgrinding processes as
none is required!
18.4.5.4 Microgrinding of Ceramics
The principles of controlled chip forma-
tion in finishing of ceramics can be extend- 18.5 Economic Aspects of
ed to emulate precision finishing processes Finishing
used in metal components. Hyperlap is a
fixed-abrasive finishing process in which Every ceramic component starts with
grinding wheels operating at low velocities the ceramic raw material, which in most
replace loose abrasive lapping. It results in cases, will be in powder form. The amount
higher MRR achievable in the grinding of ceramic material used to produce one
processes while maintaining the flatness component is termed as the material cost.
and parallelism achievable in lapping pro- This powder is often processed in one or
cesses. This technology has been extended more ways to achieve the unfinished blank.
in the finishing of ceramics using a process These processes may be hot pressing, hot
called microgrinding. Sometimes this ap- isostatic pressing, sintering, etc. We shall
plication is also described as flat honing. call these "ceramic processing" and the
Table 18-6 shows some of the results cost associated with this, the ceramic pro-

Table 18-6. Microgrinding/flat honing case studies.

Case Material Wheel MRR Surface finish Flatness Thickness

No. finish Ra: tolerances tolerances
(mm3/sec) (urn) (urn) (urn)

Sapphire 320 grit 38xlO 3 0.50 5-10 2.5

vitrified bond
Result: 5 to 10 times faster compared to lapping with conventional abrasives
r
2. Silicon nitride 320 grit 55 0.20 2.5 2.5
vitrified bond
Result: 5 times faster compared to lapping with conventional abrasives
Silicon carbide 320 grit 27 0.25 2.5 2.5
metal bond
Result: 3 times faster compared to lapping with conventional abrasives
4. Alumina/TiC 600 grit 2.2 0.125 no chipping
vitrified bond at 200 x
5. Ferrite 320 grit 55 0.40 2.5 2.5
metal bond
 18.5 Economic Aspects of Finishing 255

cessing cost. The unfinished blank may and overhead costs. Thus
then be finished, often using grinding or
machining cost = machine tool cost
other abrasive processes such as ultrasonic
machining. The cost associated with this 4- labor cost + overhead cost
will be called the "finishing cost". Thus the + tool cost (18-12)
total cost of a ceramic component will
amount to The machine tool, labor and overhead
costs are directly linked to the time it takes
total cost = finishing cost to produce one part, called the cycle time:
4- ceramic processing cost
machine tool cost (M.TC.) = machine rate
+ material cost (18-9)
($/h) x cycle time (18-13)
In reality, these elements of costs are not
independent. "Yield" is defined as the labor cost (L.C.) = labor rate ($/h)
fraction of the total number of parts pro- x cycle time (18-14)
duced to arrive at one good or usable com- overhead cost (OHC) = burden rate ($/h)
ponent. Yield in every processing step in-
x cycle time (18-15)
fluences the total cost. The yield becomes
increasingly critical as we move from the tool cost = abrasive wheel cost/no, of parts
raw material to the finished part. When a machined per wheel (18-16)
part is rejected at the finishing stage, the
total cost of a good part produced increas- Cycle time is defined as the time con-
es not only due to the loss in the finishing, sumed in finishing per part. Thus
but also due to all other costs incurred up cycle time = total time consumed (7)/no.
to this time. This compounding effect is of parts machined in time t. (N) (18-17)
shown in
total time (7) = set-up time
finishing costs
total cost = + inspection time + grinding time
(Y Fin)
(18-18)
ceramic processing cost
+
(Y Fin.) x (Y Proc.) grinding time = wheel/work contact time
+ wheel preparation time (18-19)
material cost
+
(Y Fin.) x Y Proc.) ( ' Wheel preparation time is required for
truing and dressing. Part load and unload
where Y Fin. is the yield in the finishing time is included in the set-up time.
process and Y Proc. the yield in the ceram- Grinding time directly influences the
ic processing operation. number of parts machined per wheel men-
Finishing cost consists of the cost of ma- tioned in Eq. (18-6). Thus cycle time indi-
chining, inspection, set-up or fixturing. rectly influences the tool cost as well.
Thus The above equations describe the ele-
finishing cost = machining cost ments of costs and their interrelationship
+ inspection cost in the manufacture of ceramic compo-
nents. Attempts to minimize each cost ele-
+ set-up or fixturing cost (18-11)
ment individually is tedious and often fu-
Machining costs consist of components tile. It is imperative to recognize the key
due to machine tool, labor, abrasive tools cost elements and attack them strategical-
256 18 Finishing

ly. This approach cannot be solely based be maximized by careful understanding of

on economics, but a rather careful combi- the grinding interactions and optimizing
nation of economic and technical factors them using the systems approach de-
will be required. In the following, a few of scribed earlier. Finishing cost and cycle
our experiences are described. time are two other elements which require
Yield in the finishing process, described close scrutiny in the total cost approach.
in Eq. (18-10), would appear to be a key Finishing cost can often be minimized by
cost controlling element. Table 18-7 shows controlling the inspection and set-up costs.
a hypothetical case and three situations This is pursued simultaneously with cycle
which represent the range of realistic con- time reduction using the understanding of
ditions commonly observed. While the ex- grinding process interactions described
act values are not critical, the trends are earlier. While some of these steps may ap-
important. The total cost noted in our ex- pear to be obvious, it is critical that these
amples (a), (b) and (c) represent the range steps are pursued rigorously in order to
of costs quoted for ceramic components, find early commercialization of advanced
which are 3-15 times that of ideal or antic- and emerging materials. In order to
ipated further costs for ceramic compo- achieve these results, a state-of-the-art
nents. However, it is possible to systemati- grinding system is required which increases
cally attack these total costs using step 1 the finishing rate (machine rate + labor
and step 2 also demonstrated in Table 18- rate + burden rate) by a factor of 4. How-
7. Figure 18-31 shows an actual case where ever, the cost of such increases is totally
this approach has been demonstrated suc- justified with a high return on investment
cessfully. Through these and other exam- or a short payback period, because of in-
ples, it has been learned that the yield can creased throughput (cycle time reduction)

Table 18-7. Influence of ceramic process and finishing process yield on the total cost.

Case Ceramic Finishing Material Process Finishing Total

process yield process yield costa costa costa costa

Hypothetical 100 100 5 5 10 20

Actual
(a) 20 100 25 25 10 60
(b) 100 20 25 25 50 100
(c) 20 20 125 125 50 300
Step 1 b 100 100 6 6 12 24
Step 2C 100 100 7 4 4 15
a
All costs are in relative units. As an example in the hypothetical case, the total material cost (20 units) consists
of 25% in material cost (5 units), 25% in process costs (5 units) and 50 in finishing cost (10 units). Subsequent
examples are illustrated as modifications using the hypothetical case as the reference or base. The above table
is based on our experience in a wide range of advanced ceramic components.
b
Step 1: Yield in the finishing process is optimized through systems approach to grinding which includes modi-
fications to the geometry of the unfinished blank and ceramic processing steps.
c
Step 2: Step 1 is followed by changes in component design (without altering the function) that lower the
ceramic process cost and accelerate the finishing process.
 18.7 References 257

CYCLE 1 100
TIME(HRS.)

@ >

FINISHING 1
RATE($/HR.) Figure 18-31. Grinding of
RELATIVE A I* . ^ |4
UNITS ceramics: cost/benefits
based on the systems ap-
20 proach. Symbol • indicates
YIELD results as practiced in in-
/^v|i^ii^:f:4k* !^v ;* dustry; B symbolizes im-
proved results through sys-
tems approach for grinding
25
NISHING ceramics.
(
RELATIVE
UNITS

and finishing cost reduction. In addition, 18,7 References

the component also exhibited higher
strength and reliability during perfor- Anderson, C. A., Bratton, R. J. (1979), "Effect of
Surface Finish on the Strength of Hot Pressed Sil-
mance tests. icon Nitride", in: The Science of Ceramic Machin-
Surprisingly, methods to lower finishing ing and Surface Finishing II, Special Publication
costs are also associated with sound finish- 562. Washington, DC: National Bureau of Stan-
dards, pp. 351-378.
ing practices necessary to achieve reliable Bifano, T. A., Dow, T. A., Blehe, P., Scattergood, R.
grinding of high strength parts! Thus, cost- O. (1985), "Precision Machining of Ceramic Mate-
effective finishing of ceramics to achieve rials", in: Proc. Intersociety Symp. on Machining of
Advanced Ceramics, PED-17. New York: American
low total cost per part would appear to be Society of Mechanical Engineers, pp. 99-120.
concurrent with reliable finishing methods Bifano, T. A., Dow, T. A., Scattergood, R. O. (1991),
for ceramics. "Ductile Regime Grinding: A New Technology for
Machining Brittle Materials", J. Eng. Ind. 113,
184-189.
Chandrasekar, S., et al. (1985), "Morphology of
Ground and Lapped Surfaces of Ferrite and Met-
al", in: Proc. ASME Winter Annu. Mtg., PED-17.
18.6 Acknowledgements New York: American Society of Mechanical Engi-
neers, pp. 69-74.
Conway, Jr., J. C , Kirchner, H. P. (1980), "The
The author thanks his colleagues at Mechnics of Crack Initiation and Propagation Be-
Norton Company for their inputs and neath a Moving Sharp Indentor", /. Mater. Sci. 15,
valuable discussions. Interesting and chal- 2879.
Conway, Jr., J. C , Kirchner, H. P. (1988), "Crack
lenging problems from our customers have Branching as a Mechanism of Crushing During
constantly helped to challenge common Grinding", J. Mater. Sci. 69, 603.
wisdom and seek more in-depth and com- Dudley, J. A. (1990), "Precision Finishing and Slicing
of Ceramic Materials with Diamond Abrasives", at
plete understanding of the finishing pro- 4th Int. Grinding Conf. p. 550.
cess. The help of Ms. Janice Vasalofsky Engineered Materials Handbook (1989), Vol. 16, Ma-
and Mr. Marc Tricard in the preparation chining Chapter. Materials Park, OH: ASM Int.
Engineered Materials Handbook (1991), Vol. 4: Ce-
of this manuscript is gratefully acknowl- ramics and Glass, Sec. 5: Finishing. Materials Park,
edged. OH: ASM Int.
258 18 Finishing

Engineered Materials Handbook (1994), Vol. 5: Sur- Ohta, M., Miyahara, K., Matsuo, K. (1987), "Effect
face Engineering, Sec. 3: Finishing Methods. Mate- of Grinding Parameters on the Strength of Ceram-
rials Park, OH: ASM Int. ics", /. Jpn. Soc. Precis. Eng. 753.
Ernst, H., Merchant, M. E. (1941), Surface Treatment Pukaite, L. G., Subramanian, K. (1987), "Creep Feed
of Metals. American Society of Metals, pp. 299- Grinding of Silicon Nitride Tool Material", in:
378. Proc. Soc. Carbide and Tool Mater. Annu. Mtg.,
Feng, Z., Field, J. E. (1989), "Dynamic Strengths of Phoenix, AZ.
Diamond Grits, Ind. Diamond Rev. 3, 104. Ratterman, E., Cassidy, R. (1991).
Guerra, J. M., Srinivasarao, M., Stein, R. S. (1983), Rice, R. W, Mecholsky, J. J. (1976), "The Nature of
Science 262, 3195. Strength Controlling Machining Flaws in Ceram-
Hahn, R. S., King, R. I. (1986), Handbook of Modern ics", in: The Science of Ceramic Machining and
Grinding Technology, London: Chapman and Hall. Surface Finishing II, Special Publication 562. Na-
Hebbar, R. R., Chandrasekar, S., Farris, T. N. tional Bureau of Standards, pp. 351-378.
(1992), "Ceramic Grinding Temperatures", /. Roth, P., Tonshoff, H. K. (1993), "Influence of Mi-
Amer. Ceram. Soc. 75, 2742. crostructure on Grindability of Alumina Ceram-
Ichida, Y. (1986), "Mirror Finish Grinding of (3- ics", in: Proc. Int. Conf. on Machining of Advanced
Sialon with Fine Grained Diamond Wheels". Yo- Materials, NIST Special Publication 847. NIST.
gyo Kyohoi-Shi 94, pp. 194-200. Samuel, R., et al. (1989), "Effect of Residual Stress
Inasaki, I. (1987), "Grinding of Hard and Brittle Ma- on the Fracture of Ground Ceramics", J. Am. Cer-
terials", Ann. CIRP, 36, 463. am. Soc. 72, 1960.
Lawn, B. R., Evans, A. A., Marshall, D. B. (1980), Spur, G., Stark, C , Tio, T. H. (1985), "Grinding of
"Elastic/Plastic Indentation Damage in Ceramics: Non-Oxide Ceramics Using Diamond Grinding
The Median/Radial Crack System", /. Am. Ceram. Wheels", in: Machining of Ceramic Materials and
Soc. 63, 514. Components", PED-17. New York: American Soci-
Layne, M. H., Heck, W. C. (1989), "Grinding Wheel ety of Mechanical Engineers, p. 33.
Balancing: Sources and Solutions", Abrasive Eng. Subramanian, K. (1987 a), "Advanced Ceramic Com-
28, 12-17. ponents: Current Methods and Future Needs for
Lindsay, R. P. (1989), "Principles of Grinding", in: Generation of Surfaces", Intersociety Symp. Ma-
Metals Handbook, Vol. 17: Machining. Materials chining of Advanced Ceramic Materials and Compo-
Park, OH: ASM Int. pp. 421-429. nents. Westerville, OH: American Ceramic Society,
Malkin, S. (1984), "Grinding of Metals: Theory and pp. 10-32.
Application", /. Appl. Metalwork. 3, 95. Subramanian, K. (1987 b), "Superabrasives for Preci-
Malkin, S., Ritter, J. E. (1988), "Grinding Mecha- sion Production Grinding - A Case for Interdisci-
nisms and Strength Degradation for Ceramics" in: plinary Effort", in: Proc. Symp. Interdisciplinary
Proc. Intersoc. Symp. on Machining of Advanced Issues in Materials and Manufacturing, Vol. 2,
Ceramic Materials and Components. New York: pp. 665-676.
American Society of Mechanical Engineers, Subramanian, K., Keat, P. P. (1985), "Parametric
pp. 57-72. Study on Grindability of Structural and Electronic
Mayer, Jr., J. E., Fang, G. P. (1994), Efficient High Ceramics-Part I", in: Proc. Symp. on Machining of
Strength Finish Grinding of Ceramics, Advancement Ceramic Materials and Components, Winter Annu.
of Intelligent Production: Japan Society of Preci- Mtg., New York: American Society of Mechanical
sion Engineering (Ed.). Amsterdam: Elsevier Engineers.
Miyashita, M. (1985), "Ultraprecision Centerless Subramanian, K., Lindsay, R. P. (1989), "A Systems
Grinding of Brittle Materials", at 1st Ann. Preci- Approach for the Use of Vitrified Bonded Su-
sion Eng. Conf., North Carolina State University, perabrasive Wheels for Precision Production
Raleigh, NC. Grinding", in: Proc. Symp. on Grinding Technolo-
Nakagawa, T., Suzuki, K., Uematsu, T. (1985), Three gy, Winter Annual Mtg. New York: American So-
Dimensional Creepfeed Grinding of Ceramics by ciety of Mechanical Engineers.
Machining Center", in: Proc. ASME Winter Annu. Subramanian, K., Ramanath, S. (1989), "Machine
Mtg., PED-17. New York: American Society of Tool Developments Required for Precision Pro-
Mechanical Engineers, pp. 1-8. duction Grinding of Ceramics", in: Proc. 3rd Bi-
Nakajima, L, Uno, Y, Fujiwara, T. (1989), "Cutting ennial Int. Manufacturing Technology Research Fo-
Mechanism of Fine Ceramics with a Single Point rum, Tokyo.
Diamond", Precis. Eng. 11, 19. Subramanian, K., Ramanath, S. (1992), "Mechanism
Ohta, M., Miyahara, K. (1990), "The Influence of of Material Removal in the Precision Grinding of
Grinding on the Flexural Strength of Ceramics", Ceramics", in: Proc. Symp. on Precision Machin-
in: Proc. 4th Int. Grinding Conf. Report MR90-538. ing. PED Vol. 58. New York: ASME.
 18.7 References 259

Subramanian, K., Redington, P. D. (1995), "Opti- Grinding Syrnp., Vol. 16. New York: Production
mized Grinding of Ceramics - A Systems Ap- Engineering Division, American Society of Me-
proach", in: Ceramic Technology International, chanical Engineers, pp. 209-227.
pp. 197-203. London: Sterling Publications. Yoshioka, X, Hashimoto, F., Miyashita, M. (1987),
Subramanian, K., Redington, P. D., Ramanath, S. "Application of Grinding Wheel to Ultraprecision
(1994), "A Systems Approach for Grinding of Ce- Machining: Machining for Precise Surface Genera-
ramics", Bull Am. Ceram. Soc. 73, pp. 61-66. tion on Grinding Wheel", in: Proc. Intersociety
Yoshioka, I, Hashimoto, R, Miyashita, M., Daito, Symp. on Machining of Advanced Ceramic Com-
M. (1985), "Ultraprecision Grinding Technology ponents. Westerville, OH: American Ceramic Soci-
for Brittle Materials: Application to Surface and ety, pp. 50-69.
Centerless Grinding Processes", in: Milton C. Shaw
19 Joining of Ceramics
Michael G. Nicholas

Formerly of AEA Technology, Harwell Laboratory, Oxfordshire, U.K.

List of Symbols 262

19.1 Introduction 263
19.2 Overview of Joining Processes 264
19.3 Capillarity 266
19.4 Glazing 267
19.4.1 Principles 267
19.4.1.1 Chemical Effects: Wettability and Interfacial Reactions 267
19.4.1.2 Physical Factors 268
19.4.2 Practice 269
19.4.2.1 Materials Selection 269
19.4.2.2 Workpiece Preparation 270
19.5 Brazing 270
19.5.1 Coating Prior to Brazing 271
19.5.2 Reactive Metal Brazing 272
19.5.2.1 Principles 272
19.5.2.2 Practice 275
19.6 Diffusion-Bonding 277
19.6.1 Principles and Process Parameters 277
19.6.1.1 Physical Factors 277
19.6.1.2 Chemical Effects 280
19.6.2 Practice 281
19.7 Joint Design 282
19.7.1 Glazed Joints 282
19.7.2 Brazed Joints 282
19.7.3 Diffusion-Bonded Joints 284
19.8 Joint Evaluation and Properties 284
19.8.1 Techniques for Strength and Toughness Measurements 285
19.8.2 Strength and Toughness Values 287
19.8.3 Nondestructive Evaluation 289
19.9 References 290

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
262 19 Joining of Ceramics

List of Symbols
D design constant
E elastic modulus
AG free energy
Klc plane strain fracture toughness
>H penetration (vertical, horizontal)
t time
% TM temperature, melting temperature
w capillary width
W energy

Aa difference in expansion coefficients

ys, yL surface energy (solid surface, liquid surface)
ySL, ySlS2 interface energy (solid/liquid, solid 1/solid 2)
rj viscosity
6 contact angle
Q liquid density
cr R residual stress
 19.1 Introduction 263

19.1 Introduction One of the more obvious and expected ef-

fects of this difference is the poorer electri-
Ceramics are among humanity's oldest cal conductivity of ceramics. However, ce-
technological materials and the need to ramics do not have identical electronic
join them has been recognised since time characteristics and it can be seen from
immemorial. This article focuses on the Table 19-1 that their electrical conductivity
joining of ceramics for recent technologi- increases in the order oxides -> nitrides -•
cal applications, and in particular on the carbides. Important though these trends in
materials science and engineering involved electrical conductivity may be for the sci-
when making macroscopic ceramic-ce- entific understanding of ceramic proper-
ramic and ceramic-metal joints to pro- ties, of more importance in ceramic-metal
duce engineering components. However, joining is the generally greater refractori-
insight into the principles and practices of ness of ceramics and their lower thermal
ceramic joining can be gained also from expansivities which affect the generation
descriptions of microscopic joining pro- of residual stresses in joints. It is also note-
cesses such as the fabrication of ceramic worthy that ceramics are generally lighter,
fibre and ceramic particulate reinforced more thermally insulating, and less yield-
matrices. Additionally much can be learnt ing than metals and can be transparent.
from studies of naturally occuring ce- There are, however, significant exceptions
ramic-metal interfaces such as those to these generalisations; molybdenum,
formed by oxide films on metal surfaces or niobium, and tungsten are more refractory
carbide precipitates in metal alloys. than many ceramics and diamond and
Joining technology is required to satisfy beryllia are excellent thermal conductors.
two classes of need: Reliable joining technologies offer the
designer and engineer the opportunity of
• Production of complex structures from developing devices exploiting the useful
simple shapes, as when bricks are used and even contradictory properties of ce-
to build houses. ramics and metals by incorporating both
• Incorporation into multimaterial devices, classes of material. For most of these de-
as when flints were bound to resilient velopments, the joint between the ceramic
wooden handles to make axes or WC/ and the metal must be permanent. Exam-
TiC cutting blocks are clamped in mod- ples encountered everyday include the use
ern machine tools. of sealed transparent glass bulbs to con-
Requirements for the joining of ceram- tain inert gases and protect the electrically
ics to ceramics and particularly of ceramics conductive tungsten filaments of incandes-
to metals, are expanding and feature in cent lamps from virtually instantaneous
many aspects of current technology. Most failure due to oxidation, the hermetic
of these applications require the joint to be sealing of the metal electrodes of spark
permanent, but satisfying this requirement plugs within insulating alumina bodies, the
is difficult because of the differing natures sealing of metal fillings in teeth, the bond-
of ceramics and metals. Ceramics have ing of electronic microcircuitry to insulat-
ionic or covalent structures and hence their ing ceramic substrates and the repairing of
bonding electrons are not delocalised as treasured china or porcelain with adhe-
with metals but are confined to the imme- sives. Emerging advanced technology also
diate locality of one or two ions or atoms. has its requirements, such as the incorpo-
264 19 Joining of Ceramics

Table 19-1. Some ceramics and metal properties.

Material Melting Density Coefficient Thermal Electrical

temperature of thermal conductivity resistivity
expansion
(°C) (Mgm" 3 ) (K-^xlO"6 (Wm^K-1) (Qcm)

A12O3 2030 4.0 7.9 35 1014

BeO 2530 3.1 7.4 210 1O10
MgO 2800 a 3.6 11.6 62 10 12
SiO2 1710 2.3 3.0b 1.5 10 1 2
ZrO 2 2960 5.6 7.5 19

Si 3 N 4 1900a 3.2 2.5 17 3xl0 4

TiN 2900 5.4 8.1 17 5xlO8

B4C 2350 2.5 4.3 26 3xlO" 2

SiC 2700 a 3.2 4.3 50 10"3
TiC 3140 4.9 7.2 36 5xlO" 6
we 2777 15.8 5.2 84

Ag 960 10.5 19.1 425 1.6xlO"6

Al 660 2.7 23.5 238 2.7xlO~6
Cu 1083 9.0 16.6 379 1.7xlO"6

Fe 1535 7.9 12.1 78 lOxlO" 6

Mo 2615 10.2 5.1 137 6xl0" 6
Nb 2467 8.6 7.2 54 16xlO~6
Ni 1455 8.9 13.3 89 7xlO" 6
Ti 1677 4.5 8.9 22 54xlO" 6
W 3387 19.3 4.5 174 5xlO~6
a
Does not melt but decomposes or vaporises. b Crystalline SiO 2 , for vitreous SiO2 the value is 0.5 x 10~6K~

ration of refractory wear and corrosion re- 19.2 Overview of Joining Processes
sistant ceramic inserts in automobile and
aircraft engines, the cladding of metallic A very wide range of techniques can be
hip replacements with biocompatible wear used in principle or practice to join ceram-
resistant ceramics and the attachment of ics, as illustrated in Table 19-2 and there is
electrical leads to ceramic high tempera- no one "best" technique. Each has its own
ture superconductors. advantages and limitations.
The range of possible applications for Mechanical attachment does not create
reliable joining technologies is very wide permanently bonded structures but is used
and many techniques have been developed widely for domestic applications such as
or suggested. A brief overview of these the manufacture of reading glasses and
techniques is provided in the next section, jewellery, and for important technological
followed by a more detailed discussion of functions. The range of demanding tech-
materials science and engineering aspects nological applications is illustrated by the
of three important processes; glazing, use of hooks and dog-bones to locate fur-
brazing and diffusion bonding. nace roof refractories and the clamping of
 19.2 Overview of Joining Processes 265

Table 19-2. Characteristics of joining processes (L: ture ingress which degrades their interfa-
low, M: moderate, H: high). cial bonding. Sophisticated procedures
Process Joint have been developed for the successful
joining of metals, but the application of
Integrity Use tem- Vacuum Utili- adhesives for joining ceramics is much less
perature tightness sation
developed and systematised. In particular,
Mechanical L L->H L H knowledge of bonding procedures for the
attachment newer nonoxide ceramics is sparse.
Adhesive M L M H Fusion welding is the generally pre-
bonding ferred method for producing permanently
Glazing H M H H bonded metal-metal systems and is a long
Brazing H H H H established practice for making glass-glass
Diffusion H H H L joints, albeit subject to severe constraints
bonding imposed by the need to closely match the
Fusion H H H L coefficients of thermal expansion of the
welding materials being joined. The application of
fusion welding to join ceramic-ceramic
and ceramic-metal systems is subject to
ceramics at the leading edges of the wings even more constraints. For example, ce-
of the NASA Space Shuttle. Less exotic ramics are difficult to fuse because of their
developments can be found in heat ex- refractoriness, their melting temperatures
changer and engine applications. The tech- are usually far higher than those of metals,
niques of mechanical attachment, princi- and some dissociate before they melt.
pally clamping, tying and shrink fitting, Ceramic-ceramic and ceramic-metal sys-
are among the first to be considered by tems have been fusion-welded in the labo-
design engineers. However, mechanically ratory, but the technique has yet to gain
attached joints are usually unsuitable industrial acceptance although a variant,
when the components must be leak tight or friction welding, has been receiving atten-
are subject to thermal cycling. For these tion recently. In this process, a rapidly ro-
applications, joints with permanently tating metal workpiece has its bonding
bonded interfaces are usually necessary. face forced against a static ceramic and the
The commonest process for producing resultant frictional drag generates enough
permanently bonded structures is proba- heat to melt the surface region of the metal
bly adhesive bonding. This is the most (Nicholas, 1990).
widely adopted household technique and For the most demanding applications,
also offers many advantages to the indus- the approach adopted in industry usually
trial fabricator: it requires little capital in- is to employ glazing or brazing. These pro-
vestment, is quick and can be performed at cesses are cogenerous, both depending on
ambient or low temperatures. However, capillary flow to ensure joint creation and
the service temperatures of adhesively both potentially suffering from joint deg-
bonded joints are low: epoxide and pheno- radation due to excessive chemical interac-
lic adhesives are limited to about 150°C, tion of the workpieces with the glaze or
and even polyimide and bismaleimide ad- braze. A close analog of brazing is the use
hesives are limited to 200-300 °C. Further, of ductile metal interlayers to promote dif-
adhesive joints can be susceptible to mois- fusion bonding, the achievement of a joint
266 19 Joining of Ceramics

by application of pressure at a high sub- If ys is greater than (yL + ySL)>the lic L uid
solidus temperature. will wet the solid completely. It will spread
From this brief overview it is clear that all over the surface. If ys is smaller than
the most attractive joining processes for (yL + ySL) but larger than ySL, the liquid will
applications where service temperatures wet partially and the contact angle, 0, will
may be high or environments may be ag- be less than 90°. For successful glazing and
gressive are glazing, brazing and diffusion brazing, for penetration of capillary gaps,
bonding. The next sections describe their it is essential that 9 is less than 90° and
principles and practice in more detail. desirable that it is less than about 20°.
First, however, some attention will be paid Penetration of a horizontal capillary
to the role of capillarity which underlies should be infinite for an inert system if 9 is
the practices of both glazing and brazing. less than 90°, but the penetration of a ver-
tical capillary varies progressively with 9.
For a capillary between plates so wide that
19.3 CapiUarity end effects can be ignored
The successful achievement of both 2y L cos0
glazing and brazing depends on liquid flow (19-2)
WQg
over workpiece surfaces and into narrow,
capillary, gaps between the components to where Pv is the vertical penetration, w the
be joined. The liquid must wet. For this to capillary width, Q the liquid density and g
occur naturally, there must be a net energy the acceleration due to gravity.
release when an interface is formed and In producing capillary joints, it is also
this depends on the relative sizes of the desirable that the penetration is rapid. For
surface and interfacial energies of the sys- a horizontal capillary, the penetration Pn
tem. For a small drop of liquid at rest, after time t can be calculated from
sessile, on a solid horizontal surface, the
relationship between wettability and the ^W — (19-3)
energies can be described by the Young
equation where r\ is the liquid viscosity. For a verti-
cal capillary, the time dependence of pene-
7S = 7L COS0 + 7SL (19-1) tration is more complex
2
where y is the energy in J m ~ of the solid
12*/
surface, S, the liquid surface, L, or the
solid liquid interface, SL, and 9 is the con-
tact angle assumed by the liquid at the
[- QgPy w-2y L cos01n •

drop periphery (Fig. 19-1).

even for very wide plates. It can be seen
that the effects of orientation and liquid
viscosity on joint penetration can be pro-
found. Effects of orientation on penetra-
tion kinetics for a prototypic braze and
solder are illustrated in Fig. 19-2. Similar
Figure 19-1. Sketch of a wetting sessile drop identify- effects can be calculated for glazes, but the
ing the contact angle 9. times involved to achieve a given penetra-
 19.4 Glazing 267

200
Sn horizontal ,-y Many brazes used to join ceramics are de-
eyf Cu horizontal
liberately formulated to react chemically
with the solids, and have to do so if they
are to wet.

Cu vertical

19.4 Glazing
Glazing is an old established technique
and finds important high volume applica-
tion in many industrial sectors. It is used
1 2 particularly intensely by the electrical in-
Time, seconds dustry in the manufacture of lamps and
Figure 19-2. The predicted penetrations of copper electrical feed throughs such as spark
and tin into 0.25 mm wide horizontal and vertical plugs. More generally, the utility of glazing
capillary gaps of a perfectly wetted material.
is exemplified by enamelling and the devel-
opments of glass ceramics and glass matrix
composites.
tion are typically orders of magnitude Successful glazing depends on the opti-
longer. mised exploitation of both chemical and
These calculations refer to the capillary physical effects. The study of these has
processes of ideal systems in which no been the subject of intensive scientific
chemical interactions occur and in which work during the last 50 years, which has
the solid surfaces are perfectly smooth. enhanced our understanding of both the
Real glazing and brazing systems can be principles and practice of glazing.
very different. Thus whilst a molten braze
can be regarded as a simple liquid with 19.4.1 Principles
readily characterisable physical properties,
19.4.1.1 Chemical Effects:
glazes do not have clear melting tempera-
Wettability and Interfacial Reactions
tures above which they are liquid but
merely become increasingly fluid over a The first requirement for successful
range of temperatures. Thus glazing is not glazing is that there is intimate contact be-
performed at a superheat of so many tween the glass and the workpiece, that the
degrees but at a "working" temperature glass wets the workpiece due to a contin-
which generally corresponds to a viscosity uum between the electronic structure of
of about l O N s m " 2 . (In comparison, the glass and the workpiece. Since the glass
metal viscosities are 1 0 ~ 3 t o l 0 ~ 2 N s m ~ 2 has an oxide structure, the continuum and
and that of water is 1 0 ~ 3 N s m ~ 2 ) . As hence a low ySL value is most readily
the temperature is decreased after glazing, achieved if the workpiece surface is also an
the viscosity increases and the glass ceases oxide. This is of course the case when the
to be "soft" when this reaches about workpiece is an oxide ceramic, but the sit-
1 0 7 N s m ~ 2 and becomes rigid, "set", at uation with non-oxide ceramic workpieces
about 1 0 1 1 N s m ~ 2 . Another complicat- is more complicated and will be discussed
ing factor in real joining process is that later. With metal workpieces, conditions
brazes and glazes are not chemically inert. favouring a continuum can be achieved by
268 19 Joining of Ceramics

prior oxidation of the metal. However, it is Separate oxide phase

important that adherent oxide films are -M-M-M-O-M-O-M-O-Si-O-M-O-Si-O-Si-
generated.
If an appropriate surface is produced,
then ySL will be less than ys and the contact
angle 6 will be less than 90°. This follows Oxide monolayer
from the Young equation, which relates to -M-M-M-O-Si-O-M-Si-O-Si-O-Si-O-
an equilibrium situation, but reactions
occur in many glaze-workpiece systems.
For example, a new phase may be formed
at the interface, or glass components may No oxide layer
dissolve in the solid workpiece. This will -M-M-M O-Si-O-Si-O-Si-O-Si-O-
release energy, and ySL will be decreased Figure 19-3. Schematics of glass-metal interfaces.
effectively by an amount equal to the re-
lease per unit area of interface. However,
as the reaction progresses, so the rate of
is important that the layer does not dis-
release decreases: the reaction product lay-
solve completely. If this occurs, then the
ers thicken to create more substantial dif-
interface is bound only by weak van der
fusion barriers, and the approach of solute
Waal forces as illustrated schematically in
concentrations to saturation levels dimin-
Fig. 19-3. Any oxide that dissolves in the
ishes the driving force for dissolution or
glass will diffuse to create an activity gradi-
reaction. Thus there is a time dependant
ent, which will be very steep initially but
deviation from the equilibrium energy bal-
will decrease progressively with time. The
ance (Pask et al., 1974). The Young equa-
presence of the oxide solute will affect the
tion needs to be rewritten as
thermal expansively of the glass, but more
d(AGR/dA)~ extensive diffusion will tend to create shal-
(19-5)
low, and hence less detrimental, expansiv-
ity and stress gradients. It is advantageous
where AGR is the free energy of the reac-
to preoxidise to form thick oxide films,
tion.
provided they are adherent. This also con-
While nonequilibrium conditions, chemi- fers the advantage of generating shallower
cal reactivity, can promote the wetting and
solute and expansivity gradients.
spreading of glazes, the strongest adhesion
has been associated with stable interfacial
19.4.1.2 Physical Factors
chemistries. Essentially this means that the
interface region must be saturated with a Some reference has been made already
low valent oxide of the substrate. This is to one important physical factor when dis-
particularly important when the substrate cussing dissolution effects, the coefficient
is a metal. Thus exemplary studies by Pask of thermal expansion. The close matching
(1987) have shown that good glass-metal of the coefficients of dissimilar workpieces
bonding is associated with linkage between is necessary if undue strain requirements
a "metallous" oxide and the glass. This are not to be placed on the glazed joint.
oxide layer need only be a monolayer thick Similarly it is desirable for the coefficient
in theory, but in practice significantly of the glaze to be similar to that of any
thicker oxides are usually desirable since it rigid workpieces it is desired to join. Ide-
 19.4 Glazing 269

ally, the glaze should be in slight compres- cal bonding between the workpieces and
sion because of its greater strength in this the glaze.
mode than in tension.
Since glazing is done at high tempera- 19.4.2 Practice
tures, it is necessary to achieve a close
19.4.2.1 Materials Selection
match over the range of temperatures en-
countered as the component cools. This The manufacture of glazed joints be-
can be a very demanding requirement be- tween oxide ceramics is relatively straight
cause glass expansivities are often nonlin- forward once logical material selections
ear whereas ceramic and metal expansivi- have been made. Thus it is desirable for the
ties can be approximately linear over quite glaze to have a coefficient of thermal ex-
wide temperature ranges. Thus it is possi- pansion matching that of the workpieces.
ble to have matched coefficients at the fab- Further it is necessary that the set temper-
rication temperature and at room temper- ature of the glaze exceeds those encoun-
ature, with failure occurring at some tered in service, although this requirement
intermediate temperature. When failure can be relaxed somewhat for very thin
does occur, it is due to excessive residual joints. Satisfaction of the requirements for
stresses. Hence the "set" temperature and expansivity and refractoriness can be in
the elastic modulus, E, of the glass are of conflict since the expansivity is usually less
importance. Since the plasticity of glass is for the more refractory glasses, Fig. 19-4
negligible, the residual stress, <rR9 at tem- and hence those suitable for high expan-
perature T can be described by the equa- sion workpieces will have relatively low
tion service temperatures.
Many glazed joints are made between
Jr (19-6) metal and glass workpieces and this has
required the development of ingenious de-
where Ts is the "set" temperature, Aa is the signs to overcome mismatched coeffi-
difference in expansion coefficients, and D cients. The need has also led to the use of
is a design sensitive constant. However, 7^
is not a rigid value for a particular glaze
but depends on its thermal history and can 2000

be altered by interfacial reactions. Addi-

tionally, the design constant D can vary
dramatically. Thus it is not possible to
make simplistic predictions of residual
stress levels, but some influential factors
have been identified, such as the workpiece
surface texture and adsorption behaviour.
Severe roughening of workpiece surfaces
can enhance mechanical keying of the
glaze and this can be advantageous if only
weak van der Waals bonding forces oper- CTE (10 6 K" 1 )
ate across the interface. In contrast, physi- Figure 19-4. Comparison of the refractoriness and
cal adsorption of gases is detrimental, both expansivity of some glasses. Data taken from Hol-
decreasing ys and preventing direct chemi- loway (1973).
270 19 Joining of Ceramics

metals and alloys with low coefficients braze alloy. An example is the use of
metals such as molybdenum when service CaO-TiO 2 -SiO 2 to bond Si 3 N 4 , which
conditions will not cause oxidation, and produces a bonding layer of TiN.
alloys such as Kovar and Invar.
Similarly, a new class of glasses, glass- 19.4.2.2 Workpiece Preparation
ceramics, has been developed in the last
The surfaces of workpieces intended for
few decades which can ease the problem of
glazing generally should be smooth. Ex-
mismatched coefficients by devitrifying at
ceptions occur when the interfacial bond-
high temperatures to produce interlocked
ing is expected to be very poor, when resort
crystalline precipitates. The behaviour of
may have to be made to the mechanical
this two phase material is akin to that of
keying achievable with roughened sur-
monolithic ceramics, and it is of particular
faces. The surfaces themselves should be
note that composition and thermal pro-
free of hydrocarbon contamination and
cessing optimisation can result in near lin-
also of adsorbed gases whenever possible.
ear thermal expansivities that match
When the glaze is to be contained in a
closely those of technically important ox-
capillary gap, it is essential that the sur-
ides and metals.
faces of the adjacent workpieces conform,
While glass-ceramics can have higher that they are flat in the case of butt joints
coefficients of thermal expansion than and that sleeved components can be lo-
many glasses, the coefficients of technically cated concentrically with minimal annular
important nonoxide ceramics such as SiC, variability.
A1N and, particularly, Si 3 N 4 are low, with
In the case when a ceramic is to be
coefficients of about 4.5 to 2.5 x 10" 6 K~ 1 .
joined to a metal, it is usually desirable to
This eases the problem of choosing glazes
preoxidise the metal surface. A finite thick-
with similar expansion coefficients, but it
ness of oxide is necessary if interfacial sat-
is difficult to achieve an electronic con-
uration and chemical equilibrium is to be
tinuum at glaze-nonoxide interfaces be-
achieved, but the oxide must also be adher-
cause of the higher covalency of the work-
ent to the metal substrate. In general this
pieces. Hence inadequate wetting and
means it is desirable to pre-oxidise for long
bonding can pose problems for the fabrica-
times a low temperature rather than rapidly
tor. Three approaches have been adopted
at high temperatures. Thus an oxide film
when addressing this challenge which have
grown in air on Kovar by heating at 800 °C
met with some success:
for 30 minutes is adherent and thick
• Use a glass which is a naturally occur- enough for most glazing processes. Films
ring binder phase in nonoxide ceramics. grown at 900 °C are thicker but not adher-
This has led to the successful use of an ent enough, while those formed at 600 °C
MgO-Al 2 O 3 -SiO 2 glass as a glaze for are too thin.
Si 3 N 4 .
• Use a glass which is specifically formu-
lated to be chemically compatible with 19.5 Brazing
the ceramic. Examples include borosili-
cate glasses used to form adherent coat- Many ceramic-metal systems are non-
ings on A1N. wetting, and hence the achievement of suc-
• Use a glass which reacts chemically with cessful brazing depends on adoption of
the ceramic, as does an active metal steps that induce wetting either by increas-
 19.5 Brazing 271

ing ys values or decreasing ySL values. In Many aspects of the moly-manganese

practice, the first approach of increasing ys process are cloaked in commercial secrecy
requires the use of metal coatings which and its optimum practice varies from user
also change the basic chemistry of the ce- to user, reflecting equipment characteris-
ramic surfaces prior to brazing. The sec- tics as well as the basic materials science.
ond approach of decreasing ySL without However, the process typically involves
trying to affect ys requires the use of brazes painting the surface of an alumina ceramic
that react chemically with the ceramic dur- with a powdered mixture of molybdenum,
ing the fabrication process. manganese and, for example, a calcium sil-
icate glass and then heating in hydrogen or
nitrogen-hydrogen to 1400-1500 °C.
19.5.1 Coating Prior to Brazing
The original and simplest form of the
Metal coatings, particularly of titanium process was painting of the surface of a
and nickel, have been applied to many ce- siliceous ceramic with molybdenum or
ramics prior to brazing by thin film'tech- tungsten powders and firing in hydrogen.
niques. These include physical and chemi- It was found that the addition of man-
cal vapour deposition, sputter, ion plating, ganese improved the strengths of joints
electroless deposition and hydride decom- formed with the metallised ceramic. Subse-
position. (This last process is related to quent use with high purity alumina and
reactive metal brazing, and further refer- other oxide ceramics was difficult until the
ence will be made to it in Sect. 19.5.2.2.) metal powders were mixed with glass frit to
These processes can be used to provide di- promote bond formation with the ceramic.
rectly wettable surfaces whose success de- Glass additions to the metallising paints
pends on their relative insolubility in the must bond to the ceramic as well as the
braze, or, in contrast, on their interdiffu- metal powder, they must flow over the ce-
sion with an inserted metal foil to produce ramic surface or preferably penetrate its
a wetting braze composition. A practice grain boundaries as well as flow into the
now being considered is to further elabo- capillary gaps between the metal powder
rate the interdiffusion process to achieve particles. These disparate requirements
transient liquid phase bonding. This re- can be difficult to satisfy and the composi-
quires interdiffusion initially to cause melt- tion of the glassy addition can be of critical
ing of the foil surface, producing a wetting importance. Thus laboratory studies have
liquid, and ultimately for this liquid to re- demonstrated that strongly adherent coats
solidify as further diffusion causes homog- are formed by metallisation paints con-
enisation of the foil. taining manganese silicate, manganese alu-
The coating method most widely used to mina silicate and calcium silicate glasses
promote the brazing of ceramics is the which can penetrate alumina grain bound-
moly-manganese process. It was originally aries (Twentyman, 1975). In contrast, the
developed to assist the brazing of siliceous adherence of metallisation paints con-
ceramics such as debased aluminas. Subse- taining a calcium aluminate glass, which
quently the process was developed for use spreads over the alumina surface but does
with high quality, low silica, grades of alu- not penetrate the grain boundaries, is rela-
mina. The process has been applied also to tively poor.
other oxide ceramics, and variants have Each component of the paint has a spe-
been developed for nonoxides. cific function. The glass forms the bond
272 19 Joining of Ceramics

between the ceramic and the metallic com- Overcoming poor adhesion requires re-
ponents, the manganese, and other sec- finement of process parameters, but lack
ondary additives such as titanium, form of wetting can be remedied usually by
oxides which promote the flow and bond- overplating the coating with a metal such
ing of the glass, while the molybdenum as nickel or copper, and is a common pru-
particles remain unoxidised and form a dent practice. Successfully applied coat-
metallic layer at the free surface of the ings should be brazeable, and solderable,
coating, preferably at least partially sin- using fluxes and gas torch techniques. The
tered together to form a continuous crust. usual practice, however, is to furnace braze
To fulfil the roles assigned to manganese in inert gas or evacuated environments us-
and molybdenum imposes specific envi- ing precious metal alloys such as the A g -
ronmental requirements. The oxidising po- 28%Cu eutectic.
tential of the hydrogen or nitrogen-hy- The success of the moly-manganese pro-
drogen gas in which the component being cess caused major development efforts to
metallised is heated must be such that be focused on widening its applicability
manganese is oxidised but molybdenum is (Reed et al., 1966). In particular, attention
not. The reaction has been paid to the utility of lower tem-
perature glass frit additives, to its use with
MoO 2 + 2H 2 -> 2H 2 O + Mo (19-7)
high purity oxide ceramics other than alu-
must be favourable, and the reaction mina, and to the possibility of metallising
nonoxides such as aluminium nitride. Con-
MnO + H 2 -» H 2 O + Mn (19-8)
siderable progress has been made and suc-
must be unfavourable. Thermodynamic cessful demonstrations have been con-
data for MoO 2 , MnO and H 2 O demon- ducted. However, these have not yet result-
strate that molybdenum will remain metal- ed in widescale commercial applications,
lic and manganese will be oxidised at although this may soon be true of metal-
1400 °C if the H 2 O/H 2 ratios are between lisation processes for aluminium nitride,
4.3 x 10~ 4 and 2.67. These moisture levels some of which contain copper as well as
are significant because of their chemical molybdenum as metallic components.
effects, but the wetness of the gas is charac-
terised in terms of a physical parameter: 19.5.2 Reactive Metal Brazing
the dew point. This is the temperature at
which the vapour pressure of water or ice 19.5.2.1 Principles
is equal to the partial pressure of moisture To change the wettability of a ceramic
in the hydrogen. Thus the ratios referred to by decreasing the ySL value it is necessary
above are equivalent to dew points of to change the chemistry of the interface.
- 3 0 ° C a n d +90°C. This is done during the course of the fabri-
The principal causes for rejection of cation process when reactive metal brazes
metallised ceramics are poor adhesion of are used. The technical literature contains
the coating, usually due to inadequate many descriptions of wetting promotion
grain boundary penetration by the glass by reactive metals added to braze alloys.
and poor wetting by the brazes due to in- There is evidence that beneficial effects can
adequate segregation of metal particles to be produced by additions of aluminium,
the coating surface or their failure to sinter chromium, bafnium, vanadium, titanium,
partially to form a continuous metal crust. zirconium and many other metals. In prac-
 19.5 Brazing 273

tice, however, substantial development of postoichiometric compound, it is electri-

reactive metal brazing has been limited so cally conductive and wettable by braze
far to silver-copper alloys containing tita- metals such as copper or gold (Fig. 19-6).
nium, to aluminium and its alloys, and to It is of interest, therefore, to know under
nickel alloys containing chromium used in what conditions reactions such as
vacua or low oxygen activity gaseous en-
3TiO + 2Al (19-9)
vironments. These reactive components
cause formation of reaction product lay- can proceed. If the product is assumed to
ers, whose lateral growth determines the be the stoichiometric oxide then the free
speed at which the braze front can ad- energy of the reaction, AGR, can be calcu-
vance. Thus the flow of the liquid can be lated from
controlled by slow solid state diffusion
processes and equations (19-3) and (19-4) F(A1 2 O 3 )-
describing capillary penetration and kinet-
LTiO^AlM
ics are usually not of direct relevance. = -RT\n
}aTi3 • a Al 2 0 3 J
Wetting Induced by Titanium (19-10)
Titanium forms a family of oxides as
illustrated in Fig. 19-5 of which TiO is of The activity of the various components
most interest to the brazing fabricator. is designated by the *d subscripts. Assum-
This oxide is very stable and exhibits some ing that of the oxides to be unity, forma-
metallic characteristics. Especially as a hy- tion of TiO will occur if the second brack-

Higher oxides
1200

1000

p 1200

1000

TiN
1200

Figure 19-5. Phases formed by

1000 titanium with oxygen, carbon and
nitrogen.
20 40 60
Atom % Titanium
274 19 Joining of Ceramics

160r These expectations have been confirmed

by laboratory studies and exploited in
U) practice. Titanium is less soluble in silver
120 than copper, and the levels of titanium
cn
<v needed to induce wetting of alumina and
other oxides are decreased if copper is re-
en Oxides
placed by silver-copper solvents. The sol-
80
c L Carbides ubility of titanium in tin and indium is also
o
low and improvements can be gained by
a replacing copper with copper-tin or cop-
o
per-indium solvents (Nicholas et al.,
o 1980).
While reference has been made so far
i I
30 40 50 60 70
only to the wettability of alumina, similar
effects are observed with many other ce-
T i t a n i u m (at. %)
ramic substrates. Brazes based on titanium
Figure 19-6. The wettability of titanium carbides and additions to copper or silver-copper al-
oxides by copper.
loys are effective when used with zirconia,
magnesia, silica and other oxide ceramics
(Naidich, 1981). They are also effective
eted activity ratio is below 4 x 1 0 at when used with nonoxide ceramics al-
1400 K. If the ratio rises to a higher level, though the beneficial reactions involved
(i.e., if titanium is consumed so that Ti falls are different.
and Al rises) the reaction between the Exposure of carbon and many carbides
braze and the ceramic will not cease but to brazes containing titanium leads to the
Ti 2 O 3 and other titanium lean oxides will formation of TiC, and generally hypostoi-
be produced which are not readily wet- chiometric TiC. Ttanium forms only one
table. To ensure that the bracketed activity compound with carbon, but its range of
ratio is less than 4 x 10 ~ 4 , a copper braze stoichiometry is very wide: TiC 0 4 8 is sta-
must contain more than 1 % of titanium. ble at 1400 K (Fig. 19-5). Further, the hy-
Levels for other solvents can be higher: at postoichiometric compound is wettable by
least 4% for gold and at least 10% for copper if the C/Ti ratio is less than 0.65
nickel at 1800 K. These predictions are in (Fig. 19-6), and simple thermodynamic
fair accord with the practical observations calculations show that the titanium con-
that several per cent of titanium are needed centration of, for example 8 % in copper,
to induce wetting of alumina by copper needed to induce wetting of carbon is also
and substantial amounts are needed to in- that needed to stabilise TiC 0 6 5 . The addi-
duce wetting by nickel (Naidich, 1981). tion of tin can decrease the level of tita-
However, alloys containing significant nium needed to induce wetting of carbon
concentrations of titanium can form thick to only 1 %, and similar effects have been
reaction product layers. The ideal solvent seen with carbide ceramics.
would be one in which the titanium had a Titanium also forms nitrides and TiN
very high activity coefficient (activity/con- has a wide range of hypostoichiometry,
centration) and a low solubility so there Fig. 19-5. It is not known if these hypo-
was no substantial reservoir of reactant. stoichiometric variants are better wetted as
 19.5 Brazing 275

are those of TiC and TiO, but the brazing 1988). Technological procedures for de-
of silicon nitride, aluminium nitride and creasing the wetting temperature include
other ceramics can be accomplished with using alloys containing silicon which en-
brazes containing titanium (Nicholas et al, hances the liquid fluidity or magnesium
1990) so the principles involved may be the whose vaporisation disrupts the surface
same. oxide. Even simpler is the resort of placing
the aluminium braze foil between work-
pieces and applying a pressure to disrupt
Wetting Induced by Other Reactive Metals
the encasing oxide.
Nickel chromium alloys have also been Why molten aluminium should wet ce-
used to braze engineering ceramics such as ramics is not clear, and simple approaches
silicon nitride and silicon carbide that have may be deceptive. Thus aluminium nitride
to meet demanding service conditions. Like is far more stable than silicon nitride and
gold and palladium alloys, the nickel- hence it was presumed that wetting of sili-
chromium brazes have higher melting tem- con nitride was accompanied if not in-
peratures and better oxidation resistances, duced, by the formation of aluminium ni-
than those based on the silver-copper tride. This view held for some time because
eutectic. The actual wetting processes for of the difficulties of characterising the thin
these systems involves dissolution of the ce- reaction product layers. Ultimately, TEM
ramic and the beneficial role of chromium studies were conducted and these showed
seems to be to diminish the extent of disso- that the layers were sialon and amorphous
lution. aluminium - silicon - oxygen compounds
The more widely adopted competitor (Ning e t a l , 1987). Clearly the course of
for alloys containing titanium is alumin- the reaction was influenced by the alumina
ium which has been found to wet a wide film present on the metal surface. Our un-
range of ceramics at temperatures of about derstanding of aluminium-induced wetting
900 °C and above in technical vacua. Vir- behaviour, therefore, is a topic that still
tually identical threshold temperatures requires much study.
have been observed in sessile drop experi-
ments using oxide, carbide and nitride sub-
strate and this commonality is not due to 19.5.2.2 Practice
some remarkable coincidence of thermo- Well characterised wetting behaviour
dynamic data, but to the fact that wetting has been described for a wide range of re-
at lower temperatures is prevented by the active brazes that contain titanium and
presence of a mechanically tenacious oxide some of the commercially available materi-
film on the surface of the molten alu- als are listed in Table 19-3. Particular at-
minium. Thermal and mechanical agita- tention has focused on the Ag-27Cu-2Ti
tion will cause this film to rupture from and Ag - 24 Cu -15 In -1.5 Ti compositions
time to time but the oxygen activity in the which can be produced by casting and
environment is sufficient to cause instanta- working to produce sheet or wire, or by
neous healing. Thus laboratory studies rapid solidification processes to produce
have shown that wetting of alumina by thin sheets directly from the melt. The
sessile drops of aluminium can occur at melting temperature of the Ag-28Cu eu-
700 °C if the vacuum used is improved tectic is 780 °C, and the use temperatures
from 10" 5 to 10" 7 mbar (Chatain et al., of the reactive brazes are relatively low,
276 19 Joining of Ceramics

Table 19-3. Some reactive metal brazes. surface roughness affects wettability, de-
Alloy Temperature (°C) grading that of most systems but possibly
enhancing that of very well wetting sys-
Solidus Liquidus tems. Ceramic surfaces often have to be
Ag-4Ti 970 970
ground to meet engineering requirements
Ag-35Cu-1.5Ti 770 810 for size and flatness, and while grinding
Ag-27Cu-3Ti 780 805 can produce a desired roughness, care
Ag-27Cu-2Ti 780 795 must be taken that it does not nucleate
Ag-lIn-lTi 950 960 microcracks that will degrade the strength
Ag-20Cu-5In-3Ti 730 760
605 715
of the joint. For siliceous oxides, these
Ag-23Cu-15In-1.5Ti
microcracks can be healed by firing at high
temperatures in air.

820-860°C for Ag-27 Cu-2Ti and 750- Environment

800°C for Ag-24 Cu-15 In-1.5Ti.
The commercial concentration on these The environments used during brazing
two braze compositions in part reflects the should not compete with the ceramic for
present small size of the market, but is also the reactive components because depletion
a testimony to the excellent wetting be- will negate or diminish the effectiveness of
haviour of these brazes which have been the braze. In practice, brazing in evacuated
used successfully with carbon, many car- furnaces requires pressures of less than
bides including SiC, oxides such as A12O3, 10" 3 mbar and preferably less than
ZrO 2 and BeO, and nitrides, including 10" 4 mbar. The threshold at which braz-
A1N and Si 3 N 4 . Achieving this success, ing becomes unreliable depends on many
however, has required close attention to be factors including the surface area to be
paid to processing parameters (Mizuhara brazed, the furnace volume, leak rate and
etal., 1989). Obviously these need to be pumping rate. Leak tightness is most im-
optimised for each application, but certain portant because the oxygen and nitrogen
procedures are common. partial pressures of a leaky furnace will be
far higher than those of one that is tight,
even if the total pressure is the same.
Jigging
Gas environments can be used, and have
Because of the slow spreading rates of to be used with conveyor belt processes. It
reactive metal brazes it is usually impracti- is usually necessary however to use a
cal to place them at the end of a long cap- slightly reducing atmosphere rather than
illary gap and then to depend on flow to one that is nominally inert because the
produce a joint. The common practice is to oxygen activity of a high purity gas can
preplace a foil or to use a short capillary exceed that of even a modest vacuum. Fur-
length. ther it should be noted that nitrogen is not
an inert gas when reactive metal alloys are
Surface Preparation being used.
Cleaning the surfaces of both ceramic
Thermal Cycle
and metal components is usually necessary
for the reliable achievement of successful This can be controlled closely by the
brazing. There is also ample evidence that fabricator and hence particular attention
 19.6 Diffusion-Bonding 277

can be given to its optimisation. Reactive Table 19-4. Normal process parameter ranges used
metal brazes are generally based on eutec- to diffusion bond ceramics.
tic solvents but are not themselves eutec- Process parameter Range
tics. Therefore care has to be taken to
avoid liquation effects. Common practice Pressure 10-100 MPa
Time 100-10000 s
is to heat slowly, at about 10 K min~ * to a Temperature 0.70-0.98 TMa
temperature just below the alloy solidus Surface roughness Ra<1.5 urn
and then dwell to achieve a uniform tem- Environment Vacuum, pressure <10~ 4 mbar
perature throughout the workpiece. This Inert gas, impurities < 50 ppm
dwell is followed by even slower heating, at a
r M is the melting temperature in kelvin. For dissim-
perhaps SKmin" 1 , through the solidus - ilar material bonding, r M refers to the least refractory
liquidus range to the brazing temperature. material.
Dwell times at this temperature are rela-
tively long, 10-30 minutes, to take ac-
count of the slow spreading rates of the across the contacting surfaces is required
brazes. to achieve a permanent bond. Workpieces
can be diffusion bonded directly to each
Titanium Hydride Coating other or via a metal foil inserted between
them, this latter process creating a joint
Although a coating process, this prac- structurally similar to those formed by
tice can be viewed as a variant of reactive brazing. Typical conditions used during
metal brazing in which the titanium is pre- diffusion bonding are summarised in
placed on the ceramic surface rather than Table 19-4.
having to migrate. The use of the hydride
involves painting it on the ceramic surface 19.6.1 Principles and Process Parameters
with a fugitive binder, and using a conven-
tional braze, usually the silver-copper eu- 19.6.1.1 Physical Factors
tectic. The thermal cycle is then altered to The basic driving force for diffusion
introduce a slow ramp or dwell in an evac- bonding is the minimisation of the surface
uated furnace within the range 350- energies of the contacting bodies. The en-
500 °C, during which the hydride decom- ergy released when an interface is formed
poses to deposit a thin layer of titanium. by diffusion bonding is

19.6 Diffusion-Bonding Sx and S2 refer to the two solids and S ^

to their interface. However, the mecha-
Diffusion-bonding is a solid state join- nisms that must operate if that energy is to
ing process in which contact between the be released are complex.
surfaces to be bonded is achieved initially For a metal interlayer or workpiece be-
by the application of pressure at high sub- ing diffusion bonded to a rigid ceramic,
solidus temperatures and then grows by several stages of increasing contact and
diffusional processes to minimise the sur- bonding can be visualised as illustrated
face energy of the system. Contact can be schematically in Figure 19-7. Definitive
equated to bonding in some systems but in work by Wallach and his co-workers
others a second step of interdiffusion (Derby and Wallach, 1984) has modelled
278 19 Joining of Ceramics

seven processes that can contribute to the

Metal growth of contact between similar metal
workpieces. These are: (1) plastic yielding
(A) to deform contacting asperities; (2) surface
diffusion from a surface source; (3) volume
Ceramic diffusion from a surface source; (4) vapour
phase transport; (5) grain boundary diffu-
sion from an interfacial source; (6) volume
diffusion from an interfacial source; (7)
power-law creep.
The importance of each of these mecha-
nisms varies as the material and bonding
conditions are altered, but they should be
relevant to ceramic-ceramic bonding. The
mechanisms should be relevant also to ce-
ramic-metal bonding providing mecha-
nism (5) is defined as referring to interfa-
cial diffusion, and providing the micro-
structure of the interface is not altered by
(C) the growth of a reaction product layer. The
potential relevance of these mechanisms
immediately identifies a number of process
parameters that should affect the ease of
diffusion bonding. These include tempera-
ture, time, pressure, surface finish, and en-
vironment. When metal interlayers are
used to promote bonding, their interlayer
thickness, yield strength, grain size and
chemistry also can be of importance. All
these parameters, and surface cleanliness
which was assumed to be absolute for the
Reaction layer models, have been found to be important
in practice.

Temperature
Diffusional and creep processes are very
slow at temperatures below about 0.5 TM,
Figure 19-7. Schematic illustration of the changes oc- where TM is the melting temperature in K,
curring during ceramic-metal diffusion bonding. and it is unusual to diffusion bond ce-
The initial contact is increased by the application of ramic-ceramic or ceramic-metal joints at
pressure and holding at a high temperature (B), (C) less than 0.7 TM. Thus the thresholds for
and (D). When contact is achieved, bonding across
the interface can occur (D) and is often associated diffusion-bonding copper and nickel to
with the growth of a reaction product (E). The alumina are 675 and 900 °C. Once these
hatched upper workpiece is the metal. threshold temperatures have been ex-
 19.6 Diffusion-Bonding 279

ceeded, the strengths of the bonds improve diffusion bonding workpieces, and large
rapidly and this effect of increasing tem- flashes of extruded material can be pro-
perature has been related to a dominant duced if very high pressures are applied
influence of a specific growth process for when metal interlayers are used.
some systems, for example to grain
boundary diffusion in the metal for the Surface Finish
bonding of alumina to aluminium (Derby,
Two aspects of surface finish are of prac-
1987).
tical importance, roughness and flatness,
but it is important that mechanical pro-
Time
cesses used to achieve particular surface
This parameter appears in the relation- finishes on ceramic workpieces do not
ships describing the diffusion, creep and cause microcracking. Theory predicts that
vapour phase transport processes opera- smooth flat surfaces are the most easily
tional during diffusion bonding. However, bonded, and the surface finish processes
the power terms of these relationships are such as grinding and lapping are used to
such that fabricators do not have to oper- produce ceramic surfaces that are smooth
ate within very precise time constraints. except for a few residual pits. For ceramics
Ceramic parts usually have to be heated with glassy binder phases, residual pits can
and cooled relatively slowly to avoid ther- be sealed after grinding by a high tempera-
mal shocks, and hence it is common prac- ture firing.
tice also to dwell at the bonding tempera- Surfaces of metal workpieces to which
ture for at least some minutes or even for the ceramics may be bonded are usually
several hours. machined to a fine finish or even polished.
When workpieces are bonded directly, it is
Pressure important that the bonding surfaces are
flat so macroscopic deformation to achieve
This parameter is of major importance
conformity is minimised. If metal interlay-
in determining the initial contact between
ers are used to promote diffusion bonding,
the asperities of the bonding surfaces. Dif-
their surfaces are usually abraded or
fusion bonding is usually required to be a
etched to produce fresh clean surfaces
near nett shape forming process so the
rather than a particular finish.
pressures applied in direct workpiece -
workpiece bonding are kept low to avoid
Environment
macroscopic deformation. However, when
using metal interlayers to facilitate bond- Relatively little systematic attention has
ing it is standard practice to apply pres- been paid so far to the environments used
sures that ensure the interlayers conform when diffusion bonding. Ceramic-ce-
quickly to the contours of the workpiece ramic bonding can and has been achieved
surfaces, minimising the processing time successfully in inert atmospheres and even
needed for diffusional growth of contact. in air but bonding to metal workpiece or
There are, however, upper useful limits to with metal interlayers usually requires the
the pressures that can be applied. Surface use of inert atmospheres or vacua. In fact
and subsurface cracking of ceramics can be vacuum should be the best environment
initiated at asperity contact points if very since the healing of interfacial porosity will
high pressures are applied when directly not be impeded by trapped gas.
280 19 Joining of Ceramics

19.6.1.2 Chemical Effects diffusion bonding of nickel-chromium al-

loys to silicon nitride results in an interfa-
Few models of diffusion bonds take into cial reaction product layer of chromium
account the chemical changes that can oc- nitride, but the principal benefit conferred
cur at dissimilar material interfaces even by the chromium is that it prevents the
though the effects of these changes on joint harmful formation of brittle nickel silicide.
properties can be substantial. As in the These reactions follow rather than pre-
cases of brazing and glazing, chemical cede the creation of bonded interfaces.
changes at interfaces can lead to the for- Thus microstructures can be adjusted as a
mation of reaction product layers, the separate postfabrication step. Figure 19-8
growth of which often has initially bene- shows the effect of postfabrication heat
ficial effects on joint strengths but the ulti- treatment on the strength of joints between
mate result is detrimental. While chemi- alumina and an Al-4.5Mg alloy which re-
cally reactive systems are often employed act to form an alumina-magnesia spinel.
deliberately when brazing or glazing, this The effect is initially beneficial but ulti-
is not common practice when diffusion mately detrimental, and the optimum heat
bonding. However, circumstances can oc- treatment times vary with temperature in
cur in which reactions play important accord with the activation energy for the
roles. Three main classes of such reactions diffusion of magnesium in aluminium.
have been observed. This figure also demonstrates that atten-
tion must be paid to possible effects of con-
Dissolution Processes tinued reaction product layer growth at
high service temperatures. When metal
There is a thermodynamic driving force
workpieces are joined to ceramics using
for the dissolution of oxide, and other, ce-
metal interlayers, metal-metal reactions
ramics in pure metals. This has been ob-
must be considered and the growth of
served during the bonding of alumina to
brittle intermetallic layers has been found
niobium at high temperatures in good
vacua and results in the precipitation of
alumina within the metal on cooling. Such
dissolution processes may also account for 140
the good bonding of oxides to platinum 403 °C
120 500°C
group metals (Bailey, 1980). Why such dis-
solution processes are beneficial is not 100 600 °C
clear, but it could be related to the de-
creases in interfacial energies suggested 80
to be of importance during glass-metal S 60
bonding.
40
Initial strength
Interfacial Reactions 20

Little systematic work has been done 0

0 1 2 3
until recently to develop the principles for Log( Time in hours)
applying metal-ceramic reactions to pro- Figure 19-8. The effect of post fabrication heat treat-
mote diffusion bonding and classifying re- ment on the strengths of diffusion-bonded joints be-
action effects can be deceptive. Thus the tween alumina and an Al-4.5Mg alloy.
 19.6 Diffusion-Bonding 281

to limit the strengths of joints between pansion such as niobium, titanium and
steel and alumina bonded using aluminium special ferrous alloys.
interlayer. The interlayer materials are almost in-
variably face centred cubic metals: alu-
Environment Induced Reactions minium, copper, gold, nickel and silver.
Laboratory work on the use of alloys has
Oxidising environments can enhance the
not yet been translated to practice, al-
strengths of interfaces formed by metals
though some attention has been paid to the
with oxide ceramics and this effect has
use of clad sheets in near production devel-
been exploited to promote bonding of a
opment studies. An example is the use of
wide range of systems (de Bruin et al.,
aluminium sheet clad with an aluminium-
1972). It was assumed at first that bonding
silicon-magnesium braze alloy (Yamada
was promoted by the formation of com-
etal., 1989). This permits relatively thick
plex oxides such as spinels that bridged the
interlayer joints to be produced by bond-
interfaces but more recent work shows this
ing using low pressures at temperatures be-
not to be necessary for at least the copper-
tween the melting temperatures of alu-
alumina (Ambrose etal., 1993) and nio-
minium and the braze.
bium-alumina (Ruhle etal., 1987) sys-
tems.
Surface Preparation
19.6.2 Practice The basic requirements for cleanliness
The industrial application of diffusion- and relative smoothness are similar to
bonding for joining ceramics is so far those for brazing. However, even greater
rather limited, but some practices can be care has to be taken to ensure the flatness
identified. of the workpiece surfaces because large in-
terfacial voids will be left during direct dif-
Jigging fusion bonding which cannot be sealed.
Even when interlayers are used, care has to
Jigging practices follow those used in be taken to ensure flatness, variations of
brazing when metal interlayers are em- more than about 10% of the final inter-
ployed to promote bonding. In most cases, layer thickness can lead to unreliable joint
pressure is applied uniaxially so attention strengths and even ceramic workpiece
has to be paid to the maintenance of align- damage when high pressures are used.
ment when the interlayers deform. There
is, however, an increasing interest in the Pressure
use of hot isostatic pressing as a technique
permitting a greater throughput of small The pressures used in practice are usu-
components. ally relatively high, and are applied at the
bonding temperature to ensure rapid and
Materials good contact between the workpieces and
metal interlayers.
The workpiece materials used are dic-
tated by the end applications but these
Environment
have so far been usually oxide ceramics
such as alumina or quartz and metals with The preferred environments are vacuum
near matching coefficients of thermal ex- or high purity inert gases, and no major
282 19 Joining of Ceramics

production use of a reducing gas atmo- strains of up to 0.05% do not pose major
sphere has yet emerged. Similarly the re- problems, but strains of 0.1 % or more are
cent use of oxidising environments to con- very difficult to accommodate even when
vert metal interlayers to complex oxide the componets are small.
bonds remains so far within the labora- Glazed joint designs are either seals with
tory. matched or balanced stress levels in the
glass or seals with unbalanced stress levels
Thermal Cycle such as compression or Housekeeper de-
signs. Producing the first type requires se-
The common practice is to use high tem-
lection of materials with closely matching
peratures when employing aluminium in-
coefficients of thermal expansion. A wide
terlayers to promote bonding, 550-600°C
range of seal designs could be used in prin-
corresponding to 0.88-0.93 TM. When us-
ciple, but in practice they are often sym-
ing other interlayers such as gold or nickel,
metrical. An example of such a seal is an
relatively lower temperatures are used,
electrical feed through using a Kovar pin
typical values lying in the range 0.7 to
glazed within an alumina or glass-ceramic
0.85 r M , but these sometimes reflect equip-
sleeve.
ment capabilities rather than optimised
material practice. Mismatched compression seals almost
invariably have concentric designs, the
As with brazing, relatively slow heating
higher expansion workpiece being the
and cooling rates are used to avoid thermal
sleeve so that the glaze is in compression.
shocking of the ceramics, and dwells of
This type of seal is often used to manufac-
10-100 min at the bonding temperature
ture electrical feed throughs. The most
are used commonly. These rates and dwells
widely recognised type of mismatched seal,
result in a slow throughput of diffusion
however, is that whose design was system-
bonded components because there is as yet
atised by Housekeeper (1923). The original
no practice employing a conveyor belt fur-
work related to the joining of glass work-
nace.
pieces to copper which has a high expan-
sion coefficient of about 16xlO" 6 K~ 1 <
As shown in Fig. 19-9, the joint was ef-
19.7 Joint Design fected by using a thinned copper tube
which was sealed directly to a glass tube,
The overall structures of ceramic-
the ductility of the copper permitting the
ceramic and ceramic-metal joints are de-
strains generated by the mismatched coef-
termined by the shapes and functions of
ficients to be accommodated without un-
the component. However, the joining pro-
due stress. This practice has been very suc-
cesses also impose restrictions.
cessful and is now used for other metal-
glass joints.
19.7.1 Glazed Joints
Glazed ceramic-ceramic joints gener-
19.7.2 Brazed Joints
ally have simple butt or sleeve configura-
tions, but accommodating the often mis- When joint designs are dictated by the
matched coefficients of expansion of constraints of component shape and con-
ceramic-metal systems can require con- figuration, the fabricator can make an in-
siderable skill. Differences in workpiece put only through her or his selection of
 19.7 Joint Design 283

Copper tube Thinned edge

Glass bead

Glass tubing
sealed to bead

Figure 19-9. Housekeeper seal.

joining material. If the ceramic surface is offers at least three advantages. A reactive
coated with a metal, conventional brazes metal braze foil can be placed readily in the
can be used and designs can be employed metal recess, a sound joint can be ensured
that depend on substantial capillary flow. by using a small load to drive the ceramic
These are typified by sleeve joints, but it is into the recess and the progressive thinning
still generally preferable to use thin outer of the recess wall produces a gentle com-
sleeves of a low yield stress metal so that pressive stress gradient in the ceramic.
contractional stresses are minimised.
(While the stress on the inner ceramic body Compliant Joints
will be compressive within the sleeved
This design can be exemplified by the
length, tensile shear stresses will be pro-
brazing of a thin metal cap to a thick walled
duced where the sleeving ends.)
ceramic tube illustrated in Fig. 19-10 b.
Of more challenge at present is design-
The thinness of the metal cap and its duc-
ing joints that are to be bonded using reac- tility means that mismatched contractional
tive metal brazes. Some possible solutions strains can be accommodated without
are illustrated schematically in Fig. 19-10 gross distortion and stressing in the braze
and can be described under three headings. region. A recent innovative variant of this
type of design is shown in Fig. 19-11 c
Compressive Joints which permitted a ceramic cap to be
This design uses an outer metal sleeve brazed to the nickel alloy shank of an
that contracts more than the ceramic. One automobile engine tappet (Bucklow et al.,
example is illustrated in Fig. 19-10 a in 1992). A more drastic variant is the use of
which the ceramic is tapered and the metal interlayers to produce, for example, a sili-
shank tapers to a knife edge. This design con nitride/braze/molybdenum/braze/cast
284 19 Joining of Ceramics

in Fig. 19-10d. The effects of the balancing

(a) piece are to increase the bonding area and
to change the stress distribution in the
metal so as to diminish the likelihood of
peeling being initiated at the edge of the
cap.

19.7.3 Diffusion-Bonded Joints

There has not yet been sufficient indus-
(b)
trial use of diffusion bonded ceramics for
specific families of designs to be identified
readily. In general the designs used are pat-
terned on those of brazed joints with the
additional constraint that the bonding sur-
faces must be in compression when the joint
is being made, thus favouring butt config-
urations. Individual laboratory workers
(c) and fabricators using metal interlayers to
promote bonding have used designs simi-
lar to some of those found to be effective
when reactive metal brazes are employed.
Thus tapered sleeve joints have been used
to bond quartz windows to steel flanges
n and many workers have evaluated the use
of inserts such as molybdenum when try-
(d) ing to form compliant joints. (In attempt-
ing to optimise the performance of such
joints, comment will be made later on the
Figure 19-10. Possible designs of reactive metal brazed
joints.
beneficial effects of stress relief anneals.)
Finally, the use of a balanced joint design
was found to be beneficial in the fabrica-
iron joint. This sequence is advantageous tion of an accelerator module consisting
because the relative similarity of the silicon essentially of stacks of interleaved alumina
nitride and molybdenum coefficients of rings and titanium annuli bonded together
thermal expansion shifts the region of using aluminium foil interlayers (Joy,
maximum contractional strain to the rela- 1987).
tively ductile molybdenum/braze/cast iron
side of the joint.
19.8 Joint Evaluation
Balanced Joints
and Properties
When brazing metal end caps to ceramic
tubes, it can be practical to join the cap There is a considerable scientific interest
face directly to the ceramic body if a ce- in the structure, chemistry and energies
ramic balancing piece is used as illustrated of ceramic interfaces and this has grown
 19.8 Joint Evaluation and Properties 285

with the development of instruments with they are sensitive to the presence of flaws
atomic scale resolutions. The properties of such as microcracks and unbonded regions
ceramic joints that are of interest to the as small as 1-10 |im. Changing the distri-
fabricator and user, however, relate to in- bution of these flaws, for example by alter-
service performance and are usually speci- ing the sample size, will affect strength val-
fied in terms of the technological parame- ues. In contrast, fracture toughness is a
ters of leak tightness, electrical insulation measure of mechanical performance that
and strength or toughness. takes account of flaws and hence can be
Leak tightness can be a property of di- used to characterise mechanical properties
rect technical relevance, being essential for (Elssner, 1989). At present, most mechani-
the successful performance of vacuum fit- cal property data are generated by one of
tings such as observation ports and electri- four techniques.
cal feed throughs, but it is also of more
general relevance as a measure of the Shear Tests
soundness of a joint. While a leaky glazed
The loads needed to separate work-
joint is almost certainly very weak, this is
pieces by shearing are measured for a vari-
not necessarily the case for a joint diffu-
ety of configurations. Some of these are
sion bonded using metal interlayers.
sketched in Fig. 19-11. While yielding fail-
A leak, of course, requires a continuous,
ure loads that can be used to rank the be-
unbonded region stretching right across a
haviour of a particular set of samples,
joint. Isolated unbonded regions will not
great care in gripping is needed to ensure
be detected, but they can affect electrical
that the stresses on the joints are pure
insulation. This characteristic can be of
shear.
prime importance for high voltage compo-
nents, but breakdowns in electrical insula-
Tensile Tests
tion, like leaks, can also be used to gain
structural insight. A similar family of tests was once used
The overwhelmingly specified and mea- to assess the tensile strengths of joints, but
sured characteristics of ceramic joints, how- this work was later modified to produce an
ever, are strength or toughness or both. ASTM standard (ASTM F19-64), illus-
While there is a demand for information, trated in Fig. 19-11, which was designed to
providing it is not simple and has led to assess the strengths and leak tightnesses of
vigorous technique development pro- joints produced by alumina that had been
grammes in recent years. molybdenum manganese coated and then
brazed. The test has since been widely used
to assess the strengths of ceramics joined
19.8.1 Techniques for Strength by reactive metal brazing and by diffusion
and Toughness Measurements bonding using a metal interlayer. In some
The techniques used to assess the me- work, it has also been used to assess joints
chanical properties of ceramic joints at between ceramic and metal workpieces by
first measured failure loads, then strengths using the metal workpiece as a tubular in-
and are now changing to measure fracture sert between the ceramic. So far the use of
toughness. Strength measurements are the test has been restricted mainly to the
generally simple and give a direct charac- alumina ceramics for which it was first de-
terisation of the samples tested. However signed.
286 19 Joining of Ceramics

(a) (b) (c)

loading
arms Figure 19-11. Some strength
and toughness test sample
configurations. Clockwise
from the top left, an
(d)
ASTM test piece (a), a
double cantilever bend test
(b), a "push off test (c),
i four (d) and three (e) point
bend tests, a push out
shear test (f), a two high
shear test (g).

(e)
(g)

Bend Tests Notched Bend Tests

Both three point and four point bend Three and four point bend tests with
tests are used to assess the strengths of preformed notches or cracks are used to
ceramic-ceramic and ceramic-metal joints derive fracture energy and, plane strain,
at ambient and elevated temperatures. fracture toughness values. Load-displace-
Standards akin to those for tensile test- ment curves generated by notched samples
ing have yet to be fully formulated for subjected to bend testing can be used mea-
ceramic-metal samples, but it is common sure the energy needed to propagate a sin-
practice to use symmetrical samples with gle sharp crack at a ceramic-metal inter-
rectangular cross sections. The joint is lo- face, and to derive a value for G, the work
cated in the middle of the samples which needed to initiate catastrophic failure ex-
are typically 50 mm x 6 mm x 3 mm. Par- pressed as J m" 2 . This energy is then com-
ticular care has to be taken over the prepa- pared to that for the monolithic ceramic to
ration of the test piece surface that will be characterise joint quality.
in tension, and this is usually part of the These tests can be used also to derive Klc
process of machining test pieces from and Kc values, the fracture resistance of
larger bonded blocks. monolithic ceramics and of ceramic inter-
These tests enable flexural tensile faces, which are measured in units of
strengths to be calculated from the failure MPa m 1/2 , from the failure load and sam-
loads and test piece dimensions. Because ple dimensions adjusted by correction
the test pieces are small, it is common prac- functions dependant on the failure made
tice to produce a batch from a single and test piece geometry. A particularly dif-
bonded sample and to use statistical analy- ficult term to specify can be the original
sis such as the derivation of a Weibull length of the crack that propagated, and
modulus to characterise the mechanical hence development studies have examined
properties. methods such as double cantilever beam
 19.8 Joint Evaluation and Properties 287

testing that do not require crack length MPa

100
data in order to generate plane strain frac-
ture toughness values. (Moorhead and Ti - 6AI - 4V
Becher, 1987). 2014 Al
Nb
Push-off Tests WC - 6Co 10
HPSi 3 N 4 /Hf,Zr
While not a test of joint strength, many HP - Si3N4 AI2O3/AgCu/AI2O3
SiC
workers developing the use of reactive PS - ZrO9
metal brazes for ceramics have employed AI2O3/AIMg3
the semiempirical "push-off" test sketched
in Fig. 19-11. The tensile strength of the
AI2O3/AI
metal-ceramic interface can be calculated Sodalime glass
from the contact area, the drop geometry Epoxy
and the load that has to be applied to the
equator of a nonwetting drop to cause fail-
ure. -j- 0.1
Figure 19-12. Comparison of the fracture toughnesses
19.8.2 Strength and Toughness Values of some monolithic ceramics and ceramic-metal
joints.
The measured values for the strengths of
brazed or diffusion bonded ceramic joints
depend very strongly on the type of test bonding of structural silicon nitride com-
used and hence the area stressed. Bend ponents, and are generally monitored by
tests, and the push-off test, stress areas of performance specific tests. In contrast,
a few mm 2 and yield tensile strengths typi- strength testing has been used extensively
cally of 300-800 MPa for high integrity to optimise both the materials used and the
joints. In contrast, an area of over 100 mm2 process parameters selected when brazing
is stressed in the ASTM test and tensile or diffusion bonding. These data have
strength values of 50-100 MPa character- been the foundations on which earlier
ise good quality joints. Similar low num- comments about optimum joining proce-
bers are generated by shear tests in which dures have been made. Their relevance can
the area stressed is perhaps 50 mm 2 . be illustrated by reference to three exam-
Toughness and fracture energy values are ples.
even smaller, typically 1-10 MPa m 1/2 and
5-20 Jm~ 2 for high integrity joints and are
Inter facial Reactions
comparable to those of the workpiece ma-
terials as illustrated in Fig. 19-12. These There is considerable evidence that the
values can be achieved with optimised nucleation and growth of a reaction prod-
joints but most reports of mechanical uct at ceramic-metal interfaces initially
properties have related to poorer joints causes a strengthening and toughening.
and have been part of the optimisation Figure 19-13 (Cho and Yu, 1992) is an il-
process for joining processes. lustration of the effect for a reactive metal
The requirements for glazed joints to be brazed ceramic. This effect is generally ac-
strong or tough are generally modest, al- cepted and can be related to the beneficial
though there are exceptions such as the effects that the approach to chemical equi-
288 19 Joining of Ceramics

6.0 -
400 -

4.0 - CD

c
e
-a 200
2.0 -

0.0
u.O 0.2 0.4 0.6 0.8 1.0 0 200 400 600
Fractional area reacted Free energy of formation (kJ/g mol)
Figure 19-13. The effect of the extent of reacted area Figure 19-14. Maximum strengths of interfaces
on the toughness of a reactively brazed alumina joint formed by alumina with sessile drops of nickel al-
(Cho and Yu, 1992). loyed with reactive metals (Crispin and Nicholas,
1976).

librium has on the integrity of glass-metal even be increased at modest service tem-
interfaces. The strengthening and toughen- peratures due to relaxation of stresses gen-
ing continues until the interface is com- erated during cooling after fabrication.
pletely converted, but thereafter can de- An example of these effects is shown in
crease due to volume mismatch strains. Fig. 19-15, which also illustrates effects of
Thus there is a peak interfacial strength, joint thickness.
and this has been related to chemical reac-
tivity for braze systems as illustrated in
Fig. 19-14. It follows from these micro-
structural effects, that process parameters
such as the concentration of a reactive
component, the bonding temperature and
the bonding time influence the mechanical
properties of joints.

Service Temperatures
Most strength testing is done at room
temperature, but many projected service
conditions for structural ceramics involve
high temperatures. The effect of these on
metals used to braze or diffusion bond
could be considerable, but in practice the 200 400 600
Test temperature (°C)
effects of metal softening are mitigated by
Figure 19-15. The effects of test temperature and
the thinness of the joints which constrain
joint thickness on the strengths of ASTM test pieces
yield. In fact, joint strengths can be main- of alumina brazed with aluminum (Iseki and
tained up to high temperatures, and may Nicholas, 1979).
 19.8 Joint Evaluation and Properties 289

4001
Mismatched Contractions
The differing contraction characteristics
of most desired combinations of ceramic
and metal workpieces can have major ef-
fects on joint strengths. Ways of mitigating
these effects were suggested earlier when
joint designs were discussed but several
Sialon
workers have illustrated the magnitude of 100 Silicon nitride
the problem by using one design for a spec- Silicon carbide
trum of material combinations. As illus-
trated in Fig. 19-16, joint strengths decrease 5 10 1 5
progressively as contractional characteris- Expansivity mismatch (10 6 K 1 )
tics become more mismatched and the ce- Figure 19-16. The effects of mismatches in coeffi-
ramic-metal interface is more severely cients of thermal expansion on ceramics diffusion
stressed. It is possible, however, to de- bonded with Al-lOSi (Yamada et al, 1989).
crease these stresses by low temperature
annealing and strength values can be in- 125
creased substantial as shown in Fig. 19-17.
However, as the annealing temperature is 100
raised, so strengths again decrease due Q_
once more to mismatched contractional
strains generated during cooling.

19.8.3 Nondestructive Evaluation

While measurement of a high strength

or toughness value demonstrates the me-
chanical quality of a joint, it also destroys 150 300 450 600
Annealing temperature (°C)
the joint. There is therefore a need for non-
destructive evaluation techniques. Leak Figure 19-17. The effect of postfabrication annealing
testing is a nondestructive technique, but on the tensile strength of alumina diffusion bonded
with aluminum or copper interlayers.
those usually referred to by this term are
ultrasonic and magnetic testing and X-ra-
diography. The application of such nonde- The frequency, velocity, attenuation,
structive evaluation techniques to ceramic transmission and reflection of ultrasonic
interfaces poses severe challenges because waves can all be affected by microstruc-
of the notch sensitivity of ceramics for tural effects. However there are inherent
which critical flaw sizes are usually less limitations to the applicability of ultrason-
than 100 |im as compared to several mm ics. Thus detection of flaws smaller than
for metals. These challenges are being ad- 100 |im in ceramics requires the use of fre-
dressed, and a number of techniques show quencies in the range of 100-200 MHz
promise. Those of particular note are ul- which rapidly become attenuated so that
trasonic testing, X-radiography and ther- only near-surface defects are revealed. The
mography. requirement for high frequencies to detect
290 19 Joining of Ceramics

small flaws has led to a developing interest 19.9 References

in applications of scanning acoustic mi-
croscopy which produces ultrasonic im- Ambrose, J. C , Perkins, R., Airey, R., Nicholas,
ages of microstructures. Some of these M. G. (1993), in: Designing Ceramic I/Interfaces II:
Understanding and Tailoring Interfaces for Coating,
applications have been very successful: de- Composite and Joining Applications: Peteves, S. D.
fects smaller than 100 |im have been de- (Ed.). Luxembourg: Commission of the European
tected in alumina and silicon carbide at Communities, pp. 295-303.
Bailey, F. P. (1980), Solid-State Metal-Ceramic Reac-
depths of 5 mm. However, the use of this tion Bonding: Some Applications and Properties in
instrument to characterise interfaces is still Energy and Ceramics: Vincensini, P. (Ed.). New
in the development stage. York: Elsevier.
De Bruin, H. J., Moodie, A. E, Warble, C. E. (1972),
Similar difficulties are faced when ap- /. Mater. Sci. 7, 909-918.
plying X-radiography. Flaws much smaller Bucklow, I. A., Dunkerton, S. B., Hall, W. G., Char-
than 100 jam can be detected if they are don, B. (1992), in: Proc 43th Int. Symp. Ceramic
Materials and Components for Engines: Carlsson,
conveniently located and the differences in R., Johansson, X, Kahlman, L. (Eds.). New York:
the absorption coefficients are marked. Elsevier, pp. 324-332.
However, the thinness of interfacial cracks Chatain, D., Coudurier, L., Eustathopoulos, N.
(1988), Rev. Phys. Appl. 23, 1055-1064.
and unbonded regions is such that varia- Cho, H. C , Yu, J. (1992), Scr. Metall. Mater. 26,
tions in absorption are minor and the de- 797-802.
tection of planar defects is still a serious Crispin, R. M., Nicholas, M. G. (1976), J. Mater. Sci.
11, 17-21.
problem. Derby, B., Wallach, E. R. (1984), Met. Sci. 18, 427-
Defect thinness is not a major difficulty 432.
when using pulsed video thermography. Derby, B. (1987), Mater. Sci. Res. 21, 319-329.
Elssner, G. (1989), in: Designing Interfaces for Tech-
The essence of this technique is to subject nological Applications, Ceramic-Ceramic and Ce-
one surface of a sample to a short burst of ramic-Metal Joining: Peteves, S. D. (Ed.). New
heat, such as the flash from a xenon lamp York: Elsevier.
Holloway, D. G. (1973), The Physical Properties of
and to monitor the temperature transient Glass. London: Wykeham Publications.
at the opposing surface. Any defect be- Housekeeper, W. G. (1923), /. Am. Inst. Electron.
tween the faces will slow the diffusion of Eng. 42, 870-876.
Iseki, T, Nicholas, M. G. (1979), /. Mater. Sci. 14,
heat and hence affect the transient. Iso- 687-692.
lated defects will produce local pertuba- Joy, T. (1987), in: Proc. 5th Int. Conf. on High Tech-
tions but a uniform'defect such as a contin- nology Joining, BABS/TW1, Paper 23. Cambridge:
The Welding Institute.
uous well bonded interface will merely Mizuhara, H., Heubel, E., Oyama, T. (1989), Ceram.
alter the background transient. This tech- Bull. 68, 1591-1600.
nique has considerable potential and has Moorhead, A. X, Becher, P. F. (1987), Weld. J. 66,
265-335.
been applied to diffusion bonded ceramics Naidich, Y. V. (1981), Prog. Surf Membrane Sci. 14,
(Reynolds and Wells, 1989), but the high 354-485.
capital cost of the quipment remains a Nicholas, E. D. (1990), in: Encyclopedia of Materials
Science, 2nd Suppl. Vol. Oxford: Pergamon Press,
problem. pp. 915-923.
While needed, available techniques for Nicholas, M. G., Valentine, T. M., Waite, M. J.
nondestructive evaluation have yet to (1980), J. Mater. Sci. 15, 2197-2206.
Nicholas, M. G., Mortimer, D. A., Jones, L. M.,
prove their general usefulness for charac- Crispin, R. M. (1990), J. Mater. Sci. 25, 2679-2689.
terising joints in ceramic-ceramic or ce- Ning, X. S., Suganuma, K., Morita, M., Okamoto, T.
ramic-metal components. As for many (1987), Phil. Mag. Lett. 55, 93-97.
Pask, J. A., Askay, L. A., Hoge, C. E. (1974), J. Phys.
other aspects of ceramic joining, more Chem. 12, 1178-1183.
work is needed. Pask, J. A. (1987), Ceram. Bull. 68, 1587-1593.
 19.9 References 291

Reed, L., Wade, W, Vogel, S., McRae, R., Barnes, C. Eng. Sci. Proc. 10. Westerville, OH: American Ce-
(1966), Metallurgical Research and Development ramics Society.
for Ceramic Electron Devices, AD636950. Washing- Mizuhara, H., Mally, K. (1985), "Ceramic-to-Metal
ton, DC: Clearing House for Federal Scientific and Joining with Active Brazing Filler Metal", Weld. J.
Technical Information. (Miami) 64, 27-33.
Reynolds, W N., Wells, G. M. (1989), Br. J. Nonde- • Naidich, J. V. (1981), 'The Wettability of Solids by
structive Testing 26, 40. Liquid Metals", Prog. Surf Membrane Sci. 14,
Riihle, M., Backhaus-Ricoult, M., Burger, K., 354-485.
Mader, W (1987), Mater. ScL Res. 21, 395-306. Nicholas, M. G. (Ed.) (1990), Joining of Ceramics.
Twentyman, M. E. (1975), J. Mater. Sci. 10, 765-799. London: Chapman & Hall.
Yamada, T., Yokoi, K., Kohno, A. (1989), in: Joining Pask, J. A. (1987), "From Technology to the Science
Ceramics, Glass and Metal: Kraft, W. (Ed.). of Glass/Metal and Ceramic/Metal Sealing", Ce-
Oberursel, Germany: Deutsche Gesellschaft fur ram. Bull. 66, 1587-1593.
Metallkunde, pp. 147-153. Pauling, L. (1960), The Nature of the Chemical Bond,
3rd ed. Ithaca, NY: Cornell University Press.
Peteves, S. D. (Ed.) (1989), Designing Interfaces for
General Reading Technological Applications: Ceramic-Ceramic and
Ceramic-Metal Joining. London: Elsevier Applied
Carlsson, R., Johansson, T., Kahlman, L. (Eds.) Science.
(1992), Ceramic Materials and Components for En- Schwartz, M. M. (1994), Brazing. London: Chapman
gines. London: Elsevier Applied Science. & Hall.
Loehman, R. E., Johnson, S. M., Moorhead, A. I Twentyman, M. E. (1975), "High Temperature
(Eds.) (1989), "Structural Ceramic Joining", Ceram. Metallising", J. Mater. Sci. 10, 765-799.
20 Functional Gradient Materials
Toshio Hirai

Institute for Materials Research, Tohoku University, Sendai, Japan

List of Symbols and Abbreviations 295

20.1 Introduction 297
20.2 Composites and Functional Gradient Materials 297
20.2.1 Nanocomposites 297
20.2.2 Fine Composites 299
20.2.3 Functional Gradient Materials 300
20.2.3.1 Continuously Changing the Morphology of the Dispersoid 300
20.2.3.2 Continuously Changing the State 301
20.2.3.3 Continuously Changing the Crystal Structure 301
20.2.3.4 Continuously Changing the Distribution Pattern 301
20.2.3.5 Continuously Changing the Grain Boundary Characteristics 302
20.3 Functional Gradient Materials for the Relaxation of Thermal Stress 302
20.3.1 Developmental Process 302
20.3.2 The Design 303
20.3.2.1 Compositional Distribution Functions 303
20.3.2.2 Rule of Mixtures 303
20.3.2.3 Design Procedure 304
20.3.3 Design Examples 305
20.3.3.1 Design of Functional Gradient Materials Prepared
by Chemical Vapor Deposition 305
20.3.3.2 Design of Functional Gradient Materials Prepared
by the Powder Metallurgical Process 307
20.3.3.3 Design of Functional Gradient Materials Prepared
by Self-Propagating High-Temperature Synthesis (SHS) 308
20.4 Fabrication Processes for Functional Gradient Materials 309
20.4.1 Vapor Phase Methods 309
20.4.1.1 Chemical Vapor Deposition (CVD)
and Chemical Vapor Infiltration (CVI) Methods 309
20.4.1.2 Physical Vapor Deposition (PVD) Methods 313
20.4.1.3 Surface Chemical Reaction Methods 314
20.4.2 Liquid Phase Methods 315
20.4.2.1 Solution Methods 315
20.4.2.2 Sol-Gel Method 316
20.4.2.3 Copolymerization Method 316
20.4.2.4 Plasma Spraying Methods 317

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
294 20 Functional Gradient Materials

20.4.2.5 Molten Metal Infiltration Methods 318

20.4.3 Solid Phase Methods 320
20.4.3.1 Methods of Obtaining Compositionally Graded Green Bodies 320
20.4.3.2 Sintering Methods 323
20.4.3.3 Self-Propagating High-Temperature Synthesis (SHS) Methods 324
20.4.3.4 Martensitic Transformation Technique 325
20.4.3.5 Diffusion and Reaction Techniques 326
20.5 Characteristics and Future Development of
Functional Gradient Materials 326
20.5.1 Thermal Stability (Time Variation of Compositional Distribution
Under Temperature Gradient) 327
20.5.2 Thermal Insulation Characteristics 327
20.5.3 Thermal Fatigue and Thermal Shock Resistance . 328
20.5.4 Resistance to Extreme Environmental Changes 330
20.5.4.1 High Temperature Supersonic Gas Flow 330
20.5.4.2 Bipropellant Rocket Combustion Gas Flow 331
20.5.5 Oxidation and Corrosion Resistance 332
20.5.6 Electrical Properties 332
20.5.7 Thermoelectric Properties 333
20.5.8 Magnetic Properties 334
20.5.9 Optical Properties 335
20.5.10 Other Properties 336
20.6 Final Comments 336
20.7 Acknowledgements 336
20.8 References 337
 List of Symbols and Abbreviations 295

List of Symbols and Abbreviations

c volume fraction
d thickness of FGM layer
E Young's modulus
/> /i> /o composition, at inner and outer surfaces
f(z) compositional distribution function
G shear modulus
K bulk modulus
Keff effective thermal conductivity
n parameter continuously controlled by the compositional distribution
profile
p pressure
Q Seebeck coefficient
q heat flux within the sample
r, ri9 ro radius, inner/outer radius
T temperature
Ti9 To inner/outer temperature of cylinder
Tb, Ts temperature of bottom/top surface
t thickness of base material
x relative distance from interface or surface
z, zi5 zo nondimensional thickness, inner/outer surface
Z thermoelectric figure of merit

a, ai9 ao thermal expansion coefficient, inner/outer surface of cylinder

x thermal conductivity
v Poisson's ratio
Q electrical resistivity
a stress

ARE activated reactive evaporation

BMA benzyl methacrylate
CGL compositional gradient layer
CIP cold isostatic pressing
CPU central processing unit
CVD chemical vapor deposition
CVI chemical vapor infiltration
DC direct current
EB electron beam
f.c.c. face-centered cubic
FGM functional gradient material
GI graded index
HAP hydroxy-apatite
HCD hollow-cathode discharge
h.c.p. hexagonal close-packed
296 20 Functional Gradient Materials

HIP hot isostatic pressing

HP hot pressing
HYDECS hybrid direct energy conversion system
IC integrated circuit
MM A methyl methacrylate
MMH monomethyl hydrazine
NFGM non-functional-gradient material
NTO nitrogen tetroxide
PMMA poly(methyl methacrylate)
PSZ partially stabilized zirconia
PVD physical vapor deposition
PZT Pb(Zr,Ti)O 3
RI radioisotope
RT room temperature
SHS self-propagating high-temperature synthesis
SS stainless steel
YSZ yttria stabilized zirconia
 20.2 Composites and Functional Gradient Materials 297

20.1 Introduction separate surfaces or in separate parts. An

example of a composite having different
In recent years the environments in functions in its different parts is a coated
which materials are used have become or joined material designed to improve a
more demanding. Conventional materials material's surface characteristics. However,
cannot withstand the severe environments these inhomogeneous composites possess
confronting modern technologies. Further sharp boundaries, as shown in Fig. 20-1 c.
developments in science and technology The boundary often exhibits various unde-
rely heavily upon the development of new sirable properties caused by the existence
materials that can withstand the condi- of discontinuities in the material's mechan-
tions that are created by advancing tech- ical, physical, and chemical characteristics
nology. at the boundary. An example is a separa-
In the development of these new materi- tion at the boundary due to thermal stress.
als there are two approaches. One is to For this reason a proposal to design a
develop a new material that differs com- new material aimed at eliminating the
pletely from any present materials. The macroscopic boundary in laminated-type
other option is to develop new functions materials is being examined (Niino et al.,
for existing materials. The development of 1987). That is, the approach is to synthe-
composites satisfies both of the above ap- size inhomogeneous composites, in which
proaches. Many new types of composite the material's mechanical, physical, and
have already been fabricated in accor- chemical properties change continuously,
dance with the material properties de- and which have no discontinuities within
manded by today's technology. the material. These materials are called
A general requirement for industrial functional gradient materials (FGMs) (Hi-
materials is uniformity in material proper- rai et al., 1987; Koizumi, 1993), and are
ties. That is, for industrial applications, it shown in Fig. 20-1 b. (An alternative name
is essential that every part of the material is "functionally graded materials.")
in use exhibits uniform properties. The re-
quirements for conventional composites
are no exception. Most effort in the devel- 20.2 Composites and Functional
opment of composites in the past has been Gradient Materials
put into determining how to uniformly mix
20.2.1 Nanocomposites
the dispersoid within the matrix. As a re-
sult, from a macroscopic view point, even Most common composites are com-
composites were regarded as homoge- posed of a matrix (A) and a dispersoid (B).
neous because in these materials the me-
chanical properties and other material
characteristics prove to be homogeneous,
mm
• •
••••• • ••

• •••••••
••••• • •
••••••••
•

as shown in Fig. 20-1 a.

•••:

In contrast, studies are also being con-

:::

ducted to design materials that have two (a) (b) (c)

different functions within the given materi-
Figure 20-1. Conventional composites and a func-
al. These materials are inhomogeneous tional gradient material: (a) homogeneous composite,
composites that are characterized by hav- (b) functional gradient material, and (c) coated or
ing different material characteristics on joined-type composite.
298 20 Functional Gradient Materials

Both A and B are classified as "sub- partial crystallization, grain boundary re-
stances". The composites are prepared by action, and thermal decomposition. By the
mixing or joining A and B (Hirai, 1984; CVD method, either film or plate-like
Hirai and Goto, 1986). These composites nanocomposites are obtained by co-depo-
are called phase-joined composites. In sition, that is, by simultaneous vapor de-
phase-joined composites, such as glass composition of many different source
fiber-reinforced plastic, carbon fiber-rein- gases (Hirai and Goto, 1986). Figure 20-2
forced plastic, and ceramic whisker-rein- illustrates typical nanostructures of plate
forced ceramics, the size of the dispersoid and film nanocomposites prepared by
B in the composite is of the order of CVD. In Fig. 20-2, the size of the disper-
a micrometer. The combination is: sub- soid ranges from a few to several tens of
stance (matrix) 4- substance (dispersoid: nanometers. These nanostructures can be
micrometer). The material design of these controlled by changing the kind of source
microcomposites is based on the rule of gases or the CVD conditions.
mixtures. Nanocomposites can possess very un-
In recent years, many studies have been usual material properties. These properties
conducted to prepare composites contain- are often difficult to explain by the tradi-
ing extremely small dispersoids of the or- tional rule of mixtures, mainly owing to
der of several nanometers to several tens of the extremely small size of the dispersoid
nanometers. These composites are called (Hirai and Sasaki, 1991a; Newnham,
nanocomposites (Roy, 1986; Hirai and
Sasaki, 1991a).
In nanocomposites the dispersoid B is
also considered to be a "substance", just as
in many other known composites. The
combination is: substance (matrix) + sub-
stance (dispersoid: nanometer). When the
dispersoid size reaches the order of
nanometers, the dispersion becomes very
difficult to handle. Hence a nanocompo-
site cannot easily be prepared by use of the
phase-joining method described earlier.
For preparing nanocomposites, the so-
called phase-separating (in situ) method is
employed (Hirai, 1984; Hirai and Goto,
1986). In this method the source material
containing both A and B is treated to in- Figure 20-2. Nanostructures of CVD nanocomposites,
A-B, where A and B represent matrix and dispersoid,
duce phase separation. After separation, respectively: (a) spherical particle dispersion: SiC-C,
the nanocomposite is formed. The size of Si3N4-C, Si3N4-TiN, BN-TiN, BN-Si3N4, A1N-
the dispersoid (B) in this case is of the A12O3, (b) disk-shaped particle dispersion: C-SiC,
order of nanometers. YBa2Cu3O>,-Y2O3, (c) rod-shaped particle dispersion:
BN(amorphous)-BN(hexagonal), (d) fiber dispersion:
There are several methods available for
Si3N4-TiN, SiC-TiC, YBa^UgO^-CuO, YBa2Cu3O,-
the in situ preparation of nanocomposites. Y 2 Cu 2 O 5 , (e) thin layer dispersion: C-SiC, YBa2Cu3OJ!
These include: CVD, PVD, sol-gel, co-pre- (c-axis)-YBa2Cu3Oj;(fl-axis), (f) lamella: Si3N4-BN
cipitation, hydrothermal, eutectic reaction, (Hirai, 1993).
 20.2 Composites and Functional Gradient Materials 299

surface area ultra-fine (jLlm=>nm)

interface morphology, shape

reaction

crack formation
crystal lattice
arrangement
concentration and
strain field structural variation
Figure 20-3. Effect of decreasing dispersoid size on the structural factors of composites. As the dispersoid size
decreases from micrometer to nanometer, the total surface area (interface area) of the dispersoids increases. This
affects the chemical reactions between the matrix and dispersoid and the properties of nanocomposites. In
nanocomposites, the morphology and shape of the dispersoid, the crystal lattice arrangement between the matrix
and dispersoid, and the concentration and structural variation of the dispersoid may become important factors
for the properties compared to the case of coarse dispersoids. The strain field and crack formation near fine
dispersoids differ from those in microcomposites.

1986). Figure 20-3 illustrates the effect of tion is: substance (matrix) + element (dis-
decreasing the dispersoid size. persoid). Some composites contain disper-
As will be illustrated later, FGMs pre- soids with stacking faults, lattice defects,
pared by CVD also exhibit similar nano- grain boundaries, or nanovoids.
structures. Control of these nanostructures When the dispersoid is composed of the
strongly influences the properties of the same materials as the matrix, a unique
resulting FGMs. composite can be obtained by designing
dispersoids with different crystal struc-
20.2.2 Fine Composites tures or different crystal orientations. In
Some studies are being pursued to devel-
op new composites based on a concept be- Table 20-1. Features to be considered when preparing
yond the "substance + substance" combi- fine composites (Hirai and Sasaki, 1991a)
nation illustrated earlier. The recent em-
Elements Examples
phasis is on the preparation of composites
by combining elements rather than sub- Morphology sphere, rod, fiber, flake, lamella
stances. In this case, the dispersoids are State solid, pore, liquid
not limited to the solid phase; they can in Crystal structure crystalline, turbostratic, amorphous
Distribution uniformity, continuity
fact be composed of the various "ele- Grain boundary matching, reaction, perfection
ments" listed in Table 20-1. The combina-
300 20 Functional Gradient Materials

the above case, composites are prepared by

considering the dispersoid not as a materi-
al but as an element. They are then often
referred to as "fine composites" to distin-
guish them from normal composites.
Examples of materials developed from
fine composites are FGMs having differ-
° 0 o oooo oooooo
ent crystallinity in different parts, such as O ° c o o O °o °°«ooc

a structure which changes from crystalline 0

0

o
0 j O
o ooo ooooc
to amorphous within one material. Anoth- ° O 0
ooooo<

er example would be a material having dif- o »o "To* °

o ° o °°ooo ooooo
ferent preferred orientations within the
Y
one material. FGMs having voids in the (d)
structure to increase the resistance to ther- Figure 20-5. Designs of FGMs using the concept of
mal shock are designed on the basis of the fine composites: (a) continuously changing the mor-
concept of fine composites. phology of the dispersoid, (b) continuously changing
the state, (c) continuously changing the crystal struc-
ture, (d) continuously changing the distribution pat-
20.2.3 Functional Gradient Materials tern (Hirai, 1993).
There are two approaches to obtaining
FGMs, as shown in Fig. 20-1. The first the direction of the material's thickness
method is to eliminate the boundary of but in the direction of its width as well.
laminated-type composites [from (c) to In normal FGMs, the concentration of
(b)], thereby eliminating discontinuities in the dispersoid changes gradually within
the properties at the boundary. The second the composite, but another type of FGM
method is to make nonuniform distribu- can be fabricated. An FGM can be de-
tions of dispersoids in a homogeneous signed by continuously changing the ele-
composite [from (a) to (b)], thus creating ments shown in Table 20-1 from surface X
multiple functions within the material. to surface Y of a fine composite. Examples
Figure 20-4 shows the structures of FGMs. of this can be seen in Fig. 20-5, in which
The gradient can be assigned not only in the following cases are illustrated.

20.2.3.1 Continuously Changing

(a) (b) the Morphology of the Dispersoid

Spherical and fiber-like dispersoids are

distributed over surfaces X and Y. The
shapes of these dispersoids are gradually
varied (Fig. 20-5 a). Within a material, the
dispersoid will be changed from spherical
to fiber-like. In this way a high Young's
modulus and a high toughness value can
Figure 20-4. Variations in the structures of FGMs: be obtained on surface X while also
(a) in the thickness direction, (b) in the width direc- achieving a high strength on surface Y. If
tion, (c) cylindrically, (d) from the center to both sides. surface Y is changed to a composite with a
 20.2 Composites and Functional Gradient Materials 301

flake-like dispersoid it will be a material 20.2.3.3 Continuously Changing

that is rich in lubricity. Plant systems ob- the Crystal Structure
served in trees and bamboo often have this
In an amorphous material, high me-
type of structure.
chanical strength can be obtained by crys-
tallization. If a surface X is crystallized
20.2.3.2 Continuously Changing the State while the opposite surface, Y, is main-
tained in the amorphous state, surface X
With dense materials on surface X, a
will have high mechanical strength while
porous surface Y can be created by inro-
surface Y will have good corrosion resis-
ducing small voids into that surface.
tance; a gradient of these two properties
Bioactive materials are an example of this
will develop between the two surfaces
type of composite. The advantages are that
(Fig. 20-5 c). For metallic materials, an im-
surface X has sufficient strength to be used
provement in the surface properties is ob-
as a structural material, while surface Y
tained by introducing ions to the surface,
has the good compatibility necessary for a
thereby making it amorphous. These are
biomaterial. These voids also exhibit supe-
examples of crystalline/amorphous FGMs.
rior thermal stress relaxation. This charac-
If a crystallographic axis on surface X is
teristic can be used to design unique and
oriented in one direction, and no specific
intriguing FGMs. When PSZ thermal bar-
orientation is given for surface Y, then it is
rier coatings are fabricated by electron-
possible to effectively utilize these aniso-
beam PVD, it is possible to vary the coat-
tropic properties in order to influence the
ing density by controlling the substrate
material's properties.
temperature. As a result, this coating has
superior thermal barrier characteristics By continuously changing the crystal
(Fritscher and Bunk, 1990). structure in the thickness direction of
a film from f.c.c. (surface X) to h.c.p.
When the void distribution is continu-
(surface Y), an attempt has been made to
ously changed by controlling the level of
prepare a high density perpendicular mag-
sintering from the front surface to the back
netic recording medium (Osaka et al.,
surface, the result is a simultaneous contin-
1990).
uous change in the Young's modulus from
By observing the change from paramag-
the front to the back. This phenomenon is
netic to ferromagnetic that results from the
used to assign gradients in the piezoelectric
martensitic transformation, a magnetic
properties within the material in order to
FGM that has a gradual change in its sat-
generate a flexural vibration mode (Kawa-
uration magnetization can be designed.
saki and Watanabe, 1990).
This is achieved by continuously changing
This technique of continuously chang-
the transformation level from surface X to
ing the state (Fig. 20-5 b) has some inter-
surface Y using the material to control de-
esting applications in polymeric materials.
formation (Watanabe et al., 1993).
For example, by gradually lowering the de-
gree of polymerization of a rod-like poly-
meric material, starting at the outer
20.2.3.4 Continuously Changing
perimeter and continuing toward the cen-
the Distribution Pattern
ter, it is possible to control the damping
characteristics or vary the optical proper- High strength can be obtained by uni-
ties of the final material. formly distributing the dispersoids on sur-
302 20 Functional Gradient Materials

face X. High electrical conductivity can be was a reduction in the thermal stress gener-
obtained by distributing the dispersoids in ated at the interface of the joined or coated
the form of a network pattern on surface Y materials due to the difference in their
(Fig. 20-5 d). thermal expansion coefficients. In the case
of coating, Kaczmarek et al. (1984) pre-
20.2.3.5 Continuously Changing pared a "graded coating", in which the
the Grain Boundary Characteristics composition changed from NiCrAl to
MgO-ZrO 2 , using plasma spraying in
By maintaining a nearly perfect grain order to reduce the thermal stresses. In the
boundary on surface X while introducing case of joining, Suganuma et al. (1984) in-
voids or faults into the boundary regions serted two layers at the interface between
on surface Y, it is possible to control the the metal and the ceramic; they calculated
diffusion of the atoms or molecules and the internal stress generated by the thermal
thus obtain specialized characteristics suit- expansion mismatch using a finite element
able for a variety of sensor applications, or method, and estimated an optimum thick-
for use as a catalyst. ness and a suitable material for the insert
layer. However, due to high thermal stress
in severe environments spalling or delami-
20.3 Functional Gradient nation occurs at the hetero-interface due to
Materials for the Relaxation of changes in the composition.
Thermal Stress From 1984 to 1985, the author and co-
workers proposed a new material design
20.3.1 Developmental Process
concept in which the thermal stress is re-
The surface temperature of future space- duced by continuously changing the ma-
craft is estimated to reach as high as terial-composing elements (NASTA et al.,
1700 °C, while the inside is cooled to about 1986). Changes to the material-composing
700 °C. The success of spacecraft construc- elements in these discussions included
tion relies heavily on the successful devel- changes in the concentration of the disper-
opment of thermal barrier materials which soid, as well as changes in the elements
can protect the craft from severe environ- shown in Table 20-1, such as microstruc-
ments. Since the temperature gradient of ture, crystal structure, and pores. Based on
these thermal barrier materials can often the concepts formulated during these dis-
exceed 1000 °C, a large amount of thermal cussions, the author and others produced
stress can develop within the material. A an SiC/C system FGM with an optimum
complicating factor is that these materials compositional distribution profile for re-
are preferred in the thinnest form possible ducing the thermal stress, by a CVD meth-
(about 10 mm). There is no known mono- od. In doing so, they successfully demon-
lithic material that can withstand these strated the effectiveness of an FGM for
severe thermo-mechanical loads at this thermal stress relaxation (Hirai and Sasaki,
thickness. 1991b).
Metal-ceramic laminated-type com- In 1964, Goetzel and Lavendel (1965)
posites are a possible candidate. In the ear- prepared a "graded blade" for aerospace
ly 1980s in the fields of joining and coat- systems using a liquid phase infiltration
ing, these composites were made with a process. However, the optimum composi-
stepwise change in composition. The goal tional distribution profile was not estimat-
 20.3 Functional Gradient Materials for the Relaxation of Thermal Stress 303

ed. FGMs proposed by the author and tion is presented as (Hirano et al., 1990 a)
others are designed by using an optimum
compositional distribution function ob- (20-1)
tained from numerical analyses.
The proposal to use FGMs for the re- where/is the composition, z is the nondi-
duction of thermal stress was adopted mensional thickness, and the exponent n is
from 1987 to 1992 in Japan, under the aus- a parameter continuously controlled by
pices of the Science and Technology Agen- the compositional distribution profile. As
cy, as a National Research Project (Koizu- shown in Fig. 20-6 a, the compositional
mi, 1992; Okamura, 1991). distribution can be widely varied by chang-
ing the distribution function parameters
20.3.2 The Design (/ o ,/ i? zo, zi9 n). Here the parameters zo and
z{ represent the thickness of a nongradient
20.3.2.1 Compositional Distribution
layer at the outer and inner surfaces, re-
Functions
spectively. The parameters f0 and/| are the
In the design of FGMs, the thermal corresponding compositions.
stress relaxation must be effectively at- Equation (20-1) can be simplified as
tained in the environment in which the ma-
terial is to be used. In order to achieve this x (20-2)
objective, structural shapes and thermo-
mechanical boundary conditions are first where c is the volume fraction, x is the rel-
set. Next, for a composite system, the opti- ative distance from the interface or surface,
mum combination of materials and the op- and d is the thickness of the FGM layer.
timum gradient profile of the composition This relationship is shown in Fig. 20-6 b.
or element within the material must be de-
termined. That is, the compositional distri-
20.3.2.2 Rule of Mixtures
bution profile or the structural variation
profile must be controlled. In order to obtain the temperature dis-
Of prime importance in designing an tribution or the thermal stress distribution
FGM, the compositional distribution func- for an FGM in which the composition

Figure 20-6. Compositional

distribution functions of
FGMs: (a) Eq. (20-1)
(Hirano et al., 1990 a) and
(b) Eq. (20-2).
0 0.2 0.4 0.6 0.8 1.0
x/d
(a) (b)
304 20 Functional Gradient Materials

changes continuously, it is necessary to the relationships between a composite

know first the material values for each structure and the rule of mixtures (Hirano
composite (non FGM composite). Then, etal., 1990 a).
which rule of mixtures is the most appro-
priate, out of the many known rules of
20.3.2.3 Design Procedure
mixtures, for the particular composite un-
der consideration must be known. In addi- The optimum compositional distribu-
tion, since the characteristics of com- tion profile is selected to minimize the ther-
posites depend not only on the composi- mal stress at the interface between two dif-
tion but also on the composite structures, ferent materials. This profile is usually cal-
it is necessary to fully consider the rela- culated by use of a finite element method
tionships between the material characteris- for an object with a definite size and shape
tics and structures in applying the rule of under an appropriate thermal loading con-
mixtures. dition.
The most common rule of mixtures is FGMs in practical use usually have a
based on the microstructure (Wakashima three-dimensional structure and the com-
and Tsukamoto, 1990). Table 20-2 shows positional distribution within the structure

Table 20-2. Rules of mixtures based on microstructures of composites (Hirano et al., 1990 a).

Microstructure in the Thermal Coefficient Elastic moduli Material

hypothetical layer conductivity of thermal E, K, G strength
(white; matrix, black; filler) expansion

Laminar Reuss rule modulus-based Voigt rule Voigt rule

rule

Fibrous (1) Reuss rule modulus-based Voigt rule Kelly-Tyson's

rule equation

Fibrous (2) Voigt rule Voigt rule Reuss rule (Reuss rule)

Thin layer Voigt rule Voigt rule Reuss rule (Reuss rule)

Flake-like Kerner's Eshelby's Eshelby's ?

equation theory theory

Spherical Kerner's Kerner's or Kerner's ?

particle equation Turner's equation
equation
Micro-pores Maxwell's — Mackenzie's Haynes or
equation equation Ryskewitsch's
equation
 20.3 Functional Gradient Materials for the Relaxation of Thermal Stress 305

is continuous. However, for simplicity we 20.3.3 Design Examples

will consider a flat plate in which the func-
20.3.3.1 Design of Functional Gradient
tional characteristic changes only in the
Materials Prepared by
thickness direction, or a cylinder in which
Chemical Vapor Deposition
the function changes only in the radial di-
rection. If we assume that the surface of On forming a protective coating on the
either the flat plate or the cylinder is uni- surface by the CVD method, delamination
formly heated and constrained, then the of the coating from the substrate can often
stress and strain can be considered to be observed. This delamination is due to
change only in the direction of the plate the thermal stress caused by the mismatch
thickness or in the radial direction. On the between the thermal expansion coefficients
basis of these models the thermal stress can of the substrate and the coating. In order
be analyzed for an infinite FGM plate or to prevent delamination, attempts have
cylinder (Hirano et al., 1990b). been made to use multilayered coatings in
There are several approaches to design- which each layer has a different thermal
ing this FGM. Hirano etal. (1990a) pro- expansion coefficient.
posed an "inverse design procedure" to A coating will adhere to a given sub-
obtain the optimum compositional distri- strate if the thermal expansion mismatch is
bution function. In this procedure, first the no greater than 25%. Harris etal. (1972)
proper materials are selected with the aid calculated the threshold of the mismatch
of a material characteristics database. A of thermal expansion values for the system
proper combination of materials is then of a C - C composite substrate with a CVD
examined. The resulting composite values ZrC coating. They proposed that six layers
are estimated by use of a suitable rule of of C-ZrC composites with a stepwise
mixtures. Material shapes, the boundary composition gradient would be sufficient
conditions for the equation relating tem- to keep the mismatch at less than 25%.
perature and stress, and also the composi- As shown in Fig. 20-7, Kowbel (1988)
tional distribution are determined. One-di- added a ZrC/C graded coating by CVD as
mensional steady or unsteady heat con- part of an oxidation resistant, multilayered
duction and thermal stress analyses are coating on a C - C composite. The inser-
then conducted. In this way, the material tion of a ZrC/C graded coating can be very
value distribution, the temperature distri- effective in the relaxation of the thermal
bution, and the thermal stress distribution stress. A ZrC/C graded coating was pre-
are obtained for a given compositional dis-
tribution. The compositional distribution
function can then be treated as a variable
and thermal stress analyses are repeated to ZrC-BN codeposited coating
obtain the optimum design. In this process ZrC/C FGM coating
diamond like coating
various constraints must be taken into ac- boron ion implanted C-C
count; one being that the strength/stress
ratio must be less than one. The hierarchi- C-C composite
cal optimization process developed by Hi-
rano et al. (1990 a) has given very good re- Figure 20-7. Schematic diagram of the multilayered
sults in FGM design. coating system for a C-C composite, including a ZrC/C
graded coating (Kowbel, 1988).
306 20 Functional Gradient Materials

pared by CVD in order to minimize the mal stress (a) in the circumferential direc-
stress generated in the coating of coated tion can be approximated by
particle nuclear fuels used in high tem-
perature gas-cooled reactors (Hollabaugh G = (20-3)
etal., 1975).
However, no known study has been under the assumption that a T is linearly
made on the stress distribution or on the distributed in the radial direction. Here, E
optimum compositional distribution pro- is Young's modulus, v Poisson's ratio, ao
file for these graded or multilayered coat- the thermal expansion coefficient at the
ings. The author and co-workers have suc- outer surface of the cylinder, OL{ the thermal
cessfully prepared by CVD an FGM with expansion coefficient at the inner surface
the optimum compositional distribution of the cylinder, r{ the inner radius of the
profile predicted by a design model. This cylinder ( = 95 mm), ro the outer radius of
work has provided the first step in the syn- the cylinder (105 mm), Tx the inner temper-
thesis of FGM (Sasaki et al., 1989). ature of the cylinder (1327°C), and To the
With the possible application of an outer temperature of the cylinder (27 °C).
FGM as a combustion nozzle in mind, an The stress a calculated using this equa-
infinite cylinder of thickness 10 mm and tion and the properties for CVD SiC and
inner radius 95 mm, which is fixed at one CVD C (Sasaki et al., 1989) is about three
end, is considered. This model was used to times larger than the ultimate stress for
study the stress distribution, assuming SiC (650 MPa). Due to this high stress a
zero axial stress. The inner surface of the combustion nozzle constructed out of an
cylinder was set at a temperature of SiC monolith would probably fail catastro-
1327 °C and the outer surface at 27 °C, thus phically due to the thermal stress caused
creating a temperature difference of 1300 °C by this 1300°C temperature difference.
(Fig. 20-8). In an effort to reduce this thermal stress,
When this infinite cylinder is composed calculations were made on the infinite cyl-
of an SiC monolith the inner surface ther- inder to determine the distribution profiles
of temperature, stress, Young's modulus,
105 mm
and stress/strength ratio by changing the
composition ratio of SiC/(SiC + C) in the
thickness direction. In the calculation of
the thermal conductivity, Young's modu-
27°C lus, and thermal expansion coefficient,
Kerner's rule of mixtures was employed.
The validity of using Kerner's rule for the
SiC/C FGM had been confirmed earlier in
the development of the SiC-C nongradi-
ent composite system. For the calculation
of the strength, the rule of mixtures used
was the harmonic mean rule. In estimating
SiC the internal stress of the combustion noz-
Figure 20-8. Cylinder model for the calculation of the zle, the following assumptions were made:
thermal stress in a combustion nozzle made from the internal stress is equal to the internal
SiC/C FGM (Sasaki et al., 1989). surface stress in the radial direction, and
 20.3 Functional Gradient Materials for the Relaxation of Thermal Stress 307

2000 r 400 | 1 240 H 1.2

1600 0.8

« 1200 0.4

I Figure 20-9. Distributions

of stress, temperature, ratio
I* 800 stress/strength, and Young's
-0.4 modulus calculated for
400
SiC/C FGM using the in-
finite cylinder model
0L -160 -0.8 (Sasaki et al., 1989).
100
(C) Radius, r (mm)

the stress in the radial direction at the ex- cation of PSZ/type 316L stainless steel
terior surface is zero. (PSZ/SS) FGM. These stresses were
* Figure 20-9 shows the radial distribu- grouped into three types, as shown in
tion of stress, temperature, and Young's Fig. 20-11. The residual stress of FGMs is
modulus for the SiC/C infinite cylinder found to decrease as the gradation thick-
calculated for the case where the stress-to- ness (2r/d) decreases, while the residual
strength ratio is unity or less under a given stress at the surface of the FGM coating
temperature gradient (Sasaki et al., 1989). increases as the gradation thickness de-
The assumptions used in this calculation creases.
were that from the internal surface to a Figure 20-12 shows the optimum com-
depth of 1 mm the FGM is composed positional distribution profile of PSZ/SS
of SiC monolith, and from the depth of FGM for reducing the axial stress that
1-10 mm the composition, as well as the causes surface cracks on the sample
material values, continuously change from (Watanabe and Kawasaki, 1992). From
those of SiC to C in the direction normal to the figure it can be seen that the optimum
the plate thickness. Figure 20-10 shows the
compositional distribution profile for this
case (Sasaki etal., 1989).
Using a similar procedure, an optimum
compositional distribution profile is ob-
tained for the TiC/SiC FGM coated by
CVD on an SiC-fiber-reinforced TiC sin-
tered body (Kawai et al., 1992).

20.3.3.2 Design of Functional Gradient

Materials Prepared by the
Powder Metallurgical Process
o 0.2 0.4 0.6 0.8 1.0
By using finite element method analysis, (Q conversion thickness (SiC)
Itoh and Kashiwaya (1992) analyzed the Figure 20-10. Suitable compositional distribution pro-
residual stresses induced during the fabri- file for SiC/C FGM (Sasaki et al., 1989).
308 20 Functional Gradient Materials

stress reduction. Rabin and Heaps (1993)

suggested that the optimum microstruc-
ture for a rod-shaped specimen, in order to
reduce axial stress near the surface, must
have d> 8 mm and n> 3.0.
For the development of thermal barrier
structural FGMs, it is necessary to design
FGMs consisting of more than two com-
base ponents. Matsuzaki et al. (1993) proposed
substrate material the laminated-type Al 2 O 3 -SiC-TiC-TiAl
(a) (b) (c) system, for example, Al2O3-5 %SiC/Al2O3-
Figure 20-11. Three types of FGM for calculating re- 50%TiC/Al2O3-25%TiC/50%TiAl/Al2O3-
sidual stresses: (a) compositionally graded plates from 10%TiC-80%TiAl/TiAl, as shown in Fig.
the surface to the back, (b) graded coatings formed on 20-13.
the substrate, and (c) graded joints formed between
base materials (Itoh and Kashiwaya, 1992).
20.3.3.3 Design of Functional Gradient
Materials Prepared by Self-Propagating
High-Temperature Synthesis (SHS)
Thermal stress analyses have been made
on a TiB2/Cu FGM produced by the SHS
method (Sata et al., 1990). Let us consider
an infinite cylinder of inner radius 95 mm
and thickness 10 mm, constrained at one
end. This cylinder can be a representation
of a rocket thrust chamber. The inner sur-

0.5 1.0 5.0

Exponent n in c = (x/d)n
a-Al 2 O 3
Figure 20-12. Maximum axial stress normalized to (heat insulation,
that for direct bonding as a function of the exponent oxidation
of the compositional distribution profile: where, x, d, resistance)
and n are given in Eq. (20-2) (Watanabe and Kawa-
saki, 1992).

profile is obtained at n = 0J, regardless of

the thickness of the FGM layer. \ A B C D
Williamson and Rabin (1993), using an hot side coolant side
elastic-plastic finite element model, pre- surface surface

dicted the residual stresses developed dur- Figure 20-13. FGM composed of multiple compo-
ing the cooling process for Al 2 O 3 /Ni FGM nents (Matsuzaki, 1994).
In A: a-Al2O3 + jS-SiC (comp.)
and non-FGM. They demonstrated that In B: a-Al2O3 + TiC + jg-SiC (FGM)
the plasticity is an important factor to be InC: a-Al2O3 + TiC + TiAl(FGM) Total FGM
considered in a realistic evaluation of the In D: TiAl
 20.4 Fabrication Processes for Functional Gradient Materials 309

n of the compositional distribution func-

tion. On continuously changing the com-
position or structural elements within the
material, the change can occur in two di-
rections; in the thickness and width direc-
tions, as shown in Fig. 20-4. Most of the
recent studies on FGM synthesis have
been concentrated on changes in the thick-
95
100 105 ness direction. Therefore, in this chapter
Distance (mm) the methods used in the preparation of an
Figure 20-14. The optimum compositional distribu- FGM that has continuous changes in the
tion profile for an SHS TiB 2 /Cu FGM (Sata et al, thickness direction will be discussed. The
1990). discussion also includes the method to pro-
duce an FGM having a stepwise change, as
well as a continuous change. Laminated-
face of the cylinder is composed of TiB 2 type FGMs, which are prepared by first
and is heated to 1227 °C, while the outer producing sheets of film and then layering
surface is made of Cu and is cooled to the films, will also be discussed.
227 °C. Thus the temperature difference is There are several methods for the prepa-
1000 °C. The stress distribution for the ration of an FGM (Ford, 1992; Sheppard,
zero axial stress case was calculated. As 1992). However, for convenience they will
shown in Fig. 20-14, the optimum compo- be categorized according to the source ma-
sitional distribution profile was obtained terial's state during fabrication of the
in order to minimize the stress ratio. FGM. They are: gas phase, liquid phase,
For a TiC/Ni FGM (7 layers) prepared and solid phase approaches. Typical fabri-
by gas-pressure combustion sintering (the cation methods and examples of FGMs
combined process of SHS and HIP), the are listed in Table 20-3.
optimum n values based on the thermal
stress analysis are reported to be 0.5-0.7 20.4.1 Vapor Phase Methods
(Ma et al., 1992). Design calculations were
performed for a TiC-NiAl system joined 20.4.1.1 Chemical Vapor Deposition (CVD)
by a TiC/NiAl FGM layer (Kudesia et al., and Chemical Vapor Infiltration (CVI)
1992). Methods
Chemical Vapor Deposition (CVD)
20.4 Fabrication Processes The CVD method yields a material de-
for Functional Gradient Materials posit with source gases as the feed stock.
This deposit is caused by the application of
The most important objective in the various forms of energy (heat, light, plas-
preparation of an FGM is to achieve a well ma, etc.) onto the source gases after they
controlled (according to the design) distri- are introduced into a reactor. Hydride,
bution of composition, texture, structure, bromide, and chloride are generally used
and other necessary elements. That is, the for source materials. When the source ma-
preparation method for an FGM must be terial is in liquid or solid form it is vapor-
able to accommodate an arbitrary number ized. In recent years, organometallic com-
310 20 Functional Gradient Materials

Table 20-3. Fabrication methods and examples of Table 20-3. (Continued).

FGMs.

Method Examples Method Examples

Vapor phase methods Doctor blade ZrO2/Ni, PZT(1)/PZT(2)

CVD sheet forming
SiC/C, TiC/C, SiC/TiC,
SiC/TiC-SiC, SiO2/SiO2(GeO2), Filtration Al2O3/Ni
BN/Si3N4, C/B4C/SiC, Sedimentation Al2O3/NiAl, A12O3/W
Cr3C/Cr Painting SiC(fiber)/SiC
CVI SiC/C, TiB2/SiC
EB-PVD PSZ (low density/high density) (b) Sintering methods
Ion plating TiN/Ti, TiC/Ti, CrN/Cr Sintering, HP, PSZ/SS, Ni, Ti, TiAl, YSZ/Mo,
ARE TiC/Ti, A1N/A1, Ti-Si Amor/Ti-S and HIP Al2O3/Ni, MgO/Ni, hydroxy-
Cryst, Ti-Al Amor/Ti-Al Cryst, apatite/TiSiC-AlN/W,
Al(Ti)N Amor/Al(Ti)N Cryst SiC-AlN/AlN/Mo, AIN/Ni,
Chemical gas SiC/C A1N/A1, Si3N4/Mo,
reaction Si3N4/Mo, Ni, Ni-based superalloy
Surface Ti-Al-V/nitride Plasma activated PSZ/TiAl, PSZ/Ti, YSZ/SS
treatment SHS TiB2/Cu, TiB2/Ni, TiC/NiAl,
TiC/Ni, A1N/A1, AIN/Ni,
Liquid phase methods Cr3C2/Ni, Cr3C2/TiC,
MoSi2-SiC/TiAl
Electro- Cu/CuZn, Cu/CuNi,
Transformation ss/ss
depositing ZrO 2 + Ni/Ni, SiC/C
Diffusion and ZrO 2 /Al 2 O 3 , PZT(1)/PZT(2),
Electro- CoNiReP (f.c.c.)/CoNiReP (h.c.p.)
reaction PZT/NiNb,
plating Cu/CuZn, Cu/CuNi
Al2O3/Sn/Nb/Sn/Al2O3,
Sol-gel SiO 2 /TiO 2 , SiO 2 /GeO 2 SiC/C, Mo 2 SiB 2 -Ni/SS,
Copolymeri- graded index polymer Ti5Si3/Ti, ZrSi2/Zr
zation
Plasma PSZ/NiCrAlY, PSZ/NiCrAl,
spraying PSZ/NiCr, PSZ/SS pounds have often been used as source ma-
Molten metal SiC/C, W/Cu, A12O3/A1, terials. By continuously changing the mix-
infiltration Cu and Ni, SS/Cu
ture ratio of the source gases or by con-
Centrifugal SiC/Al
trolling the CVD conditions, such as depo-
casting
sition temperatures or gas pressures, the
Solid phase methods CVD method permits relatively easy syn-
(a) Powder stacking methods
theses of various types of FGMs.
Figure 20-15 shows schematically an ex-
Pressing many
perimental setup for preparing FGMs by
(lamination)
the CVD technique (Sasaki and Hirai,
Centrifugal ZrO 2 /NiCr
1990). An SiCl 4 -CH 4 -H 2 combination
Spraying PSZ/SS
was used as the source gas. The deposits
Powder Al2O3/Ni
infiltration were obtained at 1400 and 1500°C. The
Slip casting ZrO 2 /Ni, ZrO2/SS, A12O3/W, total gas pressure range used in this case
Al 2 O 3 /ZrO 2 , Al 2 O 3 /Al 2 TiO 5 , was from 1.3 to 6.5 kPa. The SiC/C FGM
Al2O3/mullite, PZT(1)/PZT(2) (0.2-0.8 mm in thickness) was obtained on
Slurry dipping YSZ/SS a graphite substrate through an appropri-
 20.4 Fabrication Processes for Functional Gradient Materials 311

ate change of the SiC/C ratio in the source

gas. This was easily achieved by control-
ling the SiCl4 and CH 4 gas flow rates. The
change in the gas composition during de-
position at 1.3 kPa is shown in Fig. 20-16
(Sasaki and Hirai, 1990).
Figure 20-15. Schematic diagram of the CVD setup Figure 20-17 presents the compositional
for preparing SiC/C FGM: (1) H 2 gas, (2) CH 4 gas, distribution profile of the cross section of
(3) mass flow meter, (4) SiCl4 reservoir, (5) constant SiC/C FGM film prepared at 1500°C us-
temperature bath, (6) ribbon heater, (7) pressure regu- ing route (a) from Fig. 20-16. This profile
lator, (8) water-cooled reaction chamber, (9) work shows many pores throughout the film. On
coil, (10) pressure gauge, (11) optical pyrometer, (12)
graphite substrate, (13) graphite heater, and (14) pump the other hand, the SiC/C FGM obtained
(Sasaki and Hirai, 1990). at 1400 °C by use of route (b) shows pores
only in the proximity of the substrate. The
CVD conditions which gave the composi-
tional distribution providing the minimum
0.8 residual stress (see Fig. 20-10) were at the
temperatures of 1500°C using route (b)
0.6 - and 1400 °C with route (a) (Sasaki and Hi-
rai, 1990). The cross-sectional view of the
0.4 microstructure of an SiC/C FGM having
(a)/ an optimum compositional distribution
0.2
(Fig. 20-10) is shown in Fig. 20-18 (Hirai
and Sasaki, 1991a). One can see continu-
1 1
ous changes in the microstructure and
n
0 1.8 3.6 5.4 7.2 9.0 crystal orientations, in addition to compo-
deposition time (ks) sitional changes.
SiC/C FGM coatings were prepared by
Figure 20-16. Variations in the gas mixture ratio at
1.3 kPa during three deposition routes (a, b, c) for the changing the ratio of SiCl 4 /CH 4 stepwise
preparation of SiC/C FGMs by CVD (Sasaki and or continuously in a temperature range of
Hirai, 1990). 1300-1500°C and a pressure range of 1.3-

Figure 20-17. Si intensity

distributions in CVD SiC/C
FGMs prepared at 1.3 kPa
and 1500°Cor 1400 °C;
(a), (b), and (c) correspond
to these gas flow variations
shown in Fig. 20-16 (Sasaki
and Hirai, 1990).
0.2 0.4 0.6
relative thickness (mm) relative thickness (mm)
312 20 Functional Gradient Materials

lowed by a layer of TiC/SiC FGM coating,

whose composition changes gradually
from TiC to SiC. This CVD work was per-
formed using the T i C l 4 - C H 4 - H 2 source
gas system at 1350°C and 8 kPa. By
changing the gas flow rates of TiCl4 and
SiCl4 during the deposition process, an
FGM coating of thickness 150 ^m was ob-
tained (Kawai et al., 1990).
For the SiC-fiber-reinforced TiC sintered
body, in order to obtain nearly equal val-
ues for the coefficients of expansion of the
substrate and the coating, SiC-38 vol.%
TiC composite coatings of a thickness of
80 jim were placed on the substrate, and an
SiC coating of thickness 0.2 j^m was then
Figure 20-18. Cross-sectional view of the microstruc- added. The measured properties of these
ture of a CVD SiC/C FGM prepared at 1500°C and
1.3 kPa following route (b) from Fig. 20-16 (Hirai and
laminated coatings suggested that the
Sasaki, 1991a). FGM having the compositional distribu-
tion function value of n = 0.2 is the most
effective for thermal stress relaxation
13 kPa on graphite plate, carbon fibers, (Kawai etal., 1992).
and a cylindrical graphite tube (Uemura Kowbel (1988) prepared a multilayered
et al., 1990). Using similar conditions, an coating on a C - C composite, as shown in
SiC/C FGM of 120 |im thickness was coat- Fig. 20-7. He produced one of the layers
ed on a model combustor with a cylinder of the ZrC/C FGM coating using a C H 4 -
diameter of 30 mm made of three-dimen- ZrCl 4 -H 2 gas system at 1400 and 1600°C
sional C-C composite (Suemitsu et al., and 0.1 MPa. He also prepared C/SiC,
1993). An SiC/TiC FGM was coated using BN/Si 3 N 4 , C/B4C/SiC FGM coatings on a
an Si(CH 3 )Cl 3 -TiCl 4 -CH 4 -H 2 gas sys- two-dimensional C - C composite (Kowbel,
tem at 1050 °C on SS304 plate as a candi- 1993).
date material for UT-3, which is apparatus An FGM having a continuous change in
to produce hydrogen through the thermo- refractive index has been the subject of in-
chemical decomposition of water (Sasaki terest in the optical glass field since 1965
etal., 1993). (Kawakami and Nishizawa, 1965). Optical
In an effort to improve the oxidation fibers with a continuously changing refrac-
resistance of C - C composites, C-fiber-re- tive index are prepared using the following
inforced SiC sintered bodies, and C-fiber- four types of CVD method: modified
reinforced TiC bodies, a TiC/SiC FGM chemical vapor deposition (U.S.A.), out-
layer was coated on these substrates. In side vapor deposition (U.S.A.), plasma-
these cases, in order to minimize the differ- activated chemical vapor deposition (Ger-
ences between the expansion coefficients of many), and vapor phase axial deposition
the SiC coating and various composite ma- (Japan). The vapor phase axial deposition
terials, first a layer of TiC is produced on method will be described here (Fig. 20-19).
the surface of the composite material, fol- In this method, SiCl4 and GeCl4 vapors
 20.4 Fabrication Processes for Functional Gradient Materials 313

surface of the open-pore or the space in-

side of the porous substrate.
The construction of a combustor with a
carbon fiber 2-D fabric can further im-
prove its gas tightness by infiltrating its
pores with SiC/C FGM. The SiC/C FGM
is prepared by CVI using CH 4 and SiCl4
gases at 1100-1500°C and 1.3-13 kPa
(Suemitsu et al., 1993).
Agullo etal. (1993) pre-coated SiC-
Nicalon fiber with C/SiC FGM (thickens;
about 100 nm) by using continuous changes
in the gas phase composition of tetraethyl-
silane (SiEt4) and cumene (/PrC6H5) at
Figure 20-19. Vapor phase axial deposition method 770 °C and 0.8 kPa. Agullo called it a com-
for preparing graded index-type optical fiber: (1) fur- positional gradient layer (CGL). This
nace, (2) vacuum pump, (3) CVD reactor, (4) burner, CGL's composition changes continuously
(5) GeCl4, (6) SiCl4, and (7) H 2 + O 2 . from C at the fiber side to SiC at the sur-
face side. CVI is used to densify the FGM
pre-coated, SiC-Nicalon plain weave fab-
are separately introduced into a flame.
rics above 1000 °C.
Each of the oxide powders produced by
the oxidation reaction is then blasted on Chung et al. (1993) carried out a model
the surface of a quartz rod, resulting in calculation on the CVI of the dichlorodi-
porous deposits. By controlling the deposi- methylsilane-TiCl4-BCl3 system using a
tion conditions, the radial direction com- carbon fiber fabric as a substrate. The re-
positional distribution of GeO 2 , which be- sults of their calculation suggest the forma-
comes the dopant in the SiO2 matrix, is tion of an FGM in the voids of the fabric
continuously changed. With constant rota- by co-deposition of TiB2 and SiC.
tion, the deposit is taken up to the top of 20.4.1.2 Physical Vapor Deposition (PVD)
the furnace. During heating the porous de- Methods
posit reaches the glass phase. It is then
pulled to obtain a graded index-type opti- In the PVD method, a solid source ma-
terial is energized using different kinds of
cal fiber.
energy source to obtain vaporized particles
which deposit on a substrate to form a
Chemical Vapor Infiltration (CVI)
film. There is also a technique to accom-
The chemical vapor infiltration process, plish chemical reaction of the vaporized
which evolved from the chemical vapor de- particles in a gas phase. The deposition
position for surface coatings, uses porous speed of the PVD method is rather slow.
materials such as ceramic bodies and ce- Therefore only a thin FGM film can be
ramic cloths which contain many residual produced using this method. However,
pores. These materials are first placed in- since this technique does not require high
side a CVD furnace as a substrate. By de- temperature heating of the substrate, this
composition or by chemical reaction of the method is preferred for the preparation of
source gases, a coating is deposited on the the thin FGM films used in electronics.
314 20 Functional Gradient Materials

Electron-Beam PVD Technique

600 -
Partially stabilized zirconia (PSZ) with a
gradient density of 4.2-5 g/cm3 (70-84%
theoretical) from the surface to the metal/
300 -
ceramic interface has been prepared on a
superalloy substrate by an electron-beam
PVD technique (Fritscher and Bunk,
1990).
thickness 'substrate
Ion Plating Technique Figure 20-20. Change in the concentration profiles of
Al, Ti, and N in the sputtered FGM film with thick-
HCD (hollow-cathode discharge) type ness (Inoue etal., 1993).
PVD with Ar and C 2 H 2 gases has been
used to prepare a TiC/Ti FGM coating of
from the surface towards the inside by dif-
a thickness between 10 and 15 jim on a Ti
fusion. This concentration gradient can be
plate by changing the flow rate of C 2 H 2 at
controlled by regulating the reaction tem-
300 °C. High purity metallic Ti was used as
perature.
the evaporation metal source (Shinohara
Fujii et al. (1992) have prepared an SiC/
et al., 1993). A TiN/Ti FGM prepared by
C FGM with an SiC composition gradient
a HCD technique showed a superior tone
in the direction from the surface to the
of golden color (Niiyama, 1994).
inside of the graphite matrix in order to
improve the oxidation resistance of graph-
Activated Reactive Evaporation (ARE)
ite materials. This FGM was obtained by
Technique
the chemical gas reaction between graphite
Inoue etal. (1993) have successfully and gaseous silicon monoxide. The SiO
made an amorphous Al(Ti)N/hexAl(Ti)N powder is first converted into gaseous SiO
FGM film. They used Al 80 Ti 20 as the by heating it at 1200-1300 °C.
target for DC magnetron sputtering by The evaporated SiO vapor then reacts
changing the Ar and N 2 mixture ratios. with the graphite (the bulk density is
The amorphous Al (Ti) N + hex Al (Ti) N 1.75 g/cm3) heated at 13OO-138O°C to
is a nanocomposite, and the size of the form SiC. This process is schematically
hexAl(Ti)N dispersoid is abut 4 nm. Fig- shown in Fig. 20-21.
ure 20-20 presents the results of cross-sec-
tional analyses of this FGM film. SiC coating layer
surface *

20.4.1.3 Surface Chemical ®®®®®®

Reaction Methods V.V.V* ' AV.V.
Chemical Gas Reaction Technique (FGM of SiC/C) (SiC coating by
CVD)
In this technique the source gases are :graphite grain
made to react on the surface or in the voids
of a porous matrix, resulting in the chemi- Figure 20-21. Schematic model of SiC/C FGM formed
cal reaction products. The concentration in a graphite matrix by chemical reaction compared
of the end product continuously changes with SiC coating by CVD (Fujii et al., 1992).
 20.4 Fabrication Processes for Functional Gradient Materials 315

Surface Treatment Technique

An FGM can also be obtained on the
material surface by proper surface treat-
ments such as plasma treatment, nitrida-
tion, or carbonization. Ion implantation of
the material surface can result in the sur-
face having FGM characteristics.

20.4.2 Liquid Phase Methods

Figure 20-22. Schematic drawing of the apparatus for
20.4.2.1 Solution Methods electrodepositing a ZrO2/Ni FGM coating on a rocket
combustion chamber: (1) ZrO 2 powder, (2) electroly-
Electrodeposition Technique sis tank, (3) plating solution, and (4) DC supply (Mat-
sumura et al., 1993).
The electrodeposition technique is suit-
able for the production of thin-sheet, gra-
dient metallic alloys. The composition or
Two-dimensional carbon fabric prepreg
structure in the direction normal to the
sheets containing different amounts of SiC
deposition surface can be continuously
powder were fabricated by electrophoresis
changed by either controlling the concen-
in a water solution containing a mixture of
tration of the metallic ions in the elec-
SiC powder, a small amount of A12O3
trolytic solution or by controlling the elec-
powder, and the carrier substance (Kawai
trical current.
etal., 1993). SiC powder deposits in the
By using the electrodeposition tech-
inner space of the fabrics when an electric
nique, Cu/CuZn and Cu/CuNi FGM films
current is applied between the electrode
with thicknesses of 50-200 jim were pre-
and the carbon fabrics. These prepreg-
pared on a Cu substrate (Merk et al.,
sheets are laminated and hot-pressed in Ar
1993). When ceramic powder of the order
to yield the laminated-type C/SiC FGM.
of several nanometers to micrometers in
diameter is dispersed in an electrolytic so-
Electroless-Plating Technique
lution containing metallic ions, the powder
co-deposits on the cathode surface with the Osaka etal. (1990) obtained C o - N i -
metallic substance by electrolysis. R e - P film (composition ratio: 30-60-5-5)
Matsumura et al. (1993) electrolytically on a polyimide film substrate through the
prepared an Ni and ZrO 2 co-deposited film electroless-plating technique. In the C o -
of thickness 80 jim for 30 days on an SS N i - R e - P film formed on the nonmagnet-
substrate by use of a plating solution con- ic NiMoP (thickness 30 nm) of an under
taining nickel sulfamate, nickel chloride, layer, the initial deposit on the substrate
and boric acid in combination with PSZ exhibits a granular structure with a typical
particles. The Ni/Ni + ZrO 2 (25-30 vol. %) grain size of 10-20 nm. This initial deposit
FGM film is obtained by either increasing has an f.c.c. structure and its orientation is
the amount of dispersion or lowering the random. The deposit formed at a later
current density. A schematic diagram of stage (i.e., the film surface layer), on the
the experimental setup for preparing a other hand, shows a columnar structure
rocket combustion chamber is shown in with a typical diameter of about 20-
Fig. 20-22 (Matsumura et al., 1993). 30 nm. Its structure is the h.c.p. type and
316 20 Functional Gradient Materials

the <002> axis is oriented parallel to the (i) (ID

deposition surface. This type of continu- Si(OCH 3 ) 4 Si(OCH 3 ) 4
ous change of crystallographic orientation Ge(OC 2 H 5 ) 4 Ti(O-«-C 4 H 9 ) 4
in the direction of the film thickness can H2O H2O
lead to an improvement in the properties H + orOH" H + andOH"
C 2 H 5 OH «-C3H7OH
of perpendicular magnetic media.
gelling I
20.4.2.2 Sol-Gel Method
A glass rod having a radially graded re-
fractive index was prepared using a sol-gel
(D wet gel

process (Konishi et al., 1988). This process leaching I A \

leaching
employed the metal alkoxide of the two solution

binary systems Si(OCH 3 ) 4 -Ge(OC 2 H 5 ) 4 washing

and Si(OCH 3 ) 4 -Ti(O-«-C 4 H 9 ) 4 . Immer- drying
dry gel
sion of the rod-shaped wet gels in a neutral
or acidic water solution results in leaching sintering
densified
out of the dopant (Ge and Ti). The dopant
glass
left in the gels contributes to the formation
of the concentration gradient. This con- Figure 20-23. Preparation process for an FGM glass
rod by a sol-gel method (Konishi et al., 1988).
centration gradient remained in the densi-
fied glasses of SiO 2 -TiO 2 , as well as in the
SiO 2 -GeO 2 system. The FGM glass rods
were obtained by drying and sintering the
leached gels. When these rods are spun in- methacrylate (MMA) with a refractive
to a fiber, an optical fiber is obtained for index of 1.492, and benzyl methacrylate
which the refractive indices are not con- (BMA) with a refractive index of 1.562.
stant but vary throughout the cross sec- First a polymer tube is prepared using
tion. The production process for these poly-MMA. A monomer mixture of MMA
fibers is shown in Fig. 20-23 (Konishi and BMA is then added to this tube and
etal., 1988). circumferentially heated (60-80 °C) to ob-
tain polymerization. The inner wall of the
Nonreflective glass can also be prepared
polymer tube swells, interacting with the
by coating the glass with a thin layer of
monomer mixture, and eventually a thin
anti-reflection film, which has a gradient
gel phase forms on the surface of the wall.
refractive index. Asahara and Izumitani
Then the copolymerization proceeds to-
(1980) have prepared anti-reflection film
wards the center of the tube. When the
with a graded refractive index via a chemi-
polymerization is completed, a polymer
cal leaching process using a phase separa-
solid is obtained that has a radially varied
ble glass.
mixture ratio of two kinds of polymers
from the center to the outer perimeter. The
20.4.2.3 Copolymerization Method
resulting solid has a radial distribution of
Koike et al. (1989) have prepared graded refractive indices (Koike, 1991). The prep-
index-type polymer optical fiber using the aration process for this polymeric solid is
copolymerization method. The source ma- shown in Fig. 20-24 (Koike, 1992). This
terials for copolymerization are methyl graded index polymer rod can be heat-
 20.4 Fabrication Processes for Functional Gradient Materials 317

monomer liquid phase

copolymer phase
gel phase

Figure 20-24. Schematic

representation of the inter-
facial - gel polymerization
technique (Koike, 1992);
(a), (b), and (c) show the
various stages of reaction.

(a) (b) (c)

drawn at 190-280 °C into an optical fiber single-torch plasma spray reactor (Eroglu
with a graded index. etal., 1993).
Fukushima etal. (1990) developed a
20.4.2.4 Plasma Spraying Methods twin torch for use in FGM preparation. In
this method two plasma torches are placed
In the plasma spraying method, the
so that the center line of each torch is
spraying source powder is transported to
aligned to the point of spray deposition on
the plasma jet by a torch nozzle. The
the substrate. Two different source materi-
molten source material is then sprayed on-
als are sprayed from each nozzle simulta-
to the substrate to form a coating. There
neously onto the same spot on the sub-
are mainly two types of plasma-sprayed
strate to obtain a coating of composite.
FGM coatings; that is, the porosity-grad-
Since each torch can be independently con-
ed coating and the composition-graded
trolled for the required spray conditions,
coating (Steffens et al., 1990). Use of plas-
this method can easily use two source ma-
ma spraying in the preparation of FGM
terials having widely differing melting
coatings requires the clever design of an
points, such as metal and ceramic, to form
apparatus that provides a continuously
an FGM coating. Using this technique, an
changing mixture ratio of the source pow-
eleven-layered FGM coating of thickness
der transported to the torch nozzle.
0.3-0.4 mm has been prepared using an
Ni-base alloy (NiCrAlY) and YSZ. By
Atmospheric Plasma Spraying Technique
using a smaller spray angle the amount of
The MgO-ZrO 2 /NiCrAl laminated pores in the coating is increased. This con-
(7 layers) type FGM coating was prepared tributes to improved thermal barrier char-
on a steel substrate by a plasma spraying acteristics.
technique using MgO (24 wt.%)-ZrO 2
and NiCr-Al (6wt.%) powders. The
spraying took place in the air using a
318 20 Functional Gradient Materials

Low Pressure Plasma Spraying Technique signs had the lowest combined stress and
stress gradient across the FGM coating.
Low pressure plasma spraying is usually
performed in an inert gas atmosphere such 20.4.2.5 Molten Metal Infiltration
as Ar to restrain the oxidation of raw pow- Methods
der.
To prepare an FGM, Shimoda et al. Sintered Porous Body-Molten
(1990) developed a low pressure plasma Metal Infiltration Technique
spraying gun with four ports. Ceramic and This technique infiltrates molten metal
metal powders are simultaneously intro- into the voids of a porous sintered body
duced into the plasma jet using two ports where the void fraction changes continu-
each for each source material, as shown in ously from the surface of the body towards
Fig. 20-25. On spraying under an atmo- the inner body.
spheric pressure of 26.7 kPa, they obtained
During the 1960s, "graded turbine
a YSZ/NiCr FGM coating of thickness
blades" were prepared by infiltrating a
1 mm on the Cu substrate using YSZ and
molten superalloy (Ni- or Co-based) into
Ni-20 wt.% Cr powders.
the TiC porous sintered body (density:
By using vacuum plasma spraying (two 60-80% theoretical) in a vacuum. These
powder feeders) a YSZ/NiCrAlY FGM FGMs usually have high toughness at the
coating (0.625 mm thick) was prepared on outer superalloy portion and high creep
NiCrAlY-coated Inconel 718 (Mendelson resistance at the inner cermet portion
and McKechnie, 1993). It was shown that (Goetzel and Lavendel, 1965).
the parabolic and exponential gradient de-
A W/Cu FGM was prepared using this
approach, as shown in Fig. 20-26 (Taka-
hashi et al., 1993). Six different types of W
powders with an average particle diameter
ranging from 0.49-9.15 pm were used to
obtain sintered W with a graded pore con-
centration. These W powders were sintered
at 1800°C for 8h under an H 2 atmo-
sphere. The relative density of the sintered
body can be controlled within a range of
71-94% by the proper choice of powder
particle size and by controlled pre-press-
ing. The closed pores in the sintered W
were then reduced by capsule-free HIP at
1800 °C and 196 MPa for 8 h. Finally, the
molten Cu was HIP-infiltrated into the
open pores of previously HIP, sintered W
at 196 MPa. In this way a W/Cu FGM is
fabricated.
A C/(SiC + C fiber) FGM has been fab-
Figure 20-25. Schematic diagram of a four-port plasma
spraying gun: (1) ceramic powder feeder, (2) metal
ricated by first preparing laminated sheets
powder feeder, (3) cathode, (4) anode, (5) FGM coat- of carbon-fiber fabric with each fabric
ing, and (6) substrate (Shimoda et al., 1990). having a different void fraction, then im-
 20.4 Fabrication Processes for Functional Gradient Materials 319

layers of particle size A when the molten Cu is infiltrated into the

tungsten powders particle size B
particle size C pores of the sprayed body. The cross-
sectional microstructure of this FGM is
shown in Fig. 20-27 (Takahashi, 1994).
pressing
Centrifugal Casting Technique
The centrifugal force enables the ceram-
ic powder mixed with a metal to create a
gradient compositional distribution due to
reduction of the difference in the material density. A
closed pores
thick-walled ring of SiC/Al FGM was fab-
ricated by a centrifugal casting technique
infiltration of tungsten/copper (Fukui et al., 1991). In this technique, SiC
copper by HIP gradient material
powder (10vol.%) is mixed with molten
black : Cu
white : W Al alloy at 900 °C. This molten mixture is
network structure
poured into the rotating mold and solidi-
mechanical finish
fied to form an FGM.
Figure 20-26. Sintered porous body-molten metal
infiltration technique for preparing W/Cu FGMs
(Takahashi et al., 1993).

mersing the laminated sheet in molten Si at

sus 318 L
1500°C, thus converting it into SiC/C Cu
FGM (KawaietaL, 1993).
SOS 316L
Plasma Sprayed Porous Coating-Molten
Metal Infiltration Technique
This method infiltrates molten metal in- SUS 318 L
to the pores by HIP. In this case the porous Cu
body with void fractions changing contin-
uously from the surface to the inner body &£?• 3US316L
is prepared by plasma spraying.
An SS/Cu FGM was produced using
this method (Takahashi, 1994). The SS po-
rous material with graded void fractions is M^M^J^ SUS 318 L
obtained by spraying SS316L powders
with different particle sizes (average parti-
cle size 8-335 jam) onto the SS316L sub-
strate, which is heated to 550 °C. The re- u SUS 3I6L
sulting porous sprayed body is heat-treat-
Figure 20-27. Cross-sectional view of an SS/Cu FGM
ed at 1000 °C for 1 h in a vacuum and then prepared by the plasma sprayed porous coating-mol-
HIPed together with Cu at 1100°C and ten metal infiltration technique (Takahashi, 1994). SUS:
196 MPa for 0.5 h. The final FGM results stainless steel.
320 20 Functional Gradient Materials

20.4.3 Solid Phase Methods sintering. Using this method, ZrO 2 and
NiCr alloy powders are mixed to obtain an
The powder metallurgical fabrication
FGM green body. This green body is cold
process accompanied by sintering is the
pressed at 180 MPa and sintered for 3 h in
most common solid phase process for the
Ar at 1400 °C to obtain bulk FGM (Cher-
preparation of FGM. This process re-
radi etal., 1993).
quires a proper compositional distribution
In the spraying technique a powder sus-
of more than two kinds of solid source
pension having a varied mixture ratio in
materials such as powder or fiber before
ethanol solvent is sprayed using a roller
sintering. The most common method in-
pump or a compressed air nozzle on the
volves the preparation of thin green sheets
preheated substrate. The resulting deposits
with different compositions, then laminat-
are then dried to obtain the FGM green
ing them according to the pre-designed
body. Kawasaki etal. (1993) studied the
compositional distribution profile to ob-
computer control of the mixture suspen-
tain a stepwise change in the material's
sion flow rate by use of a roller pump, and
composition (thin green sheet lamination
they found the necessary conditions to ob-
method). However, in recent times various
tain a desired thickness of FGM having a
compositional gradient mixing techniques
desired compositional distribution profile,
for obtaining a continuous change in the
and obtained PSZ/SS FGM. Figure 20-28
composition have been suggested.
shows their experimental apparatus
(Watanabe et al., 1991). A TiB2/Cu FGM
20.4.3.1 Methods of Obtaining green body is obtained by this technique
Compositionally Graded Green Bodies using 40 kPa compressed air (Sata, 1992).
Powder Stacking Techniques
The pressing technique is as follows:
Using proper materials (such as polymers,
metals, or ceramics), two or more different
powders or fibers are mixed at the desired
compositional ratio. The mixture ratio is
gradually changed in a die. Then pressure
is applied to the mixture to obtain an
FGM green body.
In the centrifugal technique a mixture of
source powder is supplied to the center of
a rapidly rotating centrifuge. The mixture
ratio of the powder is computer regulated.
The mixed powder is deposited on the in-
ner wall by the centrifugal force (Ilschner,
1990). For better stabilization of the con-
centration graded layer of the mixed pow-
der, the deposited layer is preheated slight- Figure 20-28. Schematic diagram of powder spraying
and stacking apparatus: (1) powder suspension, (2)
ly and liquid, hot wax is injected before
roller pump, (3) blower, (4) heater, (5) nozzle, (6) tem-
sintering. The FGM fabrication stages in- perature regulator, (7) computer, and (8) image ana-
clude: cold compaction, de-waxing, and lyzer (Watanabe et al., 1991).
 20.4 Fabrication Processes for Functional Gradient Materials 321

Powder Infiltration Technique Ni FGM results when the mixture is

pressed isostatically at 100-300 MPa and
After the reduction of metallic chlorides
sintered for 3 h at 1300 °C. The FGMs thus
by hydrogen, the metallic fine particles
formed exhibit well controlled composi-
subsequently formed by chemical vapor
tional distribution profiles which agree
reaction were infiltrated into a ceramic
well with theoretical filtration simulation.
powder packing. The control of gradients
was accomplished easily by changing the
purge gas flow rate and packing condi- Slurry Techniques
tions. An Al 2 O 3 /Ni FGM can be obtained
During the slip casting technique source
using this technique, as shown in Fig. 20-
powders are mixed with water soluble
29 (Mori et al., 1993). A12O3 powder with
binder or solvent binder to form a slurry.
an average diameter of 50 nm was formed
This slurry is then converted to a thin film
into agglomerates of 100-150 jum, and
by slip casting. An FGM green body is
these agglomerates were packed to a thick-
obtained by layering these films. Such a
ness of 5 mm in a quartz cell having an
laminated-type ZrO 2 /Ni FGM was pre-
inner diameter of 20 mm. Then, the metal-
pared by slip casting and pressureless sin-
lic nickel monomers formed by hydrogen
tering (Takebe et al., 1992). Aqueous slips
reduction of nickel chloride powder coagu-
containing ZrO 2 , Ni, and a mixture of
late by Brownian motion and infiltrate in-
ZrO 2 and Ni were prepared using 0.1 wt.%
to the A12O3 powder-packed layer. A12O3/
of an ammonia-based polyelectrolyte. The
H overall solid content of the resulting slips
2Ar Ar H 2
was 50 wt.%. Green compacts formed
when these slips were poured into a plastic
frame. Twelve layers of these green com-
pacts formed a multilayer compact which

•i was sintered for 2 h at 1400 °C. Practically

no cracks were observed in the resulting
body.
Al 2 O 3 /ZrO 2 FGM was prepared by first
sequentially slip casting eight different
aqueous slips having 70 wt.% solids on
plaster of Paris molds, then sintering for
2 h at 1550°C (Moreno et al., 1993; Moya
et al., 1992). The relative density of the re-
sulting composite was about 99.4% theo-
retical. Using the same technique, laminat-
ed-type Al2O3/mullite FGM was also ob-
tained (Moya et al., 1993). Al 2 O 3 /Al 2 TiO 5
FGM having six layers was prepared by
H2, Ar, HC1
slip casting Al 2 O 3 -TiO 2 slurries having a
Figure 20-29. Reactor for CVD powder infiltration different Al 2 O 3 /TiO 2 ratio (Requena
to produce Al2O3/Ni FGM: (1) furnace, (2) NiCl2, (3)
Ni particles, (4) quartz tnicrofiber filter, (5) perforated
etal., 1993). Fukui and Nakanishi (1991)
base, (6) vaporizer, (7) nozzle, (8) quartz cell, (9) A12O3 suggested that ceramic powder can be dis-
layer, and (10) support tube (Mori et al., 1993). tributed with a certain gradient by cen-
322 20 Functional Gradient Materials

trifuging the slurry-containing ceramic

powder.
A YSZ/SS FGM coating of thickness
0.25 mm was obtained on an SS rod using
the slurry dipping technique (Yamaoka
et al., 1993). The synthesis process in-
volved first preparing five compositions of
ethanol slurries containing YSZ and SS
powders, and then dipping the SS rod into Figure 20-30. Equipment for preparing piezoelectric
each slurry. After CIP and HIP (under FGM plates with a cyclically changing composition:
196 MPa and 1200 °C) treatment the final (1) dispenser, (2) nozzle, (3) FGM sheet, and (4) X-Y
stage (Kawai and Miyazaki, 1990).
YSZ/SS FGM was obtained. A thermal
stress analysis calculation gives a composi-
tional distribution function exponent n of B A B A B A B A B
2 for this particular case. This method is
most suited for the formation of an FGM
coating on a curved surface. A B A B A B A B A
In the doctor blade sheet-forming tech- Figure 20-31. Structure of piezoelectric FGM film:
nique, a slurry containing source powder, each boundary is compositionally graded (Kawai and
binder, plasticizer, deflocculant, and sol- Miyazaki, 1990).
vent is pasted uniformly on the substrate
film using doctor blade equipment to ob-
tain green sheet. Multilayers of these In the filtration of a two component sys-
sheets each having different compositions tem, compositionally-graded cake layers
were dried, de-greased, and sintered to ob- can be obtained by continuously changing
tain an FGM. To avoid cracks forming the mixture ratio of two kinds of slurry as
during sintering, careful adjustment of the they are fed into the tank and filtrated.
binder content and powder particle size is Al 2 O 3 /Ni system cake layers were pre-
necessary. A laminated-type (5 layers) pared using this filtration technique (Iwata
ZrO 2 /Ni FGM was prepared by this tech- etal., 1992).
nique (Takemura et al., 1990). FGMs can be prepared by a simple sed-
Figure 20-30 shows the equipment for imentation followed by a sintering or hot
preparing a piezo-ceramic actuator FGM pressing process. The principle here is to
plate (Kawai and Miyazaki, 1990). The utilize the difference in the sedimentary ve-
first sheet is obtained using the doctor locity between ceramic and metal particles,
blade technique. The second sheet is pre- or between the large and small particles
pared on the first sheet before drying by within the slip suspension. The 50/50 vol-
alternating two kinds of cartridge filled ume ratio of A12O3 and NiAl powders is
with two kinds of source powders. Two mixed and thoroughly agitated, and sedi-
kinds of slurry are mixed at the boundary ments are obtained in the sedimentation
between the two sheets to form a composi- column. The solvent is then evaporated
tional gradient region, as shown in Fig. 20- and the remainder is dried. The specimens
31. It is easy to control the piezoelectric are then hot pressed for 4 h at 1500 °C in a
characteristics by designing such a periodi- vacuum to obtain Al 2 O 3 /NiAl FGM. The
cally changing FGM structure. moisture on the surface of the particles and
 20.4 Fabrication Processes for Functional Gradient Materials 323

the degree of agitation can strongly in- the die followed by CIP at 140 MPa. The
fluence the results (Miller etal., 1992). sintering was carried out at 1400 °C for 3 h
A12O3/W FGM is obtained using a similar in Ar, or by HP at about 1300°C for 1 h at
approach (Chu et al., 1993). a pressure less than 10 MPa, or by HIP at
about 1350 °C and 100 MPa. They con-
20.4.3.2 Sintering Methods cluded that for the best quality FGM,
careful selection and control of the powder
Normal Sintering and High Pressure
particle size ratio and the preparation of a
Sintering Techniques
proper compositional distribution profile
On preparing FGM by sintering cracks are necessary.
can often appear on the sintered body due In the case of a laminated-type MgO/Ni
to differences in the sintering characteris- FGM, the one having the compositional
tics and the mixture ratio of the two source distribution function exponent n of unity
powders. Watanabe et al. (1991) controlled has been prepared by an HP method by
the shrinkage of the powder mixture by first vacuum sintering at 1320 °C and then
blending fine and coarse particle sizes. hot pressing at 28 MPa and 1300°C (Yuan
Rabin and Heaps (1993) prepared six- et al., 1993). Tang et al. (1993) carried out
layered Al 2 O 3 /Ni FGM using the powder the thermal stress analysis of YSZ/Mo
processing method shown in Fig. 20-32. FGM and concluded that a compositional
Graded compacts were produced by se- distribution exponent n of 1.6 is best. They
quentially layering the powder mixture in produced a 15-layered FGM based on this

add sintering aids for ceramic

mix intermediate compositions

Figure 20-32. Diagram

showing the processing
steps involved in fabricat-
ing FGMs by sintering
(Rabin and Heaps, 1993).

sinter hot press HIP

324 20 Functional Gradient Materials

finding by vacuum sintering and HP at Plasma Activated Sintering Technique

1500 °C. Various ceramic/metal system lam-
In plasma activated sintering an instan-
inated-type FGMs are produced by the HP
taneous pulsed electric current is applied
method; for example, YSZ/SS (Kawasaki
to the subject powder to initiate current
and Watanabe, 1993) and Si3N4/super-
discharges in the voids between the powder
alloy (Hiilsmann and Bunk, 1993). SiC-
particles. Successful sintering can be ac-
AIN/AIN/Mo FGM was prepared by HIP
complished by use of the heat generated by
at 1850 °C and 200 MPa (Kawasaki et al.,
these discharges. Due to this current dis-
1993).
charge, the powder surface is purified and
Temperature Gradient Sintering Technique activated. This method requires a relative-
ly short time for sintering, and thus it is
In preparing an FGM using source pow-
easier to control the grain growth. This
ders having a wide difference in their sin-
method is also suitable for sintering mate-
tering temperatures (for example, ceramic/
rials with lower melting points (Bennett
metal system), one sintering temperature
etal., 1968).
cannot be used to obtain a good quality
Using this technique a laminated-type
sintered body due to the difference in
(8-layers) FGM of the YSZ/SS410 system,
shrinkage characteristics of each source
which has a wide difference in sintering
powder. To overcome this problem, the
temperature between the two sources, was
portion containing more of the higher sin-
produced. In this process the YSZ side was
tering temperature source is sintered at a
heated to 1200°C, while the SS side was
higher temperature and the portion which
heated to 1000 °C using a specially shaped
contains more of the lower sintering tem-
graphite resistance heater, in order to ob-
perature source must be sintered at a lower
tain a temperature gradient (Omori et al.,
temperature. That is, the sintering must be
1994). PSZ/TiAl FGM (5-layers) (Kimura
accomplished under a prescribed tempera-
and Kobayashi, 1993) and PSZ/Ti FGM
ture gradient. One way to achieve this tem-
(Abe and Hashimoto, 1991) were prepared
perature gradient is to apply additional
by using the same technique. Figure 20-33
heat by use of a laser beam or infrared
shows a schematic diagram of the tempera-
beam onto one side of the sample while it
ture gradient plasma sintering apparatus.
is in the sintering furnace.
It is also possible to assign a gradient to
the density by the use of temperature gra-
20.4.3.3 Self-Propagating High-
dient sintering. Kawasaki and Watanabe
Temperature Synthesis (SHS) Methods
(1990) sintered a cylindrical PZT powder
compact in air for 1 h. They heated one This technique uses an exothermic re-
side of the sample by an infrared lamp, and action at temperatures exceeding 2000-
created a temperature gradient of 150°C 3000 °C to obtain reaction products at a
within a depth of 5 mm from the sample relatively high speed. Because of its high
surface. The resulting sintered body was of speed reaction the diffusion of atoms is
higher density in the upper portion, while prevented and thus it is possible to obtain
the lower portion was of lower density. a graded composition. When pressure (wa-
Such a density difference gives a continu- ter pressure, gas pressure, etc.) is applied
ous change in the piezoelectric characteris- during the SHS process the synthesis of
tics within the material. dense composites can be achieved.
 20.4 Fabrication Processes for Functional Gradient Materials 325

Radhakrishnan et al. (1993) obtained a

compact via SHS using a mixture of Ti and
B powders at the mixture ratio of 80/20
and adding 20-30 wt.% Cu. During the
combustion synthesis, the distribution of
molten Cu changes continuously within
the compact due to gravity. By HIP treat-
ment of the resultant sample at 1080 °C
and 140 MPa, they obtained dense TiB2/
Cu FGM. A TiC/NiAl FGM layer can be
produced by SHS of Ti, C, Ni, and Al
powders at the junction of TiC and NiAl
(Kudesiaetal., 1992).

Gas-Pressure Combustion-Sintering
Technique

Figure 20-33. Schematic diagram of the temperature First the source powders are formed into
gradient plasma sintering apparatus with a specially a compact by CIP at 250 MPa and sealed in
shaped graphite susceptor: (1) graphite die, (2) FGM, a glass container under vacuum. This glass
(3) ram, (4) graphite plate, (5) pyrometer, and (6) vac- container is then embedded into the igni-
uum chamber (Omori et al., 1994).
tion agent consisting of Ti and C powders
packed in a graphite crucible placed within
the HIP apparatus. Next the container is
heated to 700 °C in Ar at 100 MPa and
Gas Reaction Sintering Technique
then the contents ignited (Miyamoto et al.,
Ni powder and Al fine powder (0.42 (im 1990). TiB2/Ni, TiC/Ni, Cr 3 C 2 /Ni, and
in diameter) are sintered in N 2 using the MoSi 2 -SiC/TiAl FGMs are some exam-
exothermic reaction of A l - N 2 to obtain ples of FGMs prepared using the gas-pres-
AIN/Ni FGM. Using a similar technique, sure combustion-sintering technique (Miya-
A1N/A1 FGM can also be produced moto et al., 1992).
(Atarashiya et al., 1993).
20.4.3.4 Martensitic Transformation
Hydrostatic Compression SHS Technique Technique
TiB 2 , Ti, B, and Cu powders are used as Watanabe et al. (1993) have attempted
the source for preparation of TiB2/Cu to prepare an FGM using crystallographic
FGM by this technique (Sata, 1993). Us- transformation. The paramagnetic phase in
ing the automatic powder spraying and austenitic stainless steel (Fe-18Cr-8Ni)
stacking device, these powders were transforms into the ferromagnetic a' mar-
sprayed onto a Cu substrate. The stacked tensitic phase by plastic deformation. The
compact was then ignited at room temper- amount of martensite increases with in-
ature under a high hydrostatic pressure of creasing deformation (strain). Thus the
58MPa. TiB2/Cu FGM with diameter saturation magnetization of the deformed
30 mm and thickness 1 mm was obtained austenitic stainless steel increases with in-
on the Cu substrate. creasing strain. Using this phenomenon a
326 20 Functional Gradient Materials

magnetic gradient function can be as- stress in the actuator, thus preventing
signed by inhomogeneously deforming the crack formation (Kawai et al., 1990).
stainless steel. Figure 20-34 shows a sam- An FGM can be prepared by means of
ple of SS304 intended for deformation, a chemical reaction on a material's surface.
and Fig. 20-35 illustrates the relationship In an attempt to improve the oxidation
between the deformation and the satura- resistance of carbon material, a carbon
tion magnetization within the sample. substrate was heated in silicon powder at
1450 °C for 3 h to form a C/SiC FGM on
20.4.3.5 Diffusion and Reaction Techniques the surface (Yamamoto et al., 1993).
A12O3 and A12O3 can be joined by using
A newly developed ceramic actuator
a transient liquid phase reaction technique
was prepared by the diffusion bonding of
(Glaeser et al., 1993). In this process C u -
two plates having different piezoelectric
N b - C u foil is inserted between the A12O3
constants at 1200°C for 3-5 h. This
layers and part of the foil is melted. As it
attempt was made using the PZT-Pb
melts, it reacts with both sides of the A12O3
(Ni 1/3 Nb 2/ 3)O 3 system. The composition-
in a vacuum hot press at 1150 °C to form a
ally-graded intermediate layer between the
graded compositional distribution at the
two plates tends to reduce the residual
boundaries.
Powder mixtures of Al with 10, 30, and
50 wt.% SiC are fed into a 2 kW continu-
ous wave CO 2 laser beam, which is focused
on an Inconel 625 alloy substrate. The
powder mixtures are melted onto the sub-
strate and FGMs are formed (Jasim et al.,
1993).

Figure 20-34. An example of the SS304 specimen 20.5 Characteristics and Future
used for deformation; dimensions are in millimeters
(Watanabe et al., 1993). Development of Functional
Gradient Materials

FGMs have been developed in an effort

to reduce the thermal stress for thermal
barrier materials which encounter a large
temperature difference across the material.
There are many studies on FGMs that fo-
cus on the thermal and mechanical aspects
of material characteristics. However, re-
cently the use of FGMs is proliferating as
their advantages are realized. FGMs have
0 10 20 30 40 50 been demonstrated to be advantageous be-
distance (mm) yond mechanical applications (such as in
Figure 20-35. The distribution of (a) the plastic strain structural material in a high temperature
and (b) the saturation magnetization of SS304 (Wata- environment) extending to electronic, opti-
nabe et al., 1993). cal, nuclear, biomedical, and other fields.
 20.5 Characteristics and Future Development of Functional Gradient Materials 327

This section introduces the various charac-

teristics of an FGM and also discusses the
future possibilities for FGMs.

20.5.1 Thermal Stability

(Time Variation of Compositional
Distribution Under Temperature Gradient)
When an FGM is used as a thermal bar-
rier material for a spacecraft it experiences
a temperature gradient exceeding 100 °C/
mm within the material. As a consequence,
a significant thermal diffusion derived by
the temperature gradient develops and the
Figure 20-36. Schematic diagram of test apparatus
compositional gradient within the FGM
for creating a large temperature difference within a
can change with time lapse. specimen: (1) liquid H 2 or N 2 tank, (2) to stack, (3) TV
The composition distribution changes camera, (4) radiation pyrometer, (5) vacuum vessel,
by thermal diffusion for TiC/Ti and V2C/V (6) mirror, (7) sample, (8) vacuum pump, (9) shutter,
FGMs were analyzed numerically using a (10) mirror, (11) Xe arc lamp, and (12) DC power
supply (Kumakawa et al., 1990 a).
finite difference method. The carbides in
the TiC/Ti FGM were condensed on the
high temperature side, and the carbides in layered samples. The thermal insulation
the V2C/V FGM were condensed on the characteristics are determined using the
lower temperature side. As a result, sepa- apparatus shown in Fig. 20-36. A xenon
ration into carbide and metallic layers has arc lamp (nominally, 1 A, maximum input
been predicted (Aihara etal., 1990). power 30 kW) was used as a heat source to
A PSZ/NiCrAlY FGM prepared by a heat one surface. The other surface was
plasma twin torches spraying technique cooled by liquid nitrogen through the cop-
(see Sec. 20.4.2.4) was found to be stable per sample holder. The heat flux in this
and provide a potentially excellent, protec- case was 0.7 MW/m 2 . An environment
tive coating in the gradient temperature with a maximum temperature of 1727 °C
field of 1200-600°C (Shinohara etal., and a temperature difference of 1000 °C
1990). Thermal stability depends strongly was simulated in this manner. The heat
on the system components, and a ceramic/ flux through the sample holder was deter-
ceramic system exhibits excellent high tem- mined using three thermocouples embed-
perature stability. For example, an SiC/C ded at equal intervals along the central axis
FGM shows good stability at a tempera- of the sample holder.
ture of 1500° C. The effective thermal conductivity, KtU,
which represents the characteristics of a
20.5.2 Thermal Insulation Characteristics thermal barrier, was measured under a
For the determination of thermal diffu- steady state heating condition (Kuma-
sivity, N. Araki etal. (1993) derived an kawa et al., 1990a). The KeH is expressed
analytical solution for the temperature re- as
sponse in an FGM under transient heating (20-4)
based on analyses conducted on the multi- Ts-TJd
328 20 Functional Gradient Materials

where q is the heat flux within the sample, 20.5.3 Thermal Fatigue
d is the thickness, Ts is the top surface and Thermal Shock Resistance
temperature, and Th is the bottom surface
temperature of the sample. When a functional gradient material is
Sasaki and Hirai (1990) measured the used as a super high temperature structural
temperature drops in the SiC/C FGM and material, an unsteady thermal stress devel-
SiC coatings that were formed on a graph- ops within the material as its surfaces expe-
ite plate by CVD (see Sec. 20.3.3.1). The rience rapid heating and cooling. Evalua-
temperature difference increased as the tion of the thermal resistance and the ther-
temperature of the specimen increased, as mal shock characteristics of an FGM re-
shown in Fig. 20-37. The temperature dif- quires a thermal environment with a high
ference for the SiC sample was 100-200 °C, heat flux. Thus a xenon lamp or laser heat-
while that for the SiC/C FGM was 230- ing is used to heat the desired portion of
310 °C. This difference demonstrates the the sample. For the evaluation of the ther-
superior thermal barrier characteristics of mal shock resistance in an oxidative envi-
the SiC/C FGM. ronment, a burner test is also sometimes
In the evaluation of thermal insulating used.
characteristics, the effective thermal con- Using the test apparatus shown in
ductivity estimated under a constant heat Fig. 20-36, a maximum heat flux of 5 MW/
flux and cooling is usually used as a criteri- m 2 can be obtained. Cyclic heating to in-
on. Matsuzaki et al. (1993), however, used duce thermal fatigue was accomplished by
a steady heat transfer model in order to opening and closing a shutter between the
simulate a more realistic environment, and reflectors. Figure 20-38 illustrates how this
evaluated the thickness of the FGM and its cyclic heating can cause a decrease in the
compositional distribution under convec- value of the effective thermal conductivity
tion heating. of a CVD SiC/C FGM and a non-FGM
SiC. When the effective thermal conductiv-
ity decreases there is a possibility of crack
,300 _ SiC/CFGM formation in the specimen. Indeed, for a
non-FGM SiC sample some cracks were
observed at the boundary between the SiC
coating and the graphite substrate after
30-40 heating cycles. The heating cycle
| 200
consisted of heating the top surface to
8, 1427-877 °C, while the bottom surface
SiC NFGM was heated to 921-621 °C under vacuum.
The heat flux for the cycle was 0.7 MW/
100 - m 2 . The cracks are believed to be due to
1150 1200 1250 1300 1350 1400 cyclic heating. In contrast, the CVD SiC/C
average temperature (°C) FGM sample showed no sign of cracks or
Figure 20-37. Relationship between the temperature damage (Sasaki and Hirai, 1990,1991; Ku-
difference in CVD SiC/C FGM and CVD SiC coat-
makawa et al., 1990b).
ings (0.4 mm in thickness), and the average tempera-
ture under steady thermal exposure at a heat flux of The relationships between thermal fa-
0.7MW/m2. Here, 7>1427°C and Tb=1027°C tigue and microstructure were studied for a
(Sasaki and Hirai, 1990). CVD SiC/C FGM coating which was de-
 20.5 Characteristics and Future Development of Functional Gradient Materials 329

Ni-20wt.% Cr alloy spray-coated under

layer on a Cu substrate showed excellent
fatigue properties. This film also had excel-
lent heat resistance, as well as excellent
thermal barrier properties, even after re-
peated exposures to a laser (5 kW CO 2 la-
ser: temperature gradient of 400 °C/mm)
(Shimoda etal., 1990). Similarly, a YSZ/
NiCrAlY FGM coating formed on an SS
substrate by plasma spraying had superior
10 20 30 40 thermal shock resistance when compared
cycle (number) to a single layer of YSZ (Fukushima et al.,
Figure 20-38. Dependence of the effective thermal con- 1990).
ductivity on the cyclic thermal exposures at a heat Watanabe and Kawasaki (1992) con-
flux of 0.7 MW/m2 for a CVD SiC/C FGM (Sasaki structed a burner heating apparatus and
and Hirai, 1990).
evaluated the thermal and mechanical
properties of FGMs. The temperature dif-
ference for the two sides of the sample was
posited onto a graphite or a C - C com- created by heating the ceramic side with a
posite substrate by CVD using SiCl4 and burner flame (H 2 + O 2 , C 3 H 8 , etc.) while
CH 4 as source gases. Numerous pores cooling the other side with water. Using
about 10 jim in size exist at 40-60 mol% C this setup the effects of repeated heating
and provide stress relaxation, thus pre- and cooling on the YSZ/SS FGM pre-
venting cracking. These pores are believed pared by HP was studied (Kawasaki and
to result from changes in the crystal Watanabe, 1993). The critical surface tem-
growth mechanism which accompany the perature at which cracks appear on the
change in concentration of the source gas. YSZ side due to the severe temperature
For this reason, the SiC/C FGM with pores gradient was found to be 1027 °C. The re-
showed a higher thermal resistance and lationship between crack occurrence and
longer thermal fatigue life than the SiC sin- compositional distribution was studied by
gle-coated specimen (Sasaki et al., 1989; thermal stress analysis. The convex compo-
Sohdaetal., 1993). sitional distribution profile (n= lA) showed
The thermal fatigue properties of four extensive crack deflections.
types of FGM specimen: TiB2/Cu by SHS, A Cr 3 C 2 /Ni FGM prepared by gas pres-
TiC/Ni by gas pressure combustion sinter- sure combustion sintering (see Sec. 20.4.3.3)
ing, PSZ/NiCr by low pressure plasma having n values of 1 and 0.5 was subjected
spray coating, and SiC/C-C by CVD, were to the burner heating test, and the results
measured using the apparatus shown in showed that the FGM withstood the sur-
Fig. 20-36. It was clearly seen that the face temperature of up to 900 °C, while the
FGM layer not only prevented the delami- temperature drop through this FGM was
nation of the coated layers but also prohib- about 750°C. (Tanihata etal., 1993).
ited propagation of the microcracks (Ku- A heating test using a C 3 H 8 /O 2 flame
makawa etal., 1993). on a PSZ/NiCr FGM formed by the cen-
FGM film consisting of a YSZ low pres- trifugal process (see Sec. 20.4.3.1) was per-
sure plasma spray-coated top layer and a formed by Cherradi et al. (1993). Thermal
330 20 Functional Gradient Materials

fatigue tests applying cyclic heating of 350 for up to several seconds using an electron
and 1450°C during an interval of 3-6 s beam irradiation system and a particle
cycle duration showed a clear advantage of beam engineering facility (M. Araki et al.,
the FGM. Functional gradient A12O3/ 1994). Both FGM coatings withstood a
NiAl produced by sedimentation in organ- temperature difference as high as 1500°C
ic solvents (see Sec. 20.4.3.1) followed by without cracking or melting. In addition,
HP at 1500 °C showed good thermal shock many other studies on thermal shock resis-
resistance against 100 cyclic heatings of tance have been conducted on various
30-815 °C using a C 3 H 8 torch (Miller types of FGM using various research tech-
etal., 1992). niques. These studies tend to suggest the
In other fatigue studies, a standard sam- effectiveness of FGMs.
ple having a homogeneous density of 70 % Erdogan and co-workers (Erdogan and
theoretical and a test PSZ FGM film sam- Qzturk, 1993; Erdogan and Wu, 1993;
ple having a continuously changing densi- Erdogan and Chen, 1993) have studied
ty from the surface to the metal-ceramic the fracture mechanics of FGM coatings.
interface of 70-84% theoretical were pre- Kokini et al. (1993) have studied the mech-
pared by EB-PVD on a superalloy sub- anisms of crack initiation or propagation.
strate (see Sec. 20.4.1.2). Both samples However, details of these studies will not
were tested using a Mach 0.4 burner rig be discussed in this chapter.
under one 1 h cyclic heating from room
temperature to 1150°C. Failure of the
20.5.4 Resistance to Extreme
standard sample is believed to be due to
Environmental Changes
continuous material loss, segmentation
cracking, and rapid spallation. In contrast, One of the main objectives of study us-
the FGM sample performed in a much ing FGMs is to develop structural materi-
more superior manner (Fritscher and als that can be used for the outer body
Schulz, 1993). surface or as an engine structural material
This ability to reduce the thermal stress of a spacecraft. For these applications it
by using an FGM was studied using a becomes very important to properly evalu-
W/Cu FGM prepared by the sintering-in- ate the material's performance in extreme-
filtration technique (Takahashi et al., 1993) ly severe environments.
(see Sec. 20.4.2.5). The heating and cooling
cycle was repeated 10 times over the tem-
20.5.4.1 High Temperature Supersonic
perature range of 800-300 °C in Ar. The
Gas Flow
results showed that W/Cu FGM, which
has good potential for reducing thermal Wakamatsu et al. (1993) have developed
stress, can be used in applications in which an apparatus capable of generating a Mach
the material is exposed to a plasma beam, 3 gas flow containing roughly 21 vol.% ox-
an ion beam, or an electron beam. ygen as a standard atmosphere. The system
In order to develop plasma-facing mate- has a stagnation temperature of 2727 °C
rials for fusion applications, the thermal and a stagnation pressure of 1.5 MPa,
resistivity of CVD SiC/C and TiC/C FGM which is produced by combusting a mix-
coatings (1 mm thick) deposited on a ture of H 2? N 2 , and O 2 . An FGM speci-
graphite substrate were evaluated. A sur- men was placed in this test apparatus and
face heat flux of up to 70 MW/m 2 was used its ability to withstand this extremely high
 20.5 Characteristics and Future Development of Functional Gradient Materials 331

speed and high temperature environment ture which allows the construction of an
was tested. actively cooled panel.
In order to test a model component of a A test sample consisting of a TiAl cool-
rocket nose cone, a 50 mm diameter, hemi- ing structure of diffusion-bonded, 11-lay-
spherical C - C composite was fabricated ered PSZ/TiAl FGM prepared by HP at
from a three-dimensionally structured fab- 1200 °C and 20 MPa for 2 h was exposed
ric resembling a net, made of pitch-based to a high temperature supersonic gas flow
carbon fibers (Sohda etal., 1993). The (a heat flux of about 4 MW/m 2 ) using the
SiC/C FGM was then coated on a C/C apparatus described above (Matsuzaki
composite by penetrative CVD, while non- et al., 1993). This test showed the excellent
FGM SiC was coated to a thickness of performance of this sample.
100 jim on SiC/C FGM by CVD (see
Sec. 20.4.1.1). These specimens were ex-
20.5.4.2 Bipropellant Rocket Combustion
posed to this supersonic gas flow at a tem-
perature of approximately 1627°C for Gas Flow
60 s. Figure 20-39 shows this test. The or- In order to evaluate the resistance of
dinary SiC-coated sample broke on only FGM to the combustion gases, as in the
one exposure to this environment. How- application as a rocket combustor, high
ever, the sample protected by layers of temperature environmental tests were con-
FGM between the C - C composite and ducted. Nitrogen tetroxide (NTO) and
SiC coating showed no discernible change monomethyl hydrazine (MMH) (NTO/
in its structure even after ten cycles. There MMH = 1.76) were used in these tests
was very little weight loss due to this test. (Kurodaetal, 1992).
When an FGM is used as part of the Rocket-combustor-shaped C - C compo-
combustion chamber wall or as part of a sites were fabricated from a fabric of pitch-
strut for a spacecraft, the far side of the based carbon fibers coated with SiC/C
high temperature wall is often actively FGM by CVD. The model combustor was
cooled by liquid hydrogen. An FGM for exposed to firing for a period of 55 s, as
this application is required to have a struc- shown in Fig. 20-40. The maximum outer

}''£'"'$&'& "'"'' ' "';

Figure 20-39. The 50-mm diameter, hemispherical, Figure 20-40. Firing test for the model rocket combus-
CVD SiC/C FGM nose cone under high-temperature tor made of C-C composite coated with CVD SiC/C
supersonic gas flow (Sohda et al., 1993). FGM (Suemitsu et al, 1993).
332 20 Functional Gradient Materials

wall temperature of the model combustor The C - C composite's cyclic oxidation

was 1377-1627 °C, while the inner wall resistance can be notably increased by
maximum temperature reached 1667- coating its surface with a coating com-
2027 °C. No damage to the test model was posed of 30 jam thick C/SiC FGM and
observed after two cycles of testing, and 100 jim thick SiC using the CVD tech-
cooling with liquid N 2 at a rate of 10 g/s nique. The heating cycle for this experi-
after each testing. The FGM rocket com- ment was in the temperature range of 350-
bustor showed excellent performance in 1525°C(Kude, 1993).
these tests (Suemitsu et al., 1993). The oxidation rates of carbon fiber-rein-
FGMs are investigated in other real life forced ceramics coated with SiC and those
environmental tests such as a plasma-arc coated with the TiC/SiC FGM were stud-
heating test, an erosion/corrosion test, an ied in air at 1300°C. The results showed
arc-heated wind tunnel test, a high temper- that the ceramics coated with the TiC/SiC
ature rotating members test, and a cylin- FGM had superior oxidation resistance
drical high pressure thrust chamber test. (Kawaietal., 1990).
The strength and oxidation resistance of
20.5.5 Oxidation and Corrosion Resistance an SiC-SiC composite prepared by CVI
on a substrate made of SiC/C FGM-coat-
When an oxidation resistant film is coat- ed SiC fiber were studied (Agullo et al.,
ed on the substrate, the graded properties 1993). The study showed that the FGM
of the coating are effective, not only in coating is best suited for improving the
relaxing the thermal stress arising at the toughness of SiC-SiC composites used in
boundary of the coating and the substrate, a high temperature oxidative atmosphere.
but also in preventing the occurrence of For use in a thermochemical dissolution
micro-cracks. This is accomplished by in- cycle (UT-3) of water, an SiC/TiC FGM
troducing a compressive stress onto the was prepared by CVD on stainless steel
surface layer during the cooling stage fol- (SS304). In the UT-3 cycle the material
lowing the completion of film synthesis. must withstand an extreme environment,
These phenomena improve the material's such as a temperature exceeding 1000°C
oxidation resistance. and a corrosive atmosphere of B r 2 - O 2 -
The oxidation resistance of SiC/C FGM HBr. A vastly improved corrosion resis-
prepared on a graphite substrate by the tance was observed for the SiC/TiC FGM-
chemical gas reaction betwen graphite and coated SS in both isothermal and cyclic
gaseous silicon monoxide was examined in tests when compared to the TiC or SiC
the atmosphere at 800 °C. The SiC/C single layer coated SS (Sasaki et al., 1993).
FGM exhibited superior oxidation resis-
tance compared to virgin graphite (Fujii
20.5.6 Electrical Properties
etal., 1992). Similarly, the SiC/C FGM
prepared on an isotropic graphite sub- A piezoelectric FGM plate having con-
strate using the reaction effect between tinuously changing material properties in
graphite and silicon powder also exhibited the direction of the plate thickness was
excellent oxidation resistance at 1400 °C. studied by wave propagation analyses. The
The weight loss rate was found to be only wave frequency spectrum, the energy
9% of that of the original graphite (Yama- propagation speed, the electromechanical
moto et al., 1993). coupling constants, and the mode shapes
 20.5 Characteristics and Future Development of Functional Gradient Materials 333

were determined by these analyses. The re- bration mode was generated on account of
sults showed that the gradient properties the density gradient.
can significantly improve the effectiveness
of surface wave devices (Tani and Liu,
20.5.7 Thermoelectric Properties
1993).
Figure 20-41 compares the structure of a The following equation defines the ther-
traditional piezoelectric actuator and that moelectric figure of merit Z
of an FGM actuator (Kawai and Miyaza-
ki, 1990). Figure 20-41 (b) shows a piezo-
(20-5)
ceramic actuator FGM plate consisting XQ
of two kinds of piezoelectric ceramics.
Plate A tPb(Ni1/3Nb2/3)0!5(Ti0.7Zr0.3)0.5O3] where Q is the Seebeck coefficient, Q is the
has a high piezoelectric modulus but electrical resistivity, and x is the thermal
a low dielectric constant, while plate B conductivity. All these parameters are
[Pb(Ni1/3Nb2/3)0.7(Ti0.7Zr0.3)0j3O3] has a strongly temperature dependent.
low piezoelectric modulus but a high One of the problems in the design of a
dielectric constant. The test plate was thermoelectric transformation device is
25 x 5 x 0.65 mm 3 in size. The bending dis- this strong temperature dependence of Z.
placement was about 60 jim when 100 V In the device tested, Z changes with tem-
was applied at room temperature. This perature, so the maximum value of Zis not
composition grading improves the durabil- obtained over the whole device. Tradition-
ity and reliability of an actuator. ally, this problem was avoided by joining
Kawasaki and Watanabe (1990) pre- several different materials, each having a
pared a density-graded PZT plate (see high value of Z over a certain temperature
Sec. 20.4.3.2) and found that a flexural vi- range, to form the device. However, this
joined-type device faces a sudden change
in the value of Z at the joint, and thus a
large amount of Peltier heat is generated
(a) and the thermoelectric transformation effi-
ciency is lowered. If a functional gradient-
type device for which Z changes with the
temperature distribution within the mate-
rial can be designed, then the generation of
Peltier heat will be much lower and thus
the thermoelectric transformation efficien-
cy will increase.
For example, as shown in Fig. 20-42, the
temperature at which Z shows a maximum
value changes with the carrier concentra-
Figure 20-41. Structures of piezoelectric actuators:" tion in n-type PbTe (Goff and Lowney,
(a) traditional bimorph-type actuator and (b) FGM 1976). It is possible to design a material
actuator; (1) electrode, (2) piezoelectric ceramics, (3) that has large Z values over the whole tem-
PZT-NiNb with higher piezoelectric constant (plate A),
(4) PZT-NiNb with lower piezoelectric constant perature gradient region by changing the
(plate B), (5) metal plate, (6) bonding agent, and (7) concentration of dopant or carrier contin-
FGM layer (Kawai and Miyazaki, 1990). uously in a material.
 334 20 Functional Gradient Materials

2.0 electric/thermionic hybrid direct energy

) FGM
conversion system (HYDECS) (Niino and
xl-5 > 4 x i o 1 0 ,. Chen, 1993). Figure 20-44 summarizes the
CO
concept HYDECS. For the energy source
o there is solar energy and nuclear energy. A
system with such an energy source could
Jl.O - 1 %^ be exposed to temperatures ranging from
1
°0.5 ! i '• room temperature to about 1700°C, thus
u*
bJ)
&> ' experiencing a temperature difference of
about 1700 °C. The application of func-
200 400 600 800 tional gradient materials in such a system
temperature (°C) is an attempt to reduce the thermal stress
Figure 20-42. Variation of the figure of merit (Z) of generated in the device at the same time
n-type PbTe with carrier concentration and tempera- while, attaining the highest energy conver-
ture; the dashed line indicates the variation of Z of the sion efficiency (exceeding 40%) over the
FGM with gradient carrier concentration (Nishida,
1993). entire temperature range. The energy con-
version mechanism of the traditional ho-
mogeneous material is effective only over a
A Bi 2 Te 3 -PbTe-Si Oi7 Ge o<3 -type FGM narrow energy band, while an FGM can
was designed to function over a tempera- accomplish this task over a much wider
ture gradient of 27-1027 °C. The Z for an energy band. A dramatic increase in con-
ideal FGM was calculated. The result version efficiency can be expected by using
showed that the FGM's thermoelectric an FGM.
transformation efficiency is 29% better
than that of Si0 7 Ge 0 3 (Hirano et al.,
1993). The modeled FGMs are shown in 20.5.8 Magnetic Properties
Fig. 20-43. Osaka et al. (1990) prepared a CoNiReP
Thermoelectric FGMs may play an im- FGM film by electroless plating (see
portant role in a future energy conversion Sec. 20.4.2.1). In this film the structure
system, that is, a high-efficiency thermo- changes from a randomly oriented f.c.c.
structure to an <002> oriented h.c.p. struc-
ture in the thickness direction. They re-
ported that by changing the composition
and the structure of the film, a high density
magnetic recording medium can be ob-
tained. A magnetic FGM was prepared by
applying a gradual inhomogeneous defor-
mation to SS304 austenitic stainless steel
(Fe-18Cr-8Ni) (see Sec. 20.4.3.4). A
magnetic FGM can be used as a position
measuring device by combining it with a
100 200 300 500 700
Temperature (°C) magnetic sensor. For example, it can be
Figure 20-43. Relationship between ZT values and used as a device that determines the focal
temperature (T) of FGMs designed from the Bi2Te3- point in an automatic focusing camera
PbTe-Si0 7 Ge 0 3 system (Hirano et al., 1993). (Watanabe et al., 1993).
 20.5 Characteristics and Future Development of Functional Gradient Materials 335

FGM (inside) FGM radiator

thermionic converter

FGM heat collector

Figure 20-44. Concept of

the hybrid direct energy
conversion system using
FGMs (Niino and Chen,
1993).

energy source
(R I and solar energy) -1700°C -1100 °C r.t.

20.5.9 Optical Properties tion distribution in the thickness direction

that allows the convergence of light in that
In 1965, Kawakami and Nishizawa direction, has been developed.
(1965) proposed an optical fiber with a Most commercially available polymeric
continuously changing refractive index in optical fibers are of the step index type.
order to increase its transmission capacity. The bandwidth of the step index polymeric
It has a higher refractive index at the core, optical fiber is about 5 MHz km. In com-
while its sheath has a lower refractive in- parison, a graded index polymeric optical
dex. This type of optical fiber is called a fiber with a refractive index that changes
graded index-type optical fiber (Gl-type). from the outer perimeter to the core of the
The Gl-type fiber has been shown to be fiber has a bandwidth of approximately
capable of carrying more than 10 times the 1 GHz km, which is roughly 200 times that
information compared to the step index- of the step index type polymeric optical
type optical fiber. Historically, Gl-type fiber. These graded index fibers are pro-
fibers are the first example of a successful duced by the interfacial-gel copolymeriza-
application of the graded property con- tion technique (see Sec. 20.4.2.3). The min-
cept. imum attenuation of transmission of the
Glass with a graded refractive index is graded index optical fiber is 56 dB/km at
used as a lens in a copier or as optical 688 nm wavelength and 94 dB/km at
components for an audio-video disk. It is 780 nm wavelength. The tensile strength of
expected in the future that these glasses these fibers was 1600 kg/cm2 (Koike,
will be fully utilized in the field of optical 1992). A simultaneously focusing bifocal
communication or in optical integrated contact lens has been developed using a
circuits (ICs). An anti-reflection film hav- graded index optical fiber produced by a
ing a graded refractive index can be used as similar technique (Koike et al., 1989).
a device to prevent reflection in laser optics These lenses are currently underoing clini-
used for nuclear fusion. An Si-O-N-type cal testing.
thin film lens, which has a certain refrac-
336 20 Functional Gradient Materials

20.5.10 Other Properties

In order that man-made materials can

coexist inside the human body, it is highly
desirable on organically joining the materi-
al, that the boundary region between the
human body and the material have a func-
tional graded structure.
The T i - A l - V alloy is known to be very
compatible with the human body and is
widely used for artificial bones and joints;
nevertheless bonding them to human bone
is still a problem. However, a new type of
implant material composed of a T i - Al-V-
0 0.2 0.4 0.6 0.8 1.0
type substrate has been made. This materi- distance from inner surface
al is coated with T i - A l - V beads to make Figure 20-45. Young's modulus of each portion of
it porous. The material is further coated bamboo (Nogata, 1993).
with the bioactive hydroxy-apatite (HAP)
by plasma spraying. Easier infiltration of
the bone tissue thus occurs with time, creat- crease in the values of tensile strength and
ing a graded structure between the bone Young's modulus from the inner part to
and the implant material. The result is bet- the outer part of the structure, as shown in
ter bonding between the human body and Fig. 20-45 (Nogata, 1993).
the inorganic material (Oonishi et al.,
1994). The bending strength of HAP/Ti is
150 MPa, which is close to that of a human
20.6 Final Comments
bone. This result suggests that the FGM is
As was discussed earlier, the FGMs pro-
effective as an implant material (Watari,
posed in Japan in 1985 were mainly devel-
1994).
oped to reduce thermal stress. However,
A phosphoric acid and amide mixture with experience the merits of FGMs in
can be impregnated into wood by a hot many other fields are being revealed as re-
pressing technique. When the phosphorus search on FGM expands. Functional gra-
content is graded from the surface to the dient materials will hopefully be developed
inner structure by phosphorylation, it be- further in the next generation. It may not
comes an excellent fire retardant material be an overstatement to say that all materi-
(Ishihara et al., 1992). als existing in nature are some sort of an
Most natural materials change their com- FGM. In the development of new materi-
position and structure continuously, so als we need to learn a lot more from what
their properties are not uniform through- exists in nature.
out. Indeed, most natural materials are
functionally graded. The cross section of a
bamboo tree shows a continuously chang- 20.7 Acknowledgements
ing bundle sheath content from the inner
epidermis to the outer epidermis. This Finally, the author gratefully acknowl-
change in the composition reflects the in- edges many helpful comments and infor-
 20.8 References 337

mation received from many FGM re- Erdogan, F , Qzturk, M. (1993), in: Ceramic Transac-
searchers in writing this article. My sincere tions, Vol. 34 - Functionally Gradient Materials
(Proc. 2nd Int. Symp. on Functionally Gradient
thanks go to Dr. Kunihiko Takeuchi of Materials): Holt, J. B., Koizumi, M., Hirai, X, Mu-
Hercules, Inc., U.S.A., for his assistance in nir, Z.A. (Eds.). Westerville, OH: American Ce-
ramic Society, pp. 31-38.
completing this article. Erdogan, F , Wu, B.H. (1993), in: Ceramic Transac-
tions, Vol. 34 - Functionally Gradient Materials
(Proc. 2nd Int. Symp. on Functionally Gradient
Materials): Holt, J.B., Koizumi, M., Hirai, X,
20.8 References Munir, Z.A. (Eds.). Westerville, OH: American
Ceramic Society, pp. 39-46.
Abe, X, Hashimoto, K. (1991), FGM News 15, 21-25 Eroglu, S., Birla, N.C., Demirci, M., Baykara, T.
(in Japanese). (1993), /. Mater. Sci. 12, 1099-1102.
Agullo, J.M., Maury, E, Jouin, J.M. (1993), J. de Ford, R. G. (Ed.) (1992), Mater. Process. Rep. 7,1 -4.
Phys. IV, Colloq. C3, Suppl. J. de Phys. II3, 549- Fritscher, K., Bunk, W. (1990), in: Proc. 1st Int.
556. Symp. on Functionally Gradient Materials (FGM
Aihara, X, Kaji, ML, Igarashi, X (1990), in: Proc. 1st '90): Yamanouchi, M., Koizumi, M., Hirai, X,
Int. Symp. on Functionally Gradient Materials Shiota, I. (Eds.). Tokyo: Functionally Gradient
(FGM '90): Yamanouchi, M., Koizumi, M., Hirai, Materials Forum and The Society of Non-Tradi-
X, Shiota, I. (Eds.). Tokyo: Functionally Gradient tional Technology, pp. 91-96.
Materials Forum and The Society of Non-Tradi- Fritscher, K., Schulz, U. (1993), in: Ceramic Coat-
tional Technology, pp. 39-43. ings, MD - Vol.44 (Proc. 1993 ASME Annu.
Araki, M., Sasaki, M., Kim, S., Suzuki, S., Naka- Mtg.): Kokini, K. (Ed.). New York: American So-
mura, K., Akiba, M. (1994), 1 Nucl Mater. 212- ciety of Mechanical Engineers, pp. 1-8.
215, 1329-1334. Fujii, K., Imai, H., Nomura, S., Shindo, M. (1992), J.
Araki, N., Makino, A., Ishiguro, T. (1993), in: Ce- Nucl. Mater. 187, 204-208.
ramic Transactions, Vol. 34 - Functionally Gradient Fukui, Y, Nakanishi, K. (1991), JSMEInt. J. Ser. Ill
Materials (Proc. 2nd Int. Symp. on Functionally 34, 144-148.
Gradient Materials): Holt, J. B., Koizumi, M., Fukui, Y, Oya-Seimiya, Y, Nakanishi, K. (1991),
Hirai, X, Munir, Z.A. (Eds.). Westerville, OH: Trans. Jpn. Soc. Mech. Eng. 57,1790-1793 (in Jap-
American Ceramic Society, pp. 305-312. anese).
Asahara, Y, Izumitani, T. (1980), /. Non-Crystalline Fukushima, T, Kuroda, S., Kitahara, S. (1990), in:
Solids 42, 269-279. Proc. 1st Int. Symp. on Functionally Gradient Mate-
Atarashiya, K., Ishida, Y, Nagai, X, Uda, M. (1993), rials (FGM '90): Yamanouchi, M., Koizumi, M.,
in: Ceramic Transactions, Vol. 34 - Functionally Hirai, X, Shiota, I. (Eds.). Tokyo: Functionally
Gradient Materials (Proc. 2nd Int. Symp. on Func- Gradient Materials Forum and The Society of
tionally Gradient Materials): Holt, J.B., Koizumi, Non-Traditional Technology, pp. 145-150.
M., Hirai, X, Munir, Z.A. (Eds.). Westerville, OH: Glaeser, A.M., Shalz, M.L., Dalgleish, B.J., Tomsia,
American Ceramic Society, pp. 141-148. A.P. (1993), in: Ceramic Transactions, Vol.34 -
Bennett, C.E.G., McKinnon, N.A., Williams, L.S. Functionally Gradient Materials (Proc. 2nd Int.
(1968), Nature 217, 1287-1288. Symp. on Functionally Gradient Materials): Holt,
Cherradi, N., Dollmeier, K., Ilschner, B. (1993), in: IB., Koizumi, M., Hirai, X, Munir, Z.A. (Eds.).
Ceramic Transactions, Vol. 34 — Functionally Gradi- Westerville, OH: American Ceramic Society,
ent Materials (Proc. 2nd Int. Symp. on Functional- pp. 341-357.
ly Gradient Materials): Holt, I B . , Koizumi, M., Goetzel, C.G., Lavendel, H. W. (1965), in: Metals for
Hirai, X, Munir, Z.A. (Eds.). Westerville, OH: The Space Age (Plansee Proceedings 1964): Be-
American Ceramic Society, pp. 229-236. nesovsky, F. (Ed.). Reutte, Austria: Metallwerk
Chu, J., Ishibashi, H., Hayashi, K., Takebe, H., Mori- Plansee, pp. 149-162.
naga, K. (1993), /. Ceram. Soc. Jpn. 101, 841-844. Goff, J.F., Lowney, J.R. (1976), in: Proc. 1st Int.
Chung, G.Y., McCoy, B.J., Smith, J.M. (1993), in: Conf. on Thermoelectric Energy Conversion: Rao,
Ceramic Transactions, Vol. 34 - Functionally Gradi- K. R. (Ed.). New York: Institute of Electrical and
ent Materials (Proc. 2nd Int. Symp. on Functional- Electronics Eng., pp. 47-49.
ly Gradient Materials): Holt, J.B., Koizumi, M., Harris, R., Kelly, E., Leeds, D.H., Kotlensky, WV.
Hirai, X, Munir, Z.A. (Eds.). Westerville, OH: (1972), in: Proc. 3rd Int. Conf. on Chemical Vapor
American Ceramic Society, pp. 83-90. Deposition: Glaski, F A . (Ed.). Westerville, OH:
Erdogan, F , Chen, Y F. (1993), in: Ceramic Coatings, American Ceramic Society, pp. 183-192.
MD - Vol. 44 (Proc. 1993 ASME Annual Meet- Hirai, T. (1984), in: Materials Science Research Se-
ing): Kokini, K. (Ed.). New York: American Soci- ries, Vol. 17 - Emergent Process Methods for High
ety of Mechanical Engineers, pp. 29-37. Technology Ceramics: Davis, R. F., Palmour, III,
338 20 Functional Gradient Materials

H., Porter, R.L. (Eds.). New York: Plenum, Iwata, M., Yi, W.-D., Nakamura, M., Toyama, S.
pp. 329-345. (1992), /. Soc. Powder Techol, Jpn. 29, 762-768 (in
Hirai, T. (1993), in: Ceramic Transactions, Vol. 34 - Japanese).
Functionally Gradient Materials (Proc. 2nd Int. Jasim, K.M., Rawlings, R.D., West, D.R.F. (1993),
Symp. on Functionally Gradient Materials): Holt, J Mater. Sci. 28, 2820-2826.
J.B., Koizumi, M., Hirai, T., Munir, Z.A. (Eds.): Kaczmarek, R., Robert, W, Jurewicz, J., Boulos,
Westerville, OH: American Ceramic Society, M.I., Dallaire, S. (1984), in: Mater. Res. Soc.
pp. 11-20. Symp. Proc, Vol. 30 - Plasma Processing and Syn-
Hirai, T, Goto, T. (1986), in: Materials Science Re- thesis of Materials: Szekely, I , Apelian, D. (Eds.).
search Series, Vol. 20 - Tailoring Multiphase and New York: North-Holland, pp. 207-215.
Composite Ceramics: Tressler, R. E., Messing, Kawai, T, Miyazaki, S. (1990), J. Ceram. Soc. Jpn.
G.L., Pantano, C.G., Newnham, R.E. (Eds.). Int. Ed. 98, 168-172.
New York: Plenum, pp. 165-177. Kawai, C , Wakamatsu, S., Sakagami, S., Igarashi, T.
Hirai, T, Sasaki, M. (1991a), Ceram. Int. 17, 275- (1990), in: Proc. 1st Int. Symp. on Functionally
281. Gradient Materials (FGM '90): Yamanouchi, M.,
Hirai, T, Sasaki, M. (1991b), JSME Int. J. Ser. 134, Koizumi, M., Hirai, T, Shiota, I. (Eds). Tokyo:
123-129. Functionally Gradient Materials Forum and The
Hirai, T, Sasaki, M., Niino, M. (1987), / Soc. Mater. Society of Non-Traditional Technology, pp 77-82.
Sci. Jpn. 36, 1205-1211 (in Japanese). Kawai, C , Teraki, I, Hirano, T, Nomura, T. (1992),
Hirano, T, Teraki, J., Yamada, T. (1990a), in: Proc. /. Ceram. Soc. Jpn. Int. Ed. 100, 1101-1105.
1st Int. Symp. on Functionally Gradient Materials Kawai, C , Wakamatsu, S., Sakagami, S., Nomura, T.
(FGM '90): Yamanouchi, M., Koizumi, M., Hirai, (1993), in: Ceramic Transactions, Vol. 34 - Func-
T, Shiota, I. (Eds.). Tokyo: Functionally Gradient tionally Gradient Materials (Proc. 2nd Int. Symp.
Materials Forum and The Society of Non-Tradi- on Functionally Gradient Materials): Holt, J.B.,
tional Technology, pp. 5-10. Koizumi, M., Hirai, T, Munir, Z.A. (Eds.). Wester-
Hirano, T., Yamada, T, Teraki, J., Kumakawa, A., ville, OH: American Ceramic Society, pp. 133-139.
Niino, M., Wakashima, K. (1990b), in: Proc. 7th Kawakami, S., Nishizawa, J. (1965), Proc. IEEE 53,
Int. Symp. on Space Technology and Science: 2148-2149.
Kuriki, K., Yamashita, M., Kozawa, H., Arakawa, Kawasaki, A., Watanabe, R. (1990), J Jpn. Soc. Pow-
Y, Azuma, H., Kibe, S. (Eds.). Tokyo: AGNE der Powder Metall. 37, 287-291 (in Japanese).
Kikaku, pp. 501-506. Kawasaki, A., Watanabe, R. (1993), in: Ceramic
Hirano, T, Whitlow, L. W, Miyajima, M. (1993), in: Transactions, Vol. 34 - Functionally Gradient Mate-
Ceramic Transactions, Vol. 34 - Functionally Gradi- rials (Proc. 2nd Int. Symp. on Functionally Gradi-
ent Materials (Proc. 2nd Int. Symp. on Functional- ent Materials): Holt, J.B., Koizumi, M., Hirai, T.,
ly Gradient Materials): Holt, J.B., Koizumi, M., Munir, Z.A. (Eds.). Westerville, OH: American
Hirai, T, Munir, Z.A. (Eds.). Westerville, OH: Ceramic Society, pp. 157-164.
American Ceramic Society, pp. 23-30. Kawasaki, A., Tanaka, M., Watanabe, R. (1993), in:
Hollabaugh, C. M., Reiswig, R. D., Wagner, P., Wah- Ceramic Transactions, Vol. 34 - Functionally Gradi-
man, L. A., White, R. W (1975), /. Nucl. Mater. 57, ent Materials (Proc. 2nd Int. Symp. on Function-
325-332. ally Gradient Materials): Holt, J. B., Koizumi, M.,
Hixlsmann, S., Bunk, W (1993), in: Ceramic Transac- Hirai, T., Munir, Z.A. (Eds.). Westerville, OH:
tions, Vol. 34 - Functionally Gradient Materials American Ceramic Society, pp. 189-196.
(Proc. 2nd Int. Symp. on Functionally Gradient Kimura, H., Kobayashi, S. (1993), / Jpn. Inst. Met.
Materials): Holt, J.B., Koizumi, M., Hirai, T, 57, 1346-1351 (in Japanese).
Munir, Z.A. (Eds.). Westervilles, OH: American Koike, Y. (1991), Polymer 32, 1737-1745.
Ceramic Society, pp. 197-202. Koike, Y. (1992), in: Mater. Res. Soc. Symp. Proc,
Ilschner, B. (1990), in: Proc. 1st Int. Symp. on Func- Vol. 247 - Electrical, Optical, and Magnetic Proper-
tionally Gradient Materials (FGM '90): Yama- ties of Organic Solid State Materials: Chiang, L. Y,
nouchi, M., Koizumi, M., Hirai, T, Shiota, I. Garito, A.F., Sandman, D.J. (Eds.). Pittsburgh,
(Eds.). Tokyo: Functionally Gradient Materials PA: Materials Research Society, pp. 817-828.
Forum and The Society of Non-Traditional Tech- Koike, Y, Tanio, N., Nihei, E., Ohtsuka, Y (1989),
nology, pp. 101-106. Polym. Eng. Sci. 29, 1200-1204.
Inoue, A., Yamagata, H., Masumoto, T. (1993), Koizumi, M. (1992), in: Ceram. Eng. Sci. Proc,
Mater. Trans. JIM 34, 548-555. Vol. 13 - 16th Annu. Conf on Composites and Ad-
Ishihara, S., Getto, H., Sumida, A. (1992), in: All- vanced Ceramic Materials: Mendelson, M. (Ed.).
Division 5 Conference "Forest Products". Nancy: Westerville, OH: American Ceramic Society,
Association pour La Recherche sur Le Bois en pp. 333-347.
Lorraine, p. 423. Koizumi, M. (1993), in: Ceramic Transactions, Vol. 34
Itoh, Y, Kashiwaya, H. (1992), J. Ceram. Soc. Jpn. - Functionally Gradient Materials (Proc. 2nd Int.
100, 476-481. Symp. on Functionally Gradient Materials): Holt,
 20.8 References 339

J.B., Koizumi, M., Hirai, T, Munir, Z.A. (Eds.). Composites and Advanced Ceramic Materials:
Westerville, OH: American Ceramic Society, Mendelson, M. (Ed.). Westerville, OH: American
pp. 3-10. Ceramic Society, pp. 356-364.
Kokini, K., Case, M., Takeuchi, Y. (1993), in: Ceram- Matsumura, S., Okada, M., Yoshikawa, I., Togawa,
ic Transactions, Vol. 34 - Functionally Gradient M., Kuroda, Y (1993), in: Ceramic Transactions,
Materials (Proc. 2nd Int. Symp. on Functionally Vol. 34 - Functionally Gradient Materials (Proc.
Gradient Materials): Holt, J. B., Koizumi, M., 2nd Int. Symp. on Functionally Gradient Materi-
Hirai, T, Munir, Z.A. (Eds.). Westerville, OH: als): Holt, J.B., Koizumi, M., Hirai, T, Munir,
American Ceramic Society, pp. 221-228. Z.A. (Eds.). Westerville, OH: American Ceramic
Konishi, S., Shingyouchi, K., Makishima, A. (1988), Society, pp. 331-338.
/. Non-Cryst. Solids 100, 511-513. Matsuzaki, Y (1994), FGM News 13, 44 (in Japa-
Kowbel, W. (1988), in: Proc. 3rd Int. Symp. on Ceram- nese).
ic Materials and Components for Engines: Tennery, Matsuzaki, Y, Fujioka, J., Okazaki, S. (1993), in:
V. J. (Ed.). Westerville, OH: American Ceramic So- Ceramic Transactions, Vol. 34 - Functionally Gradi-
ciety, pp. 290-308. ent Materials (Proc. 2nd Int. Symp. on Function-
Kowbel, W. (1993), in: Ceramic Transactions, Vol. 34 ally Gradient Materials): Holt, J.B., Koizumi, M.,
- Functionally Gradient Materials (Proc. 2nd Int. Hirai, T, Munir, Z.A. (Eds.). Westerville, OH:
Symp. on Functionally Gradient Materials): Holt, American Ceramic Society, pp. 294-300.
J.B., Koizumi, M., Hirai, T, Munir, Z.A. (Eds.). Mendelson, M.I., McKechnie, T.N. (1993), in: Ce-
Westerville, OH: American Ceramic Society, ramic Transactions, Vol. 34 - Functionally Gradient
pp. 237-244. Materials (Proc. 2nd Int. Symp. on Functionally
Kude, Y. (1993), in: Proc. 4th Symp. on High-Perfor- Gradient Materials): Holt, I B . , Koizumi, M.,
mance Materials for Severe Environments. Tokyo: Hirai, T, Munir, Z.A. (Eds.). Westerville, OH:
R&D Institute of Metals and Composites for Fu- American Ceramic Society, pp. 417-424.
ture Industries and Japan Industrial Technology Merk, N., Ding, X., Guo, X., Ilschner, B.R. (1993),
Association, supplement, pp. 1-12. in: Ceramic Transactions, Vol. 34 - Functionally
Kudesia, R., Niedzialek, S.E., Stangle, G.C., Mc- Gradient Materials (Proc. 2nd Int. Symp. on Func-
Cauley, J.W, Spriggs, R.M., Kaieda, Y (1992), in: tionally Gradient Materials): Holt, IB., Koizumi,
Ceram. Eng. Sci. Proc, Vol. 13 - 16th Annu. Conf M., Hirai, T, Munir, Z.A. (Eds.). Westerville, OH:
on Composites and Advanced Ceramic Materials'. American Ceramic Society, pp. 279-287'.
Mendelson, M. (Ed.). Westerville, OH: American Miller, D.P., Lannutti, II, Yancey, R.N. (1992), in:
Ceramic Society, pp. 374-383. Ceram. Eng. Sci. Proc, Vol. 13 - 16th Annu. Conf
Kumakawa, A., Sasaki, M., Takahashi, M., Niino, on Composites and Advanced Ceramic Materials:
M., Adachi, N., Arikawa, H. (1990a), in: Proc. 1st Mendelson, M. (Ed.). Westerville, OH: American
Int. Symp. on Functionally Gradient Materials Ceramic Society, pp. 365-373.
(FGM '90): Yamanouchi, M., Koizumi, M., Hirai, Miyamoto, Y, Nakanishi, H., Tanaka, I., Okamoto,
T, Shiota, I. (Eds.). Tokyo: Functionally Gradient T. (1990), in: Proc. 1st Int. Symp. on Functionally
Materials Forum and The Society of Non-Tradi- Gradient Materials (FGM '90): Yamanouchi, M.,
tional Technoloy, pp. 291-294. Koizumi, M., Hirai, T, Shiota, I. (Eds.). Tokyo:
Kumakawa, A., Maeda, S., Sasaki, M., Niino, A., Functionally Gradient Materials Forum and The
Sakamoto, A., Sasaki, M., Hirai, T. (1990 b), in: Society of Non-Traditional Technology, pp. 257-
Space Applications of Advanced Structural Materi- 262.
als (Proc. Int. Symposium): Burke, W.R. (Com- Miyamoto, Y, Tanihata, K., Matsuzaki, Y, Ma, X.
piler). Noordwijk, The Netherlands: ESA Publica- (1992), Int. J. Self-Propag. High-Temp. Synth. 1,
tions Division, pp. 339-344. 147-154.
Kumakawa, A., Niino, M., Kiyoto, S., Nagata, S. Moreno, R., Sanchez-Herencia, A.I, Moya, I S .
(1993), in: Ceramic Transactions, Vol. 34 - Func- (1993), in: Ceramic Transactions, Vol. 34 - Func-
tionally Gradient Materials (Proc. 2nd Int. Symp. tionally Gradient Materials (Proc. 2nd Int. Symp.
on Functionally Gradient Materials): Holt, IB., on Functionally Gradient Materials): Holt, IB.,
Koizumi, M., Hirai, T, Munir, Z.A. (Eds.). Wester- Koizumi, M., Hirai, T, Munir, Z.A. (Eds.). Wester-
ville, OH: American Ceramic Society, pp. 213-220. ville, OH: American Ceramic Society, pp. 149-
Kuroda, Y, Sato, M., Ueda, S., Niino, M., Suemitsu, 156.
T, Kude, Y (1992), in: Proc. 18th Int. Symp. on Mori, H., Nakamura, M., Toyama, S. (1993), in: Ce-
Space Technology and Science: Hirosawa, H., ramic Transactions, Vol. 34 - Functionally Gradient
Arakawa, Y, Onoda, I, Otsubo, K., Kawaguchi, Materials (Proc. 2nd Int. Symp. on Functionally
J., Toki, K., Horikawa, Y, Murata, M. (Eds.). Gradient Materials): Holt, I B . , Koizumi, M.,
Tokyo: AGNE Kikaku, pp. 87-94. Hirai, T, Munir, Z.A. (Eds.). Westerville, OH:
Ma, X., Tanihata, K., Miyamoto, Y, Kumakawa, A., American Ceramic Society, pp. 393-400.
Nagata, S., Yamada, T, Hirano, T. (1992), in: Cer- Moya, I S . , Sanchez-Herencia, A.I, Requena, I,
am. Eng. Sci. Proc, Vol. 13 - 16th Annu. Conf. on Moreno, R. (1992), Mater. Lett. 14, 333-335.
340 20 Functional Gradient Materials

Moya, I S . , Sanchez-Herencia, A.I, Moreno, R., Sasaki, M., Hirai, T. (1991), J Ceram. Soc. Jpn. 99,
Pena, P., Requena, I (1993), in: Third EURO- 1002-1013.
CERAMICS, Vol.3: Duran, P., Fernandez, I F . Sasaki, M., Wang, Y, Hirano, T, Hirai, T. (1989), J.
(Eds.). Madrid, Spain: Faenza Editrice Iberica, Ceram. Soc. Jpn. Int. Ed. 97, 530-534.
pp. 289-300. Sasaki, M., Hiratani, T, Hirai, T. (1993), in: Ceramic
NASTA (National Aerospace Science & Technology Transactions, Vol. 34 - Functionally Gradient Mate-
Agency), Hirai, T., Watanabe, R., Daikin Indus- rials (Proc. 2nd Int. Symp. on Functionally Gradi-
tries Ltd., Sumitomo Electric Industries Ltd. (1986), ent Materials): Holt, I B . , Koizumi, M., Hirai, T,
European Patent Application No. 86 310064.0. Munir, Z.A. (Eds.). Westerville, OH: American
Newnham, R. E. (1986), in: Materials Science Re- Ceramic Society, pp. 369-376.
search, Vol. 20 - Tailoring Multiphase and Com- Sata, N. (1992), in: Ceramic Engineering & Science
posite Ceramics: Tressler, R. E., Messing, G. L., Proceedings, Vol. 13 - 16th Ann. Conf. on Com-
Pantano, C.G., Newnham, R.E. (Eds.). New posites and Advanced Ceramic Materials'. Mendel-
York: Plenum, pp. 385-394. son, M. (Ed.). Westerville, OH: American Ceramic
Niino, M., Chen, L.D. (1993), in: Proc. 12th Int. Society, pp. 384-391.
Conf. on Thermoelectrics: Matsuura, K. (Ed.). Sata, N. (1993), in: Ceramic Transactions, Vol. 34 -
Yokohama: Inst. Electrical Engineers, Jpn., Functionally Gradient Materials (Proc. 2nd Int.
pp. 521-531. Symp. on Functionally Gradient Materials): Holt,
Niino, M., Hirai, T., Watanabe, R. (1987), /. Jpn. Soc. IB., Koizumi, M., Hirai, T, Munir, Z.A. (Eds.).
Compos. Mater. 13, 257-264 (in Japanese). Westerville, OH: American Ceramic Society,
Niiyama, K. (1994), private communication. pp. 109-116.
Nishida, I. (1993), FGM News 21, 17-20 (in Japa- Sata, N., Nagata, K., Yanagisawa, N., Asano, O.,
nese). Sanada, N. (1990), in: Proc. 1st Japan-U.S. Work
Nogata, F. (1993), New Ceram. 1, 55-58 (in Japa- Shop on Combustion Synthesis: Kaieda, Y, Holt,
nese). I B . (Co-Chairmen). Tokyo: National Research In-
Okamura, H. (1991), Mater. Sci. Eng. A 143, 3-9. stitute for Metals, pp. 139-146.
Omori, M., Sakai, H., Ohkubo, A., Kawahara, M., Sheppard, L.M. (Ed.). (1992), Ceram. Bull. 71, 624-
Hirai, T. (1994), /. Jpn. Soc. Powder Powder Metall. 626.
41, 649-652 (in Japanese). Shimoda, N., Kitaguchi, S., Saito, T, Takigawa, H.,
Oonishi, H., Noda, T., Ito, S., Kohda, A., Ishimaru, Koga, M. (1990), in: Proc. 1st Int. Symp. on Func-
H., Yamamoto, M., Tsuji, E. (1994), /. Appl. Bio- tionally Gradient Materials (FGM '90): Yamanou-
mater. 5, 23-37. chi, M., Koizumi, M., Hirai, T, Shiota, I. (Eds.).
Osaka, T., Matsubara, H., Homma, T., Mitamura, S., Tokyo: Functionally Gradient Materials Forum
Noda, K. (1990), Jpn. J. Appl. Phys. 29,1939-1943. and The Society of Non-Traditional Technology,
Rabin, B.H., Heaps, R . I (1993), in: Ceramic Trans- pp. 151-156.
actions, Vol. 34 - Functionally Gradient Materials Shinohara, Y, Imai, Y, Ikeno, S., Shiota, I. (1990),
(Proc. 2nd Int. Symp. on Functionally Gradient in: Proc. 1st Int. Symp. on Functionally Gradient
Materials): Holt, I B . , Koizumi, M., Hirai, T, Materials (FGM '90): Yamanouchi, M., Koizumi,
Munir, Z.A. (Eds.). Westerville, OH: American M., Hirai, T, Shiota, I. (Eds.). Tokyo: Functionally
Ceramic Society, pp. 173-180. Gradient Materials Forum and The Society of
Radhakrishnan, R., Bhaduri, S.B., Wojcicki, S. Non-Traditional Technology, pp. 225-228.
(1993), in: Ceramic Transactions, Vol. 34 - Func- Shinohara, Y, Imai, Y, Ikeno, S., Shiota, I., Fukushi-
tionally Gradient Materials (Proc. 2nd Int. Symp. ma, T. (1993), in: Ceramic Transactions, Vol. 34 -
on Functionally Gradient Materials): Holt, I B . , Functionally Gradient Materials (Proc. 2nd Int.
Koizumi, M., Hirai, T., Munir, Z.A. (Eds.). Wester- Symp. on Functionally Gradient Materials): Holt,
ville, OH: American Ceramic Society, pp. 117-123. IB., Koizumi, M., Hirai, T., Munir, Z.A. (Eds.).
Requena, I , Moya, I S . , Pena, P. (1993), in: Ceramic Westerville, OH: American Ceramic Society,
Transactions, Vol. 34 - Functionally Gradient Ma- pp. 255-262.
terials (Proc. 2nd Int. Symp. on Functionally Gra- Sohda, Y, Kude, Y, Uemura, S., Saitoh, T, Waka-
dient Materials): Holt, I B . , Koizumi, M., Hirai, T., matsu, Y, Niino, M. (1993), in: Ceramic Transac-
Munir, Z.A. (Eds.). Westerville, OH: American tions, Vol. 34 - Functionally Gradient Materials
Ceramic Society, pp. 203-210. (Proc. 2nd Int. Symp. on Functionally Gradient
Roy, R. (1986), Mater. Sci. Res. 21, 25-32. Materials): Holt, I B . , Koizumi, M., Hirai, T,
Sasaki, M., Hirai, T. (1990), in: Proc. 1st Int. Symp. Munir, Z.A. (Eds.). Westerville, OH: American
on Functionally Gradient Materials (FGM '90): Ya- Ceramic Society, pp. 125-132.
manouchi, M., Koizumi, M., Hirai, T., Shiota, I. Steffens, H.-D., Dvorak, M., Wewel, M. (1990), in:
(Eds.). Tokyo: Functionally Gradient Materials Proc. 1st Int. Symp. on Functionally Gradient Mate-
Forum and The Society of Non-Traditional Tech- rials (FGM '90): Yamanouchi, M., Koizumi, M.,
nology, pp. 83-88. Hirai, T, Shiota, I. (Eds.). Tokyo: Functionally
 20.8 References 341

Gradient Materials Forum and The Society of T, Shiota, I. (Eds.). Tokyo: Functionally Gradient
Non-Traditional Technology, pp. 139-143. Materials Forum and The Society of Non-Tradi-
Suemitsu, T., Matsuzaki, Y, Fujioka, X, Uchida, M., tional Technology, pp. 19-26.
Sohda, Y, Kude, Y, Uemura, S., Kuroda, Y, Watanabe, R., Kawasaki, A. (1992), in: Composite
Ueda, S., Moro, A. (1993), in: Ceramic Transac- Materials (Proc. Symp. A4 on Composite Materi-
tions, Vol. 34 - Functionally Gradient Materials als of The International Conference on Advanced
(Proc. 2nd Int. Symp. on Functionally Gradient Materials - CAM 91): Di Benedetto, A.T., Nico-
Materials): Holt, IB., Koizumi, M., Hirai, T., lais, L., Watanabe, R. (Eds.). New York: Elsevier,
Munir, Z.A. (Eds.). Westerville, OH: American pp. 197-208.
Ceramic Society, pp. 315-322. Watanabe, R., Kawasaki, A., Takahashi, H. (1991),
Suganuma, K., Okamoto, T, Koizumi, M. (1984), J. in: Mechanics and Mechanisms of Damage in Com-
Am. Ceram. Soc. 67, C/256-257. posites and Multi-Materials (ESISII): Baptiste, D.
Takahashi, M. (1994), private communication. (Ed.). London: Mechanical Engineering Publish-
Takahashi, M., Itoh, Y, Miyazaki, M., Takano, H., ers, pp. 285-289.
Okuhata, T. (1993), in: Plansee Proceedings, Vol. 4 Watanabe, Y, Nakamura, Y, Fukui, Y, Nakanishi,
(Proc. 13th Int. Plansee Seminar '93): Bildstein, H., K. (1993), /. Mater. Sci. Lett. 12, 326-328.
Eck, R. (Eds.). Reuttel, Austria: Metallwerk Plan- Watari, F. (1994), Bull. Ceram. Soc. Jpn. 29,191-193
see, pp. 17-28. (in Japanese).
Takebe, H., Teshima, T, Nakashima, M., Morinaga, Williamson, R.L., Rabin, B.H. (1993), in: Ceramic
K. (1992), /. Ceram. Soc. Jpn. 100, 387-391. Transactions, Vol. 34 - Functionally Gradient Mate-
Takemura, M., Yoshitake, A., Hayakawa, H., rials (Proc. 2nd Int. Symp. on Functionally Gradi-
Hyakube, T, Tamura, M. (1990), in: Proc. 1st Int. ent Materials): Holt, IB., Koizumi, M., Hirai, T,
Symp. on Functionally Gradient Materials (FGM Munir, Z.A. (Eds.). Westerville, OH: American
'90)\ Yamanouchi, M., Koizumi, M., Hirai, T, Shi- Ceramic Society, pp. 55-65.
ota, I. (Eds.). Tokyo: Functionally Gradient Mate- Yamamoto, O., Imai, K., Sasamoto, T, Inagaki, M.
rials Forum and The Society of Non-Traditional (1993), /. Fur. Ceram. Soc. 12, 435-440.
Technology, pp. 97-100. Yamaoka, H., Yuki, M., Tahara, K., Irisawa, T.,
Tang, X.F., Zhang, L.M., Zhang, Q.J., Yuan, R.Z. Watanabe, R., Kawasaki, A. (1993), in: Ceramic
(1993), in: Ceramic Transactions, Vol. 34 - Func- Transactions, Vol. 34 - Functionally Gradient Mate-
tionally Gradient Materials (Proc. 2nd Int. Symp. rials (Proc. 2nd Int. Symp. on Functionally Gradi-
on Functionally Gradient Materials): Holt, IB., ent Materials): Holt, IB., Koizumi, M., Hirai, T,
Koizumi, M., Hirai, T, Munir, Z.A. (Eds.). Wester- Munir, Z.A. (Eds.). Westerville, OH: American
ville, OH: American Ceramic Society, pp. 457-463. Ceramic Society, pp. 165-172.
Tani, I , Liu, G.-R. (1993), JSME Int. J. Ser. A 36, Yuan, R.Z., Zhang, L.M., Zhang, Q.I, Tang, X.F.
152-155. (1993), /. Mater. Synth. Process. 1, 171-179.
Tanihata, K., Miyamoto, Y, Matsushita, K., Ma, Y,
Kawasaki, A., Watanabe, R., Hirano, K. (1993),
in: Ceramic Transactions, Vol. 34 — Functionally
Gradient Materials (Proc. 2nd Int. Symp. on Func-
tionally Gradient Materials): Holt, IB., Koizumi, General Reading
M., Hirai, T., Munir, Z.A. (Eds.). Westerville, OH:
American Ceramic Society, pp. 361-368. Holt, I B . , Koizumi, M., Hirai, T., Munir, Z. A. (Eds.)
Uemura, S., Sohda, Y, Kude, Y (1990), in: Proc. 1st (1993), Ceramic Transactions, Vol. 34 - Functional-
Int. Symp. on Functionally Gradient Materials ly Gradient Materials. Westerville, OH: American
(FGM '90): Yamanouchi, M., Koizumi, M., Hirai, Ceramic Society.
T, Shiota, I. (Eds.). Tokyo: Functionally Gradient Ilschner, B., Cherradi, N. (Eds.) (1995), Proc. 3rd Int.
Materials Forum and The Society of Non-Tradi- Symp. on Structural and Functional Gradient Mate-
tional Technology, pp. 237-242. rials. Lausanne, Switzerland: Presses polytech-
Wakamatsu, Y, Ueda, S., Saito, T, Niino, M., niques et universitaires romandes.
Nomura, T, Kosaka, K., Saito, T, Kiyoto, S. Markworth, A.I, Ramesh, K.S., Parks, Jr., W.P.
(1993), in: Ceramic Transactions, Vol. 34 - Func- (1995), /. Mater. Sci. 30, 2183-2193.
tionally Gradient Materials (Proc. 2nd Int. Symp. Rabin, B.H., Shiota, I. (Eds.) (1995), "Functionally
on Functionally Gradient Materials): Holt, IB., Gradient Materials", MRS Bull. 20, No. 1.
Koizumi, M., Hirai, T, Munir, Z.A. (Eds.). Wester- Yamanouchi, M., Koizumi, M., Hirai, T, Shiota, I.
ville, OH: American Ceramic Society, pp. 263-270. (Eds.) (1990), Proc. 1st Int. Symp. on Functionally
Wakashima, K., Tsukamoto, H. (1990), in: Proc. 1st Gradient Materials (FGM '90). Tokyo: Function-
Int. Symp. on Functionally Gradient Materials ally Gradient Materials Forum and The Society of
(FGM '90): Yamanouchi, M., Koizumi, M., Hirai, Non-Tradional Technology.
21 Diamond: Its Synthesis from the Vapor Phase and Applications

Naoji Fujimori

Sumitomo Electric Industries Ltd., Itami Research Laboratories, Hyogo, Japan

List of Symbols and Abbreviations 344

21.1 Introduction 345
21.2 Characteristics of Diamond 346
21.3 Diamond Synthesis from the Vapor Phase 347
21.3.1 Hot-Filament Chemical Vapor Deposition (CVD) 348
21.3.2 Plasma-Assisted Chemical Vapor Deposition (CVD) 348
21.3.3 Plasma Jet 349
21.3.4 Combustion Flame 349
21.3.5 Conditions 349
21.3.6 Mechanism 350
21.4 Characteristics of CVD Diamond 350
21.5 Realized and Anticipated Applications of CVD Diamond 353
21.5.1 Tools 353
21.5.2 Electrical Parts 355
21.5.3 Active and Passive Devices 357
21.5.4 Other Applications 359
21.6 Future Development and Problems to be Solved 361
21.7 References 362

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
344 21 Diamond: Its Synthesis from the Vapor Phase and Applications

List of Symbols and Abbreviations

HV Vickers hardness
/ current
K dielectric constant
T temperature
V voltage
K gate voltage
^SAT electron satured drift velocity
GT thermal conductivity
cBN cubic boron nitride
CVD chemical vapor deposition
DC direct current
DLC diamond-like carbon
EMA electron microscopic analysis
FET field effect transistor
IC integrated circuit
IDT interdigital transducer
IMPATT impact avalanche transition time (diode)
IR infrared
MESFET metal-semiconductor field effect transistor
PVD physical vapor deposition
RF radio frequency
SAW surface acoustic wave
TAB tape automated bonding
YAG yttrium aluminum garnet
 21.1 Introduction 345

21.1 Introduction since. Most people may simply think of

diamond as the most expensive jewel, but
The technologies and applications of di- diamonds have also contributed to human
amond films formed from the gas phase life as a key industrial material from
are described in this chapter. Many kinds around 1960 onwards. Nowadays, more
of thin films, such as metals, ceramics, and than 95 % of the diamonds in use are for
semiconductors, are used in various fields, industrial applications, and they are
and thin film technologies are considered mostly artificial diamonds.
to be one of the present key technologies Making artificial diamonds was man's
for fabricating advanced materials. The dream, like alchemy, and was realized in
objectives for placing such films on bulk 1955 by the General Electric Company
materials are to improve the properties, (U.S.A.). The technology was based on us-
make composite materials, add another ing ultrahigh pressure to imitate the condi-
function, etc. The technologies used for tions deep underground: a pressure of
making these films are electroplating, 50000 atm (5.08 GNm" 2 ) and a tempera-
chemical vapor deposition, physical vapor ture of 1500 °C. However, the direct trans-
deposition, ion beam technologies, etc., formation from graphite to diamond is dif-
and most of them were developed during ficult to perform even under these condi-
the 1950s and 1960s. Electroplating is tions, so a metal catalyst, such as nickel,
quite an old technology and can only be cobalt, or iron, has to be used.
used to form metal films. On the other Two practical products were developed
hand, vapor deposition technologies can using this high-pressure technology. One
make various kinds of film; even artificial was abrasives, as in grinding wheels or cut-
materials can be formed using semicon- ters, or diamond abrasives bonded by
ductor superlattice technology. The initial metal or polymer for working hard materi-
applications of vapor-deposited thin films als (stone, ceramics, etc.); the other prod-
were in cutting tools, in about 1970. This uct was sintered diamonds, which are used
resulted in a major revolution as regards for cutting tools or dies for drawing metal
the tool life, giving a lifetime improvement wires. The other contribution of this tech-
of several times, or even more than ten nology was to develop cubic boron nitride
times. Not only 'functional' materials, (cBN). cBN is the second hardest material
such as magnetic materials, optical materi- and is used for machining steel or iron,
als, and electronic materials, but also which diamonds are unable to machine,
structural materials such as wear-resistant using grinding wheels or sintered cutting
materials, anticorrosion materials, and tools.
decorating materials are improved by the The development of artificial single-
use of films. Among these materials and crystal diamonds, as shown in Fig. 21.1,
technologies, diamond films may be seen resulted in new applications for diamonds,
as a technological goal or indeed they may which sintered or powder diamond had
be regarded as the alchemy of the twenti- not been able to realize. In 1984, Sumi-
eth century. tomo Electric Industries (Japan) commer-
Diamond has been called "a girl's best cialized diamond heat sinks made of
friend". The biggest diamond mine, which high-pressure synthesized single-crystal
is in South Africa, was discovered in 1866, diamonds, and these are used in electrical
and diamond has been a popular jewel ever applications such as laser diodes or micro-
346 21 Diamond: Its Synthesis from the Vapor Phase and Applications

Significant progress in the industrializa-

tion of diamond was achieved by the devel-
opment of technologies for synthesizing
diamonds from the gas phase, and these
technologies are considered to extend the
limits of diamond's application. More-
over, the size of high-pressure synthesized
diamonds is limited to less than 10 mm,
and larger diamonds would indeed be use-
ful.
In this chapter, the technologies of dia-
Figure 21-1. High-pressure-synthesized single-crystal mond synthesis from the gas phase, and
diamonds. the applications of diamond film (realized
and expected) are reviewed.
wave devices. Diamond has a higher ther-
mal conductivity than any other material, 21.2 Characteristics of Diamond
and this characteristic is best utilized by
using single-phase diamond without any Diamond is a carbon phase with the dia-
other phase. This commercialization mond crystal structure shown in Fig. 21-2.
opened the door for utilizing diamonds in The atom bonding consists of sp 3 bonds,
electronics. which are significantly different from
However, the applications of diamond those in graphite or carbon.
are limited by the shape or the cost of the The most important characteristic of di-
diamond. Its high hardness makes dia- amond is its hardness. However, diamond
mond the most difficult material to ma- has many excellent characteristics (Davice,
chine. Moreover, natural or synthesized 1979) and as such it can be expected to be
diamonds are small particles and sintered used in many application fields. Several
diamond without a binder phase has not characteristics of diamond are given in
been developed, so the technology to make Table 21-1. Diamond is the only material
pure and free shaped diamonds was de- to have more than two properties with the
sired. highest values among all the materials.

Figure 21-2. Atomic struc-

ture of diamond.
 21.3 Diamond Synthesis from the Vapor Phase 347

Table 21-1. Characteristics of diamond. oxidizing, and it is not etched by any liq-
Property
uid.
Value

Density 3.52 g/cm3

Young's modulus 1.05xl0 12 N/m 2 21.3 Diamond Synthesis from the
Hardness, HV 10000 kgf/mm2
Thermal expansion coefficient 1.6xlO~ 6 /K
Vapor Phase
Thermal conductivity 2200 W/mk
Optical refractive index 2.42 In order to make artificial diamonds,
Band gap 5.47 eV many trials were carried out from the
1800s onwards, using hydrocarbon as the
source material. The first patent for dia-
mond deposition from the gas phase was
Diamond is categorized according to filed in 1958 (Eversol, 1958). This patent
four types, as shown in Table 21-2. Type I involved using a mixed gas consisting of
diamond contains nitrogen, type II does hydrogen and methane, heated above
not. The purest diamond is type II a, which 1000 °C. This technique was accomplished
makes up less than 1 % of natural dia- by Angus et al. (1968). They reported
mond. Both type la and type Ib diamond weight gain of diamond powder using a
dissolve nitrogen. The nitrogen in type la simple thermal CVD (chemical vapor de-
diamond segregates as two or three atoms position) method. However, the growth
with a vacancy, while each nitrogen atom rate was so low that it could not be used
in type Ib dissolves separately. Type lib for industrial applications.
contains boron and is a p-type semicon- In the Soviet Union in the 1960s, several
ductor. Each type of diamond has about research groups investigated this technol-
the same hardness but the spectra in trans- ogy. One result, reported in 1968, was that
mitted light are different from one an- cyclic heating by a xenon lamp enabled
other, and this determines the color of the diamond growth on a diamond substrate,
diamond. but this method also achieved only quite a
Diamond is stable at an elevated tem- small deposition rate (Derjaguin et al.,
perature provided the atmosphere is not 1968).

Table 21-2. Diamond types and their properties.

Diamond type I II

la Ib II a lib

Impurity N (ppm) >1000 50-1000 <1 <1

other no Ni, Fe, etc., in no B ~ 150 ppm
synthesized diamonds Al -100 ppm
Absorption edge (nm) 300 300 225 225
Color colorless yellow colorless blue
Resistivity (Qcm) 10 14 -10 16 10°-10 4
Thermal conductivity (W/(cm K) >9 9-22 22 -22
348 21 Diamond: Its Synthesis from the Vapor Phase and Applications

Omarfz reactor The major innovation of this technology

was performed by Matsumoto and co-
Substrate workers in 1981, and the method was so
Holder
Tungsten filament' simple that many other researchers con-
Thermocouple firmed diamond deposition by this means
(Matsumoto et al., 1982). Their method is
so-called "hot-filament CVD", as shown
in Fig. 21.3, using hydrogen-diluted meth-
ane as the source gas, and the deposition
rate was reported to be 1 jiim or more per
VAC. —— - Reactant gas hour. They confirmed diamond formation
by Raman spectroscopy, which is quite a
reliable method for distinguishing dia-
Figure 21-3. Schematic diagram of hot-filament
mond from graphite or amorphous car-
CVD.
bon.
After Matsumoto's report on the devel-
In 1971, Aisenberg and Chabot reported opment of hot-filament CVD, many Japa-
on a very hard film deposited by carbon nese researchers rushed into this technol-
ion beam deposition (Aisenberg and ogy and succeeded in developing various
Chabot, 1971). This film was referred to as methods. Thanks to progress by the elec-
diamond-like carbon (DLC), and it was tronic industries, the technological back-
found to be possible to make such films by ground of thin film deposition encouraged
many PVD (physical vapor deposition) or the development of many technologies.
CYD methods. DLC is essentially an The technologies for making diamond
amorphous film and consists of sp 3 and from the gas phase are categorized accord-
sp2 bonds. Applications of such films have ing to the following four basic methods.
been pursued and research on DLC has
been continuing actively ever since.
21.3.1 Hot-Filament Chemical Vapor
In 1976, Derjaguin and coworkers re-
Deposition (CVD)
ported on the formation of diamond on
silicon or molybdenum substrates; the first In hot-filament CVD (Fig. 21-3) the
report of diamond formation on a sub- substrate is placed beneath a filament of
strate other than diamond (Derjaguin tungsten or tantalum, which is heated to
et al., 1976). The method was described as above 2000 °C as an activation means. A
a "chemical transport reaction", but no higher filament temperature results in a
schematic drawing of the equipment was in higher deposition rate.
their report, which made it impossible to
reproduce their results. The conventional
21.3.2 Plasma-Assisted Chemical Vapor
chemical transport reaction method in-
Deposition (CVD)
volves a vaporization region of the source
material and a redeposition region of the Many plasma sources have been per-
'object' material. The report referred to us- fected for diamond deposition, such as
ing graphite as the source material, hydro- radio frequency, microwave, and direct
gen as the carrier gas, and 'additional en- current (DC) (Sawabe and Inuzuka, 1985).
hancement means' in the reaction region. Microwave plasma CVD (Kamo et al.,
 21.3 Diamond Synthesis from the Vapor Phase 349

Reactant gas and Mituizumi, 1988). The most popular

Waveguide
t tluartz reactor

Substrate
combustion is carried out using acetylene
and oxygen in a volume ratio of about 2:3,
and small, circular diamond films can be
obtained at normal atmospheric pressure.
Since the outer part of the combustion
flame is oxidizing, the inner part of the
i flame, called the 'acetylene feather' and
Magnetron colored blue, is the part at which to make
VAC. the diamonds. This technique appears to
Figure 21-4. Schematic diagram of microwave plasma be an easy way to make diamond, and the
CVD. author considered forming diamonds on
the base of a frying pan when he knew this.

1983) is the most popular method and is

21.3.5 Conditions
shown in Fig. 21-4. The frequency of the
microwaves is typically 2.54 GHz, and a More than 30 modified methods have
silica tube with a diameter of 50 mm is been developed but they are all variants of
used as the reaction tube, because silica is CVD methods. So diamonds produced
transparent to microwaves. The reactant from the gas phase are called 'CVD dia-
gas and the reaction pressure are the same monds'. The principles of diamond deposi-
as in hot-filament CVD. tion can be summarized by the following
conditions:
21.3.3 Plasma Jet 1. At least carbon and hydrogen are
needed in the reactant gas.
Using a plasma-jet torch, such as is used
2. One or more activation means are
for welding or cutting steel, diamonds are
needed.
formed from the source gas consisting of
3. A substrate temperature from 300 to
hydrogen, methane, and argon at atmo-
1000°Cis required.
spheric pressure (Matsumoto et al., 1987;
Kurihara et al., 1988). Plasma is ignited by Many kinds of carbonaceous gas, such
direct current or radio frequency and the as hydrocarbons, alcohols, ketones, carbon
plasma temperature rises above 10000 K. oxides, carbon halides, etc., have been
As the substrate would be welded by the found to be suitable source gases. In con-
plasma, it has to be cooled from behind so trast, no one has succeeded in making dia-
that the surface temperature is kept at a mond from a reactant gas that does not
few hundred degrees Celsius. contain any hydrogen atoms. Recent re-
ports suggest the possibility of making dia-
monds at temperatures as low as 150°C.
21.3.4 Combustion Flame
Typical deposition conditions using hot-
A combustion torch, such as is used for filament CVD are as follows: Reactant
welding or cutting steel, is used with a hy- gas: 1 % CH 4 , 99% H 2 , pressure: 40 Torr
drocarbon, such as acetylene, ethane, or (5.3kNm~ 2 ); filament temperature:
methane, and oxygen, and diamonds are 2050 °C; substrate temperature: 900 °C;
deposited on the chilled substrate (Hirose and deposition rate: 1 (im/h.
350 21 Diamond: Its Synthesis from the Vapor Phase and Applications

Many reports suggest that, under these

conditions, atomic hydrogen is formed as a
result of such a high filament temperature,
and that it helps in the formation of a pre-
cursor of diamond and by etching the
graphite codeposited with the diamonds.
Differences to methods for depositing
other materials can be found in the plasma
power density and the electrode arrange-
ment. The plasma power density for dia-
mond deposition is more than 2 times that
used for making other materials. Ceramic
films, amorphous Si, and metals are de-
posited on a negatively biased electrode
(cathode), but diamond is deposited on a
positively biased electrode (anode). Using
plasma-assisted CVD for making dia- [010]
mond, the electron temperature of the hy- 2nm
drogen was found to be as high as 5000 K [100]
(Sawabe, 1988).
Figure 21-5. Scanning tunneling micrograph of an
epitaxially grown (001) surface.
21.3.6 Mechanism
The mechanism of diamond formation
from the gas phase has been studied by 21.4 Characteristics of
observing the growth of the surface film CVD Diamond
and measuring the gas phase composition.
Among those approaches, the mechanism The typical surface morphology of dia-
is readily studied during the growth of epi- mond film deposited on silicon by mi-
taxial films, and observation of the surface crowave plasma CVD using a reactant gas
structure by scanning tunneling micros- consisting of 1 % CH 4 in H 2 is shown in
copy has been carried out successfully Fig. 21-6. As the intrinsic nucleation sites
(Tsuno et al., 1991). Figure 21-5 shows the for diamond deposition are few, rubbing
atomic image of an (001) epitaxial surface the surface with diamond grit is necessary
in which a structure of many rods can be in order to make diamond film without
seen. These rods consist of two carbon using the special nucleation method de-
atoms, called a dimer, which is a recon- scribed later. Without such a pretreatment,
structed structure similar to that of the sil- separate particles are formed, as shown in
icon surface. This structure may be formed Fig. 21-7. The surface structure changes
with the cooperation of hydrogen atoms, with increasing CH 4 concentration, and
which would explain why the source gas typical surface morphologies deposited
for the CVD of diamond has to contain from a reactant gas consisting of 2 - 8 %
hydrogen atoms. The author believes that CH 4 in H 2 are shown in Fig. 21-8. Fig-
a similar growth mechanism will be found ure 21-9 shows the cross section of 30 |im
on the (111) surface. thick diamond film. A columnar structure,
 21.4 Characteristics of CVD Diamond 351

(a)

(b)

Figure 21-6. Surface structure of diamond film de-

posited from a reactant gas consisting of 1 % CH 4 in

(0

;*?#?.::•:•

Figure 21-7. SEM micrograph of the surface of dia-

mond particles. (d)

which is quite a common structure of crys-

talline films, is seen in this micrograph.
The lattice parameter of the deposited
film is determined as characteristic of the
diamond by reflective electron diffraction
and X-ray diffraction (Derjaguin et al.,
1976). CVD diamond is essentially IIa dia-
mond which contains no impurity atoms. Figure 21-8. Surface morphology of diamond films
Collins etal. (1989) reported on the ab- deposited using a reactant gas consisting of 2 - 8 %
sorption of diamond film. The absorption CH 4 in H 2 : (a) 2 %, (b) 4%, (c) 6%, and (d) 8% CH 4 .
352 21 Diamond: Its Synthesis from the Vapor Phase and Applications

(a)

(b)

1 0 um
Figure 21-9. Fracture surface of a diamond film. (c)

edge was found to be 225 nm, which is the

same as for II a diamond. Recently, Imai
and Fujimori (1991) likewise reported that
the absorption edge of CVD diamond film
with a thickness of 100 jam was 225 nm.
s (d)

Small amounts of phases other than dia-

mond have been observed by Raman spec-
troscopy (Sato et al., 1980). Figure 21-10a
shows a typical spectrum of the Raman
shift obtained from CVD diamond film de-
posited by microwave plasma CVD using (e)

a reactant gas containing 1 % CH 4 . The

Raman shift of diamond is observed at
1332 cm" 1 , and the graphite and amor-
phous phase can easily be identified be-
cause the Raman shifts of such phases are 1700 1600 1500 1300 1200 1100
observed at 1360 and 1580 cm" 1 (graph- RAMAN SHIFT (cm ) 1

ite) and 1500 cm" 1 (amorphous phase). Figure 21-10. Raman spectra of diamond films de-
The effect of the CH 4 concentration in posited using reactant gases containing (a) 1%,
the reactant gas is clearly seen in the (b) 2%, (c) 4%, (d) 6% and (e) 8% CH 4 .
Raman spectra. A typical change in the
Raman spectra on going from 1 % to 8 %
CH 4 in H 2 is shown in Fig. 21-10. Al- trations make the diamond films black and
though no peak is seen in Fig. 21-10e at low CH 4 concentrations result in transpar-
1332 cm" 1 , the presence of diamond was ent diamond films. In Fig. 21-11, a trans-
confirmed by X-ray diffraction. This parent diamond film with a thickness of
change in the characteristics causes the op- 50 |im is seen; thicker film, with a thickness
tical character of the film to change, as of 200 jLim and deposited under the same
shown in Fig. 21-11. Higher CH 4 concen- conditions as the 50 |im film, is found to be
 21.5 Realized and Anticipated Applications of CVD Diamond 353

black. The black color may be brought plasma-jet method, the combustion flame
about by the presence of quite a small method, and the DC plasma CVD method.
amount of sp2 bonding. The highest reported deposition rate was
One of the advantages of CVD diamond 1 mm/h using the plasma-jet method, but
over high-pressure diamond is feasibility the film obtained contained many pores.
of large area synthesis. Figure 21-12 shows In general, the quality of the diamond film
a photograph of diamond-coated silicon is quite closely related to the deposition
wafer and molybdenum; a diamond film as rate and the lower deposition rates prom-
large as 12 cm x 12 cm can be made. The ise better diamond films.
homogeneity in the thickness is fairly good
and the homogeneity in the characteristics
of the diamond, as examined by Raman 21.5 Realized and Anticipated
spectroscopy, is also good. Applications of CVD Diamond
Films thicker than 1 mm and continu-
ous film thinner than 0.4 |im can be fab- CVD of diamond is considered for ap-
ricated. High deposition rates of more plications with the following objectives:
than 100 |im/h were reported using the
1. Fabricating diamond-coated composite
material.
2. Fabricating three-dimensional shaped
diamonds.
3. Making high-purity diamonds.
The size limitations of single-crystal dia-
monds and the impurity of sintered dia-
monds provide the motivation to use CVD
diamonds, For example, it is necessary to
use pure diamond, containing no nondia-
mond phase, to achieve optical transpar-
ency, and single-crystal or CVD diamonds
Figure 21-11. Photograph of various diamond films are the only candidates for such applica-
of different thickness. tions. In order to make diamond IR (in-
frared) windows 50 mm in diameter, sin-
gle-crystal diamond is never used, but
transparent CVD diamond film is. The au-
thor will now review some applications us-
ing CVD diamond films.

21.5.1 Tools
Although diamond is the hardest mate-
rial, it is not a suitable material for cutting
steel or iron, because of the active reaction
with iron. The major uses of diamond cut-
Figure 21-12. Photograph of large area deposited di- ting tools are for cutting aluminum alloys
amond. and plastics. A recent requirement for
354 21 Diamond: Its Synthesis from the Vapor Phase and Applications

making automobiles light has been to use

aluminum alloys in automobile parts such
as engine blocks, transmission gear boxes,
etc. As a strong aluminum alloy is required Bonded IC Chip
Film IC Chip
for such applications, the silicon content
has been increased. The hardness of alu-
minum alloys increases with the silicon
content because of the precipitation of sili-
con-rich phases. A390, which is a typical, (b)
hard aluminum alloy containing 17wt.%
silicon, was expected to be used for engine
blocks, but neither cemented carbides
(WC + Co + TaC), which are the most pop-
ular material for cutting, nor conventional
sintered diamonds, have sufficient tool life
for such cutting operations.
A major development as regards cutting
tools occurred in the early 1970s with the
use of inserts coated with TiC, TiN,
A12O3, etc.; the tool life was improved by Figure 21-13. (a) Schematic diagram of TAB and (b)
a factor of several times. Many expected to TAB tools using diamond films.
realize 'diamond-coated tools' for cutting
aluminum alloys. Many have tried to de-
velop diamond-coated inserts and found time of this tool was found to be about 10
that the adhesion between diamond and times longer than that of conventional ma-
cemented carbide is not sufficient for prac- terials (e.g. sintered diamond). TAB tools
tical applications. The reason for this is using diamond film are shown in Fig. 21.13.
believed to be that diamond consists of The second way to apply CVD dia-
only covalent bonds and does not have any monds in tools is to use thick diamond film
compound phases. Although some practi- as if it were sintered diamond. Such tools
cal diamond-coated tools made using sili- were commercialized in 1989 (Nakamura,
con nitride and cemented carbide sub- 1991). Figure 21-14 shows the procedure
strates have nevertheless been commercial- for making such tools. Diamond films with
ized, they are not widely used. a thickness of 200 jum are made by hot-
However, applications other than cut- filament CVD. After being cut by an
ting look hopeful using diamond coating yttrium aluminum garnet (YAG) laser, the
technology, because strong adhesion is not substrates are dissolved by acid. The
required. Tape automated bonding (TAB) blanks are brazed onto a submount made
tools made of CVD diamond film have of cemented carbide. Figure 21-15 shows
been used practically since 1990 (Naka- some fabricated tools. The characteristics
mura etal., 1991). TAB is a packaging of such tools were confirmed by cutting
method, as shown in Fig. 21-13, and is uti- aluminum alloy (A390), and the tool life
lized in the fabrication of liquid crystal dis- was found to be several times longer than
plays. Precise edges to brazed electrodes that for a cemented carbide tool or a sin-
and tape carriers are required and the life- tered diamond tool, as shown in Fig. 21-16.
 21.5 Realized and Anticipated Applications of CVD Diamond 355

(a) Synthesis of CVD Diamond (b) Laser Cutting

Reaction Filament Laser

Vessel CVD Diamond
(0.1 mm thick)

~>

Feed

(c) Dissolution of Substrate (d) Freestanding CVD Diamond Pieces

Acid

z
(e) CVD Diamond Cutting Tools
(Endmill) CVD Diamond
(Insert) (0.1 mm thick)
CVD Diamond
(0.1 mm thick)

Brazing Brazing Cemented Carbide

Cemented Carbide

Figure 21-14. Procedure to fabricate diamond tools using thick diamond films.

More popular tools such as inserts for diamond-coated speaker diaphragm,

turning machines and milling machines which is the first CVD diamond product
can also be fabricated, and the characteris- (Fujimori, 1987). A diamond coating im-
tics were found to be superior to those of proved the highest frequency generated by
sintered diamond tools. an alumina diaphragm from 35 000 to
50 000 Hz.
A diamond diaphragm is considered to
21.5.2 Electrical Parts
be an ideal one. The poor machinability of
Diamond is well-known as the most natural diamond makes it impossible to
suitable material for acoustic parts be- make a natural diamond diaphragm. A
cause it has the largest sound velocity free-standing diamond diaphragm fabri-
(18 200m/s). The author has developed a cated by CVD was developed in 1990. Dia-
356 21 Diamond: Its Synthesis from the Vapor Phase and Applications

mond deposition was carried out by hot-

filament CVD on a silicon substrate. After
deposition, the silicon substrate was dis-
solved by a mixed acid of HF and H N 0 3 .
Figure 21-17 is a photograph of three such
diaphragms. They are 30 Jim thick and the
thickness distribution is controlled to with-
in 10%. The sound velocity was measured
by the vibration method as 16 500m/s,
which is a little less than the value for sin-
gle-crystal diamond. The properties of the
speaker in which one of these diaphragms
was used are shown in Fig. 21-18. The
Figure 21-15. Diamond tools fabricated according to highest frequency of this speaker was
the procedure in Fig. 21-14.
80 000 Hz, which is the highest frequency
that a dynamic speaker can generate.
Devices using high power need heat
v
Cemented Carbide sinks in order to avoid a substantial tem-
perature rise, because their characteristics
/ <Peeled Off>
change at higher temperatures. Single-
f
1
crystal diamonds, synthesized by the high-
pressure method, are used for heat sinks
Diamond Coated
for laser diodes or IMPATT diodes, which
generate quite a large amount of heat in a
small chip. Diamond's ultrahigh heat con-
CVD Diamond
j i__ ductivity enables such devices to operate
10 20 30 stably when protected by heat sinks. A
Cutting Time (mm) schematic diagram of a diamond heat sink
Figure 21-16. Characteristics of diamond tools for for a laser diode is shown in Fig. 21-19.
cutting aluminum alloy A390.
Heat sinks are the most obvious ap-
plication of CVD diamonds because of
their potential to be fabricated at a low
cost. The heat conductivity of CVD dia-
mond depends on the conditions of deposi-
tion. The author measured the heat con-
ductivity of 300 |im thick CVD diamond
film obtained by hot-filament CVD as
1600 W/mK. This value is considered to
be sufficient for application in the same
manner as single-crystal diamonds. Dia-
mond heat sinks using CVD diamonds
may be expected to become popular be-
cause of their low cost of synthesis.
Figure 21-17. Photograph of diamond speaker dia- Various types of heat sinks made from
phragms. CVD diamonds are shown in Fig. 21-19.
 21.5 Realized and Anticipated Applications of CVD Diamond 357

.DIAMOND
COATED
DIAPHRAGM
\
CO FREE STANDING
a 100 DIAMOND
LU DIAPHRAGM
LU

TITANIUM \
90 -DIAPHRAGM-^
o
CO

Figure 21-18. Characteristics of speakers

80 in which diamond diaphragms are used.
5,000 10,000 20,000 50,000 100,000

FREQUENCY [Hz]

Most of them are metallized in order to perature are summarized in Table 21-3 and
bond to laser diodes and submount ma- compared with those of Si, GaAs, and SiC.
terials. Diamond has relatively high mobility, a
low dielectric constant, and a high break-
down voltage. These characteristics sug-
21.5.3 Active and Passive Devices gest applications in refractory integrated
circuits, blue-emitting diodes, antiradia-
Diamond is a wide-band-gap semicon-
tion devices, etc.
ductor whose characteristics at room tem-
lib diamond is a p-type semiconductor,
as was reported in 1952 (Custers, 1952).
The first report on a diamond transistor
was published by Geis et al. in 1987 using
synthesized single crystal diamond and
point contacts (Geis et al., 1987). They re-
CVD Diamond ported that the transistor could operate
even at 500 °C, which suggested the feasi-
bility of high-temperature operation.
Epitaxial growth of diamond is consid-
ered to be a key technological factor in the
fabrication of diamond devices and CVD
methods are suitable growth techniques.
Shiomi et al. (1990) endeavored to obtain
high-quality epitaxial films and found that
using a (100) substrate and 6 % CH 4 in the
reactant gas for microwave plasma CVD
gave good epitaxial films.
Boron-doped epitaxial films were ob-
Figure 21-19. CVD diamond heat sinks for laser tained using B 2 H 6 in the reactant gas. The
diodes. characteristics of boron-doped epitaxial
358 21 Diamond: Its Synthesis from the Vapor Phase and Applications

Table 21-3. Characteristics of some semiconducting materials at room temperature.

Properties Si GaAs 3C-SiC Diamond

Energy gap (eV) 1.1 1.4 2.2 5.5

Dielectric constant, K 11.9 13.1 9.7 5.7
Thermal conductivity, crT (W/(cm K)) 1.5 0.5 4.9 22
Mobility (cm2/Vs) electrons 1500 8500 1000 2000
holes 450 400 70 2100
Electron saturated drift velocity 1 xlO 7 2xlO 7 2xlO 7 2.5 xlO 7
FSAT (cm/s) (7xlO 6 )

films were found to be quite similar to the donor or acceptor levels in Si, and this
those of natural lib diamonds (Fujimori results in small carrier concentrations at
e t a l , 1990). Figure 21-20 shows the tem- room temperature. The carrier mobility at
perature dependence of the carrier concen- room temperature is 600 cm2/V s, which is
tration and carrier mobility for boron- considerably less than the value in Table
doped epitaxial film. The acceptor energy 21-3, and the carrier mobility at 600 °C is
level is 0.37 eV, which is much larger than found to be 70 cm2/V s. The carrier mobil-
ity is affected by the substrate orientation
Temperature (°C)
and (100) is thought to be the most suitable
600 200 20 -70 substrate orientation.
— 1000 - As an n-type semiconductor is hard to
obtain, Schottky junctions have been stud-
ied by many researchers. Rectifying behav-
ior using polycrystalline and epitaxial film
has been reported. The surface roughness
of epitaxial films on which the metal
electrode is deposited was found to affect
the rectifying characteristics (Shiomi et al.,
1989). The surface morphology of epitax-
ial film is strongly affected by the concen-
tration of CH 4 in the reactant gas and this
affects the Schottky characteristics. The
surface of epitaxial film grown from 0.5 %
CH 4 is much rougher than that grown
from 6% CH 4 . Figure 21-21 shows the
Schottky characteristics of an aluminum
E 10 13 electrode with these films and it has been
found that using 6% CH 4 results in a
10 _
much better Schottky contact.
' ' 2 3 4
1000/7" IK"1) A field effect transistor (FET) and a
Figure 21-20. Temperature dependence of carrier light-emitting diode have been fabricated
mobility and carrier concentration for boron-doped successfully using boron-doped epitaxial
epitaxial film. films and Schottky contacts (Shiomi et al.,
 21.5 Realized and Anticipated Applications of CVD Diamond 359
10"' 1

10
0 2 4 6 8
0.5 1.0 1.5 2.0 DRAIN VOLTAGE (V)
(a) Voltage (V)
Figure 21-22. I-V characteristics of the first diamond
10" FET.

implantation. The I-V characteristics of

this FET suggest pinch-off operation,
which is the mode of operation used in
practice; however, the transconductance,
which represents the gain, was too small
for the FET to be considered for applica-
tions. Some developments in etching tech-
0.5 1.0 1.5 2.0 nology and metallization for ohmic con-
(b) Voltage (V) tact have led to progress in MESFETs and
Figure 21-21. Surface morphology of epitaxial films recent results are shown in Fig. 21-23 (Shi-
grown from reactant gas containing (a) 0.5 and (b) 6 % omi et al., 1994). However, the use of dia-
CH 4 . mond in practical applications is consid-
ered to be still far off, because the charac-
teristics of reported FET operation look
1989; Gildenblat et al., 1990; Fujimori quite primitive, as for silicon in the early
et al., 1991). Figure 21-22 shows the char- 1950s.
acteristics of the first diamond MESFET The photoconductivity of CVD dia-
(metal-semiconductor field effect transis- mond has been investigated and light sen-
tor) as reported by Shiomi et al. (1989). sors are the expected application using this
Electro-luminescent devices have also characteristic. The fabrication of a therm-
been reported using Schottky junctions. A istor using boron-doped film was reported
report of blue emission by point contact by Fujimori and Nakahata (1988). The ad-
suggests the possibility of practical blue- vantage of this device is that it can operate
emitting devices. Collins et al. (1989) re- at elevated temperatures as high as 600 °C.
ported on edge cathodoluminscence from
CVD diamonds, of which the energy was
21.5.4 Other Applications
5.3 eV. This also suggests the possibility of
ultraviolet emission. Electron microscopic analysis (EMA) is
Zeidler et al. (1991) reported on an FET a popular method of measuring the atomic
whose active layer was produced by ion composition in a small region, and this
 360 21 Diamond: Its Synthesis from the Vapor Phase and Applications

Gate voltage (V) of diamond. Windischmann and Epps

5.0 -
(1990) reported on a diamond X-ray mask
of 50 mm. A large and thin, free-standing
Gate length 4pm
4.0 — Gate width 36pm ^ ^ - * " -5
film (2 |im) is required for this application.
Channel depth 0.4 pm ^^^-~^^ The optical transparency characteristic
3.0 5 of diamond is essentially quite flat at wave-
lengths greater than 225 nm. The quality
15
2.0 — ^ — of diamond required for this application is
y^^^^'—^__ - 25 quite high and some improvement in the
1.0
- ^^—^IIIIII^—-— 35
45
quality of the film is required.
>
0 **^
^ r '. ^r~ i
— The high velocity of sound in diamond
-10 -20 -30 -40 has led to the idea of using diamond in
Voltage (V) surface acoustic wave (SAW) devices.
Figure 21-23. I-V characteristics of a recent diamond Filters, resonators, and clocks using
FET. piezoelectric materials are utilized in SAW
devices. Electrodes similar to the teeth of a
comb, so-called interdigital transducers
method needs a window for weak and low- (IDTs), generate acoustic waves and the
energy X-rays. A window made of CVD operating frequency is determined by the
diamond can also be used in practice and, dimensions of the interdigital transducer.
as carbon has a low atomic number, the The advantage in using diamond is just in
transmission of low-energy X-rays is excel- its having the highest sound velocity,
lent. This window is used as a vacuum bar- which leads to higher frequencies being at-
rier and is required to support a pressure tainable than by using other materials. Re-
difference of 1 atm (0.10 MN m" 2 ). A dia- cent progress in communication systems
mond film with a thickness of 0.4 |im is requires a high-frequency signal, so high-
fabricated and this window can transmit frequency filters are required.
low-energy X-rays characteristic of nitro- Nakahata et al. (1992) succeeded in fab-
gen, which the conventional beryllium ricating a diamond SAW filter operating at
window cannot transmit. 1 GHz. The structure of the device is
Fine lithography technology requires shown in Fig. 21-24. They used silicon for
X-ray lithography and a mask blank made the substrate on which to deposit 50 Jim

ZnO (Piezoelectric material)

Transducer
(IDT)

Acoustic wave
Diamond film Figure 21-24. Structure of a diamond SAW filter.
 21.6 Future Development and Problems to be Solved 361

diamond film and polished the diamond edge has been observed. Nitrogen contam-
surface to make the formation of 1 jum in- ination can be controlled using CVD
terdigitial transducers possible. As dia- methods, but hydrogen contamination
mond is not a piezoelectric material, ZnO also exists, as confirmed by Imai and Fuji-
film was deposited by conventional radio mori (1991). The author believes that small
frequency (RF) sputtering. The SAW ve- amounts of contamination, such as nitro-
locity observed was lOOOOm/s, which is gen, metals, or silicon, exist in the film and
about three times larger than for conven- such impurities have some effect upon the
tional materials such as SiO2 and LiNbO 3 . characteristics of diamond films. The ef-
The characteristics of SAW filters are fects are believed to be different for differ-
given in Table 21-4. ent impurities, depending on whether they
are in single-crystal or polycrystalline ma-
Table 21-4. Characteristics of diamond SAW filters. terial, and electrical or optical applications
may require purer diamond films.
Material Sound Frequency Required The lowest temperature of diamond for-
velocity using 1 urn IDT size for
IDT 2.5 GHz filter
mation accepted by most diamond re-
(m/s) (GHz) searchers is 300 °C. Almost all common
metals can be coated at this temperature.
LiNbO3 3 500 0.9 0.35 The deposition of diamond at 150°C has,
Quartz 3 200 0.8 0.32
ZnO/diamond 10000 2.5 1.00
however, been reported and this tempera-
ture would allow diamond to be coated
onto plastics. The real surface temperature
is difficult to measure and to control. The
adhesion between the film and the sub-
21.6 Future Development and strate is affected most by the substrate
Problems to be Solved temperature. The adhesion strength of dia-
mond films is lower than that of ceramic
Several commercial products have been and metal films, and one of the reasons is
developed from CVD diamonds and sig- believed to be that a compound phase con-
nificant progress has been made in new sisting of diamond and substrate is hard to
applications of diamonds. As mentioned form.
previously, CVD diamond overcomes the The cost of diamond is essentially deter-
limitations of conventional diamond's ap- mined by the deposition rate and the size
plications. However, several targets are be- of the equipment. The highest deposition
ing considered for development to allow rate reported is 1 mm/h, and this value is
CVD diamonds to be utilized practically in sufficient to make even bulk diamonds.
various fields: However, the quality of the CVD diamond
obtained at such a growth rate is not suffi-
1. Distinguish phases other than diamond.
cient for most applications. Hence im-
2. Low-temperature deposition.
provements in the growth rate or the film
3. Low-cost process.
quality are desirable.
CVD diamond film has been confirmed Study of the mechanism of diamond for-
as type II a diamond by measuring its opti- mation should help in solving these prob-
cal transmission. However, unidentified lems. Various approaches to examining the
absorption from 500 nm to the absorption mechanism, such as gas-phase analysis,
362 21 Diamond: Its Synthesis from the Vapor Phase and Applications

study of surface structure, simulation of vacuum electron-emitting devices will be

crystal growth, observation of nucleation, made using diamonds.
etc., have been tried. The role of atomic In order to achieve such difficult goals,
hydrogen in the reactant gas has become studies of surface structure, the mechanism
progressively better understood in recent of formation of diamonds, and crystal de-
studies. The author considers such studies fects in diamonds are considered to be im-
to be helpful for the better understanding portant.
of film formation other than for diamond It is common knowledge that utilizing
from the gas phase. novel materials can take quite a long time,
For electronic applications, single-crys- but it took only a few years for diamond
tal, large diamonds are desirable, as in sil- films to be used for practical applications.
icon devices, and this is considered to be However, this short history indicates that
the ultimate target of diamond technology. new fabrication technology brings new ap-
The approach toward this target involves plications which conventional technology
attempts to achieve hetero-epitaxial had never dreamed of. Diamond's excel-
growth of diamond on other substrate ma- lent properties have not yet been fully uti-
terials. The phenomenon of hetero-epitax- lized and further possibilities are sure to
ial growth was reported by Stoner and become known in the near future.
Glass (1992) with oriented single-crystal
diamond particles grown on a silicon car-
bide substrate. Some success in hetero-
epitaxial growth on an Si substrate was 21,7 References
reported and the key to the technology in- Aisenberg, S., Chabot, R. (1971), /. Appl. Phys. 42,
volves controlling the nucleation process 2953.
by appropriate electrical biasing of the Angus, I C , Will, H. A., Stanko, W. S. (1968),
/. Appl. Phys. 39, 2915.
substrate for instance (Stoner et al., 1992). Collins, A. T, Kamo, M., Sato, Y. (1989), J. Phys.;
Other observations, such as on single-crys- Condens. Matter 1, 1029.
tal Ni or on epitaxially grown Ni film on Custers, J. F. H. (1952), Physica (Utrecht) 18, 489.
Davice, G. (1979), in: Properties of Diamond: Field,
MgO, have also been reported. However, J. E. (Ed.). London: Academic, p. 165.
the crystal size was only of the order of a Derjaguin, B. V., Fedoseev, D. V., Lukyanovich, V. M.,
micrometer and the goal still looks distant. Spitsyn, B. V., Ryabov, V. A., Laverentyev, A. V.
(1968), /. Cryst. Growth 2, 380.
Another ultimate target is making n-type Derjaguin, B. V., Spitsyn, B. V., Bouilov, L. L.,
diamond. The lack of n-type semiconduc- Klochkov, A. A., Grorodetsk, A. E., Smolyanov,
tors limits the application of diamond V. (1976), Sov. Phys. - Solid State 21, 676.
Eversol, W. G. (1958), U.S. Patents 3 030187,
semiconductors to a small area. Especially 3 030188.
for microwave devices, n-type semicon- Fujimori, N. (1987), New Diamond 3 (1), 20.
ductors will be necessary. Fujimori, N., Nakahata, H. (1988), New Diamond
5 (2), 32.
Recent reports suggest that diamond Fujimori, N., Imai, T., Nakahata, H. (1990), Jpn.
can be used as an electron-emitting mate- J. Appl Phys. 29, 824.
rial. A (111) surface passivated by hydro- Fujimori, N., Nishibayashi, Y, Shiomi, H. (1991),
Jpn. J. Appl. Phys. 30, 1063.
gen may have an energy state of negative Geis, M. W, Rathman, D. D., Ehrlich, D. X, Murphy,
electron affinity and this characteristic R. A., Lindley, W. T. (1987), IEEE Electron Device
should result in cold-cathode electron Lett. EDL-8, 341.
Gildenblat, G. S., Grot, S. A., Hatfield, C. W, Wron-
emission at room temperature. If this phe- ski, C. R., Badzian, A. R., Badzian, T., Messier, R.
nomenon can be confirmed, practical (1990), MRS Symp. Proc. 162, 297.
 21.7 References 363

Hirose, Y., Mituizumi, M. (1988), New Diamond Sawabe, A., Inuzuka, T. (1985), Appl Phys. Lett. 46,
4 (3), 34. 145.
Imai, T., Fujimori, N. (1991), in: Proc. of the 4th Shiomi, H., Nishibayashi, Y, Fujimori, N. (1989),
Diamond Symposium. Tokyo: New Diamond Fo- Jpn. J. Appl. Phys. Lett. 28, L2153.
rum, p. 103. Shiomi, H., Tanabe, K., Nishibayashi, Y, Fujimori,
Kamo, M., Sato, Y, Matsumoto, S., Setaka, N. N. (1990), Jpn. J. Appl. Phys. 29, 34.
(1983), /. Cryst. Growth 62, 642. Shiomi, H., Nishibayashi, Y, Toda, N., Shikata, S.,
Kurihara, K., Sasaki, K., Kawarada, M., Hosino, K. Fujimori, N. (1994), in: Proc. 4th Int. Conf on New
(1988), Appl Phys. Lett. 52, 437. Diamond Science and Technology: Saito, S., Fuji-
Matsumoto, S., Sato, Y, Kamo, M., Setaka, N. mori, N., Fukunaga, O., Kamo, M., Kobashi, K.,
(1982), Jpn. J. Appl. Phys. 42, L183. Yoshikawa, M. (Eds.). Tokyo: Myu, p. 665.
Matsumoto, S., Hino, M., Kobayashi, T. (1987), Stoner, B. R., Glass, J. T. (1992), Appl Phys. Lett. 60,
Appl. Phys. Lett. 51, 731. 698.
Nakahata, H., Hachigo, A., Shikata, S., Fujimori, N. Stoner, B. R., Ma, G.-H., Wolter, S. D., Glass, J. T.
(1992), IEEE Ultrasonic Symp. p. 377. (1992), Phys. Rev. B45, 11067.
Nakamura, T. (1991), Machine Tool Machine-Tech- Tsuno, T, Imai, T, Nishibayashi, Y, Hamada, K.,
nol. (2) 122. Fujimori, N. (1991), Jpn. J Appl Phys. 30, 1063.
Nakamura, T., Tanaka, K., Yashiki, T., Kumazawa, Windischmann, H., Epps, G. F. (1990), /. Appl Phys.
Y, Fujimori, N., Nakai, T., Nakatani, M. (1991), 68, 5665.
in: Proc. of the 4th Diamond Symposium. Tokyo: Zeidler, J. R., Hewett, C. A., Nguye, R., Zeidler,
New Diamond Forum, p. 75. Z. R., Wilson, R. G. (1991), IEEE Electron Device
Nishibayashi, Y, Shiomi, H., Nakahata, H., Imai, T., Lett. 12, 602.
Fujimori, N. (1989), in: Proc. of the 36th Meeting
of the Society of Japanese Applied Physics. Tokyo:
Society of Japanese Applied Physics, p. 481. General Reading
Sato, Y, Kamo, M., Kanda, H., Setaka, N. (1980),
J. Surf. Sci. Soc. Jpn. 1, 60. Lettington, A. H., Steeds, J. W. (1993), Thin Film
Sawabe, A. (1988), New Diamond 4 (4), 8. Diamond. London: Chapman & Hall.