Characterization of pigments in coating formulations

for high-end ink-jet papers

Ales Hladnik*, Tadeja Muck
Institut za celulozo in papir (ICP), 1000 Ljubljana, Slovenia
Received 15 March 2002; received in revised form 19 April 2002; accepted 28 May 2002
Abstract
The eect of the two inorganic pigments widely used for coating of high-end ink-jet papersamorphous silica and
precipitated calcium carbonate (PCC)and poly(vinyl alcohol) binder on ink-jet paper and print quality was studied.
Multivariate analysis of results revealed that the type and the proportion of the pigments signicantly inuences several
paper and print characteristics as determined by wicking, mottling, striking through, ink absorption and electrical
surface resistance. Water absorption characteristics of paper are not related to ink-jet print quality but rather to the
hydrophobicity/hydrophilicity of base paper sheet. The eect of the coat weight on the printed paper performance is
considerable only in cases where 100% silica or silicaPCC combinations were applied. In coatings with 100% PCC,
this eect is much smaller in comparison to that of the pigment itself.
#2002 Elsevier Science Ltd. All rights reserved.
Keywords: Ink-jet paper; Print quality; Coating pigments; Silica; Precipitated calcium carbonate; Multivariate analysis
1. Introduction
Ink-jet (in the following text: IJ) papers that are
available on the market today are predominantly
uncoated, multipurpose papers. They are suitable
for most routine applications. However, uncoated,
surface sized grades do not allow for high-end print
and image appearance. The latter is achieved by
applying special coating formulations onto the
paper surface. In this study two coating pigments
silica and precipitated calcium carbonate (PCC)
together with dierent amounts of a binder
poly(vinyl alcohol) (PVOH)were investigated in
order to evaluate inuence of coated paper para-
meters on the IJ paper and on the resultant print
quality. The results were interpreted using multi-
variate methods, principal components analysis
(PCA) and cluster analysis.
1.1. Pigments for ink-jet paper
Demands for high-end IJ papers are manifold.
They must provide good xation of the ink to the
paper surface, quick ink solvent absorption and
minimize ink bleeding and wicking while at the
same time retaining favourable gloss, ink optical
density and colour delity [1]. Image print-through
to the back side of the paper should be kept as low
as possible. The image should also resist smearing
when wetted. Coating formulations fullling such
diverse tasks and providing high print quality
0143-7208/02/$ - see front matter # 2002 Elsevier Science Ltd. All rights reserved.
PI I : S0143- 7208( 02) 00050- 5
Dyes and Pigments 54 (2002) 253263
www.elsevier.com/locate/dyepig
* Corresponding author. Tel.: +386-1-200-28-00; fax:
+386-1-426-56-39.
E-mail address: ales.hladnik@icp-lj.si (A. Hladnik).
usually consist of silica and/or PCC as pigments
and PVOH as a binder.
Amorphous silicas (precipitated and fumed)
exhibit high porosity, hydrophilicity and surface
area [2]. Due to these properties, fast absorption
of the aqueous ink vehicle is achieved, therefore
speeding up the ink drying time and increasing the
edge sharpness. Potential drawbacks to the use of
synthetic silica are its high production costs and
the less favourable rheology that it generates in
formulation. Also, porous formulations contain-
ing silica are inherently weak. Thus, a high-
strength binder such as PVOH is needed. PVOH is
also valued for its hydrophilicity, which promotes
absorption of the aqueous ink vehicle.
Recently, another specialty pigment for IJ
papers has been oeredprecipitated calcium
carbonate (PCC). Due to its controlled manu-
facturing parameters PCC can be eectively used
as an alternative to silica-based pigments. Con-
cerning the xation of IJ inks onto the coated
paper surface, it has been proposed [3] that unlike
silica coatings, which hold the uid phase includ-
ing the dye of IJ ink, specialty PCC holds only
the dye and allows the uid phase to pass
through.
1.2. Principal components analysis
Principal components analysis (PCA) is a math-
ematical method that transforms complex multi-
variate data set into a new perspective in which the
most important information is made more obvious.
It enables identication of natural patterns or
structures in the data by visualizing the relation-
ships among the samples as well as among the
variables studied.
PCA is conceptually similar to some other mul-
tivariate methods, such as correspondence analysis,
singular value decomposition, eigenvector analysis
and factor analysis [4]. It is based on a construc-
tion (extraction) of a new set of variables called
principal components (PCs) that are completely
uncorrelatedorthogonalto each other.
A score s
ip
is a projection of sample i along a
principal component PCp and is, mathematically
speaking, a linear combination of the original
variables [5]:
s
ip

X
j
v
jp
x
ij
1
Here v
jp
is the loading of variable j on PCp and
x
ij
is the value of the ith object for original variable
j. In matrix notation one can write:
S XV 2
Here S, X and V are the matrices of scores, the
original variables and the loadings, respectively.
The equation can be rearranged into:
X SV
T
3
In practice, this means that the data matrix X
can be decomposed into a product of two matri-
ces, one about the samples (S) and the other con-
taining information about the variables (V). V
T
denotes the transposition of matrix V, obtained
from V by switching its rows and columns. Cal-
culation of these new matrices, that is, extraction
of PCs proceeds successively so that each of the
PCs captures as much as possible of the variation
that has not been explained by the former PC:
PC1 maximizes the covariance in the original data
set and the following PCs maximize the covariance
in the residual matrices left after extracting the
former PCs. The decomposition is graphically dis-
played in Fig. 1. The S matrix contains the scores
of m samples on n principal components: rows
represent individual samples and columns denote
principal components. The square V
T
matrix con-
sists of the loadings (=columns) of the n original
variables on the n latent variables (=rows). By
retaining only the rst few signicant components,
in our case PC1 to PC3, one eliminates irrelevant
information since higher components contain pre-
dominantly data noise (grey parts of matrices).
The way in which PCA is used most frequently
is by graphically displaying the calculated scores
and loadings in the coordinate system of a rst few
PCs. Such interpretation enables visualization of
patterns in the data and reveals similarities and
dissimilarities between individual samples as well
as relationships between the original variables.
254 A. Hladnik, T. Muck / Dyes and Pigments 54 (2002) 253263
Principal components can be regarded as com-
posite variables constructed from original ones as
their linear combination. Sometimes it is possible
to assign a physical meaning to the rst few prin-
cipal components.
1.3. Cluster analysis
Cluster analysis is a data analysis technique that
allows one to organize a set of observations
(objects, samples) described by variables, into
groups. It encompasses a number of dierent
classication algorithms, such as tree clustering,
two-way joining or K-means clustering [6]. In the
tree clustering algorithm objects are joined toge-
ther into successively larger clusters according to a
certain measure of similarity or distance. Wards
method [7], as applied in this study, uses an ana-
lysis of variance (ANOVA) approach to evaluate
the distances between clusters. A typical result of
tree clustering is a dendrogram or hierarchical tree
(Fig. 2). Here the horizontal axis denotes the
linkage distance. In the beginning, each object (i.e.
sample or case) represents a class by itself (left
part of the plot). By linking together more and
more objects and aggregating larger and larger
clusters of increasingly dissimilar elements we
nally come to a point where all objects are linked
together (right part of the dendrogram). The more
similar two objects or groups of objects are, the
closer to the left of the dendrogram they are
linked. As the result of a successful tree clustering
analysis, one can detect and interpret individual
clusters or branches.
2. Experimental
In the study two fundamentally dierent wood-
free base paper types were examined: an unsized
paper (samples designated as V) and a rosin sized
(samples S) paper. These samples were coated, using
a benchtop drawdown rod coater, with coating for-
mulations consisting of one or two pigments
amorphous silica and/or PCCand a PVOH bin-
der (Table 1). Rather than making optimized for-
mulations, only simple pigmentbinder blends
were set up to illustrate dierences in pigments
behaviour. Also, the amount of PVOH3050
parts per hundred parts of pigmentwas higher
than that used in real high-end IJ coatings, espe-
cially for the coating colours with PCC. All of the
formulations were prepared at 21% solids. The
coat weights ranged from 3.4 to 6 g/m
2
.
In order to evaluate the IJ print quality, a test
form was printed onto the paper sheets with an
EPSON Stylus Color 900 printer. Printing para-
meters were kept at their default values with the
resolution set to 720 dpi.
Fig. 1. Decomposition of matrix X into matrices S and V
T
.
Table 1
Coating formulations
Ingredients (parts) Formulation
F G H
Silica A 80 50
Silica B 20
Precipitated calcium carbonate (PCC) 100 50
Poly(vinyl alcohol) (PVOH) 30 50 50
A. Hladnik, T. Muck / Dyes and Pigments 54 (2002) 253263 255
Altogether 11 parameters that described paper
samples, before as well as after IJ printing process,
were measured (Table 2). The characteristics of non-
printed specimen were determined through evalua-
tion of porosity (POROS), electrical surface resis-
tance (RESIST_NP), Cobb value (COBB) and PDA
value (PDA). The latter was obtained from the
shape of ultrasound intensity vs. time curve using
a Penetration Dynamics Analyser and is an indi-
cation of water penetration rate into the paper
sheet structure. IJ print quality was assessed by
performing the classical oset K&N ink absorption
Fig. 2. Tree clustering of samples.
Table 2
Measured parameters of coated and printed papers
Variable Range
Non-printed paper samples Porosity (Gurley); s POROS 56300
Electrical resistance; Ohm RESIST_NP 1.07.2
Water absorption (Cobb), g/m
2
COBB 16100
Water penetration rate; PDA 675315
Printed paper samples K&N ink absorption;% K&N 24.954.3
Striking through; STRIKE 02
Electrical resistance; Ohm RESIST_P 0.62.1
Wicking; WICK 5.035.0
Bleeding; BLEED 11.835.3
Mottling; MOTTL 3343
Coating Coat weight; g/m
2
COATW 3.46.0
256 A. Hladnik, T. Muck / Dyes and Pigments 54 (2002) 253263
test (KN), a visual evaluation of ink striking through
(STRIKE), electrical surface resistance (RESIST_P)
measurements and using image analysis techniques
for determination of wicking (WICK), bleeding
(BLEED) and mottling (MOTTL). Since we also
wanted to examine eect of coat weight (COATW)
on IJ paper performance, this variable was also
monitored.
3. Results
The correlation coecients table, showing linear
relationships within each pair of the 11 variables
examined (Table 3), reveals several strong inter-
actions. From the corresponding values one can,
for example, see that an increase in wicking is
accompanied by a substantial reduction in K&N
ink absorption (r=0.93) and an about equally
strong increase in coated paper porosity (r=0.93).
To explore such dependencies in more depth, PCA
was run on a data matrix consisting of 12 cases
paper samples diering in base paper sizing
degree (V or S), coating pigment type (F, G and
H) and coat weight (from 3.5 to 6.0 g/m
2
)and 11
variablespaper structural, sorptive and printing
properties.
Results of the analysis indicate (Fig. 3) that the
rst four composite variablesPC1, PC2, PC3
and PC4together account for nearly 94% of
total variability in original data. The remaining
seven higher principal componentsPC5PC11
Table 3
Linear correlations among variables
POROS RES_NP COBB PDA K&N STRIKE RESIST_P WICK BLEED MOTTL COATW
POROS 1.00
RESIST_NP 0.74 1.00
COBB 0.02 0.24 1.00
PDA 0.01 0.27 1.00 1.00
K&N 0.91 0.76 0.11 0.13 1.00
STRIKE 0.78 0.64 0.42 0.44 0.85 1.00
RESIST_P 0.95 0.79 0.04 0.08 0.86 0.79 1.00
WICK 0.93 0.75 0.00 0.04 0.93 0.79 0.93 1.00
BLEED 0.53 0.37 0.20 0.16 0.46 0.35 0.39 0.35 1.00
MOTTL 0.84 0.65 0.24 0.26 0.84 0.76 0.77 0.83 0.57 1.00
COATW 0.10 0.47 0.09 0.09 0.34 0.13 0.11 0.22 0.06 0.02 1.00
Fig. 3. Plot of cumulative variance and eigenvalues for individual PCs.
A. Hladnik, T. Muck / Dyes and Pigments 54 (2002) 253263 257
Fig. 4. Variables in PC1PC2 plane.
Fig. 5. Variables in PC1PC3 plane.
258 A. Hladnik, T. Muck / Dyes and Pigments 54 (2002) 253263
Fig. 6. Variables in PC2PC3 plane.
Fig. 7. Samples in PC1PC2 plane.
A. Hladnik, T. Muck / Dyes and Pigments 54 (2002) 253263 259
explain only a very small portion of information
and since their eigenvalues are below 1.0, they can
be neglected. In fact, most of the relevant infor-
mation (87%) can be represented by the rst three
principal components. Diagrams showing corre-
lations between initial variables and the rst three
principal componentsloading plotsas well as
diagrams representing relationships within sam-
plesscore plotsare displayed in Figs. 49.
From the calculated squared loadings values,
another interesting diagram can be constructed
(Fig. 10) that displays contributions (in %) of
individual original variables to the principal com-
ponents. Another useful plot from which simila-
rities as well as dierences among samples can be
interpreted is a dendrogram (Fig. 11) produced as
a result of cluster analysis. Both diagrams will be
discussed in detail in the following section.
Fig. 8. Samples in PC1PC3 plane.
Fig. 9. Samples in PC2PC3 plane.
260 A. Hladnik, T. Muck / Dyes and Pigments 54 (2002) 253263
Constellation of individual variables in PC1PC2
plane (Fig. 4) shows some distinctive patterns. PC1
is the horizontal axis and PC2 is the vertical axis.
First of all, there is an evident grouping of points on
the left side of the diagramSTRIKE, MOTTL,
RESIST_NP, RESIST_P, POROS and WICK
indicating strong linear correlations among these
variables: papers with high porosity (POROS) exhi-
bit increased electrical surface resistance both on
non-printed samples (RESIST_NP) as well as on
printed samples (RESIST_P) and are characterized
by excessive wicking (WICK), bleeding (BLEED)
and striking through (STRIKE). These variables
all tend to have high negative loadings on PC1 and
very small contribution (i.e. orthogonal projection)
to PC2 and therefore represent PC1 to a large degree.
The variable on the opposite side of the diagram
the K&N ink absorption valuealso strongly deter-
mines PC1 but is negatively related to the above
mentioned variables: its increase is accompanied by a
decrease in STRIKE, MOTTL, RESIST_NP,
RESIST_P, POROS and WICK values.
From the position and proximity of points
COBB and PDA it is clear that these two variables
are very closely related to each other (see Table 3:
r=0.996) but not to any other variable. The
physical meaning of PC2 has obviously much to
do with these two parameters, since they are the
only ones that have high contribution to this axis.
The coat weight (COATW) parameter is located
relatively close to PC1PC2 origin meaning that it
can be related neither to PC1 nor to PC2. If the
PC1PC3 loading diagram (Fig. 5) is examined,
however, it is apparent that this variable determines
signicantly the third principal component, PC3,
since the contribution to this principal component is
considerable.
Similar conclusions can be drawn when observing
Fig. 10, which graphically shows breakdown of ori-
ginal variables by the rst four principal compo-
nents. With the exception of bleeding, whose
contribution is divided between PC1, PC3 and PC4,
all other variables seem to grasp the essence of more
or less a single principle component: variables KN,
POROS, STRIKE, WICK, MOTTL, RESIST_NP
and RESIST_P of PC1, COBB and PDA represent
PC2 and COATW represent PC3. Contribution of
variables to higher axes (PC5PC11), designated
as Residuals, is negligible.
Score diagrams give us indication of similarities/
dissimilarities and patterns among samples. PC1
in the PC1PC2 score plot (Fig. 7) separates points
into two groups: samples on the left part of the
diagram designated as G/5.7/V, G/3.4/V, G/5.5/S
and G/3.5/S form one cluster and F and H samples
on the right belong to the other one. From what
has been previously said about the variables hav-
ing highpositive or negativecontributions to
PC1 and bearing in mind the fact that G samples
were coated with PCC, it is possible to conclude
that physical meaning of the rst principal com-
ponent is related to coated paper surface char-
acteristics and, consequently, to ink behaviour.
These phenomena are decisively inuenced by the
Fig. 10. Contribution of individual variables to principal components.
A. Hladnik, T. Muck / Dyes and Pigments 54 (2002) 253263 261
coating pigment type. Wherever paper was coated
with PCC only, samples are marked by high por-
osity, ink-jet ink striking through, mottling and
wicking are pronounced, penetration rate of K&N
oset ink, on the other hand, is low. Non-printed
samples as well as ink-jet printed samples show
high electrical resistance.
Similarly, PC2 splits samples into an upper
and a lower group. The former consists entirely
of non-sized samples (V), the latter consists of
sized (S) samples only. The second principal com-
ponent is apparently related to hydrophilicity/
hydrophobicity of the base paper since the Cobb
value and the PDA integral are both an indication
of water penetration rate into the paper sheet. Note
that the principal components are, by denition,
orthogonal, i.e. uncorrelated to one another,
meaning that factors inuencing PC1 are dierent
from those representing PC2.
Eleven per cent of the total data variability can
be explained by PC3. Since the coat weight
(COATW) is the key parameter through which
this principal component can be interpreted, it is
natural that, in the PC1PC3 score plot (Fig. 8), it
separates samples according to their coat weight:
from 5.7 g/m
2
in G/5.7/V (highest positive score)
to 3.5 g/m
2
in F/3.5/V (highest negative score).
In order to study how individual samples are
similar to each other or to what extent (and why)
they dier, it isapart from looking at PCA score
plotsinstructive to observe dendrogram pro-
duced by cluster analysis (Fig. 11). Samples having
the shortest linkage distance are the most similar
to each other: G/5.5/S and G/3.5/S; F/3.6/S, H/
4.0/S and H/5.0/S; H/6.0/V and F/5.9/V; H/3.9/V
and F/3.5/Vsee also PC1PC2 score plot
(Fig. 7). The bigger the linkage distance, that is,
the further to the right, the less similar (groups of)
samples are being liked to one another. In the end,
even the two most distant clusters are linked.
A closer inspection of samples agglomeration
reveals some more interesting points. This is seen
in a study of gradual grouping of the last four
samples: H/6.0/V, F/5.9/V, H/3.9/V and F/3.5/V.
Fig. 11. Dendrogramsimilarities among samples.
262 A. Hladnik, T. Muck / Dyes and Pigments 54 (2002) 253263
Earliest linkagesH/6.0/V and F/5.9/V; H/3.9/V
and F/3.5/Vare obviously due to the similarities
in coat weights. Although not being coated with
the same coating pigment, samples H/6.0/V and F/
5.9/V have more in common to each other than
with the other partner of the same pigment type
but considerably lower coat weight. Note that all
four papers are unsized. Similar considerations
can also be applied to the middle four sized sam-
plesF/5.8/S, H/5.0/S, H/4.0/S and F/3.6/S
although the picture here is less clear. Linkage at a
distance of 1.7 indicates that there is a dis-
tinctive separation of both groups with respect to
the papers hydrophilic/hydrophobic character.
While for the paper samples coated with PCC
G/5.5/S, G/3.5/S, G/5.7/V and G/3.4/Vthe link-
age at a distance of 1.0 also connects the two sized
samples with the two unsized ones, the left-most,
earliest linkage of G samples shows an important
dierence when compared to F and H papers. A
decisive similarity criterion is the type of pigment
used in coating rather than the coat weight: G/3.5/
S is more similar to G/5.5/S than would be, for
instance, to F/3.6/S. The right-most linkage at 5.2
shows that the four samples coated with PCC
form one distinctive group compared to silica- and
PCCsilica-coated papers, in which the type of
pigment does not determine their characteristics in
such a pronounced way.
The results of this study conrm that the correct
pigment selection and pigment:binder ratio are of
vital importance for the IJ print quality. Printed
paper parameters such as wicking, mottling, striking
through, K&N absorption and electrical surface
resistance are decisively determined by the type and
proportion of pigments in a coating formulation. It
has been demonstrated that high paper porosity and
electrical resistance are accompanied by pro-
nounced mottling, wicking and ink strike through.
Markedly poorer performance of precipitated cal-
cium carbonate with respect to two silica pigments
is due to lower PCC quality and lower surface area
compared to silicas. The recommended amount of
binder lies for PCC between 7 and 15 parts per
hundred parts of pigment, while for silica these
numbers are much higherbetween 30 and 40
parts. As the PCA reveals, water absorption char-
acteristics of paper expressed through Cobb and
through PDA values are not related to IJ print
quality but rather to the hydrophobicity/hydro-
philicity of the base paper sheet, i.e. its degree of
sizing. The coat weight aects the printed paper
performance only to some extent. Only in the case
of papers that had been coated with 100% PCC, is
the inuence of coating pigment much more
important than that of coat weight.
References
[1] Sangl R, Weigl J. Kostengu nstige Herstellung ink-jet-
geeigneter Papiere bei hohen Produktionsgeschwindigkei-
ten. Das Papier 1998;10A:V1090V115.
[2] Ryu RY, Gilbert RD, Khan SA. Inuence of cationic addi-
tives on the rheological, optical and printing properties of
ink-jet coatings. TAPPI Journal 1999;82(11):12834.
[3] Donigian DW, Wernett PC, Mc Fadden MG, Mc Kay JJ.
Ink-jet dye xation and coating pigments. TAPPI Journal
1999;82(8):17582.
[4] Wold S, Esbensen K, Geladi P. Principal component ana-
lysis. Chemometrics and Intelligent Laboratory Systems
1987;2:3752.
[5] Massart DL, Vandeginste BGM, Buydens LMC, De Jong
S, Lewi PJ, Smeyers-Verbeke J. Principal Components. In:
Massart DL, editor. Handbook of chemometrics and quali-
metrics: part A. Amsterdam: Elsevier; 1997. p. 52035.
[6] StatSoft, Inc. Cluster analysis, electronic statistics text-
book. Tulsa, OK: StatSoft; 1999 Available from: http://
www.statsoft.com/textbook/stathome.html..
[7] Ward JH. Hierarchical grouping to optimize an objective
function. Journal of the American Statistical Association
1963;58:236.
A. Hladnik, T. Muck / Dyes and Pigments 54 (2002) 253263 263