Journal of Non-Crystalline Solids 282 (2001) 188±196

www.elsevier.com/locate/jnoncrysol

The di€usion mechanism of tin into glass governed by redox

reactions during the ¯oat process
Yasuo Hayashi a,*, Kiyoshi Matsumoto a, Masahiro Kudo b,1
a
Research Center, Asahi Glass Co. Ltd., 1150 Hazawa, Kanagawa-ku, Yokohama 221-8755, Japan
b
Department of Applied Physics, Seikei University, 3-3-1 Kitamachi, Kichijyoji, Musashino 180-8633, Japan
Received 14 August 2000; received in revised form 4 December 2000

Abstract
The di€usion mechanism of tin into glass was investigated using a lab-scale ¯oat apparatus, with which it was
possible to independently vary the heating parameters, in order to determine the reasons for the characteristic tin
penetration pro®le of ¯oat glass. Tin penetration pro®les of glass samples with heating at various temperatures, times
and atmospheres were measured by means of SIMS. The tin enriched inner layer, which is characteristic of the tin
penetration pro®le of ¯oat glass, was seen to be formed by heating at more than 800°C. It was found that the depth
of the tin enriched inner layer was proportional to the holding time at the maximum temperature during the heat
treatment, and was inversely proportional to the Fe3‡ concentration in the glass. It was also proven that the tin
enriched inner layer was formed by penetration of hydrogen from the atmosphere through the molten tin into the
glass. These facts indicate that the reaction between hydrogen and Fe3‡ is involved with the formation of the tin
enriched inner layer. Consequently, it has been proposed that the formation mechanism of the tin enriched inner
layer is governed by two redox reactions and the di€usion behaviors of both Sn2‡ and Sn4‡ . Namely, one of these
two redox reactions is the reduction of Fe3‡ to Fe2‡ due to hydrogen, resulting in the formation of a reduced surface
layer. Another is the oxidation of Sn2‡ to Sn4‡ due to Fe3‡ in the glass. Furthermore, it was revealed by the analysis
of commercial ¯oat glass with various thicknesses that the depth of the tin-enriched inner layer is also in¯uenced by
the stretch coecient for the attenuation of the glass during the ¯oat process. These analytical results leading to a
successful control of tin penetration into glass during the ¯oat process are discussed in detail. Ó 2001 Elsevier
Science B.V. All rights reserved.

1. Introduction ¯oat method in which molten glass is ¯oated in a

bath of molten metallic tin [1]. Therefore, the ¯oat
In order to provide optically ¯at and distortion glass has two di€erent faces. It is known that the
free surfaces, plate glass is generally produced by a bottom surface, which is in contact with the tin,
exhibits properties di€erent from those of the top
*
surfaces, such as resistance to weathering [2] and
Corresponding author. Tel.: +81-45 374 8792; fax: +81-45 the adsorption properties [3,4]. The major reason
374 8892.
E-mail addresses: hys@agc.co.jp (Y. Hayashi), kudo@apm.
for these di€erences in the surface properties of
seikei.ac.jp (M. Kudo). each face is considered to be due to the amount of
1
Tel.: +81-422 37 3779; fax: +81-422 555 471. tin that penetrates into the glass during the ¯oat
0022-3093/01/$ - see front matter Ó 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 0 2 2 - 3 0 9 3 ( 0 1 ) 0 0 3 1 9 - 2
 Y. Hayashi et al. / Journal of Non-Crystalline Solids 282 (2001) 188±196 189

process. Consequently, it is regarded as important, 2. Experimental

from the viewpoint of industrial applications, to
control the penetration of tin into the glass during 2.1. Sample preparation
the ¯oat process.
In the ¯oat process, molten glass is in contact Commercial soda-lime ¯oat glass with various
with molten tin at about 1100°C where it is formed thicknesses was adopted for the examination of the
into a continuous ribbon. The glass ribbon is tin penetration pro®les in ¯oat glass. All analyzed
cooled down while advancing on the bath of samples, which were manufactured by the same
molten tin, and it is then lifted o€ the bath at ¯oat line, had essentially the same bulk composi-
about 600°C. The atmosphere of the ¯oat bath is tion including iron content. Float glass of thick-
kept in a reducing condition by a N2 =H2 gas ¯ow nesses less than the equilibrium thickness of 6 mm
in order to prevent oxidation of the metallic tin. was shaped with an assisted direct stretch method,
Under these conditions, plate glass with uniform while for thicknesses greater than the equilibrium
thickness and a smooth surface is shaped and thickness a carbon fender method was used for
sized. During the processing of ¯oat glass, inter- shaping [1].
actions occur between the glass ribbon and the tin Using the lab-scale ¯oat apparatus, two di€erent
bath wherein components of the glass di€use out procedures were used to di€use tin into the glass.
and tin penetrates into the glass [5,6]. It is known The ®rst procedure was to heat glass placed on
that the tin penetration pro®les exhibit tin en- metallic tin set in an aluminous porcelain crucible.
riched inner layers located at depths of several The heat treatment was performed in a reducing
micrometers from the surface. These have been atmosphere of N2 =H2 using a tube furnace with a
called `humps' or `bumps' in previous papers [6,7]. large constant temperature zone. The hydrogen gas
Recently, it was recognized that most of the sur- concentration in the atmosphere was set in the
face properties of ¯oat glass correspond to the range of 0±20 vol%, and in the case of the hydrogen
characteristic tin penetration pro®les with tin en- penetration experiment, deuterium (2 D2 ) gas was
riched inner layers. For example, in the case of the used in place of hydrogen. The glass samples were
`bloom' phenomenon in which an optical haze may heated from room temperature to the holding value
be produced by a reheat treatment at about 600°C (600±900°C) at a constant heating rate of 10°C/
or above [8,9], some correlation was found be- min, and were then cooled down after being
tween the haze index after the reheat treatment maintained at the maximum temperature for 0±40
and the depth of the tin enriched inner layer. This min. This procedure makes it possible to control
phenomenon is interpreted as follows. The tin accurately the temperature, the time and the at-
enriched layer at the surface responsible for the mosphere. Soda-lime ¯oat glass (0.1 wt% iron) was
optical haze is formed during the reheat treatment used in the present experiment after mechanically
due to the migration of tin atoms, which are lo- removing at least 10 lm from each surface layer.
cated primarily at depths less than the tin enriched Another procedure was to ¯oat the glass sam-
inner layer [10]. ples in a bath of metallic tin at about 1000°C for
In this study, in order to make it possible to 10 min using the lab-scale ¯oat apparatus. The
control the characteristic tin penetration pro®le glass samples were then picked up from the molten
and the accompanying tin enriched inner layer, the tin bath after cooling down to 600°C. The lab-
di€usion mechanism of tin into glass was investi- scale ¯oat apparatus was kept in a glove-box fur-
gated using a lab-scale ¯oat apparatus with inde- nace in a reducing atmosphere of N2 =H2 (10 vol%
pendently varying heating capability. On the basis H2 ). In this procedure, although the accuracy with
of the analytical results of the dependence of the which the temperature and atmosphere could be
tin penetration pro®les on temperature, time, at- controlled was less than in the ®rst procedure, the
mosphere and glass composition, the di€usion thermal history during the tin penetration into
mechanism of tin into glass during the ¯oat pro- glass can be regarded as similar to that of the ac-
cess is discussed. tual ¯oat process. Because the redox reaction of
190 Y. Hayashi et al. / Journal of Non-Crystalline Solids 282 (2001) 188±196

iron may be strongly in¯uenced by the thermal

history [10], the in¯uence of the concentration and
oxidation state of iron in the glass on the tin
penetration pro®le was examined using this second
procedure. Except for the iron content, the series
of glass samples used here, which was prepared by
melting analytical reagent grade chemicals, had the
same bulk composition as commercial soda-lime
glass. In both procedures, tin metal of analytical
reagent grade was employed after removing tin
oxide by a pre-heat treatment at about 1000°C in a
N2 =H2 atmosphere. In the present experiment, no
essential di€erence in the tin penetration behavior
for the two above-mentioned procedures was de-
tected. Therefore, except for the detailed veri®ca-
tion experiments of the redox reactions, most of
the experiments were performed using the ®rst
procedure.

2.2. Analysis procedure

Fig. 1. SIMS depth pro®les of tin from the top (a) and bottom
faces (b) of commercial ¯oat glass with a thickness of 2 mm.
Secondary ion mass spectrometry (SIMS), with
a capability of selective isotope detection in addi-
tion to a detection limit of the order of several ppm
for almost all elements, was employed to measure thickness of 2 mm. It is con®rmed that the tin
the tin and deuterium concentration pro®les. penetration pro®le on the bottom face has a tin
Positive secondary ions of 120 Sn‡ were detected enriched inner layer located at a depth of ap-
using an 8 keV O‡ proximately 3 lm from the surface, as reported
2 primary ion beam with a beam
current of 100 nA, and negative secondary ions of previously [6,7]. In the top face, although the tin
2
D were detected using an 8 keV Cs‡ primary ion concentration is about two orders of magnitude
beam with a beam current of 100 nA. The incident lower than that of the bottom face, the tin enriched
angle of both primary ions was 60° from the nor- inner layer is also found at almost the same depth
mal to the sample surface and an area of as the bottom face. This suggests that the di€usion
100 lm  100 lm on the sample surface was mechanism of tin into glass on each face is similar.
sputtered. An electron beam of 1 keV was used to Tin penetration pro®les on the bottom face of
compensate for the positive charging of the sample commercial ¯oat glass with various thicknesses are
surface. The sputtering time to depth conversion shown in Fig. 2. It can be seen that the depth of the
was carried out using measurements of the crater tin enriched inner layer varies with thickness, and
depth made with a surface pro®lometer. that the deepest layer is for the glass near the
equilibrium thickness (6 mm). According to the
simulation of the thermal history, the residence
3. Results and discussion time of the glass samples with less than the equi-
librium thickness at the hot end of the ¯oat bath is
3.1. Tin penetration pro®les of commercial ¯oat almost the same. Consequently, the depth of the
glass tin enriched inner layers can be explained by the
stretch coecient for the attenuation of the glass
Fig. 1 shows the SIMS depth pro®les on the top during the ¯oat process. This is consistent with the
and bottom faces of commercial ¯oat glass with a proportional relationship between the thickness of
 Y. Hayashi et al. / Journal of Non-Crystalline Solids 282 (2001) 188±196 191

Fig. 3. SIMS depth pro®les of tin penetration into glass during

Fig. 2. SIMS depth pro®les of tin from the bottom face of
heat treatment at: (a) 600°C, (b) 700°C, (c) 800°C and (d) 900°C
commercial ¯oat glass with various thicknesses: (a) 3 mm, (b) 4 for 10 min in a N2 =H2 ˆ 90=10 atmosphere using the lab-scale
mm, (c) 6 mm, (d) 8 mm, (e) 10 mm, and (f) 15 mm. ¯oat apparatus.

the glass samples and the depth of the tin enriched mercial ¯oat glass. It is found that the tin enriched
inner layer. On the other hand, according to the inner layer is formed by heat treatment at tem-
simulation, in the case of glass samples with a peratures more than 800°C, which corresponds to
thickness more than the equilibrium thickness the a relatively low temperature region in the ¯oat
residence time at the hot end of the ¯oat bath is bath.
shorter than that of the glass with the equilibrium Fig. 4 gives the tin penetration pro®les of glass
thickness. As a result, the depth of the tin enriched samples with various heat treatment holding times
inner layer can be considered to have shifted in the at a temperature of 800°C. The amount of tin
direction of the surface. penetrating into the glass increases with the in-
creasing holding time, and, at the same time, the
3.2. Dependencies of the tin penetration pro®les on tin enriched inner layer shifts to greater depths.
the heating parameters Furthermore, the depth of the tin enriched inner
layer is proportional not to the square root of the
The in¯uence of temperature, time and atmo- holding time but to the ®rst-order of the holding
sphere on the di€usion of tin into glass are inves- time, as shown in Fig. 5. This fact indicates that
tigated using the lab-scale ¯oat apparatus. Fig. 3 the formation of the tin enriched inner layer is
shows the penetration behavior of tin into the glass associated with a chemical reaction.
in contact with tin during the heat treatment, i.e. The dependence of the tin penetration pro®les
the bottom face of the ¯oat glass. The penetration on the hydrogen concentration in the atmosphere
of tin into the glass increases with the increasing during the heat treatment is shown in Fig. 6. This
temperature. For temperatures more than 800°C, a proves that the tin penetration pro®les are also
tin enriched inner layer is clearly observed in the strongly in¯uenced by the hydrogen concentration
tin penetration pro®le, similar to that of com- in the atmosphere. That is, the tin enriched inner
192 Y. Hayashi et al. / Journal of Non-Crystalline Solids 282 (2001) 188±196

Fig. 4. SIMS depth pro®les of tin penetration into glass during

Fig. 6. SIMS depth pro®les of tin penetration into glass during
heat treatment at 800°C for: (a) 10 min, (b) 20 min, (c) 40 min
heat treatment at 800°C for 10 min in atmospheres of: (a)
and (d) 80 min in a N2 =H2 ˆ 90=10 atmosphere using the lab-
N2 =H2 ˆ 100=0, (b) N2 =H2 ˆ 98=2, (c) N2 =H2 ˆ 90=10 and (d)
scale ¯oat apparatus.
N2 =H2 ˆ 80=20 using the lab-scale ¯oat apparatus.

inner layer shifts to a deeper level with increasing

hydrogen concentration in the atmosphere.
The penetration of hydrogen into the glass from
the face in contact with the tin was examined by
using deuterium in place of hydrogen as a com-
ponent of the atmosphere. The use of deuterium
makes it possible to successfully distinguish the
hydrogen originally extant in the glass in a form
such as b-OH from that penetrating into the glass
during the heat treatment. The results shown in
Fig. 7 reveal that the deuterium concentration on
the face in contact with the tin was higher on the
glass sample heated in the deuterium-containing
atmosphere compared with that in a pure nitrogen
atmosphere. This means that hydrogen in the at-
Fig. 5. Relationship between the holding time during heat
treatment at 800°C in a N2 =H2 ˆ 90=10 atmosphere and the
mosphere penetrates into the glass from the face in
depth of the tin enriched inner layer. contact with tin, passing through the molten tin. It
was observed that the penetration depth of deu-
terium was greater than that of tin, probably due
layer is not formed in a pure nitrogen atmosphere, to the higher di€usion coecient of deuterium at
while it is formed by heating in a hydrogen-con- relatively low temperatures. These results show
taining atmosphere. Additionally, the tin enriched that the tin enriched inner layer is formed by the
 Y. Hayashi et al. / Journal of Non-Crystalline Solids 282 (2001) 188±196 193

Fig. 7. SIMS depth pro®les of deuterium penetration into glass Fig. 8. SIMS depth pro®les of tin penetration into glass with
during heat treatment at 800°C for 10 min in atmospheres of various Fe3‡ concentrations during heat treatment using the
N2 =D2 ˆ 100=0 (solid lines) and N2 =D2 ˆ 80=20 (dashed lines) lab-scale ¯oat apparatus: (a) total Fe ˆ less than 0.01 mol%, (b)
using the lab-scale ¯oat apparatus. Fe3‡ /total Fe ˆ 0.48/0.8 mol%, (c) Fe3‡ /total Fe ˆ 0.58/0.8
mol%.

reaction associated with hydrogen penetration into of the glass with di€erent concentration and oxi-
the glass. dation states of iron. In this experiment, tin pen-
On the other hand, it can also be seen from Fig. etration into each glass sample was achieved by a
6 that the total amount of tin penetrating into the heat treatment using a procedure with a thermal
glass decreases with the increase of hydrogen history similar to the actual ¯oat process. In Fig. 8
concentration in the atmosphere during the heat the tin enriched inner layer is clearly seen in the tin
treatment. This fact indicates that the ionization of penetration pro®les of the glass samples contain-
tin to the Sn2‡ state, which can be considered to be ing iron, while it is not formed in the iron-free
the ®rst step of tin di€usion into glass, is hindered glass sample. It was also found that the depth of
by the reducing power of hydrogen. the tin enriched inner layer varied with the Fe3‡
concentration, although the total iron concentra-
3.3. Dependencies of the tin penetration pro®les on tion was kept the same in this experiment. The
iron in the glass depth of the tin enriched inner layer was measured
for glass with various Fe3‡ concentrations, and
In the previous section, it was indicated that the this was plotted against the Fe3‡ concentration.
tin enriched inner layer was formed by the reaction The result shown in Fig. 9 proves that there is an
associated with hydrogen penetration into the inverse correlation between these factors, that is,
glass. One of the possible reactions with hydrogen the tin enriched inner layer shifts to shallower
is a redox reaction. The in¯uence of iron in the depths with increasing Fe3‡ concentration in the
glass, which has two types of oxidation state, Fe2‡ glass. Accordingly, the Fe3‡ concentration in the
and Fe3‡ , on the di€usion of tin into glass is in- glass can be regarded as one of the major factors
vestigated. Fig. 8 shows the tin penetration pro®les governing the formation of the tin enriched inner
194 Y. Hayashi et al. / Journal of Non-Crystalline Solids 282 (2001) 188±196

Fig. 9. Relationship between the Fe3‡ concentration in the

glass and the depth of the tin enriched inner layer.

layer. It is concluded that the reaction between the

hydrogen from the atmosphere and the Fe3‡ in the
glass is involved in the formation of the tin en-
riched inner layer.
Fig. 10. Schematic diagram of the di€usion mechanism of tin
3.4. Di€usion mechanism of tin into glass into glass during the ¯oat process.

On the basis of the analytical results described shown in Fig. 10. It is known, however, that the
above, the di€usion mechanism of tin into glass Sn2‡ , once it has penetrated into the glass, is easily
during the ¯oat process is discussed in this section. oxidized to the Sn4‡ state by the redox reaction
The schematic diagram in Fig. 10 summarizes tin between Sn2‡ and Fe3‡ [12,13]. In the case when all
di€usion into glass during the ¯oat process, de- of the Sn2‡ atoms have been converted to the Sn4‡
duced from this study. The initial stage of tin state, the penetration pro®le of the tin can be
di€usion into the glass can be initiated by the considered to exhibit a di€usion pro®le associated
ionization of metallic tin to the Sn2‡ state, which with the di€usion coecient of Sn4‡ . As is also
occurs in the tin bath through oxidizing species shown in Fig. 10, the concentration gradient of
such as oxygen. Although the concentration level Sn4‡ decays more rapidly with depth compared
of oxygen in the tin bath is kept at a minimum by with that of Sn2‡ , because the di€usion coecient
the use of a reducing atmosphere, a small amount of Sn4‡ is regarded to be smaller than that of Sn2‡
of residual oxygen, which is estimated by glow [14]. This is because the Sn4‡ acts as a network
discharge mass spectrometry analysis to be of the former, while the Sn2‡ acts as a network modi®er
order of several ppm, is assumed to be still present in the glass. On the other hand, hydrogen in the
in the tin bath. The Sn2‡ ions are able to penetrate atmosphere can penetrate into the glass from the
faster into the glass than Sn4‡ or Sn0 , mainly due bottom face, passing through the molten tin, as
to the ion exchange reaction with Na‡ ions. If the shown in Fig. 7, and therefore it can be assumed
tin which penetrates into the glass in the Sn2‡ state that the Fe3‡ in the surface layer of the bottom
remains as it is without being oxidized to the Sn4‡ face is reduced to the Fe2‡ state by the reaction
state, the tin penetration pro®le can be considered with hydrogen. Accordingly, the reduced layer in
to exhibit a di€usion pro®le associated with the the absence of Fe3‡ is formed at the bottom sur-
di€usion coecient of Sn2‡ [11], as schematically face of the glass. It is deemed that the penetration
 Y. Hayashi et al. / Journal of Non-Crystalline Solids 282 (2001) 188±196 195

pro®le of the tin in the reduced layer exhibits the cussed. The tin enriched inner layer can be seen to
theoretical Sn2‡ di€usion, because of the absence be formed by the heat treatment of glass in contact
of Sn4‡ . On the other hand, in the region deeper with tin using lab-scale ¯oat apparatus. It has been
than the reduced layer, most of the Sn2‡ atoms are found that the depth of the tin enriched inner layer
oxidized to the Sn4‡ state by the reaction with the is proportional to the holding time at the maxi-
Fe3‡ that exists in this region, so that the tin shows mum temperature during the heat treatment, and
a penetration pro®le that is mainly based on the is inversely proportional to the Fe3‡ concentration
theoretical Sn4‡ di€usion. The total tin penetration in the glass. It has also been proved that the tin
in the region beyond the reduced layer can be re- enriched inner layer is formed by hydrogen pene-
garded to be the same in quantity for the di€usion tration into the glass from the atmosphere, passing
pro®les associated with both Sn2‡ and Sn4‡ . through the molten tin. These results mean that
Consequently, the tin enriched inner layer due to the reaction between hydrogen and Fe3‡ is asso-
tin pile up is formed near the interface of the re- ciated with the formation of the tin enriched inner
duced layer and the deeper region. layer. Furthermore, it has been proposed that the
The di€usion mechanism described above is formation of the tin enriched inner layer is at-
consistent with the fact that the tin located on the tributed to the two redox reactions and the di€u-
shallower side of the tin enriched inner layer exists sion behavior of both Sn2‡ and Sn4‡ . That is, one
mainly in the Sn2‡ state, while the tin located on of these two redox reactions is the reduction of
the deeper side exists in the Sn4‡ state, as veri®ed Fe3‡ to Fe2‡ due to hydrogen resulting in the
by M ossbauer experiments [15]. It is worth noting formation of a reduced surface layer. Another is
that the tin enriched inner layer in the top face can the oxidation of Sn2‡ to Sn4‡ due to Fe3‡ in the
also be interpreted using the same tin di€usion glass. Consequently, it is deemed that the depth of
mechanisms. The factors governing the depth of the tin enriched inner layer of ¯oat glass is gov-
the tin enriched inner layer, the temperature his- erned by the Fe3‡ concentration in the glass, the
tory, the Fe3‡ concentration and the amount of amount of hydrogen penetration into the glass
hydrogen penetration into the glass were found to from the atmosphere and the thermal history of
be crucial, because the reaction between Fe3‡ and the ¯oat process. Besides, it is also in¯uenced by
hydrogen is associated with the formation of the the stretch coecient for the attenuation of the
tin enriched inner layer. In order to control the tin glass during the ¯oat process. The clari®cation of
concentration pro®le in the glass, it is important to this di€usion mechanism is expected to play an
control the amount not only of Fe3‡ but also those important role in controlling the penetration of tin
oxidizing species for Sn2‡ oxidation such as SO3 , into glass during the ¯oat process, so that it may
which acts analogously to Fe3‡ . Besides, in the lead to a successful improvement in the various
case of commercial ¯oat glass, the depth of the tin surface properties of ¯oat glass.
enriched inner layer is also in¯uenced by the
stretch coecient for the attenuation of the glass
during the ¯oat process. The clari®cation of the References
di€usion mechanism as well as the optimum se-
lection of production conditions such as the [1] L.A.B. Pilkington, Proc. R. Soc. London A 314 (1969) 1.
stretch coecient are essential for the successful [2] H. Franz, J. Non-Cryst. Solids 42 (1980) 529.
[3] Y. Hayashi, K. Matsumoto, M. Kudo, J. Ceram. Soc. Jpn.
control of tin penetration a€ecting the industrially 108 (2000) 1007.
important surface properties of ¯oat glass. [4] Y. Hayashi, T. Yoneda, K. Matsumoto, J. Ceram. Soc.
Jpn. 102 (1994) 206.
[5] H. Franz, Ceram. Eng. Sci. Proc. 16 (2) (1995) 221.
[6] J.S. Sieger, J. Non-Cryst. Solids 19 (1975) 213.
4. Conclusion
[7] M. Verita, F.G. Bianchini, S. Hreglich, C.G. Pantano, V.
Bojan, Bol. Soc. Esp. Ceram. Vid. 31C (6) (1992) 415.
In this paper, the di€usion mechanism of tin [8] K.F.E. Williams, M.F. Thomas, J. Greengrass, J.M.
into glass during the ¯oat process has been dis- Bradshaw, Glass Technol. 40 (1999) 103.
196 Y. Hayashi et al. / Journal of Non-Crystalline Solids 282 (2001) 188±196

[9] K.F.E. Williams, C.E. Johnson, J. Greengrass, B.P. Tilley, [13] K.F.E. Williams, M.F. Thomas, C.E. Johnson, B.P. Tilley,
D. Gelder, J.A. Johnson, J. Non-Cryst. Solids 211 (1997) J. Greengrass, J.A. Johnson, Fundamentals Glass Sci.
164. Technol. 1997, 127.
[10] Y. Hayashi, R. Akiyama, M. Kudo, Surf. Interface Anal. [14] Y. Hayashi, K. Matsumoto, M. Kudo, Glass Technol. 42
31 (2) (2001), to be published. (5) (2001), to be published.
[11] T.-j. Wang, Glass Technol. 38 (1997) 104. [15] W. Meisel, Glastch. Ber. Glass Sci. Technol. 72 (1999)
[12] A. Kumar, S.P. Singh, R. Pyare, Glastech. Ber. 64 (1991) 106. 291.