Você está na página 1de 209

THE STRUCTURE OF GLASS

PHASE-SEPARATION PHENOMENA IN GLASSES

STEKLOOBRAZNOE SOSTOY ANIE


LIKV ATSIONNYE YAVLENIY A V STEKLAKH

CTEKJI006PA3HOE COCTOSlHHE
JU1KBAlJ.HOHHbIE SlBJIEHHSl B CTEKJIAX
THE STRUCTURE OF GLASS
Volume 8

PHASE-SEPARATION
PHENOMENA IN GLASSES

PROCEEDINGS OF THE FIRST ALL-UNION SYMPOSIUM


ON PHASE-SEPARATION PHENOMENA IN GLASSES,
LENINGRAD, APRIL 16-18, 1968

Edited by E. A. Porai-Koshits
I. V. Grebenshchikov Institute of Silicate Chemistry
Academy of Sciences of the USSR, Leningrad

Translated from Russian by E. Boris Uvarov


Senior Scientific Translator

CONSULTANTS BUREAU NEW YORK-LONDON 1973


Proceedings of the First All-Union Symposium on Phase-Separation Phe-
nomena in Glasses, held in Leningrad, April 16-18, 1968, and published by
Nauka Press in Leningrad in 1969 for the State Committee for Science and
Technology of the Council of Ministers of the USSR, the Academy of
Sciences of the USSR, and the D.!. Mendeleev All-Union Chemical Society.
Editorial Board:
E. A. Porai-Koshits
V. N. Filipovich O. V. Mazurin
F. Ya. Galakhov S. V. Nemilov
R. Ya. Khodakovskaya B. G. Varshal
The original Russian text has been corrected and revised by the editor and
the present translation is published under an agreement with Mezhdunarod-
naya Kniga, the Soviet book export agency.

Library of Congress Catalog Card Number 58-44503


ISBN 978-1-4757-0159-3 ISBN 978-1-4757-0157-9 (eBook)
DOI 10.1007/978-1-4757-0157-9
1973 Consultants Bureau, New York
Softcover reprint of the hardcover 1st edition 1973
A Division of Plenum Publishing Corporation
227 West 17th Street, New York, N.Y. 10011
United Kingdom edition published by Consultants Bureau, London
A Division of Plenum Publishing Company, Ltd.
Davis House (4th Floor), 8 Scrubs Lane, Harlesden, London, NW10 6SE, England
All rights reserved
No part of this publication may be reproduced in any form
without written permission from the publisher
PREFACE

The First All- Union Symposium on Phase- Separation Phenomena in Glasses was held in Lenin-
grad from April 16 to 18, 1968. The symposium was organized by the Section of the structure
of glass and the nature of the glassy state of the Scientific Council for the problem "New Tech-
nical Materials and Coatings Based on Glass and Refractory Compounds" of the State Committee
for Science and Technology of the Council of Ministers of the USSR, by the Scientific Council for
the problem "Physicochemical Principles of Production of New Heat-Resistant Inorganic Mate-
rials" of the Division of Physical Chemistry and Technology of Inorganic Materials, Academy of
Sciences of the USSR, by the Leningrad regional board of the D. 1. Mendeleev All-Union Chemi-
cal Society, and by the 1. V. Grebenshchikov Institute of Silicate Chemistry, Academy of Sciences
of the USSR.
The present collection consists of the Proceedings of the Symposium and is No.3 in Vol.
5 of the series "The Glassy State," published by the Academy of Sciences of the USSR since 1955.
The book contains the results of theoretical and experimental investigations of phase sep-
aration of glasses into two glassy phases, and of the influence of this submicroscopically heter-
ogeneous structure on certain properties of glass.
Metastable phase separation in glasses was discovered about 10 years ago. During recent
years a considerable number of publications on the subject have appeared both in the Soviet Union
and abroad. The interest in this problem is due to the extensive occurrence of phase-separation
phenomena and to their significance in production of certain practical materials: glass-ceramics,
electrical vacuum tube and photochromic glasses, molecular sieves, etc. The book is therefore
of interest not only to scientists but also to engineers and technologists concerned with the pro-
duction of these materials.

v
CONTENTS
OPENING OF THE SYMPOSIDM
Introductory Address
F. Ya. Galakhov . . . . . . . . . . . . 0 II II II II 3
I. GENERAL ASPECTS OF PHASE SEPARATION IN GLASSES
Causes of Phase Separation in Simple Silicate Systems
F. Ya. Galakhov and B. Go Varshal . . . . . . . . . . . . . . . . 0 II 7
Theory of Phase Separation and Atomic-Ionic structure of Certain
Two-Component Systems
V. N. Filipovich and D. D. Dmitriev 12
Some Thermodynamic Aspects of Phase-Separation Phenomena
M. M. Shu! 'ts .... 0 0 0 0 II 0 II II II 23
Primary and Secondary Phase Separation in Glasses
E. A. Porai-Koshits and V. I. Aver'yanov. . . 28
Investigation of the Physical and Chemical Properties of Glasses
Undergoing Phase Separation as a Method of studying Their structure
o. V Mazurin . . .
0 II 0 0 D 0 II II 0 37
Influence of structural Heterogeneity on Physical and Chemical Properties
of Glasses
N. M. Bobkova and I. A. Trunets ... 0 0 0 II 0 43
Determination of the Nucleation Rate of New Phases in Glasses at the
Initial Stages of Phase Conversion
V. N. Filipovich and A. M. Kalinina 0 0 0 0 0 48
Surface-Active Properties of Catalysts and Phase Separation in Glasses
Ya. A. Fedorovskfi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Discussion on Section I ............. 57
II. PHASE SEPARATION IN ALKALI SILICATE AND BOROSILICATE GLASSES
Influence of the Oxides of Lithium, Potassium, and Cesium on
Phase Separation in Sodium Silicate Glasses
N. S. Andreev and G. G. Boiko ... 0 II II II II 61
Relation between the Microheterogeneity of Lithium Silicate Glasses
and the Rate of Their Dissolution in Hydrofluoric Acid
V. A. Borgman, V. K. Leko, and V. K. Markaryan . 66
study of Phase-Separation Processes in the Sodium Silicate System
by the Electromotive Force Method
A. F. Borisov and Vo I. Solov'ev. 0 II II 0 0 0 II 0 II 0 II II II 69
Influence of Heat Treatment on Viscosity of Sodium Silicate and
Sodium Borosilicate Glasses
V. P. Klyuev, Go Po Roskova, and V. I. Aver'yanov. . . 74

vii
viii CONTENTS

Regions of Metastable Phase Separation in Ternary Alkali Borosilicate Systems


F. Ya. Galakhov and 0. S. Alekseeva. 80
Composition of Coexisting Phases in Sodium Borosilicate Glasses
Exhibiting Phase Separation
B. G. Varshal . . . . . . II 0 II 0 II 0 84
Influence of Composition on Viscosity and Transformation Temperatures
of Borosilicate Glasses Exhibiting Phase Separation
M. V. Strel'tsina, 0. V. Mazurin, and A. S. Totesh. 87
Porous Glasses as a Source of Information on structural Changes in
Sodium Borosilicate Glasses Undergoing Phase Separation
S. P. Zhdanov and E. V. Koromal'di . . . 92
Investigation of Phase Separation in Glasses of the System R 20-B20 3-Si02 by
Study of Their Electrical Properties
T. P. Dgebuadze ......... 0 99
Influence of Replacement of Silica by Various Glass-Forming Oxides
on the Tendency to Phase Separation in the System
R 20-B20 3-Si0 2
D. F. Ushakov and Yu. S. Krupkin. 103
Phase Separation in Low-Alkali Borosilicate Glasses Containing
RO and Al 20 3
L. A. Bal'skaya, L. A. Grechanik, and N. M. Vaisfel'd 107
Properties of Glasses Exhibiting Phase Separation in the System
Si0 2- B20 3-AI 20 3- ZnO- Na 20
L. A. Grechanik and L. A. Bal' skaya. 114
Influence of Lead Oxide on the structure and Properties of Borosilicate Glass
N. M. Pavlushkin and Zh. A. Olobikyan . 118
Influence of the Composition and Ratio of the Separated Phases on
Stresses Arising During Leaching of Two-Phase Sodium Borosilicate Glasses
0. K. Botvinkin, M. L. Mironova, and G. L. Shpilevskaya . 122
structural Physical state of Microheterogeneous Glasses
S. I. Sil'vestrovich and V. D. Kazakov 126
Use of Differential Thermal Analysis and Dilatometric Measurements
for study of Phase-Separation Phenomena in Glasses
M. B. Usvitskii .. . . . . 0 0 0 0 0 0 0 0 0 0 131
Discussion on Section II . . . . . . . . . . . . . . . . .. . . .. . . . . . . . . . . . .. . . . .. . . . . . .. 135
III. PHASE SEPARATION IN OTHER MULTICOMPONENT GLASSES
EPR Study of structural Changes During Microphase Separation in
Titanium-Containing Glasses
N. M Pavlushkin, R. Ya. Khodakovskaya, L. A. Orlova, and
v. V. Orlov. .. . . .. . .. . . . . . . .. .... . . . . .. .. . . . . . . . .. . . . . .. . . . .. . .. .. . . . . .. 139
Investigation of Phase-Separation Processes in Titanium-Containing
Glasses in the System Si02-AI 20 3-CaO-MgO
B. G. Varshal, N. M. Vaisfel'd, G. B. Knyazher, and L. M. Yusim 145
Influence of Titanium Dioxide on Phase Separation in Lithium
Aluminosilicate Glasses
N. E. Kind and E. M. Milyukov ........................................ 0 .. ... 154
Phase-Separation Phenomena in Glasses of Aluminosilicate Systems
Containing Various Modifier Cations
E. M. Milyukov and N. E. Kind. . .. . .. .. .. .. . . . . . . . . .. .. .. .. .. .. .. .. .. .. .. .. .. . . . ... 158
CONTENTS ix

Microphase Separation in Glass Containing Certain Crystallization


Initiators
S. T. Suleimenov, M. Sh. Sharafiev, T. A. Abduvaliev,
To D. Nurbekov, and I. I. Sorokina .. 0 0 Cl " 162
Influence of Phase Separation on Physical and Chemical Properties
of Heat-Treated Fibers
M. S. Aslanova and Zo I. Shaina 0 ...... 0 " II III 0 168
Photochromism and Microphase Separation of Glass
V. A. Tsekhomskii and I. V. Tunimanova .. 172
Droplet Separation in Photosensitive Glass
A. I. Berezhnoi . . . iii 11\ CI 0 " 0 III ...... 0 0 0 176
Spectroscopic Investigation of Glasses in the System R20-La203-Si02
L. V. Labutina, N. M. Pavlushkin, and M. V. Artamonova 179
Use of the Spectral Luminescence Method for Studying Phase Separation
in Silicate Glasses
M. I. Kuz'menkov and M. B. Rzhevskii. . . 183
Influence of Microphase Separation in Fluoberyllate Glasses on the Concentration
Dependence of Certain Properties
G. T. Petrovskii, I. S. Gilev, and S. V. Nemilov. . . 187
Phase-Separation Phenomena in Glasses of the System La203-B203-P205
Z. M. Syritskaya and E. S. Kutukova . 191
Differentiation of Structural Groups in Slag Glasses
N. M. Pavlushkin and B. I. Beletskii . 196
Phase- Separation Phenomena During Thermal Aging of Multicomponent
Thermometer Glasses and Their Influence on Physical Properties
Yu. V. R.ogozhin and E. N. Kaplina. . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
Influence of Phase Separation on Certain Properties of Glasses in the System
Si0 2-Al 20 3- B20 3- Bi203-Ce02- Zr02
V. B. Brailovskii, R. I. Grichevskaya, A. I. Trushkov, and
Yu. F. Tyurnikov . . . . . . . CI CI 0 0 " 205
General Discussion . . . . . . . . . . . . ~ . . . . . . . . . . . . . . . . . . . . . . 208
OPENING OF THE SYMPOSIUM
INTRODUCTORY ADDRESS

F. Ya. Galakhov

Today we are participants and witnesses of a significant event in the development of sci-
entific research into the glassy state. This event is the present symposium, the first to be de-
voted to study of phase-separation phenomena in glasses. This is all the more noteworthy be-
cause no conferences have been held in the past speCially on the subject of phase separation not
only in glasses but in any other materials.
Investigations of phase separation in melts, initiated by J. W. Greig in the 1920s, are still
continuing, but these phenomena have attracted particularly close attention in relation to studies
of glass-ceramics and the structure of glass.
We can already speak of the history of development of research into metastable phase sep-
aration, which started with the demonstration of heterogeneous structure in certain glasses.
In later publications attention was drawn to structure resulting from phase separation in glasses.
This became possible when the electron microscope was adopted as a research tool. One could
cite a series of papers in which the role of such structure of glasses was stressed in connection
with studies of the formation of glass-ceramics. However, during this initial period phase- sep-
aration phenomena did not receive due attention and everything was reduced to establishment of
facts: either detection and study of the heterogeneous structure, or demonstration of structure
resulting from liquid-liquid phase separation in individual glasses.
Application of the theory of liquid-phase separation to explanation of observed experimen-
tal data on the heterogeneous structure of glasses came later.
The great importance of phase theory in studies of phase-separation phenomena was first
noted in the following publications. The very first was the paper by Roy [1], whose aim was
"to draw attention to forming met as tab I y in typical glass-forming systems two immiscible
liquid phases." This was followed by a paper by Hinz and Kunth [2], who examined for-
mation of glass-ceramics and opal glasses from the standpoint of phase diagrams showing re-
gions of phase separation to various degrees.
Finally, the potentialities offered to the investigator by the use of phase diagrams in
studies of glasses exhibiting liquid-liquid phase separation were demonstrated with reference
to sodium borosilicate glasses [3].
These papers not only presented earlier findings in a new light, but evidently influenced
subsequent investigations when data of the theory of heterogeneous equilibria were used in
studies of the heterogeneous structure of glass, and the phase-separation origin of heterogeneity
regions was accepted as the starting point. This applies in the first instance to studies of alkali

3
4 F. Ya. GALAKHOV

silicate glasses [4, 5], in which binodal curves for metastable liquid-liquid phase separation
were determined on the equilibrium diagrams for the first time [4, 5].
It is appropriate to emphasize in this connection that phase diagrams with established re-
gions of metastable liquid-liquid phase separation are of great practical and theoretical signif-
icance. Construction of such diagrams is a task for the immediate future.
A most important result of application of the theory of liquid-liquid phase separation to
study of the structure of glasses is that it was possible to establish the nature of the heteroge-
neities found in glasses; in the great majority of cases they owed their origin to phase separa-
tion. Ignorance of this fact in the recent past led to diffuse and sometimes even to erroneous
interpretations and to extensive controversy.
The theoretical, predominantly thermodynamic, studies of Cahn [6] and Filipovich
[7] on the spinodal mechanism of phase decomposition are of great significance in rela-
tion to the kinetics and mechanism of metastable phase separation, which determine the struc-
ture of such glasses. Regrettably, there are as yet few experimental data confirming the exis-
tence of two mechanisms of phase separation - nucleation and spinodal - and such information
as is available is not clear enough. Further studies, both theoretical and experimental, are
needed in this direction.
Phase-separation phenomena in silicate melts and glasses are much more common than
was believed until recently. They are the cause of heterogeneous structure in many glasses,
affect their properties, and in a number of cases direct the crystallization process toward for-
mation of microcrystalline products. However, such effects occur also in melts of chalcogen-
ides and of organic and other compounds which form glassy products. It is therefore to be hoped
that the work of the first symposium on phase-separation phenomena will not only assist in
further development of research and in elucidation of a number of problems in the field of sili-
cate glasses and glass-ceramics, but will also be of interest to a wide circle of specialists.
Literature Cited
1. R. Roy, J. Am. Ceram. Soc., 43:670 (1960).
2. W. Hinz and P. -0. Kunth, Glastechn. Ber., 34(9):431 (1961).
3. F. Ya. Galakhov, Izv. Akad. Nauk SSSR, Otd. Khim. Nauk, No.5, p. 743 (1962).
4. J. J. Hammel, Seventh International Congress on Glass, Brussels, No. 36 (1965).
5. N. S. Andreev, V. I. Aver'yanov, and E. A. Porai-Koshits, in: Structural Transformations
in Glasses at High Temperatures, Izd. Nauka, Moscow- Leningrad (1965), p. 59.
6. J. W. Cahn, J. Chem. Phys., 42:93 (1965).
7. V. N. Filipovich, Neorgan. Materialy, 3:1192 (1967).
I

GENERAL ASPECTS OF PHASE SEPARATION IN GLASSES


CAUSES OF PHASE SEPARATION IN
SIMPLE SILICA TE SYSTEMS

F. Ya. Galakhov and B. G. Varshal

Regions of phase separation, both stable and metastable, have now been detected in most
simple silicate systems. Various aspects of phase separation have been discussed in detail in
the literature, mainly from the thermodynamic standpoint, and many relationships in these phe-
nomena have been explained. Application of the theory of ideal and regular solutions proved ade-
quate for qualitative and in some instances also for quantitative description of phase-separation
phenomena. However, thermodynamic treatment is concerned mainly with the state of the sys-
tem and says very little about the nature of the phenomenon itself [1]. In this communication we
attempt to examine, in general form, processes of phase separation in binary silicate systems,
using known experimental data and analyzing various views of their nature.
The first study of the nature of phase separation in silicate systems, based predominantly
on the principles of crystal chemistry, was undertaken by Warren and Pincus [2]. In their
opinion, immiscibility is due to formation by the modifier cations (subsequently referred to as
cations) of independent cation- oxygen groups, differing from silicon- oxygen groups. Two dif-
ferent types arise, with formation of two liquids, only when the cation- oxygen bond energy is
sufficiently high. This bond strength, expressed in various ways [Z/r; Z/r2; Z/(r a + rc)2; Z/n, J
where Z is the valence of the cation, r is the ionic radius, and n is the coordination number of
the cation], may serve as a measure of the ability of a cation to cause immiscibility in the melt.
In a series of investigations by Levin and Block [3, 4] phase separation in oxide sys-
tems was studied and discussed in detail from various aspects. We will consider two problems
in their publications. The first refers to the possibility of calculating the composition of the
liquid phase having a low silica content. The good agreement with experimental data in most
cases appears to confirm the validity of the theoretical concepts on which the calculation is based.
However, the composition of the liquid of the lower silica content can be found in this way only
at the liquidus temperature; but as this temperature is not singular in any way either with regard
to development of the process or in its position on the binodal, the calculation is arbitrary and is
not applicable, for example, to metastable phased separation.
The second point refers to interpretation of phase separation, i.e., to its causes. In accor-
dance with views put forward earlier, Levin and Block consider the primary cause of phase sep-
aration to be the tendency of glass-former and modifier cations to form cation- oxygen coordi-
nation polyhedra. If the polyhedra formed by the modifier and glass former are similar in shape
and size, immiscibility does not occur. In absence of such Similarity between the polyhedra,
either immiscibility arises or the system exists in an unstable state, showing an S-shaped liq-
uidus curve, dependent on the relative bond strengths of the competing cations. Although this

7
8 F. Ya. GALAKHOV AND B. G. VARSHAL

interpretation supplements the idea of Warren and Pincus somewhat, it still gives an in-
complete picture of the phenomenon. For example, with regard to the glass former, such as the
silicon ion, only the structure of individual oxygen polyhedra is taken into account, and their
polymerization, i.e., the structure of the entire silicon- oxygen framework is disregarded, as is
its ability to adapt itself to the cation- oxygen polyhedron, despite its geometrical distinction
from the latter, at least in the form of the Si20 7 diortho group, as was demonstrated by Belov
with reference to crystalline silicates [5].
Esin [6], emphasizing that a considerable Si02 content in the melt is necessary for
phase separation to occur, attributes this requirement to two conditions. First, crosslinking of
silicon- oxygen tetrahedra must be sufficient to facilitate considerably conversion of cybotactic
groups into analogous regions of almost pure silica. Second, with decrease of the number of
oxygen atoms per silicon atom detachment of oxygen anions from SixO complexes becomes pro-
gressively more difficult and proceeds without separation in the liquii phase only as long as
linking of the tetrahedron chains into double chains and sheets does not go too far, Le., as the
silicon- oxygen radicals are present mainly as chain fragments and as rings.
According to De Wys [7], a possible cause of phase separation is the tendency of certain
complexes (from Si0 4 to Si4011 chains) to undergo association, leading to formation of induced
dipoles, which causes positive deviations from Raoult's law, i.e., a tendency to phase separation.
This explanation, based on the general theory of solutions, presupposes that the melt already con-
tains such groups, in addition to groups of higher complexity. Therefore it refers only to the
final and last stage of phase separation, which must be preceded by a preparatory process of
structural changes in the melt.
Weyl [8] considers that phase separation is in the first instance the result of geomet-
rical conditions; i.e., it is due to the fact that the highest concentration of XO s groups is pos-
sible in the Si02 tetrahedral structure, whereas the coordination number of most cations is great-
er than four. In all simple and multicomponent glass compositions Si 4+ is screened by four 0 2-
ions, which means that in silica glass each 0 2- ion is polarized by two Si 4+ cations. Consequent-
ly, oxygen ions of the silicon-oxygen network have low polarizability and are ineffective in
screening additional cations. A statistical ion distribution cannot persist in a liquid with a high
silica content, and the liquid separates into a high-silica phase and a phase with a higher con-
centration of the second oxide. Numerous effectively polarized 0 2- ions appear at a high con-
centration of the second oxide, and a homogeneous and stable glass can therefore be obtained.
Polarization of anions and cations plays the dominant part in formation of a homogeneous melt
or glass. The concentration of the second oxide necessary for formation of a homogeneous melt
decreases with increase of their polarization.
According to Markhasev and Sedletskii [9] a correspondence exists in a homoge-
neous melt between the coordination required by the cation and the coordination assigned to
it by the silicon- oxygen structure. With increasing silica content in the melt polymerization of
silicon- oxygen tetrahedra is intensified, and this is accompanied by increase of the space and
coordination provided to the cation by the silicon-oxygen structure. When this exceeds the
degree of coordination required by the cation, the system becomes unstable and phase separa-
tion occurs; part of the silica goes to form a silicon- oxygen structure in the amount necessary
to satisfy the coordination requirements of the cation, while excess silica forms a highly poly-
merized structure close to the structure of a melt of almost pure silica.
This brief survey of the causes of immiscibility in silicates, with the phenomenon exam-
ined from different standpoints, shows that nevertheless there are no essential differences with
regard to the nature and mechanism of liquid-liquid phase separation. Moreover, the existing
views supplement each other.
F. Ya. GALAKHOV AND B. G. VARSHAL 9

Attempts to unify them into a common scheme must be based on the most general and
universal data. The main primary cause of immiscibility must be electrostatic interaction be-
tween ions in the melt or glass, causing unlike cations to surround themselves with oxygen, form-
ing independent polyhedra. However, the suitable conditions must exist if this tendency is to
lead to phase separation. These conditions are, first, a definite degree of cation-oxygen inter-
action forces and, second, structural incompatibility of the resultant different cation- oxygen
groups, the structural characteristics of which in their turn depend on the degree of polariza-
tion of the anions, i.e., oxygen. Various combination or superpositions of these two factors
(force and structure factors) lead to various degrees of phase separation or tendency to separa-
tion. In general, the tendency to phase separation is determined by the extent of the immisci-
bility region in the phase diagram, regardless of whether it is stable or metastable. This ap-
proach to assessment of the tendency to phase separation would make it possible to find the
overall effect or the joint influence of the above-named factors producing the effect.
However, it is not possible at present to establish the relation between the extent of the
immiscibility region and the overall effect, as there is no criterion for assessment of the latter.
Therefore only the energy parameters of the cations are used. The expression Z/r2 was used
by Dietzel [10] as a measure of the tendency of cations to form two liquids. Esin [6]
noted that the tendency to phase separation depends on the radius of the cation. Glasser,
Warshaw, and Roy [11] correlated the extent of liquid immiscibility with the cationic poten-
tial Z/r. As they expressed compositions in molar percentages, which cannot be compared
for different systems, it was necessary to construct three different plots for RO, R 20 3, and
R02 Oxides causing metastable phase separation (BaO, Li 20, Na20) cannot be included in this
classification, as the comparison referred to the liquidus temperature. Moreover, they were
unable to explain experimental results for Cr203 and Zr02 and therefore had to plot hypothetical
diagrams, differing substantially from existing ones [12, 13].
We compared the extent of the immiscibility regions at temperature t o5, corresponding to
0.5 t cr (in degrees Celsius, found by calculation from the absolute scale), and not at the liquidus
temperature, so that it was possible to include data on metastable phase separation. For deter-
mination of the extent of the immiscibility region at this temperature the equations of the phase-
separation binodal were calculated for 15 systems with known immiscibility curves, and extrap-
olated to t o5'
The results of calculations for alkaline-earth cations are shown in Fig. 1. It is seen that
the binodal curves are extrapolated considerably below the liquidus curve and terminate at t o5

1800
fI--"-4'---T-~
1600 Fig. 1. Binodal immisci-
bility curves for Si02 - RO
systems. 1) MgO; 2) CaO;
1200
3) SrO; 4) BaO; a) liquidus
1000 line for systems 1-3. The
800
dashed lines represent the
extent of the immiscibility
region (6 0 5) at t o5
MeO, cat.%
10 F. Ya. GALAKHOV AND B. G. VARSHAL

ilO.5
100
90
80
70
60
Fig. 2. Dependence of the extent of
the immiscibility region (AO.5) at t o5 50
on the cationic field strength in Z/i'- 40
SiCa-RnO m
30
Z
200 1 Z 3 4 5 6 7 8 rZ

TABLE 1. Immiscibility Regions in SiCa-RnOm

Oxide RnOm ;C' A z


T~
tcr tos 6 05,
mole%
i cat.0/0
60.5' Literature
source

Li 20 0.68 2.16 1000 365 35 52 [17]


Na 20 . 0.98 1.04 850 290 24 39 [17]
MgO 0.74 3.65 2200 965 57.5 57.5 [18]
CaO 1.04 1.85 2110 920 44.5 44.5 [18]
SrO 1.2 1.39 1920 825 39.5 39.5 [18]
BaO 1.38 1.05 1600 665 30.0 30.0 [19]
Sc2OS ' 0.83 4.35 2060 895 46.0 63.0 [20]
Y 20 S 0.97 3.20 2200 965 35.2 52.0 [21]
La 2OS ' 1.04 2.77 2050 890 39.0 56.0 [22]
Nd 20 3 0.99 3.06 2000 865 45.0 62.0 [231
Sm 203 0.97 3.20 2200 965 38.6 56.0 [24]
Gd 20 3 0.94 3.40 2300 1015 38.0 55.0 [25]
DY203 0.88 3.88 2320 1025 42.5 .19.5 [26]
Er2 0 a 0.S5 4.16 2280 1005 47.0 64.0 [26)
Yb 20 3 0.81 4.58 2200 965 49.0 65.8 [241
Zr0 2 . 0.82 6.0 2430 1080 - - [12]
Ti0 2 . 0.68 8.6 - - - - [13)
Cr2OS ' 0.64 7.5 - - - - [27 )

According to a paper by Seward et al. [28J, published after this symposium, for the
system BaO -Si0 2 tcr = 1460. Calculation gives 6 0 5 = 360/0. Then the new equation
6 05 = 28.5 + 8.5 (Z/r2) describes their relationship more satisfactorily, in particular
for Ti0 2

The general data are presented in Table 1 and Fig. 2. The extent of the immiscibility re-
gion is expressed in cationic percent* (to make it possible to compare different types of oxides:
R 20, RO, R 20 a, and R02). Values of the cationic radii were taken from Zhdanov's book [41.
Calculation by the method of least squares gave the following expression for the extent of
the immiscibility region at t o5 (designated AO.5):

110 5 =25.8 +9.1(~).


This relationship is somewhat approximate, and will be corrected as more experimental data be-
come available. Points for zirconium, chromium, and titanium are indicated on the straight line
in Fig. 2.

*The content of the modifier cation as a percentage of the sum of silicon and modifier cations.
F. Ya. GALAKHOV AND B. G. VARSHAL 11

It follows from this diagram that the extent of immiscibility should be ",80% for Zr02,
",95% for Cr203, and nearly 100% for Ti0 2, which is in fairly good agreement with extrapolation
of experimental data.
These results lead to the conclusion that the cationic field strength Z/r2 is the main factor
determining the extent of the immiscibility region in simple silicate systems. It may be supposed
that the above relationship represents the upper limit of the extent of the immiscibility region,
corresponding to complete incompatibility of the structural groups. A striking example of the
influence of structural compatibility on the degree of phase separation is provided by the system
Si0 2-A1 20 3, in which, according to Fig. 2, the extent of immiscibility should be about 100%. How-
ever, immiscibility was detected experimentally [15, 16] over a much narrower range; this can
be attributed to presence of some of the aluminum ions in fourfold coordination and to structural
similarity of [Si0 4] and [A10 4] groups.
Literature Cited
1. I. N. Belyaev, Usp. Khim., 29:899 (1960).
2. B. E. Warren and A. C. Pincus, J. Am. Ceram. Soc., 23:301 (1940).
3. E. M. Levin and S. Block, J. Am. Ceram. Soc., 40:95, 113 (1957); 41:49 (1958).
4. E. M. Levin, J. Am. Ceram. Soc., 50:29 (1967).
5. N. V. Belov, Kristallografiya, 5:15 (1960).
6. 0. A. Esin, Proceedings of the Second All-Union Conference on Theoretical and Applied
Electrochemistry, Izd. Akad. Nauk Ukr. SSR, Kiev (1949), p. 215.
7. E. C. De Wys, Mineral. Mag., 32(249):471 (1960).
8. w. A. Weyl and E. C. Marboe, Glass Ind., 41:429, 487, 549, 620, 687 (1960); 42:23, 76, 123,
194 (1961).
9. B. 1. Markhasev and 1. D. Sedletskii, Dokl. Akad. Nauk SSSR, 148:916 (1963); 154:1225 (1964).
10. A. Dietzel, Z. Elektrochem., 48(19):9 (1942).
11. F. P. Glasser, 1. Warshaw, and K. Roy, Phys. Chern. Glasses, 1:39 (1960).
12. N. A. Toropov and F. Ya. Galakhov, Izv. Akad. Nauk SSSR, Otd. Khim. Nauk, No.2, p. 158
(1956).
13. M. L. Keith, J. Am. Ceram. Soc., 37:491 (1954).
14. G. S. Zhdanov, PhYSics of the Solid State, Izd. Mosk. Gos. Univ., Moscow (1962), p. 184.
15. F. Ya. Galakhov and S. F. Konovalova, Izv. Akad. Nauk SSSR, Ser. Khim., No.8, p. 1373
(1964).
16. J. F. MacDowell, Ind. Eng. Chern., 58(3):38 (1966).
17. Y. Moria, D. H. Warrington, and R. W. Douglas, Phys. Chern. Glasses, 8, p. 19 (1967).
18. Ya. 1. Ol'shanskii, Dokl. Akad. Nauk SSSR, 26:93 (1951).
19. R. J. Charles, Phys. Chern. Glasses, 8:185 (1967).
20. N. A. Toropov and V. A. Vasil 'eva, Kristallografiya, 6:968 (1961).
21. N. A. Toropov and 1. A. Bondar', Izv. Akad. Nauk SSSR, Otd. Khim. Nauk, No.4, p. 544
(1961).
22. N. A. Toropov and I. A. Bondar', Izv. Akad. Nauk SSSR, Otd. Khim. Nauk, No.5, p. 739
(1961).
23. N. A. Toropov and T. P. Kiseleva, Tr. Leningr. Tekhnol. Inst. Im. Lensoveta, 52:76 (1961).
24. N. A. Toropov and 1. A. Bondar', Izv. Akad. Nauk SSSR, Otd. Khim. Nauk, No.8, p. 1372
(1961).
25. N. A. Toropov, F. Ya. Galakhov, and S. F. Konovalova, Izv. Akad. Nauk SSSR, Otd. Khim.
Nauk, No.4, p. 539 (1961).
26. N. A. Toropov, F. Ya. Galakhov, and S. F. Konovalova, Izv. Akad. Nauk SSSR, Otd. Khim.
Nauk, No.8, p. 1365 (1961).
27. R. S. De Vries, R. Roy, and E. F. Osborn, Trans. Brit. Ceram. Soc., 53:525 (1954).
28. T. P. Seward, III, D. R. Uhlmann, and D. Turnbull, J. Am. Ceram. Soc., 51:278 (1968).
THEORY OF PHASE SEPARATION AND
ATOMIC-IONIC STR UCTURE OF CERTAIN
TWO-COMPONENT SYSTEMS

v. N. Filipovich and D. D. Dmitriev


A fairly extensive review of the literature on the nature of liquid-liquid phase separation
in silicate melts and glasses has already been given in the preceding paper by Galakhov and
Varshal. The views of various writers on the nature of phase separation are in the main
either purely qualitative or involve attempts to find formal relationships between ionic
radii, cationic field strengths, hypothetical local structures, and experimental characteristics
of phase separation. All this can be regarded merely as preliminary steps toward derivation
of a theory of phase separation, which must be a statistical physical theory based on more or
less detailed analysis of interatomic interactions. An advance in this direction is, for exam-
ple, the work of Kozheurov [1], who made an attempt to write an expression for the free
energy of a silicate melt undergoing phase separation. Owing to the complexity of his polymeric
model, he did not fully succeed in substantiating theoretically the addition of free energy postu-
lated by him, so that the final formulas are of semiempirical character. In the present paper
we derive a model, based on simplified concepts of structure and atomic- ionic interactions,
making it possible to make approximate calculations of the free energy of a silicate melt or glass
as a function of its composition and temperature. In our opinion, this model combines to a cer-
tain degree the views put forward earlier by Warren and Pincus [2], Dietzel [3], and
Markhasev and Sedletskii [4] on the nature of phase separation.
The following discussion refers mainly to the free energy of alkali and alkaline-earth
silicate melts and glasses, although cases involving oxides of tri- and quadrivalent metals are
also discussed in part.
1. Free Energy of a Melt or Glass in an Me20 - Si0 2

We first consider dissolution of an Me20 molecule in Si02 in accordance with the scheme

/O-Me
0 . Si-O-SiO.,/ -I- ~Ie-O-Me=03/ Si/. /SiO,/. (1)
'.', 2 '1I1e-0'

The energetic advantage of this process appears to be determined by two main factors.
1. The bond between a nonbridging oxygen and an Si4+ ion is stronger than the Si- bond
for a bridging oxygen, as in view of the weak polarizing action of the Me + ion in comparison with
Si 4+ the effective negative charge of the 0 2- ion approaches the Si 4+ ion, and this leads to a gain
12
V. N. FILIPOVICH AND D. D. DMITRIEV 13

of energy. The Si-O bond for a nonbridging oxygen therefore becomes more ionic. This re-
sults in a less steep potential well characterizing the Si - 0 bond and accordingly in a lowering
of the Si-O vibration frequency, observed experimentally [5].
2. As the result of rupture of Si - 0- Si bridges the multiply charged Si 4+ ions can move
away from each other somewhat, which is also energetically advantageous [6]. Incidentally, it
follows from the foregoing that the greater the charge and the smaller the radius of the metal
ion, i.e., the greater the force of repulsion from the Si4+ ion and the ability of the ion to polarize
oxygen, the less is the energy advantage of dissolution of its oxide in Si02.
The following fact should be noted. The density of vitreous silica is close to that of cris-
tobalite. Cristobalite has a honeycombed structure of six-membered rings of Si0 4/ 2 tetrahedra,
the diameter of the cavity within the ring reaching the diameter of an oxygen ion. It can there-
fore be concluded that with regard to looseness of packing of oxygen ions the structure of Si02
glass or melt is close to the structure of cristobalite. As a result, when an Me20 molecule
enters this structure in accordance with scheme (1) the Me+ ions become located at fairly con-
siderable distances from the surrounding oxygen ions, with the exception of one to three oxygen
ions directly adjacent to the Me+ ion, which interacts most strongly with the nonbridging oxygen.
It is evident from this that if the surrounding 0 2- ions are given an opportunity of approaching
close to the Me+ ion there would be a gain of potential energy. If only one Me20 "molecule"
enters Si0 2 in accordance with scheme (1), such approach of 0 2- to Me+ is impossible, as this
would lead to substantial deformations and stresses in the surrounding framework, which tends
to retain its structure unchanged. However, if the number of built-in Me20 "molecules" in-
creases and they can approach each other. then as the result of rupture of an increasing number of
Si - 0- Si bridge bonds in the vicinity of a given Me+ ion surrounding 0 2- ions can approach
closer to Me+ without significant stresses in the surrounding structure, the deformability of
which increases as rupture of Si- 0- Si bridges proceeds. A gain of energy may be correspond-
ingly achieved owing to attraction of unlike charges. This process may continue with increase
of the Me20 concentration until Me+ - 0 2- distances and a coordination number of Me+ for oxygen
ensuring minimum potential energy are attained. The average energy gain per Me+ ion is ap-
proximately

11 !1r, (2)

where KMe is the number of 0 2- ions approaching the Me+ ion and traveling through the distance
.6.r in the process; r Me , ro, eMe, and eo are the radii and effective charges of the metal and oxygen
ions; a.6.r is the work done by forces acting on the 0 2- ion from the surrounding particles, with
the exception of Me+, and opposing its approach to Me+; a and .6.r depend on the nature of the
Me+ ion and the surrounding structure. In general u (2) is a complex function of many inter-
atomic distances, of which we have defined explicitly only one: the dependence on r Me + r o and
.6.r, which in our view is the most significant, while the term a.6.r includes all the rest: change
of the energy of oxygen polarization by Me+ ions, etc.
Before searching for the dependence of internal energy on the concentration of metal ions
or nonbridging oxygens, we will calculate the entropy of the glass or melt. In accordance with
statistical physics, the free energy of the system is given by the formula
E

F = -kT In Z, Z = ~ e- k1' g (E), (3)


E

where Z is the statistical sum; g(E) is the statistical weight or number of different microstates
14 V. N. FILIPOVICH AND Do D. DMITRIEV

having the same energy E in the system. Assuming that the main role in (3) is played by terms
E E+a.
with energy close to the average energy E of the system, we have Z = e- kT ~ g (E
E-a.
+ e) and

E+a.
F=E-TS, S=kInG(E), G(E)= ~ g(E+E), (4)
E-a.

where G(E) is the number of states having energy differing from E by not more than a certain
small quantity Doc:. Formula (4) corresponds to the approximation in our calculation of free
energy. The calculation becomes even less exact in view of the fact that we will deal not with the
total energy and entropy of the systf'm but only with the parts including the dependence on the
chemical composition of the glass with which we are concerned, also omitting the temperature
dependence of E and S.
In the light of the foregoing, we can characterize varous microstates of the glass approxi-
mately (see also [7]) as states differing in the arrangement of nonbridging oxygens (0- Me
groups) at the corners of Si0 4!2 tetrahedra in the volume of the glass. Let N be the number of
Si atoms always surrounded by oxygens in tetrahedra, and xN the number of univalent metal ions
in the glass. The statistical weight of the macrostate of the glass is equal to the number of ways
(combinations) in which xN nonbridging oxygens can be distributed between 4N corners of N tet-
rahedra so that the energy E of the glass remains approximately the same in these distributions.
Tn particular, combinations leading to substantial increase of energy of the glass must be excluded
from the possible number. We will assume that attachment of more than one nonbridging oxygen
to one Si0 4!2 tetrahedron is energetically disadvantageous in the composition region under con-
sideration, where metastable phase separation occurs in alkali silicate glasses. This can be
explained as follows: attachment of two nonbridging oxygens together with two Me+ ions to the
tetrahedron intensifies repulsion between nonbridging oxygens because their negative effective
charges are closer together than in the case of neighboring bridging and nonbridging oxygens.
Moreover, dipole repulsion between polarized metal atoms located at neighboring nonbridging
oxygens of the tetrahedron is strong. Finally, preliminary calculations show that better agree-
ment between theory and experiment is obtained on the assumption that one Me atom is attached
to one tetrahedron than if it is assumed that four Me atoms can be attached to one tetrahedron.
With all the foregoing taken into account, the statistical weight and entropy of the glass are:

~N (4N-3Nx)! }
G=C4N_3Nx = (Nx)! (4N -4Nx)!' (5)
S = kN [(4 - 3x) In (4 - 3x) - 4 (1- x) In 4 (1 -- x) - x In xl,

where the Stirling formula InN! N(lnN -1) is used, and in calculation of the number of combi-
0=

nations crg it is taken into account that on introduction of each nonbridging oxygen the total num-
ber of possible sites for them diminishes by three because not more than one nonbridging oxygen
can be attached to each tetrahedron.
In the preceding discussion we assumed that all the nonbridging oxygens are mutually in-
dependent, except that they cannot be bonded to the same tetrahedron. However, if we consider
a vitreous silica structure similar, for example, to the structure of cristobalite, it is easy to
see how nonbridging oxygens can be bonded in pairs in accordance with scheme (1) and, converse-
ly, no reasons can be seen why it is advantageous for these pairs to divide into separate 0- Me
groups. Moreover, such separation must involve, at least at low metal-atom concentrations,
certain significant deformations of the structure of glassy Si02 and energetically disadvantageous

switching of Si- bonds. The possibility of paired bonding of Me atoms in yMe20. (1-y)Si0 2
glasses with low Me20 contents follows from experiments on x-ray scattering [8], giving maxima
apparently corresponding to the Me- Me(Cs- Cs) distance on the radial distribution curves. It
V. N. FILIPOVICH AND D. D. DMITRIEV 15

should be noted that Foerland [7] also admits the possibility of paired bonding of nonbridging
oxygens in alkali silicate glass. With this assumption it is possible to bring into conformity the
calculated and experimental lowering of the melting; point of Si02 on addition of an alkali oxide.
Taking all this into account, for comparison with (5) we also obtain an expression for the entropy
on the assumption that Me atoms combine in pairs not only at low but at all concentrations O~
x~l under consideration. With the assumption of paired arrangement of nonbridging oxygens it
is evident that the total initial number of possible sites for these Nx/2 pairs is the number 2N
of bridging oxygens in the original Si02 to which the oxide Me20 is added. Each added Me atom
diminishes the number of nonbridging oxygens which can be replaced by pairs by 3/2, corre-
sponding to three halves of a bridging oxygen at the three remaining corners of the Si0 4/ 2 tetra-
hedron to which the Me atom is bonded. Therefore analogously to (5)

(6)

Equations (6) can be obtained from (5) by substitution of N/2 for N. Correspondingly, en-
tropy (5) is double the entropy (6).
We now turn to the dependence of the energy E on the concentration x. At low concentra-
tions x 1 the energy of the glass is greater by E = Nxu than the energy E = 0 obtained when
the oxygen ions approach the metal ions to the optimal distance. We will assume that this close
proximity of 0 2- ions to Me+ is attained at the concentration x = 1, corresponding to the compo-
sition of the alkali disilicate. In the crystalline state, the latter consists of a system of layers
of Si04/2 tetrahedra linked by three corners through bridging oxygens while the fourth corner
has a nonbridging oxygen, with metal ions between the layers bonding them together. It may
therefore be supposed that a disilicate glass consists of flexible fragments of silicon-oxygen
layers, which results in close proximity of oxygen ions to Me+ ions. However, if the metal
oxide concentration is decreased the layers become bonded by bridging oxygens and lose their
flexibility, and the framework structure begins to dictate its coordination requirements. If
bridging oxygen atoms appear between the layers in addition to Me+ ions, this leads to loss of
energy of the order of u (2) owing to removal of oxygen ions from metal ions as the result of
formation of a rigid local framework structure. The probability of formation of such a "dis-
advatageous" ion is equal to the probability x of entry of this ion into the given volume, multiplied
by the probability (l-x)il that n other Me+ ions will not enter that volume, so that in their place
there will be bridging oxygens leading to a local framework structure. The number of such
"disadvantageous" ions is Nx(l-x)il and consequently

E= Nux (1 -x)", n ~ 2, (7)

where the minimum value of n can probably be taken as 2, as the presence of two "supports"
between the layers on the two dies of the given ion may lead, at least partially, to a local frame-
work structure.
A formula of the type of (7) can also be derived on other and more formal grounds. It may
be assumed that .6.r in (2) decreases with increase of x by the law .6.r = .6.ro(l- X)il owing to in-
creasing flexibility of the glass framework and corresponding gradual approach of oxygen ions
to the metal ions; here n(n = 1, 2, etc.) is an unknown parameter which must be chosen to give
the best agreement between theory and experiment. The results of calculations for n = 1 and
2 are given in 4.
16 V. N. FILIPOVICH AND D. D. DMITRIEV

2. Free Energy of MeO - Si0 2 Melts and Glasses


Dissolution of the first portions of MeO in Si02 probably proceeds by a scheme of the type

of (1) with formation of - 0- Me- groups instead of - 0- bridges; the energetic advantage of
this reaction may be established analogously to (1). It is known from experimental data that in
this case the immiscibility regions lie between the metasilicate composition MeO . Si02 and
Si02 It must therefore be concluded that the composition corresponding to optimal mutual ap-
proach and coordination of metal and oxygen ions must be close to the metasilicate rather than
the disilicate as in the case of Me20. The explanation is that, in distinction from Me20, in this
case the nonbridging corners of the tetrahedra are strongly bonded in pairs through bivalent
metal atoms. Therefore the rigidity of the framework and the number of "disadvantageous"
ions in the local framework environment alters with increase of the number of bivalent metal
atoms by approximately the same law as in the case of univalent metals. Accordingly, we will
consider Eqs. (7) and (2) to be valid, as before, for the energy of MeO- Si02 glasses or melts
in the immiscibility region, with the bivalent character of the metal taken into account.
For calculation of entropy we will assume, as in the case of Me20, that positions with
more than one bivalent-metal atom per tetrahedron are energetically disadvantageous and must
be excluded in calculation of the microstates corresponding to the specified energy of the melt
or glass. The required number G of microstates is equal to the number of ways in which xN

0- Me- groups can be arranged instead of 2N bridging oxygens in the original Si02 with N
Si atoms. We have

(2N-Nx)!
G=(Nx)! (2N-2Nx)!'

S =kN [(2 - x) In (2-x) -2 (1- x) In2 (1- x) - xInx), (8)

where it is taken into account that each added Me atoms decreases on the average by 2 (1/2) =
1 the number of nonbridging oxygens which can be replaced by - 0- Me- 0- groups.
It should be noted that, according to radial distribution curves, Na20- BaO- Si02 glasses
apparently have stable Ba- Ba distances even at relatively low BaO concentrations. Formation
of Ba - Ba pairs, if it occurs, may be attributed to dipole imteraction of - 0- Me- 0- groups.
It is also possible that, for some reason, when MeO dissolves in Si02 formation of groups in the
form of paired tetrahedra (a) is advantageous or pairs are formed by groups (b):

O-Me-O
I I -0, 0,
-O-Si -O-Si-O-- (a) 'Si< 'Me (b) (9)
I I -0/ 0/
O-Me-O

If these possibilities are taken into consideration, the expressions for the entropy and
energy of glasses must, of course, be affected but the overall picture is hardly likely to change.

3. Free Energy of Melts and Glasses in Other Systems


The methods used in 1 and 2 can be used, at least partially, also for finding approxi-
mate expressions for the free energy of glasses and melts containing Me203 and Me02 oxides.
However, in view of the greater complexity of the structural elements and of their interactions
with surrounding elements, the problem becomes more complicated and other methods and
models may be needed for its solution. For example, when Me203(Cr203) or Me02(Ti02) are
V. N. FILIPOVICH AND D. D. DMITRIEV 17

dissolved structural groups of the following types may be formed:

O,/-Me-O,/ 0,/ 0,/,


'I
0

(A),
I
O,/-Me-O,/ (B). O,/-Ji-Oll (el (10)
I 'I ' 'I IZ

O'/,-Me-O,/, 0,/, 0,/,

The quadrivalent groups Me203(A), Me02(B) may be regarded as certain formal equivalent
of the quadrivalent tetrahedral Si02 group (C). If model (10) is valid, the melt or glass may be
regarded as a mixture of A, C and B, C groups. A variant of the theory of regular solutions
[9] is convenient for description of such systems. If N is the total number of structural elements
A and C in a glass of the composition yA. (l-y)C, we have

N! (11)
G=C N
liN
= (yN)! (N -yN)! S =-kN lyiny + (1-y) In (1-y)].

The energy is calculated as follows, with the assumption of equal probability of neighbors
A, C, interacting with energies UAh Ucc, UAC < O. A structural element A has yK neighbors A and
(1-y)K neighbors C, where K is the coordination number, 4 in this case, in accordance with the
valences of the groups (10). The number of elements A is Ny, and their total bonding energies
with neighbors A and C are, respectively, 1I2Ny2KUM and NyK(1-y)UAC, where the factor 1/2 ap-
pears because interaction between elements A is taken into account twice. The energy of inter-
action between elements C is 1j2N(1-y)2KUc C' As a result

(12)

If the "tetrahedra" A and C (10) are similar in size, UAC,UAA,Uccmay be assumed equal

to half the sum of the Me- and Si - bond energies in the Me- 0- Si bridge, etc. If the tetra-
hedra differ greatly in size, the values of U must include the energy of structural stresses.
The energy of mixing, U, is of the most interest in (12), since on it depends the existence
of a maximum (if U> 0) of E as a function of y in (12). The quantity KU in (12) is allied to u in
(7). In general, we could at once write an expression for the energy in the form E = Nuy(l-y).
u = KU, on the basis that u is the average loss of energy due to the fact that Ny elements A have
on the average K(1-y) disadvantageous neighbors C. As an example, let us consider the system
Ti~- Si02 from this standpoint. The type B Ti0 2 tetrahedron (10) (if Ti02 is present in fourfold
coordination in the melt or glass) is probably smaller than the Si02 tetrahedron (10); this follows
from the smaller molar volume of crystalline forms of Ti02 (18.7-20.7) in comparison with the
molar volume of cristobalite, 26. Therefore, forming in a glass with a high Si02 content, this
Ti02 tetrahedron is in an extended state. The extension is released when the Ti02 tetrahedra
come close together and form a phase rich in Ti02' It may be expected that the gain of energy
u in this case will be of a form analogous to (2). Turning to Cr203 (molar volume 29.2), we
find a value of about 20 for its molar volume calculated for two oxygen atoms. Possibly in this
case also Si02 on the average pulls oxygens of the Cr203 group A (10) apart so that after mutual
approach of these groups a gain of energy is obtained, given as before by a formula of the type
of (2).
It should be noted that Eqs. (11) and (12) are applicable only to systems with very wide
immiscibility regions, lying approximately in the range 0 5 Y 51. Immiscibility regions of
preCisely this kind are found in the systems Ti02- Si02, Cr203- Si02
18 V. N. FILIPOVICH AND D. D. DMITRIEV

4. Discussion of Theoretical Results and Comparison with Experi-


mental Data
We write the equation for the free energy F1 of yMe20. (l-y)Si0 2 glasses, assuming that
formation of Me- Me pairs is insignificant and that Eq. (5) can be used for the entropy:

F 1=NU{X(1- x)n __ k: [(4-:lx)ln(4-3x)-4 (1-x)ln4(1-X)- x1nx l},


2y
X= 1 _ y , (13)

together with a formula for converting x to the molar glass composition y. To pass to the model
in which nonbridging oxygens are bonded in pairs and the entropy is given by Eq. (6), the en-
tropy term in the expression for F1 (13) should be multiplied by 1/2.
For yMeO. (l-y)Si02 glasses we have from Eqs. (7) and (8)

F2~= NU{X(l - x)"- k~ [(2-.'C)ln(2-x) --2(1-x)ln2(1-X)-x1nxl},


y
:t=l_y (14)

Plots of FtiNu for n = 1 and 2 and (3 = kT lu = 0.2 are given in Fig. 1. It has been shown
(e.g., see [10]) that the compositions of phases formed as the result of separation correspond
to points of contact of the F1 curve with a common tangent (shown in Fig. 1), which corresponds
to the condition of equal chemical potentials of the components in the phases. By varying (3 (Le,
the temperature T) we can obtain the binodal (3 (y) curve giving the concentrations Y1 and Y2 of
phases in equilibrium at a given temperature. Further, from the equation d 2F/dx 2 = 0 for dif-
ferent (3 we obtain a spinodal curve separating regions having different kinetics of phase con-
version [10, 111. The corresponding diagrams plotted from F1 and F2 (13), (14), for n = 1 and
2 are shown in Fig. 2. Figures 3 and 4 show a comparison of theoretical and experimental curves
for the case n = 2, since n = 2 gives better agreement with experimental data. The agreement

Fig. 1. Plots of F, S, and


E (13) as functions of x for
f3 = 0.2, n:::; 1 and 2.
V. N. FILIPOVICH AND D. D. DMITRIEV 19

a b
n=4
n=1
n=2
f3 \'\ fi
\ \
0.3 \ \
\ \
\\
Fig. 2. Immiscibility diagrams for the
systems yMeO (l-y) Si~ (a) and yMe20'
\\
\
\
,,
(l-y)Si~ (b).
\I
\I
0 0.1 0.4 0.3 0.4 0.5 Y 0 0.1 0.2 0.3 Y

Fig. 3. Comparison of theory with ex-


periment. 1) Experiment (Na, [13]; Li,
[15]); 2) theory.
5 10 1520
Nap, mole% Li~, mole%

/
1
/
-- " 2
.......
\
I BtL \
1200 I \ \
I \ \
1000 \ \
10 20 30 \ \
CaD, mole% 800 \ \
to 600
10 20 30
2100 BaD, mole%
to 1
1900 1900

10 20 30 40 1700 10 20
MgO, mole% SrD, mole%
Fig. 4. Comparison of theory with experiment.
1) Experiment (Ca, Mg, and Sr, [16]; Ba, [15]);
2) theory.
20 V. N. FILIPOVICH AND D. D. DMITRIEV

TABLE 1

System Ter Ber


U, keal/
mole
TMe' A KMe AT, A I
AT (K"Mf. = 4),

Li 2O-Si0 2 1430 0.38 7.5 0.60 3 0.015 0.011


Na 2 O-Si0 2 1130 0.38 5.9 0.95 5 0.010 0.012
MgO-Si0 2 2470 0.42 11.7 0.65 4 0.010 0.010
CaO-Si0 2 2370 0.42 11.2 0.99 4 0.012 0.012
SrO-Si0 2 2190 0.42 10.4 1.13 6 0.009 0.012
BaO-Si0 2 1870 0.42 8.8 1.35 6 0.009 0.012

between theory and experiment is satisfactory for alkali glasses and noticeably worse for alkaline-
earth glasses (melts). There is reason to suppose that agreement with experimental data will
improve with increase of the index n in (13) and (14).
Knowing the experimental critical temperatures Ter and the critical values of (3 = (3cr
from Figs. 2, we can find u for various melts or glasses. These data are presented in Table 1.
This table also gives values of Ar calculated from known u with the aid of Eq. (2), where it was
assumed that 0' = 0, the charges eMe are e and 2e (e is the electronic charge), rMe were taken
from Pauling's data [12], and ro = 1.4 A. The values taken for K Me were Kl (1) and K2 (2), the
coordination numbers of Me+ (K 1) and Me 2+ (K2) ions in crystalline disilicates and metasilicates.
decreased by 1 and 2, respectively (to allow for the fact that the metal atom is right from the
start in close contact at least with the oxygen atoms to which it is linked by valence bonds). It
follows from Table 1 that Ar is of the order of 0.01 A, but Ar increases if we take into account
0' in (2) and also the fact that the effective ion charges are considerably less than e and 2e. The
table also includes values found for Ar when KMeis taken to be the same (4) for all Me+ and
Me2+ ions. These values of Ar are all equal within 10-20%. This means that the critical tem-
perature Ter is proportional to the cation field strength eMe/(r Me + ro)2 to the same degree of
accuracy (regardless of any theory). There are no grounds for attaching too much importance
to this law; nevertheless the models discussed above provide qualitative and partial quantitative
confirmation of the idea that the smaller the ionic radius the greater the field strength, and the
lower the tendency of the corresponding oxide to mix with 8i02 in melts and glasses and to form
crystalline compounds with 8i02 at higher 8i02 concentrations (immiscibility in the amorphous
state is a sign of immiscibility in the crystalline state, and conversely). The tendency of a me-
tal ion to diminish its coordination number to the limit is the consequence of its tendency to at-
tract oxygen ions as near as possible. However, it is appropriate to note here that silicon-
oxygen radicals as well as metal ions play an active part in structure formation. In particular,
when the silicon- oxygen framework has not yet been sufficiently disrupted by metal ions it is
the framework which determines the structural and coordination state of the metal in the glass
or melt. In the final analysis, metal ions and silicon- oxygen radicals are both active partners,
each with its own structural requirements.
We draw attention to the following important point. If we assume that in Me20- Si02 glass-
es nonbridging oxygens group into pairs at all O:sx :sI, then the binodals and spinodals in Fig.
2a retain their form entirely but (3er = kT er lu is doubled ((3 = 0.76), which corresponds to de-
crease of u (and Ar) by half in comparison with the da~a in Table 1. The physical meaning of
this result is understandable, because as the result of pairing the number of structural units
and the number of possible random sites for them within the glass are halved. This increase of
order diminishes the influence of the entropy term, and an energy gain diminished by a half is
therefore sufficient for the given phase separation of the glass. This result also demonstrates
that entropy as well as energy factors must be taken correctly into account in order to obtain
correct quantitative data.
V. N. FILIPOVICH AND D. D. DMITRIEV 21

a b
Fig. 5. Schematic diagrams of
aggregation of anisotropic struc-
tural elements into two-dimen-
sional (a) and one-dimens ional
(b) formations.

The methods discussed above can also be applied to phase separation in multicomponent
systems. For example, it is already possible to predict the influence of addition of a third
component on the temperature of phase separation. Suppose, for example, that to a glass of
the composition yMezO. (l-y)SiO z another univalent-metal oxide is added in an amount z such
that the total content of the two oxides in the glass remains the same as before: y = Yi + zi'
Then eVidently the value of the factor (l-xf in (7) remains unchanged, as the number of breaks
in the silicon- oxygen network does not alter. However, the quantity ux changes, and should be
replaced by Uto xi + u2' x2' where xi and x2 correspond to the concentrations Yi and Zt. If the
oxide added has a lower critical temperature T cr 2 < T cr t (i.e., u2 < Ut), this leads to decrease of
Tcr< Tcrt and vice versa. These as yet tentative considerations are in agreement with experi-
mental data obtained by Andreev and Boiko [13].
With regard to the prospects of development of the theory, it must be noted that it requires
further refinement. This includes a more accurate estimate of the number of configuration
states in the glass corresponding to a specified energy. It is evident that the condition that not
more than one Me atom is attached to a tetrahedron is an approximation, especially inaccurate
at large x -1. It is also important to take into account formation of nonbridging pairs and of
any other stable structural elements or groups in the glass. Anisotropy of interaction of struc-
tural elements in the glass may also be significant. For example, it may happen that the struc-
tural elements can be schematically represented by cylinder or parallelepipeds interacting
strongly by their lateral surfaces and weakly by their bases (Fig. 5a). Such elements will tend
to form layers. Moreover, existence of two-dimensional phase separation, i.e., formation of
a kind of two-dimensional phases, is possible. It laws are formally the same as for ordinary
three-dimensional separation (e.g., see [11], [12]). However, it is difficult to detect phase sep-
aration, as layer formations are large in only two dimensions, intersect at various angles, and
form a complex microstructure in the glass.
The existence of a kind of one-dimensional phase separation may also be postulated. In
this case the structural elements are plates interacting strongly by their bases and weakly by
their lateral surfaces (Fig. 5b). These plates join to form chains, the length of which increases
with decrease of temperature. The average number of units in the chain can be estimated from
the formula N/n = eE/kT, where N is the number of units in a certain nominal long chain, n is the
number of breaks in this chain, and , is the energy of rupture, i.e., the increase of the glass
energy resulting from replacement of a unit in the chain by some other surrounding structural
element. All the glass may be penetrated by such chains, but this heterogeneity is difficult to
detect. It is possible that such borate chains and threads [with B- 0- B linkages of type A (10)]
are formed in leachable borosilicate glasses [14]. Owing to intersection of the chains complete
removal of the borate component is possible, with replacement of that component by channels
of the order of 10 A in diameter.
Formation of one- and two-dimensional structures of the kind described above may pre-
cede the usual three-dimensional phase separation.
Literature Cited
1. V. A. Kozheurov, Thermodynamics of Metallurgical Alloys, Sverdlovsk (1955).
2. B. E. Warren and A. G. Pincus, J. Am. Ceram. Soc., 23:301 (1940).
22 V. N. FILIPOVICH AND D. D. DMITRIEV

3. A. Dietzel, Z. Elektrochem., 48:9 (1942).


4. B. 1. Markhasev and I. D. Sedletskii, Dokl. Akad. Nauk SSSR, 148:916 (1963); 154:1125
(1964); B. I. Markhasev, Zh. Strukt. Khim., 6:83 (1965).
50 V. A. Kolesova, in: The Glassy State, Proceedings of the Third All-Union Conference,
Izd. Akad. Nauk SSSR, Moscow- Leningrad (1960), p. 203 [English translation: The
Structure of Glass, Vol. 2, Consultants Bureau, New York (1960), p. 1771.
6. A. Dietzel, Glastechn. Ber., 22(3-4):41 (1948).
70 T. Foerland, in: Structure of Fused Salts [Russian translationl, Mir, Moscow (1966),
p. 283.
8. C. Brosset, Eighth International Ceramic Congress, Copenhagen (1962), p. 15.
9. Ro Becker, Ann. Phys., 32:128 (1938); V. N. Filipovich, in: The Glassy State, Vol. 3,
No. I, Catalyzed Crystallization of Glass, Izd. Akad. Nauk SSSR, Moscow- Leningrad
(1963), p. 9 [English translation: The Structure of Glass, Consultants Bureau, New York
(1964), p. 91.
10. V. N. Filipovich, Neorgan. Mater., 3:993 (1967).
11. V. N. Filipovich, Neorgan. Mater., 3:1192 (1967).
12. R. C. Evans, Introduction to Crystal Chemistry, 2nd ed., Cambridge Univ. Press (1966).
13. N. S. Andreev and G. G. Boiko, this volume, p. 6].
14. S. P. Zhdanov and E. V. Koromal'di, this volume, p. 92.
15. R. J. Charles, Phys. Chem. Glasses, 8:185 (1967).
16. Ya. 1. Ol'shanskii, Dokl. Akad. Nauk SSSR, 76:93 (1951).
DISCUSSION
Replying to Dzhurinskii's question regarding the cause of the good agreement be-
tween the experimental and theoretical values of the limiting concntrations of separating sili-
cate melts, calculated by Levin, F iii po vic h observed that the agreement appears to
be of somewhat fortuitous character, as Levin's geometrical considerations of the types of
structural elements in silicate melts are insufficient as a basis for a statistical theory of melts.
Answering Nemilov, who asked how the values of ~r were calculated, Filipovich said that
purely theoretical calculation of ~r is impossible at present and therefore experimen-
tally determined critical temperatures of phase separation were used for the calculations. The
fact that the ~r values for different cations were roughly equal indicates that the temperature
of phase separation is approximately proportional to the field strength of the cation.
SOME THERMODYNAMIC ASPECTS OF
PHASESEPARATION PHENOMENA

M. M. Shul'ts

One of the causes of heterogeneity in glasses is phase separation in metls from which the
glasses are formed. Separation of supercooled metals, i.e., of stable phases with respect to the
crystalline state, is of special interest. The degree of probability of separation in supercooled
melts can be estimated from the form of the liquidus lines in phase diagrams of the correspond-
ing systems. This range of problems is directly related to the general thermodynamic theory
of heterogeneous systems, developed by Gibbs, van der Waals, and others [1-4]. Applica-
tions of thermodynamic theory to the problem under consideration may be found in the
specialist literature [5-7]. The present communication is confined to an examination of only
certain special problems. First we consider the relation between the form of the liquidus line
and the position of the critical point corresponding to phase separation.
The discussion which follows is based on thermodynamic conditions for stable equilibrium,
expressed in the following particular form:

or-, )
( oN, (1)
T.P:;;'O,

where Jli and Ni are the chemical potential and mole fraction of component i.*
Expression (1) is one of the forms of the necessary conditions for stability of equilibrium
with respect to continuous variations of state. It is presumed that the inequality sign in (1) cor-
responds to stable states (stable and metastable) while the equality sign corresponds to the
stability boundary, i.e., the spinodal curve on the phase diagram.
Since the critical phase lies on the stability boundary, it also satisfies the condition

()r-, )(k)
( -aN, (2)
-0
T,p- .

Moreover, the critical phase conforms to the equation

(3)

*Here and subsequently only two-component systems are considered.


23
24 M. M. SHUL' TS

Taking into account the known expression for the chemical potential of a component in
terms of its activity, J.Li = J.L ~ (T, p) + RTlna i, we have from Eqs. (2) and (3)

( Olll.a,)(k) =0 (4)
(ll'.. i T, p

and

( 2
a.
iJ2 ill )(k) _
-0. (5)
aN, T,p

Together with the above equations for the critical phase, we shall use the equation for the
liquidus line, confining the discussion to systems containing a two-component melt and a one-
component solid phase. In general, this equation can be written only in differential form:

(6)

where T is the temperature corresponding to the liquidus line; Ni is the mole fraction in the
melt of the component which is also present in the equilibrium solid phase; ai is its activity in
the melt; ~Hi is the heat of solution of this component in the saturated melt (the difference be-
tween the partial molar enthalpies in the melt and solid phase, ~Hi = Hf- Hlo~.
If the dependence of ~Hi on the phase-equilibrium temperature is known and the course
of the liquidus line has been determined, it is possible to find the dependence of the activities
of both metal components on concentration, to calculate the free energy of melt formation and,
presenting it as a function of the melt composition, to find the course of the phase-separation
binodal under the corresponding conditions. However, for solution of this problem it is usually
necessary to make certain simplifying assumptions, as was done, e.g., by Charles and Wagstaff
[7].

For qualitative analysis of the form of the liquidus line it is necessary to take into account
also the expression for the second derivative of T with respect to Ni, which may be written as
follows:

( dN
d2~) =RT2[C.(~)
oN T p+
(02lna i)]
~.,2i T , P ' (7)
p AH'
2
t' " to, fJI.Y

where
1
C;= AH."
2 In a.)
l- RT2. (0oTiJN.
,. T, P
In
+2RT ~ to
(a at)
T, P
-
(OdHI)
aN.
to
]
P

Examination of Eqs. (6) and (7), with (4) and (5) taken into account, leads to the following
evident result. If the critical separation point lies on the liquidus curve, then the following
conditions are satisfied at this point:

which means that an inflection appears on the liquidus line, with a horizontal tangent to the line.
By virtue of the continuity of variation of the states and properties of the system it is to be ex-
pected that if the immiscibility region (binodal) is not too far below the liquidus line the latter
may also have an inflection, but the tangent to the curve at this point will not be horizontal.
M. M. SHUL'TS 25

Evidently, a correspondence is to be expected between the slope of the tangent and the position of
the critical separation point under the liquidus line; this will be considered more specifically
somewhat later. At the same time it cannot be proved on strictly thermodynamic grounds that
the appearance of an inflection on the liquidus line is a sufficient condition for phase separation
in a metastable melt.
We will apply the equations written above to various types of melts, starting with those
defined as ideal.
A n Ide a 1 Sol uti 0 n (M e 1 t ) is characterized by the following dependence of the
chemical potentials of the components on their concentration:

(8)

If the melt exhibits ideal behavior, the course of the liquidus line is determined by the
equation

(9)

In the case under consideration the partial molar enthalpies of the components in the
melt are the same as of the pure components. Therefore ~Hi represents the difference between
the molar enthalpies of pure component i in the liquid and solid states at the phase-equilibrium
temperature. If the melt is ideal over the entire temperature range corresponding to the liq-
uidus line, it cannot separate into two layers, as the following is alwasy valid for an ideal melt:

ii In a i )
(- 1
-
iiN, T,p
=->0.
N,
(10)

However, it is known [4, p. 343] that variations in the form of the liquidus line are pos-
sible also in this case. Neglecting in the first approximation the dependence of ~Hi on temper-
ature, the following expression is easily derived from Eq. (9):

d2 T ] R T2 [2R T 1 ]
[ dN7 = N;!1H, /:;H,: - N, (11)

Equation (11) shows that the initial course of the liquidus curve may be either convex or
concave, dependent on the magnitude of ~Hi' If ~H~ < 2RT, then (d 2Tid Nn0 > 0, and the liquidus
curve is therefore concave, while with the condition ~Hii > 2RT, when (d2T/dNr) p < 0, the liquidus
curve is convex. If ~Hi does not differ very much from 2RT, the sign of the expression in brack-
ets in the right-hand side of Eq. (11) can change along the liquidus line, which implies the ap-
pearance of an inflection in absence of the necessary conditions for phase separation in the melt.
This situation appears to be very rare. However, it can be concluded from the foregoing that
the existence of an inflection on the liquidus curve is not a sufficient condition for phase separa-
tion in a metastable melt. We may note here that a necessary condition for phase separation in
melts is the occurrence of positive deviations from ideal behavior; this is considered specifical-
ly below with reference to regular solutions.
A Strictly Regular Solution (Melt) is characterized by the following dependence
of the chemical potential of each component on concentration:

p.i=p.~(T. p}+RTInN,+2(I-N;)2, (12)


26 M. M. SHUL'TS

where a is a certain constant characterizing interaction in the melt (interchange energy). If a >
0, phase separation in the melt is possible in principle.
From Eq. (12) we obtain the equation for the derivative

()U.i ) RT .
( -'-
aN, T,p
=--2a(1-N.),
Ni
(13)

from which it follows that the melt is stable if the following condition is satisfied:

(14)

It is not difficult to show that when a> 2RT there is a concentration range in which the
melt undergoes separation.
The coordinates of the critical point in the case under consideration are easily found by
solving the system of equations (2) and (3), with Eq. (12) for the chemical potential taken into
account. This solution gives the following values for the coordinates:

N~k) = 0.5 and T(k) = 2~ (15)

It was shown by Charles and Wagstaff [7] that the behavior of a B 20 3- Si0 2 melt approxi-
mates to that of a regular solution, and the critical point in this system, which lies below the
liquidus line, has the coordinate N1 "" N2 "" 0.5. The parameter a, determining the stability
boundaries of a regular melt (at a given temperature), and the coordinates of the critical point
are not difficult to calculate from data on the course of the liquidus line.
In the case under consideration the following equation is obtained for the liquidus line:

(16)

where
T

A.H; = t::.H~ + ~ ~cpdT + 0:(1 - Ny;


TO

.1.H~ is the heat of fusion of pure component i at TO; .1.c p is the difference between its specific
heats in the liquid and solid states at the equilibrium temperature; 01(1- Ni)2 = Li is the relative
partial molar enthalpy of the component in the melt.
Equation (16) makes it possible to obtain an expression for calculating the interchange
energy:

(17)

where
M. M. SHUL'TS 27

Putting this expression for Q! into Eq. (15), we obtain


l' !J.n; (()T)
,( k) _ 21i; -- 2liT iiFi; p
(18)
1 - [1-N.(ilT)J'
2(1-~N;) 1+-1'-' ilN; p

The coordinates of any point lying on the liquidus line can be put into Eq. (18). If the melt
is indeed strictly regular and the critical point is not too far down, the liquidus line will have
an inflection with a coordinate close to 0.5 along the concentration axis. Taking this into con-
sideration and neglecting in the first approximation the temperature dependence of Llc p(i) , we
can immediately obtain from Eq. (18) the following simple formula for calculation:
o
t:.H; +- -
!J.C~(;)(T(8) - TO) (01' )(B)
T(B) - 2RT(8) ilN.
T(k) - P (19)
- 1 (ill' )(B)
1 + 21'(8) aN; p

where T(s) is the temperature corresponding to the inflection on the liquidus line. Equation (19)
can be used only for estimating the region where the critical phase lies. If the heat of solution
of the given component in the satu~ated melt is known, Q! and T(k) can be calculated more ac-
curately by substitution of the corresponding value into Eq. (16).
Methods for complete calculation of the binodal and spinodal for metastable melts from
data characterizing the course of the liquidus line can be found in the general and specialist
literature cited earlier.
The methods discussed above for calculating the coordinates of the critical point can also
be extended to systems with more complex dependences of chemical potential on concentration
than in the cases discussed above.
Literature Cited
1. J. W. Gibbs, Thermodynamics [Russian translation], GITTL, Moscow-Leningrad (1950).
2. J. D. van der Waals and P. Kohnstamm, Textbook of thermostatics [Russian translation],
ONTI, Moscow (1936).
3. A. V. Storonkin, Thermodynamics of Heterogeneous Systems, Parts 1 and 2, Izd. Leningr.
Gos. Univ., Leningrad (1967).
4. I. Prigogine and R. Defay, Chemical Thermodynamics [Russian translation], Izd. Nauka,
Sibirsk. Otd., Novosibirsk (1966).
5. J. W. Cahn and R. J. Charles, Phys. Chern. Glasses, 6:181 (1965).
6. R. J. Charles, J. Am. Ceram. Soc., 50:631 (1967).
7. R. J. Charles and F. E. Wagstaff, J. Am. Ceram. Soc., 51:16 (1968).
DISCUSSION
Shu I Its, replying to Galakhov, said that the inflection on the liquidus curve need
not necessarily correspond in composition to the critical point on the spinodal. Only
regular solutions were considered in this paper, while in more general cases a shift of
the critical point may occur if the chemical potential is taken accurately into account. In reply
to Appen, Shul'ts said that a steeper liquidus curve may be associated with a lower po-
sition of the immiscibility region. Replying to Dgebuadze, Shul Its emphasized that the
criterion for distinguishing stable and metastable states from the labile state is the ab-
solute instability of the labile state with respect to infinitesimal changes of composition.
The metastable state is resistant to such changes. The stable state is resistant not only to
infinitesimal but also to finite changes.
PRIMARY AND SECONDARY
PHASE SEPARATION IN GLASSES

E. A. Porai-Koshits and V. I. Aver'yanov

Several years ago, during a study of phase separation in glasses, we detected with the aid
of electron microscopy and low-angle x-ray scattering the possibility of formation of some kind
of highly disperse structure together with the main phases [1]. Three hypotheses have been put
forward on the origin of this "fine" structure: that it is fluctuational in character [2], that a
special kind of phase separation, described as "microphase separation," exists [3], and that it
is formed during quenching of the glass to room temperature, as the result of supersaturation
of the phases followed by secondary separation in it [4]. Somewhat later Tran Thach-Lan [5],
investigating phase separation in sodium silicate glasses, found that each phase is homogeneous;
i.e., he did not detect a "fine" structure. At the present time there are no doubts regarding the
existence of this structure or the causes of its formation [6, 7]. It is described as a "background
structure" in [7], but in our opinion it is more correct to speak of "secondary separation," as
this stresses its origin.
In this paper we report the results of a more detailed study of secondary separation with
the aid of an electron microscope.
The binodal curve of metastable immiscibility in sodium silicate glasses (Fig. 1) was
plotted as follows.
The right-hand branch ofthe curve (for compositions from 20 to 5 mole% Na20) was plotted
from the temperatures of phase separation (clearing temperatures) determined earlier [8]. The
left-hand branch was plotted with the aid of the "lever rule," according to which any glass (e.g.,
b) in the immiscibility region conforms to the expression

where dNa and d Si are the macroscopic densities of glasses richer in sodium oxide and silica,
respectively (in the present instance containing 1.3 and 17 mole% Na20), and w Na and wS i are
the volume fractions of these glasses, found by a linear method from the total areas of the
drops and matrix on electron micrographs of glass b. Thus, the points represented in Fig. 1 by
black circles were obtained with the aid of electron micrographs of glasses the compositions and
temperatures of which are indicated by crosses, and the left-hand branch of the binodal curve
was drawn.
It is possible with the aid of the binodal curve to calculate the absolute values of the rms
fluctuations of electron density (Ap)~ of any glasses in the immiSCibility region; these values

28
E . A. PORAI-KOSHITS AND V. I. AVER'YANOV 29

Fig. 1. Binodal curve (t z) en-


closing the region of phase sep-
aration in sodium silicate glass-
es. AB is the line of equal phase
volumes. The liquidus curve (t L )
is shown above.

4 8 12 18 20 24 28
N afJ. mole"lo

can also be determined experimentally from low-angle x-ray scattering intensity curves - with-
out any information on the compositions and volumes of the separated phases in the glasses.
Experimental values of (b.p)~ can also be determined in absolute units by comparison of the
intensities of the incident and scattered radiation or with the aid of a standard specimen [8, 9].
In Fig. 2 curves for the dependence of (b.p)2c on composition
__ of the glasses, calculated with
the aid of the binodal curve in Fig. 1, are compared with (b.p)~ curves, determined (as in [9]) in
absolute units [101. The small divergence between the top curves is easily explained as due to
the fact that the holding time of the specimens at a relatively low temperature was insufficient
for establishment of equilibrium: with a longer holding time the experimental points should lie
higher. The general agreement between the calculated and experimental curves, nevertheless,
is evidence of the complete correctness of the experimental curves obtained by the low-angle
method, and justifies the use of the binodal curve in Fig. 1 for assessment of the composition of
the phases formed by separation of glasses (the drops and matrix), and of their changes with
variation of heat treatment or of the overall composition of the glass.

8
C>;J

~I'"
-.....:.3
~ ~ 4
~
Fig. 2. Comparison of calculated and ex-
perimental curves for the dependence of I~
~ 2
(b..p)2 (in absolute units) on composition of
sodium silicate glasses at 580 (1), 660 (2),
685 (3), and 715 (4), and at t2:tz (5).
o 10
NafJ, mole"!o
30 E. A. PORAI-KOSHITS AND V. r. AVER'YANOV

The first striking feature is the a s v m met r v of the binodal curve, indicating unequal
temperature dependences of the mutual solubility of the components. The solubility of alkali
in silica (lef-hand branch of the curve) remains almost unchanged up to about 800, and then
increases gradually up to the critical temperature (860; glass containing about 7.5 mole%
Na20), at which it reaches its maximum value. The solubility of silica in alkali (right-hand
branch), on the other hand, increases continuously with temperature. This asymmetry of the
binodal should be borne in mind in examination of individual aspects of secondary separation.

Fig. 3. Secondary separation of the


alkali-rich drop phase in glasses with
7.5 mole% Na20: original (a), and
heated at 836 for 0.5 (b), 1 (c), 2 (d),
and 4 (e) h.
E. A. PORAI-KOSHITS AND V. 1. AVER'YANOV 31

Fig. 4. Secondary separation of the alkali-rich matrix in glasses with 12.5 mole% Na20, heated
for 2 h at 766 (a) and at 797 (b).

Secondary separation occurs in one of the phases formed by primary separation as the
result of supersaturation with silica or alkali on decrease of temperature, either during cooling
(quenching) of the glass, or during repeated heat treatment at lower temperatures; with normal
cooling secondary separation occurs more easily when the composition of the phase is above
the spinodal, which it cuts on decrease of temperature.
If the influence of viscosity on secondary separation is disregarded, owing to the asym-
metry of the immiscibility region (Fig. 1) the alkali-rich phase undergoes secondary separation
more easily, as its composition varies more strongly with temperature and supersaturation is
therefore attained more rapidly. This is clearly seen from a comparison of Figs. 3 and 4; in the
glass with 7.5 mole% Na20 the drops are richer in alkali (at 836 this composition lies a little
0

to the left of the line of equal volumes AB), while in the glass with 12.5 mole % Na20 the matrix
has the higher alkali content. *
The following interesting feature of the photographs shown in Fig. 3 should be noted: the
dimensions of the larger heterogeneity regions formed as the result of primary separation in-
crease with the time of heating at 836 but the fine droplets remain unchanged in size. Both
0
,

effects are understandable. The first is consistent with ordinary phase separation. Secondary
separation on the other hand, if it occurs in the phase the composition of which has reached the
binodal, should be independent of the time of heating at 836, as secondary drops are formed not
at this temperature but when the glass is cooled from 836.
At the same time, secondary separation should depend appreciably on the difference be-
tween the temperatures of the primary and secondary treatments, as this difference determines
the degree of supersaturation with one of the components of the phase in which secondary separa-
tion occurs. Figure 5 shows micrographs of glass specimens with 12.5 mole % Na20 which were
first heated at various temperatures in the immiscibility region (Le., had primary phases of
different compositions, see Fig. 1), after which they were all heated at the same lower temper-
ature, 668 , to intenSify secondary separation; the dimensions of the secondary drops (in the
glass matrix) and their total volume increase appreciably with increasing difference between the
temperatures of primary and secondary heating, as in this case supersaturation of the primary
phase with alkali becomes greater.

*The technique of preparing glass specimens for electron microscopy is described in other
publications [4,11].
Fig. 5. Secondary separation in glasses with 12.5 mole%Na20 heated for 2 hat 766 (a), 777 (c), 809
(e), and 816 (g) (t siG o > t z) and then at 668 for 2 h (b, d, f, h).
0 0
E. A. PORAI-KOSHITS AND V. I. AVER'YANOV 33
34 E. A. PORAI-KOSHITS AND V. I. AVER'YANOV

As regards the silica-rich phase, owing to its higher viscosity and lower supersaturation
it can undergo secondary separation in the course of normal cooling only after treatment of the
glass at a sufficiently high temperature or as the result of repeated heat treatment, which can
also intensify separation if it has already occurred. These cases are demonstrated in Fig. 6.
Here the volume of the separated phase is small, which corresponds to the form of the left-hand
branch of the immiscibility curve (Fig. 1).
Thus, the properties of the heterogeneity regions arising during cooling or heat treatment
within primarily separated phases are determined entirely by the form of the binodal curve, and
therefore they are of the same phase-separation nature as the phases formed by primary separa-
tion.
During repeated low-temperature treatment of a separated glass, gradual clearing of one
of the phases from droplets of secondary origin may occur, by transfer into the second phase,
which has the same composition as the secondary droplets. Gradual clearing of the alkali-rich
phase from fine silica-rich droplets is demonstrated in Fig. 7.
We note in conclusion that secondary separation in glasses is a general phenomenon. It has
been observed by us in lithium silicate glasses (Fig. 8), and by Galakhov in alkaline-earth sili-
cate glasses [121. It should also be taken into consideration in studies of glasses containing more
than two components. It seems likely in this connection that certain sodium borosilicate glasses,

Fig. 7. Glass with 7.5 mole% Na20. Droplets of secondary origin gradually disappear with
increasing duration of additional heat treatment. a) Primary heat treatment at 849 for 4 h;
0

b-d) second heat treatment at 720 for 5 (b), 10 (c), and 60 (d) min.
0
E. A. PORAI-KOSHITS AND V. I. AVER'YANOV 35

Fig. 8. Secondary separation in the matrix of glass


containing 23 mole% Li2 0 and 77 mole% Si02 , heat-
ed at 720 and then at a lower temperature .
0

which according to many investigators have three-phase structure, are in reality two-phase
glasses, as the properties of the fine droplets taken as the thi rd phase are very similar to the
properties of droplets of secondary origin in sodium silicate glasses [13].
Finally. if a glass which has undergone phase separation at a certain temperature tl is
subjected to stepwise heat treatment at progressively lower temperatures it is possible to in-
duce phase separation of successively higher orders: secondary at temperature t2 < tl> then at
t3 < t2 third-order separation within the phase formed by secondary separation, etc. These
processes are fully controllable, so that it becomes possible to produce glasses having very
diverse structures.
The question of internal structure of individual phases and of the structure of glasses the
compositions of which lie outside the binodal curve was not considered in this paper, although
there are experimental data indicating that they have heterogeneous structure.

Literature Cited
1. V. I. Aver'yanov (Averjanov), N. S. Andreev, and E. A. Porai-Koshits, in: Physics of
Noncrystalline Solids, Amsterdam (1965), p. 580; E. A. Porai-Koshits, D. A. Goganov, and
V. I. Aver'yanov (Averjanov), ibid., p. 117.
2. V. N. Filipovich, in: Structural Transformations in Glasses at High Temperatures, Izd.
Nauka, Moscow- Leningrad (1965), p. 49.
3. F. Ya. Galakhov and S. Ya. Konovalova, Dokl. Akad. Nauk SSSR, 155:122 (1964).
4. v. I. Aver'yanov (Averjanov), N. S. Andreev, and E. A. Porai-Koshits, in: Physics of
Noncrystalline Solids, Amsterdam (1965), p. 580; V. I. Aver'yanov and E. A. Porai-Koshits,
in: Structural Transformations in Glasses at High Temperatures, Izd. Nauka, Moscow-
Leningrad (1965), p. 76.
5. Tran Thach-Lan, Verres Refractaires, 20:8 (1966).
6. D. A. Goganov and E. A. Porai-Koshits, Dokl. Akad. Nauk SSSR, 165:1037 (1965).
7. Y. Moriya, D. H. Warrington, and R. W. Douglas, Phys. Chem. Glasses, 8:19 (1967).
8. N. S. Andreev, D. A. Goganov, E. A. Porai-Koshits, and Yu. G. Sokolov, in: The Glassy
State, Vol. 3, No.1, Catalyzed Crystallization of Glass, Izd. Akad. Nauk SSSR, Moscow-
Leningrad (1963), p. 46 [English translation: The Structure of Glass, Vol. 3, Consultants
Bureau, New York (1964), p. 47J.
9. N. S. Andreev and T. I. Ershova, Dokl. Akad . Nauk SSSR, 172:1299 (1967).
36 E. A. PORAI-KOSHITS AND V. I. AVER'YANOV

10. D. A. Goganov and E. A. Porai-Koshits, Dokl. Akad. Nauk SSSR, 167:1266 (1966).
11. N. S. Andreev, V. I. Aver'yanov, and E. A. Porai-Koshits, in: Structural Transformations
in Glasses at High Temperatures, Izd. Nauka, Moscow- Leningrad (1965), p. 59.
12. F. Ya. Galakhov, in: Structural Transformations in Glasses at High Temperatures, Izd.
Nauka, Moscow- Leningrad (1965), p. 110.
13. V. P. Klyuev, G. P. Roskova, and V. I. Aver'yanova, this collection, p. 74.

DISCUSSION
Po r a i - K 0 s hit s , replying to Mazurin' s question regarding the nature of three-
phase structure in sodium borosilicate glasses, said that the occurrence of secondary
separation should be presumed in this case. The drop size is of the order of some hundreds of
angstroms and greater in primary separation, and about 100 A and less in secondary separation.
INVESTIGATION OF THE PHYSICAL AND
CHEMICAL PROPERTIES OF GLASSES UNDERGOING
PHASE SEPARATION AS A METHOD OF
STUDYING THEIR STRUCTURE

O. V. Mazurin

Greatly intensified attention to processes of metastable liquid-liquid phase separation in


glasses is a fundamental feature of the present period of research in the theory of the glassy
state. It is important to stress that Soviet scientists were the best prepared for this period,
owing both to the exceptional attention devoted by Grebenshchikov and his successors to
studies of heterogeneous borosilicate glasses [1-4] and to comprehensive analysis of various
aspects of the microheterogeneous structure of glasses in the course of lively discussions be-
tween supporters of various hypotheses of glass structure [5].
Of course, even detailed study of the phase-separation structure of glasses would contrib-
ute little to understanding of the structure of single-phase glasses, i.e., glasses outside the
region of metastable phase separation or within homogeneous portions of a glass which has un-
dergone phase separation. Nevertheless the special interest in phase-separation phenomena
in glasses, which strongly distracts investigators from the problem of structure of single-phase
glasses, appears to be fully justified. The technical means which have now become available
must be utilized for obtaining reliable information on the structure produced in glasses as the
result of microphase separation. For elucidation of many aspects of the structure of single-
phase glasses further improvements of experimental technique must be awaited.
The time has now come to develop on no less a scale investigations of the properties of
phase-separating glasses and their relation to structure, simultaneously with further studies of
their structure. This will enable us to use the results of structure studies for predicting the
influence of heat treatment on the properties of glasses, including some industrial glasses and to
develop with confidence new compositions and new heat-treatment conditions for production of
industrial glasses having improved properties. On the other hand, study of the properties of
glasses undergoing phase separation will provide additional information on their structure, in-
cluding data which cannot be at present obtained by direct methods of investigation.
The following brief review of various possibilities of study of the relation between the
properties and structure of two-phase glasses is based both on literature data and on the results
of work carried out in the Laboratory of Glass Chemistry of the Institute of Silicate Chemistry.
Interpretation of the properties of two-phase glasses in the work of this laboratory was based
on the assumption that these properties are unique functions of the properties and character of

37
38 0. V. MAZURIN

Quenched glasses

Vi?
a b
100 100
80 80
;~
~
<:::;,"'>
80 ~
c:g>
80 3
~ 40 ~40
<t:l Q;j

20 20
0 0
4- 1 Z 3 4
Nap, mole"!o Nap, mole"!o

Annealed glasses
b Fig. 1. Influence of Na20 content in glasses
100 of the system Na20-B20a-Si02 on the amount
80 of B20 S dissolved during treatment with 3N
~ 80 Hel at 50 [9]. a) In 1 h; b) in 5 days. 1) 60%
0

~ Si02; 2) 65% Si02; 3) 70% Si02.


~ 40
<t:l
20
o L----'--=_""---'-
.20o
L-....J.----L--='------'-
1 2 3 4- 1 2 3 4
Nap, mole"!o Nap, mole"lo

distribution of each of the phases in the glass. The possible influence of intermediate layers
and the specific peculiarities of the highly disperse state must be disregarded at present. This
approach, which may be described as mechanical, differs greatly from the thermodynamic ap-
proach to the problem being developed, e.g., by Nemilov [6],
All the properties of two-phase glasses can be divided into two groups: properties the
values of which depend only on the properties and volume ratio of the two phases, and properties
the values of which depend also on the character of phase distribution, i.e., on whether one of
the phases, and which particular one, forms closed inclusions within the other.
We will first consider the second group of properties, as the more important. The inves-
tigations were concerned predominantly with low-alkali borosilicate glasses. It should be noted
that such glasses are ideal objects for solving the problems under discussion. Many of them
exhibit intense phase separation and at the same time do not crystallize at all. Among a number
of studies of the chemical stability of such glasses [7, 8], special interest attaches to the work
of Zhdanov and Koromal'di [9] (Fig. 1). According to Zhdanov flO] the high stability of

4.0

......
Fig. 2. Temperature and frequency de-
0
. 3.0
'<:l
pendence of dielectric losses of glass of ..,00

the approximate composition (mole%): B


00

2Na20, 24B 20 a, 74Si~. (Heat treatment


at 550 10 h [15].) 1) 10 4 Hz; 2) 10 3 Hz;
0 ,

2.0
3) 4.10 2 Hz.
13 15 19 21
0. V. MAZURIN 39

glasses with 2-3% Na20 is due to segregation of the readily soluble alkali borate phase
into drops enclosed in a much more resistant silica matrix. These data, supplemented by
studies of the adsorption characteristics of the resultant porous glasses, enabled Zhdanov to
develop a detailed scheme representing the structure of such glasses, and thereby to utilize
for the first time information on the properties of glasses described subsequently as separating
for elucidation of their structural characteristics. The basic principles of Zhdanov's approach
are indisputable: if the chemically unstable phase forms closed inclusions, the solubility of the
glass is determined by the more stable phase; if the unstable phase is continuous the glass will
dissolve rapidly until all the unstable phase has been destroyed.
The next stage was study of the electrical properties of separating glasses in relation to
their structure. According to Maxwell's theory of heterogeneous dielectrics [11], a material con-
taining conducting inclusions enclosed in a medium of lower conductivity should give character-
istic maxima on the frequency-temperature dependences of dielectric losses. The idea that
this theory could be applied to analysis of dielectric-loss curves for heterogeneous glasses was
put forward by the present author in 1962 [12]. The first of a series of papers by Charles
[13] on the use of this method for studying structural characteristics of lithium silicate glasses
appeared a year later, followed by Leko's work [14] on the sodium silicate system and a
paper by Dgebuadze and Mazurin [15] on alkali borosilicate systems. A characteristic
example is shown in Fig. 2. It is important that such results not only demonstrate con-
clusively the presence of closed conducting inclusions in the glass but also make it possible to
determine from the temperature- frequency shift of the relaxation maximum the activation
energy for conduction of the conducting phase, to supplement the activation energy for conduc-
tion of the matrix which is obtained from data on conductivity of the two-phase glass. In the
future we must also learn how to utilize information on the position, form, and magnitude of the
relaxation maxima, which should enable us to assess the form and probably the dimensions of
the closed regions. By comparing the curves of temperature-frequency dependence of dielec-
tric losses for single-phase [16, 17] and two-phase glasses we can probably determine the "de-
gree of closure" of the conducting phase in cases where the conducting inclusions are partially
interconnected by relatively thin layers.
The theory of a heterogeneous rheological body, which differs little in principle from the
theory of heterogeneous dielectrics, can be applied to analysis of the temperature and frequency
dependence of internal friction of two-phase glasses. Here the existence of a highly viscous
continuous framework is an essential condition for the occurrence of a relaxation maximum.
Figure 3 shows the results obtained by the author on internal friction of a glass having the com-
position (mole% by analysis): 3.9 Na20, 31.2 B20 3, 64.9 Si0 2. The determinations were carried
out by the torsion pendulum method with the apparatus devised by the Department of Glass Tech-
nology, University of Sheffield [18]. Very large values of the logarithmic decrement can be
measured with the aid of this apparatus; this is absolutely essential for solution of the present
problem. Apparatus of the type described by Coenen and Amrhein [19] can also be used for
this purpose.
Like electrical measurements, measurements of internal friction permit separate deter-
mination of the temperature coefficients of viscous flow of the low-melting phase (from the fre-
quency- temperature shift of the relaxation maximum) and of the high-melting phase (from di-
rect measurements of the temperature dependence of viscosity of the two-phase glass). Since
the high-melting phase in glasses generally has the lower conductivity, glasses with closed con-
ducting inclusions generally have highly viscous frameworks. In such cases there is the pos-
sibility of independent determination of the temperature coefficients of conductivity and viscosity
of each of the phases, and correspondingly of independent determination of their chemical com-
positions in the case of tw~phase glasses in three-component systems. This permits, in par-
ticular, independent determination of tie lines in such systems. It should be noted that if the
40 O. V. MAZURIN

1/T'10 4
12 13 14 15 16

0.3

o Quenched glass 0.0


Heat treatment
600,60 h

Fig. 3. Temperature dependence of the


logarithmic decrement (log d) and the os-
cillation frequency of the torsion balance
0.0 r-----r---, 1.0 (f) for a sodium borosilicate glass exhib-
iting phase separation. Bottom left,
temperature and frequency dependences
of log d for glass heated at 600 for 60
0

h. The f values are given for tempera-


90.13
tures corresponding to internal-friction
00.06 maxima.
1.~3~.0--f.~15--~~~0-----~
1/T'IO"

immiscibility region and composition- properties relationships for one-phase glasses in a three-
component system are known, it is sufficient to know two properties of one phase and one prop-
erty of the other for determination of tie lines and hence the compositions of both phases.
It is known that, in general, the temperature and time dependences of delayed elastic de-
formation of any rheological body contain the same information as the temperature and frequen-
cy dependences of internal friction. At the same time, for a number of reasons investigation of
internal friction is greatly preferable to investigation of delayed elastic deformation.
Up to now we have discussed studies of the structure of a glass which remains unchanged
or changes very little during the experiment. However, in a number of cases, because of the
high sensitivity of viscosity to changes of glass structure and because viscosity can be measured
easily in the region of intense phase-separation processes, data on variations of the viscosity of
glasses in the course of phase separation may provide interesting information on changes in the
glass structure during the separation process [20].
The properties of two-phase glasses dependent only on the composition and volume of the
phases and not on their distribution in the glass are also very interesting. The most important
of these properties at present is the glass transition temperature. The glass transition tem-
perature of each phase should appear on the dilatometric curve, quite independently of whether
the given phase forms a continuous framework or is present in the glass in the form of closed
inclusions. Thus, in principle the glass transition temperatures of both phases should be obser-
ved on the dilatometric curve. However, in practice for a number of reasons only tg of the low-
melting phase can be found from the dilatometric curve. Nevertheless, even this possibility in
itself provides very interesting information on heterogeneous glasses [21].
The DT A method is very promising. Several publications on DTA of heterogeneous glasses
[8, 22, 23] regrettably contain random and often contradictory data. The most definite results
are reported by KUhne and Skatulla (Fig. 4). It is to be hoped that it will be possible in
the future to use DTA curves for reliable determination of tg of each phase in two-phase glasses
from the initial temperatures of each of two endotherms. A somewhat puzzling fact is that the
0. V. MAZURIN 41

At t a
I
!!
01---------1....

b
Fig. 4. DTA curves of various
glasses. a) Glass of the system 01------1...
~O-RO-SiO:!, normally anneal-
ed; b) glass of the system ~O
RO-B2~-SiO:!, normally an- 01------1..
nealed; c) glass of the composi-
tion (mole%): 5.6 Na20, 30 B2~' d
64.4 Si02, quenched; d) the same
glass, after extensive separation. 01---,

300 400 500 800 700 800 900tO

bottom curves in Fig. 4 have three endotherms. The possibility is not ruled out that here we
have a third phase, appearing as the result of secondary separation processes.
A very attractive possibility is study of phase composition in a two-phase glass by anal-
ysis of spectral absorption curves in cases where the absorption spectra of the main glass com-
ponents or coloring additives are greatly changed by variations of the glass composition in the
given system. The possibility of such investigations in principle has been recently demonstrated
by Bartenev, Gorbachev, and Udovenko [24]. However, their publication contains a number
of obscure points, and a large amount of experimental material in this field must be accumu-
lated for wide use of this method.
Here we have considered not all but only the most obvious instances of relationships be-
tween the properties and structure of two-phase glasses. However, it is obvious that compre-
hensive study of at least some of the properties discussed above will in future provide reliable
information on the distribution of both phases in two-phase glasses and on the composition of
these phases and will also make it possible to study variations of the structure of such glasses
during heat treatment. It is necessary to determine the minimum amounts of phases necessary
for demonstrating the presence of two phases in glass by study of its properties. The possibil-
ity is not excluded that these amounts may be considerably smaller than can be recorded at the
present time by most direct methods of investigation. Then, apart from anything else, we will
have an essentially new possibility for detecting and studying the initial stages of phase separa-
tion. Finally, there is no doubt that detailed study of the properties of two-phase glasses will
make it possible in the future to obtain valuable information on the degree of heterogeneity of
Single-phase glasses from data on their properties.
Literature Cited
1. I. V. Grebenshchikov and 0. S. Molchanova, Zh. Obshch. Khim., 12:588 (1942); 0. s. Mol-
chanova and M. V. Serebryakova, Tr. Gos. Optich. lnst., 23(141}:3 (1953).
2. E. A, Porai-Koshits, Dokl. Akad. Nauk SSSR, 36:285 (1942); Zh. Obshch. Khim., 12:196
(1942); in: Structure of Glass, lzd. Akad. Nauk SSSR, Moscow- Leningrad (1955), p. 145
[English translation: The Structure of Glass, Vol. 1, Consultants Bureau, New York
(1958), p. 112].
3. S. P. Zhdanov, Structure of Porous Glasses and Structural Transformations in Sodium
Borosilicate Glasses, Author's abstract of doctoral dissertation, Leningrad (1959).
42 O. V. MAZURIN

4. D. P. Dobychin, in: The Glassy State, Proceedings of the Third All-Union Conference,
Izd. Akad. Nauk SSSR, Moscow- Leningrad (1960), p. 480; in: Structure of Glass, Izd.
Akad. Nauk SSSR, Moscow- Leningrad (1955), p. 176 [English translation: The Structure
of Glass, Vol. 1, Consultants Bureau, New York (1958), p. 135].
5. Structure of Glass, Izd. Akad. Nauk SSSR, Moscow- Leningrad (1955) [English translation:
The Structure of Glass, Vol. 1, Consultants Bureau, New York (1958)]; The Glassy State,
Proceedings of the Third All-Union Conference, Izd. Akad. Nauk SSSR, Moscow- Lenin-
grad (1960) [English translation: The Structure of Glass, Vol. 3, Consultants Bureau,
New York (1964)].
6. s. V. Nemilov, Neorgan. Mater., 3:679 (1967).
7. O. S. Molchanova, in: Structure of Glass, Izd. Akad. Nauk SSSR, Moscow- Leningrad
(1955), p. 141 [English translation: The Structure of Glass, Vol. 1, Consultants Bureau,
New York (1958), p. 109].
8. K. KUhne and W. Skatulla, Silikattech., 10:105 (1959).
9. s. P. Zhdanov and E. V. Koromal'di, Izv, Akad. Nauk SSSR, otd. Khim. Nauk, No.4, p.626
(1959).
10. s. P. Zhdanov, Izv. Akad. Nauk SSSR, Otd. Khim. Nauk, No.6, p. 1011 (1959).
11. J. C. Maxwell, Treatise on Electricity and Magnetism, London (1904),
12. O. V. Mazurin, Electrical Properties of Glass, Khimizdat, Leningrad (1960), p. 110.
13. R. J. Charles, J. Am. Ceram. Soc., 46:235 (1963).
14. V. K. Leko, in: The Glassy State, Proceedings of the Fourth All-Union Conference, Izd.
Nauka, Moscow- Leningrad (1965), p. 280 [English translation: The Structure of Glass,
Vol. 7, Consultants Bureau, New York (1966), p. 96].
15. T. P. Dgebuadze and O. V. Mazurin, Neorgan. Mater., 2:1328 (1966); 3:1236 (1967).
16. J. Isard, Proc. Inst. Electr. Eng., Vol. 109B, Suppl. 22, p. 440 (1962).
17. O. V. Mazurin, Influence of Composition and Temperature on Electrical Conductivity and
Migration Dielectric Losses in Solid Glasses, Author's abstract of doctoral dissertation,
Leningrad (1962); O. V. Mazurin and V. B. Brailovskii, in: Chemistry of the Solid, Izd.
Leningr. Gos. Univ., Leningrad (1965), p. 181.
18. R. W. Douglas, P. J. Duke, and O. V. Mazurin, Phys. Chem. Glasses, 9:169 (1968).
19. M. Coenen and E. M. Amrhein, Compt. Rend. Symp. Resistance Mecan. Vevre, Moyens
Ameliorer, Charleroi (1962), p. 529.
20. V. P. Klyuev, G. P. Roskova, and V. I. Aver'yanov, this volume, p. 74. ,
21. M. V. Strel'tsina, O. V. Mazurin, and A. S. Totesh, this volume, p. 87.
22. J. Broukal, Slikaty, 3:14 (1959).
23. K. KUhne, Silikattech., 11:106 (1960).
24. G. M. Bartenev, A. A. Gorbachev, and N. G. Udovenko, Dokl. Akad. Nauk SSSR, 172:1113
(1967).
DISCUSSION
M a z uri n, replying to Borisov's question of the possibility of existence of two-
phase systems above the binodal curve, said that a heterogeneous structure but not two-phase
separation is possible here.
INFLUENCE OF STRUCTURAL HETEROGENEITY ON
PHYSICAL AND CHEMICAL PROPERTIES OF GLASSES

N. M. Bobkova and I. A. Trunets

The concept of glass as a heterogeneous system has developed greatly in recent years.
The hypothesis of chemically heterogeneous structure of Porai-Koshits [1], Vogel's con-
cept of cellular structure [2], the micellar hypothesis of Moriya [3], and the microcellular
structure hypothesis of Yoshida [4] have been developed on these lines. Microheterogeneity
regions of various kinds have been detected in glasses of both simple and complex compositions
during investigations by electron microscopy and x-ray diffraction. It is important to note that
microheterogeneity is detected in silicate glasses of almost any composition. According to
some investigators silicate glasses of all types, two-, three-, and multicomponent are not ho-
mogeneous but consist of a framework and numerous microheterogeneity regions [5J.
In most cases microheterogeneity of glass is attributed to phase-separation phenomena.
This view is justified in cases where the glass compositions lie in the region of stable or
metastable phase separation of the corresponding phase diagrams. In other cases the origin or
nature of microheterogeneity regions in glass remain unresolved.
Our electron-microscopic study of silicate glasses of various compositions also revealed
a microheterogeneous structure in most of them. The investigation included certain alkali
glasses, among them commercial window glass, the compositions of which do not lie in the im-
miscibility region. Figure 1 shows electron micrographs of quenched sodium calcium silicate
glass of the composition (wt. %): Si0 2 72, CaO 5, and Na20 23, and of commercial window glass.
It has also been shown that the degree of microheterogeneity in glasses depends on the condi-
tions of melt formation [6J. Therefore formation of structurally independent microheterogene-
ity regions bounded by interfaces (i.e., effectively regions of phase separation) in silicate glasses,
observed also in glasses which do not lie in the immiscibility region, cannot be regarded as
a random phenomenon.
In order to understand the nature of the microheterogeneity regions arising in a silicate
melt, let us examine the course of changes in its structure at temperatures above the liquidus.
The longest process during glass formation in silicate systems is dissolution of residual silica.
Since silicate glass melts which have passed through the homogenization and fining stage are
obtained at temperatures conSiderably above the liquidus, the silica lattice is eVidently destroy-
ed almost entirely, especially owing to the active influence of such mobile ions as Ca 2+ and Na+
on the lattice destruction process [7J. Therefore the microheterogeneity regions observed in
complex silicate glasses are probably unconnected with residues of the crystalline structure of
the original materials, as is confirmed by the results of electron diffraction and infrared spec-
troscopy, in contrast to the one-component quartz glass, in which residues of the quartz struc-

43
44 N. M. BOBKOVA AND I. A. TRUNETS

Fig. 1. Structure of sodium calcium silicate (a) and


window (b) glasses.

ture have been detected even in glass formed at 1900 [8]. Nevertheless, the low rate of de-
struction of the silica lattice must have an influence on the structure of silicate glasses. Owing
to the kinetic characteristics of silica dissolution, complex anionic groups of the (SixOr) type
of a very high degree of polymerization will persist in the melt for a long time. However, the
presence of high-silica regions does not in itself explain formation on microheterogeneity re-
gions with distinct interfaces.
Numerous investigations of reactions between silica and sodium oxide (or carbonate) have
shown that the primary reaction product is sodium metasilicate, Na2Si03, regardless of the
initial component ratio [9, 10]. Formation of metasilicate simultaneously with dissolution of
residual silica has also been established in sodium silicate melts [11]. Presence of groups
close in composition to sodium metasilicate was detected by Florinskaya in all sodium
silicate glasses containing from 17 to 42% Na20 [12]. The meta silicate initially formed only
gradually passes into more complex silicates [13, 14]. Because of the high viscosity of the
melt and the low mobility of silicate anions in it, equilibrium in silicate systems is reached
with great difficulty. Therefore after destruction of the residual silica lattice sodium silicate
glass melts simultaneously contain complex anionic groups of the SixOr with a very high de-
gree of polymerization of the tetrahedra, and polymerized (SiO~-)n' anions corresponding to the
meta silicate, as the result of the kinetic peculiarities of silica dissolution.
According to the phase diagram of the system Na 20- Si0 2, such anionic complexes are
mutually incompatible in presence of sodium ions, as they are separated by the region of stabil-
ity of the (Si20~-)n anion. Therefore with simultaneous presence of (SiO~-)n and SixOr anions
in the melt sites of phase separation should arise in sodium systems. The degree of phase sep-
aration will increase during gradual formation of (Si 20g-)n anions corresponding to the disili-
cate, with decrease of the amount of (SiO~-)n anions, and when diffusion and relaxation process-
es are complete phase separation will not occur in such a melt.
In the calcium silicate system initial formation of calcium orthosilicate, Ca 2Si0 4, and
rankinite, Ca3Si207 [8] is kinetically determined; this leads to appearance of (Si 20r)n and SiOt
anionic groups which are also, according to the phase diagram of the system CaO- Si02 incom-
patible with SixOr anionic complexes. In particular, calcium metasilicate is formed only from
the orthosilicate and rankinite. Therefore nonequilibrium phase separation is also possible in
N. M. BOBKOVA AND I. A. TRUNETS 45

presence of calcium ions in the silicate system. This applies even more so to the system Na20-
CaO- Si02, in which initial formation of sodium metasilicate [S] followed by triple metasilicates
[15, 16], has been demonstrated.
A similar analysis can be performed for a number of other systems.
In potassium and lithium silicate systems metasilicates are initially formed [S]; the stabil-
ity region of their groups is separated from the high-silica region by the disilicate field. How-
ever, the tendency to compatibility of anionic silicon-oxygen groups should increase with in-
crease of the cation radius, as this is accompanied by increase of the coordination number of
the cation. According to Appen's data [17] the coordination numbers of lithium, sodium,
and potassium in silicate glasses are 4, 6, and 9, respectively. According to Belov [IS]
the degree of polymerization of the silicon- oxygen groups associated with the metal cations in-
creases with increasing size and coordination number of the cation. Therefore the silicate
anions (SiO~-) n should increase in size as the result of polymerization in the sequence from
lithium to potassium glasses, becoming more compatible with the high-silica complexes. It is
to be expected from these considerations that under otherwise the same conditions potassium
glasses will have more homogeneous structure than lithium glasses. Evidently, cesium glasses
have the lowest tendency to nonequilibrium phase separation among the alkali glasses.
Therefore in our opinion the microheterogeneous structure of silicate glasses which do
not undergo liquid-liquid phase separation is due to the kinetic characteristics of silica dis-
solution and to the thermodynamic conditions for formation of particular silicates by reactions
between silica and the metal oxide. Naturally, the degree of microheterogeneity of silicate
glasses must be determined not only by the position of the corresponding point on the phase
diagram but also by the time and temperature conditions of melt formation. As we have dem-
onstrated experimentally [6], a structure without any signs of phase separation can be obtained
at the suitable temperatures and with the appropriate melting times. However, it should be
taken into account that complete mixing is not achieved in silicate systems. Only structurally
independent microheterogeneity will be absent.
Variations of the structural homogeneity of glasses have an appreciable influence on such
properties as the tendency to crystallization, viscosity, microhardness, elasticity modulus,
mechanical strength, and chemical stability.
Figure 2 represents the dependence of certain properties of sodium calcium silicate and
window glasses on the final temperature of melt formation (with a holding time of 1 h at the
final temperature). The structural heterogeneity of both glasses gradually diminishes with in-

% c

/ZI /2
a b 0.6

~z
N 7000 0.5
S NS 550
{,O500 {, 0.4
500
~1
~ ~
'-'.J~ 6000 ~
0.3
450
(300 1400 1500 f800 t" 1300 14001500 f800 to 1400 1500 f600 to
Fig. 2. Variation of properties of glasses with increase of the melting
temperature. a) Variation of elasticity modulus (E); b) variation of
microhardness (H); c) variation of chemical stability (% weight loss on
boiling in water for 1 h). 1) Sodium calcium silicate glass; 2) window
glass.
46 N. M. BOBKOVA AND I. A. TRUNETS

crease of the melting temperature, and specimens melted at 1600 show no signs of separation
when examined under the electron microscope. The lower the degree of glass microheteroge-
neity due to the possibility of nonequilibrium phase separation the more stable is the glassy
state and the higher are the mechanical strength, elasticity modulus, and chemical stability.
Literature Cited
1. E. A. Porai-Koshits, in: The Glassy State, Proceedings of the Third All-Union Conference,
Izd. Akad. Nauk SSSR, Moscow- Leningrad (1960), p. 14 [English translation: The Struc-
ture of Glass, Vol. 2, Consultants Bureau, New York (1960), p. 9J.
2. w. Vogel, in: The Glassy State, Proceedings of the Third All-Union Conference, Izd.
Akad. Nauk SSSR, Moscow- Leningrad (1960), p. 24 [English translation; The Structure
of Glass, Vol. 2, Consultants Bureau, New York (1960), p. 17J.
3. T. Moriya, J. Ceram. Assoc. Japan, 55:60 (1957).
4. U. Yoshida, X-Ray (Japan), 2:66 (1941); 3:17 (1942).
5. F. K. Aleinikov, R. P. Paulavichyus, and V. A. Slizhis, Tr. Akad. Nauk Lit. SSR, Ser. B,
No.2 (29), p. 69 (1962).
6. N. M. Bobkova and V. V. Rudakov, Steklo i Keram., No.6, 11 (1967).
7. O. K. Botvinkin and E. M. Shpil 'kov, Izv. Akad. Nauk Kaz. SSR, No.3 (14), p. 86 (1957).
8. v. A. Florinskaya and R. S. Pechenkina, in: Structure of Glass, Izd. Akad. Nauk SSSR,
Moscow- Leningrad (1955), p. 70 [English translation: The Structure of Glass, Vol. 1,
Consultants Bureau, New York (1958), p. 55J.
9. P. P. Budnikov and A. M. Ginstling, Reactions in Mixtures of Solids, Gosstroiizdat, Mos-
cow (1961).
10. v. I. Babushkin, G. M. Matveev, and O. P. Mchedlov-Petrosyan, Thermodynamics of
Silicates, Gosstroiizdat, Moscow (1965).
11. F. Ya. Kharitonov and L. G. Mel 'nichenko, Steklo i Keram., No.7, p. 5 (1963).
12. v. A. Florinskaya and R. S. Pechenkina, in: The Glassy State, Proceedings of the Third
All-Union Conference, Izd. Akad. Nauk SSSR, Moscow- Leningrad (1960), p. 156 [English
translation: The Structure of Glass, Vol. 2, Consultants Bureau, New York (1960), p. 135J.
13. E. P. Danil 'chenko, Kinetics of Formation of Soda and Sulfate Glasses, Candidate's dis-
sertation, Vses. Nauchn.-Issled. Inst. Stekla, Moscow (1952).
14. E. P. Danil'chenko, Dokl. Akad. Nauk SSSR, 36:1175 (1952).
15. C. Kroger and G. Ziegler, Glastech. Ber., 26:346 (1953).
16. C. Kroger, Glastech. Ber., 30:42 (1957).
17. A. A. Appen, V. B. Glushkova, and S. S. Kayalova, Neorgan. Mater., 1:576 (1965).
18. N. V. Belov, Crystal Chemistry of Silicates with Large Cations, Izd. Akad. Nauk SSSR,
Moscow- Leningrad (1961).
DISCUSSION
Replying to Aslanova, Bobkova noted that the elasticity modulus increases when
the glass becomes more homogeneous. In her replies to questions by Tykachinskii,
Aver'yanov, Varshal, Podushko, and others Bobkova emphasized that no significant changes
of composition were observed during high-temperature treatment; the changes did not ex-
ceed 0.3%. Changes of properties were caused mainly by variations of structure and not
of composition, since the influence of high temperatures on certain properties (such as den-
sity) was the opposite to what could be attributed to volatilization of sodium oxide. Moreover,
similar relationships are found for glasses of the calcium aluminosilicate system, which showed
no deviations of chemical composition. Experiments with increasing melting times (up to 6 h)
at a series of increasing temperatures (up to 1600) showed that temperature has a much more
marked influence than time on the structure. Bobkova noted further that the change in
the amount of heterogeneity regions is due to blurring of their boundaries, and is a gradual
rather than an abrupt process. It follows from the data presented in the paper that glasses of
N. M. BOBKOVA AND I. A. TRUNETS 47

many compositions can be obtained in practice without phase separation. However, a compara-
tive study of glasses in a system known to exhibit phase separation showed that they have heter-
ogeneous structure even after 6 h at 1600. In this case in presence of heterogeneity regions
lowering of the treatment temperature leads to further development of these regions and to
crystallization.
F iii p 0 vic h noted the novelty of the experimental facts reported by Bobkova and
Trunets. He said that estimates of the rate of dissolution of the solid phase in the melt
show that this rate is very low. The existence of a heterogeneous structure at high tem-
peratures may be attributed both to phase separation and to the low rate of dispersal of the
heterogeneity regions. This choice can only be resolved experimentally. In his opinion, high-
temperature homogenization of glass should influence the kinetics of liquid-liquid phase separa-
tion processes insofar as it influences, as is known, many processes including the crystallization
rate. It would be interesting to study experimentally the relation between the degree of homoge-
nization and the rate of phase separation.
Bob k 0 va on the whole agreed with Filipovich' s comments and said that the ob-
served heterogeneity should be assigned to the liquid type and is not necessarily attrib-
utable to the occurrence of stable or metastable phase separation. She added that high-temper-
ature homogenization of glass has also been achieved by other investigators.
T y k a chi n ski i emphasized the practical interest of the paper by Bobkova and
Trunets in relation to the variability of mechanical properties of certain commercial
glasses and to searches for methods of improving their quality. He added that in the light of the
material presented by the speakers liquid-liquid phase separation cannot be regarded as the
only cause of heterogeneity in glass. Research in this field is very promising. He therefore
proposed that a special conference should be held on methodological aspects of these problems.
DETERMINA TION OF THE NUCLEATION RATE OF
NEW PHASES IN GLASSES AT THE
INITIAL STAGES OF PHASE CONVERSION

V. N. Filipovich and A. M. Kalinina

In this paper we examine the problem of determination of the nucleation rate I and the
growth rate c of spherical particles of a new phase (glassy or crystalline) in glass from the
experimentally found distribution of circular sections of the spheres in the plane of a section of
a heterogeneous specimen. Suppose that phase conversion in glass (at a definite temperature)
is being studied from an instant T = O. The number p(R, T1)dR of particles having radii from
R to R + dR per unit volume of the specimen at time T1 is equal to the number I( T)d T of par-
t1 't't

ticles formed earlier in the time from T to T + d T, and dR = ~ C ('t ') d't' - ~ C ('t') d't' = C ('t) d't, so
,+d,
that p (R, 't j ) dR = I ('t) Cd(~) If the growth rate varies little during phase conversion, then
"
R = ~ C ('t') d't' = C ('t j - 't), where c is the average growth rate, so that

(1)

The value of c( T) or C can be found experimentally, e.g., from the rate of increase of the
maximum radius r m of circular sections of spherical particles on a polished section or an elec-
tron-microscopic replica in relation to the time T1 of phase conversion. Knowing c( T), we can
find I( T) in accordance with (1) if p(R) is known (T1 is omitted for brevity). The relation between
p(R) and the experimentally known number n(r)dr of circular traces having radii from r to r +
dr on a section of area S is found as follows. We first find the number of traces left on the
section by p(R)dR particles of the specified radius R. Traces with radii from r to r + dr will be
left by all particles the centers of which lie in layers of thickness dh = (r/h)dr at a distance
h = ..{R2_ r2 (Fig. 1) on both sides of the section plane. The number of traces is

rdr
p (R) dR2Sdh=2Sp (R) dR VR2-r 2 ' (2)

Consequently
Rm

n(r)dr=2Srdr S VR2-r2'
(R) dR
p
(3)

48
V. N. FILIPOVICH AND A. M. KALININA 49

Fig. 1. Schematic di-


dr agram of a section of
a spherical particle
cut by the plane of the
specimen.

where it is taken into account that traces of radius r can be left only by particles of radius R ~
r, R m is the maximum particle radius.
The problem consists of finding p(R) by solution of the integral equation (3). Variants of
this problem are considered in the literature [1-3]; a review of methods for approximate solu-
tion of the problem is given in Saltykov's monograph [3]. Toshev and Gutzov [4] emphasize that
it is possible to determine the nucleation rate of spherulites in glass from known n(r) and p(R),
and give theoretical relationships between n(r) and the crystal nucleation rate I for a number of
typical time laws of crystal nucleation.
We will describe a new and more exact method for solving Eq. (3); in distinction from
[1-3], due account is taken of the singular character of the integrand in (3) at R -r, which is
especially important for exact determination of p(R) at small R. We write Eq. (3) in the form

Sp
HI Hm

r SvR2-r2'
P (R) dR AI = r (R) dR
f(r)=!t(r)+A!t(r), 11 = (4)
1 VR2- r2 '
HI

where R1 is chosen fairly close to R m so that it is possible with adequate accuracy to replace
P(R) by its value p(R 1) at a certain midpoint in the interval [Rio Rm]. When r = R1

(5)

from which the value of p(R 1) can be found, which we assign to the right-hand boundary of the
interval [R lo Rm] - inthe present instance to the point Rm. Knowing p(R 1) we can, after integra-
tion analogous to (5), write 6./ 1 in the form

(6)

The meaning of 6.f1(r) is the distribution of traces left by the largest particles, with radii
from R1 to Rm. Subtracting 6.f1(r) from/(r) we obtain/1(r), the distribution of traces of the re-
maining particles, for which the maximum radius is R1 Proceeding analogously to (4)-(6), we
obtain

(7)
50 V. N. FILIPOVICH AND A. M. KALININA

30
28
28
24
22
20
18

12
Fig. 2. Graphical scheme for solu- 10
tion of Eq. (3). 8
8
4
2

o 0.1 0.2 0.3 aLI 0.5 0.8 0.7 0.8 o.g 1


rs ra ~ ro ~ ~ rs rz ~

where all the expressions differ from (4)-(6) only by replacement of Rm by R t and of R t by R 2
Continuing this process, we find the p(R) values for the required number of points R i
To test the accuracy of our method, we found the function j(r) by direct integration for
the case where P(R) = const for 0 s R s Rm and p = 0 for R > Rm' We then solved the converse
problem using the knownj(r), and found P(R) by the method described above. It is convenient
for the calculations to subdivide the interval [0, Rm] into equal parts. We divided it into 10
parts, so that the Ri values were R t = 0.9R m , R2 = 0.8R m , etc. In general, it is convenient also
in other problems to measure all lengths taking Rm as unity, i.e., in fractions of Rm. It is then
possible to use the same table of values of the function F(r, Rio Rj+t) (see Table 1) in all cases.
Calculation of p(R j ) is illustrated by Fig. 2, which showsj(r), /t(r), curves and the .6.jt(r), .
.6./2(r), curves successively subtracted from them. The calculated values of P(R) differ
from the true values by not more than 0.5%. It can therefore be concluded that subdivision of
the interval [0, Rm] into 10 parts is sufficient, and errors in experimental determination of the
function j(r) will be more significant than the errors due to the method. The accuracy of deter-
mination of the functionj(r) = n(r)j2S = Ns (r)j2Sdr, where Ns(r) is the number of traces with

TABLE 1. Values of the Function F(r, R i, R j+1)


I

~r Ri+l

0.1
1, 0.9

0.02537
0.9, 0.8

0.02909
0.8, 0.7

0.03068
0.7, 0.6

0.0467
0.6, 0.5

0.0673
0.5,0.4 0.4,0.3 0.3,0.2

0.0913 0.1303 0.2349


0.2, 0.1

1.0
0.2 0.05151 0.05818 0.08224 0.0924 0.1185 0.1827 0.2996 1.0
0.3 0.0790 0.08728 0.1231 0.156 0.2089 0.3314 1.0
0.4 0.1255 0.1233 0.1797 0.2264 0.3407 1.0
0.5 0.1476 0.1746 0.2483 0.3665 1.0
0.6 0.1988 0.2618 0.3921 1.0
0.7 0.2601 0.3666 1.0
0.8 0.3785 1.0
0.9 1.0
1.0
V. N. FILIPOVICH AND A. M. KALININA 51

radii from r to r + dr on the section of area S, depends on the magnitude of the interval dr, which
is determined by the magnification of the microscope. The accuracy of j(r) may be regarded
as. satisfactory if dr = (0.1- 0.2)r m' i.e., with j(r) determined for 5 to 10 values of r. We
used this method for determining the dependence of the rate of nucleation of spherulites on
time in lithium silicate glass (the results are not given here owing to lack of space). It should
be noted that since the particles (spherulites) are not ideally spherical, while the accuracy of
determination of j(r) diminishes with decrease of r, at low r negative values may be obtained
by subtraction of A!i(r) fromji-1(r). This is an indication that the calculation should be dis-
continued.
The form of the p(R) curve can be inferred even from the general form of the j(r) curve.
If j(r) is relatively symmetrical and convex on both sides, like the ji (r) curves in Fig. 2, then
p(R) is roughly constant in the range 0 :s R :s Rm. If j(r) is asymmetric and has a concave
left-hand branch, like the h(r) curves in Fig. 2, which approximately correspond to distribution
of traces of particles of specified radius R i in accordance with Eq. (2), then p(R) has a well-
defined maximum at a certain R.
In conclusion, we give certain integral formulas which can be used for rapid estimation of
average particle radii, average nucleation rates, and volumes of glass which have undergone
phase conversion. Integrating Eq. (2) with respect to dr from 0 to Rm and with respect to dR
from 0 to Rm , we obtain the following expressions for the total number of traces on the section
and the average radius of a trace:
Rm Rm Rm

Ns = ~ ndr=2S ~ pdR=2SNlI, N= ~ pdR, (8)


o 0 o

F=+ f Rm
Rm

8
nrdr=1-' ~2
R
,
_ 1 RSm
R2=7Y pR 2dR, - 1
R=7V SpRdR. (9)
o o o

If particles of only one radius Rm(p (R) = 0 (R-Rm}} are present, then

'It _
f=t;R. (10)

If, on the other hand, the numbers of particles of different radii from 0 to Rm are equal,
i.e., p(R) = const at 0 :s R :s R m , then

(11)

It can be concluded i!!. accorda.!!ce with (10) and (11) that in real cases the averag~ radius
r of a trace lies between 'Il'R/4 and 'Il'Rj3, i.e., it is equal to the average particle radius R with
an error of up to 25% at the most. According to (8) we have N ~ Ns/2Sf' for the total number of
particles per unit volume, with the same degree of accuracy. Therefore in order of magnitude
(in the case where p(R) varies relatively smoothly in the range 0 :s R :s R m}

N N eNs
p (R) "" Rm "" 2i' ' 1= cp "'" 4Sf 2 (12)

Analogously to (8) and (9), we calculate the total area of traces on the section:

(13)
52 V. N. FILIPOVICH AND A. M. KALININA

where Vi is the volume occupied by the particles present in unit volume V of the specimen. Of
course, Eq. (13) is valid not only in the case under consideration, when the phase heterogeneities
are spherical and discrete, but in the most general case. For completeness, we will give a
simple derivation of Eq. (13) for the general case. Suppose that we have a cube of a heteroge-
neous substance of volume V = L3. We cut it into m arbitrarily thin slices ~l thick and S = L2
in area parallel to one of the cube faces. The volume occupied by heterogeneity regions (of
m
any form, touching each other or coalescing) in the cube V is VI = ~ S/:..l, where Si is the total
i=1

area of cross sections of heterogeneity regions crossed when the i-th slice is cut. Dividing
m

i = ;~s
~s;
both sides of this equation by V = S~l, we obtain ' which proves the validity of
Eq. (13). It is also evident that this equation (like Eqs. (2)- (12 is valid for a single section
only if its area is fairly large and if the heterogeneity regions are distributed uniformly and
randomly over the volume V.
Conclusions
1. Glass is a convenient object for study and verification of the laws of phase conversion,
owing to the possibility of freezing various stages of phase conversion.
2. Study of thin glass sections with the aid of an ordinary metallographic microscope,
with suitable analysis of the experimental data, makes it possible to reproduce the picture of
the phase conversion process from its earliest stages, which are not always accessible to in-
vestigation even by highly sensitive methods.
3. A method has been devised for highly accurate determination of particle-size distribu-
tion and rate of particle nucleation in the course of phase conversion from experimentally deter-
mined growth rates of the particles and from the distribution of radii of traces of the spherical
particles on a polished section or replica (the accuracy is limited only by the accuracy of the
experimental data).
Literature Cited
1. E. ScheH, Z. Anorg. Allgem. Chern., 201(2):259 (1931).
2. A. Huber, Z. Phys., 93(4):227 (1935).
3. S. A. Saltykov, Stereometric Metallography, Metallurgizdat, Moscow (1958).
4. S. Toshev and I. Gutzov, Phys. Status Solidi, 24:349 (1967).
SURFACE-ACTIVE PROPER TIES OF CATALYSTS AND
PHASE SEPARATION IN GLASSES

Ya. A. Fedorovskii

It was formerly believed that phase separation occurs only in silica- alkaline earth oxide
systems, where the drops are fairly large and can be observed with the aid of an optical micro-
scope. Data on low-angle x-ray scattering and the results obtained by other methods indicate
that glasses contain microheterogeneity regions of the order of 10 A in size. The discovery of
microcrystalline glasses and extensive applications of electron microscopy revealed micro-
phase separation in the size range from 10 A to several microns in glasses of the most diverse
compositions.
It is assumed in this paper that all these microheterogeneity regions are of the same
nature and differ only in size and in the extent of the process. We presume that only the initial
stages of phase separation occur in microheterogeneous glasses, the drops not being able to
grow to macroscopic size for various reasons (such as high viscosity or steric hindrance). It
is also assumed that vitreous silica, which consists of a three-dimensional random network of
Si04 tetrahedra, is homogeneous. When an oxide MO is introduced into it, the cation of the ox-
ide may replace silicon in the network, and the oxide is then considered to be dissolved in the
glass. If the cation does not enter the network, elementary microheterogeneity or separation
occurs. The thermodynamic behavior of the oxide MO in the glass is represented by the mag-
nitude and sign of the energy of mixing. Exact calculation of this quantity is not possible, but we
attempted to estimate it after making a number of assumptions. The following assumptions were
made: 1) a silicate melt is regarded as ionic, and interaction between the ions is attributed to
Coulomb forces; 2) the energy of mixing is independent of concentration at low concentrations
of the oxide MO. The results of the estimates [1] are given in Table 1.
It follows from the data in Table 1 that the tendency to phase separation increases in the
sequence from oxides of high valence and low ionic radii (network formers) to alkali ions

TABLE 1
Cation Cation Energy of Cation Cation Energy of
MO radius charge mixing (Urn), MO radius charge mixing (Urn),
(R e ), A (Z) keal/mole (Re), A (z) kcal/mole

Li 20 0.68 1 445 BrOs 0.20 3 30


Na20 0.98 1 465 A 20S 0.57 3 56
K 20 1.33 1 480 Ge0 2 0.44 4 0.3
MgO 0.74 2 224 8n0 2 0.67 4 7.5
CaO 1.04 2 264 TiO a 0.64 4 6.3
BaO 1.38 2 287 Zr0 2 0.82 4 16.5

53
54 Ya. A. FEDOROVSKII

(modifiers). At first sight this appears to contradict accepted views. However, the data in
Table 1 indicate only that the tendency to phase separation is higher in alkali glasses than in
borate, titanate, or zirconate glasses, but the extent of this separation depends on the kinetics
of the process. For example, in the case of lithium glasses the small lithium ions diffuse
readily in the glass and the drops rapidly grow and become detectable. In sodium glasses
separation is observed only after lengthy heat treatment. The potassium ion is considered
larger than the sodium ion; therefore a much longer time of heat treatment is probably needed
for "development" of the drops. Accordingly, potassium glass contains very numerous drops
the size of which may be beyond the resolution limits of the electron microscope (especially if
the replica method is used), and the glass is regarded as homogeneous. An analogous effect
occurs in crystallization: glasses with a high tendency to crystallization contain numerous fine
crystals (e.g., transparent glass-ceramics), while glasses which do not give rise to glass-
ceramics contain small amounts of large crystals.
Catalyst oxides (Sn02, Ti02, Zr02) may dissolve in glass at high temperatures, but on
decrease of temperature their position in the solution becomes unstable. The mechanism of
their separation can be conveniently interpreted with the aid of Semenchenko's hypothesis, ac-
cording to which the concentration of the ions having the lower potential increases in the sur-
face layer at an interface. Accordingly, if as the result of fluctuation in glass a distribution of
oxides arises such that the central part of the given region has a higher Si02 concentration
while the peripheral parts have a higher concentration of catalyst oxides, then the process will
continue so that the concentration of catalyst oxides at the periphery will increase by diffusion
from the center. Therefore if we call the central region a drop and the peripheral regions the
surface layer, we have a surface layer enriched with catalyst, and crystallization nuclei arise
in this layer. In our view, the nuclei have the composition of the catalyst oxide or of a com-
pound of it.
Our model shows one significant discrepancy from the spinodal model of phase separation.
However, it is the consequence of a terminological ambiguity. In our view the catalyst concen-
tration increases from the center to the periphery of the drop. By the terminology of spinodal
phase separation the catalyst concentration diminishes from the center to the periphery of the
drop. In reality the situation is the same in both cases; the explanation of the discrepancy is
that in discussions of spinodal separation the drops are considered to form in the rna trix,
whereas in our model the matrix is divided by peripheral regions into drops.
Literature Cited
1. Ya. A. Fedorovskii, Investigation of Initial Stages of Catalyzed Crystallization of Lithium
Aluminosilicate Glasses, Dissertation, Inst. Obshch. Neorgan. Khim., Moscow (1966).

DISCUSSION
Fed 0 r 0 v s k ii, in reply to questions by Filipovich, Varshal, and others regarding
the mechanism of phase separation described by him, said that it consists of formation
of fluctuation drops enriched with silica at the center. An interface is formed when the
separating layer attains a definite composition and size at which it is energetically advan-
tageous. In that case, when this layer can be regarded as an independent phase, the observed
picture coincides with that of spinodal separation. The speaker noted the significant role of
kinetic factors in this process. In his opinion, the size of the alkali cations have the determining
influence on the tendency to separation in alkali silicate systems and on the rate of the process.
Filipovich said with reference to Fedorovskii's paper that in the case of a three-
component system it may happen that the third component will be present in a higher con-
centration at the interface and will decrease the free energy at the interface. However, this
Ya. A. FEDOROVSKII 55

bears no relation to spinodal decomposition, which also occurs in the two-component system.
The influence of the kinetic factor should be taken into account when the temperature of phase
separation is low while the viscosity is high, as in the potassium silicate system. Appearance
of phase separation is in itself innependent of kinetics, which determines only a higher or lower
rate of the process.
Va r s hal notes that conclusions drawn from Fedorovskii's reasoning are incon-
sistent with reliable experimental data and with comparative estimates on the tendency to
phase separation in a number of systems, including those discussed in the paper by Gal-
akhov and VarshaI. Another defect of Fedorovskii's method for assessing the tendency to
phase separation is that this tendency cannot be estimated from the energy of mixing, the dif-
ference between the free energies, or the width of the region.
In his reply Fedorovskii said that some misunderstanding arose because the fig-
ures were not to scale. Moreover, in his view the question of homogeneous structure of potas-
sium glasses is controversial, because improvement of the techniques of electron microscopy
may in prinCiple lead to detection of a droplet structure also in potassium silicate glasses,
where the droplets are possibly very small (of the order of 10 A).

* * *

In an additional discussion of Fedorovskii's paper*


Filipovich noted that Fedorovskii erroneously identifies "elementary heteroge-
neity" of glass (structural elements or structural units, the structure of which depends
on temperature) with heterogeneity due to liquid-liquid phase separation. This, in particular,
is the reason why the results of Ya. A. Fedorovskii's calculations of the energy of mixing Urn
(see Table 1) are in sharp contradiction to experimental data as they give a dependence of the
critical temperature Tcr '" Urn/R for various Me xOy- Si02, systems which contradicts the ex-
perimentally observed dependence, while the theoretical values of Tcr '" U rn/R are greater than
the experimental by 1-2 orders of magnitude. All this indicates that the physicochemical con-
cepts on which the calculations of Urn are based are erroneous. Moreover, as is known from
existing theoretical publications, the nature of phase separation in the spinodal region is, in
general, entirely unrelated to surface activity of the components, the role of which is particular-
ly stressed by Fedorovskii in his alternative explanation (not based on any theory of the
process kinetics) of the characteristic features of the geometry of heterogeneity regions formed
in glass during spinodal decomposition. As yet there are not sufficient experimental and theoret-
ical data to justify this alternative explanation.
Fed 0 r 0 v ski i retorted that the term "phase separation" of glass is not defined
in the literature. Microheterogeneous structure, microphase separation, and liquid-liquid sep-
aration are often discussed but it is by no means evident that these phenomena differ in nature.
In his opinion an oxide has dissolved in the glass if its cation has replaced silicon in the silicon-
oxygen network, and has not dissolved if its cation cannot replace silicon and is present in the
structural cavities of the silicon- oxygen network; this is the origin of phase separation. His
calculated values of Urn characterize the ability of cations to replace silicon in the network.

*In view of the controversial nature of Ya. A. Fedorovskii's paper, on the suggestion of the Edi-
torial Board the discussjon was continued while the symposium proceedings were being edited.
Members of the Editorial Board B. G. Varshal, F. Ya. Galakhov, 0. V. Mazurin, S. V. Nemilov,
E. A. Porai-Koshits, and the author of the paper Ya. A. Fedorovskii took part in the discussion
(Editorial note).
56 Ya. A. FEDOROVSKII

The critical temperatures should be referred to heterogeneity elements and not to the immis-
cibility region of the phase diagram.
With regard to this treatment of spinodal structure, Fedorovskii said that it was not
his intention to offer a new explanation of the origin of spinodal structure, but that he wanted to
emphasize that a structure of this type can be explained in a different way, on the basis of sur-
face properties of the added components.
DISCUSSION ON SECTION I

Bar ten e v pointed out that the causes of experimentally observed heterogeneous
structure of glasses need not lie only in metastable liquid-liquid phase separation. It is now
reliably known that liquid metal eutectics sometimes retain microheterogeneous structure at
temperatures considerably above the eutectic point ("quasi-eutectic structure of the liquid
state"). Admittedly, these regions are somewhat smaller, of the order of 30-50 A, than in sili-
cate systems. Since silicate systems also have phase diagrams of the eutectic type, there is
the possibility of explaining their heterogeneous structure on the same principles. This is a
debatable point.
K a rap e t y a n said that in studies of the spectroscopic properties of glasses he
met instances of the influence of the structure of glass and of its thermal history on concentra-
tion quenching of luminescence. This was attributed to nonuniform distribution of the activator
in the glass, and to its predominant presence in the ionic part. In silica-rich compositions in
binary systems the behavior of the activator is such as to give the impression that its concentra-
tion has been increased. This is due to the phenomenon of activator segregation, discovered by
Karapetyan and his coworkers, which precedes phase separation and can be used for studying
its initial stages.
With reference to Bartenev's comments, M a z uri n pointed out that the symposium
was concerned with various aspects of structure resulting from phase separation, but not
with the problem of the heterogeneity of glass structure as a whole. The term "two-
phase" (or polyphase) "glass" rather than "heterogeneous glass" is preferable for characteriza-
tion of glasses having such structure, since there are glasses which are not two-phase but are
heterogeneous. The glasses under discussion have the most distinct heterogeneous structure
and are of interest in this respect as models for formulation of theoretical concepts of the in-
fluence of heterogeneity of any kind on the properties of glass.

57
II

PHASE SEPARA TION IN ALKALI SILICA TE


AND BOROSILICATE GLASSES
INFLUENCE OF THE OXIDES OF
LITHIUM, POTASSIUM, AND
CESIUM ON PHASE SEPARATION IN
SODIUM SILICATE GLASSES

N. S. Andreev and G. G. Boiko

In studies of the structure of three-component silicate glasses containing two alkali metals
it must be taken into account that the heterogeneous structure of two-component sodium and
lithium silicate glasses results from liquid-liquid phase separation [1-7] and that the immis-
cibility region extends to an alkali oxide content of about 20 mole% in the first case and 33 mole%
in the second. The problem of the structure of glasses of the type xNa20 (1- x)R 20 . ySi02 can
then be approached by elucidation of the influence of addition of various alkali oxides on phase
separation in the sodium silicate system.
Data on variations of the temperature of phase separation in conjunction with spectral
relationships for the intensity of scattering of visible light provide information on the character
of distribution of various modifier ions in two-alkali silicate glasses containing sodium oxide.
One of the main results of the present work is demonstrated in Table 1 and Fig. 1.* If
we define ~ts as the difference between the temperature at which opalescence disappears in the
three-component glass under consideration and in a two-component sodium silicate glass [7]
with a sodium oxide content equal to the sum of the molar percentages of alkali oxides in the
three-component glass, we can subdivide oxides of other alkali metals into two groups in ac-
cordance with their influence on phase separation in sodium silicate glasses. The first group
comprises lithium oxide, introduction of which into sodium silicate glass raises the tempera-
ture of phase separation (~ts is positive). The second group contains the oxides of potassium
and cesium, which lower the temperature of phase separation (~ts is negative). Rubidium is in-
termediate between potassium and cesium in the periodic system, and in this respect its oxide
should also be assigned to the second group.
This subdivision of alkali oxides also applies to their influence on the spectral dependence
of the scattering intensity of visible light. It was shown earlier [4] that the exponent p of the
wavelength characterizing the spectral dependence of light scattering intensity depends on the
ratio of the total volume occupied by the glass by the phases into which it separates. When this

*All the results presented here refer to room temperature and quenched glasses. The time of
treatment at the corresponding temperature, required to produce opalescence, did not exceed
48 h.

61
62 N. S. ANDREEV AND G. G. BOIKO

TABLE 1
Glass composition. mole'Vo by analysis
No. ts' c II ts' C
LI,O
I K,O
I Cs,O
I Na,O
I ~R20
I SiO,

1 1.6 - - 14.9 16.5 83.5 765 +45


2 4.4 - - 13.1 17.5 82.5 760 +70
3 - 1 - 4 5 95 760 -95
4 - 1.9 - 3.1 5 95 715 -140
5 - 1.1 - 13.5 14.6 85.4 750 -20
6 - 23 - 13.2 15.5 84.5 670 -75
7 - 3.8 - 10.8 14.6 85.4 625 -150
8 - 1.5 - 15 16.5 83.5 690 -30
9 - 2.9 - 15.5 18.4 81.6 580 -80
-
10 - 0.8 14.9 15.7
I 84.3 610 -105

to
900
Fig. 1. Variations of the
temperature of phase sep-
aration ts on partial re-
placement of sodium oxide
by oxides of other alkali
metals. I) Binodal curve
for the sodium silicate sys-
tem. Glasses: 1) Li1.6 .
NaI4.9; 2) K2.3 NaI3.2;
3) KINa4; 4) K1.9 Na3.I;
5) esO.8 NaI4.9. Data of
Moriya et al. [5]: 1 ')
Lil5 .Na5; I") LilO NaIO;
10 15 20 1 In) Li5 . NaI5.

ratio is close to unity and under the condition that the scattering particles are much smaller
than the wavelength of the radiation used, the exponent p has an anomalous value close to 8 .
When the volume of one of the phases decreases to a few percent, p decreases to a value close
to 4.
For a glass of a given composition p increases with decrease of temperature, i.e., with
increasing distance from the binodal; at relatively low temperatures it reaches its highest val-
ue PM and remains almost constant on further decrease of temperature. On increase of the
sodium oxide content in the glass above 7%, PM gradually diminishes to the normal value.
The exponent p (',I for glass NaI6* is 5.2; at the same time, it follows from Fig. 2 that
for glasses having total sodium and lithium oxide contents of 16.5 and 17.5% (curves 1 and 2)
PM is 5.6. Therefore partial replacement of sodium oxide by lithium oxide led to some increase
of PM'

*The symbol Rx indicates that the two-component silicate glass contains x mole% R20; the sym-
bol Rx . R 'y indicates that the three-component silicate glass contains x mole% R20 and y mole%
H2O.
N. S. ANDREEV AND G. G. BOIKO 63

p
(j f
2
5 J
If

~
If

2
550 600 650 700 750 to
Fig. 2. Dependence of the wavelength exponent p on
temperature. Glasses: 1) Li1.6 Na14.9; 2) Li4.4
Na13.1; 3) K1.1 . Na13.5; 4) K2.3 Na13.2; 5) K3.S .
Na10.S.

Replacement of sodium oxide by potassium oxide produces the opposite result. It follows
from earlier work [4] that PM for glass Na 14.6 is in the interval between 5.6 and 6.3; at the
same time, Fig. 2 shows that for glasses K1.1 Na13.5 and K3.S . NaIO.S, having total alkali
oxide contents of 14.6% (curves 3 and 5) p \1 is 5 and 3.S, respectively. Therefore partial re-
placement of sodium oxide by potassium oxide leads to considerable decrease of the wavelength
exponent.
In interpretation of these results it must be taken into account that the model of regular
solid solutions is successfully applied to phase separation in glasses [S]. It is known from the
theory of these solutions [9] that introduction of a third component, equally soluble in the other
two components, into a two-component solution having an immiscibility region lowers the crit-
ical solution temperature, i.e., increases miscibility. Conversely, introduction of a third com-
ponent which is predominantly soluble in one of the two main components raises the critical
temperature.
Our investigation does not refer to the critical composition of the sodium silicate system,
corresponding to an Na20 content of approximately 7.5%. Therefore numerical comparison of
our results with inferences from theory is impossible. However, it is to be expected from gen-
eral considerations that the relationships described are valid not only at the critical point but
over a considerable part of the binodal curve.
With this assumption, which is supported by the nature of variations of the spectral de-
pendence of light-scattering intensity, we can in the first approximation represent the structure
of two-alkali glasses containing sodium oxide as follows.
As introduction of potassium oxide into sodium silicate glasses lowers the phase-separa-
tion temperatures along the binodal both to the left and to the right of the critical composition,
it follows that potassium ions are distributed more uniformly in the glass than sodium ions.
This is accompanied by decrease of the relative volume occupied by silica-rich regions, owing
to increase of silica solubility in regions enriched with alkali ions.
This last conclusion follows from the observed decrease of PM' i.e., weakening of the in-
terparticle interference effect. This relationship becomes more pronounced with increase of
the amount of potassium oxide introduced into the glass.
Thus, the homogenizing influence of potassium oxide on sodium silicate glasses is indis-
putable. However, the possibility of complete suppression of phase separation by introduction
of a sufficiently large amount of K20 should be considered with caution. This applies primarily
to the region of low total alkali-oxide contents, of the order of 7% and less.
64 N. S. ANDREEV AND G. G. BOlKO

This problem is intimately connected with the possible occurrence of phase separation in
glasses of the K20- Si02 system with low potassium oxide contents. There are at present no
experimental data on this question, although some indirect information is to be found in the
literature [5, 6].
We note in this connection that despite the relatively long (48 h) times of treatment at
temperatures in the 490-670 range we did not succeed in inducing opalescence in K5. Nal0 and
K3 . Na2 glasses. In the first case the result is attributable to approach of the glass structure
to a state of complete homogeneity; with further of sodium oxide by K2 0 in this glass such a
state is undoubtedly reached. Indeed, the above-mentioned decrease of PM observed on replace-
ment of sodium oxide by potassium oxide in a series of glasses haVing a total alkali content
close to the present instance, namely 14.6%, indicates, by analogy with earlier data [4], a de-
crease of the relative volume occupied by silica-rich regions from about 25% (Na14.6 glass)
to a value not exceeding 5% (Table 1, glass No.7). This tendency indicates that the impossibility
of inducing opalescence in glasses with even higher K20 contents is the result of their complete
homogeneity. The homogenizing action of potassium oxide must be even more pronounced in
glasses having a total alkali oxide content exceeding 15 %, and when the total content of alkali
oxides exceeds 20% sodium potassium silicate glasses are homogeneous (in the sense of absence
of phase separation) at all ratios of the alkali oxides.
Although in the case of glasses having total alkali oxide contents not exceeding 5% there
is no doubt that increase of the potassium oxide content results in gradual homogenization, the
limit of this process can be finally established only after the problem of phase separation in
potassium silicate glasses has been solved.
It follows from the results in Table 1 that introduction of even a small amount of cesium
oxide into sodium silicate glass lowers the temperature of phase separation conSiderably, and
therefore under given conditions it has an even stronger homogenizing effect than potassium
oxide.
A different result is produced when lithium oxide is introduced into sodium silicate glass-
es. The resultant rise of the temperature of phase separation indicates that lithium ions enter
into one of two types of heterogeneity regions. Since glasses in the system Li 20- Si02 undergo
phase separation, the more probable hypothesis is that these regions are sodium-rich forma-
tions. Increase of PM indicates that this is accompanied by increase of the relative volume oc-
cupied by silica-rich regions, i.e., by intensified phase separation.
In the present investigation the influence of lithium oxide on phase separation in sodium
silicate glasses was studied in less detail than the influence of potassium oxide. However, our
results, and also the data of Moriya et al. [5], suggest that glasses in the Na20- Li 20- Si02
system with total alkali oxide contents up to 20% undergo phase separation and have heteroge-
neous structure with any ratio of sodium and lithium oxides. The existing information is not
sufficient for drawing strictly valid conclusions regarding the structure of glasses having alkali
oxide contents over 20%; nevertheless, since the immiscibility boundary corresponds to 20%
for sodium silicate glasses and to 33% for lithium silicate glasses, it can be inferred that in this
case phase separation should be weakened with decrease of the lithium oxide content.
We note in conclusion that the aim of the present investigation was merely to elucidate the
general tendencies in variation of the structure of sodium silicate glasses on addition of relative-
ly small amount of other alkali oxides. A more detailed study of immiscibility boundaries in
two-component silicate glasses will provide a more detailed picture of their structure and as-
sist in solving the problem of the degree of influence of their heterogeneous structure on the
electrical and chemical properties [10, 11], including the polyalkali effect.
The authors thank V. S. Molchanov for his constant interest and for providing a number of
glasses for the investigation.
N. S. ANDREEV AND G. G. BOlKO 65

Literature Cited
1. V. I. Averyanov and E. A. Porai-Koshits, in: The Glassy State, Proceedings of the Fourth
All-Union Conference, Izd. Nauka, Moscow- Leningrad (1965), p. 98 [English translation:
The Structure of Glass, Vol. 6, Consultants Bureau, New York (1966), p. 98J.
2. D. A. Goganov and E. A. Porai-Koshits, Dokl. Akad. Nauk SSSR, 167 :1266 (1966).
3. J. J. Hammel, Seventh International Congress on Glass, Brussels (1965), p. 36.
4. N. S. Andreev and T. M. Ershova, Dokl. Akad. Nauk SSSR, 172:1299 (1967).
5. Y. Moriya, D. H. Warrington, and R. W. Douglas, Phys. Chem. Glasses, 8:19 (1967).
6. R. J. Charles, J. Am. Ceram. Soc., 50:631 (1967).
7. N. S. Andreev, D. A. Goganov, E. A. Porai-Koshits, and Yu. G. Sokolov, in: The Glassy
State, Vol. 3, No.1, Catalyzed Crystallization of Glass, Izd. Akad. Nauk SSSR, Moscow-
Leningrad (1963), p. 46 [English translation: The Structure of Glass, Vol. 3, Consultants
Bureau, New York (1964), p. 47J.
9. 1. Prigogine and R. Defay, Chemical Thermodynamics [Russian translationJ, Izd. Nauka,
Sibirsk. Otd., Novosibirsk (1966), p. 248.
8. V. N. Filipovich, Neorgan. Mat., 3:993 (1967).
10. O. V. Mazurin, Tr. Leningr. Tekhnol. Inst. Leningrad, No. 42 (1962).
11. T. E. Chebotareva, A. A. Pronkin, and V. S. Molchanov, Zh. Prikl. Spektrosk., 5 :241 (1966).
RELA TION BETWEEN THE MICROHETEROGENEITY OF
LITHIUM SILICATE GLASSES AND THE RATE OF
THEIR DISSOLUTION IN HYDROFLUORIC ACID

v. A. Borgman, V. K. Leko, and V. K. Markaryan


The heterogeneous character of the structure of lithium silicate glasses containing up to
33% Li 20 is well known. Studies of the dependence of electrical properties of lithium silicate
glasses on composition have been reported [1-3]. It has been shown that introduction of another
alkali oxide or Al 20 3 changes the conductivity of lithium silicate glasses in a manner suggesting
a more uniform distribution of alkali cations in comparison with the original two-component lith-
ium silicate glasses. We studied the dependence of the rate of dissolution of the same glasses
in hydrofluoric acid on their composition. The literature contains only fragmentary data on the
dissolution rates, in relation to studies of the photochemical process [4].
The following technique was used: rectangular specimens with polished surfaces were
made from annealed lithium silicate glasses. The specimens were suspended in the platinum
gauze in 10% hydrofluoric acid solution stirred at the rate of 60 rpm. The treatment time was
varied from 1 to 20 min. The dissolution rate v was expressed in grams of glass dissolved per
cm 2 per second. The results of duplicate experiments did not differ by more than 10%.
The dependence of the rate of dissolution of lithium silicate glasses on the Li 20 content
(in mole%) is shown in Fig. 1. The curve completely corresponds to the analogous dependence
of the electrical conductivity of the same glasses on concentration [1]. It follows from Fig. 1
that the dissolution rate of glass containing 8% Li 20, in which the lithium ions are present main-

5
u
~ 4
"'6
~bO 3

Fig. 1. Dependence of the


<0

,...
a
2
> rate of dissolution of lithium
1
silicate glasses in 10% HF
at room temperature on the
Li20 content.
0 10 20 30
Lip. mole"lo
66
V. A. BORGMAN, V. K. LEKO, AND V. K. MARKARYAN 67

15LiZO
Fig. 2. Dependence of solubility
on composition of glasses in the :>
series yLi 20 x K20 (100-y-
x/2)Si0 2 o

5 10
xK2 0

ly in closed alkali groups comprising about 20% of the total glass volume [1], is comparable to
the rate of dissolution of vitreous silica. Increase of the Li 20 concentration from 8 to 10%
leads to a sharp increase of solubility, apparently as the result of formation of alkali links be-
tween the alkali groups. The dissolution rate of glasses containing from 15 to 33% Li 20 is de-
termined by the dissolution rate of the alkali phase. The equal solubilities of these glasses in-
dicate that the compositions of the lithium-containing phases are similar. This result is not
consistent with Vogel's hypothesis [5] of "secondary separation" in lithium silicate glasses.
The influence of small additions of K20 or A1 20 3 depends on the degree of heterogeneity
of the original lithium silicate glass. Figure 2 shows the influence of K20 on the dissolution
rates of lithium silicate glasses containing 15, 25, and 33% Li20. Of these, the glass with 15%
Li 20 is the most heterogeneous [5, 6]; addition of K20 initially leads to a considerable decrease
of its dissolution rate. One of the main causes of this is apparently the more uniform distribu-
tion of alkali cations on introduction of potassium oxide into the glass. This lowers the local
concentrations of alkali cations, causing an increase in the resistance of the glass to the action
of HF. The effect of lowering the dissolution rate resulting from introduction of the first por-
tions of K20 diminishes with increasing homogeneity of the original glass. It does not occur at

15Li2 0
70

80

~ 50

"'5 40
"-
.
DO

"'0 30 Fig. 3. Dependence of the dis-


rl

:> solution rate on composition of


20 glasses in the series yLi20 .
x AlOt ' 5 (100 -y-x) Si02
10

0 10 20 30
xAlOl.5
68 V. A. BORGMAN, V, K. LEKO, AND V, K, MARKARYAN

all in glass corresponding to lithium disilicate in composition, and on introduction of potassium


oxide into this glass the solubility in hydrofluoric acid increases steadily.
Similar results are observed when aluminum oxide is introduced into lithium silicate
glasses (Fig. 3). Thus, introduction of Al 20 3 into heterogeneous lithium silicate glasses also
leads to homogenization. The existence of this effect is confirmed by the sharp increase of the
electrical resistance of lithium silicate glasses on introduction of a small percentage of alumi-
num oxide [31. The increase of resistance is also apparently due to homogenization of the glass
and to sharp decrease of the local concentrations of alkali ions.
Literature Cited
1. V, K, Leko, Neorgan. Mater" 3:1224 (1967).
2. V. K. Leko, Neorgan. Mater., 3:1888 (1967).
3. V. K. Leko, Neorgan. Mater., 4:121 (1967).
4. W, Hinz, G. Solow, and G. Kranz, Silikattech., 16:210 (1965).
5. W, Vogel, in: The Glassy State, Proceedings of the Fourth All-Union Conference, Izd,
Nauka, Moscow- Leningrad (1965), p. 108 [English translation: The Structure of Glass,
Vol. 6, Consultants Bureau, New York (1966), p. 1141.
6, D, A. Goganov and E, A. Porai-Koshits, in: Structural Transformation in Glasses at
High Temperatures, Izd. Nauka, Moscow- Leningrad (1965), p. 100.

DISCUSSION
Replying to questions by Mazurin and Aslanova on the paper by Borgman, Leko,
and Markaryan, L e k 0 said an influence of heat treatment on the dissolution rate was not
detected, and that the acid concentration was 10% in nearly all cases. Essentially the
same experimental relationships were found with the use of 20% acid, In reply to Er-
molenko and Galakhov, Leko said that the authors of the paper did not correlate the ob-
served relationships with the mechanism of phase separation or the phase diagram. The paper
was concerned only with the relation between the dissolution rate and the separated structure
present in the glass,
STUDY OF PHASE-SEPARATION PROCESSES IN
THE SODIUM SILICATE SYSTEM BY
THE ELECTROMOTIVE FORCE METHOD

A. F. Borisov and V. I. Solov'ev

In this paper we examine the dependence of emf on temperature and concentration in


sodium silicate melts in the composition range of 50-85 mole% Si0 2 and at temperatures of 600-
1400. Electrochemical concentration cells of the type Pt:02rNa20. 2Si0 2meltnxNa 20 ySi0 2melt!
02:Pt were investigated. The emf value is determined by the work of transfer of Na20 from
one melt to the other, and is a quantitative measure of the basicity of the test melt relatively
to the standard melt (Na20. 2Si02).
In accordance with literature data [1, 2], we can write

aNa~O
LlfLlI-I = R T In ~ (1)
Na,O

The character of emf variation in a heterogeneous system exhibiting liquid-liquid phase


separation can be depicted by consideration of the conditions for thermodynamic stability of the
phases with the aid of the dependence of the free energy Yon the concentration X (Fig. 1). In
accordance with the requirements of Eq. (1), the components in this scheme are oxides. If we
assume that the emf of the concentration cell is a function only of the oxygen ion activity (a02-)
[3, 4], it also appears to be possible to determine the direction of the change of emf from the
scheme in Fig. 1, since the assumption of proportional dependence of a 02- on the activity of the
alkali component does not contradict the main principles of the structure of sodium silicate
melts [5, 6].
Suppose that the compositions of the electrolytes forming the concentration cell are given
by the points a and b (Fig. 1). The tangents to these points cut off intercepts along the G axis
numerically equal to the chemical potentials of component A in melts a and b, and the cell emf
is determined by the difference tl/l 1 When the temperature is lowered from T 1 to T 2, tl/l may
remain constant in homogeneous melts or in cases where structural transformation processes
are accompanied by equal shifts of the chemical potentials of component A in the electrolyte
melts. If stable or metastable phase separation occurs at temperature T 3, the melt under in-
vestigation decomposes into two compositions: Xl and X2. Dependent on the amount of phase
X 1 formed, the composition X2 is shifted to some extent to the left of the original melt b. This
process leads to a change (decrease in the present instance) of the difference of chemical po-
tentials to tl/l3 at temperature T 3, and to tl/l4 at temperature T 4' It is reasonable to assume that
the dimensions of the heterogeneity regions playa subsidiary role in this process, and in that
69
70 A. F, BORISOV AND V. 1. SOLOV'EV

Fig. 1. Dependence of the spe-


cific isobaric chemical poten-
tials on composition and tem-
perature in melts exhibiting
liquid -liquid phase separation.
T1 > T2 > T3 > T 4,

case formation of composition fluctuations has the same influence on convergence of the chemi-
cal potentials of component A as does stable phase separation. The above scheme is also ap-
plicable to description of the character of emf variation during crystallization of melts, and is
in good agreement with experimental data [7] (Fig. 2, 1,2).
Figure 2 shows emf vs temperature curves for various concentration cells composed of
sodium silicate melts. The temperature coefficient of emf (0') for most of the cells studied re-
mains virtually constant over a wide temperature range. For melts differing little in chemical
composition 0' is zero within the limits of experimental error, and increases somewhat with in-
creasing difference between the concentrations of the electrolytes forming the electrochemical
cell. However, in no case does 0' exceed 0.03-0.05 mV/deg.
The temperature dependences of emf of concentration cells in which the electrolytes are
melts containing over 80 mole% silica (Fig. 2, 10-12) show a number of interesting features.
In distinction from all other concentration cells in this system, their temperature coefficients
are conSiderably greater. This type of emf variation is eVidently attributable to the structural
characteristics of this group of melts. A natural supposition is that the microheterogeneous
structure of the melt influences the results of emf measurements. Examination of our experi-
mental data shows that the composition of the matrix phase of low-alkali melts shifts toward
sodium disilicate with decrease of temperature, and at about 700 differs from the eutectic com-
0

position (74.5% Si0 2) by about 2% (76.5-76.0% Si0 2). According to Leko [8], the phase of higher
alkali content in microheterogeneous glasses (8-10% Na20) contains about 75% Si0 2.
The high tendency of low-alkali melts to crystallization causes considerable difficulties
in study of the temperature dependence of emf by the poly thermal method. Crystallization, like
separation of a supercooled melt into two liquid phases, affects the properties, but the changes
of properties accompanying these processes differ in magnitude and in some cases they can be
studied separately. Figure 3 shows variations of the emf of the cell Pt: 02 fNa20 . 2Si02 HNa20-
A. F. BORlSOV AND V. I. SOLOV'EV 71

zoo

100

800 1000 1200 1400r


O~----~------~-------L------~
~~~~~~~~~x-~~~~6

>
~ -100
4-<
E
<1l

-200

-300
~-1

-400

Fig. 2. Dependence of emf on temperature of concen-


tration cells of the type Pt: 021 Na20 . 2SiO:! II x Na20 .
y Si021 02:Pt(P02 = 0.21 atm). Circles and crosses
represent cooling and heating, respectively; the emf
values from the results of quenching experiments are
represented by squares. Na20 contents (mole% by
analysis): 1) 50.8; 2) 43.1; 3) 39.2; 4) 39.8; 5) 37.1;
6) 35.5; 7) 28.2; 8) 25.5; 9) 21.3; 10) 19.6; 11) 17.9; 12)
14.6. Curves 3 and 7 are based on Borisov's data [12].

14.6%, Si02-85.4%102:Pt during continuous cooling of the system. The results show that in the
1300-1400 temperature range equilibrium is established fairly rapidly in the system and the
structure of the melt is an equilibrium property. For example, if the system is held for 1 h at
1400 or 1300 the emf remains unchanged, whereas at 1200 it decreases somewhat to values
characteristic for a crystallized melt. Crystallization is indicated on curves 1 and 2 by inflec-
tions, the temperatures of which depend on the cooling rate. Nevertheless, a linear dependence
of emf on temperature persists over a considerable temperature range, which permits reliable
extrapolation and determination of emf values for supercooled melts at 700-800. As a check,
special experiments were carried out for ,measurement of emf of rapidly cooled electrochemical
cells. In order to diminish thermal inertia of the system, the electrolyte melts chilled sharply
72 A. F. BORISOV AND V. I. SOLOV'EV

I II
200 250
>
E
.......
E
<II 150

100
800 900 1000 1100 1200 1300 1400 to
Fig. 3. Dependence of emf on temperature of the concentration
cell pt:02INa20 2Si02I!Na20-14.6%, Si02-85.4%lo2: Pt. White
and black circles represent cooling and heating, respectively;
squares correspond to holding times of 1 h. Cooling rates: I) 12-
15,min; II) 3-5/min.

in individual experiments from 1200-1400 to 20 in the form of drops on the ends of platinum
electrodes, and determinations were then performed at 650-700. The emf values obtained in
these experiments, indicated by squares in Fig. 2, agree with extrapolated data within 10-12 mV.
As was to be expected, change of the clearing temperature (705) of the glass (17.9% Na20)
to the temperature of maximum opalescence (with the glass held at these temperatures) produces
only a slight change of emf (about 10 mV), which corresponds to a change of less than 1% in the
composition of the matrix phase. At 650 the previously transparent melt exhibited strong opal-
escence after 2 h. The emf remained unchanged. These results and literature data [9] show
that heat treatment of the glass at 650 leads to substantial changes of structure, although the
results of emf determinations indicate that the phase compositions remain constant. Differen-
tiation of compositions is apparently completed earlier, during cooling of the melt [10].
The emf vs composition isotherms (Fig. 4) reveal some characteristic features of the
structure of low-alkali melts at high temperatures. For example, heating of the melts to higher
temperatures leads to homogenization [11] while the emf increases regularly and tends to val-

I
400 /
/
/
300
1400 /

200 ....
t::>~
Cr)
1100
c::, 800
> 100 ~
~
E
...: I
E
<II I
0
50 80 90 Fig. 4. Isotherms of emf vs composi-
I Si0 2 mole"!o tion in the system Na20' Si02 - SiO:!.
-100 I
I
-200 1/

-300 I,
~
~
A. F. BORISOV AND V. 1. SOLOV'EV 73

ues corresponding to more homogeneous melts. The emf vs composition isotherm for 800 is
almost parallel to the abscissa in the region of low-alkali compositions, and the emf is indepen-
dent of the glass composition. This indicates that at the stated temperature the various original
melts decompose into phases of equal composition.
The above data demonstrate, in our view, that in low-alkali sodium silicate melts the ini-
tial stages of differential processes (in a melt with 14.6% Na20) occur at relatively high temper-
atures (about 1400).
The separated grouping in these melts can eVidently be classified as fluctuations of the
liquid-liquid phase separation type, the intensities of which are equilibrium with respect to
temperature.
These experimental data should probably be regarded as evidence of a strong tendency to
formation of separate grouping of stoichiometric composition in low-alkali sodium silicate melts.
The composition fluctuations intensify during cooling, and at a certain stage of their development
begin to acquire the characteristics of phases. Subsequent aggregation of the groupings leads to
macrophase separation. It seems likely, however, that there are no sharp boundaries with re-
spect to phase composition between the various stages of the melt differentiation process.
Literature Cited
1. G. S. Smith and G. E. Rindone, Glass Ind., 37 :437, 454, 456 (1956).
2. A. A. Appen, Temperature-resistant Inorganic Coatings, Izd. "Khimiya," Leningr. Otd.,
Leningrad (1967),
3. P. Didschenko and E. Rochow, J. Am. Chem. Soc., 76:3291 (1954).
4. v. L Minenko, S. M. Petrov, and N. S. Ivanova, Izv. Vysshikh Uchebn. Zavedenii, Chernaya
Met., No.7, p. 10 (1960).
5. M. L. Pearce, J, Am. Ceram, Soc., 47:342 (1964).
6. J. O'M, Bockris, Trans. Met. Soc. AIME, No, 5, p, 224 (1962).
7. O. M. Ivanova, Tr. Gor'kovsk. Politekhn. Inst. Im. A. A, Zhdanova, 23(4):28 (1967).
8. v. K. Leko, Neorgan. Mater., 3:1224 (1967).
9. N. S. Andreev and V. 1. Aver'yanov, in: The Glassy State, Proceedings of the Fourth All-
Union Conference, Izd. Nauka, Moscow- Leningrad (1965), p. 94 [English translation: The
Structure of Glass, Vol. 6, Consultants Bureau, New York (1966), p. 911.
10. v. N. Filipovich, Neorgan. Mater., 3:1192 (1967).
11. F. Ya. Gal akhov , in: Structural Transformations in Glasses at High Temperatures, Izd.
Nauka, Moscow- Leningrad (1965), p. 110.
12. A. F. Borisov, Study of Diffusion and Homogenization Processes and Structural Charac-
teristics of Silicate Melts by the emf Method, Dissertation, Gor'kovsk. Politekhn. Inst.
Im. A. A. Zhdanova (1959).
DISCUSSION
B 0 r i s ov, replying to questions by Mazurin and Filipovich, noted the high sensitivity
of the emf method to changes of structure; however, it is not possible to distinguish small
heterogeneity regions of the fluctuation type from larger heterogeneity regions if the phase
composition is constant in both cases. The method is sensitive to changes in the com-
position of the heterogeneity regions. In reply to questions by Galakhov, Varshal, and
Mazurin regarding the possibility of using the method for determination of the immiscibility
zone, Borisov said that only tentative correlation is possible between the anomalous effects
observed by the authors and the immiscibility boundary.
G a I a k h 0 v welcomed the fact that the emf method is highly sensitive to differen-
tiation of glass with respect to composition; this will apparently make it possible to establish
the existence of initial fluctuations at high temperatures, leading to phase separation on decrease
of temperature.
INFL UENCE OF HEA T TREA TMENT ON VISCOSITY OF
SODIUM SILICATE AND SODIUM BOROSILICATE GLASSES

V. P. KIyuev, G. P. Roskova, and V. 1. Aver'yanov

Mazurin [1] drew attention to the importance of studying the relationships between
the properties and structure of glasses undergoing phase separation. It is evident that viscosity,
which is more strongly dependent than any other known property on the composition of glass,
must also be the most sensitive to variations of glass structure. Before extensive investiga-
tions of the influence of composition of glasses exhibiting immiscibility on their viscosity it is
necessary to have a clear idea of the general relationships governing the viscosity of such
glasses during heat treatment. The increase of viscosity of sodium borosilicate glasses as the
result of heat treatment has been noted earlier [2-4]. However, even investigators who cor-
related the viscosity variations with heterogeneity of the glasses did not consider the influence
of the character of the heterogeneous structure on viscosity.
In the present investigation an attempt was made to correlate certain aspects of the tem-
perature dependence of viscosity of glasses exhibiting immiscibility with their electron-micro-
scopic structure, with reference to glasses of the compositions 16Na20. 84Si0 2 and 8Na20. 32B 20 3
60Si02 (moleo/r).*
Viscosity of the sodium silicate glass was measured by the fiber elongation method [5], by
central bending of a rod [6], and by the indentation method [7]. In all cases the deformation vs
time curve was recorded automatically [8].
The bending method was used for studying the dependence of viscosity on the time of hold-
ing at constant temperatures in the range from 300 to 600 0
Each measurement was carried out

with a fresh quenched specimen and continued until the viscosity became constant. It was found
that a constant viscosity becomes established most rapidly at 455 (this is the t g region for the
0

given glass). At 600 the tendency to increase of the viscosity persists even after 8h.
0

Examination of the results obtained by the fiber elongation methodt with the same speci-
men over the entire temperature range (Fig. 1, 1) shows that in the range from 500 to 600 the0

viscosity increases approximately from 10 14 to 10 14 7 P. Points corresponding to constant vis-


cosities determined by the bending method fit on this curve.
Determinations with quenched specimens by the indentation method (Fig. 1, 3) showed that
in this case the viscosities are lower by 1.5-2.5 orders of magnitude than the results given by

"The authors thank V. S. Molchanov and Yu. N. Kondrat'ev for providing the sodium silicate
glasses.
tThe measurements were performed by S. M. Rekhson.
74
V. P. KLYUEV, G. P. ROSKOVA, AND V. I. AVER'YANOV 75

log 1)

17

15

Fig. 1. Temperature dependence of vis-


13 cosity of the 16Na20' 84Si02 glass. 1,3)
Specimens not subjected to heat treat-
ment: squares - fiber elongation method;
11 crosses - rod bending method; black cir-
cles - indentation method; 2) specimens
held for 3 h at 700: triangles - bending
9 method; white circles - indentation method.
300 '100 500 600

the fiber elongation and bending methods. In these tests a ridge of displaced glass was formed
around the indentor imprint, whereas in the case of nonseparating glasses a crater is formed
around the indentor.
The viscosities of specimens subjected to heat treatment for 3 h at 700 (the temperature
of maximum opalescence of 16Na 20 .84Si02) glass were then measured. In this case the bend-
ing and indentation method gave results in agreement (Fig. 1, 2). It follows from the diagram
that the viscosity of the glass decreased considerably after heat treatment. In the region of 600
this decrease exceeded 5 orders of magnitude. Formation of a ridge around the indentor was
not observed on these specimens.
The temperature dependency of viscosity of the sodium borosilicate glass was also de-
termined with the use of quenched specimens and specimens subjected to heat treatment for 4
hat 700, i.e., in the region of maximum opalescence. It was found that with this glass the
indentation and bending methods give similar results. In determinations by the indentation
method the quenched specimens were put into the previously heated furnace of the viscosimeter,

log 1)

13

Fig. 2. Time dependence of viscos- 12


ity of the 8Na20' 32B 20 3 60Si02 glass.
1, 3) Specimens quenched from the
melt: 1) 472; 2) 502; 3) 644; 4,5)
specimens held at 700 for 4 hand
11
then quenched: 4) 575; 5) 644.

o 120 2~0 300 ~80 000


T. min
76 V. P. KLYUEV, G. P. ROSKOVA, AND V. I. AVER'YANOV

and the determination was started when a constant temperature had been reached. It was found
that the viscosity increases in the region from 500 to 600 by about 2 orders of magnitude. About
6-8 h is required for establishment of a nominally constant value of viscosity (Fig. 2, 2). This
time is shortened severalfold on increase of temperature (Fig. 2, 3). Constant viscosity values
are established much more rapidly in the case of specimens subjected to previous heat treat-
ment (Fig. 2, 4,5).
The curves in Fig. 3 are plotted from the final values of the glass viscosity measured
by the indentation method on specimens without preliminary heat treatment (1) and on speci-
mens held for 4 hat 700 (4). Both curves have a plateau in the region from 500 to 600, and
curve 4 is somewhat below curve 1.
The same figure shows the temperature dependence of viscosity determined by the bend-
ing method under dynamic conditions (heating rate 5-6 per min) with quenched specimens (2) and
specimens held at 700 (5). Curves 3 and 6 represent the viscosity variations of these speci-
mens on cooling (static conditions). It is seen that curves 2 and 5 also have plateaus. The
viscosity of the glass is considerably higher on cooling than on heating, and at 550 the viscosity
0

is higher by about 3 orders of magnitude (cf. curves 2 and 3).


The viscosity increase shown by these glasses in the 500-600 region should be attributed,
in our opinion, to development of phase separation with formation of a structure of two inter-

log 11

f3

Fig. 3. Temperature dependence of vis-


12 cosity of the 8Na20' 32B20 3 60Si02 glass.
1) Quenched specimens, indentation meth-
od, heated under static conditions; 2) quench-
ed specimen, bending method, heating
at the rate of 5-6/min; 3) cooling from
11
660 under static conditions: white cir-
cles - indentation method; black circles -
bending method; 4) 700, 4 h, indentation
method, heating under static conditions;
fO 5) 700, bending method, heating at the
rate of 5-6/min; 6) same specimen as in
5, cooling from 670.

9 ________ ________
L -_ _ _ _ _ _ __

400 500 600 700 t


V. P. KLYUEV, G. P. R08KOVA, AND V. I. AVER'YANOV 77

Fig. 4. Electron micrographs of the 8Na20 32B 20 3 608i02 glass. Heat


treatment: a) 700, 4 h; b) 700, 4 h, followed by 550, 10 h.

penetrating phases, each of which is continuous. The more viscous siliceous framework deter-
mines the viscosity of the glass as a whole. The existence of such a siliceous framework at the
first stages of phase sepa ration in low-alkali sodium silicate glasses close in composition to those
described in this paper is shown in electron micrographs in the paper by Andreev et a1. [9]. Heat
treatment of these glasses at higher temperatures leads not only to decrease of the amount of
the siliceous phase but also to formation of isolated inclusions [9, 10]. This produces a sharp
decrease of viscosity (cf. curves 1 and 2 in Fig. 1). The observed formation of a ridge around
the indentor imprints on specimens of sodium silicate glass not subjected to heat treatment can
also be attributed to the presence of a framework structure. Brittle destruction of the frame-
work apparently occurs in the region of contact of the indentor with the specimen surface, and
the more fusible phase is pressed out as the indentor sinks in. In such cases the indentation
method gives lower viscosity values than the fiber elongation and bending methods (Fig. I, 1,3),
and the viscosity depends on the load applied to the indentor.
Heat treatment of the sodium borosilicate glass at 700 for 4 h leads to formation of a
framework structure; this is clearly seen in the electron micrograph (Fig. 4a). A framework
structure was not detected with the aid of the electron microscope in a specimen held for 6 h at
550, but the course of the temperature dependence of viscosity (Fig. 3, 1) indicates that forma-
tion of the framework already occurs in the 500-600 region. The electron micrograph in Fig .
0

4b shows the structure of sodium borosilicate glass held at 700 for 4 h, with subsequent heat
0

treatment at 550 for 10 h. It is seen that the more fusible phase has well-defined heterogeneous
0

structure , which is probably the consequence of secondary phase separation [11]. Apparently,
when a specimen previously held at 700 is subjected to further heat treatment at 550 the excess
0 0

amount of silica-rich phase separates out in the sodium borate phase, and partially passes into
the existing highly viscous framework, increasing its cross section, while the rest forms drop-
lets within the more fusible phase. Evidently, thickening of the highly viscous framework leads
to increase of the glass viscosity. However, in this cse the amount of high-melting phase con-
stituting the framework is less than if the heat treatment proceeds from the start in the 500-600 0

region (which accounts for the difference in the positions of curves 1 and 4 and of 3 and 6 in Fig.
3).

It can be concluded from the results of these observations that the viscosity of glasses
undergoing phase separation is determined not only by the presence of two phases differing in
chemical composition (and consequently in rheological properties) in such glasses, but also by
the character of their distribution; the decisive factor is whether the less fusible phase forms a
continuous framework or is distributed in the more fusible phase in the form of separate inclu-
sions. The viscosity of the glass must be determined mainly by the viscosity of the framework
in the first case, and by the viscosity of the more fusible phase in the second.
78 V. P. KLYUEV, G. P. ROSKOVA, AND V. I. AVER'YANOV

The viscosity- temperature relationships found by us are probably characteristic to var-


ious degrees of most glasses (including commercial ones) exhibiting phase separation; this must
be taken into consideration both by investigators and by technologists.
In conclusion, it should be stressed that even now comprehensive study of the dependence
of viscosity on temperature and time can be regarded as an indirect method for studying the
heterophase structure of glass.
Literature Cited
1. O. V. Mazurin, this volume, p. 37.
2. T. Abe, J. Am. Ceram. Soc., 35 :284 (1952).
3. M. Watanabe and T. Moriya, Rev, Elec. Commun. Lab., 9:50 (1961).
4. A. L. Zijlstra, Phys. Chem. Glasses, 4:143 (1963).
5. A. Napolitano and E. G. Hawkins, J. Res. U. S. Natl. Bur. Standards, 68A :439 (1964).
6. H. E. Hagy, J. Am. Ceram. Soc., 46:93 (1963).
7. V. P. Klyuev and V. B. Sakhov, Tr. Inst. Stekla, "Steklo," No.2, p. 106 (1967).
8. B. V. Efimov, V. P. Klyuev, and V. B. Sakhov, in: Methods for Measurement of Thermal
Expansion of Glasses and of Metals Sealed to Them, Proceedings of the First All-Union
Symposium, Izd. Nauka, Leningrad (1967), p. 176.
9. N. S. Andreev, V. I. Aver'yanov, and E. A. Porai-Koshits, in: Structural Transformations
in Glasses at High Temperatures, Izd. Nauka, Moscow- Leningrad (1965), p. 59.
10. Tran Thach-Lan, Verres Refractaires, 20:8 (1966).
11. E. A. Porai-Koshits, and V. 1. Aver'yanov, this volume, p. 28.

DISCUSSION
R 0 s k 0 va said in reply to Ermolenko that at temperatures up to 455 the observed
0

deformation apparently does not correspond to viscous flow (the formation is predominant-
ly of the delayed-elastic type). In relation to Nemilov's question Roskova said that the
inflection on the curve and increase of viscosity are due to formation of a framework in
the glass. The formation of a ridge around the indentor imprint in indentation tests is
only tentatively attributed to the heterogeneous structure of glasses, as there is no direct proof
of this. However, it is significant that a ridge appears only when there are grounds for suppos-
ing that a weakened highly viscous framework is present in the glass. The appearance of a
ridge does not appear to be due to the use of a conical indentor. Volume crystallization of the
glasses was not observed during the viscosity measurements; only a crystalline layer formed
on the surface, and this was removed before each successive measurement.
N em i I ov, commenting on the paper by Klyuev, Roskova, and Aver'yanov, doubted
whether in their measurements of viscosity with the use of specimens differing in ther-
mal history they obtained viscosity curves diverging with increase of temperature. The
curves should merge rather than diverge with approach to the critical temperature. This
result can probably be attributed to inapplicability of the conical indentor method with short in-
dentation times to measurements of high viscosities, and possibly also to crystallization. With
this experimental technique the area of contact between the indentor and the specimen alters
with time, the shear stress changes, and flow ceases to be Newtonian. This may apparently also
be the cause of formation of a ridge at the edges of the imprint.
Replying to Nemilov, M a z uri n stressed the novelty and reliability of the effects
observed by the members of his laboratory. The course of the curves diverging with
increase of temperature could not be followed experimentally further owing to intensive crys-
tallization above 620 The viscosity was measured by two independent methods, and the equi-
0

librium value was reached. The specimens were quenched from their clearing temperatures.
V. P. KLYUEV, G, P. ROSKOVA, AND V. 1. AVER'YANOV 79

Mazurin doubted whether Nemilov was right in assuming that the activation energy deter-
mining the rate of formation of a phase-separation structure must be necessarily of the
same order as the activation energy of viscous flow. He agreed that with increase of the heat-
treatment temperature the viscosity curves of quenched glass and of glass treated at 700 should
0

coincide. In Mazurin's opinion, "the viscosity of a glass of a given composition, like any
other property, should be constant at a specified temperature. However, glasses the structure
of which changes substantially as the result of heat treatment may not necessarily have constant
viscosity although in principle, with a sufficiently long holding time, the structure (and, corre-
spondingly, the viscosity) should reach a state of equilibrium." The speaker stressed that with
the use of a conical indentor it was possible to perform measurements in the region where other
methods are inapplicable; in most cases where this method can be compared with others the re-
sults are in agreement. Deviations occurred only when the formation of the imprint was accom-
panied by appearance of a ridge. Interpretation of this effect is controversial.
K I y u e v emphasized that the indentation viscosimeter was calibrated against glasses
which do not form a ridge at the edges of the imprint. The formation of the ridge refer-
red to in the paper by Klyuev, Roskova, and Aver'yanov is explained by him as follows:
when a rigid high-silica framework and an interstitial liquid are present, the liquid is
pressed out by the indentor during deformation, and forms the ridge. Since the viscosimeter
was calibrated under other conditions, the results obtained during formation of this type of im-
print cannot be compared with the results given by absolute methods of measurement, as they
do not characterize the viscosity of the glass. Klyuev pointed out that the form of the
imprint must be taken into account in viscosity measurements by the indentation method.
REGIONS OF METASTABLE PHASE SEPARATION IN
TERNARY ALKALI BOROSILICATE SYSTEMS

F. Ya. Galakhov and O. S. Alekseeva

Only approximate information is available in the literature on regions of metastable phase


separation in the systems Li 20- B20 3- Si02 and Na20- B20 3- Si02 [1-31. The binodal in the
pseudobinary system Na20. 4B 20 3- Si02 was constructed by Rockett and Foster [41. The purpose
of the present work was to construct the surfaces of the metastable phase-separation volumes
in the phase diagrams of the two above-named systems.
The boundary of metastable phase separation was determined experimentally as follows.
The original glasses were made by fusion at temperatures up to 1600 of mixtures of the origi-
nal components (Li 2C0 3 Na2C03, H3B0 3, and Si0 2) in a platinum furnace, in the form of pressed
tablets in platinum foil packets; the more refractory compositions were made by fusion of mol-
ded rods in an electric arc. Small fragments of the original glasses 3-8 mm 3 in volume, wrap-
ped in platinum foil, were suspended from a thermocouple on platinum hooks, lowered into a
vertical tubular furnace heated to a known temperature, and after the required time dropped
into water in a vessel. A crucible furnace was used for prolonged heating. The quenched spec-
imens were crushed in a steel mortar to fragments 0.5 mm and smaller in size, and examined
on a slide with lateral illumination under a microscope. An azure color in the glass was a
clear indication of opalescence. If metastable phase separation was followed by crystallization,

b
a
SiOz
SiOz

82 8Z 0
Fig. 1. Isotherms of metastable phase separation on phase diagrams of the
systems. a) Li 20-B 20 3-Si02; b) Na20-B203-Si02.
80
F. Ya. GALAKHOV AND 0. S. ALEKSEEVA 81

the color of the residual glass could be distinctly observed. This method for determination of
the boundary of metastable phase separation was verified with the aid of electron microscopy.
The distinct increase in the size of the "drops" in an opalescent specimen in comparison with
a nonopalescent specimen of the same composition indicated that opalescence can serve as a
criterion of phase separation. The data obtained in this way were used for plotting isotherms
of metastable phase separation on phase diagrams of the systems (Fig. 1, a and b). The sur-
faces of the immiscibility volumes in both systems have upper critical points lying within the
triangular diagram. They correspond to the following compositions (mole%) Li 20 13.5, B20 3
19, Si02 67.5 (temperature 995) and Na20 6, B20 3 31, Si02 63 (temperature 745). A small por-
tion of the immiscibility volume in the lithium system lies above the liquidus, and the approxi-
mate position of the region of stable immiscibility is hatched.
Among the various possibilities of using the metastable immiscibility region on the phase
diagram, we consider obtaining information on the composition of coexisting glassy phases. In a
two-component system these compositions are found quite accurately from the phase-separation
binodal. Determination of the compositions of coexisting glass phases in a three-component
system is more complicated. It is not sufficient to establish the surface of the immiscibility
volume on the phase diagram of the system in order to find the compositions of equilibrium (al-
beit metastable) glass phases. It is also necessary to draw tie lines, which in the general case
run fanwise, for the given isothermal section of the ternary diagram. The direction of one of
the extreme tie lines, closest to the side of the triangle adjacent to the phase-separation iso-
therm is almost parallel to that side. The position of the other extreme line is determined by
the position of the critical point on the isothermal binodal, where the tie line degenerates to a
point. The position of the tangent to this point indicates the required direction to which its near-
est tie line tends. If the immiscibility region in the isothermal section is not adjacent to the
side of the triangle and the isotherm forms a closed curve, the positions of the two extreme tie
lines are determined by the two critical points of the isotherm. Consequently, in order to find
the tie lines in isothermal sections of the immiscibility volume the position of the critical line
on its surface must be known. This line joins the critical points in the two binary systems ad-
jacent to the immiscibility volume. If this volume has a triple critical point, the critical line
also passes through this point. For finding the critical line in the phase diagram of the system
Na20- B20 3- Si02 the triple critical point K (Fig. Ib) and the critical point Kl in the system
Na 20- Si02 [5] are known. The critical point in the system B20 3- Si02 is not known, and its posi-
tion is estimated approximately [6].*
Figure 2a, which is an isothermal section of the metastable immiscibility region of the
ternary diagram, shows the critical line K1KK 2. A high degree of accuracy cannot be claimed
in this case for the compositions of the coexisting glasses, indicated by the ends of the tie lines.
Errors in the compositions may arise mainly in consequence of the arbitrary choice of the posi-
tion of the critical point in the B 20 3- Si02 system. Nevertheless, approximate information was
obtained by this method on compositions of the glasses into which the original glass separates
during heat treatment at definite temperatures.
It is of interest to compare the direction of the tie lines found by the reasoning presented
above with the experimental data of Tran Thach-Lan [7]. who studied the chemical composition
of phases formed during heat treatment of sodium borosilicate glasses. The phases were sep-
arated by treatment of the powdered glasses with 6 N HCI solution followed by dilute HF. The
compositions of the separated phases were then determined by chemical analysis. These data

*The position of the metastable phase-separation binodal was calculated by Charles and Wagstaff
[6]. but only three experimental points are given and owing to the flat course of the binodal the
critical point was not determined.
82 F. Ya. GALAKHOV AND O. S. ALEKSEEVA

a b
SiOz SiOz

Fig. 2. Positions of tie lines on an isothermal section (700)


of the metastable immiscibility region in the system Na20-
B20 3 - Si~. a) Our data; b) after Tran Thach- Lan [7].

are shown in a ternary phase diagram (Fig. 2b) which represents an isothermal section of the
metastable immiscibility region at 700. It was at this temperature that the specimens were
held by Tran Thach-Lan [7] to produce phase separation. The straight lines drawn through the
original compositions (marked by points) are tie lines the ends of which indicate the experimen-
tally determined compositions of the phases formed. The directions of these tie lines and of
those obtained in the present investigation (Fig. 2a) are generally in fairly good agreement.
However, while the directions of the tie lines are approximately the same their sizes are dif-
ferent: the tie lines found from the phase diagram are considerably shorter. The compositions
of the coexisting phases determined from the phase diagram are eVidently in better agreement
with their true compositions than those determined experimentally. It is practically impossible
to find a solvent which, while effectively dissolving one of the coexisting glasses, would not have
some action on the other, especially in view of the high degree of subdivision of the coexisting
glassy phases and hence of their enormous interfacial area. Partial dissolution or leaching of
"insoluble" glass leads tooverestimationofthe difference between the phase compositions after
experimental separation. The error in the experimental determination, manifested in increased
length of the tie lines, becomes quite evident on examination of the positions of the points cor-
responding to compositions enriched with boric anhydride in Fig. 2b. A broken line is obtained
if these points are joined; this contradicts the theory of liquid-liquid phase separation, accord-
ing to which the ends of the tie lines should lie on an isothermal binodal having smoothly vary-
ing curvature. Incidentally, it should be noted that the interpretation given in [7] of the results
would be more general and convincing if it was based on the theory of liquid-liquid phase sep-
aration. In particular, the author's attempt to correlate the B20a/Na20 ratio in the products
formed by separation of the glass with definite stoichiometry is meaningless in the light of the
phase theory.
The positions of lines joining the compositions of coexisting liquids or glasses have been
discussed in the literature [8, 9]. It was shown experimentally that in the case of ternary bo-
F. Ya. GALAKHOV AND 0. S. ALEKSEEVA 83

rosilicate systems with barium and calcium oxides, in which phase separation is stable, the tie
lines are not parallel to the B20 a- Si02 side of the triangular diagram, although the immisci-
bility region is very close to it. On the other hand, as was shown above, in alkali borosilicate
systems the tie lines are approximately parallel to the B20 a- Si02 side. This difference in the
positions of the tie lines can be explained as follows. Binary systems, such as calcium oxide
with silica and calcium oxide with boric anhydride, have stable regions of immiscibility, form-
ing a continuous region of phase separation in the ternary system between the CaO- Si02 and
CaO- B20 a sides of the triangle; in this region the direction of the tie lines must vary from
parallel to the CaO- Si02 side to parallel to the CaO- B20 a side. This region can become ad-
jacent to the B20 a- Si02 side only at some much lower temperature, but in that case three-phase
metastable equilibrium will arise between three coexisting glassy phases. On the other hand,
in R 20- B 30 a- Si02 ternary systems the continuous region of metastable phase separation ex-
tends between the R 20- Si02 and B 20 a- Si02 sides, and merging with the region of metastable
phase separation in the R 20- B20 a system, if such a region exists in that system, can occur only
at a lower temperature. Therefore at the highest temperatures in the region of metastable
phase separation in R 20- B20 a- Si02 systems the direction of the tie lines must vary from paral-
lel to the R 20- Si02 side to parallel to the B20 a- Si02 side.
Literature Cited
1. B. S. R. Sastry and F. A. Hummel, J. Am. Ceram. Soc., 42:81 (1959).
2. B. S. R. Sastry and F. A. Hummel, J. Am. Ceram. Soc., 43:23 (1960).
3. 0. S. Molchanova, Proceedings of the Second Conference on the Structure of Glass, Izd.
Akad. Nauk SSSR, Moscow- Leningrad (1955), p. 141 [English translation: The Structure
of Glass, Vol. 1, Consultants Bureau, New York (1958), p. 109].
4. T. J. Rockett and W. R. Foster, J. Am. Ceram. Soc., 49:30 (1966).
5. Y. Moriya, D. H. Warrington, and R. W. Douglas, Phys. Chem. Glasses, 8:19 (1967).
6. R. J. Charles and F. E. Wagstaff, J. Am. Ceram. Soc., 51:16 (1968).
7, Tran Thach-Lan, Verres Refractaires, 19(6):416 (1965).
8. G. W. Morey and E. Ingerson, Am. Mineralogist, 22:38 (1937).
9. E. M. Levin and G. M. Ugrinic, J. Res. U. S. Natl. Bur. Standards, 51:37 (1953).

DISCUSSION
Gal a k h 0 v, replying to Mazurin, said that the third phase may arise at tempera-
tures below the critical temperatures for the binary systems. The configuration of the
tie lines is reminiscent of triangular and is not very complex. Secondary separation may occur
in the ternary, as in the binary, system. Nonequilibrium character of the states must be taken
into account here.
In reply to questions by Filipov-ich and Porai-Koshits, Galakhov said that the posi-
tion of the critical point within the triangular diagram was determined experimentally.
Formation of a third phase, but not as the secondary effect mentioned above, must proceed very
slowly because the temperatures are too low. The tie lines found in the present investigation are
of the equilibrium type.
COMPOSITION OF COEXISTING PHASES IN
SODIUM BOROSILICATE GLASSES
EXHIBITING PHASE SEPARATION

B. G. Varshal

The heterogeneous structure due to liquid-liquid phase separation in low-alkali borosili-


cate glasses of the Si02- B20 3- Na20 system is now undoubted, but the compositions of the coex-
isting glass phases in these glasses have not been conclusively determined. The low chemical
stability of one of the phases gives the impression that the components of this phase can be iso-
lated completely by treatment with particular chemical reagents (acid or alkali), and its chemical
composition determined. However, this disregards the possibility of precipitation of some of the
components within the porous framework, or the possibility of action on the components of the
framework, leading to dissolution and addition to the other phase.
We will consider the compositions of coexisting phases in glasses of this system, using
data on binary systems.
The system Si02- Na20 exhibits metastable phase separation [1] with a critical tempera-
ture of 850 and an immiscibility region extending to about 2 0-22 mole% Na20. Charles and Wag-
0

staff [2] showed the existence of metastable phase separation with a critical temperature of
. . . 520 in the system Si02- B20 3 Immiscibility involves almost the entire system, i.e., the co-
0

existing phases are vitreous silica and boric anhydride.


The work of Wagstaff and Charles [3] on the B203-Na20 system confirmed the sug-
gestions of Myuller [4] and Vogel [5] regarding the occurrence of phase separation in
it. Here the critical temperature is about 615 and the width of the immiscibility region corre-
0

sponds to approximately 30 mole% Na20.


Thus, all three binary systems in the ternary system have considerable immiscibility re-
gions. A part of the diagram of the system Si02- B20 3- Na20 is shown in Fig. 1. It is seen that
in this part of the diagram, adjacent to the Si0 2- B20 3 system, there are four immiscible liquids
differing substantially in structure: high-silica glass, consisting mainly of tetrahedral [Si04]
groups; high-borate glass, consisting mainly of triangular [B0 3] groups; sodium silicate glass
of the approximate composition Na20 .4Si02, consisting of tetrahedral [Si04] groups and octahe-
dral [NaOs] groups; sodium borate glass of the approximate composition Na20' 2.3B 20 3, consist-
ing of tetrahedral [B0 41Na groups and triangular [B0 3] groups (in accordance with the views of
Beakenkamp [6]).
Evidently, one important problem is incompatibility of tetrahedral [Si04] and [B0 4]Na
groups. The cause of this must be sought in the nature of the structure of the [B0 4] tetrahedron

84
B.G. VARSHAL 85

Fig. 1. Regionof immiscible glasses


in the system Si0 2- B 20 3 - Na20.
Dividing line: 1) after Charles
and Wagstaff; 2) based on our data.

in which, as was shown by Nemilov [71, the bonds are not equivalent as in the [Si0 4] tetra-
hedron but comprise two stronger and two weaker bonds. The asymmetric structure of the
(B0 41 tetrahedron makes it incompatible at low temperatures with the highly polymerized [Si0 41
structure. However, this does not rule out compatibility of the tetrahedra at elevated tempera-
tures, at a low concentration (low degree of polymerization) of [Si0 41 groups, or in crystalline
borosilicates, in which forces exist making these groups compatible.
For a condensed incompressible system the Gibbs phase rule is written as follows:

!=n-r+1.

For an isothermal section the number of degrees of freedom is decreased by one, and the
phase rule becomes

!t=n-r.

Therefore the number of coexisting phases in this three-component system under isother-
mal conditions cannot exceed three. According to Fig. 1, there are four immiscible liquids in
this part of the system. We therefore conclude that separating glasses in this part of the sys-
tem must have three-phase structure, and this region should be divided into two elementary
triangles, as was done by Charles and Wagstaff (2], by a line drawn from the Si02 corner
to the composition Na20. 2.3B 20 3 (Fig. 1, line 1). It follows that low-alkali sodium borosilicate
glasses must have a phase composition consisting of a high-silica framework containing
glassy sodium borates and boric anhydride. High-alkali glasses of the second region consist of
a high-silicate framework containing coexisting sodium silicate and sodium borate glasses.
However, this interpretation does not explain all the properties of sodium borosilicate
glasses. This ll;pplies primarily to the amount of Si0 2 dissolVing in the sodium borate component.
According to Zhdanov [81 and Tran and Sella [9], the Si02 content in this phase is not less than
about 40%.
This and certain other facts can be explained on the assumption that the dividing line runs
not through the Si02 corner (Fig. 1, line 1) but through the B20 3 corner (line 2). The region under
investigation is then divided into two parts. The first part contains coexisting high-silica glass,
86 B. G. VARSHAL

sodium silicate glass of the approximate composition Na20 4Si0 2, and high-borate glass. The
structure of high-silica glasses in this part of the diagram consists of a high-silica framework
containing two glass phases: sodium silicate and a high-borate phase. Sodium silicate, sodium
borate, and high-borate glass phases must coexist in the second part.
Our suggested alternative viewpoint on low-alkali sodium borosilicate glasses helps to
clarify the mechanism of the influence of Al 20 3 on the structure of these glasses, put forward
by Zhdanov and Koromal'di [10]: Al 20 3 introduced into the glass does not take Na20 away
from B20 3 but diminishes phase separation in the system Si02-Na20, giving rise to [AI0 4]
Na groups located in the tetrahedral network of the high-silica glass, while B20 3 passes into the
more fusible phase. In presence of sodium silicate glass B20 3 enters this phase, and when Al 20 3
has eliminated the separate sodium silicate phase B20 3 enters the newly formed sodium alumino-
silicate glass phase where it forms filaments and threads, as in glasses of the binary system
Si02 - B20 3
Literature Cited
1. Y. Moriya, D. H. Warrington, and R. W. Douglas, Phys. Chem. Glasses 8:19 (1967).
2. R. J. Charles and F. E. Wagstaff, J. Am. Ceram. Soc., 51:16 (1968).
3. F. E. Wagstaff and R. J. Charles, Ceram. Bull., Vol. 45, No.4, T126, Ref. 17-G-66 (1966).
4. R. L. Myuller, Zh. Tekhn. Fiz., 25:1868 (1955); 26:2614 (1956).
5. W. Skatulla, W. Vogel, and W. Wessel, Silikattech., 9:51 (1958).
6. P. Beakenkamp, in: Physics of Noncrystalline Solids, Amsterdam (1964), p. 512.
7. S. V. Nemilov, Neorgan. Mater., 2:349 (1966).
8. s. P. Zhdanov, Structure of Porous Glasses and Structural Transformations in Sodium
Borosilicate Glasses, Doctoral disseration, Inst. Khim. Silikatov, Akad. Nauk SSSR,
Leningrad (1959).
9. T. Tran and C. Sella, Compt. Rend., 259(6):1325 (1964).
10. S. P. Zhdanov and E. V. Koromaldi, Compt. Rend. VII Congr. Intern. du Verre, No. 302
(1965).
DISCUSSION
E v s t r 0 p , e v, commenting on Varshal' s paper, said that the concepts put forward are
valid only if the phase rule is strictly obeyed.
Replying to Evstrop'ev, Va r s h a I points out that he started with low temperatures
and equilibrium conditions and then passed to the nonequilibrium state; his discussion was
concerned not with the existence of a particular phase but with the question of whether so-
dium is combined chemically with boron or with silicon.
Po r a i - K 0 s hit s reminded that as long ago as 1940, in studies of sodium boro-
silicate glasses without preliminary leaching, x-ray diffraction data showed clearly that glasses
with low sodium contents are composed of sodium borate, i.e., that sodium is combined in them
with boron rather than silicon.
Z h dan 0 v noted the originality of the ideas put forward in Varshal's paper. How-
ever, he preferred the conventional view on the manner of entry of the alkali oxide into bo-
rosilicate glass. Sodium oxide apparently interacts with boric anhydride first, because the en-
thalpy of formation of borates is higher than of silicates.
Galakhov, with reference to Varshal's paper, emphasized the necessity of exam-
ining the phase diagram for determination of the compositions and nomenclature of dif-
ferent phases in multicomponent systems.
INFLUENCE OF COMPOSITION ON VISCOSITY AND
TRANSFORMA TION TEMPERATURES OF
BOROSILICATE GLASSES EXHIBITING PHASE SEPARATION

M. V. Strel'tsina, O. V. Mazurin, and A. S. Totesh

The extensive potentialities of studies of physical and chemical properties of glasses with
a tendency to metastable phase separation, in relation to elucidation of their structure, have
been demonstrated by Mazurin [1],
In this communication we report the results of a simultaneous study of low-temperature
viscosity and transformation temperatures.
It is known that the transformation temperatures (tg) of single-phase glasses are usually
close to the temperatures at which the viscosities are 10 13 P (as proposed by Solomin, we desig-
nate this temperature t 13 ).
We found that t 13 of borosilicate glasses exhibiting phase separation is considerably above
tg in many cases. The main purpose of the present work was systematic study of this phenome-
non.
Viscosity in the transformation range was studied by the indentation method [2], and tg
was determined from dilatometric curves [3]. All compositions are given in molar percentages
by synthesis; the deviations of analytical data are within the limits of experimental error.

to
600

Fig. 1. Comparison of the influence


of replacement of B20 a by Na20 in
sodium borosilicate and sodium bo-
rate glasses. 1) tg of x Na20 (40-
x) B20 a 60Si02 glasses; 2) 1ta of
NazO, mole%
x Na20. (40 -x) B20 a 60Si02 glass-
300 es; 3) t 1a of x Na20, (100-x)B 20 a
8 12 fj 20 25
glasses.


0 0.5 t.O 1.5
NazOjIJz 3
87
88 M. V. STREL'TSINA, 0. V. MAZURIN, AND A. S. TOTESH

Figure 1 shows the dependence of tg and t 13 on composition of borosilicate glasses con-


taining 60% Si02, in which B20 3 is progressively replaced by the alkali oxide. It is seen that in
the low-alkali region the dependences of t g and t 13 on composition follow different courses. Re-
placement of B20 3 by sodium oxide leads to steady increase of t g, whereas the viscosity (td
curve has a maximum at 8% Na20. For glasses of the composition 8Na20 32B 20 3 60Si0 2 the
difference t 13 - t g is 140 Similar results are obtained for potassium glasses, but here the
0

maximum viscosity corresponds to glasses with 6% K20 and the difference t 13 - tg is 80 0


The different characters of the dependences of tg and t13 on composition of alkali boro-
silicate glasses suggest that these relationships may be determined by properties of different
phases. It seems evident that in the case of a two-phase glass in which the highly viscous phase
forms a continuous framework (the character of distribution of the low-viscosity phase is im-
material in this case) the viscosity of the glass must be determined by the viscosity of the high-
viscosity phase, and the transformation temperature by the viscosity of the low-viscosity phase.
In distinction from one-phase glasses, the transformation temperatures of two-phase
glasses determined from the temperature dependences of different properties need not necessar-
ily coincide. For example, the break on the conductivity vs temperature curve for a two-phase
glass with closed conducting inclusions should correspond to tg of the phase having the lower
conductivity (usually the high-viscosity phase). However, the break on the dilatometric curve
must invariably correspond to the temperature of a sharp change in the coefficient of thermal
expansion (CTE), i.e., to tg of the more fusible phase: Of course, in general a break correspond-
ing to t g of the more refractory phase should also appear on the dilatometric curve at higher
temperatures. However, for a number of cases this break is indistinct and is not considered
here.
Direct experimental confirmation of the foregoing was found in a paper by Gomel'skii et
al.[4]. Thermal-expansion curves of glass fibers made from two glasses of different viscosities
clearly show that tg of the fiber is close to tg of the more fusible component.
The effect of replacement of B20 3 by Na20 in sodium borate [5] and sodium borosilicate
glasses on tg and t13 at equal Na20 j B20 3 ratios are comparE:d in Fig. 1. It is seen that up to the
ratio Na20 j B20 3 = 0.25 the viscosity of the less viscous phase in the three-component glass
tends toward the viscosity of the two-component glass. At the same time the difference between
the viscosities of the high-viscosity and low-viscosity phases increases rapidly. Both factors
indicate intensified differentiation of phase compositions with increase of the Na20 content to the
ratio Na20jB203 = 0.25. In the region of greatest phase separation the borate phase eVidently
has a low Si02 content; this is indicated by convergence of the tg curve of the three-component
glass with the t 13 curve of the two-component glass. On further increase of the Na20 contents
in the glasses the tg curve of the sodium borosilicate glass and the t 13 curve of the sodium bo-
rate glass diverge, indicating increase of the Si02 content in the more fusible component of the
three-component glass. The sharp increase of t13 of sodium borosilicate glass on introduction
of Na20 above the ratio Na20jB203 '" 0.28 probably indicates breakdown of the continuous high-
viscosity framework associated with decrease of the volume of the more viscous phase.
Figure 2 shows the temperature dependence of viscosity over a wide temperature range
for two three-component sodium borosilicate glasses, the composition of one of which lies in the
region of maximum immiscibility (8% Na20), while the other (11% Na20) has weak tendency to
phase separation. The rotary viscosimeter of the Institute of Silicate Chemistry was used for
the measurements. The sharp break in the experimental curve of the glass in which existence
of a highly viscous framework is presumed, approximately corresponding to the clearing temper-
ature, is in good agreement with our views on the relation between structure and properties of
such glasses.
M. V. STREL'TSINA, 0. V. MAZURIN, AND A. S. TOTESH 89

log 1)

fO.O
1

8.0

5.0

4.0

J'----~~
7.0 B.o 9.0
L '10'1
10.0
l/deg
11.0 12.0

T '
Fig. 2. Dependence of the viscosities of glasses on temperature.
1) 8Na20 . 32B 20 3 60Si02; 2) llNa20' 29B20 3 60Si02.

It is well known that t d, the dilatometric temperature at which deformation under load
begins, is close to tu. The difference td-tg for ordinary glasses is generally in the range of 40-
60 [6-81. As we have already noted, t i 3 of two-phase glasses with a highly viscous framework
0

is considerably higher than t g At the same time, td usually corresponds to temperatures at


which the viscosity lies in the range of 10 11_10 12 P. Therefore the difference td-tg for two-
phase glasses with a highly viscous framework should in general be considerably greater than
for one-phase glasses; this is found to be the case. For example, for glass of the composition
8Na20. 32B 20 3 60Si0 2 td-tg = 160 Thus, a large value of the difference td-t g , namely 80-
0

100 or more, can probably be regarded as fairly reliable evidence of two-phase character of the
0

glass and the presence of a highly viscous framework.


Indenbom [9] and Ananich [10] demonstrated, by study of anomalous birefringence,
the two-phase character (with a highly viscous framework) of the commercial borosilicate
glasses for electrical purposes, Nonex (S 39-1) and 3S-9 (S 38-1). We attempted to establish
the presence of two phases in certain electrical glasses by determination of the differences t 13 -
tg and td-tg' Large differences were found between td of normally annealed glasses and t11 of
the same glasses. These differences were attributed to increases of viscosity in the course of
determination at relatively high temperatures. It was therefore desirable to record dilatomet-
ric curves also after fairly lengthy heat treatment of these glasses. The results are presented
in Table 1 and Fig. 3. Two or three specimens of each glass composition and after each kind of
90 M. V. 8TREL'T8INA, O. V. MAZURIN, AND A. 8. TOTE8H

~l
T log 1]

4.0

3.0 13

2.0 12

1.0 11

o fOO zoo 300 400 SOD 000


Fig. 3. Thermal-expansion curves and temperature dependence of viscosity of glass
No. 46 (847-1). 1) Annealed glass; 2) glass held at 600 for 11 h; 3) viscosity. 0

treatment were taken for the tests. The ranges of tg and td of the glasses are indicated in Table
1. First, it follows from the results that t d, and therefore the viscosities, of normally annealed
glasses and glasses subjected to additional heat treatment differ substantially. However, tg re-
mains almost unchanged. It follows that the properties and therefore the compositions of the
more fusible phases in the glasses remain virtually unchanged during additional heat treatment.
At the same time, a framework consisting of a high-melting phase is formed (38-11, No. 46) or
reinforced (38-9) in all the glasses as the result of heat treatment. The difference td-t g for
the normally annealed glass 38-9 is very considerable. This suggests that it contains a high-
melting framework (in good agreement with the data of Ananich [10]). The other glasses
studied in the normally annealed state apparently do not have a highly viscous framework, al-
though the fact that tg is not changed by further heat treatment indicates that phase separation
in them has already started.

TABLE 1
Glasses after heat to at log 1] =
Annealed glasses treatment at 600
GlassNo. 11 and 13
for 11 h
tg
I td tg
I td t13
I t11

3S-11 (5 40-1) 505-515 540-560 510-520 600-625 600 640


3S-9(538-1) 460-465 545-565 455-460 570-590 553 613
No. 46 (5 47-1) 520-530 560-570 525-530 625-640 600 652
M. V. STREL'TSINA, O. V. MAZURIN, AND A. S. TOTESH 91

In our opinion, the data presented here may have considerable practical significance in ad-
dition to their theoretical interest. The prolonged heat treatment used in our investigation is,
of course, very different from practical conditions. Our aim was merely to choose conditions
under which structure formation in the glasses studied approaches completion. However, it fol-
lows from the work of Klyuev [11] that if a glass has a tendency to phase separation and struc-
ture formation its viscosity may increase considerably during much shorter heat treatments,
especially at temperatures higher than those used by us.
Therefore it must be taken into account that at least in some cases heat treatment (glass-
blowing, welding, etc.) of electrical borosilicate glasses may alter both their annealing tempera-
ture and the character of the CTE curve in the range between room temperature and the solidifi-
cation temperature of the glass in the weld.
Thus, methods based on simultaneous study of tg and t 13 , and of tg and td, can be used both
for demonstrating the occurrence of phase separation in glasses and for revealing some of their
structural characteristics.
Literature Cited
1. O. V. Mazurin, this volume, p. 36.
2. M. V. Strel'tsina, V. P. Klyuev, T. P. Shvaiko-Shvaikovskaya, and O. V. Mazurin, Neorgan.
Mater., 4:1129 (1968).
3. O. V. Mazurin, in: Methods for Measurement of Thermal Expansion of Glasses and of
Metals Sealed to Them, Izd. Nauka, Leningr. Otd., Leningrad (1967), p. 151.
4. M. S. Gomel'skii, V. N. Polukhin, E. A. Khabarova, and V. A. Babkina, Optiko-Mekhan.
Prom., No.5, 41 (1967).
5. S. V. Nemilov, Neorgan. Mater., 2:394 (1966).
6. W. E. S. Turner and E. Winks, J. Soc. Glass Technol., 14:84 (1930).
7. E. Seddon, W. E. S. Turner, and E. Winks, J. Soc. Glass Technol., 18:5 (1934).
8. C. A. Faick, J. C. Young, D. Hubbard, and A. N. Finn, J. Res. U. S. Natl. Bur. Standards,
14:133 (1935).
9. V. L. Indenbom, Dokl. Akad. Nauk SSSR, 89:509 (1953).
10. N. I Ananich, Study of Anomalous (Structural) Birefringence in Inorganic Glasses, Author's
abstract of candidate's dissertation, Moscow (1962).
11. V. P. Klyuev, G. P. Roskova, and V. 1. Aver'yanov, this volume, p. 71. ,
DISCUSSION
A nan i c h, commenting on the paper by Strel'tsina, Mazurin, and Totesh, said that
their method for determining the presence of two phases in glasses from the difference
between td and tg is very interesting and simple. The glasses discussed in their pa-
per have been studied in detail by other methods, including electron microscopy, and their two-
phase character has been proved. Ananich pointed out that in 1953-1955 Indenbom in the
Scientific-Research Institute of Electrical Glass observed that glasses in which Shel-
yubskii detected a heterogeneous structure with the aid of electron microscopy had two soften-
ing temperatures. It was emphasized that study of the temperature dependence of birefringence
may be an effective method in investigations of microphase separation in glasses.*

*The comments of N. I. Ananich were taken into account during preparation of the paper by
M. V. Strel'tsina, O. V. Mazurin, and A. S. Totesh for press.
POROUS GLASSES AS A SOURCE OF INFORMATION ON
STRUCTURAL CHANGES IN SODIUM BOROSILICATE
GLASSES UNDERGOING PHASE SEPARATION

S. P. Zhdanov and E. V. Koromal'di

Interest in study of chemically heterogeneous sodium borosilicate glass has increased con-
siderably in recent years owing to the use of new experimental techniques [1-6]. Information on
the structure of these glasses and on its changes with variations of the glass composition and
conditions of heat treatment was obtained earlier from the results of studies of the structure of
porous glasses by adsorption methods [7-11]. A serious defect of this information was absence
of proper correlation with the relationships of liquid-liquid phase separation in these glasses.
Development of concepts regarding phase separation as the origin of regions of chemical
heterogeneity in silicate glasses [12-17] and investigations of metastable phase separation in
sodium borosilicate glasses in relation to the phase diagram of this system [6, 14] provide a
new and general approach to interpretation of the peculiar variations of the properties both of
the original glasses and of the products formed by their leaching with variations of their com-
position and conditions of heat treatment.
One of the most interesting peculiarities of sodium borosilicate glasses is the anomalous
variation of their density with temperature [18]. This effect was not explained by Molchanova and
Serebryakova [18]. The analogous behavior of Pyrex glass was attributed by Tool [19] to non-
simultaneous solidification of two phases in this glass.
As was shown earlier [10, 20], changes in the structure of porous glasses are intimately
connected with these anomalous density variations of sodium borosilicate glasses. The sensitiv-
ity of the structure of porous glasses to the temperature at which the original sodium borosili-
cate glass 70-23* was held is illustrated by the isotherm of ethyl alcohol adsorption shown in
Fig. 1. The observed Similarity in the variations of the density of the original glass and the
structural characteristics of the porous glasses with the temperature of heat treatment, like
these variations themselves, which were not explained satisfactorily earlier[10. 20], can now,
in our opinion, be explained from a unified standpoint in relation to phase equilibria in the region
of metastable phase separation.
In accordance with the hypothesiS that regions of chemical heterogeneity in sodium boro-
silicate glasses are the consequence of liquid -liquid phase separation, when these glasses are
held in the spinodal area of the metastable immiscibility region two processes should occur: a

*In our code designations of the glasses the first number corresponds to the Si02 content and the
second to the B 20a content (mole%)j the rest corresponds to the Na20 content.

92
S. P. ZHDANOV AND E. V. KOROMAL'DI 93

500-550
o.ZZ 490 0
[70-Z3}
480 0

0.18 Original glass


500 0
550 0
..,~ 0.1" 700 0
B
u

'" 0.10

0.06

o 0.2 0.4 0.0 0.8 1.0 pips


Fig. 1. Influence of the conditions of heat treat-
ment of glass 70-23 in the temperature region of
metastable phase separation on the adsorption
properties of porous glasses made from it. a)
Adsorption; pips) relative pressure of C2R5 OR
vapor.

more rapid process of establishment of compositions (equilibrium for each temperature) of


coexisting borate and silica phases, and a much slower process - growth of the particles of the
borate phase, accompanied by increase of the opalescence of the glass. The increase of density
with the temperature of heat treatment in the opalescence region must be attributed to more
rapid first process, because the increase of opalescence during heat treatment at constant tem-
perature is not accompanied by changes in the density of the glass [18]. Equally, as has been
shown by Zhdanov [7], the structure of porous glass obtained by treatment of the original sodium
borosilicate glass with HCI is also independent of the degree of opalescence if the glass was
held at the same temperature for different times. In such cases the pore volumes, pore sizes,
and the specific surfaces of the porous glasses remained unchanged. Therefore the variations of
the structure of porous glasses with the temperature of heat treatment of the original glass
(Fig. 1) must also be attributed to changes occurring in the phase compositions during heat treat-
ment and not to differences in the degree of opalescence.
Heat treatment of glass within the immiscibility region at higher temperatures should lead
to equalization of the phase compositions and to transfer of a part of the :8:103 and N~O from the
borate into the silica phase, and of some Si02 from the silica into the borate phase; the reverse
transfers should occur on decrease of temperature. Therefore quenching of the glass at a higher
temperature must lead to formation of porous glasses with smaller pore volumes and pore sizes.
These are precisely the structural changes observed in porous glasses on increase of the tem-
perature of heat treatment from 600 to 700. Correspondingly, treatment of the original glass at
lower temperatures (480-550) leads to formation of porous glasses with larger pore volumes
and pore sizes than those obtained from the original glass not subjected to heat treatment.* These

*The fact that the adsorption isotherms of glasses treated at 500, 530, and 550 coincide may be
0

the result of insufficiently long holding times for establishment of structural equilibrium at
these temperatures.
94 S. P. ZHDANOV AND E. V. KOROMAL'DI

conclusions regarding changes in the structure of porous glasses follow from analysis of the
adsorption isotherms in Fig. 1.
The anomalous increase of the density of sodium borosilicate glasses with temperature can
also be explained in terms of changes of phase composition. The connection between this effect
and the two-phase character of the glasses is obvious, since only opalescent glasses exhibit anom-
alous variations of density [18]. The causes of this effect may be understood in the light of Zhda-
nov's concepts [21, 22] of the variable properties of negatively charged boron-oxygen tetrahedra
in glasses of different compositions, and of the nonequivalence of their contribution to the proper-
ties of glasses owing to differences in the electrostatic repulsion between them. In accordance
with these views, the boron-oxygen tetrahedra in the borate regions, with a deficience of [B0 3/ 2]
groups separating them, will increase the density of the glass to a smaller extent than [B0 4/ 2]
tetrahedra in the silica phase, where they form a common silicon- boron-oxygen network and
because of their complete separation repulsion between them will be weakened to the limit.
Therefore partial transfer of ~03 and N~O from the borate regions into the silica framework
on increase of temperature must lead to increase of the oxygen packing density in the glass, and
of the density of the glass itself. This should also be favored by simultaneous transfer of a part
of the Si02 into the borate regions, as this is accompanied by formation of numerous linkages of
I I
the -B--O-Si- type, which should also lead to greater separation of charged boron-oxygen
I I
tetrahedra and to weakened repulsion between them.
Studies of the structure of porous glasses by adsorption methods [10, 22] not only demon-
strated the presence of considerable amounts of Si0 2 in the borate regions of sodium borosilicate
glasses but also made it possible to estimate the degree of dispersion of the Si0 2 particles and
the silica content in the borate phase before such data were obtained by other methods [1, 14, 17].
Adsorption data yielded essentially new information on structural changes in low-alkali
sodium borosilicate glasses during heat treatment and with variations of the Na20 content [23,
24]. It was concluded from the results of adsorption investigations that separate borate regions
are formed in these glasses, sometimes completely isolated in the silica matrix and in other
cases linked by very fine ~03 filaments and therefore accessible to leaching by acids [24].
Closed borate regions are formed only in low-alkali sodium borosilicate glasses. This has been
confirmed by recent investigations of electrical properties of glasses [25].
Adsorption studies of the structure of porous glasses obtained by HCI treatment of ~03-
Si02 two-component glasses containing 60 mole% Si0 2 and over invariably demonstrate the pres-
ence of systems of intercommunicating fine pores (channels) the diameters of which are close to
10 A. The total channel volumes are limited and increase with the ~03 content in the original
glass; they are independent of the leaching conditions [9]. It is significant that a considerable
proportion of ~03 can be removed from such glasses by treatment with ethyl alcohol. These
results show that ~03 - Si02 glasses contain a separate structure formed by very fine ~03 fila-
ments, i.e., they are evidence of definite localization of linkages of the B-O- Band Si -O-Si
types in the glasses. The structure of porous glasses obtained from ~03-Si02 glasses is inde-
pendent of the conditions of heat treatment at 550. This effect was inexplicable in presence of
signs of heterogeneity in ~03-Si02 glasses. These results can now be reconciled with the phase
diagram of the ~03-Si02 system where, in accordance with recent data reported by Charles
and Wagstaff [26], the region of metastable phase separation lies below 520. Evidently, the
influence of heat treatment on the structure of porous glasses could be manifested only at tem-
peratures below 520. However, it is still not clear whether the differentiation of these glasses
with respect to type of linkage is the result of imperfect quenching or of deviations from the
ideal miscibility which they have at temperatures above the region of metastable phase separa-
tion.
s. P. ZHDANOV AND E. V. KOROMAL'DI 95

Investigations of structural changes produced by small amount of Al 20 3 in sodium boro-


silicate glasses are of special interest. It has been shown that Al20 3 diminishes the tendency
of these glasses to phase separation (thiS is manifested externally by suppression or weakening
of their opalescence [27, 28]), and has a considerable influence on their chemical stability.
Whereas addition of small amounts of Al20 3 to alkali silicate and sodium borosilicate glasses
with fairly high relative contents of N~O leads to substantial increase of their chemical stability
[29-32], the chemical stability of low-alkali sodium borosilicate glasses is greatly diminished by
addition of AI 20 s .
In our investigations, carried out on a series of low-alkali sodium borosilicate glasses
containing 2,3, and 4 mole% Na20 and 60-70 mole% Si0 2, Al 20 3 was introduced into the glasses
instead of B,Ps in amounts equivalent to the Na20 contents.
Adsorption investigations showed that all the porous glasses obtained from glasses with
added Al20 s had similar structure, which was independent of the composition of the glass and of
heat treatment at 550. Porous glasses obtained from such glasses have uniformly fine-pored
structure, with predominant pore diameters of about 15-20 A.
The curves in Fig. 2 present the most interesting example of the influence of added Al 20 3
in low-alkali glasses on the properties and structure of the products of their leaching. After
heat treatment at 550 for 10 h the alumina-free glass of the composition 70-27 (b) cannot be
leached appreciably with formation of a porous glass, while the glass of the composition 65-32
(a) is only partially leached after heat treatment under the same conditions, yielding a porous
glass with a limited pore volume (dashed isotherms in Fig. 2). The same glasses quenched at
high temperatures are leached out almost completely (Table 1)"
After introduction of 3% A120 S instead of B:l03 into these glasses their stability drops
sharply and becomes virtually independent of the thermal history of the glass (Table 1); the ad-

a b
0.24- 0.2'1
65 SiDz Z9BA . JNazOJAla03
70Si02 N8A3Naz O3Al z 03
O,f6 0.16
~ ~
'"a '"a
u u

'" 0.08 MSiOz"J28z 3 3Na zO


, _______________ 7
'"
I
,-,:;:::;:~
I

--- ---,.,..
/
70SOz .278 23 3NazO
------------------
0.2 0.4 0.6 0.8

01
pips
x2 eJ
0.2 0.4 0.6 0.8 f.O piPs

Fig. 2. Changes in the adsorption properties and structure of porous


glasses as the result of addition of small amount of A1 20 S to low-al-
kali sodium borosilicate glasses. The dashed lines are isotherms of
water adsorption on the products formed by leaching from low-alkali
glasses without added alumina; the continuous lines are isotherms of
water adsorption (1) and desorption (2) and of n-butanol adsorption (3)
on products formed by leaching from glasses containing A1 20 3
96 S. P. ZHDANOV AND E. V. KOROMAL'DI

TABLE 1. Changes of Chemical Stability of Low-Alkali Sodium


Borosilicate Glasses as the Result of Heat Treatment and of
Addition of Small Amounts of Al 20 3
Amounts of components dissolved in 5 days of treatment
with .'3 N HCI, % of their contents in the glass
Com ponents dissolved 70-27 70-(27-3) 65-32 65-(32-3)

I q. I ann. q. I ann. q. I ann. q.


I
ann.

Na20 57 12 72 90
I 94 42 93 100
82 2 87 96 94 33 100 98
BrO a
A 20 3 - - 73 90 I- - 98 100

Note. In the code designations of glasses containing Al 20 3 the BP3 contents are given
in parentheses. In both cases the A lP3 is 3 mole%, equivalent to the Nap content.
Quenched glasses are designated q . and glasses heat -treated at 550 0 are deSignated
"ann."

sorption isotherms (Fig. 2) show that the porous glasses obtained from them, although fine-pored,
do not exhibit an ultraporosity effect in adsor.ption of butanol.
The absence of opalescence in the original low-alkali glasses containing added Al20 3 and
of large cavities in the porous glasses made from them indicates that Al20 3 homogenizes their
structure, suppressing the tendency of phase separation.
The negatively charged boron-oxygen tetrahedra formed as the result of interaction of
Na20 with B:!03 are located predominantly in the borate phase and apparently playa most impor-
tant part in all the structural transformations occurring in low-alkali glasses. Therefore the
properties of these glasses are very strongly dependent on their N~O contents [22-24].
In distinction from low-alkali glasses without alumina, the properties and structure of
glasses containing added alumina in amounts equivalent to the alkali oxide content are insensitive
to small variations of the Na20 content and to heat treatment: Al20 3 appears to suppress the in-
fluence of Na20 on the structure of these glasses. This can be explained as follows: in pres-
ence of Al 20 3the Na20 present in the glass combines with Al 20 3 rather than with B:!03 in the first
instance, and aluminum in the aluminum-oxygen tetrahedra, together with Na+ ions balancing
their negative charges, enters the common silicon-aluminum-oxygen network of the silica-rich
phase, while free B:!03 forms a filamentary structure similar to that found in B:!03-Si02 glasses.
This is the structure revealed by studies of adsorption on porous glasses obtained from low-
alkali sodium borosilicate glasses containing alumina. Like the structure formed by B:!03 in al-
kali-free two-component borosilicate glasses, it is insensitive to heat treatment at 550 0

In our opinion, the strong influence of small amount of Al20 3 on the chemical stability of
low-alkali sodium borosilicate glasses can be satisfactorily explained by absence of borate in-
clusions enclosed in the silica framework in glasses containing A120 3, and by formation of a
filamentary boric anhydride structure more accessible to acid, without ultrapores hindering dif-
fusion during leaching.
Literature Cited
1. W. Skatulla, W. Vogel, and H. Wessel, Silikattech., 9:51 (1958).
2. K. KUhne and W. Skatulla, Silikattech., 10:105 (1959).
3. E. A. Porai-Koshits, in: Structure of Glass, Izd. Akad. Nauk SSSR, Moscow- Leningrad
(1955), p. 145 [English translation: The Structure of Glass, Vol. 1, Consultants Bureau,
New York (1958), p. 112].
S. P. ZHDANOV AND E. V. KOROMAL'DI 97

4. E. A. Porai-Koshits, S. P. Zhdanov, and D. I. Levin, lzv. Akad. Nauk SSSR, Otd. Khim.
Nauk, No.1, p. 31, No.2, p. 197, No.3, p. 395 (1955).
5. N. S. Andreev and E. A. Porai-Koshits, Dokl. Akad. Nauk SSSR, 118:735 (1958).
6. Tran Thach-Lan, Verres Refractaires, 19:416 (1965).
7. S. P. Zhdanov, Investigation of the Structure of Porous Glasses with the Aid of Isotherms
of Vapor Sorption, Author's abstract of candidate's dissertation, Leningrad (1949).
8. S. P. Zhdanov, Dokl. Akad. Nauk SSSR, 82:281 (1952).
9. S. P. Zhdanov, Dokl. Akad. Nauk SSSR, 92:597 (1953).
10. S. P. Zhdanov, in: Structure of Glass, Izd. Akad. Nauk SSSR, Moscow- Leningrad (1955),
p. 162 [English translation: The Structure of Glass, Vol. 1, Consultants Bureau, New
York (1958), p. 125].
11. D. P. Dobychin, in: Structure of Glass, lzd. Akad. Nauk SSSR, Moscow- Leningrad (1955),
p. 176 [English translation: The Structure of Glass, Vol. 1, Consultants Bureau, New York
(1958), p. 135].
12. R. Roy, J. Am. Ceram. Soc., 43:670 (1960).
13. W. Hinz and P. O. Kunth, Glastech. Ber., 9:431 (1961).
14. F. Ya. Galakhov, lzv. Akad. Nauk SSSR, Otd. Khim. Nauk, No.5, p. 743 (1962).
15. J. W. Cahn, J. Chem. Phys., 42:93 (1965).
16. J. W. Cahn and R. J. Charles, Phys. Chem. Glasses, 6:181 (1965).
17. Tran Thach-Lan, Verres Refractaires, 20:8 (1966).
18. O. S. Molchanova and M. V. Serebryakova, Tr. Gos. Optich. lnst., 23(141):3 (1953).
19. A. Q. Tool, J. Am. Ceram. Soc., 31:1177 (1948).
20. S. P. Zhdanov, Tr. Gos. Optich. Inst., 24(145):86 (1956).
21. S. P. Zhdanov, in: The Glassy State, Proceedings of the Third All- Union Conference, Izd.
Akad. Nauk SSSR, Moscow- Leningrad (1960), p. 502 [English translation: The Structure
of Glass, Vol..2, Consultants Bureau, New York (1960), p. 454].
22. S. P. Zhdanov, Structure of Porous Glasses and Structural Transformations in Sodium
Borosilicate Glasses, Doctoral dissertation, Inst. Khim. Silikatov, Leningrad (1959).
23. S. P. Zhdanov and E. V. Koromal'di, Izv. Akad. Nauk SSSR, Otd. Khim. Nauk, No.4, p.
626, No.5, p. 811 (1959).
24. S. P. Zhdanov, Izv. Akad. Nauk SSSR, Otd. Khim. Nauk, No.6, p. 1011 (1959).
25. T. P. Dgebuadze and O. V. Mazurin, Neorgan. Materialy, 2:1328 (1966); 3:1236 (1967).
26. R. J. Charles and F. E. Wagstaff, J. Am. Ceram. Soc., 51:16 (1968).
27. J. Tamura, Atti 3 Congr. Intern. del Vetro, Venice (1959), p. 426.
28. S. P. Zhdanov (Zhdanow) and E. V. Koromal'di, 8 Congr. Intern. du Verre, Bruxelles, Vol.
2, No. 302 (1965).
29. S. K. Dubrovo, Izv. Akad. Nauk SSSR, Otd. Khim. Nauk, No.2, p. 244 (1954); Zh. Prikl.
Khim., 32: 1199 (1959).
30. L. S. Yastrebova, Neorgan. Mater., 2: 749 (1966).
31. J. Yamamoto, J. Ceram. Assoc. Japan, 62:125 (1954); Chem. Abstr., 48:8503 (1954).
32. J. Tamura, Rep. Osaka Ind. Res. Inst., 299:37 (1951); Chem. Abstr., 48:10311 (1954).

DISCUSSION
In reply to Varshal, Tsekhomskii, and Porai-Koshits regarding the influence of
aluminum on the properties and structure of sodium borosilicate glasses, Z h dan 0 v
said: "When sodium oxide is introduced into two-component borosilicate glass, it first
combines with boric anhydride. Borate regions are formed, and this leads to the observed
peculiarities of the isotherms. On introduction of aluminum oxide sodium is bonded to aluminum
and the borate component loses sodium; since aluminum enters the siliceous framework, these
glasses behave like two-component borosilicate glasses. There is no certainty that the composi-
tion of the siliceous phase remains unchanged during prolonged leaching of the glass."
98 S. P. ZHDANOV AND E. V. KOROMAL'DI

Po r a i - K 0 shit s noted in connection with the paper by Zhdanov and Koromal'di


that separation of silica in the sodium borate component of sodium borosilicate glass-
es after heat treatment above 550 0 is due to secondary phase separation.
Replying to Porai-Koshits, Zhdanov emphasized that the presence of silica in the
borate regions is determined by the phase diagram and is not the result of secondary
separation.
In reply to a question by Dgebuadze regarding changes of phase composition during
isothermal treatment, Zhdanov said that in glasses with a fairly high alkali content
the compositions of the borate and siliceous phases change; silica passes partially from the
siliceous into the borate phase, while ~03 and N3.:!O are partially transferred from the borate
into the siliceous phase. This is confirmed by the curves given in the paper. Finely divided
silica is formed in the borate regions as the result of leaching, and can then be easily removed
almost without affecting the main framework. These results are in good agreement with the
phase diagram and with the position of the immiscibility region.
Replying to questions by Filipovich and Mazurin, Zhdanov emphasized that boron is
extracted completely even from quenched two-component borosilicate glasses, and there-
fore heterogeneity regions appear in these glasses even without additional heat treatment.
At the same time, the glass becomes more heterogeneous if the cooling process is retarded. It
is difficult to say at present what part is played in the formation of this heterogeneous structure
by the characteristics of glass structure above the immiscibility region and by the conditions of
cooling in passing through the immiscibility region.
In reply to Aver'yanov, Zhd a nov said that the process kinetics must be taken into
account in order to establish from which regions of sodium borosilicate glasses sodium
and aluminum are extracted during leaching. Initially the products of decomposition of the borate
regions are dissolved out, and this is followed by removal of components from the framework,
Po r a i - K 0 s hit s pointed out in connection with Zhdanov's paper that the hetero-
geneous structure of glasses in the binary borosilicate system was demonstrated by x-ray
methods as long ago as 1943 (from the additivity of the intensity and radial-distribution curves),
and this did not require heat treatment of the glass at 520 for 1000 h, which was needed by
0

Charles for evaluation of the phase diagram.


INVESTIGATION OF PHASE SEPARATION IN GLASSES OF
THE SYSTEM R20-B20 3-Si02 BY STUDY OF
THEIR ELECTRIC PROPERTIES

T. P. Dgebuadze

Glasses of the system ~o- ~03- Si02 exhibit metastable phase separation, which dimin-
ishes with increaSing radius of the alkali-metal ion in the glass [1-7]. Because of the different
rates of metastable phase separation processes, it is possible to obtain various intermediate
structures [8, 9].
The course of metastable phase separation can be followed by analysis of the conductivity
values and of the character of dielectric losses [4, 10]. The appearance of relaxation maxima
on the curves for total dielectric losses indicates that relaxation processes occur at the phase
boundaries between the enclosed conducting inclusions and the poorly conducting medium [11-
14]. Subsequent heat treatment of a glass with a metastable phase-separation structure results
in formation of a stable crystalline structure.
Methods of measuring the electrical properties and of analysis of results obtained for
glasses studied in the present investigation are described elsewhere [4, 10].
Table 1 gives the experimental results obtained in studies of electrical properties (p, the
specific resistance, and E C' activation energy of conduction) of glasses of the R20- ~03-Si02
system, revealing some characteristic features of their structure.
Glasses cast in molds from melts and annealed at 300-400 are described as "untreated,"
in distinction from glasses subjected to subsequent heat treatment (Table 1).
A phase-separation structure close to equilibrium was formed in all "untreated" lithium
glasses. Subsequent heat treatment either did not alter Significantly the type of microhetero-
geneities (glass 1) or facilitated crystallization (glass 2).
Prolonged heat treatment is needed to produce metastable phase equilibrium in sodium and
all the more so in potassium glasses. A fine heterogeneous structure can be detected in "un-
treated" sodium glasses (Fig. la). Distinct phase separation occurs during heat treatment
(Fig. Ib).
Glass 3 before heat treatment exhibits relaxation losses at the phase boundaries of the
inclusions; this is indicated by the appearance of a relaxation maximum on the total-loss curve
(Table 1). Glasses with 2% N~O containing 19.6 and 24.5% ~03* gave a similar result. The
presence of isolated regions of the conducting component in "untreated" glass 3 is confirmed by
*All the compositions are given in mole%.

99
100 T.P.DGEBUADZE

TABLE 1

No. GI'~"mpo'''I~e" ne"mem


log
p300
Presence
Ec ' eV of max-
im um*
Character of mi cro -
heterogeneities
Notes

Untreated H. 4 1.4::3 Yes Enclosed inclusions Opalescence


1 4Li 2 0 . 36B 2 0 3 60Si0 2
{ of conducting phase
700, 2 h 11 .6 1.46
. The sa me .
Untreated fi.2 0.85 - Continuous conducting Milky
2 !OLizO 36B 2 0 3 54Si0 2 { 700, 2 h 8.2 1.1 :~ -
phase
The sam e Milky (crystal-
lization
Untreated HI.8 1. :3,) Yes Fine heterogeneous Transparent
3 2Na 2 0 29.4D 2 0 a 68.6Si0 2
{ 550, 10 h [ 1.1 1.17 Of
structure t
Enclosed inclusions o pa 1escen ce
I of conducting phase
Untreated 8.8 1.tJ7 - Fine h eterogeneou s Transparent
4 4Na20 26D 20 3 70Si0 2
{ 600, 20 h 10. 1 1. 25 Yes
structure
Enclosed incl usions Milky
of conducting phase
Untreated 7.0 0.01 - Fine heterogeneo us Transparent
5 lONa 20 22.5B 20 3 67.5Si0 2
{ 600 0
, 20 h (i. 8 (Ull) -
structure
Continuous conducting Opalescen ce
I ph ase

6 4K 2 0 26B 2 0 3 70Si0 2 { Untreated


500, 45 h
I I. 7
Il. I
!.4 1
1.10
I -
-
-
Continuous conducting
Transparent
Opa lescence
phase

I
"Refers to relaxation maximum on the curve for total dielectri c losses.
t Observed under the e lectron microscope.

comparison with Zhdanov's adsorption data [15, 16]. According to these data, glass of the
composition 2 % Na20. 28 % B:P3 ' 70 %Si02, which is close to our glass 3, is not leached apprecia-
bly and has high chemical stability both in the quenched and in the annealed (550 10 h) state. 0
,

The explanation given is that the leaching agent (hydrochloric acid) does not have access to the
soluble sodium borate component because the latter forms enclosed inclusions.

Fig. 1. Electron micrographs of glass of the composition 4% Na20 26%B 20 3 . 70 %Si02


a) "Untreated" glass; b) heat treatment at 600 for 75 h. 0
T. P.DGEBUADZE 101

Therefore the observed change of resistance of glass 3 (Table 1) must be attributed to


continuing differentiation of the components in the phases into which the glass separates. During
establishment of phase equilibrium alkali-metal ions pass from the silica-rich phase into the
phase rich in boric anhydride and alkali; this leads to decrease of the alkali-metal ion content
in the matrix and to increase of the glass resistance.
Analogous variations of phase composition in the course of phase separation must evident-
ly occur in other glasses containing sodium and potassium, in which the conducting phase be-
comes continuous owing to increase of its volume (Table 1). However, it is noteworthy that the
activation energy of conduction of glasses 5 and 6, corresponding to the activation energy of con-
duction of the phase having higher conductivity, remains almost unchanged after heat treatment.
The constancy of the activation energy of conduction in glasses in which the phase of high-
er conductivity was presumed continuous led to the conclusion in a number of investigations that
the phase compositions remained unchanged. This conclusion was reached in comparisons both
of different glasses and of glasses of the same composition subjected to different heat treatments
[2,10, 13,17J.
Evidently, conclusions based on calculations of the activation energy of conduction need not
al ways be correct [2, 14]. For example, it is not clear how the components of the effective ac-
tivation energy (dissociation energy and activation energy [18]) should vary in the course of phase
separation. With different temperatures of heat treatment, the use of the values found for the
activation energy of conduction may be difficult owing to possible secondary phase separation [19].
The influence of phase volumes is reflected in the electrical properties not only when sep-
arated structures close to equilibrium are formed but also at the initial stages of development
of the phase-separation processes. "Untreated" glass containing 2% Na20 gives a maximum in-
dicating the presence of enclosed inclusions, whereas "untreated" glass with 4% Na 20, which has
the same fine heterogeneous structure as the glass with 2% Na20, does not give a maximum (Ta-
ble 1). In all probability, this is due to the fact that because of the greater volume ofthe conduct-
ing phase formed in comparison with the glass with 2% Na20 at the early stage of phase separa-
tion interconnected regions of the conducting component are formed in the glass with 4% Na20.
The conducting phase forms enclosed inclusions only at later stages of phase separation [10].
Therefore formation of the heterogeneous structure in glasses of the system R20- ~03-
Si02 during establishment of metastable phase equilibrium is accompanied, in addition to changes
of the phase compositions (as was shown above for glass 3), by changes of the particle shape
from elongated "vermiform" to spherical [10] and increase of particle size, with distinct phase
boundaries.
Literature Cited
1. F. Ya. Galakhov, Steklo i Keram., No.7, p. 5 (1962).
2. R. J. Charles, Jo Am. Ceramo Soc., 47:559 (1964).
3. J. W. Cahn and E. J. Charles, Phys. Chem. Glasses, 6:181 (1965).
4. T. P. Dgebuadze, Investigation of Phase Separation in Alkali Borosilicate Glasses by Study
of Their Electrical Properties, Author's abstract of candidate's dissertation, Leningrad
(1967) 0
50 T. P. Dgebuadze and 0. V. Mazurin, Neorgan. Mater., 4:1526 (1968).
6. F. Ya. Galakhov and 0. S. Alekseeva, this volume, p. 80.
70 M. Vo Strel'tsina, 0. V. Mazurin, and A. S. Totesh, this volume, p. 87.
8. J. W. Cahn, J. Chem. Phys., 42:93 (1965).
9. V. N. Filipovich, Neorgan. Mater., 3:993, 1192 (1967).
10. T. P. Dgebuadze and 0. V. Mazurin, Neorgan. Mater., 3:1236 (1967).
11. R. J. Charles, J. Am. Ceram. Soc., 46:235 (1963).
102 T. P. DGEBUADZE

12. V. K. Leko and M. L. Dorokhova, in: Electrical Properties and Structure of Glass, Izd.
Khimiya, Moscow- Leningrad (1964), p. 84.
13. O. V. Mazurin and V. B. Brailovskii, in: The Glassy State, Proceedings of the Fourth
All- Union Conference, Izd. Nauka, Moscow- Leningrad (1965), p. 277 [English translation:
The Structure of Glass, Vol. 7, Consultants Bureau, New York (1966), p. 93].
14. T. P. Dgebuadze and O. V. Mazurin, Neorgan. Mater., 2:1328 (1966).
15. S. P. Zhdanovand E. V. Koromal'di, Izv. Akad. Nauk SSSR, Otd. Khim. Nauk, No.4, p. 626
(1959)
16. S. P. Zhdanov, Izv. Akad. Nauk SSSR, Otd. Khim. Nauk SSSR, Otd. Khim. Nauk, No.6,
p. 1011 (1959).
17. V. K. Leko, Neorgan. Mater., 2:1224 (1967).
18. O. V. Mazurin, Electrical Properties of Glass, Goskhimizdat, Leningrad (1962).
19. E. A. Porai-Koshits and V. 1. Aver'yanov, this volume, p. 28.
INFLUENCE OF REPLACEMENT OF SILICA BY VARIOUS
GLASS-FORMING OXIDES ON THE TENDENCY TO
PHASE SEPARATION IN THE SYSTEM R20-B20 3-Si02

D. F. Ushakov and Yu. S. Krupkin

The influence of equimolecular replacement of silica by ZnO, PbO, CaO, A120 3 , and Ge02
on the tendency of alkali borosilicate glasses to phase separation was studied. The original glass
contained 4 mole% Li 20, 36 mole% B,P3' and 60 mole% Si02 The influence of replacement of 5
mole% Si02 by ZnO, CaO, PbO, and A1 20 3 , and of equimolecular replacement of up to 20% Si0 2 by
Ge02, on the tendency of the original glass to phase separation was investigated. The glasses
were made at 1450-1500 in vitreous silica crucibles fitted with lids and molded into disks 30 mm
in diamter and 4 mm thick, which permitted the most rapid cooling of the specimens to the an-
nealing temperature, In addition, glasses in which Si0 2 was replaced completely by ZhO, PbO, or
Ge02 were investigated; these glasses were made at 900-950, All the glasses were annealed at
400, The glasses were investigated by the method of induced crystallization, by differential
thermal analysis, and by determination of their electrical characteristics. The potentialities of
this last method are discussed in the literature [1, 2]. The resistance was measured correct to
0.1 log p with aid of the E6-3 instrument, and dielectric losses were determined at 10 3Hz fre-
quency with the MLE-l instrument. The electrodes were attached by brazing with the silver
paste.
It is known from studies of alkali borosilicate glasses by adsorption methods [3, 4] and by
determinations of their electrical characteristics [5] that these glasses separate into a polar
phase consisting predominantly of alkali borate with a certain amount of silica, and a high-silica
nonpolar phase with small contents of alkali oxides and boric anhydride. The composition of the
original glass was chosen so that the polar phase forms enclosed inclusions as the result of
phase separation and the curve for the temperature dependence of total dielectric losses has a
distinct relaxation maximum. The results of measurements of the electrical characteristics of
the original glass and of glasses in which Si02 was replaced completely by ZnO, PbO, or Ge02 are
shown in Fig. 1. Heat treatment of the original glass at 600 for 10 h leads to increase of log p
at 300 by 1.3 and of the activation energy of conduction from 1.3 to 1.6 eV, and results in the
appearance of a well-defined relaxation maximum on the dielectric loss vs temperature curve.
The electrical characteristics of the other glasses in this series are almost unchanged after
heat treatment in the 400-600 range for 10-50 h. Phase separation could not be detected by
means of induced crystallization or DTA either; the explanations that the compositions studied
are close to the eutectic compositions in the systems ZnO- B,P3 [6] and PbO - B,P3 [7]. The
phase diagram of the system Ge02 - BP3 is not known.

103
104 D. F. USHAKOV AND Yu. S. KRUPKIN

12.0
11.0
Q. 10.0
1>0
.3 9.0 4.0
B.O
7.0

3.0 0()

....1>0

2.0

H ~ " m _ u u u n M
liT tO It
Fig. 1. Electrical characteristics of glasses of various composi-
tion (mole%). LBS) 4% Li20, 36% B20a, 60% Si~; LBP, LBZ, LBG)
glasses in which Si~ was replaced by PbO, ZnO, and Ge02' respec-
tively. The dashed line corresponds to LBS glass after heat treat-
ment for 10 h at 600 0

Investigation of glasses in which 5 mole% Si~ was replaced by another oxide showed that
the influence of an oxide on the tendency of the original glass to phase separation corresponds to
the phase diagram for the system formed by this oxide with silica. The results of electrical
measurements are shown in Fig. 2. Glasses containing ZnO and CaO enhibit greater opalescence
and have higher specific resistance than the original glass, and give a more pronounced relaxa-
tion maximum on the dielectric-loss curve. Intensification of phase separation of the original
glass is due in this case to the existence of immiscibility regions in the systems CaO-Si02 [8],
ZnO-Si02 [91, CaO- ~Oa [10], and ZnO- ~03 [6]. The glass containing 5 mole% Al20 a is quite
transparent and its electrical characteristics are altered insignificantly by heat treatment.
Weakening of phase separation on replacement of Si02 by A120 a is also consistent with the phase
diagrams of the systems A120 a-Si02 [11] and ~Oa-A120a [12]. The glass containing 5 mole%
PbO is more opalescent than the original, but no inferences can be drawn with regard to the in-
fluence on the tendency to phase separation, as the form of the dielectric-loss curve indicates
that the polar phase is continuous. The phase diagram of the system PbO-Si02 in the high-sil-
ica region is not known.
Replacement of up to 10 mole% Si02 in the original glass by Ge02 leads to decrease of
opalescence, and glasses containing 15 and 20 mole% Ge02 are quite transparent. The results
of determinations of the electrical characteristics of these glasses are given in Fig. 3. It was
shown that replacement of Si~ by Ge02 leads to decrease of opalescence and of specific resis-
tance, and to disappearance of the relaxation maximum on the dielectric-loss curve. Heat treat-
ment of these glasses at 600 0 for 10 h leads to appearance or intensification of opalescence,
increase of specific resistance, and appearance of a relaxation maximum on the dielectric-loss
curve.
D. F. USHAKOV AND Yu. S. KRUPKIN 105

11.0
Q..10.0
go 9.0
- 8.0
7.0
4.0~
.-I

3.0

L----L_--L._.-L-_.L----:':-----.:!::--::-~;;:_---'z.O
H 6 m " m ~ M U n
IjTl0*
Fig. 2. Electrical characteristics of glasses in which
5 mole% Si02 was replaced by PbO, ZnO, CaO, Ge~, or
AI 20 a

It can be concluded from the results of this investigation that in the case of multicompo-
nent glasses the tendency to phase separation is also determined by the phase diagram of the
corresponding oxide system.

Original glass
IZ.0 ,... ...... ,...,....~

--
11.0
Co. 10.0
........ ,....------
be
oS 9.0
5.0
8.0
7.0

L----1._--'-_-L...._...L.-_L.----1._-L_.-L------.!Z.O
H 6 m " ~ M U m n
fiT- 10*
Fig. 3. Influence of replacement of Si~ by Ge02 on
electrical characteristics of alkali borosilicate glasses.
106 D. F. USHAKOV AND Yu. S. KRUPKIN

Literature Cited
1. O. V. Mazurin and V. B. Brailovskii, in: The Glassy State, Proceedings of the Fourth
All- Union Conference, Izd. Nauka, Moscow- Leningrad (1965), p. 277 [English translation:
The Structure of Glass, Vol. 7, Consultants Bureau, New York (1966), p. 93].
2. R. J. Charles, J. Am. Ceram. Soc., 49:55 (1966).
3. D. P. Dobychin, in: The Glassy State, Proceedings of the Third All-Union Conference,
Izd. Akad. Nauk SSSR, Moscow- Leningrad (1960), p. 480 [English translation: The Struc-
ture of Glass, Vol. 2, Consultants Bureau, New York (1960), p. 432].
4. S. P. Zhdanov, in: The Glassy State, Proceedings of the Third All-Union Conference, Izd.
Akad. Nauk SSSR, Moscow- Leningrad (1960), p. 502 [English translation: The Structure
of Glass, Vol. 2, Consultants Bureau, New York (1960), p. 454].
5. T. P. Dgebuadze and O. V. Mazurin, Neorgan. Mater., 3:1236 (1967).
6. N. A. Toropov and P. F. Konovalov, Dokl. Akad. Nauk SSSR, 66:1105 (1962).
7. R. Geller and E. Bunting, J. Res. U. S. Natl. Bur. Standards, 18:585 (1937).
8. F. P. Glasser, J. Am. Ceram. Soc., 45:242 (1962).
9. E. Bunting, J. Am. Ceram. Soc., 13:8 (1930).
10. E. Carlson, J. Res. U. S. Natl. Bur. Standards, 9:825 (1932).
11. N. A. Toropov and F. Ya. Galakhov, Izv. Akad. Nauk SSSR, Otd. Khim. Nauk, No.1, p. 8
(1958) .
12. P. J. M. Gielisse and H. R. Foster, Nature, 195(4936):69 (1962).

DISCUSSION
In reply to a comment by Mazurin in the paper by Ushakov and Krupkin, Us ha kov
said that on the basis of the available data it may be inferred that the influence of ger-
manium dioxide differs in accordance with its content in the glass.
PHASE SEPARATION IN LOW-ALKALI BOROSILICATE
GLASSES CONTAINING RO AND A1 20 3

L. A. Bal'skaya, L. A. Grechanik, and N. M. Vaisfel'd

It is well known that opalescence caused by phase separation is diminished by introduction


of alumina into sodium borosilicate glasses [1-3]. However, in investigations of low-alkali boro-
silicate glasses containing even small amounts of oxides of the RO type it was found that in a
number of cases opalescence of these glasses is intensified by introduction of A120 3 [4, 5]. It
was therefore considered interesting and desirable to investigate phase separation in borosilicate
glasses containing both A120 3 and RO, especially as in most cases both these oxides are actually
present.
The low-alkali region close to industrial glasses in the system Si02- B:P3- A120 3- ZnO-
N~O, previously studied by Grechanik and Bal'skaya [6], was chosen for the investigation. The
glasses were designed in series in accordance with the ZnO contents, introduced to replace Si02
in amounts of 1, 2, 3, 4, 8, 12, 16, and 20% (all compositions are given in mole%). Each series
included several sections with various B20 3 contents (from 0 to 30% in steps of 2 or 4%; for some
sections also 40 and 50%); the A120 3 concentration in each section was varied from 0 to 8% in
steps of 1%. The Na20 content of all the glasses was constant at 5 %.
The glasses were made from reagents of "pure" or cp grade in a gas-fired furnace in vit-
reous silica crucibles, in 5-g lots; up to 60-70 crucibles were put into the furnace simultaneous-
ly (high-borate compositions were synthesized in a laboratory furnace with Silit heaters). The
following temperature schedule was used for the synthesis: rise of temperature to 1400-1480
in 5-6 h, followed by a holding time of 3 h. The glasses were cooled in the turned-off furnace; in
the temperature range of maximum phase separation (500-800) the cooling rate was 3-5/' min.
This cooling procedure is polythermal treatment in which each glass passes through the temper-
ature range of maximum phase separation during 30-40 min, as the decrease of temperature is
continuous, the degree of separation attained does not alter. Thus, the degree of opalescence ob-
served in the glasses is the maximal for each under the given conditions.
Opalescence was estimated in reflected light visually with the naked eye or with the aid of
the MBS-2 microscope at x 50 magnification.
Figure 1 shows that in all the sections studied in two series of the system Si0 2- B:P3-
A1 20 3 - ZnO- N~O (with the exception of high-borate glasses, which were opaque) there is a com-
plex dependence of opalescence on the alumina content: complete opacification is characteristic
of the original glass without A120 3 ; 1-2% A120 3 causes sharp weakening of opalescence; further
increase of the alumina content (3-5%) leads to substantial intensification of opalescence, which
reaches a maxima in presence of about 5% A120 3 Larger amounts of A120 3 (6-8%) decrease
opalescence again, up to total disappearance.

107
108 L. A. BAL'SKAYA, L. A. GRECHANIK, AND N. M. VAISFEL'D

a b

o 5 fO f5 o .; 10 15
AlzOa , %
Fig. 1. Opalescence regions of glasses in the system SiO:!-
B2O:J-AI20a-ZnO-Na20 and ZnO contents (%): a) 4; b) 12. Code
symbols for Figs. 1, 2, and 4: 1) No opalescence; 2) weak opal-
escence; 3) strong opalescence; 4) weak opacification; 5) com-
plete opacification; 6) crystallization.

This relationship is observed in the system with various amounts of ZnO and ~Oa, and
persists even in the boron-free boundary four-component system Si02- A120 3- ZnO- Na20 (Fig. 2).
Thus, two opalescence regions were found in the system Si02- ~Oa- A1 20 3- ZnO- Na20:
the first is narrow, adjacent to the boundary system Si02- ~03- ZnO- Na20 and extending only
to 1-2% Al 20 3 into the system; the second lies approximately between Al20 a concentrations of 2
to 7%, extends along the plane corresponding to the constant ratio AI20 a : Na20 = 1, and emerges
on the boron-free boundary system Si02 - A120 3- ZnO- Na20. The two regions overlap when the
~03 content exceeds 30%.

In one of the sections of the system (containing Si0 2 69-61 %, ~Oa 22%, ZnO 4%, Na 20 5%,
Al20 a 0-8%), in addition to visual estimation of opalescence, the clearing temperature and the
extent of the temperature range of opalescence were determined. Variations of these charac-
teristics with the Al20 a content are shown in Fig. 3; it can be seen that they are similar in char-
acter.
The existence of two regions of opalescence is also observed in the low-alkali ranges of
Si02-~03-AI203-RO-Na20 systems where other oxides, MgO or CaO, represent RO (Fig. 4,
a and b); however, this effect was not observed in glasses containing SrO, BaO, or PbO, where
increase of the Al 20 3 content leads to steady decrease of opalescence (Fig. 4, c-e).
The structure produced in glasses by phase separation was detected by electron-micro-
scopic studies of opalescent and clear specimens after polythermal treatment. The investiga-
tion was carried out by the platinum-carbon replica method with the "Tesla" electron micro-
scope.
L. A. BAL'SKAYA, L. A. GRECHANIK, AND N. M. VAISFEL'D 109

Fig. 2. Opalescence region


of glasses in the system Si02-
Al 20 3- Na20. Symbols as in
Fig. 1.

Fig. 3. Clearing temperatures


and extents of the temperature
range of opalescence of glasses
of the composition (69-x) Si02
22B20 3 . xAl 20 3 . 4ZnO' 5Na20
(x=O-S).

a b c d e

,f 10 5 10 AlgOJ'5 to
%
0 j 10 5 fO

Fig. 4. Opalescence of glasses in Si02-B203-A1203-RO-Na20 sys-


tems. RO content 4%. a) MgO; b) CaO; c) SrO; d) BaO; e) PbO. other
symbols as in Fig. 1.
110 L. A. BAL'SKAYA, L. A. GRECHANIK, AND N. M. VAISFEL'D

This investigation showed that all the glasses studied have a structure resulting from liq-
uid -liquid phase separation (the occurrence of specifically liquid -liquid separation phenomena
in these glasses was also confirmed by x-ray structure analysis of three completely opacified
glasses from different sections; no crystalline phases were detected in them). The investigation
also revealed a sharp difference of the heterogeneous structures in the two opalescence regions.
Opalescent glasses of the first region have a two-phase interconnected labyrinth structure, most
pronounced in glasses without Al20 3 (Fig. 5a); the presence of a considerable amount of ~03
(50%) leads to independent separation within each phase. The first additions of Al 20 3 (1-2%) lead
to breakdown of this structure and conversion into a system of irregular vermiform particles,
isolated from each other in most cases (Fig. 5, b and c). The structure of opalescent glasses of
the second region consists of isolated droplets of a separated phase distributed in the matrix
(Fig. 5, c-f). The amount and size of the drops depend on the Al20 3 content and show correlation
with the degree of opalescence. At 7-8% Al20 3 concentration the drops become so small that the
heterogeneous structure visible under the electron microscope almost disappears (Fig. 5, g, h).
The presence of other bivalent-metal oxides (instead of ZnO) in glasses not containing
Al20 3 does not alter essentially the labyrinth character of the heterogeneous structure (Fig. 6a);
this is consistent with literature data [6]. Glasses with calcium oxide contain isolated particles

Fig. 5. Electron micrographs of glasses in the system SHl.!-B203-AI203-ZnO-Na20.


Contents (%): ZnO, 4; B20 3: I) 14; II) 22; 1m 30; IV) 40; V) 50; A1 20 3 : a) 0; b) 1; c) 2;
d) 3; e) 4; f) 5; g) 6; h) 8.
L. A. BAL'SKAYA, L. A. GRECHANIK, AND N. M. VAISFEL'D 111

Fig. 6. Electron micrographs of glasses in the system SiO:!-B 20 3 - A1203-


RO-Na20. Contents (%): RO, 4; B20 3, 22; A1 20 3: a) 0; b) 1; c) 2; d) 3; e)
4; 1) 5; g) 6. I) MgO; II) CaO; III) SrO; IV) BaO; V) PbO.

of irregular shape (Fig. 6, II, a), formed by breakdown of the continuity of the channels as the
result of polythermal treatment. The presence of Al20 3 (1%) in glasses containing various RO
leads to the same kind of breakdown of the structure as in glasses with ZnO (Fig. 6b). On in-
crease of the Al20 3 concentration (to 2-5%) in glasses containing MgO and CaO, as in glasses
containing zinc oxide a droplet-type structure is formed (Fig. 6, I, II, c-1), while in glasses with
SrO, BaO, and PbO the particle size of the separated structure gradually diminishes until the
heterogeneous structure visible under the electron microscope disappears entirely (Fig. 6, III
and IV).
study of opalescence and heterogeneous structure showed that a most important feature of
the low-alkali range of the system Si02-B20a-A120a-ZnO-Na20 is the existence of two regions
of phase separation.
The relationships found for low-alkali glasses of the system Si02- ~Oa- Na20 are char-
acteristic of glasses in the first region. As in the case of sodium borosilicate glasses [7], on
introduction of 1-2% A120 a in the first phase-separation region considerable weakening of opales-
cence occurs and the clearing temperature is lowered, i.e., phase separation is diminished. In
presence of 1-2% A120 a the minimum degree of phase separation is observed in all sections of
the system Si02- ~03-AI20a- ZnO-N~O.
112 L. A. BAL'SKAYA, L. A. GRECHANIK, AND N. M. VAISFEL'D

The existence of the second region of phase separation is probably attributable to the
specific characteristics of the system Si02- A120 3- ZnO- Na20, as this region is detected even
in the boron-free boundary system. This system has one region of phase separation which ex-
tends, as for glasses of the boron-containing system, along the plane corresponding to the ratio
A1203:Na20 = 1 (Fig. 2). The first region of phase separation was not detected; this can be at-
tributed primarily to the presence of 5% of sodium oxide, which diminishes substantially phase
separation in glasses in the Si02-RO systems (MgO, CaO, ZnO, etc.) [8]; moreover the system
in question does not contain ~03' Introduction of Al20 3 into the system Si02- ZnO- Na20 leads,
as in other glasses, to interaction of Al20 3 with Na20 in the first instance and to localization of
sodium ions around [AI04/ 2]- tetrahedra which are incorporated in the silicon-oxygen network.
this diminishes the amount of "free" N~O tending to eliminate phase separation. On further in-
crease of the alumina content the amount of "free" N~O becomes insufficient and phase separa-
tion characteristic of the binary system Si02- ZnO occurs; it intensifies and reaches a maximum
at 5% alumina concentration, i.e., at the ratio A120 3 : Na20 = 1, when all the sodium oxide is com-
bined with alumina.
It has been shown [8] that Al20 3 is an active homogenizer of glasses in systems of the
Si02- RO type. In presence of N~O in the system in question, alumina could not have a homoge-
nizing effect in amounts up to 5%. When the Al20 3 content exceeds 5% it begins to act as a homog-
enizer, and 2-3% A1 20 3 uncombined with Na20 is sufficient to prevent phase separation; in pres-
ence of 7-8% Al 20 3 phase separation does not occur in the region of the system Si02-AI20 3- ZnO-
Na20 studied.
The causes of phase separation in the boron-free system discussed above, connected with
the contents and interaction of alumina and sodium oxide, persist in the five-component system
containing boron oxide and are manifested in the second region of phase separation. This follows
from the constancy of its position along the plane corresponding to the constant ratio Al 20 3 :
Na20 = 1 and from boundary values of the Al20 3 concentration in it. Introduction of ~03 into the
system Si0 2- Al20 3- ZnO - N~O intensifies phase separation, which is analogous to the substan-
tial increase of phase separation in the ternary system Si02 - ~03 - ZnO in comparison with the
binary system Si02- ZnO; this accounts for the certain broadening of the second region of phase
separation with increase of the ~03 concentration.
Thus, intensive interaction between A1 20 3 and N~O, leading to mutual suppression of their
homogenizing effects, is the dominant factor determining phase separation in the given system.
The presence of a second region of phase separation in systems containing MgO, ZnO, or
CaO, and its absence in systems with SrO, BaO, and PbO, indicate that the appearance of the
second region is attributable primarily to the individual characteristics of these cations. The
second region of phase separation appears in presence of cations having field strength in excess
of 1.5 and coordination type A according to Levin and Block [8].
It must be noted that these results were obtained under the specified conditions of heat
treatment. The possibility is not ruled out that under different treatment conditions the second
region could be detected in systems with cations having field strength below 1.5; primarily with
SrO. The important point is that in presence of these cations the tendency to phase separation
is lowered.

Literature Cited
1. I. V. Grebenshchikovand 0. S. Molchanova, Zh. Obshch. Khim., 12:588 (1942).
2. M. E. Nordberg, J. Am. Ceram. Soc., 27:299 (1944).
3. T. Abe, J. Am. Ceram. Soc., 35:284 (1952).
4. L. A. Grechanik, Inform.-Tekhn. Sb. Tsentr. Nauchn.-Issled. Elektrotekhn. Stekla, No.1,
p. 3 (1954).
L. A. BAL'SKAYA, L. A. GRECHANIK, AND N. M. VAISFEL'D 113

5. L. A. Grechanik, Inform.-Tekhn. Sh. Tsentr. Nauchn.-Issled. Elektrotekhn. Stekla, No.8,


p.3 (1957).
6. L. A. Grechanik and L. A. Bal'skaya, Elektron. Tekhn., Sere 14, Mater., No.5, p. 3 (1967).
7. S. P. Zhdanov and E. V. Koromaldi, Compt. Rend. VII Congr. Intern. du Verre, No. 302
(1965)
8. E. M. Levin and S. Block, J. Am. Ceram. Soc., 40:95, 113 (1957); 41:49 (1958).
PROPERTIES OF GLASSES EXHIBITING PHASE SEPARATION
IN THE SYSTEM Si02-B203-A1203-ZnO-Na20
L. A. Grechanik and L. A. Bal'skaya
Investigations have been carried out concerning the relationships between the composition,
heterogeneous structure, and various physicochemical composition of glasses in one of the sec-
tions (A) in the system Si 02 - ~03 - Al 20 3- ZnO - Na20, intersecting the two immiscibility regions
found in it [11. Glasses in this section have the following chemical composition: (69 - x)Si0 2
22~03 xA1 20 3 4ZnO 5Na20 (x = 0-8), and are close to a number of commercial and previously
investigated sodium borosilicate glasses.
The glasses were made from pure reagents in vitreous silica crucibles in a gas-fired fur-
nace; the results of chemical analyses of the glasses showed only slight deviations from the cal-
culated compositions. Annealed glass specimens were used for all the determinations. Stan-
dard methods were used for determination of the physicochemical properties of the glasses:
the coefficient of thermal expansion, the deformation temperature, and the glass transformation
temperature were determined with the aid of the DKV-1 vertical quartz dilatometer; viscosity
in the softening range was determined from the rate of extension of a rod extended under con-
stant load at constant temperature. The dielectric properties were measured at 106 Hz at room
temperature with the E9-4 instrument; the E9-3 instrument was used for measuring the elec-
trical resistance of glasses at various temperatures; the activation energy of conduction was
calculated from the formula PT = poe- E / 2kT . Chemical stability was determined by the standard
method from the loss in weight (in %) of a weighed portion of glass grains after treatment with
the particular reagents, in accordance with State All-Union Standard GOST 10134-62; the leach-
ing rate was determined by a conductometric method, from variations of the specific conduc-
tance of 0.001 N HCl solution during treatment of the glass powder. The method used for the
electron-microscopic investigations is described in the preceding paper [I].
The experimental results are presented in Figs. 1-3. The following characteristics should
be noted:
1) variations of the properties are generally represented by curves without sharp breaks;
2) most curves have smooth inflections near the point corresponding to 5% Al20 3 in the
glass, i.e., at the ratio A120 3 : N~O = 1.
3) the transformation from an interconnected labyrinth structure to a structure consisting
of isolated droplets observed at 2% Al20 3 [1] has a sharp influence only on the chemical stability
and leaching rate of the glasses.
The interpretation of these results was based on the concepts put forward by Zhdanov and
Koromaldi [2] on interaction of Al 20 3 with Na20 in low-alkali borosilicate glasses, with
the assumption that certain physicochemical properties of two-phase glasses are determined
predominantly by the corresponding properties of one of the phases [3].
114
L. A. GRECHANIK AND L. A. BAVSKAYA 115

Fig. 1. Electron micrographs of glasses in section A after poly thermal


treatment. 0, 1, 2 .. 8) % Al20 3 in glass.

800 .00
....
..::..
500 v
t!J .'"
....
'100
0...
fO IO >.
tOll .~
0

1O f2 .~
:>

~ :>
v
Cl..
be 9 t.o ~.

.Q

.
0
,.... '10
5.5 Fig. 2. Physicochemical properties
of glasses in section A. The in-
<u
'<l 5.2 fluence of A1 20 3 on viscosity is rep-
....
be resented by isokoms. The corre-
!f.B sponding viscosity values are indi-
cated on the right.
116 L. A. GRECHANIK AND L. A. BAL'SKA YA

320
... t.O
2
I

Eiu
...
I
c:
'"0 8
~ 7
x 0
<l 5.~
3

0 'fO 80 tZO t50


T, min
Fig. 3. Kinetics of leaching of glasses in section A.
Symbols as in Fig. 1.

The results do not contradict the scheme put forward by us [1] for the phase-separation
process in this system, and make it possible to draw certain inferences with regard to the com-
positions of the coexisting phases.
The first glass (without A120 3) is a sodium borosilicate glass of the type stuc;Iied in detail
earlier (e.g., Na 7/23, Na &,25) [4], containing 4 % ZnO. The heterogeneous structure consists
of a labyrinth of continuous channels (Fig. 1); the volume of the second phase is about 20-30%.
By analogy with literature data [4-6] it may be supposed that the matrix consists predominantly
of a silicon-oxygen framework containing a very small amount of Na20 together with a little
~03. The second phase consists of sodium borate complexes and it contains nearly all the
ZnO [7]. The physicochemical properties of this glass (Figs. 2 and 3) are consistent with this
supposition: it has the highest deformation temperature and viscosity and the lowest transforma-
tion temperature; the low dielectric losses and high electrical resistance, which are determined
by the second phase in this glass, indicate fairly stable trapping of Na+ ions near the [B0 4d-
tetrahedra.
Introduction of 1-2% A1 20 3 (region of decreasing phase separation) results in considerable
lowering of the deformation temperature and decrease of viscosity, and increase of the transfor-
mation temperature. This can be attributed to enrichment of the silicon-oxygen matrix with the
oxides of aluminun, sodium, and boron owing to formation of [Al0 4 /2]-Na+, groups which are in-
corporated into the silicon- oxygen framework, with breakdown of borate complexes and transfer
of "liberated" B20 3 into the matrix, i.e., loss of the oxides of sodium and boron from the second
phase. The change in the position of Na+ ions, which become located near the [AlO4/2]- tetra-
hedra results in increase of dielectric losses and decrease of electrical resistance and activa-
tion energy of conduction. Entry of A1 20 3 into the matrix probably raises its surface tension so
that the second phase loses its continuity and a fine vermiform and predominantly isolated struc-
ture is formed; this is accompanied by a substantial decrease of the volume of the second phase
(Fig. 1). The chemical stability in H20 reaches Virtually the maximum value and the stability in
HCI increases considerably; there is a sharp change in the kinetics of glass leaching, indicated
by variations ~ 'X of the conductivity of the solution (Fig. 3).
On further increase of the Al 20 3 content to 3-5% (region of increasing phase separation) the
main physicochemical properties of the glasses continue to vary in the same direction as in pres-
ence of 1-2% Al 20 3 up to the composition containing 5% A1 20 3, i.e., corresponding to the ratio
A1 20 3 : Na20 = 1. It may be supposed that here the alumina continues to bind sodium oxide and
transfer it and the "liberated" boron oxide from the second phase into the matrix. However, the
properties vary less rapidly, and the volume of the second (droplike) phase distribution in the
matrix increases in this range, reaching a maximum at the ratio A1 20 3 : Na20 = 1. The drop size
increases, the clearing temperature rises [1], and the transformation temperature ceases to in-
crease. These effects may be attributed to occurrence of another process in this range of A1 20 3
L. A. GRECHANIK AND L. A. BAL'SKAYA 117

concentrations - probably transfer of Si02 from the matrix into the drops of the second phase
with formation of silicate- zinc complexes, i.e., the composition of the second phase changes
even more significantly. Increase of the Al 20 3 content in the matrix increases its chemical sta-
bility in acid as before and lowers the leaching rate.
Glasses containing 6-8% Al 20 3 (region of decreasing phase separation) show considerable
decrease of the particle size of the second phase and of its total volume. In this region intensive
homogenization of the glass by alumina occurs, probably involving combination of the alumina
with zinc oxide accompanied by breakdown of the silicate- zinc complexes and formation of a
unified aluminum zinc silicate structure. The changes of physicochemical properties are con-
sistent with this hypothesis: the deformation temperature, electrical resistance, and activation
energy of conduction increase somewhat while the chemical stability in acid diminishes substan-
tially (Fig. 2).
The explanation that homogenization in this range of Al 20 3 concentrations (6-8 %) is due to
formation of complexes of the oxides of aluminum and zinc is also confirmed by the fact that
with higher ZnO contents (than present in glasses of the section studied) a larger amount of
Al 20 3 is needed for homogenization of the glass [11.
The results of this investigation of the influence of alumina on phase separation and proper-
ties of low-alkali borosilicate glasses containing oxides of the RO type suggest that among the
complex processes of component interaction in these glasses the processes of interaction of
Al 20 3 with Na20 and then with RO are predominant. The intensification and weakening of phase
separation accompanying these processes make it possible to detect their possible occurrence
with greater certainty.
Literature Cited
1. L. A. Bal'skaya, L. A. Grechanik, and N. M. Vaisfel'd, this volume, p. 107.
2. S. P. Zhdanov and E. V. Koromaldi, Compt. Rend. VII Congr. Intern. du Verre, No. 302
(1965).
3. 0. V. Mazurin, this volume, p. 37.
4. S. P. Zhdanov, structure of Porous Glasses and structural Transformations in Sodium
Borosilicate Glasses, Doctoral dissertation, Inst. Khim. Silikatov, Akad. Nauk SSSR,
Leningrad (1959).
5. T. Tran and C. Sella, Compt. Rend., 256(14):3063 (1963); 259(6):1325 (1964).
6. R. J. Charles, J. Am. Chern. Soc., 47:559 (1964).
7. L. A. Grechanik and L. A. Bal'skaya, Elektron. Tekhn., Ser. 14, Mater., No.5, p. 3 (1967).

DISCUSSION
In reply to questions relat ing to the paper by Grechanik and Bal' skaya, G r e c han i k re-
ferred to difficulties in determining the number of phases in a multicomponent system and
the occurrence of secondary separation from electron micrographs.
INFLUENCE OF LEAD OXIDE ON THE STRUCTURE
AND PROPERTIES OF BOROSILICATE GLASS

N. M. Pavlushkin and Zh. A. Olobikyan

Considerable attention is still being devoted to investigations of borosilicate glasses, which


have peculiar properties, but the results are controversial in most cases. For example, it is
known that addition of oxides of the RO type (CaO, MgO, PhO, etc.) and of Al20 3 in amounts up to
5 moleo/c sharply raise the chemical stability of three-component borosilicate glasses [1-3], but
while additives of the RO type cause strong opalescence up to complete opacification of the
glasses Al 20 3 has the opposite effect. On the other hand, the same cations differ in their effects
on the coordination number of boron [4],
The subject of the present investigation was the three-component glass No. 8 of the follow-
ing composition (mole%): Na20 5, B20 3 25, Si02 70. According to the phase diagram, this com-
position lies in the region of glasses with a tendency to phase separation after prolonged heat
treatment. Lead oxide was introduced in amounts from 5 to 50 mole% of the original glass:

GlassNo. PhO GlassNo. PhO


8 95 25.0
79 5.0 96 30.0
80 10.0 97 35.0
81 15.0 98 40.0
84 20.0 111 50.0

The glasses were made in sintered corundum crucibles in a SHit furnace. According to
the results of chemical analysis, the amount of Al20 3 dissolved out of the crucibles was about
0.26%. The melting temperature was varied from 1100 to 1300, dependent on the lead oxide con-
tent. Joint volatilization of boric anhydride and alkali, amounting to 10%, was taken into account
in batch preparation. The prepared glasses were slightly opalescent and had a yellow tinge,
which became deeper with increase of the lead oxide content.
Chemical stability of the glasses was investigated by a conductometric method, based on
changes in the conductivity of a suspension consisting of the powdered glass in 0.001 N HCl. The
change of conductivity of the suspension after 15 min from the start of the experiment was taken
as the measure of the chemical stability of all the glasses. This change is designated ~ 'X15'
The variation of chemical stability of the glasses with composition is shown in Fig. 1, 3.
The graph shows that with 20 mole% PbO in the glass the dependence of ~}(15 on composi-
tion passes through a maximum, Le., this glass has the minimum chemical stability.
The dilatometric curves, which have an inflection at temperatures in the region of 475-
480, are of special interest (Fig. 2). The coefficient of thermal expansion of the glasses studied
before the inflection point lies in the range of (43 -70) . 10- 7 deg- 1 , dependent on composition.

118
N. M. PAVLUSHKIN AND Zh. A. OLOBIKYAN 119

~
I

a
u
~
I
C

::l. . ::
0
....'"
\'0
,....,

;"'
~bO
.... 720 <I
....I 16
...
~
640
14
300 560
Fig. 1. Variations of the length of the
12
480
portion of the curve between t f and t g
100
10 (1), the softening temperature (2), and
~ 'X15 (3) with composition.

0 10 20 30 40 50
PbO, mole"!o

Above the inflection it increases to (250-300) .10- 7 deg- 1 The region of the curve between the
inflection and the softening temperature, corresponding to a high coefficient of thermal expan-
sion of the glass, is designated here as l(tf-t~. The dependence of the length of this region on
composition is shown in Fig. 1, curve 1. The great length of the region between tf and tg on the
dilatometric curves of certain boron-containing glasses has been noted by Tool [5], Old-
field [6], and Mori [7]. This effect was attributed to phase separation. Liquid-liquid
phase separation also occurs in the glasses studied by us. Studies with the aid of the electron
microscope confirm the supposition that introduction of lead oxide into the original glass alters
its structure, leading to formation of microheterogeneity regions of the liquid-liquid separation
type. The greatest degree of heterogeneity is found in glass containing 20 mole% PbO, which is
in good agreement with the minimum of chemical stability noted above. The same glass has the
longest interval between the inflection and softening points on the dilatometric curve and the
highest softening temperature (Fig. 1, 1 and 2).

The infrared absorption spectra of some of the glasses studied were recorded, in order
to study variations of the coordination number of boron. It is known from investigations of crys-
talline borates [8] that boron in threefold coordination gives strong absorption in the region of

~OO

tf
300 84

...
::l.

ZOO
<I

Fig. 2. Dilatometric curves of


glasses containing 5 (No. 79) and tOO
20 (No. 84) mole% PbO.

0 400 800 to
120 N. M. PAVLUSHKIN AND Zh. A. OLOBll(YAN

111
r::
.~
., 98
'6;g
C<!
.... 8'1
E-<
79
8

1350 1080

1'100 1200 1000 800 700 000 500 '100


Wave num ber. em -1
Fig. 3. Infrared absorption spectra of the glasses
studied.

1100-1300 cm- 1 and additional absorption at 700-7S0 cm- 1 Boron in fourfold coordination is
characterized by strong absorption at SOO-1100 cm- 1 and additional absorption in the 600-S00
cm- 1 region. The infrared spectra of glasses containing 5, 20, 40, and 50 mole% PbO and of the
original glass are shown in Fig. 3. Introduction of PbO causes first weakening and then total
disappearance of the peak at 930 cm- 1 , corresponding to boron in fourfold coordination. It should
be noted that the change in the coordination number of boron is observed only in compositions
which undergo phase separation. No significant changes are observed in the IR spectra of ho-
mogeneous glasses having high lead contents (40-50 mole%).
Introduction of lead oxide into sodium borosilicate glass has a complex effect: in small
amounts the large lead cations cause loosening of the glass network. Since the original glass
has a tendency to phase separation, loosening of the network increases this tendency, with re-
sultant appearance of small heterogeneity regions. This is manifested in a decrease of chemical
stability, and is confirmed by electron micrographs. The most heterogeneous glass contains
20 mole% PbO and has the lowest chemical stability and the longest softening range. Homogeniza-
tion of glasses with higher PbO contents can probably be attributed to the fact that lead begins
to act as a glass former.

Literature Cited
1. J. Yamamoto, J. Ceram. Assoc. Japan, 61:255 (1954).
2. J. Tamura, Atti III Congr. Intern. del Vetro, Venice (1954), p. 426.
3. O. S. Molchanova, Sodium Borosilicate and Porous Glasses, Oborongiz, Leningrad (1954).
4. A. A. Appen, Zh. Prikl. Khim., 26:569 (1953).
5. A. Q. Tool, J. Am. Ceram. Soc., 29:240 (1946).
6. L. F. Oldfield, Glass Technol., 5:150 (1964).
7. Yoshio Mori, Bull. Govt. Ind. Res. Inst. Osaka, 15:194 (1964).
S. C. E. Weir and R. A. Schroeder, J. Res. U. S. Natl. Bur. Standards, 6SA:465 (1964).
N. M. PAVLUSHKIN AND Zh. A. OLOBIKYAN 121

DISCUSSION
Referring to the anomalous variation of the coefficient of thermal expansion observed by
Pavlushkin and Olobikyan, Klyuev and Mazurin asked whether any connection or correla-
tion exists between the observed effect and the glass transformation temperature. 0 lob i k-
y a n replied that, in her opinion, the anomalous bend of the dilatometric curve is not con-
nected with the transformation temperature.
INFLUENCE OF THE COMPOSITION AND RATIO
OF THE SEPARATED PHASES ON STRESSES ARISING
DURING LEACHING OF TWO-PHASE
SODIUM BOROSILICATE GLASSES

O. K. Botvinkin, N. L. Mironova, and G. L. Shpilevskaya

It is known that certain glasses in the system Na20-~03-Si02 separate into two phases
during heat treatment: a low-melting phase (L), containing mainly Na20 and ~03' and a high-
melting phase (H), with a high silica content. The low-melting component of these glasses is re-
moved by treatment with acids, while the remaining high-silica skeleton forms a glass penetrated
by minute pores [1, 2].
During leaching of the heterogeneous glass stresses arise in the inner zone, not yet ex-
tracted; the magnitude of these stresses approach the tensile strength of the glass. This often leads
to crac king.
All investigators attribute these stresses to volume changes during conversion of the alkali
borosilicate into porous glasses. At the same time, the causes of the volume changes themselves
are not clear, and have been interpreted differently by different workers [3-6]. The main con-
clusion reached by previous investigators is that shrinkage and tensile deformations arising
during leaching are the consequence of relaxation of microstresses existing in the glass after
heat treatment. Analysis of data on this subject shows that the following factors may be respon-
sible for the volume change of the porous part of the glass during leaching: differences between
the coefficients of thermal expansion and of the interfacial tensions of the microphases; possible
volume changes due to adsorption of water vapor in the narrow capillaries of the porous glass.
As sone of the above-mentioned factors are to a certain extent determined by the composi-
tions and ratio of the separated phases, it is of interest to examine the effect of variations of
these on stresses arising during leaching.
Glasses in the system Na20. 4B20 3-Si02 [7], previously studied by Rockett, were investi-
gated. The positions of the glasses studied and the phase diagram are shown in Fig. 1. The in-
fluence of the phase ratio was studied with six compositions, limited by Si02 contents of 60-77%
(region aa1)' All these glasses, heat-treated at 560 have the same composition, determined by
0

the end points of the isothermal line I-I', at different phase ratios. The ratio of the equilibrium
phases in the two extreme compositions studied, a and a1' varies from L/ H == 1/1.4 at point a
to L/H = 1/5 at point a1; thus, the amount of low-melting phase changes approximately threefold
between the extreme compositions of the region studied in the equilibrium state.

122
0. K. BOTVINKIN, N. L. MIRONOVA, AND G. L. SHPILEVSKAYA 123

1600

flfOO

1Z00

tODD

Fig. 1. Phase diagram of the sys-


tem Na20'4B203-Si02 [7] and posi-
tions of the compositions studied.
400

200

LIZOL2
NazOlfBzDs
wt.%

The time required for establishment of equilibrium at the chosen temperature of heat treat-
ment (560), determined from changes of structural birefringence arising in sodium borosilicate
glasses during cooling from a temperature above tg under load [8, 9], was 5-10 h.
Polished plates 25 x 10 x 2 mm in size were used for the stress measurements. They
were leached in 3 N Hel at 20-22 in a thermostat. The stresses (path difference in mJ1) were
measured with the aid of a polarizing microscope and a rotary ring compensator. During mea-
surements of the optical path in the porous and nonporous layers their thickness was recorded.
In every case compressive stresses were observed in the porous layer and tensile stresses
in the unleached layer during leaching. In order to eliminate the influence of leached layers per-
pendicular to the optic axis (25 x 2 plane) the measured path difference was calculated for a stan-
dard thickness of 1 cm.
Figure 2 shows the results of determinations of stresses in the unleached and porous por-
tions of glasses in region aa1' The measured stress was referred to 2d/ d, where d 1 is the
thickness of one porous layer and d is the thickness of the specimen leached.
In the range of glasses studied the stresses increase to 4500 mp/cm with increasing thick-
ness of the leached layer; as the photoelastic constant of these glasses is 3.4' 10- 7 cm 2jkg, this
corresponds to '" 1300 kg/ cm2.
Figure 2a shows the dependence of stresses arising in the unleached parts of the glasses
on the ratio of the separated phases (Si02 contents in the glasses) at various 2d/ d ratios. It is
seen that variation of the phase ratio in the range studied has only a slight influence on the
stresses and is not the determining factor.
The influence of variation of the composition of the separated phases on the stresses was
tested with glass b (Fig. 1) treated at various temperatures in the 500-700 range. These tem-
peratures are indicated by points on the section II-II'. Each temperature corresponds to definite
phase compositions. Variations of the microphase compositions with the temperature of heat
124 O. K. BOTVINKIN, N. L. MffiONOVA, AND G. L. SHPILEVSKA YA

a b

5000 %SiOz 5000


075
4000 () 72.5 4000 Zdt
8u " 70.5
"8B3000
~ T
"8 3000 x 08
0.4
.
o
::l.. .. 05 ::l..

o ZOOO 00 15 2000 ~ 0.0


() 0.8

fOOD fOOO ~
Zdt 0
0 0.2 0.4 0.0 0.8 tOT 05 70 75 SiOz,%
f:Z f:3 1:4 1:5 L:H
Fig. 2. Dependence of stresses on the phase ratio (ratio of Si~ con-
tents) in glasses of region aa1' a) stress variations in the unleached
parts of the glasses; b) data of Fig. 2a for different thicknesses of the
porous layer.

treatment are illustrated by the points L t , L:!. L3 and Hi. H2 H3 (Fig. 1). The time required for
establishment of equilibrium ranged from 1.5 h at 700 to 10 days at 500. 0

Variations of stresses in the unleached and porous parts of glass b are shown in Fig. 3.
The dependence of the stresses on the temperature of heat treatment at various 2d/ d ratios is
effectively illustrated in Fig. 3b. It is seen that the stresses increase with the heat-treatment
temperature up to 600 and then decrease. 0

a, mlllcm
a, mil/em
a b
'2 2dt
4000
0
'1;\
.500 T
=
G) 05'10 0.3
~ x 500
G) 3000 ... 3000 o 0.5
13>. (J 600
G)
>.
!!
(J 0.8
!! .. 650 "0
"0 2000 g 700
G)
G)

'tic<I 'tic<I
G)

~
G)
'2
'2 fOOO ::J fOOD
::J

~
0
f.Od
Zdt
a
:JJ700t'
~
>. ....
en
:: i! -fOOO
c<I
I-fooO
:::l en
e008
0. :::l

0. U ~
0.
~ -2000 -2000

Fig. 3. Variations of stresses in glass b with the chemical composi-


tions of the microphases (see Fig. 1, section IT-n') , a) Variations of
stresses in unleached and porous parts of glass b after heat treatment
at different temperatures; b) the data of Fig. 3a for different thickness-
es of the porous layer.
O. K. BOTVINKIN, N. L. MffiONOVA, AND G. L. SHPILEVSKAYA 125

The observed relationship cannot be attributed only to changes in the compositions of the
separated phases, as it has been shown that heat treatment also influences the structure (size
and shape of the microphases [10]) of the glasses, and this undoubtedly affects the magnitude of
the stresses.
It follows that the magnitude of the stresses arising during leaching of two-phase sodium
borosilicate glasses depends on the structure and composition of the microphases.
The phase ratio in the composition range studied has no significant influence on the
stresses.
Literature Cited
1. O. S. Molchanova, Steklo i Keram., No.5, p. 5 (1967).
2. H. P. Hood and M. E. Nordberg, U. S. Pat. No. 2,106,744; No. 2,215,039.
3. K. T. Bondarev, Steklo i Keram., No.1, p. 19 (1961).
4. Hayashi Jiro, Nishiraku Takashi, and Ohtani Katsuya, Mitsubishi Denki Giho, 39(9) :44 (1965).
5. o. S. Molchanova, Optika i Spektroskopiya, No.1, p.917 (1956).
6. K. Kiihne, Z. Phys. Chern., 204:20 (1955); Silikattech., No.8, p. 338 (1954).
7. N. 1. Ananich and O. K. Botvinkin, Steklo i Keram., No. 10, p. 10 (1962); Optiko-Mekhan.
Prom., No.8, p. 45 (1962).
8. N. 1. Ananich, O. K. Botvinkin, and M. L. Mironova, Steklo i Keram., No.1, p. 15 (1965).
9. T. J. Rockett, W. R. Foster, and R. G. Ferguson, J. Am. Ceram. Soc., 48:329 (1965).
10. D. P. Dobychin and N. N. Kiseleva, Dokl. Akad. Nauk SSSR, 113:372 (1957); S. P. Zhdanov,
in: Structure of Glass, Izd. Akad. Nauk SSSR, Moscow- Leningrad (1955), p. 162 [English
translation: The Structure of Glass, Vol. 1, Consultants Bureau, New York (1958), p.125].
STRUCTURAL PHYSICAL STATE OF
MICROHETEROGENEOUS GLASSES

S. T. Sil'vestrovich and V. D. Kazakov

It is known [1, 2] that double-network borosilicate glasses have distinct microheteroge-


neous structure and accordingly have a tendency to microphase separation [3-5]; this is con-
vincingly confirmed, in particular, by the character of leaching of sodium borosilicate glasses
[2, 6].
Formation of etched cavities in commercial Pyrex glass during chemical etching is also
attributed predominantly to the presence of regions of chemical heterogeneity as the result of
microphase separation in this glass [7].
Anomalous birefringence may arise in microheterogeneous glasses as the result of inter-
action between the high-melting (silicate) and low-melting (borate) networks [8], as the result
of orientation of asymmetric structural elements [9] or as the consequence of freezing of high-
elastic deformations [10].

It has been shown [11] that the magnitude and character of anomalous birefringence depend
on the chemical nature (composition) of the glass and the conditions of interaction of the struc-
tural networks. Special interest attaches in this connection to investigations [12, 13] of the
microheterogeneous structure of glass on its physical and mechanical properties, and in partic-
ular to data [13] on the loss of strength of glass under the influence of microphase separation
caused by increase of the temperature of heat treatment of sodium borosilicate glass.
Our experiments in which the surface layers of heat-strengthened Pyrex glass of indus-
trial manufacture (cylindrical rods 4 mm in diameter) were progressively etched with hydro-
fluoric acid solutions, as in the investigation cited [7], distinctly revealed (Fig. 1) the selective
character of chemical etching, due to the microheterogeneous character of this glass and to de-
velopment of microphase separation in it. At the same time, our micrographs of the etched
glass surface (Fig. 1) lead to the conclusion that the structure of the microheterogeneous glass
is influenced considerably by the forming (drawing) conditions; in particular, by thermal and
technological heterogeneities (stratification in the rod) caused by the specific forming conditions.
Therefore formation of bottle-shaped cavities located concentrically about the rod axis is ob-
served at the end of the glass rod as the result of etching (Fig. Ib).
Formation of separate microphase (liquid -liquid phase separation) differing in chemical
nature, density, and thermal properties, is naturally a source of structural imbalence in the
glass, the effects of which will be especially pronounced under external thermomechanical in-
fluences. To test this hypothesis, Pyrex glass was heat-strengthened in sodium polyethylsili-
conate (GKZh-IO), quenching in which is especially unfavorable, as it produces predominantly

126
S. T. SIVVESTROVICH AND V. D. KAZAKOV 127

Fig. 1. Photomicrographs of the surface of Pyrex glass tempered in GKZh-10


silicone liquid, after chemical etching in 40% hydrofluoric acid (x 25). a, b)
Cylindrical surface after dissolution of layers 570 p, (a) and 820 p, (b) thick; c-e)
end face of rod after dissolution of layers 15 p, (c) 350 p, (d), and 720 p, (e) thick;
f, g) development of cracks on the surface after dissolution of layers 102 p, (f)
and 290 p, (g) thick; h-j) development of crack profile after dissolution of layers
15p, (h), lOOp, (i), and 290p, 0) thick.
128 S. T. SIL'VESTROVICH AND V. D. KAZAKOV

2.6
Z.t,.
2.2
0' 2.0 Fig. 2. Variations of the coefficient
~ of strengthening (oq+/uq ) as the re-
G) 1.8
6- sult of chemical etching of heat-
tl 1.6
strengthened Pyrex glass with slight
1.4 (1) and considerable (2) changes of
1.2 surface relief after etching (based
1.0 _ - - . L - - _ . L - _ - - - ' on the results of bending-strength
300 tests on rods).
Thickness of layer removed,ll.

temporary thermoelastic stresses in the glass in the brittle state, which must naturally favor
active development of microcracks in the heterogeneous glass, especially on its surface. In
fact, such cracks were clearly revealed during subsequent etching in hydrofluoric acid solutions
of glass quenched in this way; it was found that these cracks on the glass surface (Fig. Ic) are
similar in character to those formed on glass as the result of thermal shock [14].
As chemical etching proceeds the profile of these cracks alters and their edges become
smoother (Fig. Id), but on the whole the surface relief of microheterogeneous glass subjected to
this drastic quenching (thermal shock) becomes much more distinct during etching than in the
case of ordinary homogeneous glass. Naturally, this must affect the strength of the glass. The
data in Fig. 2 represent variations of the coefficient of strengthening during etching of specially
selected specimens of heat-strengthened Pyrex glass varying in surface defectiveness. It is
seen that the more defective (relief) glass has a lower capacity for chemical strengthening.
In addition to this weakening effect of increased fissuring of microheterogeneous glasses,
it is also necessary to take into account another characteristic of these glasses - structural
anisotropy, which develops as the result of thermomechanical treatment under certain conditions
and is accompanied by appearance of anomalous birefringence [11]. The cylindrical rods (3-5
mm in diameter) of commercial MKR-l glass, similar in composition to Pyrex, used in our
experiments showed negative anomalous birefringence, varying from 0 to 180 m/1/cm, even after
the usual factory production (hand working) and annealing. Nonex commercial electrical glass
was taken in order to produce a greater degree of structural anisotropy in a microheterogeneous

MKR-l
() Nonex

Fig. 3. Influence of structural an-


isotropy (~an> of MKR-1 and Nonex
glasses on their bending strength.
3a~ ____~~ ____- L _ _ _ _ _ _~
o 300 sao 900
S. T. SIVVESTROVICH AND V. D. KAZAKOV 129

W,% MKR-l w,% Nonex


100 100
1 1

50 50

o 50 100 a 50 fDa
Bending strength, kg/mm 2
Fig.4. Integral functions of the probability
of destruction of MKR-1 and Nonex glasses
differing in structural physical states. 1)
Original defective glass; 2) low degree of
surface defectiveness, etched to depths of
50/J (MKR-1) and 100/J (Nonex); 3) low
degree of surface defectiveness, etched to
depths of 50/J (MKR-1) and 100/J (Nonex),
exhibiting anomalous birefringence of
182 m/J/cm (MKR-1) and 900 m/J/cm (Nonex).

glass. Cylindrical rods (4-6 mm in diameter) made from this glass and subjected to additional
thermomechanical treatment and low-temperature annealing exhibited negative anomalous bire-
fringence in the range from 0 to 900 m/J/cm. * After chemical removal of the fissured surface
layer (50-100/.1 thick) from these specimens of glasses exhibiting structural anisotropy their
bending strength was determined. The results (Fig. 3) show that structural anisotropy arising
as the consequence of microphase separation has a weakening effect; the strength of glass with
a low degree of surface defectiveness decreases steadily (by a half at the maximum) with increase
of structural anisotropy.
The weakening influence of structural anisotropy in a microheterogeneous glass is objec-
tively confirmed by statistical strength characteristics - curves for the integral probability func-
tions of destruction W (Fig. 4) of the original glass with a defective surface, glass with a low
degree of surface defectiveness (chemically etched), and similar glass with structural anisotropy.
The mechanism of weakening of glass as the result of microphase separation appears to
be of complex character, determined by both the physicochemical and the structural incompatibil-
ity of microphases developing during phase separation in the glass. Therefore, as was shown
above, the strength decrease of the glass in the present instance may be the consequence of a
predominant influence either of surface defects or of structural anisotropy, i.e., at least two dif-
ferent phenomena characteristic of the structural physical state of microheterogeneous glasses.
Literature Cited
1. E. A. Porai-Koshits and Yu. G. Sokolov, Dakl. Akad. Nauk SSSR, 72:477 (1950).
2. S. P. Zhdanov, Dokl. Akad. Nauk SSSR, 92:597 (1953).
3. W. Vogel, Silikattech., 10:241 (1959).
4. M. Watanabe, J. Am. Ceram. Soc., 42:593 (1959).
5. K. T. Bondarev and V. A. Minakov, Steklo i Keram., No. 12, p. 22 (1960).

*The authors thank N. I. Ananich for help in preparing the Nonex glass specimens and determina-
tion of birefringence.
130 S. T. SIVVESTROVICH AND V. D. KAZAKOV

6. D. P. Dobychin, in: The Glassy State, Proceedings of the Third All- Union Conference on
the Structure of Glass, Izd. Akad. Nauk SSSR, Moscow-Leningrad (1960), p. 480 [English
translation: The Structure of Glass, Vol. 2, Consultants Bureau, New York (1960), p. 432J.
7. D. G. Holloway and D. M. Schlapp, Nature, 189(4762):385 (1961).
8. V. L. Indenbom, Dokl. Akad. Nauk SSSR, 89:509 (1953).
9. B. Lang, M. Kantzer, and M. Orlu, Trav. IV Congr. Intern. du Verre, Paris (1957), p. 292.
10. G. M. Bartenev and A. S. Eremeeva, Vysokomol. Soedin., 11: 508 (1960).
11. O. K. Botvinkin and N. 1. Ananich, Steklo i Keram., No.9, p. 6 (1959); No. 10, p. 10 (1962);
Optiko-Mekhan. Prom., No.8, p. 45 (1962).
12. M. Watanabe and T. Moriya, Rev. Electr. Commun. Lab. Nippon Telegraph and Telephone
Publ. Corps., 9:1 (1961),
13. K. T. Bondarev and S. E. Dvorkina, in: The Glassy State, Vol. 3, No.2, Mechanical Prop-
erties and Structure of Glass, Izd. OMP, Leningrad (1963), p. 38.
14. Ya. B. Fridman, E. M. Morozov, and S. S. Solntsev, Steklo. Byull. Vses. Nauch.-Issled.
Inst. Stekla, No.4, p. 44 (1963).

DISCUSSION
A nan i c h, commenting on the paper by Sil'vestrovich and Kazakov, said that the
effects observed by them are very weak and may be masked by other more prominent
factors. In evaluation of the data it is necessary to take into account the stresses applied
to the glass during drawing, the elastic properties of the constituent microphases, and other ef-
fects. Discussion of strength characteristics must be based on statistical variance of the exper-
imental data.
In reply to these comments, S iI' v est r 0 vic h emphasized that the aim of the in-
vestigation was to elucidate the influence of heterogeneous structure on the mechanical properties
of glass, with the influence of surface defects eliminated as far as possible. Definite conclusions
can be drawn from the results, with statistical variance of the data taken into account: The au-
thors consider that direct estimation of microstresses between the phases must be undertaken in
subsequent work. The comment of Ananich regarding the influence of drawing on strength
is valid; such effects were observed by the authors.
USE OF DIFFERENTIAL THERMAL ANALYSIS AND
DILATOMETRIC MEASUREMENTS FOR STUDY OF
PHASE-SEPARATION PHENOMENA IN GLASSES*

M. B. Usvitskii

Observations of changes in the thermal properties of glasses during heating for study of
phase transformations are being used increasingly in research work.
Dilatometric measurements and differential thermal analysis are among the most usual
methods of this kind. Both methods have the same physical basis and are mutually supplemen-
tary.
It has already been shown [11 that when glass is heated in the softening range the differen-
tial thermal analysis (DTA) curve shows a sharp break due to a change in the specific heat of
the glass; the temperature of this break coincides with that of an analogous break on the curve of
linear thermal expansion (LTE).
In an investigation of the thermal properties of glass by the DT A method [21 the depen-
dence of the temperature at which the DT A curve shows breaks on the heating rate was determined.
This dependence showed that the temperature at which the course of the curve changes in the
softening range corresponds to the glass transformation temperature. The temperature limits
of the break on the DTA curve in the softening range vary in accordance with the same law, hav-
ing a dependence on the heating rate determined earlier from LTE curves [3, 41.
In studies of DTA curves for glasses corresponding to certain compounds of definite
stoichiometric composition [51 it was found that the DTA curve, like the LTE curve, has a single
change of direction in the softening range. It was shown that over the entire temperature range,
up to the point where the glass becomes soft and the dilatometric curve no longer characterizes
linear thermal expansion, the breaks on the DTA and LTE curves occur at the same tempera-
tures and are caused by structural transformations occurring in the glass.
Figure 1 shows curves for the dependence of the temperature and elongation of the speci-
men on the temperature of the ambient medium during heating at a constant rate (curves 1 and 3),
and also the DTA and LTE curves (2 and 4). In absence of endo- and exothermic effects the
breaks on the DTA curve are due to changes of the specific heat; therefore in regions free from
such effects the DTA curve and the LTE curve represent the first derivatives of the enthalpy and
volume, respectively.

*In view of the fact that M. B. Usvitskii's contribution to the discussion contained experimental
material of general interest, it was decided to publish it as a separate brief communication
(editorial note).
131
132 M. B. USVITSKII

3.5

3.0

2.5

2.0
.d
as
.....s
E-< 1.5

1.0

0.5

0 200 400 600 800 1000 1200 1400


I I I I I I

-20 -10 0 10 20 30
Temperature difference (tsp-tst), deg
I I I I !

0 1 Z 3 " '10 2
Elongation, 11
I I I I I I

0 50 100 150 ZOO 250


Coefficient of linear expansion, ex.1 07, deg- 1
Fig. 1. DTA curves and dilatometric curves for
glass corresponding to cordierite in composition.
1) Temperature of the specimen; 2) temperature
difference between the specimen and the reference
standard; 3) elongation of the specimen; 4) coef-
ficient of linear thermal expansion.

Comparison of the course of the curves shows that they have breaks at the same tempera-
tures.
The DTA curves of glasses in which one liquid phase is formed during softening (Fig. 1)
show a single change of course. This portion of the curve can be described by parameters char-
acterizing the temperature difference arising with increase of the lag between the temperature
of the glass and the temperature of the medium,~ta' and the time 'fa during which the glass
passes from the brittle into the viscofluid state [2].
M. B. USVITSKII 133

3.0

2.5

2.0

1.5

1.0

o ZOO 400 600 800 fOOO 1Z00 1400 to


o 5 10 15
Temperature difference (tsp-tst), deg
I I

o
!

2 3 4-
Elongation, J.l
! !

o 50 100 150 200 250


Coefficient of linear expansion, ex .107, deg- 1

Fig. 2. DTA curves and dilatometric curves for so-


dium borosilicate glass (determined jointly with M. L.
Mironova). Symbols as in Fig. 1.

It can be inferred theoretically that if several liquid phases are formed in the glass on
heating the DTA curve will show several changes of direction (corresponding to the number of
phases) in the phase-separation region.
This effect is clearly seen in the DTA curves given by Botvinkin and Mironova [6]. Curves
obtained during heating of a borosilicate glass are shown in Fig. 2. Over the entire tempera-
ture range where dilatometric measurements are possible the breaks on the DTA curve corre-
spond to analogous breaks on the LTE curve. In the softening range the DTA curve shows three
changes of direction, corresponding to three distinct maxima on the LTE curve; the intensity of
these maxima diminishes relatively to the DTA data with decrease of the glass viscosity.
Without going into details of the investigation of the nature of the breaks in the curves in
relation to study of liquid-liquid phase separation in glasses (as this is the subject of an exten-
sive independent study), we may note that combined differential-thermal and dilatometric in-
vestigations may prove very promising in study of problems associated with phase-separation
effects.
134 M. B. USVITSKII

Literature Cited
1. G. Tammann, The Vitreous State, ONTI, Leningrad-Moscow (1935).
2. M. B. Usvitskii, Tr. Inst. Stekla, "Steklo," No.2, p. 128 (1966).
3. G. M. Bartenev, Dokl. Akad. Nauk SSSR, 76:227 (1951).
4. G. M. Bartenev and I. A. Luk'yanov, Zh. Fiz. Khim., 29:1486 (1955).
5. I. I. Kitaigorodskii and M. B. Usvitskii, Tr. Inst. Stekla, "Steklo," No.3, p. 72 (1965).
6. O. K. Botvinkin and L. M. Mironova, Tr. Inst. Stekla, "Steklo," No.6, p. 17 (1959).
DISCUSSION ON SECTION II

F iIi P 0 vic h commented that in three-component systems the compositions of


glasses formed during phase separation can be determined from tie lines in pseudo binary sec-
tions of the immiscibility region only if they coincide with (or are very close to) the true tie
lines. He went on to say that it is important in principle to determine, with the aid of suitable
experiments on rapid and slow cooling of borosilicate glasses, whether the heterogeneous struc-
ture of these glasses with heterogeneity regions of about 10 'A in size is inherent in the glass it-
self or whether it appears when the glass passes through the immiscibility region on cooling.
Galakhov drew attention to the importance of finding experimental criteria for
determination of true equilibrium compositions corresponding to metastable liquid-liquid
separation on the phase diagrams. The true positions of the tie lines can be found in this way.

135
III

PHASE SEPARATION IN
OTHER MULTICOMPONENT GLASSES
EPR STUDY OF STRUCTURAL CHANGES
DURING MICRO PHASE SEPARATION IN
TITANIUM-CONTAINING GLASSES

N. M. Pavlushkin, R. Ya. Khodakovskaya, L. A. Orlova,


and V. V. Orlov

It is now commonly believed that processes preceding crystallization of titanium-contain-


ing glasses yielding glass-ceramics are of the liquid -liquid phase separation type.
We attempted to follow the course of structural changes accompanying phase separation
in glasses of the system A120 3 -MgO-Si02 -Ti02 , reflected in changes of the nearest environ-
ment of cations present in the glass, and primarily of titanium.
The investigation was concerned with two main variables having the greatest influence on
microphase separation in glass [1, 21: the Ti02 content and heat treatment during the precrys-
tallization period. The Ti02 content of the glass was varied from 0 to 20 wt.% (in steps of 3-5%).
The heat treatment was carried out in the 650-800 temperature range with exposure times from
2 to 8 h.

The main methods of investigation were EPR and electron microscopy. The EPR spectra
were recorded with the RE-1301 radio spectrometer.
Since the glasses studied do not contain natural paramagnetic centers, they were subjected
to gamma irradiation at the temperature of liquid nitrogen. The absolute amount of paramag-
netic centers formed was determined by comparison of the integral line intensities of the spec-
imen and the standard with the line form factor taken into account. The EPR spectra gave infor-
mation on the coordination state of titanium in the glass. This possibility is based on the follow-
ing principles.
1. Gamma irradiation of titanium-containing glass results in radiochemical reduction of
titanium to the trivalent state [31.
2. The EPR signal of Ti3+ can be recorded at room temperature only if the nearest en-
vironment of Ti 3+ consists of a deformed octahedron of six oxygen ions [4-61.
3. Only ionization and excitation processes occur in glass subjected to small doses of
gamma radiation at 77K. Therefore information obtained from EPR spectra on the coordina-
tion state of Ti 3+ after irradiation can be extended to the coordination state of quadrivalent ti-
tanium present in the glass before irradiation.

139
140 N. M. PAVLUSHKIN ET AL.

Fig. 1. Electron micrographs of glasses. a) Original glasses with different Ti02


contents (5, 10, 15, 20%); b) glasses after heat treatment in the pre crystallization
period (at 750).

It was shown earlier [2] that in the specified range of Ti02 concentrations the glasses can
be subdivided into three groups, differing sharply in the character of their crystallization:
coarsely crystallizing glasses, incapable of forming glass-ceramics (up to 10% Ti0 2 ) (I); glasses
yielding microcrystalline opaque glass-ceramics on crystallization (12-15% Ti02) (II); glasses
which are converted into transparent crystalline glass materials at an early crystallization
stage (18-20% Ti02) (III). These differences in crystallization character correspond to differ-
ences in their microheterogeneous structure (Fig. 1); with up to 15% Ti02 the droplike micro-
heterogeneity regions in the glass are isolated and distributed randomly in the matrix, and they
become appreciably more numerous at a Ti02 content of 12-15%. In presence of 20% Ti02 the
character of phase separation changes radically, the glass structure as a whole becomes micro-
heterogeneous and individual droplets are virtually impossible to detect in the micrograph. In
this case it is preferable to speak of a "quasi-homogeneous" structure in the glass.
Heat treatment in the precrystallization period produces sharp changes in the degree and
character of microphase separation in the glasses; the direction of these changes is similar to
that observed on increase of the Ti02 content (Fig. 1, a and b). The character of glass crystal-
lization undergoes corresponding changes: after preliminary heat treatment the glass with 10%
Ti02 can be converted into a glass-ceramic, and the glass with 15% Yields a transparent crys-
talline glass material. It must be emphasized that with the treatment times used these changes
of microphase structure in all the glasses occur only at one temperature (750 in the present in-
stance) (Fig. 2); after other heat treatments its character remains the same as before and only
the number and size of the drop like regions alter.
N. M. PAVLUSHKIN ET AL. 141

Fig. 2. Electron micrographs of glass containing 15% Ti02 after heat treat-
ment at 700,750,775, and 800 (treatment time 2 h).

Two absorption lines appear in the EPR spectra of the glasses after irradiation. The first
has a g factor of 1.94 and its width varies from 73 to 93 Oe; it has been identified [7, 8] as cor-
responding to Ti 3+ ions in sixfold coordination. The second line, having geff = 2.01 and tlH =
20-33 Oe, is the result of capture centers of two types: [SiO,d and [AIOs ] groups [3].
The intensity of these lines depends very strongly on the Ti02 content of the glass (Fig. 3,
curves 1 and 2). In this respect the entire range of Ti02 concentrations studied can be subdi-
vided into three regions: 1) up to 10%, with an almost constant amount of reduced Ti3+ ions; 2)
12-15%, where the amount of these ions increases sharply; 3) above 15%, where the content of
radiochemically reduced Ti3+ decreases just as sharply (becoming lower than in the first region).
It also follows from Fig. 3, curve 3 that on increase of the Ti02 content to 12% the number
of radiation defects on [Si04 ] and [AlO s] groups decreases, after which a certain degree of stabil-
ization is attained. The width of this line decreases somewhat, from 33 to 20 Oe, with increase
of the Ti02 concentration in the glass.
Preliminary heat treatment of the glass in the precrystallization period at the optimal tem-
perature * leads to changes in the integral intensity of the EPR spectra, the character of which
varies as a function of the Ti02 content of the glass.
Comparison of curves 1 and 4 in Fig. 3 shows that the form of the concentration depen-
dence of the number of Ti 3 + ions detected is not altered by precrystallization heat treatment,
but the whole curve shifts toward lower Ti0 2 concentrations.
For a more detailed study of the changes occurring in glass during precrystallization
phase separation, two compositions were chosen, containing 10% and 15% Ti0 2 , which lie on the
boundaries of the above-mentioned regions and which differ sharply in the degree and character
of liquid-liquid phase separation and subsequent crystallization (Figs. 1 and 2). It follows from
the EPR spectra (Fig. 4) the influence of precrystallization heat treatment on these glasses is
extremely strong, and depends on the temperature and time conditions.
The number of Ti3+ ions detected in the first of these glasses (Fig. 4a) passes through a
maximum and then falls to values lower than in the original state.
In the glass with 15% Ti02 (Fig. 4b) the number of Ti3+ ions recorded diminishes, and the
temperature curve has a minimum at 750. The width of the Ti3+ line is constant within the lim-
*The "optimal" temperature in this case refers to a temperature near tf. heat treatment at
which leads to the greatest changes in the microphase structure of the glasses and in the
character of their subsequent crystallization (Fig. 2).
142 N. M. PA VLUSHKIN ET AL.

Fig. 3. Dependence of the num-


2.0
ber of paramagnetic centers on
capture centers (npc) in y-ir-
radiated glasses on the TiO:! con-
tent (from EPR data). 1) Num-
ber of paramagnetic Ti3 + ions;
:I 1.0 9
'" 2) number of paramagnetic Ti3 +
o 4 E-< ions per g TiO:!; 3) number of
.-.
~ capture cente:rs on [Si04], [A106];
Z 0.5 0.3 g,
0.2 : 4) number of paramagnetic Ti 3 +
0.1 ~ ions after heat treatment of the
o
.-. glasses in the precrystallization
o 5 to 15 20 Z period (750).
TiOg %

its of experimental error. The intensity of the total line having geff = 2.01 also diminishes after
heat treatment, but its value depends little on the temperature of the precrystallization period.
Comparison of the results of EPR and electron-microscopic investigations (Figs. 3 and 1)
shows that the observed complex character of variations of the number of paramagnetic centers
is not accidental but is determined by the structural changes occurring in the glass under the
influence of Ti02

o 2 4 8 8
Time of heat treatment, h Fig. 4. fufluence of precrystal-
lization heat treatment on the
b number of paramagnetic centers
in glasses with different TiO:!
contents. a) 10% TiO:!; b) 15%
!:l 1.00
TiO:!. 1) Number of paramag-
'"
~
netic Ti3 + ions; 2) number of
~
u 0.75 capture centers on [Si04]'
c..
d
C>
[A106]
1,
.-. 0.50

650 700 750 800 to


N. M. PAVLUSHKIN ET AL. 143

In fact, concentration region I corresponds to glasses incapable of yielding glass-ceramics,


while regions II and III relate to glasses which can be converted into glass-ceramics, but they
differ substantially in the structures produced by liquid -liquid separation and subsequent crys-
tallization: region II corresponds to glasses yielding opaque glass-ceramics, in which liquid-
liquid separation produces separate droplets, and region III to glasses having "quasi-homoge-
neous" structure which crystallize to form transparent glass-ceramics.
The disproportionately large increase of the number of Ti 3+ ions recorded and the equally
sharp decrease of the number of radiation defects on [Si04 ] and [AIO s ] groups occurring when a
certain Ti02 concentration (12%) is reached indicate conclusively that the glass structure dnes
not vary monotonically with increase of the Ti0 2 content and that the structures of glasses which
form glass-ceramics differ radically.
These structural changes are also reflected in the dependence of properties of the glasses
(Young's modulus, density, etc. [9]) on concentration: at a Ti0 2 concentration ensuring liquid-
liquid separation and subsequent formation of glass-ceramics characteristic inflections (pla-
teaus) or minima appear on the curves.
Variations of the glass structures with the Ti0 2 contents are also clearly manifested in the
influence of precrystallization heat treatment on the number of Ti3+ ions detected in the glasses
(Fig. 3).
The relationship between the number of Ti 3 + ions recorded by the EPR method and the
microheterogeneous structure of the glasses becomes apparent on comparison of all the data pre-
sented above, namely: the increase of the number of Ti3+ ions recorded is associated to phase
separation in the form of iso lated drops, while the decrease of the number of paramagnetic cen-
ters is due to a change in the character of the phase separation, with formation of a "quasi-
homogeneous" structure (Figs. 1, 2 and 3, 4).
In addition to providing indirect information on the structural characteristics of titanium-
containing glasses, EPR data make it possible to draw certain conclusions regarding changes in
the nearest environment of the cations during phase separation in the glass. The observed num-
ber of paramagnetic Ti 3+ ions detected by the EPR method can be correlated with two factors: 1)
changes in the number of radiochemically reduced titanium ions in sixfold coordination in the
glass; 2) changes in the symmetry of the sixfold coordination sphere of titanium [S].
As the width of the Ti3+ absorption line varies little, regardless of the percentage Ti02
content and the conditions of heat treatment, it may be concluded that the role of the first factor
is predominant, i. e., that a relative change in the number of [TiOs1 groups occurs, although this
does not exclude qualitative changes in the nearest environment of these complexes.
It can be concluded from comparison of EPR and electron-microscopic data that processes
preparing the glass structure for finely dispersed volume separation (dropler separation) are
accompanied by increase of the amount of titanium in sixfold coordination. Processes preparing
the glass structure for possible formation of transparent glass-ceramics ("quasi-homogeneous"
structure) consist mainly of decrease of the number of [TiO s ] groups. It is interesting to note
that the structural transformations accompanying the change in the character of phase separa-
tion occur mainly in the Ti0 2-rich phase, since the number of paramagnetic centers on [Si04 ]
and [AlOs] groups remains constant and independent of the Ti02 content (Fig. 3).
Literature Cited
1. 1. 1. Kitaigorodskii and R. Ya. Khodakovskaya, in: The Glassy State, Vol. 3, No.1, Cat-
alyzed Crystallization of Glass, Izd. Akad. Nauk SSSR, Moscow-Leningrad (1963), p. 31
[English translation: The Structure of Glass, Vol. 3, Consultants Bureau, New York
(1964), p. 271.
144 N. M. PAVLUSHKIN ET AL.

2. I. I. Kitaigorodskii, R. Ya. Khodakovskaya, and M. D. Beus, Tr. Mosk. Khim.-Tekhnol.


Inst. 1m. D. I. Mendeleeva, No. 50, p. 10 (1966).
3. G. O. Karapetyan, Yu. N. Kondrat'ev, and D. M. Yudin, Fiz. Tverd. Tela, 6:1554 (1964).
4. J. Siegel, Phys. Rev., 134(1A):A193 (1964).
5. H. J. Gerritsen and H. R. Lewis, Phys. Rev., 119:1010 (1960).
6. J. C. W. Chien and C. R. Boss, J. Am. Chern. Soc., 83:3767 (1961).
7. N. R. Yafaev and Yu. V. Yablokov, Fiz. Tverd. Tela, 4:1529 (1962).
8. N. S. Garif'yanov, M. I. Rubtsov, and Yu. M. Ryzhmanov, Steklo i Keram., No.3, p. 11
(1962).
9. R. Ya. Khodakovskaya and M. N. Pavlushkin, in: Thermal and Mechanical Properties and
Structure of Inorganic Glasses, Abstracts of Symposium Papers, Moscow (1967), p. 66.

DISCUSSION
With reference to the paper by Pavlushkin, Khodakovskaya, Orlova, and Orlov, Gal a-
khov and Podushko expressed interest in the sensitivity of the EPR method in assess-
ment of structural changes due to liquid-liquid separation and in quantitative determina-
tion of titanium in different valence states. 0 r 1 ov a replied that the reported results
are in good agreement with the results obtained with the aid of electron microscopy, but
the existence of microphase separation cannot be inferred from EPR data alone, because
a combination of properties must be studied before conclusions with regard to structure
can be drawn. Absolute determination of the content of titanium dioxide in fourfold coordination
is impossible; only comparative assessments are meaningful.
INVESTIGATION OF PHASE-SEPARATION PROCESSES IN
TITANIUM-CONTAINING GLASSES IN THE
SYSTEM Si02 -Ah03 -CaO-MgO

B. G. Varshal, N. M. Vaisfel'd, G. B. Knyazher,


and L. M. Yusim

The purpose of this work was to study the influence of titanium dioxide on the heteroge-
neous structure and certain properties of glass in the system Si02-Al20 3 -CaO-MgO, and also
the role of alkali oxides (Na20 in this instance).
The original titanium-free glass in the system Si02-Al20 3 -CaO-MgO was used for prep-
aration of three series of glasses in each of which the Ti02 content was varied from 0 to 30%
in steps of 2%, and in some cases of 0.5% (percentages by weight). The titanium dioxide was
introduced in addition to the contents of the main components. The three series differed in the
Na20 contents, which were 0.25,1.0, and 5.0%.

Fig. 1. Variations of the struc-


ture of glasses under the influ-
ence of heat treatment in relation
b
to the Ti02 content, with Na20
contents of 0.25% (a), 1% (b), and
5% (c) (from data on gradient
crystallization). 1) Transparent
glass; 2) surface crystallization;
3) opalescence; 4) darkening; 5)
750
coarse volume crystallization; 6)
fine-grained volume crystalliza-
o2 6 tion.

_0
10 It, 18 22 26 30
TiOz,%
Of ~2 If:I 13 lIlTIJ~ ~5

145
~
>P-
O'>

'{, nOz
8.5

9.0

10.0 IJ:j
-1:.'1 F) P:"~ __~ ~1 ~~ ~J ~~,:"/:1 rA~ ~_~ _~ ~: ~.~"1
0

:;
~
12.0 l".:\ : ~,;;&'.~' ~;'I:~ :..;;~~ 'S~"j ~Zl"~ t~;i.~.,:r:-~ I....~:'~~~-~.~.!i . .'-~ A-~"""""""" tr ~~ '!II 'h":. Y..:.~:r,.. en
::t:
>
t"'
tr:l
I-j

13.0 "~ ~'P-!';;'~ E.-'!.:g'..~ ~-viOr~ M.fi~lt'~ ~jptAl.1 ~~ ~fiIIt az:Iii8AGiAii.' ~~ ,~


>
r

15.0

25.0

Fig. 2. Electron micrographs of the structure of glasses with different TiOz contents at various stages of heat treatment (1 %Na20).
B. G. VARSHAL ET AL. 147

The glasses were made from technical materials. In the case of glasses with 0.25% Na20
the Na20 was present as an impurity in the original batch components, while in the other series
(with 1 and 5% Na20) Na20 was introduced in the form of soda ash.
The glasses were melted in I-liter vitreous silica pots in a gas-fired crucible furnace at
1500-1550.
Density was determined by the hydrostatic weighing method, refractive index by the im-
merison method, and Young's modulus by th.? static method with the aid of the IZS-7 spherome-
ter, with disks 32 mm in diameter and 1. 5 mm thick. The microstructure of the glasses was
investigated with the "Tesla" electron microscope at initial magnification of x 2500-3000, fol-
lowed by photographic enlargement to xl0,000, by the platinum -carbon replica method. The
optical characteristics of the glasses were recorded with the SF-4 quartz spectrophotometer in
the visible region, and with the SF-4 instrument and an FE U-39 device in the ultraviolet region.
The results were calculated for a thickness of 0.1 mm with a correction for reflected light. The
thickness was not taken into account in calculations of the scattering spectral characteristics.
Comparison of data on gradient heat treatment with the results of electron-microscopic
investigations (Figs. 1 and 2) shows that the glass can be subdivided sharply into three groups
in accordance with the Ti02 content (Table 1).
Variations of the physical properties of glasses in the series with 1 % Na20 with the Ti02
content after heat treatment at 750 are shown in Fig. 3. There is a clear correlation between
the inflections on the curves and the subdivision of the glasses into groups in accordance with
their structure, given in Table 1: composition region with proportional variation of properties;
region of glass compositions showing anomalous variations; region of glass compositions in which
the variation of properties again becomes proportional.
Variations of the position of the absorption edge with the temperature of heat treatment
(Fig. 4) are of special interest: glasses of the first group do not show a shift, while in the sec-
ond and third groups the edge is shifted into the shortwave and longwave regions of the spectrum,
respectively.
The spectral characteristics of light scattering in specimens treated at 800 and 850 for
2 h are shown in Fig. 5. First, it must be noted that homogeneous glasses of the first group do
not exhibit light scattering even after prolonged heat treatment (curves 0-4). Scattering is ob-
served in specimens containing over 8% Ti02 The spectral characteristics of scattering differ
substantially for glasses of the second and third groups. The spectral characteristics of glass-
es of the second group (curves 8 and 10) correspond to Rayleigh scattering, with D proportional
to A-'. Glasses of the third group (curves 12, 13, and 15) give scattering curves of quite unusual

TABLE 1. Boundaries of Groups of Glasses with


Different Structures
'Ti0 2 contents, 0/0
Glass group characteristics
0.25'10 Na,oil% Na,O 5% Na 2 0

Homogeneous glasses,
surface crystallization 0-3 0-7.5 0-11
Heterogeneous glasses with appreciable
opalescence, volume crystallization,
coarse structure 4-8 8.0-12 12-15
Heterogeneous glasses, darkening at
first, volume crystallization, fine-
grained structure 9-20 13-25 16-30
148 B. G. VARSHAL ET AL.

5 350
I'l
5

~ ff.O~
a 4
~~ 9.0 ~
~ \.../'
..... 7.0
3
4.2
"'5
~ 4.0
~

3.8 2
Fig. 3. Dependence of the proper-
2.68
ties of glasses (with 1% Na20) on
"'5
the Ti~ content after heat treat- ~2.64
ment at 750 for 2 h. 1) Refrac-
"l:J~ 2.60
tive index; 2) density of glass; 3)
molar refraction of oxygen; 4) 2.56
Young's modulus; 5) position of
absorption edge. N~I
1.63
f.58

1. 540 4 8 12 16 20 24
TiOz,%

form, apparently as the result of a sharp increase of absorption by glasses of this group after
heat treatment. While in the case of glasses of the second group the error of absorption is dis-
regarded if less than 10%, which can be estimated from nonscattering specimens, in the third
group absorption makes the main contribution to the spectral characteristics. Scattering can be
assessed at A = 700 nm, where absorption is almost unchanged. It should be noted that the spec-
tral curves for all glasses of this group follow an absolutely identical course up to a treatment
temperature of 800 for 6 h; the spectral coefficient of the glass containing 15% Ti02 is altered
only by heat treatment at 850.

15
f3
12

Fig. 4. Dependence of lI.so%on the temper- f2a 0-4


ature of heat treatment of glasses with
8
different Ti~ contents. The Ti~ con- 9
tents of the glasses are indicated by the -10 fO
numbers on the corresponding curves.
-15

700 720 750 770


B. G. VARSHAL ET AL. 149

f),mm -1 15

1.5

f), mm- 1

1.5 1.01------

1.0

Fig. 5. Spectral dependence of light


scattering of the glasses. Symbols
0.5 as in Fig. 4.

500 700 800 900 1000


A,nm

Let us consider a three-component system with an internal immiscibility region (Fig. 6).
According to Schreinemakers [1], binodal boundary lines, spinodal curves, and tie lines
can be drawn in any isothermal section of the immiscibility region. Examining a certain pseudo-
binary section in such a ternary system, we can obtain several variants of passing through the
immiscibility region, and hence the corresponding changes of glass properties.
Case 1: line I-II does not touch the immiscibility region; the whole section corresponds
to homogeneous glasses only.
Case 2: line I-III passes only through the binodal region; this section has two regions of
homogeneous glasses and one of heterogeneous glasses.
Case 3: line I-IV cuts the binodal and spinodal regions; the region of heterogeneous
glasses has a binodal and spinodal section, the structural differences in which have been con-
sidered by Cahn and Charles [2, 3],

Fig. 6. Isothermal section of a ter-


nary phase diagram with an immis-
cibility region. acb, adb) Binodals;
aeb, afb) spinodals; tie lines are con-
tinuous, and variants of pseudo binary
sections are represented by dashed
lines.
A III W IV c
150 B. G. VARSHAL ET AL.

Case 4: line V-VI crosses the immiscibility region through the critical point, and from
the region of homogeneou6 glasses we pass directly into the spinodal region.
The possible pseudobinary sections in the ternary system are probably confined to these
four variants.
Examination of the structures of the glasses studied both before and after heat treatment
shows that introduction of titanium dioxide into the original glass, e. g., in the series with 1%
Na20, first leads to formation of homogeneous glasses, and when the Ti02 content exceeds 8%
heterogeneous glasses are obtained. There are two distinct groups of heterogeneous glasses,
with 8-12% and with 13-25% Ti02.
Since in our investigations we did not reenter a region of homogeneous compositions, it
must be assumed that the pseudo binary section studied corresponds to the third variant discussed
above, the difference being that not the whole immiscibility region but only the region of homo-
geneous glasses, the binodal region, and part of the spinodal region were crossed. It follows that
our three groups of glasses can probably be correlated with the corresponding regions of the
pseudobinary section on the phase diagram, namely: first group (0-8% Ti0 2), homogeneous glasses;
second group (8-12% Ti02), heterogeneous glasses of the binodal region; third group (12-25%
Ti02) , heterogeneous glasses of the spinodal region. The structure of a cooled melt will depend
significantly on the position of its composition on the phase diagram and on the cooling rate.
The properties of our glasses can now be examined in the light of relationships character-
istic of glasses in these regions of the phase diagram.
Region I is the group of glasses containing from 0 to 7.5% Ti02 These glasses have homo-
geneous structure. The properties of glasses in this group depend directly on the titanium diox-
ide content, with proportional variations of density, refractive index, Young's modulus, and fun-
damental absorption edge. As was to be expected, heat treatment at temperatures up to 800 does0

not lead to appreciable changes of glass structure in this region.


Region II is the group of glasses containing from 8 to 13% Ti02 Phase separation by a
binodal mechanism occurs in the region. Glasses of this region must have homogeneous struc-
ture after forming (which is essentially quenching), while heat treatment should produce a het-
erogeneous structure. Binodal separation may lead to changes in the macroscopic properties
of the glasses. Experimental data show that the transition from surface to volume crystalliza-
tion is abrupt, occurring as the result of a small change of the Ti0 2 content in the glass.
Phase separation in glasses of this group at a given temperature is a function of time:
prolonged treatments at low temperatures lead to appearance and intensification of opalescence.
In this group of glasses phase separation occurs in the 750-850 region and introduces a number
0

of peculiarities into variations of the physicochemical properties, dependent on the composition


and size of the drops formed.
Region III is the group of glasses containing from 13 to 25% Ti02 The structure of glasses
in this group is determined by the cooling rate and by the coefficient of diffusion. As the vis-
cosity of high-titanium glasses in the fusion zone is low, phase separation already occurs during
forming, when the cooling rate is considerably below the permissible limit [2, 3J. Spinodal phase
separation leads to a more uniform heterogeneous structure, with periodicity in the positions of
the particles of the secondary phase. This explains the very uniform microdisperse structure
of the crystallized glasses, in distinction from glasses in the binodal region. The observed pro-
portional variations of density, refractive index, and Young's modulus can also be attributed to
greater uniformity of the glass structure, since these properties depend on the phase ratio in
the quasi-homogeneous structure.
B. G. VARSHAL ET AL. 151

It should be noted that in glasses the compositions of which lie on the boundary between the
immiscibility regions heterogeneous structures characteristic either of the binodal or the spi-
nodal region may be formed, dependent on the cooling conditions or slight variations of the con-
tents of the main components.
As an example, we consider a glass containing 10% Ti02 and 0.25% Na20 (this composition
lies in the region of spinodal separation) cooled in a large mass (about 5 tons) at the highest pos-
sible rate from 1550 to 900 and then taken slowly through the immiscibility region. The glass
0

was present mainly in the binodal region, above the spinodal, which led to formation of drops of
the second phase (Fig. 6a). Additional low-temperature treatment led to secondary spinodal
separation [4], characteristic for the given composition. Heat treatment at 1050 (Fig. 7) led to
0

fine-grained crystallization of the matrix without affecting the large drops or the diffusion re-
gions surrounding them.
An example of the influence of slight variations of composition and melting conditions is
provided by glass 12a, in which the contents of the main components were varied somewhat (with-
in the technological tolerance limits), with the exception of Ti02, the content of which remained
12%; this led to a shift of the glass structure from the spinodal into the binodal region of phase
separation. A shift of glass structure from one region into the other is accompanied by changes
of properties characteristic for the given region (Figs. 4 and 5).
We examine the results of optical measurements in the light of the foregoing concepts of
the heterogeneous structure of the glasses studied. As was to be expected, heat treatment of
homogeneous glass does not lead to changes of absorption. When we pass into the region of
binodal phase separation drops of a second phase, enriched with Ti02 separate out; their refrac-
tive index is higher than that of the matrix, which has a reduced content of this component. It is
to be expected that scattering is determined by reflection at the drop boundaries, and the optical
path passes mainly through the matrix. This accounts for the shift of the absorption edge into
shortwave region (clearing) for glasses of the second group. Treatment of the glasses at higher
temperatures (850) leads to sharp increase of light scattering.
The slope of the curves in Fig. 5 is determined by the power of the degree of scattering
= 4) the particles of the separated phase are smaller
'A.- m; in the case of Rayleigh scattering (m
than the wavelength of light. Decrease of the slope (m) * means that the particles causing scat-
tering increase in size (curve 12a in Fig. 5).
The structure of glasses of the third group consists of two continuous interpenetrating
matrices. In this case the light path may be presumed to pass mainly through the absorbing high-
titanium phase. This leads to sharp increase of absorption, darkening of the glass, and unusual
spectral characteristics, which in this case are determined mainly by absorption and to a lesser
degree by scattering. The identity of the course of the curves and even of the spectral scattering
coefficient m indicate identity of structure in the spinodal region, and deviations from these laws
arise only at 850 and a high Ti02 content (15%). Probably scattering by crystallites of the tita-
0

nium-rich phase occurs in glasses treated at high temperatures, and in accordance with literature
data [5, 6] the change of the slope of curve 15 is due to increase of the size of these crystallites.
It must be specially noted that glass 12a, having a slightly different composition and con-
taining 12% Ti02 , exhibits all the optical properties characteristic of the binodQ.l region (shift of
the absorption edge, spectral dependence of scattering), despite the fact that all the characteris-
tics of glass of the usual composition and containing 12% Ti02 place it in the spinodal region (Fig.
4). It may be supposed that these changes of properties are more characteristic of the corre-
spondingheterogeneous structure than of a definite Ti0 2 content in the glass.
*At a heat-treatment temperature of 800 (2 h), m for glasses 12, 13, and 15 is 4.9, 4.8, and 5.0;
0

at 850 (2 h), m for glasses 8, 10, 12, 12a, 13, and 15 is 4.2, 4.2, 5.1,2.0, 5.4, and 4.9.
0
152 B. G. VARSHAL ET AL.

Fig. 7. Structure of glass containing 10%


Ti0 2 a) Cooled slowly during forming;
b-e) after heat treatment: b) 750, 2 h;
c) 800, 2 h; d) 900, 2 h; e) 1050, 3 h.

The role of Na20 in these glasses must be considered. Experimental data show clearly
that increase of the Na20 content shifts all phase-separation processes toward higher Ti02
concentrations (Table 1) and slows down the development of these processes. This effect must
apparently be considered from the standpoint of stabilization of titanium ions in fourfold coor-
dination by sodium oxide [7], which must have a significant influence on phase separation. In
any event, it follows from the results of the present investigation that the structure of titanium-
containing alkali-free glasses can be regulated so as to produce the required combination of
properties by addition of small amounts of Na20.
Literature Cited
1. F. A. H. Schreinemakers, Z. Phys. Chern., 22:93 (1897).
2. J. W. Cahn, J. Chern. Phys., 42:93 (1965).
3. J. W. Cahn and E. J. Charles, Phys. Chern. Glasses, 6:181 (1965).
4. E. A. Porai-Koshits, this volume, p. 28.
5. R. D. Maurer, J. Appl. Phys., 33:2139 (1962).
6. B. G. Varshal, Elektron. Tekhn., Ser. "Tekhnol. i Organizatsiya Proizv.," No.9, p. 13
(1966).
7. A. A. Appen, in: Structure of Glass, Izd. Akad. Nauk SSSR, Moscow-Leningrad (1955),
p. 96 [English translation: The Structure of Glass, Vol. 1, Consultants Bureau, New York
(1958), p. 75].
B. G. VARSHAL ET AL. 153

DISCUSSION
Varshal, replying to questions by Mazurin, Filipovich, and others regarding the re-
lation between spinodal decomposition and subsequent formation of glass-ceramics, said
that liquid-liquid phase separation has a dual effect: it increases the tendency to crystal-
lization and creates an interface. No comparison was made with the phase diagram in the pre-
sent instance.
INFLUENCE OF TITANIUM DIOXIDE ON
PHASE SEPARATION IN LITHIUM
ALUMINOSILICATE GLASSES

N. E. Kind and E. M. Milyukov

Titanium dioxide is widely used as a catalyst in production of glass-ceramics from a num-


ber of aluminosilicate systems. Most investigators consider that the influence of Ti02 is mani-
fested primarily in separation of the melt into two glassy phases, and that the metastable phase
separation which occurs ensures fine-grained volume crystallization. However, the structural
changes occurring during the period preceding volume crystallization have not been studied or
fully explained.
Studies of titanium-containing lithium aluminosilicate glasses show that crystalline glass
materials can be obtained only in the composition region where the molecular ratio A1 20 3 : Li20 >
1. This shows that an excess of alumina relative to lithium oxide in the glass composition plays
a very important part in formation of the crystalline glass structure. It must also be pointed out
that the catalytic action of added Ti02 , leading to formation of a microheterogeneous structure
and to volume crystallization, begins to be manifested in lithium alumino silicate glasses only af-
ter introduction of Ti02 in amounts above a certain minimum.
It was of interest to investigate the relationship between phase separation and the amount
of Ti02 added to a series of lithium aluminosilicate glasses with various excess alumina contents.

Fig.!. Positions of the glass composi-


tions studied in the triangular diagram.

154
N. E. KIND AND E. M. MIL YUKOV 155

2.400
o 2 8

Fig. 2. Variations of the densities of glasses


with various AI 20 3/Li20 ratios with the titanium
dioxide content.

We started with the hypothesis that changes of the glass structure as the result of separation into
two glassy phases should be expressed in anomalous variations of a number of properties of the
glasses, including anomalous variations of density. Density determination was therefore used as
the main method in the present investigation. Densities of the glasses were measured by weigh-
ing them in piece or powder form in toluene.
The investigations were carried out on a series of glasses (Fig. 1, 1-5) having a constant
Li 20 content (18 mole%) and variable Si02 and Al20 3 contents, with additions of 0-8 mole% Ti02
on 100% of the original glass. The compositions of these glasses lie on a line cutting the trans-
versal corresponding to an equimolecular ratio of Al 20 3 to Li20 (Fig. 1, line 1:1).
The glasses were made in 3-liter vitreous silica pots in a flame furnace, cast on a cold
plate, and annealed in a muffle. Double melting was used in some cases. The original glasses
were transparent, either colorless or colored yellow or brown.
156 N. E. KIND AND E. M. MILYUKOV

Fig. 3. Structure of glass No.5. Titanium dioxide content:


a) 4 mole%; b) 6 mole%.

Study of the densities of these glasses showed that in most cases the curve for the density
as a function of the Ti02 content has an inflection. As the glass specimens were free from crys-
talline inclusions, the observed inflections most probably represent structural changes due to
microphase separation. Figure 2 shows the dependence of the densities of the original glasses
(Fig. 2, 1-5) on the titanium dioxide content. The titanium dioxide content corresponding to the
inflection differs for glasses of different compositions and depends on the amount of alumina
present in excess relative to Li 20 in the glass. For glass compositions to the right of the line
1:1 the inflection is accompanied by anomalously high increase of density. The glass of the com-
position lying on the line 1:1 does not give an inflection. The curve for the glass of the composi-
tion to the left of the line 1:1 has an inflection, but in this case the rate of density increase is
higher in presence of small than of large amounts of added titanium dioxide. Electron micro-
graphs of glasses Nos. 3-5 showed distinct structural heterogeneity in specimens with high Ti0 2
contents (above the inflection), indicating phase separation. The size of the heterogeneous re-
gions was 250-300 A (Fig. 3). The structure of the same glasses with low Ti02 contents was
homogeneous.
The graphical relationship between the molecular concentration of titanium dioxide corre-
sponding to the inflection on the density curves for glasses Nos. 2-5 and the excess alumina con-
centration is represented by a straight line passing through the origin (Fig. 4, 1-5).

10

6
t;t.
" 6
'0
6

Fig. 4. Relationship between


g: If
<
the titanium dioxide content 5u 2
corresponding to the inflection M
"-I

on the density curve and the


2 J If 5
amount of excess alumina in \
TIO z mole%
\
the glasses studied. -2 \
\ f
-If
N. E. KIND AND E. M. MIL YUKOV 157

The ratio of excess alumina to titanium dioxide was found to be 2:1. Glass No.1, the com-
position of which lies to the left of the line 1:1, does not conform to the linear relationship; this
is fully justified.
The observed experimental facts can be explained as follows. In the modern view, intro-
duction of alumina in excess relative to Li 20 into the silicate glass must lead to formation of
[AIO s] groups. The formation of glasses over a relatively wide region, extending to compositions
with high contents of excess alumina, indicates that these groups can be distributed in their
structure. The uniform increase of density of the original titanium-free glasses with increase
of their alumina content can be attributed to statistical distribution of the [AIOs ] groups in the
glass structure. Further increase of the alumina content above a certain limit leads to sharp in-
crease of the crystallization tendency of the glass. Castings from these glasses after annealing
are found to be fully crystallized with formation of coarse-grained aggregates.
Examination of the states of castings of the corresponding titanium-containing glasses shows
that the higher the concentration of excess [AIO s] groups in the original glass the greater are the
amounts of Ti0 2 which can enter stably into its structure without causing effects indicating struc-
tural transformations (crystallization, opalescence, cracking). The same conclusion follows from
an examination of the density curves for these glasses.
At high temperatures the coordination number of the Ti' + ion may be four, and the [TiO,]
ion may be four, and the [TiO,] group then becomes compatible with the structural network com-
posed of [SiO,] tetrahedra. At the same time, at low temperatures six is the stable coordination
number of the Ti'+ ion. During cooling the melt apparently becomes supersaturated with [AIO s ]
and [TiO s ] groups in sixfold coordination, and this produces favorable conditions for formation
of a second glassy phase, enriched with these elements.
Thus, there must be a direct relation between the change of the coordination state of Ti02
and phase separation in the glasses. This process requires a definite ratio of excess alumina
to titanium dioxide. Our study of densities of a series of titanium-containing lithium aluminosil-
icate glasses confirmed that separation into two phases may already occur in them during cool-
ing in melts.

DISCUSSION
Kin d, replying to questions by Podushko and Ermolenko, confirmed that the glasses
used in these studies had not been subjected to any preliminary heat treatment. The ob-
served inflections on the density curves are evidence only of structural transformation,
i.e., phase separation, as crystallization was not observed. Electron micrographs confirm this
particular interpretation.
With reference to the second paper by the same authors, Bal' skaya and Filipovich
asked them to define more preCisely the immiscibility region in the system studied and
the nature of the phase separation. M i1 y uk 0 v confirmed that phase separation was
observed in all the compositions studied but, according to electron micrographs, it cannot be
assigned to the stable type. In reply to a question by Podushko, the speaker confirmed absence
of crystallization.
PHASE-SEPARATION PHENOMENA IN GLASSES OF
ALUMINOSILICA TE SYSTEMS CONTAINING
VARIOUS MODIFIER CATIONS

E. M. Milyukov and N. E. Kind

This communication is a brief account of the results of a study of phase separation in alu-
minosilicate glasses containing oxides of Li, Na, Ca, Ba, Cd, Sn, and Zn. The need for such an in-
vestigation became obvious in studies of formation of glass-ceramics in these systems. A total
of about 200 glass compositions were studied in the course of the investigation by electron micros-
copy and other methods both with and without addition of Ti02 as a catalyst of volume crystal-
lization.
The investigation showed that all the titanium-containing systems studied include composi-
tion regions where the first stage in the phase transformations in the transition from a melt to
a glass is metastable phase separation leading to subsequent volume crystallization in the form

Fig. 1. Positions of the compositions studied in the


system ZnO- Al203-Si~. 1) Transparent glass; 2)
opalescent glass; 3) sinter.
158
E. M. MILYUKOV AND N. E. KIND 159

of fine crystals. In some instances, e.g., in glasses containing Na and Sr, the heterogeneous
structure disappeared during subsequent heat treatment, the drops were resorbed and the
glasses either crystallized from the surface or melted without crystallizing.
The most interesting studies were obtained in studies of zinc aluminosilicate glasses. In
the composition region studied (Fig. 1) the surface (dome) of metastable immiscibility descended
smoothly on progressive addition of alumina to glasses of the binary zinc silicate system. This
can be deduced qualitatively from a visual examination of the specimens and from electron micro-
graphs. Glasses close to the region of stable immiscibility and containing 5 mole% A1 20 3 ex-
hibited intense opalescence even in rapidly quenched drops of the melt. This fact indicates
that when the melt is cooled phase separation occurs at relatively high temperatures and low
melt viscosities.
The consequence of further additions of A1 20 3 , which lower both the liquidus temperature
and the immiscibility!?urface, is that when the melt is cooled and passes through the tempera-
ture region corresponding to metastable separation its viscosity is high enough to prevent sepa-
ration. Therefore glasses containing 10-13 mole% A1 20 a in thick castings exhibit dense opales-
cence which intensifies toward the center, while when cast in drop form they are transparent
and do not opalesce. Glasses having even higher A1 20 3 contents give castings which are trans-
parent throughout even after annealing. Nevertheless, despite their apparent homogeneity and
absence of opalescence, examination of these glasses under the electron microscope showed that
they have heterogeneous structure with drops 500-2000 A in size. Therefore the boundary in
Fig. 1 between the regions of opalescent and clear glasses is arbitrary, as there is no difference
in principle between the structure of these two groups of glasses. Heat treatment of glasses
to the right of this boundary intensifies phase separation, which is manifested in some increase
of drop size and in appearance of opalescence. Additions of titanium also intensified phase sep-
aration sharply in all cases. This is due, first, to the tendency of titanium itself to form a sec-
ond phase in silicate melts and, second, apparently to the fact that TiOz lowers the melt vis-
cosity, which facilitates separation.
The results of studies of phase separation in other aluminosilicate systems indicate that
all the titanium-containing systems studied and some systems without titanium have extensive
regions of metastable immiscibility.

fO

7.5
5

o
Fig. 2. Dependence of the
average drop diameter on
the temperature of heat
treatment and on the Ti02
content. Glass composition:
SrO 20, A1 20 3 5, and Si02
75 mole%. The TiOz con-
tents (mole%) are indicated
on the curves.
160 E. M. MILYUKOV AND N. E. KIND

Fig. 3. Microstructure of strontium aluminosilicate (a-c) and calcium aluminosilicate


(d-f) glasses at various stages of heat treatment. a) Original; b) 800; c) 1000; d) 750;
e) 800; f) 850. Glass compositions (mole%): SrO 20, Al20 3 5, Si02 75; CaO 20, Al 20 3 5,
Si02 75.

Figure 2 shows the dependence of drop size on temperature and Ti02 content in
glasses of the strontium aluminosilicate system (SrO 20, Al 20 3 5, and Si0 2 75 mole%).
It can be seen that with increase of the treatment temperature the drop size increases steadily
from a few hundred 'A in the original glasses to 2500-4000 'A in glasses treated at 1000. On
further heating, resorption of the drops occurs and the glass becomes homogeneous and melts.
Additions of titanium intensify phase separation, but the form of the curves remains unchanged.
The structure of this glass is shown in Fig. 3, a-c.
The structure of one of the calcium aluminosilicate glasses at various stages of heat treat-
ment is shown in Fig. 3, d-f. The drop size increases with temperature and at a certain tem-
perature coalescence into larger regions consisting of two or three drops occurs.
Figure 4a shows the structure of the original lithium aluminosilicate glass close to spod-
um.e ne in composition, with addition of 10 mole% Ti02 The photograph clearly shows rounded
drops up to 40,000 'A in size formed as the result of separation, and acicular crystals, identified
by x-ray diffraction as p-eucryptite, around their peripheries.
Figure 4b is a micrograph of a glass of complex composition containing barium; this also
clearly shows coexisting drops formed by phase separation and crystal growing as the result of
their coalescence.
E. M. MILYUKOV AND N. E. KIND 161

Fig. 4. Microstructure of glasses. a) Original lithium alumino-


silicate glass close to spodumene in composition, with addition of
10 mole% Ti02; b) barium-containing glass of complex composition.

This study shows that all the titanium-containing aluminosilicate systems investigated have
regions of metastable phase separation. The phase changes occurring in such glasses under the
influence of heat treatments at progressively increasing temperatures may proceed by the schemes
shown below.

1. separation .... melt

homogeneous glass }
2. Separation -> homogeneous { .... melt
glass.... surface crystallization

3. Separation Isepa:ation
surface
) .... 1:~~:~allizationl"" melt
" (starts at surface)
crysta 11lzatlOn

separation )
4. Separation 1VOlu:'e ....
volume
crystallization melt
crystalliz ation

This diversity of possible phase transitions from a liquid -liquid separation structure in
the glass to the melt, and especially the possibility of a transition in accordance with scheme 2,
where the liquid-liquid structure disappears before crystallization begins, confirms the view
that liquid -liquid phase separation, like crystallization, is an independent phase transition. In
most of the aluminosilicate systems studied by us the liquid -liquid structure persists to the
start of crystallization and the two partially overlap.
MICRO PHASE SEPARATION IN GLASS CONTAINING
CERTAIN CRYSTALLIZATION INITIATORS

S. T. Suleimenov, M. She Sharafiev, T. A. Abduvaliev,


T. D. Nurbekov, and I. I. Sorokina

Formation of micro heterogeneity regions is characteristic of most natural and artificial


silica melts [1-5].
It is noted in the literature [6-8] that glasses for conversion into glass-ceramics should be
chosen in regions of immiscibility in the given systems. The explanation is that microhetero-
geneity regions in glass are in most cases enriched with structural elements of future crystals.
Our investigation was concerned with microphase separation in glasses made from
tephrite basalt.
The original tephrite basalt glass had the following chemical composition (wt. %): Si02 49.5;
Al 2 0 a 16.78; Fe203 9.30; Ca011.28; MgO 6.77; R20 4.83; Ti02 1.54.
Investigations showed that glass-ceramic materials can be obtained from the pure tephrite
basalt glass by one-step heat treatment at 800 for 3 h [9]. Electron-microscopic studies showed
that crystallization of this glass is associated with microphase separation. It follows from Fig.
la that a freshly fractured surface of the original glass is homogeneous and smooth, whereas
after heat treatment (600, 3 h) drop like heterogeneity regions arise (Fig. Ib). Heat treatment
of this glass at 700-750 for 3 h leads to formation of a crystalline structure (Fig. lc). confirmed
by x-ray phase analysis. This behavior of the glasses can be explained as follows.
During heat treatment in the 685-700 range a differentiation process begins in the glass,
\\lith formation of regions enriched with iron ions. These regions appear to be close in chemical
composition to magnetite, FeO Fe203. On further increase of temperature, magnetite begins to
separate out from these regions.
It was noted by Galakhov [3] that microphase separation occurs not only during cooling of
the melt but can also develop at temperatures above the liquidus.
We used the isothermal heating method for studying crystallization of a tephrite basalt
melt. The quenched specimens were investigated by optical and electron-microscopic methods.
The original glass has refractive index ND = 1.584. This glass is converted completely into a
homogeneous melt in 1 hat 1300. However, individual pieces of glass with the higher refrac-
tive index ND = 1.598 can be detected by the immersion method in quenched specimens. The re-
fractive index of the main mass of the glass is ND = 1.~81. On further cooling to 1280 with a
holding time of 60 min magnetite crystallized from the glass having the higher refractive index
(Fig. 2a). Decrease of temperature to 1200 leads to crystallization of plagioclase from the
162
s. T. SULEIMENOV ET AL. 163

Fig. 1. Structure of pure tephrite


basalt glass. a) Original glass; b)
after heat treatment at 600, 3 h;
c) after heat treatment at 750, 3 h.

Fig. 2. Sepa ration of crystalline


phases from tephrite basalt melt.
a) Magnetite; b) plagioclase; c)
augite.
164 S. T. SULEIMENOV ET AL.

FiF:. 3. Structure of titanium-con-


taining glasses. Ti0 2 contents of
glasses: a) 3%; b) 7%; c) 15%.

main mass of the melt (Fig. 2b), while at 1150 monoclinic pyroxene (augite) also crystallizes
out (Fig. 2c) .
These data show that the original melt has heterogeneous structure and contains regions
close to the future crystalline phases, including magnetite.
These regions undergo microphase separation during heat treatment (600 , 3 h) while in
the 700-750 range magnetite is formed from them.
Investigations of tephrite basalt glasses with additions of 0.5 to 15 wt.%of titanium dioxide
showed that separation of the liquid phase already occurs in the original titanium-containing
glasses. It follows from Fig. 3 that in glasses containing 3-7% titanium dioxide and not subjected
to heat treatment droplike heterogeneity regions 0.2-0.3J.l in size are present (Fig. 3a); in glass
containing 7% Ti02 these drops are considerably larger, up to IJ.l (Fig. 3b) , than in glass with
3% TiC>:!. The heterogeneity regions are even more pronounced in glasses containing 10-15%
titanium dioxide (Fig. 3c).
Thermograms of titanium-containing tephrite basalt glasses have two endotherms (Fig. 4).
This appears to be associated with phase separation; the first endotherm, at 720-725 , corre:-
sponds to the softening range of the matrix while the second, at 800-820, the intensity of which
depends on the TiO:! content in the glass, is due to softening of the titanium-rich second phase
of the glass. These regions of phase separation in the entire volume of the glass give rise to
an extensive interface, and nucleation is fa~ilitated considerably.
The tendency of glasses containing titanium dioxide to separate into two phases can be at-
tributed to a change of the coordination number of titanium from six to four on increase of tem-
perature and to "freezing" of the high-temperature fourfold coordination on cooling of the melt.
S. T. SULEIMENOV ET AL. 165

O ~~~L-~~--~~~L--I

600 650 700 750 800 850 900 9sa 1000


Fig. 4. Thermograms of titanium-containing
glasses.

In nephrite basalt glasses containing up to 2 wt.% of phosphorus pentoxide droplike hetero-


geneity regions were found even before heat treatment (Fig. 5, a and b). Investigations with the
aid of an optical microscope showed that original glasses of high P205 contents (6-8 wt.%) con-
tain glass regions having different refractive indices, from 1.578 to 1.627. This effect may be a
consequence of macrophase separation; 1.578 is close to the refractive index of pure tephrite
basalt glass. Glasses having higher refractive indices are evidently enriched with phosphoric
anhydride. Study of the glass with the lowest refractive index under the electron microscope re-
veals a microheterogeneous structure (Fig. 5c).

Fig. 5. Structure of the original phos-


phorus containing glasses. a) 1% P 20 5;
b) 2% P 20 5; c) 8% P 20 5
166 S. T. SULEIMENOV ET AL.

It is known [10] that phosphorus pentoxide is a characteristic glass-forming oxide; the


phosphorus ion is present in glass in fourfold coordination. Conditions for phase separation arise
owing to the difference between the structure-forming quadrivalent silicon and quinquevalent
phosphorus ions. Apparently, a necessary condition for occurrence of microphase separation in
a tephrite basalt melt is a low P 20 5 content (up to 2-4 wt.%), so that ring anions of the (P 40d4-
type and chain anions can be formed [11]. On increase of the P 20 5 content above 4 wt.% ribbons
of (P04)3- tetrahedra are formed and macrophase separation occurs.
It was found that tephrite basalt glasses differ in their tendency to crystallization in accor-
dance with the phosphoric anhydride content; this is associated with phase-separation processes.
For example, glasses with low P 20 5 contents (up to 1 wt.%) undergo complete volume crystalliza-
tion at 800 (3 h) while in absence of phosphorus pentoxide this effect is achieved at 950 (3 h).
The same crystalline phases separate out: magnetite, monoclinic pyroxene of the diopside-augite
series, and basic plagioclase.
The structure of heat-treated glasses containing 2 wt.% P 20 5 is shown in Fig. 6. On heat
treatment (800, 3 h) of the original glass containing microheterogeneity regions (Fig. 5b) the
drops become more numerous and ill honeycomb structure is produced (Fig. 6a). No crystalline
phases separate out as yet. When the temperature of heat treatment is raised to 900 (3 h) indi-
vidual droplets begin to lose their regular outlines, apparently owing to separation of a primary
crystalline phase.
The number and size of the droplets also increase in heat-treated glasses (650, 3 h) con-
taining 8 wt.% P 20 5 (Fig. 6c). In specimens heated at 800 heatherlike formations appear (Fig. 6d);

Fig. 6. Structure of heat-treated phosphorus-containing glasses. a) 2% P 20 5, 800, 3 h; b)


the same, 900, 3 h; c) 8% P 20 5, 650, 3 h; d) the same, 800, 3 h.
S. T. SULEIMENOV ET AL. 167

x-ray phase studies showed that this represents separation of apatite. The size of individual
apatite crystals reaches 5Jl.. On increase of the treatment temperature to 900 small amounts of
0

monoclinic pyroxene and magnetite appear, detected by x-ray phase analysis. Therefore crystal-
lization of glasses containing 8 wt.% P 20 5 begins at lower temperatures than that of glasses with-
out additions or containing 2-4 wt.%of phosphorus pentoxide.
Thus, microphase separation in the original tephrite basalt glass is intensified in presence
of small amount of P 20 5 (0.5-1 wt.%), and this improves the crystallizability of these glasses.
Although increase of the P 20 5 content from 2 to 4% intensifies phase separation, it has an
adverse effect on crystallizability. This is apparently due to change of the coordination of the
aluminum cation from sixfold to fourfold, which favors crystallization of gehlenite, Ca2AI[AISi071,
which separates out at a higher temperature than magnetite. On further increase of the P 20 5 con-
tent to 6-8% [P04]3- groups are formed and macrophase separation occurs. Heat treatment of these
glasses causes crystallization of apatite from the phosphate glass. This phase separates out at
a lower temperature than the other crystalline phases (gehlenite, monoclinic pyroxene), and this
improves the crystallizability of these tephrite basalt glasses.
Literature Cited
1. D. P. Grigor'ev, Zap. Vseros. Mineralog. Obshchestva, Ser. 2, Part 64, No.1, p. 250
(1935).
2. V. V. Lapin, Tr. Inst. Geol. Nauk, Issue 106, Petrograf. Ser., No. 30, p. 28 (1949).
3. F. Ya. Galakhov, in: Structural Transformations in Glasses at High Temperatures, Izd.
Nauka, Moscow-Leningrad (1965), p. 110.
4. I. I. Kitaigorodskii, N. M. Pavlushkin, Yu. 1. Kolesov, Z. V. Zhitkevich, and S. V. Petrov,
in: Vitreous Systems and Materials, Izd. Zinatne, Riga (1967), p. 217.
5. Yu. D. Kruchinin, T. V. Bakhireva, L. P. Kruchinina, and T. A. Ust'yantseva, in: Boron-
Free, Alkali-Free, and Low-Alkali Glassy Systems and New Glasses Made from Them,
TsNIITEstrom, Moscow (1967), p. 240.
6. F. Ya. Galakhov, in: The Glassy State, Proceedings of the Fourth All-Union Conference,
Izd. Nauka, Moscow- Leningrad (1965), p. 113 [English translation: The structure of Glass,
Vol. 1, Consultants Bureau, New York (1966), p. 121].
7. V. N. Filipovich, in: The Glassy State, Proceedings of the Fourth All-Union Conference,
Izd. Nauka, Moscow- Leningrad (1965), p. 38 [English translation: The Structure of Glass,
Vol. 1, Consultants Bureau, New York (1966), p. 32].
8. W. Vogel, in: The Glassy State, Proceedings of the Fourth All-Union Conference, Izd.
Nauka, Moscow- Leningrad (1965), p. 108 [English translation: The Structure of Glass,
Vol. 1, Consultants Bureau, New York (1966), p. 114].
9. N. M. Pavlushkin, S. Y. Suleimenov, T. A. Abduvaliev, M. Sh. Sharafiev, and T. D. Nur-
bekov, in: Chemistry and Chemical Technology, Vol. 5, Alma-Ata (1966), p. 234.
10. J. Partridge and P. W. McMillan, Glass Technol., 4:173 (1963).
11. O. A. Esin and P. V. Gel'd, Physical Chemistry of Pyrometallurgical Processes, Vol. 2,
Izd. Metallurgiya, Moscow (1966).
INFLUENCE OF PHASE SEPARATION ON PHYSICAL AND
CHEMICAL PROPERTIES OF HEAT-TREATED FIBERS

M. S. Aslanova and Z. I. Shaina

In distinction from "massive" glass, the strength of thin glass fibers decreases after low-
temperature treatment in the 100-500 range.
It was shown earlier [1-3] for fibers made from lead and cadmium glasses with limited
silica contents, and also for iron-containing silicate fibers that the strength decrease after treat-
ment in the 100-500 range is due to microphase separation preceding crystallization.
In addition to the strength, it was of interest to study in more detail how other phYSical and
chemical properties of glass fibers alter during heat treatment. Lithium aluminosilicate glass
fibers were prepared for this purpose, and their properties were compared after heat treatment
with those of fibers made from alkali-free aluminoborosilicate glass.
The physical properties (low-temperature viscosity, density, refractive index, strength)
of the fibers were studied both in the original state and after heat treatment.
It was found that viscosity increases after heat treatment of the fibers for 5-10 h at 400;
the curve shifts by about 40-50. The viscosity increases further on increase of the treatment
temperature.
The observed increase of fiber viscosity after heat treatment is due to microphase separa-
tion, leading to formation of a high-melting phase.
It must be pointed out that this is accompanied by an increase of the sintering temperature
by 200.
The density and refractive index of the fibers are lower than those of "massive" glass,
and they increase only after heat treatment. At the same time, it was found that the density and
refractive index of fibers made from lithium aluminosilicate glasses used in the original state
for production of glass-ceramics are closer to the values for "massive" glass and depend less
on the fiber diameter than those of fibers made from alkali-free glass.
Heat treatment of fibers made from lithium aluminosilicate glasses leads to sharp in-
creases of their density and refractive index (Table 1).
The increase of denSity and refractive index after heating and cooling of the fibers, regard-
less of the glass composition and forming conditions, indicate that increase of the density of the
fiber structure after heat treatment is due to increase of the degree of microheterogeneity of the
glasses from which they are made.

168
M. S. ASLANOVA AND Z. I. SHAINA 169

TABLE 1
Density,g/cm 3 Refractive index
Glass Form of specimen
original 16000, 2 h original 16000, 2 h
-

Lithium alumino-
f Glass. 2.437 2.451 - 1.539
silicate A I Fiber, lOll : 2.427 2.451 1.531 1.535
Lithi urn alumino- -
.i Glass. 2.537 2.554 1.558
silicate B I Fiber, lOll . 2.496 2.514 1.545 1.549
A luminoboro- -
f Glass. 2.546 2.558 1.552
silicate I Fiber, lOll . . 2.518 2.558 1.540 1.549

Determinations of fiber strength after heat treatment showed that the strength begins to
fall sharply at 400 (Fig. 1).
The influence of prolonged heat treatment for 120 h at lower temperatures (250-300) on
the strength of glass fibers was studied. Experimental data show that the fiber strength remains
almost unchanged under these conditions, but the variability of the strength values, especially
low ones, diminishes. This indicates that redistribution of imperfections at the interface between
the microphases occurs even at low temperatures.
Processes occurring during heating were investigated with the aid of differential thermal
analysis both of the lithium aluminosilicate glasses and of fibers made from them.
Examination of the thermograms shows that the scale factor influences not only the tem-
perature at which the new phases separate out but also the number of such phases formed during
heat treatment.
It should be noted that the character of the processes occurring when the fibers are heated
depends to a considerable degree also on temperature at which the fibers were drawn (Fig. 2).
The DTA curve of fibers drawn above ts have three low-temperature effects: at 370, 435, and
495.
The character of the structural changes detected with the aid of differential thermal anal-
ysis was investigated with the aid of x-ray structural analysis and electron microscopy.

20
18
18
';> 14
~
""'I'~ 12
""I""
-I",,'" 10
Fig. 1. Influence of the temperature of heat "
r-;, 8
treatment on the strength of fibers of d = 10M.
8
1) Original fiber; 2-4) heat treatment for 2 h
at 400 (2), 500 (3), and 600 (4). 4
Z

50 100 150 300


P, kg/mm 2
170 M. S. ASLANOVA AND Z. I. SHAINA

880 0

Fig. 2. Thermograms of fibers.


2 1) Drawn below the upper crys-
tallization limit; 2) drawn above
the upper crystallization limit.

Crystallization is not detected by x-ray diffraction in lithium aluminosilicate fibers heat-


treated at 400-500. The diffractograms of fibers heat-treated above 600 show distinct crystal-
lization (Fig. 3). Electron micrographs of fibers after heat treatment at 400 show presence of
droplike regions; these become more numerous with increase of temperature.
Interesting data were obtained in determinations of the chemical stability of the fibers in
water. The dependence of the degree of leaching of the fibers on boiling for 3 h on the tempera-
ture of heat treatment was determined (Fig. 4). It follows from the figure that phase separation
occurs in the fibers during heat treatment in the 300-400 range; the droplets enriched with al-
kali oxides formed in the process are dissolved on boiling, and the fibers have lower chemical
stability .
The chemical stability increases with the temperature of heat treatment of the fibers.

3.40

-Angle e, deg

Fig. 3. Diffractograms of glass (1) and fiber (2) after the same heat treatment. The numbers
at the peaks indicate the interplanar spacings (in A).
M. S. ASLANOVA AND Z. I. SHAINA 171

Fig. 4. Dependence of the chem-


ical stability of fibers in water on
the temperature of heat treatment
for 2 h.

200 400 800 800 to

Thus, the observed changes of the physical, chemical, and mechanical properties of glass
fibers after low-temperature heat treatment are due to precrystallization microphase separation,
characteristic of lithium aluminosilicate, borosilicate, and certain other types of fibers, and to
subsequent crystallization. All these processes occur at lower temperatures in finer fibers.
Literature Cited
1. M. S. Aslanova, Steklo i Keram., No. 11, 10 (1960).
2. M. S. Aslanova and S. Z. Vol'skaya, in: The Glassy State, Proceedings of the Fourth
All-Union Conference, Izd. Nauka, Moscow- Leningrad (1965), p. 426 [English translation:
The Structure of Glass, Vol. 6, Consultants Bureau, New York (1966), p. 227].
3. A. A. Myasnikov and M. S. Aslanova, Steklo i Keram., No.5, p. 15 (1964).

DISCUSSION
With reference to the paper by Aslanova and Shaina, Mazurin asked whether the
results of low-temperature heat treatment of glasses might be influenced by the heat-
ing involved in viscosity measurements, performed at considerably higher temperatures.
S h a ina replied that the time of heat treatment was 5-10 h whereas the viscosity measurement
took only 1 min. Replying to Filipovich, she agreed that the extensive surface of glass
fibers tends to intensify crystallization.
PHOTOCHROMISM AND MICROPHASE
SEPARATION OF GLASS

V. A. Tsekhomskii and I. V. Tunimanova

Photochromism is the ability of a material to alter its optical density reversibly under the
influence of activating light [1]. It must be pointed out at once that two classes of photochromic
glasses are known: glasses sensitized [2, 3] by ions of variable valence (Eu, Ce, W, Mn, etc.)
and glasses the properties of which are due to the presence of silver halide (AgHal) crystals.
Other photosensitive crystals, such as ZnS, Ti02 , etc., apparently may also produce this effect.
This paper is not concerned with the first class, and deals only with photochromic glasses
containing silver halide crystals.
The known glasses sensitized by silver halide crystals are heterogeneous sodium alumi-
noborosilicate glasses containing from 0.2 to 0.8%silver [4].
An interesting point is that, despite the fairly large number of published patents [5], the
compositions of the photochromic glasses given in them lie in very narrow ranges. This alone
suggested to us that phase separation plays an important part in the phenomenon of photochro-
mism.
Examination of electron micrographs of photo chromic glasses (Fig. 1) showed that these
glasses are two-phase systems. The drops vary in size in accordance with the temperature and
time of heat treatment. Even in completely transparent glass these drops are up to 300 A in
size. It would be quite wrong to suppose that these spherical heterogeneities, up to 2000 A in
size and occupying 20-30% of the area of the figure, are silver halide crystals, because the
silver content of the glass is only 0.42%. It was accordingly supposed that during heat treatment
the glass separates into two phases (it should be pointed out that in order to produce photochro-
mic properties the original glasses must be subjected to special heat treatment, known as "strik-
ing"). The drop phase consists of sodium borate glass. This was confirmed experimentally.
If the alkali content of the basic "closed" phase decreases as the result of phase separation,
the ionic conductivity should decrease; this was observed in practice (Fig. 2).
In the light of the foregoing a hypothesis was put forward on the mechanism whereby photo-
chromic properties are produced in glasses or, more correctly, the mechanism of growth of
silver halide crystals in them.
It should be noted that the sodium alumino borosilicate glass chosen as the basis for photo-
chromic glasses, free from silver and halides, does not undergo phase separation in the tempera-
ture range used for striking.
Phase separation is induced by introduction of silver into these glasses and dissociation of

172
V. A. TSEKHOMSKII AND I. V. TUNIMANOVA 173

Fig. 1. Electron micrographs. a) Transparent


photochromic glass, treated for 2 h at 620; b)
opalescent glass, treated at 650 for 2 h.

8 Fig. 2. Influence of strik-


ing on the temperature de-
pendence of resistance of
the glass. 1) Original glass;
2) the same glass after heat
treatment for 2 h at 600.
f5 f7 19 21
I/T fO~
174 v. A. TSEKHOMSKII AND I. V. TUNIMANOVA

silver oxide with formation of colloidal metallic silver during heat treatment in the 500-550
range.
We showed that phase separation does not occur in glasses melted under drastic oxidizing
conditio'1s, and these glasses cannot be "struck," and therefore do not acquire photo chromic
properties, as dissociation of silver oxide does not occur in them. Only forced reduction as the
result of 'Y -ray or x-ray treatment of these glasses leads first to formation of metallic silver
crystals, then to phase separation, and finally to appearance of photo chromic properties during
heat treatment.
Thus, our main hypothesis is that metastable phase separation with formation of sodium
borate form in drop form, catalyzed by metallic silver, occurs in photochromic glasses in the
temperature range of "striking." This sodium borate phase, which has a much higher solvent
power than the aluminum - silicon -oxygen matrix, collects all the "impurities" in the glass
(silver, chlorine, bromine, copper, etc.). Just as in zone melting, where all the impurities are
collected in the liquid phase, when phase separation occurs in these glasses all the components
which could not enter the silica network are collected in the microphase. Such concentration of
silver, chlorine, copper, and other components in individud microregions evidently favor growth
of silver halide crystals.
Within certain limits, th3 size of the drops formed determines the size of the silver halide
crystals. We found that th8 rate of decolorization of photochromic glasses depends on the size
of the AgHal crystals: the smaller the crystals the higher the rate [6). It is well known that
alumina decreases the size of the microheterogeneity regions. We found that increase of the
alumina content in the glasses raises the decolorization rate.
If our hypotheSiS is correct (it will be shown below that it was confirmed by certain exper-
iments), we have very interesting potentialities for creation of new materials. utilizing the ef-
fect of phase separation in which the droplet phase has a much higher solvent power than the
matrix, which leads to considerable concentration and supersaturation in that phase, we will be
able to grow various crystals having magnetic, hminescent, and many other properties in the

Fig. 3. Electron micrograph of glass of the composition BaO 2%,


B20 3 98%.
V. A. TSEKHOMSKII AND 1. V. TUNIMANOVA 175

glass. As the crystal size is limited by the "reserves" of such microregions, the glass may re-
main transparent.
We chose glasses in the system BaO- B20 3 to verify our hypothesis.
It has been shown [7] that glasses containing from 0.7 to 16% BaO have a considerable ten-
dency to phase separation. Introduction of small amounts of Al 20 3 leads to formation of trans-
parent glasses.
An electron micrograph of glass consisting of 2 mole% BaO and 98 mole% B20 3 is shown
in Fig. 3. It undergoes extensive separation, forming enclosed drops. It is interesting to note
that separation is weakened considerably by introduction of Al 20 3 Introduction of 2 mole% Al20 3
into glasses containing 2 mole% BaO results in formation of transparent glasses. The micro-
phase separating out in such glasses is the barium borate component, which apparently has
higher solvent power than pure vitreous boric anhydride. On introduction of silver and halogens
into glasses of this system we succeed in growing silver halide crystals and obtaining photo-
chromic glasses.
Literature Cited
1. W. Marckwald, Z. Phys. Chern., 30:140 (1899).
2. A. J. Cohen and H. L. Smith, Science (USA), 137(3534):981 (1962).
3. U. S. Pat. No.3 ,255,026.
4. W. H. Armistead and S. D. Stookey, Science, 10(4):150 (1964).
5. V. A. Tsekhomskii, Optiko-Mekhan. Prom., No.7, p. 58 (1967).
6. V. V. Vargin, A. Ya. Kuznetsov, S. A. Stepanov, and V. A. Tsekhomskii, Optiko-Mekhan.
Prom., No.1, p. 43 (1968).
7. E. M. Levin and S. Block, J. Am. Ceram. Soc., 40:95 (1957).

DISCUSSION
Answering numerous questions provoked by the paper, T s e k hom ski i put forward
the view that photochromic glasses cannot be obtained from glasses which do not undergo phase
separation, as aggregation and growth of the silver halide crystals occur within the heterogeneity
regions having higher solvent power. Intense phase separation is also undesirable, as such
glasses become opaque. Irradiation leads to formation of F center, favoring appearance of me-
tallic silver.
DROPLET SEPARATION IN PHOTOSENSITIVE GLASS
A. I. Berezhnoi

It is noted in the literature that microphase separation in droplet form is possible not only
in the usual glasses for production of glass-ceramics containing Ti02 or other crystallization
catalysts, but also in photosensitive glasses containing gold, silver, or copper ions as catalysts
[1-4]. However, there are no experimental data in the literature confirming this suggestion.
According to Kleber [5], formation of crystallization nuclei is a kinetic process and
depends on two main factors: 1) the work of nucleation proper, which he defines as the energy
consumed for formation of a nucleus under the condition that all the structural elements con-
stituting the nucleus are already present at its site; 2) the activation energy which must be ex-
pended in order to supply the structural elements to the nucleation site. The actual work of
nucleation depends on the difference between the energy expended to overcome interfacial ten-
sion forces during formation of a new surface of the nucleus, and the energy liberated during
the glass-crystal transformation. If the structural elements are already present at the nuclea-
tion site, as in the case of uniform distribution of small amounts of Au, Ag, or Cu throughout
the glass volume or enrichment with these elements of the droplet microphases which, accord-
ing to Vogel's data [1, 2] are formed in photosensitive glass as the result of microphase sepa-
ration, the work of nucleation becomes very small.
Vogel et al. [1, 2] consider that crystallization of photosensitive glasses may involve
phase-separation phenomena analogous to those occurring in titanium-containing glasses, with
the only difference that in this case the droplet phase contains Cu+ , Ag+ , or Au3+ and Ce H
ions stead of Ti~; like Ti02 , these ions have a dual function: they accelerate phase separation
and, after ultraviolet irradiation of the photosensitive glass, form nuclei for epitaxial growth
of crystals of the main separating phase.
The ability of noble metals to cause phase separation in photosensitive glasses may involve
noted by Weyl [6], who considers that artificial formation of numerous ultramicroscopic par-
ticles of Au, Ag, Cu, Pt, or Ti02 in the glasses produces interfaces which intensify crystalliza-
tion of compounds the components of which predominate in the glass. Weyl's explanation of for-
mation of crystallization nuclei from noble metals in glasses is that in the metallic state the
electrons are utilized most effectively for screening the positively charged nuclei. During inter-
action with the glass melt containing tetrahedra differing in the degree of screening the minute
colloidal particles of Pt, Au, Ag, and Cu interact predominantly with the tetrahedra which con-
tain the largest number of polarized ions and which are the most effectively screened, i.e.,
tetrahedra containing the largest number of nonbridging 0 2- ions. Preferential adsorption of
these structural units by colloidal Pt or Au, Ag, and Cu particles results in formation of a phase
having higher basiCity and polarizability than the matrix phase. This separation of the system
into volume elements of higher and lower basicity intensifies formation of crystallization nuclei
and crystallization itself during phase separation.

176
A. I. BEREZHNOI 177

Fig. 1. Initial stages of formation of crystallization nuclei in photosensi-


tive lithium aluminosilicate glass subjected to different treatments. a) Ir-
radiation with UV light for 90 min, heating to 500, holding time 3 h; b) not
irradiated, heated to 570 and held for 5 h.

Thus, the demand for screening colloidal particles of the noble metal leads to preferential
absorption of the more polarizable and more basic structural units, and this separation induces
nucleation.
We have demonstrated experimentally for the first time droplet microphase separation in
a photosensitive lithium aluminosilicate glass which had been subjected to precrystallization
heat treatment in order to create the most favorable conditions for nucleation and to increase
the amount of crystalline phase. It was found that intermediate cooling of the photosensitive
glass from 500 to 250 and preliminary treatment in the low-temperature region, e. g., at 500
for 3 h, determine the "precrystalline state" of the glass and facilitate microphase separation
in droplet form. The initial stages of formation of droplet regions of separation in the photo-
sensitive glass subjected to low-temperature precrystallization heat treatment can be clearly
seen in the electron micrograph in Fig. 1. In accordance with the literature data cited above
[1, 2], it may be supposed that these microphase droplets contain Ag+ ions, which initially ac-
celerate phase separation, and after irradiation of the photosensitive glass with ultraviolet light
form crystallization nuclei on which epitaxial growth of the main nonmetallic separating phase
begins. However, another mechanism is more probable, where the main initial and independently
controllable process is formation, under the influence of ultraviolet radiation and heat treatment,
of crystallization nuclei from colloidal particles of metallic silver; when these nuclei reach the
critical size, droplike spherical regions of the separating vitreous matrix grow in them. The
separation process, which occurs in lithium alumino silicate glasses even in absence of silver,
is intensified considerably in presence of photographically reduced silver particles, and this
subsequently facilitates crystallization of the photosensitive glass considerably.

Fig. 2. Electron micrographs of specimens of photosensitive lithium aluminosilicate glass


irradiated for 10 min and heated at 560 for 120 (a), 240 (b), 375 (c), 480 (d), and 660 min (e).
178 A. I. BEREZHNOI

Fig. 3. Electron micrographs of specimens of photosensitive lithium aluminosilicate glass


irradiated for 0 (a), 10 (b), 60 (c), 100 (d), and 419 min (e) and heated at 570 for 300 min.
0

It is interesting to note that point droplet microseparation regions are formed both in
irradiated (Fig. la) and in nonirradiated (Fig. Ib) photosensitive glass subjected to heat treat-
ment. The temperature and duration of heat treatment are greater for the nonirradiated glass.
This suggests that ultraviolet irradiation of photosensitive glass favors microphase separation
during the precrystallization period in the course of heat treatment. This is in good agreement
with the phenomenon of photocatalysis or photonucleation, i.e., with the well-known activating in-
fluence of shortwave ultraviolet radiation, manifested in lowering of the temperature at which
color begins to appear ("striking") in irradiated photosensitive glass in comparison with nonir-
radiated glass, with formation of considerably more numerous uniformly distributed crystalliza-
tion nuclei.
Droplet regions of phase separation, formation of crystallization nuclei, and crystal growth
in photosensitive glass subjected to different heat treatments after the same irradiation and to
the same heat treatment after different exposure to radiation are shown in Figs. 2 and 3, respec-
tively.
Literature Cited
1. W. Vogel and K. Gerth, in: Symposium on Nucleation and Crystallization in Glasses and
Melts, Am. Ceram. Soc. Inc., Ohio (1962), p. 11.
2. W. Vogel, Glass Technol., 7:1 (1966).
3. A. 1. Berezhnoi, Pyrocerams and Photocerams, Izd. Mashinostroenie, Moscow (1966),
p.70.
4. A. 1. Berezhnoi and Yu. M. Po lukhin , Neorgan. Mater., 3:986 (1967).
5. W. Kleber, Silikattech., 13:5 (1962).
6. W. A. Weyl, Sprechsaal Keramik-Glas-Email, 93:128 (1960).

DISCUSSION
Replying to questions by Filipovich, Bobkova, and Yusim, Be r e z h n 0 i detailed
the mechanism of the observed processes. Irradiation results in formation of a certain
amount of reduced silver particles. Subsequent heat treatment intensifies aggregation
and colloidal particles are formed. Available literature data suggest that introduction of
photosensitive additives does not alter the initial glass structure, but on general considerations
presence of silver chloride should have an influence.
SPECTROSCOPIC INVESTIGATION OF
GLASSES IN THE SYSTEM Rp-La203-Si02

L. V. Labutina, N. M. Pavlushkin,
and M. V. Artamonova

Spectroscopic investigation of glasses containing rare-earth elements as active additives


("probes") is one method of studying structural changes in glasses, as their absorption and lu-
minescence spectra depend on interaction of the active additive with the surrounding ligands.
This interaction, in its turn, is determined not only by the nature of the additives but also by
their structural distribution.
The purpose of the present work was to investigate the effects of composition and heat
treatment on structural changes in glasses, manifested in changes of the absorption and lumi-
nescence spectra of neodymium ions present in small amounts in the glasses.
Our choice of the original composition was guided by the following considerations:
1) neodymium must replace isomorphously one of the components of the original glass;
2) the glass must not give characteristic absorption and luminescence bands in the region
of neodymium absorption and luminescence bands;
3) the glasses must not crystallize or undergo phase separation during the forming, as
this may cause light scattering.
On the basis of literature data [1-3] the following glass compositions were taken for study
of structural changes occurring in glasses during heat treatment: 1) Li20 1.53K20 . La203 .
8.2Si02; 2) Li 20' 1.53K20. La203 ~ 7.7 Si02 0.5Ti02 Neodymium oxide was introduced into these
glasses in amounts of 2.0 wt.% on 100% of the main glass composition.
The glasses were made in corundum crucibles in a SHit furnace at 1550. They were held
for 2 h at this temperature.
Heat treatment of the glasses was carried out at 675 for 3,12, 24, 48, 60, and 72 h. The
temperature of heat treatment was chosen in the light of DTA data.
The investigations of absorption and luminescence spectra and x-ray diffraction and elec-
tron-microscopic analyses were carried out by known methods.
Figures 1 and 2 show variations of the coefficient of absorption (K), calculated as K :;
DA/O .43 . a . c , of the luminescence intensity at A = 1.06 /-L, of the integral luminescence intensity
(I), and of the halfwidth of the luminescence band (Ll) at A :; 1.06 J1 with the duration of heat treat-
ment at 675.

179
180 L. V. LABUTINA, N. M. PAVLUSHKIN, AND M. V. ARTAMANOVA
o
.1,A K, cm- 1
400 11....,( 0.0

Time of heat treatment, h


Fig. 1. Dependence of the coefficient of absorption
at A = 0.589 J..I. (1) and of the halfwidth of the lumines-
cence band of Nd:H- ions at A = 1.06 J..I. (2) on the dura-
tion of heat treatment of glasses I and II at 293 K.

I. cm 2

Fig. 2. Dependence of the lumines-


cence intensity of Nd3+ ions on the
duration of heat treatment of glasses
I and ll. 1) Luminescence inten-
sity of Nd3+ ions at A;;; 1.06 J..I. and
293K; 2) integral luminescence in-
tensity of Nd:H- ions at A = 0.9, 1.06,
and 1.3 J..I. and 293K; 3) luminescence
intensity of Nd3+ ions at A = 1.06 J..I.
and 77K.

i2~--~----~----~--~
o 20 60 80
Time of heat treatment, h
L. V. LABUTINA, N. M. PAVLUSHKIN, AND M. V. ARTAMANOVA 181

A=I.0D fl

A=f.06 fl

3~=f.09fl
,
, I
--

A= 1.06 fl I
II
'\
2 I
I
\
I I
I I
I I
I I
Fig. 3. Luminescence spectra of glass I (1, I I
I
2) and of a crystal (3), containing Nd3+ ions at \
\
293K (dashed lines) and 77K (continuous \
\
lines). 1) Original glass I; 2) glass I after \
heat treatment for 24 h at 675C; 3) CaW04 I

"
I
\
crystal. \
I
\
\
\

Luminescence spectra of the original glass and of the glass after heat treatment for 24 h,
recorded at 293 and 77K, are given in Fig. 3 (1 and 2). These figures show that variations of
the coefficients of absorption, luminescence intensity, and halfwidth of the luminescence band
of Nd 3+ ions depend on the time of treatment at 675; at the start of heat treatment the proper-
ties either remain almost unchanged (K) or show some decrease (I at 293K), increasing only
after prolonged heat treatment, whereas the luminescence intensity of Nd 3+ ions at 77K in-
creases continuously.
It follows from the variations of the spectroscopic properties of glasses with heat treat-
ment that the luminescent properties of glasses containing Nd3+ ions are the most sensitive to
structural changes occurring in the glasses during heat treatment. It should also be noted that
the variations of K;\=O.589 fl' of the luminescence intensity of Nd 3 + ions at 293 and 77K, and of
the halfwidth of the luminescence band of Nd 3+ are more pronounced in the case of the glass con-
taining TiO:!.
These relationships are determined by changes of glass structure during heat treatment.
The structure of glass is determined to a considerable extent by the structure of the melt,
because it is in the melt that formation of submicroregions occurs; these regions are silicate
compounds in which one type of cations predominates, having close values of field strength [4,
5].
In melt of the glasses studied at least two phases differing in chemical composition are
formed, namely nLa203 mSi02 and n'La203 m'SiO:! (where n > m and n' < m') with alkali-metal
cations distributed at random in the compositions of both phases. As was shown earlier [6], the
equilibrium chemical composition of the phases is determined by temperature.
It may be supposed that the active addition of neodymium oxide is present in the phase
182 L. V. LABUTINA, N. M. PAVLUSHKIN, AND M. V. ARTAMANOVA

richer in lanthanum oxide, phase separation being of the liquid -liquid type. The following facts
confirm absence of structural order in these chemical formations:
a) the luminescence intensity and the structure of the luminescence bands of the original
glasses not subjected to heat treatment do not change significantly if the experimental tempera-
ture is lowered from 293 to 77K;
b) considerable changes in the properties of the glasses are observed only after prolonged
heat treatment ("'40 h), necessary for the OC0u:crence of processes producing structural order.
Diffuse absorption and luminescence bands and absence of changes of luminescence inten-
sity on decrease of temperature to 77K are characteristic of ordinary glasses (Fig. 3, 1); the
explanation is that the probability of nonradiative transfer of energy to the glass network is high
in these glasses. It is known that owing to the random structure of glass the energy barriers
between neighboring positions which may be occupied by modifier ions vary in height. At the
same time, the average height of the energy barriers is lower in glasses than in crystals of
equivalent composition. This results in higher mobility of the modifier ions in the glass and, in
consequence, produces a closer relationship between the energy state of the excited ion and ther-
mal vibrations of the glass network.
Interaction of Nd 3+ ions with the surrounding matrix and the extent of splitting of energy
levels depend both on the structure of the first coordination sphere and on the degree of struc-
tural order of the material as a whole. Considerable changes of intensity and structure of the
Nd 3 + ion luminescence bands on decrease of the experimental temperature are characteristic
for crystalline compounds (Fig. 3, 3).
Thus, the existence of long-range order tends to increase the luminescence intensity ow-
ing to decreased probability of nonradiative transfer of excitation energy to the crystal lattice.
Variations of the luminescence spectra of heat-treated glasses are of the same character as
those for crystalline media. This leads to the conclusion that as the result of prolonged heat
treatment (at 675) an ordered structure, which cannot be detected by x-ray diffraction or elec-
tron microscopy, is produced in the glass. Presence of titanium dioxide in the glass facilitates
ordering of the structure.
Literature Cited
1. N. M. Pavlushkin, M. V. Artamonova, and L. V. Labutina, Tr. Mosk. Khim.-Tekhnol. Inst.,
No. 40, p. 267 (1967).
2. E. I. Semenov, Mineralogy of the Rare Earths, Izd. Akad. Nauk SSSR, Moscow (1963).
3. I. A. Bondar', Silicates and Aluminates of the Rare Earths and Yttrium, doctoral disser-
tation, Inst. Khim. Silikatov, Leningrad (1965).
4. J. Krogh-Moe, Glass Ind., 47:308 (1966).
5. D. A. Goganov and E. A. Porai-Koshits, in: Structural Transformations in Glasses at
High Temperatures, Izd. Nauka, Moscow- Leningrad (1965), p. 100.
6. N. M. Pavlushkin, M. V. Artamonov, and O. L. Al'takh, Elektron. Tekhn., Ser. 14, No.5,
p. 83 (1967).
DISCUSSION
Replying to questions on the paper by Pavlushkin, Artamonova, and Labutina, Lab u-
tin a said that the luminescence method is structure-sensitive with respect to changes
of symmetry of luminescence centers and of their similarity in local regions, and can
therefore characterize relative changes of composition in these regions.
USE OF THE SPECTRAL LUMINESCENCE METHOD FOR
STUDYING PHASE SEPARATION IN SILICATE GLASSES

M. I. Kuz'menkov and M. B. Rzhevskii

One important problem in microphase separation in glasses is determination of the chem-


ical composition and structure of the droplet phases.
Since the spectral luminescence characteristics of glasses activated by rare-earth ele-
ments are highly sensitive to various structural changes in the medium containing the activator
ions, we studied decay of luminescence of glasses in the systems R 20-Si02 (where R is Li, Na
or K) and Si02- CaO- MgO-Na20- AIF 3, activated by terbium [1].
The glasses were made by melting batches made up from cp grade materials in open sin-
tered corundum crucibles in a Silit furnace at temperatures from 1400 to 1500. The activator
was Tb 20 3, introduced into the batches in amounts of 0.5 wt.%on 100%. The iron content was
kept to a minimum. When the melts were completely clear they were cast on a cold steel slab.
Decay of luminescence was investigated in an apparatus assembled on the basis of a UF-2
monochromator and operating on the double-disk phosphoroscope principle. Luminescence was
excited in the glasses by radiation of 365 J.1 wavelength from a DRSSh-250 mercury vapor lamp,
passed through a UFS-2 filter and copper sulfate solution. An FEU-27 photomultiplier was used
as the radiation detector. Emission from the specimen was recorded in the region of the lu-
minescence band with a maximum at436 J.1 for 5 .l(j3 sec from the end of excitation. The resolv-
ing power of the apparatus was about 10- 4 sec. Decay of luminescence was studied by points.
The decay curves of binary glasses are complex in character and in the range of 1.5 .
10- 4 - 5 .1(J'"3 sec are described by the sum of two exponents with different values of the duration
of the excited state T, which is evidence of heterogeneous structure of the glasses. The lumi-
nescence centers found are designated luminescence centers of the first (I) and second (II) types.
Recording of two emissions, differing in duration, is possible because type I centers have much

Fig. 1. Dependence of the spectral


0.30
luminescence characteristics of
~0.25
~"'0.20 glasses of the composition ~O . 3Si02
0.15 on the radius of the ion in the modify-
3
ing oxide. 1) Duration of the excited
~ 2.0 -02 state of luminescence centers of the
~ 1.5 -01 first type (Tt); 2) duration of the ex-
!- f. 0 La:::'----O:..::.....J..,--..L..L:------,-'--_--'::c_---L.IL.K_+o cited state of luminescence centers
0.8 1.0 1.1 1.2 1.3 R;A of the second type (T2) j 3) At re1
183
184 M. I. KUZ'MENKOV AND M. B. RZHEVSKII

0.35
0.30
4l
...... 0.25
<l 0.20
0.15 3
1.8
Fig. 2. Dependence of the spec- u 1.4 '-~--~'-----
-.------~2
<U
1.2
tral luminescence characteristics E. 1.0 .-~--~r~~------1
on the Li20 content in glasses of I-
the system Li20-Si02. Symbols 0.8
as in Fig. 1. 0.8
15 20 25 30

higher emission intensity than type II centers. It was thus possible to introduce the quantity
At reI , defined as the ratio of the duration of recording of emission from centers of type I to the
duration of recording of emission from centers of both types. From the variation of this quan-
tity with the glass composition or conditions of heat treatment it is possible to determine the
phase to which type I centers, recorded first, belong.
Thus, T of type I and II centers increase with increasing radius of the modifier ion (R+)
in the series Li20 - Na20 - K20 (Fig. 1). The duration T1 of the excited state of type I centers
(Fig. 1, 1) alters little, indicating that changes on the coordination of terbium ions are small.
The decrease of At reI (Fig. 1, 3) in the sequence from lithium to potassium glasses is due to de-
crease of the emission intensity of type I centers, which is the consequence of decrease of their
total amount and therefore of the total volume of the phase containing them.
It can be inferred from the foregoing that type I centers are associated with microhetero-
geneity regions, and the nature of variation of the dependence of T1 on R+ suggests that they are
enriched with the quartzlike component. Type II centers are associated with the glass matrix,
enriched with silicates; this follows from the considerable variation of T2 with the type of mod-
ifying oxide.
Variation of the Li20 content in glasses of the system Li 20- Si~ has different effects on
Tof the two types of luminescence centers (Fig. 2, 1,2). It has been shown [2] that in glasses of
this system the composition of the microphase droplets alters with decrease of the Li20 content.
For example, in glasses containing 7.5-17.5 mole% Li20 the droplets are enriched with disilicate,
and in glasses with higher Li20 concentrations, up to 33.3 mole% and over, with silica. The
droplet size increases with decrease of the Li20 content from 33.3 to 20 mole%, and then de-
creases sharply with further decrease of the Li20 content [3]. Moreover, it has been suggested
that in glasses of the system Li20-Si~ the volume ratio of the "structural elements" Li20'
2Si~ and Si~ becomes unity at a certain Li20 concentration [4].

The character of variation of Atrel (Fig. 2, 3) with decrease of the Li20 content from 33.3
to 14.3 mole% apparently represents variation of the total number of type I centers, which also
supports the view that they are associated with microphase droplets. It may be supposed that
the maximum on this curve at compositions in the region of 20 mole% corresponds to equaliza-
tion of the volumes of the two phases. On decrease of the Li20 content in the glass from 25 to
14.3 mole% T1 decreases while T2 increases, from which it can be concluded (Fig. 2,1,2) that
the microphase droplets are enriched with silicates in low-alkali glasses and with silica in high-
alkali glasses. This conclusion is in good agreement with literature data [2-4].
M. I. KUZ'MENKOV AND M. B. RZHEVSKII 185

Qj 0.5
::f 0.3
0.1 ~
r
.3
Fig. 3. Influence of the temperature
1.5 2
u of heat treatment on the spectral lu-
~
1.3
S
minescence characteristics of glass
1.1
.: 0.9
I
'1 of the composition Si02- CaO- MgO-
./" Na20- AIF 3 Symbols as in Fig. 1.
850 700 750 800 850 900 to

At the eutectic composition (33.3 mole% Li20) centers of types I and II were detected, which
is also evidence of a microheterogeneous structure in this glass, although this could not be de-
tected with the aid of electron microscopy [2, 3]. Admittedly, it follows from Fig. 2, 3 that the
total number of type I centers is small, indicating that the total volume of the microheterogeneity
regions is also small. It is possible that micro heterogeneity regions of fluctuation type [5] are
formed in this case. In any event, the observed fact is evidence of the high sensitivity of the
spectral luminescence method.
Phase transformations in glasses of the system Si02- CaO- MgO- Na20- AIF 3 , which ex-
hibit volume crystallization, were studied with the use of glass 240, which has been fully inves-
tigated by other physical methods [6]. Curves representing decay of luminescence in specimens
of this glass, heated for 2 h in the 650-900 temperature range, are of complex character and
are the result of superposition of two luminescences differing in duration: 71 and 72. The value
of 71 remains almost unchanged for specimens heat-treated at 650-725, and then increases
slightly when the temperature is raised to 750 (Fig. 3, 1). The dependence of 72 on the temper-
ature of heat treatment is similar, except that it increases more sharply at 735 (Fig. 3, 2).
To determine the phase to which type I centers should be assigned, we analyze the course
of variation of ~trel with the temperature of heat treatment. Since the volume of the glassy phase
decreases as crystallization of the glass proceeds, it can be concluded that type I centers are
associated with the glassy phase (matrix). This view is also supported by the slight variation
of 71 (Fig. 3, 1). Correspondingly, type II centers should be assigned to the microphase droplets,
the presence of which in the pre crystallization period has been detected with the aid of electron
microscopy [6]. Since 72 remains unchanged up to 725, this is evidence of structural similarity
of the droplet, although growth and coalescence of the droplets were detected in this temperature
range under the electron microscope.
In the 725-750 range crystallization begins, as the result of which the structure of type II
centers already corresponds to the structure of the primary crystalline phase at 750. This fol-
lows from the constancy of 7 for specimens heat-treated at 800-900, when the glass is com-
pletely crystallized, which is in good agreement with x-ray diffraction data indicating that the
nature of the crystalline phase, a solid solution of the diopside type, remains unchanged [6].
Literature Cited
1. M. A. EI'yashevich, Spectra of the Rare Earths, Gosizdat, Moscow (1963).
2. V.1. Aver'yanov and E. A. Porai-Koshits, in: Structural Transformations in Glasses at
Elevated Temperatures, Izd. Nauka, Moscow- Leningrad (1965), p. 76.
3. W. Vogel, Struktur und Kristallization der Glaser, Leipzig (1965), p. 97.
4. V. 1. Aver'yanov and E. A. Porai-Koshits, in: The Glassy State, Proceedings of the Fourth
All-Union Conference, Izd. Nauka, Moscow- Leningrad (1965), p. 98 [English translation:
The Structure of Glass, Vol. 6, Consultants Bureau, New York (1966), p. 98].
186 M. I. KUZ'MENKOV AND M. B. RZHEVSKII

5. V. N. Filipovich, in: The Glassy State, Vol. 3, No.1, Catalyzed Crystallization of Glass,
Izd. Akad. Nauk SSSR, Moscow - Leningrad (1963), p. 9 [English translation: The Struc-
ture of Glass, Vol. 3, Consultants Bureau, New York (1964), p. 9].
6. M. I. Kuz'menkov, Synthesis and Investigation of Wear-Resistant and Chemically Stable
Glass-Ceramics in the System CaO-MgO-Si02 , Author's abstract of candidate's disser-
tation, Minsk (1966).
DISCUSSION
R z h e v s k ii, replying to questions by Filipovich and Yus im on the paper by Kuz'-
menkov and Rzhevskii, said that, according to electron micrographs obtained by the au-
thors of the paper, potassium glasses are virtually homogeneous. Karapetyan's inves-
tigations of a series of terbium luminescence bands in the shortwave region showed that
the probability of nonradiative transitions depends on the structure of the glassy medium.
INFLUENCE OF MICROPHASE SEPARATION IN
FLUOBERYLLATE GLASSES ON THE
CONCENTRATION DEPENDENCE OF CERTAIN PROPERTIES

G. T. Petrovskii, I. S. Gilev, and S. V. Nemilov

Fluoberyllate glasses in simple systems are often regarded as weakened models of sili-
cate glasses, with a higher tendency to microphase separation. Many processes of structural
regrouping proceed at higher rates and lower temperatures in fluoberyllate than in silicate
glasses. Accordingly, it becomes easier to study certain properties and structure-formation
processes, characteristic of inorganic glasses, in fluoberyllate glasses.
Electron-microscopic studies of the structure of fluoberyllate glasses in simple systems,
carried out by Vogel [1], show that microheterogeneity regions may be detected even in op-
tically homogeneous specimens of these glasses. Vogel carried out a systematic investiga-
tion of phase separation in glasses of the lithium fluoberyllate system. The investigation showed
that: 1) phase separation occurs on addition of the smallest amount of alkali fluoride to glassy
beryllium fluoride; 2) one of the two interpenetrating phases decreases in extent with increase
of the alkali fluoride content; it forms droplike regions, which gradually diminish in size on fur-
ther increase of the alkali fluoride content.
This was the basis of Vogel's hypothesis regarding the composition of the phases
formed as the result of separation: the droplike regions comprise the high-beryllium phase while
lithium fluoride is present mainly in the matrix. The high-beryllium phase has the higher sur-
face tension.
It was also shown by Vogel's investigations that the tendency to microphase separation
in simple fluoberyllate glasses depends on the type of the alkali cation: it diminishes with weak-
ening of the field strength and increasing size of the cation in the sequence from Li to Na, K,
Rb, andes. The highest tendency to microphase separation is exhibited by Li, Na, and K fluober-
yllate glasses. In glasses of the rubidium and cesium systems minute microheterogeneity re-
gions could be detected only on electron micrographs of the optically homogeneous specimens.
The viscosities [2, 3] and elasticity moduli of binary glasses in the potassium fluoberyl-
late system were investigated, and plotted as functions of composition. The curves (Fig. 1) show
regions of anomalous variation of these properties. We attempted to correlate the occurrence
of such regions with changes in the volume ratio of the two phases formed in glasses of the po-
tassium system as the result of separation.
For this purpose we examined the possible variants of the compositions of both phases by
simple approximate calculations and comparison of the results with Vogel's experimental
data. The phase diagram of the potassium system (Fig. 2) contains two compounds in the com-
187
.0
188 G. T. PETROVSKII, I. S. GILEV, AND S. V. NEMILOV

8 8

"'8
u ~Jf
~
,:,' 3.5 ",'
I
3.5
o o
...... ......

3.0 3.0

10 20 30 Refs 10 20 30
mole,,/oKF- mole,,/oCsF-

Fig. 1. Variations of viscosity and elasticity mod-


ulus of alkali fluoberyllate glasses with the alkali
fluoride content.

position region under consideration: KBe2F5 and KBeF 3 The compound KBe2F5 is unstable at
high temperatures and does not exist above 270, decomposing into KBeF 3 + BeF2. A microhet-
erogeneous structure in the solidifying glass is observed at temperatures above the region of
stable existence of the compound KBe2F5' Therefore it can only be concluded that the alkali phase
is close in composition to the compound KBeFa. For estimation of the volumes occupied in the
glass by the two phases, we assumed that the alkali phase consists of the compound KBeF 3 and
the high-beryllium phase is BeF2.

to
600 L

500

400 rx.-BeFz+ L

300 rx.-BeFz +{3-KBeFa


rx.-BeFz +K8ezFs
l.i...""'
200 oj. ~ Fig. 2. Portion of the phase
l.i.....,~
fl-BeFz+ ~~ diagram of the system KF-
100 /(Bez~ ~
I BeF2 adjacent to BeF2.
<Q,.

Befa fO 20 30 40 50
mole"/oKF-
G. T. PETROVSKII, I. S. GILEV, AND S. V. NEMILOV 189

TABLE 1.
Mole fractions Volume fractions
KF
mole%
alkali
phase
I BeF phase
2
alkali
phase
I
IBeF 2 phase

5 0.05 0.9 0.1 0.9


10 0.1 0.8 0. 2 0.8
15 0.15 0.7 0.3 0.7
20 0.2 0.6 0. 4 0.6
25 0. 25 0. 5 0.5 0.5
30 0.3 0.4 0.6 0.4

Assuming that the compositions of the phases formed in the glass do not change with in-
crease of the KF content, and that all the added KF merely alters the phase ratio, we can calcu-
late the amount of alkali phase and the volume occupied by it in glasses of various compositions.
The sizes of the respective ions were taken into account in conversion of mole fractions into
volume fractions. The calculation was simplified by the fact that the sizes of the I{'- and F- ions
(according to Pauling) are almost equal.
The results of calculations for glasses containing from 5 to 30 mole% KF are given in
Table 1.
Constancy of the compositions of the two phases, on which the calculations were based, is
the cause of the continuous and regular variation of the volume ratio of these phases in the glass
with variation of the glass composition. The volumes of the two phases must be equal when the
total KF content in the glass is 25 mole%, i.e., the volume occupied by the alkali phase can be
the greater only in glasses containing over 25 mole% KF.
If the phase volume ratio in the glasses varies uniformly, variations of such properties as
viscosity and elasticity modulus should be monotonic in character. The fact that the experimen-
tal curves for the dependence of viscosity and elasticity modulus on the alkali fluoride content
in glasses of the potassium system are nonmonotonic indicates that the volume ratio of the coex-
isting phases in the glass varies nonuniformly. The nonuniform variation of the phase ratio may
be a consequence of changes of the phase compositions with variations of the KF content in the
glass .

Fig. 3. Electron micrograph of potas-


sium fluoberyllate glass of the composi-
tion 15 mole% KF + 85 mole% BeF2 (af-
ter Vogel [lJ).
190 G. T. PETROVSKII, I. S. GILEV, AND S. V. NEMILOV

Vogel's electron micrograph of glass containing 15 mole% KF (Fig. 3) shows that the
volumes occupied by the two phases are already almost equal (48 and 52%) in this glass. This
means that the change of the predominant phase in the glass must occur when the total potas-
sium fluoride content exceeds 15% (but not above 25% as was assumed from calculations). It is
precisely in this composition region that the viscosity and elasticity modulus of the glasses
studied in the potassium fluoberyllate system decrease sharply. Therefore the composition of
the alkali phase in glass containing 15 mole% KF does not correspond to the compound KBeF 3
but is somewhat enriched with BeF2
For comparison, Fig. 1 shows the corresponding variations of viscosity and elastiCity
modulus of glasses in the system CsF-BeF2 As already mentioned, glasses in this system are
optically homogeneous over the entire composition range under consideration; the minute drop-
like formations observed by Vogel [1] on electron micrographs can no longer have an appre-
ciable influence on variations of the viscosity and elasticity modulus of the glasses. According-
ly, the viscosity and elasticity modulus of cesium fluoberyllate glasses vary smoothly, without
abrupt changes which might be caused by a change of the phase predominating in the glass.
Literature Cited
1. w. Vogel and K. Gerth, Glastech. Ber., 31(8):15 (1958).
2. S. V. Nemilov, G. T. Petrovskii, and L. A. Krylova, Neorgan. Mater., 4:1664 (1968).
3. 1. S. Gilevand G. T. Petrovskii, Neorgan. Mater., 4:1942 (1968).

DISCUSSION
Replying to questions by Mazurin regarding the causes of a maximum of the elasticity
modulus in the composition region corresponding to 15 mole% KF and the relation between this
e
effect and phase separation, Gil v said that the cause of the maximum is not clear, but
this composition corresponds to the boundary of the region of stable glasses. The connection
with phase separation was inferred from the qualitative difference between variation of the prop-
erties of the potassium and cesium systems, which differ significantly in structure.
PHASE-SEPARATION PHENOMENA IN GLASSES
OF THE SYSTEM Lai)3-B203-P205

z. M. Syritskaya and E. S. Kutukova


., The structure of glass is not an equilibrium characteristic. It is determined to a con-
siderable extent by the melting conditions, cooling rate, and subsequent heat treatment of the
glass.
If the cations present in the melt cannot achieve the necessary environment of oxygen
anions, formation of a second glassy component in the melt becomes probable on thermodynamic
grounds. According to Weyl and Marboe [1] if the conditions of screening of two strong
cations present in the melt are satisfied this not only favors formation of two dissimilar anionic
complexes but also produces two locally separated microgroups, initiating microphase separa-
tion. Processes of structure formation by two different atomic groups, bound to different glass
formers, will be mutually independent [2].
If, in agreement with Pauling and with Gordy and Thomas [3], we take the degree of cova-
lence of the bond of a cation with oxygen as a criterion of its glass-forming ability, we can see
that phosphorus (60% covalent bond character) and boron (58% covalent bond character) are
very similar in glass-forming ability. Therefore phase-separation phenomena of various inten-
sities may be expected in borophosphate glasses. It is known that borophosphate glasses are
difficult to prepare because of their high tendency to crystallization. In order to obtain stable
glasses we studied the system La203 - B20 3- P 20 5 as lanthanum oxide diminishes crystallization
of phosphate glasses.
The compositions of the glasses studied lie on the lanthanum metaphosphate-boron meta-
phosphate and lanthanum metaphosphate - boron orthophosphate lines (Fig. 1).
Lanthanum metaphosphate was obtained in glassy form. The tendency of the glasses to
exhibit opalescence increased with increasing boric anhydride content. Melts corresponding to
boron orthophosphate in composition formed milky white opaque products on casting (1450).
The glasses were subjected to heat treatment in the softening range. They were held for
8 h at temperatures t:=: t g, t:=: t g-50, and tg < t < t f. The temperatures tg and tf were deter-
mined from dilatometric curves.
These particular conditions of heat treatment were chosen on the following grounds. It
seems likely that in the softening range the glass network becomes stabilized and the glass
structure changes so as to tend to equilibrium. Moreover, changes in the electronic structure
of the glass framework may occur in the softening range. In any event, Grjotheim and
Krogh-Moe [4] showed that during softening of glassy B20 3 a considerable number of boron atoms

191
192 Z. M. SYRITSKA YA AND E. S. KUTUKOVA

Fig. 1. Region of glass compositions studied.

change from fourfold to threefold coordination, with simultaneous increase of the B-O bond
energy. According to Myuller [5], in the softening rate the bonds of the glass network
change from more ionic to more ionic-covalent, i.e., the bond energy increases. A similar
concept is supported by Tarasov [6].
It was found in our experiments on heat treatment that glasses along the lanthanum meta-
phosphate-boron metaphosphate line containing less than 20% B20 3 (here and subsequently the
glass compositions are expressed in molar percentages) remained transparent at all tempera-
tures. Heat treatment of glasses containing over 20% B20 3 produced opalescence. No signs of
crystallization could be detected by x-ray phase analysis in any of the glasses subjected to heat
treatment.
The most interesting result was obtained for glasses along the lanthanum metaphosphate-
boron orthophosphate line. Glasses containing less than 20% B2 0 3 acquired a yellowish color
after treatment at t = t g -50, while treatment at t = tg and tg < t < tf produced an intense yellow-
orange color. The color of the specimens did not depend on the cooling conditions - either in
air directly after the heating or cooled slowly in the disconnected furnace. Glasses containing
over 20% ~03 underwent complete phase separation during heat treatment, becoming quite
opaque, and acquired color. No signs of crystallization were detected in this case either.
The transmission spectra of the glasses before and after heat treatment were recorded in
the visible and ultraviolet regions of the spectrum in the GIS Physics Laboratory with the aid
of the SF-4 spectrophotometer with a quartz optical system. The transmission cutoff of glasses
heated at t = tg and tg < t < tf is shifted considerably into the longwave region, although no
characteristic absorption bands appear in the spectra.
At the same time, the infrared transmission spectra of glasses of the compositions
[La203' 3P20 5] 4[B 20 3 ' P 20 5] and 2[La203' 3P 20 5] 3[B 20 3 P 20 5], and, for comparison, of glass of
composition 2[La2Oa . 3P20 5] 3[B203 . 3P20 5] , before and after heat treatment were recorded with
the aid of the IKS-14 instrument with LiBr and NaCI prisms, by sedimentation of glass powders
on KBr plates. These spectra are shown in Figs. 2-4. Comparison of these figures shows that
all the original glasses give similar spectra. However, the spectra of glasses after heat treat-
ment show considerable differences, dependent on their position in the system La203-B203-
P 20 5 The spectra of glass the composition of which lies on the lanthanum metaphosphate-
boron metaphosphate line, either in the original state or after heat treatment at various tempera-
z. M. SYRITSKAYA AND E. S. KUTUKOVA 193

R,%
90

445478 511 547 577 810 643678 8781042120713721537170219882032


v, em-1

Fig. 2. Reflection spectra of glass of the composition


2[La203 . 3P2051 . 3[B20s 3P20 51 before and after heat
treatment (the curves are shifted arbitrarily along the
ordinate axis).

R,%
1,11 III IV

70

80
80
90 70 50
40
80 80
70 50 30

80 40 20
30 10

20
10

10

445478511 544577810 843878 8781042120713721537170219882032


v, em -1
Fig. 3. Reflection spectra of glass of the composition [La203 .
3P20 5] . 4 [B20 3 . P 20 5] before and after heat treatment.
194 Z. M. SYRITSKAYA AND E. S. KUTUKOVA

R,%
!,II III IV

100 70
80
80
70
50
80
70 40
50
30
40
50 20
30
40 fO
20
30
10
20

10

412 4454785(1 544 574 6fO 643 676 8781042 1207137Z 15371702 1888
v, em-1
Fig. 4. Reflection spectra of glass of the composition 2[La203 .
3P205] . 3 [B20 3 . P 20 5] before and after heat treatment.

tures, do not differ significantly in the positions and intensities of the principal absorption bands.
This is in all probability due to the isostructural character of these metaphosphates. The glass
structure does not undergo extensive transformations in the specified temperature range.
This does not apply to glasses along the lanthanum metaphosphate-boron orthophosphate
line. Heat treatment produces substantial changes in the spectra of these glasses: the intensity
of the 1270 cm- i band increases and a new absorption band appears at 676 cm- i . Kolesova [7]


showed that the absorption band at 1250-1300 cm- i in the spectra of phosphate glasses charac-
terize the P = double bond and that the intensity of this band increases with the degree of poly-
merization of the glass network. It was shown by Weir and Lippincott [8] that the absorption band
at 676 cm- i is characteristic of borate groups of the LnB03 type (where Ln is a rare-earth ele-
ment) and corresponds to doubly degenerate symmetrical deformation vibrations of the B-O
bond, corresponding to boron in threefold coordination. It is known that boron is present in its
orthophosphate mainly in fourfold coordination [9]. Therefore it can be claimed with reasonable
confidence that some of the boron atoms in the glasses studied change from fourfold to threefold
coordination as the result of heat treatment.
It is not unreasonable to suppose from a comparison of the changes in the spectra of all
the glasses investigated that deep structural changes occur in glasses containing lanthanum
metaphosphate and boron metaphosphate as the result of heat treatment in the softening range.
These changes essentially consist of separation of a boron-rich component and consequent in-
crease of the degree of polymerization of the phosphate network of the glass.
Z. M. SYRITSKA YA AND E. S. KUTUKOV A 195

Literature Cited
1. W. A. Weyl and C. E. Marboe, Glass Ind., 41:429, 463 (1960).
2. O. K. Botvinkin, in: The Glassy State, Proceedings of the Fourth All-Union Conference,
Izd. Nauka, Moscow- Leningrad (1965), p. 59 [English translation: The Structure of Glass,
Vol. 6, Consultants Bureau, New York (1966), p. 47].
3. W. Gordy and W. J. Thomas, J. Chem. Phys., 24:439 (1956).
4. K Grjotheim and J. Krogh-Moe, Kgl. Norske Vidensk. Selsk. Forh., 27(18):1 (1954).
5. R. L. Myuller, Zh. Fiz. Khim., 28:1193 (1954).
6. V. V. Tarasov, New Problems of the Physics of Glass, Gosstroiizdat, Moscow (1959), p. 17.
7. V. A. Kolesova, Optika i Spektroskopiya, 2:165 (1957).
8. C. E. Weir and E. R. Lippincott, J. Res. U. S. Natl. Bur. Standards, A5(3):251 (1961).
9. P. J. Bray, in: The Glassy State, Proceedings of the Fourth All-Union Conference, Izd.
Nauka, Moscow-Leningrad (1965), p. 237 [English translation: The Structure of Glass,
Vol. 7, Consultants Bureau, New York (1966), p. 52].
DISCUSSION
Replying to questions by Filipovich, K u t u k ova said that the change in the coor-
dination of boron found in borophosphate is not the cause of microphase separation; the two
phenomena accompany each other.
DIFFERENTIATION OF STRUCTURAL
GROUPS IN SLAG GLASSES

N. M. Pavlushkin and B. I. Beletskii


The purpose of the present work was to study variations of certain properties of glasses,
from which it might be possible to draw inferences regarding their structure and structural
changes occurring during heat treatment near the softening temperature for various times. Vari-
ations of the properties of glasses during heat treatment at low temperatures have been studied
by many workers. It has been shown [1, 2] that changes of glass properties during heat treatment
in the vitrification region are due to structural transformations. others correlate changes of
mechanical properties with the location of the glass composition on the phase diagram [3]. De-
mishev [4] showed that displacement and slippage of structural complexes in glass may occur at
the vitrification temperature.
Glass of the following composition (wt.%) was taken for the investigation: CaO 25.15, MgO
1.83, Al 20 3 6.54, Si02 60.24, Na20 5.16, Mn01.08. All the calcium oxide and a certain propor-
tion of the other oxides were introduced into the glass with Konstantinovka blast-furnace slag.
Some of the glasses also contained 10% each of zinc and iron oxides. The glasses were made in
a SHit furnace at 1450 and held for 2 h at the maximum melting temperature. The glasses were
cast on a steel plate to freeze the high-temperature structural state and then put into asbestos
packing to cool.
Glass close to ours in composition was studied by Toropov and Barzakovskii [5], who
showed that under the influence of CaF2 the main glass separates out in the form of drops in a
matrix enriched with Si02. In our case it was of interest to find how the properties of the glass
alter during low-temperature heat treatment and to determine whether separation is possible in
glass without additives producing this effect.
Electron-microscopic investigation of glasses subjected to heat treatment at temperatures
causing microphase separation with droplets about 1 f.1 in size in glass with added fluoride did
not reveal any separation which could be detected with the attainable magnification. However,
certain properties of slag glasses alter as the result of treatment in the softening range (at the
temperature of the endothermic effect, 680).
Variations of the density (Ad) of glasses held at 680 for 8 h are shown in Fig. 1. All three
glasses behave Similarly. During the first period of treatment the density increases; this is
followed by a decrease. After 6 h of treatment the density increases again. Changes of the tem-
perature at which softening begins (ti. g) correspond to the density changes of the glasses (Fig. 1).
The dependence of the density and t i . s) of the glasses on the time of heat treatment is analogous
to the variations of density of borosilicate glasses with temperature [6] and of the internal fric-
tion of vitreous arsenic trisulfide as a function of its thermal history [7], which are represented

196
N. lVI. PAVLUSHKIN AND B. 1. BELETSKII 197

1.0

3
ti2;.
~ 0.5

8 T, h

Fig, 1. Variations of density


and t i. s of slag glasses, 1)
Glass without additive; 2) with
0'':: 0 r---:~-'----TI-'--/i-'-- added ZnO (10 wt.%); 3) with
<i added Fe203 (10 wt.%).

-5

by similar curves. It may be supposed that slag glasses undergo the same structural changes.
In that case the increase of density and t i . s of the glasses may be attributed to densification as
the result of annealing. The decreases of these characteristics during heat treatment between
2 and 6 h may possibly be due to development of a micro heterogeneous structure of phase sepa-
ration, as in borosilicate glasses. The subsequent increase of density and t Ls is evidently at-
tributable to densification preceding crystallization.
Determinations of density and t Ls involve interruption of heat treatment every 2 h. In or-
der to observe the behavior of the glasses under conditions of continuous treatment, they were
subjected to isothermal heat treatment in a dilatometric furnace. This made it possible to fol-
low variations of the linear dimensions of the specimens during densification of the glasses, and
also to determine t Ls immediately after heat treatment without creating tempering stresses
caused by rapid cooling. It was found that isothermal heat treatment is accompanied by shrink-
age of the specimens (Fig. 2). The temperature of heat treatment was chosen for each glass
in the region of maximum thermal expansion, and was 655-670 10, which is 25-30 lower than
ti. s of each of the glasses studied. After 6 h of treatment t i . s was determined.
Slight secondary elongation of the specimens was observed when the temperature was
raised from the constant value (Fig. 2). After softening (temperature t 1) the indicator returned
to the level corresponding to the end of isothermal treatment. Further increase of temperature
was accompanied by a second and final softening of the glasses (temperature ~). It may be con-
cluded that low-temperature heat treatment is accompanied by separation of the glass into two
phases differing in the initial softening temperatures. The experimental results are given in
Table 1.
198 N. M. PA VLUSHKIN AND B. I. BELETSKII

1 to
t2 800
tf
700

800

500
8 T, h

-
<l
400

Fig. 2. Shrinkage of slag glasses


300
during isothermal treatment. 1)
Treatment conditions; 2) glass with- 200
out additives; 3) with addition of
ZnO; 4) with addition of Fe203' 100

TABLE 1.
ti.s of glass phase after
tLs of original isothermal treatment:C
GlassNo.
glass, c I
glass phase! glass phase II

701 722 744


690 718 739
679 718 742

The density, t i . s ' and rate of shrinkage of the glasses all vary during the first 6 h of heat
treatment. By the end of this time the specimens shrink at an approximately equal rate of 175 /.l/h
for a 100-mm specimen. Crystalline phases could not be detected in the specimens by x-ray
phase analysis after isothermal treatment.
Thus, the properties of glasses alter during heat treatment at temperatures close to ti. s;
the character of the changes can be tentatively correlated with formation of a microphase struc-
ture. Evidently, spontaneous separation occurs in the glass by combination of groupings with
formation of two amorphous phases.
Literature Cited
1. N. N. Ermolenko and Z. F. Manchenko, in: Thermal and Mechanical Properties and Struc-
ture of Inorganic Glasses, Abstracts of Symposium Paper, Moscow (1967), p. 67.
2. A. S. Makhnavetskii, ibid., p. 69.
3. N. M. Bobkova, I. S. Kachan, L. M. SHich, and V. 1. Rusak, ibid., p. 60.
4. G. K. Demishev, in: The Glassy State, Proceedings of the Fourth All-Union Conference,
Izd. Nauka, Moscow- Leningrad (1965), p. 57 [English translation: The Structure of Glass,
Vol. 6, Consultants Bureau, New York (1966), p. 50].
5. N. A. Toropov and V. P. Barzakovskii, High-Temperature Chemistry of Silicate and other
Oxide Systems, Izd. Akad. Nauk SSSR, Moscow- Leningrad (1963), p. 17.
6. 0. S. Molchanova, in: Structure of Glass, Izd. Akad. Nauk SSSR, Moscow- Leningrad (1955),
N. M. PA VLUSHKIN AND B. I. BELETSKII 199

p. 141 [English translation: The Structure of Glass, Vol. 1, Consultants Bureau, New York
(1958), p. 109].
7. V. V. Tarasov, G. M. Bartenev, A. S. Eremeeva, and V. A. Ratobyl'skaya, in: The Glassy
State, Proceedings of the Fourth All-Union Conference, Izd. Nauka, Moscow- Leningrad
(1965), p. 168 [English translation: The Structure of Glass, Vol. 6, Consultants Bureau,
New York (1966), p. 183J.
DISCUSSION
Replying to questions regarding reproducibility of the experimental data and precision of
the measurements, Bel e t ski i said that the general relationships presented in the paper
were obtained from the results of numerous experiments, with the precision of tempera-
ture and expansion measurements and other factors fully taken into account.
PHASE-SEPARATION PHENOMENA DURING THERMAL
AGING OF MULTICOMPONENT THERMOMETER GLASSES
AND THEIR INFLUENCE ON PHYSICAL PROPERTIES

Yu. V. Rogozhin and E. N. Kaplina


Glasses are subjected to prolonged heat treatment (artificial aging) for stabilization of
their thermometric properties. Artificial aging comprises gradual heating to the aging temper-
ature, prolonged treatment at that temperature, and subsequent gradual cooling.
Various investigators [1-12] demonstrated the presence of microheterogeneous regions in
borosilicate glasses, appearing as the result of prolonged heat treatment and tempering. It is
reasonable to suppose that aging of thermometer glasses, which belong to the sodium calcium
borosilicate system, involves phase separation. Study of the variation of Rayleigh scattering
with the time of aging of thermometer glass confirmed that this glass contains microheteroge-
neous regions resulting from phase separation, which alter in size during aging [13].
In this paper we present the results of a study of phase-separation phenomena during arti-
ficial aging of commercial thermometer glass (All-Union State Standard GOST 1224-41) of the
following chemical composition (wt.%): Si~ 67.3~~:~, Al20 3 2.5 0.3, B 20s 2.0-:!:~:~, CaO 7.0~~:~,
ZnO 7.0 -:!:~:~, Na20 14.0 ~~:~.
It has been reported that curves for the temperature dependence of properties (including
density and coefficient of expansion) of borosilicate glasses are of anomalous character [14-16].
We investigated the dynamics of intra structural transformations with the aid of the DKV-M
dilatometer of the Institute of Glass. The character of variation of the coefficient of thermal ex-
pansion of thermometer glass with the aging time was investigated earlier [17]. A tendency to
an increase of the maximum of the thermal-expansion curve in the tg-tfrange, also reported by
others [18], was detected at the same time, In the present investigation we found that the in-
crease of this maximum (Fig. 1, a and b) during aging is not monotonic but has a complex cyclic
course. In each cycle the maximum on the AI, vs t curve in the tg-tf range increases to a certain
limit (Fig. la, 1-3) and then drops sharply (4); this is followed by increase.
It must be pointed out here that the decrease of the maximum on the curve in the tg-tf
range corresponds to the point when relative shrinkage of the specimen ceases (Fig. 2, 1,2).
It can be concluded from the foregoing that the complex variation of Rayleigh scattering
[13, 19] and of the height of the maximum on the thermal-expansion curve in the tg-tf range can
be attributed to variations in the degree of differentiation of the separated droplets and the ma-
trix glass. The fusible sodium borosilicate and possibly the sodium zinc calcium borate compo-
nents separate from the matrix and the silicon-oxygen framework is strengthened. Formation

200
Yu. V. ROGOZHIN AND E. N. KAPLINA 201

a b

to
Fig. 1. Variations of the height of the maximum on
the t!J, vs t curve in the tg-tjrange. a) Continuous
aging (500): 1) Without heat treatment; 2) heat treat-
ment for 10 h; 3) 60 h; 4) 70 h; 5) 120 h; 6) 200 h. b)
Cyclic aging (500): 1) Without heat treatment; 2) heat
treatment for 5 h; 3) 10 h; 4) 15 h; 5) 20 h; 6) 25 h;
7) 30 h; 8) 45 h .

.1l
T
50

40

30

20 2

10

O~--~--J----L---L--~--~----L---~
10 20 30 40 50 80 70 80
T,h

Fig. 2. Relative change of specimen length as a function


of aging time. 1) Continuous aging; 2) cyclic aging.
202 Yu. V. ROGOZHIN AND E. N. KAPLINA

TABLE 1
Time of Time of
tog t~ heat tog t~
heat
treatment, h treatment,h

0 545 585 20 540 582


5 535 586 25 545 594
10 545 580 30 550 592
15 540 589 45 555 589

of the fusible component as the result of phase separation altered tg while the change of the sili-
con-oxygen framework (matrix) affected t f (Table 1).
Two methods, continuous and cyclic, were used for aging of the glass. The latter com-
prised a series of cycles, in each of which the specimen was heated to the aging temperature (50 0

below tg)' held at that temperature for 5 h and cooled rapidly. Aging is much more rapid by this
procedure. It follows from Fig. 2, 2 that with cyclic aging the relative change of specimen length
becomes constant in one quarter of the time required in the continuous process.
Figure 1b shows variation of the maximum on the thermal-expansion curve in the tg- t f
range during cyclic aging as a function of the aging time.
It follows from the figure that this variation conforms to the same complex relationship as
in the case of continuous aging.
Apart from its theoretical interest, the cyclic method of aging is also of practical signifi-
cance, as it makes it possible to shorten the aging time of accurate thermometers considerably
under production conditions, with improvement of their thermometric characteristics and de-
crease of cost.
The variation of glass density during cyclic aging is shown in Fig. 3; it is also of complex
character.
The change of relative density with the time of cyclic aging is represented by the equation

1.004

.~1.003
c:Q,) 0

'"
.~... 1.002
I 0
Ql
IX
1.001 0

1.000 0
10 20 30 40 50 00
T, h
Fig. 3. Variation of glass denSity during cyclic aging.
Yu. V. ROGOZHIN AND E. N. KAPLINA 203

where a, b, and c are constants, dependent on the chemical composition of the glass, for a given
aging temperature.
A simplified physicochemical model of the aging process is put forward on the basis of the
above experimental data and of electron-microscopic investigations.
The fluctuation structure of the glass melt is frozen in quenched glass. This fluctuation
structure determines to a certain extent the initial conditions of the diffusion stage of phase
separation.
Thus, the first stage of aging during heat treatment of the glass includes diffusion growth
of fluctuation regions of heterogeneity, formed at higher temperatures. This is manifested in
increases of the density, relative shrinkage, and maximum on the dilatometric curve in the tg-
tf range. The end of growth of the microphase droplets corresponds to the greatest rise of the
maximum on the dilatometric curve in the tg-t} range. This completes the first stage in estab-
lishment of the "quasi-equilibrium" state. We call this point equilibrium of the first order.
The subsequent growth of the microphase regions proceeds not by differentiation but by
coalescence as the result of reverse diffusion from existing heterogeneity regions along the short-
est path between them. As follows from electron micrographs, this leads to formation of some-
what elongated heterogeneity regions. Equilibrium of the second order is established. This is
accompanied by further differentiation as the result of diffusion separation, as the coefficient of
diffusion at the interfaces between the heterogeneity regions and the matrix is not zero.
This is followed by further growth of the microheterogeneity region until equilibrium of the
third order is established. Undulating variation of this type was also observed by us in the case
of Rayleigh scattering [13] during aging, indicating not only growth of the microheterogeneity re-
gions but also decrease of their size at a certain stage.
Literature Cited
1. 1. V. Grebenshchikov and O. V. Molchanova, Zh. Obshch. Khim., 12:588 (1942).
2. E. A. Porai-Koshits, Dokl. Akad. Nauk SSSR, 36:285 (1942).
3. v. A. Indenbom, Dokl. Akad. Nauk SSSR, 83:509 (1953).
4. E. A. Porai-Koshits, Zh. Obshch. Khim., 12:396 (1942).
5. s. P. Zhdanov, Dokl. Akad. Nauk SSSR, 92:593 (1953).
6. S. P. Zhdanov, Izv. Akad. Nauk SSSR, Otd. Khim. Nauk, No.6, p. 1011 (1959).
7. S. P. Zhdanov, Structure of Porous Glasses and Structural Transformations in Sodium
Borosilicate, Doctoral dissertation, Inst. Khim. Silikatov, Leningrad (1959).
8. N. S. Andreev and E. A. Porai-Koshits, Dokl. Akad. Nauk SSSR, 118:735 (1958).
9. N. I. Ananich, Study of Anomalous (Structural) Birefringence in Inorganic Glasses [in
Russian], Dissertation, Mosk. Khim.-Tekhnol. Inst., Moscow (1962).
10. Yu. V. Rogozhin and L. A. Zaionts, in: Boron-Free, Alkali-Free, and Low-Alkali Glassy
Systems and New Glasses Based on Them, Moscow (1967), p. 78.
11. D. I. Levin, Study of Rayleigh Scattering in Optical Glasses, Author's abstract of disserta-
tion, Leningrad (1951).
12. s. P. Zhdanov, in: Structure of Glass, Izd. Akad. Nauk SSSR, MOSCOW-Leningrad (1955),
p. 162 [English translation: The Structure of Glass, Vol. 7, Consultants Bureau, New York
(1958), p. 125].
13. Yu. V. Rogozhin and E. N. Kaplina, Steklo, No.2, 76 (1967).
14. A. Q. Tool, J. Am. Ceram. Soc., 29:235 (1946); 31:171 (1948).
15. A. 1. Stozharov, in: Structure of Glass, Izd. Akad. Nauk SSSR, Moscow- Leningrad (1955),
p. 120 [English translation: The Structure of Glass, Vol. 7, Consultants Bureau, New York
(1958), p. 93].
204 Yu. V. ROGOZHIN AND E. N. KAPLINA

16. O. V. Molchanova, in: Structure of Glass, Izd. Akad. Nauk SSSR, Moscow- Leningrad (1955),
p. 141 [English translation: The Structure of Glass, Vol. 7, Consultants Bureau, New York
(1958), p. 109J.
17. E. N. Kaplina, Steklo, No.1, p. 75 (1967).
18. G. K. Sentyurin, Zh. Vses. Khim. Obshchestva Im. D. I. Mendeleeva, 6:643 (1961).
19. E. A. Porai-Koshits, S. P. Zhdanov, and D. r. Levin, Izv. Akad. Nauk SSSR, Otd. Khim.
Nauk, No.3, p. 395 (1955).
INFLUENCE OF PHASE SEPARATION ON
CERTAIN PROPERTIES OF GLASSES IN THE

v. B. Brailovskii, R. I. Grichevskaya,
A. I. Trushkov, and Yu. F. Tyurnikov

In this paper we report a study of phase separation of glass in the system Si02- A1 20 3 -
B20 3 - Bi20 3 - Ce02- Zr02
Microstructure of the glass was investigated with the uEMV-I00 electron microscope.
Platinum-carbon replicas were made by the preliminary shadowing method.
Figure 1 shows photomicrographs of the glass before heat treatment (a) and after heat
treatment in the temperature region of 830-960 (b, c). The glass was heated for 4 h in a gra-
dient furnace. It is seen in Fig. la that the original glass is appreciably heterogeneous even
before heat treatment; on the general relief background with individual grains of the order of
0.03- 0.05 f.L in size heterogeneity regions up to 0.1 f.L in size, consisting of grain aggregates, can
be seen. Heat treatment of the glass changes its microstructure sharply. Distinct phase separa-
tion can be seen. At temperatures of 830-860 (Fig. Ib) many droplets are still interconnected.
At higher temperatures the drops become less numerous and larger (Fig. lc). The drop sizes
and number per f.L2 of glass surface are given in Table 1.

Fig. 1. Electron micrographs of glass structure before heat treatment (a) and after heat
treatment for 4 h in the 830-860 (b) and 930-960 (c) temperature ranges.
205
206 V. B. BRAILOVSKII ET AL.

TABLE 1

Temperature Number of drops


Drop siz es, /1
range, C per /1 2

830-860 200 0.02-0.06


860-900 59 0.03-0.1
900-930 28 0.03-0.2
930-960 21 0.03-0.25
960-1000 14 0.06-0.25

Al3
JlZ IZ0
Al 1 ,Z0
IZ0 80
80
80 40
40
40

0~~---L--~--L-~---L-L~~80~0~t~O
100 ZOO 300 400 500
Fig. 2. Thermal-expansion curves of glass be-
fore heat treatment (1) and after heat treatment
for 1 h at 840 (2) and 950 (3).

It should be noted that the glass matrix has more homogeneous structure after heat treat-
ment than the original glass in all cases. Tests on specimens of the glass with the aid of the
DRS-501 instrument showed absence of crystalline formations even in glass treated for a long
time at a temperature below 1200.
In determinations of the coefficient of thermal expansion of the glass before and after heat
treatment it was found that the glass transformation temperature tg is lowered after heat treat-
ment. It follows from Fig. 2 that while tg of the glass before heat treatment was 720, after heat
treatment at 840 (it fell) by 50 (t g2 ), and after heat treatment at 950 by 110 (t g3 ).
The large difference tg-tf indicates, as follows from the paper by Strel'tsina et al. [1],
that the highly viscous phase forms a continuous matrix while the fusible phase passes into drops.
The large variation of tg in the course of heat treatment probably indicates considerable changes
in the composition of the fusible phase.
Some of the glass specimens were subjected to stepwise heat treatment, first at 750 for
2 h and then at 950 for 4 h. The electrical conductivity and dielectric losses of these specimens
were determined.* It was found that the conductivity of the glass decreased while the dielectric
losses increased after this heat treatment (Fig. 3).
It may be supposed that this occurs when the oxides causing conduction pass into the drop-
like second phase. The drops have a much higher concentration of charge carriers than the glass
matrix. Despite the fall of conductivity, the total losses increase, apparently owing to increase
of the relaxation component of dielectric losses in the drops.

*By N. G. Suikovskaya.
V. B. BRAILOVSKII ET AL. 207

Fig. 3. Temperature dependence of


log p and log tan (j .10 4 of the glass
before and after heat treatment. 1)
Log P before heat treatment; 2) log
p after heat treatment for 2 h at
750 and 4 hat 950 3) log tan {j .10 4
0 0;

(atJ= 103 Hz) before heat treat-


8 ment; 4) log tan 0 . 104 (at J =
400 350 300 250 to
I I I I 103 Hz) after heat treatment
15 16 17 18 19 for 2 h at 750 and 4 h at 950
0 0

!-10"

The following conclusions can therefore be drawn. Heat treatment of glass in the system
Si02- A1 2O:i - B2 O:i - Bi20 3 - CeO:! - Zr02 leads to formation of a two-phase glass, with separation
of the more fusible phase in drop form. The electrical resistance and dielectric losses increase
considerably and the transformation temperature is lowered.
Literature Cited
1. M. V. Strel'tsina, 0. V. Mazurin, and A. S. Totesh, this volume, p. 87.
GENERAL DISCUSSION

Va is f e 1 ' d reminded the audience that electron microscopy is highly sensitive to


the surface finish of the specimens and to replica preparation. In order to obtain a reproducible
picture of the structure of transparent glasses comprehensive studies are necessary, because
individual photographs may, as the result of chance, not reflect the real structure of the glass
characteristic of the given composition.
Agreeing with these comments, M a z uri n proposed that a special symposium, limited
to a fairly small number of participants, should be organized to discuss methods of electron-
microscopic study of glasses.
M a z uri n then expressed satisfaction with the results of the Symposium, emphasiz-
ing that studies of phase separation in glasses are very important from the practical stand-
point. He drew attention to the necessity of solving a number of terminological and method-
ological problems.
T s e kh 0 m ski i stressed the importance of studying phase separation in relation
to the possibility of regulating the structure of glass, and also the necessity of studying funda-
mentally new systems exhibiting phase separation, of interest for development of new materials.
Galakhov agreed with Mazurin's opinion of the value of the Symposium. He added
that study of phase separation in glasses should be regarded also as an indirect method for
investigation of single-phase glasses. The combined information on phase separation, both
of theoretical and of experimental character, presenting a picture of the initial stage of
phase separation and of the influence of phase separation on the properties of glasses, is close-
ly connected with the properties of single-phase glass, and its study may shed light on the origin
and development of phase separation in single-phase glass. The interest aroused by the paper of
Bobkova and Trunets is understandable in this connection.
About 40 papers were presented at the Symposium, of which three were theoretical and the
rest were partly theoretical or purely experimental. Only one of the last group was concerned
with determination of the region of metastable phase separation. Work in this field must be ex-
tended, since the practical value of such data was noted in several papers. They are also needed
for unequivocal interpretation of the dependence of properties on concentration.

208

Você também pode gostar