STRUCTURE AND KINETICS OF GLASS CORROSION

BY

DAVID M. SANDERS

A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL
OF THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

1973

ACKNOWLEDGEMENTS

The author thanks R. Condrate for sparking an interest

in materials spectroscopy while at Alfred. At the Univer-

sity of Florida, the author became indebted to W. B. Person

for the use of his spectrometers and for aid in the inter-

pretation of many of the results of this work related to

infrared spectroscopy. The author is especially indebted

to his advisor, L. L. Hench, who provided invaluable

encouragement, advice, and assistance throughout the entire

study. Finally, the author wishes to thank his wife,

Marilyn, without whose help and encouragement this research

would not have been possible.

This work was supported in part by the Office of Naval

Research and in part by the Glass Container Corporation.

TABLE OF CONTENTS

Page

ACKNOWLEDGEMENTS . . . . . . ... iii

LIST OF TABLES. .. . . . . . .. vi

LIST OF FIGURES . . . . . . . ... vii

ABSTRACT . . . . . . . .... . xii

Chapter

I INTRODUCTION . . . . . . . 1

II METHODS FOR STUDYING GLASS CORROSION
KINETICS . . . . . . . . 10

Infrared Reflection . . . . .. 10
Continuous Atomic Emission . . .. 12
The Optical System . . . . .. 12
The Minirig System . . . ... 17
The Automated System . . . ... 24
Continuous Atomic Emission . . .. 33

III QUANTITATIVE ANALYSIS OF GLASS
STRUCTURE USING INFRARED REFLECTION
SPECTRA . . . . . . . .. 37

Methods . . . ... . . . . 38
Results . . . . . . ... 40
Discussion of Results . . . .. 49
Conclusions . ... . . . . 68

IV ANALYSIS OF SILICA-RICH GELS FORMED
DURING GLASS CORROSION . . . . .. 70

Methods . . . . . . . .. 76
Results . . . . . . . . 78
Discussion . . . . . ... 93
Conclusions . . . . . . . 101

TABLE OF CONTENTS (continued)

Chapter Page

V CORROSION OF A LTTHIA-SILICA GLASS
IN VARIOUS AQUEOUS ENVIRONMENTS ... .103

Methods . . .. . . . . 104
Results . . .... . . . 105
Discussion . . .... . . . 117
Conclusions ...... .. . . . 131

VI SURFACE ROUGHNESS EFFECTS ON GLASS
CORROSION . ... . . . . . 133

Results . . . . . . . . 135
Discussion of Results . . . .. 152
Conclusions ..... . . . 154

VII CONCLUSIONS AND NEW STUDIES . . .. 156

BIBLIOGRAPHY . ... . . . . . . 162

BIOGRAPHICAL SKETCH . . . .... . . 167

LIST OF TABLES

Table Page

1 Examples of General Corrosion Conditions
and Sample States Which Might Influence
Corrosion Behavior . . . . . . 5

2 X-ray Analysis of Corroded and Auto-
claved 33L Glass . . . . . . 114

LIST OF FIGURES

Figure Page

1 The ray diagram for the optical system
used in reflection studies made using the
expanded sample compartment of the Perkin-
Elmer model 621 spectrometer . . . 14

2 Sample-holding device ("minirig") shown
mounted in aluminum holder, which is in
turn permanently fixed to the Plexiglas
mounting block . . . . . . .. 16

3 Minirig sample holder showing a sample
mountedd under stress being applied
through the loading bars . . . . 19

4 Infrared reflection spectra of a cor-
roded and of a freshly abraded glass of
a composition corresponding to lithium
disilicate . . . . . . ... 21

5 Corrosion cells used in this study as
described in the text . . . ... 23

6 "Automated system" showing one sample
rotated from the corrosion bath position
into the position for reflectance
measurement . . .... . . 26

7 Reflectance at 985 cm-I of a lithium
disilicate glass (A) as a [unction of
time, compared to that from a standard
composed of non-corroding vitreous
silica .. . . . . . . . 30

8 Schematic drawing of atomic emission
sampling techniques used in this study . 34

9 Compositional dependent changes in
infrared reflection spectra in the
lithia-silica system .. . . . .. 42

10 Compositional dependent changes in
infrared reflection spectra in the
soda-silica system . . . . ... 44

LIST OF FIGURES (continued)

Figure Page

11 Compositional dependent changes in
infrared reflection spectra in the
potash-silica system . . . . ... 46

12 A comparison of infrared spectra of
50 mole % Na2O-Si02 glasses . . .. 48

13 The reflectance of the silicon-oxygen
stretching maximum as a function of
composition for the three systems
studied . . . . . . . ... 51

14 The reflectance of the silicon-oxygen
rocking maximum as a function of compo-
sition for the three systems studied . 52

15 The reflectance of the silicon non-bridging
oxygen vibration associated with alkali
ion as a function of composition .... 53

16 The comparison between the recorded and
calculated spectra of phase separation
in a 25 mole % Li20-SiO2 glass as
described in the text ......... 56

17 Spectra of the silicon-oxygen rocking
peak for the three systems studied
showing reflectance due to lithium vibra-
tion (L) as described in the text . . 62

18 Plot of the difference in reflectance at
520 cm-1 between the spectrum of vitreous
silica and those of lithia-silica glasses. 64

19 Infrared spectrum of a 50 mole % SiO2
glass . . . . . . . ... 67

20 Two simplified models of the composition
profile of the silica-rich corrosion
layer. . . . . . . . .. 74

21 An example of data obtained from solu-
tion analysis of glass corrosion and
its relation to the calculated quantities
a, E, and B. . . . . . . ... 80

viii

LIST OF FIGURES (continued)

Figure Page

22 Changes in infrared reflection spectra
of 33L glass on exposure to static
water at 40C . . . . . ... 82

23 A comparison of the corroded and non-
corroded glass spectra in the soda-
silica system . . . . . ... 84

24 A summary of the plots of E and a cal-
culated from all the solution analysis
data reported in this work . . ... 86

25 An S.E.M. micrograph of the edge of the
corroded layer of 33L glass exposed to
static water for 350 minutes at 79.50C 88

26 A composition profile of the corroded
33L glass film obtained with an electron
microprobe . . . . . . ... 91

27 A comparison of normalized infrared
reflectance plots with normalized e
behavior for corrosion of 33L at 400C 92

28 A comparison of normalized infrared
reflectance plots with normalized c
behavior for corrosion of 33L at 79.50C 94

29 Plots of the various reaction coordi-
nates described in this work to obtain
apparent activation energies . . ... 96

30 Changes in the infrared spectrum of
33L glass with exposure to 100% relative
humidity at room temperature . . ... .107

31 Comparison of the extent of possible
reaction determined by the reflectance
of the stretching (S) and non-bridging
oxygen (NSL) peaks of 33L glass
exposed to both 0.9 ml of static
liquid water and 100% relative humidity 110

32 Infrared spectra of 33L glass given
the crystallization treatments shown . 112

LIST OF FIGURES Continued)

Figure Page

33 Changes in the infrared reflection
spectra of 33L glass on exposure to
85% relative humidity .. . .... 116

34 Comparison of the extent of reaction,
determined by the magnitude of the
stretching peak reflectance (S), for
33L glass exposed to water under static
and flowing conditions . . . ...119

35 Changes in infrared reflection spectra
upon exposure of 33L to water contain-
ing corrosion cells of different volumes
and to the residual water present in a
desiccator filled with Drierite . .. .121

36 Changes in infrared reflection spectra
upon exposure of 33L glass to hydro-
chloric acid, to a basic solution used
to corrode 33L powder, and to hydro-
fluoric acid 123

37 S.E.M. micrographs of 33L glass having
the following corrosion histories:
A. freshly abraded with dry 600 grit SiC
B. 4.5 hours with 0.9 ml static water
C. 120 hours with 0.9 ml static water
D. 216 hours with 0.9 ml static water
E. 195 hours with 100% R.H.
F. 239 hours with 100% R.H.
The corrosion temperature was 23.5C
and all surfaces were initially abraded
with dry 600 grit SiC . . . ... 125

38 Variation of B and concentration of
dissolved SiO2 with corrosion time for
33L glass abraded with dry 120, 320,
and 600 grit SiC . . . .. . . 137

39 Variation of E with corrosion time for
33L glass abraded with dry 120, 320,
and 600 grit SiC . . . . . . 140

40 Variation of a with corrosion time for
33L glass abraded with dry 120, 320,
and 600 grit SiC . . . . . . 142

LIST OF FIGURES (continued)

Figure Page

41 Variation of pH with corrosion time
for 33L glass abraded with dry 120,
320, and 600 grit SiC . . . ... 144

42 S.E.M. micrographs of freshly abraded
and corroded 33L glass. The initial
surface was abraded with dry 600 grit
SiC and the corrosion treatment was
with static water at 79.50C in a
0.9 ml cell . . . . . . ... 147

43 S.E.M. micrographs of freshly abraded
and corroded 33L. The initial surface
was abraded with dry 320 grit SiC and
the corrosion treatment was with static
water at 79.5C in a 0.9 ml cell ... 149

44 S.E.M. micrographs of freshly abraded
and corroded 33L glass. The initial
surface was abraded with dry 120 grit
SiC and the corrosion treatment was
with static water at 79.5C in a
0.9 ml cell . . . . . . . 151

Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

STRUCTURE AND KINETICS OF GLASS CORROSION

By

David M. Sanders

June 1973

Chairman: L. L. Hench
Major Department: Materials Science and Engineering

A general procedure is established for studying the

effects of chemical reactions on the structure of glass

surfaces. This work is designed to serve as a basis for

future studies relating corroded glass structures with

mechanical behavior, biological applications, and the

design of amorphous materials having greater chemical resis-

tance.

To accomplish these goals, a systems approach is used

to establish the variables required to completely specify

a corrosion reaction. New techniques including infrared

reflection spectroscopy are introduced and a theoretical

basis provided for interpretation of results obtained from

these techniques. In contrast to most previous work in

glass corrosion, a variety of techniques are used to observe

both morphological and chemical changes in the glass surface

during exposure. Corrosion conditions systematically varied

include pH, solution composition, relative humidity, solu-

tion replenishment, temperature, and pressure. The effect

of surface roughness is also studied. Details of glass

corrosion mechanisms are discussed and related to the

effects of sample state and corrosion condition on result-

ing corroded structures.

The infrared reflection techniques developed are sensi-

tive measures of both silica and alkali concentrations of

the gel formed on the glass surface by the corrosion process.

These concentrations, used in conjunction with new parameters

obtained from solution data, define a composition profile

which determines future corrosion behavior of glasses. It

is shown that infrared reflectance is proportional to the

concentration of the reflection species. This result has

general application in the use of reflection spectroscopy

for analytical as well as for structural studies. Both

glass surface roughness and the corrosion condition used

are seen to strongly influence the chemical and morphological

nature of the corroded surface in addition to affecting the

kinetics of the corrosion process.

xiii

CHAPTER I

INTRODUCTION

Reactions of silicate glass surfaces with aqueous solu-

tions are important as an engineering design parameter for

many applications. The mechanical strength,(1) optical

properties,(2) surface electrical conductivity, and

resistance to structural deterioration are all influenced

by the nature of these reactions. In addition, liquids

whose beneficial properties are destroyed by the release

of ions are frequently stored in glass containers. One

example of a problem area which has occurred in the past is

the leaching of certain lead glass glazes by acidic fruit

juices causing lead poisoning; another is the destruction

of human blood by defective glass containers. (2

Reactions of glasses with aqueous solutions can also

have beneficial effects. Recently a calcium and phosphate

containing glass(4) was discovered which becomes chemically

bonded to living tissues, promising to make possible the

design of a new generation of rejection-free prosthetic

devices. This behavior is intimately connected with chemical

reactions between the glass surface and its surroundings.

In addition, certain corrosion reactions increase resistance

to further attack of the glass. (2) Related reactions have

been proposed to lead to the formation of low index of

refraction antireflection films,(3) although other methods

are normally used commercially to produce these films.

Finally, the soluble silicate cement industry and the deter-

gent industry depend on an intimate knowledge of the condi-

tions under which binary silicates dissolve.(5)

The traditional approaches used to study the corrosion

kinetics of glasses have been based either on the rate at

which ions from the glass are taken into solution(6-8) or

on the rate of change of corrosion film thickness, surface

roughness, or some other property associated with the bulk

material. 9-13) The first approach is exemplified by the

work of Rana and Douglas, (13) who chose to follow the corro-

sion reactions of a series of binary glass compositions with

water by measuring changes in both the alkali and the silicon

concentrations in the corrosion media. These workers fre-

quently used powdered samples in order to increase the sur-

face area exposed to the corrosion media, thereby increasing

the sensitivity of their measurements. One of the most

informative studies using the second approach was made by

Charles,(14) who found that superheated stean and water

cause layers of corroded glass to form on soda lime glass.

These layers are of sufficient thickness that they could

easily be measured on sectioned samples.

Some of the difficulties encountered by workers using

the first approach are related to the need for sufficient

ion concentration in the corrosion solution for accurate

measurement. The use of powders and grains to increase the

surface area of the corroding glass complicates the inter-

pretation of the results obtained when applying these results

to the bulk systems frequently encountered in normal use.

Also, using powders and grains as the starting material

makes it difficult to take surface roughness of the freshly

prepared glass into account. In addition, the need for a

solution for analysis eliminates the possibility of studying

the corrosion of glass surfaces by gas phase reactants such

as water vapor.

Both conventional approaches share additional drawbacks.

One of the most serious is the lack of information concern-

ing the chemical makeup of the corroded glass surface. The

techniques used to measure film thickness based on optical

interference or microscopic inspection do not yield this

information at all and the solution analysis experiments

only supply indirect evidence. As a result, both the dis-

tribution and bonding of the atoms at the corroded glass

surface have remained open to speculation.(2) Another diffi-

culty with all the approaches used previously is the use of

single techniques for corrosion studies. It is unusual in

the corrosion literature for more than one tool to be

applied to a corrosion problem. As a result, much of the

careful solution analysis data was carried out using simple

model glasses, (15) most of the film thickness measurement

using a wide variety of commercial glasses, 3) and the

B.E.T. pore structure analysis using phase separated boro-

silicate glasses.16) In addition, there is scant informa-

tion about the effect of the corrosion condition adopted

and the sample state chosen on the resulting corrosion

reactions. Finally, no workers, to the author's knowledge,

have applied a systems approach to determine the variables

which must be controlled and explored to completely under-

stand the corrosion process of even one glass composition.

An ideal corrosion study would involve the use of

multiple experimental approaches to obtain complete infor-

mation about the effect of all possible corrosion variables

on all possible resulting corrosion structures and kinetics.

Some of the corrosion conditions requiring specification are

listed in Table 1. The size, shape and distribution of

microstructures and chemical species produced in the corroded

glass surface, as well as the concentration of the ions

released into the corrosion solution, are all needed to

specify the resulting corrosion structures and kinetics.

At the initiation of this research it was found that

there was no satisfactory experimental tool in current use

to obtain direct information about the concentration and

bonding of atoms present in the corroded glass surface.

Early results by Pfund(17) and later results by Anderson(18)

Table 1

Examples of General Corrosion Conditions and
Sample States Which Might Influence Corrosion Behavior

Corrosion Conditions

Ratio of glass surface to corrosion solution volume
Replenishment rate
Solution volume
Flow patterns at glass surface
Initial solution composition
Initial solution pH
Relative humidity exposure
SO2 exposure
SnC14 exposure
Temperature
Pressure

Sample States

Phase separation
Cords
Initial composition of glass
Roughness
Stress
Prior corrosion

suggested that infrared reflection spectroscopy might pro-

vide such a tool. Since the infrared absorbtion spectra

contain information about the concentration, position, and

strength of chemical bonds,119) and since reflectance spectra

can be related to absorbtion spectra,(20) it follows that it

should be possible to obtain the same information from

reflection spectroscopy. Consequently, suitable techniques

were developed in this research to enable infrared spectros-

copy to be applied to kinetic and structural studies of

glass surfaces. These techniques are discussed in detail

in Chapter II.

Also of interest is the initial stage of reaction of

glass surfaces with water. Since corrosion reactions can be

very rapid in the first few minutes of exposure compared to

later corrosion times, it was necessary to develop automated

techniques to monitor these early stages to see if there was

any unusual behavior. One of the techniques developed was

based on atomic emission spectroscopy, while the other was

based on infrared reElection spectroscopy. These techniques

are also discussed in Chapter II.

Chapter III describes the relationships between infra-

red reflection spectra and glass structure. Reflection

spectra were recorded for a series of glasses to determine

spectral trends as a function of glass composition. Several

observations are made about the structure of silicate based

glasses, independent of corrosion responses. As with all

succeeding chapters, the literature pertinent to this chapter

is reviewed there. The effect of glass corrosion is only

mentioned in passing in Chapter II[, but that work forms the

basis for all the infrared related corrosion studies which

follow. In Chapter IV the structural relations developed

in Chapter III are applied to the glass corrosion problem in

a systematic manner.

In addition to the new techniques developed to supply

specific information about the corrosion process, the more

conventional solution and microscopy techniques were also

applied where appropriate. At the onset it was clear that

it would not be possible to explore all of the variables

possible and several of these variables were held constant

throughout the entire study in order to explore the others.

One of these variables was the composition of the starting

glass which was chosen to be 33 mole % lithia and 67 mole %

silica (33L). This particular composition was chosen because

it was found to have an intermediate chemical resistance

compared to a variety of glass compositions prepared in an

exploratory study. Also, the glass was not phase separated

and its glassy structure has been studied extensively.(21)

One additional glass composition, 22 mole % soda and 78

mole % silica, was chosen for a preliminary comparison and

its behavior was found to be qualitatively similar as dis-

cussed in Chapter IV.

Another variable held constant throughout the study

was the bulk planar surface of the glass to be corroded as

discussed in Chapter II. This surface was prepared by

abrading the glass with dry silicon carbide paper, a treat-

ment which is thought to produce scratches similar to those

produced during normal use.(22) With these constraints,

some of the other possible corrosion parameters listed in

Table 1 were explored and the results discussed in Chapters

V and VI.

In Chapter V various modes of corrosion are explored

using the techniques developed earlier. Various corrosion

conditions are chosen to be models for actual exposure

environments and, while the Full analysis described in Chap-

ter IV is not possible, it is possible to project results

From that analysis to the more realistic cases found in

Chapter V. This trend toward applications is continued in

Chapter VI, where the effect of surface roughness on the

corrosion behavior is explored. This is of practical inter-

est because glass in most uses becomes scratched. Therefore

surface roughness must be studied to relate experimental

results to actual corrosion problems.

While no attempt was made to explore the silicate glass

corrosion problem in its entirety, an attempt was made to

define what would be required to make such a study. Then

specific areas of the total problem were chosen to form a

foundation upon which a more complete study could be based.

9

Rather than varying glass composition as many previous
(23)
workers have done,23) composition was held constant and

both the corrosion conditions and surface roughness were

varied in ways designed to model actual systems. The use

of the new techniques developed in this work in combination

with techniques used more conventionally in glass corrosion

studies is quite general and should be applicable to other

glass compositions and corrosion conditions.

CHAPTER II

METHODS FOR STUDYING GLASS CORROSION KINETICS

Infrared Reflection

Actual reflection can be described in terms of two

idealized components. In the first (specular reflection)

all the reflected light leaves an ideally smooth surface

at an angle equal to the angle of incidence; in the second

(diffuse reflection) the light is reflected from an ideally
(24)
matte surface with equal intensity in all directions.

Because of the manner in which samples were prepared (to be

described in a later section), the spectra obtained in this

study are largely specular in nature with a very small dif-

fuse component. This can be demonstrated experimentally

by comparing the reflection spectrum of a glass sample

polished to 600 grit with the reflection spectrum of the

same sample polished to 1/4v diamond paste. The resulting

spectra have peak positions that are identical, with the

reflectance of the maxima for the two spectra differing by

less than 0.8% because of light scattering in the 600 grit

sample.

General discussions of the reasons for resorting to

infrared specular reflection as opposed to infrared

absorption were given by Harrick 25 and Simon. (26)

Briefly, there are three primary reasons why the reflection

method is especially useful in corrosion studies. The most

obvious is that with strongly absorbing materials the

reflection method eliminates the necessity of preparing very

thin samples. In addition, the area of investigation is

limited to a depth of penetration determined by the large

extinction coefficient of silica-based glasses in the wave-

length region studied. Using an equation given by Born and

Wolf,(27) this depth at normal incidence is estimated to be

on the order of 0.5p at 1,000 cm-. As one increases this

angle, approaching grazing incidence, the penetration depth

decreases to a very small fraction of this value. A logical

extension of the present work is to use a variable angle of

incidence in order to vary this depth of penetration and to

increase the specular component reflected from rougher sur-

faces. 28) Thirdly, the spectra, as in the case of absorp-

tion spectra, are characteristic of the vibrations oE cer-

tain bonds. Thus the reflection spectra can be used both

to determine the rate of reactions taking place in the

reacting layer and, at the same tine, to provide information

about the structural changes that take place as a result of

these reactions. Also, whenever there may be an advantage

in interpretation, reflection spectra can be converted to

absorption spectra via methods discussed in detail by Simon

and others. (29,30)

Continuous Atomic Emission

Atomic emission spectroscopy has been used by many

investigators to study the amount of alkali ion corroded

from powdered glass samples as a function of time. A

recent review of these investigations was given by Das.(31)

In contrast to most of the work done in this area, the

samples used in this study were in bulk form. The planar

surfaces were ground with silicon carbide paper to simulate

the abrading effect of normal use on the glass surfaces.

Atomic emission measures the concentration of alkali

ion taken into solution, which is a direct reaction coordi-

nate for one phase of the corrosion process. Understanding

the alkali ion loss rate aids in the interpretation of the

changes which occur in the infrared reflection spectra due

to corrosion. Also, the use of continuous or semicontinuous

atomic emission sampling, described later in this chapter,

allows one to see subtle changes in the early stages of the

corrosion process that are characteristic of the state of

the surface before it begins to corrode to be compared

with the corresponding changes in the infrared reflection.

The Optical System

The optical system used for the infrared reflection

study in a Perkin-Elmer 621 spectrometer with an expanded

sample compartment is shown in Fig. 1. It consists of a

Fig. 1. The ray diagram for the optical system used in reflection studies made
using the expanded sanple compartment of the Perkin-Elmer model 621
spectrometer.

reference
f pj,^^cavity

E C..

S.----'oI of 4 toroid
E m sample

Mirrors

double beam configuration of mirrors that focus a source

image onto the sample and reference surfaces with a reduc-

tion in image size by approximately four times. The angle

of reflection, 6, was fixed at 250 in this system.

It was soon discovered that special means were needed

to permit repositioning of the front surfaces of the glass

samples at the exact point of focus after each corrosion

treatment. Two alternative experimental designs have been

employed for this purpose -- one providing greater precision

(especially in rapid reactions), the other providing more

versatility. This positioning was accomplished by milling

a cavity into the plate supporting the sample beam optics.

This cavity accepts identical Plexiglas mounting blocks

(Fig. 2) for the different positioning systems.

The reference beam has a holder similar to the one in

Fig. 2 to accept interchangeable sample-holding devices,

called "minirigs." In most of the experiments, a plane

front-surfaced aluminum mirror was placed in a minirig

which in turn was placed in the reference beam position

(Fig. 1). However, it was at times desirable to place a

glass in the minirig in the reference beam in place of the

mirror in order to take "difference" spectra.

.Fi. 2. Samnle-holding device ("iniirig") shown mounted
in aluminum holder, vwhicn is in turn permanently
fixed to the PLexiglas mounting block.

The Minirig System

Further detail of the minirig sample holder is shown

in Fig. 3. The holder is designed to hold a glass plate

under various loads to determine the effects of strain on

the corrosion process. The load is applied using the set

screw located in the back of the minirig, and the level of

strain in the glass is determined using standard stress-

optical methods. In the experiments described in this

work, howev-r, the minirig was used without the loading

roids and therefore served simply as a means of holding the

sample glass and reference glass or mirror in the infrared

b -.; between corrosion treatments. The results of the study

of the effects of stress on corrosion will be reported in

a waterr wor.. The minirigs, in conjunction with the

"'liniiig holder" show-n in Fig. 2, comprise what will be

termed the "minirig system."

As an example of the application of this minirig system

we shoe in Fig. 4 two sample reflection spectra obtained

from a glass initially of lithium disilicate composition

(33 mole % Li20-67 mole % Si02). The detailed assignment

of the peaks will be discussed in a later work, but one can

see the extent of spectral change which can be expected.(32)

Briefly, the sharp reflection peak Forming at the higher

frequency (1,090 cm- ) upon corrosion is attributed to a

transformation oF a part of the reactive layer to a

Fig. 3. Minirig sample holder showing a sample mounted under stress being
applied through the loading bars.

FRONT VIEW OF MINIRIG
WITHOUT LOADING FIXTURE

li
o I

4

Fig. 4. Infrared reflection spectra of a corroded and of a freshly abraded glass
of a composition corresponding to lithium disilicate. The freshly
abraded sample was prepared by grinding the glass surface with dry sili-
con carbide paper. The corroded sample was prepared by subjecting the
surface of a freshly abraded glass sample to distilled water in the
corrosion cell described in Fig. 5A for 24 hours at 230C.

0
0

L? O

0 0 0 0 0
0 0 0 0 0
j -..freshiy
[ .Q abraded

WV 0 0E en co

WAVE NUMBER (CM-')

configuration more closely resembling vitreous silica,

while the peak at 910 cm-1 is attributed to a lithia-rich

compound. The spectrum of a sample subjected to corrosion

for a short time can be changed to that of the freshly

polished sample by several 6-inch strokes over 600 grit

silicon carbide paper, confirming the statement above that

the reflection spectrum is characteristic of a surface

layer less than a few microns thick. For the over-all study,

many pairs of spectra similar to those in Fig. 4 were re-

corded under differing experimental conditions. Each pair

consisted of a spectrum of a corroded glass sample and

reference spectrum of the same glass sample before corro-

sion, recorded consecutively.

The corrosion cells employed in the minirig system are

shown in Fig. 5. The cell labeled A, used with corrosive

still media, is a Teflon block with a cavity containing the

medium. Figure SB shows how this cell is clamped against

the glass sample during the corrosion treatment. Only a

known and reproducible portion of the glass surface (deter-

mined by cavity shape) is exposed to the corrosive medium,

which can be replenished at predetermined time increments

if desired.

The continuous flow cell, shown in Fig. SC, consists

of a Teflon block with two perpendicular holes. Two mini-

rigs containing glass samples that have had identical pre-

vious treatments are pressed against the cell as shown in

A.

I I-
I I
I I

teflon still cell

B.

C.
-II

from to collection
flow-e- system
meter
teflon flow cell

D.

teflon finger
teflon finger

Fig. 5. Corrosion cells used in this study as
described in the text.

Fig. 5D. A controlled flow of water or other corrosive

fluid is then introduced through the tube to the sample

surfaces as indicated. The fluid can be collected from the

outflow tube for studies in atomic emission spectroscopy

using procedures described below. After a specified length

of corrosion time, the minirig is removed from the cell and

mounted in the spectrometer to record the spectrum.

The Automated System

It is sometimes desirable to compare the spectral

changes with time occurring upon corrosion for several sam-

ples subjected to different applied stresses, or subjected

to different surface treatments. Also, the kinetics of the

rapid corrosion reactions that occur at elevated tempera-

tures are of interest. A system that permits such studies

is pictured in Fig. 6 and is termed the "automated system"

in the following discussion.

The automated system consists of a "Ferris-wheel" type

of arrangement (Fig. 6) where the glasses under study are

alternately subjected to a corrosion treatment, dried, and

placed in the sample beam of the spectrometer automatically.

This is accomplished by means of the rotor which holds up

to four minirigs. The rotor is mounted in a Plexiglas block

of dimensions allowing it to be placed snugly in the cavity

of the optical bench in Fig. 1. The rotor is driven at

1 rpm by a Hurst synchronous motor.

Fig. 6. "Automated system" showing one sample rotated from the corrosion bath
position into the position for reflectance measurement.

POSITION
RECORDING

FOR
SPECTRA

DRYING
POSITION

CORROSION
POSITION

1

motor
driven
axle

plexiglas
mounting
block

'ihen a glass sample in the minirig is covered by the

corrosive medium in the "down" (or corrosion) position, the

corrosion reaction occurs for a small increment of time,

controlled by the angle t in Fig. 6. This angle can be

varied with the different modes of corrosion to be dis-

cussed shortly. As the minirig rotates to the drying posi-

tion, the corrosion reaction stops as the glass is dried

(and cooled) by blowing a stream of dry nitrogen or air

onto the surface. When the minirig reaches the "up" (or

recording) position, the intensity of reflection at a par-

ticular wavenumber is recorded by the spectrometer. Blinds

or flags on the minirigs block the infrared beam when the

rotor is not positioned so that one of the ninirigs is

exactly vertical, in order to eliminate stray reflections

from the ninirig holders.

The wavenumbers chosen for the reflectance study are

held constant during the entire kineLic study, so that there

is a maximum change in reflectance from the sample for the

particular reaction under study. This wavenumber is chosen

prior to any kinetic study after examining the spectra from

a corroded specimen and from a freshly polished sample oF

the same material (as shown in Fig. 4). The spectrometer

gain is adjusted so that when the minirig reaches the "up"

position the pen rises to its maximum with an "overshoot"

(refer to PE manual) and then Calls to zero when the minirig

goes out of the recording position. The chart speed is

adjusted so that these maxima occur next to one another to

result in a record of intensity as a function of time.
-I
Results are shown in Fig. 7 for the reflectance at 985 cm-

from an experimental lithium disilicate glass in a minirig

mounted on the rotor opposite a minirig containing a stan-

dard glass, in this case vitreous silica. This particular

wavenumber (985 cm-1) was selected because the variation in

the reflectance was relatively large after very short per-

iods of time, as seen in Fig. 4.

The relatively constant intensity spikes in Fig. 7

indicate the reflectance from vitreous silica, the standard

chosen because it does not corrode on exposure to water.

This set serves the twofold purpose of marking the exact

time of recording and of indicating the noise level of the

spectrometer. The spikes which monotonically decrease in

height form the corrosion envelope for the experimental

sample of lithium disilicate glass. The changes in reflec-

tance are amplified by using a scale expansion of five

times, and, as shown in Fig. 7, the noise is still at an

acceptable level. For the spectrum shown in Fig. 7, the

sample was corroded by water at approximately 900C flowing

through a Teflon tube, as described later. The actual cor-

rosion time added on each revolution was 10 seconds, indi-

cating that the corrosion causing the changes occurred in

less than 5 minutes, causing a drop in absolute reflectance

-1
Fig. 7. Reflectance at 985 cm of a lithium disilicate glass (A) as a function
of time, compared to that from a standard composed of non-corroding
vitreous silica. This corrosion curve was obtained using the automated
system described in the text and shown in Fig. 6. The reflectance
scale is expanded five times with respect to that shown in Fig. 4.

LU
(.)

0 7
F-
LUI VITREOUS
J (6 SILICA STANDARD
CL A

LU

5 10 15 20 25 30

MEASUREMENT TIME
IN MINUTES

intensity of roughly 1%. The fluctuation in the absolute

reflectance of the standard is seen to be less than 0.15%.

This illustration shows both the advantages and dis-

advantages of the technique. It is clearly most useful for

studies of the detailed structures of kinetic curves of

very rapid reactions. The fact that it took 30 minutes to

obtain 5 minutes of reaction time makes the method less

desirable for studying reactions that take greater than

1 h. By increasing the angle of the arc (C in Fig. 6),

the length of time the sample is exposed to the corroding

fluid can be increased to about 20 seconds of each revolu-

tion. However, other methods to be described shortly are

more appropriate for kinetic studies of slower reactions.

If all of the positions in the rotor of Fig. 6 are

filled, reflectance measurements from four different sam-

ples are recorded each revolution. One of the four minirigs

always contains a standard glass for reference, chosen so

that its reflectance at the wavenumber in question is nearly

identical with that of the glasses of interest and so that

its reflectance does not change when subjected to the cor-

rosive medium.

Several possible modes of corrosion of the glass sam-

ples can be chosen depending on the conditions required.

In the example cited in Fig. 7, a Teflon tube shown in

Fig. 5E was held with its end just touching the glass sur-

face of the minirig in the "down" position. As the rotor

revolved, the glass surface was exposed for roughly 10

seconds to water from the tube and then to the blast of air

that dried the surface prior to the reflectance measurement.

This configuration is useful for several reasons. It allows

the study of the effect of flow rates of the corrosive media

from 0.05 ml/min to more than 10 ml/min. The cell area is

identical with the other corrosion cells described in Fig.

5 Small increments of corrosion are possible because the

corrosion time for each cycle of the rotor is small.

One of the disadvantages of the automated system is

that the spectrum is monitored at only one wavenumber for

the entire experiment. This disadvantage is not serious if

peak frequencies do not shift during the corrosion process,

or if the extent of shift is known. Another disadvantage

is that the automated system ties up the infrared spectrom-

eter for the duration of the corrosion experiment.

For slightly slower reactions, the cell shown in Fig.

6 is used in a somewhat different mode. The entire spec-

trum of the untreated glass sample is recorded first. The

rotor is then turned manually so that the sample dips into

the corrosion solution. After an accurately determined

corrosion time, the rotor is again turned quickly to place

the sample in the infrared beam, the glass is wiped dry,

and the reflection spectrum of the corroded sample is then

recorded. This procedure is repeated at appropriate inter-

vals to establish a set of reflection spectra as a function

of corrosion time. From these spectra one can plot reflec-

tances at individual wavenumbers as a function of time.

Continuous Atomic Emission

The last two procedures for studying corrosion to be

described here involve modifications of standard atomic

emission techniques, using the Teflon cell shown in Fig.

5B A schematic drawing of both techniques is shown in

Fig. 8. In the first procedure (option 1), the corrosion

fluid is drawn by the suction produced by the atomizer of

the atomic emission machine. The fluid is pulled succes-

sively through a flow meter, a metering valve, and finally

through the corrosion cell before reaching the instrument.

The whole sample assembly is submerged in a water bath with

temperature controlled to 0.10C. The concentration of

analyte is recorded continuously on a chart recorder con-

nected to the photomultiplier output.

In this experiment it is necessary to monitor and con-

trol the flow rate carefully since it controls the amount

of dilution experienced by a given quantity of lithium ion

leached from those surfaces, and the flow into the atomizer

influences the intensity of emission due to a given concen-

tration of analyte. It has been found experimentally that

fluctuations must be controlled to about 1% of the flow

rate to avoid these difficulties. A continuously variable

to collection
system

Corrosion cell
shown in detail
in figure 5 D.

TEMPERATURE BATH

COLLECTION
SYSTEM

option I

VARIABLE FLOW
NEBULIZER (AUTOMATED
r OPTION)

from corrosion cell

option

from corrosion
cell
2 (semi-aulomated)

0 microbeoker
S CH holder

SYNCHRONOUS MOTOR

Fig. 8. Schematic drawing of atomic emission sampling
techniques used in this study.

atomizer and a micrometer valve provide the necessary con-

trol. Since all of the corrosion solution is analyzed

continuously, this method is particularly useful for rapid

reactions in their initial stages.

For longer periods of time and slower reaction rates a

second technique is used. This involves a positive pressure

head established by an elevated fluid supply with automatic

collection of the corrosion microbeakers (Fig. 8, option 2).

The microbeakers are rotated by a synchronous motor under

the outlet from the corrosion cell at a rate determined by

the flow rate chosen for the corrosion treatment. Each

beaker collects a sample of the analyte that is an average

of the solution concentrations over the time required for

collection. If the concentration is changing very rapidly,

the previous technique is to be preferred; if not, the

technique involving microbeakers provides certain advan-

tages.

The use of this second option in Fig. 8 allows a

blank, in most cases distilled water, to be aspirated

between the measurement of each microbeaker solution, in

order to help eliminate the errors due to machine drift.

Also, solutions flowing at rates too slow to be measured

directly using the technique involving option 1 of Fig. 8

can be collected in microbeakers, thus alleviating some-

what the problem of flow rate sensitivity discussed earlier.

Finally, this second option permits one to analyze for more

36

than one element, and to measure pHl and electrical conduc-

tivity of each solution. Thus, as is the case with the

infrared techniques, the automatic measurement of corrosion

is best suited to short experiments while the semiautomatic

method is more versatile.

CHAPTER III

QUANTITATIVE ANALYSIS OF GLASS STRUCTURE
USING INFRARED REFLECTION SPECTRA

Infrared spectroscopy has been used extensively in the

study of glass largely because it is sensitive to the local

positioning of atoms and the strength of the chemical bonds

between them.(19) An excellent review article by Simon(29)

shows that traditional emphasis has been on the study of

hydroxyl groups in glass. However, there is considerable

current interest in quantitative studies of the vibrations

of the atoms making up the glassy network, as reviewed by

Wong.(33) This shift in emphasis has been encouraged by the

introduction of infrared spectrometers having greater spec-

tral wavelength ranges(34) and by the widespread use of

laser Raman spectrometers permitting the study of vibrations

not seen in infrared. (35) Recent theoretical studies by

Bell and Dean,(36) and Gaskell(37) greatly facilitate the

interpretation of spectra from simple silicate glasses having

compositions corresponding to stoichiometric crystalline

compounds.

The first objective of this chapter is to introduce the

use of plots of reflectance vs. composition for the analysis

of the concentrations of specific vibrational species present

in three series of binary silicate glasses. It will be

demonstrated experimentally that reflectance is proportional

to the concentration of the vibrational species causing it.

Breaks in this linear behavior can be explained in terms

of abrupt changes in the nature of the species involved.

A second objective is the re-examination of the spectra

of the binary lithia-silica, soda-silica, and potash-silica

glass systems to include the silicon-oxygen rocking vibra-

tions. The spectrometers used by earlier workers who inves-

tigated these systems did not have sufficient wavelength

range to observe these peaks.(33)

in addition, it has been found that infrared reflection

spectra from glass surfaces are extremely sensitive to the

attack of the surface by water vapor and aqueous solutions

as discussed later. Therefore, a third objective of this

study is to examine the influence of various preparation

conditions on the infrared spectra of binary alkali silicate

glasses.

Methods

The glasses, except for vitreous silica and the 10 mole %

lithia composition, were prepared in platinum crucibles as

described in an earlier work by one of the authors.(3) The

glasses were cast between graphite blocks, and given a

minimal anneal between 4000-500C for 4 hours. Glass

compositions of 45 and 50 mole % Na20 and 38 % Li20 were

quenched between steel blocks to prevent crystallization.

Infrared analysis of specimens with and without annealing

and with the different quench rates showed equivalent IR

spectra.

The vitreous silica samples, GE125, were obtained from

the General Electric Company. The 10 mole % lithia-silica

samples were prepared by presintering a compact of lithium

carbonate, methyl cellulose and 5 micron silica at 1250C

and fusing the compact using an oxyacetylene torch, a method

similar to that used by Kumar.(39)

The samples were wet ground with 600 grit silicon car-

bide paper. The resulting corroded layer was removed immedi-

ately before taking the reflection spectrum using dry silicon

carbide 600 grit paper. It was found that for the 50 mole %

Na20-Si02 glass this procedure did not produce an uncorroded

surface due to the presence of water vapor in the air. It

was therefore necessary to grind the samples of that compo-

sition in a glove box sealed to the spectrometer while con-

tinuously purging both with nitrogen obtained by boiling

liquid nitrogen. This method excluded even momentary expo-

sure to atmospheric water. Only in this way was it possible

to obtain the true spectrum for the 50 mole % Na20-SiO2 glass

composition.

Selected glass compositions sent to Sharp-Schurtz Com-

pany for wet chemical analysis produced measured compositions

within 1 mole % of the nominal composition. The 50 mole %

glass was analyzed twice in this lab using the well-known

ammonium molybdate colorimetric analysis for silicon.(40)

These determinations were also within 1 mole % of the

nominal composition.

Results

Infrared reflection spectra of glasses having selected

compositions in the lithia-silica, soda-silica, and potash-

silica systems are given in Figs. 9, 10, and 11. This

three-dimensional mode of representation has been chosen to

illustrate trends in spectral bond shapes with increasing

alkali addition. The grid network drawn on each spectrum

permits the determination of both reflectance and wavenumber

for any point on a given spectrum. In Fig. 9 the spectra of

the compositions 20, 25, 30, and 35 mole % lithia are shown

broken for increased clarity. In Fig. 10 the spectrum of

16 mole % is shown broken for the same reason.

The three figures reveal similarities in the changes

which take place on the addition of alkali ions to vitreous

silica. The vitreous silica peaks tend to broaden and to

decrease in intensity. In addition, a new peak between 900
-l
and 950 cm- forms on the low frequency side of the prominent

reflection peak near 1,100 cm-i
reflection peak near 1,100 cm

Fig. 9. Compositional dependent changes in infrared reflection spectra in the
lithia-silica system.

0

540

WU 30
-j
LL
L)20

2 -- - -. -- - -- 10
I i,,
I LI II- 'IL

12001100 1000 900 800 700 600 500 400
WAVENUMBER(CM-1 )

Fig. 10. Compositional dependent changes in infrared reflection spectra in the
soda-silica system.

---Y. I I I~-~~

I I I I I',
I5 I I I
_; I I i_ i..;_ I;
100X 8G03

WJAVE N~UrvBEP(CN'fl)

Fig. 11. Compositional dependent changes in infrared reflection spectra in the
potash-silica system.

800 600

400

)D 800 600 400
WAVE NUMBER(CM 1 )

1200

There are also subtle differences in the trends seen

in Figs. 9-11 depending upon the kind of alkali ion which

is introduced. It can be seen that the exact amount of

change from the spectrum of vitreous silica occurring for a

given mole % of alkali ion increases in the order Li
Two spectra are shown in Fig. 10 for the 50 mole % soda-

silica glass. The dotted line is the spectrum obtained from

a sample whose surface is ground with dry 600 grit silicon

carbide paper immediately before taking the spectrum. The

solid line is the spectrum obtained using the dry nitrogen

technique described in the methods section. It can be seen

that there are significant changes in both peak position and

peak intensity due to the brief two-minute exposure of the

first sample to air at 50% relative humidity, thus emphasiz-

ing the importance of careful control of the environment

during the study.

As can be seen in Fig. 12, the spectrum reported for

the 50 mole % soda-silica glass kept in the dry nitrogen

atmosphere differs from the two spectra reported in the

literature for glasses of that nominal composition.(7,41

We originally thought that this difference could be explained

in terms of reactions caused by trace amounts of water pres-

ent in the grinding media used by the other authors. A

sample of 50 mole % soda-silica glass was therefore ground

under wet acetone to see the nature of the change in the

resulting spectrum. As can be seen in Fig. 12, the reaction

45mol/o Na20
\ This work

S50molo oNa2
Gaskell

"50mrol o/" N20
Flo-inskaya etal

wet
50 mdl / ,0 (ACE TONE)
This work

II

0I -L I III
1200 10C 10CO 900 so 700 S00 500 400 300
WAVENUMBER (CM 1)

Fig. 12. A comparison of infrared spectra of 50 mole %
ai20-SiO2 glasses.

3 i

explains not only the disappearance of the third hump near

800 cm but it also causes a shift of the silicon-oxygen

stretching vibration to lower frequencies. A glass of

45 mole % soda composition was then prepared and its spectrum

(see Fig. 12) was found to be quite similar to the spectra

of the nominal 50 mole % soda samples presented by the other

authors. We suggest, therefore, that the procedure used by

the other workers(42) of melting the glass samples in quartz

crucibles produced glasses whose actual compositions were

richer in silica than the nominal compositions by as much as

5 mole %. This postulate is consistent with the findings of

Vogel(21) that binary glasses having alkali concentrations

greater than 33 mole % tend to dissolve quartz crucibles.

Discussion of Results

The assignment of each of the prominent features in the

alkali-silicate glass reflection spectra is possible using

results From recent investigations. The high frequency peak

in vitreous silica near 1,100 cm-1 (S) is attributed to the

silicon-oxygen stretching vibration.(43) The new peak that

forms between 900 and 950 cm on the low frequency side of

the S peak as the concentration of alkali ion increases is

assigned to the silicon non-bridging oxygen stretching

vibration (NSR). (44)
-I
The peak found between 400 and 600 cm1 is due to rock-

ing motions of the silicon-oxygen bonds and the peak near

800 cm- is due to bending motions of the same bonds. The

last two assignments are based upon the lattice dynamical

calculations of a vitreous silica model developed by Bell

and Dean.(36)

In previous work done on binary silicate glasses, the

frequencies of various peaks for a given binary system have

been plotted as a function of composition, yielding irregular

curves. The discontinuities in these plots have been inter-

preted as indicating heterogeneous structures.(42) In the

present work, the reflectance of an individual peak maximum

rather than its wavenumber has been emphasized because it

has been found to be directly proportional to composition.

However, when frequencies are plotted against composition

and compared to those obtained by Cherneva and Florinskaya,(45)

the agreement is excellent. Differences in the 50 mole %

soda glass are also present, as discussed earlier.

Plots of the maxima of the spectra found in Figs. 9, 10,

and 11 vs. composition are given in Figs. 13, 14, and 15.

It can be seen that in all cases the behavior of the reflec-

tance is linear with composition of bulk glass. In each

case breaks occur in the otherwise linear plots. It is

argued that IF reflectance can be shown to be proportional

to the concentration of the species causing it and that it

is additive, then these breaks imply abrupt changes in the

nature of the -pecies causing the reflectance.

Rmax (Si-O STRETCH)

workThis Florinskayai
work

Gaskell

Li I 1 ___
Na 1 0
K +

10 20

30 40

MOL 0/o ALKALI OXIDE
Fig. 13. The reflectance of the silicon-oxygen stretching
maximum as a function of composition for the
three systems studied.

0
0

u 50
z

U
, 40
LL
W :
r

u 50 O
S0 LITHIA-SlLICA

S40- SODA-SILICA
LLU
-J
LLJ
LU 0h

20L POTASH -SILICA

10"

0 10 20 30 40 5
MOL O/o R20

Fi_. 14. The reflectance of the silicon-oxygen rocking
maximum as a function of composition for the
three systems studied.

0

\

W A Lithium
z dSodium
< OPotassium
Ill -

U7 0.
09 0

7 I
w

04
0 3-

MOL /% R 0

Fig. 15. The reflectance of the silicon non-bridging
oxygen vibration associated with alkali ion
as a function of composition.

Figure 16 is a demonstration of an analysis which is

possible using the principles introduced above. The solid

line is the measured spectrum of 25 mole % lithia-silica

glass. This glass is phase-separated and known to contain,

using the lever principle on the miscibility dome, approxi-

mately 75% of the reflectance of 33 mole % lithia-silica

glass and 25% of the reflectance of vitreous silica. The

dotted curve was obtained by multiplying the spectrum

reflectance of vitreous silica by .25 while the dashed

line was obtained by multiplying the reflectances of 33

mole 2 lithia-silica glass (33L) by .75. The triangles are

obtained by adding, at each wavenumber, .25 times the

reflectance of SiO2 plus .75 times the reflectance of 33L.

It is seen that the agreement between the calculated and

measured spectra of 25 mole % lithia-silica glass is excel-

lent. The slight discrepancy that is seen may be due to the

fact that the end points of the miscibility gap at the

annealing temperature are not exactly those given. Also,

the errors in the calculated spectrum may be due to errors

in measured reflectances compounded by the addition process

used to obtain it.

By separating the 25 mole % lithia-silica glass spectrum

into its component parts, as described in the previous para-

graph, it is seen that the recorded spectrum is simply the

sum of two simpler spectra. The peaks in the simpler spectra

have already been assigned. Thus, the prominent peak in the

Fig. 16. The comparison between the recorded and calculated spectra of phase
separation in a 25 mole % Li20-SiO2 glass as described in the text.

---MEASURED 25L
A CALCULATED 25L
SLL5 -- 75%" of 33 L
U 25% ot Si02
<40-

-30
i- 30 I &.
LL.
r20 - "
S,. NS 1
10 ../.. o ' ,

/ ..
1200 1000 800
WAV ENUMBER(CM )

recorded spectrum is the combination of the stretching peaks

due to 33L and silica. The shoulder at 950 cm- is largely

due to the silicon non-bridging oxygen stretching vibration

of 33 mole % lithia-silica glass.

In Fig. 13 the reflectance of this combination band

comprised of two stretching peaks for the three systems

studied is plotted against composition. The two stretching

peaks are the S peak already discussed and a new LS peak

which is the silicon-oxygen stretching peak of lithium

disilicate glass. The data of Gaskell(37) and Cherneva and

Florinskaya(45) are also plotted for comparison. It can be

seen in Fig. 13 that the reflectance of the bridging silicon-

oxygen stretching band varies linearly with the addition of

alkali with an abrupt change in slope between 18 and 33

mole % alkali depending on the kind of alkali ion added.

This linear behavior makes the technique an excellent non-

destructive analytical tool For the determination of the

alkali concentration of high-silica glass; the basis of a

technique used to follow the progress of glass corrosion

reactions discussed in Chapter II.

The break in the lithia-silica plot in Fig. 13, denoted

by the dashed line, is due to the domination of the combina-

tion band by the LS vibration as seen in Fig. 16. At high

silica concentrations, this combination band is controlled

by the S peak because the proportional contribution from

vitreous silica is higher. The S and LS contributions to

the combination band are illustrated in Fig. 9 with straight

lines drawn through the portions of the combination band due

to each respective peak. In the lithia-silica system, the

break in the plot of the reflectance of this combination

band vs. composition, designated by a dashed line, occurs

when the reflectance of the hump due to the LS mode is

greater than that due to the S vibration. The break is an

artifact due to the way in which the reflectance of a com-

pound band was defined. If, instead of plotting the highest

reflectance of the combination band, the reflectances due to

the S vibration alone are plotted, the break then occurs at

33 mole % lithia. This is taken to be the true reflectance

plot for this vibration and is shown ii Fig. 13 as a solid

line. In the bands for the potassia-silica and soda-silica

svstens the two stretching peaks are not separated (see

Figs. 10 and 11) and therefore do not exhibit this compli-

cat on.

The breaks for the three systems occur at 18, 21, and

33 mole % alkali oxide for the potash-siLica, soda-silica,

and lithia-silica systems, respectively. At the transforma-

tion temperature the boundaries of the miscibility gaps of

the soda-silica and lithia-silica systems are known to be
(46) (46)
20 and 33 mole 7, respectively. ( Charles(41 arg"'es that

the consulate temperature of the potassia-silica miscibiLity

gap is below the glass transformation temperature and hence

not observable experimentally. But by the same argument,

such a gap is thermodynamically probable. The gap boundary

should be closer to vitreous silica than for the soda-

silica system. (46) The breaks in the plots of reflectance

vs. glass composition are thus seen to be due to the deple-

tion of a vibrational species, in this case, the stretching

vibration of the high silica phase. The linear decrease in

reflectance corresponds to the expected linear decrease in

the volume fraction of the vitreous silica matrix as the

total glass composition moves across the miscibility region.

The fact that miscibility does not take place in the potash

system for kinetic reasons does not prevent the local

segregation of non-bridging oxygen around the alkali ions

described by Weyl.(2) Thus, the plots of reflectance vs.

composition detect both phase separation on a macroscopic

scale and the tendency towards phase separation on a micro-

scopic scale.

Figure 14 shows plots of the maximum reflectances of

the silicon-oxygen rocking peaks shown in Figs. 9, 10, and

11. The breaks occur at the same compositions as were

observed for the stretching plots just discussed. The fact

that the break in clots of the R peaks is not as abrupt as

observed for the plots of the S peaks indicates that the

effect of alkali ions on rocking vibrations is not as great

as on stretching vibrations. In fact, the break in the

lithia-silica plot is barely detectable in Fig. 13. The

fact that the spectra of strongly phase separated 20 and 25

mole % lithia-silica glass (given in Fig. 9) do not have

bimodal rocking peaks further strengthens the argument that

the rocking vibration is quite similar in vitreous silica

and lithium disilicate glass.

Figure 17 shows a detailed comparison of the rocking

neaks for the three systems studied. It is seen that the

snectra of all the glasses in the potash-silica and soda-
-I
silica systems exhibit a minimum near 520 cm The spectra

of the lithia-silica glass, on the other hand, exhibit no

such ninima, having additional reflectance in this area

(designated by the letter L). The greatest difEcrence

between the spectra of vitreous silica and 33 mole 2 lithia-

silica glass occurs near 520 cii It is concluded that

this Tma.tiumln is d;e to the motion of lithium ions. The

corrcsoondi;ng sodium and potassium-oxygen vibrations are

not seen because lhey are found at lo;: 'r frequencies and

therefore buried in the continuum.l. This assignment is

in agreement IJi.Lh the work of Ri.sen, who has reported tThis

vibration in the same frequency, range for a variety of

1 1hiio.; on containing system:-;.(48-50

FiIurc iS is o plot of the differ-eice In reflectance

at 520 cn; betvseen the spectrum of vi.troous silica and the

snectra of various' 1 it.hia-silica o lasscr. It is seen thkt

the beh'ovior con be approxi maed by stIraight lines having

breaks at 1.0 and 33 iaole % ithLia. Thei breaks indicate

abrupt chan 'es ii the envi ronme nt aroudn the li thi. u ions

Fig. 17. Spnctra of the silicon-oxygen rocking peak for the three systems
studied showing reflectance due to lithium vibration (L) as described
in the text.

POTAS H-SILICA

600 500 4.00
VWV/ENL CB,,lvR
CMf'O

600 500 400
WAVtE NiUiMBER
CM1'

600 500 400
WAiV E N UMBER
CM,1

LITHiA-SILICA

SODA-S ILICA

-1
Fig. 18. Plot of the difference in reflectance at 520 cr1
between the spectrum of vitreous silica and those
of lithia-silica glasses.

20 30 40
MOL /o Li20
'2_

30,

25-

20

ljI

J15-
LL
Lj

10o1.

/

'0

10

50

1111~11~

and coincide with breaks found in plots of the activation

energy for electrical conduction as a function of composi-

tion.. () Also, Kumar and Maitra(39) have shown that glasses

having lithia concentrations less than 10 mole % do not

release lithium ions into aqueous solutions, indicating that

the silica-rich phase completely surrounds the small volume

fraction of lithia-rich phase present in these phase

separated glasses.

Again, this linear behavior of the reflectance of

lithia-silica glasses at 520 cm- with mole lithia permits

the neasuroment of the concentration of lithium ion in

lithia-silica glasses. Since lithium is not easily detected

using x-ray techniques because of its lo.,; atomic number, the

use of infrared reflectance spectroscopy provides a pco(lsiny

analytical tool for the study of lithia-based glasses.

A plot of the re.lectance o the: nron-bridging silicon-

oxygcen stretching reflectance (NSR) vs. composition is g.ven

in Fli. 13. It is seen that he behavior is fiuite similar

in all three systems with a ina.i'mum ticlectance occurrin.g

at roughly 36 nole %, R70. The explanatior [or this .i;u

illustrates the nature of the species cause i .;_ this band.

It caon be shown(52) thiat a b:and due to a specific vibra-

tional species produced by the combination of two reactants

caln 1'e plotted against the conLccnt rat on oE orn of those

Creactarts to )ield ai maxi num. This naiaum corresponds to

th, stoichiometry of vibrational species causing the

reflectance. In the present case, the alkali ion reacts

sith the silicon-oxygen network to produce non-bridging

silicon-oxygen bonds. At 33 mole % R20, there is one non-

bridging oxygen per silicon-oxygen tetrahedron. Since the

maximum in the plot of the NSR peak vs. composition occurs

at compositions slightly greater than 33 mole % R20, it is

seen that the NSR peak is due to a vibrational species which,

on the average, has slightly more than one non-bridging

oxygen per tetrahedron.

At concentrations greater than the 36 mole %, the number

of totrahedra having more than one non-bridging oay'gen

becomes substntiatl Those tetrahedra are linked by the

remaining bridging : -iiicon-oxygen bonds forming chainl ik;

structures. The two non-bridging ox-ygen;is in a tctrahedron

can then cou-ie to vibrate in a synmmetric:al anL d asyn irnetri ca

mn;ner, forrting peaks on either side of the io, -b ridgj;in

oxygen peak. This is, in fact, seen iri the spectrum for

50 m;oe t soda-silica glass shoi-n in Fi 1. 9. Coup in. of

the two NSR vibrations leads to theb too scial. n aks diagran.ed

sch.m:I ic(.al] v en either side of 900 cm-. Th. e'; pe'ks 'ad

to bo';th 1i. tilinrg in of the valley between the S and iSR

rpcai,;s snr th:, now ; shoulder formed on the low freq'uency s: do

of tiL NS a at S -20 cm
ofr tK'. 1150 "-e"' at %820 en

4; -
A!.
/ CI

1000

N

800 60.
WAVE N UMBER (CM')

~4i. 19, Infrared snectrun of a 50 role % SiO2 glass. The dotted lines
represent possiblee peaks forming this spectrum as described in
the t--t,

LjJ ?
u I

- .

w* 20,

LL
L 10

0

I-

1200L;;;Y'"~

400

"IL II-~3_L~IYI-Y~iYI-~ L-LI~II~PUPLI~U*l ~Y

s-^~lr. ,,~imrr~-yrrrmi~~~;~rr -ur;uuar-. .rrur~ -rrlr~';uu-r~wrurar;u

---

Conclusions

It has been shown experimentally that infrared reflec-

tances are proportional to the concentrations of the vibra-

tional species causing them. Further, these reflectances

can be added w.avenumber by wavenumber to show more clearly

the origin of the broad bands frequently encountered in the

spectra of amorphous substances (see Fig. 16). These char-

acteristics allow the extension of assignments made on

stoichiometric glasses to more complex glass systems.

In addition, breaks in the linear plots of reflectance

vs. bulk composition in a series of glasses imply changes in

the nature of the vibrational species involved. These

changes can be a depletion of a vibrational species as was

seen in the stretching neak behavior near the miscibility

gao boundaries (see Fig. 13), a change in the local geometry

as was seen in the behavior of the lithium vibration near

]0 mole ,, or a partial coupling of several vibrations as

was seen in the behavior of the non-bridging silicon-oxygont

bond at hi gh soda concentrations (see Figs. 15 and 19).

In the comnos ition sequences the rocking peak follows

much the same changes as does the stretching peak. This is

expected because both neaks arc due to dificrent motions of

the same atomic group, the silicon--oxygen tetrahedron. The

phase separation seen in the lithiu-silica system in the

20 and 25 wole t glasses does not cause shoulders on the

rocking peak as it does on the stretching peak (see Fig. 9).

This implies that the rocking peak behavior in the lithia-

silica system is not as sensitive to local composition

fluctuations caused by phase separation.

The high frequency base of the rocking peak at 520 cm-

contains reflectance due to the motion of lithium ions in

the glassy network. This reflectance is not seen to the

same extent (see Fig. 17) in the potash and soda-silica

systems due to the increased mass of those cations, which

causes the peaks due to them to be buried in the continuum.

In searching for an optimal method of preparation for

the glass surfaces, it was found that grinding with 600 grit

dry silicon carbide paper produced a very reproducible sur-

face free of reaction products due to the interactions of

the glass surface with its environment. For 50 mole %

soda-silica glass, this process had to be carried out in a

totally dry environment without even momentary exposure of

the sample to the water vapor in the atmosphere. For this

reason, the author feels that the spectrum of 50 mole

soda-silica glass given in this work is the most accurate

to date.

CHAPTER IV

ANALYSIS OF SILICA-RICH GELS
FORMED DURING GLASS CORROSION

Charles described the destruction of alkali silicate

glass by water in terms of three chemical reactions.(54)

The first involved the penetration of a proton from the

water into the glassy network, replacing an alkali ion which

is, in turn, released into solution:

-Si-O-R+H 20-Si-O-H+R +OH- (1)

This reaction produces an OH ion and a non-bridging oxygen

bond attached to a hydrogen ion (NSII). A second reaction,

with the hydroxyl ions obtained in Equation (1), destroys

silicon-oxygen-silicon bonds to form non-bridging silicon-

oxygen bonds (NS) as:

-Si- -Si-+ OH --Si--OH+-Si- (2)
I I I I

The NS bond formed in Equation (2) interacts with the water

molecule forming another NSII bond and a hydroxyl ion, which

is free to repeat reaction 2 over again, e.g.,

-Si-O + I12 O-Si-O-H+O-0H (3)
I 2 1

The silicic acid formed in all three reactions is soluble

in water under the appropriate conditions of pH, time, tem-

perature and ion concentrations.

Following the discussion by Douglas et al.,(55) the

relationship between the activity of NS groups in the aqueous

solution is derived as:

[NS] [-Si-O 1 = 0-9.8 (4)
NSH [-Si-O-Hi]

assuming that the pH of the solution is less than nine and

hence the LiOH and -Si-O-R (NSR) are completely dissociated.

Equation (4) shows that in acid and in neutral solutions

I -
[-Si-O-H] >> [-Si- ]

Also, if one assumes equilibrium conditions, [NSH] decreases

rapidly with increasing pH. When [NSH] is very low, [NSL]

increases and Equation (4) no longer holds. One expects,

from equilibrium considerations, that the amount of R

release would decrease with increasing pll above pH 9.

Douglas argues(55) that as the pH exceeds 9.8, the number

oF NS groups increases, leading to a net charge and breakup

of the structure and increased rate oF silica release.

The above relationships do not specify the location of

the various corrosion species, e.g., the silanol (NSH)

groups mny either go into solution or form a Film on the

glass surfaces. Likewise, NSR groups may be present in

films on the glass surface that have been exposed to high

pH values. The extent to which a glass forms a film and

the coherence of that film has a strong influence on corro-

sion behavior.

Shnidt(56) introduced a parameter, a, defined as

Si2 1
m NNa2

where R20-mSiO2 is the composition of the glass and N is

the moles/cm2 of glass surface which dissolves into solution.

a is a measure of the extent of selective dissolution, when

a = 0, compared to total dissolution, when a = 1.0. At

intermediate a values, a mixture or selective and total dis-

solution occurs concurrently. Shmidt found in binary soda-

silica glasses that at a given temperature a remained unity

with increasing silica concentration to a critical concen-

tration where it decreased precipitously to zero. At higher

temperatures the critical composition occurred at higher

concentrations of silica. Thus for the ono-hour reat!Ment

used by Shnidt, at 100"C, only compositions having silica

concentrations >80 mole % formed films.

In the present study the dependence of c on exposure

tiie is determined for 33 mole % Li 20-SiO2 (33L) and 31

mole % Na20-SiO2 (31N) glasses. In addition, a new parameter

which is a measure of the amount of sitica available for film

formation is defined. This quant ity, referred to as "excess

silica," or c, is the difference between the amount of

silica which would go into solution if the ratio between

the alkali concentration and the silica concentration in

the original glass were maintained (balanced silica or 3)

and the amount of silica which actually does go into solu-

tion. The relationships between these quantities are:

Y solution 'Y glass (6)
X solution X glass

mPM Si02 i mwR iF Pm

1PPM R+ 2

and [ + .

a (balanced silica) = (---P 7- = RmwR J

and

c (excess silica, PPM) = PP;i SiO2 = PPM Si021 ]

(9)

where Y = moles SiO X = moles R20, Pm = mole fraction

iR20 in glass, mw = moleculare weight, PPM SiO02 = concentra-

tion of silica in solution, and PPM R = concentration of

R' in solution.

Figure 20 shows two hypoLthetical plots of silica con-

centration vs. its distance from the film surface, winch

can be calculated from the parranters described in this work.

The area under the composition proFile (S) shaded in Fig. 20

is equal to one-half times the product of the ordinate and

Z
O

7 A

0 D

I A B. --

DISTANCE

Fig. 20 Two simplified models of the coiposi tion profile
of the silica-rich corrosion layer. Part A is a
profile oF a more resistant layer while Part B
represents a less resistant layer.

abscissa of the shaded triangle. The ordinate is the difEer-

once between the measured surface silica concentration (Y)

and the silica concentration of the original glass, tIus

having units of grains of excess silica divided by cm of

Film volume. The abscissa is the profile thickness (X) in

cm. The product, therefore, has units of grams of excess
2
silica divided by the film surface area (A) given in cm .

The quantity S can also be obtained from the parameter

c if the solution volume (V) and V/A ratio is known.
-3
S grams excess silica grams excess silica x 10
cm2 film surface 1,000 cm solution

V CIcm.3 solution
x-
xA 2
S cm film surface

In the present study, the V/A ratio is simply the depth of

the cylindrical corrosion cell cavity (d) and therefore the

area under the composition profile (S) is proportional to

the product of e and the corrosion cell depth (d) in cm,

e.g., S = edlO6. Both plots in Fig. 20 have the same area

and hence the same e values For the surface film, but the

excess silica in film A is distributed much closer to the

surface thin in film B. Thus, Film A is characteristic of

a resistant film. Film B, on the other hand, is thicker

because it has the same c value as filn A but a lower sur-

face SiO2 concentration, a characteristic of less resistant

films.

Methods

The glass samples were corroded using the static cor-

rosion cells as described in Chapter IL. After exposure,

the cell solution was immediately diluted for later analysis

for lithium or sodium concentration using atomic emission

spectroscopy, and for dissolved silica using the well-known

molybdenum blue calorintotric method.(40) At low tempera-

tures and/or short times, it was necessary to corrode a

number of samples in the same manner in order to obtain

sufEicient concentrations of silica for analysis. The pH

was determined for each run using short range plI paper

(10.5) calib rated with a pll electrode.

The exposed glass surfaces were analyzed using 1RRS as

described in Chaptor II. Lach corroded glass sample was

run with a freshly polished 55L glass for comparison. The

diCffrence in reflectance between specific Features of the

uncorrod-d glass and that of the corroded samples was

measured for each spectral feature of interest. These

fe..tiures wore the omaximiumo valuo of Lhe silicon-oxygen

stretching peak at L,100 cm- (S), Lho maximum value of the

silicon non-brlidling oxygen stretching peak at 425 cm (NS)

the reflectance at 520 cm I due to Ithe 1 thium ion (L), and

thIe maximum value of the rocking peak at 475 cm- (T) The

greatest extent of deviation in reclectaince fromn the uncor-

roded re rectc ancer was taken as 100'l ract-ion. 1Thus, the

reflectance values of corrosion events loss than this maxi-

mum value were divided by the maximum value and multiplied

by 100 to obtain the value of the percent reaction reported

herein.

An Acton Laboratories, Model MS-67 electron microprobe

was used to determine the composition profile of one of the

films. The glass sample was exposed to iatcr at 79.5C for

17 hours. Its in [rared spectra wore recorded and the sample

imbedded in epoxy. The imbedded sample was then cut perpen-

dicular to the corrosion film and the resulting exposed

edge polished using standard metallographic techniques to

0.25 pm diamond paste. Both the corroded section and a

standard silica sample were plated side-by-side with a gold

film.

The 33L sample w-as scanned for silicon at 10 microns

per inch from the interior portion out to the edge of the

samnle. Tn addition, the filu was scanned point by point

manual at 2-micron intervals from the edge of the sample

inward. This procedure was repeated five times and an aver-

age taken. A linear re Lationship between the number of

counts nor second and the concentration oF the silica was

assumed as a first approximation.

Results

Figure 21 is the solution data obtained for 33L at 400C.

It is representative of results obtained over a temperature

range of 24.5 to 79.50C. In Fig. 21 the solid line with

circles is the plot of balanced silica, 4, computed from the

lithium concentration of the corrosion solution as discussed

previously. Since g is proportional to the lithium concen-

tration, as with lithium release, its plot is a straight

line on the log-log scales as shown in this figure.

The concentration of silica in solution is best approx-

imated by three line segments with breaks corresponding to

pH values of 7 and 9. a values are shown as a dotted line.

As with the behavior of silica in solution, a has three

linear portions and two breaks at the same values of pH, 7

and 9. Figure 21 shows that, upon exposure of a freshly

abraded 33L sample to water, the initial a value decreases

with time, indicating an increasing tendency for film forma-

tion. The initially higher value of a, which implies a

greater tendency For total destruction of the glassy network,

is probably due to the presence of high energy silica network

sites produced by the abrasion process, as discussed later.

In Fig. 21 the first break in the a curve occurs at

about 170 minutes, after which the slope of the a curve

becomes positive, indicating a continually decreasing ten-

dency for Film formation. Finally, at the second break,

Fig. 21. An example of data obtained from solution analysis of glass corrosion
and its relation to the calculated quantities a, e, and B.

-....... alpha
----excess silica (2)
- --silica in solution
-0--balanced silica (P)

1
\ 1
\ 1

1 10 100 1000
TIME IN MINUTES

0.1

0.01
0.01

10,000

occurring at 3,000 minutes, the a-time slope becomes still

steeper and rapidly approaches unity. When unity is reached

there is no tendency for film formation and the whole reac-

tion is one of total dissolution of the silicon-oxygen net-

work. However, the value of ,a = 1 does not preclude the

possibility of a high silica film being present on the glass

if elsewhere in the system there are insoluble corrosion

products having higher alkali content than the glass.

The e curve, representing the silica that can be avail-

able for film formation, is shown in Fig. 21 as a dashed

line. Because the excess silica is the difference between

a relatively large quantity, e.g., balanced silica, and a

relatively small quantity, silica in solution, its behavior

closely parallels the former. Only when the balanced silica

and the solution silica approach the same order of magnitude

does deviate significantly from 6, as can be seen in Fig.

21. The E curve reaches a maximum when the value of a is

0.45, indicating that roughly half of the total reaction

has led to film formation.

In order to interpret the solution behavior, infrared

reflection spectra of the same corroded 33L glasses were

recorded and are summarized in Fig. 22. The silicon-oxygen

stretching peak of 33L (LS) is seen to sharpen and increase

in reflectance with corrosion time while the non-bridging

silicon-oxygen peak associated with alkali ion (NSL) is seen

NSL

a: zB
L

0.1

10

1,00
O, O.

10,0000
1200 1000 800 600
WAVENUMBER(CM"1)
Fig. 22. Changes in infrared reflection spectra of 33L glass on exposure to
static water at 400C. The letters are explained in the text.

to decrease. Both these changes show that the corrosion

process leads to a structure similar to that of vitreous

silica.

The similarity of the vitreous silica reflection spec-

tra and a 33L glass corroded in water for 22 hours at 50.50C

is shown in Fig. 23. Concurrently, the silicon-oxygen rock-

ing peak (R) in Fig. 22 increases in reflectance and

sharpens during corrosion. This change can also be attrib-

uted to an increase in the silica content of the structure

sampled by the reflectance method as discussed in Chapter

III. After longer periods of time, the peak B is seen to

form, which is attributed to the bending mode of a silicon-

oxygen bond. All of the above changes are expected in a

leaching process with the increase in the silica concentra-

tion resulting from the loss of alkali ions.

Direct evidence for the decrease in alkali concentra-
-I
tion can be seen as a decrease in reflectance at 520 cm

This reflectance, labeled L in Fig. 22, is due to the vibra-

tion of lithium ions and its decrease signifies a decrease

in the average concentration of the lithia. Another change,

the decrease in reflectance between peaks LS and NSL, due to

coupling between them, is also related to the loss of the

lithium ions responsible for this coupling.

The shoulder () in Fig. 22) formed at 1,010 cm- after

exposure to water For 3,000 minutes at 400C has not been

conclusively assigned. It may be due to an interference

gl2aS5
SiO2
-.-- 31N

treatment
as abraded
as abraded

60- ........ 33L 22Hrs. 50.50
S-- 31N 15Min. 50.5
50- i\ 31 N 17Hrs. 50.5

/ AC
z 40-i
- 0 -, c \

20 -
, 0u / / ,' V. c \f'
W # I \ '*" \

10 / /, ,; ... ..... : .- :- .. ,
," # //* ...... -*'"

1200 1000 800 600 400
WAV ENUMBER(CM )
Fig. 23. A comparison of the corroded and noncorroded glass spectra in the soda-
silica system. The siectra of vitreous silica and corroded 33L are
included for comparison.

fringe caused by the corroded layer, but more probably is

due to the silicon non-bridging oxygen vibration associated

with bonded hydrogen ions in a polymerized gel structure.

A similar assignment was given by Naudin for an absorption
-1 (57)
peak found at 952 cm1

Figure 24 is a compilation of the solution data

obtained at four temperatures. Part II of Fig. 24 contains

plots of a values for the 33L glass. Also included is a

dotted line showing the behavior of a for 31N at 50.50C.

In part I of Fig. 24, the behavior of excess silica is shown

for the same glass compositions and temperatures. For

clarity, letters coded for the appropriate temperature are

placed at the ends and break points of the solid lines for

the 33L curves.

Figure 25 is a micrograph of the corroded layer of 33L

treated for 350 minutes at 79.5C. The area seen is the

fractured edge of the corroded layer made by drawing a

diamond stylus through the film. The gage ball marked with

the arrow is 0.87 microns in diameter. The white, jagged

flakes are debris produced by the diamond stylus. The

corrosion layer in the foreground has grinding scratches

with pits roughly 0.5 microns in diameter running along them.

These pits appear to be the same depth as the corrosion film.

They apparently arc due to the total dissolution of the

glass in the scratch valleys.

TIME IN MINUTES

Fig. 24. A summary of the plots of e and a calculated
from all the solution analysis data reported
in this work. The letters are used to
designate the corrosion temperature involved.

Fig. 25. An S.E.M. micrograph of the edge of the corroded layer of 33L glass
exposed to static water for 350 minutes at 79.50C. The gage ball
is 0.87 microns in diameter.