Citation
A durability evaluation of soda-lime-silica glasses

Material Information

Title:
A durability evaluation of soda-lime-silica glasses using electron microprobe analysis, infrared reflection spectroscopy and other techniques
Creator:
Clark, David Edward, 1946-
Publication Date:
Language:
English
Physical Description:
xii, 156 leaves : ill., graphs ; 28 cm.

Subjects

Subjects / Keywords:
Corrosion ( jstor )
Durability ( jstor )
Infrared reflection ( jstor )
Infrared spectrum ( jstor )
Ions ( jstor )
Music analysis ( jstor )
pH ( jstor )
Precipitates ( jstor )
Spectral reflectance ( jstor )
Weathering processes ( jstor )
Dissertations, Academic -- Materials Science and Engineering -- UF
Glass -- Corrosion ( lcsh )
Materials Science and Engineering thesis Ph. D
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis--University of Florida.
Bibliography:
Bibliography: leaves 149-154.
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by David Edward Clark.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
This item is presumed in the public domain according to the terms of the Retrospective Dissertation Scanning (RDS) policy, which may be viewed at http://ufdc.ufl.edu/AA00007596/00001. The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact the RDS coordinator (ufdissertations@uflib.ufl.edu) with any additional information they can provide.
Resource Identifier:
025457761 ( ALEPH )
02952067 ( OCLC )

Downloads

This item has the following downloads:


Full Text













A DURABILITY EVALUATION OF SODA-LIME-SILICA GLASSES USING ELECTRON MICROPROBE ANALYSIS,
INFRARED REFLECTION SPECTROSCOPY AND OTHER TECHNIQUES








By

DAVID EDWARD CLARK


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


1976

























For Dagney















ACKNOWLEDGEMENTS


The author gratefully recognizes the expert

guidance of his committee members and wishes to especially acknowledge the valuable research experience gained through his association with Dr. Larry Hench. He also thanks Morris Dilmore, Ed Ethridge and Carlo Pantano for their assistance in data interpretation, Teresa Ne Smith for her well executed drawings, and Wayne Acree for his invaluable electron microprobe assistance. With deepest gratitude, he acknowledges the editing assistance and patient support of his wife, Sue.














TABLE OF CONTENTS



Page


ACKNOWLEDGEMENTS ................................... iii

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

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

ABSTRACT ........................................... xi

CHAPTERS

I INTRODUCTION ................................. .

General Discussion ........................... 1
Summary of Accomplishments .................. 3
Glass Durability Review- Aqueous Corrosion .. 4 Glass Durability Review- Weathering ......... 14 Research Objectives ......................... 16

II A TECHNIQUE FOR ELECTRON MICROPROBE ANALYSIS
OF CORRODED SODA-LIME-SILICA GLASSES ..... 20

Introduction-A Discussion of the Problem . . 20 Experimental ................................ 22
Results and Discussion-Bulk Glasses ......... 24 Results and Discussion-Corroded Glasses ..... 31 Summary ..................................... . 37

III INFRARED REFLECTION SPECTROSCOPY OF CORRODED
SODA-LIME-SILICA.GLASSES UTILIZING FREQUENCY
SHIFTS .................................... 39

Introduction-A Review of Infrared Reflection
Spectroscopy of Glasses ................... 39
Experimental ................................ 41
Results and Discussion-Technique
Development ............................... 42
Summary ..................................... 64












IV APPLICATION OF THE ELECTRON MICROPROBE,
INFRARED REFLECTION SPECTROSCOPY AND
OTHER TECHNIQUES FOR EVALUATING
AQUEOUS CORROSION OF SODA-LIME-SILICA
GLASSES ...... ............................... 66

Introduction -Review of Corrosion
Evaluation Techniques ..... ................... 66
Experimental -Corrosion and Instrument
Techniques ............................... 68
Results and Discussion of Aqueous
Corrosion Data for the Various
Techniques ..... ............................. 72
Summary ..... .................................. 110

V APPLICATION OF THE ELECTRON MICROPROBE,
INFRARED REFLECTION SPECTROSCOPY AND
OTHER TECHNIQUES FOR EVALUATING THE
WEATHERING BEHAVIOR OF SODA-LIME-SILICA
GLASSES .... ................................. 113

Introduction-Review of Weathering and
Comparison with Aqueous Corrosion ........ 113
Experimental -Weathering Parameters ........ 113
Results and Discussion of Weathering
Data for the Various Techniques ........... 117
Summary .... ................................... 133

VI SUMMARY AND CONCLUSIONS ... ................... 141)

Glass Durability .... .......................... 14)
Evaluation Techniques ... ..................... 146

REFERENCES .... ........................................ 149

BIOGRAPHICAL SKETCH .................................... 155















LIST OF TABLES


Table Page


1 Coefficient of Variation in X-ray Intensity as a Function of Surface Roughness for a 20 Mole Na2l0-10 Mole CaO70 Mole SiO2 Glass. 36

2 Glass Compositions Investigated. 69

3 Solution Data for the 20 Na2O-80 Si02
and 20 Na20-lO CaO-70 Si02 Glasses
Corroded in Static Aqueous Solution
at 1000C. 89

4 Solution Data for Various Soda-LimeSilica Glasses Corroded in Static
Aqueous Solution at lO0�C. 93

5 Solution Data for the Commercial Glass
Corroded in Static Aqueous Solution at
1000C. 95

6 Reaction Rate Constants of the Glasses
Calculated from Equation (1) Using lO0�C
Solution Data. 97















LIST OF FIGURES


Figure Page


1 Mechanisms of Glass Corrosion. 5

2 Sampling Depths for Various Glass Durability Analysis Techniques. 19

3 Effect of Stationary Electron Beam on
Na X-ray Intensity. 25

4 Effect of Electron Beam Diameter on Na
X-ray Intensity. 27

5 Na X-ray Intensity as a Function of
Electron Beam Impingement Time and Sample
Velocity. 29

6 Na X-ray Intensity as a Function of
Electron Beam Impingement Time for
Several Glasses. 30

7 Na X-ray Intensity as a Function of Corrosion Time for Three Soda-Lime-Silica Glasses. 32

8 SEMs for Soda-Lime-Silica Glasses Corroded for 12 h at 1000C. 34

9 Effect of Surface Roughness on the Infrared Reflected Intensity. 43

10 SEM of a 20 Na20-lO CaO-70 SiO2 Glass Polished to 120 Grit and Exposed to Atmosphere (= 70 Per Cent R.H.) for 24 h
at 250C (1300x). 44

11 Infrared Spectra for a Freshly Abraded (f.a.) Glass and the Same Glass Corroded
for Various Times. 46









Figure Page

12 SEMs of Corroded Glasses. 49

13 Infrared Spectra for Several Na 20SiO2 Glasses. 50

14 Infrared Spectra for Several Na20--5 CaO-SiO2 Glasses. 51

15 Calibration Curve for the Si-O Stretch Peak Maxima of Na2 0-SiO2 and Na20O
10 CaO-SiO2 Glasses. 52

16 Wavenumber of Si-O Stretch Maxima as a Function of Corrosion Time. 55

17 Na 0 Concentration as a Function of Corrosion Time. 58

18 Na20 Concentration as a Function of the Square Root of Corrosion Time. 62

19 Infrared Reflection Spectra for Corroded Glass Specimens. 74

20 Infrared Reflection Spectra for the Later Stage of Corrosion for the 20
Na20-10 CaO-70 SiO2 Glass. 77

21 SEMs of Corroded Glass Surfaces. 79

22 Infrared Reflection Spectra for Freshly Abraded and Corroded l0Na20-lO CaO80 SiO2 Glass. 80

23 Infrared Reflection Spectra for Freshly Abraded and Corroded 10 Na20-20 CaO70 SiO2 Glass. 81

24 Infrared Reflection Spectra for Freshly Abraded and Corroded 30 Na20-l0 CaO60 SiO2 Glass. 82

25 Infrared Reflection Spectra for Freshly Abraded and Corroded 20 Na2 0-20 CaO60 Si02 Glass. 83


viii








Figure Page


26 Infrared Reflection Spectra for Freshly Abraded and Corroded Commercial Glass. 86

27 Infrared Reflection Spectra for Freshly Abraded Soda-Silica, Soda-Lime-Silica
and Commerical Glasses with a Constant
Na20 to SiO2 ratio. 88

28 EMP X-ray Intensities as a Function of Corrosion Time for the 20 Na20-10 CaO70 SiO2 glass. 98

29 EMP X-ray Intensities as a Function of Aqueous Corrosion Time for the Commercial
Glass. 100

30 Auger Signals at Various Depths in the Freshly Abraded 20 Na2Ol0 CaO-70 SiO2
Glass, 02

31 Auger Signals at Various Depths in the 20 Na20-10 CaO-70 Si02 Glass Corroded for
2 d at 1000C. 103

32 Auger Signals at Various Depths in the 20 Na20-10 CaO-70 Si02 Glass Corroded for
5 d at 1000C. 104

33 Auger Signals at Various Depths in the 20 Na20-10 CaO-70 Si02 Glass Corroded for
9 d at 100'C. 105

34 Auger Signals for Ca as a Function of Depth for Various Samples Corroded at 100'C. 107

35 Auger Signals at Various Depths in the Freshly Abraded Commercial Glass. 108

36 Auger Signals at Various Depths in the Commercial Glass Corroded for 10 d at 100'C. 109

37 Infrared Spectra for Weathered 20 Na2-0-10 CaO-70 SiO2 Glass. 118

38 Infrared Spectra for Weathered 10 Na2-0-11 CaO-80 Si02 Glass. 119









Figure Pa


39 Infrared Spectra for Weathered 10 Na2 0-1 20 CaO-70 SiO2 Glass. 120

40 Infrared Spectra for Weathered 30 Na20-1 10 CaO-60 SiO2 Glass. 121

41 Infrared Spectra for Weathered 70 Na2 01 20 CaO-60 SiO2 Glass. 122

42 Na20 Concentration as a Function of the Square Root of Corrosion Time for
Weathered Glass. 125

43 Infrared Spectra of Commercial Glass for Type 1 Weathering at 100'C-100 Per Cent
R.H. 127

44 Infrared Spectra of Commercial Glass for Type 1 Weathering at 50�C-100 Per Cent
R.H. 128

45 Infrared Spectra of Commercial Glass for Type 2 Weathering at IOO0C-100 Per Cent
R.H. 129

46 Infrared Spectra of Commercial Glass for Type 2 Weathering at 50�C-I00 Per Cent
R.H. 130

47 Infrared Spectra of Commerical Glass for Type 2 Weathering at 250C-100 Per Cent
R.H. 131

48 SEMs of Weathered Commercial Glass. 132

49 EMP X-ray Intensities as a Function of Exposure Time for Weathered Commercial Glass. 135

50 Auger Signals at Various Depths in Type 2 Weathered Commercial Glass, 100�C-100 Per
Cent R.H., 15 d. 137

51 Schematic Comparing Static Aqueous Corrosion and Weathering. 143

52 Schematic of Plausible Corrosion Mechanisms. 145

X















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 A DURABILITY EVALUATION OF SODA-LIME-SILICA GLASSES
USING ELECTRON MICROPROBE ANALYSIS,
INFRARED REFLECTION SPECTROSCOPY AND OTHER TECHNIQUES


By

David Edward Clark


March, 1976


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


Traditionally, solution analysis has provided the bulk of glass corrosion data. Although this technique permits an indirect determination of glass dissolution kinetics, it provides little information concerning the structural and chemical changes occurring within the glass surface during corrosion. Techniques have been developec , employed infrared reflection spectroscopy and electron microprobe analysis, for measuring directly the surface compositional changes that occur during glass corrosion. These methods are combined with Auger electron spectroscopy, scanning electron microscopy and improved solutior









analysis techniques for examining the durability of a systematic series of soda-lime-silica glasses. The mechanisms and kinetics of both aqueous corrosion and weathering for these glasses are discussed in terms of preferential Na decoupling from the glass network, SiO 2rich film development and Ca-rich precipitate formation. A binary soda-silica and a soda-lime-silica commercial glass are also investigated and the results illustrate the effects on durability of CaO and other commercial glass additives.


xii















CHAPTER I

INTRODUCTION



General Discussion

The subject of glass durability is certainly of

no small importance when one considers the number of glass containers produced for consumer use and the various environments to which these glasses are exposed. Although visual observation of most commercial glass reveals no visible reaction, all glasses are subject to some degree of degradation. The extent of degradation is affected by numerous factors: the composition and phase (i.e., gas, liquid, solid) of the reacting environment, the composition and thermal history of the glass, surface composition of the glass, water affinity of the glass surface, the ratio of the glass surface area to the solution volume and concentration, temperature, pressure, and duration of exposure. This list is not exhaustive and should serve only to illustrate a few of the variables which influence glass durability. "Durability" refers to the characteristic of the glass which determines the resistance to interaction












with its environment. A durable glass reacts very little with its environment, while a glass exhibiting poor durability is very reactive. "Aqueous corrosion" will refer to the durability of a glass in solution (where in the study the solution is pure water before reaction begins) and "weathering" will refer to the durability of glass exposed to various humidity conditions.

The mechanisms of reaction between glass and

water are not elementary. As soon as the reaction begins, products due to reaction between the two alter the corrosion solution and affect the course of further reactions. The study of these reactions is further complicated by precipitate formation as well as chemical and structural changes occurring within the surface of the glass during corrosion. After more than a century of endeavor, only a few of the mechanisms associated with glass durability have been identified. Progress has been impeded primarily by the absence of adequate surface chemistry tools. Until recently, the most widely used technique for evaluating glass durability has been solution analysis. At best, this method provides an indirect determination of glass reaction kinetics, but yields little information concerning the chemical gradients or structural modifications produced in the glass by these reactions. Presently, there are several research tools











which, when used in combination, are capable of providing a direct and more precise evaluation of glass durability than was previously possible. In the present investigation, the techniques of adapting these tools for durability evaluations are demonstrated for the commercially important soda-lime-silica glasses.



Summary of Accomplishments


1. An electron microprobe technique utilizing electron

beam enlargement and specimen translation has been developed that is suitable for evaluating chemical

durability.

2. A method involving infrared reflection peak frequency shifts has been developed that permits

quantitative analysis of corroded glasses independent of their surface roughness.

3. A chemical durability evaluation (aqueous corrosion and weathering) of a systematic series of

soda-lime-silica glasses has been performed using electron microprobe analysis, infrared reflection

spectroscopy, Auger electron spectroscopy, scanning

electron microscopy and solution analysis. Both the usefulness and limitations of each technique are discussed in terms of the depth within the surface

from which information is obtained.












4. The chemical durability of a commercial sodalime-silica glass has been evaluated and compared

to that of the pure soda-lime-silica glasses.



Glass Durability Review-Aqueous Corrosion


It is important to recognize that when a glass

reacts with an aqueous solution both chemical and physical changes occur within its surface structure. Consider a simple binary soda-silica glass whose structure is shown in Figure 1. The addition of I mole of Na20 to 2 moles of SiO2 produces 2 moles ofnonbridging oxygen atoms, and provides 2 moles of Na atoms to associate with these oxygen atoms. Each bridging oxygen is shared by two Si atoms and each Si atom is tetrahedrally coordinated with four oxygen atoms, three of which are shared with other Si atoms. The addition of CaO to a SiO2 glass produces the same number of nonbridging oxygen atoms as an equivalent mole fraction of Na20. However, CaO provides only one Ca atom per two nonbridging oxygen atoms.

Initially the corrosion solution consists of pure water (pH = 7.0), but as the reaction proceeds, accumulation of corrosion products causes both the chemical composition and pH of the solution to change. Several










0
I I
-Si-O-Si-O Na I I
ONa ONaO I I
-Si-O-SiONa I I
0


0
I I
-Si-O-Si-OH
I I
+ OH
+ H20--ONa
I I
- Si-O-Si-OH
I I
0


I
-Si-ONa
I


NaOH--OH
NaOH~~Na
I
-Si-OH
I
0


+ NaOH
+ glass dissolution
products


t>O
pH _9
selective Na dissolution


t))0
pH>9
total dissolution


Figure 1. Mechanisms of Glass Corrosion.


t=O pH=7












investigators 1,2,3 have suggested that the reactions should be considered in two stages:

1. the initial stage, involving ion-exchange between Na ions from the glass and H ions from

the solution, during which the remaining

constituents of the glass are not altered.

2. the second stage, in which breakdown of the

silica structure occurs and uniform corrosion

ensues.

During the initial stage of corrosion, Na ions diffuse from the glass into the solution. Wang and Tooley2 and Lyle4 have postulated that the driving force for this diffusion process is the Na ion concetration gradient. In order to maintain electrical neutrality, H ions diffuse from the solution into the glass and occupy those sites vacated by the sodium ions. The recognition of this second diffusion process has led to the hypothesis that the driving force for the initial stage involves more than just the Na ion concentration gradient.5'6 Budd7 discusses the initial stage of corrosion in terms of an electrophilic reaction whereby the electrophilic reagent (H ions in this case) attacks sites of electron excess (nonbridging oxygen atoms).











Bacon and Calcamuggio3 suggest that the rate controlling mechanism in the first stage is H ion diffusion since the H ion diffusivity is much lower than the Na ion diffusivity in glass. This is attributed to the tight bonding of the H ion in the hydroxyl group. However, Douglas and Isard8 have shown that the rate of corrosion is determined by the rate of Na ion diffusion in
9
glass. Douglas and El-Shamy later found that there was no agreement between the diffusion of Na ion in the bulk glass and the apparent Na ion diffusion coefficient determined from corrosion experiments. Furthermore, it was found that the apparent diffusion coefficient for binary glasses, calculated from chemical attack, may be as much as 106 times greater than that for diffusion in bulk glass. Das and Douglas'I0 statement, ". . . the removal of alkali ions from the glass by water is more complicated than a single ion exchange reaction in which the rate determining step is the diffusion of ions through the corrosion layer," accurately describes the complexities of the first stage reaction mechanism. Thus, at present it can only be concluded that the diffusion of Na ion through the corroded layer (away from the glass) is an important contributory factor in the initial stage of corrosion.











Douglas and El-Shamy9 have developed the equations for determining the kinetics during the first stage of corrosion. At constant temperature,


Q = K � ty (1) where Q = quantity of Na ions from the glass

t = duration of experiment

K = reaction rate constant assuming constant

glass area

y = 1/2

Lyle4 derived a similar expression with the inclusion of a temperature function.


y log Q = log t - c
T

where T = absolute temperature

b, c = empirically determined constants.



Although not immediately obvious, equations (1) and (2) require that the area of glass undergoing corrosion remain constant. The importance of surface area on reaction kinetics has been previously recognized by El-Shamy and Douglas. However, in most durability











experiments the glasses were ground into fine powders to increase the surface area and therefore effectively reduce the time required to conduct the experiment. This necessitates the inclusion of a geometrical factor in the kinetics equations since the area of the glass capable of supplying Na ions is continuously decreasing as corrosion time increases. The error in spherical surface area is 25 per cent for a 20 pm diameter particle with a corrosion layer of 1 pm, and increases as the thickness of the corrosion layer increases. Therefore, equations (1) and (2) are not valid for powder durability studies. Barrer12 derived equations allowing for changing surface area of reacting spheres in solid state kinetics, and similar expressions can be derived for corrosion of powders. This will not be attempted in the present investigation since all data were taken from bulk (planar) surfaces.

Several other characteristics of the first stage of corrosion are

1. the effective area of SiO2 exposed to the corrosion

solution is increased by the production of surface

micropores resulting from Na ion removal

2. the Na to Si ratio in solution is greater than that

for the bulk glass (this suggests selective leaching of the Na ion)












3. the pH of the solution increases as a result of

Na ions replacing H ions in solution. EI-Shamy

et al. 13 proposed the following equilibrium

equation for the initial stage of corrosion at

the surface of a soda-silica glass.

SiONa(g) + H20 SiOH(g) + NaOH (3)



For pH < 9 the NaOH and SiONa(g) are assumed to be completely dissociated and the corresponding acid constant is given by


[SiO-] [H+]
= K a(4) [SiOH] a

where the brackets represent the activities of the species. The value of K was assumed to be that of dissociation
a
step of orthosilicic acid, H4SiO4, which was found to be 10-9.8 Thus equation (4) becomes


[SiO-] = 10-9.8 + pH (5)
[SiOH]


Equation (5) provides a means of calculating the mole fraction of nonbridging oxygen sites occupied by H ions. Using this equation, Douglas and EI-Shamy9 show that for pH < 9 practically all the nonbridging oxygen sites are occupied by H ions at 300C. For pH > 9, less than one-half of











the sites are occupied by H ions, leaving the other sites to be occupied by Na ions. The increased Na ion surface concentration tends to hinder further Na ion diffusion from the glass.

Although no sharp line of demarcation exists between the first and second stages of reaction, the latter effectively begins at pH > 9 for soda-silica glasses. This reaction involves complete breakdown of the silica network and all species dissolve at approximately a uniform rate. El-Shamy et al. have shown that even vitreous silica dissolves in solution where pH > 9. This reaction is illustrated schematically in Figure 1 and shows that the presence of OH ions at the glass surface is a pre requisite. The equilibrium equation for the reaction can be represented by

I I
- Si-O- Si-OH + OH
I I (g) (soln.)



-Si-OH + 0--Si-OH()
- (g) 0 (soln.) (6)



It can be seen that increasing the OH ion concentration will favor dissolution of silica. However, it was demonstrated earlier that an increase in pH suppressed the diffusion of Na ion from the glass. From this information











it would appear that Na ion diffusion from the glass would decrease at pH > 9. However, experiments have shown that selective leaching of Na ions occurs even at pH = 12, although not to the same extent as at pH < 9. This has been explained by Douglas and El-Shamy9to be due to the decrease in width of the siliceous layer at high pH. The reaction kinetics for this stage can be determined from equations (1) and (2) for planar surfaces. The value of y has been experimentally determined to be slightly greater than one, thus suggesting some selective leaching even during the second stage of reaction. Budd7 discusses the second stage reaction in which a nucleophilic reagent, such as OH ions, attacks positions of electron deficiency. Accompanying this reaction is also an electrophilic reaction where Na ions in solution attack the nonbridging oxygen atoms created by the nucleophilic attack. The equilibrium conditions established by these reactions determine the value of y obtained for the second stage.

Sanders and Hench14 discuss the two stages of reaction in terms of three corrosive parameters, a. , and c. These parameters are calculated from the quantities of various species in solution. The dissolution parameter, a, measures the extent of selective leaching. It is equal to zero for the dissolution of only one component (first stage) and is equal to one for total











dissolution (second stage). The balanced silica parameter, , is the amount of silica which would be in solution if selective leaching occurred. The film parameter, e, is a measure of the SiO2 available on the surface to form a silica-rich film and is directly proportional to the film thickness. By measuring the time dependence on these parameters, composition profiles and information on film behavior can be obtained.15

The selective leaching parameter, a, and film parameter, e, equations are given below.

PPM SiO2
MW SiO2 1
+ P (7)
PPM R + PPM R SiO2
MW R + MW R2+

= PPM SiO (1-- ) (8)

2

Symbols in the above equation are:

PSi02 = mole fraction of SiO2 in glass

PPM = parts per million MW = molecular weight

R + = alkali ion

R 2+ = alkaline-earth ion


Similar solution parameters have also been derived by Sen and Tooley.16










Glass Durability ReviewWeathering

The degradation of a glass surface due to interaction with the atmosphere is referred to as "weathering." Weathering has been classified into two types:

1. condensation-runoff, in which moisture collects

on the glass surface until natural runoff occurs,

carrying away the reaction products

2. condensation-evaporation, in which a thin layer

of "fog" forms on the glass surface but evaporates before droplets form.
Type 1 weathering is very similar to aqueous corrosion in which the solution is continuously replenished at a specified rate. During the time that the water droplet remains on the surface, dealkalization of the glass occurs with a simultaneous increase in the pH of the water. If the surface area to solution volume ratio is high in the water droplet, rapid pH increases will cause total breakdown of the glass in contact with the droplet. The resulting localized deterioration and roughening of the glass surface leads to solution entrapment during runoff. Further degradation of the surface continues where the solution is pooled in the corroded pores. The pH in subsequent water droplets will not change as rapidly since the majority of the Na ion leaching occurs in the initial droplets. However, as discussed previously,











for pH > 9 the total dissolution significantly decreases the thickness of the leached layer. Sanders and Hench17 have shown that for a large surface area to solution volume ratio, (simulates water droplet on glass) large quantities of the alkali remain in the glass surface at high pH values. Apparently very little alkali is required to increase the pH of the small volume to the region where total dissolution occurs.

Type 2 weathering is characterized by the presence of reaction products on the glass surface. This type of weathering,produced in cyclic temperature or cyclic humidity environments, is the greatest source of concern to glass manufacturers. Attack is initially manifested by a slight dimming of the glass surface, followed by the formation of a noticeable white film which varies in thickness according to the composition of the glass. Tichane and Carrier18 have studied these films extensively using transmission electron microscopy and X-ray diffraction analysis, and have shown that they are rich in Na. Furthermore, Tichane19 has shown that the Na-rich films react with most acidic gases such as CO2 and SO2 to form their respective salts at temperatures below 300'C. Since CO2 is present in significant quantities in our atmosphere, Na2CO3 should be a reaction product on most weathered glasses. Simpson20 has also verified the












presence of both Na2CO3 and CaCO3 by X-ray diffraction analysis on the surface of weathered commercial glasses. Unlike type 1 weathering, the precipitate formed in type 2 can act as a protective barrier for retarding further corrosion.

Little quantitative data are available for either type of weathering because the research methods for extracting the appropriate data have been developed only recently. Simpson 2122 developed a "haze meter" for quantitatively measuring film growth and appears to have successfully applied this technique to commercial glasses. However, the method does not yield any useful information concerning the structural modification of glass during corrosion.



Research Objectives


There are several newly developed techniques for

characterizing the chemical durability of glasses. Infrared reflection spectroscopy (IRRS) is an exceptional tool for determining structural and chemical changes in sodasilica glass surfaces exposed to various environments.23 Similarly, corrosion solutions have been analyzed by atomic absorption (AA) and atomic emission (AE), and the results used to indicate the uniformity of corrosion.











Moreover, recent advances promise to make electron microprobe analysis (EMP) an attractive technique for durability studies on soda-lime-silica glass.24 Each of these instruments yields averaged information which is characteristic of a volume extending from the surface to a specific depth (usually > 0.5 Vm) within the sample. Thus, an accurate chemical analysis of extremely thin surface films, multilayered surface films, and surface reaction layers containing composition gradients may require additional techniques. Rynd and Rastogi25 have demonstrated that Auger electron spectroscopy (AES)isa particularly effective means of investigating the surface compositions of commercial glass fibers because the escape depth of the
0
Auger electron is small (<5 A). Furthermore, Pantano
__a.26
et al. have shown that AES coupled with ion milling is a reliable method for determining chemical profiles on a submicron level in corroded bioglass specimens.

The primary objective of this investigation was to characterize the durability of soda-lime-silica glass and a similar commercial glass through the joint use of IRRS, EMP, SEM (scanning electron microscopy), solution analysis, and AES. Comparison of these glasses with a soda-silica glass also yielded information concerning the roles of CaO and Al203 in glass durability. For this investigation, surface analysis is defined according to the











depth within the sample from which information was obtained: the "near" surface determined by AES refers to
0
depth < 5 A from the glass-air interface; the "middle" surface, measured by IRRS, extends to =.05 1m;and the "far" surface is defined by the effective EMP electron penetration depth as = 1.5 jim. Using ion milling with AES yields information from all three regions, whereas in solution analysis the depth from which information is obtained depends on the extent of reaction (Figure 2). These techniques, with the exception of solution analysis, allow for a direct measure of glass durability. Comparison of the results from the various techniques provides a better understanding of glass corrosion.




















*These values vary slightly according to the density, composition and corrosion extent of the glass.













AES
IRRS -'- PmI Corrosion Depth [Soln. Anal.] EMP


Uncorroded Glass


Figure 2. Sampling Depths for Various Glass Durability Analysis Techniques.















CHAPTER II

A TECHNIQUE FOR ELECTRON MICROPROBE
ANALYSIS OF CORRODED SODA-LIME-SILICA GLASSES


IntroductionA Discussion of the Problem

Electron microprobe analysis (EMP) of alkali glasses has been recognized as a problem by previous
27-32
investigators. Generally, the X-ray intensity for the alkali species changes as a function of time when an electron beam remains stationary on a glass specimen. An accompanying change in the X-ray intensities of other atomic species in the glass is also observed. 29,30 This X-ray intensity variation suggests a compositional change in the sample volume irradiated by the electron beam. Pantano et al.33 have reported similar changes in the Na Auger electron intensities for alkali glasses subjected to a stationary beam. Plausible mechanisms associated with these local compositional changes are discussed by Lineweaver,27 Borom and Hanneman,31 and Tanaka and Warrington.28

The surface condition of the glass specimen is

also an important consideration in compositional analysis.











Both abraded and corroded surfaces can cause considerable variation in X-ray intensities when the dimensions of the surface perturbations and electron beam size are comparable. However, the capability for such analyses is essential, especially in the container industry where glasses are subjected to corrosive environments.

The rate of Na X-ray intensity change is proportional to the electron beam current and specimen temperature, and is related to the bulk composition of the glass under consideration. Small electron beam currents, low specimen temperatures, and the presence of Ca, Al, and Mg give correspondingly lower rates of alkali X-ray intensity changes.34 Douglas and EI-Shamy9 report that the diffusion rate of Na ions is much greater in the corroded layer than in the bulk glass. Thus, the rate change of X-ray intensity should be correspondingly greater for corroded specimens than for uncorroded glass. Several methods for reducing the rate of change in alkali intensity are discussed in the literature. 30, 32, 34-36 An electron raster technique has been employed to stabilize the X-ray intensities of K20 -SrO-SiO2 glass for a sufficient time to obtain qualitative analysis. 30,35 Moreover, it has been suggested that the change in glass X-ray intensities can be decreased by translating the specimen at a constant rate under a stationary beam.31











Presented here is a method for obtaining reliable compositional analyses for a variety of soda-lime-silica glasses by utilizing electron beam enlargement and specimen translation. Because of its macroscopic nature, this method is especially useful for analysis of semi-rough surfaces obtained by abrading or corroding glass specimens.37 This technique is used to evaluate the earlystage corrosion behavior of three soda-lime-silica glasses.


Experimental

Several glasses were prepared by melting mixtures of reagent grades Na2CO3, CaCO3, and 5.0 pm Min-u-sil* at 1400-15000C for 24 hour (h) in a Pt crucible. This melting procedure was previously found to be satisfactory for batch homogenization and small bubble elimination. The molten glass was pressed between two graphite plates to a thickness of 3/8 in. and annealed at 4500C for

4 h under prevailing atmospheric conditions. Several investigators have noted the effect of the annealing atmosphere on the durability of soda-lime-silica glasses.3' 38, 39, 40 McVay and Farnum38 have reported that the affected region of glass extends = 200 Pm into



*Pennsylvania Glass and Sand Company, Pittsburgh, Pa.











the glass surface for annealing conditions similar to the one used for this investigation. Thus, specimens 3/4 in. square were polished through 600 grit with SiC paper, removing at least 1000 pm, before quantitative microprobe analysis was performed. The polishing also eliminated thin films and compositional variations which might have formed at room temperature due to reaction with atmospheric moisture. 17 In order to evaluate the effects of surface roughness on X-ray intensity, similar specimens were ground to 120, 240, 320, and 400 grits. (The surface roughness decreases as the grit number increases.) Dilmore41 et al. have shown that a specimen polished to 120 grit has approximately 1.5 times the surface area as the same specimen polished to 600 grit. This is due to the large surface perturbation created by the coarse polishing. The prepared samples were maintained in a desicator until examined with the EMP.*

The specimens were coated by vacuum evaporation
0
with approximately 100 A of C and electrically connected to Al disks with Ag paint. The disks were placed on a stage inside the EMP vacuum chamber, which could be electrically driven perpendicularly to the electron beam



*Electron Probe X-ray Microanalyzer, Model Ms-64, Acton Labs, Acton, Mass.











at specified rates between 0 pm/min and 150 pm/min. The electron beam diameter was varied between 1 Pm and 200 pm for various specimen speeds while maintaining a constant beam current of l0-7 amps and voltage difference of 20 kv. An Ar-methane flow proportional detector with a KAP crystal was used to measure the Na and Si Ka radiation intensities. A PET crystal was employed in obtaining Ca Ka radiation intensities.

Corroded specimens were prepared from glasses containing 10, 20, and 30 per cent* Na20, each containing 10 per cent CaO, with the remaining constituent being SiO2. The samples were ground to 600 grit before corroding in a Teflon R ** cell at 1000C for up to 12 h. A major feature of the corrosion cell-assembly is that it provides a constant surface area to solution volume ratio during corrosion. The cell assembly and corrosion technique are described more fully in Chapter IV.


Results and Discussion-Bulk Glasses

Figure 3 illustrates the effect of allowing the electron beam to remain in a stationary position on a



*Per cent will refer to mole per cent throughout this dissertation unless otherwise designated.
**E. I. Dupont de Nemours & Co.,Inc.,Wilmington, Del.









50
20 Na2O-lOCaO- 70SiO2 Stationary Beam


40
40 200pm n








30



0




10
' ' lpm25prm

12 3 4 5 6 Time (min.) Figure 3. Effect of Stationary Electron Beam on Na X-ray Intensity.











20 Na20-lOCaO-7OSi02* glass. Every data point represents an average of five measurements on different areas of the sample. The intensities in counts per second (cps) were obtained by placing the sample under the beam and taking fixed time counts (10 s each) at 1 min intervals. The X-ray intensities given for 0 time correspond to those obtained during the first 10 s of electron beam impinge-ment. Although the Na X-ray intensity is not time independent for any beam size investigated, the rate of intensity decrease is much less for the larger beam sizes. Figure 4 is a 10 s cross section of Figure 3, illustrating the effects of beam diameter on the Na X-ray intensity. The X-ray intensity is significantly reduced as the beam size is decreased below 100 Pm. Furthermore, the data suggest that a quantitative analysis is possible using a beam diameter > 100 Pm if the measurement is made within 10 s after the electron beam impinges upon the sample. Vassamillet and Caldwell34 show that significant changes in the local composition occur when the electron beam is allowed to remain in a stationary position on a sodasilica glass for sufficient times to make a complete analysis (i.e., > 1 min). A similar effect is observed



*The number preceding each constituent represents its mole per cent in the glass, unless otherwise noted.










50
Na
" i///' Stat ionary Beam


40 100 nanoamps
20 soda-101ime-70 silica / Freshly abraded, 63 grit








I.
_ 20



Ci
10I




40 80 120 160 200

Beam Size (pm) Figure 4. Effect of Electron Beam Diameter on Na X-ray Intensity.











for the soda-lime-silica system. The extremely low intensity obtained with the 1 pm beam is probably due to a combination of surface roughness-beam interference and Na ion diffusion from under the beam. In order to minimize compositional analysis errors, the samples were moved at various constant speeds under a stationary 100 pm electron beam. Na X-ray intensity as a function of beam impingement times for various specimen speeds and a 100 pm beam diameter is shown in Figure 5. This graph demonstrates that the variation in X-ray intensity decreases as the specimen speed increases. Using a 150 pm/ min speed, the X-ray intensity becomes almost independent of time and the variation in intensity is within the experimental scatter band normally encountered in EMP analysis. The X-ray intensities of Si and Ca were also independent of time for this beam diameter and specimen velocity.

Other soda-lime-silica compositions with the

Na20 ranging from 10 to 30 per cent and the CaO ranging from 0 to 20 per cent were analyzed using this technique. Figure 6 shows that a 100 pm beam diameter and 150 pm/ min specimen speed were sufficient to give time independent X-ray intensities for all the compositions investigated. Of further importance is the time independent X-ray intensity for the corroded specimen. Although the diffusion rate for Ma in corroded layers has been reported
















































time (min.)


Figure 5.


Na X-ray Intensity as a Function of Electron Beam Impingement Time and Sample Velocity.



















60 lli ,.




50

20-10-70 UA 40

>,, 20-10-70 (31-100�C)
0

E 30' 10-10-80

20


100pm Beam Dia. 10
1501m/min. specimen vel.





0 1 2 4 time [min.] Figure 6. Na X-ray Intensity as a Function of Electron
Beam Impingement Time for Several Glasses.











to be much greater than for the bulk glass, the beam enlargement and specimen speeds employed are sufficient to compensate for this increased diffusion rate. Therefore, this technique is suitable for evaluating the corrosion behavior of glasses where the surface compositions are time dependent. Figure 6 also shows that the replacement of SiO2 with CaO decreases the Na X-ray intensity. This is probably due to the higher mass attenuation of Ca compared to Si, for Na X-rays.


Results and Discussion-Corroded Glasses


Figure 7 presents the Na X-ray intensity as a

function of corrosion for three soda-lime-silica glasses. Since the X-ray intensity is related to the mole per cent of Na present in the glass, the slopes of these curves represent the rates of dealkalization within the volume of analysis. This graph illustrates that the rate of Na diffusion from the glass is proportional to the bulk composition. As the per cent of Na in the glass is increased, the rate of Na in the glass is also increased. The increased dealkalization rates are probably due to both the larger Na concentration gradients encountered in the high soda glasses and the more open structure produced during corrosion. Severe surface roughening and flaking were observed on the glass containing 30 per cent Na20 after 6



























3-40 filmremoved

C
20-10-70
C30- A,.

S.A./V.-.17cmrj1 ~.


20 10=10-80




10


Severe flaking


2 4 6 8 10 12


Corrosion Time [hrs.]

Figure 7. Na X-ray Intensity as a Function of Corrosion
Time for Three Soda-Lime-Silica Glasses.











h of corrosion. The same glass corroded for 12 h produced a flakey surface which was too rough for EMP analysis. After mechanically removing some of the flakes, a Na X-ray intensity was obtained for the underlying surface which was = 2/3 the bulk glass value. Scanning electron micrographs (SEMs)* for these glasses are shown in Figure

8. The cracks and flakes observed in the 30 per cent Na20 glass, not evident immediately after removal from the corrosion solution, are thought to be due to dehydration while in the desiccator. The effect of surface area to solution volume ratio is also illustrated in Figure

7 for the 20 per cent Na20 glass. Decreasing this ratio by a factor of = 4.5 yields no significant difference in the dealkalization rate. This result, suggesting that the rate of dealkalization in the first stage (pH < 9.0) of corrosion is independent of Na concentration in the solution, is consistent with that of Rana and Douglas5'6 11
and El-Shamy and Douglas.

Although not shown graphically, the X-ray intensities for Ca and Si were also monitored as a function of corrosion time for the three glasses. The Ca X-ray intensity remained constant while the Si X-ray intensity



*Cambridge Stereoscan, Kent Cambridge Scientific, Inc., Morton Grove, Ill.















(a)









(b)









(c)











Figure 8. SEMs for Soda-Lime-Silica Glasses Corroded for
12 h at 1O0C: (a) 10 Na2O-l0 CaO, 80 Si02 (6000x) (b) 30 Na 0-10 CaO, 60 SiO2 (55x) (c) 30 Na20-10 CaO, 0 SiO2. Fractured surface
(2000x)











increased in a manner related to the amount of Na depletion for all corrosion times < 12 h. As Na is removed from the glass, increasing quantities of Si are exposed to the electron beam and a corresponding increase in Si X-ray intensity is obtained. The relative constance of the Ca X-ray intensity is thought to be due to two factors:

1. the smaller mass attenuation of Na on Ca

2. the greater effective escape depth of Ca, compared

to Si, from the glass.

Both of these factors would tend to decrease the Ca X-ray sensitivity to Na variations in the glass.

The effect of sample surface roughness on the coefficient of variation in X-ray intensity is shown in Table 1. These measurements were made using a 100 pm beam diameter and sample speed of 150 pm/min. In general, as the surface roughness increases, the variation in X-ray intensity also increases. This beam diameter and specimen speed were sufficient to permit an accurate analysis for a sample polished with 400 grit or finer. IRRS and SEM illustrate that the surface finishes produced from this range of grit sizes are comparable to those encountered for the 10 per cent Na20 and 20 per cent Na20 glasses corroded up to 10 h and for the 30 per cent Na20 corroded for 1 h. The 30 per cent Na20 corroded for 6 h gave a










Table 1. Coefficient of Variation in X-ray Intensity as a Function of Surface
Roughness for a 20 Mole Na20-10 Mole CaO-70 Mole Si02 Glass.




Grit Size % A Intensity (roughness) Na Ca Si


120 12.0 11.0 1.0 240 12.0 2.0 1.4 320 11.0 2.0 1.3 400 2.0 1.0 1.3 600 1.5 1.0 0.5











surface roughness comparable to a 120 grit polish or less with a corresponding large error in X-ray intensity.



Summary

An EMP technique has been developed which provides for an accurate and reproducible compositional analysis of both bulk and corroded soda-lime-silica glass. Application of this technique to several glasses reveals that the rate of dealkalization is dependent on the composition of the glass. Within the depth of analysis (= 1.5 pm), the glass composition containing 30 per cent Na20 exhibits a much greater Na depletion rate than the glasses containing less Na20. This is a result of the greater Na concentration gradient and hence a greater driving force for Na ion diffusion from the glass. The majority of the Na ions are depleted from the analysis volumes of the 30 Na2010 CaO-60 SiO2 during the first hour of corrosion. The low X-ray intensities and the small change in intensities between 1 h and 6 h are due to the finite depth of analysis of the EMP. The high concentration of Na ions remaining in the uncorroded glass is at a depth below the range of the analysis and essentially only the tail of the Na concentration gradient is detected by the EMP.

The glass surface area to solution ratio does

not alter the rate of dealkalization for the two ratios











investigated. This is due to the relatively small change of Na ion concentration in solution for pH < 9.

Sample surface roughness is also an important parameter in EMP analysis. The error in analysis increases as the surface roughness increases due to X-ray scattering. The surfaces of samples containing 30 per cent Na20 are too rough for EMP analysis after 12 h of corrosion due to extensive cracking of the Na-depleted film while samples containing < 20 per cent Na have surfaces comparable to a 400 grit polish or finer. Thus, the latter surfaces provide more reliable EMP analysis.













CHAPTER III

INFRARED REFLECTION SPECTROSCOPY OF CORRODED
SODA-LIME-SILICA GLASSES UTILIZING FREQUENCY SHIFTS

Introduction-A Review of Infrared Reflection Spectroscopy of Glasses

Traditionally, chemical analyses of solutions in

contact with glass powders (large surface area to solution volume ratio) have provided the bulk of glass corrosion data. 1,2,4,5-11,13,16 This method allows accurate chemical analysis of dissolved species, but provides little informition concerning the surface structural changes accompanying dissolution.

Scanning electron microscopy (SEM), electron microprobe analysis (EMP), Auger electron spectroscopy (AES) and infrared reflection spectroscopy (IRRS) are currently being utilized for characterizing glass corrosion.37 It has been demonstrated that the latter is sensitive to boti the chemical and structural changes in the spectral region from 1200 cm-1 to 600 cm- 141-43 Furthermore, minimum specimen preparation, no vacuum requirements and rapid sample analysis make this an attractive tool for on-line quality control.











Sanders et al.23 have developed a method of

quantitative glass compositional analysis based on the proportionality between IRRS peak amplitudes (intensities). In some glass systems, notably the binary silicates and
37
high alkali soda-lime-silica glasses, corrosion can lead to severe surface roughening with a simultaneous decrease in the infrared reflected intensity. Precipitate formation on weathered glasses can also give a decrease in infrared reflected intensity.17 Thus, compositional analysis of corroded glass surfaces using an analysis of peak amplitudes, may yield misleading results.

Several investigators have demonstrated that infrared peak frequency (or wavenumber) shifts can be used for compositional analysis of binary sodium silicate
44-48
glasses. , Other researchers have applied this technique to more complicated systems.49,50 Frequency shifts are a result of structural modifications in the glass, and are not affected by the physical condition of the surface.

The objectives of this chapter are to

1. demonstrate the usefulness of infrared peak frequency shifts for monitoring glass corrosion
2. apply this technique quantitatively to the corrosion of soda-silica and soda-lime-silica glass

systems.











Experimental

Silicate glasses containing 10-30 Na20 and upto 10 per cent CaO were prepared by melting reagent grade carbonates and 50 pm Min-u-sil in covered Pt crucibles between 14001550'C for 24 h. Glass slabs 1/4 in. thick were cast between graphite plates and annealed at 450'C for 4 h. All specimens were polished to 600 grit with SiC paper except those noted. The 600 grit surface provides a common base for data normalization.51

Samples 3/4 in. square of 20 Na20-10 CaO70 SiO2 glasses were corroded in a static aqueous environment by the technique described by Sanders et al.42 Briefly, the samples were sealed onto Teflon R corrosion cells which were filled with deionized water and maintained in a hot water bath at constant temperatures, for various times. The procedure utilized a constant exposure area to solution volume ratio of 0.77 cm-1. Corroded specimens were stored in a desiccator while awaiting analysis.

IRRS spectra* for the freshly abraded (standards) and corroded samples were obtained for the spectral region 1400 cm -l to 600 cm-l using a medium scan rate and 230 angle of incidence. The spectrum for vitreous silica was



*Grating Infrared Spectrophotometer, Model 467, Perkin-Elmer Corp., Norwalk, Conn.











used for adjusting the data acquired over a period of several months.

Samples were prepared for the SEM and EMP analysis
a
by vapor coating with = 100 A of C and electrically connected to Al mounts with Ag paint. Quantitative EMP analysis was performed by traversing a 100 pm diameter electron beam at 150 pm/min parallel to the sample surface to 24
minimize alkali diffusion.


Results and Discussion.

Technique Development
The variation in infrared reflection intensity

with surface roughness is shown in Figure 9. The intensity increases significantly as the sample surface is polished with finer grit paper. A maximum reflected intensity of = 25 per cent at 1045 cm- l is obtained for the 20 Na20

-10 CaO-70 Si02 glass when the sample is polished with 400 grit or finer SiC paper. It should be noted that the locations of the peak maxima shown in this figure are not altered by the surface condition. Figure 10 presents an SEM of this glass polished to 120 grit and exposed to a humid atmosphere for 24 h at 25�C. Surface roughness due to both polishing scratches and precipitate formation can be observed. EMP analysis revealed that the surface was. rich in both Na and Ca. The precipitate is probably the carbonates of these elements.19' 20




















400)


20 0





1400 1200 1000 800 600


Wavenumber [cmR Figure 9. Effect of Surface Roughness on the Infrared Reflected Intensity.



















































Figure 10.


SEM of a 20 Na2O-10 CaO-70 Si02 Glass Polished to 120 Grit and Exposed to Atmoshere (= 70 Per Cent R.H.) for 24 h at 250 C (1300x).











The peak maximum in Figure 11 occurring at 1110 cm-1 for freshly abraded (f.a.) silica has been assigned

4
to the Si-O stretch vibration within the SiO4 tta hedra.45' 46 The exact location of this peak varies slightly with different investigators and Hanna52 has suggested that its position is dependent upon the presence of nonbridging oxygen atoms necessary to accomodate the random orientation of the Si044 groups. This peak is shifted to lower wavenumbers and the intensity is simultaneously decreased as Na20 is added to the SiO2 network.

A satellite peak at = 960 cm-l appears as a

shoulder on the Si-O stretch peak for the glass containing 20 per cent Na20 (Figure 11). This peak has been assigned to a 42
Si-O nonbriding oxygen stretch in an alkali environment. The exact position of the peak maximum is dependent on both the alkali species and its concentration in the

glass.

The infrared spectra of a 20 Na20- 80 SiO2 glass corroded at 1000C for various times are also shown in Figure 11. The Si-O stretch maximum (= 1070 cm-1 for the freshly abraded sample) shifts to higher wavenumbers and increases in intensity during the first hour of corrosion. The Si-O nonbridging maximum shifts to lower wavenumbers and disappears after 1 h of corrosion. These peak shifts


















































1400 1200 1000 800


Wavenumber


[cmil]


Figure 11.


Infrared Spectra for Glass and the Same G1 Times.


a Freshly Abraded (f.a.) ass Corroded for Various











have been associated with diffusion of Na ions from the glass into solution.37

It is interesting that the amplitude of the Si-O nonbridging oxygen stretch maximum decreases to zero at the frequency corresponding to the lowest wavenumber (820 cm1) in the original distribution of the freshly abraded sample. If this peak is representative of nonbriding oxygen atoms associated (to various degrees) with Na ion, then the maximum corresponds to the frequency where the largest number of Si-O--Na occur with a particular

0..Na association strength. Hanna and Su53 have suggested that the 0.Na association weakens the Si-O terminal bond and shifts its maximum to lower wavenumbers. Thus, the Na ions with the weakest O--Na association will first diffuse from the glass and the Na ions with the strongest association will diffuse last from the glass. This diffusion sequence should shift the Si-O nonbridgingmaximum to lower frequencies, decrease the width of frequency distribution and decrease the maximum intensity as the corrosion time increases. These suggested spectra changes are consistent with the experimental results (Figure 11).

The Si-O peak maximum continues to shift to higher wavenumbers for up to 3 h of corrosion suggesting the continuous development of a SiO2 rich layer. However, the amplitude decreases significantly between I h and 3 h and











this could be interpreted as a decrease in SiO2 if the surface roughness contribution is not considered. Since the amplitude of this peak is directly related to the SiO2 concentration and inversely proportional to the surface roughness, peak heights cannot be used for any stage of quantitative corrosion analysis without a separate evaluation of surface roughness. Apparently, the surface roughness becomes the dominant factor after 1 h or corrosion for the 20 Na2O-80 SiO2 glass. A SEM illustrating surface roughening on this glass after 2.5 h at 1000C is shown in Figure 12a. Although not as severe, surface roughening can also be observed on the 20 Na20-10 CaO-70 SiO2 glass corroded for 12 h at 100'C (Figure 12b). However, roughening does not become the controlling factor to amplitude until after 5 days (d) of corrosion (Figure 20).

The Si-O stretch maximum was used for quantitative analysis due to its greater sensitivity to compositional variations below 20 per cent Na20. Figures 13 and 14 illustrate the infrared reflection spectra for the Na2O-SiO2 and Na20-10 CaO-SiO2 glasses freshly abraded to 600 grit. Figure 15 presents in graphical form the Si-O stretch maxima wavenumbers as functions of per cent Na20 in the glass for these two figures. The graph for the binary glasses is linear up to 22 per cent Na20 in agreement with the results of other investigators.45'48 For a given concentration of




















(a)




















(b)






Figure 12. SEMs of Corroded Glasses for (a) 20 Na20-80
Si02 2.5 h-100�C (1000x) (b) 20 Na20-10 CaO70 SiO2, 12 h-lO0�C (1200x).











































1400


Figure 13.


1200 1000 800


Wavenumber [c m-1]

Infrared Spectra for Several Na2O-SiO2 Glasses.


















0,
�10Mole% Na20 20- 15
2503




1400 1200 1000 Soo

Wavenumber [cm"]
Figure 14. Infrared Spectra for Several Na2O-CaO-SiO2 Glasses.





















0
.E 1100

0

E \0
Z51080- \




> 106010401
0 10 20 30 Mole % Na20



Figure 15. Calibration Curve for the Si-O Stretch Peak
Maxima of Na2O-SiO2 and Na20-10 CaO-SiO2
Glasses.











Na 20, the replacement of SiO2 with CaO shifts the Si-O stretch peak to lower wavenumbers. A reversal occurs in the Si-O stretch peak at = 22 per cent Na2 0 for the binary glass and at = 20 per cent for the glass containing 10 CaO. Similar trends are observed in the data of other investigators and are thought to be related to the critical nonbridging oxygen concentration (O/Si > 2.5).47 Since most commercial glasses contain < 20 per cent Na20, problems which arise from reversals in peak location can be avoided in this approach to quantitative analysis of corrosion.

The variations in wavenumber of the Si-O stretch maximum with corrosion time and temperature for the 20 Na20-80 SiO2 and 20 Na20-10 CaO-70 Si02 glasses are illustrated in Figure 16. The peak maxima increased from 1070 cm-1 for the freshly abraded surface to 1100 cm-I after corroding the binary glass at 50'C for 24 h and at l0�0C for 3 h (Figure 16a). The maxima increased from 1045 cm-I to 1090 cm-1 after corroding the ternary glass for 9 d at 1000C (Figure 16b).

The composition versus corrosion time graphs shown in Figure 17 were constructed from Figures 15 and 16 by assuming that the CaO to Si02 ratio remains constant in the ternary glass during corrosion. This assumption is based on solution analysis showing that the dissolution of Ca from the ternary glass is nearly stoichiometric with
















Figure 16. Wavenumber of Si-O Stretch Maxima as a Function of Corrosion
Time for (a) 20 Na20-80 SiO2 Glass (b) 20 Na20-10 CaO70 SiO2 Glass.












1120


-7 1000C o 1100 50 . 30 O

Z51080
fa.
E .20mole % Na20-8Omole % SiO2 S1060 S. A./V.-0.77cm1'


10400

0 4 8 12 16 20 24 Corrosion Time [hrs.]


(a)











1120


a


0



L
E


0 2 4 6 8


Corrosion


Time [days]


Figure 16 continued.


( b)





















Figure 17. Na20 Concentration as a Function of Corrosion Time for.
(ai 20 Na20-80 SiO2 Glass (b) 20 Na2O-lO CaO-70 SiO2
Glass.











20
f.a.
20mole % Na20-80mole % SiO2

S.A. /V.- 0.77cm"1 O 15
N
a
z

. 1 0

0
CL E 50C
0




01
0 4 8 12 16 20 24

Corrosion Time [hrs.]
(a)


















0.

z

C
0

0

E
0
U


2 4 6 8


Corrosion
(b)


Time [days]


Figure 17 continued











respect to the dissolution of Si.37 The graphs in Figure 17a show that the rate of Na ion depletion from the binary glass increases as the corrosion solution temperature increases. For all three temperatures there is an initial rapid rate of Na ion depletion from the glass followed by a gradual decrease in the depletion rate. When the Na20 concentration in the glass is plotted as a function of the square root of corrosion time, two linear regions are obtained for each temperature (Figure 18a). This suggests a change in the diffusion controlling mechanism for Na ions leaving the glass. The change in mechanisms occurs at pH > 9 for the three temperatures investigated. Douglas and El-Shamy9 have found similar changes in the dealkalization rate and interpreted these decreases in terms of H ion availability at the surface of the glass. For pH > 9, the H ions in solution are associated with only a small fraction of the nonbridging oxygen atoms on the glass surface (equation 5). These remaining sites become occupied by Na ions, which effectively decreases the Na ion concentration gradient between the glass and solution. This results in a decrease in the Na ion diffusion rate from the glass. Infrared analysis indicates there is = 3.5 per cent Na2 0 remaining within the depth of IRRS analysis of the 3 h1000C sample. The EMP analysis of the same sample yielded 2 per cent Na20 based on comparison with a NaCl standard.

























Figure 18. Na20 Concentration as a Function of the Square
Root of Corrosion Time for: (a) 20 Na2O-80
Si02 Glass (b) 20 Na20-l0 CaO-70 SiO2,
10 Na20--10 CaO-80 SiO2 Glasses.















20


0


zO

0
,z..-10 -00C ...
0 pH)0.-10.5 E O�C

5


0.
0 1 2 3 4 5

[Corrosion time- hrs.]Y

(a)


























N

C
0


0
0~
E
0
U


0 4 8 12


[Corrosion time- hrs.]Y

(b)


Figure 18 continued











Figure 17b illustrates that when 10 per cent CaO is substituted for SiO2 in the glass the Na ion depletion rate is decreased by a factor of = 3. Again, two linear regions are observed in the plot of Na20 concentration versus the square root of corrosion time for the 20 Na2 0

-10 CaO-70 SiO2 glass. Although the corrosion time at which the change in mechanisms occurs is greater for the ternary than for the binary glass at the same temperature, the pH at which the change in mechanisms occurs is approximately the same for both glasses. Figure 18b also shows that the initial rate of Na ion diffusion from the 10 Na 2O10 CaO-80 SiO2 glass is very slow and approximately equal to that of the second mechanism in the ternary glass containing 20 per cent Na20. A change in the diffusion rates is not observed in this glass even after 9 d. Infrared analysis of the 20 Na2 0-10 CaO-70 SiO2 glass corroded for

9 d at 1000C indicated that 2 per cent Na20 remained within the depth of analysis. The EMP analysis for this sample also yielded = 2 per cent Na20.



Summary

The infrared frequency maxima associated with various vibrational modes in sodium silicate glasses are sensitive to compositional changes but are unaffected by the surface condition. The Si-O stretch maxima shift linearly with











Na20 content up to 22 per cent Na20. Similar changes are found for sodium silicate glasses containing 10 per cent CaO. The same frequency shifts are observed when these glasses are corroded, and are attributed to Na ion diffusion from the glass into solution. The Si-O stretch peak has been used to quantitatively determine the per cent Na20 remaining in these glasses as a function of time and temperature. The results show that decreasing the solution temperature from IO0�C to 300C greatly decreases the rate of Na ion depletion from the binary glass. There is a similar decrease in Na ion depletion rate by substituting 10 per cent CaO for SiO2 in the glass. A change in diffusion mechanisms is observed at pH > 9 for all the glasses investigated except the 10 Na2Ol0 CaO-80 SiO2 glass. The surface compositions determined from infrared analysis are in satisfactory agreement with the EMP analysis.

Although quantitative analysis of multicomponent commercial glasses appears to be possible, the determination of calibration curves is more complicated. However, qualitative analysis of commercial glass durability using this technique is superior to peak height analysis, especially in weathering studies. This will be demonstrated in Chapter V.













CHAPTER IV
APPLICATION OF THE ELECTRON MICROPROBE,
INFRARED REFLECTION SPECTROSCOPY AND OTHER TECHNIQUES FOR EVALUATING AQUEOUS CORROSION
OF SODA-LIME-SILICA GLASSES

Introduction-Review of Corrosion Evaluation Techniques

Evaluation of glass aqueous corrosion requires specific and accurate information concerning the accompanying surface chemical and structural changes. The existence of surface films, concentration gradients, and precipitates with compositions different from those of the bulk glass influences its mechanical, electrical and overall interfacial behavior. Previous studies have relied primarily on glass powder-solution analysis techniques which, at best, provide only an indirect measure of corrosion. Furthermore, the geometrical parameters necessary for precise corrosion evaluation of powders have not been incorporated into the solution analysis equations.

Several new techniques are presently available

for characterization of glass corrosion. These tools are attractive because they provide a direct measure of corrosion extent on the "whole item" or bulk surface. Three techniques for evaluating glass durability employing EMP, IRRS and SEM are described in the preceding chapters. A











fourth method, AES, has been demonstrated by previous investigators to be an excellent tool for glass surface analysis.25'54 Furthermore, Pantano et al.26 have shown that AES coupled with ion milling is a useful method for determining chemical profiles in glass corrosion films. These tools, together with the more sensitive solution analysis equipment now available, can yield a thorough characterization of glass durability. Moreover, when applied throughout a systematic analysis of a series of glass compositions and environments, optimum batch compositions and thermal processing requirements can be determined.

Aqueous corrosion of soda-lime-silica glasses has previously been studied by numerous researchers employing solution analysis.1,2,5,6,55-59 The objective here is to combine the five techniques discussed above, EMP, IRRS, SEM, AES and solution analysis, into a packaged system for evaluating aqueous corrosion of these glasses. The data obtained from each techniqueare interpreted singularly and in comparison with those obtained using the other techniques. This permits, for each technique, an evaluation of the type of information, and the depth within the sample from which this information is obtained.

The corrosion characteristics of a soda-silica

and a soda-lime-silica commercial glass are evaluated and











the results are compared to those of the soda-lime-silica glasses. This provides information concerning the role of CaO and other commercial additives in glass corrosion.


ExperimentalCorrosion and Instrument Techniques

The compositions of the glasses investigated are listed in Table 2. The glass melts containing only Na20, CaO and SiO2 were prepared from reagent grades Na2CO3, CaCO3 and 5.0 Pm Min-u-Sil. The mixtures were homogenized in covered Pt crucibles maintained at 1400-15500C for 24 h in an electric muffle furnace. The commercial composition was prepared from raw materials commonly used in the glass industry.* The glass was melted in a refractory clay crucible** at 13500C for 24 h in order to simulate, as closely as possible, those procedures used for producing glass containers. The strain point of this glass was 518'C and the annealing point was 559�C.***

Slabs of all the glasses, = 3/8 in. thick, were cast between two graphite plates and specimens 3/4 in. square were diced and used for corrosion studies. The



*Anchor Hocking Glass, Jan. 1975, Jacksonville, Fla.
**DFC, Ceramics, Inc., Canon City, Colo.

***Anchor Hocking Laboratories, Lancaster, Ohio.








Table 2

Glass Compositions Investigated


% Na20 CaO % SiO2 % Al203 % MgO %k20 20 80 10 10 80 10 20 70 20 10 70 20 20 60 30 10 60

*14.5 10 72.5 1.8 0.7 0.5

*Determined by X-ray Spectrochemical Analysis, Siemens X-ray Fluorescence Unit, Anchor Hocking Glass, Lancaster, Ohio.











commercial glass was annealed at 500'C for 6 h and all other glasses were annealed at 450'C for 4 h.

The specimens were stored in a desiccator before corroding in a controlled static environment. Prior to exposure, the samples were polished to 600 grit with SiC paper, to remove any surface films which might have formed during casting or from atmospheric reactions. The corrosion cell 42 consisted of a Teflon R cube with a cavity = 3/8 in. in diameter and = 3/8 in. deep, filled with distilled, deionized water (pH = 6.5). These cell dimensions provide an effective exposure surface area to solution volume ratio of 0.77 cm-1. The samples were sealed over the cavity with Teflon R gaskets and the assembly was immersed in a hot water bath maintained at 1000C for various times, up to 20 d. After removal from the corrosion cell, the corroded side of the sample was placed on tissue paper and allowed to dry before being placed in the desiccator.

Freshly abraded and corroded specimens were submitted to IRRS. The samples were scanned from 1400 cm-l to 400 cm-l using a medium scan rate. A freshly abraded fused quartz sample was scanned along with the test specimen to determine the effect of the various glass additives on the Si-O stretching peak.











The EMP studies were conducted using the method

described by Clark et al. 24 for soda-lime-silica glasses. An electron beam 100 pm in diameter was used to scan parallel to the glass surface at 150 pm/min. A fixed-time (10 s) count was taken at 1 min intervals for a period of 5 min to obtain the average Ka X-ray intensities for Na, Ca, Al and Si. A KAP crystal was used for Na, Al and Si detection and a PET crystal was used for Ca detection, all in conjunction with an Ar-methane proportional detector.

Solution measurements were made on the diluted

solutions prepared from the corrosion cells. The concentrations of Si, Ca and Al were determined by atomic absorption (AA),* and the concentrations of Na and K were determined by atomic emission (AE)** Indicator paper*** was used to check the pH of these solutions.

The AES**** was performed on the glass specimens using the technique described by Pantano et al.26



*Atomic Absorption Spectrophotometer, Model 303, Perkin-Elmer Corp., Norwalk, Conn.
**Atomic Emission Spectrophotometer, Model B, Beckman Instruments, Fullerton, Calif.
***Gallard-Schlesinger Chemical Mfg. Corp., Carle Place, N.Y.
****Cylindrical mirror analyzer, Physical Electronics Industries, Inc., Edina, Minn.











Chemical profiles were determined for Si (1619 eV), Ca (291 eV), and 0 (510 eV) at various corrosion times using an Ar ion-milling apparatus to remove the surface material while simultaneously analyzing with AES. The rate of material removed by this particular ion miller was found
0
to be = 30 A/min for most oxides previously investigated. Ion milling was continued until the Si to 0 ratio became constant, at which point it was assumed that the bulk glass had been reached.


Results and Discussion
of Aqueous Corrosion Data for the Various Techniques

The infrared reflection spectra for the 20 Na2080 SiO2 and 20 Na2Ol0 CaO-70 SiO2 glasses after identical corrosion conditions are compared in Figure 19. The peak occurring near 1050 cm- (Figure 19a) for the freshly abraded (f.a.) glass is assigned to a symmetric SiO-stretching vibration in an alkali environment.23 The other peak, appearing on the shoulder of the first peak at 950 cm 1, is assigned to the Si-O nonbridging oxygen (NS). The overlap of these peaks is caused by coupling interactions between these two vibrations. After corrosion of the binary glass for 1 h, there was a large increase in the height and a shift to higher wavenumbers of the Si-O stretching peak whereas the NS peak height























Figure 19. Infrared Reflection Spectra for Corroded
Glass Specimens: (a) 20 Na20-80 Si02 (b)
20 Na2 0-10 CaO-70 SiO2.









































Wavenumber (cm")


(a)























20 mo 70 mol Corroded I 8o S.A./ V.-0.




60

Si-O NS
f.a.
40



~12 lhrs. 20- 2hs 48 hrs



1400 1200 1000

Wavenumber (cm')
(b)


e% No2O- lOmole % CaO 0% SiO2

Od'C- static aqueous soln. T7cm-I


goo


600


Figure 19 continued











decreased and shifted to lower wavenumbers during this same period. This behavior indicates the formation of a SiO2 rich layer on the glass surface.14 Longer corrosion times (> 3 h) indicate dissolution and eventually complete breakdown of the SiO2 layer causing surface roughening as illustrated by the 12 h spectrum shown earlier in Figure 11. In contrast, the dealkalization of the soda-lime-silica glass with an equivalent Na20 content proceeds at a much slower rate as evidenced by the presence of a strong NS peak after 12 h of corrosion (Figure 19b).

Figure 19 also illustrates the effect on the Si-O and NS vibrations of adding CaO to the glass. The IRRS maxima for the freshly abraded ternary glass are shifted to slightly lower wavenumbers, and the height of the coupled region is greater than for the binary glass. The rate of change in wavenumber for the maxima is slower during corrosion for the ternary glass than for the binary glass during corrosion.

Reflection spectra corresponding to the later stage of corrosion for the soda-lime-silica glass are shown in Figure 20. The gradual shifting of the Si-O peak to higher wavenumbers indicates that the SiO2 layer was still developing after 9 d of exposure. This peak did not attain the magnitude that it did for the soda-silica glass even though the NS peak completely vanished after 9 d of corrosion. This apparently results from the presence of Ca





































Wavenumber [cm"]


Figure 20. Infrared Reflection Spectra for the Later Stage of Corrosion
for the 20 Na20- 10 CaO-70 SiO2 Glass.











in the Na depleted zone, as indicated by the comparison of the 2 d sample and the vitreous silica spectra in Figure 20 to the 1 h spectrum in Figure 19a. Thus, the Na depleted layer on the ternary glass is rich in both SiO2 and CaO. The decrease in the magnitude of the Si-O peak between 5 d and 9 d is probably the result of surface roughening. However, roughening does not alter the wavenumber of the peak maxima. These spectra show that the surface breakdown and roughening is not as extensive in the soda-lime-silica glass after 9 d as it is in the sodasilica glass with an equivalent quantity of Na20, corroded for 12 h. SEMs reveal the extent of surface damage to each of these glasses in Figure 21.

The peaks occurring at = 450 cm 1 in Figure 20 result from the rocking vibrations of the Si-O tetrahedra, and are not as sensitive to composition changes as the stretching vibrations.

Infrared reflection spectra for a variety of freshly abraded and corroded soda-lime-silica glasses are shown in Figures 22-25. The NS peak is not observed for the 10 Na20

-10 CaO-80 SiO2 glass, but the small Si-O peak shift after 9 d of corrosion suggests some dealkalization (Figure 22). Figure 18b shows that the 5 per cent Na20 remains in the glass surface (within a depth of 0.5 pm) after this period of time. Increasing the quantity of CaO














(a)











(b)










(c)











Figure 21. SEMs of Corroded Glass Surfaces: (a) 20 Na20
-80 SiO9, 12-100�C (lO00x) (b) 20 Na2010 CaO-70 SiO2, 9 d-lO0C (1000x) (c) Commercial Glass, 20 d-1000C (2500x).









80


60-


10mole% Na20 -10 mole% C aO-80mole% SiO2 Corroded; 100' C Static Aqueous Solution SA./V-0.77cm1


2 days
9 days
5 days


401


201


1300


1100


900


700


Wavenumber [cm"l] Fioure 22. Infrared Reflection Spectra for Freshly Abraded and Corroded 10 Na2 0-10
CaO-80 SiO2 Glass.


v v









80


60.

1 hr.
12hrs;2days "
c: 40- 3hrs; 9days/ 1,0 mi n,
400




20





1300 1100 900 700

Wavenumber [cm"1] Figure 23. Infrared Reflection Spectra for Freshly Abraded and Corroded 10 Na20
-20 CaO-70 SiO2 Glass.



















40m lhr. 2hrs.
rlOmin. fL a . ZL:L- f. a .



20





1300 1100 900 700 Wavenumber [cm-1]
Figure 24. Infrared Reflection Spectra for Freshly Abraded and Corroded 20 NaO020 CaO-60 SiO2 Glass.









80


1300 1100 900 700 Wavenumber [cm-l] Fiqure 25. Infrared Reflection Spectra for Freshly Abraded and Corroded 30 Na 0
-10 CaO-60 SiO2 Glass.











to 20 per cent in the 10 per cent Na20 glass does not significantly affect the rate of dealkalization even though a NS peak is observable in this glass (Figure 23). However, it does significantly increase the extent of surface roughening.

Figures 24 and 25 illustrate the spectra for the 20 Na2O-20 CaO-60 SiO2 and 30 Na2 0-lO CaO 60 SiO2 glasses in which total modifier ion concentration is the equivalent. Although the spectra for both freshly abraded specimens are nearly identicial, the corrosion behavior of the glasses is clearly different. Analysis of the Si-O peak amplitude suggests a SiO2-rich layer develops in the 30 per cent Na20 glass after 10 min and continues to develop for 3 h. After

6 h the NS peak disappears and the Si-O peak begins to decrease. The Si-O peak vanishes after 12 h suggesting severe surface roughening due to the breakdown of the SiO2-rich film. The SEM of this surface is shown in Figure 8b. In contrast, the amplitudes of the Si-O and NS peaks for the 20 Na20-20 CaO-60 SiO2 glass remained constant during the first 12 h of corrosion indicating no dealkalization or SiO2 film development. Between 2 d and 9 d of corrosion the entire spectra decrease in amplitude significantly, but with both the Si-O and NS peaks remaining discernible. The decrease in amplitude can be interpreted either as a surface roughening phenomena or redeposition of a new phase on the surface of the glass. The peak shifts that occur during the











corrosion of both the 30 Na20-10 CaO-60 SiO2 and 20 Na 20

-20 CaO-60 SiO2 glass are complicated and not as easily interpreted as the peak shifts in the glasses containing < 20 per cent Na20 and < 10 per cent CaO glasses. IRRS peak reversals occur in noncorroded glasses in this compositional range (Figure 15).

Several general conclusions can be made from

Figures 19, 22-25. The NS peak becomes more discernible from the Si-O peak as the total percentage of Na20 and CaO increases in the glass. Smaller quantities of Na20 than CaO are required to obtain equivalent Si-O and NS peak shifts and peak separation. In general, the NS peak decreases and shifts to lower wavenumbers while the Si-O peak increases in wavenumber during the first stage (dealkalization) of corrosion. During the later stage (> 2d)of corrosion, the Si-O peak decreases, suggesting surface roughening. There is no significant difference in the corrosion behavior of the 10 per cent Na20 glasses containing < 20 per cent CaO. There is a definite improvement in the corrosion behavior of the 20 per cent Na20 glass when 10 per cent CaO is added. However, 20 per cent CaO reduces the corrosion resistance of the 20 per cent Na20 glass.

Figure 26 presents the infrared reflection spectra for the commercial glass corroded at 1000C. The trends observed in these spectra are relatively easy to interpret since the total modifier in content is < 30 per cent



































Wavenumber [cm"1]


Figure 26. Infrared Reflection Spectra for Freshly Abraded and Corroded Commercial Glass.












i.e. no peak reversals). A gradual decline in the NS peak is observed during the 20 d of corrosion with a corresponding shift in the Si-O peak to slightly higher wavenumbers. These observations suggest a much reduced dealkalization rate and SiO2 film development than for the soda-silica and soda-lime-silica glasses. The relative constancy of the Si-O peak indicates very little surface roughening after 20 d of corrosion. Figure 27 illustrates the changes in the infrared reflection spectra when CaO, Al203 and other constituents are added to a glass containing a constant Na20 to SiO2 ratio. The Si-O peak is shifted to lower wavenumbers as the modifier ion concentration is increased. Figures 19 and 22-27 illustrate the spectral changes, which are indicative of structural variations, as the quantity and type of modifier ions are varied in the glass. In general, the corrosion resistance increases as the Si-O wavenumber (f.a.) decreases for samples exhibiting poorly defined NS peaks, whereas most glasses with well defined NS peaks exhibit inferior corrosion resistance.

Solution data presented in Table 3 for the 20 Na2O080 SiO2 and 20 Na2Ol0 CaO-70 SiO2 glasses show that the pH and solution ion concentrations increase at a more rapid rate for the binary glass. The results of Wang and Tooley2,60 show that the major ions in solution


























C
20




1200 lw goo 600 400 Wavenumber [cm"]



Figure 27. Infrared Reflection Spectra for Freshly Abraded Soda-Silica, Soda-LimeSilica and Commercial Glasses with a Constant Na20 in SiO2 ratio.




Full Text

PAGE 1

A DURABILITY EVALUATION OF SODA-LIME-SILICA GLASSES USING ELECTRON MICROPROBE ANALYSIS, INFRARED REFLECTION SPECTROSCOPY AND OTHER TECHNIQUES By DAVID EDWARD CLARK A IN DISSERTATION PRESENTED TO THE GRADUATE OF THE UNIVERSITY OF FLORIDA PARTIAL FULFILLMENT OF THE REQUIREMENTS DEGREE OF DOCTOR OF PHILOSOPHY COUNCIL FOR THE UNIVERSITY OF FLORIDA 1976

PAGE 2

7 Vofi Vagmy

PAGE 3

ACKNOWLEDGEMENTS The author gratefully recognizes the expert guidance of his committee members and wishes to especially acknowledge the valuable research experience gained through his association with Dr. Larry Hench. He also thanks Morris Dilmore, Ed Ethridge and Carlo Pantano for their assistance in data interpretation, Teresa Ne Smith for her well executed drawings, and Wayne Acree for his invaluable electron microprobe assistance. With deepest gratitude, he acknowledges the editing assistance and patient support of his wife. Sue. 1 1 i

PAGE 4

TABLE OF CONTENTS Page ACKNOWLEDGEMENTS iii LIST OF TABLES v1 LIST OF FIGURES vii ABSTRACT . CHAPTERS I INTRODUCTION ] General Discussion -j Summary of Accomplishments ..I.'!!!!!.*!!.’!!.'!! 3 Glass Durability Revi ew— Aqueous Corrosion .. 4 Glass Durability Review— Weathering 14 Research Objectives II A TECHNIQUE FOR ELECTRON MICROPROBE ANALYSIS OF CORRODED SODA-LIME-SILICA GLASSES 20 Introduction — A Discussion of the Problem ... 20 Experimental 22 Results and Di scussion — Bui k Glasses..'.*.’.'.'.*.’.’ 24 Results and Di scuss i on — Corroded Glasses 31 Summary 37 III INFRARED REFLECTION SPECTROSCOPY OF CORRODED SODA-LIME-SILICA GLASSES UTILIZING FREQUENCY SHIFTS 39 Introduction A Review of Infrared Reflection Spectroscopy of Glasses 39 Experi mental ' ‘ * 41 Results and Discussion — Technique Development 42 S umma ry !.*!.'.*!!.*.'!.*!!.'!'* 64 i v

PAGE 5

Page IV APPLICATION OF THE ELECTRON MICROPROBE, INFRARED REFLECTION SPECTROSCOPY AND OTHER TECHNIQUES FOR EVALUATING AQUEOUS CORROSION OF SODA-LIME-SILICA GLASSES Introduction — Review of Corrosion Evaluation Techniques Experimental — Corrosion and Instrument Techniques Results and Discussion of Aqueous Corrosion Data for the Various Techniques Summary V APPLICATION OF THE ELECTRON MICROPROBE, INFRARED REFLECTION SPECTROSCOPY AND OTHER TECHNIQUES FOR EVALUATING THE WEATHERING BEHAVIOR OF SODA-LIME-SILICA GLASSES Introduction — Review of Weathering and Comparison with Aqueous Corrosion Experimental — Weathering Parameters Results and Discussion of Weathering Data for the Various Techniques Summary VI SUMMARY AND CONCLUSIONS Glass Durability Evaluation Techniques !! REFERENCES BIOGRAPHICAL SKETCH 66 66 68 72 110 1 1 3 1 1 3 115 1 1 7 133 140 140 146 140 156 v

PAGE 6

LIST OF TABLES X_ a b 1 , e Page s i ty as a Function of Surface Roughness for a 20 Mole Na 9 0 — 10 Mole CaO — 70 Mol e S i 0 2 G1 ass . 6 36 2 Glass Compositions Investigated. 69 3 Solution Data for the 20 Na20 — 80 S i 0 2 and 20 Na20— 10 CaO — 70 S i 02 Glasses Corroded in Static Aqueous Solution at 1 00°C. gg 4 Solution Data for Various Soda-LimeSilica Glasses Corroded in Static Aqueous Solution at 100°C. 93 5 Solution Data for the Commercial Glass Corroded in Static Aqueous Solution at 100°C. 95 6 Reaction Rate Constants of the Glasses Calculated from Equation (1) Using 100°C Solution Data. 97 vi

PAGE 7

LIST OF FIGURES Figure Page 1 Meehan isms of Glass Corros ion. 5 2 Sampling Depths for Various Glass Durability Analysis Techniques. 19 3 Effect of Stationary Electron Beam on Na X-ray Intensity. 25 4 Effect of Electron Beam Diameter on Na X-ray Intensity. 27 5 Na X-ray Intensity as a Function of Electron Beam Impingement Time and Sample Velocity. 29 6 Na X-ray Intensity as a Function of Electron Beam Impingement Time for Several G1 asses . 30 7 Na X-ray Intensity as a Function of Corrosion Time for Three Soda-Lime-Silica Glasses. 32 8 SEMs for Soda-Lime-Silica Glasses Corroded for 12 h at 100°C. 34 9 Effect of Surface Roughness on the Infrared Reflected Intensity. 43 10 SEM of a 20 Na 2 0 — 10 CaO — 70 Si0 ? Glass Polished to 120 Grit and Exposed to Atmosphere (70 Per Cent R.H.) for 24 h at 25°C (1 300x ) . 44 11 Infrared Spectra for a Freshly Abraded (f.a.) Glass and the Same Glass Corroded for Various Times. 46 v i i

PAGE 8

Figur e Page 12 SEMs of Corroded Glasses. 49 1 3 Infrared Spectra for Several Na 9 0 — Si Op G1 asses . L 50 14 Infrared Spectra for Several Na 9 0 — CaO — Si02 Glasses. ^ 51 15 Calibration Curve for the Si-0 Stretch Peak Maxima of Na 9 0 — Si0 9 and Na 9 0 — 10 CaO — S i 0 2 Glasses. 52 16 Wavenumber of Si-0 Stretch Maxima as a Function of Corrosion Time. 55 17 Na 2 0 Concentration as a Function of Corrosion T ime. 58 18 Na20 Concentration as a Function of the Square Root of Corrosion Time. 62 19 Infrared Reflection Spectra for Corroded Glass Specimens. 74 20 Infrared Reflection Spectra for the Later Stage of Corrosion for the 20 Na 2 0 — 10 CaO — 70 Si0 2 Glass. 77 21 SEMs of Corroded Glass Surfaces. 79 22 Infrared Reflection Spectra for Freshly Abraded and Corroded 1 0 Na^O — 10 CaO — 80 S i 0 2 Glass. 80 23 Infrared Reflection Spectra for Freshly Abraded and Corroded 10 Na 9 0 — 20 CaO — 70 Si02 Glass. ^ 81 24 Infrared Reflection Spectra for Freshly Abraded and Corroded 30 Na 9 0 — 10 CaO — 60Si02Glass. ^ 82 25 Infrared Reflection Spectra for Freshly Abraded and Corroded 20 Na 9 0 — 20 CaO — 60 S i 0 2 Glass. * 83 v 1 1 1

PAGE 9

Figure Page 26 Infrared Reflection Spectra for Freshly Abraded and Corroded Commercial Glass. 86 27 Infrared Reflection Spectra for Freshly Abraded Soda-Silica, Soda-Lime-Silica and Commerical Glasses with a Constant Na^O to S i 0 2 ratio. 88 28 EMP X-ray Intensities as a Function of Corrosion Time for the 20 Na o 0— 10 CaO — 70 Si0£ glass. 2 98 29 EMP X-ray Intensities as a Function of Aqueous Corrosion Time for the Commercial Glass. 100 30 Auger Signals at Various Depths in the Freshly Abraded 20 Na,0 — 10 CaO — 70 Si0 o Glass. 2 2 102 31 Auger Signals at Various Depths in the 20 N a 2 0 — 10 CaO — 70 S i 0 2 Glass Corroded for 2 d at 100°C. 103 32 Auger Signals at Various Depths in the 20 Na20 — 10 CaO — 70 S i 0 o Glass Corroded for 5 d at 100°C. 104 33 Auger Signals at Various Depths in the 20 Na20 — 10 CaO — 70 Si Op Glass Corroded for 9 d at 100°C. 105 34 Auger Signals for Ca as a Function of Depth for Various Samples Corroded at 100°C. 1 07 35 Auger Signals at Various Depths in the Freshly Abraded Commercial Glass. 108 36 Auger Signals at Various Depths in the Commercial Glass Corroded for 10 d at 100°C. 109 37 Infrared Spectra for Weathered 20 Na o 0— 10 CaO — 70 Si02 Glass. 2 118 38 Infrared Spectra for Weathered 10 Na o 0— 10 CaO — 80 S i 0 2 Glass. i x 119

PAGE 10

LLMie Page 39 Infrared Spectra for Weathered 1 0 Na o 0 — 20Ca0 — 70 Si02 Glass. 2 120 40 Infrared Spectra for Weathered 30 Na.,0 — lOCaO — 60 Si02 Glass, . ^ 121 41 Infrared Spectra for Weathered 70 Na o 0— 2 0 C a 0 — 60 Si02 Glass. ^ 122 42 Na20 Concentration as a Function of the Square Root of Corrosion Time for Weathered Glass. 125 43 Infrared Spectra of Commercial Glass for Type 1 Weathering at 100°C-100 Per Cent R * H 127 44 Infrared Spectra of Commercial Glass for Type 1 Weathering at 50°C-100 Per Cent R.H. 128 45 Infrared Spectra of Commercial Glass for Type 2 Weathering at 100°C-100 Per Cent R.H. 1 29 46 Infrared Spectra of Commercial Glass for Type 2 Weathering at 50°C-100 Per Cent R.H. 130 47 Infrared Spectra of Commerical Glass for Type 2 Weathering at 25°C-100 Per Cent R.H. 131 48 SEMs of Weathered Commercial Glass. 132 49 EMP X-ray Intensities as a Function of Exposure Time for Weathered Commercial Glass. 135 50 Auger Signals at Various Depths in Type 2 Weathered Commercial Glass, 100°C-100 Per Cent R.H. , 15 d. 137 51 Schematic Comparing Static Aqueous Corrosion andWeathering. 143 52 Schematic of Plausible Corrosion Mechanisms. 145 x

PAGE 11

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 A DURABILITY EVALUATION OF SODA-LIME-SILICA GLASSES USING ELECTRON MICROPROBE ANALYSIS, INFRARED REFLECTION SPECTROSCOPY AND OTHER TECHNIQUES By David Edward Clark March, 1976 Chairman: Larry L. Hench Major Department: Materials Science and Engineering Traditionally, solution analysis has provided the bulk of glass corrosion data. Although this technique permits an indirect determination of glass dissolution kinetics, it provides little information concerning the structural and chemical changes occurring within the glass surface during corrosion. Techniques have been developed, employed infrared reflection spectroscopy and electron microprobe analysis, for measuring directly the surface compositional changes that occur during glass corrosion. These methods are combined with Auger electron spectroscopy, scanning electron microscopy and improved solutior XT

PAGE 12

analysis techniques for examining the durability of a systematic series of soda-lime-silica glasses. The mechanisms and kinetics of both aqueous corrosion and weathering for these glasses are discussed in terms of preferential Na decoupling from the glass network, S i 0 2 rich film development and Ca-rich precipitate formation. A binary soda-silica and a soda-lime -silica commercial glass are also investigated and the results illustrate the effects on durability of CaO and other commercial glass additives. XT i

PAGE 13

CHAPTER I INTRODUCTION General Discussion The subject of glass durability is certainly of no small importance when one considers the number of glass containers produced for consumer use and the various environments to which these glasses are exposed. Although visual observation of most commercial glass reveals no visible reaction, all glasses are subject to some degree of degradation. The extent of degradation is affected by numerous factors: the composition and phase (i.e., gas, liquid, solid) of the reacting environment, the composition and thermal history of the glass, surface composition of the glass, water affinity of the glass surface, the ratio of the glass surface area to the solution volume and concentration, temperature, pressure, and duration of exposure. This list is not exhaustive and should serve only to illustrate a few of the variables which influence glass durability. "Durability" refers to the characteristic of the glass which determines the resistance to interaction 1

PAGE 14

2 with its environment. A durable glass reacts very little with its en vi ronment, whi 1 e a glass exhibiting poor durability is very reactive. "Aqueous corrosion" will refer to the durability of a glass in solution (where in the study the solution is pure water before reaction begins) and "weathering" will refer to the durability of glass exposed to various humidity conditions. The mechanisms of reaction between glass and water are not elementary. As soon as the reaction begins, products due to reaction between the two alter the corrosion solution and affect the course of further reactions. The study of these reactions is further complicated by precipitate formation as well as chemical and structural changes occurring within the surface of the glass during corrosion. After more than a century of endeavor, only a few of the mechanisms associated with glass durability have been identified. Progress has been impeded primarily by the absence of adequate surface chemistry tools. Until recently, the most widely used technique for evaluating glass durability has been solution analysis. At best, this method provides an indirect determination of glass reaction kinetics, but yields little information concerning the chemical gradients or structural modifications produced in the glass by these reactions. Presently, there are several research tools

PAGE 15

3 which, when used in combination, are capable of providing a direct and more precise evaluation of glass durability than was previously possible. In the present investigation the techniques of adapting these tools for durability evalu ations are demonstrated for the commercially important soda-lime-silica glasses. Summary of Accomplishments 1. An electron microprobe technique utilizing electron beam enlargement and specimen translation has been developed that is suitable for evaluating chemical durabi 1 i ty . 2. A method involving infrared reflection peak frequency shifts has been developed that permits quantitative analysis of corroded glasses independent of their surface roughness. 3. A chemical durability evaluation (aqueous corrosion and weathering) of a systematic series of soda 1 imes i 1 i ca glasses has been performed using electron microprobe analysis, infrared reflection spectroscopy. Auger electron spectroscopy, scanning electron microscopy and solution analysis. Both the usefulness and limitations of each technique are dis cussed in terms of the depth within the surface from which information is obtained.

PAGE 16

4 4. The chemical durability of a commercial sodalime-silica glass has been evaluated and compared to that of the pure soda-lime-silica glasses. Glass Durability Review — Aqueous Corrosion It is important to recognize that when a glass reacts with an aqueous solution both chemical and physical changes occur within its surface structure. Consider a simple binary soda-silica glass whose structure is shown in Figure 1. The addition of 1 mole of Na^O to 2 moles of S i 0 2 produces 2 mol es of nonbri dg i ng oxygen atoms, and provides 2 moles of Na atoms to associate with these oxygen atoms. Each bridging oxygen is shared by two Si atoms and each Si atom is tetrahedra 1 1 y coordinated with four oxygen atoms, three of which are shared with other Si atoms. The addition of CaO to a S i 0 2 glass produces the same number of nonbridging oxygen atoms as an equivalent mole fraction of Na£0. However, CaO provides only one Ca atom per two nonbridging oxygen atoms. Initially the corrosion solution consists of pure water (pH = 7.0), but as the reaction proceeds, accumulation of corrosion products causes both the chemical composition and pH of the solution to change. Several

PAGE 17

5 Figure 1. Mechanisms of Glass Corrosion.

PAGE 18

6 1 2 3 investigators Â’ Â’ have suggested that the reactions should be considered in two stages: 1. the initial stage, involving ion-exchange between Na ions from the glass and H ions from the solution, during which the remaining constituents of the glass are not altered. 2. the second stage, in which breakdown of the silica structure occurs and uniform corrosion ensues. During the initial stage of corrosion, Na ions diffuse from the glass into the solution. Wang and 2 4 Tooley and Lyle have postulated that the driving force for this diffusion process is the Na ion concetration gradient. In order to maintain electrical neutrality, H ions diffuse from the solution into the glass and occupy those sites vacated by the sodium ions. The recognition of this second diffusion process has led to the hypothesis that the driving force for the initial stage involves more than just the Na ion con5 6 7 centration gradient. Â’ Budd discusses the initial stage of corrosion in terms of an electrophilic reaction whereby the electrophilic reagent (H ions in this case) attacks sites of electron excess (nonbridging oxygen atoms ) .

PAGE 19

7 3 Bacon and Calcamuggio suggest that the rate controlling mechanism in the first stage is H ion diffusion since the H ion diffusivity is much lower than the Na ion diffusivity in glass. This is attributed to the tight bonding of the H ion in the hydroxyl group. HowO ever, Douglas and Isard have shown that the rate of corrosion is determined by the rate of Na ion diffusion in Q glass. Douglas and El-Shamy later found that there was no agreement between the diffusion of Na ion in the bulk glass and the apparent Na ion diffusion coefficient determined from corrosion experiments. Furthermore, it was found that the apparent diffusion coefficient for binary glasses, calculated from chemical attack, may be as much C as 10 times greater than that for diffusion in bulk glass. Das and Douglas'^ statement, ". . . the removal of alkali ions from the glass by water is more complicated than a single ion exchange reaction in which the rate determining step is the diffusion of ions through the corrosion layer," accurately describes the complexities of the first stage reaction mechanism. Thus, at present it can only be concluded that the diffusion of Na ion through the corroded layer (away from the glass) is an important contributory factor in the initial stage of corrosion.

PAGE 20

8 g Douglas and El-Shamy have developed the equations for determining the kinetics during the first stage of corrosion. At constant temperature, Q = K • t Y (1) where Q = quantity of Na ions from the glass t = duration of experiment K = reaction rate constant assuming constant glass area y * 1/2 4 Lyle derived a similar expression with the inclusion of a temperature function. y log Q = log t ^ + c ( 2 ) where T = absolute temperature b, c = empirically determined constants. Although not immediately obvious, equations (1) and (2) require that the area of glass undergoing corrosion remain constant. The importance of surface area on reaction kinetics has been previously recognized by El-Shamy and Douglas. However, in most durability

PAGE 21

9 experiments the glasses were ground into fine powders to increase the surface area and therefore effectively reduce the time required to conduct the experiment. This necessitates the inclusion of a geometrical factor in the kinetics equations since the area of the glass capable of supplying Na ions is continuously decreasing as corrosion time increases. The error in spherical surface area is 25 per cent for a 20 ym diameter particle with a corrosion layer of 1 ym, and increases as the thickness of the corrosion layer increases. Therefore, equations (1) and (2) are not valid for powder durability studies. Barrer derived equations allowing for changing surface area of reacting spheres in solid state kinetics, and similar expressions can be derived for corrosion of powders. This will not be attempted in the present investigation since all data were taken from bulk (planar) surfaces . Several other characteristics of the first stage of corrosion are 1. the effective area of S i 0 2 exposed to the corrosion solution is increased by the production of surface micropores resulting from Na ion removal 2. the Na to Si ratio in solution is greater than that for the bulk glass (this suggests selective leaching of the Na ion)

PAGE 22

10 3. the pH of the solution increases as a result of Na ions replacing H ions in solution. El-Shamy 1 3 et_ a_l_. proposed the following equilibrium equation for the initial stage of corrosion at the surface of a soda-silica glass. S i 0 N a ( g ) + h 2 o + si 0 H (g) + NaOH (3) For pH < 9 the NaOH and Si ON a ^ ^ are assumed to be completely dissociated and the corresponding acid constant is given by [SiCf] [H + ] K, < 4 ) [SiOH] a where the brackets represent the activities of the species. The value of K was assumed to be that of dissociation a step of orthosilicic acid, H^SiO^, which was found to be 9 8 10 " . Thus equation (4) becomes [SiO ] [SiOH] 10 -9.8 + pH (5) Equation (5) provides a means of calculating the mole fraction of nonbridging oxygen sites occupied by H ions. Using g this equation, Douglas and El-Shamy show that for pH £ 9 practically all the nonbridging oxygen sites are occupied by H ions at 30°C. For pH > 9, less than one-half of

PAGE 23

11 the sites are occupied by H ions, leaving the other sites to be occupied by Na ions. The increased Na ion surface concentration tends to hinder further Na ion diffusion from the glass. Although no sharp line of demarcation exists between the first and second stages of reaction, the latter effectively begins at pH > 9 for soda-silica glasses. This reaction involves complete breakdown of the silica network and all species dissolve at approximately a uniform 1 3 rate. El-Shamy e_t aj_. have shown that even vitreous silica dissolves in solution where pH > 9. This reaction is illustrated schematically in Figure 1 and shows that the presence of OH ions at the glass surface is a pre requisite. The equilibrium equation for the reaction can be represented by — Si— 0 — Si— OH ( 9 ) + OH ( so 1 n . ) J — SI— OH (g) + 0— SI— 0H (aoln _, (6) It can be seen that increasing the OH ion concentration will favor dissolution of silica. However, it was demonstrated earlier that an increase in pH suppressed the diffusion of Na ion from the glass. From this information

PAGE 24

12 it would appear that Na ion diffusion from the glass would decrease at pH > 9. However, experiments have shown that selective leaching of Na ions occurs even at pH = 12, although not to the same extent as at pH < 9. This has been explained by Douglas and El-Shamy 9 to be due to the decrease in width of the siliceous layer at high pH. The reaction kinetics for this stage can be determined from equations (1) and (2) for planar surfaces. The value of y has been experimentally determined to be slightly greater than one, thus suggesting some selective leaching even during the second stage of reaction. Budd* 7 discusses the second stage reaction in which a nucleophili reagent, such as OH ions, attacks positions of electron deficiency. Accompanying this reaction is also an electro phi lie reaction where Na ions in solution attack the nonbridging oxygen atoms created by the nucleophilic attack. The equilibrium conditions established by these reactions determine the value of y obtained for the second stage. Sanders and Hench^ discuss the two stages of reaction in terms of three corrosive parameters, a , g, and e. These parameters are calculated from the quantities of various species in solution. The dissolution parameter, a, measures the extent of selective leaching. It is equal to zero for the dissolution of only one component (first stage) and is equal to one for total

PAGE 25

13 dissolution (second stage). The balanced silica parameter, 6, is the amount of silica which would be in solution if selective leaching occurred. The film parameter, e, is a measure of the S i 0 2 available on the surface to form a silica-rich film and is directly proportional to the film thickness. By measuring the time dependence on these parameters, composition profiles and information on film behavior can be obtained.^ The selective leaching parameter, a, and film parameter, e, equations are given below. PPM Si0 2 MW ST07 a = + TT PPM R + PPM IT MW R + MW R 2 + e = PPM S i 0 0 (1—^) 2 a ' 1 P c . n S l 0, Si 0, (7) ( 8 ) Symbols in the above equation are: P S i 0 2 PPM MW R + R 2 + mole fraction of S i 0 2 in glass parts per million mol ecul ar wei ght alkali ion alkaline-earth ion Similar solution parameters have also been derived by Sen and Tool ey . ^

PAGE 26

14 Glass Durability Review W eathering The degradation of a glass surface due to interaction with the atmosphere is referred to as “weathering." Weathering has been classified into two types: 1. condensation-runoff, in which moisture collects on the glass surface until natural runoff occurs, carrying away the reaction products 2. condensation-evaporation, in which a thin layer of "fog" forms on the glass surface but evaporates before droplets form. Type 1 weathering is very similar to aqueous corrosion in which the solution is continuously replenished at a specified rate. During the time that the water droplet remains on the surface, dealkalization of the glass occurs with a simultaneous increase in the pH of the water. If the surface area to solution volume ratio is high in the water droplet, rapid pH increases will cause total breakdown of the glass in contact with the droplet. The resulting localized deterioration and roughening of the glass surface leads to solution entrapment during runoff. Further degradation of the surface continues where the solution is pooled in the corroded pores. The pH in subsequent water droplets will not change as rapidly since the majority of the Na ion leaching occurs in the initial droplets. However, as discussed previously.

PAGE 27

for pH > 9 the total dissolution significantly decreases the thickness of the leached layer. Sanders and Hench 17 have shown that for a large surface area to solution volume ratio, (simulates water droplet on glass) large quantities of the alkali remain in the glass surface at high pH values. Apparently very little alkali is required to increase the pH of the small volume to the region where total dissolution occurs. Type 2 weathering is characterized by the presence of reaction products on the glass surface. This type of weatheri ng, produced in cyclic temperature or cyclic humidity environments, is the greatest source of concern to glass manufacturers. Attack is initially manifested by a slight dimming of the glass surface, followed by the formation of a noticeable white film which varies in thickness according to the composition of the glass. Tichane and Carrier have studied these films extensively using transmission electron microscopy and X-ray diffraction analysis, and have shown that they are rich in Na. Furthermore, Tichane has shown that the Na-rich films react with most acidic gases such as C 0 £ and S 0 ^ to form their respective salts at temperatures below 300°C. Since CO^ is present in significant quantities in our atmosphere, Na^CO^ should be a reaction product on 20 most weathered glasses. Simpson has also verified the

PAGE 28

16 presence of both Na 2 C0 3 and CaC0 3 by X-ray diffraction analysis on the surface of weathered commercial glasses. Unlike type 1 weathering, the precipitate formed in type 2 can act as a protective barrier for retarding further corros i on . Little quantitative data a re a va i 1 abl e for either type of weathering because the research methods for extracting the appropriate data have been developed only recently. Simpson^ developed a "haze meter" for quantitatively measuring film growth and appears to have successfully applied this technique to commercial glasses. However, the method does not yield any useful information concerning the structural modification of glass during corrosion . Research Objectives There are several newly developed techniques for characterizing the chemical durability of glasses. Infrared reflection spectroscopy (IRRS) is an exceptional tool for determining structural and chemical changes in sodasilica glass surfaces exposed to various environments.^ Similarly, corrosion solutions have been analyzed by atomic absorption (AA) and atomic emission (AE), and the results used to indicate the uniformity of corrosion.

PAGE 29

17 Moreover, recent advances promise to make electron microprobe analysis (EMP) an attractive technique for durability studies on soda-lime-silica glass. Each of these instruments yields averaged information which is characteristic of a volume extending from the surface to a specific depth (usually _> 0.5 pm) within the sample. Thus, an accurate chemical analysis of extremely thin surface films, multilayered surface films, and surface reaction layers containing composition gradients may require additional techniques. Rynd and Rastogi have demonstrated that Auger electron spectroscopy (AES ) is a particul arly effective means of investigating the surface compositions of commercial glass fibers because the escape depth of the o Auger electron is small (<5 A). Furthermore, Pantano 2 6 et al . have shown that AES coupled with ion milling is a reliable method for determining chemical profiles on a submicron level in corroded bioglass specimens. The primary objective of this investigation was to characterize the durability of soda-lime-silica glass and a similar commercial glass through the joint use of IRRS, EMP, SEM (scanning electron microscopy), solution analysis, and AES. Comparison of these glasses with a soda-silica glass also yielded information concerning the roles of CaO and Al^O^ in glass durability. For this investigation, surface analysis is defined according to the

PAGE 30

18 depth within the sample from which information was obtained; the "near" surface determined by AES refers to o a depth < 5 A from the glass-air interface; the "middle" surface, measured by IRRS, extends to -.05 ym;*and the "far" surface is defined by the effective EMP electron penetration depth as 1.5 y m . Using ion milling with AES yields information from all three regions, whereas in solution analysis the depth from which information is obtained depends on the extent of reaction (Figure 2). These techniques, with the exception of solution analysis, allow for a direct measure of glass durability. Comparison of the results from the various techniques provides a better understanding of glass corrosion. *These values vary slightly according to the density, composition and corrosion extent of the glass.

PAGE 31

19 Q_ a> o Q c < c O _c io O O CO o u -« V W 00 ^ < at: LU a£ Figure 2. Sampling Depths for Various Glass Durability Analysis Techniques.

PAGE 32

CHAPTER II A TECHNIQUE FOR ELECTRON MICROPROBE ANALYSIS OF CORRODED SODA-LIME-SILICA GLASSES Introduction — A Discussion of the Problem Electron microprobe analysis (EMP) of alkali glasses has been recognized as a problem by previous investigators. ^ ^ Generally, the X-ray intensity for the alkali species changes as a function of time when an electron beam remains stationary on a glass specimen. An accompanying change in the X-ray intensities of other atomic species in the glass is also ob29 30 served. ’ This X-ray intensity variation suggests a compositional change in the sample volume irradiated by the electron beam. Pantano ejt a_l_. ^ have reported similar changes in the Na Auger electron intensities for alkali glasses subjected to a stationary beam. Plausible mechanisms associated with these local compos i tional changes are discussed by Line weaver ,^' 7 Borom and Hanneman,^ and Tanaka and Warrington.^ The surface condition of the glass specimen is also an important consideration in compositional analysi 20

PAGE 33

21 Both abraded and corroded surfaces can cause considerable variation in X-ray intensities when the dimensions of the surface perturbations and electron beam size are comparable. However, the capability for such analyses is essential, especially in the container industry where glasses are subjected to corrosive environments. The rate of Na X-ray intensity change is proportional to the electron beam current and specimen temperature, and is related to the bulk composition of the glass under consideration. Small electron beam currents, low specimen temperatures, and the presence of Ca, A1 , and Mg give correspondingly lower rates of alkali X-ray intensity 34 q changes. Douglas and El-Shamy report that the diffusion rate of Na ions is much greater in the corroded layer than in the bulk glass. Thus, the rate change of X-ray intensity should be correspondingly greater for corroded specimens than for uncorroded glass. Several methods for reducing the rate of change in alkali intensity are discussed in the literature. 30, 32, 3/ *~ 36 An electron raster technique has been employed to stabilize the X-ray intensities of K^O — SrO — S i 0 2 glass for a on o c sufficient time to obtain qualitative analysis. * Moreover, it has been suggested that the change in glass X-ray intensities can be decreased by translating the 3 1 specimen at a constant rate under a stationary beam.

PAGE 34

22 Presented here is a method for obtaining reliable compositional analyses for a variety of soda-lime-silica glasses by utilizing electron beam enlargement and specimen translation. Because of its macroscopic nature, this method is especially useful for analysis of semi -rough surfaces obtained by abrading or corroding glass speci37 mens. This technique is used to evaluate the earlystage corrosion behavior of three soda-lime-silica glasses. Experimental Several glasses were prepared by melting mixtures of reagent grades Na 2 C0 3 , CaC0 3 , and 5.0 ym Min-u-sil* at 1400-1500°C for 24 hour (h) in a Pt crucible. This melting procedure was previously found to be satisfactory for batch homogenization and small bubble elimination. The molten glass was pressed between two graphite plates to a thickness of 3/8 in. and annealed at 450°C for 4 h under prevailing atmospheric conditions. Several investigators have noted the effect of the annealing atmosphere on the durability of soda-lime-silica glasses. ’ ^ McVay and Farnum^ have reported that the affected region of glass extends = 200 ym into Pa. Pennsylvania Glass and Sand Company, Pittsburgh,

PAGE 35

23 the glass surface for annealing conditions similar to the one used for this investigation. Thus, specimens 3/4 in. square were polished through 600 grit with SiC paper, removing at least 1000 pm, before quantitative microprobe analysis was performed. The polishing also eliminated thin films and compositional variations which might have formed at room temperature due to reaction with atmospheric moisture. 17 In order to evaluate the effects of surface roughness on X-ray intensity, similar specimens were ground to 120, 240, 320, and 400 grits. (The surface roughness decreases as the grit number increases.) Dilmore et a^. have shown that a specimen polished to 120 grit has approximately 1.5 times the surface area as the same specimen polished to 600 grit. This is due to the large surface perturbation created by the coarse polishing. The prepared samples were maintained in a desicator until examined with the EMP.* The specimens were coated by vacuum evaporation o with approximately 100 A of C and electrically connected to A1 disks with Ag paint. The disks were placed on a stage inside the EMP vacuum chamber, which could be electrically driven perpendicularly to the electron beam *Electron Probe X-ray Microanalyzer, Model Ms-64, Acton Labs, Acton, Mass.

PAGE 36

24 at specified rates between 0 ym/min and 150 ym/min. The electron beam diameter was varied between 1 ym and 200 ym for various specimen speeds while maintaining a constant beam current of 10 ^ amps and voltage difference of 20 kv. An Arme thane flow proportional detector with a KAP crystal was used to measure the Na and Si Ka radiation intensities. A PET crystal was employed in obtaining Ca Ka radiation intensities. Corroded specimens were prepared from glasses containing 10, 20, and 30 per cent* Na^O, each containing 10 per cent CaO, with the remaining constituent being S i 0 2 . The samples were ground to 600 grit before corroding in a Teflon ** ** cell at 100°C for up to 12 h. A major feature of the corrosion cel 1 -assembly is that it provides a constant surface area to solution volume ratio during corrosion. The cell assembly and corrosion technique are described more fully in Chapter IV. Results and Discussion — Bulk Glasses Figure 3 illustrates the effect of allowing the electron beam to remain in a stationary position on a *Per cent will refer to mole per cent throughout this dissertation unless otherwise designated. **E. I. Dupont de Nemours & Co. , I nc ., Wi 1 mi ngton , Del.

PAGE 37

25

PAGE 38

26 20 NaoO — lOCaO — 7 0 S i 0 2 * glass. Every data point represents an average of five measurements on different areas of the sample. The intensities in counts per second (cps) were obtained by placing the sample under the beam and taking fixed time counts (10 s each) at 1 min intervals. The X-ray intensities given for 0 time correspond to those obtained during the first 10 s of electron beam impingement. Although the Na X-ray intensity is not time independent for any beam size investigated, the rate of intensity decrease is much less for the larger beam sizes. Figure 4 is a 10 s cross section of Figure 3, illustrating the effects of beam diameter on the Na X-ray intensity. The X-ray intensity is significantly reduced as the beam size is decreased below 100 urn. Furthermore, the data suggest that a quantitative analysis is possible using a beam diameter 100 urn if the measurement is made within 10 s after the electron beam impinges upon the sample. Vassamillet and Caldwell"^ show that significant changes in the local composition occur when the electron beam is allowed to remain in a stationary position on a sodasilica glass for sufficient times to make a complete analysis (i.e., > 1 min), A similar effect is observed *The number preceding each constituent represents its mole per cent in the glass, unless otherwise noted.

PAGE 39

27 CO c O) >> sI X fO c o ifT3 0)Q . 2 e v/} ro cu CQ Ec Oo 0) s00 +J u (U LU 0 J (Sd>) A|ISUd|U| Figure

PAGE 40

28 for the soda-1 ime-sil ica system. The extremely low intensity obtained with the 1 ym beam is probably due to a combination of surface roughness-beam interference and Na ion diffusion from under the beam. In order to min imize compositional analysis errors, the samples were moved at various constant speeds under a stationary 100 pm electron beam. Na X-ray intensity as a function of beam impingement times for various specimen speeds and a 100 ym beam diameter is shown in Figure 5. This graph demonstrates that the variation in X-ray intensity decreases as the specimen speed increases. Using a 150 ym/ min speed, the X-ray intensity becomes almost independent of time and the variation in intensity is within the experimental scatter band normally encountered in EMP analysis. The X-ray intensities of Si and Ca were also independent of time for this beam diameter and specimen vel oci ty . Other soda-lime-silica compositions with the Na 2 0 ranging from 10 to 30 per cent and the CaO ranging from 0 to 20 per cent were analyzed using this technique. Figure 6 shows that a 100 ym beam diameter and 150 ym/ min specimen speed were sufficient to give time independent X-ray intensities for all the compositions investigated. Of further importance is the time independent X-ray intensity for the corroded specimen. Although the diffusion rate for Na in corroded layers has been reported

PAGE 41

29 Figure 5. Na X-ray Intensity as a Function of Electron Beam Impingement Time and Sample Velocity.

PAGE 42

30 Na X-ray Intensity as a Function of Electron Beam Impingement Time for Several Glasses. Figure 6.

PAGE 43

31 to be much greater than for the bulk glass, the beam enlargement and specimen speeds employed are sufficient to compensate for this increased diffusion rate. Therefore, this technique is suitable for evaluating the corrosion behavior of glasses where the surface compositions are time dependent. Figure 6 also shows that the replacement of S i 0 2 with CaO decreases the Na X-ray intensity. This is probably due to the higher mass attenuation of Ca compared to Si, for Na X-rays. Results and Discussion — Corroded Glasses Figure 7 presents the Na X-ray intensity as a function of corrosion for three soda-lime-silica glasses. Since the X-ray intensity is related to the mole per cent of Na present in the glass, the slopes of these curves represent the rates of dealkalization within the volume of analysis. This graph illustrates that the rate of Na diffusion from the glass is proportional to the bulk composition. As the per cent of Na in the glass is increased, the rate of Na in the glass is also increased. The increased dealkalization rates are probably due to both the larger Na concentration gradients encountered in the high soda glasses and the more open structure produced during corrosion. Severe surface roughening and flaking were observed on the glass containing 30 per cent Na^O after 6

PAGE 44

32 Corrosion Time [hrs.] Figure 7. Na X-ray Intensity as a Function of Corrosion Time for Three Soda-Lime-Silica Glasses.

PAGE 45

33 h of corrosion. The same glass corroded for 12 h produced a f 1 a key surface which was too rough for EMP analysis. After mechanically removing some of the flakes, a Na X-ray intensity was obtained for the underlying surface which was 2/3 the bulk glass value. Scanning electron micrographs (SEMs)* for these glasses are shown in Figure 8. The cracks and flakes observed in the 30 per cent Na£0 glass, not evident immediately after removal from the corrosion solution, are thought to be due to dehydration while in the desiccator. The effect of surface area to solution volume ratio is also illustrated in Figure 7 for the 20 per cent Na20 glass. Decreasing this ratio by a factor of 4.5 yields no significant difference in the dealkalization rate. This re suit, suggesting that the rate of dealkalization in the first stage (pH _< 9.0) of corrosion is independent of Na concentration in the solution, is consistent with that of Rana and Douglas 5,6 and El-Shamy and Douglas.^ Although not shown graphically, the X-ray intensities for Ca and Si were also monitored as a function of corrosion time for the three glasses. The Ca X-ray intensity remained constant while the Si X-ray intensity *Cambridge Stereoscan, Kent Cambridge Scientific, Inc . , Morton Grove , 111.

PAGE 46

34 Figure 8. SEMs for Soda L i me-S i 1 i ca Glasses Corroded for 12 h at 1 00°C : (a) 10 Na 2 0-10 CaO, 80 SiO? ( 6000x ) (b) 30 Na 2 0 — 10 CaO, 60 Si0 ? (55x) (c) 30 Na 2 0 10 CaO, 60 Si0 2 . Fractured surface ( 2 00 0 x ) .

PAGE 47

35 increased in a manner related to the amount of Na depletion for all corrosion times _< 12 h. As Na is removed from the glass, increasing quantities of Si are exposed to the electron beam and a corresponding increase in Si X-ray intensity is obtained. The relative Constance of the Ca X-ray intensity is thought to be due to two factors: 1. the smaller mass attenuation of Na on Ca 2. the greater effective escape depth of Ca, compared to Si , from the glass. Both of these factors would tend to decrease the Ca X-ray sensitivity to Na variations in the glass. The effect of sample surface roughness on the coefficient of variation in X-ray intensity is shown in Table 1. These measurements were made using a 100 pm beam diameter and sample speed of 150 pm/min. In general, as the surface roughness increases, the variation in X-ray intensity also increases. This beam diameter and specimen speed were sufficient to permit an accurate analysis for a sample polished with 400 grit or finer. IRRS and SEM illustrate that the surface finishes produced from this range of grit sizes are comparable to those encountered for the 10 per cent Na^O and 20 per cent Na^O glasses corroded up to 10 h and for the 30 per cent Na20 corroded for 1 h. The 30 per cent Na20 corroded for 6 h gave a

PAGE 48

Table 1. Coefficient of Variation in X-ray Intensity as a Function of Surface Roughness for a 20 Mole Na,0 — 10 Mole CaO — 70 Mole Si0 o Glass. 36 OO CO CO m o < CL) CO N CO Cl) OO c m cn •rZ3 SO CD o o o o o OJ C\J o o ' — C\J CO *3"

PAGE 49

37 surface roughness comparable to a 120 grit polish or less with a corresponding large error in X-ray intensity. Summary An EMP technique has been developed which provides for an accurate and reproducible compositional analysis of both bulk and corroded soda-lime-silica glass. Application of this technique to several glasses reveals that the rate of dealkalization is dependent on the composition of the glass. Within the depth of analysis (=; 1.5 y m ) , the glass composition containing 30 per cent Na 2 0 exhibits a much greater Na depletion rate than the glasses containing less Na^O. This is a result of the greater Na concentration gradient and hence a greater driving force for Na ion diffusion from the glass. The majority of the Na ions are depleted from the analysis volumes of the 30 Na 2 0 — 10 CaO — 60 S i 0 2 during the first hour of corrosion. The low X-ray intensities and the small change in intensities between 1 h and 6 h are due to the finite depth of analysis of the EMP. The high concentration of Na ions remaining in the uncorroded glass is at a depth below the range of the analysis and essentially only the tail of the Na concentration gradient is detected by the EMP. The glass surface area to solution ratio does not alter the rate of dealkalization for the two ratios

PAGE 50

38 investigated. This is due to the relatively small change of Na ion concentration in solution for pH < 9. Sample surface roughness is also an important parameter in EMP analysis. The error in analysis increases as the surface roughness increases due to X-ray scattering. The surfaces of samples containing 30 per cent Na20 are too rough for EMP analysis after 12 h of corrosion due to extensive cracking of the Na-depleted film while samples containing 20 per cent Na have surfaces comparable to a 400 grit polish or finer. Thus, the latter surfaces provide more reliable EMP analysis.

PAGE 51

CHAPTER III INFRARED REFLECTION SPECTROSCOPY OF CORRODED SODA-LIME-SILICA GLASSES UTILIZING FREQUENCY SHIFTS Introduction — A Review of Infrared Reflection Spectroscopy of Glasses Traditionally, chemical analyses of solutions in contact with glass powders (large surface area to solution volume ratio) have provided the bulk of glass corrosion aa t a This method allows accurate chemical analysis of dissolved species, but provides little information concerning the surface structural changes accompanying dissolution. Scanning electron microscopy (SEM), electron microprobe analysis (EMP), Auger electron spectroscopy (AES) and infrared reflection spectroscopy (IRRS) are currently being utilized for characterizing glass corrosion. ^ It has been demonstrated that the latter is sensitive to botn the chemical and structural changes in the spectral region from 1200 cm ^ to 600 cm ^ ^ ^ Furthermore, minimum specimen preparation, no vacuum requirements and rapid sample analysis make this an attractive tool for on-line qua! i ty control . 39

PAGE 52

40 2 3 Sanders ejt jfk have developed a method of quantitative glass compositional analysis based on the proportionality between IRRS peak amplitudes (intensities). In some glass systems, notably the binary silicates and high alkali soda-lime-silica glasses, corrosion can lead to severe surface roughening with a simultaneous decrease in the infrared reflected intensity. Precipitate formation on weathered glasses can also give a decrease in infrared reflected intensity. 7 Thus, compositional analysis of corroded glass surfaces using an analysis of peak amplitudes, may yield misleading results. Several investigators have demonstrated that infrared peak frequency (or wavenumber) shifts can be used for compositional analysis of binary sodium silicate 44-48 glasses. T Other researchers have applied this . 49 50 technique to more complicated systems. Frequency shifts are a result of structural modifications in the glass, and are not affected by the physical condition of the surface. The objectives of this chapter are to 1. demonstrate the usefulness of infrared peak frequency shifts for monitoring glass corrosion 2. apply this technique quantitatively to the corrosion of soda-silica and soda-lime -silica glass systems .

PAGE 53

41 E xperimental Silicate glasses containing 10-30 Na20 and up to 10 per cent CaO were prepared by melting reagent grade carbonates and 50 pm Min-u-sil in covered Pt crucibles between 1 4001550°C for 24 h. Glass slabs 1/4 in. thick were cast between graphite plates and annealed at 450°C for 4 h. All specimens were polished to 600 grit with SiC paper except those noted. The 600 grit surface provides a common base for data normalization.^ Samples 3/4 in. square of 20 Na 2 C— 10 CaO — 70 S i 0 2 glasses were corroded in a static aqueous environment by the technique described by Sanders et a 1 . ^ Briefly, the samples were sealed onto Teflon ^ corrosion cells which were filled with deionized water and maintained in a hot water bath at constant temperatures, for various times. The procedure utilized a constant exposure area to solution volume ratio of 0.77 cm ^ . Corroded specimens were stored in a desiccator while awaiting analysis. IRRS spectra* for the freshly abraded (standards) and corroded samples were obtained for the spectral region 1 400 cm to 600 cm ^ using a medium scan rate and 23° angle of incidence. The spectrum for vitreous silica was *Grating Infrared Spectrophotometer, Model 467, Perkin-Elmer Corp., Norwalk, Conn.

PAGE 54

42 used for adjusting the data acquired over a period of several months. Samples were prepared for the SEM and EMP analysis by vapor coating with * 100 A of C and electrically connected to A1 mounts with Ag paint. Quantitative EMP analysis was performed by traversing a 100 pm diameter electron beam at 150 pm/min parallel to the sample surface to minimize alkali diffusion. 24 Results and Discussion Technique Development The variation in infrared reflection intensity with surface roughness is shown in Figure 9. The intensity increases significantly as the sample surface is polished with finer grit paper. A maximum reflected intensity of ~ 25 P er cent at 1045 cm 1 is obtained for the 20 Na^O — 10 CaO — 70 Si0 2 glass when the sample is polished with 400 grit or finer SiC paper. It should be noted that the locations of the peak maxima shown in this figure are not altered by the surface condition. Figure 10 presents an SEM of this glass polished to 120 grit and exposed to a humid atmosphere for 24 h at 25°C. Surface roughness due to both polishing scratches and precipitate formation can be observed. EMP analysis revealed that the surface was. rich in both Na and Ca. The precipitate is probably the carbonates of these elements. 19, 20

PAGE 55

20mole%Na2O-10mole%CaO 43 d% Wavenumber [cm'l] Figure 9. Effect of Surface Roughness on the Infrared Reflected Intensity.

PAGE 56

44 Figure 10. SEM of a 20 NaoO — 10 CaO — 70 S i 02 Glass Polished to 120 Grit and Exposed to Atmos here {70 Per Cent R.H.) for 24 h at 25° C ( 1 300x) .

PAGE 57

45 The peak maximum in Figure 11 occurring at 1110 cm ^ for freshly abraded (f.a.) silica has been assigned -4 to the Si-0 stretch vibration within the S i 0 ^ tetra45 46 hedra. * The exact location of this peak varies 5 2 slightly with different investigators and Hanna has suggested that its position is dependent upon the presence of nonbridging oxygen atoms necessary to accomodate the random orientation of the SiO^ groups. This peak is shifted to lower wavenumbers and the intensity is simultaneously decreased as Na 2 0 is added to the Si0 o network . A satellite peak at = 960 cm ^ appears as a shoulder on the Si-0 stretch peak for the glass containing 20 per cent Na 2 0 (Figure 11). This peak has been assigned to a 42 Si-0 nonbriding oxygen stretch in an alkali environment. The exact position of the peak maximum is dependent on both the alkali species and its concentration in the glass. The infrared spectra of a 20 Na 2 0 — 80 S i 0 2 glass corroded at 100°C for various times are also shown in Figure 11. The Si-0 stretch maximum (= 1070 cm’^ for the freshly abraded sample) shifts to higher wavenumbers and increases in intensity during the first hour of corrosion. The Si-0 nonbridging maximum shifts to lower wavenumbers and disappears after 1 h of corrosion. These peak shifts

PAGE 58

46 Figure 11. Infrared Spectra for a Freshly Abraded (f.a.) Glass and the Same Glass Corroded for Various T imes .

PAGE 59

47 have been associated with diffusion of Na ions from the 3 7 glass into solution. It is interesting that the amplitude of the Si-0 nonbridging oxygen stretch maximum decreases to zero at the frequency corresponding to the lowest wavenumber (820 cm !) in the original distribution of the freshly abraded sample. If this peak is representative of non briding oxygen atoms associated (to various degrees) with Na ion, then the maximum corresponds to the frequency where the largest number of S i 0 • • N a occur with a particular C o 0 • • N a association strength. Hanna and Su have suggested that the 0**Na association weakens the Si-0 terminal bond and shifts its maximum to lower wavenumbers. Thus, the Na ions with the weakest 0**Na association will first diffuse from the glass and the Na ions with the strongest association will diffuse last from the glass. This diffusion sequence should shift the Si-0 nonbri dging maxi mum to lower frequencies, decrease the width of frequency distribution and decrease the maximum intensity as the corrosion time increases. These suggested spectra changes are consistent with the experimental results (Figure 11). The Si-0 peak maximum continues to shift to higher wavenumbers for up to 3 h of corrosion suggesting the continuous development of a S i 0 2 rich layer. However, the amplitude decreases significantly between 1 h and 3 h and

PAGE 60

48 this could be interpreted as a decrease in S i 0 2 if the surface roughness contribution is not considered. Since the amplitude of this peak is directly related to the S i 0 2 concentration and inversely proportional to the surface roughness, peak heights cannot be used for any stage of quantitative corrosion analysis without a separate evaluation of surface roughness. Apparently, the surface roughness becomes the dominant factor after 1 h or corrosion for the 20 N a 2 0 — 80 S i 0 2 glass. A SEM illustrating surface roughening on this glass after 2.5 h at 100°C is shown in Figure 12a. Although not as severe, surface roughening can also be observed on the 20 Na,,0 — 10 CaO — 70 S i 0 2 glass corroded for 12 h at 100°C (Figure 12b). However, roughening does not become the controlling factor to amplitude until after 5 days (d) of corrosion (Figure 20). The Si-0 stretch maximum was used for quantitative analysis due to its greater sensitivity to compositional variations below 20 per cent Na 2 0. Figures 13 and 14 illustrate the infrared reflection spectra for the N a 2 0 — S i 0 2 and Na 2 0 — 10 CaO — S i 0 2 glasses freshly abraded to 600 grit. Figure 15 presents in graphical form the Si-0 stretch maxima wavenumbers as functions of per cent Na 2 0 in the glass for these two figures. The graph for the binary glasses is linear up to 22 per cent Na 2 0 in agreement with the results of other investigators. ^ » ^8 For a gj ven concentration of

PAGE 61

49 Figure 12. SEMs of Corroded Glasses for (a) 20 NapO — 80 S i 0 2 2.5 h 1 00 ° C (lOOOx) (b) 20 NapO — 10 CaO — 70 S i 0 2 » 12 h-1 00°C ( 1 200x) .

PAGE 62

50 Wavenumber [cm 1] Figure 13. Infrared Spectra for Several Na 2 0 — S i 0 2 Glasses.

PAGE 63

51 n% Wavenumber [cm" 1 ] Figure 14. Infrared Spectra for Several Na 9 0 — CaO — Si0 9 Glasses.

PAGE 64

52 Figure 15. Calibration Curve for the Si-0 Stretch Peak Maxima of Na 9 0 — S i 0 o and Na o 0 — 10 CaO — SiO„ Glasses. 1 2 1 2

PAGE 65

53 Na 2 0, the replacement of Si0 2 with CaO shifts the Si-0 stretch peak to lower wavenumbers. A reversal occurs in the Si-0 stretch peak at = 22 per cent Na 2 0 for the binary glass and at 20 per cent for the glass containing 10 CaO. Similar trends are observed in the data of other investigators and are thought to be related to the critical nonbridging oxygen concentration (0/Si > 2. 5). 47 Since most commercial glasses contain £ 20 per cent Na 2 0, problems which arise from reversals in peak location can be avoided in this approach to quantitative analysis of corrosion. The variations in wavenumber of the Si-0 stretch maximum with corrosion time and temperature for the 20 Na 2 0 — 80 S i 0 2 and 20 Na 2 0 — 10 CaO — 70 S i 0 2 glasses are illustrated in Figure 16. The peak maxima increased from 1070 cm 1 for the freshly abraded surface to 1100 cm" 1 after corroding the binary glass at 50°C for 24 h and at 100°C for 3 h (Figure 16a). The maxima increased from 1045 cm 1 to 1090 cm 1 after corroding the ternary glass for 9 d at 100°C (Figure 16b). The composition versus corrosion time graphs shown in Figure 17 were constructed from Figures 15 and 16 by assuming that the CaO to S i 0 2 ratio remains constant in the ternary glass during corrosion. This assumption is based on solution analysis showing that the dissolution of Ca from the ternary glass is nearly stoichiometric with

PAGE 66

c o to O O 03 C-> s~ o o CJ I— 4O O CXI C 03 o z: •r* +J O U OsJ c: Li03 (/) to to 03 03 03 CD E •rCNJ X o 03 *r5: oo -C o O CO 4-> CD iO 4-> CNJ CO) 03 o I o •rCNJ 00 4tO O 03 to 03 CCD CD -Q SE O CNJ Z5 4o C -r(D CD CO > E 03 *rO 2hN to CD S13 cn

PAGE 67

55 ["X Dili] OiS JsqujnuaADyyy Corrosion Time [hrs.]

PAGE 68

1120 56 00 CO CM o < 1 ) E c o o o u [xdiu] OlS Figure 16 continued.

PAGE 69

CM so o •<400 CD O E r" £ I o C CO o o < A O O r— 1 1 o o O C\J 03 O O C C\J O +> -Q 13 to U_ to 03 03 r— CJ3 to 03 CM O C *rO GO 4-> O 03 00 s4-» c o 03 OJ O 03 c z o o o • C\J to O to CM"-" 03 03 03 r— Z'—'O 03 S3 cn •rLl_

PAGE 70

58 [0^ D N%] uoijisoduuo^ Corrosion Time [hrsj

PAGE 71

59 [0^ D N%] uoijisodiuo^ Figure 17 continued

PAGE 72

60 3 7 respect to the dissolution of Si. The graphs in Figure 17a show that the rate of Na ion depletion from the binary glass increases as the corrosion solution temperature increases. For all three temperatures there is an initial rapid rate of Na ion depletion from the glass followed by a gradual decrease in the depletion rate. When the Na^O concentration in the glass is plotted as a function of the square root of corrosion time, two linear regions are obtained for each temperature (Figure 18a). This suggests a change in the diffusion controlling mechanism for Na ions leaving the glass. The change in mechanisms occurs at pH >_ 9 for the three temperatures investigated. Douglas and g El-Shamy have found similar changes in the dealkalization rate and interpreted these decreases in terms of H ion availability at the surface of the glass. For pH > 9, the H ions in solution are associated with only a small fraction of the nonbridging oxygen atoms on the glass surface (equation 5). These remaining sites become occupied by Na ions, which effectively decreases the Na ion concentration gradient between the glass and solution. This results in a decrease in the Na ion diffusion rate from the glass. Infrared analysis indicates there is 3.5 per cent Na 2 0 remaining within the depth of IRRS analysis of the 3 h100°C sample. The EMP analysis of the same sample yielded = 2 per cent Ma^O based on comparison with a NaCl standard.

PAGE 73

Figure 18. Na£0 Concentration as a Function of the Root of Corrosion Time for: (a) 20 Na20 S i 0 2 Glass (b) 20 Na20— 10 CaO — 70 Si02, 10 Na£0 — 10 CaO — 80 SiO^ Glasses. Square — 80

PAGE 74

62 (a)

PAGE 75

63 (b) Figure 18 continued

PAGE 76

64 Figure 17b illustrates that when 10 per cent CaO is substituted for S i 0 ^ in the glass the Na ion depletion rate is decreased by a factor of = 3. Again, two linear regions are observed in the plot of N a 2 0 concentration versus the square root of corrosion time for the 20 Na^O 10 CaO — 70 S i 0 2 glass. Although the corrosion time at which the change in mechanisms occurs is greater for the ternary than for the binary glass at the same temperature, the pH at which the change in mechanisms occurs is approximately the same for both glasses. Figure 18b also shows that the initial rate of Na ion diffusion from the 10 Na^O — 10 CaO — 80 S i 0 2 glass is very slow and approximately equal to that of the second mechanism in the ternary glass containing 20 per cent Na 2 0 . A change in the diffusion rates is not observed in this glass even after 9 d. Infrared analysis of the 20 Na 2 0 — 10 CaO — 70 S i 0 2 glass corroded for 9 d at 100°C indicated that 2 per cent Na 2 0 remained within the depth of analysis. The EMP analysis for this sample also yielded = 2 per cent Na 2 0. Summary The infrared frequency maxima associated with various vibrational modes in sodium silicate glasses are sensitive to compositional changes but are unaffected by the surface condition. The Si-0 stretch maxima shift linearly with

PAGE 77

65 N a 2 0 content up to 22 per cent N a 2 0 . Similar changes are found for sodium silicate glasses containing 10 per cent CaO. The same frequency shifts are observed when these glasses are corroded, and are attributed to Na ion diffusion from the glass into solution. The Si-0 stretch peak has been used to quantitatively determine the per cent Na^O remaining in these glasses as a function of time and temperature. The results show that decreasing the solution temperature from 100°C to 30°C greatly decreases the rate of Na ion depletion from the binary glass. There is a similar decrease in Na ion depletion rate by substituting 10 per cent CaO for S i 0 2 in the glass. A change in diffusion mechanisms is observed at pH > 9 for all the glasses investigated except the 10 Na 2 0 — 10 CaO — 80 Si0 2 glass. The surface compositions determined from infrared analysis are in satisfactory agreement with the EMP analysis. Although quantitative analysis of multicomponent comme rcial glasses appears to be possible, the determination of calibration curves is more complicated. However, qualitative analysis of commercial glass durability using this technique is superior to peak height analysis, especially in weathering studies. This will be demonstrated in Chapter V.

PAGE 78

CHAPTER IV APPLICATION OF THE ELECTRON MICROPROBE, INFRARED REFLECTION SPECTROSCOPY AND OTHER TECHNIQUES FOR EVALUATING AQUEOUS CORROSION OF SODA-LIME-SILICA GLASSES Introduction — Review of Corrosion Evaluation Techniques Evaluation of glass aqueous corrosion requires specific and accurate information concerning the accompanying surface chemical and structural changes. The existence of surface films, concentration gradients, and precipitates with compositions different from those of the bulk glass influences its mechanical, electrical and overall interfacial behavior. Previous studies have relied primarily on glass powder-solution analysis techniques which, at best, provide only an indirect measure of corrosion. Furthermore, the geometrical parameters necessary for precise corrosion evaluation of powders have not been incorporated into the solution analysis equations. Several new techniques are presently available for characterization of glass corrosion. These tools are attractive because they provide a direct measure of corrosion extent on the "whole item" or bulk surface. Three techniques for evaluating glass durability employing EMP, IRRS and SEM are described in the preceding chapters. A 66

PAGE 79

67 fourth method, AES, has been demonstrated by previous investigators to be an excellent tool for glass surface 25 54 25 analysis. Â’ Furthermore, Pantano ejt a_l_. have shown that AES coupled with ion milling is a useful method for determining chemical profiles in glass corrosion films. These tools, together with the more sensitive solution analysis equipment now available, can yield a thorough characterization of glass durability. Moreover, when applied throughout a systematic analysis of a series of glass compositions and environments, optimum batch compositions and thermal processing requirements can be determined. Aqueous corrosion of soda-lime-silica glasses has previously been studied by numerous researchers employing i o c c c c _ c q solution analysis. Â’ Â’ Â’ Â’ The objective here is to combine the five techniques discussed above, EMP, IRRS, SEM, AES and solution analysis, into a packaged system for evaluating aqueous corrosion of these glasses. The data obtained from each techniqueare interpreted singularly and in comparison wi th those obta i ned using the other techniques. This permits, for each technique, an evaluation of the type of information, and the depth within the sample from which this information is obtained. The corrosion characteristics of a soda-silica and a soda 1 imes i 1 i ca commercial glass are evaluated and

PAGE 80

68 the results are compared to those of the soda 1 i mes i 1 i ca glasses. This provides information concerning the role of CaO and other commercial additives in glass corrosion. Experimental — Corrosion and Instrument Techniques The compositions of the glasses investigated are listed in Table 2. The glass melts containing only N a 0 , CaO and S i 0 2 were prepared from reagent grades Na^CO^, CaCOg and 5.0 pm Min-u-Sil. The mixtures were homogenized in covered Pt crucibles maintained at 1400-1550°C for 24 h in an electric muffle furnace. The commercial composition was prepared from raw materials commonly used in the glass industry.* The glass was melted in a refractory clay crucible** at 1350°C for 24 h in order to simulate, as closely as possible, those procedures used for producing glass containers. The strain point of this glass was 518°C and the annealing point was 559°C.*** Slabs of all the glasses, 3/8 in. thick, were cast between two graphite plates and specimens 3/4 in. square were diced and used for corrosion studies. The *Anchor Hocking Glass, Jan. 1975, Jacksonville, Fla. **DFC, Ceramics, Inc., Canon City, Colo. ***Anchor Hocking Laboratories, Lancaster, Ohio.

PAGE 81

Glass Compositions Investigated 69 o CM LO o o CD IE o 00 CsJ o •p00 IO O O O O O O CM co 00 r*^ vo vo o 03 CJ o C\J o 00 o 00 03 O OO o CM O CM O CO VO * Determined by X-ray Spectrochemi cal Analysis, Siemens X-ray Fluorescence Unit, Anchor Hocking Glass, Lancaster, Ohio.

PAGE 82

70 commercial glass was annealed at 500°C for 6 h and all other glasses were annealed at 450°C for 4 h. The specimens were stored in a desiccator before corroding in a controlled static environment. Prior to exposure, the samples were polished to 600 grit with SiC paper, to remove any surface films which might have formed during casting or from atmospheric reactions. The cor. 42 . R rosion cell consisted of a Teflon cube with a cavity 3/8 in. in diameter and 3/8 in. deep, filled with distilled, deionized water (pH = 6.5). These cell dimensions provide an effective exposure surface area to solution volume ratio of 0.77 cm \ The samples were sealed over the cavity with Teflon gaskets and the assembly was immersed in a hot water bath maintained at 100°C for various times, up to 20 d. After removal from the corrosion cell, the corroded side of the sample was placed on tissue paper and allowed to dry before being placed in the desiccator. Freshly abraded and corroded specimens were submitted to IRRS. The samples were scanned from 1400 cm ^ to 400 cm ^ using a medium scan rate. A freshly abraded fused quartz sample was scanned along with the test specimen to determine the effect of the various glass additives on the Si-0 stretching peak.

PAGE 83

71 The EMP studies were conducted using the method described by Clark e_t al^. for soda-lime-silica glasses. An electron beam 100 ym in diameter was used to scan parallel to the glass surface at 150 ym/min. A fixed-time (10 s) count was taken at 1 min intervals for a period of 5 min to obtain the average Kct X-ray intensities for Na, Ca, A1 and Si. A KAP crystal was used for Na, A1 and Si detection and a PET crystal was used for Ca detection, all in conjunction with an Arme thane proportional detector. Solution measurements were made on the diluted solutions prepared from the corrosion cells. The concentrations of Si, Ca and A1 were determined by atomic absorption (AA),* and the concentrations of Na and K were determined by atomic emission (AE).** Indicator paper*** was used to check the pH of these solutions. The AES**** was performed on the glass specimens 2 using the technique described by Pantano e_t aj_. *Atomic Absorption Spectrophotometer, Model 303, Perkin-Elmer Corp., Norwalk, Conn. **Atomic Emission Spectrophotometer, Model B, Beckman Instruments, Fullerton, Calif. ***Gal 1 ard-Schl es i nger Chemical Mfg. Corp., Carle Place, N . Y . ****Cyl i ndrical mirror analyzer. Physical Electronics Industries, Inc., Edina, Minn.

PAGE 84

72 Chemical profiles were determined for Si (1619 eV), Ca (291 eV), and 0 (510 eV) at various corrosion times using an Ar ion-milling apparatus to remove the surface material while simultaneously analyzing with AES. The rate of material removed by this particular ion miller was found o to be 30 A/min for most oxides previously investigated. Ion milling was continued until the Si to 0 ratio became constant, at which point it was assumed that the bulk glass had been reached. Results and Discussion of Aqueous Corrosion Data for the Various Techniques The infrared reflection spectra for the 20 Na 2 0 — 80 S i 0 2 snd 20 N a 2 0 — 10 CaO — 70 S i 0 2 glasses after identical corrosion conditions are compared in Figure 19. The peak occurring near 1050 cm" 1 (Figure 19a) for the freshly abraded (f.a.) glass is assigned to a symmetric SiO-stretching vibration in an alkali environment. 23 The other peak, appearing on the shoulder of the first peak at 950 cm \ is assigned to the Si-0 nonbridging oxygen (NS). The overlap of these peaks is caused by coupling interactions between these two vibrations. After corrosion of the binary glass for 1 h, there was a large increase in the height and a shift to higher wavenumbers of the Si-0 stretching peak whereas the NS peak height

PAGE 85

Figure 19. Infrared Reflection Spectra for Corroded Glass Specimens: (a) 20 NaoO— 80 SiO? (b) 20 Na 2 0 — 10 CaO — 70 S i 0 2 .

PAGE 86

74 (a)

PAGE 87

75 Wavenumber (cm' 1 ) (b) Figure 19 continued

PAGE 88

76 decreased and shifted to lower wavenumbers during this same period. This behavior indicates the formation of a S i 0 ^ rich layer on the glass surface.^ Longer corrosion times [>_ 3 h) indicate dissolution and eventually complete breakdown of the S i 0 ^ layer causing surface roughening as illustrated by the 12 h spectrum shown earlier in Figure 11. In contrast, the dealkalization of the soda-lime-silica glass with an equivalent Na^O content proceeds at a much slower rate as evidenced by the presence of a strong NS peak after 12 h of corrosion (Figure 19b). Figure 19 also illustrates the effect on the Si-0 and NS vibrations of adding CaO to the glass. The IRRS maxima for the freshly abraded ternary glass are shifted to slightly lower wavenumbers, and the height of the coupled region is greater than for the binary glass. The rate of change in wavenumber for the maxima is slower during corrosion for the ternary glass than for the binary glass during corrosion. Reflection spectra corresponding to the later stage of corrosion for the soda1 i mes i 1 i ca glass are shown in Figure 20. The gradual shifting of the Si-0 peak to higher wavenumbers indicates that the SiO^ layer was still developing after 9 d of exposure. This peak did not attain the magnitude that it did for the soda-silica glass even though the NS peak completely vanished after 9 d of corrosion. This apparently results from the presence of Ca

PAGE 89

20mole% Na 2 O~ 10 mole%CaO 70 mole% SiO» 77 E L_ a> -O E 3 C a) > o £ o 00 Figure 20. Infrared Reflection Spectra for the Later Stage of Corrosion for the 20 N a 0 0 — 10 CaO — 70 S i 0 0 Glass.

PAGE 90

78 in the Na depleted zone, as indicated by the comparison of the 2 d sample and the vitreous silica spectra in Figure 20 to the 1 h spectrum in Figure 19a. Thus, the Na depleted layer on the ternary glass is rich in both S i 0 2 and CaO. The decrease in the magnitude of the Si-0 peak between 5 d and 9 d is probably the result of surface roughening. However, roughening does not alter the wavenumber of the peak maxima. These spectra show that the surface breakdown and roughening is not as extensive in the soda-lime-silica glass after 9 d as it is in the sodasilica glass with an equivalent quantity of Na^O, corroded for 12 h. SEMs reveal the extent of surface damage to each of these glasses in Figure 21. The peaks occurring at 450 cm"^ in Figure 20 result from the rocking vibrations of the Si-0 tetrahedra, and are not as sensitive to composition changes as the stretching vibrations. Infrared reflection spectra for a variety of freshly abraded and corroded soda-lime-silica glasses are shown in Figures 22-25. The NS peak is not observed for the 10 Na^O — 10 CaO — 80 S i 0 2 glass, but the small Si-0 peak shift after 9 d of corrosion suggests some dealkalization (Figure 22). Figure 18b shows that the 5 per cent Na^O remains in the glass surface (within a depth of 0.5 ym) af ter th i s per i od of time. Increasing the quantity of CaO

PAGE 91

79 Figure 21. SEMs of Corroded Glass Surfaces: (a) 20 N a o 0 — 80 Si 09, 1 2 1 00 ° C (lOOOx) (b) 20 Na20 — 10 CaO — 70 S i 0 2 •> 9 d-100°C (lOOOx) (c) Commercial Glass, 20 d 1 00 ° C ( 2500x).

PAGE 92

80 o o C\J O O O O 00 CD
PAGE 93

10mole%Na 2 O-20mole%CaO-70mole%SiO Corroded; 100°C 81 o C\J O d% Wavenumber [cm " 1 Figure 23. Infrared Reflection Spectra for Freshly Abraded and Corroded 10 Na — 20 CaO — 70 S i 0 0 Glass.

PAGE 94

82 o C\J 0 3 O CNJ “O
PAGE 95

30mole% NaiO-10mole%CaO" , 60mole%SiO 83 1 O O a% Figure 25. Infrared Reflection Spectra for Freshly Abraded and Corroded 30 Na„

PAGE 96

84 to 20 per cent in the 10 per cent N a 2 0 glass does not significantly affect the rate of dealkalization even though a NS peak is observable in this glass (Figure 23). However, it does significantly increase the extent of surface roughening. Figures 24 and 25 illustrate the spectra for the 20 Na 2 0 — 20 CaO — 60 S i 0 2 and 30 Na 2 0 — 10 CaO 60 Si 0 2 gl asses in which total modifier ion concentration is the equivalent. Although the spectra for both freshly abraded specimens are nearly identicial, the corrosion behavior of the glasses is clearly different. Analysis of the Si-0 peak amplitude suggests a Si0 2 -rich layer develops in the 30 per cent Na 2 0 glass after 10 min and continues to develop for 3 h. After 6 h the NS peak disappears and the Si-0 peak begins to decrease. The Si-0 peak vanishes after 12 h suggesting severe surface roughening due to the breakdown of the Si0 2 -rich film. The SEM of this surface is shown in Figure 8b. In contrast, the amplitudes of the Si-0 and NS peaks for the 20 Na 2 0 — 20 CaO — 60 S i 0 2 glass remained constant during the first 12 h of corrosion indicating no dealkalization or S i 0 2 film development. Between 2 d and 9 d of corrosion the entire spectra decrease in amplitude significantly, but with both the Si-0 and NS peaks remaining discernible. The decrease in amplitude can be interpreted either as a surface roughening phenomena or redeposition of a new phase on the surface of the glass. The peak shifts that occur during the

PAGE 97

85 corrosion of both the 30 Na 2 0 — 10 CaO — 60 Si0 2 and 20 Na 2 0 — 20 CaO — 60 S i 0 2 glass are complicated and not as easily interpreted as the peak shifts in the glasses containing £ 20 per cent Na 2 0 a n d £ 10 per cent CaO glasses. I R R S peak reversals occur in noncorroded glasses in this compositional range (Figure 15). Several general conclusions can be made from Figures 19, 22-25. The NS peak becomes more discernible from the Si-0 peak as the total percentage of Na 2 0 and CaO increases in the glass. Smaller quantities of Na 2 0 than CaO are required to obtain equivalent Si-0 and NS peak shifts and peak separation. In general, the NS peak decreases and shifts to lower wavenumbers while the Si-0 peak increases in wavenumber during the first stage (dealkalization) of corrosion. During the later stage (> 2d) of corrosion, the Si-0 peak decreases, suggesti ng surface roughening. There is no significant difference in the corrosion behavior of the 10 per cent Na 2 0 glasses containing £ 20 per cent CaO. There is a definite improvement in the corrosion behavior of the 20 per cent Na 2 0 glass when 10 per cent CaO is added. However, 20 per cent CaO reduces the corrosion resistance of the 20 per cent Na 2 0 glass. Figure 26 presents the infrared reflection spectra for the commercial glass corroded at 100°C. The trends observed in these spectra are relatively easy to interpret since the total modifier in content is < 30 per cent

PAGE 98

86 Figure 26. Infrared Reflect ion Spectra for Freshly Abraded and Corroded Commerci al Glass.

PAGE 99

87 i.e. no peak reversals). A gradual decline in the NS peak is observed during the 20 d of corrosion with a corresponding shift in the Si-0 peak to slightly higher wavenumbers. These observations suggest a much reduced dealkalization rate and S i 0 2 film development than for the soda-silica and soda-lime-silica glasses. The relative constancy of the Si-0 peak indicates very little surface roughening after 20 d of corrosion. Figure 27 illustrates the changes in the infrared reflection spectra when CaO, Al^O^ and other constituents are added to a glass containing a constant Na 2 0 to S i 0 2 ratio. The Si-0 peak is shifted to lower wavenumbers as the modifier ion concentration is increased. Figures 19 and 22-27 illustrate the spectral changes, which are indicative of structural variations, as the quantity and type of modifier ions are varied in the glass. In general, the corrosion resistance increases as the Si-0 wavenumber (f.a.) decreases for samples exhibiting poorly defined NS peaks, whereas most glasses with well defined NS peaks exhibit inferior corrosion resistance. Solution data presented in Table 3 for the 20 Na^O — 80 S i 0 2 and 20 N a 2 0 — 10 CaO — 70 S i 0 2 glasses show that the pH and solution ion concentrations increase at a more rapid rate for the binary glass. The results of 2 gQ Wang and Tooley ’ show that the major ions in solution

PAGE 100

88 4 > -Q E 3 C a> > Figure 27. Infrared Reflection Spectra for Freshly Abraded Soda-Silica, Soda-LimeSilica and Commercial Glasses with a Constant N a 0 0 in S i 0 0 ratio.

PAGE 101

89 Table 3 Solution Data for the 20 Na?0 — 80 S i 0 2 and 20 Na20 — 10 CaO — 70 S i 0 2 Glasses Corroded in Static Aqueous Solution at 100°C Glass Time (h) . PPm pH Na Ca Si a e 20-80 0.083 7 31 — 36 .49 75 0.5 9 50 — 85 .70 73 1 10.5 204 — 391 .79 208 2 10.5 393 — 683 .72 531 3 10.5 1 . 6x1 0 4 6.0xl0 4 c t; precipitate 12 10.5 2.5x10° 8.5x10° formation 20-10-70 0.167 6.5 10.5 4 9.7 .96 23 3 8.0 38.0 11 268 .37 91 12 9.7 73.3 37 50.0 .30 233 48 10.1 500 47 143 .20 9144 216 10.8 566 108 566 .58 507

PAGE 102

90 during aqueous corrosion of soda-lime-silica glass are 2 + + 2 + S i 0 ^ , OH , H , Na and Ca and that the major reaction products are mixtures of H^SiO^, Na^SiO^ and CaSiO^. The relative quantities of these constituents in solution are dependent on their solubilities, bulk glass composition and duration of exposure. Solution parameters, a and e, were calculated for both glasses using equations 7 and 8. The solution parameter, a, is a measure of the extent of preferential corrosion. As a 0 selective leaching of one ion, usually Na, is indicated and as a -* 1 uniform total dissolution is favored. The parameter, e, is directly proportional to the leached layer thickness and positive changes in e represent an increase in thickness while negative changes in e suggest layer de Struction. However, the parameters a and e should only be used for analysis of selective leaching and film development in solutions where the solubility limits of the various ions are not exceeded. Corrosion solutions involving precipitate formation can give misleading values of a and e. Table 3 shows that a increases from 0.49 after 5 min of corrosion to 0.72 after 2 h, indicating selective Na dissolution for the binary glass. During this period, e continuously increases indicating an increase in thickness for the Si02~rich layer. After 3 h the Si to Na

PAGE 103

91 ratio in solution is greater than for the bulk glass and the values of a and e are meaningless. This suggests the precipitation of a Na-rich compound from solution. Evidence of a precipitate on the 12 h sample, probably 2 g Q NaSiO^, Â’ is shown in Figure 21a. The a and e values for the ternary glass in Table 3 show that preferential Na dissolution and film development are the primary corrosion mechanisms during the first 2 d. The value of a ( 0 . 58 ) suggests that some selective Na leaching is still occurring after 9 d of corrosion, but the decline in the value of e indicates that total dissolution is also an important corrosion mechanism during this period. The decline in the infrared spectra between 2 d and 9 d, shown in Figure 20, could be suggestive of a decrease in the SiC^-rich layer thickness. The concentration of Ca in solution increases during the 9 d of corrosion (Table 3). During the first 2 d, the Ca to Si ratio is greater than for the bulk glass, suggesting the preferential dissolution of Ca during this period. After 9 d, the ratio decreases to a value slightly less than for the bulk glass. Wang and Tooley 2,60 observed similar behavior in the Ca concentration for aqueous corrosion of an analogous soda-lime-silica glass, and suggested a counter diffusion of Ca from the solution to the glass. Ca diffusion to the glass could result in a

PAGE 104

92 protective CaSiO^ film precipitating onto the glass surface. A precipitate could also account for the decrease in infrared spectra between 2 d and 9 d (Figure 20). Table 4 presents solution data for a variety of soda-lime -silica glasses. The 10 Na 2 0 — 10 CaO — 80 S i 0 2 and 30 Na 2 0 — 10 CaO — 60 S i 0 2 exhibit corrosion behavior similar to the ternary glass in Table 3. However, the rates of pH change, Na dissolution and film growth are lower for the 10 per cent Na 2 0 and higher for the 30 per cent Na 2 0 glass than for the 20 per cent Na 2 0 glasses.* Selective leaching is the dominant mode of corrosion during the 9 d for the 10 per cent Na 2 0 glass, as evidenced by the relatively constant value of a(=0.25) and the large increases in e. The 30 per cent Na 2 0 glass changes from a primary selective leaching mode after 2 d to a combination selective leaching — total dissolution made after 9 d. Comparison of the e values for equivalent corrosion times reveals that the 30 per cent Na 2 0 has a much thicker leached layer than the other glasses investigated. The rate of dealkalization and film growth for the glasses as determined from solution analysis decreases as the CaO content is increased to 20 per cent. However, the large decrease in the infrared spectra of the glass containing 20 N a 2 0 — 20 N a 2 0 — 60 S i 0 2 revealed either *The relative rates of film growth can also be seen in the infrared reflection spectra (Figures 20b, 22, 24).

PAGE 105

93 Table 4 Solution Data for Various Soda-Lime-Silica Glasses Corroded in Static Aqueous Solution at 100°C Glass Time (h) PPm pH Na Ca Si a e 10-10-80 1 6.5 16 1.3 15 .35 56 12 8.9 27 2.8 25 .34 97 48 9.3 175 15.6 110 .23 737 216 9.5 659 10 412 .25 2472 10-20-70 3 6.5 22 4 12 .32 51 12 8.0 28 10 32 .57 48 48 9.2 142 22 95 .40 285 216 10.0 369 21 492 .88 134 20-20-60 0.167 6.5 6 2 7 .93 1 1 8.0 15 4 16 .90 4 12 10.2 43 28 69 1 .0 0 48 10.3 389 65 147 .35 546 216 10.4 325 102 325 .81 152 30-10-60 0.167 10 64 2 32 .53 57 1 11.0 168 2 40 .26 228 12 12.0 433 12 123 .30 574 48 13.0 2. 3x1 0 4 14 1 . 2xl0 3 .06 37600 216 13.5 3. 9x1 0 4 27 9. 6x1 0 3 .27 44800

PAGE 106

94 significant surface roughening or surface precipitation after 2 d of corrosion. This demonstrates that solution analysis alone does not always yield an accurate evaluation of glass corrosion. The Ca to Si ratios in solution are either equal to or slightly greater than the ratios in the bulk glass during the first 2 d of corrosion for all glasses investigated. However, after 9 d, the ratio in the solution is much less than that for the glass in all but the 20 Na 2 0 — 20 CaO — 60Si0 2 glass. These results are consistent with those of Das and Douglas 10 for a series of alkalialkaline earth glasses. Solution data shown in Table 5 illustrate the effects on corrosion of adding small additions of ^ 1 2 0 ^ and K 2 0 to a soda-lime-silica glass. The rates of change in pH, film formation and ion dissolution are significantly reduced in comparison with a comparable soda-lime-silica glass (20 Na 2 0 — 10 CaO — 70 S i 0 2 ) . Both the values of a and pH (<9) suggest that the major mechanism of corrosion after 20 d is selection dissolution of the Na. The concentrations of Ca, Si and K do not change significantly and no A1 was detected in the solution dur i ng th i s per i od of time. Based on the Na solution concentration and density of the glass (2.502 g/cc)* the affected surface thickness *Agr density comparator. Anchor Hocking Labs, Lancaster, Ohio.

PAGE 107

95 o o O in o co . — «3 CD 1X3 r— C <0 o •r— *r— O +J S3 CD i— E O E c/1 O O co UD 3 3 -Q ofO sc (— o -t-> (O 1 X 3 -(-> q cr> c c O T+-> -O 3 i. s_ o c_> (J CM r— i OO 00 oo CM i CM CM «5tO'* LO CO CM 3 CO CT> i <3^3" *=3i 00 CO oo CO oo OO oo o O o o O O o r^ 1 LO VO LO r — 1 r— CM m r^ CD OO co CD •3CO Q. • • • . • • CO VO c-00 00 -C —
PAGE 108

96 due to selective leaching was 1.14 pm after 20 d. The presence of a concentration gradient will increase this value slightly. Table 6 gives the reaction rate constant, K, of selective leaching for all the glasses investigated. These 1 3 constants were determined from solution data for pH £ 9 using equation (1).* The replacement of S i 0 2 with up to 20 per cent CaO significantly reduces (a factor of 5) the rate of selective leaching in glasses containing 10 per cent and 20 per cent Na^O. The reaction rate is further reduced (a factor of 10) by the addition of A 1 2 0 ^ and 1^0 (commercial glass). The EMP X-ray intensities for Na, Si and Ca in the 20 Na^O— 10 CaO — 70 S i 0 2 are shown in Figure 28 as a function of corrosion time. These data suggest that a majority of the Na is depleted from the sample surface after two days of corrosion. This is consistent with the infrared spectra which shows a similar Na depletion rate. The Na X-ray intensity, which remains approximately constant at a low value after 2 d, is probably due to the Na concentration gradient extending from the bulk glass through the leached layer. The corresponding increase in The log Q versus log t plots of this equation yielded slopes of = 1/2 for all the glasses, and the intercepts at log t = 0 provided the K for each glass.

PAGE 109

Reaction Rate Constants of the Glasses Calculated from Equation (1) Using 100°C. Solution Data. 97 {/) {/) o o o o ro r-. CO r— i i i 1 CD o o o o o 00 r ~ r— CM 1 CM 1 o o l o 1 o O CM <\J r— 1 CM o O r— I o CO Commercial 1 .25

PAGE 110

98 Corrosion Time (days) Figure 28. EMP X-ray Intensities as a Function of Corrosion Time for the 20 Na 2 0— 10 CaO — 70 Si0 2 Glass.

PAGE 111

99 the Si X-ray intensities, up to 2 d, suggesting a S i 0 2 rich layer on the glass, is the result of their greater escape depth caused by the absence of Na in the surface. Na is a relatively strong absorber of Si X-rays. The Ca X-ray intensity for samples corroded for < 4 d is not significantly different from that of the freshly abraded sample. The Ca intensity does increase after 5 d and remains at this new level throughout 9 d of exposure. This behavior suggests that the surface layer of the glass is being enriched in Ca either as a result of diffusion from the bulk or by the deposition onto the glass as a Ca-rich precipitate. The EMP X-ray intensities of Na, Ca, Si and A1 for the commerical glass are shown in Figure 29 as a function of corrosion time. The major feature of this graph is the steady decrease in the Na X-ray intensity during the 20 d of exposure. The relative constancy of the Ca, Si and A1 X-ray intensities suggest that preferential Na dissolution is the principal corrosion mechanism during this period. These results are in agreement with those of solution analysis which also suggest selective Na leaching throughout 20 d of corrosion. The Na intensity decreases to approximately 40 per cent of the original value after 20 d. In the absence of a Na concentration gradient, the effective thickness of the leached layer

PAGE 112

1000 loo [SCb] , Corrosion Time [days] Figure 29. EMP X-ray Intensities as a Function of Aqueous Corrosion Time for the Commercial Glass.

PAGE 113

1 01 is approximately 1.0 ym based on an EMP sampling depth of 1.5 ym. This is in good agreement with the thickness value (1.14 ym) calculated from solution data. The depth chemical profiles, determined by AES and Ar ion milling, for the freshly abraded and several corroded 20 Na20 — 10 CaO — 70 S i 0 2 glasses, are shown in Figures 30-33. The Si signal attained a constant value o after the first minute of ion milling (=30A) for all of the samples. Furthermore, the Si signals appear to be independent of corrosion time since the corroded samples yield Si signals that are equivalent to that from the freshly abraded sample. This result implies that the amount of Si in the surfaces of the corroded specimens is the same as in the freshly abraded specimens and that corrosion produces no concentration gradient. Apparently, the absence of Na does not enhance the Auger Si signal, as it does the EMP X-ray intensity, due to the small escape depth of the Auger electrons. The Ca signal is high at the "near" surface for fully abraded and corroded glasses but decreases rapidly suggesting the presence of a thin Ca-rich precipitate on the surface of the glasses.* The Ca signal reaches *The 0 signal decreases simultaneously with the Ca signal implyino the existence of a thin surface film (possibly CaSi03)2>60 rich in both Ca and 0 compared to the bul k gl ass .

PAGE 114

Auger Signal ( Pea k-toPeak) 102 Depth of Analysis [£) Ion Milling Time (min) Figure 30. Auger Signals at Various Depths in the Freshly Abraded 20 Na„0 — 10 CaO — 70 Si0 o Glass. <2

PAGE 115

1 03 Depth of Analysis (>£) Figure 31. Auger Signals at Various Depths in the 20 Na 9 0 — 10 CaO — 70 Si0 9 Glass Corroded for 2 d at 100°C. ^

PAGE 116

104 Depth of Analysis (A°) Ion Milling Time (min.) Figure 32. Auger Signals at Various Depths in the 20 Na 9 0 10 CaO — 80 Si0 9 Glass Corroded for 5 d at 100°C. ^

PAGE 117

uger Signal (Peak-to-Peak ) 105 Depth of Analysis (A) Figure 33. Auger Signals at Various Depths i-n the 20 Na 9 0 — 10 CaO — 70 Si0 9 Glass Corroded for 9 d at ^ 100°C. L

PAGE 118

a minimal value 200 A below the surface and remains at this value to a depth depending on the extent of corrosion. The data illustrated in Figure 34 show that the o Ca depleted region extends to 1000 A for the 2 d sample o and to = 2200 A for the 9 d sample. A Ca-rich zone develops directly beneath the depleted region as evidenced by the increase in the Auger signal. The concentration of Ca in this region increases with corrosion time and apparently after 5 d is sufficient to give a detectable increase in EMP X-ray intensity (Figure 28). The low Ca regions are thought to be caused by the precipitate forming on the near surface and depleting the Ca in the adjacent reg ions. The Ca-rich and Ca-depleted regions observed in the corroded samples are also found in the freshly abraded sample, but to a much smaller extent. This is probably due to the reaction of the abraded surface with atmospheri moi sture . The chemical depth profiles for freshly abraded and 10 d corroded glass specimens are illustrated in Figures 35-36. An excess of Ca is observed on the freshly O abraded glass and extends a few hundred A below the surface. As with the ternary glass, a Ca-depleted region exists directly beneath the Ca-rich zone and extends o through several hundred A. Excess Ca is not observed on

PAGE 119

Auger Signal (Peak-to-Peak) 107 Depth of Analysis (A) Figure 34. Auger Signals for Ca as a Function of Depth for Various Samples Corroded at 100°C.

PAGE 120

Depth (a) 108 |eu3js J03n\/ Figure 35. Auger Signals at Various Depths in the Freshly Abraded Commercial Glass.

PAGE 121

1 09 -o (^ead-oj-^ead) |eu3|s jaSny Figure 36. Auger Signals at Various Depths in the Commercial Glass Corroded for 10 at 100°C.

PAGE 122

no the near surface of the 10 d sample but a Ca-depleted region is present. The absence of excess Ca could be due to its diffusion into solution during corrosion. Its absence may also be related to a strong A1 signal observed on the near surface during the first 30 s of ion milling. Unlike the ternary glass, the Si signal for the commercial glass behaved in a manner opposite to that of the Ca signal. The Si signal was low on the near surface and increased in the Ca-depleted zones. The depth of the Ca-depleted zone for the commercial glass is also dependent upon the corrosion time. This depleted region extends to twice the depth for the 10 d as for the freshly abraded sample. The Ca-rich zones present below the depleted zone for the ternary glass were not found in the commercial glass. This could be due to a differential in milling rate between the commercial and ternary glasses. Summary Preferential Na dissolution and simultaneous pH increases were observed for all the glasses in the early stage of corrosion. The rates of dealkalization and pH change are much greater for the soda-silica than for the soda-lime-silica glass containing an equivalent quantity

PAGE 123

1 1 1 of Na 2 0. After 1 h of corrosion, the majority of the Na has diffused from the surface of the binary glass and complete breakdown of the Si0 2 ~rich layer has begun. IRRS and EMP analysis reveal that a signifcant quantity of Na is present in the 20 Na 2 0 — 10 CaO — 70 S i 0 2 glass surface after 2 d of corrosion. IRRS and solution analysis demonstrate that S i 0 2 rich films due to Na depletion develop on all of the sodalime-silica glasses during aqueous corrosion. Furthermore, the films formed on these glasses (with the exception of the 30 Na 2 0 — 10 CaO — 60 Si0 2 and 20 Na 2 0 — 20 CaO — 60 Si0 2 ) are relatively resistant to attack at pH > 9 where total dissolution occurs for the binary glass. Film parameters and reaction rate constants calculated from solution data show that the rate of dealkalization and subsequent S i 0 2 film growth are reduced as the quantity of CaO in the glass is increased for glasses containing <_ 20 per cent Na 2 0. The additions of small quantities of K 2 0 and A^O^ to a soda-lime-silica glass (commercial) reduces its dealkalization rate by a factor of 20. Compared with the binary glass, the decrease in the early stage corrosion rate for the soda-lime-silica and commerical glasses could be interpreted as an increase in coupling interactions between Si-0 and NS bonds influenced by the presence of the modifier ions (Ca, A 1 ,

PAGE 124

112 K). The stronger structural coupling retards the dealkalization of the glasses in the early stage (pH < 9) and provides a stable Si0 2 film that resists attack during the later stage of corrosion (pH > 9). The excess Ca observed on the glasses with AES suggests the presence of a Ca-rich precipitation such as CaSiO^.^Â’^ El Shamy ejt a 1 . ^ have suggested the possibility of the surface nonbridging oxygen atoms preferring Ca ions to either H ions or N ions, which would also explain the excess Ca on the glass surface. Regardless of the form (i.e. precipitation or ionic), the presence of Ca on the near surface strongly influences the corrosion characteristics of the glass.

PAGE 125

CHAPTER V APPLICATION OF THE ELECTRON MICROPROBE, INFRARED REFLECTION SPECTROSCOPY AND OTHER TECHNIQUES FOR EVALUATING THE WEATHERING BEHAVIOR OF SODA-LIME-SILICA GLASSES Introduction — Review of Weathering and Comparison with Aqueous Corrosion The weathering behavior of glass is interesting to the archeologist because it provides a means for determining the historical background of glass fragments.^ In contrast, the attention of the engineer is focused on systematic evaluations of weathering behavior in order to predict the suitability of glass for particular applications. Weathering is categorized by the extent of aqueous precipitation that accumulates on the glass surface during exposure to humid atmospheres. Type 1 weathering occurs when sufficient quantities of moisture collect on the surface to form visible droplets which eventually lead to "runoff." During type 2 weathering, a thin layer of fog on ? o forms on the glass but evaporates before droplets form. ’ Type 2 weathering is encountered when glass is subjected to mild humi di tytempera tu re changes, such as during storage or shipment, while type 1 weathering occurs on glasses exposed to extreme climate changes or temperature 1 1 3

PAGE 126

114 differentials. These conditions would be experienced by glass of exposed window panes in the northern climates. As in aqueous corrosion, both structural and chemical changes occur within the surface of glass during weathering. In both forms of corrosion, Na ions diffuse from the bulk to the surface of the glass and H ions diffuse from the surface into the glass to occupy positions vacated by the Na ions. In static aqueous corrosion, the Na ions dissolve into solution at a rate independent of the Na solution concentrations for pH < 9. 9,13 However, for type 2 weathering, large quanti ti ties of sol uti on are not available to remove the Na ions from the surface. Consequently, these ions remain on the glass surface and eventually react with COg present in the atmosphere to form stable Na^COg. Â’ The presence of the precipitate on the surface reduces Na ion diffusion due to the decreased area of glass exposed, and to the decreased Na concentration gradient. During type 1 weathering, Na ions are dissolved into solution but large glass surface area to solution volume ratios and continuous solution replenishment alter the reaction kinetics from those of static aqueous corrosion. 20-22 Simpson developed a "haze meter" technique and demonstrated a correlation between the quantity of light scattered from a glass surface and the duration of

PAGE 127

exposure for type 2 weathering. The scattered light was attributed to an increase in surface roughness due to solid precipitate formation on the glass. A similar correlation was shown for type 1 weathering, but the extent of light scattering was reduced due to precipitate "washoff." Sanders and Hench^ investigated the two types of weathering using infrared reflection spectroscopy and showed that the reflected amplitude was dependent on both the extent and type of weathering. The accumulation of corrosion products on the surface of the glass (type 2) was demonstrated to influence the reflected amplitude in a fashion similar to compositional changes. Thus, infrared amplitude analysis alone appears insufficient for characterizing glass weathering. An extensive evaluation of the structural and chemical changes accompanying weathering has been hindered by the absence of adequate methods for analysis. Several quantitative techniques discussed in the previous chapters provided direct analysis of static aqueous glass corrosion. These techniques are now used for evaluating the weathering behavior of a sy sterna ti c series of soda-lime-silica glasses and a soda-lime-silica commerc i a 1 glass.

PAGE 128

116 Experimental Weathering Parameters Silicate glasses containing from 10 to 30 per cent Na^O and from 10 to 20 per cent CaO, and a commercial soda-lime-silica glass were prepared by the methods discussed in Chapter 4. Samples 3/4 in. square, and 3/8 in. thick, were tilted at 45° on a Teflon R sheet raised 2 in. above a tray of water. The assembly was enclosed in an oven and maintained at room temperature (25°C + 2°C). Specimens of the commercial glass were similarly weathered at 50°C + 4°C and at 100°C + 4°C in the oven. The water vapor volumes associated with the three temperatures invesitaged are 703 ft /lb , 194 ft/ lb and 26.8 ft 3 / 1 b m for 25°C, 50°C and 100°C, respectively. Water precipitation runoff was not observed on the glasses weathered in the oven. Samples of the commercial glass were also weathered by placing on the Teflon sheet raised 2 in. above water level in a covered pan. The water in the container was maintained at 50°C + 4°C and at 100°C (boiling). The large pools of water that formed on the glasses were due to the lower specimen temperature ( 10°C) in relation to the temperature of the water, and from condensation falling onto the samples from the cover. All of the glasses were analyzed with IRRS, and selected samples were analyzed with AES, SEM and EMP. The operating conditions for these instruments

PAGE 129

117 and the analytical techniques employed were the same as those outlined in the previous chapter. Results and Discussion of Weathering Data Infrared reflection spectra for the soda-limesilica glasses weathered up to 28 d at 25°C-100 per cent R.H. are shown in Figures 37-41. Significant differences can be seen between these spectra and the aqueous corrosion spectra of similar specimens (Figures 22-25). The total infrared reflected intensities continuously decrease as the exposure time increases for all of the weathered glasses. In contrast, the Si-0 stretch peak intensities increased during the first stages of corrosion while the nonbridging oxygen peak (NS) decreased during the entire course of exposure for the aqueou scor roded glasses. Generally, an increase in the Si-0 stretch peak intensity is associated with an increase in the volume fraction of S i 0 2 present in the surface of the glass. However, it was demonstrated previously that an increase in surface roughness due to either abrading or precipitate formation can also decrease the intensity of this peak even in the presence of a Si02~rich film. An important similarity between the weathered glass spectra and the aqueouscorroded glass spectra is the systematic peak position

PAGE 130

118

PAGE 131

119 y% Figure 38. Infrared Spectra for Weathered 10 Na 9 0— 10 CaO — 80 Si0 o Glass.

PAGE 132

120 Figure 39. Infrared Spectra for Weathered 10 Na o 0 — 20 CaO — 70 SiO Glass. L

PAGE 133

121 a% Figure 40. Infrared Spectra for Weathered 30 Na o 0— 10 CaO— 60 S i 0 o Glass.

PAGE 134

122 Figure 41. Infrared Spectra for Weathered 20 Na 9 0 — 20 CaO— 60 Si0 o Glass.

PAGE 135

123 shifts that occur during exposure. The Si-0 stretch peaks shift to higher wavenumbers and the NS peaks shift to lower wavenumbers, relative to the freshly abraded peaks, for both forms of corrosion. The Si-0 stretch peak positions have been shown to be directly related to the composition of the glasses and not influenced by the surface roughness. Increases in the wavenumber for this peak correspond to increases in the volume fraction of S i 0 2 present in the glass surface. The simultaneous increase in wavenumber and decrease in amplitude of the Si-0 stretch peak for the weathered glasses suggest the formation of a precipitate over a SiO^-rich film. The rates of precipitate formation and S i 0 2 film development are dependent on the composition of the glass. Generally, these rates increase as the percentage of Na^O increases for glasses containing 10 CaO (Figures 37, 38, 40). Increasing the percentage of CaO from 10 per cent CaO to 20 per cent CaO also increases the rates of precipitate formation and S i 0 2 film development (Figures 39, 41). Hence, the weathering resistance is directly related to the per cent S i 0 2 in the glass. The infrared spectra for the glasses containing equivalent total quantities of Na^O and CaO are similar (Figures 37, 39 and 40, 41). The spectra for the 28 d weathered glasses containing 40 per cent total Na^O and CaO are equivalent in roughness to a glass surface abraded with 120 grit SiC paper. A white haze formed on most of

PAGE 136

124 the weathered glasses and a white powder was observed on the high soda glasses. The powders on the weathered 30 Na 2 0 — 10 CaO — 60 S i 0 2 glass, determined from X-ray diffraction analysis, were Na 2 C0 3 and CaC0 3> The rates of Na ion depletion from the 20 Na 2 0— 10 CaO — 70 S i 0 2 and 10 Na 2 0 — 10 CaO — 80 S i 0 2 glasses are shown in Figure 42. These graphs were determined from the Si-0 stretch peak shifts using Figures 15, 37 and 38. Comparison of Figures 18b and 42 illustrate that the rate of Na ion depletion from the aqueous corroded 20 Na 2 0— 10 CaO — 70 S i 0 2 glass is very rapid during the first 4 h of exposure in comparison with the weathered glass. However, after longer periods of exposure, the rate of Na ion depletion becomes more rapid for the weathered glass. This reversal in depletion rates is thought to be related to the pH increases that accompany aqueous corrosion and slow down Q 1 O Na ion diffusion for pH _> 9. ’ Since no solution is present at the glass surface during type 2 weathering, the rate of Na ion depletion is not dependent on pH changes. However, the kinetics of Na ion depletion during weathering are influenced by the rate of moisture adsorption on the glasses from the environment, the rate of precipitate formation and exposure temperature. The constant slope for the 10 Na 2 0 — 10 CaO — 80 S i 0 2 glass in Figure 42 during the first 5 h of exposure is thought to be associated with

PAGE 137

125 CM <0 M 00 k_ U "O _C C O) 1 (d •*— to CD c «d 3: 0 C SCO 0 0 0 •r4L_ -l-> L_ rd CD 0 E u, 4 _> c CD u c c 0 O *r0 to O O SC\J £rd O 2 : o C\1 CD sCT.

PAGE 138

126 the rate of moisture adsorption by the glass. Sanders and Hench 1 ^ observed a similar incubation period during weathering of 1 ithia-sil ica glass. Infrared spectra for the commercial glass, weathered in both the oven (type 2) and pan (type 1), are shown in Figures 43-47. During both types of weathering at 100°C the amplitude of the Si-0 stretch peak increases and the position of the peak shifts to high wavenumbers. This corresponds to the development of a SiO^-rich film on the glass surfaces. The Si-0 stretch peak amplitude, for the glass weathered in the oven for 20 d, is below that for the freshly abraded sample, but its maxima are at a higher wavenumber (Figure 45). The simultaneous wavenumber increase and amplitude decrease are suggestive of surface roughening due to either precipitate formation or network breakdown. The extent of surface roughening on the 20 d specimen is equivalent to that obtained by abrading the same glass to 320 grit with SiC paper. The amplitudes of the Si-0 stretch peak for the samples weathered in the pan for > 3 d are the same as the amplitude of the freshly abraded sample (Figure 43). However, the wavenumber position of the Si-0 stretch maximum increases continuously through the course of exposure (Figure 43). SEMs of glass surfaces for both types of weathering are shown in Figure 48. The extensive precipitate formation

PAGE 139

127 Figure 43. Infrared Spectra of Commercial Glass for Type 1 Weathering at 100°C-100 Per Cent R.H.

PAGE 140

128 Figure 44. Infrared Spectra of Commercial Glass for Type 1 Weathering at 50°C-100 Per Cent R.H.

PAGE 141

129 60 40 Od a? 20 Figure 45. Commercial Glass Oven,100"C 100% R.H. Si-O 1 1200 1000 800 Wavenumber [cm 1 ] Infrared Spectra of Commercial Glass for Type 2 Weathering at 100°C-100 Per Cent R.H.

PAGE 142

130 Figure 46. Infrared Spectra of Commercial Glass for Type 2 Weathering at 50°C-100 Per Cent R.H.

PAGE 143

131 CSJ Figure 47. Infrared Spectra of Commercial Glass for Type Weathering at 25°C-100 Per Cent R.H.

PAGE 144

132 Figure 48. SEMs of Weathered Commercial Glass: (a) Type 2, 100°C-100 Per Cent R.H., 20 d (1300x) (b) Type 1, 100°C-100 Per Cent R.H., 1 d ( 620x ) (c) Type 1, 100°C-100 Per Cent R.H., 10 d ( 1 2 OOx ) .

PAGE 145

133 observed on the 20 d oven -weathered sample accounts for the decreased infrared amplitude obtained from this specimen. Evidence of precipitate formation is also seen on the 1 d pan-weathered specimen, but water condensation and runoff removes this precipitate for exposure times > 3 d ( Figure 48b, c) . This accounts for the relative constancy of the infrared amplitudes obtained from these samples (Figure 43). The relative wavenumber positions of the Si-0 stretch maxima shown in Figures 43-46 indicate that the rate of Na ion decoupling from the glass is more rapid for type 1 than for type 2 weathering, at both 50°C and 100°C. The extent of Na ion decoupling at 25°C-100 per cent R.H. is not significant even after 20 d of exposure (Figure 47). Furthermore, these figures (43-47) demonstrate that commercial glass durability is dependent on the type of weathering, temperature and duration of exposure. The rate of Na ion decoupling from the glass network is slightly greater for specimens weathered in the pan at 100°C than for those aqueous -corroded at 100°C (Figures 26, 43). The replenished supply of water, maintaining a high Na ion concentration gradient, is thought to be partially responsible for the greater Na ion dissolution rate for the former.

PAGE 146

1 34 The EMP analysis for commercial glass specimens weathered in the pan (type 1) and in the oven (type 2) is presented in Figure 49. An initial increase in Ca X-ray intensity, observed for both types of weathering, is attributed to the formation of a Ca-rich precipitate on the surface of the glass (Figure 48). The Ca X-ray intensity for the specimen weathered in the pan attains a maximum value after 1 d of exposure and decreases sharply during the remaining 9 d. The Ca X-ray intensity for the oven-weathered samples decreases only slightly after 4 d, but remains at a higher value than for the freshly abraded glass thoughout the duration of exposure. The Na X-ray intensity continuously decreases for the pan-weathered specimens, but remains constant for the oven-weathered samples as corrosion time increases. The decrease in both the Na and Ca X-ray intensities for the pan-weathered specimens is due to the "runoff" effect. These ions are continuously removed from the glass surface by the flow of water for type 1 weathering. However, these ions remain on the surface during type 2 weathering, since there are no media present to remove them. The high Ca X-ray intensity and constant Na and Si X-ray intensities indicate that the surface reaction product produced for oven weathering is Ca-rich.

PAGE 147

1000 135 (sdD) I Corrosion Time (Days) Figure 49. EMP X-ray Intensities as a Function of Exposure Time for Weathered Commercial Glass.

PAGE 148

136 The mechanism of corrosion for type 1 weathering appears to be different from that for static aqueous corrosion (Figures 29, 49). The Na X-ray intensity decreases as exposure time increases for both forms of corrosion, but at a more rapid rate for type 1 weathering. The Ca X-ray intensity remains constant during static aqueous corrosion, but decreases significantly during type 1 weathering. These data suggest that the Ca ions remain on the surface during static aqueous corrosion (perhaps as a CaSiO^ film) and slow down Na ion diffusion from the glass. During type 1 weathering the Ca ions react with C0 2 from the atmosphere and from CaCO^. This carbonate coalesces into discrete particles which are removed from the surface by the water flow. The depth chemical profile of the commercial glass weathered in the pan for 15 d is shown in Figure 50. The surface of the glass was similar in appearance to that shown in Figure 48a. The presence of a coalesced precipitate complicates the interpretation of the profile due to heterogeneous ion milling. However, the significant portion of this profile is the presence of the high Ca signal in the weathered gl ass , in comparison with the Ca signal in the bulk glass of the freshly abraded sample (Figure 35). These data are in agreement with the EMP data for the ovenweathered specimens and suggest the presence of a Ca-rich region near the surface.

PAGE 149

Depth (a) 137 o CCD E E O o CD CD CO CD CM C o “O a; iCD 03 CD CM a) CL >, • l — “O C LO CO 4-> CL • a> cl: o 4-> co c =3 CD O O •r— Cs~ 03 CD > Cl 4-> O 03 O to I r— O tO o c o 0)0 •I — f— U1 Cto CD CO cn ru 3 I— C CD o LO CD (MB0d-O}->ie0d) |Bu8!s J9§n V S Li.

PAGE 150

138 Summary The weathering rate at 25°C-100 per cent R.H. for soda 1 i mes i 1 i ca glasses is dependent upon composition. The rate of type 2 weathering increases as the percentage of ^£0 in the glass increases. Furthermore, the total quantity of Na20 plus CaO is an important consideration in glass weathering. For instance, the extent of weathering after 28 d is approximately the same for both the 20 Na 2 0 — 10 CaO — 70 S i 0 2 and 10 Na 2 0 — 20 CaO — 70 Si0 2 glasses. Increasing the total quantity of Na 2 0 plus CaO above 20 per cent drastically reduces the weathering resistance of the glass. The kinetics of Na ion decoupling from the glass network are different for the 20 Na 2 0 — 10 CaO — 70 S i 0 2 glass weathered (type 2) at 25°C-100 per cent R.H. and for the same glass aqueouscorroded at 100°C. However, the extent of Na ion decoupling, as determined by IRRS, after 9 d is approximately the same for both forms of corrosion. This could be due to the finite sampling depth (= 0.5 ym) of the infrared spectrometer. The affected depth is probably greater for the aqueous corroded glass. The rate of Na ion decoupling in commercial glass increases as the temperature (and hence the absolute humidity) increases for both types of weathering. However,

PAGE 151

the decoupling rate is greater for type 1 (pan) than for type 2 (oven) weathering. This is due to the washoff effect characteristics of type 1, which continuosuly removes Ca and Na ions from the glass surface. These ions remain on the surface in the form of precipitates during type 2 weathering. The greater Na ion decoupling rate for type 1 weathering in comparison to aqueous corrosion is attributed to the coalescense and washoff of the protective Ca-rich film during weathering. Furthermore, the continuous removal of the Na ions during type 2 weathering maintains the constant high Na ion concentration gradient that aids ion exchange.

PAGE 152

CHAPTER VI SUMMARY AND CONCLUSIONS Glass Durability The usual problem of alkali ion diffusion encountered during EMP analysis of glasses has been minimized by electron beam enlargement and specimen translation. Furtiermore, this technique has been shown to be applicable for evaluating compositional changes that occur in glass surfaces during corrosion, although the surface roughness and alkali diffusion rate are greater for corroded glass than for bulk glass. EMP analysis of three soda-lime -silica glasses demonstrated that the initial rate of Na ion depletion during aqueous corrosion is dependent on the concentration gradient between the glass and the solution. "he rate of Na ion dissolution increases as the percentage of Na^O in the glass increases for glasses containing 10 per cent CaO. However, the initial rate (pH £ 9) of Na ion d ffusion from the glass appears to be independent of the surface area of the glass to the solution volume ratio. The frequency position of the infrared reflection maxima for soda-lime -silica glasses is dependent upon the

PAGE 153

141 glass composition. Unlike the peak amplitude, the frequency position is unaffected by roughness of the glass surface. Thus, accurate analysis of glass corrosion is possible even for glasses where extensive surface breakdown occurs. These characteristics were used for quantitative analysis of corroded glasses containing _< 20 per cent ^£0. The dealkalization rates for two soda-limesilica glasses were determined and compared to that for a binary soda-silica glass. Again, it was found that the rate of dealkalization increased as the percentage of Na^O increased for glasses containing 10 per cent CaO. Moreover, the dealkalization rate for the binary glass is greater than for the ternary glass containing an equivalent quantity of Na^O. The rate of Na ion decoupling from the binary glass increases as the solution temperature increases with definite kinetic changes occurring at pH _> 9 for all temperatures. The kinetic change is attributed to an increase in Na ion concentration at the glass surface. Quantitative compositional analysis for corroded specimens analyzed by both the IRRS and EMP techniques were in satisfactory agreement. A durability evaluation of a systematic series of soda-lime-silica glasses was performed using IRRS, EMP, SEM, AES and solution analysis. The results show that the corrosion resistance decreases as the percentage of Na 2 0

PAGE 154

142 in the glass increases. The total quantity of Na^O plus CaO is an important factor in glass durability for glasses containing _> 20 per cent Na^O. The addition of 10 per cent CaO to glasses containing 20 per cent Na^O improves the chemical durability whereas the addition of 20 per cent CaO to glasses containing _> 20 per cent Na^O is detrimental. The role of Ca in glass durability is to increase the coupling interaction between the Na ions and the glass network, and to form a protective Ca-rich film on the glass surface. Both of these conditions retard the diffusion of Na ions from the glass, and the Ca-rich film is apparently resistant to high pH. Commercial additives to soda-lime-silica glasses improve glass durability by further tightening the glass structure and increasing the coupling interaction between the Na ion and glass network. The weathering of commercial glasses further illustrates the importance of Ca in glass durability. During both types of weathering, the Ca-rich surface film coalesces into discrete particles through reaction with the atmosphere (Figures 51). The removal of these particles from the glass surface is partially responsible for the increased dealkalization rate observed for type 1 weathering as compared to aqueous corrosion. Type 1 weathering is also more severe than type 2 weathering due to the large quantities of Na and Ca ions remaining

PAGE 155

143 Protective Precipitate Aq ueous Corrosion Non-Protective Precipitate Depleted Zone Bulk Glass > Weathering Figure 51 . Schematic Comparing Static Aqueous Corrosion and Weathering. I

PAGE 156

144 on the surface of type 2 weathered glass. These ions decrease the surface exposure area and ion concentration gradients, thus retarding Na ion diffusion. Several processes that occur during glass corrosion of soda-lime-silica glasses are shown in Figure 52. The order in which the reactions occur and the relative magnitude of the reaction rate constants, K, depend upon the glass composition, reaction temperatures and type of corrosion. The reaction rate constants, as determined from various rate equations (such as equation (1)), are independent of corrosion time for a particular mechanism. During the first stage of corrosion (pH £ 9), the ratecontrolling reaction is the ion exchange between H ions from solution and Na ions from the glass. At some pH value > 9, the ion exchange mechanism is still rate controlling, but the rate of exchange has been decreased due to a Na ion concentration increase in the surface of the glass. Hence, the reaction rate constant is correspondingly decreased. When the ion exchange process has proceeded to a point where large quantities of OH ions are present in solution, the rate-controlling mechanism is the attack of OH ions on Si-0 bonds. During this period of time, total dissolution of the glass occurs and the reaction rate constant is determined by the rate at which Si-0 bonds are broken. Ca-rich films can form on the near surface of

PAGE 157

145 c o o co ~ CT> O A -5? I <1> Q. CO C o o CO CO CO o c o E — CO — E 0 ) o — »*n co W co o c T3 O Q) 2 o o CO c <0 o T3 <-* jto o CO o> T 3 0 ) c IE o o k. *-* 03 c ( 1 ) o O CM 13 CD o VI o (/) I a> CL |) JUDJSUCQ 0|Dy UOJ4 DDd^J Figure 52. Schematic of Plausible Corrosion Mechanisms.

PAGE 158

146 the glass either by diffusion processes or by precipitation from solution. The reaction rate constant during this period of time is dependent on the rate of Na ion diffusion through the Ca-rich film and on the resistance of the film to OH ion attach. The experimental evidence shows that the rates for both of these processes are low for the soda-limesilica glasses. Durability Evaluation Techniques The combination of EMP, IRRS, AES-ion milling, SEM and solution analysis permits a more accurate evaluation of glass corrosion than can be obtained by any of the techniques employed singularly. Each technique provides unique information concerning the composition within a specified depth of the glass surface. Thus, each technique yields an averaged compositional analysis of the sampled volume, and the usefulness of each is limited by its sampling depth. IRRS analyzes most glasses to a depth of 0.5 pm. The combination of spectra frequency shifts and amplitude variations illustrates both glass dissolution mechanisms and surface feature alterations. This technique becomes relatively insensitive to surface chemical changes, due to diffusion, when the depth of the corroded layer becomes large relative to the sampled depth. When this condition

PAGE 159

147 exists, the IRRS is effectively sampling only the tail of the diffusion profile which remains relatively constant. Both precipitate formation and surface roughening during corrosion diminish the penetration depth of the infrared beam into the glass surface. In some instances, its capability of quantitively evaluating surface roughening and precipitate formation is a valuable asset. However, the SEM is required to distinguish between these two surface features. Of crucial importance for many applications is the nondestructive evaluation of the specimen that is permitted by IRRS. The EMP provides averaged compositional changes from a greater depth in the glass surface, and therefore is useful for analyzing thicker corrosion films than is IRRS. Furthermore, the data provided by the techni que are not affected to as great an extent by surface feature changes as is the latter. Unlike IRRS, the EMP has the capability for providing complete compositional changes that occur during corrosion. Since solution analysis is independent of beam penetration depth, it provides chemical dissolution data during the entire course of corrosion. Accurate kinetics analysis for the various corrosion mechanisms, film growth and destruction, reaction rate constants and the tendency for selective leaching, can be obtained from solution

PAGE 160

148 parameters. However, interpretation of solution data can be misleading if stable films precipitate from the solution, onto the glass surface. SEM can demonstrate the presence of most precipitate, but extremely thin continuous films may require the use of AES. Since the escape depth of the Auger electron is O small (< 5 A), AES coupled with ion milling can provide chemical depth profiles of corroded glass surfaces, or a submicron resolution level, that cannot be obtained with other techniques. Verifying the presence of thin surface films with compositions different from the bulk glass is routine for AES. However, variable ion milling rates, heterogeneous ion milling, precipitate formation, and matrix effects can alter the shape and dimensions of the experimentally obtained profiles. Fortunately, these effects can be minimzed when evaluating the corrosion of a systematic series of similar glasses that has been well characterized by the other techniques.

PAGE 161

REFERENCES 1. S. Tsuchihashi and E. Sekido, "On the Dissolution of Na20 — CaO — S i 0 2 Glass in Acid and in Water," Bull. Chem. Soc. Japan, 32(8) 868-72 (1959). 2. F. F. Wang and F. V. Tooley, "Influence of Reaction Products on Reaction Between Water and Soda-LimeSilica Glass," J. Amer. Ceram. Soc., 9J|^( 12) 521-24 ( 1 958) . 3. F. R. Bacon and G. L. Calcamuggio, "Effect of Heat Treatment in Moist and Dry Atmospheres on Chemical Durability of Soda-Lime Glass Bottles," Am. Ceram. Soc. Bull., 46(9) 850-55 (1967) . 4. A. K. Lyle, "Theoretical Aspects of Chemical Attack of Glasses by Water," J. Amer. Ceram. Soc., 26(6) 201-4 (1943). 5. M. A. Rana and R. W. Douglas, "The Reaction Between Glass and Water. Part 1. Experimental Methods and Observations," Phys. Chem. Glasses, 2 { 6 ) 1 79-1 95 ( 1 961 ). 6. M. A. Rana and R. W. Douglas, "The Reaction Between Glass and Water. Part 2. Discussion of Results," Phys. Chem. Glasses, 2(6) 196-205 (1961). 7. S. M. Budd, "The Mechanisms of Chemical Reaction Between Silicate Glasses and Attacking Agents. Part I. Electrophilic and Nucleophilic Mechanisms of Attack," Phys. Chem. Glasses, 2(4) 111-14 (1961). 8. R. W. Douglas and J. 0 Isard, "The Action of Water and Sulphur Dioxide on Glass Surfaces," J. Soc. Glass Tech., 33 288-334 (1949). 9. R. W. Douglas and T. M. M. El-Shamy, "Reactions of Glasses with Aqueous Solutions," J. Amer. Ceram. Soc., 50( 1 ) 1 -8 ( 1 967 ). 149

PAGE 162

1 50 10. C. R. Das and R. W. Douglas, "Studies on the Reaction Between Water and Glass. Part 3," Phys. Chem. Glass, 8(5) 178-84 (1967). 11* T • M • M . El-Shamy and R. W. Douglas, "Kinetics of the Reaction of Water with Glass," Glass Tech., H(3) 77-80 ( 1 972 ). 12. R. M. Barrer, "Diffusion in Spherical Shells, and a New Method of Measuring the Thermal Diffusivity Constant," Phil. Mag., 35(12) 802 (1944). 13. T. M. N. El-Shamy, J. Lewins and R. W. Douglas, "The Dependence on the pH of the Decomposition of Glasses by Aqueous Solutions," Glass Tech., 13(3) 81-87 ( 1 972 ) . 14. D. M. Sanders and L. L. Hench, "Mechamisms of Glass Corrosion," J. Amer. Ceram. Soc., 56(7) 373-77 (1973). 15. L. L. Hench, "Characterization of Glass Corrosion and Durability," Journal of Non-Crystalline Solids, in press . 16. S. Sen and F. V. Tooley, "Effects of Na20/K20 Ratio on Chemcial Durability of A1 kal i -Lime-Si 1 ica Glasses," J. Amer. Ceram. Soc., 38(5) 1 75-77 ( 1 955 ). 17. D. M. Sanders and L. L. Hench, "Environmental Effects on Glass Corrosion Ki net i c s ," Amer. Ceram. Soc. Bull., 52(9) 662-65, 669 (1973). 18. R. M. Tichane and G. B. Carrier, "Microstructure of a Soda-Lime Glass Surface," J. Amer. Ceram. Soc., 44(12) 606-10 (1961). 19. R. M. Tichane, "Initial Stages of the Weathering Process on a Soda-Lime Glass Surface," Glass Tech., 7 ( 1 ) 26-29 ( 1 966) . 20. H. E. Simpson, "Study of Surface Structure of Glass as Related to Its Durability," J. Amer. Ceram. Soc., 4]_( 2 ) 43-49 ( 1 958). 21. H. E. Simpson, "Measuring Surface Durability of Glass," Am. Ceram. Soc. Bull., 30(2) 41-45 (1951). 22. H. E. Simpson, "Study of Surface Durability of Container Glasses," J. Amer. Ceram. Soc., 42(7) 337-43 ( 1 959) .

PAGE 163

151 23. D. M. Sanders, W. B. Person and L. L. Hench, "Quantitative Analysis of Glass Structure Using Infrared Reflection Spectra," Appl . Spectrosc., 26(5) 530-36 (1972). 24. D. E. Clark, L. L. Hench and W. A. Acree, "Electron Microprobe Analysis of Na?0 — CaO — SiO? Glass," J. Amer. Ceram. Soc., 58(11-12) 531-32 (1975). 25. J. P. Rynd and A. K. Rastogi, "Auger Electron Spectroscopy — a New Tool in the Characterization of Glass Fiber Surfaces," Amer. Ceram. Soc. Bull., 53(9) 412-13 (1974). 26. C. G. Pantano, Jr., A. E. Clark, Jr., and L. L. Hench, "Multilayered Corrosion Films on Bioglass Surfaces," J. Amer. Ceram. Soc., 57(9) 412-13 (1974). 27. J. L. Lineweaver, "Oxygen Outgassing Caused by Electron Bombardment of Glass," J. Appl. Phys., i34(6) 178691 (1963). 28. S. Tanaka and D. H. Warrington, "Some Electron Microscope Observations of Thin Glass Foils," Phys. Chem. Glasses, 5(3) 87-89 (1964). 29. R. V. Adams, H. Rawson, D. G. Fisher and P. Worthington, "The Use of the Electron Probe X-ray Microanalyser for the Identification of Inhomogeneity in Glass," Glass Tech. , 7 ( 3 ) 98-1 05 ( 1 966). 30. A. K. Varshneya,A. R. Cooper and M. Cable, "Changes in Composition During Electron Microprobe Analysis of KoO — SrO — S i 0? Glass," J. Appl. Phys., 3_7(5) 2199 ( T 966 ) . 31. M. P. Borom and R. E. Hanneman, "Local Compositional Changes in ALkali Silicate Glasses during Electron Microprobe Analysis," J. Appl. Phys., 38^( 5 ) 2406-07 ( 1 967 ) . 32. W. T. Kane, "Application of the Electron Microprobe in Ceramics and Glass Technology," Microprobe Analysis, C. A. Anderson, Ed., J. Wiley & Sons, 241-69 (1973). 33. C. G. Pantano, Jr., D. B. Dove and G. Y. Onoda, Jr., "AES Analysis of Sodium in a Corroded Bioglass Using a Low Temperature Technique," Appl. Phys. Lett., 2^6 ( 1 1 ) 601 -02 ( 1 975 ) .

PAGE 164

1 52 34. L. F. Vassamillet and V. F. Caldwell, " El ec tron -Probe Microanalysis of Alkali Metals in Glasses," J. Appl . Phys. , 40(4) 1637-43 ( 1 969) . 35. A. R. Cooper and A. K. Varshneya, "Diffusion in the System K20 — SrO — S i 0 2 : I. Effective Binary Diffusion Coefficients," J. Amer. Ceram. Soc., 51(2) 103-6 (1968). 36. P. J. Goodhew and J. E. C. Gulley, "The Determination of Alkali Metals in Glasses by Electron Probe Microanalysis," Glass Tech., ]_5( 5 ) 1 23-26 ( 1 974). 37. D. E. Clark, M. F. Dilmore, E. C. Ethridge and L. L. Hench, "Aqueous Corrosion of Soda-Silica and Soda-LimeSilica Glass," J. Amer. Ceram. Soc., 59(1-2) 97-100 (1976). 38. G. L. McVay and E. H. Farnum, "Atmospheric Effects on Na Diffusion in Glass," J. Amer. Ceram. Soc., 55 (5) 275 (1972). 39. G. L. McVay and E. H. Farnum, "Anomalous Effects of H 2 O on Na Diffusion in Glass," J. Amer. Ceram. Soc., 57 ( 1 ) 43-4 ( 1 974 ) . 40. D. L. Rothermel , "Durability of Annealed and Unannealed Soda-Lime Glass Powders," J. Amer. Ceram. Soc. , 5j4 ( 4 ) 218 (1971 ) . 41. M. F. Dilmore, D. E. Clark and L. L. Hench., "Correlation of Grit Size with Surface Roughness," in preparation. 42. D. M. Sanders, W. B. Person and L. L. Hench, "New Methods for Studying Glass Corrosion Kinetics," Appl. Spectrosc., 26(6) 530-36 (1972). 43. S. Anderson, "Investigation of Structure of Glasses by Their Infrared Reflection Spectra," J. Amer. Ceram. Soc. , 33 (2 ) 45-51 ( 1 950) . 44. P. E. Jellyman and J. P. Procter, "Infrared Reflection Spectra of Glasses," J. Soc. Glass Tech., 39, 173-192 (1955) . J. R. Sweet and W. B. White, "Study of Sodium Silicate Glasses and Liquids by Infrared Reflectance Spectroscopy," Phys. Chem. Glasses, 1_0(6) 246-51 ( 1 969). 45 .

PAGE 165

153 46. J. R. Ferraro and M. H. Manghnam, "Infrared Absorption Spectra of Sodium Silicate Glasses at High Pressures," J. Appl . Phys., 43(11) 4595-99 (1972). 47. I. Simon and H. 0. McMahon, "Study of Some Binary Silicate Glasses by Means of Reflection in the Infrared," J. Amer. Ceram. Soc., 36^(5 ) 1 60-64 ( 1 953 ). 48. D. Crozier and R. W. Douglas, "Study of Sodium Silicate Glasses in the Infrared by Means of Thin Films," Phys. and Chem. Glasses, 6(6) 740-45 (1965). 49. M. H. Manghani , J. R. Ferraro and L. J. Basile, "A Study of Na20 — HO 2 — S i 0 2 Glasses by Infrared Spectroscopy," Appl. Spectrosc., 28 [ 3 ) 256-59 (1974). 50. D. A. Duke and J. D. Stephens, "Infrared Investigation of the Olivine Group Minerals," Am. Miner., 49^ 1388-1406 (1964). 51. D. M. Sanders and L. L. Hench, "Surface Roughness and Glass Corrosion," Amer. Ceram. Soc. Bull., 52(9) 666-69 (1973). 52. R. Hanna, "The Structure of Sodium Silicate Glasses and Their Far-Infrared Absorption Spectra," J. Phys. Chem. , 69(1 1 ) 3846-49 ( 1 965). 53. R. Hanna and G. J. Su, "Infrared Absorption Spectra of Sodium Silicate Glasses from 4 to 30," J. Amer. Ceram. Soc., 47(12) 597-601 (1964). 54. B. Goldstein and D. E. Carlson, "Determination of the Composition of Glass Surfaces by Auger Spectroscopy," J. Amer. Ceram. Soc., 5J5 ( 1 ) 51 -52 ( 1 972 ). 55. S. M. Budd and J. Frackiewicz, "The Mechamisms of Chemical Reaction Between Silicate Glass and Attacking Agents. Part 2. Chemical Equilibria at GlassSolution Interfaces," Phys. Chem. Glasses, 2^(4) 11518 (1961 ). 56. S. M. Budd and J. Frackiewicz, "The Mechanisms of Chemical Reaction Between Silicate Glass and Attacking Agents. Part 3. Equilibrium pH of Some Na20 — CaO — S i 0 2 Glasses and Its Relationship with Chemical Reactivity," Phys. Chem. Glasses, 3(4) 116-20 (1962).

PAGE 166

154 57. S. Sen and F. V. Tooley, "Determination of Calcium, Sodium, and Silicate Ions in Extracts from Chemical Durability Tests on Glass," J. Amer. Ceram. Soc., 33(5) 178-80 (1950). 58. G. W. Morey, The Properties of Glass, pp. 101-131, Reinheld, New York, (1954). 59. F. R. Bacon, "The Chemical Durability of Silicate Glass," Glass Ind., 49(10) 554-59 (1968). 60. F. F. Wang and F. V. Tooley, "Detection of Reaction Products Between Water and Soda-Lime-Silica Glass," J. Amer. Ceram. Soc., 41(H) 467-69 (1 958). 61. R. G. Newton, "Some Problems in the Dating of Ancient Glass by Counting the Layers in the Weathering Crust," Glass Tech. , 7 ( 1 ) 22-25 ( 1 966 ). 62. R. G. Newton, "Stereoscan Views of Weathering Layers on a Piece of Ancient Glass," Glass Tech., 13(2) 54-56 (1972).

PAGE 167

BIOGRAPHICAL SKETCH David Edward Clark was born July 3, 1946, in Marianna, Florida. He received his elementary education in Florida and was graduated from Gainesville High School. Gainesville, Florida, in 1964. He entered the University of Florida in September, 1964, and worked part time as a lab technician and researcher in the Department of Metallurgical and Materials Engineering while working towards his degree. He received his Bachelor of Science in June, 1969, and his Master of Science in August, 1970 Both degrees were from the Department of Metallurgical and Materials Engineering at the University of Florida. He was chief metallurgical research engineer for 3M Corporation in Chattanooga, Tennessee from June, 1970 through September, 1972, at which time he returned to the University of Florida to pursue a Ph.D. in Materials Science and Engineering. He has published papers in several research areas, including solid-state kinetics, electronic ceramics, ceramic-metal composites, electron microprobe analysis. Auger electron analysis and glass corrosion.

PAGE 168

1 56 He is a member of A I ME , ASE, Alpha Sigma Mu honorary, Sigma Xi honorary, American Ceramic Society and the National Institute of Ceramic Engineers.

PAGE 169

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Larr/ L . Hench , Chairman Professor and Head of Ceramics Di vision I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. <• George Y. Onoda , Jr. Associate Professor, Materials Science and Engineering I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. 3>.2>.3>*~c „ Derek B. Dove Professor, Materials Science and Engineering

PAGE 170

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. This dissertation was submitted to the Graduate Faculty of the College of Engineering and to the Graduate Council, and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. March, 1976 Dinesh 0. Shah Professor, Chemical Engineering D&a n .College of Engineering Dean, Graduate School

PAGE 171

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. This dissertation was submitted to the Graduate Faculty of the College of Engineering and to the Graduate Council, and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. March, 1976 Dinesh 0. Shah Professor, Chemical Engineering D&a n .College of Engineering Dean, Graduate School