KINETIC AND MICROSTRUCTURAL ASPECTS OF
THE AQUEOUS ALTERATION OF GLASS
BRUCE K. ZOITOS
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
UNIVERSITY OF FLORIDA LIRAIES
Bruce K. Zoitos
For Mom and Dad, of course.
But mostly, for me.
I wish to thank my advisor, Professor David Clark, for introducing me to the field
of glass corrosion and for the many growth opportunities he made available to me during
my graduate school career. I also thank Professors Stan Bates, Emmett Bolch, Robert
DeHoff, Alex Lodding and E. Dow Whitney for their guidance and participation as
I am very grateful to those individuals whose time and expertise contributed
materially to this work. Guy Latorre provided numerous hours of assistance with both
ICP and FTIRRS studies. I thank Professor Lodding and his staff at Chalmers University
of Technology for countless SIMS profiles and for helping to meet those ever-present
publication deadlines. Professor J.F. Quinson of Universite Claude Bernard was very
generous to provide porosimetric analysis using his "one-of-a-kind" calorimeter. Dr.
David Hensley of the Ceramic Engineering Department at Rutgers University performed
the XPS measurements discussed in this work. A very special thanks goes to Dr. Steve
Spooner of the National Center for Small Angle Scattering Research at Oak Ridge
National Laboratory for giving up several evenings to insure that my samples were
completed within the allotted 48 hours beam time. Thanks go also to Richard Crockett,
Augusto Morrone and Andy Duncan for expert TEM and SEM work.
None of this would have been possible without the financial support of
Westinghouse Savannah River Company and its predecessor, E.I. du Pont de Nemours
and Company, nor would it have been possible without the enthusiasm and personal
support of Dr. George Wicks of WSRC. I am endebted to George for his friendship and
encouragement. It is with much pleasure that I thank Drs. Phil Bennett and Vince Puglisi
of Gates Energy Products for tolerating the periodic absence of a certain GEP scientist
during the busier stages of this work.
Finally, I wish to thank all my friends and fellow students for providing the love
and moral support that helped me to stay afloat during some very difficult and painful
times. You are too numerous to mention, and I trust you will not be offended if your
names are omitted. You know who you are. I would like to extend very special thanks
to Beree Darby, who was a sympathetic listener at a time when I needed someone to
listen and Anne Ricker, whose friendship has been a reassuring constant for the last two
years. Lastly, I wish to thank my former supervisor, Bob Stosky, for a particular
conversation in December 1990, in which I was reminded of my true reason for coming
to graduate school. That conversation resulted in the appearance of this dissertation
approximately one year later.
TABLE OF CONTENTS
A B STRA CT....................................................................................................... viii
1 INTRODUCTION........................................................................................... 1
2 OVERVIEW OF AQUEOUS CORROSION IN SILICATE GLASSES......... 4
Reactions of Amorphous Silica with Water............................... ....... 5
Solubility of Amorphous Silica at 250C......................................... 5
Effect of Temperature on the Solubility of Amorphous Silica...... 5
Effect of pH on the Solubility of Amorphous Silica................... 7
Effect of Particle Size on the Solubility of Amorphous Silica...... 7
Kinetics of Silica Dissolution and Deposition................................ 9
Structure and Corrosion of Alkali-Silicate Glasses.................................. 10
The Structure of Alkali-Silicate Glasses.................................. 12
Corrosion of Alkali-Silicate Glasses........................... ........... 14
Corrosion of Commercially Important Ternary Glasses.......................... 16
Corrosion of Soda-Lime-Silica Glass.............................................. 17
Corrosion of Sodium Borosilicate Glass.................................... 18
Corrosion of Nuclear Waste Glasses..................................................... 23
Characteristics of Nuclear Waste Glass Alteration Layers............ 24
The Savannah River Leachability Model................................... 26
The Thermodynamic Model....................................................... 29
Geochemical Modelling of Glass Corrosion................................. 30
3 OBJECTIVES AND ACCOMPLISHMENTS........................................................ 34
4 STRUCTURAL, MICROSTRUCTURAL AND KINETIC ASPECTS
OF CORROSION IN BINARY ALKALI-SILICATE GLASSES........... 36
Experimental.................................................... ..................................... 36
R esults.......... .............. ............................................................................ 42
Visual Inspection.............................................................. ............. 42
Scanning Electron Microscopy.................................... ........... 42
Transmission Electron Microscopy................................................. 61
Small Angle Neutron Scattering................................. .......... 66
BET Gas Adsorption Analysis................................... ........... 71
Fourier Transform Infrared Reflection Spectroscopy................ 74
X-ray Photoelectron Spectroscopy.............................. .......... 78
Leachate Analysis.................................................................... 80
Surface/Bulk Solution Studies..................................... .......... 87
Summary, Model and Discussion................... ................................ 90
Corrosion of Li20.2SiO2.......................................................... 90
Corrosion of Na20-2SiO,......................................................... 92
Corrosion of K20-2SiO2.......................................................... 95
A Model for Corrosion in Binary Alkali-Silicate Glasses......... 95
5 AQUEOUS ALTERATION OF
165/TDS SIMULATED NUCLEAR WASTE GLASS............................ 98
Introduction...................................................... .................................... 98
Experim ental........................................................................................... 99
Glass Processing and Composition............................. .......... 99
Leachants................................................... ............................... 99
Leach T ests ....................................................................................... 10 1
Field Leaching Studies..................................................................... 104
R esults.................................................................................................... 105
Scanning Electron Microscopy....................................................... 105
Transmission Electron Microscopy............................................ 113
Therm oporom etry............................................................................. 113
Fourier Transform Infrared Reflection Spectroscopy................... 118
Secondary Ion Mass Spectrometry................................................. 121
Leaching K inetics............................................................................. 134
Sum m ary and D iscussion............................................................................. 138
Corrosion of 165/TDS in Deionized Water................................... 143
Corrosion of 165/TDS in Granitic Groundwater............................ 144
Corrosion of 165/TDS in WIPP-A Brine........................................ 144
Correlation of Laboratory and Field Studies................................. 145
6 CONCLUSIONS AND SUGGESTIONS FOR FUTURE WORK.................... 147
C onclusions.................................................................................................. 147
Future W ork........................................................................................... 149
REFEREN CES................................................................................................... 151
BIOGRAPHICAL SKETCH.................................................................................... 162
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
KINETIC AND MICROSTRUCTURAL ASPECTS OF
THE AQUEOUS ALTERATION OF GLASS
Bruce K. Zoitos
Chairman: David E. Clark
Major Department: Materials Science and Engineering
The effect of aqueous alteration was examined in R20-2SiO2 glasses, where R
represents Li, Na or K. Glasses were leached at 700C in flowing deionized water for
periods ranging from five minutes to 40 days. Leached surfaces were examined visually
and by scanning electron microscopy (SEM), transmission electron microscopy (TEM),
Brunauer-Emmett-Teller nitrogen adsorption analysis (BET), small angle neutron
scattering (SANS), Fourier transform infrared reflection spectroscopy (FTIRRS) and x-ray
photoelectron spectroscopy (XPS). Leaching kinetics were evaluated by inductively
coupled plasma spectroscopy (ICP) analysis of leachate solutions.
Drastically different alteration surfaces and kinetics were observed for each glass.
Depending upon the glass, surface layers were composed of the original silica network,
a porous colloidal silica precipitate or were completely absent. Leaching kinetics ranged
from diffusion-limited to linear.
A model was developed to explain these differences, based on the solution
chemistry of the hydrodynamic boundary layer adjacent to the glass surface and its effect
on glass alteration. Model predictions were validated by direct measurement of solution
parameters within this boundary layer.
A similar methodology was used to evaluate alteration product formation in a
simulated nuclear waste glass leached at 900C in deionized water and in granitic
groundwater and synthetic brine, which are considered to be "repository-relevant"
leachants. Corrosion in deionized water and granitic groundwater leads to a porous,
precipitate-like alteration product. Kinetic measurements indicate that this product serves
to decrease the alteration rate of the underlying glass. In brine, a thin brine precipitate
layer forms on the glass surface which appears to drastically reduce the corrosion rate of
At the time of this writing, nearly 1.3 billion Curies (4.4 x 10'9 Becquerel) of
defense high level radioactive waste (HLW), mostly in liquid form, await processing at
four U.S. sites [Wic92]. The first U.S. HLW vitrification plant, the Defense Waste
Processing Facility (DWPF), is scheduled to begin operation this year at Savannah River
Site using an optimized glass composition designated 165/TDS. One of the primary
design targets for this glass was that it have a low aqueous corrosion rate.
Glass is the material of choice for most applications which require chemical
durability. It is inexpensive, raw materials are readily available in large quantities, it is
easily manufactured and its composition may be tailored to suit a given application. By
far, the most critical application of a glass product is that of HLW immobilization. Such
use requires a product capable of maintaining its integrity for several hundred thousand
years under "worst case scenario" conditions. The need to understand glass performance
in such an application has motivated most glass corrosion research over the last two
Prior to interest in nuclear waste glasses, most investigations of durability were
directed at simpler glasses, mostly the alkali-silicates and soda-lime-silica glasses. This
early work was somewhat limited, in that comparatively primitive analytical techniques
were available at that time. In addition, much of the work done on simple glass systems
during the 1970s was based on static test methods. In a static test, glass corrodes in a
solution in which dissolved glass components accumulate over time. As a result,
corrosion effects due to intrinsic glass behavior and effects due to the accumulation of
leached glass components in solution cannot be unambiguously determined. As analytical
methods improved during the 1980s and research emphasis shifted to more complex
nuclear waste glasses, few efforts were made to correlate the advancing waste glass
corrosion database with that of simpler, more easily understood glasses. These limitations
in understanding corrosion in simple glass systems have led to a general lack of
consensus on corrosion mechanisms in more complicated glasses.
The primary objective of this dissertation was to address deficiencies in
understanding of glass corrosion through improved studies on binary alkali-silicate
glasses. These studies introduce two innovations to this subject. First, corrosion tests
were designed to limit any effects due to accumulation of leach products in solution. This
was accomplished by leaching glasses in a flowing leachant. Additional test
modifications were made which allowed direct measurement of concentration gradients
between bulk solution and solution adjacent to the glass surface. The existence of such
gradients has long been speculated but has never been verified.
In addition, an effort was made to identify and utilize new analytical methods for
alteration product characterization. This led to a collaboration with workers at the
National Center for Small Angle Scattering Studies at Oak Ridge National Laboratory to
study leached glass microstructure using small-angle neutron scattering techniques, as well
as extensive utilization of in-house analytical expertise such as scanning and transmission
electron microscopy. Where possible, these techniques were applied to nuclear waste
glasses in addition to the binary glasses. In some cases, additional techniques were
utilized to circumvent analytical problems inherent to nuclear waste glasses. For example,
complex alteration product compositional profiles were measured quantitatively and to a
high degree of accuracy using secondary ion mass spectrometry techniques developed by
co-workers at Chalmers University of Technology. Thermoporometry, conducted in
collaboration with workers at Universite Claude Bernard in Villeurbanne, France, provided
pore size analysis on very limited amounts of leached nuclear waste glass and did not
require that the sample be dried. The success of these efforts was made possible not only
by utilization of advanced techniques and equipment but also by the expertise and
experience of the individuals who developed and apply them.
Analysis of the information developed over the course of this study demonstrates
that aqueous corrosion can lead to drastically different leaching kinetics and alteration
products depending upon glass composition. These diverse behaviors have been unified
by considering the effect of the inherent glass leach rate on solution chemistry.
Specifically, it suggested that interaction of the glass with the chemistry of the solution
adjacent to the glass surface is the factor which determines the mode in which the glass
OVERVIEW OF AQUEOUS CORROSION
IN SILICATE GLASSES
Silicate glasses encompass a family of materials with a wide range of chemical
durability. Within the silicate glasses, composition and leaching conditions exert a strong
influence on durability and corrosion. For instance, specimens of K20-2SiO2 placed in
deionized water at 700C dissolve completely within minutes, while nuclear waste glasses
corrode to only 10 15 .tm depth after two years under aggressive laboratory conditions.
While some aspects of glass corrosion are well understood, an all-encompassing
explanation of corrosion in these materials does not currently exist. Instead, several basic
mechanistic phenomena (ion exchange and network dissolution) have been described and
used to explain aspects of the corrosion process, and several models thermodynamicc,
geochemical, surface layer effects) have been proposed to explain and predict corrosion.
Thus, considerable information has been accumulated regarding atomic-level aspects of
alteration as well as global behavior of glass-water systems. Very little information exists
regarding the microstructure of glass ateration products.
The objective of this chapter is to examine previous work conducted on corrosion
in silicate glasses. For purposes of clarity, these glasses will be considered in order of
complexity: amorphous silica, alkali-silicate glasses, commercially important ternary
glasses and multicomponent nuclear waste glasses.
Reactions of Amorphous Silica with Water
The fundamental reaction between silica and water, as described by Eiler [Eil74],
(SiO2)x + 21H20 < (SiO2)x,- + Si(OH)4 (2-1)
The net reaction direction is dictated by conditions such as solubility (supersaturated or
undersaturated), temperature, pH, and silica particle size. The effect of each of these
parameters will be discussed.
Solubility of Amorphous Silica at 250C
In assessing the solubility of silica at 250C, Morey et al. [Mor64] noted that
solubility was a strong function of particle size, state of hydration and adsorbed impurities
and that several months were typically required for the silica-water system to reach
equilibrium. After allowing initial concentration transients to subside, a saturation value
of 115 ppm SiO2 was noted for dissolved silica in contact with silica gel. A similar value
was obtained by the same authors by extrapolating to 250C solubility data taken at higher
temperatures by other authors.
Effect of Temperature on the Solubility of Amorphous Silica
The work of Okamoto et al. [Oko57], Lehner and Merril [Lenl7], Elmer and
Nordberg [Elm58], Alexander et al. [Ale54], Krauskopf [Kra56], Kitahara [Kit60], and
White [Whi56] on the solubility of amorphous silica at elevated temperatures was
compiled by Morey et al. [Mor64]. The results are shown in Figure 2-1. While these
The solubility of amorphous silica in water between 00 and 1000C.
Adapted from Morey [Mor64]. Reprinted by permission of Journal of
data show considerable spread, it is clear that silica solubility increases from between 75
and 150 ppm SiO2 at 250C to between 300 and 400 ppm at 900C.
Effect of pH on the Solubility of Amorphous Silica
Alexander et al. [Ale54] found that the solubility of silica was approximately
constant from pH 1 to 8; however, solubility increased rapidly above pH 8. Using
colorimetric methods, they were able to determine that the increase in dissolved silica
above pH 8 was due to the presence in solution of the H3SiO4 ion in addition to Si(OH)4.
They concluded that the reaction
Si(OH)4 + OH- (HO)3SiO- +H20 (2-2)
dictated silica solubility at high pH, with the stipulation of constant Si(OH)4
concentration. Silica solubility as a function of pH is shown in Figure 2-2.
Effect of Particle Size on the Solubility of Amorphous Silica
Since silica occurs in a number of forms, from monolithic pieces to nanometer-
sized colloids, it is important to understand the effect of particle size on solubility. In
general, convex surfaces have a higher solubility, while concave surfaces have a lower
solubility. This effect, known as the Gibbs-Thompson effect, is described by the
S(r)=S,e, ( )
where S(r) is the solubility of a particle of radius r, S, is the solubility of a flat surface,
2 4 6 8 10 12
Solubility of silica in water at 250C as a function of pH. Adapted from
Alexander et al. [Ale541. Reprinted by permission of the American
E is the interfacial energy, V is the molar volume, R is the universal gas constant and T
is the absolute temperature.
For collections of small particles, points of particle-particle contact represent
regions of high negative curvature and will be preferred sites for deposition. Similarly,
small particles will dissolve due to their high radius of curvature and silica will deposit
at regions of negative curvature or on particles with larger radii, a phenomenon referred
to as Ostwald ripening.
Kinetics of Silica Dissolution and Deposition
Rimstidt and Barnes [Rim80] derived a general rate equation for the hydration and
dehydration of silica described in equation (2-1). Their work was based on the
assumption that reactants must form an activated complex of arbitrary composition as an
intermediate step during their conversion to products. Starting with the reactions
Si02 + 2H20 <- (SiO2-2H20)* (2-4)
(SiO2-2H20)* <-> Si(OH)4 (2-5)
and assuming that reaction rate is proportional to the concentration of the activated
complex (denoted with an asterisk) they derived the following rate equation
(as(o), A / 2 (2-6)
I ----- I [1 = -- towi)n kt4casmoran o- asion)
where a, is activity of i, P is the pressure, T is the absolute temperature, A is the
interfacial area, M is the mass of water in system, yi is the activity coefficient of i, k, is
the dissolution rate constant and k is the precipitation rate constant. The parameters k+
and k. were determined by measuring silica reaction rates at various temperatures and are
logk_ =-0.707- 2598 (2-7)
logk = -0.369-(7.890 x10-4)T-343 (2-8)
This set of equations expresses the observation that the rate of dissolution decreases as
the activity of dissolved silica silicicc acid) increases in solution. Dissolution rate also
is proportional to the ratio of silica surface area to solution mass. Thus, a high surface
area silica submerged in a limited amount of water will reach equilibrium more quickly
than a low surface area silica in a large solution volume. The area-to-mass ratio,
commonly expressed in glass corrosion literature as the surface area-to-solution volume
ratio (SA/V), is a critical parameter which dictates kinetics and sometimes corrosion
mechanisms. Important factors and their effect on silica-water reaction rates are presented
in Table 2-1.
Structure and Corrosion of Alkali-Silicate Glasses
Next we consider the slightly more complex case of silica doped with a modifying
group I alkali oxide, R20, where R = Li, Na, K. Introduction of alkali atoms into the
glass results in a decrease in the glass transition temperature and the aqueous durability
of the glass. These effects are related to the alteration of the atomic-level structure of
silica by the conversion of bridging oxygen, or siloxane (Si-O-Si) bonds, to non-bridging
Table 2-1. Factors that Influence the Rate
of Silica-Water Reactions
Variable Effect on Reaction Rate Valid Extrapolation Range
Very little effect
0 500 bars
Extent of System
Activity of H2SiO4
Silica phase present
Rate proportional to A,
inversely proportional to
Rate proportional to (1-
Q/K). (Q = activity
product, K = equilibrium
constant for formation of
Rate controlled by
breaking Si-O bonds
Determines K and
Reduce aH20 and thus
Small particles have
(A/M) < 104
At large Q/K high free
energy phases may
At low (A/M) nucleation
controls precipitation rate;
at T > 3000C diffusion
may control rates
Valid for quartz,
Slightly acid to neutral
(?) No evidence of effect
For particles < 0.1 lim, K
should be corrected for
surface free energy
Adapted from Rimstidt and Barnes [Rim80].
oxygen bonds (NBO), which are more susceptible to aqueous attack. In addressing the
issue of corrosion, the situation is now complicated by the effect of leached alkali ions
on solution composition and pH, as well as possible barrier effects of any silicious
material remaining behind on the glass surface. The problem now becomes that of
considering a non-equilibrium solid having a constantly changing surface composition
reacting with a solution of constantly changing pH and composition. The following two
sections will discuss the structure of binary alkali glasses as a function of composition
and examine leaching data and models which have been used to explain corrosion
The Structure of Alkali Silicate Glasses
The basic structural unit of silica is the SiO2 tetrahedron. Tetrahedra are joined
at their covers through bridging oxygen atoms (siloxane or Si-O-Si bonds). In order to
accommodate the introduction of R20, siloxane bonds are broken, the oxygen from R20
bonds to Si, and R' atoms take up positions to charge compensate Si-O groups. Such
bonds are referred to as non-bridging oxygen bonds. In a study using magic angle
spinning nuclear magnetic resonance, Maekawa et al. [Mae91] measured the occurrence
of bridging and non-bridging oxygen in R20-SiO2 glasses for compositions between 20
and 56 mol% R20. Their results, summarized in Figure 2-3, express glass structure in
terms of Qn groups, where n designates the number of bridging oxygens per Si04
tetrahedron. The important point of this work is that binary glasses should
S 40 \ /
20 30 40 50
Figure 2-3. Experimentally determined Qn distributions in lithium (triangles), sodium
(squares) and potassium (circles) silicate glasses as a function of
composition. Adapted from Maekawa et al. [Mae91]. Reprinted by
permission of Elsevier Science Publishers.
be considered a mixture of Q, groups, rather than being comprised of only one or two
types of Q groups.
Corrosion of Alkali-Silicate Glasses
Many of the studies performed on the leaching of alkali-silicate glasses were
conducted in the early 1960s, before much of the current arsenal of analytical equipment
was available. As a result, most of the data developed on corrosion of these glasses is
based on the analysis of leachate solutions.
Charles [Cha58] and Budd [Bud61 ] described the corrosion of alkali-silicate glass
by water in terms of three reactions. The first reaction is essentially an ion exchange
process in which a proton, produced by dissociation of a water molecule, replaces an
alkali atom in the glass
=Si-O-'R + H20 =Si-OH + R+ + OH (2-9)
Hydronium or a hydrogen ion could serve as the reacting species rather than water.
Another possible reaction involves the attack by OH on bridging oxygen bonds
=Si-O-Si= + OH -- =Si-OH + -Si-O (2-10)
The terminal O group created here is very reactive and may further interact with water
=Si-O- + H20 -- =Si-OH + OH (2-11)
As this reaction continues, Si may be progressively hydrated and removed from the glass
network, forming silicic acid.
It is also possible for terminal silanol (=Si-OH) groups to undergo condensation
to form siloxane bonds [Eil74]
=Si-OH + HO-Si- =-Si-O-Si= + H20 (2-12)
These reactions form much of the current basis of understanding of glass
corrosion. With this understanding of the reactions involved, Rana and Douglas
[Ran61a,b] produced what is considered the definitive work on the kinetics of alkali-
Their study consisted of corroding crushed binary and ternary glass compositions
in a soxhlet extraction system in such a way that the glass was in contact with nearly pure
distilled water. Their analysis considered only the rate of appearance of glass constituents
in the leachate solution over the duration of the experiment.
In all cases, they observed two distinct leaching patterns. In the early stages of
leaching, the cumulative leached amounts of glass constituents increased with a t0-5
dependence. This was designated Stage I leaching. At later times, leaching showed a
linear time dependence, which was termed Stage II leaching.
Mathematically, Stage I was expressed as
and Stage II was written
where Q(t) is the total amount of a given element released up to time t, a and b are
adjustable parameters, and k and c are rate constants.
Based on these results, they proposed a model in which alkali diffused out of the
glass by ion exchange, creating a silica-rich surface layer. Since this was a diffusional
process, a t0'5 dependence would be expected. However, the silica surface layer also is
subject to dissolution, and as the alkali removal rate is reduced due to the growing silica-
rich surface layer, a steady state is ultimately reached where both silica dissolution and
alkali removal proceed at the same rate.
Corrosion of Commercially Important Ternary Glasses
The binary glasses discussed previously are of generally low durability and of little
commerical importance. However, addition of a third component to these compositions
may drastically improve durability. The addition of approximately 10% CaO results in
a composition known as soda-lime-silica glass, which is the basis for most window and
container glass. Addition of 10%-30% B203 results in alkali borosilicate glass. Known
as Pyrex, this extremely durable glass is used in laboratory and household applications
where resistance to aqueous corrosion is required. Alkali borosilicate glasses owe their
corrosion resistance in part to phase separation, in which an easily corroded sodium borate
phase separates from a more durable silica-rich phase. Leaching of these glasses is
entirely dependent upon the distribution of the soluble phase. If this phase occurs in a
non-interconnected fashion, the corrosion properties of the glass will be dictated by those
of the silica-rich phase, as is the case with Pyrex; however, if the leachable phase
develops as an interconnected network, then it is entirely available for corrosion. In this
case, leaching will result in a porous, silica rich skeleton. This is the basis for the Vycor
process which is used to manufacture nearly pure silica glasses at temperatures well
below the melting point of silica [Kin76].
Corrosion of Soda-Lime-Silica Glass
In early studies of soda-lime-silica glasses, Rana and Douglas [Ran61a,b] observed
that as SiO2 was replaced by CaO in sodium-silicate glass, the glass became more durable
and the proportion of alkali to silica in the leachate approached their ratio in the glass,
suggesting that the corrosion mechanism had changed from one of selective removal of
alkali to one of congruent dissolution. Leached surfaces of this glass were still shiny and
showed no visual indications of alteration.
Soda-lime-silica glass leaching as a function of solution pH was studied by El-
Shamy et al. [E1S72] and by Smets and Tholen [Sme85]. El-Shamy et al. found that
below pH 9 no silica was extracted and alkali extraction was independent of pH; however,
at pH values greater than 9, silica extraction increased while alkali extraction decreased.
In all cases extraction of calcium was negligible. At pH values less than 3, glasses with
greater than 10 mol% CaO demonstrated a much greater dissolution rate and calcium was
observed in the leachate. At these low pH values the ratio of extracted sodium to calcium
was the same as that in the glass.
Smets and Tholen [Sme85] examined the corrosion of 20Na20 10CaO-70SiO2 in
buffered solutions at pH values from 4.5-13. Secondary ion mass spectrometry (SIMS)
analysis of the leached glasses indicated that calcium depletion was limited to the first 10
nm of the glass surface. Sodium depletion was deepest in acidic solutions (approximately
700 nm after 16 hours) and only 100 and 10 nm in solutions of pH 9.1 and 13,
respectively. Their results are compiled in Table 2-2. These results suggest that between
pH 4.5 and 9, corrosion is a combination of congruent dissolution (linear rate) and
diffusion-controlled leaching of alkali ions. At pH 13, the glass is leached almost
congruently, but very slowly.
A SIMS profile of a leached soda-lime-silica glass surface is shown in Figure 2-4.
This result, obtained from a 10CaO-16Na0-74SiO2 glass [Ric84], shows a surface
depleted in Na, enriched in Si and a near-surface depletion of Ca with an enrichment of
Ca at deeper levels. This result is consistent with those obtained by other workers
[Cla76a, Sme85, Gos78].
Corrosion of Sodium Borosilicate Glass
Probably the most definitive work to date on the corrosion of sodium borosilicate
glass was performed by Bunker et al. [Bun88] who used Raman spectroscopy and nuclear
magnetic resonance spectroscopy (NMR) to characterize atomic-level structure in several
sodium borosilicate glasses before and after leaching at pH 1, 9 and 12.
Raman spectroscopy of Na2O-B203.SiO2 glasses with component ratios of
30-10-60, 20-20-60 and 10-30-60 revealed significantly fewer non-bridging oxygen bonds
than would be predicted if all sodium were associated with NBO sites, indicating that
some sodium is associated with anionic, tetrahedral boron sites. In the 30-10-60 and
20-20-60 glasses Raman spectra indicated that borate structures were interspersed
throughout a silicate matrix, as evidenced by the presence of Si-O-B bonds. However,
Table 2-2. Dissolution Rates and Sodium Depletion Depths
for Soda-Lime-Silica Glass
Buffer pH Dissolution Sodium Depletion
Rate (nm/hr) Depth (nm)
H20 5.8- 6.1 <10 720 (16hr)
KH2PO4 4.7 5.0 <10 670 (16hr)
K2HPO4 9.1 8.8 155 100 (16hr)
KOH 13.0 12.8 240 10 (Ihr)
Adapted from Smets and Tholen [Sme85].
I / sp --------B
0 100 200 x/nm
Figure 2-4. SIMS profile of 16Na20l10CaO.74SiO2 leached 30 minutes at 850C.
Adapted from Richter et al. [Ric84]. Reprinted by permission of
Revista Stazione Sperimentale della Vitro.
in the borate rich 10-30-60 glass no Si-O-B signal was detected. This suggests that the
glass is phase-separated and that borate groups tend to occur in borate rather than silicate
After the glasses were leached at pH 1 no non-bridging oxygen sites were found,
and the glasses contained substantially fewer silanol groups than would be predicted by
the conversion of NBO sites to silanols. This was taken to indicate that silanol
condensation had taken place. The leached material was nearly identical in composition
to silica, but differed from vitreous silica in that it was comprised of four-fold rings in
contrast to vitreous silica where higher order rings predominate.
For 10-30-60 glass leached at pH 9 and 12, Raman spectra were identical to that
of vitreous silica. A small proportion of Si-OH was found, and no 3- or 4-fold rings were
noted. The structure was composed of higher-order silica rings. Raman spectra of
30-10-60 and 20-20-60 glasses were unaffected by leaching at pH 9 and 12.
Nuclear magnetic resonance spectroscopy of 29Si is capable of distinguishing the
local environment of silicon in terms of the number of bridging oxygen bonds (i.e. Q4,
Q3, Q2 and Q' groups). In 30-10-60 glass prior to leaching, 75% of Si was present in Q3
groups and 25% in Q4 groups. After leaching at pH 1, 75% of Si existed in Q4 groups,
20% in Q3 groups and 5% in Q2 groups. The Q3 and Q2 groups terminated in silanol
groups rather than sodium, corroborating the results of Raman spectroscopy regarding
silanol condensation. No changes were noted in the glasses leached at pH 9 and 12 based
Acid leaching of 20-20-60 and 10-30-60 doubled the number of Q3 sites from 15%
NBO present initially to 30% silanols following leaching. In 10-30-60 glass leached at
pH 9 and 12, only Q4 sites were present after leaching.
Bunker's group also examined the amount of 170 incorporated into silanol and
siloxane bonds from leaching in H2170 labelled water. They found that '70 was
incorporated into Si-O-Si sites for all acid-leached glasses except 10-30-60, indicating
silanol condensation. The 30-10-60 glass leached at pH 12 was the only sample in which
NBO sites contained labelled oxygen and also had the lowest amount of 170 in Si-O-Si
sites. Analysis by 23Na NMR indicated that for all compositions studied, leaching at pH
1 resulted in the removal of all Na from the glass, while leaching at pH 9 and 12 did not
affect the sodium content.
Transmission electron microscopy of leached material showed that the leached
glass was comprised of a network of particles ranging in size from 10 to 160 A in
diameter. Leaching under acidic conditions favored formation of small particles, while
basic conditions resulted in larger particles. BET surface areas ranged from 350 400
m2/g with pore diameters of <30 A.
Bunker et al. concluded that the formation of silanol groups promoted the
hydrolysis of adjacent siloxane bonds, which resulted in the depolymerization and
dissolution of the silicate network. Simultaneously, silanol groups condensed, reforming
Si-O-Si bonds and repolymerizing the silica network, resulting in a leached layer similar
to colloidal silica. This behavior was used to rationalize the transition from t.05 to t'
kinetics described by Rana and Douglas [Ran61a,b] as the result of a leached layer which
evolved with time from being protective into being nonprotective.
Corrosion of Nuclear Waste Glasses
Assessing the mechanisms of corrosion in nuclear waste glasses presents a
substantial technical challenge. These glasses were designed to optimize multiple
properties simultaneously [Wic85a, Sop83]. They must be sufficiently durable to
immobilize radioactive waste for several hundred thousand years and be remotely
processed at temperatures sufficiently low to prevent evaporation of easily volatilized
components such as cesium. They must be insensitive to variations in waste stream
composition, be thermally stable against devitrification and have good mechanical
integrity. Glasses suitable for such applications contain numerous elements. The
composition of choice of U.S. defense waste applications, 165/TDS, contains 15 different
oxide components. The complexity of this material precludes most types of structural
analysis such as those described previously. However, an extensive empirical database
has been established for the corrosion performance of nuclear waste glass in both
laboratory and field environments [Tac90, Bra90, Sas90, Lod90, Ram90]. Significant
progress was made in the early 1980s by the Department of Energy-sponsored Leaching
Mechanisms Program [Men84] in elucidating general principles of corrosion in these
complex materials and delineating major characteristics of glass alteration layers. During
the last decade, three models have emerged which are capable of explaining and
predicting aspects of corrosion in nuclear waste glasses. These are the Savannah River
leachability model and the thermodynamic model, both developed at Savannah River
Laboratory, and the geochemical model, developed at the Hahn-Meitner Institut in Berlin.
Characteristics of Nuclear Waste Glass Alteration Layers
The Leaching Mechanisms Program, an intensive, multi-investigator research effort
[Men84], examined layer formation under a variety of leaching conditions. They
determined that altered glass tended to form highly structured, multi-layer structures
which had a sharp interface with the underlying glass. The alteration product retained the
dimensions of the parent glass but was of lower density. The degree of porosity varied
as a function of glass and leaching conditions, but all layers tended to have a honeycomb-
shaped surface structure. Frequently, both crystalline and amorphous structures were
found at the surface. Layer compositions were depleted in soluble elements such as Na,
B and Li but retained the less soluble elements Si, Fe, Al, Mn, Ni, Ca and Mg.
These observations were combined into the surface layer schematic shown in
Figure 2-5. This diagram shows three layers typically seen in altered nuclear waste glass.
The reaction zone at the boundary between the main layer and the unaltered glass is
typically less than a micron thick and represents the advance of aqueous attack into the
glass. The main alteration layer is found beyond the reaction zone and may be several
microns thick. The precipitate layer is at the outer surface and consists of both
amorphous and crystalline deposits. More recent work [Lod89, Abr90] supports and
Precipitates \Magnesite, CaC03)
Glass Reaction Zone
(B, Na, Li)
Figure 2-5. Schematic of alteration layer on a nuclear waste glass showing the reaction
zone, gel layer and precipitate layer. Adapted from Mendel [Men84].
extends the observation of layering. As many as six distinct layers have been observed
in some cases.
The Savannah River Leachability Model
The Savannah River model, also known as the Wallace-Wicks model [Wal83,
Wic92], describes the effect on glass leach rate of a growing, semipermiable precipitate
layer on the glass surface. It is assumed that the glass leaches initially by an
interdiffusional process, followed by matrix dissolution. In addition to these reactions,
the SR model considers a third process--precipitation. As glass components are released
to solution, they may saturate and precipitate onto the glass surface, forming a barrier
which may or may not have a protective effect.
The model assumes steady-state conditions and that the rate of silica dissolution
at the glass-precipitate interface, RdiS, is equivalent to the rate of diffusion of silica
through the precipitated layer, Rdif. Mathematically,
R= K(C, -C) (2-15)
Rd (CoC) (2-16)
where K is a rate constant, D is the diffusion coefficient of silica in the precipitated layer,
1 is the thickness of the precipitate layer and C, C0, and C, are the silica concentrations
in the bulk leachate, the leachate within the layer immediately adjacent to the gel layer,
and a saturated solution, respectively.
If L is the amount of silica leached up to time t and q is an appropriate constant,
then 1 = qL and, since R = dL/dt, the above equations may be rearranged to yield
dL_ KC C (2-17)
where p = Kq/D represents the silica dissolution from the glass and C/C, represents the
degree of saturation of silica in the system.
If the system is closed then leaching products accumulate in solution and C/C, =
aL, where a = SA/V(1/C,), SA is the glass surface area and V is solution volume.
Integrating 2-17 using the conditions that at t = 0, L = 0 the result is
(1+ +)O(n(l-aL))+1L=-aKCt (2-18)
In the case that the system is open and leach products do not accumulate then as V --> ,
a -> 0 and
L+ 2=KCSf (2-19)
These results are summarized in Figure 2-6 which plots the slope of a log-log plot
of silica release vs. time against the log of the reaction extent. Curve A shows that in the
early stages of leaching the slope is unity (congruent dissolution) and as the precipitate
layer forms the dissolution reaction becomes diffusion limited and the slope approaches
1/2. As saturation is approached (i.e. C/C, -- 1), the slope approaches 0. Curve B
I STAGE 3 I
Layer Diffusion Controlled
Slope of logL vs. logt as a function of reaction extent based on Savannah
River leachability model. Curve A shows behavior in a system where
leaching is diffusion-controlled; Curve B shows saturation-controlled
leaching. Adapted from Wicks [Wic92]. Reprinted by permission of
Noyes Publishing Co.
demonstrates the effect of leaching at high SA/ where solution saturation effects stop
the corrosion process before a precipitate layer may form.
The Thermodynamic Model
The thermodynamic model of glass corrosion was developed by Plodinec and
Jantzen of Savannah River Laboratory [Plo84, Jan92] and is based upon the earlier work
of Newton and Paul [Pau77, Pau82]. This model treats the glass as a mechanical mixture
of silicate and oxide components and assumes the overall free energy of hydration is
simply the sum of the free energy of hydration of the individual components. Calculation
of a composite Gibbs free energy of hydration is made by distributing modifying elements
into available silicates and treating glass-formers and leftover SiO2 individually.
A Gh= x(A G hd) (2-20)
where AGhyd is the Gibbs free energy of hydration of the glass, xi is the mole fraction of
component i, (AGhyd)i is the Gibbs free energy of hydration of component i.
Since many glass corrosion reactions involve pH excursions, allowance has been
made for the dissociation of silicic and boric acid at high pH [Jan92]. For H2SiO'
A(AG,=1.3f 1 ++ l 10+ _- (2-21)
10-p" 10-*pH 1)
and for H2BO3
10( -9.18 10-21.89 10-35.69 \\
A(AG,)=1.364 -log 1+ + +-. 10- 10 (2-22)
10-pH 10-2p 10-3pH J
Based on studies of over 300 glass compositions, Jantzen determined that a linear
relationship existed between the pH-corrected free energy of hydration of the glass and
the log of the normalized silicon mass loss from a 28 day MCC-1 [MCC83] leach test.
This is shown in Figure 2-7.
Jantzen's work is of particular importance in predicting the long-term performance
(-105 years) of nuclear waste glasses via thermodynamic correlation with naturally-
occurring glasses which are known to be stable in the environment on a geologic
timescale [Jan86, Ewi87].
Geochemical Modelling of Glass Corrosion
Computer modelling has been used to predict corrosion of mineral systems [Gra84,
Gra92] and was first applied to nuclear waste glasses by Bernd Grambow. This approach
treats congruent dissolution of the glass as a source term for entry of glass components
into solution. Once components enter solution, the model considers saturation and
precipitation and predicts the thermodynamically favored precipitation products and
corresponding solution concentrations. Because the model is thermodynamic in nature,
the occurrence of these processes is expressed as a function of reaction extent, rather than
10000 x BASALT
2 LogSi--0.1636G(hyd) + 0.3557 = 0.73 a CHINESE
2 I EGYPTIAN
ix B0 GLCERAM
0 0 GRANITE
S 100 2 4 HIGHSI
L/) I/ 1 ISLAMIC
o 10' a MISC
N a MIXALK
1 x ROMAN
+ /'+ w 0 SRL
0 a TEKTITE
Z f VENETIAN
C5 + WASTE
10 5 0 -5 -10 -15 -20
FREE ENERGY OF HYDRATION ( kcal / mole )
Figure 2-7. Linear regression plot of over 300 leach tests relating AGhyd calculated
from glass composition and pH to log silicon release for 28 day static
leach tests. Adapted from Jantzen [Jan92]. Reprinted by permission of
Noyes Publishing Co.
In Grambow's early work, he noted that for glasses leached in buffered solutions
at various pH values, the concentration of Ca" behaved as though it were in equilibrium
with solid CaCO3 [Gra82]. Grambow later adapted a geochemical computer code
(PHREEQE) to model the formation of glass alteration products. The output of this
model is shown in Figure 2-8 and shows a close correspondence to measured values.
PHREEQE is also capable of accounting for the presence of reaction products from
corrosion of other parts of a waste package, such as the stainless steel canister.
The geochemical model considers the rate-limiting step in corrosion to be the
irreversible decomposition of an activated surface complex. The overall reaction rate of
the glass is given by the product of the concentration of the activated complex and a
decomposition frequency factor. The affinity of the reaction is given by the ratio of the
ion activity product in solution to the equilibrium constant for the reaction. According
to Grambow, the cessation of leaching in high silica solutions is due to saturation of the
solution with respect to the activated surface complex, rather than with respect to the
REACTION PROGRESS. g/m3
Reaction of nuclear waste glass with 0.001 MgCl2 solution. Solid lines are
the result of modelling using PHREEQE. Data points are experimental
results. Adapted from Grambow and Strachan [Gra84]. Reprinted by
permission of the Materials Research Society and Bernd Grambow.
OBJECTIVES AND ACCOMPLISHMENTS
The principal aim of this study was to evaluate mechanisms of glass corrosion
using a variety of surface and solution analytical techniques. This strategy was applied
first to simple binary alkali-silicate glasses leached in deionized water and is discussed
in Chapter 4. Information gained from study of this system was then applied to simulated
nuclear waste glass alteration in deionized water and "repository-relevant" leachants. This
is discussed in Chapter 5.
Specifically, the objectives of this work were to
1) develop and apply an integrated characterization methodology to study
composition, structure and microstructure in leached glasses
2) examine in detail the microstructure of glass alteration products
3) develop a model relating alteration product microstructure to glass leaching
mechanisms and kinetics.
As a result of this effort, three characteristic leaching phenomena were identified
in alkali-silicate glasses, based on analysis of microstructure and leaching kinetics. These
1) alkali removal with silica network preservation
2) congruent dissolution with precipitation of colloidal silica
3) congruent dissolution.
Based on these results, a model was proposed which explains the evolution of
these structures and the associated leaching kinetics. The model is based on
considerations of solution chemistry within the hydrodynamic boundary layer (HDBL)
adjacent to the glass surface and its effect on alteration product microstructure. The
validity of the model was confirmed by direct measurement of HDBL chemistry using a
novel sampling approach in order to obtain solution samples adjacent to the glass surface.
The methodology developed in this effort was further utilized to study the
alteration of simulated nuclear waste glass in deionized water, granitic groundwater and
synthetic brine. Studies of alteration product microstructure identified dissolution and
precipitation as the most likely mode of alteration. Diffusion-limited leaching kinetics
were identified in some cases and were explained based on saturation effects within the
pores of the alteration layer.
STRUCTURAL, MICROSTRUCTURAL AND KINETIC ASPECTS
OF CORROSION IN BINARY ALKALI-SILICATE GLASSES
The objective of the work described in this chapter was to utilize surface analytical
tools to assess changes due to corrosion in R20-2SiO2 (R = Li, Na, K) glasses. In
particular, microstructural alteration and its effect on leaching kinetics were examined.
Glass specimens were leached in flowing deionized water and analyzed by visual
inspection, scanning electron microscopy (SEM), transmission electron microscopy
(TEM), BET gas adsorption analysis and small-angle neutron scattering (SANS) to assess
microstructural features over five orders of magnitude, from approximately 10 A to 100
pm. Leaching-induced changes in nearest-neighbor atomic structure were assessed using
Fourier transform infrared reflection spectroscopy (FTIRRS) and x-ray photoelectron
spectroscopy (XPS). Leaching kinetics were examined by measuring the rate of
appearance of glass constituent elements in solution using inductively coupled plasma
Alkali-silicate glasses were made by combining appropriate amounts of reagent-
grade alkali carbonates and silica and mixing them for a minimum of 24 hours in a jar
mill. Mixtures were melted in a platimum crucible in an electric furnace between 1200
and 15000C for a minimum of 24 hours and cast into 1 cm x 1 cm x 5 cm bars.
Solidified bars were annealed at 5000C for one hour and allowed to furnace cool to room
temperature. Bars were stored in a desiccator until needed.
Individual glass specimens were cut from the cast bars using a low-speed diamond
saw. Bars of Li2O-2SiO2 and Na20-2SiO2 glasses were cut into 1 cm x 1 cm x 0.2 cm
specimens; K20.2SiO2 specimens were 1 cm x 1 cm x 1 cm. Samples were polished on
all sides using 400 and 600 grit silicon carbide paper, followed by 1 pm diamond paste.
Polished samples were wiped with a soft cloth and washed ultrasonically for 15 minutes
in ethanol, as specified by MCC protocol [Mat83]. Alcohol used in cleaning was replaced
with fresh ethanol at five minute intervals. Prepared specimens were stored in a
desiccator until used in corrosion studies.
Glasses were leached in teflon containers in flowing deionized water at 700C,
similar to the MCC-4 test [Mat83]. Photographs of the corrosion system are shown in
Figure 4-1. Leachant flow rates and glass surface area-to-solution volume ratios (SA/V)
were controlled in order to maintain the pH as close to neutral as possible for each glass.
This is a particularly important aspect of the study, since accumulation of corrosion
products and pH excursions may impact both the mechanisms and kinetics of corrosion
[San73, Eth77, Wan58]. The duration of each study was adjusted to provide sufficient
leached layer for study and varied from six minutes in the case of the least durable glass
(KO2-2Si2O) to 35 days for the most durable glass (Li2O-2SiO2). Leachant flow rate,
residence time, SA/V, pH and test duration are shown in Table 4-1 for each glass studied.
Photograph of (a) corrosion system overview showing dionized water
reservoir, peristaltic pump, constant-temperature cabinet, leaching vessel
and collection bottle and (b) corrosion cell.
m. ~C *
Leachate samples were collected at appropriate intervals, acidified with nitric acid
[MCC83] and analyzed using ICP. At the conclusion of the leaching study, glass samples
were removed from the leaching vessel, rinsed in deionized water and transferred into
increasingly alcoholic mixtures of water and ethanol. Samples were hypercritically dried
under CO2 [Lau86, Fil86] using ethanol as a transfer liquid. Hypercritical drying is a
particularly important (and largely neglected) aspect of corrosion sample treatment.
Drying in this manner avoids the formation of a meniscus in the pores of the leached
layer. Drying stresses are thus avoided and fine-scale microstructural features are
preserved [Sch90]. It was also found that samples dried in this way had better structural
integrity than those dried in air. Leached specimens were stored in a desiccator pending
In addition to MCC-4-type testing, tests were conducted in a novel leaching cell
devised to permit separate, simultaneous solution specimens to be obtained from bulk
solution and the solution directly adjacent to the glass surface. This was accomplished
by adding a third port to the leaching flow cell described previously, as shown in Figure
4-2. Water was pumped into the cell through one port and allowed to flow out via the
second. Small-bore tubing was passed through the third port into the container and its
end firmly held against the glass surface. Water was very slowly pumped out of this port
in order to sample the solution at the plane where it contacted the glass.
Corrosion cell modified to allow sampling at the surface of glass: (a)
photograph; (b) schematic.
The macroscopic appearance of the glasses before and after leaching is shown in
Figure 4-3. Prior to leaching all glasses were transparent and clear. After leaching for
35 days, the Li2O-2SiO2 glass was clear and shiny, but contained numerous surface
cracks. These cracks began forming shortly after the glass had been removed from the
leaching solution and continued to form over a period of several days. The Na20O2SiO2
glass was opaque and extremely delicate after 10 hours leaching. Mechanical integrity
was so low that several samples were destroyed during handling. Finally, the overall
appearance of K20-2SiO2 glass was unchanged by leaching; however, the dimensions of
the sample were drastically reduced by leaching for approximately six minutes.
Scanning Electron Microscopy
Microstructure on the order of several millimeters down to approximately 0.05 p.m
was examined using SEM. Specimens were examined both in cross-section (obtained by
fracturing the sample) and face-on.
Figure 4-4 shows an SEM of a cross-section of Li20.2SiO2. The leached layer can
be easily discerned from the underlying glass. Large cracks and surface "channels" where
preferential dissolution has occurred are also visible. Higher magnification of this region
showed numerous small cracks throughout the layer. These cracks were not present in
the unleached glass.
Figure 4-3. Macroscopic appearance of binary glasses in leached and unleached states.
Figure 4-4. Scanning electron micrographs of fracture surface of leached
Li20.2SiO2: (a) 35x; (b) 1,000x; (c) 35,000x.
The surface of leached Li2O-2SiO2 glass is shown in Figure 4-5. Cracks are
visible on the surface along with small pits and channels where preferential corrosion
appears to have taken place. It is significant that polishing scratches are still visible on
the surface, since this indicates that appreciable dissolution of the silica network has not
Micrographs of a cross-sectional fracture surface of Na2O-2SiO2 are shown in
Figure 4-6. At low magnification it is clear that the leached layer has a very disrupted
microstructure, with numerous cracks and extensive pitting compared to the underlying
glass. High magnification in Figure 4-6c shows that, on a small scale, the altered glass
is not homogeneous but is composed of densely packed particles with diameters on the
order of 0.07 L.m.
The surface of leached Na2O-2SiO2 is shown in Figure 4-7. The appearance of
this glass is drastically different from the Li20-2SiO2 glass. Macroscopically, it is roughly
textured and opaque. Microscopically it has a structure which is dominated by large,
hemispherical pits. At high magnifications (Figure 4-7c) the pit edges show a terraced
structure. The mechanism of formation of these terraced pits is uncertain.
SEM examination of the leached K20-2SiO2 glass, shown in Figures 4-8 and 4-9,
revealed an extremely thin (<0.1 pm) reacted layer with no discerable microstructural
features. Some regions of layer formation on the order of 1 gtm were found in places, as
noted in Figure 4-8b. This layer, when present, was cracked and partially separated from
the underlying glass.
Figure 4-5. Scanning electron micrographs of leached Li20O2SiO2 surface:
(a) 35x; (b) 1,000x; (c) 35,000x.
Figure 4-6. Scanning electron micrographs of fracture surface of leached
Na2O.2SiO2: (a) 35x; (b) 200x; (c) 35,000x.
Figure 4-7. Scanning electron micrographs of leached Na2O-2SiO2 surface: (a)
35x; (b) 200x; (c) 600x.
Scanning electron micrographs of fracture surface of leached
K20.2SiO2: (a) 35x; (b) 30,000x.
, 9111 (b)
Figure 4-9. Scanning electron micrographs of leached K20-2SiO2 surface: (a)
200x; (b) 750x.
Transmission Electron Microscopy
Small-scale microstructure, on the order of 5-500 nm was examined using
transmission electron microscopy (TEM). Specimens were prepared using a technique
described by Bunker et al. [Bun84] and Randall [Ran87] in which leached layers were
scraped from the glass using a diamond-tipped stylus into a container of 1,1,1
trichoroethane. Suspended particles were dispersed by high intensity sonication for 15 -
20 minutes. The resulting slurry was placed dropwise onto a warm formvar-coated
copper TEM grid, and the trichloroethane was allowed to evaporate.
Examination of Li20-2SiO2 revealed numerous particles such as the one shown in
Figure 4-10. These particles had smooth surfaces and a reasonably homogeneous
microstructure on the scales studied here. Selected area electron diffraction (SAED),
shown in Figure 4-10b, showed diffraction rings, indicating some degree of crystallinity
in this material; however, x-ray diffraction analysis of the leached glass did not indicate
crystallinity, suggesting that the ordering found by SAED occurs on a small scale,
possibly as isolated crystallites.
In contrast, the leached Na2O-2SiO2 glass is roughly textured and appears to
consist of connected particles approximately 7-16 nm in diameter with void spaces in
between, as shown in Figure 4-11. SAED showed this material to be completely
amorphous. Leached K20-2SiO2 was not examined, since it lacked sufficient layer for
Transmission electron microscopic anllN 'is of fragment of leached
Li2O-2SiO2: (a) bright-field image; (b) selected area electron
100 nm I
Figure 4-11. Transmission electron microscopic analysis of leached Na2O-2SiO2
fragment: (a) bright-field image; (b) selected area electron
Small Angle Neutron Scattering
Small angle neutron scattering (SANS) measurements were used to determine
microstructure correlation lengths and to investigate any fractal properties in the leached
material. All measurements were made on the 30-meter neutron scattering camera at the
National Center for Small Angle Scattering Research at Oak Ridge National Laboratory.
SANS is sensitive to structural inhomogeneities and is useful in the study of small-
scale microstructure. Scattering occurs as a result of neutron interaction with spatial
variations in the distribution of matter. In the case of leached glass, this variation is due
to the presence of silica and pore space [Tom83, Sch87, Sch89].
The central problem of SANS work involves determination of the structure
responsible for a given scattering pattern. Several models are available to address this
issue. The Debye-Bueche model [Deb49, Deb57, Lon74, Wig90] was chosen for this
work because it is applicable to a random, interpenetrating two-phase system and is
capable of determining correlation lengths in porous material. According to the Debye-
Bueche model, scattering curves may be described as
I(Q)= 1 +A2e 4 (4-1)
where Q is the scattering vector, I(Q) is the scattering intensity at a given value of Q, A,
and A2 are fitting parameters and a, and a2 are correlation lengths.
SANS has also been used extensively to probe fractal aspects of materials [Bal84,
Cra86, Hen90, Sch86, Sch89a, Sch89c]. A fractal object is one which displays "dilation
symmetry" or "scale invariance." This means simply that the object appears similar at
varying length scales. Coastlines, trees and river deltas are familiar examples.
Mathematically, such structures are characterized by a fractional dimensionality. Surface
fractals obey the relation
where S is surface area, R is the object size and D, is the surface dimension (D, = 2 for
a smooth surface). Alternatively, an object is said to be mass fractal if it obeys the
where M is the mass of the object, R is the object size and D is the fractal dimension.
Measurement of the fractal aspects of an object by small angle scattering
techniques is made possible by a power law dependence of scattering intensity vs.
where all variables have the same meaning as previously discussed. At large values of
Q, scattering is due to interaction with primary particle surfaces. This region of the curve
is known as the Porod regime and displays a -4 power law dependence for a smooth
surface. A fractally rough surface will result in a deviation from this behavior and will
display a power law exponent between -3 and -4.
At smaller Q values, collections of primary particles give rise to scattering. This
region of the curve contains information from larger microstructures and is known as
the Guinier regime. On this scale of measurement, the surfaces of the primary particles
appear smooth and power law exponents between -1 and -3 would be expected.
The scattering curve of leached Li20-2SiO2 is shown in Figure 4-12. This sample
was a weak scatterer and required a 12-hour collection time to accumulate sufficient data.
The scattering curve is presented in two segments, since two camera lengths were used
to cover a larger range of Q. The region of the curve corresponding to the smallest scale
(Q values ranging from 1.25 to 0.25 nm ') shows no structure. At scattering vectors less
than 0.25 nm ', results demonstrate power law scattering with an exponent of -4.24.
There is sufficient spread in the results that an exponent of -4.0 cannot be ruled out
within a reasonable confidence interval. This appears to be Porod scattering behavior,
indicative of smooth surfaces; however, the lengths being probed in this region correspond
to length scales of several hundred nanometers, which is considerably larger than a
primary particle. This data may be the result of scattering from the fracture surfaces
observed by SEM.
Figure 4-13 presents the scattering curve of leached Na20.2SiO2. This curve
clearly demonstrates Porod scattering behavior from smooth particle surfaces at high
scattering vector values. At these values, length scales of a few nanometers are being
probed, corresponding to surfaces of the primary particles which were observed by TEM.
At intermediate Q values, the scattering exponent is approximately -1, and at the largest
Q values, a slope of -4.12 is observed. At large Q values, scattering is due to neutron
interactions with the surfaces of the 700 nm particles which were identified by SEM. No
clear surface fractal aspects were observed in any of the curves. However,
-1.6 -1.4 -1.2 -1.0 -0.8 -0.6
-0.4 -0.2 0
Log Q (nm"1)
Small angle neutron scattering curve of leached Li20-2SiO2. The two data
segments are from different camera lengths. Low-Q portions of the
curves have slopes of approximately -4.24.
CPO t %
I I I I I I I I I I I I
-1.6 -1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0
Log Q (nm1)
Small angle neutron scattering curve of leached Na20.2SiO2. Slope in the
high Q region is approximately -4.0 and is -4.12 in the low Q region.
a a a a I
because of statistical variations present in linear curve fitting, small deviations from
integer dimensionality cannot be ruled out.
Further analysis of the Na2O-2SiO2 curve by Debye-Bueche methods detected
correlation lengths of 29 A and 642 A, which correspond approximately to the structure
scales observed by TEM and SEM, respectively.
BET Gas Adsorption Analysis
Results from BET nitrogen adsorption analysis of leached Na2O-2SiO2 indicate that
the surface area of a single sample increased from 7.26 cm2 for an unleached specimen
to approximately 80,000 cm2 after leaching, an increase of over four orders of magnitude.
These results are consistent with those of Walker [Wal77] for Bioglass leached in 0.1M
TRIS buffer at 370C. Pore size analysis is shown in Figures 4-14 and 4-15. Results
presented here have been normalized to the original geometric surface area of the glass
instead of the more traditional sample mass. This was done because both leached
surfaces and unleached interiors were present in these samples. Thus, these measurements
indicate the amount of pore volume present per unit of original geometrical surface area.
These results show a narrow pore size distribution centered around a diameter of 35 A,
with a second, wider distribution ranging from 60 to 300 A. Bunker et al. [Bun88] found
a similar distribution of pores under 30 A diameter in studies of leached alkali
borosilicate glasses. Corrosion of Li2O-2SiO2 and K20-2SiO2 did not produce sufficient
surface area to allow study by gas adsorption methods.
0.00408 Na2 02S102
0.00374 Cumulative Desorption
aU 0.00238 -
10 100 1000
PORE DIAMETER, (A)
Figure 4-14. Cumulative desorption pore volume of leached Na20-2SiO2 measured by
PORE DIAMETER, (A)
Figure 4-15. Pore size distribution of leached Na20-2SiO2 as measured by BET.
Fourier Transform Infrared Reflection Spectroscopy
Glass surface structure and composition were probed using specular reflectance
Fourier transform infrared spectroscopy (FTIRRS) [And50, Swe69, Ger87]. Binary
glasses exhibit two broad reflectance peaks in the range of 800 to 1200 cm-'. The peak
centered at 1050 cm-1 is due to linear vibrations of siloxane bonds. Non-bridging oxygen
(NBO) bonds give rise to the other peak at 975 cm-' [San74, Cla77, Cla79, Che81,
Hus90]. Since modifier content correlates with the number of NBO sites in the glass, the
relative intensity of these two peaks is indicative of both structure and composition. In
addition, since a rough surface tends to scatter rather than reflect incident radiation,
decreases in the overall intensity of the spectrum indicate an increase in the degree of
Spectra of leached and unleached binary glasses are shown in Figures 4-16 through
4-18. The spectrum of vitreous silica is also shown in each figure. In the case of
Li20.2SiO2 glass, upon leaching, the NBO peak disappears, the bridging oxygen peak
intensifies, and the overall shape of the spectrum very closely approaches that of vitreous
Leached Na2O-2SiO2 shows very similar behavior; however the intensity of its
spectrum is much less than that of vitreous silica, indicating a rough surface. These
results demonstrate that the composition of the leached surfaces of LizO.2SiOz and
Na2O-2SiO2 are both similar in composition and structure to vitreous silica, due to the
formation of bridging oxygen bonds at the expense of NBO bonds during leaching.
Vitreous Silica S
O 20.1 I \, Unleached
ac 13.4 -
1400 1275 1150 1025 900 775 650 525 400
Figure 4-16. FTIRRS spectrum of Li2O-2SiO2 before and after leaching. The spectrum
of vitreous silica is shown for reference.
: 13.4 -
1275 1150 1025 900 775 650 525 400
FTIRRS spectrum of Na20-2SiO2 before and after leaching. The spectrum
of vitreous silica is shown for reference.
1400 1275 1150 1025 900 775 650 525 400
Figure 4-18. FTIRR spectrum of K20-2SiO2 before and after leaching. The spectrum
of vitreous silica is included for reference.
The two glass surfaces differ, however, in that NaO2-2Si02 is rougher and scatters more
strongly than Li20-2Si02.
FTIRRS measurements of K20-SiO2 show that leaching has little effect on the
shape of the spectrum, demonstrating that the leached surface has a composition and
structure nearly identical to the unleached surface. However, the leached glass spectrum
has a lower intensity than the unleached glass, due to surface roughening.
X-ray Photoelectron Spectroscopy
Near-surface atomic structure was studied using XPS. XPS is sensitive to the
bonding state of a given element. Thus, it was possible to measure the binding energy
of Ols electrons in order to determine the relative proportions of oxygen atoms involved
in bridging and non-bridging bonds [Vea80,Spr90]. Also, because XPS peak area is
proportional to concentration, it was possible to determine surface concentrations of each
glass element based on Ols, Si2p, Lils, Nakll and K2p peaks.
Since this technique measures only the outer few atomic layers, and because
simple exposure to atmosphere is sufficient to induce changes in this region of the glass,
unleached glasses were fractured in the vacuum chamber of the XPS system to provide
a fresh, unreacted surface for analysis. Leached samples were examined in "as received"
The binding energy distribution of the Ols electron is shown in Figure 4-19 for
unleached and leached Li20-2SiO2 glass. Peaks from bridging and non-bridging oxygen
bonds are not completely resolved in these spectra and appear as an asymetry in the
Figure 4-19. XPS of Ols line of Li2-02SiO2 glass: (a) unleached; (b) leached.
overall band. The individual contributions of each type of bond have been resolved by
fitting Gaussian distributions to the composite curve. In XPS, peak area scales with
concentration so the relative amounts of each type of bond may be determined. A
contribution from NBO bonds clearly can be seen in the spectrum of the unleached glass.
After leaching this contribution is lower, and the spectrum is dominated by the lower
binding energy peak. Virtually identical results were obtained for the Na20.2SiO2 glass,
as shown in Figure 4-20.
In the case of K20-2SiO2, shown in Figure 4-21, the Ols spectrum of the leached
sample is very similar to that of the unleached sample. The NBO peak has lost only a
small portion of its area as a result of leaching. This suggests that surface alteration
leaves a leached surface very nearly identical in structure and composition to that of the
unleached surface. Surface compositions of leached and unleached samples were also
determined [Esc75] and are summarized in Table 4-2.
The rate at which glass component elements appear in solution provides insight
into the mechanism and rate of reactions occurring at the glass surface. The leach rates
of silicon and lithium from Li2O-2SiO2 are shown in Figure 4-22. Small amounts of
silicon appeared in solution during the first few sampling intervals but rapidly fell below
the detection limits of ICP. The leach rate of lithium decayed rapidly over the first few
days of the study. Since these glasses contain equimolar amounts of modifier and silicon,
preferential or congruent dissolution may be assessed by comparing the leach rates of
540 538 536 534 532
540 538 536 534 532 53
Figure 4-20. XPS of Ols line of Na SiO glass: (a) leached; (b) unleashed.
538 536 534 532 530
Figure 4-20. XPS of Ols line of Na2O02SiO2 glass: (a) leached; (b) unleached.
Figure 4-21. XPS of Ols line of KO2-2SiO2 glass: (a) unleached; (b) leached.
Table 4-2. Surface Composition Determined by
X-Ray Photoelectron Spectroscopy
Sample O Si R
atomic% atomic% atomic%
Li20-2SiO2 57 23 20
Li20-2SiO2 69 28 3
Na20-2SiO2 53 26 21
Na20-2SiO2 65 27 8
K2O-2SiO2 56 20 24
K2O*2SiO2 64 17 19
SiO2 68 32 --
0.4 L20 2Si02
Cd Leach Rate
E 0.3 Predicted result
%w from Q = at.535
C 0.2 -
0 10000 20000 30000 40000 50000
CORROSION TIME (min)
Figure 4-22. Leach rates of Li and Si from Li20-2SiO2 during a 35-day corrosion test.
Data points are measured leach rates. Solid line is fit to Q = at0535
the two elements. If the glass were dissolving congruently, the leach rates of the two
elements should coincide. A difference in leach rates between the two components
indicates selective leaching. Based on this explanation, it is clear that lithium is
selectively removed from this glass.
The overall shape of the lithium leach rate curve suggests a diffusion-limited
process, like that proposed by Rana and Douglas [Ran61a,b]. To test this possibility, data
were fit to equation 2-13 by plotting logQ vs. logt and determining the slope, which is
predicted to be 0.5 for a diffusion-limited process. This plot is shown in Figure 4-23.
Data taken at later times are linear and have a slope of 0.535, an intercept value of
0.00525 mol/m2, and a correlation coefficient of 0.999, indicating a high quality fit. It
was noted that data obtained at short times are not linear and have a slope greater than
0.5. This deviation from t.05 behavior may be explained by mixing effects within the
leaching cell and the fact that the time required for sampling is long compared to the rate
of change of the glass leach rate at early times in the experiment. Within the first few
hours of the test, the glass leach rate is changing rapidly during the 30-minute sampling
interval. Thus, solution analysis is measuring a weighted average of glass leach rate over
the course of the sampling time, resulting in skewed data. To test this hypothesis, a
simple BASIC computer program was written to calculate the solution values which
would be measured throughout the experiment, based on mixing and sampling effects and
assuming a leach rate based on parameters measured at long times where logQ vs. logt
was linear. The predicted results are plotted in Figure 4-22 along with the data and show
a very close match. For this reason, it is concluded that the glass leached with a t0.535
time dependence throughout the study.
Li2 0 2SiO
Cumulative Leached Li j
'"Slope = 0.535
! I I I I I I I I
1.4 1.8 2.2 2.6 3 3.4 3.8 4.2 4.6
log t (min)
Logarithmic plot of cumulative amount of lithium leached from glass vs.
time. The slope of approximately 0.5 indicates a diffusion-limited process.
Leaching results for Na2O-2SiO2 are presented in Figure 4-24. This graph shows
that after an initial mixing transient, sodium is removed from the glass at a nearly
constant rate. Silicon, however, is very slow to appear in solution, but after
approximately 30 minutes, its leach rate rises rapidly to approximately 9.5 mol-m2-d-1 and
then gradually increases further until it matches the sodium leach rate. Thus, leaching is
initially by selective removal of sodium and ultimately evolves to congruent dissolution.
These results do not correlate with any established kinetic model.
Leaching of K20-2SiO2 is shown in Figure 4-25. In this glass, silicon and
potassium appear in solution at the same rate, indicating congruent dissolution. This
figure also shows that leach rate is nearly constant over the duration of the experiment.
Surface/Bulk Solution Studies
In a few instances has it been recognized that solution conditions immediately
adjacent to the surface of a corroding glass may differ from conditions in the bulk
solution [Cla83, Sa184, Gra86, Bun88, Wic92, Gra92]. This issue is particularly
important, since it is the solution within the hydrodynamic boundary layer (HDBL) which
reacts with the glass. To date, no direct measurements have been made on this
hypothetical entity. Obtaining such measurements was the objective of this particular
portion of the work.
Samples were obtained at early leaching times from both the bulk solution and
also from solution next to the glass surface as described previously. Solutions were
analyzed for pH and glass constituents. The results of these analyses are summarized
f14 O *
E 12 o
o o Na20 2S102
So o o Leach Rate
0 IdI I I I I I-
0 200 400 600
CORROSION TIME (min)
Figure 4-24. Time dependence of the leach rates of Na and Si from Na202SiO2 glass.
4000 0600o o
E K20 2S1O2
S Leach Rate
0 2000 oSI
0 -I I I I II
0 2 4 6
CORROSION TIME (min)
Time dependence of the leach rate of K and Si from K20O2SiO2.
in Table 4-3. In this table, surface values are expressed as the minimum value for that
parameter, since the measured values are a function of sampler flow rate. Ideally, such
a sample should be collected infinitely slowly to avoid dilution by the bulk solution.
Since this is not practical, it is likely that the true values were reduced somewhat by
dilution from non-ideal sampling; therefore, the figures presented here represent a lower
limit to the actual value.
As shown in Table 4-3, the difference between conditions in the bulk and surface
solutions is greatest in the K20-2SiO2 glass, which corrodes most rapidly, followed by
Na2O-2SiO2. Measurements of Li20-2SiO2 show the smallest difference between surface
and bulk values.
Summary, Model and Discussion
Corrosion of Li,02SiO2
This glass proved to be the most durable of the three compositions studied.
Polishing scratches, still visible on the glass surface after corrosion, demonstrate that the
silica network was not significantly disrupted during corrosion. XPS spectra show that
lithium is depleted at the glass surface, and infrared reflection spectra indicate that the
corroded surface is similar in structure to vitreous silica. The numerous cracks which
formed during drying suggest that silanol condensation occurred, since this is known to
result in tension and subsequent cracking in the leached surface [Bun83, Bun88].
Microscopic studies using SEM and TEM showed that the alteration layer had a
homogeneous structure but with numerous cracks. This sample was a weak neutron
So 9 o
o 0 0 0
d U- b^ (
oX< ^ ^
rt 6 0 ^ e
Ll^^o^ o (
Al A~L A