Journal of Non-Crystalhne Sohds 67 (1984) 193-205 North-Holland, Amsterdam

193

Sectton IV. Corroston and leachmg

GLASS CORROSION

A record of the past? A predictor of the future?

P.B. A D A M S
Cormng Glass Works, Cormng, New York, USA

The pnnclples of chenucal corroston of glasses are reviewed. The influence of temperature, glass composition, glass homogeneity and corrosion medta are related to observed corrosion rates. A comparison of the durabthty of man-made and natural glasses ~s offered and the corrosion rates of tekUtes and obstdians are d~scussed. Natural glasses are not hkely to advance our present understanding of corrosion theory

1. Introduction
Since glass was first encountered, m a n ' s use of glass has been influenced whether he realized it or not - b y its corrosion resistance. The bulk features of glass tell us m u c h about the process of its formation and hence its origin. The surface of a glass item also has a story to tell, sometimes about its origin but most often about its intervening environmental history, Rain water and dish water wash the surface of glass. Wetting and drying proceed. Temperature rises and falls. Some of these changes affect glass only slightly or not al all; others affect glass drastically. Some affect them uniformly; others affect them selectively. The glass technologist has watched, puzzled, and theorized about the changes that chemical reactions produce on glass. In the past he has relied heavily on empirical tests to help him define glass corrosion processes. H e knows from these studies that corrosion can generally be described either as an etching process, or as a leaching process, or as a combination of both. Recently, efforts to describe the glass corrosion process have been accelerated. Glass appears to be the best host for trapping nuclear waste and insuring its safe disposal. The urgency of this issue has focused the efforts of m a n y competent scientists on refining glass corrosion theory in order to predict how a nuclear waste glass will survive in a natural environment for thousands of years. This paper briefly reviews the principles of glass corrosion and c o m m e n t s on the role of glass composition. It deals primarily with obsidians and tektites since lunar glasses have not been exposed to a history of corrosion. M u c h of the discussion is taken from m o d e r n commercial glass technology experience, recognizing that m a n - m a d e glasses represent a mere few hundred years of corrosion history, c o m p a r e d to several million years for natural glasses. 0022-3093/84/$03.00 Elsevaer Science Pubhshers B.V (North-Holland Physics Pubhsbang Division)

194

P B. Adams / Glass corroston

2x NaOH + (SiOz)x----- xNa 2 SiO3 + xH20

0 I / S f .,.
0

0 I ~SJ
0 '~'S~ 0 I S! 0

/ ~

Na

"Sl
\

~0
Xsl ....0
"~OH

0 - - Si

/ 0
\

I o'Si \ 0 ""
0

0 ~ S =/
\

Na

O-.--Si .... O ' - s i / I

Fig. 1 SchemaUc of etcbang process.

2. Principles of corrosion

Corrosion mechanisms can be classified into two categories, etching or leaching [1-8]. Etching is usually characterized as alkaline attack (fig. 1). As the silica network is destroyed, other glass components are released. If there is no accumulation of reaction products on the remaining glass surface and no change in the activity of the contacting solution, the process proceeds at a constant rate. In the ideal case, temperature influences corrosion rate as described by the Arrhenius equation. Fig. 2 shows empirical curves relating temperature to the corrosion rate of a borosilicate glass. The slope of these curves indicates that corrosion rate increases by a factor of 2 to 2 per 10 C temperature rise.

ao

I/r
NaC %
J

'Na, CO~

z_ot

.,o/

.~ o ~ = ~' o m ~
o.ool 40

60

80 ioo 12o 14o 16o 18o 200 TEMPERATURE, = C

Fig. 2. Solubihty of a boroslhcate glass m vanous solutions versus temperature.

P B Adams / Glass corrosion

195

Accumulation of reaction products in solution may depress the reaction rate. In fact silica saturation can depress the rate to essentially zero. The presence of other elements may mitmte precipitation of salts which in turn may restore the activity of the contacting solvent or may adhere tenaciously to the glass surface, shielding it from attack. It has been shown that many elements such as beryllium (fig. 3), zinc, aluminum, lead, calcium, strontium and barium strongly inhibit alkaline attack [9]. Etching leaves a smooth surface if dissolution is total and no reaction products remain. However, the surface coated with reaction products, or grossly roughened, is probably more frequently encountered in real environments. Leaching ~s usually characterized by acid attack as schematically shown in fig. 4. Hydrogen ions, or more probably hydronium ions, exchange for alkali and other positively charged mobile ions in the bulk glass. The remaining glass network, principally silica, usually retains its integrity. It may become hydrated if the network is relatively unstable; or tt may become even more dense and stable than the parent glass. If the leached layer is not substantially altered, the reaction rate decreases with time and is proportional to the square root of time. In this simple case, the layer becomes, in essence, a protection against continuing attack (fig. 5). Often this layer consolidates, shrinking the reaction rate to zero (fig. 6). If the layer is loosely constituted, it may detach itself so that its protective effect is lost. Acids react much more slowly with glass than strong alkali and the p H effect is much less pronounced. The rate at low Ph may not be noticeably different from p H 5 or 6.

3.0 q

-tO z >-20 ,rt

Ls

3
Na*HCI ~
o

I-I++NaCl
0 Na CI

Sl
u. I0! 0 UJ In.-

$i
O,,si ~ o _

--~
"". 0

.~0

o. I

--o..__~__o

o-

( s

Na

2"o,

H*

o s=
0

OO10 .0020 CONCENTRATION OF BeO= (moles /liter)

0 "-. S i " O - - S i " I ~ 0 0

Si-- 0" / 0

Fig 3 Effect of berylhum on corrosion rate by05NNaOHat100C.

F~g. 4. Schematic of leachmg process

196
LEACHED LAYER

P B Adams / Glass corroswn

ht-~
;-'No + f - - , - BULK GLASS

:.,.:?..
,C:.
j . . .

I I

-;

I I

'.2-'.:.."-: ?j:-'
GE FRONT SiO z DISSOLUTION FRONT

f
J
TIME

Fig. 5. Acid dlssolut]on process.

Fig. 6 Dlssolutmn m acid with Ume

Attack by water and other near neutral solutions is less than for alkali and often is similar to acid attack. Usually it involves both ion exchange and sdica network destruction. In closed systems, or in systems where there is no

i
v

n/IKCI

Hs0
m

o2

H20 ADSORPTION

L
T

b w

i1_ o tLI
i.-

n.-

j
/
I 4 I0 TIME

ION EXCHANGE aZ;

J nllO KCI LEACHING

ETCHING DEPOSITS HYDRATION DEHYDRATION

Fig. 8. Schematic of weathering process.

(hours)

16

22

Fig. 7. Rate of attack by various concentrations of potassmm chlonde [5].

P B Adams / Glass corroston

197

detachment of reaction products, water attack may in fact become alkali attack as alkali is extracted into solution. If reaction products are removed, it retains its ion exchange character. Often it is more properly described as a hybrid system where ion exchange and silica network destruction proceed simultaneously. Diffusion is the initial rate controlling factor so that the usual square root dependence on time is found. As the sdica layer thickens, the reaction with silica may become equivalent to alkali extraction and the reaction tends to be strictly proportional to t~me. Attack by salts may be, in some cases, similar to water. However, the deviations abound, and they are not always readily explained (fig. 7). Weathering of glass generally refers to reactions that occur between glass and various atmospheric components, principally water [10]. Fig. 8 illustrates the process. It involves a variety of steps, including those previously discussed, which occcur in sequence or s~multaneously, such as adsorption of water, ion exchange of hydronium ion for alkali ion in the glass, hydration to form a thin layer, as well as physical changes resulting from wetting and drying. Without atmospheric water, there is no weathering. Among commercial glasses, none but the poorly durable glasses will show any detectable weathering below about 20 to 30% relative humidity (fig. 9). At the other extreme, at humidities above 90 to 95% RH, weathering may also cease because of dilution and run-off of reaction products (fig. 10),

I00

E
6 B

3O

0
0

IO

i "t
30 50 75 90

33 %z
z
0

0
N

16
B

30

50

75

90

% RELATIVE HUMIDITY
Flg. 9. Sodmm oxide found on surface of a soda-hme glass versus relatwe hurm&ty and time [10]

% RELATIVEHUMIDITY
F]g 10. Water and sodmm oxade found on surface of a soda-brae glass versus relative humadity [10].

198

P B Adams / Glass corroston

Table 1 Corrosnon of typical commercial glasses Glass type Weight loss ") m: 5% N a O H Soda-hme Borosfl]cate High zircoma Alurmno slhcate High silica Vitreous sihca a) m g / c m 2 after 24 h at 95 o C. 2 5 0.4 4 2 2 5% HC1 0.02 0.005 0 0l 0.2 0.001 0 O00X

3. Composition and corrosion resistance

The corrosion rates of six commercial glasses, a soda-lime, a borosilicate, a high zirconia silicate, an alumino silicate, a high silica and a 100% silicate glass are shown in table 1. Their resistance to strong alkali differs about 10-fold, whereas resistance to acid varies about 1000-fold. In qualitative terms, this relationship, as exhibited for the broad family of silicate glasses, is shown in fig. 11. Experience indicates that a durable commercial glass will survive water attack for a long period of time. One would predict that a chemically resistant borosilicate glass would dissolve at the rate of about one millimeter in 25 000 years or more. But this is still several orders of magnitude faster than is believed to be the case for many natural glasses. The predominant compositional variable influencing solubility is generally silica content. The reason is evident from the stability diagram of fused silica in aqueous solution at various p H (fig. 12) [11]. Noting that a low activity of

\ /X

// / / /

!o
-2 i

TYPICAL C0 MMERCIA L'-~j~,,, , GLASSES ~ . . . ~ . . . . , . ~

i

.,sio3
n
13

I
I0

= pH

I
12

I
14

H 7

II

Fig. 11. Relatlonslup between p H and corrosion rate for a variety of glasses,

Fig. 12. Stabihty chagram of fused sdlca m aqueous soluUon versus p H [11]

P.B Adams / Glass corroston

199

AIs+
).I--

/ AlOe//

,.)

-6
i I/% I I I I I

6pH

I0

12

14

Fng. 13. Stabthty &agram of alumina m aqueous solution versus p H [11].

- 4 is equivalent to less than one ppm SIO 2 in solution, it becomes apparent why, in the absence of ion exchange, a high silica glass is so extremely resistant. Stability diagrams for other glass components provide additional insight into glass corrosion resistance. Fig. 13 shows alumina, commonly found in almost all natural glasses [11]. As predicted here, the glass technologist knows that alumina enhances durability in neutral solution, at the expense of resistance to strong alkali and strong acid. Thus, a glass that has good weathering resistance may be a poor laboratory vessel. The alkaline earth elements which are common to almost all natural glasses also impart exceptional stability in neutral solution to a silicate glass. Hydrogen ion exchange is strongly dependent on p H and Ca 2+ activity (fig. 14) [11]. Calcium hydroxide is extremely insoluble in water so that reactions are suppressed. Furthermore, silicic acid probably provides a diffusion barrier at low pH. So the effect of the alkaline earths derives more from their presence in solution than their presence in the glass. Adsorption of water, a key factor in weathering, is dependent on glass composition. A commercial soda-lime glass responds quite differently in this respect than a borosilicate glass (fig. 15). In fact there is a wide spectrum of

)I'-

7I-.3

TM

-4

SILICA
I I I I ~ I I

pH

I0

14

Frog 14. Stability dnagram of calcmm slhcate in aqueous solution versus p H [11].

200

P.B. Adams / Glass corrosion

121

/SODA-LIME

r
=.
30

/
/&.o
50 75 90

-;- J
% RELATIVE HUMIDITY

Fig. 15. Water adsorbed on some glasses after 7 days versus relative hunudity [10]. resistance to weathering (fig. 16) such that high alumina glasses may be hydrophobic. Variance in hydration rate, as a consequence of differences in silica content, as for obsidian is a most important factor in natural weathering (fig. 17) [12]. Ancient m a n - m a d e glasses are sometimes viewed as a potential source of
WORST GLASSES

"

~,,,'

i I I "//d~?/~'X,,~2+ ~ ~

o LI////.$~.~,o.. ~.,% FI I////J~',~,,,_," ~ ~ "

AF
0

i
I

t "NG

T
2

4 6 8 TIME (WEEKS)

I0

12

Fig. 16. Weathering of various glasses at t, gh humidity

P B Adams / Glass corroston

201

1500

,ooo

x, q,

.J

[ I

Io IS TEMPERATURE

20 (% G)

25

30

Fig. 17. Hydration rates of obsidian with a range of slhca content [12]

information about how and why silicate glass survives in the natural environment. The earliest of these glasses date to about 1500 BC, so that they have survived only a brief 3500 years, rather insignificant when compared to the millions of years for some tektites. Most ancient glasses have weathered significantly [13]. A few have survived virtually unaffected. CompositionaUy they are almost all soda-lime glasses, which are relatively easy to melt and not highly durable. For the most part, those excavated in places with drier climates such as Egypt, are less weathered than similar glasses from-wetter areas throughout the Near East and Europe. Some bear thick crusts of weathering products while others are simply iridescent. There are a few peculiar glasses from India and the Black Sea area which have survived noticeably better than other glasses from the same regions. Interestingly, these are soda-lime glasses of relatively high A120 3 content, about 6-10%. Estimates of long term, ambient corrosion are shown m table 2. Some commercial glasses should survive for long periods of time, especially if during that time there are periods when water is absent, when temperatures are low or when solvents are saturated.
Table 2 Calculated corroston rate m pure water Loss m thickness a) (ram per 106 y) Soda-hme Borosihcate High slhca
a) A t 25 o C

1000 10 1

202

P B Adams / Glass corrosion

Table 3 Typical composttlonsa) of some commeroal and natural glasses SO2 Soda hme Borosdlcate High slhca Alununo slhcate Obsidians Tektites
a) In wt%.

A1203 2 2 1 25 10-20 5-35

R O b)

R 2 c) 13 5 0 0 5-10 0-7

73 81 96 65 50-75 50-88

12 0 0 10 0-15 1-25

FeO+ Fe203 0 0 0 0 0-12 0-12

B203 0 13 3 0 0 0

b) IncludesMgO, CaO, and BaO c) IncludesNa20, K20 and Li20.

Table 3 shows a few commercial and natural glass compositions [14,15]. Seldom do we find high A1203 or high alkaline earth commercial glasses unless significant amounts of alkali or boron - needed for ease of melting - are also present. On the other hand, the natural glasses invariably contain no boron, little alkali and relatively large amounts of A1203 a n d / o r alkaline earth oxides, all of which impart exceptional stability in neutral solutions. Only on rare occasions do we find a commercial glass of such composition, because technologists are usually trying to design a material for a m o d e m environment that includes acids and bases. But if a glass was needed to survive for a long time where near neutral conditions prevailed the technologist would start with silica and add generous amounts of alumina and lime or magnesia. Tektites in general, have apparently survived much longer than obsidians; and obsidians are often reported to be substantially hydrated at the surface, in contrast to tektites. This may relate to structural differences associated with original water content when formed. The extremely low water content of tektites may make them virtually inpregnable to hydration.
4. l n h o m o g e n e i t y

All glass products formed by man or nature are inhomogeneous on some scale. Surface composition often differs from bulk composition. Sometimes, the surface will nucleate subsurface crystal growth. Reboil leaves the surface peppered with tiny bubbles. Also present are bulk inhomogeneities, such as bubbles, cords, stones and crystals. Some inhomogeneities may be erased by strong corrosive action. But other fairly subtle inhomogeneities may influence the course of corrosion, or persist in spite of it. Hence, the inhomogeneities in the original glass will leave their mark. Then, there is the corrosion environment itself. Even in a laboratory situation, it is rarely homogeneous unless meticulously controlled. The products of reaction continually alter the reaction interface chemistry. A glass of water left on the shelf is changing as alkali is

P.B Adams / Glass corroswn

203

extracted. Silica dissolves, and then precipitation occurs. A sea bed is changrag; the annual weathering rings, found on ancient glass retrieved from the ocean floor, testify to cyclical changes [16]. Reactions are constantly altering the stated conditions. Silica saturates the solution, and then if iron is present, iron silicates may precipitate. Clays such as bentonite react with active species such as sodium. Surface deposits, as well as pits and scratches, lead to localized attack. Even ordinary window glass weathers in a non-homogeneous manner. As alkali is extracted, the p H rises and falls. Changing relatwe humidity causes the solvent layer on the glass to become dilute and then concentrated. Crystals form which act as nuclei for enhanced corrosion centers. Runoff flushes the surface and then collects along the edge. So natural weathering is a dynamac, ever changing, process that varies point to point on any surface leawng a record that may in turn be obliterated by the next cycle or may remain to attest to its presence.

5. Recent work For about two decades, work has been underway to develop a glassy host for nuclear waste. A key property is long term corrosion resistance in a disposal site. Various sites are under consideration, ranging from sea-bed to burial in granite, basalt, shale or salt. Prediction of glass durability has become a topic of intensive study. The national urgency of this task has attracted m a n y competent scientists. The result has been that advances are being made at an accelerating pace. Three papers presented at the 85th Annual American Ceranuc Society Meeting are indicative of the progress being made. Sullivan and Machiels [17] modeled the hydration kinetics of waste glass m

ISURFACE[

INTO SOLUTION

LEACHANT ONTO SURFAGE

F~g. 18. Representation of potential rate controlhng mechamsm [17].

204
*-GLASS SURFACE

P.B Adams / Glass corroston

//
/ / ='-- TRANSITION / / ZONE // // // ~// DRY GLASS // // / /
/ /

SOLUTION H

HYDRATED GEL LAYER

Fig. 19. Representationof aqueous attack process[19].

contact with an aqueous solution (fig. 18). They concluded that " I n the absence of appreciable network dissolution.., growth of the gel layer becomes eventually proportional to the square root of time; however as long as processes in the transition zone are rate controlling, growth is proportional to time." Tlus is of interest since it coincides with the linear hydration rate found for obsidians. Plodinec et al. [18] have extended the thermodynamic approach to a wide variety of glasses including obsidian. Their results are reported in these proceedings. This approach seems exceptionally promising for ranking glasses in order of reactivity. Pescatore and Machiels [19] have consolidated the mathematics of "four Basic Mechanisms, which are potentially rate controlling", into a single approach relying on thermodynamics and kinetic principles (fig. 19). The result is a computer program, which incorporates a description of the solution rate at the glass surface, redeposition of species back onto the surface dissolution of the solid network at the ion exchange front, and ion exchange of species into the bulk glass. In the final analysis a comprehensive view of near neutral corrosion must consider composition, kinetics, water adsorption rates, changing solution activity and, also, mechanical events which disturb or destroy protective layers, as well as the chemistry and physics of "soil" retention on the surface of a material [20]. The latter includes the effects of particle size, pH, relative humidity and glass composition. Finally, there are also biological agents which interact with glass [21].

6. Summary The essential mechanisms of glass corrosion have been understood for some time. Tens of thousands of experimental tests leave few frontiers of composition research untouched. Glass technologists understand quite well how to formulate a combination of oxides to make a glass that will resist neutral corrosion. So, it does not appear that these natural glasses can teach us much about corrosion theory.

P B Adams / Glass corroston

205

N a t u r e has, however, d e m o n s t r a t e d that certain glasses can resist her chemical forces for extremely long periods of time, perhaps longer than we w o u l d have dared predict. This is n o t because we did n o t u n d e r s t a n d corrosion theory, b u t because we did not a n d stdl do n o t have a n accurate record of the e n v i r o n m e n t to which these glasses have b e e n exposed. It seems that we are often puzzled by the different a p p e a r a n c e of similar materials a n d a t t e m p t to relate them to some p e c u h a r i t y of the glass. It m a y be that the record o n these n a t u r a l glasses for which we should be looking is the record of climate and weather. Perhaps we can make better use of these differences as clues to past e n v i r o n m e n t . The d r a m a t i c difference between tektite survival a n d o b s i d i a n survival, despite the fact that m a n y are similar in gross c o m p o s m o n , deserves a final c o m m e n t . Tektites reportedly c o n t a i n a few p p m water. O b s i d i a n s c o n t a i n at least a few tenths of a per cent of water. This m a y result in a different structure which is more a m e n a b l e to diffusion of water. I n spite of the fact that corrosion theory is well d o c u m e n t e d , predicting a 30 m i l h o n year life for a typical tektite stretches our theories. But those theories are n o w b e i n g fine t u n e d b y a group of nuclear waste dtsposal explorers. Let us hope that they can certify them, perhaps with the aid of these a n c i e n t materials, so that they can insure that a million years from n o w there will be other scientists who will study the nuclear waste glasses b e i n g formed today a n d conclude that the T w e n t i e t h C e n t u r y scientist knew his business well.

References
[1] L. Holland, The Properties of Glass Surfaces (Chapman and Hall, London, 1964). [2] C R. Das, Glass. Ind. (1969) 422. [3] S M. Bond and J. Fracklewlcz, Phys. Chem. Glasses 2 (4) (1961) 111 [4] R W. Douglas and T.M.M.' EI-Shamy,J. Amer. Cer. Soc. 50 (1) (1967) 1 [5] E.B. Shand, Glass Engineenng Handbook (McGraw-Hdl, New York) [6] P.B. Adams, Ultra Purity (Dekker, New York) p 293. [7] D M. Saunders and L.L. Hench, J Am. Cer. Soc. 56 (7) (1973) 373. [8] M B Volf, Techmcal Glasses (Pitman, London, 1961) [9] G.A Hudson and F R. Bacon, Bull Amer. Cer. Soc 37 (4) (1958) 185 [10] H V Waiters and P B. Adams, J Non-Crystalhne Sohds 19 (1975) 183. [11] A. Paul and A. Youssefi, J. Mat. Scl 13 (1978) 97; A. Paul, J. Mat. Scl 12 (1977) 2246 [12] R.E Taylor, Advances m Obsidian Glass Studies (Noyes Press, 1976) [13] R.H Bnll, Archeology 14 (1) (1961) 18. [14] B.P Glass, Report CR 170518 NASA Contract NAS 526273 (1982) [15] J A O'Keefe, Tektites and Their Ong/n (Elsevier, New York, 1976). [16] R.H. Bnll, Sol Am. 209 (5) (1963) 120 [17] T.M Sulhvan and A.J. Machlels, presented at Am. Cer. Soc 85th Mtg, Clucago (1983) [18] J. Plodmec, C.M. Jantzen and G.C. Wicks, presented at Am. Cer. Soc 85th Mtg, Cbacago (1983). [19] C Pescatore and A J Machtels, presented at Am. Cer. Soc 85th Mtg, Clucago (1983). [20] E.F Cuddlhy, Theoretical Consideration of Soil Reaction, Jet Prop. Lab, Cal. Tech, Pasadena