You are on page 1of 11

Water Penetration—Its Effect on the Strength and Toughness

of Silica Glass
SHELDON M. WIEDERHORN , THEO FETT, GABRIELE RIZZI,
MICHAEL J. HOFFMANN, and JEAN-PIERRE GUIN

When a crack forms in silica glass, the surrounding environment flows into the crack opening,
and water from the environment reacts with the glass to promote crack growth. A chemical
reaction between water and the strained crack-tip bonds is commonly regarded as the cause of
subcritical crack growth in glass. In silica glass, water can also have a secondary effect on crack
growth. By penetrating into the glass, water generates a zone of swelling and, hence, creates a
compression zone around the crack tip and on the newly formed fracture surfaces. This zone of
compression acts as a fracture mechanics shield to the stresses at the crack tip, modifying both
the strength and subcritical crack growth resistance of the glass. Water penetration is especially
apparent in silica glass because of its low density and the fact that it contains no modifier ions.
Using diffusion data from the literature, we show that the diffusion of water into silica glass can
explain several significant experimental observations that have been reported on silica glass,
including (1) the strengthening of silica glass by soaking the glass in water at elevated tem-
peratures, (2) the observation of permanent crack face displacements near the crack tip of a
silica specimen that had been soaked in water under load, and (3) the observation of high
concentrations of water close to the fracture surfaces that had been formed in water. These
effects are consistent with a model suggesting that crack growth in silica glass is modified by a
physical swelling of the glass around the crack tip. An implication of water-induced swelling
during fracture is that silica glass is more resistant to crack growth than it would be if swelling
did not occur.

DOI: 10.1007/s11661-012-1333-z
 The Minerals, Metals & Materials Society and ASM International 2012

I. INTRODUCTION Other applications for glass include chemical vessels, flat


screen TVs, PC monitors, bottles, fiber glass insulation,
GLASSES form an important class of industrial jars, and vessels for cooking and other kitchen use.
material for a variety of reasons. Generally, they are Compositions of glasses used for these applications can
transparent, chemically durable, and inexpensive to be found in Reference 2.
make, and they can be easily formed into flat sheets For most applications, glass has to maintain its
for window panes. The most common type of glass for structural integrity; as glass is brittle, fracture is always a
window applications is soda-lime-silicate glass; over matter of concern. For this reason, substantial amounts
90 pct of the glasses made today have compositions that of research funding are spent each year to characterize
are characteristic of soda-lime-silicate glass. Glass is also the strength of glass, to make it stronger and to improve
used in optical fibers for information and voice trans- its reliability. Glass almost always fails from surface
mission. Silica glass is the glass of choice in this cracks. Under a large enough applied stress, a surface
application because it has a low concentration of crack will start to grow, slowly at first, accelerating as it
scattering centers that would otherwise interfere with gets longer, until failure occurs. The process of crack
light transmission. The magnitude of this application growth in glass has been reviewed recently by a number
can be sensed by the fact that enough optical fiber has of authors.[3–5] Water in the atmosphere causes crack
been installed in the world since 1976 to stretch back growth by acting as a stress-corrosion agent, attacking
and forth between the Earth and the Moon 160 times.[1] strained -Si-O- bonds at the tips of cracks and causing
them to rupture. The rate of this stress enhanced
SHELDON M. WIEDERHORN, NIST Research Associate, is with chemical reaction controls the rate of crack growth,
the National Institute of Standards and Technology, 100 Bureau Dr., the higher the stress the faster the cracks grow.
Gaithersburg, MD 20899-8520. Contact e-mail: sheldon.wiederhorn Generally, two approaches have been used to improve
@nist.gov THEO FETT, Guest Scientist, GABRIELE RIZZI,
Mathematical Technical Assistant, and MICHAEL J. HOFFMANN, the strength and reliability of glasses. The first is to
Head of Institute, are with the Institute of Applied Materials, eliminate surface cracks and then to protect the surface
Karlsruhe Institute of Technology, Haid-und-Neu Str. 7, 76131 of the glass with a polymer so that new cracks will not
Karlsruhe, Germany. JEAN-PIERRE GUIN, CNRS Researcher, is be generated. This approach is used for optical fibers.
with the Larmaur ERL-CNRS 6274, University of Rennes 1, 35000
Rennes, France.
Silica glass is forced through a nozzle in a highly viscous
Manuscript submitted January 5, 2012. but fluid state, so that any damage to the glass fiber
Article published online September 8, 2012 while it is forming will self-heal. Once through the

1164—VOLUME 44A, MARCH 2013 METALLURGICAL AND MATERIALS TRANSACTIONS A


nozzle and sufficiently cooled, the glass is coated with a treatment may be as high as –700 MPa,[2] and the glass
polymeric coating to protect the surface of the fiber from can be as much a six times as strong as a fully annealed
further damage. Glass fibers made in this manner are glass of the original composition. For ordinary articles,
essentially free of cracks and possess strengths close to say beverage bottles, the penetration depth may be about
14 GPa.[6] This method of strengthening glass works very 40 lm, but for special applications (windscreens for
well with optical fibers, but it is too expensive to apply to airplanes, smart-phone windows, etc.) the penetration
glass in ordinary use, say as in wind-screens or windows depth may be as much as 300 lm.[2] This technique
for glass doors. Flat screens on television sets and requires mobile ions in the glass to work, so it too cannot
computer monitors, on the other hand, are made by a be used for silica glass, which contains no mobile ions.
flow process and hence are very strong, being relatively
free from flaws, but flaws can be introduced by handling,
C. Transformation Toughening
which reduces the strength of the glass with time.
The second approach to strengthening glass is to A possible way of strengthening or toughening silica
apply a compressive stress to the surfaces of the glass. glass depends on the local expansion in volume that
This technique is most easily applied to flat glass articles occurs when water enters into the structure of silica
such as glass doors for dwellings and offices, transparent glass. This local expansion can be understood by
panels for smartphones, and windows for automobiles. discussing transformation toughening of ceramic mate-
The reason that the method works so well is that the rials. Ceramics can be toughened if they contain a
compressive stress keeps the cracks closed; for the glass crystalline phase that undergoes a volume increase due
to fail, a high enough tensile stress has to be applied to to a phase transformation when subjected to the high
overcome the compressive stress in the surface of the stresses present at the tips of cracks. The phase
glass. The two methods of applying such compressive transformed zone behind and around the tip of the
stresses are thermal tempering and chemical tempering. crack forms a compressive layer that is constrained by
the surrounding material. For the crack to move, it must
propagate against this compressive stress field; hence, a
A. Thermal Tempering higher stress is required to propagate the crack than in
In thermal tempering, the finished glass article is the absence of the phase transformation.
heated well above the glass transition temperature The equation that is used to quantify this process was
(viscosity  1012 PaÆs).[2] Both sides of the glass plate first derived by McMeeking and Evans[8]:
are then cooled with a blast of cold air. A parabolic
ev  E pffiffiffiffiffiffiffiffi
temperature gradient develops across the thickness of Ksh ¼ 0:22 xeff ½1
the plate as a consequence of the cooling. Until the 1m
temperature falls below the glass transition temperature, where Ksh is the shielding stress intensity factor, ev is the
no stresses develop within the glass plate. As the glass unconstrained volume expansion of the transformed
plate approaches room temperature, the material near material, E is Young’s modulus, m is Poisson’s ratio, and
the outside surfaces of the glass plate stops contracting 2xeff is the width of the transformation zone. Figure 1
first, because they are cooler than the glass in the center shows these terms with reference to the crack geometry.
of the plate. Continued contraction on the inside of the
plate puts the outside volume of the glass into a state of
compressive stress, while the inside is subject to a tensile
stress. During this process, the parabolic temperature
gradient is converted into a parabolic stress gradient.
Thermal tempering is an excellent way of strengthen-
ing glass. Fully tempered commercial glasses can have a
surface stress as high as 100 MPa[2] and can be four
times as strong as the fully annealed glass from which it
is made. Glasses can be thermally tempered only if it has
a high thermal expansion coefficient, such as (80 to 100)
9107/C. As silica glass has a thermal expansion
coefficient of only 5.5 9 107/C, thermal tempering is
not an effective method for improving its strength.

B. Chemical Tempering
Chemical tempering works by an ion exchange
process taking advantage of the fact that potassium
ions K+ in glass are much larger than lithium Li+ ions, Fig. 1—Schematic diagram of transformation toughening. As the
0.133 nm vs 0.060 nm, respectively.[7] The process works crack propagates, a transformation zone is formed around the tip of
by holding the glass piece in a molten bath of potassium the crack of width 2xeff. The volume of expansion within the trans-
formation zone creates a compressive stress on the surfaces of the
nitrate at a temperature slightly less that the strain point crack and around the crack tip that requires an additional applied
of the glass (viscosity = 1013.5 PaÆs) for a period of load for crack propagation. The toughening comes primarily from
several hours. The compressive stress developed by this the compression in the wake of the crack tip.[8]

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 44A, MARCH 2013—1165


The crack-tip stress intensity factor has to be equal to crack tip that enhance diffusion at the crack tip during
the sum of the applied stress intensity factor Kapplied and the period that the load is applied to the crack. After the
the shielding stress intensity factor Ksh. crack is forced to propagate through the compression
zone (by increasing the mechanical force on the speci-
Ktip ¼ Kapplied þ Ksh ½2 men), the stresses due to water diffusion cause an elastic
This equation is applied in a situation for which a phase rebound (in the vicinity of the crack tip) that can be
transformation occurs around the crack tip. In silica glass, quantified and used to estimate the mass fraction of
however, a phase transformation does not occur so that water in the glass and the activation volume for
toughening cannot occur in this manner. In silica, the diffusion of water in the silica glass.
volume expansion occurs as a consequence of water Case III—a load is applied to the crack that then
penetration into the glass structure. As demonstrated by propagates through the zone of compression surrounding
Shelby,[9] Bruckner,[10,11] and Shackelford et al.[12] when the crack tip. The shielding stress intensity factor now
water enters into the structure of silica glass, the glass causes a substantial shift of the position of the v-KI curve,
suffers a small increase in volume; the volume increase is which greatly improves resistance of silica to crack
proportional to the mass of water entering the glass. growth. Taking advantage of the shielding zone, it may
Water entry into the silica glass structure is particularly be possible to design a better glass composition that will
strong in the vicinity of a moving crack.[13,14] In studies by enhance the process of water penetration around a crack
Tomozawa et al.[13] and by Lechenault et al.[14] fracture tip and, thus, increase the toughness of the silica glass.
surfaces were formed by the passage of a crack through In this article, we review recent evidence of water
double cleavage drilled compression specimens penetration in the vicinity of the crack tip and discuss
(DCDC),[15–17] while the specimen was exposed to either some of its consequences. We start with case I, for which
water or deuterium oxide. Measurements[13,14] demon- no load is applied to a surface crack, and water
strated that the concentration of water was significantly penetrates into the glass surrounding the crack by
higher near the fracture surface than for silica glass merely simple diffusion, thus increasing the strength of the
exposed to water for a time equal to that required for the glass. Experiments that quantify the effect of this
passage of the crack through the specimen.[13,14] In the process on the strength of glass have already been
absence of high tensile stresses at the crack tip, water carried out for silica glass.[19,20] These experiments will
diffusion into silica glass is very slow.[18] be analyzed; predictions of strength before and after
Three cases of diffusion in the vicinity of a crack tip water penetration agree with experimental measure-
can be envisioned, Figure 2: ment, within the scatter of the measurement.
Case I—No load is applied to the crack and water In case II, a fracture mechanics specimen (double
penetrates uniformly along the fracture surfaces and the cantilever beam) was used for the study. A load was
external surface of the specimen; swelling, as a conse- applied to the specimen, which was submerged in water,
quence of water penetration, affects the strength of the and held there for 80 days at an applied stress intensity
silica glass by providing a shielding stress intensity factor of 0.245 MPaÆm1/2. After the hold period, the
factor that tends to close the crack. specimen was split in two and the surfaces were
Case II—a load is applied to the crack, but the load is examined with an atomic force microscope. Profiles of
low enough that the crack does not move; as a the crack from the two fracture surfaces were shown to
consequence, there are high tensile stresses near the match over the entire fracture surface, with the excep-
tion of the point where the crack was arrested. Down-
stream of this point, extra material was found at the
crack surface and was attributed to a volume expansion
as a consequence of elastic relaxation of a water-
penetrated zone. These conclusions will be justified and
their consequences will be discussed below.
Finally, in case III, we predict the depth of penetra-
tion around a moving crack and compare the calcula-
tion with actual measurements. Quantification of water
diffusion around the tip of a moving crack requires a full
two dimensional solution of the diffusion equation.
Within scatter, the results of the calculation agree with
the two experiments that have been carried out to date.

II. CASE I—WATER PENETRATION INTO


SILICA GLASS SURROUNDING A SURFACE
CRACK
A. Method

Fig. 2—Illustration of water penetration into the region near a crack


This is by far the simplest of the three cases shown in
tip. Each of these three cases of water diffusion at a crack tip in sil- Figure 2. Water diffuses into both fracture surfaces and
ica glass is discussed in the text. into the external surface on which the crack terminates.

1166—VOLUME 44A, MARCH 2013 METALLURGICAL AND MATERIALS TRANSACTIONS A


In both kinds of surface, water diffusion into the glass negative shielding contribution to the edge crack; see
causes swelling, which in turn causes a shielding stress Appendix.
intensity factor Ksh at the crack tip. For diffusion
directly into the fracture surfaces, the shielding is given ev  E 1:465  b
Ksh2 ffi  pffiffiffi ½7
by a slight variation of Eq. [1]. 3ð 1  m Þ a
ev  E pffiffiffiffiffiffiffiffi where a is the length of the surface crack and b is the
Ksh1 ¼ w xeff ½3 thickness of the diffusion layer from the free surface.
1m
The total shielding stress intensity factor for the two
where Ksh1 is the shielding from the crack surface and surface layers is given by adding Eq. [3] to Eq. [7].
the coefficient w is 0.255; see Appendix. On the frac-
ture surfaces and on the external surface of the glass ev  E pffiffiffiffiffiffiffiffi ev  E 1:465  xeff
Ksh ¼ 0:255 xeff  pffiffiffi ½8
specimen, we assumed that water diffusion into the 1m 3ð 1  m Þ a
glass occurs in the same way as at a stress-free, flat
surface. This assumption yields a one-dimensional dif-
fusion problem, the solution of which is a complimen-
tary error function of water concentration, Cw(z, t). B. Experimental data
The water concentration depends on the diffusion
time, t, and distance, z, from the glass surface. Strength data that can be used to test our model
  have been published by Ito and Tomozawa[19] and by
z Hirao and Tomozawa.[20] These authors used abraded
Cw ðz; tÞ ¼ Cw0  erfc pffiffiffiffiffiffiffiffiffiffiffiffiffi ½4 rods of annealed Vycor (C7900)*,** 3 mm in diameter
2 Dw  t

The effective penetration distance for Eq. [3] is given *The use of commercial names is for identification only and does not
by[21] imply endorsement by the National Institute of Standards and Tech-
pffiffiffiffiffiffiffiffiffiffiffiffiffi nology.
xf ¼ z ¼ Dw  t ½5
where Dw is the diffusivity and t is the time for diffusion.
We now assume that Eq. [5] gives the distance r from the
crack surface to the diffusion penetration boundary,
r = xf, including the crack tip, where the diffusion **Vycor glass contains about 96 pct SiO2, 3 pct B2O3 and 1 pct
boundary now forms a cylinder-like zone around the tip Al2O3 by weight.
of the crack. In an earlier paper,[21] the zone was
assumed to be a circular cylinder. When Eq. [5] is used and 60 mm long for their study. Sixty rods were
to define the diffusion distance, the value of the shielding tumbled in a small, evacuated ball mill (1 L volume)
from Eq. [3] is identical to that obtained by using the for 2 hours, with 100 g of 600 grit SiC. After abrad-
entire error function in the analysis to obtain Ksh1. In ing, the specimens were washed in water at room
the analysis for Eq. [5], we assumed a constant concen- temperature, cleaned three times in an ultrasonic bath,
tration of water from the surface to xf.[21] and then dried at 403 K (130 C) for 3 hours. A
The diffusivity Dw is given by the following equation: minimum of 20 specimens were used for each four-
Dw ¼ A0 expðQ=RHÞ expðDVw rh =RHÞ ; point bending strength measurement.
½6 This procedure supplied a set of uniformly treated
D0 ¼ Dw ðrh ¼ 0Þ specimens for their study. For as received strengths, one
set of specimen was tested as is. Another set was heated
where Q is the activation energy and DVw is the in water at 361 K (88 C) for either 112 hours[20] or
activation volume for diffusion of water in silica glass. 240 hours.[19] All strengths were measured in four-point
The diffusion process depends on temperature h in K bending in N2 (l). Previous experience[22] shows that
and the hydrostatic stress rh at the crack tip. when fracture mechanics specimens are broken in N2 (l),
A solution for Ksh1 can be obtained for an edge- subcritical crack growth does not occur. The as received
cracked half-space (this is our model) by combining Eqs. strengths were 181 MPa for 112 hours[20] and 134 for
[3], [5], and [6], yielding Ksh1 as a function of Dw, t, and 240 hours.[19] The initial size a of a crack when tested in
water concentration. Experimental measurements of the liquid nitrogen is given by
diffusivity and the water concentration in silica glass
may be found in Reference 18 and the conversion of 1h i
water concentration to volume expansion is found in a¼ ðKIc =1:12rÞ2 ½9
p
Reference 9.
As shown in case I, Figure 2, for an edge-cracked where the critical stress intensity factor of Vycor KIc in
specimen, a swollen layer of thickness, xeff, extends from liquid nitrogen is 0.741 MPaÆm1/2.[22] For the 112 hours
the surfaces of the crack and from the free surfaces exposure,[20] a crack length of 4.26 lm is obtained; for
of the specimen. Expansion of the penetrated layer the 240 hours exposure[19] the crack length is 7.76 lm.
along the free surface is constrained by the underlying In the calculations of a, we assume that shallow wide
material. The expansion of this penetrated layer causes a cracks are formed by the abrasion process so that Eq. [9]

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 44A, MARCH 2013—1167


Ktip pffiffi
rh;tip  rh ðKtip Þ ¼ ð2=3Þð1 þ mÞ pffiffiffiffiffiffiffiffi cosðu=2Þ ¼ 2j= r
2p r
½11
where r and u are the polar coordinates with the origin
at the crack tip, and rh,tip is the hydrostatic stress caused
by the external loading. In the crack tip region, the
diffusion coefficient is strongly affected by this stress.
Substituting Eq. [11] into Eq. [10] yields the following
equation for r1/2
pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi
xeff ¼ r1=2 ¼ D0  t expðDVw j=RH r1=2 Þ ½12

Equation [12] is a transcendental equation and has to


be solved by numerical methods.[23] For the purposes of
this article, a complete solution for r1/2 is unnecessary;
Eq. [12] alone will be used for comparison with
experimental data.

2. Comparison of theory with experiment


Fig. 3—A comparison of the strengths of silica rods tested in N2 (l) In order to investigate the effect of crack tip stress on
both before and after exposure to water at 361 K (88 C). The error diffusion, we carried out the following experiment[23]:
bars give the standard deviation of the measurements. The dark tri- Using a double cantilever specimen, a static load was
angles are the predicted strengths after high temperature exposure. applied to the ends of the specimen and held for 80 days.
Data are taken from Refs. [19] and [20].
During this period, the specimen was submerged in
room-temperature water. The applied stress intensity
for an edge crack can be used. The calculated strengths factor was low for silica glass, 0.245 MPaÆm1/2, low
obtained from these crack lengths after exposure are enough that the crack did not move during the loading
197 MPa[20] and 150 MPa.[19] These values are the dark period. At the end of the loading period, the stress
triangles plotted in Figure 3. The simpler but less accurate intensity factor was increased to 0.45 MPaÆm1/2; crack
evaluation with the approximation in Eqs. [7] and [8] yields growth started immediately upon increasing the load to
slightly lower values of 196 MPa[20] and 148 MPa.[19] The the crack. The specimen was then broken in two and
measured strengths after exposure were 205 MPa for the both fracture surfaces were examined by atomic force
112 hours exposure and 155 for the 240 hours exposure. microscopy (AFM).
As can be seen from Figure 3, the increases of The profiles of both fracture surfaces are shown in
calculated strengths are from 67 pct to 76 pct of the Figure 4(a). The crack propagated from right to left (see
final increase of measured strength. So, they are a bit on arrow), and both profiles matched over the scanned
the low side, but considering the scatter in the strength region with the exception of a small segment down-
(Figure 3), diffusivity (standard deviation  20 pct to stream from the point of crack arrest, point a in the
30 pct of the mean[18]), and concentration (standard figure. Over this segment, the lower and upper profiles
deviation  15 pct to 20 pct of the mean[18]), the overlapped, indicating the presence of extra material
comparison between experimental strengths and calcu- along the crack front that was not present over the rest
lated strengths seems to be reasonable. of the profile. We attributed the extra material to an
elastic stress relaxation once the specimen was broken in
C. Case II—Diffusion enhanced by an applied stress two and the constraints at the crack tip were removed.
Surface displacements can be measured directly by use
1. Deriving the pertinent equations of the AFM, i.e., Figure 4(a). The displacements can
In case II, Eq. [5] is used to calculate the diffusion also be calculated by carrying out a finite-element
depth. However, because the crack tip is now associated analysis to determine how much surface displace-
with high stresses, the diffusivity in Eq. [5] is considered ment should occur as a result of the relaxation of the
to be stress dependent. Substituting Eq. [6] into Eq. [5] stress.
and rearranging terms, we obtain the following equation The profile given in Figure 4(b) is obtained from
for the distance at which the concentration decreased to Figure 4(a); the profile given in Figure 4(c) is a calcula-
50 pct of maximum: tion using a finite-element analysis.[23] Assuming that the
pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi strain due to water diffusion was constrained to the
r1=2 ¼ Dw  t ¼ D0  t expðDVw rh =2RHÞ ½10
cardioid shape shown in Figure 4(d) and that the crack
split the cardioid shape into equal parts when it propa-
If Ktip denotes the stress intensity factor at the tip of a gated, half the measured displacement in Figure 4(a)
crack, then the singular near-tip hydrostatic stress field went with each part. By comparing the results of the
is given by the following equation: finite-element analysis with the AFM measurement, we

1168—VOLUME 44A, MARCH 2013 METALLURGICAL AND MATERIALS TRANSACTIONS A


Fig. 4—Determination of glass expansion due to water penetration of a crack tip: (a) AFM scan of a crack tip that was held under load in
water at room temperature, Kappl = 0.254 MPaÆm1/2 for 80 days. (b) Profile of region of swelling on the fracture surface. (c) Finite-element anal-
ysis of cardioid diffusion volume subject to expansion. (d) Cardioid diffusion volume. Taken from Ref. [23].

obtained the equations shown in Figure 4, and from these the known variables into Eq. [12], we obtained
equations and the AFM measurements near the point of DVw = 15 9 106 m3/mol. Thus, the activation volume
crack arrest, point a, we reached the conclusion that the for the diffusion of water through silica glass is of the
penetration distance x = 80 nm and the strain within same order as the molecular volume of the water, 18 9
the water penetrated region is eo = 13.7 pct. This strain 106 m3/mol. The method presented here is a novel way
is close to the strain of 11.7 pct measured by Tomozawa of determining the activation volume for the diffusion of
et al.[13] as a consequence of water penetration into glass water in silica glass.
during crack growth. These values are about a factor of
50 times larger than that due to water penetration into
D. Case III—Diffusion through the tip of a moving crack
silica glass in the absence of an applied stress, Zouine
in silica glass
et al.[18] The dilatational stresses at the crack tip probably
create the extra volume needed to accommodate the Two experiments on silica glass[13,14] have been
additional water molecules. carried out to show that during crack growth, water
The activation volume DVw for the diffusion of water diffuses through the crack tip and into the glass to leave
through silica glass can be calculated from Eq. [12], a layer of water penetrated glass at the fracture surface
using the swelling data presented in Figure 4. For the of the test specimen as the crack moves on. Both studies
case of a stress-induced weak phase transformation, the used the DCDC specimens[15–17] to produce fracture
surfaces that could be analyzed for water penetration

after being fractured in two. Crack growth was carried
The effect of back stresses on the swelling volume is not considered. out either in water[13] or in N2 (g) containing a high
See Reference 23 for a discussion of this point.
concentration of deuterium oxide.[14] In Tomozawa
et al.[13] the water concentration was analyzed as a
zone shape was derived by McMeeking and Evans.[8] function of depth using the nuclear reaction analysis
For the polar angle, u = 0, the relation between r and x technique, whereas Lechenault et al.[14] used the neutron
is given by r1/2 = (8/27)Æx. All of the other parameters reflection technique for the same purpose. Both papers
in Eq. [12] are known, or are experimental variables. concluded that a layer of glass containing absorbed
The only unknown is the activation volume, DVw, for water was present at the surface. Diffusive penetration
the diffusion of water in silica glass. Substituting all of of water into the glass was deeper than could be

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 44A, MARCH 2013—1169


1

C/C0 σh σh(Ktip)
3000
(MPa) t=10-4 s

ϕ=0
2000
ϕ=0
t=1 s t=1 s
0.5
0.3 s
1000 σh,total
0.1 s
2 4 6 8
-4
0.01 s 10 s r (nm)
σh,swell
0.001 s
-1000
t=1s σh(Ktip)+σh,swell(r=0)

ε0=0.13, ΔV=15 cm3/mol


0 2 4 6 8 -2000
r (nm)
(a) (b)
Fig. 5—This figure shows (a) the concentration of water on the projected crack plain as a function of distance r from the crack tip, and the (b)
stresses on that plain as a function of r. The upper part of the figure shows the combined stresses from the Kapplied, whereas the lower part of
the diagram shows the stresses due to swelling.

explained from known values of the diffusivity at room vapor. This difference may explain the difference
temperature. Both sets of authors attributed the between our calculation and the measurements of
enhanced penetration to the very high tensile stresses Lechenault et al.[14]
normally present at the tips of cracks in glass and the Another interesting feature of the curves in
sensitivity of the diffusivity to these stresses (Eq. [6]). Figure 5(a) is the dependence of the curve spacing on
In Reference 21, a two-dimensional diffusion equation hold time. At or near the plateau, C/Co = 1, the
was solved and the thickness of the diffusion layer distance between the curves is small; virtually all the
around the crack tip was calculated. The diffusivity was curves have the same penetration distance r. After that,
assumed to be stress dependent (Eq. [6]), which pre- the curves fan out; the penetration distance increases as
cluded simplification of the partial differential equations the concentration decreases. At C/Co = 0.5, the differ-
used to solve the problem. The solution of the differen- ence between the curves marked t = 0.0001 seconds
tial equation was obtained numerically by application of and t = 1.0 seconds is about 1 nm, while at C/
the procedure NDSolve in Mathematica,[24] as discussed Co = 0.9, the difference is about 3.5 nm. Such behavior
in Reference 21. Some of the results obtained in is very different from that of a complimentary error
Reference 21 are shown in Figure 5. function and is a result of the assumed exponential stress
In Figure 5(a), we show the concentration of water vs dependence of the diffusivity.
the penetration distance as a function of hold-time for a Figure 5(b) shows the stress distribution on the
static crack. The curves in the figure can be converted projected crack plane as a consequence of crack tip
into velocity curves by dividing the penetration distance, swelling. One can see two regions of behavior, connected
i.e., the half height distance in the figure, by the hold by a transition between the two. The dash-dotted curve
time for each curve.[21] In this way, the velocities range rh(Ktip) represents the hydrostatic stress caused exclu-
from 2 9 109 m/s (the curve furthest to the right) to 1 sively by the crack tip, stress intensity factor Ktip
9 106 m/s (the curve furthest to the left). An interesting according to Eq. [11]. This stress dominates at large
feature of these curves is the plateau at C/Co = 1, for distances from the crack tip. Close in to the crack tip the
which the concentration is independent of the crack stress is given by the sum of the crack tip stress and the
velocity. Penetration distances of as much as 1 nm can maximum hydraulic stress due to swelling, rh(Ktip) +
be seen on the plateau. The plateau is similar to the one rh, swell(r = 0). The transition between these two curves
measured by Lechenault et al.[14]; however, their plateau depends on the hold time: For a short hold time, the
was larger, 4.5 nm instead of  1 nm. The size of the transition between the two curves occurs at higher
plateau may depend on the position of the v–K-I curve, stresses than it does for long hold times. The transition
which can be very different depending on whether the distance is relatively short, regardless of the hold time.
test is carried out in water or N2 (g) containing water For all hold times studied, the transition between the

1170—VOLUME 44A, MARCH 2013 METALLURGICAL AND MATERIALS TRANSACTIONS A


two curves was complete by 8 nm, which means that for glass has been solved and the concentration and strain
distances greater than 8 nm, the stresses around the profiles determined,[21] The interaction between the
crack tip are controlled by the crack tip stress intensity crack motion due to the chemical reaction rate at the
factor alone (Eq. [11]). This finding is consistent with the crack tip and the enhanced stress at the crack tip are
AFM observation that displacements around the tip of taken into account in the solution.
an emerging crack are all in the elastic region for
distances greater than 10 nm.[25]
B. Water Penetration: Case I
In the current paper, we explored the possibility that
water diffusion into the surface of silica glass can
III. DISCUSSION produce stresses that are high enough to significantly
affect the strength and toughness of glass. Three types of
A. Chemical Reaction or Water Penetration?
experimental data were analyzed; each showed that the
Water is a well-known stress-corrosion agent for volume expansion associated with the diffusion of water
glasses, which causes delayed failure in glass and a into silica glass surfaces provides a plausible explanation
loading rate dependence of the strength. Without the for experimentally observed increases in the strength
water present, shortened lifetimes and lower strengths and toughness of silica glass. In the first experiment,
would not be experienced by glass. The dominant belief case I, the strength of silica glass was enhanced by
is that stress corrosion in silicate glasses is chemical in soaking it in hot water for extended periods of
nature. Water reacts chemically with the strained Si-O time.[19,20] We showed that swelling of the glass surface
bonds at crack tips causing them to rupture, with the provided a reasonable explanation for the change in
production of two Si–OH groups on the fracture surface strength. Calculating the strength from an edge crack,
of the glass. As the Si-O bonds rupture, the crack we accounted for 67 pct to 76 pct of the strength gain
advances and the rate of crack motion is related to the due to surface swelling. As the cracks in the surface of
rate of the chemical reaction. The kinetics of such bond- the glass were more likely elliptical cracks than edge
breaking chemical reactions have been discussed and a cracks, we feel that the comparison provided a semi-
substantial literature exists on the subject detailing the quantitative explanation for the strength. The problem
types of interactions that are believed to occur. Recent of elliptical cracks in a silica glass surface will be treated
reviews[3–5] clearly describe the experiments and the in a subsequent paper.
experimental results that support the chemical model of
crack growth.
The present paper gives a very different view of how C. Water Penetration: Case II
water reacts with glass. We look at the possibility that Here, we explored the possibility that water diffusion,
water not only acts as a stress corrosion agent but also enhanced by stresses at the crack tip, could explain the
can act as potential toughening agent for silica glass, a swelling observed at the tip of a crack held at an applied
situation that is probably unique in the stress corrosion stress intensity factor of 0.245 MPaÆm1/2 in water for
literature. The diffusion of water into the silica glass is 80 days.[21,23] After the hold period, crack tip profiles
analyzed and its potential action as a swelling agent is were obtained from each of the fracture surfaces. The two
described. The mechanisms of the two processes are profiles matched perfectly except at the point of crack
believed to be totally different. As a stress-corrosion arrest during the hold period. There, extra material was
agent, water reacts chemically as soon as it comes in observed (Figure 4). We assumed that the apparent extra
contact with the Si-O bond. The rate of reaction in material was the result of swelling of the glass as a
region 1[3–5] is determined entirely by chemical reaction consequence of water penetration into the near crack tip
rate theory.[26,27] Diffusion within the glass structure region of the specimen. By measuring the surface lift at the
never enters the picture. point of crack arrest with an atomic force microscope and
As a toughening agent, the water must diffuse into the analyzing the surface lift by finite-element analysis, we
glass structure and cause the glass to swell. The process concluded that the water penetrated a distance of 80 nm
for the diffusion of water into the silica glass structure is into the glass from the crack tip and that the volume strain
not simple because it involves both (1) the migration of at the end of 80 days was 13.7 pct. From these results and
water molecules through the glass structure and (2) a an analysis of the diffusion problem at a crack tip, we
reaction of the water molecules with the glass.[28] The found that the activation volume DVw for water diffusion
effective diffusion coefficient is a parameter that is in silica glass was 15 9 106 m3/mol. The method used
determined by both of these processes,[28] and thus, it here is a novel technique of determining activation
describes the effective diffusion rate of water through volumes for diffusion.
silica glass. To complicate the situation even further, the
parameters that control the rates of diffusion and
D. Water Penetration: Case III
chemical reaction depend on the local temperatures
and stresses in the silica glass near the crack tip. The In the final experiment, we used a two-dimensional
crack-tip stresses can be very high (GPa range) and can finite-element analysis of the diffusion of water at the tip
change very rapidly with distance from the crack tip, of a moving crack to obtain concentration, stress, and
further complicating the situation. A complete two- displacement distributions around the crack tip. Two
dimensional diffusion problem at a crack tip in silica sets of data[13,14] were compared with our theoretical

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 44A, MARCH 2013—1171


analysis of the diffusion problem. The data set obtained have had an effect on the concentration profiles for the
by Tomozawa et al.[13] used the method of nuclear two environments. Clearly, additional experimentation
reaction analysis to measure the depth of penetration of will be needed to clarify the measured differences in
water into the glass, whereas the data obtained by concentration profile.
Lechenault et al.[14] used the method of neutron reflec-
tion to calculate the penetration of deuterium oxide into
E. Effect of Crack Tip Shielding on the v-Kappl Curves
silica glass. The data given by Tomozawa et al. suggest a
shallow penetration of water into the glass surface, One of the more important findings of the article is
about 2 nm; the theoretical prediction suggested a half- the level of the crack tip shielding that develops as a
height penetration distance of 2 nm to 4 nm depending consequence of swelling. This effect was quantified and
on the crack velocity. Our theoretical analysis is discussed in Reference 21. In that article, the effective
consistent with the data collected by Tomozawa et al. shielding stress intensity factor Ksh was calculated from
The experimental results obtained by Lechenault et al. the solution of the diffusion problem and added to the
predict a concentration profile consisting of two parts: a applied stress intensity factor Kappl to yield the crack tip
plateau extending about 4.5 nm from the glass surface stress intensity factor Ktip for each value of the crack
followed by an exponentially decreasing concentration velocity; i.e., Ktip = Kappl + Ksh. The curves shown in
beyond the plateau. The same sort of curve is predicted Figure 6[29] were then plotted. The shift in the position
theoretically, but the plateau only extends about 1 nm of the v-KI curve was about 0.15 MPaÆm1/2 for the most
from the free surface and the half-height distance likely curves, which are shown by the heavy dashed
extends about 1 nm beyond the plateau. The decay is line, and the experimental v–Kappl curve, which is
not exponential but can probably be approximated as shown by the solid line. Shifts of this magnitude in the
such. The Tomozawa et al. curves give no indication of a position of the v-Ktip curve amount to a difference of
plateau, perhaps because the technique used for the about 30 pct in load-carrying ability of the silica glass.
analysis of the water concentration was incapable of the To date, little is known about the dependence of the v-
fine resolution needed to observe this detail. The Kappl curve on the structure of the silica glass or how
differences between the Tomozawa et al. and the the glass structure affected the swelling and, hence, the
Lechenault et al. data may be a consequence of the fact degree of shielding for the glass structure. Additional
that different glasses were used in the two studies, or experiments in this area might prove to be very fruitful
that some characteristic of the glasses differed, for for producing glasses with greater resistance to crack
example, the fictive temperature. Also, the Tomozawa growth.
et al. experiments were carried out in liquid water, One final comment should be made in the discussion
whereas the Lechenault et al. experiments were carried of water as a toughening or strengthening agent for
out in N2 (g) containing a high concentration of silica glass. The method we discuss is shielding of the
deuterium oxide. As the v-KI plots for these two crack tip as a consequence of a volume expansion in
environments are also different, this difference may the glass when water diffuses into the glass structure.

Fig. 6—Crack growth curves. The v-Kappl was obtained from Ref. [29]; calculation of v-Ktip was from the shielding stress intensity factor Ksh in
Ref. [21]. (a) v-Ktip as a function of swelling strain eo. (b) v-Ktip as a function of activation volume, DVw. The most likely swelling strain and acti-
vation volume were 13 pct and 15 9 106 m3/mol, respectively.

1172—VOLUME 44A, MARCH 2013 METALLURGICAL AND MATERIALS TRANSACTIONS A


The predicted strengths, surface displacements, or water APPENDIX
penetration into the glass were of the same magnitude as
those measured experimentally. Based on this observa- Shielding Stress Intensity Factor for the Crack-Face
tion, we feel that the suggested toughening or strength- Swelling Zone
ening process plays an important role in the crack In a finite-element study, the diffusion problem for
growth process. However, we feel it is also important to water-soaked specimens was solved for a semi-infinite
note that this process does not exclude other toughening crack in the half space. Figure A1(a) shows the water
or strengthening processes from occurring at the same concentration contours near the end of a semi-infinite
time. Crack tip shielding by water penetration is based crack. In Figure A1(b), the local water concentration C,
on fracture mechanics and, hence, is by its nature a normalized on the maximum value at the crack surfaces
subset of the theory of elasticity. Therefore, other elastic C0, is plotted vs the crack tip distance. It can be seen that
processes can occur simultaneously with the water the zone size ahead of the tip is reduced. There is also a
penetration and swelling process; the effect of these height reduction visible near the tip (x = 0). Fig-
processes must then be added to the swelling process to ure A1(b) gives the concentration profiles for the x-
assess the total effect. Thus, we might have bond- direction ahead of the crack tip and in y-direction for a
breaking mechanisms that differ from the chemical one large crack-tip distance xfi-¥ and x = 0.
suggested early in this article.[19,20] We might also have From the stress intensity factor evaluation by ABAQUS
other types of shielding, such as crack tip bridging by a 6.2 a shielding coefficient of w = 0.255 was obtained in
precipitate from the crack tip solution.[19,20,30,31] Each Eq. [3]. For a circular diffusion zone, McMeeking and
potential mechanism would have to be tested separately Evans[8] found w = 0.25. Although the zone shape
to assess the probability of its participation in the around the tip deviates from a circular cylinder, the
toughening process; this has not been done to date. resulting stress intensity factor is hardly affected.

Shielding Stress-Intensity Factor for the Swelling Zone


IV. SUMMARY from the Free Surface
In this article, we explored the effect of crack tip In a finite body, a diffusion and swelling layer of same
swelling on the strength of glass. We found that swelling thickness b and volume strain ev extends from the free
due to water penetration into the silica glass can explain surface, see Figure A2. The contribution of this layer to
several experimental observations reported on the the shielding stress intensity factor can be computed as
strength of silica glass and the structure of crack tips
in silica glass.
1. Silica glass increases its strength when soaked in
hot water.[19,20] In this article, we have shown that y
the strength increase can be explained by swelling 0.1
0.2
of the glass around the crack tip, which imposes a
shielding stress intensity factor Ksh on the crack.
0.5
2. Because of a stress-enhanced diffusion coefficient, 0.8
the size of the swelling zone in the vicinity of a
crack tip is greater for a crack under an applied C/C0=1
Crack
x
stress than for one held free of any stress. a
3. For a moving crack in silica glass, the shielding (a)
stress intensity factor is approximately 1
0.15 MPaÆm1/2, depending on the crack velocity, the
activation volume, and the strain increase due to C/C0

water penetration. This shielding process is suffi- y-direction


x→−∞
cient to account for a 30 pct increase in strength
when failure occurs as a consequence of subcritical y-direction
x=0
crack growth. 0.5
4. AFM profiles across points of crack arrest under
load indicate an overlap of the images from oppo-
x-direction
site fracture surfaces, presumably due to swelling of x>0
the glass that has absorbed water into the glass
structure.
0
5. The activation volume DVw for the diffusion of 0.5 1 1.5
water through silica was determined to be 15 9 x/√D t ,
×
y/√D t×

106 m3/mol. (b)


6. The exploration of glass compositions that can take
advantage of crack tip swelling as a means of Fig. A1—Water diffusion through crack surfaces of a semi-infinite
improving their mechanical behavior has yet to be crack: (a) contour lines for constant water concentration and (b)
carried out. concentration profiles in length and height directions.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 44A, MARCH 2013—1173


4 stress free silica specimens,[18] or at fresh fracture

h√a surfaces in silica glass,[13] using the relation between
b concentration of water in silica and volume expansion
3 found in Reference 9.
x
σn
2

REFERENCES
a 1. D.W. Richerson: The Magic of Ceramics, The American Ceramic
1 h ≅ const.
Society, Westerville, OH, 2000.
2. A.K. Varshneya: Fundamentals of Inorganic Glasses, Academic Press,
Inc., Harcourt Brace & Company, Publishers, San Diego, CA, 1994.
3. S.W. Freiman, S.M. Wiederhorn, and J.J. Mecholsky: J. Am.
0.2 0.4 0.6 0.8 1 Ceram. Soc., 2009, vol. 92 (7), pp. 1371–82.
x/a 4. M. Ciccotti: J. Phys. D: Appl. Phys., 2009, vol. 42, p. 214006.
5. B.R. Lawn: Fracture of Brittle Solids, 2nd ed., Cambridge Uni-
versity Press, Cambridge, U.K., 1993.
Fig. A2—Weight function h for an edge-cracked half-space. 6. C.R. Kurkjian, P.K. Gupta, R.K. Brow, and N. Lower: J. Non-
cryst. Solids, 2003, vol. 316, pp. 114–24.
Z b 7. L. Pauling: The Nature of the Chemical Bond, 3rd ed., Cornell
University Press, Ithaca, NY, 1960.
K¼ rn hðx; aÞdx ½A1 8. R.M. McMeeking and A.G. Evans: J. Am. Ceram. Soc., 1982,
0
vol. 65, pp. 242–46.
with the normal stress rn on the crack 9. J.E. Shelby: J. Non-Cryst., 2004, no. 349, pp. 331–36.
10. R. Bruckner: Glastech. Ber., 1970, vol. 43, pp. 8–12.
11. R. Bruckner: J. Non-Cryst. Solids, 1971, vol. 5, pp. 281–85.
ev  E 12. J.F. Shackelford, J.S. Masaryk, and R.M. Fulrath: J. Am. Ceram.
rn ¼  ½A2
3ð1  mÞ Soc., 1970, vol. 53, p. 417.
13. M. Tomozawa, W.-T. Han, and W.A. Lanford: J. Am. Ceram.
and the fracture mechanics weight function h given for Soc., 1991, vol. 74 (10), pp. 2573–76.
the edge-cracked half-space (an edge crack is consid- 14. F. Lechenault, D.L. Rountree, F. Cousin, J.-P Bouchaud, L. Ponson,
and E. Bouchaud: Phys. Rev. Lett., 2011, vol. 106, p. 165504.
ered in our model) by[30] 15. C. Janssen: Proceedings of Tenth International Congress on Glass,
rffiffiffiffiffiffi ! Kyoto, Japan 1974, Ceramic Society of Japan, Tokyo, Japan, 1974,
2 1 X5  xnþ12 pp. 10–23.
h¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ D 1 ½A3 16. T.A. Michalske, W.L. Smith, and E.P. Chen: Eng. Fract. Mech.,
pa 1  x=a n¼0 n a 1993, vol. 45, pp. 637–42.
17. M.Y. He, M.R. Turner, and A.G. Evans: Acta Metall. Mater.,
with the coefficients 1995, vol. 43, pp. 3453–58.
18. A. Zouine, O. Dersch, G. Walter, and F. Rauch: Phys. Chem.
D0 = 0.58852, D1 = 0.031854, D2 = 0.463397, Glasses, 2007, vol. 48, pp. 85–91.
D3 = 0.22721, D4 = -0.82853, and D5 = 0.35138. 19. S. Ito and M. Tomozawa: J. Am. Ceram. Soc., 1982, vol. 65 (8),
The weight function is shown in Figure A2. pp. 368–71.
In an extended range at the surface 0 £ x/a £ 1/3, the 20. K. Hirao and M. Tomozawa: J. Am. Ceram. Soc., 1987, vol. 70
value of h is nearly independent on x/a (6), pp. 377–82.
21. S.M. Wiederhorn, T. Fett, G. Rizzi, M.J. Hoffmann, and J.-P.
Guin: Eng. Fract. Mech., 2012, in press.
1:465 22. S.M. Wiederhorn: J. Am. Ceram. Soc., 1969, vol. 52 (2), pp. 99–105.
h ffi pffiffiffi ½A4 23. S.M. Wiederhorn, T. Fett, G. Rizzi, S. Fünfschilling, M.J.
a Hoffmann, and J.-P. Guin: J. Am. Ceram. Soc., 2011, vol. 94 (S1),
pp. S196–S203.
24. Mathematica: Wolfram Research, Champaign, IL.
Consequently, the shielding stress intensity factor 25. K. Han, M. Ciccotti, and S. Roux: EPL, 2010, vol. 89, p. 66003.
caused by the side-surface layer is simply 26. S.M. Wiederhorn, S.W. Freiman, E.R. Fuller, Jr., and C.J.
Simmons: J. Mater. Sci., 1982, vol. 17, pp. 3460–78.
27. S.M. Wiederhorn, E.R. Fuller, Jr., and R. Thomson: Met. Sci.,
ev E 1:465 b 1980, vol. 14, pp. 450–8.
Ksh ffi 1=3 pffiffiffi ½A5
1m a 28. R.H. Doremus: J. Mater. Res., 1995, vol. 10, pp. 2379–89.
29. S.M. Wiederhorn and L.H. Bolz: J. Am. Ceram. Soc., 1970, vol. 53
(10), pp. 543–48.
The method of determining the unconstrained volume 30. G.W. Weidmann and D.G. Holloway: Phys. Chem. Glasses, 1974,
expansion due to water penetration into the silica glass, vol. 15, pp. 68–75.
31. D.H. Roach, S. Lathabai, and B.R. Lawn: J. Am. Ceram. Soc.,
em, is discussed in Reference 21. For unstressed silica at 1988, vol. 71 (2), pp. 97–105.
361 K (88 C), em = 0.00325; for stressed glass at a crack 32. T. Fett and D. Munz: Stress Intensity Factors and Weight Func-
tip, em = 0.13. These values are calculated from direct tions, Computational Mechanics Publications, Southampton,
measurements of water concentration at the surface of U.K., 1997.

1174—VOLUME 44A, MARCH 2013 METALLURGICAL AND MATERIALS TRANSACTIONS A

You might also like