Journal of the European Ceramic Society 21 (2001) 561±567

www.elsevier.com/locate/jeurceramsoc

Fatigue limit in fused silica

Vincenzo M. Sglavo a,*, David J. Green b
a
Dipartimento di Ingegneria dei Materiali, UniversitaÁ di Trento, Via Mesiano 77, I-38050 Trento, Italy
b
Department of Materials Science and Engineering, The Pennsylvania State University, University Park, PA 16802, USA

Received 14 February 2000; received in revised form 27 July 2000; accepted 8 August 2000

Abstract
The threshold stress intensity factor for sub-critical crack growth in fused silica was measured using the interrupted static fatigue
(ISF) test and from observations of indentation crack growth under load. Good agreement was obtained between the two techni-
p p
ques with values of 0.31±0.34 MPa m and 0.300.02 MPa m, respectively. The relative advantages of the two experimental
approaches are discussed. # 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Crack growth; Fatigue; Glasses; Indentation; SiO2

1. Introduction minimum load below which sub-critical growth of

defects does not occur.3 18 This lower limit has been
Fused silica glass is an essential material in several quanti®ed in terms of a stress intensity factor, Kth,
modern technological applications. The optical ®bres called the threshold stress intensity factor or the fatigue
used in telecommunications or the dielectric present in limit, below which, the crack velocity is zero. Values of
many microelectronic devices are two such examples. In Kth have been proposed for several glasses in humid
these various applications, the silica is subjected to environments and for some ceramics.5 19 Unfortu-
mechanical loading that can be static or cyclic. These nately, in spite of the technological importance of silica
stresses can lead to the premature failure of the compo- glass, no fatigue limit data has been clearly identi®ed or
nent, especially when loading occurs in a corrosive measured for this material. Recently, Ritter et al. mea-
environment. Since even humid air represents a corro- sured crack arrest in DCDC specimens at about 0.4
p
sive environment for silica, this is a common situation. MPa m.15 Similar values were calculated by Cook on
The phenomenon which determines the premature fail- the basis of indentation±strength measurements.15
ure of a component under the contemporaneous action Michalske and Fuller20 and Lawn et al.21 have sug-
of load and of a corrosion process is known as fatigue gestedp that crack healing can occur around 0.10±0.13
and is common for most ceramics and glasses.1 3 The MPa m for silica glass in humid air and presumably
design of devices or structural components with such this occurs below the fatigue limit. Studies have also
materials is usually performed by the assumption of a been performed for the analysis of the fatigue behaviour
minimum lifetime which, in turn, determines the max- of silica glass, though some discrepancies between pro-
imum applied load. If higher loads are applied, failure posed fatigue laws are evident. Sub-critical crack
will occur in shorter times. An important aspect of this growth kinetics have been established for both bulk
process is that failure occurs in a subtle manner, in the fused silica3,5,8,15,16,22 28 and silica optical ®bres29 36 in
sense that the sub-critical growth of the critical defect humid environments. For example, Muraoka and AbeÂ36
only becomes evident in the last instants of the fracture have measured crack velocities in optical ®bres as low as
process, without any prior warning. 10 13 m/s without detecting a threshold. These authors
In some cases, the fatigue phenomenon has been showed their data ®tted an exponential law based on
shown to possess a lower bound corresponding to a chemical kinetics better than the ubiquitous empirical
power law. This is in agreement with earlier observa-
* Corresponding author. Tel.: +49-0461-882468; fax: +49-0461- tions by Michalske et al., who employed this idea to
881977. explain the discrepancies observed between the fatigue
0955-2219/01/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved.
PII: S0955-2219(00)00241-7
562 V.M. Sglavo, D.J. Green / Journal of the European Ceramic Society 21 (2001) 561±567

behaviour of bulk silica and high-strength silica

®bres.34
In recent papers, Sglavo and Green proposed two
di€erent methodologies for the determination of Kth in
glasses and ceramics.13,14,18,37 In the ®rst approach, the
fatigue limit is determined from the strength of samples
subjected to an interrupted static fatigue (ISF) test. In
the second, Kth is calculated from the length of median
cracks generated during the loading phase of Vickers
indentation tests. In the current study, these two tech-
niques are applied to fused silica samples with the aim
of the determination of Kth in a water environment. For
the indentation approach in the current study, the crack
geometry was distinctly di€erent from that observed in
other silicate glasses38,39 and this necessitated a mod-
i®cation of the data analysis. The results obtained from
the study are also used to compare the relative advan-
tages of the two approaches. Fig. 1. Schematic of the apparatus used for ISF tests.

2. Experimental procedure samples to be tested simultaneously. The dead weight

was applied such that the stress on the specimens during
Silica glass (99.7 wt% pure) in the form of bars and the constant stress phase of the ISF test was equal to 85,
rods was used in this work. Two di€erent procedures 94, 100 and 105 MPa. The load was maintained for
were used to determine the fatigue limit of silica glass in times ranging from 1 h to 50 days. Fifteen to 30 samples
a water environment. In the ®rst approach, specimens were tested at each load or time condition. The samples
were subjected to interrupted static fatigue in deionized that survived the ISF test were then broken in deionized
water and the strength of the surviving samples was water using a displacement rate of 20 mm/min.
compared to that measured in an inert environment. In
the alternative approach, Vickers indentations were 2.2. Indentation tests
introduced into glass surfaces in the presence of water.
The cracks so formed were then compared to those Silica plates with a nominal thickness of 3 mm were
produced in the inert environment. used for the indentation tests. Bars of 4±5 mm in width
were machined from the as-received sheets. The edges of
2.1. Interrupted static fatigue tests the bars were polished with diamond paste in order to
remove macroscopic defects. Vickers indentations were
Silica glass in the form of rods, nominal diameter of introduced on the surface of the bars. Care was taken to
2 mm and length of about 90 mm, were used. The place the indentation in the middle of the bar and to
strength was measured by four point bending tests both align one of the indenter diagonals perpendicular to the
in deionized water and in silicone oil. This latter liquid sample edges. The indenter movement was controlled
was considered to be the inert environment, preventing such that the maximum load was reached in about 10 s.
the access of water to the crack tip.13,17,39 In this latter Indentations were performed in both silicone oil and
case, samples were dried at 120 C for 2 h before being deionized water. In the former case, samples were dried
completely immersed and fractured in a silicone oil according to the procedure previously described and a
bath. Analogously, a deionized water bath was used for drop of oil was immediately placed on the surface of the
the tests as the active environment. Bending tests were bar at the location where the indentation was to be
conducted with inner and outer span equal to 40 and introduced. The indenter was then maintained on the
80 mm, respectively, using a displacement rate of surface of the sample for 15 s at the maximum load.
20 mm/min. For the strength testing, 30 specimens were Maximum loads ranging from 1.96 to 39.2 N were used.
broken in both environments. Indentations in water were obtained after placing a drop
Interrupted static fatigue (ISF) tests were conducted of the water on the surface of the bar. In this case, the
in deionized water using the apparatus shown schema- load was maintained for times ranging from 15 s to 8
tically in Fig. 1. Both the frame and the pulleys were days. For the longest times, additional water drops were
fabricated from stainless steel. The apparatus consisted added to the indentation site during the test. After the
of a series of frames like the one shown in Fig. 1. These removal of the indenter, samples were cleaned and bro-
frames were joined side by side in order to allow 15 ken manually using the indentation crack as the critical
 V.M. Sglavo, D.J. Green / Journal of the European Ceramic Society 21 (2001) 561±567 563
 
defect. The fracture surface was then observed using an 
KW ˆ KC …2†
optical microscope. S0

where  is pthe hold stress, KC is the fracture toughness

3. Results (0.75 MPa m)3 and S0 is the inert strength for sample
with the same failure probability. This calculation is
3.1. Interrupted static fatigue tests performed for increasing duration in the ISF test and
when KW becomes independent of time, it can be con-
The strength distributions of the glass rods measured sidered as the fatigue limit.13,14,37
in silicone oil and deionized water are reported in Fig. 2. The results of this computing procedure are shown in
The strength data are shown as a Weibull plot in which Fig. 4. The data corresponding to applied hold stress of
the failure probability, F, is de®ned as: 85 MPa were not considered in the present discussion
because only two, three and one specimens failed during
j the ISF tests for times of 10, 20 and 50 days, respec-
Fˆ …1†
N‡1 tively. For shorter times no specimens failed during the
stress hold. It is considered that such situations give rise
where j is the rank in the ascending ordered strength to large errors in the KW calculation.37 It is important to
distribution and N the total number of samples (30). point out that KW evaluation is theoretically una€ected
The Weibull modulus values in silicone oil and deio- by the particular choice of failure probability [Eq. (1)]
nized water are 6.01.2 and 6.91.3, respectively. since the characteristic inert strength depends only on
The strength of the specimens that survived the ISF the rank in the ordered distribution.37 Calculation
tests was also reported as a Weibull plot. Fig. 3 shows a
comparison of the strengths with and without the static
hold for one of the test conditions. For the construction
of the Weibull plot, the complete set of specimens initi-
ally subjected to the interrupted static fatigue test was
considered and a strength of zero was assigned to sam-
ples which fail during the hold period of the ISF
test.13,14,17 In a previous paper, Sglavo and Green per-
formed a theoretical analysis of the ISF test and pro-
posed a procedure for the determination of the fatigue
limit from the strength distribution of samples subjected
to interrupted static fatigue.14,37 The potential Kth is
evaluated as the stress intensity factor, KW, that is
applied at the start of the static hold to the weakest
specimen that survives this phase of the test and is given
by13,37
Fig. 3. Weibull plot of the strength measured in deionized water for
samples subjected to ISF test with duration of 5 days and applied
stress of 94 MPa. The weakest survivor is highlighted by a circle. The
initial strength distribution (solid symbols) is shown for comparison.

Fig. 2. Weibull plot of the strength measured in deionized water and

silicone oil. Fig. 4. KW data as function of dwell time for the constant stress.
564 V.M. Sglavo, D.J. Green / Journal of the European Ceramic Society 21 (2001) 561±567

performed by using F=(j 0.5)/N furnished results shown in Fig. 8. Similar values of K are obtained at the
equal to those shown in Figs. 2±4, except for the Wei- same testing time for the two indentation loads. This
bull moduli, which gave results equal to 7.61.6 and represents a further con®rmation of Eq. (3). For time
7.71.6 in water and silicone oil, respectively. Typical durations longer than 100 ks an invariant stress intensity
uncertainty in K is 5% and this was used to calculate factor is reached. This allows one to de®ne a fatigue limit
p
the error bars shown in Fig. 4. The KW data in Fig. 4, of approximately 0.29±0.31 MPa m for the fused silica
that are threshold estimates, clearly approach a limiting glass used in this work. More precisely, if the complete set
p
value in the range of 0.31±0.34 MPa m. of crack lengths recorded at t52 days is considered, the
threshold stress intensity factor is determined to be
3.2. Indentation tests

The sub-surface pattern of deformation and fracture

produced by Vickers indentation in fused silica is shown
in Fig. 5. The main crack is conical and it initiates out-
side the plastic zone. This observation is in agreement
with previous studies in which silica was de®ned as an
``anomalous'' glass.38,40 For these glasses, plastic defor-
mation is associated with densi®cation processes occur-
ring within the plastic zone, which ``blunts'' the contact.
This behaviour promotes formation of cone cracks
instead of the median±radial cracks normally associated
with sharp indentations.
The size of cone crack formed in water increased with
the duration of the indentation test. In order to deter-
mine the possibility of a fatigue limit, the growth of the
cone cracks was determined as a function of indentation
time. Fig. 6 shows the length of the cone crack, l, as a Fig. 5. Typical cone crack obtained by Vickers indentation in silica
function of the duration of the maximum indentation glass (P=9.8 N, t=120 s).
load. For both indentation loads, 9.8 and 39.2 N, the
crack length increases as testing time increases at least
up to 100 ks. For longer indentation times crack arrest
appears as the fatigue limit is reached, most noticeably
for the indentation force of 39.2 N. For quantitative
evaluation of the fatigue limit, the stress intensity factor
applied to the cone crack during the indentation test
must be analytically de®ned. Fig. 7 shows the plot of log
P vs log l data determined from indentations performed
in silicone oil. A straight line with slope equal to 1.5 ®ts
the experimental data. Therefore, the stress intensity
factor applied to the cone crack can be de®ned as:

P
Kˆ …3†
l 1:5

where the factor  can be evaluated by linear regression

of the data in Fig. 7 and the fracture toughness of silica
glass. The value obtained by this procedure is =0.046.
The relation expressed by Eq. (3) is similar to the stress
intensity factor proposed for Hertzian cone cracks and
median±radial cracks produced by Vickers indentation.3
The factor  de®nes the intensity of the stress ®eld
around the indentation site upon loading and, therefore,
can be considered as invariant as the crack length
increases during a constant load test. On the basis of Fig. 6. Cone crack length as a function of loading time for indentation
Eq. (3), the stress intensity factor applied to the cone loads of 9.8 and 39.2 N. Solid symbols represent crack length obtained
cracks reported in Fig. 6 was evaluated. The results are in silicone oil.
 V.M. Sglavo, D.J. Green / Journal of the European Ceramic Society 21 (2001) 561±567 565

con®dence higher than 80% that crack arrest has been

detected during indentation tests performed with dura-
tion equal to or in excess of 691.2 ks.
Additional information on the fatigue behaviour of
the glass can be gathered from the data in Fig. 6. Crack
length values recorded at times less than 100 ks corre-
spond to the sub-critical growth of cracks under a spe-
ci®c stress ®eld. The sub-critical crack velocity can be
expressed as:
 n
dl K
ˆ v0 …4†
dt KC

where v0 and n are parameters which depend on the

material and the environment. Eq. (4) can be combined
with Eq. (3) to obtain a relation between the crack
length and indentation time, t, as:39
 1=h
t
l  l0 hv0 …5†
l0
Fig. 7. Plot of log P as a function of log l from indentation tests per-
formed in silicone oil.
where l0 is the cone length in silicone oil and h=1.5 n
+1. Eq. (5) can be easily transformed into a linear
relation:
"   #
hv0 1=h 1
log l ˆ log l0 ‡ log t: …6†
l0 h

The experimental data shown in Fig. 6 for t<100 ks

were expressed in terms of log l vs log t and linear
regression was used to determine n and v0. Values equal
to 19.72.9 and 3.91.4 mm/s were obtained for n and
v0, respectively, from crack length data obtained with
P=39.2 N. Correspondingly, values equal to 22.03.4
and 26.912.6 mm/s were calculated for P=9.8 N.
It is worth comparing the fatigue parameters obtained
here with previous results. Two distinct ranges of n data
Fig. 8. Stress intensity factor applied to indentation cone cracks as a
can be found in the literature for silica glass. Values of n
function of time at maximum load.
ranging from 14 to 26 have been obtained by static or
p dynamic fatigue tests on silica ®bres in humid environ-
0.300.02 MPa m. In order to point out the validity of ments.29,31,32,34,36 Conversely, direct measurements of
the results obtained by indentation, a statistical test was the large crack velocities in bulk silica specimens furn-
performed on K data gathered at times equal to 50.4 ks ished n values between 36 and 44. Dabbs and Lawn
(14 h), 172.8 ks (2 days) and 691.2 ks (8 days). The aim determined values of n equal to 25 and 40 by dynamic
was to demonstrate that K data corresponding to the fatigue tests on silica ®bres in which sub-threshold and
two highest duration represent the same number and are post-threshold indentation cracks, respectively, had
di€erent from K results obtained at t=14 h. Typical been produced.32 The e€ect of the indentation residual
statistical approaches based on hypothesis testing were stress ®eld, which is much stronger for sub-threshold
used.41 For both indentation loads, K values at t=172.8 cracks, was used to account for the reduction of n from
ks andp 691.2 ks represent the same number (0.30 40 to 25. An alternative explanation for the di€erent
MPa m) with a probability of a Type II p error lower values of n was advanced by Muraoka and AbeÂ.36 These
than 20% for a di€erence of 0.03 MPa m from the researchers measured the sub-critical growth of inden-
average. Conversely, values calculated at t=50.4 ks are tation cracks in silica ®bres by direct observation and
larger than K data at t=691.2 ks at a 0.5% level of sig- calculated values of n around 23 and 40 for K intervals
p
ni®cance (error Type I). Therefore, one can claim with a 0.3 to 0.45 and 0.3 to 0.6 MPa m, respectively.36 This
566 V.M. Sglavo, D.J. Green / Journal of the European Ceramic Society 21 (2001) 561±567

proposed dependence of n on the sub-critical crack tical growth of the cracks can be directly observed.
growth velocity or on K is in agreement with previous Unfortunately, this technique can be applied only to
results by Michalske et al.34 Similarly, Cook has pro- materials in which indentation cracks can be easily
posed that the values of n may be lower when the lower observed. The interrupted static fatigue test presents the
part of the v(K) curve is explored.40 On this basis, one advantage of being applicable to any material and at
would expect the n values in the indentation tests to be high temperature. The diculty is that it can involve
in the lower range since the measurements were made at testing a large number of specimens.
low crack velocities, in agreement with the current data
for the n values.
The calculated values for v0 reported earlier are dif- 5. Conclusions
ferent for the two indentation loads. This should not be
too surprising. In most papers dealing with fatigue, v0 Two techniques were used to determine whether a
values are not reported and the fatigue behaviour is fatigue limit occurs in fused silica. In the ®rst approach
described only by the n parameter. The calculation of v0 the interrupted static fatigue test was used. In this tech-
uses an evaluation of a power (ex or 10x) where small nique, determining the strength of the weakest sample
di€erences in the argument can develop into large var- that survives the constant stress phase of the test iden-
iations in the ®nal result. The di€erence in the v0 values ti®es the threshold. A p value for the threshold in the
is mainly a re¯ection of the di€erence in the n values. range 0.31±0.34 MPa m was obtained. In the second
approach, the growth of cone cracks under indentation
load was used to determine the sub-critical crack growth
4. Discussion parameters and the threshold value. For this latter
p
technique, a threshold value of 0.300.02 MPa m was
There was good agreement between the two techni- obtained. Of the two techniques, the indentation
ques used for the determination of the fatigue limit of approach involved less experimental scatter, fewer spe-
silica glass in the water environment. Interrupted static cimens and direct observation of crack growth. Unfor-
fatigue
p tests furnished values of Kth around 0.31±0.34
p tunately, this technique, cannot be readily applied to all
MPa m while a value of 0.300.02 MPa m was materials, especially at high temperature and it was
obtained by the indentation procedure. At this point, it found to require longer testing times.
is useful to discuss the results shown in Figs. 4 and 8 in
more detail.
The ISF tests were performed using quite long times Acknowledgements
(up to 50 days) and a clear trendp(Fig. 4) in KW towards
a range from 0.31 to 0.34 MPa m was identi®ed. For Dr Edi Degasperi and Miss Alexia Conci are
hold stresses 94 MPa or greater, the threshold is reached acknowledged for their experimental work. NATO is
at t100 ks. The choice for the constant stress in the acknowledged for the ®nancial support (Grant number
ISF test plays an important role in this test. For the CRG 950160).
smallest sampling errors, a similar number of samples
need to fail in the stress hold compared to those that
survive and this only occurs for the higher hold stresses. References
The main disadvantage of the ISF test is the large 1. Davidge, R. W., Mechanical Behaviour of Ceramics. Cambridge
number of specimens that are needed, especially when University Press, Cambridge, UK, 1979.
there is large strength variability. The theoretical analy- 2. Green, D. J., Introduction to Mechanical Properties of Ceramics.
Cambridge University Press, Cambridge, UK, 1998.
sis of Sglavo and Green37 showed that a hold time of 20
3. Lawn, B. R., Fracture of Brittle Solids, 2nd ed. Cambridge Uni-
ks would be necessary for a specimen (of average versity Press, Cambridge, UK, 1993.
strength) to fail without the presence of a threshold 4. Charles, R. J. and Hillig, W. B., Symposium on Mechanical
p
when subjected to K=0.30 MPa m at the start of the Strength of Glass and Ways of Improving It. Union Scienti®que
hold period. This suggests that the hold times used in Continentale du Verre, Charleroi, Belgium, 1962, pp. 511±527.
the ISF were long enough to detect the threshold if this 5. Wiederhorn, S. M. and Bolz, L. H., Stress corrosion and static
fatigue of glass. J. Am. Ceram. Soc., 1970, 53(10), 543±548.
were its value. 6. Evans, A. G., A method for evaluating the time-dependent failure
The indentation data clearly has the advantage of characteristics of brittle materials Ð and its application to poly-
involving less experimental scatter and involves using a crystalline alumina. J. Mater. Sci., 1972, 7, 1137±1146.
small number of specimens. It is, however, clear from 7. Evans, A. G., A simple method for evaluating slow crack growth
Fig. 4 that testing times in excess of a 105±106 s are in brittle materials. Int. J. Fract., 1973, 9(3), 267±275.
8. Wiederhorn, S. M., Subcritical crack growth in ceramics. In
needed for fused silica. The indentation method is a Fracture Mechanics of Ceramics, Vol. 2, ed. R. C. Bradt, D. P. H.
simple technique whose results are easily analysed and Hasselman and F. F. Lange. Plenum Press, New York, 1974, pp.
give less experimental scatter. In addition, the sub-cri- 613±646.
 V.M. Sglavo, D.J. Green / Journal of the European Ceramic Society 21 (2001) 561±567 567

9. Michalske, T. A., The stress corrosion limit: its measurement and 26. Michalske, T. A. and Bunker, B. C., A chemical kinetics model
application. In Fracture Mechanics of Ceramics, Vol. 5, ed R. C. for glass fracture. J. Am. Ceram. Soc., 1993, 76(10), 2613±
et al. Bradt Plenum Press, New York, 1983, pp. 277±289. 2618.
10. Cook, R. F., Crack propagation thresholds: a measure of surface 27. Li, H. and Tomozawa, M., Mechanical strength increase of
energy. J. Mater. Sci., 1986, 1, 852±856. abraded silica glass by high pressure water vapor treatment. J.
11. Wan, K., Lathabai, S. and Lawn, B. R., Crack velocity functions Non-Cryst. Solids, 1994, 168, 287±292.
and threshold in brittle solids. J. Eur. Ceram. Soc., 1990, 6, 259± 28. Tomozawa, M., Stress corrosion reaction of silica glass and
268. water. Phys. Chem. Glasses, 1998, 39(2), 65±69.
12. Cook, R. F. and Liniger, E. G., Kinetics of indentation cracking 29. Ritter, J. E. Jr, Sullivan, J. M. Jr and Jakus, K., Application of
in glass. J. Am. Ceram. Soc., 1993, 76, 1096±1106. fracture mechanics theory to fatigue failure of optical glass ®bers.
13. Sglavo, V. M. and Green, D. J., Threshold stress intensity factor J. Appl. Phys., 1978, 49(9), 4779±4782.
for soda lime silicate glass by interrupted static fatigue test. J. 30. Sakaguchi, S. and Kimura, T., In¯uence of temperature and
Eur. Ceram. Soc., 1996, 16, 645±651. humidity on dynamic fatigue of optical ®bers. J. Am. Ceram.
14. Sglavo, V. M., Green, D. J., Martz, S. W. and Tressler, R. E., In Soc., 1981, 61(5), 259±262.
Fracture Mechanics of Ceramics, vol. 12, ed. R. C. Bradt et al. 31. Mitsunaga, Y., Katsuyama, Y., Kobayashi, H. and Ishida, Y.,
Plenum Press, New York, 1996, pp. 167±177 Failure prediction for long length optical ®ber based on proof
15. Ritter, J. E., Fox, J. R., Hutko, D. I. and Lardner, T. J., Moist- testing. J. Appl. Phys., 1982, 53(7), 4847±4853.
ure-assisted crack growth at epoxy-glass interfaces. J. Mater. Sci., 32. Dabbs, T. P. and Lawn, B. R., Strength and fatigue properties of
1998, 33, 4581±4588. optical ®bers containing microindentation ¯aws. J. Am. Ceram.
16. Cook, R. F., Thermally-Activated Cracking and Failure of Fused Soc., 1985, 68(11), 563±569.
Silica, Paper SXVII-029-98 presented at the 100th Annual Meet- 33. Mathewson, M. J. and Kurkjian, C. R., Environmental e€ects on
ing of the American Ceramic Society, Cincinnati, OH, USA, Mat the static fatigue of silica optical ®ber. J. Am. Ceram. Soc., 1988,
3-6, 1998. 71(3), 177±183.
17. Sglavo, V. M. and Renzi, S., Fatigue limit in borosilicate glass by 34. Michalske, T. A., Smith, W. L. and Bunker, B. C., Fatigue
interrupted static fatigue test. Phys. Chem. Glasses, 1999, 40(2), mechanisms in high-strength silica-glass ®bers. J. Am. Ceram.
79±84. Soc., 1991, 74(8), 1993±1996.
18. Sglavo, V. M. and Green, D. J., Indentation determination of 35. Ostojic, P., Review. Stress enhanced environmental corrosion and
fatigue limits in silicate glasses. J. Am. Ceram. Soc., 1999, 82(5), lifetime prediction modelling in silica optical ®bres. J. Mater.
1269±1274. Sci., 1995, 30, 3011±3023.
19. Hayashi, K., Easler, T. E. and Bradt, R. C., A fracture statistics 36. Muraoka, M. and AbeÂ, H., Subcritical crack growth in silica
estimate of the fatigue limit of a borosilicate glass. J. Eur. Ceram. optical ®bers in a wide range of velocities. J. Am. Ceram. Soc.,
Soc., 1993, 12, 487±491. 1996, 64(5), 51±57.
20. Michalske, T. A. and Fuller, E. R. Jr, Closure and repropagation 37. Sglavo, V. M. and Green, D. J., The interrupted static fatigue test
of healed cracks in silicate glass. J. Am. Ceram. Soc., 1985, for evaluating threshold stress intensity factor in ceramic materi-
68(11), 586±590. als: a numerical analysis. J. Eur. Ceram. Soc., 1995, 15, 777±785.
21. Lawn, B. R., Roach, D. H. and Thompson, R. M., Threshold 38. Cook, R. F. and Pharr, G. M., Direct observation and analysis of
and reversibility in brittle cracks: an atomistic surface force indentation cracking in glasses and ceramics. J. Am. Ceram. Soc.,
model. J. Mater. Sci., 1987, 22, 4036±4050. 1990, 73, 787±817.
22. Ito, S. and Tomozawa, M., Crack blunting of high-silica glass. J. 39. Sglavo, V. M. and Green, D. J., Subcritical growth of indentation
Am. Ceram. Soc., 1982, 65(8), 368±371. median cracks in soda-lime-silica glass. J. Am. Ceram. Soc., 1995,
23. Sakaguchi, S., Sawaki, Y., Abe, Y. and Kawasaki, T., Delayed 78(3), 650±656.
failure in silica glass. J. Mater. Sci., 1982, 17, 2878±2886. 40. Arora, A., Marshall, D. B. and Lawn, B. R., Indentation defor-
24. Hibino, Y., Sakaguchi, S. and Tajima, Y., Crack growth in silica mation/fracture of normal and anomalous glasses. J. Non-Cryst.
glass under dynamic loading. J. Am. Ceram. Soc., 1984, 67(1), Solids, 1979, 31, 415±428.
64±68. 41. Miller, I. and Freund, J. E., Probability and Statistics for Engi-
25. Tomozawa, M. and Hirao, K., Mechanical fatigue of silica glass. neers, 3rd edn. Prentice-Hall, Englewood Cli€s, NJ, 1985, pp.
J. Non-Cryst. Solids, 1987, 95&96, 149±160. 185±228.