Professional Documents
Culture Documents
JACS Borosilicate
JACS Borosilicate
Journal
Characterization of Confined Intact and Damaged Borosilicate Glass
Sidney Chocronw, Charles E. Anderson Jr., Arthur E. Nicholls, and Kathryn A. Dannemann
Engineering Dynamics Department, Southwest Research Institute, P.O. Drawer 28510, San Antonio, Texas 78228-0510
This article describes two different techniques utilized to char- Dannemann et al.13 In general, the above references confine
acterize intact and damaged borosilicate glass at pressures up to the specimens at low to moderate pressures (100–300 MPa). In
B2 GPa: triaxial compression and confined sleeve. The results the present work, the pressures achieved—on the order of
of the characterization experiments—for intact and damaged 2 GPa—are significantly higher. We also performed a limited
glass as a function of confinement pressure—are described; the number of tests at high strain rates.
results are interpreted in terms of two pressure-dependent con- Triaxial compression and confined sleeve techniques are com-
stitutive models: Drucker–Prager and Mohr–Coulomb (MC). plementary because one explores lower pressures than the other.
The MC model is successful at predicting the damage pattern. The fact that they overlap at confining pressures of 300–400
An observation is that the slopes of the two models appear to be MPa increases the confidence in the interpretation of the con-
independent of the degree of damage (intact, predamaged, and fined sleeve technique. The high pressures achieved in the tests
severely damaged specimens). Lastly, these data are compared also can be compared with results obtained using flyer-plate
with flyer-plate impact data of intact and damaged glass. impact tests, which interrogate material response at very high
strain rates. This will be done near the end of the article.
I. Introduction
II. Experimental Technique
1
Fluid at Pressure P 0
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0
Axial Strain (%)
Plastic sleeve
Alumina anvils Fig. 2. Two intact specimens (BF-61 and BF-63) and two predamaged
specimens (BF-49 and BF-53) tested in the bomb at 250 and 400 MPa
Fig. 1. Schematic of the hydraulic bomb experiment. nominal confinement pressures.
3392 Journal of the American Ceramic Society—Chocron et al. Vol. 93, No. 10
2000
Hoop Strain (%)
0.0 0.1 0.2 0.3 0.4
3500
3000
1500
Axial Stress (MPa)
2500
1500
1000
1000
Axial Strain
500 Hoop Strain
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
500
Axial Strain (%)
Fig. 3. Stress versus axial and hoop strain for a predamage specimen in
a confined sleeve test under monotonic load (test BF-17).
0
or ‘‘failure avalanches’’2 occur (the maximum load attained in a
2500
2000
1500
1000
500
0
test is controlled by the operator, not specimen strength). The
jumps probably indicate the formation of a shear plane that is
suddenly propagated. The planes slide against each other, but Equivalent Stress (MPa)
after a few microns of motion, the propagation is stopped by the
sleeve. Slip planes are apparent in the posttest analysis of the
specimens as described in Dannemann et al.13 The hoop strain in
2000
the sleeve is very sensitive to specimen motion; therefore, when
failure occurs in the specimen and two planes slide, the relative
change in the hoop strain is large, as shown in Fig. 3.
Confined sleeve experiments were conducted on 18 predam-
aged specimens that were subjected to a maximum of 10 load
1500
cycles. An example with nine load cycles is shown in Fig. 4(a),
1000
Esl b2 a2
~r ¼
s ey (2)
2 a2
500
where Esl is the elastic modulus of the sleeve, a and b are the
internal and external radii, respectively, and ey is the hoop strain
of the sleeve. The elastic constants of the specimens can also be
determined but a more elaborate analysis is required; this was
done for the predamaged specimens in Chocron et al.15 showing
that if the material is well confined, a severely cracked specimen
0
2500
2000
1500
1000
500
has elastic constants that are the same, within measurement un-
certainties, as an intact specimen.
The interpretation of the jumps observed in Fig. 4(a) and how Equivalent Stress (MPa)
these jumps are ‘‘translated’’ to a constitutive model is also shown
in Fig. 4. A jump is a sudden discontinuity in the pressure applied Fig. 4. Interpretation of the sleeve tests with predamaged specimens.
to the specimen, probably due to the creation and slippage of an (a) Test BF-21, (b) Jumps recorded for all the tests performed.
internal shear plane. Jumps occur while the applied axial load is
increasing. It is thought that jumps provide fundamental infor-
mation regarding how the specimen fails at different confinement flow phenomena in glasses can be explained as densification and
pressures. Consequently, all the jumps recorded in each of the argues that some glasses show properties compatible with classic
tests—four jumps in the one shown in Fig. 4(a) for test BF-21— plasticity. Other arguments for plastic flow can be found in
are placed on an equivalent stress versus pressure graph. The Ernsberger,19 and Lankford et al.20 Here, whenever DP or MC
equivalent stress is computed from Eq. (1) once the radial stress is models are discussed here, it is understood that they represent a
calculated from the hoop strain gage. The stress–pressure jumps surface that, when used in numerical simulations, are used to
recorded in all the experiments conducted on predamaged speci- limit material strength. For intact material, this might be
mens in the confined-sleeve tests are shown in Fig. 4(b). thought of as a failure surface; for predamaged material, it
might be a flow surface. We use the term ‘‘failure’’ to signify a
change in the load-carrying capability of the material. We do
IV. Constitutive Model Parameters not provide a prescription for how the strength might change as
a function of damage, as done by others, e.g., Holmquist and
(1) Constitutive Models colleagues.21,22 Rather, the data will be analyzed to determine
It is known that some glasses or ceramics under confinement can the constants that describe the limits in the load-carrying capa-
flow plastically. For example, Peter18 mentions that not all the bility of intact glass and damaged glass.
October 2010 Characterization of Borosilicate Glass 3393
4000 aged specimens are also included in the plot. A linear regression
fit performed using only the confined data is shown in Fig. 5.
Equivalent Stress (MPa) The unconfined data were not included in the linear fit as a very
3000 Y = 1588 + 1.20P small confinement pressure, like the one used for test BF-57 (25
MPa), significantly increases the strength of the specimen from
Y = 423 + 1.22P approximately 150–600 MPa.
2000 The residual strength constants for the hydraulic bomb tests
Y = 140 + 1.3P
were obtained from the same tests that were used to estimate the
predamaged constants, but using the sawtooth portions of the
1000 response. It was decided to select three ‘‘characteristic’’ points
along the curve: a local maximum, a local minimum, and an
BF-57 average. These points are plotted as seq vs P in Fig. 5. The
0
scatter is very large due to the oscillatory nature of the test re-
0 500 1000 1500 2000 sults. The linear regression performed did not force the intercept
Hydrostatic Pressure (MPa) to the (0,0) point, although it would be reasonable to think that
at zero pressure, these specimens have a very small strength.
Fig. 5. Equivalent stress at failure for specimens tested in the hydraulic (B) MC Analysis: Predamaged specimens tested in the
bomb.
bomb systematically showed a shear plane at an angle between
551 and 701; see Fig. 6. The angle appeared to be independent of
The DP model23 has the following form: the confinement pressure applied to the specimen. A character-
istic angle can arise only if the flow surface on the p-plane is not
Y0 þ bP P < Ycap Y0 =b a circle but a polygon, i.e., the third invariant J3 enters in the
Y¼ (3) flow surface equation. The MC model includes J3, and a char-
Ycap P Ycap Y0 =b
acteristic failure angle independent of the confinement pressure
where Y0 is the zero-pressure strength, b is the slope of Y vs P, P arises naturally from the model.24
is the hydrostatic pressure (negative of the mean stress), and The same intact data shown in Fig. 5 are now analyzed and
Ycap is the limiting flow stress. plotted from an MC perspective in Fig. 7. The MC model gives
Predamaged specimens tested in the bomb systematically the maximum shear stress t a solid can support on any plane,
showed a shear plane at an angle between 551 and 701. The an- Eq. (4). A linear least-squares regressiony in the (s1s3)/2 vs
gle seems to be independent of the applied confinement pressure. (s1+s3)/2 representation was performed to the intact test data
The DP model is based on the first invariant of the stress tensor, to obtain an intercept of a 5 651 MPa and the slope b 5 0.506.
I1, and the second invariant of the stress deviator tensor, J2. The The relation between these curve-fit constants and MC param-
flow surface in the p-plane is a circle, and thus, the DP model eters, Eq. (4), is given by
can never have a characteristic failure angle. Incorporating the
third invariant J3 into the description of failure results in the ðfÞ ¼ arcsinðbÞ m ¼ tanðfÞ c ¼ a=cosðfÞ (5)
flow surface on the p-plane being a polygon, which then has a
characteristic failure angle. The MC model incorporates J3, and and so for the intact specimens, an MC model can be written as
has a characteristic failure angle independent of the confinement t ¼ 0:755 þ 0:587~sn ðGPaÞ. Similarly, linear regressions were
pressure24; thus, it was felt that the MC model could be an ap- conducted on the predamaged response and the sawtooth
propriate candidate for describing the response of glass. portions of Fig. 3. The fits are shown in Fig. 7; using Eq. (5),
The data from the hydraulic bomb and confined sleeve tests the MC constants for predamaged specimens is t ¼ 0:201þ
were reanalyzed from the perspective of an MC model. The MC 0:594~sn ðGPaÞ, and that for the residual damaged specimens
model gives the maximum shear stress, t, that the glass can is t ¼ 0:063 þ 0:66~sn ðGPaÞ.
support on any plane:
(3) Confined Sleeve Test Results
t ¼ c þ m~
sn (4) Sleeve test results were presented in Chocron et al.16 However,
because more experimental data are now available, resulting
where c is the cohesion, mtan(f) is the friction coefficient (f is in a refinement of the previous analyses, the data will be briefly
~ n is the normal stress (positive in com-
the friction angle), and s revisited.
pression). (A) DP Analysis: As described in Section IV(2), all the
jumps recorded in the confined sleeve tests were transferred to
(2) Hydraulic Bomb Tests Results an equivalent stress versus pressure plot. Results for intact and
predamaged specimens confined in the sleeve are shown in
(A) DP Analysis: Failure data obtained from tests like Fig. 8. The unconfined specimens are also included in the graph.
the ones shown in Fig. 2 were plotted in an equivalent stress The scatter in results for the intact and predamaged confined
versus hydrostatic pressure graph, as shown in Fig. 5. The open sleeve tests is much larger than in the hydraulic bomb, and
triangles indicate intact unconfined tests, i.e., uniaxial stress so the linear fits through the data have smaller correlation
tests, for which the load path is a straight line with a slope of coefficients.
3. For unconfined specimens, the scatter in equivalent stress is The intact data should be used with care because testing in-
very large, ranging roughly from 1.4 to 2.6 GPa. This scatter is tact specimens inside the sleeve was very challenging. Even
inherent to brittle materials like glass, but the scatter seems to though the sleeve was honed to fit the specimens, any small
decrease when the specimen is confined. The results of a linear gap, misalignment, or eccentricity affects the results of the intact
regression for the intact confined specimens (solid triangles) are specimens significantly. Three of the four confined sleeve exper-
shown in Fig. 5. iments with intact specimens fail at a seq considerably less than
Predamaged specimens show a very different behavior, be- the intact specimens in the hydraulic bomb. Nevertheless,
cause after failing, the specimen is still able to carry load,z as the slope of the seqP response is essentially the same. The
shown in Fig. 2. The peak load of the 16 tests performed is resultant regression analyses for the intact and predamaged
designated with the solid inverted triangles in Fig. 5. Three un- responses are shown in Fig. 8.
confined tests (open inverted triangles) performed on predam- It can also be observed that, for the predamaged material, the
z
Technically, the specimens that were initially intact and then failed could also carry a strength (equivalent stress) appears to have reached a plateau at
load, but failure is so catastrophic that the plastic sleeve is ripped apart and the comminuted
y
specimen is dispersed in the hydraulic fluid. Regression fits are done in the form of y 5 a1bx.
3394 Journal of the American Ceramic Society—Chocron et al. Vol. 93, No. 10
Fig. 6. Two predamaged specimens that were tested in the bomb at 250 and 400 MPa confinement pressures. Shear angles of approximately 551–701
were very similar for the entire range of confinement pressures, from 25 to 400 MPa.
2500 1600
Confined Intact 1400
2000 Unconfined Intact
Predamaged Bomb 1200
(1–3)/2 (MPa)
(1–3) /2 (MPa)
Predamaged Unconfined
1500 Residual Bomb 1000
y = 651 + 0.506x
800
1000 y = 172.7 + 0.511x
600
y = 52.8 + 0.552x 400 y = 210 + 0.520x
500
200
0
0 500 1000 1500 2000 2500 0
0 500 1000 1500 2000
(1+3)/2 (MPa)
(1+3) /2 (MPa)
Fig. 7. Mohr–Coulomb failure points for specimens tested in the
hydraulic bomb. s1 and s3 are the maximum and the minimum prin- Fig. 9. Mohr–Coulomb constitutive models inferred for intact and
cipal stresses, respectively. predamaged specimens, confined sleeve tests.
Y = 1280 + 1.10P results of the confined sleeve experimental procedures and in-
3000
terpretation; the advantage of the confined sleeve test is that
higher confinement pressures can be achieved.
2000
Table II. Drucker–Prager Parameters for Intact and
Predamaged Borosilicate Glass (Y 5 Y01bP)
1000
Y = 535 + 1.23P Bomb tests Sleeve tests
Table III. Mohr–Coulomb Parameters for Intact and duce the later time penetration velocities.28 This suggests that
Predamaged Specimens ðs ¼ c þ l~
rn Þ details of the transition of intact to damaged glass are important
at lower impact velocities, and that a more comprehensive glass
Bomb tests Sleeve tests
model (intact, damage initiation, damage propagation) is re-
quired in order to model projectile penetration over the full
Specimen c (GPa) m c (GPa) m
range of impact velocities.
Intact 0.755 0.587 0.583 0.553
Predamaged 0.201 0.594 0.246 0.609
Residual 0.063 0.66 — — V. Characterization Data and Flyer-Plate Impact Data
(1) Introduction
(5) High Strain Rate Effects The characterization experiments described in the previous sec-
Data on strain-rate effects on damaged glass or damaged brittle tions were performed at quasi-static strain rates (except for the
materials are scarce in the literature. It was demonstrated indi- limited strain-rate investigation discussed in Section IV(5)). It is
rectly that there was no strengthening of the damaged glass at interesting to compare the results from these quasi-static exper-
high strain rates.26 A limited number of tests were conducted on iments with flyer-plate impact experiments, where strain rates
predamaged borosilicate glass, confined in the steel sleeve, at are on the order of 105 s1. Borosilicate and soda–lime glass
1 and 1000 s1. Because it was difficult to determine when failure have been the object of extensive plate impact characterization
occurs, particularly for the split-Hopkinson bar experiments be- since the pioneering work by Cagnoux,30 Rasorenov et al.,31 and
cause of the oscillations inherent in a dynamic test, it was decided Rosenberg et al.32 Of particular interest for our discussion are
to simulate the experiments using the constitutive model devel- the papers where the strength of both intact and damaged glass
oped from static tests. The simulations were shown to reproduce under confinement is presented. Brar et al.,33 as early as in 1991,
numerically the waves observed during the tests in the split-Hop- estimated 2 GPa as the strength of damaged soda–lime glass
kinson pressure bar. Further evidence for the lack of strain rate at pressures of 4–6 GPa. Bourne et al.34,35 showed tests with
strengthening for damaged glass will be discussed in Section V. strength values of 1.8 GPa for damaged borosilicate and
soda–lime for pressures from 4 to 8 GPa. Recently, Alexander
(6) Numerical Simulations of Impact and Penetration et al.,36 presented hugoniot elastic limit (HEL) and equation of
The results of the characterization experiments were used to state data for borosilicate and soda–lime glass. The flyer-plate
simulate long gold rods impacting borosilicate glass cylinders impact data considerably extends the confining pressures that
over a range of impact velocities (0.8–2.8 km/s),27 based on ex- can be achieved using the hydraulic bomb and confining sleeve
periments in Behner et al.28 It was assumed that the projectile techniques. Therefore, we will summarize the results of Bourne
penetrated failed material; thus, details of the transition of intact and colleagues and Alexander and colleagues and then provide
glass to failed glass were avoided. The rationale for this ap- comparisons between these data and our data.
proach was that the failure front propagates much more rapidly
than the projectile penetrates28; thus, the projectile penetrates
failed material. (2) Plate-Impact Experiments
Previous work29 demonstrated that the cap controls the pen- The flyer-plate data from Bourne et al.34,35 and the HEL from
etration velocity at high impact velocities, therefore, parametric Alexander et al.36 are shown in Fig. 10. In this figure, the black
simulations were conducted to investigate the dependence of the squares denote soda–lime glass, the white triangles denote boro-
penetration velocity on the cap (Ycap and tcap). The parametric silicate
s
(Pyrex) glass, and the circle shows borosilicate (Boro-
studies showed that the penetration velocity of an Au rod into float 33) glass. Some explanation of the Bourne and colleagues,
borosilicate glass is relatively insensitive to quite large variations data is required.z Bourne and colleagues used lateral and lon-
in the value of the cap. An B40% increase in the cap resulted in gitudinal gauges to measure the stresses in the glass specimens.
only an B3% decrease in the penetration velocity. Nevertheless, As they increased the impact velocity, the lateral stress showed a
the value deduced for the cap from numerical simulations was in second increase, which is interpreted as the arrival of a failure
agreement with the experimental characterization results for the front,34,35 which lowers the strength. The arrival of the second
damaged glass (Figs. 8 and 9). increase in the lateral stress corresponds to the arrival of a lon-
Parametric studies were required to deduce the zero-pressure gitudinal recompression wave observed in a VISAR signal in the
strength (Y0) and the cohesion (c) for the DP and MC models to experiments by Rasorenov et al.31
reproduce the penetration velocities of the gold rod at lower The equivalent stress, for uniaxial strain conditions, is
impact velocities (npo1.5 km/s). The values deduced from the seq 5 sxsy, where sx is the longitudinal stress and sy is
simulations were significantly lower than those obtained from the lateral stress. The pressure (P), for these uniaxial strain
the characterization experiments.27 It was concluded that the experiments, is given byJ (sx12sy)/3, and the lateral stress is
material beneath the penetrator is more highly damaged (co- related to the longitudinal stress through Poisson’s ratio, n, by
mminuted) than the damaged glass characterized in the labora- sy 5 nsx/(1n). Thus, seq 5 (12n)sx/(1n).
tory experiments. This highly comminuted region corresponds Bourne et al.35 use the lowest value of sx (the longitudinal
to the Mescall zone, as described by Shockey et al.1 The DP and stress) where there is a jump in the lateral stress gage (the arrival
MC constitutive constants for failed borosilicate glass that of the failure wave) to obtain an estimate of the HEL.ww The
matched the position-time data are as follows: long dashed horizontal line through the damaged material in
Fig. 10 is seq 5 1.85 GPa, an average of the strength of the
0:038 þ 1:2P P 1:72 GPa damaged glass. Letting the symbol sfail represent the dashed line
DruckerPrager: Y ¼ (i.e., 1.85 GPa), then the longitudinal stress at failure is given by
2:1 GPa P > 1:72 GPa
sx 5 P12sfail/3. Therefore, the estimate for the HEL of soda–
(6) lime glass is B4.5 GPa (P 5 B3.25 GPa) and that for Pyrex is
B5.2 GPa (P 5 B4.0 GPa). It should be noted that this value
0:012 þ 0:6~ sn s ~ n 1:65 GPa for the HEL for Pyrex is less than the 8 GPa reported in Table I
MohrCoulomb: t ¼
1:0 GPa ~ n >1:65 GPa
s of Bourne et al.34
(7) z
Bourne et al.,34,35 show uncertainty (‘‘error’’) bars of approximately 0.5 GPa centered
on their data values, for clarity, we have omitted these error bars.
It was also observed at the lowest impact velocities, the con- J
For convenience, for this discussion of flyer-plate impact, we assume that the stress is
positive in compression.
stitutive model underestimates the penetration resistance of the ww
Presumably, the glass remains elastic below the HEL, but fails upon reaching the
glass at early penetration times; however, the simulations repro- HEL.
3396 Journal of the American Ceramic Society—Chocron et al. Vol. 93, No. 10
7.0 through the Bourne data. But, as will be discussed in the next
Bomb Intact section, this does not satisfactorily explain the inconsistencies
6.0 Bomb Predamaged Alexander between the Bourne data and the characterization data that
Sleeve Predamaged
Alexander have been reported here.
Equivalent Stress (GPa)
Unconfined Intact
5.0 Adjusted
4.0 Bomb
Intact (3) Comparison of Characterization Experiments with
Bourne Intact
Flyer-Plate Impact Experiments
3.0
(A) Confined Compression Tests on Intact Borosilicate
2.0 Glass: The confined compression tests from the hydraulic
bomb (open diamonds) as well as the unconfined compression
1.0 Sleeve Predamaged tests (open hexagons) are also plotted in Fig. 10. An important
Bourne Damaged
Bomb Predamaged distinction between these data and the flyer-plate data is that the
0.0 characterization tests are used to define a failure surface,
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0
whereas some of the responses of the flyer-plate experiments
Hydrostatic Pressure (GPa)
(particularly at the lower pressures) appear to be elastic. None of
Fig. 10. Comparison of characterization data with flyer-plate impact the intact confined sleeve tests are plotted in Fig. 10 because a
data. small gap between the intact specimen and the sleeve can make a
large difference in the confinement pressure and hence strength,
and the glass can fail before it contacts the confining sleeve.
The slope of the elastic response in a seq–P graph is A dash-dot line, with a slope of 3, designates the elastic re-
seq 5 3(12n)P/(11n). Bourne and colleagues report that sponse of an unconfined uniaxial stress specimen, because for
n for soda–lime glass is 0.23, giving a slope of 1.32, which is uniaxial stress, P 5 sx/3. The three hexagon data points fall on
plotted as the medium-length dashed line in Fig. 10. Bourne and this curve. These points are plotted where the specimens failed.
colleagues soda–lime glass data fall on this line up to seq of ap- As mentioned previously, unconfined tests on brittle materials
proximately 3 GPa, at which point the data begin to deviate have large scatter in their compressive strengths.
from the line. The first Pyrex datum also lies on the medium- The hydraulic bomb tests (open diamonds) provide very re-
length dashed line, and not on the solid line (described in the liable failure data as a function of hydrostatic pressure. The long
next paragraph), even though Bourne and colleagues, report dashed line through the data is given by the fit in Fig. 5; also see
that n 5 0.20 for Pyrex.35 We will return to this observation Table II. Note that as the failure envelope is extrapolated to
later. s
high pressures, it runs almost through the ‘‘elastic’’ datum from
Alexander et al.,36 determined that the HEL for Borofloat Alexander et al.,36 but not the adjusted datum point. With re-
33—defined as the point where the stress-particle velocity load- spect to the hydraulic bomb data, there is no requirement for the
ing path is no longer linear, i.e., the elastic limit—to be 8.7 GPa. extrapolation of the failure surface to be linear with pressure.
This gives a value of 6.5 GPa for the equivalent stress (with This failure surface could easily have some curvature and go
n 5 0.20, from Table I). The corresponding pressure of 4.35 GPa through Alexander’s adjusted point; however, much of Bourne
is calculated from P 5 (11n)sHEL/[3(1n)]. This point is plotted and colleagues data lie below the strengths determined form the
as the open circle in Fig. 10, with the callout ‘‘Alexander.’’ The hydraulic bomb tests, and thus lie below a smooth envelope that
elastic slope is 1.5, denoted by the solid line
s
in Fig. 10. The solid would pass through the hydraulic bomb data and the adjusted
line goes through the HEL for Borofloat 33; i.e., the load path Alexander data point.
is elastic until the HEL. However, an elastic analysis has been (B) Predamaged Compression Tests on Borosilicate
assumed, i.e., the bulk modulus and shear modulus—and equiv- Glass: The ambiguities that exist in the interpretation of the
alently, Poisson’s ratio—are assumed to be constant. That is, the intact flyer-plate data do not seem to be evident in a comparison
fact that the solid line goes through the data point is a state- of the predamaged compression data and damaged flyer-plate
ment of self-consistency of the assumption of a perfectly elastic data. The characterization data from the hydraulic bomb (open
response. squares) and the confined sleeve experiments (open very small
A couple of interpretations will be explored in the paragraphs squares) are plotted with the other data in Fig. 10. The hydraulic
below. As already indicated, Bourne and colleagues data lie on bomb and confined sleeve data are triaxial tests; thus, the data
the dashed line (n 5 0.23) until approximately seq 5 3.0 GPa fall between the uniaxial stress loading path (the dash-dot line)
(P 5 B2.5 GPa), and above this stress, the data lie below the and the uniaxial strain line (the solid line).
dashed line. This suggests that soda–lime glass is no longer The hydraulic bomb data and the confined sleeve data are
responding elastically. Additionally, the Pyrex data deviate con- seen to overlap in pressure, and where they overlap, the equiv-
siderably from the n 5 0.23 line (and lie below the n 5 0.20 line), alent stresses as a function of pressure are in agreement. As al-
also suggesting that the Pyrex is not responding elastically. This ready stated, much higher confining pressures, and hence
could, perhaps, mean that the glass Bourne and colleagues were hydrostatic pressures, can be achieved using the confined sleeve
testing had some predamage, caused by the placement of the experimental technique. As observed in Fig. 8, the confining
lateral stress gages. This could also explain why a value of the pressures achieved were sufficiently high to observe a maximum
Pyrex HEL, which Bourne and colleagues inferred from their stress that can be carried by the damaged material. This is de-
data, is substantially lower than that determined by Alexander noted by the short horizontal dashed line in Fig. 10. This cap is
et al.36 in very good agreement with the cap observed in the Bourne and
However, there is another possible explanation/interpretation colleagues data. The two lines differ by only B12% (1.85 vs
of the intact flyer-plate data. The data in Fig. 10 Alexander and 2.1 GPa). The difference between these two values lies within the
colleagues data point has been adjusted to reflect a softening of scatter of the experimental data.
the bulk and shear moduli, as determined by Holmquist (T. J. There are several significant conclusions resulting from a
Holmquist, private communication). This softening is based on comparison of the data in Fig. 10: (1) there exists a maximum
the volumetric strain data from Alexander (C. S. Alexander, load-carrying capability, i.e., a cap, for the damaged glass; (2)
private communication), and a hydrostat for borosilicate glass the failure surface for the damaged glass is independent of strain
by Holmquist (T. J. Holmquist, private communication) using rate; and (3) there is not any significant decrease in the value of
data from Cagnoux.30 This point is denoted as ‘‘Alexander ad- the cap as a function of damage. This last conclusion results
justed.’’ The hydrostat indicates that there is a softening of the from the observation that a cap generated from the quasi-
bulk modulus and the shear modulus. This ‘‘adjusted’’ data static laboratory experiments does not decrease as a result of
point is more in line with a nonlinear envelope that passes further load–reload cycling (and hence, comminution) of the
October 2010 Characterization of Borosilicate Glass 3397
glass during testing, and from the flyer-plate experiments, which ment with the cap determined for damaged glass from the char-
presumably result in a very high degree of damage. acterization experiments. This agreement over many orders of
(C) MC Representation: The data in Fig. 10 are plotted magnitude in strain rate shows that a cap exists for damaged
in as seq vs P, which is convenient for representing the exper- (failed) glass, and that this cap is strain rate independent.
imental results in terms of a DP model. These data can At this point, the DP and MC constitutive models do equally
be readily converted to a MC representation by the following well in predicting the penetration response of a gold rod into
expressions: damaged borosilicate glass. Two of the three constitutive con-
stants (the slope and the cap) required for each model were de-
ðs1 þ s3 Þ seq rived from laboratory characterization experiments, but a third
¼Pþ (8) parameter—one that appears to be associated with the degree of
2 6
damage—had to be inferred from matching simulations to bal-
ðs1 s3 Þ seq listic experiments.24 These observations could potentially sim-
¼ (9) plify a more comprehensive glass model: damage seems to affect
2 2 only the zero-pressure intercept (DP model) or cohesion (MC
model) of the glass. A potential advantage of the MC model,
which may be more relevant for the intact material, is that the
The graph is not shown as it looks similar to Fig. 10.
MC model provides a characteristic failure angle due to the third
invarient, whereas for the DP model, damage is isotropic.
VI. Summary
Laboratory characterization experiments were conducted to
gather experimental data that could be used for developing a Acknowledgments
numerical constitutive model for borosilicate glass that was ap- The authors would like to thank Dr. Doug Templeton from TARDEC for
plicable to ballistic experiments. The strength of glass is pressure funding this work, and the administrative support provided by Mr. Rick Rickert
dependent; therefore, the characterization experiments were de- of TARDEC. Thanks are also due to the ARL for providing ultrasonic modulus
measurements of intact samples. Additionally, the authors thank Mr. Tim Holm-
vised to explore the strength of the glass as a function of con- quist (Southwest Research Institute) and Mr. Dennis Orphal (International
fining pressure. Interest was in both intact and damaged glass Research Associates) for their review of early versions of this article. In particu-
because simulations require a description of both intact and lar, we thank Dennis for his very careful reading and his many helpful comments
failed material. Three types of characterization experiments and suggestions.
were conducted:
(1) Unconfined compression on intact specimens (uniaxial
stress experiments). References
1
(2) Hydraulic compression using the hydraulic bomb appa- D. A. Shockey, D. Bergmannshoff, D. R. Curran, and J. W. Simons, ‘‘Physics
ratus (triaxial compression, giving equivalent stress versus con- of Glass Failure During Rod Penetration’’; pp. 23–32 in Advances in Ceramic
Armor IV: Ceramic Engineering and Science Proceedings, Vol. 29(6), Edited by
finement or hydrostatic pressure) on intact and predamaged L. P. Franks. Wiley-Blackwell, Hoboken, NJ, 2008.
specimens. 2
D. R. Curran, D. A. Shockey, and J. W. Simons, ‘‘Mesomechanical Consti-
(3) Confined sleeve compression experiments (triaxial com- tutive Relations for Glass and Ceramic Armor’’; pp. 3–14 in Advances in Ceramic
pression, where the confining pressure increases with axial load) Armor IV: Ceramic Engineering and Science Proceedings, Vol. 29(6), Edited by
L. P. Franks. Wiley-Blackwell, Hoboken, NJ, 2008.
on intact and predamaged specimens. 3
C. S. Desai and H. J. Siriwardane, Constitutive Laws for Engineering Materials
The focus of the characterization experiments was largely on with Emphasis on Geologic Materials. Prentice-Hall, Englewood Cliffs, NJ, 1984.
4
the latter two test procedures, with the objective of developing W. Chen and G. Ravichandran, ‘‘Static and Dynamic Compressive Behavior of
failure maps, as a function of pressure, for intact and predam- Aluminum Nitride Under Moderate Confinement,’’ J. Am. Ceram. Soc., 79, 579–
84 (1996).
aged material. The confined sleeve test, in principle, can explore 5
W. Chen and G. Ravichandran, ‘‘Dynamic Compressive Failure of a Glass
higher pressures than possible with the hydraulic bomb test. Ceramic Under Lateral Confinement,’’ J. Mech. Phys. Solids, 45, 1303–28 (1997).
6
However, for intact material, small gaps and/or misalignment of Z. Ma and K. Ravi-Chandar, ‘‘Confined Compression: A Stable Homogeneous
the specimen with the sleeve typically resulted in glass failure Deformation for Constitutive Characterization,’’ Exp. Mech., 40 [38] 38–45 (2000).
7
J. Lu and G. Ravichandran, ‘‘Pressure-Dependent Behavior of Zr41.2Ti13.8
before full potential confinement could be exerted by the steel Cu12.5Ni10Be22.5 Bulk Metallic Glasses,’’ J. Mater. Res., 18, 2039–49 (2003).
sleeve. However, the confined sleeve test worked very well for 8
W. Chen and H. Luo, ‘‘Dynamic Compressive Responses of Intact and Dam-
predamaged specimens, and higher confinement pressures were aged Ceramics from a Single Split Hopkinson Pressure Bar Experiment,’’ Exp.
achieved. The results for the hydraulic bomb and the confined Mech., 44 [3] 295–9 (2004).
9
P. Forquin, A. Árias, and R. Zaera, ‘‘An Experimental Method of Measuring
sleeve overlapped at pressures of about 350 MPa, providing the Confined Compression Strength of High-Performance Concretes to Analyse
increased confidence in the results. Their Ballistic Behaviour,’’ J. Phys. IV, 134, 629–34 (2006).
10
The experimental data were interpreted in terms of two pres- W. Chen and G. Ravichandran, ‘‘Failure Mode Transition in Ceramics Under
sure-dependent constitutive models, the DP model and the MC Dynamic Loading,’’ Int. J. Fract., 101, 141–59 (2000).
11
W. Chen and H. Luo, ‘‘Dynamic Compressive Responses of Intact and Dam-
model. Constitutive parameters were determined from the char- aged Ceramics from a Single Split Hopkinson Pressure Bar Experiment,’’ Exp.
acterization experiments. The slope (b) for the DP model and Mech., 44, 295–9 (2004).
12
the friction angle (f) for the MC model, determined from the H. Luo, W. Chen, and A. M. Rajendran, ‘‘Dynamic Compressive Response of
characterization experiments, were found to be nominally inde- Damaged and Interlocked SiC–N Ceramics,’’ J. Am. Ceram. Soc., 89, 266–73 (2005).
13
K. A. Dannemann, S. Chocron, A. E. Nicholls, J. D. Walker, and C. E.
pendent of the degree of damage to the glass (intact, predam- Anderson Jr., ‘‘Compression Testing and Response of SiC–N Ceramics: Intact,
aged, residual damage). However, the zero-pressure strength Damaged and Powder’’; pp. 109–16 in Ceramic Engineering and Science Proc.,
(Y0) and the cohesion (c) for the DP and MC models, respec- Advances in Ceramic Armor, 29th International Conference on Advanced Ceramics
tively, depended upon the degree of damage, with these param- and Composites, Vol. 26, Edited by J. J. Swab. American Ceramic Society,
Westerville, OH, 2005.
eters decreasing as damage increased. Additionally, the 14
K. A. Dannemann, S. Chocron, A. E. Nicholls, and C. E. Anderson Jr.,
confining pressures in the confined sleeve experiments were suffi- ‘‘Compressive Damage Development in Confined Borosilicate Glass,’’ Mater. Sci.
ciently high to achieve a saturation of the load-carrying ability Eng. A, 478, 340–50 (2008).
15
of the damaged glass, i.e., a cap. S. Chocron, J. D. Walker, A. E. Nicholls, K. A. Dannemann, and C. E.
Anderson Jr., ‘‘Analytical Model of the Confined Compression Test Used to
The data from the characterization experiments were Characterize Brittle Materials,’’ J. Appl. Mech., 75, 021006, 7pp (2008).
compared with flyer-plate impact experiments conducted on 16
S. Chocron, K. A. Dannemann, J. D. Walker, A. E. Nicholls, and C. E.
borosilicate and soda–lime glass. There appear to be some in- Anderson Jr., ‘‘Constitutive Model for Damaged Borosilicate Glass Under Con-
consistencies between the results of the characterization exper- finement,’’ J. Am. Ceram. Soc., 90 [8] 2549–55 (2007).
17
S. Timoshenko and J. N. Goodier, Theory of Elasticity, 1st edition, McGraw-
iments and the flyer-plate impact experiments for intact Hill, New York, 1951.
material. However, the equivalent stress derived for failed glass 18
K. W. Peter, ‘‘Densification and Flow Phenomena of Glass in Indentation
from the flyer-plate impact experiments is in very good agree- Experiments,’’ J. Non-Cryst. Solids, 5, 103–15 (1970).
3398 Journal of the American Ceramic Society—Chocron et al. Vol. 93, No. 10
19
F. M. Ernsberger, ‘‘Mechanical Properties of Glasses,’’ J. Non-Cryst. Solids, Predictive Modeling of Dynamic Processes: A Tribute to Klaus Thoma, Edited by S.
25, 293–321 (1977). Hiermaier. Springer Science1Business Media LLC, NY, 2009.
20 28
J. Lankford, C. E. Anderson Jr., A. J. Nagy, J. D. Walker, A. E. Nicholls, and Th. Behner, C. E. Anderson Jr., D. L. Orphal, V. Hohler, M. Moll, and D. W.
R. A. Page, ‘‘Inelastic Response of Confined Aluminum Oxide Under Dynamic Templeton, ‘‘Penetration and Failure of Lead and Borosilicate Glass Against Rod
Loading Conditions,’’ J. Mater. Sci., 33 [6] 1619–25 (1998). Impact,’’ Int. J. Impact Eng., 35 [6] 447–56 (2008).
21 29
T. J. Holmquist, G. R. Johnson, D. E. Grady, C. M. Lopatin, and E. S. Jr. C. E. Anderson Jr., S. Chocron, and T. Behner, ‘‘A Constitutive Model
Hertel, ‘‘High Strain Rate Properties and Constitutive Modeling of Glass’’; pp. for In-Situ Comminuted Silicon Carbide,’’ J. Am. Ceram. Soc., 93 [6] 1280–6
237–244, in Proceedings of 15th International Symposium, Ballistics, Jerusalem, Is- (2009).
30
rael, 1995. J. Cagnoux, ‘‘Shock-Wave Compression of a Borosilicate Glass up to 170
22
T. J. Holmquist and G. R. Johnson, ‘‘The Failed Strength of Ceramics Sub- kbar’’; pp. 392–6 in Shock Waves in Condensed Matter—1981, AIP Conf. Proc. 78,
jected to High-Velocity Impact,’’ J. Appl. Phys., 104, 013533, 11pp (2008). Edited by W. J. Nellis, L. Seaman, and R. A. Graham. AIP, NY, 1982.
23 31
D. C. Drucker and W. Prager, ‘‘Soil Mechanics and Plastic Analysis of Limit S. V. Rasorenov, G. I. Kanel, and V. E. Fortov, ‘‘The Fracture of Glass Under
Design,’’ Q. Appl. Math., 10 [2] 157–75 (1952). High Pressure Impulsive Loading,’’ High Press. Res., 6, 225–32 (1991).
24 32
R. M. Nedderman, Statics and Kinematics of Granular Materials. Cambridge Z. Rosenberg, D. Yaziv, and S. Bless, ‘‘Spall Strength of the Shock-Loaded
University Press, Cambridge, U.K., 1992. Glass,’’ J. Appl. Phys., 59 [8] 3249–51 (1985).
25 33
J. C. Jaeger and N. G. W. Cook, Fundamentals of Rock Mechanics. Chapman N. S. Brar, S. J. Bless, and Z. Rosenberg, ‘‘Impact-Induced Failure Waves in
and Hall Ltd. and Science Paperbacks, London, 1969. Glass Bars and Plates,’’ J. Appl. Phys., 59 [26] 3396–8 (1991).
26 34
S. Chocron, K. A. Dannemann, J. D. Walker, A. E. Nicholls, and C. E. N. Bourne, J. Millett, Z. Rosenberg, and N. Murray, ‘‘On the Shock Induced
Anderson Jr., ‘‘Static and Dynamic Confined Compression of Borosilicate Glass’’; Failure of Brittle Solids,’’ J. Mech. Phys. Solid, 46 [10] 1887–908 (1998).
35
pp. 67–72, in DYMAT 2009, 9th International Conference Mechanical and Physical N. K. Bourne, J. C. F. Millet, and J. E. Field, ‘‘On the Strength of Shocked
Behaviour of Materials under Dynamic Loading, EDP Sciences, PA de Court- Glasses,’’ Proc. R. Soc. Lond. A, 455, 1275–82 (1999).
36
aboeuf, France, 2009. C. S. Alexander, L. C. Chhabildas, W. D. Reinhart, and D. W. Templeton,
27
S. Chocron and C. E. Anderson Jr., ‘‘Numerical Simulations of the Penetra- ‘‘Changes to the Shock Response of Fused Quartz Due to Glass Modification,’’
tion of Glass Using Two Pressure-Dependent Constitutive Models’’; pp. 167–87 in Int. J. Impact Eng., 35 [12] 1376–85 (2008). &