You are on page 1of 4

Prediction of Fracture Energy and Flaw Size in Glasses

from Measurements of Mirror Size


J. J. MECHOLSKY, R. W. RICE, and S. W. FREIMAN
Engineering Materials Division, Naval Research Laboratory, Washington, DC 20375

Fracture strengths (u), fracture-initiating flaw sizes, and mir- r H n C K L E REGION


ror radii ( r ) ,as outlined by either the mist or the hackle bound-
ary, were measured for silicate and nonsilicate glasses (e.g.
AsZ&and glassy carbon). For all glasses, ur112= constant. The

-
average ratios of inner and outer mirror radii to flaw size were
-1O:l and 13:1, respectively, for most of the glasses. Critical
fracture energies calculated from either flaw or mirror size
agreed very well with those obtained by double-cantilever-beam
measurements.

1. Introduction

F OUR definitive regions surrounding fracture-initiating flaws in


silicate glasses have been observed (Figs. 1 and 2). The mirror
(a flat, smooth region) is bounded by the onset of mist (a region of
Fig. 1. Shape and general appearance of fracture mirror and
related features on typical glass fracture surface.
small radial ridges), which is bounded in turn by hackle (a region of
larger radial ridges), which is bounded by macroscopic crack
branching. It has been extensively dernon~trated'-~ that the product
of strength, u ,and the square root ofthe distance from the origin to
the onset of mist (i.e. the mirror radius, r M ) is constant in silicate
glasses:
ury112=
A (1)
Similar relations have been demonstrated in soda-lime glass for
radii determined by the onset of hackle ( r H ) and of crack branching
(re). A t t e m p t P have been made to relate these radii to the initial
flaw depth, u, or half-width, b, which can be obtained from the
Griffith-Orowan-Irwin equation,
a= Y 2 E y , / u 2 (h)
uu =Y d F y c (2b)
where Y is a geometric constant, E the modulus of elasticity, and y r
the critical fracture energy.* Estimates of the ratio of mirror size to
initial flaw size can range from 3: 1 (Ref. 4) to 28: 1 (Ref. 3 ) . The
work of Krohn and Hasselman6 indicates that the mirror-to-flaw
size ratio in soda-lime glass is = 10:1, and calculations using data of
McKinney' also lead to a ratio of 10:1.
It will be shown that both u ( r y ) 1 1 2 and u ( r H ) 1 1 2 areconstant for
silicate glasses; these relations are extended to nonsilicate glasses,
e.g. As& and glassy carbon. The rlb ratios were determined for all Fig. 2. Fracture surface of a borosilicate glass with features
the glasses studied, and it is demonstrated that the critical fracture like those identified in Fig. 1. Mist and hackle form an incom-
energies can be predicted from such measurements. plete semicircle, as is characteristic of specimens having mirror
sizes ? ' / 5 the thickness of bend specimens because of stress
gradient effects, as discussed in Ref. 1.
11. Experimental Procedure
In general, as-ground glassbars - 0 . 2 by 0.1 in. in cross section nique and that of a leached high-SiO, glass by an ultrasonic reso-
with a span length of 0.5 in. were tested in 3-point bending: at a nance technique. Fracture energy was determined by a constant-
loading rate of 0.05 in./min, Flaws introduced during grinding moment modification of the double-cantilever-beam technique',
usually acted as the source for failure; some silica glass rods were (Fig. 3), typically using specimens 2 by 0.5 by 0.04 in. with a center
flame-polished to achieve the higher strengths. Fracture surfaces groove to guide the sharp crack that was introduced before testing.
were examined optically, and the features (Fig. 1) were measured Critical fracture energies were determined by loading these speci-
with a microscope using a filar eyepiece. Flaw and mirror sizes were mens at a rate of 8x10-, in./min* until the crack propagated
always measured along the tensile surface (at constant stress), as in rapidly. The fracture energy was then calculated from
measurements described e l ~ e w h e r e The
. ~ ~ largest
~ sources of error
in these measurements were found at the extremes. The error in the yc= T2L2/2E/t (3)
measurement of large mirrors (low stress) results primarily from
diffuse mist and hackle boundaries, whereas that for the small Presented at the Fall Meeting of the Glass Division of The American Ceramic
Society, Bedford, P A , October 12, 1973 (Paper No. 36-G-73F). Received January
mirrors (high stress), probably the greatest error, results from the 18, 1974: revised copy received May 25, 1974.
small dimensions. For ease of analysis, only approximately Supported by the Advanced Research Projects Agency under Order No. 2029 (J. J.
Mecholsky and R. W . Rice) and by the Office of Naval Research under Contract No.
semicircular flaws, i.e. depth-to-width ratios between 0.35 and 84-F0l-23 (S. W. Freiman).
0.65, were considered. Elastic moduli were in most cases obtained *This quantity can be expressed in terms of the sfrain-energy release rate, G,, i.e.
yc = GJ2.
from a handbooklo or the literature.Il However, the modulus of tTensile testing machine, Ametek, Inc., Lansdale. PA.
0.3PbSe-0.7Ge,,5Aso.sSe3 was measured by a load-deflection tech- $A testing machine from Instron Corp.. Canton, MA, was used

440
October 1974 Prediction of Fracture Energy and Flaw Size in Glasses from Mirror Size 44 1
T d these materials as in silicate glasses, independent measurements by
several observers on the same samples yielded similar results.
Average values for measured inner and outer mirror radii, as well
as the flaw half-width, b, are presented in Table I . For most of the
glasses, the measured mirror-to-flaw size ratio was 2= 10:1 for the
inner mirror, as was calculated by Krohn and Hasselman6 for
soda-lime glass, and 13:l for the outer mirror. The greatest devia-
tions from these values occurred in glassy carbon and the leached
i high-SiO, glass, both porous materials.
b Logarithmic plots of fracture stress vs inner and outer mirror size
are presented in Fig. 5. The curves are best-fit straight lines of slope
i - 0.5. Lower slopes are observed in plots of strength vs mirror size
SECTION for specimens with residual s t r e s ~ . ~However,
J~ by replotting the
A-A
actual stress along the tensile surface (i.e. the applied stress minus
the surface stress), a slope of - 0.5 is again obtained. The limited

w 2 h
!
u TYPICAL DIMENSIONS-mrn
h-6.0
b- 1.0
1 - 0.5
L - 19.0
data for some of these glasses prevented a more accurate determina-
tion of the slope, but, as will be shown, the agreement between
fracture energies calculated from these plots and measured values
justifies the use of this slope. In addition, the mirror constants
determined from these curves agree with existing literature
value^.^*^^^ The proportionality constant, A , determined from the
intercepts of the plots in Fig. 5, is shown for each glass in Table II.
Fig. 3. Schematic of modified double-cantilever-beam speci- Critical fracture energies were calculated from these data and the
men. corresponding flaw-to-mirror size ratio from Table I by equating the
stress in Eqs. (1) and ( 2 ) , thus obtaining
where I is the moment of inertia of the specimen about its longitudi-
nal axis, t the specimen thickness at the groove, T the load on one y c= ( b / r ) ( A , / ( Y Z E ) ] (4)
arm, and L the length of the moment arm. All measurements were
made in air (=40% rh) at room temperature. A = Y [ (E y c ) / ( b / r ) ] 1 / 2

where Y = (l! 1.12) ( ~ / 2 ) for


' / ~a penny-shaped surface crack,15 i.e.
111. Results
a = b . Table II shows that the calculated values compare well with
Figure 4 shows mist and hackle about the source of failure in measured ones.
As,S, and glassy carbon as well as curved (Wallner) lines.13Figure That Eqs. (1) and (2) are expressions of the same phenomenon is
4(B) illustrates the large difference between inner and outer mirror supported by the agreement of the data with Eq. (4). However,
sizes in glassy carbon compared with that in the other materials. The when, for example, the outer mirror is used as the flaw size in Eq.
flaw half-width is =O. 1 the hackle radius and 2 0 . 2 the mist radius. ( 2 ) , the calculated fracture energy is no longer the critical fracture
Although the demarcation between mirror and mist is not as clear in energy but a larger value corresponding to branching associated

Fig. 4. Fracture surfaces of (A) As,S3 and ( B ) glassy carbon showing source of failure ( F ) , mist ( M ) , and hackle ( H ) regions. The
fracture surface of As2S3is similar to those of silicate glasses, whereas in glassy carbon the separation of mist and hackle is greater.
Table I. Flaw-to-Mirror Size Ratio for Glasses Studied
Mirror radii ( m m ) t t
Flaw half-width, lu t$ Ln tt
Glass Inner, rM Outer, rH b (mm)tt b b
SiO,* 1.02 1.31 0.08 12.5 16.7
Ahminosilicate* 0.71 0.86 .06 12.5 14.3
Borosilicate* .78 1.04 .08 10.0 12.5
Soda-lime? .59 0.85 .05 12.5 16.7
Lead silicate* .61 0.77 .06 10.0 12.5
Leached high-SiO,* .75 1.40 .04 14.3 25.0
Glassy Cg .18 0.32 .04 4.6 7.7
As,S,V .95 1.25 .ll 8.3 12.5
0. fPke-O.7Ge ,5Aso.,Se,'1 .28 0.37 .05 7.1 9.1
Ge,,As &5e5,** .75 1.10 .07 10.7 15.7
'Coming Glass Works, Coming, NY; Code Nos. (in order listed) 7940, 1723, 7740, and 7930. tPPG Industries, lnc., Pittsburgh, PA. SBausch & Lomb, Rochester, NY; No.
617366. $Beckwith Carbon Corp., Van Nuys, CA. 7Servo Corp., Hicksville, NY. "Catholic University of America, Washington, DC.**Texas Instruments, lnc., Houston, TX;No.
TI-20. ttAvg for all observed mirrors. $$Avg for semicircular flaws only,
442 Journal of The American Ceramic Society-Mecholsky et al. Vol. 57, No. 10
ure of brittle materials. After enough energy has been supplied to
0 GLASSY CARBON the crack, it will propagate at a velocity which increases as its length
BOROSILICATE increases. Since there is no known inflection in the
SODA LIME crack-velocity-crack-length curve at the critical flaw size, it is
2001
F L l M E \ 3 3-49
v AL~MINOSILICATE
0
0
LEAD SILICATE
LEACHED HIGH-
expected that the flaws observed are preexisting flaws of near-
critical size. They are observed because any slight change in the
direction of stress from that which initially produced the flaw will
lead to a demarcation line.I6 At the fairly high strain rates (0.05
in./min) at which the strengths were measured, this initial flaw
would enlarge only slightly in size before becoming “critical.”
Since the driving force depends on crack length, crack velocity will

40F increase until it approaches a terminal velocity, V,, determined by


the shear-wave velocity, V,?,in that material, i.e. V,==0.6Vs.As the

“110 (A)
crack approaches V,, the sum of the potential energy resulting from
its increasing length and the kinetic energy resulting from its motion
becomes greater than the energy that can be used to increase the
velocity of the crack. Small cracks are nucleated around the tip of
the main crack,” forming mist, but there is insufficient energy to
propagate these secondary cracks very far. Limited velocity in-
creases allow propagation of such secondary cracks to form hackle.
I I I 1 I I l l 1 I I I l l - 30.0 Finally, when enough energy is available, the crack can branch
- x A52S3
macroscopically.
0 GLASSY CARBON
.SILICA7 BOROSILICATE Two major reasons account for much of the variation in mirror-
SODA L
SODA
A 302
~ME
LIME
GLASS
~ - 20.0 flaw size ratios observed by various investigators; one is the use of
crack velocity to measure this ratio. Most of the velocity-flaw-size
relations are of the form V 2 m [ 1 - (b/r)], where V is the velocity of
- 10.0; the crack front of length r and b is the critical flaw size.18 Because
b l r is small compared to I , however, there is little difference in the
velocity for a wide range o f h l r ratios. Therefore, estimates of ratios
of (hackle) radius to flaw size from this type of data cannot be
differentiated within the limits imposed by this criterion. When the
crack velocity data of Congleton and Petch4 are used to obtain the
(hackle or branching) mirror-to-flaw size ratio, the estimates could
be as low as 3.1. However, Congleton and Petch4 point out that
crack velocity is not the critical parameter in determining when
branching will occur; rather, crack branching occurs at a particular
stress intensity factor, i.e. a particular mirror constant. Although
I I I I 1 1 l 1 1 I I I I I I
0.2 03 0.4 0.50.6 0.8 1.0 20 4.0 6.0 these workers did not make the calculations, when the (hackle)
OUTER MIRROR RADIUS ( m m ) mirror-to-flaw size ratio is calculated directly from their mea-
Fig. 5. Strength vs ( A ) inner (mist) mirror radius and ( B ) surements of flaw size, mirror constant, and strength, e . g .
outer (hackle) mirror radius of silicate and nonsilicate glasses. r,/b= ( A / C ) ~l/b),
( a ratio of = 12: I is obtained, in agreement with
Solid lines represent slopes of -0.5. Error bars indicate true the present work.
scatter about average for number of tests given. The second reason for variations in mirror-to-flaw size ratios is
the use ofdifferent experimental techniques, e.g. sample geometry,
with hackle. Equations ( I ) and (2) are equivalent when the mirror internal stress, and especially the method of measurement. An
constant is as expressed in Eq. ( 5 ) , where yc is the critical fracture excellent example is the work of Levengood,3 who measured mirror
energy. depth, r 2 , as opposed to radius (or diameter) along the tensile
The chalcogenide glasses, Ge,,As,,Se,, and 0 . 3 P b S e - surface, r , . Because of the stress gradient in bend bars, this tech-
0.7Ge,,,As,,,Se3, fit Eq. ( I ) but are not represented in Figs. 3 nique is likely to lead to high values of the mirror size. By correcting
and 4 because of limited data. Because it was difficult to handle and Levengood’s mirror (mist) depth-to-flaw size ratio (28: 1) for a
obtain specimens of sufficient size, fracture energy data were not radius along a constant tensile stress surface, i.e. r , / r 2 = ( A , / A # ,
taken for these glasses. where A l and A 2 are the mirror constants corresponding tor, andr,,
IV., Discussion respectively, a value of = 12: 1 is obtained, in good agreement with
the present work.
The Griffith-Orowan-Irwin expression presents a criterion for Kerkhof and Richter5 report a value of 6:1 for G,,,/G,, (the
propagation of preexisting flaws that generally determines the fail- strain-energy release rates for mist and flaw). It can be misleading to

Table 11. Calculated and Measured Fracture Energies


Fracture energy (J/m2)
Avg fracture Modulus of Mirror constants (MN/rn3t2) yc (talc.)*
stress, u elasticity, E
Glass ( M N / m 2 x lo-’) ( M N / r n 2 x lo-‘) A, (inner) A,, (outer) Inner Outer yc (meas.)

SiO, 7.82 7.21 2.23 2.42 4.4 3.9 3.7


Aluminosilicate 9.74 8.91 2. I4 2.40 3.3 3.6 3.7
Boros i licate 8.76 6.37 1.87 2.10 4.4 4.4 4.0
Soda-lime 8.37 7.34 1.92 2.21 3.2 3.2 3.5
Lead silicate 6.48 5.24 1.61 1.78 4.0 3.9 3.8
Leached high-Si02 2.82 1.79 0.91 1.19 2.2 1.9 2.5
Glassy C 7.51 2.48 1.17 1.67 9.7 0.8 8.5
As&, I .86 1.66 0.56 0.65 1.8 1.6 2.1
0.3PbSe-0.7Ge,,,As,,,,Se:, 2.28 1.38 .48 .55 1.9 1.9
Ge,,,,As ,2Sejj 1.98 2.21 .55 .65 1.o 1 .O
* y , = [ A 2 ( b / r ) ] / ( Y 2 EY)2;= 1.25. b = flaw half-width, r = radius of mirror.
October 1974 Single-Crystal Diffraction Patterns from Vitreous Carbon 443
predict r,vlb ratios from this type of data since, in this case, G,v was glasses over a wide range ofcomposition. The exceptions are glassy
calculated from mirror measurements on rods, whereas G,. was carbon and leached high-SiO, glass, both of which contain >20%
obtained from plate measurements. Thus, G,v/G, is equal to r,vlb fine porosity. A comparison of fracture energies calculated from the
only when the 2 strength distributions are equal, which they are not mirror data with measured values supports the accuracy of these
for such different test techniques. results.
The ratios of outer (hackle) mirror-to-flaw size (13: I ) and inner Acknowledgments: W e thank J . Breen and G . Kendrick for machining the
(mist) mirror-to-flaw size (1O:l) obtained in the present study thus samples and acknowledge Coming Glass Works, Bausch & Lomb. and Catholic
University for providing some of the material.
agree very well with previous estimates, within the limits imposed
by the different experimental techniques. The accuracy of the val- References
ues measured is confirmed by the excellent agreement of the critical ’ J . W . Johnson and D. G . Holloway, “Shape and Size of Fracture Zones on Glass
Fracture Surfaces,” Phil. Mag.. 14 (1301 731-43 (1966).
fracture energies determined through the use of these ratios with E. B. Shand. “Breaking Stresses of Glass Determined from Dimensions of
those measured directly. Fracture Mirrors,” J. Amer. Cerum. Soc., 42 [lo]474-77 (1959).
W. C. Levengood, “Effects of Original Flaw Characteristics on Glass Strength,”
Although much past research in fracture (mirror) surfaces has J. pf pl Ph s , 29 [5]820-26 (1958).
been clarified herein, many questions remain to be answered. In the d o n g i t o n and N . J. Petch. “Crack Branching,”Phil. Mag.. 16[142]749-60
present study, only circular flaws were considered. Whereas in (1967).
F. Kerkhof and H. Richter; pp. 463-74 in Fracture 1969.Edited by E. I . Pratt and
practice many flaws can be idealized by an ellipse, many others are P. L. Pratt. Chapman and Hall Ltd.. London, 1969.
irregularly shaped. For elliptical flaws, it is important to measure ‘ D. A . Krohn and D. P. H. Hasselman. “Relation of Flaw Size to Mirror in the
Fracture of Glass,” J. Amer. Cerum. Soc., 54 [8]41 I (1971).
the flaw and mirror along the same line in a constant stress field. ’ K. R. McKinney, “Relation of Fracture Mechanics to Mirror Surfaces of Fractured
Preliminary results indicate that calculation of elliptical flaw sizes Brittle Materials.” {bid..56 [4]225 (1973).
from mirror size data, using the material constants established in the ” E. F. Poncelet, “Markings on Fracture Surfaces,” J. Glass Techno/.. 42,
279-88T (1958).
present study, will give accurate results for glasses. More complex ’ Leighton Orr, “Practical Analysis of Fractures in Glass Windows,” Marer. Res.
sources of failure, e.g. 2 or more interacting flaws, may alter the Stand., 12111 21-23, 47 (1972).
l o E. B. Shand, Glass Engineering Handbook. McGraw-Hill Book Co., New York,
ratio of the mirror size to the size of an idealized flaw producing the .
105x
,“”.
same strength. However, this effect must be limited, since speci- ‘ I S . M . Wiederhom, “Fracture Surface Energy of Glass,” J Amer. Cerum. Soc..
52,\2]99-105.(1969).
mens of a given strength had similar mirror sizes (i.e. within the S. W . Freiman, D. R . Mulville, and P. W. Mast. “Crack Propagation Studies in
typical scatter bands of Fig. 5 ) . Whether environment or loading Brittle Materials.” J. Murer. Sci.. 8 III 1 1527 (1973).
l 3 E. B . Shand, “Experimental Slidy bf Fraciure of Glass: 11,” J. Amer. Cerum.
rate affects crack propagation, and hence the mirror-to-flaw size Soc., 37 [I21 559-72 (1954).
ratios, has not been fully elucidated, but work on the problem is I ‘ M . J. Kerper and T . G . Scuderi, ”Relation of Strength of Thermally Tempered

c~ntinuing.’~ Glass to Fracture Mirror Size,” Amer. Ceram. Soc. Bull.. 44 1121 953-55 (1965).
j 5 G . R. Irwin and P. C . Paris; pp. 1-46 in Fracture, Vol. 111. Edited by H.
Liebowitz. Academic Press, New York, 1971.
V. Summary “ S . M . Wiederhom;pp. 503-28 in Materials Science Research, Vol. 111. Edited by
W . W . Kriegel and Hayne Palmour 111. Plenum Press, New York, 1966.
The literature contains a wide range of estimates for fracture “J. E. Field, “Brittle Fracture: Its Study and Application,”Conremp. Phys.. 12[I]
1-31 (1971).
mirror-to-flaw size ratios in glasses. However, when all measure- ’” E. N. Dulaney and W . F. Brace, “Velocity Behavior of a Growing Crack,” J.
ments are made in the same manner and under the same environ- Appl. Phys., 31 1121 2233-36 (1960).
l o H. Kirchner and R. M . Gruver; pp. 309-21 in Fracture Mechanics ofceramics,
mental conditions, the ratio is = 10:1 for the inner mirror and = 13: 1 Vol. 1. Edited by R. C . Bradt, D. P. H. Hasselman, and F. F. Lange. Plenum Press.
for the outer mirror. These ratios appear to be constant for most New York, 1974.

Single-Crystal Diffraction Patterns from Vitreous Carbon


A. GREENVILLE WHITTAKER
The Aerospace Corporation, El Segundo, California 90009
B. TOOPER
Electron Microscopy Service, Pasadena, California 91 101

In contrast to the broad X-ray diffraction maxima typical of tallite size, which was reported to range from 40 to 80 I\, as the layer
vitreous materials, electron diffraction studies revealed many diameter. With crystallites of this size, it is often possible to obtain
regions that give single-crystal patterns. Although some of the single-crystal patterns by electron diffraction. Since such a study
patterns are those of graphite, as expected, many that were had evidently not been made, the electron diffraction work was
obtained corresponded to other carbon crystal forms, e.g. cha- continued to determine what crystal forms could be found in vitre-
oite and carbon VI. In addition, ring patternsof lonsdaleite and ous carbons.
diamond were found, and patterns that do not correspond to 11. Experimental Procedure
any known carbon allotrope were obtained.
The vitreous carbons used were R292* (heated to 1800°C) and
L-2000 and L-3000’ (heated to 2000°C for 2 h and to 3000°C for 8
1. Introduction h , respectively, during manufacture).9 For electron diffraction ex-
amination, the entire sample was crushed, and a few particles were

V ITREOUS carbon was used in a study of the effect on carbons of


intense heating by a continuous C 0 2 laser.’ Both chaoite* and
carbon V13 were produced from vitreous carbon, as indicated by
selected from the interior region. These particles were mulled in a
collodion dispersant, mounted on standard 200-mesh grids, and
studied, using selected-area diffraction conditions, in an electron
electron diffraction patterns; many faceted crystals of both carbon microscope .$
forms occurred in the laser-heated region. However, during exami-
nation of vitreous carbons that had not been laser-heated, single- Presented at the Pacific Coast Regional Meeting. The American Ceramic Society.
crystal electron diffraction patterns also occurred rather frequently. Anaheim,CA, November 3,1971 (CarbonandGraphire, No. 13-CG-71P). Received
March 13, 1974.
A search of the literature on the structure of vitreous carbon revealed Supported by the United States Air Force under Contract No. F04701-73-C-0074.
that several X-ray diffraction studies have been In *Beckwith Carbon Corp., Van Nuys, CA.
general, that work indicated that vitreous carbon gave broad diffrac- t2000 and 3000 grades, respectively, Lockheed Missiles& Space Co., Sunnyvale,
CA .
tion lines which provided information on bond distances and crys- SHU-11, Hitachi Ltd., Tokyo, Japan.

You might also like