ICCM 11v3
ICCM 11v3
ICCM 11v3
INTRODUCTION
In the absence of a matrix crack, the interface becomes important only in the presence of a
fiber break. Fairly high shear and compressive radial stresses are generated at the interface
next to the break [2-4], and the former can be sufficient to cause interface shear failure and
relative fiber-matrix sliding. The sliding resistance at the interface is usually represented by
an effective frictional sliding stress, τ. The prevailing understanding is that when τ is high,
there is high elastic stress concentration in the adjacent fiber to cause it to fail in a cooperative
III - 238
Proceedings of ICCM–11, Gold Coast, Australia, 14th-18th July 1997
The exact role of the interface on load sharing behavior was recently quantified in [9], based
on the ratio of the stress concentration in nearest and next nearest neighbors. For a 35
volume percent SiC/Ti-alloy TMC, the model indicates that GLS would occur when τ was
less than about 10% of the fiber strength. Using available data on interface friction stress
from push-out tests and extracted fiber strength, this implies that GLS would prevail at room
temperature for most composites of interest. On the other hand, except for reference [10]
where there are concerns regarding the analysis used, available strength data and fiber break
density and locations in a number of TMCs [11,12] suggest that LLS is the dominant
mechanism. At the load at which the first fiber fails randomly, the stresses in all fibers are
quite high, so that even small stress concentrations can set up successive failures of both
nearest and next-nearest neighbors. Also, plasticity from a fiber break can produce significant
stress concentration and lead to LLS, as will be demonstrated in this paper.
Past experimental efforts for evaluating the influence of the interface on longitudinal
properties have typically relied on heat treatments of a composite [13,14], such that the
interface is altered by adverse fiber-matrix reactions. Such reactions serve to change
(increase) the fiber-matrix bond strength and frictional resistance, which is required for a
comparative study, but which have the undesirable consequence of significantly degrading
both fiber and the matrix properties. Consequently, strength loss of the composite cannot
simply be ascribed to changes in interface conditions and load sharing behavior. This work
differs from previous efforts in that attention was focused on the mechanisms of progressive
fiber failures at the micro and meso scales, rather than on a macroscopic parameter such as the
composite strength. Attention was confined to composites with only limited number of fibers,
in single-fiber and multiple-fiber single-ply configurations. In particular, the single-fiber
experiments and analysis methodology [15,16] were used to characterize the interface under
fragmentation conditions, and to subsequently use that data to evaluate fiber failures in
multiple-fiber specimens. The interface was varied by selecting SiC fibers with different
coatings, and associated different interface properties. This overall approach using
mini/micro-composites is part of our continuing effort at establishing a comprehensive
methodology for developing optimized interfaces in TMCs.
EXPERIMENTS
Single-fiber and multi-fiber single-ply specimens were fabricated by hot pressing SiC fibers
between two Ti-6Al-4V sheets at 950°C. Tensile specimens with a 25.4 mm gage length
x15.2 mm width x 0.7 mm thickness were prepared by electric discharge machining. Three
different fibers were considered: (i) SCS-0 fiber, which is a 140 m diameter uncoated SiC
fibers from Textron Specialty Metals (TSM), (ii) SCS-6 fiber, which is essentially the SCS-0
fiber coated with turbostatic carbon (graphitic) with graded amounts of Si, and (iii) Trimarc
III - 239
Volume III: Metal Matrix Composites and Physical Properties
fiber, which is a 125 m diameter SiC fiber from ARC Amercom, with a "soft-hard-soft"
turbostatic carbon coating. These fibers were selected because the interfaces represent a
rather wide range of interface tensile (normal) strength [17-19] and frictional push-out
stresses (see Table 1), so that the influence of the interface on load sharing behavior can be
assessed. Multiple fiber specimens were prepared in two different configurations. One was a
15-fiber specimen with a 0.7 mm fiber spacing, and was used primarily to understand the
basic deformation mechanisms. The other multiple-fiber configuration was a 40-fiber
specimen with a center-to-center fiber spacing of 200 m, that is characteristic of typical 30
volume percent TMCs. These 40-fiber samples had smaller specimen dimensions, with a
fiber volume percent of approximately 11.6 for the SCS-6 and SCS-0 fibers, and
approximately 9 for the Trimarc fibers. Tensile tests were conducted at room temperature
(RT), and acoustic emission (AE) sensors were used to detect fiber breaks. The break
locations were determined using an ultrasonic shear wave back reflection (SBR) technique
[20-22], and those results correlated extremely well with the break locations determined
metallographically.
Table 1: Interface and Fiber Strength Data for the SiC/Ti-6Al-4V Systems
DATA ANALYSIS
Past research in single-fiber fragmentation (SFF) testing has relied primarily on the well-
known Kelly-Tyson equation [23]: τ = (σf r)/Lc, where Lc is the critical length of fibers at
break saturation, and σf is the fiber strength. Since fiber strengths generally follow Weibull
statistics, this requires an independent measurement of fiber strengths. Also, the procedure
assumes that the insitu strength of the fibers in the composite remain the same as when fibers
are tested independently, an assumption that may be violated in TMCs. In addition, the
formula does not provide a methodology to assess stress concentration effects and load
sharing in multiple fiber samples. After surveying a number of recent models [24,25], the
analytical fiber fragmentation model of Curtin [26] was selected for assessing the
fragmentation data. We also performed Monte Carlo simulation of the fragmentation
experiment, and found excellent correlation with Curtin's model. The three parameters that
emerge from the analysis are: the insitu Weibull modulus (m) and Weibull strength (σo) of
the fiber, and the shear stress, τ, at the interface under fiber-fracture conditions. The details of
the application of the model in experiments are described in [15]. Essentially, two plots are
obtained from the fragmentation test: (i) the stress plot, which contains the probability
(LnLn{1/(1-Pf)}) of fiber breaks plotted versus the logarithm of the applied stress, and (ii)
cumulative number of breaks plotted versus the fragment length; here Pf is the fiber failure
III - 240
Proceedings of ICCM–11, Gold Coast, Australia, 14th-18th July 1997
probability. The analysis involves selection the three parameters that best match the two
experimental plots. Among them , m is the simplest to evaluate, since it represents the slope
of the stress plot. Also, if log(N) is plotted versus log(stress), the curve is linear in the initial
stages with a slope close to m (see additional details in [15]); here N is the cumulative number
of fiber breaks. Comparison of single and multiple fiber results provide a means of evaluating
whether stress concentration effects are operative.
RESULTS
Fig. 1 shows SBR scans for the single fiber specimens. The location of breaks are indicated
by the short white line (only a two are marked for the Trimarc fiber). From such scans and
the stress-strain-AE data, the stress and fragment length plots were obtained.
Fig. 1: SBR images for the single fiber specimens: (a) SCS-6 fiber, (b) Trimarc fiber.
Trimarc/Ti-6Al-4V
Single-Fiber MMCs
3 30
2
25
Theoretical:
Cumulative Fragments
so=3050 MPa 20
Theoretical:
m=9 so=3050 MPa
lnln(1/1-Pf)
Lo=25.4 mm m=9
-1 15 Lo=25.4 mm
Experimental
Experimental
-2
Experimental 10 Experimental
-3 Experimental Experimental
Pf=i/(Nf+1) 5
-4 i=rank, Nf=
total breaks.
-5 0
7 7.5 8 8.5 9 9.5 10 0 0.5 1 1.5 2
25a1a229athtrim.plot1 Ln(Stress, MPa) Frag Length/Avg. Length (x/x )
avg
Fig. 2: (a) Weibull stress plot for the single-fiber Trimarc/Ti-6Al-4V sample
(b) Corresponding fragment length distribution plot
Figs. 2a and 2b illustrate the results for the Trimarc/Ti-6Al-4V system. Referring to Fig. 2a,
it may be noted that the fiber stress was determined by multiplying the measured strain by the
fiber modulus and then subtracting the residual stress in the fiber. This assumes isostrain
conditions, which is violated when substantial gaps develop between the ends of the broken
fiber. The deviation of the data from the theoretical curve in Fig. 2a is believed to arise from
this effect. Hence, only the initial part of the theoretical curve was used to match with the
data in Fig. 2a. The second point to note is that the axial compressive residual stress in the
fiber was determined by an etching technique to lie between 600 and 900 MPa for both the
III - 241
Volume III: Metal Matrix Composites and Physical Properties
single and multiple-fiber specimens. This is well below that based on the assumption of a
stress-free fiber at the processing temperature, and is believed to be due to loading of the fiber
at the processing temperature (because of the very small amount of fibers).
Using plots such as Fig. 2, the insitu Weibull strength, σo, the Weibull modulus, m, and the
frictional sliding stress, τ, were determined. They are listed in the last three coulmns of Table
1, where σo is determined at the gage length Lo=25.4 mm. The table also contains data from
tensile tests on extracted fibers, frictional stress from push-out tests, and the tensile (normal)
strength measured by using the cruciform specimen geometry [17-19].
A common feature of the data in Table 1 is the significantly higher friction stress obtained
from the fragmentation experiment compared with the push-out test. For example, in the case
of the Trimarc fiber, the fragmentation frictional stress is nearly six times that obtained from
push-out tests. In the case of SCS-6 fibers, the frictional sliding stress determined from the
fragmentation experiments is 390 MPa, compared with only 160-190 MPa obtained from
push-out tests; the latter is slightly higher than reported values for a 30 volume percent SCS-
6/Ti-6Al-4V composite [27]. These high friction stresses under fragmentation conditions
suggest that there may be significant clamping of the fiber at the fiber break, which may
prevent propagation of the interface crack. If a friction coefficient of 0.5 is assumed for the
SCS-6 fiber (based on push-out tests), it implies that in the fragmentation test there is a radial
clamping on the fiber of 780 MPa, which is of the order of the yield stress of the matrix,
approximately 850 MPa. The analysis in [2,3] suggest that this level of stress can indeed be
generated in the proximity of the break. A more recent elastic-plastic analysis [4] indicates
that even if a small split was assumed along the interface (representative of a debond crack),
the local radial stress at the tip of that crack is negative and singular, suggesting that it would
always be very difficult to propagate the interface debond crack under fragmentation
conditions.
Fig. 3 illustrates the interface region next to a fiber break in a SCS-6/Ti-6Al-4V specimen.
Theoretical:
2 3050-9-190-25.4mm
Experimental, 25.4 mm
1 Experimental, 25.4 mm
Experimental, 25.4 mm
Theoretical:
0
3050-9-190-381mm
Experimental, 15 fiber, 381 mm
-1
7 7.5 8 8.5 9 9.5 10
Ln (Stress, MPa)
Fig. 3: Micrograph showing coating damage Fig. 4: Cumulative number of breaks versus
near a fiber break in a SCS-6/Ti-6Al-4V the stress for the single and 15-fiber Trimarc
sample. composite. The predictions for 25.4 mm
length and 381 mm length are indicated
III - 242
Proceedings of ICCM–11, Gold Coast, Australia, 14th-18th July 1997
Fig. 4 is a stress plot for a 15-fiber Trimarc/Ti-6Al-4V specimen, where ln(N) is plotted
versus the logarithm of the breaking stress, ln(σ). This method of plotting is necessary
because the specimen cannot be taken to fiber fracture saturation. Fig. 4 includes data from
the SFF tests, as well as the theoretical curve for a gage length of 25.4 mm, using parameters
that were determined earlier, i.e., σo=3050 MPa, m=9, and τ=190 MPa; the designation in the
plot is 3050-9-190. Also included in Fig. 4 is the theoretical curve for the same set of
parameters, but for a total fiber length of 381 mm, which is the total length available for
fracture in the 15-fiber/25.4 mm gage-length specimen. The basis for this latter curve is that
it represents the failure stress distribution if there was no interaction between the 15 fibers.
Thus, any interaction between neighboring fibers would manifest in the form of a lack of
agreement with the theoretical 15-fiber curve.
Fig. 4 shows that although the 15-fiber specimen approached the theoretical curve at higher
stresses, the experimental data significantly deviated from the 15-fiber theoretical curve at
lower values of stresses. During the initial stage, when fiber strain is well represented by the
measured strain, the slope for the experimental data is higher than the theoretical slope by
about a factor of 2. Since the slope of the curve represents the Weibull modulus, the 15-fiber
data suggests a higher apparent Weibull modulus for the fibers. Since there is no reason why
the insitu fiber statistics should differ from the SFF samples, the higher apparent Weibull
modulus may be interpreted as being due to a stress-concentration effect between neighboring
fibers. This is because the occurrence of a break under stress concentration conditions would
quickly load up fibers in adjacent locations, thus setting up a succession of breaks with very
little increase required in the applied stress/strain. In essence, this would manifest as a higher
Weibull modulus.
Fiber Crack
(a) (b)
Fig. 5:(a) Face of the 15-fiber Trimarc specimen showing macroscopic slip bands on the
specimen face. (b) SBR image, illustrating that the intersection of bands with the fiber
correspond to the fiber break locations.
III - 243
Volume III: Metal Matrix Composites and Physical Properties
Figs. 5a and 5b correspond to the 15-fiber specimen whose data were plotted in Fig. 4. Fig. 5a
is a macroscopic view of the specimen face and shows a criss-cross pattern of slip bands that
are concentrated in a relatively small section of the gage length. Figure 5b illustrates the fiber
break locations, as imaged by the SBR technique, and shows a one-to-one correspondence
between the slip band intersections with the fiber and the fiber break locations. The intense
but macroscopic slip bands are not observed in the unreinforced material, nor can they be seen
near the specimen edges in Fig. 5a because the fibers are located a small distance away from
the edges. Together, Figs. 5a and 5b indicate that fiber fractures have occurred cooperatively,
with localized matrix plasticity being responsible for cumulative failure of fibers. This type
of LLS behavior is consistent with the higher Weibull modulus for the 15-fiber specimen
compared with the single fiber specimen (see Fig. 4).
Interrupted tests, where specimens were periodically unloaded to obtain SBR and plasticity
images [15], have confirmed that fiber fractures are strongly influenced by slip bands which
propagate from a fiber break and impinge on the neighboring fiber. Note that the LLS
behavior that is observed is not inconsistent with the results in [9], since that model predicts
that LLS will always occur in single-ply composites. However, the mechanism of stress
concentration is different here, and needs to be modeled rigorously.
In order to illustrate how these results translate to higher volume fraction composites, Fig. 6 is
presented for the 40-fiber specimens that had a center-to-center fiber distance of 200 m. The
stress-strain curve for the Ti-6Al-4V matrix is also plotted here.
40-Fiber SiC/Composites
1600
vf = 11.6% for SCS-6, SCS-0
1400 vf = 9%, for Trimarc
1200 D
A
1000 B
C
800 E
600
Ti-6Al-4V Matrix
400
SCS-0/Ti-6Al-4V
200 SCS-6/Ti-6Al-4V
Trimarc/Ti-6Al-4V
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
28abcmat.plot1 Strain, %
Fig. 6: Stress-strain curves for the 40-fiber single-ply composites and the neat Ti-6Al-4V
matrix.
III - 244
Proceedings of ICCM–11, Gold Coast, Australia, 14th-18th July 1997
In this figure, point D corresponds to failure of the SCS-6/Ti-6Al-4V specimen, and point A
corresponds to an instability point in the Trimarc/Ti-6Al-4V specimen. For the latter
specimen, instability ended at point B, and the specimen was subsequently unloaded at point
C, by which time the crack in the specimen had almost fully propagated through the specimen
width. The SCS-0/Ti-6Al-4V specimen exhibited a constant flow stress well below that of
the matrix, and it was unloaded at approximately 1% strain. The stress-strain curve for the
SCS-0 case became non-linear at approximately 0.5 % strain, which was substantially lower
than the flow stress of the Ti-6Al-4V matrix. Additional important parameters about the data
are provided in Table 2.
Table 2: Tensile Data and Predictions for the 40-Fiber Single-Ply Specimens
The most revealing aspect of the above specimens is obtained from optical observation and
SBR images. Fig. 7 shows the face of the specimens, wherein it may be observed that an
intense slip band has traversed the entire width of the specimen for the SCS-6 and Trimarc
fibers, both of which possess relatively weak interfaces. Final fracture took place inside these
intense slip zones, and SBR images showed that the fibers were severely fragmented within
the zone. No fiber fractures are determined outside the band, consistent with slip not being
observed at other locations, as well as lack of any significant AE event prior to specimen
fracture. The fiber failure locations confirm LLS in these materials with relatively weak
interfaces. The specimen with SCS-0 fibers shows a rather unanticipated behavior, in that the
entire gage length is filled with a criss-cross arrangement of localized slip, presumably
connecting regions of severely fragmented fibers.
DISCUSSION
The fragmentation experiments and analysis show that useful data can be extracted from the
single-fiber fragmentation test, and that the analysis can be used to assess load sharing
behavior in multiple-fiber specimens. Thus, local load sharing manifests as a higher insitu
Weibull modulus for the multiple fiber specimen compared with the single fiber sample.
III - 245
Volume III: Metal Matrix Composites and Physical Properties
Fig. 7: Macroscopic slip traces on the faces of the 40-fiber specimens. (a) and (b) show that
fracture was contained inside the intense slip band that traversed the width of the SCS-6 and
Trimarc fiber specimens. Fig. 7(c) shows a criss-cross arrangement of slip bands in the
entire gage section of the sample with SCS-0 fibers.
The tests show that the friction stress is much higher under fragmentation conditions than in a
push-out test, and this is believed to be due to substantial clamping of the fiber at the location
of the break. The microstructures also showed very little evidence of debonding. The high
friction stresses are a real effect, and indicate that shear stresses obtained from push-out tests
are not applicable when considering fiber failures in longitudinal loaded composites.
A more difficult item to resolve is the discrepancy in the Weibull parameters (see Table 1)
obtained from the fragmentation tests versus those obtained by testing extracted fibers,
although the difference is modest for the SCS-0 and Trimarc fibers. The worst agreement is
for the SCS-6 fibers, and a possible reason is damage to the fibers during testing. This is
discussed in more detail in [15] and is based on the SCS-6 coating structure [28] It may be
noted that the SCS-6 fiber has been found to exhibit strength degradation under fatigue crack
growth conditions [29-31], whereas Trimarc fibers do not show that behavior[29]. Thus,
SCS-6 fibers appear to be more prone to damage than the Trimarc fibers, and explanation may
lie in the interlayer next to the SiC surface and the profile of that surface.
As final part of the discussion, we shall attempt to rationalize the strengths obtained from the
40-fiber specimens. In a previous paper [11], we have shown the existence of LLS at RT in a
4-ply SCS-6/Ti-25Al-17V composite. The evidence was based on significant strength over-
prediction by Curtin's GLS model [6], and lack of random fiber breaks outside a thin 2 mm
band on either side of the fracture surface. It was found [11] that Zweben and Rosen's LLS
model [7], based on the occurrence of the second fiber break, provided excellent correlation
with the measured strengths. In this model, the critical event is the formation of a doublet,
where the second break is due to stress concentration from the first break.
In reference [11], a stress concentration factor of 1.146 was assumed, based on the elastic
calculations in [32]. On the other hand, the current observations suggest the scenario shown
in Figure 8, where slip bands emanating from the break allow the high modulus fibers to pick
up the load at four equivalent locations, making those locations vulnerable to fracture. Since
III - 246
Proceedings of ICCM–11, Gold Coast, Australia, 14th-18th July 1997
the load is equally transmitted along the slip bands, at most one fourth of the load in the center
fiber is picked up at any one location by the unbroken high modulus fiber. In other words, as a
first approximation we obtain a stress concentration factor of 1.25 at the possible fiber
fracture location, although this needs to be confirmed by rigorous analysis. The strength
prediction follows from the following equation for the fiber stress, σf [7,11]:
where δ is the shear lag distance (= σfd/4τ ), d is the fiber diameter, Lo is the length on which
σo is based, L is the gage length of the sample, Nt is the total number of fibers, and C is a
constant, given by C = 4. { km - 1}, where k is the stress concentration factor, and is assessed
to be 1.25 for the single-ply specimens. Substituting in σc=vf.σf + (1-vf).σmatrix, and noting
that σmatrix is the matrix yield stress for the specimens with SCS-6 and Trimarc fibers, the
composite strength, σc , can be obtained. Here vf is the volume fraction of fibers. In the
predictions, the friction stress was based on the fragmentation tests, and the Weibull
parameters corresponded to those of the extracted fibers. The predicted strengths were 1180
MPa and 1064 MPa for the SCS-6 and Trimarc composites, respectively, which show
excellent correlation with the corresponding measured values of 1160 MPa and 1076 MPa for
these two materials (see Table 2). The corresponding GLS predictions using Curtin's model
[6] are 1207 MPa and 1096 MPa, respectively. Although these numbers are also not very far
from the measured strengths, the lack of random fiber breaks away from the intense slip zone
suggests that the premise of the GLS model is violated in the current samples. In the case of
the SCS-0 fibers, the second fiber break model leads to LLS becoming operative at a fiber
stress of only 950 MPa (approx. 0.26%, using a fiber modulus of 370 GPa). Adding to this
strain the residual strain (approx. 0.24%) in the fiber, the critical condition is predicted at an
applied strain of 0.5 % strain, which also correlates extremely well with the experimental data
(Fig.6). The constant flow stress for the sample with SCS-0 fibers is not predicted by the
model, but the measured value is consistent with that of the yielded matrix modified by 11.6
percent of voided area.
Fig. 8. Sketch illustrating how a fiber break trasnmits load through local slip zones. The
intersection of the lobes with the adjacent fibers are potential fiber failure sites.
CONCLUSIONS
1. A methodology has been established for extracting the interface friction stress and the
insitu Weibull parameters from fragmentation tests. The friction stress is much higher under
fragmentation conditions than in push-out tests, and the microstructures showed very little
evidence of debonding.
III - 247
Volume III: Metal Matrix Composites and Physical Properties
ACKNOWLEDGMENTS
This research was performed at the USAF Wright Laboratory Materials Directorate, under
Contract Nos. F33615-96-C-5258 (BSM) and F-33615-94-C-5213 (TEM). The assistance of
T. Campbell in specimen preparation is sincerely acknowledged.
REFERENCES
2. M.J. Iremonger & W.G. Wood, J. Strain Anal., Vol. 5, pp.212-222 (1970)
3. A.S. Carara and F.J. McGarry, J. Compos. Materials, Vol.2, p.222 (1968)
4. T. Nicholas and J.A. Ahmad, Compos. Sc. and Tech., Vol.52, pp.29-38 (1994)
7. Zweben and B.W. Rosen, J. Mech. Phys. Solids, Vol.18, pp.189-206 (1970)
8. S.B. Batdorf and R. Ghaffarian, J. Reinforced Plastics, Vol.1, No.4, pp.165-176 (1982)
9. M.Y. He, A.G. Evans, and W.A. Curtin, Acta Metall., Vol.41, No.3, pp.871-878 (1993)
10. C.H. Weber, X. Chen, S.J. Connell, and F.W. Zok, Acta Met., 42, pp.3443-3450 (1994)
12. D.B. Gundel and F.E. Wawner, Compos. Science and Tech., in press (1996)
13. A.G. Metcalfe and M.J. Klein, Interfaces in Metal Matrix Composites, Ed. A.G.
Metcalfe, Academic Press, New York, pp.125-166 (1974)
III - 248
Proceedings of ICCM–11, Gold Coast, Australia, 14th-18th July 1997
14. M.G. Watson and T.W. Clyne, Composites, Vol.24, No.3, pp.222-228 (1993)
15. B.S. Majumdar, T.E. Matikas, & D.B. Miracle, J.Composites B:Engg., in press (1997)
16. B.S. Majumdar, S.G. Warrier, and D.B. Miracle, MRS Proc., Oct. 1996, in press (1997)
17. D.B. Gundel, B.S. Majumdar & D.B. Miracle, Scripta Metall., 33, p.2057 (1995)
18. D.B. Gundel, B.S. Majumdar & D.B. Miracle, Proc. ICCM-10, 2, Ed. A. Poursartip,
Woodhead Publishing Ltd. ,U.K., pp. 703-710, (1995)
19. S.G. Warrier, D.B. Gundel, B.S.Majumdar & D.B. Miracle, Scripta Met., 34, 293
(1995)
21. M.C. Waterbury, P. Karpur, T.E. Matikas, S. Krishnamurthy, and D.B. Miracle,
Composite Science and Technology, Vol. 52, pp. 261-266 (1994)
22. P. Karpur, T.E. Matikas, S. Krishnamurthy, and D.B. Miracle, Proc. ICCM-10, Ed. A.
Poursartip and K. Street, Woodhead Publishing Limited, Vol. V, pp. 479-486 (1995)
23. A. Kelly and W.R. Tyson, J. Mech. Phys. of Solids, Vol.13, pp.329-350 (1965)
24. L.T. Drzal, M.J. Rich, and P.F. Lloyd, J. Adhesion, Vol.16, pp.1-30 (1982)
25. B. Yavin, H.E. Gallis, J. Scherf, A. Eitan, & H.D. Wagner, Polym. Comp., 12, (1991)
27. C.J. Yang, S.M. Jeng, and J.M. Yang, Scripta Metall., Vol.24, pp.469-474 (1990)
28. X.J. Ning and P. Pirouz, J. Mater. Res., Vol.6, No. 10, pp.2234-2248 (1991)
29. S.G. Warrier and B.S. Majumdar, submitted to Mat. Sc. and Engg. (1997)
30. J.M. Larsen, J.R. Jira, R. John, & N.E. Ashbaugh, ASTM STP 1253 (1995)
31. P. Kantzos, J.I. Eldridge, D.A. Koss, & L.J. Ghosn, MRS Proc., 273, pp.325-330 (1992)
32. J.M. Hedgepeth & P. Van Dyke: J. Composite Materials, vol. 1, p. 294 (1967)
III - 249