Fracture Mechanics of Ceramics Volume 7 Composites, Impact, Statistics, and High-Temperature Phenomena
Volume 1 Concepts, Flaws, and Fractography Volume 2 Microstructure, Materials, and Applications Volume 3 Flaws and Testing Volume 4 Crack Growth and Microstructure Volume 5 Surface Flaws, Statistics, and Microcracking Volume 6 Measurements, Transformations, and High-Temperature
Fracture Volume 7 Composites, Impact, Statistics, and
High-Temperature Phenomena Volume 8 Microstructure, Methods, Design, and
Fatigue
Fracture Mechanics of Ceramics Volume 7 Composites, Impact, Statistics, and High-Temperature Phenomena
Edited by R. C. Bradt College of Engineering University of Washington Seattle, Washington
A.G.Evans Department of Materials and Mineral Engineering University of California Berkeley, California
n P. H. Hasselman Department of Materials Engineering Virginia Polytechnic Institute and State University Blacksburg, Virginia
and FFLange Rockwell International Science Center Thousand Oaks, California
PLENUM PRESS· NEW YORK- LONDON
ISBN 978-1-4615-7025-7 ISBN 978-1-4615-7023-3 (ebook) DOl 10.1007/978-1-4615-7023-3
Library of Congress Catalog Card Number 83-641076
First part of the proceedings of the Fourth International Symposium on the Fracture Mechanics of Ceramics, held June 19-21, 1985, at the Virginia Polytechnic Institute and State University, Blacksburg, Virginia
© 1986 Plenum Press, New York Softcover reprint of the hardcover I st edition 1986
A Division of Plenum Publishing Corporation 233 Spring Street, New York, N.Y. 10013
All rights reserved
No part of this book may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording, or otherwise, without written permission from the Publisher
PREFACE
These volumes, 7 and 8, of Fracture Mechanics of Ceramics constitute the proceedings of an international symposium on the fracture mechanics of ceramic materials held at Virginia Polytechnic Institute and State University, Blacksburg, Virginia on June 19, 20 and 21, 1985. These proceedings constitute the fourth pair of volumes of a continuing series of conferences.
The theme of this conference, as the previous three, focused on the mechanical behavior of ceramic materials in terms of the characteristics of cracks, particularly the roles which they assume in the fracture process. The 78 contributed papers by over 100 authors and co-authors represent the current state of the field. They address many of the theoretical and practical problems of interest to those concerned with brittle fracture.
The program chairmen gratefully acknowledge the financial assistance for the Symposium provided by the EXXON Foundation, the Army Research Office, the Natio~al Science Foundation, and the Office of Naval Research. Without their support, this conference simply would not have been possible. The suggestions of Drs. J. C. Hurt, R. C. Pohanka, and L. Toth were particularly helpful in assuring the success of this symposium.
Special appreciation is extended to Professor J. I. Robertson, C. P. Miles Professor of History. whose presentation following the banquet on the American Civil War was very well received by the audience.
Finally, we wish to also thank our joint secretaries, especially Karen Snider, for their patience and help in finally bringing these proceedings to press.
June, 1985
v
CONTENTS
D.B. Marshall and A.G. Evans
Applicability of Fracture Mechanics to Fiber-Reinforced CVD-Ceramic Composites
M. Gomina, J. L. Chermant, F. Osterstock G. Bernhart, and J. Mace
Fracture of SiC Fiber/Glass - Ceramic Composites As A Function of Temperature
R.L. Stewart, K. Chyung, M.P. Taylor and R.F. Cooper
Strength and Toughness Measurements of Ceramic Fiber Composites ••••
C.Cm Wu, D. Lewis and K.R. Mckinney
Toughening of Ceramics by Whisker Reinforcement P.F. Becher, T.N. Tiegs, J.C. Ogle and
W.H. Warwick
Fracture Behavior of Brittle Matrix, Particulate Composi tes with Thermal Expansion Mismatch
N. Miyata, S. Ichikawa, H. Monji and H. Jinno
High-Temperature Mechanical Properties
1
17
33
53
61
75
87
of A1Z0 3-SiC Composi tes • • • . • . . • • • • • " • • • •• 103 K. Nihara, A. Nakahira, T. Uchiyama
and T. Hirai
J. Homeny, D. Lewis, R.W. Rice and T. Garino
Graphical Methods for Determining the Nonlinear Fracture Parameters of Silica Graphite Refactory Composites .•••••••.
M. Sakai and R.C. Bradt lZ7
vii
Fiber - Matrix Bonding in Steel Fiber­ Reinforced Cement - Based Composites
R.J. Gray
Fracture Mechanics and Failure Processes in Polymer Modified and Blended Hydraulic Cements •••••••
J.E. Bailey, S. Chanda and N.B. Eden
Fracture of Brittle Rock Under Dynamic Loading Conditions • • • • •
E.D. Chen and L.M. Taylor
An Impact Damage Model of Ceramic Coatings B.M. Liaw, A.S. Kobayashi, A.F. Emery
and J.J. Du
M.M. Abou-el-leil
E. D. Case
On the Statistical Theory of Fracture Location Combined with Competing Risk Theo ry . . • • • • • • . . •
Y. Matsuo and K. Kitakami
The Use of Exploratory Data Analysis for the Saftey Evaluation of Structural Ceramics
F.E. Buresch and H. Meyer
Proof Testing to Assure Reliability of Structural Ceramics
T.R. Service and J.E. Ritter
Assessment of Flaws in Ceramics Materials on the Basis of Non - Destructive Evaluation • • • • .
D. Munz, O. Rosenfelder, K. Goebbels and H. Reiter
Mechanical Properties and Dependence Hith Temperature of Tetragonal Polycrystalline Zirconia Materials . • . • • • • • .
G. Orange and G. Fantozzi
Prediction of delayed Fracture from Crack Coalescence - Alemina .••.
T. Okada and G. Sines
Damage Accumulation in Rot Pressed Alumina During Flexural Creep and Anneals in Air . . • • . • • •
A.G. Robertson and D.S. Hilkinson
Creep Cavitation Behavior in Polycrystalline Ceramics
J. Lankford, K.S. Chan, and R.A. Page
viii
143
157
175
187
197
211
223
237
255
265
285
297
311
327
Matrix Representation of the Crack - Enhanced Creep of Ceramics Under Conditions of Multi - Axial Loading •••••••••
A. Venkateswaran and D.P.H. Hasselman
CONTRIB UTORS
CERAMIC FIBER COMPOSITES
D.B. Marshall Rockwell International Science Center Thousand Oaks, CA 91360
A.G. Evans Department of Materials Science University of California Santa Barbara, CA 93106
Recent developments in understanding failure mechanisms and in applying fracture mechanics to cerami.c fiber composites are reviewed. Direct observations of failure mechanisms in a uniaxially reinforced SiC/glass-ceramic composite are first summarized, thereby establishing a basis for a fracture mechanics analysis. The key observation is that frictional forces exerted by the fibers on the matrix oppose the opening of matrix cracks. The fracture mechanics analysis defines transitions between several failure mechanisms, provides strength/crack-size rela­ tions for each mechanism, and relates strength and/or toughness to micro­ structural properties of the composite. Implications of the results for designing composites with optimum properties are discussed.
1.0 INTRODUCTION
The resistance of brittle materials to tensile failure can be en­ hanced considerably by reinforcing with high strength fibers. The most dramatic improvements in properties have been achieved in composites that contain continuous unbonded fibers aligned parallel to the tensile axis. This class of composites includes glasses and glass-ceramics reinforced by carbon l - 4 and SiC fibers. 5- 7
Mechanisms of failure in these composites and in monolithic ceramics can differ substantially. Monolithic ceramics generally fail by the growth of a single crack on a plane normal to the maximum principal stress. Fiber composites, on the other hand, can fail by a variety of mechanisms, dependent upon the applied stress state and the geometry and microstructural characteristics of the composite. S Moreover, mechanisms that do not involve failure by growth of a single crack have been ob­ served. S In, that case fracture toughness cannot be defined in the usual sense.
Despite these complications, fracture mechanics can be applied to analyze failure of fiber composites, provided that the detailed mecha-
nisms of failure are identified for each combination of composite and stress state. Such analyses provide insight into failure processes and allow definition of alternative material properties which characterize the mechanical response. Furthermore, by relating these properties to microstructural parameters, the fracture mechanics analyses provide a means of designing optimum microstructures and anticipating microstruc­ tural changes that lead to changes in failure mechanisms. 1,9,10
The purpose of the present paper is to review recent progress8-11 in understanding failure mechanisms and in applying fracture mechanics to ceramic composites. Specifically, direct observations of the failure process in a composite material* consisting of approximately 50% uniaxi­ ally aligned SiC fibers in a lithium-alumino-silicate (LAS) glass-ceramic matrix are first described. These observations are then used as a basis for developing a fracture mechanics analysis which provides further in­ sight into the mechanics of failure as well as relating strength, tough­ ness, and changes in failure mechanism to microstructural properties.
2.0 FAILURE MECHANISMS
The general features of room-temperature load/deflection curves for flexure or tension tests in the SiC/LAS composite are shown in Fig. 1. 8 In both cases an initial linear elastic region is followed by nonlinear load increase to a maximum, then a continuous load decrease. The non­ catastrophic decrease in load gives these materials the appearance of being very "tough." Similar curves have been reported for flexure tests of a SiC/magnesium-alumino-silicate glass ceramic 7 and carbon-fiber/ glass-ceramic composites. 2
400 600
Fig. 1 Load-deflection curve for a SiC/glass ceramic composite.
Direct observation of the tensile surfaces during loading has allowed the damage processes corresponding to each portion of the load/ deflection curve to be identified. 8 In both flexure and tension tests
*United Technologies Research Center
2
the onset of nonlinear deflection occurs at a stress of ~ 300 MPa and coincides with the formation of a single matrix crack. In a tension test this crack passes completely through the central test section, and the applied load is supported entirely by the intact fibers bridging the crack. Further small increase of load causes formation of multiple regularly spaced cracks in the matrix throughout the central test area (Fig. 2). The spacing of the cracks is about 400~. Most of the addi­ tional deflection after the onset of matrix cracking is due to pullout of the fibers from the matrix and the associated increase in crack open­ ing. The peak load (~sOO MPa) is dictated by fiber failure. At deflec­ tions beyond the peak, the opening of one of the cracks becomes very large, and final separation involves the pulling of broken fibers through the blocks of matrix formed by multiple cracking.
Fig. 2 Tensile surface of a flexure specimen, loaded beyond the linear region of the load-deflection curve. Width of field 1.5 mm. Brightly reflecting regions are polished cross sections of fibers, grey regions are the matrix. After Ref. 8.
In a flexure test the matrix cracks that form at the onset of non­ linear deflection penetrate only to about the midplane of the beam. These cracks destroy the macroscopic uniformity of the beam causing re­ distribution of stresses. In particular the neutral axis moves towards the compressive surface, resulting in an enhancement of compressive stress. Further loading increases the opening of the cracks. However, the peak load is determined by a kinking instability on the compressive side of the beam. The importance of the stress redistribution was illus­ trated by a comparison of the compressive strength calculated from beam bending formulae (780 MPa) with the strength indicated by a strain gauge on the compressive surface (1330 MPa).8 Thus, the peak load-bearing capacity in flexure is dictated by a combination of tensile and compres­ sive failure mechanisms.
3
At higher temperatures (~ 1000°C) a change in failure mechanism has been observed. 11 Matrix cracking is accompanied by fiber fracture, re­ sulting in catastrophic failure of the composite and loss of the capacity to support high strains to failure. Fracture surface observations indi­ cate that the majority of fibers extend relatively short distances out of the surfaces, thus implying that fiber failure occurs behind the crack tip (fiber failure coincident with, or ahead of the crack tip would yield either no fiber pullout or equal numbers of protruding fibers and holes).
3.0 INFLUENCE OF FIBERS ON MATRIX CRACKING
The tensile stress at which the first matrix crack forms is an i~ portant characteristic of the composite. If all of the fibers bridging the matrix crack remain intact, matrix fracture signifies the onset of permanent damage, the loss of protection provided by the matrix against corrosion and oxidation of the fibers, and the likelihood of an enhanced susceptibility to degradation due to cyclic loading. On the other hand, if fiber failure accompanies matrix cracking, catastrophic failure ensues.
Separation of the surfaces of a matrix crack that is bridged by uni­ axially aligned reinforcing fibers requires some sliding of the matrix over the fibers. In general, this process entails debonding at the fiber/matrix interface followed by sliding against frictional forces. However, in composites that exhibit the failure mechanisms discussed in the previous section there is no chemical bond between the fibers and matrix. S A direct indication of the role of frictional forces in such composites was obtained from observations of matrix cracks during load cycling. Measurements of the crack opening displacements (Fig. 3) indi­ cated that the separations of the crack surfaces were larger during un­ loading than during loading. These observations imply that the fibers
6 I
51- • - t:;.~ /
E 4
t:;.~ A ::L I- - ci /{ z z w 3r- t:;. / -c. 0 e/ ~
~ (,J <t
/ t:;. UNLOADING
A RELOADING
0 0.005 0.01
STRAIN ON COMPRESSION SURFACE
Fig. 3 Plot of separation of crack surfaces in tensile surface of flexure beam during loading, unloading and reloading. After Ref. 8.
4
exert frictional forces on the matrix, which tend to oppose crack closure during unloading and resist crack opening during loading. Therefore, frictional forces must play an important role in inhibiting the initial extension of the first crack through the matrix.
4.0 FRACTURE MECHANICS ANALYSIS OF MATRIX CRACKING
4.1 Formulation of Problem
The influence of the fibers on the stress for matrix cracking can be evaluated using a stress intensity approach, in which the frictional forces that resist sliding are viewed as crack closure tractions. 9 The influence of these tractions is evaluated by imagining the crack to be formed in two steps. First, all of the bonds across the prospective crack plane (in the fibers as well as the matrix) are cut and stress cr= is applied (Fig. 4a), causing the crack to open. In the second step tractions, T, are applied to the end of each fiber that lies within a distance d of the crack tip. The magnitude of T is chosen so that the
1 0 00
r-
Fig. 4 Hypothetical steps used to evaluate the closure effect of fibers bridging a matrix crack.
fiber ends displace relative to the matrix and allow the fibers to be rejoined (Fig. 4b). In a continuum approximation (c » fiber spacing), this procedure is equivalent to applying a distribution of closing pressure p(x) to the crack surfaces:
p(x) T(x) f o
(x > c-d) (x < c-d)
(1)
5
where x represents the position on the crack surface (Fig. 4b) and f is the volume fraction of fibers. The closure induced by the pressure p(x) opposes the opening due to the applied stress croo• The influence of the applied stress on the crack tip stress intensity can be evaluated by re­ garding the stresses as a uniform opening pressure, croo' acting at the crack surfaces. Therefore, with the crack surfaces being subject to net pressure (croo-p(x)), a composite stress intensity factor can be defined as (for a penny crack* embedded in an infinite medium):
K 2(c/ 1f,) 1/2 / [croo-p(X) ]XdX
o I 1-X2 (2)
where X x/c.
The stress intensity K characterizes the composite stress and strain fields in the region immediately ahead of the matrix crack. In this re­ gion, the matrix and, fiber strains are expected to remain compatible, whereupon the stresses exhibit the usual composite relationship
cr /E = cr /E m m 00 (3a)
where crm is the matrix stress and E is the composite modulus,
with ~ and Ef referring to the Young's modulus of the matrix and fibers, respectively. The matrix and composite stress intensities scale with the stresses, so that
K = K (E/E ) m m
(3b)
where Km is the stress intensity factor in the matrix. The condition for equilibrium crack growth (in the absence of environmental effects) is given by setting ~ equal to the critical stress intensity factor, Ko' for the matrix. Therefore, the criterion for crack growth can be ex­ pressed in terms of K as;
Thus, Eqs. (2) and (4) relate the matrix cracking condition to the applied stress croo•
(4)
Evaluation of K in Eq. (2) requires a separate calculation of the pressure distribution p(x). Analysis of fiber pullout from the matrix 9 reveals that the closure pressure is related to the crack opening, u, at a given location by
(5)
where ~ = Ef f/Em(1-f), R is the fiber radius, and ~ is the sliding fric­ tional stress at the interface. However, th~ crack opening at a given position is determined by the entire distribution of surface tractions. For a penny crack, 13
*In the analysis for multiple matrix cracking, 9 penny cracks and straight cracks yielded almost identical results. Therefore, for convenience, only penny cracks are considered explicitly in this paper.
6
Is2- t 2 (6)
where sand t are normalized position coordinates and v is the Poisson's ratio of the composite. Therefore, analysis of matrix cracking by the stress intensity approach requires solution of Eqs. (5) and (6) to obtain the crack surface tractions, followed by evaluation of the integral in Eq. (2) and combination with the crack growth criterion, Eq. (4).
4.2 Closure Effect of Fibers
Rigorous solutions for u(X) from Eqs. (5) and (6) can only be ob­ tained numerically. However, an analytical solution that closely resemL bles the exact numerical result 9 can be obtained by assuming an approxi­ mate form for the crack profile. This solution has the attraction that the final result can be expressed in simple mathematical form. The ap­ proximate crack profile is taken as the solution of Eq. (6) for a crack subject to uniform pressure, with the magnitude of the opening governed by the net stress intensity factor K (Eq. 2);
(7)
The actual pressure distribution is obtained by combining Eqs. (5) and (7) to give
(x > c - d)
where
With this pressure distribution, the net stress intensity factor (Eq. (2» is given by
where
(8a)
(8b)
(9a)
(9b)
(9c)
and 0 = 2/~. The terms Kco and ~ represent the contributions to the crack tip stress intensity due to the applied load and the fiber closure tractions, respectively.
4.3 Multiple Matrix Cracking
If all of the fibers that intersect the crack plane remain intact the traction-induced stress intensity (Eq. (9c» becomes
(10)
Thus, the closure effect of the fibers increases indefinitely with crack length. The mechanics of crack growth is most conveniently investigated by combining Eqs. (9) and (10), setting K = K , and solving for 0'", to obtain an equilibriumLstress/crack-size function;
(11)
7
Thi.s function can be e~pressed conveniently in normal.1.zea rorm,
(1 / (1 -1/2 + (2/3)(c/c )1/4 (c < c ) == (l/3)(c/c ) c m m m 0
(12)
where
and 2 1/3
(1 = m (3/Q)(4cd<.c /9'Jt) (13b)
Equation (12) provides a relation between normalized stress and crack length parameters, (1c/am and c/Sn' without explicit reference to material and microstructural properties (these properties enter only in their in­ fluence on the normalizing factors cm and am). Thus, the mechanics of crack growth may be examined independently of the specific composite system.
Further progress requires that large and small cracks be distin­ guished. Large cracks must experience a crack opening which asymptotic­ ally approaches (but cannot exceed) the equilibrium separation of the completely failed matrix (i.e., two half planes connected by fibers). However, the crack opening expressed by Eq. (7) is unbounded at large c. Therefore, the preceding analysis is used only for cracks smaller than a transition crack length, co' defined by setting p = (100 at X = 0 in Eq. (Sa):
c o
2 2 «(1/aK) 00 c (14)
For larger cracks, the net force on the fibers that bridge the crack in the region of asymptotic opening (i.e., X < c-co) must balance the ap­ plied load. Consequently, the crack-tip stress concentration is induced exclusively over the length Co and the stress required to extend the crack must be independent of the total crack length. The resultant steady-state stress, given by Eq. (11) with c = co' is equal to am.
The complete equilibrium-stress/crack-size function is plotted in Fig. 5. Also plotted for comparison is a solution obtained by numerical integration. 9 It is noted that the stress required to propagate a matrix crack is almost independent of crack length for cracks larger than ~ cm/3. This defines the range of crack sizes over which steady-state conditions apply. The crack response in this region contrasts with the behavior of cracks in unreinforced brittle materials, for which the strength de­ creases with c- 1/2•
4.4 Fiber Failure Behind Crack Tip
When fibers fail behind the crack tip, 10 rigorous evaluation of Kn would involve consideration of the statistical nature of fiber strengths. However, in the present analysis, a single-valued fiber strength, S, is assumed.* Then the position within the crack at which fiber failure
*A single-valued fiber strength implies that fiber failure occurs between the crack surfaces, so that broken fibers do not exert closure forces on the crack. On the other hand, a statistical distribution of fiber strengths would allow fiber failure within the matrix and continued cIa­ ure effect until the broken fiber pulls out of the matrix. Therefore, the present calculations yield lower bound values of composite strengths.
B
I/) I/) w a;;
COMPOSITE: APPROXIMATE ANALYTICAL SOLUTION I I NUMERICAL SOLUTION
I I \ \ \ \ \ '\." MONOLITHIC ....... __ r MATRIX WITH
----- ___ -.l __ Ef = Em -------- O~----------_7------------~----------~ o 2 3
NORMALIZED CRACK LENGTH, CICm
Fig. 5 Equ±librium-stress/crack-size functions for penny-shaped matrix cracks in a composite containing high strength fibers and in a monolithic material. After Ref. 9.
occurs is defined
(15)
where K has been equated to Kc' Substitution of Eq. (15) into Eq. (9c) then yields
K = (413h) (Sf)3/ aK (16) p c
In this case, the closure effect of the fibers is manifest as a constant decrease in stress intensity factor (independent of crack length), so that the effect of the fibers is to increase the fracture toughness by &.C = Kp'
Strength/crack-size relations pertinent to this crack configuration can be conveniently compared with the results for the case where fibers do not fail behind the crack tip by normalizing the stresses and crack lengths with the parameters om and ~ defined in the previous section. The relative toughness increase becomes
!$.. /K = 2(Sf/ cr )3 c c m
and the strength/crack-size relation becomes (Eqs. (9) and (7))
(cr/crm) = [1 + 2(Sf/cr )3]/3(c/c )1/2 m m
(17)
(18)
Equation (18) is plotted in Fig. 6a for several values of the parameter Sf/om' The result from Section 4.3 for multiple matrix cracking (i.e., a fully bridged crack) is also shown. It is noted that, in these normal­ ized coordinates, the crack response is determined by the parameter Sf/crm, i.e., the relative magnitudes of the fiber strength and the steady-state matrix cracking stress.
9
C W N ::; <t :2 a: o z
2.----------.-----------r----------,-~-----.
Sf/urn = 1.0
OL-__________ L-__________ L-__________ L-______ ~
o 2 3 4
NORMALIZED CRACK LENGTH. clcm
Fig. 6a Strength/crack-size relations for cracks fully bridged by fibers (Eq. 12) and cracks with fiber failure occurring behind the crack tips (Eq. 18).
4.5 Influence of Initial Crack Configuration on Strength
The stength/crack-size relation defined by Eq. (18) for composites with Sf < am corresponds to a special crack configuration in which the trailing edge of the bridging zone is specified by fiber failure. More generally, the crack configuration at instability depends on both the size of the pre-existing matrix crack and the initial fiber bridging state associated with the crack. Insight into the influence of initial crack state on the strength can be obtained by considering two extreme configurations: an initially fully-bridged crack and a crack that initially has no bridging zone.
4.5.1 Initially Fully Bridged Crack
Matrix cracks that are initially fully bridged by fibers show three regions of behavior, depending on the size of the crack. For small cracks (i.e. c < d*) the crack opening is insufficient to cause fiber failure before the matrix crack becomes unstable at an applied stress given by Eq. (12). As the crack extends fiber failure occurs in the wake so that Eq. (12) also defines the strength of the composite. For inter­ mediate sized cracks, the opening at the crack mouth exceeds that re­ quired for fiber failure before the matrix crack extends. Then the equi­ librium bridging zone (i.e., d = d* defined by Eq. (15)) develops, and the strength of the composite is given by Eq. (18). The crack lengths at which this transition first occurs are given by setting c = d* in Eq. (15);
(19)
The transition crack lengths are also defined in Fig. 6a by the intersec­ tions of the strength curve for fully bridged cracks with the set of curves for cracks with broken fibers. For large cracks, the crack open­ ing approaches an asymptotic value (as discussed in Section 4.3) which is smaller than the opening defined by the approximate crack profile that
10
underlies Eq. (18). In this case the net force on the fibers in the asymptotic region balances the applied stress. Therefore, the crack re­ mains fully bridged until the applied stress exceeds Sf, whereupon fiber failure is followed by catastropic failure of the composite. The transi­ tion to this long crack limit occurs when the stress defined by Eq. (18) is (Sf. Strength/crack-size relations for these three regions of behavior for initially fully-bridged cracks are shown in Fig. 6b.
E tl
2r-------------.-------------~------------~_,
Fig. 6b Strength/crack-size relations for cracks that are initially fully bridged.
4.5.2 Initially Unbridged Crack
Matrix cracks that exist initially with a fiber bridging zone that is smaller than d* can extend stably with increasing applied stress prior to failure. This response is revealed by analyzing the growth of an ini­ tially unbridged crack of length Co which extends so that a fiber bridg­ ing zone of length d develops and the total crack length becomes Co + d (Fig. 7). The equilibrium-stress/crack-size function for this crack is obtained from Eqs. (9) and (13) with K = Kc;
a /a = (1/3) (c /c)1/2 {I + 2(d/c )3/4 [2 - (d/c )/(c/c )]3/4} (20) co m m m m m
where c = Co + d. The equilibrium stress is plotted as a function of the normalized crack extension, d/Cm, for various values of co/Cm in Fig. 7 (solid curves). The broken curves in Fig. 7 indicate the critical zone sizes d* for each value of co/cm and Sf/~, obtained by solving the nor­ malized form of Eq. (15);
4 (d*/c ) [2 - (d*/c )/(c/c )] = (Sf/a) m m m m (21)
The curves in Fig. 7 indicate that crack growth is always stable with in­ creasing applied stress for co/cm ~ 0.15 and d < d*; stable crack growth occurs until d = d*, whereupon fiber failure accompanies further matrix
11
r r I
o~_' __ ~ ________ ~~ ____________ ~~ ______ -" o 0.5 1.0
CRACK EXTENSION. diem
Fig. 7 Variation of equilibrium stress with extension of a partly-bridged crack (solid curves), for several values of the initial unbridged crack length co. Broken curves represent loci of the critical zone size, d*, for the onset of fiber failure at the end of the bridging zone.
crack extension, and failure is catastropic. The failure stress (defined by the intersection of the appropriate solid and broken curves in Fig. 7) is given by Eq. (18) with c = Co + d*. For small initial cracks (cohm < 0.15) instability of the matrix crack can occur at d < d*. In this case the strength of the composite exceeds the value given by Eq. (18).
The fracture response depicted in Fig. 7 can be characterized alter­ natively in terms of a crack-growth-resistance that increases with crack extension (R-curve).lO The R-curve is defined by Eq. (9) with KR = Koo at K = Kc ' and can be expressed in the normalized form
(22)
The R-curves for various values of co/cm are plotted in Fig. 8. Also plotted are the limiting toughnesses, obtained from Eq. (17), for several values of Sf/~. The intersections of these two sets of curves define the critical bridging zone size d*, for each combination of co/cm and Sf/om·
The condition for failure (i.e., unstable crack growth) is defined by Koo = KR and dKoo/dc = dKR/dc. Thus, the crack stability depends on the slope of the R-curve, which in turn is dictated by the initial unbridged crack length, co. For large cracks, stable growth occurs until d = d* and KR equals the limiting toughness. For smaller initial cracks,
12
~ o a: CI ::.: (.)
CRACK EXTENSION, dlCm
Fig. 8 Crack growth resistance curves for partly-bridged cracks. Horizontal lines represent the limiting toughnesses (i.e., the onset of fiber failure at the end of the bridging zone). After Ref. 10.
instability may be achieved at d < d*. Fully bridged matrix cracks exhibit instability without precursor stable growth.
5.0 DISCUSSION
5.1 Failure Mechanism in Frictionally Bonded Composites
The use of normalized strengths and crack lengths in the fracture mechanics analysis of Section 4.0 has enabled the mechanics and mecha­ nisms of crack growth to be examined independently of the specific ma­ terial and microstructural properties. The results of the analysis, summarized in Fig. 6, specify strength/crack-size relations for several failure mechanisms, as well as defining conditions for transitions between the mechanisms.
In composites containing fibers with sufficient strength to remain intact after a crack extends completely through the matrix (i.e., Sf > am)' the formation of periodic matrix cracks precedes failure of the com­ posite. Then, the tensile strength of the composite can substantially exceed the matrix-cracking stress and large strains-to-failure can be achieved (Fig. 1). Moreover, the stress for matrix cracking is indepen­ dent of pre-existing crack-size for cracks longer than a characteristic length. Under this condition the matrix cracking stress is an intrinsic property of the composite and is, therefore, both damage tolerant and independent of specimen size. Furthermore, it is noted that a fracture toughness cannot be defined with reference to either the matrix cracking event or the ultimate failure.
13
If the relative strength of the fibers is smaller (i.e., Sf < ~), failure of the composite coincides with matrix fracture and the strength of the composite becomes sensitive to pre-existing cracks. The failure response is dependent on both the size of the pre-existing matrix crack and the fiber bridging state associated with the crack. The response of cracks that are initially fully bridged by fibers is characterized by a fracture toughness which is enhanced by the reinforcing fibers. On the other hand, a crack that is initially unbridged (e.g., a notch) encoun­ ters an increasing resistance (R-curve) as the crack extends in the matrix and develops a bridging zone.
The reinforcing effect of the fibers can be inferred directly from Fig. 6. The lowest curve (i.e., Eq. (18) with S = 0) represents the product of the strength of the unreinforced matrix and the modulus ratio Ec/Em. The differences between this curve and the other solutions repre­ sent the reinforcing effects of the fibers for composites with equal fiber and matrix moduli (Ef = Em). For typical composites, the fiber modulus is the larger, and the matrix-cracking stress of the composite is always higher than the strength of the unreinforced matrix for a given crack length. For composites with Ef < Em' the matrix cracking stress is higher than the unreinforced strength within certain ranges of crack lengths and fiber strengths. Thus, it is evident that in general it is not necessary for the fiber modulus to be higher than that of the matrix in order to obtain reinforcement.
5.2 Microstructural Influences
The influence of microstructural properties on strength, toughness and transitions between failure mechanisms can be readily assessed by evaluating the normalizing parameters ~ and ~ (Eqs. 8 and 13):
crm (12(1 - v2)K~~Eff2(1 - f)(1 + n)2/EmR]I/3 (23)
3/2 2 2 2 2/3 c = (91]; /32) K E (1 - f) (1 + n) R/~f Ef(l - v ] mom (24)
For Sf > crm multiple matrix cracking occurs at the steady-state stress ao ~ 0.8 ~, provided the pre-existing flaws are larger than about cm/3. For the SiC/glass-ceramic composite, Ko = 2 MPa ml/2 , Ef = 200 GPa, ~ 85 GPa, f = 0.5, R = 8 ~, and ~ = 2 MPa. Substitution of these values into Eqs. (23) and (24) yields ~ = 313 ~ and cro = 265 MPa •. Thus, cm/3 represents several fiber spacings. Since the sizes of inherent flaws in ceramics are usually about the same as microstructural dimensions, this result implies that the condition for steady-state matrix-cracking will be generally satisfied for this composite. Moreover, the predicted stress, ao' is consistent with measured values of 290 ± 20 MPa. 8 More importantly, Eq. (23) provides a basis for design of optimum microstruc­ tures. The critical stress increases with the toughness of the matrix, the modulus and volume fraction of fibers, the frictional stress at the fiber/matrix interface, and decreasing fiber diameter.
The transition to the failure mechanism involving simultaneous fiber failure and matrix cracking is dictated by the relative values of Sf and am. If steady-state matrix cracking 'is desired, an increase in the volume fraction of fibers aligned in the principal stress axis benefici­ ally influences all of the parameters that determine optimum steady-state properties (i.e., cm decreases, while both am and Sf increase). However, the allowable increases in other parameters are limited. Increasing ~
increases om and decreases cm' but the maximum increase in ~ is limited by the fiber-failure stress. Increasing Ko increases ~ but also has the detrimental effect of increasing cm• Thus, the maximum acceptable Ko
14
could be dictated either by the fiber-failure stress or by the require­ ment that cm be less than a pre-existing flaw size. These restrictions account for the brittle response observed in a number of fiber or whisker-reinforced brittle systems, and place important bounds on the design of optimum microstructures.
For the failure mechanisms that involve simultaneous fiber failure and matrix cracking the limiting fracture toughness increase is
&. = S3fE R/6Ef 'tK (l + ... )(1 - i) c m 0 'I (2S)
and the range of crack lengths for which the limiting toughness applies (i.e., c ) d*) is defined by Eq. (21)
c > (1t/8) [S2E R/(l - i)K 'tEf{l + 1)]2 (26) m 0
It is interesting to note that the influences of all material parameters on &.c are opposite to their influence on the stress for steady-state multiple matrix cracking (i.e., am)' This arises because, for a given fiber strength (which does not influence am), the ratio Sf/~ decreases with increasing am thus leading to a smaller fiber-bridging zone (Eq. 21) and a decreased toughness increment (Eq. 17).
ACKNOWLEDGEMENT
Funding for this work was supplied by the u.s. Office of Naval Research, Contract numbers NOOOI4-8S-C-0416 and N00014-79-C-OlS9.
REFERENCES
1. J. Aveston, G.A. Cooper, and A. Kelly, pp. IS-26 in the Properties of Fiber Composites, Conf. Proc. Nat. Physical Lab., IPC Science and Technology Pres Ltd., Surrey, England, 1971.
2. R.A.J. Sambell, A. Briggs, D.C. Phillips, and D.H. Bowen, J. Mater. Sci. 7[6], 676-681 (1972).
3. D.C. Phillips, J. Mater. Sci. 9[11], 1847-S4 (1974). 4. D.C. Phillips, J. Mater. Sci. 7[10] 117S-91 (1972). S. K.M. Prewo and J.J. Brennan, J:-Mater. Sci. IS[2] 463-8 (1980). 6 K.M. Prewo and J.J. Brennan, J. Mater. Sci. ][7[4] 1201-6 (1982). 7. J.J. Brennan and K.M. Prewo, J. Mater. Sci. lr7[8] 2371-83 (1982). 8. D.B. Marshall and A.G. Evans, J. Amer. Ceram:--Soc. 68[S] 22S-31
(198S). - 9. D.B. Marshall, B.N. Cox and A.G. Evans, Acta. Met., in press.
10. D.B. Marshall and A.G. Evans, in proceedings of the Fifth Interna­ tional Conference on Composite Materials.
11. A.G. Evans, M.D. Thouless, D.B. Johnson-Walls, E. Luh, and D.B. Marshall, in proceedings of Fifth International Conference on Composite Materials.
12. G.C. Sih, Handbook of Stress Intensity Factors, Lehigh University, Bethlehem, Pennsylvania, 1973.
13. I.N. Sneddon and M. Lowengrub, "Crack Problems in the Classical Theory of Elasticity," Wiley, New York, 1969.
15
COMPOSITES
ABSTRACT
M. Gomina*. J.L. Chermant*. F. Osterstock*. G. Bernhart** and J. Mace*~'
*Equipe Materiaux-Microstructure du LA 251,ISMRa-Universite 14032 CAEN Cedex, France **SEP, Etablissement de Bordeaux, Le Haillan, B.P. 37, 33165 St Medaed en Jalles. France
Investigation of mechanical behavior of C-SiC and SiC-SiC composite ceramic materials, using SENB and CT specimens at room temperature, is discussed according to the relative arrangements of the fibers, the orientation of the layers and the applied stress.
R, G and J-curves have been plotted. The values depend on the orien­ tation of the layers prior to the applied stress. Different mechanisms are involved. For CT specimens, crazing has been observed, showing the importance of the multidirectionality of these materials. G and J values corresponding to those of good ceramic materials were obtained for inter­ laminar crack propagation and higher values of these parameters for the other crack propagation orientations.
INTRODUCTION
The use of fibrous reinforced ceramic materials instead of monoli­ thic ceramics in structures for high technology (aeronautics, arms, cars and sports factories, medicine, ••• ) is advantageous: it allows higher mechanical and thermal performances, weight-saving, parts all in one piece, biocompatibility, and often an increase of the energetic efficiency. But their utilization is hindered by the knowledge still incomplete of their thermomechanical properties and mechanisms involved in their degradation under different sollicitations and environments. Until now their mechanical investigations were only concerned with the application of the linear elastic fracture mechanics (LEFM) (1)(2). The concepts of critical stress intensity factor, K, and critical strain energy release rate, G (3), were proposed for tsotropic and homogeneous materials with brittleCfracture. Further, the concepts of R-curve and of J-integral have been introduced to account for the anelastic character of the rupture of some of these materials (4). But the basical hypothesa of the LEFM are not always compatible with very anisotropic structure of the fiber composites.
The aim of this paper deals with the application of the R, G, J and K concepts of the LEFM to the rupture of C-SiC and SiC-SiC materials. An
17
~"31
.'", (1,2)
Fig. 1. Schematic of the test specimens used.
empirical method is proposed to measure the crack length in any point of the load-displacement curve, allowing calculation of the rupture parame­ ters in the case when, an anelastic rupture of the material occurs.
MATERIALS AND EXPERIMENTAL PROCEDURES
Materials
The materials are made of carbon fibers embedded in a matrix of silicon carbide (C-SiC) and of silicon carbide fibers in a matrix of silicon carbide (SiC-SiC), manufactured by Societe Europeenne de Propul­ sion (S.E.P., Bordeaux). Carbon fibers are arranged in bundles and woven according to a bidirectional arrangement. These woven cloths are pilled up and the rema1n1ng pores are closed by a CVD, process (5). SiC-SiC materials were prepared from SiC fibers produced by pyrolisis of carbosi­ lane precursors.
Specimens of different orientations have been machined according to two orientations, noted (1,3) and (3,1) for bending specimens, and one orientation, noted (1,2), for CT specimens (Fig. 1). The orientations of the layers prior to the applied stress have been also reported on this figure.
Fig. 2. Optical micrograph of a C-SiC composite.
18
Experimental procedure
The specimens of sizes L = 50 mm, B = 10 mm and W = 5 mm, were notch~d with diamond saws of 0.1 mm, 0.3 mm and 0.5 mm thick. The three point bending specimens have been tested in the configuration L/W = 4, using an Instron 1185 apparatus with a cross head speed of 0.1 mm/mn. The displacement of the loading point was given by an inductive gage Schae­ vitz 200 DC-D with an amplification system allowing displacements of 0.001 mm to be measured.
For CT specimens (of size 24x25xl0 mm3 ) the crack opening displace­ ment, v, has been converted to a load line displacement, 6 , using a computer. The crack extension during the test was monitored using the partial unloading compliance method (at the level of approximately 20% of the last maximum value).
The CT specimens have been tested with an Instron 1165 apparatus, with a cross head speed of 0.5 mm/mn.
The density of the C-SiC and SiC-SiC materials are similar (of the order of 2.4).
Measurement methods
For SENB specimens with orientation (3,1) we have observed that the load-unloading curves show a singularity : the tangents to the increasing part of the loops from the unloading points meet always at a same point, M. This point is located just below the zero value of the displacement (Fig. 3). The compliance associated to any point of the load displacement curve is given by the slope of the line joining this point to the fixed point M.
1 Q.
C-SiC (3,1)
MP' J
Fig. 3. Experimental load-displacement curve with several load­ unloading loops.
19
For CT specimens it has been verified for materials with similar densities, that the presence of the notch can be described by the classi­ cal analytical expression, C = f(a/W), given for isotropic materials (6). The notch extension, ~ a, was determined from the compliance, measured during successive unloadings. As pointed out by J.A. Clarke (7) the compliance data are not used to calculate crack lengths but rather rela­ tive crack extension values. Between two successive unloadings, the crack extension, ~a, is given by :
where
~C (W - a) (a ~a = -C 2 g W) [1]
For SENB and CT specimens the J-curves were determined using the classical expression
with A B W-a
A J = TJ B(W _ a)
the area under the load-displacement curve, the thickness of the specimen, the remaining ligament length,
[2J
a corrective factor given by J.G. Merkle and H.J. Corten (8). TJ = 2 for 3 point bending and 2 (1 + 0.261 (W-a)/W) for CT specimen.
In the calculation of the areas, we do not substract the area due to the elastic contribution, because it has been shown (9) that for crack lengths (a/W) ~ 0.5, the elastic contribution is small. It is then possi­ ble to perform better J values over a large crack extension in taking into account the total area monitored during the test.
The determination of the R-curves was performed using the classical expression :
p.2 dC Ri = 2~ da (a i )
with P. 1.
C-SiC (1,3)
20
RESULTS
Morphological investigations by image analysis have shown that the largest pores are located preferentially at the intersection of the fibers bundles and that the small pores are between the fibers in the bundles (10).
We have tried to use fracture mechanics to describe the rupture behavior of these materials according to the three orientations.
C-SiC, orientation (1,3)
A typical load-displacement curve is given on figure 4 for C-SiC (1,3). The deviation from the linearity before the maximum loading point indicates a crack growth, corroborated by the presence of a residual deformation observed during the test.
The values of the compliance of the notched specimens were deter­ mined from the linear portion of the load-displacement curves. The rela­ tionship between the compliance C and the relative crack length, a/W, is reported on figure 5 with the theoretical curve C = f(a!W) obtained from the equations proposed by A.V. Virkar and R.S. Gordon (6). From the cali­ bration curve of figure 5, an effective crack length, a ff' was determi­ ned. This crack length value reflects the damaging of Ene material at that stage of the test.
The elastic strain energy release rate was calculated from the expression :
G = E
with K max
P L 3 max - ----- for a three-point bending test, 2 BW2
Pmax the maximum applied load, v the Poisson coefficient, E the elastic modulus.
t C-SiC (1,3) i Z 0.8 -- 500 I'm ~ ---- 300 I'm
E 100 I'm -- tho
[4]
Fig. 5. Change in the compliance, C, as a function of the a/W ratio of specimens notched with diamond saws of diffe­ rent thickness. Theoretical curve calculated from the equations proposed by A.V. Virkar and R.S. Gordon (1975) is also plotted.
21
The change of the two parameters Rand G as a function of the increase of the crack length is plotted on figure 6. According to the crack extension in a specimen by application of LEFM, the crack insta­ bility occurs when the elastic energy release rate reaches the resistance of the material to the crack extension :
R = G
aR < aG aa aA [5J
But one can observe on figure 6 that beyond the intersection point of R and G-curves, the R-curve remains always higher than the G-curve. This indicates that it is not possible to define a G value to characterize
c
22
Table I. Critical values of the fracture parameters calculated by application of linear fracture mechanics as fonc­ tion of the initial crack lengths, for C-SiC (1,3) composites. a corresponds to the maximum value of the load, P , rof the first maximum (see Fig. 3)
max
t I
C-SiC (1,3)
I
590
6 8mm-'
Fig. 6. Loading curves, G (in dotted line), and R-curves (in solid line) for two specimens with initial crack length a = 2.6 mm and 4.7 mm. They have been calculated for tRe maximum value of the tension stress, Pmax ' on the outside fiber.
the crack instability, consistant with the LEFM hypothesa, by a measure­ ment of the compliance from the origin. The values of the critical stress intensity factor calculated from the G values using relationship [3] are markedly dependent on the initial crac~ length (Table I).
C-SiC, orientation (3,1)
The load-displacement curve for C-SiC (3,1), exhibits a linear portion close to the maximum load led extension of the crack from this maximum point.
shown on figure 7, point and a control-
But, as in the case of specimens (1,3), crack extension appears before the maximum load. One can also observe the presence of a residual deformation during unloading.
t z C-SiC (3.1) ~
t ":' C-SiC (3.1) 52 il20 ~ (.) iI w
15
10
5
o 0.2 0.4 0.6a/w ......
Fig. 8. Experimental change of E*C(a) as a function of the relative notch depth a/W.
23
On figure B, we have plotted the quantity E*C(a) (which represents the product of the elastic modulus with the value of the compliance) a~ a function of the relative crack length, a/W. This is the calibration curve to obtain the a ff values. This plot allows to free from the change in the elastic modulUs of the specimens.
In three point bending test, E*C(a) is given by :
L 2 1 [9 2 i a /w 2 ] E * C(a) = (W) B 1 + '2 (l - \I ) 0 Y XdX [6]
with X = a/W.
The measurement of E*C(a) values allows the determination of the po­ lynomial Y(a/W) introduced in the LEFM relationships to take into account the size effect of the crack on the finite dimensions on the test speci­ mens (y = JK for a/W - 0) :
y(!.)=~[_l_ d(E*C) 2B2 ]1/2 [7] W L a/W d(a/W) 9(1-\1 )
The variation of the experimental polynomial Y(a/W) as a function of the relative crack length a/W is reported on figure 9, with that of the theoretical expression of the polynomial proposed by B. Gross and J.E. Srawley (11). The fitted analytical expression of the experimental Y(a/W) values is :
Y(~) = B.IB - 53(~) + 510.B(~)2 - 157.4(~)3 + 57.3(~)4 [B]
for 0.25 ~ a/W~ 0.6.
0 0.1 0.2 0.3 0.4 0.5 a/w .....
Fig. 9. Change in the Y polynomial as a function of the relative notch depth, a/W. Curve in solid line corresponds to the experimental curve for the composite and that in dotted line corresponds to the theoretical curve for monolithic ceramics.
ao w
Table II.
(.!) w Fmax
0.37 0.45 0.50 0.50 0.53
Values of the critical strain energy release rate (G = R ) and the critical stress intensity factor, K ; forme~SiC (3,1). a and a are respectively tfie initial crack lengtR and tK~a~ffective crack length at the maximum load value.
(.!) w Rmax
F max N
FRmax N
Rmax
J/m2
77 90 64 64 65
2.3 2.5 2.0 2.0 2.0
The experimental Y(a/W) values lead to stress intensity factor values, K = aY.Ja, higher than those given by the classical Y(a/W) polyno- mial.
On figure 10 we have reported the Rand G-curves as a function of the crack extension. The critical values, G and K, calculated for the maximum load and at the tangent point of th~ R andc G-curves are reported on Table II. In this case the curves obtained from two initial crack lengths exhibit the same shape, which was not the case for (1,3) compo­ sites.
The presence of a residual deformation justifies the rupture ap­ proach of these composites by the J-integral method. This parameter describes the strain energy close to the crack tip even when anelastic phenomena are largely present (12)(13). The calculation method of the J = ,f(8a) ,curve (14) necessitates the measurement of the total energy (area, A, under the load displacement curve P(h)) dissipated to perform an extension 8a of the crack. The energy release rate to create this crack was calculated from the expression [2].
t '1 ~ 75 .,
0.1 0.2 0.3 0.4 0.5~
Fig. 10. Rand G-curves as a function of the notch for C-SiC (3,1) composites (R in solid dotted line).
depth, a/W, line, G in
25
o 0.2 0.4 0.6 0.8 4amm ....
Fig. 11. Change in J-curve as a function of the crack displace­ ment, Aa, for three ,values of notch depth, a/W, for C­ SiC (1,2) composite.
The J = f(Aa) plot is in fact a crack resistance curve. The critical value, J, is given by the intersection of this curve with the crack bluntingCline. In this investigation we have measured the crack extension from the compliance values, using our proposed empirical method.
The change of J as a function of the crack extension, Aa, is shown on-figure 11, for three initial relative crac~2 lengths. The J IC values obtained are located between 40 and 50 J.m
C-SiC, orientation (1,2)
The orientation (1,2) was only tested in the CT configuration. An illustration of Rand J-curves is shown on figures 12 and 13 for a specimen wit~2 a/W = 0.6. The initiation energy (Fig. 12) is the same - 40 to 60 J.m - than those measured for the C-SiC with the orientation 0,1).
t N
26
20000
15000
10000
C-SiC (1,2)
Fig. 13. J-curve for C-SiC (1,2) composite.
For the metallic materials (Fig. 14a) the fracture process asso­ ciated with the shape of the J-curve has been described by J.D. Landes and J.A. Begley (13). J 1 corresponds to the point of the first crack extension. The first linea~ part of the J curve is the blunting line (J = 20 Aa). an
For ceramic composite materials, the machined notch initiates a process zone of matrix microcracking. When the load increases, multiple fracture of the matrix occurs until the first stage of fiber fracture arises. When the maximum load occurs, the slope of the J-curve changes and becomes less pronounced. This point of the J-curve can be related to the propagation of the primary macroscopic crack (matrix crazing corres­ ponds to secondary cracks) (Fig. l4b).
We can propose an energetic microcracking law in the form
t Log J = n Log Aa + c [9]
to describe the first part of the curve. The coefficient n is a material characteristic .which describes the capability of the composite to develop microcracks. It qualifies physically the multidirectionality of the frac­ ture process : - n 1 for metallic materials (straight crack), - n > 1 for materials exhibiting multiple cracking process at the crack
tip.
The transition point on figure 14b, denoted J 1T , between the zone of predominant matrix microcracking process and the predominant fiber frac­ ture process, is not a composite material characteristic. For specimen~
with a/W = 0.6, values of n are 2.3 and J lT value is of 5000 J.m -2 (Fig. 13).
Regarding the R-curves (Fig. the energy necessary to propagate propagation is always c~2trolled release rate is 2200 J.m • The K then 15 MPa.Jffi. c
12), when the maximum load is reached, the crack decreases. However the crack during the test. The critical energy value calculated from this G value is
c
27
crack blunting
first loading
fatigue crack
'-=(~ \~\~
machined crack Ji~ ________________________ ___
b
Fig. 14. Schematic of J-curve J.D. Landes and J.A. ceramic materials.
68'"
SiC-SiC, orientation (1,2)
SiC-SiC materials present high mechanical characteristics and others properties the C-SiC materials can not afford (15). The same parameters have been measured for SiC-SiC specimens with the orientation (1,2) as for C-SiC specimens with the same orientation. A typical curve for an initial crack length a/W =20.6 is shown on figure 15. The initiation energy is J 1i_2 250 J.m and the energy at the transition point is J 1T = 17500 J.m • The value of the parameter n is 2.1.
To observe whether the R curve is an intrinsic characteristic of the material, specimens with 4 different initial crack lengths (a/W = 0.3, 0.4, 0.5 and 0.6) were tested. The corresponding R curves are shown on figure 16. The mean curve is closed to the one with a/W = 0.4. This mean curve have then been used to determine critical G values for different crack lengths (Table III). c
28
Fig. 15. J-curve for SiC-SiC (1,2) composite.
t ~ E ...,
SiC-SiC (1.2)
6 6a mm ....
Fig. 16. R-Curve for SiC-SiC (1,2) composite of different rela­ tive notch depths, a/W.
a W
0.3 0.4 0.5 0.6
Table III. Experimental values for SiC-SiC orientation (1,2) specimens with 4 different initial crack length, a:
E GPa
180 156 149 166
E, elastic modulus ; P . : calculated instability load ; P : experimental·measured load G : cri­ tical stf~~n energy release rate; K gritical
. . f c stress ~ntens~ty actor.
7000 5600 3800 2300
29
DISCUSSIONS
For small size specimens (50x10x5mm3 ) of C-SiC (1,3), a non-coplanar crack extension is observed (Fig. 17a).
This is as a consequence of the coarse microstructure of these mate­ rials. The laminates are approximately 0.75 mm thick and about only ten laminates are present in the section of a test specimen. This is not enough to allow the full development of a process zone at the front of the crack tip. As for example, in the case of alumina, the remaining ligament contains thousands of grains.
Proceeding with larger specimens (100x20x1Omm3 ), we have observed that the crack path remains in the plane of the initial crack, though it is still zigzagging (Fig. 17b). But we still observe that the primary crack extends along a plane inclined to the original crack plane. This shows that the initial straight crack is really under a combined mode I and mode II loading for C-SiC (1,3).
M. Ichikawa and S. Tanaka (16) proposed an expression of the energy release rate G for initiation of an infinitesimal kink developing at the tip of the principal crack. Calculations involving the deviation angle and_ 2 the elastic constants of the C-SiC (1,3) materials lead to G = 3000 J.m •
The crack path on the surface of the C-SiC orientation (3,1) meets preferentially the large pores located not far from the initial crack's plane. Even when the crack tip lies in a region rich in SiC, the devia­ tion of the crack by the pores or by the fibers leads it in the bundles of fibers. It is the effects of the longitudinal fibers which modify the elastic conditions of the material. They act as a very effective stress concentrator parameter. This explains the high Y values measured. The calculation of the Y polynomial ,from the detailed analysis of the com­ pliance is then a realistic calibration coefficient for the determination of the stress intensity factor, K.
The R-curve plot is linked to the measurement of the crack exten­ sion. The empirical method we proposed to determine the compliance at each point on the P(h) curve leads to G values (40 to 60 J.m-2 ) compati­ ble with those measured similarly usin~ the J-integral method. It is to be noted that same values are obtained for C-SiC materials with the two orientations (1,2) and (3,1) : this corresponds, in fact, to the fracture of the SiC matrix. These G values result from the anelastic rupture of the material treated by c the LEFM method, but including a shift in the load value. Comparison of the J values (40 - 50 J.m-2 ) to the Rand RF values shows that the ru~ture is initiated at the point of max non­ l~W~~rity apparition on the P(h) curve. But the material stands higher loads. The instability of the crack occurs for the maximum load: that is
30
Fig. 17. Macrographies of tested spe3imens C-SiC (1,3) 30f two different size: 50x10x5 mm and 100x20x10 mm •
the point of failure of the material. It corresponds to a K value of 1.8 ~a~. c
For C-SiC orientations (1,3), the specimens sizes are critical parameters. The coarse microstructure of the materials did not allow a stable extension of the crack. The size of the damage zone was very often not negligible compared to that of the remaining ligament.
Results corresponding to C-SiC and SiC-SiC CT specimens indicate that the use of Rand J-curves are complementary for orientation (1,2).
The R-curves allow the prediction of the instability stresses in a part. Th~2e instability stresses correspond to _2ifferent energy values : 2200 J.m for the C-SiC material and 10000 J.m for the SiC-SiC mate­ rials.
The J-curves, essentially due to the determination of the multidi­ rectional propagation coefficients, indicate that the damage, prior to the fiber rupture, rises similarly:
- n = 2.3 for C-SiC materials, - n = 2.1 for SiC-SiC materials.
The difference is due to the behaviour of the woven cloths used - high mechanical properties fibers for the SiC-SiC cloth on the contrary of the thermal type cloth of C-SiC - which constrains much more the damage because the monofilaments support better the load transfert from the matrix microcracking.
The damage in the two kinds of materials is due, as shown by the SEM observations, to the same elementary processes:
rupture in the matrix parallely to the longitudinal fibers in the sheath of matrix around the fibers,
- matrix-fiber decohesion, - rupture of the fibers and work loose.
ACKNOWLEDGMENTS
This work has been performed under a contract MIR-CNRS-DRET, AlP-ASP W19.84.54.
REFERENCES
1. Randolph J.F., Philips D.C., Beaumont P.W.R., Tetelman A.S., J. Mat. Sci., 7, 289 (1972).
2. Marom---C;:: White E.F., J. Mat. Sci., 7 : 1299 (1972). 3. Irwin G.R., Tra~s. ASM-,-40A : 147 (1948). 4. Rice J.R., J. Appl.JMech., Trans. ASME, 35 : 379 (1968). 5. Christin F.:- Naslain R., Bernard~ Proc. 7th Int. Conf. CVD,
edited by Sedwick T.O. and Lydtin M., The ElectroCheffiicar-soCie- ty, Princeton, p. 499 (1979). ---
6. Virkar A.V., Gordon R.S., J. Amer. Ceram. Soc., 59 : 68 (1976). 7. CLarke J.A., ASTM - STP N°743:-P:-553 (198~ 8. Merkle J.G., Corten R.J., Trans ASME, Nov: 286 (1974). 9. Underwood J.R., ASTM - STP n0601~ 312 (1976). 10. Gomina M., Chermant J.L., Coster M., Acta Stereol., 2 179
(1983) •
31
32
11. Gross B., Srawley J .E., "Stress intensity factors for single­ edge-notch specimens in bending of combined bending and tension by boundary collocation of a stress function", Tech. Note NASA­ TN-D-2603, (Janv. 1965).
12. Begley J.A., Landes J.D., ASTM - STP N°514, p. 24 (1972). 13. Landes J.D., Begley J.A., ASTM - STP N°632, p. 57 (1977). 14. Rice J.R., Paris P.C., Merkle J.G., ASTM - STP n0536, p. 231
(1973) • 15. Dauchier M., Bernhart G., Bonnet C., 30th National SAMPE Symp.,
March 19-21, p. 1519 (1985). 16. Ichikawa M., Tanaka S., Int. J. Fract., 22, 125 (1983). 17. Gomina M., Chermant J.L-.-,--Osterstock F., Proc. lInd Conference
on "Creep and Fracture of Engineering Materials and Structures", Edited by Wilshire B. and Owen D.R.J., Pinebridge Press, Vol. I, p. 541 (1984).
FRACTURE OF SiC FIBER/GLASS-CERAMIC COMPOSITES
AS A FUNCTION OF TEMPERATURE
ABSTRACT
R.L. Stewart, K. Chyung, M.P. Taylor, and R.F. Cooper
Corning Glass Works Sullivan Park, Rand D Division Corning, NY 14831
SiC fiber reinforced LAS and BMAS glass-ceramic matrix composites have been recently developed. They show a combination of high strength and exceptional toughness over a wide temperature range. However, they suffer from significant embrittlement over the temperature range of 800 o -1100°C when tested in air. The objective of the work was to understand this embrittlement behavior.
INTRODUCTION
The development of melt spun polycarbosiline derived fibers with high strength and small diameter by Yajima and co-workers(I-3) has offered the possibility of high temperature fiber reinforced ceramic composites with good strength and greatly superior fracture toughness compared to mono­ lithic ceramics. Fibers, now produced and sold under the trade name Nicalon® by Nippon Carbon Company, can be obtained in either a completely amorphous mixture of SiC, Si02 and free carbon, or as a microcrystalline S-SiC with Si02 and free carbon. For the remainder of this report, the term "SiC fib~r" will refer to the lower oxygen content version of the mi crocrysta 11 i ne S-Si C, Si Or C mi xture that has become known as lice rami c grade". Simon and Bunsell( ) have recently described the structure and chemistry of this fiber. They also reported that the microcrystalline fiber maintained its strength at high temperatures better than the amor­ phous form. The chemistry of this fiber was given as 49, 40 and 11 mole
33
percent of SiC, C, and Si02, respectively, and the structure was described as 1.7 nm S-SiC grains distributed evenly across the fiber cross-section with Si02 and C. Carbon grain radii were centered at 1.5 and 2.2 nm.
The lithium aluminosilicate matrix used for this work is the LAS-III composition developed jointly for SiC fiber composites by Corning and UTRC. The composition and phases of consolidated and cerammed composites has been reported by Brennan(5). The matrix phases in well crystallized composites include S-spodumene SS, mullite, and NbC. The NbC has been shown by Brennan(6) and Bender et al.(7) to surround the SiC fibers at the matrix/ fiber boundary with a discontinuous appearing grainy structure. This NbC layer is thought to form from a reaction between the Nb205 in the matrix precursor glass and the free carbon in the fibers during the hot-pressing consolidation.
We have been developing SiC fiber reinforced glass-ceramics using more refractory compositions in the magnesium aluminosilicate family. One is based on the double ring silicate osumilite which is structurally similar to cordierite. Like cordierite, osumilite may be stuffed with various cations which provide charge compensation for the replacements of A1 3+ for Si 4+ stoichiometry in the framework. We have been developing matrices based on Ba2+ stuffing, known as barium osumilite which when "fully stuffed" has the composition BaMg2A16Si9030. The composition reported on here is designated BMAS-III. The improved refractoriness ve~us LAS-III is illustrated by bending beam viscosity data using the temperature at a vis­ cosity of 1013 poises. These temperatures are 1089°C, for LAS-III matrix and 1190°C for the BMAS-III matrix, indicating about a 100°C higher use temperature for BMAS-III.
Our composite fabrication procedure involves melting the appropriate glass, grinding and sizing it, then preparing a slurry mixture with an organic binder. The fiber tows are pulled from the spool through a flame to burn off a protective sizing, and then are pulled through the glass pow­ der slurry and wound onto a flat sided drum where it is dried. These glass-impregnated fiber mats are cut to shape, stacked up and then burned out to remove the binder. The prepregged stacks are then loaded into graphite molds and hot-pressed to achieve consolidation. After pressing, a crystallization treatment is followed to raise the refractoriness of the matrices.
34
Because we have been in the materials development mode to this point, we have limited our composite evaluations to X-ray diffraction phase iden­ tification and flexural testing as a function of temperature (25-1250°C) in air, with subsequent optical, scanning electron and transmission electron" microscopy. Included in all the flexural evaluations is an estimate of the
matrix microcracking stress and strain as well as the highest stress sup­ ported and its associated strain (all calculated using standard linear elastic bending formulae). The flexural testing also involves a visual inspection for the failure mode (tension, compression or shear), and char­
acterization of the rupture surfaces in terms of fibrous, wood-like or brittle appearances. This visual inspection plus the area under the stress-strain curve is used to compare toughness.
During the high temperature flexure testing in air it was noted that the SiC/LAS-III composites that had demonstrated high ultimate stress and toughness at room temperature, began to show a significant loss of these properties in the 800-I100°C temperature range. Accompanying this degrada­ tion in a I5-20-minute time span was a smoothing of the failure surface morphology and change in failure mode. Composites that had failed in com­
pression at 25°C with fibrous morphologies, displayed a shear or tensile phenomenon failure mode with woody to brittle appearing features. The same
was also observed in SiC/LAS-III composites by Prewo(8) at UTRC. Later we discovered that SiC/BMAS-III composites are also subject to degradation in air in the same temperature range. Mah(9) has also documented this be­
havior with both a Ba-osumilite and Ba-cordierite matrix.
Tests conducted in inert environments reported by prewo(8), and in vacuum by Mah(9) implied that exposure to high temperature air was respon­
sible for the degradation, because under these non-air environments no loss of strength or toughness and fibrosity occurred. A recent study by Mah(lO)
has shown that oxygen is the active specie involved in the loss of strength and toughness. It is this oxygen embrittlement degradation which is ad­ dressed in this report by a further description of it in two different glass-ceramic matrices, and with possible explanations as to how it occurs
in them.
Mechanical Tests
Samples were prepared for the four-point flexural testing by blanchard grinding of the hot-pressed billets or of separate samples to the final thickness of either 0.20 or 0.15 cm. The 0%° specimens were cut on a diamond saw to final dimensions of either 9.5xO.46xO.20 cm or 4.5xO.46x 0.15 cm. No edge bevelling was performed because prior experimentation indicated that tough composites were not sensitive to edge or surface flaws from this specimen preparation.
The testing spans used were either 6.4 and 2.0 cm for the support and load spans or for the shorter specimens 4.0x1.3 cm. These sample and load­ ing geometries result in span to depth (L/h), ratios of 32 and 26.7 for the longer and shorter samples, respectively. Room temperature testing indi­ cated little difference in properties or failure modes between these two
flexural situations. The fixtures at room temperature were stainless-steel while for elevated temperature they were machined high purity alumina. Provision was made to ensure equal loading of the upper span in both cases.
The high temperature tests included some at two strain rates. Most testing was carried out at ~=5xl0-5sec-l, but at elevated temperatures tests were also run at ~=2.5xl0-3sec-l, all on similar mechanical testing machines. The high temperature experiments were run in an SiC heated box furnace controlled by a solid state controller from a Pt.-l0% Rh. thermo­ couple located near the specimen mid-span. Samples were introduced into the furnace and then allowed to equilibrate for at least 15 minutes. Samples were removed from the furnace within five minutes of initiating the
test.
One other study was done to help identify the physical processes taking place during the rapid oxidation embrittlement of LAS-III composites. This study involved making an LAS-III composite that was heavily microcracked in the matrix in the as-processed condition. Flexure samples were made and
some were exposed to 900°C air for either 20 minutes or 24 hours. The as­ formed and heat-treated specimens were compared by 25°C flexural testing. Other specimens of all three conditions were subjected to a 49% HF soak for
24 hours to dissolve the LAS-III matrix, followed by ultrasonic cleansing
in dionized water, leaving the fibers behind. Single filament tensile
tests were conducted, using the ASTM 03379 Standard Test Method for Tensile
36
Strength and Young's Modulus for High Modulus Single-Filament Materals,
with a 2.54 cm gauge length on the fibers from the as-formed, and 24-hour exposed specimens. Fibers from all three conditions were also analyzed with ESCA. Fiber diameters for the tensile strength tests were estimated by optical microscopy of the remaining gauge sections which were recovered by using petroleum jelly coatings of the fiber gauge length.
Microscopy
Optical microscopy was conducted to observe the microstructure of the as-formed compos:; tes and specifically to determine if matrices were micro­ cracked in the as-formed state. This was done using polished sections and a Zeiss Ultraphot microscope using Nomanshift interference optics. Fiia­ ment diameter measurements for the fiber strength tests were made in trans­ mission mode.
Scanning electron microscopy was used to document fracture morphology and fiber/matrix interfaces. Specimens were coated with gold in all cases, and the sample orientation from bend bars was noted to keep tensile and compression sides separate.
Transmission electron microscopy was performed by preparing thin foils by ion milling.
_160 ~ 150
~ :::J x lLJ ...J u..
Figure 1
SiC FIBER REINFORCED BMAS ill GLASS- CERAMIC (0°/0° , 30VOL %, 25°C)
I TENSILE i-=MA=T=RI---:X c=M=ICccRO-::-;C:-::RA:-C_KI_NG ______ • r FAILURE
~\~~~ ~6~gl~~NG _--r-~I ~ FIBER FAILURE
0.5 \.0 STRAIN (%)
2.0
Flexural stress-strain curve for SiC/BMAS-III at 25°C -5 -1 and ~=5x10 sec .
37
RESULTS AND DISCUSSION
Room Temperature Fracture
Figure 1 is a 4-point flexural stress-strain curve for a BMAS-III com­
posite tested on the 6.4 cm support span at room temperature. It illu­
strates the behavior of a strong and tough unidirectional composite where the matrix failure strain is much less than the fiber failure strain
(E «Ef ). -The matrix microcrack point corresponds to the first detected mu u non-linearity after an initial composite linear elastic region. As the
flexural strain is increased the matrix further microcracks and fibers begin debonding in the tensile half of the specimen. In the case of BMAS­
III composites, just as the compression surface begins buckling a large
crack grows on the tension surface followed by fiber failure which allows a rapid load drop.
LAS-III composite curves are similar until the ultimate stress region.
Here the composites fail from the compression ~ide before any large cracks
can open on the tensile surface of flexure specimens, and the drop-off of stress after the maximum is less rapid.
(a) (b)
Figure 225°C failure morphologies at g=5xlO- 5sec-T for (a)
SiC/LAS-III showing matrix microcracks on tensile surface,
and (b) SiC/BMAS-III fibrous rupture (tensile side on
bottom).
38
Scanning electron micrographs depicting typical fracture morphologies
are shown in Figure 2(a) for SiC/LAS-III and in Figure 2(b) for SiC/BMAS­
III tested at room temperature. These low magnification photomicrographs
indicate that there is weak interfacial bonding, matrix microcracking
[2(a)], and apparent fiber debonding occurring during the flexure tests.
Since the LAS-III composites fail from the compression side full advantage
of fiber debonding and pull-out is not realized. Because BMAS-III compo­
sites generally rupture from the tensile side with fiber failure following
the opening of a large crack more debonding and pull-out can occur.
Figure 3(a) and (b) are views of the fiber/matrix interface zones for
SiC/LAS-III [3(a)] and SiC/BMAS-III [3(b)]. Note the grainy appearance at
the LAS-III interface, which is probably due to the NbC formed during pressing. The BMAS-III sample has a thin layer that seems to have peeled off of the fiber surface as the fiber pulled out of the matrix. Brennan(6)
has also reported observing small pieces of thin films adhering to SiC fiber surfaces for room temperature fractured LAS-III composites.
(a) (b)
Figure 3 (a) SiC/LAS-III flexure sample showing NbC grains in
matrix next to fiber, (b) SiC/BMAS-III with thin layer
peeled off fiber.
have concluded that whenever a composite demonstrates good toughness at
room temperature, with rupture surfaces similar to those shown above, a
thin 10-20 nm layer structure is present in the interface region. There
is agreement that this layer is primarily carbon, and we believe it to be
at least partly graphitic. This layer has not been seen for composites
that are weak and brittle, which were consolidated at lower maximum hot­
pressing temperatures than those which are strong and tough. We also found
39
that strong and tough SiC fiber/Code 1723 glass composites require development of this carbon-rich layer. Brennan(6) independently made the
same conclusion for both LAS-III composites and for SiC/Code 7930 Vycor® glass composites. It is likely that the peeled film seen in the interface region of Figure 3(b) for the BMAS-III material corresponds to this layer.
These results imply that tough, strong composites have formed a carbon­ rich layer through reaction between the SiC fiber and silicate matrices. Recall that free carbon is available in the fiber. The formation of this carbon layer must provide the weak interface that allows fiber debonding and crack stoppage to occur.
·iii -'"
SICILAS m COMPOSITE (0°/0", 30 VOL %J STRAIN RATE DEPENDENCE OF STRENGTHS IN AIR
120
110
..J 80 .<C
60 ..J lJ..
20 8 10
I I 1200 1400
Figure 4 Ultimate flexural stress vs. temperature for two strain rates in air for SiC/LAS-III.
High Temperature Fracture
Figure 4 is a plot of the ultimate stress versus temperature at test strain rates of 5xlO- 5sec- l and 2.5xlO- 3sec- l for an LAS-III composite.
The slow strain rate (~) ultimate composite stress (crcu ) drops to a minimum in the region of 900-1000°C, and it exemplifies the oxidation embrittlement problem. Although thls is the "slow" strain rate test, the flexure experi-
40
ment takes only about 1 1/2 minutes to complete after the 15-minute equili­ bration period. and the active oxidation reaction really does not begin until the matrix has microcracked. This has been demonstrated by long-term thermal aging studies in air for SiC/LAS materials by Prewo(8) and by ourselves on SiC/BMAS-III composites. The slight upturn in this property beyond 11000 C we believe is due to matrix softening that delays the onset of matrix microcracking.
Figure 5 SiC/LAS-III tested at 900°C. ~=5xlO-5sec-1. Less
fi brous fail ure morphology ill ustrates oxi dati on degradation effects.
Figure 5 is a scanning electron micrograph of an LAS-III composite -5 -1 specimen tested at 900°C at ~=5xlO sec • near the 0 CU minimum. It is
obvious that the degradation in 0 CU and associated toughness corresponds to a loss of the fibrous character of the rupture surface (compare to Figure 2(a)). Apparently either the fibers are catastrophically weakened and embrittled or as a result of the oxidation reaction an intimate bond between fiber and matrix has formed so that crack stopping weak layers are no longer present.
Figure 4 shows that when the flexural tests were performed at the faster -3 -1 speed (~=2.5xlO sec ) however. only a slight decrease of 0 CU was found
at lOOuC, and it increased slightly at 800°C and 900°C before again dropping off at >lOOO°C.
41
Figure 6 SiC/LAS-III tested at 900°C, €=2.5xlO-3sec- l . Note thin embrittled region adjacent to tensile surface before transition to fibrous character of majority of rupture surface.
An SEM micrograph shown in Figure 6 of a specimen tested at the fast strain rate at 900°C displays a fibrous character more like the 25°C tested material except for a thin region adjacent to the tensile surface. Appar­ ently at this rapid testing speed the oxidation reaction has not proceeded fast enough to degrade the composite performance. A comparison of matrix microcrack strains between the fast and slow strain rates indicated the testing speed had no effect on this point. Apparently either the diffusion of oxygen to the carbon-rich interface is not rapid enough at the fast testing speed or the composite embrittling reaction hasn't proceeded fast enough to affect the composite.
Figure 7 has the 0cu versus temperature results of flexural tests at the strain rates of 5xlO-5 and 2.5xlO-3sec- l for a BMAS-III composition composite. This composite unfortunately had a little cordierite mixed with Ba-osumilite and hence more residual glass in the matrix so that its refractoriness suffers. This doesn't appear to affect the results until ~llOO°C. For comparison the ° versus temperature of a more fully developed Ba-osumilite BMAS-II¥ucomposite tested in air at €=5xlO- 5sec- l "
is listed in Table 1, along with these data for the composite of Figure 7.
Referring to Figure 7, the ° data as a function of temperature for the "slow" test speed (t=5xl 0-5se~~1) ill ustrates that again the oxidation
42
SiC/BMAS m (0%°, 30 VOL %1 STRAIN RATE DEPENDENCE OF ULT STRESS IN AIR
150
~ 120
W 70
!;t 60 E =5.0 x 1O-5 00c- I ' \ ::!:
5 50 :::l 40 ~O w '= 30 If) 0 20 (L
::!: 10 0
u
Figure 7 Ultimate flexural stress vs. temperature at two strain rates in air for SiC/B~lAS-III.
Table 1: Dependence of Ultimate Stress on Crystallization
Phases --> Cordierite + Osumilite Osumilite Temp. (OC) er + St. Dev. er + St. Dev. cu- cu-
(Ksi) (Ksi)
25 136 + 0.7 149 + 7 900 37.5+ 0.5 41.0+ 3.4
1100 44.8+ 3.0 51.6+ 0.2 1200 42.0+ 1.9 68.4+ 0.6 1250 64.7
embrittlement of the composite dominates the 700-1100°C fracture with an ultimate composite stress minimum at 900°C. An SEM micrograph of a BMAS­ III composite tested at 900°C in air at this strain rate is shown in Figure 8. A comparison of this micrograph with the room temperature micrograph (Figure 2(b)) illustrates the dramatic effect of the oxidation embrittlement.
43
morphology (no fiber debond or pullout) due to oxidation effects.
The "fast" speed (~=2.5xlO-3) test data indicates that again when less time is given for the oxidation degradation reaction the composites appear tougher and stronger. However, the high-speed ° data for this material did cu decrease with increasing temperature. Also the strain rate 0 CU data displays more sensitivity for this BMAS-III matrix composite at 700°C compared to the SiC/LAS-III. Thus either a faster oxidation reaction may occur for this composite (SiC/BMAS-III) or the embrittling phenomena are not the same. Figure 9 is an SEM micrograph of a BMAS-III composite sample tested at 900°C at ~=2.5xlO-3sec-l. Adjacent to the tensile surface there is again an embrittled zone as seen in LAS-III composites tested at this condition (Figure 6); however, this brittle region appears more extensive for SiC/ BMAS-III and more planer (less pull-out).
Figure 9 SiC/BMAS-III tested at 900°C at ~=2.5xlO-3sec-l.
44
Scanning electron microscope examinations of the interface region for both composites do suggest some difference of behavior resulting from the
oxidation reaction. Figure 10(a) is a SiC/LAS-III sample tested at 900°C at ~=5xlO-5sec-l. There is an obvious space around many fibers, such as
this one, that may result from the removal of the carbon-rich layer by oxidation. Figure 10(b) is a SiC/BMAS-III specimen from the same test condition. The region between fiber and matrix is apparently bridged by material in place of the carbon-rich zone. Perhaps this difference in behavior at the fiber/matrix interface accounts for the slight amount of
fiber debonding and pull-out seen for LAS-III composites whereas the BMAS­ III materials display more plane features, suggesting strong fiber to matrix bonding has developed.
(a) (b)
Figure 10 Close-ups of interface regions of (a) SiC/LAS-III and (b) SiC/BMAS-III both tested at 900°C at ~=5xlO-5sec-l.
A variance in behavior between these composite systems should not be
unexpected. Not only are the matrix chemistries and phases obviously different, but the LAS-III composite also has the NbC crystals along the fiber/matrix boundary. Brennan(6) indicated that the Nb205 additions could also increase the thickness of the carbon-rich, weak interface layer. In discussion of the possible embrittlement mechanism in SiC/LAS-III materials Brennan's analytical results led him to conclude that during high temperature flexure testing in air the matrix microcracking allows air to penetrate into the composite where it very quickly reacts with the Nicalon® fibers, forming an oxide layer on them that renders them either extremely brittle or bonds them very strongly to the matrix. This very rapid oxide
formation is only seen for these SiC fibers that have been incorporated into matrices where the carbon-rich interface has formed. Oxide growth on
45
as-received Nica10n® SiC fibers in the 800o -1100°C range in air was not significant for up to 24 hours nor were the fibers embritt1ed.(6}
Fiber Studies from LAS-III Composites
To better understand how the LAS-III composites could show an embrit­ t1ement, despite the gaps between fiber and matrix observed in 800o -1100°C air fractures, an experiment was devised to study the behavior of fibers after incorporation into the matrix. The mechanical results of this study are summarized in Table 2 and Figure 11.
Table 2: SiC/LAS-III Flexural Results on Microcracked Samples Thermally Aged in Air vs. As-Processed
Condition O"cu (Ksi) ~u(%} Fracture Appear. Mode As-Processed 58.6* .56 Fibrous Shear 900°C/20 min. 60.3 .87 Fibrous Shear 900oC/24 hrs. 24.4 .16 Brittle Tensile
* Heavily microcracked composites fail in compression or shear giving
much lower as-processed O"cu' s and €cu's.
Matrix microcracked composite specimens (as-processed) that were heat­ treated at 900°C in air for 20 minutes or 24 hours are compared to the as­ processed condition by 25°C flexural testing in Table 2. An optical micro­ graph of a polished section of the composite in Figure 12 shows the micro­ cracks in the as-processed materi~. The Table 2 results indicate that a 20-mi"nute 900°C air exposure has no degrading effect on this microcracked composite; however, for 24 hours at 900°C the ultimate stress measured was less than half the as-processed result, and the failure mode and appearance changed to a brittle character. The rupture surface looked just like those tested at 900°C in flexure. It appears the microcracks formed on process­ ing are not as effective a diffusion path for oxygen as are cracks formed
46
~
o As- PROCESSED m=4.69 HEAT TREATED m=3.IO
Figure 11 Wei bull failure probability distribution of single filament strengths on fibers extracted from matrix microcracked SiC/LAS-III samples.
during flexural stressing at the test temperature, so the degradation reaction is much slower for thermally aged specimens versus those under­ going flexure at temperature.
Figure 11 compares the Weibu11 strength distributions of fibers extrac­ ted from the as-processed composite and from samples thermally aged at 900°C for 24 hours. The measured average tensile strength of fibers from the thermally aged composite specimens is over 40% lower than the as­ processed composite fibers (193 vs. 329 Ksi). The weaker fibers also seemed to be more brittle while handling, and this is reflected in the lowering of the Wei bull modulus from 4.7 for fibers from the as-processed composite to 3.1 for fibers from the air exposed samples. The average streng