Advances in ceramic composites reinforced by continuous fibers Brian N Cox* and Frank W Zokt Ceramic matrix composites reinforced with continuous fibers loads. The serious difficulties of ensuring durability are on the verge of insertion into hot engineering structures cratures are being confronted; oxidation Yet current research is only beginning to attack some of pesting of SiC fibers at intermediate temperatures, fiber the most critical problems. Key developments in the last eep at higher temperatures, and the chemical stability 24 months include the formulation of constitutive laws for of interfaces arc all hot topics. Textile reinforcement, continuum mechanics analyses: the discovery of stable weak especially with 3D architecture, has appeared as the xide-oxide interface systems: the analysis of how fiber creep solution to the unavoidable vulnerability of brittle matrix s life at high temperatures; confrontation of the problem composites to delamination. And even the central axiom of oxidation pesting at intermediate temperatures in Sic that CMCs cannot be tough unless the fiber/matrix based systems: re-examination of the maxim that interfaces interfaces are weak is now being challenged must be weak; and the advent of textile reinforcement as the olution to delaminat Modeling the inelastic regime Major progress has been made in the last year or two in devcloping design and reliability codes suitable for Addresses Rockwell Science Center, 1049 Camino Dos Rios PO Box 1085 field use from the wealth of micromechanical models in Thousand Oaks, CA 91358, USA the CMC literature. Effort has focused on generating Materials Department,University of California at Santa Barbara, CA constitutive laws for insertion into finite element models, ith the goal of reducing Current Opinion in Solid State Materials Science 1996 CMCs to standard continuum mechanics Current Chemistry Ltd ISSN 1359-0286 Nonlinearity in CMCs at room temperature involves matrix cracking, stochastic fiber fracture, damage local Abbreviation CMCs ceramic matrix composites ization, and fiber pullout. Two groups have presented exhaustive studies for unidirectional composites of the relation between micromechanical properties (including the interfacial friction stress, residual stresses, constituent Introduction elastic moduli. fiber radius, and fiber volume fraction) The period covered by this review (1995 and the and the macroscopic stress-strain response under aligned beginning of 1996, with selected inclusion of papers from loads prior to damage localization and ultimate failur 1994)marks a major epoch in the history of research [3-9]. Prior and well established models of matrix cracks into continuous fiber reinforced ceramic matrix composites bridged by sliding fibers are used as the physical basis (CMCs). From the carly 1980s, when CMC research for modeling. Micromechanical properties are deduced first enjoyed large scale funding and the attention of directly from experimental hysteresis loops, obviating any significant groups all over the world, effort has been detailed tests of interface conditions, for cxample fibcr concentrated on a simple paradigm of the ideal CMC. It pullout or pushout tests. One group has couched its work must have a weak fiber/matrix interface to allow energy in the language of micromechanics more familiar ro the absorption during fracture by the deflection of cracks, in CMC community [3, 451; the other in the language of the complete absence of any dislocation based toughening. continuum damage mechanics, but with a thermodynat Freed of stress concentration when the matrix cracked, potential function derived from the same micromechanics strong fibers would continue to bear high loads. This 16, 7, 8, 9]. They offer equivalent treatments of nonlinear approach to protecting CMCs from intrinsic faws, notches, ity up to localization, with some variations in the poi and damage was pursued almost entirely in the context of of view and in the level of micromechanical detail used unidirectionally reinforced CMCs, with aligned loads; and in fitting data. Both sets of work are essential readi mostly in terms of room temperature phenomena. It is now More empirical (probably unnecessarily so)treatments of very well understood. (See [1, 2] for recent articles covering nonlinearity in unidirectional CMCs have also appeared many aspects of work up to 1995.) Structural applications almost never involve uniaxial Recent extensions of the continuum damage approach stresses; and the long sought pay-off for CMCs will also deal with predicting the onset of damage localization, certainly come at high temperatures. Now we see at which is required to model ultimate failure and the notch ast the reduction of micromechanical models and our sensitivity of strength [11]. Localization and subsequent dctailcd understanding of matrix cracking and statistical fiber pullout involve distributions of Aaw strengths and fiber failure to constitutive laws suitable for use in stress redistribution effects which are complex and not finite element calculations of structures under complex generally well known in a particular material.There
666 Advances in ceramic by continuous fibers Brian N Cox* and Frank W Zokt Ceramic matrix composites reinforced with continuous fibers are on the verge of insertion into hot engineering structures. Yet current research is only beginning to attack some of the most critical problems. Key developments in the last 24 months include the formulation of constitutive laws for continuum mechanics analyses; the discovery of stable weak oxide-oxide interface systems; the analysis of how fiber creep limits life at high temperatures; confrontation of the problem of oxidation pesting at intermediate temperatures in Sic based systems; re-examination of the maxim that interfaces must be weak; and the advent of textile reinforcement as the solution to delamination problems. Addresses *Rockwell Science Center, 1049 Camino DOS Rios, PO Box 1065, Thousand Oaks, CA 91356, USA +Materials Department, University of California at Santa Barbara, CA 93106, USA Current Opinion in Solid State & Materials Science 1996, 1:666-673 0 Current Chemistry Ltd ISSN 1359-0266 Abbreviation CMCs ceramic matrix composites Introduction The period covered by this review (1995 and the beginning of 1996, with selected inclusion of papers from 1994) marks a major epoch in the history of research into continuous fiber reinforced ceramic matrix composites (ChlCs). From the early 198Os, when CMC research first enjoyed large scale funding and the attention of significant groups all over the world, effort has been concentrated on a simple paradigm of the ideal ChlC. It must have a weak fiber/matrix interface to allow energy absorption during fracture by the deflection of cracks, in the complete absence of any dislocation based toughening. Freed of stress concentration when the matrix cracked, strong fibers would continue to bear high loads. This approach to protecting CMCs from intrinsic flaws, notches, and damage was pursued almost entirely in the context of unidirectionally reinforced ChlCs, with aligned loads; and mostly in terms of room temperature phenomena. It is now very well understood. (See [1,‘2] for recent articles covering many aspects of work up to 1995.) Structural applications almost never involve uniaxial stresses; and the long sought pay-off for ChlCs will certainly come at high temperatures. Now we see at last the reduction of micromechanical models and our detailed understanding of matrix cracking and statistical fiber failure to constitutive laws suitable for use in finite element calculations of structures under complex loads. The serious difficulties of ensuring durability at high temperatures are being confronted; oxidation pesting of SIC fibers at intermediate temperatures, fiber creep at higher temperatures, and the chemical stability of interfaces are all hot topics. Textile reinforcement, especially with 3D architecture, has appeared as the solution to the unavoidable vulnerability of brittle matrix composites to delamination. And even the central axiom that CMCs cannot be tough unless the fiber/matrix interfaces are weak is now being challenged. Modeling the inelastic regime Major progress has been made in the last year or two in developing design and reliability codes suitable for field use from the wealth of micromechanical models in the Ch,iC literature. Effort has focused on generating constitutive laws for insertion into finite element models, with the goal of reducing the treatment of nonlinearity in CIVICS to standard continuum mechanics. Nonlinearity in CMCs at room temperature involves matrix cracking, stochastic fiber fracture, damage localization, and fiber pullout. Two groups have presented exhaustive studies for unidirectional composites of the relation between micromechanical properties (including the interfacial friction stress, residual stresses, constituent elastic moduli, fiber radius, and fiber volume fraction) and the macroscopic stress-strain response under aligned loads prior to damage localization and ultimate failure [3-91. Prior and well established models of matrix cracks bridged by sliding fibers are used as the physical basis for modeling. Micromechanical properties are deduced directly from experimental hysteresis loops, obviating any detailed tests of interface conditions, for example fiber pullout or pushout tests. One group has couched its work in the language of micromechanics more familiar to the ChlC community [3*,4’,5]; the other in the language of continuum damage mechanics, but with a thermodynamic potential function derived from the same micromechanics [6,7,8”,9]. They offer equivalent treatments of nonlinearity up to localization, with some variations in the point of view and in the level of micromechanical detail used in fitting data. Both sets of work are essential reading. hlore empirical (probably unnecessarily so) treatments of nonlinearity in unidirectional CMCs have also appeared 1101. Recent extensions of the continuum damage approach also deal with predicting the onset of damage localization, which is required to model ultimate failure and the notch sensitivity of strength [ 111. Localization and subsequent fiber pullout involve distributions of flaw strengths and stress redistribution effects which are complex and not generally well known in a particular material. There
Ceramie composites reinforeed by continuous fibers Cox and Zok 667 is a commensurate increase in the number of material under most stress is small (as it is ahead of a notch) parameters to be determined by calibrating experiments. If so, the strength should follow weakest-link scaling Applications to circumferentially reinforced rotors have Comparisons of strengths in tension and bending (taking been presented, but proof of the predictive power of into account the nonlinear stress distributions) support localization models is not yet convincing [11] this hypothesis(MeNulty JC, Zok FW, unpublished data) Issues related to strength variability are addressed in more ling with more complex fiber architectures is also detail elsewhere in this journal [22] ch more challenging. The response of a 0/90"lamina prior to localization under loads aligned with the 0'fibers The degree of notch sensitivity is influenced by the nature the easiest case, since the cracking evolution is of the inelastic deformation occurring ahead of the notches best understood(5]. Damage in textile CMCs involves ( Fig. 1). In some materials (e. g. Nicalon TM/calcium much more complicated cracking patterns, for aluminosilicate), a damage zone of multiple matrix cracks micromechanical models are relatively crude (and not forms ahead of the notch, which has an analogous effec certain to improve, because of the difficulties of dealing to the plastic zone in metals(designated Class II behavior with the tortuous heterogeneity of textiles). Continuum by Evans [231). In others(e. g. C/C), nonlinearity arises damage approaches are necessarily more empirical from shear bands oriented parallel to the tensile direction (Class IIi behavior [23]). In more brittle CMCs, fracture Likewise, highly empirical approaches arc most credible occurs by the propagation of a dominant mode I crack, for multiaxial or off-axis loading, even in unidirectional with fiber failure and pullout in the crack wake, but with CMCs. A general method for developing multiaxial minimal inelastic deformation elsewhere(Class I behavior constitutive laws up to localization has been demonstrated [(23.1). Models of strength for Classes I and In have been for plane stress cases, using a combination of standard developed, based on line-spring representations of the tension, compression, and shear test data [12 ]. when the inelastic processes (24, 25 ]. Models that take into account constitutive laws are embedded in finite element calcula- large scale sliding [26] indicate that the maximum fber encouraging agreement is obtained with measured strain predicted from the line-spring models; the latter are ficlds. Stiffness changes under off-axis loading have been thus expected to provide conservative predictions for the measured ultrasonically [13]. stresses at the onset of fiber failure This area of work represents the culmination of efforts Some censure is due to several authors over loose claims to qualify CMCs as structural materials. Current activity that a given material has been found to be notch focuses on dealing with rate dependent behaviour at hi insensitive. This generally fallacious conclusion has been temperature, fatigue effects, and weakest link fract based on tests performed with relatively small notches statistics(volume effects) typically 1-5 mm. Moreover, there has been almost discussion of the effects of notch shape (circular holes versus sharp slits). In the presence of sufficiently large, Fracture and notch sensitivity sharp notches, the strength must follow the Griffith Tensile tests performed on specimens containing holes relation and the material must be notch-sensitive (as is or notches have demonstrated that many CMCs are even the most ductile metal). Researchers should identify relatively notch-insensitive [14-16, 17,18, 19, 20, 21). The the net-section stress at fracture is typically 80-100% of the i head scales associated with the bridging processes the notch sizes and shapes for which notch unnotched strength: considerably higher than the value sensitivity will occur calculated on the basis of the elastic stress concentration factor. Indeed, in some instances, there appears to be evidence of notch strengthening [14]. Measurements of Compressive failure in-plane strains (using moire interferometry (18)and Compressive failure of CMcs has remained largely unex stresses(using SPATE (14-16, 17.))have shown that plored. Some evidence exists that com strain concentrations are essentially unchanged by the [27]fall below tensile strengths [28 ] In CMCs with weak inelastic deformation but stress concentrations are reduced or porous matrices, observations to date [27]show that dramatically. However, even in the most notch-insensitive compressive failure involves kink band formation within materials, stress concentrations are not eliminated alto- fiber bundles(plies or tows), similar to the prevalent fail gether, yet the net section strength is essentially equal ure mechanisms in polymer matrix composites(laminates to the unnotched strength. Similar conclusions have been and textiles). In this case, compressive strength will be reached from finite clement simulations which incorporate governed by the initial misalignment of segments of fiber the inelastic deformation [12] bundles and the shear strength of the matrix. Compressive failure also involves interply and intraply delamination, These results suggest that the failure stress should exhibit which will probably be the principal mechanisms of failure volume dependence, being highest when the volume in CMCs with nonporous, relatively strong matrices
Ceramic composites reinforced by continuous fibers Cox and Zok 667 is a commensurate increase in the number of material parameters to be determined by calibrating experiments. Applications to circumferentially reinforced rotors have been presented, but proof of the predictive power of localization models is not yet convincing [ll]. Dealing with more complex fiber architectures is also much more challenging. The response of a 0/9O’laminate prior to localization under loads aligned with the O’fibers is the easiest case, since the cracking evolution is best understood [S]. Damage in textile CMCs involves much more complicated cracking patterns, for which micromechanical models are relatively crude (and not certain to improve, because of the difficulties of dealing with the tortuous heterogeneity of textiles). Continuum damage approaches are necessarily more empirical. Likewise, highly empirical approaches are most credible for multiaxial or off-axis loading, even in unidirectional CMCs. A general method for developing multiaxial constitutive laws up to localization has been demonstrated for plane stress cases, using a combination of standard tension, compression, and shear test data [l?]. When the constitutive laws are embedded in finite element calculations of strain distributions around a stress concentrator, encouraging agreement is obtained with measured strain fields. Stiffness changes under off-axis loading have been measured ultrasonically [13]. This area of work represents the culmination of efforts to qualify CMCs as structural materials. Current activity focuses on dealing with rate dependent behaviour at high temperature, fatigue effects, and weakest link fracture statistics (volume effects). Fracture and notch sensitivity Tensile tests performed on specimens containing holes or notches have demonstrated that many CMCs arc relatively notch-insensitive [14-16,17*,18*,19,20,21’]. The net-section stress at fracture is typically 80-100% of the unnotched strength: considerably higher than the value calculated on the basis of the elastic stress concentration factor. Indeed, in some instances, there appears to be evidence of notch strengthening [14]. Measurements of in-plane strains (using moire interferometry [18*]) and stresses (using SPATE [14-16,17*]) have shown that strain concentrations are essentially unchanged by the inelastic deformation but stress concentrations are reduced dramatically. However, even in the most notch-insensitive materials, stress concentrations are not eliminated altogether, yet the net section strength is essentially equal to the unnotched strength. Similar conclusions have been reached from finite element simulations which incorporate the inelastic deformation [12*]. These results suggest that the failure stress should exhibit volume dependence, being highest when the volume under most stress is small (as it is ahead of a notch). If so, the strength should follow weakest-link scaling. Comparisons of strengths in tension and bending (taking into account the nonlinear stress distributions) support this hypothesis (MeNulty JC, Zok FW, unpublished data). Issues related to strength variability are addressed in more detail elsewhere in this journal [22]. The degree of notch sensitivity is influenced by the nature of the inelastic deformation occurring ahead of the notches (Fig. 1). In some materials (e.g. NicalonTVcalcium aluminosilicate), a damage zone of multiple matrix cracks forms ahead of the notch, which has an analogous effect to the plastic zone in metals (designated Class II behavior by Evans [23’]). In others (e.g. C/C), nonlinearity arises from shear bands oriented parallel to the tensile direction (Class III behavior [23-l). In more brittle CMCs, fracture occurs by the propagation of a dominant mode I crack, with fiber failure and pullout in the crack wake, but with minimal inelastic deformation elsewhere (Class I behavior [23*]). Models of strength for Classes I and III have been developed, based on line-spring representations of the inelastic processes [24*.,25]. Models that take into account large scale sliding [26”] indicate that the maximum fiber stress in the bridging zone is somewhat lower than that predicted from the line-spring models; the latter are thus expected to provide conservative predictions for the stresses at the onset of fiber failure. Some censure is due to several authors over loose claims that a given material has been found to be notch insensitive. This generally fallacious conclusion has been based on tests performed with relatively small notches: typically 1-5 mm. Moreover, there has been almost no discussion of the effects of notch shape (circular holes versus sharp slits). In the presence of sufficiently large, sharp notches, the strength must follow the Griffith relation and the material must be notch-sensitive (as is even the most ductile metal). Researchers should identify the length scales associated with the bridging processes and hence the notch sizes and shapes for which notch sensitivity will occur. Compressive failure Compressive failure of CMCs has remained largely unexplored. Some evidence exists that compressive strengths [27’] fall below tensile strengths [28-l. In CMCs with weak or porous matrices, observations to date [27*] show that compressive failure involves kink band formation within fiber bundles (plies or tows), similar to the prevalent failure mechanisms in polymer matrix composites (laminates and textiles). In this case, compressive strength will be governed by the initial misalignment of segments of fiber bundles and the shear strength of the matrix. Compressive failure also involves interply and intraply delamination, which will probably be the principal mechanisms of failure in CMCs with nonporous, relatively strong matrices
668 Ceramics, composites and intergrowths Figure 1 While fatigue effects in CMCs with SiC or oxide fibers appear to be mainly related to interfacial degradation, with minimal fiber damage, carbon fibers can be entirely worn away by fatigue. Thus generally severe fatigue effects are found in CMCs based on carbon fibers [32 Delamination Whether toughness is achieved in a CMC by incorporating weak fiber matrix interfaces or a porous or otherwise weak matrix around fiber bundles, strength under loads acting normal to the fiber direction will be seriously impaired Thus while laminated CMCs exhibit encouraging in-plane toughness, they remain vulnerable to delamination. De lamination can be resisted to some degree by in-plane hbers crossing the fracture plane obliquely, but the work of fracture remains well below 1 k]m-2[33, 34]. Laminates of 2D fabrics (e. g, plain or satin weave)arc equally vulnerable The delani CMCs are not yet preferred for load bearing components in high performance applications such as turbine engines Schematics of three classes of cracking found in unidirectional or Current designs require them to sustain mainly thermal cross-plied CMCs (a)Class 1, matrix cracking plus fiber failure: (b) loads. Even then, thermal gradients present severe delani Class l, matrix cracking, no fiber failure: (e) Class l, shear damage. nation risks. Hutchinson and Lu(35)have explored how in textile CMCs, because of the strong role played by heterogene lamination crack interrupts hcat flow, producing theimal on the scale of fiber tows or bundles. Their identification is a topic of stresses and crack tip stress intensities. The design limit current research. implied for thermal gradients is quite stringent for realisti parameter values The obvious way to suppress delamination is by in corporating through-thickness reinforcement, for exam ple, by stitching, 3D weaving, or inserting short rods. tigue Very encouraging precedents exist in polymer compos Fatigue failure occurs in most CMCs. The dominant ites [36, 37]. While the processing challenges remain echanism at room temperature in CMCs based on largely unaddressed by the ceramics community, data oxide or SiC fibers involves matrix cracking on the first for other composite systems and theory lead the way loading cycle, followed by debonding and cyclic sliding Through-thickness reinforcement bridges delamination along the fiber-matrix interfaces. Repeated sliding causes cracks, often creating an analogue of the steady state wear of the fiber coatings, leading to a reduction in the matrix crack familiar from Mode I aligned loading of interface sliding stress and a corresponding reduction in CMCs [38]. A lower bound therefore exists for the critical e fiber bundle strength [29, 30]. It can also lead to load, regardless of delamination crack length. This allows higher permanent inelastic strains and a reduction in the simple design rules to be formulated. The minimum hysteresis modulus. The latter effects may be important volume fraction of through-thickness reinforcement re for dimensional stability The wear process is also likely quired to suppress delamination is usually only a few to produce flaws in the fibers, further reducing thc bundle percent or less [39]. For the thermal gradicnt problem in trength. Such effects have been seen in fiber-reinforced particular, through-thickness reinforcement also transports fibers extracted from composites before and after fatigue), thermal stresses 35 ion crack. reducing crack-induced titanium matrix composites(by comparing the strengths of heat across a delamin but not in CMCs, partly because of the diffculty of extracting fibers from ceramic matrices. Typically, the Thermal properties fatigue thresholds are 275% of the ultimate tensile To minimize thermal stress, CMCs must have high strength (UTS)and the retained strengths following thermal conductivity along with low thermal expansion fatigue loading are almost equal to the UTS. However, and stiffness. Despite their importance determining limited data suggest that the fatigue threshold is reduced the performance of CMC structures, thermal piopeitics by notches(relative to the notched tensile strength)and have received relatively little attention, with a few notable negative stress ratios [31] theoretical studies. Models have been developed to
666 Ceramics, composites and intergrowths Figure 1 r- (a) (b) Schematics of three classes of cracking found in unidirectional or cross-plied CMCs. (a) Class I, matrix cracking plus fiber failure; (b) Class II, matrix cracking, no fiber failure; (c) Class III, shear damage by matrix cracking. (Additional stress redistribution mechanisms exist in textile CMCs, because of the strong role played by heterogeneity on the scale of fiber tows or bundles. Their identification is a topic of current research.) Fatigue Fatigue failure occurs in most CMCs. The dominant mechanism at room temperature in CMCs based on oxide or SIC fibers involves matrix cracking on the first loading cycle, followed by debonding and cyclic sliding along the fiber-matrix interfaces. Repeated sliding causes wear of the fiber coatings, leading to a reduction in the interface sliding stress and a corresponding reduction in the fiber bundle strength [29,30*]. It can also lead to higher permanent inelastic strains and a reduction in the hysteresis modulus. The latter effects may be important for dimensional stability. The wear process is also likely to produce flaws in the fibers, further reducing the bundle strength. Such effects have been seen in fiber-reinforced titanium matrix composites (by comparing the strengths of fibers extracted from composites before and after fatigue), but not in CMCs, partly because of the difficulty of extracting fibers from ceramic matrices. Typically, the fatigue thresholds are 275% of the ultimate tensile strength (UTS) and the retained strengths following fatigue loading are almost equal to the UTS. However, limited data suggest that the fatigue threshold is reduced by notches (relative to the notched tensile strength) and negative stress ratios [31]. While fatigue effects in CMCs with Sic or oxide fibers appear to be mainly related to interfacial degradation, with minimal fiber damage, carbon fibers can be entirely worn away by fatigue. Thus generally severe fatigue effects are found in CMCs based on carbon fibers [32]. Delamination Whether toughness is achieved in a CMC by incorporating weak fiber/matrix interfaces or a porous or otherwise weak matrix around fiber bundles, strength under loads acting normal to the fiber direction will be seriously impaired. Thus while laminated CMCs exhibit encouraging in-plane toughness, they remain vulnerable to delamination. Delamination can be resisted to some degree by in-plane fibers crossing the fracture plane obliquely, but the work of fracture remains well below 1 kJm-2 [33,34]. Laminates of 2D fabrics (e.g., plain or satin weave) are equally vulnerable. The delamination problem is one of the main reasons CMCs are not yet preferred for load bearing components in high performance applications such as turbine engines. Current designs require them to sustain mainly thermal loads. Even then, thermal gradients present severe delamination risks. Hutchinson and Lu [35*] have explored how a delamination crack interrupts heat flow, producing thermal stresses and crack tip stress intensities. The design limit implied for thermal gradients is quite stringent for realistic parameter values. The obvious way to suppress delamination is by incorporating through-thickness reinforcement, for example, by stitching, 3D weaving, or inserting short rods. Very encouraging precedents exist in polymer composites [36,37*]. While the processing challenges remain largely unaddressed by the ceramics community, data for other composite systems and theory lead the way. Through-thickness reinforcement bridges delamination cracks, often creating an analogue of the steady state matrix crack familiar from Mode I aligned loading of ChlCs [38*]. A lower bound therefore exists for the critical load, regardless of delamination crack length. This allows simple design rules to be formulated. The minimum volume fraction of through-thickness reinforcement required to suppress delamination is usually only a few percent or less [39]. For the thermal gradient problem in particular, through-thickness reinforcement also transports heat across a delamination crack, reducing crack-induced thermal stresses [35*]. Thermal properties To minimize thermal stress, CMCs must have high thermal conductivity along with low thermal expansion and stiffness. Despite their importance in determining the performance of CMC structures, thermal properties have received relatively little attention, with a few notable theoretical studies. Models have been developed to
Ceramic composites reinforced by continuous fibers Cox and Zok 669 account for degradation in the thermal expansion and structural applications under aligned loads and whe conductivity of cross-ply laminates in the presence of environmental degradation has been controlled. Current periodic matrix cracks [40, 41]. The models highlight the research focuses on notch sensitivity, where the combi 2.portance of the Biot numbers associated with fiber nation of matrix cracking and the stress concentration of trix interfaces, bridged matrix cracks in longitudinal the notch accelerate creep rupture of fibers [46, 55]. Fiber plies, and unbridged cracks in transverse plies, along with creep encourages the dominance of a single matrix crack, the corresponding crack densities. The through-thickness since it tends to relieve stresses on parallel matrix cracks conductivity can also be impaired by the presence of which have initiated upon the first loading. This contrasts orosiry. The effects of porosity and fiber waviness with cracking at room temperature, where many cracks are have been incorporated into a cell model for plain usually found, even next to very sharp notches [ 14] the through-thickness thermal properties is particularly The creep properties of the fibers are obviously critical important in design because of the low delamination in setting design limits and for determining lifetime resistance of CMCs. Much is yet understood. For example, early experiments have shown challengingly complex A review of the models and the experimental work relationships between morphological changes and creep covering the past decade can be found in [43]. Despite rates in NicalonTM fibers [56, 57]; and both creep rates and progress in the development of models for the thermal strength are likely to be affected by interactions berween properties of composites, a critical assessment of the fibers and either interphases or the matrix (Morscher models has been hampered by the lack of experimental GN, unpublished data). Developing creep resistant fibers data on the thermal properties of the constituents and establishing confidence in their performance remain (especially the fibers)and the conductance of fiber-matrix central problems in the CMc field A peripheral field is developing in the potential (distant Creep future! )application of SiC-bascd CMCs in fusion reactors Environment and phase stability aside, the design bounds SiC is favoured for its low nuclear activation rate. Creep f CMCs under aligned loads are set by creep failure remains central, but is now coupled with radiation In glass matrix composites, creep occurs predominant in the matrix. In unidirectional glass matrix composites, Oxidation embrittlement high creep rates in the matrix under transverse loads The problem of oxidation embrittlement continues to cause considerable creep anisotropy. In asymmetrically plague SiC-based CMCs. The embrittlement involves laid-up laminates or in the presence of stress gradients, oxygen ingress through matrix cracks and the subsequent creep anisotropy within plies will compromise dimensional reaction of oxygen with both the fiber coatings and the stability [44]. In CMCs designed for higher temperatures, fibers [60,61]. It occurs as a so-called pest phenomenon, hich have nonglass matrices, the situation is reversed. being worst at temperatures lower than those of intended Polycrystalline fibers are fabricated with fine grains for service. At higher temperatures, oxidation products near trength, which are usually smaller than the grains in the the external surfaces tend to seal cracks and inhibit matrix. It is therefore the fibers that creep first. Consider- further oxygen ingress. (Yet high temperature interfacial able progress has now been made towards understanding degradation is still a potential problem [62]. )The pest the important consequences of this. Under sustained temperature can be determined by mechanical testing loads, matrix cracks, which would be arrested and remain of tensile specimens subject to a temperature gradient stable at room temperature, exhibit stable, time dependent along the specimen length: the gradient being produced growth as fiber creep degrades the shielding effects by (localized)induction heating near the gauge center of bridging fibers [45-47, 48]. The steady state matrix [61]. Typically, the pest temperatures are in the range cracking stress is no longer a lower bound for nonlinearity 600-800.C. These temperatures are considerably lower and ultimate failure [49]. Constitutive laws have been than those usually used for high temperature testing of derived for bridging fibers that creep [50, 51](as well CMCs. Consequently, the embrittlement phenomenon as for creeping interfaces [521). the incubation of crack can be readily overlooked growth from prior matrix faws has been modeled [51] and crack growth trends have been detailed [53]. Global Additional complications arise because of the inherent is now well understood [47, 50]. Possible failure modes currently used in CMCs). Experiments on SiC/Nicalon TMI for 0/90laminates with creeping fibers have been mapped minicomposites (single tow composites) confirm that in terms of fundamental material parameters [49, 54]. failure at elevated temperatures dominated by the strength degradation of the fibers alone Matrix cracking moderated by creeping fibers appears to (Morscher Gn, unpublished data) Composites containing be the critical failure path for CMCs in high temperature Hi-Nicalon TaI fibers with bn coatings appear to be more
Cemmic composites reinforced by continuous fibers Cox and Zok 669 account for degradation in the thermal expansion and conductivity of cross-ply laminates in the presence of periodic matrix cracks [40,41’]. The models highlight the importance of the Biot numbers associated with fiber matrix interfaces, bridged matrix cracks in longitudinal plies, and unbridged cracks in transverse plies, along with the corresponding crack densities. The through-thickness conductivity can also be impaired by the presence of porosity. The effects of porosity and fiber waviness have been incorporated into a cell model for plain weave architectures [42]. As noted earlier, knowledge of the through-thickness thermal properties is particularly important in design because of the low delamination resistance of CMCs. A review of the models and the experimental work covering the past decade can be found in [43]. Despite progress in the development of models for the thermal properties of composites, a critical assessment of the models has been hampered by the lack of experimental data on the thermal properties of the constituents (especially the fibers) and the conductance of fiber-matrix interfaces. Creep Environment and phase stability aside, the design bounds of CMCs under aligned loads are set by creep. In glass matrix composites, creep occurs predominantly in the matrix. In unidirectional glass matrix composites, high creep rates in the matrix under transverse loads cause considerable creep anisotropy. In asymmetrically laid-up laminates or in the presence of stress gradients, creep anisotropy within plies will compromise dimensional stability [44]. In CMCs designed for higher temperatures, which have nonglass matrices, the situation is reversed. Polycrystalline fibers are fabricated with fine grains for strength, which are usually smaller than the grains in the matrix. It is therefore the fibers that creep first. Considerable progress has now been made towards understanding the important consequences of this. Under sustained loads, matrix cracks, which would be arrested and remain stable at room temperature, exhibit stable, time dependent growth as fiber creep degrades the shielding effects of bridging fibers [45*117*,48]. The steady state matrix cracking stress is no longer a lower bound for nonlinearity and ultimate failure [49]. Constitutive laws have been derived for bridging fibers that creep [50*,51] (as well as for creeping interfaces [52]), the incubation of crack growth from prior matrix flaws has been modeled [51], and crack growth trends have been detailed [53]. Global creep plasticity in the presence of multiple matrix cracks is now well understood [47*,50*]. Possible failure modes for 0/90’laminates with creeping fibers have been mapped in terms of fundamental material parameters [49,54]. Matrix cracking moderated by creeping fibers appears to be the critical failure path for CMCs in high temperature structural applications under aligned loads and when environmental degradation has been controlled. Current research focuses on notch sensitivity, where the combination of matrix cracking and the stress concentration of the notch accelerate creep rupture of fibers [46’,55]. Fiber creep encourages the dominance of a single matrix crack, since it tends to relieve stresses on parallel matrix cracks which have initiated upon the first loading. This contrasts with cracking at room temperature, where many cracks are usually found, even next to very sharp notches [14]. The creep properties of the fibers are obviously critical in setting design limits and for determining lifetime. Much is yet to be understood. For example, some early experiments have shown challengingly complex relationships between morphological changes and creep rates in NicalonTbl fibers (56,571; and both creep rates and strength are likely to be affected by interactions between fibers and either interphases or the matrix (Morscher GN, unpublished data). Developing creep resistant fibers and establishing confidence in their performance remain central problems in the CMC field. A peripheral field is developing in the potential (distant future!) application of Sic-based CMCs in fusion reactors. SIC is favoured for its low nuclear activation rate. Creep failure remains central, but is now coupled with radiation damage [58,59]. Oxidation embrittlement The problem of oxidation embrictlement continues to plague Sic-based CMCs. The embrittlement involves oxygen ingress through matrix cracks and the subsequent reaction of oxygen with both the fiber coatings and the fibers [60,61’]. It occurs as a so-called pest phenomenon, being worst at temperatures lower than those of intended service. At higher temperatures, oxidation products near the external surfaces tend to seal cracks and inhibit further oxygen ingress. (Yet high temperature interfacial degradation is still a potential problem [62].) The pest temperature can be determined by mechanical testing of tensile specimens subject to a temperature gradient along the specimen length: the gradient being produced by (localized) induction heating near the gauge center (61.1. Typically, the pest temperatures are in the range 600-SOO’C. These temperatures are considerably lower than those usually used for high temperature testing of CMCs. Consequently, the embrittlemenc phenomenon can be readily overlooked. Additional complications arise because of the inherent instability of NicalonThl fibers (the most common fiber currently used in CMCs). Experiments on SiC/NicalonThl minicomposites (single tow composites) confirm that failure at elevated temperatures is, in some instances, dominated by the strength degradation of the fibers alone (Morscher GN, unpublished data). Composites containing Hi-NicalonThl fibers with BN coatings appear co be more
670 Ceramics, composites and intergrowths stable, although a critical assessment of their performance porous matrix that offers easy splitting paths [67,68] under conditions that accentuate the embrittlement has (Other wood-like ceramics presented recently are really be performed porous monoliths [69). )The propensity for splitting in CMCs can be enhanced by compressive residual stresses A rudimentary model has been developed to predict in the matrix [70]. Weak matrices and fiber entanglement the rupture time under static loading [60]. More refined within fiber bundles must also favour splitting [28] odcls are nceded to take into account in a more Precedents in polymer and carbon-carbon composites realistic way the nature of the oxidation processes and suggest that especially effective toughening mechanisms the mechanisms responsible for fiber strength degradation. for strong interface CMCs exist in textile composites Some thermodynamic calculations have been performed [71]: fiber bundles fail as units, but neighboring bundles to identify the dominant reaction products formed when are protected from stress concentration by easy splitting Bn and Sic are present with oxygen [63]. These have between bundles; and 3D architectures bind failed fiber been limited to a temperature of 1100"C, considerably bundles together to large strains, giving exceptional values higher than the pest temperature. The kinetics and of work of fracture Model brittle/brittle composites have thermodynamics of these reactions and their effects been devised to demonstrate the benefits of interlocking on strength degradation in the pest Icginc are poorly 3D architectures [72] understood The growing recognition of the prevalence of the em- Figure 2 brittlement phenomenon in virtually all SiC-based CMC has led to the development of all-oxide CMCs. These are described in a subsequent section of this review. Barring where SiC-based systems must endure long term exposure, the design stress will have to be limited to the matrix cracking stress,as this represents the threshold below which embrittlement is suppressed. Tw o strategie fully dense matrices, produced, for example, by melt infiltration, rather than chemical vapor infiltration(which usually results in large pores at which cracks initiate); or the use of hybrid laminates comprising alternating layers of fiber reinforced CMCs and fully dense ceramic sheets The latter approach has been demonstrated [64], although 2 it is expected to be limited in its use to components with those which can be using a tile construction Bending fracture in an oxide-oxide CN with no fiber coating (strong interface crack pal Materials development BD Dalgleish, U Ramamurty, and CG Levi. The most eye-catching advance of the last two years in chemistry has been the advent of monazites and closely related structures, for example, xenotimes, as terface coatings and matrices in all-oxide composites. Wood-like fracture has now also been observed in When prototypical LaPO4 is deposited on Al2O3 fibers weak-interface systems, including fibrous Si3N4-BN [73] (with care to maintain accurate 1: 1 La: P stoichiometry), and monazite composites a weak interface is formed which is extremely stable up to at least 1600C [65, 66]. The processing required While the progress with all-oxide CMCs is exciting, thei appears to be simple and reproducible enough and the inherent disadvantages persist. Oxides generally exhibit base materials sufficiently low in cost that monazites may higher thermal expansion and lower thermal conductivity well eliminate degradation by interphase reactions as a life than SiC-based CMCs and will therefore have to sustain imiting process in all-oxide composites. If so, this is a higher stresses and temperatures in thermal applications major breakthrough. There is substantial new activity in Moreover, the creep resistance of currently available oxide monazite processing and applications fibers is markedly inferior to that of SiC fibers (e.g Tough oxide-oxide CMCs with strong fber/matrix inter- ccs have now also bccn dcmonstratcd. Crack defection Problcms remain with chemical stability in carbon fiber characteristics like those seen in wood have been achieved CMCs [74-76 and oxide-oxide Cmcs with bn interface without fiber/matrix debonding by bundling fibers in a coatings [77]
670 Ceramics, composites and intergrowths stable, although a critical assessment of their performance under conditions that accentuate the embrittlement has yet to be performed. A rudimentary model has been developed to predict the rupture time under static loading (60). More refined models are needed to take into account in a more realistic way the nature of the oxidation processes and the mechanisms responsible for fiber strength degradation. Some thermodynamic calculations have been performed to identify the dominant reaction products formed when BN and SIC are present with oxygen [63]. These have been limited to a temperature of llOo”C, considerably higher than the pest temperature. The kinetics and thermodynamics of these reactions and their effects on strength degradation in the pest regime are poorly understood. The growing recognition of the prevalence of the embrittlement phenomenon in virtually all Sic-based CMCs has led to the development of all-oxide CMCs. These are described in a subsequent section of this review. Barring a significant materials breakthrough, in cases where Sic-based systems must endure long term exposure, the design stress will have to be limited to the matrix cracking stress, as this represents the threshold below which embrittlement is suppressed. Two strategies for improving upon the cracking stress could be adopted: stronger, fully dense matrices, produced, for example, by melt infiltration, rather than chemical vapor infiltration (which usually results in large pores at which cracks initiate); or the use of hybrid laminates comprising alternating layers of fiber reinforced CMCs and fully dense ceramic sheets. The latter approach has been demonstrated [64], although it is expected to be limited in its use to components with relatively simple geometry or those which can be produced using a tile construction. Materials development The most eye-catching advance of the last two years in chemistry has been the advent of monazites and closely related structures, for example, xenotimes, as interface coatings and matrices in all-oxide composites. When prototypical Lap04 is deposited on A1203 fibers (with care to maintain accurate 1:l La:P stoichiometry), a weak interface is formed which is extremely stable up to at least 16OO’C [65**,66*]. The processing required appears to be simple and reproducible enough and the base materials sufficiently low in cost that monazites may well eliminate degradation by interphase reactions as a life limiting process in all-oxide composites. If so, this is a major breakthrough. There is substantial new activity in monazite processing and applications. Tough oxide-oxide CMCs with strong fiber/matrix interfaces have now also been demonstrated. Crack deflection characteristics like those seen in wood have been achieved without fiber/matrix debonding by bundling fibers in a porous matrix that offers easy splitting paths [67”,68]. (Other wood-like ceramics presented recently are really porous monoliths [69].) The propensity for splitting in CMCs can be enhanced by compressive residual stresses in the matrix [70]. Weak matrices and fiber entanglement within fiber bundles must also favour splitting [ZS]. Precedents in polymer and carbon-carbon composites suggest that especially effective toughening mechanisms for strong interface CMCs exist in textile composites [71]: fiber bundles fail as units, but neighboring bundles are protected from stress concentration by easy splitting between bundles: and 3D architectures bind failed fiber bundles together to large strains, giving exceptional values of work of fracture. Model brittle/brittle composites have been devised to demonstrate the benefits of interlocking 3D architectures [72*]. Figure 2 Bending fracture in an oxide-oxide CMC made from woven fiber tows with no fiber coating (strong interfaces). Note the irregular crack path and the extent of fiber bundle pullout, reminiscent of wood. (Courtesy BD Dalgleish, U Ramamurty, and CG Levi.) Wood-like fracture has now also been observed in weak-interface systems, including fibrous SixNh-BN [73] and monazite composites. While the progress with all-oxide CMCs is exciting, their inherent disadvantages persist. Oxides generally exhibit higher thermal expansion and lower thermal conductivity than Sic-based CMCs and will therefore have to sustain higher stresses and temperatures in thermal applications. Moreover, the creep resistance of currently available oxide fibers is markedly inferior to that of Sic fibers (e.g., Hi-NicalonThr). Problems remain with chemical stability in carbon fiber CMCs [74-761 and oxide-oxide CMCs with BN interface coatings [77]