2604 Journal of the American Ceramic Socien-Kerans et aL. Vol. 85. No. 1I Consideration of protection of fibers by residual coating lay on interfacial stresses and sliding friction. Realization that rough- es the issue of the degree of protection that might be expecte ness misfit effects can be substantial in oxide coatings has led to nra has discussed the issue of Sic-fiber protection from reexamination of conventional composites for coating thicknes oxidation in some detail. It is evident that very thin coatings can compliance effects. Modeling has shown that roughness in slow oxidation only to a limited degree. Small-diameter fibers- creases the compressive radial stress in a hypothetical uncoated Sic filaments are typically 8-12 um in diameter--are desirable Nicalon fiber/SiC composite from "150 MPa before sliding to 450 for easy handling, weaving, and shape-making, but the surface/ MPa after sliding. These stresses are decreased by 1/3 by including volume ratio is very high. Consequently, oxidation depths that are a 0.5 um thick carbon coating, therefore, changes in coating thickness can be expected to affect debond length and composite properties. In general, oxides are less compliant than carbon and BN; therefore, thicker coatings are required to similarly accom- n modate misfit stresses. Assuming a Nicalon/SiC composite and a a. MC behavior also depends strongly on the fiber/matrix sliding practical lower limit of 70 GPa for the elastic modulus of a porous iction. The ultimate strength, strain-to-failure, matrix crack oxide, the compliance provided by 500 nm of carbon requires 2 spacing, and toughness are affecte Coulomb friction is um of oxide. If coatings of such thickness are not practical proportional to the radial clamping stress on the fiber, which can suitable friction levels may need to be engineered in other ways be caused by residual stress from differential thermal expansion or e.g., by controlling roughness, matrix compliance, and residual els and experiments focus on residual stresses, ,/- but, re- thickness are also a large volume fraction of the composite and can cently, more attention has been given to roughness-induced stress- affect other composite properties, such as modulus, thermal A large roughness effect on sliding friction has been conductivity, and thermal expansion. Astute design allows for the shown by fiber push back or"seating drop"measurements. 3 effects on composite properties. 6 nitial modeling of the roughness effectis based on an approx- mation that debond roughness of amplitude h causes a mismatch strain of h/R, where R, is the fiber radius, that adds to the thermal (6) Effects of Coating Properties on Composite Analysis mismatch strain. Experiments show that this captures major Many calculations of radial clamping stress during fiber/matrix debonding and sliding consider only the thermoelastic properties spects of the behavior for many interfacial crack roughness of the fiber and matrix. The discussion above implies that serious geometries and. for most systems. during sliding of long fiber lengths. However, modeling has shown that the effect of roug errors may result. A rigorous treatment of the coating elastic ness in the early stages of debond crack propagation(Fig. 7)can effects exists 7 but the results are not easily incorporated into be much more pronounced and can have a significant influence on existing models of behavior. An approach that utilizes an approx omposite properties. This effect is due to the initial unseating of imation of this work in a method that represents the coated fiber by the matching rough surfaces just behind the crack tip. In this an"effective "( transversely isotropic) fiber in simple fiber/matrix region, the work required to further compress the fiber and matrix composites allows simple inclusion of coating elasticity in exi to accommodate the misfit is done. Furthermore the sliding analyses. This work also indicates that many conventional urfaces are not parallel to the fiber axis therefore, there is a analyses that have neglected carbon and Bn coatings in a Nicalon/ Sic system are significantly in error, Plots of normalized elastic beehponent of applied force that increases the friction. Perhap the modulus and coefficient of thermal expansion(CTE)for isotro treated-fiber Sic composite system discussed earlier, a rough interface model is necessary to decrease pushout data, and rough ometries for which th work well for compliant(carbon, BN) coating thickness up to 6 ness appears to be the primary source of the high friction that of the fiber radius, and they give reasonable approximations for dictates the very good fracture properties. ,0 Models of such thickness up to 10%. The thickness constraints relax somewhat A-Tocesses are now available and can be used to study debonding with incr, sg S coating stiffness. Other limitations are discussed ughness contributions to composite behavior. 7, 4 Effects pre- dicted for oxide fiber coatings are discussed later elsewhere This approach is applicable to many models that assume transversely isotropic fibers. For example, effective fiber pr (5) Interfacial Layer Compliance ties can be directly used in the shear-lag models of fiber pullout Although the coating is not often explicitly considered in pushout, ,8 as well as the Budiansky-Hutchinson-Evans nalysis, the compliance of the coating can have significant effects (BHE)model for matrix-cracking stress (7 Necessary Values of Interfacial Toughness and Friction Many CMCs fit in one of two categories: those with negligible interfacial strength, moderate to low interfacial friction, and toug (bond Crack-tip behavior, and those with high interfacial strength and elastic behavior. From these categories, it often has been inferred that toughness. s, u, When combined with the ease of using one parameter to describe the interface, this practice has led to the ssumption of zero interfacial strength and constant low interfacial friction(T)in most fracture models. 75.,90 Nicalon/C/SiC composites made with fibers treated to enhand Matrix oating/fiber bond strength",o evidence interface properties that defy common assumptions regarding what is required for good te behavior. Composites made with treated fibers have 30% higher tensile strength(from 250 to 350 MPa) at the same strain-to-failure. much finer matrix crack Fig. 7. Illustration of the effect of interfacial cantly different stress-strain behavior(Fig. 9). The change is e debonding progressing away from a matrix e attributed to interfacial friction(T)that increases from -5 to -150 sion. Three different reg labeled I.II MPa. Strong and tough composites with high strain-to-failure Roughness amplitude, h, period, 2d, and R. are the (0.5%)are observed even when T 370 MPa. The high mportant parameters that influence interfacial friction. has been attributed to the decrease in effective
Consideration of protection of fibers by residual coating layers raises the issue of the degree of protection that might be expected. Luthra57 has discussed the issue of SiC-fiber protection from oxidation in some detail. It is evident that very thin coatings can slow oxidation only to a limited degree. Small-diameter fibers— SiC filaments are typically 8–12 m in diameter—are desirable for easy handling, weaving, and shape-making, but the surface/ volume ratio is very high. Consequently, oxidation depths that are insignificant in monolithics damage fibers. (4) Interfacial Friction CMC behavior also depends strongly on the fiber/matrix sliding friction. The ultimate strength, strain-to-failure, matrix crack spacing, and toughness are affected.75,76 Coulomb friction is proportional to the radial clamping stress on the fiber, which can be caused by residual stress from differential thermal expansion or misfit from roughness at the debonding interface.73,77 Most models and experiments focus on residual stresses,73,78–80 but, recently, more attention has been given to roughness-induced stresses.71,81–83 A large roughness effect on sliding friction has been shown by fiber push back or “seating drop” measurements.82,83 Initial modeling of the roughness effect73 is based on an approximation that debond roughness of amplitude h causes a mismatch strain of h/Rf , where Rf is the fiber radius, that adds to the thermal mismatch strain. Experiments show that this captures major aspects of the behavior for many interfacial crack roughness geometries and, for most systems, during sliding of long fiber lengths.77 However, modeling has shown that the effect of roughness in the early stages of debond crack propagation (Fig. 7) can be much more pronounced and can have a significant influence on composite properties. This effect is due to the initial unseating of the matching rough surfaces just behind the crack tip. In this region, the work required to further compress the fiber and matrix to accommodate the misfit is done. Furthermore, the sliding surfaces are not parallel to the fiber axis; therefore, there is a component of applied force that increases the friction. Perhaps the best example of a system where this effect is important is the treated-fiber SiC composite system discussed earlier; a roughinterface model is necessary to decrease pushout data, and roughness appears to be the primary source of the high friction that dictates the very good fracture properties.25,68 Models of such processes are now available and can be used to study debonding roughness contributions to composite behavior.71,84 Effects predicted for oxide fiber coatings are discussed later. (5) Interfacial Layer Compliance Although the coating is not often explicitly considered in analysis, the compliance of the coating can have significant effects on interfacial stresses and sliding friction. Realization that roughness misfit effects can be substantial in oxide coatings has led to reexamination of conventional composites for coating thickness/ compliance effects.85 Modeling has shown that roughness increases the compressive radial stress in a hypothetical uncoated Nicalon fiber/SiC composite from 150 MPa before sliding to 450 MPa after sliding. These stresses are decreased by 1/3 by including a 0.5 m thick carbon coating; therefore, changes in coating thickness can be expected to affect debond length and composite properties. In general, oxides are less compliant than carbon and BN; therefore, thicker coatings are required to similarly accommodate misfit stresses. Assuming a Nicalon/SiC composite and a practical lower limit of 70 GPa for the elastic modulus of a porous oxide, the compliance provided by 500 nm of carbon requires 2 m of oxide.86 If coatings of such thickness are not practical, suitable friction levels may need to be engineered in other ways, e.g., by controlling roughness, matrix compliance, and residual stress state, or by other deformation mechanisms. Coatings of such thickness are also a large volume fraction of the composite and can affect other composite properties, such as modulus, thermal conductivity, and thermal expansion. Astute design allows for the effects on composite properties.86 (6) Effects of Coating Properties on Composite Analysis Many calculations of radial clamping stress during fiber/matrix debonding and sliding consider only the thermoelastic properties of the fiber and matrix. The discussion above implies that serious errors may result. A rigorous treatment of the coating elastic effects exists,87 but the results are not easily incorporated into existing models of behavior. An approach that utilizes an approximation of this work in a method that represents the coated fiber by an “effective” (transversely isotropic) fiber in simple fiber/matrix composites allows simple inclusion of coating elasticity in existing analyses.88 This work also indicates that many conventional analyses that have neglected carbon and BN coatings in a Nicalon/ SiC system are significantly in error. Plots of normalized elastic modulus and coefficient of thermal expansion (CTE) for isotropic “effective” fibers are given in Fig. 8. There are limits to the geometries for which this approach yields good results. The plots work well for compliant (carbon, BN) coating thickness up to 6% of the fiber radius, and they give reasonable approximations for thickness up to 10%. The thickness constraints relax somewhat with increasing coating stiffness. Other limitations are discussed elsewhere.88 This approach is applicable to many models that assume transversely isotropic fibers. For example, effective fiber properties can be directly used in the shear–lag models of fiber pullout and pushout,72,81 as well as the Budiansky–Hutchinson–Evans (BHE) model89 for matrix-cracking stress. (7) Necessary Values of Interfacial Toughness and Friction Many CMCs fit in one of two categories: those with negligible interfacial strength, moderate to low interfacial friction, and tough behavior; and those with high interfacial strength and elastic behavior. From these categories, it often has been inferred that negligible interfacial strength and low friction are necessary for toughness.75,90,91 When combined with the ease of using one parameter to describe the interface, this practice has led to the assumption of zero interfacial strength and constant low interfacial friction () in most fracture models.75,90 Nicalon/C/SiC composites made with fibers treated to enhance coating/fiber bond strength25,68 evidence interface properties that defy common assumptions regarding what is required for good composite behavior.25 Composites made with treated fibers have 30% higher tensile strength (from 250 to 350 MPa) at the same strain-to-failure, much finer matrix crack spacing, and significantly different stress–strain behavior (Fig. 9). The change is attributed to interfacial friction () that increases from 5 to 150 MPa.67 Strong and tough composites with high strain-to-failure (0.5%) are observed even when 370 MPa. The high composite strength has been attributed to the decrease in effective Fig. 7. Illustration of the effect of interfacial roughness during progressive debonding progressing away from a matrix crack in a composite under tension. Three different regions, labeled I, II, and III, can be envisioned. Roughness amplitude, h, period, 2d, and fiber radius, R, are the most important parameters that influence interfacial friction.71 2604 Journal of the American Ceramic Society—Kerans et al. Vol. 85, No. 11
November 2002 Interface Design for Oxidation-Resistant Ceramic Composites 2605 1.3 0.9 10(c) t/R<6% 0.8 20(BN) s.12/Q。=12575100C 0.7 0.75 0.6 0 0.5 t<0.5μm:E≤E 04 00.2040.60.8 -0.1-0.0500.050.10.150.20.25 R}{(Ect4}1+0.5h}tR}{(axa)-1}{1+0.5(E,/E} Fig. 8. Universal plots can be used to obtain the properties of an"effective" transverse fiber that can be substituted for a fiber pl odels that use transversely isotropic moduli and CTE: plots of (a)effective modulus and CTE in the transverse direction for fibers lots are a good approximation for up to 0.5 um thick(n)coating on an 8 um fiber radiL bols e and a refer to the modulus and Cte, subscripts t, c, and f stand for the transverse, coating, and fiber, respectively. Asterisk(* ffective properties.(Plot(b) was corrected for an er in the oniginal reference, where the matrix on the right-hand side in Eq. (8)should be gauge length of bridging fibers resulting from short debond lengths oating-it can be as high as 0.7 for an elastic anisotropy that are, in turn, a consequence of high T. As discussed earlier. om the He and Hutchinson%,60 analysis is not satisfied matrix cracks in high-strength material deflect into multiple the coating. A similar discrepancy has been noted for interfacial cracks, rather than a single debond. Therefore, crack on criteria using a laminate geometry. Although this deflection for this CMC is decided primarily by fracture anisotropy result is not well understood, it is encouraging with regard to the within the coating, rather than at the coating/fiber or coating/ development of alternative coatings in that the fracture energies matrix interface. Unusual fiber pushout load-deflection curves and the sliding friction may not be required to be as low as uggest substantial effects of rough interfaces, and subsequent previously thought. In any event, many of the coating approaches analysis implies that the critical strain energy to propagate cracks discussed later are likely to exhibit sufficiently high fracture in this interfacial region may be as high as 25 J/m". This is more energy and friction to greatly restrict debond lengths. It is helpful than half the fracture energy across the strongest graphite planes to know that, although the composites discussed above exhibit The criterion of fracture energy anisotropy of 1/4 or less( for an matrix crack spacings of from one to three fiber diameters, J d prc fiber 0.6 LONGITUDINAL TENSILE STRAIN(%) Fig 9. Tensile stress-strain behaviors in tension measured on the two-dimensional SiC/SiC composites fabricated from ()untreated or()treated Nicalon (ceramic grade)fibers. Complex crack deflection within the coating on treated fibers(schematic upper left)leads to higher friction than smooth interfacial ailure with untreated fibers (lower right)
gauge length of bridging fibers resulting from short debond lengths that are, in turn, a consequence of high . As discussed earlier, matrix cracks in high-strength material deflect into multiple interfacial cracks, rather than a single debond.67 Therefore, crack deflection for this CMC is decided primarily by fracture anisotropy within the coating, rather than at the coating/fiber or coating/ matrix interface. Unusual fiber pushout load–deflection curves suggest substantial effects of rough interfaces, and subsequent analysis implies that the critical strain energy to propagate cracks in this interfacial region may be as high as 25 J/m2 . 66 This is more than half the fracture energy across the strongest graphite planes. The criterion of fracture energy anisotropy of 1/4 or less (for an isotropic coating—it can be as high as 0.7 for an elastic anisotropy of 6) from the He and Hutchinson59,60 analysis is not satisfied, even in the coating. A similar discrepancy has been noted for deflection criteria using a laminate geometry.64 Although this result is not well understood, it is encouraging with regard to the development of alternative coatings in that the fracture energies and the sliding friction may not be required to be as low as previously thought. In any event, many of the coating approaches discussed later are likely to exhibit sufficiently high fracture energy and friction to greatly restrict debond lengths. It is helpful to know that, although the composites discussed above exhibit matrix crack spacings of from one to three fiber diameters, Fig. 8. Universal plots can be used to obtain the properties of an “effective” transversely isotropic fiber that can be substituted for a fiber plus coating in models that use transversely isotropic moduli and CTE: plots of (a) effective modulus and (b) effective CTE in the transverse direction for fibers with coatings. Plots are a good approximation for up to 0.5 m thick (t) coating on an 8 m fiber radius (R). Symbols E and refer to the modulus and CTE, respectively; subscripts t, c, and f stand for the transverse, coating, and fiber, respectively. Asterisk () denotes effective properties.88 (Plot (b) was corrected for an error in the original reference, where the matrix on the right-hand side in Eq. (8) should be inverted.) Fig. 9. Tensile stress–strain behaviors in tension measured on the two-dimensional SiC/SiC composites fabricated from (I) untreated or (J) treated Nicalon (ceramic grade) fibers. Complex crack deflection within the coating on treated fibers (schematic upper left) leads to higher friction than smooth interfacial failure with untreated fibers (lower right).350 November 2002 Interface Design for Oxidation-Resistant Ceramic Composites 2605
2606 Journal of the American Ceramic Socien-Kerans et aL. Vol. 85. No. 1I implying very short debond lengths, they also demonstrate high and coating surface roughnesses. o Therefore, if debonding is strength and toughness. Nevertheless, there is such a thing as within a coating and the crack meanders in the coating, a thinner debond lengths that are too short, even though that value is coating may decrease the fracture surface roughness and, there- considerably less than has been widely assumed before analysis of fore, increase toughness. If debonding initiates and remains at the these composites. coating/fiber interface, fracture surface roughness can be varied If the coating cracks ultimately reach the coating/fiber interface, only by modifying the fiber surface roughness as discussed in Section I(), the result is apparently benign. That However, if the debonding crack tends to approach th is, either(i)the interface, although stronger than the coating itself, surface via Mode I steps as it propagates and the interfa is weak enough to fail before the fiber, (ii) the changed local stress debond criterion is not satisfied( the situation discussed in state and short crack do not pose substantial stress concentration Il(), then greater coating thickness leads to longer debond on the fiber, or (iii) the resulting failure event is sufficiently late to ngths and higher toughness. yet allow excellent composite properties There are conflicts between some coating design parameters For example, a thicker coating can provide a route to lower friction by decreasing the compressive residual stresses, but it counters that Ill. Coating System Design and Evaluation effect by allowing higher fracture surface roughness, conversely, a thin coating may contribute to decreasing friction by minimizing (I General Interface Considerations roughness, but it may fail to relieve compressive residual stress In Ideally, the choice of composite constituents and geometry such cases, two coating layers might be considered. The weak should lead to the best balance of properties throughout the crack-deflecting layer would be thin, and the compliant layer for omponent service lifetime. In fact, many possibilities must await relief of residual compressive stress would be thick. The added the development of more constituent options, and optimizing complexity and expense is not desirable, but it may not be properties requires more highly sophisticated models. Eventually, prohibitive there may be more fibers, coatings, and matrices to choose from, but, presently, composite design is constrained by constituents for which there are no viable alternatives. Likewise. mechanistic (2) CMC Design Steps understanding is incomplete and often speculative. Nevertheless, it The first step in a logical CMC design sequence might be the is useful to take a logical approach that develops a framework into choices of fiber, coating, and matrix that are thermochemical which new tools can be fitted as they become available and that stable individually and in combination in the temperature range an provide insight for the refinement of approaches and environment of interest. In practice, that condition is often The first function of the coating or interface is that it must fail elaxed to include materials that react at acceptably slow rates. In before the fiber fails, thereby removing matrix -imposed stress fact, almost no structural materials are at thermodynamic equili concentrations on the fiber. The second function is that the coating rium in their use environments. a common example of acceptable must allow some sliding along the fiber/matrix debond after environmental instability is SiC 20,- Sio,+ co,, whe deflection. As discussed earlier. results from carbon- and bN- oxidation of SiC is defined by the low diffusion rates of oxygen in interface CMCs and models for their behavior suggest that the the SiO, scale. -s The second step that must be considered in debond may be at either the fiber/coating interface or within the design is processing. Processing should not excessively degrade coating. Coating design strategies can be based on either possibil the fiber or coating, therefore, matrix choice can be, and often is For debonding at the fiber/coating interface, allowable T-/ limited by the processing requirements values based on the He and Hutchinson criterion 9,60 vary with Excessive thermal stress in the coating may cause it to sp fiber/coating elastic modulus mismatch from -0.25 for matrix processing. This is particularly important for CMC mismatch to almost 0.7 when the fiber is 6 times stiffer than the atings, because they are designed to be weak, or weaki coating or matrix, as in SiC-reinforced glass-matrix CMCs.A to the fiber. Many excellent review articles discuss similar criterion based on interface strengths also can be used. 9 debonding of coatings from thermal stress(see, for example, Ref. For debonding within the coating, fracture anisotropy of the 96). If possible, choice of a fiber-coating combination with coating is the most important parameter. Although the He and minimal thermal stress should be considered. Debonding of Hutchinson criterion is a very useful guide as discussed earlier, it coatings during handling or weaving of coated fibers might be may not always be relevant because of effects such as debonding decreased by eliminating steps that bend fibers excessively ahead of the matrix crack. Excessive handling can be avoided by applying fiber coatings to Once debonding starts. it continue to propagate as a woven cloth or, better yet, the final fiber preform, as is often done cylindrical Mode Il crack between the fiber and matrix. The length in chemical vapor infiltration(CVD) processing, rather than to fiber of the debond crack(distance from the matrix crack plane to the tows. Preform-coating processes using other than cvi or in situ debond crack tip) depends on the interfacial sliding friction. The ocesses using fiber constituents have not been demonstrated lower the friction, the longer the crack and the greater the distance opposites that perform poorly may require careful evaluation to from the matrix crack plane required to transfer the excess load on determine if an ineffective coating, a damaged coating, or a the fiber back to the matrix. Higher friction along this Mode l damaged fiber is responsible ack causes the fiber stress to decrease faster with distance from Thermal expansion mismatch, roughness, and coating com the matrix crack plane. That is, the highly stressed portion of the ance interplay determine the postslidin ber is shorter, and there is a higher probability that fibers fracture and they should be considered simultaneously. For example, if the at or near the matrix crack plane. Therefore, toughne fiber is known to have a comparatively rough surface, residual decrease with increasing friction. Friction is controlled by residual stresses should be low and coating compliance should be high and applied stress, the fracture surface and the coeffi- cient of friction Residual stress is d CTEs. the coating thickness the fiber volup tion and the use ( Coating Evaluation emperature. In many systems, the coating is the most compliant The properties a coating must possess to provide good compo component; therefore, coating thickness can provide some adjust ite properties are not well-known. Hence, coating evaluation is ment of residual stresses. Specifically, where the coating is more most convincingly done via behavior of a composite that is compliant and/or has higher thermal expansion than the other analogous to a practically usable material form: for example, in constituents, thicker coatings can be expected to provide higher sheet form with fiber volume fraction >25%. This process can be toughness development of new fiber-coating and matrix-processing metha c time consuming and expensive. Each new approach can requi Potential opposite effects of coating thickness on crack pat should be considered. The maximum fracture surface roughness is Replacement of the CMC matrix with a glass matrix that is easier bounded by the sum of the coating thickness as well as the fiber to process also can be considered for coating evaluation, although
implying very short debond lengths, they also demonstrate high strength and toughness. Nevertheless, there is such a thing as debond lengths that are too short, even though that value is considerably less than has been widely assumed before analysis of these composites. If the coating cracks ultimately reach the coating/fiber interface, as discussed in Section II(3), the result is apparently benign. That is, either (i) the interface, although stronger than the coating itself, is weak enough to fail before the fiber, (ii) the changed local stress state and short crack do not pose substantial stress concentration on the fiber, or (iii) the resulting failure event is sufficiently late to yet allow excellent composite properties. III. Coating System Design and Evaluation (1) General Interface Considerations Ideally, the choice of composite constituents and geometry should lead to the best balance of properties throughout the component service lifetime. In fact, many possibilities must await the development of more constituent options, and optimizing properties requires more highly sophisticated models. Eventually, there may be more fibers, coatings, and matrices to choose from, but, presently, composite design is constrained by constituents for which there are no viable alternatives. Likewise, mechanistic understanding is incomplete and often speculative. Nevertheless, it is useful to take a logical approach that develops a framework into which new tools can be fitted as they become available and that can provide insight for the refinement of approaches. The first function of the coating, or interface, is that it must fail before the fiber fails, thereby removing matrix-imposed stress concentrations on the fiber. The second function is that the coating must allow some sliding along the fiber/matrix debond after deflection. As discussed earlier, results from carbon- and BNinterface CMCs and models for their behavior suggest that the debond may be at either the fiber/coating interface or within the coating. Coating design strategies can be based on either possibility. For debonding at the fiber/coating interface, allowable i r/ f z values based on the He and Hutchinson criterion59,60 vary with fiber/coating elastic modulus mismatch from 0.25 for zero mismatch to almost 0.7 when the fiber is 6 times stiffer than the coating or matrix, as in SiC-reinforced glass-matrix CMCs. A similar criterion based on interface strengths also can be used.91 For debonding within the coating, fracture anisotropy of the coating is the most important parameter. Although the He and Hutchinson criterion is a very useful guide, as discussed earlier, it may not always be relevant because of effects such as debonding ahead of the matrix crack. Once debonding starts, it must continue to propagate as a cylindrical Mode II crack between the fiber and matrix. The length of the debond crack (distance from the matrix crack plane to the debond crack tip) depends on the interfacial sliding friction. The lower the friction, the longer the crack and the greater the distance from the matrix crack plane required to transfer the excess load on the fiber back to the matrix. Higher friction along this Mode II crack causes the fiber stress to decrease faster with distance from the matrix crack plane. That is, the highly stressed portion of the fiber is shorter, and there is a higher probability that fibers fracture at or near the matrix crack plane. Therefore, toughness may decrease with increasing friction. Friction is controlled by residual and applied stress, the fracture surface roughness, and the coefficient of friction.81 Residual stress is determined by constituent CTEs, the coating thickness, the fiber volume fraction, and the use temperature.88 In many systems, the coating is the most compliant component; therefore, coating thickness can provide some adjustment of residual stresses. Specifically, where the coating is more compliant and/or has higher thermal expansion than the other constituents, thicker coatings can be expected to provide higher toughness. Potential opposite effects of coating thickness on crack path should be considered. The maximum fracture surface roughness is bounded by the sum of the coating thickness as well as the fiber and coating surface roughnesses.86 Therefore, if debonding is within a coating and the crack meanders in the coating, a thinner coating may decrease the fracture surface roughness and, therefore, increase toughness. If debonding initiates and remains at the coating/fiber interface, fracture surface roughness can be varied only by modifying the fiber surface roughness. However, if the debonding crack tends to approach the fiber surface via Mode I steps as it propagates and the interface/fiber debond criterion is not satisfied (the situation discussed in Section II(3)), then greater coating thickness leads to longer debond lengths and higher toughness. There are conflicts between some coating design parameters. For example, a thicker coating can provide a route to lower friction by decreasing the compressive residual stresses, but it counters that effect by allowing higher fracture surface roughness; conversely, a thin coating may contribute to decreasing friction by minimizing roughness, but it may fail to relieve compressive residual stress. In such cases, two coating layers might be considered. The weak, crack-deflecting layer would be thin, and the compliant layer for relief of residual compressive stress would be thick. The added complexity and expense is not desirable, but it may not be prohibitive. (2) CMC Design Steps The first step in a logical CMC design sequence might be the choices of fiber, coating, and matrix that are thermochemically stable individually and in combination in the temperature range and environment of interest. In practice, that condition is often relaxed to include materials that react at acceptably slow rates. In fact, almost no structural materials are at thermodynamic equilibrium in their use environments. A common example of acceptable environmental instability is SiC 2O2 3 SiO2 CO2, where oxidation of SiC is defined by the low diffusion rates of oxygen in the SiO2 scale.93–95 The second step that must be considered in design is processing. Processing should not excessively degrade the fiber or coating; therefore, matrix choice can be, and often is, limited by the processing requirements. Excessive thermal stress in the coating may cause it to spall during matrix processing. This is particularly important for CMC fiber coatings, because they are designed to be weak, or weakly bonded, to the fiber. Many excellent review articles discuss debonding of coatings from thermal stress (see, for example, Ref. 96). If possible, choice of a fiber–coating combination with minimal thermal stress should be considered. Debonding of coatings during handling or weaving of coated fibers might be decreased by eliminating steps that bend fibers excessively. Excessive handling can be avoided by applying fiber coatings to woven cloth or, better yet, the final fiber preform, as is often done in chemical vapor infiltration (CVI) processing, rather than to fiber tows. Preform-coating processes using other than CVI or in situ processes using fiber constituents have not been demonstrated. Composites that perform poorly may require careful evaluation to determine if an ineffective coating, a damaged coating, or a damaged fiber is responsible. Thermal expansion mismatch, roughness, and coating compliance interplay to determine the postsliding stresses and friction, and they should be considered simultaneously. For example, if the fiber is known to have a comparatively rough surface, residual stresses should be low and coating compliance should be high. (3) Coating Evaluation The properties a coating must possess to provide good composite properties are not well-known. Hence, coating evaluation is most convincingly done via behavior of a composite that is analogous to a practically usable material form: for example, in sheet form with fiber volume fraction 25%. This process can be time consuming and expensive. Each new approach can require development of new fiber-coating and matrix-processing methods. Replacement of the CMC matrix with a glass matrix that is easier to process also can be considered for coating evaluation, although 2606 Journal of the American Ceramic Society—Kerans et al. Vol. 85, No. 11
November 2002 Interface Design for Oxidation-Resistant Ceramic Co 2607 the change in chemistry, and probably elastic properties, may environmental resistance have been studied. Periodic matrix introduce some ambiguity in interpretation of results. cracks, nonlinear load displacement, and hysteresis during unload- Porous-matrix CMCs without fiber coatings can have attractive reload cycles have been observed, from which debond energies properties via distributed damage mechanisms, because cracks and the average friction(T)have been estimated. o However, full deflect around fibers without need for a coating (see Section confidence in validity of the results for property prediction in a full Tv(9). Matrix pore volume fractions at which significant tough- CMC has not been established. 100 ening is observed range from >30% to 15%.98,9%Hence, porous Oxide/oxide microcomposites have been fabricated and tested matrices complicate evaluation of fiber coatings, because the to evaluate the effectiveness of monazite (LapO,)and hibonite porous matrix and the coating can contribute to toughening. CaAl12Ojg)as interlayers in sapphire reinforced/Al,O,matrix matrix composites may be necessary for complete understanding as the control composites, the fractography and fracture strengths of damage mechanisms in coated-fiber composites with imper- were compared. For interlayer thicknesses of 0.3-0.5 um, bot fectly densified matrices-usually the case interlayers showed evidence of crack deflection; however debond lengths in hibonite-coated specimens were limited to just a smal (4) Micro- and Mini-CMCs fraction of the fiber diameter. Monazite-coated specimens showed Use of micro- or mini-CMCs for more-rapid evaluation has multiple matrix cracks and extensive debonding at the coating atrix interface. In both cases, the load-displacement curves were cylindrical matrix reinforced with one fiber, while a mini-CMC almost linear to failure, therefore, there was no unload-reload ses one or multiple fiber tows(200-3000 fibers/tow and up to hysteresis from which to measure interfacial friction. Failure four tows). The mechanical behavior of a mini-CMC is more strength was the only measurable mechanical parameter. The difficult to interpret, but it includes the statistical nature of fiber extent of nonlinearity in tension of specimens of any type with fracture and is more representative of a real composite. These high fiber modulus, straight fibers d low matrix volume fractio micro- and mini-CMCs are easier to fabricate than full cMcs. and must be small. The evaluation of the results was based on the relaxed sintering constraints on matrix densification can allow hypothesis that, even if the coating and matrix volume fraction is denser matrices to be more easily made. 4 03 Most such tests very low, there is severe degradation in apparent fiber strength if ave been limited to carbon and bn fiber/matrix interfaces and there is no mechanism to deflect cracks. The matrix and coating ostly CVI-SiC matrices. Effects of fiber surface treatments or crack at relatively low strain, and, unless the crack deflects, it ac coating procedures on interface properties and evaluation of as a large flaw in the fiber. In this experiment, composite strengths Sapph I mm Hibonite TM-DAR :bonded Su √ atrix dislodge AlumIt 2 2 CMC-Control 1. 18 GPa 0 1.18 G -1 Eiber-1450.C,2h 2 2.33 GPa m-lI. Fiber-Hibonite 3 Fiber CMCMonazite 2.25 GPa m=103 m=577 -10.500.51 1.5 1-0.500.511.5 Ln I Stress, GPa I Ln Stress, GPa Fig. 10. Single-filament sapphire fiber reinforced/Al,O, matrix microcomposites tested in tension (a) cracks deflect within the hibonite interface but by matrix regions that fell off during the test; (c)and(d) ites with coatings have almost the same mean strengths as the control composite tes with coatings, but the Weibull modulus is higher, about the same as the coated fibers. Results imply that the matrix is not sufficiently dense for evaluation of the coatings, because even the control samples have high microcomposite strength
the change in chemistry, and probably elastic properties, may introduce some ambiguity in interpretation of results. Porous-matrix CMCs without fiber coatings can have attractive properties via distributed damage mechanisms, because cracks deflect around fibers without need for a coating (see Section IV(9)). Matrix pore volume fractions at which significant toughening is observed range from 30% to 15%.98,99 Hence, porous matrices complicate evaluation of fiber coatings, because the porous matrix and the coating can contribute to toughening. Therefore, better understanding of damage mechanisms in porousmatrix composites may be necessary for complete understanding of damage mechanisms in coated-fiber composites with imperfectly densified matrices—usually the case. (4) Micro- and Mini-CMCs Use of micro- or mini-CMCs for more-rapid evaluation has received increasing attention.100,101 A micro-CMC is defined as a cylindrical matrix reinforced with one fiber, while a mini-CMC uses one or multiple fiber tows (200–3000 fibers/tow and up to four tows). The mechanical behavior of a mini-CMC is more difficult to interpret, but it includes the statistical nature of fiber fracture and is more representative of a real composite. These micro- and mini-CMCs are easier to fabricate than full CMCs, and relaxed sintering constraints on matrix densification can allow denser matrices to be more easily made.102,103 Most such tests have been limited to carbon and BN fiber/matrix interfaces and mostly CVI-SiC matrices. Effects of fiber surface treatments or coating procedures on interface properties100 and evaluation of environmental resistance have been studied.101 Periodic matrix cracks, nonlinear load displacement, and hysteresis during unload– reload cycles have been observed, from which debond energies and the average friction () have been estimated.100 However, full confidence in validity of the results for property prediction in a full CMC has not been established.100 Oxide/oxide microcomposites have been fabricated and tested to evaluate the effectiveness of monazite (LaPO4) and hibonite (CaAl12O19) as interlayers in sapphire reinforced/Al2O3 matrix composites.99 Using sapphire monofilaments in an Al2O3 matrix as the control composites, the fractography and fracture strengths were compared. For interlayer thicknesses of 0.3–0.5 m, both interlayers showed evidence of crack deflection; however debond lengths in hibonite-coated specimens were limited to just a small fraction of the fiber diameter. Monazite-coated specimens showed multiple matrix cracks and extensive debonding at the coating/ matrix interface. In both cases, the load–displacement curves were almost linear to failure; therefore, there was no unload–reload hysteresis from which to measure interfacial friction.99 Failure strength was the only measurable mechanical parameter. The extent of nonlinearity in tension of specimens of any type with high fiber modulus, straight fibers, and low matrix volume fraction must be small. The evaluation of the results was based on the hypothesis that, even if the coating and matrix volume fraction is very low, there is severe degradation in apparent fiber strength if there is no mechanism to deflect cracks. The matrix and coating crack at relatively low strain, and, unless the crack deflects, it acts as a large flaw in the fiber. In this experiment, composite strengths Fig. 10. Single-filament sapphire fiber reinforced/Al2O3 matrix microcomposites tested in tension: (a) cracks deflect within the hibonite interface but debond lengths are very short, much less than a fiber diameter, because of the roughness; (b) debonds present at the monazite/matrix interface are revealed by matrix regions that fell off during the test; (c) and (d) microcomposites with coatings have almost the same mean strengths as the control composites with no coatings, but the Weibull modulus is higher, about the same as the coated fibers. Results imply that the matrix is not sufficiently dense for evaluation of the coatings, because even the control samples have high microcomposite strength.99 November 2002 Interface Design for Oxidation-Resistant Ceramic Composites 2607
2608 Journal of the American Ceramic Socien-Kerans et aL. Vol. 85. No. 11 were relatively high for both coatings, considering the fiber the nature of the process. Although the desired Al,O, phase strengths were not significantly different from that of the fiber/ How ever, the stability of the interface coating during CVD matrix control specimens, although coated-fiber composites had processing is unknown and likely to be a major issue because of higher Weibull moduli. The lack of difference in strength is the use of gaseous hydrogen and Co in the process attributed to the porosity in the matrix; porous-matrix composites This has led to the use of glass matrices to test coating concepts are known to perform well without interface treatments(see next section). The results imply that the matrix density needs to be Allied Signal, Inc(now 85% to evaluate novel interface strategies reliably. 9 Honeywell), Morristown, N)polymer-derived glass as matrix shows some promise, although the matrices remain far from ideal The processing of even minicomposites having a dense oxide blackglas yields a matrix that is locally dense but filled with matrix can be challenging. Use of chemical vapor deposition (CVD) to deposit oxides remains in the developmental stage array of shrinkage microcracks. Oxide-fiber-reinforced minicom- CVD-deposited Al2O, matrices are amorphous and do not bond posites having a dense but microcracked glassy matrix of Black- readily to coated or uncoated fiber tows, which causes debonding, glas have been used in two studies to test oxidation-resistant even in control specimens. There is no known work on In one study, hnique was used t polycrystalline-oxide-matrix composites with high enough matrix 610(3M Corp, St Paul, MNyBlackglas composites with and densities to definitively suppress the mechanism of debonding vi matrix cracking(say 90%) CVI of dense stable polycrystallin minicomposites with the fiber coatings had significar oxides is made difficult by the formation of amorphous ultimate strengths than the uncoated control specime metastable oxides(which later crystallize or transform, introducin another study, porous oxide(zrO2-SiO2 mixture) and significant stresses and cracking)and by the inability to reach were evaluated in Nextel 720m-reinforced Blackglas porosity levels below the permeation threshold (-15%)because of coated and uncoated fibers were used as controls for comparison (a Control (uncoated b)BN CMC- Control (uncoated) 265MPa;m=88 0 742MP 383 MPa 55 5.5 5.5 65 Ln stress, MPa 1 Ln Stress, MPa] c) Porous ZrO2-SiO2 (d Monazite CMC. Control cMc· Contro coated) 265MPa;m=8.8 MPa; m=8.8 234 己5,5 CMC-porous Zro-sio cMc· Monazite 356MPa;m=6.0 353MPa;m=88 55 65 5.5 6.5 Ln Stress, MPa Ln stress, MPa] Fig. 11. Weibull plots of the strengths of minicomposites using dense Blackglas as the matrix show that porous(c)ZrO2-SiO2 and(d) monazite coatings on Nextel 720 fibers are as effective as the(b) BN-coated fibers (a) Control is significantly weaker than the fiber, showing that Blackglas might be a good model matrix to evaluate interface coatings
were relatively high for both coatings, considering the fiber strength degradation during processing; the strengths were greater than the matrix-cracking stresses (Fig. 10). However, the mean strengths were not significantly different from that of the fiber/ matrix control specimens, although coated-fiber composites had higher Weibull moduli. The lack of difference in strength is attributed to the porosity in the matrix; porous-matrix composites are known to perform well without interface treatments (see next section). The results imply that the matrix density needs to be 85% to evaluate novel interface strategies reliably.99 The processing of even minicomposites having a dense oxide matrix can be challenging. Use of chemical vapor deposition (CVD) to deposit oxides remains in the developmental stage. CVD-deposited Al2O3 matrices are amorphous and do not bond readily to coated or uncoated fiber tows, which causes debonding, even in control specimens.104 There is no known work on polycrystalline-oxide-matrix composites with high enough matrix densities to definitively suppress the mechanism of debonding via matrix cracking (say 90%).99 CVI of dense stable polycrystalline oxides is made difficult by the formation of amorphous or metastable oxides (which later crystallize or transform, introducing significant stresses and cracking) and by the inability to reach porosity levels below the permeation threshold (15%) because of the nature of the process. Although the desired Al2O3 phase remains difficult to process, work has been reported where almost 85%-dense ZrO2 has been deposited around woven preforms.105 However, the stability of the interface coating during CVD processing is unknown and likely to be a major issue because of the use of gaseous hydrogen and CO in the process. This has led to the use of glass matrices to test coating concepts. Preliminary work using BlackglasTM (Allied Signal, Inc. (now Honeywell), Morristown, NJ) polymer-derived glass as matrix shows some promise, although the matrices remain far from ideal. Blackglas yields a matrix that is locally dense but filled with an array of shrinkage microcracks. Oxide-fiber-reinforced minicomposites having a dense but microcracked glassy matrix of Blackglas have been used in two studies to test oxidation-resistant coatings. In one study, the technique was used to evaluate Nextel 610TM (3M Corp., St. Paul, MN)/Blackglas composites with and without porous lanthanum hexaluminate fiber coatings.106 The minicomposites with the fiber coatings had significantly higher ultimate strengths than the uncoated control specimens. In a another study, porous oxide (ZrO2–SiO2 mixture) and monazite were evaluated in Nextel 720TM-reinforced Blackglas.107 BNcoated and uncoated fibers were used as controls for comparison. Fig. 11. Weibull plots of the strengths of minicomposites using dense Blackglas as the matrix show that porous (c) ZrO2–SiO2 and (d) monazite coatings on Nextel 720 fibers are as effective as the (b) BN-coated fibers. (a) Control is significantly weaker than the fiber, showing that Blackglas might be a good model matrix to evaluate interface coatings.107 2608 Journal of the American Ceramic Society—Kerans et al. Vol. 85, No. 11