ournal Am. Ceram.Soc.85m2599-632(2002) Interface Design for Oxidation-Resistant Ceramic Composites Ronald J. Kerans, *f Randall S Hay, f Triplicane A. Parthasarathy, and Michael K. Cinibulk*f Materials and Manufacturing Directorate, AFRL/MLLN, Air Force Research Laboratory Wright-Patterson Air Force Base, Ohio 45433 fibers and matrices also suffer environmental through distributed damage mechanisms. These mechanisms r, changes in the mechanical properties of carbon- or lent on matrix cracks deflecting into fiber/matrix rolled interfaces after oxidation or enhanced oxidation of interfacial debonding cracks. Oxidation resistance of the fiber fibers or matrices after interface oxidation usually dominates CMC coatings often used to enable crack deflection is an importal behavior(see, for example, Refs. 5 and 18-20). This has mot limitation for long-term use in many applications. Research on vated research on more oxidation-resistant fiber coatings, viscous alternative, mostly oxide, coatings for oxide and non-oxide alant phases, and porous-matrix systems that do not require composites is reviewed. Processing issues, such as fiber coat ific interface control constituents (for concise reviews, see ngs and fiber strength degradation, are discussed. Mechanics Refs. 21 and 22). From a mechanistic standpoint, the substitution work related to design of crack deflecting coatings is also of Bn for carbon has been relatively straightforward; they have reviewed, and implications on the design of coatings and of very similar structures and elastic and fracture properties. BN and composite systems using alternative coatings are discussed. carbon are used as solid lubricants and can be expected to provide Potential topics for further research are identified low sliding friction. Substitution of oxides is a very different matter, and, unfortunately, lack of well-defined interface property L. Introduction equirements complicates the design and evaluation of alternative viable approaches for use in composites IE discovery that brittle ceramics can be made highly having non-oxide constituents can be further complicated by the tolerant by combining them in fiber/matrix composi sIte form need for stability and compatibility in strongly reducing processin (ceramic-matrix composite or CMC, continuous-fiber environments. In fact, most oxide-coating work to date has been on e or CFCC, and ceramic-fiber matrix composite or oxide fibers to be used in oxide matrices. Research on fiber CFMC) has spawned research spanning approximately three de- coating processes is also required. For example, coated fibers often ades. Early work revealed that deflection of matrix cracks to the dis splay severely degraded tensile strength, 3, 24 which has moti- fiber/matrix interface, leaving intact fibers behind the matrix crack vated research on mechanisms of degradation. tip, was essential for tough behavior. Crack deflection in mos Although development of oxidation-resistant interface control is CMCs has been effected by a relatively weak and compliant complex, there has been progress carbon coating applied to the fibers before matrix processing o (1) There are many interface design parameters, and they are formed in situ by fiber decomposition during matrix processing better understood However, long-term use of CMCs has been limited by several (2) Several more oxidation-resistant alternatives to carbon and forms of environmental degradation, the most pervasive of which bn have the correct crack deflection behavior. and some show as been oxidation of the fiber coatings promise for the correct fiber pullout behavior. 2-30 To improve oxidation resistance, BN has been substituted for (3) There has been progress toward viable fiber-coating carbon(see, for example, Refs. 7-17). Progress has been made on process ystems using BN, and the best Bn coatings demonstrate very (4) Definitive evidence of oxide coatings effecting character good properties. Nevertheless, although BN is a much better istic composite fracture and properties in true yarm-reinforced coating than carbon, it has much poorer oxidation resistance than composites has been observed for two different oxide coatings most candidate fiber and matrix constituents (Fig. 1). In this review, progress is summarized in a manner intended to ssist in developing guidelines for the design and evaluation of B. Marshall-contributing editor fiber coatings and to highlight the most interesting areas for further esearch. Strategies for oxidation-resistant coatings and relevant interface mechanics are critically reviewed. Progress and problems in coating of fibers are summarized. Section Il provides back- Manuscript No 188122 Received November 29, 2000, approved June 13, 2002. ground in the form of a brief review of historical aspects of interface oxidation. a discussion of the mechanics of crack Also affiliated with UES, Inc, Dayton, OH, under U.S. Air Force Contract No. deflection and sliding, the effects of coating properties F33615-96-C5258 posite behavior, and target values for interface parameters. Section Feature
Interface Design for Oxidation-Resistant Ceramic Composites Ronald J. Kerans,* ,† Randall S. Hay,* ,† Triplicane A. Parthasarathy,* ,‡ and Michael K. Cinibulk* ,† Materials and Manufacturing Directorate, AFRL/MLLN, Air Force Research Laboratory, Wright-Patterson Air Force Base, Ohio 45433 Fiber-reinforced ceramic composites achieve high toughness through distributed damage mechanisms. These mechanisms are dependent on matrix cracks deflecting into fiber/matrix interfacial debonding cracks. Oxidation resistance of the fiber coatings often used to enable crack deflection is an important limitation for long-term use in many applications. Research on alternative, mostly oxide, coatings for oxide and non-oxide composites is reviewed. Processing issues, such as fiber coatings and fiber strength degradation, are discussed. Mechanics work related to design of crack deflecting coatings is also reviewed, and implications on the design of coatings and of composite systems using alternative coatings are discussed. Potential topics for further research are identified. I. Introduction THE discovery that brittle ceramics can be made highly damage tolerant by combining them in fiber/matrix composite form (ceramic-matrix composite or CMC, continuous-fiber ceramic composite or CFCC, and ceramic-fiber matrix composite or CFMC) has spawned research spanning approximately three decades. Early work revealed that deflection of matrix cracks to the fiber/matrix interface, leaving intact fibers behind the matrix crack tip, was essential for tough behavior.1–6 Crack deflection in most CMCs has been effected by a relatively weak and compliant carbon coating applied to the fibers before matrix processing or formed in situ by fiber decomposition during matrix processing. However, long-term use of CMCs has been limited by several forms of environmental degradation, the most pervasive of which has been oxidation of the fiber coatings. To improve oxidation resistance, BN has been substituted for carbon (see, for example, Refs. 7–17). Progress has been made on systems using BN, and the best BN coatings demonstrate very good properties. Nevertheless, although BN is a much better coating than carbon, it has much poorer oxidation resistance than most candidate fiber and matrix constituents. CMC fibers and matrices also suffer environmental degradation. However, changes in the mechanical properties of carbon- or BN-controlled interfaces after oxidation or enhanced oxidation of fibers or matrices after interface oxidation usually dominates CMC behavior (see, for example, Refs. 5 and 18–20). This has motivated research on more oxidation-resistant fiber coatings, viscous sealant phases, and porous-matrix systems that do not require specific interface control constituents (for concise reviews, see Refs. 21 and 22). From a mechanistic standpoint, the substitution of BN for carbon has been relatively straightforward; they have very similar structures and elastic and fracture properties. BN and carbon are used as solid lubricants and can be expected to provide low sliding friction. Substitution of oxides is a very different matter, and, unfortunately, lack of well-defined interface property requirements complicates the design and evaluation of alternative interfaces. Identifying viable approaches for use in composites having non-oxide constituents can be further complicated by the need for stability and compatibility in strongly reducing processing environments. In fact, most oxide-coating work to date has been on oxide fibers to be used in oxide matrices. Research on fibercoating processes is also required. For example, coated fibers often display severely degraded tensile strength,23,24 which has motivated research on mechanisms of degradation. Although development of oxidation-resistant interface control is complex, there has been progress. (1) There are many interface design parameters, and they are better understood.25,26 (2) Several more oxidation-resistant alternatives to carbon and BN have the correct crack deflection behavior, and some show promise for the correct fiber pullout behavior.27–30 (3) There has been progress toward viable fiber-coating processes.23,31–37 (4) Definitive evidence of oxide coatings effecting characteristic composite fracture and properties in true yarn-reinforced composites has been observed for two different oxide coatings (Fig. 1). In this review, progress is summarized in a manner intended to assist in developing guidelines for the design and evaluation of fiber coatings and to highlight the most interesting areas for further research. Strategies for oxidation-resistant coatings and relevant interface mechanics are critically reviewed. Progress and problems in coating of fibers are summarized. Section II provides background in the form of a brief review of historical aspects of interface oxidation, a discussion of the mechanics of crack deflection and sliding, the effects of coating properties on composite behavior, and target values for interface parameters. Section D. B. Marshall—contributing editor Manuscript No. 188122. Received November 29, 2000; approved June 13, 2002. *Member, American Ceramic Society. † Air Force Research Laboratory. ‡ Also affiliated with UES, Inc., Dayton, OH, under U.S. Air Force Contract No. F33615-96-C-5258. journal J. Am. Ceram. Soc., 85 [11] 2599–632 (2002) Feature
2600 Journal of the American Ceramic Socien-Kerans et aL. Vol. 85. No. 11 fiber surfaces(a thin carbon layer over a thin SiO, layer) that is weak enough to deflect matrix cracks and protect the fibers from matrix crack stress concentrations. ,s Sliding between fiber and matrix, before and after the fibers fracture, further dissipates energy via friction. These mechanisms give CMCs the tolerance to local overload that makes them useful as structural materials Composites with no carbon layer fail catastrophically with low strength in the manner of poor-quality monolithics cally, had the Nicalon fiber actually been stoichiometric crystalline SiC, carbon layers would not have formed in situ, and attaining mechanically viable ceramic composites would have been more problematic, but perhaps hastened more-detailed understanding of the mechanics governing composite desi Early CMC studies measured strength and load-deflection behavior at room temperature, CMCs with carbon layers on the fibers demonstrated high strength, high strain-to-failure, and non- linear load-deflection behavior. However. when tested at (b temperatures, there was a substantial loss in strength above 900C (Fig. 2).,,+Initially, this was attributed to replacement of the carbon layer by Sio, that strongly bonded fibers to the matrix and allowed matrix cracks to propagate directly through fibers. 7 48 Recent work suggests that oxidative degradation of Nicalon fiber may contribute to composite strength loss to a degree comparable to direct effects of interface property changes. 49-52Nevertheles in either case, carbon interface oxidation allowing oxygen acces to the entire fiber surface area in a Cmc is the first degradation step. Above 1000oC, a self-sealing SiO, layer can prevent acces of oxygen to the interface. 48,53 However, at intermediate temper occurs from uninterrupted oxidation(Fig. 3). 20,48 so ength los atures, typically between 700 and 900C, significant st experiments, analytical modeling s and experiments on Nica- lon/C/SiC composites have contributed to the current under standing of this intermediate-temperature degradation. It has been argued that fibers(and coatings)do not oxidize in a crack-free CMC used at design stresses less then the matrix- cracker Fig. 1. (a)Fracture surface of Nextel 720 fiber/monazite fiber coating/ stress. Such an approach might be acceptable for preservation of the aluminosilicate matrix indicating that crack deflection occurred at or near interface when overloads are infrequent and design stresses are low ber/coating interfaces. Energy dispersive spectroscopy(EDS)analysis enough that cracks are not held open, or if there are mechanisms to indicates that the light phase is monazite and that it is essentially al ways al lightly loaded cracks. A sensible design using this approach left in the trough.( Fiber coating by AFRL/ML; composite by Composi trives to have the regions most likely to crack, the more highly Optics, Inc )(b) Fracture surface of Nextel 610/scheelite fiber coating/ stressed regions, at temperatures that are relatively benign alumina CerablakM matrix indicating that crack deflection occurred at or Although this approach has merit if the cra near fiber/coating interfaces( Coating and composite by Mc Dermott, In and Applied Thin Films, Inc. made sufficiently high and the application environment is well- known. it seems far from an ideal solution. All design stress calculations are approximations based on an idealized situation including mating of perfectly matching surfaces, absence of Ill discusses the design and evaluation of coatings and composites lefects and foreign matter, and predictable environments. These Section IV discusses specific approaches to interface control. For approximations work for metals, because ductile materials blunt completeness, BN coatings and porous-matrix composites also are flaws by local plastic deformation that otherwise cause local stress briefly reviewed in Section IV. Section V discusses coating concentrations. For CMCs, the equivalent local deformation is process technology and fiber degradation. Section VI summarizes local matrix cracking and a few broken fibers, which allows acces and speculates on future options. This review is intended to be a of the atmosphere to the composite interior. Furthermore, there is comprehensive critical review and to provide some thought- evidence that matrix cracking occurs in some CMCs well below provoking speculation on composite design and useful future the proportional limit. The fact that introduction of monolithic ceramics into structural applications has been slow and limited despite very high strength and thorough proof testing, provides circumstantial evidence for this point of view. At least occasional Il. Interface Properties and Mechanics local stress concentrations greater than the matrix-cracking stress almost 's exist in practice. Hence, the ideal composite Initial interest in CFCCs was generated by marketing of equires all constituents to be oxidation resistant, including the Nicalon fiber(Nippon Carbon Co Japan) and the fiber/matrix interface erceived availability of a fiber that had the nsity, creep, and oxidation resistance of sic and the high and fabrication ease of small-diameter filaments in a fiber tow However Nicalon (2) Initiation of Interfacial Cracks and Deflection of is not crystalline SiC, but instead is carbon-and oxygen-rich and atrix cracks Although in most respects Ni Crack deflection is the most important event for achievin excellent fiber, when exposed to high temperatures, it crystallizes tough composites; however, the complexities of the problem and to SiC, rejects carbon and oxygen, and shrinks slightly. ,4During of real materials require simplification for analysis, and confirma matrix processing, this decomposition can form a coating on the tion by experiment is problematic. The details of crack deflection
III discusses the design and evaluation of coatings and composites. Section IV discusses specific approaches to interface control. For completeness, BN coatings and porous-matrix composites also are briefly reviewed in Section IV. Section V discusses coating process technology and fiber degradation. Section VI summarizes and speculates on future options. This review is intended to be a comprehensive critical review and to provide some thoughtprovoking speculation on composite design and useful future work. II. Interface Properties and Mechanics Initial interest in CFCCs was generated by marketing of NicalonTM fiber (Nippon Carbon Co., Tokyo, Japan) and the perceived availability of a fiber that had the low density, creep, and oxidation resistance of SiC and the high strength and fabrication ease of small-diameter filaments in a fiber tow. However, Nicalon is not crystalline SiC, but instead is carbon- and oxygen-rich and nearly amorphous.38–41 Although in most respects Nicalon is an excellent fiber, when exposed to high temperatures, it crystallizes to SiC, rejects carbon and oxygen, and shrinks slightly.40,42 During matrix processing, this decomposition can form a coating on the fiber surfaces (a thin carbon layer over a thin SiO2 layer) that is weak enough to deflect matrix cracks and protect the fibers from matrix crack stress concentrations.3,38 Sliding between fiber and matrix, before and after the fibers fracture, further dissipates energy via friction. These mechanisms give CMCs the tolerance to local overload that makes them useful as structural materials. Composites with no carbon layer fail catastrophically with low strength in the manner of poor-quality monolithics.38,43–45 Ironically, had the Nicalon fiber actually been stoichiometric crystalline SiC, carbon layers would not have formed in situ, and attaining mechanically viable ceramic composites would have been more problematic, but perhaps hastened more-detailed understanding of the mechanics governing composite design. (1) Oxidation History Early CMC studies measured strength and load–deflection behavior at room temperature.3,4 CMCs with carbon layers on the fibers demonstrated high strength, high strain-to-failure, and nonlinear load–deflection behavior. However, when tested at high temperatures, there was a substantial loss in strength above 900°C (Fig. 2).18,46,47 Initially, this was attributed to replacement of the carbon layer by SiO2 that strongly bonded fibers to the matrix and allowed matrix cracks to propagate directly through fibers.47,48 Recent work suggests that oxidative degradation of Nicalon fiber may contribute to composite strength loss to a degree comparable to direct effects of interface property changes.49–52 Nevertheless, in either case, carbon interface oxidation allowing oxygen access to the entire fiber surface area in a CMC is the first degradation step. Above 1000°C, a self-sealing SiO2 layer can prevent access of oxygen to the interface.48,53 However, at intermediate temperatures, typically between 700° and 900°C, significant strength loss occurs from uninterrupted oxidation (Fig. 3).20,48,50,53 Model experiments,54 analytical modeling,55 and experiments on Nicalon/C/SiC composites20 have contributed to the current understanding of this intermediate-temperature degradation. It has been argued that fibers (and coatings) do not oxidize in a crack-free CMC used at design stresses less then the matrix-cracking stress.56 Such an approach might be acceptable for preservation of the interface when overloads are infrequent and design stresses are low enough that cracks are not held open, or if there are mechanisms to seal lightly loaded cracks.57 A sensible design using this approach strives to have the regions most likely to crack, the more highly stressed regions, at temperatures that are relatively benign. Although this approach has merit if the cracking stress can be made sufficiently high and the application environment is wellknown, it seems far from an ideal solution. All design stress calculations are approximations based on an idealized situation, including mating of perfectly matching surfaces, absence of defects and foreign matter, and predictable environments. These approximations work for metals, because ductile materials blunt flaws by local plastic deformation that otherwise cause local stress concentrations. For CMCs, the equivalent local deformation is local matrix cracking and a few broken fibers, which allows access of the atmosphere to the composite interior. Furthermore, there is evidence that matrix cracking occurs in some CMCs well below the proportional limit.58 The fact that introduction of monolithic ceramics into structural applications has been slow and limited, despite very high strength and thorough proof testing, provides circumstantial evidence for this point of view. At least occasional local stress concentrations greater than the matrix-cracking stress almost always exist in practice. Hence, the ideal composite requires all constituents to be oxidation resistant, including the fiber/matrix interface. (2) Initiation of Interfacial Cracks and Deflection of Matrix Cracks Crack deflection is the most important event for achieving tough composites; however, the complexities of the problem and of real materials require simplification for analysis, and confirmation by experiment is problematic. The details of crack deflection Fig. 1. (a) Fracture surface of Nextel 720 fiber/monazite fiber coating/ aluminosilicate matrix indicating that crack deflection occurred at or near fiber/coating interfaces. Energy dispersive spectroscopy (EDS) analysis indicates that the light phase is monazite and that it is essentially always left in the trough. (Fiber coating by AFRL/ML; composite by Composite Optics, Inc.) (b) Fracture surface of Nextel 610/scheelite fiber coating/ alumina CerablakTM matrix indicating that crack deflection occurred at or near fiber/coating interfaces. (Coating and composite by McDermott, Inc., and Applied Thin Films, Inc.) 2600 Journal of the American Ceramic Society—Kerans et al. Vol. 85, No. 11
Interface Design for Oxidation-Resistant Ceramic Composites 2 (a) 30l -0.0 25 800C Time(10 s Sio cOcO 250 CROSS-HEAD DISPLACEMENT ( IN L 0254 (b) PYROCARBON Fig. 3. (a) Measured oxidative weight change of a Nicalon/carbon/SiC omposite. Oxidation starts with a weight loss from carbon oxidation bove 900C, the oxidation of SiC results in plugs that limit further carbon oxidation; however, at 700 and 800C, the sealing does not happen, as indicated in the(b) schematic and associated statistics of the nterfaces are not well-known. The second problem is that there are no proven local failure criteria. That is, even if the stresses can be U 2。0 calculated, there is no appropriate failure criterion for a very small volume of material where bulk c flaw distribution is unknown. Energy-based analyses assume particular virtual crack extensions and may not be appropriate for 20 ng behavior on this scale. Continued progress can be expected with increasing comparison of analyses to experiments It is also not yet understood what level of detail a model must 10 apture to properly predict actual behavior. Treating coating layers as being properly described by properties of an infinitesimally thin nterface surely must be misleading in many situations. For example, the He and Hutchinson analysis considered the cri- terion for deflection of a mode i matrix crack to an interfacial crack in an ideal planar interface perpendicular to the matrix crack plane. They found that deflection should be expected only for ratios of interface toughness to "fiber" toughness less than a 005 certain value dependant on elastic properties but genera than about 1/4 This often has beel died without CROSS-HEAD DISPLACEMENT (IN) the details of crack deflection However if cracks deflect inside the coating, propagation of the crack in either sense is determined Fig. 2. Effect of oxidation on the mechanical behavior at 900%C in air of by coating fracture properties, i.e., the ratio of coating debond Nicalon-reinforced lithium aluminosilicate(LAS) matrix composite is mode to coating transmission mode fracture energies (T/I shown from the load-stress-displacement behavior at(a)room temperature where c indicates coating, r and z indicate crack surface normals in and at(b)900°cin lindrical coordinates with z along the fiber axis, and applied tractions are along +z )25 Fiber toughnesses are typically a few determine the interface property that must be engineered MPam, therefore, coating toughness can be higher than fiber mechanics analyses have contributed much to the level of toughness. A coating can fail the test of (debond fracture energy M standing of composite behavior, but there are two fiber fracture energy)< /4, but can deflect cracks, because the problems that limit the utility of the analyses in guiding co ratio of coating toughnesses for the two types of cracks do satisfy design. The first of these problems is that the properties, ge the criterion; that is, the coating is sufficiently anisotropic in
determine the interface property that must be engineered. Micromechanics analyses have contributed much to the level of understanding of composite behavior, but there are two pervasive problems that limit the utility of the analyses in guiding composite design. The first of these problems is that the properties, geometry, and associated statistics of the thin coatings and associated interfaces are not well-known. The second problem is that there are no proven local failure criteria. That is, even if the stresses can be calculated, there is no appropriate failure criterion for a very small volume of material where bulk properties do not apply and the flaw distribution is unknown. Energy-based analyses assume particular virtual crack extensions and may not be appropriate for predicting behavior on this scale. Continued progress can be expected with increasing comparison of analyses to experiments. It is also not yet understood what level of detail a model must capture to properly predict actual behavior. Treating coating layers as being properly described by properties of an infinitesimally thin interface surely must be misleading in many situations. For example, the He and Hutchinson analysis59 considered the criterion for deflection of a Mode I matrix crack to an interfacial crack in an ideal planar interface perpendicular to the matrix crack plane. They found that deflection should be expected only for ratios of interface toughness to “fiber” toughness less than a certain value dependant on elastic properties but generally less than about 1/4.59–61 This often has been applied without regard to the details of crack deflection. However, if cracks deflect inside the coating, propagation of the crack in either sense is determined by coating fracture properties, i.e., the ratio of coating debond mode to coating transmission mode fracture energies (c r/ c z, where c indicates coating, r and z indicate crack surface normals in cylindrical coordinates with z along the fiber axis, and applied tractions are along z.).25 Fiber toughnesses are typically a few MPam1/2; therefore, coating toughness can be higher than fiber toughness. A coating can fail the test of (debond fracture energy)/ (fiber fracture energy) 1/4, but can deflect cracks, because the ratio of coating toughnesses for the two types of cracks do satisfy the criterion; that is, the coating is sufficiently anisotropic in Fig. 2. Effect of oxidation on the mechanical behavior at 900°C in air of a Nicalon-reinforced lithium aluminosilicate (LAS) matrix composite is shown from the load-stress–displacement behavior at (a) room temperature and at (b) 900°C in air.18 Fig. 3. (a) Measured oxidative weight change of a Nicalon/carbon/SiC composite. Oxidation starts with a weight loss from carbon oxidation. Above 900°C, the oxidation of SiC results in plugs that limit further carbon oxidation; however, at 700° and 800°C, the sealing does not happen, as indicated in the (b) schematic.48 November 2002 Interface Design for Oxidation-Resistant Ceramic Composites 2601
Journal of the American Ceramic Socien-Kerans et aL. Vol. 85. No. 1I fracture energy to deflect cracks itself. This discussion also failure criterion leaves the matter open to speculation. This suggests that coatings that are intended to deflect cracks by failure sequence has been observed in laminates and in model fracture within the coating can be evaluated independently of the composites Nicalon/C/SiC composites made with fibers treated to promote rack direction from perpendicular to parallel to a fiber surface dence of interfacial failure preceding matrix crack impinge- Fig. 4(a), but there are other possibilities. Mode I interface cracks can form in the tensile stress field normal to the fiber surface ahead connect to debonded coating/fiber interfaces, with no deflection in f a matrix crack z(Fig 4(b). Modeling of an annular matrix the coating itself, i.e., the interface, not the coating, fails. In crack has predicted that, for most reasonable choices of properties, otherwise identical composites made with treated fibers, matrix cracks connect to diffuse cracks in the coating without debonding other failure event(e. g, fiber fracture)intervenes before the coating/fiber interface; ie, the coating itself fails.66, 68The the matrix crack can be driven to the interface. unless the inter interfacial toughness, friction, and composite strength are higher face is debonded ahead of it. 63 Interface stresses can be high for the treated-fiber CMC If the matrix crack runs through the enough to make interfacial debond ahead of the matrix crack a plausible mechanism, but lack of a completely understood local coatings on untreated fibers before the interface debonds, it deflects in the coating, as it does in the identical coating on the treated fiber; hence, the coating/fiber interface must fail before the matrix crack enters the coating. When the crack does pass through the coating, the elastic constraint of the fiber is mostly removed by the preexisting debond, and the crack runs directly to the debonded interface. If there are truly no material difference besides interface strength, a definitive sequence of events consistent with the model is implied. These composites have other interesting properties that are discussed in later sections Another deflection mechanism preceding matrix crack impinge ment has been suggested for a composite comprising SiC mono- filaments with successive coatings of carbon and TiB, in a glass matrix.9 The coating was calculated to be in triaxial compressio d model ested that matrix cracks would run to th coating only after coating or interfacial failure. Experiment re- LATRIX vealed that debond cracks ran very near the fiber surface except for ular-section C rings around the fiber with their peaks at the matrix crack planes. Shear stresses on planes in the approximate orientation of the sides of the C rings(about 45. from the matrix crack plane) were calculated to be the highest coating stresses that could lead to fracture. The suggested failure sequence was(i)the matrix crack approached the coating, (ii) the coating failed in shear on planes +45 from the crack plane, ahead of the crack, and formed C rings, (iii) the coating cracks turned parallel to the fiber urface at or near the fiber surface, and(iv) the matrix crack advanced until it joined the coating shear cracks at their intersec- tion(see Fig. 4(c)) A further possibility is growth of periodic echelon cracks In an analysis of thin laminates with a Mode I crack normal to the nterface. the maximum tensile stresses in the coatings were 45o allel to a "half-turn of the nto eries of parallel"periodic echelon"microcracks on these hi stress planes at the center of the coating. As the microcrack approached the coating/plate interfaces, they turned parallel to the interface and joined to form a debond. Evidence for a similar- appearing but different sequence of events has been observed in monazite interlayers in Al, O /Al,O3 laminates(Fig. 5).27In that case, the echelon cracks appeared after initial deflection of the main crack into the coating/ laminate interface. [二A Detailed fracture observations are difficult: therefore. the se- quence of events for fiber/matrix debonding in CMCs speculative Debonding mechanisms may vary with the particular composite and global stress state. The coating most ortant to crack deflection depends on subtle differences in onstituent elastic properties and residual stresses. In the ideal case, coatings are engineered material "components'" of a compo ite system selected for phase, microstructure, and geometry to promote a specific failure mechanism. Enhanced understanding of crack deflection is necessary to allow such a priori design Simple Fig. 4. Three possible sequences leading to crack deflection:(a)matrix crack grows into the coating and then bifurcates and turn models are often useful, but they can sometimes be misleading running parallel to the fiber surface in each direction(as well hence, detailed analysis and comparison of microstructure, crack on in the matrix),(b) coating or interface fails in the tensile deflection behavior, and analytical models may be necessary. The he matrix crack before arrival of the matrix crack at the interface region same considerations apply to the interpretation of micromechani- and (c)in the matrix crack at the coating, the crack bifurcates and turns as cal tests. For example, fiber pushout/pullout tests may not direct the coating fails in shear at an intermediate angle, then turns parallel to the measure the parameters that actually determine debonding during fiber surface at or near the fiber surface composite failure. It is even possible that debonding in single
fracture energy to deflect cracks itself. This discussion also suggests that coatings that are intended to deflect cracks by fracture within the coating can be evaluated independently of the fiber. Crack deflection is usually assumed to be a local change in crack direction from perpendicular to parallel to a fiber surface (Fig. 4(a)), but there are other possibilities. Mode I interface cracks can form in the tensile stress field normal to the fiber surface ahead of a matrix crack62 (Fig. 4(b)). Modeling of an annular matrix crack has predicted that, for most reasonable choices of properties, some other failure event (e.g., fiber fracture) intervenes before the matrix crack can be driven to the interface, unless the interface is debonded ahead of it.63 Interface stresses can be high enough to make interfacial debond ahead of the matrix crack a plausible mechanism, but lack of a completely understood local failure criterion leaves the matter open to speculation. This failure sequence has been observed in laminates64 and in model composites.65 Nicalon/C/SiC composites made with fibers treated to promote higher coating/fiber interface strengths also provide indirect evidence of interfacial failure preceding matrix crack impingement.66,67 In composites made with untreated fibers, matrix cracks connect to debonded coating/fiber interfaces, with no deflection in the coating itself; i.e., the interface, not the coating, fails. In otherwise identical composites made with treated fibers, matrix cracks connect to diffuse cracks in the coating without debonding the coating/fiber interface; i.e., the coating itself fails.66,68 The interfacial toughness, friction, and composite strength are higher for the treated-fiber CMC. If the matrix crack runs through the coatings on untreated fibers before the interface debonds, it deflects in the coating, as it does in the identical coating on the treated fiber; hence, the coating/fiber interface must fail before the matrix crack enters the coating. When the crack does pass through the coating, the elastic constraint of the fiber is mostly removed by the preexisting debond, and the crack runs directly to the debonded interface. If there are truly no material differences besides interface strength, a definitive sequence of events consistent with the model is implied.63 These composites have other interesting properties that are discussed in later sections. Another deflection mechanism preceding matrix crack impingement has been suggested for a composite comprising SiC monofilaments with successive coatings of carbon and TiB2 in a glass matrix.69 The coating was calculated to be in triaxial compression, and modeling suggested that matrix cracks would run to the coating only after coating or interfacial failure. Experiment revealed that debond cracks ran very near the fiber surface except for triangular-section C rings around the fiber with their peaks at the matrix crack planes. Shear stresses on planes in the approximate orientation of the sides of the C rings (about 45° from the matrix crack plane) were calculated to be the highest coating stresses that could lead to fracture. The suggested failure sequence was (i) the matrix crack approached the coating, (ii) the coating failed in shear on planes 45° from the crack plane, ahead of the crack, and formed C rings, (iii) the coating cracks turned parallel to the fiber surface at or near the fiber surface, and (iv) the matrix crack advanced until it joined the coating shear cracks at their intersection (see Fig. 4(c)). A further possibility is growth of periodic echelon cracks. In an analysis of thin laminates with a Mode I crack normal to the interface, the maximum tensile stresses in the coatings were 45° to the interface plane, that is, parallel to a “half-turn” of the impinging crack into the interface plane.70 Failure initiated as a series of parallel “periodic echelon” microcracks on these highstress planes at the center of the coating. As the microcracks approached the coating/plate interfaces, they turned parallel to the interface and joined to form a debond. Evidence for a similarappearing but different sequence of events has been observed in monazite interlayers in Al2O3/Al2O3 laminates (Fig. 5).27 In that case, the echelon cracks appeared after initial deflection of the main crack into the coating/laminate interface. Detailed fracture observations are difficult; therefore, the sequence of events for fiber/matrix debonding in CMCs remains speculative. Debonding mechanisms may vary with the particular composite and global stress state. The coating property most important to crack deflection depends on subtle differences in constituent elastic properties and residual stresses. In the ideal case, coatings are engineered material “components” of a composite system selected for phase, microstructure, and geometry to promote a specific failure mechanism. Enhanced understanding of crack deflection is necessary to allow such a priori design. Simple models are often useful, but they can sometimes be misleading; hence, detailed analysis and comparison of microstructure, crack deflection behavior, and analytical models may be necessary. The same considerations apply to the interpretation of micromechanical tests. For example, fiber pushout/pullout tests may not directly measure the parameters that actually determine debonding during composite failure.71 It is even possible that debonding in singleFig. 4. Three possible sequences leading to crack deflection: (a) matrix crack grows into the coating and then bifurcates and turns with fronts running parallel to the fiber surface in each direction (as well as continuing on in the matrix); (b) coating or interface fails in the tensile field ahead of the matrix crack before arrival of the matrix crack at the interface region; and (c) in the matrix crack at the coating, the crack bifurcates and turns as the coating fails in shear at an intermediate angle, then turns parallel to the fiber surface at or near the fiber surface. 2602 Journal of the American Ceramic Society—Kerans et al. Vol. 85, No. 11
Interface Design for Oxidation-Resistant Ceramic Composites 2603 Alumina Alumina ng, or deflection, of a matrix crack can occur through the formation of"echelon"cracks, as shown in the optical micrograph of the phenomenon taking place within a monazite interlayer separating two Al,O, regions.(b)and(c)are schematics of the mechanisms filament microcomposites can be different from that in full debond length because of insufficient axial strength On the other composites because of the different constraints hand, the local stress state changes and this short coating crack nay not provide sufficient stress concentration to greatly influence ( Interfacial Crack Propagation fiber fracture. In the latter case. the "last" debond crack continues If debonding is along a fiber/coating or coating/matrix interface, to grow while the last coating layer develops multiple Mode I then debond propagation is determined by the interfacial energ cracks that are benign in the short term but presumably cause some d the friction generated by shear traction 72,73 If matrix crack decrease in apparent fiber strength. In the former case, this are deflected in the coating, debonding criteria and crack propa- scenario implies that(i)even if the coating deflects cracks, debond gation can be expected to be more complex. Deflection within the lengths may be short because of a nondeflecting coating/fit coating is attractive, because a layer of coating remains on the nterface, (ii) long debond lengths may require coatings with high fiber, slowing environmental degradation of the fiber. However, it xial strain to failure, as does fiber oxidation protection by a seems that the remaining coating is unlikely to remain intact coating,(iii) failure characterization may find coating/fiber inter- beyond some critical level of strain. This limits the protective face cracks even though crack deflection occurs in the coating, and function and may limit debond I Athought experiment "can (iv)residual coating layers should not be expected to"seal" fibers be illustrative(Fig. 6). We imagine that a matrix crack impinges or throughout their entire strain range. Although this discussion is through the coating away from the matrix crack plane. Eventually, the later sectore, it is consistent with the behavior observed in larg on easy-cleaving oxides, and it comprises the matrix crack bypasses ypothesis for comparison of fracture evidence coating; therefore, the matrix crack is bridged by a fiber with a thinner coating. (This remaining thickness continues to function to slow oxidation and other environmental degradation. )As the mposite is loaded further, the coated fiber is strained until the coating fails in Mode I via a surface-initiated crack. However,a coating that deflects cracks can be expected to again deflect a Mode I crack to Mode Il, leaving the fiber with a yet thinner intact coating. The strain-to-failure of thin coatings often increases with decreasing thickness: 4 therefore, the now thinner coating segment can tolerate higher strain before the deflection process repeats chaps many times. Even if the strain-to-failure does not increase as the layers become thinner. successive mode i cracks can be expected to initiate in a noncoplanar fashion, either because of random flaw distribution or biased strain fields at the tips of the debonding cracks. In either case, eventually, this Mode I coating crack impinges the fiber, where deflection is governed by a different criterion. T_/T where i refers to the coating/fiber Fig. 6. Schematic of a matrix crack impinging on a coated fiber in a interface and f to the fiber, r and z to the normals of crack planes mposite under increasing tension along the axis of the fiber(vertical):(a) initial crack deflection within a coating,(b) subsequent Mode I failure of in cylindrical coordinates with z along the fiber axis. Hence, a the coating, followed by a second deflection; and(c) additional Mode I coating can successfully deflect cracks but not provide sufficient failures and deflections. until the fiber/matrix interface is reached
filament microcomposites can be different from that in full composites because of the different constraints.72 (3) Interfacial Crack Propagation If debonding is along a fiber/coating or coating/matrix interface, then debond propagation is determined by the interfacial energy and the friction generated by shear traction.72,73 If matrix cracks are deflected in the coating, debonding criteria and crack propagation can be expected to be more complex. Deflection within the coating is attractive, because a layer of coating remains on the fiber, slowing environmental degradation of the fiber. However, it seems that the remaining coating is unlikely to remain intact beyond some critical level of strain. This limits the protective function and may limit debond length. A “thought experiment” can be illustrative (Fig. 6). We imagine that a matrix crack impinges on a coated fiber, is deflected in the coating (a debond), and advances through the coating away from the matrix crack plane. Eventually, the matrix crack bypasses the fiber, and the debond advances in the coating; therefore, the matrix crack is bridged by a fiber with a thinner coating. (This remaining thickness continues to function to slow oxidation and other environmental degradation.) As the composite is loaded further, the coated fiber is strained until the coating fails in Mode I via a surface-initiated crack. However, a coating that deflects cracks can be expected to again deflect a Mode I crack to Mode II, leaving the fiber with a yet thinner intact coating. The strain-to-failure of thin coatings often increases with decreasing thickness;74 therefore, the now thinner coating segment can tolerate higher strain before the deflection process repeats, perhaps many times. Even if the strain-to-failure does not increase as the layers become thinner, successive Mode I cracks can be expected to initiate in a noncoplanar fashion, either because of random flaw distribution or biased strain fields at the tips of the debonding cracks. In either case, eventually, this Mode I coating crack impinges the fiber, where deflection is governed by a different criterion, i r/ f z, where i refers to the coating/fiber interface and f to the fiber, r and z to the normals of crack planes in cylindrical coordinates with z along the fiber axis. Hence, a coating can successfully deflect cracks but not provide sufficient debond length because of insufficient axial strength. On the other hand, the local stress state changes and this short coating crack may not provide sufficient stress concentration to greatly influence fiber fracture. In the latter case, the “last” debond crack continues to grow while the last coating layer develops multiple Mode I cracks that are benign in the short term but presumably cause some decrease in apparent fiber strength. In the former case, this scenario implies that (i) even if the coating deflects cracks, debond lengths may be short because of a nondeflecting coating/fiber interface, (ii) long debond lengths may require coatings with high axial strain to failure, as does fiber oxidation protection by a coating, (iii) failure characterization may find coating/fiber interface cracks even though crack deflection occurs in the coating, and (iv) residual coating layers should not be expected to “seal” fibers throughout their entire strain range. Although this discussion is largely speculative, it is consistent with the behavior observed in the later section on easy-cleaving oxides, and it comprises a hypothesis for comparison of fracture evidence. Fig. 5. (a) Blunting, or deflection, of a matrix crack can occur through the formation of “echelon” cracks, as shown in the optical micrograph of the phenomenon taking place within a monazite interlayer separating two Al2O3 regions. (b) and (c) are schematics of the mechanisms.27 Fig. 6. Schematic of a matrix crack impinging on a coated fiber in a composite under increasing tension along the axis of the fiber (vertical): (a) initial crack deflection within a coating; (b) subsequent Mode I failure of the coating, followed by a second deflection; and (c) additional Mode I failures and deflections, until the fiber/matrix interface is reached. November 2002 Interface Design for Oxidation-Resistant Ceramic Composites 2603