《复合材料 Composites》课程教学资源（学习资料）第五章 陶瓷基复合材料_SiC-SiC-40

CATERTALIA Pergamon Acta mater.48(200046094618 www.elsevier.com/locate/actamat THE CONCEPT OF A STRONG INTERFACE APPLIED TO SiC/SIC COMPOSITES WITH A BN INTERPHASE F. REBILLAT, J. LAMONT and A. GUETTE Laboratoire des Composites Thermostructuraux, UMR 5801, CNRS-SNECMA-CEA-UB1, 3, allee de la Boetie 33600 Pessac. france natrix composites(CMC). The concept of a strong interface has been established in SiC/SiC composit with pyrocarbon(Py C)or multilayered(PyC/SiC) fiber coatings(also referred to as interphases). The present reports an attempt directe applying the concept of a strong interface to SiC/SiC composites with BN coating(referred to as SiC/BN/SiC). Fiber bonding and frictional sliding were investigated by means of push-out tests performed on 2D-composites as well as on microcomposite samples, and tensile tests perfor- d on microcomposites. The stress-strain behavior of the SiC/BN/SiC composites and microcomposites is discussed with respect to interface characteristics and location of debonding either in the coating or in the fiber/coating interface. 2000 Acta Metallurgica Inc. Published by Elsevier Science Ltd. AlI rights reserved. 1 INTRODUCTION expansion mismatch. Fiber/matrix interfaces in the most advanced ceramic matrix composites consist of of composites because load transfers from the matrix a thin coating layer(less than l um thick )of one or to the fiber and vice versa must occur through the several materials deposited on the fiber (interphase) interface. Therefore, it exerts a profound infuence Recently, SiC/SiC composites with strong interfaces upon the mechanical behavior and the lifetime. Thus, have been developed. The coating/fiber bond was sig- as a function of end use applications through optimiz. treated 3-5). Features of the mechanical behavior of SiC/SiC composites with strong fiber/coating inter- 6-1 In fiber-reinforced ceramic composites, most Experiments as well as models have demonstrated increase fracture toughness. The major contribution that a strong interface is beneficial to the strength, the to toughness is attributed to crack bridging and fiber toughness, the lifetime and the creep resistance 14, 6 composite strength. A high strength requires efficient mental trast, weak interfaces are shown to be det- he once ept of strong interfaces has been estab- load transfers which are obtained with strong inter- lished on Sic/C/SiC composites with PyC and multi- faces. This implies short debond cracks and/or sig- layered(Py C/SiC)fiber coatings. In the present paper, nificant sliding friction. These latter requirements, to be met for strong composites, are therefore incompat- it is applied to SiC/BN/SiC composites with boron ible with the former ones for tough composites, if nitride fiber coatings. BN is foreseen to be an alterna- toughening is based solely upon the above mentioned tive fiber coating to improve the oxidation resistance weak interface-based mechanisms of ceramic matrix composites at high temperature Fiber/matrix bonding results from diffusion or chemical reactions(chemical bonding)or from fiber 2. FEATURES OF STRONG INTERFACES VS WEAK clamping by residual stresses induced by thermal INTERFACES n prc uce of a hom all correspondence should be addressed. Tel: strong interface, we recall first the basic features of 844-703;fax:+33-556-841-225 interface phenomena in CMCs subject to an essen- ddress: admin@lcts. u-bordeaux fr(. Lamon) tially tensile load. These phenomena influence the 1359-6454100/520.00@ 2000 Acta Metallurgica Inc. Published by Elsevier Science Ltd. All rights reserved PI:S1359-6454(00)00247-0

Acta mater. 48 (2000) 4609–4618 www.elsevier.com/locate/actamat THE CONCEPT OF A STRONG INTERFACE APPLIED TO SiC/SiC COMPOSITES WITH A BN INTERPHASE F. REBILLAT, J. LAMON† and A. GUETTE Laboratoire des Composites Thermostructuraux, UMR 5801, CNRS-SNECMA-CEA-UB1, 3, alle´e de la Boe´tie, 33600 Pessac, France Abstract—Strong interfaces have been shown to allow improvement of the mechanical properties of ceramic matrix composites (CMC). The concept of a strong interface has been established in SiC/SiC composites with pyrocarbon (PyC) or multilayered (PyC/SiC) fiber coatings (also referred to as interphases). The present paper reports an attempt directed at applying the concept of a strong interface to SiC/SiC composites with a BN coating (referred to as SiC/BN/SiC). Fiber bonding and frictional sliding were investigated by means of push-out tests performed on 2D-composites as well as on microcomposite samples, and tensile tests perfor￾med on microcomposites. The stress–strain behavior of the SiC/BN/SiC composites and microcomposites is discussed with respect to interface characteristics and location of debonding either in the coating or in the fiber/coating interface.  2000 Acta Metallurgica Inc. Published by Elsevier Science Ltd. All rights reserved. Keywords: Interphase; Interface; Composites 1. INTRODUCTION The fiber–matrix interfacial domain is a critical part of composites because load transfers from the matrix to the fiber and vice versa must occur through the interface. Therefore, it exerts a profound influence upon the mechanical behavior and the lifetime. Thus, it may be expected that composites could be tailored as a function of end use applications through optimiz￾ation of interfaces. In fiber-reinforced ceramic composites, most authors promote the concept of weak interfaces to increase fracture toughness. The major contribution to toughness is attributed to crack bridging and fiber pull-out [1, 2]. Weak interfaces are detrimental to composite strength. A high strength requires efficient load transfers which are obtained with strong inter￾faces. This implies short debond cracks and/or sig￾nificant sliding friction. These latter requirements, to be met for strong composites, are therefore incompat￾ible with the former ones for tough composites, if toughening is based solely upon the above mentioned weak interface-based mechanisms. Fiber/matrix bonding results from diffusion or chemical reactions (chemical bonding) or from fiber clamping by residual stresses induced by thermal † To whom all correspondence should be addressed. Tel.: 133-556-844-703; fax: 133-556-841-225. E-mail address: admin@lcts.u-bordeaux.fr (J. Lamon) 1359-6454/00/$20.00  2000 Acta Metallurgica Inc. Published by Elsevier Science Ltd. All rights reserved. PII: S13 59-6454(00)00247-0 expansion mismatch. Fiber/matrix interfaces in the most advanced ceramic matrix composites consist of a thin coating layer (less than 1 µm thick) of one or several materials deposited on the fiber (interphase). Recently, SiC/SiC composites with strong interfaces have been developed. The coating/fiber bond was sig￾nificantly stronger when fibers had been previously treated [3–5]. Features of the mechanical behavior of SiC/SiC composites with strong fiber/coating inter￾faces has been examined in several papers [4, 6–10]. Experiments as well as models have demonstrated that a strong interface is beneficial to the strength, the toughness, the lifetime and the creep resistance [4, 6– 11]. By contrast, weak interfaces are shown to be det￾rimental. The concept of strong interfaces has been estab￾lished on SiC/C/SiC composites with PyC and multi￾layered (PyC/SiC) fiber coatings. In the present paper, it is applied to SiC/BN/SiC composites with boron nitride fiber coatings. BN is foreseen to be an alterna￾tive fiber coating to improve the oxidation resistance of ceramic matrix composites at high temperature. 2. FEATURES OF STRONG INTERFACES VS WEAK INTERFACES In order to properly introduce the concept of a strong interface, we recall first the basic features of interface phenomena in CMCs subject to an essen￾tially tensile load. These phenomena influence the

REBILLAT et al: SiC/SIC COMPOSITES mechanical response of composites described by the tress -strain curve Fiber debonding results from the deflection of the cracks that initiate in the matrix(Fig. 1). Then sliding 5 of the fiber debonded in the interface determines the load transfers from the fiber to the matrix and vice 5 200 versa. The fiber sliding is influenced by the misfit E strain, the associated radial component of the ther mally induced residual stress-field, surface roughness Weak interfaces debond easily. A single long debond crack is located at the surface of the fibers in o LONGITUDINAL TE small interface shear stresses, load transfers through Fig. 2: Typical tensile stress-strain behaviors measured on 2D the debonded interfaces are poor. The matrix becomes ricated from untreated or treated Nicalon(ceramic grade)fit subjected to lower stresses and the volume of matrix ers: (a)strong fiber/coating interfaces and (b) that may experience further cracking is reduced the presence of long debonds. The cracks are gener- ally widely opened, whereas the crack spacing dis- tance at saturation as well as the pull out length tend (cohesive failure type, Fig. 1), into short and branched to be rather long (100 um). Toughening results multiple cracks [4, 12]. Short debonds as well as essentially from sliding friction along the debonds. Improved load transfers allow further cracking of the matrix via a scale effect 6, 7] leading to a hig However, due to poor load transfers and long density of matrix cracks(which are slightly opened debonds,the fibers carry most of the load, which Sliding friction within the coating as well as multiple reduces the composite strength. The corresponding cracking of the matrix increase energy absorption, tensile stress-strain curve exhibits a short curved domain limited by a stress at matrix cracking satu- leading to toughening. Limited debonding and fibers, leading to strengthening. The associated tensile In the presence of stronger fiber/coating bonds, the stress strain curve exhibits a wide curved domain and the stress at matrix cracking saturation is close to ulti mate failure(Fig. 2). Table 1 gives various values of the interfacial shear stresses measured using various debond crack methods on SiC/SiC composites with PyC-based fiber coating. It can be noticed that the interfacial shear stresses range between 10 and 20 MPa for the weak interfaces whereas they are larger than 100-300 MPa for the strong interfaces. Additional data can be found in[4,7,8,24,29] 3. SiC/BN/SiC COMPOSITES: TESTING METHODOLOGY AND MICROSTRUCTURAL ANALYSES Fiber 3. 1. Specimen preparation debond crack SIC/BN/SIC microcomposites and woven com posites were manufactured via chemical vapor infil- tration [13]. They were reinforced with either as received or treated(proprietary treatment, SNECMA/SEP, Bordeaux) SiC Nicalon fibers (NL 202 grade). The SiC/BN/SIC mIcrocosmos tes consist of a single fiber (15 um diameter ), coated with a boron nitride layer(0.3-0.9 um thick) and a Sic matrix deposited by CVD. They are tive of Fiber their counterparts in the 2D woven con they are produced using identical deposition conditions [13] Fig 1 Schematic diagram showing crack deflection when the A single or a bilayered BN fiber coating was fiber coating/interface is (a) strong or(b) weak. deposited from a BF3, NH3, Ar gas mixture (Table

4610 REBILLAT et al.: SiC/SiC COMPOSITES mechanical response of composites described by the stress–strain curve. Fiber debonding results from the deflection of the cracks that initiate in the matrix (Fig. 1). Then sliding of the fiber debonded in the interface determines the load transfers from the fiber to the matrix and vice versa. The fiber sliding is influenced by the misfit strain, the associated radial component of the ther￾mally induced residual stress-field, surface roughness and debond length. Weak interfaces debond easily. A single long debond crack is located at the surface of the fibers in those composites exhibiting weak interfaces (adhesive failure type, Fig. 1). As a consequence of small interface shear stresses, load transfers through the debonded interfaces are poor. The matrix becomes subjected to lower stresses and the volume of matrix that may experience further cracking is reduced by the presence of long debonds. The cracks are gener￾ally widely opened, whereas the crack spacing dis￾tance at saturation as well as the pull out length tend to be rather long (>100 µm). Toughening results essentially from sliding friction along the debonds. However, due to poor load transfers and long debonds, the fibers carry most of the load, which reduces the composite strength. The corresponding tensile stress–strain curve exhibits a short curved domain limited by a stress at matrix cracking satu￾ration which is significantly smaller than ultimate strength (Fig. 2). In the presence of stronger fiber/coating bonds, the matrix cracks are deflected within the coating Fig. 1. Schematic diagram showing crack deflection when the fiber coating/interface is (a) strong or (b) weak. Fig. 2. Typical tensile stress–strain behaviors measured on 2D SiC/SiC composites possessing PyC based interphases and fab￾ricated from untreated or treated Nicalon (ceramic grade) fib￾ers: (a) strong fiber/coating interfaces and (b) weak fiber/coating interfaces. (cohesive failure type, Fig. 1), into short and branched multiple cracks [4, 12]. Short debonds as well as improved load transfers allow further cracking of the matrix via a scale effect [6, 7] leading to a higher density of matrix cracks (which are slightly opened). Sliding friction within the coating as well as multiple cracking of the matrix increase energy absorption, leading to toughening. Limited debonding and improved load transfers reduce the load carried by the fibers, leading to strengthening. The associated tensile stress strain curve exhibits a wide curved domain and the stress at matrix cracking saturation is close to ulti￾mate failure (Fig. 2). Table 1 gives various values of the interfacial shear stresses measured using various methods on SiC/SiC composites with PyC-based fiber coating. It can be noticed that the interfacial shear stresses range between 10 and 20 MPa for the weak interfaces whereas they are larger than 100–300 MPa for the strong interfaces. Additional data can be found in [4, 7, 8, 24, 29]. 3. SiC/BN/SiC COMPOSITES: TESTING METHODOLOGY AND MICROSTRUCTURAL ANALYSES 3.1. Specimen preparation SiC/BN/SiC microcomposites and woven com￾posites were manufactured via chemical vapor infil￾tration [13]. They were reinforced with either as￾received or treated (proprietary treatment, SNECMA/SEP, Bordeaux) SiC Nicalon fibers (NL 202 grade). The SiC/BN/SiC microcomposites consist of a single fiber (15 µm diameter), coated with a boron nitride layer (0.3–0.9 µm thick) and a SiC matrix deposited by CVD. They are representative of their counterparts in the 2D woven composites, since they are produced using identical chemical vapor deposition conditions [13]. A single or a bilayered BN fiber coating was deposited from a BF3, NH3, Ar gas mixture (Table

REBILLAT et al: SiC/SIC COMPOSITES 4611 Table 1. Interfacial shear stress (MPa) measured using various methods on 2D-SiC/SiC composites with Py C based fiber coatings and reinforced with either as-received or treated fibers SiC/C/SiC )n/SiC loops)[24] domain Untreated fibers 21-115 4080 12-10 Treated fibers PyC(o 1) 165-273 PyC/SiC)4 2). The selected processing conditions have been shown to cause minimum damage to the fibers and to improve adhesion of the BN coating onto the fib- ers, and the microstructure [13]. In the bi-layered coating(referred to as BN4, Table 2 and Fig. 3), the first sublayer on the fiber is BN2 type(poorly crystallized), whereas the second one is BNI type (highly crystallized). Processing of BN2 involved the less aggressive gaseous phase, which led to a better contact between the fiber and the coating. The pro- essing conditions of BNI were found to be aggress- ive against the fibers [13]- (a) 3.2. Push-out tests Fiber bonding and frictional sliding in the 2D SiC/BN/SiC composites were investigated by means of single fiber push-out tests [14-17].500 um thick wedges were prepared using standard metallographic ONERA, France)was used. The load was applied to the top of the fiber using a flat bottom diamond cone (at a constant displacement rate of 0.1 um/s). The nterface characteristics were extracted from the experimental stress-fiber end displacement curves by fitting the push-out model of Hsueh [ 15], as discussed n a previous paper [16] Only a few push-out experiments could be carried out on the microcomposites owing to the difficulties involved in microcomposite handling, preparation and testing. Parallel-faced strips were cut out of the Fig 3. Images of BN coatings:(a)TEM-image and DEAS. microcomposites which had been previously embed- picture of the BNI coating showing the three-dimensional ded in glass [18](microcomposites 2)or in a ceramic xagonal structure; (b) SEM-image of bi-layered BN ement [19](microcomposites 2 and 4). A first series BN4)in a SiC/BN/SiC microcomposite showing dered BN2 layer and the BN1 layer with a three dimensional ordered hexagonal structure Table 2. Main characteristics ber coatings [13] Batch BN coating Number Coating of push out tests on microcomposites 2 used a Vickers zation adhesion diamond probe [18).A correction for indentor dis- placement was done. Most of the tests were push in rather than push-out tests, Push-out could not be BN1+BN2 Strong achieved on the specimens with a thickness exceeding 200 um. A number of push-in curves

REBILLAT et al.: SiC/SiC COMPOSITES 4611 Table 1. Interfacial shear stress (MPa) measured using various methods on 2D-SiC/SiC composites with PyC based fiber coatings and reinforced with either as-received or treated fibers SiC/C/SiC Interphase Crack spacing Crack spacing Tensile tests Tensile tests Push-out tests Push-out tests SiC/(C/SiC)n/SiC [30] [31] (hysteresis (curved (curved (plateau) loops) [24] domain) domain) [8, 29] [7, 24] [8, 29] Untreated fibers 2D woven PyC (0.1) 12 8 0.7 Microcomposites PyC(0.1) 3 4–20 Minicomposites PyC(0.1) 21–115 40–80 2D woven PyC(0.5) 4 14–16 12–10 (PyC/SiC)2 2 31 19.3 (PyC/SiC)4 9 28 12.5 Treated fibers 2D woven PyC(0.1) 203 140 190 165–273 PyC(0.5) 370 100–105 (PyC/SiC)2 150 133 (PyC/SiC)4 90 90 2). The selected processing conditions have been shown to cause minimum damage to the fibers and to improve adhesion of the BN coating onto the fib￾ers, and the microstructure [13]. In the bi-layered coating (referred to as BN4, Table 2 and Fig. 3), the first sublayer on the fiber is BN2 type (poorly crystallized), whereas the second one is BN1 type (highly crystallized). Processing of BN2 involved the less aggressive gaseous phase, which led to a better contact between the fiber and the coating. The pro￾cessing conditions of BN1 were found to be aggress￾ive against the fibers [13]. 3.2. Push-out tests Fiber bonding and frictional sliding in the 2D￾SiC/BN/SiC composites were investigated by means of single fiber push-out tests [14–17]. 500 µm thick wedges were prepared using standard metallographic techniques. An interfacial test system (designed by ONERA, France) was used. The load was applied to the top of the fiber using a flat bottom diamond cone (at a constant displacement rate of 0.1 µm/s). The interface characteristics were extracted from the experimental stress–fiber end displacement curves by fitting the push-out model of Hsueh [15], as discussed in a previous paper [16]. Only a few push-out experiments could be carried out on the microcomposites owing to the difficulties involved in microcomposite handling, preparation and testing. Parallel-faced strips were cut out of the microcomposites which had been previously embed￾ded in glass [18] (microcomposites 2) or in a ceramic cement [19] (microcomposites 2 and 4). A first series Table 2. Main characteristics of the BN fiber coatings [13] Batch BN coating Number of Degree of Coating BN layers crystallization adhesion 1 BN1 1 High Strong 2 BN2 1 Low Weak 4 BN4 2 BN11BN2 Strong Fig. 3. Images of BN coatings: (a) TEM-image and DEAS￾picture of the BN1 coating showing the three-dimensional ordered hexagonal structure; (b) SEM-image of bi-layered BN interphase (BN4) in a SiC/BN/SiC microcomposite showing poorly ordered BN2 layer and the BN1 layer with a three￾dimensional ordered hexagonal structure. of push out tests on microcomposites 2 used a Vickers diamond probe [18]. A correction for indentor dis￾placement was done. Most of the tests were push￾in rather than push-out tests. Push-out could not be achieved on the specimens with a thickness exceeding 200 µm. A number of push-in curves was unusable

REBILLAT et al: SiC/SIC COMPOSITES for analysis in that they exhibited features that were (a) inconsistent with the model. A second series of push- in tests on microcomposites 2 was then performed on thicker samples (290 um) using a flat-bottomed cone microcomposites 4(thickness 190 um). The inter faces characteristics were extracted by fitting the a Hsueh's model [15] to the push-in curves or to the curved domain and to the plateau of the push-out 3.3. Tensile tests Five microcomposites per batch were tested in ten sion by using a specific table-model testing machine designed and developed for fiber testing [20]). The sin- gle fiber tensile test procedure based on window frames with appropriate gauge lengths(generally 10 mm)was employed [21, 22]. The microcomposites were loaded up to failure, either monotonically or with unloading-reloading cycles at a low strain rate (0.1%mn-) The interface characteristics including the shear stress(T), the debond energy (Gis)and the debond length (la)were extracted from the stress-strain curves [22-24] and from hysteresis loops on Stain (o unloading-reloading [23, 24). Independent models Fig. 4. Tensile stress-strain curves measured on the were used in order to assess the results. These models sic/BN icrocomposites reinforced with: (a) are referred to as LRLC. ClR and lre according to fibers and(b) treated fiber authors s[22-24]. They derive from modelling the tensile, load-displacement behavior (LRLC microcomposites I and 4 reinforced with as-received ers, suggesting the presence of rather weak models)of microcomposites: the CLR model deter- fiber/matrix interactions, short debonds and small mines the energy dissipated in the friction phenomena densities of matrix cracks at saturation when com whereas the LRE one determines the crack opening ing with the microcomposites 2 which exhibit a displacement during unloading-reloading cycles After ultimate failure, the microcomposites were widely curved stress-strain behavior up to ultimate examined using scanning electron microscopy (SEM failure and larger stresses. saturation of matrix crack The composition of the surface of fibers was determ- (0.6%) ing generally occurred at rather large deformations ined from Auger electron spectroscopy(AES) depth- profile analyses of the pulled-out fibers 4.1.2. SEM fractography of microcomposites. The o The tensile tests on the SiC/BN/SiC woven com- numbers of matrix cracks identified on the microcom- posites(three test specimens per batch) were perfor- posites after ultimate failure were generally compara- med at a constant strain rate of 0.05% min-I Defor- ble with the numbers of load drops or of slope mations were measured using an extensometer decreases on the force-displacement curves(Table 3) length 25 mm). The dir ensions of the test speci The higher density of matrix cracks was observed in were as follows: thickness 3 mm width 8 mm microcomposites 2(Table 3) 100mn In the microcomposites reinforced with untreated fibers debonding was observed mainly at the fiber/BN nterface. The free surface of untreated fibers RESULTS which the BN interphase was deposited, is at least 4.1. Tensile tests on the SiC/BN/SiC microcomposites partly made of silica 3). The resulting fiber/BN inter 4.1.. Stress-strain curves. Most of the tensile bond 3, 13, 25, 26]. Similar features have been tressstrain curves( Fig. 4) exhibited a curved observed on SiC/C/SiC composites with a fiber coat- domain over a wide range of deformations(0.2- ing of anisotropic PyC [51 0.9%), and rather large strains-to-failure up to 1. 2% In the microcomposites reinforced with treated (Table 3). However, most of the microcomposites fibers, debonding was detected in the BN coating ith treated fibers essentially experienced premature only in microcomposites 4 with a bi-layered BN coat ing, which indicates that the weakest link is now a plateau-like behavior was observed for located in the interface between the BN sublayers. As

4612 REBILLAT et al.: SiC/SiC COMPOSITES for analysis in that they exhibited features that were inconsistent with the model. A second series of push￾in tests on microcomposites 2 was then performed on thicker samples (290 µm) using a flat-bottomed cone. Push-out tests were successfull on samples of microcomposites 4 (thickness 0.6%). 4.1.2. SEM fractography of microcomposites. The numbers of matrix cracks identified on the microcom￾posites after ultimate failure were generally compara￾ble with the numbers of load drops or of slope decreases on the force–displacement curves (Table 3). The higher density of matrix cracks was observed in microcomposites 2 (Table 3). In the microcomposites reinforced with untreated fibers debonding was observed mainly at the fiber/BN interface. The free surface of untreated fibers, on which the BN interphase was deposited, is at least partly made of silica [3]. The resulting fiber/BN inter￾face has been reported to correspond to a very weak bond [3, 13, 25, 26]. Similar features have been observed on SiC/C/SiC composites with a fiber coat￾ing of anisotropic PyC [5]. In the microcomposites reinforced with treated fibers, debonding was detected in the BN coating only in microcomposites 4 with a bi-layered BN coat￾ing, which indicates that the weakest link is now located in the interface between the BN sublayers. As

REBILLAT et al: SiC/SIC COMPOSITES Table 3. Main features of the stress-strain curves for the SiC/BN/SiC microcomposites and 2D woven composites Number of cracks at Interphase thickne Pr Failure stress(MPa) Failure strain (%) Microcomposites 0.850.18 BN4 S=as-received fibers, T=treated fibers. SEM previously reported for Pyrocarbon fiber coatings [4 1, treated fibers seem to give stronger fiber/BN bonds. However, the microcomposites with a single layer BN coating appear to be an exception to this 2 rule. since the interface crack was detected at the fiber/BN interface. This was attributed to the presence of a weakly bonded sublayer of carbon that formed on the fibers [27] 4.1.3. Auger electron spectroscopy analyses. AES depth-profile analyses of the pulled out fibers in nicrocomposites reinforced with untreated fibers, showed that the fiber surface is rich in free carbon A layer enriched in carbon and oxygen(probably con- sisting of silica) is present under this carbon layer Such a complex interfacial sequence has been already observed in 2D SiC/BN/SiC composites [26, 28]. The8 曰 CLR (Ioo 原cLR( envelop very thin carbon layer results from the attack of the o LRE (loops) fiber surface during BN processing [271 器LRE( envelop) 4.1.4. Extraction of interfacial properties from the stress-strain curve. The models provide compara ble estimates of interfacial shear stresses for the icrocomposites reinforced with untreated fibers (Fig. 5). A certain discrepancy may be observed for microcomposites 2. The interfacial shear stresses can be grouped into two distinct families(Fig. 5) [5 MPa for microcomposites 4, Fig. 5. Interfacial characteristics estimated using various mod- T210 MPa for microcomposites I and reinforced with untreated fibers The debond energy estimates range between I and 4.2. Tensile tests on the 2D SiC/BN/SiC composites The interfacial characteristics determined for the The stress-strain curves of the 2D sic/bn/sic microcomposites reinforced with treated fibers are composites also display a curved domain( Fig. 7). The Fig. 6. The interfacial shear stresses strains-to-failure are smaller than those measured on obtained for microcomposites I and 2 are larger than the microcomposites(Table 3). They are close those measured for the microcomposites reinforced 0.6% for composites I(reinforced with untreated with as-received fibers. A certain discrepancy is fibers) and 4(reinforced with as-received or treated observed on the data extracted using the LRE model fibers ), whereas the other composites failed at defor- [24]: T=400 MPa and Gie =70 J/m2 seem to be mations <0.2% overestimations although microcomposites I experi- enced a premature failure. The characteristics pro. 4.3. Push-out tests on the 2D SiC/BN/SiC composites vided by the other models seem to be more realistic: 43. 1. Composites reinforced with as-received 10<<50MPa,0<G<7Jm2(Fg6) fibers. The stresses to initiate and propagate the

REBILLAT et al.: SiC/SiC COMPOSITES 4613 Table 3. Main features of the stress–strain curves for the SiC/BN/SiC microcomposites and 2D woven composites Interphase Number of cracks at Failure stress (MPa) Failure strain (%) Interphase thickness Vf saturationb (µm) Sa Ta STST Microcomposites BN1 0.28 0.42 792 680 0.55 0.18 9 2 BN2 0.27 0.76 1368 1813 0.85 1.27 55 46 BN4 0.29 0.47 970 670 0.99 0.2 14 1 Composites BN1 0.5 0.40 220 32 0.58 0.06 BN2 0.3 0.40 210 110 0.38 0.11 BN4 0.5 0.40 200 210 0.5 0.064 a S5as-received fibers, T5treated fibers. b Determined by SEM. previously reported for Pyrocarbon fiber coatings [4, 5], treated fibers seem to give stronger fiber/BN bonds. However, the microcomposites with a single layer BN coating appear to be an exception to this rule, since the interface crack was detected at the fiber/BN interface. This was attributed to the presence of a weakly bonded sublayer of carbon that formed on the fibers [27]. 4.1.3. Auger electron spectroscopy analyses. AES depth-profile analyses of the pulled out fibers in microcomposites reinforced with untreated fibers, showed that the fiber surface is rich in free carbon. A layer enriched in carbon and oxygen (probably con￾sisting of silica) is present under this carbon layer. Such a complex interfacial sequence has been already observed in 2D SiC/BN/SiC composites [26, 28]. The very thin carbon layer results from the attack of the fiber surface during BN processing [27]. 4.1.4. Extraction of interfacial properties from the stress–strain curve. The models provide compara￾ble estimates of interfacial shear stresses for the microcomposites reinforced with untreated fibers (Fig. 5). A certain discrepancy may be observed for microcomposites 2. The interfacial shear stresses can be grouped into two distinct families (Fig. 5): t<5 MPa for microcomposites 4, t$10 MPa for microcomposites 1 and 2, The debond energy estimates range between 1 and 8 J/m2 (Fig. 5). The interfacial characteristics determined for the microcomposites reinforced with treated fibers are shown on Fig. 6. The interfacial shear stresses obtained for microcomposites 1 and 2 are larger than those measured for the microcomposites reinforced with as-received fibers. A certain discrepancy is observed on the data extracted using the LRE model [24]: t 5 400 MPa and Gic 5 70 J/m2 seem to be overestimations although microcomposites 1 experi￾enced a premature failure. The characteristics pro￾vided by the other models seem to be more realistic: 10,t,50 MPa, 0,Gic,7 J/m2 (Fig. 6). Fig. 5. Interfacial characteristics estimated using various mod￾els for various BN interphases in SiC/BN/SiC microcomposites reinforced with untreated fibers. 4.2. Tensile tests on the 2D SiC/BN/SiC composites The stress–strain curves of the 2D SiC/BN/SiC composites also display a curved domain (Fig. 7). The strains-to-failure are smaller than those measured on the microcomposites (Table 3). They are close to 0.6% for composites 1 (reinforced with untreated fibers) and 4 (reinforced with as-received or treated fibers), whereas the other composites failed at defor￾mations ,0.2%. 4.3. Push-out tests on the 2D SiC/BN/SiC composites 4.3.1. Composites reinforced with as-received fibers. The stresses to initiate and propagate the

4614 REBILLAT et al: SiC/SIC COMPOSITES 400十 bN 4 80 S CLR (loops) BN 4 Fig. 7. Tensile stress-strain curves measured on the Fig. 6. Interfacial characteristics estimated using various mod- SiC/BN/SiC 2D composites reinforced with:(a)as-received els for various BN interphases in SiC/BN/SiC microcomposites fibers and(b) treated fibers. reinforced with treated fibers those estimated for the composites reinforced with debond crack as well as the interfacial shear stresses untreated fibers. Unfortunately, interfacial properties are much higher than those measured on 2D woven could not be extracted from the curved domain of the h-out curve to bending of the sample as a Sic/C/SiC composites with a PyC-based fiber coating result of the high load required to cause debonding and as received fibers [8, 291(Table 4 and Fig. 8). The interfacial shear stress that was extracted from previously with the microcomposites, the largest the plateau is large([=140 MPa). Smaller interfacial stresses and the interfacial shear stresses thus appear shear and debond stresses were obtained for com- to be insensitive to the conditions of BN processing. posites 4(Table 4). These interfacial shear stresses are close to those measured on the composites However, the effect of friction seems to be the most reinforced with untreated fibers efficient in the composites 2, as indicated by the com- The magnitudes of debond stresses may be related parison of the respective values of the following spe- to the respective stress-strain curves displayed by cific parameters(Table 4): the applied maximum composites 2 and 4. The premature ultimate failure stress, the fiber-end displacement and the roughness of composites 2 may correspond to the large oavalue amplitude. The smallest interfacial shear stress and The stress-strain behavior of composites 4, which roughness amplitude() were estimated for the bi- close to that observed for the composites reinforced layered bn coat ble 4) ith untreated fibers, may reflect the similarity in the SEM revealed the following interestis ather rough SEM revealed that debonding took place at the g features:(i) in composites 1, the fiber surface was respective interface characteristics t and oa an ace shov composites 2, the slid surface was very smooth and between the coating and the fiber was not continuous fiber/coating interface(as is usually observed in such probably, as a result of the attack of the fiber by the opposites [18, 26), (ii)in composites 4, the debond gaseous phase crack was located in the interface between the Bn In composites 4, debonding occurred in the inter face between the BN sublayers. Figure 9 shows that a bn sublayer remains bonded to the fiber. SEM 4.3.2. Composites reinforced with treated fibers. examination of the fracture surface of 2D specimens A large debond stress was estimated for composites tested in tension also showed a BN sublayer stuck on 2(oa=2000 MPa, Table 4). It is much larger than the fibers

4614 REBILLAT et al.: SiC/SiC COMPOSITES Fig. 6. Interfacial characteristics estimated using various mod￾els for various BN interphases in SiC/BN/SiC microcomposites reinforced with treated fibers. debond crack as well as the interfacial shear stresses are much higher than those measured on 2D woven SiC/C/SiC composites with a PyC-based fiber coating and as received fibers [8, 29] (Table 4 and Fig. 8). As previously with the microcomposites, the largest t were obtained for composites 1 and 2. The debond stresses and the interfacial shear stresses thus appear to be insensitive to the conditions of BN processing. However, the effect of friction seems to be the most efficient in the composites 2, as indicated by the com￾parison of the respective values of the following spe￾cific parameters (Table 4): the applied maximum stress, the fiber-end displacement and the roughness amplitude. The smallest interfacial shear stress and roughness amplitude (A) were estimated for the bi￾layered BN coating (Table 4). SEM revealed the following interesting features: (i) in composites 1, the fiber surface was rather rough and the matrix surface showed a lot of pores, (ii) in composites 2, the slid surface was very smooth and debonding seemed to have occurred at the fiber/coating interface (as is usually observed in such composites [18, 26]), (iii) in composites 4, the debond crack was located in the interface between the BN sublayers (Fig. 9). 4.3.2. Composites reinforced with treated fibers. A large debond stress was estimated for composites 2 (sd 5 2000 MPa, Table 4). It is much larger than Fig. 7. Tensile stress–strain curves measured on the SiC/BN/SiC 2D composites reinforced with: (a) as-received fibers and (b) treated fibers. those estimated for the composites reinforced with untreated fibers. Unfortunately, interfacial properties could not be extracted from the curved domain of the push-out curve, owing to bending of the sample as a result of the high load required to cause debonding. The interfacial shear stress that was extracted from the plateau is large (t<140 MPa). Smaller interfacial shear and debond stresses were obtained for com￾posites 4 (Table 4). These interfacial shear stresses are close to those measured on the composites reinforced with untreated fibers. The magnitudes of debond stresses may be related to the respective stress–strain curves displayed by composites 2 and 4. The premature ultimate failure of composites 2 may correspond to the large sd value. The stress–strain behavior of composites 4, which is close to that observed for the composites reinforced with untreated fibers, may reflect the similarity in the respective interface characteristics t and sd. SEM revealed that debonding took place at the fiber/BN interface in composites 2. The contact between the coating and the fiber was not continuous, probably, as a result of the attack of the fiber by the gaseous phase. In composites 4, debonding occurred in the inter￾face between the BN sublayers. Figure 9 shows that a BN sublayer remains bonded to the fiber. SEM examination of the fracture surface of 2D specimens tested in tension also showed a BN sublayer stuck on the fibers

REBILLAT et al. SiC/SIC COMPOSITES 3 8 An,avE 8 三8 豆三二

REBILLAT et al.: SiC/SiC COMPOSITES 4615 Table 4. Interfacial characteristics extracted from the push-out curves for the 2D-SiC/BN/SiC composites reinforced with as-received or treated fibers, and from push-in or push-out curves for the SiC/BN/SiC microcomposites reinforced with as-received fibers (standard deviation is given in brackets) Sample thickness Displacement t (plateau) Interphase sd (MPa) smax (MPa) t (MPa) m A (nm) ld (µm) Gic (J/m2) (µm) (µm) (MPa) Composites BN1 (Sa) 200 (10) 1130 (200) 3680 (1315) 1.35 (0.3) 66 (9) 87 (22) 0.11 (0.02) 32 (14) 120 (20) 2.1 (1.1) BN2 (S) 90 (15) 1250 (170) 3770 (530) 1 (0.2) 95 (20) 83 (15) 0.09 (0.03) 45 (20) 95 (9) 2.8 (1.4) BN4 (S) 160 (50) 1200 (330) 2720 (760) 1.1 (0.6) 48 (9) 41 (13) 0.09 (0.03) 16 (10) 105 (15) 1.5 (1.9) BN2 (Ta) 60 (15) 2000 (440) 2770 (350) 140 (26) BN4 (T) 120 (10) 740 (100) 2670 (300) 0.9 (0.1) 77 (9) 66 (5) 0.06 (0.01) 68 (7) 120 (8) 1.1 (0.4) Microcomposites BN2 (S) push-in 280 (10) 1600 (400) 5500 (900) 1.6 (0.4) 100 (40) 0.11 (0.02) 50 (20) 110 (35) 16 (8) BN4 (S) push-out 190 (10) 1100 (500) 2500 (250) 0.6 (0.3) 48 (9) 40 (15) 0.04 (0.005) 45 (23) 110 (30) 0.5 (0.4) a S5as-received fibers, T5treated fibers

REBILLAT et al: SiC/SIC COMPOSITES The debond energy estimates fall within the range 2 composite: t-75 m a 10 um thk obtained from the tensile tests(Fig. 5) The debond stress in microcomposites 2 is q igl 1500 MPa), as well as the interfacial shear stress (t= 90 MPa), which indicates a high resist- plateau ance to debonding. This may be related to the fact that the fiber could not be pushed out the matrix, even For microcomposites 4, a very good agreement can extracted from respectively the curved domain and the plateau of the push out curves. As previously, lower interfacial shear stress is obtained for this 4 o microcomposite with a double layer coating. Further more, push out of the fiber did not require a load as nd displacement single fiber push-out high as that(omac)applied during fiber pushing in the IC/BN/SIC and a 2D microcomposite SIC/BN (interphase BN4)reinforced with as- SEM showed that debonding essentially received fibers at the fiber/coating interface. Ger v smooth slid surfaces were observed. A was detected 5. DISCUSSION The difficulty to perform push-out tests on thin samples becomes still worse with microcomposites Large load transfers require very thin embedded lengths in order to permit push-out. The parallel-faced strip is not the most convenient test specimen because it is too fragile, but it remains the only usable sample geometry. During a push-in, unknown maximum debond length, damage of the fiber due to high com pressive stresses, bending of the sample and diamond meeting with the matrix are factors that affect the stress-strain curves, and the validity of the extracted characteristics. Then, getting well polished thin strips with a thickness <200 um is rare. This explains why the push-out tests could be carried out only on microcomposites 4 which possessed the weakest Interfaces The less aggressive gaseous phase was used for processing the BN2 coating. This coating has been 5um/ shown to adhere well to the fiber[13]. The resistance to propagation of the interface crack during the push- fiber pu 二 out tests appeared to be high, which suggests efficient load transfers under a tensile load. During the tensile as-received fibers and(b)com- tests, the composites 2 experienced a premature fail- ure whereas the microcomposites 2 exhibited a high strain at saturation. Both features are compatible with efficient load transfers. Ultimate failure of 2D 4.4. Push-out and push-in tests on microcomposites posites involves additionnal phenomena and it The few push-out and push-in curves that were affected by variability in fiber strength degradation suitable for analysis exhibited the features usually during processing. The debond stresses and th observed with 2D composites(Fig. 8). The interface interfacial shear stresses are larger than those parameters extracted from these curves are summar- obtained for composites and microcomposites 4 with ized in Table 4 a bi-layered coating, and they are comparable to those The interfacial shear larger than those estimated for comp xtracted from the tensile ain curves(Fig. cessed in more aggressive conditions. As previously 5). They are in excellent agreement with those esti- mentioned, the propagation of interface cracks during mated on the 2D-SiC/BN/SiC composites(Table 4). the push-out tests was easier for composites I than

4616 REBILLAT et al.: SiC/SiC COMPOSITES Fig. 8. Stress–fiber end displacement single fiber push-out curves measured on a SiC/BN/SiC microcomposite and a 2D SiC/BN/SiC composite (interphase BN4) reinforced with as￾received fibers. Fig. 9. SEM micrographs showing protruding fibers after single fiber push-out tests performed on 2D-SiC/BN/SiC composites with an interphase made of two layers of boron nitride (BN4): (a) composite reinforced with as-received fibers and (b) com￾posite reinforced with treated fibers. 4.4. Push-out and push-in tests on microcomposites The few push-out and push-in curves that were suitable for analysis exhibited the features usually observed with 2D composites (Fig. 8). The interface parameters extracted from these curves are summar￾ized in Table 4. The interfacial shear stresses are larger than those extracted from the tensile stress–strain curves (Fig. 5). They are in excellent agreement with those esti￾mated on the 2D-SiC/BN/SiC composites (Table 4). The debond energy estimates fall within the range obtained from the tensile tests (Fig. 5). The debond stress in microcomposites 2 is quite high (sd 5 1500 MPa), as well as the interfacial shear stress (t 5 90 MPa), which indicates a high resist￾ance to debonding. This may be related to the fact that the fiber could not be pushed out the matrix, even with thin samples. For microcomposites 4, a very good agreement can be noticed between the interfacial shear stresses extracted from respectively the curved domain and the plateau of the push out curves. As previously, the lower interfacial shear stress is obtained for this microcomposite with a double layer coating. Further￾more, push out of the fiber did not require a load as high as that (smax) applied during fiber pushing in the microcomposites 2. SEM showed that debonding occurred essentially at the fiber/coating interface. Generally, really smooth slid surfaces were observed. A certain roughness was detected. 5. DISCUSSION The difficulty to perform push-out tests on thin samples becomes still worse with microcomposites. Large load transfers require very thin embedded lengths in order to permit push-out. The parallel-faced strip is not the most convenient test specimen because it is too fragile, but it remains the only usable sample geometry. During a push-in, unknown maximum debond length, damage of the fiber due to high com￾pressive stresses, bending of the sample and diamond meeting with the matrix are factors that affect the stress–strain curves, and the validity of the extracted characteristics. Then, getting well polished thin strips with a thickness ,200 µm is rare. This explains why the push-out tests could be carried out only on microcomposites 4 which possessed the weakest interfaces. The less aggressive gaseous phase was used for processing the BN2 coating. This coating has been shown to adhere well to the fiber [13]. The resistance to propagation of the interface crack during the push￾out tests appeared to be high, which suggests efficient load transfers under a tensile load. During the tensile tests, the composites 2 experienced a premature fail￾ure whereas the microcomposites 2 exhibited a high strain at saturation. Both features are compatible with efficient load transfers. Ultimate failure of 2D com￾posites involves additionnal phenomena and it is affected by variability in fiber strength degradation during processing. The debond stresses and the interfacial shear stresses are larger than those obtained for composites and microcomposites 4 with a bi-layered coating, and they are comparable to those estimated for composites 1 with an interphase pro￾cessed in more aggressive conditions. As previously mentioned, the propagation of interface cracks during the push-out tests was easier for composites 1 than

REBILLAT et al: SiC/SIC COMPOSITES for composites 2. This result is also indicated by the behavior. The microcomposites 2 possessed the thinn- tress-strain tensile curves of microcomposites I est matrix layer (Table 3) which display a plateau-like feature when comparing Comparable interfacial characteristics and an ident th those of microcomposites 2. This suggests that ical tensile stress-strain behavior were obtained for the fiber BNI bond is weaker. The plateau-like fea- earlier SiC/BN/SiC composites with as-received ture was not observed with the 2D composites. Fur- Nicalon fibers but a different boron nitride [26] thermore, microcomposites I failed at a smaller A difference in ultimate failure can be noticed deformation, which may be attributed to the degra- when comparing the microcomposites and the com- dation of the fiber during Bn processing posites. This difference must be attributed to the The interface between the BN sublayers in com- mechanisms involved. As previously mentioned, th posites 4(with as-received or treated fibers), and in ultimate failure of 2D composites involves fiber inter microcomposites 4 with treated fibers, was found to actions and the individual break of the weakest fibers be comparatively weaker. The fact that the associated prior to instability. The ultimate failure of microcom interfacial shear stresses are the smallest is not posites is thus dictated by the filament strength reflected by the stress-strain curves of the 2D com-(which exhibits a wide statistical distribution), posites nor of the microcomposites (premature whereas that of 2D composites is determined by fiber failure). The interfacial shear stresses are not distinct tow strength(which exhibits a limited scatter) enough to influence the stress-strain behavior. This conclusion agrees with predictions [6, 71 With treated fibers, the fiber/coating bond was 6. CONCLUSION strengthened, as shown by the estimated interfacial In the SiC/BN/SiC microcomposites reinforced shear stresses and the debond stresses. In the com- with untreated fibers, debonding occurred between posites 2 reinforced with treated fibers, oa(which the fiber free surface and the coating. The fiber/BN ay be considered to be commensurate with the inter- interface is the weakest bond in the interfacial face bond strength) and t were significantly sequence. The presence of a very thin layer of carbon increased. In the composites 4 with bi-layered fiber between a SiO2-C mixture layer and the BN coating atings, debonding occurred in the interface between was detected by aeS depth profile analyses. This sub- the BN sublayers. Furthermore, the corresponding layer seems to affect the interface bond shear and debond stresses are small when comparing When the Bn coating was deposited on treated with those pertinent to the other SiC/BN/SiC com- fibers, the fiber/coating interface was stronger, but posites for which debonding took place in the crack deflection did not occur within the BN coating fiber/BN coating interface. It is particularly interest- Only bi-layered BN coatings with a weak interface ing to compare composites 2 and 4 for which a BN2 between the sublayers experienced crack deviation layer is on the fiber. It may be considered that the within the coating. The interface between the bn sub- strength of the fiber/BN2 bond was the same in both layers becomes the weakest link in the interfacial composites. Therefore it may be concluded that the sequence between the fiber and the matrix interface between the bn sublayers was weaker than The micromechanics based models used for the the fiber/BN2 bond analysis of the stress-strain behavior of th The tensile behavior of the 2D SiC/BN/SiC com- microcomposites provided comparable interfacial posites may be considered to agree with the interface shear stresses, despite a certain scatter in some cases haracteristics measured using push-out tests, since A certain discrepancy was observed for the interface close interfacial shear stresses were obtained for the fracture energy estimates, which fell within a range omposites that exhibited a comparable stress-strain of small values(1-10 J/m). Although there might be behavior, whereas the largest interfacial shear stresses some uncertainly in the t data estimated from tensile were observed for the composites 2 with treated fibers tests on microcomposites, they indicated stronger that experienced a premature failure. This conclusion fiber/matrix bonds in the microcomposites fabricated can be also drawn for the microcomposites reinforced with treated fibers with as-received fibers The data determined using the Hsueh's pushout The interfacial shear stresses are in a good corre- model were about one order of magnitude larger thar lation with the stress-strain behavior, except for those extracted from the tensile stress-strain behavior stresses were determined for microcomposites 2 has not been elucidated in the papes discrepancy which possessed the highest density of matrix cracks The tensile stress/strain curves determined with at saturation(Table 3)and gave the most pronounced microcomposites were similar to those obtained for curvature in the stress-strain curve. However, the the 2D composites. Tensile tests on microcomposite thermally-induced residual stresses, resulting from the may be regarded as an interesting approach to inter thermal expansion mismatch between the fibers and face design. Despite the difficulties to perform the the matrix, exert a certain influence. They are push-out tests on microcomposites, this sample increased by a low volume fraction of matrix, which geometry seems to be appropriate to estimate realistic xplain discrepancies in the stress-strain data on the fiber/matrix interfaces in composites

REBILLAT et al.: SiC/SiC COMPOSITES 4617 for composites 2. This result is also indicated by the stress–strain tensile curves of microcomposites 1 which display a plateau-like feature when comparing with those of microcomposites 2. This suggests that the fiber BN1 bond is weaker. The plateau-like fea￾ture was not observed with the 2D composites. Fur￾thermore, microcomposites 1 failed at a smaller deformation, which may be attributed to the degra￾dation of the fiber during BN processing. The interface between the BN sublayers in com￾posites 4 (with as-received or treated fibers), and in microcomposites 4 with treated fibers, was found to be comparatively weaker. The fact that the associated interfacial shear stresses are the smallest is not reflected by the stress–strain curves of the 2D com￾posites nor of the microcomposites (premature failure). The interfacial shear stresses are not distinct enough to influence the stress–strain behavior. This conclusion agrees with predictions [6, 7]. With treated fibers, the fiber/coating bond was strengthened, as shown by the estimated interfacial shear stresses and the debond stresses. In the com￾posites 2 reinforced with treated fibers, sd (which may be considered to be commensurate with the inter￾face bond strength) and t were significantly increased. In the composites 4 with bi-layered fiber coatings, debonding occurred in the interface between the BN sublayers. Furthermore, the corresponding shear and debond stresses are small when comparing with those pertinent to the other SiC/BN/SiC com￾posites for which debonding took place in the fiber/BN coating interface. It is particularly interest￾ing to compare composites 2 and 4 for which a BN2 layer is on the fiber. It may be considered that the strength of the fiber/BN2 bond was the same in both composites. Therefore it may be concluded that the interface between the BN sublayers was weaker than the fiber/BN2 bond. The tensile behavior of the 2D SiC/BN/SiC com￾posites may be considered to agree with the interface characteristics measured using push-out tests, since close interfacial shear stresses were obtained for the composites that exhibited a comparable stress–strain behavior, whereas the largest interfacial shear stresses were observed for the composites 2 with treated fibers that experienced a premature failure. This conclusion can be also drawn for the microcomposites reinforced with as-received fibers. The interfacial shear stresses are in a good corre￾lation with the stress–strain behavior, except for microcomposites 1. Thus, the largest interfacial shear stresses were determined for microcomposites 2 which possessed the highest density of matrix cracks at saturation (Table 3) and gave the most pronounced curvature in the stress–strain curve. However, the thermally-induced residual stresses, resulting from the thermal expansion mismatch between the fibers and the matrix, exert a certain influence. They are increased by a low volume fraction of matrix, which may explain discrepancies in the stress–strain behavior. The microcomposites 2 possessed the thinn￾est matrix layer (Table 3). Comparable interfacial characteristics and an ident￾ical tensile stress–strain behavior were obtained for earlier SiC/BN/SiC composites with as-received Nicalon fibers but a different boron nitride [26]. A difference in ultimate failure can be noticed when comparing the microcomposites and the com￾posites. This difference must be attributed to the mechanisms involved. As previously mentioned, the ultimate failure of 2D composites involves fiber inter￾actions and the individual break of the weakest fibers prior to instability. The ultimate failure of microcom￾posites is thus dictated by the filament strength (which exhibits a wide statistical distribution), whereas that of 2D composites is determined by fiber tow strength (which exhibits a limited scatter). 6. CONCLUSION In the SiC/BN/SiC microcomposites reinforced with untreated fibers, debonding occurred between the fiber free surface and the coating. The fiber/BN interface is the weakest bond in the interfacial sequence. The presence of a very thin layer of carbon between a SiO2–C mixture layer and the BN coating was detected by AES depth profile analyses. This sub￾layer seems to affect the interface bond. When the BN coating was deposited on treated fibers, the fiber/coating interface was stronger, but crack deflection did not occur within the BN coating. Only bi-layered BN coatings with a weak interface between the sublayers experienced crack deviation within the coating. The interface between the BN sub￾layers becomes the weakest link in the interfacial sequence between the fiber and the matrix. The micromechanics based models used for the analysis of the stress–strain behavior of the microcomposites provided comparable interfacial shear stresses, despite a certain scatter in some cases. A certain discrepancy was observed for the interface fracture energy estimates, which fell within a range of small values (1–10 J/m2 ). Although there might be some uncertainly in the t data estimated from tensile tests on microcomposites, they indicated stronger fiber/matrix bonds in the microcomposites fabricated with treated fibers. The data determined using the Hsueh’s pushout model were about one order of magnitude larger than those extracted from the tensile stress–strain behavior of microcomposites. The origin of this discrepancy has not been elucidated in the paper. The tensile stress/strain curves determined with microcomposites were similar to those obtained for the 2D composites. Tensile tests on microcomposites may be regarded as an interesting approach to inter￾face design. Despite the difficulties to perform the push-out tests on microcomposites, this sample geometry seems to be appropriate to estimate realistic data on the fiber/matrix interfaces in composites

REBILLAT et al: SiC/SIC COMPOSITES Finally, the deviation of matrix cracks within the 13. Rebillat, F, Guette, A, Naslain, R and Robin-Brosse, C BN fiber coatings was observed only when a weak interface had been created in the coating. From this 14. Kerans, R J and Parthasarathy, TA,J.Am.CeramSoc. 1991,74(7),1585 e ewpoint, single layered BN coatings which showed 15.Hsueh, C.H,J Am Ceram Soc., 1993, 76(12),3041 limited sensitivity to delamination, were not found 16. Lara Curzio, E and Ferber, M. K, J. Mater. Sci., 1994, as efficient as the pyrocarbon ones. 29,6152 17. Lara Curzio, E, Ferber, M.K. and Lowden, R. A, Ceram. Sci.Eng,1994,15(5)989 Acknowledgements-This work has been supported by the K. J, Rebillat, F and Lamon, J, J. Am. Ceram. n and research an 97,800),506 grant given to F. R and by CNRs. The authors F, LCTS internal report, 1992 uve, J. F, Mocaer, D, Pailler, R, Naslain, R and Lahaye for his hel the AEs J. Mater.Sci,l993,28,1227. Brosse for his contribution to the preparation of 2D composites, issart, N, Rechiniac, C, Roach, D. M, an R. Naslain for valuable discussions and to J. Forget for the uin,J. M, Ceram. Eng. Sci. Proc., 1993, Sept-Oct preparation of the manuscript 22. Lamon, J, Rechiniac, C, Lissart, N and Corne, P in Pr REFERENCES Materials(ECCM5), ed A. R. Bunsell, J. F Jamet and A. I. Evans, A. G, J.A. Ceram. Soc., 1990, 73, 187 Massiah, EACM, Bordeaux, 1992, p. 895 2. Kerans, R J, Hay, R. S, Pagano, J and Parthasarathy T. 23. Come, P, Rechiniac, C. and Lamon, J, in Proceedi A. Am. Cera. Soc. Bu 198 JNC 8, ed O. Allix, J. P. Favre and P Ladeveze. AMAC 3. Naslain, R, Comp. Interf, 1993, 1(3), 253. Paris,1992,p.213 4. Droillard, C. and Lamon, J.,J. Am. Ceram. Soc.. 1996. 24. Lamon, J, Rebillat, F. and Evans, A.G., J. Am. Ceram 79(4)849. Soc.,1995,78(2),401-405. 5. Droillard, C, Elaboration et le composites 25. Naslain, R, Cera a matrice Sic et a interphase sequence C/SiC, Ph. D the- 26. Prouhet,S,CI de la cvd du nitrure de bore dans s, Univ. Bordeaux(France), 1993, no 91 le systeme BF3-NH3-Ar. Application aux composites 6. Guillaumat, L and Lamon, J, IntJ. fract., 1996, 82, 297. nterphase BN. Ph. D. thesis, University of Bor- 7. Lissart, N. and Lamon, J, Acta Metall, 1997, 45(3), 1025 l,1991,no.662 8. Rebillat, F, Lamon, J, Naslain, R, Lara Curzio, E, 27. Rebillat, F, Guette, A, Robin-Brosse, C and Naslain, R Ferber, M. K, and Besmann, T.J. 4m. Ceram. Soc., 1998 Am. Ceram.Soc,1994,77(3),649 81(4),965 28. Naslain, R, Dugne, O, Guette, A, Sevely, S, Robin 9. Pasquier, S, Lamon, J and Naslain, R, Composites Part Rocher. J. P. and Cotteret. ... Am. Ceram. A,1998,29,1157 1991,74,2482. 10. Bertrand, S, Germain, F, Pailler, R. and Lamon, J., in 29. Rebillat, F, Lamon, J, Naslain, R, Lara Curzio, E, Proceedings of the 12th International Conference on Con Ferber, M, K. and Besmann, T. ,J. Am. Ceram Soc, 1998 osite Materials(CCM12, Paris, 5-9 July 1999, in press 81(9)2315 11. Rugg, K. L. Tressler, R. E and Lamon J, J. Eur. Ceram. 30. Marshall, D, B, Cox, B N and Evans, A G, Acta Metall. 985,33(11),2013. 12. Droillard,C, Lamon, J, and Bourret, X, Mat, Res. Soc, 31, cedings of the National Physical Laboratory. IPC Scier ce pro- Symp. Proc., 1994 Socie,1995,365,371 and Technology Press, Surrey, UK

4618 REBILLAT et al.: SiC/SiC COMPOSITES Finally, the deviation of matrix cracks within the BN fiber coatings was observed only when a weak interface had been created in the coating. From this viewpoint, single layered BN coatings which showed a limited sensitivity to delamination, were not found as efficient as the pyrocarbon ones. Acknowledgements—This work has been supported by the french ministry of education and research and SEP through a grant given to F. R. and by C.N.R.S.. The authors are grateful to R. Kerans for his assistance in the push-in tests, to M. Lahaye for his help in the AES experiments, to C. Robin￾Brosse for his contribution to the preparation of 2D composites, to R. Naslain for valuable discussions and to J. Forget for the preparation of the manuscript. REFERENCES 1. Evans, A. G., J. Am. Ceram. Soc., 1990, 73, 187. 2. Kerans, R. J., Hay, R. S., Pagano, J. and Parthasarathy, T. A., Am. Ceram. Soc. Bull., 1989, 68, 429. 3. Naslain, R., Comp. Interf., 1993, 1(3), 253. 4. Droillard, C. and Lamon, J., J. Am. Ceram. Soc., 1996, 79(4), 849. 5. Droillard, C., Elaboration et caracterisation de composites a` matrice SiC et a` interphase se´quence´e C/SiC. Ph.D. the￾sis, Univ. Bordeaux (France), 1993, no. 913. 6. Guillaumat, L. and Lamon, J., Int. J. Fract., 1996, 82, 297. 7. Lissart, N. and Lamon, J., Acta Metall., 1997, 45(3), 1025. 8. Rebillat, F., Lamon, J., Naslain, R., Lara Curzio, E., Ferber, M. K. and Besmann, T., J. Am. Ceram. Soc., 1998, 81(4), 965. 9. Pasquier, S., Lamon, J. and Naslain, R., Composites Part A, 1998, 29, 1157. 10. Bertrand, S., Germain, F., Pailler, R. and Lamon, J., in Proceedings of the 12th International Conference on Com￾posite Materials (ICCM12), Paris, 5–9 July 1999, in press. 11. Rugg, K. L., Tressler, R. E. and Lamon J., J. Eur. Ceram. Soc., in press. 12. Droillard, C., Lamon, J. and Bourrat, X., Mat. Res. Soc. Symp. Proc., 1994 MRS Fall Meeting, Materials Research Society, 1995, 365, 371. 13. Rebillat, F., Guette, A., Naslain, R. and Robin-Brosse, C., J. Eur. Ceram. Soc., 1997, 17, 1403. 14. Kerans, R. J. and Parthasarathy, T. A., J. Am. Ceram. Soc., 1991, 74(7), 1585. 15. Hsueh, C. H., J. Am. Ceram. Soc., 1993, 76(12), 3041. 16. Lara Curzio, E. and Ferber, M. K., J. Mater. Sci., 1994, 29, 6152. 17. Lara Curzio, E., Ferber, M. K. and Lowden, R. A., Ceram. Sci. Eng., 1994, 15(5), 989. 18. Kerans, K. J., Rebillat, F. and Lamon, J., J. Am. Ceram. Soc., 1997, 80(2), 506. 19. Rebillat, F., LCTS internal report, 1992. 20. Villeneuve, J. F., Mocaer, D., Pailler, R., Naslain, R. and Olry, P., J. Mater. Sci., 1993, 28, 1227. 21. Lamon, J., Lissart, N., Rechiniac, C., Roach, D. M. and Jouin, J. M., Ceram. Eng. Sci. Proc., 1993, Sept–Oct, 1115. 22. Lamon, J., Rechiniac, C., Lissart, N. and Corne, P., in Pro￾ceedings of the Fifth European Conference on Composite Materials (ECCM5), ed. A. R. Bunsell, J. F. Jamet and A. Massiah. EACM, Bordeaux, 1992, p. 895. 23. Corne, P., Rechiniac, C. and Lamon, J., in Proceedings JNC 8, ed. O. Allix, J. P. Favre and P. Ladeveze. AMAC, Paris, 1992, p. 213. 24. Lamon, J., Rebillat, F. and Evans, A. G., J. Am. Ceram. Soc., 1995, 78(2), 401–405. 25. Naslain, R., Ceram. Trans., 1995, 58, 23. 26. Prouhet, S., Cinetique de la CVD du nitrure de bore dans le systeme BF3–NH3–Ar. Application aux composites SiC/SiC a` interphase BN. Ph.D. thesis, University of Bor￾deaux I, 1991, no. 662. 27. Rebillat, F., Guette, A., Robin-Brosse, C. and Naslain, R., J. Am. Ceram. Soc., 1994, 77(3), 649. 28. Naslain, R., Dugne, O., Guette, A., Sevely, S., Robin￾Brosse, C., Rocher, J. P. and Cotteret, J., J. Am. Ceram. Soc., 1991, 74, 2482. 29. Rebillat, F., Lamon, J., Naslain, R., Lara Curzio, E., Ferber, M. K. and Besmann, T., J. Am. Ceram. Soc., 1998, 81(9), 2315. 30. Marshall, D. B., Cox, B. N. and Evans, A. G., Acta Metall., 1985, 33(11), 2013. 31. Aveston, J., Cooper, A. and Kelly, A., in Conference Pro￾ceedings of the National Physical Laboratory. IPC Science and Technology Press, Surrey, UK, 1971, p. 15