E≈S Journal of the European Ceramic Society 20(2000)1929-1938 Sic/SiC minicomposites with structure-graded BN interphases S Jacques, A. Lopez-Marure, C. Vincent, H. Vincent, J. Bouix Laboratoire des multimateriaux et Interfaces- UMR 5615 CNRS/University of Lyon I 43. boulevard du // Novembre 1918. F-69622 Villeurbanne Cedex france Received 28 October 1999: received in revised form 10 February 2000: accepted 14 February 2000 Abstract BN interphases in SiC/Sic minicomposites were produced by infiltration of fibre tows from BF3-NH3-H2 gaseous system dur- ing interphase one-step processing, the tow travels through a reactor containing a succession of different hot areas. By tem char- acterization, the Bn interphases were found to be made of a structural gradient from isotropic to highly anisotropic. The very first coating is poorly organised and allows to protect the fibre from a further chemical attack by the reactant mixture. The mini- composites were tensile tested at room temperature with unloading-reloading cycles. The Bn interphases act as mechanical the fibre/matrix bonding intensity ranges from weak to rather strong depending on the tow travelling rate during interphase infill tration. The specimen lifetimes at 700C under a constant tensile loading were measured in dry and moist air. Compared to a pyr ocarbon reference interphase, the bn interphases significantly improve the oxidation resistance of the Sic/Sic minicomposit C 2000 Elsevier Science Ltd. All rights reserved Keywords: BN; Composites: Electron microscopy; Interphase: Oxidation resistance; SiC/SiC 1. Introduction 2. Experimental procedure Sic/SiC type ceramic matrix composites, a good 2.1. Samples toughness can be achieved by adding between the fibre ind the brittle matrix a thin film of a compliant material The samples used in this study were SiC/SiC mini- called"interphase. If highly anisotropic pyrolytic(ex- composites. A minicomposite is a ID model composite BF3)boron nitride can replace advantageously the poor It consists of a tow in which the Bn or PyC interphase oxidation resistance pyrocarbon(PyC), its chemical and then the SiC matrix are infiltrated by CVI. The tow vapour infiltration(CVI)processing requires to protect is composed of around 500 Hi-Nicalon fibres(mono the fibre from a chemical attack with an isotropic Bn filaments)(from Nippon Carbon, Japan)(Fig. 1) pre-coating obtained with less aggressive conditions The geometry of the minicomposite is simpler than Thus, in the case of a classical static isothermal CVI, the that of the actual composite. It can properly simulate its N interphase must be infiltrated by following several behaviour and specially enables to determine the inter separate steps (i.e. changing the boron precursor gas phase properties and CVI parameters during the experiment.) In our case, the fibre volume fraction Vr of the mini- The aim of the present work was to prepare within ID composites was about 20%(measured by weighing) minicomposites bn interphases with structural gra- dients by using a one-step continuous dynamic CVI 2.2. Interphase processing process and to characterize these minicomposites The process used for the interphase preparatic new process called gradient-chemical vapour infiltration ( TG-CVi) le reactor was a hot wal furnace: a graphite 35 cm) placed inside a silica tube(inner diameter: 24 mm) is heated by dm.univ-lyonlfr(S Jacques). induction. The fibre tow was unwound from a first spool 0955-2219/00/S-see front matter C 2000 Elsevier Science Ltd. All rights reserved PII:S0955-2219(00)00064-9
SiC/SiC minicomposites with structure-graded BN interphases S. Jacques *, A. Lopez-Marure, C. Vincent, H. Vincent, J. Bouix Laboratoire des MultimateÂriaux et Interfaces Ð UMR 5615 CNRS/University of Lyon 1, 43, boulevard du 11 Novembre 1918, F±69622 Villeurbanne Cedex, France Received 28 October 1999; received in revised form 10 February 2000; accepted 14 February 2000 Abstract BN interphases in SiC/SiC minicomposites were produced by in®ltration of ®bre tows from BF3±NH3±H2 gaseous system. During interphase one-step processing, the tow travels through a reactor containing a succession of dierent hot areas. By TEM characterization, the BN interphases were found to be made of a structural gradient: from isotropic to highly anisotropic. The very ®rst coating is poorly organised and allows to protect the ®bre from a further chemical attack by the reactant mixture. The minicomposites were tensile tested at room temperature with unloading-reloading cycles. The BN interphases act as mechanical fuses; the ®bre/matrix bonding intensity ranges from weak to rather strong depending on the tow travelling rate during interphase in®ltration. The specimen lifetimes at 700C under a constant tensile loading were measured in dry and moist air. Compared to a pyrocarbon reference interphase, the BN interphases signi®cantly improve the oxidation resistance of the SiC/SiC minicomposites. # 2000 Elsevier Science Ltd. All rights reserved. Keywords: BN; Composites; Electron microscopy; Interphase; Oxidation resistance; SiC/SiC 1. Introduction In SiC/SiC type ceramic matrix composites, a good toughness can be achieved by adding between the ®bre and the brittle matrix a thin ®lm of a compliant material called ``interphase''.1 If highly anisotropic pyrolytic (exBF3) boron nitride can replace advantageously the poor oxidation resistance pyrocarbon (PyC), its chemical vapour in®ltration (CVI) processing requires to protect the ®bre from a chemical attack with an isotropic BN pre-coating obtained with less aggressive conditions.2,3 Thus, in the case of a classical static isothermal CVI, the BN interphase must be in®ltrated by following several separate steps (i.e. changing the boron precursor gas and CVI parameters during the experiment...). The aim of the present work was to prepare within 1D minicomposites BN interphases with structural gradients by using a one-step continuous dynamic CVI process and to characterize these minicomposites. 2. Experimental procedure 2.1. Samples The samples used in this study were SiC/SiC minicomposites. A minicomposite is a 1D model composite. It consists of a tow in which the BN or PyC interphase and then the SiC matrix are in®ltrated by CVI. The tow is composed of around 500 Hi-Nicalon ®bres (mono- ®laments) (from Nippon Carbon, Japan) (Fig. 1). The geometry of the minicomposite is simpler than that of the actual composite. It can properly simulate its behaviour and specially enables to determine the interphase properties. In our case, the ®bre volume fraction Vf of the minicomposites was about 20% (measured by weighing). 2.2. Interphase processing The process used for the interphase preparation was a new process called thermal gradient-chemical vapour in®ltration (TG-CVI).4,5 The reactor was a hot wall furnace: a graphite susceptor (length: 35 cm) placed inside a silica tube (inner diameter: 24 mm) is heated by induction. The ®bre tow was unwound from a ®rst spool 0955-2219/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved. PII: S0955-2219(00)00064-9 Journal of the European Ceramic Society 20 (2000) 1929±1938 * Corresponding author. E-mail address: sylvain.jacques@adm.univ-lyon1.fr (S. Jacques)
1930 S Jacques et al. Journal of the European Ceramic Society 20(2000)1929-1938 Three batches of minicomposites called 1, 2 and 3 containing BN interphases obtained respectively with a tow rate r of 2, 2.5 and 3 m/h were prepared. A fourth batch called 0 with a pyrocarbon(PyC) interphase was also prepared in order to be used as reference. The pyrocarbon was infiltrated by CVI but using a classical induction device (i.e. constant pitch of coil turns and consequently one hot area) from propane (@cHs=9.6 cm'min-)under a pressure of 10 kPa and with r=6 m/h. Argon was used as the gas vector (@Ar=38.5 cm3 min-). The maximum temperature in the centre of the hot area was 1050oc. unlike the bn interphase infiltration, gas inlet and outlet were permuted in order to have a gas flux and a tow displacement in reverse Fig 1. Schema matics of a minicomposite prepared by CVi(for clarity, only direction and thus a good adhesion between the fibre a few fibres out of the 500 that constitute a whole tow are represented) and the coating through the susceptor. The outlet reactor winding around 2.3. SiC matrix processing a second spool controlled the fibre travelling rate r Thus, the reactor contained three distinct hot areas with flow rate was 70 cm p e For BN interphase processing, a specific temperature The Sic matrix was infiltrated by classic profile was obtained inside the susceptor thanks to second reactor from CH3 SiCl3/H2 precursor gases at variable pitch of induction heating coil turns(Fig. 2). 1100C. The ratio @1 CH,SiCh, was 0.75, the d 5 The respectively a"low temperature(<1100C), a medium duration of matrix infiltration was 5 h. maximum temperature (1150C) and a high one (1250C). The control of the fibre displacement (r) 2. 4. Charaterization through the hot areas allowed to determine the duration of the CvI treatment and consequently the interphase 2. 1. Mechanical behaviour thickness. BF3/NH3 were the gas precursors of BN. The minicomposites were tensile tested at room tem- argon was used as the gas vector but also to dilute the perature with unloading-reloading cycles using a MTS reactive species. The total gas flow rate Q was 60 cm Adamel DY-22 machine(Ivry sur Seine, France)equip- then gripped into the testing machine h an extensometer (Ingstrom, Cat. No. 2620-602 travel: 2.5 mm, type dynamic, code value 136)directly gripped on the minicomposite itself. The extensometer 7 A 90. 70060 gauge length(Lg) was 40 mm. Five tests per batch were carried out fibre tow In order to evaluate the fibre/ nding. the interfacial shear stress t was estimated by following the same methods as those used by bertrand et al. They are presented in outline below The first method considers, for unloading-reloading cycles, the width SA of the hysteresis loop at a given en by Eq(1) b2N(1-a1)R/ Er Fig. 2. Schematics of the TG-CVI reactor(a) and susceptor tempera d
through the susceptor. The outlet reactor winding around a second spool controlled the ®bre travelling rate r. For BN interphase processing, a speci®c temperature pro®le was obtained inside the susceptor thanks to a variable pitch of induction heating coil turns (Fig. 2). Thus, the reactor contained three distinct hot areas with respectively a ``low'' temperature (<1100C), a medium maximum temperature (1150C) and a high one (1250C). The control of the ®bre displacement (r) through the hot areas allowed to determine the duration of the CVI treatment and consequently the interphase thickness. BF3/NH3 were the gas precursors of BN, argon was used as the gas vector but also to dilute the reactive species. The total gas ¯ow rate Q was 60 cm3 minÿ1 . The gas composition was: QNH3 /QBF3 =0.5 and QAr/(QBF3 +QNH3 )=0.222. The gas pressure was 20 kPa. Three batches of minicomposites called 1, 2 and 3 containing BN interphases obtained respectively with a tow rate r of 2, 2.5 and 3 m/h were prepared. A fourth batch called 0 with a pyrocarbon (PyC) interphase was also prepared in order to be used as a reference. The pyrocarbon was in®ltrated by CVI but using a classical induction device (i.e. constant pitch of coil turns and consequently one hot area) from propane (QC3H8=9.6 cm3 minÿ1 ) under a pressure of 10 kPa and with r=6 m/h. Argon was used as the gas vector (QAr=38.5 cm3 minÿ1 ). The maximum temperature in the centre of the hot area was 1050C. Unlike the BN in®ltration, gas inlet and outlet were permuted in order to have a gas ¯ux and a tow displacement in reverse direction and thus a good adhesion between the ®bre and the coating.6 2.3. SiC matrix processing The SiC matrix was in®ltrated by classical CVI in a second reactor from CH3SiCl3/H2 precursor gases at 1100C. The ratio QH2 /QCH3SiCl3 was 0.75, the total gas ¯ow rate was 70 cm3 minÿ1 and pressure 5 kPa. The duration of matrix in®ltration was 5 h. 2.4. Charaterization 2.4.1. Mechanical behaviour The minicomposites were tensile tested at room temperature with unloading-reloading cycles using a MTSAdamel DY-22 machine (Ivry sur Seine, France) equipped with a 1 kN load cell. The minicomposite ends were glued within 50 mm distant metallic tubes that were then gripped into the testing machine jaws. The crosshead speed was 0.1 mm/min. The strain was measured with an extensometer (IngstroÈm, Cat. No. 2620-602, travel: 2.5 mm, type dynamic, code value 136) directly gripped on the minicomposite itself. The extensometer gauge length (Lg) was 40 mm. Five tests per batch were carried out. In order to evaluate the ®bre/matrix bonding, the interfacial shear stress was estimated by following the same methods as those used by Bertrand et al.7 They are presented in outline below. The ®rst method considers, for unloading-reloading cycles, the width of the hysteresis loop at a given stress '. is given by Eq. (1):8 b2N 1 ÿ a1Vf ÿ 2 Rf 2V2 fEm 2 p 0 p 1 ÿ 0 p 1 with a1 Ef E0 and Fig. 1. Schematics of a minicomposite prepared by CVI (for clarity, only a few ®bres out of the 500 that constitute a whole tow are represented). Fig. 2. Schematics of the TG-CVI reactor (a) and susceptor temperature pro®le (b). 1930 S. Jacques et al. / Journal of the European Ceramic Society 20 (2000) 1929±1938
S Jacques et al. Journal of the European Ceramic Society 20(2000)1929-1938 (1+vEmEr +(1-2v)Eol 2.43. B microstructure and texture Erl(+DEr+(I-v)Eol Thin longitudinal sections of minicomposites after tensile test were studied by transmission electron microscopy(TEM: Topcon 002B, Japan) using bright- where o is the maximum stress of the cycle, Eo is the field, high resolution(HR) and selected area diffraction initial Youngs modulus of the minicomposite, Er(280 (SAD) techniques. The samples were embedded in a GPa) and Em (400 GPa)are the Youngs moduli of ceramic cement and mechanically thinned. The thin he fibre and the matrix, respectively, v is the SiC Pois- sheets were then ion-milled(600 Duo Mill from Gatan sons ratio (vavevm0. 2), and R is the fibre radius USA)until the electron transparency (R7 um).84 was measured on the last cycle befor the failure. N is the number of matrix cracks. It has be measured by optical microscopy on polished long itudinal sections of the failed minicomposites after che- mical etching(Murakami reactant) in order to reveal 3. 1. Mechanical behaviour the matrix microcracks closed during unloading The second method considers the matrix crack spacing 3. .. Tensile tests distance Is(=N/L,. t is given by Eq( 2). 9. 0 Average tensile mechanical characteristics at ambient temperature for the four batches of minicomposites are os Re Err (2) Presented in Table 1. Fig 3 displays a typical force-strain Table verage tensile mechanical characteristics of each batch of mini- where os is the applied stress at matrix cracking satura composites. ion and Vm the matrix volume fraction(Vml-Ve) In Batch r (m/h) EPL (%) FPL(N EF(% FF(N) he present case, because of a low fibre volume fraction (Ve s0. 2), the minicomposites failed before reaching the 0(PyC) 6 56 cracking saturation. Therefore, os was similar to the 2 After ultimate failure, the morphology of the fracture EPL and FPL are the proportional limit and force, EF and FF are surfaces was observed with a scanning electron microscope the strain and force at failure (SEM)(HITACHI S800) 2. 2. Oxidation test 180 Interphase 0 In order to study the oxidation resistance of the inter- phases, the minicomposites were submitted to thermal ageing in air under a static loading following a procedure 120 similar to that described by Lebrun et al. Each speci- F(N men, with a 50 mm gauge length, was vertically main- tained in the hot area of a furnace between two alumina tubes with an alumina-based adhesive(Armco ref. 603 SA). At each alumina rod ends, hooks allowed device self-alignment. Once the temperature reached the set point of 700oC, a 9.5 kg load(corresponding to a force E(%) E(%) 180 of 93 N)was very carefully hooked to the bottom end of Interphase 2 the alumina tube in the cold area. The specimen lifetime (i.e. the time before failure) was automatically measured 120 by means of a switch connected with a timer which F(N FON detects the load fall. The test originality lied in the pos sibility to expose the minicomposite either to a dry air stream(from the top to the bottom obtained from liquid air evaporation or to a moist air obtained by sparging in 40 C liquid water. In the case of moist air, the air inlet pipe was maintained at 50 C between the sparger and the furnace; this device ensured constant Fig 3. Typical tensile force-strain curves with unloading-reloading moisture content for each experiment cycles for minicomposites with different kinds of interphase
b2 1 Em Ef 1 ÿ 2E0 Ef 1 Ef 1 ÿ E0 where p is the maximum stress of the cycle, E0 is the initial Young's modulus of the minicomposite, Ef (280 GPa) and Em (400 GPa) are the Young's moduli of the ®bre and the matrix, respectively, is the SiC Poisson's ratio (nnfnm0.2), and Rf is the ®bre radius (Rf7 mm). was measured on the last cycle before the failure. N is the number of matrix cracks. It has been measured by optical microscopy on polished longitudinal sections of the failed minicomposites after chemical etching (Murakami reactant) in order to reveal the matrix microcracks closed during unloading. The second method considers the matrix crack spacing distance ls (=N/Lg). is given by Eq. (2).9,10 sRf 2Vfls 1 EfVf EmVm 2 where s is the applied stress at matrix cracking saturation and Vm the matrix volume fraction (Vm1ÿVf). In the present case, because of a low ®bre volume fraction (Vf 0.2), the minicomposites failed before reaching the cracking saturation. Therefore, s was similar to the failure stress. After ultimate failure, the morphology of the fracture surfaces was observed with a scanning electron microscope (SEM) (HITACHI S800). 2.4.2. Oxidation tests In order to study the oxidation resistance of the interphases, the minicomposites were submitted to thermal ageing in air under a static loading following a procedure similar to that described by Lebrun et al.11 Each specimen, with a 50 mm gauge length, was vertically maintained in the hot area of a furnace between two alumina tubes with an alumina-based adhesive (Aremco ref. 603, USA). At each alumina rod ends, hooks allowed device self-alignment. Once the temperature reached the setpoint of 700C, a 9.5 kg load (corresponding to a force of 93 N) was very carefully hooked to the bottom end of the alumina tube in the cold area. The specimen lifetime (i.e. the time before failure) was automatically measured by means of a switch connected with a timer which detects the load fall. The test originality lied in the possibility to expose the minicomposite either to a dry air stream (from the top to the bottom) obtained from liquid air evaporation or to a moist air obtained by sparging in 40C liquid water. In the case of moist air, the air inlet pipe was maintained at 50C between the sparger and the furnace; this device ensured constant moisture content for each experiment. 2.4.3. BN microstructure and texture Thin longitudinal sections of minicomposites after tensile test were studied by transmission electron microscopy (TEM: Topcon 002B, Japan) using bright- ®eld, high resolution (HR) and selected area diraction (SAD) techniques. The samples were embedded in a ceramic cement and mechanically thinned. The thin sheets were then ion-milled (600 Duo Mill from Gatan, USA) until the electron transparency. 3. Results 3.1. Mechanical behaviour 3.1.1. Tensile tests Average tensile mechanical characteristics at ambient temperature for the four batches of minicomposites are presented in Table 1. Fig. 3 displays a typical force±strain Fig. 3. Typical tensile force±strain curves with unloading-reloading cycles for minicomposites with dierent kinds of interphases. Table 1 Average tensile mechanical characteristics of each batch of minicomposites.a Batch r (m/h) "PL (%) FPL (N) "F (%) FF (N) 0 (PyC) 6 0.09 111 0.56 169 1 2 0.04 59 0.35 128 2 2.5 0.10 119 0.53 167 3 3 0.06 83 0.55 137 a "PL and FPL are the proportional limit and force, "F and FF are the strain and force at failure S. Jacques et al. / Journal of the European Ceramic Society 20 (2000) 1929±1938 1931
1932 S Jacques et al. /Journal of the European Ceramic Society 20(2000)1929-1938 curve for each kind of interphase. In every case, an fibre/matrix bonding. Here, the interphase remains extended non-linear domain evidencing matrix micro- fixed on the fibre as shown in Fig. 5(a) where the rough cracking and fibre/matrix debonding follows the initial and tortured BN coating surface appears. A highest inear elastic region. Therefore, the four kinds of inter- magnification observation [Fig. 5(b)]shows a debonding phase act as mechanical fuses that occurs within the interphase itself. Minicomposites of batches 0 and 2 exhibit sim laI Concerning batch 3 [Fig. 60 the pull-out lengths tensile behaviours with small residual stresses. The are medium(reaching 100 Hm). If the surface of the identical average failure forces for both batches are the pulled out fibres seems to be smooth as in the case of highest. Therefore, BN interphase 2 processed with batch 1, a meticulous observation shows that they are r=2.5 m/h is as good at room temperature as reference not bare but coated with the whole BN interphase [Fig PyC interphase. Batch 3 obtained with r=3 m/h exhi- 6(b)]. Here, the debonding occurs near the matrix bits weakest mechanical properties with important resi dual strains evidencing that the fibre/matrix load 3. 2. TEM characterization transfer is less good. The characteristics decrease more for batch I interphase prepared with the slowest rate of 3. 2. Batch 2 minicomposite ow travelling. Thus, with these processing conditions, Bright-field observation of interphase 2(Fig. 7)shows the intermediate r of 2.5 m/h appears to result in an that it is made of two sublayers with different textures timum The average proportional limit force for each batch evolves in the same way. These results need not mean debond that for the best composites(batches 0 and 2)the matrix begins to crack later during the composite tension. But as the load transfer is better, the very first crack coming out does not lead to a deflection from linearity detect- able on curves at low load level atrIx 3. 1.2. Interfacial shear stress t Interfacial shear stresses measured by following the different methods(Table 2) bear out the different inter- Fig 4. SEM obsevations of the failure surface of a batch I mini phase characteristics. However, a distinction appears composite: (a)important fibre pull-out; (b)debonding in the fibre/ between batches 0 and 2. If the fibre/matrix bond is interphase interface strong with Pyc interphase, the lowest T values obtained with batch 2 show that the bn coating results in an intermediate bond fibre 3. 1.3. SEM observation of fracture surfaces SEM observation of batch I minicomposite fracture surfaces after tensile tests [Fig. 4(a)] shows an important fibre pull-out. The pull-out lengths can exceed 200 um matrix The surface of the pulled out fibres is smooth and free of any coating. The debonding occurs in the fibre/inter- phase interface as evidenced by highest magnification observation [Fig 4(b). This large debonding corrobo rates the measured low value of t Fig. 5. SEM observations of the failure of a batch 2 minicomposite In the case of batch 2, the fibre pull-out is small and (a)part of BN coating fixed on pull-out fibre; (b)debonding within the generally lower than 50 um evidencing the strongest Average interfacial shear stresses r (%) Is(um) N Eo(GPa) as(MPa) 0.(MPa) o(MPa) I(MPa) 0 (PyC) 73 19.2 17.5 128
curve for each kind of interphase. In every case, an extended non-linear domain evidencing matrix microcracking and ®bre/matrix debonding follows the initial linear elastic region. Therefore, the four kinds of interphase act as mechanical fuses. Minicomposites of batches 0 and 2 exhibit similar tensile behaviours with small residual stresses. The identical average failure forces for both batches are the highest. Therefore, BN interphase 2 processed with r=2.5 m/h is as good at room temperature as reference PyC interphase. Batch 3 obtained with r=3 m/h exhibits weakest mechanical properties with important residual strains evidencing that the ®bre/matrix load transfer is less good. The characteristics decrease more for batch 1 interphase prepared with the slowest rate of tow travelling. Thus, with these processing conditions, the intermediate r of 2.5 m/h appears to result in an optimum. The average proportional limit force for each batch evolves in the same way. These results need not mean that for the best composites (batches 0 and 2) the matrix begins to crack later during the composite tension. But, as the load transfer is better, the very ®rst crack coming out does not lead to a de¯ection from linearity detectable on curves at low load level. 3.1.2. Interfacial shear stress Interfacial shear stresses measured by following the dierent methods (Table 2) bear out the dierent interphase characteristics. However, a distinction appears between batches 0 and 2. If the ®bre/matrix bond is strong with PyC interphase, the lowest values obtained with batch 2 show that the BN coating results in an intermediate bond. 3.1.3. SEM observation of fracture surfaces SEM observation of batch 1 minicomposite fracture surfaces after tensile tests [Fig. 4(a)] shows an important ®bre pull-out. The pull-out lengths can exceed 200 mm. The surface of the pulled out ®bres is smooth and free of any coating. The debonding occurs in the ®bre/interphase interface as evidenced by highest magni®cation observation [Fig. 4(b)]. This large debonding corroborates the measured low value of . In the case of batch 2, the ®bre pull-out is small and generally lower than 50 mm evidencing the strongest ®bre/matrix bonding. Here, the interphase remains well ®xed on the ®bre as shown in Fig. 5(a) where the rough and tortured BN coating surface appears. A highest magni®cation observation [Fig. 5(b)] shows a debonding that occurs within the interphase itself. Concerning batch 3 [Fig. 6(a)], the pull-out lengths are medium (reaching 100 mm). If the surface of the pulled out ®bres seems to be smooth as in the case of batch 1, a meticulous observation shows that they are not bare but coated with the whole BN interphase [Fig. 6(b)]. Here, the debonding occurs near the matrix. 3.2. TEM characterization 3.2.1. Batch 2 minicomposite Bright-®eld observation of interphase 2 (Fig. 7) shows that it is made of two sublayers with dierent textures Table 2 Average interfacial shear stresses batch V0 p % ls(mm) N E0(GPa) s(MPa) p(MPa) 0 (MPa) (mm) (MPa) Eq. (1) Eq. (2) 0 (PyC) 22.9 87 462 350 503 455 223 31 117 73 1 18.4 408 98 348 306 311 155 41 15 12 2 19.2 206 194 305 417 409 200 24 75 32 3 17.5 279 143 328 269 255 128 32 29 20 Fig. 4. SEM obsevations of the failure surface of a batch 1 minicomposite: (a) important ®bre pull-out; (b) debonding in the ®bre/ interphase interface. Fig. 5. SEM observations of the failure of a batch 2 minicomposite: (a) part of BN coating ®xed on pull-out ®bre; (b) debonding within the interphase. 1932 S. Jacques et al. / Journal of the European Ceramic Society 20 (2000) 1929±1938
S Jacques et al./Journal of the European Ceramic Society 20(2000)1929-1938 and equivalent in thickness. The total interphase thick form. In any cases, this interface observation (per- ness is about 450 nm. While the interface between both formed after mechanical test) does not reveal any sublayers is regular and linear, the interphase /matrix debonding interface appears on this scale irregular ("serrated") Close to the fibre surface, the deposited BN appears first owing to the growth of extended coherent domains(cor- poorly organised with limited coherent domains and 002 responding to high bright/dark contrast areas)where bn planes orientated at random. Then, the degree of struc organisation is very high. On the contrary, the weak tural anisotropy increases progressively from the fibre to contrast of the second sublayer located near the fibre the interface of both interphase sublayers (i.e. Fig 9, from evidences a poorer crystallographical orga nisation the left to the right): the coherent domains expand along The SAD patterns performed in both sublayers(Fig. stacking c axis(ranging from 10 to 15 fringes)as well as are typical, with their 002 arcs, of an anisotropic tur- longitudinally (parallel to the fibre axis)(fringes length matrix, spots occurrence in the SAD pattern reveals the denced within the first BN interphase sublayer ent is evi- bostratic material. For the sublayer located near the 10 nm)(Fig. 10, left). Thus, a structural gradi local existence of bn submicrometric polycrystalline structures. Furthermore, the smaller arc length(or arc- opening angle in azimuth) evidences a better 002 planes 200nm average orientation than near the fibre HR observation of fibre/Bn interface(Fig. 9, left) shows that it is made of a thin amorphous sublayer (thickness <5 nm). This layer can(i) either result from a Hi-Nicalon fibre surface carbon enrichment, 12-14 or(ii)be Bn that begins to depositate on the fibre in amorphous matrixes SAD SAD 、 Fig. 6. SeM observations of the failure surface batch 3 mini Fig. 8. Bright-field TEM image of the interfacial zone of a batch 2 composite:(a) smooth pulled out fibres;(b)BN interphase fixed minicomposite and corresponding SAd patterns(negatives) pulled out fibre 400nm Fig. 7. Bright field TEM image of the interphase of a batch 2 mini- Fig 9. HR-TEM image of the interphase of a batch 2 minicomposite
and equivalent in thickness. The total interphase thickness is about 450 nm. While the interface between both sublayers is regular and linear, the interphase/matrix interface appears on this scale irregular (``serrated'') owing to the growth of extended coherent domains (corresponding to high bright/dark contrast areas) where BN organisation is very high. On the contrary, the weak contrast of the second sublayer located near the ®bre evidences a poorer crystallographical organisation. The SAD patterns performed in both sublayers (Fig. 8) are typical, with their 002 arcs, of an anisotropic turbostratic material. For the sublayer located near the matrix, spots occurrence in the SAD pattern reveals the local existence of BN submicrometric polycrystalline structures. Furthermore, the smaller arc length (or arcopening angle in azimuth) evidences a better 002 planes average orientation than near the ®bre. HR observation of ®bre/BN interface (Fig. 9, left) shows that it is made of a thin amorphous sublayer (thickness<5 nm). This layer can (i) either result from a Hi-Nicalon ®bre surface carbon enrichment,12ÿ14 or (ii) be BN that begins to depositate on the ®bre in amorphous form. In any cases, this interface observation (performed after mechanical test) does not reveal any debonding. Close to the ®bre surface, the deposited BN appears ®rst poorly organised with limited coherent domains and 002 planes orientated at random. Then, the degree of structural anisotropy increases progressively from the ®bre to the interface of both interphase sublayers (i.e. Fig. 9, from the left to the right): the coherent domains expand along stacking c axis (ranging from 10 to 15 fringes) as well as longitudinally (parallel to the ®bre axis) (fringes length 10 nm) (Fig. 10, left). Thus, a structural gradient is evidenced within the ®rst BN interphase sublayer. Fig. 6. SEM observations of the failure surface batch 3 minicomposite: (a) smooth pulled out ®bres; (b) BN interphase ®xed on pulled out ®bre. Fig. 7. Bright ®eld TEM image of the interphase of a batch 2 minicomposite. Fig. 8. Bright-®eld TEM image of the interfacial zone of a batch 2 minicomposite and corresponding SAD patterns (negatives). Fig. 9. HR-TEM image of the interphase of a batch 2 minicomposite near the ®bre. S. Jacques et al. / Journal of the European Ceramic Society 20 (2000) 1929±1938 1933
1934 S Jacques et al. Journal of the European Ceramic Society 20(2000)1929-1938 fibre Fig 10. HR-TEM image of the middle of the interphase of a batch 2 The observation of the second sublayer located near the matrix(Fig. 10, right) reveals a very high structural organisation even if some 002 fringe distortions still remain. The fringes can exceed 50 nm in length. These large coherent domain sizes explain the sublayer thick debonding ness inhomogeneity and consequently the interphase/ matrix interface irregularity. 800nm 3.2.2. Batch I minicomposite As in the case of batch 2, batch I minicomposite study expected because of the slower rate of toy ayers. As Fig. 11. Bright-field TEM image of the interphase of a batch I mini- Fig. 11)evidences the occurrence of two su (r=2 m/h for batch I against 2.5 m/h for batch 2)and (located near the matrix) is made of an isotropic BN, hence a higher reactor resident time, the total interphase 002 diffraction arcs meet together and form continuous thickness is higher: about 950 nm. But here, the second rings; there is no more preferential orientation. Here sublayer located near the matrix is homogeneous in again, this difference with batch 2 can be explained by thickness. Another important difference: in addition, two the slower rate of tow displacement. The gaseous phase dark"edges"appear at the fibre/interphase interface infiltrated in the tow porosity is drag off more slowly to with a large debonding located between the both; this the hottest areas, which allows it to be raised to a high debonding extends over long distances est temperature. Homogeneous nucleation can be then From a comparison between a SAd performed on favoured to the detriment of heterogeneous chemical these dark edges and a Sad performed more inside the surface reactions. This phenomenon associated with a fibre(Fig. 12, bottom), it is clear that the SiC fibre sur- high yield of homogeneous reaction, results in an face has been crystallised. Both SAD patterns are typical important production of HF and intermediate species of Sic but, whereas diffraction rings are quasi-con- and consequently leads to a material disorganisation tinuous within the fibre (Sic nanometric grains size) these are spotted at the fibre surface(more extended Sic 3. 2.3. Batch 3 minicomposite grains size). This crystallisation has not been observed Batch 3 interphase is also made of two sublayers(Fig. in batch 2. For batch l, it can be due to a slower rate of 13): a first preponderant one located near the fibre tow displacement, which gives the fibre enough time to identical with those of the other batches, about 250 nm be raised to a highest temperature while going through in thickness, and a very thin one located near the matrix he reactor hottest area(1250C). This thermal treat- (about 20 nm in thickness)where BN organisation is ment is responsible for the fibre crystallisation that very pronounced. The faster tow rate (3 m/h)can starts from its surface, causes local embrittlement and explain this disproportion. The fibre resident time in the results in a large debonding hottest area is reduced; the second organised sublayer SAD observations of interphase(Fig. 12, top) show has no more time to grow as in the case of batch 2. But, that the first bn sublayer (located near the fibre) is contrary to batch 2, this restricted growth allows this identical with that of batch 2: it is an anisotropic tur- sublayer to keep a constant thickness and therefore to bostratic BN. On the contrary, the second sublayer form a regular and linear interface with the matrix
The observation of the second sublayer located near the matrix (Fig. 10, right) reveals a very high structural organisation even if some 002 fringe distortions still remain. The fringes can exceed 50 nm in length. These large coherent domain sizes explain the sublayer thickness inhomogeneity and consequently the interphase/ matrix interface irregularity. 3.2.2. Batch 1 minicomposite As in the case of batch 2, batch 1 minicomposite study (Fig. 11) evidences the occurrence of two sublayers. As expected because of the slower rate of tow travelling (r=2 m/h for batch 1 against 2.5 m/h for batch 2) and hence a higher reactor resident time, the total interphase thickness is higher: about 950 nm. But here, the second sublayer located near the matrix is homogeneous in thickness. Another important dierence: in addition, two dark ``edges'' appear at the ®bre/interphase interface with a large debonding located between the both; this debonding extends over long distances. From a comparison between a SAD performed on these dark edges and a SAD performed more inside the ®bre (Fig. 12, bottom), it is clear that the SiC ®bre surface has been crystallised. Both SAD patterns are typical of SiC but, whereas diraction rings are quasi-continuous within the ®bre (SiC nanometric grains size), these are spotted at the ®bre surface (more extended SiC grains size). This crystallisation has not been observed in batch 2. For batch 1, it can be due to a slower rate of tow displacement, which gives the ®bre enough time to be raised to a highest temperature while going through the reactor hottest area (1250C). This thermal treatment is responsible for the ®bre crystallisation that starts from its surface, causes local embrittlement and results in a large debonding. SAD observations of interphase (Fig. 12, top) show that the ®rst BN sublayer (located near the ®bre) is identical with that of batch 2: it is an anisotropic turbostratic BN. On the contrary, the second sublayer (located near the matrix) is made of an isotropic BN, 002 diraction arcs meet together and form continuous rings; there is no more preferential orientation. Here again, this dierence with batch 2 can be explained by the slower rate of tow displacement. The gaseous phase in®ltrated in the tow porosity is drag o more slowly to the hottest areas, which allows it to be raised to a highest temperature. Homogeneous nucleation can be then favoured to the detriment of heterogeneous chemical surface reactions. This phenomenon associated with a high yield of homogeneous reaction, results in an important production of HF and intermediate species and consequently leads to a material disorganisation. 3.2.3. Batch 3 minicomposite Batch 3 interphase is also made of two sublayers (Fig. 13): a ®rst preponderant one located near the ®bre identical with those of the other batches, about 250 nm in thickness, and a very thin one located near the matrix (about 20 nm in thickness) where BN organisation is very pronounced. The faster tow rate (3 m/h) can explain this disproportion. The ®bre resident time in the hottest area is reduced; the second organised sublayer has no more time to grow as in the case of batch 2. But, contrary to batch 2, this restricted growth allows this sublayer to keep a constant thickness and therefore to form a regular and linear interface with the matrix. Fig. 10. HR-TEM image of the middle of the interphase of a batch 2 minicomposite. Fig. 11. Bright-®eld TEM image of the interphase of a batch 1 minicomposite. 1934 S. Jacques et al. / Journal of the European Ceramic Society 20 (2000) 1929±1938
S Jacques et al. /Journal of the European Ceramic Society 20(2000)1929-1938 ” SAD SAD SAD SAD fibre 300nm fibre SAD SAD Fig. 12. Bright-field TEM image of the interfacial zone of a batch I minicomposite and corresponding SAD patterns(negatives). Fig. 13 exhibits a deflection of a matrix crack: it occurs close to the matrix within this thin orientated ig. 13. Bright-field TEM image of the interfacial zone of a batch 2 te and corresponding SAd patterns(negatives) sublayer. Because of interface regularity, the crack can propagates along the matrix in mode Il over very long distances. This phenomenon explains SEM observations Table 3 where the whole bn coating was found to remain fixed Average lifetimes, at 700C, under a 9.5 kg load, of the mini- on the fibre with a smooth and regular surface composites Batch Lifetime(h) 3.3. Lifetime at high temperature under tensile loading Dry air Moist air The average values of the lifetimes at 700oC in dry moist air under a 9.5 kg load are reported in Table 3 As a general rule, in these conditions, the mini- composite lifetimes depend on the ability of the inter- >400 phase to protect the fibre from a mechanism of stress corrosion. Indeed. Bertrand et al. 5 have evidenced thanks to similar tests performed on as-received bare For all the minicomposites with BN interphases, the fibre tows, that the Hi-Nicalon fibre lifetime becomes lifetimes are significantly improved which does prove extremely short(below one minute for a tow submitted the advantage of bn with respect to pyrocarbon both in to a load of 4.5 kg! ). Within a cracked matrix of a dry air and moist air. Among the three batches, batch 3 composite, if the interphase is oxidised the fibre is allows to achieve the best results. In dry air the experiment directly exposed to air and breaks, which results in the had to be stop after 400 h of exposure, which evidences failure of the whole composite the very efficient protection
Fig. 13 exhibits a de¯ection of a matrix crack: it occurs close to the matrix within this thin orientated sublayer. Because of interface regularity, the crack can propagates along the matrix in mode II over very long distances. This phenomenon explains SEM observations where the whole BN coating was found to remain ®xed on the ®bre with a smooth and regular surface. 3.3. Lifetime at high temperature under tensile loading The average values of the lifetimes at 700C in dry or moist air under a 9.5 kg load are reported in Table 3. As a general rule, in these conditions, the minicomposite lifetimes depend on the ability of the interphase to protect the ®bre from a mechanism of stress corrosion. Indeed, Bertrand et al.15 have evidenced, thanks to similar tests performed on as-received bare ®bre tows, that the Hi-Nicalon ®bre lifetime becomes extremely short (below one minute for a tow submitted to a load of 4.5 kg!). Within a cracked matrix of a composite, if the interphase is oxidised the ®bre is directly exposed to air and breaks, which results in the failure of the whole composite. For all the minicomposites with BN interphases, the lifetimes are signi®cantly improved which does prove the advantage of BN with respect to pyrocarbon both in dry air and moist air. Among the three batches, batch 3 allows to achieve the best results. In dry air the experiment had to be stop after 400 h of exposure, which evidences the very ecient protection. Fig. 13. Bright-®eld TEM image of the interfacial zone of a batch 2 minicomposite and corresponding SAD patterns (negatives). Fig. 12. Bright-®eld TEM image of the interfacial zone of a batch 1 minicomposite and corresponding SAD patterns (negatives). Table 3 Average lifetimes, at 700C, under a 9.5 kg load, of the minicomposites Batch Lifetime (h) Dry air Moist air 0 (PyC) 3 5 1 47 7 2 89 62 3 >400 187 S. Jacques et al. / Journal of the European Ceramic Society 20 (2000) 1929±1938 1935
S Jacques et al. /Journal of the European Ceramic Society 20(2000)1929-1938 to the pivot point of 1300C)limits the crystallisation to the fibre surface. This local inhomogeneity involves a 4.1. Relation between structures of fibre/matrix interfaces local embrittlement instead of global reinforcement and id minicomposite mechanical behaviour so an important fibre pull-out. The fibre/matrix bonding becomes very weak (low t values)and the load transfer From the above results, the processing parameter from the matrix to the fibre is no more ensured appears to play a very important role in the structure of A schematic representation, which summarises the the resulting BN interphase and so in the control of the different structures and the fracture behaviour of the BN fibre/matrix bonding. Indeed, the resident time of the interphases, is given in Fig. 14 tow inside the hot areas and the gap between the actual fibre temperature and the susceptor temperature depends 4. 2. Lifetime on r. Thus, with a high rate(r>2 m/h), the fibre tem- perature does not have time to reach the susceptor tem- As expected, the lifetime of the reference batch 0 with perature and does not follow exactly the temperature a pyrocarbon interphase is not very dependent on the profile presented in Fig. 2 moisture content and is the shortest despite an applied Concerning batch 2, the bn coating located close to the matrix is made of very well organised BN domains a slow enough rate of displacement tow (r=2.5 m/h) allows the large extension of the domains. The result is a rough interphase surface and an irregular BN/matrix interface and consequently a good bonding of the matrix During deflections of matrix cracks from mode I to mode Il occurring in the course of unloading-reloading cycles, intense frictions(cf. t values) between two rough sur- faces restrict the sliding. These frictions and tearing result in a SEM observation of a tortured interphase surface fixed on pulled out fibres. These features explain that, among the three bn interphases, interphase 2 confers the best mechanical properties. When r increases up to 3 m/h(batch 3), the wel matrix organised sublayer does not have time to thicken remains thin and regular. Hence, debonding and sliding can occur over long distances without any obstacles. As a matter of fact, the surface coating of the pulled out fibres appears smooth on SEM examinations. The interfacial shear stress is weak and so are the mechanical properties B Previous studies on CMCs with a PyC interphase1617 ive already reported the importance of tailoring the interphase thickness and orientation. But, whereas a thick well-organised pyrocarbon results in a too weak Interphase 2 fibre/matrix bonding, here a sufficient highly crystallised Bn thickness allows, on the contrary, to limit the sliding For slowest rates of displacement tow (r=2 m/h) batch 1), the second bn sublayer located near the matrix becomes isotropic because of a more important homogeneous nucleation; but above all, the fibre surface starts to crystallise by thermal treatment. This feature is rising for the Hi-Nicalon fibre is ex →, thermodynamically stable owing to its poor content of oxygen 18-20 Nevertheless, a recent study 2 performed on heat-treated Hi-Nicalon fibres has shown that a growth of the B-SiC crystallite size occurs in the bulk fibre crystallised 450 nm from and above 1300 c resulting in a mechanical interphase 1 enhancement of the fibre. In the present case, the part Fig. 14. Schematic representation of the BN interphases structures cular treatment (rather short time at a temperature close and the matrix cracks deflection
4. Discussion 4.1. Relation between structures of ®bre/matrix interfaces and minicomposite mechanical behaviour From the above results, the processing parameter r appears to play a very important role in the structure of the resulting BN interphase and so in the control of the ®bre/matrix bonding. Indeed, the resident time of the tow inside the hot areas and the gap between the actual ®bre temperature and the susceptor temperature depends on r. Thus, with a high rate (r>2 m/h), the ®bre temperature does not have time to reach the susceptor temperature and does not follow exactly the temperature pro®le presented in Fig. 2. Concerning batch 2, the BN coating located close to the matrix is made of very well organised BN domains. A slow enough rate of displacement tow (r=2.5 m/h) allows the large extension of the domains. The result is a rough interphase surface and an irregular BN/matrix interface and consequently a good bonding of the matrix. During de¯ections of matrix cracks from mode I to mode II occurring in the course of unloading-reloading cycles, intense frictions (cf. values) between two rough surfaces restrict the sliding. These frictions and tearing result in a SEM observation of a tortured interphase surface ®xed on pulled out ®bres. These features explain that, among the three BN interphases, interphase 2 confers the best mechanical properties. When r increases up to 3 m/h (batch 3), the wellorganised sublayer does not have time to thicken; it remains thin and regular. Hence, debonding and sliding can occur over long distances without any obstacles. As a matter of fact, the surface coating of the pulled out ®bres appears smooth on SEM examinations. The interfacial shear stress is weak and so are the mechanical properties. Previous studies on CMCs with a PyC interphase16,17 have already reported the importance of tailoring the interphase thickness and orientation. But, whereas a thick well-organised pyrocarbon results in a too weak ®bre/matrix bonding, here a sucient highly crystallised BN thickness allows, on the contrary, to limit the sliding. For slowest rates of displacement tow (r=2 m/h) (batch 1), the second BN sublayer located near the matrix becomes isotropic because of a more important homogeneous nucleation; but above all, the ®bre surface starts to crystallise by thermal treatment. This feature is surprising for the Hi-Nicalon ®bre is expected to be thermodynamically stable owing to its poor content of oxygen.18ÿ20 Nevertheless, a recent study12 performed on heat-treated Hi-Nicalon ®bres has shown that a growth of the b-SiC crystallite size occurs in the bulk from and above 1300C resulting in a mechanical enhancement of the ®bre. In the present case, the particular treatment (rather short time at a temperature close to the pivot point of 1300C) limits the crystallisation to the ®bre surface. This local inhomogeneity involves a local embrittlement instead of global reinforcement and so an important ®bre pull-out. The ®bre/matrix bonding becomes very weak (low values) and the load transfer from the matrix to the ®bre is no more ensured. A schematic representation, which summarises the dierent structures and the fracture behaviour of the BN interphases, is given in Fig. 14. 4.2. Lifetime As expected, the lifetime of the reference batch 0 with a pyrocarbon interphase is not very dependent on the moisture content and is the shortest despite an applied Fig. 14. Schematic representation of the BN interphases structures and the matrix cracks de¯ection. 1936 S. Jacques et al. / Journal of the European Ceramic Society 20 (2000) 1929±1938
S Jacques et al. Journal of the European Ceramic Society 20(2000)1929-1938 force(93 N) below the apparent proportional limit (111 deflect the matrix cracks. The key role of the tow tra- N). In these conditions, the composite failure results velling rate has been confirmed. Indeed, the variation of from a rapid oxidation of the Pyc interphase by air the substrate temperature and the infiltrated gaseous oxygen into CO(g)and/or CO2(g).21 phase depends upon this rate. As evidenced by tEM Concerning the other batches, the protection of the the thickness, the textures, the homogeneity and the N interphase and consequently of the minicomposite morphology of the different sublayers that constitute fibres toward oxidation is due to the formation of a the whole final interphase hence depend on it, and, as a B2O3 layer. This glass acts as a physical barrier against mater of fact, the global composite properties. Further oxygen. But, this protection depends on the moisture more, a too slow passing through raises the fibre to a content. Indeed, water reacts with B2O, to give volatile too high temperature, the fibre surface undergoes a lo boric acids HBO2(g), H3BO3 (g), and H3,O6(g), which crystallisation which results in an intense debonding decreases the protecting effect of the oxide layer Finally, thermal ageing tests in dry or moist air under For batch 3, the matrix cracks are deflected close to a static loading have confirmed that the BN interphases at about 250 nm(Fig 14, top). Although this behaviour compared with a classical pyrocarbon ntepnomposites ne interphase/matrix interface i.e. far"from the fibre significantly improve the lifetimes of the minico is not favourable to the mechanical properties; it allows in return, the fibre to be protected from air. For batch 2, the interphase/matrix bonding involves important tear- References ng. Finally, cracks allows air to come nearer to the fibre (middle of Fig. 14)than in the case of batch 3, which 1. Evans, A. G and Zok, F w, The physics and mechanics of decreases the lifetime. as for batch 1. the weak link is fibre-reinforced brittle matrix composites. J. Mater. Sci., 1994, ocated in the fibre surface(Fig 14, bottom); the fibre is 29,3857-3896 of course quickly exposed; which results in an even 2. Rebillat, F, Guette, A, Spit shorter lifetime Naslain, R, Oxidation resistance of SiC/SiC minicomposites with a highly crystallised BN interphase. J. Eur. Ceram Soc., 1998, 18, At this point, it is not possible to determine the best Ninterphase. Indeed, even if the oxidation resistance is 3. Rebillat, F,Guette, A.and Brosse, C.R,Chemical and considerably improved compared with the reference, the echanical alterations of Sic Nicalon fiber properties during the best minicomposites at room temperature(batch 2)de CVD/CVI process for boron nitride. Acta Mater. 1999. 47(5) not result in the best lifetimes at high temperature( those of batch 3). Hot dynamic cycle fatigue tests would allow Bouix. J. Continuous elaboration of boron nitride on SiC (Hi- to settle because the mechanical behaviour would play an Nicalon) fiber. Textures of deposited BN and properties of coated even more important part than in static test fibers. In Proceedings of the IIth Journees Nationales sur les opposites, J. Lamon, and D. Baptiste, D. 18-20 Novemb Arcachon, France, 1998, pp 393-404 5. Lopez-Marure, A, La Realisation et le Comportement dinter. 5. Conclusion phases BN a gradient de Proprietes, Application a des Composites Ceramiques. PhD thesis, University of Claude Bernard Lyon 1 ith a structural gradient were France. 1999 pared by using a one-step continuous dynamic CVI 6. Olivier, C, Elaboration et Etude du comportement mecanique de process within 1-D SiC/SiC minicomposites. The fibre tow travelled through a reactor that contains different Institut National des Sciences Appliquees of Lyon, France, 1998 7. Bertrand, S, Forio, P, Pailler, R and Lamon, J, Hi-Nicalon / SiC hot areas(TG-Cvi process). The infiltration occurred minicomposites with(Pyrocarbon/SiC)n Nanoscale multilayered while keeping the same conditions of gaseous phase nterphases. J. Am. Ceram Soc., 1999. 82(9), 2465-2476 (pressure, flow rates, composition) during the whole 8. Lamon, J, Rebillat, F. and Evans, A G, Microcomposite test experiment. A judicious choice of travelling rate com- procedure for evaluating the interface properties of ceramic bined with a well adapted temperature profile in the matrix composites. J. Am. Ceram. Soc., 1995, 78(2), 401-405. 9. Marshall. D. B. and Evans. A. G. Failure Mechanisms in Cera. sceptor allowed to obtain minicomposites with good Fiber/Ceramic-Matrix Composites. J. Am. Cera. Soc properties 1985.68(5),225-231 Indeed, tensile tests at room temperature have shown 10. Aveston, J, Cooper, G.A. and Kelly, A. Single and multiple that the BN interphases act as mechanical fuses and that fracture. In Proceedings of the Conference of the National Physical Laboratory, the Properties of Fiber Composites, IPC Science and the fibres are not chemically degraded. These good results are obtained thanks to a fibre coating starting in a 11. Lebrun, G.A. and Lamon, J, Influence de la distribution spatiale lowtemperature area. This non-aggressive treatment des fibres au sein d un composite sur le comportement mecanique conditions protect the fibre. Then, while going through a en traction. In Annales des Composites, 1995/4. AMAC, 1995, pp hottest area, the structural anisotropy of the bn infil 121-130. 12. Bertrand, S, Lamon, J, Pailler, R. and Goujard, S, Effect of trated in the tow can increase without fibre chemical eat-treatments on the microstructure and the properties of Hi- attack. This organised bn allows the interphase to Nicalon fibres. J. Eur. Ceram. Soc., submitted for publication
force (93 N) below the apparent proportional limit (111 N). In these conditions, the composite failure results from a rapid oxidation of the PyC interphase by air oxygen into CO(g) and/or CO2(g).21 Concerning the other batches, the protection of the BN interphase and consequently of the minicomposite ®bres toward oxidation is due to the formation of a B2O3 layer. This glass acts as a physical barrier against oxygen. But, this protection depends on the moisture content. Indeed, water reacts with B2O3 to give volatile boric acids HBO2(g), H3BO3(g), and H3B3O6(g), which decreases the protecting eect of the oxide layer.22 For batch 3, the matrix cracks are de¯ected close to the interphase/matrix interface i.e. ``far'' from the ®bre at about 250 nm (Fig. 14, top). Although this behaviour is not favourable to the mechanical properties; it allows, in return, the ®bre to be protected from air. For batch 2, the interphase/matrix bonding involves important tearing. Finally, cracks allows air to come nearer to the ®bre (middle of Fig. 14) than in the case of batch 3, which decreases the lifetime. As for batch 1, the weak link is located in the ®bre surface (Fig. 14, bottom); the ®bre is of course quickly exposed; which results in an even shorter lifetime. At this point, it is not possible to determine the best BN interphase. Indeed, even if the oxidation resistance is considerably improved compared with the reference, the best minicomposites at room temperature (batch 2) do not result in the best lifetimes at high temperature (those of batch 3). Hot dynamic cycle fatigue tests would allow to settle because the mechanical behaviour would play an even more important part than in static tests. 5. Conclusion BN interphases with a structural gradient were prepared by using a one-step continuous dynamic CVI process within 1-D SiC/SiC minicomposites. The ®bre tow travelled through a reactor that contains dierent hot areas (TG-CVI process). The in®ltration occurred while keeping the same conditions of gaseous phase (pressure, ¯ow rates, composition) during the whole experiment. A judicious choice of travelling rate combined with a well adapted temperature pro®le in the susceptor allowed to obtain minicomposites with good properties. Indeed, tensile tests at room temperature have shown that the BN interphases act as mechanical fuses and that the ®bres are not chemically degraded. These good results are obtained thanks to a ®bre coating starting in a ``low'' temperature area. This non-aggressive treatment conditions protect the ®bre. Then, while going through a hottest area, the structural anisotropy of the BN in®ltrated in the tow can increase without ®bre chemical attack. This organised BN allows the interphase to de¯ect the matrix cracks. The key role of the tow travelling rate has been con®rmed. Indeed, the variation of the substrate temperature and the in®ltrated gaseous phase depends upon this rate. As evidenced by TEM, the thickness, the textures, the homogeneity and the morphology of the dierent sublayers that constitute the whole ®nal interphase hence depend on it, and, as a mater of fact, the global composite properties. Furthermore, a too slow passing through raises the ®bre to a too high temperature, the ®bre surface undergoes a local crystallisation which results in an intense debonding. Finally, thermal ageing tests in dry or moist air under a static loading have con®rmed that the BN interphases signi®cantly improve the lifetimes of the minicomposites compared with a classical pyrocarbon interphase. References 1. Evans, A. G. and Zok, F. W., The physics and mechanics of ®bre-reinforced brittle matrix composites. J. Mater. Sci., 1994, 29, 3857±3896. 2. Rebillat, F., Guette, A., Espitalier, L., Debieuvre, C. and Naslain, R., Oxidation resistance of SiC/SiC minicomposites with a highly crystallised BN interphase. J. Eur. Ceram Soc., 1998, 18, 1809±1819. 3. Rebillat, F., Guette, A. and Brosse, C. R., Chemical and mechanical alterations of SiC Nicalon ®ber properties during the CVD/CVI process for boron nitride. Acta Mater., 1999, 47(5), 1685±1696. 4. Vincent, H., Lopez-Marure, A., Lamouroux, F., Vincent, C., and Bouix, J., Continuous elaboration of boron nitride on SiC (HiNicalon) ®ber. Textures of deposited BN and properties of coated ®bers. In Proceedings of the 11th JourneÂes Nationales sur les composites, J. Lamon, and D. Baptiste, D. 18±20 November, Arcachon, France, 1998, pp. 393±404. 5. Lopez-Marure, A., La ReÂalisation et le Comportement d'Interphases BN aÁ Gradient de ProprieÂteÂs, Application aÁ des Composites CeÂramiques. PhD thesis, University of Claude Bernard Lyon 1, France, 1999. 6. Olivier, C., Elaboration et Etude du Comportement MeÂcanique de Composites Unidirectionnels C/Si3N4 et SiC/Si3N4. PhD thesis, Institut National des Sciences AppliqueÂes of Lyon, France, 1998. 7. Bertrand, S., Forio, P., Pailler, R. and Lamon, J., Hi-Nicalon/SiC minicomposites with (Pyrocarbon/SiC)n Nanoscale multilayered interphases. J. Am. Ceram. Soc., 1999, 82(9), 2465±2476. 8. Lamon, J., Rebillat, F. and Evans, A. G., Microcomposite test procedure for evaluating the interface properties of ceramic matrix composites. J. Am. Ceram. Soc., 1995, 78(2), 401±405. 9. Marshall, D. B. and Evans, A. G., Failure Mechanisms in Ceramic-Fiber/Ceramic-Matrix Composites. J. Am. Ceram. Soc., 1985, 68(5), 225±231. 10. Aveston, J., Cooper, G.A. and Kelly, A., Single and multiple fracture. In Proceedings of the Conference of the National Physical Laboratory, the Properties of Fiber Composites, IPC Science and Technology Press Ltd, Surrey, UK, 1971, pp. 15±26. 11. Lebrun, G.A. and Lamon, J., In¯uence de la distribution spatiale des ®bres au sein d'un composite sur le comportement meÂcanique en traction. In Annales des Composites, 1995/4. AMAC, 1995, pp. 121±130. 12. Bertrand, S., Lamon, J., Pailler, R. and Goujard, S., Eect of heat-treatments on the microstructure and the properties of HiNicalon ®bres. J. Eur. Ceram. Soc., submitted for publication. S. Jacques et al. / Journal of the European Ceramic Society 20 (2000) 1929±1938 1937
1938 S Jacques et al. Journal of the European Ceramic Society 20(2000)1929-1938 13. Bansal. N. P. Effects of hf treatments on tensile strength of Hi- Proceedings of the 17th Annual Conference on Composites and Nicalon fibres. J Mater. Sci., 1998. 33(17). 4287-429 Advanced Ceramic Materials. in Ceram. Eng. Sci. Proc.,ed 14. Bansal. N. P. and Chen. Y. L. Chemical. mechanica D.C. Cranner. 10-15 January, Cocoa Beach. American Ceramic microstructural characterization of low-oxygen containing Society,1993,pp.540-547 carbide fibers with ceramic coatings. J. Mater. Sci.. 1998 19. Shimoo. T. Tsukada. I. Narisawa. M. Seguchi, T and Oka. 5277-5289 ura.K, Change in properties of polycarbosilane-derived Sic 15. Bertrand, S. Lamon. J. and Pailler, R. Oxidation resistance in fibers at high temperatures. J. Ceram. Soc. of Japan, 1997. 105(7). static fatigue conditions of Hi-Nicalon fibre tows and SiC/(Pyc 559-563 Sic),/SiC minicomposites. J. Eur. C 20. Hollon. G. Pailler. R. Naslain, R. Laanani. F. Monthioux publication M. and Olry. P, Thermal stability of a PCS-derived Sic fiber 16. Nguyen Van Sang-Trouvat, B, Analyse et Optimisation des with a low oxygen content(Hi-Nicalon).J Mater. Sci. 1997. 32, Interfaces dans les Composites Carbone/Carbone a Ren 327-347 Fibreux. PhD thesis no. 1635. University of Bordeaux I, france. 21. Fillipuzi, L. Camus. G. Naslain, R and Thebault, J, Oxidation 1996. mechanisms and kinetics of ID-SiC/C/SiC composite materials 17. Olivier. C. Veyret. J. B. and Vidal Setif. M. H. Mechanical an experimental approach. J. A. Ceram. Soc., 1994. 77(2) sites. Key Engineering Materials, 1997, 127-131(2), 753-760. 18. Takeda, M.. Imai, Y. Ichikawa, H, Ichikawa. T, Kasai R. E, High-temperature oxidation of boron nitride: Il, boron guchi T and Okamura, K. Thermomechanical analysis of the nitride layers in composites. J. Am. Ceram. Soc., 1999, 82(6 low oxygen silicon carbide fibers derived from polycarbosilane. In 1473-1482
13. Bansal, N. P., Eects of HF treatments on tensile strength of HiNicalon ®bres. J. Mater. Sci., 1998, 33(17), 4287±4295. 14. Bansal, N. P. and Chen, Y. L., Chemical, mechanical and microstructural characterization of low-oxygen containing silicon carbide ®bers with ceramic coatings. J. Mater. Sci., 1998, 33(22), 5277±5289. 15. Bertrand, S., Lamon, J. and Pailler, R., Oxidation resistance in static fatigue conditions of Hi-Nicalon ®bre tows and SiC/(PyC± SiC)n/SiC minicomposites. J. Eur. Ceram. Soc., submitted for publication. 16. Nguyen Van Sang-Trouvat, B., Analyse et Optimisation des Interfaces dans les Composites Carbone/Carbone aÁ Renfort Fibreux. PhD thesis no. 1635, University of Bordeaux I, France, 1996. 17. Olivier, C., Veyret, J. B. and Vidal Setif, M. H., Mechanical properties of Hi-Nicalon ®bre-reinforced silicon nitride composites. Key Engineering Materials, 1997, 127±131(2), 753±760. 18. Takeda, M., Imai, Y., Ichikawa, H., Ichikawa, T., Kasai, N., Seguchi T. and Okamura, K., Thermomechanical analysis of the low oxygen silicon carbide ®bers derived from polycarbosilane. In Proceedings of the 17th Annual Conference on Composites and Advanced Ceramic Materials, in Ceram. Eng. & Sci. Proc., ed. D.C. Cranner. 10±15 January, Cocoa Beach. American Ceramic Society, 1993, pp. 540-547. 19. Shimoo, T., Tsukada, I., Narisawa, M., Seguchi, T. and Okamura, K., Change in properties of polycarbosilane-derived SiC ®bers at high temperatures. J. Ceram. Soc. of Japan, 1997, 105(7), 559±563. 20. Chollon, G., Pailler, R., Naslain, R., Laanani, F., Monthioux, M. and Olry, P., Thermal stability of a PCS-derived SiC ®ber with a low oxygen content (Hi-Nicalon). J. Mater. Sci., 1997, 32, 327±347. 21. Fillipuzi, L., Camus, G., Naslain, R. and Thebault, J., Oxidation mechanisms and kinetics of 1D-SiC/C/SiC composite materials: 1 Ð an experimental approach. J. Am. Ceram. Soc., 1994, 77(2), 459±466. 22. Jacobson, N. S., Morscher, G. N., Bryant, D. R. and Tressler, R. E., High-temperature oxidation of boron nitride: II, boron nitride layers in composites. J. Am. Ceram. Soc., 1999, 82(6), 1473±1482. 1938 S. Jacques et al. / Journal of the European Ceramic Society 20 (2000) 1929±1938