and manufacturing ELSEVIER Composites: Part A 30(1999)537-547 Micro/minicomposites: a useful approach to the design and development of non-oxide cmcs Roger Naslain", Jacques Lamon, Rene Pailler, Xavier Bourrat, Alain Guette, Francis Langlais aboratory for Thermostructural Composites, UMR-47(CNRS-SEP-UB1), University of bordeaux. 3 Allee de la boetie, 33600 Pessac, france Micro(one single filament)and mini (one single fiber tow) non-oxide composites(C/C; C/SiC and SiC/SiC)with simple(Pyc or BN)or complex interphases [C (B)or(Pyc-SiC) multilayers] are fabricated in a short time by CVD/CVI. The fiber/matrix interfacial zone is characterized by aEs and TEM. Tensile tests are used to assess the mechanical properties and the weibull statistical parameters of both the fiber and matrix, as well as the fiber-matrix interfacial parameters(Ti Id, Gis). The tensile stress-strain behaviour has been modelled. The tensile curves exhibit the same features as those previously reported for real nD-composites. Lifetime at high temperatures in air is characterized through static/cyclic fatigue tests and modelled. It is improved by replacing conventional pyrocarbon by highly engineered interphases. The micro/mini composite approach is used in the optimization of processing conditions and to derive parameters necessary fo the modelling of the thermomechanical and chemical behaviour of composites with more complex fiber architectures. o 1999 Elsevier Science ltd. All rights reserved Keywords: A Ceramic matrix composites(CMCs); Model composites; B. Interface/interphase 1. Introduction CMC, which requires several processing/characterization be tir Non-oxide ceramic matrix composites(CMCs), such as fabrication technique. Moreover, the complexity of real C/C, C/SiC or SiC/SiC composites, usually exhibit a fiber architectures often precludes the derivation of simple complex fiber architecture(2D, 2.5D or 3D). They are correlations between composite properties and processing produced according to liquid or gas phase routes requiring conditions relatively long processing times. In the liquid phase routes, within the scope of the design and development of new the starting fibrous material is impregnated with a liquid materials it can be more appropriate to use ID model precursor of the matrix, e.g. a slurry or an organic/organo- composites, such as the microcomposites or the minicom metallic polymer, and pyrolysed at high temperature. The posites(comprising one single fiber or one single tow, impregnation/pyrolysis sequence is repeated several times respectively), in order to conduct several processing/char- in order to achieve a high densification level. In the ga acterization iterative loops in a relatively short time phase routes, such as the isothermal/isobaric chemical Furthermore, for such very simple fiber architectures vapor infiltration process(I-CVI), a porous fiber preform micromechanics-based models exist which can be used to is infiltrated with the matrix deposited from a gaseous derive useful material parameters, such as load transfer precursor(a hydrocarbon for carbon and a chlorosilane for parameters, from simple mechanical tests [5,6]. Examples SiC). The deposition process should be conducted at low of studies conducted via the use of the micro/mini compo- temperature and low pressure, in order to avoid an early site approach, have been already reported in the field of non- sealing of the pore entrances. The densification duration oxide materials, however it is not by far a generalised way can be relatively long, depending on the size of the preform for designing CMCs and for optimising their processing and the residual porosity [1]. Although accelerated CVI- conditions [7-191 processes have been proposed, the densification duration Micro/mini model composites have been used during is still of several tens of hours [2-4]. Thus, the optimizatio almost one decade at LCts to optimize the fiber-matrix of the composition and the processing conditions for a given ( FM)interfacial zone in SiC/SiC composites and to generate micromechanical data necessary for modelling the mechan- en四+3006 ical behaviour. More recently, the approach has been tended to cc co 1359-835X/99/S- see front matter @1999 Elsevier Science Ltd. All rights reserved P:S1359-835X(98)00147-X
Micro/minicomposites: a useful approach to the design and development of non-oxide CMCs Roger Naslain*, Jacques Lamon, Rene´ Pailler, Xavier Bourrat, Alain Guette, Francis Langlais Laboratory for Thermostructural Composites, UMR-47 (CNRS-SEP-UB1), University of Bordeaux, 3 Alle´e de La Boe¨tie, 33600 Pessac, France Abstract Micro (one single filament) and mini (one single fiber tow) non-oxide composites (C/C; C/SiC and SiC/SiC) with simple (PyC or BN) or complex interphases [C (B) or (PyC-SiC)n multilayers] are fabricated in a short time by CVD/CVI. The fiber/matrix interfacial zone is characterized by AES and TEM. Tensile tests are used to assess the mechanical properties and the Weibull statistical parameters of both the fiber and matrix, as well as the fiber–matrix interfacial parameters (ti; ld; Gic). The tensile stress–strain behaviour has been modelled. The tensile curves exhibit the same features as those previously reported for real nD-composites. Lifetime at high temperatures in air is characterized through static/cyclic fatigue tests and modelled. It is improved by replacing conventional pyrocarbon by highly engineered interphases. The micro/mini composite approach is used in the optimization of processing conditions and to derive parameters necessary for the modelling of the thermomechanical and chemical behaviour of composites with more complex fiber architectures. q 1999 Elsevier Science Ltd. All rights reserved. Keywords: A. Ceramic matrix composites (CMCs); Model composites; B. Interface/interphase 1. Introduction Non-oxide ceramic matrix composites (CMCs), such as C/C, C/SiC or SiC/SiC composites, usually exhibit a complex fiber architecture (2D, 2.5D or 3D). They are produced according to liquid or gas phase routes requiring relatively long processing times. In the liquid phase routes, the starting fibrous material is impregnated with a liquid precursor of the matrix, e.g. a slurry or an organic/organometallic polymer, and pyrolysed at high temperature. The impregnation/pyrolysis sequence is repeated several times in order to achieve a high densification level. In the gas phase routes, such as the isothermal/isobaric chemical vapor infiltration process (I-CVI), a porous fiber preform is infiltrated with the matrix deposited from a gaseous precursor (a hydrocarbon for carbon and a chlorosilane for SiC). The deposition process should be conducted at low temperature and low pressure, in order to avoid an early sealing of the pore entrances. The densification duration can be relatively long, depending on the size of the preform and the residual porosity [1]. Although accelerated CVIprocesses have been proposed, the densification duration is still of several tens of hours [2–4]. Thus, the optimization of the composition and the processing conditions for a given CMC, which requires several processing/characterization loops, can be time consuming whatever the nature of the fabrication technique. Moreover, the complexity of real fiber architectures often precludes the derivation of simple correlations between composite properties and processing conditions. Within the scope of the design and development of new materials it can be more appropriate to use 1D model composites, such as the microcomposites or the minicomposites (comprising one single fiber or one single tow, respectively), in order to conduct several processing/characterization iterative loops in a relatively short time. Furthermore, for such very simple fiber architectures, micromechanics-based models exist which can be used to derive useful material parameters, such as load transfer parameters, from simple mechanical tests [5,6]. Examples of studies conducted via the use of the micro/mini composite approach, have been already reported in the field of nonoxide materials, however it is not by far a generalised way for designing CMCs and for optimising their processing conditions [7–19]. Micro/mini model composites have been used during almost one decade at LCTS to optimize the fiber–matrix (FM) interfacial zone in SiC/SiC composites and to generate micromechanical data necessary for modelling the mechanical behaviour. More recently, the approach has been extended to C/C composites [20]. The aim of the present Composites: Part A 30 (1999) 537–547 1359-835X/99/$ - see front matter q 1999 Elsevier Science Ltd. All rights reserved. PII: S1359-835X(98)00147-X * Corresponding author. Tel.: 133-5-56844706; fax: 133-5-56841225; e-mail: admin@lcts.u-bordeaux.fr
R. Naslain et al/Composites: Part A 30(1999)537-547 Fig 1. Nicalon/SiC microcomposites:(a)samp ing CVD-processing,(b)morphology of a SiC/BN/SiC microcomposite failed under tensile loading at room temperature showing fiber pull-out. contribution is to show how such model composites can be prepared and characterized, on the one hand, and to give the limits of the approach, on the other hand 2. Experimental SiC/SiC micro-or minl-composites comprise one straight single filament or a tow, respectively. Most experiments were performed with Si-C-O ex-polycarbosilane Nicalon fibers or Si-C oxygen-free Hi-Nicalon fibers, both display ing 500 filaments per yarn (with a filament diameter of 14 um)and being manufactured by Nippon Carbon. Some experiments were also conducted with high strength ex- Pan carbon fibers (filament diameter of 7 um). Micro 4≥s and minicomposites were fabricated according to I-CVI or pressure pulsed CVI (P-CVI)techniques, which have been described in detail elsewhere [1, 21]. In the preparation of minicomposite, the starting material was a length of the as- received multifilament yarn, slightly twisted and maintained in straight configuration with a ceramic holder. In that of a microcomposite, a length of one single filament was care- acted from the yarn and mounted on a ceramic holder with a high temperature cement, as shown in Fi 1. In a second step, the interphase was deposited onto the fiber surface from suitable gaseous precursors including: (i) hydrocarbons, such as CH4, C3 H8, C3H, for anisotropic pyrocarbon interphases; ( ii) BF3-NH3 or BCly-NH3-H2 mixtures for the deposition of boron nitride; (iii) hydrocar- bons and CHa SiCl, /H for multilayered(Pyc-SiC)n inter- phases and (iv)C3Hg-BCl3-H2 mixtures for composition graded C(B)interphases. The interphase, with an overall 186R thickness of a few 100 nm, was usually deposited(micro- composites)or infiltrated (minicomposites) by P-CVD/P- CVI rather than by conventional I-CVI, inasmuch as the Fig. 2. C/C minicomposites:(a) failure surface; (b)multiple matrix micro- former allows a better control of the morphology and texture acking, according to Ref. [24] of the interphase. As a matter of fact, complex multilayered
contribution is to show how such model composites can be prepared and characterized, on the one hand, and to give the limits of the approach, on the other hand. 2. Experimental SiC/SiC micro- or mini-composites comprise one straight single filament or a tow, respectively. Most experiments were performed with Si-C-O ex-polycarbosilane Nicalon fibers or Si-C oxygen-free Hi-Nicalon fibers, both displaying 500 filaments per yarn (with a filament diameter of 14 mm) and being manufactured by Nippon Carbon. Some experiments were also conducted with high strength exPAN carbon fibers (filament diameter of 7 mm). Micro and minicomposites were fabricated according to I-CVI or pressure pulsed CVI (P-CVI) techniques, which have been described in detail elsewhere [1,21]. In the preparation of a minicomposite, the starting material was a length of the asreceived multifilament yarn, slightly twisted and maintained in straight configuration with a ceramic holder. In that of a microcomposite, a length of one single filament was carefully extracted from the yarn and mounted on a ceramic holder with a high temperature cement, as shown in Fig. 1. In a second step, the interphase was deposited onto the fiber surface from suitable gaseous precursors including: (i) hydrocarbons, such as CH4; C3H8; C3H6, for anisotropic pyrocarbon interphases; (ii) BF3-NH3 or BCl3-NH3-H2 mixtures for the deposition of boron nitride; (iii) hydrocarbons and CH3SiCl3/H2 for multilayered (PyC-SiC)n interphases and (iv) C3H8-BCl3-H2 mixtures for composition graded C (B) interphases. The interphase, with an overall thickness of a few 100 nm, was usually deposited (microcomposites) or infiltrated (minicomposites) by P-CVD/PCVI rather than by conventional I-CVI, inasmuch as the former allows a better control of the morphology and texture of the interphase. As a matter of fact, complex multilayered 538 R. Naslain et al. / Composites: Part A 30 (1999) 537–547 Fig. 1. Nicalon/SiC microcomposites: (a) sample holder used during CVD-processing, (b) morphology of a SiC/BN/SiC microcomposite failed under tensile loading at room temperature showing fiber pull-out. Fig. 2. C/C minicomposites: (a) failure surface; (b) multiple matrix microcracking, according to Ref. [24]
R. Naslain et al/Composites: Part A 30(1999)537-547 fixed epoXy mini SiC CVD calibrated 000 masses switches Fig. 3. Mechanical testing of micro/minicomposites: (a) tensile specimens; (b) tensile device for minicomposite tests; and (c) apparatus for lifetime urements in air at high temperatures of SiC/SiC microcomposites, according to Ref. [ 9, 12, 26] interphases with elementary layer thicknesses of a few nm Vr=0.50, corresponding to that commonly observed for the can only be deposited (infiltrated) in a controlled manner by corresponding real nD-composites). The chemical analysis P-CVD/P-CVI [22, 23]. In a last step, the Sic (or carbon) of the FM-interfacial zone was performed according to two was deposited (infiltrated) by conventional CVD complementary techniques: (i) by Auger electron spectro- (CVI)in the apparatus used for the fabrication of real nd scopy (AES)depth profiles, recorded from fracture composites and under similar T-P conditions(typically at surfaces both on debonded fibers [see Fig. I(b) and on 00'C-1000C and under few kPa or few 10 kPa, depend- their trough in the matrix, and (i by parallel electron ing on the nature of the precursor). Hence, micro/mini energy loss spectroscopy(P-EELS)in a transmission elec tron microscope(TEM). Longitudinal thin foils of micro- ques closely related to those used for the corresponding real composites were prepared according to a technique which composites, with however two important differences: (i)the has been depicted elsewhere [11] processing time is much shorter and; (ii)the handling of the Micro-and mini-composites were tensile tested at room specimens requires specific care. Examples of micro/mini- temperature with tensile devices built in-house and composites are shown in Figs. I and 2 described elsewhere [25-27]. Microcomposites were tested The characterization of micro/mini composites requires specific experimental procedures owing to the small radial size of the specimens (a Nicalon/SiC microcomposite has an I Auger PHI 5590, from Physical Electronics PEELS 666-3K, from Gatan (USA). overall diameter of about 20 um for a fiber volume fraction, CM 30 ST, from Philips (NL)
interphases with elementary layer thicknesses of a few nm can only be deposited (infiltrated) in a controlled manner by P-CVD/P-CVI [22,23]. In a last step, the SiC (or carbon) matrix was deposited (infiltrated) by conventional CVD (CVI) in the apparatus used for the fabrication of real nD composites and under similar T–P conditions (typically at 9008C–10008C and under few kPa or few 10 kPa, depending on the nature of the precursor). Hence, micro/mini composites are fabricated according to processing techniques closely related to those used for the corresponding real composites, with however two important differences: (i) the processing time is much shorter and; (ii) the handling of the specimens requires specific care. Examples of micro/minicomposites are shown in Figs. 1 and 2. The characterization of micro/mini composites requires specific experimental procedures owing to the small radial size of the specimens (a Nicalon/SiC microcomposite has an overall diameter of about 20 mm for a fiber volume fraction, Vf 0.50, corresponding to that commonly observed for the corresponding real nD-composites). The chemical analysis of the FM-interfacial zone was performed according to two complementary techniques: (i) by Auger electron spectroscopy (AES)1 depth profiles, recorded from fracture surfaces both on debonded fibers [see Fig. 1(b)] and on their trough in the matrix, and (ii) by parallel electron energy loss spectroscopy (P-EELS)2 in a transmission electron microscope (TEM)3 . Longitudinal thin foils of microcomposites were prepared according to a technique which has been depicted elsewhere [11]. Micro- and mini-composites were tensile tested at room temperature with tensile devices built in-house and described elsewhere [25–27]. Microcomposites were tested R. Naslain et al. / Composites: Part A 30 (1999) 537–547 539 Fig. 3. Mechanical testing of micro/minicomposites: (a) tensile specimens; (b) tensile device for minicomposite tests; and (c) apparatus for lifetime measurements in air at high temperatures of SiC/SiC microcomposites, according to Ref. [9, 12, 26]. 1 Auger PHI 5590, from Physical Electronics. 2 PEELS 666-3 K, from Gatan (USA). 3 CM 30 ST, from Philips (NL)
R. Naslain et al/Composites: Part A 30(1999)537-547 MATRI 「atB①P Sputter time (min M 500nm nd TEM analysis of complex interphases in SiC/SiC microcomposites:(a)C (B) graded com ase; (b)AES depth profile (B)graded composition type A interphase(a being similar to A but wi sublayer V),(c)bright field TEM of a Nicalon/SiC with a type A'C(B)interphase, and(d) bright field TEM image of a Hi-Nicalon/SiC microcomposite with a(Py C-SiC)o multilayered interphase, according to Refs. [11, 12, 15, 23] according to a procedure similar to that one used for single elsewhere [5,6]. The FM-interfacial parameters were also filaments. A length of microcomposite was pasted with an assessed from push-in or push-through experiments epoxy cement on a paper holder(gauge length, L, ranging performed on polished cross-sections [14. Finally, the fail from 10 to 50 mm), as shown in Fig 3(a). The whole assem- ure surfaces were observed with a high-resolution scanning bly was then attached to the tensile tester grips and finally, microscope(HR-sEM) to identify the failure mode and to the paper holder was cut immediately before applying the calculate the in-situ failure stress of the fibers from mirror load. In a similar manner, minicomposites were attached radius measurements with the epoxy cement to two steel tubes(gauge lengt he effect of the environment, e.g. the ambient air, on the ranging from 50 to 75 mm)and strained at a constant mechanical behaviour and lifetime, was studied through speed(0.085% per min), the displacement being measured fatigue tests(static or cyclic)performed at high tempera either optically or with inductive transducers [Fig. 3(b)]. tures on either micro-or mini-composites The minicompo- Unloading-reloading hysteresis loops were systematically site specimens were prepared as described above for tests recorded in order to measure the Youngs modulus, E, and performed at room temperature, the epoxy cement being the residual permanent strain, ep, (at o=0), of the material replaced by a high temperature alumina-based cement. In as it is progressively damaged and to derive the FM-inter- the static(or cyclic)fatigue tests run on microcomposites, a facial parameters, i.e. the debond length, Id, the interfacial length of microcomposite was attached with the alumina- shear stress, Ti, and the debond energy, Ti, according to micromechanics-based models, which have been reported S4500, from Hitachi (Japan)
according to a procedure similar to that one used for single filaments. A length of microcomposite was pasted with an epoxy cement on a paper holder (gauge length, L, ranging from 10 to 50 mm), as shown in Fig. 3(a). The whole assembly was then attached to the tensile tester grips and finally, the paper holder was cut immediately before applying the load. In a similar manner, minicomposites were attached with the epoxy cement to two steel tubes (gauge length ranging from 50 to 75 mm) and strained at a constant speed (0.085% per min), the displacement being measured either optically or with inductive transducers [Fig. 3(b)]. Unloading–reloading hysteresis loops were systematically recorded in order to measure the Young’s modulus, E, and the residual permanent strain, ep, (at s 0), of the material as it is progressively damaged and to derive the FM-interfacial parameters, i.e. the debond length, ld, the interfacial shear stress, ti, and the debond energy, Gi, according to micromechanics-based models, which have been reported elsewhere [5,6]. The FM-interfacial parameters were also assessed from push-in or push-through experiments performed on polished cross-sections [14]. Finally, the failure surfaces were observed with a high-resolution scanning microscope (HR-SEM)4 to identify the failure mode and to calculate the in-situ failure stress of the fibers from mirror radius measurements. The effect of the environment, e.g. the ambient air, on the mechanical behaviour and lifetime, was studied through fatigue tests (static or cyclic) performed at high temperatures on either micro- or mini-composites. The minicomposite specimens were prepared as described above for tests performed at room temperature, the epoxy cement being replaced by a high temperature alumina-based cement. In the static (or cyclic) fatigue tests run on microcomposites, a length of microcomposite was attached with the alumina- 540 R. Naslain et al. / Composites: Part A 30 (1999) 537–547 Fig. 4. AES and TEM analysis of complex interphases in SiC/SiC microcomposites: (a) C (B) graded composition type A 0 interphase; (b) AES depth profile analysis of a C (B) graded composition type A interphase (A being similar to A 0 but without sublayer V); (c) bright field TEM image of a Nicalon/SiC microcomposite with a type A 0 C (B) interphase; and (d) bright field TEM image of a Hi-Nicalon/SiC microcomposite with a (PyC-SiC)10 multilayered interphase, according to Refs. [11,12,15,23]. 4 S 4500, from Hitachi (Japan)
R. Naslain et al/Composites: Part A 30(1999)537-547 based cement to two 100 um SIC CVD filaments, acting elementary thicknesses are of the order of 100 nm as tensile rods, and the lifetime was measured with a timer as expected; and (ii) a matrix mode-I crack, which has been shown in Fig 3(c) deflected in mode Il(parallel to the fiber surface) near th boundary between sublayers Il and ll [11, 12]. The analys 3. Results and discussion of complex multilayered interphases becomes more difficult when the thickness of the elementary sublayer is extremely The objective of the present contribution being to show small, typically a few nm. Under such conditions, the AES microprobe resolution may be insuficient and the TEM/P- the potential and limits of the micro/mini composites EEls analysis the only appropriate technique. As an exam- approach, rather than to give a detailed analysis of the mate- als, a few examples will be presented and discussed to ple, Fig. 4(d)shows a TEM-image recorded from a lon illustrate how this approach can be used to assess the struc. itudinal thin foil of a Hi-Nicalon/(Py C-SiC)Sic tural and mechanical behaviour of selected composites as microcomposite failed under tensile loading. First, the 10 well as the effect of an oxidizing atmosphere PyC-SiC sequences deposited by P-CVD(each comprising a 20 nm thick pyrocarbon and a 30 nm thick SiC +Cnano 3.1. Chemical and structural analysis crystalline sublayers)are clearly apparent. Second, the TEM-image also shows a mode-I matrix microcrack The most complex region in a CMC is the FM-interfacial which has been multideflected parallel to the fiber surface zone CMCs display a non-brittle mechanical behaviour in the interphase [23]. When necessary, selected area when the FM-bonding is not too strong. The control of the diffraction(SAD), lattice fringe(LF) image and P-EELS FM-bonding is achieved during processing, via the use of a elementary analysis are used to get an insight in the struc- thin layer of a compliant material with a low shear strength ture and composition of the FM-interfacial zone in micro/ red to as the interphase. It has been suggested that minI composites appropriate interphase materials might be those with a layered crystal structure(pyrocarbon or hex-BN)or a 3. 2. Mechanical behaviour at room temperature ayered microstructure, such as the(PyC-SiC )n interphase [22] with typically n= 1-10. The main function of the As shown in Fig. 5, the tensile stress-strain curves fo interphase is to act as mechanical fuse, i.e. to deflect the micro- and minl-composites display the same general matrix microcracks parallel to the fiber surface. The chemi features as those reported for their multidirectional counter cal composition, morphology and structure of highly engi- parts. Beyond the proportional limit, both model composites peered interphases, e.g. multilayered interphases, can be undergo damage phenomena, I.e. mainly matrix multiple extremely complex. However, their analysis can be cracking and fiber debonding, which are responsible for performed in a rather straightforward manner on either the non-linear feature of the stress-strain behaviour. As micro-or mini-composite specimens, provided appropriate strain increases, the model composites are more damaged experimental procedures are used, as already said in the with the result that: (1) the stiffness of the composites, preceding section. An example of such an analysis is assessed through secant modulus measurements, decreases shown in Fig. 4, for two SiC/SiC microcomposites with (1i) simultaneously, the width, 84, and the area,S, of the different fibers and interphases. In the Nicalon/C( B)/Sic unloading-reloading hysteresis loops increase and finally microcomposite, the interphase consists of 4(or 5)boron- (ii) the permanent residual strain(at O =O),Ep, increases doped 100 nm pyrocarbon sublayers. The first sublayer All these features depend on the intrinsic failure properties deposited on the fiber surface is a film of pure anisotropic of the brittle matrix and the characteristics of the FM-bond- pyrocarbon(deposited from propane) whereas the followi ing and are thus observed for both model and real compo- sublayers exhibit a boron concentration which increases sites. However, there are also differences which are related when moving towards the matrix(the gaseous precur to the nature of the fiber architecture and processing consid- being C3 Hg-BCI3-H2). Boron increases the anisotropy of erations. As an example, in the 2D-Nicalon/PyC/Sic pyrocarbon at low doping levels(with a maximum effect composites processed with a relatively strong FM-bonding for =8 at. %B; sublayer II)and improves its oxidation and exhibiting a high failure strain, the presence of three resistance at high doping levels(sublayers Ill-V). The families of matrix cracks has been reported, which are omposition gradient in the multilayered interphase is successively formed as the strain increases. The first family clearly apparent from the AES depth profiles, which have consists of cracks initiated at the large residual intertow een recorded from the coated fiber surface prior to the sic pores left by the CVI-process. The second family comprises matrix deposition [Fig 4(b)]. Furthermore, the TEM-image cracks formed in the transverse tows( those oriented at 90 recorded on a longitudinal thin foil of a microcomposite to the load assumed to be applied along the O longitudinal failed under tensile loading [ Fig 4(c) 1, shows:(i) the inter- tows). Finally, the third is made of transverse cracks present nal structure of the interphase i.e. five sublayers whose in the O tows. It has been shown that the longitudinal tows behave as physical entities, and dictate the mechanical beha SIGMA, Germany viour of the composite once the first and the second familie
based cement to two 100 mm SiC CVD filaments5 , acting as tensile rods, and the lifetime was measured with a timer as shown in Fig. 3(c). 3. Results and discussion The objective of the present contribution being to show the potential and limits of the micro/mini composites approach, rather than to give a detailed analysis of the materials, a few examples will be presented and discussed to illustrate how this approach can be used to assess the structural and mechanical behaviour of selected composites as well as the effect of an oxidizing atmosphere. 3.1. Chemical and structural analysis The most complex region in a CMC is the FM-interfacial zone. CMCs display a non-brittle mechanical behaviour when the FM-bonding is not too strong. The control of the FM-bonding is achieved during processing, via the use of a thin layer of a compliant material with a low shear strength referred to as the interphase. It has been suggested that appropriate interphase materials might be those with a layered crystal structure (pyrocarbon or hex-BN) or a layered microstructure, such as the (PyC-SiC)n interphase [22] with typically n 1–10. The main function of the interphase is to act as mechanical fuse, i.e. to deflect the matrix microcracks parallel to the fiber surface. The chemical composition, morphology and structure of highly engineered interphases, e.g. multilayered interphases, can be extremely complex. However, their analysis can be performed in a rather straightforward manner on either micro- or mini-composite specimens, provided appropriate experimental procedures are used, as already said in the preceding section. An example of such an analysis is shown in Fig. 4, for two SiC/SiC microcomposites with different fibers and interphases. In the Nicalon/C (B) /SiC microcomposite, the interphase consists of 4 (or 5) borondoped 100 nm pyrocarbon sublayers. The first sublayer deposited on the fiber surface is a film of pure anisotropic pyrocarbon (deposited from propane) whereas the following sublayers exhibit a boron concentration which increases when moving towards the matrix (the gaseous precursor being C3H8-BCl3-H2). Boron increases the anisotropy of pyrocarbon at low doping levels (with a maximum effect for < 8 at.% B; sublayer II) and improves its oxidation resistance at high doping levels (sublayers III–V). The composition gradient in the multilayered interphase is clearly apparent from the AES depth profiles, which have been recorded from the coated fiber surface prior to the SiCmatrix deposition [Fig. 4(b)]. Furthermore, the TEM-image recorded on a longitudinal thin foil of a microcomposite failed under tensile loading [Fig. 4(c)], shows: (i) the internal structure of the interphase i.e. five sublayers whose elementary thicknesses are of the order of 100 nm as expected; and (ii) a matrix mode-I crack, which has been deflected in mode II (parallel to the fiber surface) near the boundary between sublayers II and III [11,12]. The analysis of complex multilayered interphases becomes more difficult when the thickness of the elementary sublayer is extremely small, typically a few nm. Under such conditions, the AES microprobe resolution may be insufficient and the TEM/PEELS analysis the only appropriate technique. As an example, Fig. 4(d) shows a TEM-image recorded from a longitudinal thin foil of a Hi-Nicalon/(PyC-SiC)10/SiC microcomposite failed under tensile loading. First, the 10 PyC-SiC sequences deposited by P-CVD (each comprising a 20 nm thick pyrocarbon and a 30 nm thick SiC 1 C nanocrystalline sublayers) are clearly apparent. Second, the TEM-image also shows a mode-I matrix microcrack which has been multideflected parallel to the fiber surface in the interphase [23]. When necessary, selected area diffraction (SAD), lattice fringe (LF) image and P-EELS elementary analysis are used to get an insight in the structure and composition of the FM-interfacial zone in micro/ mini composites. 3.2. Mechanical behaviour at room temperature As shown in Fig. 5, the tensile stress–strain curves for micro- and mini-composites display the same general features as those reported for their multidirectional counterparts. Beyond the proportional limit, both model composites undergo damage phenomena, i.e. mainly matrix multiple cracking and fiber debonding, which are responsible for the non-linear feature of the stress–strain behaviour. As strain increases, the model composites are more damaged with the result that: (i) the stiffness of the composites, assessed through secant modulus measurements, decreases; (ii) simultaneously, the width, dD, and the area, S, of the unloading–reloading hysteresis loops increase and finally; (iii) the permanent residual strain (at s 0), ep , increases. All these features depend on the intrinsic failure properties of the brittle matrix and the characteristics of the FM-bonding and are thus observed for both model and real composites. However, there are also differences which are related to the nature of the fiber architecture and processing considerations. As an example, in the 2D-Nicalon/PyC/SiC composites processed with a relatively strong FM-bonding and exhibiting a high failure strain, the presence of three families of matrix cracks has been reported, which are successively formed as the strain increases. The first family consists of cracks initiated at the large residual intertow pores left by the CVI-process. The second family comprises cracks formed in the transverse tows (those oriented at 908 to the load assumed to be applied along the 08 longitudinal tows). Finally, the third is made of transverse cracks present in the 08 tows. It has been shown that the longitudinal tows behave as physical entities, and dictate the mechanical behaviour of the composite once the first and the second families R. Naslain et al. / Composites: Part A 30 (1999) 537–547 541 5 SIGMA, Germany
R. Naslain et al/Composites: Part A 30(1999)537-547 Table 1 FM-interfacial parameters for Nicalon/SiC microcomposites with a bn or PyC interphases, as calculated with the LRE model, according to Ref. 6] Materials T (MPa) G1(J/m2) SiC/BN/SIC 5510 0,08N SiC/PyC/SIC 97 2 4 4 0 2 displacement(μm) FM-bonding, i.e. the interfacial shear stress, Ti, the debond energy, Ii, and the debond length, Ia, can be derived from data recorded during tensile tests performed on microcom posites with unloading-reloading hysteresis loops, utilising the Lamon-Rebillat-Evans model [6]. Table I gives the 2500E3 corresponding interfacial parameter values for Nicalon/ BN/SIC and Nicalon/PyC/SiC microcomposites. Generally 冒60 2000 speaking, the debond energy observed for most CMCs is low(from nearly 0 to a few J/m), as expected for a mechan ical fuse. Furthermore, for a relatively strong FM-bonding, both the crack number at saturation. N. and the interfacial shear stress, Ti, are high and conversely, the debond length, 00.1020.30.40.50.60.70.809 Strain (%0) Id, is small, and vice versa for a weak FM-bonding [ 32, 33] (b) Although these micromechanics-based models could still be Fig. 5. Tensile curves for model Nicalon/SiC(CVI)composites: (a)micro- refined, e.g. to take into account the complexity of highly composites with a BN-interphase and(b) minicomposites with a pyrocar engineered interphases, they already provide a relatively rding to Refs. 6, 27, respectively simple way to compare the FM-bonding characteristics resulting from change in processing conditions and of matrix cracks have reached saturation (for a strain understand the related evolution of the mechanical beha- <0. 2%) Since in model composites, there are no residual viour of the composites macropores(at least for microcomposites) and one single The second important material data which can be fiber orientation, only matrix cracks similar to the third obtained through tensile tests performed on model ID- crack family observed in nD-composites (and which are composites are the statistical strength parameters pertinent esponsible for the majority of the non-linear o-e domain) to the matrix itself. In CMCs, both constituents are essen- are actually present. The minicomposites represent the infil- tially brittle and cracking involves defect-induced random trated longitudinal bundles in actual composites. Further- failures. The statistical distribution of strength data can be more, for C/SiC composites the matrix microcracking is described using a Weibull equation expected to be somewhat different from micro to mini to real composites ha=1-s-(x) Tensile tests performed on model composites provide important data justifying the use of micro/mini composites where PR is the failure probability, V the volum for processing optimization, interphase design and as inter- stresses, o is the applied stress and vo a reference mediate scale materials in the modelling of the mechanical and m and o are respectively the shape and the scale para- behaviour of real composites with more complex fiber archi- meters. The Weibull parameters pertinent to the fiber(mf tectures. First, they provide a rather straightforward way to and oof) are usually derived from tensile tests performed assess the values of the FM interfacial parameters, under on either single filaments or bundles [27]. Those character loading conditions more representative of those in real izing the matrix are assessed through tensile tests performed composite than e. g. the compressive loading in push-in or on a batch of models ID-composites for which the volume push-out tests. Different micromechanics-based models can of uncracked matrix, V, is known(which is not the case for a be used to extract the FM-interfacial parameters from the real composite with a complex fiber architecture). They are tensile tests [5,6,, 28-31]. As an example, the three derived from the statistical distribution of the stresses nterfacial parameters commonly used to characterize the measured at onset of cracking in the matrix 32, 33]
of matrix cracks have reached saturation (for a strain , 0.2%). Since in model composites, there are no residual macropores (at least for microcomposites) and one single fiber orientation, only matrix cracks similar to the third crack family observed in nD-composites (and which are responsible for the majority of the non-linear s-e domain) are actually present. The minicomposites represent the infiltrated longitudinal bundles in actual composites. Furthermore, for C/SiC composites the matrix microcracking is expected to be somewhat different from micro to mini to real composites. Tensile tests performed on model composites provide important data justifying the use of micro/mini composites for processing optimization, interphase design and as intermediate scale materials in the modelling of the mechanical behaviour of real composites with more complex fiber architectures. First, they provide a rather straightforward way to assess the values of the FM interfacial parameters, under loading conditions more representative of those in real composite than e.g. the compressive loading in push-in or push-out tests. Different micromechanics-based models can be used to extract the FM-interfacial parameters from the tensile tests [5,6,26,28–31]. As an example, the three interfacial parameters commonly used to characterize the FM-bonding, i.e. the interfacial shear stress, ti, the debond energy, Gi, and the debond length, ld, can be derived from data recorded during tensile tests performed on microcomposites with unloading–reloading hysteresis loops, utilising the Lamon–Rebillat–Evans model [6]. Table 1 gives the corresponding interfacial parameter values for Nicalon/ BN/SiC and Nicalon/PyC/SiC microcomposites. Generally speaking, the debond energy observed for most CMCs is low (from nearly 0 to a few J/m2 ), as expected for a mechanical fuse. Furthermore, for a relatively strong FM-bonding, both the crack number at saturation, N, and the interfacial shear stress, ti, are high and conversely, the debond length, ld, is small, and vice versa for a weak FM-bonding [32,33]. Although these micromechanics-based models could still be refined, e.g. to take into account the complexity of highly engineered interphases, they already provide a relatively simple way to compare the FM-bonding characteristics resulting from change in processing conditions and to understand the related evolution of the mechanical behaviour of the composites. The second important material data which can be obtained through tensile tests performed on model 1Dcomposites are the statistical strength parameters pertinent to the matrix itself. In CMCs, both constituents are essentially brittle and cracking involves defect-induced random failures. The statistical distribution of strength data can be described using a Weibull equation: PR 1 2 exp 2 V Vo s so m 1 where PR is the failure probability, V the volume under stresses, s is the applied stress and Vo a reference volume, and m and so are respectively the shape and the scale parameters. The Weibull parameters pertinent to the fiber (mf and sof) are usually derived from tensile tests performed on either single filaments or bundles [27]. Those characterizing the matrix are assessed through tensile tests performed on a batch of models 1D-composites for which the volume of uncracked matrix, V, is known (which is not the case for a real composite with a complex fiber architecture). They are derived from the statistical distribution of the stresses measured at onset of cracking in the matrix [32,33]. 542 R. Naslain et al. / Composites: Part A 30 (1999) 537–547 Table 1 FM-interfacial parameters for Nicalon/SiC microcomposites with a BN or PyC interphases, as calculated with the LRE model, according to Ref. [6] Materials N ti (MPa) Gi (J/m2 ) ld (mm) SiC/BN/SiC 55 10 < 0 170 14 4 3 600 9 7 7 490 5 3 4 1200 SiC/PyC/SiC 9 7 2 390 6 21 4 200 4 4 < 0 890 2 1 1 3600 Fig. 5. Tensile curves for model Nicalon/SiC (CVI) composites: (a) microcomposites with a BN-interphase and (b) minicomposites with a pyrocarbon interphase, according to Refs. [6,27], respectively
R. Naslain et al/Composites: Part A 30(1999)537-547 The third useful material data derived from tensile tests n model 1D-composites are related to the in-situ failure stength of the fibers. As in nD real composites, the failure of model 1 D-composites involves fiber pull-out and the fiber A failure surface displays the mirror-mist-hackle features classically observed on brittle materials. The mirror radius Im, is related to the in-situ tensile failure stress of the fiber OR, through the following empirical equation OR=A/(/m) where A is a constant depending on fiber fracture toughnes A=2.45 MPa mfor Nicalon fibers [27]. Mirror radii have been measured from the failure surfaces of Nicalon/PyC/ 5.05.1 SiC(CVI)minicomposites with Vf=0.27: 0.60 and 0.70 Om(Imm) in the SEm. The corresponding in-situ fiber failure stresses are presented in Fig. 8 together with the data for bare Nica Fig. 6. Weibull plot for the matrix failure stress in C(ex-PANYPyC/SiC lon single filaments and bare Nicalon fiber bundles (L= minicomposites, as derived from measurements performed with a tensile 75 mm). The value of the Weibull shape parameter, mt stage set in a scanning electron microscope, according to Ref [26] derived from the minicomposite data is 4.1-4.4. Further, he fact that the failure stress distributions for the minicom Furthermore, by performing a tensile test on a single mini- posites are almost identical to that for the Nicalon bundle omposite in a scanning electron microscope, the stresses at suggests that some degradation of the fibers may have which the successive cracks appear in the uncracked matrix occurred during minicomposite processing. The failure fragments have been measured. The statistical parameters stress distribution curves for the minicomposites are shifted were derived from the statistical distribution of the matrix towards higher stresses if the effective length is much cracking stresses taking into account the successive sizes of shorter than the gauge length(L= 75 mm)used in the the matrix fragments created at each step of the matrix bare bundle tests [27]. Therefore, these data may also cracking process. For the example shown in Fig. 6, the confirm that the fiber debonding was complete at failure statistical parameters for the SiC(Cvn)matrix, are mm and that the minicomposite failure is similar to the one of 8.5 and oom= 216 MPa(Vo=l mm). Statistical-probabil- fiber bundles istic models have been worked out to predict the stress- Although most mechanical tests have been pe train behaviour for model 1D-composites [26, 27, 32, 33]. under tensile loading in the present and related An example of such a prediction is shown in Fig. 7, for a other micromechanical tests are also of interest. Such is Nicalon/PyC/SiC minicomposite. The best fit between the the case for push-in or push-out tests conducted under model and the experimental data is observed for T. compressive loading. Such local tests (only one fiber is 80 MPa, providing thus an alternative method to get char- loaded) can be applied to either real nD-composites acteristics of the FM bonding [27, 33] model 1D-composites provided suitable specimen T=80 MP Strain (%o) Fig. 7. Simulation of the tensile stress-strain curve for a Nicalon/PyC/SiC minicomposite based on a probabilistic/statistical model, according to Ref, [27]
Furthermore, by performing a tensile test on a single minicomposite in a scanning electron microscope, the stresses at which the successive cracks appear in the uncracked matrix fragments have been measured. The statistical parameters were derived from the statistical distribution of the matrix cracking stresses taking into account the successive sizes of the matrix fragments created at each step of the matrix cracking process. For the example shown in Fig. 6, the statistical parameters for the SiC (CVI) matrix, are mm 8.5 and som 216 MPa (Vo 1 mm3 ). Statistical–probabilistic models have been worked out to predict the stress– strain behaviour for model 1D-composites [26,27,32,33]. An example of such a prediction is shown in Fig. 7, for a Nicalon/PyC/SiC minicomposite. The best fit between the model and the experimental data is observed for ti 80 MPa, providing thus an alternative method to get characteristics of the FM bonding [27,33]. The third useful material data derived from tensile tests on model 1D-composites are related to the in-situ failure stength of the fibers. As in nD real composites, the failure of model 1D-composites involves fiber pull-out and the fiber failure surface displays the mirror–mist–hackle features classically observed on brittle materials. The mirror radius, rm, is related to the in-situ tensile failure stress of the fiber, sR, through the following empirical equation: sR A= rm 1=2 2 where A is a constant depending on fiber fracture toughness. A 2.45 MPa m1/2 for Nicalon fibers [27]. Mirror radii have been measured from the failure surfaces of Nicalon/PyC/ SiC (CVI) minicomposites with Vf 0.27; 0.60 and 0.70 in the SEM. The corresponding in-situ fiber failure stresses are presented in Fig. 8 together with the data for bare Nicalon single filaments and bare Nicalon fiber bundles (L 75 mm). The value of the Weibull shape parameter, mf, derived from the minicomposite data is 4.1–4.4. Further, the fact that the failure stress distributions for the minicomposites are almost identical to that for the Nicalon bundle suggests that some degradation of the fibers may have occurred during minicomposite processing. The failure stress distribution curves for the minicomposites are shifted towards higher stresses if the effective length is much shorter than the gauge length (L 75 mm) used in the bare bundle tests [27]. Therefore, these data may also confirm that the fiber debonding was complete at failure and that the minicomposite failure is similar to the one of fiber bundles. Although most mechanical tests have been performed under tensile loading in the present and related studies, other micromechanical tests are also of interest. Such is the case for push-in or push-out tests conducted under compressive loading. Such local tests (only one fiber is loaded) can be applied to either real nD-composites or model 1D-composites provided suitable specimen R. Naslain et al. / Composites: Part A 30 (1999) 537–547 543 Fig. 6. Weibull plot for the matrix failure stress in C (ex-PAN)/PyC/SiC minicomposites, as derived from measurements performed with a tensile stage set in a scanning electron microscope, according to Ref. [26]. Fig. 7. Simulation of the tensile stress–strain curve for a Nicalon/PyC/SiC minicomposite based on a probabilistic/statistical model, according to Ref. [27]
R. Naslain et al/Composites: Part A 30(1999)537-547 10 120 25°C750°900C1050° 100 乙 1000 3000 Tensile failure stress(MPa) Fig. 8. Tensile failure stress distributions for Nicalon fibers as measured from: bare single filaments with a gauge length of 75 mm(O), bare fiber 0.0 3 bundle with a gauge length of 75 mm(e), mirror radii in Nicalon/PyC/SiC Strain(%0) (CVI)minicomposites with Vr=0.27(4),Vr=0.60(%)and Vr=0.70 Fig. 10. Stress-strain tensile curves for C(ex-PANVPyC/SiC O); according to Ref. [27] minicomposites, recorded in argon(PAr =65 kPa) and at incre temperatures, showing a rigidification effect related to crac preparation and test procedures are used [14, 34] and assum- according to Ref. 26 ing that the effects of the surrounding fibers are the same (minicomposites) or can be neglected(microcomposites) be used on a comparative basis in processing condition Stress-displacement push-out curves, recorded for Nica- lon/BN/SiC (CVi) real(2D)-and micro-composites, are shown in Fig 9. They obviously exhibit similar features. 3. 3. Mechanical behaviour at high temperature and efect of The data have been used to assess the FM-interfacial para meters according to available micromechanics-based models [32]. The interface properties determined using The effect of temperature on the mechanical behaviour of push-in tests on single filament microcomposites were in non-oxide composites has been assessed through test agreement with those obtained by analysis of tensile tests performed on ID-composites, under an atmosphere of [14]. As an example, the interfacial parameters calculated argon. As an example, Fig. 10 shows the tensile curves for Nicalon/BN/SiC (CVI) microcomposites were T,= recorded at temperature ranging from 25"C to 1050C, for 18 MPa and Gic=3. 1 J/m2 from push-in data[14] versus carbon(ex-PAN)PyC/Sic (CVI) minicomposites. Such T,=24 MPa and Gic= 2.3 J/m2'from tensile test data[6]. materials undergo microcracking during processing when Push-in or push-out tests on micro/minicomposites can cooled from the CVI-temperature (1000oC)to room temperature, due to coefficient of thermal expansion (CTE) mismatch. This feature explains the rigidification effect observed at high temperatures(T> 750C)and low applied stresses. Under such conditions, the matrix cracks 2.0 due to processing are actually closed and the composite Microc stiffness is equal le of mixture Conversely, at low temperatures(T750C for high applied stresses the matrix cracks are open and hence the stifness is lower 1.0 Crack closing as temperature is raised has been observed posites, with a heated tensile stage set in the SEM, and modelled [26] The combined effects of temperature and oxidizing atmo- sphere, i.e. the ambient air, have been studied via static(or cyclic) fatigue tests run on both micro-and mini-model Displacement(um) composites. The potential of the micro/ mini composite approach in this field is shown through a few examples Fig 9. Stress-displacement push-out curves recorded for Nicalon/BN/SIC illustrating the effect of engineered interphases on the life- (CVD)real (2D)-and micro- composites, according to Refs. [14, 32]- time under load. First, a series of Nicalon/C(B)SIC (CVI
preparation and test procedures are used [14,34] and assuming that the effects of the surrounding fibers are the same (minicomposites) or can be neglected (microcomposites). Stress–displacement push-out curves, recorded for Nicalon/BN/SiC (CVI) real (2D)- and micro-composites, are shown in Fig. 9. They obviously exhibit similar features. The data have been used to assess the FM-interfacial parameters according to available micromechanics-based models [32]. The interface properties determined using push-in tests on single filament microcomposites were in agreement with those obtained by analysis of tensile tests [14]. As an example, the interfacial parameters calculated for Nicalon/BN/SiC (CVI) microcomposites were ti 18 MPa and Gic 3.1 J/m2 from push-in data [14] versus ti 24 MPa and Gic 2.3 J/m2 from tensile test data [6]. Push-in or push-out tests on micro/minicomposites can be used on a comparative basis in processing condition screening. 3.3. Mechanical behaviour at high temperature and effect of the environment The effect of temperature on the mechanical behaviour of non-oxide composites has been assessed through test performed on 1D-composites, under an atmosphere of argon. As an example, Fig. 10 shows the tensile curves recorded at temperature ranging from 258C to 10508C, for carbon (ex-PAN)/PyC/SiC (CVI) minicomposites. Such materials undergo microcracking during processing when cooled from the CVI-temperature (<10008C) to room temperature, due to coefficient of thermal expansion (CTE) mismatch. This feature explains the rigidification effect observed at high temperatures (T . 7508C) and low applied stresses. Under such conditions, the matrix cracks due to processing are actually closed and the composite stiffness is equal to that predicted by the rule of mixture. Conversely, at low temperatures (T , 7508C) and whatever the applied stress, or at T . 7508C for high applied stresses, the matrix cracks are open and hence the stiffness is lower. Crack closing as temperature is raised has been observed and measured through tensile tests performed on minicomposites, with a heated tensile stage set in the SEM, and modelled [26]. The combined effects of temperature and oxidizing atmosphere, i.e. the ambient air, have been studied via static (or cyclic) fatigue tests run on both micro-and mini-model composites. The potential of the micro/mini composite approach in this field is shown through a few examples illustrating the effect of engineered interphases on the lifetime under load. First, a series of Nicalon/C(B)/SiC (CVI) 544 R. Naslain et al. / Composites: Part A 30 (1999) 537–547 Fig. 8. Tensile failure stress distributions for Nicalon fibers as measured from: bare single filaments with a gauge length of 75 mm (B), bare fiber bundle with a gauge length of 75 mm (QR), mirror radii in Nicalon/PyC/SiC (CVI) minicomposites with Vf 0.27 (O); Vf 0.60 (V) and Vf 0.70 (X); according to Ref. [27]. Fig. 9. Stress–displacement push-out curves recorded for Nicalon/BN/SiC (CVI) real (2D) -and micro- composites, according to Refs. [14,32]. Fig. 10. Stress–strain tensile curves for C (ex-PAN)/PyC/SiC pre-craked minicomposites, recorded in argon (PAr 65 kPa) and at increasing test temperatures, showing a rigidification effect related to crack closing, according to Ref. [26]
R. Naslain et al/Composites: Part A 30 (1999)537-547 1009 803 (b) 10KV 0ym37 40 I(PyC) <lifetime at 600 C air 800MPa 192 time (h) (c), a matrix crack as formed at room temperature and as filled with a low melting glass after the fatigue test, according to Ref le lifetime diagram, (b), ig. 11. Static fatigue tests performed under tensile loading(o= 800 MPa)in air at 600"C, for Nicalon/C (B)SiC microcomposites:(a) microcomposites with C(B)graded composition inter- and the applied load modelled on the basis of the Sic phases, A and A, depicted in a previous section were fati- crack opening(which controls the oxygen diffusion) and gue tested in air at 600C under a constant tensile load the oxidation kinetics of the carbon fibers, pyrocarbon inter higher than the proportional limit. Under such conditions, phase and SiC-matrix [26, 35]. The resulting lifetime the Sic-matrix was microcracked and the cracks maintained diagram presented in Fig. 12, shows that: (i)at low tempera open allowing oxygen to diffuse towards the interphase. The tures, e.g. 600C, the lifetime does not depend on the time to failure was measured for each specimen and the applied load since oxidation is rate limited by surface reac percentage of specimens remaining unbroken for a given tions and the cracks open whatever the applied load; and(ii) ageing duration, calculated. As shown in Fig. Il, replacing at high temperatures, the lifetime is strongly dependent on the conventional PyC-interphase by a B(C) interphase the applied load: at low applied load, the matrix cracks increases significantly the composite lifetime. As a matter remain closed as already mentioned, the oxygen diffusion of fact, when boron is present in the interphase, a low melt- across the SiC-matrix is slow and hence, the lifetime is long, ing glass is formed by oxidation, which fills the microcracks whereas, at high applied load, the matrix cracks are open. and slows down the oxygen diffusion towards the fiber [12]. the oxygen diffusion and the chemical reactions are fast and In a second series of experiments, Hi-Nicalon/SiC (Cvi) thus the lifetime short. These conclusions are consistent sites with either a 100 nm PyC-interphase or(Pyc- with those previously drawn for 2D-C(ex-PAN)/C/SIC multilayered interphases were tested in cyclic fatigue related composites [36,37] at 700C in air[23]. The value of n was equal to 10 or 30, the thickness of the elementary SiC sublayer equal to 10 30 nm and that of PyC to 3, 10 or 50 nm. The resul ts showed 4. Conclusion that all the microcomposites with a single 100 nm PyC nterphase failed within less than I h. Conversely, the 1. Micro and mini non-oxide composites can be prepared in a reproducible manner and in a short time by chemical luch more resistant to oxidation, as it was expected [22] vapor deposition or infiltration. They bear basic informa- As an example, 20% of the specimens were still unbroken tion about fiber/matrix interactions and can be easily after 24 h, for n= 10; e(Py C)=3 nm and e(siC)=10 or modelled owing to their extremely simple fiber config- 30 nm. In a last series of experiments, a batch of C(ex uration and morphology PAN)C/SiC minicomposites have been fatigue tested in 2. The fiber/matrix interfacial zone can be analyzed at the air at 600< T<1200C and under a constant tensile nm-scale by Auger electron spectroscopy and transmis- load ranging from 0 to 10 N. The lifetime was measured sion electron microscopy, even when it is complex and its variations as a function of both the temperature 3. Important mechanical parameters, such as the Weibu
microcomposites with C (B) graded composition interphases, A and A 0 , depicted in a previous section were fatigue tested in air at 6008C under a constant tensile load higher than the proportional limit. Under such conditions, the SiC-matrix was microcracked and the cracks maintained open allowing oxygen to diffuse towards the interphase. The time to failure was measured for each specimen and the percentage of specimens remaining unbroken for a given ageing duration, calculated. As shown in Fig. 11, replacing the conventional PyC-interphase by a B (C) interphase increases significantly the composite lifetime. As a matter of fact, when boron is present in the interphase, a low melting glass is formed by oxidation, which fills the microcracks and slows down the oxygen diffusion towards the fiber [12]. In a second series of experiments, Hi-Nicalon/SiC (CVI) composites with either a 100 nm PyC-interphase or (PyCSiC)n multilayered interphases were tested in cyclic fatigue at 7008C in air [23]. The value of n was equal to 10 or 30, the thickness of the elementary SiC sublayer equal to 10 or 30 nm and that of PyC to 3, 10 or 50 nm. The results showed that all the microcomposites with a single 100 nm PyCinterphase failed within less than 1 h. Conversely, the microcomposites with a multilayered interphase were much more resistant to oxidation, as it was expected [22]. As an example, 20% of the specimens were still unbroken after 24 h, for n 10; e (PyC) 3 nm and e (SiC) 10 or 30 nm. In a last series of experiments, a batch of C (exPAN)/C/SiC minicomposites have been fatigue tested in air at 600 , T , 12008C and under a constant tensile load ranging from 0 to 10 N. The lifetime was measured and its variations as a function of both the temperature and the applied load modelled on the basis of the SiC crack opening (which controls the oxygen diffusion) and the oxidation kinetics of the carbon fibers, pyrocarbon interphase and SiC-matrix [26,35]. The resulting lifetime diagram presented in Fig. 12, shows that: (i) at low temperatures, e.g. 6008C, the lifetime does not depend on the applied load since oxidation is rate limited by surface reactions and the cracks open whatever the applied load; and (ii) at high temperatures, the lifetime is strongly dependent on the applied load: at low applied load, the matrix cracks remain closed as already mentioned, the oxygen diffusion across the SiC-matrix is slow and hence, the lifetime is long, whereas, at high applied load, the matrix cracks are open, the oxygen diffusion and the chemical reactions are fast and thus the lifetime short. These conclusions are consistent with those previously drawn for 2D-C (ex-PAN)/C/SiC related composites [36,37]. 4. Conclusion 1. Micro and mini non-oxide composites can be prepared in a reproducible manner and in a short time by chemical vapor deposition or infiltration. They bear basic information about fiber/matrix interactions and can be easily modelled owing to their extremely simple fiber configuration and morphology. 2. The fiber/matrix interfacial zone can be analyzed at the nm-scale by Auger electron spectroscopy and transmission electron microscopy, even when it is complex. 3. Important mechanical parameters, such as the Weibull R. Naslain et al. / Composites: Part A 30 (1999) 537–547 545 Fig. 11. Static fatigue tests performed under tensile loading (s 800 MPa) in air at 6008C, for Nicalon/C (B)/SiC microcomposites: (a) lifetime diagram, (b), (c), a matrix crack as formed at room temperature and as filled with a low melting glass after the fatigue test, according to Ref. [12]
slain et al. / Composites: Part A 30(1999)537-547 trc e tida at diffusion rate ea ox dation 00000 100 600700800900100011001200 Temperature(°C) Fig. 12. Lifetime of C(ex-PANVPyC/SiC(CVI)minicomposites in static fatigue in air, as a function of the temperature and the applied tensile load, according to Ref. 26 statistical parameters of both the fiber and matrix, as well composite materials In: Tressler RE, Messing GL, Pantano CG, as the fiber-matrix interfacial parameters and the effect Newnham re, editors. Tai Multiphase and Compo of the environment, can be derived from tensile tests mics, Mater. Sci. Research, 20. New York: Plenum Press [2] Stinton DP, Caputo AJ, Lowden RA. Synthesis of fiber-reinforced SiC 4. Conversely, the micro/mini composite approach has also its limit. It does not take into account the macro defects chemical vapor infiltration. Amer Ceram Soc Bull 986;65(2):347 (macropores), the interactions between the elementary 3 David P, Narcy B, Lulewicz JD, Ravel F, Schulmann S. In: P tows, the tow undulation(woven fiber architectures ), the R and street K editors. Proc. 10th Int Conf Composite Mate in-depth diffusion of oxygen in the macroporosity, etc IV. Abington, Cambridge: Woodhead Publishing, 1995: 611 4]Golecki l, Morris RC, Narasimhan D, Clements N. Rapid which are present in real composites. Nevertheless, they tion of carbon-carbon composites by thermal-gradient chemical are useful model materials for the optimization of real apor infiltration. In: Evans AG, Naslain R editors. High-temperature composites. Finally, the data generated from tests on eramic-matrix composites Il, Ceram. Trans. 58. Westerville, OH: micro/mini composites are often necessary for the model Am. Ceramics Soc. 1995: 231 ing of the thermomechanical and chemical behaviour of [] Lamon J, Rechiniac C, Lissart N, Come P Determination of inter- facial properties in ceramic matrix composites using microcomposite composites with more complex fiber architectures imens. In: Bunsell AR, Jamet JF, Massiah A editors. Proc. ECCM 5, Developments in the Science and Technology of Composite Mate- als. Bordeaux EACM. 1992 895-900 6 Lamon J, Rebillat F, Evans AG. Assessment of a microcomposite test Acknowledgement: procedure for evaluating constituent properties of ceramic matri composites. J Am Ceram Soc 1995: 78(2): 401 This study is part of a long term research program on the [7 Lamon J, Rechiniac C, Roach DH, Jouin JM. Micromechanical and synthesis, characterization and modelling of non-oxide atistical approach to the behavior of CMCs Ceram Engng Sci Proc CMCs, funded by CNRS and SEP. The authors are grateful 993;14(9-10):1115 [8 Lamon J, Guillaumat L. Influence of interfacial characterization on to Drs. S. Jacques, L. Guillaumat, F. Heurtevent, GA he non-linear deformations of ceramic matrix composites. In: Pour Lebrun, N. Lissart and F. Rebillat for the data that they sarti R, Street K editors. Proc. 10th Int. Conf Composite Materials, provided through their PhD work and for fruitful discussion, vol. IV. Abington, Cambridge: Woodhead Publishing, 1995: 649- as well as to M. Saux and J. Forget for their assistance in the preparation of the manuscript F, Naslain R, Goujard S High ure lifetime in air of Sic/C (B)SiC microcomposites prepared by LPCVD. In: Evans AG, Naslain R, editors. High-temperature cera- osites I. Ceram. Trans. 57. Westerville. OH: Am. References eram. Soc. 1995: 381 [10 Rebillat F, Crepin-Leblond J, Mougin S, Lamon J. Experimental [1] Naslain R, Langlais F. CVD-processing of ceramic-ceramic determination of the stress field in fiber-matrix interphases of
statistical parameters of both the fiber and matrix, as well as the fiber–matrix interfacial parameters and the effect of the environment, can be derived from tensile tests. 4. Conversely, the micro/mini composite approach has also its limit. It does not take into account the macro defects (macropores), the interactions between the elementary tows, the tow undulation (woven fiber architectures), the in-depth diffusion of oxygen in the macroporosity, etc..., which are present in real composites. Nevertheless, they are useful model materials for the optimization of real composites. Finally, the data generated from tests on micro/mini composites are often necessary for the modelling of the thermomechanical and chemical behaviour of composites with more complex fiber architectures. Acknowledgements This study is part of a long term research program on the synthesis, characterization and modelling of non-oxide CMCs, funded by CNRS and SEP. The authors are grateful to Drs. S. Jacques, L. Guillaumat, F. Heurtevent, G.A. Lebrun, N. Lissart and F. Rebillat for the data that they provided through their PhD work and for fruitful discussion, as well as to M. Saux and J. Forget for their assistance in the preparation of the manuscript. References [1] Naslain R, Langlais F. CVD-processing of ceramic–ceramic composite materials In: Tressler RE, Messing GL, Pantano CG, Newnham RE, editors. Tailoring Multiphase and Composite Ceramics, Mater. Sci. Research, 20. New York: Plenum Press, 1986:145. [2] Stinton DP, Caputo AJ, Lowden RA. Synthesis of fiber-reinforced SiC composites by chemical vapor infiltration. Amer Ceram Soc Bull 1986;65(2):347. [3] David P, Narcy B, Lulewicz JD, Ravel F, Schulmann S. In: Poursartip R and Street K editors. Proc. 10th Int. Conf. Composite Materials, vol. IV. Abington, Cambridge: Woodhead Publishing, 1995:611–616. [4] Golecki I, Morris RC, Narasimhan D, Clements N. Rapid densification of carbon–carbon composites by thermal-gradient chemical vapor infiltration. In: Evans AG, Naslain R editors. High-temperature ceramic–matrix composites II, Ceram. Trans. 58. Westerville, OH: Am. Ceramics Soc., 1995:231. [5] Lamon J, Rechiniac C, Lissart N, Corne P. Determination of interfacial properties in ceramic matrix composites using microcomposites specimens. In: Bunsell AR, Jamet JF, Massiah A editors. Proc. ECCM 5, Developments in the Science and Technology of Composite Materials. Bordeaux: EACM, 1992:895–900. [6] Lamon J, Rebillat F, Evans AG. Assessment of a microcomposite test procedure for evaluating constituent properties of ceramic matrix composites. J Am Ceram Soc 1995;78(2):401. [7] Lamon J, Rechiniac C, Roach DH, Jouin JM. Micromechanical and statistical approach to the behavior of CMCs. Ceram Engng Sci Proc 1993;14(9–10):1115. [8] Lamon J, Guillaumat L. Influence of interfacial characterization on the non-linear deformations of ceramic matrix composites. In: Poursartip R, Street K editors. Proc. 10th Int. Conf. Composite Materials, vol. IV. Abington, Cambridge: Woodhead Publishing, 1995:649– 656. [9] Jacques S, Guette A, Langlais F, Naslain R, Goujard S. High temperature lifetime in air of SiC/C (B)/SiC microcomposites prepared by LPCVD. In: Evans AG, Naslain R, editors. High-temperature ceramic–matrix composites I. Ceram. Trans, 57. Westerville, OH: Am. Ceram. Soc., 1995:381. [10] Rebillat F, Cre´pin–Leblond J, Mougin S, Lamon J. Experimental determination of the stress field in fiber–matrix interphases of 546 R. Naslain et al. / Composites: Part A 30 (1999) 537–547 Fig. 12. Lifetime of C (ex-PAN)/PyC/SiC (CVI) minicomposites in static fatigue in air, as a function of the temperature and the applied tensile load, according to Ref. [26]