PHILOSOPHICAL MAGAZINE A. 1998. VOL. 78. No. 1. 189-202 Sic Nicalon fibre-glass matrix composites: interphases and their mechanism of formation By E. LE STRAT Laborator d Etudes des materiaux, 29 avenue de la Division Leclerc, Office National d Etudes et de recherches Aerospatiale, Chatillon, France M. LANCIN.N. FOURCHES-COULON and C. marhic ntre de Micro caracteris2m比Hom4 channe, s [Received 16 June 1997 and accepted 17 November 1997] ABSTRACT The fibre-matrix reaction that occurs during the processing of a Nicalon SiC fibre-Pyrex glass matrix composite is analysed. Interphases are characterized by means of various complementary techniques: electron diffraction, HRTEM, EDX, EELS, SIMS and XPS. Neither the results of the present study nor those previously obtained for Nicalon SiC fibre-LAS(Li2O-Al2O-SiO,) glass or available model of reaction. An alternative reaction mechanism is suggested and oxidize SiC and Sio Cy in the fibre. The reaction yields carbon and silicon oxycarbide in the fibre and SiO which dissolves into the matrix. When the oxygen in excess in the matrix is consumed, the reaction stops and the phases on in the reaction layer which generates two interphases one carbon rich and the other silicon oxycarbide rich. These interphases are observed at the fibre periphery in all glass or glass-ceramic matrix composit §1. INTRODUCTION Glass or glass-ceramic matrix composites are developed for thermostructural applications such as in aircraft parts submitted to severe thermomechanical stresses. Glass-ceramic is reinforced by means of long fibres that are essentially stronger and less brittle than the matrix. This reinforcement yields composites with high strength and a'pseudoplastic' strain sufficient to alleviate the catastrophic rupture mode of a brittle material. Amongst commercial fibres, the Sic nicalon 202 fibre has been successfully used in the last years. The mechanical properties of Nicalon-glass matrix composites depend on the nature of the interphases that develop during material processing(Brennan 1988, Lancin 1991). A good characterization of the reaction between the fibre and the matrix together with the understanding of this reaction are essential in order to elaborate composites with required specifications Studies of the interphases produced in the course of the fibre-matrix reaction have been conducted in a number of composites based on various matrices(Brennan 1988, Chen et al. 1989, Bonney and Cooper 1990 ). The most currently used are Li2O-Al2O-SiO,(LAS)or Cao-Al2O3-SiO2( CAS)containing some additives such as niobium, arsenic or yttrium. In order to explain the formation of interphases, one usually refers to the model proposed by Cooper and Chyung(1987) which 0141-8610/98 $1200 C 1998 Taylor Francis Ltd
PHILOSOPHICAL MAGAZINE A, 1998, VOL. 78, NO. 1, 189±202 SiC Nicalon ®bre±glass matrix composites: interphases and their mechanism of formation By E. Le Strat Laboratore d’Etudes des Materiaux, 29 avenue de la Division Leclerc, O ce National d’Etudes et de Recherches Ae rospatiales, Chatillon, France M. Lancin, N. Fourches-Coulon and C. Marhic Centre de Micro Caracte risations, Laboratoire de Physique Cristalline, Institut des Materiaux de Nantes, 2 rue de la HoussinieÁre, 44 322 Nantes, France [Received 16 June 1997 and accepted 17 November 1997] Abstract The ®bre±matrix reaction that occurs during the processing of a Nicalon SiC ®bre±Pyrex glass matrix composite is analysed. Interphases are characterized by means of various complementary techniques: electron di raction, HRTEM, EDX, EELS, SIMS and XPS. Neither the results of the present study nor those previously obtained for Nicalon SiC ®bre±LAS (Li2O±Al2O3±SiO2) glass or LAS + MAS (MgO±Al2O3±SiO2) glass±ceramic matrix composites support the available model of reaction. An alternative reaction mechanism is suggested whereby the dissolved and non-bridging oxygen atoms of the matrix di use and oxidize SiC and SiOxCy in the ®bre. The reaction yields carbon and silicon oxycarbide in the ®bre and SiO2 which dissolves into the matrix. When the oxygen in excess in the matrix is consumed, the reaction stops and the phases undergo a reorganization in the reaction layer which generates two interphases, one carbon rich and the other silicon oxycarbide rich. These interphases are observed at the ®bre periphery in all glass or glass±ceramic matrix composites. § 1. Introduction Glass or glass±ceramic matrix composites are developed for thermostructural applications such as in aircraft parts submitted to severe thermomechanical stresses. Glass±ceramic is reinforced by means of long ®bres that are essentially stronger and less brittle than the matrix. This reinforcement yields composites with high strength and a `pseudoplastic’ strain su cient to alleviate the catastrophic rupture mode of a brittle material. Amongst commercial ®bres, the SiC Nicalon 202 ®bre has been successfully used in the last years. The mechanical properties of Nicalon±glass matrix composites depend on the nature of the interphases that develop during material processing (Brennan 1988, Lancin 1991). A good characterization of the reaction between the ®bre and the matrix together with the understanding of this reaction are essential in order to elaborate composites with required speci®cations. Studies of the interphases produced in the course of the ®bre±matrix reaction have been conducted in a number of composites based on various matrices (Brennan 1988, Chen et al. 1989, Bonney and Cooper 1990). The most currently used are Li2O-Al2O3-SiO2 (LAS) or CaO±Al2O3-SiO2 (CAS) containing some additives such as niobium, arsenic or yttrium. In order to explain the formation of interphases, one usually refers to the model proposed by Cooper and Chyung (1987) which 0141±8610/98 $12.00 Ñ 1998 Taylor & Francis Ltd
E Le strat et al involves the oxidation of Sic grains of the fibre by 'excess oxygen of the matrix. This oxidation would occur according to the following reaction: SiC+O→C+SiO The Cooper-Chyung model is based upon the two fundamental hypotheses: firstly that the Nicalon fibre contains SiC, carbon and Sio and secondly that the fibres are hermodynamically stable, which in turn implies that the activities of the different phases are equal to unity. The diff usion of silicon, carbon and oxygen from the fibre to the matrix is controlled by gradients of activity between these. Several studies have subsequently shown that the fibre is actually made up of Sic and carbon but that it also contains a significant amount of silicon oxycarbide and very little Sioz (Lipowitz et al. 1987, Laffon et al. 1989, Porte and Sartre 1989, Bleay et al. 1992) Other studies have proved that the Nicalon fibre is not thermodynamically stable (Mah et al. 1984, Clark et al. 1985, Johnson et al. 1988, Le Coustumer et al. 1993) Every published study shows that, given the elaboration conditions, a carbon interphase develops whose thickness and structure depend on characteristics of the matrix( Brennan 1988, Cooper and Chyung 1987, Ponthieu et al. 1990). These stu- dies generally agree as to the formation of a NbC interphase between the carbon layer and the Nb2 Os-containing matrix. Other results are more controversial, how- ever. The formation of silica is difficult to demonstrate. The presence of Sio between the carbon phase and the matrix has been identified only rarely by the electron diffraction pattern(EDP) or energy dispersive X-ray spectroscopy DXS)( Cooper and Chyung 1987, Bonney and Cooper 1990, Doreau 1995, Hahnel et al. 1995) but silica could be incorporated in the majority of matrices and then be undetectable. Kumar and Knowles(1996) claim to have localized silica between the carbon phase and the fibre but the edP, on which this analysis relies, cannot discriminate carbon from SiO nor can electron-energy-loss spectroscopy (EELS)differentiate SiO Cy from Sio2+ SiC. By means of high-resolution trans- mission electron microscopy(HRTEM), EDXS, secondary-ion mass spect (SIMS), EELS, Auger electron spectroscopy(AES)and X-ray photoelectro roscopy(XPS), Ponthieu et al. (1990, 1994)and Lancin et al. (1994 have ide an interphase between carbon and the fibre in Nicalon-LAS composites, whose structure and composition differ from those of the fibre. This zone is enriched with oxygen compared with the fibre, it does not contain Sio but essentially SiO Cy as demonstrated by XPS analysis. Since then, Pippel et al.(1995)b EELS, Doreau(1995) by HRTEM and EDXS and Lancin et al.(1993)by HRTEM, EDXS, SIMS and eels have identified this interphase in other compo- sites. It is worth emphasizing that this oxygen-rich interphase can be spotted in every published AES or EDXs analysis even though its existence had not actually been noted(Brennan 1988, Chen et al. 1989, Bonney and Cooper 1990, Homeny et al. 1990). It is, however, not yet proved that the composition of this oxygen-enriched interphase, as revealed by XPS in Nicalon-LAS composites, remains the same gardless of the matrix. The nature and the localization of the phases resulting from the fibre-matrix reaction are thus a matter of debate and this holds true also for their kinetics of formation. According to Cooper and Chyung(1987), Homeny et al.(1990)and Hahnel et al.(1995), the thickness of the carbon interphase increases with increasing hot-pressing temperature or time. By contrast, Brennan(1988), Ponthieu(1990)and
involves the oxidation of SiC grains of the ®bre by `excess’ oxygen of the matrix. This oxidation would occur according to the following reaction: SiC + O2 ® C + SiO2. The Cooper±Chyung model is based upon the two fundamental hypotheses: ®rstly that the Nicalon ®bre contains SiC, carbon and SiO2 and secondly that the ®bres are thermodynamically stable, which in turn implies that the activities of the di erent phases are equal to unity. The di usion of silicon, carbon and oxygen from the ®bre to the matrix is controlled by gradients of activity between these. Several studies have subsequently shown that the ®bre is actually made up of SiC and carbon but that it also contains a signi®cant amount of silicon oxycarbide and very little SiO2 (Lipowitz et al. 1987, La on et al. 1989, Porte and Sartre 1989, Bleay et al. 1992). Other studies have proved that the Nicalon ®bre is not thermodynamically stable (Mah et al. 1984, Clark et al. 1985, Johnson et al. 1988, Le Coustumer et al. 1993). Every published study shows that, given the elaboration conditions, a carbon interphase develops whose thickness and structure depend on characteristics of the matrix (Brennan 1988, Cooper and Chyung 1987, Ponthieu et al. 1990). These studies generally agree as to the formation of a NbC interphase between the carbon layer and the Nb2O5-containing matrix. Other results are more controversial, however. The formation of silica is di cult to demonstrate. The presence of SiO2 between the carbon phase and the matrix has been identi®ed only rarely by the electron di raction pattern (EDP) or energy dispersive X-ray spectroscopy (EDXS) (Cooper and Chyung 1987, Bonney and Cooper 1990, Doreau 1995, HaÈhnel et al. 1995) but silica could be incorporated in the majority of matrices and then be undetectable. Kumar and Knowles (1996) claim to have localized silica between the carbon phase and the ®bre but the EDP, on which this analysis relies, cannot discriminate carbon from SiO2 nor can electron-energy-loss spectroscopy (EELS) di erentiate SiOxCy from SiO2 + SiC. By means of high-resolution transmission electron microscopy (HRTEM), EDXS, secondary-ion mass spectrometry (SIMS), EELS, Auger electron spectroscopy (AES) and X-ray photoelectron spectroscopy (XPS), Ponthieu et al. (1990, 1994) and Lancin et al. (1994) have identi®ed an interphase between carbon and the ®bre in Nicalon±LAS composites, whose structure and composition di er from those of the ®bre. This zone is enriched with oxygen compared with the ®bre; it does not contain SiO2 but essentially SiOxCy as demonstrated by XPS analysis. Since then, Pippel et al. (1995) by EELS, Doreau (1995) by HRTEM and EDXS and Lancin et al. (1993) by HRTEM, EDXS, SIMS and EELS have identi®ed this interphase in other composites. It is worth emphasizing that this oxygen-rich interphase can be spotted in every published AES or EDXS analysis even though its existence had not actually been noted (Brennan 1988, Chen et al. 1989, Bonney and Cooper 1990, Homeny et al. 1990). It is, however, not yet proved that the composition of this oxygen-enriched interphase, as revealed by XPS in Nicalon±LAS composites, remains the same regardless of the matrix. The nature and the localization of the phases resulting from the ®bre±matrix reaction are thus a matter of debate and this holds true also for their kinetics of formation. According to Cooper and Chyung (1987), Homeny et al. (1990) and HaÈhnel et al. (1995), the thickness of the carbon interphase increases with increasing hot-pressing temperature or time. By contrast, Brennan (1988), Ponthieu (1990) and 190 E. Le Strat et al
SiC Nicalon fibre-glass matrix composites Lancin et al.(1993) did not report any significant modification of the carbon inter phase with hot-pressing time in Nicalon-LAS The present study was initiated in order to shed further light on the properties of the interphases and on their formation kinetics in Nicalon-Pyrex composites. A Pyrex matrix was selected because of the simplicity of its composition. The absence of additions such as Nb O enables one to avoid, between carbon and the matrix. any secondary reaction which could complicate the interpretation of the experimen al data. §2. MATERIALS Pyrex consists essentially of Sio,(80 wt%)and B2O(13. 1 wt%) which are arranged as a network of tetrahedra connected by bridging oxygen atoms. Pyrex also contains modifying oxides, namely Na2O(3.5 wt%), Fe2O+ AlO(2.5 wt%) and Ko(1.15 wt%), which involve polar bondings between cations and non-brid- ging oxygen atoms, that is oxygen atoms linked to only one silicon or boron atom. Oxygen is also dissolved in the glass in the form of interstitial molecular oxygen. In Pyrex, the non-bridging oxygen concentration, as determined by the corresponding concentrations of sodium, iron, aluminium and potassium, is low(several atomic per cent). The concentration of dissolved oxygen is more difficult to estimate because it depends both on the glass and on its preparation conditions. Typically, this concen- tration should be higher than that of the non-bridging oxygen (Qi et al. 1993) Two composites were prepared at the Office National d Etudes et de recherches Aerospatiale de Chatillon by hot pressing, at 1398 K, two-dimensional Nicalon woven fibres impregnated with a suspension containing the Pyrex powder. In order to obtain a uniform temperature in the materials, the preforms were main- tained for 25 nm at 1398 K before hot pressing. Two times for hot pressing were chosen which were sufficiently different(5 and 235 min) for the expected thickness ariation in the reaction zone to be larger than the scatter of the measurements, by at least one order of magnitude 83. EXPERIMENTAL TECHNIQU ts, Electron microscopy studies were conducted on thin foils prepared by mechan- al grinding and subsequent ion thinning. The bright-field and high-resolution ere obtained with a Philips CM20 and a Hitachi field emission gun HF 2000. EDXS and EELS analyses were realized on the FEG equipped with a Si-Li Super Quantek diode(Kevex)and a parallel EELS detector( Gatan). The probe siz was 6nm and 2 nm for EDXS and EELS respectively SIMS was performed on fibres mechanically extracted from the composite hot pressed for 5 min. Analyses were realized with a Cameca IMS 4F spectrometer. The primary beam consisted of Cs ions extracted under a voltage of 10kv. The proce- dure adopted during the analyses has been described by Lancin et al. (1997). XPS analysis, alternating with ion erosion, was conducted on extracted fibres in order to reveal the possible depth-dependent phase modifications. The apparatus was a Leybold Heraeus LHS 12 spectrometer equipped with a Mg Ka source (1253.6eV). The fibres were extracted by matrix dissolution in hydrofluoric acid. By comparing mechanically and chemically extracted fibres we found that the che- mical treatment did not modify the surface composition of the fibres. Analysis was realized on fibre tows because the analysed area was approximately 4 mm x 10 mm. SiC and Sio Cy standards were used in order to obtain the energy of the Si 2p
Lancin et al. (1993) did not report any signi®cant modi®cation of the carbon interphase with hot-pressing time in Nicalon±LAS. The present study was initiated in order to shed further light on the properties of the interphases and on their formation kinetics in Nicalon±Pyrex composites. A Pyrex matrix was selected because of the simplicity of its composition. The absence of additions such as Nb2O5 enables one to avoid, between carbon and the matrix, any secondary reaction which could complicate the interpretation of the experimental data. § 2. Materials Pyrex consists essentially of SiO2 (80 wt%) and B2O3 (13.1 wt%) which are arranged as a network of tetrahedra connected by `bridging’ oxygen atoms. Pyrex also contains modifying oxides, namely Na2O (3.5 wt%), Fe2O3 + Al2O3 (2.5 wt%) and K2O (1.15 wt%), which involve polar bondings between cations and non-bridging oxygen atoms, that is oxygen atoms linked to only one silicon or boron atom. Oxygen is also dissolved in the glass in the form of interstitial molecular oxygen. In Pyrex, the non-bridging oxygen concentration, as determined by the corresponding concentrations of sodium, iron, aluminium and potassium, is low (several atomic per cent). The concentration of dissolved oxygen is more di cult to estimate because it depends both on the glass and on its preparation conditions. Typically, this concentration should be higher than that of the non-bridging oxygen (Qi et al. 1993). Two composites were prepared at the O ce National d’Etudes et de Recherches Ae rospatiales de Chatillon by hot pressing, at 1398 K, two-dimensional Nicalon woven ®bres impregnated with a suspension containing the Pyrex powder. In order to obtain a uniform temperature in the materials, the preforms were maintained for 25 nm at 1398 K before hot pressing. Two times for hot pressing were chosen which were su ciently di erent (5 and 235 min) for the expected thickness variation in the reaction zone to be larger than the scatter of the measurements, by at least one order of magnitude. § 3. Experimental techniques Electron microscopy studies were conducted on thin foils prepared by mechanical grinding and subsequent ion thinning. The bright-®eld and high-resolution images were obtained with a Philips CM20 and a Hitachi ®eld emission gun HF 2000. EDXS and EELS analyses were realized on the FEG equipped with a Si±Li Super Quantek diode (Kevex) and a parallel EELS detector (Gatan). The probe size was 6 nm and 2 nm for EDXS and EELS respectively. SIMS was performed on ®bres mechanically extracted from the composite hot pressed for 5 min. Analyses were realized with a Cameca IMS 4F spectrometer. The primary beam consisted of Cs+ ions extracted under a voltage of 10 kV. The procedure adopted during the analyses has been described by Lancin et al. (1997). XPS analysis, alternating with ion erosion, was conducted on extracted ®bres in order to reveal the possible depth-dependent phase modi®cations. The apparatus was a Leybold Heraeus LHS 12 spectrometer equipped with a Mg Ka source (1253.6 eV). The ®bres were extracted by matrix dissolution in hydro¯ uoric acid. By comparing mechanically and chemically extracted ®bres we found that the chemical treatment did not modify the surface composition of the ®bres. Analysis was realized on ®bre tows because the analysed area was approximately 4 mm ´ 10 mm. SiC and SiOxCy standards were used in order to obtain the energy of the Si 2p SiC Nicalon ®bre±glass matrix composites 191
E. Le strat et al transition and its full width at half-maximum. Nicalon fibres were also analysed against depth. No change in the Si 2p and C Is transitions was detected, showing that the sputtering did not induce any phase modification of the fibres 4. RESULTS The two composites have similar microstructures with slight differences, how ever Between the fibre and the matrix, we always observe a reaction layer consisting of two distinct interphases(figure 1). The interphase next to the matrix, namely the carbon layer(CL), is essentially composed of carbon. The following interphase exhibits varying microstructure and composition from the Cl to the unmodified fibre core(this intermediate interphase is called the transition layer(TL)). The thick ness of the Cl interphase varies from one fibre to the next, just as that of the Tl does but to a lesser extent than the CL. Typically, the Cl thickness is 30-40 nm and the TL thickness is 60-70nm. It is worth noting that we detect no thickness dependence of the reaction layer(RL)(which equals the Cl plus the Tl)upon the time t of hot The CL microstructure exhibits the characteristic features of microporous tur bostratic carbon (figure 2). The TL microstructure is similar to that of the fibre although these can be easily differentiated. Just as the fibre, the Tl contains Sic grains(figure 3). As distinct from what is found in the fibre, the density of Sic grains decreases in the Tl from the fibre inside towards the CL. sic grains are larger in the TL than in the fibre as shown in the high-resolution images and EDPs(figure 4) Moreover, the images and patterns reveal the existence of carbon structural units (SUs)in the Tl which are not detected in the fibre. The density and size of the carbon SUs in the CL, which features in its organization, slightly increases with Figure 1. Typical interface. The bright-field image shows two interphases with different contrast. The CL is the carbon interphase and the tl a transition layer where the microstructure and the composition vary from the unmodified fibre core to the Cl
transition and its full width at half-maximum. Nicalon ®bres were also analysed against depth. No change in the Si 2p and C 1s transitions was detected, showing that the sputtering did not induce any phase modi®cation of the ®bres. § 4. Results The two composites have similar microstructures with slight di erences, however. Between the ®bre and the matrix, we always observe a reaction layer consisting of two distinct interphases (®gure 1). The interphase next to the matrix, namely the carbon layer (CL), is essentially composed of carbon. The following interphase exhibits varying microstructure and composition from the CL to the unmodi®ed ®bre core (this intermediate interphase is called the transition layer (TL)). The thickness of the CL interphase varies from one ®bre to the next, just as that of the TL does but to a lesser extent than the CL. Typically, the CL thickness is 30±40 nm and the TL thickness is 60±70 nm. It is worth noting that we detect no thickness dependence of the reaction layer (RL) (which equals the CL plus the TL) upon the time t of hot pressing. The CL microstructure exhibits the characteristic features of microporous turbostratic carbon (®gure 2). The TL microstructure is similar to that of the ®bre although these can be easily di erentiated. Just as the ®bre, the TL contains SiC grains (®gure 3). As distinct from what is found in the ®bre, the density of SiC grains decreases in the TL from the ®bre inside towards the CL. SiC grains are larger in the TL than in the ®bre as shown in the high-resolution images and EDPs (®gure 4). Moreover, the images and patterns reveal the existence of carbon structural units (SUs) in the TL which are not detected in the ®bre. The density and size of the carbon SUs in the CL, which features in its organization, slightly increases with 192 E. Le Strat et al. Figure 1. Typical interface. The bright-®eld image shows two interphases with di erent contrast. The CL is the carbon interphase and the TL a transition layer where the microstructure and the composition vary from the unmodi®ed ®bre core to the CL
n Figure 2. Microstructure of the carbon interphase as shown by the HRTEM image of microporous turbostratic carbon increasing hot-pressing time. By contrast, the TL microstructure does not undergo significant changes The composition variations in the Rl is revealed by EDXS, SIMS and EELS EDXS analyses show that the CL contains about 2 at. %Si and 10 at. %O. The CL also contains traces of sodium(1 at. or less ) whereas traces of iron, aluminium a Common to all our analyses, the main characteristic of the Tl is its high oxygen content, which is actually markedly larger than in the fibre. In spite of the local concentration fluctuations, the eDXS analyses always show a more prominent oxygen concentration in the Tl than in the fibre(figure 5). The carbon, silicon and oxygen contents in the Tl amount to 45 at. % 23 at and 30 at. respectively. In the fibre, they amount to 50at %, 32 at and 18 at. % respectively. In electron- energy-loss spectra, the Si L2,3 edge structure at 116ev, characteristic of the Si-O- Si bond, is clearly more pronounced in the tl than in the fibre(figure 6). On SIMS profiles, the intensities of the C, Si and O signals are constant in the fibre whereas they vary significantly at the fibre periphery. The CL is revealed by the maximum in the C signal. The Tl is characterized by a peak in the O signal,a minimum in the C signal compared with the fibre and a shoulder in the Si signal (arrowed in figure 7). The TL can be characterized by a higher oxygen concentration and lower carbon and silicon contents than to those in the fibre The asymmetry of the C and o profiles shows that the variation in composi tion is more progressive between the CL and the tl than between the Cl and the matrix, which is well supported by HRTEM observations. The width of the tl car be derived from SIMS profiles. The uncertainty of this estimation mainly result from the location of the tl- fibre interface because the o-to-c and si-to-c ratios reach constant values at different depths(figure 7). We choose to take int account the variation in the Si -to-C ratio. Even so, the width(90 nm) is larger than that derived from hrtEM observations. SIMS profiles show in addition that the tl contains a low percentage of boron while this element is not detectable in the CL. The other matrix elements could not be identified by SIMs because inherent
increasing hot-pressing time. By contrast, the TL microstructure does not undergo signi®cant changes. The composition variations in the RL is revealed by EDXS, SIMS and EELS. EDXS analyses show that the CL contains about 2 at.%Si and 10 at.%O. The CL also contains traces of sodium (1 at.% or less), whereas traces of iron, aluminium and potassium are within the experimental uncertainty. Common to all our analyses, the main characteristic of the TL is its high oxygen content, which is actually markedly larger than in the ®bre. In spite of the local concentration ¯ uctuations, the EDXS analyses always show a more prominent oxygen concentration in the TL than in the ®bre (®gure 5). The carbon, silicon and oxygen contents in the TL amount to 45 at.%, 23 at.% and 30 at.% respectively. In the ®bre, they amount to 50 at.%, 32 at.% and 18 at.% respectively. In electronenergy-loss spectra, the Si L2,3 edge structure at 116 eV, characteristic of the Si±O± Si bond, is clearly more pronounced in the TL than in the ®bre (®gure 6). On SIMS pro®les, the intensities of the C - , Si- and O - signals are constant in the ®bre whereas they vary signi®cantly at the ®bre periphery. The CL is revealed by the maximum in the C - signal. The TL is characterized by a peak in the O - signal, a minimum in the C - signal compared with the ®bre and a shoulder in the Si- signal (arrowed in ®gure 7). The TL can be characterized by a higher oxygen concentration and lower carbon and silicon contents than to those in the ®bre. The asymmetry of the C - and O - pro®les shows that the variation in composition is more progressive between the CL and the TL than between the CL and the matrix, which is well supported by HRTEM observations. The width of the TL can be derived from SIMS pro®les. The uncertainty of this estimation mainly results from the location of the TL±®bre interface because the O - -to-C- and Si- -to-Cratios reach constant values at di erent depths (®gure 7). We choose to take into account the variation in the Si- -to-C- ratio. Even so, the width (90 nm) is larger than that derived from HRTEM observations. SIMS pro®les show in addition that the TL contains a low percentage of boron while this element is not detectable in the CL. The other matrix elements could not be identi®ed by SIMS because inherent SiC Nicalon ®bre±glass matrix composites 193 Figure 2. Microstructure of the carbon interphase as shown by the HRTEM image of microporous turbostratic carbon