Acta mater. vol 44 Copyright C 1996 Acta Metallurgica 09567151(95)00386X Printed in Great Dr 1359645496515.00+0.00 MICROSTRUCTURE-PROPERTY RELATIONSHIPS OF SIC FIBRE-REINFORCED MAGNESIUM ALUMINOSILICATES-I. MICROSTRUCTURAL CHIARACTERISATION A KUMAR+ and K M. KNOWLES University of Cambridge, Department of Materials Science and Metallurgy, Pembroke Street, Cambridge CB2 3QZ, England (Received 31 October 1994: in revised form 6 September 1995) Abstract-The microstructure of two magnesium aluminosilicates unidirectionally reinforced with SiC fibres(Nicalon) has been examined. a diphasic interlayer having a higher O/Si ratio than in the fibres was found on the surface of the fibres in both composites. This interlayer could be identified as an amorphous mixture of silica and carbon in the composite hot-pressed just below the liquidus temperature of stoichiometric cordierite(composite A). In the other composite hot-pressed at 920 C and subsequently craned at 1150C (composite B), a relatively thicker diphasic interlayer was observed, consisting of lentified in composite A. A thin interlayer consisting mostly of matrix elements was also identified etween the diphasic interlayer and the discrete carbon interlayer in this cor le. Dierences in structure and morphology of interfacial regions in the two composites could clearly be attributed differences in the hot-pressing schedules. The basal planes of turbostratic carbon were aligned parallel he fibre-matrix interfaces in both composites. Copyright c 1996 Acia Metallurgica Inc. 1 INTRODUCTION It is evident from the early work of Brennan [1, 2 Glass-ceramics reinforced with continuous Sic fibres offer a combination of properties which are desirable related to the nature of interfacial layers or high-temperature structural applications. the an [1, 2] demonstrated that a carbon-rich layer high strength and high modulus of these composites at the fibre matrix interface is always present in is a direct consequence of the high strength and tough SiC/lithium aluminosilicate(LAs)composites modulus of the nanocrystalline SiC fibres, while the investigate the nature of the number of researchers have since attempted to damage tolerance of these composites arises from the ese interfacial layers and ability of fibres first to carry load after transverse ve tried to correlate this with the mechanical pull out from the matrix properties of other potential glass-ceramic matrix materials. For these toughening mechanisms to oper- aluminosilicates (MAS). Chaim et al. B3 observed a relatively weak in shear to allow debonding at these Sic/MAS composite which they interpreted as silica interfaces, yet strong enough to give good load because of the brittle nature of the composite. They transfer between the fibre and the matrix. Therefore. were unable to identify any carbon-rich layers unam the mechanical properties of these composites, and in oguously, even in the damage tolerant composites Metcalfe et al. [4] have recentl particular their damage tolerance, depend largely on demonstrated that strong and tough composites with the mechanical properties of the fibre-matrix inter faces, which in turn are controlled by the structure MAS as the matrix can be produced, but they did not analyse the matrix-fibre interfaces in sufficient and chemistry of the interfaces. Hence, in order to detail to be able to identify the structure and chem- there is a need to characterise both the chemistry and istry of the thin reaction layer they. observed by the microstructure of the fibre-matrix interface and scanning electron microscopy. It is, therefore, the phases present in the matrix apparent that an unambiguous identification of the structure and the chemistry of the interfacial layers Present address: Department of Mechanical mate performance of Naval Postgraduate School, Montere U.S.A literature
Pergamon 0956-7151(95)00386-X Acta mater. Vol. 44, No. 7, PP. 2901-2921. 1996 Copyright 0 1996 Acta Metallurgica Inc. Published by Elsevier Science Ltd Printed in Great Britain. All rights reserved 1359-6454/96 $15.00 + 0.00 MICROSTRUCTURE-PROPERTY RELATIONSHIPS OF Sic FIBRE-REINFORCED MAGNESIUM ALUMINOSILICATES-I. MICROSTRUCTURAL CHARACTERISATION A. KUMARt and K. M. KNOWLES University of Cambridge, Department of Materials Science and Metallurgy, Pembroke Street, Cambridge CB2 342, England (Received 31 October 1994; in revised form 6 September 1995) Abstract-The microstructure of two magnesium aluminosilicates unidirectionally reinforced with Sic fibres (Nicalon) has been examined. A diphasic interlayer having a higher OjSi ratio than in the fibres was found on the surface of the fibres in both composites. This interlayer could be identified as an amorphous mixture of silica and carbon in the composite hot-pressed just below the liquidus temperature of stoichiometric cordierite (composite A). In the other composite hot-pressed at 920°C and subsequently ceramed at 1150°C (composite B), a relatively thicker diphasic interlayer was observed, consisting of turbostratic carbon together with amorphous silica. A distinct interlayer of turbostratic carbon was identified in composite A. A thin interlayer consisting mostly of matrix elements was also identified between the diphasic interlayer and the discrete carbon interlayer in this composite. Differences in the structure and morphology of interfacial regions in the two composites could clearly be attributed to differences in the hot-pressing schedules. The basal planes of turbostratic carbon were aligned parallel to the fibre-matrix interfaces in both composites. Copyright 0 1996 Acfa Metallurgica Inc. 1. INTRODUCTION Glass-ceramics reinforced with continuous SIC fibres offer a combination of properties which are desirable for high-temperature structural applications. The high strength and high modulus of these composites is a direct consequence of the high strength and modulus of the nanocrystalline SIC fibres, while the damage tolerance of these composites arises from the ability of fibres first to carry load after transverse matrix cracking and then to pull out from the matrix beyond the peak in the load-deflection curve for such materials. For these toughening mechanisms to operate successfully, the fibre-matrix interfaces must be relatively weak in shear to allow debonding at these interfaces, yet strong enough to give good load transfer between the fibre and the matrix. Therefore, the mechanical properties of these composites, and in particular their damage tolerance, depend largely on the mechanical properties of the fibre-matrix interfaces, which in turn are controlled by the structure and chemistry of the interfaces. Hence, in order to optimise the mechanical properties of composites, there is a need to characterise both the chemistry and the microstructure of the fibre-matrix interface and the phases present in the matrix. TPresent address: Department of Mechanical Engineering, Naval Postgraduate School, Monterey. CA 93943, U.S.A. It is evident from the early work of Brennan [l, 21 that the mechanical properties of composites are closely related to the nature of interfacial layers. Brennan [l, 21 demonstrated that a carbon-rich layer at the fibre-matrix interface is always present in tough Sic/lithium aluminosilicate (LAS) composites. A number of researchers have since attempted to investigate the nature of these interfacial layers and have tried to correlate this with the mechanical properties of other potential glass-ceramic matrix composites, such as BC fibre-reinforced magnesium aluminosilicates (MAS). Chaim et al. [3] observed a reaction layer at matrix-fibre interfaces in a SiC/MAS composite which they interpreted as silica because of the brittle nature of the composite. They were unable to identify any carbon-rich layers unambiguously, even in the damage tolerant composites they produced. Metcalfe et al. [4] have recently demonstrated that strong and tough composites with MAS as the matrix can be produced, but they did not analyse the matrix-fibre interfaces in sufficient detail to be able to identify the structure and chemistry of the thin reaction layer they observed by scanning electron microscopy. It is, therefore, apparent that an unambiguous identification of the structure and the chemistry of the interfacial layers and their effects on the ultimate performance of SiC/MAS composites has not yet been reported in the literature
KUMAR and KNOWLES: SiC REINFORCED ALUMINOSILICAtESI In the work reported here, two different mag- sitional maps. Electron beam energy and beam cur- nesium aluminosilicate matrices unidirectionally rent were 5kv and 4nA, respectively. A digital reinforced with SiC (Nicalon)-fibres have been exam- resolution of 128 x 128 was used to collect all the ned to determine the relationship between mechan- compositional images using a conventional eds de ical properties and microstructure. In this paper(Part tector (detection limit atomic number >11).Dwell D), the evolution of microstructure during processing time per pixel was 2 s for composite A and 3 s for and post-processing heat-treatments is described for composite B No matrix correction was applied to the both composites. Qualitative and quantitative analy- digitised maps and therefore these maps are used only ses of interfacial layers and the microstructure of the for comparison purpose composites have been carried out using conventional 2.2. Interface analysis. The fibre-matrix inter- and analytical electron microscopy, optical faces were analysed using thin foils in transmission microscopy and X-ray diffraction techniques. The electron microscopes. Both a Philips 400T operating mechanical behaviour of these composites is reported at 120 kV and a JEOL 2000FX operating at 200kV in Part I [S were used. A JEOL 4000EX-lI operating at 400 kv was used to perform high-resolution electron mi- 2. MATERIALS AND EXPERIMENTAL METHODS croscopy(HREM). An ultra thin window energy dispersive X-ray analyser(EDS, Link model 6284) and a parallel electron energy-loss spectrometer Composite plates of sizes 10 x 10 cm in the form of (PEELS, Gatan model 666) on the JEOL 2000FX layers of unidirectional fibre-reinforced matrices, were used to carry out the compositional and struc. ea[0l, configuration, were received from Pilkington tural analysis of the fibres, the matrices and th Technology Centre, Lathom, U.K. in two batches. fibre-matrix interfaces SiC(Nicalon, NL 202, Nippon Carbon Company Thin foils for TEM examination were prepared by of Japan)fibres were used in both batches. A mechanically thinning the small rectangular samples glass of stoichiometric cordierite composition to thicknesses of A 20-50 um. The mechanical thin (2MgO2Al2O, 5SiO2)was chosen as the matrix ning was carried out only on high grade SiC papers material for composite A, whereas a glass of compo- (600 and 800)and the pressure applied was kept to weight with small amounts of P2 Os(2.0 wt%)and time than the routine grinding process, it reduced the B,O, (1.0 wt%)was used for composite B. P2O, acts erosion of fibre-matrix interfacial areas during mech anical thinning. Both surfaces of the mechanically B,O, lowers the melt temperature and delays particle surface crystallisation [6, 7]. As will be discussed thinned samples were polished for approx 15 min using first 6 um diamond paste and then l um Section 3. 1, the rationale for choosing this compo- diamond paste. This was essential for uniform thin sition is to reduce the hot-pressing temperature ning of samples in an ion-beam thinner. The polished glass for both composites was melted in air in a central hole of size 2000 or 1000 u u using sil gas-fired furnace. The composite plates were made by the slurry impregnation procedure for making con- as adhesive. This was followed by ion-beam tinuous fibre-reinforced composites [8]. The plates ( Gatan dual ion mill model 600) at 5kv were hot-pressed inert atmosphere at 1500 and incidence to perforation, and finally thinning at a low 920C for composites A and B, respectively 9 Plates of composite B were ceramed in air for I h at 1150C after consolidation, whereas composite a did not siC(220) receive any heat-treatment after hot-pressing SiC (11) 2. 2. Experimental c(110) 2.2. 1. Microstructural characterisation, The micro- structure of both composites was assessed using an X-ray diffractometer(Philips PW 3719), an optical microscope (Olympus, model BHM) and scanning electron microscopes (SEM, Camscan S2 and Radial Camscan $4). Polished composite cross-sections were coated with either carbon or gold before SEM exam ination. Energy dispersive X-ray analysis(EDS)of fibre and matrix was carried out on a sem(camscan S4)using a conventional Link EDS analyser with a detection limit of atomic number >11 Fig. 1 Schematic SADP illustrating overlapping rings from a dedicated clectron-probe microanalyser Sio,, C and Sic along with the aperture positions for radial ( Cameca, SX-50) was used to obtain digital compo- oration of the diffraction pattern
2902 KUMAR and KNOWLES: Sic REINFORCED ALUMINOSILICATES-I In the work reported here, two different magnesium aluminosilicate matrices unidirectionally reinforced with SIC (Nicalon)-fibres have been examined to determine the relationship between mechanical properties and microstructure. In this paper (Part I), the evolution of microstructure during processing and post-processing heat-treatments is described for both composites. Qualitative and quantitative analyses of interfacial layers and the microstructure of the composites have been carried out using conventional and analytical electron microscopy, optical microscopy and X-ray diffraction techniques. The mechanical behaviour of these composites is reported in Part II [5]. 2. MATERIALS AND EXPERIMENTAL METHODS 2.1. Materials Composite plates of sizes 10 x 10 cm in the form of n layers of unidirectional fibre-reinforced matrices, i.e. a [0], configuration, were received from Pilkington Technology Centre, Lathom, U.K. in two batches. SIC (Nicalon, NL 202, Nippon Carbon Company of Japan) fibres were used in both batches. A glass of stoichiometric cordierite composition (2Mg0.2A1203.5Si02) was chosen as the matrix material for composite A, whereas a glass of composition 22.3% MgO, 21.3% Al,O,, 53.4% SiO, by weight with small amounts of P,O, (2.0 wt%) and B,03 (1.0 wt%) was used for composite B. P,O, acts as a nucleating agent and reduces viscosity, whereas B,O, lowers the melt temperature and delays particle surface crystallisation [6,7]. As will be discussed in Section 3.1, the rationale for choosing this composition is to reduce the hot-pressing temperature required to consolidate the composites. The precursor glass for both composites was melted in air in a gas-fired furnace. The composite plates were made by the slurry impregnation procedure for making continuous fibre-reinforced composites [8]. The plates were hot-pressed in an inert atmosphere at 1500 and 920°C for composites A and B, respectively [9]. Plates of composite B were ceramed in air for 1 h at 1150°C after consolidation, whereas composite A did not receive any heat-treatment after hot-pressing. 2.2. Experimental 2.2.1. Microstructural characterisation. The microstructure of both composites was assessed using an X-ray diffractometer (Philips PW 3719), an optical microscope (Olympus, model BHM) and scanning electron microscopes (SEM, Camscan S2 and Camscan S4). Polished composite cross-sections were coated with either carbon or gold before SEM examination. Energy dispersive X-ray analysis (EDS) of fibre and matrix was carried out on a SEM (Camscan, S4) using a conventional Link EDS analyser with a detection limit of atomic number 2 11. A dedicated electron-probe microanalyser (Cameca, SX-50) was used to obtain digital compositional maps. Electron beam energy and beam current were 5 kV and 4nA, respectively. A digital resolution of 128 x 128 was used to collect all the compositional images using a conventional EDS detector (detection limit atomic number > 11). Dwell time per pixel was 2 s for composite A and 3 s for composite B. No matrix correction was applied to the digitised maps and therefore these maps are used only for comparison purposes. 2.2.2. Interface analysis. The fibre-matrix interfaces were analysed using thin foils in transmission electron microscopes. Both a Philips 400T operating at 120 kV and a JEOL 2000FX operating at 200 kV were used. A JEOL 4000EX-II operating at 400 kV was used to perform high-resolution electron microscopy (HREM). An ultra thin window energy dispersive X-ray analyser (EDS, Link model 6284) and a parallel electron energy-loss spectrometer (PEELS, Gatan model 666) on the JEOL 2000FX were used to carry out the compositional and structural analysis of the fibres, the matrices and the fibre-matrix interfaces. Thin foils for TEM examination were prepared by mechanically thinning the small rectangular samples to thicknesses of x20-50 pm. The mechanical thinning was carried out only on high grade SIC papers (600 and 800) and the pressure applied was kept to a minimum. Although this procedure took a longer time than the routine grinding process, it reduced the erosion of fibre-matrix interfacial areas during mechanical thinning. Both surfaces of the mechanically thinned samples were polished for approximately 15 min using first 6 pm diamond paste and then 1 pm diamond paste. This was essential for uniform thinning of samples in an ion-beam thinner. The polished sections were then mounted onto copper rings with a central hole of size 2000 or 1000 pm using silver paint as adhesive. This was followed by ion-beam thinning (Gatan dual ion mill model 600) at 5 kV and 15” incidence to perforation, and finally thinning at a low silica Fig. 1. Schematic SADP illustrating overlapping rings from SiO,, C and Sic along with the aperture positions for radial exploration of the diffraction pattern
KUMAR and KNOWLEs: SIC REINFOKCED ALUMINOSILICA'TES-I 2%U3 angle(a10)for 5-30 min to extend the thin area. Table 1. Crystal structure and lattice parameters of the phases When required, the thin foils for hREM were coated with a thin layer of amorphous carbon using a carbon Pha Structure evaporator unit(Edwards E306) Quantitative and qualitative analyses of a micro- a-cordieriteb structural feature require a detailed correlation Orthorhombic 9.2 between the diffraction pattern and the image. The 'A.S.T. M index card no. 15-776 centred dark field technique was therefore used to a.s.t.m index card no, 11-273 dentify the phases present at interfacial layers. This vas achieved by radial exploration of reciprocal space following the procedure described by Oberlin was lower than that which could be readily detected [10 by the diffractometer(realistically an amount <10 A schematic diagram depicting the relative pos- wt %, even if a sophisticated analysis of the x-ray pattern(SADP), from SiO2, SiC and C is shown in were to have been used). Cordierite was present in the Fig.I.Amorphous silica gives a diffuse ring at high temperature hexagonal form in both composites 0.41 nm and a faint plateau out to 0.12 nm[11]. The The crystal structure and lattice parameters of the B-SiC in the nanocrystalline Nicalon fibres gives 111 relevant phases are given in Table 1. It was also noted 251 nm), 220(0. 154 nm)and 311(0.131 nm)rings. that in composite A the diffraction peaks of B-siC 02(0.344 nm), 100(0.212 nm)and 110(0.11 nm)(from the fibres)were better defined than those seen rings of turbostratic carbon are also shown in Fig. 1. in composite B. This suggests that in composite A the Turbostratic carbon consists of aromatic layer stacks average grain size of crystals of p-Sic in the fibres that are piled up in parallel but rotated slightly at Optical micrographs of polished cross-sections of ndom relative to one another [12 ] This structure is characterised by the presence of the 002 reflection composites A and B are shown in Figs 2(a)and(b), 44 A)and hko instead of hkI diffraction spots respectively. Qualitatively, the fibre distribution is or diffraction rings from graphite [12]. When the relatively uniform, but some matrix-rich regions are bjective aperture is in position 1(Fig. 1), C and Sio, also evident. Mean fibre volume fractions estimated will appear bright in a dark field image, whereas in using an optical microscope were 0. 47 and 0.40 for position 2(Fig. 1)crystals of Sic will appear bright. composites A and B, respectively. In general, the When imaging interfacial regions it is important polished sections of composite A are pore-free, that the interfacial layer is parallel to the electron whereas micro-pores are clearly seen on the polished beam.The specimen needs to be tilted so that the surfaces of composite B. Back-scattered electron terfacial layer is aligned parallel to the electron images of the polished sections of composites A and beam(edgc-on imaging). If the interface layer is not B are shown in Figs 3(a) and(b), respcctivcly. A seen edge-on, misinterpretation of the interface 1-2 um thick bright layer was consistently observed gions is possible because of overlapping of around the fibres in composite a [Fig 3(a)l, whereas fibre-interface-matrix regions (see for example a very thin dark layer was seen around the fibres in the discussion in Ref. [13]) composite B[Fig. 3(b)]. The different morphologies of the crystalline phases within the matrix are also 3. RESULTS AND INTERPRETATION readily seen in Figs 2 and 3 The observed differences in the microstructure of 3. 1. Microstructural characterisation the two composites arise from the difference in Analysis of X-ray diffraction (XRD) traces of processing temperature and the composition of glass compositesAandBrevealedthefollowingmulliteprecursorsusedtofabricatethecomposites.com (AlO3 2SiO2) was the major crystalline phase in posite A was hot-pressed at a temperature just below which small amounts of the liquidus temperature of stoichiometric cordierite a-cordierite(2MgO. 2A1,O3. 5SiO,), a-cristobalite glass. The formation of different phases on cooling (SiO,)and relatively large amounts of residual glass from the hot-pressing temperature can be understood were present, and (i)a-cordierite and protoenstatite by considering the pseudobinary section going (MgSiO,)were the major crystalline phases in com. through cordierite(2MgO. 2Al2O3. 5SiO2) and the osite B, although some diffraction peaks corre- theoretical compound "Mg-beryl" in the sponding to orthoenstatite were also present in the MgO-Al2Oj-Sio, ternary phase diagram(Fig. 4, XRD trace of composite B. However, it is difficult to [15D. At temperatures below 1523 C mullite and gla distinguish the different pyroxenes because of the (i.e. supercooled liquid) will be present (point A overlapping of various peaks in XRD. The absence of Fig. 4). On further cooling to 1465 C (point B in Fig a hump, characteristic of a siliceous glassy phase, 4)all glass and mullite should react to form cordier between 20 and 30 20 in the XRD trace of com- ite. The co-existence of mullite, glass and,most istobalite with cordierite at room ss or that, more likely, the amount of residual glass temperature in composite a would suggest that the
KUMAR and KNOWLES: Sic REINFORCED ALUMINOSILICATES-I 2903 angle (x 100) for 5-30 min to extend the thin area. When required, the thin foils for HREM were coated with a thin layer of amorphous carbon using a carbon evaporator unit (Edwards E306). Quantitative and qualitative analyses of a microstructural feature require a detailed correlation between the diffraction pattern and the image. The centred dark field technique was therefore used to identify the phases present at interfacial layers. This was achieved by radial exploration of reciprocal space following the procedure described by Oberlin [lOI. A schematic diagram depicting the relative positions of various hkl rings in a selected area diffraction pattern (SADP), from SiOz, SIC and C is shown in Fig. 1. Amorphous silica gives a diffuse ring at 0.41 nm and a faint plateau out to 0.12 nm [ll]. The /?-Sic in the nanocrystalline Nicalon fibres gives 111 (0.251 nm), 220 (0.154 nm) and 311 (0.131 nm) rings. 002 (0.344 nm), 100 (0.212 nm) and 110 (0.11 nm) rings of turbostratic carbon are also shown in Fig. 1. Turbostratic carbon consists of aromatic layer stacks that are piled up in parallel but rotated slightly at random relative to one another [12]. This structure is characterised by the presence of the 002 reflection (d > 3.44 A) and hk0 instead of hkl diffraction spots or diffraction rings from graphite [12]. When the objective aperture is in position 1 (Fig. 1), C and SiO, will appear bright in a dark field image, whereas in position 2 (Fig. 1) crystals of SIC will appear bright. When imaging interfacial regions it is important that the interfacial layer is parallel to the electron beam. The specimen needs to be tilted so that the interfacial layer is aligned parallel to the electron beam (edge-on imaging). If the interface layer is not seen edge-on, misinterpretation of the interface regions is possible because of overlapping of fibre-interface-matrix regions (see for example the discussion in Ref. [13]). 3. RESULTS AND INTERPRETATION 3.1. Microstructural characterisation Analysis of X-ray diffraction (XRD) traces of composites A and B revealed the following: (i) mullite (3Al,0,.2Si02) was the major crystalline phase in composite A, in addition to which small amounts of cr -cordierite (2MgO. 2Al,O,. 5SiO,), cr -cristobalite (SiO,) and relatively large amounts of residual glass were present, and (ii) c( -cordierite and protoenstatite (MgSiO,) were the major crystalline phases in composite B, although some diffraction peaks corresponding to orthoenstatite were also present in the XRD trace of composite B. However, it is difficult to distinguish the different pyroxenes because of the overlapping of various peaks in XRD. The absence of a hump, characteristic of a siliceous glassy phase, between 20” and 30” 28 in the XRD trace of composite B suggested that either there is no residual glass or that, more likely, the amount of residual glass Table 1. Crystal structure and lattice parameters of the phases present in the as-received composites Lattice parameters (A) Phase structure a b c Mullit? Orthorhombic 1.54 1.86 2.884 a-cordier@ Hexagonal 9.11 9.71 9.352 Protoenstatitec Orthorhombic 9.25 8.75 5.320 ‘A.S.T.M index card no. 15-716. bA.S.T.M index card no. 13-293. ‘A.S.T.M index card no. 1 l-273. was lower than that which could be readily detected by the diffractometer (realistically an amount < 10 wt%, even if a sophisticated analysis of the X-ray spectrum such as a modified Rietveld analysis [14] were to have been used). Cordierite was present in the high temperature hexagonal form in both composites. The crystal structure and lattice parameters of the relevant phases are given in Table 1. It was also noted that in composite A the diffraction peaks of B-Sic (from the fibres) were better defined than those seen in composite B. This suggests that in composite A the average grain size of crystals of /I-Sic in the fibres was higher than that in the fibres in composite B. Optical micrographs of polished cross-sections of composites A and B are shown in Figs 2(a) and (b), respectively. Qualitatively, the fibre distribution is relatively uniform, but some matrix-rich regions are also evident. Mean fibre volume fractions estimated using an optical microscope were 0.47 and 0.40 for composites A and B, respectively. In general, the polished sections of composite A are pore-free, whereas micro-pores are clearly seen on the polished surfaces of composite B. Back-scattered electron images of the polished sections of composites A and B are shown in Figs 3(a) and (b), respectively. A l-2 pm thick bright layer was consistently observed around the fibres in composite A [Fig. 3(a)], whereas a very thin dark layer was seen around the fibres in composite B [Fig. 3(b)]. The different morphologies of the crystalline phases within the matrix are also readily seen in Figs 2 and 3. The observed differences in the microstructure of the two composites arise from the difference in the processing temperature and the composition of glass precursors used to fabricate the composites. Composite A was hot-pressed at a temperature just below the liquidus temperature of stoichiometric cordierite glass. The formation of different phases on cooling from the hot-pressing temperature can be understood by considering the pseudobinary section going through cordierite (2MgO. 2A1, 03. 5Si02) and the theoretical compound “Mg-beryl” in the Mg@Al,03-Si02 ternary phase diagram (Fig. 4, [ 151). At temperatures below 1523°C mullite and glass (i.e. supercooled liquid) will be present (point A in Fig. 4). On further cooling to 1465°C (point B in Fig. 4) all glass and mullite should react to form cordierite. The co-existence of mullite, glass and, most importantly, cristobalite with cordierite at room temperature in composite A would suggest that the
2904 KUMAR and KNOWLES: SiC REINFORCED ALUMINOSILICATES-I 88 ● °° 50 um Fig. 2. Optical micrographs of polished cross-sections of (a) composite A and (b)composite B composition of the original glass precursor was non- mullite and liquid will be very sluggish and will stoichiometric and that it should be closer to the require very slow cooling rates. The two main reasons .2O-Sior side of the Mgo-AlO Sio, ternary for the co-existence of mullite, cordierite, cristobalite phase diagram(the glass composition would be in the and liquid in composite a are therefore: (i a small phase field on the extreme right of the pseudobinary deviation from the exact 2: 2: 5 cordierite compo- section in Fig. 4). This assumption is supported by sition and (i)a non-equilibrium cooling rate. he fact that the projection point of the 2: 2: 5 compo- Cristobalite and cordierite precipitate on further sition is completely surrounded by fields containing cooling. Cristobalite is undesirable because it under liquids as seen in an isothermal section at 1460.c goes a displacive phase transformation around Fig. 5, [15]. Therefore, if a glass composition closer 200-250'C to form low cristobalite [16]. This involves to the Al2O-SiO2 side of the ternary system is a substantial volume reduction (e 3.9%)which can chosen, mullite, cordierite and liquid would be pre- cause extensive cracking within cristobalite and sent at 1460'C. Furthermore, peritectic reactions are therefore in the matrix [17, 18 very slow and even in metallic systems they require A glass of non-stoichiometric cordierite compo very slow cooling rates to go to completion. There- sition in the MAs phase field was used to fabricate fore, the kinetics of the formation of cordierite from composite B. The two main reasons for selecting a
290 4 KUMAR and KNOWLES: Sic REINFORCED ALUMINOSILICATES-I F ?g. 2. Optical micrographs of polished cross-sections of (a) composite A and (b) composite B. composition of the original glass precursor was nonstoichiometric and that it should be closer to the A&O,-SiO, side of the Mg&Al,OrSiO, ternary phase diagram (the glass composition would be in the phase field on the extreme right of the pseudobinary section in Fig. 4). This assumption is supported by the fact that the projection point of the 2 : 2 : 5 composition is completely surrounded by fields containing liquids as seen in an isothermal section at 1460°C (Fig. 5, [ 151). Therefore, if a glass composition closer to the Al,03-SiO, side of the ternary system is chosen, mullite, cordierite and liquid would be present at 1460°C. Furthermore, peritectic reactions are very slow and even in metallic systems they require very slow cooling rates to go to completion. Therefore, the kinetics of the formation of cordierite from mullite and liquid will be very sluggish and will require very slow cooling rates. The two main reasons for the co-existence of mullite, cordierite, cristobalite and liquid in composite A are therefore: (i) a small deviation from the exact 2: 2: 5 cordierite composition and (ii) a non-equilibrium cooling rate. Cristobalite and cordierite precipitate on further cooling. Cristobalite is undesirable because it undergoes a displacive phase transformation around 200-250°C to form low cristobahte [ 161. This involves a substantial volume reduction (z 3.9%) which can cause extensive cracking within cristobalite and therefore in the matrix [17, 181. A glass of non-stoichiometric cordierite composition in the MAS phase field was used to fabricate composite B. The two main reasons for selecting a
KUMAR and KNOWLES: SIC REINFORCED ALUMINOSILICATES-I non-stoichiometric glass composition were: (i) to [20] proposed that cooling from the protoenstatite decrease the hot-pressing temperature and (ii)to stability field results in an intergrowth structure avoid the formation of mullite and cristobalite. The consisting of regions of clinoenstatite and orthoen composites were hot-pressed at 920.C to avoid crys- statite, the relative amounts of which will depend on tallisation during pressing, which would increase the the cooling rate Slower cooling rates are predicted to viscosity of the glass, Cordierite and protoenstatite produce more regions of orthoenstatite [20]. On the formed on ceraming the as-pressed composites at other hand, Schreyer and Schairer [15]used their own 1 150C. The presence of protoenstatite in the matrix experimental evidence to propose that protoenstatite at room temperature is surprising because this form could be retained at room temperature with very low of enstatite is known to be stable only at high amounts of clinoenstatite and orthoenstatite-like temperature [19-21]. Other forms of enstatite are: (i) phases when Al2O, was present in the bulk compo clinoenstatite, which forms on quenching proton- sition(see compositions no 35 and 37 in their paper) statite to room temperature, and which may be The formation of protoenstatite in similar MAS metastable at all temperatures; and (i)orthoenstatite, glasses has also been observed by many other re- a low temperature phase [19, 20]. Buseck and Iijima searchers, and yet none of them commented on the a b Fig. 3. Back-scattered electron images of polished cross-sections of (a)composite A and (b)
KUMAR and KNOWLES: Sic REINFORCED ALUMINOSILICATES-I 2905 non-stoichiometric glass composition were: (i) to decrease the hot-pressing temperature and (ii) to avoid the formation of mullite and cristobalite. The composites were hot-pressed at 920°C to avoid crystallisation during pressing, which would increase the viscosity of the glass. Cordierite and protoenstatite formed on ceraming the as-pressed composites at 1150°C. The presence of protoenstatite in the matrix at room temperature is surprising because this form of enstatite is known to be stable only at high temperature [19-211. Other forms of enstatite are: (i) clinoenstatite, which forms on quenching protoenstatite to room temperature, and which may be metastable at all temperatures; and (ii) orthoenstatite, a low temperature phase [19,20]. Buseck and Iijima [20] proposed that cooling from the protoenstatite stability field results in an intergrowth structure consisting of regions of clinoenstatite and orthoenstatite, the relative amounts of which will depend on the cooling rate. Slower cooling rates are predicted to produce more regions of orthoenstatite [20]. On the other hand, Schreyer and Schairer [ 151 used their own experimental evidence to propose that protoenstatite could be retained at room temperature with very low amounts of clinoenstatite and orthoenstatite-like phases when A&O, was present in the bulk composition (see compositions no. 35 and 37 in their paper). The formation of protoenstatite in similar MAS glasses has also been observed by many other researchers, and yet none of them commented on the Fig. 3. Back-scattered electron images of polished cross-sections of (a) composite A and (b) composite B
KUMAR and KNOWLES: SIC REINFORCED ALUMINOSILICATESI matrix, but in addition the XRD pattern suggested that orthoenstatite existed 1500 observations showed the presence of a few twinned rains typical of clinoenstatite and orthoenstatite in P+L he matrix. Therefore, while it is possible that pre L toenstatite remains stable at room temperature fo the matrix composition used to fabricate comp 1400 B, further work is clearly required to be able to C(ss)+P+L understand the mechanism by which it can happen C(ss) Hot-pressing of composite A at high temperature ields a pore-free microstructure because of increased C(ss)+P+S viscous fow due to the formation of a liquid How- 1300 C+P+s increase markedly the diffusion of matrix elements 80 Cordierite in the back-scattered electron image [Fig. 3(a)] would (2: 2: 5) suggest that the averag omic number of the layer is higher than the average atomic number of the rest cortical compound"Mg- beryl"[15]. L, P, M, C, C(ss) elsewhere from SEM work on a SiC fibre-reinforced refer to liquid, protoenstatite, mullite, stoichiometric cordierite composite [26]. EDS analysis by SEM lierite, cordierite solid solution and silica respectively. revealed higher amounts of Al and Mg in the bright layer compared to the remaining fibre in composite stability of this phase at room temperature [6, 22-24]. A. further evidence for this local increase in concen It is interesting to note that the possibility of the tration of Al and Mg is obtained from the digital stabilisation of protoenstatite at room temperature compositional maps of these elements in Figs 6(a) by incorporation of a related silicate or a foreign and(b), respectively. A composition gradient is ob oxide into solid solution in the crystal lattice of served from the centre of the fibre to the matrix protoenstatite was suggested some time ago by Foster However, it should be noted that an apparent compo. [25] sitional gradient could also tocnstatite were the major crystalline phases in the interaction volume of X-rays at an interface between PE+SiO, +Liq. f Cordierite +Lig. a-b PE+Fo+Liq e Co+Mu+Liq. a Forsterite Mu+Sa+s Mullite Forsterite+ Spinel+Corund 0 2 Fig. 5. Isothermal(1460C)section of the AlO-Mgo- Sio, phase diagram [15]
2906 KUMAR and KNOWLES: Sic REINFORCED ALUMINOSILICATES-I L C(ss) + P + s C+P+S Mg’-beryl io $0 io Cokerite (2:2:5) Fig. 4. Pseudobinary section going through cordierite and the theoretical compound “Mg-beryl” [15]. L, P, M, C, C(ss) and S refer to liquid, protoenstatite. mullite, stoichiometric cordierite, cordierite solid solution and silica respectively. stability of this phase at room temperature [6,22-241. It is interesting to note that the possibility of the stabilisation of protoenstatite at room temperature by incorporation of a related silicate or a foreign oxide into solid solution in the crystal lattice of protoenstatite was suggested some time ago by Foster [25]. As we have already stated, cordierite and protoenstatite were the major crystalline phases in the matrix, but in addition the XRD pattern suggested that orthoenstatite existed in trace quantities. TEM observations showed the presence of a few twinned grains typical of clinoenstatite and orthoenstatite in the matrix. Therefore, while it is possible that protoenstatite remains stable at room temperature for the matrix composition used to fabricate composite B, further work is clearly required to be able to understand the mechanism by which it can happen. Hot-pressing of composite A at high temperature yields a pore-free microstructure because of increased viscous flow due to the formation of a liquid. However, hot-pressing at high temperature will also increase markedly the diffusion of matrix elements into the fibre. The bright layer around the fibres seen in the back-scattered electron image [Fig. 3(a)] would suggest that the average atomic number of the layer is higher than the average atomic number of the rest of the fibre. Similar layers have also been reported elsewhere from SEM work on a SIC fibre-reinforced cordierite composite [26]. EDS analysis by SEM revealed higher amounts of Al and Mg in the bright layer compared to the remaining fibre in composite A. Further evidence for this local increase in concentration of Al and Mg is obtained from the digital compositional maps of these elements in Figs 6(a) and (b), respectively. A composition gradient is observed from the centre of the fibre to the matrix. However, it should be noted that an apparent compositional gradient could also arise from overlapping of interaction volume of X-rays at an interface between SiO, Protoenstatite h- PE+Fo+Liq. e p/ Forsterite+ Periclase+Spinel A_ SiOp+Mu+Liq.g MgO 20 40 60 Spine1 80 A’203 Fig. 5. Isothermal (1460°C) section of the AlzO,-Mg&SiO, phase diagram [15]
KUMAR and KNOWLES: SiC REINFORCED ALUMINOSILICATES-I 2907 Al (c) Fig. 6.(a)X-ray Al map;(b)X-ray Mg map osite A; and (c)colour intensity scale: white and black correspond aximum and minimum x ray intensities respectively
KUMAR and KNOWLES: Sic REINFORCED ALUMINOSILICATES-I 2907 Fig. 6. (a) X-ray Al map; (b) X-ray Mg map in composite A; and (c) colour intensity scale: white al black correspond to maximum and minimum X-ray intensities respectively
yuE KUMAR and KNOWLES: SIC REINFORCED ALUMINOSILICATES-I 10 Fig. 7(a)X-ray Mg map;(b)X-ray Al map; (c)X-ray Si map; and (d)X-ray P map in composite B The colour intensity scale is the same as in Fig. 6
2908 KUMAR and KNOWLES: Sic REINFORCED ALUMINOSILICATES-I Fig. 7. (a) X-ray Mg map; (b) X-ray Al map; (c) X-ray Si map; and (d) X-ray P map in composite B. The colour intensity scale is the same as in Fig. 6
KUMAR and KNOWLES: SIC REINFORCED ALUMINOSILICATES-l 2909 Table 2. Comparison of X- ray spatial resolution at different electron beam energie Critical excitation energy Electron beam energy E (keV) 1.559 1.838 8666 o different materials. It is therefore important to be typically 10 2-104 poise[29, 30], the glass still have an idea of the extent of X-ray spread, i. e the has very high viscosity (10-10 2 poise) and there- spatial resolution of X-rays in the sample. This can fore the composites may not densify fully [ 8]. The be estimated from the formula given by Recd [27]: prcscncc of the dark laycr around the fibres [ Fig 3(b) might suggest the presence of lighter elements in 0.077 the layer. However, observations at different tilt dark layer was, in fact, due to surface relief caused where d is the spatial resolution (in um), p is the by debonding at the matrix-fibre interface during density of material (in gm/cm), Eo and Ec are the polishing beam energy and the critical ionisation energy in kev, Digital compositional maps of Mg, Al, Si and P in espectively. Estimated values of d are in a cross-section of composite B are shown in Figs Table 2. For these, the density was assumed to have 7(aH(d), respectively. A small circumferential layer a constant value of 2.60 gm/cm, a reasonable (<0.3 um, red in colour) is seen around the fibres assumption considering that the density of crystalline Since the thickness of this layer is below the resol- ordicritc is 2.50 gm/cmand that of mullite is ution limit of microprobe analysis in the prcscnt caso 23.0 gm/cm. It is evident from the above equation it is most likely that this"layer""arises from that d increases with an increase in E for a given e ping of interaction volume of X-rays generated at the nd p. Therefore, when analysing small microstruc- interface between the fibre and the matrix. This ral features of sizes in the range of 1-2 um, an would in turn suggest that there is no segregation of electron beam of low energy and a small probe size Al and Mg along the circumference of the fibres. It (small beam currents)should be used. Thus, a choice is interesting to note that some"hot-spots"of P [the of electron beam energy of 5 kv and beam current of white areas in Fig. 7(d)] are present in the matrix. The 4 nA is reasonable because the spatial resolution d, amount of P2O5 added in the matrix was 2.0 wt%.At emains below 0.30 um(the maximum value for the mass concentration levels of 1-3 wt%, the back- elements of interest). At a low electron beam energy, ground contributes as much intensity as the charac X-rays are mostly generated at the surface and there- teristic X-ray peak [31 which would explain the fore errors duc to X-ray fluorcscencc arc minimise. prcscnce of a high lcvcl of background noisc in thi We can therefore be confident in deducing that the map composition maps in Fig. 6 are real maps and n No phase containing either phosphorus or boron apparent maps down to at least 1 um resolution. It was identified in the Xrd trace of composite B. This follows that the bright regions in the Al map corre- is likely to be because of both the small amounts of pond to mullite crystals, consistent with the expected phosphorus- and bord morphology of mullite [Fig. 6(a). Segregation of Al and the resolution limit of the diffractometer used in and Mg in the hright layer around the fibres is evident this study. Metcalfe et al. [4] have reported the and consistent with the EDS analysis. Bleay and Scott formation of farringtonite [magnesium phosphate, [28] have observed a similar bright layer, in back-scat- Mg(PO4)] at low temperatures(below the liquidus ered SEM images, around the fibres in a Nicalon temperature of x 1357 C of the glass precursor they fibre-reinforced barium osumilite matrix composite, used) in SiC fibre-reinforced magnesium aluminosili which they attributed to the es having A 5 mol% of PzO,. It is ther fore possible that the P-rich regions seen in Fig. 7(d) The micro-porosity observed in composite B is correspond to magnesium phosphate. Although the likely to be because of hot-pressing at relatively low matrix was not analysed in detail in TEM, isolated temperatures. Although the composites were hot- pockets of glassy phase in the matrix were observed above the glass transition temperature, during which, while dependent on the rate of cooling, will TEM. Such an area of the matrix containing glassy
KUMAR and KNOWLES: Sic REINFORCED ALUMINOSILICATES-I 2909 Table 2. Comparison of X-ray spatial resolution at different electron beam energies Critical excitation energy Electron beam energy d Element E, (keV K, (ke’4 (PI Mg 1.303 20 2.60 15 1.68 10 1.23 5 0.29 Al 1.559 20 2.59 15 I .66 IO 0.88 5 0.27 Si 1.838 20 2.58 15 I .65 10 0.86 5 0.26 two different materials. It is therefore important to have an idea of the extent of X-ray spread, i.e. the spatial resolution of X-rays in the sample. This can be estimated from the formula given by Reed [27]: where d is the spatial resolution (in pm), p is the density of material (in gm/cm3), E, and EC are the beam energy and the critical ionisation energy in keV, respectively. Estimated values of d are given in Table 2. For these, the density was assumed to have a constant value of 2.60 gm/cm’, a reasonable assumption considering that the density of crystalline cordierite is x2.50 gm/cm3 and that of mullite is z 3.0 gm/cm3. It is evident from the above equation that d increases with an increase in E, for a given EC and p. Therefore, when analysing small microstructural features of sizes in the range of l-2 pm, an electron beam of low energy and a small probe size (small beam currents) should be used. Thus, a choice of electron beam energy of 5 kV and beam current of 4 nA is reasonable because the spatial resolution d, remains below 0.30 pm (the maximum value for the elements of interest). At a low electron beam energy, X-rays are mostly generated at the surface and therefore errors due to X-ray fluorescence are minimised. We can therefore be confident in deducing that the composition maps in Fig. 6 are real maps and not apparent maps down to at least 1 pm resolution. It follows that the bright regions in the Al map correspond to mullite crystals, consistent with the expected morphology of mullite [Fig. 6(a)]. Segregation of Al and Mg in the bright layer around the fibres is evident and consistent with the EDS analysis. Bleay and Scott [28] have observed a similar bright layer, in back-scattered SEM images, around the fibres in a Nicalon fibre-reinforced barium osumilite matrix composite, which they attributed to the presence of Ba in the layer. The micro-porosity observed in composite B is likely to be because of hot-pressing at relatively low temperatures. Although the composites were hotpressed above the glass transition temperature, which, while dependent on the rate of cooling, will be typically 10’2-10’4 poise [29,30], the glass still has very high viscosity (109-10’2 poise) and therefore the composites may not densify fully [8]. The presence of the dark layer around the fibres [Fig. 3(b)] might suggest the presence of lighter elements in the layer. However, observations at different tilt angles in the SEM revealed that the supposed dark layer was, in fact, due to surface relief caused by debonding at the matrix-fibre interface during polishing. Digital compositional maps of Mg, Al, Si and P in a cross-section of composite B are shown in Figs 7(a)-(d), respectively. A small circumferential layer (< 0.3 pm, red in colour) is seen around the fibres. Since the thickness of this layer is below the resolution limit of microprobe analysis in the present case, it is most likely that this “layer” arises from overlapping of interaction volume of X-rays generated at the interface between the fibre and the matrix. This would in turn suggest that there is no segregation of Al and Mg along the circumference of the fibres. It is interesting to note that some “hot-spots” of P [the white areas in Fig. 7(d)] are present in the matrix. The amount of P,O, added in the matrix was 2.0 wt%. At mass concentration levels of l-3 wt%, the background contributes as much intensity as the characteristic X-ray peak [31], which would explain the presence of a high level of background noise in this map. No phase containing either phosphorus or boron was identified in the XRD trace of composite B. This is likely to be because of both the small amounts of phosphorus- and boron-containing phases present and the resolution limit of the diffractometer used in this study. Metcalfe et al. [4] have reported the formation of ferringtonite [magnesium phosphate, Mg2(P04)3] at low temperatures (below the liquidus temperature of z 1357°C of the glass precursor they used) in Sic fibre-reinforced magnesium aluminosilicate matrices having z 5 mol% of P,O, It is therefore possible that the P-rich regions seen in Fig. 7(d) correspond to magnesium phosphate. Although the matrix was not analysed in detail in TEM. isolated pockets of glassy phase in the matrix were observed during examination of fibre-matrix interfaces in TEM. Such an area of the matrix containing glassy
2910 KUMAR and KNOWLES: SiC REINFORCED ALUMINOSILICATES--I pockets (some are indicated by arrows in Fig. 8)is Table 3. Chemical com ±2wt%)of shown in a bright field TEM micrograph in Fig. 8 glassy pockets Quantitative EDS analysis of these pockets revealed Glass Pocket Sio, Mgo P,O, AL, O, the presence of oxides of P, Mg, Si and Al. The weight 9.016.54.5 of these oxides varied from one glassy pocket to 4103272164 another. Examples of quantitative EDS analyses of from the eds data that some of the glassy pockets 3. 2. Interface analysis ontain a large amount of P, Os compared with the a bright field image of composite A is shown in urrounding matrix (no P was observed in grains Fig. 9(a)and selected area diffraction patterns surrounding the glassy pockels). The sizes of these (SADP)from the fibre and fibre- interlayer 1 pockets vary between a0 1 and 0.2 um while the shown in Figs 9(b)and (c), respectively. Three dis- hot-spots"in Fig. 7(d) have sizes in the range of tinct interphase layers are seen at the fibre-matrix 0-3.0 um. The sizes of the glassy pockets are below interface. The layer next to the fibre [layer I in Fig the resolution limit of the X-ray analysis and, there- 9(a)] is about 100 nm thick and the layer adjacent to fore, at first sight it is difficult to correlate the the matrix [layer 3 in Fig 9(a)] is about 20 nm thick phosphorus-rich glassy pockets seen in TEM with the Interfacial layers 1 and 3 are separated by a thin interphase layer [lay 9(a)] shosphorus. However, overlapping of the interaction appear brighter than the fibre, suggesting that these volumes of X-rays between two or more adjacent layers consist of light elements. Interlayer 3 appears glassy pockets will give rise to at least some of the brighter than interlayer 1. It was observed that layer maller " hot-spots"seen in the X-ray map of 2 was crystalline because it changed contrast when horu t the e Finally, the examination of the compositional width of the fibre-matrix interface regions varied maps in composite b suggested that the effects of between 120 and 140 nm from one region to another overlapping interaction volumes at the fibre-matrix within the same thin foil. interface under the present experimental conditions Three well-defined and reasonably sharp rings of are minimal. This is further confirmation that the p-Sic can be seen in the selected area diffraction composition gradients seen in compositional maps pattern[Fig 9(b)]of the fibre. This suggests that the f composite a are due to real variations in the fibre is microcrystalline. In addition to the three rings concentrations of the matrix elements across the of B-SiC, a very broad diffuse ring is present around fibre-matrix interfaces he central spot in the selected area diffraction pattern Fig 8. A bright field TEM image showing glassy pockets in matrix of composite B Sotne of thesc glassy ockets are arrowed
2910 KUMAR and KNOWLES: SIC REINFORCED ALUMINOSILICATES-I pockets (some are indicated by arrows in Fig. 8) is shown in a bright field TEM micrograph in Fig. 8. Quantitative EDS analysis of these pockets revealed the presence of oxides of P, Mg, Si and Al. The weight % of these oxides varied from one glassy pocket to another. Examples of quantitative EDS analyses of the glassy pockets are given in Table 3. It is obvious from the EDS data that some of the glassy pockets contain a large amount of P,05 compared with the surrounding matrix (no P was observed in grains surrounding the glassy pockets). The sizes of these pockets vary between ~0.1 and 0.2pm while the “hot-spots” in Fig. 7(d) have sizes in the range of l&3.0 pm. The sizes of the glassy pockets are below the resolution limit of the X-ray analysis and, therefore, at first sight it is difficult to correlate the phosphorus-rich glassy pockets seen in TEM with the presence of the “hot-spots” in the X-ray map of phosphorus. However, overlapping of the interaction volumes of X-rays between two or more adjacent glassy pockets will give rise to at least some of the smaller “hot-spots” seen in the X-ray map of phosphorus. Finally, the examination of the compositional maps in composite B suggested that the effects of overlapping interaction volumes at the fibre-matrix interface under the present experimental conditions are minimal. This is further confirmation that the composition gradients seen in compositional maps of composite A are due to real variations in the concentrations of the matrix elements across the fibre-matrix interfaces. Table 3. Chemical composition (within k2 wt%) of glassy pockets in the matrix of composite B Glass Pocket SO, MgO P,O, A’,O, GLI 11.6 48.6 31.5 8.3 CL2 79.0 16.5 4.5 0.0 CL3 41.0 32.7 21.6 4.7 3.2. Interface analysis A bright field image of composite A is shown in Fig. 9(a) and selected area diffraction patterns (SADP) from the fibre and fibre-interlayer 1 are shown in Figs 9(b) and (c), respectively. Three distinct interphase layers are seen at the fibre-matrix interface. The layer next to the fibre [layer 1 in Fig. 9(a)] is about 100 nm thick and the layer adjacent to the matrix [layer 3 in Fig. 9(a)] is about 20 nm thick. Interfacial layers 1 and 3 are separated by a thin darker interphase layer [layer 2 in Fig. 9(a)] and appear brighter than the fibre, suggesting that these layers consist of light elements. Interlayer 3 appears brighter than interlayer 1. It was observed that layer 2 was crystalline because it changed contrast when the sample was tilted. It should be noted that the width of the fibre-matrix interface regions varied between 120 and 140 nm from one region to another within the same thin foil. Three well-defined and reasonably sharp rings of /l-Sic can be seen in the selected area diffraction pattern [Fig. 9(b)] of the fibre. This suggests that the fibre is microcrystalline. In addition to the three rings of P-Sic, a very broad diffuse ring is present around the central spot in the selected area diffraction pattern Fig. 8. A bright field TEM image showing glassy pockets in matrix of composite B. Some of these glassy pockets are arrowed