Acta mater. ve Pergamon Copyright C 1996 Acta Met 09567151(95003878 Printed in Great Bri 1359-645496 MICROSTRUCTURE-PROPERTY RELATIONSHIPS OF SIC FIBRE-REINFORCED MAGNESIUM ALUMINOSILICATES--II. MECHANICAL PROPERTIES AND FAILURE CHARACTERISTICS A KUMAR+ and K M. KNOWLES University of Cambridge, Department of Materials Science and Metallurgy, P Cambridge CB2 3QZ, England Received 31 October 1994: in revised form 6 September 1995) Abstract--Interfacial frictional shear stresses, flexural properties and failure mechanisms are reported for two magnesium aluminosilicates unidirectionally reinforced with Nicalon SiC fibres Composites A and 00and920°c spcctivcly. High valucs of intcrfacial frictional shear stres inferred from Marshall,'s analysis of the micro-indentation technique could be attributed in part to presence of compressive radial stresses at the fibre-matrix interfaces. Although both comp symmetrical four point bend testing at roo different. Extensive matrix microcracking, fibre failure and then fibre pull-out were commonly observed in composite A. Failure modes in composite B included the formation of a limited number demonstrate that the mechanical properties, the interfacial frictional shear st mechanisms of both composites are governed by their microstructural features, in particular the and structure of the matrix-fibre interfacial region. Copyright c 1996 Acta Metallurgica ine 1 INTRODUCTION microstructural features have been described in Part I [2]. Estimates of the interfacial frictional It is now well established that the structure and shear stress have been made using Marshall's micro properties of the fibre-matrix interf nant role in controlling the mechanical properties ot the failure mechanisms of the composites have been comin ned using a symmetrical four-point bend tes matrix composites[I]. In Part I of this paper [2], it Finally, the properties and the fracture behaviour of was demonstrated that the structure and morphology these composites have been correlated with the of the interfacial layers in two different magnesium microstructure of the composites aluminosilicate glass-ceramic matrices unidirection ally reinforced with SiC fibres were quite complex and different. As a result of this, interpretation of the 2. MATERLALS AND EXPERIMENTAL mechanical properties and fracture process is not as 2.1 Materials straightforward as is often presumed Details of material and processing parameters Several attempts have been made in the past to have been given in Part I of this paper [2 ]. Briefly, estimate the mechanical properties of Sic fibre einforced glass-ceramics. In particular, the Sic/ composite A was hot-pressed at 1500oC,whereas lithium aluminosilicates (Las)and Sic/calci composite B was hot-pressed at 920.C. Composite B aluminosilicates(CAS)systems have received a lot of was ceramed in air at 1150 C for I h, whereas no attention [3-5]. Although the mechanical properties of SiC/magnesium aluminosilicate(MAS)composites posite A. Fibre volume fractions were 0.47 and 0.40 have been reported in the literature, attempts have in composites A and B, respectively not been made to correlate the properties with the microstructural fcaturcs of thc composites [69] 2.2. Experimental The first objective of this paper is to report the 2.2. 1. Microhardness and interfacial friction stres mechanical properties of the composites whose A Vickers diamond on a Leitz microhardness tester was used to indent the fibres for evaluating both the Department of Mechanica hardness and the interfacial friction stress. A load of raduate School. Monterey, CA 93943, 245, 25 mN was used to measure the hardness of the fibres in be mposites. The total indentati 2923
Pergamon 09%7151(95)00387-8 Acta mater. Vol. 44, No. I, Pp. 2923-2934. 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-II. MECHANICAL PROPERTIES AND FAILURE CHARACTERISTICS 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-Interfacial frictional shear stresses, flexural properties and failure mechanisms are reported for two magnesium aluminosilicates unidirectionally reinforced with Nicalon Sic fibres. Composites A and B were hot-pressed at 1500 and 92O”C, respectively. High values of interfacial frictional shear stresses inferred from Marshall’s analysis of the micro-indentation technique could be attributed in part to the presence of compressive radial stresses at the fibre-matrix interfaces. Although both composites failed non-catastrophically in symmetrical four point bend testing at room temperature, the failure modes were different. Extensive matrix microcracking, fibre failure and then fibre pull-out were commonly observed in composite A. Failure modes in composite B included the formation of a limited number of matrix cracks, the failure of fibres in the matrix crack front and progressive delamination. Our observations demonstrate that the mechanical properties, the interfacial frictional shear stresses and the failure mechanisms of both composites are governed by their microstructural features, in particular the chemistry and structure of the matrix-fibre interfacial region. Copyright 0 1996 Acta Metallurgica Inc. 1. INTRODUCTION It is now well established that the structure and properties of the fibre-matrix interface play a dominant role in controlling the mechanical properties of continuous fibre-reinforced glass and glass-ceramic matrix composites [l]. In Part I of this paper [2], it was demonstrated that the structure and morphology of the interfacial layers in two different magnesium aluminosilicate glass-ceramic matrices unidirectionally reinforced with SIC fibres were quite complex and different. As a result of this, interpretation of the mechanical properties and fracture process is not as straightforward as is often presumed. Several attempts have been made in the past to estimate the mechanical properties of SIC fibrereinforced glass-ceramics. In particular, the Sic/ lithium aluminosilicates (LAS) and Sic/calcium aluminosilicates (CAS) systems have received a lot of attention [3-51. Although the mechanical properties of Sic/magnesium aluminosilicate (MAS) composites have been reported in the literature, attempts have not been made to correlate the properties with the microstructural features of the composites [6-91. The first objective of this paper is to report the mechanical properties of the composites whose TPresent address: Department of Mechanical Engineering, Naval Postgraduate School, Monterey, CA 93943, U.S.A. microstructural features have been described in Part I [2]. Estimates of the interfacial frictional shear stress have been made using Marshall’s microindentation technique [lo]. Flexural properties and the failure mechanisms of the composites have been determined using a symmetrical four-point bend test. Finally, the properties and the fracture behaviour of these composites have been correlated with the microstructure of the composites. 2. MATERIALS AND EXPERIMENTAL 2.1. Materials Details of material and processing parameters have been given in Part I of this paper [2]. Briefly, composite A was hot-pressed at 15OO”C, whereas composite B was hot-pressed at 920°C. Composite B was ceramed in air at 1150°C for 1 h, whereas no post-processing heat-treatment was given to composite A. Fibre volume fractions were 0.47 and 0.40 in composites A and B, respectively. 2.2. Experimental 2.2.1. Microhardness and interfacial friction stress. A Vickers diamond on a Leitz microhardness tester was used to indent the fibres for evaluating both the hardness and the interfacial friction stress. A load of 245.25 mN was used to measure the hardness of the fibres in both composites. The total indentation time 2923
2924 KUMAR and KNOWLES: SIC REINFORCED ALUMINOSILICA'TES-II plates using a high speed diamond saw. The as-cut Indenter stub specially designed for grinding and polishing large composite samples. All four sides of the samples ere first polished with 6 um diamond paste and finally with I um diamond paste. The composite samples were 2 a tiff vo-hydraulic machine with a l kn load cell t investigate their fexural properties and their fracture 2 b behaviour. The tests were carried out with the machine under displacement control. The ramp-rate was 0.025 mm/ mples of dimensions approxi Matrix Matrix mately50×10×06 mm and50×5×3 mm were tes A and B, respectively. The Fig. 1. Schematic of the indentati separation between the inner loading points was (10D). The relevant parameters have been defined in the text. 16 mm for both composites, whereas the separation 10 and 12 mm for composites A and B, respectively. was 30s(15 s for load application and 15s for The exact span to depth ratio varied from sample to dwelling). Diagonals of the indentation impression sample because of small variations in the thicknesses on the fibres and the matrix were measured on the of the samples, but was in the range of 14-16 and 4-5 micrographs recorded using both a scanning electron for composites A and B, respectively microscope(SEM, Camscan-S2)and a Zeiss optical Load-deflection curves were plotted using an x The interfacial frictional shear stress was estimated mechanisms, the test was interrupted at different load sing Marshall's method [10]. The fibres were loaded levels and the samples were examined ex situ in an so that indentation impressions were seen in the optical microscope. A minimum of four samples was matrix(Fig. 1). The equations tested for both composites. The maximum nominal Hat Hexural stress and the elastic modulus were calculated (I) from the linear elastic theory of uniform beams Fracture surfaces were carefully cut and mounted derived by Marshall [10]were used to calcula ( 2) onto an aluminium stub for examination by SEM (b-a)cotψ Sample surfaces were gold coated to reduce specimen in the sem. Th interfacial frictional shear stress. In these equations, SEM was operated in secondary electron mode t is the interfacial frictional shear stress (assumed in this analysis to be constant ), 2a is the mean diagonal of the indentation on the fibre. h is the fibre hard- 3. RESULTS AND INTERPRETATION ess, u is the fibre displacement, 2b is the mean onal of the indentatio on on the matrix 3.1. Microhardness and interface friction stress a(a148 )is the angle between the opposite edges of The values of microhardness obtained using a the indenter, r is the fibre radius and Er is the elastic 245. 25 mn load were 19-24 and 26-32 GPa for modulus of the fibres composites A and B, respectively. An example of According to the suppliers specifications, the elastic micro-indentations used to calculate the micro- modulus of the Nicalon fibres (NL 202) used to hardness of fibres in both composites is shown for fabricate both composites was 184 GPa [11]. This composite B in Fig. 2(a), Debonding at the alue was therefore used to estimate the interfacial fibre-matrix interface was observed when a load frictional shear stresses via equation (1). However, it higher than 245.25 mn was used to indent the fibres should be noted that the values of the elastic modulus in both sites,The fibre hardness obtained of Nicalon fibres reported in the literature vary from experimentally in composite a is lower than that in 182 to 210 GPa [e.g. 12]and that authors often quote composite B and also the hardness in both a value of 200 GPa instead [10, 13]. The interface composites is higher than the typical Nicalon fibre friction stress in equation( 1)is inversely proportional hardness(13 GPa)available in the literature [3, 10 to the elastic modulus of the fibre and therefore a Bleay et al. [ 13] have recently reported a micro- relatively small value assumed for the elastic modulus hardness value of 8.9 GPa for Nicalon fibres, and of the fibres will produce a relatively high frictional although they did not mention the load required shear stress and vice-versa obtain this value. it can be inferred that the load used d fractography. was less than 0.5N, the load used in their work to Samples for flexural testing were cut from composite cause the fibres to slide within the matrix
2924 KUMAR and KNOWLES: SIC REINFORCED ALUMINOSILICATES-II Indenter I ! 2b h 2r ! I * Matrix’ Fibre Matrix Fig. 1. Schematic of the indentation method for measurement of matrix-fibre interfacial friction stress (after Ref. [lo]). The relevant parameters have been defined in the text. was 30 s (15 s for load application and 15 s for dwelling). Diagonals of the indentation impression on the fibres and the matrix were measured on the micrographs recorded using both a scanning electron microscope (SEM, CamscanS2) and a Zeiss optical microscope. The interfacial frictional shear stress was estimated using Marshall’s method [lo]. The fibres were loaded so that indentation impressions were seen in the matrix (Fig. 1). The equations and H2a4 z- n 2ur3E, (1) u =(b -a)cotti (2) derived by Marshall [lo] were used to calculate the interfacial frictional shear stress. In these equations, z is the interfacial frictional shear stress (assumed in this analysis to be constant), 2a is the mean diagonal of the indentation on the fibre, H is the fibre hardness, u is the fibre displacement, 26 is the mean diagonal of the indentation impression on the matrix, 2$(x 148”) is the angle between the opposite edges of the indenter, r is the fibre radius and Ef is the elastic modulus of the fibres. According to the suppliers specifications, the elastic modulus of the Nicalon fibres (NL 202) used to fabricate both composites was x 184 GPa [l 11. This value was therefore used to estimate the interfacial frictional shear stresses via equation (1). However, it should be noted that the values of the elastic modulus of Nicalon fibres reported in the literature vary from 182 to 210 GPa [e.g. 121 and that authors often quote a value of 200 GPa instead [lo, 131. The interface friction stress in equation (1) is inversely proportional to the elastic modulus of the fibre and therefore a relatively small value assumed for the elastic modulus of the fibres will produce a relatively high frictional shear stress and vice-versa. 2.2.2. Mechanical testing and fractography. Samples for flexural testing were cut from composite plates using a high speed diamond saw. The as-cut samples were mounted on a well polished aluminium stub specially designed for grinding and polishing large composite samples. All four sides of the samples were first ground with graded SIC grit papers, polished with 6pm diamond paste and finally with 1 pm diamond paste. The composite samples were tested in symmetrical four point bend on a stiff servo-hydraulic machine with a 1 kN load cell to investigate their flexural properties and their fracture behaviour. The tests were carried out with the machine under displacement control. The ramp-rate was 0.025 mm/min. Samples of dimensions approximately 50 x 10 x 0.6 mm and 50 x 5 x 3 mm were used for composites A and B, respectively. The separation between the inner loading points was 16mm for both composites, whereas the separation between the inner and the outer loading points was 10 and 12 mm for composites A and B, respectively. The exact span to depth ratio varied from sample to sample because of small variations in the thicknesses of the samples, but was in the range of 14-16 and 4-5 for composites A and B, respectively. Load-deflection curves were plotted using an x-y recorder. In order to study the damage initiation mechanisms, the test was interrupted at different load levels and the samples were examined ex situ in an optical microscope. A minimum of four samples was tested for both composites. The maximum nominal flexural stress and the elastic modulus were calculated from the linear elastic theory of uniform beams. Fracture surfaces were carefully cut and mounted onto an aluminium stub for examination by SEM. Sample surfaces were gold coated to reduce specimen charging and to enhance contrast in the SEM. The SEM was operated in secondary electron mode. 3. RESULTS AND INTERPRETATION 3.1. Microhardness and interface friction stress The values of microhardness obtained using a 245.25 mN load were 19-24 and 2632GPa for composites A and B, respectively. An example of micro-indentations used to calculate the microhardness of fibres in both composites is shown for composite B in Fig. 2(a). Debonding at the fibre-matrix interface was observed when a load higher than 245.25 mN was used to indent the fibres in both composites. The fibre hardness obtained experimentally in composite A is lower than that in composite B and also the hardness in both composites is higher than the typical Nicalon fibre hardness (13 GPa) available in the literature [3, lo]. Bleay et al. [13] have recently reported a microhardness value of 8.9 GPa for Nicalon fibres, and although they did not mention the load required to obtain this value, it can be inferred that the load used was less than 0.5 N, the load used in their work to cause the fibres to slide within the matrix
KUMAR and KNOWLES: SiC REINFORCED ALUMINOSILICATES--II 2925 It is likely that the low microhardness of fibres in Values of ise indices for a number of crystalline composite a compared with composite B is due to the ceramics have been given by Sargent [16] and an large-scale diffusion of matrix elements into the fibres explanation of the effect in terms of a mixed elastic- in composite A [2], producing a softening effect which plastic materials deformation response has been given composite B. The size of the Sic grains in the si co by Bull et al.[17]. An ISE index, n, for Nicalon fibres ble. but if Nicalon fibres is on a nanometre level [2], approxi data collected by Sargent for hot-pressed SiC, where ately three orders of magnitude smaller than the n 1.7, can be considered to be a first approximation mean diagonal of the indentations used for estimating to what might be expected for the fibres, a simple the fibre hardness, and so a useful analogy is with the calculation combining the equations for Vickers hardness behaviour of glasses function of hardness with the ISE force-diagonal power law [14] network modifiers [14]. Although the hardness of shows that hardness values H, and H2 measured at crystalline ceramics increases generally with a loads Fi and Fx. respectively, are related through the decrease in grain size because dislocations generated equation by the indenter are blocked by the grain boundaries [5, this effect is clearly not relevant in the present case because of the grain sizes involved de and so if F2/F=4, H,/HR0.784 for n=1.7. Thu discrepancies reported for the experimentally ifferent laboratories were to have used loads measured hardness values of Nicalon fibres: (i) an significantly higher than the ones we have used, some intrinsic variation of hardness between the different of the discrepancies could be explained at least in part batches of fibres; (ii) an indentation-size effect; and by an ISE. However, this is not in accord with the iii) systenatic experimental errors. Neither (i) nor lvads used for measuring fibre hardness reported in ii)can account for the differences we have observed the literature ween the fibres in composites a and b because the It is therefore more likely that a major reason hardness values were taken at a single load and any for the discrepancies in the hardness values of systematic errors inherent in the measurement will be Nicalon fibres lies in the calibration and use of the same for both batches different equipment by various researchers generating When comparing results from different labora- systematic measurement errors. To examine this tories, it is important to recognise that in general the possibility, the hardness of fibres in composite b was microhardness of ceramics is load dependent(known measured on indentation equipment at the School of as the indentation-size effect, ISE)and that in general Materials Science, University of Bath, Claverton it increases with a decrease in applied load [14]. Down, England. In order to obtain a direct compari- a 15m ig. 2. Nomarski interference contrast images showing (a, b) indentations obtained Microhardness tester il cUinpusite B and (b) indentations (indicated b microhardness tester in composite B
KUMAR and KNOWLES: Sic REINFORCED ALUMINOSILICATES-II 2925 It is likely that the low microhardness of fibres in composite A compared with composite B is due to the large-scale diffusion of matrix elements into the fibres in composite A [2], producing a softening effect which reduces the hardness compared with the fibres in composite B. The size of the SIC grains in the SIC-0 Nicalon fibres is on a nanometre level [2], approximately three orders of magnitude smaller than the mean diagonal of the indentations used for estimating the fibre hardness, and so a useful analogy is with the hardness behaviour of glasses as a function of network modifiers [14]. Although the hardness of crystalline ceramics increases generally with a decrease in grain size because dislocations generated by the indenter are blocked by the grain boundaries [15], this effect is clearly not relevant in the present case because of the grain sizes involved. There are three possible reasons for the wide discrepancies reported for the experimentally measured hardness values of Nicalon fibres: (i) an intrinsic variation of hardness between the different batches of fibres; (ii) an indentation-size effect; and (iii) systematic experimental errors. Neither (ii) nor (iii) can account for the differences we have observed between the fibres in composites A and B, because the hardness values were taken at a single load and any systematic errors inherent in the measurement will be the same for both batches. When comparing results from different laboratories, it is important to recognise that in general the microhardness of ceramics is load dependent (known as the indentation-size effect, ISE) and that in general it increases with a decrease in applied load [14]. Values of ISE indices for a number of crystalline :eramics have been given by Sargent [16] and an :xplanation of the effect in terms of a mixed elasticplastic materials deformation response has been given by Bull et al. [17]. An ISE index, n, for Nicalon fibres IS, to the best of our knowledge, not available, but if data collected by Sargent for hot-pressed SIC, where ‘I z 1.7, can be considered to be a first approximation to what might be expected for the fibres, a simple :alculation combining the equations for Vickers hardness with the ISE force-diagonal power law [14] shows that hardness values H, and H, measured at loads F, and F2, respectively, are related through the equation n-2 H, F2 T -=(-> HI F, ’ and so if F,/F, = 4, H,IH, e 0.784 for n = 1.7. Thus, if different laboratories were to have used loads significantly higher than the ones we have used, some of the discrepancies could be explained at least in part by an ISE. However, this is not in accord with the loads used for measuring fibre hardness reported in the literature [lo, 131. It is therefore more likely that a major reason for the discrepancies in the hardness values of Nicalon fibres lies in the calibration and use of different equipment by various researchers generating systematic measurement errors. To examine this possibility, the hardness of fibres in composite B was measured on indentation equipment at the School of Materials Science, University of Bath, Claverton Down, England. In order to obtain a direct compariFig. 2. Nomarski interference contrast images showing (a, b) indentations obtained on the Leitz microhardness tester in composite B and (b) indentations (indicated by arrows) obtained on the Leco microhardness tester in composite B
KUMAR and KNOWLES: SIC KEINFURCED ALUMINOSILICA'TES-II 10 um Fig. 3. No ashed in son with our own measurements the hardness of the were used to calculate these values of interfacial fibres was measured on the optical microscope on frictional shear stresses and the assumption has been their Leco (M-400) microhardness tester (as in the made that there is no indentation size effect in esults quoted by the Bath group [18) using a vickers Nicalon SiC fibres, as other workers in this area have diamond and an indentation load of 245. 25 mN. a assumed implicitly hardness valuc of 10-13 GPa was obtained. these It was observed that most of the fibres at or near hardness values are considerably lower than the the edge of the samples of composite B did not slide values obtained using our own Leitz microhardness in the matrix at 981 mN, suggesting higher interfacial ster. The indentations made on the two different frictional shear stresses at or near the edges in ieces of equipment are shown in Fig. 2. The comparison with the bulk of the sample. Values of indentations made on the Leco arrowed in Fig. 2(b) interfacial frictional shear stresses of 136-158 MPa are representative of those reported elsewhere by the were obtained from those fibres which could b Bath group [19] and are used both for hardness pushed-in at or near the edge of the samples. These measurements and in the relevant Marshall equation apparently high shear stresses can arise because of for interfacial frictional shear stress measurements. It localised oxidation of surface layers during ceraming should be noted that these indentations are not sharp of composite B in air. Similar trends in results have and that the diagonals are longer than the diagonals been reported for a Sic/barium osumilite composite btaincd on the Leitz using the same load morcover by Bleay and Scott [201 the values of diagonals measured on the optical The interfacial frictional shear stresses in these nicroscope on the Leco equipment equipped with an composites are considerably higher than those eye-piece were always higher than the values reported for SiC/LAS composites [10, 21-23]. This measured in a SEM, leading to an underestimate of can be attributed at least in part to the differences in fibre hardness. Thus, it is clear thal, at least for this residual thermal stresses at the fibre-matrix articular comparison, the considerable differences interfaces. Residual thermal stresses arise from the reported in the apparent fibre hardness on the Leitz thermal contraction mismatch between matrix and and the leco equipment are due purely to the system fibre, and also from the unrelaxed volume changes atic errors generated when introducing and then associated with any phase transformation an measuring indentation diagonals using different crystallisation in the matrix, Brun and Singh [22] equipment. have shown that the sliding friction stress is nearly A total load on both the fibre and matrix of zero when the coefficient of thermal expansion of 981 mN was sufficient to push-in all the fibres in both fibre(ar)is greater than the coefficient of thermal composites. A fibre pushed into the matrix in expansion of matrix (am )and that it increases composite A is shown in Fig. 3. Interfacial frictional linearly with the thermal expansion mismatch when shear stresses estimated using equation( 1)were 24-33 a< and 49 71 MPa in composites A and B, respectively. The coefficient of thermal expansion of LAS is The values of fibre hardness measured on the Leitz smaller than that of Nicalon fibres and therefore
2926 KUMAR and KNOWLES: Sic REINFORCED ALUMINOSILICATES-II Fig. 3. Nomarski interference contrast image of a fibre pushed in the matrix in composite A. son with our own measurements, the hardness of the fibres was measured on the optical microscope on their Leco (M-400) microhardness tester (as in the results quoted by the Bath group [18]) using a Vickers diamond and an indentation load of 245.25 mN. A hardness value of l&13 GPa was obtained. These hardness values are considerably lower than the values obtained using our own Leitz microhardness tester. The indentations made on the two different pieces of equipment are shown in Fig. 2. The indentations made on the Leco arrowed in Fig. 2(b) are representative of those reported elsewhere by the Bath group [19] and are used both for hardness measurements and in the relevant Marshall equation for interfacial frictional shear stress measurements. It should be noted that these indentations are not sharp and that the diagonals are longer than the diagonals obtained on the Leitz using the same load. Moreover, the values of diagonals measured on the optical microscope on the Leco equipment equipped with an eye-piece were always higher than the values measured in a SEM, leading to an underestimate of fibre hardness. Thus, it is clear that, at least for this particular comparison, the considerable differences reported in the apparent fibre hardness on the Leitz and the Leco equipment are due purely to the systematic errors generated when introducing and then measuring indentation diagonals using different equipment. A total load on both the fibre and matrix of 981 mN was sufficient to push-in all the fibres in both composites. A fibre pushed into the matrix in composite A is shown in Fig. 3. Interfacial frictional shear stresses estimated using equation (1) were 2433 and 49-71 MPa in composites A and B, respectively. The values of fibre hardness measured on the Leitz were used to calculate these values of interfacial frictional shear stresses and the assumption has been made that there is no indentation size effect in Nicalon Sic fibres, as other workers in this area have assumed implicitly. It was observed that most of the fibres at or near the edge of the samples of composite B did not slide in the matrix at 98 1 mN, suggesting higher interfacial frictional shear stresses at or near the edges in comparison with the bulk of the sample. Values of interfacial frictional shear stresses of 136-l 58 MPa were obtained from those fibres which could be pushed-in at or near the edge of the samples. These apparently high shear stresses can arise because of localised oxidation of surface layers during ceraming of composite B in air. Similar trends in results have been reported for a Sic/barium osumilite composite by Bleay and Scott [20]. The interfacial frictional shear stresses in these composites are considerably higher than those reported for Sic/LAS composites [lo, 21-231. This can be attributed at least in part to the differences in residual thermal stresses at the fibre-matrix interfaces. Residual thermal stresses arise from the thermal contraction mismatch between matrix and fibre, and also from the unrelaxed volume changes associated with any phase transformation and crystallisation in the matrix. Brun and Singh [22] have shown that the sliding friction stress is nearly zero when the coefficient of thermal expansion of fibre (a,) is greater than the coefficient of thermal expansion of matrix (a,) and that it increases linearly with the thermal expansion mismatch when tlf< CI,. The coefficient of thermal expansion of LAS is smaller than that of Nicalon fibres and therefore
KUMAR and KNOWLES: SIC REINHORCED ALUMINOSILICA'TES--Il able 1. Assumed materials parameters and estimated values of residual thermal stresses in composites A and B Composite CC MPa)(MPa ratio, linear thermal exp coefticient and elastic modulus, respectively. The f and m reter to the fibre and matrix respectively. (e.g. the glass transition temperature) and room temperature, m is the res g s moduli of the mullite and glass [26]. whereas the value of a, for composite R is ei residual thermal radial tensile stresses will be present where at the fibre-matrix interface in the Nicalon-LAS composites, consistent with low interfacial frictional Em ve (4) shear stresses. In contrast. the coefficients of thermal expansion of the matrices in both the composites that In these equations, vr and vm are the Poisson's ratios compressive stresses will be present at the r is the radius of the fibres, u the coefficient of fric fibre-matrix interface. It should be noted that if at the fibre matrix interface arro the residual radial either of the composites were to have had only stress at the fibre-matrix interface, and to is a con- a-cordierite in the matrix, there would have been stant sliding resistance term justified by Weihs and residual tensile stresses at the fibre-matrix interface, Nix on the basis of remnant fibre surface roughness rather than compressive stresses, because the co- after fibre-matrix debonding cfficient of thermal expansion of -cordierite is If we assume that the logarithm term in equation smaller than that of Nicalon fibres. As explained in (3)is of the form In(1+x)for small x, equation(3) Part I (2 the matrices in both the composites consist can be rewritten in the form of a-cordierite and other phases. Phases with high coefficients of thermal expansion, such as mullite in (1-2vk composite A and enstatite in composite B, help increase the effective thermal expansion coefficient of The term 2vk is small in comparison with I as it is the matrix in both cases of the order of v?(which is 0.0225 using the value of An estimation of an upper limit to the residual 0.15 for vr quoted for Nicalon fibres [21, 23). If we thermal stresses arising solely from the thermal follow Marshall [10] and let F-2a2H then to a good expansion mismatch can be made using the classical approximation Lame solution of a coaxial fibre and surrounding ix [24, 25]. The relevant cquations are given in H the appendix. The residual stresses estimated assum 1. The residual radial stresses in composites A and b replacing t. Coefficients of friction, H, assumed for at the fibre-matrix interfaces are compressive and SiC/LAS composites, in which there is good evidence comparable in magnitude, .-28 and -27MPa, of a carbon layer at the fibre-matrix interface, are respectively.If,instead of Marshall's simple model quoted in the range 0.01-0.32[21-23). If we therefore [10], the more sophisticated model of the push assume for SiC/MAS that a value of u=0. 2 is not technique of Weihs and Nix [23] is used, in which unreasonable in lieu of any firm experimental data, account is taken not only of a constant sliding then a 6 MPa contribution to any"constant "inter- resistance term, but also of the residual stress arising facial frictional shear stress would arise from the from thermal expansion mismatch and a Poisson on effect, then the displacement of the top of presence of thermal stresses on these calculations and would be independent of errors arising from, in the fibre,u, for a given load F on the fibre takes the particular, the measurement of fibre hardness form The values of interfacial frictional shear stresses calculated in equation()are very sensitive to the F22n:(0+m) values of fibre hardness and any ISE effect, if present Thus, if a hardness value of 10-13 GPa were to have een used instead of 26-32 GPa for the fibres in (3) compositc B in cquation(1), the inter facial frictional hear stresses would be reduced to 8-13 MPa before
KUMAR and KNOWLES: Sic REINFORCED ALUMINOSILICATES-II 2921 Table 1, Assumed materials parameters and estimated values of residual thermal stresses in composites A and B Comuosite Y_ Yr a, x 10-s AT Em Er Qw fl,,m (‘C’) (K) (GPaj (GPa) tMPa) (MPa) A 0.47 0.20 0.15 3.1 4.0 800 110 184 56 -28 B 0.40 0.20 0.15 3.1 4.1 800 80 184 45 -27 Y. G( and E refer to Poisson’s ratio. linear thermal expansion coefficient and elastic modulus, respectively. The subscripts, f and m refer to the fibre and matrix respectively. Vr is the volume fraction of the fibres. AT is the difference in temperature between the temperature at which the composite can be assumed to be stress-free (e.g. the glass transition temperature) and room temperature. o,,, is the residual axial stress in the matrix and Q,,, is the residual radial stress in the matrix at the fibre-matrix interface in a Lam& coaxial cylinder model of the composite [24]. The value of a, for composite A is based on the known linear thermal coefficients of Mg0_Al,O,-SiO, glasses and mullite. the relative volume fractions of mullite and glass and reasonable values for the Young’s moduli of the mullite and glass [26], whereas the value of OL, for composite B is given by the suppliers [I l] residual thermal radial tensile stresses will be present at the fibre-matrix interface in the Nicalon-LAS composites, consistent with low interfacial frictional shear stresses. In contrast, the coefficients of thermal expansion of the matrices in both the composites that we have examined here are higher than that of Nicalon fibres and, therefore, residual radial compressive stresses will be present at the fibre-matrix interface. It should be noted that if either of the composites were to have had only a-cordierite in the matrix, there would have been residual tensile stresses at the fibre-matrix interface, rather than compressive stresses, because the coefficient of thermal expansion of a-cordierite is smaller than that of Nicalon fibres. As explained in Part I [2], the matrices in both the composites consist of cc-cordierite and other phases. Phases with high coefficients of thermal expansion, such as mullite in composite A and enstatite in composite B, help to increase the effective thermal expansion coefficient of the matrix in both cases. An estimation of an upper limit to the residual thermal stresses arising solely from the thermal expansion mismatch can be made using the classical Lame solution of a coaxial fibre and surrounding matrix [24,25]. The relevant equations are given in the Appendix. The residual stresses estimated assuming the various material parameters are given in Table 1. The residual radial stresses in composites A and B at the fibre-matrix interfaces are compressive and comparable in magnitude, -28 and -27 MPa, respectively. If, instead of Marshall’s simple model [lo], the more sophisticated model of the push-down technique of Weihs and Nix [23] is used, in which account is taken not only of a constant sliding resistance term, but also of the residual stress arising from thermal expansion mismatch and a Poisson expansion effect, then the displacement of the top of the fibre, u, for a given load F on the fibre takes the form (1 - 2vrk) F tl= ~ 4 - r t%l+ P%,,) 2pkzr 2p2k’ xln 1 @F ’ +(z,+~(T,,,) 7cr2 II where k = Emvi -w + %I) In these equations, vr and v, are the Poisson’s ratios for the fibre and matrix respectively, E, and Em are the Young’s moduli of the fibre and matrix respectively, Y is the radius of the fibres, p the coefficient of friction at the fibre-matrix interface, orro the residual radial stress at the fibre-matrix interface, and z, is a constant sliding resistance term justified by Weihs and Nix on the basis of remnant fibre surface roughness after fibre-matrix debonding. If we assume that the logarithm term in equation (3) is of the form ln(1 + x) for small x, equation (3) can be rewritten in the form (1 - 2v,k) F= 24% Ef 4712r3(7, + pi,,,)’ (5) The term 2v,k is small in comparison with 1 as it is of the order of vf (which is 0.0225 using the value of 0.15 for vr quoted for Nicalon fibres [21,23]). If we follow Marshall [lo] and let F = 2&H, then to a good approximation, H2a4 70 + w-Jr,, =-. n”ur3Ef This is the same as equation (l), but with 7, + purr0 replacing 7. Coefficients of friction, p, assumed for Sic/LAS composites, in which there is good evidence of a carbon layer at the fibre-matrix interface, are quoted in the range 0.014.32 [21-231. If we therefore assume for SiC/MAS that a value of p = 0.2 is not unreasonable in lieu of any firm experimental data, then a 6 MPa contribution to any “constant” interfacial frictional shear stress would arise from the presence of thermal stresses on these calculations and would be independent of errors arising from, in particular, the measurement of fibre hardness. The values of interfacial frictional shear stresses calculated in equation (1) are very sensitive to the values of fibre hardness and any ISE effect, if present. Thus, if a hardness value of lo-13 GPa were to have been used instead of 2632 GPa for the fibres in composite B in equation (1) the interfacial frictional shear stresses would be reduced to 8-13 MPa before
KUMAR and KNOWLES: SiC REINFORCED ALUMINOSILICATES--II Table 2. Flexural properties of composite matrix in composite B also contained enstatite, the elastic modulus of which could not be found in the Elastic modulus literature The elastic modulus of the matrix in composite B calculated by the rule of mixtures is much lower than the elastic modulus of cordierite reported in the literature. It was demonstrated in Part I [2] tha alowance was made for the effect of any residual composite B contained micro-pores in the matrix due compressive stresses at the fibre-matrix interface. to hot-pressing at a low temperature In addition to This reinforces the doubts that can arise about this. some surface delamination cracks were also confidence in the absolute magnitudes of interfacial present in composite B. It is reasonable to assume frictional shear stresses quoted in the literature, and that these cracks arise from poor interlaminar hile obviously we have conlidence in our own bonding in these composites. The elastic modulus of measurements and in the observations of the trends monolithic ceramics decreases with an increase in in interfacial frictional shear stress that we have porosity [15, 28]. It is therefore most likely that the measured, the above analysis shows that we should presence of pores and cracks reduces the elastic of interfacial shear stresses in terms of absolute modulus of the matrix(see also Ref. n9 e elastic also quite properly be cautious in the interpretation modulus of composite B by decreasing the Typical load-displacement curves for composites 3. 2. fracture characteristics A and B are shown in Figs 4(a) and(b), respectively In both composites, the load increases linearly with The flexural properties of both composites are displacement at first, and this is then followed hy a given in Table 2. The average maximum nominal non-linear regime. The load drops after reaching a flexural stress (as inferred from the linear elastic peak value. The characteristic difference in the theory of uniform beams) and the elastic modulus of fracture behaviour of the composites is illustrated by composite A are higher than those of composite B. the variation in the load after the first load drop In This is, in part, a direct consequence of the higher composite A, the load continues to drop with very fibre volume fraction of composite A compared to small or no build up of load, whereas the load builds composite B. Applying the rule of mixtures, the up considerably before falling again in composite B values of the elastic moduli of the matrices lie in the Thus, the damage initiation process(such as matrix range of 111-130 and 40-50 GPa (assuming the cracking or failure in compression)may be the same fibre modulus is in the range 180-200 GPat) for in both composites, but the mechanisms leading to composites A and B, respectively. These values of the the final failure are different clastic modulus of matrix in composite A are slightly Observations during flexural testing of composit lower than that of mullite(E x 145 GPa, [27D. This A indicated that the first damage always occurred is because an appreciable amount of residual glass either in tension or in compression but never in and small amounts of silica and cordierite were shear. It was difficult to judge whether the damage in present in the matrix [2], and the elastic moduli of compression occurred after the matrix cracking in these phases are generally lower than that of mullite t tension or without any damage on the tension side The elastic modulus of the composites depends on If the first damage process involves matrix cracking elastic moduli of both the fibres and the matrix. in the tension side of the flexural beam, subsequent iven that composite B was hot-pressed at a much failure in compression will be possible because the lower temperature than composite A, the elastic actual stresses on the compression side will be modulus of the fibres in the former is not likely to be higher than those on the tension side(e.g. [3], [30D) lower than that in the latter. The elastic modulus of Fractographic examination of the samples showed cordierite (61-117 GPa)5 is lower than the elastic extensive matrix cracking on the tension surface of modulus of mullite, and therefore, to a first approxi- the flexural beam(Fig. 5). Also seen in Fig. 5 are fibres bridging lower than that of composite A. However, the that the fibres have failed away from the plane of the matrix crack. This clearly demonstrates the phenomenon of interface debonding and sliding taccording to the suppliers specifications, the elastic modu- the crack wake leading to fibre fracture Pa [I11 tThe elastic modulus of Mgo-Al2O,, glasses varies The final failure of the composite occurs ositions propagation of the matrix cracks on the tension between 103 and i11 GPa for a numbes ic c omp,3 GPa through the thickness and/or by buckling of fibres on c[26 the compression side. The fractograph in Fig. 6 shows SThe elastic modulus of cordierite glass-ceramics will dc- that the composite sample has failed both d on the type of nucleating agents. The values quote here are taken from Ref [27 ] Although the reported compression and tension. Fibre buckling is seen on rystalline phase was orthorhombic cordierite, the com- the compression side and the fibres pulled-out from position of the parent gla3s-ccramic was not mentioned. the matrix are seen on the tension side of the sample
2928 KUMAR and KNOWLES: SIC REINFORCED ALUMINOSILICATES-II Table 2. Flexural properties of composites A and B Comoosite Elastic modulus (GPa) Nominal ultimate flexural stress WPa) A 153+ 19 590* 18 R 104+4 415+69 allowance was made for the effect of any residual compressive stresses at the fibre-matrix interface. This reinforces the doubts that can arise about confidence in the absolute magnitudes of interfacial frictional shear stresses quoted in the literature, and while obviously we have confidence in our own measurements and in the observations of the trends in interfacial frictional shear stress that we have measured, the above analysis shows that we should also quite properly be cautious in the interpretation of interfacial shear stresses in terms of absolute magnitudes. 3.2. Fracture characteristics The flexural properties of both composites are given in Table 2. The average maximum nominal flexural stress (as inferred from the linear elastic theory of uniform beams) and the elastic modulus of composite A are higher than those of composite B. This is, in part, a direct consequence of the higher fibre volume fraction of composite A compared to composite B. Applying the rule of mixtures, the values of the elastic moduli of the matrices lie in the range of 111-130 and 4&50GPa (assuming the fibre modulus is in the range 180-200 GPat) for composites A and B, respectively. These values of the elastic modulus of matrix in composite A are slightly lower than that of mullite (E z 145 GPa, [27]). This is because an appreciable amount of residual glass and small amounts of silica and cordierite were present in the matrix [2], and the elastic moduli of these phases are generally lower than that of mul1ite.J The elastic modulus of the composites depends on elastic moduli of both the fibres and the matrix. Given that composite B was hot-pressed at a much lower temperature than composite A, the elastic modulus of the fibres in the former is not likely to be lower than that in the latter. The elastic modulus of cordierite (z 61-l 17 GPa)§ is lower than the elastic modulus of mullite, and therefore, to a first approximation, the elastic modulus of composite B will be lower than that of composite A. However, the TAccording to the suppliers specifications, the elastic modulus of Nicalon NL 202 fibres is z 184 GPa [l 11. fThe elastic modulus of Mg0-A&O,-Si02 glasses varies between 103 and 111 GPa for a number of compositions [26]. The elastic modulus of vitreous silica is ~73 GPa at room temperature [26]. §The elastic modulus of cordierite glass-ceramics will depend on the type of nucleating agents. The values quoted here are taken from Ref. [27]. Although the reported crystalline phase was orthorhombic cordierite, the composition of the parent glass-ceramic was not mentioned. matrix in composite B also contained enstatite, the elastic modulus of which could not be found in the literature. The elastic modulus of the matrix in composite B calculated by the rule of mixtures is much lower than the elastic modulus of cordierite reported in the literature. It was demonstrated in Part I [2] that composite B contained micro-pores in the matrix due to hot-pressing at a low temperature. In addition to this, some surface delamination cracks were also present in composite B. It is reasonable to assume that these cracks arise from poor interlaminar bonding in these composites. The elastic modulus of monolithic ceramics decreases with an increase in porosity [15,28]. It is therefore most likely that the presence of pores and cracks reduces the elastic modulus of composite B by decreasing the elastic modulus of the matrix (see also Ref. [29]). Typical load-displacement curves for composites A and B are shown in Figs 4(a) and (b), respectively. In both composites, the load increases linearly with displacement at first, and this is then followed by a non-linear regime. The load drops after reaching a peak value. The characteristic difference in the fracture behaviour of the composites is illustrated by the variation in the load after the first load drop. In composite A, the load continues to drop with very small or no build up of load, whereas the load builds up considerably before falling again in composite B. Thus, the damage initiation process (such as matrix cracking or failure in compression) may be the same in both composites, but the mechanisms leading to the final failure are different. Observations during flexural testing of composite A indicated that the first damage always occurred either in tension or in compression, but never in shear. It was difficult to judge whether the damage in compression occurred after the matrix cracking in tension or without any damage on the tension side. If the first damage process involves matrix cracking in the tension side of the flexural beam, subsequent failure in compression will be possible because the actual stresses on the compression side will be higher than those on the tension side (e.g. [3], [30]). Fractographic examination of the samples showed extensive matrix cracking on the tension surface of the flexural beam (Fig. 5). Also seen in Fig. 5 are fibres bridging an opened matrix crack. It can be seen that the fibres have failed away from the plane of the matrix crack. This clearly demonstrates the phenomenon of interface debonding and sliding in the crack wake leading to fibre fracture. The final failure of the composite occurs by propagation of the matrix cracks on the tension side through the thickness and/or by buckling of fibres on the compression side. The fractograph in Fig. 6 shows that the composite sample has failed both in compression and tension. Fibre buckling is seen on the compression side and the fibres pulled-out from the matrix are seen on the tension side of the sample
KUMAR and KNOWLES: SiC REINFORCED ALUMINOSILICATES-II (b) Stress (MPa)200 Stress (MPa) D Fig. 4. Typical load-deflection plots of (a te indicated based on the linear elastic theory of uniform beams Fibre pull-ol s commonly observed when the away"during testi amples were finally separated by hand(Fig. 7). SEM observed on the observations revealed smooth fibre surfaces and This may give rise to apparently long fibre pull-out imprints of fibres on the matrix, suggesting weak lengths bonding between the matrix and the fibre. There are A systematic study of the fracture process in matrix regions that are still attached to fibres(Fig. 7). composite B was carried out by interrupting the test Therefore, it is possible that the matrix was"blown at different loads and examining the samples in an Fig. 5. Fractograph showing extensive matrix cracking and fibres bridging an opened matrix crack in composite A
KUMAR and KNOWLES: SIC REINFORCED ALUMINOSILICATES-II 2929 (a) 250 Stress (MPa) 200’ 150 100 50 1 I I I I 2 3 Displacement (mm) (b) Stress (MPa) 2”” 100 0.25 nsn 0.75 1.0 Displacement (mm) 1.25 Fig. 4. Typical load-deflection plots of (a) composite A and (b) composite B. Values of nominal stress are indicated based on the linear elastic theory of uniform beams. Fibre pull-out was commonly observed when the away” during testing as was suggested by matrix dust samples were finally separated by hand (Fig. 7). SEM observed on the flexural testing rig after each test. observations revealed smooth fibre surfaces and This may give rise to apparently long fibre pull-out imprints of fibres on the matrix, suggesting weak lengths. bonding between the matrix and the fibre. There are A systematic study of the fracture process in matrix regions that are still attached to fibres (Fig. 7). composite B was carried out by interrupting the test is possible that the matrix was “blown at different loads and examining the samples in an Therefore, it Fig. 5. Fractograph showing extensive matrix cracking and fibres bridging an opened matrix crack in composite A
KUMAR and KNoWLEs: SiC REINFORCED ALUMINOSILICATES--lI Compression sid 1000y門 Fig. 6. Fractograph showing both compression and tension failures in composite A. Fibre buckling is also optical microscope. It was observed that the first crossed over the fibres because of misalignment of the damage always occurred on the tensile side of fibres [31] the beam. However, only a few matrix cracks, The load built up again as the delamination cracks perpendicular to the loading plane, were observed on propagated. The load then dropped abruptly the tension side when the test was stopped at a load suggesting either the failure of bridging fibres and or corresponding to the onset of non-linear deflection in the initiation of further delamination cracks(Fig. 8) the load-deflection curve. It is possible that some in a different laminate. The process of load build-up matrix cracks close upon unloading, but observations and drop continued. In some samples, damage on fter loading beyond the onset of non- linear the compression side was also observed at higher deflection did not reveal any increase in the crack loads. In addition to these failure processes, crushing population. No damage on the compression side of below the loading rollers and shear failure between the beam was observed at this stage Matrix cracks the inner and the outer loading points were also also seen to have propagated partially through observed. Shear failure between the outer and inner thickness of the sample On further loading of loading points can be attributed to small span to mples, delamination cracks were seen to depth ratios(24). Only two samples were loaded to orm and the load dropped abruptly. Fibres were the extent that it was possible to separate them by also seen bridging these delamination cracks. This hand. These samples were then used for detailed suggests that the delamination cracks in the matrix fractography. Fig. 7. Fractograph showing fibre pull-out in composite A. Matrix debris attached to the fibres is scen at some places
2930 KUMAR and KNOWLES: SIC REINFORCED ALUMINOSILICATES-II Fig. 6. Fractograph showing both compression and tension failures in composite A. Fibre buckling is also seen. optical microscope. It was observed that the first damage always occurred on the tensile side of the beam. However, only a few matrix cracks, perpendicular to the loading plane, were observed on the tension side when the test was stopped at a load corresponding to the onset of non-linear deflection in the load-deflection curve. It is possible that some matrix cracks close upon unloading, but observations after loading beyond the onset of non-linear deflection did not reveal any increase in the crack population. No damage on the compression side of the beam was observed at this stage. Matrix cracks were also seen to have propagated partially through the thickness of the sample. On further loading of the samples, delamination cracks were seen to form and the load dropped abruptly. Fibres were also seen bridging these delamination cracks. This suggests that the delamination cracks in the matrix crossed over the fibres because of misalignment of the fibres [3 11. The load built up again as the delamination cracks propagated. The load then dropped abruptly, suggesting either the failure of bridging fibres and/or the initiation of further delamination cracks (Fig. 8) in a different laminate. The process of load build-up and drop continued. In some samples, damage on the compression side was also observed at higher loads. In addition to these failure processes, crushing below the loading rollers and shear failure between the inner and the outer loading points were also observed. Shear failure between the outer and inner loading points can be attributed to small span to depth ratios (~4). Only two samples were loaded to the extent that it was possible to separate them by hand. These samples were then used for detailed fractography. Fig. 7. Fractograph showing fibre pull-out m composite A. Matrix debris attached to the fibres is seen at some places
KUMAR and KNOWLES: SiC REINFORCED ALUMINOSILICATES-ll 0.5mm Fig 8. Fractograph showing a matrix crack(indicated by arrows) and the formation of delamination composite B The delamination cracks originated at a matrix edge of the sample compared with the bulk sample rack that had propagated partially through the (see Section 3. 1). Similar fracture behaviour has thickness, as in the example shown arrowed in Fig. 8. been reported in the SiC/MAS literature [7, 20 Additional matrix cracks were opened perpendicular Metcalfe et al. [7] have observed brittle failure and to the delamination cracks and emanate from the delamination in SiC fibre(Tyranno)-reinforced MAs lamination cracks as the delamination cracks glass-ceramic composites. They attributed the brittle propagated [32]. The final failure resulted from the failure to the non-uniform distribution of fibres, umulative effect of all these failure processes. The suggesting that the flat regions were relatively devoid fracture surface of a sample failed by the process of of fibres. On the other hand, bleay and Scott [20] matrix cracking and delamination is shown in reported that brittle failure in the matrix crack front ig. 9(a). The flat region [shown by an arrow in was due to localised oxidation at the surface Fig. (a) represents matrix cracking without any the composites (Nicalon fibre- reinforced barium obvious fibre pull-out in this region. However, a osumilite)during the heat-treatment. Our experi closer examination of the flat region showed some mental results here are in accord with the work of holes in the matrix, indicating some fibre pull-out; Bleay and Scott [20] ome fibres can also be seen protruding from the Figure 10 shows a delaminated surface of the surface [Fig. 9(b). The pull-out lengths were, how- sample shown in Fig. 9(a). The bare fibres and the ever, smaller than 10 um. The failure of fibres in the fibre imprints in the matrix clearly indicate poor sulting in the on can the matrix and the fibres. Thc fibre attributed to high interfacial friction stresses at the surfaces appear to be fairly smooth, except for some Fig. y,(a) SEM micrograph showing fracture surface of a sample ol composite B and (b)fractograph showing fibres protruding from the flat surface shown by an arrow in (a)
KUMAR and KNOWLES: SIC REINFORCED ALUMINOSILICATES-II 2931 Fig. 8. Fractograph showing a matrix crack (indicated by arrows) and the formation of delamination cracks in composite B. The delamination cracks originated at a matrix crack that had propagated partially through the thickness, as in the example shown arrowed in Fig. 8. Additional matrix cracks were opened perpendicular to the delamination cracks and emanate from the delamination cracks as the delamination cracks propagated [32]. The final failure resulted from the cumulative effect of all these failure processes. The fracture surface of a sample failed by the process of matrix cracking and delamination is shown in Fig. 9(a). The flat region [shown by an arrow in Fig. 9(a)] represents matrix cracking without any obvious fibre pull-out in this region. However, a closer examination of the flat region showed some holes in the matrix, indicating some fibre pull-out; some fibres can also be seen protruding from the surface [Fig. 9(b)]. The pull-out lengths were, however, smaller than 10 pm. The failure of fibres in the matrix crack front resulting in the flat region can be attributed to high interfacial friction stresses at the edge of the sample compared with the bulk sample (see Section 3.1). Similar fracture behaviour has been reported in the SiC/MAS literature [7,20]. Metcalfe et al. [7] have observed brittle failure and delamination in Sic fibre (Tyranno)-reinforced MAS glass-ceramic composites. They attributed the brittle failure to the non-uniform distribution of fibres, suggesting that the flat regions were relatively devoid of fibres. On the other hand, Bleay and Scott [20] reported that brittle failure in the matrix crack front was due to localised oxidation at the surface of the composites (Nicalon fibre-reinforced barium osumilite) during the heat-treatment. Our experimental results here are in accord with the work of Bleay and Scott [20]. Figure 10 shows a delaminated surface of the sample shown in Fig. 9(a). The bare fibres and the fibre imprints in the matrix clearly indicate poor bonding between the matrix and the fibres. The fibre surfaces appear to be fairly smooth, except for some Fig. 9. (a) SEM micrograph showing fracture surface of a sample of composite B and (b) fractograph showing fibres protruding from the flat surface shown by an arrow in (a)
2932 KUMAR and KNOWLES: SiC REINFORCED ALUMINOSILICATES-II Fig. 10. Fractograph showing the smooth fibre surfaces and fbre imprints in the matrix. A matrix crack is also seen going around the fibres indicating debonding at weak fibre-matrix interfaces. small debris attached to the surface. A matrix crack tip This mode I crack extended through is also seen going around the fibres and this again thickness and was followed by the formation of suggests poor bonding between the fibre and the delamination cracks (mixed-mode I and Ii) matrix. Figure 11 shows a similar delaminated perpendicular to the notch. More mode-I cracks form surface of the same sample. On this surface a number on propagation of the delamination crack. Similarly, of broken fibres are seen and the fractured fibre ends in composite B, mode-I matrix cracking starts at the are generally flat, indicating tensile brittle failure. tension surfacc, and then a matrix crack extends Fibres generally fracture at defects such as those through the thickness, followed by mixed-mode shown by arrows in Fig. Il The fracture process in composite B is similar to delamination cracks extend, more mode-I cracks the failure of a notched unidirectional SiC fibre- form perpendicular to the delamination cracks reinforced borosilicate glass matrix composite [31]. Two important differences in the failure Bordia et al. [31]observed that the first damage was mechanisms of composites A and B are thus: (i) the formation of a matrix crack(mode-I)at the notch extensive matrix cracking of composite A Fig, I1. Fractograph showing the broken fibres in a delaminated surfacc. The fat fracture su
2932 Fig. KUMAR and KNOWLES: Sic REINFORCED ALUMINOSILICATES-II 10. Fractograph showing the smooth fibre surfaces and fibre imprints in the matrix. A matrix crack is also seen going around the fibres indicating debonding at weak fibre-matrix interfaces. small debris attached to the surface. A matrix crack is also seen going around the fibres and this again suggests poor bonding between the fibre and the matrix. Figure 11 shows a similar delaminated surface of the same sample. On this surface a number of broken fibres are seen and the fractured fibre ends are generally flat, indicating tensile brittle failure. Fibres generally fracture at defects such as those shown by arrows in Fig. 11. The fracture process in composite B is similar to the failure of a notched unidirectional SIC fibrereinforced borosilicate glass matrix composite [31]. Bordia et al. [31] observed that the first damage was the formation of a matrix crack (mode-I) at the notch tip. This mode-1 crack extended through the thickness and was followed by the formation of delamination cracks (mixed-mode I and II) perpendicular to the notch. More mode-1 cracks form on propagation of the delamination crack. Similarly, in composite B, mode-1 matrix cracking starts at the tension surface, and then a matrix crack extends through the thickness, followed by mixed-mode delamination at the matrix crack tip. As the delamination cracks extend, more mode-1 cracks form perpendicular to the delamination cracks. Two important differences in the failure mechanisms of composites A and B are thus: (i) extensive matrix cracking of composite A in Fig. 11. Fractograph showing the broken fibres in a delaminated surface. The flat fracture surfaces indicate tensile brittle fracture at defects shown by arrows