M.L. Antti et al. Journal of the European Ceramic Society 24(2004)565-578 These changes were not observed for shorter times at the fibre/matrix bonding is also provided by the fact l100° C or in a sample exposed for 8 h at 1200°C that the strength of the +45 composite, which is domi The results of the HT-XRD study of as-received nated by the matrix properties, increased while its strain Nextel 720 fibres revealed, as expected, the thermal to failure decreased with thermal exposure. Thus, after expansion of the mullite and alumina phases. However, 100 h at 1100C the 0/90 and +45 materials had similar beginning around 900C, the expansion coefficient of mechanical properties approaching the behaviour of a the a-axis of the mullite crystal structure(the crystal monolithic ceramic structure being close to tetragonal in the as-received There were some indications of phase transformation, fibre)began to decrease with increasing temperature namely the formation of cristobalite after 100 h at and subsequently became negative, implying shrinkage 1100C. This was observed by XRD but not by of the a-axis. The resulting change in axial ratio was spectroscopy. However, in the latter method spectra largely retained on cooling. At the same time changes in were taken from a restricted number of spots in the relative peak-intensities indicated an increase in the sample and could therefore have overlooked cristobalite amount of alumina. These observations suggest the if this existed locally Phase studies of the alumina-silica initiation of a permanent change in the fibre involving a (A-S) system show that mullite can exist over a rela reduction in aluminium content in the mullite associated tively wide range of A: S ratios. The composition of with a shift in crystal structure from near-tetragonal to the matrix in the present composites is considered to lie orthorhombic, an effect reported earlier by Wilson et al.26 well within this range. However, if the matrix mixture The Raman spectroscopy indicated no change in the were inhomogeneous, as was indicated by the metallo- composition of the fibres after heat-treatment for 100 h graphic studies, regions of high silica content lying out at 1100 C. Also the spectra for the matrix were very side the mullite range and therefore having a potential similar before and after heat-treatment, the only differ- for cristobalite formation could have existed locally. An ence detected being a slightly higher content of alumina alternative explanation for the occurrence of cristobalite after heat-treatment. No cristobalite was detected in the is that the matrix, prepared from an alumina-silica matrix or the fibres either before or after heat-treatment. mixture, had not fully reacted to mullite at all places The formation of mullite matrices in similar composites has been observed to occur via cristobalite formation at 5. Discussio temperatures as low as 1240oC8 Cristobalite is gen- erally considered to be a high-temperature form of silica The mechanical test results revealed that thermal occurring above about 1450C- but could well be xposure of the composites at 1000C and above caused encouraged to form at lower temperatures by the nature embrittlement and, in the case of 0/900 composites, loss of the precursor reactants and or impurity species such of strength. This degradation was particularly marked as Na and K. In view of the local and transient nature after 100 h at 1100C when the 0/90 material exhibited of cristobalite in the composites it is considered that its a fracture toughness and notch sensitivity that would be formation did not contribute directly to the composite expected of a monolithic ceramic. A number of reasons degradation for such degradation can be sought including: (i) sinter Regarding the strength retention of Nextel 720 fibres ing of the matrix leading to reduction in porosity and after thermal exposure, the present authors are not consequently a loss of the supposed crack deflection aware of any reported measurements for exposures cor- behaviour (ii)increased bonding between fibre and responding to the thermal treatments applied to the matrix leading to a similar effect (iii) strength reduction composites in this work. However, the hT-XRD results of the matrix due to phase changes and/ or grain growth presented here and the few reported results of strength (iv) degradation of the fibre. loss of fibres indicate that permanent structural changes porosity measurements, supported by micro- can occur in the fibre and that this would be accom hardness measurements indicated clearly that matrix panied by some strength loss. Such strength loss could densification occurred during thermal exposure. The lead to the reduction in fibre pull-out observed on the overall shrinkage of the samples was negligible due to fracture surfaces but this could equally well be attrib the constraint of the continuous fibre skeleton but the uted to the increased fibre-matrix bonding. Resolution shrinkage occurred internally leading to opening of of this question awaits further studies of fibre strength matrix cracks and growth of existing voids. An degradation due to thermal exposure increased bonding between fibres and matrix also occurred as evidenced by an increased adherence of matrix to the fibres observed on fracture surfaces. Both 6. Conclusions matrix strengthening and increased fibre /matrix bond ing was indicated by the increasing stifness of the com The effects have been investigated of thermal expo- posites. Evidence for a strengthening of the matrix and sure on the microstructure and stress-strain behaviourThese changes were not observed for shorter times at 1100
C or in a sample exposed for 8 h at 1200
C. The results of the HT-XRD study of as-received Nextel 720 fibres revealed, as expected, the thermal expansion of the mullite and alumina phases. However, beginning around 900
C, the expansion coefficient of the a-axis of the mullite crystal structure (the crystal structure being close to tetragonal in the as-received fibre) began to decrease with increasing temperature and subsequently became negative, implying shrinkage of the a-axis. The resulting change in axial ratio was largely retained on cooling. At the same time changes in relative peak-intensities indicated an increase in the amount of alumina. These observations suggest the initiation of a permanent change in the fibre involving a reduction in aluminium content in the mullite associated with a shift in crystal structure from near-tetragonal to orthorhombic, an effect reported earlier by Wilson et al.26 The Raman spectroscopy indicated no change in the composition of the fibres after heat-treatment for 100 h at 1100
C. Also the spectra for the matrix were very similar before and after heat-treatment, the only differ￾ence detected being a slightly higher content of alumina after heat-treatment. No cristobalite was detected in the matrix or the fibres either before or after heat-treatment. 5. Discussion The mechanical test results revealed that thermal exposure of the composites at 1000
C and above caused embrittlement and, in the case of 0/90
composites, loss of strength. This degradation was particularly marked after 100 h at 1100
C when the 0/90 material exhibited a fracture toughness and notch sensitivity that would be expected of a monolithic ceramic. A number of reasons for such degradation can be sought including: (i) sinter￾ing of the matrix leading to reduction in porosity and consequently a loss of the supposed crack deflection behaviour (ii) increased bonding between fibre and matrix leading to a similar effect (iii) strength reduction of the matrix due to phase changes and/or grain growth (iv) degradation of the fibre. The porosity measurements, supported by micro￾hardness measurements indicated clearly that matrix densification occurred during thermal exposure. The overall shrinkage of the samples was negligible due to the constraint of the continuous fibre skeleton but the shrinkage occurred internally leading to opening of matrix cracks and growth of existing voids. An increased bonding between fibres and matrix also occurred as evidenced by an increased adherence of matrix to the fibres observed on fracture surfaces. Both matrix strengthening and increased fibre/matrix bond￾ing was indicated by the increasing stiffness of the com￾posites. Evidence for a strengthening of the matrix and the fibre/matrix bonding is also provided by the fact that the strength of the 45 composite, which is domi￾nated by the matrix properties, increased while its strain to failure decreased with thermal exposure. Thus, after 100 h at 1100
C the 0/90 and 45 materials had similar mechanical properties approaching the behaviour of a monolithic ceramic. There were some indications of phase transformation, namely the formation of cristobalite after 100 h at 1100
C. This was observed by XRD but not by Raman spectroscopy. However, in the latter method spectra were taken from a restricted number of spots in the sample and could therefore have overlooked cristobalite if this existed locally. Phase studies of the alumina–silica (A–S) system show that mullite can exist over a rela￾tively wide range of A:S ratios.27 The composition of the matrix in the present composites is considered to lie well within this range. However, if the matrix mixture were inhomogeneous, as was indicated by the metallo￾graphic studies, regions of high silica content lying out￾side the mullite range and therefore having a potential for cristobalite formation could have existed locally. An alternative explanation for the occurrence of cristobalite is that the matrix, prepared from an alumina–silica mixture, had not fully reacted to mullite at all places. The formation of mullite matrices in similar composites has been observed to occur via cristoballite formation at temperatures as low as 1240
C.28 Cristobalite is gen￾erally considered to be a high-temperature form of silica occurring above about 1450
C 27 but could well be encouraged to form at lower temperatures by the nature of the precursor reactants and/or impurity species such as Na and K.29 In view of the local and transient nature of cristobalite in the composites it is considered that its formation did not contribute directly to the composite degradation. Regarding the strength retention of Nextel 720 fibres after thermal exposure, the present authors are not aware of any reported measurements for exposures cor￾responding to the thermal treatments applied to the composites in this work. However, the HT-XRD results presented here and the few reported results of strength loss of fibres indicate that permanent structural changes can occur in the fibre and that this would be accom￾panied by some strength loss. Such strength loss could lead to the reduction in fibre pull-out observed on the fracture surfaces but this could equally well be attrib￾uted to the increased fibre-matrix bonding. Resolution of this question awaits further studies of fibre strength degradation due to thermal exposure. 6. Conclusions The effects have been investigated of thermal expo￾sure on the microstructure and stress–strain behaviour 576 M.-L. Antti et al. / Journal of the European Ceramic Society 24 (2004) 565–578