M. Schmicker et al /Journal of the European Ceramic Society 20(2000)2491-2497 TEM, however, showed that the mullite nucleation sponds to 2 h firing at 152 respectively. The occurred inside the glassy phase rather than in direct microstructural observation is in good accordance to contact with the a-Al2O3 grains. Microstructural this estimation(Figs. 7 and 3c): samples annealed for evidence suggested that Al,O3 was solved in the coex 000 h at 1300 C do display reaction zones at the fibre isting non-crystalline silica phase and mullite nucleation rims but the zones are smaller than these of samples occurs as soon as Al2O3 supersaturation is reached. >In thermally treated for 2 h at 1600C the present study, mullite formation inside the glassy Load/ deflection curves of the porous mullite matrix pockets of the fibres has never been observed composites heat-treated at various temperatures signa mullite grains a priori are in contact with both lize a damage tolerant fracture behaviour up to 1500oC (matrix)and a-Al2O3(fibres), it is concluded that no (Fig 8). A 25% decrease in maximum strength occurs mullite nucleation took place. We believe, instead, that after firing at 1500C with respect to the starting growth of the pre-existing mullite grains by interdiffu- material. Interestingly enough, Nextel 720 fibres alone sion of Si+ and Al+ occurred. The driving force for undergo a 60%strength degradation after a 1500C interdiffusion of Si++ and AP+ in mullite crystals is heat-treatment. 17 Obviously, the strength of the com- obviously the occurring concentration gradient ranging posites fired up to 1500C is not controlled by direct from A72 wt. Al2O3(mullite in contact with SiO2)to fibre strength but fibre/matrix delamination will occur 4 wt. Al2O3(mullite in contact with a-Al2O3; see as the first step of failure. The damage tolerant fracture Fig. 11). According to this model, mullitization starts at behaviour of the composite after reaching maximum he interfacial area of fibres and matrix. The matrix load is controlled by crack bridging, multiple cracking thereby acts as a silica reservoir when the Sio2 located and fibre pull-out, which is demonstrated by a typical directly at the fibre/matrix boundary is consumed. fracture surface(Fig. 9a). Firing at 1600C, on the other During the reaction process, the viscous silica-rich hand, leads to the above described reactions in the fibre/ phase is presumably transported towards the fibre sur- matrix interfacial area and hence bonding between face by capillary forces. Reaction kinetics can be esti- matrix and fibres drastically increases. As a con- mated assuming @=700 kJ/mol as a typical value sequence, the material becomes brittle as can be clearly of activation energy for diffusional processes in mul- derived from the load /deflection curves(Fig. 8)and by lite. 6 Using this activation energy value, a temperature the resulting flat fracture surface( Fig. 9b) increment of 25 C corresponds to a doubling of the reaction rate. and 1000 h annealing at 1300 oc corre e It is an important result of this study that only little rmally-induced degradation of the porous matrix composite occurs unless interactions between fibres and matrix take place. This favorable behavior is explained by the fact that thermal activation of reactions between vitreous SiO and a-Al2O3 is considerably high According to the present study, the thermal stability of mullite matrix composites is estimated to be 1500C in the case of short-term application long-term applications(1000 h and more 1800- mullite Acknowledgements silica mullite alumina We thank the german Research Foundation. DFG mullite (Schn 297/15-1)for financial support 1700 References 1. Chawla, K. K, Composite Materials, Science and Engineerin Springer-Verlag, New York, 1987(pp. 134-149) 2. Chawla, K. K, Ceramic Matrix Composites. Chapman and Hall, London,1993(pp.162-195) 3. Morgan, P. E. D. and Marshall, D. B, Funct terraces for 76 wt %AL O 4. Keller. K. A, Mah, T, Parthasarathy, T.A poke. C. M. Fugitive interfacial carbon coatings for oxide/oxide composites. Fig. Il. Mullite stability region of the SiOx-Al2O3 phase diagram Ceram. Eng. and Sci. Proc., 1997, 14, 878-879. (after Klug et al.) 5. Lange, F. F. Tu. W. and Evans. C. A. G. Processing of damage.TEM, however, showed that the mullite nucleation occurred inside the glassy phase rather than in direct contact with the a-Al2O3 grains. Microstructural evidence suggested that Al2O3 was solved in the coex￾isting non-crystalline silica phase and mullite nucleation occurs as soon as Al2O3 supersaturation is reached.15 In the present study, mullite formation inside the glassy pockets of the ®bres has never been observed. Since mullite grains a priori are in contact with both, SiO2 (matrix) and a-Al2O3 (®bres), it is concluded that no mullite nucleation took place. We believe, instead, that growth of the pre-existing mullite grains by interdi€u￾sion of Si4+ and Al3+ occurred. The driving force for interdi€usion of Si4+ and Al3+ in mullite crystals is obviously the occurring concentration gradient ranging from 72 wt.% Al2O3 (mullite in contact with SiO2) to 74 wt.% Al2O3 (mullite in contact with a-Al2O3; see Fig. 11). According to this model, mullitization starts at the interfacial area of ®bres and matrix. The matrix thereby acts as a silica reservoir when the SiO2 located directly at the ®bre/matrix boundary is consumed. During the reaction process, the viscous silica-rich phase is presumably transported towards the ®bre sur￾face by capillary forces. Reaction kinetics can be esti￾mated assuming Q=700 kJ/mol as a typical value of activation energy for di€usional processes in mul￾lite.16 Using this activation energy value, a temperature increment of 25C corresponds to a doubling of the reaction rate, and 1000 h annealing at 1300 C corre￾sponds to 2 h ®ring at 1525 C, respectively. The microstructural observation is in good accordance to this estimation (Figs. 7 and 3c): samples annealed for 1000 h at 1300C do display reaction zones at the ®bre rims but the zones are smaller than these of samples thermally treated for 2 h at 1600C. Load/de¯ection curves of the porous mullite matrix composites heat-treated at various temperatures signa￾lize a damage tolerant fracture behaviour up to 1500C (Fig. 8). A 25% decrease in maximum strength occurs after ®ring at 1500C with respect to the starting material. Interestingly enough, Nextel 720 ®bres alone undergo a 60% strength degradation after a 1500C heat-treatment.17 Obviously, the strength of the com￾posites ®red up to 1500C is not controlled by direct ®bre strength but ®bre/matrix delamination will occur as the ®rst step of failure. The damage tolerant fracture behaviour of the composite after reaching maximum load is controlled by crack bridging, multiple cracking and ®bre pull-out, which is demonstrated by a typical fracture surface (Fig. 9a). Firing at 1600C, on the other hand, leads to the above described reactions in the ®bre/ matrix interfacial area and hence bonding between matrix and ®bres drastically increases. As a con￾sequence, the material becomes brittle as can be clearly derived from the load/de¯ection curves (Fig. 8) and by the resulting ¯at fracture surface (Fig. 9b). It is an important result of this study that only little thermally-induced degradation of the porous matrix composite occurs unless interactions between ®bres and matrix take place. This favorable behavior is explained by the fact that thermal activation of reactions between vitreous SiO2 and a-Al2O3 is considerably high. According to the present study, the thermal stability of the Nextel 720 ®bre reinforced porous mullite matrix composites is estimated to be 1500C in the case of short-term application (several hours) or 1300C for long-term applications (1000 h and more). Acknowledgements We thank the German Research Foundation, DFG (Schn 297/15-1) for ®nancial support. References 1. Chawla, K. K., Composite Materials, Science and Engineering. Springer-Verlag, New York, 1987 (pp. 134±149). 2. Chawla, K. K., Ceramic Matrix Composites. Chapman and Hall, London, 1993 (pp. 162±195). 3. Morgan, P. E. D. and Marshall, D. B., Functional interfaces for oxide/oxide composites. Mat. Sci. Engg., 1993, A162, 15±25. 4. Keller, K. A., Mah, T., Parthasarathy, T. A. and Cooke, C. M., Fugitive interfacial carbon coatings for oxide/oxide composites. Ceram. Eng. and Sci. Proc., 1997, 14, 878±879. 5. Lange, F. F., Tu, W. and Evans, C. A. G., Processing of damage￾Fig. 11. Mullite stability region of the SiO2±Al2O3 phase diagram (after Klug et al.18). 2496 M. SchmuÈcker et al. / Journal of the European Ceramic Society 20 (2000) 2491±2497