J. She et al. Materials Science and Engineering 4325 (2002)19-24 (b) 10 um um ig. 5. SEM micrographs of the fracture surface of a porous mullite/mullite composite after 3 cycles of infiltration, showing(a) extensive fiber pullout and (b) a broken fiber with matrix adhered to it. bserved over the whole fracture surface. This is proba- occur in a spatially random fashion, as indicated in Fig bly due to the fact that porous matrix between the 6(b) by the broken fibers with different crack planes. In fibers disintegrated into smaller pieces during the frac- fact, the fibers fracture sequentially, according to the ture process, as evidenced in Fig. 5(b) by a noticeable probabilistic distribution of their strengths. This pro- amount of the matrix debris attached to the fibers cess results in the gradual load decreases in Fig 4 With five to seven infiltrations, the composites exhib- When the composites were infiltrated for 8-10 cycles, ited a non-catastrophic failure, but the load dropped he initial cracking event in the surface region leads to suddenly to a certain level after the maximum load. The an abrupt drop in the nominal stress to a relatively low sudden load drop is believed to be associated with level, followed immediately by a load decrease. This is surface embrittlement. As observed in Fig. 2(b), the attributed primarily to the reduction in the thickness of matrix porosity is considerably decreased in the surface the porous interior region, where the fiber pullout may region. Especially, the bonds between the fibers and the occur during fracture matrix are significantly promoted by the alumina Furthermore. it was observed that all the infiltrated bridges. In this case, the matrix crack, which initiates composites remained intact after testing, but the degree on the tensile surface at the peak stress, would propa- of load retention decreased with increasing number of gate through the fibers rather than deflect between the infiltrations. As presented in Fig 4, the composites with fibers. As can be seen in Fig. 6(a), the fracture topogra- three, six and eight infiltrations have the ability to phy of the surface region is nearly planar, with almost withstand a flexural stress of 30, 22 and 10 MPa at a no fiber pullout. This cracking event causes the sharp cross-head displacement of 0.4 mm, while the com- load drops in Fig. 4. However, such a crack is arrested posite with ten infiltrations has almost no load-bearing ind deflected by the porous matrix in the interior capability even at a cross-head displacement of 0. 2 mm region. This allows the measured load to rise again with On the other hand, it is worth to note in the stress ncreasing cross-head displacement, until some cracks displacement curves of Fig. 4 that the slope of the initiate within the fibers. Subsequently, fiber failures initial linear region increases slightly with the number22 J. She et al. / Materials Science and Engineering A325 (2002) 19–24 Fig. 5. SEM micrographs of the fracture surface of a porous mullite/mullite composite after 3 cycles of infiltration, showing (a) extensive fiber pullout and (b) a broken fiber with matrix adhered to it. observed over the whole fracture surface. This is proba￾bly due to the fact that porous matrix between the fibers disintegrated into smaller pieces during the frac￾ture process, as evidenced in Fig. 5(b) by a noticeable amount of the matrix debris attached to the fibers. With five to seven infiltrations, the composites exhib￾ited a non-catastrophic failure, but the load dropped suddenly to a certain level after the maximum load. The sudden load drop is believed to be associated with surface embrittlement. As observed in Fig. 2(b), the matrix porosity is considerably decreased in the surface region. Especially, the bonds between the fibers and the matrix are significantly promoted by the alumina ‘bridges’. In this case, the matrix crack, which initiates on the tensile surface at the peak stress, would propa￾gate through the fibers rather than deflect between the fibers. As can be seen in Fig. 6(a), the fracture topogra￾phy of the surface region is nearly planar, with almost no fiber pullout. This cracking event causes the sharp load drops in Fig. 4. However, such a crack is arrested and deflected by the porous matrix in the interior region. This allows the measured load to rise again with increasing cross-head displacement, until some cracks initiate within the fibers. Subsequently, fiber failures occur in a spatially random fashion, as indicated in Fig. 6(b) by the broken fibers with different crack planes. In fact, the fibers fracture sequentially, according to the probabilistic distribution of their strengths. This pro￾cess results in the gradual load decreases in Fig. 4. When the composites were infiltrated for 8–10 cycles, the initial cracking event in the surface region leads to an abrupt drop in the nominal stress to a relatively low level, followed immediately by a load decrease. This is attributed primarily to the reduction in the thickness of the porous interior region, where the fiber pullout may occur during fracture. Furthermore, it was observed that all the infiltrated composites remained intact after testing, but the degree of load retention decreased with increasing number of infiltrations. As presented in Fig. 4, the composites with three, six and eight infiltrations have the ability to withstand a flexural stress of 30, 22 and 10 MPa at a cross-head displacement of 0.4 mm, while the com￾posite with ten infiltrations has almost no load-bearing capability even at a cross-head displacement of 0.2 mm. On the other hand, it is worth to note in the stress– displacement curves of Fig. 4 that the slope of the initial linear region increases slightly with the number