S. Zhu et al. /Composites Science and Technology 59(1999)833-851 in 0 bundles at crossover points initiated from large 3.3. Crack propagation and connection pores for cyclic fatigue at low stresses at 1000%C.This implies that there may exist a critical stress for the Cracks first propagate in the matrix or along transition of the dominant damage mode. At high faces between the matrix and the fibers in the 90 bun- stresses,matrix cracking is considered to be saturated dles and then interact with predominantly intact fibers during first loading. Since extensive matrix cracks and in the 0o bundles, as shown in Fig 8. The debonding of debonding of interfaces decrease stress concentrations, interfaces occurs in 0 bundles. Both the pores and the he 90 bundles become easy paths for crack propagation. interfaces cause cracks to deflect. The propagating However, at low stress, matrix cracking is unsaturated cracks sometimes stop at the pores If there is another and therefo trix cracking continues during cyclic crack initiated at the pore in the direction of crack pro- loading. Stress concentration points at pores take an pagation, crack interconnection occurs. The interfaces important role for fracture path orientation. More- in 90 bundles promote crack propagation. However, over, creep of the 0o fibers and the reduced interfacial the interfaces in 0o bundles can deflect cracks by sliding resistance also promote fracture in the 0 bun debonding along the interfaces between fibers and dles at high temperatures. Therefore, the saturation matrix. Debonding is a prerequisite for fiber-bridging stress for matrix cracking may be the critical stress for and fiber-pull-out. When the transverse cracks propa the transition of fracture mode. This will be studied in gate into 0 bundles, they may occasionally cut some the future fibers which are strongly bonded with the matrix so that Both the pores and the architecture are important for no debonding occurs, and then propagate along the crack initiation and propagation. As a consequence of weak, debonded interfaces in the middle of 0 bundles the woven 2D fiber architecture. non-uniform stresses At room temperature, the first loading leads to and strains develop in the axial, width and thickness cracking and at the same time leaves intact 0 fibers in the directions of the specimen. Because of the non-uni- wake of the cracks. If the stress remains constant, the formity in stress and strain near the crossover points of bridging force keeps the cracks stationary. Therefore, no the fiber bundles, there is extensive fracture and spalla- stress rupture occurs at room temperature [64]. How on of the SiC matrix at these locations [19]. In the ever, under cyclic loading the debonded interface sliding specimens, cracks commonly initiated at the sharp cor- resistance decreases, which in turn promotes interface ners of large pores at the crossover point debonding further. The average fiber strength depends The non-uniform stress and strain distributions on the length tested: longer fibers are more likely to develop at different scales along three orthogonal direc- have more detrimental, strength-limiting defects and so tions [ 19]. At the macroscopic scale, the 2D fiber bundle are weaker. Therefore, the decrease of the interface configuration produces a non-uniform stress distribu- sliding resistance can reduce the fiber bridging stress tion along the axial and width directions within indivi- and is a dominant cyclic fatigue damage mechanism in dual plies; the random stacking of the plies produces a SiC/SiC [62] non-uniform stress distribution in the thickness direc t high temperature, creep of fibers can relax the fiber tion of the specimen. Non-uniform stretching of plies bridging forces and promote subcritical crack propaga produces a shear stress and contact pressure at the con- tion. Moreover, relaxation of the residual stress at the act points between them. This leads to delamination interface between fibers and matrix at 1000C can between plies. Within individual fiber bundles, cracking decrease frictional stress of the interface. The time- of the matrix occurs as the 0 bundles attempt to align dependent decrease of the fiber bridging stress promotes with the tensile loading direction. Non-aligned fibers crack propagation. It was found that the creep fracture suppress pull-out by virtue of bending strains. The life also sharply decreased at 1000@C like cyclic fatigue longest pull-out length in 2D composite is the width at the stresses below 160 MPa [73] between the two 90 bundles A random distribution of 0 fiber failure under fati Evans et al. [67] reported that the strength of 2D gue loading and bundle rupture near the crossover SiC/SiC was half of the strength of unidirectional points as found at room temperature in a carbon/SiC NicalonM/SiC and pointed out that hence the beha- composite [19]. This differs from the present results on vior of 2D materials must be dominated by the 0o bun- cyclic fatigue at room temperature. However, fracture dles. In fact, it is not only the effect of 90 bundles to by cyclic fatigue at 1000C shows a similar distribution reduce the strength by half, since the architecture and of 0 bundle failure. The delamination cracking was he pores strongly decrease the strength too. As a result, severe under room temperature fatigue in carbon/ SiC the stress on 0% fibers, calculated from the maximum composite [19]. Evident delamination was also found load over the cross section area of 0 fibers without under high temperature cyclic fatigue in SiC/SiC com considering the contribution of 90 fibers and the posite. Considering carbon fibers do not bond to the matrix, is lower than the strength of the fibers(as men- SiC matrix like NicalonTM fibers, the interface sliding tioned in Section 2.1) resistance of carbon/Sic is expected to be lower thanin 0 bundles at crossover points initiated from large pores for cyclic fatigue at low stresses at 1000C. This implies that there may exist a critical stress for the transition of the dominant damage mode. At high stresses, matrix cracking is considered to be saturated during ®rst loading. Since extensive matrix cracks and debonding of interfaces decrease stress concentrations, the 90 bundles become easy paths for crack propagation. However, at low stress, matrix cracking is unsaturated and therefore matrix cracking continues during cyclic loading. Stress concentration points at pores take an important role for fracture path orientation. More￾over, creep of the 0 ®bers and the reduced interfacial sliding resistance also promote fracture in the 0 bun￾dles at high temperatures. Therefore, the saturation stress for matrix cracking may be the critical stress for the transition of fracture mode. This will be studied in the future. Both the pores and the architecture are important for crack initiation and propagation. As a consequence of the woven 2D ®ber architecture, non-uniform stresses and strains develop in the axial, width and thickness directions of the specimen. Because of the non-uni￾formity in stress and strain near the crossover points of the ®ber bundles, there is extensive fracture and spalla￾tion of the SiC matrix at these locations [19]. In the specimens, cracks commonly initiated at the sharp cor￾ners of large pores at the crossover points. The non-uniform stress and strain distributions develop at di€erent scales along three orthogonal direc￾tions [19]. At the macroscopic scale, the 2D ®ber bundle con®guration produces a non-uniform stress distribu￾tion along the axial and width directions within indivi￾dual plies; the random stacking of the plies produces a non-uniform stress distribution in the thickness direc￾tion of the specimen. Non-uniform stretching of plies produces a shear stress and contact pressure at the con￾tact points between them. This leads to delamination between plies. Within individual ®ber bundles, cracking of the matrix occurs as the 0 bundles attempt to align with the tensile loading direction. Non-aligned ®bers suppress pull-out by virtue of bending strains. The longest pull-out length in 2D composite is the width between the two 90 bundles. Evans et al. [67] reported that the strength of 2D SiC/SiC was half of the strength of unidirectional NicalonTM/SiC and pointed out that hence the beha￾vior of 2D materials must be dominated by the 0 bun￾dles. In fact, it is not only the e€ect of 90 bundles to reduce the strength by half, since the architecture and the pores strongly decrease the strength too. As a result, the stress on 0 ®bers, calculated from the maximum load over the cross section area of 0 ®bers without considering the contribution of 90 ®bers and the matrix, is lower than the strength of the ®bers (as men￾tioned in Section 2.1). 3.3. Crack propagation and connection Cracks ®rst propagate in the matrix or along inter￾faces between the matrix and the ®bers in the 90 bun￾dles and then interact with predominantly intact ®bers in the 0 bundles, as shown in Fig. 8. The debonding of interfaces occurs in 0 bundles. Both the pores and the interfaces cause cracks to de¯ect. The propagating cracks sometimes stop at the pores. If there is another crack initiated at the pore in the direction of crack pro￾pagation, crack interconnection occurs. The interfaces in 90 bundles promote crack propagation. However, the interfaces in 0 bundles can de¯ect cracks by debonding along the interfaces between ®bers and matrix. Debonding is a prerequisite for ®ber-bridging and ®ber-pull-out. When the transverse cracks propa￾gate into 0 bundles, they may occasionally cut some ®bers which are strongly bonded with the matrix so that no debonding occurs, and then propagate along the weak, debonded interfaces in the middle of 0 bundles. At room temperature, the ®rst loading leads to cracking and at the same time leaves intact 0 ®bers in the wake of the cracks. If the stress remains constant, the bridging force keeps the cracks stationary. Therefore, no stress rupture occurs at room temperature [64]. How￾ever, under cyclic loading the debonded interface sliding resistance decreases, which in turn promotes interface debonding further. The average ®ber strength depends on the length tested: longer ®bers are more likely to have more detrimental, strength-limiting defects and so are weaker. Therefore, the decrease of the interface sliding resistance can reduce the ®ber bridging stress and is a dominant cyclic fatigue damage mechanism in SiC/SiC [62]. At high temperature, creep of ®bers can relax the ®ber bridging forces and promote subcritical crack propaga￾tion. Moreover, relaxation of the residual stress at the interface between ®bers and matrix at 1000C can decrease frictional stress of the interface. The time￾dependent decrease of the ®ber bridging stress promotes crack propagation. It was found that the creep fracture life also sharply decreased at 1000C like cyclic fatigue at the stresses below 160 MPa [73]. A random distribution of 0 ®ber failure under fati￾gue loading and bundle rupture near the crossover points as found at room temperature in a carbon/SiC composite [19]. This di€ers from the present results on cyclic fatigue at room temperature. However, fracture by cyclic fatigue at 1000C shows a similar distribution of 0 bundle failure. The delamination cracking was severe under room temperature fatigue in carbon/SiC composite [19]. Evident delamination was also found under high temperature cyclic fatigue in SiC/SiC com￾posite. Considering carbon ®bers do not bond to the SiC matrix like NicalonTM ®bers, the interface sliding resistance of carbon/SiC is expected to be lower than 840 S. Zhu et al. / Composites Science and Technology 59 (1999) 833±851