w. Yang et al. Materials Science and Engineering 4345(2003)28-35 the 90 bundles are regarded as matrix. Thus the volume that further marked improvement of the strength of the fraction of the 'matrixis the volume fractions of the real Tyranno-SA/SiC composites is possible through the CVI-SiC matrix and the 90 direction fibers given br optimization of the CVi process for dense and pure V"=1-V-V crystalline matrix deposition and appropriate woven type of the reinforcements where Vp is the volume fraction of pores in the Another main assumption of the model [21] is that composites,as given in Table 2. Vf is the volume multiple matrix cracking occur with constant distance fraction of the 0o fiber bundles(half of ve between two consecutive cracks when the matrix crack Fig. 8 shows the calculated PLS, as well as the initiated, followed by interfacial debonding and fiber experiment results, via the ISS. Compared with the sliding without further damage of the matrix until the experiment results, the mode over estimated the PLs failure of the fibers. This is not always true in practical with a simple increase with the ISS. As, mentioned materials, especially for composites with strong inter- before, the mode was developed based on unidirectional facial bonding, because when the interfacial bonding is composite cell, assuming defect-free fibers and matrix. strong, interfacial debonding or matrix crack being In above calculation, the 2d plain-woven fabrics were deflected at the interfaces or bridged by the fibers are simply regarded as unidirectional reinforcement fibers difficult to occur. Instead, significant stress concentra- (O bundles)andmatrix(90 bundles). Clearly, crack is tion will take place at the crack tip. Extremely, when the easier to initiate and propagate in the 90 bundle fiber/matrix bonding is so strong that before interfacial matrix. Therefore, the load-bearing ability of the debonding and matrix crack deflections occur, a single matrix under such simple treatment was over esti- main matrix crack(s)forms and propagates through the thickness of the composite, resulting a brittle failure CVI-matrix due to the mechanical and thermal residual (without fiber pullout) behavior with lower PLS, as stresses resulting from the fabrication process. In addi- exhibited by composites T-NL. This is likely to be the are, more or less, curved. This phenomenon, wich practical composites, the reinforcement fibers reason that the mode predicted a simple increase with the iss while the experimental results displayed a not considered in the mode, is enhanced in a 2D plain contrary tendency when the ISs is over certain value woven or more complex woven fabric preforms, result- This indicates that there is a threshold of ISS beyond ing in a decrease of the load-sharing ability of the fibers. which the mode becomes invalid. For the present These factors are believed to be the main reasons Tyranno-SA/SIC composites, this threshold seems to causing the over estimation of the Pls by the mode be 340 MPa at which the highest Pls was achieved Assuming that the effects of above factors on the PLs among all the composites can be expressed by simply applying a constant coeffi- cient,K, in Eq (1). Then a value of 0.373 for the K ielded a much better agreement between the mode 5. Conclusion estimation and experimental results, as shown in Fig 8 (modified estimations were slightly shifted towards Several Tyranno-SA fiber reinforced SiC/SiC compo- smaller ISS for clarity). The original mode(K-1)gives sites with various PyC and SiC/PyC interlayers we a simple theoretical prediction of the strength of a fat fabricated and the mechanical properties and fracture composite from the properties of the constitutive(the behaviors were studied using three-point bending tests fiber, the matrix, etc. ) By fitting the model calculation The conclusions are into the experiment results(K=0.373), it is indicative The isothermal CVI process was confirmed to be able to deposit PyC and Sic/C interlayers on small diameter Tyranno-SA fibers in SiC/SiC composites with sufficient Calculation (K= 1) thickness and uniformity control alculation (K=0.373) The flexural strengths of the composites demonstrated close dependence on the PyC interlayer. The ultimate strength increased with the increasing of the PyC layer thickness to 100 nm, and then kept at a similar level till t300 200nm. The materials demonstrated high PLSs but relatively small displacements at load maximums. High PLS is a Interfacial shear strength /MPa direct benefit from the large tensile modulus of the fiber while relatively small displacements at load maximums Fig. 8. Calculated and experimental PLS via ISS of the composites are likely due mainly to the strong interfacial bondingthe 908 bundles are regarded as matrix. Thus the volume fraction of the ‘matrix’ is the volume fractions of the real CVI-SiC matrix and the 908 direction fibers given by: V? m1VpV? f; where Vp is the volume fraction of pores in the composites, as given in Table 2. V? f is the volume fraction of the 08 fiber bundles (half of Vf). Fig. 8 shows the calculated PLS, as well as the experiment results, via the ISS. Compared with the experiment results, the mode over estimated the PLS with a simple increase with the ISS. As, mentioned before, the mode was developed based on unidirectional composite cell, assuming defect-free fibers and matrix. In above calculation, the 2D plain-woven fabrics were simply regarded as unidirectional reinforcement fibers (08 bundles) and ‘matrix’ (908 bundles). Clearly, crack is easier to initiate and propagate in the 908 bundle ‘matrix’. Therefore, the load-bearing ability of the ‘matrix’ under such simple treatment was over esti￾mated, let along the possible pre-exist cracks in the real CVI-matrix due to the mechanical and thermal residual stresses resulting from the fabrication process. In addi￾tion, in practical composites, the reinforcement fibers are, more or less, curved. This phenomenon, which is not considered in the mode, is enhanced in a 2D plain￾woven or more complex woven fabric preforms, result￾ing in a decrease of the load-sharing ability of the fibers. These factors are believed to be the main reasons causing the over estimation of the PLS by the mode. Assuming that the effects of above factors on the PLS can be expressed by simply applying a constant coeffi￾cient, K, in Eq. (1). Then a value of 0.373 for the K yielded a much better agreement between the model estimation and experimental results, as shown in Fig. 8 (modified estimations were slightly shifted towards smaller ISS for clarity). The original mode (K/1) gives a simple theoretical prediction of the strength of a composite from the properties of the constitutive (the fiber, the matrix, etc.). By fitting the model calculation into the experiment results (K/0.373), it is indicative that further marked improvement of the strength of the Tyranno-SA/SiC composites is possible through the optimization of the CVI process for dense and pure crystalline matrix deposition and appropriate woven type of the reinforcements. Another main assumption of the model [21] is that multiple matrix cracking occur with constant distance between two consecutive cracks when the matrix crack initiated, followed by interfacial debonding and fiber sliding without further damage of the matrix until the failure of the fibers. This is not always true in practical materials, especially for composites with strong inter￾facial bonding, because when the interfacial bonding is strong, interfacial debonding or matrix crack being deflected at the interfaces or bridged by the fibers are difficult to occur. Instead, significant stress concentra￾tion will take place at the crack tip. Extremely, when the fiber/matrix bonding is so strong that before interfacial debonding and matrix crack deflections occur, a single main matrix crack(s) forms and propagates through the thickness of the composite, resulting a brittle failure (without fiber pullout) behavior with lower PLS, as exhibited by composites T-NL. This is likely to be the reason that the mode predicted a simple increase with the ISS while the experimental results displayed a contrary tendency when the ISS is over certain value. This indicates that there is a threshold of ISS beyond which the mode becomes invalid. For the present Tyranno-SA/SiC composites, this threshold seems to be
/340 MPa at which the highest PLS was achieved among all the composites. 5. Conclusion Several Tyranno-SA fiber reinforced SiC/SiC compo￾sites with various PyC and SiC/PyC interlayers were fabricated and the mechanical properties and fracture behaviors were studied using three-point bending tests. The conclusions are: The isothermal CVI process was confirmed to be able to deposit PyC and SiC/C interlayers on small diameter Tyranno-SA fibers in SiC/SiC composites with sufficient thickness and uniformity control. The flexural strengths of the composites demonstrated a close dependence on the PyC interlayer. The ultimate strength increased with the increasing of the PyC layer thickness to 100 nm, and then kept at a similar level till 200 nm. The materials demonstrated high PLSs but relatively small displacements at load maximums. High PLS is a direct benefit from the large tensile modulus of the fiber while relatively small displacements at load maximums Fig. 8. Calculated and experimental PLS via ISS of the composites. are likely due mainly to the strong interfacial bonding. 34 W. Yang et al. / Materials Science and Engineering A345 (2003) 28 /35