S. Bertrand et al. /Journal of the European Ceramic Society 20(2000)1-13 Matri Matrix Fibre Fibre 6 um Fig. 15. TEM cross-section of a minicomposite reinforced with a pristine Hi-Nicalon fibre(arrows: debonding due to fibre contraction) 300nm Matrix Fibre Fig. 17. TEM longitudinal section(treated Hi-Nicalon reinforced deflection in the Pyc/Sic multilayered inter- tilayer (multideflection) as shown in Fig. 17. The microcrack does not propagate through the multilayer as in an homogeneous material. Multideflection is Fibre meanwhile not abundant and also, final deflection sys tematically occurs at the first interface. These latter fea- tures when compared to those obtained with the strong interface in Section 3. 4 are supposed to be due to a Fig. 16. SEM longit inal section showing the defection of a large interphase) mixed regime(adhesive and cohesive failure of the (non treated Hi-Nicalon reinforced minicomposite)(according to re 5. Concluding remark result, the interface is effectively weak on most of the This work has compared(PyC/SiC)n multilayers at fibre surface. The multilayer cannot work. Matrix two different scales: micrometric scale materials were cracks deflection is systematically occurring at the fibre obtained by I-CvI and nanometric ones by P-CVI surface: but also at the matrix/multilayer interface as Morphology and behaviour of these two multilayers hown in Fig. 16 under mechanical loading, were compared by means of When the fibre has been treated, the CtE mismatch a TEM examination before and after tensile loading etween fibre and matrix is correct. The electron test From the data reported in Sections 3 and 4, several microscopy shows that the chemical bonding is weaker. remarks can be drawn Despite this weakest bonding the composite behaves in a different manner. The very thin residual matrix cracks 1. Silicon carbide sublayers morphology obtained are deflected on the many interfaces present in the mul by P-CVI are much more homogeneous andresult, the interface is e€ectively weak on most of the ®bre surface. The multilayer cannot work. Matrix cracks de¯ection is systematically occurring at the ®bre surface; but also at the matrix/multilayer interface as shown in Fig. 16. When the ®bre has been treated, the CTE mismatch between ®bre and matrix is correct. The electron microscopy shows that the chemical bonding is weaker. Despite this weakest bonding the composite behaves in a di€erent manner. The very thin residual matrix cracks are de¯ected on the many interfaces present in the mul￾tilayer (multide¯ection) as shown in Fig. 17. The microcrack does not propagate through the multilayer as in an homogeneous material. Multide¯ection is meanwhile not abundant and also, ®nal de¯ection sys￾tematically occurs at the ®rst interface. These latter fea￾tures when compared to those obtained with the strong interface in Section 3.4 are supposed to be due to a mixed regime (adhesive and cohesive failure of the interphase). 5. Concluding remarks This work has compared (PyC/SiC)n multilayers at two di€erent scales: micrometric scale materials were obtained by I-CVI and nanometric ones by P-CVI. Morphology and behaviour of these two multilayers under mechanical loading, were compared by means of a TEM examination before and after tensile loading test. From the data reported in Sections 3 and 4, several remarks can be drawn. 1. Silicon carbide sublayers morphology obtained by P-CVI are much more homogeneous and Fig. 15. TEM cross-section of a minicomposite reinforced with a pristine Hi-Nicalon ®bre (arrows: debonding due to ®bre contraction). Fig. 16. SEM longitudinal section showing the de¯ection of a large matrix crack at the ®bre surface and at matrix/multilayer interface (non treated Hi-Nicalon reinforced minicomposite) (according to Ref. 20). Fig. 17. TEM longitudinal section (treated Hi-Nicalon reinforced minicomposite): multide¯ection in the PyC/SiC multilayered inter￾faces. S. Bertrand et al. / Journal of the European Ceramic Society 20 (2000) 1±13 11