can ceramic either by an increase in the interfacial shear stress"). The predicted appearance of the fracture sur formation of strong interface bonding. If globa sharing consistent with the observations in Figs 4 and 5 cannot occur, the composite behaves like a brittle material in the It is plausible that it may take a certain temperature and time core; matrix cracking can penetrate the fibers before the interface damage reaches the critical state, where fiber But why has the damage developed differently in the center of slip can no longer occur. Thus, in the initial stages, fatigue damage the cross section than at the edges? The answer is not obvious, may increase slowly, without causing fiber failures ince for ID composites the macroscopic stress state(uniaxial The fatigue mechanism postulated above may in principle oncentration,and temperature field. The latter may vary across the specimen cross section, since energy is lost at the surface by may be understood in terms of the local temperature at the fiber/matrix interface, which may scale roughly with the energy dissipation. Both a larger stress range and a higher loading frequency increase the frictional energy dissipation(per unit time), (3) A Fatigue Damage Mechanism: Embrittlement Due to resulting in a higher temperature rise. This may accelerate the Internal Heating increase in T, and shorten the fatigue life Assume that the interfacial shear stress can increase or a strong (2)It has been found that thicker (32 ply) test specimens bonding can form with increasing temperature and time. Next, possess a lower fatigue life than thinner(8-ply) specimens. The imagine a situation where composite cross section (in the localized proposed fatigue mechanism indicates that the occurrence of region) consists of three domains(see Fig. 9): The core area atigue failure will depend on geometry and the thermal boundary (domain D), in which the fibers are broken(initially, there may be conditions. It is likely that, for identical loading conditions, thicker no broken fibers in domain I; it may start from a crack in a specimens possess a higher temperature rise in the center of the matrix-rich region). Outside the core area is a transition zone cross section than thinner specimens; temperature-induced damage (domain In) where the stresses at the fibers are now higher than the is likely to progress at a faster rate nominal value, since there is local load sharing(stress concentra- tion), and domain Ill, where intact fibers experience global load sharing due to a lower value of T Fatigue damage may evolve as follows. During cyclic loading (4 Temperature-Driven Interface Changes the energy dissipation is highest in domain Il, near the fibers The scenario described above is based on the assumption that located at the edge of domain I(at this location the fibers the interfacial shear stress or bond decreases initially but then perience the highest stress concentration). If the resulting increases with increasing temperature and time. There may be temperature rise is sufficiently high, then after some time interfa- several causes for changes in interfacial properties as the tempe cial slip may be hindered and the fibers in the vicinity of domain ature increases. The test environment, such as humidity and I fail, transferring more stress on the surviving fibers in domain ll. oxygen, affects the sliding behavior of a CMC possessing a carbon These fibers are now subjected to the highest stresses and interphase. 40. Obviously, such phenomena may depend on tem- temperature. In this manner, domain I can extend across the cross perature. Probably the most documented phenomenon for em- section of the specimen; the mechanism is self-sustaining. The composite fails when the remaining cross section cannot carry the brittlement of CMCs is oxidation damage. If the temperature at the nterface rises above approximately 200C, the C-interphase layer applied load. Then the fracture surface will show an to co CO interior region with no fiber pull-out(corresponding to domain D), Thomas and Sanches 2). If the temperature at the interface exceeds and an exterior region displaying usual fiber pull-out(note that bout 700C, the Sic fibers may decompose and form SiO, at the shorter fiber pull-out lengths reflect a higher interfacial shear interface. Since burn-off of the C-layer would reduce the mismatch between the fibers and matrix, it is likely that the interfacial shear stress T decreases. On the other hand. the formation of Sio, results in strong bonding It is well known that strong interfacial bonding results in brittle behavior of CMCs. 423,25,42 It cannot be ruled out that fatigue failure may be caused by the formation of a strong SiO, bond ( note that locally at contact points along the fiber/matrix interface, where the frictional energy dissipation takes place, the temperature may be much higher than the surrounding bulk temperature) IAFT Broken Fibers No Slip For the particular material system examined( 8-ply unidirec- tional SiC/CAs II)and test conditions(room-temperature cycling between g 240 MPa and omin 10 MPa at 200 Hz) the following were found DomainI Domain l Domain Ill (1) The residual strength of specimens cycled to 10 cycle was similar to the tensile strength of virgin specimens. Conse- quently, measurements of residual strength(e.g, after 10 cycles) cannot be used as a predictor of fatigue failure, since the strength decrease seems to take place only shortly before the occurrence of fatigue failure. Also, the retained strength raises doubts about the X alidity of GLs models, which predict a significant Fig 9. Schematic drawing of the pr reduction if the interfacial shear stress decreases Fibers fail when the interfacial shear becomes so high that fiber/ (2) The results suggest that frictional interface ng was matrix slip is hindered. The assumed va of the interfacial shear stress hindered within the specimen core, causing brittle fracture in the center of cross sections of the specimenseither by an increase in the interfacial shear stress or by the formation of strong interface bonding. If global load sharing cannot occur, the composite behaves like a brittle material in the core; matrix cracking can penetrate the fibers. But why has the damage developed differently in the center of the cross section than at the edges? The answer is not obvious, since for 1D composites the macroscopic stress state (uniaxial tension) does not vary across the cross section. Other parameters may vary across the cross section, such as humidity, oxygen concentration, and temperature field. The latter may vary across the specimen cross section, since energy is lost at the surface by radiation and convection. (3) A Fatigue Damage Mechanism: Embrittlement Due to Internal Heating Assume that the interfacial shear stress can increase or a strong bonding can form with increasing temperature and time. Next, imagine a situation where composite cross section (in the localized region) consists of three domains (see Fig. 9): The core area (domain I), in which the fibers are broken (initially, there may be no broken fibers in domain I; it may start from a crack in a matrix-rich region). Outside the core area is a transition zone (domain II) where the stresses at the fibers are now higher than the nominal value, since there is local load sharing (stress concentra￾tion), and domain III, where intact fibers experience global load sharing due to a lower value of t. Fatigue damage may evolve as follows. During cyclic loading the energy dissipation is highest in domain II, near the fibers located at the edge of domain I (at this location the fibers experience the highest stress concentration). If the resulting temperature rise is sufficiently high, then after some time interfa￾cial slip may be hindered and the fibers in the vicinity of domain I fail, transferring more stress on the surviving fibers in domain II. These fibers are now subjected to the highest stresses and temperature. In this manner, domain I can extend across the cross section of the specimen; the mechanism is self-sustaining. The composite fails when the remaining cross section cannot carry the maximum applied load. Then the fracture surface will show an interior region with no fiber pull-out (corresponding to domain I), a transition region with short fiber pull-out lengths (domain II), and an exterior region displaying usual fiber pull-out (note that shorter fiber pull-out lengths reflect a higher interfacial shear stress8 ). The predicted appearance of the fracture surface is consistent with the observations in Figs. 4 and 5. It is plausible that it may take a certain temperature and time before the interface damage reaches the critical state, where fiber slip can no longer occur. Thus, in the initial stages, fatigue damage may increase slowly, without causing fiber failures. The fatigue mechanism postulated above may in principle explain the following experimental observations: (1) A larger stress range and higher loading frequency shorten the fatigue life of CMCs with weakly bonded interfaces.20,38 This may be understood in terms of the local temperature at the fiber/matrix interface, which may scale roughly with the energy dissipation.22 Both a larger stress range and a higher loading frequency increase the frictional energy dissipation (per unit time), resulting in a higher temperature rise. This may accelerate the increase in t, and shorten the fatigue life. (2) It has been found that thicker (32 ply) test specimens possess a lower fatigue life than thinner (8-ply) specimens.39 The proposed fatigue mechanism indicates that the occurrence of fatigue failure will depend on geometry and the thermal boundary conditions. It is likely that, for identical loading conditions, thicker specimens possess a higher temperature rise in the center of the cross section than thinner specimens; temperature-induced damage is likely to progress at a faster rate. (4) Temperature-Driven Interface Changes The scenario described above is based on the assumption that the interfacial shear stress or bond decreases initially but then increases with increasing temperature and time. There may be several causes for changes in interfacial properties as the temper￾ature increases. The test environment, such as humidity and oxygen, affects the sliding behavior of a CMC possessing a carbon interphase.40,41 Obviously, such phenomena may depend on tem￾perature. Probably the most documented phenomenon for em￾brittlement of CMCs is oxidation damage. If the temperature at the interface rises above approximately 200°C, the C-interphase layer may begin to disappear by oxidation to CO or CO2 (see, e.g., Thomas and Sanches42). If the temperature at the interface exceeds about 700°C, the SiC fibers may decompose and form SiO2 at the interface.42 Since burn-off of the C-layer would reduce the mismatch between the fibers and matrix, it is likely that the interfacial shear stress t decreases. On the other hand, the formation of SiO2 results in strong bonding.10,11,42 It is well known that strong interfacial bonding results in brittle behavior of CMCs.14,23,25,42 It cannot be ruled out that fatigue failure may be caused by the formation of a strong SiO2 bond (note that locally at contact points along the fiber/matrix interface, where the frictional energy dissipation takes place, the temperature may be much higher than the surrounding bulk temperature). V. Conclusions For the particular material system examined (8-ply unidirec￾tional SiCf /CAS II) and test conditions (room-temperature cycling between smax 5 240 MPa and smin 5 10 MPa at 200 Hz) the following were found: (1) The residual strength of specimens cycled to 105 cycles was similar to the tensile strength of virgin specimens. Conse￾quently, measurements of residual strength (e.g., after 105 cycles) cannot be used as a predictor of fatigue failure, since the strength decrease seems to take place only shortly before the occurrence of fatigue failure. Also, the retained strength raises doubts about the validity of GLS models, which predict a significant strength reduction if the interfacial shear stress decreases. (2) The results suggest that frictional interface sliding was hindered within the specimen core, causing brittle fracture in the center of cross sections of the specimens. Fig. 9. Schematic drawing of the proposed fatigue damage mechanism. Fibers fail when the interfacial shear stress becomes so high that fiber/ matrix slip is hindered. The assumed variation of the interfacial shear stress is indicated. 1474 Journal of the American Ceramic Society—Sørensen et al. Vol. 83, No. 6