September 1998 Multilayered Interphases in SiC/SiC CH Composites with"Weak"and" Strong" Interfaces 2325 trongly bonded fibers, however, significant wear was ol lear decrease in the interfacial sh served only when the fibers slid over long distances(approxi- om push-back tests, the load transf mately ten fiber diameters) t than with a weak bond. Moreover the load transfer ma Discussion to be enhanced by the presence of fragments of phase that interfere with fiber ()Composites Reinforced with Weakly Bonded Fibers As commonly observed for the SiC/SiC composites pro- V. Conclusion interfacial sequence is, in fact, the fiber/first-carbon-layer in The single-fiber push-out test seems to be a useful technique terface, 9 The presence of a carbon interphase exerts a signifi- not only for determining the characteristics of interfaces and cant influence on the residual stresses. The clamping stress nterphases but also for giving insight into the micromechanics of fiber-interphase-matrix interactions. This technique has cluding the compressive effect of surface roughness. A rough- been applied to SiC/SiC composites with multilayered inter- ness amplitude A of 50-60 nm was estimated from the results ases and reinforced with either as-received or treated fibers of push-out tests, assuming a residual stress of -50 MPa. The By modifying the fiber surface, it is possible to develop a value of A measured on as-received Nicalon fibers using strong fiber-coating bond. Thus, the coating itself becomes the atomic force microscopy(AFM) is -30 nm. 32 Therefore, the weakest constituent, favoring deflection of the incoming matrix contribution of surface roughness derived from Eq (3)seems cracks within the coating and further branching into an array of to be realistic. The results indicate that the effect of roughness nanometer-scale cracks along parallel paths that correspond to cannot be neglected. Equation(3)shows that, even for fibers the carbon sublayers ith a radius as small as 7 pm, the contribution of roughne For composites reinforced with as-received fibers(weakly significantly greater than the effect of thermal expansion bonded fibers). several interfacial characteristics were ex mismatch. A somewhat smaller amplitude of 40-50 nm for a tracted from the push-out curves using the Hsueh model radial residual stress of -205 MPa was determined using the However, the clamping residual stresses need to be corrected, Kerans and Parthasarathy interface model, 8 which confirms because they are highly overestimated. Introduction of rough that the contribution of the thermally induced residual stress in ness effects seemed to be a realistic solution. Nevertheless, this Hsueh's clamping stress is not dominant model requires refinement to properly account for the phenom When ce with t data measured for a carbon ena associated with the presence of ar terphase(T 5 MPa for coatings thicker than 0 Some trends in the influence of(C-SiC)layers were estab- the interfacial shear stresses estimated for multil Inter- phases may be regarded as rather high. Pertinent comparisons ished. First, Increasing the number of (C-Sic) lae xe he resistance to fiber sliding and the compressive must refer to thinner coatings(=0. 1 um), because the frictional operate on the fibers. Second, the interfacial charac sliding parameters for a multilayered interphase are determined seem to be related to the thickness of the first carbon la by the first carbon layer surrounding the fiber the fiber The trends in the influence of multilayered interphases on The phenomena involved in the sliding of an initiall the residual stress, as evidenced by the push-out tests( Fig. 11) strongly bonded fiber are beginning to be understood. The with the residual stresses dependence on carbon inter upward curvature of the push-out curve was attributed to in- hase thickness. 29 The residual stresses on the fiber decrease creasing compression in the fiber due to very limited propaga when SiC sublayers are introduced in the carbon interphase tion of the debond, with the presence of rough crack surfaces because the total thickness of the carbon layers decreases. For reventing sliding. Interfacial properties could not be extracted two and four layers, the carbon thickness was 0.2 um. There- from the nonlinear domain of these push-out curves, using fore, the residual stresses are not influenced by the number of interface models that describe the behavior of weak interfaces sublayers However, a few sliding characteristics were extracted from the plateau of the curves (2) Composites Reinforced with Relatively Strongly The contribution of crack surface rough Bonded fibers sive stresses that are acting on the fiber during fiber sliding For the composites with a strong fiber-carbon coating bond cannot be neglected. The contribution of roughness is sub- it was shown previously-that there is not a single interfacial stantially greater than that of thermally induced residual crack running along the fiber; instead, there are several cracks stresses On the basis of the rough surfaces that were observed that branch within the interphase. The push-out behavior is in on the debonded fibers in composites with initially strongly agreement with these crack patterns, which exhibited several bonded fibers, it may be concluded that the contribution iginal features when compared with composites reinforced roughness due to a partial tearing of the interphase has been vith as-received fibers: the load necessary for fiber debonding underestimated 4 times higher, and much-higher loads are required to propa Finally, it is worth noting that the T estimates are in agree- gate the debond cracks. These features illustrate how difficult ment with those derived from hysteresis loops measured during it is for the fiber to slide. Fibers could not be pushed through tensile tests. ,6 Agreement is excellent for the composites re- samples thicker than 190 um inforced with treated fibers. The presence of short debonds The interfacial shear stresses(extracted from the plateau )are when the fiber/coating bond is strong, as observed on test 10 times larger than those obtained for the composites with specimens under tensile loading, is confirmed by the present weakly bonded fibers. This phenomenon can be logically at- tributed to sliding of the fiber along a tortuous surface that seen for composites with strong interfaces involves locking-unlocking phenomena. The push-out curves show stiffness increases that reflect higher load transfer after Acknowledgment: The authors are undebted to C. Droillard for pro- the initiation of debonding and a larger Poisson expansion the fiber in the matrix sheath In addition to crack branching, fiber sliding may also be evented by locking-unlocking, because of the roughness of the crack surfaces. Indeed. it has been demonstrated that the References force increases at the tip of a tilted and/or twisted IA. G. Evans and D. B. Marshall. ."The Mechanical Behavior of Ceramic la metall.,37[o]2567-83(1989) R.J. Kerans, R.S. Hay, N J. Pagano, and T. A. Parthasarathy, The Rolestrongly bonded fibers, however, significant wear was ob￾served only when the fibers slid over long distances (approxi￾mately ten fiber diameters).14 V. Discussion (1) Composites Reinforced with Weakly Bonded Fibers As commonly observed for the SiC/SiC composites pro￾cessed with as-received Nicalon fibers, the weakest link in the interfacial sequence is, in fact, the fiber/first-carbon-layer in￾terface.8,9 The presence of a carbon interphase exerts a signifi￾cant influence on the residual stresses. The clamping stress derived from push-out tests was corrected accordingly by in￾cluding the compressive effect of surface roughness. A rough￾ness amplitude A of 50–60 nm was estimated from the results of push-out tests, assuming a residual stress of −50 MPa. The value of A measured on as-received Nicalon fibers using atomic force microscopy (AFM) is ∼30 nm.32 Therefore, the contribution of surface roughness derived from Eq. (3) seems to be realistic. The results indicate that the effect of roughness cannot be neglected. Equation (3) shows that, even for fibers with a radius as small as 7 mm, the contribution of roughness is significantly greater than the effect of thermal expansion mismatch. A somewhat smaller amplitude of 40–50 nm for a radial residual stress of −205 MPa was determined using the Kerans and Parthasarathy interface model,18 which confirms that the contribution of the thermally induced residual stress in Hsueh’s clamping stress16 is not dominant. When comparing with t data measured for a single carbon interphase (t # 5 MPa for coatings thicker than 0.3 mm15,26,27), the interfacial shear stresses estimated for multilayered inter￾phases may be regarded as rather high. Pertinent comparisons must refer to thinner coatings (#0.1 mm), because the frictional sliding parameters for a multilayered interphase are determined by the first carbon layer surrounding the fiber. The trends in the influence of multilayered interphases on the residual stress, as evidenced by the push-out tests (Fig. 11), agree with the residual stresses dependence on carbon inter￾phase thickness.29 The residual stresses on the fiber decrease when SiC sublayers are introduced in the carbon interphase because the total thickness of the carbon layers decreases. For two and four layers, the carbon thickness was 0.2 mm. There￾fore, the residual stresses are not influenced by the number of sublayers. (2) Composites Reinforced with Relatively Strongly Bonded Fibers For the composites with a strong fiber–carbon coating bond, it was shown previously5–7 that there is not a single interfacial crack running along the fiber; instead, there are several cracks that branch within the interphase. The push-out behavior is in agreement with these crack patterns, which exhibited several original features when compared with composites reinforced with as-received fibers: the load necessary for fiber debonding is 4 times higher, and much-higher loads are required to propa￾gate the debond cracks. These features illustrate how difficult it is for the fiber to slide. Fibers could not be pushed through samples thicker than 190 mm. The interfacial shear stresses (extracted from the plateau) are ∼10 times larger than those obtained for the composites with weakly bonded fibers. This phenomenon can be logically at￾tributed to sliding of the fiber along a tortuous surface that involves locking–unlocking phenomena. The push-out curves show stiffness increases that reflect higher load transfer after the initiation of debonding and a larger Poisson expansion of the fiber in the matrix sheath. In addition to crack branching, fiber sliding may also be prevented by locking–unlocking, because of the roughness of the crack surfaces. Indeed, it has been demonstrated that the driving force increases at the tip of a tilted and/or twisted crack.33 Despite a clear decrease in the interfacial shear stress ex￾tracted from push-back tests, the load transfer across the debonded surfaces remains more efficient than in composites with a weak bond. Moreover, the load transfer may be expected to be enhanced by the presence of fragments of a torn inter￾phase that interfere with fiber sliding. VI. Conclusion The single-fiber push-out test seems to be a useful technique not only for determining the characteristics of interfaces and interphases but also for giving insight into the micromechanics of fiber–interphase–matrix interactions. This technique has been applied to SiC/SiC composites with multilayered inter￾phases and reinforced with either as-received or treated fibers. By modifying the fiber surface, it is possible to develop a strong fiber–coating bond. Thus, the coating itself becomes the weakest constituent, favoring deflection of the incoming matrix cracks within the coating and further branching into an array of nanometer-scale cracks along parallel paths that correspond to the carbon sublayers. For composites reinforced with as-received fibers (weakly bonded fibers), several interfacial characteristics were ex￾tracted from the push-out curves using the Hsueh model.16 However, the clamping residual stresses need to be corrected, because they are highly overestimated. Introduction of rough￾ness effects seemed to be a realistic solution. Nevertheless, this model requires refinement to properly account for the phenom￾ena associated with the presence of an interphase. Some trends in the influence of (C–SiC) layers were estab￾lished. First, increasing the number of (C–SiC) layers increases the resistance to fiber sliding and the compressive stresses that operate on the fibers. Second, the interfacial characteristics seem to be related to the thickness of the first carbon layer on the fiber. The phenomena involved in the sliding of an initially strongly bonded fiber are beginning to be understood. The upward curvature of the push-out curve was attributed to in￾creasing compression in the fiber due to very limited propaga￾tion of the debond, with the presence of rough crack surfaces preventing sliding. Interfacial properties could not be extracted from the nonlinear domain of these push-out curves, using interface models that describe the behavior of weak interfaces. However, a few sliding characteristics were extracted from the plateau of the curves. The contribution of crack surface roughness to the compres￾sive stresses that are acting on the fiber during fiber sliding cannot be neglected. The contribution of roughness is sub￾stantially greater than that of thermally induced residual stresses. On the basis of the rough surfaces that were observed on the debonded fibers in composites with initially strongly bonded fibers, it may be concluded that the contribution of roughness due to a partial tearing of the interphase has been underestimated. Finally, it is worth noting that the t estimates are in agree￾ment with those derived from hysteresis loops measured during tensile tests.5,6 Agreement is excellent for the composites re￾inforced with treated fibers. The presence of short debonds when the fiber/coating bond is strong, as observed on test specimens under tensile loading, is confirmed by the present analysis. Therefore, an improved fatigue lifetime may be fore￾seen for composites with strong interfaces. Acknowledgment: The authors are undebted to C. Droillard for pro￾viding the samples and fruitful discussion. References 1 A. G. Evans and D. B. Marshall, ‘‘The Mechanical Behavior of Ceramic Matrix Composites,’’ Acta Metall., 37 [10] 2567–83 (1989). 2 R. J. Kerans, R. S. Hay, N. J. Pagano, and T. A. Parthasarathy, ‘‘The Role September 1998 Multilayered Interphases in SiC/SiC CVI Composites with ‘‘Weak’’ and ‘‘Strong’’ Interfaces 2325