ournal Am Ceram So, 81 14 965-78(1998 Interfacial Bond Strength in SiC/C/SiC Composite Materials, As Studied by Single-Fiber Push-Out Tests Francis Rebillat, Jacques Lamon, and Roger Naslain Laboratoire des Composites Thermostructuraux, UMR 47, CNRS-SEP-UB1, LCTS, 33600 Pessac, France Edgar Lara-Curzio, Mattison K. Ferber, and Theodore M. Besmann Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831-6064 The interfacial characteristics of SiC/C/SiC composites characterized by a transverse anisotropic microstructure(typi- with different fiber-coating bond strengths have been ir cally pyrolytic carbon(pyC)or boron nitride(BN))and a low vestigated using single-fiber push-out tests. Previous stud shear modulus(40 GPa). Carbon seems to be the most effec- ies have shown that weak or strong bonds can be obtained tive. However, in Nicalon M(Nippon Carbon, Tokyo, Japan) fiber/SiC matrix combinations. the weakest link is not neces the stress-strain behavior is improved with the treated fi- sarily the interphase. Depending on the fiber-coating bond, 9 bers. This effect results from multiple branching of the the fiber/coating interface, instead of the interphase, may act as cracks within the interphase. The model used to extract a mechanical fuse. In addition, the thermally induced radial interfacial characteristics from nanoindentation and micro- stresses are partially relieved by the interphase. 0, I The inter- indentation tests does not consider the presence of an in- facial bond ultimately influences various features of the com- terphase. However, the results highlight the significant ef- posite mechanical behavior, including ultimate strength, non fect of the interphase on the interfacial parameters, as well linear deformations, Youngs modulus(E), interlaminar shear as the effect of roughness along the sliding surfaces. For the strength, fracture toughness, and compressive strength. Thus, composite with treated fibers, the uncommon upward cur- the properties of the interfacial bond should be tailored to vature of the push-out curves is related to different modes maximize composite performance of crack propagation in the interphase. Different tech- It has been shown, on the basis of fracture statistics, that the niques are required to analyze the interfacial properties, number of matrix cracks increases as the interfacial shear stress such as nanoindentation and microindentation with push increases. As a consequence, saturation of matrix cracking out and push-back tests. has a tendency to occur close to ultimate failure. Furthermore, the highest stresses have been obtained when the interfacial L. Introduction shear stress is relatively high. In contrast, for low interfacial shear stresses, the tensile strain-strain behavior exhibits a pla B ASED on many experimental and theorical studies, it has Following the idea that strong interfacial bonding enhances been established that the mechanical behavior of ceramic composite properties, a new family of SiC/C/SiC materials was matrix composites is dependent not only on the intrinsic prop designed and processed with strengthened fiber-interphase interface. 1-3 Indeed, the nonlinear stress-strain behavior of a treated surface, 14, Is and the composites exhibit higher strength ceramic-fiber-ceramic-matrix composite is related to the de- and toughness, in comparison to their counterparts that have models-6 note the importance of frictional sliding along the been fabricated from as-received fibers 13, 16, 1 In the present paper, the interface and interphase character- debonded interface. Thus, a tough composite is described as istics of strongly bonded two-dimensional(2D)SiC/SiC com- ng fiber frictional sliding over long debond dis- tances, which results in significant fiber pullout and bridging of technique is the most appropriate method to directly measure the cracks propagating through the matrix. Generally, to en- the interfacial properties of these 2D woven composites. How- ourage the deviation of matrix cracks. the fiber-matrix bond is equired to be weak. Improved control of the fiber/matrix in- small-diameter fibers. These difficulties are compounded by terface is achieved via deposition of a layer of a compliant the sample-preparation requirements. The main advantages of material on the fibers. Appropriate interphase materials are ush-out tests are the relative simplicity of the test method and he fact that specific fibers in the composite can be probed Furthermore, the mathematics involved in the pull-out and R. Kerans--contributing editor ush-out models are equivalent, despite a difference in the sign of the Poisson's effect (i.e, during a push-out test, the fiber expands under compression, thus increasing the sliding stress, whereas it contracts under tension, thus reducing the sliding No. 191641. Received July The intent of the present paper is(i)to estimate ssistant Secretary les(as part of the either with as-received or treated fibers using single-f out tests, (ii) to establish correlations between pr lo. DE-AC05-960R22464 with and the stress-strain behavior determined previously, (iii) and Member. American mic Society to assess the concept of strong fiber-matrix bonding
Interfacial Bond Strength in SiC/C/SiC Composite Materials, As Studied by Single-Fiber Push-Out Tests Francis Rebillat, Jacques Lamon,* and Roger Naslain* Laboratoire des Composites Thermostructuraux, UMR 47, CNRS-SEP-UB1, LCTS, 33600 Pessac, France Edgar Lara-Curzio, Mattison K. Ferber, and Theodore M. Besmann* Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831–6064 The interfacial characteristics of SiC/C/SiC composites with different fiber-coating bond strengths have been investigated using single-fiber push-out tests. Previous studies have shown that weak or strong bonds can be obtained by using as-received or treated fibers, respectively, and that the stress–strain behavior is improved with the treated fibers. This effect results from multiple branching of the cracks within the interphase. The model used to extract interfacial characteristics from nanoindentation and microindentation tests does not consider the presence of an interphase. However, the results highlight the significant effect of the interphase on the interfacial parameters, as well as the effect of roughness along the sliding surfaces. For the composite with treated fibers, the uncommon upward curvature of the push-out curves is related to different modes of crack propagation in the interphase. Different techniques are required to analyze the interfacial properties, such as nanoindentation and microindentation with pushout and push-back tests. I. Introduction BASED on many experimental and theorical studies, it has been established that the mechanical behavior of ceramicmatrix composites is dependent not only on the intrinsic properties of the constituents but also largely on the fiber/matrix interface.1–3 Indeed, the nonlinear stress–strain behavior of a ceramic-fiber–ceramic-matrix composite is related to the deflection of matrix cracks into the fiber/matrix interface. Many models4–6 note the importance of frictional sliding along the debonded interface. Thus, a tough composite is described as experiencing fiber frictional sliding over long debond distances, which results in significant fiber pullout and bridging of the cracks propagating through the matrix.7 Generally, to encourage the deviation of matrix cracks, the fiber–matrix bond is required to be weak. Improved control of the fiber/matrix interface is achieved via deposition of a layer of a compliant material on the fibers. Appropriate interphase materials are characterized by a transverse anisotropic microstructure (typically pyrolytic carbon (pyC) or boron nitride (BN)) and a low shear modulus (∼40 GPa). Carbon seems to be the most effective. However, in NicalonTM (Nippon Carbon, Tokyo, Japan) fiber/SiC matrix combinations, the weakest link is not necessarily the interphase. Depending on the fiber–coating bond,8,9 the fiber/coating interface, instead of the interphase, may act as a mechanical fuse. In addition, the thermally induced radial stresses are partially relieved by the interphase.10,11 The interfacial bond ultimately influences various features of the composite mechanical behavior, including ultimate strength, nonlinear deformations, Young’s modulus (E), interlaminar shear strength, fracture toughness, and compressive strength. Thus, the properties of the interfacial bond should be tailored to maximize composite performance. It has been shown, on the basis of fracture statistics, that the number of matrix cracks increases as the interfacial shear stress increases.12 As a consequence, saturation of matrix cracking has a tendency to occur close to ultimate failure. Furthermore, the highest stresses have been obtained when the interfacial shear stress is relatively high. In contrast, for low interfacial shear stresses, the tensile strain–strain behavior exhibits a plateau, and saturation of matrix cracking occurs at lower stresses. Following the idea that strong interfacial bonding enhances composite properties, a new family of SiC/C/SiC materials was designed and processed with strengthened fiber–interphase bonds.13 These were obtained using Nicalon™ fibers with a treated surface,14,15 and the composites exhibit higher strength and toughness, in comparison to their counterparts that have been fabricated from as-received fibers.13,16,17 In the present paper, the interface and interphase characteristics of strongly bonded two-dimensional (2D) SiC/SiC composites13 are investigated using push-out tests. This testing technique is the most appropriate method to directly measure the interfacial properties of these 2D woven composites. However, the single-fiber push-out test is difficult to perform on small-diameter fibers. These difficulties are compounded by the sample-preparation requirements. The main advantages of push-out tests are the relative simplicity of the test method and the fact that specific fibers in the composite can be probed. Furthermore, the mathematics involved in the pull-out and push-out models are equivalent, despite a difference in the sign of the Poisson’s effect (i.e., during a push-out test, the fiber expands under compression, thus increasing the sliding stress, whereas it contracts under tension, thus reducing the sliding stress). The intent of the present paper is (i) to estimate the interfacial characteristics for 2D SiC/C/SiC composites reinforced either with as-received or treated fibers using single-fiber pushout tests, (ii) to establish correlations between push-out data and the stress–strain behavior determined previously, (iii) and to assess the concept of strong fiber–matrix bonding. R. J. Kerans—contributing editor Manuscript No. 191641. Received July 30, 1996; approved June 24, 1997. Supported by the LCTS (Pessac, France) and SEP through a grant given to author FR, as well as, at Oak Ridge National Laboratory, by the U.S. Department of Energy, Assistant Secretary for Energy Efficiency and Renewable Energy, Office of Transportation Technologies (as part of the HTML User Program), and the U.S. Department of Energy, Office of Fossil Energy, Advanced Research and Technology Development Materials Program (under Contract No. DE-AC05-96OR22464 with Lockheed Martin Energy Research). *Member, American Ceramic Society. J. Am. Ceram. Soc., 81 [4] 965–78 (1998) Journal 965
Journal of the American Ceramic Sociery-Rebillat et al. Vol 81. No 4 Il. Description of the Materials The interfacial parameters were extracted from the experi- mental-stress-versus-fiber-end-displacement curves, using the O Processing and Chemical Analysis The carbon interphase and the Sic matrix were infiltrated placement(u) predicted by the Hsueh model(after subtracting into porous fiber preforms by using the isothermal/isobaric the contribution of the loading-frame deformation) for the chemical vapor infiltration(I-CVI) process. 17 Two series of applied stress, or, is given by previously by h(a++2 Droillard: 13 the first used as-received Nicalon TM fabric(de (1 noted as " I"' in Droillard), and the second used treated Nicalon TM fabric(denoted as"J"and"P"in Droillard where aa is the stress required to initiate debonding and sliding, The free surface of untreated Nicalon TM fibers contains silica h the sliding length, and a, the axial residual stress SiO2 ). Therefore, the fiber/interphase interface is the silica amorphous )anisotropic pyC interface, whereas the surface of The sliding length h and the interfacial shear stress T can be separated into a constant and a variable component treated NicalonTM fibers is rich in free carbon and, as a result, the fiber/interphase interface may be regarded as a carbon/ T=-u(o +on) carbon interface. Such an interface exhibits a stronger bond than the silica/anisotropic pyC interface, which is generall described as weak 8,9 Two carbon thicknesses have been de ted on the fibers: 0. 1(material P)and 0.5 um(materials I and J)(Table 1) where oe is the clamping residual stress, u the coefficient of friction, and o, the average contribution to the clamping stress (2) Stress-Strain Behavior due to Poissons effect. The variable o, is given by Tensile stress-strain curves for the materials were previously ported by Droillard. Two distinct behaviors, depending on l「/ V Em fv the fiber surface, were observed(Fig. 1). A plateaulike behav 2D( E I )+(m) ior was obtained for the as-received fiber-reinforced material ( material 1). However, the mechanical behavior was signifi cantly improved(material J) when fiber-matrix bonding was modified by fiber surface treatment( Fig. 1), as indicated by the following features: (i)a 50% higher failure stress, (ii)a satu- I-f+Vm+ E ration stress similar to that of ultimate failure, (iii)a very small where ve and v are the Poissons ratios for the fiber and crack-spacing distance at saturation(\s 20-30 um, versus matrix, respectively. A uniform Poisson's effect, characterized 115-300 um for the composite reinforced with untreated fi- by an average stress o, is assumed along the sliding interface bers),(iv)small residual deformation, and (v) narrow hyster- Hence, an average interfacial shear stress, T, is determined and esis loops when unloading-reloading was performed(approxi he stress in the fiber, or varies linearly along the sliding mately one-tenth of the width observed for the composites with length, from the applied stress o at the surface to the debond as-received Nicalon TM fibers). Short crack-spacing distances, stress od at the end of the sliding zone: 0 =(0+ gdv as well as small hysteresis loop widths, reflect strong fiber matrix bonding. Examination of interfacial regions via trans- varies with the sliding distance. During crack propagation,the mission electron microscopy(TEM) revealed a complex net axial stress in the fiber at the end of the sliding zone is always work of cracks branching into microcracks within the in equilibrium with o interphase in the treated fiber-reinforced composites(upper in B)Push-Back Tests: By turning over the sample, push- set in Fig. 1), whereas a single crack that propagated along the back tests can be conducted on the pushed-out fibers. Because fiber surface over a long distance was generally observed thes nly frictional for the osites with as-received fibers(lower inset characterized from the stress-versus-displacement curves. This technique is an alternative method to extract the sliding char- acteristics. However, during reverse sliding of the fibers, wear lIL. Determination of Interfacial Characteristics underestimation of interphase/interface properties. Neve of the sliding interphases may occur, which leads to significar Microindentation Tests theless, these tests can provide useful data on the compos- ites reinforced with treated fibers that complement the data (A) Push-Out Tests: A wedge specimen geometry was rmined by the push-out the specimens only needed to be half that commonly used for (2) Nanoindentation Tests oush-in tests, 9(-500 um) to push the fibers out. The tests Another method for fiber push testing uses a nanoindentor (ORNL) Interfacial Test System 20 at a constant displacement (MPM)(Nano Instrument, Knoxville, TN). The MPM that has rate of 0. 1 um/s. A flat-end diamond tip was used to push the been used for the present study allows an accurate application fl of loads in the millinewton range via a Berkovich pyramidal Table L. Investigated Materials and Main Tensile Mechanical Properties Failure Maximum Material cMi Nontreated 241 1.07 183 185 15.2-21.3 Treate 356 24-293 Treated 353 20-30
II. Description of the Materials (1) Processing and Chemical Analysis The carbon interphase and the SiC matrix were infiltrated into porous fiber preforms by using the isothermal/isobaric chemical vapor infiltration (I-CVI) process.17 Two series of 2D-SiC/C/SiC composites have been prepared previously by Droillard:13 the first used as-received Nicalon™ fabric (denoted as ‘‘I’’ in Droillard13), and the second used treated Nicalon™ fabric14 (denoted as ‘‘J’’ and ‘‘P’’ in Droillard13). The free surface of untreated Nicalon™ fibers contains silica (SiO2). Therefore, the fiber/interphase interface is the silica (amorphous)/anisotropic pyC interface, whereas the surface of treated Nicalon™ fibers is rich in free carbon and, as a result, the fiber/interphase interface may be regarded as a carbon/ carbon interface.13 Such an interface exhibits a stronger bonding than the silica/anisotropic pyC interface, which is generally described as weak.8,9 Two carbon thicknesses have been deposited on the fibers: 0.1 (material P) and 0.5 mm (materials I and J) (Table I). (2) Stress–Strain Behavior Tensile stress–strain curves for the materials were previously reported by Droillard.13 Two distinct behaviors, depending on the fiber surface, were observed (Fig. 1). A plateaulike behavior was obtained for the as-received fiber-reinforced material (material I). However, the mechanical behavior was significantly improved (material J) when fiber–matrix bonding was modified by fiber surface treatment (Fig. 1), as indicated by the following features: (i) a 50% higher failure stress, (ii) a saturation stress similar to that of ultimate failure, (iii) a very small crack-spacing distance at saturation (ls 4 20–30 mm, versus 115–300 mm for the composite reinforced with untreated fibers), (iv) small residual deformation, and (v) narrow hysteresis loops when unloading–reloading was performed (approximately one-tenth of the width observed for the composites with as-received Nicalon™ fibers). Short crack-spacing distances, as well as small hysteresis loop widths, reflect strong fiber– matrix bonding. Examination of interfacial regions via transmission electron microscopy (TEM) revealed a complex network of cracks branching into microcracks within the interphase in the treated fiber-reinforced composites (upper inset in Fig. 1), whereas a single crack that propagated along the fiber surface over a long distance was generally observed for the composites with as-received fibers (lower inset in Fig. 1).16,17 III. Determination of Interfacial Characteristics (1) Microindentation Tests (A) Push-Out Tests: A wedge specimen geometry was used for the indentation tests.18 However, the thickest part of the specimens only needed to be half that commonly used for push-in tests18,19 (∼500 mm) to push the fibers out. The tests were conducted using the Oak Ridge National Laboratory (ORNL) Interfacial Test System20 at a constant displacement rate of 0.1 mm/s. A flat-end diamond tip was used to push the fibers. The interfacial parameters were extracted from the experimental-stress-versus-fiber-end-displacement curves, using the push-out model proposed by Hsueh.21,22 The fiber-end displacement (u) predicted by the Hsueh model (after subtracting the contribution of the loading-frame deformation) for the applied stress, s, is given by u = h~s + sd + 2sz! 2Ef (1) where sd is the stress required to initiate debonding and sliding, h the sliding length, and sz the axial residual stress. The sliding length h and the interfacial shear stress t can be separated into a constant and a variable component: t = −m~sc + sp! (2) h = r~sd − s! 2t (3) where sc is the clamping residual stress, m the coefficient of friction, and sp the average contribution to the clamping stress due to Poisson’s effect. The variable sp is given by sp = 1 2D FS nf Em Ef − f nm 1 − f Ds + S nf Em Ef − fnm 1 − f DsdG (4) with D = 1 + f 1 − f + nm + ~1 − nf!Em Ef (5) where nf and nm are the Poisson’s ratios for the fiber and matrix, respectively. A uniform Poisson’s effect, characterized by an average stress sp, is assumed along the sliding interface. Hence, an average interfacial shear stress, t, is determined and the stress in the fiber, sf , varies linearly along the sliding length, from the applied stress s at the surface to the debond stress sd at the end of the sliding zone: s 4 (s + sd)/2. However, it is noted that the average interfacial shear stress varies with the sliding distance. During crack propagation, the axial stress in the fiber at the end of the sliding zone is always in equilibrium with sd. (B) Push-Back Tests: By turning over the sample, pushback tests can be conducted on the pushed-out fibers. Because these fibers are already debonded, only frictional sliding is characterized from the stress-versus-displacement curves. This technique is an alternative method to extract the sliding characteristics. However, during reverse sliding of the fibers, wear of the sliding interphases may occur, which leads to significant underestimation of interphase/interface properties. Nevertheless, these tests can provide useful data on the composites reinforced with treated fibers that complement the data determined by the push-out test. (2) Nanoindentation Tests Another method for fiber push testing uses a nanoindentor that is also called the Mechanical Properties Microprobe (MPM) (Nano Instrument, Knoxville, TN). The MPM that has been used for the present study allows an accurate application of loads in the millinewton range via a Berkovich pyramidal Table I. Investigated Materials and Main Tensile Mechanical Properties† Material Nature of the fabric Interphase thickness (mm) Failure stress (MPa) Failure strain (%) Young’s modulus (GPa) Matrix crack spacing at saturation (mm) Interfacial shear stress (MPa) Maximum strain energy (kJ/m2 ) I Nontreated 0.5 241 1.07 183 185 4 15.2–21.3 J Treated 0.5 356 1.00 170 20 370 24–29.3 P Treated 0.1 P Treated 0.1 353 0.9 250 20 210 20–30 † From Droillard.13 966 Journal of the American Ceramic Society—Rebillat et al. Vol. 81, No. 4
April 1998 Interfacial Bond Strength in SiC/C/SiC Composite Materials 3z 学[m 厂hr LONGITUDINAL TENSILE STRAIN(%) Fig. 1. Tensile stress-strain reinforced with as-received (I)or treated ( NicalonTM fibers, ac diamond indentor that has the same depth area ratio as a Vick (3) Alternate Methods of Estimation of ers diamond indentor. The accuracy of measurements with this Interfacial Parameters from Push-Out Curves quipment is considerably higher than that for microindenta (A) Interfacial Sliding Stress: Bright et al.23 measured the tion. However, the maximum load that can be applied is-0 interface debonding load and the maximum sliding load for N. The fiber-loading history involves the application of a pre- various embedded lengths of fiber. The observed dependence determined constant displacement (or loading)rate, up to a of the maximum sliding load on the initial embedded lengths naximum displacement(or force), followed by a constant un fibers was satisfactorily described by the nonlinear shear lag ding until 95% of the force is removed. Magnitudes of the model. The interfacial properties, including the sliding stress, force and displacement are continuously measured with reso- are derived from the maximum stress(omax)using a regression tions of 2, 4 un and 0.4 nm, respectively. The tests are procedure based on the following equati performed under displacement control as well as load control The sliding distance during nanoindentation tests is gener- ally much smaller than the specimen thickness. The fiber is pushed into the matrix but not through it( the same sample Values for the coefficient of friction, H, of the clamping with geometry as that for microindentation was used esidual stress and for the axial residual stress, oa, can be extracted by fitting the Hsueh equations to the loader k=Emr unloading curves. During unloading. fiber recovery as terized by the ratio of the residual to the maximum displace ependent on alues axial residual stress in the fiber and the poisson's ratio of the fiber(ve. The fiber-end displacement is predicted by the Hsueh model, using an expression similar to Eq. (1) The analysis is independent of the shape of the indentation Because of the pyramidal shape of the diamond indentor, the curve; therefore, it can also be applied to composites that have measured displacement must be corrected, due to indentor pen- strong interfacial bonds (B) Interfacial Shear Strength: Because the stress field surface of the loaded fibers). The depth of diamond penetration along the debond must be approximated in most cases, the applied load. Because the latter parameters evolve during the interfacial cracking, is extracted directly from the debonding test, there is no way of establishing a simple relation for diamond penetration versus the applied load tions:24-27(i) the interfacial shear stress decreases from the A method based on the indentation of transverse fibers was surface of the sample, (ii) debonding initiates when the maxi- used to determine the load-versus-diamond-penetratio mum shear stress exceeds the shear strength of the interface distance curve. Many reproducible indentation tests were per and(iii) the examined cell consists of a fiber embedded in ar formed perpendicular to the fiber axis. Then, an empirical re infinite matrix with a free surfa ace lation for diamond penetration distance was obtained by fitting Assuming that the stresses in the fiber are proportional to the ifference between the axial displacements of the fiber and unloading curves. Thus, it was possible to subtract the diamond of the matrix, considered independently, a simple analytica penetration distance from the displacements that were mea- equation for the initial debond stress Ts has been derived: 25,26 sured during the nanoindentation tests. Only the unloading curves for which the deviation from linearity was the largest ogEr+vm)./R
diamond indentor that has the same depth:area ratio as a Vickers diamond indentor. The accuracy of measurements with this equipment is considerably higher than that for microindentation. However, the maximum load that can be applied is ∼0.5 N. The fiber-loading history involves the application of a predetermined constant displacement (or loading) rate, up to a maximum displacement (or force), followed by a constant unloading until 95% of the force is removed. Magnitudes of the force and displacement are continuously measured with resolutions of 2.4 mN and 0.4 nm, respectively. The tests are performed under displacement control as well as load control. The sliding distance during nanoindentation tests is generally much smaller than the specimen thickness. The fiber is pushed into the matrix but not through it (the same sample geometry as that for microindentation was used). Values for the coefficient of friction, m, of the clamping residual stress and for the axial residual stress, sa, can be extracted by fitting the Hsueh equations22 to the loading and unloading curves. During unloading, fiber recovery, as characterized by the ratio of the residual to the maximum displacements (U0/Umax) is strongly dependent on the values of the axial residual stress in the fiber and the Poisson’s ratio of the fiber (nf ). The fiber-end displacement is predicted by the Hsueh model, using an expression similar to Eq. (1). Because of the pyramidal shape of the diamond indentor, the measured displacement must be corrected, due to indentor penetration into the fiber (a pyramidal mark is observed on the top surface of the loaded fibers). The depth of diamond penetration is directly related to the fiber hardness, the indent size, and the applied load. Because the latter parameters evolve during the test, there is no way of establishing a simple relation for diamond penetration versus the applied load. A method based on the indentation of transverse fibers was used to determine the load-versus-diamond-penetrationdistance curve. Many reproducible indentation tests were performed perpendicular to the fiber axis. Then, an empirical relation for diamond penetration distance was obtained by fitting a second-order polynomial to the experimental loading and unloading curves. Thus, it was possible to subtract the diamond penetration distance from the displacements that were measured during the nanoindentation tests. Only the unloading curves for which the deviation from linearity was the largest were analyzed. (3) Alternate Methods of Estimation of Interfacial Parameters from Push-Out Curves (A) Interfacial Sliding Stress: Bright et al.23 measured the interface debonding load and the maximum sliding load for various embedded lengths of fiber. The observed dependence of the maximum sliding load on the initial embedded lengths of fibers was satisfactorily described by the nonlinear shear lag model. The interfacial properties, including the sliding stress, are derived from the maximum stress (smax) using a regression procedure based on the following equation: smax = sc k FexpS 2mkt r D − 1G (6) with k = Emnf Ef~1 + nm! (7) and t = −msc (8) The analysis is independent of the shape of the indentation curve; therefore, it can also be applied to composites that have strong interfacial bonds. (B) Interfacial Shear Strength: Because the stress field along the debond must be approximated in most cases, the interfacial shear strength, ts, which measures the resistance to interfacial cracking, is extracted directly from the debonding stress. The few proposed models are based on similar assumptions:24–27 (i) the interfacial shear stress decreases from the surface of the sample, (ii) debonding initiates when the maximum shear stress exceeds the shear strength of the interface, and (iii) the examined cell consists of a fiber embedded in an infinite matrix with a free surface. Assuming that the stresses in the fiber are proportional to the difference between the axial displacements of the fiber and of the matrix, considered independently, a simple analytical equation for the initial debond stress ts has been derived:25,26 ts = − sd 2 F Ef~1 + nm! Em ln S R rDG −1/2 1 tanh (bt) (9) Fig. 1. Tensile stress–strain curves for 2D SiC/SiC composites with carbon interphases (0.5 mm thick) reinforced with as-received (I) or treated (J) Nicalon™ fibers, according to Droillard.13 Associated mechanisms of interface cracking are also indicated. April 1998 Interfacial Bond Strength in SiC/C/SiC Composite Materials 967
Journal of the American Ceramic Sociery-Rebillat et al. Vol 81. No 4 weak fiber/matrix interfaces, 18, 19 21, 22, 28,29 that are addressed in the push-out models. 22 28-30 It exhibits well-defined features (10) After the elastic deformation of the fiber(a-"b in Fig. 2), debonding or fiber sliding of unbonded fibers initiates at point EdI +vm)In b. In the case of unbonded fibers, the applied stress overcomes static friction. The nonlinear portion of the curve(b-c)is where r and R refer to the radius of the fiber and the matrix, characterized by a continuously decreasing slope due to pro- respectively, and t is the thickness of the sample gressive fiber debonding or the fiber overcoming static friction, Hsueh introduced the effect of Poisson s expansion of the followed by sliding. The stable growth of the interfacial crack fiber, which increases the radial compressive stresses. The in- continues up to a maximum stress at point c. Unstable debond- terfacial shear strength(for t > R)is given by the following ing of the remaining portion of bonded fiber is accompanied by equation: a sudden load drop that is observed in the decreasing portion of the curve(c-d"). Finally, the fiber is pushed out of the matrix, yielding a pseudo-plateau(d-e t== (+()会[=(=2 y 8 The average interfacial sliding stress may be estimated from plateau stre (14) where plateau is the compressive applied stress, t the embedded (12) E(1+Vm)l R In (R out model of Hsueh to the nonlinear portion of the curve("b c" in Fig. 2)through the coefficient of friction, H, the re In a more refined model. Hsueh27 added the effects of the sidual clamping stress, o and the residual axial stress in the mismatch strains between the fiber and the matrix in the radial fiber, o,(Eqs. (1H5)) and the axial directions (2) Push-Out Curves for Composites Reinforced with Treated fibers B2lexp(mt)+exp(-mt)-2] (2[exp(mt)-exp(-m)l In contrast to the classical push-out behavior shown in F 2, the push-out curves measured for the CMCs with treated A[exp(mt)+exp(mt)-2 fibers exhibit features that are not described by the interfacial aaf exp(mt)+exp(-mt)+ models. The curves(Fig 3)consist of a linear portion("a b")that corresponds to the elastic deformation of the fiber,a (13) steep load drop(b'c) that indicates the occurrence of fiber debonding, a short plateau(c) that is followed by a where B, m, and A are elastic constants and B2 is a parameter nonlinear portion("c"'d")with reversed curvature(in con- that is dependent on the thermal expansion mismatch between trast to the curves in Fig. 2), a downward small concave portion the fiber and the matrix. 27 d-e), a load drop(e-f) that coincides with fiber rotrusion,and finally a pseudo-plateau when the fiber slides IV. Results out of the matrix. As indicated above. the Hsueh model27 can- not be fitted to the nonlinear portion of the push-out curves, ( Push-Out Curves for Composites Reinforced with because of the existence of the reversed curvature. The fric- As-Received fibers tional shear stress was estimated from the plateau via Eq (15) Figure 2 shows an example of the typical stress-versus-fiber- (3)Interfacial Parameters for the end-displacement curves measured for the standard SiC/C/SiC Untreated Fiber-Reinforced Composite composite with untreated fibers. This curve is similar to those Results of the series of tests performed on composite I are obtained for other ceramic-matrix composites(CMCs)with ummarized in Table Il. previous results on the effect of 2000 1500 0 D Fig. 2. Single-fiber push-out curve measured for a SiC/C/SiC composite(sample I)reinforced with as-received NicalonTM fibers(F is the applied stress)
with b = 1 r H Em Ef~1 + nm! ln S R rDJ 1/2 (10) where r and R refer to the radius of the fiber and the matrix, respectively, and t is the thickness of the sample. Hsueh26 introduced the effect of Poisson’s expansion of the fiber, which increases the radial compressive stresses. The interfacial shear strength (for t >> R) is given by the following equation: ts = −sd 1 S R2 r 2 − 1DSEm Ef D coth ~at! + 2 exp~at! − exp~−at! S 2 rDH~1 + nm!F1 + S R2 r 2 − 1DSEm Ef DGFR2 ln S R r D − ~R2 − r 2 ! 2 GJ 1/2 2 (11) with a = 1 r 5 r 2 Ef + ~R2 − r 2 !Em Ef~1 + nm!FR2 ln S R r D − ~R2 − r 2 ! 2 G6 1/2 (12) In a more refined model, Hsueh27 added the effects of the mismatch strains between the fiber and the matrix in the radial and the axial directions: ts = S rm 2@exp~mt! − exp~−mt!# DSB2@exp~mt! + exp~−mt! − 2# B1 − sdHexp~mt! + exp~−mt! + A@exp~mt! + exp~−mt! − 2# B1 JD (13) where B1, m, and A are elastic constants and B2 is a parameter that is dependent on the thermal expansion mismatch between the fiber and the matrix.27 IV. Results (1) Push-Out Curves for Composites Reinforced with As-Received Fibers Figure 2 shows an example of the typical stress-versus-fiberend-displacement curves measured for the standard SiC/C/SiC composite with untreated fibers. This curve is similar to those obtained for other ceramic-matrix composites (CMCs) with weak fiber/matrix interfaces2,18,19,21,22,28,29 that are addressed in the push-out models.22,28–30 It exhibits well-defined features. After the elastic deformation of the fiber (‘‘a’’–‘‘b’’ in Fig. 2), debonding or fiber sliding of unbonded fibers initiates at point b. In the case of unbonded fibers, the applied stress overcomes static friction. The nonlinear portion of the curve (‘‘b’’–‘‘c’’) is characterized by a continuously decreasing slope due to progressive fiber debonding or the fiber overcoming static friction, followed by sliding. The stable growth of the interfacial crack continues up to a maximum stress at point c. Unstable debonding of the remaining portion of bonded fiber is accompanied by a sudden load drop that is observed in the decreasing portion of the curve (‘‘c’’–‘‘d’’). Finally, the fiber is pushed out of the matrix, yielding a pseudo-plateau (‘‘d’’–‘‘e’’). The average interfacial sliding stress may be estimated from the plateau stress: tplateau = splateaur 2t (14) where splateau is the compressive applied stress, t the embedded length of fiber, and r the fiber radius. As discussed previously, the interfacial shear stress is also extracted by fitting the pushout model of Hsueh to the nonlinear portion of the curve (‘‘b’’– ‘‘c’’ in Fig. 2) through the coefficient of friction, m, the residual clamping stress, sc, and the residual axial stress in the fiber, sz (Eqs. (1)–(5)).21 (2) Push-Out Curves for Composites Reinforced with Treated Fibers In contrast to the classical push-out behavior shown in Fig. 2, the push-out curves measured for the CMCs with treated fibers exhibit features that are not described by the interfacial models. The curves (Fig. 3) consist of a linear portion (‘‘a’’– ‘‘b’’) that corresponds to the elastic deformation of the fiber, a steep load drop (‘‘b’’–‘‘c’’) that indicates the occurrence of fiber debonding, a short plateau (‘‘c’’) that is followed by a nonlinear portion (‘‘c’’–‘‘d’’) with reversed curvature (in contrast to the curves in Fig. 2), a downward small concave portion (‘‘d’’–‘‘e’’), a load drop (‘‘e’’–‘‘f’’) that coincides with fiber protrusion, and finally a pseudo-plateau when the fiber slides out of the matrix. As indicated above, the Hsueh model27 cannot be fitted to the nonlinear portion of the push-out curves, because of the existence of the reversed curvature. The frictional shear stress was estimated from the plateau via Eq. (15). (3) Interfacial Parameters for the Untreated Fiber-Reinforced Composites Results of the series of tests performed on composite I are summarized in Table II. Previous results on the effect of Fig. 2. Single-fiber push-out curve measured for a SiC/C/SiC composite (sample I) reinforced with as-received Nicalon™ fibers (F is the applied stress). 968 Journal of the American Ceramic Society—Rebillat et al. Vol. 81, No. 4
April 1998 Interfacial Bond Strength in SiC/C/SiC Composite Materials 0 6000 4000 3000 thickness 186 um 2000 ,f,,,,f,,, 3 Displacement (um) ig. 3. Single-fiber push-out curve measured for a SiC/C/SiC composite(sample J)reinforced with treated Nicalon TM fibers ample thickness on the measured interfacial properties pro- with weak interfaces. However, evidence of a very small am- ided data on the test zone to be selected for the push-oI plitude roughness is detected via SEM at higher magnification tests. The thickest part of the samples seemed to be the most(Fig. 6(b)) appropriate to minimize the scatter in data, because of the to the debonding stress. This obser yamax"Plateau)was similar Finally, the load drop(Aodrop =omax presence of fibers that had been debonded during sam expected in such a preparation. The embedded length of the tested fibers(320-400 um)did not seem to influence the measured properties, which emain essentially constant as the sample thickness varies Interfacial Parameters for However, the standard deviation for almost all the parameters ated Fiber-Reinforced C is quite large(100 observations. The scanning electron microscopy(SEM)image MPa, which is an order of magnitude higher than those deter- of a pushed-out fiber( Fig. 6(a)) indicates that debonding oc- mined for the untreated fiber composites. Such high T values curred at the fiber surface (lower inset in Fig. 1). The micro- Iggest rgy dissipation by friction during fiber sliding ph shows that the fiber surface has rather smooth features, SEM images show that the surface of the pushed-out treated as is usually observed for Nicalon TM-reinforced composites fibers appears to be rather rough, which shows that the inter-
sample thickness on the measured interfacial properties provided data on the test zone to be selected for the push-out tests.31 The thickest part of the samples seemed to be the most appropriate to minimize the scatter in data, because of the presence of fibers that had been debonded during sample preparation. The embedded length of the tested fibers (320–400 mm) did not seem to influence the measured properties, which remain essentially constant as the sample thickness varies. However, the standard deviation for almost all the parameters is quite large (#30%), and, the maximum stress at the load drop (Fig. 4) did not increase with the embedded length. Comparable estimates of the interfacial shear stress were derived from the nonlinear part of the curves and from the plateau (Table II and Fig. 5). These values are close to the upper bound in the range of interfacial shear stresses commonly measured by this technique on Nicalon™-fiber/SiC-matrix composites with a carbon interphase of 0.5 mm.18,19 The magnitude of the derived axial and radial clamping stresses are too high, in comparison to those generally computed for this fiber–matrix combination (less than or equal to −200 MPa).10,11,28 A logical reason for the high magnitude of sc is that the Hsueh model does not consider the effect of surface roughness during fiber sliding. The effect of surface roughness may be characterized through a radial compressive stress at the interface that is superimposed on the radial residual stress component. The clamping stress sc becomes the sum of the thermally induced residual clamping stress and the compressive stress due to roughness. A magnitude for the roughness (A) was extracted using the following expression:28 A = H− scF Ef~1 + nm! + Em~1 − nf! qEmEf G − DaDTJr (15) where parameter q 4 1 for an infinite matrix. A q value of 1 was used, which may lead to an underestimation of A; in the presence of a finite matrix, a value of q that is commensurate with the matrix fraction may be preferred. An amplitude of 35 nm (±15 nm) was thus obtained for the roughness A (Table II). This value is in agreement with the fiber roughness measured on Nicalon™ fibers by atomic force microscopy (AFM) (30 nm (see More32)). Examination of the fiber surface also seems to confirm these observations. The scanning electron microscopy (SEM) image of a pushed-out fiber (Fig. 6(a)) indicates that debonding occurred at the fiber surface (lower inset in Fig. 1). The micrograph shows that the fiber surface has rather smooth features, as is usually observed for Nicalon™-reinforced composites with weak interfaces. However, evidence of a very small amplitude roughness is detected via SEM at higher magnification (Fig. 6(b)). Finally, the load drop (Dsdrop 4 smax − splateau) was similar to the debonding stress. This observation is expected in such a system (Table II). (4) Interfacial Parameters for Treated Fiber-Reinforced Composites Because of the lack of a pertinent model, only the following parameters could be extracted from the push-out curves: the debonding stress, the maximum load, the displacement at the maximum load, and the frictional shear stress from the plateau stress (Table II). The large force needed to push the fibers out of the matrix indicated a high resistance to the propagation of interfacial cracks. Hence, it was possible to satisfactorily push out the fibers (without fiber damage or the indentor contacting the matrix) only when the sample thickness was 2500 MPa) and for pushing out the fibers (>4000 MPa) are uncommonly high for test specimens as thin as 150 mm.31 The associated displacements are also very small (∼1.1 mm). Under similar displacements, the untreated fibers exhibit maximum stresses that are a factor of ∼3 smaller. The sliding shear stresses for the treated fibers are >100 MPa, which is an order of magnitude higher than those determined for the untreated fiber composites. Such high t values suggest large energy dissipation by friction during fiber sliding. SEM images show that the surface of the pushed-out treated fibers appears to be rather rough, which shows that the interFig. 3. Single-fiber push-out curve measured for a SiC/C/SiC composite (sample J) reinforced with treated Nicalon™ fibers. April 1998 Interfacial Bond Strength in SiC/C/SiC Composite Materials 969
Journal of the American Ceramic Society-Rebillat et al. Vol 81. No 4 G守F守B 十--r-rrrr十rrr 8了曾 月225228283 5 vd)ssns: aqs pepa叫 G 导二 G守了 岩 05 F日4 8日8> s5 家点 〔ed) sans
Table II. Interfacial Characteristics Measured by Push-Out Tests† Thickness (mm) Debonding stress (MPa) Maximum stress (MPa) Drop of stress at pushout (MPa) Displacement of the top of the fiber (mm) t (MPa) Coefficient of friction, m Axial stress (MPa) Clamping stress (MPa) Debonded length (mm) Magnitude of roughness (nm) Ratio of debond stress to drop in stress Plateau Fitting Material I 390 (30) 510 (280) 1380 (730) 540 (350) 0.8 (0.6) 10 (5) 14 (8) 0.022 (0.012) −340 (300) −720 (240) 250 (80) 35 (15) 1.1 (0.5) 320 (10) 460 (110) 1580 (400) 630 (230) 0.8 (0.3) 12 (3) 16 (7) 0.025 (0.002) −360 (260) −760 (270) 275 (40) 35 (16) 0.8 (0.4) Material J 180 (5) 3840 (110) 6040 (980) 666 (440) 1 (0.4) 105 (9) 7.3 (4.7) 140 (15) 3720 (410) 4860 (1400) 900 (390) 1 (0.5) 100 (40) 5 (2.4) Material P 80 (10) 2980 (890) 7350 (2860) 1155 (310) 0.75 (0.23) 273 (115) 2.7 (0.8) 45 (5) 1830 (260) 2900 (860) 980 (630) 0.28 (0.2) 165 (76) 2.3 (1) †Values given in parentheses are standard deviations. Fig. 4. Debonding and maximum stresses measured experimentally versus the embedded length in SiC/C/SiC for various interphase thicknesses and strengths of the fiber–coating bond. Fig. 5. Interfacial shear stresses extracted from the plateau region versus the embedded length in SiC/C/SiC for various interphase thicknesses and strengths of the fiber–coating bond. 970 Journal of the American Ceramic Society—Rebillat et al. Vol. 81, No. 4
April 1998 Interfacial Bond Strength in SiC/C/SiC Composite Materials (a) tal 10 um (b) (b) I um Fig. 6. SEM micrographs of (a) protruding fibers and(b)a matrix Fig. 7. SEM micrographs of protruding fibers after push-out tests sliding surface after a push-out test on a SiC/C/SiC composite rein- performed on a SiC/C/SiC composite reinforced with treated fibers forced with as-received fibers(material 1). phase has been torn during fiber sliding(Fig. 7); this is in ontrast to the featureless surfac untreated fibe (6) Nanoindentation Tests Performed on 6(a). The sliding surface on th side. as well SiC/SiC Composites Reinforced with Treated Fibers protruded fiber surface, shows crack initiated Figure 8(a) shows a comparison of experimental nanoinden- the carbon layer(top inset in Fig. 1) tation data on a Nicalon TM fiber and a second-order polynomial xpression of that relationship. The good fit indicates that the 5) Results of Alternate Approaches to polynomial can be used to subtract the effect of indentor pen- Deriving Interfacial Characteristics relatively indepen. etration into the fiber. Figure 8(b) shows an example of nanoindentation curve derived by subtracting the indentor pen- dent of the expression that was used. The interfacial shear etration distance from the raw experimental data. The end of strengths(Ts)increase modestly as the complexity of the rela- the linear portion marks the initiation of debonding. Tables Il tionship increases. The largest Ts values were obtained for the and IV show that the debonding stress that is determined by composites reinforced with treated fibers; Ts remained near nanoindentation is much smaller than that measured by micro- 150-185 MPa for the composite reinforced with as-received indentation(1000 MPa versus -3800 MPa). The fiber dis- fibers, and Ts 1000 or 1300 MPa for those reinforced with placement obtained after subtraction of the penetration distance treated fibers(composites P and J, respectively ) Furthermore, is small (150 nm). Even at maximum load (50 g), the applied s was not significantly affected by the sample thickness, ex- stress is less than the debonding stress during microindentation cept for composite P specimens that had a thickness that was Such a small displacement cannot be measured by microinden- not significantly larger than the fiber diameter anymore, as tation, mainly because of significant background pertubations assumed in the models However, the interfacial characteristics measured during un- The clamping stress and the interfacial shear stress, deter- loading(Table IV) are in agreement with those determined by mined using Eq(6)(Table III), are in excellent agreement with microindentation(Table Il). The discrepancy between loading those obtained using the plateau stress(Table II). They confirm and unloading may result from the very large residual stresses the increase in interfacial characteristics due to the use of calculated from the fitting method Furthermore. when unload- treated fibers is performed, influence of the diamond on fiber deforma-
phase has been torn during fiber sliding (Fig. 7); this is in contrast to the featureless surface of the untreated fibers (Fig. 6(a)). The sliding surface on the matrix side, as well as the protruded fiber surface, shows that the crack initiated within the carbon layer (top inset in Fig. 1). (5) Results of Alternate Approaches to Deriving Interfacial Characteristics The results given in Table III seem to be relatively independent of the expression that was used. The interfacial shear strengths (ts) increase modestly as the complexity of the relationship increases. The largest ts values were obtained for the composites reinforced with treated fibers; ts remained near 150–185 MPa for the composite reinforced with as-received fibers, and ts ≈ 1000 or 1300 MPa for those reinforced with treated fibers (composites P and J, respectively). Furthermore, ts was not significantly affected by the sample thickness, except for composite P specimens that had a thickness that was not significantly larger than the fiber diameter anymore, as assumed in the models. The clamping stress and the interfacial shear stress, determined using Eq. (6) (Table III), are in excellent agreement with those obtained using the plateau stress (Table II). They confirm the increase in interfacial characteristics due to the use of treated fibers. (6) Nanoindentation Tests Performed on SiC/SiC Composites Reinforced with Treated Fibers Figure 8(a) shows a comparison of experimental nanoindentation data on a Nicalon™ fiber and a second-order polynomial expression of that relationship. The good fit indicates that the polynomial can be used to subtract the effect of indentor penetration into the fiber. Figure 8(b) shows an example of a nanoindentation curve derived by subtracting the indentor penetration distance from the raw experimental data. The end of the linear portion marks the initiation of debonding. Tables II and IV show that the debonding stress that is determined by nanoindentation is much smaller than that measured by microindentation (∼1000 MPa versus ∼3800 MPa). The fiber displacement obtained after subtraction of the penetration distance is small (∼150 nm). Even at maximum load (∼50 g), the applied stress is less than the debonding stress during microindentation. Such a small displacement cannot be measured by microindentation, mainly because of significant background pertubations. However, the interfacial characteristics measured during unloading (Table IV) are in agreement with those determined by microindentation (Table II). The discrepancy between loading and unloading may result from the very large residual stresses calculated from the fitting method. Furthermore, when unloading is performed, influence of the diamond on fiber deformaFig. 6. SEM micrographs of (a) protruding fibers and (b) a matrix sliding surface after a push-out test on a SiC/C/SiC composite reinforced with as-received fibers (material I). Fig. 7. SEM micrographs of protruding fibers after push-out tests performed on a SiC/C/SiC composite reinforced with treated fibers (material J). April 1998 Interfacial Bond Strength in SiC/C/SiC Composite Materials 971
Journal of the American Ceramic Society-Rebillat et al. Vol 81. No 4 Table ll. Interfacial Parameters Calculated from the measured Debonding and Maximum Stresses from Microindentation Push-Out Tests Coefficient of Thickness Interfacial shear strength, Ts (MPa) Material I 0.03 -460 320(10) 170(40) 0.03 39030) 200(100) 180(100) 190(70)2 170(70)产 160(60) Material J 140(13) 360(150) 0.08 1100 180(5) 00(40) 1380(40) 380(30)2 1360(30)2 Material P 675(85) 615(80) Values given in parentheses are standard deviations. Values for H, the clamping stress, and T were obtained from Takuku and Arridge. 4 1Average value: 90(100) 0.2 80(10) 1000(100) 990(100) 875(190) 870(180)2 16 Model 1200 Extracted 30 0 l000 Displacement (nm) (b) Fig 8. Nanoindentation tests on SiC/C/SiC composites(composite J)((a) penetration of the diamond in the fiber and(b)extraction of the
Table III. Interfacial Parameters Calculated from the Measured Debonding and Maximum Stresses from Microindentation Push-Out Tests† Coefficient of friction, m Clamping stress (MPa) Interfacial shear stress, t (MPa) Thickness (mm) Interfacial shear strength, ts (MPa) Eq. (13) Eq. (11) Eq. (9) Material I 0.03 −460 13 320 (10) 185 (40) 170 (40) 150 (40) 0.03 −460 13 390 (30) 200 (100) 180 (100) 170 (90) 190 (70)‡ 170 (70)‡ 160 (60)‡ Material J 0.08 −1100 88 140 (13) 1360 (150) 1340 (150) 1210 (130) 0.08 −1100 88 180 (5) 1400 (40) 1380 (40) 1250 (40) 1380 (30)‡ 1360 (30)‡ 1230 (30)‡ Material P 0.2 −1300 160 45 (5) 680 (95) 675 (85) 615 (80) 0.2 −1300 160 80 (10) 1000 (100) 990 (100) 900 (100) 875 (190)‡ 870 (180)‡ 790 (160)‡ † Values given in parentheses are standard deviations. Values for m, the clamping stress, and t were obtained from Takuku and Arridge.24 ‡Average value. Fig. 8. Nanoindentation tests on SiC/C/SiC composites (composite J) ((a) penetration of the diamond in the fiber and (b) extraction of the load-versus-fiber-end-displacement curve from a nanoindentation test). 972 Journal of the American Ceramic Society—Rebillat et al. Vol. 81, No. 4
April 1998 Interfacial Bond Strength in SiC/C/SiC Composite Materials Table IV. Interfacial Characteristics Measured by anoindentation Tests Performed on Composite j value Displacement Fiber radius(um) 7.9(0.5) Debonding stress(MPa) 1024(128) 1013(129 Clamping stress(MPa) 1426(40 Coefficient of friction, u 0.15(0.03) 0.1700.02) Interfacial shear stress, T(MP 247(138) 51(100) 0033(0.055) 0061(0.043) Maximum displacement(nm) 171(73) 67(56 Maximum stress(MPa) 2233(266) 2355(98 24(8) Roughness, A(from Eq (3))(um 0.08(0.04) 008(0.03) TFor a sample thickness of-200 um. Values given in parentheses are standard deviations tions and sliding is less significant than during loading. 21 The rectly related to the wavelength of the roughness along the high friction coefficient calculated for the loading phase is sliding surface similar to that which has been measured for two sliding carbon The values measured in push-back tests are naturally smaller surfaces(0.11-0.13).3 than those measured by push-out tests(Table V). The frictional sliding resistance is still larger than the shear stresses obtaine (7 Interfacial Parameters from from push-out tests on composites reinforced with untreated Single- Fiber Push-Back Tests on Treated fibers fibers(Fig. 5). Furthermore, all the interfacial characteristics. The curves recorded during the push-back tests are typical ncluding the debonding and the maximum stresses, the re- and include a linear domain followed by a nonlinear domain sidual clamping stress, and the magnitude of roughness(Table vith a downward curvature(Fig. 9). An initial linear domain of V), confirm the higher resistance to fiber sliding in composites elastic deformation is still present at applied stresses that are reinforced with treated fibers. Thus, comparison with the re smaller than the critical stress required for sliding. The non- sults of push-out tests performed on material I indicates that the inear domain(b'c" in Fig 9)is attributed to the progres stress at the end of the initial linear-elastic domain and the sive increase in fiber sliding length, The curves exhibit a load maximum stress of the push-back curve of material J are much drop before the plateau, as observed during push out, whic higher--1000 MPa vs 500 MPa and 2100 MPa vs 1400 MPa. reflects the catastrophic sliding of the rest of the fiber. The respectively. Yet, the displacement at the load drop to obtain debonding stress now represents the resistance to sliding. Fi- sliding of the entire fiber is smaller: 0.57 um vs 0.8 um nally, the reseating peak load in the plateau indicates that the The surface roughness of the fibers is estimated from the fiber recovers its initial position before protruding clamping stresses and the reseating load: amplitude of -50 nm The reseating peak load has been often described as evidence (+30 nm)and wavelength of -500 nm(*100 nm). The mag of the contribution of surface roughness to fiber sliding 34, 35 nitude of roughness(A)is larger than that estimated for mate Therefore, the wavelength of the reseating peak should be di- rial I from push-out tests. The roughness of the sliding surface 2500+ Thickness= 150 HI 1500 Fig 9. Stress-displacement from a push-back test performed on composite J
tions and sliding is less significant than during loading.21 The high friction coefficient calculated for the loading phase is similar to that which has been measured for two sliding carbon surfaces (0.11–0.13).33 (7) Interfacial Parameters from Single-Fiber Push-Back Tests on Treated Fibers The curves recorded during the push-back tests are typical and include a linear domain followed by a nonlinear domain with a downward curvature (Fig. 9). An initial linear domain of elastic deformation is still present at applied stresses that are smaller than the critical stress required for sliding. The nonlinear domain (‘‘b’’–‘‘c’’ in Fig. 9) is attributed to the progressive increase in fiber sliding length. The curves exhibit a load drop before the plateau, as observed during push out, which reflects the catastrophic sliding of the rest of the fiber. The debonding stress now represents the resistance to sliding. Finally, the reseating peak load in the plateau indicates that the fiber recovers its initial position before protruding. The reseating peak load has been often described as evidence of the contribution of surface roughness to fiber sliding.34,35 Therefore, the wavelength of the reseating peak should be directly related to the wavelength of the roughness along the sliding surface.35 The values measured in push-back tests are naturally smaller than those measured by push-out tests (Table V). The frictional sliding resistance is still larger than the shear stresses obtained from push-out tests on composites reinforced with untreated fibers (Fig. 5). Furthermore, all the interfacial characteristics, including the debonding and the maximum stresses, the residual clamping stress, and the magnitude of roughness (Table V), confirm the higher resistance to fiber sliding in composites reinforced with treated fibers. Thus, comparison with the results of push-out tests performed on material I indicates that the stress at the end of the initial linear-elastic domain and the maximum stress of the push-back curve of material J are much higher—1000 MPa vs 500 MPa and 2100 MPa vs 1400 MPa, respectively. Yet, the displacement at the load drop to obtain sliding of the entire fiber is smaller: 0.57 mm vs 0.8 mm. The surface roughness of the fibers is estimated from the clamping stresses and the reseating load: amplitude of ∼50 nm (±30 nm) and wavelength of ∼500 nm (±100 nm). The magnitude of roughness (A) is larger than that estimated for material I from push-out tests. The roughness of the sliding surface Table IV. Interfacial Characteristics Measured by Nanoindentation Tests Performed on Composite J† Characteristic Value Displacement control Load control Fiber radius (mm) 7.9 (0.5) 8.2 (0.2) Debonding stress (MPa) 1024 (128) 1013 (129) Clamping stress (MPa) −1525 (664) −1426 (404) Axial stress (MPa) −831 (219) −950 (268) Loading Coefficient of friction, m 0.15 (0.03) 0.17 (0.02) Interfacial shear stress, t (MPa) 247 (138) 251 (100) Unloading mn 0.033 (0.055) 0.061 (0.043) tn (MPa) 36 (49) 93 (71) Maximum displacement (nm) 171 (73) 167 (56) Maximum stress (MPa) 2233 (266) 2355 (98) Debonded length (mm) 22 (9) 24 (8) Roughness, A (from Eq. (3)) (mm) 0.08 (0.04) 0.08 (0.03) † For a sample thickness of ∼200 mm. Values given in parentheses are standard deviations. Fig. 9. Stress–displacement curve from a push-back test performed on composite J. April 1998 Interfacial Bond Strength in SiC/C/SiC Composite Materials 973
974 Journal of the American Ceramic Sociery-Rebillat et al. Vol 8I. No 4 Table v, Interfacial Characteristics Measured by Push-Back Tests for Material j value 3630(590) Maximum stress(MPa) 4345(1570) 2040(650) Decrease in stress at push out(MPa) 935(240 f the fiber (um) 0.7800.31) Interfacial shear stress.T 80(30) 40(8) Fitting(MPa) Coefficient of friction, H 0.034(0.017 Axia S(MP 850(390) Clamping stress(MPa) 030(50) 114 avele 0.4900.11) Ratio of debond st stress decrease 4.2(1.5) 2.5(1.8) Embedded length of -140 um. Values given in parentheses are standard deviations. clearly observed via SEM(Fig. 10). Although the rou V. Discussion mplitude cannot be measured in the micrograph in Fig. 10,it seems comparable to that observed on the pushed-out fibers ()Composites Reinforced with As-Received Fibers (Fg,7 The frictional shear stresses extracted from the push-out tests are similar to those estimated by Droillard 3 for material I from the hysteresis loops of unloading-reloading under tension and (a) by other authors for 2D SiC/SiC composites and even for microcomposites tested under tension. 36,37 A microcomposite consists of a single fiber coated with an interphase and with the matrix. Values of-10 MPa seem to be characteristic of these SiC/SiC composite materials reinforced with as-received Nicalon TM fibers 3, 7, 8, 12, 13,15 Moreover, the approach by Bright et al.23 that was based only on the maximum stresses gave interfacial parameters (u, T) that were similar to those extracted from the nonlinea domain of the push-out curve using the Hsueh model, 18,21 except that the derived clamping residual stress from the Hsueh model was significantly smaller. Furthermore, comparable values of interfacial shear strengths(Ts) were determined using various equations which do and do not account for the presence of residual stresses. 24-/These Ts values are one order of magnitude higher than the interfacial shear stresses(T) As commonly observed in composites with weakly bonded fibers. the fiber/carbon-coating interface is the weakest link in the interfacial sequence(lower inset in Fig. 1). Examination of both sides of samples after push-out tests reveals the smooth surface of the fiber and does not indicate any tearing of the in- terphase, which suggests the presence of a single crack that propa- (b) C gates at the surface of the fiber. Hence, the role of the inter- cau the matrix, only contributing to the relief of residual stress. The analysis shows that, despite the presence of a layer of soft carbon, the effect of surface roughness cannot be This effect is greater than the thermal residual stress; carbon layer is subjected to compression during fib (2) Composites Reinforced with Treated Fibers (A) Debonding: TEM examination of the cracked inter phases after tensile tests has revealed the presence of cracks branching within the interphase(top inset in Fig. 1). 12, 16 The interfacial damage behavior, as well as the interfacial proper ties extracted from the push-out curves, seem to be consistent with these crack patterns The load required for initiation of fiber sliding, as measured by the so-calleddebond stress(oa), is more than a factor of 4 higher for the composites reinforced with treated fibers. which is also reflected by the maximum stress. This stress seems to increase exponentially as the embedded length of the micrographs of (a)protruding fibers and (b)the fiber fiber increases and, thus, is sensitive to very small increases in back tests performed on a Sic/C/SiC com- the thickness of the tested zone(-10 um). Fibers could not be ated fibers(material J) pushed out when the samples were thicker than 180 um, be-
is clearly observed via SEM (Fig. 10). Although the roughness amplitude cannot be measured in the micrograph in Fig. 10, it seems comparable to that observed on the pushed-out fibers (Fig. 7). V. Discussion (1) Composites Reinforced with As-Received Fibers The frictional shear stresses extracted from the push-out tests are similar to those estimated by Droillard13 for material I from the hysteresis loops of unloading–reloading under tension and by other authors for 2D SiC/SiC composites19 and even for microcomposites tested under tension.36,37 A microcomposite consists of a single fiber coated with an interphase and with the matrix. Values of ∼10 MPa seem to be characteristic of these SiC/SiC composite materials reinforced with as-received Nicalon™ fibers.3,7,8,12,13,15 Moreover, the approach by Bright et al.23 that was based only on the maximum stresses gave interfacial parameters (m, t) that were similar to those extracted from the nonlinear domain of the push-out curve using the Hsueh model,18,21 except that the derived clamping residual stress from the Hsueh model was significantly smaller. Furthermore, comparable values of interfacial shear strengths (ts) were determined using various equations which do and do not account for the presence of residual stresses.24–27 These ts values are one order of magnitude higher than the interfacial shear stresses (t). As commonly observed in composites with weakly bonded fibers, the fiber/carbon-coating interface is the weakest link in the interfacial sequence (lower inset in Fig. 1). Examination of both sides of samples after push-out tests reveals the smooth surface of the fiber and does not indicate any tearing of the interphase, which suggests the presence of a single crack that propagates at the surface of the fiber. Hence, the role of the interphase may be regarded as limited, because it remains bonded to the matrix, only contributing to the relief of residual stress. The analysis shows that, despite the presence of a layer of soft carbon, the effect of surface roughness cannot be neglected. This effect is greater than the thermal residual stress; thus, the carbon layer is subjected to compression during fiber sliding. (2) Composites Reinforced with Treated Fibers (A) Debonding: TEM examination of the cracked interphases after tensile tests has revealed the presence of cracks branching within the interphase (top inset in Fig. 1).12,16 The interfacial damage behavior, as well as the interfacial properties extracted from the push-out curves, seem to be consistent with these crack patterns. The load required for initiation of fiber sliding, as measured by the so-called ‘‘debond stress’’ (sd), is more than a factor of 4 higher for the composites reinforced with treated fibers, which is also reflected by the maximum stress. This stress seems to increase exponentially as the embedded length of the fiber increases and, thus, is sensitive to very small increases in the thickness of the tested zone (∼10 mm). Fibers could not be pushed out when the samples were thicker than 180 mm, beTable V. Interfacial Characteristics Measured by Push-Back Tests for Material J† Characteristic Value Zero push back (push-out) One push back Debonding stress (MPa) 3630 (590) 970 (350) Maximum stress (MPa) 4345 (1570) 2040 (650) Decrease in stress at push out (MPa) 935 (240) 490 (220) Displacement of the top of the fiber (mm) 0.78 (0.31) 0.57 (0.33) Interfacial shear stress, t Plateau (MPa) 80 (30) 40 (8) Fitting (MPa) 35 (20) Coefficient of friction, m 0.034 (0.017) Axial stress (MPa) −850 (390) Clamping stress (MPa) −1030 (50) Debonded length (mm) 114 (35) Magnitude of roughness (nm) 47 (29) Wavelength of roughness (mm) 0.49 (0.11) Ratio of debond stress to stress decrease 4.2 (1.5) 2.5 (1.8) † Embedded length of ∼140 mm. Values given in parentheses are standard deviations. Fig. 10. SEM micrographs of (a) protruding fibers and (b) the fiber sliding surface after push-back tests performed on a SiC/C/SiC composite reinforced with treated fibers (material J). 974 Journal of the American Ceramic Society—Rebillat et al. Vol. 81, No. 4