(1998 Published by Elsevier Science Ltd. All rights reserve PII:s0266·3538(97)00150-4 66-353898s1900 MECHANICAL BEHAVIOR AND DAMAGE DEVELOPMENT DURING CYCLIC FATIGUE AT HIGH-TEMPERATURE OF A 2.5D CARBON/SIC COMPOSITE A Almaz,P. Reynaud, D. Rouby, G. Fantozzi& F Abbe ' Groupe d' Etudes de metallurgie Physique ue et de physique des materiaux, UMR CNRS 5510, Institut National des sciences appliquees, 20 av. A. Einstein, 69 62/ Villeurbanne, france h Societe Europeenne de Propulsion, Propulsion d Poudre et Composites. Les Cing Chemins, Le hailan 33 165 St-Medard-en-Jalles Cedex, france Received 29 October 1996; accepted 24 July 1997) Abstract and 2D C/SiC composites. These studies describe the This article describes part of a project to investigate mechanical behavior of these materials (non-linear high-temperature fatigue behavior of a 2. 5D C/SiC mul- behavior under tension, fracture stress independent of bilayer composite. This material is particularly interesting temperature, life- independent of frequency) the origin ince it has been found that damage exists in the as-sup- and mechanisms leading to damage development and plied material owing to the residual stresses produced fracture of the 2.SD C/SiC composite is still poorly understood describes the effect of the number of cycles and the tem- The aim of this work is to describe the damage perature on the composite, and attempts to explain the mechanism for a 2. 5D C/SiC composite, and thus define behavior during testing. The parameters controlling the mechanical laws controlling fracture and cyclic damage development and therefore fatigue life, are fatigue at various temperatures. Moreover, since this believed to be both thermal constraint in the composite, composite contains cracks from processing, it is impor- nd the damage produced. The nature of the damage is tant to determine their effect on fatigue behavior. such thut the material is best considered in lerms Hence, the change in mechanical behavior in relation to damage occuring on the scale of the yarns, rather than at the number of the cycles and the temperature are the smaller fiber level. C 1998 Published by Elsevier described in this paper, and the specific rk of Science Ltd. All rights reserved cracks developed under such loading is identified Keywords: A ceramic-matrix composite, B fatigue, B interface, high-temperature, damage 2 EXPERIMENTAL PROCEDURE 2.1 Material 1 INTRODUCTION Tests were conducted on a long-carbon-fiber silicon- carbide matrix 2, 5D C/SiC composite provided by the This study has concentrated on a multilayer long-fiber- Societe Europeenne de Propulsion (SEP, Bordeaux reinforced ceramic-matrix composite of carbon fibers in France). This composite is composed of cloth made a silicon carbide matrix (2. 5D C/SiC), supplied by the from long woven yarns(see Fig. 7). The preform is Societe Europeenne de Propulsion. This thermo- infiltrated by a pyrocarbon interlayer in order to opti structural composite is intended for use at high-tem- mize the fiber /matrix shear lag, and induce a non-brittle perature under substantial mechanical loads. Hence, it mechanical macroscopi n-ca is important to understand the cyclic fatigue behavior of matrix is deposited by chemical vapor infiltration this material at high-temperature and to study the effect (C.v. I.). The as-received composite exhibits a porosity of environment level below 10% and a density of 2200 kg m . The Previous studies of the tensile mechanical behavior characteristic of this composite is that matrix cracks are under cyclic fatigue at high-temperature have con- created during cooling just after processing, owing to centrated on composites similar, but not identical, the high tensile thermal residual stresses the 2.5D C/SiC. For example the work of Arendts'ane Camus2 on the characterization of the tensile mechan 2.2 Mechanical tests ical behavior at high-temperature of a 2D C/SiC and the The aim of this work is to determine the characteristics ork of Shuler and Iee on cyclic fatigue of [C/SiClo of the 2. 5D C/SiC composite under cyclic tensile/tensile
LLSEVIER Compo.srres Suence and Twhno1og.v 58 (1998) 693 699 P‘s 1998 Published by Elsevier Saence Ltd. All rights reserved Printed in Great Britam PII: SO266-3538(97)00150-4 0266.3538:98 $19.00 MECHANICAL BEHAVIOR AND DAMAGE DEVELOPMENT DURING CYCLIC FATIGUE AT HIGH-TEMPERATURE OF A 23D CARBON/Sic COMPOSITE A. Dalmaz,a P. Reynaud,a D. Rouby,” G. Fantozzia & F. Abbeh “Groupe d’Etudes de MCtallurgie Physique et de Physique des Mathiaux, UMR CNRS 5510, Institut National des Sciences AppliquPes, 20 av. A. Einstein, 69 621 Villeurhanne, France hSoeiPtP EuropPenne de Propulsion, Propulsion d Poudre et Composites, Les Cinq Chemins, Le HailIan, 33 165 St-Mhdard-en-Jalles Cedex, France (Received 29 October 1996: accepted 24 July 1997) Abstract This article describes part of a project to investigate the high-temperature fatigue behavior of a 2.50 C/Sic multilayer composite. This material is particularly interesting since it has been,found that damage exists in the as-supplied material owing to the residual stresses produced during cooling of the composite. This section of work describes the t$ect of the number of cycles and the temperature on the composite, and attempts to explain the behavior during testing. The parameters controlling damage development and therefore fatigue life, are believed to be both thermal constraint in the composite, and the damage produced. The nature of the damage is such that the material is best considered in terms of damage occuring on the scale of the yarns, rather than at the smaller jiber level. 0 1998 Published by Elsevier Science Ltd. All rights reserved and 2D C/Sic composites. These studies describe the mechanical behavior of these materials (non-linear behavior under tension, fracture stress independent of temperature, life-independent of frequency) the origin and mechanisms leading to damage development and fracture of the 2.5D C/Sic composite is still poorly understood. The aim of this work is to describe the damage mechanism for a 2.5D C/Sic composite, and thus define the mechanical laws controlling fracture and cyclic fatigue at various temperatures. Moreover, since this composite contains cracks from processing, it is important to determine their effect on fatigue behavior. Hence, the change in mechanical behavior in relation to the number of the cycles and the temperature are described in this paper, and the specific network of cracks developed under such loading is identified. Keyw’ords: A. ceramic-matrix composite, B. fatigue, B. interface, high-temperature, damage 2 EXPERIMENTAL PROCEDURE 2.1 Material 1 INTRODUCTION This study has concentrated on a multilayer long-fiberreinforced ceramic-matrix composite of carbon fibers in a silicon-carbide matrix (2.5D C/Sic), supplied by the Soci&i. Europkenne de Propulsion. This thermostructural composite is intended for use at high-temperature under substantial mechanical loads. Hence, it is important to understand the cyclic fatigue behavior of this material at high-temperature and to study the effect of environment. Previous studies of the tensile mechanical behavior under cyclic fatigue at high-temperature have concentrated on composites similar, but not identical, to the 2.5D C/Sic. For example, the work of Arendts’ and Camus on the characterization of the tensile mechanical behavior at high-temperature of a 2D C/Sic and the work of Shuler’ and Lee4 on cyclic fatigue of [C/SiC]0,90 Tests were conducted on a long-carbon-fiber siliconcarbide matrix 2.5D C/Sic composite provided by the Soci& Europlenne de Propulsion (SEP, Bordeaux, France). This composite is composed of cloth made from long woven yarns (see Fig. 7). The preform is infiltrated by a pyrocarbon interlayer in order to optimize the fiber/matrix shear lag, and induce a non-brittle mechanical macroscopic behavior. The silicon-carbide matrix is deposited by chemical vapor infiltration (C.V.I.). The as-received composite exhibits a porosity level below 10% and a density of 2200 kgm-‘. The characteristic of this composite is that matrix cracks are created during cooling just after processing, owing to the high tensile thermal residual stresses. 2.2 Mechanical tests The aim of this work is to determine the characteristics of the 2.5D C/Sic composite under cyclic tensile/tensile 693
694 1. Alma- et al fatigue loading, and to define the changes in behavior stresses in the material during cooling after manu lused by test temperature. The mechanical load facture. These stresses are probably partially relaxed applied as a sinusoidal signal from 0 to a given max- generating a network of cracks in the matrix(Fig. 7) imum amplitude. The frequencies of the load We can note three types of pre-existent cracks unloading cycle were I and 10 Hz. Two types of test have been conducted at room temperature at 35 and I. Matrix cracks perpendicular to the longitudina 75% of the fracture stress. The effect of the temperature yarns: the intercrack distance evolves from has also been studied for a load of 75% of the fracture 240±117mto506±32m, for a matrix thick stress at 600. 1000 and 1500%( ess of 40 um and 100 um, respectively. No evi- a quantitative analysis of these phenomena was con lence of crack propagation parallel to the fiber ducted on the basis of the different parameters(Youngs matrix interface was noted modulus, residual deformation. hysteresis loop)repre- 2. Cracks within the transversal yarns: all the yarns Senting the macroscopic mechanical behavior of the are cracked in a direction parallel to the short axis of the ellipse and exhibit, on average, 2. 5 crack mposite Tests were conducted on tensile specimens with a per yarn, across the whole thickness of the yarns parallel section of 16 x 50 mm. The thickness of 2 a tew cracks parallel to the long axis were also corresponds to that of the standard composite. Tests present (0.2 per yarn) with a mean length of were conducted in an Instron(8502servo-hydraulic test 400 um. Macroporosity does not seem to initiate rig and aA E.T. furnace. Force was measured by an the crack network Instron load cell ( 50 kN)and the deformation by an Instron high-temperature extensometer (+%). The 3.1.2 Thermal stress field specimen was placed in cooled grips and positioned in Pre-existent cracks in the matrix and within the yarns the center of the furnace. The latter was heated by an are due to the partial or total relaxation of the thermal induction healing element with thermal insulation. A stresses induced during fabrication special type of insulation allowed tests to be conducted A finite-element analysis of the composite was con up to 2000C under an inert atmosphere or 1600C ducted in order to estimate the thermal stresses in each component and to study the crack propagation mechanisms after fabrication or during tensile loading The mechanical and dilatometric properties of each 3 RESULTS AND DISCUSSION constituent are presented in table First, a study was conducted at the microscopic fiber 3. 1 Characterization of the as-received composite matrix level in order to determine the stress in a cylin- The transverse and longitudinal yarns each consist of drical system made up of carbon fiber and matrix 1000 carbon fibers of diameter 7 um (Fig. 7). These according to the constitution of the composite. The yarns form cylinders of elliptical cross-section of results by finite-element analysis, after cooling fromm the 1200 um long axis by 200 um short axis. The archi- elaboration to room temperature, are tecture. the specific infiltration of pyrocarbon and the atrix of the composite produce macroporosities alongitudinal in matrix =350MPa between the longitudinal yarns. The fibers are highl Longitudinal in fiber =-600MPa, compact (hexagonal or cubic arrangements) and just Ordial at the interface 36MPa allow the infiltration of pyrocarbon. A small amount of microporosity within the yarns is also noted. The matrix The matrix is in tension in the longitudinal direction therefore exclusively deposited around the yarn and and the interface in the radial direction. Since the ther constitutes a coating. It should be noted that. in the mal stress is higher than the Sic fracture strength composite. the longitudinal yarns are parallel to the (or: 294 MPa), matrix cracks are genet loading direction. analytical calculations, taking into account radial and longitudinal effects, are in accordance with these results 3.1.1 Crack network Since the pyrocarbon interlayer ry im t to C he difference in the thermal expansion coefficient the mechanical behavior of the composite. a cylindrical system with three elements was considered. The thermal Table 1. Constituent properties Longitudinal GPa) longitudinal(C) (C1) Carbon fiber T300 430 48 48×106
694 .4. Dultnu: fatigue loading, and to define the changes in behavior caused by test temperature. The mechanical load was applied as a sinusoidal signal from 0 to a given maximum amplitude. The frequencies of the loading; unloading cycle were 1 and IO Hz. Two types of test have been conducted at room temperature at 35 and 750/o of the fracture stress. The effect of the temperature has also been studied for a load of 75% of the fracture stress at 600. 1000 and 1500°C. A quantitative analysis of these phenomena was conducted on the basis of the different parameters (Young’s modulus, residual deformation. hysteresis loop) representing the macroscopic mechanical behavior of the composite. Tests were conducted on tensile specimens with a parallel section of 16 x 50 mm. The thickness of 2 mm corresponds to that of the standard composite. Tests were conducted in an lnstron (8502) servo-hydraulic test rig and a A.E.T. furnace. Force was measured by an Instron load cell (+ 50 kN) and the deformation by an lnstron high-temperature extensometer ( + 5%). The specimen was placed in cooled grips and positioned in the center of the furnace. The latter was heated by an induction heating element with thermal insulation. A special type of insulation allowed tests to be conducted up to 2000°C under an inert atmosphere or 1600°C in air. 3 RESULTS AND DISCUSSION 3.1 Characterization of the as-received composite The transverse and longitudinal yarns each consist of 1000 carbon fibers of diameter 7Llrn (Fig. 7). These yarns form cylinders of elliptical cross-section of 1200~m long axis by 200pm short axis. The architecture, the specific infiltration of pyrocarbon and the matrix of the composite produce macroporosities between the longitudinal yarns. The fibers are highly compact (hexagonal or cubic arrangements) and .just allow the infiltration of pyrocarbon. A small amount of microporosity within the yarns is also noted. The matrix is therefore exclusively deposited around the yarn and constitutes a coating. It should be noted that. in the composite, the longitudinal yarns are parallel to the loading direction. 3.1.1 Crack nrtrtwrk The difference in the thermal expansion coefficient between the constituents produces thermal residual et al. stresses in the material during cooling after manufacture. These stresses are probably partially relaxed. generating a network of cracks in the matrix (Fig. 7). We can note three types of pre-existent cracks: Matrix cracks perpendicular to the longitudinal yarns: the intercrack distance evolves from 24Oi 117 I_cm to 506~t 321_~m, for a matrix thickness of 40pm and 1001_~m, respectively. No evidence of crack propagation parallel to the fiber:’ matrix interface was noted. Cracks within the transversal yarns: all the yarns are cracked in a direction parallel to the short axis of the ellipse and exhibit, on average, 2.5 cracks per yarn, across the whole thickness of the yarns. A few cracks parallel to the long axis were also present (0.2 per yarn) with a mean length of 4001_~m. Macroporosity does not seem to initiate the crack network. 3. I .2 Thrrnzal .stre.s.s firid Pre-existent cracks in the matrix and within the yarns are due to the partial or total relaxation of the thermal stresses induced during fabrication. A finite-element analysis of the composite was conducted in order to estimate the thermal stresses in each component and to study the crack propagation mechanisms after fabrication or during tensile loading. The mechanical and dilatometric properties of each constituent are presented in Table 1. First, a study was conducted at the microscopic fiber: matrix level in order to determine the stress in a cylindrical system made up of carbon fiber and matrix, according to the constitution of the composite. The results by finite-element analysis, after cooling from the elaboration to room temperature, are: %ngltudlnal I” matnx - - - ISOMPa, ~longltudinal I” fiber = -600MFk ~radlal at the mterl’ace = - 36MPa. The matrix is in tension in the longitudinal direction, and the interface in the radial direction. Since the thermal stress is higher than the SIC fracture strength ~~~~~~~~~~ = 294 MPa), matrix cracks are generated. The analytical calculations,5 taking into account radial and longitudinal effects, are in accordance with these results. Since the pyrocarbon interlayer is very important to the mechanical behavior of the composite,’ a cylindrical system with three elements was considered. The thermal Table 1. Constituent properties Carbon fiber T300 Pyrocarbon SIC %>“gltudlnal K ‘) %,dl;il (“C ’ ) 1x10 h 7x10 h 3x10 h 28x10 h7 4.8x 10 h 4.8x10 h
Mechanical behavior and damage development during cyclic fatigue 695 stresses after cooling calculated by finite-element analy longitudinal in matrix= 50MPa longitudinal in fiber=-150MPa longitudinal in interface = 250MPa =-175MPa, radial interface pyrocarbon/ matrix= 225MPa Second, a finite-element analysis as been performed 00.2040.60.81.01.21.4 a mesoscopic yarn/ matrix level, more representative of STRAIN (%O the composite architecture. In fact, after loading lests, specific types of cracks such as decohesion at the long Fig. 2. Cyclic fatigue tests at ambient temperature (omax=220 MPa, failure at 7.18x 10cycles) itudinal and transverse yarn interface and at the yarn/ matrix interface are observed in specimens. It appears 3.2 Thermomechanical tests necessary therefore to consider the yarn as a whole and to identify the interaction between the transverse and 3. 2. 1 The infuence of temperature on cyclic tensile longitudinal yarns in order to point out some weak loadin areas. Hence, a unit cell representative of the composite As previously observed, 2 the mechanical behavior is is represented schematically in Fig. I following the non-linear up to failure. This non-linearity is less actual proportions of the material. The yarn is con- marked below 600C, leading to quasi-linear elastic sidered as an element composed of carbon fiber and behavior, and becomes increasingly pronounced as the pyrocarbon, with a matrix coating. The equivalent temperature rises(to 1500C). The main mechanica properties of the yarn were then estimated from the properties, for the temperatures used, are given aracteristics of each constituent(Table 1)by using the Table 2 rule of mixtures and Shapery calculations Table 2 shows, primarily, an increase in stifness with Residual thermal stress analysis produces, in the x temperature, with a peak at 600 C (E, rises by 180%at and Y directions, during post-fabrication cooling, the 600 C as compared to room temperature) llowing situations Observation of the microstructure reveals that the (a)Longitudinal and transverse yarns are in com initial network of cracks changes. Under mechanical pression( 500 MPa)and tension (500 MPa), load, new matrix cracks are created, oriented perpend respectively. in the X direction. This stress field cular to the longitudinal yarns. Cracking is more divided by3 than in the bulk( crack spacing onlf吃令 marked near the specimen surface(the crack spacing transverse yarns, developed along the short axis tiated of the ellipse. Cracks might be initiated at a died by 2). In addition, micropore within the yarn nside the transversal yarns with an average of 0.5 cracks per yarn more. This was observed only after tests (b) The discontinuity in stress, o between the long- conducted at 1000 and 1500C itudinal and the transversal yarns can produce a shear phenomenon or a crack propagation at the 3.2.2 The effect of the number of cycles during fatigue longitudinal/ transverse, yarn interface. (c) The matrix stress, x, changes along the long test itudinal yarn from 900 MPa to 200 M The changes in mechanical behavior as a function of the number of cycles at various temperatures are analyse in relation to the following parameters: E(Youngs modulus), Er (residual strain) and w(area of the hys 100um telesis loop) proceeds can be due to either the number of load- Table 2. Ultimate stress (od), ultimate strain (Ed, and initial tensile modulus (Fo), under cyclic tensile loading at different emperatures Room temperature600°C1000°1500°C 二四n or(MPa) 376294 0.690.84 Eo(GPa) 125 Fig. 1. Unit cell for the finite-element analysis
Mechanical behavior and damage development during cyclic fatigue 695 stresses after cooling calculated by finite-element analysis are: %ngitudinal in matrix = SOMPa, %ngitudinal in fiber = - 1 SOMPa, %ngitudmal in interface = 250MPa, ondial interface liber/pyrocarbon = -175MPa, OradA interface pyrocarbon/matrix - - 225MPa. Second, a finite-element analysis as been performed at a mesoscopic yarn/matrix level, more representative of the composite architecture. In fact, after loading tests, specific types of cracks such as decohesion at the longitudinal and transverse yarn interface and at the yarn/ matrix interface are observed in specimens. It appears necessary therefore to consider the yarn as a whole and to identify the interaction between the transverse and longitudinal yarns in order to point out some weak areas. Hence, a unit cell representative of the composite is represented schematically in Fig. 1 following the actual proportions of the material. The yarn is considered as an element composed of carbon fiber and pyrocarbon, with a matrix coating. The equivalent properties of the yarn were then estimated from the characteristics of each constituent (Table 1) by using the rule of mixtures and Shapery calculations.* Residual thermal stress analysis produces, in the X and Y directions, during post-fabrication cooling, the following situations: (4 (b) (c) Longitudinal and transverse yarns are in compression (- 500 MPa) and tension (500 MPa), respectively, in the X direction. This stress field may be the main cause of cracks within the transverse yarns, developed along the short axis of the ellipse. Cracks might be initiated at a micropore within the yarn. The discontinuity in stress, ox, between the longitudinal and the transversal yarns can produce a shear phenomenon or a crack propagation at the longitudinal/transverse, yarn interface. The matrix stress, ox, changes along the longitudinal yarn from 900MPa to 200MPa. 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 STRAIN (%) Fig. 2. Cyclic fatigue tests at ambient temperature (cmax = 220 MPa, failure at 7.18 x 1 O4 cycles). 3.2 Thermomechanical tests 3.2.1 The in.uence of temperature on cyclic tensile loading As previously observed, in the mechanical behavior is non-linear up to failure. This non-linearity is less marked below 600°C leading to quasi-linear elastic behavior, and becomes increasingly pronounced as the temperature rises (to 15OO’C). The main mechanical properties, for the temperatures used, are given in Table 2. Table 2 shows, primarily, an increase in stiffness with temperature, with a peak at 600°C (E, rises by 180% at 600°C as compared to room temperature). Observation of the microstructure reveals that the initial network of cracks changes. Under mechanical load, new matrix cracks are created, oriented perpendicular to the longitudinal yarns. Cracking is more marked near the specimen surface (the crack spacing is divided by 3) than in the bulk (crack spacing only dividied by 2). In addition, few new cracks were initiated inside the transversal yarns with an average of 0.5 cracks per yarn more. This was observed only after tests conducted at 1000 and 1500°C. 3.2.2 The efSect qf the number qf cycles during fatigue tests The changes in mechanical behavior as a function of the number of cycles at various temperatures are analysed in relation to the following parameters: E (Young’s modulus), E, (residual strain) and W (area of the hysteresis loop). The changes in mechanical behavior as fatigue cycling proceeds can be due to either the number of loadTable 2. Ultimate stress (ar), ultimate strain (er), and initial tensile modulus (E,,), under cyclic tensile loading at different temperatures Room temperature 600°C 1000°C 1500°C OF (MPa) 305 376 294 294 El‘(%) 0.86 0.69 0.84 I .23 Eo (GPa) 70 125 85 100 Fig. 1. Unit cell for the finite-element analysis
696 A Dalma- et al Table 3. Conditions of cyclic fatigue tests in tension-tension loading Room temperature 600C 1000°C Peak stress(MPa) Number of cycles (1)at failure(2)without failure 710)4(1) 07(2) 107(2) 2105(1) unload loops, or to the time needed to reach that num- 3.2.3 Change in fatigue behavior as a function of ber of cycles. In order to evaluate the effect of cycling temperature frequency fatigue tests have been performed (at room Under the same peak load, corresponding to 75%of temperature. 600 and 1000C)with a frequency jump ultimate stress for room temperature, 1000 and 1500C after the first 50 cycles from 0-05 to I Hz. Comparing test, and 60% of the ultimate stress at 600 C. the life the change of the tangent moduli. with number of and the fatigue limit increase both as the temperature cycles, no significant difference is observed. Thus. we rises. at least up to 1000 C. A shorter life is always can conclude that time-dependent phenomena are not obtained at room temperature preponderant in the fatigue mechanisms under an inert when the peak load is high enough to lead to fatigue atmosphere. This is also reported by Shulerfor tests failure. the residual elongation increases at room tem performed at room temperature at a frequency lower perature and very sharply at 1500 C. Below the fatigue than 10 Hz. We have. therefore. studied only the effect limit, the residual strain is virtually stable, but tends to of the number of cycles on the behavior of the material. decrease at room temperature and to increase at 600oC The fatigue tests are divided in two types(Table 3) and 1000C. Increasing test temperature leads to lower loop area: there is a factor of 4 between the area at 3.2.2.1 Tests conducted be ond /(r cycles withour room temperature and that at 1500C As shown in Fig. 3, the modulus increases sig- nificantly, especially at 600C. with the number of cycles. This stiffening changes with increase in the resi 2.210° dual strain(Fig. 5). The hysteresis(loop area)becomes ower at clavated temperature( Fig. 6)and decreases as the fatigue cycling proceeds 15() 3. 2.2.2 Tests leading to failure before 10 cycles In all cases, the tangent modulus decreases up to failure (Figs 2 and 4). At room temperature, the slope at the 50 end of loading decreases after 2000 cycles and the resid al strain decreases(Fig. 5). This phenomenon must be confirmed by more systematic tests. For the test at 1500C, the principal observation is a very sharp increase in the residual strain up to failure(Er-1.96%at STRAIN (%) failure). As before. the area of the loops decreases as the Fig 4. Cyclic fatigue tests at 15009 230 MPa. failure cycling proceeds at22×105c 500°C 20 14 z150 乙∽5 2 口100 4 600C 0.00.20.40.60.8 oy(NUMBER OF CYCLES STRAIN (%) Fig. 5. Residual strains during fatigue cycling at different Fig 3. Cyclic fatigue tests at 600C(omax =230 MPa, spec temperatures (omax=230 MPa (600, 1000 and 1500C) en not failed after 10 cycles) omax-220 MPa(room temperature)
696 A. Lkdtw~ et al. Table 3. Conditions of cyclic fatigue tests in tension-tension loading Room temperature 600°C Peak stress (MPa) 220 I IO 230 Number of cycles (I) at failure (2) without failure 710J (I) IO” (2) IO’ (2) 1000”c i5OO”C 230 230 IO’ (2) 210’ (I) unload loops, or to the time needed to reach that number of cycles. In order to evaluate the effect of cycling frequency. fatigue tests have been performed (at room temperature. 600 and IOOO’C) with a frequency jump after the first 50 cycles from 0.05 to I Hz. Comparing the change of the tangent moduli, with number of cycles, no significant difference is observed. Thus. we can conclude that time-dependent phenomena are not preponderant in the fatigue mechanisms under an inert atmosphere. This is also reported by Shulerj for tests performed at room temperature at a frequency lower than 1OHz. We have, therefore, studied only the effect of the number of cycles on the behavior of the material. The fatigue tests are divided in two types (Table 3). 3.2.2. I Tests conductrd hqvnd IO” c~,~~~it~.s withour ,fhilurr As shown in Fig. 3. the modulus increases significantly, especially at 6OO”C, with the number of cycles. This stiffening changes with increase in the residual strain (Fig. 5). The hysteresis (loop area) becomes lower at elavated temperature (Fig. 6) and decreases as the fatigue cycling proceeds. 3.2.2.2 Tests leuding to,fitiluw lwffore 10” c~~~ic~.s In all cases, the tangent modulus decreases up to failure (Figs 2 and 4). At room temperature, the slope at the end of loading decreases after 2000cycles and the residual strain decreases (Fig. 5). This phenomenon must be confirmed by more systematic tests. For the test at 15OO”C, the principal observation is a very sharp increase in the residual strain up to failure (or== 1.96% at failure). As before, the area of the loops decreases as the cycling proceeds. 0.0 0.2 0.4 0.6 0.8 1 .o STRAIN (%) Fig. 3. Cyclic fatigue tests at 600°C (urnax = 230 MPa, specimen not failed after IO7 cycles). 3.23 C’hungr in,fktigur hrhuvior as u,fimc.tion of’ tciqwruturc Under the same peak load, corresponding to 75% ot ultimate stress for room temperature, 1000 and 1500°C test. and 60% of the ultimate stress at 6OO”C, the life and the fatigue limit increase both as the temperature rises. at least up to 1000°C. A shorter life is always obtained at room temperature. When the peak load is high enough to lead to fatigue failure. the residual elongation increases at room temperature and very sharply at 1500°C. Below the fatigue limit, the residual strain is virtually stable, but tends to decrease at room temperature and to increase at 600°C and 1000°C. Increasing test temperature leads to lower loop area; there is a factor of 4 between the area at room temperature and that at 1500°C. 0 0.0 0.5 1.0 1.5 2.0 2.5 STRAIN (%) Fig. 4. Cyclic fatigue tests at 1500°C (crmax = 230 MPa, failure at 2.2 x 105cycles). 2 1500 “C ; 1.X . + I IO I 0’ IO’ IO1 IO‘ IIf log(NtiMlEKOFCYCLES) Fig. 5. Residual strains during fatigue cycling at different temperatures ((J,,, = 230 MPa (600, 1000 and 1 SOOT), (T,,, = 220 MPa (room temperature))
Mechanical behavior and damage development during cyclic fatigue 3.2.4 Microstructural observations plane parallel to the load axis in the middle of the spe- The study of the microstructure is madc on spccimcns cimcn width, the damage bcing cxamined along the tested by cyclic fatigue at different temperatures with a specimen thickness peak stress at 230 MPa. The observations are made on a Matrix cracking(perpendicular to the tensile direc tion). The broken and unbroken specimens, after face between a fifth and a third of that of the untested material (267 117 um initial crack spacing ≌30 to 80=+50 um after cyclic tensile testing at 1500C) If we compare this crack spacing with that exhibited by a specie men s objected to a tensile rve 100c 10 a very small change, a factor of I·2(89±80um and 122+80 um, respectively, after cyclic tensile loading and fatigue tests at 1000C)for the unfailed specimens(tested at 600 and 1000oC)and lOg (NUMBER OF CYCLES) smaller by a factor 0-65 for the fatigue failed spe cimens (tested at room temperature and 1500 Fig. 6. Hysteresis changes during cyclic fatigue tests at differ (126±82{mand80±50um, respectively, after 220 MPa) tensile cyclic loading and fatigue testing at 1500oC) Fig. 9. Damage accumulated during cyclic fatigue testing at Fig. 7. The as-received 2- SD C/SiC composite 600C(omax=230 MPa, in a horizontal loading direction) Fig 8. Damage accumulated during cyclic fatigue testing at 230 MPa, in a horizontal loading Fig 10. Damage accumulated during cyclic fatigue testing at
Mechanical behavior and dumage development during cyclic fafigue 697 3.2.4 Microstructural observations The study of the microstructure is made on specimens tested by cyclic fatigue at different temperatures with a peak stress at 230 MPa. The observations are made on a em T EW ‘C 0 / ~~ ~~_ 1 10 10’ 10’ 101 IO’ IOh log (NUMBER OF CYCLES) Fig. 6. Hysteresis changes during cyclic fatigue tests at different temperatures (umaX = 220 MPa). Fig. 7. The as-received 2-5D C/Sic composite. Fig. 8. Damage accumulated during cyclic fatigue testing at room temperature ((T,,, = 230MPa, in a horizontal loading direction). plane parallel to the load axis in the middle of the specimen width, the damage being examined along the specimen thickness. Matrix cracking (perpendicular to the tensile direction). The broken and unbroken specimens, after fatigue testing, exhibit crack spacings at the surface between a fifth and a third of that of the untested material (267 f 117 pm initial crack spacing to 80 l 50 firn after cyclic tensile testing at 1500°C). If we compare this crack spacing with that exhibited by a specimen subjected to a tensile test, we observe a very small change, a factor of 1.2 (89i 80pm and 122 f 80 Km, respectively, after cyclic tensile loading and fatigue tests at 1000°C) for the unfailed specimens (tested at 600 and 1OOOC) and smaller by a factor 0.65 for the fatigue failed specimens (tested at room temperature and 15OO’C) (126 f 82 pm and 80 f 50 pm, respectively, after tensile cyclic loading and fatigue testing at 1 SOO’C). Fig. 9. Damage accumulated during cyclic fatigue testing at 600°C (o,,, = 230 MPa, in a horizontal loading direction). Fig. 10. Damage accumulated during cyclic fatigue testing at 1500°C (urnax = 230 MPa, in a horizontal loading direction)
A Dalmas et al Crack density inside the transverse yarns(i. e. yarns elevated temperature. Cracks produced during the perpendicular to the loading direction). The estima- cooling phase of material manufacture have less effect tion has been conducted for cracks that cross the on material behavior as the temperature is increased. At whole thickness of the yarns. This value increases the same time the tendency to produce new cracking is from the initial crack density (2.4+0.7)by a factor reduced of 1. 3 for unfailed specimens(600 and 1000 C) An increase in the modulus of the material is noted and by a factor of 1-7 for those failed by fatigue during cyclic loading at 600C, accompanied by an (room temperature and 1500oC). It should be increase in the residual deformation within the material noted that after the test at 1500 C. these cracks This results from fact that fatigue produces a were well open, up to 50 um(Fig. 10) reduction in the stress transfer responsible for crack Debonding between the entities of the architecture. propagation, most notably where the interfacial shear We observe the first debonding cracks at the stress is sufficiently high. However, the complex archi- nterfaces between weft and warp yarns. mainly at tecture of the matcrial mcans that the crack surfaces are 600 and 1000"C(Fig. 9). The same damage not fat( because of cracking across the transverse yarns has been observed by Lee after failure testing of following fiber contours). Also, their displacement is a 2D C/SiC at 1000C. This kind of damage is multi-directional, producing both crack opening and more likely to occur as the weave is more pro- shearing If this lateral displacement is too large(greater nounced. Some debonding appears between the than the period of the surface roughness of the crack longitudinal fiber bundles and the matrix in the faces). contact is made between the crack faces. stop specimen tested and failed at 1500oC, as well as ping load redistribution racks around transverse yarns at the yarn/ matrix interface Influence of the macroporosity Macropores have 5 CONCLUSIONS an effect on the initiation of cracks parallel to the longitudinal yarns, this is observed at all tempera- The cyclic fatigue behavior of a 2. 5D C/SiC composite tures(Figs 8 10). This cracking tends to separate has been studied over a range of temperatures; 20, 600, the adjacent woven plies and propagates along the 1000 and 1500oC. The analysis of the cyclic fatigue interface between longitudinal and transverse behavior and the damage accumulated has allowed us to yarns(Fig. 9) or through the transverse yarn describe the fatigue mechanism operating within the Figs 8 and 9) material. The effect of cyclic fatigue is, in part, depen- dent on the shear strength at the fibre/ matrix interface (in this work considered on a mesoscopic scale, for the 4 DISCUSSION yarn/yarn interaction). For temperatures below 1000oC cyclic fatigue behavior is governed by both the friction First, it should be noted that, for this material, fatigue between the yarn/yarn or the yarn/ matrix and the resi behavior is independent of temperature below 1000oC. dual stress produced when cooling during manufacture and is linked to the number of cycles and not to the At temperaturcs above 1000C the effect of thermal period of the test. This behavior is in turn, dependent stresses appears to be masked by other effects: these on mechanical effects at the fibre/matrix interface. At merit further investigation temperatures above 1000C, the test period appears to The hypothesis proposed to explain the stiffening influence the fatigue behavior. but this effect requires observed with an increasing number of cycles is to be urther observation further investigated, notably by high-temperature ten Residual stresses in the material are in tension in the sile tests in situ in the sem nteriors of the fiber bundles transverse to the fibers which produces the observed cracking. It is also in ten- s and their coat. ACKNOWLEDGEMENTS ing, promoting debonding at the interface. The stress redistribution that is produced controls thc macroscopi The authors are grateful to the SEp and CNRS for their mechanical behavior nterest and the support in this work. Thanks are parti Cyclic loading produces a cyclic sliding that degrades cularly due to Dr M. Bourgeon, Dr F. Abbe and Dr J the interface. It is this process that leads to a reduction Ph. Richard from SEP, for numerous fruitful discus in fatigue damage since it is more difficult for the matrix sions and for providing the material cracks to propagate into the fiber ncreasing test temperature augments both the life and the elastic modulus of the material. Up to 1000C REfEReNces he fatigue mechanisms operating are attenuated by an increase in temperature. This effect is due, in most part Arends. F.J. The Mail. K. and Kuhnle. 3 lo the reduction of residual stresses at the interface at Mechanical behavior nt sized C/C-SiC tubes under
698 A. Dulmuz et al. l Crack density inside the transverse yarns (i.e. yarns perpendicular to the loading direction). The estimation has been conducted for cracks that cross the whole thickness of the yarns. This value increases from the initial crack density (2.4 f 0.7) by a factor of I.3 for unfailed specimens (600 and IOOO’C) and by a factor of I .7 for those failed by fatigue (room temperature and I 500°C). It should be noted that after the test at 1_5OO”C, these cracks were well open, up to 50 pm (Fig. 10). l Debonding between the entities of the architecture. We observe the first debonding cracks at the interfaces between weft and warp yarns, mainly at 600 and 1000°C (Fig. 9). The same damage has been observed by Lee4 after failure testing of a 2D CjSiC at 1000°C. This kind of damage is more likely to occur as the weave is more pronounced. Some debonding appears between the longitudinal fiber bundles and the matrix in the specimen tested and failed at 15OO”C, as well as cracks around transverse yarns at the yarn/matrix interface. l Influence of the macroporosity. Macropores have an effect on the initiation of cracks parallel to the longitudinal yarns, this is observed at all temperatures (Figs 8%10). This cracking tends to separate the adjacent woven plies and propagates along the interface between longitudinal and transverse yarns (Fig. 9) or through the transverse yarn (Figs 8 and 9). 4 DISCUSSION First, it should be noted that, for this material, fatigue behavior is independent of temperature below lOOo”C, and is linked to the number of cycles and not to the period of the test. This behavior is. in turn, dependent on mechanical effects at the fibrelmatrix interface. At temperatures above lOOO”C, the test period appears to influence the fatigue behavior. but this effect requires further observation. Residual stresses in the material are in tension in the interiors of the fiber bundles, transverse to the fibers. which produces the observed cracking. It is also in tension at the interface between the yarns and their coating, promoting debonding at the interface. The stress redistribution that is produced controls the macroscopic mechanical behavior. Cyclic loading produces a cyclic sliding that degrades the interface. It is this process that leads to a reduction in fatigue damage since it is more difficult for the matrix cracks to propagate into the fibers. Increasing test temperature augments both the life and the elastic modulus of the material. Up to 1000°C the fatigue mechanisms operating are attenuated by an increase in temperature. This effect is due, in most part, to the reduction of residual stresses at the interface at elevated temperature. Cracks produced during the cooling phase of material manufacture have less effect on material behavior as the temperature is increased. At the same time the tendency to produce new cracking is reduced. An increase in the modulus of the material is noted during cyclic loading at 6OO”C, accompanied by an increase in the residual deformation within the material. This results from the fact that fatigue produces a reduction in the stress transfer responsible for crack propagation, most notably where the interfacial shear stress is sufficiently high. However, the complex architecture of the material means that the crack surfaces are not flat (because of cracking across the transverse yarns, following fiber contours). Also, their displacement is multi-directional. producing both crack opening and shearing. If this lateral displacement is too large (grcnter than the period of the surface roughness of the crack faces). contact is made between the crack faces, stopping load redistribution. 5 CONCLUSIONS The cyclic fatigue behavior of a 2.5D C/Sic composite has been studied over a range of temperatures; 20, 600, 1000 and 1500°C. The analysis of the cyclic fatigue behavior and the damage accumulated has allowed us to describe the fatigue mechanism operating within the material. The effect of cyclic fatigue is, in part, dependent on the shear strength at the fibre/matrix interface (in this work considered on a mesoscopic scale, for the yarn/yarn interaction). For temperatures below 1000°C cyclic fatigue behavior is governed by both the friction between the yarn/yarn or the yarn/matrix and the residual stress produced when cooling during manufacture. At temperatures above 1000°C the effect of thermal stresses appears to be masked by other effects; these merit further investigation. The hypothesis proposed to explain the stiffening observed with an increasing number of cycles is to be further investigated, notably by high-temperature tensile tests in situ in the SEM. ACKNOWLEDGEMENTS The authors are grateful to the SEP and CNRS for their interest and the support in this work. Thanks are particularly due to Dr M. Bourgeon, Dr F. Abbt and Dr J. Ph. Richard from SEP, for numerous fruitful discussions and for providing the materials. REFERENCES I Arendts, F. J.. Theuer, A., Mail, K. and Kuhnle. J., Mechanical behavior of different sized C/C-SC tubes under
Mechanical behavior and damage development during cyclic fatigue multiaxial load and temperature up to 1600C Ceramic- 5. Hutchinson, J. w. and Jensen, H. M, Models of fiber Belgium, 18-20 October, 1 debonding and pull-out in brittle composites with fric 2. Camus. G. and Barbier. J. E. Tensile behavior of 2D tion. Mech. Mater. 1990. 9. 139-163 oven C/Sic composites at ambient and elevated tem- 6. Bobet, J. L. and Lamon, J. Thermal residual stresses in perature. In Proc. High Temperature Ceramic Matrix ceramic matrix composites-L. Axisymmetrical model Composites, Santa Barbara, CA, 1995, pp 407-412 and finite element analysis. Acta Metall. Mater., 1995, 43 3. Shuler. S. F. Holmes. J. w. and Wu.x. Influence of 4|-2253 loading frequency on the room temperature fatigue of 7. Marion, S, Influence de I' interphase de carbone pyroli- carbon fibre Sic matrix composite. J. Am. Ceram Soc tique sur I'endommagement et le comportement macro- 1994,77,792 des matcriaux composites ceramique C/Sic. 4. Lee, S.S., Nicholas, T. and Zawada, L. P, Fatigue Ph. D. Thesis. Universite Paris XVIll. 1993 damage mechanisms and environmental effects 8. Shapery, R. A, Thermal expension coefficients of com- he long-term performance of matrix composites. In posites materials based on energy principles. J. Compos Mechanical Testing of Ceramics and Ceramic Composites Mater.,1986,2,380 AMD, Vol. 197, ASME, 1994, pp. 117-155
Mechanical behavior and damage development during cyclic fatigue 699 multiaxiaJ load and temperature up to 1600°C. CeramicCeramic Composites ItI, Belgium, 18-20 October, 1994. 2. Camus, G. and Barbier, J. E., Tensile behavior of 2D woven CjSiC composites at ambient and elevated temperature. In Proc. High Temperature Ceramic Matri.w Composites, Santa Barbara, CA, 1995, pp. 407412. 3. Shuler, S. F., Holmes, J. W. and Wu, X., Influence of loading frequency on the room temperature fatigue of carbon fibre/SiC matrix composite. J. Am. Ceram. Sot., 1994, 77, 792. 4. Lee, S. S., Nicholas, T. and Zawada, L. P., Fatigue damage mechanisms and environmental effects on the long-term performance of matrix composites. In Mechanical Testing of Ceramics and Ceramic Composites. AMD, Vol. 197, ASME, 1994, pQ. 117-155. 5. Hutchinson, J. W. and Jensen, H. M., Models of fiber debonding and pull-out in brittle composites with friction. Mech. Mater., 1990, 9, 139-163. 6. Bobet, J. L. and Lamon, J., Thermal residual stresses in ceramic matrix composites-I. Axisymmetrical model and finite element analysis. Acta MetaN. Mater., 1995, 43, 2241-2253. 7. Marion, S., Influence de t’interphase de carbone pyrolitique sur I’endommagement et le comportement macroscopique des materiaux composites ceramiques C/Sic. Ph.D. Thesis, Universite Paris XVIII. 1993. 8. Shapery, R. A., Thermal expension coefficients of composites materials based on energy principles. J. Compos. Muter., 1986, 2, 38k385