《复合材料 Composites》课程教学资源（学习资料）第五章 陶瓷基复合材料_Effect of fiber architectures on thermal cycling damage of

MATERIALS HIENGE& ENGIEERING ELSEVIER Materials Science and Engineering A 460-461(2007)306-313 www.elseviercom/locate/msea Effect of fiber architectures on thermal cycling damage of C/SiC composites in oxidizing atmosphere Hui Mei, aifei Cheng, Litong Zhang, Yongdong xu onal Key Laboratory of Thermostructure Composite Materials, Northwestem Polytechnical University. Xi'an Shaanxi 710072, People's Republic of China Received 22 November 2006: received in revised form 14 January 2007; accepted 27 February 2007 Abstract Mechanical response of two and three-dimensional carbon-fiber-reinforced Sic-matrix composites(2D and 3D C/SiCs) subjected simultaneously to thermal cycling and mechanical fatigue in oxidizing atmosphere was compared. Damage was assessed by residual strengths and microstructural characterization. Compared with 2D architecture, the braided 3D composites were shown to possess larger strain increment and strain rate during testing, higher retained strength after 50 thermal cycles, and better damage resistance against oxidation and thermal shock. Differences in oxidation regimes and in thermal shock resistance were ascribed in large part to differences in the fiber architectures. It is actually observed that the fiber architectures have critical influences on the orientations of coating cracking the constraints between neighboring fiber bundles, and the matrix crack propagating resistance, which can result eventually in the different damage resistance of the composites @2007 Elsevier B v. All rights rese Keywords: Ceramic matrix composites; Fiber architectures; Damage; Mechanical properties; Microstructure; Thermal cycling 1. Introduction icant mechanical stress in oxidizing atmosphere has not been reported yet, and under the environments mentioned above no Carbon-fiber-reinforced SiC-matrix composites(C/SiCs)are description involving damage comparison of the two compos- currently considered for applications as structural materials in ites with different fiber architectures can be found in the recent both aerospace and other industries [1-3. There are two main literature kinds of C/SiC composites: 2D C/SiC is usually used to fabricate The paper here provides experimental results on the response components such as plates, tubes and shells, and braided 3D of 2D and braided 3D C/SiC composites to thermal cycling C/SiC is used to fabricate items such as nozzles, combustor liners under mechanical fatigue in wet oxygen atmosphere. Residual and thrusters[4, 5]. In many of the instances, a part made of the properties and microstructures of the thermally cycled compos- composite materials is likely to undergo simultaneously thermal ite specimens are presented, and damage induced in the two cycling and mechanical fatigue in severe service environments. architecture composites is evaluated and compared. Efforts were Many previous efforts have been devoted to investigation on made in this investigation to correlate different preform struc- thermal shock/cycling behav the C/SiC composites with tures (i.e, fiber architectures) with damage resistance of the the same preform structures(either 2D or 3D)in absence of the composites against the oxidation and thermal shock external mechanical stress [6-10]. Recently, thermal properties of 2D and 3D C/SiC composites have been basically reported 2. Experimental from room temperature to 1400C by Cheng et al. [11], and the static mechanical responses of 2D and 3D C/C composites 2. Materials preparation were compared by Aly-Hassan et al. [12]. However, mechanical response of the C/SiC composites to thermal cycles and signif- Three-dimensional(braided 3D) preforms were braided by a four-step method using lk T-300 carbon fibers(braiding angle N 220), and two-dimensional(2D)preforms were lami- Corresponding author. Tel: +86 29 88494616: fax: +86 29 88494620 nated with the same lK T-300 carbon fiber fabrics(10/90D) E-mailaddress:phdhuimei@yahoo.com(H.Mei) The composites were prepared by isothermal CVI technolog 0921-5093/S-see front matter O 2007 Elsevier B v. All rights reserved doi:10.1016/msea.2007.02.104

Materials Science and Engineering A 460–461 (2007) 306–313 Effect of fiber architectures on thermal cycling damage of C/SiC composites in oxidizing atmosphere Hui Mei ∗, Laifei Cheng, Litong Zhang, Yongdong Xu National Key Laboratory of Thermostructure Composite Materials, Northwestern Polytechnical University, Xi’an Shaanxi 710072, People’s Republic of China Received 22 November 2006; received in revised form 14 January 2007; accepted 27 February 2007 Abstract Mechanical response of two and three-dimensional carbon-fiber-reinforced SiC-matrix composites (2D and 3D C/SiCs) subjected simultaneously to thermal cycling and mechanical fatigue in oxidizing atmosphere was compared. Damage was assessed by residual strengths and microstructural characterization. Compared with 2D architecture, the braided 3D composites were shown to possess larger strain increment and strain rate during testing, higher retained strength after 50 thermal cycles, and better damage resistance against oxidation and thermal shock. Differences in oxidation regimes and in thermal shock resistance were ascribed in large part to differences in the fiber architectures. It is actually observed that the fiber architectures have critical influences on the orientations of coating cracking, the constraints between neighboring fiber bundles, and the matrix crack propagating resistance, which can result eventually in the different damage resistance of the composites. © 2007 Elsevier B.V. All rights reserved. Keywords: Ceramic matrix composites; Fiber architectures; Damage; Mechanical properties; Microstructure; Thermal cycling 1. Introduction Carbon-fiber-reinforced SiC-matrix composites (C/SiCs) are currently considered for applications as structural materials in both aerospace and other industries [1–3]. There are two main kinds of C/SiC composites: 2D C/SiC is usually used to fabricate components such as plates, tubes and shells, and braided 3D C/SiC is used to fabricate items such as nozzles, combustor liners and thrusters [4,5]. In many of the instances, a part made of the composite materials is likely to undergo simultaneously thermal cycling and mechanical fatigue in severe service environments. Many previous efforts have been devoted to investigation on thermal shock/cycling behavior of the C/SiC composites with the same preform structures (either 2D or 3D) in absence of the external mechanical stress [6–10]. Recently, thermal properties of 2D and 3D C/SiC composites have been basically reported from room temperature to 1400 ◦C by Cheng et al. [11], and the static mechanical responses of 2D and 3D C/C composites were compared by Aly-Hassan et al. [12]. However, mechanical response of the C/SiC composites to thermal cycles and signif- ∗ Corresponding author. Tel.: +86 29 88494616; fax: +86 29 88494620. E-mail address: phdhuimei@yahoo.com (H. Mei). icant mechanical stress in oxidizing atmosphere has not been reported yet, and under the environments mentioned above no description involving damage comparison of the two compos￾ites with different fiber architectures can be found in the recent literature. The paper here provides experimental results on the response of 2D and braided 3D C/SiC composites to thermal cycling under mechanical fatigue in wet oxygen atmosphere. Residual properties and microstructures of the thermally cycled compos￾ite specimens are presented, and damage induced in the two architecture composites is evaluated and compared. Efforts were made in this investigation to correlate different preform struc￾tures (i.e., fiber architectures) with damage resistance of the composites against the oxidation and thermal shock. 2. Experimental 2.1. Materials preparation Three-dimensional (braided 3D) preforms were braided by a four-step method using 1K T-300 carbon fibers (braiding angle ≈ 22◦), and two-dimensional (2D) preforms were lami￾nated with the same 1K T-300 carbon fiber fabrics ([0◦/90◦]). The composites were prepared by isothermal CVI technology 0921-5093/$ – see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.msea.2007.02.104

H Mei et al. /Materials Science and Engineering A 460-461(2007)306-313 p su ide surfac 150kv120mmx35 Fig. 1. SEM micrographs showing fiber architectures of the as-fabricated C/SiC composites, (a) 2D and(b)braided 3D preform structures at about 1000oC. The volume fractions of fibers for the 3d 23. Measurements and observations braided architecture and 2d architecture were about 40 and 37%0, respectively. The dog-bone shaped test samples were cut Mechanical strengths of the composite specimens before from the fabricated composite plates and further coated with Sic and after thermal cycles were measured using an Instron y I-CVI under the same conditions(thickness a 50 um). Fiber tester(Model 1196, Instron Ltd, High Wycombe, England) at architectures of the as-received 2D and braided 3D composite room temperature. Fractured sections and coating surfaces were mples are shown in Fig. la and b The virgin properties of the observed with a scanning electron microscope(SEM, HITACHI s-received composite samples are listed in Table 1 S-4700). 2.2. Thermal cycling test 3. Results and discussion Thermal cycling experiments were conducted with an 3.1. Monotonic tensile behaviors integrated system (see the details in Fig. 2 of Ref. [13)), including an induction heating furnace and a servo-hydraulic The static tensile monotonic stress-strain behaviors of the tester(Instron 8801 d ). 2D and braided 3D C/SiC composite materials were measured The dimensions of the dog-bone shaped specimens were to rupture on the Instron tester with a loading rate of 0.001 mm/s 185 mm x 3 mm x 3 mm as illustrated in the middle of the fur- at room temperature. Typical monotonic tensile stress-strain nace. As shown later in Fig 3, thermal cycling was carried out curves, obtained from one example of several 2D and braided between 900 and 1200C over a period of 120s(temperature 3D C/SiC composite samples, are shown in Fig. 2. The com- gradient ATN300C) Only the middle parts of the specimens posites behave as a typical damageable material, exhibiting an (40 mm long, 3 mm wide, and 3 mm thick) were kept in the hot extensive non-linear stress-strain domain up to rupture because zone and wet oxygen atmosphere: 7.90 vol %o O2/14.85 vol %o of the presence of the processing-induced microcracks and of H,O/77.25 vol. Ar Loading mode was tension-tension fatigue the damage accumulation nature of materials during testing (sine wave, frequency: 1 Hz, stress: 60+20 MPa, and stress ratio The linear deformation of the braided 3D C/SiC composite is R=0.5). Strains were assessed directly from gauge length of clearly limited up to about 50 MPa(referred to as"proportional specimen by a contact Instron extensometer with a gauge length limit"or first-matrix cracking stress oM), after which the behav of 10mm. Coefficient of thermal expansion(CTEof the com- iors become non-linear. Moreover, the slope of the tensile curve posite was assessed by a dilatometer(Model DIL 402C, Netszch continuously decreases as the stress increases. However, for the Ltd, Selb, Germany) 2D architecture composite, the non-linearity starts almost from Table Properties of the as- 3D C/SiC composites Materials Density (GPa) Strength(MPa) Failure strain(%6) Porosity(%) CTE (x10-6oC-I 900°C1000°C1100°C1200°C 9123 52.45 0.71 4.2785452654.6125 364153.77664.32934.3719 fficient of thermal expansion

H. Mei et al. / Materials Science and Engineering A 460–461 (2007) 306–313 307 Fig. 1. SEM micrographs showing fiber architectures of the as-fabricated C/SiC composites, (a) 2D and (b) braided 3D preform structures. at about 1000 ◦C. The volume fractions of fibers for the 3D braided architecture and 2D architecture were about 40 and 37%, respectively. The dog-bone shaped test samples were cut from the fabricated composite plates and further coated with SiC by I-CVI under the same conditions (thickness ≈ 50m). Fiber architectures of the as-received 2D and braided 3D composite samples are shown in Fig. 1a and b. The virgin properties of the as-received composite samples are listed in Table 1. 2.2. Thermal cycling test Thermal cycling experiments were conducted with an integrated system (see the details in Fig. 2 of Ref. [13]), including an induction heating furnace and a servo-hydraulic tester (Instron 8801, Instron Ltd., High Wycombe, England). The dimensions of the dog-bone shaped specimens were 185 mm × 3 mm × 3 mm as illustrated in the middle of the fur￾nace. As shown later in Fig. 3, thermal cycling was carried out between 900 and 1200 ◦C over a period of 120 s (temperature gradient T ≈ 300 ◦C). Only the middle parts of the specimens (40 mm long, 3 mm wide, and 3 mm thick) were kept in the hot zone and wet oxygen atmosphere: 7.90 vol.% O2/14.85 vol.% H2O/77.25 vol.% Ar. Loading mode was tension–tension fatigue (sine wave, frequency: 1 Hz, stress: 60 ± 20 MPa, and stress ratio R = 0.5). Strains were assessed directly from gauge length of specimen by a contact Instron extensometer with a gauge length of 10 mm. Coefficient of thermal expansion (CTE) of the com￾posite was assessed by a dilatometer (Model DIL 402 C, Netszch Ltd., Selb, Germany). 2.3. Measurements and observations Mechanical strengths of the composite specimens before and after thermal cycles were measured using an Instron tester (Model 1196, Instron Ltd., High Wycombe, England) at room temperature. Fractured sections and coating surfaces were observed with a scanning electron microscope (SEM, HITACHI S-4700). 3. Results and discussion 3.1. Monotonic tensile behaviors The static tensile monotonic stress–strain behaviors of the 2D and braided 3D C/SiC composite materials were measured to rupture on the Instron tester with a loading rate of 0.001 mm/s at room temperature. Typical monotonic tensile stress–strain curves, obtained from one example of several 2D and braided 3D C/SiC composite samples, are shown in Fig. 2. The com￾posites behave as a typical damageable material, exhibiting an extensive non-linear stress–strain domain up to rupture because of the presence of the processing-induced microcracks and of the damage accumulation nature of materials during testing. The linear deformation of the braided 3D C/SiC composite is clearly limited up to about 50 MPa (referred to as “proportional limit” or first-matrix cracking stress σM), after which the behav￾iors become non-linear. Moreover, the slope of the tensile curve continuously decreases as the stress increases. However, for the 2D architecture composite, the non-linearity starts almost from Table 1 Properties of the as-received 2D and braided 3D C/SiC composites Materials Density (g/cm3) Modulus (GPa) Strength (MPa) Failure strain (%) Porosity (%) CTE (×10−6 ◦C−1) 900 ◦C 1000 ◦C 1100 ◦C 1200 ◦C 2D C/SiC 2.20 91.23 252.45 0.71 10.66 4.2785 4.5265 4.6125 4.1855 3D C/SiC 2.16 142.85 413.76 0.92 13.74 3.6415 3.7766 4.3293 4.3719 CTE, Coefficient of thermal expansion.

H Mei et aL. Materials Science and Engineering A 460-461(2007)306-313 1600 3D C/SIC 0.33 0.30 2D CSIC g150 024 100 060112011801240130013601420 Time(s) Fig. 3. Correlation of the cyclic temperature and the fatigue stress with the strain 0.00.1020.30.40.5060.70.80.91.0 of the braided 3D C/SiC composites subjected to thermal cycling and mechanical Tensile strain(%) Fig.2. Typical tensile stress-strain curves of 2D and braided 3D C/SiC com- due to loading/unloading followed by a time-dependent damage posites at room temperature with a loading rate of 0.001 mm/s strain baseline. As shown in Fig. 3, thermal strain of the tested specimen increases gradually, reaching a peak as the tempera- the onset of loading. Furthermore, the stress-strain relationship ture ascends to the upper limit of 1200.C, and then decreases tends to have apparent slope recovery above 150 MPa, indicat- with cooling back to the lower limit of 900CAs thermal cycles ing that the o fiber bundles in 2D architecture can enhance proceed, the quasi-triangular strain is repeated periodically with the transverse compression resistance to hinder the longitudinal a fixed range magnitude as the same period of 120s as the extension at higher stress level. It is interesting to note that near cyclic temperature, independent upon the external fatigue stress the top of the loading curves of the braided 3D composite, at [13, 14]. It is believed that the thermal strain range depends sig the point where a saturated matrix cracking state was believed nificantly on the temperature gradient AT. On the other hand to have been reached and no more matrix cracks and interfacial it can be also seen in Fig 3 that the mechanical fatigue stress debonding were believed to form, higher modulus fibers and provides a significant influence on the strain induced. It is wor- bundles split in a sudden manner leading to an apparent"stiff- thy to be noted that the stress-induced strain is repeated with ening"on the stress-strain curve On the other hand, the apparent the cyclic fatigue stress between 40 and 80MPa. No matter stiffening of the 3D braided composite is likely due to reorienta- how the temperature is high or low, the mechanical strain range tion of the fibers angle during loading. Initially the fibers are at an also always sustains a constant range magnitude, dependent angle substantially different from the stress axis. As the stress is only on the stress difference Ao and independent upon the applied, the braid may stretch in such a way that the fibers move cyclic temperature. Furthermore, the mechanical strains are reg- closer to the same orientation of the stress axis( see the details ularly distributed on the quasi-triangular thermal strain wave as in Fig. 8 of Ref [14)). As this happens, the fibers will carry the same period as the fatigue stress. Evidently, the mechan- more of the load and the matrix less, thus resulting in""stiffen- ical strain is considerably small in comparison to the thermal ing". Later, as fibers begin to fracture, accumulation of damage strain will cause the curve to bend over prior to failure. Nevertheless, The entire strain versus time curves for the 2d and braided right before the sudden failure of the braided 3D composite, the 3D C/Sic composites subjected to fatigue loading and thermal lope of the stress-strain curve decreases again. The average ten- cycling(N=50) in wet oxygen are compared in Fig. 4. The sile strengths and failure strains are 413.76 MPa,0.92% for the fitting strain of 2D composites varies approximately from the braided 3D composites, and 252. 45 MPa, 0.71%o for the 2D com- initial transient strain of 0.23% to the final nonreversible dam- posites, respectively. The average Youngs modulus obtained age strain of 0.35%, whereas for the 3d braided architecture by the linear fitting of the initial stress-strain curves from 0 to these two strain values are about 0. 21 and 0.47%o, respectively 60MPa is 142.85 GPa for the braided 3D and 91.23 GPa for 2d It should be noted that the two strain curves become parted from C/SiC composites each other after experiencing the initial nearly same transient increased stage. Compared with 2D architecture, the 3D braided 3.2. Strain response of the composites architecture composite has a larger strain in a higher strain rate under the same testing conditions Thus the deviation between Strain response of the 2D and braided 3D composites to ther- the 2D and braided 3D composite strain curves becomes greater mal cycling and mechanical fatigue stress is found to be the and greater with increasing thermal cycles. The fiber architec similar. Taking from an example of several braided 3D com- tures in the 2D and braided 3D fiber preforms must be considered posite samples, Fig. 3 indicates the correlation of the cyclic to be responsible for this result In 2D architecture, volume frac temperature and fatigue stress with the strain induced. It is tions of the longitudinal(90%)and transverse(0%)fibers are the apparent that the measured strain should be a combined result same, and equal to half the total fiber volume fraction. Fur- of thermal strain due to heating/cooling and mechanical strain thermore the longitudinal extension strain of the fibers can be

308 H. Mei et al. / Materials Science and Engineering A 460–461 (2007) 306–313 Fig. 2. Typical tensile stress–strain curves of 2D and braided 3D C/SiC com￾posites at room temperature with a loading rate of 0.001 mm/s. the onset of loading. Furthermore, the stress–strain relationship tends to have apparent slope recovery above 150 MPa, indicat￾ing that the 0◦ fiber bundles in 2D architecture can enhance the transverse compression resistance to hinder the longitudinal extension at higher stress level. It is interesting to note that near the top of the loading curves of the braided 3D composite, at the point where a saturated matrix cracking state was believed to have been reached and no more matrix cracks and interfacial debonding were believed to form, higher modulus fibers and bundles split in a sudden manner leading to an apparent “stiff￾ening” on the stress–strain curve. On the other hand, the apparent stiffening of the 3D braided composite is likely due to reorienta￾tion of the fibers angle during loading. Initially the fibers are at an angle substantially different from the stress axis. As the stress is applied, the braid may stretch in such a way that the fibers move closer to the same orientation of the stress axis (see the details in Fig. 8 of Ref. [14]). As this happens, the fibers will carry more of the load and the matrix less, thus resulting in “stiffen￾ing”. Later, as fibers begin to fracture, accumulation of damage will cause the curve to bend over prior to failure. Nevertheless, right before the sudden failure of the braided 3D composite, the slope of the stress–strain curve decreases again. The average ten￾sile strengths and failure strains are 413.76 MPa, 0.92% for the braided 3D composites, and 252.45 MPa, 0.71% for the 2D com￾posites, respectively. The average Young’s modulus obtained by the linear fitting of the initial stress–strain curves from 0 to 50 MPa is 142.85 GPa for the braided 3D and 91.23 GPa for 2D C/SiC composites. 3.2. Strain response of the composites Strain response of the 2D and braided 3D composites to ther￾mal cycling and mechanical fatigue stress is found to be the similar. Taking from an example of several braided 3D com￾posite samples, Fig. 3 indicates the correlation of the cyclic temperature and fatigue stress with the strain induced. It is apparent that the measured strain should be a combined result of thermal strain due to heating/cooling and mechanical strain Fig. 3. Correlation of the cyclic temperature and the fatigue stress with the strain of the braided 3D C/SiC composites subjected to thermal cycling and mechanical fatigue. due to loading/unloading followed by a time-dependent damage strain baseline. As shown in Fig. 3, thermal strain of the tested specimen increases gradually, reaching a peak as the tempera￾ture ascends to the upper limit of 1200 ◦C, and then decreases with cooling back to the lower limit of 900 ◦C. As thermal cycles proceed, the quasi-triangular strain is repeated periodically with a fixed range magnitude as the same period of 120 s as the cyclic temperature, independent upon the external fatigue stress [13,14]. It is believed that the thermal strain range depends sig￾nificantly on the temperature gradient T. On the other hand, it can be also seen in Fig. 3 that the mechanical fatigue stress provides a significant influence on the strain induced. It is wor￾thy to be noted that the stress-induced strain is repeated with the cyclic fatigue stress between 40 and 80 MPa. No matter how the temperature is high or low, the mechanical strain range also always sustains a constant range magnitude, dependent only on the stress difference σ and independent upon the cyclic temperature. Furthermore, the mechanical strains are reg￾ularly distributed on the quasi-triangular thermal strain wave as the same period as the fatigue stress. Evidently, the mechan￾ical strain is considerably small in comparison to the thermal strain. The entire strain versus time curves for the 2D and braided 3D C/SiC composites subjected to fatigue loading and thermal cycling (N= 50) in wet oxygen are compared in Fig. 4. The fitting strain of 2D composites varies approximately from the initial transient strain of 0.23% to the final nonreversible dam￾age strain of 0.35%, whereas for the 3D braided architecture, these two strain values are about 0.21 and 0.47%, respectively. It should be noted that the two strain curves become parted from each other after experiencing the initial nearly same transient increased stage. Compared with 2D architecture, the 3D braided architecture composite has a larger strain in a higher strain rate under the same testing conditions. Thus, the deviation between the 2D and braided 3D composite strain curves becomes greater and greater with increasing thermal cycles. The fiber architec￾tures in the 2D and braided 3D fiber preforms must be considered to be responsible for this result. In 2D architecture, volume frac￾tions of the longitudinal (90◦) and transverse (0◦) fibers are the same, and equal to half the total fiber volume fraction. Fur￾thermore, the longitudinal extension strain of the fibers can be

H Mei et al. Materials Science and Engineering A 460-461(2007)306-313 Thermal cycle number, N After 50 thermal cycles in the oxidizing atmosphere, resid- 10 al strengths of the C/SiC specimens were measured with a loading rate of 0.001 mm/s at room temperature. The statisti 9808 cal mean strengths of the thermally cycled 2D and braided 3D composite specimens retained 83.2 and.6% of the initial prop- erties. Obviously, related to the 3D braided architecture, the 2D C/SiC composites suffer greater loss in the strength. The differ ent reduction in the retained strengths could be better interpreted by the following observed microstructures, which caused differ s ent damage resistance of the composites against oxidation and thermal shock SiC fitting 33. Microstructural observations SEMmicrographs of the outer coatings and fractured sections 0 1200 2400 3600 4800 6000 of the thermally cycled composites are presented in Figs. 5- Time(s) It can be seen from Figs. 5 and 6 that the irregular netty cracks and the highly oriented wavy cracks could be found on the top Fig. 4. Strain vs time curves of the 2D and braided 3D C/SiC composites surfaces(as defined in Fig. la and b)of the 2D and braided 3D ubjected to thermal cycling and mechanical fatigue, and their fitting curves specimens, whilst the coating cracks on the side surfaces for the two architectures exhibit the same transverse direction(verti resisted and hindered by the transverse fibers. As a result, it is cal to tensile axis). It is strongly believed that thermal cycles more and more difficult for the 2D architecture to be elongated made the coating cracks strictly arranged in the direction per- with increasing transverse compression resistance. By contrast, pendicular to the fibers in composites since the SiC matrix has the 3D braided architecture can easily extend in the longitudi- a greater CtE than the longitudinal fiber. As a result, the netty nal direction since all the fibers are laid at a small angle(22%) cracks and wavy cracks were formed on the top surface coatings along the tensile axis and the porous CVI-Sic matrix has a poor of the 2D architecture and 3D braided architecture composites, transverse compression resistance. As mentioned previously in respectively. Simultaneously, both side surface coatings always Section 3. 1, when the 3D braided architecture preform is loaded generated the transverse cracks. It is more important that the with a longitudinal stress, all the neighboring two fiber bundles arrangement orientation of the cracks have a significant effect will move closer and rotate around their contacts, and the braid- on the oxidation resistance of the composites since the coating ing angle decreases gradually. Consequently, the longitudinal cracks will serve as avenues for the ingress of the environment strain of the braided 3D composites seems not to be limited into the composite. The highly ordered orientations are con- until the fibers are stretched straightly up to final rupture. Fur- sidered to be of advantage to the crack closure for the braided thermore, under the effect of the reloading and unloading in the 3D composites upon heating. Comparatively, the irregular netty tensile axis, the coating regions between the two neighboring crack distributions are of impediment for the crack closure for fiber bundles can be compressed and corrugated, leading to the the 2D composites. As a consequence, because oxygen has an wave-shaped coating cracking at relatively regular spacing(as easier path into the 2D composite, the fibers in the 2D composite are more exposed to oxidation than the fibers in the braided 3D 5.0Kv 101mm x45 SE Fig. 5. Typical micrographs showing the outer coating cracks on the 2D C/SiC composites after 50 thermal cycles: (a)top surface and (b)side surface

H. Mei et al. / Materials Science and Engineering A 460–461 (2007) 306–313 309 Fig. 4. Strain vs. time curves of the 2D and braided 3D C/SiC composites subjected to thermal cycling and mechanical fatigue, and their fitting curves. resisted and hindered by the transverse fibers. As a result, it is more and more difficult for the 2D architecture to be elongated with increasing transverse compression resistance. By contrast, the 3D braided architecture can easily extend in the longitudi￾nal direction since all the fibers are laid at a small angle (∼22◦) along the tensile axis and the porous CVI-SiC matrix has a poor transverse compression resistance. As mentioned previously in Section 3.1, when the 3D braided architecture preform is loaded with a longitudinal stress, all the neighboring two fiber bundles will move closer and rotate around their contacts, and the braid￾ing angle decreases gradually. Consequently, the longitudinal strain of the braided 3D composites seems not to be limited until the fibers are stretched straightly up to final rupture. Fur￾thermore, under the effect of the reloading and unloading in the tensile axis, the coating regions between the two neighboring fiber bundles can be compressed and corrugated, leading to the wave-shaped coating cracking at relatively regular spacing (as shown later in Fig. 6). After 50 thermal cycles in the oxidizing atmosphere, resid￾ual strengths of the C/SiC specimens were measured with a loading rate of 0.001 mm/s at room temperature. The statisti￾cal mean strengths of the thermally cycled 2D and braided 3D composite specimens retained 83.2 and 91.6% of the initial prop￾erties. Obviously, related to the 3D braided architecture, the 2D C/SiC composites suffer greater loss in the strength. The differ￾ent reduction in the retained strengths could be better interpreted by the following observed microstructures, which caused differ￾ent damage resistance of the composites against oxidation and thermal shock. 3.3. Microstructural observations SEM micrographs of the outer coatings and fractured sections of the thermally cycled composites are presented in Figs. 5–8. It can be seen from Figs. 5 and 6 that the irregular netty cracks and the highly oriented wavy cracks could be found on the top surfaces (as defined in Fig. 1a and b) of the 2D and braided 3D specimens, whilst the coating cracks on the side surfaces for the two architectures exhibit the same transverse direction (verti￾cal to tensile axis). It is strongly believed that thermal cycles made the coating cracks strictly arranged in the direction per￾pendicular to the fibers in composites since the SiC matrix has a greater CTE than the longitudinal fiber. As a result, the netty cracks and wavy cracks were formed on the top surface coatings of the 2D architecture and 3D braided architecture composites, respectively. Simultaneously, both side surface coatings always generated the transverse cracks. It is more important that the arrangement orientation of the cracks have a significant effect on the oxidation resistance of the composites since the coating cracks will serve as avenues for the ingress of the environment into the composite. The highly ordered orientations are con￾sidered to be of advantage to the crack closure for the braided 3D composites upon heating. Comparatively, the irregular netty crack distributions are of impediment for the crack closure for the 2D composites. As a consequence, because oxygen has an easier path into the 2D composite, the fibers in the 2D composite are more exposed to oxidation than the fibers in the braided 3D Fig. 5. Typical micrographs showing the outer coating cracks on the 2D C/SiC composites after 50 thermal cycles: (a) top surface and (b) side surface

310 H Mei et aL. Materials Science and Engineering A 460-461(2007)306-313 1sa√10tmm×60sEN Fig. 6. Typical micrographs showing the outer coating cracks on the braided 3D C/SiC composites after 50 thermal cycles:(a) top surface and (b)side surface. Fig. 7. Micrographs showing the fractured sections of the 2D C/SiC composites after 50 thermal cycles in oxidizing atmosphere: (a)uniform oxidation of the fibers and (b)fiber pullouts composite. As also confirmed by the microstructural observa- specimens while in the braided 3D specimens only a superficial tions the 3d braided architecture has better oxidation resistand oxidation takes place beneath the coatings. Both thermal cycling than the 2D architecture. Figs. 7a and &a indicate that a large uni- and mechanical fatigue stress can cause opening-closing eff form oxidation can be found on the fractured section of the 2D on the microcracks periodically. As illustrated in Figs. 7b and /Coating 5K93m×110sEM Fig 8. Micrographs showing the fractured sections of the braided 3D C/SiC composites after 50 thermal cycles in oxidizing atmosphere:(a)superficial oxidation of the fibers and (b) fiber pullo

310 H. Mei et al. / Materials Science and Engineering A 460–461 (2007) 306–313 Fig. 6. Typical micrographs showing the outer coating cracks on the braided 3D C/SiC composites after 50 thermal cycles: (a) top surface and (b) side surface. Fig. 7. Micrographs showing the fractured sections of the 2D C/SiC composites after 50 thermal cycles in oxidizing atmosphere: (a) uniform oxidation of the fibers and (b) fiber pullouts. composite. As also confirmed by the microstructural observa￾tions, the 3D braided architecture has better oxidation resistance than the 2D architecture. Figs. 7a and 8a indicate that a large uni￾form oxidation can be found on the fractured section of the 2D specimens while in the braided 3D specimens only a superficial oxidation takes place beneath the coatings. Both thermal cycling and mechanical fatigue stress can cause opening–closing effects on the microcracks periodically. As illustrated in Figs. 7b and Fig. 8. Micrographs showing the fractured sections of the braided 3D C/SiC composites after 50 thermal cycles in oxidizing atmosphere: (a) superficial oxidation of the fibers and (b) fiber pullouts

H Mei et al. Materials Science and Engineering A 460-461(2007)306-313 ac Exterior Fig 9. SEM micrograph a fractured fiber due to the constraint stress t between longitudinal (90 )and transverse(0 )fiber bundles in 2D C/SiC composites during thermal cycling. The external fatigue stress a is vertical. 8b, it appears that large quantities of fibers and/or bundles are that the longer fiber debonding and sliding during the testing pulled out together with the covered SiC matrix, and the pullout resulted eventually in the longer fiber pullout lengths in the next length of the fibers are longer for the braided 3D composites monotonic tension 500 um) than that for the 2D composites(100 um). Under The constraints between longitudinal(90%)and transverse thermal cycling and fatigue stress, the cyclic unloading and (0%)fiber bundles in 2D C/SiC composites are likely to enhance reloading increase the debonding length by decreasing interface physical destruction resulting from the cyclic thermal mismatch sliding resistance given by [15] Fig 9 shows that a bridging fiber was fractured by the con- (1) fiber bundles in 2D C/SiC composites during thermal cycin .y between longitudinal (90%)and transverse(0 It is easy for the matrix cracks to be formed at the intersec- where of is the fiber strength, d the diameter of the carbon fiber tions(crossovers)of the neighboring fiber bundles, where the (7 um as indicated later in Fig 9)and Le is the critical length thermal stress can be generated in the two perpendicular direc- of the broken fiber. Normally, it is thought that the fiber pullout tions (i.e, 90 and 0o)and the constrained thermal stress is length is equal to Lc/2 eventually relaxed by shearing the bridging fibers(Fig 9)along The cyclic fatigue stress or repetitive temperature could the propagating cracks For the braided 3D composites, all the reduce the sliding resistance tr of the interface by the friction and fibers are laid at a small angle(22)along the longitudinal wear effect between fiber and matrix. This process increased the axis. This fiber architecture is helpful for relaxation of thermal length of the debonded interface and enabled the broken fibers stress via deforming composites longitudinally and adjusting the to slide along the interfaces, leading to long fiber pullouts when braiding angle properly. As also illustrated in Fig. 4, it is actu- the fiber strength of is assumed to be constant(3.05 GPa)in ally observed that the 3D braided architecture exhibits better (1). As mentioned earlier in Fig 4, the braided 3D compos- deformability than the 2D architecture. As we know, the physi- ite have a larger strain incremental amount and strain rate than cal damage created by the cyclic thermal mismatch can facilitate the 2D composites during thermal cycling. It is not surprising fiber oxidation leading to mechanical degradation of the com-

H. Mei et al. / Materials Science and Engineering A 460–461 (2007) 306–313 311 Fig. 9. SEM micrographs showing a fractured fiber due to the constraint stress τ between longitudinal (90◦) and transverse (0◦) fiber bundles in 2D C/SiC composites during thermal cycling. The external fatigue stress σ is vertical. 8b, it appears that large quantities of fibers and/or bundles are pulled out together with the covered SiC matrix, and the pullout length of the fibers are longer for the braided 3D composites (≈500m) than that for the 2D composites (≈100m). Under thermal cycling and fatigue stress, the cyclic unloading and reloading increase the debonding length by decreasing interface sliding resistance given by [15] τr = σfd 2Lc (1) where σf is the fiber strength, d the diameter of the carbon fiber (≈7m as indicated later in Fig. 9) and Lc is the critical length of the broken fiber. Normally, it is thought that the fiber pullout length is equal to Lc/2. The cyclic fatigue stress or repetitive temperature could reduce the sliding resistance τr of the interface by the friction and wear effect between fiber and matrix. This process increased the length of the debonded interface and enabled the broken fibers to slide along the interfaces, leading to long fiber pullouts when the fiber strength σf is assumed to be constant (≈3.05 GPa) in Eq. (1). As mentioned earlier in Fig. 4, the braided 3D compos￾ite have a larger strain incremental amount and strain rate than the 2D composites during thermal cycling. It is not surprising that the longer fiber debonding and sliding during the testing resulted eventually in the longer fiber pullout lengths in the next monotonic tension. The constraints between longitudinal (90◦) and transverse (0◦) fiber bundles in 2D C/SiC composites are likely to enhance physical destruction resulting from the cyclic thermal mismatch. Fig. 9 shows that a bridging fiber was fractured by the con￾straint stress τ between longitudinal (90◦) and transverse (0◦) fiber bundles in 2D C/SiC composites during thermal cycling. It is easy for the matrix cracks to be formed at the intersec￾tions (crossovers) of the neighboring fiber bundles, where the thermal stress can be generated in the two perpendicular direc￾tions (i.e., 90◦ and 0◦) and the constrained thermal stress is eventually relaxed by shearing the bridging fibers (Fig. 9) along the propagating cracks. For the braided 3D composites, all the fibers are laid at a small angle (∼22◦) along the longitudinal axis. This fiber architecture is helpful for relaxation of thermal stress via deforming composites longitudinally and adjusting the braiding angle properly. As also illustrated in Fig. 4, it is actu￾ally observed that the 3D braided architecture exhibits better deformability than the 2D architecture. As we know, the physi￾cal damage created by the cyclic thermal mismatch can facilitate fiber oxidation leading to mechanical degradation of the com-

312 H Mei et aL. Materials Science and Engineering A 460-461(2007)306-313 posites. The greater the extents of the thermal cycling damage the more the matrix cracks acting as the oxygen tunnels and the severer the fiber oxidation of the composites. Therefore, the 2D architecture has a poorer thermal shock resistance than the 3D braided architecture. which in turn results in the lower oxidation resistances of 2D composites It is interesting to discuss the influence of the fiber architec 90 tures in composites on the crack propagation resistance, which Cracking is also believed to be a contributing factor for the different oxidation regimes. According to the oxidation kinetics mech developed by Eckel et al. [16], the reaction-controlled and diffusion-controlled kinetics can be used to interpret the crAcking different oxidation regimes from Figs. 7a and &a. On the fractured sections of the 2D C/SiC composite specimens,a great number of fibers became thinner and thinner, indicat- ing that the oxidizing atmosphere diffuses inwards along the opening and propagating cracks largely and easily, leading Fig 10. A propagating crack penetrates easily through the transverse%fiber to a uniform oxidation governed by the reaction-controlled bundles vertical to the tensile axis, and eventually is deflected and hindered by kinetic as the high strength longitudinal 90. fibers. PeRT / Ko ex 2) mainly deflect longitudinally or arrested beneath the coatings to form a superficial oxidation governed by diffusion-controlled where x is termed the recession distance of carbon phase from kinetics the surface into the center. Under the same environmental condi- tions, a typical superficial oxidation governed by the diffusion- 4. Conclusions ontrolled kinetics occurs in the braided 3D C/SiC composites owing to the self-closure of the highly oriented transverse During thermal cycling between 900 and 1200.C and fatigue cracks, as stress of 60+ 20 MPa, the braided 3D C/SiC composites exhibit (1+x)Dk/D)+1 larger strain increment, better oxidation resistance and thermal (3) shock resistance than the 2D composites. As indicated by the D/D+1 residual strength measurements, the thermally cycled braided the parabolic rate constant, respectively. x is the oxidant par- 2D composites. Differences in the fiber architectures are taken tial pressure, P the total pressure(Pa), Pe the molar density into account to be response for the results with respect to the of carbon(mol/m ), R the gas constant(J/mol K), T the abso- following lute temperature (K), ko a constant(m/s), e the activation energy (/mol), t the duration of the test, and Dk and D are (1)Thermal cycling and fatigue stress resulted in the irregu- the Knudsen diffusion coefficient and Fick diffusion coefficient. lar netty cracks and regular wavy cracks on the top surface respectively. coatings of the 2D architecture and 3d braided architecture The crack initiation and propagation depend significantly on composites, respectively. Simultaneously, both side surface the arrangements of the fibers in composites. Generally, the coatings always generated the highly oriented transverse external stress-induced cracks initiate on the brittle ceramic cracks. Differences in the self-closure capacity of these coatings disregarding the fiber architectures in the preforms cracks during heating lead to the different oxidation resis- Subsequently, the substantial crack propagation pattern is tance of the two architectures strongly relevant to the fiber architectures and has a key influence (2) The constraints between longitudinal(90) and transverse on oxidation regimes. As shown in Fig. 10, a propagating crack (0)fiber bundles in 2D architecture are likely to enhance penetrates easily through the transverse 0o fiber bundles vertical physical destruction during thermal cycling, through which to the tensile axis, and eventually is deflected and hindered by the fiber oxidation can be aggravated. In contrast, the the high strength longitudinal 90 fibers. Thus, only one half of 3d braided architecture exhibits excellent thermal shock the fibers in 2D architecture (i.e, longitudinal fibers )can effec resistance by the collective and highly oriented transverse tively hinder the transverse crack propagation. As a result, the cracking. cracks in 2D composites can propagate rapidly from the outer (3) Only one half of the fibers in the 2D architecture can effec- coatings into the cores through which the fibers are oxidized uni- tively hinder the transverse crack propagation from the formly according to the reaction-controlled kinetics mechanism. surface to the center whereas almost all the fibers in the By contrast, almost all the fibers in the 3d braided architecture 3D braided architecture can deflect longitudinally the trans can resist the crack growth transversely. The transverse cracks verse cracks, which result in the uniform oxidation governed hardly propagate across so many longitudinal fibers, and are by a linear reaction-controlled kinetics mechanism for the

312 H. Mei et al. / Materials Science and Engineering A 460–461 (2007) 306–313 posites. The greater the extents of the thermal cycling damage, the more the matrix cracks acting as the oxygen tunnels and the severer the fiber oxidation of the composites. Therefore, the 2D architecture has a poorer thermal shock resistance than the 3D braided architecture, which in turn results in the lower oxidation resistances of 2D composites. It is interesting to discuss the influence of the fiber architec￾tures in composites on the crack propagation resistance, which is also believed to be a contributing factor for the different oxidation regimes. According to the oxidation kinetics mech￾anisms developed by Eckel et al. [16], the reaction-controlled and diffusion-controlled kinetics can be used to interpret the different oxidation regimes from Figs. 7a and 8a. On the fractured sections of the 2D C/SiC composite specimens, a great number of fibers became thinner and thinner, indicat￾ing that the oxidizing atmosphere diffuses inwards along the opening and propagating cracks largely and easily, leading to a uniform oxidation governed by the reaction-controlled kinetic as x = Klt = χP ρcRT k0 exp − Q RT t (2) where x is termed the recession distance of carbon phase from the surface into the center. Under the same environmental condi￾tions, a typical superficial oxidation governed by the diffusion￾controlled kinetics occurs in the braided 3D C/SiC composites owing to the self-closure of the highly oriented transverse cracks, as x2 = Kpt = 4D P ρcRT ln (1 + χ)(Dk/D) + 1 Dk/D + 1 t (3) where Kl, Kp are referred to as the linear rate constant and the parabolic rate constant, respectively. χ is the oxidant par￾tial pressure, P the total pressure (Pa), ρc the molar density of carbon (mol/m3), R the gas constant (J/mol K), T the abso￾lute temperature (K), k0 a constant (m/s), Q the activation energy (J/mol), t the duration of the test, and Dk and D are the Knudsen diffusion coefficient and Fick diffusion coefficient, respectively. The crack initiation and propagation depend significantly on the arrangements of the fibers in composites. Generally, the external stress-induced cracks initiate on the brittle ceramic coatings disregarding the fiber architectures in the preforms. Subsequently, the substantial crack propagation pattern is strongly relevant to the fiber architectures and has a key influence on oxidation regimes. As shown in Fig. 10, a propagating crack penetrates easily through the transverse 0◦ fiber bundles vertical to the tensile axis, and eventually is deflected and hindered by the high strength longitudinal 90◦ fibers. Thus, only one half of the fibers in 2D architecture (i.e., longitudinal fibers) can effec￾tively hinder the transverse crack propagation. As a result, the cracks in 2D composites can propagate rapidly from the outer coatings into the cores, through which the fibers are oxidized uni￾formly according to the reaction-controlled kinetics mechanism. By contrast, almost all the fibers in the 3D braided architecture can resist the crack growth transversely. The transverse cracks hardly propagate across so many longitudinal fibers, and are Fig. 10. A propagating crack penetrates easily through the transverse 0◦ fiber bundles vertical to the tensile axis, and eventually is deflected and hindered by the high strength longitudinal 90◦ fibers. mainly deflect longitudinally or arrested beneath the coatings to form a superficial oxidation governed by diffusion-controlled kinetics. 4. Conclusions During thermal cycling between 900 and 1200 ◦C and fatigue stress of 60 ± 20 MPa, the braided 3D C/SiC composites exhibit larger strain increment, better oxidation resistance and thermal shock resistance than the 2D composites. As indicated by the residual strength measurements, the thermally cycled braided 3D C/SiC composites suffer less loss in the strength than the 2D composites. Differences in the fiber architectures are taken into account to be response for the results with respect to the following: (1) Thermal cycling and fatigue stress resulted in the irregu￾lar netty cracks and regular wavy cracks on the top surface coatings of the 2D architecture and 3D braided architecture composites, respectively. Simultaneously, both side surface coatings always generated the highly oriented transverse cracks. Differences in the self-closure capacity of these cracks during heating lead to the different oxidation resis￾tance of the two architectures. (2) The constraints between longitudinal (90◦) and transverse (0◦) fiber bundles in 2D architecture are likely to enhance physical destruction during thermal cycling, through which the fiber oxidation can be aggravated. In contrast, the 3D braided architecture exhibits excellent thermal shock resistance by the collective and highly oriented transverse cracking. (3) Only one half of the fibers in the 2D architecture can effec￾tively hinder the transverse crack propagation from the surface to the center whereas almost all the fibers in the 3D braided architecture can deflect longitudinally the trans￾verse cracks, which result in the uniform oxidation governed by a linear reaction-controlled kinetics mechanism for the

H Mei et al. Materials Science and Engineering A 460-461(2007)306-313 former and the superficial oxidation governed by a parabolic F Christin, in: w. Krenkel, R. Naslain, H. Schneider(Eds ) High Temper- diffusion-controlled kinetics mechanism for the latter ature Ceramic Matrix Composites, vol 4, WILEY-VCH Press, Weinheim, 2001,pp.731-743. Acknowledgements [5] SSchmidt, S. Beyer, H. Knabe, H. Immich, R. Meistring, A. Gessler, Acta [6] J.E. Webb, R.N. Singh, J Am Ceram Soc. 79(11)(1996)2857. Financial support for this work was provided by the Nat- [7 R.N. Singh, H Wang Compos. Eng. 5(10-11)(1995)1287 ural Science Foundation of China( Contract No. 90405015) [8]X W Yin, L.F. Cheng. Carbon 40(2002)905 nd the National Young Elitists Foundation (Contract No. [9]K. Biemacki, W. Szyszkowski, S. Yannacopoulos, Compos. Part A 30 50425208). The authors also gratefully acknowledge the Pro- (1999)1027 ram for Changjiang Scholars and Innovative Research Team [10] H. Mei, L F Cheng, L.T. Zhang, X.G. Luan, J. Zhang, Carbon 44(2006) university(PCSIrT) [11] L.F. Cheng, Y D. Xu, L.T. Zhang, Q Zhang, Carbon 411(2002)1645 [12] M.S. Aly-Hassan, H. Hatta, S. Wakayama, M. Watanabe, K. Miyagawa, References Carbon41(2003)1069. [13]H. Mei, L.F. Cheng, L.T. Zhang, Y.D. Xu, Mater. Sci. Eng A430(2006) [1] F Lamouroux, X. Bourrat, J. Sevely, R. Naslain, Carbon 31(8)(1993) 314 [14] H. Mei, L F Cheng, L.T. Zhang, Scr. Mater. 54(2006)163 [2] J.C. Cavalier, A. Lacombe, J.M. Rouges, in: A.R. Bunsell, P. Lamicq, A. 1151 S Zhua, M. Mizunob, Y. Kagawac, Y. Mutoh, Compos. Sci. TechnoL. 59 Massiah(Eds ) Developments in the Science and Technology of Composite (1999)833 Materials, Elsevier, London, UK, 1989, Pp. 99-110. [16] A.J. Eckel, J D. Cawley, T.A. Parthasarathy. J Am Ceram Soc. 78(4) [3] K.A. Appiah, ZL Wang, w. Lackey, Carbon 38(2000)83

H. Mei et al. / Materials Science and Engineering A 460–461 (2007) 306–313 313 former and the superficial oxidation governed by a parabolic diffusion-controlled kinetics mechanism for the latter. Acknowledgements Financial support for this work was provided by the Nat￾ural Science Foundation of China (Contract No. 90405015) and the National Young Elitists Foundation (Contract No. 50425208). The authors also gratefully acknowledge the Pro￾gram for Changjiang Scholars and Innovative Research Team in university (PCSIRT). References [1] F. Lamouroux, X. Bourrat, J. Sevely, R. Naslain, Carbon 31 (8) (1993) 1273. [2] J.C. Cavalier, A. Lacombe, J.M. Rouges, in: A.R. Bunsell, P. Lamicq, A. Massiah (Eds.), Developments in the Science and Technology of Composite Materials, Elsevier, London, UK, 1989, pp. 99–110. [3] K.A. Appiah, Z.L. Wang, W.J. Lackey, Carbon 38 (2000) 831. [4] F. Christin, in: W. Krenkel, R. Naslain, H. Schneider (Eds.), High Temper￾ature Ceramic Matrix Composites, vol. 4, WILEY-VCH Press, Weinheim, 2001, pp. 731–743. [5] S. Schmidt, S. Beyer, H. Knabe, H. Immich, R. Meistring, A. Gessler, Acta Astronautica 55 (2004) 409. [6] J.E. Webb, R.N. Singh, J. Am. Ceram. Soc. 79 (11) (1996) 2857. [7] R.N. Singh, H. Wang, Compos. Eng. 5 (10–11) (1995) 1287. [8] X.W. Yin, L.F. Cheng, Carbon 40 (2002) 905. [9] K. Biernacki, W. Szyszkowski, S. Yannacopoulos, Compos. Part A 30 (1999) 1027. [10] H. Mei, L.F. Cheng, L.T. Zhang, X.G. Luan, J. Zhang, Carbon 44 (2006) 121. [11] L.F. Cheng, Y.D. Xu, L.T. Zhang, Q. Zhang, Carbon 411 (2002) 1645. [12] M.S. Aly-Hassan, H. Hatta, S. Wakayama, M. Watanabe, K. Miyagawa, Carbon 41 (2003) 1069. [13] H. Mei, L.F. Cheng, L.T. Zhang, Y.D. Xu, Mater. Sci. Eng. A430 (2006) 314. [14] H. Mei, L.F. Cheng, L.T. Zhang, Scr. Mater. 54 (2006) 163. [15] S. Zhua, M. Mizunob, Y. Kagawac, Y. Mutoh, Compos. Sci. Technol. 59 (1999) 833. [16] A.J. Eckel, J.D. Cawley, T.A. Parthasarathy, J. Am. Ceram. Soc. 78 (4) (1995) 972