JAm. Ceram.Soc.83[8l1999-2005(2000 urna Stress-Corrosion Cracking of Silicon Carbide Fiber/Silicon Carbide Composites Russell H. Jones. Charles H. Henager Jr * Charles A. Lewinsohn, and Charles F. Windisch Jr Pacific Northwest National Laboratory, Richland, Washington 99352 Ceramic-matrix composites are being developed to operate at specimens are tested at room temperature after high-temperature elevated temperatures and in oxidizing environments. Consid- exposure without the application of stress. /This environmental erable improvements have been made in the creep resistance of process results in bulk embrittlement and loss of fracture strength SiC fibers and, hence in the high-temperature properties of A time-dependent, environmentally induced crack growth pro- SiC fiber/SiC (SiC SiC) composites; however, more must be cess that results from removal of the fiber-matrix interphase has known about the stability of these materials in oxidizing been demonstrated and modeled by Henager and Jones, Jones et environments before they are widely accepted. Experimental al, and Windisch et al. o This process shares many features with weight change and crack growth data support the conclusion the OEM, except that crack growth is controlled by the time- that the oxygen-enhanced crack growth of SiC/Sic occurs by dependent reduction of the fiber-bridging stress instead of the ore than one mechanism, depending on the experimental formation of a brittle glass layer. An interphase removal mecha- conditions. These data suggest an oxidation embrittlement nism(IRM) results in a K and in stage Il growth, but the mechanism(OEM) at temperatures dd) and the temperature characteristics. Brief descriptions of the important features of each below a critical value(T TR for d> d. This paper summarizes the can result in both removal of the interphase and formation of Sio vidence for the existence of these two mechanisms and from reaction with the fiber or the matrix. For composites attempts to define the conditions for their operation. containing BN interphases, oxidation also can result in removal of the interphase and formation of a borosilicate-glass phase. a key feature of the oxidation behavior observed by Windisch et al. for L. Introduction SiC/SiC with a I um thick interphase, in oxygen at 2.4 X 10Pa NVIRONMENTALLY induced crack growth of ceramic-matrix of pressure(atmospheric pressure of oxygen)and 1373 K, for nay result from several mechanisms. An oxidation mes up to 10- s, was weight loss alone. Little or no SiO2 embrittlement mechanism(OEM), as proposed by Evans et al l formation occurred in any of the materials with carbon interphases, and observed by Heredia et al.,Lin and Becher, and Raghuraman although some boron-containing glass phase was observed for the et al, results from the reaction of the environment with the fiber material with a BN interphase or fiber-matrix interphase to cause local embrittlement. This The thermogravimetric analysis (TGA)results for the kinetics mechanism requires porosity or microcracks produced by applica- of mass loss are shown as a function of pressure and temperature tion of a stress before or during exposure to the environment in Figs. I and 2 for material with a carbon interphase. Complete allow ingress of the environment and formation of a brittle glass small test samples at a pressure of 2.4 X 10" Pa and a temperature stress-intensity threshold, Kth, below which crack growth does not of 1373 K. Clearly, the reaction rate increased with increasing occur, and of two growth stages, consistent with the stress- temperature. An activation energy of -50 kJ/mol was reported by Windisch et al, which could be explained as diffusion con- K-dependent regime for K values greater than the threshold value, nterphase recession rate(RR)was determined from the weight. trolled, through a boundary layer, or reaction-rate controlled. An stage II is the K-independent regime following stage I. In porous loss measurements and by optical microscopy. Both methods gave vapor infiltration, this embrittlement has been observed when recession-rate equations very similar to the physically measured log(rr)=0.9 log(Po)-99 (1) F. Zok-contributing editor where RR is the interphase recession rate, in m pascals. Greater recession rates were measured by by optical microscopy, which could suggest weig Manuscript No. 190054. Received July 2, 1998; approved January 18, 2000 bers, as well as from oxidation of the carbon interphase material Member, American Ceramic Society However, the forms derived for the equation by both methods wer 1999
Stress-Corrosion Cracking of Silicon Carbide Fiber/Silicon Carbide Composites Russell H. Jones,* Charles H. Henager Jr.,* Charles A. Lewinsohn,* and Charles F. Windisch Jr. Pacific Northwest National Laboratory, Richland, Washington 99352 Ceramic-matrix composites are being developed to operate at elevated temperatures and in oxidizing environments. Considerable improvements have been made in the creep resistance of SiC fibers and, hence, in the high-temperature properties of SiC fiber/SiC (SiCf /SiC) composites; however, more must be known about the stability of these materials in oxidizing environments before they are widely accepted. Experimental weight change and crack growth data support the conclusion that the oxygen-enhanced crack growth of SiCf /SiC occurs by more than one mechanism, depending on the experimental conditions. These data suggest an oxidation embrittlement mechanism (OEM) at temperatures dc) and the temperature below a critical value (T Tg for d > dc. This paper summarizes the evidence for the existence of these two mechanisms and attempts to define the conditions for their operation. I. Introduction ENVIRONMENTALLY induced crack growth of ceramic-matrix composites may result from several mechanisms. An oxidation embrittlement mechanism (OEM), as proposed by Evans et al. 1 and observed by Heredia et al.,2 Lin and Becher,3 and Raghuraman et al.,4 results from the reaction of the environment with the fiber or fiber–matrix interphase to cause local embrittlement. This mechanism requires porosity or microcracks produced by application of a stress before or during exposure to the environment to allow ingress of the environment and formation of a brittle glass layer.5 Modeling of this mechanism suggests the presence of a stress-intensity threshold, Kth, below which crack growth does not occur, and of two growth stages, consistent with the stresscorrosion cracking of ceramics.1 Stage I is the strongly K-dependent regime for K values greater than the threshold value; stage II is the K-independent regime following stage I. In porous materials, such as SiC fiber/SiC (SiCf /SiC), produced by chemical vapor infiltration, this embrittlement has been observed when specimens are tested at room temperature after high-temperature exposure without the application of stress.6,7 This environmental process results in bulk embrittlement and loss of fracture strength. A time-dependent, environmentally induced crack growth process that results from removal of the fiber–matrix interphase has been demonstrated and modeled by Henager and Jones,8 Jones et al.,9 and Windisch et al. 10 This process shares many features with the OEM, except that crack growth is controlled by the timedependent reduction of the fiber-bridging stress instead of the formation of a brittle glass layer. An interphase removal mechanism (IRM) results in a Kth and in stage II growth, but the characteristics of stage II for the IRM differ from those for the OEM. Local embrittlement does not necessarily occur and, in fact, the composite strength typically is unaffected by cracking, as demonstrated by Jones et al. 11 The purpose of this paper is to summarize the evidence for the existence of the OEM and the IRM in SiCf /SiC and to define their characteristics. Brief descriptions of the important features of each mechanism are provided. Key microstructural parameters controlling each mechanism are emphasized. II. Oxidation of SiCf /SiC Composites Oxidation of an SiCf /SiC composite with a carbon interphase can result in both removal of the interphase and formation of SiO2 from reaction with the fiber or the matrix. For composites containing BN interphases, oxidation also can result in removal of the interphase and formation of a borosilicate-glass phase. A key feature of the oxidation behavior observed by Windisch et al. 10 for SiCf /SiC with a 1 mm thick interphase, in oxygen at 2.4 3 104 Pa of pressure (atmospheric pressure of oxygen) and 1373 K, for times up to 104 s, was weight loss alone. Little or no SiO2 formation occurred in any of the materials with carbon interphases, although some boron-containing glass phase was observed for the material with a BN interphase. The thermogravimetric analysis (TGA) results for the kinetics of mass loss are shown as a function of pressure and temperature in Figs. 1 and 2 for material with a carbon interphase. Complete burnout of the carbon interphase occurred within ,104 s in the small test samples at a pressure of 2.4 3 104 Pa and a temperature of 1373 K. Clearly, the reaction rate increased with increasing temperature. An activation energy of ;50 kJ/mol was reported by Windisch et al.,10 which could be explained as diffusion controlled, through a boundary layer, or reaction-rate controlled. An interphase recession rate (RR) was determined from the weightloss measurements and by optical microscopy. Both methods gave recession-rate equations very similar to the physically measured equation, as follows: log ~RR! 5 0.9 log ~ pO2! 2 9.9 (1) where RR is the interphase recession rate, in m/s, and pO2 is in pascals. Greater recession rates were measured by weight loss than by optical microscopy, which could suggest weight loss from the fibers, as well as from oxidation of the carbon interphase material. However, the forms derived for the equation by both methods were F. Zok—contributing editor Manuscript No. 190054. Received July 2, 1998; approved January 18, 2000. *Member, American Ceramic Society. J. Am. Ceram. Soc., 83 [8] 1999–2005 (2000) 1999 journal
Journal of the american Ce via chemical vapor deposition, onto a pitch-based carbon-fiber Oxidation of SiC!Sic composites with carbon interphases can sult in the formation of SiO2 temperature. Tortorelli and More observed an initial weight loss, followed by a weight gain, in a Nicalon- fiber-reinforced Sic sic composite with a 0.3 um thick carbon layer exposed to dry air (Po, =2X 10 Pa)at 1223 K. Following the initial weight loss from oxidation of the carbon, they found that Sio, formation ccurred within the interfacial region previously occupied by the carbon. For the Nicalon-reinforced composite material, complete 0E+001E+042E+04 carbon depletion occurred within 900 s at 1223 K, followed by weight gain from SiO, formation. Unal et al. observed their largest Time, s weight loss(5%)at 1223 K, for an exposure of 1.8 X 10 s in dry xygen, and a decreasing weight loss with increasing temperature Fig.1.TGA mass loss, as a function of time, for Sic/sic with carbon up to 1673 K (2%). Their material was a Nicalon-reinforced interfaces exposed to various Po, values at 1373 K(O)2.4 x 10, (o) SiC/SiC with a 0.5 um thick fiber-matrix carbon interphase Kleykamp et al. observed the following reactions of air with Sic-fiber-reinforced SiC composites: (1)oxidation of free carbon at 800-965K, (2)a rapid exothermic reaction and weight gain 0.20 beginning at 1073 K and continuing up to 1773 K, and for times up to 3.6X 10 s, followed by (3)the diffusion-controlled 0.15 oxidation of bulk SiC. Sebire-Lhermitte et al. 7 identified the presence and location of SiO, formation in SiC/SiC composites using transmission electron microscopy(TEM). They noted the g0.10 presence of 15 nm thick SiO, layers at both the fiber-carbon and the matrix-carbon interfaces following exposure for 3.6 X 10sat 1123 K in air That Windisch et al. o observed a weight loss alone. whereas others observed a weight gain following the weight loss, could be the result of lower oxygen pressures, shorter exposure time, and, perhaps, greater carbon-layer thickness. The lower oxygen IE+042E+04 pressure would decrease the SiO, formation rate and, therefore, the chance for a measurable weight gain during the 1. 8X 10-s Time. s xposures. Tortorelli et al.5,6 used exposures of up to 5.4X 105 s ig. 2. TGA mass loss, as a function of time for SiC/SiC with carbon but observed weight loss alone for times <1. X 10 s, with temperatures((D)1373, (O)1273, (O)1173, and(4)l0? e at various exposure at 1223 K, at Po2=2 X 10 Pa, for a composite with a interfaces exposed to 2.5 x 10 Pa of oxygen press 0.3 um thick carbon interphase. Unal et al. used 1.8X 105 s exposures. Measurable SiO, formation at Po. <2 X 10 Pa would require a much greater exposure time than that used by Windisch identical, suggesting that the same chemical reaction was control- et al. and even greater than the time used by Tortorelli and More ng. Conceivably, some glass formation occurred during these The existence of subcritical crack growth, as described in the next measurements, although for the weight-loss measurements to giv section, which coincides with weight loss alone or interphase greater recession rates than those derived by optical microscop removal without the embrittling effect of SiO, or other solid- would have required that the weight gain from glass formation roduct formation, is the primary difference between the IRM and I. Weight loss alone also was observed over the temperature the IRM only. 1073-1373 K, which borders on the temperature range 873-1073 K suggested by Evans et al. for the OEM The material studied by windisch et al. u was reinforced with Ill. Interphase Removal Mechanism ceramic-grade Nicalon fibers, coated with a 1.0 um thick carbon nterphase. Lewinsohn et al. measured the rate of interphase 1) Subcritical Crack Growth behavior oxidation for up to 7.2 X 10 s, at 1073 K in air, for effect of oxygen on the subcritical crack growth velocity of reinforced by Hi-Nicalon' fibers and with a I um thick SiC/ SiC is clearly demonstrated by the data given in Fig. 3 The interphase recession distance increased linearly Oxygen has little effect on the midpoint displacement (i.e, crack velocity) for -2 X 10- s, but a marked increase in the crack under these conditions. Based on Eq. (1) by Windisch et al. the velocity is noted for longer times. These tests were performed in carbon interphase RR was predicted to be 0. 28 Hum/s at 1073 K in air. the O2 pressure, temperature, and time regime where weight loss The experimental results agree alone was observed during oxidation studies. Therefore, the recession rate, considering the po of slight differences in embrittling effect of a solid reaction product should not have been the carbon interphase materials car y differences in process- ing. Furthermore, the ion rate and linear time should have contributed to the crack growth rate. However, even if nce also are lent with the values measured b SiO or other glassy phases had been present, they would have had Eckel et al. for the of the carbon core of an SCs-6 low viscosity at this high temperature and would not likely have fiber (SCS-6 fibers ar ated by depositing silicon carbide, affected the crack growth behavior or caused brittle fracture The dependence of the crack velocity on oxygen partial pressure up to Po, =2 X 10 Pa is given in Fig. 4 for tests at 1373K.A sharp increase in the crack velocity occurred at low pressure and a ippon Carbon Co., Tokyo, Japan. slower increase at pressures of -0.25 x 10- to 2.5X 10- Pa
identical, suggesting that the same chemical reaction was controlling. Conceivably, some glass formation occurred during these measurements, although for the weight-loss measurements to give greater recession rates than those derived by optical microscopy would have required that the weight gain from glass formation be compensated for by weight loss from the fibers. The thermodynamic and kinetic results of the study by Windisch et al. are in agreement with those reported by Filipuzzi et al. 12,13 and Eckel et al.14 Weight loss alone also was observed over the temperature range 1073–1373 K, which borders on the temperature range 873–1073 K suggested by Evans et al. 1 for the OEM. The material studied by Windisch et al. 10 was reinforced with ceramic-grade Nicalon† fibers, coated with a 1.0 mm thick carbon interphase. Lewinsohn et al. 15 measured the rate of interphase oxidation for up to 7.2 3 103 s, at 1073 K in air, for materials reinforced by Hi-Nicalon† fibers and with a 1 mm thick interphase. The interphase recession distance increased linearly with time, at a rate of 0.19 mm/s. There was no evidence of SiO2 formation under these conditions. Based on Eq. (1) by Windisch et al., the carbon interphase RR was predicted to be 0.28 mm/s at 1073 K in air. The experimental results agree quite well with the predicted recession rate, considering the possibility of slight differences in the carbon interphase materials caused by differences in processing. Furthermore, the measured recession rate and linear time dependence also are in agreement with the values measured by Eckel et al. 14 for the recession of the carbon core of an SCS-6 fiber. (SCS-6 fibers are fabricated by depositing silicon carbide, via chemical vapor deposition, onto a pitch-based carbon-fiber core.) Oxidation of SiCf /SiC composites with carbon interphases can also result in the formation of SiO2 and a weight gain, following an initial weight loss,5,6 or a decreased weight loss, with increasing temperature.7 Tortorelli and More5 observed an initial weight loss, followed by a weight gain, in a Nicalon-fiber-reinforced SiCf /SiC composite with a 0.3 mm thick carbon layer exposed to dry air (pO2 5 2 3 104 Pa) at 1223 K. Following the initial weight loss from oxidation of the carbon, they found that SiO2 formation occurred within the interfacial region previously occupied by the carbon. For the Nicalon-reinforced composite material, complete carbon depletion occurred within 900 s at 1223 K, followed by weight gain from SiO2 formation. Unal et al. 7 observed their largest weight loss (5%) at 1223 K, for an exposure of 1.8 3 105 s in dry oxygen, and a decreasing weight loss with increasing temperature, up to 1673 K (2%). Their material was a Nicalon-reinforced SiCf /SiC with a 0.5 mm thick fiber–matrix carbon interphase. Kleykamp et al. 16 observed the following reactions of air with SiC-fiber-reinforced SiC composites: (1) oxidation of free carbon at 800–965 K, (2) a rapid exothermic reaction and weight gain, beginning at 1073 K and continuing up to 1773 K, and for times up to 3.6 3 103 s, followed by (3) the diffusion-controlled oxidation of bulk SiC. Sebire-Lhermitte et al. 17 identified the presence and location of SiO2 formation in SiCf /SiC composites using transmission electron microscopy (TEM). They noted the presence of 15 nm thick SiO2 layers at both the fiber–carbon and the matrix–carbon interfaces following exposure for 3.6 3 103 s at 1123 K in air. That Windisch et al. 10 observed a weight loss alone, whereas others5–7 observed a weight gain following the weight loss, could be the result of lower oxygen pressures, shorter exposure time, and, perhaps, greater carbon-layer thickness. The lower oxygen pressure would decrease the SiO2 formation rate and, therefore, the chance for a measurable weight gain during the 1.8 3 104 s exposures. Tortorelli et al. 5,6 used exposures of up to 5.4 3 105 s but observed weight loss alone for times ,1.4 3 104 s, with exposure at 1223 K, at pO2 5 2 3 104 Pa, for a composite with a 0.3 mm thick carbon interphase. Unal et al. 7 used 1.8 3 105 s exposures. Measurable SiO2 formation at pO2 , 2 3 103 Pa would require a much greater exposure time than that used by Windisch et al. and even greater than the time used by Tortorelli and More. The existence of subcritical crack growth, as described in the next section, which coincides with weight loss alone or interphase removal without the embrittling effect of SiO2 or other solidproduct formation, is the primary difference between the IRM and the OEM. The oxidation results of Windisch et al. demonstrate that the results of Henager and Jones8 and Jones et al. 9 at temperatures ranging from 1073 to 1473 K and pO2 , 2 3 103 Pa occurred by the IRM only. III. Interphase Removal Mechanism (1) Subcritical Crack Growth Behavior The effect of oxygen on the subcritical crack growth velocity of SiCf /SiC is clearly demonstrated by the data given in Fig. 3. Oxygen has little effect on the midpoint displacement (i.e., crack velocity) for ;2 3 104 s, but a marked increase in the crack velocity is noted for longer times. These tests were performed in the O2 pressure, temperature, and time regime where weight loss alone was observed during oxidation studies.10 Therefore, the embrittling effect of a solid reaction product should not have been a factor; only the effect of fiber creep and interfacial removal should have contributed to the crack growth rate. However, even if SiO2 or other glassy phases had been present, they would have had low viscosity at this high temperature and would not likely have affected the crack growth behavior or caused brittle fracture. The dependence of the crack velocity on oxygen partial pressure up to pO2 5 2 3 103 Pa is given in Fig. 4 for tests at 1373 K. A sharp increase in the crack velocity occurred at low pressure and a slower increase at pressures of ;0.25 3 102 to 2.5 3 103 Pa. † Nippon Carbon Co., Tokyo, Japan. Fig. 1. TGA mass loss, as a function of time, for SiC/SiC with carbon interfaces exposed to various pO2 values at 1373 K ((M) 2.4 3 104 , (e) 2.5 3 103 , (E) 1.2 3 103 , (‚) 6.3 3 102 and (ƒ) 3.1 3 102 Pa). Fig. 2. TGA mass loss, as a function of time, for SiC/SiC with carbon interfaces exposed to 2.5 3 103 Pa of oxygen pressure at various temperatures ((M) 1373, (e) 1273, (E) 1173, and (‚) 1073 K). 2000 Journal of the American Ceramic Society—Jones et al. Vol. 83, No. 8
august 2000 2001 (2) Subcritical Crack Growth Mechanism argon The IRM, as proposed by Henager and Jones, is based 000 ppm o primarily on the creep of bridging fibers and the effect of the nterphase removal rate on fiber relaxation. Available creep data for Nicalon fibers at 1 C in pure argon were used to construct a constitutive equation for the stress dependence and time 0.15 dence of creep in Nicalon fibers at this temperature. Based on this approach, the discrete micromechanics model was used to calculate the crack velocity, assuming a quasi-static approximation. The process envisioned for decoupling the bridging fibers from the matrix is shown schematically in Fig. 6. In dependence of the fiber-bridging stress can be related to the rate of 1101.510 interphase removal by oxidation, o through the implied time dependence of the debonded region, A, and the fiber-matrix shear gerin ests omen. nicalonreisforce cement during subcritical crack growth trength. t. The hase removal process(oxidation process ed composites at 1473 K in argon and occurs at all fibers that intersect the crack and at the composite argon/oxygen mixture urface. Fiber-interphase material ahead of the crack is not subje to oxidation until mechanical debonding between the fiber and an oxygen transpo 6 a definition for the stress intensity for an equilibrium-bridging 1° zone was used to derive an expression for the velocity of a crack in a composite at equilibriun lting in a quasi-static approxi- mation to the crack velocity. For this approximation, the bridging zone was assumed to be in equilibrium, by virtue of a balance 0° between crack advance and relaxation of bridging-zone stresses Minimum CG.BN s the crack advanced. it would bridge additional fibers whie would retard its growth. As the stresses in the bridging zone ning would be decreased, and the would advance C-interface minimum in A Decreasing the crack-closure(fiber-bridging) forces as a func- tion of time. because of either stress relaxation in the fibers or BN-interface minimun in Ar removal of the interface, would allow the crack to extend during 0.00.51.0 recession would control removal of the fiber-bridging stresses Oxygen Partial Pressure (10 Pa) because the activation energy for interphase oxidation is much Fig. 4. Minimum or limiting crack velocity for CG-C and CG-BN lower than for fiber creep. The interphase-recession-induced de materials, as a function of Po, at 1373 K. crease of fiber-bridging stresses would be more rapid than for fiber creep alone, because oxidation simultaneously removes the fiber matrix interface. This process would decrease the fiber-matrix Material with a BN interface exhibited slower crack velocity, by a interfacial shear strength, as a function of time. A more rapid rate factor of -5-7. Some glass-phase formation was noted in this of fiber-stress relaxation would shift the onset of accelerated aterial, as a result of these exposures, but there was no evidence cracking(stage Ill) to lower Ka values and increase the relative crack velocities in the stage II region. A similar increase in effective crack velocity with increasing Po, was observed at 1073 K, as shown in Fig The time dependence of the midpoint displacement was characteristic of IV. Oxidation Embrittlement Mechanism IRM-type crack growth for all of these experiments. Examination (1) Subcritical Crack Growth Behavior of the fracture surfaces of these specimens also revealed fiber A dynamic(sample stressed during exposure)OEM has been pullout, although limited, in agreement with the IRM mechanism. observed-320 in SiC/SiC at 1073-1223 K in air(Po,=2 X 10 The time at which the effective crack velocity rapidly increased diminished with increasing oxygen concentration. The effective Pa), whereas a static(sample unstressed during exposure) OEM crack velocity was less sensitive to the oxygen concentration. Slip effective 175000 Debonded velocity Fiber/ Matrix 010 110 2104310441045104 Fig. 6. Schematic diagram of fiber-debond model, showing A, T, k, an gs. where x is the debond length. t the interfacial Fig. 5. Crack velocity at 800C versus time and oxygen con crack opening, and a the acture strength argon for SiC/SiC reinforced with cel grade nicalon an along the fiber-matrix inte
Material with a BN interface exhibited slower crack velocity, by a factor of ;5–7. Some glass-phase formation was noted in this material, as a result of these exposures, but there was no evidence that it induced the OEM. A similar increase in effective crack velocity with increasing pO2 was observed at 1073 K, as shown in Fig. 5. The time dependence of the midpoint displacement was characteristic of IRM-type crack growth for all of these experiments. Examination of the fracture surfaces of these specimens also revealed fiber pullout, although limited, in agreement with the IRM mechanism. The time at which the effective crack velocity rapidly increased diminished with increasing oxygen concentration. The effective crack velocity was less sensitive to the oxygen concentration. (2) Subcritical Crack Growth Mechanism The IRM, as proposed by Henager and Jones,8 is based primarily on the creep of bridging fibers and the effect of the interphase removal rate on fiber relaxation. Available creep data for Nicalon fibers at 1100°C in pure argon18 were used to construct a constitutive equation for the stress dependence and time dependence of creep in Nicalon fibers at this temperature.19 Based on this approach, the discrete micromechanics model was used to calculate the crack velocity, assuming a quasi-static approximation. The process envisioned for decoupling the bridging fibers from the matrix is shown schematically in Fig. 6. In principle, the time dependence of the fiber-bridging stress can be related to the rate of interphase removal by oxidation,10 through the implied time dependence of the debonded region, l, and the fiber–matrix shear strength, t. The interphase removal process (oxidation process) occurs at all fibers that intersect the crack and at the composite surface. Fiber-interphase material ahead of the crack is not subject to oxidation until mechanical debonding between the fiber and matrix provides an oxygen transport path. A definition for the stress intensity for an equilibrium-bridging zone was used to derive an expression for the velocity of a crack in a composite at equilibrium, resulting in a quasi-static approximation to the crack velocity. For this approximation, the bridging zone was assumed to be in equilibrium, by virtue of a balance between crack advance and relaxation of bridging-zone stresses. As the crack advanced, it would bridge additional fibers, which would retard its growth. As the stresses in the bridging zone relaxed, the crack-tip screening would be decreased, and the crack would advance. Decreasing the crack-closure (fiber-bridging) forces as a function of time, because of either stress relaxation in the fibers or removal of the interface, would allow the crack to extend during the load step. In oxygen-containing environments, interphase recession would control removal of the fiber-bridging stresses, because the activation energy for interphase oxidation is much lower than for fiber creep. The interphase-recession-induced decrease of fiber-bridging stresses would be more rapid than for fiber creep alone, because oxidation simultaneously removes the fiber– matrix interface. This process would decrease the fiber–matrix interfacial shear strength, as a function of time. A more rapid rate of fiber-stress relaxation would shift the onset of accelerated cracking (stage III) to lower KA values and increase the relative crack velocities in the stage II region. IV. Oxidation Embrittlement Mechanism (1) Subcritical Crack Growth Behavior A dynamic (sample stressed during exposure) OEM has been observed1–3,20 in SiCf /SiC at 1073–1223 K in air (pO2 5 2 3 104 Pa), whereas a static (sample unstressed during exposure) OEM Fig. 3. Specimen midpoint displacement during subcritical crack growth experiments on Hi-Nicalon-reinforced composites at 1473 K in argon and argon/oxygen mixture. Fig. 4. Minimum or limiting crack velocity for CG-C and CG-BN materials, as a function of pO2 at 1373 K. Fig. 5. Crack velocity at 800°C versus time and oxygen concentration in argon for SiC/SiC reinforced with ceramic-grade Nicalon and with a 150 nm carbon interphase. Fig. 6. Schematic diagram of fiber-debond model, showing l, t, m, and sf , where l is the debond length, t the interfacial shear strength, m the crack opening, and sf the fiber fracture strength. Oxygen ingress occurs along the fiber–matrix interphase and increases l; thus l and t become time-dependent. August 2000 Stress-Corrosion Cracking of Silicon Carbide Fiber/Silicon Carbide Composites 2001
2002 Journal of the was observed in room-temperature tests following elevated- exponent is very close to r.Hence, the flaw size in the oxide temperature exposures at 1223 K in air,1273 K in air, and layer appears to scale directly with its thickness, in agreement with to 1673 K in dry oxygen at a pressure of 10 Pa. Heredia et al the assumption made by Lara-Curzio reported an upper temperature of 1073 K for dynamic OEM in air; If the methods proposed by Evans and Zok2 and Singh et al. 8 Lin and Becher and Raghuraman et al.observed dynamic OEM are used to estimate the preexisting flaw size for the Hi-Nicalon in air at 1223 K. Tortorelli et al. and Unal et al. found the OEM fibers studied by Takeda et al., 4 one obtains flaw sizes in the operate in room-temperature tests following elevated- range 10-40 nm for fiber fracture-toughness values of 0.5-1.0 temperature exposure tests conducted in air without applied stress. MPam". Helmer et al. investigated the influence of the In summary, the upper temperature limit reported for a dynamic thickness of a pyrolytic carbon coating on polyacrylonitrile oEM is between 1073 and 1223 K whereas the formation of a derived carbon fibers. They observed that the strength value for the glass phase at temperatures >1223 K can cause a static OEM fibers decreased for coatings thicker than 17 nm. This coating when specimens are tested at lower temperatures. Because the thickness corresponded to the observed pore size in the fibers results from the formation of a brittle glass phase, this 1-20 nm. measured via mechanism must depend on the growth rate and viscosity of the The results of Helmer et al. 29 suggest that the fiber strength is glass phase. The growth rate increases with increasing temperature not decreased if the thickness of the fiber coating is less than the nd Po, but the viscosity decreases with increasing temperature trinsic flaw size. In fact, if fiber failure is controlled by surface Therefore, a temperature most likely exists at which the effective- ness of the oEM is maximum lead to an observed increase in fiber strength. thus the critical oxide-layer thickness, de, above which the OEM occurs is con- trolled by the preexisting fiber flaw size. An upper limit on the (2) Subcritical Crack Growth Mechanism critical oxide-layer thickness would be the fiber preexisting flaw the fiber, degrading the fiber strength. A glass-layer thickness >d oxide-layer thickness would scale by the ratio of the oxide-layer produces a surface flaw of sufficient size to decrease the fracture toughness to that of the fiber stress of the fiber. This result presumes that the glass layer cracks This progression of fiber fracture and the growth rate of SiO, at a lower stress than does the fiber. fibers farthest from the crack accounts for time-dependent crack growth in the OEM. Reaction- tip experience the greatest decrease in fracture stress, because they controlled stage I and diffusion-controlled stage Il regimes are have been exposed the longest time to the environment and have proposed by Evans et al., with stage I resulting when crack the thickest oxide layer (i.e, the largest fiber-surface crack). propagation is slow, such that the oxygen concentration is constant Fracture of the outer fibers shifts the load to other fibers and to the in the crack, and stage Il resulting when crack propagation crack tip, causing further crack extension and exposure of addi- sufficiently rapid that the oxygen concentration decreases from the tional fibers crack mouth to the tip. The oxygen concentration gradient results Lara-Curzio analyzed the lifetime of specimens undergoing in a variable SiO, growth rate along the length of the crack. The the OEM, assuming that the fiber flaw size was controlled by the crack velocity is proportional to k in stage I and independent of xide-layer thickness and that the oxide-layer thickness was K, but with a plateau proportional to o, in stage Il. As mentioned proportional to the square root of time(as for processes controlled earlier, the temperature dependence results from the effect on the by atomic diffusion). According to these assumptions, the strength SiO, growth rate, but an upper temperature is expected from the of the fibers, at a fixed temperature and oxygen concentration, transition of the glass from brittle to viscous behavior. hanges with time in proportion to r 4. The time-to-failure of a Although Evans et al. modeled the OEm as resulting from the CVI SiC/SiC composite predicted by this model is in good formation of a brittle glass layer that forms on the bridging fibers, agreement with that measured experimentally and consistent with several other OEMs could be operative. These include (1)the the OEM model and results described in this paper. Zhu et al formation of a reaction product within the fiber-matrix interface showed that the dependency of the oxide-layer thickness on time is that increases the interfacial bond strength and decreases the fiber alid for a limited time, beyond which the decrease in reaction area debonding and pullout, as demonstrated by glime and Cawley, of oxidized cylindrical fibers causes the oxide-layer thickness to or(2)a decrease in the fiber strength, as a result of reaction with become linearly dependent on time. Therefore, Lara-Curzio's the environment, as noted by Gogotsi and Yoshimura. o The first analysis may not be accurate for long times mechanism would have a similar dependence on the glass viscosity Numerous observations have demonstrated that a glass layer on to that of the Evans et al. model, in that the interfacial bond the surface of Sic fibers decreases the fiber strength. Included in strength induced by a glass interphase would decrease with this list are reports by Takeda et al., 4 Glime and Cawley 25and increasing temperature. However, with the second mechanism, the Parthasarathy et al. Takeda et al. measured the strength, at fiber strength would continue to decrease with increasing tempe temperature, of Hi-Nicalon and Hi-Nicalon-type S fibers ature from either thermal or environmental effects after annealing in argon or air for 10 h at different temperatures The model of evans et al. considers a decrease in the fiber These authors also measured thicknesses, obtained by measure- fracture stress caused by a surface flaw without involving a loss in ments using scanning electron microscopy images, of oxide layers fiber strength. Okabe et al. observed evidence of the oEm in grown onto the fibers. This information made it possible to SiCSiC tested by three-point bending in air at 1473 K, which investigate the effect of the oxide layer on the fiber strength above the temperature suggested by others for the OEM Assuming a semi-elliptical surface flaw, the expected strength Strength data were not provided; however, flat fracture without value for fibers with the proposed fracture-toughness value and fiber pullout was noted. This result could indicate a decreased fiber flaw size corresponding to the thickness of the oxide layer can be strength from interaction with the environment, as suggested in predicted from the method proposed by Evans and Zok7 and mechanism( 2). However, further information is needed to identify Singh et al. Takeda ef al. measured oxide-layer thicknesses of whether only one OEM mechanism exists 0.979 and 1 229 um for Hi-Nicalon fibers oxidized in dry air fo 10 h at 1573 and 1673 K, respectively. The strength values calculated using these oxide-layer thicknesses were 315 and 281 Environmental Parameter Space for the IRM and OEM MPa. The strength values of fibers( tested at room temperature) oxidized in dry air for 10 h at 1573 and 1673 K were 1. 38 and 0.71 A summary of the key features of the OEM and IRM GPa, respectively. Clearly, the oxide layer does not act as a flaw stress-corrosion cracking of ceramic composites is given in Table in itself; rather, flaws in the oxide layer must control the fiber I. Only a few characteristics are similar these two strength. On the other hand, if the data of strength versus processes: (1)time-dependent behavior, (2)th ce of a kths oxide-layer thickness are fit by a power-law function, the resulting and (3)subcritical crack growth stages. The
was observed in room-temperature tests following elevatedtemperature exposures at 1223 K in air,6 1273 K in air,21 and up to 1673 K in dry oxygen at a pressure of 105 Pa.7 Heredia et al. 2 reported an upper temperature of 1073 K for dynamic OEM in air; Lin and Becher3 and Raghuraman et al. 4 observed dynamic OEM in air at 1223 K. Tortorelli et al. 6 and Unal et al. 7 found the OEM to operate in room-temperature tests following elevatedtemperature exposure tests conducted in air without applied stress. In summary, the upper temperature limit reported for a dynamic OEM is between 1073 and 1223 K, whereas the formation of a glass phase at temperatures .1223 K can cause a static OEM when specimens are tested at lower temperatures. Because the OEM results from the formation of a brittle glass phase, this mechanism must depend on the growth rate and viscosity of the glass phase. The growth rate increases with increasing temperature and pO2 , but the viscosity decreases with increasing temperature. Therefore, a temperature most likely exists at which the effectiveness of the OEM is maximum. (2) Subcritical Crack Growth Mechanism The OEM, as proposed by Evans et al.,1 results from the formation of a “brittle” glass layer that creates a surface crack on the fiber, degrading the fiber strength. A glass-layer thickness .dc produces a surface flaw of sufficient size to decrease the fracture stress of the fiber. This result presumes that the glass layer cracks at a lower stress than does the fiber. Fibers farthest from the crack tip experience the greatest decrease in fracture stress, because they have been exposed the longest time to the environment and have the thickest oxide layer (i.e., the largest fiber-surface crack). Fracture of the outer fibers shifts the load to other fibers and to the crack tip, causing further crack extension and exposure of additional fibers. Lara-Curzio22 analyzed the lifetime of specimens undergoing the OEM, assuming that the fiber flaw size was controlled by the oxide-layer thickness and that the oxide-layer thickness was proportional to the square root of time (as for processes controlled by atomic diffusion). According to these assumptions, the strength of the fibers, at a fixed temperature and oxygen concentration, changes with time in proportion to t 1/4. The time-to-failure of a CVI SiCf /SiC composite predicted by this model is in good agreement with that measured experimentally and consistent with the OEM model and results described in this paper. Zhu et al. 23 showed that the dependency of the oxide-layer thickness on time is valid for a limited time, beyond which the decrease in reaction area of oxidized cylindrical fibers causes the oxide-layer thickness to become linearly dependent on time. Therefore, Lara-Curzio’s analysis may not be accurate for long times. Numerous observations have demonstrated that a glass layer on the surface of SiC fibers decreases the fiber strength. Included in this list are reports by Takeda et al.,24 Glime and Cawley,25 and Parthasarathy et al. 26 Takeda et al. 24 measured the strength, at room temperature, of Hi-Nicalon and Hi-Nicalon-type S fibers after annealing in argon or air for 10 h at different temperatures. These authors also measured thicknesses, obtained by measurements using scanning electron microscopy images, of oxide layers grown onto the fibers. This information made it possible to investigate the effect of the oxide layer on the fiber strength. Assuming a semi-elliptical surface flaw, the expected strength value for fibers with the proposed fracture-toughness value and a flaw size corresponding to the thickness of the oxide layer can be predicted from the method proposed by Evans and Zok27 and Singh et al. 28 Takeda et al. measured oxide-layer thicknesses of 0.979 and 1.229 mm for Hi-Nicalon fibers oxidized in dry air for 10 h at 1573 and 1673 K, respectively. The strength values calculated using these oxide-layer thicknesses were 315 and 281 MPa. The strength values of fibers (tested at room temperature) oxidized in dry air for 10 h at 1573 and 1673 K were 1.38 and 0.71 GPa, respectively. Clearly, the oxide layer does not act as a flaw in itself; rather, flaws in the oxide layer must control the fiber strength. On the other hand, if the data of strength versus oxide-layer thickness are fit by a power-law function, the resulting exponent is very close to t 21/2. Hence, the flaw size in the oxide layer appears to scale directly with its thickness, in agreement with the assumption made by Lara-Curzio.22 If the methods proposed by Evans and Zok27 and Singh et al. 28 are used to estimate the preexisting flaw size for the Hi-Nicalon fibers studied by Takeda et al.,24 one obtains flaw sizes in the range 10–40 nm for fiber fracture-toughness values of 0.5–1.0 MPazm1/2. Helmer et al. 29 investigated the influence of the thickness of a pyrolytic carbon coating on polyacrylonitrilederived carbon fibers. They observed that the strength value for the fibers decreased for coatings thicker than 17 nm. This coating thickness corresponded to the observed pore size in the fibers, 1–20 nm, measured via TEM. The results of Helmer et al. 29 suggest that the fiber strength is not decreased if the thickness of the fiber coating is less than the intrinsic flaw size. In fact, if fiber failure is controlled by surface porosity, a thin fiber coating may blunt the preexisting flaws and lead to an observed increase in fiber strength. Thus, the critical oxide-layer thickness, dc, above which the OEM occurs is controlled by the preexisting fiber flaw size. An upper limit on the critical oxide-layer thickness would be the fiber preexisting flaw size. Because the fracture toughness of a brittle oxide layer probably is lower than that of the fiber, the actual critical oxide-layer thickness would scale by the ratio of the oxide-layer toughness to that of the fiber. This progression of fiber fracture and the growth rate of SiO2 accounts for time-dependent crack growth in the OEM. Reactioncontrolled stage I and diffusion-controlled stage II regimes are proposed by Evans et al.,1 with stage I resulting when crack propagation is slow, such that the oxygen concentration is constant in the crack, and stage II resulting when crack propagation is sufficiently rapid that the oxygen concentration decreases from the crack mouth to the tip. The oxygen concentration gradient results in a variable SiO2 growth rate along the length of the crack. The crack velocity is proportional to K13/3 in stage I and independent of K, but with a plateau proportional to s, 5 in stage II. As mentioned earlier, the temperature dependence results from the effect on the SiO2 growth rate, but an upper temperature is expected from the transition of the glass from brittle to viscous behavior. Although Evans et al. 1 modeled the OEM as resulting from the formation of a brittle glass layer that forms on the bridging fibers, several other OEMs could be operative. These include (1) the formation of a reaction product within the fiber–matrix interface that increases the interfacial bond strength and decreases the fiber debonding and pullout, as demonstrated by Glime and Cawley,25 or (2) a decrease in the fiber strength, as a result of reaction with the environment, as noted by Gogotsi and Yoshimura.30 The first mechanism would have a similar dependence on the glass viscosity to that of the Evans et al. model, in that the interfacial bond strength induced by a glass interphase would decrease with increasing temperature. However, with the second mechanism, the fiber strength would continue to decrease with increasing temperature from either thermal or environmental effects. The model of Evans et al. 1 considers a decrease in the fiber fracture stress caused by a surface flaw without involving a loss in fiber strength. Okabe et al. 31 observed evidence of the OEM in SiCf /SiC tested by three-point bending in air at 1473 K, which is above the temperature suggested by others1–3 for the OEM. Strength data were not provided; however, flat fracture without fiber pullout was noted. This result could indicate a decreased fiber strength from interaction with the environment, as suggested in mechanism (2). However, further information is needed to identify whether only one OEM mechanism exists. V. Environmental Parameter Space for the IRM and OEM A summary of the key features of the OEM and IRM for stress-corrosion cracking of ceramic composites is given in Table I. Only a few characteristics are similar between these two processes: (1) time-dependent behavior, (2) the presence of a Kth, and (3) subcritical crack growth stages. The differences between 2002 Journal of the American Ceramic Society—Jones et al. Vol. 83, No. 8
august 2000 Tablel similarities and differenees between oem and IrM M Solid reaction produc Gaseous reaction product Brittle glass layer Reduced interface strength arrow temperature range ncreasing Local strength reduction No or little effect on local strength and, on engineering strength dependent behavior hold stress intensity for crack growth itical crack growth in stages i and nl these two mechanisms result primarily from the fundamentally and both 0.15 and 1.0 um thick interphases.2, 19, The fiber different crack growth mechanisms. The proposed formation of a relaxation-mechanism(FRM) regime includes that area where brittle glass phase that locally decreases the fiber strength accounts crack growth occurs in the absence of a significant environmental for the OEM characteristics occurring within a narrow temperature effect and is controlled by fiber creep, as described in Section range, the decreased local and engineering strength, and the IV(B). The fiber-debond model described in Fig. 6 also describes absence of a need for a dynamic stress. On the other hand, the the conditions controlling the FRM, except that the debond length removal of the interphase and the resulting decrease in crack- in the latter case is controlled by the stress in the fiber. and there closure forces by the bridging fibers during the IRM accounts for is no time-dependent change in the interface caused by oxidation the temperature dependence and the lack of an effect on engineer- The effect of coatings, glass-forming additions such as BC, and ing strength up to the time for total burnout of the interphase environmental species other than oxygen were not considered in developing the failure- mechanism map in Fig. 7. The map can be both the toughness and the dynamic crack velocity of a material sed to identify the type of crack growth process leading to failure and higher velocities. The OEM has the potential to affect both needed to consider the effects of coatings, glass-forming addition processes, whereas the IRM appears to affect only the subcritical crack growth process. Hydrog and other environmental species, such as H,O; alkali elements duced crack growth of metals and alloys can affect both the fracture toughness and the subcritical uch as sodium or potassium; or combustion gases. The devel crack velocity, as does the OEM whereas aqueous stress corrosion present work, because the boundaries shown in Fig. 7 most likely implications of this phenomenon are that a process that both will shift, and some regimes may not exist, depending on the decreases the K and ses da/dt in stage lI can have a much conditions greater impact on material performance than one that affects only The SCC mechanism map shows the T and Po, values at which the subcritical da/dr the oEM and the IRM have been observed. In all cases. for Two primary variables that define the operative OEM or IRM samples exposed to Po, values equal to or greater than those of air, environmental parameter space are(1)temperature and(2) Po, as but tested at <1073 K, the OEM type of mechanism occurred.For map given in Fig. 7. The IRM section was determined with at 1073 K and 2 x 10 Pa at 1373 K, the IRM type of mechanism composite materi c-grade Nicalon fibers occurred. Other variables that may be involved include time, the fiber-matrix interphase thickness, and the composition of the glas phase that forms on the fiber, the presence of fluxing agents, such as boron in BN interphases, alter glass-flow properties It is possible that the IRM operates at short times and the OEm at longer times. For instance, weight change tests of specimens with a I um thick arbon interphase show that all of the carbon air oxidizes before significant SiO, forms, therefore, the IRM could be operative until the SiO, formed is thick enough to embrittle the fiber or to bond the fiber to the matrix, leading to the oEm. Of course, this 1423 and 1773 K, based on measurement Gf thich perature liley iso embrittlement or bonding would only occur below a critical ten G ature, T, where the oxide is brittle. The critical ten related to the glass-transition temperature, T. nges between an amorphous scale formed between silicon and SiC, obtained by Futakawa and Stein- 14731673 brech, and those for an amorphous SiO2, reported by Bansal and The observations by Lara-Curzio et al. o and Becher et al.of 999):C, 0.15 micron the IRM in air at 698 K, where the growth of the glass layer would 04 1-0izr m B-enhanced matrix be very slow, support the idea of the IRM operating at short times, th the potential for the oEm at longer times, i.e., at lower Lara-Curzio and Ferber (1997): icons d Heredia et al.(1995): C.0.1-3.0 tresses or slower failure mechanisms. At 1073-1373K and low o Ishikawa(1994): C, 0.3 microns ition from irm to oem ailure. however at high stresses. it has been demonstrated Failure-mechanism map for continuous-fiber ceramic composites (O)data obtained from posttreatment, room-temperature experi- point is further supported by the oxidation-kinetic data reported by er symbols designate OEM observations, closed symbols des- Costello and Tressler for Sic in pure, dry FRM) show the time required to form an SiO2
these two mechanisms result primarily from the fundamentally different crack growth mechanisms. The proposed1 formation of a brittle glass phase that locally decreases the fiber strength accounts for the OEM characteristics occurring within a narrow temperature range, the decreased local and engineering strength, and the absence of a need for a dynamic stress. On the other hand, the removal of the interphase and the resulting decrease in crackclosure forces by the bridging fibers during the IRM accounts for the temperature dependence and the lack of an effect on engineering strength up to the time for total burnout of the interphase. Many environmentally induced cracking processes can affect both the toughness and the dynamic crack velocity of a material, which in effect shifts the da/dt–K curve to lower stress intensities and higher velocities. The OEM has the potential to affect both processes, whereas the IRM appears to affect only the subcritical crack growth process. Hydrogen-induced crack growth of metals and alloys can affect both the fracture toughness and the subcritical crack velocity, as does the OEM, whereas aqueous stress corrosion normally affects only the subcritical crack growth behavior.32 The implications of this phenomenon are that a process that both decreases the KIc and increases da/dt in stage II can have a much greater impact on material performance than one that affects only the subcritical da/dt. Two primary variables that define the operative OEM or IRM environmental parameter space are (1) temperature and (2) pO2 , as summarized in the stress-corrosion cracking (SCC) mechanism map given in Fig. 7. The IRM section was determined with composite material reinforced with ceramic-grade Nicalon fibers and both 0.15 and 1.0 mm thick interphases.2,19,33–37 The fiberrelaxation-mechanism (FRM) regime includes that area where crack growth occurs in the absence of a significant environmental effect and is controlled by fiber creep, as described in Section IV(B). The fiber-debond model described in Fig. 6 also describes the conditions controlling the FRM, except that the debond length in the latter case is controlled by the stress in the fiber, and there is no time-dependent change in the interface caused by oxidation. The effect of coatings, glass-forming additions such as B4C, and environmental species other than oxygen were not considered in developing the failure-mechanism map in Fig. 7. The map can be used to identify the type of crack growth process leading to failure of a material, assuming coating failure. Separate maps would be needed to consider the effects of coatings, glass-forming additions, and other environmental species, such as H2O; alkali elements, such as sodium or potassium; or combustion gases. The development of maps that consider these factors is beyond the scope of the present work, because the boundaries shown in Fig. 7 most likely will shift, and some regimes may not exist, depending on the conditions. The SCC mechanism map shows the T and pO2 values at which the OEM and the IRM have been observed. In all cases, for samples exposed to pO2 values equal to or greater than those of air, but tested at ,1073 K, the OEM type of mechanism occurred. For tests conducted at 1073–1373 K, but at pO2 values of #19.5 3 103 Pa at 1073 K and 2 3 103 Pa at 1373 K, the IRM type of mechanism occurred. Other variables that may be involved include time, the fiber–matrix interphase thickness, and the composition of the glass phase that forms on the fiber; the presence of fluxing agents, such as boron in BN interphases, alter glass-flow properties. It is possible that the IRM operates at short times and the OEM at longer times. For instance, weight change tests10 of specimens with a 1 mm thick carbon interphase show that all of the carbon in the interphase oxidizes before significant SiO2 forms; therefore, the IRM could be operative until the SiO2 formed is thick enough to embrittle the fiber or to bond the fiber to the matrix, leading to the OEM. Of course, this embrittlement or bonding would only occur below a critical temperature, Tc, where the oxide is brittle. The critical temperature likely is related to the glass-transition temperature, Tg, which ranges between 1423 and 1773 K, based on measurements for an amorphous scale formed between silicon and SiC, obtained by Futakawa and Steinbrech,38 and those for an amorphous SiO2, reported by Bansal and Doremus.39 The observations by Lara-Curzio et al. 40 and Becher et al. 41 of the IRM in air at 698 K, where the growth of the glass layer would be very slow, support the idea of the IRM operating at short times, with the potential for the OEM at longer times, i.e., at lower stresses or slower failure mechanisms. At 1073–1373 K and low stresses, a transition from IRM to OEM may occur before sample failure; however, at high stresses, it has been demonstrated10,42,43 that failure occurs by the IRM before a transition to the OEM. The point is further supported by the oxidation-kinetic data reported by Costello and Tressler44 for SiC in pure, dry oxygen. Those data show the time required to form an SiO2 layer thick enough to Table I. Similarities and Differences between OEM and IRM OEM IRM Differences Solid reaction product Gaseous reaction product Brittle glass layer Reduced interface strength Narrow temperature range Local strength reduction Increasing da/dt with increasing temperature No or little effect on local strength and, on engineering strength Similarities Time-dependent behavior Threshold stress intensity for crack growth Subcritical crack growth in stages I and II Fig. 7. Failure-mechanism map for continuous-fiber ceramic composites ((ƒ) and (E) data obtained from posttreatment, room-temperature experiments; other symbols designate OEM observations; closed symbols designate IRM or FRM). August 2000 Stress-Corrosion Cracking of Silicon Carbide Fiber/Silicon Carbide Composites 2003
Journal of the bridge the fiber-matrix gap for material with a 0. 15 um thick F. E. Heredia, J. C. MeNulty, F. w. Zok, and A. G. Evans, Oxidatio carbon layer as -170 h at 1073K and-7h at 373K. This upper 2097-100(1995) Embrittlement Probe for Ceramic-Matrix Composites, J. Am. Ceram. Soc., 78 temperature limit for the IRM, shown in Fig. 7, was determined by H-T. Lin and P. F. Becher, ""Effect of Fiber Coating on Lifetime of Nicalon Henager and Jones'and Lewinsohn et al. using tests that lasted Fiber-Silicon Carbide Composites in Air,"Mater. Sci Eng. A,231 a few hours. so that the transition from irm to oem would not S. Raghuraman, M. K. Ferber, J. F. Stubbins, and A.A. Oxidation Tests in SiC/SiC Composites",pp. 1015-26 in Co /ol The boundaries of the IRM and OEM regions, shown in Fig. 7, Ban sa ames ca ceramic societ westerwelle ohd egg have been defined by the available experimental data. However, Microstructural Chang the IRM region may extend to both lower and high temperatures and to lower Po values. The experimentally determined upper Proceedings of the 20th Anmal Conference on Composites, Advanced Ceramics, temperature limit results from the lack of fiber stability at higher Westerville, OH, 1996 Structures-B. Edited by V Greenhut. American Ceramic Society, temperatures, such that the FRM becomes dominant at higher Po values In other words, the fiber creep rate dominates sample failure long before carbon oxidation can play a role. A composite the 17th Annual Conference on Composites with a more thermally stable, creep-resistant fiber would support a by D. Cranmer, Ameri O. Unal, A. J. Eckel, and F. C. Laabs, "Mechanical Properties and Microstructure higher IRM regime. The experimentally determined lower temper- n to expect the IRM to continue Sorbide.mR下Cm由四cwSm ature limit is primarily a result of the limits of reasonable at lower temperatures, except that the oem boundary is likely to So277192381-94(1994 Jones, C. H. Henager Jr, and C. F. Windisch Jr, "High-Temperature love to lower Po, values with increased time, as discussed above Corrosion and Crack Growth of SiC-SiC at Variable Oxygen Partial Pressures, Likewise, the IRM could extend to lower Po, values, but the FRM becomes the dominant crack growth mechanism at low DC. F. Windisch Jr, C. H. Henager Jr, G. D. Springer, and R H. Jones, "Oxidation of the Carbon Interface in Nicalon-Fiber-Reinforced Silicon Carbide Composite, In summary, regardless of the temperature and oxygen pressure, J. dm. Ceram Soc, 80 [3]569-74(199 transition temperature(T> T )or at an oxide thickness lower than during Slow Crack Growth and the Result Jones, "Crack Bridging by SiC Fibers C. R. Jones. C. H. He tant Fracture Toughness of SiC/SiCr a critical value(dd) L. Filipuzzi, G. Camus, lain, and J. Thebault, "Oxidation Mechanisms and Composite Materials: I Ceram.Soc,77[2]467-80(1994) VI. Conclusions A, J. Eckel, J, D. Cawley, and T. A. Parthasarathy, "Oxidation Kinetics of a arbon Phase in a Nonreactive Matrix,J.Am. Ceran. Soc., 78 (41 Experimental weight loss and crack growth data support the 72-80(1995) conclusion that the oxygen-enhanced crack growth of Sic/Sic ISC. A. Lewinsohn, J. 1. Eldridge, and R. H. Jones, "Techniques for Measuring occurs by more than one mechanism, depending on the exper Interfacial Recession in CFCCs and th ications on Subcritical Crack Growth nental conditions. An OEM operates at temperatures s1373 Ceram, Eng. Sci. Proc., 19 [3] 19-26(1998). okan dation behavior of Sic-Fiber nd at high oxygen pressures; an IRM operates at temperatures Reinforced SiC, " J Nuc. Mater, 227, 130-37(1995). >700 K and low oxygen pressures. The IRM may operate at short litte, M. Gomina, and J. vicens, "TEM Observations of SiC-Sic times, with a transition to the oem at longer times, if the stress is Composites with a Carbon Interphase Layer Annealed in Air at High Temperatures, low enough that sample failure does not occur first. reep Behavior and Structural Characterization at The OEm results from the reaction of oxygen w m High Temperatures of Nicalon SiC Fibers,J Mater. Sci., 19, 3658-70 a glass layer on the fiber. The fracture stress of the fiber is I9C. H. Henager Jr. and R. H. Jones, "High-Temperature Plasticity decreased if this layer is thicker than a critical value(d> d) and and Subcritical Crack Growth in Ceramic Composites, the temperature below a critical value(TT, and d> de. the IrM process has only been Degradation Behavior of Silicon Carbide Fiber Hi-Nicalon Type-S, "JNacl. Mater. observed in SiC/SiC with either carbon or bn as the fiber-matrix 258-63,1594-99(1998 interface material. Further research is needed to identify the W. H. Glime and J.D. Cawle ss Concentration Due to Fiber-Matrix Fusion eram. Soc,8loj2597-604(1998) specific temperature and Po, region in which the IRM and the OEM operate in SiCSic and whether there are conditions Exposure of Salt(NaCI) Water and Oxidation on the Strength of Uncoated and where both mechanisms also operate in glass-and oxide-matrix BN-Coated Nicalon Fibers, J Am Ceram. Soc-81171812-18(1998) Physics and Mechanics of Acknowledgments Temperatures, "J. Am. Cera. Soc., 79[3]591-96(1996) 2T. Helmer, H. Peterlik, and K. Kromp, "Coating of Carbon Fibers-The Strength the Office of Basic Department of Energy, under Contract No. DE-ACO6-75RLO 1830, with Battelle ies Degradation of sic Memorial Institute, which operates Pacific Northwest al Laboratory for the Fibers Below 850%C Mate ,13,680-83(1994) Department of Energy N. Okabe, I Murakami, I Y. Yoshioka, and H. Ichikawa, "Environmen- Deterioration and Damage of Ceramic Matrix Composites", in Proceedings of the by G. Pfendt American Ceramic Society, Westerville, OH, 1995. References 32R. H Jones and R. E. Ricker. "Mechanisms of Stress-Corrosion Cracking". in IA. G. Evans, F. w. Z M. McMecking, and Z. Z. Du "Models of Jones. .merican Socicty for Meals, Materils park. od, 1992. on. Edited by R. 3C. A Lewinsohn, C H. Henager Jr, and R H. Jones, "Environmentally Induced Composites,J. Am. Ceram 92345-52(199 Failure-Mechanism Mapping for Continuous-Fiber, Ceramic Composites", pp
bridge the fiber–matrix gap for material with a 0.15 mm thick carbon layer as ;170 h at 1073 K and ;7 h at 1373 K. This upper temperature limit for the IRM, shown in Fig. 7, was determined by Henager and Jones19 and Lewinsohn et al.,33 using tests that lasted a few hours, so that the transition from IRM to OEM would not require a much longer duration test. The boundaries of the IRM and OEM regions, shown in Fig. 7, have been defined by the available experimental data. However, the IRM region may extend to both lower and high temperatures and to lower pO2 values. The experimentally determined upper temperature limit results from the lack of fiber stability at higher temperatures, such that the FRM becomes dominant at higher pO2 values. In other words, the fiber creep rate dominates sample failure long before carbon oxidation can play a role. A composite with a more thermally stable, creep-resistant fiber would support a higher IRM regime. The experimentally determined lower temperature limit is primarily a result of the limits of reasonable experimental times. There is reason to expect the IRM to continue at lower temperatures, except that the OEM boundary is likely to move to lower pO2 values with increased time, as discussed above. Likewise, the IRM could extend to lower pO2 values, but the FRM becomes the dominant crack growth mechanism at low pO2 values. In summary, regardless of the temperature and oxygen pressure, the IRM appears to operate at temperatures exceeding the glasstransition temperature (T . Tg) or at an oxide thickness lower than a critical value (d , dc), and the OEM appears to operate at temperatures below the glass-transition temperature (T , Tg) and at an oxide thickness above a critical value (d . dc). VI. Conclusions Experimental weight loss and crack growth data support the conclusion that the oxygen-enhanced crack growth of SiCf /SiC occurs by more than one mechanism, depending on the experimental conditions. An OEM operates at temperatures &1373 K and at high oxygen pressures; an IRM operates at temperatures .700 K and low oxygen pressures. The IRM may operate at short times, with a transition to the OEM at longer times, if the stress is low enough that sample failure does not occur first. The OEM results from the reaction of oxygen with SiC to form a glass layer on the fiber. The fracture stress of the fiber is decreased if this layer is thicker than a critical value (d . dc) and the temperature below a critical value (T , Tg), such that a sharp crack can be sustained in the layer. Other possible, but not demonstrated, OEMs include (1) glass-phase formation in the fiber–matrix interface and (2) fiber-strength reduction by reaction with oxygen. The IRM results from the oxidation of the interfacial layer and the resulting relaxation of the bridging fibers. Interface removal contributes to the stress relaxation of the fiber that occurs by creep. The IRM occurs over a wide range of temperatures for d , dc and may occur at T . Tg and d . dc. The IRM process has only been observed in SiCf /SiC with either carbon or BN as the fiber–matrix interface material. Further research is needed to identify the specific temperature and pO2 region in which the IRM and the OEM operate in SiCf /SiC and whether there are conditions where both mechanisms also operate in glass- and oxide-matrix composites. Acknowledgments This research was supported by the Office of Basic Energy Sciences of the U.S. Department of Energy, under Contract No. DE-AC06-75RLO 1830, with Battelle Memorial Institute, which operates Pacific Northwest National Laboratory for the Department of Energy. References 1 A. G. Evans, F. W. Zok, R. M. McMeeking, and Z. Z. Du, “Models of High-Temperature, Environmentally Assisted Embrittlement in Ceramic-Matrix Composites,” J. Am. Ceram. Soc., 79 [9] 2345–52 (1996). 2 F. E. Heredia, J. C. McNulty, F. W. Zok, and A. G. Evans, “Oxidation Embrittlement Probe for Ceramic-Matrix Composites,” J. Am. Ceram. Soc., 78 [8] 2097–100 (1995). 3 H-T. Lin and P. F. Becher, “Effect of Fiber Coating on Lifetime of Nicalon Fiber–Silicon Carbide Composites in Air,” Mater. Sci. Eng. A, 231 [1–2] 143–50 (1997). 4 S. Raghuraman, M. K. Ferber, J. F. Stubbins, and A. A. Wereszcak, “StressOxidation Tests in SiCf/SiC Composites”; pp. 1015–26 in Ceramic Transactions, Vol. 46, Advances in Ceramic-Matrix Composites II. Edited by J. P. Singh and N. P. Bansal. American Ceramic Society, Westerville, OH, 1994. 5 P. F. Tortorelli and K. L. More, “Time Dependence of Oxidation-Induced Microstructural Changes in Nicalon- and Nextel-Reinforced SiC”; pp. 366–74 in Proceedings of the 20th Annual Conference on Composites, Advanced Ceramics, Materials, and Structures—B. Edited by V. Greenhut. American Ceramic Society, Westerville, OH, 1996. 6 P. F. Tortorelli, S. Nijhawan, L. Riester, and R. A. Lowden, “Influence of Fiber Coatings on the Oxidation of Fiber-Reinforced SiC Composites”; in Proceedings of the 17th Annual Conference on Composites and Advanced Ceramic Materials. Edited by D. Cranmer. American Ceramic Society, Westerville, OH, 1993. 7 O. Unal, A. J. Eckel, and F. C. Laabs, “Mechanical Properties and Microstructure of Oxidized SiC/SiC Composites”; see Ref. 8, pp. 333–41. 8 C. H. Henager Jr. and R. H. Jones, “Subcritical Crack Growth in CVI Silicon Carbide Reinforced with Nicalon Fibers: Experiment and Model,” J. Am. Ceram. Soc., 77 [9] 2381–94 (1994). 9 R. H. Jones, C. H. Henager Jr., and C. F. Windisch Jr., “High-Temperature Corrosion and Crack Growth of SiC–SiC at Variable Oxygen Partial Pressures,” Mater. Sci. Eng. A, 198, 103–12. 10C. F. Windisch Jr., C. H. Henager Jr., G. D. Springer, and R. H. Jones, “Oxidation of the Carbon Interface in Nicalon-Fiber-Reinforced Silicon Carbide Composite,” J. Am. Ceram. Soc., 80 [3] 569–74 (1997). 11C. R. Jones, C. H. Henager Jr., and R. H. Jones, “Crack Bridging by SiC Fibers during Slow Crack Growth and the Resultant Fracture Toughness of SiC/SiCf Composites,” Scr. Metall. Mater., 33, 2067–72 (1995). 12L. Filipuzzi, G. Camus, R. Naslain, and J. Thebault, “Oxidation Mechanisms and Kinetics of One-Dimensional SiC/SiC Composite Materials: I, An Experimental Approach,” J. Am. Ceram. Soc., 77 [2] 459–66 (1994). 13L. Filipuzzi, G. Camus, R. Naslain, and J. Thebault, “Oxidation Mechanisms and Kinetics of One-Dimensional SiC/SiC Composite Materials: II, Modeling,” J. Am. Ceram. Soc., 77 [2] 467–80 (1994). 14A. J. Eckel, J. D. Cawley, and T. A. Parthasarathy, “Oxidation Kinetics of a Continuous Carbon Phase in a Nonreactive Matrix,” J. Am. Ceram. Soc., 78 [4] 972–80 (1995). 15C. A. Lewinsohn, J. I. Eldridge, and R. H. Jones, “Techniques for Measuring Interfacial Recession in CFCCs and the Implications on Subcritical Crack Growth,” Ceram. Eng. Sci. Proc., 19 [3] 19–26 (1998). 16H. Kleykamp, V. Schauer, and A. Skokan, “Oxidation Behavior of SiC-FiberReinforced SiC,” J. Nucl. Mater., 227, 130–37 (1995). 17I. Sebire-Lhermitte, M. Gomina, and J. Vicens, “TEM Observations of SiC–SiC Composites with a Carbon Interphase Layer Annealed in Air at High Temperatures,” J. Microsc. (Oxford), 169, 97–205 (1993). 18G. Simon and A. R. Bunsell, “Creep Behavior and Structural Characterization at High Temperatures of Nicalon SiC Fibers,” J. Mater. Sci., 19, 3658–70 (1984). 19C. H. Henager Jr. and R. H. Jones, “High-Temperature Plasticity Effects in Bridged Cracks and Subcritical Crack Growth in Ceramic Composites,” Mater. Sci. Eng. A, 166, 211–20 (1993). 20H-T. Lin, P. F. Becher, and P. F. Tortorelli, “Elevated-Temperature Static Fatigue of a Nicalon-Fiber-Reinforced SiC Composite,” Mater. Res. Soc. Symp. Proc., 365, 435–40 (1995). 21D. A. Woodford, D. R. Van Steele, J. A. Brehm, L. A. Timms, and J. E. Palko, “Testing the Tensile Properties of Ceramic-Matrix Composites,” JOM, 57 [5] 63 (1993). 22E. Lara-Curzio, “Stress-Rupture of Nicalon/SiC Continuous Fiber Ceramic Composites in Air at 950°C,” J. Am. Ceram. Soc., 80 [12] 3268–72 (1997). 23Y. T. Zhu, S. T. Taylor, M. G. Stout, D. P. Butt, and T. C. Lowe, “Kinetics of Thermal, Passive Oxidation of Nicalon Fibers,” J. Am. Ceram. Soc., 81 [3] 655–60 (1998). 24M. Takeda, A. Urano, J. Sakamoto, and Y. Imai, “Microstructure and Oxidative Degradation Behavior of Silicon Carbide Fiber Hi-Nicalon Type-S,” J. Nucl. Mater., 258–63, 1594–99 (1998). 25W. H. Glime and J. D. Cawley, “Stress Concentration Due to Fiber-Matrix Fusion in Ceramic Matrix Composites,” J. Am. Ceram. Soc., 81 [10] 2597–604 (1998). 26T. A. Parthasarathy, C. A. Folsom, and L. P. Zawada, “Combined Effects of Exposure of Salt (NaCl) Water and Oxidation on the Strength of Uncoated and BN-Coated Nicalon Fibers,” J. Am. Ceram. Soc., 81 [7] 1812–18 (1998). 27A. G. Evans and F. W. Zok, “Review: The Physics and Mechanics of Fiber-Reinforced Brittle Matrix Composites,” J. Mater. Sci., 29, 3857–96 (1994). 28D. Singh, J. P. Singh, and M. J. Wheeler, “Mechanical Behavior of SiC/SiC Composites and Correlation to in Situ Fiber Strength at Room and Elevated Temperatures,“ J. Am. Ceram. Soc., 79 [3] 591–96 (1996). 29T. Helmer, H. Peterlik, and K. Kromp, “Coating of Carbon Fibers—The Strength of the Fibers,” J. Am. Ceram. Soc., 78 [1] 133–36 (1995). 30Y. Gogotsi and M. Yoshimura, “Oxidation and Properties Degradation of SiC Fibers Below 850°C,” J. Mater. Sci. Lett., 13, 680–83 (1994). 31N. Okabe, I. Murakami, H. Hirata, Y. Yoshioka, and H. Ichikawa, “Environmental Deterioration and Damage of Ceramic Matrix Composites”; in Proceedings of the 19th Annual Conference on Composites, Advanced Ceramics, Materials, and Structures—B. Edited by G. Pfendt. American Ceramic Society, Westerville, OH, 1995. 32R. H. Jones and R. E. Ricker, “Mechanisms of Stress-Corrosion Cracking”; in Stress-Corrosion Cracking: Materials Performance and Evaluation. Edited by R. H. Jones. American Society for Metals, Materials Park, OH, 1992. 33C. A. Lewinsohn, C. H. Henager Jr., and R. H. Jones, “Environmentally Induced Failure-Mechanism Mapping for Continuous-Fiber, Ceramic Composites”; pp. 2004 Journal of the American Ceramic Society—Jones et al. Vol. 83, No. 8
August 2000 EbPS面 N. Bansal. Ar抽 cier i w Ceramic composites E. Lara-Curzio, P F. Tortorelli, and K. L More, ""Stress-R E. Lara-Curzio,"Analysis of Oxidation-Assisted Stress-Rupture of Continuous 以 ntermediate Temperatur” eram. Eng. Sci. Pro18120 F Becher, H-T. Lin, and K. L More, "Lifetime-Applied Stress Response Fiber-Reinforced Ceramic-Matrix Composites at Intermediate Temperatures, "Com- Air of a SiC-Based Nicalon-Fiber-Reinforced Composite with a Carbon Interfacial Layer: Effects of Temperature(300°to1150°C),”J.Am. Ceran Soc,81四7 sSE. Lara-Curzio, personal communication, 1997. -25(1998) bE. Lara-Curzio and M. K. Ferber "Stress-Rupture of Continuous-Fibre Ceramic 4-C. H. Henager Jr,R. H. Jones, C. F. Windisch Jr, M. M. Stackpoole, and R Composites at Intermediate Temperatures, "J. Mater. Sci. Letf, 1 [6]23-26(1997). Bordia,"I pendent, Environmentally Assisted Crack Growth in Nicalon-Fiber- 37T Ishikawa, "Recent Developments of the SiC Fiber Nicalon and Its Composites, Reinforced SiC Composites at Elevated Temperatures, Metall. Mater. Trans. A, 27A, ncluding Properties of the SiC Fiber Hi-Nicalon for Ultra-High Temperature, Compos. Sci. Technol, 5[1] 135-44(1994) 3C. A. Lewinsohn, C. H. Henager Jr, and R. H. Jones, "Subcritical Crack Growth n CVI SiC/SiC Composites at Elevated Temperatures: Effect of Fiber Creep Rate, SiSiC at Elevated Temperatures, "J. Am. Ceram Soc., 81 [7 1819-23(1998)- 3N. P. Bansal and R H. Doremus, Handbook of Glass Properties; pp. 227-305 J. A. Costello and R. E. Tressler, "Oxidation Kinetics of Silicon Carbide Crystals Academic Press, New York, 1986 and Ceramics: 1, In Dry Oxygen, "JAm Ceram Soc., 69[9]674-81(1986). D
351–59 in Ceramic Transactions, Vol. 96, Advances in Ceramic Composites IV. Edited by J. P. Singh and N. Bansal. American Society, Westerville, OH, 1999. 34E. Lara-Curzio, “Analysis of Oxidation-Assisted Stress-Rupture of ContinuousFiber-Reinforced Ceramic-Matrix Composites at Intermediate Temperatures,” Composites A (Guildford, U.K.), 30 [4] 549–54 (1999). 35E. Lara-Curzio, personal communication, 1997. 36E. Lara-Curzio and M. K. Ferber, “Stress-Rupture of Continuous-Fibre Ceramic Composites at Intermediate Temperatures,” J. Mater. Sci. Lett., 1 [6] 23–26 (1997). 37T. Ishikawa, “Recent Developments of the SiC Fiber Nicalon and Its Composites, Including Properties of the SiC Fiber Hi-Nicalon for Ultra-High Temperature,” Compos. Sci. Technol., 5 [1] 135–44 (1994). 38M. Futakawa and R. W. Steinbrech, “Viscosity of Amorphous Oxide Scales on SiSiC at Elevated Temperatures,” J. Am. Ceram. Soc., 81 [7] 1819–23 (1998). 39N. P. Bansal and R. H. Doremus, Handbook of Glass Properties; pp. 227–305. Academic Press, New York, 1986. 40E. Lara-Curzio, P. F. Tortorelli, and K. L. More, “Stress-Rupture of NicalonTM/ SiC at Intermediate Temperatures,” Ceram. Eng. Sci. Proc., 18 [4] 209–19 (1997). 41P. F. Becher, H-T. Lin, and K. L. More, “Lifetime-Applied Stress Response in Air of a SiC-Based Nicalon-Fiber-Reinforced Composite with a Carbon Interfacial Layer: Effects of Temperature (300° to 1150°C),” J. Am. Ceram. Soc., 81 [7] 1919–25 (1998). 42C. H. Henager Jr., R. H. Jones, C. F. Windisch Jr., M. M. Stackpoole, and R. Bordia, “Time-Dependent, Environmentally Assisted Crack Growth in Nicalon-FiberReinforced SiC Composites at Elevated Temperatures,” Metall. Mater. Trans. A, 27A, 839–49 (1996). 43C. A. Lewinsohn, C. H. Henager Jr., and R. H. Jones, “Subcritical Crack Growth in CVI SiCf/SiC Composites at Elevated Temperatures: Effect of Fiber Creep Rate,” Acta Metall., in press. 44J. A. Costello and R. E. Tressler, “Oxidation Kinetics of Silicon Carbide Crystals and Ceramics: I, In Dry Oxygen,” J. Am. Ceram. Soc., 69 [9] 674–81 (1986). M August 2000 Stress-Corrosion Cracking of Silicon Carbide Fiber/Silicon Carbide Composites 2005