ournal /. An. Ceram Soc, 81[7] 1797-811(1998) Mechanical Behavior and High-Temperature Performance of a Woven Nicalon TM/Si-N-C Ceramic-Matrix Composite S. Steven Lee, t Larry P Zawada, James M. Staehler, and Craig A Folsom Metals and Ceramics Division, USAF Wright Laboratory Materials Directorate, Wright-Patterson Air Force Base, Ohio 45433-7817 A modern ceramic-matrix composite(CMC) has been ex- the matrix are deflected along the fiber/matrix interface rather tensively characterized for a high-temperature aerospace han continuing through the fiber turbine-engine application. The CMC system has a silicon- In many early CMC systems, the weak interface between the nitrogen-carbon(Si-N-C)matrix reinforced with Nicalon fiber and the matrix contained carbon. However. the carbon fibers woven in a balanced eight-harness satin weave fab layer is oxidized in oxygen-rich environments at temperatures ric. Tensile tests have demonstrated that this CMC exhibits as low as 400 C. After the carbon is gone, the oxygen reacts excellent strength retention up to 1100C. The roor with the fiber to form a silica(SiO2) layer on the surface of the temperature fatigue limit was 160 MPa, -80% of the roo fiber - The SiO2 layer significantly weakens the fiber and temperature tensile strength. The composite reached run- also allows strong bonding to the matrix, which results in a out conditions under cyclic(105 cycles at 1 Hz) and significant decrease in fracture toughness for the CmC sustained tension(100 h) conditions at a stress of 110 MPa, As an alternative to carbon, boron nitride(Bn-2 has been which was -35 MPa above the proportional limits at tem considered as a new fiber coating, and several CMC manufac cyclic load up to 1100C in air At stress levels >110 MPa, turers have switched to a BN-containing interface between the pel fiber and the matrix. However, BN that is not fully crystalline life, based on time, compared with sustained tension. Fur- (turbostratic with a d (002)spacing of >3.33 A)is still suscep- ther life degradation was observed in the 1000 C fatigue tible to oxidation at temperatures 2650 C, 15 In addition, non- specimens that were exposed to a salt-fog environment. crystalline bn has been shown to be sensitive to moisture and This degradation decreased the fatigue life -85% at the oxidation. 3 Crystalline BN has been deposited via chemical stress levels that were tested vapor deposition(CVD)only at deposition temperatures 1400C, however, this high of a temperature damages the current ceramic fibers and the deposition rates are difficult to . Introduction control. The oxidation process is accelerated whenever the ap- plied tensile stress is high enough to cause matrix cracking in DVANCED high-temperature material systems are a key CMCs. Such damage-enhanced oxidation is typically the dom- technological tool in the evolution of current aerospace inant degradation mechanism, which limits the high ropulsion systems. In the past decade, the aerospace design community has greatly improved the understanding of the ther- hel me impnt investigation involves the evaluation of the du- mal environment within an aerospace turbine engine, thus al- rability improvements offered by CMCs, compared with metal lowing for new concepts in airfoil design, cooling, and com- parts, for a demanding military turbine-engine application. Ad- bustion. However, major advances can only be realized with vanced military aerospace turbine engines use an afterburner greatly improved material and a thorough understanding of with vectoring nozzles to produce supersonic flight. Compared their high-temperature mechanical behavior and performance to the design life, the superalloy nozzle components have a Among the available high-temperature material systems cu greatly reduced service life. CMCs are being analyzed not only rently being developed, fiber-reinforced ceramic-matrix com- to meet the design life but also to advance the design life by a posites(CMCs) have attracted great attention for aerospace factor of 3. This increase in life is one of several advantages to remarkable damage tolerance, with a fracture toughness several using CMCs and is expected to offset an anticipated greater applications. Compared to monolithic ceramics, CMCs exhibit acquisition cost over the superalloy parts imes higher than most monolithic ceramics Defect sensitivity The main focus of the investigation was to characterize the and lack of toughness continue to be the two main technica basic mechanical response as well as the long-term high hurdles that exclude monolithic ceramics from aerospace tur- temperature performance and durability of the CMC system bine engines. In addition, these limitations are compounded by NicalonTM/SINC. Furthermore, the components of interest are damage from foreign objects. However, high-toughness CMCs located at the end of the exhaust nozzle. At this location they require interfaces between the fiber and the matrix that possess are exposed to rain and sodium (Na*)and other ions from low fracture toughness, such that cracks propagating throug ocean mists and exhaust deposits. In consideration of these conditions, the effect of salt-fog exposure on the CMC was considered and simulated using an environment-controlled The test conditions used in the investigation were chosen in G. Evans--contributing editor such a way that the feasibility of using the present CMC for the nce and material characteristics: (i) the fiber/matrix interfa Manuscript No. 192004. Received March 13, 1997; approved October 14, 1997. cial strength, (i) short-term mecha sponse at room and ture long-term uld be addressed. Now with Structures Tech- WAg8124g2Sopzeed Civil Transport, Boeing Commerical Airplane Group, Scale formance and durability, and (iv) ce to environmental 1797
Mechanical Behavior and High-Temperature Performance of a Woven Nicalon™/Si-N-C Ceramic-Matrix Composite S. Steven Lee,*,† Larry P. Zawada,* James M. Staehler,* and Craig A. Folsom* Metals and Ceramics Division, USAF Wright Laboratory Materials Directorate, Wright-Patterson Air Force Base, Ohio 45433–7817 A modern ceramic-matrix composite (CMC) has been extensively characterized for a high-temperature aerospace turbine-engine application. The CMC system has a silicon– nitrogen–carbon (Si-N-C) matrix reinforced with Nicalon fibers woven in a balanced eight-harness satin weave fabric. Tensile tests have demonstrated that this CMC exhibits excellent strength retention up to 1100°C. The roomtemperature fatigue limit was 160 MPa, ∼80% of the roomtemperature tensile strength. The composite reached runout conditions under cyclic (105 cycles at 1 Hz) and sustained tension (100 h) conditions at a stress of 110 MPa, which was ∼35 MPa above the proportional limits at temperatures up to 1100°C in air. At stress levels >110 MPa, cyclic loading at 1000°C caused a more severe reduction in life, based on time, compared with sustained tension. Further life degradation was observed in the 1000°C fatigue specimens that were exposed to a salt-fog environment. This degradation decreased the fatigue life ∼85% at the stress levels that were tested. I. Introduction ADVANCED high-temperature material systems are a key technological tool in the evolution of current aerospace propulsion systems. In the past decade, the aerospace design community has greatly improved the understanding of the thermal environment within an aerospace turbine engine, thus allowing for new concepts in airfoil design, cooling, and combustion. However, major advances can only be realized with greatly improved material and a thorough understanding of their high-temperature mechanical behavior and performance. Among the available high-temperature material systems currently being developed, fiber-reinforced ceramic-matrix composites (CMCs) have attracted great attention for aerospace applications. Compared to monolithic ceramics, CMCs exhibit remarkable damage tolerance, with a fracture toughness several times higher than most monolithic ceramics. Defect sensitivity and lack of toughness continue to be the two main technical hurdles that exclude monolithic ceramics from aerospace turbine engines. In addition, these limitations are compounded by damage from foreign objects. However, high-toughness CMCs require interfaces between the fiber and the matrix that possess low fracture toughness, such that cracks propagating through the matrix are deflected along the fiber/matrix interface rather than continuing through the fiber. In many early CMC systems, the weak interface between the fiber and the matrix contained carbon. However, the carbon layer is oxidized in oxygen-rich environments at temperatures as low as 400°C. After the carbon is gone, the oxygen reacts with the fiber to form a silica (SiO2) layer on the surface of the fiber.1–3 The SiO2 layer significantly weakens the fiber and also allows strong bonding to the matrix, which results in a significant decrease in fracture toughness for the CMC. As an alternative to carbon, boron nitride (BN)4–12 has been considered as a new fiber coating, and several CMC manufacturers have switched to a BN-containing interface between the fiber and the matrix. However, BN that is not fully crystalline (turbostratic with a d(002) spacing of >3.33 Å) is still susceptible to oxidation at temperatures $650°C.15 In addition, noncrystalline BN has been shown to be sensitive to moisture and oxidation.13 Crystalline BN has been deposited via chemical vapor deposition (CVD) only at deposition temperatures >1400°C;14 however, this high of a temperature damages the current ceramic fibers and the deposition rates are difficult to control. The oxidation process is accelerated whenever the applied tensile stress is high enough to cause matrix cracking in CMCs. Such damage-enhanced oxidation is typically the dominant degradation mechanism, which limits the hightemperature long-term behavior and performance of CMCs. The current investigation involves the evaluation of the durability improvements offered by CMCs, compared with metal parts, for a demanding military turbine-engine application. Advanced military aerospace turbine engines use an afterburner with vectoring nozzles to produce supersonic flight. Compared to the design life, the superalloy nozzle components have a greatly reduced service life. CMCs are being analyzed not only to meet the design life but also to advance the design life by a factor of 3. This increase in life is one of several advantages to using CMCs and is expected to offset an anticipated greater acquisition cost over the superalloy parts. The main focus of the investigation was to characterize the basic mechanical response as well as the long-term hightemperature performance and durability of the CMC system Nicalon™/SiNC. Furthermore, the components of interest are located at the end of the exhaust nozzle. At this location, they are exposed to rain and sodium (Na+) and other ions from ocean mists and exhaust deposits. In consideration of these conditions, the effect of salt-fog exposure on the CMC was considered and simulated using an environment-controlled chamber. The test conditions used in the investigation were chosen in such a way that the feasibility of using the present CMC for the high-temperature aerospace application could be efficiently evaluated. The investigation concerns the following performance and material characteristics: (i) the fiber/matrix interfacial strength, (ii) short-term mechanical response at room and high temperatures, (iii) the high-temperature long-term performance and durability, and (iv) resistance to environmental degradation. A. G. Evans—contributing editor Manuscript No. 192004. Received March 13, 1997; approved October 14, 1997. *Member, American Ceramic Society. † Author to whom correspondence should be addressed. Now with Structures Technology, High Speed Civil Transport, Boeing Commerical Airplane Group, Seattle, WA 98124–2207. J. Am. Ceram. Soc., 81 [7] 1797–811 (1998) Journal 1797
1798 Journal of the American Ceramic SocieryLee et al. Vol 81. No. 7 IL. Material (al The CMC system studied in this program was manufactured by Kaiser Ceramic Composites, which is a joint venture of Dow Corning(Midland, MI)and Kaiser Aerotech Engine Company(San Leandro, CA), and is sold under the tradename SylramicTM 202 Composite. For proprietary reasons, only a brief description of the CMC can be given. The silicon- carbon-oxygen(Si-C-O) fibers used in the composite were harness satin weave(8HSW)cloth. Individual sections of cloth were placed in a CVD furnace, and a boron-containing coating as applied to the fibers Once coated, a total of eight sections cloth were prepregged with a mixture of polymer, fillers, and solvent. The individual sections of prepregged cloth were then placed in a vacuum bag using a hand lay-up procedure and a yarp aligned stacking se e. Once in the bag. the was warm molded and cured in an autoclave to produce a flat 35 tile, Curing was followed by pyrolysis under nitrogen at tem atures>1000C. During pyrolysis, the polymer was pyro- d to an amorphous SiNC ceramic matrix. After pyrolysis, the assembly was reinfiltrated with polymer solution and then pyrolyzed again; this procedure was repeated several times to increase the matrix density. Test coupons were then machined from the tiles(done via laser), reinfiltrated with polymer solu ion, and pyrolyzed several times to densify the matrix further However, it was speculated that laser cutting might generate a elatively high temperature along the cutting edges and in crease the matrix density locally, which may diminish the ef- fectiveness of the above-stated procedures of increasing the matrix density after the coupons are machined from the tiles Optical micrographs of transverse cross sections that show the typical fiber distribution, porosity, and fillers are given in Fig. 1(a). Based on density measurements using immersion and helium-pycnometer techniques, the closed and op comprised-1% and 10% of the composite volume, respec tively. Be cause the porosity in this composite was predomi nantly open, this may suggest that either additional infiltrations with the polymeric precursor would have been advisable or the viscosity or molecular weight of the final precursor was too Much of the total porosity was associated with the large total and large pores in the 8 HSW Nicalon'Silc and filler and(b) res that are readily observable in Fig. 1 (b). The appearance of these large pores suggests that they formed during the in filtration cycles, with some evidence of shrinkage in the sur face layer of the pore during pyrolysis. Concentrating primarily OH), which is designed specifically for fiber interface testing. on the larger pores and cracks, which were readily distinguish The probe velocity during push-in was 12.7 um/min. The load olume fraction of 0.08(8% total porosity)was measured vie was recorded as a function of time. An applied stress was digital image analysis of the composite. This value compares calculated from the probe force and the cross-sectional area of easonably well with the above-referenced total each respective fiber. The areas were derived from the radii of each fiber measured immediately prior to its push-in. The raw 11%.Much of the discrepancy may be attributed to the finer displacement data also includes elastic components that re- rosity and cracks, which were not distinguished from the res of the matrix during the image-analysis work. The presence of sulted from the specimen and the test machine, in addition to large pores may present a problem when the CMC comes into debonding. During reduction of the data(prior to analysis of In addition, the inhomogeneous nature of matrix and the pres- ence of microcracks and fillers in the matrix makes it very ment were removed in a manner consistent with that proposed difficult to estimate the elastic properties of the matrix from by Jero et al. b The resulting displacement is called the corrected effective displacement(CED). The data was then measurements made on the CMC system analyzed using the continuous debonding model proposed by Kerans and parthasarath IlL. Experiments (2) Mechanical-Test Apparatus All the mechanical-behavior tests were conducted on a hori- O Fiber Push-In Test in the As-Received material zontal servohydraulic machine using water-cooled rigid hy Microanalysis of the fiber/matrix interface was conducted or draulic clamping and quartz-lamp heating. Test control s-received material using a fiber push-in technique. The push data acquisition, and interactive data-analysis functions we in probe used in these experiments had a ground conical dia provided by the MATE Programs installed on an IBM- mond tip with a 55 inclusion angle and a 10 um flat base. All compatible personal computer that was linked to the test frame the interface data was collected at room temperature using the via an analog-to-digital board. The MATE software was used Micro Measure Machine(Process Equipment Co., Tipp City for all testing in this investigation. Temperature was measured
II. Material The CMC system studied in this program was manufactured by Kaiser Ceramic Composites, which is a joint venture of Dow Corning (Midland, MI) and Kaiser Aerotech Engine Company (San Leandro, CA), and is sold under the tradename Sylramic™ 202 Composite. For proprietary reasons, only a brief description of the CMC can be given. The silicon– carbon–oxygen (Si-C-O) fibers used in the composite were ceramic-grade Nicalon™ fibers (Nippon Carbon Co., Tokyo, Japan). The fibers were supplied in the form of an eightharness satin weave (8HSW) cloth. Individual sections of cloth were placed in a CVD furnace, and a boron-containing coating was applied to the fibers. Once coated, a total of eight sections of cloth were prepregged with a mixture of polymer, fillers, and solvent. The individual sections of prepregged cloth were then placed in a vacuum bag using a hand lay-up procedure and a warp aligned stacking sequence. Once in the bag, the assembly was warm molded and cured in an autoclave to produce a flat tile. Curing was followed by pyrolysis under nitrogen at temperatures >1000°C. During pyrolysis, the polymer was pyrolyzed to an amorphous SiNC ceramic matrix. After pyrolysis, the assembly was reinfiltrated with polymer solution and then pyrolyzed again; this procedure was repeated several times to increase the matrix density. Test coupons were then machined from the tiles (done via laser), reinfiltrated with polymer solution, and pyrolyzed several times to densify the matrix further. However, it was speculated that laser cutting might generate a relatively high temperature along the cutting edges and increase the matrix density locally, which may diminish the effectiveness of the above-stated procedures of increasing the matrix density after the coupons are machined from the tiles. Optical micrographs of transverse cross sections that show the typical fiber distribution, porosity, and fillers are given in Fig. 1(a). Based on density measurements using immersion and helium-pycnometer techniques, the closed and open porosity comprised ∼1% and 10% of the composite volume, respectively. Because the porosity in this composite was predominantly open, this may suggest that either additional infiltrations with the polymeric precursor would have been advisable or the viscosity or molecular weight of the final precursor was too high to penetrate and fill the open pores effectively. Much of the total porosity was associated with the large pores that are readily observable in Fig. 1(b). The appearance of these large pores suggests that they formed during the infiltration cycles, with some evidence of shrinkage in the surface layer of the pore during pyrolysis. Concentrating primarily on the larger pores and cracks, which were readily distinguishable from a low-magnification optical image (200×), a pore volume fraction of 0.08 (8% total porosity) was measured via digital image analysis of the composite. This value compares reasonably well with the above-referenced total porosity of 11%. Much of the discrepancy may be attributed to the finer porosity and cracks, which were not distinguished from the rest of the matrix during the image-analysis work. The presence of large pores may present a problem when the CMC comes into direct contact with water, and this issue will be discussed later. In addition, the inhomogeneous nature of matrix and the presence of microcracks and fillers in the matrix makes it very difficult to estimate the elastic properties of the matrix from measurements made on the CMC system. III. Experiments (1) Fiber Push-In Test in the As-Received Material Microanalysis of the fiber/matrix interface was conducted on as-received material using a fiber push-in technique. The pushin probe used in these experiments had a ground conical diamond tip with a 55° inclusion angle and a 10 mm flat base. All the interface data was collected at room temperature using the Micro Measure Machine (Process Equipment Co., Tipp City, OH), which is designed specifically for fiber interface testing. The probe velocity during push-in was 12.7 mm/min. The load applied to the fiber during push-in, as well as the displacement, was recorded as a function of time. An applied stress was calculated from the probe force and the cross-sectional area of each respective fiber. The areas were derived from the radii of each fiber measured immediately prior to its push-in. The raw displacement data also includes elastic components that resulted from the specimen and the test machine, in addition to the displacement of the fiber relative to the matrix during debonding. During reduction of the data (prior to analysis of interfacial properties), the later contributions to the displacement were removed in a manner consistent with that proposed by Jero et al.16 The resulting displacement is called the corrected effective displacement (CED). The data was then analyzed using the continuous debonding model proposed by Kerans and Parthasarathy.17 (2) Mechanical-Test Apparatus All the mechanical-behavior tests were conducted on a horizontal servohydraulic machine using water-cooled rigid hydraulic clamping grips and quartz-lamp heating. Test control, data acquisition, and interactive data-analysis functions were provided by the MATE Program18 installed on an IBMcompatible personal computer that was linked to the test frame via an analog-to-digital board. The MATE software was used for all testing in this investigation. Temperature was measured Fig. 1. (a) Typical distributions of fiber, porosity, and filler and (b) total and large pores in the 8 HSW Nicalon/SiNC. 1798 Journal of the American Ceramic Society—Lee et al. Vol. 81, No. 7
July 1998 Mechanical Behavior and High-Temperature Performance of a Woven NicalonTMSi-N-C CMC 1799 by bonding five type thermocouples to specimen tested specimen. The temperature scatter between the dummy using an alumina-based ceramic adhesive. A descrip- specimen and tested specimen was measured to be -5"C Dur- ion of the test equipment and specimen design has been given ing the salt-fog interrupted 1000C fatigue test, the mechanical cycling and environmental exposure was applied alternately to In the high-temperature tension tests, each specimen was the test specimen. The 1000oC cycling was periodically inter heated to the test temperature in 15 min and then held there for rupted, and the specimen was removed from the test machine to -20 min to allow the specimen to equilibrate. When the 20 min the corrosion chamber for salt-fog exposure. Runout for the ak was completed, the load was applied. A high-temperature tested specimens included eight blocks of cycling at 1000C at extensometer with alumina rods was used to measure strain the selected stress level with seven interruptions of 24 h expo- All high-temperature tension, creep rupture, and fatigue tests sures to the salt-fog. The cycle count for each block was ob- followed this heating procedure tained from earlier 1000C fatigue tests without interruptions (A) Tension: Room-temperature, 1000C, and 1100C or salt-fo tension tests were conducted under stroke control with a con The specimen was initially fatigue tested at 1000C at a tant displacement rate of 0.05 mm/s. Residual-strength tests elected stress level for 5% of the fatigue life that was obtained were also conducted on all specimens that reached runout dur from the 1000.C fatigue test at the same applied stress level ing creep rupture and fatigue testing. These residual strength The specimen was then removed from the test frame and placed tests were conducted at room temperature using a displacement in the chamber for salt-fog exposure. Each salt-fog exposure te of 0.05 mm lasted for 24 h. After the salt-fog exposure the specimen was (B) Fatigue: Fatigue tests were conducted under load laced in a drying oven at 37 C for 12 h. The specimen was control with a stress ratio of 0.05. The maximum applied stress then placed back in the test machine for fatigue cycling. The levels for the fatigue tests were above or below the proportional entire procedure was repeated until the total accumulation of imits determined from the tension tests, to determine the role cycles in the tested specimen was 25% of the fatigue life. when of matrix cracking on room-temperature and high-temperature 25% of the fatigue life was reached in the tested specimen, the fatigue performance. The room-temperature tests were allowed cycle count for each fatigue block was increased to 25% of to operate for 10 cycles. The room-temperature tests were fatigue life to accelerate the test By the end of the eighth block cycled at a frequency of I Hz for the first 105 cycles and then of fatigue loading, the total accumulat on of cycle would be at 5 Hz for an additional 9 x 10 cycles or until failure 100% of the fatigue life and the test would be terminated at this The run-out condition for the 1000C fatigue tests was se- owever. the tested lected to approximately duplicate the projected number of he fatigue life was reached In the former case tension test wa cycles at maximum temperature for a desired design lifetime. conducted to measure residual tensile strength at room tem- The high-temperature fatigue tests were all conducted at a fre perature. Scanning electron microscopy(SEM) was used to quency of I with a run-out condition of 105 cycle characterize and document damage and failure modes in all During all fatigue tests, the load-versus-displacement trace cle was periodically recorded. The load-versus-displacement loops were converted to stress- ersus-strain loops. The stress-strain data collected between Results and discussions 5% and 15%of the stress range during loading Wa Aus-strain Fiber Push-In Test on As-Received Material calculate a modulus value. The area of the stress-versi was calculated and divided by the gauge length to gen The fiber/matrix interfacial properties were determined from hysteretic energy density(HED)values. The HED data is fiber push-in tests on a section cut and prepared from as- essentially hysteretic energy per volume of gauge section. This calculation produced an energy per unit volume and allowed a samples cut and prepared from fatigued and creep-tested speci- direct comparison of different specimens. All fatigue speci- mens mens, and those results will be discussed in a future paper mens that reached runout were tension tested at room tempera- Figure 2 is a compilation of applied stress and CED for sev- ture to determine residual mechanical properties enteen individual fiber push-in experiments. The scatter in the (C) Creep Rupture: Creep rupture tests were conducted data could be due to several factors. some of which occur under load control at 1000 and 1100%C. The run-out condition during composite processing, such as local variations in poros- was defined as 100 h and approximates the time at maximum temperature for the application using the increased design life The applied stress levels were selected to be both below and above the proportional limit determined from the high temperature tension tests. Creep strain was recorded from the instant the specimen reached the test stress level (D) Interrupted Fatigue Plus Salt-Fog Exposure: A cy clic corrosion chamber was used for salt-fog exposure in this investigation. The NaCl salt concentration was 0.05 wt% and was selected after detailed discussion with the engine manu- 3 facturers. Salt water has a salt concentration of.2 wt%.a E concentration of 5 wt%, which is typically used in corrosion studies on marine paints, I was initially considered as a worst case condition; however, the engine manufacturer suggested that the concentration be changed to that found at a routinely ed altitude. Atmospheric charts identified the salt concentra- tion at an altitude of 1000 ft to be only 0. 2 ppm, and this concentration would be overwhelmed by simple contamination during handling. Therefore, 0.05 wt% was finally selected to haps more realistically represent levels found near coastal cations Deionized water was mixed with salt to produce the CED (um) salt concentration. Temperature distributions on specimen surface were monitored through five therme Fig. 2. Stress versus corrected effective displacement(CED) for 17 placed on a dummy specimen that was positioned ne fibers obtained from the push-in test in the 8 HSw Nicalon/SiNC
by bonding five ‘‘S’’-type thermocouples to each specimen using an alumina-based ceramic adhesive. A detailed description of the test equipment and specimen design has been given elsewhere.19,20 In the high-temperature tension tests, each specimen was heated to the test temperature in 15 min and then held there for ∼20 min to allow the specimen to equilibrate. When the 20 min soak was completed, the load was applied. A high-temperature extensometer with alumina rods19 was used to measure strain. All high-temperature tension, creep rupture, and fatigue tests followed this heating procedure. (A) Tension: Room-temperature, 1000°C, and 1100°C tension tests were conducted under stroke control with a constant displacement rate of 0.05 mm/s. Residual-strength tests were also conducted on all specimens that reached runout during creep rupture and fatigue testing. These residual strength tests were conducted at room temperature using a displacement rate of 0.05 mm/s. (B) Fatigue: Fatigue tests were conducted under load control with a stress ratio of 0.05. The maximum applied stress levels for the fatigue tests were above or below the proportional limits determined from the tension tests, to determine the role of matrix cracking on room-temperature and high-temperature fatigue performance. The room-temperature tests were allowed to operate for 106 cycles. The room-temperature tests were cycled at a frequency of 1 Hz for the first 105 cycles and then at 5 Hz for an additional 9 × 105 cycles or until failure. The run-out condition for the 1000°C fatigue tests was selected to approximately duplicate the projected number of cycles at maximum temperature for a desired design lifetime. The high-temperature fatigue tests were all conducted at a frequency of 1 Hz, with a run-out condition of 105 cycles. During all fatigue tests, the load-versus-displacement trace for a complete fatigue cycle was periodically recorded. The load-versus-displacement loops were converted to stressversus-strain loops. The stress–strain data collected between 5% and 15% of the stress range during loading was used to calculate a modulus value. The area of the stress-versus-strain loops was calculated and divided by the gauge length to generate hysteretic energy density (HED) values. The HED data is essentially hysteretic energy per volume of gauge section. This calculation produced an energy per unit volume and allowed a direct comparison of different specimens. All fatigue specimens that reached runout were tension tested at room temperature to determine residual mechanical properties. (C) Creep Rupture: Creep rupture tests were conducted under load control at 1000° and 1100°C. The run-out condition was defined as 100 h and approximates the time at maximum temperature for the application using the increased design life. The applied stress levels were selected to be both below and above the proportional limit determined from the hightemperature tension tests. Creep strain was recorded from the instant the specimen reached the test stress level. (D) Interrupted Fatigue Plus Salt-Fog Exposure: A cyclic corrosion chamber was used for salt-fog exposure in this investigation. The NaCl salt concentration was 0.05 wt% and was selected after detailed discussion with the engine manufacturers. Salt water has a salt concentration of ∼3.2 wt%. A concentration of 5 wt%, which is typically used in corrosion studies on marine paints,21 was initially considered as a worstcase condition; however, the engine manufacturer suggested that the concentration be changed to that found at a routinely used altitude. Atmospheric charts identified the salt concentration at an altitude of 1000 ft to be only 0.2 ppm, and this concentration would be overwhelmed by simple contamination during handling. Therefore, 0.05 wt% was finally selected to perhaps more realistically represent levels found near coastal locations. Deionized water was mixed with salt to produce the desired salt concentration. Temperature distributions on the tested specimen surface were monitored through five thermocouples placed on a dummy specimen that was positioned next to the tested specimen. The temperature scatter between the dummy specimen and tested specimen was measured to be ∼5°C. During the salt-fog interrupted 1000°C fatigue test, the mechanical cycling and environmental exposure was applied alternately to the test specimen. The 1000°C cycling was periodically interrupted, and the specimen was removed from the test machine to the corrosion chamber for salt-fog exposure. Runout for the tested specimens included eight blocks of cycling at 1000°C at the selected stress level with seven interruptions of 24 h exposures to the salt-fog. The cycle count for each block was obtained from earlier 1000°C fatigue tests without interruptions for salt-fog. The specimen was initially fatigue tested at 1000°C at a selected stress level for 5% of the fatigue life that was obtained from the 1000°C fatigue test at the same applied stress level. The specimen was then removed from the test frame and placed in the chamber for salt-fog exposure. Each salt-fog exposure lasted for 24 h. After the salt-fog exposure, the specimen was placed in a drying oven at 37°C for 12 h. The specimen was then placed back in the test machine for fatigue cycling. The entire procedure was repeated until the total accumulation of cycles in the tested specimen was 25% of the fatigue life. When 25% of the fatigue life was reached in the tested specimen, the cycle count for each fatigue block was increased to 25% of fatigue life to accelerate the test. By the end of the eighth block of fatigue loading, the total accumulation of cycle would be 100% of the fatigue life and the test would be terminated at this point. However, the tested specimen may fail before 100% of the fatigue life was reached. In the former case, tension test was conducted to measure residual tensile strength at room temperature. Scanning electron microscopy (SEM) was used to characterize and document damage and failure modes in all failed specimens. IV. Results and Discussions (1) Fiber Push-In Test on As-Received Material The fiber/matrix interfacial properties were determined from fiber push-in tests on a section cut and prepared from asreceived material. The same analysis was also conducted on samples cut and prepared from fatigued and creep-tested specimens, and those results will be discussed in a future paper. Figure 2 is a compilation of applied stress and CED for seventeen individual fiber push-in experiments. The scatter in the data could be due to several factors, some of which occur during composite processing, such as local variations in porosFig. 2. Stress versus corrected effective displacement (CED) for 17 fibers obtained from the push-in test in the 8 HSW Nicalon/SiNC. July 1998 Mechanical Behavior and High-Temperature Performance of a Woven Nicalon™/Si-N-C CMC 1799
1800 Journal of the American Ceramic Sociery-Lee et al. Vol 81. No. 7 ty, filler concentrations or microcracks, curvature of the fiber ue to the woven architecture, or the close proximity of (or even contact with) adjacent fibers. There will also be exper 197 MPa mental errors and inaccuracies, which occur during the push-in test itself. Similar scatter was observed in the push-out data or BN-coated Nicalon fibers in a barium magnesium aluminosili te( BMAS)matrix and was attributed to variations in the interfacial microstructure l50 The Kerans and Parthasarathy 7 model was derived for a 2 unidirectional composite that consisted of a single fiber and a surrounding matrix of infinite radius. For the woven composite t hand, this was obviously not the case. The fibers chosen for oush-in were those for which no other fibers were in direct ontact, at least as far as this could be determined from the 50 lished surface. However there were generally adjacent fibers within one diameter band around the pushed fiber. As a con E s107 GPa sequence, these fibers contributed to the elastic properties of the matrix region around the pushed fiber. One of the require ments of the Kerans and Parthasarathy model is knowledge of the modulus of this surrounding material. Because this su Strain (% ounding material was not purely a matrix, an effective modu which includes some contribution from adjacent fibers would seem to be the most appropriate. Complicating things further is the fact that the modulus of the sinc matrix was o.= 21 MPa unknown. Because of the uncertainty of the matrix modulus two significantly different moduli were used in the model. The first value was based on the assu NC matrix is similar to that of Nicalon fibers(-200 GPa) 22 because the fibers and the matrix both were amorphous and similar in composition. A slightly lower value of 170 GPa was used for the matrix modulus, to account for the porosity. In the second case. an effective modulus was calculated for the matrix L using a simple rule-of-mixtures approach and the composite perature tension tests. In thi 75M approximation, the fibers oriented parallel to the load axis were stiffness of 107 GPa. which was derived from room DI GPa temperature tension tests. For this reason, only half of the total me fract contribute to the composite stiffness. The transverse fibers and 0.0 0.1 rosity were assumed to contribute to an effective matrix Strain 0.5V where E. and Er are the composite and fiber moduli, respec tively, and V is the total fiber volume fraction. The resulting effective matrix modulus was calculated to be 81 GPa The Kerans and Parthasarathy model 7 considers residual (c) F =205 MPR stresses and Poisson s expansion. In this model, the frictional shear stress, Tt, is based on the Coloumbic friction law, which is given by Tr WoN, where u is the friction coefficient and on is the normal clamping stress on the fiber For the assumed a matrix modulus of 170 GPa, the average values for H and oN 0. 10 and 125 MPa, respectively. Using the Em value of 81 GPa p derived as shown earlier from Ec, the corresponding values were 0.18 and 80 MPa. The interfacial shear stresses were calculated for each fiber test and then averaged the results 70 MPa were 9 and 10 MPa, for the high-and low-matrix moduli cases respectively. The standard deviations for these shear stresses ere 4.8 and 5.6 MPa, respectively, which indicates that the averages are indistinguishable. Therefore, the proper choice of matrix modulus turns out to be somewhat of a moot point. The interfacial friction coefficient H and the residual clamping stress on the fiber are sensitive to the modulus used for the Strain (%) matrix, whereas the interfacial sliding stress, which is the prod uct of the two, is less sensitive. In either case, the reasonably Fig 3.(a)room ature(23°C,(b)1000,and(c)1100°C ensile response of 8 HSW Nicalon/SINC low interfacial shear stress for the system should, by itself, be conducive to some fiber pullout (2) Monotonic Tension tively. The elastic modulus was 107 GPa, and the proportional limit was-85 MPa. The stress-versus-strain trace is extremely The room-te tensile stress-strain behavior is linear up to the proportional limit. At the proportional limit, shown in Fig. 3(a) average temperature tensile obvious nonlinear behavior caused by matrix cracking occurs strength and strain to e were 197 MPa and 0.33% over a short strain range, after which the development of non-
ity, filler concentrations or microcracks, curvature of the fibers due to the woven architecture, or the close proximity of (or even contact with) adjacent fibers. There will also be experimental errors and inaccuracies, which occur during the push-in test itself. Similar scatter was observed in the push-out data on BN-coated Nicalon fibers in a barium magnesium aluminosilicate11 (BMAS) matrix and was attributed to variations in the interfacial microstructure. The Kerans and Parthasarathy17 model was derived for a unidirectional composite that consisted of a single fiber and a surrounding matrix of infinite radius. For the woven composite at hand, this was obviously not the case. The fibers chosen for push-in were those for which no other fibers were in direct contact, at least as far as this could be determined from the polished surface. However, there were generally adjacent fibers within one diameter band around the pushed fiber. As a consequence, these fibers contributed to the elastic properties of the matrix region around the pushed fiber. One of the requirements of the Kerans and Parthasarathy model is knowledge of the modulus of this surrounding material. Because this surrounding material was not purely a matrix, an effective modulus, which includes some contribution from adjacent fibers, would seem to be the most appropriate. Complicating things further is the fact that the modulus of the SiNC matrix was unknown. Because of the uncertainty of the matrix modulus, two significantly different moduli were used in the model. The first value was based on the assumption that the modulus of the SiNC matrix is similar to that of Nicalon fibers (∼200 GPa),22 because the fibers and the matrix both were amorphous and similar in composition. A slightly lower value of 170 GPa was used for the matrix modulus, to account for the porosity. In the second case, an effective modulus was calculated for the matrix using a simple rule-of-mixtures approach and the composite modulus obtained from room-temperature tension tests. In this approximation, the fibers oriented parallel to the load axis were assumed to be the greatest single contribution to the composite stiffness of 107 GPa, which was derived from roomtemperature tension tests. For this reason, only half of the total volume fraction of fibers in the composite was assumed to contribute to the composite stiffness. The transverse fibers and porosity were assumed to contribute to an effective matrix modulus that is defined as Em 4 (Ec − 0.5Vf Ef )/(1 − 0.5Vf ), where Ec and Ef are the composite and fiber moduli, respectively, and Vf is the total fiber volume fraction. The resulting effective matrix modulus was calculated to be 81 GPa. The Kerans and Parthasarathy model17 considers residual stresses and Poisson’s expansion. In this model, the frictional shear stress, tf , is based on the Coloumbic friction law, which is given by tf 4 msN, where m is the friction coefficient and sN is the normal clamping stress on the fiber. For the assumed matrix modulus of 170 GPa, the average values for m and sN derived from the model for each of the 17 fibers tested were 0.10 and 125 MPa, respectively. Using the Em value of 81 GPa derived as shown earlier from Ec, the corresponding values were 0.18 and 80 MPa. The interfacial shear stresses were calculated for each fiber test and then averaged; the results were 9 and 10 MPa, for the high- and low-matrix moduli cases, respectively. The standard deviations for these shear stresses were 4.8 and 5.6 MPa, respectively, which indicates that the averages are indistinguishable. Therefore, the proper choice of matrix modulus turns out to be somewhat of a moot point. The interfacial friction coefficient m and the residual clamping stress on the fiber are sensitive to the modulus used for the matrix, whereas the interfacial sliding stress, which is the product of the two, is less sensitive. In either case, the reasonably low interfacial shear stress for the system should, by itself, be conducive to some fiber pullout. (2) Monotonic Tension The room-temperature tensile stress–strain behavior is shown in Fig. 3(a). The average room-temperature tensile strength and strain to failure were 197 MPa and 0.33%, respectively. The elastic modulus was 107 GPa, and the proportional limit was ∼85 MPa. The stress-versus-strain trace is extremely linear up to the proportional limit. At the proportional limit, obvious nonlinear behavior caused by matrix cracking occurs over a short strain range, after which the development of nonFig. 3. (a) Room-temperature (23°C), (b) 1000°C, and (c) 1100°C tensile response of 8 HSW Nicalon/SiNC. 1800 Journal of the American Ceramic Society—Lee et al. Vol. 81, No. 7
July 1998 Mechanical Behavior and High-Temperature Performance of a Woven NicalonTMSi-N-C CMC 1801 Table L. Average Tensile Properties Obtained at Room Temperature and at gh Temperatu Proportional limit Temperature E (GPa) 0.36 1100° 8万0 214 0.42 205 0.4 linear behavior seems to slow down. and the behavior is almost he system resistance to environmental oxidation may be weak- linear before failure ened at lower temperatures, however, the limited amount of The 1000 and 1100C tensile stress-strain responses both data restricts making any conclusions at this time. The run-out were similar in shape to the room-temperature tensile response, stress level is substantially above (-35 MPa) the proportional as shown in Figs. 3(b)and(c), respectively. The averaged limit that has been identified in the high-temperature tension tensile properties obtained from all three temperatures are tests; this observation is a very important material-performance listed in Table I. There are minor differences between the characteristic. because it shows that. even with extensive ma- room-temperature and high-temperature tensile properties. trix cracks, this composite exhibits remarkable creep life However, statistically significant variations in these properties, Creep-strain-versus-time traces for all the creep-rupture tests as a function of temperature, could not be determined, because are shown in Figs. 6(a)and(b). For all tests, the creep strain of the limited number of specimens. Nevertheless, the test re- was <0.15%. This small amount of creep strain is typically cults suggest that the tensile properties of this CMC are stable observed for Nicalon-containing CMCs of the same fiber ar- after short-term exposure at temperatures up to 1100oC chitecture tested at 1000oC. As the applied stress level in- ,The typical fracture morphology of roo temperature and creases, the strain to failure remains relatively constant at significant differences in the failure modes were observed portant point is that all the specimens experienced very little the examined samples. In general, a ragged fracture surface ith widespread fiber pullout but relatively short pull-out the primary damage mechanism is not creep damage but envi- ngths was observed in all samples. These results are consis- ronmental attack of the fiber/matrix inte nt with the fiber push-in result, which demonstrated that fiber fiber itself, or both. Another important observation fro allout should occur. The similarity of fracture characteristics 6(a)and(b)is that, for the lower stress levels, the rate indicates that short-term high-temperature exposure does not accumulation occurs rapidly during the first 10-25 h and ther significantly change the failure process; this observation is con- sistent with the similarities in the tensile properties listed in The observation of widespread fiber pullout is consiste ith the fiber push-in results In the latter case, the magnitude of the average fiber/matrix interfacial shear stress predicts that fiber pullout should occur. The wide variation in pullout lengths is also consistent with the push-in results. The large standard deviation associated with the calculation of the inter- facial shear stress suggests that some fibers will be more con- ducive to pullout during a tension test than will others. Thou less et al. 3 reported that the distribution in fiber pull-out matrix interfacial sliding stress, fiber diameter, and Weibull modulus of the fiber. Among these factors, the fiber/matrix interfacial sliding stress is easily influenced by the homogene ity of the surrounding material. For instance, inhomogeneou matrix properties, which may occur due to local density and oppositional variations, could result in a fiber-to-fiber varia- A tion of the stress field that exists along the fiber/matrix inter- faces. This phenomenon could, in turn, lead to scatter in the fiber-to-fiber pullout(push-in)response. The local stress field E around the fibers be further complicated by the woven (b) fiber architecture, which may also have a strong influence on the fiber pull-out length ( Creep Rupture Creep-rupture life versus applied stress, for the 1000 and 1100.C test conditions, are summarized in Fig. 5. No signifi- cant differences are observed in the test data between these two test temperatures. The run-out stress level in each curve is 110 MPa. In addition, the data for both temperatures essentially fall on top of each other. Although this twin behavior may suggest that the same damage mechanism is operating at both test temperatures, it is important to note that failure typically oc- urred outside of the gauge section of the extensometer. Tal lI summarizes the test conditions failure locations as measured from the center of the specimen, and the approximate tempera- ture at the failure location. In reviewing this information, the Fig 4. Fracture surface of 8 HSW Nicalon/SiNC tension tested at(a) failure location and corresponding temperature suggests that 23. and(b)1100C( Bar in each figure represents 1.00 mm
linear behavior seems to slow down, and the behavior is almost linear before failure. The 1000° and 1100°C tensile stress–strain responses both were similar in shape to the room-temperature tensile response, as shown in Figs. 3(b) and (c), respectively. The averaged tensile properties obtained from all three temperatures are listed in Table I. There are minor differences between the room-temperature and high-temperature tensile properties. However, statistically significant variations in these properties, as a function of temperature, could not be determined, because of the limited number of specimens. Nevertheless, the test results suggest that the tensile properties of this CMC are stable after short-term exposure at temperatures up to 1100°C. The typical fracture morphology of room-temperature and 1100°C tested specimens are shown in Figs. 4(a) and (b). No significant differences in the failure modes were observed in the examined samples. In general, a ragged fracture surface with widespread fiber pullout but relatively short pull-out lengths was observed in all samples. These results are consistent with the fiber push-in result, which demonstrated that fiber pullout should occur. The similarity of fracture characteristics indicates that short-term high-temperature exposure does not significantly change the failure process; this observation is consistent with the similarities in the tensile properties listed in Table I. The observation of widespread fiber pullout is consistent with the fiber push-in results. In the latter case, the magnitude of the average fiber/matrix interfacial shear stress predicts that fiber pullout should occur. The wide variation in pullout lengths is also consistent with the push-in results. The large standard deviation associated with the calculation of the interfacial shear stress suggests that some fibers will be more conducive to pullout during a tension test than will others. Thouless et al.23 reported that the distribution in fiber pull-out lengths in unidirectional CMCs should be related to the fiber/ matrix interfacial sliding stress, fiber diameter, and Weibull modulus of the fiber. Among these factors, the fiber/matrix interfacial sliding stress is easily influenced by the homogeneity of the surrounding material. For instance, inhomogeneous matrix properties, which may occur due to local density and compositional variations, could result in a fiber-to-fiber variation of the stress field that exists along the fiber/matrix interfaces. This phenomenon could, in turn, lead to scatter in the fiber-to-fiber pullout (push-in) response. The local stress field around the fibers may be further complicated by the woven fiber architecture,24–30 which may also have a strong influence on the fiber pull-out length.30 (3) Creep Rupture Creep-rupture life versus applied stress, for the 1000° and 1100°C test conditions, are summarized in Fig. 5. No significant differences are observed in the test data between these two test temperatures. The run-out stress level in each curve is 110 MPa. In addition, the data for both temperatures essentially fall on top of each other. Although this twin behavior may suggest that the same damage mechanism is operating at both test temperatures, it is important to note that failure typically occurred outside of the gauge section of the extensometer. Table II summarizes the test conditions, failure locations as measured from the center of the specimen, and the approximate temperature at the failure location. In reviewing this information, the failure location and corresponding temperature suggests that the system resistance to environmental oxidation may be weakened at lower temperatures; however, the limited amount of data restricts making any conclusions at this time. The run-out stress level is substantially above (∼35 MPa) the proportional limit that has been identified in the high-temperature tension tests; this observation is a very important material-performance characteristic, because it shows that, even with extensive matrix cracks, this composite exhibits remarkable creep life. Creep-strain-versus-time traces for all the creep-rupture tests are shown in Figs. 6(a) and (b). For all tests, the creep strain was <0.15%. This small amount of creep strain is typically observed for Nicalon-containing CMCs of the same fiber architecture tested at 1000°C. As the applied stress level increases, the strain to failure remains relatively constant at 1000°C and increases only slightly at 1100°C. The most important point is that all the specimens experienced very little strain accumulation. This observation highlights the fact that the primary damage mechanism is not creep damage but environmental attack of the fiber/matrix interphase, the Nicalon fiber itself, or both. Another important observation from Figs. 6(a) and (b) is that, for the lower stress levels, the rate of strain accumulation occurs rapidly during the first 10–25 h and then Table I. Average Tensile Properties Obtained at Room Temperature and at High Temperatures Temperature Elastic modulus, Exx (GPa) Proportional limit (MPa) Ultimate tensile strength (MPa) Strain to failure (%) Room temperature 107 85 197 0.36 1100°C 101 75 214 0.42 1200°C 95 70 205 0.4 Fig. 4. Fracture surface of 8 HSW Nicalon/SiNC tension tested at (a) 23° and (b) 1100°C. (Bar in each figure represents 1.00 mm.) July 1998 Mechanical Behavior and High-Temperature Performance of a Woven Nicalon™/Si-N-C CMC 1801