ournal 1. Am Ceram Soc, s1 [8]2077-86 (1998) Processing and Performance of an All-Oxide Ceramic Composite Carlos G. Levi, James Y. Yang, Brian J Dalgleish, Frank W. Zok, and Anthony G. Evans"T High Performance Composites Center, Departments of Materials and Mechanical E University of California, Santa Barbara, California 93160-5050 Continuous fiber ceramic composites(CFCCs) based on rous interlayers as crack deflection paths,0 and extends the oxides are of interest for high-temperature applications ow concept to utilize a porous matrix as a surrogate. The concept ing to their inherent oxidative stability. An enabling ele- has been successfully demonstrated, 14, I8 but only limited in- ment is a matrix with an optimum combination of tough formation is available in the open literature. The design and ness and strength, which may be achieved by incorporating stability of the matrix microstructure are arguably more critical a controlled amount of fine, well-distributed porosity aken to explore both a concept for a stable, porous oxide mas in the latter approach. The present investigation was und Implementation of this concept by vacuum infiltration of aqueous mullite-alumina slurries into two-dimensional wo- and the mechanical performance of the resulting comy fhe ven preforms of alumina fibers has been investigated The information is organized in the following manner Evaluation of these materials shows stress-strain chara derlying microstructural design concept for the matrix is teristics similar to other CFCCs, especially carbon-matrix elaborated, followed by an identification of the materials to be composites. Moreover, promising notch and creep proper- used. Subsequently, key elements of the processing science and ties have been found. Microstructural and processing issues technology are addressed. These include a specification of the relevant to the attainment of these behaviors are discussed manufacturing sequence, a sintering study that guides the choice of materials to be used for the matrix, as well as a . Introduction characterization of the composite microstructure. Thereafter, Ca ONTINUoUS fiber ceramic composites(CFCCs) with suit several essential thermomechanical properties are measured ably tailored interfaces can exhibit inelastic deformation and analyzed. Initially, the capacity of the composite to exhibit haracteristics, which enable the composites to retain streng inelastic deformation is determined in the0°/90°and±45° in the presence of holes and notches. They also render them orientations. Such results reveal that these composites exhibit amenable to design and life prediction strategies developed for damage tolerance comparable to other fiber-dominated CFCCs metals. This damage tolerance, coupled with their inherent re Moreover, the effects of thermal exposure on the 0/90 tensile fractoriness, has enabled CFCCs to emerge as candidates for rties are determined in order to characterize fiber de dation effects that might arise either during manufacture or in icular interest is their use in combustors wherein the ability service. Finally, some preliminary results are obtained concern- to operate at high temperatures with reduced need for cool air can yield substantial benefits in efficiency and also provide creep strength and notch performance, which relate to desigr control of deleterious emissions. such as No. .4 and lifing issues Most CFCC systems are based on SiC fibers, with either The general conclusion reached is that these materials have oxide or non-oxide matrices, and interphases consisting of vari- echanical characteristics comparable to those established for ous combinations of carbon, BN, and SiC. The interphas carbon-matrix materials(such as SiC/carbon and carbon/ tailored to enable interfacial debonding and crack bridging to arbon), with attendant implications for thermostructural per- occur upon matrix cracking accompanied by internal fric ormance. The key differences with the carbon-matrix materi- tion,However, these systems are susceptible to embrittle als are their superior oxidative stability and their differing nent by oxygen ingress through the matrix cracks, followed by creep response eaction with the interphase and the fibers. 6 -The kinetics ar particularly debilitating at intermediate temperatures(500 IL. Microstructural Design 900 C)and upon cyclic loading. 7,9 The embrittlement problem The microstructure must be designed to have a sufficientl y imposes major design limitations by requiring that the stresses low toughness to enable crack deflection through the matrix remain below the matrix cracking stress. This deficiency has while maintaining enough strength for adequate off-axis and stable oxide constituents 10-18 e Development of all-oxide composites has followed two dis- amount of fine, uniformly distribute a rporating a controlled ct microstructural design paths. The first is based weak interface concept, typical of most CFCCs. It uses eithe matrix performance dictates a stable and well-bonded particle fugitive layers or stable oxide interphases with suitably low network with substantial vold space, of order%, fracture toughness. 7 The second implicitly accepts the forma- comparable to the interparticle spacing. Fine matrix tion of strong interfaces. It builds on the experience with po- form. as well as the nominal strength of the matrix. However fine particles also reduce the stability of the matrix against densification during processing and service, promoting the B. N. Cox--contributing editor evolution of undesirable flaws under the constraint imposed by he fibers 22,23 Manuscript No. 191156. Received March 7 pproved October 6 Q Mullite emerges as an attractive matrix material, owing to its cellent creep resistance, low modulus, and, as noted beloy generally sluggish sintering kinetics below -1300oC. The sin- monitored by Dr.S. G. Fishman of the Office tering kinetics suggests adequate microstructural stability for applications in the gas-turbine engine, where initial target wall Now with the Division of Applied Sciences, Harvard University, Cambridge, MA. temperatures are in the range -1000o to -1200C. However
Processing and Performance of an All-Oxide Ceramic Composite Carlos G. Levi,* James Y. Yang, Brian J. Dalgleish, Frank W. Zok,* and Anthony G. Evans*,† High Performance Composites Center, Departments of Materials and Mechanical Engineering, University of California, Santa Barbara, California 93160–5050 Continuous fiber ceramic composites (CFCCs) based on oxides are of interest for high-temperature applications owing to their inherent oxidative stability. An enabling element is a matrix with an optimum combination of toughness and strength, which may be achieved by incorporating a controlled amount of fine, well-distributed porosity. Implementation of this concept by vacuum infiltration of aqueous mullite–alumina slurries into two-dimensional woven preforms of alumina fibers has been investigated. Evaluation of these materials shows stress–strain characteristics similar to other CFCCs, especially carbon-matrix composites. Moreover, promising notch and creep properties have been found. Microstructural and processing issues relevant to the attainment of these behaviors are discussed. I. Introduction CONTINUOUS fiber ceramic composites (CFCCs) with suitably tailored interfaces can exhibit inelastic deformation characteristics, which enable the composites to retain strength in the presence of holes and notches.1 They also render them amenable to design and life prediction strategies developed for metals. This damage tolerance, coupled with their inherent refractoriness, has enabled CFCCs to emerge as candidates for many high-temperature thermostructural applications.2 Of particular interest is their use in combustors,3 wherein the ability to operate at high temperatures with reduced need for cooling air can yield substantial benefits in efficiency and also provide control of deleterious emissions, such as NOx. 4 Most CFCC systems are based on SiC fibers, with either oxide or non-oxide matrices, and interphases consisting of various combinations of carbon, BN, and SiC. The interphases are tailored to enable interfacial debonding and crack bridging to occur upon matrix cracking accompanied by internal friction.1,5 However, these systems are susceptible to embrittlement by oxygen ingress through the matrix cracks, followed by reaction with the interphase and the fibers.6–9 The kinetics are particularly debilitating at intermediate temperatures (500°– 900°C) and upon cyclic loading.7,9 The embrittlement problem imposes major design limitations by requiring that the stresses remain below the matrix cracking stress. This deficiency has motivated the search for CFCCs based on environmentally stable oxide constituents.10–18 Development of all-oxide composites has followed two distinct microstructural design paths. The first is based on the weak interface concept, typical of most CFCCs. It uses either fugitive layers12 or stable oxide interphases with suitably low fracture toughness.17 The second implicitly accepts the formation of strong interfaces. It builds on the experience with porous interlayers as crack deflection paths19,20 and extends the concept to utilize a porous matrix as a surrogate. The concept has been successfully demonstrated,11,14,18 but only limited information is available in the open literature. The design and stability of the matrix microstructure are arguably more critical in the latter approach. The present investigation was undertaken to explore both a concept for a stable, porous oxide matrix and the mechanical performance of the resulting composites. The information is organized in the following manner. The underlying microstructural design concept for the matrix is elaborated, followed by an identification of the materials to be used. Subsequently, key elements of the processing science and technology are addressed. These include a specification of the manufacturing sequence, a sintering study that guides the choice of materials to be used for the matrix, as well as a characterization of the composite microstructure. Thereafter, several essential thermomechanical properties are measured and analyzed. Initially, the capacity of the composite to exhibit inelastic deformation is determined in the 0°/90° and ±45° orientations. Such results reveal that these composites exhibit damage tolerance comparable to other fiber-dominated CFCCs. Moreover, the effects of thermal exposure on the 0°/90° tensile properties are determined in order to characterize fiber degradation effects that might arise either during manufacture or in service. Finally, some preliminary results are obtained concerning the interlaminar shear properties as well as the in-plane creep strength and notch performance, which relate to design and lifing issues. The general conclusion reached is that these materials have mechanical characteristics comparable to those established for carbon-matrix materials (such as SiC/carbon and carbon/ carbon), with attendant implications for thermostructural performance. The key differences with the carbon-matrix materials are their superior oxidative stability and their differing creep response. II. Microstructural Design The microstructure must be designed to have a sufficiently low toughness to enable crack deflection through the matrix while maintaining enough strength for adequate off-axis and interlaminar properties.18 These seemingly contradictory requirements are achievable by incorporating a controlled amount of fine, uniformly distributed porosity.18,21 Acceptable matrix performance dictates a stable and well-bonded particle network with substantial void space, of order ∼30%, on a scale comparable to the interparticle spacing. Fine matrix particles enhance packing density and uniformity within the fiber preform, as well as the nominal strength of the matrix. However, fine particles also reduce the stability of the matrix against densification during processing and service, promoting the evolution of undesirable flaws under the constraint imposed by the fibers.22,23 Mullite emerges as an attractive matrix material, owing to its excellent creep resistance, low modulus, and, as noted below, generally sluggish sintering kinetics below ∼1300°C. The sintering kinetics suggests adequate microstructural stability for applications in the gas-turbine engine, where initial target wall temperatures are in the range ∼1000° to ∼1200°C. However, B. N. Cox—contributing editor Manuscript No. 191156. Received March 7, 1997; approved October 6, 1997. Supported by the Defense Advanced Research Projects Agency under University Research Initiative Grant N00014-92-J-1808. Supervised by Dr. W. Coblenz and monitored by Dr. S. G. Fishman of the Office of Naval Research. *Member, American Ceramic Society. † Now with the Division of Applied Sciences, Harvard University, Cambridge, MA. J. Am. Ceram. Soc., 81 [8] 2077–86 (1998) Journal 2077
2078 Journal of the American Ceramic Society-Levi et al. Vol 81. No 8 the sluggish sintering kinetics presents a challenge in process- scopic technique to monitor the fiber stresses. 2 For the N610 ing. That is, temperatures above -1300oC are required to fibers, these measurements yield a dry bundle strength of ob achieve the requisite bonding between the matrix particles, yet 2.6 GPa and a corresponding failure strain of -0.7%. 9The nost commercial oxide fibers are susceptible to microstructural measured bundle strength is essentially identical to the value degradation at these temperatures. 16, 24 Ideally, neck formation calculated on the basis of the monofilament data(2.3=obs vith minimal shrink y operating 2.7 GPa), indicating negligible degradation in fiber streng egimes dominated by surface and vapor transport mecha- upon weaving. In contrast, similar measurements on the N720 nisms.2>However, this approach has not provided the requisite fibers reveal significant degradation. Notably, the strength and strengths. Liquid precursor impregnation and pyrolysis provide failure strain of the extracted tows are.9 GPa and.3%, another avenue to build the interparticle necks, but the initial respectively, approximately half of the calculated values for the matrix must have sufficient strength to withstand handling. 4 pristine(unwoven) fibers(-1.6 GPa and.7%). The proper The proposed matrix design concept is depicted in Fig. I ties obtained on the extracted tows are thus the appropriate Relatively large(I um)mullite particles are packed between baseline on which the properties of the CFCCs must be as- vs to form a touching ffects associated with composite Alumina particles that fit within the void spaces of this network processing identified. Indeed, degradation effects associate 200 nm) are added in a proportion limited primarily by the with high-temperature exposure have been reported for earlier requisite levels of porosity. Since submicrometer alumina sin- generations of the N610 fibers, with dry bundle strengths di ters readily above 800C, 2 the fine particles form bridges be- ishing from -1. 6 to -1. 1 GPa after 4 h at 1200%C, and to tween the larger mullite particles, as well as between the mul 0.97 GPa after I h at 1300C, and the accompanying failure lite particles and the fibers, at processing temperatures that minimize fiber degradation. Interparticle voids may open lo- trained from shrinking by the rigid mullite network. The ma- cused primarily on nzA ed temperature properties have fo trix is further strengthened by adding material to the alumina The mullite powder selected is MU-107 (Showa Denko K bridges using precursor impregnation and pyrolysis, as dis Tokyo, Japan) with mean particle size of -l um and BET surface area of 7.5 m2/g. The chemical analysis shows 75.5% 12O3 and 24% Sio,(by weight), with only trace amounts of Ill. Implementation TiO,, Fe,O3, and Na2O. The particle-size distribution ranges from <0.5 to >2.5 um, which is rather broad from a micro- O Materials structural design perspective(Fig. 1), but offers the advantage enforcements used in this investigation are Nextel 610TM of promoting reasonable packing densities. The alumina pow- and 720TM(N610 and N720, 3M Corp, Minneapolis, MN der is AKP-50(Sumitomo Chemicals, Tokyo, Japan) with a fibers woven in eight-harness satin fabric(float length 6.5 mean particle size of.2 um(0. 1-0.3 um)and BET surface mm). The tows in the fabric contain -400 filaments with di area of 10.6 m /g. Its chemistry is essentially pure(99.995%) meters of 10-12 um. The tow denier is 1500 and the yield is a-Al2O3. The precursor used to strengthen the matrix is a so- 6000 m/kg. The aspect ratio of the tow cross section is-10 lution of aluminum hydroxyl chloride, Al2CI(OH)s, which N610 fibers are essentially pure (99%)polycrystalline a- yields 8. 2% of Al2O3 by weight Al,O3, 26,27 whereas N720 fibers consist of a mixture of fine d (2) Manufacturing Process strength and stability against coarsening relative to the N610 The geometry of interest is a flat panel, 200 mm x 125 mm x 3 mm, with a two-dimensional(2-D) laminate architecture. The The pristine, unwoven N610 and N720 fibers are reported to reinforcement content, f, is-36t 4 vol%, comparable to that of have mean filament strengths of-3 and-2. 1 GPa, respectively, other fabric CFCCs. The fabric is arranged in two different for a 25 mm gauge ength, with a similar Weibull shape pa- orientations: one with the fibers parallel to the panel edges ( designated0990°) and the other rotated by±45°. The pro and 260 GPa for N610 and N720, respectively2728odulus(380 cessing route is summarized in the flow chart of Fig.2.A anied by a similar difference in y matrix slurr aning 30% solids is prepared by mixing the ated mean failure strain, Er, for pristine fibers is -0.8% in mullite and alumina powders in the desired proportions into both cases. However, some degradation may be expected to deionized water, using HNO, to adjust the pH to-3. The HNO3 g. Such effects have been probed by allows repulsive interactions to develop between the oxide par- performing fiber bundle tests on tows extracted from the as- ticles, providing the appropriate rheology to facilitate their received fabric, 29 using a recently developed piezo-spectro- flow into the fiber preform and their subsequent packing. 30 Fibe structural design of all-oxide continuous fiber ceramic composite Matrix consists etwork of mullite particles bonded to the fibers and among particles and ed by precursor impregnation and pyrolysis
the sluggish sintering kinetics presents a challenge in processing. That is, temperatures above ∼1300°C are required to achieve the requisite bonding between the matrix particles, yet most commercial oxide fibers are susceptible to microstructural degradation at these temperatures.16,24 Ideally, neck formation with minimal shrinkage should be promoted by operating in regimes dominated by surface and vapor transport mechanisms.25 However, this approach has not provided the requisite strengths. Liquid precursor impregnation and pyrolysis provide another avenue to build the interparticle necks, but the initial matrix must have sufficient strength to withstand handling.14 The proposed matrix design concept is depicted in Fig. 1. Relatively large (∼1 mm) mullite particles are packed between and within tows to form a touching, nonshrinking network. Alumina particles that fit within the void spaces of this network (∼200 nm) are added in a proportion limited primarily by the requisite levels of porosity. Since submicrometer alumina sinters readily above 800°C,22 the fine particles form bridges between the larger mullite particles, as well as between the mullite particles and the fibers, at processing temperatures that minimize fiber degradation. Interparticle voids may open locally owing to the sintering, but the overall matrix is constrained from shrinking by the rigid mullite network. The matrix is further strengthened by adding material to the alumina bridges using precursor impregnation and pyrolysis, as discussed below. III. Implementation (1) Materials Reinforcements used in this investigation are Nextel 610™ and 720™ (N610 and N720, 3M Corp., Minneapolis, MN) fibers woven in eight-harness satin fabric (float length 6.5 mm). The tows in the fabric contain ∼400 filaments with diameters of 10–12 mm. The tow denier is 1500 and the yield is 6000 m/kg. The aspect ratio of the tow cross section is ∼10. N610 fibers are essentially pure (>99%) polycrystalline aAl2O3, 26,27 whereas N720 fibers consist of a mixture of finegrained mullite and alumina that exhibits improved creep strength and stability against coarsening relative to the N610 fiber.28 The pristine, unwoven N610 and N720 fibers are reported to have mean filament strengths of ∼3 and ∼2.1 GPa, respectively, for a 25 mm gauge length, with a similar Weibull shape parameter, m ≈ 10.27,28 Since the difference in strengths is accompanied by a similar difference in Young’s modulus (380 and 260 GPa for N610 and N720, respectively27,28), the anticipated mean failure strain, «f , for pristine fibers is ∼0.8% in both cases. However, some degradation may be expected to take place upon weaving. Such effects have been probed by performing fiber bundle tests on tows extracted from the asreceived fabric,29 using a recently developed piezo-spectroscopic technique to monitor the fiber stresses.24 For the N610 fibers, these measurements yield a dry bundle strength of sb ≈ 2.6 GPa and a corresponding failure strain of ∼0.7%.29 The measured bundle strength is essentially identical to the value calculated on the basis of the monofilament data (2.3 # sb # 2.7 GPa), indicating negligible degradation in fiber strength upon weaving. In contrast, similar measurements on the N720 fibers reveal significant degradation. Notably, the strength and failure strain of the extracted tows are ∼0.9 GPa and ∼0.3%,29 respectively, approximately half of the calculated values for the pristine (unwoven) fibers (∼1.6 GPa and ∼0.7%). The properties obtained on the extracted tows are thus the appropriate baseline on which the properties of the CFCCs must be assessed and the degradation effects associated with composite processing identified. Indeed, degradation effects associated with high-temperature exposure have been reported for earlier generations of the N610 fibers, with dry bundle strengths diminishing from ∼1.6 to ∼1.1 GPa after 4 h at 1200°C, and to 0.97 GPa after 1 h at 1300°C, and the accompanying failure strains dropping to ∼0.3%. Because of this susceptibility to strength degradation and the lower creep strength of the N610 fiber, studies on the elevated temperature properties have focused primarily on N720. The mullite powder selected is MU-107 (Showa Denko KK, Tokyo, Japan) with mean particle size of ∼1 mm and BET surface area of 7.5 m2 /g. The chemical analysis shows 75.5% Al2O3 and 24% SiO2 (by weight), with only trace amounts of TiO2, Fe2O3, and Na2O. The particle-size distribution ranges from <0.5 to >2.5 mm, which is rather broad from a microstructural design perspective (Fig. 1), but offers the advantage of promoting reasonable packing densities. The alumina powder is AKP-50 (Sumitomo Chemicals, Tokyo, Japan) with a mean particle size of ∼0.2 mm (0.1–0.3 mm) and BET surface area of 10.6 m2 /g. Its chemistry is essentially pure (99.995%) a-Al2O3. The precursor used to strengthen the matrix is a solution of aluminum hydroxyl chloride, Al2Cl(OH)5, which yields 8.2% of Al2O3 by weight. (2) Manufacturing Process The geometry of interest is a flat panel, 200 mm × 125 mm × 3 mm, with a two-dimensional (2-D) laminate architecture. The reinforcement content, f, is ∼36 ± 4 vol%, comparable to that of other fabric CFCCs. The fabric is arranged in two different orientations: one with the fibers parallel to the panel edges (designated 0°/90°) and the other rotated by ±45°. The processing route is summarized in the flow chart of Fig. 2. A matrix slurry containing ∼30% solids is prepared by mixing the mullite and alumina powders in the desired proportions into deionized water, using HNO3 to adjust the pH to ∼3. The HNO3 allows repulsive interactions to develop between the oxide particles, providing the appropriate rheology to facilitate their flow into the fiber preform and their subsequent packing.30 Fig. 1. Microstructural design of all-oxide continuous-fiber ceramic composite. Fibers are typically Nextel 610 or 720 from 3M. Matrix consists of a continuous network of mullite particles bonded to the fibers and among themselves by bridges consisting of smaller alumina particles and alumina produced by precursor impregnation and pyrolysis. 2078 Journal of the American Ceramic Society—Levi et al. Vol. 81, No. 8
August 1998 Processing and Performance of an Al-Oxide Ceramic Composite 2079 e-sizing H Cutting 120. D) Lay-up 1000°c Matrix Bond Treatment (90 sdg Drying Pyrolysis Treatment (1200C) Product Fig. 2. Processing approach used for the fabrication of damage- Volume Fraction of Alumina in Matrix(%) tolerant all-oxide CFCC panels without fiber coatings Composites with 3-D preforms can be fabricated in the same manner Fig. 3. Evolution of porosity during heat treatment as a function of the alumina content in unreinforced mullite-alumina compacts pro- duced by vacuum filtration under conditions similar to those used for Dispersion of soft agglomerates is promoted by a combination he composites. The average particle sizes are -l um for mullite and hanical and ultrasonic agitation 2 um for alumina. Details of the heat treatment are given in the cloth is cut into 200 mm x 125 mm pieces having the desired orientation(0°90°or±45°). The pieces are stacked in a loose preform and de-sized by heating to 550oC for 0.5 h. The 5C/min (The slow temperature ramping was aimed at mini- fiber content of the cloth is 9.3 mm/cm2, thus a 3 mm plate mizing the risk of cracking or chipping, which would impair with a fiber volume fraction of-36% requires-12 layers. After the accuracy of the measurements. Examination of these de-sizing,the fiber preform is placed between two perforated samples revealed no microcracking at any stage. )The packing stainless-steel plates that constrain the cloth within the desired densities were measured after each step following the ASTM placed in the middle of a chambe cess. The assembly is then C20-92 standard, t as were the external dimensions of the thickness during the filtration pre ation through both top and bottom. An amount of slurry cal- The results are summarized in Fig 3. The differing behav culated to fill the preform is poured on top, whereupon the iors of the pure-alumina and pure-mullite compacts lend sup- chamber is closed and evacuated from both top and bottom to port to the proposed microstructural design concept. The po- eliminate air bubbles trapped within the preform. The to rosites for the as-consolidated(wet) compacts are estimated ion of the chamber is then open to the atmosphere while from the packing densities after drying at 120%C and the cor- vacuum is maintained in the bottom to drive the filtration pro- responding change in volume during drying. The packing de- ess. Full infiltration is typically accomplished in a few hours, creases slightly for alumina contents 250%, because of the depending on the fiber packing and the rheology of the slurry narrower distribution of particle sizes in the finer alumina pow- After consolidation, the panels, still within the stainless steel der. For each composition, the change in porosity relative to the frame are removed from the filtration chamber and dried over as-consolidated condition is reflected in the shrinkage. All a period of-48 h, with the last 4 h in an oven at 120 C. Once specimens experience a-1.7%#.2% linear contraction upon drying, reflected as an -3% change in the apparent porosity given an initiai sintering treatment at 900C for 2 h to promote Specimens with <50% alumina undergo little additional change the development of alumina bridges between the mullite par- during sintering at temperatures up to 1200oC. Since the pro- ticle network. The composites are subsequently impregnated cessing temperatures are of this order, there should be no sub- with the alumina precursor solution under vacuum, dried in stantial impact of the alumina content on the evolution of po- open air under an infrared lamp, and heated to 900oC for 2 h to rosity(microstructural stability)within the mullite-rich range yrolyze the precursor. The materials discussed in this paper were impregnated four times. Following the last cycle, the term properties. These data suggest that the shrinkage increases anels are given a final sintering treatment, typically 2 h at more rapidly for compositions above -20% alumina (This may 1200 C, which also serves to stabilize the precursor-derived be related to the evolution of percolation paths between the alumina to the corundum() structure. The implications of this alumina particles). Consequently, the initial composites are final step with regard to potential fiber degradation are dis- based on a 20% alumina matrix. cussed in Sections Iv and V It is noteworthy that all mullite-rich mixtures maintain po- 3) Matrix Sintering rosity levels above -30%, even after heating to 1400C, with attendant implications for the long-term crack deflection capa- Studies of the effect of alumina additions on the sintering of bilities of the matrix under service conditions. The porosity mullite were used to select a suitable matrix composition. content is likely to decrease after precursor impregnation and Samples of unreinforced matrix, 90 mm x 30 mm x 4 pyrolysis, but the initial matrix packing densities are also were consolidated by vacuum filtration from slurries similar to expected to be lower in the composites owing to the presence those used to fabricate the composites. Nominally pure alumina of the fibers. 3 and mullite compacts were prepared, as well as mixtures con- taining 10%, 20%, 30%, and 50% Al2O, by(solid) volume (4 Composite Microstructure The samples were first dried in ambient air for 24 h and then Views of the composite microstructure at different scales slowly heated to 120 C and held at this temperature ove Fig. 4, reveal that the tows are well infiltrated and there are Once dried, each specimen was given a 24 h treatment at 550oC vals. For each step, the samples were heated at -3C/min up to water raisi nd Bulk Density p Bumt d refractory Brick b sorption, apparent and then sequential treatments at 9000-1400C in 100oC inter- the desired temperature, held for 2 h, then cooled down at ignation C-20. 1992 Book of ASTM Standards. American
Dispersion of soft agglomerates is promoted by a combination of mechanical and ultrasonic agitation. The cloth is cut into 200 mm × 125 mm pieces having the desired orientation (0°/90° or ±45°). The pieces are stacked in a loose preform and de-sized by heating to 550°C for 0.5 h. The fiber content of the cloth is 9.3 mm3 /cm2 ; thus a 3 mm plate with a fiber volume fraction of ∼36% requires ∼12 layers. After de-sizing, the fiber preform is placed between two perforated stainless-steel plates that constrain the cloth within the desired thickness during the filtration process. The assembly is then placed in the middle of a chamber with capabilities for evacuation through both top and bottom. An amount of slurry calculated to fill the preform is poured on top, whereupon the chamber is closed and evacuated from both top and bottom to eliminate air bubbles trapped within the preform. The top portion of the chamber is then open to the atmosphere while vacuum is maintained in the bottom to drive the filtration process. Full infiltration is typically accomplished in a few hours, depending on the fiber packing and the rheology of the slurry. After consolidation, the panels, still within the stainless steel frame, are removed from the filtration chamber and dried over a period of ∼48 h, with the last 4 h in an oven at 120°C. Once dry, the green panels are separated from the metallic frame and given an initial sintering treatment at 900°C for 2 h to promote the development of alumina bridges between the mullite particle network. The composites are subsequently impregnated with the alumina precursor solution under vacuum, dried in open air under an infrared lamp, and heated to 900°C for 2 h to pyrolyze the precursor. The materials discussed in this paper were impregnated four times. Following the last cycle, the panels are given a final sintering treatment, typically 2 h at 1200°C, which also serves to stabilize the precursor-derived alumina to the corundum (a) structure. The implications of this final step with regard to potential fiber degradation are discussed in Sections IV and V. (3) Matrix Sintering Studies of the effect of alumina additions on the sintering of mullite were used to select a suitable matrix composition. Samples of unreinforced matrix, 90 mm × 30 mm × 4 mm, were consolidated by vacuum filtration from slurries similar to those used to fabricate the composites. Nominally pure alumina and mullite compacts were prepared, as well as mixtures containing 10%, 20%, 30%, and 50% Al2O3 by (solid) volume. The samples were first dried in ambient air for 24 h and then slowly heated to 120°C and held at this temperature overnight. Once dried, each specimen was given a 24 h treatment at 550°C and then sequential treatments at 900°–1400°C in 100°C intervals. For each step, the samples were heated at ∼3°C/min up to the desired temperature, held for 2 h, then cooled down at ∼5°C/min. (The slow temperature ramping was aimed at minimizing the risk of cracking or chipping, which would impair the accuracy of the measurements. Examination of these samples revealed no microcracking at any stage.) The packing densities were measured after each step following the ASTM C20-92 standard,‡ as were the external dimensions of the specimens. The results are summarized in Fig. 3. The differing behaviors of the pure-alumina and pure-mullite compacts lend support to the proposed microstructural design concept. The porosites for the as-consolidated (wet) compacts are estimated from the packing densities after drying at 120°C and the corresponding change in volume during drying. The packing decreases slightly for alumina contents $50%, because of the narrower distribution of particle sizes in the finer alumina powder. For each composition, the change in porosity relative to the as-consolidated condition is reflected in the shrinkage. All specimens experience a ∼1.7% ± 0.2% linear contraction upon drying, reflected as an ∼3% change in the apparent porosity. Specimens with <50% alumina undergo little additional change during sintering at temperatures up to 1200°C. Since the processing temperatures are of this order, there should be no substantial impact of the alumina content on the evolution of porosity (microstructural stability) within the mullite-rich range. However, the 1400°C data may be more relevant to the longterm properties. These data suggest that the shrinkage increases more rapidly for compositions above ∼20% alumina. (This may be related to the evolution of percolation paths between the alumina particles). Consequently, the initial composites are based on a 20% alumina matrix. It is noteworthy that all mullite-rich mixtures maintain porosity levels above ∼30%, even after heating to 1400°C, with attendant implications for the long-term crack deflection capabilities of the matrix under service conditions. The porosity content is likely to decrease after precursor impregnation and pyrolysis,31 but the initial matrix packing densities are also expected to be lower in the composites owing to the presence of the fibers.32 (4) Composite Microstructure Views of the composite microstructure at different scales, Fig. 4, reveal that the tows are well infiltrated, and there are ‡ ‘‘Standard Test Methods for Apparent Porosity, Water Absorption, Apparent Specific Gravity, and Bulk Density of Burned Refractory Brick and Shapes by Boiling Water.’’ ASTM Designation C-20. 1992 Book of ASTM Standards. American Society for Testing and Materials, Philadelphia, PA. Fig. 3. Evolution of porosity during heat treatment as a function of the alumina content in unreinforced mullite–alumina compacts produced by vacuum filtration under conditions similar to those used for the composites. The average particle sizes are ∼1 mm for mullite and ∼0.2 mm for alumina. Details of the heat treatment are given in the text. Fig. 2. Processing approach used for the fabrication of damagetolerant all-oxide CFCC panels without fiber coatings. Composites with 3-D preforms can be fabricated in the same manner. August 1998 Processing and Performance of an All-Oxide Ceramic Composite 2079
Journal of the American Ceramic Society-Levi et al. Vol 8l. No 8 only occasional instances of large-scale voids that arise fro reduce the scale of the unreinforced matrix regions by filling air bubbles or pockets of unconsolidated slurry trapped by the he larger spaces in the preform with short fibers. This idea was filtration front. The large particles within the matrix are mul explored by coating the cloth with a paste of chopped alumina lite, and the finer particles are a mixture of alumina and mullite fiber with an average aspect ratio of -10. The paste fills the ( Fig. 4(a)). It is also evident upon analysis of the fracture cross-over regions, as depicted in Figs. 5(a)and (b), and at surfaces that the matrix is bonded to the fibers, presumably taches randomly aligned short fibers to the cloth surface. La through the same alumina bridges and/or precursor necks pres- ers of such cloth were assembled into a preform and processed ent within the mullite network(see Section IV) in th ne same manner. The results in Figs. 5(c) and(d)show that Examination at lower magnifications reveals the presence of the flaws have been largely suppressed. This approach, how- cracklike shrinkage flaws within the matrix, especially in re- ever, further decreases the packing efficiency and limits the gions devoid of fibers( Fig. 4(d)). The flaws are typically per- volume fraction of continuous reinforcement to -25% which is pendicular to the fibers and tend to form 2-D arrays on planes substantially below the levels typical of current CFCC com- parallel to the fiber cloth. The results in Fig 3 suggest that the posites(235%). Efforts to find suitable ways to minimize flaws evolve primarily upon matrix shrinkage during drying flaws without sacrificing the volume fraction of reinforcement The phenomenon is induced by the biaxial constraint imposed are continuing e reinforcements and enhanced by the presence of rather large unreinforced matrix regions, owing to the limitations to fiber packing when using woven cloth. Such cracking does not IV. Tensile Behavior at Room Temperature occur in unidirectional composites. )Because of the rela Stress-strain curves for both materials have been measured tively large openings of these drying cracks, they cannot be healed in tension in the 0/900 and +45 orientations, as needed for by the subsequent precursor and pyrolysis implementation in numerical design codes. The test procedures One way to alleviate the formation of shrinkage flaws is to have been described elsewhere 33 Periodic unload-reload mea- d m 100pm Fig. 4. Microstructural views of al on N610 woven fiber preforms: (a) matrix containing 20% AL2O3 and-37% porosity, (b )and(c) show good levels of inf ween the fiber layers, respectively, (d) matrix cracks produced during drying. Samples in(c)and(d)were given 10 impre
only occasional instances of large-scale voids that arise from air bubbles or pockets of unconsolidated slurry trapped by the filtration front. The large particles within the matrix are mullite, and the finer particles are a mixture of alumina and mullite (Fig. 4(a)). It is also evident upon analysis of the fracture surfaces that the matrix is bonded to the fibers, presumably through the same alumina bridges and/or precursor necks present within the mullite network (see Section IV). Examination at lower magnifications reveals the presence of cracklike shrinkage flaws within the matrix, especially in regions devoid of fibers (Fig. 4(d)). The flaws are typically perpendicular to the fibers and tend to form 2-D arrays on planes parallel to the fiber cloth. The results in Fig. 3 suggest that the flaws evolve primarily upon matrix shrinkage during drying. The phenomenon is induced by the biaxial constraint imposed by the reinforcements and enhanced by the presence of rather large unreinforced matrix regions, owing to the limitations to fiber packing when using woven cloth. (Such cracking does not occur in unidirectional composites.14,18) Because of the relatively large openings of these drying cracks, they cannot be healed by the subsequent precursor impregnation and pyrolysis. One way to alleviate the formation of shrinkage flaws is to reduce the scale of the unreinforced matrix regions by filling the larger spaces in the preform with short fibers. This idea was explored by coating the cloth with a paste of chopped alumina fiber with an average aspect ratio of ∼10. The paste fills the cross-over regions, as depicted in Figs. 5(a) and (b), and attaches randomly aligned short fibers to the cloth surface. Layers of such cloth were assembled into a preform and processed in the same manner. The results in Figs. 5(c) and (d) show that the flaws have been largely suppressed. This approach, however, further decreases the packing efficiency and limits the volume fraction of continuous reinforcement to ∼25%, which is substantially below the levels typical of current CFCC composites ($35%). Efforts to find suitable ways to minimize flaws without sacrificing the volume fraction of reinforcement are continuing. IV. Tensile Behavior at Room Temperature Stress–strain curves for both materials have been measured in tension in the 0°/90° and ±45° orientations, as needed for implementation in numerical design codes. The test procedures have been described elsewhere.33 Periodic unload–reload meaFig. 4. Microstructural views of all-oxide composites based on N610 woven fiber preforms: (a) matrix containing 20% Al2O3 and ∼37% porosity, (b) and (c) show good levels of infiltration within and between the fiber layers, respectively, (d) matrix cracks produced during drying. Samples in (c) and (d) were given 10 impregnation cycles to facilitate polishing. 2080 Journal of the American Ceramic Society—Levi et al. Vol. 81, No. 8
August 1998 Processing and Performance of an All-Oxide Ceramic Composite (b) 1 mm (c) (d) 溶答容20m 200m Fig. 5. An approach to minimize matrix cracking in slurry-infiltrated composies: (a) depicts the topography of the fiber cloth in the as-received condition and(b)after coating with a paste of short alumina fiber. (c) and(d) show in-plane and cross-sectional views of the resulting composites where only minimal microcracking can be detected surements have been used to assess the incidence of damage also of the above order, but these strains are much closer to the evolution(from modulus changes), as well as the occurrence of values measured in tows extracted from the woven fabric. internal friction(from hysteresis strains ). 34, 35 The results are (Section Ill(I). These observations are in agreement with the reminiscent of those found for fiber-dominated carbon-matrix superior stability of the N720 fiber. They also suggest that the composites, such as carbon/carbon and SiC/carbon. 5.,33, 36-39 composite manufacturing process do not cause significant In the 0 /90 orientation, the tensile response is essentially degradation of the reinforcement beyond that expected from linear with only small inelastic strains, Fig. 6(a), similar to that hermal of the pristine fibers. Notwithstanding the observed in SiC/carbon systems, ,39 Fig. 6(b). The behavior is higher stability of the N720 fiber, these composites have sig- quite reproducible, with stress-strain curves for a given volume nificantly lower strengths at ambient temperature than those tively narrow band. The initial moduli (-100 and-60 GPa for reflecting the initially lower strength of the N720 fiber and the N610 and N720 composites, respectively) are consistently indicating that strength retention in the N610 fiber is still ad- larger than those expected from the axial fibers alone (fEr /2 of equate after processing. Further work is clearly needed to un- -75 and-50 GPa, respectively ), reflecting a contribution from derstand and quantify the effects of processing on fiber degra- the matrix and the transverse reinforcements. The average fail ure strain for 14 specimens from five different panels of N610 In the 0o/90 orientation, the fracture plane is ill-defined composite(33%39% fiber) sintered at 1200.C is 0.27% with the fiber tows breaking over a wide range of axial loca- These failure strains are smaller than those expected for pris tions, spanning a distance of-I cm(Fig. 7(a)). The locations of ine tows of comparable gauge length, even after weaving, but the fiber breaks within an individual tow also exhibit a broad are consistent with the values reported for N610 fibers sub- distribution, typically -l mm in length(Fig. 7(b). These ob- jected to thermal cycles similar to those used in composite servations validate the efficacy of the porous matrix as a crack processing. 24 The failure strains of the N720 composites are deflection medium both within and between the fiber tows The
surements have been used to assess the incidence of damage evolution (from modulus changes), as well as the occurrence of internal friction (from hysteresis strains).34,35 The results are reminiscent of those found for fiber-dominated carbon-matrix composites, such as carbon/carbon and SiC/carbon.5,33,36–39 In the 0°/90° orientation, the tensile response is essentially linear with only small inelastic strains, Fig. 6(a), similar to that observed in SiC/carbon systems,5,39 Fig. 6(b). The behavior is quite reproducible, with stress–strain curves for a given volume fraction and set of processing conditions falling within a relatively narrow band. The initial moduli (∼100 and ∼60 GPa for the N610 and N720 composites, respectively) are consistently larger than those expected from the axial fibers alone ( f Ef/2 of ∼75 and ∼50 GPa, respectively), reflecting a contribution from the matrix and the transverse reinforcements. The average failure strain for 14 specimens from five different panels of N610 composite (33%–39% fiber) sintered at 1200°C is 0.27%. These failure strains are smaller than those expected for pristine tows of comparable gauge length, even after weaving, but are consistent with the values reported for N610 fibers subjected to thermal cycles similar to those used in composite processing.24 The failure strains of the N720 composites are also of the above order, but these strains are much closer to the values measured in tows extracted from the woven fabric29 (Section III(1)). These observations are in agreement with the superior stability of the N720 fiber. They also suggest that the composite manufacturing process does not cause significant degradation of the reinforcement beyond that expected from thermal exposure of the pristine fibers. Notwithstanding the higher stability of the N720 fiber, these composites have significantly lower strengths at ambient temperature than those made with N610 fibers (∼140 MPa vs >200 MPa, respectively), reflecting the initially lower strength of the N720 fiber and indicating that strength retention in the N610 fiber is still adequate after processing. Further work is clearly needed to understand and quantify the effects of processing on fiber degradation and its effects on composite performance. In the 0°/90° orientation, the fracture plane is ill-defined, with the fiber tows breaking over a wide range of axial locations, spanning a distance of ∼1 cm (Fig. 7(a)). The locations of the fiber breaks within an individual tow also exhibit a broad distribution, typically ∼1 mm in length (Fig. 7(b)). These observations validate the efficacy of the porous matrix as a crack deflection medium both within and between the fiber tows. The Fig. 5. An approach to minimize matrix cracking in slurry-infiltrated composies: (a) depicts the topography of the fiber cloth in the as-received condition and (b) after coating with a paste of short alumina fiber. (c) and (d) show in-plane and cross-sectional views of the resulting composites where only minimal microcracking can be detected. August 1998 Processing and Performance of an All-Oxide Ceramic Composite 2081