J. A Cera Soc.,86[0]1733-4002003) ournal Processing and Properties of a Porous Oxide Matrix Composite Reinforced with Continuous Oxide Fibers Magnus G. Holmquist'-and Fred F. Lange*, Materials Department, University of California, Santa Barbara, California 93106 Volvo Aero Corporation, 461 81 Trollhattan, Sweden A process to manufacture porous oxide matrix/polycrystalline nvironment, such as nitrogen oxides, carbon monoxide and oxide fiber composites was developed and evaluated. The unburned hydrocarbon nethod uses infiltration of fiber cloths with an aqueous slurry Most CFCCs that are commercially available are based on SiC of mullite/alumina powders to make prepregs. By careful fibers, with either oxide or non-oxide matrixes, and interphases anipulation of the interparticle pair potential in the slurry, a consisting of carbon, BN, SiC or combinations thereof. The consolidated slurry with a high particle density is produced interphases are designed to provide a crack-deflecting layer be with a sufficiently low viscosity to allow efficient infiltration of tween the matrix and fibers that prevents matrix cracks from the fiber tows. Vibration-assisted infiltration of stacked, cloth extending through the fibers, thus allowing crack bridging to occ prepregs in combination with a simple vacuum bag technique on matrix cracking enabling damage tolerance via notch insensi- produced composites with homogeneous microstructures. The method has the additional advantage of allowing complex tivity. SiC fiber based composites have attractive high- reep resistance, microstructural shapes to be made. Subsequent infiltration of the powder temperature properties such mixture with an alumina precursor was made to strengthen stability, high tensile strength, and high thermal conductivity. the matrix. The porous matrix, without fibers, possessed good phase will cause embrittlement of the composite after service at treatment at 1200C. Mechanical properties were evaluated in high temperature for long times. Embrittlement is most severe with flexural testing in a manner that precluded interlaminar shear cyclic loading beyond the proportional limit because oxygen that failure before failure via the tensile stresses. It was shown tha penetrates via the matrix cracks will react with the interphase and he composite produced by this method was comparable to the fibers. ,0 This effect is most pronounced for carbon coatings orous ox kide matrix composites manufactured by other pre but the introduction of bn coatings and boron additives has cesses using the same fibers(N610 and N720). The ratio of improved the situation in oxidizing environments, where BI notch strength to unnotch strength for a crack to width ratio of oxidation products (liquid boron oxide)help in healing matrix cracks. However, in wet environments the problem persists since 0.5 was 0.7-0.9, indicating moderate notch sensitivity. Inter- the boron oxidation products volatilize as boron hydroxide aminar shear strength, which is dominated by matrix strength, changed from 7 to 12 MPa for matrix porosity avoid degradation in oxidizing (especially wet)environments. ranging from 38% to 43%, respectively. The porous micro- tructural design strategies therefore usually require that the structure did not change after aging at 1200%C for 100 h. Heat stresses remain below the matrix cracking stress. End-user treatment at 1300C for 100 h reduced the strength for the experience indicates that stress excursions above the matrix N610 and N720 composites by 35% and 20%, respectively, and cracking stress is very difficult to avoid, and thus local ncreased their brittle nature embrittlement will be one of the dominant life-limiting phe nomena of non-oxide composites. These shortco promoted the development of environmentally stable all-oxide L. Introduction composites, i.e., materials where all constituents(fiber, inter C ONTINUOUS fiber ceramic composites(CFCCs) have attracted Two approaches have been used to develop damage-tolerant interest for a variety of high-temperature thermostructural all-oxide composites. The more traditional approach requires a applications in gas turbine engines, rocket engines, heat ex- crack-deflecting interface between the matrix and fibers. This can be hangers, and hot gas filters. The reason is that they offer achieved by adding an interphase which either forms a crack uperior refractoriness compared with conventional metal alloys deflecting interface with the fibers. -l5 has itself a low fracture combined with an inelastic deformation behavior rendering them toughness(e.g, "cleavable"oxides or a porous layer-9), or forms damage tolerant. Of particular interest is their use in combustor a gap between fiber and matrix(fugitive coating) king advantage of the FcCS ability to op The use of a porous matrix to isolate fibers from matrix cracks at high temperatures with reduced need for cooling air, it is a second, more recent approach for developing damage-tolerant possible to increase the efficiency and also control the combus- opposites. In this approach, the crack does not have a contin- tion process to minimize formation of species harmful to the uous front, but, instead, the matrix is held together by grai Matrix failure by the sequential failure pairs. Fibers are isolated from the stress singularity of a matrix E. Lara-Curzio--contributing editor crack because the matrix is not sufficiently continuous to support a crack. There are several examples of CFCCs which rely on a porous matrix for damage tolerance 3,21 24-27 The failure mech- anisms have been examined in some detail. 22-24,28-3On loading has scpt h o as 5 0. Red ebv d Agust r 200ch appice d nagessbes-51 3095. the matrix experiences continuous microcracking during loading oundation for financial support. and appears to have completely disintegrated at the onset of fiber failure. Contrary to the conventional weak interface CFCCs where ation: Advanced Engineering, SAAB, 461 80 Trollhattan, Sweden the fibers slide out of the matrix, leaving distinct holes, when Califonia fibers fail in a porous matrix, they release a large volume of
Processing and Properties of a Porous Oxide Matrix Composite Reinforced with Continuous Oxide Fibers Magnus G. Holmquist†,‡ and Fred F. Lange* ,§ Materials Department, University of California, Santa Barbara, California 93106 Volvo Aero Corporation, 461 81 Trollha¨ttan, Sweden A process to manufacture porous oxide matrix/polycrystalline oxide fiber composites was developed and evaluated. The method uses infiltration of fiber cloths with an aqueous slurry of mullite/alumina powders to make prepregs. By careful manipulation of the interparticle pair potential in the slurry, a consolidated slurry with a high particle density is produced with a sufficiently low viscosity to allow efficient infiltration of the fiber tows. Vibration-assisted infiltration of stacked, cloth prepregs in combination with a simple vacuum bag technique produced composites with homogeneous microstructures. The method has the additional advantage of allowing complex shapes to be made. Subsequent infiltration of the powder mixture with an alumina precursor was made to strengthen the matrix. The porous matrix, without fibers, possessed good thermal stability and showed linear shrinkage of 0.9% on heat treatment at 1200°C. Mechanical properties were evaluated in flexural testing in a manner that precluded interlaminar shear failure before failure via the tensile stresses. It was shown that the composite produced by this method was comparable to porous oxide matrix composites manufactured by other processes using the same fibers (N610 and N720). The ratio of notch strength to unnotch strength for a crack to width ratio of 0.5 was 0.7–0.9, indicating moderate notch sensitivity. Interlaminar shear strength, which is dominated by matrix strength, changed from 7 to 12 MPa for matrix porosity ranging from 38% to 43%, respectively. The porous microstructure did not change after aging at 1200°C for 100 h. Heat treatment at 1300°C for 100 h reduced the strength for the N610 and N720 composites by 35% and 20%, respectively, and increased their brittle nature. I. Introduction CONTINUOUS fiber ceramic composites (CFCCs) have attracted interest for a variety of high-temperature thermostructural applications in gas turbine engines,1 rocket engines,2 heat exchangers,3 and hot gas filters.4 The reason is that they offer superior refractoriness compared with conventional metal alloys combined with an inelastic deformation behavior rendering them damage tolerant.5 Of particular interest is their use in combustor walls.6–8 By taking advantage of the CFCCs ability to operate at high temperatures with reduced need for cooling air, it is possible to increase the efficiency and also control the combustion process to minimize formation of species harmful to the environment, such as nitrogen oxides, carbon monoxide, and unburned hydrocarbons.9 Most CFCCs that are commercially available are based on SiC fibers, with either oxide or non-oxide matrixes, and interphases consisting of carbon, BN, SiC or combinations thereof. The interphases are designed to provide a crack-deflecting layer between the matrix and fibers that prevents matrix cracks from extending through the fibers, thus allowing crack bridging to occur on matrix cracking enabling damage tolerance via notch insensitivity.3,5 SiC fiber based composites have attractive hightemperature properties such as creep resistance, microstructural stability, high tensile strength, and high thermal conductivity. However, the oxidation sensitivity of the crack-deflecting interphase will cause embrittlement of the composite after service at high temperature for long times. Embrittlement is most severe with cyclic loading beyond the proportional limit because oxygen that penetrates via the matrix cracks will react with the interphase and the fibers.3,10 This effect is most pronounced for carbon coatings, but the introduction of BN coatings and boron additives has improved the situation in oxidizing environments, where BN oxidation products (liquid boron oxide) help in healing matrix cracks. However, in wet environments the problem persists since the boron oxidation products volatilize as boron hydroxides.3,11 To avoid degradation in oxidizing (especially wet) environments, structural design strategies therefore usually require that the stresses remain below the matrix cracking stress. End-user experience indicates that stress excursions above the matrix cracking stress is very difficult to avoid, and thus local embrittlement will be one of the dominant life-limiting phenomena of non-oxide composites. These shortcomings have promoted the development of environmentally stable all-oxide composites, i.e., materials where all constituents (fiber, interphase, and matrix) are oxides.3,12,13 Two approaches have been used to develop damage-tolerant all-oxide composites. The more traditional approach requires a crack-deflecting interface between the matrix and fibers. This can be achieved by adding an interphase which either forms a crackdeflecting interface with the fibers,13–15 has itself a low fracture toughness (e.g., “cleavable” oxides16 or a porous layer17–19), or forms a gap between fiber and matrix (fugitive coating).18,20 The use of a porous matrix to isolate fibers from matrix cracks is a second, more recent approach for developing damage-tolerant composites.21 In this approach, the crack does not have a continuous front, but, instead, the matrix is held together by grain pairs.22,23 Matrix failure occurs by the sequential failure of grain pairs. Fibers are isolated from the stress singularity of a matrix crack because the matrix is not sufficiently continuous to support a crack. There are several examples of CFCCs which rely on a porous matrix for damage tolerance.13,21,24–27 The failure mechanisms have been examined in some detail.22–24,28–31 On loading, the matrix experiences continuous microcracking during loading and appears to have completely disintegrated at the onset of fiber failure. Contrary to the conventional weak interface CFCCs where the fibers slide out of the matrix, leaving distinct holes, when fibers fail in a porous matrix, they release a large volume of E. Lara-Curzio—contributing editor Manuscript No. 187504. Received August 22, 2001; approved November 8, 2002. This research was supported by the Army Research Office, DAAG55-98-1-0455. M.G.H. thanks the Hans Werthe´n Foundation for financial support. *Member, American Ceramic Society. † Volvo Aero Corp. ‡ Current affiliation: Advanced Engineering, SAAB, 461 80 Trollha¨ttan, Sweden. § University of California. J. Am. Ceram. Soc., 86 [10] 1733–40 (2003) 1733 journal
1734 Journal of the American Ceramic Sociery-Holmquist and lange Vol. 86. No. 10 disintegrated matrix in the form of powder. Thus, the ability mullite grains and have higher creep resistance and high porous matrix to isolate the fibers from matrix cracks wi temperature stability, but lower room temperature strength, com- the fibers to fail in a manner similar to what is seen pared with N610 fibers, which are high purity (99%) polycrys- bundles. The high failure strain of the fiber bundle wi talline a-alumil become the failure strain of the composite The mullite powder used was MU-107(Showa Denko KK, Two methods have been developed at University of California Tokyo, Japan), which has a mean particle size of -I um, a particle at Santa Barbara(UCSB) to produce porous matrix composites. size distribution of 0.5-2.5 um, and a Bet surface area of 7.5 The first uses pressure filtration to pack particles around fibers m-g. It has a chemistry of 75.5% Al,O3 and 24% SiOz(by within a preform. The powder surrounding the fibers is ther weight) with only trace amounts of TiO,, Fe2O3, and Na, strengthened by the cyclic infiltration and pyrolysis of a precur- (manufacturers' data). AKP-50(Sumitomo Chemicals, Tokyo sor.2,24, 25,28For this method, the slurry is formulated so that the Japan)was selected as the alumina powder and has a mean particle particles are repulsive to themselves and also to the fibers. Mullite size of -0 2 um, a more narrow particle size distribution(0.1-0 densification at temperatures below -1300C25 Levi et al28 um), and a BET surface area of 10.6 m2/g(manufacturers'data) has been chosen as the matrix material because of its lack Its chemistry is essentially pure a-Al2O3(99.995%). An Al, O3 reported that alumina powder with a much smaller particle size precursor, aluminum(Ill) sec-butoxide, C12H27O3Al(Gelest, Inc (-200 nm) could be added to the mullite powder to aid in Tullytown, PA), was used to strengthen the matrix. The recurs engthening the powder matrix. At processing temperatures ound C(chosen to avoid degradation of the fibers) the infiltration the precursor was diluted with 25%(by volume)of alumina sinters to form bridges between the larger mullite particles sec-butyl alcohol(Sigma-Aldrich, Milwaukee, Wn) and between the mullite particles and the fibers. In this case, the persed, aqueous slurries containing 20 vol% solids(mullite/ volume fraction of mullite powder, which does not shrink at alumina proportions, 70/30)were prepared The mullite to alumina 1200C, is sufficient (20.70)to prevent shrinkage of the mixed ratio was selected to achieve high packing densities of the powder powder matrix body and low shrinkage in the following sintering step2 428Lower The second method was first introduced by Haslam et al.- alumina content would give fewer sintering necks between mullite stead of packing the particles around the fibers via pressure particles, whereas mixtures with higher alumina content will filtration, the powder is first consol to a very high volume densify relatively fast above 1200oC. Tetraethylammonium hy- fraction and then infiltrated into the fiber preform via vibration- droxide (TEA-OH) was used to maintain the pH above 11 assisted flow. As detailed elsewhere, the initial slurry must be allowing electrostatic repulsive interactions to develop between formulated so that the particles are weakly attracted to one another the oxide particles. A 2 wt% amount(relative to the solids) of The formulation of the weakly attractive particle network requires poly(ethylene oxide) urethane silane(PEG-silane, Gelest, Inc the development of a short-range repulsive particle potential-that was added to induce a steric dispersing effect. The PEG-silane events the particles from being pushed into contact during molecules chem-absorbed to the pa acting wit oressure filtration, thus allowing the network to retain its interpar--M-OH (M= metal atom)surface sites. 37,38 Dispersion of soft ticle potential in the consolidated state gglomerates was promoted by ultrasonic agitation for 5 min. The In the present investigation we used a powder slurry with a rry was placed on a mechanical roller for 12 h, and then special interparticle pair potential weakly attractive produced by a tetramethy lammonium nitrate(TMA-N)salt was added (0. 25M)to hort-range repulsive")that allows a powder compact, which has orm weakly attractive pair potentials between the particles. A been previously consolidated by pressure filtration, to be fluidized. described elsewhere, TMA counter ions aid in producing a The preconsolidated slurry with a relative density of 0.54 was weakly attractive particle network. The network can be packed placed on a fiber cloth, and since it shows shear rate thinning, Itration. and the so-for added vibration reduces the viscosity and allows rapid and effi- consolidated body can subsequently be fluidized again via vibra cient intrusion into the fiber tows. The technique to make com- tion. The slurry was consolidated by pressure filtration at 4 MPa te posites, called Vibrolntrusion, has been described elsewhere.",- form disk-shaped bodies that were stored in sealed plastic bags The vibration of the preconsolidated slurry was conducted betweer The consolidation pressure of 4 MPa was lower than the critical plastic sheets to avoid evaporation. Prepregs made in this manner pressure where a large number of particles are pushed into contact, could be frozen and stored. Once thawed they were flexible and which would obviate fluidization after consolidation. 32,37,38Using could be bent, cut, and formed much like an epoxy/fiber prepreg. the weight difference method, the volume fraction of solids within Prepregs could be stacked and formed into complex geometries the consolidated bodies was determined to 54 1% like T-joints, doubly curved shapes, and tubes. 33,34 Drying only The consolidated powder compact was fluidized by subjecting it causes minimal shrinkage since a high dry content could be to mechanical vibration. Fiber cloths(cut to -60 X 60 mm2)were reached in the matrix slurry. Subsequent precursor impregnation/ put in separate plastic bags, and an excess of the preconsolidated pyrolysis cycles were done to strengthen the matrix slurry was dispensed to both sides of each fiber cloth. Assisted by This new method to manufacture CFCCs produced material vibration, the slurry was manually rolled across the surface of the with similar matrixes as a previous route, 24,25,128-30 comparisons fiber cloth with a piece of aluminum rod until the cloth was fully between this new method and the older method to produce porous infiltrated. Since the preconsolidated slurry exhibits shear-rate thinning, the vibration reduces the viscosity and allows rapid difference is that the mullite/alumina volume ratio, 70/30 used intrusion of the particles into the fiber cloth.22,23 here, is lower than that(80/20)used earlier. The purpose was to Prepregs were frozen to aid removal from the plastic bag, or show that the process produces composite material with similar they could be stored for later use. To produce the composite, properties but with the added advantage of allowing complex frozen prepregs were removed from the plastic bags and stacked shapes to be made. on top of each other. The pile of prepregs was packed in a plastic ed between two flat steel plates usin spacer bars to fix the thickness. The number of prepregs was IL. Experimental Procedure chosen to give the desired fiber volume fraction of the composite (in this case 13 layers of prepregs and a nominal thickness of 3.18 (I Materials and Composite Processing mm). After thawing, the assembled layers were put on a vibrating enforcement fibers used in this work were Nextel 610 table and pressed lightly to cause the preconsolidated slurry to Nextel 720(N610 and N720, 3M Corp, St Paul, MN) flow and complete the infiltration. To remove trapped air as much nto eight-harmess satin fabrics. The tows in the fabric as possible(which later on could give rise to large-scale porosity 400 filaments with diameters between 10 and 12 um the vacuum was kept on a level of -10 torr during the vibration bers are composed of a mixture of sub-micrometer alumina and step(-5 min). It should be noted that the so-formed green ceramic
disintegrated matrix in the form of powder. Thus, the ability of the porous matrix to isolate the fibers from matrix cracks will allow the fibers to fail in a manner similar to what is seen for dry bundles. The high failure strain of the fiber bundle will also become the failure strain of the composite. Two methods have been developed at University of California at Santa Barbara (UCSB) to produce porous matrix composites. The first uses pressure filtration to pack particles around fibers within a preform. The powder surrounding the fibers is then strengthened by the cyclic infiltration and pyrolysis of a precursor.21,24,25,28 For this method, the slurry is formulated so that the particles are repulsive to themselves and also to the fibers. Mullite has been chosen as the matrix material because of its lack of densification at temperatures below 1300°C.25 Levi et al.28 reported that alumina powder with a much smaller particle size (200 nm) could be added to the mullite powder to aid in strengthening the powder matrix. At processing temperatures around 1200°C (chosen to avoid degradation of the fibers) the alumina sinters to form bridges between the larger mullite particles and between the mullite particles and the fibers. In this case, the volume fraction of mullite powder, which does not shrink at 1200°C, is sufficient (0.70) to prevent shrinkage of the mixed powder matrix. The second method was first introduced by Haslam et al.23 Instead of packing the particles around the fibers via pressure filtration, the powder is first consolidated to a very high volume fraction and then infiltrated into the fiber preform via vibrationassisted flow. As detailed elsewhere, the initial slurry must be formulated so that the particles are weakly attracted to one another. The formulation of the weakly attractive particle network requires the development of a short-range repulsive particle potential32 that prevents the particles from being pushed into contact during pressure filtration, thus allowing the network to retain its interparticle potential in the consolidated state. In the present investigation we used a powder slurry with a special interparticle pair potential (weakly attractive produced by a short-range repulsive22) that allows a powder compact, which has been previously consolidated by pressure filtration, to be fluidized. The preconsolidated slurry with a relative density of 0.54 was placed on a fiber cloth, and since it shows shear rate thinning, added vibration reduces the viscosity and allows rapid and efficient intrusion into the fiber tows. The technique to make composites, called VibroIntrusion, has been described elsewhere.22,23 The vibration of the preconsolidated slurry was conducted between plastic sheets to avoid evaporation. Prepregs made in this manner could be frozen and stored. Once thawed they were flexible and could be bent, cut, and formed much like an epoxy/fiber prepreg. Prepregs could be stacked and formed into complex geometries like T-joints, doubly curved shapes, and tubes.33,34 Drying only causes minimal shrinkage since a high dry content could be reached in the matrix slurry. Subsequent precursor impregnation/ pyrolysis cycles were done to strengthen the matrix. This new method to manufacture CFCCs produced material with similar matrixes as a previous route;24,25,28–30 comparisons between this new method and the older method to produce porous matrix CFCCs will be made throughout the text. The main difference is that the mullite/alumina volume ratio, 70/30 used here, is lower than that (80/20) used earlier. The purpose was to show that the process produces composite material with similar properties but with the added advantage of allowing complex shapes to be made. II. Experimental Procedure (1) Materials and Composite Processing Reinforcement fibers used in this work were Nextel 610TM and Nextel 720TM (N610 and N720, 3M Corp., St. Paul, MN) woven into eight-harness satin fabrics. The tows in the fabric contain 400 filaments with diameters between 10 and 12 m. N720 fibers are composed of a mixture of sub-micrometer alumina and mullite grains and have higher creep resistance and hightemperature stability, but lower room temperature strength, compared with N610 fibers, which are high purity (99%) polycrystalline -alumina.35,36 The mullite powder used was MU-107 (Showa Denko KK, Tokyo, Japan), which has a mean particle size of 1 m, a particle size distribution of 0.5–2.5 m, and a BET surface area of 7.5 m2 /g. It has a chemistry of 75.5% Al2O3 and 24% SiO2 (by weight) with only trace amounts of TiO2, Fe2O3, and Na2O (manufacturers’ data). AKP-50 (Sumitomo Chemicals, Tokyo, Japan) was selected as the alumina powder and has a mean particle size of 0.2 m, a more narrow particle size distribution (0.1–0.3 m), and a BET surface area of 10.6 m2 /g (manufacturers’ data). Its chemistry is essentially pure -Al2O3 (99.995%). An Al2O3 precursor, aluminum(III) sec-butoxide, C12H27O3Al (Gelest, Inc., Tullytown, PA), was used to strengthen the matrix. The precursor is 95% pure and has a yield of 4% of Al2O3 by volume. Before infiltration the precursor was diluted with 25% (by volume) of sec-butyl alcohol (Sigma-Aldrich, Milwaukee, WI). Dispersed, aqueous slurries containing 20 vol% solids (mullite/ alumina proportions, 70/30) were prepared. The mullite to alumina ratio was selected to achieve high packing densities of the powder body and low shrinkage in the following sintering step.24,28 Lower alumina content would give fewer sintering necks between mullite particles, whereas mixtures with higher alumina content will densify relatively fast above 1200°C. Tetraethylammonium hydroxide (TEA-OH) was used to maintain the pH above 11, allowing electrostatic repulsive interactions to develop between the oxide particles. A 2 wt% amount (relative to the solids) of poly(ethylene oxide) urethane silane (PEG-silane, Gelest, Inc.) was added to induce a steric dispersing effect. The PEG-silane molecules chem-absorbed to the particles by reacting with the –M–OH (M metal atom) surface sites.37,38 Dispersion of soft agglomerates was promoted by ultrasonic agitation for 5 min. The slurry was placed on a mechanical roller for 12 h, and then tetramethylammonium nitrate (TMA-N) salt was added (0.25M) to form weakly attractive pair potentials between the particles. As described elsewhere, TMA counter ions aid in producing a weakly attractive particle network.37,38 The network can be packed to a high density via pressure filtration, and the so-formed consolidated body can subsequently be fluidized again via vibration. The slurry was consolidated by pressure filtration at 4 MPa to form disk-shaped bodies that were stored in sealed plastic bags. The consolidation pressure of 4 MPa was lower than the critical pressure where a large number of particles are pushed into contact, which would obviate fluidization after consolidation.32,37,38 Using the weight difference method, the volume fraction of solids within the consolidated bodies was determined to 54 1%. The consolidated powder compact was fluidized by subjecting it to mechanical vibration. Fiber cloths (cut to 60 60 mm2 ) were put in separate plastic bags, and an excess of the preconsolidated slurry was dispensed to both sides of each fiber cloth. Assisted by vibration, the slurry was manually rolled across the surface of the fiber cloth with a piece of aluminum rod until the cloth was fully infiltrated. Since the preconsolidated slurry exhibits shear-rate thinning, the vibration reduces the viscosity and allows rapid intrusion of the particles into the fiber cloth.22,23 Prepregs were frozen to aid removal from the plastic bag, or they could be stored for later use. To produce the composite, frozen prepregs were removed from the plastic bags and stacked on top of each other. The pile of prepregs was packed in a plastic bag, evacuated, and placed between two flat steel plates using two spacer bars to fix the thickness. The number of prepregs was chosen to give the desired fiber volume fraction of the composite (in this case 13 layers of prepregs and a nominal thickness of 3.18 mm). After thawing, the assembled layers were put on a vibrating table and pressed lightly to cause the preconsolidated slurry to flow and complete the infiltration. To remove trapped air as much as possible (which later on could give rise to large-scale porosity), the vacuum was kept on a level of 10 torr during the vibration step (5 min). It should be noted that the so-formed green ceramic 1734 Journal of the American Ceramic Society—Holmquist and Lange Vol. 86, No. 10
October 2003 Porous Oxide Matrix Composite Reinforced with Oxide Fibers 735 matrix composite (CMC) body is very flexible and could be shaped much like an epoxy/fiber prepreg After shaping, the water was removed from the powder matrix by drying at 70C. An initial sintering treatment was done at 900C tor 2 h to promote the development of alumina bridges (a)h impregnated with the alumina precursor solution under vacuum. The impregnation step was performed in a dry nitrogen atmo- sphere to prevent premature gelation of the precursor due to atmospheric water vapor. The composites were left in the precur- sor solution for 2 h at atmospheric pressure and transferred into ammoniated water(pH 10)to gel the precursor throughout the body and prevent it from redistributing to the surface during the evaporation of the solvent. After 4 h the removed, dried, and heated to 900C to pyrolyze the precursor This was repeated eight times and following the last cycle, the composites were given a final sintering treatment at 1200.C for 2 h, which served to crystallize the precursor to the corundum(a) a structure. In this way, the strength of the connection between mullite particles could be increased without any shrinkage of the mullite network taking place Knowing the volume fraction of fibers per unit area of cloth (data obtained from the manufacturer), the volume fraction of fibers in the composite was calculated by measuring the volume of the composite and counting the number of fiber layers in each specimen. The porosity of the composites was measured using the Archimedes technique Studies of the shrinkage of the matrix were done ately by casting thin rods(~2×2×10 flon surface. Linear shrinkage was measure different heat treatments For each step the mens were treated in a way similar to the desired temperature, held for 2 h, and then cooled down at 5°Cmin) Test bar geometries used for(a) in-plane bend testing for flexural rk of fracture, and(c) interlaminar shear strength (2) Mechanical Testing Fiber dominated composite properties were evaluated using ar in-plane three-point flexural test shown in Fig. 1(a). The speci composites. A solution(which also was used in this work) was mens(58 mm long, 3 mm wide, 3.5 mm high) were processed as to place a rubber sheet between the loading pins and the specimen described above and were tested using a loading span of 53 mm urface to minimize stress concentrations 23, 29 Interlaminar shear This configuration and loading mode precluded interlaminar shear strength, T was calculated from the maximum load, Pmax, and test ailure before failure via the tensile stresses on one surface. A bar dimensions using the equation for shear stress at the midplane servoelectric testing machine(Instron, Inc, Model 8562) with a of a flexural bar specimen described by beam theory high stiffness loading frame was used. A crosshead speed of 0.1 mm/min was used. Nylon rods were used as loading pins to reduce contact stress. Strain was calculated based on crosshead displace ment and by correcting for the compliance in the load train. Since the mode of testing is not pure tension, and because the matrix of The tensile stress in the outer fibers in three-point bending is given the composites are known to continually fracture during loader which changes the modulus of the material during testing, the 3/P results of these tests should be considered as qualitative rather than quantitative. Although qualitative, the results will serve to com- pare the different specimens in this study Thus, the midplane shear stress to maximum tensile stress(T /o)is The in-plane notch sensi given by otched specimens shown in Fig. 1(b), 58 mm long and 3 mm wide d 7 mm A notch with a length ao =3.5 mm(nominally alf of the test bar height, ao/w=0.50 0.002 mm)and a width of 0.65 mm was introduced by diamond machining. The net- section notch strength was compared with the unnotched strength To ensure failure by delamination(shear) rather than a tensile to assess the degree of notch sensitivity. Calculations of the energy ailure originating from the surface, the span(s)to thickness ratio uired to break the specimens were also conducted to further s/L, is kept small, Failure modes of the test bars were examined in characterize the work required for fracture. This was done by different microscopes. easuring the area under the load-displacement curve Fracture surfaces and the microstructure of the composites were esults and discussion studied by optical microscopy and scanning electron microscop Interlaminar shear strength (a matrix dominated property (I Composite Matrix and Ce Characteristics these composites) was determined using a short beam shear The results from the shown in Fig. I(c). The specimens were 30 mm long, 5 mm wide summarized in Fig. 2. During and 3 mm high, and the loading span was 15 mm. It has been 1.2% was observed. The change on sintering at bserved that local stress concentrations due to the loading pins temperatures up to 1200.C is -0.9%. Since the processin can give premature failure at low loads in porous oxide matrix temperature is of this order and the observed shrinkage is smal
matrix composite (CMC) body is very flexible and could be shaped much like an epoxy/fiber prepreg. After shaping, the water was removed from the powder matrix by drying at 70°C. An initial sintering treatment was done at 900°C for 2 h to promote the development of alumina bridges between the mullite network. The composites were subsequently impregnated with the alumina precursor solution under vacuum. The impregnation step was performed in a dry nitrogen atmosphere to prevent premature gelation of the precursor due to atmospheric water vapor. The composites were left in the precursor solution for 2 h at atmospheric pressure and transferred into ammoniated water (pH 10) to gel the precursor throughout the body and prevent it from redistributing to the surface during the evaporation of the solvent.39,40 After 4 h the composites were removed, dried, and heated to 900°C to pyrolyze the precursor. This was repeated eight times and following the last cycle, the composites were given a final sintering treatment at 1200°C for 2 h, which served to crystallize the precursor to the corundum () structure. In this way, the strength of the connection between mullite particles could be increased without any shrinkage of the mullite network taking place. Knowing the volume fraction of fibers per unit area of cloth (data obtained from the manufacturer), the volume fraction of fibers in the composite was calculated by measuring the volume of the composite and counting the number of fiber layers in each specimen. The porosity of the composites was measured using the Archimedes technique. Studies of the shrinkage of the matrix slurry were done separately by casting thin rods (2 2 10 mm3 ) of slurry on a Teflon surface. Linear shrinkage was measured after drying and different heat treatments. For each step, the specimens were treated in a way similar to the composites (heated at 5°C/min up to the desired temperature, held for 2 h, and then cooled down at 5°C/min). (2) Mechanical Testing Fiber dominated composite properties were evaluated using an in-plane three-point flexural test shown in Fig. 1(a). The specimens (58 mm long, 3 mm wide, 3.5 mm high) were processed as described above and were tested using a loading span of 53 mm. This configuration and loading mode precluded interlaminar shear failure before failure via the tensile stresses on one surface. A servoelectric testing machine (Instron, Inc., Model 8562) with a high stiffness loading frame was used. A crosshead speed of 0.1 mm/min was used. Nylon rods were used as loading pins to reduce contact stress. Strain was calculated based on crosshead displacement and by correcting for the compliance in the load train. Since the mode of testing is not pure tension, and because the matrix of the composites are known to continually fracture during loading30 which changes the modulus of the material during testing, the results of these tests should be considered as qualitative rather than quantitative. Although qualitative, the results will serve to compare the different specimens in this study. The in-plane notch sensitivity was assessed using the edgenotched specimens shown in Fig. 1(b), 58 mm long and 3 mm wide and 7 mm high. A notch with a length a0 3.5 mm (nominally half of the test bar height, a0/W 0.50 0.002 mm) and a width of 0.65 mm was introduced by diamond machining. The netsection notch strength was compared with the unnotched strength to assess the degree of notch sensitivity. Calculations of the energy required to break the specimens were also conducted to further characterize the work required for fracture. This was done by measuring the area under the load–displacement curve. Fracture surfaces and the microstructure of the composites were studied by optical microscopy and scanning electron microscopy. Interlaminar shear strength (a matrix dominated property for these composites) was determined using a short beam shear test shown in Fig. 1(c). The specimens were 30 mm long, 5 mm wide, and 3 mm high, and the loading span was 15 mm. It has been observed that local stress concentrations due to the loading pins can give premature failure at low loads in porous oxide matrix composites.29 A solution (which also was used in this work) was to place a rubber sheet between the loading pins and the specimen surface to minimize stress concentrations.23,29 Interlaminar shear strength, i , was calculated from the maximum load, Pmax, and test bar dimensions using the equation for shear stress at the midplane of a flexural bar specimen described by beam theory: i 3 4 Pmax bt (1) The tensile stress in the outer fibers in three-point bending is given by 3 2 Ps bt2 (2) Thus, the midplane shear stress to maximum tensile stress ( i / ) is given by i 1 2 t s (3) To ensure failure by delamination (shear) rather than a tensile failure originating from the surface, the span (s) to thickness ratio, s/t, is kept small. Failure modes of the test bars were examined in different microscopes. III. Results and Discussion (1) Composite Matrix and Composite Characteristics The results from the sintering studies of pure matrix rods are summarized in Fig. 2. During drying, a mean linear shrinkage of 1.2% was observed. The additional change on sintering at temperatures up to 1200°C is 0.9%. Since the processing temperature is of this order and the observed shrinkage is small, Fig. 1. Test bar geometries used for (a) in-plane bend testing for flexural strength and elastic modulus, (b) in-plane bend testing for notch sensitivity and work of fracture, and (c) interlaminar shear strength. October 2003 Porous Oxide Matrix Composite Reinforced with Oxide Fibers 1735
1736 Journal of the American Ceramic Sociery-Holmquist and lange Vol. 86. No. 10 Drying shrinkage Panel C Panel D pn 1000 Number of precursor impregnations, N 2乙101 lume fraction Fid, &i M arcx poro imy fer ht: co mpo tes meas rew ciestiall-c a te three of solids in the slurry was 54+ 1%. porosity as derived from Eq (4)in text.) the results indicate that the chosen composition will produce and 6.14 after eight impregnations, compared with 5.67 for the y stable matrix. However. at 1400.C the shrink N720 fiber. 5 This suggests that the N720 fiber and the matrix increases more rapidly, suggesting reduced long-term stability of have similar coefficients of thermal expansion(CTEs)and the the material at temperatures above 1200%C. These measurement esidual stresses might be very small. The composites containing are in agreement with observations made by others made for the N610 fibers had similar crack patterns, strongly suggesting that similar materials the cracks were caused by constrained drying/densification shrink A summary of the composite panels and their fiber content and age and not differential CTE orosity is given in Table I. Fiber volume fraction did not vary a lot between the panels, with values Vr e 40%-42%. The porosity levels were less uniform; the first two manufactured panels(A and B)had an initial matrix porosity of.5%, whereas the last two panels(C and D)had a matrix porosity of -46.5%(Fig 3). This was attributed to a processing improvement made in the vacuum agging step, the vacuum level was reduced from -300 to -10 torr. The change in porosity with the following impregnation/ pyrolysis cycles is shown in Fig. 3. Assuming that all the available voids in the composite are filled in each impregnation cycle, the emaining porosity pr after N cycles should be given by Pm=Pm(1-yp) where Pm is the initial matrix porosity and yp is the volume yield of the precursor solution(measured to 3%). Equation(4)is plotted in Fig. 3 and agrees well with measurements of matrix porosity after three impregnations. The decrease in measured porosity after 100um eight impregnations is lower than the calculations predict, suggest ing the formation of closed pores that would prevent subsequent precursor infiltration. Micrographs of the composite structure reveal were well-infiltrated and only a few large-scale pores produced from trapped air were evident, as shown in Fig 4. As expected, the large pores were more frequent in panels A and B. Cracklike flaws. erpendicular to the fibers and with regular spacing, were also bserved. These were more than likely due to the constraint the fibers impose on the matrix shrinkage during drying and slight densification.28.4 The matrix had an Al, O,SiO, weight ratio of 5.41 before the multiple impregnations with the Al,O,precursor Table I. Summary of Composite Panels Panel 100um A N720 42.1 23.2 42.0 B N720 23.4 23.6 N610 22.0 Vr fiber volume fraction. ' Pe composite porosity. 'p Fig. 4. Microstructural views of cross sections of N720 composite matrix porosity panels:(a) as processed and(b)after 1300.C/(100 h) heat treatment
the results indicate that the chosen composition will produce a microstructurally stable matrix. However, at 1400°C the shrinkage increases more rapidly, suggesting reduced long-term stability of the material at temperatures above 1200°C. These measurements are in agreement with observations made by others made for similar materials.24,28 A summary of the composite panels and their fiber content and porosity is given in Table I. Fiber volume fraction did not vary a lot between the panels, with values Vf 40%–42%. The porosity levels were less uniform; the first two manufactured panels (A and B) had an initial matrix porosity of 51.5%, whereas the last two panels (C and D) had a matrix porosity of 46.5% (Fig. 3). This was attributed to a processing improvement made in the vacuum bagging step; the vacuum level was reduced from 300 to 10 torr. The change in porosity with the following impregnation/ pyrolysis cycles is shown in Fig. 3. Assuming that all the available voids in the composite are filled in each impregnation cycle, the remaining porosity pm, after N cycles should be given by pm pm 0 1 yp N (4) where pm 0 is the initial matrix porosity and yp is the volume yield of the precursor solution (measured to 3%). Equation (4) is plotted in Fig. 3 and agrees well with measurements of matrix porosity after three impregnations. The decrease in measured porosity after eight impregnations is lower than the calculations predict, suggesting the formation of closed pores that would prevent subsequent precursor infiltration. Micrographs of the composite structure reveal that the tows were well-infiltrated and only a few large-scale pores produced from trapped air were evident, as shown in Fig. 4. As expected, the large pores were more frequent in panels A and B. Cracklike flaws, perpendicular to the fibers and with regular spacing, were also observed. These were more than likely due to the constraint the fibers impose on the matrix shrinkage during drying and slight densification.28,41 The matrix had an Al2O3/SiO2 weight ratio of 5.41 before the multiple impregnations with the Al2O3 precursor and 6.14 after eight impregnations, compared with 5.67 for the N720 fiber.35 This suggests that the N720 fiber and the matrix have similar coefficients of thermal expansion (CTEs) and the residual stresses might be very small. The composites containing the N610 fibers had similar crack patterns, strongly suggesting that the cracks were caused by constrained drying/densification shrinkage and not differential CTE. Fig. 2. Linear shrinkage after heat treatments for 2 h at various temperatures of cast mullite/alumina (70/30) powder slurry. The volume fraction of solids in the slurry was 54 1%. Fig. 3. Matrix porosity for the composites measured initially, after three and eight precursor impregnation and pyrolysis cycles. (() indicates porosity as derived from Eq. (4) in text.) Fig. 4. Microstructural views of cross sections of N720 composite panels: (a) as processed and (b) after 1300°C/(100 h) heat treatment. Table I. Summary of Composite Panels Panel Fiber Vf (%)† pc (%)‡ pm (%)§ A N720 42.1 23.2 42.0 B N720 42.4 23.4 43.2 C N610 40.6 23.6 39.8 D N610 39.8 22.0 38.2 † Vf fiber volume fraction. ‡ pc composite porosity. § pm matrix porosity. 1736 Journal of the American Ceramic Society—Holmquist and Lange Vol. 86, No. 10
October 2003 Porous Oxide Matrix Composite Reinforced with Oxide Fibers 1737 Matrix porosity 38. 2%, 1200.C/100 h vDN6101200°c Matrix porosity 38.2%, 1200'C/2h QDN6101200° cooH Matrix porosity 43.2%, 1200.C/100h EEEx三 Matrix porosity 43. 2%, 1200.C/2 h mml Variation in maximun stress with matrix porosity in essed condition and after Is sed spec shear behavior for all-oxide composites with two filled symbols, after agin C(100h) ty levels. An increase in shear strength is seen after 00 h compared with as-processed specimens(sintered that high matrix strength can be obtained by strengthening the SEM observations showed no significant changes in the matrix structure after heat treatment at 1200C for 100 h. However, aging (3) In-Plane Flexure Testing at 1300C/(100 h)promoted an evolution of the flaws caused by matrix densification (Fig. 4(b). Others have reported matri notched composite specimens as proces e n-plane mec erties for un aging. The densification with associated matrix damage at 1200C for longer strength of the N720 specimens was >170 d the n610 periods(1000 h) in similar materials pecimens was >280 MPa in as-processed condition, which is (2) Interlaminar Shear Strength consistent with data presented by Levi, Zok, and co-workers" for the previously described(see Section m(I) method of pro- Typical stress/displacement respons for the inter- cessing porous mullite/alumina matrix composites. Figure 7 shows laminar shear tests are shown in Fig. 5 n initial linear representative stress-strain curves for the rtion, a slight reduction in stiffness rved after the Figure 8 shows fracture surfaces of a N720 composite. The 0 maximum load was obtained, followed by a number of load drops. fiber tows break in a random fashion over a wide range of axial In general the observed behavior is similar to the phenomenon of locations producing fiber brushes. The locations of fracture within equential delamination failure. The failure mode was delani individual tows also show a distribution( Fig. 8(a)). These obser nation in all tests, and the maximum interlaminar shear stresses vations show that the porous matrix is an efficient crack deflector calculated according to Eq. (3)are a measure of the matrix both within and between fiber tows. In more conventional CMcs strength. Interlaminar shear strength as a function of matrix with crack-deflecting matrix/fiber interfaces, one can observe porosity is shown in Fig. 6. These data suggest that the shear holes that contained fibers that fracture within the matrix. In strength increases with decreasing porosity. Heat treatment at composites with porous matrixes, no such holes can be observed 1200.C for 100 h also seems to increase the interlaminar shear and the matrix rather disintegrates into smaller pieces, some of strength. Although no matrix densification could be observed after which still are bonded to the fibers. 23, 24,28 The amount of fiber this heat treatment(see Section Ill(I), the results imply that the ull-out appears to be relatively uniform over the fracture surface, mullite/alumina matrix network has strengthened. The delamina whereas others have observed more coplanar fracture near the material(38 vol%)was tion stress measured for the least porous d wnlolattoni et al2for redistribution of the precursor(used to strengthen the matrix)in edges of the test bar.24 29 This behavior has been explained by 11-12 MPa, which is close to that measure a composite with a more dense matrix (34 vol% and -12 MPa). the absence of a gelling step in the manufacturing process The composite used by Mattoni et al had a matrix with a Thermal stability of the composites was studied by aging mullite/alumina composition of 80/20. This observation suggests specimens in air for 100 h at 1200 and 1300@C and then tested at Table Il. Properties of In-Plane Unnotched Composite Three-Point Bend Tests Flexure strength Elastic modulus Panel Fibe Heat treatment (MPa) Failure modet N720 200°C/2h 163 ABABABCDCDCD N720 200°C/100h) 179 N720 l89 N720 300°C100h) N610 200°C(2h N610 200°C(2h N610 TTTTTTTMMTTT CN6101300°C/(100h) l86 N6101300°C(100h) l81 Failure mode: T, tensile, M, mixed mode(e. g, buckling, delamination)
SEM observations showed no significant changes in the matrix structure after heat treatment at 1200°C for 100 h. However, aging at 1300°C/(100 h) promoted an evolution of the flaws caused by matrix densification (Fig. 4(b)). Others have reported matrix densification with associated matrix damage at 1200°C for longer periods (1000 h) in similar materials.31 (2) Interlaminar Shear Strength Typical stress/displacement responses obtained for the interlaminar shear tests are shown in Fig. 5. After an initial linear portion, a slight reduction in stiffness was observed after the maximum load was obtained, followed by a number of load drops. In general the observed behavior is similar to the phenomenon of sequential delamination failure.42 The failure mode was delamination in all tests, and the maximum interlaminar shear stresses calculated according to Eq. (3) are a measure of the matrix strength. Interlaminar shear strength as a function of matrix porosity is shown in Fig. 6. These data suggest that the shear strength increases with decreasing porosity. Heat treatment at 1200°C for 100 h also seems to increase the interlaminar shear strength. Although no matrix densification could be observed after this heat treatment (see Section III(1)), the results imply that the mullite/alumina matrix network has strengthened. The delamination stress measured for the least porous material (38 vol%) was 11–12 MPa, which is close to that measured by Mattoni et al.29 for a composite with a more dense matrix (34 vol% and 12 MPa). The composite used by Mattoni et al. had a matrix with a mullite/alumina composition of 80/20. This observation suggests that high matrix strength can be obtained by strengthening the network without decreasing porosity. (3) In-Plane Flexure Testing Table II reports the in-plane mechanical properties for unnotched composite specimens as processed and after aging. The strength of the N720 specimens was 170 MPa, and the N610 specimens was 280 MPa in as-processed condition, which is consistent with data presented by Levi, Zok, and co-workers28–30 for the previously described (see Section III(1)) method of processing porous mullite/alumina matrix composites. Figure 7 shows representative stress–strain curves for the composites. Figure 8 shows fracture surfaces of a N720 composite. The 0° fiber tows break in a random fashion over a wide range of axial locations producing fiber brushes. The locations of fracture within individual tows also show a distribution (Fig. 8(a)). These observations show that the porous matrix is an efficient crack deflector both within and between fiber tows. In more conventional CMCs with crack-deflecting matrix/fiber interfaces, one can observe holes that contained fibers that fracture within the matrix.17 In composites with porous matrixes, no such holes can be observed and the matrix rather disintegrates into smaller pieces, some of which still are bonded to the fibers.23,24,28 The amount of fiber pull-out appears to be relatively uniform over the fracture surface, whereas others have observed more coplanar fracture near the edges of the test bar.24,29 This behavior has been explained by redistribution of the precursor (used to strengthen the matrix) in the absence of a gelling step in the manufacturing process. Thermal stability of the composites was studied by aging specimens in air for 100 h at 1200° and 1300°C and then tested at Fig. 5. Short-beam shear behavior for all-oxide composites with two different matrix porosity levels. An increase in shear strength is seen after aging at 1200°C for 100 h compared with as-processed specimens (sintered at 1200°C for 2 h). Fig. 6. Variation in maximum shear stress with matrix porosity in as-processed condition and after aging: open symbols, as-processed specimens; filled symbols, after aging at 1200°C/(100 h). Table II. Properties of In-Plane Unnotched Composite Three-Point Bend Tests Panel Fiber Heat treatment Flexure strength (MPa) Elastic modulus (GPa) Failure mode† A N720 1200°C/(2 h) 163 62 T B N720 1200°C/(2 h) 177 60 T A N720 1200°C/(100 h) 179 65 T B N720 1200°C/(100 h) 189 64 T A N720 1300°C/(100 h) 129 70 T B N720 1300°C/(100 h) 139 71 T C N610 1200°C/(2 h) 289 85 T D N610 1200°C/(2 h) 226 85 M C N610 1200°C/(100 h) 251 92 M D N610 1200°C/(100 h) 276 91 T C N610 1300°C/(100 h) 186 95 T D N610 1300°C/(100 h) 181 96 T † Failure mode: T, tensile; M, mixed mode (e.g., buckling, delamination). October 2003 Porous Oxide Matrix Composite Reinforced with Oxide Fibers 1737
1738 Journal of the American Ceramic Sociery-Holmquist and lange Vol. 86. No. 10 N720,1200°c2h N720.1300°c/00h (a) 300pr Nominal Flexure Strain(%) N610.1200cPh 0.1200°c/100 200 300pm condition and after aging. he mechanical properties of the notched specimens are room temperature. The results are also summarized in table ll and in Table Ill. Net-section strengths were evaluated by Fig. 7. There is no significant effect on either stiffness or strength that the bar dimensions used to calculate the after thermal exposure at 1200C for 100 h. The fracture surfaces ss at failure(Eq (2)do not include the notched portion are similar to those of the as-processed specimens, showing a of the bar; i. e, the length of the notch is subtracted from the height fibrous nature. However, the pull-out lengths of the broken fibers of the beam, and this value is used as the height used to determin appear to be somewhat shorter. Such a trend was also observed by the strength. A measure of the notch sensitivity of the material is Mattoni et al-9 when the matrix strength was increased obtained by comparing the net-section strengths of the notched repeated precursor impregnations to reduce the porosity. The specimens to the strength of the unnotched specimens proposed explanation was that the reduced matrix porosity induced ratio, Table Ill). The strength ratio was -0.75-0.90 for higher stress concentrations around fiber breaks and thereby the ailure probability of adjacent fibers was increased. 4 A similar composites and -0.70-0.85 for the N610 composites moderate notch sensitivity for the material. These values are in behavior was observed for the N610 composites, although the agreement with values reported for similar materials made wit as-processed specimens tended to fail in a mixed mode rather than other processes. ,>o Load-displacement data in Fig 9 shows that n a pure tensile mode the fracture occurs in a stable manor after the load maximum has Despite the higher thermal stability of the n720 fibers, these opposites exhibited a significantly lower strength at room tem- rature compared with those fabricated with N610 fibers, the lower strength of the N720 fibers is expected to be the cause of the Table Ill. Properties of In-Plane Notched Composit lower strength composite Three- Point Bend Tests creasing the heat treatment to 1300C/(100 h) had a more Net-sect ducing strength by about 20% for N720 com- osites and -35% for N610 composites. The degree of fiber Panel Fiber Heat treatment (J/m2) ratio pull-out was significantly reduced, as shown in Fig 8(b). A small AN7201200°C(2h) 6660.9 increase in stiffness could be observed for the N720 composite The N610 fibers are known to loose strength rapidly at high 7201200°C/(100h)1565 temperatures, whic BN7201200°C/(100h)1819 strength. Also the N720 composites are affected, but to less extent CN610 because of the higher thermal stability of N720 fibers compared CN6101200°C/(100h)25420.75 20 with N610 fibers. The heat treatment will also give a denser and DN6101200°C/100h)25800.71 197 ronger matrix as described above
room temperature. The results are also summarized in Table II and Fig. 7. There is no significant effect on either stiffness or strength after thermal exposure at 1200°C for 100 h. The fracture surfaces are similar to those of the as-processed specimens, showing a fibrous nature. However, the pull-out lengths of the broken fibers appear to be somewhat shorter. Such a trend was also observed by Mattoni et al., 29 when the matrix strength was increased by repeated precursor impregnations to reduce the porosity. The proposed explanation was that the reduced matrix porosity induced higher stress concentrations around fiber breaks and thereby the failure probability of adjacent fibers was increased.44 A similar behavior was observed for the N610 composites, although the as-processed specimens tended to fail in a mixed mode rather than in a pure tensile mode. Despite the higher thermal stability of the N720 fibers, these composites exhibited a significantly lower strength at room temperature compared with those fabricated with N610 fibers; the lower strength of the N720 fibers is expected to be the cause of the lower strength composite. Increasing the heat treatment to 1300°C/(100 h) had a more dramatic effect, reducing strength by about 20% for N720 composites and 35% for N610 composites. The degree of fiber pull-out was significantly reduced, as shown in Fig. 8(b). A small increase in stiffness could be observed for the N720 composite. The N610 fibers are known to loose strength rapidly at high temperatures,43 which might explain the decrease in composite strength. Also the N720 composites are affected, but to less extent because of the higher thermal stability of N720 fibers compared with N610 fibers.35 The heat treatment will also give a denser and stronger matrix as described above. In-plane mechanical properties of the notched specimens are reported in Table III. Net-section strengths were evaluated by assuming that the bar dimensions used to calculate the maximum tensile stress at failure (Eq. (2)) do not include the notched portion of the bar; i.e., the length of the notch is subtracted from the height of the beam, and this value is used as the height used to determine the strength. A measure of the notch sensitivity of the material is obtained by comparing the net-section strengths of the notched specimens to the strength of the unnotched specimens (strength ratio, Table III). The strength ratio was 0.75–0.90 for the N720 composites and 0.70–0.85 for the N610 composites, indicating moderate notch sensitivity for the material. These values are in agreement with values reported for similar materials made with other processes.29,30 Load-displacement data in Fig. 9 shows that the fracture occurs in a stable manor after the load maximum has Fig. 7. Flexural stress versus flexural strain plots for unnotched composites reinforced with (a) N720 fibers and (b) N610 fibers, in as-processed condition and after aging. Fig. 8. SEM micrograph of fracture surfaces of N720 composites; (a) as processed, and (b) heat-treated at 1300°C for 100 h. Table III. Properties of In-Plane Notched Composite Three-Point Bend Tests† Panel Fiber Heat treatment WOF (J/m2 ) Strength ratio Net-section strength (MPa) A N720 1200°C/(2 h) 1666 0.90 154 B N720 1200°C/(2 h) 1954 0.77 131 A N720 1200°C/(100 h) 1565 0.76 140 B N720 1200°C/(100 h) 1819 0.91 167 C N610 1200°C/(2 h) 3601 0.83 239 C N610 1200°C/(100 h) 2542 0.75 206 D N610 1200°C/(100 h) 2580 0.71 197 † a0/W 0.50 0.002. 1738 Journal of the American Ceramic Society—Holmquist and Lange Vol. 86, No. 10
October 2003 Porous Oxide Matrix Composite Reinforced with Oxide Fibers 1739 National Materials Advisory Board, Committee on Advanced Fibers for Higl Temperature Ceramic Composites, Publication NMAB-494. National Academy Press, 00∞080 dvanced Hot Gas Filter Development, "Ceram. Eng. Sci. Proc., 0p tme 21(314757(20 R. Warren(Ed ) Comprehensive Composite Materials. Elsevier, Amsterdam, M. van Roode, w. D. Brentnall, K. O Smith, B D, Edwards, J. McClain, and J.R. 忍 Price, "Ceramic Stationary Gas Turbine D nent Program--Fourth Annual 720.1200°c/100h WOF= 1819/n Iz, S. Wittig, and G. Andrees, "Experimental Assessment of Fiber Reinforced Ceramics for Combustor Walls. m. Soc. Mech. WO K. Nishio, K-, Igashira, K. Take, and T. Suemitsu, "Development of a Combustor Liner Composed of Ceramic Matrix Composite( CMC),"Am Soc. Mech. Pa198GT:104(199 A. G. Razzell, M. Holmquist, L. Molliex, and O. Sudre, "Oxide/Oxide Ceramic Matrix Composites in Gas Turbine Combustors, Am. Soc. Mech. Eng,/Pap/ 98GT-30(1998) loW. H. Glime and J. D ""Stress Concentration Due to Fiber-Matrix Fusion in Ceramic-MatrIx s,J.Am. Ceram. Soc., 81[10] 2597-604 (1998) R. H. Jones, C. H. Henger Jr, C. A. Lewinsohn, and C. F. Windisch Jr. ig. 9. Load versus displacement plots for notched composites reinforced racking of Silicon Carbide Fiber/Silicon Carbide Composi with N720 or N610 fibers, in as-processed condition and after aging J. Am. Ce Recent Developments in Fibers and Interphases for High Matrix Composites, "Composites, Part A, 30, 429-37(1999) D. B. Marshall, J. B. Davis, P. E. D. Morgan, and J.R. Porter,"Interface Materials for Damage-Tolerant Oxide Composites, Key Eng. Mater, 127-13 been reached and, consequently, the area under the curve can be used as a measure of the work of fracture(WOF). WOF was Butler, and L. Al-Dawery, "Development of Interfaces measured to-16-2.0 kJ/m? for the N720 composite. Correspond. in spide Matrix. ” Key Eng Mater.,l64-165,351-56(1999) ing values for the N610 composite were -2.5-3.6 kJ/m". A slight Fiber Coatings fo Oxide CFCC, Ceram. Eng. Sci. Proc., 18 [3] 279-86 decrease in WoF could be seen after heat treatment at 1200%C fo (1997) 100 h. as shown in Table Ill. IbM. K. Cinibulk, "Hexaluminates as a Cleavable Fiber-Matrix Interphase: Syn- Development, and Phase Compatibility, "J. Eur. Ceram. Soc., 20, 17M. Holmquist, R Lundberg, O. Sudre, A. G.Razzell, L. Molliex,J.Benoit, IV. Conclusions J. Adlerborn,Alumina/Alumina Composite with a Porous Zirconia Interphak& ocessing, Properties and Component Testing, J. Eur. Ceram Soc., 20, 599-606 a process to manufacture porous oxide matrix/po oxide fiber composites was developed and evaluated. Th O Sudre, A. G. Razzell, L. Molliex, and M. Holmquist, "Alumina Single-Crystal Fibre Reinforced Alumina Matrix for Combustor Tiles. Ceran Eng. Sci. Proc. 19 uses a preconsolidated slurry with a very high volume f powder to infiltrate fiber cloths. These infiltrated fiber "M. J. O'Brien and B. W. Sheldon, "Porous Alumina Coating with Tailored be frozen and used latter to fabricate engineering shapes. Mechan- Fracture Resistance for Alumina Composites,J. Am. Ceram Soc., 82[12]3567-74 ical property measurements suggest that the processing method (299). used here was comparable to porous oxide matrix composites 20K. A. Keller, T-I. Mah, T. A. Parthasarathy and C. M. Cooke, F Interfacial Carbon Coatings for Oxide/Oxide Composites,J. Am. Ceram Soc., 832) manufactured by other processes using the same fibers. and mullite/alumina N20 fibers showed nonbrittle fracture behav- Ceramic Compos叫Em,℃m以Dm,m In-plane flexural tests of composites based on alumina N610 and strengths of >280 and >170 MPa, respectively, and moderate notch sensitivity. Interlaminar shear strength, which is F F. Lange, C G. Levi, and F. w. Zok, "Processing Fiber Reinforced Ceramics dominated by the porous matrix, ranged between 7 and 12 MPa for s,2000 matrix porosity ranging from 38% to 43%, respectively roth, and F. F. Lange, "Processing and Properties of an The composites possessed good thermal stability; the micro- All-Oxide Composite with a Porous Matrix,J. Eur. Cera. Soc., 20, 607 structure was stable after aging at 1200 C for 100 h, showing no C. G. Levi, F. W. Zok, J-Y. Yang, M. Mattoni, and J. P. A. Lofvander visible signs of densification. This heat treatment was found to Microstructural Design of Stable Porous Matrices for All-Oxide Ceramic Compos- slightly increase the interlaminar shear strength, which was ites” Z Metalled,90121037-47(1999) attributed to a strengthening of the matrix network, and was F. F. Lange, w. C. Tu, and A. G. Evans, U.S. Pat. No. 5856 252 accompanied by a reduction in composite toughness. However, Several the 1200 C/(100 h) treatment did not significantly change the Ceram. Eng. Sci. Proc., 19 [3]327-39(1 composite strength or strain to failure. Heat treatment at 1300C J.99GT-190(1999) for 100 h reduced the strength for the N610 and N720C G Levi, J-Y. Yang, B J Dalgleish, F. W Zok, and A G. Evans,"Processin composites by 35% and 20%, respectively, and increased their 2M. Mattoni, J -Y. Yang, C. G. Levi, and F. W. Zok, "Effects of a Precursor ived Alumina on the Mechanical Properties of a Porous-Matrix, All-Oxide Ceramic Composite,JAm, Ceram Soc., in review Acknowledgments In-Plane Mechanical Properties of an Al-Oxide Ceramic Composite, "J.Am. Ceram We thank R. Harrysson for SEM work and Professor F, w. Zok, Dr. J.-Y. Yang Soc,82]2721-301999 nd M. Mattoni for useful discussion E.A. V.Carelli, H Fujita, J. Y. Yang, and F. w. Zok,"Effects of Thermal Aging on the Mechanical Properties of a Porous-Matrix Ceramic Composite,"in revie 32G. W. Franks and F. F. Lange, "Plastic Clay-like Flow Stress of Saturated Advanced Ceramic Powder Compacts,".. Eur. Cera. Soc., in press References 3M. Holmquist, T C. Radsick, O. Sudre, FF. Lange, and F. W. Zok, "Fabrication esting of All-Oxide CFCC IC. P. Beesley, "The Application of CMC's in High Integrity Gas Turbine ange, T. C. Radsick, Engines, Key Eng Mater, 127-131, 165-74(1997 Control of Microstructure and Properties", Pp. 587-99 in Proceedings of the 4th H. Knabe, and F, Strobel, "Development and Testing of C/SiC omponents for Liquid Rocket Propulsion Applications, ALAA Pap, 99-2896 Edited by W. Krenkel, R. Naslain, and H, Schne eramic Matrix Composites International Conference on High Temperature Wiley-VCH, New Yor
been reached and, consequently, the area under the curve can be used as a measure of the work of fracture (WOF). WOF was measured to 1.6–2.0 kJ/m2 for the N720 composite. Corresponding values for the N610 composite were 2.5–3.6 kJ/m2 . A slight decrease in WOF could be seen after heat treatment at 1200°C for 100 h, as shown in Table III. IV. Conclusions A process to manufacture porous oxide matrix/polycrystalline oxide fiber composites was developed and evaluated. The method uses a preconsolidated slurry with a very high volume fraction of powder to infiltrate fiber cloths. These infiltrated fiber cloths can be frozen and used latter to fabricate engineering shapes. Mechanical property measurements suggest that the processing method used here was comparable to porous oxide matrix composites manufactured by other processes using the same fibers. In-plane flexural tests of composites based on alumina N610 and mullite/alumina N720 fibers showed nonbrittle fracture behavior and strengths of 280 and 170 MPa, respectively, and moderate notch sensitivity. Interlaminar shear strength, which is dominated by the porous matrix, ranged between 7 and 12 MPa for matrix porosity ranging from 38% to 43%, respectively. The composites possessed good thermal stability; the microstructure was stable after aging at 1200°C for 100 h, showing no visible signs of densification. This heat treatment was found to slightly increase the interlaminar shear strength, which was attributed to a strengthening of the matrix network, and was accompanied by a reduction in composite toughness. However, the 1200°C/(100 h) treatment did not significantly change the composite strength or strain to failure. Heat treatment at 1300°C for 100 h reduced the strength for the N610 and N720 composites by 35% and 20%, respectively, and increased their brittle nature. Acknowledgments We thank R. Harrysson for SEM work and Professor F. W. Zok, Dr. J.-Y. Yang, and M. Mattoni for useful discussions. References 1 C. P. Beesley, “The Application of CMC’s in High Integrity Gas Turbine Engines,” Key Eng. Mater., 127-131, 165–74 (1997). 2 S. Beyer, H. Knabe, and F. Strobel, “Development and Testing of C/SiC Components for Liquid Rocket Propulsion Applications,” AIAA Pap., 99-2896 (1999). 3 National Materials Advisory Board, Committee on Advanced Fibers for High Temperature Ceramic Composites, Publication NMAB-494. National Academy Press, Washington, DC, 1998. 4 T. J. McMahon, “Advanced Hot Gas Filter Development,” Ceram. Eng. Sci. Proc., 21 [3] 47–57 (2000). 5 R. Warren (Ed.), Comprehensive Composite Materials. Elsevier, Amsterdam, Netherlands, 2000. 6 M. van Roode, W. D. Brentnall, K. O. Smith, B. D. Edwards, J. McClain, and J. R. Price, “Ceramic Stationary Gas Turbine Development Program—Fourth Annual Summary,” Am. Soc. Mech. Eng., [Pap.], 97-GT-317 (1997). 7 D. Filsinger, S. Mu¨nz, A. Schulz, S. Wittig, and G. Andrees, “Experimental Assessment of Fiber Reinforced Ceramics for Combustor Walls,” Am. Soc. Mech. Eng., [Pap.], 97-GT-154 (1997). 8 K. Nishio, K.-I. Igashira, K. Take, and T. Suemitsu, “Development of a Combustor Liner Composed of Ceramic Matrix Composite (CMC),” Am. Soc. Mech. Eng., [Pap.], 98-GT-104 (1998). 9 A. G. Razzell, M. Holmquist, L. Molliex, and O. Sudre, “Oxide/Oxide Ceramic Matrix Composites in Gas Turbine Combustors,” Am. Soc. Mech. Eng., [Pap.], 98-GT-30 (1998). 10W. H. Glime and J. D. Cawley, “Stress Concentration Due to Fiber–Matrix Fusion in Ceramic–Matrix Composites,” J. Am. Ceram. Soc., 81 [10] 2597–604 (1998). 11R. H. Jones, C. H. Henager Jr., C. A. Lewinsohn, and C. F. Windisch Jr., “Stress-Corrosion Cracking of Silicon Carbide Fiber/Silicon Carbide Composite,” J. Am. Ceram. Soc., 83 [8] 1999–2005 (2000). 12R. E. Tressler, “Recent Developments in Fibers and Interphases for High Temperature Ceramic Matrix Composites,” Composites, Part A, 30, 429–37 (1999). 13D. B. Marshall, J. B. Davis, P. E. D. Morgan, and J. R. Porter, “Interface Materials for Damage-Tolerant Oxide Composites,” Key Eng. Mater., 127–131, 27–36 (1997). 14M. H. Lewis, A. Tye, E. Butler, and I. Al-Dawery, “Development of Interfaces in Oxide Matrix Composites,” Key Eng. Mater., 164–165, 351–56 (1999). 15R. W. Goettler, S. Sambasivan, and V. P. Dravid, “Isotropic Complex Oxides as Fiber Coatings for Oxide–Oxide CFCC,” Ceram. Eng. Sci. Proc., 18 [3] 279–86 (1997). 16M. K. Cinibulk, “Hexaluminates as a Cleavable Fiber-Matrix Interphase: Synthesis, Texture Development, and Phase Compatibility,” J. Eur. Ceram. Soc., 20, 569–82 (2000). 17M. Holmquist, R. Lundberg, O. Sudre, A. G. Razzell, L. Molliex, J. Benoit, and J. Adlerborn, “Alumina/Alumina Composite with a Porous Zirconia Interphase— Processing, Properties and Component Testing,” J. Eur. Ceram. Soc., 20, 599–606 (2000). 18O. Sudre, A. G. Razzell, L. Molliex, and M. Holmquist, “Alumina Single-Crystal Fibre Reinforced Alumina Matrix for Combustor Tiles,” Ceram. Eng. Sci. Proc., 19 [4] 273–80 (1998). 19M. J. O’Brien and B. W. Sheldon, “Porous Alumina Coating with Tailored Fracture Resistance for Alumina Composites,” J. Am. Ceram. Soc., 82 [12] 3567–74 (1999). 20K. A. Keller, T.-I. Mah, T. A. Parthasarathy, and C. M. Cooke, “Fugitive Interfacial Carbon Coatings for Oxide/Oxide Composites,” J. Am. Ceram. Soc., 83 [2] 329–36 (2000). 21W.-C. Tu, F. F. Lange, and A. G. Evans, “Concept for a Damage Tolerant Ceramic Composite with ”Strong“ Interfaces,” J. Am. Ceram. Soc., 79 [2] 417–24 (1996). 22F. F. Lange, C. G. Levi, and F. W. Zok, “Processing Fiber Reinforced Ceramics with Porous Matrices”; Ch. 14 in Comprehensive Composite Materials. Edited by R. Warren. Elsevier, Amsterdam, Netherlands, 2000. 23J. J. Haslam, K. E. Berroth, and F. F. Lange, “Processing and Properties of an All-Oxide Composite with a Porous Matrix,” J. Eur. Ceram. Soc., 20, 607–18 (2000). 24C. G. Levi, F. W. Zok, J.-Y. Yang, M. Mattoni, and J. P. A. Lo¨fvander, “Microstructural Design of Stable Porous Matrices for All-Oxide Ceramic Composites,” Z. Metallkd., 90 [12] 1037–47 (1999). 25F. F. Lange, W. C. Tu, and A. G. Evans, U.S. Pat. No. 5 856 252, 1999. 26L. P. Zawada, “Longitudinal and Transthickness Tensile Behavior of Several Oxide/Oxide Composites,” Ceram. Eng. Sci. Proc., 19 [3] 327–39 (1998). 27R. A. Jurf and S. C. Butner, “Advances in Oxide–Oxide CMC,” Am. Soc. Mech. Eng., [Pap.], 99-GT-190 (1999). 28C. G. Levi, J.-Y. Yang, B. J. Dalgleish, F. W. Zok, and A. G. Evans, “Processing and Performance of an All-Oxide Ceramic Composite,” J. Am. Ceram. Soc., 81 [8] 2077–86 (1998). 29M. Mattoni, J.-Y. Yang, C. G. Levi, and F. W. Zok, “Effects of a Precursor Derived Alumina on the Mechanical Properties of a Porous-Matrix, All-Oxide Ceramic Composite,” J. Am. Ceram. Soc., in review. 30J. A. Heathcote, X.-Y. Gong, J.-Y. Yang, U. Ramamurty, and F. W. Zok, “In-Plane Mechanical Properties of an All-Oxide Ceramic Composite,” J. Am. Ceram. Soc., 82 [10] 2721–30 (1999). 31E. A. V. Carelli, H. Fujita, J. Y. Yang, and F. W. Zok, “Effects of Thermal Aging on the Mechanical Properties of a Porous-Matrix Ceramic Composite,” in review. 32G. W. Franks and F. F. Lange, “Plastic Clay-like Flow Stress of Saturated Advanced Ceramic Powder Compacts,” J. Eur. Ceram. Soc., in press. 33M. Holmquist, T. C. Radsick, O. Sudre, F. F. Lange, and F. W. Zok, “Fabrication and Testing of All-Oxide CFCC Tubes,” to be submitted. 34F. F. Lange, T. C. Radsick, and M. Holmquist, “Oxide/Oxide Composites: Control of Microstructure and Properties”; pp. 587–99 in Proceedings of the 4th International Conference on High Temperature Ceramic Matrix Composites. Edited by W. Krenkel, R. Naslain, and H. Schneider. Wiley-VCH, New York, 2001. Fig. 9. Load versus displacement plots for notched composites reinforced with N720 or N610 fibers, in as-processed condition and after aging. October 2003 Porous Oxide Matrix Composite Reinforced with Oxide Fibers 1739
1740 Journal of the American Ceramic Sociery-Holmquist and lange Vol. 86. No. 10 35D. M. Wilson, S. L. Lieder, and D. C. Lueneberg, "Microstructure and Hi w-C. Tu and F. F. Lange, "Liquid Precursor Infiltration Processing of Powder loosprature properties of Nextel 720 Fibers, Ceram. Eng- Sa. Proc, I6 Compacts: IL, Fracture Toughness and Strength,J. Am. Ceram. Soc., 78[1 D. M. Wilson, ""Statistical Tensile Strength of Nextel 610 and 720 Fibres, D. C. C. Lam and F. F. Lange. ""Microstructural Observations on Constrained J.Maer.Sci,32,2535-42(1997) Periodic Array of Sapphire Fibers, 37M. Colic, G. Franks, M. Fisher, and F. F. Lange,"Chemisorption of Organo- JAm. Ceram.soe,771976-78(1994) functional Silanes on Silicon Nitride for Improved Aqueous Processing,J.Am. 4J. B. Davis, A. Kristoffersson, E. Carlstrom, and W.J. Clegg,"Fabrication and Ceram,Soc,8182157-63(1998 Crack Deflection in Ceramic Laminates with Porous Interlayers, J. An. Ceram. Soc. Colic, and F. Lange, "Effect of the 83[10]2369-74(2000) erparticle Pair Potential on the Rheological Behavior of Zirconia Powders: Il, The 3J. He and D. R. Clarke."Determination of Fiber Strength Distribution from Bundl ptical Fluorescence,Proc. R Soc. London, Ser. A, 453, 1881-901(1997) M. Ibnabdeljalil and w.A. Curtin,"Strength and Reliability of Fiber-Reinforce Kinetic Studies and Microstructure Development, " J. A Ceram. Soc. Localized Load-sharing and Associated Size Effects, Int. J. Solids 782]3277-82(1995) Saat,34[21]12649-68(1997)
35D. M. Wilson, S. L. Lieder, and D. C. Lueneberg, “Microstructure and High Temperature Properties of Nextel 720 Fibers,” Ceram. Eng. Sci. Proc., 16 [5] 1005–14 (1995). 36D. M. Wilson, “Statistical Tensile Strength of Nextel 610 and 720 Fibres,” J. Mater. Sci., 32, 2535–42 (1997). 37M. Colic, G. Franks, M. Fisher, and F. F. Lange, “Chemisorption of Organofunctional Silanes on Silicon Nitride for Improved Aqueous Processing,” J. Am. Ceram. Soc., 81 [8] 2157–63 (1998). 38S. Klein, M. Fisher, G. Franks, M. Colic, and F. Lange, “Effect of the Interparticle Pair Potential on the Rheological Behavior of Zirconia Powders: II, The Influence of Chem-Adsorbed Silanes,” J. Am. Ceram. Soc., in press. 39W.-C. Tu and F. F. Lange, “Liquid Precursor Infiltration Processing of Powder Compacts: I, Kinetic Studies and Microstructure Development,” J. Am. Ceram. Soc., 78 [12] 3277–82 (1995). 40W.-C. Tu and F. F. Lange, “Liquid Precursor Infiltration Processing of Powder Compacts: II, Fracture Toughness and Strength,” J. Am. Ceram. Soc., 78 [12] 3283–89 (1995). 41D. C. C. Lam and F. F. Lange, “Microstructural Observations on Constrained Densification of Alumina Powder Containing a Periodic Array of Sapphire Fibers,” J. Am. Ceram. Soc., 77 [7] 1976–78 (1994). 42J. B. Davis, A. Kristoffersson, E. Carlstro¨m, and W. J. Clegg, “Fabrication and Crack Deflection in Ceramic Laminates with Porous Interlayers,” J. Am. Ceram. Soc., 83 [10] 2369–74 (2000). 43J. He and D. R. Clarke, “Determination of Fiber Strength Distribution from Bundle Tests Using Optical Fluorescence,” Proc. R. Soc. London, Ser. A, 453, 1881–901 (1997). 44M. Ibnabdeljalil and W. A. Curtin, “Strength and Reliability of Fiber-Reinforced Composites: Localized Load-Sharing and Associated Size Effects,” Int. J. Solids Struct., 34 [21] 2649–68 (1997). 1740 Journal of the American Ceramic Society—Holmquist and Lange Vol. 86, No. 10