E驅≈3S Journal of the European Ceramic Society 20(2000)607-618 Processing and properties of an all-oxide composite with a porous matrix .. Haslam, K.e. berroth*ff la Materials Department, University of California at Santa Barbara, Santa Barbara, CA 93106, US.A 18 August 199 Abstract Processing and mechanical properties of an all-oxide fiber composite with a porous matrix are presented here. The processing approach for an all-oxide composite was developed to be simple and involve one sintering process. The composite uses a porous matrix instead of riser coatings to deflect cracks from the fibers. a processing method involving recently developed methods for reshaping and forming saturated high-volume fraction(> 50 vol%) particle bodies was used to form the composite Good infiltra- tion of the woven fiber tows was obtained. Sintering in a pure HCl gas atmosphere was used to produce a porous matrix without shrinkage during processing. The sintering process also produced coarsening which makes the microstructure stable against densi- fication during use and thereby prevents forming cracklike voids and retains sufficient porosity for crack deflection. Measurements of interlaminar shear strength and strength of the composite show that composite produced by this processing method is compar- able to previous all-oxide materials produced using the oxide fibers used here. The mechanical properties are rationalized in terms of the features on the fracture surfaces. Disintegration of the matrix to allow energy dissipation during fracture was apparent and correlates with the measurements of the fracture toughness of the material. Moderate notch insensitivity was demonstrated with a net section strength in the presence of a notch being 700% of the unnotched strength. c 2000 Elsevier Science Ltd. All rights reserved Keywords: Aluminosilicate fibres; Composites; Mechanical properties; Porosity: Sintering: ZrO2 matrix 1. Introduction matrix. A fiber within a good CMC is only expected to break when the applied load exceeds its strength An important property of any ceramic matrix com- In the late 1960s Phillips recognized that brittle, but posite(CMC) is that its strength should be relatively strong fibers could be isolated from one another within insensitive to the presence of notches. If the riders a brittle matrix by providing a path for cracks prova within a CMC are effective, the strength of a body with gating through the matrix to bypass the fibers. a"weak a notch(or hole)of any size or shape will be the same as interface between the matrix and fiber provides the path the unnotched strength of a body with same net (or for crack deflection, thus allowing the crack to propa- reduced) cross-section. That is, for an ideal CMC, one gate along the fiber /matrix interface instead of through drill the fiber. As described by He and Huchinson, the con reducing the failure load other than the effect of redu- dition for crack deflection depends on the ratio of the cing its cross sectional critical strain energy release rate for the interface and Since the failure strain of a strong fiber is generally fiber, and the elastic properties of the two materials. For much larger than a dense matrix, cracks generally first a few fiber/matrix combinations, the crack defecting extend within the matrix. In terms of crack extension, interface needs no special processing conditions. For notch insensitivity requires that the fibers must be iso- example. the carbon fibers in the CMCs produced by lated from the very high stress field of a crack within the Phillips et al. did not bond to the glass matrix. For most other CMCs. the fibers must be coated with either carbon or boron nitride films to achieve a crack deflecting inter- Researcher. High Performance face. Not only do fiber coatings introduce cost and proces- Ceramic Section, Swiss Federal Laboratories for Materials Testing sing complexity, but they are not stable in oxidizing and Research. EMPA. Dube environments and they can cause composite embrittlement 0955-2219/00/S- see front matter o 2000 Elsevier Science Ltd. All rights reserved PII:S0955-2219(99)00259-9
Processing and properties of an all-oxide composite with a porous matrix J.J. Haslam, K.E. Berroth*, F.F. Lange Materials Department, University of California at Santa Barbara, Santa Barbara, CA 93106, USA Accepted 18 August 1999 Abstract Processing and mechanical properties of an all-oxide ®ber composite with a porous matrix are presented here. The processing approach for an all-oxide composite was developed to be simple and involve one sintering process. The composite uses a porous matrix instead of riser coatings to de¯ect cracks from the ®bers. A processing method involving recently developed methods for reshaping and forming saturated high-volume fraction (>50 vol%) particle bodies was used to form the composite. Good in®ltration of the woven ®ber tows was obtained. Sintering in a pure HCl gas atmosphere was used to produce a porous matrix without shrinkage during processing. The sintering process also produced coarsening which makes the microstructure stable against densi- ®cation during use and thereby prevents forming cracklike voids and retains sucient porosity for crack de¯ection. Measurements of interlaminar shear strength and strength of the composite show that composite produced by this processing method is comparable to previous all-oxide materials produced using the oxide ®bers used here. The mechanical properties are rationalized in terms of the features on the fracture surfaces. Disintegration of the matrix to allow energy dissipation during fracture was apparent and correlates with the measurements of the fracture toughness of the material. Moderate notch insensitivity was demonstrated with a net section strength in the presence of a notch being 700% of the unnotched strength. # 2000 Elsevier Science Ltd. All rights reserved. Keywords: Aluminosilicate ®bres; Composites; Mechanical properties; Porosity; Sintering; ZrO2 matrix 1. Introduction An important property of any ceramic matrix composite (CMC) is that its strength should be relatively insensitive to the presence of notches. If the riders within a CMC are eective, the strength of a body with a notch (or hole) of any size or shape will be the same as the unnotched strength of a body with same net (or reduced) cross-section. That is, for an ideal CMC, one should be able to drill a hole without signi®cantly reducing the failure load other than the eect of reducing its cross sectional area. Since the failure strain of a strong ®ber is generally much larger than a dense matrix, cracks generally ®rst extend within the matrix. In terms of crack extension, notch insensitivity requires that the ®bers must be isolated from the very high stress ®eld of a crack within the matrix. A ®ber within a good CMC is only expected to break when the applied load exceeds its strength. In the late 1960s Phillips1 recognized that brittle, but strong ®bers could be isolated from one another within a brittle matrix by providing a path for cracks propagating through the matrix to bypass the ®bers. A `weak' interface between the matrix and ®ber provides the path for crack de¯ection, thus allowing the crack to propagate along the ®ber/matrix interface instead of through the ®ber. As described by He and Huchinson,2 the condition for crack de¯ection depends on the ratio of the critical strain energy release rate for the interface and ®ber, and the elastic properties of the two materials. For a few ®ber/matrix combinations, the crack defecting interface needs no special processing conditions. For example. the carbon ®bers in the CMCs produced by Phillips et al. did not bond to the glass matrix. For most other CMCs, the ®bers must be coated with either carbon or boron nitride ®lms to achieve a crack de¯ecting interface. Not only do ®ber coatings introduce cost and processing complexity, but they are not stable in oxidizing environments and they can cause composite embrittlement. 0955-2219/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved. PII: S0955-2219(99)00259-9 Journal of the European Ceramic Society 20 (2000) 607±618 * Corresponding author. Visiting Researcher, High Performance Ceramic Section, Swiss Federal Laboratories for Materials Testing and Research, EMPA, DuÈbendorf, Switzerland
J. Haslam et al. Journal of the European Ceramic Society 20(2000)607-618 Cracking phenomena for the tensile loading of a uni within the fibers and matrix is identical. It is possible directional CMC containing crack deflecting interfaces that the failure strain(Em) of the porous matrix can be can be related to the composite's stress/strain behavior. equal or even larger than the failure strain of the fibers During the initial loading, the behavior is linear and (Ee). Applying Hook's Law, one can show that characterized by the combined elastic modulus of the fiber and matrix weighted by the appropriate volume Er= E (1) raction of each. As loading proceeds, matrix cracking initiates without fiber failure. Matrix cracking is char- where Emf and Em f are the failure stress and elastic acterized by a decreased slope of the stress-strain curve. modulus of the matrix(m) and fibers(f). Using proper Multiple matrix cracking generally occurs prior to the ties of a low density Al2O3 matrix material and Al2O3 initiation of fiber failure. Prior to and during fiber fail- fibers(e.g. 0m 200 MPa, o 2000 MPa, Emx40 GPa ure, the cracked matrix is held together by the fibers, and Er= 400 GPa) we can see that it is reasonable to be which now supports nearly all of the applied load. Fiber able to fabricate a porous matrix with a failure strain failure, and thus CMC failure occurs at a high strain that approaches that of a strong fiber. Therefore, in (0.5 to 1.0%), indicative of a strong fiber with a low tension, a large fraction of the strength and strain to elastic modulus. For many commercial and experi- failure of strong fibers can be achieved in a ceramic mental CMCs the stress for matrix cracking lies between composite that contains a porous matrix 40 and 100 MPa whereas composite failure(fiber fail- The second role of the porous matrix is to allow fibers ure)does not occur until the stresses exceed 150 to 300 to be isolated from cracks within the matrix In porous MPa. Thus, CMCs that have been developed over the materials the crack front can be non-continuous and last 25 years to contain crack deflecting interfaces can crack extension must occur by the continued breaking not only be relatively notch insensitive, but they can of the solid phase units, i.e. fracture has to be reinitiated also exhibit higher strains to failure relative to mono- in the solid phase within the high stress field of the pro- lithic ceramics(e.g Si3N4, with a mean tensile strength pagating crack. A comparable example of this fracture of 1000 MPa, has a strain to failure of 0.3%) mode is the extension of a crack within cloth. where the Approximately 5 years ago another type of CMC was fracture of each fiber is independent of the last to fail inadvertently discovered. 3.4 Unlike the CMCs with This mode of crack extension occurs in powder com- weak ' fiber matrix interfaces, the matrix and fibers are pacts that have been heated to produce necks between bonded together in these ' new'CMCs. The second touching particles. Observing the fr acture change is that the matrix in the 'new'CMCs is purpo- these very porous materials one can see that fracture(or sely made to be porous. Despite these two major chan- crack extension,)occurred by the breaking of grain ges, both of which are not taught by mechanics of pairs at grain boundaries. A continuous crack front conventional CMCs, the new CMC with well-bonded does not exist in these porous materials fiber/matrix interfaces and porous matrixes are notch The lack of a crack front in a porous matrix means insensitive. In addition, although not as high as the that embedded fibers never see an extending crack front conventional CMCs, their failure strain is larger than as the matrix fails. Fiber fracture within a very porous conventional monolithic ceramics. 5 matrix must initiate within the fiber itself. ie. from flaws The mode of failure of these new composites is dif- either on the surface or within the fiber, and not by the ferent from the older CMCs. The stress/strain behavior propagation of a crack within the matrix. Thus, fibers in of tensile specimens is nearly linear to failure, indicating a very porous matrix can fracture in the same manner as that both the matrix and fibers fail at about the same they do when they exist as a bundle, without a matrix. failure strain. Tensile failure can occur at 200 MPa: The high failure strain of the fibers becomes the failure CMngth is also relatively notch insensitive. 5 These strain of the composite because the matrix will have a oxide matrix and can be very stable in air to tempera- As detailed elsewhere, -5 one method to produce the tures where the fibers begin to degrade. It can be fiber composites described above is to pack particles expected that the processing of the new CMCs is much around the fibers within a fiber preform by pressure fil less complex and less costly. Eliminating fiber coatings is tration and then strengthen the porous matrix. In this a significant advantage in processing and reducing cost. method, a fiber preform(3-D weave, stacked layers of The porous matrix appears to play a critical role in cloth, etc. ) is mounted on a filter within a die cavity. A achieving a notch insensitive strength and a high failure pressure is exerted to a dispersed slurry to cause the strain. One role concerns the strain to failure. When a particles to stream though the preform to become trap- composite is loaded in tension, the fibers will support ped at the filter, and to build up a consolidated layer much of the load due to their much larger elastic mod- within the fiber preform. The slurry must be formulated ulus relative to the porous matrix. Although the fiber such that the particles are repulsive with respect to must carry the major portion of the stress, the strain themselves and the fibers. The particles must also be
Cracking phenomena for the tensile loading of a unidirectional CMC containing crack de¯ecting interfaces can be related to the composite's stress/strain behavior. During the initial loading, the behavior is linear and characterized by the combined elastic modulus of the ®ber and matrix weighted by the appropriate volume fraction of each. As loading proceeds, matrix cracking initiates without ®ber failure. Matrix cracking is characterized by a decreased slope of the stress±strain curve. Multiple matrix cracking generally occurs prior to the initiation of ®ber failure. Prior to and during ®ber failure, the cracked matrix is held together by the ®bers, which now supports nearly all of the applied load. Fiber failure, and thus CMC failure occurs at a high strain (0.5 to 1.0%), indicative of a strong ®ber with a low elastic modulus. For many commercial and experimental CMCs the stress for matrix cracking lies between 40 and 100 MPa whereas composite failure (®ber failure) does not occur until the stresses exceed 150 to 300 MPa. Thus, CMCs that have been developed over the last 25 years to contain crack de¯ecting interfaces can not only be relatively notch insensitive, but they can also exhibit higher strains to failure relative to monolithic ceramics (e.g. Si3N4, with a mean tensile strength of 1000 MPa, has a strain to failure of 0.3%). Approximately 5 years ago another type of CMC was inadvertently discovered.3,4 Unlike the CMCs with `weak' ®ber matrix interfaces, the matrix and ®bers are bonded together in these `new' CMCs. The second change is that the matrix in the `new' CMCs is purposely made to be porous. Despite these two major changes, both of which are not taught by mechanics of conventional CMCs, the new CMC with well-bonded ®ber/matrix interfaces and porous matrixes are notch insensitive. In addition, although not as high as the conventional CMCs, their failure strain is larger than conventional monolithic ceramics.5 The mode of failure of these new composites is different from the older CMCs.4 The stress/strain behavior of tensile specimens is nearly linear to failure, indicating that both the matrix and ®bers fail at about the same failure strain.5 Tensile failure can occur at 200 MPa; the strength is also relatively notch insensitive.5 These new CMCs can be processed with oxide ®bers in an oxide matrix and can be very stable in air to temperatures where the ®bers begin to degrade. It can be expected that the processing of the new CMCs is much less complex and less costly. Eliminating ®ber coatings is a signi®cant advantage in processing and reducing cost. The porous matrix appears to play a critical role in achieving a notch insensitive strength and a high failure strain. One role concerns the strain to failure. When a composite is loaded in tension, the ®bers will support much of the load due to their much larger elastic modulus relative to the porous matrix. Although the ®ber must carry the major portion of the stress, the strain within the ®bers and matrix is identical. It is possible that the failure strain ("m) of the porous matrix can be equal or even larger than the failure strain of the ®bers ("f). Applying Hook's Law, one can show that "f "m; or f Ef m "m ; 1 where "m;f and Em,f are the failure stress and elastic modulus of the matrix (m) and ®bers (f). Using properties of a low density Al2O3 matrix material and Al2O3 ®bers (e.g. m200 MPa, f2000 MPa, Em40 GPa and Ef 400 GPa) we can see that it is reasonable to be able to fabricate a porous matrix with a failure strain that approaches that of a strong ®ber. Therefore, in tension, a large fraction of the strength and strain to failure of strong ®bers can be achieved in a ceramic composite that contains a porous matrix. The second role of the porous matrix is to allow ®bers to be isolated from cracks within the matrix. In porous materials the crack front can be non-continuous and crack extension must occur by the continued breaking of the solid phase units, i.e. fracture has to be reinitiated in the solid phase within the high stress ®eld of the propagating crack. A comparable example of this fracture mode is the extension of a crack within cloth, where the fracture of each ®ber is independent of the last to fail. This mode of crack extension occurs in powder compacts that have been heated to produce necks between touching particles. Observing the fracture surface of these very porous materials one can see that fracture (or `crack extension') occurred by the breaking of grain pairs at grain boundaries.6 A continuous crack front does not exist in these porous materials. The lack of a crack front in a porous matrix means that embedded ®bers never see an extending crack front as the matrix fails. Fiber fracture within a very porous matrix must initiate within the ®ber itself. i.e. from ¯aws either on the surface or within the ®ber, and not by the propagation of a crack within the matrix. Thus, ®bers in a very porous matrix can fracture in the same manner as they do when they exist as a bundle, without a matrix. The high failure strain of the ®bers becomes the failure strain of the composite because the matrix will have a comparable strain to failure. As detailed elsewhere,3±5 one method to produce the ®ber composites described above is to pack particles around the ®bers within a ®ber preform by pressure ®ltration and then strengthen the porous matrix. In this method, a ®ber preform (3-D weave, stacked layers of cloth, etc.) is mounted on a ®lter within a die cavity. A pressure is exerted to a dispersed slurry to cause the particles to stream though the preform to become trapped at the ®lter, and to build up a consolidated layer within the ®ber preform. The slurry must be formulated such that the particles are repulsive with respect to themselves and the ®bers. The particles must also be 608 J.J. Haslam et al. / Journal of the European Ceramic Society 20 (2000) 607±618
J. Haslam et al. Journal of the European Ceramic Society 20(2000)607-618 much smaller than the fiber diameter to ensure good tows produced by the 3M Corporation( St Paul, MN) particle packing. 7 To avoid large cracklike voids from Each tow nominally contains 420 fibers. The Nextel 720 developing within the matrix, the powder should not fiber is an experimental fiber composed of a mixture of densify during subsequent heat treatments and at applica- submicron alumina and mullite grains. The two inter tion temperatures. 8.9 For this reason, in our previous work penetrating phases ensure a small grain size during pro- we used a mullite powder that did not begin to shrink until cessing. The mullite in the fiber contributes to high A1300oC, the maximum fiber application temperature. creep resistance compared to a similar all-alumina fiber After removing the liquid via evaporation, the powder (Nextel 610). The strength of the Nextel 720 fiber matrix was strengthened by infiltrating the composite with about 30% less than the 610 for single filament proper a solution containing precursor molecules. After evapor- ties, but it was selected because of its greater creep ating the liquid, heating causes the precursor molecules to resistance. I decompose to form an inorganic material that bonds the as detailed below laminated ceramic cloth was infil particles together. The inorganic phase that bonds and trated with a previously consolidated mixture of 70 strengthens the powder matrix also bonds the particles vol% cubic zirconia(solid solution with 8 mol%Y203 (matrix)to the fibers. Cyclic solution precursor infiltration, TZ8YS, Toso Ceramics, average particle diameter of evaporation, and decomposition further strengthens the(0.4 um) and 30 vol% mullite(MU-107, Showa Denko) matrix phase. Care must be taken to avoid precursor As detailed below and elsewhere, zirconia was used as molecules from migrating to the surface during evapora- the matrix because it can be sintered, without shrinkage, tion. This condition produces surface cracking during when heat treated in HCI at temperatures as low as drying due to a thin layer of precursor molecules that 1 C. 3 Mullite was introduced because previous form on the surface. 0 An all-oxide. fiber reinforced cera- work has shown that mullite does not allow the sintered mic composite can be processed in this method. Extensive and coarsened zirconia to shrink(densify)after expo- mechanical testing by Levi et al. has shown that this sure to air at 1200 C for 100 h.6 The zirconia was com- type of composite can exhibit a significant notch insen- osed of agglomerated particles which contained sitive strength in tensile loading. It also has all the primary particles of 50-100 nm in diameter. Infrared attributes found for fiber reinforced ceramics fabricated spectroscopy indicated that the mullite contained an with dense matrixes and weak fiber/matrix interfaces organic contaminant that had to be removed from the Here we report a much simpler and less time consuming powder before it was formulated as a slurry. A 10 h heat fabrication method for processing these new CMCs with treatment in air at 800C was sufficient to remove the porous matrices. In the new method, the powder is treated contaminant. The particle size(average=0.7um) did not to produce a special interparticle pair potential which change during the heat treatment. llows the powder compact(previously consolidated by Dispersed, aqueous slurries containing 20 vol% of the pressure filtration)to be fluidized. It can then be formed two powders were formed by adding 1.3 vol% poly nto a thin sheet by vibrating between plastic sheets. The ethylene oxide urethane silane(PEg-silane, Gelest, Inc) plastic sheets help to avoid evaporation and the con- at pH 10.5. This was found to be sufficient to coat the sequent drying of the thin particle layers. The ceramic particle surfaces. As detailed elsewhere, the PEG-silane sheet is then frozen to enable removal from between the molecules chem-adsorb to the particles by reacting with plastic sheets. The frozen ceramic sheet of powder is the -M-OH(M=metal atom) surface sites. 14, 15The then sandwiched between sheets of ceramic fibers(e.g. zirconia slurry was attrition milled for 15 min after the woven cloth). After thawing, the powder sheet is flui- powder was added. The mullite slurry was sonicated dized by vibrating. It then flows to surround all fibers in with an ultrasonic horn for 5 min prior to the final the adjacent fiber sheets After evaporation, the powder adjustment of the pH. Tetraethylammonium chloride urrounding the fibers can be made strong either by the TEACI)salt(0. 1 molar) was added to form weakly use of precursors described above or by an HCl eva- attractive pair potentials between the particles. As poration/condensation treatment described below. Pre- detailed elsewhere, TMA+ counter ions aid in produ- liminary mechanical measurements show that this new cing a weakly attractive particle network which can be route can result in similar properties as the previous packed to a high density via pressure filtration and route to manufacture CMCs with porous matrices allow the consolidated body to be fluidized via vibra- tion 16-18 TEA+ counter ions were used in this work these counter ions are slightly larger than TMA. Other 2. Experimental methods can be used to produce weakly attractive net- works such as surfactants or chemi-sorption of alco- 2.1. Composite processing hols. 9 The PEG-silane plus TEACI approach was appropriate here due to the two different powders used Composites were formed from layers of two dimen- to form a composite slurry The two slurries were mixed sional, 8 harness woven cloth of Nextel M 720 fiber in appropriate portions described above. The mixed
much smaller than the ®ber diameter to ensure good particle packing.7 To avoid large, cracklike voids from developing within the matrix, the powder should not densify during subsequent heat treatments and at application temperatures.8,9 For this reason, in our previous work we used a mullite powder that did not begin to shrink until 1300C, the maximum ®ber application temperature. After removing the liquid via evaporation, the powder matrix was strengthened by in®ltrating the composite with a solution containing precursor molecules. After evaporating the liquid, heating causes the precursor molecules to decompose to form an inorganic material that bonds the particles together. The inorganic phase that bonds and strengthens the powder matrix also bonds the particles (matrix) to the ®bers. Cyclic solution precursor in®ltration, evaporation, and decomposition further strengthens the matrix phase. Care must be taken to avoid precursor molecules from migrating to the surface during evaporation. This condition produces surface cracking during drying due to a thin layer of precursor molecules that form on the surface.10 An all-oxide, ®ber reinforced ceramic composite can be processed in this method. Extensive mechanical testing by Levi et al.5 has shown that this type of composite can exhibit a signi®cant notch insensitive strength in tensile loading. It also has all the attributes found for ®ber reinforced ceramics fabricated with dense matrixes and weak ®ber/matrix interfaces. Here we report a much simpler and less time consuming fabrication method for processing these new CMCs with porous matrices. In the new method, the powder is treated to produce a special interparticle pair potential which allows the powder compact (previously consolidated by pressure ®ltration) to be ¯uidized. It can then be formed into a thin sheet by vibrating between plastic sheets. The plastic sheets help to avoid evaporation and the consequent drying of the thin particle layers. The ceramic sheet is then frozen to enable removal from between the plastic sheets. The frozen ceramic sheet of powder is then sandwiched between sheets of ceramic ®bers (e.g. woven cloth). After thawing, the powder sheet is ¯uidized by vibrating. It then ¯ows to surround all ®bers in the adjacent ®ber sheets After evaporation, the powder surrounding the ®bers can be made strong either by the use of precursors described above or by an HCl evaporation/condensation treatment described below. Preliminary mechanical measurements show that this new route can result in similar properties as the previous route to manufacture CMCs with porous matrices. 2. Experimental 2.1. Composite processing Composites were formed from layers of two dimensional, 8 harness woven cloth of NextelTM 720 ®ber tows produced by the 3M Corporation (St. Paul, MN). Each tow nominally contains 420 ®bers. The Nextel 720 ®ber is an experimental ®ber composed of a mixture of submicron alumina and mullite grains. The two interpenetrating phases ensure a small grain size during processing. The mullite in the ®ber contributes to high creep resistance compared to a similar all-alumina ®ber (Nextel 610). The strength of the Nextel 720 ®ber is about 30% less than the 610 for single ®lament properties,11 but it was selected because of its greater creep resistance.12 As detailed below, laminated ceramic cloth was in®ltrated with a previously consolidated mixture of 70 vol% cubic zirconia (solid solution with 8 mol% Y2O3, TZ8YS, Toso Ceramics, average particle diameter of (0.4 mm) and 30 vol% mullite (MU-107, Showa Denko). As detailed below and elsewhere, zirconia was used as the matrix because it can be sintered, without shrinkage, when heat treated in HCl at temperatures as low as 1100C.13 Mullite was introduced because previous work has shown that mullite does not allow the sintered and coarsened zirconia to shrink (densify) after exposure to air at 1200C for 100 h.6 The zirconia was composed of agglomerated particles which contained primary particles of 50±100 nm in diameter. Infrared spectroscopy indicated that the mullite contained an organic contaminant that had to be removed from the powder before it was formulated as a slurry. A 10 h heat treatment in air at 800C was sucient to remove the contaminant. The particle size (average=0.7mm) did not change during the heat treatment. Dispersed, aqueous slurries containing 20 vol% of the two powders were formed by adding 1.3 vol% polyethylene oxide urethane silane (PEG-silane, Gelest, Inc.) at pH 10.5. This was found to be sucient to coat the particle surfaces. As detailed elsewhere, the PEG-silane molecules chem-adsorb to the particles by reacting with the -M±OH (M=metal atom) surface sites.14,15 The zirconia slurry was attrition milled for 15 min after the powder was added. The mullite slurry was sonnicated with an ultrasonic horn for 5 min prior to the ®nal adjustment of the pH. Tetraethylammonium chloride TEACl) salt (0.1 molar) was added to form weakly attractive pair potentials between the particles. As detailed elsewhere, TMA+ counter ions aid in producing a weakly attractive particle network which can be packed to a high density via pressure ®ltration and allow the consolidated body to be ¯uidized via vibration.16±18 TEA+ counter ions were used in this work; these counter ions are slightly larger than TMA+. Other methods can be used to produce weakly attractive networks such as surfactants or chemi-sorption of alcohols.19 The PEG-silane plus TEACI approach was appropriate here due to the two dierent powders used to form a composite slurry The two slurries were mixed in appropriate portions described above. The mixed J.J. Haslam et al. / Journal of the European Ceramic Society 20 (2000) 607±618 609
J. Haslam et al. Journal of the European Ceramic Society 20(2000)607-618 slurry was then consolidated by pressure filtration at 5 wiched between two layers of woven, fiber cloth. Sand- MPa to form disc shaped bodies that were fully satu- wiches with up to 27 layers, (14 weaves and 13 frozen rated with water. The saturated bodies were stored in ceramic tapes) were piled up, packed in plastic, evac- sealed plastic bags containing a small paper towel satu- uated, and sealed in plastic. After thawing, the assem- rated with water to help prevent drying. The volume bled layers were vibrated and pressed lightly in between fraction of powder within the saturated, consolidated two steel plates with appropriate spacers to cause the bodies was determined by weight difference method as fluidized powder to flow and intrude the finer layers 52%. 5 At a later time the consolidated powder com- Multi-layer composites in sizes of 40x 100x 3 mm could pact (or a portion cut with a razor blade) was placed be fabricated by this vibration supported single step between two plastic sheets(e. g. a bag) and fluidized with impregnation. The multi-layer cloth composite could an air-powered vibrator into uniform 300 um thick then be dried or frozen for later use. It should be noted esof consolidated particles that the layers of fiber cloth, impregnated with the flui- An illustration of the composite processing steps is dized powder ceramic compact as described were very shown in Fig. 1. Initially, tapes are formed by pressing flexible and could be shaped much like a sheet of un- the fluidized sheets in between two flat steel plates using cross-linked carbon fiber/epoxy prepreg wo spacer bars to fix the thickness. The pressed tape Further processing requires removing the water from were flexible due to the weakly attractive particle poten- the saturated powder matrix by drying in an oven at tial. The tapes, still between the plastic sheets, were fro- 70C, and then sintering the ZrO 2 in a dry HCl gas zen to facilitate composite processing and/or storage environment at temperatures between 1200 and To produce the composite, the frozen tapes were 1300 C20 As reported elsewhere the HCI gas heat removed from between the plastic sheets and sand- treatment did not affect the strength of fiber bundles With the knowledge of the volume of fibers per unit area of cloth the volume fraction of foors within the Vibro Impregnation Process composite was determined by measuring the volume of with Tape Freezing the composite and counting the number of fiber layers sll ach specimen. For composites fabricated for this ly, the average volume fraction of fibers was 0.37±0.02 2.2 Interlaminar shear tests queeze For some design considerations, a desirable property of a woven, layered composite is to have sufficient interlaminar shear strength to resist delamination. This type of failure might be encountered in a bending type pile + pack of loading through the thickness as encountered with a through-thickness temperature gradient. Interlaminar shear strength was determined with 0/90 bar speci mens(3.5x7x20 mm nominal dimensions) diamond cut from larger plates fabricated with 12 or more cloth lay ers. The specimen edges were diamond ground(400 grit) vibrate squeeze to remove a minimum 300 um of damage introduced by the diamond cutting. 3-Point flexural tests were the span was changed for reasons dis- cussed below. The fiber weave orientation wa horizontal with the loading in the vertical direction as layered composite shown in Fig. 2(a). Nylon rods(6.45 mm diameter)were Matrix material Fiberweave 2 Impregnated composite 口 Steel plates o Plastic baa (b) Fig. 2. Schematic of fiber orientation of composite for bending tests. Fig 1. This illustration shows the processing steps used to form the (a)Interlaminar Shear Strength tests.(b)In-plane bend testing composite. flexural strength and elastic modulus
slurry was then consolidated by pressure ®ltration at 5 MPa to form disc shaped bodies that were fully saturated with water. The saturated bodies were stored in sealed plastic bags containing a small paper towel saturated with water to help prevent drying. The volume fraction of powder within the saturated, consolidated bodies was determined by weight dierence method as 52%.15 At a later time, the consolidated powder compact (or a portion cut with a razor blade) was placed between two plastic sheets (e.g. a bag) and ¯uidized with an air-powered vibrator into uniform 300 mm thick `tapes' of consolidated particles. An illustration of the composite processing steps is shown in Fig. 1. Initially, tapes are formed by pressing the ¯uidized sheets in between two ¯at steel plates using two spacer bars to ®x the thickness. The pressed tapes were ¯exible due to the weakly attractive particle potential. The tapes, still between the plastic sheets, were frozen to facilitate composite processing and/or storage. To produce the composite, the frozen tapes were removed from between the plastic sheets and sandwiched between two layers of woven, ®ber cloth. Sandwiches with up to 27 layers, (14 weaves and 13 frozen ceramic tapes) were piled up, packed in plastic, evacuated, and sealed in plastic. After thawing, the assembled layers were vibrated and pressed lightly in between two steel plates with appropriate spacers to cause the ¯uidized powder to ¯ow and intrude the ®ner layers. Multi-layer composites in sizes of 401003 mm could be fabricated by this vibration supported single step impregnation. The multi-layer cloth composite could then be dried or frozen for later use. It should be noted that the layers of ®ber cloth, impregnated with the ¯uidized powder ceramic compact as described were very ¯exible and could be shaped much like a sheet of uncross-linked carbon ®ber/epoxy prepreg. Further processing requires removing the water from the saturated powder matrix by drying in an oven at 70C, and then sintering the ZrO2 in a dry HCl gas environment at temperatures between 1200 and 1300C.20 As reported elsewhere21 the HCI gas heat treatment did not aect the strength of ®ber bundles. With the knowledge of the volume of ®bers per unit area of cloth, the volume fraction of ¯oors within the composite was determined by measuring the volume of the composite and counting the number of ®ber layers in each specimen. For composites fabricated for this study, the average volume fraction of ®bers was 0.370.02. 2.2. Interlaminar shear tests For some design considerations, a desirable property of a woven, layered composite is to have sucient interlaminar shear strength to resist delamination. This type of failure might be encountered in a bending type of loading through the thickness as encountered with a through-thickness temperature gradient. Interlaminar shear strength was determined with 0/90 bar specimens (3.5720 mm nominal dimensions) diamond cut from larger plates fabricated with 12 or more cloth layers. The specimen edges were diamond ground (400 grit) to remove a minimum 300 mm of damage introduced by the diamond cutting. 3-Point ¯exural tests were performed where the span was changed for reasons discussed below. The ®ber weave orientation was horizontal with the loading in the vertical direction as shown in Fig. 2(a). Nylon rods (6.45 mm diameter) were Fig. 1. This illustration shows the processing steps used to form the composite. Fig. 2. Schematic of ®ber orientation of composite for bending tests. (a) Interlaminar Shear Strength tests. (b) In-plane bend testing for ¯exural strength and elastic modulus. 610 J.J. Haslam et al. / Journal of the European Ceramic Society 20 (2000) 607±618
J. Haslam et al. Journal of the European Ceramic Society 20(2000)607-618 used as loading pins to accommodate the inherent composite strength in bending without interlaminar roughness of the woven fibers on the specimen surface shear type failures. Strain was calculated based on and to reduce contact loading stress. No permanent measurements of the bottom beam displacement. In deformation of the loading rods was observed after addition, based on a technique used by Heathcote et testing, which indicates that they remained elastic under al., 22 notched bending tests were performed in a similar the stresses encountered in testing. A servo-electric test- manner. Two diferent bar specimens(3.5x7x90 mm ing machine (Instron, Inc. model 8562) with a high and 3. 5x7x45 mm nominal dimensions)were tested in stifness frame was used to load the specimens at a cross 3-point flexural load with an outer span of 35 mm and head speed of 0. 1 mm/ min 89 mm for the shorter and longer specimens respec The shear stress at the mid-plane of a flexural bar tively. The cross head displacement rate was 0. 1 mm/ pecimen can be calculated from beam theory as min. two different fiber with0°/90° and one with+/-45° fiber directions rela r=3/4L/(W*1) (2) tive to the direction of the bar lengtH Notches were cut in the center of bars with a diamond where L=load. w=width, and t= thickness The mea wheel with a resulting notch thickness of 0.55 mm. The surement of delamination stress for CMCs usually pre- notch depth was nominally one half of the sample scribes a span(S)to thickness ratio(S/ n)of greater than height, a/W=0.485+0.015. The high stifness of the 10 to help insure that the specimen fails by delamination testing machine/load cell, and the non-catastrophic fail- (shear)rather than a tensile failure on the surface, given ure of the specimens allowed for careful measurement of the energy required to break the notched specimens The measured projected surface area of the sample was 3/2PS (3) used to calculate energy per unit area to produce frac- ture The stress-strain response of the un-notched compo- The midplane shear stress to maximum tensile stress sites was nearly linear elastic with some significant ratio(t/o)is given by deviations in the +/-45 fiber direction tests. Due to the low loads encountered in testing specimens of this r/a=1/2(/S) (4) length, no damage was observed at the loading po The specimens were tested in the same high-stifiness Therefore, for a given specimen thickness, the shorter universal testing machine used for the interlaminar the span, the greater the probability that failure will shear strength tests take place by a delamination of the cloth layers, rather than crack extension through the layers. Flexural testing with small values of S/t is called short beam bend test esults ing and is used to characterize the interlaminar shear trength 3. 1. Interlaminar shear strength 2.3. In-plane flexure testing of notched and un-notched Fig. 3 reports the apparent interlaminar shear strength as a function of span to thickness ratio(S/n) for individual CMC specimens heat treated in HCI for dif- Attempts to perform tensile tests on 100 mm long ferent time periods and temperatures. As shown, the specimens(same material as above) with a reduced delamination stress was 10+2 MPa for all heat treat gauge section(5. 1 mm wide, 40 mm long, produced with ments, and that specimens produced from one heat a 152 mm diameter diamond grinding wheel) were not treatment(1250C/5 h) failed in tension and did not successful with our limited amount of material. Despite delaminate. Fig. 3 also reports the delamination stress the use of fiberglass tabs that were epoxied to the ends reported by Levi et al. for a CMC with a porous of the specimen and double knife-edge universal joints matrix, but fabricated by the older method(pressure within the tensile train, most specimens failed either in filtration, multiple precursor infiltration and pyrolysis the non-reduced gauge section or adjacent to the cycles ). Its delamination strength of 8 MPa is a little clamping grip lower than most of the values reported for our newer Because of the limited amount of fiber cloth available method but Levi et al. used harder, steel loading pins, for fabricating specimens, the implementation of an which could have produced a stress concentration and a improved tensile test was not possible. The testing mode lower delamination stress. was changed to an in-plane flexural test, subjected to 3- Figs. 4 and 5 illustrate typical stress versus strain plot point flexural loading as shown in Fig. 2(b). This con- for specimens that delaminated prior to tensile failure figuration and loading mode allowed for testing of the In general, one or two load drops were observed similar
used as loading pins to accommodate the inherent roughness of the woven ®bers on the specimen surface and to reduce contact loading stress. No permanent deformation of the loading rods was observed after testing, which indicates that they remained elastic under the stresses encountered in testing. A servo-electric testing machine (Instron, Inc. model 8562) with a high stiness frame was used to load the specimens at a cross head speed of 0.1 mm/min. The shear stress at the mid-plane of a ¯exural bar specimen can be calculated from beam theory as: 3=4 L= Wt; 2 where L=load, w=width, and t=thickness. The measurement of delamination stress for CMCs usually prescribes a span (S) to thickness ratio (S/t) of greater than 10 to help insure that the specimen fails by delamination (shear) rather than a tensile failure on the surface, given by 3=2PS bt2 : 3 The midplane shear stress to maximum tensile stress ratio (t/s) is given by = 1=2 t=S: 4 Therefore, for a given specimen thickness, the shorter the span, the greater the probability that failure will take place by a delamination of the cloth layers, rather than crack extension through the layers. Flexural testing with small values of S/t is called short beam bend testing and is used to characterize the interlaminar shear strength. 2.3. In-plane ¯exure testing of notched and un-notched specimens Attempts to perform tensile tests on 100 mm long specimens (same material as above) with a reduced gauge section (5.1 mm wide, 40 mm long, produced with a 152 mm diameter diamond grinding wheel) were not successful with our limited amount of material. Despite the use of ®berglass tabs that were epoxied to the ends of the specimen and double knife-edge universal joints within the tensile train, most specimens failed either in the non-reduced gauge section or adjacent to the clamping grip. Because of the limited amount of ®ber cloth available for fabricating specimens, the implementation of an improved tensile test was not possible. The testing mode was changed to an in-plane ¯exural test, subjected to 3- point ¯exural loading as shown in Fig. 2(b). This con- ®guration and loading mode allowed for testing of the composite strength in bending without interlaminar shear type failures. Strain was calculated based on measurements of the bottom beam displacement. In addition, based on a technique used by Heathcote et al.,22 notched bending tests were performed in a similar manner. Two dierent bar specimens (3.5790 mm and 3.5745 mm nominal dimensions) were tested in 3-point ¯exural load with an outer span of 35 mm and 89 mm for the shorter and longer specimens respectively. The cross head displacement rate was 0.1 mm/ min. Two dierent ®ber alignments were tested, one with 0/90 and one with +/ÿ45 ®ber directions relative to the direction of the bar length. Notches were cut in the center of bars with a diamond wheel with a resulting notch thickness of 0.55 mm. The notch depth was nominally one half of the sample height, a/W=0.4850.015. The high stiness of the testing machine/load cell, and the non-catastrophic failure of the specimens allowed for careful measurement of the energy required to break the notched specimens. The measured projected surface area of the sample was used to calculate energy per unit area to produce fracture. The stress±strain response of the un-notched composites was nearly linear elastic with some signi®cant deviations in the +/ÿ45 ®ber direction tests. Due to the low loads encountered in testing specimens of this length, no damage was observed at the loading points. The specimens were tested in the same high-stiness universal testing machine used for the interlaminar shear strength tests. 3. Results 3.1. Interlaminar shear strength Fig. 3 reports the apparent interlaminar shear strength as a function of span to thickness ratio (S/t) for individual CMC specimens heat treated in HCI for different time periods and temperatures. As shown, the delamination stress was 102 MPa for all heat treatments, and that specimens produced from one heat treatment (1250C/5 h) failed in tension and did not delaminate. Fig. 3 also reports the delamination stress reported by Levi et al.5 for a CMC with a porous matrix, but fabricated by the older method (pressure ®ltration, multiple precursor in®ltration and pyrolysis cycles). Its delamination strength of 8 MPa is a little lower than most of the values reported for our newer method but Levi et al. used harder, steel loading pins, which could have produced a stress concentration and a lower delamination stress.5 Figs. 4 and 5 illustrate typical stress versus strain plot for specimens that delaminated prior to tensile failure. In general, one or two load drops were observed similar J.J. Haslam et al. / Journal of the European Ceramic Society 20 (2000) 607±618 611
J. Haslam et al. Journal of the European Ceramic Society 20(2000)607-618 to the phenomenon of sequential, delamination failure A slight reduction in stiffness was observed near the Tensile Failures peak loads. At the first load drop, cracking was usually cracks became apparent after loading beyond the initial load drop Levi et al, referene 3. 2. In-plane flexure testing 200 C1 hr. /HCI Table I reports the in-plane mechanical properties for 1200 C/5 hrs /HCI un-notched 00/90 and +/-45 oriented composite 1050C/2 hrs /Air, 1250C/5 hrs. /HCI pecimens processed in HCI for different conditions The strength of the 0/90 specimens was >160 MPa Span to Thickness 9 10 and the strength of the +/-45o specimens was >80 MPa consistent with those reported by Levi et al. for the pre- Fig3. Plot of interlaminar shear strength against span to vious method of processing the porous matrix composites thickness ratio for a variety of samples. The 1250C/5 h/HCI samples Fig. 6 shows representative stress-strain curves for 00/90 failed in a tensile failure mode which implies that the interlaminar and +/-450 composites(1250C/5h/HCD); their respec shear strength is greater than these values tive failure strains were 0.25 and 0.2% Table 2 reports the in-plane mechanical properties of the notched specimens 0/90 and +1-45 oriented composite specimens processed with the same condi tions as the un-notched specimens (Table 1). Strength values are reported for a'net cross-sectional area'spe cimen, i.e. assuming that the bar dimensions used to E:5 calculate the maximum tensile stress at failure [Eq. (2) does not include the notched portion of the bar. As shown, the strength of the notched bars, based on the net cross-sectional area is 70% of the un-notched bars, regardless of fiber orientation. Fig. 7 shows the load/ displacement curves for representative 0 /90 and +/-45 notched specimens. It can be seen that the failure was not catastrophic, and the work of fracture, reported 0,0% 20% in Table 2, could be determined by measurements of the Nominal flexure strain area under the curve. The work of fracture for the 0o/ 90 specimens was 3-4 times greater than the +/-45o orientation sintered in H 200c strength test result: 0/90 fiber oriented Fig 8, a photograph of one of fractured, notched 0/90 oriented specimen, illustrates that interlocking fibers still hold the bar together after the applied load is removed. It was observed that both notched and un-notched specimens did not fall apart after the load was removed Table I Properties of in-plane un-notched composite 3-point bend tests Fiber Flexure Strength Modulu let stopped GPa 1250°c/5h/HCl 1250°C/5h/HCl 1200°c/5h/HCl Fig. 5. "Nominal Interlaminar shear strength test result: 0/90 fiber +/-45 1250°C/5h/HCl +/-45 orientation sintered in Hcl at 12500C. 5h 1250°C/5b/HCl
to the phenomenon of sequential, delamination failure.23 A slight reduction in stiness was observed near the peak loads. At the ®rst load drop, cracking was usually not evident under low magni®cation (2±3X), but the cracks became apparent after loading beyond the initial load drop. 3.2. In-plane ¯exure testing Table 1 reports the in-plane mechanical properties for un-notched 0/90 and +/ÿ45 oriented composite specimens processed in HCI for dierent conditions. The strength of the 0/90 specimens was >160 MPa, and the strength of the +/ÿ45 specimens was >80 MPa consistent with those reported by Levi et al.5 for the previous method of processing the porous matrix composites. Fig. 6 shows representative stress±strain curves for 0/90 and +/ÿ45 composites (1250C/5h/HCl); their respective failure strains were 0.25 and 0.2%. Table 2 reports the in-plane mechanical properties of the notched specimens 0/90 and +/ÿ45 oriented composite specimens processed with the same conditions as the un-notched specimens (Table 1). Strength values are reported for a `net cross-sectional area' specimen, i.e. assuming that the bar dimensions used to calculate the maximum tensile stress at failure [Eq. (2)] does not include the notched portion of the bar. As shown, the strength of the notched bars, based on the `net cross-sectional area' is 70% of the un-notched bars, regardless of ®ber orientation. Fig. 7 shows the load/displacement curves for representative 0/90 and +/ÿ45 notched specimens. It can be seen that the failure was not catastrophic, and the work of fracture, reported in Table 2, could be determined by measurements of the area under the curve. The work of fracture for the 0/ 90 specimens was 3±4 times greater than the +/ÿ45 oriented specimens. Fig. 8, a photograph of one of the fractured, notched 0/90 oriented specimen, illustrates that interlocking ®bers still hold the bar together after the applied load is removed. It was observed that both notched and un-notched specimens did not fall apart after the load was removed. Fig. 4. `Nominal' Interlaminar shear strength test result: 0/90 ®ber orientation sintered in HCl at 1200C, 5s. Fig. 3. Plot of `nominal' interlaminar shear strength against span to thickness ratio for a variety of samples. The 1250C/5 h/HCl samples failed in a tensile failure mode which implies that the interlaminar shear strength is greater than these values. Fig. 5. `Nominal' Interlaminar shear strength test result: 0/90 ®ber orientation sintered in HCl at 1250C, 5h. Table 1 Properties of in-plane un-notched composite 3-point bend tests Fiber Sintering Flexure Strength MPa Elastic Modulus GPa 0/90 1250C/5 h/HCl 165 64.4 0/90 1250C/5 h/HCl 168 61.6 0/90 1200C/5 h/HCl 152 50.2 +/ÿ45 1250C/5 h/HCl 85 45.9 +/ÿ45 1250C/5 h/HCl 88 41.9 612 J.J. Haslam et al. / Journal of the European Ceramic Society 20 (2000) 607±618
J. Haslam et al. Journal of the European Ceramic Society 20(2000)607-618 35 30 150 0/90°a/W=0.48 0/9D° WOF= 1373) 20 15 4+/-45°,a/W=049 飞wNOF=38/m2 0%0.05%0.1%0.15%0.2%0.25%0.3% 0.5 Flexure strain Displacement (mm) Fig. 6. Flexure stress versus Nominal Flexure Strain plot showing Fig. 7. Load versus displacement plots of notched composite sampl elastic modulus of composite with final sintering at 1250'C for 5 h in tested in the in-plane configuration. Final sintering of composite was HCL. The composite was tested in the in-plane fiber orientation. Tests at 1250C for 5 h in HCl. Composite samples with 0/90 and +/-45o for fiber weave orientations of 0/90 and +/-450 are shown. fiber orientations are shown. The normalized notch depth for these samples was afw=0.48 and 0.49, respectively. Table 2 Properties of in-plane notched composite 3-point bend tests Fiber WOF J/mm2 Strength ratio Net-section stress MPa KI at peak load MPam/2 Ks MPam/2 1250°C/5hHC11373 0.69 115 1250C/5 h/HCI 3.89 /-45°1250c/5h/HC1438 0.72 1250°C/5h/HCl 338 53.9 3.8 W≈0.49 3.3. Composite microstructure many fiber tows contained pairs of fibers that exhibited this frequent fracture origin. Fig. 9 illustrates the fracture region of a 0/90 speci- Fig 9(a)and(b)shows the planar fracture of the 90 men. Fig. 9(a)and(b) show that fibers in 0 tows in fiber tows in each cloth layer. Here, the crack top each cloth layer exhibit random failure to produce graphy can be characterized as nearly planar, and the brushes,, or what is known as ' fiber pull-out. Since no surface of this 'planar' region contains relatively long holes are observed in the matrix from where the fibers lengths of the 0 fibers extending from the fracture sur- could have pulled from, it must be concluded that por- face [see arrow in Fig 9(b). Fig. 10 shows the fracture ous matrix between the fibers disintegrated into smaller surface of the +/-45 oriented specimens. Large areas pieces, and that fibers did not slide out of matrix holes are seen where the crack path propagated along the as observed for composites produced with'weak'inter- cylindrical fiber surfaces(arrows); these cracks jog faces. Some of the matrix debris, and matrix still bon- across fiber tows. The surface produced when the crack ded to the fibers are seen in Fig 9(c). Fig. 9(d) shows propagates across the tow certainly does not form fib that some fracture regions in the 0 tows have a flatter, 'brushes, but as shown in the enlarged view, Fig. 10(b). more coordinated fracture topography. Close examina- it can be seen that the same crack front did not cause all tion of the fibers in this region shows that most of them fibers to fail fracture on different planes, indicating that one crack Generally, as shown in Figs. 9 and 10, the ZrO2 and front did not cause this fracture topography. One can mullite matrix fills all of the interstices between the see a few pairs of fibers(arrows )which do exhibit planar fibers. As detailed elsewhere, the HCl treatment at fracture; examination of these fiber pairs shows that temperatures between 1200 and 1300C is effective in they have a common fracture origin where they touch. producing a strong matrix without shrinkage via an As detailed elsewhere, 2I this common fracture origin evaporation/ condensation sintering and coarsening was produced during fiber processing. This occurs when phenomenon for the Zro, portion of the matrix. The adjacent fibers in the bundle (all floors are spun from lack of shrinkage of the powder matrix is evident in solution concurrently) stick to each other and sinter Figs. 9 and 10 by the lack of crack-like voids in the together along their cylindrical axis. It was shown that matrix. If the powder matrix were to densify its shrinkage
3.3. Composite microstructure Fig. 9 illustrates the fracture region of a 0/90 specimen. Fig. 9(a) and (b) show that ®bers in 0 tows in each cloth layer exhibit random failure to produce `brushes', or what is known as `®ber pull-out'. Since no holes are observed in the matrix from where the ®bers could have pulled from, it must be concluded that porous matrix between the ®bers disintegrated into smaller pieces, and that ®bers did not slide out of matrix holes as observed for composites produced with 'weak' interfaces. Some of the matrix debris, and matrix still bonded to the ®bers are seen in Fig. 9(c). Fig. 9(d) shows that some fracture regions in the 0 tows have a ¯atter, more coordinated fracture topography. Close examination of the ®bers in this region shows that most of them fracture on dierent planes, indicating that one crack front did not cause this fracture topography. One can see a few pairs of ®bers (arrows) which do exhibit planar fracture; examination of these ®ber pairs shows that they have a common fracture origin where they touch. As detailed elsewhere,21 this common fracture origin was produced during ®ber processing. This occurs when adjacent ®bers in the bundle (all ¯oors are spun from solution concurrently) stick to each other and sinter together along their cylindrical axis. It was shown that many ®ber tows contained pairs of ®bers that exhibited this frequent fracture origin. Fig. 9(a) and (b) shows the planar fracture of the 90 ®ber tows in each cloth layer. Here, the crack topography can be characterized as nearly planar, and the surface of this `planar' region contains relatively long lengths of the 0 ®bers extending from the fracture surface [see arrow in Fig. 9(b)]. Fig. 10 shows the fracture surface of the +/ÿ45 oriented specimens. Large areas are seen where the crack path propagated along the cylindrical ®ber surfaces (arrows); these cracks jog across ®ber tows. The surface produced when the crack propagates across the tow certainly does not form ®ber `brushes', but as shown in the enlarged view, Fig. 10(b), it can be seen that the same crack front did not cause all ®bers to fail. Generally, as shown in Figs. 9 and 10, the ZrO2 and mullite matrix ®lls all of the interstices between the ®bers. As detailed elsewhere,6 the HCl treatment at temperatures between 1200 and 1300C is eective in producing a strong matrix without shrinkage via an evaporation/ condensation sintering and coarsening phenomenon for the ZrO2 portion of the matrix. The lack of shrinkage of the powder matrix is evident in Figs. 9 and 10 by the lack of crack-like voids in the matrix. If the powder matrix were to densify its shrinkage, Fig. 6. Flexure stress versus Nominal Flexure Strain plot showing elastic modulus of composite with ®nal sintering at 1250C for 5 h in HCl. The composite was tested in the in-plane ®ber orientation. Tests for ®ber weave orientations of 0/90 and +/ÿ45 are shown. Table 2 Properties of in-plane notched composite 3-point bend testsa Fiber Sintering WOF J/mm2 Strength ratio Net-section stress MPa KI at peak load MPa m1/2 Kss MPa m1=2 0/90 1250C/5 h/HCl 1373 0.69 115 4.46 9.3 0/90 1250C/5 h/HCl 1287 0.66 100 3.89 8.0 +/ÿ45 1250C/5 h/HCl 438 0.72 62.4 2.32 4.38 +/ÿ45 1250C/5 h/HCl 338 0.62 53.9 2.06 3.85 a a/W0.49. Fig. 7. Load versus displacement plots of notched composite samples tested in the in-plane con®guration. Final sintering of composite was at 1250C for 5 h in HCl. Composite samples with 0/90 and +/ÿ45 ®ber orientations are shown. The normalized notch depth for these samples was a=W 0:48 and 0.49, respectively. J.J. Haslam et al. / Journal of the European Ceramic Society 20 (2000) 607±618 613
J.. Haslam et al. / Journal of the European Ceramic Society 20(2000)607-618 constrained by the fibers, would produce regions that are dense and others that are less dense. Eventually crack-like voids form if it is heated to a high enough 4. Discussion 10 mm 4. Interlaminar shear strength According to finite element analysis, the shear stresses are much higher than calculated by the classical beam theory equation 25 They are higher principally at the loading lines and along almost a diagonal line between that was tested with a notch in the upper load point and the lower load points. Close to the in-plane 3-point bending configuration. The sample still holds together due to interlocking fibers even after considerable permanent below the surface(approximately 0. 13-0. 2 of the normal ized sample height). 5 It is suggested that the localized 1m 100u 81 100 Fig. 9. Fracture surface near the tensile surface of a 0/90 fiber orientation composite containing 70v% cubic ZrO2(8m%Y2O3)/30v% sintered at 1200C for 5 h inHCL. The composite was tested in the in-plane 3-point bending configuration. Large amounts of fiber pullout are with little coordinated fracture. (a) Low magnification showing general extent of fiber pullout. (b) Higher magnification, arrows show fibers that failed independently and pulled free with disintegration of the surrounding matrix.(c)Micrograph showing fiber pullout and showing portions of the matrix bonded to the fibers indicating a strongly bonded matrix. Matrix between the fiber disintegrated during the fracture process. (d) Micr graph showing a region of fairly coordinated fiber fracture in the 0/90 composite sample although, except for fibers that were bonded together (arrows), the crack plane is unique implying that a continuous crack plane did not exist during fracture even in this region
constrained by the ®bers, would produce regions that are dense and others that are less dense. Eventually, crack-like voids form if it is heated to a high enough temperature.8,24 4. Discussion 4.1. Interlaminar shear strength According to ®nite element analysis, the shear stresses are much higher than calculated by the classical beam theory equation.25 They are higher principally at the loading lines and along almost a diagonal line between the upper load point and the lower load points. Close to the top anvil, the maximum shear stress occurs just below the surface (approximately 0.13±0.2 of the normalized sample height).25 It is suggested that the localized Fig. 8. Photograph of 0/90 composite that was tested with a notch in the in-plane 3-point bending con®guration. The sample still holds together due to interlocking ®bers even after considerable permanent strain, a/W0.5. Fig. 9. Fracture surface near the tensile surface of a 0/90 ®ber orientation composite containing 70v% cubic ZrO2 (8m% Y2O3)/30v% mullite sintered at 1200C for 5 h inHCl. The composite was tested in the in-plane 3-point bending con®guration. Large amounts of ®ber pullout are evident with little coordinated fracture. (a) Low magni®cation showing general extent of ®ber pullout, (b) Higher magni®cation, arrows show ®bers that failed independently and pulled free with disintegration of the surrounding matrix. (c) Micrograph showing ®ber pullout and showing portions of the matrix bonded to the ®bers indicating a strongly bonded matrix. Matrix between the ®ber disintegrated during the fracture process. (d) Micrograph showing a region of fairly coordinated ®ber fracture in the 0/90 composite sample although, except for ®bers that were bonded together (arrows), the crack plane is unique implying that a continuous crack plane did not exist during fracture even in this region. 614 J.J. Haslam et al. / Journal of the European Ceramic Society 20 (2000) 607±618
J. Haslam et al. Journal of the European Ceramic Society 20(2000)607-618 requires long beams and assumes pure bending. The general conclusion is that the short beam bend test may be used for comparison of similar materials for inter laminar shear strength The results for interlaminar shear strength showed apparent shear strength dependence of the span to thickness ratio, Work of others has also shown this and concluded that the short beam bend test does not char- acterize a quantitative interlaminar shear strength. 25.28 This is due to the localized loading conditions at the loading anvils and due to the short beam length. It could be pointed out that the short beam prevents 100pm application of St. Venant's principle, which states that stresses become uniformly distributed in a body at a (a sufficient distance from the loading points. A short beam does not provide sufficient distance to allow the stress distribution to be uniform in the body and so the measured shear strength changes with the span length for a given thickness. Also, as discussed above, the absolute magnitude of the shear strength is uncertain due to the nature of the localized loading The 'nominal interlaminar shear strength for the materials tested varied from 8 to 11 MPa. This is com- parable to the alumina and mullite matrix composites made by others. One observation was that lower strengths were measured in specimens that appeared to have more flaws on side edges of the beam. This sug- gests the possibility that flaws in the matrix micro- structure(voids or cracks) may reduce the interlaminar shear strength, but this trend is not clear nor readily quantifiable. In work by others, lower porosity (or Fig. 10. Fracture surface near the tensile surface of a +/-450 fiber higher densities in the matrix)caused increased inter orientation composite containing 70v% cubic ZrO2(8m%Y203)/ 30v% mullite sintered at 1250C for 5 h in HCL. The composite was laminar shear strength at the cost of notch insensitivity. 2 tested in the in-plane 3-point bending configuration. Very little fiber The effort to characterize and improve the inter- pullout is evident for in-plane testing in this fiber orientation. (a) laminar shear strength of all-oxide composites was due Arrows show where the crack propagated along fibers before cutting to initial measurements of the properties of the compo- across part of the tow of fibers. (b)Arrows show that fracture of the site in bending to measure a tensile strength. In early fibers was not on a common plane but still independently failed fibers with less disintegration of surrounding matrix tests, not reported here, the composite failed in an interlaminar shear manner. By reducing the thickness of the powder layer between the fiber cloth layers, which is loading can be reduced by placing compliant materials not characterized here, the interlaminar shear strength between the beam and the loading anvil, as attempted was improved to that reported in Fig 3. It appears that here, so that stresses will be distributed more uniformly with improved processing tensile failure may now limit over a larger area at the loading point 26 This can be the performance of the composite for potential applica tolerated in this test because it is the shear stress that is tions. Very short spans(or intense localized bending) of interest rather than the tensile stress on the lower are needed to produce delamination failure. Further work may determine that 3-D architectures and weaving The short beam bending test that has been used to of layers of the fibers could be effective ways of characterize the interlaminar shear strength of fiber improving this property of the composite if it is necessary composites is ASTM Standard D2344. However, based on finite element modeling experiments and analytical 4.2. Notch sensitivity and in-plane bend testing analysis it is suggested that this type of testing is not accurate for comparing substantially different materials The flexure test provides a measure of the tensile or for design purposes. 25.27.28 This is due to the localized strength of the material. The 0/90 orientation had and complicated loading at the loading anvils and due flexure strength of about 165 MPa which is comparable to to differences from Euler-Bernouilli beam theory, which the strengths obtained previously with a porous alumina
loading can be reduced by placing compliant materials between the beam and the loading anvil, as attempted here, so that stresses will be distributed more uniformly over a larger area at the loading point.26 This can be tolerated in this test because it is the shear stress that is of interest rather than the tensile stress on the lower beam. The short beam bending test that has been used to characterize the interlaminar shear strength of ®ber composites is ASTM Standard D2344. However, based on ®nite element modeling experiments and analytical analysis it is suggested that this type of testing is not accurate for comparing substantially dierent materials or for design purposes.25,27,28 This is due to the localized and complicated loading at the loading anvils and due to dierences from Euler±Bernouilli beam theory, which requires long beams and assumes pure bending. The general conclusion is that the short beam bend test may be used for comparison of similar materials for interlaminar shear strength. The results for interlaminar shear strength showed an apparent shear strength dependence of the span to thickness ratio. Work of others has also shown this and concluded that the short beam bend test does not characterize a quantitative interlaminar shear strength.25,28 This is due to the localized loading conditions at the loading anvils and due to the short beam length. It could be pointed out that the short beam prevents application of St. Venant's principle, which states that stresses become uniformly distributed in a body at a sucient distance from the loading points. A short beam does not provide sucient distance to allow the stress distribution to be uniform in the body and so the measured shear strength changes with the span length for a given thickness. Also, as discussed above, the absolute magnitude of the shear strength is uncertain due to the nature of the localized loading. The `nominal' interlaminar shear strength for the materials tested varied from 8 to 11 MPa. This is comparable to the alumina and mullite matrix composites made by others.26 One observation was that lower strengths were measured in specimens that appeared to have more ¯aws on side edges of the beam. This suggests the possibility that ¯aws in the matrix microstructure (voids or cracks) may reduce the interlaminar shear strength, but this trend is not clear nor readily quanti®able. In work by others, lower porosity (or higher densities in the matrix) caused increased interlaminar shear strength at the cost of notch insensitivity.26 The eort to characterize and improve the interlaminar shear strength of all-oxide composites was due to initial measurements of the properties of the composite in bending to measure a tensile strength. In early tests, not reported here, the composite failed in an interlaminar shear manner. By reducing the thickness of the powder layer between the ®ber cloth layers, which is not characterized here, the interlaminar shear strength was improved to that reported in Fig. 3. It appears that with improved processing tensile failure may now limit the performance of the composite for potential applications. Very short spans (or intense localized bending) are needed to produce delamination failure. Further work may determine that 3-D architectures and weaving of layers of the ®bers could be eective ways of improving this property of the composite if it is necessary. 4.2. Notch sensitivity and in-plane bend testing The ¯exure test provides a measure of the tensile strength of the material. The 0/90 orientation had a ¯exure strength of about 165 MPa which is comparable to the strengths obtained previously with a porous alumina/ Fig. 10. Fracture surface near the tensile surface of a +/ÿ45 ®ber orientation composite containing 70v% cubic ZrO2 (8m% Y2O3)/ 30v% mullite sintered at 1250C for 5 h in HCl. The composite was tested in the in-plane 3-point bending con®guration. Very little ®ber pullout is evident for in-plane testing in this ®ber orientation. (a) Arrows show where the crack propagated along ®bers before cutting across part of the tow of ®bers. (b) Arrows show that fracture of the ®bers was not on a common plane but still independently failed ®bers with less disintegration of surrounding matrix. J.J. Haslam et al. / Journal of the European Ceramic Society 20 (2000) 607±618 615
J. Haslam et al. Journal of the European Ceramic Society 20(2000)607-618 mullite matrix using the 720 fibers.' The fact that the length, b'is the beam width, and't'is the beam thick- composites with the mullite/zirconia matrix have ness. As shown in Table 2, the values of Kl, are about strengths similar to those with a mullite/alumina matrix half of KIss suggesting that the notch lengths in the bend is a reasonable result if the point of view is taken that tests are not long enough to reach the steady state the tensile properties in the 0/90 orientation are domi- toughness of the material. The notch sensitivity of the nated by the fiber properties. Relative to CMCs with composite therefore will have a size dependent effect weak'interfaces, matrix cracking is not apparent until which was not examined here. This crack length effect near the fracture stress, but the porous matrix compo- on the notch sensitivity is similar to that in brittle sites appear to have a somewhat smaller strain to fail materials with ductile reinforcements ure. It must be remembered that in most CMCs multiple matrix cracking occurs at a stress much below the frac- 43. Composite microstructure ture stress. most of these materials have either carbon or bn interphases between the fiber and matrix and In general, the microstructure is consistent with the contain non-oxide fibers. Multiple matrix cracking mechanical properties. In the case of the 0/90 compo would therefore dramatically limit their application in sites, improved toughness is obtained from a micro- oxidizing conditions. Therefore the allowable design structure where the fibers can absorb energy of stress strain could be similar to current cmcs oncentrations by bridging crack surfaces and dissip The net-section stress at failure in the presence of the ing energy as the matrix disintegrates during fiber pull- notch can be calculated and compared to the strength at out as shown in Fig 9. Fig 9(c) shows that the particles failure of an un-notched sample as a measure of notch are bonding to the fibers. This supports the conclusion ensitivity. The ction strength can be approxi- that the energy absorption during fracture comes from mated by assuming that all the material above the plane disintegration of the matrix surrounding the fibers of the notch tip acts to apply load to the remaining un- Where these features are absent in the fracture surface notch portion of the beam. This can be done by simply lower toughness is observed as in the case of the subtracting the length of the notch from the height of -45 composite samples shown in Fig. 10. However, the the beam and using this as the new height. The net-sec- 80 MPa strength in the +/-45 floor orientation is tion strength of the notched beams compared to an un igher than the 50 MPa obtained by Levi et al.5 even notched beam gives a strength ratio that is an indication for the higher strength 610 fiber. Clearly, the lower fiber of the degree of notch sensitivity of the material. This is strength in the 720 fibers is not reducing the strength fo shown in Table 2 for the current material. The strength this fiber orientation. The absence of extensive disin ratio for the notched beams was a 0.7 indicating mod- tegration of the matrix in the +/-45% samples may erate notch insensitivity. This level of notch sensitivity is provide for a higher strength at a cost of a lower strain greater than that observed with the notched bend test to failure than is obtained with the 610 fibers. Poten- with the mullite/alumina matrix where the ratio is tial matrix(zirconia/mullite) combined ~0.922 with a lower fiber strength(720 fiber) may be the cause In addition, the notched beam specimen can be used for this result. It should be pointed out, however, that to determine the work of fracture. Additionally, the no testing of the mullite/alumina matrix composites work of fracture(WOF) can be related to a steady state with 720 fibers in the +/-45 fiber orientation has been oughness(Kiss)of the composite through the simplified performed. So, a direct comparison is not available. An advantage of the microstructure containing coarsened zirconia and mullite is the coarsening and strengthening KIss =(WOF*E)/2 (5) without shrinkage. It is also resistant to densification during high temperature, long term use in air. These where E is the elastic modulus of the un-notched sample. added benefits make this processing method useful in Because there is a crack length over which this fracture addition to the benefit of a simplified processing toughness develops in composite materials (R-curve method behavior), it is important to compare the steady state fracture toughness to the magnitude of the stress intensity caused by the notch at the failure load. The applied stress 5. Summary intensity at the notch tip of a beam in bending can be determined using the following equation A simple single-step processing method was intro- duced and evaluated for an all-oxide layered woven K1=2.66Ma1//(b°t2) (6) fiber composite in this paper. This composite and of its mechanical properties were also described in this where KI is the applied stress intensity factor, M is the paper. The method uses an infiltration of the all-oxide applied bending moment=Load"Span/4, a'is the notch fibers with sub-micron ceramic particles that results in a
mullite matrix using the 720 ®bers.5 The fact that the composites with the mullite/zirconia matrix have strengths similar to those with a mullite/alumina matrix is a reasonable result if the point of view is taken that the tensile properties in the 0/90 orientation are dominated by the ®ber properties. Relative to CMCs with `weak' interfaces, matrix cracking is not apparent until near the fracture stress, but the porous matrix composites appear to have a somewhat smaller strain to failure. It must be remembered that in most CMCs multiple matrix cracking occurs at a stress much below the fracture stress. Most of these materials have either carbon or BN interphases between the ®ber and matrix and contain non-oxide ®bers. Multiple matrix cracking would therefore dramatically limit their application in oxidizing conditions. Therefore the allowable design strain could be similar to current CMCs. The net-section stress at failure in the presence of the notch can be calculated and compared to the strength at failure of an un-notched sample as a measure of notch sensitivity. The net-section strength can be approximated by assuming that all the material above the plane of the notch tip acts to apply load to the remaining unnotch portion of the beam. This can be done by simply subtracting the length of the notch from the height of the beam and using this as the new height. The net-section strength of the notched beams compared to an unnotched beam gives a strength ratio that is an indication of the degree of notch sensitivity of the material. This is shown in Table 2 for the current material. The strength ratio for the notched beams was 0.7 indicating moderate notch insensitivity. This level of notch sensitivity is greater than that observed with the notched bend test with the mullite/alumina matrix where the ratio is 0.9.22 In addition, the notched beam specimen can be used to determine the work of fracture. Additionally, the work of fracture (WOF) can be related to a steady state toughness (KIss) of the composite through the simpli®ed equation: KIss WOF E 1=2 5 where E is the elastic modulus of the un-notched sample. Because there is a crack length over which this fracture toughness develops in composite materials (R-curve behavior), it is important to compare the steady state fracture toughness to the magnitude of the stress intensity caused by the notch at the failure load. The applied stress intensity at the notch tip of a beam in bending can be determined using the following equation: KI 2:66Ma1=2 = bt 2 ; 6 where KI is the applied stress intensity factor, M is the applied bending moment=Load Span/4, `a' is the notch length, `b' is the beam width, and `t' is the beam thickness. As shown in Table 2, the values of KI, are about half of KIss suggesting that the notch lengths in the bend tests are not long enough to reach the steady state toughness of the material. The notch sensitivity of the composite therefore will have a size dependent eect which was not examined here. This crack length eect on the notch sensitivity is similar to that in brittle materials with ductile reinforcements.29 4.3. Composite microstructure In general, the microstructure is consistent with the mechanical properties. In the case of the 0/90 composites, improved toughness is obtained from a microstructure where the ®bers can absorb energy of stress concentrations by bridging crack surfaces and dissipating energy as the matrix disintegrates during ®ber pullout as shown in Fig. 9. Fig. 9(c) shows that the particles are bonding to the ®bers. This supports the conclusion that the energy absorption during fracture comes from disintegration of the matrix surrounding the ®bers. Where these features are absent in the fracture surface lower toughness is observed as in the case of the +/ ÿ45 composite samples shown in Fig. 10. However, the 80 MPa strength in the +/ÿ45 ¯oor orientation is higher than the 5O MPa obtained by Levi et al.5 even for the higher strength 610 ®ber. Clearly, the lower ®ber strength in the 720 ®bers is not reducing the strength for this ®ber orientation. The absence of extensive disintegration of the matrix in the +/ÿ45 samples may provide for a higher strength at a cost of a lower strain to failure than is obtained with the 610 ®bers. Potentially, a stronger matrix (zirconia/mullite) combined with a lower ®ber strength (720 ®ber) may be the cause for this result. It should be pointed out, however, that no testing of the mullite/alumina matrix composites with 720 ®bers in the +/ÿ45 ®ber orientation has been performed. So, a direct comparison is not available. An advantage of the microstructure containing coarsened zirconia and mullite is the coarsening and strengthening without shrinkage. It is also resistant to densi®cation during high temperature, long term use in air. These added bene®ts make this processing method useful in addition to the bene®t of a simpli®ed processing method. 5. Summary A simple single-step processing method was introduced and evaluated for an all-oxide layered woven ®ber composite in this paper. This composite and some of its mechanical properties were also described in this paper. The method uses an in®ltration of the all-oxide ®bers with sub-micron ceramic particles that results in a 616 J.J. Haslam et al. / Journal of the European Ceramic Society 20 (2000) 607±618