E驅≈3S ournal of the European Ceramic Society 20(2000)599-606 Alumina alumina composite with a porous zirconia interphase Processing, properties and component testing M. Holmquist * R. Lundberg, O. Sudreb, A.G. Razzell, L. Molliexd, J. Benoit J. Adlerborn° volvo Aero Corporation, 46181 Trollhattan, Sweden ONERA BP7292322 Chatillon Cedex france c Rolls-Royce plc, P.O. Box 31, Derby DE24 8B/, UK d Snecma, villaroche Centre, 77550 Moissy-cranmavel, france AC Cerana AB, Box 501, 915 23 Robertsfors, Sweden ccepted 18 August 1999 Abstract Novel oxide ceramic composites(NOCC) was a four year European programme aimed to develop an all-oxide ceramic matrix osite(CMC)and processing route, carry out a characterisation programme on the material and demonstrate it in a combustor rig at conditions representative of a gas turbine engine. The fibre used was a single crystal monofilament (Saphikon Inc ) which was chosen for its temperature and creep resistance. Alumina(aluminium oxide) was chosen for the fibre and matrix, and zirconia as a weak interphase coating on the fibre. Tape casting followed by hot pressing was chosen as the manufacturing route for the com- posite, with hot isostatic pressing(HIPping)as an alternative densification process. Cross-ply material with fibre volume fractions of around 30% was found to have moderate strength(100-130 MPa), but retained composite properties at elevated temperatures d after extended periods at elevated temperatures (1000 h at 1400C). In addition, the material was found to withstand thermal cycling(>1300 cycles to 1200C), retaining its as-fabricated properties. Computational fluid dynamics(CFD) calculations were carried out for a combustor rig, and a Cmc tile was designed The temperatures, stresses and strains in the tile were predicted using finite element(FE)analysis and combustor tiles were manufactured. a tile was successfully tested in a rig at temperatures > 1260C and up to 46 cycles. Some of the issues that remain to be addressed with the material and manufacturing method are cost, dela mination during manufacture, and consistency. It is likely that, due to the high cost of the fibre and relatively modest usable strength, the material will remain as a model material. The promising results on long term static and cyclic ageing proves that the concept of an all-oxide CMc is valid and points the way to future development of this class of material. c 2000 Elsevier Science Ltd. all rights reserved Keywords: Al2O2 fibre: Al2O3 matrix; Composites; Gas turbine; Interphase; Mechanical properties; Thermal shock resistance 1. Introduction (Rich burn-Quick quench-Lean burn) combustors which lower the emissions by controlling the combus- Great efforts are being made among gas turbine tion temperature within a narrow temperature range. manufacturers throughout the world to pursue tech- These methods require the usage of hot uncooled com nology for reducing pollutant emissions (nitrogen bustor liner walls to prevent (1)incoming cooling air oxides- NOx, carbon monoxide-CO and unburned from locally quenching the combustion (i.e. increase hydrocarbons- UHC)in the exhaust. Three promising UHC and Co) or (2)raising the temperature by methods for reducing emissions are LP(Lean Pre- decreasing fuel to air ratio (i. e. increase NOx). Candi mixed), LPP(Lean Pre-mixed Pre-vaporised) and rQl date materials for this application are ceramic matrix composites( CMCs)that can withstand the temperature s Corresponding author. of the hot gas in the reaction zone, without film air Now at: Rockwell International Corp, Science Centre, 1049 cooling used in conventional combustors made from Camino dos rios. Thousand Oaks. CA 93106 USA nickelbase superalloys today. With only back-cooling 0955-2219/00/S-see front matter C 2000 Elsevier Science Ltd. All rights reserved PII:S0955-2219(99)00258-7
Alumina/alumina composite with a porous zirconia interphase Ð Processing, properties and component testing M. Holmquista,*, R. Lundberga , O. Sudreb,1, A.G. Razzellc , L. Molliexd, J. Benoitd, J. Adlerborne a Volvo Aero Corporation, 461 81 TrollhaÈttan, Sweden bONERA, B.P. 72, 92322 ChaÃtillon Cedex, France c Rolls-Royce plc, P.O. Box 31, Derby DE24 8BJ, UK dSnecma, Villaroche Centre, 77550 Moissy-Cramayel, France e AC Cerama AB, Box 501, 915 23 Robertsfors, Sweden Accepted 18 August 1999 Abstract Novel oxide ceramic composites (NOCC) was a four year European programme aimed to develop an all-oxide ceramic matrix composite (CMC) and processing route, carry out a characterisation programme on the material and demonstrate it in a combustor rig at conditions representative of a gas turbine engine. The ®bre used was a single crystal mono®lament (Saphikon Inc.), which was chosen for its temperature and creep resistance. Alumina (aluminium oxide) was chosen for the ®bre and matrix, and zirconia as a weak interphase coating on the ®bre. Tape casting followed by hot pressing was chosen as the manufacturing route for the composite, with hot isostatic pressing (HIPping) as an alternative densi®cation process. Cross-ply material with ®bre volume fractions of around 30% was found to have moderate strength (100±130 MPa), but retained composite properties at elevated temperatures and after extended periods at elevated temperatures (1000 h at 1400C). In addition, the material was found to withstand thermal cycling (>1300 cycles to 1200C), retaining its as-fabricated properties. Computational ¯uid dynamics (CFD) calculations were carried out for a combustor rig, and a CMC tile was designed. The temperatures, stresses and strains in the tile were predicted using ®nite element (FE) analysis and combustor tiles were manufactured. A tile was successfully tested in a rig at temperatures >1260C and up to 46 cycles. Some of the issues that remain to be addressed with the material and manufacturing method are cost, delamination during manufacture, and consistency. It is likely that, due to the high cost of the ®bre and relatively modest usable strength, the material will remain as a model material. The promising results on long term static and cyclic ageing proves that the concept of an all-oxide CMC is valid and points the way to future development of this class of material. # 2000 Elsevier Science Ltd. All rights reserved. Keywords: Al2O2 ®bre; Al2O3 matrix; Composites; Gas turbine; Interphase; Mechanical properties; Thermal shock resistance 1. Introduction Great eorts are being made among gas turbine manufacturers throughout the world to pursue technology for reducing pollutant emissions (nitrogen oxides Ð NOx, carbon monoxide Ð CO and unburned hydrocarbons Ð UHC) in the exhaust. Three promising methods for reducing emissions are LP (Lean Premixed), LPP (Lean Pre-mixed Pre-vaporised) and RQL (Rich burn±Quick quench±Lean burn) combustors, which lower the emissions by controlling the combustion temperature within a narrow temperature range.1 These methods require the usage of hot uncooled combustor liner walls to prevent (1) incoming cooling air from locally quenching the combustion (i.e. increase UHC and CO) or (2) raising the temperature by decreasing fuel to air ratio (i.e. increase NOx). Candidate materials for this application are ceramic matrix composites (CMCs) that can withstand the temperature of the hot gas in the reaction zone, without ®lm air cooling used in conventional combustors made from nickelbase superalloys today. With only back-cooling 0955-2219/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved. PII: S0955-2219(99)00258-7 Journal of the European Ceramic Society 20 (2000) 599±606 * Corresponding author. 1 Now at: Rockwell International Corp., Science Centre, 1049 Camino Dos Rios, Thousand Oaks, CA 93106, USA
M. Holmquist et al. / Journal of the European Ceramic Society 20(2000)599-606 the wall temperature of a CMC liner may be kept in the associated with creep and sintering due to the high range of 1200-1400C while ensuring a suitable uniform fusivities of oxides compared to SiC. Single crystal oxi- combustion des, particularly complex oxides(e.g. YAG), are known The envisaged application of CMC in combustor to have significantly better creep behaviour than poly liners requires that the component resist thermal loads. crystals, although fabricating fibres remains a challenge. These applications are designed to have minimal However, a single crystal alumina fibre(SaphikonM)is requirements for components to withstand pressure or commercially available having a higher temperature other mechanical loads. In these cases, the failure strain capability and better creep resistance than polycrystal of the composite is the most important measure of its line alumina fibres. This fibre, which is produced by the damage tolerance. The causes of thermally induced edge film fed growth method from molten alumina, has strains include thermal mismatch with surrounding the disadvantage of being monofilament(diameter 125 components, temperature gradients within the compo- um)which prevents weaving and forming into complex nent and transient strains during temperature cycling.2 shapes, and a very high cost(due to the manufacturing he expected properties needed are temperature stabi- method). Several concepts for fibre coatings that would lity to 1400oC, oxidation resistance, mechanical stabi- improve the damage tolerance of oxide CMCs have been demonstrated in model composites. 1oe have only lity, chemical stability, damage tolerance and thermal been proposed but the performance of these have only shock resistance. These properties must be maintained for long times(>10, 000 h)and under cyclic conditions Two classes of CMCs are considered for this applica tion: oxide and non-oxide materials. Most development 2. Objective work has been done on non-oxide materials and com- mercial variants are usually based on silicon carbide The objective of the"Novel Oxide Ceramic Compo- (SiC)fibres (i.e. Nicalon"M or Tyranno TM)with a Sic sites "programme was to develop an all-oxide composite or an oxide (i.e. Al2O3)matrix. They have a fibre/matri for long life-time applications(>10,000 h)at tempera- interphase of carbon(C)or boron nitride(Bn) that tures above 1400 C in oxidising environments. Design contain weakly bonded planes of atoms, providing a and development of an oxide interphase has been weak debond layer which imparts a non-brittle fracture reported previously. 14-16 paper reports on the behaviour. Non-oxide CMCs have attractive high tem- development and scale-up of a composite fabrication perature properties, such as creep resistance and micro- process, results from mechanical testing as well as fab- structural stability. They also show high thermal rication and combustor rig tests of a model component conductivity and low thermal expansion, which reduces ermally induced strains. However, the oxidation sen- sitivity of the interphase will cause embrittlement of the 3. Experimental composite after service at high temperatures for long times. Embrittlement is most severe with cyclic mechan-.. Materials ical and thermal loading beyond the proportional limit, because oxygen that penetrates via the matrix crack Single crystal continuous a-Al2O3 sapphire fibres with created will react locally with the fibres and fibre coat- a nominal diameter of 125 um( Saphikon, USA)were ings to form oxide products. These reaction products chosen for this project since they have a thermodyn will create strong bonds between the fibre and matrix, stability compatible with the temperature goal and which prevent crack deflection and suppress internal forming limitations were not an issue for flat combustor frictional mechanisms that otherwise give toughness tiles. However, this single crystal fibre may not be the A way to avoid the oxidation problem is to use all-oxide ultimate candidate for this application due to its limited composites(i.e. the composites consist of oxide fibres, high-temperature temperature properties and sensitivity to slow oxide interfacial coatings and oxide matrices ). The com- crack growth, 3 but rather provide a model material on ponents are fully oxidised and further damage by oxida- to which to base an oxide/oxide CMC. Alumina was tion can be avoided, even at high temperatures and after chosen as a matrix to minimise thermally induced stres matrix cracking. Oxide composites also have the attrac- ses between the fibre and the matrix and limit undesired tive feature of potentially low cost. The primary difficul- chemical reactions. A small amount of zirconia was ties with this approach are the lack of suitable oxide fibre added to control matrix grain growth at elevated tem- reinforcement and the lack of oxide based "debond peratures. Zirconia has been identified as a suitable layer analogous to carbon or boron nitride. Most oxides interphase material in alumina based composites since also have low thermal conductivity and high thermal the system is thermochemically stable. 3 However, in expansion, leading to high thermally induced strains order for the interphase to behave as a debond layer, Polycrystalline oxide based fibres(e.g. Nextel 610 or is necessary to reduce the strength by introducing por- Nextel 720)have temperature limitations <1100 C osity (a porosity level of 30% has been suggested").An
the wall temperature of a CMC liner may be kept in the range of 1200±1400C while ensuring a suitable uniform combustion temperature. The envisaged application of CMC in combustor liners requires that the component resist thermal loads. These applications are designed to have minimal requirements for components to withstand pressure or other mechanical loads. In these cases, the failure strain of the composite is the most important measure of its damage tolerance. The causes of thermally induced strains include thermal mismatch with surrounding components, temperature gradients within the component and transient strains during temperature cycling.2 The expected properties needed are temperature stability to 1400C, oxidation resistance, mechanical stability, chemical stability, damage tolerance and thermal shock resistance. These properties must be maintained for long times (>10,000 h) and under cyclic conditions. Two classes of CMCs are considered for this application: oxide and non-oxide materials. Most development work has been done on non-oxide materials and commercial variants are usually based on silicon carbide (SiC) ®bres (i.e. NicalonTM or TyrannoTM) with a SiC or an oxide (i.e. Al2O3) matrix. They have a ®bre/matrix interphase of carbon (C) or boron nitride (BN) that contain weakly bonded planes of atoms, providing a weak debond layer which imparts a non-brittle fracture behaviour. Non-oxide CMCs have attractive high temperature properties, such as creep resistance and microstructural stability. They also show high thermal conductivity and low thermal expansion, which reduces thermally induced strains. However, the oxidation sensitivity of the interphase will cause embrittlement of the composite after service at high temperatures for long times. Embrittlement is most severe with cyclic mechanical and thermal loading beyond the proportional limit, because oxygen that penetrates via the matrix cracks created will react locally with the ®bres and ®bre coatings to form oxide products. These reaction products will create strong bonds between the ®bre and matrix, which prevent crack de¯ection and suppress internal frictional mechanisms that otherwise give toughness. A way to avoid the oxidation problem is to use all-oxide composites (i.e. the composites consist of oxide ®bres, oxide interfacial coatings and oxide matrices). The components are fully oxidised and further damage by oxidation can be avoided, even at high temperatures and after matrix cracking. Oxide composites also have the attractive feature of potentially low cost. The primary diculties with this approach are the lack of suitable oxide ®bre reinforcement and the lack of oxide based ``debond'' layer analogous to carbon or boron nitride. Most oxides also have low thermal conductivity and high thermal expansion, leading to high thermally induced strains. Polycrystalline oxide based ®bres (e.g. NextelTM 610 or Nextel 720) have temperature limitations <1100C associated with creep and sintering due to the high diffusivities of oxides compared to SiC. Single crystal oxides, particularly complex oxides (e.g. YAG), are known to have signi®cantly better creep behaviour than polycrystals, although fabricating ®bres remains a challenge. However, a single crystal alumina ®bre (SaphikonTM) is commercially available having a higher temperature capability and better creep resistance than polycrystalline alumina ®bres. This ®bre, which is produced by the edge ®lm fed growth method from molten alumina, has the disadvantage of being mono®lament (diameter 125 mm) which prevents weaving and forming into complex shapes, and a very high cost (due to the manufacturing method). Several concepts for ®bre coatings that would improve the damage tolerance of oxide CMCs have been proposed but the performance of these have only been demonstrated in model composites.3±10 2. Objective The objective of the ``Novel Oxide Ceramic Composites'' programme was to develop an all-oxide composite for long life-time applications (>10,000 h) at temperatures above 1400C in oxidising environments.11 Design and development of an oxide interphase has been reported previously.14±16 This paper reports on the development and scale-up of a composite fabrication process, results from mechanical testing as well as fabrication and combustor rig tests of a model component. 3. Experimental 3.1. Materials Single crystal continuous a-Al2O3 sapphire ®bres with a nominal diameter of 125 mm (Saphikon, USA) were chosen for this project since they have a thermodynamic stability compatible with the temperature goal and forming limitations were not an issue for ¯at combustor tiles. However, this single crystal ®bre may not be the ultimate candidate for this application due to its limited high-temperature properties12 and sensitivity to slow crack growth,13 but rather provide a model material on to which to base an oxide/oxide CMC. Alumina was chosen as a matrix to minimise thermally induced stresses between the ®bre and the matrix and limit undesired chemical reactions. A small amount of zirconia was added to control matrix grain growth at elevated temperatures. Zirconia has been identi®ed as a suitable interphase material in alumina based composites since the system is thermochemically stable.3 However, in order for the interphase to behave as a debond layer, it is necessary to reduce the strength by introducing porosity (a porosity level of 30% has been suggested4 ). An 600 M. Holmquist et al. / Journal of the European Ceramic Society 20 (2000) 599±606
M. Holmquist et al. / Journal of the European Ceramic Society 20(2000)599-606 optimised zirconia coating based on a powder slurry 3.3. Evaluation technique was developed using a proprietary pro- cess.5,6 The zirconia slurry contai Tensile testing was done using 200 mm long test bars, carbon black powder(ZrO2/C volume ratio was 1: 1). gauge length 40 mm, at room temperature and elevated The matrix slurry was prepared by dispersing alumina temperatures(800, 1200 and 1400 C). An induction powder(SM8, Baikowski, France)and 5 vol% unstabi- furnace with a susceptor was used lised zirconia powder (Degussa, Germany) in water ture testing. The test temperature was controlled by a using classical dispersion techniques thermocouple on the specimen in its centre. Tensile testing was also carried out on 100 mm long samples 3.2. Composite processing with 20 mm gauge length at room temperature in the as- rece ed condition, after static thermal ageing for 100 a process based on prepreg technique was developed 1000 h at 1400 C in air and after cyclic thermal Fig. 1. Single crystal alumina fibres were first passed ageing tests. The cyclic thermal ageing was carried out through the zirconia slurry. After drying, the coated using a 17 min cycle: 20%C-20oC. Forced fibre was wound around an alumina /zirconia powder air-cooling was used and the sample experienced tape placed on top of a large diameter spool. An alu severe cooling regime because it was only cooled from mina/zirconia layer was tape-cast directly onto the one side. A servo-hydraulic MTs testing system with spool and allowed to dry. Prepregs were then cut and hydraulic collet grips and a cross-head displacement stacked to form cross-ply green composite preforms. rate of 0.5 mm/min was used. Strain was monitored by Composites were hot pressed in a graphite die under uniaxial or biaxial extensometers. Microstructural fea- nitrogen atmosphere according to the temperature tures of composite cross-sections and fracture surfaces schedule established previously(1400C, 10 MPa, 70 were characterised using optical and scanning electron min)and finally heat-treated at 1250 C to remove the microscopy carbon in order to form the porous zirconia interphase. Composite plates ranging in size from 50x50 mm- to 3. 4. Component testing 180x 200 mm were made using this manufacturing method. Results from two full size plates(plates 4 and The applicability of the composite as a material for 5, 180x200 mm2)are presented and discussed in this uncooled combustor walls was assessed by evaluation in report. Interphase coating process conditions were a combustor test rig operating at conditions realistic of slightly changed from plate 4 to plate 5. The purpose a gas turbine combustor. The rig consists of a square was to make a thinner coating(but keeping the same frame with effusion cooled nickel alloy walls which are porosity level) in order to increase the fibre/matrix load attached with bolts. One side of the combustor has cut transfer. This was achieved by decreasing the slurry outs where flat tiles can be mounted. Ceramic tiles viscosity. A [0/90].s stacking sequence was used to (approximately 90x 50x3 mm)were fabricated using obtain a symmetrical and balanced composite plate. the process described above and tested in these cut-outs, Hot isostatic pressing was evaluated as an alternative to the front(higher temperature)tile having no holes and pressing process. Specialised tooling was he rear (lower temperature) having two air dilution designed to keep the plates fat during processing and holes(Fig 9). In order to predict the temperature dis- suitable encapsulation technique developed. Results ribution in the combustor walls, a computational fluid from HIPing are reported separately. dynamics (CFD)calculation was carried out using FLUent code. The thermal stresses in the tiles were then calculated by finite element(FE)methods(ANSYS code)using the temperature predictions AlO tape precasting Furnace 4. Results and discussion 4.1. Microstructure of composites wheel Fibre volume fractions around 30% were achieved and the plates were approximately 2.80 mm thick with densities around 87% of theoretical. Initially there were some delamination problems. A solution was found by Hot-pressing increasing the pressure during final sintering to 15 MPa Fig 1. Schematic illustration of fabrication process of ceramic matrix and performing a gentle heat-treatment to 1250"C after the plates had been cut to test bars. A cross-section is
optimised zirconia coating based on a powder slurry technique was developed using a proprietary process.15,16 The zirconia slurry contained a binder and carbon black powder (ZrO2/C volume ratio was 1:1). The matrix slurry was prepared by dispersing alumina powder (SM8, Baikowski, France) and 5 vol% unstabilised zirconia powder (Degussa, Germany) in water using classical dispersion techniques. 3.2. Composite processing A process based on prepreg technique was developed, Fig. 1. Single crystal alumina ®bres were ®rst passed through the zirconia slurry. After drying, the coated ®bre was wound around an alumina/zirconia powder tape placed on top of a large diameter spool. An alumina/zirconia layer was tape-cast directly onto the spool and allowed to dry. Prepregs were then cut and stacked to form cross-ply green composite preforms. Composites were hot pressed in a graphite die under nitrogen atmosphere according to the temperature schedule established previously (1400C, 10 MPa, 70 min15) and ®nally heat-treated at 1250C to remove the carbon in order to form the porous zirconia interphase. Composite plates ranging in size from 5050 mm2 to 180200 mm2 were made using this manufacturing method. Results from two full size plates (plates 4 and 5, 180200 mm2 ) are presented and discussed in this report. Interphase coating process conditions were slightly changed from plate 4 to plate 5. The purpose was to make a thinner coating (but keeping the same porosity level) in order to increase the ®bre/matrix load transfer. This was achieved by decreasing the slurry viscosity. A [0/90]8,s stacking sequence was used to obtain a symmetrical and balanced composite plate. Hot isostatic pressing was evaluated as an alternative to the hot pressing process. Specialised tooling was designed to keep the plates ¯at during processing and suitable encapsulation technique developed. Results from HIPing are reported separately.17 3.3. Evaluation Tensile testing was done using 200 mm long test bars, gauge length 40 mm, at room temperature and elevated temperatures (800, 1200 and 1400C). An induction furnace with a susceptor was used in the high temperature testing. The test temperature was controlled by a thermocouple on the specimen in its centre. Tensile testing was also carried out on 100 mm long samples with 20 mm gauge length at room temperature in the asreceived condition, after static thermal ageing for 100 and 1000 h at 1400C in air and after cyclic thermal ageing tests. The cyclic thermal ageing was carried out using a 17 min cycle; 20C!1200C!20C. Forced air-cooling was used and the sample experienced a severe cooling regime because it was only cooled from one side. A servo-hydraulic MTS testing system with hydraulic collet grips and a cross-head displacement rate of 0.5 mm/min was used. Strain was monitored by uniaxial or biaxial extensometers. Microstructural features of composite cross-sections and fracture surfaces were characterised using optical and scanning electron microscopy. 3.4. Component testing The applicability of the composite as a material for uncooled combustor walls was assessed by evaluation in a combustor test rig operating at conditions realistic of a gas turbine combustor.1,19 The rig consists of a square frame with eusion cooled nickel alloy walls which are attached with bolts. One side of the combustor has cutouts where ¯at tiles can be mounted. Ceramic tiles (approximately 90503 mm3 ) were fabricated using the process described above and tested in these cut-outs, the front (higher temperature) tile having no holes and the rear (lower temperature) having two air dilution holes (Fig. 9). In order to predict the temperature distribution in the combustor walls, a computational ¯uid dynamics (CFD) calculation was carried out using FLUENT code. The thermal stresses in the tiles were then calculated by ®nite element (FE) methods (ANSYS code) using the temperature predictions. 4. Results and discussion 4.1. Microstructure of composites Fibre volume fractions around 30% were achieved and the plates were approximately 2.80 mm thick with densities around 87% of theoretical. Initially there were some delamination problems. A solution was found by increasing the pressure during ®nal sintering to 15 MPa and performing a gentle heat-treatment to 1250C after the plates had been cut to test bars. A cross-section is Fig. 1. Schematic illustration of fabrication process of ceramic matrix composites. M. Holmquist et al. / Journal of the European Ceramic Society 20 (2000) 599±606 601
shown in Fig. 2. The fibre spacing is relatively well controlled. Zirconia inclusions in the matrix appear to have limited the alumina grain size as they were mostly Plate n5-grip failure located at grain boundaries and triple grain junctions Some porous regions with lower density could be observed between fibres in the plane of plies. In addi- Plate n4- gauge failure tion, the zirconia fibre coating has been deformed into ""Mickey Mouse ears"on either side of the fibres and in some cases even detached from the fibre. This is a result of the hot-pressing conditions that will enhance the axial compressive deformation (which is vertical in Fig. 2). The zirconia interphase had an average thick ness of 5-10 Hm. No microstructural differences were seen between plates 4 and 5. 4.2. Mechanical testing rature tensile stress/st curves of as-received [0/90]ss Al2O3/Al2O3 composite. Tensile testing at room temperature of as-received samples from plates 4 and 5 showed similar UTS (ulti- mate tensile strength) values, although the modulus for outside the gauge length. Smaller pull-out lengths( 5mm) he samples from plate 5 was higher(193 GPa instead of were noted. This indicates a stronger fibre /matrix 152 GPa)(Fig 3). Both sets of samples showed classical bonding increasing the interfacial shear stress between stress/strain curves with a knee at w100 MPa and a low fibre and matrix, which also was the purpose of the modulus section to the point of failure. For plate 4, very changes made to the interphase process long(up to 50 mm) pull-out lengths were observed and Remains of the porous zirconia interphase layer were only a few matrix cracks appeared after the matrix visible on both the fibre and the internal surfaces of the cracking stress had been reached. This behaviour was matrix hole, Fig 4. Round shape, loose individual zir hought to originate from the low load transfer to raise conia grains were also observed. It has been suggested the stress on the composite around a main crack and that a damage zone propagates in the porous zirconia eventually initiate more cracks within the matrix. The interphase. Relative sliding between fibre and matrix UTS, 110 MPa was approximately halved compared to will then further crush the porous structure by breaking similar material with unidirectional fibre lay-up. 15, 16 sintered necks between the grains. Eventually the por This was attributed partly to a reduction in fibre volume ous sintered structure is transformed into individual fraction in the tensile direction to M15%. The strain to round powder grains that are rolled between the fibre failure was 0.45% which is comparable with the cur- and matrix forming small" ball-bearings"that promote rently commercially available CMCs. For samples from sliding. 15 Close examination of the fibre surface revealed plate 5 slightly higher UTS were observed. The low indicated strain to failures were caused by rupture 999228KV 58818FmlD17 99928KU8198198琴mW Fig. 4. Fracture surface of as-received [0/90]8.s Al2O3/Al2O3 composite fibre with remains of porous zirconia interphase
shown in Fig. 2. The ®bre spacing is relatively well controlled. Zirconia inclusions in the matrix appear to have limited the alumina grain size as they were mostly located at grain boundaries and triple grain junctions. Some porous regions with lower density could be observed between ®bres in the plane of plies. In addition, the zirconia ®bre coating has been deformed into ``Mickey Mouse ears'' on either side of the ®bres and in some cases even detached from the ®bre. This is a result of the hot-pressing conditions that will enhance the axial compressive deformation (which is vertical in Fig. 2). The zirconia interphase had an average thickness of 5±10 mm. No microstructural dierences were seen between plates 4 and 5. 4.2. Mechanical testing Tensile testing at room temperature of as-received samples from plates 4 and 5 showed similar UTS (ultimate tensile strength) values, although the modulus for the samples from plate 5 was higher (193 GPa instead of 152 GPa) (Fig. 3). Both sets of samples showed classical stress/strain curves with a knee at 100 MPa and a low modulus section to the point of failure. For plate 4, very long (up to 50 mm) pull-out lengths were observed and only a few matrix cracks appeared after the matrix cracking stress had been reached. This behaviour was thought to originate from the low load transfer to raise the stress on the composite around a main crack and eventually initiate more cracks within the matrix. The UTS, 110 MPa was approximately halved compared to similar material with unidirectional ®bre lay-up.15,16 This was attributed partly to a reduction in ®bre volume fraction in the tensile direction to 15%. The strain to failure was 0.45% which is comparable with the currently commercially available CMCs. For samples from plate 5 slightly higher UTS were observed. The low indicated strain to failures were caused by rupture outside the gauge length. Smaller pull-out lengths (5 mm) were noted. This indicates a stronger ®bre/matrix bonding increasing the interfacial shear stress between ®bre and matrix, which also was the purpose of the changes made to the interphase process. Remains of the porous zirconia interphase layer were visible on both the ®bre and the internal surfaces of the matrix hole, Fig. 4. Round shape, loose individual zirconia grains were also observed. It has been suggested that a damage zone propagates in the porous zirconia interphase. Relative sliding between ®bre and matrix will then further crush the porous structure by breaking sintered necks between the grains. Eventually the porous sintered structure is transformed into individual round powder grains that are rolled between the ®bre and matrix forming small ``ball-bearings'' that promote sliding.15 Close examination of the ®bre surface revealed Fig. 2. Microstructure of hot pressed [0/90]8,s composite. Fig. 3. Room temperature tensile stress/strain curves of as-received [0/90]8,s Al2O3/Al2O3 composite. Fig. 4. Fracture surface of as-received [0/90]8,s Al2O3/Al2O3 composite showing pulled-out ®bre with remains of porous zirconia interphase layer. 602 M. Holmquist et al. / Journal of the European Ceramic Society 20 (2000) 599±606
strength being retained. The modulus also dropped off as a function of temperature(Fig. 6). Examination of Plate n°4 the fracture surfaces showed a significant reduction in pu ed to rt and little evidence of pull-out at 1400 C(Fig. 7), which is con- sistent with reduced fibre strength at high tempera tures.20,2 Scratching of the zirconia off the fibre surfaces was more evident at 1400%C. and an accumu lation of zirconia was also seen on some fibres. one explanation may be that zirconia is exhibiting some plasticity at the testing temperature, modifying the 50101 ball-bearing "mechanisI Samples from plate 4 were aged at 1400 C for 100 and Temperature (C) 1000 h in air and tested at rt to measure residual strength(Fig. 8). Samples aged at 1400C, 100 h, had Fig.5. UTS as a function of temperature for [0/90k, Al 03/Al203 UTS values above 100 MPa(134 and 107 MPa) similar to the as-received sample (110 MPa) and the Youngs modulus had increased slightly (to 202 and 187 GPa, respectively). No dramatic loss of properties was Plate n°4 observed after ageing for 1000 h, one specimen even retained the initial UTS. The pull-out lengths were smaller compared to the as-received samples but still about 10 mm in length. This confirms previous bend test results5.7 that composite behaviour is retained after ageing at 1400oC. However, coarsening of the zirconia interphase grains had clearly taken place during the 50 ageing. 15 Grain size grew from below a micron to a few microns during the 1000-h heat-treatment. This coarsening effect could be detrimental in the long run. It has been shown that a porous structure, which is pre- vented from densifying, will evolve via a de-sintering mechanism(i.e the breaking of sintered necks leading Youngs modulus as a function of temperature for [0/90)s s to pore coalescence). Eventually a gap may form between fibre and matrix. The coarsening effect would therefore. alter the load transfer mechanism but embrittlement should not occur roughness features like sapphire surface facetting and One sample was subjected to thermal cycling and tes cusps from initially sintered zirconia grains. However, ted after 1367 cycles(corresponding to an accumulated these features were not thought to be too detrimental to time of 230 h at 1200oC)were reached. An UtS of 148 the fibre strength. I5 MPa and an elastic modulus of 150 GPa was noted. The No significant differences between plates 4 and 5 were UTS is plotted in Fig. 8 for comparison with the other seen in the high temperature tensile tests(Fig. 5). At thermally aged samples. The results demonstrates very 800C, the rate of property drop-off was 40% compared good thermal shock resistance for the composite mate to the initial UTS. The Uts did not change between rial, a property that monolithic oxides otherwise do not 1200 and 1400C with 50% of room temperature(RT) exhibit 800°c 1200°C Fig. 7. Reduced fibre pull-out lengths at elevated temperatures (all samples are from plate 5)
roughness features like sapphire surface facetting and cusps from initially sintered zirconia grains. However, these features were not thought to be too detrimental to the ®bre strength.15 No signi®cant dierences between plates 4 and 5 were seen in the high temperature tensile tests (Fig. 5). At 800C, the rate of property drop-o was 40% compared to the initial UTS. The UTS did not change between 1200 and 1400C with 50% of room temperature (RT) strength being retained. The modulus also dropped o as a function of temperature (Fig. 6). Examination of the fracture surfaces showed a signi®cant reduction in ®bre pull-out at 1200C compared to RT and very little evidence of pull-out at 1400C (Fig. 7), which is consistent with reduced ®bre strength at high temperatures.20,21 Scratching of the zirconia o the ®bre surfaces was more evident at 1400C, and an accumulation of zirconia was also seen on some ®bres. One explanation may be that zirconia is exhibiting some plasticity at the testing temperature, modifying the ``ball-bearing'' mechanism. Samples from plate 4 were aged at 1400C for 100 and 1000 h in air and tested at RT to measure residual strength (Fig. 8). Samples aged at 1400C, 100 h, had UTS values above 100 MPa (134 and 107 MPa) similar to the as-received sample (110 MPa) and the Young's modulus had increased slightly (to 202 and 187 GPa, respectively). No dramatic loss of properties was observed after ageing for 1000 h, one specimen even retained the initial UTS. The pull-out lengths were smaller compared to the as-received samples but still about 10 mm in length. This con®rms previous bend test results15,17 that composite behaviour is retained after ageing at 1400C. However, coarsening of the zirconia interphase grains had clearly taken place during the ageing.15 Grain size grew from below a micron to a few microns during the 1000-h heat-treatment. This coarsening eect could be detrimental in the long run. It has been shown that a porous structure, which is prevented from densifying, will evolve via a de-sintering mechanism18 (i.e. the breaking of sintered necks leading to pore coalescence). Eventually a gap may form between ®bre and matrix. The coarsening eect would, therefore, alter the load transfer mechanism but embrittlement should not occur. One sample was subjected to thermal cycling and tested after 1367 cycles (corresponding to an accumulated time of 230 h at 1200C) were reached. An UTS of 148 MPa and an elastic modulus of 150 GPa was noted. The UTS is plotted in Fig. 8 for comparison with the other thermally aged samples. The results demonstrates very good thermal shock resistance for the composite material, a property that monolithic oxides otherwise do not exhibit. Fig. 5. UTS as a function of temperature for [0/90]8,s Al2O3/Al2O3 composites. Fig. 6. Young's modulus as a function of temperature for [0/90]8,s Al2O3/Al2O3 composites. Fig. 7. Reduced ®bre pull-out lengths at elevated temperatures (all samples are from plate 5). M. Holmquist et al. / Journal of the European Ceramic Society 20 (2000) 599±606 603