MIATERIALS ENE S ENGINEERING ELSEVIER Materials Science and Engineering A272(1999)38-44 www.elsevier.com/locate/msea Deformation and failure processes during tensile creep of fibre and whisker reinforced SiC/Al2O3 composites B. Wilshire f. Carreno I Department of Materials Engineering, University of Wales, Swansea SA2 8PP, UK abstract The tensile creep and creep fracture properties in air from 1473 to 1673 k are reported for an alumina composite, reinforced with interwoven bundles of silicon carbide(Nicalon) fibres aligned parallel and normal to the stres The creep trength of this 0/90 SiC/Al,O3 composite is determined by the longitudinal fibres, with creep of the fibres acco by cra development in the weak porous matrix. Oxygen penetration then promotes failure of the fibres bridging the ing cracks with creep fracture occurring when a crack becomes long enough to cause sudden failure by fibre pull-out. These behaviour nade between the effects of whisker and fibre reinforcement on the high-temperature creep and creep fracture characteristics of reserve Keywords:Continuous fibre-reinforced composites: SIC/AL,O, composites; Creep: Creep fracture 1. Introduction at the reinforcement-matrix interfaces Thus for exam- ple, compared with monolithic alumina, reinforcement or high-temperature aeroengine applications, major of alumina with 10-33 vol. of silicon carbide research efforts continue to be directed towards the whiskers results in substantial improvements in tough- development of ceramic-matrix composites( CMCs) ness and strength [1, 2 as well as significant increases in with improved toughness and damage tolerance, seek- high-temperature creep and creep fracture resistance g to overcome the engineering design constraints im- [3-12]. Yet, while numerous studies have targeted Sic posed by the inherently brittle nature of monolithic whisker reinforced alumina (SICw/Al2O3)products, re- ceramics. However, before these advanced CMCs can cent materials development programmes have focused be considered for applications requiring component particularly on continuous fibre-reinforced CMCs lives of many thousands of hours, their long-term be- whose strength characteristics depend primarily on load haviour under tensile loading at elevated temperatures transfer to large volume fractions of high-modulu must be assessed and understood fibres with weak fibre-matrix interfaces [13-15] CMCs produced with dispersions of ceramic whiskers A study has therefore been made of the deformation rely on toughening mechanisms which include whisk and failure processes controlling the tensile creep and bridging or pull out, crack deflection or microcracking creep fracture properties of a SiC fibre reinforced alu- mina-matrix(SiCAl,O3) composite. In addition to al This paper is dedicated to Professor Herbert Herman on the lowing comparisons to be made between the creep behaviour patterns displayed by Sic whisker and Sic 295244 onding author.Tel:+4-1792-295243;fax:+4-1792 fibre reinforced alumina composites, by completing long-term tests in air from 1473 to 1673 K. the I Present address: Centro Nacional de Investigaciones Metalurgi. gramme was also designed to provide information cas, CSIC, 28040 Madrid, Spain. vant to aeroengine component design 0921-5093/99/S- see front matter c 1999 Elsevier Science S.A. All rights reserved. PI:s09215093099)00451·7
Materials Science and Engineering A272 (1999) 38–44 Deformation and failure processes during tensile creep of fibre and whisker reinforced SiC/Al2O3 composites B. Wilshire *, F. Carren˜o 1 Department of Materials Engineering, Uni6ersity of Wales, Swansea SA2 8PP, UK Abstract The tensile creep and creep fracture properties in air from 1473 to 1673 K are reported for an alumina matrix composite, reinforced with interwoven bundles of silicon carbide (Nicalon™) fibres aligned parallel and normal to the stress axis. The creep strength of this 0/90° SiCf /Al2O3 composite is determined by the longitudinal fibres, with creep of the fibres accompanied by crack development in the weak porous matrix. Oxygen penetration then promotes failure of the fibres bridging the developing cracks, with creep fracture occurring when a crack becomes long enough to cause sudden failure by fibre pull-out. These behaviour patterns are discussed in relation to tensile creep data available for SiC whisker-reinforced alumina, allowing comparisons to be made between the effects of whisker and fibre reinforcement on the high-temperature creep and creep fracture characteristics of SiC/Al2O3 composites. © 1999 Elsevier Science S.A. All rights reserved. Keywords: Continuous fibre-reinforced composites; SiC/Al2O3 composites; Creep; Creep fracture www.elsevier.com/locate/msea 1. Introduction For high-temperature aeroengine applications, major research efforts continue to be directed towards the development of ceramic–matrix composites (CMCs) with improved toughness and damage tolerance, seeking to overcome the engineering design constraints imposed by the inherently brittle nature of monolithic ceramics. However, before these advanced CMCs can be considered for applications requiring component lives of many thousands of hours, their long-term behaviour under tensile loading at elevated temperatures must be assessed and understood. CMCs produced with dispersions of ceramic whiskers rely on toughening mechanisms which include whisker bridging or pull out, crack deflection or microcracking at the reinforcement–matrix interfaces. Thus, for example, compared with monolithic alumina, reinforcement of alumina with 10–33 vol.% of silicon carbide whiskers results in substantial improvements in toughness and strength [1,2], as well as significant increases in high-temperature creep and creep fracture resistance [3–12]. Yet, while numerous studies have targeted SiC whisker reinforced alumina (SiCw/Al2O3) products, recent materials development programmes have focused particularly on continuous fibre-reinforced CMCs, whose strength characteristics depend primarily on load transfer to large volume fractions of high-modulus fibres with weak fibre–matrix interfaces [13–15]. A study has therefore been made of the deformation and failure processes controlling the tensile creep and creep fracture properties of a SiC fibre reinforced alumina-matrix (SiCf /Al2O3) composite. In addition to allowing comparisons to be made between the creep behaviour patterns displayed by SiC whisker and SiC fibre reinforced alumina composites, by completing long-term tests in air from 1473 to 1673 K, the programme was also designed to provide information relevant to aeroengine component design. This paper is dedicated to Professor Herbert Herman on the occasion of his 65th birthday. * Corresponding author. Tel.:+44-1792-295243; fax: +44-1792- 295244. E-mail address: b.wilshire@swansea.ac.uk (B. Wilshire) 1 Present address: Centro Nacional de Investigaciones Metalu´rgicas, CSIC, 28040 Madrid, Spain. 0921-5093/99/$ - see front matter © 1999 Elsevier Science S.A. All rights reserved. PII: S0921-5093(99)00451-7
B. Wilshire, F. Carrerio/ Materials Science and Engineering 4272(1999)38-44 2. Experimental procedures The silicon carbide fibre-reinforced alumina(SiCd/ AL,O3)composite, having a fibre volume fraction of 0.38, was manufactured by Dupont Lanxide Com posites. Essentially, the material consisted of layers or plys of interwoven bundles of NicalonTM (SiC soO.15) fibres, with the 2-D layers of satin-woven fabric aligned and stacked to obtain a balanced 0/90 fibre archite ture. The Nicalon nl202 fibres were coated boron nitride, which formed an interface layer between the approximately 15 um diameter fibres and the ap. proximately 5 um thick SiC fibre coatings deposited by chemical vapour infiltration. The alumina matrix was hen introduced via the diMoxM proces b in-situ directional oxidation of liquid aluminium Fig. 2. Tensile creep curves for the 0/90% Sic. composite at [16, 17]. A metal removal operation was used to extract 1473, 1573 and 1673 K, plotted as normalized (c/e) against residual alloy after the DIMOXTM process, resulting in 180 ks realized time (t/te) for stresses resulting in less than 4%/ residual metal in the finished composite The final matrix porosity was around 15%, arising mm long specimen to the water-cooled pull rods, were mainly from micropores, but with larger pores within located just outside the hot zone of a furnace fitted with and between the fibre bundles, as illustrated in Fig. I. molybdenum disilicide heating elements. With this type This results in an effective matrix modulus substantiall of furnace construction, temperature uniformity along lower than that of the coated fibre tows the gauge length of the testpiece was within +2 K Testpieces were produced with ga from 1473 to 1673 K o one fibre direction. The flat tensile specimens had gauge lengths of 40 mm and widths of 8 mm, with means of an external cage extensometer [19], with the hicknesses of 2.8 mm. These specimens were surface cage fixed firmly to the upper and lower pull rods. The ground prior to creep exposure to give a maximum of cage design allowed changes in specimen length to be seven undamaged plys across the testpiece thickness. measured to better than I um by means of a pair The tensile creep properties of the SiCAlO, com- differential capacitance transducers located on opposite posite were determined in air using high-precision con- sides of the testpiece, with the transducers linked to a stant-stress machines described elsewhere [18] but with computerized data logging unit the gripping system modified to accept flat rather than Sets of full creep curves were obtained at tempera cylindrical samples. The grips, which attached the 200 tures from 1473 to 1673 K, selecting the stress ranges to give creep rupture lives of up to about 2500 h. Few problems were encountered during the test programme 10um although, occasionally, a specimen was found to fail outside the gauge length. Such tests were not included in the present analyses. However, to assess data repro- ducibility, three tests were carried out at 1573 K under a stress of 90 MPa. The individual measurements of the minimum creep rate, the time to fracture and the total strain to failure were within + 50% of the mean values for the three tests, a degree of scatter which appears reasonable in relation to the material complexity(Fig 3. Experimental results Fig. 1. Scanning electron micrograph showing the Nicalon M fibre The forms of the creep curves recorded for the 0/90o produced with boron nitride interface layers between the fibres and SiCdAL,O, composite are illustrated in Fig. 2. After the the silicon carbide fibre coatings, in the porous polycrystalline alu. initial strain obtained on loading at the creep tempera mina matrix of the 0/90 SicrAl2O3 composite ture, a normal primary curve was observed during
B. Wilshire, F. Carren˜o / Materials Science and Engineering A272 (1999) 38–44 39 2. Experimental procedures The silicon carbide fibre-reinforced alumina (SiCf / Al2O3) composite, having a fibre volume fraction of 0.38, was manufactured by Dupont Lanxide Composites. Essentially, the material consisted of layers or plys of interwoven bundles of Nicalon™ (SiC0.85O0.15) fibres, with the 2-D layers of satin-woven fabric aligned and stacked to obtain a balanced 0/90° fibre architecture. The Nicalon™ NL202 fibres were coated with boron nitride, which formed an interface layer between the approximately 15 mm diameter fibres and the approximately 5 mm thick SiC fibre coatings deposited by chemical vapour infiltration. The alumina matrix was then introduced via the DIMOX™ process, i.e. by in-situ directional oxidation of liquid aluminium [16,17]. A metal removal operation was used to extract residual alloy after the DIMOX™ process, resulting in less than 4% residual metal in the finished composite. The final matrix porosity was around 15%, arising mainly from micropores, but with larger pores within and between the fibre bundles, as illustrated in Fig. 1. This results in an effective matrix modulus substantially lower than that of the coated fibre tows. Testpieces were produced with gauge lengths parallel to one fibre direction. The flat tensile specimens had gauge lengths of 40 mm and widths of 8 mm, with thicknesses of 2.8 mm. These specimens were surface ground prior to creep exposure to give a maximum of seven undamaged plys across the testpiece thickness. The tensile creep properties of the SiCf /Al2O3 composite were determined in air using high-precision constant-stress machines described elsewhere [18], but with the gripping system modified to accept flat rather than cylindrical samples. The grips, which attached the 200 Fig. 2. Tensile creep curves for the 0/90° SiCf /Al2O3 composite at 1473, 1573 and 1673 K, plotted as normalized strain (o/of ) against normalized time (t/tf ) for stresses resulting in creep lives of around 180 ks. mm long specimen to the water-cooled pull rods, were located just outside the hot zone of a furnace fitted with molybdenum disilicide heating elements. With this type of furnace construction, temperature uniformity along the gauge length of the testpiece was within 92 K from 1473 to 1673 K. Changes in specimen length were monitored by means of an external cage extensometer [19], with the cage fixed firmly to the upper and lower pull rods. The cage design allowed changes in specimen length to be measured to better than 1 mm by means of a pair of differential capacitance transducers located on opposite sides of the testpiece, with the transducers linked to a computerized data logging unit. Sets of full creep curves were obtained at temperatures from 1473 to 1673 K, selecting the stress ranges to give creep rupture lives of up to about 2500 h. Few problems were encountered during the test programme although, occasionally, a specimen was found to fail outside the gauge length. Such tests were not included in the present analyses. However, to assess data reproducibility, three tests were carried out at 1573 K under a stress of 90 MPa. The individual measurements of the minimum creep rate, the time to fracture and the total strain to failure were within 950% of the mean values for the three tests, a degree of scatter which appears reasonable in relation to the material complexity (Fig. 1). 3. Experimental results The forms of the creep curves recorded for the 0/90° SiCf /Al2O3 composite are illustrated in Fig. 2. After the initial strain obtained on loading at the creep temperature, a normal primary curve was observed during Fig. 1. Scanning electron micrograph showing the Nicalon™ fibres, produced with boron nitride interface layers between the fibres and the silicon carbide fibre coatings, in the porous polycrystalline alumina matrix of the 0/90° SiCf /Al2O3 composite
B. Wilshire, F. Carrerio/ Materials Science and Engineering 4272(1999)38-44 which the creep rate decayed with time until a mini mum or secondary rate was attained. However, under all test conditions covered, little or no tertiary stage was apparent before fracture occurred r each creep curve, the values of the minimum creep rate(Em), the time to fracture (te) and the total creep strain to failure (e)were determined. In this way, he results now reported for the SiCr Al2O3 samples can be discussed in relation to equivalent tensile data sets obtained in air from 1473 and 1573k for a sic 巴=E whisker-reinforced alumina [12]. Although this SiCw/ 1,O, composite was produced with only 25 vol% whiskers[12], compared with the 38 vol. fibres in the 1016 present SiC AlO3 material, increasing the whisker con- 80100120 tent above approximately 25 vol. leads to little addi- stress(MPa) tional improvement in creep performance [7 The creep properties of whisker-reinforced alumina Fig 4. The dependence of the temperature-compensated creep rate on products have been determined in flexure [3-9] and stress for the from 1473 to 1673 K compression [10, 1l], as well as in tension [12]. In all cases, the results were analysed by describing the varia Q。≈600+50 tions of the minimum creep rate with stress(o) and in units of kJ mol-l when o is expressed in MPa. Thus, emperature(T) using power law relationships of the at the creep temperatures studied for the 0/90 SiC/ form Al2O3 composite, n and 2 decrease with decreasing Em= Ao"exp-Oc/RT a ay (d applied stress, in line with the general behaviour pat- terns reported for variouS SICwAl2O3 materials [3-121 where A is a constant, n is the stress exponent The results included in Fig. 5 allow the relative creep is the activation energy for creep. When the stress/creep strengths in tension to be compared at 1573 K for the rate data for the 0/90 SiCAl,O3 material are plotted present 0/90 SiC Al,O3 material, for the 25 vol. SiC according to Eq (1), particularly at the higher creep whisker-reinforced alumina (SiCw/Al,O3)composite mperatures,the log a/log 'm relationships in Fig. 3 [12] and for Nicalon TM NL202 fibres [20]. Clearly, the curve such that the large n values(approximately 50) creep resistance of the continuous fibre-reinforced alu- recorded at high stresses decrease with decreasing mina is considerably superior to that of the whisker-re- stress. Furthermore, as evident from Fig. 4, standard inforced product, but inferior to that of the NicalonTM curve fitting procedures indicate that the log o/log Em fibre. Since the creep strength of the porous alumina lots at the various creep temperatures can be superim- matrix of the present SiCdAlO3 composite must posed by temperature compensation of the creep rate through an arrhenius term with SiCf/AbO3 CwlAho3 a Nicalon argon 5060708090100120 stress(MPa) stress(MPa) Fig. 5. The variations of the minimum creep rate(Em) with stress ir at 1573 K for the 0/900 SiC-Al,O, composite, for a SiC whisker- Fig. 3. The stress dependence of the minimum creep rate(Em)for the reinforced alumina (SICwlAl2O3)composite [12] and for Nicalon M 0/90 SiCH/ Al2O3 composite from 1473 to 1673K NL202 fibres tested in argon [20]
40 B. Wilshire, F. Carren˜o / Materials Science and Engineering A272 (1999) 38–44 which the creep rate decayed with time until a minimum or secondary rate was attained. However, under all test conditions covered, little or no tertiary stage was apparent before fracture occurred. For each creep curve, the values of the minimum creep rate (o; m), the time to fracture (tf ) and the total creep strain to failure (of ) were determined. In this way, the results now reported for the SiCf /Al2O3 samples can be discussed in relation to equivalent tensile data sets obtained in air from 1473 and 1573 K for a SiC whisker-reinforced alumina [12]. Although this SiCw/ Al2O3 composite was produced with only 25 vol.% whiskers [12], compared with the 38 vol.% fibres in the present SiCf /Al2O3 material, increasing the whisker content above approximately 25 vol.% leads to little additional improvement in creep performance [7]. The creep properties of whisker-reinforced alumina products have been determined in flexure [3–9] and compression [10,11], as well as in tension [12]. In all cases, the results were analysed by describing the variations of the minimum creep rate with stress (s) and temperature (T) using power law relationships of the form: o; m=Asn exp−Qc/RT (1) where A is a constant, n is the stress exponent and Qc is the activation energy for creep. When the stress/creep rate data for the 0/90° SiCf /Al2O3 material are plotted according to Eq. (1), particularly at the higher creep temperatures, the log s/log o; m relationships in Fig. 3 curve such that the large n values (approximately 50) recorded at high stresses decrease with decreasing stress. Furthermore, as evident from Fig. 4, standard curve fitting procedures indicate that the log s/log o; m plots at the various creep temperatures can be superimposed by temperature compensation of the creep rate through an Arrhenius term with: Fig. 4. The dependence of the temperature-compensated creep rate on stress for the 0/90° SiCf /Al2O3 composite from 1473 to 1673 K. Qc$600+5s (2) in units of kJ mol−1 when s is expressed in MPa. Thus, at the creep temperatures studied for the 0/90° SiCf / Al2O3 composite, n and Qc decrease with decreasing applied stress, in line with the general behaviour patterns reported for various SiCw/Al2O3 materials [3–12]. The results included in Fig. 5 allow the relative creep strengths in tension to be compared at 1573 K for the present 0/90° SiCf /Al2O3 material, for the 25 vol.% SiC whisker-reinforced alumina (SiCw/Al2O3) composite [12] and for Nicalon™ NL202 fibres [20]. Clearly, the creep resistance of the continuous fibre-reinforced alumina is considerably superior to that of the whisker-reinforced product, but inferior to that of the Nicalon™ fibre. Since the creep strength of the porous alumina matrix of the present SiCf /Al2O3 composite must be Fig. 5. The variations of the minimum creep rate (o; m) with stress in air at 1573 K for the 0/90° SiCf /Al2O3 composite, for a SiC whiskerreinforced alumina (SiCw/Al2O3) composite [12] and for Nicalon™ NL202 fibres tested in argon [20]. Fig. 3. The stress dependence of the minimum creep rate (o; m) for the 0/90° SiCf /Al2O3 composite from 1473 to 1673 K
B. Wilshire, F. Carrerio/ Materials Science and Engineering 4272(1999)38-44 1000m Fig. 6. Scanning electron micrograph showing crack development in Fig. 7. Scanning electron micrograph showing the zone characterized the alumina matrix of the 0/90 SiC Al,O, composite by oxidized fibres(marked A) and the final fibre pull-out zone on the creep fracture surface of a 0/90% SiC Al,O, crept to failure in even lower than that of the dense whisker-reinforced under a tensile stress of 79 MPa at 1573 K alumina, it appears that the matrix makes little contri- indicating that the product emtf increases as the strain bution to the strength of the fibre-reinforced CMC. In to failure increases with decreasing stress at each tem- contrast, stresses about five times higher must be ar perature(Figs. 10 and l1). Interestingly, by incorporat plied to the Nicalon TM fibres in order to achieve creep ing the tensile data quoted for the SIcw/AlO3 ceramic rates comparable with those recorded for the Nicalon [12] in Figs. 9-ll inclusive, similar trends in the rela fibre-reinforced alumina. The creep performance of the tionships between Em'tr and Er are revealed for the present 0/900 SiC, composite is therefore gov- whisker and fibre reinforced SiC/Al2O, materials erned by the longitudinal fibres, which occupy approxi tely one fifth of the testpiece cross-section With 0/90 CMCs, creep of the longitudinal fibres 4. Discussion transfers stress to the matrix, causing intergranular crack development, as shown in Fig. 6. Cracking re When power law approaches are used to describ duces matrix stiffness, reloading the fibres and inducing reep behaviour (Eq. (I)), it is common practice to further creep. As crack growth occurs, the developing approximate the curving log a/log im plots by a series of tangents, assuming that changes in n and @e signify cracks become bridged by the longitudinal fibres [21, 22] that different creep mechanisms become dominant but, as oxygen penetrates during tests in air, oxidation within different stress-temperature regimes. Conversely, promotes failure of the crack-bridging fibres. Zones showing oxidized fibres therefore extend over substan- tial fractions of the final fracture surfaces before the cracks grow to the length required to cause sudden failure of the testpieces by fibre pull out(Fig. 7) The rate of crack development and the time to 60 eventual failure of the SiC/Al,O, material depend on he rate of creep strain accumulation. As a result, the shapes of the log o/log tr plots in Fig. 8 seem to mirror 1473K he forms of the log o/log Em relationships in Fig 3 but, unlike the behaviour observed for many crystalline 1673K materials [23], the product(im' tr) of the minimum creep 031473K rate and the creep rupture life is not a constant. In 9 SiCw/Al203 1573K stead, for stresses giving the same creep rate, Emtr increases with increasing test temperature(Fig 9). Fur 103104105105 thermore, over the range of test conditions studied for time to fracture(s) the SICrAl2O3 composite, the time to fracture varies as Fig 8. The stress dependence of the time to fracture(td) for the 0/90 tr∝叫ln SiCdAlO3 composite from 1473 to 1673 K, with data also included for a SiCw Al2O3 ceramic tested in tension at 1473 and 1573K[12]
B. Wilshire, F. Carren˜o / Materials Science and Engineering A272 (1999) 38–44 41 Fig. 6. Scanning electron micrograph showing crack development in the alumina matrix of the 0/90° SiCf /Al2O3 composite. Fig. 7. Scanning electron micrograph showing the zone characterized by oxidized fibres (marked A) and the final fibre pull-out zone on the creep fracture surface of a 0/90° SiCf /Al2O3 crept to failure in air under a tensile stress of 79 MPa at 1573 K. even lower than that of the dense whisker-reinforced alumina, it appears that the matrix makes little contribution to the strength of the fibre-reinforced CMC. In contrast, stresses about five times higher must be applied to the Nicalon™ fibres in order to achieve creep rates comparable with those recorded for the Nicalon™ fibre-reinforced alumina. The creep performance of the present 0/90° SiCf /Al2O3 composite is therefore governed by the longitudinal fibres, which occupy approximately one fifth of the testpiece cross-section. With 0/90° CMCs, creep of the longitudinal fibres transfers stress to the matrix, causing intergranular crack development, as shown in Fig. 6. Cracking reduces matrix stiffness, reloading the fibres and inducing further creep. As crack growth occurs, the developing cracks become bridged by the longitudinal fibres [21,22] but, as oxygen penetrates during tests in air, oxidation promotes failure of the crack-bridging fibres. Zones showing oxidized fibres therefore extend over substantial fractions of the final fracture surfaces before the cracks grow to the length required to cause sudden failure of the testpieces by fibre pull out (Fig. 7). The rate of crack development and the time to eventual failure of the SiCf /Al2O3 material depend on the rate of creep strain accumulation. As a result, the shapes of the log s/log tf plots in Fig. 8 seem to mirror the forms of the log s/log o; m relationships in Fig. 3 but, unlike the behaviour observed for many crystalline materials [23], the product (o; m·tf ) of the minimum creep rate and the creep rupture life is not a constant. Instead, for stresses giving the same creep rate, o; m·tf increases with increasing test temperature (Fig. 9). Furthermore, over the range of test conditions studied for the SiCf /Al2O3 composite, the time to fracture varies as: tf8of /o; m (3) indicating that the product o; m·tf increases as the strain to failure increases with decreasing stress at each temperature (Figs. 10 and 11). Interestingly, by incorporating the tensile data quoted for the SiCw/Al2O3 ceramic [12] in Figs. 9–11 inclusive, similar trends in the relationships between o; m·tf and of are revealed for the whisker and fibre reinforced SiC/Al2O3 materials. 4. Discussion When power law approaches are used to describe creep behaviour (Eq. (1)), it is common practice to approximate the curving log s/log o; m plots by a series of tangents, assuming that changes in n and Qc signify that different creep mechanisms become dominant within different stress-temperature regimes. Conversely, Fig. 8. The stress dependence of the time to fracture (tf ) for the 0/90° SiCf /Al2O3 composite from 1473 to 1673 K, with data also included for a SiCw/Al2O3 ceramic tested in tension at 1473 and 1573 K [12]
B. Wilshire, F. Carrerio/ Materials Science and Engineering 4272(1999)38-44 0.05 9 SiCw Al203 1573K v SiCw/A203 1573K 003 107 s0.02 0.01 time to fracture(s stress(MPa) Fig. 9. The dependence of the time to fracture (I on the minim Fig. Il. The stress dependence of the total strain to failure(=) for the creep rate(im) for the o/90 SiCdAl2O3 composite from 1473 to 167 0/90 SiC/AlO, composite from 1473 to 1673 K, with data also a also included for a Sicw/Al O, composite tested in included for a SiCwALO, ceramic tested in tension at 1473 and 1573 tension at 1473 and 1573 K[12 K[2] it is then assumed that the dominant creep mechanism creep temperatures(Fig. 3), but also for the high can be identified by comparing the n and @c values stress-dependent activation energies recorded(Eq.(2)) observed experimentally with the values predicted the- The stress and temperature dependences of the mea- oretically for various creep processes. Yet, with the sured n and Q values for the 0/900 SiCd/ O,com- oresent 0/90 SiC/AlO3 composite, no change in posite are therefore a consequence of microstructural creep mechanism need be invoked to explain the de- instability of the Nicalon TM fibres at high temperatures creases in n and 2 which occur with decreasing stress rather than a result of any change in basic creep over the temperature range studied(Figs. 3 and 4) nechanism The results presented in Fig. 5 demonstrate that the As with the present 0/90 SiCr/Al,O3 composite, the creep resistance of the fibre-reinforced alumina is de- values of n and Q have also been found to vary with ermined by the fibres, but the creep strength of the stress and temperature for a variety of SICw/Al,o NicalonTM fibres deteriorates with time and tempera materials. Specifically, the stress exponent has been ure [24-26]. This weakening of the fibres then ac- observed to increase from n e l or 2 towards n=5 or counts not only for the gradual decrease in n with more with increasing stress [3-12], although n values ncreasing test duration, particularly at the higher up to 17 have been reported [8]. Stress exponents of I or 2 have been interpreted in terms of diffusional 0.035 creep and/or grain boundary sliding, possibly facili g0030 tated by intergranular viscous glass flow [5, 7,101 whereas na 5 or more has been taken as evidence for 0.025 31573K diffusion-controlled dislocation creep processes [3, 4, 6] However, several investigations have shown that creep of SiCw/,O3 materials does not involve dislocation 0.015 generation and movement [8, 10]. Consequently, n val ues of 5 or more have often been attributed to pro- 0010 cesses such as sliding and cavitation [5, 8, 9]. Indeed almost all studies completed for SICwAlO, com 0.005 posites provide either direct or indirect evidence for 0000·x grain boundary sliding, with the incorporation of Sic 0. 03 0.04 whiskers improving creep resistance by impeding the sliding process. In turn, sliding results in the formation of intergranular cavities and cracks, leading to ever Fig. 10. The variation of the product of the minimum creep rate and tual failure. It, therefore, appears that sliding and as the time to fracture('m t with the total creep strain to failure(e)for sociated cavitation represent the dominant the 0/90 SiCAl,O3 composite from 1473 to 1673 K, with data also included for a SiCwAlO3 ceramic tested in tension at 1473 and 1573 deformation and damage modes at all stress levels, irrespective of the n and @e values recorded
42 B. Wilshire, F. Carren˜o / Materials Science and Engineering A272 (1999) 38–44 Fig. 9. The dependence of the time to fracture (tf ) on the minimum creep rate (o; m) for the 0/90° SiCf /Al2O3 composite from 1473 to 1673 K, with data also included for a SiCw/Al2O3 composite tested in tension at 1473 and 1573 K [12]. Fig. 11. The stress dependence of the total strain to failure (of ) for the 0/90° SiCf /Al2O3 composite from 1473 to 1673 K, with data also included for a SiCw/Al2O3 ceramic tested in tension at 1473 and 1573 K [12]. it is then assumed that the dominant creep mechanism can be identified by comparing the n and Qc values observed experimentally with the values predicted theoretically for various creep processes. Yet, with the present 0/90° SiCf /Al2O3 composite, no change in creep mechanism need be invoked to explain the decreases in n and Qc which occur with decreasing stress over the temperature range studied (Figs. 3 and 4). The results presented in Fig. 5 demonstrate that the creep resistance of the fibre-reinforced alumina is determined by the fibres, but the creep strength of the Nicalon™ fibres deteriorates with time and temperature [24–26]. This weakening of the fibres then accounts not only for the gradual decrease in n with increasing test duration, particularly at the higher creep temperatures (Fig. 3), but also for the high stress-dependent activation energies recorded (Eq. (2)). The stress and temperature dependences of the measured n and Qc values for the 0/90° SiCf /Al2O3 composite are therefore a consequence of microstructural instability of the Nicalon™ fibres at high temperatures rather than a result of any change in basic creep mechanism. As with the present 0/90° SiCf /Al2O3 composite, the values of n and Qc have also been found to vary with stress and temperature for a variety of SiCw/Al2O3 materials. Specifically, the stress exponent has been observed to increase from n$1 or 2 towards n$5 or more with increasing stress [3–12], although n values up to 17 have been reported [8]. Stress exponents of 1 or 2 have been interpreted in terms of diffusional creep and/or grain boundary sliding, possibly facilitated by intergranular viscous glass flow [5,7,10], whereas n$5 or more has been taken as evidence for diffusion-controlled dislocation creep processes [3,4,6]. However, several investigations have shown that creep of SiCw/Al2O3 materials does not involve dislocation generation and movement [8,10]. Consequently, n values of 5 or more have often been attributed to processes such as sliding and cavitation [5,8,9]. Indeed, almost all studies completed for SiCw/Al2O3 composites provide either direct or indirect evidence for grain boundary sliding, with the incorporation of SiC whiskers improving creep resistance by impeding the sliding process. In turn, sliding results in the formation of intergranular cavities and cracks, leading to eventual failure. It, therefore, appears that sliding and associated cavitation represent the dominant deformation and damage modes at all stress levels, irrespective of the n and Qc values recorded. Fig. 10. The variation of the product of the minimum creep rate and the time to fracture (o; m·tf ) with the total creep strain to failure (of ) for the 0/90° SiCf /Al2O3 composite from 1473 to 1673 K, with data also included for a SiCw/Al2O3 ceramic tested in tension at 1473 and 1573 K [12]
B. Wilshire, F. Carrero/ Materials Science and Engineering 4272 (1999)38-44 With the SiCw/Al,O3 ceramics, as with the present the creep strain/ time plots become less pronounced as SiCr AlO, composite, the decrease in n value with the creep temperature is raised(Fig. 2), the im' t e ratio decreasing applied stress can be explained in terms of increases as the primary strain (Ep) decreases with in- microstructural instability rather than mechanism creasing test temperature(Fig. 10) changes. Oxidation of the sic whiskers can increase the The fact that a increases with decreasing stress(Fig quantities of intergranular viscous phases present, 11)can then be explained on the basis that, for both the which would reduce the creep strength with increasing whisker and fibre reinforced CMCs, fracture occurs by est duration, especially at high creep temperatures. the development of a crack long enough to propagate Grain boundary and interfacial amorphous phases are rapidly under the prevailing test conditions, with the detrimental to creep resistance [5, 7], accounting for the critical crack length increasing with decreasing applied creep strength of SICw/Al2O3 materials being lower in stress. At high stresses, when the critical crack lengths air than in inert atmospheres [10]. This sensitivity to are relatively short, low ductility failures are recorded oxidation is then consistent with the observation that. In contrast formation of the critical crack involves for the 25 vol. whisker reinforced alumina tested in progressive growth and link up of shorter cracks as the tension, the alumina-rich oxidized regions at the speci test duration increases. Since cracks develop more uni- men surfaces increased in thickness with increasing test formly throughout the testpieces at lower stresses and duration and increasing test temperature [12]. On this temperatures, creep must continue for longer times to basis, in line with the results obtained for the present reach higher strains in order to create a crack long SiCHAl2O3 material, the information reported for SiCwl enough to cause failure by fibre pull-out, so that er and Al2O3 composites supports the view [27] that compari- Em'tr increase with decreasing stress at each creep tem- sons of measured and theoretical n and @c values perature(Figs. 10 and ll) cannot provide an unambiguous indication of the dom For the SiCr/Al2O3 composite, while tr is a function inant creep mechanism. of im(Eq (3)), the gradients of the log im/log tr relation Although traditional methods for identifying creep ships are approximately 0.75(Fig 9). Consequently, the mechanisms are unsatisfactory, the data comparisons in n values derived from the stress/creep rate plots(Fig 3) Fig. 5 show that creep of the 0/90 SiC Al2O3 com- are lower than the equivalent stress dependences calcu posite is controlled by the fibres. In turn, creep of the lated from the stress rupture data(Fig 8). For example fibres results in matrix cracking, with the longer cracks at the highest stress levels imposed at 1473, 1573 and developing normal to the tensile axis being bridged by 1673 K, the maximum n values are around 50 for the longitudinal fibres. While oxygen penetration promotes log a/log Em plots in Fig. 3, compared with the equiva failure of the crack-bridging fibres, the dependence of lent maximum values of about 65 for the log a/log tr the rupture life on the minimum creep rate(Eq.(3)) data in Fig 8. Even so, especially at the higher creep confirms that the rate of crack growth and the time to temperatures, both the log a/log Em and log a/log tr fracture are determined by the rate of creep strain plots curve such that the large stress exponents and accumulation. Clearly, as the crack-bridging fibres fail, activation energies found in short-duration tests de- he numbers of fibres supporting the creep loads de- crease with decreasing stress. For this reason, linea crease with increasing creep strain. However, the local- extrapolation of results obtained at high stresses overes- ized increases in stress due to crack development are timates low-stress performance, emphasizing the engi not sufficient to give a discernible acceleration in creep neering design requirement for reliable determination of rate before fracture occurs. Consequently, failure takes long-term property values place by the development of a crack long enough to Other information may be derived from the curving cause fracture by fibre pull-out(Fig. 7), terminating the log o/log Em and log a/log tr plots. For example, since decaying primary curve with no clearly-defined tertiary the n value usually increases with increasing stress, the age(Fig. 2) maximum recorded n value must depend on the stress- While tertiary stage tually absent with both bearing capabilities of the material. Thus, for polycrys- the fibre and whisker orced alumina-matrix ceram- talline magnesia, the maximum measured n values cs, distinct primary stages are found with the Sic increased from approximately 2 to approximately 7 as AlO3 composite(Fig. 2) but not with the SiCwAlO3 the strength and ductility were improved by modifying material [12]. Even so, for both types of creep curve, the fabrication routes[28]. On this basis, the attainment the total creep strain to failure can be expressed as of n values of 50 or more at high stresses(Figs. 3 and f=印+-mr 8)emphasizes the impressive creep and creep fracture resistance of the present 0/90% SiCH/Al2O, composite where ap is the primary creep strain. Thus, with the high temperatures. whisker-reinforced alumina, sp=0 so that = Em'tr ractical viewpoint, the creep and creep (Fig. 10). With the fibre-reinforced alumina, Em'tr also fracture strengths of the fibre-reinforced alumina are increases with increasing ar but, since the curvatures of markedly superior to those of whisker-reinforced prod
B. Wilshire, F. Carren˜o / Materials Science and Engineering A272 (1999) 38–44 43 With the SiCw/Al2O3 ceramics, as with the present SiCf /Al2O3 composite, the decrease in n value with decreasing applied stress can be explained in terms of microstructural instability rather than mechanism changes. Oxidation of the SiC whiskers can increase the quantities of intergranular viscous phases present, which would reduce the creep strength with increasing test duration, especially at high creep temperatures. Grain boundary and interfacial amorphous phases are detrimental to creep resistance [5,7], accounting for the creep strength of SiCw/Al2O3 materials being lower in air than in inert atmospheres [10]. This sensitivity to oxidation is then consistent with the observation that, for the 25 vol.% whisker reinforced alumina tested in tension, the alumina-rich oxidized regions at the specimen surfaces increased in thickness with increasing test duration and increasing test temperature [12]. On this basis, in line with the results obtained for the present SiCf /Al2O3 material, the information reported for SiCw/ Al2O3 composites supports the view [27] that comparisons of measured and theoretical n and Qc values cannot provide an unambiguous indication of the dominant creep mechanism. Although traditional methods for identifying creep mechanisms are unsatisfactory, the data comparisons in Fig. 5 show that creep of the 0/90° SiCf /Al2O3 composite is controlled by the fibres. In turn, creep of the fibres results in matrix cracking, with the longer cracks developing normal to the tensile axis being bridged by longitudinal fibres. While oxygen penetration promotes failure of the crack-bridging fibres, the dependence of the rupture life on the minimum creep rate (Eq. (3)) confirms that the rate of crack growth and the time to fracture are determined by the rate of creep strain accumulation. Clearly, as the crack-bridging fibres fail, the numbers of fibres supporting the creep loads decrease with increasing creep strain. However, the localized increases in stress due to crack development are not sufficient to give a discernible acceleration in creep rate before fracture occurs. Consequently, failure takes place by the development of a crack long enough to cause fracture by fibre pull-out (Fig. 7), terminating the decaying primary curve with no clearly-defined tertiary stage (Fig. 2). While tertiary stages are virtually absent with both the fibre and whisker reinforced alumina-matrix ceramics, distinct primary stages are found with the SiCf / Al2O3 composite (Fig. 2) but not with the SiCw/Al2O3 material [12]. Even so, for both types of creep curve, the total creep strain to failure can be expressed as: of=op+o; m·tf (4) where op is the primary creep strain. Thus, with the whisker-reinforced alumina, op$0 so that of$o; m·tf (Fig. 10). With the fibre-reinforced alumina, o; m·tf also increases with increasing of but, since the curvatures of the creep strain/time plots become less pronounced as the creep temperature is raised (Fig. 2), the o; m·tf /of ratio increases as the primary strain (op) decreases with increasing test temperature (Fig. 10). The fact that of increases with decreasing stress (Fig. 11) can then be explained on the basis that, for both the whisker and fibre reinforced CMCs, fracture occurs by the development of a crack long enough to propagate rapidly under the prevailing test conditions, with the critical crack length increasing with decreasing applied stress. At high stresses, when the critical crack lengths are relatively short, low ductility failures are recorded. In contrast, formation of the critical crack involves progressive growth and link up of shorter cracks as the test duration increases. Since cracks develop more uniformly throughout the testpieces at lower stresses and temperatures, creep must continue for longer times to reach higher strains in order to create a crack long enough to cause failure by fibre pull-out, so that of and o; m·tf increase with decreasing stress at each creep temperature (Figs. 10 and 11). For the SiCf /Al2O3 composite, while tf is a function of o; m (Eq. (3)), the gradients of the log o; m/log tf relationships are approximately 0.75 (Fig. 9). Consequently, the n values derived from the stress/creep rate plots (Fig. 3) are lower than the equivalent stress dependences calculated from the stress rupture data (Fig. 8). For example, at the highest stress levels imposed at 1473, 1573 and 1673 K, the maximum n values are around 50 for the log s/log o; m plots in Fig. 3, compared with the equivalent maximum values of about 65 for the log s/log tf data in Fig. 8. Even so, especially at the higher creep temperatures, both the log s/log o; m and log s/log tf plots curve such that the large stress exponents and activation energies found in short-duration tests decrease with decreasing stress. For this reason, linear extrapolation of results obtained at high stresses overestimates low-stress performance, emphasizing the engineering design requirement for reliable determination of long-term property values. Other information may be derived from the curving log s/log o; m and log s/log tf plots. For example, since the n value usually increases with increasing stress, the maximum recorded n value must depend on the stressbearing capabilities of the material. Thus, for polycrystalline magnesia, the maximum measured n values increased from approximately 2 to approximately 7 as the strength and ductility were improved by modifying the fabrication routes [28]. On this basis, the attainment of n values of 50 or more at high stresses (Figs. 3 and 8) emphasizes the impressive creep and creep fracture resistance of the present 0/90° SiCf /Al2O3 composite at high temperatures. From a practical viewpoint, the creep and creep fracture strengths of the fibre-reinforced alumina are markedly superior to those of whisker-reinforced prod-
B. Wilshire, F. Carrero/ Materials Science and Engineering 4272 (1999)38-44 cts,( Figs. 5 and 8). Moreover, with the 0/90% fibre Acknowledgements architecture, large cracks can develop without causing immediate failure(Fig. 7), especially when high creep The provision of a post-doctoral research fellow he test durations increase ship for Dr Carreno under the EC Human Capital towards the service lives expected for aeroengine com- and Mobility Programme( Contract No. ERBCH ponents(Fig. 11). Hence, crack tolerance is combined BGCT9330303)and the support received through the with impressive stress-bearing capabilities at high tem- British Council UK/Spain research collaboration ini- peratures, illustrating the potential of continuous tiative are gratefully acknowledged. The authors also fibre-reinforced CMCs for high-performance engineer- wish to thank Dr M.J. L. Percival for helpful discus- ng applications. sions and Rolls-Royce plc for financial support and materials supplies. 5. Conclusions References Under high-temperature tensile creep conditions, [PFBecher, G.C.Wei,JAm. c.67C(1984)267 similar behaviour patterns are displayed by SiCt T.N. Tiegs. J. Am. AlO, and SICw/Al,O, composites, although the Ceram.Soc.71(1988)1050. mechanisms controlling deformation and damage ac 3AH. Chokshi, J. R. Porter, J. Am. Ceram Soc. 68C(1985)144 4J.R. Porter, FF. Lange, A H. Chokshi, Am. Ceram Soc. Bull cumulation differ. With SiCw/ALO3 products, creep (1985)343. deformation appears to occur by grain b 5S.R. Nutt, P. Lipetzky, P F. Becher, Mater. Sci. Eng. A126 g, with the resulting cavity formation and crack de- (1990)165 velopment leading to eventual failure. In contrast [6 K. Xia, T.G. Langdon, Mater. Res. Soc. Symp Proc. 120(1988) over the stress-temperature ranges investigated, the [7 H.T. Lin, P.F. Becher, J. Am. Ceram Soc. 74(1991)1886 tensile creep and creep fracture properties of the o/ 8J.C. Romero. R.J. Arsenalt, R.F. Krause Jr, Mater. Sci. Eng. 90 SIC:/Al,O3 composite are determined by the creep A201(1995)13 strength of the NicalonTM fibres 9]SR. Nutt, P. Lipetzky, Mater. Sci. Eng. A166 Creep of the fibres is accompanied by cracking of [0]AR. de Arellano-Lopez, F.Cumbrera,A driguez. K.C. Goretta, J. L. Routbort, J. Am. the weak porous matrix. Cracks developing normal to (1990)1297 the tensile axis are bridged by the longitudinal fibres, [l P. Lipetzky, S.R. Nutt, D.A. Koester, R.F. Davies, J. Am with the rate of crack growth and the time to frac Ceran.Soc.74(1991)1240 [2C. OMeara, T. Suihkonen, T. Hansson, R. Warren, Mater. Sci. ture determined by the rate of strain accumulation EngA209(1996)251 allowed by the fibres. However, oxygen penetration [13] K M. Prewo, J Mater. Sci. 21(1986)3590 during tests in air promotes fibre failure, so zones [14] J.J. Brennan, Mater. Sci. Res. 20(1986)546 characterized by oxidized fibres are therefore observed [15]AG. Evans, B.D. Marshall, Acta Metall. 37(1989)2657 [16 M.S. Newkirk, A.w. Urquhart, H.R. Zwicker, E. Breval, J over substantial areas of the fracture surfaces before Mater Res. I(1986)81. failure finally occurs by fibre pull out [17 M.S. Newkirk, H D. Lesher, D.R. White, C.R. Kennedy, A w The creep and creep fracture strengths of the 0/900 Urquhart, T D. Clear, Ceram. Eng. Sci. Proc. 8(1987)879. SiCH Al2O3 composite decrease with increasing test du- [18 B. Wilshire, H. Jiang, Br. Ceram. Trans. 93(1994)213 ration, particularly at high temperatures, reflecting the [19 F.A. Kandil, B F. Dyson, in: B. Wilshire, R W. Evans(Eds ) Creep and Fracture of Engineering Materials and Structures, deterioration in the creep resistance of the fibres with The Institute of Metals, London, 1990, p. 409 time and temperature. However, the ability to with- [20]G. Simon, A.R. Bunsell, J. Mater. Sci. 19(1984)3670 stand extensive crack development without immediate 21C.H. Henager, R.H. Jones, Mater. Sci Eng. A166(1993)211 failure taking place at low stresses, combined with a [23]EC Monkman, N J. Grant, Proc. ASTM. 56(1956)593 tensile creep strength markedly superior to that of [24]G. Simon, A. Bunsell, J Mater. Sci. 19(1984)3649 whisker-reinforced materials, emphasizes the potential (25 T Mah, N. Hecht, D. McCullum, et al., J Mater. Sci. 19(192 for development of continuous fibre-reinforced CMCs I191 26K. Luthra, J. Am. Ceram. (1986)231 for advanced engineering applications involving long 27 B. Wilshire, Br. Ceram. TI (1995)57 term service exposure at elevated temperatures 28 P C. Dokko, J.A. Pask, Mater. Sci. Eng. 25(1976)77
44 B. Wilshire, F. Carren˜o / Materials Science and Engineering A272 (1999) 38–44 ucts, (Figs. 5 and 8). Moreover, with the 0/90° fibre architecture, large cracks can develop without causing immediate failure (Fig. 7), especially when high creep ductilities are displayed as the test durations increase towards the service lives expected for aeroengine components (Fig. 11). Hence, crack tolerance is combined with impressive stress-bearing capabilities at high temperatures, illustrating the potential of continuous fibre-reinforced CMCs for high-performance engineering applications. 5. Conclusions Under high-temperature tensile creep conditions, similar behaviour patterns are displayed by SiCf / Al2O3 and SiCw/Al2O3 composites, although the mechanisms controlling deformation and damage accumulation differ. With SiCw/Al2O3 products, creep deformation appears to occur by grain boundary sliding, with the resulting cavity formation and crack development leading to eventual failure. In contrast, over the stress-temperature ranges investigated, the tensile creep and creep fracture properties of the 0/ 90° SiCf /Al2O3 composite are determined by the creep strength of the Nicalon™ fibres. Creep of the fibres is accompanied by cracking of the weak porous matrix. Cracks developing normal to the tensile axis are bridged by the longitudinal fibres, with the rate of crack growth and the time to fracture determined by the rate of strain accumulation allowed by the fibres. However, oxygen penetration during tests in air promotes fibre failure, so zones characterized by oxidized fibres are therefore observed over substantial areas of the fracture surfaces, before failure finally occurs by fibre pull out. The creep and creep fracture strengths of the 0/90° SiCf /Al2O3 composite decrease with increasing test duration, particularly at high temperatures, reflecting the deterioration in the creep resistance of the fibres with time and temperature. However, the ability to withstand extensive crack development without immediate failure taking place at low stresses, combined with a tensile creep strength markedly superior to that of whisker-reinforced materials, emphasizes the potential for development of continuous fibre-reinforced CMCs for advanced engineering applications involving longterm service exposure at elevated temperatures. Acknowledgements The provision of a post-doctoral research fellowship for Dr Carren˜o under the EC Human Capital and Mobility Programme (Contract No. ERBCHBGCT9330303) and the support received through the British Council UK/Spain research collaboration initiative are gratefully acknowledged. The authors also wish to thank Dr M.J.L. Percival for helpful discussions and Rolls-Royce plc for financial support and materials supplies. References [1] P.F. Becher, G.C. Wei, J. Am. Ceram. Soc. 67C (1984) 267. [2] P.F. Becher, C.H. Hsueh, P. Angelini, T.N. Tiegs, J. Am. Ceram. Soc. 71 (1988) 1050. [3] A.H. Chokshi, J.R. Porter, J. Am. Ceram. Soc. 68C (1985) 144. [4] J.R. Porter, F.F. Lange, A.H. Chokshi, Am. Ceram. Soc. Bull. 66 (1985) 343. [5] S.R. Nutt, P. Lipetzky, P.F. Becher, Mater. Sci. Eng. A126 (1990) 165. [6] K. Xia, T.G. Langdon, Mater. Res. Soc. Symp. Proc. 120 (1988) 265. [7] H.T. Lin, P.F. Becher, J. Am. Ceram. Soc. 74 (1991) 1886. [8] J.C. Romero, R.J. Arsenalt, R.F. Krause Jr, Mater. Sci. Eng. A201 (1995) 13. [9] S.R. Nutt, P. Lipetzky, Mater. Sci. Eng. A166 (1993) 199. [10] A.R. de Arellano-Lo´pez, F. Cumbrera, A. Domı´nguez-Rodrı´guez, K.C. Goretta, J.L. Routbort, J. Am. Ceram. Soc. 73 (1990) 1297. [11] P. Lipetzky, S.R. Nutt, D.A. Koester, R.F. Davies, J. Am. Ceram. Soc. 74 (1991) 1240. [12] C. O’Meara, T. Suihkonen, T. Hansson, R. Warren, Mater. Sci. Eng. A209 (1996) 251. [13] K.M. Prewo, J. Mater. Sci. 21 (1986) 3590. [14] J.J. Brennan, Mater. Sci. Res. 20 (1986) 546. [15] A.G. Evans, B.D. Marshall, Acta Metall. 37 (1989) 2657. [16] M.S. Newkirk, A.W. Urquhart, H.R. Zwicker, E. Breval, J. Mater. Res. 1 (1986) 81. [17] M.S. Newkirk, H.D. Lesher, D.R. White, C.R. Kennedy, A.W. Urquhart, T.D. Clear, Ceram. Eng. Sci. Proc. 8 (1987) 879. [18] B. Wilshire, H. Jiang, Br. Ceram. Trans. 93 (1994) 213. [19] F.A. Kandil, B.F. Dyson, in: B. Wilshire, R.W. Evans (Eds.), Creep and Fracture of Engineering Materials and Structures, The Institute of Metals, London, 1990, p. 409. [20] G. Simon, A.R. Bunsell, J. Mater. Sci. 19 (1984) 3670. [21] C.H. Henager, R.H. Jones, Mater. Sci. Eng. A166 (1993) 211. [22] A.G. Evans, C. Weber, Mater. Sci. Eng. A208 (1996) 1. [23] F.C. Monkman, N.J. Grant, Proc. ASTM. 56 (1956) 593. [24] G. Simon, A. Bunsell, J. Mater. Sci. 19 (1984) 3649. [25] T. Mah, N. Hecht, D. McCullum, et al., J. Mater. Sci. 19 (1984) 1191. [26] K. Luthra, J. Am. Ceram. Soc. 69 (1986) 231. [27] B. Wilshire, Br. Ceram. Trans. 94 (1995) 57. [28] P.C. Dokko, J.A. Pask, Mater. Sci. Eng. 25 (1976) 77.