《复合材料 Composites》课程教学资源（学习资料）第五章 陶瓷基复合材料_6High-temperature mechanical behaviour of flaw tolerant

Availableonlineatwww.sciencedirect.co ScienceDirect Acta materialia ELSEVIER Acta Materialia 55(2007)4891-4901 High-temperature mechanical behaviour of flaw tolerant alumina-zirconia multilayered ceramics R. Bermejo,, A.J. Sanchez-Herencia", L Llanes, C. Baudin Instituto de Ceramica y Vidrio(CSIC), C/ Kelsen Institut fiir Struktur- tend Funktionskeramik (ISFK), Peter-Thnner StraBe 5, 8700 Leoben, Austria Departamento de Ciencia de los Materiales e Ingenieria Metalirgica, ETSEIB, Universidad Politecnica de Cataluiia (UPC) Io. Diagonal 647, 08028 Barcelona, spain Received 3 April 2007: received in revised form 3 May 2007: accepted 7 May 2007 vailable online 26 June 2007 Abstract The mechanical behaviour of alumina-zirconia multilayered ceramics designed with thin internal compressive layers has been inves- tigated under flexural loading at room and high temperature Youngs modulus and the sintering evolution of each layer have been exper- imentally determined up to 1200C, to account for the residual stress distribution in the layered composite. The fracture behaviour has been assessed by indentation- strength experiments at different temperatures and by a fracture mechanics analysis. Experimental find ings showed that improvement in mechanical properties of the laminate at high temperatures in comparison to the alumina-based mono- lithic material was essentially related to the distinct modes of failure observed as a function of the temperature, in the presence of energy release mechanisms such as crack bifurcation and/or delamination that may be used as a tool for designing tolerant materials at higl temperatures o 2007 Published by Elsevier Ltd on behalf of Acta Materialia Inc Keywords: Layered structures; Residual stresses; High temperature; Toughness; Flaw tolerance 1. Introduction ceramic materials. The so-called flaw elimination approach envisages the production of the highest degree Ceramic materials are the most suitable substitutes for of homogeneity in bulk monophase ceramics with very metals in structural applications that involve high temper- small flaws. However, new strategies fundamentally differ- atures in severe erosive and corrosive environments and/or ent from this conventional approach have emerged aiming under compressive loads. However, the main drawback of to achieve"flaw tolerant'"materials by designing special ceramics is associated with their brittle mode of failure that microstructures that improve the toughness(or apparent implies an extreme variation of the strength of different toughness) of ceramics. One of the most attractive propos- components within the same batch as a function of the flaw als for this latter approach consists of layered architectures size distribution. Therefore, the main requirement for the that combine materials with different properties. As a extension of ceramic use in structural applications is to result, laminates with mechanical behaviour superior to avoid this lack of strength reliability that of the individual constituents can be fabricated. In this In the last three decades, remarkable advances have regard, different fracture mechanics appi roaches been achieved in improving the mechanical behaviour of attempted for the design of ic layered composites. Residual stress free laminates constituted by layers sepa nding author. Address: Instituto de Ceramica y Vidrio rated by weak interfaces between themselves and laminates CSIC), C/Kelsen 5. 28049 Madrid, Spain that combine stiff and high strength external layers with E-mail address: raul bermejo(@mu-leoben at(R. Bermejo) flaw tolerant internal layers have been proposed to 1359-6454530.00 o 2007 Published by Elsevier Ltd on behalf of Acta Materialia Inc. do: 10. 1016/j.actamat. 2007.05.005

High-temperature mechanical behaviour of flaw tolerant alumina–zirconia multilayered ceramics R. Bermejo a,b,*, A.J. Sa´nchez-Herencia a , L. Llanes c , C. Baudı´n a a Instituto de Cera´mica y Vidrio (CSIC), C/ Kelsen 5, 28049 Madrid, Spain b Institut fu¨r Struktur- und Funktionskeramik (ISFK), Peter-Tunner Straße 5, 8700 Leoben, Austria c Departamento de Ciencia de los Materiales e Ingenierı´a Metalu´rgica, ETSEIB, Universidad Polite´cnica de Catalun˜a (UPC), Av. Diagonal 647, 08028 Barcelona, Spain Received 3 April 2007; received in revised form 3 May 2007; accepted 7 May 2007 Available online 26 June 2007 Abstract The mechanical behaviour of alumina–zirconia multilayered ceramics designed with thin internal compressive layers has been inves￾tigated under flexural loading at room and high temperature. Young’s modulus and the sintering evolution of each layer have been exper￾imentally determined up to 1200 C, to account for the residual stress distribution in the layered composite. The fracture behaviour has been assessed by indentation – strength experiments at different temperatures and by a fracture mechanics analysis. Experimental find￾ings showed that improvement in mechanical properties of the laminate at high temperatures in comparison to the alumina-based mono￾lithic material was essentially related to the distinct modes of failure observed as a function of the temperature, in the presence of energy release mechanisms such as crack bifurcation and/or delamination that may be used as a tool for designing tolerant materials at high temperatures. 2007 Published by Elsevier Ltd on behalf of Acta Materialia Inc. Keywords: Layered structures; Residual stresses; High temperature; Toughness; Flaw tolerance 1. Introduction Ceramic materials are the most suitable substitutes for metals in structural applications that involve high temper￾atures in severe erosive and corrosive environments and/or under compressive loads. However, the main drawback of ceramics is associated with their brittle mode of failure that implies an extreme variation of the strength of different components within the same batch as a function of the flaw size distribution. Therefore, the main requirement for the extension of ceramic use in structural applications is to avoid this lack of strength reliability. In the last three decades, remarkable advances have been achieved in improving the mechanical behaviour of ceramic materials. The so-called ‘‘flaw elimination’’ approach envisages the production of the highest degree of homogeneity in bulk monophase ceramics with very small flaws. However, new strategies fundamentally differ￾ent from this conventional approach have emerged aiming to achieve ‘‘flaw tolerant’’ materials by designing special microstructures that improve the toughness (or apparent toughness) of ceramics. One of the most attractive propos￾als for this latter approach consists of layered architectures that combine materials with different properties. As a result, laminates with mechanical behaviour superior to that of the individual constituents can be fabricated. In this regard, different fracture mechanics approaches have been attempted for the design of ceramic layered composites. Residual stress free laminates constituted by layers sepa￾rated by weak interfaces between themselves and laminates that combine stiff and high strength external layers with flaw tolerant internal layers have been proposed to 1359-6454/$30.00 2007 Published by Elsevier Ltd on behalf of Acta Materialia Inc. doi:10.1016/j.actamat.2007.05.005 * Corresponding author. Address: Instituto de Cera´mica y Vidrio (CSIC), C/ Kelsen 5, 28049 Madrid, Spain. E-mail address: raul.bermejo@mu-leoben.at (R. Bermejo). www.elsevier.com/locate/actamat Acta Materialia 55 (2007) 4891–4901

R Bermejo et al. Acta Materialia 55(2007)4891-4901 promote sufficient strength with significant flaw tolerance The temperature effect can be of extreme importance for to delamination and/or crack deflection processes [1-11]. the performance of laminates designed on the basis of Another large family comprises materials designed on residual stresses when the temperature of use approaches the basis of residual stress development in the layers during that of stress relaxation. In this sense, although room tem- cooling from the sintering temperature [12-24], alumina- perature mechanical data in flexure have been reported for zirconia being the most studied layered architecture alumina-yttria tetragonal zirconia polycrystal (YTZP [12, 13, 15, 16, 20-24]. In this system, a broad interval of ther- laminates, claiming high strength and/or toughening and mal strains can be reached owing to differences between the R-curve behaviour, the high-temperature mechanical thermal expansion coefficient of the constituent layers [12- response has only been characterised and compared to that ion, significant impro ers of the laminate are in compres- [13, 15, 16) In de unie als. with the same composition as mation[20,21,24 the layers, by uniaxial compression creep experiments cases, laminates were more creep resistant ements in strength can be achieved when strained in the direction parallel to the interfaces than [22,23]. This is the so-called strengthening approach, simi- when deformed in the perpendicular direction, owing to the lar to that traditionally used in glasses [17, 25]. Moreover, constraints imposed by the more creep resistant constituent R-curve behaviour revealing flaw tolerance is also observed layers. Creep exponents for the alumina/YTZP laminates this type of laminate for flaws embedded in the external were similar to those determined for the monoliths, indicat ayer [23]. On the other hand, a challenging approach is ing coincident mass transport mechanisms. The overall that of using layered ceramics with thin internal layers mechanical behaviour of these laminates was either better under high compressive stresses and thick external layers or fell in between that of the monoliths. An important under low tensile ones [14, 19, 21, 24, 26]. When tested in flex- characteristic of the alumina/YTZP layered ceramics stud- ure, the compressive layers would hinder the crack propa- ied was that the interfaces between layers maintained their gation through the rest of the material yielding as a result a structural integrity after testing sole'critical flaw size, increasing the material reliability In a recent investigation [24], an Al2O3-5 vol % t-ZrO2/ as well as its apparent fracture toughness and work of Al2O3-30 voL m-ZrO2(t= tetragonal and m=mono- acture clinic) layered composite designed with high compressive Extensive work has been reported on the design, pro- internal layers showed a threshold strength under flexural cessing and room temperature mechanical behaviour of loading at room temperature, where the thin compressive ceramic laminates. In contrast, even though the main envis- layers acted as a barrier to crack extension. Moreover, aged applications for ceramics as metal substitutes would the further propagation of the critical flaw throughout involve relatively high temperatures, works aiming to the layered structure took place through deflection/bifurca investigate the mechanical behaviour of ceramic laminates tion mechanisms, yielding as a result an increase in appar- under these conditions are rather scarce [3-6, 9, 12, 13, 15, 16]. ent fracture toughness and fracture energy in comparison Within this context, residual stress free laminates with to a monolithic material that has the same composition weak interfaces between layers, designed for crack deflec- as the external layer, taken as a reference. The purpose of tion and delamination, have been tested at high tempera- the present investigation is to assess the mechanical behav tures. Such experiments showed how the reaction iour of this layered system under high-temperature condi between the layers and/or interactions with the atmosphere tions. Four-point bending tests were performed at at high temperatures can dramatically change the expected different temperatures on indented laminated specimens mechanical behaviour of the laminate. For instance, the and on the reference material for comparative purposes. crack deflection capability and associated"graceful fail- Young,'s modulus over the temperature range of study ure"in flexure(four-point bending) of a laminate consti- and deformation during cooling from the sintering temper tuted by alternated thick Si3 N4 and thin bn layers ature of monoliths of the same compositions as those of the decreased, and appreciable plastic flow occurred at high constituent layers were determined in order to analyse the temperature(1400C)owing to oxidation [5]. Also the evolution of the distribution of residual stresses through capability for crack deflection at the interfaces disappeared the laminate with temperature. The effect of temperature at 1300C in laminates consisting of alumina/MoSi2+ on the mechanical response of the multilayered material Mo2Bs and alumina/TiC-MoSi2+Mo2 Bs layers tested in is discussed in terms of the variation with the temperature flexure(three-point bending), indicating that the strength of:(i)the elastic properties of the layers, (ii) the residua of the bonding between layers increased with temperature stress profile and (iii) the apparent fracture toughness [3, 6]. To overcome the degradation of mechanical behav iour in laminates designed for crack deflection caused by 2. Experimental the strengthening of the interfaces alumina based lami- nates formed by dense alumina layers linked by porous alu- 2. 1. Materials mina interfaces have been proposed. Full crack deflection and graceful failure have been observed at testing temper- A laminar ceramic consisting of alternating layers of atures up to1200°C[4] Al2O3-5 vol t-ZrO2, named ATZ, and Al,O3-30 vol%

promote sufficient strength with significant flaw tolerance to delamination and/or crack deflection processes [1–11]. Another large family comprises materials designed on the basis of residual stress development in the layers during cooling from the sintering temperature [12–24], alumina– zirconia being the most studied layered architecture [12,13,15,16,20–24]. In this system, a broad interval of ther￾mal strains can be reached owing to differences between the thermal expansion coefficient of the constituent layers [12– 16,22,23] and/or different levels of zirconia phase transfor￾mation [20,21,24]. When the external layers of the laminate are in compres￾sion, significant improvements in strength can be achieved [22,23]. This is the so-called strengthening approach, simi￾lar to that traditionally used in glasses [17,25]. Moreover, R-curve behaviour revealing flaw tolerance is also observed in this type of laminate for flaws embedded in the external layer [23]. On the other hand, a challenging approach is that of using layered ceramics with thin internal layers under high compressive stresses and thick external layers under low tensile ones [14,19,21,24,26]. When tested in flex￾ure, the compressive layers would hinder the crack propa￾gation through the rest of the material yielding as a result a ‘‘sole’’ critical flaw size, increasing the material reliability as well as its apparent fracture toughness and work of fracture. Extensive work has been reported on the design, pro￾cessing and room temperature mechanical behaviour of ceramic laminates. In contrast, even though the main envis￾aged applications for ceramics as metal substitutes would involve relatively high temperatures, works aiming to investigate the mechanical behaviour of ceramic laminates under these conditions are rather scarce [3–6,9,12,13,15,16]. Within this context, residual stress free laminates with weak interfaces between layers, designed for crack deflec￾tion and delamination, have been tested at high tempera￾tures. Such experiments showed how the reaction between the layers and/or interactions with the atmosphere at high temperatures can dramatically change the expected mechanical behaviour of the laminate. For instance, the crack deflection capability and associated ‘‘graceful fail￾ure’’ in flexure (four-point bending) of a laminate consti￾tuted by alternated thick Si3N4 and thin BN layers decreased, and appreciable plastic flow occurred at high temperature (1400 C) owing to oxidation [5]. Also the capability for crack deflection at the interfaces disappeared at 1300 C in laminates consisting of alumina/MoSi2+ Mo2B5 and alumina/TiC–MoSi2+Mo2B5 layers tested in flexure (three-point bending), indicating that the strength of the bonding between layers increased with temperature [3,6]. To overcome the degradation of mechanical behav￾iour in laminates designed for crack deflection caused by the strengthening of the interfaces, alumina based lami￾nates formed by dense alumina layers linked by porous alu￾mina interfaces have been proposed. Full crack deflection and graceful failure have been observed at testing temper￾atures up to 1200 C [4]. The temperature effect can be of extreme importance for the performance of laminates designed on the basis of residual stresses when the temperature of use approaches that of stress relaxation. In this sense, although room tem￾perature mechanical data in flexure have been reported for alumina–yttria tetragonal zirconia polycrystal (YTZP) laminates, claiming high strength and/or toughening and R-curve behaviour, the high-temperature mechanical response has only been characterised and compared to that of monolithic materials, with the same composition as the layers, by uniaxial compression creep experiments [13,15,16]. In all cases, laminates were more creep resistant when strained in the direction parallel to the interfaces than when deformed in the perpendicular direction, owing to the constraints imposed by the more creep resistant constituent layers. Creep exponents for the alumina/YTZP laminates were similar to those determined for the monoliths, indicat￾ing coincident mass transport mechanisms. The overall mechanical behaviour of these laminates was either better or fell in between that of the monoliths. An important characteristic of the alumina/YTZP layered ceramics stud￾ied was that the interfaces between layers maintained their structural integrity after testing. In a recent investigation [24], an Al2O3–5 vol.% t-ZrO2/ Al2O3–30 vol.% m-ZrO2 (t = tetragonal and m = mono￾clinic) layered composite designed with high compressive internal layers showed a threshold strength under flexural loading at room temperature, where the thin compressive layers acted as a barrier to crack extension. Moreover, the further propagation of the critical flaw throughout the layered structure took place through deflection/bifurca￾tion mechanisms, yielding as a result an increase in appar￾ent fracture toughness and fracture energy in comparison to a monolithic material that has the same composition as the external layer, taken as a reference. The purpose of the present investigation is to assess the mechanical behav￾iour of this layered system under high-temperature condi￾tions. Four-point bending tests were performed at different temperatures on indented laminated specimens and on the reference material for comparative purposes. Young’s modulus over the temperature range of study and deformation during cooling from the sintering temper￾ature of monoliths of the same compositions as those of the constituent layers were determined in order to analyse the evolution of the distribution of residual stresses through the laminate with temperature. The effect of temperature on the mechanical response of the multilayered material is discussed in terms of the variation with the temperature of: (i) the elastic properties of the layers, (ii) the residual stress profile and (iii) the apparent fracture toughness. 2. Experimental 2.1. Materials A laminar ceramic consisting of alternating layers of Al2O3–5 vol.% t-ZrO2, named ATZ, and Al2O3–30 vol.% 4892 R. Bermejo et al. / Acta Materialia 55 (2007) 4891–4901

ermejo et al. Acta Materialia 55(2007)4891-4901 m-ZrO2, referred to as AMZ, as well as the corresponding under displacement control, at a rate of 0.05 mm min- ATZ and AMZ monoliths were fabricated by slip casting using a universal testing machine model Instron 8562 following a procedure described elsewhere [27]. Samples Great Britain) with an electrical furnace. Mechanical test were sintered at 1550C for 2 h using heating and cooling ing was performed at different temperatures and after dif- rates of 5C min. As a result, symmetrical laminates ferent thermal histories, i.e. room temperature after with four thin AMZ layers sandwiched between five thick sintering, 800C reached on heating(before m-t trans- ATZ layers as well as monolithic specimens of composi- formation), 1200C reached on heating, 800C reached tions ATZ and AMZ were obtained. After sintering, spec- on cooling from 1200C (after the m-t transformation imens were polished with diamond paste down to I um for on heating and before the reverse transformation on cool- SEM observation. The density of the sintered ATZ and ing), and 650C cooling(after t-m transformation). AMZ monolithic samples was measured by the archimedes All the fractured specimens were inspected by reflected method in water. Additionally, an XRD analysis was also light optical microscopy and ng electron microscopy carried out in the monoliths for composition and phase(DSM-950, Zeiss, Germany) to determine the type, size and identification. Finally, bars of approximately 3. 6 mm x location of the failure-controlling flaws The load-displace- 3.2 mm x 45 mm were diamond machined for mechanical ment curves were recorded using the software coupled to characterisation the testing set-up, and the engineering stress was calculated using the load values and the dimensions of the specimens 2. 2. Elastic properties evaluation and the spans, assuming linear elastic behaviour. Since the elastic properties of the laminate vary through the different Youngs modulus, E; of both ATZ and AMZ monoliths layers, the failure stress for the indented specimens, or, was was evaluated between room(20C) and high (1250C) calculated using the following equation temperature using the impulse excitation technique (IET) e8, following the guidelines provided by ASTM E 1876- GR,E.(-yna 99 and ENV-843-2. Dimensional changes in sintered monolithic samples(10 mm length and cross section of where Ei is the Young's modulus of the corresponding 5x5mm)were determined on heating up to 1250C layer, M is the moment for the case of four-point bending and on cooling using a dilatometer(D1-24, Adamel Lhom- tests(M=Fa, where F is the applied load at failure and a argy, France)with alumina rod and calibrated with a plat- the distance between inner and outer spans), yna is the po- inum standard of similar size as that of the testing sition of the neutral axis and EI the flexural rigidity of the composite calculated for bending perpendicular to the lay ers considering the corresponding Youngs modulus of 23. Residual stress estimation each layer [29-31]. n order to evaluate the residual stress profile within the 3. Results and discussion tridimensional multilayered structure owing to the thermal strain mismatch between adjacent layers, a three-dimen- 3. 1. Microstructural characterisation sional(3D)finite element model developed elsewhere [20] was implemented using the FE code ANSYS 10.0. Pris- Densities of the sintered atz and amz monoliths were matic bar-shaped specimens were modelled as representa-.5% and 98.5% of the theoretical values, respectively tive of the specimens utilised in the experiments for The XRD analysis revealed a 95.5 vol. of alumina and material characterisation, and the elastic and thermal prop 4.5 vol. of tetragonal zirconia(t-ZrO2) in the ATZ com- erties previously determined were introduced as a function pacts, while for the AMz samples a 70 vol % of alumina of the temperature and a 30 vol% of monoclinic zirconia (m-ZrO2)were detected [32]. a phase analysis of the zirconia content in 2.4. Flexural tests the AMz compacts indicated that most of the zirconia had transformed into the monoclinic phase, and only 5- Indentation tests were performed using a Vickers inden- 10% remained as the tetragonal phase in the sintered com ter (Microtest, Spain) with a displacement rate of pacts [32]. Further microstructural observations in the bulk 0. 1 mm min up to reach a maximum load of 100N and AMZ compacts showed small microcracks in the alumina- holding time of 10 s. Three indentations were placed in zirconia grain boundaries, as observed by other authors the middle of one of the outer layers along the longitudinal [33; the volume expansion associated with the zirconia direction of the sample with an offset separation of 2 mm to transformation from tetragonal to monoclinic phase on avoid any crack interaction. Indented samples were frac- cooling led to the formation of radial microcracks which tured in a fully articulated alumina four-point bending emerged from the transformed zirconia grains in the alu- device(the loading axis normal to the layer plane and the mina matrix. Regarding the laminar composite, uniform indented layer in tension) with inner and outer spans of layer thickness was obtained for the ATZ and AMZ layers, 20 mm and 40 mm, respectively. Tests were carried out resulting in 530+ 10 um and 100+5 um, respectively

m-ZrO2, referred to as AMZ, as well as the corresponding ATZ and AMZ monoliths were fabricated by slip casting following a procedure described elsewhere [27]. Samples were sintered at 1550 C for 2 h using heating and cooling rates of 5 C min1 . As a result, symmetrical laminates with four thin AMZ layers sandwiched between five thick ATZ layers as well as monolithic specimens of composi￾tions ATZ and AMZ were obtained. After sintering, spec￾imens were polished with diamond paste down to 1 lm for SEM observation. The density of the sintered ATZ and AMZ monolithic samples was measured by the Archimedes method in water. Additionally, an XRD analysis was also carried out in the monoliths for composition and phase identification. Finally, bars of approximately 3.6 mm · 3.2 mm · 45 mm were diamond machined for mechanical characterisation. 2.2. Elastic properties evaluation Young’s modulus, Ei, of both ATZ and AMZ monoliths was evaluated between room (20 C) and high (1250 C) temperature using the impulse excitation technique (IET) [28], following the guidelines provided by ASTM E 1876- 99 and ENV-843-2. Dimensional changes in sintered monolithic samples (10 mm length and cross section of 5 · 5 mm2 ) were determined on heating up to 1250 C and on cooling using a dilatometer (DI-24, Adamel Lhom￾argy, France) with alumina rod and calibrated with a plat￾inum standard of similar size as that of the testing specimens. 2.3. Residual stress estimation In order to evaluate the residual stress profile within the tridimensional multilayered structure owing to the thermal strain mismatch between adjacent layers, a three-dimen￾sional (3D) finite element model developed elsewhere [20] was implemented using the FE code ANSYS 10.0. Pris￾matic bar-shaped specimens were modelled as representa￾tive of the specimens utilised in the experiments for material characterisation, and the elastic and thermal prop￾erties previously determined were introduced as a function of the temperature. 2.4. Flexural tests Indentation tests were performed using a Vickers inden￾ter (Microtest, Spain) with a displacement rate of 0.1 mm min1 up to reach a maximum load of 100 N and holding time of 10 s. Three indentations were placed in the middle of one of the outer layers along the longitudinal direction of the sample with an offset separation of 2 mm to avoid any crack interaction. Indented samples were frac￾tured in a fully articulated alumina four-point bending device (the loading axis normal to the layer plane and the indented layer in tension) with inner and outer spans of 20 mm and 40 mm, respectively. Tests were carried out under displacement control, at a rate of 0.05 mm min1 , using a universal testing machine model Instron 8562 (Great Britain) with an electrical furnace. Mechanical test￾ing was performed at different temperatures and after dif￾ferent thermal histories, i.e. room temperature after sintering, 800 C reached on heating (before m ! t trans￾formation), 1200 C reached on heating, 800 C reached on cooling from 1200 C (after the m ! t transformation on heating and before the reverse transformation on cool￾ing), and 650 C cooling (after t ! m transformation). All the fractured specimens were inspected by reflected light optical microscopy and scanning electron microscopy (DSM-950, Zeiss, Germany) to determine the type, size and location of the failure-controlling flaws. The load–displace￾ment curves were recorded using the software coupled to the testing set-up, and the engineering stress was calculated using the load values and the dimensions of the specimens and the spans, assuming linear elastic behaviour. Since the elastic properties of the laminate vary through the different layers, the failure stress for the indented specimens, rRi , was calculated using the following equation: rRi ¼ EiM EI ðy ynaÞ ð1Þ where Ei is the Young’s modulus of the corresponding layer, M is the moment for the case of four-point bending tests (M = Fa, where F is the applied load at failure and a the distance between inner and outer spans), yna is the po￾sition of the neutral axis and EI the flexural rigidity of the composite calculated for bending perpendicular to the lay￾ers considering the corresponding Young’s modulus of each layer [29–31]. 3. Results and discussion 3.1. Microstructural characterisation Densities of the sintered ATZ and AMZ monoliths were 99.5% and 98.5% of the theoretical values, respectively. The XRD analysis revealed a 95.5 vol.% of alumina and 4.5 vol.% of tetragonal zirconia (t-ZrO2) in the ATZ com￾pacts, while for the AMZ samples a 70 vol.% of alumina and a 30 vol.% of monoclinic zirconia (m-ZrO2) were detected [32]. A phase analysis of the zirconia content in the AMZ compacts indicated that most of the zirconia had transformed into the monoclinic phase, and only 5– 10% remained as the tetragonal phase in the sintered com￾pacts [32]. Further microstructural observations in the bulk AMZ compacts showed small microcracks in the alumina– zirconia grain boundaries, as observed by other authors [33]; the volume expansion associated with the zirconia transformation from tetragonal to monoclinic phase on cooling led to the formation of radial microcracks which emerged from the transformed zirconia grains in the alu￾mina matrix. Regarding the laminar composite, uniform layer thickness was obtained for the ATZ and AMZ layers, resulting in 530 ± 10 lm and 100 ± 5 lm, respectively. R. Bermejo et al. / Acta Materialia 55 (2007) 4891–4901 4893

R Bermejo et al. Acta Materialia 55(2007)4891-4901 3.2. Young s modulus, thermal expans d residual stress evolution with temperature ng. I shows the Youngs modulus of the AMz an ATZ monolith, Young's modulus varied following the clas- 兰 sical relationship(1% every 100C)found by Watchman et al. for a large number of ceramics [34] ranging from 9390 GPa at room temperature to 344 GPa at 1200C AT7 and did not show hysteresis in the heating-cooling cycle AMZ For the AMz specimens, the elastic modulus decreased from room temperature(≈290GPa)upto900°C 2 25050075010001250 221 GPa)during heating, this decrease(22.6% for every 100C) being significantly higher than for ATZ. The rela tive low value determined at room temperature(as com- Fig. 2. Dilatometry curves for the ATZ and AMz monoliths. Circles that obtained with the rule of mixtures correspond to the different temperatures(indicated by th considering 400 GPa for alumina [34] and 244 GPa for flexural tests were carried out monophase monoclinic zirconia [35] may be related to the presence of microcracks caused by the high content of monoclinic zirconia in these compacts, as discussed heating and cooling can be attributed to experimental above. From about 860C, values started to increase, first uncertainties). This behaviour, together with the observed moothly and then sharply from 950C up to 1150C. The variation of Youngs modulus proves that no significant initial increase might be associated with stiffening of the microstructural changes occurred during the thermal material caused by closure of the microcracks whereas cycling of the sintered ATZ compacts. Conversely, the vol the sharp increase has to be attributed to the transforma- ume changes associated with the reversible zirconia phase ion of zirconia from the monoclinic(Ex 164 GPa at transformation, from monoclinic to tetragonal during 950C)to the tetragonal phase(E N 220 GPa at 1200C) heating above 1150C and below 725C on cooling, con- 65]. The effect of the inverse transformation was observed ditioned the dimensional changes of the AMZ monoliths. on cooling where, after a slight increase, an abrupt decrease The expected differences in thermal strain of the layers, occurred at the temperature of the tetragonal to monoclinic referred to as Ae, which will condition the residual stress zirconia phase transformation(725C). After this sharp state in the layered structure, can be drawn from the decrease, the Youngs modulus continued rising as the tem- dimensional changes experienced by the monoliths perature fell to room temperature, where a similar value to(Fig. 2). At high temperature, where mass transport mech the initial room temperature was achieved anisms are active, the strain mismatch between layers The dilatometer curves during heating and cooling of caused by differences in thermal expansion would be the sintered atz and Amz monoliths are shown in accommodated. Therefore, the strain mismatch between Fig. 2. For ATZ, both curves were monotonous and with- layers at room temperature can be derived by taking out any significant hysteresis(the small differences between 1200C as zero point for both materials and considering the dimensional changes from 1200C to room ture. In doing so, the thermal strain mismatch between AE=EAMZ-EATZ 0.00212 [271, where EAMz and Eat refer to the thermal strains of the amz and atz mono- ATZ liths from Fig. 2. as discussed in the introduction. the mechanical esponse of the layered material at a given temperature will AMZ be determined by the residual stress field within the layers which, in turn, would be associated with two features: ( E modulus heating the elastic properties at the given temperature, i.e. ET E modulus cooling and (ii) the thermal strain mismatch between layers namely AE(T). Hence, an evaluation of the residual stress 020040060080010001200 state at every testing temperature was carried out. In doing modulus variation with temperature of the AtZ and esidual stress calculation over the temperature range The corresponding elastic and thermal properties of each curves s. The abrupt change in slope on the heating and cooling layer, l.e. Young s modulus and thermal strain, were intro- AMz is due to the m - t and t zirconia phase transformations, respectively. duced as a function of temperature

3.2. Young’s modulus, thermal expansion and residual stress evolution with temperature Fig. 1 shows the Young’s modulus of the AMZ and ATZ monoliths as a function of temperature. For the ATZ monolith, Young’s modulus varied following the clas￾sical relationship (1% every 100 C) found by Watchman et al. for a large number of ceramics [34], ranging from 390 GPa at room temperature to 344 GPa at 1200 C and did not show hysteresis in the heating–cooling cycle. For the AMZ specimens, the elastic modulus decreased from room temperature (290 GPa) up to 900 C (221 GPa) during heating, this decrease (2.6% for every 100 C) being significantly higher than for ATZ. The rela￾tive low value determined at room temperature (as com￾pared to that obtained with the rule of mixtures considering 400 GPa for alumina [34] and 244 GPa for monophase monoclinic zirconia [35]) may be related to the presence of microcracks caused by the high content of monoclinic zirconia in these compacts, as discussed above. From about 860 C, values started to increase, first smoothly and then sharply from 950 C up to 1150 C. The initial increase might be associated with stiffening of the material caused by closure of the microcracks whereas the sharp increase has to be attributed to the transforma￾tion of zirconia from the monoclinic (E  164 GPa at 950 C) to the tetragonal phase (E  220 GPa at 1200 C) [35]. The effect of the inverse transformation was observed on cooling where, after a slight increase, an abrupt decrease occurred at the temperature of the tetragonal to monoclinic zirconia phase transformation (725 C). After this sharp decrease, the Young’s modulus continued rising as the tem￾perature fell to room temperature, where a similar value to the initial room temperature was achieved. The dilatometer curves during heating and cooling of the sintered ATZ and AMZ monoliths are shown in Fig. 2. For ATZ, both curves were monotonous and with￾out any significant hysteresis (the small differences between heating and cooling can be attributed to experimental uncertainties). This behaviour, together with the observed variation of Young’s modulus proves that no significant microstructural changes occurred during the thermal cycling of the sintered ATZ compacts. Conversely, the vol￾ume changes associated with the reversible zirconia phase transformation, from monoclinic to tetragonal during heating above 1150 C and below 725 C on cooling, con￾ditioned the dimensional changes of the AMZ monoliths. The expected differences in thermal strain of the layers, referred to as De, which will condition the residual stress state in the layered structure, can be drawn from the dimensional changes experienced by the monoliths (Fig. 2). At high temperature, where mass transport mech￾anisms are active, the strain mismatch between layers caused by differences in thermal expansion would be accommodated. Therefore, the strain mismatch between layers at room temperature can be derived by taking 1200 C as zero point for both materials and considering the dimensional changes from 1200 C to room tempera￾ture. In doing so, the thermal strain mismatch between the layers of the sintered laminates at 25 C is De = eAMZ eATZ  0.00212 [27], where eAMZ and eATZ refer to the thermal strains of the AMZ and ATZ mono￾liths from Fig. 2. As discussed in the introduction, the mechanical response of the layered material at a given temperature will be determined by the residual stress field within the layers which, in turn, would be associated with two features: (i) the elastic properties at the given temperature, i.e. E(T), and (ii) the thermal strain mismatch between layers, namely De(T). Hence, an evaluation of the residual stress state at every testing temperature was carried out. In doing so, a 3D finite element analysis was carried out for the residual stress calculation over the temperature range. The corresponding elastic and thermal properties of each layer, i.e. Young’s modulus and thermal strain, were intro￾duced as a function of temperature. Fig. 1. Young’s modulus variation with temperature of the ATZ and AMZ monoliths. The abrupt change in slope on the heating and cooling curves of the AMZ is due to the m ! t and t ! m zirconia phase transformations, respectively. Fig. 2. Dilatometry curves for the ATZ and AMZ monoliths. Circles correspond to the different temperatures (indicated by the squares) where flexural tests were carried out. 4894 R. Bermejo et al. / Acta Materialia 55 (2007) 4891–4901

R Bermejo et al. Acta Materialia 55(2007)4891-4901 4895 Table I Maximum residual stress values in the ATZ and AMZ layers of the laminate at the testing temperatures indicated by the circles in Fig. 2 Residual stress(MPa) 800°C↑ heating 800°C↓( cooling) 650°C↓( cooling Gres.max in ATZ layers +105 +70 +154 A minus sign indicates that the layer is under compression. In order to compare the residual stress distribution in the AMz and ATZ layers at the temperatures of study men,across the layers. The results showed a biaxial stress =150 e inate rT (see full squares in Fig. 2), the maximum stress developed in the laminate was determined at the centre of the speci L.650°C L.800°C distribution within the aTz and AMz layers parallel to the layer plane, which was constant far from the edges 9100 Results are listed in table l It is inferred from the referred table that there is a strong influence, not only of the temperature, but also of the pre-50 vious thermal cycle, on the level of residual stresses within Indent. pop-t the layers. In this regard, if the temperature of 800C is reached on heating from room temperature, the internal 0.000.030.060.090.120.150.18 AMZ layers would be in compression and the atZ ones in tension as it occurs in the " as sintered" laminate. In Displacement [ mm] contrast, cooling from a temperature over 1150C results Fig. 3. Stress-displacement curves corresponding to the indentation- in ATZ layers under compression and AMZ layers under strength tests in the laminates and in the reference monolith at several nsion. For specimens tested oling from 1200 oC temperatures. The"pop-in" events indicate the initial growth of the dentation cracks efore reaching the temperature for the t-m phase trans the ATz/AMZ interface. Curves are shifted ong the displacement axis for clearer observation formation(a725C), residual stresses would be solely due to thermal expansion mismatch between the layers, being The first thin AMZ layer, with high compressive stresses significantly smaller than strain differences caused by the at room temperature (Table 1), acted as a barrier to crack phase transformation(Fig. 1). However, once transforma- propagation yielding a threshold strength, characteristic of tion occurs on further cooling, the previous residual stress this layered configuration. The effect of the tensile residual state turns into tensile stress in the ATZ layers and into stresses in the outer ATZ layers was reflected by the fact compressive stress in the AMz ones (Table 1), the magni- that the first pop-in events, corresponding to the propaga tude of such stresses being mainly affected by the thermal tion of the indentations up to the first ATZ/AMZ interface, strain between layers coming from the zirconia phase trans- occurred at a stress levels significantly lower than those for ormation. As a consequence, the response of the layered the ATZ monolithic specimens pecimens under flexure will change as a function of the As temperature increases, a decrease of strength is thermal history, as it will be discussed below expected in the materials associated with the decrease in Youngs modulus(Fig. 1)and fracture energy. However, 3.3. Indentation -strength tests at different temperatures the stress levels of the first fracture events(280 MPa)in the layered specimens tested at 800C on heating were Fig 3 shows the stress-displacement response of the lam- higher than those corresponding to the specimens tested inate specimens at the temperatures of study. The corre- at room temperature(x45 MPa). Therefore, in the former ponding characteristic curve for Atz monolithic specimens, the detrimental effect of temperature on specimens at room temperature is also represented for com- strength was counterbalanced by the decrease in tensile parative purposes. The variations in Young's modulus with stresses in the outer layer(Table 1). Nevertheless, compres- temperature( Fig. I)are reflected in these curves by a change sive residual stresses were still active in the Az layers and in the slope. Differences between the fracture behaviour at thus, the effectiveness of the layered configuration in arrest room temperature of the indented monolithic and layered ing the crack propagation was not lost, as evidenced by the pecimens are also apparent. The former presented linear "pop-in"events in the stress-displacement curves(Fig 3) stress-displacement behaviour up to fracture, which In fact, post-mortem examinations of the specimens occurred at a maximum stress value of 145+10 MPa, and showed crack bifurcation at the first AMz layer, as the failure was catastrophic as it corresponds to brittle mate- occurred in the specimens tested at room temperature rials. In contrast, laminate failures were preceded by three [24]. Moreover, additional"pop-in"events were observed pop-in"events associated with the initial growth of the in the 800C T test, corresponding to energy dissipation three indentation cracks up to the ATZ/ AMZ interface, as mechanisms occurring after the indentation cracks reached evidenced in a recent work [24]. the first ATZ/AMZ interface as it is discussed below

In order to compare the residual stress distribution in the AMZ and ATZ layers at the temperatures of study (see full squares in Fig. 2), the maximum stress developed in the laminate was determined at the centre of the speci￾men, across the layers. The results showed a biaxial stress distribution within the ATZ and AMZ layers parallel to the layer plane, which was constant far from the edges. Results are listed in Table 1. It is inferred from the referred table that there is a strong influence, not only of the temperature, but also of the pre￾vious thermal cycle, on the level of residual stresses within the layers. In this regard, if the temperature of 800 C is reached on heating from room temperature, the internal AMZ layers would be in compression and the ATZ ones in tension, as it occurs in the ‘‘as sintered’’ laminate. In contrast, cooling from a temperature over 1150 C results in ATZ layers under compression and AMZ layers under tension. For specimens tested on cooling from 1200 C before reaching the temperature for the t ! m phase trans￾formation (725 C), residual stresses would be solely due to thermal expansion mismatch between the layers, being significantly smaller than strain differences caused by the phase transformation (Fig. 1). However, once transforma￾tion occurs on further cooling, the previous residual stress state turns into tensile stress in the ATZ layers and into compressive stress in the AMZ ones (Table 1), the magni￾tude of such stresses being mainly affected by the thermal strain between layers coming from the zirconia phase trans￾formation. As a consequence, the response of the layered specimens under flexure will change as a function of the thermal history, as it will be discussed below. 3.3. Indentation – strength tests at different temperatures Fig. 3 shows the stress–displacement response of the lam￾inate specimens at the temperatures of study. The corre￾sponding characteristic curve for ATZ monolithic specimens at room temperature is also represented for com￾parative purposes. The variations in Young’s modulus with temperature (Fig. 1) are reflected in these curves by a change in the slope. Differences between the fracture behaviour at room temperature of the indented monolithic and layered specimens are also apparent. The former presented linear stress–displacement behaviour up to fracture, which occurred at a maximum stress value of 145 ± 10 MPa, and the failure was catastrophic as it corresponds to brittle mate￾rials. In contrast, laminate failures were preceded by three ‘‘pop-in’’ events associated with the initial growth of the three indentation cracks up to the ATZ/AMZ interface, as evidenced in a recent work [24]. The first thin AMZ layer, with high compressive stresses at room temperature (Table 1), acted as a barrier to crack propagation yielding a threshold strength, characteristic of this layered configuration. The effect of the tensile residual stresses in the outer ATZ layers was reflected by the fact that the first pop-in events, corresponding to the propaga￾tion of the indentations up to the first ATZ/AMZ interface, occurred at a stress levels significantly lower than those for the ATZ monolithic specimens. As temperature increases, a decrease of strength is expected in the materials associated with the decrease in Young’s modulus (Fig. 1) and fracture energy. However, the stress levels of the first fracture events (80 MPa) in the layered specimens tested at 800 C on heating were higher than those corresponding to the specimens tested at room temperature (45 MPa). Therefore, in the former specimens, the detrimental effect of temperature on strength was counterbalanced by the decrease in tensile stresses in the outer layer (Table 1). Nevertheless, compres￾sive residual stresses were still active in the AMZ layers and thus, the effectiveness of the layered configuration in arrest￾ing the crack propagation was not lost, as evidenced by the ‘‘pop-in’’ events in the stress–displacement curves (Fig. 3). In fact, post-mortem examinations of the specimens showed crack bifurcation at the first AMZ layer, as occurred in the specimens tested at room temperature [24]. Moreover, additional ‘‘pop-in’’ events were observed in the 800 C › test, corresponding to energy dissipation mechanisms occurring after the indentation cracks reached the first ATZ/AMZ interface, as it is discussed below. Table 1 Maximum residual stress values in the ATZ and AMZ layers of the laminate at the testing temperatures indicated by the circles in Fig. 2 Residual stress (MPa) 20 C 800 C › (heating) 1200 C 800 C fl (cooling) 650 C fl (cooling) rres,max. in ATZ layers +105 +40 – 26 +70 rres,max. in AMZ layers 695 245 – +154 415 A minus sign indicates that the layer is under compression. Fig. 3. Stress–displacement curves corresponding to the indentation– strength tests in the laminates and in the reference monolith at several temperatures. The ‘‘pop-in’’ events indicate the initial growth of the indentation cracks up to the ATZ/AMZ interface. Curves are shifted along the displacement axis for clearer observation. R. Bermejo et al. / Acta Materialia 55 (2007) 4891–4901 4895

4896 R Bermejo et al. Acta Materialia 55(2007)4891-4901 The specimens tested at 1200C and at 800C where Kic is the intrinsic fracture toughness of each individ- (reached on cooling) showed linear stress-displacement ual layer, x is the distance along the crack length measured curves and catastrophic failure( Fig 3), similar to the brit- from the surface, a is the crack length and h(a, x) is the mens failed at stress levels higher than those of the first assumed geomety evelope tle fracture of the aTz monoliths. moreover these weight function, developed by Fett and Munz [36] for the pop-in events in specimens tested at lower temperatures In the present investigation, the same approach has been These facts reflect the low level of residual stresses in the employed to calculate the expected apparent fracture laminate once the large strain mismatch due to zirconia toughness of the laminate at different temperatures. The transformation is no longer present ( Table 1). The absence Kapt.c is now dependent on the residual stress state, ores of compressive stresses in the AMZ internal layers caused given for each temperature and previous thermal history the cracks not being arrested nor bifurcated. Moreover, Integrating the weight function with the corresponding tensile stresses within the external layers did not add to residual stress distribution, the apparent fracture toughne he external stress to decrease the strength. In fact, lami- was determined as a function of the crack length parameter nates tested at 800C during cooling showed higher 8, defined as a/w, where w is the total thickness. We cau- strength (135 MPa) than those tested during heating tion the reader that the weight function here employed (≈120MPa) applies to a homogeneous material(with constant elastic For specimens tested at 650C on cooling, the develop- properties). In our case, this approximation may lead to ment of residual stresses after transformation(Table 1)led an overestimation of the calculated apparent fracture to the characteristic fracture of the laminate, showing toughness in the ATZ layers, owing to the lower stiffness op-in"events previous to complete failure associated with of the adjacent AMZ ones [37, 38]. In this regard, an alter- the crack arrest provided by the AMz compressive layers native procedure has been proposed elsewhere [39], to pre Following the above discussion, it can be inferred that dict in a more accurate way the fracture toughness of for testing temperatures where the residual stress distribu- multilayered ceramics with various elastic properties. Nev- tion yielded a compressive stress state in the internal ertheless, this level of accuracy is not within the scope of AMZ layers of the laminate, a barrier to crack propagation this work. Fig. 4 represents the calculated Kapt.c curve cor was successfully achieved under flexural testing. Hence, the responding to the layered material at the temperatures of so-called"threshold strength"concept found at room tem- study. The intrinsic fracture toughness of each layer(KId) perature can be extended to applications at temperatures has been determined at room temperature by the single monoclinic form. This fact has to be taken into account monoliths(2.6+0.1 MPa m/ 2 and 3.2+0.1 MPa m /2for hen materials are expected to work at rather low temper- the AMz and ATZ layers, respectively ), and has been atures(e.g. 800C) but after being cycled from tempera- assumed to be constant through the temperature range of tures higher than the transformation one. The effectiveness study of the compressive layers in hindering the crack propaga- It can be observed that Kapt.c decreases in the layers tion determines the value of the threshold strength which which have a tensile residual stress state and increases rap would be associated with the apparent fracture toughness idly within the layers with compressive residual stresses, as of the laminate at the testing temperature the crack length increases. Analysing the different testing conditions used in Fig. 4, the most pronounced R-curve 3. 4. Apparent fracture toughness with temperature haviour would be found at room temperature, where The fracture behaviour of the studied laminated struc- ture at room temperature, in which a steep R-curve behav- OC(RT) lour was observed, was described in a previous work [24].It was shown how the residual stress state determines the 1200°C apparent fracture toughness of the laminate at a given posi- tion. In such a study, a fracture mechanics weight function analysis was effectively used to estimate the crack growth resistance(R-curve behaviour )as a function of the position within a multilayered system, for a given residual stress dis- tribution, Ores(x). The geometry assumed for the crack was that of an edge crack (through-the-thickness crack ) com- monly employed in the evaluation of R-curve behaviour for multilayered systems. The so-called apparent fracture 0.40.5 toughness, Kapt. c, may be defined as folle Kapt.c=Klc- h(a, x) res(x). dr (2) Fig. 4. Apparent fracture toughness, Kapt c, calculated analytically using the weight function approach, considering the residual stresses in the laminate at the different temperatures

The specimens tested at 1200 C and at 800 C fl (reached on cooling) showed linear stress–displacement curves and catastrophic failure (Fig. 3), similar to the brit￾tle fracture of the ATZ monoliths. Moreover, these speci￾mens failed at stress levels higher than those of the first pop-in events in specimens tested at lower temperatures. These facts reflect the low level of residual stresses in the laminate once the large strain mismatch due to zirconia transformation is no longer present (Table 1). The absence of compressive stresses in the AMZ internal layers caused the cracks not being arrested nor bifurcated. Moreover, tensile stresses within the external layers did not add to the external stress to decrease the strength. In fact, lami￾nates tested at 800 C during cooling showed higher strength (135 MPa) than those tested during heating (120 MPa). For specimens tested at 650 C on cooling, the develop￾ment of residual stresses after transformation (Table 1) led to the characteristic fracture of the laminate, showing ‘‘pop-in’’ events previous to complete failure associated with the crack arrest provided by the AMZ compressive layers. Following the above discussion, it can be inferred that for testing temperatures where the residual stress distribu￾tion yielded a compressive stress state in the internal AMZ layers of the laminate, a barrier to crack propagation was successfully achieved under flexural testing. Hence, the so-called ‘‘threshold strength’’ concept found at room tem￾perature can be extended to applications at temperatures where the zirconia in the internal layers remains in the monoclinic form. This fact has to be taken into account when materials are expected to work at rather low temper￾atures (e.g. 800 C) but after being cycled from tempera￾tures higher than the transformation one. The effectiveness of the compressive layers in hindering the crack propaga￾tion determines the value of the threshold strength which would be associated with the apparent fracture toughness of the laminate at the testing temperature. 3.4. Apparent fracture toughness with temperature The fracture behaviour of the studied laminated struc￾ture at room temperature, in which a steep R-curve behav￾iour was observed, was described in a previous work [24]. It was shown how the residual stress state determines the apparent fracture toughness of the laminate at a given posi￾tion. In such a study, a fracture mechanics weight function analysis was effectively used to estimate the crack growth resistance (R-curve behaviour) as a function of the position within a multilayered system, for a given residual stress dis￾tribution, rres (x). The geometry assumed for the crack was that of an edge crack (through-the-thickness crack), com￾monly employed in the evaluation of R-curve behaviour for multilayered systems. The so-called apparent fracture toughness, Kapt,c, may be defined as follows: Kapt;c ¼ KIc Z a 0 hða; xÞ rresðxÞ dx ð2Þ where KIc is the intrinsic fracture toughness of each individ￾ual layer, x is the distance along the crack length measured from the surface, a is the crack length and h(a,x) is the weight function, developed by Fett and Munz [36] for the assumed geometry. In the present investigation, the same approach has been employed to calculate the expected apparent fracture toughness of the laminate at different temperatures. The Kapt,c is now dependent on the residual stress state, rres, given for each temperature and previous thermal history. Integrating the weight function with the corresponding residual stress distribution, the apparent fracture toughness was determined as a function of the crack length parameter d, defined as a/W, where W is the total thickness. We cau￾tion the reader that the weight function here employed applies to a homogeneous material (with constant elastic properties). In our case, this approximation may lead to an overestimation of the calculated apparent fracture toughness in the ATZ layers, owing to the lower stiffness of the adjacent AMZ ones [37,38]. In this regard, an alter￾native procedure has been proposed elsewhere [39], to pre￾dict in a more accurate way the fracture toughness of multilayered ceramics with various elastic properties. Nev￾ertheless, this level of accuracy is not within the scope of this work. Fig. 4 represents the calculated Kapt,c curve cor￾responding to the layered material at the temperatures of study. The intrinsic fracture toughness of each layer (KIc) has been determined at room temperature by the single￾edge V-notch beam (SEVNB) method in the corresponding monoliths (2.6 ± 0.1 MPa m1/2 and 3.2 ± 0.1 MPa m1/2 for the AMZ and ATZ layers, respectively), and has been assumed to be constant through the temperature range of study. It can be observed that Kapt,c decreases in the layers which have a tensile residual stress state and increases rap￾idly within the layers with compressive residual stresses, as the crack length increases. Analysing the different testing conditions used in Fig. 4, the most pronounced R-curve behaviour would be found at room temperature, where Fig. 4. Apparent fracture toughness, Kapt,c, calculated analytically using the weight function approach, considering the residual stresses in the laminate at the different temperatures. 4896 R. Bermejo et al. / Acta Materialia 55 (2007) 4891–4901

R Bermejo et al./ Acta Materialia 55(2007)4891-4901 4897 the first AMZ layer would reach an apparent fractur Table 2 toughness of 8.6 MPa m. For the other two cases where Critical stress inter factor calculated experimentally for several high compressive stresses are located in the internal AMz temperatures in the laminates layers, i.e. 800C on heating and 650C on cooling, R- Temperature or ac(um) Ke(MPa m") kapp curve behaviour would also be observed through the (MPa) MPam AMZ layers, yielding maximum apparent fracture tough ness values at the first AMZ layer lower than those reached 8000c↑121±450±549±0 5.1 650°C, respectively). Since residual stresses are negligible650°↓156±5530±562±0.1 at 1200C, the apparent toughness of the laminate at this The maximum apparent fracture toughness values predicted analytically temperature would correspond to the intrinsic fracture through the weight function analysis are also presented for comparison toughness of the atz and aMz layers, with no contribu- tion of the second term of Eq .(2)to Kapt. c. The inverse sense of the residual stresses developed in the laminate flexure method [40], using a geometrical factor Y corre- tested at 800C during cooling leads to crack growth resis- sponding to the solution of a semi- radial crack, i.e tance through the first ATZ layer, owing to the contribu- Y=2/VI tion of the compressive stresses in this layer. As these Following the above considerations, the critical stress stresses are relatively low compared to those developed in intensity factors were calculated experimentally for the dif- the thin AMz layers (Table 1), a maximum value of only ferent temperatures, assuming mode I of fracture, and are 4.5 MPa m would be reached(Fig. 4) listed in Table 2. The maximum apparent fracture tough Overall, the model used predicts a higher apparent frac- ness values predicted analytically through the weight func- ture toughness for the layered composite through the tem- tion analysis are also presented for comparison perature range of study than that corresponding to th The values reported in Table 2 show good agreement reference ATZ monolith. For the cases where the compres- between the apparent fracture toughness predicted with sive stresses are located in the internal AMZ layers, the the analytical approach and the critical stress intensity fac- weight function analysis yields different threshold strength tor evaluated in the laminates for specimens tested at values depending on the thermal history, the room temper- 800C on heating and 800C and 650C on cooling, ature conditions being the most suitable for a high thresh- whereas a relative discrepancy is found at room tempera old strength level, as was observed experimentally in the ture. In this regard, a further inspection of the fracture sur- flexural tests( Fig 3) faces revealed a different crack propagation mode in the In order to correlate the analytical approach with the laminates for every testing temperature(owing to deflec experimental results for the laminated specimens tested at tion/bifurcation mechanisms and in some cases delamina different temperatures, a linear elastic fracture mechanics tion processes occurring at fracture). These energy analysis was implemented. An apparent fracture toughness dissipating mechanisms are associated with the overall value was assessed at the different temperatures using the crack resistance of the laminate. Thus, a more detailed experimental values of failure stress, af, and assuming the analysis of the crack propagation was conducted at every critical flaw size, ac, as (i) the thickness of the outer Atz temperature, considering not only the corresponding resid- yer for the cases were the initial cracks were arrested at ual stress state but also the change in the elastic properties the first ATZ/AMZ interface, i.e. after the tetragonal to of each layer which might have influenced the crack prop- monoclinic phase transformation has occurred, and (ii) agation mode [42] the size of the indentation crack measured at the surface of the post-mortem specimens(analogue to the indentation in flexure method [40], where the survival cracks yield the 3.5. Crack propagation features at different temperatures critical flaw size at the moment of fracture)for the case f specimens tested before the t-m phase transformation In a previous work [21]it was shown that, for the partic- has taken place, i.e. 800C during cooling. A value of the ular conditions of compressive stresses and layer thickness critical stress intensity factor Kc for the first cases may be of the laminated structure studied here, tested at room tem obtained using the following equation perature, the crack bifurcated as it entered the thin com- pressive layer of the laminate (Fig. 5a). Further Ke=Y(sorv (3) propagation through the multilayered structure implied with Y(o) given in Ref. [41] first the propagation of the bifurcated crack throug central part of the thin layer in compression, which was (=19=0-0215-392+26 ssociated with energy dissipation during the fracture pro- (1+26)(1 (4)cess yielding an increase in the apparent fracture toughness of the material. In contrast, for the flexural tests performe

the first AMZ layer would reach an apparent fracture toughness of 8.6 MPa m1/2. For the other two cases where high compressive stresses are located in the internal AMZ layers, i.e. 800 C on heating and 650 C on cooling, R￾curve behaviour would also be observed through the AMZ layers, yielding maximum apparent fracture tough￾ness values at the first AMZ layer lower than those reached at room temperature (5.1 and 6.5 MPa m1/2, for 800 and 650 C, respectively). Since residual stresses are negligible at 1200 C, the apparent toughness of the laminate at this temperature would correspond to the intrinsic fracture toughness of the ATZ and AMZ layers, with no contribu￾tion of the second term of Eq. (2) to Kapt,c. The inverse sense of the residual stresses developed in the laminate tested at 800 C during cooling leads to crack growth resis￾tance through the first ATZ layer, owing to the contribu￾tion of the compressive stresses in this layer. As these stresses are relatively low compared to those developed in the thin AMZ layers (Table 1), a maximum value of only 4.5 MPa m1/2 would be reached (Fig. 4). Overall, the model used predicts a higher apparent frac￾ture toughness for the layered composite through the tem￾perature range of study than that corresponding to the reference ATZ monolith. For the cases where the compres￾sive stresses are located in the internal AMZ layers, the weight function analysis yields different threshold strength values depending on the thermal history, the room temper￾ature conditions being the most suitable for a high thresh￾old strength level, as was observed experimentally in the flexural tests (Fig. 3). In order to correlate the analytical approach with the experimental results for the laminated specimens tested at different temperatures, a linear elastic fracture mechanics analysis was implemented. An apparent fracture toughness value was assessed at the different temperatures using the experimental values of failure stress, rf, and assuming the critical flaw size, ac, as (i) the thickness of the outer ATZ layer for the cases were the initial cracks were arrested at the first ATZ/AMZ interface, i.e. after the tetragonal to monoclinic phase transformation has occurred, and (ii) the size of the indentation crack measured at the surface of the post-mortem specimens (analogue to the indentation in flexure method [40], where the survival cracks yield the critical flaw size at the moment of fracture) for the case of specimens tested before the t ! m phase transformation has taken place, i.e. 800 C during cooling. A value of the critical stress intensity factor Kc for the first cases may be obtained using the following equation: Kc ¼ Y ðdÞrf ffiffiffiffi ac p ð3Þ with Y(d) given in Ref. [41]: Y ðdÞ ¼ 1:99 dð1 dÞð2:15 3:93d þ 2:7d2 Þ ð1 þ 2dÞð1 dÞ 3=2 " # ð4Þ For the second case, the critical stress intensity factor Kc was determined using Eq. (3) based on the indentation in flexure method [40], using a geometrical factor Y corre￾sponding to the solution of a semi-radial crack, i.e. Y ¼ 2= ffiffiffi p p . Following the above considerations, the critical stress intensity factors were calculated experimentally for the dif￾ferent temperatures, assuming mode I of fracture, and are listed in Table 2. The maximum apparent fracture tough￾ness values predicted analytically through the weight func￾tion analysis are also presented for comparison. The values reported in Table 2 show good agreement between the apparent fracture toughness predicted with the analytical approach and the critical stress intensity fac￾tor evaluated in the laminates for specimens tested at 800 C on heating and 800 C and 650 C on cooling, whereas a relative discrepancy is found at room tempera￾ture. In this regard, a further inspection of the fracture sur￾faces revealed a different crack propagation mode in the laminates for every testing temperature (owing to deflec￾tion/bifurcation mechanisms and in some cases delamina￾tion processes occurring at fracture). These energy dissipating mechanisms are associated with the overall crack resistance of the laminate. Thus, a more detailed analysis of the crack propagation was conducted at every temperature, considering not only the corresponding resid￾ual stress state but also the change in the elastic properties of each layer which might have influenced the crack prop￾agation mode [42]. 3.5. Crack propagation features at different temperatures In a previous work [21] it was shown that, for the partic￾ular conditions of compressive stresses and layer thickness of the laminated structure studied here, tested at room tem￾perature, the crack bifurcated as it entered the thin com￾pressive layer of the laminate (Fig. 5a). Further propagation through the multilayered structure implied first the propagation of the bifurcated crack through the central part of the thin layer in compression, which was associated with energy dissipation during the fracture pro￾cess yielding an increase in the apparent fracture toughness of the material. In contrast, for the flexural tests performed at 1200 C, the crack path was found to be straight as for the case of brittle ceramics, owing to the absence of resid￾Table 2 Critical stress intensity factor calculated experimentally for several temperatures in the laminates Temperature rf (MPa) ac (lm) Kc (MPa m1/2) (experimental) Kappt,c (MPa m1/2) (analytic) 20 C 167 ± 5 530 ± 5 6.7 ± 0.1 8.6 800 C › 121 ± 4 530 ± 5 4.9 ± 0.1 5.1 800 C fl 132 ± 10 460 ± 20 4.2 ± 0.2 4.5 650 C fl 156 ± 5 530 ± 5 6.2 ± 0.1 6.5 The maximum apparent fracture toughness values predicted analytically through the weight function analysis are also presented for comparison. R. Bermejo et al. / Acta Materialia 55 (2007) 4891–4901 4897

4898 R Bermejo et al. Acta Materialia 55(2007)4891-4901 a b c d Fig. 5. Characteristic crack patterns in specimens with different thermal histories. Optical micrographs showing:(a) crack bifurcation in the thin essive layers of laminates tested at room temperature, (b) crack propagating straight through the specimens with negligible residual stresses (1200C), (c)small crack deflection in laminates tested at 650C during cooling and (d)crack bifurcation and interface delamination in laminates tested at ual compressive stresses inside the test bars(Fig. 5b). The which presented internal compressive stresses (Table 1) same straight crack pattern was also observed for the spec- Nevertheless, the branches of the bifurcated crack did not imens with compressive residual stresses in the external propagate along the centre of the AMZ layers, as in the layer, such as those tested at 800C on cooling. In such specimens tested at room temperature(Fig 5a), but tra cases, the shielding provided by the compressive stresses versed them and reached the next AMZ layer impinging acting at the crack tip yielded a relative high apparent with a certain angle(Fig. 5c and d and Fig. 6). As stated toughness value(Table 2)in comparison to the toughness in previous works [21, 43-46), crack propagation along of ATZ monolithic specimens. However, when the failure the central part of the layers in compression may be asso- stress was reached at the first compressive layer, a cata- ciated with the presence of edge cracks in the materials strophic fracture was observed. As described above, crack These cracks will be present in the"as fabricated"speci bifurcation at the internal ATZ layers occurred in speci- mens but, most probably, thermal expansion of the speci mens tested at 650C on cooling and at 800C on heating, mens during heating might lead to their closure. Fig. 6. SEM micrograph showing interface delamination at the AMZ/ATZ interfaces after bifurcation occurred in laminated specimens tested at 800C on heating. The bifurcated crack impinges on the interface at an angle of 25

ual compressive stresses inside the test bars (Fig. 5b). The same straight crack pattern was also observed for the spec￾imens with compressive residual stresses in the external layer, such as those tested at 800 C on cooling. In such cases, the shielding provided by the compressive stresses acting at the crack tip yielded a relative high apparent toughness value (Table 2) in comparison to the toughness of ATZ monolithic specimens. However, when the failure stress was reached at the first compressive layer, a cata￾strophic fracture was observed. As described above, crack bifurcation at the internal ATZ layers occurred in speci￾mens tested at 650 C on cooling and at 800 C on heating, which presented internal compressive stresses (Table 1). Nevertheless, the branches of the bifurcated crack did not propagate along the centre of the AMZ layers, as in the specimens tested at room temperature (Fig. 5a), but tra￾versed them and reached the next AMZ layer impinging with a certain angle (Fig. 5c and d and Fig. 6). As stated in previous works [21,43–46], crack propagation along the central part of the layers in compression may be asso￾ciated with the presence of edge cracks in the materials. These cracks will be present in the ‘‘as fabricated’’ speci￾mens but, most probably, thermal expansion of the speci￾mens during heating might lead to their closure. Fig. 5. Characteristic crack patterns in specimens with different thermal histories. Optical micrographs showing: (a) crack bifurcation in the thin compressive layers of laminates tested at room temperature, (b) crack propagating straight through the specimens with negligible residual stresses (1200 C), (c) small crack deflection in laminates tested at 650 C during cooling and (d) crack bifurcation and interface delamination in laminates tested at 800 C on heating. Fig. 6. SEM micrograph showing interface delamination at the AMZ/ATZ interfaces after bifurcation occurred in laminated specimens tested at 800 C on heating. The bifurcated crack impinges on the interface at an angle of 25. 4898 R. Bermejo et al. / Acta Materialia 55 (2007) 4891–4901

R Bermejo et al. Acta Materi 2007)4891-4901 4899 As inferred from Table 2, the experimental values of to the applied stress field. In layered ceramics, the nor apparent fracture toughness of the specimens tested at parameter(related to the normal stresses at the interface 650C on cooling were in good agreement with the analyt- is usually zero, and the incidence of interface debonding ical approximation; this is associated with the relatively is dominated by tag, which accounts for the tensile or com- raight crack propagation with short propagation dis- pressive in-plane residual stresses in the layers and repre- tances across the energy consuming AMZ layers(Fig. 5c). sents the boundary region between crack deflection and In the specimens tested at 800C during heating, a crack penetration delamination process took place as the branched cracks For the case of thin layers with a high relative elastic reached the AMZ/ATZ interface(Figs. 5d and 6). This modulus(large ax)and a low relative thermal expansion fracture process was reflected in the stress-displacement coefficient that results in a negative ntag, interface debond curves by a"pop-in"event located at a stress level signifi- ing effects are favoured. On the other hand, when the elas- intly higher than that corresponding to the pop-in" tic mismatch is not so significant(low a) crack penetration events from the indentation cracks( Fig 3). This observa- is enhanced At room temperature, the E modulus values of tion is associated with the re-initiation of fracture in the the adjacent AtZ and AMz layers are 390 GPa and ATZ layer after delamination as observed in laminates with 290 GPa, respectively(Fig. 1). Assuming the same Pois- weak interfaces [2]. In order to rationalise the delamination son's ratio of 0. 22 for ATZ and AMZ compacts, the first process in the specimens tested under heating conditions, Dundurs' parameter, oz, gives a value of t0. 15. On the the crack deflection/penetration criterion proposed by He other hand, taking into account the elastic modulus of and Hutchinson [47]at the interface of dissimilar materials the layers at 800C, i.e. 360 GPa for ATZ and 220 GPa is recalled. The tendency of a crack meeting the interface for AMZ (Fig. 1), the parameter a results in A+0. 24 etween dissimilar materials at 90 either to deflect into The minus sign corresponds to the impinging crack from the interface or penetrate through it into the next layer material AtZ to AMZ and vice versa. The variable i can depends on the elastic properties of the materials and be evaluated as a function of o as given by He and Hutch- whether the ratio sa/lp(deflection energy/penetration inson [49], the corresponding values being listed in Table 3 energy) is either higher or lower than the ratio of the frac- Considering the 2 parameter and the corresponding resid ture energy of the interface and that of the adjoining layer, ual stresses(atag) at room temperature and at 800C, the Slaver. The variables of interest depend only on two non- tag parameters can be calculated using Eq. (6)for bot dimensional combinations of the material parameters, the cases. The length of the potential crack branch ap is consid so-called Dundurs' parameters, a and B [48]. The first ered to be the critical flaw size at the interface, which is and more important parameter can be easily interpreted assumed to be of the order of minimum microstructural when expressed as units,i.e.Al um, while the parameter KI represents the fracture toughness of the corresponding layer. Values at 5)room temperature for AMZ and ATZ layers are given above. At 800C, the Ki values were determined by the where E=Ey/(I-vi) are the plain strain tensile modulus, indentation flexure method [40] in ATZ monolithic speci- Ei the Youngs modulus and y the Poissons ratio of the mens tested at this temperature, giving approximately layers I and 2. In the presence of normal(ono)and/or tan- 3.0+0.1 MPa m /2. For the AMz, K i was assumed to be gential(tag) residual stresses, two additional non-dimen- that determined at room temperature. The ntag values are sional length parameters, i. e. "nor and tag, become also listed in Table 3 mportant and are defined as [49] The ntag curves are represented in Fig. 7 in a He-Evans- Hutchinson plot [49]. The ATZ and AMZ fracture energy, (6) i.e. SATz and CAMz, was calculated as 5=K/E' for each layer, considering the Youngs modulus and toughness of where a is the length of the crack branch either at the inter- the corresponding layers at the given temperature. In order face(ad)or in the next layer (ap), i is a stress singularity to evaluate the fracture energy of the ATZ/AMZ interface, exponent for the main crack and K, is a factor proportional indentation cracks at room temperature were analysed. No Table 3 Youngs moduli and Dundurs parameter corresponding to room temperature and 800C Material E(GPa) (20°C) (800°C (800°C) (20°C) (800°) ATZ AMZ ATZu)→AMz(2 0.15 24 0.527 0.545 -0.18 -0.05 AMz(→AT2(2) +0.24 0.478 +0.03 The exponent A is calculated by interpolating the tabulated values given in Ref.[49] and the ntag curves are determined using Eq.(6)

As inferred from Table 2, the experimental values of apparent fracture toughness of the specimens tested at 650 C on cooling were in good agreement with the analyt￾ical approximation; this is associated with the relatively straight crack propagation with short propagation dis￾tances across the energy consuming AMZ layers (Fig. 5c). In the specimens tested at 800 C during heating, a delamination process took place as the branched cracks reached the AMZ/ATZ interface (Figs. 5d and 6). This fracture process was reflected in the stress–displacement curves by a ‘‘pop-in’’ event located at a stress level signifi- cantly higher than that corresponding to the ‘‘pop-in’’ events from the indentation cracks (Fig. 3). This observa￾tion is associated with the re-initiation of fracture in the ATZ layer after delamination as observed in laminates with weak interfaces [2]. In order to rationalise the delamination process in the specimens tested under heating conditions, the crack deflection/penetration criterion proposed by He and Hutchinson [47] at the interface of dissimilar materials is recalled. The tendency of a crack meeting the interface between dissimilar materials at 90 either to deflect into the interface or penetrate through it into the next layer depends on the elastic properties of the materials and whether the ratio fd/fp (deflection energy/penetration energy) is either higher or lower than the ratio of the frac￾ture energy of the interface and that of the adjoining layer, fi/flayer. The variables of interest depend only on two non￾dimensional combinations of the material parameters, the so-called Dundurs’ parameters, a and b [48]. The first and more important parameter can be easily interpreted when expressed as a ¼ E0 1 E0 2 E0 1 þ E0 2 ð5Þ where E0 i ¼ Ei=ð1 m2 i Þ are the plain strain tensile modulus, Ei the Young’s modulus and m the Poisson’s ratio of the layers 1 and 2. In the presence of normal (rnor) and/or tan￾gential (rtag) residual stresses, two additional non-dimen￾sional length parameters, i.e. gnor and gtag, become important and are defined as [49]: gnor ¼ rnor ak d KI ; gtag ¼ rtag ak p KI ð6Þ where a is the length of the crack branch either at the inter￾face (ad) or in the next layer (ap), k is a stress singularity exponent for the main crack and KI is a factor proportional to the applied stress field. In layered ceramics, the gnor parameter (related to the normal stresses at the interface) is usually zero, and the incidence of interface debonding is dominated by gtag, which accounts for the tensile or com￾pressive in-plane residual stresses in the layers and repre￾sents the boundary region between crack deflection and crack penetration. For the case of thin layers with a high relative elastic modulus (large a) and a low relative thermal expansion coefficient that results in a negative gtag, interface debond￾ing effects are favoured. On the other hand, when the elas￾tic mismatch is not so significant (low a) crack penetration is enhanced. At room temperature, the E modulus values of the adjacent ATZ and AMZ layers are 390 GPa and 290 GPa, respectively (Fig. 1). Assuming the same Pois￾son’s ratio of 0.22 for ATZ and AMZ compacts, the first Dundurs’ parameter, a, gives a value of ±0.15. On the other hand, taking into account the elastic modulus of the layers at 800 C, i.e. 360 GPa for ATZ and 220 GPa for AMZ (Fig. 1), the parameter a results in ±0.24. The minus sign corresponds to the impinging crack from material ATZ to AMZ and vice versa. The variable k can be evaluated as a function of a as given by He and Hutch￾inson [49], the corresponding values being listed in Table 3. Considering the k parameter and the corresponding resid￾ual stresses (rtag) at room temperature and at 800 C, the gtag parameters can be calculated using Eq. (6) for both cases. The length of the potential crack branch ap is consid￾ered to be the critical flaw size at the interface, which is assumed to be of the order of minimum microstructural units, i.e. 1 lm, while the parameter KI represents the fracture toughness of the corresponding layer. Values at room temperature for AMZ and ATZ layers are given above. At 800 C, the KI values were determined by the indentation flexure method [40] in ATZ monolithic speci￾mens tested at this temperature, giving approximately 3.0 ± 0.1 MPa m1/2. For the AMZ, KI was assumed to be that determined at room temperature. The gtag values are also listed in Table 3. The gtag curves are represented in Fig. 7 in a He–Evans– Hutchinson plot [49]. The ATZ and AMZ fracture energy, i.e. fATZ and fAMZ, was calculated as f = K2 /E0 for each layer, considering the Young’s modulus and toughness of the corresponding layers at the given temperature. In order to evaluate the fracture energy of the ATZ/AMZ interface, indentation cracks at room temperature were analysed. No Table 3 Young’s moduli and Dundurs’ parameter a corresponding to room temperature and 800 C Material E (GPa) akgtag (20 C) (800 C) (20 C) (800 C) (20 C) (800 C) (20 C) (800 C) ATZ 390 360 AMZ 290 220 ATZ(1) ! AMZ(2) 0.15 0.24 0.527 0.545 0.18 0.05 AMZ(1) ! ATZ(2) +0.15 +0.24 0.478 0.467 +0.03 +0.03 The exponent k is calculated by interpolating the tabulated values given in Ref. [49], and the gtag curves are determined using Eq. (6). R. Bermejo et al. / Acta Materialia 55 (2007) 4891–4901 4899

R Bermejo et al. Acta Materialia 55(2007)4891-4901 A possible explanation for this debonding at 800C lies penetration in the assumption made when estimating the tag curves for 18 ATzm+=0.03 crack propagates in a straight line, i.e. 90, within a given layer. This was the case for the crack propagating from layer ATZ to layer AMZ(Fig. 6). However, experimental 800° observations showed that when the impinging crack pene nt=905 trated into the AMZ compressive layer bifurcation 0.5 occurred and thus the propagating crack faced the new AMZ/ ATZ interface with angles smaller than90°(≈25° deflection in this case), as can be seen in Fig. 6. As demonstrated by He and Hutchinson [47]. the tendency for a crack to delaminate increases for small angles. Therefore, Fig. 7 -0.5 00 cannot describe this particular case of interface debonding a=(E2E1)(E2+E1) from the AMz to the Atz layer, and thus an"upwards Fig. 7. Comparison of the material parameters with the crack deflec. correction of the ntag=+0.03 curve should be recalled, tion/penetration He-Hutchinson criteria at room temperature and although it is out of the scope of the present investigation. 800C, assuming straight crack propagation, ie. 90, of the impinging It should be highlighted that the delamination process taking place in the AMZ/ATZ interfaces would the fracture energy of the system, changing the propagating crack from Mode I to Mode II of fracture. In this regard, preferential crack growth was observed along the ATz/ the Mode II crack propagation should be considered to AMZ interfaces, which indicated that, in spite of the shear make a significant contribution to the fracture energy of stresses present at the interfaces between layers caused by the system, similar to the enhancement provided by crack the different stress states, the interfaces between layers were trong. In this regard, strong interfaces have been reported bifurcation along the centre of the AMZ compressive layer by other authors for alumina/YTZP layered ceramics. under room temperature conditions[21]. The strength reliability of the multilayered system stud which maintained their structural integrity after high-tem- ied here together with the capability of fracture energy dis- perature compression testing [12, 13, 15, 16] In this case, as sipation through bifurcation and/or delamination no preferential crack growth was discerned, it can be spec- mechanisms emphasise the suitability of this layered archi ulated that the minimum value of the interface fracture tecture as a substitute for alumina-based monolithic com- energy is of the same order as that of the most brittle layer, ponents for high-temperature applications, always being i.e. KI(AMZ. In Table 4, the corresponding fracture ware of the temperature hysteretic conditions under energy values are listed for room temperature and 800C. ervic Following these ideas and representing the C/laver ratios in Fig. 7(where Slayer corresponds to the fracture energy of 4. conclusions deflecting along)it can be inferred that no debonding The mechanical behaviour of an alumina-zirconia mul ould take place at room or high temperature during the tilayered ceramic designed with thin internal compressive flexural tests. However, the experimental results at 800C layers and strong interfaces has been investigated under during heating are clearly in conflict with understanding of the parameters which crack flexural loading at nd high temperatures. Experi- delamination. This contradictory behaviour has also beer mental findings showed that improvement in mechanical observed by other authors for a Si, N,/BN multilayered properties at high temperatures in comparison to the alu- system [5]. mina-based reference material is essentially related to the maintenance of the compressive stresses developed during ering in the internal thin layers of the laminate struc- Table 4 ture, acting as an effective barrier to crack propagatio Fracture toughness and fracture energies of the ATz and aMz layers at In such cases, a steep R-curve behaviour is found, leading room temperature and at 800C to a greater reliability in terms of structural design com- Material K(MPam 2) (J/m2) s,(/m2) pared to the brittle behaviour of monolithic ceramics. (20°C)(8 (20°C)(800°C) Additionally, step-wise fracture is observed under these 3.2+0.1 3.0+0.1 24+1 24+1 22+1 conditions, owing to crack bifurcation mechanisms taking 士 place in the thin compressive layers. In some cases, i.e. at Interface fracture toughness is assumed to be the amz fracture toughness 800C during heating, bifurcation within the first compres- sive layer is followed by interface delamination in the next s, is assumed as the fracture energy of the interface at every interface owing to the change in the elastic properties of the temperature. layers with the temperature. Special attention must be paic

preferential crack growth was observed along the ATZ/ AMZ interfaces, which indicated that, in spite of the shear stresses present at the interfaces between layers caused by the different stress states, the interfaces between layers were strong. In this regard, strong interfaces have been reported by other authors for alumina/YTZP layered ceramics, which maintained their structural integrity after high-tem￾perature compression testing [12,13,15,16]. In this case, as no preferential crack growth was discerned, it can be spec￾ulated that the minimum value of the interface fracture energy is of the same order as that of the most brittle layer, i.e. KI(AMZ). In Table 4, the corresponding fracture energy values are listed for room temperature and 800 C. Following these ideas and representing the fi/flayer ratios in Fig. 7 (where flayer corresponds to the fracture energy of the layer where the impinging crack is penetrating to or deflecting along) it can be inferred that no debonding should take place at room or high temperature during the flexural tests. However, the experimental results at 800 C during heating are clearly in conflict with the current understanding of the parameters which govern crack delamination. This contradictory behaviour has also been observed by other authors for a Si3N4/BN multilayered system [5]. A possible explanation for this debonding at 800 C lies in the assumption made when estimating the gtag curves for crack deflection/penetration criteria, which are valid if the crack propagates in a straight line, i.e. 90, within a given layer. This was the case for the crack propagating from layer ATZ to layer AMZ (Fig. 6). However, experimental observations showed that when the impinging crack pene￾trated into the AMZ compressive layer bifurcation occurred and thus the propagating crack faced the new AMZ/ATZ interface with angles smaller than 90 (25 in this case), as can be seen in Fig. 6. As demonstrated by He and Hutchinson [47], the tendency for a crack to delaminate increases for small angles. Therefore, Fig. 7 cannot describe this particular case of interface debonding from the AMZ to the ATZ layer, and thus an ‘‘upwards’’ correction of the gtag = +0.03 curve should be recalled, although it is out of the scope of the present investigation. It should be highlighted that the delamination process taking place in the AMZ/ATZ interfaces would increase the fracture energy of the system, changing the propagating crack from Mode I to Mode II of fracture. In this regard, the Mode II crack propagation should be considered to make a significant contribution to the fracture energy of the system, similar to the enhancement provided by crack bifurcation along the centre of the AMZ compressive layer under room temperature conditions [21]. The strength reliability of the multilayered system stud￾ied here together with the capability of fracture energy dis￾sipation through bifurcation and/or delamination mechanisms emphasise the suitability of this layered archi￾tecture as a substitute for alumina-based monolithic com￾ponents for high-temperature applications, always being aware of the temperature hysteretic conditions under service. 4. Conclusions The mechanical behaviour of an alumina–zirconia mul￾tilayered ceramic designed with thin internal compressive layers and strong interfaces has been investigated under flexural loading at room and high temperatures. Experi￾mental findings showed that improvement in mechanical properties at high temperatures in comparison to the alu￾mina-based reference material is essentially related to the maintenance of the compressive stresses developed during sintering in the internal thin layers of the laminate struc￾ture, acting as an effective barrier to crack propagation. In such cases, a steep R-curve behaviour is found, leading to a greater reliability in terms of structural design com￾pared to the brittle behaviour of monolithic ceramics. Additionally, step-wise fracture is observed under these conditions, owing to crack bifurcation mechanisms taking place in the thin compressive layers. In some cases, i.e. at 800 C during heating, bifurcation within the first compres￾sive layer is followed by interface delamination in the next interface owing to the change in the elastic properties of the layers with the temperature. Special attention must be paid Fig. 7. Comparison of the material parameters with the crack deflec￾tion/penetration He–Hutchinson criteria at room temperature and at 800 C, assuming straight crack propagation, i.e. 90, of the impinging crack. Table 4 Fracture toughness and fracture energies of the ATZ and AMZ layers at room temperature and at 800 C Material K (MPa m1/2) f (J/m2 ) fi (J/m2 ) (20 C) (800 C) (20 C) (800 C) ATZ 3.2 ± 0.1 3.0 ± 0.1 24 ± 1 24 ± 1 22 ± 1a AMZ 2.6 ± 0.1 — 22 ± 1 29 ± 1 Interface fracture toughness is assumed to be the AMZ fracture toughness at 20 C and to be constant at every temperature. a fi is assumed as the fracture energy of the interface at every temperature. 4900 R. Bermejo et al. / Acta Materialia 55 (2007) 4891–4901