《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_brittleness10

Availableonlineatwww.sciencedirect.com Science Direct E噩≈RS ELSEVIER Joumal of the European Ceramic Society 28(2008)1961-1971 www.elsevier.comlocate/jeurceramsoc fracture behaviour of microcrack -free alumina-aluminium titanate ceramics with second phase nanoparticles at alumina grain boundaries S Bueno a m.H. Berger b.R. Moreno a c.Baudin a a Instituto de Ceramica y Vidrio(CSIC). C. Kelsen 5, 28049 Madrid, Spain Ecole des Mines de Paris, Centre des Materiaur, 91003 Evry Cedex, france Received 12 October 2007: received in revised form 19 December 2007; accepted 4 January 2008 Available online 4 March 2008 Alumina+10 vol %o aluminium titanate composites were obtained by colloidal filtration and reaction sintering of alumina and titania. The materials were dense with aluminium titanate grains of average sizes 2.2-2. 4 um located mainly at alumina triple points. The reaction sintering schedule promoted the formation of additional nanometric grains, identified as aluminium titanate using STEM-EDX analysis between the alumina grains. This special microstructure led to a change of the toughening mechanism from the typical crack bridging reported for microcrack-free composites fabricated from alumina and aluminium titanate powders to microcracking The identification of microcracking as the main toughening mechanism was done from the analysis of stable fracture tests of SENVB samples in three points bending and fractographic observations. Monophase alumina materials with similar grain sizes were used as referer Different fracture toughness parameters were derived from the load-displacement curves: the critical stress intensity factor, KiC, the critical energy release rate, Gic, the J-Integral and the work of fracture, ywoF, and the R curves were also built. The comparison between the linear elastic fracture arameters and the non-linear ones revealed significant toughening and faw tolerance o 2008 Elsevier Ltd. all rights reserved. Keywords: D. Al2O3; D. Al2 TiOs; C Mechanical properties: C. Toughening: B. Nanocomposites Introduction 10-6°C-1,a25-100c=-2:7×10-6°C-)13and mina shows limited anisotropy(aa25-10000C=8.4 x 10-6 6oc-I The use of ceramic materials in structural applications is ac25-10000C=9.2 x 10-6oC-), 4 thus, high tensile or com- limited by the"ilaw sensitive"fracture, occurring sponta- pressive stresses, depending on the particular crystallographic neously from natural flaws, inherent to the brittle behaviour. orientation of the grains, would develop during cooling from The"flaw tolerance"approach deals with the development of the sintering temperature at the grain-matrix interfaces due to microstructures that originate toughening mechanisms to reduce thermal expansion mismatch. Depending on grain size and the the sensitivity of the strength to the size of any processing or characteristics of the grain boundaries, microcracking might induced flaw, thus improving the reliability of the materials. occur during cooling from sintering and/or during fracture. Such mechanisms originate an increasing resistance with con- In the early 90s alumina-aluminium titanate composites tinued crack extension, rising R-curve behaviour, and most of with aluminium titanate contents 20-30 vol% obtained from them are caused by localized internal residual stresses in the alumina and aluminium titanate mixtures, were studied by materials other authors. Crack bridging by second phase agglomerates Alumina(Al2O3-aluminium titanate(Al2TiO5) materials nd by lar was e n offer improved flaw tolerance and toughness 4-12 Ther- ing mechanism leading to R-curve behaviour, assessed by hal expansion of aluminium titanate is highly anisotr- the indentation-strength method; no toughness values were opic(a25-100c010.9×10-60C-1,ab25-100c20.5× Corresponding author. Tel: +3491 7355840: fax: +3491 7355843 In this work. B-Al2TiOs orthorhombic crystal is described by a b-face cen- E-mail address: cbaudin @icv csices(C. Baudin) tered unit cell, space group Bbmm, a=9.439A, b=9.647A c=3.593A 0955-2219/S-see front matter o 2008 Elsevier Ltd. All rights reserved. doi: 10.1016/j-jeurceramsoc 2008.01.01

Available online at www.sciencedirect.com Journal of the European Ceramic Society 28 (2008) 1961–1971 Fracture behaviour of microcrack-free alumina–aluminium titanate ceramics with second phase nanoparticles at alumina grain boundaries S. Bueno a, M.H. Berger b, R. Moreno a, C. Baud´ın a,∗ a Instituto de Cer ´amica y Vidrio (CSIC). C. Kelsen 5, 28049 Madrid, Spain b Ecole des Mines de Paris, Centre des Mat´eriaux, 91003 Evry Cedex, France Received 12 October 2007; received in revised form 19 December 2007; accepted 4 January 2008 Available online 4 March 2008 Abstract Alumina + 10 vol.% aluminium titanate composites were obtained by colloidal filtration and reaction sintering of alumina and titania. The materials were dense with aluminium titanate grains of average sizes 2.2–2.4 m located mainly at alumina triple points. The reaction sintering schedule promoted the formation of additional nanometric grains, identified as aluminium titanate using STEM–EDX analysis between the alumina grains. This special microstructure led to a change of the toughening mechanism from the typical crack bridging reported for microcrack-free composites fabricated from alumina and aluminium titanate powders to microcracking. The identification of microcracking as the main toughening mechanism was done from the analysis of stable fracture tests of SENVB samples in three points bending and fractographic observations. Monophase alumina materials with similar grain sizes were used as reference. Different fracture toughness parameters were derived from the load–displacement curves: the critical stress intensity factor, KIC, the critical energy release rate, GIC, the J-Integral and the work of fracture, γWOF, and the R curves were also built. The comparison between the linear elastic fracture parameters and the non-linear ones revealed significant toughening and flaw tolerance. © 2008 Elsevier Ltd. All rights reserved. Keywords: D. Al2O3; D. Al2TiO5; C. Mechanical properties; C. Toughening; B. Nanocomposites 1. Introduction The use of ceramic materials in structural applications is limited by the “flaw sensitive” fracture, occurring sponta￾neously from natural flaws, inherent to the brittle behaviour. The “flaw tolerance” approach deals with the development of microstructures that originate toughening mechanisms to reduce the sensitivity of the strength to the size of any processing or induced flaw, thus improving the reliability of the materials.1–3 Such mechanisms originate an increasing resistance with con￾tinued crack extension, rising R-curve behaviour, and most of them are caused by localized internal residual stresses in the materials. Alumina (Al2O3)–aluminium titanate (Al2TiO5) materials can offer improved flaw tolerance and toughness.4–12 Ther￾mal expansion of aluminium titanate is highly anisotr￾opic (αa25–1000 ◦C = 10.9 × 10−6 ◦C−1, αb25–1000 ◦C = 20.5 × ∗ Corresponding author. Tel.: +34 91 7355840; fax: +34 91 7355843. E-mail address: cbaudin@icv.csic.es (C. Baud´ın). 10−6 ◦C−1, αc25–1000 ◦C = −2.7 × 10−6 ◦C−1) 1 13 and alu￾mina shows limited anisotropy (αa25–1000 ◦C = 8.4 × 10−6 ◦C−1, αc25–1000 ◦C = 9.2 × 10−6 ◦C−1),14 thus, high tensile or com￾pressive stresses, depending on the particular crystallographic orientation of the grains, would develop during cooling from the sintering temperature at the grain–matrix interfaces due to thermal expansion mismatch. Depending on grain size and the characteristics of the grain boundaries, microcracking might occur during cooling from sintering and/or during fracture. In the early 90s alumina–aluminium titanate composites with aluminium titanate contents 20–30 vol.% obtained from alumina and aluminium titanate mixtures, were studied by other authors.4–7 Crack bridging by second phase agglomerates and by large alumina grains was identified as the toughen￾ing mechanism leading to R-curve behaviour, assessed by the indentation–strength method; no toughness values were 1 In this work, -Al2TiO5 orthorhombic crystal is described by a b-face cen￾tered unit cell, space group Bbmm, a = 9.439 A, ˚ b = 9.647 A, ˚ c = 3.593 A. ˚ 0955-2219/$ – see front matter © 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.jeurceramsoc.2008.01.017

S. Bueno et al. /Journal of the European Ceramic Society 28 (2008)1961-1971 provided for the fine-grained materials with homogeneous where E=E/(1-v)is the generalized Youngs modulus for microstructures. All materials presented different levels of plane strain(E is the Youngs modulus and v is the Poisson's microcracks in the"as sintered"state. A latter work on ratio) microcrack-free and fine-grained alumina+ 10 vol %o aluminium The activation of toughening mechanisms during the frac- titanate fabricated from alumina and aluminium titanate mix- ture of ceramic materials gives rise to inelastic strain processes tures showed that second phase grains as well as matrix grains that produce additional release of the elastic energy accumulated could act as bridges in the wake of the propagating crack. 8 in the material at the moment of fracture initiation and/or con- This material presented increased thermal shock resistance than tributing to the retardation of crack growth.The inelastic strain a monophase alumina of similar grain size while aining levels achieved in ceramic materials can be enough to restrict the strength direct utilization of linear elastic fracture toughness parameters The initial objective of this work was to investigate the since they become dependent on testing and specimen geometry possibilities of crack bridging in fine-grained, homogeneous for non-linear materials. 18-20 and microcrack -free alumina-10 vol %o aluminium titanate com The rising R-curve behaviour, increasing Kic or GIc posites for flaw tolerance. Reaction sintering of alumina and with crack extension(Aa), has traditionally been the most titania was used as processing route. I 5 The microstruc- utilized approach to analyze deviations from the linear tures of the reaction sintered materials were different than behaviour induced by toughening in dense and fine-grained nat of the previously studied material, with a bimodal dis- ceramics. 9, 21-22 In equilibrium conditions, the applied stress tribution of aluminium titanate grains with nanoparticles intensity factor, KI, is balanced by the crack growth resistance, located at the alumina grain boundaries. The characteriza- K, and maximum values of this latter, Ko, are reached when tion of the fracture process in the composites and monophase the process zone is completely developed. alumina materials, combining different fracture parameters In order to build the R curve of the materials, crack growth gether with fractographic observations, has allowed determin- resistance and crack length values during crack extension are ing the extreme effect of the grain boundary characteristics needed. The"in situ" measurement of crack length can be a prob- in the fracture process. The major toughening mechanism lem especially for materials such as alumina-aluminiumtitanate identified in the composite studied here has been microcrack- composites, constituted by phases with large differences in hard- ness and in which residual stresses are present. The low quality of polished surfaces of relatively large specimens(e.g: bending 2. Quantification of fracture toughness bars with lateral face dimension 50mm x 6 mm)of such mate- rials makes the identification and monitoring of the propagating In general, the linear elastic fracture behaviour of ceramic crack enormously difficult Alternatively, the R curves can be determined by the indirect critical stress intensity factor in mode I, KiC, and critical strain method that defines an equivalent crack length as a function energy release rate, GIC. For three-point-bend beams, the val- of the elastic compliance of the specimen, C 23-25 For par- ues of Kic can be determined from the notch depths and the allelepiped bars with straight through notches tested in three maximum loads reached in the tests according to the general points bending, the expression provided by Guinea et al. can stress intensity formulation, valid for any notch depth, a, in line be utilized(eq (4)) elastic materials(Eq(1)6 (CEB) 3PL 2BW3/2 X Y(a) [CEB+qI(CEB)2+q2(CEB)+g3 where P is the maximum load, L is the span, B and w are the where E, a and B have the same meaning as before(eq.(1) width and the thickness of the bars, a is the normalized notch and qi(i=1, 2, 3) are parameters that depend on the L/W ratio length(a=alW) and Y(a)is a shape function depending on the (2.5 <(L/W)< 16) span to thickness ratio(L/W, Eq(2)) In a lesser extent, the non-linear fracture toughness paran J-integral and work of fracture, ywoF, have been used br (199+0.83a-0.31a2+0.14a3+4W/L) and co-workers. 26 Bradt and co-workers 27 and Sakai et al. 18 Y(a)=x(-0.09-0.42a+0.82a2-0.31a2) (1-a)32×(1+3a) (2) to characterize ceramic materials with coarser microstructures and higher levels of non-linearity such as refractories and fiber From Kic and Young s modulus, Gic can be calculated according The J-integral is an energy term that generalizes the energy to the analysis of Irwin that relates the stress-derived fracture release rate, G, to include non-linear elastic and inelastic toughness(Kic)and the energy-derived fracture toughness( Gic) behaviours and that describes the total energy of the crack-tip for plane strain conditions(Eq ( 3): stress-strain field. The critical value, JIC, constitutes a fracture criterion for materials where the toughening occurs along lim- ited crack propagation such as those that present small bridging

1962 S. Bueno et al. / Journal of the European Ceramic Society 28 (2008) 1961–1971 provided for the fine-grained materials with homogeneous microstructures. All materials presented different levels of microcracks in the “as sintered” state. A latter work on microcrack-free and fine-grained alumina + 10 vol.% aluminium titanate fabricated from alumina and aluminium titanate mix￾tures showed that second phase grains as well as matrix grains could act as bridges in the wake of the propagating crack.8 This material presented increased thermal shock resistance than a monophase alumina of similar grain size while maintaining strength. The initial objective of this work was to investigate the possibilities of crack bridging in fine-grained, homogeneous and microcrack-free alumina–10 vol.% aluminium titanate com￾posites for flaw tolerance. Reaction sintering of alumina and titania was used as processing route. 15 The microstruc￾tures of the reaction sintered materials were different than that of the previously studied material, with a bimodal dis￾tribution of aluminium titanate grains with nanoparticles located at the alumina grain boundaries. The characteriza￾tion of the fracture process in the composites and monophase alumina materials, combining different fracture parameters together with fractographic observations, has allowed determin￾ing the extreme effect of the grain boundary characteristics in the fracture process. The major toughening mechanism identified in the composite studied here has been microcrack￾ing. 2. Quantification of fracture toughness In general, the linear elastic fracture behaviour of ceramic materials is quantified by the following toughness parameters: critical stress intensity factor in mode I, KIC, and critical strain energy release rate, GIC. For three-point-bend beams, the val￾ues of KIC can be determined from the notch depths and the maximum loads reached in the tests according to the general stress intensity formulation, valid for any notch depth, a, in linear elastic materials (Eq. (1)) 16: KI = 3PL 2BW3/2 × Y(α) (1) where P is the maximum load, L is the span, B and W are the width and the thickness of the bars, α is the normalized notch length (α = a/W) and Y(α) is a shape function depending on the span to thickness ratio (L/W, Eq. (2)). Y(α)= √α(1.99 + 0.83α − 0.31α2 + 0.14α3 + 4(W/L) ×(−0.09 − 0.42α + 0.82α2 − 0.31α3)) (1 − α) 3/2 × (1 + 3α) (2) FromKIC and Young’s modulus,GIC can be calculated according to the analysis of Irwin that relates the stress-derived fracture toughness (KIC) and the energy-derived fracture toughness (GIC) for plane strain conditions (Eq. (3)): GIC = K2 IC E (3) where E = E/(1 − ν2) is the generalized Young’s modulus for plane strain (E is the Young’s modulus and ν is the Poisson’s ratio). The activation of toughening mechanisms during the frac￾ture of ceramic materials gives rise to inelastic strain processes that produce additional release of the elastic energy accumulated in the material at the moment of fracture initiation and/or con￾tributing to the retardation of crack growth.17 The inelastic strain levels achieved in ceramic materials can be enough to restrict the direct utilization of linear elastic fracture toughness parameters since they become dependent on testing and specimen geometry for non-linear materials.18–20 The rising R-curve behaviour, increasing KIC or GIC with crack extension (a), has traditionally been the most utilized approach to analyze deviations from the linear behaviour induced by toughening in dense and fine-grained ceramics.19,21–22 In equilibrium conditions, the applied stress intensity factor, KI, is balanced by the crack growth resistance, KR, and maximum values of this latter, K∞, are reached when the process zone is completely developed. In order to build the R curve of the materials, crack growth resistance and crack length values during crack extension are needed. The “in situ” measurement of crack length can be a prob￾lem especially for materials such as alumina–aluminium titanate composites, constituted by phases with large differences in hard￾ness and in which residual stresses are present. The low quality of polished surfaces of relatively large specimens (e.g.: bending bars with lateral face dimension 50 mm × 6 mm) of such mate￾rials makes the identification and monitoring of the propagating crack enormously difficult. Alternatively, the R curves can be determined by the indirect method that defines an equivalent crack length as a function of the elastic compliance of the specimen, C. 23–25 For par￾allelepiped bars with straight through notches tested in three points bending, the expression provided by Guinea et al.16 can be utilized (Eq. (4)): α = (CE B) 1/2 [CE B + q1(CE B) 1/2 + q2(CE B) 1/3 + q3] 1/2 (4) where E , α and B have the same meaning as before (Eq. (1)) and qi (i = 1, 2, 3) are parameters that depend on the L/W ratio (2.5 ≤ (L/W) ≤ 16). In a lesser extent, the non-linear fracture toughness parameter J-integral and work of fracture, γWOF, have been used by Li and co-workers,26 Bradt and co-workers27 and Sakai et al.18 to characterize ceramic materials with coarser microstructures and higher levels of non-linearity such as refractories and fiber reinforced ceramic matrix composites. The J-integral is an energy term that generalizes the energy release rate, G, to include non-linear elastic and inelastic behaviours and that describes the total energy of the crack-tip stress–strain field.28 The critical value, JIC, constitutes a fracture criterion for materials where the toughening occurs along lim￾ited crack propagation such as those that present small bridging zones.29

S. Bueno et al. / Journal of the European Ceramic Society 28(2008)1961-1971 1963 There are two main different procedures to determine 1389: 2003)and relative densities were calculated from these JiC,26-27,30 either based on the determination of the energy values and those of theoretical densities calculated taking values absorbed by the specimen, given by the area under the corre- of 3.99 gcm-3for alumina(ASTM 42-1468)and 3.70 g cm-3 sponding load-crack opening displacement curve, 30 or from for aluminium titanate(ASTM 26-0040 load-displacement curves by conducting tests on two spec Microstructural characterization on polished and thermally imens with different crack lengths. Both methods require etched(20C below the sintering temperature during 1 min) the identification of the propagating crack. In this work, a surfaces was performed by field emission gun scanning elec- graphical procedure was used- in which Jic is calculated tron microscopy(FEG-SEM, Hitachi, S-4700, Japan). The (Eq (5))from the difference between the areas under the load true average grain size was determined by the linear inter- (P)displacement(8)curves of the notched non-linear speci- cept method considering at least 200 grains for each phase and mens(Al) and an unnotched linear elastic specimen of the same g the correction factor 4.Chemical profiles across grain material(AE) for equal maximum loads(Pmax) boundaries were achieved by STEM-EDX (energy dispersed x- ray spectroscopy, coupled with scanning transmission electron X(Al-Ae) (5) microscopy, Tecnai F20-ST, The Netherlands) at 200 kV. Thin (W-a)B (W一a)B foils were prepared by mechanical polishing of a 3 mm diame ere a, W and B have the same meaning as before(Eq (1)). ter disk up to 15 um in thickness followed by Ar milling(PlPs The work of fracture is defined as the mean energy con- Gatan, USA, operating at 5 kV with a beam incidence of 6%0) sumption required forming the unit fracture surface area and Bars of 25 mm x 2 mm x 2.5 mm were diamond machined the additional process zone. It accounts an average value of from the sintered blocks for bend strength tests(three points, the whole fracture process that does not require any as 20 mm span, 0.5 mm min; Microtest, Spain). Engineering tion on the constitutive equation of the cracked body to deal stress-strain curves were calculated from the load values and with crack growth problems as discussed by Sakai et al. 8 The the displacement of the central part of the samples recorded work of fracture is obtained by dividing the work done on the during the bending tests and static Youngs modulus was deter specimen to propagate the crack, given by the area under the mined from the initial linear part of the curves. Given results load-displacement curves, by the area of the newly created for strength and static Youngs modulus are the average of five surfaces. For parallelepiped bars with straight through notches determinations and the standard deviation tested in flexure, this area is twice the area of the unnotched part Strength was also determined for specimens of a previously of the cross-section of the specimens studied A10 composite(named A10AT)fabricated from pow ders of Al2O3(90 vol %)and Al2TiO5 (10 vol %o)obtained by 3. Experimental procedure reaction of Al2O3 and TiO2 powders and sintered at 1500C, the starting Al2O3 and TiO2 powders used were the same as in Monoliths of monophase alumina (A) and alu- this work. nina+10 vol 9 aluminium titanate(AlO)composites were Single-Edge-V-Notch-Beams(SEVNB )of 4 mm x 6mm x manufactured by colloidal filtration from aqueous alumina, 50 mm were tested in a three points bending device using a span Al2O3, and titania, TiO2, suspensions using the optimum green of 40 mm and a cross-head speed of 0.005 mm min(Microtest, processing conditions previously established. 5.31 A mixture Spain). The compliance of the whole testing system(machine, of alumina ( =95 wt % )and titania (5 wt %)was used to supports, load cell and fixtures) was determined by testing a obtain the sintered composition with 10 vol. of aluminium thick(25 mm x 25 mm x 100 mm)unnotched alumina bar. The titanate, Al2TiO5. The starting materials were commercial obtained value was 1. 5x 10-m/N. The notches were initially a-Al2O3( Condea, HPAO5, USA) and TiO2-anatase(Merck, cut with a 150 um wide diamond wheel. Using this slot as a 808, Germany) powders. The powders were dispersed in guide, the remaining part of the notch was done with a razor deionised water by adding 0.5 wt %(on a dry solids basis) blade sprinkled with diamond pastes of successively 6 and 1 um of a carbonic acid-based polyelectrolyte (Dolapix CE64, Three relative notch depths, a, with approximately 0.4, 0.5 and Zschimmer-Schwarz, Germany). Suspensions were prepared to 0.6 of the thickness of the samples(w) were tested. The tip a solids loading of 50 vol. and ball milled with alumina radii of all notches were determined from optical observations and balls during 4 h. and they were always found to be below 20 um. The curves Plates of the materials with 70 mm x 70 mm x 10 mm dimen- load-displacement of the cross-head of the load frame were sions were obtained by slip casting, removed from the moulds recorded. All curves were corrected by subtracting the com- and dried in air at room temperature for at least 24 h. Sinter- pliance of the testing set up. ing of the green plates was performed in air in an electrical Additional tests were performed with unnotched specimens box furnace(Termiber, Spain) at heating and cooling rates of up to loads(=20N) well below the starting of the non-linear 2Cmin-I,with 4h, dwell at 1200C during heating and two behaviour and the obtained values of stiffness were used to different treatments at the maximum temperature: 2h, dwell at calculate JIc following the procedure described above(Eq (5)) 1450C and 3 h, dwell at 1550C. For all tests, samples were The fracture toughness parameters, i.e., critical stress inten- diamond machined from the sintered blocks sity factor, KiC, critical strain energy release rate, GIC, critical Densities of the sintered compacts were determined by J-integral, JiC, and work of fracture, ywoF, were calculated the Archimedes's method in water(European Standard en from the curves obtained during the sEvnb tests for the three

S. Bueno et al. / Journal of the European Ceramic Society 28 (2008) 1961–1971 1963 There are two main different procedures to determine JIC, 26–27,30 either based on the determination of the energy absorbed by the specimen, given by the area under the corre￾sponding load–crack opening displacement curve,30 or from load–displacement curves by conducting tests on two spec￾imens with different crack lengths.26 Both methods require the identification of the propagating crack. In this work, a graphical procedure was used27 in which JIC is calculated (Eq. (5)) from the difference between the areas under the load (P)–displacement (δ) curves of the notched non-linear speci￾mens (AI) and an unnotched linear elastic specimen of the same material (AE) for equal maximum loads (Pmax): JIC= 2 (W − a)B × δmax 0 Pdδ= 2 (W − a)B × (AI − AE) (5) where a, W and B have the same meaning as before (Eq. (1)). The work of fracture is defined as the mean energy con￾sumption required forming the unit fracture surface area and the additional process zone. It accounts an average value of the whole fracture process that does not require any assump￾tion on the constitutive equation of the cracked body to deal with crack growth problems as discussed by Sakai et al.18 The work of fracture is obtained by dividing the work done on the specimen to propagate the crack, given by the area under the load–displacement curves, by the area of the newly created surfaces. For parallelepiped bars with straight through notches tested in flexure, this area is twice the area of the unnotched part of the cross-section of the specimens. 3. Experimental procedure Monoliths of monophase alumina (A) and alu￾mina + 10 vol.% aluminium titanate (A10) composites were manufactured by colloidal filtration from aqueous alumina, Al2O3, and titania, TiO2, suspensions using the optimum green processing conditions previously established.15,31 A mixture of alumina (∼=95 wt.%) and titania (∼=5 wt.%) was used to obtain the sintered composition with 10 vol.% of aluminium titanate, Al2TiO5. The starting materials were commercial α-Al2O3 (Condea, HPA05, USA) and TiO2-anatase (Merck, 808, Germany) powders. The powders were dispersed in deionised water by adding 0.5 wt.% (on a dry solids basis) of a carbonic acid-based polyelectrolyte (Dolapix CE64, Zschimmer-Schwarz, Germany). Suspensions were prepared to a solids loading of 50 vol.% and ball milled with alumina jar and balls during 4 h. Plates of the materials with 70 mm × 70 mm × 10 mm dimen￾sions were obtained by slip casting, removed from the moulds and dried in air at room temperature for at least 24 h. Sinter￾ing of the green plates was performed in air in an electrical box furnace (Termiber, Spain) at heating and cooling rates of 2 ◦C min−1, with 4 h, dwell at 1200 ◦C during heating and two different treatments at the maximum temperature: 2 h, dwell at 1450 ◦C and 3 h, dwell at 1550 ◦C. For all tests, samples were diamond machined from the sintered blocks. Densities of the sintered compacts were determined by the Archimedes’s method in water (European Standard EN 1389:2003) and relative densities were calculated from these values and those of theoretical densities calculated taking values of 3.99 g cm−3 for alumina (ASTM 42-1468) and 3.70 g cm−3 for aluminium titanate (ASTM 26-0040). Microstructural characterization on polished and thermally etched (20 ◦C below the sintering temperature during 1 min) surfaces was performed by field emission gun scanning elec￾tron microscopy (FEG-SEM, Hitachi, S-4700, Japan). The true average grain size was determined by the linear inter￾cept method considering at least 200 grains for each phase and using the correction factor 4/π. 32 Chemical profiles across grain boundaries were achieved by STEM–EDX (energy dispersed X￾ray spectroscopy, coupled with scanning transmission electron microscopy, Tecnai F20-ST, The Netherlands) at 200 kV. Thin foils were prepared by mechanical polishing of a 3 mm diame￾ter disk up to 15m in thickness followed by Ar+ milling (PIPS Gatan, USA, operating at 5 kV with a beam incidence of 6%). Bars of 25 mm × 2 mm × 2.5 mm were diamond machined from the sintered blocks for bend strength tests (three points, 20 mm span, 0.5 mm min−1; Microtest, Spain). Engineering stress–strain curves were calculated from the load values and the displacement of the central part of the samples recorded during the bending tests and static Young’s modulus was deter￾mined from the initial linear part of the curves. Given results for strength and static Young’s modulus are the average of five determinations and the standard deviation. Strength was also determined for specimens of a previously studied A10 composite8 (named A10AT) fabricated from pow￾ders of Al2O3 (90 vol.%) and Al2TiO5 (10 vol.%) obtained by reaction of Al2O3 and TiO2 powders33 and sintered at 1500 ◦C; the starting Al2O3 and TiO2 powders used were the same as in this work. Single-Edge-V-Notch-Beams (SEVNB) of 4 mm × 6 mm × 50 mm were tested in a three points bending device using a span of 40 mm and a cross-head speed of 0.005 mm min−1 (Microtest, Spain). The compliance of the whole testing system (machine, supports, load cell and fixtures) was determined by testing a thick (25 mm × 25 mm × 100 mm) unnotched alumina bar. The obtained value was 1.5 × 10−7 m/N. The notches were initially cut with a 150 m wide diamond wheel. Using this slot as a guide, the remaining part of the notch was done with a razor blade sprinkled with diamond pastes of successively 6 and 1 m. Three relative notch depths, α, with approximately 0.4, 0.5 and 0.6 of the thickness of the samples (W) were tested. The tip radii of all notches were determined from optical observations and they were always found to be below 20 m. The curves load–displacement of the cross-head of the load frame were recorded. All curves were corrected by subtracting the com￾pliance of the testing set up. Additional tests were performed with unnotched specimens up to loads (∼=20 N) well below the starting of the non-linear behaviour and the obtained values of stiffness were used to calculate JIC following the procedure described above (Eq. (5)). The fracture toughness parameters, i.e., critical stress inten￾sity factor, KIC, critical strain energy release rate, GIC, critical J-integral, JIC, and work of fracture, γWOF, were calculated from the curves obtained during the SEVNB tests for the three

S. Bueno et al. /Journal of the European Ceramic Society 28 (2008)1961-1971 Table I Properties of the materials: average grain size(G), relative density (p), static Young s modulus(E) and three points bending strength(of) GA(S.D. )(um) GAT(.D)(um) p(S D )( theoretical) E(S D )(GPa) ar(SD )(MPa) 3.5(0.3) 981(0.3) 379(8) 456(29) 5.5(0.6) 981(0.5) 349(31) A10-1450 3.2(0.4 2.2(0.1) 973(0.5) 301(4) 261(6 10-1550 3.9(0.3 24(0.2) 97200.3) 272(10) 230(1) AlOAT 985 367(5) 360(31) A: alumina, AT: aluminum titanate: S D. standard deviation notch depths utilized. Reported values are the average of three(0.85-0.87)was slightly lower than the stoichiometric(0.89) determinations and errors are the standard deviations. R curves Nevertheless, results of these semi quantitative analyses are valid were determined from the load versus displacement curves cor- for comparative purposes. No Ti was detected inside the alumina responding to tests performed with a relative notch depth of 0.6 grains of the composites(Fig 2a), whose analyses were simi- of the thickness of The fracture surfaces of tested strength and SEVNB speci mens were characterized by FEG-SEM. Also small samples of the lateral faces(face dimension 50 mm x 6 mm)containing the notches and the cracks were polished and chemically etched (HF-10 vol %0-3 min) in order to observe the zones surround ing the propagating cracks to characterize the process zones. In order to complement the fractographic observations, polished surfaces of composite samples indented with a Vickers point using 50N during 10s, were also observed. 4. Results and discussion 4. Microstructure The microstructures of both aluminas were typical of mate- rials fabricated from high-purity submicron alumina powders The material sintered at 1450C was constituted by equidimen sional grains with a narrow distribution of relatively small sizes whereas that sintered at 1550C presented a coarser microstruc ture with a wide distribution of sizes and pore trapping associated 04m with exaggerated grain growth. The microstructural parameters together with the density, static Youngs modulus and strength values are reported in Table 1 The composites presented micrometer sized(2.2-2. 4 um, Fig. la and b, Table 1)aluminium titanate grains homoge neously distributed and located mainly at alumina triple points and grain boundaries and alumina grains of sizes similar to those of the monophase alumina sintered at 1450C(3. 2-3.9 um, Fig. la, Table 1). Submicrometric second phase grains were also observed inside the alumina grains and occasionally at grai boundaries(Fig. la). Additional nanometer sized grains were bserved at grain boundaries by SEM(Fig. lb) In Fig. 2 characteristic STEM observations for the ites sintered at 1550C together with EDX chemical analysis are shown. The ratios(wt %)Al/O(E1.4)and T1/O(E1.3)in the than those corresponding to the stoichiometric, 0.68 and 0.60, electron micrographs of polishedand s of the stud grains of aluminium titanate( Fig. 2a)were always well higher Fig. 1. Characteristic micr AlO composites. Scanning ermally etched surfaces. Alumina grains respectively. The Ka. B radiations emitted by light elements have appear with dark grey colour whereas micrometer sized aluminium titanat lower energies and are preferentially absorbed by carbon con- cromenc. ghter gray shade.(a)Composite A10 sintered at 1450CSubmi- tamination formed during the spot analyses. This induces an (b)Composite A10 sintered at 1550C. Detail of nanosized(arrows)aluminium underestimation of oxygen concentration. Also, the ratio Ti/al titanate grains located at the boundaries between the alumina grains

1964 S. Bueno et al. / Journal of the European Ceramic Society 28 (2008) 1961–1971 Table 1 Properties of the materials: average grain size (G), relative density (ρ), static Young’s modulus (E) and three points bending strength (σf) GA (S.D.) (m) GAT (S.D.) (m) ρ (S.D.) (% theoretical) E (S.D.) (GPa) σf (S.D.) (MPa) A-1450 3.5 (0.3) – 98.1 (0.3) 379 (8) 456 (29) A-1550 5.5 (0.6) – 98.1 (0.5) 376 (6) 349 (31) A10-1450 3.2 (0.4) 2.2 (0.1) 97.3 (0.5) 301 (4) 261 (6) A10-1550 3.9 (0.3) 2.4 (0.2) 97.2 (0.3) 272 (10) 230 (1) A10AT 98.5 (0.1) 367 (5) 360 (31) A: alumina, AT: aluminum titanate; S.D.: standard deviation. notch depths utilized. Reported values are the average of three determinations and errors are the standard deviations. R curves were determined from the load versus displacement curves cor￾responding to tests performed with a relative notch depth of 0.6 of the thickness of the samples. The fracture surfaces of tested strength and SEVNB speci￾mens were characterized by FEG-SEM. Also small samples of the lateral faces (face dimension 50 mm × 6 mm) containing the notches and the cracks were polished and chemically etched (HF-10 vol.%–3 min) in order to observe the zones surround￾ing the propagating cracks to characterize the process zones. In order to complement the fractographic observations, polished surfaces of composite samples indented with a Vickers point using 50 N during 10 s, were also observed. 4. Results and discussion 4.1. Microstructure The microstructures of both aluminas were typical of mate￾rials fabricated from high-purity submicron alumina powders. The material sintered at 1450 ◦C was constituted by equidimen￾sional grains with a narrow distribution of relatively small sizes whereas that sintered at 1550 ◦C presented a coarser microstruc￾ture with a wide distribution of sizes and pore trapping associated with exaggerated grain growth. The microstructural parameters together with the density, static Young’s modulus and strength values are reported in Table 1. The composites presented micrometer sized (2.2–2.4m, Fig. 1a and b, Table 1) aluminium titanate grains homoge￾neously distributed and located mainly at alumina triple points and grain boundaries and alumina grains of sizes similar to those of the monophase alumina sintered at 1450 ◦C (3.2–3.9m, Fig. 1a, Table 1). Submicrometric second phase grains were also observed inside the alumina grains and occasionally at grain boundaries (Fig. 1a). Additional nanometer sized grains were observed at grain boundaries by SEM (Fig. 1b). In Fig. 2 characteristic STEM observations for the compos￾ites sintered at 1550 ◦C together with EDX chemical analysis are shown. The ratios (wt.%) Al/O (∼=1.4) and Ti/O (∼=1.3) in the grains of aluminium titanate (Fig. 2a) were always well higher than those corresponding to the stoichiometric, 0.68 and 0.60, respectively. The K, radiations emitted by light elements have lower energies and are preferentially absorbed by carbon con￾tamination formed during the spot analyses. This induces an underestimation of oxygen concentration. Also, the ratio Ti/Al (0.85–0.87) was slightly lower than the stoichiometric (0.89). Nevertheless, results of these semi quantitative analyses are valid for comparative purposes. No Ti was detected inside the alumina grains of the composites (Fig. 2a), whose analyses were simi￾Fig. 1. Characteristic microstructures of the studied A10 composites. Scanning electron micrographs of polished and thermally etched surfaces. Alumina grains appear with dark grey colour whereas micrometer sized aluminium titanate grains have lighter gray shade. (a) Composite A10 sintered at 1450 ◦C. Submi￾crometric second phase grains inside the alumina matrix are pointed by arrows. (b) Composite A10 sintered at 1550 ◦C. Detail of nanosized (arrows) aluminium titanate grains located at the boundaries between the alumina grains.

S. Bueno et al. /Journal of the European Ceramic Sociery 28(2008)1961-1971 1965 3 0100020003000400050006000 Energy (ev C 1000200030 400050006000 100nm 95 Fig. 2. Characteristic scanning transmission electron microscopy(STEM)observations for the A10 composites sintered at 1550C together with EDX chemical nalysis(au -arbitrary units). (a) Alumina and aluminium titanate grains. No Ti was detected inside the alumina grains (b)Chemical profile along a line traversing an alumina/alumina grain boundary showing enrichment in Ti Negligible Si contents are detected. lar to those of the monophase specimens. However, the eDx should be aluminium titanate, formed by reaction of the thermo- line profiles across alumina grain boundaries in the composites dynamically incompatible compounds alumina and titania. The (Fig. 2b) showed a systematic evidence of Ti segregation at the fact that such particles were not observed by STEM should be alumina/alumina grain boundaries. Values from 0.5 to 2.5 Ti due to the relatively small portions of material characterized by wt% were detected with no systematic variation with alumina this method(two samples were observed) graIn size. The presence of the major impurity in the starting pow- 4.2. Toughness parameters ders, Si, was also investigated and only no Si or negligible Si contents were found in the grain boundaries(Fig. 2b). More- The load-displacement curves for both composites and for over, STEM-EDX analysis evidenced diffusion of titanium ions the three relative notch sizes showed stable fracture In Fig. 3 across the alumina grain boundaries during sintering. Thus, the characteristic curves for specimens with a relative notch length composition of the nanosized particles found by SEM(Fig. 1b) a/W=0.5 are shown. Controlled fracture was difficult to achieve

S. Bueno et al. / Journal of the European Ceramic Society 28 (2008) 1961–1971 1965 Fig. 2. Characteristic scanning transmission electron microscopy (STEM) observations for the A10 composites sintered at 1550 ◦C together with EDX chemical analysis (a.u. = arbitrary units). (a) Alumina and aluminium titanate grains. No Ti was detected inside the alumina grains. (b) Chemical profile along a line traversing an alumina/alumina grain boundary showing enrichment in Ti. Negligible Si contents are detected. lar to those of the monophase specimens. However, the EDX line profiles across alumina grain boundaries in the composites (Fig. 2b) showed a systematic evidence of Ti segregation at the alumina/alumina grain boundaries. Values from 0.5 to 2.5 Ti wt.% were detected with no systematic variation with alumina grain size. The presence of the major impurity in the starting pow￾ders, Si, was also investigated and only no Si or negligible Si contents were found in the grain boundaries (Fig. 2b). More￾over, STEM–EDX analysis evidenced diffusion of titanium ions across the alumina grain boundaries during sintering. Thus, the composition of the nanosized particles found by SEM (Fig. 1b) should be aluminium titanate, formed by reaction of the thermo￾dynamically incompatible compounds alumina and titania. The fact that such particles were not observed by STEM should be due to the relatively small portions of material characterized by this method (two samples were observed). 4.2. Toughness parameters The load–displacement curves for both composites and for the three relative notch sizes showed stable fracture. In Fig. 3 characteristic curves for specimens with a relative notch length a/W = 0.5 are shown. Controlled fracture was difficult to achieve

S. Bueno et al. /Journal of the European Ceramic Society 28 (2008)1961-1971 A10-1550 30 0000010020,030,040,05 Displacement [mm] Displacement [mm] the procedure followed for R-curve deter- Fig. 3. Characteristic load-displacement curves recorded during the SENVB mination and Jc calculation. 27 The curve correspond to the A10 composite ending tests. Specimens tested with an initial relative notch length a/w=0.5. sintered at 1550C and tested with a relative notch depth a/W=0.6.For R table fracture is observed for both composites and semi-stable fracture is determination, the arrow marks the point where the non-linear beha observed for both aluminas selected as the onset of crack propagation. From this point, the increments in the Ci+1). For Jic calculation(Eq. (5)), the areas under the curves corresponding for the monophase alumina specimens and only semi-stable frac- to specimen tested with the notch(A)and to an unnotched specimen(AE)are ture was obtained for a limited number of tests of specimens shown with relative notch depths of 0.5(Fig. 3). The introduction of larger notches led to the failure of the specimens during machin- the compliance( Ci, Ci+1, Fig 4). An example of the graphics used to calculate Jic is depicted in Fig 4 The fracture toughness parameters are summarized in For the monophase materials, the Kic values were lower Table 2. Kic values were calculated using Eq (1)and the values than those reported for aluminas with similar grain sizes of the maximum loads attained during the tests and Gic was (KIC 235-4. MPam/), determined from unstable tests 35-37 calculated according to Eq (3)from Kic values and Young's and, in some cases, using specimens with notch radii larger modulus(E, Table 1)and using the poisson ratio of dense and than those utilized in this work, o and similar to those deter- fine-grained alumina(v=0.22) mined by Sbaizero et al. 38(KIC MPam)for hot-pressed In order to determine the increments in crack length aluminas using stable fracture tests. As discussed by Bar-On et (Aa=di+1-ai)to build the R curves, the common criterion al., unstable crack extension results in apparent increases of relating the onset of crack propagation in the load-displacement fracture toughness values compared to those determined during curves with the point where the non-linear behaviour starts quasi-static crack growth. Therefore, the semi-stable crack prop- (arrow in Fig 4)18. 24. 34 was used. From this point, the increments agation obtained in this work for the alumina specimens(Fig 3) in crack length(Aa=ai+1-ai) would produce the changes in would give values closer to the actual fracture toughness Table 2 Fracture toughness parameters of the materials: critical stress intensity factor (Kic), critical energy releasing rate(Gic), critical J-integral (ic) and work of fracture (ywoF) KiC(S D )MPam/2 GIC(S D)(J/m-) JiC(S D )(/m-) JIC/GIC (SD) yWoF(SD)(J/m2) 29 1.0 2.8 196 189 1.0 A-1550 3.2(0.1) 26.2(0.7) 29.9(3.0) 1.1(0.1) 20.1(20)2 A10-1450 3.5(0.1) 384(0.8) 21(3.7) 1.1(0.1) 34.7(1.3) 3.5(0.2) 38.6(2.9) 1.1(0.1) 33.4(23) 3.5(0. 37.6(1.2) 45.9(3.3) 1.2(0.1) 35.1(1.7) A10-1550 33(0.1) 38.4(1.2) 50.8(6.0) 1.3(0.2) 40.6(1.2) 0.5 33(0.1) 374(26) 554(4.1) .5(0.1) 41.9(3.0) 0.6 33(0.2) 379(20) 53.1(2.3) .4(.1) 39.6(2.1) S D. standard deviation. For monophase alumina materials valid tests were obtained only with a relative notch depth of 0-5. The values of the two tests obtained on pecimens of alumina sintered at 1450C are shown Semi-stable tests

1966 S. Bueno et al. / Journal of the European Ceramic Society 28 (2008) 1961–1971 Fig. 3. Characteristic load–displacement curves recorded during the SENVB bending tests. Specimens tested with an initial relative notch length a/W = 0.5. Stable fracture is observed for both composites and semi-stable fracture is observed for both aluminas. for the monophase alumina specimens and only semi-stable frac￾ture was obtained for a limited number of tests of specimens with relative notch depths of 0.5 (Fig. 3). The introduction of larger notches led to the failure of the specimens during machin￾ing. The fracture toughness parameters are summarized in Table 2. KIC values were calculated using Eq. (1) and the values of the maximum loads attained during the tests and GIC was calculated according to Eq. (3) from KIC values and Young’s modulus (E, Table 1) and using the Poisson ratio of dense and fine-grained alumina (ν = 0.22). In order to determine the increments in crack length (a = ai + 1 − ai) to build the R curves, the common criterion relating the onset of crack propagation in the load–displacement curves with the point where the non-linear behaviour starts (arrow in Fig. 4) 18,24,34 was used. From this point, the increments in crack length (a = ai + 1 − ai) would produce the changes in Fig. 4. Schematic representation of the procedure followed for R-curve deter￾mination and JIC calculation.27 The curve correspond to the A10 composite sintered at 1550 ◦C and tested with a relative notch depth a/W = 0.6. For R-curve determination, the arrow marks the point where the non-linear behaviour starts, selected as the onset of crack propagation. From this point, the increments in the crack length (a = ai+1 − ai) would produce the changes in the compliance (Ci, Ci + 1). For JIC calculation (Eq. (5)), the areas under the curves corresponding to specimen tested with the notch (AI) and to an unnotched specimen (AE) are shown. the compliance (Ci, Ci + 1, Fig. 4). An example of the graphics used to calculate JIC is depicted in Fig. 4. For the monophase materials, the KIC values were lower than those reported for aluminas with similar grain sizes (KIC ∼= 3.5–4.5 MPa m1/2), determined from unstable tests35–37 and, in some cases, using specimens with notch radii larger than those utilized in this work,36 and similar to those deter￾mined by Sbaizero et al.38 (KIC ∼= 3 MPa m1/2) for hot-pressed aluminas using stable fracture tests. As discussed by Bar-On et al.,39 unstable crack extension results in apparent increases of fracture toughness values compared to those determined during quasi-static crack growth. Therefore, the semi-stable crack prop￾agation obtained in this work for the alumina specimens (Fig. 3) would give values closer to the actual fracture toughness. Table 2 Fracture toughness parameters of the materials: critical stress intensity factor (KIC), critical energy releasing rate (GIC), critical J-integral (JIC) and work of fracture (γWOF) KIC (S.D.) MPa m1/2 GIC (S.D.) (J/m2) JIC (S.D.) (J/m2) JIC/GIC (S.D.) γWOF (S.D.) (J/m2) A-1450 0.5 2.9 20.4 19.2 1.0 10.5a 2.8 19.6 18.9 1.0 9.8a A-1550 0.5 3.2 (0.1) 26.2 (0.7) 29.9 (3.0) 1.1 (0.1) 20.1 (2.0)a A10-1450 0.4 3.5 (0.1) 38.4 (0.8) 42.1 (3.7) 1.1 (0.1) 34.7 (1.3) 0.5 3.5 (0.2) 39.2 (0.6) 38.6 (2.9) 1.1 (0.1) 33.4 (2.3) 0.6 3.5 (0.1) 37.6 (1.2) 45.9 (3.3) 1.2 (0.1) 35.1 (1.7) A10-1550 0.4 3.3 (0.1) 38.4 (1.2) 50.8 (6.0) 1.3 (0.2) 40.6 (1.2) 0.5 3.3 (0.1) 37.4 (2.6) 55.4 (4.1) 1.5 (0.1) 41.9 (3.0) 0.6 3.3 (0.2) 37.9 (2.0) 53.1 (2.3) 1.4 (0.1) 39.6 (2.1) S.D.: standard deviation. For monophase alumina materials valid tests were obtained only with a relative notch depth of 0.5. The values of the two tests obtained on specimens of alumina sintered at 1450 ◦C are shown. a Semi-stable tests

S. Bueno et al. / Journal of the European Ceramic Society 28(2008)1961-1971 Kic values determined for the fine grain sized alumina The classical linear fracture toughness parameters, KIC and are inside the variability for crack-tip toughness, Ko,(1.5- Gic, are adequate to characterize fracture of the fine alumina, as 3.0MPam)reported by Seidel and Rodel+0 and Fett et al. +I revealed by the coincidence between JIc and GIC(Table 2),as for a series of aluminas with grain sizes in the range of 1-20 um. occurs for perfectly linear materials. 28 On the contrary, the KiC These authors determined Ko from"in situ"crack opening dis- increase for the alumina with larger grain size was accompanied placement(COD)measurements in a SEM and attributed the by Jic being slightly higher than GIC, revealing toughening. An significant scatter of data to charging of the crack edges. Also, increase in Kic with grain size has been reported for other alumi charging might lead to observed COD at the point of fracture nas with similar microstructures.35-36However, coarse-grained and, consequently, calculated Ko values, smaller than the real materials with mean grain sizes larger than 10-20 um, where ones intergranular fracture occurs for the largest grains(50 um) act Apart from the experimental facts discussed above, differ brid ges,25.- are required for significant toughening and ences between crack-tip toughness determined for different rising R-curve behaviour. total toughness would be the sum of contributions of intergran- monophase alumina materials with a mean grain size lower lia aluminas can be attributed to differences in fracture mode. as There are no reported values of work of fracture for dens ular grain boundary fracture, cleavage across the easy fracture 5 um. Even though the work of fracture values determined in the planes, and the increase in fracture surface due to deflection. semi-stable tests reached in this work might be slightly overesti- A significant amount of transgranular fracture, independent mated, to the authors knowledge they are the lowest everreported from the grain size of the material, was reported by Seidel and for dense fine-aluminas For the alumina with the smallest aver- Rodel-o(20%), whereas only the largest grains(>5 um) pre- age grain size(GA=3.5 um, Table 1), the value determined in sented transgranular fracture in the fine-grained alumina studied this work (ywoF 2 10J/m, Table 2)is higher than the value here(Fig 5a) (rr =6J/m-)reported by wiederhorn for the rombohedral plane AT g. 5. Fractographic observations showing the mode of fracture of the materials. Scanning electron micrographs of fracture surfaces of SENVB specimens(a and b), strength specimens(c) and indentation cracks (d).(a) Monophase alumina sintered at 1450C. The largest grains(>5 um) show transgranular fracture and the mall ones show intergranular fracture.(b) Characteristic intergranular fracture in A10 composites with microcracks perpendicular to the fracture surfaces pointed by arrows. Specimen sintered at 1450C.(c) Mostly transgranular fracture in A10 composites previously obtained without no nanoparticles at grain boundaries (AlOAT).(d) Characteristic paths of indentation cracks in the Al0 composites sintered at 1450C Intact alumina(A)and aluminium titanate(At) grains were bserved along the crack trace

S. Bueno et al. / Journal of the European Ceramic Society 28 (2008) 1961–1971 1967 KIC values determined for the fine grain sized alumina are inside the variability for crack-tip toughness, K0, (1.5– 3.0 MPa m1/2) reported by Seidel and Rodel ¨ 40 and Fett et al.41 for a series of aluminas with grain sizes in the range of 1–20m. These authors determined K0 from “in situ” crack opening dis￾placement (COD) measurements in a SEM and attributed the significant scatter of data to charging of the crack edges. Also, charging might lead to observed COD at the point of fracture and, consequently, calculated K0 values, smaller than the real ones. Apart from the experimental facts discussed above, differ￾ences between crack-tip toughness determined for different aluminas can be attributed to differences in fracture mode, as total toughness would be the sum of contributions of intergran￾ular grain boundary fracture, cleavage across the easy fracture planes, and the increase in fracture surface due to deflection. A significant amount of transgranular fracture, independent from the grain size of the material, was reported by Seidel and Rodel ¨ 40 (20%), whereas only the largest grains (>5 m) pre￾sented transgranular fracture in the fine-grained alumina studied here (Fig. 5a). The classical linear fracture toughness parameters, KIC and GIC, are adequate to characterize fracture of the fine alumina, as revealed by the coincidence between JIC and GIC (Table 2), as occurs for perfectly linear materials.28 On the contrary, the KIC increase for the alumina with larger grain size was accompanied by JIC being slightly higher than GIC, revealing toughening. An increase in KIC with grain size has been reported for other alumi￾nas with similar microstructures.35–36 However, coarse-grained materials with mean grain sizes larger than 10–20 m, where intergranular fracture occurs for the largest grains (>50m) act￾ing as bridges,25,42–43 are required for significant toughening and rising R-curve behaviour. There are no reported values of work of fracture for dense monophase alumina materials with a mean grain size lower than 5m. Even though the work of fracture values determined in the semi-stable tests reached in this work might be slightly overesti￾mated, to the authors knowledge they are the lowest ever reported for dense fine-aluminas. For the alumina with the smallest aver￾age grain size (GA = 3.5m, Table 1), the value determined in this work (γWOF ∼= 10 J/m2, Table 2) is higher than the value (γf ∼= 6 J/m2) reported by Wiederhorn for the rombohedral plane Fig. 5. Fractographic observations showing the mode of fracture of the materials. Scanning electron micrographs of fracture surfaces of SENVB specimens (a and b), strength specimens (c) and indentation cracks (d). (a) Monophase alumina sintered at 1450 ◦C. The largest grains (>5m) show transgranular fracture and the small ones show intergranular fracture. (b) Characteristic intergranular fracture in A10 composites with microcracks perpendicular to the fracture surfaces pointed by arrows. Specimen sintered at 1450 ◦C. (c) Mostly transgranular fracture in A10 composites previously obtained without no nanoparticles at grain boundaries (A10AT).8 (d) Characteristic paths of indentation cracks in the A10 composites sintered at 1450 ◦C. Intact alumina (A) and aluminium titanate (AT) grains were observed along the crack trace.

S. Bueno et al. /Journal of the European Ceramic Society 28 (2008)1961-1971 which is the preferred cleavage plane in alumina monocrystals at 43. Toughening mechanisms in the composites room temperature. The fracture energies generally determined for polycrystals are higher than those for monocrystals due to the Toughness increase with crack extension(AKR =54% and contribution of intergranular fracture, in the same way as crack- 38% for A10-1450 and A10-1550, respectively, Fig. 6)was tip toughness values in polycrystals are higher than those of observed for both composites. Moreover, they presented inter the easy cleavage planes, as discussed above. Nevertheless, the granular fracture(Fig 5b), revealing the micrometric aluminium coincidence between the values of 2ywoF and Gic for this mate- titanate grains located mainly at alumina triple points and grain rial(A-1450)would reveal the absence of significant crack-size boundaries, and numerous microcracks, perpendicular to the dependent toughening phenomena fracture surfaces, surrounding alumina and aluminium titanate For the coarser alumina(Ga=5.5 um, Table 1), the work of grains. Microcracks were not observed at the polished sur- fracture values (20J/m", Table 2)also coincide with those faces and the materials presented reversible thermal expansion ported for aluminas with similar microstructures.3 In this behaviour, 45,46 thus, the microcracks observed at the fracture material, the fact that ywoF was slightly higher than Gic reveals surfaces should be formed during the fracture process. From the action of limited additional energy consuming processes due the fracture surfaces of fine-grained materials it is not possible to the interaction of the growing crack with the microstructure. to ascertain whether such features are due to pull out of grains Taking into account the materials properties and the stiffness of that have acted as bridges during the fracture process or actual the testing device, the selected loading geometry imen and crack sizes should lead to unstable crack growth for Bridges are easily differentiated along the path of indenta the four studied materials, according to Bar-On et al.39 Nev- tion cracks such as those shown in Fig. 5d in which relatively ertheless, semi-stable tests were obtained for the aluminas and small grains of alumina(2-3 um) and aluminium titanate were easier to obtain for A-1550 specimens than for the finer =1-2 um) that acted as frictional sliding bridges during frac- ained alumina. However, it was not possible to build the R- ture are observed. In general, crack bridging efficiency, in curve because it is not possible to calculate compliance at each terms of AKR and of the crack extension along which AKl unloading point for semi-unstable tests ases with bridge size. Conversely, the smaller For the composites, the brittle fracture parameters, KIc and KR increase and crack extension for the material with the GIc, were similar, and higher than those corresponding to the largest grain sizes(AKR=1.1 MPa malong a crack extension alumina with similar grain size(A-1450, Table2). On the con-△a≈360mand△KR1.5 MPam2 along△a全480m trary, the ratio JIC/GIC (Table 2)was slightly higher than l for the for A10-1550 and A10-1450, respectively, Fig. 6) suggest that composite sintered at 1450C and increased for that fabricated the main toughening mechanism was not bridging 1550C. Moreover, they presented rising R-curve behaviour Post-fracture examinations of the zones that surrounded the (Fig. 6)with KR increasing to the steady state value(Koo)over notch and crack-tip regions in tested bend bars(Fig. 7) showed a crack extension, Aa, of about 360 and 480 um for A10-1550 irregular shaped damaged zones(Fig. 7a)of widths =15-30 and A10-1450, respectively. Therefore, the fracture behaviour and 20-40 um for the composites sintered at 1450 and 1550C of these materials would be more adequately described by the respectively. Detailed observations of these damaged zones non-brittle fracture parameters, Jic and R curve than by Kic or revealed microcracking along grain boundaries( Fig. 7b). These GIC. Extrapolation from the R curves(Fig. 6)showed crack-tip observations demonstrate that microcracking acted as toughen- toughness, Ko, values similar for both composites and of the ing mechanism during fracture of the composites. In general same order as that of the alumina material with similar grain microcracking is associated with low resistance of the materials size(A-1450, Table 2) to the propagation of small defects and, therefore, low strength values. Data in Table I clearly show how the composites devel oped here present significantly lower strength values than the A10-1450 one fabricated from already reacted aluminium titanate. This material, in which no titanium segregation occurred at the alu- Al-1550 mina grain boundaries, presented mostly transgranular fracture and no microcracking, crack bridging being the only toughening mechanism observed. According to the microcracking model by Evans and Faber+ 0 and the work by Lutz et al. +o it is possible to relate the width of he microcracked zones and the value of the crack-tip toughness Frontal Ko, with the critical stress for microcrack initiation according to Extended zone △a[pm h ×(1+u)2× ad-displacement curves of notched specimens with a relative notch depth of 0.6 and considering the onset of crack propagation at the point where the where Ko can be taken as the constant matrix crack-tip intensity on-linear behaviour starts in the load-displacement curves factor, equal to Kic for the monophase alumina with similar

1968 S. Bueno et al. / Journal of the European Ceramic Society 28 (2008) 1961–1971 which is the preferred cleavage plane in alumina monocrystals at room temperature.44 The fracture energies generally determined for polycrystals are higher than those for monocrystals due to the contribution of intergranular fracture, in the same way as crack￾tip toughness values in polycrystals are higher than those of the easy cleavage planes, as discussed above. Nevertheless, the coincidence between the values of 2γWOF and GIC for this mate￾rial (A-1450) would reveal the absence of significant crack-size dependent toughening phenomena. For the coarser alumina (GA = 5.5m, Table 1), the work of fracture values (∼=20 J/m2, Table 2) also coincide with those reported for aluminas with similar microstructures.38 In this material, the fact that γWOF was slightly higher than GIC reveals the action of limited additional energy consuming processes due to the interaction of the growing crack with the microstructure. Taking into account the materials properties and the stiffness of the testing device, the selected loading geometry and the spec￾imen and crack sizes should lead to unstable crack growth for the four studied materials, according to Bar-On et al.39 Nev￾ertheless, semi-stable tests were obtained for the aluminas and were easier to obtain for A-1550 specimens than for the finer grained alumina. However, it was not possible to build the R￾curve because it is not possible to calculate compliance at each unloading point for semi-unstable tests. For the composites, the brittle fracture parameters, KIC and GIC, were similar, and higher than those corresponding to the alumina with similar grain size (A-1450, Table 2). On the con￾trary, the ratio JIC/GIC (Table 2) was slightly higher than 1 for the composite sintered at 1450 ◦C and increased for that fabricated at 1550 ◦C. Moreover, they presented rising R-curve behaviour (Fig. 6) with KR increasing to the steady state value (K∞) over a crack extension, a, of about 360 and 480 m for A10-1550 and A10-1450, respectively. Therefore, the fracture behaviour of these materials would be more adequately described by the non-brittle fracture parameters, JIC and R curve than by KIC or GIC. Extrapolation from the R curves (Fig. 6) showed crack-tip toughness, K0, values similar for both composites and of the same order as that of the alumina material with similar grain size (A-1450, Table 2). Fig. 6. Characteristic R curves determined for the composites from the load–displacement curves of notched specimens with a relative notch depth of 0.6 and considering the onset of crack propagation at the point where the non-linear behaviour starts in the load–displacement curves. 4.3. Toughening mechanisms in the composites Toughness increase with crack extension (KR ∼= 54% and 38% for A10-1450 and A10-1550, respectively, Fig. 6) was observed for both composites. Moreover, they presented inter￾granular fracture (Fig. 5b), revealing the micrometric aluminium titanate grains located mainly at alumina triple points and grain boundaries, and numerous microcracks, perpendicular to the fracture surfaces, surrounding alumina and aluminium titanate grains. Microcracks were not observed at the polished sur￾faces and the materials presented reversible thermal expansion behaviour,45,46 thus, the microcracks observed at the fracture surfaces should be formed during the fracture process. From the fracture surfaces of fine-grained materials it is not possible to ascertain whether such features are due to pull out of grains that have acted as bridges during the fracture process or actual microcracks developed during fracture. Bridges are easily differentiated along the path of indenta￾tion cracks such as those shown in Fig. 5d in which relatively small grains of alumina (∼=2–3m) and aluminium titanate (∼=1–2m) that acted as frictional sliding bridges during frac￾ture are observed. In general, crack bridging efficiency, in terms of KR and of the crack extension along which KR occurs, increases with bridge size.4–5 Conversely, the smaller KR increase and crack extension for the material with the largest grain sizes (KR ∼= 1.1 MPa m1/2 along a crack extension a ∼= 360m and KR ∼= 1.5 MPa m1/2 along a ∼= 480m, for A10-1550 and A10-1450, respectively, Fig. 6) suggest that the main toughening mechanism was not bridging. Post-fracture examinations of the zones that surrounded the notch and crack-tip regions in tested bend bars (Fig. 7) showed irregular shaped damaged zones (Fig. 7a) of widths ∼=15–30 and 20–40m for the composites sintered at 1450 and 1550 ◦C, respectively. Detailed observations of these damaged zones revealed microcracking along grain boundaries (Fig. 7b). These observations demonstrate that microcracking acted as toughen￾ing mechanism during fracture of the composites. In general, microcracking is associated with low resistance of the materials to the propagation of small defects and, therefore, low strength values. Data in Table 1 clearly show how the composites devel￾oped here present significantly lower strength values than the one fabricated from already reacted aluminium titanate.8 This material, in which no titanium segregation occurred at the alu￾mina grain boundaries, presented mostly transgranular fracture and no microcracking, crack bridging being the only toughening mechanism observed.8 According to the microcracking model by Evans and Faber47 and the work by Lutz et al.48 it is possible to relate the width of the microcracked zones and the value of the crack-tip toughness, K0, with the critical stress for microcrack initiation according to Eq. (6): h = √3 12π × (1 + ν) 2 × K0 σc 2 (6) where K0 can be taken as the constant matrix crack-tip intensity factor, equal to KIC for the monophase alumina with similar

S. Bueno et al. / Journal of the European Ceramic Society 28(2008)1961-1971 1969 Values of Gic in the R curves determined for the composites AlO with a relative notch depth of 0.6 Go(S D)(/m2) Goo(S D)(/m2) Jc(S D)(/m2) A10-1450214(1.6) 585(2.7) A10-1550 94(2.1 569(29) 53.1(23) Go: initial values and Goo: steady state values. The Jc values correspond Table 2. S D. standard deviatior The different non-brittle mechanical parameters calculated in this work did not follow the same trend as a function of the microstructure of the composites. Toughness values from the R curves for completely developed process zones, Koo and Goo seem to be slightly higher for the fine-grained material, A10-1450, that presented smaller Go and process zone width, h, but significantly higher Aa across which the toughness increase (AKR)occurred. On the contrary, Jic was significantly higher for the coarse-grained composite, A10-1550, with larger Go and h and smaller Aa. This discrepancy is due to the fact that Jiccon- stitutes a fracture criterion for materials where the toughening occurs along limited crack propagation. 29 Therefore, Jic will be closer to the toughness of composite Al0-1550, for which the major part(76%)of the total toughness increase(38%)occurred along one half (180 um) of the total crack growth before the steady state was reached(Fig. 6). On the contrary, significantly larger crack growth had to take place in the composite A10-1450 to reach the steady state. In this latter material, a crack growth of 230 um occurred before the 76% of the total KR increase(54%) was reached(Fig. 6) Fig. 7. Post-fracture observations of the zones that surrounded the notch and The work of fracture values for the composites(Table 2)were crack-tip region in the bend bars of Al0. Scanning electron micrographs of considerably higher than those for the monophase aluminas in polished and chemically etched (HF 10 vol So-3 min)surfaces. (a) Damaged agreement with the toughening mechanisms described, and sim ite sintered at 1450 C(b) Detail of microcracks in the composite sintered 1550°C grain size of 25 um( J/m2)0 and in porous aluminas with a mean grain size of 15-20 um(40 J/m-). In those coarse- grained aluminas the main toughening mechanisms identified, grain size as those of the composites, A-1450 0e is the critical crack bridging and crack branching, 2 were related to extensive stress to trigger the microcracked process zone grain boundary microcracking, at the expense of lower strength Inserting in Eq.(6) the average width values from Fig. 7 values (300 MPa, 25% inferior to strength for fine-graine (h=20 and 30 um, for A10-1450 and A10-1550, respectively) alumina). The same drawback strength-microcracking occurs and the Ko values from Table 2(2.8 MPam), critical stresses in the composites studied here, that present lower strengths (oc)of 163 and 133 MPa for A10-1450 and A10-1550, respec- than the previously studied composite material (28-36% lower tively, are obtained. These values are one order of magnitude Table 1) higher than that determined by acoustic emission (20 MPa) Moreover, work of fracture values substantially exceeded for a coarse-grained(A123, 20-40 um, dso=16 um) alumina those of both energy-fracture toughness that resulted from the material+/49 and for composite materials of zirconia-alumina extension of the principles of linear elastic fracture to situations in which significantly larger process zones(h=mm) than the where the inelastic deformation occurs prior to fracture, JIc and ones observed in the materials studied here were observed. Goo, (YwoF >Jic/2, Goo/2, Tables 2 and 3). This fact suggests On the contrary, they are similar to those reported for that there are additional non-linear phenomena occurring during zirconia-alumina composites with small microcrack process fracture of the composites that significantly contribute to the total zones(h=60-160 um). o Even though the size of the process energy consumption but not to resistance to crack initiation and zone for the above mentioned coarse-grained alumina was not that can be envisaged as follows. The efficiency of microcraks reported, its size should be larger than those observed in this in the toughening of the studied composites will be the result work. In fact, using Eq (6) and for Ko the range of values deter- of a compromise between their crack shielding and weakening mined by Seidel and Rodel#0 and Fett et al. 4:1.5-3MPam2, effects. For sufficient levels of microcrack density, microcrack process zones of h=384-1539 um are found could coalesce and link together with the crack front, leading to

S. Bueno et al. / Journal of the European Ceramic Society 28 (2008) 1961–1971 1969 Fig. 7. Post-fracture observations of the zones that surrounded the notch and crack-tip region in the bend bars of A10. Scanning electron micrographs of polished and chemically etched (HF 10 vol.%–3 min) surfaces. (a) Damaged zone adjacent to one side of the crack (upper part of the image) in the compos￾ite sintered at 1450 ◦C. (b) Detail of microcracks in the composite sintered at 1550 ◦C. grain size as those of the composites, A-1450. σc is the critical stress to trigger the microcracked process zone. Inserting in Eq. (6) the average width values from Fig. 7 (h ∼= 20 and 30m, for A10-1450 and A10-1550, respectively) and the K0 values from Table 2 (2.8 MPa m1/2), critical stresses (σc) of 163 and 133 MPa for A10-1450 and A10-1550, respec￾tively, are obtained. These values are one order of magnitude higher than that determined by acoustic emission (20 MPa) for a coarse-grained (A123, 20–40m, d50 = 16m) alumina material47,49 and for composite materials of zirconia–alumina in which significantly larger process zones (h ∼= mm) than the ones observed in the materials studied here were observed.48 On the contrary, they are similar to those reported for zirconia–alumina composites with small microcrack process zones (h ∼= 60–160m).48 Even though the size of the process zone for the above mentioned coarse-grained alumina was not reported, its size should be larger than those observed in this work. In fact, using Eq. (6) and for K0 the range of values deter￾mined by Seidel and Rodel ¨ 40 and Fett et al.41: 1.5–3 MPa m1/2, process zones of h = 384–1539m are found. Table 3 Values of GIC in the R curves determined for the composites A10 with a relative notch depth of 0.6 G0 (S.D.) (J/m2) G∞ (S.D.) (J/m2) JIC (S.D.) (J/m2) A10-1450 21.4 (1.6) 58.5 (2.7) 45.9 (3.3) A10-1550 29.4 (2.1) 56.9 (2.9) 53.1 (2.3) G0: initial values and G∞: steady state values. The JIC values correspond to Table 2. S.D.: standard deviation. The different non-brittle mechanical parameters calculated in this work did not follow the same trend as a function of the microstructure of the composites. Toughness values from the R curves for completely developed process zones, K∞ and G∞ seem to be slightly higher for the fine-grained material, A10-1450, that presented smaller G0 and process zone width, h, but significantly higher a across which the toughness increase (KR) occurred. On the contrary, JIC was significantly higher for the coarse-grained composite, A10-1550, with larger G0 and h and smaller a. This discrepancy is due to the fact that JIC con￾stitutes a fracture criterion for materials where the toughening occurs along limited crack propagation.29 Therefore, JIC will be closer to the toughness of composite A10-1550, for which the major part (76%) of the total toughness increase (38%) occurred along one half (180 m) of the total crack growth before the steady state was reached (Fig. 6). On the contrary, significantly larger crack growth had to take place in the composite A10-1450 to reach the steady state. In this latter material, a crack growth of 230m occurred before the 76% of the total KR increase (54%) was reached (Fig. 6). The work of fracture values for the composites (Table 2) were considerably higher than those for the monophase aluminas in agreement with the toughening mechanisms described, and sim￾ilar to those determined in dense alumina materials with a mean grain size of 25 m (∼=50 J/m2) 50 and in porous aluminas with a mean grain size of 15–20m (∼=40 J/m2).51 In those coarse￾grained aluminas the main toughening mechanisms identified, crack bridging and crack branching,52 were related to extensive grain boundary microcracking, at the expense of lower strength values (∼=300 MPa, 25% inferior to strength for fine-grained alumina).53 The same drawback strength-microcracking occurs in the composites studied here, that present lower strengths than the previously studied composite material (28–36% lower, Table 1). Moreover, work of fracture values substantially exceeded those of both energy–fracture toughness that resulted from the extension of the principles of linear elastic fracture to situations where the inelastic deformation occurs prior to fracture, JIC and G∞, (γWOF > JIC/2, G∞/2, Tables 2 and 3). This fact suggests that there are additional non-linear phenomena occurring during fracture of the composites that significantly contribute to the total energy consumption but not to resistance to crack initiation and that can be envisaged as follows. The efficiency of microcraks in the toughening of the studied composites will be the result of a compromise between their crack shielding and weakening effects. For sufficient levels of microcrack density, microcracks could coalesce and link together with the crack front, leading to

S. Bueno et al. /Journal of the European Ceramic Society 28 (2008)1961-1971 a decrease of the fracture toughness of the material. Neverthe- 10. Bueno, S and Baudin, C, Layered materials with high strength and less, such microcracks could lead to the branching of the main tolerance based on alumina and aluminium titanate. J. Eur Ceram. crack and, consequently, to the increase of the fracture surface, 2007,27,1455-1462. leading to the increase of the total energy consumed during crack 11. Dakskobler, A and Kosmac, T, Preparation and properties of aluminium propagation. This second phenomenon would contribute to the 12 Manurung, P, Low, L M and o' Connor, B H, Effect of beta-spodumene on the phase development in an alumina/aluminium-titanate system Mater high crack driving forces such as thermal shock Res.Bul.,2005,40,2047-2055 13. Taylor, D. Thermal expansion data. XI. Complex oxides, A2 BOs, and the garnets. Br. Ceram. Trans. J., 1987, 86, 1-6. 5. Conclusions 14. Taylor, D, Thermal expansion data. Ill. Sesquioxides, M2O3 with the corun- dum and the A B- and C-M O3 structures. Br. Ceram. Trans. J, 1984, 83 Alumina+10 vol. aluminium titanate composites were 15. Bueno. S. Moreno, R and Baudin, C, Reaction sintered Al2O3/Al2TiOs 92-98 obtained by reaction sintering of alumina and titania. The reac microcrack-free composites obtained by colloidal filtration. J. Eur Ceram. ion sintering process promoted the formation of aluminium Soc.,2004,24.2785-2791 titanate nanometric grains at grain boundaries between the alu- 16. Guinea, G. V, Pastor, J. Y, Planas, J and Elices, M. Stress intensity factor, mina graIn ompliance and CMOD for a general three-point-bend beam. Int J. fract. 1998,89,103-118 as the main toughening mechanism in the composites, which 17 Gogotsi, G A The use of brittleness measure(x)to represent mechanical behaviour of ceramics Ceram. Int. 1989. 15. 127-129 showed significant increments in work of fracture and faw tol- 18.Sakai, M, Yoshimura, J, Goto, Y and Inagaki, M. R-curve behaviour of erance as compared with monophase alumina materials with a polycrystalline graphite: microcracking and grain bridging in the wake similar microstructures egion. J. Am. Ceram. Soc. 1988, 71, 609-616 The classical linear fracture toughness parameters, KIC and 19. Steinbrech R. W, Reichl. A and Schaarwaichter, W.J. R-curve behaviou of long cracks in alumina. J. An. Ceram. Soc. 1990, 73, 2009-2015 IC, have demonstrated not to be adequate to characterize frac- 20. Fett, T. Munz, D, Geraghty, R D and White, K W, Influence of specimen ture of the composites, each fracture parameter analyzed, JIC, R geometry and relative crack size on the R-curve. Eng. Fract. Mech., 2000 curve and work of fracture gave different information about the 66.375-386 fracture behaviour of the material 21. Wachtman, J. B, Stable crack propagation and R-curve behaviour. Mechan- ical Properties of Ceramics. John Willey Sons Inc, New York, NY, 1996, pp.141-15 Acknowledgments 22. Fett, T and Munz D. Evaluation of R-curve effects in ceramics. J Mater Sci.,1993,28,742-752. This work has been supported by the EC Human Poten- 23. Hubner, H and Jillek, w, Sub-critical crack extension and crack resistance tial Programme HPRN-CT-2002-00203, by the Project CICYT 24. Tanaka. K. Akiniwa. Y. Kimachi. H and Kita. YR MAT2006-13480(Spain) and the Postdoctoral Fellowship MEC fracture of notched porous ceramics. Eng. Fract. Mech., 2003, 70, 1101 EX-2006-0555 (Spain 25. Ebrahimi, M. E, Chevalier. J. and Fantozzi, G, R-curve evaluation and bridging stress determination in alumina by compliance analysis. J. Eur References Ceram.Soc.,2003,23,943-949 26. Hashida, T, Li, C. and Takahashi, H, New development of the J-based 1. Harmer, M. P, Chan, H. M. and Miller, G. A, Unique opportunities for fracture testing r ceramic matrix composites. J. Am. Ceram. Soc.,1994,77,1553-156 J.Am.Cerm.Soc,1992,75,1715-1728 27. Homeny, J, Darroudi, T and Bradt, R C, J-integral measurements of the Steinbrech, R. w, Toughening mechanisms for ceramic materials. J. Eur fracture of 50% alumina refractories. J. Am. Ceram Soc. 1980. 63. 326- Ceran.Soc.,1992,10,131-142 3. Evans. A G. P ve on the development of high-toughness ceramics. 28. Rice, J.R, A path independent integral and the approximate analysis 丿Am. Ceram.Soc.,1980,73,187-206 of strain concentration by notches and cracks. J. AppL. Mech., 1968, 3 4. Lawn, B. R, Padture, N. P. Braun, L. M. and Bennison, S.J., Model for 379-386. 29. Stevens, R. N and Guiu, F, The application of the J-integral to problems of Ceran.Soc.,1993,76,2235-2240 crack bridging. Acta Metall. Mater, 1994, 42, 1805-1810 5.Padture,NP Runyan, J. L. Bennison, SJ, Braun, L.M. and Lawn, B. 30. Droillard, C and Lamon, J.J. Fracture toughness of 2-D woven Sic/Sic R. Model for toughness curves in two-phase ceramics. Il. Microstructural CVl-composites with multilayered interphases. J Am Ceram Soc., 1996 variables. J. Am. Ceram Soc. 1993.76.2241-2247 79,849858 6. Padture, N. P. Bennison. S. J. and Chan. H. M. Flaw-tolerance and 31. Bueno, S, Moreno, R and Baudin, C, Design and processing of Al2O3 crack-resistance properties of alumina-aluminium titanate I2TiOs layered structures. J. Eur Ceram Soc., 2005, 25, 847-856 with tailored microstructures. J. Am. Ceram. Soc., 1993, 76. 2312- 32. Fullmann, R. L. Measurement of particle sizes in opaque bodies. Trans AME,J.Met.,1953,197,447. 7. Runyan, I L and Bennison, S J, Fabrication of flaw-tolerant aluminium- 33. Uribe, R and Baudin, C Aluminium titanate formation by solid-state reac tion of alumina and titania. Bol. Soc. Esp Ceram. vid, 2000, 39, 221- 8. Uribe, R. and Baudin, C, Infuence of a dispersion of aluminium titanate 228 particles of controlled size on the thermal shock resistance of alumina. J. 34. wieninger, H, Kromp, k and Pabst, R. F. Crack resistance curves of Am Ceram Soc. 2003. 86.846 alumina and zirconia at room temperature. J Mater. Sci, 1986, 21, 411-418 9.Baudin, C, Sayir, A. and Berger, M. H, Mechanical behaviour of direction- 35. Rice, R. W, Freiman, S. w and Becher, P. F- Grain-size dependence of ally solidified alumina/aluminium titanate ceramics. Acta Mater, 2006, 54, fracture energy in ceramics. I. Experiment. J. Am. Ceram. Soc., 1981, 64 3835-3841

1970 S. Bueno et al. / Journal of the European Ceramic Society 28 (2008) 1961–1971 a decrease of the fracture toughness of the material. Neverthe￾less, such microcracks could lead to the branching of the main crack and, consequently, to the increase of the fracture surface, leading to the increase of the total energy consumed during crack propagation. This second phenomenon would contribute to the resistance of the materials under loading conditions that imply high crack driving forces such as thermal shock. 5. Conclusions Alumina + 10 vol.% aluminium titanate composites were obtained by reaction sintering of alumina and titania. The reac￾tion sintering process promoted the formation of aluminium titanate nanometric grains at grain boundaries between the alu￾mina grains. This special microstructure led to extensive microcracking as the main toughening mechanism in the composites, which showed significant increments in work of fracture and flaw tol￾erance as compared with monophase alumina materials with similar microstructures. The classical linear fracture toughness parameters, KIC and GIC, have demonstrated not to be adequate to characterize frac￾ture of the composites, each fracture parameter analyzed, JIC, R curve and work of fracture gave different information about the fracture behaviour of the material. Acknowledgments This work has been supported by the EC Human Poten￾tial Programme HPRN-CT-2002-00203, by the Project CICYT MAT2006-13480 (Spain) and the Postdoctoral Fellowship MEC EX-2006-0555 (Spain). References 1. Harmer, M. P., Chan, H. M. and Miller, G. A., Unique opportunities for microstructural engineering with duplex and laminar ceramics composites. J. Am. Ceram. Soc., 1992, 75, 1715–1728. 2. Steinbrech, R. W., Toughening mechanisms for ceramic materials. J. Eur. Ceram. Soc., 1992, 10, 131–142. 3. Evans, A. G., Perspective on the development of high-toughness ceramics. J. Am. Ceram. Soc., 1980, 73, 187–206. 4. Lawn, B. R., Padture, N. P., Braun, L. M. and Bennison, S. J., Model for toughness curves in two-phase ceramics. I. Basic fracture mechanics. J. Am. Ceram. Soc., 1993, 76, 2235–2240. 5. Padture, N. P., Runyan, J. L., Bennison, S. J., Braun, L. M. and Lawn, B. R., Model for toughness curves in two-phase ceramics. II. Microstructural variables. J. Am. Ceram. Soc., 1993, 76, 2241–2247. 6. Padture, N. P., Bennison, S. J. and Chan, H. M., Flaw-tolerance and crack-resistance properties of alumina–aluminium titanate composites with tailored microstructures. J. Am. Ceram. Soc., 1993, 76, 2312– 2320. 7. Runyan, J. L. and Bennison, S. J., Fabrication of flaw-tolerant aluminium￾titanate-reinforced alumina. J. Eur. Ceram. Soc., 1991, 7, 93–99. 8. Uribe, R. and Baudin, C., Influence of a dispersion of aluminium titanate particles of controlled size on the thermal shock resistance of alumina. J. Am. Ceram. Soc., 2003, 86, 846–850. 9. Baudin, C., Sayir, A. and Berger, M. H., Mechanical behaviour of direction￾ally solidified alumina/aluminium titanate ceramics. Acta Mater., 2006, 54, 3835–3841. 10. Bueno, S. and Baudin, C., Layered materials with high strength and flaw tolerance based on alumina and aluminium titanate. J. Eur. Ceram. Soc., 2007, 27, 1455–1462. 11. Dakskobler, A. and Kosmac, T., Preparation and properties of aluminium titanate–alumina composites. J. Mat. Res., 2006, 21, 448–454. 12. Manurung, P., Low, I. M. and O‘Connor, B. H., Effect of beta-spodumene on the phase development in an alumina/aluminium-titanate system. Mater. Res. Bull., 2005, 40, 2047–2055. 13. Taylor, D., Thermal expansion data. XI. Complex oxides, A2BO5, and the garnets. Br. Ceram. Trans. J., 1987, 86, 1–6. 14. Taylor, D., Thermal expansion data. III. Sesquioxides, M2O3 with the corun￾dum and the A-, B- and C-M2O3 structures. Br. Ceram. Trans. J., 1984, 83, 92–98. 15. Bueno, S., Moreno, R. and Baudin, C., Reaction sintered Al2O3/Al2TiO5 microcrack-free composites obtained by colloidal filtration. J. Eur. Ceram. Soc., 2004, 24, 2785–2791. 16. Guinea, G. V., Pastor, J. Y., Planas, J. and Elices, M., Stress intensity factor, compliance and CMOD for a general three-point-bend beam. Int. J. Fract., 1998, 89, 103–118. 17. Gogotsi, G. A., The use of brittleness measure (x) to represent mechanical behaviour of ceramics. Ceram. Int., 1989, 15, 127–129. 18. Sakai, M., Yoshimura, J., Goto, Y. and Inagaki, M., R-curve behaviour of a polycrystalline graphite: microcracking and grain bridging in the wake region. J. Am. Ceram. Soc., 1988, 71, 609–616. 19. Steinbrech, R. W., Reichl, A. and Schaarwachter, W. J., R-curve behaviour ¨ of long cracks in alumina. J. Am. Ceram. Soc., 1990, 73, 2009–2015. 20. Fett, T., Munz, D., Geraghty, R. D. and White, K. W., Influence of specimen geometry and relative crack size on the R-curve. Eng. Fract. Mech., 2000, 66, 375–386. 21. Wachtman, J. B., Stable crack propagation and R-curve behaviour. Mechan￾ical Properties of Ceramics. John Willey & Sons Inc., New York, NY, 1996, pp. 141–157. 22. Fett, T. and Munz, D., Evaluation of R-curve effects in ceramics. J. Mater. Sci., 1993, 28, 742–752. 23. Hubner, H. and Jillek, W., Sub-critical crack extension and crack resistance ¨ in polycrystalline alumina. J. Mater. Sci., 1977, 12, 117–125. 24. Tanaka, K., Akiniwa, Y., Kimachi, H. and Kita, Y., R-curve behaviour in fracture of notched porous ceramics. Eng. Fract. Mech., 2003, 70, 1101– 1113. 25. Ebrahimi, M. E., Chevalier, J. and Fantozzi, G., R-curve evaluation and bridging stress determination in alumina by compliance analysis. J. Eur. Ceram. Soc., 2003, 23, 943–949. 26. Hashida, T., Li, C. and Takahashi, H., New development of the J-based fracture testing technique for ceramic matrix composites. J. Am. Ceram. Soc., 1994, 77, 1553–1561. 27. Homeny, J., Darroudi, T. and Bradt, R. C., J-integral measurements of the fracture of 50% alumina refractories. J. Am. Ceram. Soc., 1980, 63, 326– 331. 28. Rice, J. R., A path independent integral and the approximate analysis of strain concentration by notches and cracks. J. Appl. Mech., 1968, 35, 379–386. 29. Stevens, R. N. and Guiu, F., The application of the J-integral to problems of crack bridging. Acta Metall. Mater., 1994, 42, 1805–1810. 30. Droillard, C. and Lamon, J. J., Fracture toughness of 2-D woven SiC/SiC CVI-composites with multilayered interphases. J. Am. Ceram. Soc., 1996, 79, 849–858. 31. Bueno, S., Moreno, R. and Baud´ın, C., Design and processing of Al2O3- Al2TiO5 layered structures. J. Eur. Ceram. Soc., 2005, 25, 847–856. 32. Fullmann, R. L., Measurement of particle sizes in opaque bodies. Trans. AIME, J. Met., 1953, 197, 447. 33. Uribe, R. and Baudin, C., Aluminium titanate formation by solid-state reac￾tion of alumina and titania. Bol. Soc. Esp. Ceram. Vidr., 2000, 39, 221– 228. 34. Wieninger, H., Kromp, K. and Pabst, R. F., Crack resistance curves of alumina and zirconia at room temperature. J. Mater. Sci., 1986, 21, 411–418. 35. Rice, R. W., Freiman, S. W. and Becher, P. F., Grain-size dependence of fracture energy in ceramics. I. Experiment. J. Am. Ceram. Soc., 1981, 64, 345–350