《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_Fabrication and residual stresses characterization of novel non-oxide

Availableonlineatwww.sciencedirect.com ScienceDirect E≈RS ELSEVIER Joumal of the European Ceramic Society 26(2006)3415-342 www.elsevier.comlocate/jeurceramsoc Fabrication and residual stresses characterization of novel non -oxide multilayer ceramics D. Scitia, C,*M. Nagliatia, c.S. Tochino D, C.G. PezzottiD, c.S. guicciardi a, c a CNR-ISTEC, Institute of Science and Technology for Ceramics, Via granarolo 64, 1-48018 Faenza, Italy b Department of Materials, Kyoto Institute for Technology. Matsugasaki, Sakio-ku, Kyoto 606-8585, Japan e Research Institute for Nanoscience, RIN, Kyoto Instinte of Technology, Sakyo ku, Matsugasaki, Kyoto 606-8585, Japan Received 7 May 2005; received in revised form 26 September 2005; accepted 30 September 2005 Available online 14 November 2005 A multilayer composite was produced by the tape-casting technique in the AIN-SiC-MoSiz system, coupling insulating and electrically conductive layers. The composite was highly dense and delamination-free. The values of residual stresses due to the coupling of layers of different composition vere assessed using the indentation technique and Raman piezo-spectroscopy. The experimental techniques agreed in respect to the sign and magnitude of the residual stresses, revealing compression in the insulating layers and tension in the conductive layers. The experimental results were lower than the expected values calculated by lamination theory. The choice of the parameter values needed for the theoretical calculations appears to be critical from this point of view o 2005 Elsevier Ltd. All rights reserved. Keywords: Composites; (Raman piezo-)spectroscopy; Mechanical properties; Tape casting: Laminates; AIN; SiC; MoSi2 Introduction A number of techniques can be used to measure residual stresses in ceramic materials, including X-ray diffraction, 7.8 Laminar ceramic composites have been widely studied since neutron diffraction,9. 0 indentation technique, 2 and piezo- they offer the possibility of obtaining superior properties in spectroscopic analyses 4.5, 3, 4 In particular, Raman piezo to the design of specific architectures in which different lam- been shown to be a reliable technique for characterizing residual inated materials are coupled and the layers'thickness is care- stress maps in alumina/zirconia multilayers. 4.5, fully controlled. -6 However, it must be considered that, when In this work, a new kind of non-oxide multilayer ceramic coupling two different materials in a multilayer configuration, composite in the AIN-SiC-MoSi2 system is presented. As residual stresses can arise mainly due to the different thermal a bulk material, this system has already shown remarkable expansion of the materials. The magnitude of these residual room-temperature and high-temperature properties along with stresses depends on the elastic and thermal expansion coeffi- a high oxidation resistance up to 1500-1600C16-18The cients(CTEs)mismatch of the layers, the layers'thickness and main characteristic of this system is the ability to tailor elec- the stress-free temperature -In order to avoid cracking and trical conductivity by varying the MoSi2 content.Electro- delamination, a precise control of both the magnitude and dis- conductivity measurements have shown that for MoSi2 content tribution of the residual stresses is mandatory. In particular, an higher than 25-30 vol %o the composite is electro-conductive experimental assessment of these residual stresses can help to with resistivity of the order of 10-S2 cm, whilst for lower design better multilayer composites MoSi2 content (10-15 vol %)the composite resistivity is over 10--10 cm. Furthermore, these types of composites have been successfully produced by tape-casting technique which should allow multilayer composites to be easily produced. Corresponding author. Tel. +39 0546699748: fax: +39 054646381 One interesting application of these materials is in fact as E-mailaddress:dile@istec.cnr.it(D.Sciti) heaters and igniters where, in most of these devices, thin layers 0955-2219/S-see front matter o 2005 Elsevier Ltd. All rights reserved. doi: 10. 1016/j-jeurceramsoc. 2005.09.106

Journal of the European Ceramic Society 26 (2006) 3415–3423 Fabrication and residual stresses characterization of novel non-oxide multilayer ceramics D. Sciti a,c,∗, M. Nagliati a,c, S. Tochino b,c, G. Pezzotti b,c, S. Guicciardi a,c a CNR-ISTEC, Institute of Science and Technology for Ceramics, Via Granarolo 64, I-48018 Faenza, Italy b Department of Materials, Kyoto Institute for Technology, Matsugasaki, Sakio-ku, Kyoto 606-8585, Japan c Research Institute for Nanoscience, RIN, Kyoto Institute of Technology, Sakyo-ku, Matsugasaki, Kyoto 606-8585, Japan Received 7 May 2005; received in revised form 26 September 2005; accepted 30 September 2005 Available online 14 November 2005 Abstract A multilayer composite was produced by the tape-casting technique in the AlN–SiC–MoSi2 system, coupling insulating and electrically conductive layers. The composite was highly dense and delamination-free. The values of residual stresses due to the coupling of layers of different composition were assessed using the indentation technique and Raman piezo-spectroscopy. The experimental techniques agreed in respect to the sign and magnitude of the residual stresses, revealing compression in the insulating layers and tension in the conductive layers. The experimental results were lower than the expected values calculated by lamination theory. The choice of the parameter values needed for the theoretical calculations appears to be critical from this point of view. © 2005 Elsevier Ltd. All rights reserved. Keywords: Composites; (Raman piezo-)spectroscopy; Mechanical properties; Tape casting; Laminates; AlN; SiC; MoSi2 1. Introduction Laminar ceramic composites have been widely studied since they offer the possibility of obtaining superior properties in term of strength, fracture toughness and wear resistance; thanks to the design of specific architectures in which different lam￾inated materials are coupled and the layers’ thickness is care￾fully controlled.1–6 However, it must be considered that, when coupling two different materials in a multilayer configuration, residual stresses can arise mainly due to the different thermal expansion of the materials. The magnitude of these residual stresses depends on the elastic and thermal expansion coeffi- cients (CTEs) mismatch of the layers, the layers’ thickness and the stress-free temperature.1–3 In order to avoid cracking and delamination, a precise control of both the magnitude and dis￾tribution of the residual stresses is mandatory. In particular, an experimental assessment of these residual stresses can help to design better multilayer composites. ∗ Corresponding author. Tel.: +39 0546699748; fax: +39 054646381. E-mail address: dile@istec.cnr.it (D. Sciti). A number of techniques can be used to measure residual stresses in ceramic materials, including X-ray diffraction,7,8 neutron diffraction,9,10 indentation technique11,12 and piezo￾spectroscopic analyses.4,5,13,14 In particular, Raman piezo￾spectroscopy, due to its very small spot size (∼1
m), has already been shown to be a reliable technique for characterizing residual stress maps in alumina/zirconia multilayers.4,5,15 In this work, a new kind of non-oxide multilayer ceramic composite in the AlN–SiC–MoSi2 system is presented. As a bulk material, this system has already shown remarkable room-temperature and high-temperature properties along with a high oxidation resistance up to 1500–1600 ◦C.16–18 The main characteristic of this system is the ability to tailor elec￾trical conductivity by varying the MoSi2 content. Electro￾conductivity measurements have shown that for MoSi2 content higher than 25–30 vol.% the composite is electro-conductive with resistivity of the order of 10−3
cm, whilst for lower MoSi2 content (10–15 vol.%) the composite resistivity is over 102–103
cm. Furthermore, these types of composites have been successfully produced by tape-casting technique19 which should allow multilayer composites to be easily produced. One interesting application of these materials is in fact as heaters and igniters where, in most of these devices, thin layers 0955-2219/$ – see front matter © 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.jeurceramsoc.2005.09.106

3416 D. Sciti er al. / Joumal of the European Ceramic Society 26 (2006)3415-3423 Table I Slurry composition(vol %)at different X, Y and plasticizers ratio volume X: organic to inorganic ratio, Y: binder to plasticizer ratio Formulation Ceramic(vol %) Deflocculant(vol %) PVB(vol %) BBP+PEG(voL%o) Solvent(vol %o) x Y Plasticizers ratio .7 26.84 (100-200 um) having different electrical resistivities are after indicated as BI and BC. a heterogeneous laminated com- coupled posite was produced alternating C and I layers with sequence In this contribution, homogenous and heterogeneous lami- 1//C//C//C//C//, hereafter indicated as ICl. A schematic dia- nates in the AIN-SiC-MoSi2 system are produced. In hetero- gram of the specimens preparation is shown in Fig. 1. All the geneous structures, electrically insulating layers are coupled to laminates were warm-pressed at 75Cfor 15 min with an appliec onductive layers and the resulting microstructure is studied. pressure of 17 MPa and then heated at 80C for 15 min without Macroscopic residual stresses developed in the heterogeneous pressure. The burnout stage(150 C/h from 25 to 600 C, 30 min composite are measured by an indentation technique and by of holding time) was chosen on the basis of previous thermo- Raman piezo-spectroscopy. Experimental results are then dis- gravimetric analysis. 9 Sintering was carried out in a graphite cussed and compared to theoretical models furnace, at 1850C for 30 min(heating up: 600C/h), under flowing nitrogen. In order to prevent dissociation of AIN, the 2. Experimental green samples were placed in a powder bed with composition 80wt. AIN+20 wt.o bN. The relative densities of the sin- 2.1. Material preparation tered samples were measured using Archimedes method. On polished surfaces, the microstructure was analysed by scannin he characteristics of the starting powders are the follow. electron microscopy (SEM, Cambridge S360, Cambridge, UK) ing: Hexagonal AIN, grade C(H.C. Starck, Germany), mean and energy dispersive microanalySis (EDS, INCA Energy 300 particle size 0.85 um, oxygen content 0.25 wt %; B-SiC (BF- Oxford Instruments, UK) 12, H C. Starck), mean particle size 0. 15 um, oxygen content On the homogeneous laminates, bI and bC, the linear 0.88 wt%; tetragonal MoSi2(Aldrich, USA), mean particle thermal-expansion coefficient( CTE)was measured with dilato- dimension 2. um, oxygen content 1 wt% Y203, grade C(HC. metric tests(Netzsch Geraetebau Dil E 402, Germany )up to Starck, Germany). The as-received MoSi2 powder was ball- 1400C in air, with a heating rate of 5 C/min On the same milled for 120 h in ethanol and dried to reduce the mean particle samples, the electrical conductivity measurements were carried diameter from 2.7 to 1.8 um. The compositions in vol. of the out by a four-probe DC method at room temperature, induc ceramic layers are the following ing a current in 2 mm x 2.5 mm x 25 mm bar specimens. The current and the voltage reading were detected at the same time 55AIN+15SiC +30MoSi, labelled as C. with two different digital high-resolution multimeters. The con- 80AIN +10SiC+10MoSi2, labelled as l. ductivity values were determined from the electrical resistance measurement, taking into account the test leads distance and cross-sectional area of the samples Each composition was treated with 2 wt %o of Y2O3 as sin tering aid. The powders were dispersed in an azeotropic mixture of methyl-ethyl-ketone and ethanol (MEK-EtoH, 66-34 vol %o) with I wt %o of deflocculant(Ep, Emphos PS 610, Crompton, France) and ball-milled with Si3 N4 milling media. The binder (poly-vinyl-butyral PVB, B98, Monsanto, USA)and the plas- ticizers(polyethylene glycol PEG-400, Merck, Germany and benzyl-butyl-phtalate BBP, S160, Monsanto, USA)were sub- sequently added, according to the procedure described in ref. 19. The formulations of the slurries are reported in Table 1.The formulation of composition C was set on the basis of the work reported in ref. 9. The formulation of composition I adopted the same conditions, except for the plasticizers ratio, which was modified to favour the peeling away of the tape and to avoid BC adhesion phenomena on the support. Circular samples with a thickness of 250 um and a dian ter of 40 mm were punched out from the as-cast Fig. 1. Schematic of the laminates production. Legend: 1: insulating layers; Homogeneous laminates were obtained by stacking insulating C: conductive layers; ICL: heterogeneous laminate; BI: homogeneous laminate or conductive monolayers, i.e. I.. I/I and C/C... C/C, here- containing I layers; BC: homogeneous laminate containing C layers

3416 D. Sciti et al. / Journal of the European Ceramic Society 26 (2006) 3415–3423 Table 1 Slurry composition (vol.%) at different X, Y and plasticizers ratio volume X: organic to inorganic ratio, Y: binder to plasticizer ratio Formulation Ceramic (vol.%) Deflocculant (vol.%) PVB (vol.%) BBP + PEG (vol.%) Solvent (vol.%) X Y Plasticizers ratio I 28.42 0.91 5.63 5.63 59.41 0.7 1.0 0.7:0.3 C 26.84 1.00 5.25 5.25 61.66 0.7 1.0 0.5:0.5 (∼100–200
m) having different electrical resistivities are coupled. In this contribution, homogenous and heterogeneous lami￾nates in the AlN–SiC–MoSi2 system are produced. In hetero￾geneous structures, electrically insulating layers are coupled to conductive layers and the resulting microstructure is studied. Macroscopic residual stresses developed in the heterogeneous composite are measured by an indentation technique and by Raman piezo-spectroscopy. Experimental results are then dis￾cussed and compared to theoretical models. 2. Experimental 2.1. Material preparation The characteristics of the starting powders are the follow￾ing: Hexagonal AlN, grade C (H.C. Starck, Germany), mean particle size 0.85
m, oxygen content 0.25 wt.%; -SiC (BF- 12, H.C. Starck), mean particle size 0.15
m, oxygen content 0.88 wt.%; tetragonal MoSi2 (Aldrich, USA), mean particle dimension 2.7
m, oxygen content 1 wt.% Y2O3, grade C (H.C. Starck, Germany). The as-received MoSi2 powder was ball￾milled for 120 h in ethanol and dried to reduce the mean particle diameter from 2.7 to 1.8
m. The compositions in vol.% of the ceramic layers are the following: 55AlN + 15SiC + 30MoSi2, labelled as C. 80AlN + 10SiC + 10MoSi2, labelled as I. Each composition was treated with 2 wt.% of Y2O3 as sin￾tering aid. The powders were dispersed in an azeotropic mixture of methyl-ethyl-ketone and ethanol (MEK–EtOH, 66–34 vol.%) with 1 wt.% of deflocculant (Ep, Emphos PS 610, Crompton, France) and ball-milled with Si3N4 milling media. The binder (poly-vinyl-butyral PVB, B98, Monsanto, USA) and the plas￾ticizers (polyethylene glycol PEG-400, Merck, Germany and benzyl-butyl-phtalate BBP, S160, Monsanto, USA) were sub￾sequently added, according to the procedure described in ref. 19. The formulations of the slurries are reported in Table 1. The formulation of composition C was set on the basis of the work reported in ref. 19. The formulation of composition I adopted the same conditions, except for the plasticizers ratio, which was modified to favour the peeling away of the tape and to avoid adhesion phenomena on the support. Circular samples with a thickness of 250
m and a diam￾eter of 40 mm were punched out from the as-cast green tape. Homogeneous laminates were obtained by stacking insulating or conductive monolayers, i.e. I/I ... I/I and C/C ... C/C, here￾after indicated as BI and BC. A heterogeneous laminated com￾posite was produced alternating C and I layers with sequence I/I/C/I/C/I/C/I/C/I/I, hereafter indicated as ICI. A schematic dia￾gram of the specimen’s preparation is shown in Fig. 1. All the laminates were warm-pressed at 75 ◦C for 15 min with an applied pressure of 17 MPa and then heated at 80 ◦C for 15 min without pressure. The burnout stage (150 ◦C/h from 25 to 600 ◦C, 30 min of holding time) was chosen on the basis of previous thermo￾gravimetric analysis.19 Sintering was carried out in a graphite furnace, at 1850 ◦C for 30 min (heating up: 600 ◦C/h), under flowing nitrogen. In order to prevent dissociation of AlN, the green samples were placed in a powder bed with composition 80 wt.% AlN + 20 wt.% BN. The relative densities of the sin￾tered samples were measured using Archimedes method. On polished surfaces, the microstructure was analysed by scanning electron microscopy (SEM, Cambridge S360, Cambridge, UK) and energy dispersive microanalysis (EDS, INCA Energy 300, Oxford Instruments, UK). On the homogeneous laminates, BI and BC, the linear thermal-expansion coefficient (CTE) was measured with dilato￾metric tests (Netzsch Geraetebau Dil E 402, Germany) up to 1400 ◦C in air, with a heating rate of 5 ◦C/min. On the same samples, the electrical conductivity measurements were carried out by a four-probe DC method at room temperature, induc￾ing a current in 2 mm × 2.5 mm × 25 mm bar specimens. The current and the voltage reading were detected at the same time with two different digital high-resolution multimeters. The con￾ductivity values were determined from the electrical resistance measurement, taking into account the test leads distance and cross-sectional area of the samples. Fig. 1. Schematic of the laminates production. Legend: I: insulating layers; C: conductive layers; ICI: heterogeneous laminate; BI: homogeneous laminate containing I layers; BC: homogeneous laminate containing C layers.

D. Sciti et al / Joumal of the European Ceramic Society 26 (2006)3415-3423 2.2. Measurement of residual stress by indentation technique The presence of a residual stress field in a material mod ies the indentation crack length with respect to the length observed in the same stress-free material The influence of resid- ual stresses on the length of the indentation cracks is such that the true fracture toughness, KIc, of the material can be calculated dentation by considering the contributions due to the indentation stress and ICI CII CIII the residual stresses field for an indentation crack length c in a residual stress field Ores, this implies" k=0.016/E)2 H (1) Fig. 2. Schematic of the Raman and indentation experiments carried out on the ICI material where p is the applied indentation load, E and H the You modulus and the hardness of the material, respectively, rescence peaks occurs. The relationship is given by the following is a constant related to the aspect ratio of the crack and tensorial expression taken as 1. 29 for surface half-penny crack in a uniform stress ieldvIce versa, knowing the true fracture toughness of a △u= material, for example, by measuring it in the stress-free state, where Av is the stress-induced peak shift, Ii is the piezo- and measuring the indentation crack length when the material spectroscopic tensor and oi is the stress tensor. When the applied contains aresidual stresses field, the value of the residual stresses stress is uniaxial, the above equation reduces to a proportional can be obtained by a non-linear regression of c over p according ity relation between the applied uniaxial stress and the shift of a to Eq (1). selected Raman peak. As a first approximation, we will assume For this purpose, the true fracture toughness of the layer I, that under a hydrostatic stress field h the relation can be written which is also the external layer in the ICI laminate, was measured as by Vicker's indentation technique on the polished top surface of the stress-free material BI with applied loads of 49.05, 68.67, 4v=avOn 98. 10 and 147 15N using a Zwick hardness tester mod. 3212. where Iay is the average piezo-spectroscopic coefficient, Different values of applied load were used to assess the con- nay=3/u lu is the piezo-spectroscopic coefficient measured stancy of the Kle values, which should ensure that the crack under uniaxial stress system developed does not change by increasing the applied load. The formula of Anstis et al. was used for the calculation a wavelength of 488 nm was used as the excitation source of Klc. 22 The Young's modulus of the layer I was calculated Specimens were placed on a mapping device(lateral resolution using the generalization of the Hashin and Strikman bounds for 0.1 um), which allows very precise displacements on the speci three-phase composites set up by Walpole, taking literature men surface. The region of interest was selected using an optical values for the Young's modulus of the three constituent phases microscope. Scattered frequencies were analyzed with a triple AIN=316 GPa,+SiC=440GPa, MoSi2 =440 GPa26)and monochrometer equipped with a charge-coupled device cam- neglecting the contribution of any additional grain boundary era. Bands from an Hg/Ne lamp were used as internal reference phases(see Section 3. 1). The value obtained for the I composi- for spectral calibrations. The recorded spectra were examined tion is 339 GPa. Then, indentation cracks were generated with by fitting with a Gauss-Lorentz mixed function using commer- the same loads on the outer layer of the ICl composite(plane Yz cially available software(LABSPEC 4.02, Horiba/Jobin-Ivon) in Fig. 2). The indentation marks were introduced on the top sur- On the polished cross-section of the ICI composite(plane XY face of the ICI composite, which is under a biaxial stress state, in Fig. 2), linear profiles were collected in steps of 1 um and a and not in its cross-section (plane XY, Fig. 2), where the non- laser spot size of about 1 um(objective lens x100). Five lin- biaxial stress state would have ger nerated an asy symmetric cracks ear profiles were recorded, starting from the conductive layer system. For the calculation of the residual stress, the non-linear (C), as schematically indicated in Fig. 2. With the same condi regression of c over p was performed with commercial math- tions, 7 um x 7 um maps were recorded in the bulk composites ematical software(MATHEMATICA 5.0, Wolfram Research I and BC), which will be considered as zero-stress reference Inc, USA) materials In order to determine the piezo-spectroscopic coefficient 2.3. Measurement of residual stress by raman Hu, calibrations of spectral shift versus externally plez0-spectroscopy stress were carried out using a miniature four-point bending jig connected with a load-cell in which 3 mm x 4 mm x 20mm According to the studies carried out by Grabner, when a bars cut from BC and BI reference materials were stressed material is under stress a shift of its characteristic Raman fluo- The bending jig was fixed in turn to a mapping device(lateral

D. Sciti et al. / Journal of the European Ceramic Society 26 (2006) 3415–3423 3417 2.2. Measurement of residual stress by indentation technique The presence of a residual stress field in a material mod￾ifies the indentation crack length with respect to the length observed in the same stress-free material. The influence of resid￾ual stresses on the length of the indentation cracks is such that the true fracture toughness, KIc, of the material can be calculated by considering the contributions due to the indentation stress and the residual stresses field. For an indentation crack length c in a residual stress field σres, this implies11: KIc = 0.016
E H 1/2 p c3/2 + yσres√c (1) where p is the applied indentation load, E and H the Young’s modulus and the hardness of the material, respectively, and y is a constant related to the aspect ratio of the crack and can be taken as 1.29 for surface half-penny crack in a uniform stress field.21 Vice versa, knowing the true fracture toughness of a material, for example, by measuring it in the stress-free state, and measuring the indentation crack length when the material contains a residual stresses field, the value of the residual stresses can be obtained by a non-linear regression of c over p according to Eq. (1). For this purpose, the true fracture toughness of the layer I, which is also the external layer in the ICI laminate, was measured by Vicker’s indentation technique on the polished top surface of the stress-free material BI with applied loads of 49.05, 68.67, 98.10 and 147.15 N using a Zwick hardness tester mod. 3212. Different values of applied load were used to assess the con￾stancy of the KIc values, which should ensure that the crack system developed does not change by increasing the applied load. The formula of Anstis et al. was used for the calculation of KIc. 22 The Young’s modulus of the layer I was calculated using the generalization of the Hashin and Strikman bounds for three-phase composites set up by Walpole,23 taking literature values for the Young’s modulus of the three constituent phases (AlN = 316 GPa,24 SiC = 440 GPa,25 MoSi2 = 440 GPa26) and neglecting the contribution of any additional grain boundary phases (see Section 3.1). The value obtained for the I composi￾tion is 339 GPa. Then, indentation cracks were generated with the same loads on the outer layer of the ICI composite (plane YZ in Fig. 2). The indentation marks were introduced on the top sur￾face of the ICI composite, which is under a biaxial stress state, and not in its cross-section (plane XY, Fig. 2), where the non￾biaxial stress state would have generated an asymmetric cracks system. For the calculation of the residual stress, the non-linear regression of c over p was performed with commercial math￾ematical software (MATHEMATICA 5.0, Wolfram Research Inc., USA). 2.3. Measurement of residual stress by Raman piezo-spectroscopy According to the studies carried out by Grabner,20 when a material is under stress a shift of its characteristic Raman fluo￾Fig. 2. Schematic of the Raman and indentation experiments carried out on the ICI material. rescence peaks occurs. The relationship is given by the following tensorial expression: ν = Πijσij (2) where ν is the stress-induced peak shift, Πij is the piezo￾spectroscopic tensor and σij is the stress tensor. When the applied stress is uniaxial, the above equation reduces to a proportional￾ity relation between the applied uniaxial stress and the shift of a selected Raman peak. As a first approximation, we will assume that under a hydrostatic stress field σh the relation can be written as: ∆ν = Πavσh (3) where Πav is the average piezo-spectroscopic coefficient, Πav = 3Πu. Πu is the piezo-spectroscopic coefficient measured under uniaxial stress. For the Raman investigation, an Ar-ion laser operating at a wavelength of 488 nm was used as the excitation source. Specimens were placed on a mapping device (lateral resolution 0.1
m), which allows very precise displacements on the speci￾men surface. The region of interest was selected using an optical microscope. Scattered frequencies were analyzed with a triple monochrometer equipped with a charge-coupled device cam￾era. Bands from an Hg/Ne lamp were used as internal reference for spectral calibrations. The recorded spectra were examined by fitting with a Gauss-Lorentz mixed function using commer￾cially available software (LABSPEC 4.02, Horiba/Jobin-Ivon). On the polished cross-section of the ICI composite (plane XY in Fig. 2), linear profiles were collected in steps of 1
m and a laser spot size of about 1
m (objective lens ×100). Five lin￾ear profiles were recorded, starting from the conductive layer (C), as schematically indicated in Fig. 2. With the same condi￾tions, 7
m × 7
m maps were recorded in the bulk composites (BI and BC), which will be considered as zero-stress reference materials. In order to determine the piezo-spectroscopic coefficient, Πu, calibrations of spectral shift versus externally applied stress were carried out using a miniature four-point bending jig connected with a load-cell in which 3 mm × 4 mm × 20 mm bars cut from BC and BI reference materials were stressed. The bending jig was fixed in turn to a mapping device (lateral

3418 D. Sciti er al. / Joumal of the European Ceramic Society 26 (2006)3415-3423 350 (c) Fig 3. Microstructure of the ICI laminate: (a) polished cross-section; (b)detail showing the interface between adjacent layers; (c)enlarged view of the conductive layers; and(d) enlarged view of the insulating layers. Bright contrast particles belong to the MoSi2 phase. resolution of 0.01 um) connected to a personal computer to 170-180 um for the inner layers and 370 um for the outer layers drive highly precise displacements in order to scan the bar (see Table 3). lateral surface. i.e. from the bar edge under tension to the bar In Fig 3c and d, detailed views of the single layers are edge under compression reported. The bright contrast particles are molybdenum disili- cide and the grey regions consist of AIN and Sic phases, not 3. Results distinguishable(even by backscattered imaging)due to the very close atomic number. The MoSi2 particle distribution is homo- 3.1. Microstructure and properties geneous inside each layer. The mean MoSig diameter evaluated by image analysis on polished surface is in the range 0.9-1 um Highly dense materials were obtained after sintering at in both the I and C layers. Some aggregates of MoSiz particles 1850C, confirming the good sinterability of this non-oxide sys- are observed in the C layers due to the higher volumetric frac- tem. The relative densities of the BI and BC laminates were in the tion. Intergranular phases consisting of crystalline YAG and/or range 98-99%. The relative density of ICI laminate after sinter- amorphous silicate phases are present in all the systems. SEM ing was >97%. All the theoretical densities were calculated with observations evidence that the amount of secondary phases in I the rule of mixture, considering the starting nominal composi- layers is considerably lower than in C layers. At the same time, tions. The presence of MoSi2 particles is considered to be the key I layers present a higher level of porosity. These two factors are factor that improves the densification of this composite. MoSi2 again related to the Mosi content and confirm that a higher particles are in fact originally coated with an amorphous silica content of this phase is beneficial for improving the densifica- layer that enhances sintering kinetics acting as grain bound- tion. The microstructural features(not shown) of samples BI ary lubricant and favouring the formation of liquid phase in the and BC are similar to those reported for I and C layers. Despite AIN-Y203-SiO2 system. 6, I7 he lamination technique used, no traces of the junction between The main crystalline phases in the dense materials are the overlapping layers are observed, ie. these layered composites starting ones, as confirmed by previous studies. b, l/ Some resemble bulk materials. The mean MoSiz particle dimensions microstructural features of the ICI composite are illustrated are in the range 0.9-l um, likewise for I and C layers in Fig. 3a-d, showing a polished cross-section. Perfect adhe- As reported in Table 2, the BC laminate is a good conductor sion was found between the layers and no delamination was as a result of the interconnectivity of MoSi2 particles, whilst observed through the composite(Fig. 3b), very likely due to in the BI sample, the amount of electro-conductive phase is too the close similarity of composition of the layers containing the low to reach the percolation limit. As expected, the CTEs of the same starting phases. The final layers'thickness was around two materials are different as a result of the different starting

3418 D. Sciti et al. / Journal of the European Ceramic Society 26 (2006) 3415–3423 Fig. 3. Microstructure of the ICI laminate: (a) polished cross-section; (b) detail showing the interface between adjacent layers; (c) enlarged view of the conductive layers; and (d) enlarged view of the insulating layers. Bright contrast particles belong to the MoSi2 phase. resolution of 0.01
m) connected to a personal computer to drive highly precise displacements in order to scan the bar lateral surface, i.e. from the bar edge under tension to the bar edge under compression. 3. Results 3.1. Microstructure and properties Highly dense materials were obtained after sintering at 1850 ◦C, confirming the good sinterability of this non-oxide sys￾tem. The relative densities of the BI and BC laminates were in the range 98–99%. The relative density of ICI laminate after sinter￾ing was >97%. All the theoretical densities were calculated with the rule of mixture, considering the starting nominal composi￾tions. The presence of MoSi2 particles is considered to be the key factor that improves the densification of this composite. MoSi2 particles are in fact originally coated with an amorphous silica layer that enhances sintering kinetics acting as grain bound￾ary lubricant and favouring the formation of liquid phase in the AlN–Y2O3–SiO2 system.16,17 The main crystalline phases in the dense materials are the starting ones, as confirmed by previous studies.16,17 Some microstructural features of the ICI composite are illustrated in Fig. 3a–d, showing a polished cross-section. Perfect adhe￾sion was found between the layers and no delamination was observed through the composite (Fig. 3b), very likely due to the close similarity of composition of the layers containing the same starting phases. The final layers’ thickness was around 170–180
m for the inner layers and 370
m for the outer layers (see Table 3). In Fig. 3c and d, detailed views of the single layers are reported. The bright contrast particles are molybdenum disili￾cide and the grey regions consist of AlN and SiC phases, not distinguishable (even by backscattered imaging) due to the very close atomic number. The MoSi2 particle distribution is homo￾geneous inside each layer. The mean MoSi2 diameter evaluated by image analysis on polished surface is in the range 0.9–1
m in both the I and C layers. Some aggregates of MoSi2 particles are observed in the C layers due to the higher volumetric frac￾tion. Intergranular phases consisting of crystalline YAG and/or amorphous silicate phases are present in all the systems. SEM observations evidence that the amount of secondary phases in I layers is considerably lower than in C layers. At the same time, I layers present a higher level of porosity. These two factors are again related to the MoSi2 content and confirm that a higher content of this phase is beneficial for improving the densifica￾tion. The microstructural features (not shown) of samples BI and BC are similar to those reported for I and C layers. Despite the lamination technique used, no traces of the junction between overlapping layers are observed, i.e. these layered composites resemble bulk materials. The mean MoSi2 particle dimensions are in the range 0.9–1
m, likewise for I and C layers. As reported in Table 2, the BC laminate is a good conductor as a result of the interconnectivity of MoSi2 particles, whilst in the BI sample, the amount of electro-conductive phase is too low to reach the percolation limit. As expected, the CTEs of the two materials are different as a result of the different starting

D. Sciti et al / Joumal of the European Ceramic Society 26 (2006)3415-3423 Table 2 Experimental values of CTE and electrical resistivity measured on the reference materials BI and BC. Material CTE20-750°C(x10-6/C)CTE20-1000C(×10-6/°C)CTE20-1200°(×10-6/°C resistivity(Scm) Iu(cm GPa -0.90±0.05 composition. The values are determined by the synergetic action of the three phases present. The mechanical properties of the heterogeneous laminated composite will be the subject of future work; however, just as an indication, the 4-pt bending strength of a pressureless sintered material (not cast) with the same composition as bC is about 380 8 and the fracture toughness, measured by indentation, is in the range 3-4 MPam /2. These values can also be considered significant g for the homogenous laminates, since their microstructure is not distinguishable from that of bulk materials. In contrast, the properties of the ICI composite are expected to change significantly due to the presence of the residual stress field 1.5-50 Indentation load, p (N) 3. 2. Lamination theory Fig. 4. Fracture toughness of the stress-free bl material as a function of the indentation load The values of residual stresses in a multilayer composite can ature, with 1200C being a quite common choice. However, be a priori calculated by considering the elastic constants, the in the present case, each individual layer contains MoSi2 for thermal expansion coefficients and the thickness of the different layers along with the stress-free temperature by the equations of which a brittle-to-ductile transition is known to occur at about the lamination theory. For a symmetrical composite with 2n+1 1000C. Taking 1000oC as the stress-free temperature,the alternated layers, the equations which should be solved in order calculated residual stresses in the ICI composite are -11MPa to obtain the values of the residual stresses are n the I layers and +200 MPa in the C layers (Table 3). Er oi+aj Ar=constant (4) 33. Indentation technique 0 The plot of true fracture toughness, Klc, of the stress-free bI material as a function of the indentation load is shown in where ai is the stress developed in the layer of thickness t. Fig. 4. As can be seen, the value of fracture toughness is almost Respectively, ai, Ei and vi are the thermal expansion coeffi- constant in the range of applied loads used, indicating that the ents, the Youngs modulus and the Poisson ratio of the ith layer. crack system developed under the indentation marks does not ATis the temperature range over which elastic stress develops significantly change by increasing the applied load. The crack due to thermal strain mismatch. The values of Youngs modu- lengths as a function of the indentation load in the stress-free BI lus were calculated for I and C layers using the generalization material and in the stressed outer I layer of the ICI composite of the Hashin and Strikman bounds for three phase composites are shown in Fig. 5. The maximum crack length measured in set up by Walpole, as already mentioned. The Poissons ratios the ICI composite was less than 220 um, i. e shallower than the were determined with the rule of mixture on the basis of starting external layer thickness. Applying Eq.(1), the regression of the ompositions. The CTEs of the BI and BC materials are those indentation crack length on indentation load in the stressed layer reported in Table 2 for selected temperatures. A crucial parame- gives-79 MPa as the value of residual stress in the outer I layer. ter for the theoretical calculations is the stress-free temperature, As for the precision of this value, the standard error of +4 MPa i.e. the temperature below which stresses are accumulated elas- calculated by the regression can be assumed. tically. This temperature is difficult to determine experimentally As shown in Section 3. 2. the residual stress value in the con- Usually, it is taken somewhat lower than the sintering temper- ductive C layers, d res, can be calculated using the equilibrium Table 3 Comparison of residual stress values with different techniques and according to theoretical models Layer Layer thickness(um) Indentation technique(MPa) Theoretical(stress free ee T=1000°C)(MPa) T=750°C)(MPa) uter370±2, Inner175±5 -(79±4) C +(102±15) +(140±4)2 By equilibrium criterion

D. Sciti et al. / Journal of the European Ceramic Society 26 (2006) 3415–3423 3419 Table 2 Experimental values of CTE and electrical resistivity measured on the reference materials BI and BC. Material CTE 20–750 ◦C (×10−6/ ◦C) CTE 20–1000 ◦C (×10−6/ ◦C) CTE 20–1200 ◦C (×10−6/ ◦C) Electrical resistivity (
cm) Πu (cm GPa−1) BI 5.46 5.90 6.12 5 × 103 −1.80 ± 0.07 BC 6.14 6.54 6.77 3 × 10−3 −0.90 ± 0.05 composition. The values are determined by the synergetic action of the three phases present. The mechanical properties of the heterogeneous laminated composite will be the subject of future work; however, just as an indication, the 4-pt bending strength of a pressureless sintered material (not cast) with the same composition as BC is about 380 MPa18 and the fracture toughness, measured by indentation, is in the range 3–4 MPa m1/2. These values can also be considered significant for the homogenous laminates, since their microstructure is not distinguishable from that of bulk materials. In contrast, the properties of the ICI composite are expected to change significantly due to the presence of the residual stress field. 3.2. Lamination theory The values of residual stresses in a multilayer composite can be a priori calculated by considering the elastic constants, the thermal expansion coefficients and the thickness of the different layers along with the stress-free temperature by the equations of the lamination theory.3 For a symmetrical composite with 2n + 1 alternated layers, the equations which should be solved in order to obtain the values of the residual stresses are: εi = 1 − νi Ei σi + αiT = constant (4) Σiσiti = 0 (5) where σi is the stress developed in the layer of thickness ti. Respectively, αi, Ei and νi are the thermal expansion coeffi- cients, the Young’s modulus and the Poisson ratio of the ith layer. T is the temperature range over which elastic stress develops due to thermal strain mismatch. The values of Young’s modu￾lus were calculated for I and C layers using the generalization of the Hashin and Strikman bounds for three phase composites set up by Walpole, as already mentioned. The Poisson’s ratios were determined with the rule of mixture on the basis of starting compositions. The CTEs of the BI and BC materials are those reported in Table 2 for selected temperatures. A crucial parame￾ter for the theoretical calculations is the stress-free temperature, i.e. the temperature below which stresses are accumulated elas￾tically. This temperature is difficult to determine experimentally. Usually, it is taken somewhat lower than the sintering temper￾Fig. 4. Fracture toughness of the stress-free BI material as a function of the indentation load. ature, with 1200 ◦C being a quite common choice. However, in the present case, each individual layer contains MoSi2 for which a brittle-to-ductile transition is known to occur at about 1000 ◦C.26 Taking 1000 ◦C as the stress-free temperature, the calculated residual stresses in the ICI composite are −113 MPa in the I layers and +200 MPa in the C layers (Table 3). 3.3. Indentation technique The plot of true fracture toughness, KIc, of the stress-free BI material as a function of the indentation load is shown in Fig. 4. As can be seen, the value of fracture toughness is almost constant in the range of applied loads used, indicating that the crack system developed under the indentation marks does not significantly change by increasing the applied load. The crack lengths as a function of the indentation load in the stress-free BI material and in the stressed outer I layer of the ICI composite are shown in Fig. 5. The maximum crack length measured in the ICI composite was less than 220
m, i.e. shallower than the external layer thickness. Applying Eq. (1), the regression of the indentation crack length on indentation load in the stressed layer gives −79 MPa as the value of residual stress in the outer I layer. As for the precision of this value, the standard error of ±4 MPa calculated by the regression can be assumed. As shown in Section 3.2, the residual stress value in the con￾ductive C layers, σC res, can be calculated using the equilibrium Table 3 Comparison of residual stress values with different techniques and according to theoretical models Layer Layer thickness (
m) σres by Raman spectroscopy (MPa) Indentation technique (MPa) Theoretical (stress free T = 1000 ◦C) (MPa) Theoretical (stress free T = 750 ◦C) (MPa) I Outer 370 ± 2, inner 175 ± 5 −(62 ± 3) −(79 ± 4) −113 −85 C 178 ± 10 +(102 ± 15) +(140 ± 4)a +200 +150 a By equilibrium criterion.

D. Sciti er al. / Joumal of the European Ceramic Society 26 (2006)3415-3423 Applied load(N) aman shift(cm-1) Fig. 5. Crack lengths as a function of the indentation load in the stress-free B material and in the stressed outer I layer of the ICI composite. criterion: ares =-cores where dres is the residual stress in the I layer and tot and fot are the total thickness of I and C layers, respectively. With a o 79 MPa, the calculated value of ores is +140 MPa(Table 3) C layer: 797.3 cm 3.4. Raman spectroscopy The Raman spectra of these materials show the typical bands aman shift(cm1) of hexagonal AIN at 656cm- 27B-Sic at 796cm-128and tetragonal MoSi2 at 430cm-129(Fig. 6). Since the peaks of Fig. 7. Raman shift of the B-SiC Raman band (796cm )in(a) I layers of ICI AIN and MoSi, had quite a low intensity, the B-SiC Raman in respect with BI reference material and (b) in C layers of ICI in respect with the reference bc material band 796 cm was chosen as stress sensor due to its sharp ness. A representative Raman shift for the systems considered of their respective composite. These values are also different is illustrated in Fig. 7 a and b, where the Sic peak shifts of I from the piezo-spectroscopic coefficient found for B-SiC single and C layers of the ICI laminate are compared to the SiC peak crystal. 30 Using these values, the residual stresses profile of the positions in the reference BI and BC materials, respectively multilayer composite was calculated according to Eq (2). For the The calibration data for the Sic peak relative to the BI and sake of clarity, in the materials of the present work the Raman BC samples are reported in Fig. 8a and b. The slopes of the laser has a very shallow penetration(see Section 3.5).There regression lines, i.e. the piezo-spectroscopic coefficients, are fore, the line profiles in the ICI laminate were collected from 1.79 and -0.88 cm" /GPa for material BI and BC, respec- a near-edge region where the stress state in non- biaxial due to tively Table 2). These values are significantly different from the presence of a normal component. 5.I This component, which each other confirming that they are affected by the composition has the opposite sign of the stress state inside the layer, vanishes going toward the core of the layer on a length comparable to the 5,31 amely 180 um. Owing to this triaxial stress field, the constant Ilay in Eq. (2)was taken as three times llu and not two as in the case of a pure biaxial stress field The measured Raman shift in the I or C layers of the SiC phase is the result of both a lamination-induced macro-stres and a micro-stress due to the different ctes of the constituent phases present in each individual layer(namely, AIN, SiC and MoSi2). aving measured th the Sic phase in bulk materials with composition C and l,1.e the same composition of the layers forming the laminate com- ed in the laminat is uniquely caused by the residual stresses induced by the lami- nation process. The resulting stress profile, averaged on the five Raman spectrum(cm-1) scanning lines, is reported in Fig. 9. As expected, the insulat Fig. 6. Raman spectrum of the laminate. ing layers are under compressive stress while the conductive

3420 D. Sciti et al. / Journal of the European Ceramic Society 26 (2006) 3415–3423 Fig. 5. Crack lengths as a function of the indentation load in the stress-free BI material and in the stressed outer I layer of the ICI composite. criterion: σC res = −t I tot t C tot σI res (6) where σI resis the residual stress in the I layer and t I tot and t C tot are the total thickness of I and C layers, respectively. With a σI res of −79 MPa, the calculated value of σC res is +140 MPa (Table 3). 3.4. Raman spectroscopy The Raman spectra of these materials show the typical bands of hexagonal AlN at 656 cm−1, 27 -SiC at 796 cm−1 28 and tetragonal MoSi2 at 430 cm−1 29 (Fig. 6). Since the peaks of AlN and MoSi2 had quite a low intensity, the -SiC Raman band 796 cm−1 was chosen as stress sensor due to its sharp￾ness. A representative Raman shift for the systems considered is illustrated in Fig. 7 a and b, where the SiC peak shifts of I and C layers of the ICI laminate are compared to the SiC peak positions in the reference BI and BC materials, respectively. The calibration data for the SiC peak relative to the BI and BC samples are reported in Fig. 8a and b. The slopes of the regression lines, i.e. the piezo-spectroscopic coefficients, are −1.79 and −0.88 cm−1/GPa for material BI and BC, respec￾tively (Table 2). These values are significantly different from each other confirming that they are affected by the composition Fig. 6. Raman spectrum of the laminate. Fig. 7. Raman shift of the -SiC Raman band (796 cm−1) in (a) I layers of ICI in respect with BI reference material and (b) in C layers of ICI in respect with the reference BC material. of their respective composite. These values are also different from the piezo-spectroscopic coefficient found for -SiC single￾crystal.30 Using these values, the residual stresses profile of the multilayer composite was calculated according to Eq.(2). For the sake of clarity, in the materials of the present work the Raman laser has a very shallow penetration (see Section 3.5). There￾fore, the line profiles in the ICI laminate were collected from a near-edge region where the stress state in non-biaxial due to the presence of a normal component.5,31 This component, which has the opposite sign of the stress state inside the layer, vanishes going toward the core of the layer on a length comparable to the layer thickness,5,31 namely 180
m. Owing to this triaxial stress field, the constant Πav in Eq. (2) was taken as three times Πu and not two as in the case of a pure biaxial stress field. The measured Raman shift in the I or C layers of the SiC phase is the result of both a lamination-induced macro-stress and a micro-stress due to the different CTEs of the constituent phases present in each individual layer (namely, AlN, SiC and MoSi2).32 However, having measured the piezo-coefficient of the SiC phase in bulk materials with composition C and I, i.e. the same composition of the layers forming the laminate com￾posite, the SiC peak shift measured in the laminate composite is uniquely caused by the residual stresses induced by the lami￾nation process. The resulting stress profile, averaged on the five scanning lines, is reported in Fig. 9. As expected, the insulat￾ing layers are under compressive stress while the conductive

D. Sciti et al / Joumal of the European Ceramic Society 26 (2006)3415-3423 3421 7980 =096y=7 796.38·0.00179*x phases, particularly abundant in the conductive layers, and the presence of the microscopic stress field due to microstructural compositionthat is superimposed to the lamination-induced macroscopic one. The average of the values reported in the plot of Fig 9 are-62 MPa in the insulating layer and +102 MPa in the conductive one(Table 3). For the precision of these mea surements, the standard error of the experimental values, either in I or in C layers, was considered (Table 3) 3.5. Comparison of the results 200-150100-500501001 The experimental results and theoretical values are summa pplied stress, MPa rized in Table 3. Despite the different technique used, the exper imental values obtained for the layers in compression agreed 798,0 P=0.92;y=795.40075 both in magnitude and sign of the residual stresses, with a dif- 7975 ference between them of about 25%0. For sake of precision, however, it must be clearly stated that results by the indenta 7970 top surface of the multilayer composite where the stress felf a tion technique are based on measurements carried out on purely biaxial, whilst Raman measurements were carried out on the cross-section where a normal component is present, as previ ously mentioned. In the materials of the present work the Raman ponent, has a strong infuence on the estimated values of residual stresses. This could be the justification for the slightly higher 79400150-10050050100150200 values measured by indentation with respect to those measured by Raman. The penetration depth of the laser in these materi als can be evaluated to be less than a few micrometers due to Fig.8.Calibration data for the SiC peak in(a) BI and(b)BC reference material. the strong absorption of the reinforcing MoSi2 phase. This is The slopes of the regression lines represent the piezo-spectroscopic coefficients, based on the empirical observation that when the laser scans the Mosi2 particles, the ain and Sic Raman peaks are greatly layers are under tensile stress. The plot also clearly evidences reduced. 32 As a very rough approximation, one can estimate that in the conductive C layers, the data are quite scattered with the mean penetration depth of the Raman laser in our compos- respect to the data in the I layers. This is very likely due to ites as given by 1/(sn), where s is the cross-section between the higher volumetric fraction of the MoSi2 phase which is a the laser and the particles and n is the number of particles per very strong absorbing phase as a previous study confirmed. 32 unit volume Using the value of particles radius of 0.5 um and Further sources of experimental error can be, for example, the volumetric fraction of the reinforcing phase, the penetratic microstructural inhomogeneities, such as the presence of silicate depth is about 0.6 um in the conductive layer and 1.3 um in the insulating layers. Even if this were a very rough approxima tion, it would be difficult to think that the Raman laser could go deeper than the layer thickness, about 180 Hm materials, Raman mal component vanishes. Therefore, in these spectroscopy investigates the stress at the very surface of each la The theoretical estimations were in any case higher than the experimental results(Table 3). For the Raman measurements, again the main reason could be the fact that the theoretical mod els consider the stress in the bulk of the laminate. where the stress field is purely biaxial, whilst Raman measurements are affected by the normal component. Further sources of discrepancy between theoretical and experimental values could stem from the values used for the calculation themselves. In fact. the values of elastic constants were approximated without taking into account the presence of 0.8-0.6-04-02002040.60.81121.41.61.8 nases in the microstructure of both C and I layers. In normalized distance. x/ addition, the choice of the stress-free temperature is somehow Fig 9. Residual stress profile in the ICI laminat ambiguous. The presence of amorphous silicate phases could

D. Sciti et al. / Journal of the European Ceramic Society 26 (2006) 3415–3423 3421 Fig. 8. Calibration data for the SiC peak in (a) BI and (b) BC reference material. The slopes of the regression lines represent the piezo-spectroscopic coefficients, Πu. layers are under tensile stress. The plot also clearly evidences that in the conductive C layers, the data are quite scattered with respect to the data in the I layers. This is very likely due to the higher volumetric fraction of the MoSi2 phase which is a very strong absorbing phase as a previous study confirmed.32 Further sources of experimental error can be, for example, microstructural inhomogeneities, such as the presence of silicate Fig. 9. Residual stress profile in the ICI laminate. phases, particularly abundant in the conductive layers, and the presence of the microscopic stress field due to microstructural composition32 that is superimposed to the lamination-induced macroscopic one. The average of the values reported in the plot of Fig. 9 are −62 MPa in the insulating layer and +102 MPa in the conductive one (Table 3). For the precision of these mea￾surements, the standard error of the experimental values, either in I or in C layers, was considered (Table 3). 3.5. Comparison of the results The experimental results and theoretical values are summa￾rized in Table 3. Despite the different technique used, the exper￾imental values obtained for the layers in compression agreed both in magnitude and sign of the residual stresses, with a dif￾ference between them of about 25%. For sake of precision, however, it must be clearly stated that results by the indenta￾tion technique are based on measurements carried out on the top surface of the multilayer composite where the stress field is purely biaxial, whilst Raman measurements were carried out on the cross-section where a normal component is present, as previ￾ously mentioned. In the materials of the present work the Raman laser has a very shallow penetration, therefore, the normal com￾ponent, has a strong influence on the estimated values of residual stresses. This could be the justification for the slightly higher values measured by indentation with respect to those measured by Raman. The penetration depth of the laser in these materi￾als can be evaluated to be less than a few micrometers, due to the strong absorption of the reinforcing MoSi2 phase. This is based on the empirical observation that when the laser scans the MoSi2 particles, the AlN and SiC Raman peaks are greatly reduced.32 As a very rough approximation, one can estimate the mean penetration depth of the Raman laser in our compos￾ites as given by 1/(sn), where s is the cross-section between the laser and the particles and n is the number of particles per unit volume. Using the value of particles radius of 0.5
m and the volumetric fraction of the reinforcing phase, the penetration depth is about 0.6
m in the conductive layer and 1.3
m in the insulating layers. Even if this were a very rough approxima￾tion, it would be difficult to think that the Raman laser could go deeper than the layer thickness, about 180
m, where the nor￾mal component vanishes. Therefore, in these materials, Raman spectroscopy investigates the stress at the very surface of each layer. The theoretical estimations were in any case higher than the experimental results (Table 3). For the Raman measurements, again the main reason could be the fact that the theoretical mod￾els consider the stress in the bulk of the laminate, where the stress field is purely biaxial, whilst Raman measurements are affected by the normal component. Further sources of discrepancy between theoretical and experimental values could stem from the values used for the calculation themselves. In fact, the values of elastic constants were approximated without taking into account the presence of remnant phases in the microstructure of both C and I layers. In addition, the choice of the stress-free temperature is somehow ambiguous. The presence of amorphous silicate phases could

D. Sciti et al. Journal of the European Ceramic Society 26(2006)3415-3423 have contributed to lower the chosen stress-free temperature of 3. Chartier, T, Merle, D. and Besson, J. L, Laminar ceramic composites 1000C, for example. Moreover, the brittle-to-ductile transition J. Eur Ceram.Soc.,1995,15,101-107 temperature of the MoSi2, which was taken as 1000C, can be as 4. Sergo, V. Room-temperature aging of laminate composites of alu- low as 750. if the material has been previously prestrained.33 mina/ mols%o-yttria-stabilized tetragonal zirconia polycrystals.J.Am Ceram.Soc.,2004,87,247-253. This could be the case, since the MoSi2 particles were ball mille 5. Sergo, V, Lipkin, D. M, de Portu, G. and Clarke, D. R, Edge stresses for 120 h as indicated in the Experimental section Inserting this in alumina/zirconia laminates. J. Am. Ceram. Soc.. 1997.. 1633-1638. latter value of stress-free temperature in Eq. (4), one obtains 6. Tarlazzi, A,Roncari,E.Pinasco,P,Guicciardi,S,Melandri,Cand de 85 MPa and +150 MPa as values of residual stresses in the l Portu, G, Tribological behaviour of Al, O3/ZrOz-ZrOz laminated com- and C layer, respectively, which are closer to the experimental 7. Immelmann, S, Welle, E. and Reimers, W. x-ray residual stress analy values, especially to the value obtained by indentation sis on machined and tempered HPSN-ceramics. Mater Sci. Eng, 1997, Lastly, it must be said that the difference between the the- A238,287-292 retical estimations and the experimental values could have an 8. Soares, M.R. Belmontea, M and Silva, R E, Low incident angle and explanation that is merely technological. In fact, in the calcu lassical X-ray diffraction analysis of residual stresses in diamond coated lation of the theoretical residual stresses it was not considered SiN4J. Appl. Phys,2003,94.5633-5638. that the different layers could have different shrinkages during [9].Kupperman, D.S. Singh, J.P,Faber,.and Hitterman,RL,Application of neutron diffraction to the characterization of residual thermal strains sintering. This shrinkage mismatch generates residual stresses in YBa2 Cu3 Ag. J. Appl. Phys., 1989, 66, 3396-3398 in the multilayer composite that overlap the thermal residual 10. Lukas, P, Vrana, M, Mikula, P, Vleugels, J, Anne, G and Van der Biest, stresses. It is likely that in our system the shrinkage mismatch O, Neutron diffraction study of the phase-specific stresses in graded alumina/zirconia ceramics. Physica B. 2004. 350. e517-e520 was such as to reduce the residual thermal stresses calculated by 11. Widjaja. S, Jakus, K, Ritter, J.E.Atri,Rand Battacharaya, S,Resid- ual surface stress by localized contact creep. J. Mater. Res, 1997, 12, 210-217 4. Conclusions 12. Zhang, T.-Y, Chen, L.-Q. and Fu, R. Measurements of residual stresses in thin films deposited on silicon wafers by indentation fracture. Acta Mater,1999,47,3869-3878. Dense delamination-free multilayer composites were pro- 13. De Wolf, I Micro-Raman spectroscopy to study local mechanical stress duced in the AIN-SiC-MoSi2 system. Heterogeneous samples in silicon integrated circuits. Semicond. Sci. Technol, 1996, 11, 139-154 ere produced, by alternating insulating and electrically cor 14. Dietrich, B and Dombrowski, K. F, Experimental challenges of stress ductive layers. Raman spectroscopy and an indentation tech- measurements with resonant micro-Raman spectroscopy. J. Raman Spec- osc.,1999,30,893-897 nique were employed to evaluate the residual stresses among 15 Micele. L. Pezzotti, G. de Portu, G. and Sekiguchi, Y, Measurements adjacent layers. Compressive residual stresses were found in of residual stress distributions in Al]O3/3Y-TZP multilayered composites the insulating layers whilst conductive layers were subjected to by fluorescence and Raman microprobe piezo-spectroscopy. Acta Mate 2005.53.1511-1520. and magnitude of the residual stresses values. The compari- 16. sect, D, Guicciand s meland c and Belost,, High-temperature son between experimental results and theoretical expectations 03,86,1720 provided evidence that the results of the theoretical model are 17. Sciti, D. and Guicciardi, S, Microstructure and mechanical properties greatly influenced by parameter values that are difficult to obtain ternary electroconductive ceramics. J. Mater. Res, 2004, 19, xperimentally, like the stress-free temperature. Moreover, the 3343-3352 18. Krnel. K. Sciti D. and Bellosi. A.J. Eur Cera. Soc. 2003. 23 3135 different shrinkages of the layers with different compositions 19. Roncari, E, Pinasco, P, Nagliati, M. and Sciti, D. Tape casting of could have superimposed sintering-derived residual stresses to AIN-SIC-Mo opposites. J. Eur Ceram. Soc., 2004, 24, 2303- lamination process-derived stresses. The presence of a compres sive residual stress in the external layer of the heterogeneous 20 Grabner, L, Spectroscopic technique for the measurement of residual aterial is expected to improve the mechanical properties of this system. 21. Murakami. Y. ed.. Stress Intensity Factors Handbook. Vol. /. Pergamon Press, Oxford, 1987, p. 42 22. Anstis. G. R. Chantiful, P, Lawn, B. R. and Marshall, D. B, J.A. Acknowledgements Ceram.Soc.,1981,64,533 23. Walpole, L J, On bounds for the overall elastic moduli of inhomoge- The two authors( D.S. and s.G.) gratefully acknowledge the 24. Talwar, D N, Sofranko, D, Mooney, Cand Tallo, S,Elastic,structural financial support of the JSPS (Japan Society for the Promotion bonding and defect properties of zinc-blende Bn, AIN, GaN, InN and of Science) during their stay at Rin in Kyoto J their alloys. Mater. Sci. Eng, 2002, B90, 269-277 25. Shackelford, J. F. and Alexander. W. ed, CRC Materials Science and Engineering Handbook. CRC Press, Boca Raton, 2001 Refe 26. Jang, Y.L. and Lavernia, E.J. Review, processing of molybdenum dis- icide.J. Mater:Sci,1994,29,2557-2571 I.Rao,M.P and Lange, F. F, Factors affecting threshold strength in lam- 27. Ichimaru, H. and Pezzotti, G, Raman microprobe mapping of residual aining thin compressive layers. J. Am. Ceram. So and bridging stress fields in AIN ceramics. Mater. Sci. Eng, 2002, A326 2. Requena, J, Moreno, R and Moya, J.S., Alumina and zirconia 28. Olego, D. and Cardona, M, Temperature dependance of the optical multilayer composites obtained by slip casting. J. Am. Ceram. Soc., 1989 phonons and traverse effective charge in 3C-SiC. Phys. Rev. B, 1982

3422 D. Sciti et al. / Journal of the European Ceramic Society 26 (2006) 3415–3423 have contributed to lower the chosen stress-free temperature of 1000 ◦C, for example. Moreover, the brittle-to-ductile transition temperature of the MoSi2, which was taken as 1000 ◦C, can be as low as 750 ◦C if the material has been previously prestrained.33 This could be the case, since the MoSi2 particles were ball milled for 120 h as indicated in the Experimental section. Inserting this latter value of stress-free temperature in Eq. (4), one obtains −85 MPa and +150 MPa as values of residual stresses in the I and C layer, respectively, which are closer to the experimental values, especially to the value obtained by indentation. Lastly, it must be said that the difference between the the￾oretical estimations and the experimental values could have an explanation that is merely technological. In fact, in the calcu￾lation of the theoretical residual stresses it was not considered that the different layers could have different shrinkages during sintering. This shrinkage mismatch generates residual stresses in the multilayer composite that overlap the thermal residual stresses. It is likely that in our system the shrinkage mismatch was such as to reduce the residual thermal stresses calculated by the lamination theory. 4. Conclusions Dense delamination-free multilayer composites were pro￾duced in the AlN–SiC–MoSi2 system. Heterogeneous samples were produced, by alternating insulating and electrically con￾ductive layers. Raman spectroscopy and an indentation tech￾nique were employed to evaluate the residual stresses among adjacent layers. Compressive residual stresses were found in the insulating layers whilst conductive layers were subjected to tensional stresses. The experimental techniques agreed in sign and magnitude of the residual stresses values. The compari￾son between experimental results and theoretical expectations provided evidence that the results of the theoretical model are greatly influenced by parameter values that are difficult to obtain experimentally, like the stress-free temperature. Moreover, the different shrinkages of the layers with different compositions could have superimposed sintering-derived residual stresses to lamination process-derived stresses. The presence of a compres￾sive residual stress in the external layer of the heterogeneous material is expected to improve the mechanical properties of this system. Acknowledgements The two authors (D.S. and S.G.) gratefully acknowledge the financial support of the JSPS (Japan Society for the Promotion of Science) during their stay at RIN in Kyoto (J). References 1. Rao, M. P. and Lange, F. F., Factors affecting threshold strength in lam￾inar ceramic containing thin compressive layers. J. Am. Ceram. Soc., 2002, 85, 1222–1228. 2. Requena, J., Moreno, R. and Moya, J. S., Alumina and alumina/zirconia multilayer composites obtained by slip casting. J. Am. Ceram. Soc., 1989, 72, 1511–1513. 3. Chartier, T., Merle, D. and Besson, J. L., Laminar ceramic composites. J. Eur. Ceram. Soc., 1995, 15, 101–107. 4. Sergo, V., Room-temperature aging of laminate composites of alu￾mina/3 mol%-yttria-stabilized tetragonal zirconia polycrystals. J. Am. Ceram. Soc., 2004, 87, 247–253. 5. Sergo, V., Lipkin, D. M., de Portu, G. and Clarke, D. R., Edge stresses in alumina/zirconia laminates. J. Am. Ceram. Soc., 1997, 80, 1633–1638. 6. Tarlazzi, A., Roncari, E., Pinasco, P., Guicciardi, S., Melandri, C. and de Portu, G., Tribological behaviour of Al2O3/ZrO2–ZrO2 laminated com￾posites. Wear, 2000, 244, 29–40. 7. Immelmann, S., Welle, E. and Reimers, W., X-ray residual stress analy￾sis on machined and tempered HPSN-ceramics. Mater. Sci. Eng., 1997, A238, 287–292. 8. Soares, M. R., Belmontea, M. and Silva, R. F., Low incident angle and classical X-ray diffraction analysis of residual stresses in diamond coated Si3N4. J. Appl. Phys., 2003, 94, 5633–5638. [9]. Kupperman, D. S., Singh, J. P., Faber, J. and Hitterman, R. L., Application of neutron diffraction to the characterization of residual thermal strains in YBa2Cu3O7 − x/Ag. J. Appl. Phys., 1989, 66, 3396–3398. 10. Luka´s, P., Vr ˇ ana, M., Mikula, P., Vleugels, J., Ann ´ e, G. and Van der Biest, ´ O., Neutron diffraction study of the phase-specific stresses in graded alumina/zirconia ceramics. Physica B, 2004, 350, e517–e520. 11. Widjaja, S., Jakus, K., Ritter, J. E., Atri, R. and Battacharaya, S., Resid￾ual surface stress by localized contact creep. J. Mater. Res., 1997, 12, 210–217. 12. Zhang, T.-Y., Chen, L.-Q. and Fu, R., Measurements of residual stresses in thin films deposited on silicon wafers by indentation fracture. Acta Mater., 1999, 47, 3869–3878. 13. De Wolf, I., Micro-Raman spectroscopy to study local mechanical stress in silicon integrated circuits. Semicond. Sci. Technol., 1996, 11, 139–154. 14. Dietrich, B. and Dombrowski, K. F., Experimental challenges of stress measurements with resonant micro-Raman spectroscopy. J. Raman Spec￾trosc., 1999, 30, 893–897. 15. Micele, L., Pezzotti, G., de Portu, G. and Sekiguchi, Y., Measurements of residual stress distributions in Al2O3/3Y-TZP multilayered composites by fluorescence and Raman microprobe piezo-spectroscopy. Acta Mater., 2005, 53, 1511–1520. 16. Sciti, D., Guicciardi, S., Melandri, C. and Bellosi, A., High-temperature resistant composites in the AlN–SiC–MoSi2 system. J. Am. Ceram. Soc., 2003, 86, 1720. 17. Sciti, D. and Guicciardi, S., Microstructure and mechanical properties of novel ternary electroconductive ceramics. J. Mater. Res., 2004, 19, 3343–3352. 18. Krnel, K., Sciti, D. and Bellosi, A., J. Eur. Ceram. Soc., 2003, 23, 3135. 19. Roncari, E., Pinasco, P., Nagliati, M. and Sciti, D., Tape casting of AlN–SiC–MoSi2 composites. J. Eur. Ceram. Soc., 2004, 24, 2303– 2311. 20. Grabner, L., Spectroscopic technique for the measurement of residual stress in sintered Al2O3. J. Appl. Phys., 1978, 49, 580. 21. Murakami, Y. ed., Stress Intensity Factors Handbook, Vol. 1. Pergamon Press, Oxford, 1987, p. 42. 22. Anstis, G. R., Chantiful, P., Lawn, B. R. and Marshall, D. B., J. Am. Ceram. Soc., 1981, 64, 533. 23. Walpole, L. J., On bounds for the overall elastic moduli of inhomoge￾neous systems. J. Mech. Phys. Solids, 1966, 14, 151. 24. Talwar, D. N., Sofranko, D., Mooney, C. and Tallo, S., Elastic, structural, bonding and defect properties of zinc-blende BN, AlN, GaN, InN and their alloys. Mater. Sci. Eng., 2002, B90, 269–277. 25. Shackelford, J. F. and Alexander, W., ed., CRC Materials Science and Engineering Handbook. CRC Press, Boca Raton, 2001, p. Fl. 26. Jang, Y.-L. and Lavernia, E. J., Review, processing of molybdenum dis￾ilicide. J. Mater. Sci., 1994, 29, 2557–2571. 27. Ichimaru, H. and Pezzotti, G., Raman microprobe mapping of residual and bridging stress fields in AlN ceramics. Mater. Sci. Eng., 2002, A326, 261–269. 28. Olego, D. and Cardona, M., Temperature dependance of the optical phonons and traverse effective charge in 3C–SiC. Phys. Rev. B, 1982, 25, 3889–3896

D. Sciti et al. Jounal of the European Ceramic Sociery 26(2006)3415-3423 29. Srinivas, G. and Vankar, V. D, Raman spectroscopy of polycrystalline ements analysis of stresses in joints. J. Am. Ceram. Soc., 1987, 70 molybdenum silicide thin 30. Colombi, C. L, Gregori, specroscopy: a materials G, Lugh ater.lett,1997,30,209-215. hi, V, Rossi, A. and Sergo, V, Piezo- [32]. Citi, D, Guicciardi, S, Celotti, G, Tochino, S and Pezzotti, G, Analysis perspectiv of residual stresses in novel ele petroso.,1999,2,243-272 2006,82(2),317-324 31. Kirchner, H, Conway, J. and Segall, A. E, Effect of joint thickness and 33 Gibala, R, Chang, H and Czamik, C M, Mater. Res. Soc. Symp. Proc. residual stresses on the properties of ceramic adhesive joints: I, finite 1994,322,175

D. Sciti et al. / Journal of the European Ceramic Society 26 (2006) 3415–3423 3423 29. Srinivas, G. and Vankar, V. D., Raman spectroscopy of polycrystalline molybdenum silicide thin films. Mater. Lett., 1997, 30, 209–215. 30. Colombi, C. L., Gregori, G., Lughi, V., Rossi, A. and Sergo, V., Piezo￾specroscopy: a materials science perspective. Recent Res. Dev. Appl. Spectrosc., 1999, 2, 243–272. 31. Kirchner, H., Conway, J. and Segall, A. E., Effect of joint thickness and residual stresses on the properties of ceramic adhesive joints: I, finite elements analysis of stresses in joints. J. Am. Ceram. Soc., 1987, 70, 104–109. [32].Sciti, D., Guicciardi, S., Celotti, G., Tochino, S. and Pezzotti, G., Analysis of residual stresses in novel electroconductive composites. Appl. Phys. A, 2006, 82(2), 317–324. 33. Gibala, R., Chang, H. and Czamik, C. M., Mater. Res. Soc. Symp. Proc., 1994, 322, 175