J Mater Sci(2006)41:7425-7436 DOI10.1007/10853-0060808y Oxide ceramic laminates with highly textured a-alumina interlayers: I. Texture control and laminate formation Ming wei· Dan Zhi· David G. Brandon Received: 18 August 2005/ Accepted: 29 November 2005/Published online: 20 September 2006 +Business media. Lle Abstract Three kinds of texture-reinforced oxide component, could be used to tailor electronic and ceramic laminates with strongly bonded interfaces have structural properties and obtain some of the singl been fabricated. All three were based on highly textured crystal anisotropy. Such texture could be used to a-alumina interlayers, but each with a different oxide reduce thermal expansion mismatch, and hence the laminate matrix and a correspondingly different ther- risk of microcracking 1, while texture control offers a mal mismatch between the textured interlayers and the useful alternative to aligned second-phase reinforce- matrix. Alginate-based, aqueous gel casting was used to ment [2]. Most recently, it has been reported that the produce a flexible tape for all the compositions studied. formation of texture in a-alumina can effectively lower The highly textured alumina(TA) interlayers were the residual stress and make the stress distribution derived from gel-cast tapes containing aligned alumina narrower 3]. In general, processing to develop texture seed platelets. It has been found that the strongest in ceramics differs significantly from the conventional texture derived from g1 vol of the seed platelets in powder processing of dense, fine-grained and equiaxed the gelled precursor tape. Using the March-Dollase microstructures. Textured ceramics have been fabri model for texture analysis, the oriented volume frac- cated by several techniques: sinter-forging [4, hot- tion in the highly textured interlayers was estimated to pressing [5], slip-casting [6], gel-casting in a magnetic vary from 60 to 80% field 1,7,8, and, most commonly, by templating using seeded grain growth [2, 9-13]. Textured micro- tructures have been reported in a single-phase iron titanate ceramic laminate, containing a predetermine Introduction configuration of the textured and non-textured layer In a conventional ceramic the anisotropic single crystal The design of ceramic composites with layered properties are averaged in the isotropic, randomly macrostructures is also receiving considerable research oriented polycrystal. Texturing a polycrystalline attention because they exhibit decreased sensitivity material, by controlling the orientation distribution of surface defects and have demonstrated non-cata- crystallites with respect to the coordinate system of a strophic failure in some laminate systems [14-16 key feature in multilayer ceramic systems is the ability to deflect propagating cracks and two different mechanisms of crack deflection have previously been of Materials Science and metall employed. Propagating cracks can be deflected either Cambridge, Cambridge CB along weak interfaces with adjacent layers [14, 15]or wei@@gmail.com into layers exhibiting residual biaxial compressive D. G. Brandon stress[17-19. Essentially, the first mechanism depends epartment of Materials Engineering, Technion-Israel on matrix/interface strength ratio and has had varying degrees of success. However, controlling the strength 2 Springer
Abstract Three kinds of texture-reinforced oxide ceramic laminates with strongly bonded interfaces have been fabricated. All three were based on highly textured a-alumina interlayers, but each with a different oxide laminate matrix and a correspondingly different thermal mismatch between the textured interlayers and the matrix. Alginate-based, aqueous gel casting was used to produce a flexible tape for all the compositions studied. The highly textured alumina (TA) interlayers were derived from gel-cast tapes containing aligned alumina seed platelets. It has been found that the strongest texture derived from 9.1 vol% of the seed platelets in the gelled precursor tape. Using the March–Dollase model for texture analysis, the oriented volume fraction in the highly textured interlayers was estimated to vary from 60 to 80%. Introduction In a conventional ceramic the anisotropic single crystal properties are averaged in the isotropic, randomly oriented polycrystal. Texturing a polycrystalline material, by controlling the orientation distribution of crystallites with respect to the coordinate system of a component, could be used to tailor electronic and structural properties and obtain some of the single crystal anisotropy. Such texture could be used to reduce thermal expansion mismatch, and hence the risk of microcracking [1], while texture control offers a useful alternative to aligned second-phase reinforcement [2]. Most recently, it has been reported that the formation of texture in a-alumina can effectively lower the residual stress and make the stress distribution narrower [3]. In general, processing to develop texture in ceramics differs significantly from the conventional powder processing of dense, fine-grained and equiaxed microstructures. Textured ceramics have been fabricated by several techniques: sinter-forging [4], hotpressing [5], slip-casting [6], gel-casting in a magnetic field [1, 7, 8], and, most commonly, by ‘templating’, using seeded grain growth [2, 9–13]. Textured microstructures have been reported in a single-phase iron titanate ceramic laminate, containing a predetermined configuration of the textured and non-textured layers [8]. The design of ceramic composites with layered macrostructures is also receiving considerable research attention because they exhibit decreased sensitivity to surface defects and have demonstrated non-catastrophic failure in some laminate systems [14–16]. A key feature in multilayer ceramic systems is the ability to deflect propagating cracks and two different mechanisms of crack deflection have previously been employed. Propagating cracks can be deflected either along weak interfaces with adjacent layers [14, 15] or into layers exhibiting residual biaxial compressive stress [17–19]. Essentially, the first mechanism depends on matrix/interface strength ratio and has had varying degrees of success. However, controlling the strength M. Wei (&) Æ D. Zhi Department of Materials Science and Metallurgy, University of Cambridge, Cambridge CB2 3QZ, UK e-mail: dr.m.wei@gmail.com D. G. Brandon Department of Materials Engineering, Technion – Israel Institute of Technology, Haifa 32000, Israel J Mater Sci (2006) 41:7425–7436 DOI 10.1007/s10853-006-0808-y 123 Oxide ceramic laminates with highly textured a-alumina interlayers: I. Texture control and laminate formation Ming Wei Æ Dan Zhi Æ David G. Brandon Received: 18 August 2005 / Accepted: 29 November 2005 / Published online: 20 September 2006 Springer Science+Business Media, LLC 2006
J Mater Sci(2006)41:7425-7436 of such ceramics is technologically difficult, and fur- Experimental procedure thermore weak interfaces are known to compromise certain properties. The use of residual stress patterns Powder preparation associated with thermal expansion mismatch within ceramic laminate materials therefore offers an attrac- A high purity, submicron a-Al2O3 powder(Ceralox, tive alternative mechanism to improve fracture HPA-0.5) and a known volume percent of a-Al2O3 behaviour seed platelets(10-15 um diameter, ELF Atochem) Q In the present research, texture-reinforced oxide were used to prepare the TA layers. The oxide matrix amic laminates have been prepared by an alginate- precursor powder compositions and their sources are based, gel-cast, doctor-blade process, in which tape listed in Table 1. The rBao powder mixture was lay-ups to develop the laminated structure. Three attrition-milled in acetone with 3 mm-diameter high laminate systems were studied; all based on strongly purity Al2O, beads(99.9%, Union Process) for 6-8 h bonded interfaces between the matrix composition and at 500-600 rpm. After drying in a ventilation hood, the highly textured alumina(TA)interlayers. The control powder mixture was sieved with a 60-mesh sieve. The bonded and textured interlayers is independent of, and attrition milling was monitored by scanning electa of mechanical anisotropy by the presence of strongly particle size of the rBao powder before and af quite different from, the principle of reinforcement by microscopy(LEO 982)and laser particle size analysis weak-interface crack deflection, which is used to inhibit (HORIBA, LA-910) through-thickness crack propagation in many aligned ceramic composites. Gel casting and lamination In this paper, the texture and microstructure of the alumina interlayers were optimized by controlling the Water-based gel casting [2] was used to fabricate both volume fraction and alignment of alumina seed plate- the textured and texture-free layers. An in situ sol-gel lets in a precursor tape. Meanwhile, three types of reaction based on ion exchange between water-soluble matrix layer were used in three laminate systems: sodium alginate and polyvalent cations yields an 1. A reaction-bonded aluminum oxide (RBAO) insoluble, cation cross-linked alginate gel near-zero residual thermal stress in the fully sin- Nanalginate n/2Ca--nNa+ Can/2 alginate (1) tered, single-phase alumina/TA laminate 2. An alumina-toughened zirconia (ATZ), designed A 0.15 M Ca solution of calcium nitrate-4-hydrate for minimum tetragonal zirconia grain-size and (AR purity, Riedel-de Haen), served as the gelling with controlled residual tension in the zirconia solution. The seeded alumina layers were prepared by matrix layers of the laminate first mixing the alumina powder with alginate and a 3. A reaction-bonded(zircon-based) mullite(RBM), dispersant in distilled water by ball milling in a plastic designed for maximum mullite conversion and bottle for at least 12 h(99.9% alumina balls, 5 mn residual compression in the mullite matrix layers of diameter--Union Process). The alumina seed platelets the laminate were then incorporated and ball-milled for a further 5 h. The homogeneously mixed slip was transferred to This paper reports the processing methodology and a second plastic bottle to remove the alumina balls, and texture control for these three texture-reinforced then rolled slowly on the ball mill for about 2 h to ceramic laminates. It will be followed by a further remove trapped air bubbles. The de-gassed slips were report on microstructural development and mechanical poured into the tape caster and cast with the gelling behaviour of the three laminate systems. solution onto a Mylar sheet substrate(Dupont) which Table 1 Precursor powders for RBAO. ATZ and rBM ayer type Material Powder characteristic and source Composition(wt% RBAO 99.36%,45-90mm Miller Thermal. In AlO3 TZ3Y TSK, Y-stabilized tetragonal ZrO2(75.2)Y2O3(4.2) s Another zircon powde ZrO+a-AlO3, <100 n Al2O3(20) submicron Rami submicron (5 mm) was used for AlO3 Ceralox, HPA-05 2 Springer
of such ceramics is technologically difficult, and furthermore weak interfaces are known to compromise certain properties. The use of residual stress patterns associated with thermal expansion mismatch within ceramic laminate materials therefore offers an attractive alternative mechanism to improve fracture behaviour. In the present research, texture-reinforced oxide ceramic laminates have been prepared by an alginatebased, gel-cast, doctor-blade process, in which tape lay-ups to develop the laminated structure. Three laminate systems were studied; all based on strongly bonded interfaces between the matrix composition and highly textured alumina (TA) interlayers. The control of mechanical anisotropy by the presence of strongly bonded and textured interlayers is independent of, and quite different from, the principle of reinforcement by weak-interface crack deflection, which is used to inhibit through-thickness crack propagation in many aligned ceramic composites. In this paper, the texture and microstructure of the alumina interlayers were optimized by controlling the volume fraction and alignment of alumina seed platelets in a precursor tape. Meanwhile, three types of matrix layer were used in three laminate systems: 1. A reaction-bonded aluminum oxide (RBAO), designed for minimum sintering shrinkage and a near-zero residual thermal stress in the fully sintered, single-phase alumina/TA laminate. 2. An alumina-toughened zirconia (ATZ), designed for minimum tetragonal zirconia grain-size and with controlled residual tension in the zirconia matrix layers of the laminate. 3. A reaction-bonded (zircon-based) mullite (RBM), designed for maximum mullite conversion and residual compression in the mullite matrix layers of the laminate. This paper reports the processing methodology and texture control for these three texture-reinforced ceramic laminates. It will be followed by a further report on microstructural development and mechanical behaviour of the three laminate systems. Experimental procedure Powder preparation A high purity, submicron a-Al2O3 powder (Ceralox, HPA-0.5) and a known volume percent of a-Al2O3 seed platelets (10–15 lm diameter, ELF Atochem) were used to prepare the TA layers. The oxide matrix precursor powder compositions and their sources are listed in Table 1. The RBAO powder mixture was attrition-milled in acetone with 3 mm-diameter highpurity Al2O3 beads (99.9%, Union Process) for 6–8 h at 500–600 rpm. After drying in a ventilation hood, the powder mixture was sieved with a 60-mesh sieve. The particle size of the RBAO powder before and after attrition milling was monitored by scanning electron microscopy (LEO 982) and laser particle size analysis (HORIBA, LA-910). Gel casting and lamination Water-based gel casting [2] was used to fabricate both the textured and texture-free layers. An in situ sol-gel reaction based on ion exchange between water-soluble sodium alginate and polyvalent cations yields an insoluble, cation cross-linked alginate gel: Nanalginate þ n=2Ca2þ ! nNaþ þ Can=2 alginate ð1Þ A 0.15 M Ca2+ solution of calcium nitrate-4-hydrate (AR purity, Riedel-deHae¨n), served as the gelling solution. The seeded alumina layers were prepared by first mixing the alumina powder with alginate and a dispersant in distilled water by ball milling in a plastic bottle for at least 12 h (99.9% alumina balls, 5 mm diameter—Union Process). The alumina seed platelets were then incorporated and ball-milled for a further 5 h. The homogeneously mixed slip was transferred to a second plastic bottle to remove the alumina balls, and then rolled slowly on the ball mill for about 2 h to remove trapped air bubbles. The de-gassed slips were poured into the tape caster and cast with the gelling solution onto a Mylar sheet substrate (Dupont) which Table 1 Precursor powders for RBAO, ATZ and RBM layers * Another zircon powder (~5 mm) was used for comparison Layer type Material Powder characteristic and source Composition (wt%) RBAO Al Globular, 99.36%, 45–90 mm, Miller Thermal, Inc. 20–40 Al2O3 Submicron, Ceralox, HPA-0.5 60–80 ATZ TZ3Y20A TSK, Y-stabilized tetragonal ZrO2 + a-Al2O3, < 100 nm ZrO2 (75.2) Y2O3 (4.2) Al2O3 (20) RBM ZrSiO4* submicron, Rami Submicron, Ceralox, HPA-0.5 54.5 Al2O3 45.5 7426 J Mater Sci (2006) 41:7425–7436 123
J Mater Sci(2006)41:7425-7436 7427 was supported on an inclined glass plate to drain away low heating rate(1C/min) and low temperature hold excess gelling solution and inhibit premature gelling of (600C, 4-10 h). The green RBAO/TA laminated the slip beneath the doctor blade. The thickness of the samples were placed on a porous alumina tile to ensure cast tape was controlled by adjusting both the caster free oxygen access beneath the sample during reaction velocity and the gap between the extended doctor bonding. blade and the casting plate. Strong, flexible green tapes The alumina-toughened zirconia/textured 100-200 um in thickness were obtained with a doctor (ATZ/TA)and reaction-bonded mullite/t blade gap of 250 um alumina(RBMTA)laminates were sintered The non-textured oxide layers(RBAO, ATZ, and tionally. Dissociation of zircon and its reaction wit RBM) were cast by an identical gel-casting process, but alumina to form zirconia and mullite only occurs above without incorporation of the seed platelets. The slip 1400C [24], so the heating cycle selected for the RBM compositions for the TA layers and the three types of laminates was 5/min up to 1400C, 2/min from 1400 non-textured oxide layers are listed in Table 2. to1600C, followed by holding for 5 h at 1600C and After casting, the tapes were detached from the then cooling to room temperature at 5C/min. The Mylar sheet, by inverting the substrate in a distilled effect of the zircon particle size on the phase evolution water bath. The tapes were then washed to remove of the RBM layers was evaluated by X-ray diffraction excess cation, and 1.25"diameter disks were stamped phase analysis and the final density of the laminate from the washed tapes using a polished steel punch composite samples was determined by Archimedes (Buehler). Green laminate specimens were pressed density measurements from a stacked lay-up of the textured and non-textured disks in a stainless steel die(also 1. 25"in diameter) using a single filter paper top and bottom to prevent Texture characterization dhesion and allow excess water to escape. The spec imens were uniaxially pressed in the die at -80 Mpa for The chemical homogeneity and composition of indi- 20 min and subsequently cold isostatically pressed vidual grains were determined from carbon-coated (CIP)at 250-300 MPa for 30 min. polished and thermally etched samples using conven- tional scanning electron microscopy(SEM) and energy dispersive X-ray spectroscopy(EDS-JEOL 840). The microstructure of cross-sections taken from laminated Residual water was removed by drying in a 2.45 GHz samples was also examined by diffraction contrast microwave oven(EM-S301, SANYO)operated at 100- transmission electron microscopy (TEM-JEOL 2000- 150 W for 30-60 min. The progress of microwave FX)at 200 kV. TEM samples were prepared by dia- drying was assessed by weighing the samples. The small mond-sectioning perpendicular to the original quantity of organic additives (<1 vol% of dispersant the precursor tape, followed by mechanical grinding, and alginate) was removed during the sintering/reac- dimpling, and precision ion polishing(PIPS-Gatan tion-bonding cycle and not as a separate burnout Model 691) treatment The Harris method [25-27 was used to quantify orted for monolithic RBAo the degree of crystallographic alignment(texture)in ceramic processing [20-23], a somewhat complex all specimens. X-ray diffraction spectra were taken heating cycle is required to sinter the BAOTa from polished surfaces prepared parallel to the laminates to high density(avoiding swelling)and with direction of tape casting. Reference diffraction data 100% conversion to oxide. Oxidation of the aluminum were obtained from randomly oriented high purity metal particles at low temperatures(below the melting alumina powder(HPA-0.5, Ceralox). All X-ray data point) is essential and was accomplished by an initially were collected on an X-ray diffractometer with Table 2 Slip comp osition for Tape gel casting of textured and Matrix powder AlO, platelet Na alginate non-textured oxide layers TA 25-30 1-6 0.15-0.2 0.5-0.6 66-72 RBAO 000 0.5-0.6 0.2-0.25 06-0.7 DISPEX A40. Allied RBM 0.2-0.25 0.5-0.6 Colloid, england 2 Springer
was supported on an inclined glass plate to drain away excess gelling solution and inhibit premature gelling of the slip beneath the doctor blade. The thickness of the cast tape was controlled by adjusting both the caster velocity and the gap between the extended doctor blade and the casting plate. Strong, flexible green tapes 100–200 lm in thickness were obtained with a doctor blade gap of 250 lm. The non-textured oxide layers (RBAO, ATZ, and RBM) were cast by an identical gel-casting process, but without incorporation of the seed platelets. The slip compositions for the TA layers and the three types of non-textured oxide layers are listed in Table 2. After casting, the tapes were detached from the Mylar sheet, by inverting the substrate in a distilled water bath. The tapes were then washed to remove excess cation, and 1.25¢¢ diameter disks were stamped from the washed tapes using a polished steel punch (Buehler). Green laminate specimens were pressed from a stacked lay-up of the textured and non-textured disks in a stainless steel die (also 1.25¢¢ in diameter), using a single filter paper top and bottom to prevent adhesion and allow excess water to escape. The specimens were uniaxially pressed in the die at ~80 Mpa for 20 min and subsequently cold isostatically pressed (CIP) at 250–300 MPa for 30 min. Sintering Residual water was removed by drying in a 2.45 GHz microwave oven (EM-S301, SANYO) operated at 100– 150 W for 30–60 min. The progress of microwave drying was assessed by weighing the samples. The small quantity of organic additives (<1 vol% of dispersant and alginate) was removed during the sintering/reaction-bonding cycle and not as a separate burnout treatment. As previously reported for monolithic RBAO ceramic processing [20–23], a somewhat complex heating cycle is required to sinter the RBAO/TA laminates to high density (avoiding swelling) and with 100% conversion to oxide. Oxidation of the aluminum metal particles at low temperatures (below the melting point) is essential and was accomplished by an initially low heating rate (1 C/min) and low temperature hold (600 C, 4–10 h). The green RBAO/TA laminated samples were placed on a porous alumina tile to ensure free oxygen access beneath the sample during reaction bonding. The alumina-toughened zirconia/textured alumina (ATZ/TA) and reaction-bonded mullite/textured alumina (RBM/TA) laminates were sintered conventionally. Dissociation of zircon and its reaction with alumina to form zirconia and mullite only occurs above 1400 C [24], so the heating cycle selected for the RBM laminates was 5/min up to 1400 C, 2/min from 1400 to1600 C, followed by holding for 5 h at 1600 C and then cooling to room temperature at 5 C/min. The effect of the zircon particle size on the phase evolution of the RBM layers was evaluated by X-ray diffraction phase analysis and the final density of the laminate composite samples was determined by Archimedes density measurements. Texture characterization The chemical homogeneity and composition of individual grains were determined from carbon-coated, polished and thermally etched samples using conventional scanning electron microscopy (SEM) and energy dispersive X-ray spectroscopy (EDS—JEOL 840). The microstructure of cross-sections taken from laminated samples was also examined by diffraction contrast transmission electron microscopy (TEM-JEOL 2000- FX) at 200 kV. TEM samples were prepared by diamond-sectioning perpendicular to the original plane of the precursor tape, followed by mechanical grinding, dimpling, and precision ion polishing (PIPS—Gatan, Model 691). The Harris method [25–27] was used to quantify the degree of crystallographic alignment (texture) in all specimens. X-ray diffraction spectra were taken from polished surfaces prepared parallel to the direction of tape casting. Reference diffraction data were obtained from randomly oriented high purity alumina powder (HPA-0.5, Ceralox). All X-ray data were collected on an X-ray diffractometer with a Table 2 Slip composition for gel casting of textured and non-textured oxide layers (vol%) * DISPEX A40, Allied Colloid, England Tape Matrix powder Al2O3 platelet Dispersant* Na alginate Water TA 25–30 1–6 0.15–0.2 0.5–0.6 66–72 RBAO 32–38 0 0.15–0.2 0.5–0.6 62–68 ATZ 25–30 0 0.2–0.25 0.6–0.7 70–75 RBM 28–32 0 0.2–0.25 0.5–0.6 68–72 J Mater Sci (2006) 41:7425–7436 7427 123
J Mater Sci(2006)41:7425-7436 parallel-beam assembly (PW1710 based). The data SEM. Surface preparation for the EBSD sample is were recorded for 20 values from 40 to 154 using Cu critical since the diffraction information originates in a Kan radiation with a step size of 0.05 and a 3-s scan 20 nm layer at the surface, corresponding to the pen time per step. The lower limit of 40 was chosen to etration depth for back-scattered electrons. In our ensure a constant irradiated volume for all data. The process, the sample surface was polished, etched, and integrated intensities for all reflections were calcu- ultrasonically cleaned before EBSD examination. The lated using PhILLIPs APD software EBSD data were collected sequentially by positioning The March-Dollase function [28, P,(n), was used to the focussed electron beam on each grain individually estimate the probability of observing diffraction from The normal to the sample surface was tilted 70 to the the hkl planes at a given angle incident beam and the eBsd pattern was captured at a beam voltage of 20 kV. The individual crystal orien Pr(n)=(cos2n+r-lsin2n)-3/2 (2) tations were analyzed using commercial software to generate discrete pole figures for each localized where n is the angle between the basal plane(001) and microtexture data se the(hkl) planes. The parameter r is the ' compaction ratio'(final thickness/initial thickness) in the original March model. For the present purpose, r was used to Results and discussion indicate the degree of texture: r= l for a random sample and rI F (greater than random). The oriented volume fraction was thus defined as: Oriented=/(P(m)-1)sin ndn where is given by P(1)=1. Fig. 1 SEM image of a cross-section of gel-cast textured green Microtexture was also investigated using electron tape showing the distribution of seed platelets(marked by back-scatter diffraction(EBSD) in the LEO 982-FEG- arrows)in the tap 2 Springer
parallel-beam assembly (PW1710 based). The data were recorded for 2h values from 40 to 154 using Cu Ka1 radiation with a step size of 0.05 and a 3-s scan time per step. The lower limit of 40 was chosen to ensure a constant irradiated volume for all data. The integrated intensities for all reflections were calculated using PHILLIPS APD software. The March–Dollase function [28], Pr(g), was used to estimate the probability of observing diffraction from the hkl planes at a given angle: PrðÞ¼ð g r 2 cos2 g þ r 1 sin2 gÞ 3=2 ð2Þ where g is the angle between the basal plane (001) and the (hkl) planes. The parameter r is the ‘compaction ratio’ (final thickness/initial thickness) in the original March model. For the present purpose, r was used to indicate the degree of texture: r = 1 for a random sample and r 1 (greater than random). The oriented volume fraction was thus defined as: Voriented ¼ Z g1 0 ðPðgÞ 1Þsin gdg ð5Þ where g1 is given by P(g1) = 1. Microtexture was also investigated using electron back-scatter diffraction (EBSD) in the LEO 982-FEGSEM. Surface preparation for the EBSD sample is critical since the diffraction information originates in a 20 nm layer at the surface, corresponding to the penetration depth for back-scattered electrons. In our process, the sample surface was polished, etched, and ultrasonically cleaned before EBSD examination. The EBSD data were collected sequentially by positioning the focussed electron beam on each grain individually. The normal to the sample surface was tilted 70o to the incident beam and the EBSD pattern was captured at a beam voltage of 20 kV. The individual crystal orientations were analyzed using commercial software to generate discrete pole figures for each localized, microtexture data set. Results and discussion Texture control in c-axis TA interlayers A cross-section of gel-cast green tape is shown in Fig. 1 and demonstrates alignment of the alumina seed platelets by the gel casting process. Good alignment of the seed alumina platelets is a key factor in attaining high particle packing density in the green bodies, and the aligned, high aspect ratio platelets did not prevent the achievement of high green densities in the uniaxially and cold-isostatically pressed compacts. High sintered densities were obtainable in the TA samples after sintering at 1550 C for 2 h, despite the presence of the aligned platelets and the limitations of pressureless sintering. As expected, under the same sintering conditions lower final densities were associated with higher initial platelet contents. Fig. 1 SEM image of a cross-section of gel-cast textured green tape showing the distribution of seed platelets (marked by arrows) in the tape 7428 J Mater Sci (2006) 41:7425–7436 123
J Mater Sci(2006)41:7425-7436 7429 The linear shrinkage ratios(the ratio of the sintering Ka) shrinkage of the compact perpendicular and parallel to the original seed platelets) were also measured, as shown in Fig. 2. Increasing the initial platelet content leads to increased constraint in the plane of the platelets, so that more densification occurs normal to the platelets. Thus, platelet alignment is accompanied by anisotropic densification, as has been reported fo an alumina-platelet-reinforced Ce-ZrO2/Al2O3 com- The volume of dramatically during sintering through growth of the aligned seed platelets at the expense of the surround um ing, randomly oriented, fine Al2O, particles [2]. The microstructure of a cross-section of a sintered alumina compact with 4.8 vol% initial seed platelet content is (b) shown in Fig. 3 and the microstructural anisotropy of the monolithic ta is evident in both the micrograph in Fig 3a and the SEM image in The seed platelets have acted as templates for both sintered grain morphology and crystalline texture development, so that the strong 0001 texture(c-axis perpendicular to the sample surface and normal to the tape-casting direction) is accompanied by morpholog ical anisotropy(flattened grains parallel to the surface) Figure 3c shows a bright-field TEM image taken from a cross-section of TA layer, in which a grain boundary between two Ta grains is accurately parallel to both basal planes(see inserted SAD pattern) with no sign of 10 Hm any glassy phase or second-phase precipitation X-ray diffraction patterns from a green body and c sintered (1550C, 2 h) samples with different initial latelet contents are shown in Fig 4. The 00.6 and 00. 12 peaks for the sintered, platelet-containing zA.[210 sintered textured alumina (TA)with 4.8 vol% initia (b) SEM micrograph of the cross-section of fully sir with 4.8 vol% initial platelets:(c) bright-field TEM basal grain boundary between two Ta grains Platelet Content(vol%) les are much more prominent than those from either Fig 2 Linear shrinkage ratios for the textured alumina(TA) the green body or a sintered, platelet-free sample samples with different initial seed platelet contents Similarly, the near-basal-plane peaks, such as 10.10 and
The linear shrinkage ratios (the ratio of the sintering shrinkage of the compact perpendicular and parallel to the original seed platelets) were also measured, as shown in Fig. 2. Increasing the initial platelet content leads to increased constraint in the plane of the platelets, so that more densification occurs normal to the platelets. Thus, platelet alignment is accompanied by anisotropic densification, as has been reported for an alumina-platelet-reinforced Ce–ZrO2/Al2O3 composite [30]. The volume of textured (aligned) crystals increased dramatically during sintering through growth of the aligned seed platelets at the expense of the surrounding, randomly oriented, fine Al2O3 particles [2]. The microstructure of a cross-section of a sintered alumina compact with 4.8 vol% initial seed platelet content is shown in Fig. 3 and the microstructural anisotropy of the monolithic TA is evident in both the optical micrograph in Fig. 3a and the SEM image in Fig. 3b. The seed platelets have acted as templates for both sintered grain morphology and crystalline texture development, so that the strong 0001 texture (c-axis perpendicular to the sample surface and normal to the tape-casting direction) is accompanied by morphological anisotropy (flattened grains parallel to the surface). Figure 3c shows a bright-field TEM image taken from a cross-section of TA layer, in which a grain boundary between two TA grains is accurately parallel to both basal planes (see inserted SAD pattern) with no sign of any glassy phase or second-phase precipitation. X-ray diffraction patterns from a green body and sintered (1550 C, 2 h) samples with different initial platelet contents are shown in Fig. 4. The 00.6 and 00.12 peaks for the sintered, platelet-containing samples are much more prominent than those from either the green body or a sintered, platelet-free sample. Similarly, the near-basal-plane peaks, such as 10.10 and 1.0 1.5 2.0 2.5 3.0 0 5 10 15 20 Linear Shrinkage Ratio (Y/X Shrinkage Ratio) Platelet Content (vol%) X X Y Fig. 2 Linear shrinkage ratios for the textured alumina (TA) samples with different initial seed platelet contents Fig. 3 (a) Optical micrograph of the cross-section of fully sintered textured alumina (TA) with 4.8 vol% initial platelets; (b) SEM micrograph of the cross-section of fully sintered TA with 4.8 vol% initial platelets; (c) bright-field TEM image of a basal grain boundary between two TA grains J Mater Sci (2006) 41:7425–7436 7429 123
J Mater Sci(2006)41:7425-7436 10.14(17.5 and 12.7 from the basal plane, respec- degree of crystallographic alignment initially increases tively), also appear stronger in the platelet-containing with increasing platelet content but eventually reaches samples than in the green body or the platelet-free a maximum due to increasing interference between sample. The results confirm that large volume fractions adjacent platelets as they rotate into the plane of the of strong c-axis textured material are formed in sin- tape during tape-casting. Interference between neigh tered platelet-containing samples. Diffraction data boring platelets is also the probable cause of the lower from samples with different initial platelet contents densification at high platelet contents. were compared with the diffraction data from a In traditional texture determination by X-ray dif platelet-free, random sample. The MRD for each fraction, large numbers of grains are sampled simul reflection was calculated and fitted to a March-Dollase taneously and no information is available regarding the model function Phkl(n), and is shown as a function of n crystal location responsible for individual X-ray in Fig 5a. The parameter r in the March-Dollase reflections. Local grain orientation cannot be identified function and the oriented volume fraction calculated by the X-ray method. Transmission electron micros- from Eq 5 are plotted as a function of n in Fig 5b, copy (TEM) is capable of providing diffraction data from which the smallest r value and the highest from grains down to 10 nm diameter with an angular ented volume fraction are seen to correspond to precision, in convergent beam diffraction, as high as initial platelet fraction of 9.1 vol %. In effect, 0.1. However, tEM specimens must be electron 00.12 geen body (9. tvo 91v0% ▲--24w ::48vc% m丰→才 Sintered (9.1 vol%) Angle from basal plane(degree (b) Sintered(4.8 yol/) 鲁一 Chemed voLme factor;% Mrch-oclase Parameter 0.8 6 AA人A~人 40 04 0.2 5 Initial platelet content (vol%) A△人 ultiple of random distribution (MRD) fo flections from textured alumina(TA) samples itial platelet contents, together with the March- Dollase function fitting:(b)the March-Dollase parameter r and Fig 4 XRD patterns for green and sintered alumina samples the oriented volume fractions for TA samples as a function the with different initial seed platelet content initial platelet content 2 Springer
10.14 (17.5 and 12.7 from the basal plane, respectively), also appear stronger in the platelet-containing samples than in the green body or the platelet-free sample. The results confirm that large volume fractions of strong c-axis textured material are formed in sintered platelet-containing samples. Diffraction data from samples with different initial platelet contents were compared with the diffraction data from a platelet-free, random sample. The MRD for each reflection was calculated and fitted to a March–Dollase model function Phkl(g), and is shown as a function of g in Fig. 5a. The parameter r in the March–Dollase function and the oriented volume fraction calculated from Eq. 5 are plotted as a function of g in Fig. 5b, from which the smallest r value and the highest oriented volume fraction are seen to correspond to an initial platelet fraction of 9.1 vol%. In effect, the degree of crystallographic alignment initially increases with increasing platelet content but eventually reaches a maximum due to increasing interference between adjacent platelets as they rotate into the plane of the tape during tape-casting. Interference between neighboring platelets is also the probable cause of the lower densification at high platelet contents. In traditional texture determination by X-ray diffraction, large numbers of grains are sampled simultaneously and no information is available regarding the crystal location responsible for individual X-ray reflections. Local grain orientation cannot be identified by the X-ray method. Transmission electron microscopy (TEM) is capable of providing diffraction data from grains down to 10 nm diameter with an angular precision, in convergent beam diffraction, as high as 0.1. However, TEM specimens must be electron Fig. 4 XRD patterns for green and sintered alumina samples with different initial seed platelet contents Fig. 5 (a) The multiple of random distribution (MRD) for observed (hkl) reflections from textured alumina (TA) samples with different initial platelet contents, together with the March– Dollase function fitting; (b) the March–Dollase parameter r and the oriented volume fractions for TA samples as a function the initial platelet content 7430 J Mater Sci (2006) 41:7425–7436 123
J Mater Sci(2006)41:7425-7436 7431 transparent, typically less than 100 nm thick, and the standard eBsd pattern, shown in Fig. 6b, was first area of the sample thin enough for electron transmis- taken from a polished sapphire substrate for compari sion is only of the order of the typical grain size of TA, son. Figure 7 shows the EBSD patterns taken from the so that TEM unsuitable for comparing grain orienta- TA grains and the EBSD results demonstrate that the tions in these materials. In the present work EBSD was local orientation of each grain relative to the surface used to examine the regional grain orientation rela- normal deviates only slightly from the 0001 zone axis tions(microtexture)in the c-axis textured a-alumina These results were plotted in the orientation of the pellets. Figure 6a shows a forward-scatter secondary plane of the sample as discrete and inverse pole figures electron(FSE)image taken from an area examined in shown in Fig 8. The above EBSD measurement results a tilted in-plane surface of the TA with 9.1% initial confirm the X-ray diffraction macrotexture results and platelet content. A dozen alumina grains from this area show the strong out-of-plane c-axis texture. In addi- (marked by letters) were examined by EBSD and a tion, no preferred in-plane orientation was detected from the ebsd measurement similar results can also be obtained from ebsd characterization of the cross. section of the same TA sample(not shown). Formation of rbaota. atta and rbm/ta laminates Based on the above texture analysis, we decided to 9.1 vol% alumina platelets in the TA interlayers in all D three types of laminates. These TA tapes were first used,together with texture-free, fine-grain, RBAO alumina layers, to fabricate the rBaoTA laminate, in which the two'different'layers contain the same phase alumina, but in different morphologies. The thermal mismatch between the textured and non-textured la ers in this laminate is small and is associated with the 10% difference in thermal expansion coefficient of a-alumina parallel and perpendicular to the c axis. The textured layers in this laminate structure are expected to reinforce the alumina by crack deflection along the grain boundaries formed with the basal planes of the 山m6 textured grains, which leads to the formation of low 百21 energy basal plane free surfaces In order to adapt traditional rBAo processing for the texture-free layers in this texture-reinforced, sin gle-phase alumina laminate, a fine(usually submicron) 231022113 o precursor powder mixture of aluminum and alumina was prepared by attrition milling. The particle size 565416 distribution after attrition milling is shown in Fig. 9a, 2423 110u while Fig 9b and c compare the as-received aluminum powder (plasma-sprayed, 45-90 um) with the submi cron particle size and irregular grains in the attrition 45-13 5413 milled RBAO powder mixture. The heating cycle for RBAO samples requires a low heating rate below 1100oC to ensure sufficient oxidation of the metal 2201 particles Phase evolution in a series of RBAO samples was evaluated by X-ray diffraction. After heating at taken from an area examined in a tilted in-plane surface of the 1 C/min to 1100 Fig 6(a) Forward-scatter secondary electron (FSE) image the rAO samp fully textured alumina(TA)with 9. 1% initial platelet content;(b oxidized to a-alumina, but a large fraction of the met indexed standard electron back-scatter diffraction(EBSD) oxidized below the melting point (660C). Figure 10a pattern from polished a basal sapphire substrate shows the X-ray diffraction patterns from a RBAO 2 Springer
transparent, typically less than 100 nm thick, and the area of the sample thin enough for electron transmission is only of the order of the typical grain size of TA, so that TEM unsuitable for comparing grain orientations in these materials. In the present work EBSD was used to examine the regional grain orientation relations (microtexture) in the c-axis textured a-alumina pellets. Figure 6a shows a forward-scatter secondary electron (FSE) image taken from an area examined in a tilted in-plane surface of the TA with 9.1% initial platelet content. A dozen alumina grains from this area (marked by letters) were examined by EBSD and a standard EBSD pattern, shown in Fig. 6b, was first taken from a polished sapphire substrate for comparison. Figure 7 shows the EBSD patterns taken from the TA grains and the EBSD results demonstrate that the local orientation of each grain relative to the surface normal deviates only slightly from the 0001 zone axis. These results were plotted in the orientation of the plane of the sample as discrete and inverse pole figures shown in Fig. 8. The above EBSD measurement results confirm the X-ray diffraction macrotexture results and show the strong out-of-plane c-axis texture. In addition, no preferred in-plane orientation was detected from the EBSD measurement. Similar results can also be obtained from EBSD characterization of the crosssection of the same TA sample (not shown). Formation of RBAO/TA, ATZ/TA, and RBM/TA laminates Based on the above texture analysis, we decided to use 9.1 vol% alumina platelets in the TA interlayers in all three types of laminates. These TA tapes were first used, together with texture-free, fine-grain, RBAO alumina layers, to fabricate the RBAO/TA laminate, in which the two ‘different’ layers contain the same phase, a-alumina, but in different morphologies. The thermal mismatch between the textured and non-textured layers in this laminate is small, and is associated with the 10% difference in thermal expansion coefficient of a-alumina parallel and perpendicular to the c axis. The textured layers in this laminate structure are expected to reinforce the alumina by crack deflection along the grain boundaries formed with the basal planes of the textured grains, which leads to the formation of low energy basal plane free surfaces. In order to adapt traditional RBAO processing for the texture-free layers in this texture-reinforced, single-phase alumina laminate, a fine (usually submicron) precursor powder mixture of aluminum and alumina was prepared by attrition milling. The particle size distribution after attrition milling is shown in Fig. 9a, while Fig. 9b and c compare the as-received aluminum powder (plasma-sprayed, 45–90 lm) with the submicron particle size and irregular grains in the attrition milled RBAO powder mixture. The heating cycle for RBAO samples requires a low heating rate below 1100 C to ensure sufficient oxidation of the metal particles. Phase evolution in a series of RBAO samples was evaluated by X-ray diffraction. After heating at 1 C/min to 1100 C, the RBAO sample was fully oxidized to a-alumina, but a large fraction of the metal oxidized below the melting point (660 C). Figure 10a shows the X-ray diffraction patterns from a RBAO Fig. 6 (a) Forward-scatter secondary electron (FSE) image taken from an area examined in a tilted in-plane surface of the textured alumina (TA) with 9.1% initial platelet content; (b) indexed standard electron back-scatter diffraction (EBSD) pattern from polished a basal sapphire substrate J Mater Sci (2006) 41:7425–7436 7431 123
J Mater Sci(2006)41:7425-7436 Fig. 7 Electron back-scatter diffraction(EBSD) pattern from the 12 grains(A-L) ndicated in Fig. 6 sample(20 vol% Al) before and after reaction bond- contrast to the above RBAO/TA laminates, the aTZ ng, confirming complete oxidation after the rBao layers are now in tension while the TA layers are in process. The final density of the RBAO/TA laminates compression, due to thermal expansion mismatch reached 97%TD. Figure 10b shows a typical optical between the adjacent layers. The residual stress micrograph of a sintered RBAO/TA laminate, in distribution is therefore expected to enhance crack which the thickness ratio of RBAO and TA layer is 3: 1 deflection at the interface between the zirconia and TA (alternating three stacks of gel-cast green RBAO tapes layers and one gel-cast green TA tape ). It has been found that Figure 1la shows an optical micrograph of a sin the resulting laminar interfaces are fairly straight in the tered ATZTA laminate, in which the thickness ratio laminate sample of ATZ and TA layer is 1: 1(alternating one gel-cast The second type of the texture-reinforced ceramic green ATZ tape and one gel-cast green TA tape). The laminate was developed from a fine-grain, texture-free, ATZ layer thickness has been found to be slightly ATZ matrix interleaved with highly TA interlayers In smaller than that of TA layer, which is due to larger 2 Springer
sample (20 vol% Al) before and after reaction bonding, confirming complete oxidation after the RBAO process. The final density of the RBAO/TA laminates reached 97% TD. Figure 10b shows a typical optical micrograph of a sintered RBAO/TA laminate, in which the thickness ratio of RBAO and TA layer is 3:1 (alternating three stacks of gel-cast green RBAO tapes and one gel-cast green TA tape). It has been found that the resulting laminar interfaces are fairly straight in the laminate sample. The second type of the texture-reinforced ceramic laminate was developed from a fine-grain, texture-free, ATZ matrix interleaved with highly TA interlayers. In contrast to the above RBAO/TA laminates, the ATZ layers are now in tension while the TA layers are in compression, due to thermal expansion mismatch between the adjacent layers. The residual stress distribution is therefore expected to enhance crack deflection at the interface between the zirconia and TA layers. Figure 11a shows an optical micrograph of a sintered ATZ/TA laminate, in which the thickness ratio of ATZ and TA layer is 1:1 (alternating one gel-cast green ATZ tape and one gel-cast green TA tape). The ATZ layer thickness has been found to be slightly smaller than that of TA layer, which is due to larger Fig. 7 Electron back-scatter diffraction (EBSD) patterns from the 12 grains (A–L) indicated in Fig. 6 7432 J Mater Sci (2006) 41:7425–7436 123
J Mater Sci(2006)41:7425-7436 7433 100 DIN Particle size (um) 12叮 g1-10 Fig8 Discrete 0001 pole figure (a) and inverse pole figure (b), both measured from electron back-scatter diffraction (EBSD) data 25um sintering shrinkage through original tape thickness in ATZ layer. Figure 1lb-d show a SEM image and EDX omposition maps taken from fine grain ATZ layer The results indicate that the alumina grains are uni- formly distributed among the zirconia matrix in the ATZ For the rBm/TA laminate, the mullite layers were produced by an in situ mullitization reaction. In con trast to the rbaoita and atta laminates the TA layers are now designed in biaxial tension due to the thermal mismatch between adjacent layers, placing 500nm the mullite in residual compression. The expectation as that the mullite layers would be reinforced by the Fig9(a)Particle size distribution for the reaction imposed compressive stresses, while the TA micro- aluminum oxide(RBAO)powder mixture after attrition structure would still promote crack deflection along the (b) The as received Al metal powder,(e)The RBAO mixture after attrition milling basal plane grain boundaries. Figure 12a shows an ptical micrograph of the cross-section of a sintered between zircon and alumina will be difficult to com RBM/TA laminate, in which the thickness ratio of plete, and Fig 12b shows a large unreacted zircon RBM and TA layer is 1: 1(alternating one gel-cast grain in the RBM layer. Phase evolution in RBM green RBM tape and one gel-cast green TA tape). samples with different zircon precursor particle sizes As in the rBAo process, the precursor particle size was therefore studied by X-ray diffraction analysis critical for the successful completion of the rbm The phase contents of both zirconia(M1)and mullite reaction. With a coarse zircon powder, the reaction (M2) were estimated using the following equations 2 Springer
sintering shrinkage through original tape thickness in ATZ layer. Figure 11b–d show a SEM image and EDX composition maps taken from fine grain ATZ layer. The results indicate that the alumina grains are uniformly distributed among the zirconia matrix in the ATZ layer. For the RBM/TA laminate, the mullite layers were produced by an in situ mullitization reaction. In contrast to the RBAO/TA and ATZ/TA laminates, the TA layers are now designed in biaxial tension due to the thermal mismatch between adjacent layers, placing the mullite in residual compression. The expectation was that the mullite layers would be reinforced by the imposed compressive stresses, while the TA microstructure would still promote crack deflection along the basal plane grain boundaries. Figure 12a shows an optical micrograph of the cross-section of a sintered RBM/TA laminate, in which the thickness ratio of RBM and TA layer is 1:1 (alternating one gel-cast green RBM tape and one gel-cast green TA tape). As in the RBAO process, the precursor particle size is critical for the successful completion of the RBM reaction. With a coarse zircon powder, the reaction between zircon and alumina will be difficult to complete, and Fig. 12b shows a large unreacted zircon grain in the RBM layer. Phase evolution in RBM samples with different zircon precursor particle sizes was therefore studied by X-ray diffraction analysis. The phase contents of both zirconia (M1) and mullite (M2) were estimated using the following equations: 0 2 4 6 8 10 12 0.10 1.00 10.00 Frequency % Particle size (µm) (a) (b) (c) Fig. 9 (a) Particle size distribution for the reaction-bonded aluminum oxide (RBAO) powder mixture after attrition milling; (b) The as received Al metal powder; (c) The RBAO powder mixture after attrition milling Fig. 8 Discrete 0001 pole figure (a) and inverse pole figure (b), both measured from electron back-scatter diffraction (EBSD) data J Mater Sci (2006) 41:7425–7436 7433 123
743 J Mater Sci(2006)41:7425-7436 5000 Mi I{m-ZrO2(111+{t-ZrO2(111 ZrO2(111+I{-ZrO2(111+I{ ErIc4(200)} 27 2000 MUllite(210)1 IMullite(210))+IZrSio4(200)1 The results are summarized in Table 3. with a sub-micron zircon as starting powder, the reaction between zircon and alumina was much closer to completion than for the case of coar zIrcon Dow der. As shown in Fig. 12c, the rBM layer made fror RBAOTA finer zircon powder is composed of only mullite (matrix) and zirconia(brighter and in round shape) grains. From the XRD results, it has been found that up to 95 of the zircon was reacted to form mullite mated to be 93-95 TD% by comparing the results of X-ray phase analysis with the density measurements However, it should be noted that some glassy phase may be present in the rBM layers, which would not etected by XRD and would reduce the residual Fig 10(a) XRD powder patterns from a reaction-bonded porosity. aluminum oxide(RBAo)sample before and after reaction The above results show that we have successfully bonding:(b)optical micrograph of the cross-section of a RBAO/ fabricated three types of oxide ceramic laminates, TA laminate which were designed for different stress states and are micrograph of the cross- (b) section of a sintered alumina- toughened zirconia/textured alumina(ATZ/TA)laminate b) back scattering electron image of the ATZ layer; the atz layer; (d)Zr EDs map of the ATZ laye 2 Springer
M1 ¼ I mf gþ ZrO2ð111Þ I t f g ZrO2ð111Þ I mf gþ ZrO2ð111Þ I t f gþ ZrO2ð111Þ If g ZrSiO4ð200Þ ð6Þ M2 ¼ If g Mullite ð210Þ If g Mullite ð210Þ þ If g ZrSiO4ð200Þ ð7Þ The results are summarized in Table 3. With a sub-micron zircon as starting powder, the reaction between zircon and alumina was much closer to completion than for the case of coarse zircon powder. As shown in Fig. 12c, the RBM layer made from finer zircon powder is composed of only mullite (matrix) and zirconia (brighter and in round shape) grains. From the XRD results, it has been found that up to 95% of the zircon was reacted to form mullite and zirconia. The final laminate density was estimated to be 93–95 TD% by comparing the results of X-ray phase analysis with the density measurements. However, it should be noted that some glassy phase may be present in the RBM layers, which would not be detected by XRD and would reduce the residual porosity. The above results show that we have successfully fabricated three types of oxide ceramic laminates, which were designed for different stress states and are Fig. 11 (a) Optical micrograph of the crosssection of a sintered aluminatoughened zirconia/textured alumina (ATZ/TA) laminate; (b) back scattering electron image of the ATZ layer; (c) Al energy dispersive X-ray spectroscopy (EDS) map of the ATZ layer; (d) Zr EDS map of the ATZ layer 0 1000 2000 3000 4000 5000 30 40 50 60 70 80 90 RBAO20/green RBAO/final Intensity (a.u.) 2θ(degree) (a) (b) Fig. 10 (a) XRD powder patterns from a reaction-bonded aluminum oxide (RBAO) sample before and after reaction bonding; (b) optical micrograph of the cross-section of a RBAO/ TA laminate 7434 J Mater Sci (2006) 41:7425–7436 123