Availableonlineatwww.sciencedirect.com °scⅰ ence Direct COMPOSITES SCIENCE AND TECHNOLOGY ELSEVIER Composites Science and Technology 67(2007)1930-1938 w.elsevier. com/locate/compscitech Processing optimisation and fracture behaviour of layered ceramic composites with highly compressive layers R. Bermejo,, C. Baudin ,R. Moreno, L Llanes, A.J. Sanchez-Herencia epartamento de clew b Instituto de Ceramica y Vidrio(CSIC). CIKelsen 5, 28049 Madrid, Spain Departamento de Ciencia de Materiales e Ingenieria Metalurgica (UPC), Aoda. Diagonal 647(ETSEIB)08028 Barcelona, spain Received 23 June 2006: received in revised form 3 October 2006: accepted 8 October 2006 Available online 20 November 2006 Abstract Multilayer ceramics with five thick layers of alumina-YTZP and four thin layers of alumina-pure zirconia have been fabricated by slip casting and mechanically characterized in order to evaluate the influence of residual stresses on the fracture behaviour. In doing so, pro- essing parameters are optimised in terms of suspension stabilization and well-dispersion of the zirconia particles within the alumina ayers. Two different compressive residual stress values are attained through sequential slip casting of specimens with thick/thin layer thickness ratios of 5.4 and 9.5, by controlling the stability and casting time of aqueous slurries Flexural tests conducted on indented monoliths and multilayered materials show a stepwise failure in the laminates associated with the bifurcation of the impinging crack when interacting with the internal layers, this phenomenon being evidenced even for the case of laminates with very thin compressive layers where edge cracking is not observed. c 2006 Elsevier Ltd. All rights reserved. Keywords: E Powder processing: A Layered structures; Phase transformation; C Residual stress; B Fracture 1. Introduction design approach. If mechanical resistance is sought, com- pressive stresses are to be located at the surface [7, 8] hence In the last years layered configurations have been stud- fracture strength increases as a result of the corresponding ied as a structural option for improving the mechanical stress state superposition acting on the intrinsic critical sur behaviour and reliability of ceramics [1]. Some of the lam- face flaw On the other hand, if the compressive residual inates proposed include reinforcement mechanisms based stresses are induced in the internal layers [9-11], damage on either the deflection of opagating crack by weak tolerance and, consequently, reliability is the property pri- interlayers [24] or through the variation of the elastic marily favoured. From this perspective, attainment of a modulus within the material [5, 6]. Another case refers to threshold strength, i.e. a failure stress that is independent layered structures with strongly joined interfaces where of the original processing flaw size, is a sound evidence of an alternate tensile-compressive residual stress state may the potential effectiveness of this approach [ll] arise during cooling from sintering. Such stress field can Residual stresses in ceramic laminates can be due to dif- be controlled by designing the proper thickness, composi- ferent factors, either intrinsic like epitaxial growth, varia- tion and/or distribution of the layers in order to tailor tions of density or volume, densification or oxidation at the mechanical properties depending on the attempted the surface etc or extrinsic such as thermal or thermo- plastic strains developed during cooling or external forces and momentums. The most common approach is that asso- Corresponding author. Address: Institut fur Struktur- und Fun ionskeramik. Montanuniversitat Leoben. Peter-Tunner Strasse 5. A-8700 ciated with the differences in the coefficient of thermal expansion(Cte) between adjacent layers. Here, although E-mail address: raul bermejo@mu-leoben at(R. Bermejo) during sintering it is considered that stresses are negligible 0266-3538S. see front matter a 2006 Elsevier Ltd. All rights reserved doi: 10.1016/j. compscitech. 2006.10.010
Processing optimisation and fracture behaviour of layered ceramic composites with highly compressive layers R. Bermejo a,*, C. Baudı´n b , R. Moreno b , L. Llanes a , A.J. Sa´nchez-Herencia b a Departamento de Ciencia de Materiales e Ingenierı´a Metalu´rgica (UPC), Avda. Diagonal 647 (ETSEIB) 08028 Barcelona, Spain b Instituto de Cera´mica y Vidrio (CSIC), C/Kelsen 5, 28049 Madrid, Spain Received 23 June 2006; received in revised form 3 October 2006; accepted 8 October 2006 Available online 20 November 2006 Abstract Multilayer ceramics with five thick layers of alumina-YTZP and four thin layers of alumina-pure zirconia have been fabricated by slip casting and mechanically characterized in order to evaluate the influence of residual stresses on the fracture behaviour. In doing so, processing parameters are optimised in terms of suspension stabilization and well-dispersion of the zirconia particles within the alumina layers. Two different compressive residual stress values are attained through sequential slip casting of specimens with thick/thin layer thickness ratios of 5.4 and 9.5, by controlling the stability and casting time of aqueous slurries. Flexural tests conducted on indented monoliths and multilayered materials show a stepwise failure in the laminates associated with the bifurcation of the impinging crack when interacting with the internal layers, this phenomenon being evidenced even for the case of laminates with very thin compressive layers where edge cracking is not observed. 2006 Elsevier Ltd. All rights reserved. Keywords: E. Powder processing; A. Layered structures; Phase transformation; C. Residual stress; B. Fracture 1. Introduction In the last years layered configurations have been studied as a structural option for improving the mechanical behaviour and reliability of ceramics [1]. Some of the laminates proposed include reinforcement mechanisms based on either the deflection of a propagating crack by weak interlayers [2–4] or through the variation of the elastic modulus within the material [5,6]. Another case refers to layered structures with strongly joined interfaces where an alternate tensile–compressive residual stress state may arise during cooling from sintering. Such stress field can be controlled by designing the proper thickness, composition and/or distribution of the layers in order to tailor the mechanical properties depending on the attempted design approach. If mechanical resistance is sought, compressive stresses are to be located at the surface [7,8], hence fracture strength increases as a result of the corresponding stress state superposition acting on the intrinsic critical surface flaw. On the other hand, if the compressive residual stresses are induced in the internal layers [9–11], damage tolerance and, consequently, reliability is the property primarily favoured. From this perspective, attainment of a threshold strength, i.e. a failure stress that is independent of the original processing flaw size, is a sound evidence of the potential effectiveness of this approach [11]. Residual stresses in ceramic laminates can be due to different factors, either intrinsic like epitaxial growth, variations of density or volume, densification or oxidation at the surface, etc., or extrinsic such as thermal or thermoplastic strains developed during cooling or external forces and momentums. The most common approach is that associated with the differences in the coefficient of thermal expansion (CTE) between adjacent layers. Here, although during sintering it is considered that stresses are negligible 0266-3538/$ - see front matter 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.compscitech.2006.10.010 * Corresponding author. Address: Institut fu¨r Struktur- und Funktionskeramik, Montanuniversita¨t Leoben, Peter-Tunner Strasse 5, A-8700 Leoben, Austria. E-mail address: raul.bermejo@mu-leoben.at (R. Bermejo). www.elsevier.com/locate/compscitech Composites Science and Technology 67 (2007) 1930–1938 COMPOSITES SCIENCE AND TECHNOLOGY
R Bermejo et al. Composites Science and Technology 67(2007)1930-1938 due to the accommodation of strain mismatch by mass ponent as they oppose to crack growth [18, 19] and/or may transport mechanisms, as the temperature decreases, the develop a threshold strength(high reliability)[ll]. On the differences in the Cte (ai) should promote a differential other hand, tensile stresses should be subtracted to the strain between layers. In addition to this strain source, strength of the material, and if they overpass a critical other strain differences should be considered as those due value tunnelling cracks will appear and consequently the to phase transformations(Act[7, 10]or chemical reactions mechanical response will degrade [20]. Following these (Aer)[12]inside one particular layer. Hence, the final strain ideas, thin compressive layers are desirable, as they will cre difference between two given layers A and B, after cooling ate an additional reinforcement as well as diminish the from a reference temperature Trer(above which residual associated residual tensile stress. Moreover, the thickness stresses are negligible) down to a temperature T, may be of the layers are referred to other observations related to the residual stresses, such as edge crack and crack bifurca where AT indicates the difference in temperaure be (1) tion, for which a critical thickness"te"has to be surpassed △E=(x4-xB)△T+△a+Aer [3,16,21 the reference state(Tref) and the actual temperature i en In order to satisfy the particular design requirements the number, thickness and composition of the layers should be In ceramic laminates with strong interfaces the differ- controlled. In this regard, colloidal processing methods tion [7, 14] have been commonly used to develop residual aiming to manufacture reliable laminated structures. Those stresses. Regarding the latter, as zirconia cools down from methods include tape-casting [22-24], centrifugal casting the sintering temperature, it transforms from the tetragonal [25, 26], sequential slip casting [27] and electrophoretic to the monoclinic phase with an expansion of about deposition (EPD)[28, 29] among others.All these vol%. The magnitude of the transformation can be con- approaches are based on the preparation of stable slurries trolled by either adding small amounts of stabilizers like with specific compositions that are piled up by adding a Y2 O3 or Cao among others [lo] or varying the amount layer to a previously formed one Stable slurries that ensure of zirconia included inside the composite [15]. For the case a homogeneous and well-dispersed composition are of alumina-zirconia based ceramics, the zirconia expansion obtained by controlling the interparticle potentials devel inside the alumina matrix has been used as a stress devel- oped within the liquid media [30, 31]. The layer thickness oper to change the material fracture behav In this is controlled by adjusting the processing parameter associ sense, the crack propagation has been varied by tension- ated with the corresponding technique(e.g. casting time, compression states [14], crack bifurcation [16] or threshold blades gap, amount of slurry, etc ength development through compressive residual stres- In this work, multilayered materials composed of thick layers with low residual tensile stresses and thin internal For a multilayer system composed of n layers of compo- layers with high residual compressive stresses are fabricated sition A and thickness ta and (n-1)layers of composition by sequential slip casting, aiming to study the influence of B and thickness tb, the residual stress within each layer can the layers thickness on the fracture behaviour of the lay- be calculated by ered architecture. Processing parameters are optimised △E and layer thickness is controlled by adjusting the casting (2) time of stable slurries. The value of the residual stresses is tailored by varying the thickness of the thin layers. Mate- △Eb rials are tested by the indentation-strength method. M lithic specimens are used to collect data for residual stress evaluation and for comparative purposes where E=E/(I-vi), being E; the Young's modulus and vi the Poisson ratio of a given layer. The stress in one layer 2. Experimental is related to the stress in the adjacent one by Starting powders were submicron-sized Al2O3(HPA 0.5, 0b=-(n-1)hb (4) Condea, USA)with a mean particle size(dso) of 0.3 um, tetragonal zirconia polycrystals(Y-TZP)with 3 mol% of For the case of tb ta, then oa-0, i.e. if thin layers are Y2O3(TZ-3YS, Tosoh, Japan)with dso=0. 4 um, and pure inserted between thick ones, the stresses inside the latter are ZrO2(TZ-0, Tosoh, Japan) with dso=0.3 um. Slurries negligible. This suggests the fabrication of laminar ceram- were prepared to a solid loading of 36.5 vol% by mixing ics with thin layers subjected to high internal compressive starting powders with DI water, containing the desired stresses combined with thick layers exhibiting tensile resid- amount of a commercial acrylic based polyelectrolyte ual stresses whose effect on the final strength of the material (Duramax D-3021, Rohm& Haas, USA). For dispersion, slurries were ultrasonicated using a 400 w sonotrode It has been shown that compressive stresses are usually (UP400S, Hielscher, Germany) with magnetic stirring for beneficial for the mechanical performance of a given com- 2 min, and left stirring for at least 4 h before use. Colloidal
due to the accommodation of strain mismatch by mass transport mechanisms, as the temperature decreases, the differences in the CTE (ai) should promote a differential strain between layers. In addition to this strain source, other strain differences should be considered as those due to phase transformations (Det) [7,10] or chemical reactions (Der) [12] inside one particular layer. Hence, the final strain difference between two given layers A and B, after cooling from a reference temperature Tref (above which residual stresses are negligible) down to a temperature Ti, may be expressed as De ¼ ðaA aBÞDT þ Det þ Der ð1Þ where DT indicates the difference in temperature between the reference state (Tref) and the actual temperature Ti. In ceramic laminates with strong interfaces the differences in a [11,13] as well as the zirconia phase transformation [7,14] have been commonly used to develop residual stresses. Regarding the latter, as zirconia cools down from the sintering temperature, it transforms from the tetragonal to the monoclinic phase with an expansion of about 4 vol%. The magnitude of the transformation can be controlled by either adding small amounts of stabilizers like Y2O3 or CaO among others [10] or varying the amount of zirconia included inside the composite [15]. For the case of alumina–zirconia based ceramics, the zirconia expansion inside the alumina matrix has been used as a stress developer to change the material fracture behaviour. In this sense, the crack propagation has been varied by tension– compression states [14], crack bifurcation [16], or threshold strength development through compressive residual stresses [11,17]. For a multilayer system composed of n layers of composition A and thickness ta and (n 1) layers of composition B and thickness tb, the residual stress within each layer can be calculated by ra ¼ DeE0 a 1 þ E0 anta E0 bðn1Þtb ð2Þ rb ¼ DeE0 b 1 þ E0 bðn1Þtb E0 anta ð3Þ where E0 i ¼ Ei=ð1 miÞ, being Ei the Young’s modulus and mi the Poisson ratio of a given layer. The stress in one layer is related to the stress in the adjacent one by rb ¼ ra nta ðn 1Þtb ð4Þ For the case of tb ta, then ra ! 0, i.e. if thin layers are inserted between thick ones, the stresses inside the latter are negligible. This suggests the fabrication of laminar ceramics with thin layers subjected to high internal compressive stresses combined with thick layers exhibiting tensile residual stresses whose effect on the final strength of the material is negligible. It has been shown that compressive stresses are usually beneficial for the mechanical performance of a given component as they oppose to crack growth [18,19] and/or may develop a threshold strength (high reliability) [11]. On the other hand, tensile stresses should be subtracted to the strength of the material, and if they overpass a critical value tunnelling cracks will appear and consequently the mechanical response will degrade [20]. Following these ideas, thin compressive layers are desirable, as they will create an additional reinforcement as well as diminish the associated residual tensile stress. Moreover, the thickness of the layers are referred to other observations related to the residual stresses, such as edge crack and crack bifurcation, for which a critical thickness ‘‘tc’’ has to be surpassed [13,16,21]. In order to satisfy the particular design requirements the number, thickness and composition of the layers should be controlled. In this regard, colloidal processing methods have proved to be useful to tailor the mentioned variables aiming to manufacture reliable laminated structures. Those methods include tape-casting [22–24], centrifugal casting [25,26], sequential slip casting [27] and electrophoretic deposition (EPD) [28,29] among others. All these approaches are based on the preparation of stable slurries with specific compositions that are piled up by adding a layer to a previously formed one. Stable slurries that ensure a homogeneous and well-dispersed composition are obtained by controlling the interparticle potentials developed within the liquid media [30,31]. The layer thickness is controlled by adjusting the processing parameter associated with the corresponding technique (e.g. casting time, blades gap, amount of slurry, etc.). In this work, multilayered materials composed of thick layers with low residual tensile stresses and thin internal layers with high residual compressive stresses are fabricated by sequential slip casting, aiming to study the influence of the layers thickness on the fracture behaviour of the layered architecture. Processing parameters are optimised and layer thickness is controlled by adjusting the casting time of stable slurries. The value of the residual stresses is tailored by varying the thickness of the thin layers. Materials are tested by the indentation-strength method. Monolithic specimens are used to collect data for residual stress evaluation and for comparative purposes. 2. Experimental Starting powders were submicron-sized Al2O3 (HPA 0.5, Condea, USA) with a mean particle size (d50) of 0.3 lm, tetragonal zirconia polycrystals (Y-TZP) with 3 mol% of Y2O3 (TZ-3YS, Tosoh, Japan) with d50 = 0.4 lm, and pure ZrO2 (TZ-0, Tosoh, Japan) with d50 = 0.3 lm. Slurries were prepared to a solid loading of 36.5 vol% by mixing starting powders with DI water, containing the desired amount of a commercial acrylic based polyelectrolyte (Duramax D-3021, Rohm& Haas, USA). For dispersion, slurries were ultrasonicated using a 400 W sonotrode (UP400S, Hielscher, Germany) with magnetic stirring for 2 min, and left stirring for at least 4 h before use. Colloidal R. Bermejo et al. / Composites Science and Technology 67 (2007) 1930–1938 1931
R. Bermejo et aL Composites Science and Technology 67(2007)1930-1938 stability was studied through zeta potential measurements was determined on monolithic materials from the reso- by using the laser Doppler velocimetry principle(Zetasizer nance frequency of bars tested in flexure by impact [33] NanoZS, Malvern, UK). For these measurements the sus-( Grindo Sonic MK5, J W. Lemmens-Electronica N.V., Bel pensions were diluted to a powder concentration of gium) following the guidelines provided by ASTM E 1876- 99 and ENV-843-2. Indentation tests were performed using Suspensions were slip cast in a plaster of Paris mould a Vickers indenter(Microtest, Spain)with a displacement with only one filtrating surface in order to obtain rate of 0. 1 mm/min up to reach a maximum load of 7 cm x 7 cm plates. Monoliths of Al2O3 with 5 vol% Y- 100 N and holding time of 10 s. Three indentations were TZP (labelled as ATZ) and Al,O3 with 30 vol% of Tz-0 placed in the middle of the outer layers along the longitu- (referred to as AMZ) were prepared. Sequential slip casting dinal direction of the sample with an offset separation of [27, 32] was used to fabricate laminates composed of 5 thick 2 mm to avoid any crack interaction. Indented samples layers of ATZ alternated with 4 thin layers of AMZ. The were fractured under four-point bendingthe loading axis thickness of the layers was controlled from the measure- normal to the layer plane) with inner and outer spans of ment of the wall thickness after different casting times for 15 mm and 30 mm, respectively. Tests were carried out both AtZ and AMZ individual suspensions. According under displacement control using a universal testing to filtration kinetic mechanisms, the thickness of a cast machine(Microtest, Spain)with a load cell of 5 Kn at a body is related to the square root of the casting time by cross head speed of 0.5 mm/min. The corresponding (5) oad-displacement curves were recorded by a software cou- pled to the testing set-up. The stress distribution under where e is the wall thickness, k is a casting constant and t is four-point bending on a prismatic bar formed by different With these data a first approximation was layers was calculated following the expression given by made regarding the casting time of each layer to obtain the [34] desired thicknesses in the green laminate. The laminate was EM cast, then sintered, and finally cut and polished in order to al y=rl measure the thickness of the layers by means of scanning electron microscopy (DSM-950, Zeiss, Germany). The where Ei is the Young's modulus of the corresponding measured thicknesses were used to recalculate new casting layer, M is the moment for the case of four-point bending constants for both slurry compositions. Using the recalcu- tests(M=FL, where Fis the applied load and the distance lated casting constants, laminates were cast to obtain thick between inner and outer spans), yna is the position of the layers of 630 um and intermediate thin layers of AMZ with neutral axis according to thicknesses of 126 um and 63 um, i.e. laminates with thick ness ratios of about 1/5 and 1/10 respectively. In all cases 1+t the outer ATZ layers were calculated to be thicker than J 2.∑=1E1·1·B the final desired dimensions in order to allow further grind- and eI the flexural rigidity of the multilayer calculated for g processes during preparation of beams for bending bending perpendicular to the interfaces between the layers Samples were pre-sintered at 900C for 30 min and cut as given by into bars of 5 mm x 5 mm x 50 mm before sintering. The resulting testing bars were then sintered at 1550 C for 2h El=3 2Er. B using heating and cooling rates of 5 C/min. The sintering =1 shrinkage and the phase transformation of the composites containing pure ZrO2 were determined on monolithic sam ples of 10 mm length and a cross section of 5 x 5 mm, using where ti is the corresponding layer thickness and B the a dilatometer(DI-24, Adamel Lhomargy, France)with an specimen width [35,36 alumina rod and selecting the same thermal cycle as the one used to sinter the laminates. Density was measured by 3. Results and discussion the Archimedes method using mercury for the green sam ples and water for the sintered ones. X-Ray diffraction 3. 1. Preparation and characterization of monoliths (XRD) was performed for phase identification by means of a diffractometer(D-5000, Siemens, Germany) with the As the transformation of ZrO,(TZ-0) from tetragonal Kocu radiation. After sintering, specimens were polished to monoclinic during cooling from sintering occurs sponta with diamond paste down to I um for SEM observation. neously in a sharp temperature range, the presence of inho- In order to avoid structural changes due to the zirconia mogeneities can generate local stress concentration that transformation, no thermal etching was applied to the may diminish the mechanical properties of the material samples For this reason ZrOz particles have to be well-dispersed For mechanical characterization bars of approximately inside the Al2O3 matrix. Such dispersion is possible 3.6 mm x 3. 2 mm x 40 mm were prepared. Elastic modulus through the addition of a deflocculant that develops an
stability was studied through zeta potential measurements by using the laser Doppler velocimetry principle (Zetasizer NanoZS, Malvern, UK). For these measurements the suspensions were diluted to a powder concentration of 100 mg/l. Suspensions were slip cast in a plaster of Paris mould with only one filtrating surface in order to obtain 7 cm · 7 cm plates. Monoliths of Al2O3 with 5 vol% YTZP (labelled as ATZ) and Al2O3 with 30 vol% of TZ-0 (referred to as AMZ) were prepared. Sequential slip casting [27,32] was used to fabricate laminates composed of 5 thick layers of ATZ alternated with 4 thin layers of AMZ. The thickness of the layers was controlled from the measurement of the wall thickness after different casting times for both ATZ and AMZ individual suspensions. According to filtration kinetic mechanisms, the thickness of a cast body is related to the square root of the casting time by e ¼ kt1=2 ð5Þ where e is the wall thickness, k is a casting constant and t is the casting time. With these data a first approximation was made regarding the casting time of each layer to obtain the desired thicknesses in the green laminate. The laminate was cast, then sintered, and finally cut and polished in order to measure the thickness of the layers by means of scanning electron microscopy (DSM-950, Zeiss, Germany). The measured thicknesses were used to recalculate new casting constants for both slurry compositions. Using the recalculated casting constants, laminates were cast to obtain thick layers of 630 lm and intermediate thin layers of AMZ with thicknesses of 126 lm and 63 lm, i.e. laminates with thickness ratios of about 1/5 and 1/10 respectively. In all cases the outer ATZ layers were calculated to be thicker than the final desired dimensions in order to allow further grinding processes during preparation of beams for bending tests. Samples were pre-sintered at 900 C for 30 min and cut into bars of 5 mm · 5 mm · 50 mm before sintering. The resulting testing bars were then sintered at 1550 C for 2 h using heating and cooling rates of 5 C/min. The sintering shrinkage and the phase transformation of the composites containing pure ZrO2 were determined on monolithic samples of 10 mm length and a cross section of 5 · 5 mm, using a dilatometer (DI-24, Adamel Lhomargy, France) with an alumina rod and selecting the same thermal cycle as the one used to sinter the laminates. Density was measured by the Archimedes method using mercury for the green samples and water for the sintered ones. X-Ray diffraction (XRD) was performed for phase identification by means of a diffractometer (D-5000, Siemens, Germany) with the KaCu radiation. After sintering, specimens were polished with diamond paste down to 1 lm for SEM observation. In order to avoid structural changes due to the zirconia transformation, no thermal etching was applied to the samples. For mechanical characterization bars of approximately 3.6 mm · 3.2 mm · 40 mm were prepared. Elastic modulus was determined on monolithic materials from the resonance frequency of bars tested in flexure by impact [33] (GrindoSonic MK5, J.W. Lemmens-Electronica N.V., Belgium) following the guidelines provided by ASTM E 1876- 99 and ENV-843-2. Indentation tests were performed using a Vickers indenter (Microtest, Spain) with a displacement rate of 0.1 mm/min up to reach a maximum load of 100 N and holding time of 10 s. Three indentations were placed in the middle of the outer layers along the longitudinal direction of the sample with an offset separation of 2 mm to avoid any crack interaction. Indented samples were fractured under four-point bending (the loading axis normal to the layer plane) with inner and outer spans of 15 mm and 30 mm, respectively. Tests were carried out under displacement control using a universal testing machine (Microtest, Spain) with a load cell of 5 KN at a cross head speed of 0.5 mm/min. The corresponding load–displacement curves were recorded by a software coupled to the testing set-up. The stress distribution under four-point bending on a prismatic bar formed by different layers was calculated following the expression given by [34]: ri;y ¼ EiM EI ðy ynaÞ ð6Þ where Ei is the Young’s modulus of the corresponding layer, M is the moment for the case of four-point bending tests (M = Fl, where F is the applied load and l the distance between inner and outer spans), yna is the position of the neutral axis according to yna ¼ Pn i¼1Ei ti B 2 Pi1 j¼1tj þ ti 2 Pn i¼1Ei ti B ð7Þ and EI the flexural rigidity of the multilayer calculated for bending perpendicular to the interfaces between the layers as given by EI ¼ 1 3 Xn i¼1 Ei B Xi j¼1 tj yna !3 þ yna Xi1 j¼1 tj !3 0 @ 1 A ð8Þ where ti is the corresponding layer thickness and B the specimen width [35,36]. 3. Results and discussion 3.1. Preparation and characterization of monoliths As the transformation of ZrO2 (TZ-0) from tetragonal to monoclinic during cooling from sintering occurs spontaneously in a sharp temperature range, the presence of inhomogeneities can generate local stress concentration that may diminish the mechanical properties of the material. For this reason ZrO2 particles have to be well-dispersed inside the Al2O3 matrix. Such dispersion is possible through the addition of a deflocculant that develops an 1932 R. Bermejo et al. / Composites Science and Technology 67 (2007) 1930–1938
R Bermejo et al. Composites Science and Technology 67(2007)1930-1938 1933 electrical charge on the surface of the powders, leading to a AtZ compacts recorded during a complete sintering cycle good stability of the final slurry. Fig. I shows the zeta In this figure a close-up view of the transformation zone is potential vs. deflocculant content for both Al_O3 and also plotted. It can be observed that the ATZ compact ZrO, powders in water. The TZ-0 material presents a neg- starts to sinter at 1000C and shrinks 13% whereas the ative zeta potential for all the measurement range, whereas AMZ starts to sinter around 1100C and shrinks 14% the sign of the surface charge of Al2O3 changes for deflocc- These small differences in the starting temperature for sin- ulant contents of 0.5-0.6%, thus indicating that 0.6% is the tering and the final shrinkage can be explained due to the lower content of deflocculant required to avoid heterocoag- higher packing density of the atz green bodies. At ulation. It can also be observed that over a deflocculant 1330C both samples reach the same shrinkage rate. The content of 0. 8 wt% the absolute value of zeta potential is most interesting differences between those curves are higher than 30 mV for both powders, which is accepted observed in the cooling ramp. While AMz sample rapidly as a limit value that indicates a high stabilization of casting expands at 730C due to the martensitic phase transfor Following the above ideas, suspensions of ATZ and mation of the zirconia particles from tetragonal to mono- AMZ with 36.5 vol%/(approximately 70 wt%)solids dis- clinic, the AtZ sample maintains linear shrinkage during persed with 0. 8 wt% polyelectrolyte were prepared and slip cooling. The magnitude of the expansion due to phase cast in order to obtain the kinetic equations used to calcu- transformation in the AMZ samples is of 0. 16%. XRD late the casting time for the laminates. Kinetic constants analysis indicates that most of the zirconia present in the for ATZ and AMZ slurries, determined as the slope of AMZ samples has transformed into the monoclinic phase the wall thickness- casting time curve, were measured as and only a 5-10% remaining tetragonal zirconia was 0.00303 and 0.00216 mm/s, respectively. Green densities detected in this sample. Density measurements of A of cast monolithic samples were 69%th for ATZ andand AMZ compacts are listed in Table l, along with the 62%th for AMZ compacts. data obtained from the characterization of the sintered Monolithic samples were used to separately study the monolithic samples [36]. ntering behaviour of the layers forming the laminates Fig. 2 shows the dilatometric curves for both AMZ and 3. 2. Sequential slip casting of laminates with residual stresses On the basis of the kinetic constants calculated from the monolithic sample characterization, a multilayer piece with o Zrt predicted layer thicknesses of 630 um for ATZ and 63 um for AMZ was cast. Fig. 3 plots the variation of wall thick wt%D-3021 ness as a function of casting time for both AMz and atZ suspensions, as well as the step-like procedure used to cast 1.0 the multilayers. In order to form the first ATZ layer of a certain thickness, the casting time was fixed following the ATZ casting curve(ATZ equation in Fig. 3). After that, the suspension left was poured out of the mould and the new slip to form the AMz layer poured on top of the Fig.I.Variation of zeta potential with the dispersant content for the already formed ATZ layer. The thickness of this new powders used to prepare the slurries. AMZ layer corresponds to a fixed casting time given by the aMz casting curve(AMZ equation in Fig. 3). This procedure was repeated to alternatively form the rest of the layers. Hence, for casting a new layer with a thickness eAtz, the atZ slurry was maintained in the mould for a casting time tATz, dictated by the AtZ casting curve, as shown in the referred figure. Likewise, attempting to form an AMz layer of thickness eAMz, the AMz slurry was drain cast for a time tamz, following the AMz casting curve(Fig 3). After casting the whole multilayer, the green plate was dried in air inside the mould for two days until it shrank enough to free from the mould walls. Aft 02004006008001000120014001600 ing the test-piece laminate, the thickness of the resulting layers was assessed and the kinetic constants then refined yielding new values of 0.00433 and 0.01233 for ATZ and Fig. 2. Dilatometric curves from monolithic samples of ATz and AMz AMz layers, respectively. Considering those recalculated materials. A close-up of the 900-600C interval during cooling is also constants and the sintering shrinkage of each compact plotted to show the tetragonal to monoclinic zirconia phase Fig. 2), two sets of laminates composed of 5 thick ATZ transfe layers of 550 um and 4 thin AMZ layers of 110 um and
electrical charge on the surface of the powders, leading to a good stability of the final slurry. Fig. 1 shows the zeta potential vs. deflocculant content for both Al2O3 and ZrO2 powders in water. The TZ-0 material presents a negative zeta potential for all the measurement range, whereas the sign of the surface charge of Al2O3 changes for deflocculant contents of 0.5–0.6%, thus indicating that 0.6% is the lower content of deflocculant required to avoid heterocoagulation. It can also be observed that over a deflocculant content of 0.8 wt% the absolute value of zeta potential is higher than 30 mV for both powders, which is accepted as a limit value that indicates a high stabilization of casting slips. Following the above ideas, suspensions of ATZ and AMZ with 36.5 vol% (approximately 70 wt%) solids dispersed with 0.8 wt% polyelectrolyte were prepared and slip cast in order to obtain the kinetic equations used to calculate the casting time for the laminates. Kinetic constants for ATZ and AMZ slurries, determined as the slope of the wall thickness – casting time curve, were measured as 0.00303 and 0.00216 mm2 /s, respectively. Green densities of cast monolithic samples were 69%th for ATZ and 62%th for AMZ compacts. Monolithic samples were used to separately study the sintering behaviour of the layers forming the laminates. Fig. 2 shows the dilatometric curves for both AMZ and ATZ compacts recorded during a complete sintering cycle. In this figure a close-up view of the transformation zone is also plotted. It can be observed that the ATZ compact starts to sinter at 1000 C and shrinks 13% whereas the AMZ starts to sinter around 1100 C and shrinks 14%. These small differences in the starting temperature for sintering and the final shrinkage can be explained due to the higher packing density of the ATZ green bodies. At 1330 C both samples reach the same shrinkage rate. The most interesting differences between those curves are observed in the cooling ramp. While AMZ sample rapidly expands at 730 C due to the martensitic phase transformation of the zirconia particles from tetragonal to monoclinic, the ATZ sample maintains linear shrinkage during cooling. The magnitude of the expansion due to phase transformation in the AMZ samples is of 0.16%. XRD analysis indicates that most of the zirconia present in the AMZ samples has transformed into the monoclinic phase, and only a 5–10% remaining tetragonal zirconia was detected in this sample. Density measurements of ATZ and AMZ compacts are listed in Table 1, along with the data obtained from the characterization of the sintered monolithic samples [36]. 3.2. Sequential slip casting of laminates with residual stresses On the basis of the kinetic constants calculated from the monolithic sample characterization, a multilayer piece with predicted layer thicknesses of 630 lm for ATZ and 63 lm for AMZ was cast. Fig. 3 plots the variation of wall thickness as a function of casting time for both AMZ and ATZ suspensions, as well as the step-like procedure used to cast the multilayers. In order to form the first ATZ layer of a certain thickness, the casting time was fixed following the ATZ casting curve (ATZ equation in Fig. 3). After that, the suspension left was poured out of the mould and the new slip to form the AMZ layer poured on top of the already formed ATZ layer. The thickness of this new AMZ layer corresponds to a fixed casting time given by the AMZ casting curve (AMZ equation in Fig. 3). This procedure was repeated to alternatively form the rest of the layers. Hence, for casting a new layer with a thickness eATZ, the ATZ slurry was maintained in the mould for a casting time tATZ, dictated by the ATZ casting curve, as shown in the referred figure. Likewise, attempting to form an AMZ layer of thickness eAMZ, the AMZ slurry was drain cast for a time tAMZ, following the AMZ casting curve (Fig. 3). After casting the whole multilayer, the green plate was dried in air inside the mould for two days until it shrank enough to free from the mould walls. After sintering the test-piece laminate, the thickness of the resulting layers was assessed and the kinetic constants then refined, yielding new values of 0.00433 and 0.01233 for ATZ and AMZ layers, respectively. Considering those recalculated constants and the sintering shrinkage of each compact (Fig. 2), two sets of laminates composed of 5 thick ATZ layers of 550 lm and 4 thin AMZ layers of 110 lm and Fig. 1. Variation of zeta potential with the dispersant content for the powders used to prepare the slurries. Fig. 2. Dilatometric curves from monolithic samples of ATZ and AMZ recorded using the thermal cycle employed to sinter the multilayered materials. A close-up of the 900–600 C interval during cooling is also plotted to show the tetragonal to monoclinic zirconia phase transformation. R. Bermejo et al. / Composites Science and Technology 67 (2007) 1930–1938 1933
R Bermejo et al Composites Science and Technology 67(2007)1930-1938 Table l ic physical and mechanical characterization of the monolithic samples after sintering at 1550C for 2 h Density (% Young modulus(GPa) m-ZrO2(vol % ATZ 390±10 3.2±0.2 士30 AM 280±30 90-95 士20 Predicted and obtained thicknesses for the two laminated systems fabricated in this work. Predicted thickness has been calculated with the casting time and shrinkage Thick Thin Ratio Thick Thin Ratio Thick Thin 3±999±5543 560±859±69.52 ATZ Thick Layer MZ Thin laver the ones predicted for the thin layers(between 7% and l0%0) Fig 3. Kinetic curves for the ATZ and aMz slips and representation of A close-up of the thin layers is shown in Fig. 5 for both the predicted times to obtain thick ATZ and thin amz layers of 630 um laminate systems. The sharp interface between layers and and 63 um, respectively the homogeneous dispersion of the different phases can also be inferred. In the micrograph showing laminate B 55 um were prepared. They are referred to as systems B and (Fig. 5a) the edge crack can be clearly observed, thus sug- C respectively, in this investigation. A general view of the gesting the presence of high compressive residual stresses cross section for laminates B and C is presented in Fig. 4, within such layer. This phenomenon also indicates that where the bright thin layers correspond to AMZ and the the thickness of the internal AMZ layers in laminate B dark thick ones to ATZ composition. It can be appreciated over the critical thickness(te) value, above which the edge the uniform thickness achieved for both thin and thick lay cracks are induced, and thus crack bifurcation phenome- ers.Table 2 lists the predicted and measured thickness of non at failure is likely to occur [13, 21]. On the other hand, the ATZ and AMZ layers, and the corresponding thickness the SEM micrograph of laminate C(Fig. 5b) shows no ratio for each type of layered system. The predicted thick- edge crack in the centre of the compressive layer, i.e. the ness was calculated taking into account the shrinkage mea- AMZ layer thickness in this case must be lower than"tc sured in the monolithic samples. After SEM examination In order to evaluate and quantify the residual stresses of the laminates. it can be seen that the thickness of the developed in the laminates during cooling, the difference thick layers presents a lower relative error(about 2-3%), in strain for ATZ and AMZ was analysed from the data with respect to the experimentally obtained values, than recorded in the dilatometric experiments(Fig. 6).It was ATZ ATZ AMz ATZ ATZ AMZ AIZ ATZ AMZ AMZ AMZ Fig 4. Cross section of the laminates fabricated with a thickness ratio of (a)5.4 and (b)9.5. The bright thin layers are of AMz whereas the thick dark ones
55 lm were prepared. They are referred to as systems B and C respectively, in this investigation. A general view of the cross section for laminates B and C is presented in Fig. 4, where the bright thin layers correspond to AMZ and the dark thick ones to ATZ composition. It can be appreciated the uniform thickness achieved for both thin and thick layers. Table 2 lists the predicted and measured thickness of the ATZ and AMZ layers, and the corresponding thickness ratio for each type of layered system. The predicted thickness was calculated taking into account the shrinkage measured in the monolithic samples. After SEM examination of the laminates, it can be seen that the thickness of the thick layers presents a lower relative error (about 2–3%), with respect to the experimentally obtained values, than the ones predicted for the thin layers (between 7% and 10%). A close-up of the thin layers is shown in Fig. 5 for both laminate systems. The sharp interface between layers and the homogeneous dispersion of the different phases can also be inferred. In the micrograph showing laminate B (Fig. 5a) the edge crack can be clearly observed, thus suggesting the presence of high compressive residual stresses within such layer. This phenomenon also indicates that the thickness of the internal AMZ layers in laminate B is over the critical thickness (tc) value, above which the edge cracks are induced, and thus crack bifurcation phenomenon at failure is likely to occur [13,21]. On the other hand, the SEM micrograph of laminate C (Fig. 5b) shows no edge crack in the centre of the compressive layer, i.e. the AMZ layer thickness in this case must be lower than ‘‘tc’’. In order to evaluate and quantify the residual stresses developed in the laminates during cooling, the difference in strain for ATZ and AMZ was analysed from the data recorded in the dilatometric experiments (Fig. 6). It was Table 1 Basic physical and mechanical characterization of the monolithic samples after sintering at 1550 C for 2 h Material Density (%) Young modulus (GPa) m-ZrO2 (vol.%) KIC (MPam1/2) Strength (MPa) ATZ 99.5 390 ± 10 – 3.2 ± 0.2 422 ± 30 AMZ 98.5 280 ± 30 90–95 2.6 ± 0.2 90 ± 20 Fig. 3. Kinetic curves for the ATZ and AMZ slips and representation of the predicted times to obtain thick ATZ and thin AMZ layers of 630 lm and 63 lm, respectively. Fig. 4. Cross section of the laminates fabricated with a thickness ratio of (a) 5.4 and (b) 9.5. The bright thin layers are of AMZ whereas the thick dark ones are of ATZ. Table 2 Predicted and obtained thicknesses for the two laminated systems fabricated in this work. Predicted thickness has been calculated with the casting time and shrinkage System Predicted Measured % Error Thick Thin Ratio Thick Thin Ratio Thick Thin B 550 110 5 533 ± 9 99 ± 5 5.4 3 10 C 550 55 10 560 ± 8 59 ± 6 9.5 2 7 1934 R. Bermejo et al. / Composites Science and Technology 67 (2007) 1930–1938