J.Am. Ceran.So,9同1901-1920(2009) DOI:10.11111551-2916.200903278.x urna The Tetragonal-Monoclinic Transformation in Zirconia Lessons learned and future trends Jerome Chevalier and Laurent gremillard' University of Lyon. INSA-Lyon, MATEIS, Villeurbanne FR-69621, France Anil v.Ⅴirka Department of Material Science Engineering. University of Utah, Salt Lake City, Utah 84112 David R. Clarke School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 02138 Zirconia ceramics have found broad applications in a variety L. Introduction energy and biomedical applications because of their unusua combination of strength, fracture toughness, ionic conductivity ZiRals fo hwsll eterne oe the m ost ime disct era moir trans. and low thermal conductivity. These attractive characteristics formation toughening in 1975 heralded new visions for new are largely associated with the stabilization of the tetragonal and cubic phases through alloying with aliovalent ions. The large high-performance applications of zirconia, ranging from bearing and wear applications to thermal barrier coatings (TBCs)to the aliovalent alloying is responsible for both the exceptionally most recently, biomedical applications. The subsequent discov high ionic conductivity and the unusually low, and temperature ery that zirconia could also be toughened by ferroelastic switch- independent, thermal conductivity. The high fracture toughnes g gave further confidence in the application of zirconia exhibited by many of zirconia ceramics is attributed to the con- ceramics in critical applications. Nevertheless, despite the suc- straint of the tetragonal-to-monoclinic phase transformation cess of zirconia in many new applications, it has become appar- and its release during crack propagation. In other zirconia ce- ent that certain zirconia compositions can also have an achilles the tett phase, the high fracture tough heel, namely their propensity to low-temperature degradation ness is associated with ferroelastic domain switching. However (LTD)in the presence of moisture. This is a kinetic phenomenon many of these attractive features of zirconia, especially fracture in which polycrystalline tetragona toughness and strength, are compromised after prolonged expo monoclinic zirconia over a rather narrow but important tem In p iess referm t intermediate temperatures(- c) perature range, typically room temperature to around 400oC, sure to water v to as low-temperature degradation (LTD), depending on the stabilizer, its concentration, and the grain size and initially identified over two decades ago. This is particularly of the ceramic. The transformation occurs by a nucleation and growth process and typically begins at the surfaces of polycrys for zirconia in biomedical applications, such as hip implants talline ceramics. It has all the characteristics of being an iso- and dental restorations. Less well substantiated is the possibility that the same process can also occur in zirconia used in other thermal martensite. Also. although there continues to remain applications, for instance, zirconia thermal barrier coatings af- some uncertainty as to the precise mechanism by which moisture ter long exposure at high temperature. Based on experience with causes destabilization of the tetragonal phase, the observation the failure of zirconia femoral heads as well as studies of LtD. diffusion suggests that the transformation occurs by the hcy that the kinetics of Ltd are similar to those of oxygen vaca it is shown that many of the problems of ltd can be mitigate by the appropriate choice of alloying and or process control diffusion of a moisture species with an activation energy similar to that of oxygen vacancy diffusion. In practical terms, LTD is in effect, an alternative to crack propagation, stress-induced transformation for the transformation from metastable tetra onal to monoclinic(I-m)zirconia(see Fig. 1). In this feature article we describe the role of phase trans- rmations responsible for the impressive combination of me- chanical properties of zirconia, their relationship to equilibriu and metastable phase diagrams, and the phenomenon of Ltd ript No 26208 Received April 27, 2009, ap uthor to whom correspondence she We include the effects of transformations at free surfaces be- cause these affect the surface finish that is important for many Feature
The Tetragonal-Monoclinic Transformation in Zirconia: Lessons Learned and Future Trends Je´roˆme Chevalier and Laurent Gremillardw University of Lyon, INSA-Lyon, MATEIS, Villeurbanne FR-69621, France Anil V. Virkar Department of Material Science & Engineering, University of Utah, Salt Lake City, Utah 84112 David R. Clarke School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 02138 Zirconia ceramics have found broad applications in a variety of energy and biomedical applications because of their unusual combination of strength, fracture toughness, ionic conductivity, and low thermal conductivity. These attractive characteristics are largely associated with the stabilization of the tetragonal and cubic phases through alloying with aliovalent ions. The large concentration of vacancies introduced to charge compensate of the aliovalent alloying is responsible for both the exceptionally high ionic conductivity and the unusually low, and temperature independent, thermal conductivity. The high fracture toughness exhibited by many of zirconia ceramics is attributed to the constraint of the tetragonal-to-monoclinic phase transformation and its release during crack propagation. In other zirconia ceramics containing the tetragonal phase, the high fracture toughness is associated with ferroelastic domain switching. However, many of these attractive features of zirconia, especially fracture toughness and strength, are compromised after prolonged exposure to water vapor at intermediate temperatures (B301–3001C) in a process referred to as low-temperature degradation (LTD), and initially identified over two decades ago. This is particularly so for zirconia in biomedical applications, such as hip implants and dental restorations. Less well substantiated is the possibility that the same process can also occur in zirconia used in other applications, for instance, zirconia thermal barrier coatings after long exposure at high temperature. Based on experience with the failure of zirconia femoral heads, as well as studies of LTD, it is shown that many of the problems of LTD can be mitigated by the appropriate choice of alloying and/or process control. I. Introduction ZIRCONIA has been one of the most important ceramic materials for well over a century but the discovery of transformation toughening in 19751 heralded new visions for new high-performance applications of zirconia, ranging from bearing and wear applications to thermal barrier coatings (TBCs) to, most recently, biomedical applications. The subsequent discovery that zirconia could also be toughened by ferroelastic switching2 gave further confidence in the application of zirconia ceramics in critical applications. Nevertheless, despite the success of zirconia in many new applications, it has become apparent that certain zirconia compositions can also have an Achilles heel, namely their propensity to low-temperature degradation (LTD) in the presence of moisture. This is a kinetic phenomenon in which polycrystalline tetragonal material slowly transforms to monoclinic zirconia over a rather narrow but important temperature range, typically room temperature to around 4001C, depending on the stabilizer, its concentration, and the grain size of the ceramic. The transformation occurs by a nucleation and growth process and typically begins at the surfaces of polycrystalline ceramics. It has all the characteristics of being an isothermal martensite. Also, although there continues to remain some uncertainty as to the precise mechanism by which moisture causes destabilization of the tetragonal phase, the observation that the kinetics of LTD are similar to those of oxygen vacancy diffusion suggests that the transformation occurs by the indiffusion of a moisture species with an activation energy similar to that of oxygen vacancy diffusion. In practical terms, LTD is, in effect, an alternative to crack propagation, stress-induced transformation for the transformation from metastable tetragonal to monoclinic (t–m) zirconia (see Fig. 1). In this feature article we describe the role of phase transformations responsible for the impressive combination of mechanical properties of zirconia, their relationship to equilibrium and metastable phase diagrams, and the phenomenon of LTD. We include the effects of transformations at free surfaces because these affect the surface finish that is important for many Feature D. J. Green—contributing editor w Author to whom correspondence should be addressed. e-mail: laurent.gremillard@ insa-lyon.fr Manuscript No. 26208. Received April 27, 2009; approved 1 July 2009. Journal J. Am. Ceram. Soc., 92 [9] 1901–1920 (2009) DOI: 10.1111/j.1551-2916.2009.03278.x r 2009 The American Ceramic Society
1902 Journal of the American Ceramic Society--Chevalier et al Vol. 92. No 9 zirconia polycrystals. In FSZ zirconia is in its cubic form, the Tetragonal zirconia ceramics form most commonly used in oxygen sensors and fuel cell elec- olytes. It is generally obtained with large concentration of sta bilizers (ie, more than 8 mol% Y2O3). The PSZ consists of anosized tetragonal or monoclinic particles that have precipi tated out in a cubic matrix. Such zirconia ceramics are generally Moist Atmosphere obtained with the addition of lime or magnesia. TZPs are often onsidered as monoliths of tetragonal phase, although they may contain a secondary cubic phase( see Panel A). The majority of TZPs that have been investigated are those stabilized with yttria or ceria ow Temperature Toughening Il. Stabilization and Transformation of the Tetragonal phase Fig 1. Crack propagation-induced transformation and intermediate temperature exposure to moisture are two alternative means by which As mentioned above, stabilization of powders and sintered metastable tetragonal phase can transform to monoclinic phase. ceramics can be achieved by alloying pure zirconia with other oxides. Investigations of the stability of different phases wit applic atages as wvo a thr bine s mals desease app ressons phase diagrams such as those compiled by the American Ce- learnt from investigation of LTD in femoral implants and the ramic Society-NIST project Is echanisms that control it. Much of this article is concerned Alloy stabilization not only enables fabrication of crack-free with the properties of the zirconia-yttria material system be- zrconia but as demonstrated by Gupta et al., sintered bodies cause the majority of the research and development that has of polycrystalline tetragonal zirconia can be prepared and the been performed on zirconia in the last three decades has been tetragonal phase retained down to room temperature even n yttria-stabilized zirconia (YSZ). YSZ ceramics have the best ough the equilibrium phase is monoclinic. These metastable combination of toughness and strength of any of the stabi- trigonal ceramics exhibited tional fracture toughness when the transformation to the monoclinic phase was triggered lized zirconias. Also, and this cannot be overemphasized, it is by a propagating crack. The toughening that can be achieved for undoubtedly due to the early and continued availability of high purity, uniform submicrometer particle size powders from com- different concentrations of yttria is illustrated in Fig. 2, which panies such as Tosoh in Japan summarizes data for transformation-toughened zirconia and ferroelastic toughening. Also shown, superimposed, is the sen- sitivity to aging as a function of yttria concentration. The frac- II. Experimental Procedure ture toughness of monoclinic and cubic zirconia, which is similar o that of window glass, provides a refe The principal properties of zirconia are well known and various toughening through the I-m phase transformation can be co aspects have been reviewed in detail many times since the dis- pared. It is emphasized that the data are obtained from standard covery of transformation toughening by garvie et al. in calcia acture toughness tests such as indentation and double canti stabilized zirconia. For this reason in this section we summarize ever beam tests. in fast fracture conditions under which ltd is and ea principal features of the stabilization of zirconia, the avoided. As indicated in Fig. 2, the attainable fracture toughness raphy of stabilized zirconia, and the relationship be- decreases as the yttria concentration increases. In the context of rmations the metastable phase diagram, the toughening is proportional to The panels describe the essentials of phase equilibria the magnitude of the undercooling below the To(t/m) temper ature(see Panel A for a more detailed explanation). Further Pure zirconium oxide exhibits three allotropes: monoclin more. transformation toughening is restricted to moderate m), which is the stable phase up to 1170.C, where it transforms nperatures, becoming ineffective when the stable phase is to tetragonal (0), and then cubic (c) at temperatures above tragonal and not monoclinic. for these reasons the most 2370C. A comprehensive review of the different structures for tractive compositions for transformation toughening are those zirconia can be found in green et al. The t-m transformation, with low yttria concentrations(but high enough to prevent which is martensitic, has been the subject of the most careful spontaneous f-m transformation during cooling), typically 2- attention, because it usually occurs during the sintering and on 3 mol%Y2O3 both heating and cooling. The t-m transformation is accompa In the ce of mo he transformation of metastable nied by a large shear strain and a large volume increase(see -m can alternatively occur without the passage of a crack In Panel B). Together these create large internal stresses on cooling. this sense, LTD is a competing process to transformation tough large, in fact, that pure zirconia sintered above 1170C inev- ning with the two providing limiting behaviors by which the can either sinter at low temperature for it to remain monoclinic a propagating crack, then one can get enhanced toughening(see during sintering-which leads to a low-strength and toughness "Section Ill(2)"). On the other hand, the transformation ma ceramic--or stabilize the tetragonal or the cubic phases at room be triggered"chemically "by the infusion of water-derived spe temperature by alloying, thereby avoiding the f-m transforma- cies from the surface. The process on a surface is complex tion during cooling. The fundamental approach to the engineer (Fig 3)and results not only in the undesirable transformation ng use of zirconia and avoiding the transformation-induced but also surface roughening, microcracking, and grain pull-out cracking described by Ruff and Ebert" almost a century ago re- Is well as loss of strength-all processes detrimental to struc- mains valid today alloying pure zirconia with another oxide to tural applications. The dilemma facing the alloy designer is that ully or partially stabilize the tetragonal and or the cubic pha the Ysz compositions that are the most attractive for their Calcium, magnesium, yttrium, and cerium oxides have been the fracture toughness are also those that are most susceptible to most widely used stabilizers and lead to a number of different LTD. This is illustrated by the comparison of the fracture microstructures In general, zirconia ceramics may conveniently toughness data with the ltd data in Fig. 2. be classified into three major types according to their micro- The stabilization of the tetragonal phase in polycrystalline structure: FSZ, PSZ, and TZP, standing, respectively, for fully ceramics is, btedly, largely dependent on the mutual elas- stabilized zirconia, partially stabilized zirconia, and tetragonal vided by the surrounding, untransformed
applications as well as the kinetics. We also describe approaches being taken to avoid LTD, or minimize it, based on lessons learnt from investigation of LTD in femoral implants and the mechanisms that control it. Much of this article is concerned with the properties of the zirconia–yttria material system because the majority of the research and development that has been performed on zirconia in the last three decades has been on yttria-stabilized zirconia (YSZ). YSZ ceramics have the best combination of toughness and strength of any of the stabilized zirconias. Also, and this cannot be overemphasized, it is undoubtedly due to the early and continued availability of highpurity, uniform submicrometer particle size powders from companies such as Tosoh in Japan. II. Experimental Procedure The principal properties of zirconia are well known and various aspects have been reviewed in detail many times since the discovery of transformation toughening by Garvie et al. 1 in calciastabilized zirconia. For this reason, in this section we summarize only the principal features of the stabilization of zirconia, the crystallography of stabilized zirconia, and the relationship between mechanical properties and the phase transformations in zirconia. The panels describe the essentials of phase equilibria and the transformation crystallography. Pure zirconium oxide exhibits three allotropes: monoclinic (m), which is the stable phase up to 11701C, where it transforms to tetragonal (t), and then cubic (c) at temperatures above 23701C. A comprehensive review of the different structures for zirconia can be found in Green et al. 3 The tm transformation, which is martensitic, has been the subject of the most careful attention, because it usually occurs during the sintering and on both heating and cooling. The tm transformation is accompanied by a large shear strain and a large volume increase (see Panel B). Together these create large internal stresses on cooling. So large, in fact, that pure zirconia sintered above 11701C inevitably disintegrates by cracking upon cooling. To maintain the integrity of sintered zirconia bodies at room temperature, one can either sinter at low temperature for it to remain monoclinic during sintering—which leads to a low-strength and toughness ceramic—or stabilize the tetragonal or the cubic phases at room temperature by alloying, thereby avoiding the tm transformation during cooling. The fundamental approach to the engineering use of zirconia and avoiding the transformation-induced cracking described by Ruff and Ebert4 almost a century ago remains valid today: alloying pure zirconia with another oxide to fully or partially stabilize the tetragonal and/or the cubic phase. Calcium, magnesium, yttrium, and cerium oxides have been the most widely used stabilizers and lead to a number of different microstructures. In general, zirconia ceramics may conveniently be classified into three major types according to their microstructure: FSZ, PSZ, and TZP, standing, respectively, for fully stabilized zirconia, partially stabilized zirconia, and tetragonal zirconia polycrystals. In FSZ zirconia is in its cubic form, the form most commonly used in oxygen sensors and fuel cell electrolytes. It is generally obtained with large concentration of stabilizers (i.e., more than 8 mol% Y2O3). The PSZ consists of nanosized tetragonal or monoclinic particles that have precipitated out in a cubic matrix. Such zirconia ceramics are generally obtained with the addition of lime or magnesia. TZPs are often considered as monoliths of tetragonal phase, although they may contain a secondary cubic phase (see Panel A). The majority of TZPs that have been investigated are those stabilized with yttria or ceria. III. Stabilization and Transformation of the Tetragonal Phase As mentioned above, stabilization of powders and sintered ceramics can be achieved by alloying pure zirconia with other oxides. Investigations of the stability of different phases with different stabilizers led to the development of the equilibrium phase diagrams such as those compiled by the American Ceramic Society—NIST project.15 Alloy stabilization not only enables fabrication of crack-free zirconia but as demonstrated by Gupta et al.,16 sintered bodies of polycrystalline tetragonal zirconia can be prepared and the tetragonal phase retained down to room temperature even though the equilibrium phase is monoclinic. These metastable tetragonal ceramics exhibited exceptional fracture toughness when the transformation to the monoclinic phase was triggered by a propagating crack. The toughening that can be achieved for different concentrations of yttria is illustrated in Fig. 2, which summarizes data for transformation-toughened zirconia and ferroelastic toughening. Also shown, superimposed, is the sensitivity to aging as a function of yttria concentration. The fracture toughness of monoclinic and cubic zirconia, which is similar to that of window glass, provides a reference against which the toughening through the tm phase transformation can be compared. It is emphasized that the data are obtained from standard fracture toughness tests, such as indentation and double cantilever beam tests, in fast fracture conditions under which LTD is avoided. As indicated in Fig. 2, the attainable fracture toughness decreases as the yttria concentration increases. In the context of the metastable phase diagram, the toughening is proportional to the magnitude of the undercooling below the T0 (t/m) temperature (see Panel A for a more detailed explanation). Furthermore, transformation toughening is restricted to moderate temperatures, becoming ineffective when the stable phase is tetragonal and not monoclinic. For these reasons, the most attractive compositions for transformation toughening are those with low yttria concentrations (but high enough to prevent spontaneous tm transformation during cooling), typically 2– 3 mol% Y2O3. In the presence of moisture, the transformation of metastable t–m can alternatively occur without the passage of a crack. In this sense, LTD is a competing process to transformation toughening with the two providing limiting behaviors by which the metastable tetragonal phase transforms to the more stable monoclinic phase (Fig. 1). If the transformation is triggered by a propagating crack, then one can get enhanced toughening (see ‘‘Section III(2)’’). On the other hand, the transformation may be triggered ‘‘chemically’’ by the infusion of water-derived species from the surface. The process on a surface is complex (Fig. 3) and results not only in the undesirable transformation but also surface roughening, microcracking, and grain pull-out as well as loss of strength—all processes detrimental to structural applications. The dilemma facing the alloy designer is that the YSZ compositions that are the most attractive for their fracture toughness are also those that are most susceptible to LTD. This is illustrated by the comparison of the fracture toughness data with the LTD data in Fig. 2. The stabilization of the tetragonal phase in polycrystalline ceramics is, undoubtedly, largely dependent on the mutual elastic constraint provided by the surrounding, untransformed Fig. 1. Crack propagation-induced transformation and intermediate temperature exposure to moisture are two alternative means by which metastable tetragonal phase can transform to monoclinic phase. 1902 Journal of the American Ceramic Society—Chevalier et al. Vol. 92, No. 9
September 2009 The Tetragonal-Monoclinic Transformation in Zirconia 1903 Panel a. Zirconia-Yttria phase diagram The zirconia-yttria phase diagram has been significantly refined many times since it was first introduced in 1951 by duwez et al (The phase diagram book devoted to zirconia includes 30 different variations in the phase diagram. ) At first, the succession revisions might be a surprise but the essential difficulty is that cation diffusion in zirconia is so slow that it has proved Furthermore, the slow diffusion kinetics means that metastable extensions of the phases can readily occur. Interestingly, it was the prospect of diffusion-limited processes in zirconia that led Pol Duwez to investigate the system in his pioneering studies of rapid solidification and glass formation. In addition, the characteristic features of the martensitic transformation, such as the start and finish temperatures, have further complicated the interpretation of the diagram and the interpretation of microstructures. This uncertainty is shown in Scott's phase diagram by the hatched region. As a result, there has been considerable confusion in the literature about the details of the phase diagram. Yashima et al. present a graphic illustration of particularly pronounced for the region pertinent to transformation toughening and the low-temperature degradation are he confusion by superimposing many of the published diagrams on one another in their Fig. Al. The disagreements are 2000 2000 T 1500 02 YO Mol fraction Y, O Mole Fraction Y,O, Mole Fraction 0026 0053 111 0026 0053 0081 0.111 2400 2400 2100 C 51500 T+C g T+C CICm = 900 600M C+M T(T/M) 25 25 0 YO,s Mole Fraction YO O. Mole Fraction Oxygen Site Fraction Vacancies Oxygen Site Fraction Vacancies Fig. Al. Evolution in our knowledge of the zirconia-yttria phase diagram: (a) original diagram by Duwez in 1951,(b)diagram presented by Scott in 1975.(c)and (d) most recent diagrams.(d)is the metastable phase diagram. The evolution in our knowledge of the phase diagram can be illustrated by the diagrams in Fig. Al, showing the original diagram by Duwez, the diagram presented by Scott in 1975 and the most recent diagrams that combines experimental and putational studies. At the time of writing, there are indications that even this version may not be quite correct and that the tetragonal boundary may exhibit retrograde curvature, as occurs in the ZrOz-CeO system. Considera ble clarification has been obtained from computer determination of the phase diagram, particularly the position of the metastable To lines. However, it has to be emphasized that the metastable lines themselves, as well as the phase boundary lines are obtained from optimization
Panel A. Zirconia-Yttria Phase Diagram5,6 The zirconia–yttria phase diagram has been significantly refined many times since it was first introduced in 1951 by Duwez et al. 7 (The phase diagram book devoted to zirconia includes 30 different variations in the phase diagram.) At first, the succession of revisions might be a surprise but the essential difficulty is that cation diffusion in zirconia is so slow8 that it has proved particularly difficult to establish equilibrium and hence the phase boundaries at temperatures below about 14001C. Furthermore, the slow diffusion kinetics means that metastable extensions of the phases can readily occur. Interestingly, it was the prospect of diffusion-limited processes in zirconia that led Pol Duwez to investigate the system in his pioneering studies of rapid solidification and glass formation. In addition, the characteristic features of the martensitic transformation, such as the start and finish temperatures, have further complicated the interpretation of the diagram and the interpretation of microstructures. This uncertainty is shown in Scott’s phase diagram by the hatched region.9 As a result, there has been considerable confusion in the literature about the details of the phase diagram. Yashima et al. 5 present a graphic illustration of the confusion by superimposing many of the published diagrams on one another in their Fig. A1. The disagreements are particularly pronounced for the region pertinent to transformation toughening and the low-temperature degradation. The evolution in our knowledge of the phase diagram can be illustrated by the diagrams in Fig. A1, showing the original diagram by Duwez, the diagram presented by Scott in 19758 and the most recent diagrams that combines experimental and computational studies.10,11 At the time of writing, there are indications that even this version may not be quite correct and that the tetragonal boundary may exhibit retrograde curvature, as occurs in the ZrO2–CeO2 system.12 Considerable clarification has been obtained from computer determination of the phase diagram, particularly the position of the metastable T0 lines. However, it has to be emphasized that the metastable lines themselves, as well as the phase boundary lines are obtained from optimization Fig. A1. Evolution in our knowledge of the zirconia–yttria phase diagram: (a) original diagram by Duwez in 1951,6 (b) diagram presented by Scott in 1975,8 (c) and (d) most recent diagrams9,10 (d) is the metastable phase diagram. September 2009 The Tetragonal-Monoclinic Transformation in Zirconia 1903
1904 Journal of the American Ceramic Society--Chevalier et al Vol. 92. No 9 Panel A. Continued procedures that use the experimentally determined equilibrium phase boundaries as input parameters. For instance, the temperature of the monoclinic to tetragonal and the tetragonal to cubic, as well as the t/c boundary line are used to determine the metastable To(/m) boundary so if the t/e boundary line is inaccurate, then the computed To(/m) may not be fully correct The necessity of considering the metastable phase diagram is that cation diffusion in zirconia is exceptionally slow at all but the highest temperature. This is illustrated by the graph in Fig. A2, which shows the estimated time for diffusion to occur to homogenize the y content of a 0. 5-or 3-Hm-diameter grain at different temperatures(calculated from Kilo et al. and ).Even at a temperature of 1500C, a sintering temperature commonly used for zirconia ceramics in the past, it is estimated to take several weeks to achieve compositional homogeneity for the 3-um-grain size material. For a 0.5-um-grain size, this would represent days/hours instead of weeks( time roughly divided by 36 compared with the 3 um case). This explains why phase and yttria partitioning was observed in 3Y-TZP sintered for 5 h at 1550C in previous work 2O, Mole Fraction 0025 005 0075 Cation (Zr, Y 2400 2000 01500+-9 900 600 300 MOnoclinic. Time(s) 25 Estimated time for diffusion to homogenize the y 005 0.15 of 0.5.and 3-Hm-diameter grains at different temperatures in YO, Mole Fraction 3Y-TZP. TZP, tetragonal zirconia polycrystals; Y, yttria. Fig A3. Metastable zirconia-yttria phase diagram. To illustrate the consequences of the very slow cation diffusion, and the crucial importance of the metastable phase diagram in understanding LTD, we take as an example, a 3 mlo Y,,(6.0 m/o YO,s)material sintered at 1500C, composition C in Fig. A3. At equilibrium at this temperature, the sintered sample should consist of two phases, a tetragonal phase of composition 2.4 mo Y,O3(4.5 YO, 5)(point A)and a cubic phase of composition of 7.5 m/o Y,O,(point B). At room temperature. the equilibrium phases, from Fig. Al, would be a monoclinic phase with a yttria concentration of almost zero and a cubic phase with a yttria concentration of 18 m/o Y203. However, for this equilibrium condition to occur the yttrium ions must diffuse to partition into the yttria-poor monoclinic and yttria-rich cubic phases as shown by the horizontal arrows in the figure. As the ndicated by the diffusion distance figure, this would take many years. Instead, under typical cooling conditions, little or no yttrium ion partitioning occurs and the compositions obtained on cooling to room temperature, will be given by the intersection of the vertical dashed lines with the composition axis. At temperatures below the To (t/m). the tetragonal phase is metastable with respect to transformation to the monoclinic but the transformation is kinetically limited For instance, if there has been no diffusion, the To(/m) temperature is given by the intersection point E whereas if diffusional partitioning is complete at the sintering temperature, then the To (t/m)temperature is given by the intersection point D. Consequently, before low-temperature aging, the phases observed will then be a metastable tetragonal phase and a cubic phase, both with the same yttria concentrations as they have at the sintering temperature. Then, as the transformation occurs below the To(t/m) temperature, the nonoclinic phase will have the same yttria concentration as the tetragonal phase, namely 2.4 m/o Y,O3 (4.5 YO,5)in this example. Interestingly, as the yttria stabilizer concentration is increased and no partitioning occurs, the To(t/m) temperature decreases until at about 3.6 mo Y2O3(7.0 m/o YO1.5), it falls to below room temperature. So, unless the material is first transformed to the equilibrium cubic and tetragonal phases during sintering and subsequent heat treatment, it will not be susceptible to transformation until lower temperature is attained A further consequence of the equilibrium t/m phase boundary is that its steep slope implies that the composition of the tetragonal phase formed by diffusional partitioning at high temperatures is not very sensitive to the average composition of the powders, and hence there is little variation in the To (/m) temperature with the yttria content. What does change are the relative volume fractions of the tetragonal and cubic phases at the sintering temperature, and hence the maximum volume fraction of tetragonal that can transform to monoclinic by either transformation toughening or moisture-mediated LTD
Panel A. Continued procedures that use the experimentally determined equilibrium phase boundaries as input parameters. For instance, the temperature of the monoclinic to tetragonal and the tetragonal to cubic, as well as the t/c boundary line are used to determine the metastable T0 (t/m) boundary so if the t/c boundary line is inaccurate, then the computed T0 (t/m) may not be fully correct. The necessity of considering the metastable phase diagram is that cation diffusion in zirconia is exceptionally slow at all but the highest temperature. This is illustrated by the graph in Fig. A2, which shows the estimated time for diffusion to occur to homogenize the Y content of a 0.5- or 3-mm-diameter grain at different temperatures (calculated from Kilo et al. 8 and13). Even at a temperature of 15001C, a sintering temperature commonly used for zirconia ceramics in the past, it is estimated to take several weeks to achieve compositional homogeneity for the 3-mm-grain size material. For a 0.5-mm-grain size, this would represent days/hours instead of weeks (time roughly divided by 36 compared with the 3 mm case). This explains why phase and yttria partitioning was observed in 3Y-TZP sintered for 5 h at 15501C in previous work.14 To illustrate the consequences of the very slow cation diffusion, and the crucial importance of the metastable phase diagram in understanding LTD, we take as an example, a 3 m/o Y2O3 (6.0 m/o YO1.5) material sintered at 15001C, composition C in Fig. A3. At equilibrium at this temperature, the sintered sample should consist of two phases, a tetragonal phase of composition 2.4 m/o Y2O3 (4.5 YO1.5) (point A) and a cubic phase of composition of 7.5 m/o Y2O3 (point B). At room temperature, the equilibrium phases, from Fig. A1, would be a monoclinic phase with a yttria concentration of almost zero and a cubic phase with a yttria concentration of B18 m/o Y2O3. However, for this equilibrium condition to occur the yttrium ions must diffuse to partition into the yttria-poor monoclinic and yttria-rich cubic phases as shown by the horizontal arrows in the figure. As the indicated by the diffusion distance figure, this would take many years. Instead, under typical cooling conditions, little or no yttrium ion partitioning occurs and the compositions obtained on cooling to room temperature, will be given by the intersection of the vertical dashed lines with the composition axis. At temperatures below the T0 (t/m), the tetragonal phase is metastable with respect to transformation to the monoclinic but the transformation is kinetically limited. For instance, if there has been no diffusion, the T0 (t/m) temperature is given by the intersection point E whereas if diffusional partitioning is complete at the sintering temperature, then the T0 (t/m) temperature is given by the intersection point D. Consequently, before low-temperature aging, the phases observed will then be a metastable tetragonal phase and a cubic phase, both with the same yttria concentrations as they have at the sintering temperature. Then, as the transformation occurs below the T0 (t/m) temperature, the monoclinic phase will have the same yttria concentration as the tetragonal phase, namely 2.4 m/o Y2O3 (4.5 YO1.5) in this example. Interestingly, as the yttria stabilizer concentration is increased and no partitioning occurs, the T0 (t/m) temperature decreases until at about 3.6 m/o Y2O3 (7.0 m/o YO1.5), it falls to below room temperature. So, unless the material is first transformed to the equilibrium cubic and tetragonal phases during sintering and subsequent heat treatment, it will not be susceptible to transformation until lower temperature is attained. A further consequence of the equilibrium t/m phase boundary is that its steep slope implies that the composition of the tetragonal phase formed by diffusional partitioning at high temperatures is not very sensitive to the average composition of the powders, and hence there is little variation in the T0 (t/m) temperature with the yttria content. What does change are the relative volume fractions of the tetragonal and cubic phases at the sintering temperature, and hence the maximum volume fraction of tetragonal that can transform to monoclinic by either transformation toughening or moisture-mediated LTD. Fig. A2. Estimated time for diffusion to occur to homogenize the Y content of 0.5- and 3-mm-diameter grains at different temperatures in 3Y-TZP.7,11 TZP, tetragonal zirconia polycrystals; Y, yttria. Fig. A3. Metastable zirconia–yttria phase diagram. 1904 Journal of the American Ceramic Society—Chevalier et al. Vol. 92, No. 9
eptember 2009 The Tetragonal-Monoclinic Transformation in Zirconia 1905 While the stabilization of pure zirconia can be understood in erms of the balance between chemical and surface energy, the IP ceram reason that different aliovalent ions are effective as stabilizers nd also, perhaps, why moisture causes destabilization and iso- thermal transformation from tetragonal to monoclinic remains 6 oo to be understood. Many researchers have argued that stabilize- EEyY 9 Kc TZP ceramics tion is a direct consequence of the presence of the oxygen va cancies introduced by the aliovalent alloying element rather than the aliovalent dopant itself. 0.2 However, this hypothesis could 4 not be tested until the advent of large-scale computations. Now it has been shown computationally that the tetragonal phase 20 can be produced with lower energy than the monoclinic phase by introducing oxygen vacancies and without any aliovalent ions into the unit cell. This form of stabilization alone is unlikely 2 Kr Cubic to be the complete explanation because the solubility of the te- tragonal and cubic phases at different temperatures depends or the alloying dopant. Otherwise, all the phase diagrams for differ Aging sensitivity ent stabilizers would collapse onto one when plotted as a func- ion of oxygen vacancies. Nevertheless, it attractive explanation that has been used to rationalize the pre- 8 vailing explanation of LTD(see""Section IV): that moisture Yttria content(mol %Y,O3 pecies enter into the tetragonal lattice, annihilating oxygen ion Fig. 2. Fracture toughness and aging sensitivity of yttria-stabilized vacancies and thereby destabilizing the tetragonal(and cubic zirconia as a function of yttria stabilizer concentration. Toughness phases) data for cubic and monoclinic zirconia is indicated together with the The most systematic study of the role of different stabilizer rroelastic toughness in a densified thermal barrier coating. aging (dopant) ions on the stability of tetragonal and cubic zirconia sensitivity is here described by the degree of tetragonal phase transfo has been performed using X-ray absorption spectroscopy and mation results published in a series of papers by li et al. 20.23 They ex tetragonal fraction) after 3 h at 134.C in water vapor. mined the effect of trivalent and tetravalent dopant ions, and the effect of undersized and oversized dopants on the local en grains, whether these are also tetragonal grains, as in TzP, or ronment of zirconium ions Local atomic structures around the cubic matrix material in which the tetragonal precipitates are the Zr and around dopant cations in zirconia solid solution mbedded for the psz ceramics. as the I-mm transformation is were determined. These included undersized(Fe, GaT) and ccompanied by a volume increase, the transformation is con 3+ Gd trivalent ions as well as undersized strained by the surrounding grains. In these cases, the therme (Ge)and oversized(Ce)tetravalent ions dynamic framework that includes mechanical work arguments In the case of trivalent dopants, oxyge provides a rationale for the stabilization. The most general de- ated for charge com vas concluded tha scription of the energetics involved is reproduced below cies are associated with the zr cations in the case of oversized Recently with detailed study of zirconia nanoparticles, Gar dopants, and with the two dopant cations in the case of under- ie's claimthat pure, zirconia powders could be retained in the zed dopants. Both configurations favor sevenfold coordinated tetragonal state provided that their size was below a critical size xygen ions around the Zr cations and stabilize the tetragonal or has been extensively validated. Garvie's concept was that stabi lization could occur by surface energy alone and a series of en- gen vacancies to Zr is responsible for the more effective stabi ergy cross-overs between monoclinic, tetragonal, and cubic ation effects of oversized trivalent dopants. In essence, the been demonstrated as being consistent with stabilization of tetragonal zirconia with oversized trivalent cat- the dominant role of surface energy at nanometer particle sizes For example, a 4-24 nm stability range of tetragonal zirconia at From the results and the analysis performed by Li and col- room temperature can be extrapolated from Pitcher et al.s leagues, it is evident that doping by trivalent oversized cations, work,while monoclinic phase is stable above this size. This such as Y3+, is most efficient in relieving the overcrowd ole of surface energy is confirmed by Suresh et al.,who found ing(via both oxygen vacancies generation and dilatation of the a decrease of the f-m transformation temperature upon cooling tion network). The conceptual idea being that oxygen over- crowding around the small zirconium Zr cation is responsib for th the tetragonal phase may be stabilized by oxygen vacancies ad jacent to the Zr t ion and introduced by aliovalent doping. this lso provides a conceptual explanation for the stabilization by dilatation of the cation structure, and explains why 1.5 mol% of Y,O3 is sufficient to stabilize tetragonal zirconia, whereas Stresses Roughening Micro-cracking Path for wate 10 mol% of CeO2 is needed to achieve the same stability. Recent k- using Y NMR has provided more direct experimen support for the preference of oxygen vacancies to reside in lattice sites adjacent to the zr ion in YSz alloy Increased wear Slow Crack Growth (1) The Energetics of Transformation The foregoing discussion describes the stabilization of tetrago nal and cubic phases of zirconia under stress-free conditions Wear debris Loss in strength For polycrystalline ceramics, where the tetragonal phase is retained, transformation is mechanically constrained under Fig 3. Potential effect of aging on the integrity of zirconia devices. metastable conditions. The condition for transformation can ontributions to the mechanisms
grains, whether these are also tetragonal grains, as in TZP, or the cubic matrix material in which the tetragonal precipitates are embedded for the PSZ ceramics. As the tm transformation is accompanied by a volume increase, the transformation is constrained by the surrounding grains. In these cases, the thermodynamic framework that includes mechanical work arguments provides a rationale for the stabilization. The most general description of the energetics involved is reproduced below. Recently with detailed study of zirconia nanoparticles, Garvie’s claim17 that pure, zirconia powders could be retained in the tetragonal state provided that their size was below a critical size has been extensively validated. Garvie’s concept was that stabilization could occur by surface energy alone and a series of energy cross-overs between monoclinic, tetragonal, and cubic phases has since been demonstrated as being consistent with the dominant role of surface energy at nanometer particle sizes. For example, a 4–24 nm stability range of tetragonal zirconia at room temperature can be extrapolated from Pitcher et al.’s work,18 while monoclinic phase is stable above this size. This role of surface energy is confirmed by Suresh et al.,19 who found a decrease of the tm transformation temperature upon cooling with grain size. While the stabilization of pure zirconia can be understood in terms of the balance between chemical and surface energy, the reason that different aliovalent ions are effective as stabilizers and also, perhaps, why moisture causes destabilization and isothermal transformation from tetragonal to monoclinic remains to be understood. Many researchers have argued that stabilization is a direct consequence of the presence of the oxygen vacancies introduced by the aliovalent alloying element rather than the aliovalent dopant itself.20,21 However, this hypothesis could not be tested until the advent of large-scale computations. Now, it has been shown computationally22 that the tetragonal phase can be produced with lower energy than the monoclinic phase by introducing oxygen vacancies and without any aliovalent ions into the unit cell. This form of stabilization alone is unlikely to be the complete explanation because the solubility of the tetragonal and cubic phases at different temperatures depends on the alloying dopant. Otherwise, all the phase diagrams for different stabilizers would collapse onto one when plotted as a function of concentration of oxygen vacancies. Nevertheless, it is an attractive explanation that has been used to rationalize the prevailing explanation of LTD (see ‘‘Section IV’’): that moisture species enter into the tetragonal lattice, annihilating oxygen ion vacancies and thereby destabilizing the tetragonal (and cubic phases). The most systematic study of the role of different stabilizer (dopant) ions on the stability of tetragonal and cubic zirconia has been performed using X-ray absorption spectroscopy and results published in a series of papers by Li et al. 20,23 They examined the effect of trivalent and tetravalent dopant ions, and the effect of undersized and oversized dopants on the local environment of zirconium ions. Local atomic structures around the Zr41 and around dopant cations in zirconia solid solutions were determined. These included undersized (Fe31, Ga31) and oversized (Y31, Gd31) trivalent ions as well as undersized (Ge41) and oversized (Ce41) tetravalent ions. In the case of trivalent dopants, oxygen vacancies are generated for charge compensation. It was concluded that the vacancies are associated with the Zr cations in the case of oversized dopants, and with the two dopant cations in the case of undersized dopants. Both configurations favor sevenfold coordinated oxygen ions around the Zr cations and stabilize the tetragonal or even the cubic phases. However, the different availability of oxygen vacancies to Zr is responsible for the more effective stabilization effects of oversized trivalent dopants. In essence, the stabilization of tetragonal zirconia with oversized trivalent cations is twice as efficient as with undersized trivalent cations. From the results and the analysis performed by Li and colleagues, it is evident that doping by trivalent oversized cations, such as Y31, is most efficient in relieving the oxygen overcrowding (via both oxygen vacancies generation and dilatation of the cation network). The conceptual idea being that oxygen overcrowding around the small zirconium Zr41 cation is responsible for the poor stability of undoped tetragonal zirconia, and that the tetragonal phase may be stabilized by oxygen vacancies adjacent to the Zr41 ion and introduced by aliovalent doping. This also provides a conceptual explanation for the stabilization by dilatation of the cation structure, and explains why 1.5 mol% of Y2O3 is sufficient to stabilize tetragonal zirconia, whereas 10 mol% of CeO2 is needed to achieve the same stability. Recent work24 using 89Y NMR has provided more direct experimental support for the preference of oxygen vacancies to reside in lattice sites adjacent to the Zr41 ion in YSZ alloys. (1) The Energetics of Transformation The foregoing discussion describes the stabilization of tetragonal and cubic phases of zirconia under stress-free conditions. For polycrystalline ceramics, where the tetragonal phase is retained, transformation is mechanically constrained under metastable conditions. The condition for transformation can be expressed in terms of the different energy contributions to the overall energy, as discussed by Lange,25 who considered the Fig. 2. Fracture toughness and aging sensitivity of yttria-stabilized zirconia as a function of yttria stabilizer concentration. Toughness data for cubic and monoclinic zirconia is indicated together with the ferroelastic toughness in a densified thermal barrier coating. Aging sensitivity is here described by the degree of tetragonal phase transformation toward monoclinic (i.e., the ratio monoclinic fraction/initial tetragonal fraction) after 3 h at 1341C in water vapor. Fig. 3. Potential effect of aging on the integrity of zirconia devices. Arrows indicate coupling of aging with crack propagation and wear mechanisms. September 2009 The Tetragonal-Monoclinic Transformation in Zirconia 1905
1906 Journal of the American Ceramic Society-Chevalier et al Vol. 92. No 9 ey of a tetragonal particle embedded in an infinite matrix. AGcl AUSE +AUs. When AGr-m becomes negative, the ugh a rather idealized configuration that does not take tetragonal particle is metastable or unstable and may transform grains, this simple analysis provides considerable insight into the AUSE, the addition of Y203 decreases the driving force of l-m factors affecting the transformation of the particle to its mono- transformation, hence its temperature(see yttria-zirconia phase clinic allotrope. The change of total free energy (4Gi-m)for the diagram in Panel A), making possible the retention of meta I-m transformation of the particle can be expressed by stable tetragonal phase in dense bodies at room temperatures. The elastic self-energy AUsE is directly related to the surround- △Gc+△UsE+△Us (1) ing matrix modulus, so having the matrix of a stiffer material such as alumina, increases AUse, stabilizing the tetragonal here AGe(0) refers to the change in sequences of these contributions is that the driving force for the strain energy associated with the transformation of par m transformation will not be the same inside the bulk and on This is dependent on the modulus of the surrounding urface(or even for powders). because neither AUSE nor AUs the size and shape of the particle, and the presence of are the same. In particular, there can be configurations at the internal or external stresses. The final term, AUs(>0)is the surface where the volume change of the transformation can be change in energy associated with the formation of new interfaces accommodated by a surface uplift( Panel B). This accommoda when the transformation occurs, for instance, cracks and mono- tion is not possible in the bulk. (The main features of the clinic variants. The particle remains in its tetragonal state tetragonal to monoclinic transformation at the surface and the if the overall thermodynamic driving force AGr-m>0, i.e. if bulk are schematized in Panel B. There is also the possibility Main Features of the Tetragonal to Monoclin inic Transformation in Zirconia Crystallography of the transformation The tetragonal to monoclinic phase transformation in zirconia is martensitic in nature. Even if alternative approaches have been recently developed, it is most often described by the phenomenological theory of martensitic crystallography(PTMC) The reader may refer to the work of Kelly and Rose> or of Deville et al. for a comprehensive description Crystallographic correspondences exist between the parent( tetragonal)and the product(monoclinic) phase, as schematized in Fig. Bl. They can be described by habit planes and directions(shape strain) as summarized in Table Bl. Thr lattice cor dences exist, called ABC, BCA, and CAB, which correspond to a change of the(ar, br c,) lattice axis of the gonal phase changes into(a, b e a,m) and(cm, anm bm), respectively. Each of these lattice correspondences may occur along two different habit planes. This leads to the six different configurations given in Table Bl and schematized in Fig. B2. The configurations depicted in Fig. B2 take into account the fact that four variants may occur for each crystallographic orrespondence(indeed, in the tetragonal symmetry, a, b, -a and -b are crystallographically equivalent). For each, the shear strain associated to the tetragonal to monoclinic(I-m) transformation is around 0. 16 and the volume expansion around 0.05 Parent phase(t) Fig. B1. Schematic illustration of crystallographic correspondences between the tetragonal(parent) and the monoclinic(product) phases during the martensitic (-m transformation f-m, tetragonal to monoclinic
energy of a tetragonal particle embedded in an infinite matrix. Although a rather idealized configuration that does not take into account the presence of a free surface or irregular shapes grains, this simple analysis provides considerable insight into the factors affecting the transformation of the particle to its monoclinic allotrope. The change of total free energy (DGtm) for the tm transformation of the particle can be expressed by DGtm ¼ DGc þ DUSE þ DUS (1) where DGc (o0 at temperatures below the equilibrium MS temperature) is the difference in chemical free energy between the tetragonal and monoclinic phases. This is dependent on temperature and composition, implicitly including the oxygen vacancy content. The term DUSE (40) refers to the change in elastic strain energy associated with the transformation of particles. This is dependent on the modulus of the surrounding matrix, the size and shape of the particle, and the presence of internal or external stresses. The final term, DUS (40) is the change in energy associated with the formation of new interfaces when the transformation occurs, for instance, cracks and monoclinic variants. The particle remains in its tetragonal state if the overall thermodynamic driving force DGtm40, i.e. if jDGcj < DUSE þ DUS. When DGtm becomes negative, the tetragonal particle is metastable or unstable and may transform into its monoclinic state. By decreasing jDGcj and increasing DUSE, the addition of Y2O3 decreases the driving force of tm transformation, hence its temperature (see yttria–zirconia phase diagram in Panel A), making possible the retention of metastable tetragonal phase in dense bodies at room temperatures. The elastic self-energy DUSE is directly related to the surrounding matrix modulus, so having the matrix of a stiffer material, such as alumina, increases DUSE, stabilizing the tetragonal phase. It is also directly influenced by applied or internal stresses: tensile hydrostatic stress will act to reduce DUSE, destabilizing the tetragonal phase, whereas hydrostatic pressure favors the retention of the tetragonal phase. One of the consequences of these contributions is that the driving force for the tm transformation will not be the same inside the bulk and on its surface (or even for powders), because neither DUSE nor DUS are the same. In particular, there can be configurations at the surface where the volume change of the transformation can be accommodated by a surface uplift (Panel B). This accommodation is not possible in the bulk. (The main features of the tetragonal to monoclinic transformation at the surface and the bulk are schematized in Panel B.) There is also the possibility Panel B. Main Features of the Tetragonal to Monoclinic Transformation in Zirconia Crystallography of the transformation. The tetragonal to monoclinic phase transformation in zirconia is martensitic in nature. Even if alternative approaches29 have been recently developed, it is most often described by the phenomenological theory of martensitic crystallography (PTMC). The reader may refer to the work of Kelly and Rose30 or of Deville et al. 31 for a comprehensive description. Crystallographic correspondences exist between the parent (tetragonal) and the product (monoclinic) phase, as schematized in Fig. B1. They can be described by habit planes and directions (shape strain) as summarized in Table B1.31 Three possible lattice correspondences exist, called ABC, BCA, and CAB, which correspond to a change of the (at, bt, ct) lattice axis of the tetragonal phase changes into (am, bm, cm), (bm, cm, am) and (cm, am, bm), respectively. Each of these lattice correspondences may occur along two different habit planes. This leads to the six different configurations given in Table B1 and schematized in Fig. B2. The configurations depicted in Fig. B2 take into account the fact that four variants may occur for each crystallographic correspondence (indeed, in the tetragonal symmetry, a, b, –a and –b are crystallographically equivalent). For each, the shear strain associated to the tetragonal to monoclinic (t–m) transformation is around 0.16 and the volume expansion around 0.05. Fig. B1. Schematic illustration of crystallographic correspondences between the tetragonal (parent) and the monoclinic (product) phases during the martensitic tm transformation. tm, tetragonal to monoclinic. 1906 Journal of the American Ceramic Society—Chevalier et al. Vol. 92, No. 9
September 2009 The Tetragonal-Monoclinic Transformation in Zirconia Table B1. Crystallographic Features of the Tetragonal-Monoclinic Martensitic Transformation in Zirconia correspondence shear Shape amplitude ABC I 0.0344 -0.9537 0.0026 0.1640 0.1556 0.0518 0.0055 0.0028 0.3005 0.1640 ABC 2 0.0344 0.0915 0.1597 0.1640 0.1556 0.0518 0.0171 0.0007 -0.9956 0.0373 BCA I (1)010J 0.0344 0.0034 0.0030 0.1761 0.1683 0.0518 0.1751 0.9193 0.0186 BCA 2 (1l)o 0.0344 0.0168 0.0004 0.176l 0.1683 0.0518 0.9996 0.0558 0.0241 0.1670 CAB I [00.0027 0.3006 0.1640 0.164 0.1556 0.0518 0.9537 0.0026 0.0001 0.0002 CAB 2 (101)[o 0.0027 0.9958 -0.0373 0.1640 0.1556 0.0518 0.0915 0.1597 0.000 iput parameters: I-phase: a, =5.128 A, C,=5.224 A; m-phase: am=5.203 A, bm=5.217 A, c,=5.388 A, Bm,=98.91. Expression in the lattice axis system of the tragonal parent phase ' Expression in the orthogonal axis system bounded to the tetragonal lattice axis system ABC1 ABC2 BCA2 CAB1 ACt Fig B2. Self-accommodating variant pairs deduced from the different lattice correspondences with the effect of t-m transformation on a surface Features of the transformation at the surface Recently, atomic force microscopy(AFM) brought new insights into the transformation induced relief. A typical example of surface uplift associated to the onset of transformation in a ceria-stabilized zirconia is given in Fig. B3. The relief exhibit fourfold symmetry, with a set of four variants. This indicates that the free surfaces where the observations are done are nearly perpendicular to the craxis. Among the six possible configurations of Fig. B2, only ABCI and BCA2 present the most mportant shape strain along the craxis. In the case of BCA2, however, a significant strain takes place along the braxis to which would lead to important internal stresses. Therefore ABCI is the most likely to occur in practice at the surface, because all the volume increase associated to the transformation is relaxed through a surface uplift. In other words, for such configuration
Features of the transformation at the surface Recently, atomic force microscopy (AFM) brought new insights into the transformation induced relief. A typical example of surface uplift associated to the onset of transformation in a ceria-stabilized zirconia is given in Fig. B3. The relief exhibits fourfold symmetry, with a set of four variants. This indicates that the free surfaces where the observations are done are nearly perpendicular to the ct-axis. Among the six possible configurations of Fig. B2, only ABC1 and BCA2 present the most important shape strain along the ct-axis. In the case of BCA2, however, a significant strain takes place along the bt-axis too, which would lead to important internal stresses. Therefore ABC1 is the most likely to occur in practice at the surface, because all the volume increase associated to the transformation is relaxed through a surface uplift. In other words, for such configuration Fig. B2. Self-accommodating variant pairs deduced from the different lattice correspondences with the effect of tm transformation on a surface perpendicular to the junction plane.31 tm, tetragonal to monoclinic. Table B1. Crystallographic Features of the Tetragonal–Monoclinic Martensitic Transformation in Zirconia31 Lattice correspondence Lattice invariant shearw Magnitude of g Habit plane normalz Shape strainz Shape strain amplitude Shear component Volume change ABC 1 011 ð Þ 011 0.0344 0:9537 0:0055 0:3005 2 4 3 5 0:0026 0:0028 0:1640 2 4 3 5 0.1640 0.1556 0.0518 ABC 2 011 ð Þ 011 0.0344 0:0915 0:0171 0:9956 2 4 3 5 0:1597 0:0007 0:0373 2 4 3 5 0.1640 0.1556 0.0518 BCA 1 110 ð Þ 110 0.0344 0:0034 0:3935 0:9193 2 4 3 5 0:0030 0:1751 0:0186 2 4 3 5 0.1761 0.1683 0.0518 BCA 2 110 ð Þ 110 0.0344 0:0168 0:9996 0:0241 2 4 3 5 0:0004 0:0558 0:1670 2 4 3 5 0.1761 0.1683 0.0518 CAB 1 ð Þ 101 101 0.0027 0:3006 0:9537 0:0001 2 4 3 5 0:1640 0:0026 0:0002 2 4 3 5 0.1640 0.1556 0.0518 CAB 2 101 ð Þ 101 0.0027 0:9958 0:0915 0:0003 2 4 3 5 0:0373 0:1597 0:0001 2 4 3 5 0.1640 0.1556 0.0518 Input parameters: t-phase: at 5 5.128 A˚ , ct 5 5.224 A˚ ; m-phase: am 5 5.203 A˚ , bm 5 5.217 A˚ , cm 5 5.388 A˚ , bm 5 98.911. w Expression in the lattice axis system of the tetragonal parent phase. z Expression in the orthogonal axis system bounded to the tetragonal lattice axis system. September 2009 The Tetragonal-Monoclinic Transformation in Zirconia 1907
1908 Journal of the American Ceramic Society--Chevalier et al Vol. 92. No 9 Panel B. Continued Crack ation dir Fig. B3. surface uplifts associated to the onset of transformation in a uplifts associated to the t-m transformation in a (craxis perpendicular to the surface and ABCI correspondence), the term AUsE in Eq (1)(elastic strain energy change due to the transformation)is close to zero. It is then obvious that the grains likely to transform first in zirconia are surface grains, and that these grains have their craxis close to the normal to the surface. However, grains with their araxis perpendicular to the surface may also transform following lattice correspondence CABl, because shape strain for this correspondence is parallel to ar. For this onfiguration, only two variants are susceptible to occur, leading to a topography change with a twofold symmetry only Features of the transformation at the surface, in the presence of stresses nature and the magnitude of the stress field and on the orientation of the potential habit plane with respect to the stress field Grains with their craxis perpendicular to the surface and their junction planes parallel to the maximum tensile stress are the most likely to transform. Figure B4 shows an AFM picture of a transformed zone around a propagating crack. Large stresses around the crack favor the transformation, and this occurs first for the grains subjected to the largest tensile stresses, with an adequate orientation Features of the transformation in the bulk f-m transformation in the bulk, in the presence of a stress field, is the main source of toughening in zirconia systems at room emperature Transmission electron microscopy (TEM) images are so far the only way to obtain the features of the transformation in the bulk and the majority of TEM studies have been performed in Mg-PSZ. Lenticular tetragonal precipitates transform into a stack of monoclinic variants, which accommodate the shear component of the transformation. Therefore, in first good approximation, only the dilatant component of the transformation contributes to toughening. that the surface energy change AUs is lower in the presence of (2) Mechanics of Transformation Toughening moisture or water vapor pressure. Similarly, just as a critical size As described earlier, the energy required to propagate a crack exists for the I-m transformation exists in powders, it may through a dense nla-contain ng metastable zirconia is well be modified on the surface27 or in the bulk. 5 The critical increased if the crack relieves some or all of the mechanical size often reported for bulk transformation is on the order of a constraint on the metastable tetragonal and allows it to trans micrometer, whereas it falls to around a nanometer at the form to the monoclinic phase. This can only occur below the To urface. This is of major implication for LTD, which may (r/m) temperature, which in some papers is identified as the occur even in very fine zirconia ceramics martensite start temperature, TMs. Two equivalent descriptions
that the surface energy change DUS is lower in the presence of moisture or water vapor pressure. Similarly, just as a critical size exists for the tm transformation exists in powders,26 it may well be modified on the surface27 or in the bulk.25 The critical size often reported for bulk transformation is on the order of a micrometer, whereas it falls to around a nanometer at the surface.28 This is of major implication for LTD, which may occur even in very fine zirconia ceramics.28 (2) Mechanics of Transformation Toughening As described earlier, the energy required to propagate a crack through a dense zirconia-containing metastable zirconia is increased if the crack relieves some or all of the mechanical constraint on the metastable tetragonal and allows it to transform to the monoclinic phase. This can only occur below the T0 (t/m) temperature, which in some papers is identified as the martensite start temperature, TMS . Two equivalent descriptions Panel B. Continued (ct-axis perpendicular to the surface and ABC1 correspondence), the term DUSE in Eq. (1) (elastic strain energy change due to the transformation) is close to zero. It is then obvious that the grains likely to transform first in zirconia are surface grains, and that these grains have their ct-axis close to the normal to the surface. However, grains with their at-axis perpendicular to the surface may also transform following lattice correspondence CAB1, because shape strain for this correspondence is parallel to at. For this configuration, only two variants are susceptible to occur, leading to a topography change with a twofold symmetry only. Features of the transformation at the surface, in the presence of stresses In the presence of a stress field, the net driving force for the transformation can be modified. This change depends both on the nature and the magnitude of the stress field and on the orientation of the potential habit plane with respect to the stress field. Grains with their ct-axis perpendicular to the surface and their junction planes parallel to the maximum tensile stress are the most likely to transform.32 Figure B4 shows an AFM picture of a transformed zone around a propagating crack. Large stresses around the crack favor the transformation, and this occurs first for the grains subjected to the largest tensile stresses, with an adequate orientation. Features of the transformation in the bulk tm transformation in the bulk, in the presence of a stress field, is the main source of toughening in zirconia systems at room temperature. Transmission electron microscopy (TEM) images are so far the only way to obtain the features of the transformation in the bulk and the majority of TEM studies have been performed in Mg-PSZ. Lenticular tetragonal precipitates transform into a stack of monoclinic variants, which accommodate the shear component of the transformation. Therefore, in first good approximation, only the dilatant component of the transformation contributes to toughening. Fig. B4. Surface uplifts associated to the tm transformation in a ceria-stabilized zirconia in the vicinity of a propagating crack. tm, tetragonal to monoclinic. Fig. B3. surface uplifts associated to the onset of transformation in a ceria-stabilized zirconia.30 1908 Journal of the American Ceramic Society—Chevalier et al. Vol. 92, No. 9
eptember 2009 The Tetragonal-Monoclinic Transformation in Zirconia 1909 of the toughening afforded by crack-tip-induced transformation have been formulated. one in terms of stress intensities and the △ Water75℃ other in terms of energies. A comprehensive review of the models and the experimental data on transformation tougher △ Water25℃ ing is beyond the scope of this review. The reader interested may o··A25℃ refer to the book by green et al.for details. Phase transforma- 口ol25℃ tion toughening originates in the presence of large tensile stresses 口 Vacuum25℃ round a crack, which can destabilize the tetragonal phase in the vicinity of the crack, forming a transformation zone. McMeek ing and Evansdeveloped a model of phase transformation ughening at the beginning of the 1980s in which the induced transformation leads to a shielding Kish of the applied stress intensity factor K. This means that the real stress intensity factor at the crack tip Kltin is lower than that applied by the external forces according Kltip=K1-KIsh 2) Theoretical modelsand experimental results show that the stress intensity factor, the larger the trans- ormation d the larger the shielding effect, leading to 10-12 well-known equation K,(MPa. m) KIsh= Chk (3) Fig 4. -KI laws in 3Y-TZP. TZP, tetragonal zirconia polycrystals: deposition of tetragonal-prime zirconia, for instance by 0214Eve(1+y) tering or electron-beam deposition. In the former case when a cubic crystal or individual grain transforms to tetra gonal, six different, crystallographically equivalent orientations In this equation, E is the Young modulus, Vr the volume of the c-axis of the tetragonal can result. Each of these variants fraction of transformable particles, e is the volume dilatation ("domains") has the same energy but can re-orient when a stress ciated to the transformation, v the Poisson ratio, and om is applied. In the case of tetragonal crystals formed directly, each the critical local stress leading to phase transformation. grain can be a single domain. Then, individual grains or portions The toughening capability of a irconia is directly thin each grain can be switched to a different orientation by dependent on the critical local stress leading to phase transfor- an applied stress. This can occur as a result of an applied stress mation, om. This value, om, depends in turn on the magnitude or in the presence of a propagating crack. of the undercooling below the To(t/m) temperature: the larger or domains to reorient, an applied stress of the correct sign the undercooling below To (/m) the lower critical stress for and nature needs to exceed the coercive stress. The depender stress assisted phase transformation and thus the larger the of coercive stress upon a number of factors such as composition transformation toughening. lattice parameter, and temperature is poorly known but the As mentioned in * Section I". transformation induced by ortant distinguishing feature is that the crack propagation is one of a number of competing pathw effect is related to change from one equilibrium state(variant) to by which the tetragonal phase can transform to the monoclinic another equilibrium state unlike in transformation toughening. form. At fast crack velocities the rate at which atomic bonds at The magnitude of the toughening effect is estimated to be AKc~2-3 MPa. m 2 in yttria-doped zirconia and as high as ne environment can reach the tip. This is region ll of the crack AKe-5-7 MPa in Ce-doped zirconia, depending upon elocity vs stres intenisity characteristic shown in Fig. t,e whens electron beam deposition the toughening is rather similar g 2 given by the extrapolation of region Ill to lower stress intensities, MPa.m/-(Fig. 2)and is produced by switching within indivi- as illustrated by the data in Fig. 4 obtained in testing in vacuum nd silicon oil. in those cases where the environment reacts with need for additional work to expand our knowledge of the values e crack-tip region I, higher crack velocities than the extra of the coercive stress and validate these estimates. In principle, lation of the region Ill behavior result as shown in Fig. 4 for both domain switching and transformation toughening can YSZ. Similar behavior is well established for fracture of glasses occur during crack propagation, as has been discussed in greater ind several other ceramic materials, including alumin detail elsewhere 35 Data on the depende for 3 mfo Y2O3 has been obtained at temperatures IV LTD nd exhibit an apparent threshold at 3.2 MPa.m Detailed measurements of the kinetics of the moisture-induced e data at this stress intensity, the crack velocity is commensu- transformation of both sintered ceramics and coatings, obtained rate with the rate at which the degradation front moves into ceramic. Indeed, the temperature de ace data suggest that either X-ray diffraction or Raman spectrosco the activation energy for moisture-assisted crack growth is similar ndicate that the kinetics can be fit with the standard mehk to that for moisture-induced ltd discussed in "Section Iv Avrami-Johnson equations for a nucleation and growth process see Fig. 5) (3) Ferro- Elastic Toughenin x=1-exp(-(bn)") Another possible source of toughening in tetragonal zirconias is crack propagation-induced ferro-elastic domain switching where a is the fraction of tetragonal that has transformed to sometimes referred to as ferroelastic toughening. This is possible monoclinic phase, t is the time of exposure to moisture and the in tetragonal zirconias produced by cooling from the cubic value of the constant, b, and the exponent, n, depends on phase by a composition invariant displacive reaction or by the mperature. Experimental data and simulations show that the
of the toughening afforded by crack-tip-induced transformation have been formulated, one in terms of stress intensities and the other in terms of energies. A comprehensive review of the models and the experimental data on transformation toughening is beyond the scope of this review. The reader interested may refer to the book by Green et al. 3 for details. Phase transformation toughening originates in the presence of large tensile stresses around a crack, which can destabilize the tetragonal phase in the vicinity of the crack, forming a transformation zone. McMeeking and Evans33 developed a model of phase transformation toughening at the beginning of the 1980s in which the stressinduced transformation leads to a shielding KIsh of the applied stress intensity factor KI. This means that the real stress intensity factor at the crack tip KItip is lower than that applied by the external forces, according to KItip ¼ Kl KIsh (2) Theoretical models33 and experimental results34 show that the higher the applied stress intensity factor, the larger the transformation zone and the larger the shielding effect, leading to the well-known equation KIsh ¼ CshKI (3) with Csh ¼ 0:214EVfeTð1 þ nÞ ð1 nÞsc m ffiffiffi 3 p 12p ! In this equation, E is the Young modulus, Vf the volume fraction of transformable particles, e T is the volume dilatation associated to the transformation, n the Poisson ratio, and sm c is the critical local stress leading to phase transformation. The toughening capability of a given zirconia is directly dependent on the critical local stress leading to phase transformation, sm c . This value, sm c , depends in turn on the magnitude of the undercooling below the T0 (t/m) temperature: the larger the undercooling below T0 (t/m) the lower critical stress for stress assisted phase transformation and thus the larger the transformation toughening. As mentioned in ‘‘Section I‘‘, transformation induced by crack propagation is one of a number of competing pathways by which the tetragonal phase can transform to the monoclinic form. At fast crack velocities, the rate at which atomic bonds at the crack tip are ruptured is much greater than the rate at which the environment can reach the tip. This is region III of the crack velocity vs stress intensity characteristic shown in Fig. 4.34 When a crack is propagated in a nonreactive environment, its velocity is given by the extrapolation of region III to lower stress intensities, as illustrated by the data in Fig. 4 obtained in testing in vacuum and silicon oil. In those cases where the environment reacts with the crack-tip, region I, higher crack velocities than the extrapolation of the region III behavior result as shown in Fig. 4 for YSZ. Similar behavior is well established for fracture of glasses and several other ceramic materials, including alumina. Data on the dependence of crack velocity on stress intensity for 3 m/o Y2O3 has been obtained at temperatures up to 751C and exhibit an apparent threshold at B3.2 MPa m1/2. Based on the data at this stress intensity, the crack velocity is commensurate with the rate at which the degradation front moves into a ceramic. Indeed, the temperature dependence data suggest that the activation energy for moisture-assisted crack growth is similar to that for moisture-induced LTD discussed in ‘‘Section IV.’’ (3) Ferro-Elastic Toughening Another possible source of toughening in tetragonal zirconias is crack propagation-induced ferro-elastic domain switching, sometimes referred to as ferroelastic toughening. This is possible in tetragonal zirconias produced by cooling from the cubic phase by a composition invariant displacive reaction or by the direct deposition of tetragonal-prime zirconia, for instance by sputtering or electron-beam deposition. In the former case, when a cubic crystal or individual grain transforms to tetragonal, six different, crystallographically equivalent orientations of the c-axis of the tetragonal can result. Each of these variants (‘‘domains’’) has the same energy but can re-orient when a stress is applied. In the case of tetragonal crystals formed directly, each grain can be a single domain. Then, individual grains or portions within each grain can be switched to a different orientation by an applied stress. This can occur as a result of an applied stress or in the presence of a propagating crack. For domains to reorient, an applied stress of the correct sign and nature needs to exceed the coercive stress. The dependence of coercive stress upon a number of factors such as composition, lattice parameter, and temperature is poorly known but the most important distinguishing feature is that the toughening effect is related to change from one equilibrium state (variant) to another equilibrium state unlike in transformation toughening. The magnitude of the toughening effect is estimated to be DKcB2–3 MPa m1/2 in yttria -doped zirconia and as high as DKcB5–7 MPa m1/2 in Ce-doped zirconia, depending upon composition. For the 7 YSZ tetragonal material produced by electron beam deposition, the toughening is rather similar,19 B2 MPa m1/2 (Fig. 2) and is produced by switching within individual grains. These values are approximate and there is a clear need for additional work to expand our knowledge of the values of the coercive stress and validate these estimates. In principle, both domain switching and transformation toughening can occur during crack propagation, as has been discussed in greater detail elsewhere.35 IV. LTD Detailed measurements of the kinetics of the moisture-induced transformation of both sintered ceramics and coatings, obtained by either X-ray diffraction36 or Raman spectroscopy,37 all indicate that the kinetics can be fit with the standard Mehl– Avrami–Johnson equations for a nucleation and growth process (see Fig. 5): a ¼ 1 exp ð Þ bt n ð Þ (4) where a is the fraction of tetragonal that has transformed to monoclinic phase, t is the time of exposure to moisture and the value of the constant, b, and the exponent, n, depends on temperature. Experimental data and simulations show that the Fig. 4. V–KI laws in 3Y-TZP. TZP, tetragonal zirconia polycrystals; Y, yttria. September 2009 The Tetragonal-Monoclinic Transformation in Zirconia 1909
1910 Journal of the American Ceramic Society--Chevalier et al Vol. 92. No 9 Aging time in-vivo(years, simulated) 32y 3000 (a)400 3 e ◆◆ ◆ 0%A2O31.2%A2O3 A10Z0Y -- 20%6 monoclinic lines 0 E-ruInnkrtttttts⊥AH 1000 (b) Aging time at 134C n kinetics of 10 mol% ceriastabl. 2Y(Isubakino) ■3Y( Tsubaki) ia(A80Z3Y) zirconia-toughened alumina(A10ZoY)2 measured at 134Cand pected at 37C(considering an activation energy of 106 k/mol for materials, which is probably not completely accurate). The shadowed 4Y(Tsubakino) areas give uncertainty ranges when they can be evaluated. 300 exponent has a value between 0.5 and 4. Combining the data 2200 btained at different temperatures, the transformation kinetics form"C"-shaped curves on a time-temperature plot. A number f examples of this time-temperature transformation (TTT behavior are shown in Fig. 6. At temperatures well below the nose, the kinetics follow an Arrhenius dependence Q b=bo rt 309 monoclinic lines here bo is a constant, Q is an apparent activation energy, R is the gas constant, and T is the absolute temperature. The c)400 reported activation energies are close to 100 k/mol (l ev To Line a value similar to the activation energy for oxygen vacancy diffusion extrapolated from higher temperatures. As with other nucleation and growth transformations. the C-shaped curve can be interpreted in terms of the competition between the driving forces for nucleation and the growth rate At low temperatures, below the nose, there is a high nucleation rate and the growth rate is kinetically limited by the growth velocity of the interface between the tetragonal and monoclinic 200 hases At high ten es. the nucleation rate is limiting with the driving force for the nucleation of the monoclinic pha being related to the undercooling relative to the To( /m) temperature. This is also consistent with the observations that 9096 mon the transformation rate decreased rapidly as the To( /m) temperature was approached, as well as the observation of the reverse monoclinic-to-tetragonal transformation on heating above the To(t/m) temperature Microstructural observations of the surface of polycrystalline tetragonal zirconia exposed to water show clearly the nucleation and growth of small regions of monoclinic phase, fully consis- 0210310410°10 tent with the Maj kinetics. Confocal Raman spectroscopy Time to transformation(s) confirms that the transformed regions are indeed monoclinic and, furthermore, that they do not extend deep into the surface. Time-temperature transformation (TTT) curves of sintered typically less than a few micronmeters before the whole surface yn from Tsu baking (b)TTT curves for YSZ with is transformed. Careful observations also indicate that there is a ria-concentrations.(c) TTT curves of 5-YSZ (2.8 mol% preference for nucleation at the grain junctions and corners, and al barrier coating
exponent has a value between 0.5 and 4.38 Combining the data obtained at different temperatures, the transformation kinetics form ‘‘C’’-shaped curves on a time–temperature plot. A number of examples of this time–temperature transformation (TTT) behavior are shown in Fig. 6. At temperatures well below the ‘‘nose,’’ the kinetics follow an Arrhenius dependence b ¼ b0 exp Q RT (5) where b0 is a constant, Q is an apparent activation energy, R is the gas constant, and T is the absolute temperature. The reported activation energies are close to 100 kJ/mol (B1 eV), a value similar to the activation energy for oxygen vacancy diffusion extrapolated from higher temperatures. As with other nucleation and growth transformations, the ‘‘C’’-shaped curve can be interpreted in terms of the competition between the driving forces for nucleation and the growth rates. At low temperatures, below the nose, there is a high nucleation rate and the growth rate is kinetically limited by the growth velocity of the interface between the tetragonal and monoclinic phases. At high temperatures, the nucleation rate is limiting with the driving force for the nucleation of the monoclinic phase being related to the undercooling relative to the T0 (t/m) temperature. This is also consistent with the observations that the transformation rate decreased rapidly as the T0 (t/m) temperature was approached, as well as the observation of the reverse monoclinic-to-tetragonal transformation on heating above the T0 (t/m) temperature.44 Microstructural observations of the surface of polycrystalline tetragonal zirconia exposed to water show clearly the nucleation and growth of small regions of monoclinic phase, fully consistent with the MAJ kinetics.36 Confocal Raman spectroscopy confirms that the transformed regions are indeed monoclinic and, furthermore, that they do not extend deep into the surface, typically less than a few micronmeters before the whole surface is transformed. Careful observations also indicate that there is a preference for nucleation at the grain junctions and corners, and that the transformation then extends across individual grains.45 Fig. 5. Low-temperature degradation kinetics of 10 mol% ceria-stabilized zirconia,39 3 mol% yttria-stabilized zirconia,36 Magnesium partially stabilized zirconia,40 alumina-toughened zirconia (A80Z3Y),41 and zirconia-toughened alumina (A10Z0Y)42 measured at 1341C and expected at 371C (considering an activation energy of 106 kJ/mol for all materials, which is probably not completely accurate). The shadowed areas give uncertainty ranges when they can be evaluated. Fig. 6. (a) Time–temperature transformation (TTT) curves of sintered yttria-stabilized zirconia (YSZ) (3 mol% Y2O3) and alumina doped YSZ, redrawn from Tsubakino et al. 43 (b) TTT curves for YSZ with different yttria-concentrations. (c) TTT curves of 5-YSZ (2.8 mol% Y2O3) thermal barrier coating.44 1910 Journal of the American Ceramic Society—Chevalier et al. Vol. 92, No. 9