PHASE STABILITY ISSUES IN EMERGING TBC SYSTEMS
Noemí R. Rebollo, Ashutosh S. Gandhi and Carlos G. Levi
University of California, Santa Barbara Materials Department, Engineering II Santa Barbara, CA 93106-5050 ABSTRACT The high temperature stability of metastable single-phase zirconias codoped with Y plus trivalent rare earth cations is discussed. The motivation arises from strategies to enhance the insulating efficiency and/or higher temperature capability of thermal barrier systems by co-doping the conventional ZrO2-7.6%YO1.5 composition. The fundamental issue is the resistance to partitioning of the metastable tetragonal (t') or cubic (c') phases into the equilibrium assemblage dictated by the phase diagram, whereupon the stabilizer-depleted tetragonal phase becomes susceptible to the disruptive monoclinic transformation. The experiments rely on compositions synthesized by precursor pyrolysis, all of which yield initially supersaturated single-phase solid solutions. Clear trends are noted in the phase stability of single and co-doped compositions, which increases with decreasing the size of the rare earth cation. Thermodynamic arguments are offered to rationalize these trends, with additional comments on the potential roles of diffusion and clustering phenomena. INTRODUCTION Zirconias co-doped with yttria and one or more rare earth oxides (REO) offer potential benefits in reducing the thermal conductivity of thermal barrier coatings (TBCs) below current levels [1, 2]. In general, these materials are based on REO additions to the standard ~7.6 mole% YO1.5 partially stabilized zirconia (7YSZ), and occur as metastable single phases instead of the tetragonal + cubic assemblage expected from the equilibrium phase diagram (Fig. 1). In this metastable condition, the material can be thermally cycled without undergoing the disruptive tetragonal-monoclinic transformation that compromises the mechanical integrity of the coating. Because this transformation is generally diffusionless, "non-transformability" requires selecting a coating composition whose T0(t/m) temperature is below ambient. However, prolonged exposure to high temperature (1200°C) drives the microstructure to its equilibrium two-phase configuration [3], yielding a Y-depleted tetragonal form whose T0(t/m) is now above ambient (Fig. 1). This renders the coating "transformable" to monoclinic degrading the tolerance of the coating to thermal cycling. In this context, "phase stability" reflects the resistance of the coating material to this partitioning process at high temperature, and is thus a requisite (albeit not sufficient) condition for TBC durability in gas turbine applications. The present investigation aims to evaluate and understand the effects of rare earth co-doping on the high temperature stability of the t' phase, using the standard 7YSZ material as a baseline.
Published in "High Temperature Corrosion and Materials Chemistry IV " E. Opila, P. Hou, T. Maruyama, B. Pieraggi, M. McNallan, D. Shifler, and E. Wuchina (eds.) Electrochemical Society Proceedings vol. PV-2003-16, pp. 431-442 (2003).. EXPERIMENTAL Two groups of materials based on 7YSZ were investigated. In one the Y was fully replaced by a RE cation and the stabilizer content was kept constant at 7.6% MO1.5. In the other, 7.6% REO (or yttria) was added to the standard 7YSZ, bringing the total stabilizer content to 15.2%. All compositions were synthesized by reverse co-precipitation of precursor powders from mixed solutions containing the requisite cations [4]. Starting materials were Zr acetate solution in dilute acetic acid (Aldrich) and RE nitrates in powder form (Alfa Fig. 1. Zirconia-rich end of the ZrO2-YO1.5 phase diagram showing the temperatures for diffusionless Aesar), with a minimum purity of transformation from cubic to tetragonal, T (c/t), 0 99.9%. The nitrates were prepared as and from tetragonal to monoclinic, T0(t/m). The aqueous stock solutions, assayed for dotted line at 7.6% represents the standard TBC oxide content, and then mixed with composition comprising the metastable t' phase. the Zr acetate solution in the proportions needed to yield the desired compositions. Precipitation was effected by drop-wise addition of the mixed solution to aqueous NH4OH (pH = 9). This approach is more effective than straight co-precipitation (i.e. by adding the base to the precursor solution) when homogeneity of the powders is a concern [4]. The precipitates were dried and pyrolyzed at 900°C for 2 h yielding single-phase supersaturated solid solution powders. The crystallite size, calculated by the Scherrer method [5, 6] was reasonably constant for all pyrolyzed powders: ~31 nm on average with a standard deviation of ~4 nm. Specimens were heat treated primarily as loose particulate, without compaction, to minimize constraint effects that may hinder the tetragonal-monoclinic transformation. (Comparative studies with compacted and coating specimens of selected compositions are underway to evaluate these effects.) The heat treatment schedule consisted of 24 h cycles, the first four at 1200°C with increments of 50°C in each subsequent cycle. For these treatments the samples were (i) placed within a covered crucible, (ii) introduced into a furnace pre-heated to the prescribed temperature, (iii) held isothermally for 24h and, (iv) withdrawn to cool in air. Phase analysis by X-ray diffraction at room temperature was performed after each cycle, with Raman spectroscopy in selected cases. The approximate fraction of monoclinic was estimated from:
Xm = I( 1 11) m + I(111) m I( 1 11) m + I(111) m + I(111) t / c (1) wherein I is the integrated area of the (111) peak for the monoclinic (m), tetragonal (t) or cubic (c) phases. The heating schedule was designed to rank the compositions according to their resistance to de-stabilization, rather than to quantify the partitioning kinetics of a specific composition [7]. Studies of the latter type are in progress. m RESULTS AND DISCUSSION
m Results for powders of the baseline 7YSZ material are given in Fig. 2. Because of their fine crystallite size (~30 nm), it is not im+1450C/24h mediately obvious whether the powders after pyrolysis are tetragonal or cubic. Their c single phase nature, however, is conclu+1400C/24h sively ascertained from the XRD pattern after the first 24h cycle at 1200°C (not in Fig. +1350C/24h 2), which clearly shows the peaks characteristic of single-phase t-ZrO2. Evidence of supersaturation is provided by the appear+1300C/24h ance of a (200) cubic peak between the (200)/(002) tetragonal peaks in the patterns +1250C/24h for 1350-1450°C in Fig. 2. It is noted that monoclinic does not appear concurrently t' with the first observation of this cubic peak, t' +1200C/96h presumably because the associated depletion of Y from the t' phase is still insufficient to render it transformable. (The absence of m900C/2h ZrO2 at these temperatures was confirmed by Raman spectroscopy.) At some point 25 30 40 2 35 small monoclinic peaks may appear, e.g. Fig. 2. XRD patterns for ZrO2-7.6%YO1.5 after 1450°C in Fig. 2. These may develop after various stages in the heat treatment. t' fully in the next stage of the heat treatment denotes the (200)/(002) tetragonal peaks, c the (1500°C), but in other compositions they (200) cubic peak and m the (111) monoclinic grow only slightly and may even disappear peaks. at higher temperatures. The latter is a consequence of the reduction in the relative amount of equilibrium tetragonal phase with increasing temperature, and hence in the amount of monoclinic that can form from it. Since a small amount of monoclinic is presumably tolerable, the "practical" onset of destabilization was set when partitioning leads to Xm in Equation (1) to be >10%. (An exception is the La co-doped 7YSZ, wherein extensive partitioning is observed but no evidence of monoclinic is detected, as discussed later.)
+1500C/24h 7.6Y Singly Doped Compositions The temperature of the heat treatment stage immediately prior to "substantial" partitioning, e.g. 1450°C in Fig. 2, is designated as the "stability limit" for the purposes of the present comparison. These temperatures are plotted in Fig. 3 for the two groups of materials investigated. It is evident in this figure that the effectiveness of a single dopant in preserving t' stability decreases systematically with increasing ionic size within the range studied. All materials exhibit essentially the same XRD pattern in the as-pyrolyzed condition (cf. Fig. 2), with La showing the least evidence of tetragonality. Only the slightest hint of monoclinic was detectable at the highest temperatures for the 7.6Yb composition, whereas materials based on Nd and La partitioned extensively during the first 24h cycle at 1200°C (i.e. their "stability limit" would be below the range investigated). In both Nd and La the products of partitioning were the pyrochlore zirconate and Fig. 3. Relative resistance to de-stabilization in the systems investigated, plotted as the temperature of the last heat treatment stage before the onset of substantial monoclinic formation, as a function of ionic size (8-fold coordination) [8]. The normalization length scale, r , is the radius of the interstitial "cubic" site coordinated by 8 oxygen anions. The circles represent singly doped compositions (Y or rare earth), while the diamonds represent ternary compositions containing equal proportions of Y and a RE dopant, with the 15.2%YO1.5 composition added for comparison. The empty symbols denote compositions that did not exhibit any signs of partitioining. The radii of Sc and In, two alternate trivalent stabilizers, are given for reference. tetragonal zirconia, subsequently transformed to monoclinic. Sm and Gd behaved qualitatively like Y, except for the lower stability temperature and the absence of the pretransition (200) cubic peak. The reasons for the trend of decreasing stability with ionic size are not immediately evident. The decomposition of t' into the equilibrium tetragonal + cubic forms requires long-range cation diffusion. Since the host structure is the same and the nature of the rate-controlling defects for cation diffusion [9] is expected to be similar, one may argue that the diffusion rates should vary inversely with cation size. The implication is that phase stability would increase with increasing dopant size, which is clearly at variance with the experimental observations. Conversely, the phase equilibria information on these systems, albeit limited and still in need of verification, suggests the driving force may also change systematically with ionic size. Figure 4 shows tentative binary diagrams for ZrO2 and Lanthanide oxides, including YO1.5 for comparison. Notable are the systematic trends with decreasing ionic size, including the increase in the extent of the cubic (fluorite) field and the associated t + c equilibrium, as well as the decrease in the stability of the pyrochlore zirconate M2Zr2O7. The ZrO2-YO1.5 diagram fits very well within this trend if placed according to ionic size, but not if placed according to atomic number/mass (smaller than La). Two different groups can be identified in the context of the problem at hand. One comprises Sm, Gd, Y and Yb, wherein de-stabilization involves the precipitation of cubic phase from the supersaturated t' at all temperatures of interest (1200-1500°C). The other group includes La and Nd, which would tend to precipitate the zirconate instead of the cubic phase to relieve the supersaturation of t', albeit only at the lower temperatures in the case of Nd. Fig. 4. Binary phase diagrams for the stabilizers investigated, in order of decreasing ionic size, adapted from calculations by Yokokawa [10]. The S-P ionic sizes are indicated in parenthesis. The tetragonal + cubic field is shaded in gray. The hatched region represents the relevant twophase fields and temperature range for the materials investigated. Note the well behaved trend in phase equilibria including that pertaining to Y, which does not belong to the Lanthanide period but fits the trend based on ionic size. It can be shown that the maximum driving force for the separation of t' into t + c increases with the width of the corresponding two-phase field [7]. Prior reports in the literature [11, 12] suggest that the the t + c field widens systematically from Y to Nd. The inference is that the driving force for phase separation may increase with increasing ionic size, as elaborated below. It is noted, however, that there are inconsistencies between the cited studies and the t + c boundaries are still under debate, even for Y. Notably, the equilibrium tetragonal boundaries in [11] exhibit a retrograde shape with a maximum in solid solubility at temperatures much higher than those suggested by more comprehensive phase diagram evaluations, e.g. [13, 14] for Y--cf. Fig. 1. In the context of the problem at hand, the phase diagram information may be cast in the form of two distinct thermodynamic scenarios, illustrated in Fig. 5. The baseline is the ZrO2-YO1.5 system, for which assessed Gibbs free energy descriptions of the relevant phases are available--curves marked Y in Fig. 5(a). The width of the two-phase field is determined by the common tangent, in the usual fashion, and the overall driving force for partitioning is the distance between the tetragonal free energy curve, G(t), and the point on the common tangent corresponding to the same composition. The structural model for oversized trivalent dopants in ZrO2 includes 8-fold coordination for the dopant with the anion vacancies located preferentially in the vicinity of Zr ions to relieve the "oxygen crowding" around them [15-17]. The dopant is compressed by the surrounding lattice [16] owing presumably to its size difference with Zr (from ~16% for Yb to ~40% for La). Arguably, the associated strain energy in the lattice may induce the free energy vs. composition curve to rise faster for a larger cation, as illustrated schematically for Gd in Fig. 5(a). In consequence, the equilibrium tetragonal and cubic compositions would shift to lower and higher value, respectively, relative to Y, reflecting the widening of the twophase field with increasing cation size discussed above [11].1 More importantly, the driving force for de-stabilization would likely increase as the intersection of the free energy curves shifts upward with increasing ionic size (inset in Fig. 5a). Fig. 5. Schematic Gibbs free energy curves showing the thermodynamic scenarios that may account for differences in the driving force for partitioning among the systems investigated. The curves for the ZrO2-YO1.5 system were calculated from revised thermodynamic models for the cubic and tetragonal phases developed by Fabrichnaya [14]. The rest of the curves are schematic, drawn to be qualitatively consistent with the phase diagrams and the expected effects of ionic size on the free energy of the relevant phases.
1 Katamura et al. [11] proposed a continuous free energy curve wherein the tetragonal form is viewed as an "ordering" of the oxygen lattice relative to the cubic array in ideal fluorite. The difference is not significant for the purposes of the present discussion, but may become an issue in elucidating the partitioning mechanism. For systems in which the pyrochlore is preferred over cubic as the precipitate from a supersaturated tetragonal phase, the difference in driving force is likely to be more substantial. The general situation is illustrated in Fig. 5(b). The relative positions of the free energy curves below the c (fluorite) t + Py eutectoid (Fig. 4) is such that the overall driving force for pyrochlore precipitation should be significantly larger than that for the cubic phase, as inferred from the relative positions of the respective common tangents. A significant degradation in phase stability should then be expected as one moves from the systems in which the preferred precipitate is cubic (Yb, Y, Gd) to those in which it is pyrochlore (Nd, La), in agreement with the experimental results. Sm represents an intermediate case in which the lowest heat treatment temperature (1200°C) is quite close to the c t + Py eutectoid and seems to be only marginally more stable than those compositions forming pyrochlore (Fig. 3). A thermodynamic argument is thus more consistent with the experimental observations than one based on differences in diffusion kinetics. However, diffusion kinetics is clearly important in that it prevents partitioning at the lower temperatures (where "lower" depends on the relative driving force and hence on the dopant). Moreover, limited experiments on Y and Gd single-doped systems reveal that phase stability is enhanced by increasing composition [7], a result that cannot be fully explained with thermodynamics and requires improved understanding of the partitioning kinetics. For example, Fig. 5 shows that the driving force for partitioning at a constant temperature should increase monotonically from the tetragonal boundary to the composition corresponding to the T0(c/t) curve, and then decrease monotonically to vanish at the cubic boundary. However, the partitioning resistance of Y-doped t' increases in the order 7.6%