High-resolution electron microscopy of grain boundary structures in yttria-stabilized cubic zirconia

Size: px
Start display at page:

Download "High-resolution electron microscopy of grain boundary structures in yttria-stabilized cubic zirconia"

Transcription

1 Mat. Res. Soc. Symp. Proc. Vol Materials Research Society High-resolution electron microscopy of grain boundary structures in yttria-stabilized cubic zirconia K. L. Merkle, L. J. Thompson, G.-R. Bai, and J. A. Eastman Materials Science Division, Argonne National Laboratory, Argonne, IL 60439, U. S. A. ABSTRACT The atomic-scale structure of grain boundaries (GBs) in yttria-stabilized cubic zirconia (YSZ) was investigated by high-resolution electron microscopy (HREM). Non-stoichiometric oxides have found a wide range of applications and therefore it is of importance to explore the role of GBs and their atomic-scale relaxation modes. [001] and [110] tilt GBs were examined by HREM in highly textured thin films of YSZ grown by metal-organic chemical vapor deposition (MOCVD). In addition, a special technique was developed to also allow the HREM study of twist and general GBs. GBs and triple junctions show quite dense arrangements of cation atomic columns. The GB core structures in YSZ can be contrasted to the more open structures in stoichiometric cubic oxides, such NiO, which are characterized by a relatively large GB excess volume. This appears to be due to several factors, including the necessary rearrangement of the oxygen sublattice near GBs in a CsCl 2 type structure, the redeployment of oxygen vacancies near GBs, and the segregation of Y to the GB. Relative to stoichiometric oxides, such mechanisms provide additional degrees of freedom for atomic relaxations at GBs and the development of low-energy GBs. These additional relaxation modes, which result in GB cation arrangements more akin to metallic systems, are also reflected by Burgers vector dissociations observed in low-angle YSZ GBs. INTRODUCTION Zirconia based materials are of great importance in a broad range of technological applications. In many of these areas GBs play an important and often dominant role in determining the properties of the material and functionality of the technical systems. In the Y 2 O 3 -ZrO 2 system the cubic phase can be stabilized to low temperature over a wide range of composition, from roughly 8 to 30 mol.% Y 2 O 3. Figure 1 illustrates the cubic unit cell of the YSZ lattice. The Y 3+ and Zr 4+ cations are arranged on an fcc lattice and consequently a high concentration of oxygen vacancies is present in the bulk to maintain charge neutrality. When two rigid blocks of YSZ material are put together to form a grain boundary, it is immediately clear, as illustrated in figure 2, that a considerable amount of restructuring of the oxygen sublattice is necessary to avoid energetically impossible configurations associated with having ions with like charges in close proximity to each other. To study the atomic-scale structure of GBs by HREM it is necessary that adjoining grains are closely aligned to low index zone axes. In the oxides, the scattering power of oxygen typically is too weak to give any distinctive features that can be related to the position of the oxygen sublattice, however cation columns can under favorable conditions give rise to motifs that can be directly related to their position. A detailed analysis of model structures can be performed through comparisons between experimental and simulated images, however in the AA1.6.1

2 Figure 1. Cubic YSZ unit cell. Cations Zr 4+ and Y 3+ are located on an fcc lattice. Anions occupy tetrahedral interstices. The oxygen sublattice contains a high concentration of vacancies. Figure 2. Rigid models of low angle [001] tilt GBs for NiO on the left and for YSZ on the right. In NiO anions and cations alternate within the atomic columns, whereas for YSZ anions (light) and cations (dark) occupy separate columns. present work we only consider qualitative features that can be directly extracted from HREM images taken under optimum defocus conditions. EXPERIMENTAL YSZ films in the range 10 to 15 mol% Y 2 O 3 were deposited by MOCVD in a low-pressure, horizontal, cold-wall reactor. A variety of epitaxial grade substrates was used, including MgO, YSZ, and Al2O3. At growth temperatures near 550 C and below nanocrystalline films were produced on most substrates. Anneals near 1300 C in Ar+ 0.2%O 2 typically resulted in wellconnected, [001] textured columnar grains, suitable for HREM whenever the zone axes were well enough aligned between neighboring grains. Thin sections for HREM were obtained by dimpling and ion milling. HREM in the axial illumination mode was performed on a JEOL 4000EX operated at 400 kv. Images of GBs between well-aligned grains were recorded near optimum defocus. AA1.6.2

3 DISCUSSION 1 nm Figure 3. Low-angle [001] YSZ tilt GB, θ = 8. Burgers circuits are indicated for one a[100] edge dislocation and dissociated partials. Arrows indicate dissociation width. 1 nm Figure 4. = 13, θ =22.6 [001] asymmetric tilt GB. Note curvature of (020) planes at (100)(12,5,0) GB. A detail of a symmetrical low-angle [001] tilt GB consisting of a wall of a[100] edge dislocations is shown in figure 3. The arrangement of the atomic columns within the core of the a[100] edge dislocation appears to be quite compact. Moreover, as indicated by the Burgers circuits and arrows, the a[100] edge dislocation is dissociated by approximately 1.5 nm. The exact nature of the dissociation is not known, since the HREM image gives only information of the projection in [001]. However, such a dissociation is quite unexpected and has never been observed in a stoichiometric oxide, such as NiO. High-angle YSZ GBs show Y segregation. It has also been found that YSZ [001] tilt GBs share many common structural features with other oxide GBs [1], such as for example the tendency to form coherent regions where low-index atomic planes are connected across the GB. Figure 4, however shows an additional feature of connectivity between lattice planes near a =13, θ = 22.6 asymmetrical tilt GB, where is the reciprocal density of coincident lattice sites [2]. Here the lattice planes display continuous bending over a short distance in one of the grains. This is quite surprising and atypical for high-angle grain boundaries in oxides. In metals a similar behavior is found for the connection between (111) planes across the asymmetric (557)(113), [110] tilt GB in Au [3]. In this case the GB dissociates by the incorporation of periodically spaced short lengths of stacking faults. It appears that the GB in figure 4 similarly increases its structural width over a finite region that gradually relaxes the structure to the misorientation of the second grain. Among the [001] tilt GBs, the =5, (210) symmetric tilt GB has the shortest structural repeat unit. The GB depicted in figure 5 displays several symmetric facets. The atomic-scale structure appears to be identical on the different (210) facets. This is in contrast to observations for the =5, (310) GB in NiO [4], where structural multiplicity was found, i. e. 2 different structures on facets of the same macroscopic geometry. Steps between GB facets often have dislocation character [5,6]. The symmetric facets in figure 5 are all connected by exactly one AA1.6.3

4 1 nm 1 nm Figure 5. HREM image of the = 5 tilt GB. Three (210) facets are composed of identical structural units. The connecting steps consist of one structural unit of the (430)(100) asymmetric GB. Figure 6. HREM image of high-angle YSZ [001] tilt GB, misorientation θ = nm 1 nm Figure 7. HREM image of triple junction between (110) grains. Note atomically well-structured apex of grains. Figure 8. HREM image of high-angle YSZ [110] tilt GB, θ = 86. Note connection between (111) planes Figure 9. High-angle YSZ [110] tilt GB. Note stacking faults (arrows). structural unit of the (430)(100) asymmetric GB. Thus the steps are strain-free and do not include a dislocation character. These observations are clear indications that the =5 GBs are structurally unique and form well-relaxed low-energy configurations. Depending on the misorientation and the GB plane high-angle GBs may be short period, long-period, or aperiodic. The boundary in figure 6 translates towards a different facet near the top of the figure and clearly shows a compact, dense atomic column structure throughout, with no indications for large open spaces at the GB core. In contrast high-angle GBs in stoichiometric cubic oxides, such as NiO, typically can be characterized by quite large (compared to metals) AA1.6.4

5 a 10 nm SiO2 (110) YSZ b (110) YSZ Figure 10. Schematic for preparations of general GBs. (a) Substrate preparation (b) MOCVD growth. 1 nm Figure 11. HREM image of YSZ GB with tilt and twist components. Note well-connected, coherent interface. excess volumes [4, 7]. The open structure in these materials are thought to be due to the strong ionic interactions which prevent the close proximity of like ions. Well-structured triple junctions without amorphous pockets were also found in these samples. Figure 7 shows a triple junction at the union of three [110] oriented grains. The GBs directly join at the triple junction without the formation of a second phase. Although [001] was the typical texture in the MOCVD grown films, the [110] texture was obtained in a singular case of a film grown on [001] MgO. Just as in the [001] tilt GBs, the [110] tilt GBs that were observed were well structured at the atomic scale. The tendency to form coherency between low-index planes and the development of asymmetrical GBs, that incorporate a low index plane at one side of the boundary can be well recognized in figure 8 which depicts a high-angle (θ=86 ) [110] tilt GB. In [110] YSZ tilt GBs we have even found indications for the formation of stacking faults on (111) planes as indicated by arrows in figure 9. The incorporation of stacking faults in high-angle [110] tilt GBs is a common feature in low stacking-fault-energy metals. In this manner low-energy GB relaxations are often achieved by GB dissociations, involving stacking faults [3,8-10]. The picture that evolves from the above direct observations suggests that the atomic structure of high-angle YSZ GBs is much more metal-like than in stoichiometric cubic oxides, such as NiO. Of course, our observations only relate to the positions of the cation atomic columns. Major open questions relating to the oxygen sublattice, the oxygen vacancy distribution, and the details of the Y segregation as well as the ionic charge distribution at GBs remain and need to be addressed by analytical techniques. We conclude that the relaxation processes that are possible within the multidimensional parameter space of nonstoichiometric oxides allow the formation of low-energy GB configurations that are substantially more relaxed than is possible in stoichiometric oxides. HREM OF GENERAL GRAIN BOUNDARIES HREM typically is limited to the direct observation of tilt GBs along a low-index zone axis. However, twist and general GBs can also be studied by HREM, if the two grains forming AA1.6.5

6 the GB are aligned along two different low-index zone axes. Samples of this type can be produced by a special epitaxy technique [11] or in the present system by combining the textured [001] growth with epitaxial regrowth on an [110] YSZ substrate. A schematic of the growth geometry is illustrated in figure 10. Vacuum evaporation of SiO 2 onto a (110) YSZ substrate through a physical mask is used to prepare a fine pattern of 10 nm thick layers of SiO/SiO 2. This is followed by the MOCVD deposition of YSZ, which grows epitaxially on (110) YSZ and with a (001) columnar texture on the regions covered by SiO 2. In this manner general GBs suitable for HREM can be fabricated. Here GBs with finite tilt and twist components are formed between adjoining grains with parallel but distinct zone axes, such as in figure 11. Clearly, this boundary is also quite well structured at the atomic scale and in addition shows coherence between the (200) planes crossing the interface. CONCLUSIONS Direct observation by HREM has shown that high-angle GBs in YSZ are well-structured at the atomic scale without evidence for an amorphous phase. Well-annealed triple junctions are also devoid of extraneous phases. The core structures of YSZ GBs appear quite dense, in contrast to the relatively open structures in NiO. Moreover, several GB atomic-scale structural features indicate that the GBs are very well relaxed to low-energy configurations. It appears that, more generally, the cation substructure of GBs in non-stoichiometric oxides may be more akin to metallic GBs than the stoichiometric cubic oxides, such as NiO. We believe that the origin of this behavior lies in the additional degrees of freedom that are available for atomic relaxation, involving rearrangements of the oxygen sublattice and intrinsic point defects, as well as GB segregation. For the first time a technique was developed that allows the atomic-scale investigation of general oxide GB, i. e. GBs in which the misorientation vector does not lie in the GB plane. Nevertheless, great challenges remain in understanding the atomic-scale nature of YSZ GBs, which includes the determination of GB chemical, point defect, and ionic charge distributions. ACKNOWLEDGEMENTS This work was supported by the U. S. Department of Energy, Basic Energy Science - Materials Science, under contract W Eng-38. REFERENCES 1. K. L. Merkle, R.-R. Bai, Z. Li, C.-Y. Song, and L. J. Thompson, phys. stat. sol. (a) 166, 73 (1998). 2. W. Bollmann, Crystal defects and crystalline interfaces, (Springer, Berlin, 1970). 3. K. L. Merkle, Microscopy and Microanalysis 3, 339 (1997). 4. K. L. Merkle, and D. J. Smith, Phys. Rev. Lett. 59, 2887 (1987). 5. A. P. Sutton, and R. W. Balluffi, Interfaces in Crystalline Materials, (Clarendon Press, Oxford, 1995). 6. R. C. Pond. Interfaces and Dislocations, Oxford. Oxford University Press 1985, K. L. Merkle, and D. J. Smith, Ultramicroscopy 22, 57 (1987). 8. K. L. Merkle, and D. Wolf, Materials Science Forum , 65 (1993). 9. K. L. Merkle, Interface Science 2, 311 (1995). 10. J. D. Rittner, and D. N. Seidman, Phys. Rev. B 54, 6999 (1996). 11. K. L. Merkle, and L. J. Thompson, Phys. Rev. Lett. 83, 556 (1999). AA1.6.6