J. Phys. Chem. B 2001, 105, 12481-12489
12481
Structural Characterization of the CeO2/YSZ(111) Catalytic System Synthesized Using Simulated Amorphization and Recrystallization S. Andrada Maicaneanu,† Dean C. Sayle,*,† and Graeme W. Watson‡ Department of EnVironmental and Ordnance Systems, Cranfield UniVersity, Royal Military College of Science, ShriVenham, Swindon, United Kingdom, SN6 8LA, and Department of Chemistry, Trinity College, Dublin 2, Ireland ReceiVed: March 19, 2001; In Final Form: June 22, 2001
A “simulated amorphization and recrystallization” technique is employed to explore the structural modifications and epitaxial relationships, which evolve within a CeO2 thin-film when supported on yttrium-stabilized zirconia (YSZ). The epitaxial relationship identified within the CeO2/YSZ(111) system facilitates a reduction in the lattice misfit from ca. 6.1% (bulk misfit) to ca. 1% thereby, in part, stabilizing the system. In addition, edge dislocation networks and defects, including vacancies, substitutions, and interstitials evolve in both the CeO2 and the underlying YSZ to reduce further this residual misfit. Graphical techniques are employed to characterize and present, the epitaxial relationships and atomistic structure of the dislocation networks and defects that evolve within the system. Because experimentation is, at present, unable to explore the structural features that evolve within supported thin films with three-dimensional atomistic resolution, simulation provides an invaluable complement with which to study supported catalysts.
Introduction Ceria, its structure, properties, and applications (catalysis, fuel cell technology) has received considerable attention over the past few years.1,2 As a catalyst and catalyst promoter, ceria and ceria containing materials are used in many processes such as the removal of sulfurous oxides from flue gases, removal of organics from wastewaters, ethylbenzene dehydrogenation and many oxidation reactions.1,3 However, perhaps, the most important application of ceria is as a component in three-way catalysts (TWC) for the treatment of automobile exhaust gases.1 The utilization of ceria in this type of catalyst is based mainly upon its so-called oxygen storage capacity (OSC), which represents the ability of ceria to shift from Ce4+ to Ce3+ in reducing atmospheres and from Ce3+ to Ce4+ under oxidizing conditions with charge compensation facilitated via oxygen vacancies
CeO2 a CeO2-x + 1/2xO2
(1)
Accordingly, ceria based catalysts can work in both oxidizing and reducing conditions, converting carbon monoxide, nitrogen oxides, and hydrocarbons into nontoxic products.1 Specifically, under reducing conditions, ceria forms oxygen-deficient nonstoichiometric oxides, CeO2-x (0 < x e 0.5), while maintaining a fluorite type crystal structure, even at elevated temperatures.1,2 Experimental and theoretical studies have shown that the reducibility of cerium oxide can be modified by introducing dopant cations into the material through the preparation of solid solutions. For example, the introduction of Zr4+ into the ceria lattice results in a defective fluorite structure with improved oxygen mobility.1,4,5 * To whom correspondence should be addressed. E-mail: sayle@ rmcs.cranfield.ac.uk. † Department of Environmental and Ordnance Systems, Cranfield University, Royal Military College of Science. ‡ Department of Chemistry, Trinity College.
An additional mechanism for enhancing the catalytic properties of CeO2 is to support a thin film of the material on an oxide substrate. This can have a profound effect on the structure and hence material properties of the supported material.6 For example because there is likely to be a lattice misfit associated with the system, structural modifications arise within the thin film, which act to reduce the lattice misfit thereby stabilizing the system. These may include for example dislocation arrays, point defects, and reduced interfacial densities. Although such defects may be deleterious to certain applications (notably dislocations within superconducting materials7), within catalytic systems they can be advantageous, enhancing the catalytic properties of the material. For example, recent theoretical studies of crystal growth8 show enhanced binding at the surface termination of a dislocation indicating enhanced reactivity at low coordinated sites. In addition, a theoretical study by Sayle and co-workers9 suggests that CeO2(111), when supported on Al2O3(001), exhibits a lower oxygen vacancy formation energy compared with the unsupported CeO2. Experimental studies have identified weakly bound surface oxygen species on ceria films deposited on Al2O3, which are not present on the unsupported CeO2(111).10 A comparison between the properties of ceria thin films vapor deposited onto polycrystalline zirconia and zirconia based substrates such as yttrium-stabilized zirconia (YSZ), showed that these supports are more efficient than R-Al2O3 in promoting ceria reducibility.11,12 The properties of the system can be enhanced further via the addition of metals such as Rh and Pt.1,3 It has also been shown that the OSC can be improved by thermal treatment of the catalyst.1 It is well-known that catalytically active centers are linked with low coordinative saturation of the surface ions of the supported thin film.13,14 These may arise from defects located at the surface including steps, edges and vacancies, or by defects terminating at the surface, such as dislocations and grainboundaries. All these types of defects can appear during the preparation and thermal treatments of the material. For ceria,
10.1021/jp011033j CCC: $20.00 © 2001 American Chemical Society Published on Web 11/21/2001
12482 J. Phys. Chem. B, Vol. 105, No. 50, 2001
Maicaneanu et al.
the highly defective surfaces that include grain boundaries, defects or surface regions with small ceria crystallites, all contribute to the OSC of ceria and facilitate oxygen transfer, on model Rh/CeO2 catalytic systems, from ceria to the metal.15 Moreover, recent advances in materials fabrication, has enabled almost atomistic control over the structure of the material.13 Clearly, one aspires to generate materials with optimum catalytic properties. However, to realize this objective one must first characterize structurally the defects, which may give rise to the catalytic properties of the material. Experimentally, it is very difficult to obtain a complete structural characterization of a supported thin film. Highresolution electron microscopy techniques can provide (almost atomistic) images of specific regions of a material.16,17 However, a complete three-dimensional atomistic description of the structure is not possible at present. (We note however that transmission electron microscopy techniques are currently being developed to enable three-dimensional imaging18). Accordingly, atomistic computer simulation offers opportunities to complement experiment in the exploration of interfacial structures since the position of all the atoms is explicitly defined. We consider in this present study CeO2 thin films supported on an YSZ substrate. Specifically, we consider the model system, CeO2/YSZ(111) with a level of yttrium doping equal to 9.24 mol. % Y2O3, which enables high ionic conductivity19,20 and stability of the cubic phase at low temperatures.21 The focus of the study is to explore the influence the YSZ support may have with respect to the overlying CeO2 and to characterize the structural defects, which evolve within the CeO2. Because it has been shown that by supporting CeO2 on a substrate material the catalytic properties may be enhanced, elucidating the structural features, which evolve upon supporting the material will have important implications for catalytic systems. Most atomistic simulations exploring interfacial structures proceed by defining the basic configuration of the interface, which is then simulated using static or dynamical methods. Accordingly, the investigation of ceria thin films deposited on an YSZ support may proceed in two ways. First, the ceria thin film may be placed directly on top of the YSZ support, with CeO2 thin film lying directly above their respective counterions of the underlying support, which corresponds to a coherent structure. Here, the YSZ support will maintain its natural lattice parameter (5.1 Å), whereas the CeO2 thin film must be compressed from 5.4 to 5.1 Å to ensure coherence. The misfit, (F), associated with this configuration is +6.1% following
F)
(naCeO2 - maYSZ) (naCeO2 + maYSZ)/2
(2)
where aCeO2 and aYSZ represent the lattice parameters for CeO2 and YSZ respectively, and n and m, the number of unit cells. To maintain coherence the CeO2 needs to be compressed, (C), by +5.9% following
C)
(naCeO2 - maYSZ) naCeO2
(3)
An alternative construction is to generate a commensurate structure. Here, the incommensurate relationship between the two materials can be improved by placing, for example, a 13 × 13 CeO2 thin film on top of a 14 × 14 YSZ support, resulting in a lattice misfit of -1.3%. To accommodate the misfit, the CeO2 thin film must be compressed by -1.3%. This can be extended using a near coincidence site lattice (NCSL)22,23
approach as applied to some particular CeO2(hkl)/YSZ(h′k′l′) system. From this, a variety of different commensurate configurations can be constructed with low associated lattice misfits. To generate models of supported oxide thin-films that are more realistic, various structural features, including the epitaxial relationships, various defects, and reduced interfacial ion densities must also be introduced within the model. The defects, which evolve in response to misfit accommodation, may include dislocations arrays, vacancies, substitutions and interstitials including clustering of such defects. All these types of defects will evolve and act to reduce the lattice misfit thereby stabilizing the thin film. Indeed, it is these defects, which will, in part, be responsible for the catalytic properties of the material. However, simulations based upon the initial configurations generated by some particular NCSL are very unlikely, under dynamical simulation, to surmount the energy barriers required to facilitate the formation of these defects. This is because the energy barriers to evolve for example a dislocation within a crystalline material are likely to be very high, and therefore, the duration of the dynamical simulation required to evolve such a structure would be prohibitively large. Consequently, this approach becomes untenable in generating a realistic model for the CeO2/ YSZ catalytic system or indeed any supported thin film. To address this problem, one might consider introducing each of the structural features ‘by hand’ and then simulating this ‘trial system’ using energy minimization or dynamical methods. However, such an approach is no longer appropriate because, for example, the nature, location and concentration of the dislocation arrays and associated defects presents a prohibitively large number of permutations to consider. In addition, the inclusion of a dislocation will influence the structure of neighboring dislocations. The model must therefore also include the synergy of interaction of each of the structural features present in the system and the effect such synergistic interactions have on each of the structural features present within the system. Therefore, we require an approach whereby the simulation will allow a natural evolution of all structural features that would exist within a real system. In modeling supported thin films, there are two primary requirements for the simulation. First, the simulation cell must be sufficiently large to accommodate the incommensurate nature of the system24 and second, the final structure should not be influenced by the starting structure. For example, if the initial configuration were erroneous, any final structure, which reflects structurally the (erroneous) preparatory configuration, would also be suspect. Accordingly, we employ a simulated “amorphisation and recrystallization” methodology,25,26 which resolves each of the problems discussed above. Simulated amorphization and recrystallization involves forcing the thin film to undergo a transformation to an amorphous state prior to recrystallizing, and therefore, the recrystallization process rather than the initial structure will dictate the final structure. In addition, the recrystallization process is controlled by the interaction between the amorphous thin film and the substrate, which does not undergo an amorphous transition. Central to this methodology is that dynamical simulation, as applied to amorphous structure, allows a more comprehensive exploration of the configurational space due to the high energy amorphous starting point and the conformational freedom this gives rise to. In particular, the initial strain under which the thin film is constrained is critical in generating the desired highenergy amorphous transition. If the strain is too low, the system may not go amorphous and therefore no structural modifications will evolve. Alternatively, for greater but still inadequate initial
Characterization CeO2/YSZ(111) Catalytic System strain the system will go amorphous, although the amorphous structure may have insufficient energy to allow complete exploration of the potential hypersurface. Here, the procedure may result in certain regions, which remain amorphous.27 Conversely, if the initial amorphous inducing strain is too large, the velocities of the ions will be so high such that the thin film will loose all integrity and the simulation will fail catastrophically. Preliminary calculations, using various trial starting structures have to be performed for each system because it was observed that the initial strain required to induce adequate amorphization is system dependent. For simple binary oxides, the tolerances are high, and a wide range of strains are valid, while for systems with fluorite structures, it was found that the tolerances are narrow. Although the initial strain imposed upon the supported thin film is the main driving force to amorphization, the temperature at which the dynamical simulation is performed plays an additional but important role. For example, the procedure can be performed equally as well at 20 K as at 2000 K. However, the recrystallization process at 20 K is much slower. In essence, the optimum temperature is one that allows the structure to evolve but that falls short of melting the thin film. This would be detrimental as it would prevent recrystallization and require an additional quenching step.26 Because the amorphous solid can explore its flatter potential energy surface more readily than starting from a crystalline solid, the dynamical simulation is capable of surmounting the energy barriers to enable the film to accommodate appropriate (energetically favorable) structures. These include the epitaxial relationships with respect to the underlying support together with a range of defects, such as dislocations, grain-boundaries, vacancies, interstitials, and reduced interfacial ion densities including intermixing of ions across the interface, which evolve to lower the energy.28,29 Methodology The calculations presented in this study are based on the Born model for ionic solids, with potential parameters taken from Lewis and Catlow30 for Y3+, Dwivedi and Cormak31 for Zr4+ and O2- ions and Sayle32 for Ce4+ ions. These potentials have been employed to model lattice parameters,33 thermal expansivities,34 conductivity and diffusion properties34 for CeO2ZrO2, ZrO2-MxOy and CeO2-MxOy solid solutions, in accord with experiment. We have not attempted to model Ce3+ due to the complexity of randomly adding Ce3+ (in addition to oxygen vacancies to maintain charge neutrality), the concentration of which would be dependent experimentally on the partial pressure of Oxygen. We have thus assumed that the cerium is fully oxidized and stoichiometric. In addition, a mean field strategy35 was employed to represent the fixed ions comprising the yttriumstabilized zirconia support (considering the standard two region methodology36). The ZrxY1-xO2-x/2 system is assumed to contain a single, ‘hybrid’ cationic species, between Zr4+ and Y3+. Considering that 10% (x ) 0.1) of the zirconium ions are replaced with yttrium, and oxygen vacancies are created to preserve electroneutrality of the system, the new ‘hybrid’ charges will be +(4 - x) and -(2 - x/2) or +3.9 and -1.95, respectively. Accordingly, the short-range potentials parameters describing ‘hybrid’-‘hybrid’ interactions and ‘hybrid’-ion interactions can be calculated35,37 (Table 1). New lattice parameters for yttriumstabilized zirconia were then calculated, using the GULP code,38 based upon these ‘hybrid’ potentials. The approximation of the rigid ion model was also imposed to reduce computational expense.
J. Phys. Chem. B, Vol. 105, No. 50, 2001 12483 TABLE 1: Potential Parameters of the Form V(r) ) A exp(-r/G) - Cr-6 Employed in This Study; M3.9+ Represents the Hybrid Yttrium/Zirconium Species ionic pair Ce4+-O2Zr4+-O2Y3+-O2O2--O2-
A [eV]
F [Å]
short-range potential parameters 1986.83 0.3511 985.87 0.3760 1345.10 0.3491 22764.30 0.1490
C [eV Å-6] 20.40 0.00 0.00 27.89
short-range potential parameters using mean field approach M3.9+-O21021.79 0.3733 0.00 M3.9+-O1.95996.25 0.3733 0.00 Ce4+-O1.951937.16 0.3511 19.89 Zr4+-O1.95961.22 0.3760 0.00 Y3+-O1.951311.47 0.3491 0.00 O2--O1.9522195.20 0.1490 27.19 O1.95--O1.9521626.10 0.1490 26.50
The dynamical simulations, which employ three-dimensional periodicity, were performed using the DL_POLY code,39 and therefore, a void, 60 Å in size, normal to the surface was introduced to represent the free surface. The simulation cell contains ions distributed in two regions:36 Region I comprises the thin film and the first six repeat units of the support and ions within this region are allowed to move under dynamical simulation, whereas region II comprises ions (four repeat units of the support) that are kept fixed to ensure the correct crystalline environment. In this study, the CeO2/YSZ(111) system is explored using a simulated amorphization and recrystallization methodology.25,26 Here, the ceria thin film is constrained initially under considerable pressure and placed on top of the support. The thin film responds, under dynamical simulation, to this huge initial pressure and relieves it via a transition of the thin film from a crystalline to an amorphous structure. Prolonged dynamical simulation results in the recrystallization of the thin film together with the evolution of structural modifications as the system responds to the lattice misfit and interaction potential of the support. The evolution of the amorphization and recrystalization processes was monitored using radial distribution function (RDF) and mean square displacements (MSD) of the ions during the dynamical simulation. It is pertinent to suggest that the simulated recrystallization parallels recrystallization, which exists for real systems such as vapor deposition.12 However, the amorphous starting configurations are of high energy and do not therefore reflect real systems. Moreover, the time scales required to effect the recrystallization are much smaller compared with real crystallization. Typically, they are of the order of a nano-second. The simulation is therefore a technique to derive a range of low energy configurations, which comprise various structural features observed in real systems and the evolution or dynamical recrystallization bears little physical significance. Rather, it is only the final structures, which provide useful models to compare directly with experiment. One important criterion identified above was that the simulation cell must be sufficiently large to accommodate the incommensurate nature of the system. Accordingly, a single meso-scale simulation has been performed facilitating the evolution of a multitude of structural features, including dislocation networks, vacancies and interstitials, within the simulation cell. This is in contrast to performing many smaller simulations, which would necessarily comprise fewer structural features. Previous simulations on the SrO/MgO(001) system26 using different simulation cells revealed equivalent thin film energies, epitaxial relationships, dislocation densities and struc-
12484 J. Phys. Chem. B, Vol. 105, No. 50, 2001
Maicaneanu et al.
Figure 1. Schematic representation of the three stages involved in constructing the thin-film interface.
tural configurations, suggesting that a single very large simulation cell is sufficiently representative. The YSZ(111) surface was chosen specifically because it is the most stable fluorite surface; future studies will focus on supporting CeO2 thin films on alternative surfaces. In addition, for this present study, we have considered a simple bulk termination as the simulated amorphization and recrystallization will allow the substrate to respond structurally to the overlying thin film. Consequently, performing an initial relaxation on the substrate was deemed unnecessary. For the overlying CeO2 thin film, the analogous CeO2(111) was chosen. This was chosen arbitrarily (albeit based upon ease of initial construction) as the amorphization and recrystallization will enable the CeO2 to evolve into any orientation it desires with respect to the underlying YSZ(111). Specifically, in a previous study, exploring SrO/MgO(110), the SrO was generated initially with the SrO(110) exposed at the surface and interfacial planes. Upon recrystallization, the SrO transformed from the preparatory SrO(110) orientation to the SrO(100), resulting in the generation of the SrO(100)/MgO(110) system, demonstrating clearly that the methodology is capable of allowing such a transformation to occur.40 Construction of the Model To generate the CeO2/YSZ(111) interface three steps were followed (Figure 1). First, a CeO2/ZrO2(111) system was constructed by placing two CeO2(111) repeat units (thick) directly on top of 10 repeat units of ZrO2(111) support, (six repeat units included in region I and four repeat units included in region II), using a “cube-on-cube” methodology.40 In particular, a 54 × 54 (which corresponds to 54 cerium atoms or 27 CeO2 units for each side of the simulation cell) CeO2 thin film was placed directly above a 40 × 40 ZrO2(111) support, giving an interfacial area of 17849 Å2 and 65496 ions within the simulation cell. To ensure that the CeO2 fitted exactly on top of the substrate (coincidence), the CeO2 thin film was compressed (eq 3) by ca. 31%. Simulation trials on the CeO2/ ZrO2 system where the thin film was compressed by only 15%,27 revealed that the initial pressure was insufficient to generate an appropriate amorphous structure. Consequently, the energy of the system was deemed insufficient to facilitate adequate recrystallization of the thin film and the final structure demonstrated regions, which remained amorphous. The second step was to introduce the mean field ‘hybrid’ ions into region II of the support and to scale the entire simulation cell at the new lattice parameters from 5.075 Å corresponding to cubic ZrO2 to 5.106 Å corresponding to YSZ MF. Accordingly, the interfacial area increased to 18064 Å2. Finally, yttrium species were introduced into region I by replacing 10% of the Zr4+ ions with Y3+ (960 yttrium ions), corresponding to 9.24 mol. % Y2O3. 480 oxygen ions were then removed to maintain charge neutrality of the system. To facilitate such a procedure, an additional program was written, in which yttrium ions and oxygen vacancies were introduced at random into the zirconia lattice. The whole simulation cell comprises finally 65016 ions.
Figure 2. Calculated cerium and oxygen mean square displacements (MSD) measured in Å2 within the CeO2/YSZ(111) system as a function of time during dynamical simulation performed at 3400 K.
All simulations were performed within the NVE ensemble: constant Number of particles, constant Volume and constant Energy with instantaneous velocity scaling to the simulation temperature used throughout. This prevents the rapid and large build up of excess kinetic energy as the thin film evolves from the highly strained initial configuration, via an amorphous transition, to a crystalline phase with reduced strain and a range of defects. Dynamical simulation, with a time step of 5 × 10-3 ps, was applied to the system for 150 ps at 3400 K, 40 ps at 2500 K, 25 ps at 2000 K, 5 ps at 1500 K, 15 ps at 1000 K, 5 ps at 500 K and 85 ps at 0 K; the latter acts effectively as an energy minimization and was performed until the energy converged. For each temperature, the dynamical simulation was performed until the system was no longer evolving structurally or energetically. Results Amorphization and Recrystallization. The MSD of cerium and oxygen ions for dynamical simulation at 3400 K is presented in Figure 2. It can be seen that for cerium, the MSD increase rapidly at the beginning of the simulation during the amorphization process reflecting high mobility of the cerium ions. Conversely, after ca. 50 ps the MSD plateaus, indicating little diffusion once recrystallization nears completion. For the oxygen ions, similar behavior is observed at the beginning of the amorphization, whereas toward the end, the curve maintains a positive gradient indicating that the oxygen species retain significant mobility even after recrystallization. This is to be expected due to the high mobility of oxygen within CeO2 at high temperature. Calculated Ce-O RDF within the CeO2 thin film during dynamical simulation performed at 3400 K are shown in Figures 3(a)-(e) after 0.005, 0.25, 25, 150 ps and at the end of the simulation (final structure). Initially, the RDF peaks are sharp (Figure 3(a)) reflecting the high crystallinity of the initial structure; the calculated RDF is consistent with the fluorite structure and reflects the 31% compression, imposed on the thin film to initiate amorphization, via a (left) shift of the peaks. After 0.25 ps, Figure 3(b), the CeO2 thin film loses long-range order, indicating an amorphous transition. At 25 ps, Figure 3(c), the ceria thin film starts to regain long-range order, indicating recrystallization of the CeO2 thin film. At the end of the first stage of dynamical simulation (3400 K), after 150 ps, the thin film regains much of its crystallinity as depicted in Figure 3(d). However, it is clear that the structure has rearranged significantly as shown by a shift of the RDF peaks from 1.9, 3.2, and 4.1 Å to 2.2. 4.3, and 6.0 Å, respectively. The RDF for the final structure, Figure 3(e), shows sharp peaks indicating a highly
Characterization CeO2/YSZ(111) Catalytic System
J. Phys. Chem. B, Vol. 105, No. 50, 2001 12485
Figure 3. Calculated Ce-O radial distribution functions within the CeO2 thin film during dynamical simulation performed at 3400 K after: (a) 0.005 ps; (b) 0.25 ps; (c) 25 ps; (d) 150 ps and (e) at the end of the simulation. Interatomic separations are measured in Å.
crystalline structure obtained after cooling the system at 0 K. The peaks in Figure 3(e) are again consistent with the fluorite structure (as supported also by inspection of the final atom positions (Figure 4(c)) using graphical techniques). For a visual link to the thin film structure, we show in Figures 4(a)-(c) part of the simulation cell comprising the thin film and three layers of the support. Figure 4(a) represents the initial structure with the ceria thin film constrained under considerable pressure. As the dynamical simulation progress, the thin film expands along the surface normal in attempt to eliminate the huge strain present in the initial structure and causes the CeO2 thin film to go amorphous after ca. 0.25 ps, Figure 4(b). After prolonged dynamical simulation, the ceria thin film starts to recrystallize realizing, at the end of the simulation, a cubic fluorite structure, Figure 4(c). The success of the simulated amorphization and recrystallization methodology in generating the CeO2 structure from an amorphous solid, suggests that the methodology is applicable to study fluorite-structured systems in addition to the supported rocksalt systems considered previously.25,26
Inspection of the MSD, RDF, and structures for our system and comparison with those obtained by Sayle and Watson26 for the SrO/MgO(001) system suggests that the thin film proceeds from a highly strained but crystalline initial structure, through an amorphous (solid) transition and then a final crystalline form. We find no evidence that at any stage does the CeO2 become molten. We will consider now the structural modifications that evolve during the recrystallization process in the thin film and the support in the final structure of the CeO2/YSZ(111) system. Structural Characterization. The final CeO2 thin film exposes the (111) plane at both interface and surface and comprises ca. five CeO2 repeat units with an incomplete surface layer and a final thickness of approximately 20 Å. The surface layer has an occupancy of about 25% and comprises small clusters ranging from, for example, Ce2O4 and Ce4O8, to larger clusters up to ca. 600 Å2 in size (Figure 5). These “clusters”, as depicted in Figure 5, lie flat with respect to the underlying CeO2(111) thin film, resulting in an almost atomically flat surface layer.
12486 J. Phys. Chem. B, Vol. 105, No. 50, 2001
Maicaneanu et al.
Figure 6. Representation of the interfacial CeO2 and YSZ planes, illustrating the epitaxy between the two materials at the interface. Only the cation sublattices and yttrium dopants are shown for clarity. Color notation as Figure 4.
Figure 4. Representation of the CeO2/YSZ(111) interface during dynamical simulation performed at 3400 K: (a) starting structure; (b) after 0.25 ps of dynamical simulation and (c) final structure. Zirconium is colored light blue, cerium is magenta, yttrium is yellow, oxygen (YSZ) is red and oxygen (CeO2) is green. To preserve clarity of the figure only a small part of the full simulation cell is shown together with three unit cells of the underlying support.
Figure 7. Schematic of the final structure of the CeO2/YSZ interface, highlighting the epitaxial relationships at the interfacial region.
Figure 5. Ball and stick representation of the CeO2 surface layer (fifth layer from the interface) depicting (a) all clusters distributed over the surface; (b) an enlarged view of one particular cluster (far left of Figure 5(a)). In addition, part of the underlying CeO2 layer (fourth from the interface) is included as a stick model. Only the cerium sublattice is shown for reasons of clarity.
Epitaxial Relationships. Inspection of the epitaxial relationships that exist between the CeO2 thin film and the YSZ support, reveals that the thin film lies almost coherent (Figure 6) with the underlying YSZ support with the CeO2 accommodating a 37 × 38 pattern with no rotation of the thin film with respect to the support. The 37 × 38 refers to 37 atoms along the [1h 1 0] and 38 along the [1h 0 1], which are along the surface lattice vectors defining the simulation cell (Figure 7). The lattice misfit (based upon a 40 × 40 lattice for the underlying YSZ) associated with such a configuration is therefore reduced from +6.1% (bulk misfit) to -1.7% along the [1h 1 0] and +1.0% along [1h 0 1]. We suggest that the misfit is reduced further (via relaxation of the lattice) by the CeO2 being constrained under compression along the [1h 0 1] in conjunction with tension along the [1h 1 0]. Surprisingly, the interfacial YSZ layer of the support also changes and presents a 39 × 39 pattern in contrast to the initial 40 × 40 configuration of the preparatory configuration; the remaining YSZ below maintains a 40 × 40 configuration. Experimentally, Putna et al., who explored the structure and reducibility of ceria deposited on YSZ by vapor deposition,12
found that deposition on the YSZ(111) surface, resulted in the formation of semicoherent, epitaxial CeO2 films oriented in the direction of the support. In addition, Dmowski et al.41 found, for the same system (ca. 20 Å CeO2 thin film deposited on yttrium stabilized zirconia), that the YSZ lattice parameter was 5.23 Å. This is larger than the value obtained for bulk diffraction (5.13 Å) of YSZ with yttrium doping levels of ca. 9.5%. Moreover, the authors suggest that there is some inhomogeneity in the YSZ crystal, which results in a change in lattice parameter as a function of depth. In this present study, the calculated lattice parameter for the 39 × 39 YSZ interfacial layer is ca. 5.24 Å in accord with the experimental results of Dmowski et al. The values assigned to the misfit have been based purely upon geometrical considerations. However, the presence of defects in the system can help to reduce further any residual strain in the system. We now explore the structure of various defects observed within the CeO2 thin film and underlying YSZ support. Defects. A layer-by-layer analysis of the system, using graphical techniques, revealed that the CeO2 comprises ca. 0.8% cerium vacancies. However, no cerium vacancies were observed in the interfacial layer. Figure 8 shows the third CeO2 plane from the interface depicting 16 such cerium vacancies. For the underlying YSZ support, 0.2% zirconium vacancies were identified within the first three layers from the interfacial YSZ plane. Figure 9 shows zirconium vacancies within the second YSZ layer from the interfacial plane, together with the random distribution of yttrium ions. Clearly, the evolution of cerium and zirconium vacancies within the system disrupts the charge neutrality. Accordingly, oxygen vacancies are also present to restore charge balance. In addition, careful inspection of each
Characterization CeO2/YSZ(111) Catalytic System
J. Phys. Chem. B, Vol. 105, No. 50, 2001 12487
Figure 8. Ball and stick representation of the third layer in the ceria thin film showing the distribution of cerium vacancies. Only the cerium sublattice is shown. Figure 12. Ball and stick representation of the interfacial cerium sublattice depicting the dislocation network within the layer.
Figure 9. Ball and stick representation of the second YSZ layer from the interface highlighting the random distribution of yttrium ions (dark spheres) within the zirconium lattice (light spheres). The oxygen ions are not shown to maintain clarity of the figure.
Figure 10. Representation of a small segment of the cerium (grey spheres) interfacial layer illustrating the nature of the interdiffused cluster of zirconium ions (light spheres) and yttrium dopants (black spheres). Oxygen ions have been omitted to preserve clarity.
Figure 13. Enlarged view of neighboring dislocations depicting more clearly their core structure. Only the cerium sublattice is shown for clarity. Cerium is colored magenta and white (dislocation) and zirconium, light blue.
Figure 11. Ball and stick representation of the interfacial YSZ layer, showing the complex dislocation network. To preserve clarity of the figure only the cation sublattice of the YSZ interfacial plane is depicted.
layer revealed that the density of cerium vacancies within the CeO2 thin film increases within planes further from the interface. Analysis of the system also revealed interdiffusion of cerium, zirconium and yttrium across the interfacial planes. Cerium ions
were observed to occupy zirconium lattice sites within the YSZ support and zirconium and yttrium occupying both cerium lattice sites and interstitial positions within the CeO2 thin film. The interdiffusion is modest; ca. 3% of the interfacial cerium ions are displaced. Defect association as dimeric and trimeric cerium clusters was also observed. However, one particular cluster as shown in Figure 10, comprises, surprisingly, 23 zirconium and 3 yttrium ions, which may have evolved to help relieve high strain within this particular region. In addition, the interdiffusion was observed to occur across only the interfacial CeO2 and YSZ layers, which is in line with the experimental work of Lind et al.42 They investigated the Fe3O4/NiO system using oxygen-
12488 J. Phys. Chem. B, Vol. 105, No. 50, 2001
Maicaneanu et al.
Figure 14. Cross-section of the CeO2(111)/YSZ(111) interface: (a) across the entire simulation cell; (b) enlarged view of the dislocation far left in Figure 14(a) depicting more clearly the core structure. Only the cationic sublattice is shown for clarity. Cerium ions are magenta, zirconium ions are light blue and red (dislocation core region) and yttrium, yellow.
plasma-assisted molecular beam epitaxy and observed the interdiffusion across the interface, “of the order of 1 or 2 atomic layers”. Dislocations. Inspection of the interfacial YSZ and CeO2 layers (Figure 11 and Figure 12 respectively) revealed the presence of a network of edge dislocations, which lie parallel to the interfacial plane. In Figure 13 an enlarged segment of two neighboring dislocations is depicted, showing more clearly their core structure. Similar dislocation networks were observed, using transmission electron microscopy (TEM), in the GaAs/ InP system; in GaSb/GaAs system, a square network of dislocations was found.16 Figure 14 depicts a side view of the simulation cell, which shows the presence of a regular array of dislocations with calculated inter-dislocation separations of ca. 61 Å, 50, 47, and 65 Å. An enlarged segment of one such dislocation is shown in Figure 14(b) to illustrate more clearly the atomistic structure of the core region. In contrast to the dislocations identified above, the lines of the dislocations identified here do not lie parallel to the interfacial plane. The figure also demonstrates that the dislocation cores reside in regions of poor match, halfway between the regions of good match (or coherent regions). A plan view of the simulation cell (Figure 6) reiterates this argument. Across interfacial regions, which are closely lattice matched (coherent), there will be strong favorable interactions between the CeO2 and underlying YSZ. However, because of the difference in lattice constants between these two materials, the coherent regions will be small. This is because the strain energy terms, associated with bringing the two lattices into coherence across the whole of the interfacial plane, outweigh the energy
returned in maintaining such coherency. Epitaxial relationships can be identified at points from which the structures change from coherent to regions of poor match (and back to coherent). At the poorly matched regions, like-ions will come into close contact, destabilizing the structure. We suggest that the dislocation “cores” reside in these poorly matched regions where they help, in conjunction with ionic relaxation, to facilitate a reduction of like-ions in close proximity and additionally, reduce the distance between coherent regions. We also suggest that the evolution of dislocations, as identified in Figures 11 and 12, help contribute in maximizing regions of coherency. An investigation of the atomistic structure of the SrZrO3/ SrTiO3 interface, fabricated using metal-organic deposition of SrZrO3 layers on SrTiO3 single-crystal substrates, showed that the dislocation core resides in regions of poor match17 with a periodicity, between dislocation arrays, of ca. 60 Å. In addition, Wang et al. who explored the structure of the CeO2/YSZ(001) system using metal-organic vapor deposition43 observed dislocation arrays with a periodicity of ca. 44 Å. Here, we observe an average periodicity of ca. 53 Å. In a previous study, on the CeO2/ZrO2(111) system,44 both pure edge and mixed screw-edge dislocations were identified to have evolved within the supported CeO2 thin-film. However, after careful analysis, we found no evidence of mixed screwedge dislocations within either the supported CeO2 nor the underlying YSZ support for this present system. The epitaxial relationship between the CeO2 and underlying YSZ, enables the bulk lattice misfit and therefore the strain energy, associated with the system, to be reduced (based upon geometrical criterion) from ca. 6% to -1.7% along the [1h 1 0] and +1.0% along [1h 0 1]. However, this lattice misfit remains
Characterization CeO2/YSZ(111) Catalytic System significant because the strain energy terms are additive for each layer comprising the thin film. At some particular (critical) thickness,45 the strain energy terms will outweigh the energy associated with interfacial interaction and structural modifications, as we have identified above, will evolve. Conclusions Simulated amorphization and recrystallization, has been employed to observe the evolution of many structural modifications including dislocation networks, vacancies, interstitials and substitutions including defect clustering within the CeO2/YSZ system. The driving force to the evolution of such structural features is the reduction in lattice misfit and hence increased stability associated with the system. In addition, the detailed relaxation will have relieved further the lattice misfit thereby stabilizing the thin film; an important role of energy minimization is to maximize interfacial interactions across the interfacial planes. One particular feature of this methodology is that it accommodates the synergy of interaction that exists between, for example neighboring dislocations. Specifically, each dislocation will have evolved structurally in response to interactions between neighboring dislocations and defects. Moreover, the structural features identified here have been determined with atomistic detail that is not yet possible using experimentation thereby demonstrating its value as a complimentary tool. It is likely that the considerable structural modifications identified here will result in a change of the material properties of the CeO2. For example, there are many cerium and oxygen ions on the surface of the CeO2 thin film with reduced coordinative saturation, which may influence their lability or alternatively the lability of neighboring ions. For example, some oxygen ions may be more weakly bound to the surface of the CeO2 thin film compared with the analogous unsupported CeO2 enabling it to more readily leave the surface and participate in catalysing some particular chemical reaction. In addition, the CeO2 is populated with cerium and oxygen vacancies together with zirconium species occupying interstitial and cerium lattice sites. Such defects were found to increase the OSC of the material1,4,5 although it should be noted that the presence of Ce3+, not accounted for in this study, may affect the defects to some extent. Future studies will focus on addressing the binding energies of such surface oxygen ions and the formation of Ce3+ together with associated oxygen vacancies to ensure charge neutrality, which is required for oxygen storage and release, using atomistic and quantum mechanical techniques. In addition, we intend to explore the migration of oxygen within the supported CeO2 lattice. Although many simulation studies of oxygen lability32 and migration4,33 have been explored previously within CeO2, the mobility is likely to be influenced (as it will within a real system) by the strain within the CeO2 associated with accommodating the lattice misfit, dislocation networks and the range of defects observed in this present model. References and Notes (1) Trovarelli, A. Catal. ReV. Sci. Eng. 1996, 38, 439. (2) Mogensen, M.; Sammes, N. M.; Tompsett, G. A. Solid State Ionics 2000, 129, 63. (3) Trovarelli, A.; de Leitenburg, C.; Boaro, M.; Dolcetti, G. Catal. Today 1999, 50, 353. (4) Balducci, G.; Kasˇpar, J.; Fornasiero, P.; Graziani, M.; Islam, M. S.; Gale, J. D. J. Phys. Chem. B 1997, 101, 1750.
J. Phys. Chem. B, Vol. 105, No. 50, 2001 12489 (5) Kasˇpar, J.; Fornasiero, P.; Graziani, M. Catal. Today 1999, 50, 285. (6) Sutton, A. P.; Balluffi, R. W. Interfaces in Crystalline Materials, Monographs on the Physics and Chemistry of Materials; Oxford University Press Inc.: New York, 1995; vol. 51. (7) Ramesh, R.; Hwang, D.; Ravi, T. S.; Inam, A.; Barner, J. B.; Nazar, L.; Chan, S. W.; Chen, C. Y.; Dutta, B.; Venkatesan, T.; Wu, X. D. Appl. Phys. Lett. 1990, 56, 2243. (8) Watson, G. W.; Olivier, P. M.; Parker, S. C. Surface Science 2001, 474, L185. (9) Sayle, D. C.; Sayle, T. X. T.; Parker, S. C.; Harding, J. H.; Catlow, C. R. A. Surf. Sci. 1995, 334, 170. (10) Putna, E. S.; Vohs, J. M.; Gorte, R. J. J. Phys. Chem. 1996, 100, 17 862. (11) Putna, E. S.; Vohs, J. M.; Gorte, R. J. Catal. Lett. 1997, 45, 143. (12) Putna, E. S.; Bunluesin, T.; Fan, X. L.; Gorte, R. J.; Vosh, J. M.; Lakis, R. E.; Egami, T. Catal. Today 1999, 50, 343. (13) Schwarz, J. A.; Contescu, C.; Contescu, A. Chem. ReV. 1995, 95, 477. (14) Somorjai, G. A. Introduction to Surface Chemistry and Catalysis; John Wiley & Sons: New York, 1994. (15) Cordatos, H.; Bunluesin, T.; Stubenrauch, J.; Vohs, J. M.; Gorte, R. J. J. Phys. Chem. 1996, 100, 785. (16) Rocher, A.; Snoeck, E. Mater. Sci. Eng. B 1999, 67, 62. (17) Ernst, F.; Recˇnik, A.; Langjahr, P. A.; Nellist, P. D.; Ru¨hle, M. Acta Mater. 1999, 47, 183. (18) Koster, A. J.; Ziese, U.; Verkleij, A. J.; Janssen, A. H.; de Jong, K. P. J. Phys. Chem. B 2000, 104, 9368. (19) Brinkman, H. W.; Briels, W. J.; Verweij, H. Chem. Phys. Lett. 1995, 247, 386. (20) Yamamura, Y.; Kawasaki, S.; Sakai, H. Solid State Ionics 1999, 126, 181. (21) Li, P.; Chen, I.-W.; Penner-Hahn, J. E. J. Am. Ceram. Soc. 1994, 77, 118. (22) Sutton, A. P.; Balluffui, R. W. Acta Metall. 1987, 35, 2177. (23) Sayle, T. X. T.; Catlow, C. R. A.; Sayle, D. C.; Parker, S. C.; Harding, J. H. Philos. Mag. A 1993, 68, 565. (24) Schnitker, J.; Srolovitz, D. J. Model. Simul. Mater. Sci. Eng. 1998, 6, 153. (25) Sayle, D. C. J. Mater. Chem. 1999, 9, 2961. (26) Sayle, D. C.; Watson, G. W. Surf. Sci. 2001, 473, 97. (27) Sayle, D. C.; Catlow, C. R. A.; Dulamita, N.; Healy, M. J. F.; Maicaneanu, S. A.; Slater, B.; Watson, G. W. Accepted (Molecular Simulation). (28) Sayle, D. C.; Watson, G. W. J. Mater. Chem. 2000, 10, 2241. (29) Sayle, D. C.; Watson, G. W. Phys. Chem. Chem. Phys. 2000, 2, 5491. (30) Lewis, G. V.; Catlow, C. R. A. J. Phys. C: Solid State Phys. 1985, 18, 1149. (31) Dwivedi, A.; Cormack, A. N. Philos. Mag. A 1990, 61, 1. (32) Sayle, T. X. T.; Parker, S. C.; Catlow, C. R. A. Surf. Sci. 1994, 316, 329. (33) Balducci, G.; Islam, M. S.; Kasˇpar, J.; Fornasiero, P.; Graziani, M. Chem. Mater. 2000, 12, 677. (34) Khan, M. S.; Islam, M. S.; Bates, D. R. J. Mater. Chem. 1998, 8, 2299. (35) Morris, B. C.; Flavell, W. R.; Mackrodt, W. C.; Morris, M. A. J. Mater. Chem. 1993, 3, 1007. (36) Watson, G. W.; Kelsey, E. T.; de Leeuw, N. H.; Harris, D. J.; Parker, S. C. J. Chem. Soc., Faraday Trams. 1996, 92, 433. (37) Balducci, G.; Kasˇpar, J.; Fornasiero, P.; Graziani, M.; Islam, M. S. J. Phys. Chem. B 1998, 102, 557. (38) Gale, J. D. J. Chem. Soc., Faraday Trans. 1997, 93, 629. (39) Smith, W.; Forester, T. R. The DL_POLY Molecular Simulation Package, URL: http://www.dl.ac.uk/TCSC/Software/DL_POLY. (40) Sayle, D. C.; Catlow, C. R. A.; Harding, J. H.; Healy, M. J. F.; Maicaneanu, S. A.; Parker, S. C.; Slater, B.; Watson, G. W. J. Mater. Chem. 2000, 10, 1315. (41) Dmowski, W.; Mamontov, E.; Egami, T.; Putna, S.; Gorte, R. Physica B 1998, 248, 95. (42) Lind, D. M.; Berry, S. D.; Chern, G.; Mathias, H.; Testardi, L. R. Phys. ReV. B 1992, 45, 1838. (43) Wang, A.; Belot, J. A.; Marks, T. J.; Markworth, P. R.; Chang, R. P. H.; Chudzik, M. P.; Kannewurf, C. R. Physica C 1999, 320, 154. (44) Maicaneanu, S. A.; Sayle, D. C.; Watson, G. W. Chem. Commun. 2001, 289. (45) Dong, L.; Schnitker, J.; Smith, R. W.; Srolovitz, D. J. J. Appl. Phys. 1998, 83, 217.
