|
|
|
Model Studies of Automobile Exhaust Catalysis Using Single Crystals of Rhodium and Ceria/ZirconiaC. H. F. Peden, G. S. Herman, and D. N. Belton(a) Supported by the U.S. Department of Energy (DOE), Office of Basic Energy Sciences, Division of Chemical Sciences. The 1990 Clean Air Act Amendment mandates that automobile manufacturers produce cars with significantly lower hydrocarbon (HC), carbon monoxide (CO), and nitrogen oxide (NOx) emissions to be phased in over the next 10 years (Taylor 1990, 1993; Calvert et al. 1993). In recognition of the serious pressures on the current emission control technology described above, the recent DOE-cosponsored workshop, Basic Research Needs for Vehicles of the Future (Eisenberger 1995), identified a critical need for greatly improved catalyst activity and durability. To effect these dramatic improvements that are so crucial for meeting the very stringent future regulatory requirements, the need for detailed chemical kinetics information on realistic and model catalyst materials, directly coupled with extensive catalyst characterization cannot be overemphasized. Indeed, more specific recommendations from the Vehicles of the Future workshop include the need for fundamental advances in the modeling of transient NOx kinetics, and a determination of the mechanisms of catalyst deactivation in the current generation of catalyst materials. Dramatic improvements have been made in automobile exhaust converter catalysts by the incorporation of oxygen storage (OS) materials, usually consisting of ceria or modified ceria, that can effectively dampen deviation in the exhaust air/fuel (A/F) ratio bringing the gas phase closer to the stoichiometric point. Although there is some limited understanding of the activity and durability of these 1990-vintage catalysts, virtually no fundamental understanding of this type exists for the current generation of catalysts that incorporate ceria-zirconia solid solutions as the OS material. In the work to be described here, our overall objective is to obtain detailed chemical kinetics data on idealized but well-characterized catalyst systems useful for understanding the important elementary catalytic converter reactions, and how they are affected by the oxygen uptake, storage, and release processes of the OS material. The catalytic activity measurements were performed in a custom-built system that combines a UHV analysis chamber connected to a high pressure (<760 Torr) reactor (Herman et al. in press). The UHV analysis chamber is equipped with several surface analytical techniques. For this study we used x-ray photoelectron spectroscopy (XPS), Auger electron spectroscopy (AES), and low energy electron diffraction (LEED). The sample was mounted to a retractable transfer rod that can be moved between the reactor and analysis chamber when both are under UHV conditions. Elsewhere (Herman et al. in press) we describe the mounting and preparation of the single crystal Rh samples, and details about the methods used for CO+NO reaction rate measurements over these model catalysts. Epitaxial growth and in-situ characterization of Ce1-xZrxO2 films were carried out in a dual-chamber MBE system specially designed for growth of oxide thin films and described in detail elsewhere (Chambers et al. 1994). Polished, pure and 0.15 wt.% Nb-doped SrTiO3 (001) single crystals were used as substrates for epitaxial growth of Ce1-xZrxO2 solid solutions as the oxygen sublattices of these two crystals are very well matched. Post-growth compositional and structural measurements included XPS and LEED. The film composition was also determined ex situ by Rutherford backscattering spectrometry (RBS). The film composition determined by XPS and RBS agreed with that measured by a quartz crystal oscillator (QCO) deposition rate monitor during growth to within a few percent. In addition, the Ce1-xZrxO2 epitaxial films were characterized ex situ by x-ray diffraction (XRD), ion channeling, and atomic force microscopy (AFM). The combination of these probes allows us to determine film structure and composition, oxidation states of both Ce and Zr, the crystal quality, surface morphology, and Zr incorporation in the CeO2 lattice. Studies of the surface chemistry of these films were performed in a separate ultrahigh vacuum (UHV) system that contains capabilities for XPS, AES, temperature programmed desorption (TPD), LEED, and a sputter gun for sample cleaning. This system also has a sample holder that allowed sample temperatures to be varied from 80 K to 1600 K by liquid nitrogen cooling and resistive heating. Angle-resolved mass-spectroscopy of recoiled ions (AR-MSRI) measurements were performed in a separate, custom UHV chamber described elsewhere (Herman 1999a, 1999b), where we also discuss sample mounting and temperature measurement. Based on the position of the water multilayer desorption peak, the TPD peak temperatures reported here should be close to the actual sample temperature.
1. Characterization and Reactivity of Single Crystal Rh CatalystsIn collaboration with S. J. Schmeig (General Motors Research and Development Center). A direct comparison of the NO consumption rate over the Rh(100), Rh(111) and Rh(110) surfaces represents the best relative measure of overall activity of the NO-CO reaction since it does not depend on the selectivity of the reaction (Herman et al. in press). While the Rh(111) and Rh(110) data, plotted in Arrhenius fashion, can be described by a single straight line with Ea values of 34.8 and 27.2 kcal/mol, respectively, the Rh(100) data has a low (LT) and high (HT) temperature region with Ea values of 35.3 and 20.5 kcal/mol, respectively. In Figure 7.7, the temperature sensitivity of the N2O selectivity for NO-CO reaction over the same three single crystal Rh surfaces is plotted. In this case, the N2O selectivity is defined as S(N2O) = 100 x [moles N2O/(moles N2O + moles N2)]. These results can be understood in terms of the relative surface coverages of adsorbed NO and N-atoms on the three surfaces. Notably, catalyst surfaces with higher steady-state N-atom coverages favor N-atom recombination (N2 formation) more than the NO+N reaction (N2O formation). Figure 7.8 shows post-reaction XPS of the three different surfaces where it is clear that the three different surfaces have different relative concentrations of N(ad) and NO(ad). The Rh(111) surface consists of only NO(ad), while the Rh(100) and Rh(110) surfaces contain two peaks that can be assigned to NO(ad) and N(ad) with binding energies of 400.3 and 397.6 eV, respectively. The relative intensity of the N(ad) feature increases with decreasing rhodium surface atomic density (surface atomic density: Rh(111) > Rh(100) > Rh(110)).
To better understand the effect of surface structure on reaction kinetics, it is useful to discuss the reaction mechanism that we favor for the NO-CO reaction, which we take to be the simplest of the proposed NO-CO reaction mechanisms (Belton et al. 1995). The reaction can be broken down into six elementary steps where S represents a site for adsorption on the surface: 1. CO(g) + S <=> CO(ad) The adsorption or dissociation of the reactants is intimately related to the availability of sites at the surface (i.e., reactions 1, 2, and 3). In this model, the relative rates of reactions 5 and 6 dictate the relative N2O selectivity, and are the only two pathways for the removal of N(ad) from the surface. Reactions 5 and 6 are dependent on the relative surface coverages of NO and N. The only available source of N(ad) is from reaction 3, which is dependent on both the presence of NO(ad) and binding sites. The formation of CO2 (reaction 4) requires the presence of O(ad) from reaction 3 and CO(ad) from reaction 1. Modeling (Zhdanov and Kasemo 1997) and FTIR (Permana et al. 1995) studies for Rh(111) surfaces at moderate NO and CO pressures (8 Torr) have indicated that at temperatures below ~650 K the surface is predominately covered by adsorbed NO. Above 650 K, the amount of NO on the surface is reduced, and modeling indicates that the relative concentration of N(ad) and surface sites increase substantially. Although these calculations and measurements have not been carried out for either the Rh(100) or Rh(110) surfaces, some generalizations related to the rate limiting steps may be obtained. In our prior study, we proposed that the differences observed between the activity towards NO consumption for the Rh(111) and Rh(110) surfaces is related to the manner in which adsorbed NO inhibits the dissociation reaction (Peden et al. 1995). At low temperatures (~528 K) and moderate NO and CO pressures (~8 Torr), the three different surfaces do have increasing concentrations of adsorbed N relative to NO based on post reaction XPS (e.g., Figure 7.8), which are directly related to the surface atomic density in the top layer [Rh(110)<Rh(100)<Rh(111)]. The results that were obtained for the Rh(100) surface are fully consistent with our prior model in which adsorbed NO inhibits the dissociation reaction and the overall NO consumption kinetics. To understand the origin of the change in the Ea for the reaction products for the Rh(100) surface, we compare to other studies that have observed a similar change in their data. For polycrystalline films of Rh, a similar break, at ~1.7 x 10-3 K-1, in the Arrhenius plot for the NO-CO reaction has been observed (Tolia et al. 1995). Surface-enhanced Raman spectroscopy (SERS) was used in this study and it was observed that a vibration feature at 315 cm-1, assigned to the Rh-N stretch of N(ad), is significantly reduced in intensity in the temperature range of 573 to 623 K (i.e., where the break occurs) (Tolia et al. 1995). The authors suggested that the reduction of N(ad) at the surface at higher temperatures results in a change in the apparent activation energy. The difference in the likely surface structure and reaction conditions between these two studies makes direct comparison difficult. However, the similarity between these two different studies suggest that the concentration of N(ad) may have a strong impact on the NO-CO at low temperatures. Likewise, it was noted that N2 formation from N(ad) recombination occurs at higher temperatures on the Rh(100) surface (770 K), compared with the Rh(111) (660 K) and Rh(110) (580 K) surfaces (Peden et al. 1988). These differences may have implications that are related to the change in Ea for Rh(100). To investigate the concentration of N(ad) under different reaction conditions, we have performed post reaction XPS, shown in Figure 7.9, for the Rh(100) surface at 528, 635, and 700 K. The 528 K value corresponds to a reaction in the LT regime as defined earlier, and the 635 and 700 K values correspond to reactions in the HT regime. It is clear from this data that as the temperature increases the relative intensity of the N(ad) component in the XPS is reduced. This data appears to suggest that the concentration of N(ad) may be related to the changes observed in the apparent activation energies in these two temperature regimes. This may be due in part to the increased N(ad) concentration resulting in blocked sites for NO(ad) dissociation in the LT regime. Once the sample temperature reaches the HT regime, the concentration of N(ad) is reduced and the number of available vacant sites increases. It is interesting to note that the N2O selectivity in the LT region (Figure 7.7) is nearly constant, and once the sample reaches the HT region the selectivity rapidly decreases. It appears that the reduction in N(ad) and increase in available sites in the HT region results in more facile dissociation of the adsorbed NO. Furthermore, the increased rate for NO dissociation in the HT regime results in an increase in the recombination of N(ad) to form N2, hence the reduction in N2O selectivity. The low concentration of N(ad) and increase in sites at 635 K can also help explain why with increasing NO partial pressure the N2O selectivity increases. This is likely due to the decrease of NO(ad) at the surface due to stearic hindrance from other NO molecules. The high NO(ad) and low N(ad) coverages suggest that NO(ad) dissociation will result in an almost immediate reaction to form N2O. The reason that the N2O selectivity does not increase with increasing temperature in Figure 7.7 is due to the reduced amount of NO(ad) available with increasing temperature but constant NO partial pressure.
2. Preparation and Characterization of Model Ceria Thin FilmsIn collaboration with Y. J. Kim,(a) Y. Gao, G. S. Herman, S. Thevuthasan, W. Jiang,(b) D. E. McCready, and S. A. Chambers Supported by the US DOE, Office of Basic Energy Sciences, Division of Chemical Sciences. The growth of epitaxial thin films of insulating materials is currently of significant interest and relevance in several areas of technology. One example is the growth of high-dielectric materials on semiconductors. Ceria (CeO2) is an attractive insulator since it has a large band gap (6.0 eV), a high dielectric constant (26), and is very stable at high temperatures. One potential application for CeO2 is as an insulating layer in silicon-on insulator (SOI) devices, such as the storage capacitors in dynamic random access memory devices. As a second example, ceria is often used as a buffer layer between high temperature superconducting oxide films. The heteroepitaxy of these materials is thus of interest in order to preserve the structural quality of the superconducting oxide layers. In addition, ceria is being used in catalytic reactions as a reducible oxide support material in emission control catalysis. In three-way automotive catalysts, it is believed that the reducibility of ceria contributes to oxygen storage/release capability, which plays an important role in the oxidation of CO and hydrocarbons catalyzed at the surface of Pt particles. Knowledge of the surface structure and reactivity of ceria is very important for understanding the role of this material in automotive catalysis. Model crystalline surfaces would be of considerable value in attempting to gain such knowledge. Unfortunately, like many oxides, ceria single-crystals of sufficient quality for surface science investigations are not readily available. Therefore, epitaxial growth of such materials is of interest as a way to generate the needed surfaces. However, most of the reported epitaxial growth of ceria thin films is focused on the understanding of the interface formed between CeO2 and Si and the associated device applications. The surface structure and crystalline quality of CeO2 epitaxial films has attracted little attention. The immediate goal of this work has been to synthesize well-ordered single-crystalline ceria thin films in order to perform surface science measurements. Our ultimate scientific goal is to better understand the role of ceria for the CO oxidation and NOx reduction. Figure 7.10 shows the crystal structures of CeO2(001) and SrTiO3(001) in plan view. CeO2 has a cubic fluorite structure whereas SrTiO3 exhibits a cubic perovskite structure. The lattice parameters are 5.411 Å and 3.905 Å for CeO2 and SrTiO3, respectively. It is now generally accepted that the lattice mismatch and crystal symmetry of the oxygen sublattices are more important than that of the cation sublattices for growing high-quality epitaxial oxide films on oxide substrates. The oxygen sublattices are very well matched for these two materials provided one cube is rotated 45° about [001] with respect to the other. The oxygen sublattice constants of 2.705 Å along [100] in CeO2 and 2.761 Å along [110] in SrTiO3 [110] result in a lattice mismatch of 2.0%, after performing a 45° rotation about [001]. We thus expect that high-quality epitaxial CeO2 film growth should be possible on SrTiO3(001) under the appropriate conditions.
Experimentally a Ce metal beam from an electron beam evaporator was directed at the substrate along with a beam of activated oxygen created by the ECR plasma source. The substrate temperature, Ce deposition rate, oxygen partial pressure, and ECR plasma power level were systematically varied and the resulting films were characterized to obtain the best crystalline CeO2 films possible. Successful synthesis of pure epitaxial CeO2 was found to be critically dependent on the choice of substrates and relatively less sensitive to the growth parameters. The total film thickness varied from ~100 Å to ~2000 Å from sample to sample. For well-ordered CeO2 films on SrTiO3(001), we used a ~0.15 Å/sec Ce deposition rate in an oxygen pressure of ~5x10-6 Torr with the ECR plasma source running at 200 W total power. Substrate temperatures of 550°C, 650°C, and 700°C were used, and the total film thickness was 500 Å in all cases. Films grown at all three temperatures consisted of stoichiometric CeO2, as judged by the XPS Ce 3d3/2 lineshape. Films grown at temperatures above 650°C showed a very small amount of Ti outdiffusion, with a Ti/Ce of at most 1 at. % as determined by XPS quantitative analysis. We show in Figure 7.11 XRD q-2q scans for 500 Å CeO2 epitaxial films grown on SrTiO3(001) grown at substrate temperatures of 550°C, 650°C, and 700°C. The peaks at 22.8° and 46.5° are reflections associated with SrTiO3(001) and (002), respectively. An intense CeO2(002) reflection is present at 33.1° in all three scans, indicating that CeO2(001) is the majority phase, independent of substrate temperature. The film grown at 550°C is pure CeO2(001), with no minority orientations detected. Two very weak peaks appear at 28.6° and 47.5° for the film grown at 650°C. These two peaks can be indexed as CeO2 (111) and CeO2 (220), respectively, indicating the existence of small amounts of (111) and (220)-oriented crystallites at this growth temperature. In addition, there is a weak feature at 30°, indicating the presence of a small amount of amorphous-like material in the film grown at 650°C. The film grown at 700°C shows a CeO2(111) reflection that is considerably weaker than that seen in the film grown at 650°C, but no CeO2(220) reflection is present.
We show in Figure 7.12 RHEED patterns obtained for a typical SrTiO3(001) substrate (a), and CeO2 epitaxial films at growth temperatures of 550°C (b), 650°C (c), and 700°C (d). The primary beam alignment along [100] is shown, and the primary beam voltage was 15 kV. The substrate pattern is of good quality, showing streaks rather than arrays of spots, indicative of reasonably flat surfaces with large terrace widths. In contrast, the diffraction pattern is dominated by transmission, as indicated by arrays of diffraction spots, when the film is grown at 550°C. However, the pattern becomes more streaked as the substrate temperature is increased, indicating a transition from rougher to smoother surface morphology. The RHEED patterns for the film grown at 650°C (Figure 7.12c) are a mixture of transmission and reflection (i.e., spots and streaks), whereas those for the film grown at 700°C (Figure 7.12d) are essentially pure reflection (i.e., streaks). Moreover, the streaks are sharper at 700°C than at 650°C, indicating superior surface crystallographic quality at the higher substrate temperature. The streak spacings in the zeroth-order Laue zone are smaller by the expected ratio of lattice parameters in the SrTiO3(001) pattern relative to the CeO2(001) pattern, indicating that the films have achieved the bulk lattice spacing of CeO2 shown in Figure 7.10. In addition, there is no evidence for surface reconstruction, which is consistent with LEED and XPD results to be discussed later. The RHEED patterns reveal the expected epitaxial orientation relationship of CeO2[100] || SrTiO3[110].
The epitaxial growth of CeO2 films has been investigated over a wide range of growth parameters by oxygen-plasma-assisted MBE on three different substrates. The resulting films and surfaces were thoroughly characterized by a number of techniques in order to determine their structures and compositions. Pure-phase epitaxial CeO2(001) readily grew on SrTiO3(001) for substrate temperatures ranging from 550° C to 700° C. The resulting epitaxial film surfaces are unreconstructed and exhibit the structure of bulk CeO2(001) with an oxygen termination. In general, the film surface becomes smoother and the overall crystalline quality of the film improves as the growth temperature increases due to thermally driven island coalescence. The use of Si(111) and MgO(001) as substrates results in poorer crystalline quality and surface morphology compared to the use of SrTiO3(001). The poorer ordering of CeO2 on Si(111) and the rougher film surfaces on MgO(001) are explained by the presence of amorphous silicon oxides at interface and a large lattice mismatch, respectively. The ability to synthesize CeO2 films as pure, single-crystalline phase paves the way for determining the surface chemistry of this oxide with surfaces that are very well defined compositionally and structurally.
3. Preparation and Characterization of Model Ceria/Zirconia Thin FilmsIn collaboration with Y. Gao and S. Thevuthasan We have synthesized epitaxial thin films of CeO2(001) and Ce1-xZrxO2(001) on SrTiO3(001) substrates. CeO2 has a cubic fluorite structure whereas SrTiO3 exhibits a cubic perovskite structure. The lattice parameters are 5.411 Å and 3.905 Å for CeO2 and SrTiO3, respectively. The oxygen sublattices of these two materials are very well matched (~2% mismatch) when one cube is rotated 45° about [001] with respect to the other (Figure 7.13). The ionic radius of Zr4+ (0.81 Å) is smaller than that of Ce4+ (0.97 Å), leading to a smaller lattice constant (a = 5.14 Å). When a Ce4+ ion is replaced by a Zr4+ ion in the fluorite lattice, the mismatch of the metal-oxygen bond lengths will result in lattice distortion, which determines how much Zr can be substitutionally incorporated into the CeO2 lattice without changing the fluorite structure. The maximum amount of Zr in bulk CeO2 is about 40 at. % (x = 0.4), at which the Ce1-xZrxO2 material exhibits a cubic-tetragonal structural phase transition. The oxygen sublattice mismatch between Ce0.6Zr0.4O2 and SrTiO3 is still relatively small, ~4%.
Epitaxial Ce1-xZrxO2 FilmsEpitaxial Ce1-xZrxO2 films were deposited at 600°C to prevent any Ti out-diffusion from SrTiO3 substrates. RHEED patterns showed streaks during the first ~7 Å, indicating layer-by-layer growth, and then started to show transmission diffraction spots as a result of surface roughening. The RHEED patterns became completely transmission-like at ~30 Å. These observations suggest Stranski-Krastanov growth (layer-by-layer growth followed by three-dimensional islands) for growth of Ce1-xZrxO2 films on SrTiO3(100). The 3-D RHEED patterns remained unchanged during the rest of film growth (~320-470 Å). After growth, the temperature was raised to 700°C within 5 seconds. The transmission diffraction spots completely disappeared after the films were held at 700° C for about 20 seconds. The reappearance of streaks and associated disappearance of transmission diffraction spots were the result of surface diffusion, leading to an atomically flat surface. Annealing at 700°C for a longer period of time appeared to improve the crystalline quality near the surface region, as indicated by sharpening of the RHEED streaks. However, such long-time annealing also resulted in Ti out-diffusion from the SrTiO3 substrates, as revealed by XPS. The structure of the Ce1-xZrxO2 films was determined by x-ray diffraction (Figure 7.14). In addition to the substrate peaks (22.8° and 46.5°), only a single diffraction peak from the films is observed at about 33.3°, corresponding to the (002) plane of cubic Ce1-xZrxO2. The 2q value increases with increasing Zr content in the Ce1-xZrxO2 films, resulting in reduction of the lattice constant (the inset of Figure 7.14). The reduction in the lattice constant is presumably due to both the substrate effect (e.g., lattice and thermal mismatch) and the incorporation of smaller Zr+4 cations in the CeO2 lattice. For example, a smaller lattice constant was observed even for pure CeO2 films compared to the bulk value. The lattice constant (5.387 Å) of the pure CeO2 film determined by XRD in Figure 7.15 is slightly smaller than that of bulk CeO2 (5.411 Å), indicating that the pure CeO2 film is under in-plane tensile stress due to the lattice mismatch between CeO2 and SrTiO3 because the oxygen sublattice of SrTiO3 is ~2% larger than that of CeO2 (see Figure 7.13).
The crystalline quality of the Ce1-xZrxO2 films in the growth direction was determined using x-ray q rocking curves and ion-beam channeling. The rocking curves of the (002) diffraction peak (the growth plane) reveal a full width at half maximum (FWHM) of about 0.2°, 0.2°, 0.3°, and 0.45° for the Ce1-xZrxO2 films with x = 0, 0.1, 0.2, and 0.3, respectively. The value for the pure CeO2 films is similar to that for the Ce0.9Zr0.1O2 films, suggesting that there is no significant lattice distortion for 10% Zr doping. However, a slight increase in the FWHM value (~0.3°) for the Ce0.8Zr0.2O2 films indicates that 20% Zr doping leads to notable distortion of the lattice planes. Relatively large lattice distortion was observed for the 30% Zr doped CeO2 films, which exhibit a FWHM of about 0.45°. This is not unexpected because the lattice distortion induced by ~40 at.% Zr doping is large enough to result in a phase transition from cubic to tetragonal structure. The minimum yields in ion channeling for a Ce0.9Zr0.1O2 epitaxial film were 23.0 ± 0.3% for Zr, and 17.6 ± 0.2% for Ce, compared to 15.2 ± 0.3% for substrate Sr. The small values for Zr and Ce indicate a high degree of atomic alignment in the film growth direction. Based on these values, the fraction of Zr substitution for Ce cations in the film was calculated to be about 93%. In other words, about 93% of the Zr atoms substitutionally occupy the cation sites in the CeO2 lattice. The normalized film Zr and Ce angular yield curves indicate that the Zr atoms are slightly displaced from the Ce atomic positions. The Zr displacement from the Ce atomic position is most likely due to the difference between Ce-O and Zr-O bond lengths. The variation of the yield as a function of the polar angle for Zr and Ce in the film is similar to Sr in the substrate, which indicates that the cation atomic rows in the film are parallel to the cation atomic rows in the substrate.
The oxidation state of Zr in the Ce1-xZrxO2 epitaxial films was determined by the binding energy of the Zr 3d5/2 core level (Figure 7.15a). Core level binding energies listed in Table 7.2 include Ce 3d3/2, O 1s, and Zr 3d5/2, along with the Zr 3d5/2 and O 1s binding energies for a ZrO2 on Zr for comparison. The unique shape of the Ce 3d spectra in Figure 7.15b is typical for Ce+4. However, the Ce 3d spectrum for the Ce0.7Zr0.3O2 film shows considerable increase in intensity at the binding energies at v’ and u’, as compared to those for pure CeO2 and Ce1-xZrxO2 with x £ 0.2. These two peaks are typical for Ce+3, suggesting that mixing a large amount of Zr in the CeO2 lattice resulted in formation of Ce+3. Though formation of Ce+3 is still not yet understood, one possible explanation is the formation of oxygen vacancies in the surface region due to oxygen displacements along the c axis from the ideal fluorite position caused by incorporation of Zr in the CeO2 lattice. The Zr 3d5/2 binding energy decreases from 182.69 eV for Ce0.9Zr0.1O2 to 182.45 eV for Ce0.7Zr0.3O2, as shown in Table 7.2. These measured Zr 3d5/2 values for the Ce1-xZrxO2 films are slightly higher than the literature value of 182.2 eV for pure ZrO2. The higher Zr 3d5/2 values for the Ce1-xZrxO2 films are presumably due to the "alloy" effect. Thus, the XPS results revealed a +4 oxidation state for both Ce and Zr in the Ce1-xZrxO2 films.
4. Structure of Model Ceria Thin FilmsIn collaboration with Y. J. Kim(a) and S. A. Chambers Supported by the US DOE, Office of Basic Energy Sciences, Division of Chemical Sciences. Low energy ion scattering (LEIS) and direct recoil spectroscopy (DRS) are powerful techniques for the determination of the surface composition and structure. The kinetic energies of the scattered and recoiled particles are usually on the order of several hundred to several thousand electron volts (eV), resulting in the detection of only elemental species (i.e., no molecular fragments). The energies of these scattered and recoiled ions can be described by sequences of two-body elastic scattering events that depend on the energy and mass of the projectile, the mass of the target atom, and the scattering or recoiling angles. An inherent problem with these techniques is that the energy and mass resolution is strongly dependent on the mass ratio of the target and projectile atoms. To circumvent this limitation, mass-spectroscopy of recoiled ions (MSRI) was developed in which time-of-flight methods and a time-focusing electrostatic analyzer are used to separate the scattered and recoiled particles resulting in an isotopically resolved mass spectrum. The mass resolution for the MSRI technique is strongly influenced by the experimental setup. Attempts have been made in the past to use electrostatic analysis of the recoiled and scattered particles, but poor mass resolution was obtained. However, recent developments show that isotopic resolution for the entire periodic table can be realized by reducing the ion source pulse width (<30 nsec) and using a custom reflectron-based analyzer. Although there have been several advances in the MSRI technique, there has been no prior MSRI work performed in the area of surface structure determination. However, the high-mass resolution and the inherent surface sensitivity available with the MSRI technique provides a unique and powerful new capability for surface structure measurements. In this study, AR-MSRI was used to examine the structure of the CeO2(001) surface. This surface has several potentially important applications including buffer layers for the growth of high Tc superconductors on substrates with large lattice mismatches and reactive interfaces, dielectric layers on silicon-on-insulator structures, and as a model surface for the characterization of chemistry occurring in automotive three-way exhaust catalysts. Knowing the surface structure of CeO2(001) should allow for a better understanding of the epitaxial relationship between the substrate and the overlayer, and the observed surface chemistry. Little surface structural work of any type has been performed on this surface due in part to the large band gap (6 eV) and the limited availability of high-quality crystals. However, the results that exist do not agree on the surface termination or atomic structure. For example, prior experimental results suggested that the surface consists of an ~50% ratio of bulk-like cerium and oxygen terminated domains as shown in Figure 7.16a and b, respectively. Early theoretical simulations predicted that these bulk terminated CeO2(001) surfaces were intrinsically unstable due to a nonzero surface dipole. More recent simulations suggested that the surface would be stable if the surface oxygen concentration was reduced by 50%, as shown in Figure 7.16c. This stabilization was explained by a model in which the stacking sequence is made up of units that have no dipole moment, and the resulting 1/2 ML of oxygen at the surface is then removed to neutralize and stabilize the structure. Figure 7.17 shows the integrated intensities of the MSRI Ce+ feature versus azimuthal angle. In the experiment an entire MSRI spectrum was collected at each angle and analyzed afterwards. The solid and dashed lines are for the experimental and calculated data, respectively. Both sets of data have been mirror averaged along the [010] azimuth at an azimuthal angle (f) equal to 45°, and the calculated data have been smoothed with a binomial algorithm to reduce statistical noise. The experimental data has a major peak at f = 10° and 80°. The calculations to simulate the experimental data were performed using the scattering and recoiling imaging code (SARIC) that is based on the binary collision approximation and uses the Ziegler-Biersack-Littmark universal potential to describe the interactions between atoms and includes both single- and multiple-scattering effects. The calculated AR-MSRI curves are labeled in Figure 7.17 with the topmost corresponding to the oxygen-terminated structure represented in Figure 7.16b, the second corresponding to the cerium-terminated structure represented in Figure 7.16a, and the bottom most corresponding to the oxygen-terminated structure with 0.5 monolayers of oxygen missing represented in Figure 7.16c. The peak observed at f = 45° in the calculations for the oxygen-terminated surface is associated with channeling of the incoming argon ions and the recoiling cerium ions by the rows of oxygen for the model in Figure 7.16b. Removal of 0.5 ML of oxygen from the oxygen-terminated surface results in a wider peak at f = 45°
due to a reduction of the channeling effect. This peak is not observed for the cerium-terminated model. The two peaks at f =
In conclusion, it was found that neither the bulk-terminated oxygen nor cerium surfaces gave good agreement to the AR-MSRI data. The best fit to the data is obtained for an oxygen-terminated surface that has 0.5 ML of oxygen removed. This structural model is consistent with a prior theoretical investigation that was based on reducing the surface dipole moment. The structural model proposed for CeO2(001) in this study has two separate domains which may possibly complicate thin film growth. Experimentally, it has been shown that YBa2Cu3O7-x(100) grown on CeO2(001) buffer layers results in an epitaxial relationship with the a/b axes oriented along the <110> azimuthal directions. Although the epitaxial relationship appears to be well understood for these buffer layers, further analysis may be required in light of these results. Finally, with the CeO2(001) surface structure better characterized, it should be possible to further understand the chemistry that occurs at the surface and the role of oxygen vacancies. For example, the availability of such a large number of surface vacancies may help in the bulk reduction and oxidation of ceria-based materials used in automotive catalytic-converter-based applications. Finally, the MSRI technique has much higher mass resolution than either LEIS or DRS, and can be applied to systems in which isotopic resolution is necessary.
5. Defects of Model Ceria Thin FilmsIn collaboration with Y. Gao Supported by the US DOE, Office of Basic Energy Sciences, Division of Chemical Sciences. Ceria (CeO2) or doped forms of ceria are currently used to modify the properties of the chemistry in automotive three-way exhaust catalysts. The oxygen storage properties of CeO2 allow automobiles to operate over a wider range of air to fuel ratios (A/F). Under rich conditions (i.e., low A/F), the CeO2 provides oxygen so that carbon monoxide can be fully oxidized to carbon dioxide and hydrocarbons can be fully oxidized to carbon dioxide and water. Under lean conditions (i.e., high A/F), the CeO2 can remove oxygen to reduce the formation of nitrogen oxides. Required properties for CeO2 to be effective for these types of reactions are high oxygen mobilities in the bulk and a fast molecular dissociation rate at the surface. The availability of vacancy sites for the dissociation of molecular oxygen may be an important factor in determining the reaction mechanisms on these materials. Prior experimental work has indicated that polycrystalline ceria samples are relatively easy to reduce; however, single crystals and epitaxial thin films require heating above 800 K in vacuum before any reduction is observed. These very different reduction behaviors for polycrystalline and single crystal CeO2 suggest that the crystal size has an important role in the ability to reduce these materials. As such, we have used MSRI and DRS to obtain a better understanding of the degree of surface reduction for high quality thin films. Both DRS and MSRI are very sensitive to the atomic concentrations at the topmost surface region and are thus ideally suited to detect any changes due to reduction. MSRI spectra are presented in Figure 7.18 for the sample at 295 K. The spectra in Figure 7.18 are for three different sample preparations. Figure 7.18a is a spectrum taken after annealing to 825 K in oxygen (PO2 = 1 x 10-7 Torr) for 10 min., then slowly cooling to 295 K, and finally pumping out the background oxygen. The sample after this preparation will be termed the "defect-free" surface. Figure 7.18b is a spectrum taken after annealing the defect-free sample at 775 K in UHV for 60 sec., with a subsequent 10 Langmuir (L) dose of 18O2 (P18O2 = 1 x 10-7 Torr for 100 sec.) at 295 K. Figure 7.18c is a spectrum taken after annealing the defect-free sample at 775 K in UHV for 60 sec. and cooling to 295 K. The major feature observed in the MSRI spectra, shown in Figure 7.18, was due to Ce+; however, a significant concentration of impurities can be observed as well. The major impurities were due to iron, aluminum, carbon, and silicon (not labeled). Although these impurities gave MSRI signals as large or larger than the oxygen signal, they were not detected in AES. This suggests that the impurities were at concentrations below a few percent in the surface region. The oxygen MSRI signal was small compared to cerium (and the impurities) due to the low yield for the production of O+. These spectra also illustrate a significant difference between MSRI and secondary-ion mass spectroscopy (SIMS). Due to the high energy of the recoiled particles in MSRI, no molecular fragments were observed, whereas, for the low energy of the secondary ions in SIMS, the dominant features are typically due to molecular fragments.
DRS spectra are shown in Figure 7.19 using identical preparation procedures as described for Figure 7.18. The DRS spectra contain far fewer resolved species as compared to MSRI. The species that are clearly detected in DRS are recoiled hydrogen, oxygen, and iron, argon single-scattered from cerium [Ce(SS)], and a multiply scattered component. The multiply scattered peak contains events involving small-angle scattering from both oxygen and cerium at the surface. The flight times for the recoiled and scattered particles are in excellent agreement with those predicted from the binary collision model. Two major changes are observed when comparing the DRS spectra for the defect-free surface (Figure 7.19a) and the annealed surface (Figure 7.19c). The first change is a significant increase in signal for argon scattered from cerium, observed as both a broadening of the peak to shorter TOF and an increase in intensity. For the annealed surface, the peak due to argon scattering from cerium can be resolved as two components. The component at longer TOF is due to single scattering of argon from cerium, and the component at shorter TOF is due to argon undergoing multiple small-angle scattering events with cerium. The second change is the substantial increase in the iron recoil signal. Under the different sample preparation conditions, there are significant differences in the total and relative intensities of the features observed in both MSRI and DRS.
Figure 7.20 shows the MSRI Ce+ and O+ intensities (solid lines and solid symbols) and the DRS O and Ce(SS) intensities (dashed lines and open symbols) versus sample annealing temperature. The oxygen and cerium MSRI and oxygen DRS signals are integrated peak areas and were normalized to one for spectra taken from the defect-free surface. The Ce(SS) intensities are peak heights and were normalized to 0.5 (to fit on the same graph) for a spectrum taken from the defect-free surface. The experimental procedure was to heat the defect-free sample to the indicated temperatures for two minutes in UHV, then cooled to 295 K, and finally acquire MSRI and DRS spectra concurrently. As shown in Figure 7.20, the oxygen and cerium MSRI signals and oxygen DRS signal were significantly reduced even for annealing temperatures as low as 375 K. However the Ce(SS) signal increased for the same temperature region. After annealing to 775 K, the MSRI cerium and oxygen intensities were reduced by 40% and 75% from their original values, respectively. Likewise, the DRS oxygen intensity was reduced by 12% from the original value, while the Ce(SS) intensity increased by 110% above the original value.
These results indicate that there are significant changes occurring at the CeO2(001) surface even at low temperatures. Both DRS and MSRI were found to be extremely sensitive to these changes. MSRI measurements using post-anneal oxygen isotopic labeling experiments indicated that the surface defect concentration was on the order of 11-12%. These results are in good agreement with the DRS measurements. Finally, these measurements suggest that prior studies on single crystal surfaces were not sensitive to the reduction of ceria surfaces due to their limited surface sensitivity. The pathway for the removal of oxygen from these surfaces, at low temperature (>375 K), is likely through a process involving surface exchange and bulk diffusion instead of molecular recombination of oxygen adatoms and subsequent desorption.
6. Interaction of Water with CeO2(001)In collaboration with Y. J. Kim(a) and S. A. Chambers Supported by the US DOE, Office of Basic Energy Sciences, Division of Chemical Sciences. The effect that ceria has regarding CO oxidation and the water-gas-shift reactions on precious metals can be substantial. In the automobile exhaust stream, CO, H2, and hydrocarbons can reduce ceria, while O2, H2O, and NO can oxidize ceria. It has been found that water can reoxidize reduced ceria; however, this interaction with water results in a significant reduction in the oxygen storage capacity of the material. The water-gas-shift reaction oxidizes CO to CO2 and provides H2 for the reduction of NOx. These two reactions are very important in automotive exhaust catalysis. It has been suggested that the water-gas-shift reaction occurs by a mechanism in which the noble metal-adsorbed CO was oxidized by ceria, and ceria was then oxidized by water. The reduction of ceria by hydrogen occurs by a mechanism in which oxygen is removed at the surface as water, and that it is the reduced form of ceria (CeO2-x) that can subsequently be reoxidized by water. Further studies on the interaction of water with ceria surfaces may provide insight into these proposed mechanisms. The termination of the clean and hydroxylated CeO2(001) surfaces are shown in Figure 7.21a and Figure 7.21b, respectively. The cerium, oxygen and hydrogen ions are represented by gray, white, and black circles, respectively. In Figure 7.21a, the oxygen ions in the third layer are shown as smaller circles with respect to those in the first layer. The clean surface has 50% of the top layer oxygen removed to eliminate the dipole moment at the surface, as has been previously proposed. Another method of removing the surface dipole is by replacing all the oxygen ions in the top layer with hydroxyls, Figure 7.21b.
Figure 7.22 shows the D2O TPD data for various exposures of D2O on the CeO2(001) surface held at 85 K. At low exposures, a broad feature at ~275 K continues to grow with increasing exposures. After a 30-second exposure, a new feature at 200 K appears. Increasing the exposure time to 75 seconds results in a third state at 170 K, that shifts to 152 K for longer exposures. It appears that the states at 200 and 275 K quickly saturate and that the increase in height for higher exposures was due to the tail on the 152 K state. No signal was observed for D2 desorption for any of the TPD results. The TPD data were collected up to 950 K and the data were flat and featureless above 400 K. The sample was not reoxidized after each run and this may account for the presence of some thermally induced surface defects. The TPD results were very reproducible, suggesting that the substrate surface was not changing significantly during the experimental series. A plot of the water peak area versus water exposure (not shown) gives a straight line suggesting that water does not irreversibly decompose on the surface or lead to the production of other desorption products.
In Figure 7.23 are the XPS results for O 1s emission with the sample at 85 K before (lower) and after (upper) a 220-second D2O exposure. The open circles are the experimental data, the solid line running through the open circles is the fit to the data, and the solid lines with the solid circles are the G-L peaks. The background for the fitting was a Shirley type. The O 1s peak for the clean surface was fit with a single G-L peak with a Gaussian width of 1.60 eV and with a 76% Gaussian component. After a 220-second D2O exposure at 85 K, we find that there was a shoulder at slightly higher binding energy for the O 1s XPS spectrum. Using the same fitting parameters as used for the clean surface, we find that it was necessary to include two peaks at 530.1 eV and 531.6 eV that are due to emission from lattice oxygen and hydroxyls, respectively. At 85 K the intensity ratio of the hydroxyl to lattice oxygen for our data was 12.9%. An estimation of the surface coverage for hydroxyls can be performed assuming a non-attenuating overlayer. This analysis results in a initial hydroxyl coverage of 0.89 ± 0.31 monolayers (ML) at 85 K, and with 0.70 ± 0.25 ML of the hydroxyls remaining on the surface after heating to 203 K (the point at which most of the water should have desorbed from the surface). The formation of a hydroxyl on the surface requires dissociation of water to form OD– and D+. The D+ then reacts with a surface oxygen anion forming a second hydroxyl. Therefore this suggests that 0.45 ± 0.16 ML of water dissociates on the surface at 85 K. The formation of the hydroxylated surface can be described by a mechanism in which water binds to the site above the third layer oxygen ion (i.e., the oxygen vacancy) shown in Figure 7.21a. The water then transfers a proton to the lattice oxygen ion at the surface due to the close proximity between the water and lattice oxygen ion (2.7 Å). This mechanism would then result in the model shown in Figure 7.21b.
In conclusion, we have found that there is initially a strong interaction for 1/2 ML of D2O with the CeO2(001) surface. It was found with TPD that D2O desorption occurs in three states with temperatures of 152, 200, and 275 K which are defined as multilayer D2O, weakly bound surface D2O, and hydroxyl recombination, respectively. Ce 3d XPS measurements suggest that the ceria surfaces are stoichiometric CeO2. Furthermore, it was found that even for high D2O exposures, where multilayer water desorption was observed in the TPD, only O 1s emission from the substrate and hydroxyl oxygen are observed. This is likely due to a non-wetting behavior of D2O on this surface with the formation of water nanoclusters. ReferencesBelton, D. N., C. L. DiMaggio, S. J. Schmeig, and K. Y. S. Ng, J. Catal. 157, 59 (1995). Calvert, J. G., J. B. Heywood, R. E. Sawyer, and J. H. Seinfeld, Science 261, 37 (1993). Chambers, S. A., T. T. Tran, and T. A. Hileman, J. Mater. Res. 9, 2944 (1994). Eisenberger, P. M. (ed.), Basic Research Needs for Vehicles of the Future. Princeton Materials Institute (Princeton, New Jersey) 1995. Herman, G. S., Phys. Rev. B 59, 14,899 (1999a). Herman, G. S., Surf. Sci. 437, 207 (1999b). Herman, G. S., C. H. F. Peden, S. J . Schmieg, and D. N. Belton, Catal. Lett., in press. Peden, C. H. F., D. N. Belton, and S. J. Schmieg, J. Catal. 155, 204 (1995). Peden, C. H. F., D. W. Goodman, D. S. Blair, P. J. Berlowitz, G. B. Fisher, and S. H. Oh, J. Phys. Chem. 92, 1563 (1988). Permana, H., K. Y. S. Ng, C. H. F. Peden, S. J. Schmieg, and D. N. Belton, J. Phys. Chem. 99, 16,344 (1995). Taylor, K. C., Chemtech 20, 551 (1990). Taylor, K. C., Catal. Rev.-Sci. Eng. 35, 457 (1993). Tolia, A. T., C. T. Williams, C. G. Takoudis, and M. J. Weaver, J. Phys. Chem. 99, 4599 (1995). Zhdanov, V. P., and B. Kasemo, Surf. Sci. Rep. 29, 31 (1997), and references therein.
William R. Wiley Environmental Molecular Sciences Laboratory Feedback: webmaster@emsl.pnl.gov Revised: April 17, 2000 Security & Privacy PNNL-13147 |
] azimuthal directions. These blocking cone features can be seen more clearly in the calculations for the Ce-terminated surface where there is an increase in intensity from f = 0°-10°. Neither of the two upper curves shown in Figure 7.17, nor a linear combination of the two curves, gave a reasonable fit to the experimental AR-MSRI data. These data suggest that the model proposed based on prior experimental results is not correct. The calculated results for the oxygen-terminated surface with 0.5 ML of oxygen missing gives by far the best agreement to the experimental data.
