J. Phys. Chem. B 2000, 104, 9653-9660
9653
Structure of Pd/CeOx/Al2O3 Catalysts for NOx Reduction Determined By in Situ X-ray Absorption Spectroscopy Joseph H. Holles and Robert J. Davis* Department of Chemical Engineering, UniVersity of Virginia, CharlottesVille, Virginia 22904-4741 ReceiVed: April 7, 2000; In Final Form: August 8, 2000
In situ X-ray absorption spectroscopy at the Ce LIII edge and Pd K edge was used to characterize the oxidation state and structural parameters of CeOx/Al2O3 and Pd/CeOx/Al2O3 catalysts during the reduction of NO by CO. The samples were exposed to oxidizing (5% NO in He), reducing (5% CO in He), and stoichiometric (5% NO/5% CO in He) gaseous environments. The Ce LIII edge structure indicated that one-third of the cerium readily changed oxidation state between 3+ and 4+ upon exposure to various mixtures of NO and CO at 673 K. However, structural parameters derived from EXAFS analysis indicated that cerium remained in the CeO2 crystal structure regardless of gaseous environment. The average oxidation state of Pd was also affected by gaseous environment with an average oxidation state between 0 and 2+ for a stoichiometric mixture of NO and CO. Exposure of Pd particles to NO resulted in the formation of chemisorbed oxygen and/or a surface oxide layer that can be completely removed by exposure to CO at 573 K.
Introduction Automobile exhaust catalysts are designed to reduce emissions of carbon monoxide, nitrogen oxides, and uncombusted hydrocarbons. These catalysts typically contain noble metals such as Pt, Pd, and Rh with a ceria promoter supported on alumina. Traditionally, the principal function of the Rh is to control emissions of nitrogen oxides (NOx)1 by reaction with carbon monoxide. However, use of Pd to control NOx emissions has increased. As a result, the NO + CO reaction has been studied recently on high surface area Pd/Al2O3 powder catalysts.2-6 Although ceria has been shown to effectively promote Rh for the NO + CO reaction,7 its use as a promoter for Pd has not been as widely studied.6,8 Nevertheless, substantial improvements in reaction rates are observed when ceria is added to supported Pd catalysts. Investigations on the role of ceria in automobile exhaust catalysts have demonstrated its capacity to store and release oxygen during the catalytic reaction.9 Whether this oxidationreduction cycle results in CeO2 conversion to Ce2O3 or simply a disordered suboxide under reaction conditions is still unclear. X-ray absorption spectroscopy (XAS) provides a valuable tool for investigating ceria since it allows determination of both oxidation state and atomic structural parameters. The technique has been used previously to elucidate the valence of cerium in ceria-supported Pd and Rh catalysts.8,10 Unfortunately, these studies were not performed in situ and the catalysts were reduced in H2 and oxidized in air instead of CO and NO. A later paper also reported on the in situ reduction of ceria reduction by H2 as measured by XAS.11 Since the oxidation state of cerium in these catalysts is dependent on the operating conditions, in situ experiments using NO and CO are needed to derive conclusions that are relevant to automotive catalysis. * Author to whom correspondence should be addressed at Department of Chemical Engineering, Chemical Engineering Bldg., Room 117A, University of Virginia, 102 Engineers Way, Charlottesville, VA 229044741. Phone: (804) 924-6284. Fax: (804) 982-2658. E-mail: rjd4f@ virginia.edu.
The Pd component of NOx reduction catalysts can also be easily probed by XAS. For example, Matsumoto and Tanabe have shown Pd in zeolite Y consists of small clusters under operating conditions for the reaction of NO reduction by propane.12 Additionally, Ali et al. have used XAS to show that Pd is transformed into dispersed Pd2+ ions on acidic supports and to Pd oxide clusters on nonacidic supports after exposure to the reaction mixture of methane, nitrogen monoxide, and oxygen.13 The present study uses XAS to investigate in situ the oxidation state and structural parameters of both Ce and Pd in catalysts that have demonstrated activity for NO reduction by CO. Typical three-way catalysts operate at temperatures around 900 K and do not promote NOx reduction except in a narrow temperature window in the vicinity of catalyst lightoff.1 This is a concern during cold start since the catalyst has not yet warmed to operating temperature and a significant amount of total emissions is released during this period. Thus, the temperatures used in this work are less than 673 K to simulate the state of the catalyst relevant to cold-start conditions. Experimental Section Sample Preparation. A CeOx/Al2O3 sample was prepared by stirring γ-Al2O3 (Alfa Aesar, 99.97%) and cerium(III) acetylacetonate (acac) hydrate (Aldrich) in toluene for 2 h at 353 K, removing excess solvent under vacuum in a rotary evaporator, and drying for 24 h in air at 473 K. The catalyst was then calcined in flowing air (BOC gases) by heating to 673 K at 0.5 K min-1 and then remaining at 673 K for 4 h. A Pd/Al2O3 sample was prepared similarly using palladium acetylacetonate (Aldrich, 99%) and then drying and calcining as described above. To prepare a ceria-promoted sample (Pd/ CeOx/Al2O3), palladium was deposited on the CeOx/Al2O3 catalyst using the above procedure. Subsequent reduction of the Pd catalysts occurred at 673 K for 2 h in flowing dihydrogen (99.999% from BOC gases, passed through a Matheson model 8371v purifier).
10.1021/jp001347r CCC: $19.00 © 2000 American Chemical Society Published on Web 09/22/2000
9654 J. Phys. Chem. B, Vol. 104, No. 41, 2000 The percent of the Pd atoms exposed was determined using H2 chemisorption (99.999% from BOC gases passed through a Supelco OMI purifier) in a Coulter Omnisorp 100CX system. Adsorption isotherms were obtained at 303 K. The samples were evacuated for 40 min at 303 K after total chemisorption to decompose the β-phase Pd hydride. The hydrogen uptake on the surface was calculated from the difference between total chemisorption and reversible chemisorption. Elemental analysis was performed by Galbraith Laboratories Inc., Knoxville, TN. The reaction experiments were performed in a quartz reactor containing approximately 50 mg of catalyst diluted in chromatographic silica gel (Fisher) supported on a quartz frit. Reactant feed gas was a mixture of 5.00%NO/5.07%CO/He from BOC gases. Gas flow rates were controlled using mass flow controllers (Brooks Series 5850C) and varied from 5 to 250 mL min-1. Total pressure in the reactor was near atmospheric. Reactant and product analyses involved a combination of gas chromatography (HP 5890 Series II with an Alltech CTR I column) and mass spectrometry (Dycor MA 100 model). The standard catalyst pretreatment consisted of heating at approximately 523 K for 1 h under flowing He. Product analysis was performed after a period of 45 min on stream at each set of reaction conditions to allow the reaction to reach steady state. X-ray Absorption Spectroscopy. The X-ray absorption spectra associated with the Ce LIII edge (5723 eV) and the Pd K edge (24350 eV) of the materials were recorded in the transmission mode on beamline X23A2 at the National Synchrotron Light Source, Brookhaven National Laboratory, Upton, NY. In addition to examining the catalysts, the EXAFS associated with cerium(III) nitrate (Aldrich), cerium(IV) oxide (Aldrich), and palladium metal foil (Goodfellow) were recorded and used as standards. The reference spectrum of palladium(II) oxide was furnished by Prof. D. Resasco, School of Chemical Engineering, Oklahoma University. The storage ring operated with an electron energy of 2.8 GeV with beam currents ranging from 150 to 250 mA. As energy calibrations standards, appropriate metal foils (Goodfellow) were placed between the second and third ionization chambers. For operation at the Ce LIII edge, the first chamber was filled with He and the second and third chambers with N2. All chambers were filled with Ar for collection at the Pd K edge. Sample powders were diluted in boron nitride (Alfa Aesar) and pressed into self-supporting wafers and placed in an in situ cell capable of both heating and cooling the wafer in controlled atmospheres. The loading of sample in BN was adjusted to obtain ∆µx ) 0.3 for Ce and 0.75 for Pd to maximize the signalto-noise. The sample cell was a stainless steel chamber with water-cooled Kapton windows and a copper sample holder that contained electrical cartridge heaters. Standard pretreatment of the catalyst consisted of heating the wafer at 5 K min-1 in flowing helium. The sample was then exposed to oxidizing, reducing, or stoichiometric conditions of 5% NO (98.5%, Aldrich), 5% CO (99%, Aldrich), or 5% NO and 5% CO in He (99.999%, Air Products). The gas mixture was prepared from the individual gases using calibrated electronic mass flow controllers (Brooks series 5850 C). The reactant gases were then introduced to the sample cell via a stainless steel sample line next to the sample wafer. At least three X-ray absorption spectra were collected at a specified temperature. The EXAFS data were processed with Macintosh version of the University of Washington analysis program. The raw EXAFS data were extracted from the absorption spectra and normalized by dividing the absorption spectra by the height of the edge jump. Post-edge background subtraction was done with
Holles and Davis TABLE 1: Elemental Analyses and Chemisorption Results sample
wt % Pda
Pd/Al2O3 Pd/CeOx/Al2O3 CeOx/Al2O3
4.92 4.12
wt % Cea 10.66 10.83
H/Mb 0.18 0.46
a Quantitative analysis by Galbraith Laboratories Inc. b Total chemisorption-reversible chemisorption at 303 K.
Figure 1. Arrhenius-type plots for the NO + CO reaction over Pd/ Al2O3 and Pd/CeOx/Al2O3 for reactant mixture of 5.07 kPa of each NO and CO.
a cubic spline and 3 knots. The EXAFS were k3 weighted before additional processing. Errors associated with the structural parameters for Pd-Pd absorber-backscatterer pairs are assumed to be about the same as those generally accepted for the EXAFS technique, (20% for N (coordination number), (10% for ∆σ2 (change in Debye-Waller factor), and (0.01 Å for R (interatomic distance). However, due to the smaller data range used for Pd-O, Ce-Ce, and Ce-O, correlation between N and ∆σ2 decreased the reliability of those structural parameters. In some cases (denoted in the Results section), the coordination number was set to a fixed value. Results Table 1 summarizes results from elemental analysis and H2 chemisorption. Assuming a spherical particle shape, the average Pd particle size for the Pd/Al2O3 and Pd/CeOx/Al2O3 samples are approximately 5 and 2 nm, respectively. The cerium loading (∼11 wt %) on both supported samples is on the order of that needed for a theoretical monolayer (∼9 wt % Ce based on the Ce surface density of 4.83 Ce/nm2 derived from the CeO2(110) plane). Very small and broad X-ray diffraction (XRD) peaks associated with CeO2 indicate that the promoter was well dispersed on the alumina. Figure 1 shows the catalytic activity for unpromoted and ceria-promoted Pd supported on alumina for the NO + CO reaction. The addition of the ceria promoter resulted in an activity increase of almost 2 orders of magnitude. Ceria also decreased the selectivity to N2 (N2O being the other nitrogencontaining product) from 30% to 20% at 500 K. Although alumina-supported ceria catalyzed the NO + CO reaction with no metal present, the reaction rate, on a per gram basis, was a factor of 4 lower than the unpromoted Pd catalyst. A more detailed discussion of the reaction kinetics over these catalysts was reported earlier.6 The Ce LIII edge XANES spectrum for cerium(III) nitrate (Ce(NO3)3), cerium(IV) oxide (CeO2), and the CeOx/Al2O3 sample at room temperature are shown in Figure 2. These spectra
Pd/CeOx/Al2O3 Catalysts for NOx Reduction
Figure 2. Cerium LIII edge spectra for Ce(NO3)3 and CeO2 as Ce3+ and Ce4+ standards and the CeOx/Al2O3 sample.
Figure 3. Cerium LIII edge spectra for CeOx/Al2O3 at 673 K as a function of atmosphere.
demonstrate that the double peak in the white line region is associated with a Ce(IV) compound. For a sample having an average Ce oxidation state between 3+ and 4+, changes in the magnitude of the second peak can be used to detect subtle changes in the oxidation state of the sample. In addition, the sample with ceria supported on alumina is very similar to the CeO2 standard. Figure 3 shows the XANES spectrum for the CeOx/Al2O3 catalyst exposed to a series of oxidizing, reducing, and stoichiometric NO and/or CO atmospheres at 673 K. The sample was first exposed to 5% NO in He, then 5% NO and 5% CO in He, then 5% CO in He, and finally 5% NO and 5% CO in He. The magnitude of the second peak is largest for the oxidizing NO atmosphere, smallest for the reducing CO atmosphere, and intermediate for the stoichiometric NO/CO atmosphere. The results obtained in stoichiometric NO + CO
J. Phys. Chem. B, Vol. 104, No. 41, 2000 9655
Figure 4. Cerium LIII edge difference spectra for cerium standards and CeOx/Al2O3 as a function of atmosphere.
did not depend on the sample being previously exposed to oxidizing or reducing conditions. Although the changes in oxidation state can be observed in the XANES, similar shifts in edge energy were not detectable. The edge energy (determined by the maximum in the first derivative) for CeO2 was 5727.0 eV with the edge energies for the samples at 5725.55726.0 eV. The differences in XANES structure are more clearly illustrated in Figure 4. This figure compares the LIII XANES difference spectra of the cerium standards and the sample under various conditions. The difference spectra for the samples are similar to that of the standards with features at about 3 and 13 eV; however, these features are less intense for the samples. The peak at about 13 eV shows that the difference between oxidizing and reducing conditions is about one-third the difference between the 4+ and 3+ standards. The difference between the room temperature and 673 K in He conditions is slightly smaller. The relative magnitude of these difference peaks provides evidence for the quantity of cerium atoms that change oxidation states as the gaseous environments change. The k3-weighted EXAFS function associated with the Ce LIII edges of CeO2 and the CeOx/Al2O3 sample exposed to various atmospheres are shown in Figure 5. The EXAFS functions (k3χ) of the catalyst as a function of atmosphere are quite similar to each other but differ somewhat from that of the CeO2 standard. The similarity of the EXAFS function among the samples in various environments can also be seen in the radial structure functions of Figure 6. The peak in the transform at about 2 Å results from the nearest neighbor oxygen atoms whereas the peak between 3 and 4 Å results from the higher shell backscattering from Ce. Each contribution to the Fourier transform was isolated by backtransforming over the region of interest. The r-space Ce-O fitting range was 1.21-2.46 Å. A Ce-O distance of 2.34 Å and a coordination number of 8 were used as the standard.14 For Ce-Ce, the r-space fitting range was 2.79-4.73 Å. A Ce-Ce distance of 3.83 Å and coordination number of 12 was used.15 A summary of structural parameters including coordination numbers, interatomic distances, and Debye-Waller factors from the Ce-O and Ce-Ce fitting as a function of conditions are summarized in Table 2. Representative
9656 J. Phys. Chem. B, Vol. 104, No. 41, 2000
Holles and Davis TABLE 2: Cerium EXAFS Fitting Results for Ce-O and Ce-Ce in CeOx/Al2O3 Catalyst treatment He (303 K) He (673 K) 5%NO (673 K) 5%NO/5%CO (673 K) 5%CO (673 K) He (303 K) He (673 K) 5%NO (673 K) 5%NO/5%CO (673 K) 5%CO (673 K)
backabsorber scatterer Na Ce Ce Ce Ce Ce Ce Ce Ce Ce Ce
O O O O O Ce Ce Ce Ce Ce
5.3 5.5 5.5 5.0 4.2 6.7 6.7e 6.7e 6.7e 6.7e
Rb (Å)
∆σ2 c (Å2)
∆Eo d (eV)
2.33 2.32 2.34 2.33 2.33 3.81 3.81 3.83 3.81 3.82
0.0005 0.0027 0.0029 0.0015 0.0002 0.0000e 0.0015 0.0019 0.0020 0.0023
2 0 -2 -1 0 0 -1 -4 -2 -2
a
Coordination number. b Interatomic distance. c Change in DebyeWaller factor relative to bulk CeO2 at 303 K. d Shift in edge energy required to optimize curve fit. e Fixed value due to strong correlation between N and ∆σ2.
Figure 5. Cerium LIII EXAFS for CeO2 and CeOx/Al2O3 as a function of atmosphere at 673 K.
Figure 7. Fourier-filtered EXAFS function of the Ce-O shell and resulting curve fit used to calculate the structural parameters for CeOx/ Al2O3 at 673 K in 5% NO in He.
Figure 6. Radial structure function (not corrected for phase shifts) derived from the Fourier transform of the Ce-LIII edge EXAFS of CeO2 at room temperature and CeOx/Al2O3 at room temperature and at 673 K as a function of atmosphere.
Ce-O and Ce-Ce fits for the sample in a 5% NO in He atmosphere are shown in Figures 7 and 8, respectively. The parameters presented in Table 2 indicate that oxidizing or reducing conditions at 673 K had little effect on the overall structure of the supported cerium oxide. The Pd K edge XANES spectrum for Pd metal foil and palladium(II) oxide (PdO) are shown in Figure 9. This figure demonstrates that the near edge spectrum differs as a function of metal oxidation. As in the Ce case, the XANES spectrum can be used to discern subtle changes in the average oxidation state of the Pd particles. Figure 10 shows the Pd XANES spectrum for the Pd/CeOx/Al2O3 catalyst exposed to oxidizing, reducing, and stoichiometric atmospheres at 573 K. The sample was first exposed to 5% NO in He, then 5% CO in He, and
Figure 8. Fourier-filtered EXAFS function of the Ce-Ce shell and resulting curve fit used to calculate the structural parameters for CeOx/ Al2O3 at 673 K in 5% NO in He.
finally, 5% NO and 5% CO in He. When exposed to NO, the first feature in the XANES spectrum at about 15 eV above the edge increases significantly to be more like PdO. After exposure of this “oxidized” sample to CO, this feature diminished to an intensity typical of metallic Pd. Finally, exposure to both NO and CO results in a spectrum intermediate between the oxide and the metal.
Pd/CeOx/Al2O3 Catalysts for NOx Reduction
J. Phys. Chem. B, Vol. 104, No. 41, 2000 9657
Figure 9. Palladium K edge spectra for Pd metal and PdO standards. Figure 11. Palladium K edge difference spectra for palladium standards and Pd/CeOx/Al2O3 as a function of atmosphere.
Figure 10. Palladium K edge spectra for Pd/CeOx/Al2O3 at 573 K as a function of atmosphere. Figure 12. Palladium K edge EXAFS for Pd metal and PdO standards.
Figure 11 more clearly illustrates the differences in XANES structure. Again, the difference spectra for the sample are similar to those of the standards with major features at 0, 20, and 40 eV. The peak at 20 eV shows that the difference between the oxidizing and reducing conditions is approximately one-half the difference between the oxide and metal standards. In addition, the difference between stoichiometric and reduced conditions is about one-quarter the magnitude. These results indicated that under stoichiometric reaction conditions at 673 K, an appreciable amount of Pd was oxidized. However, even under the most severe oxidizing environment at 673 K, not all of the Pd was oxidized. The k3-weighted EXAFS function associated with the Pd K edges of the standards and samples are presented in Figures 12 and 13, respectively. Due to the noise level at high k values, analysis was done over the range of 3.0-13.0 Å-1. The radial structure functions in Figure 14 show a major peak at about 2.5 Å associated with the backscattering from Pd neighbors. The other feature of note is a peak at about 1.5 Å (marked by the arrow) which suggests backscattering from oxygen nearest
neighbors since it is also prominent in PdO. The low intensity of this peak in the samples is due to the k3 weighting of the EXAFS data. Interestingly, this feature is only present in the sample when it is exposed to stoichiometric and oxidizing conditions. The radial structure functions were backtransformed over the range of 1.63-3.20 Å and fit to Pd metal having the first Pd-Pd shell at 2.75 Å and a coordination number of 12.16,17 For Pd-O, the backtransform was from 0.96 to 2.03 Å, which was fit to the Pd-O standard having an interatomic distance of 2.01 Å with a coordination number of 4.18 Single-shell curve fits for Pd-Pd and Pd-O are shown in Figures 15 and 16, respectively. The structural parameters derived from the fits are summarized in Table 3. Discussion The catalytic activity for the ceria-promoted palladium was approximately 2 orders of magnitude larger than the unpromoted palladium. Ceria has been previously shown to promote the NO
9658 J. Phys. Chem. B, Vol. 104, No. 41, 2000
Holles and Davis
Figure 15. Fourier-filtered EXAFS function of the Pd-Pd shell and resulting curve fit used to calculate the structural parameters for Pd/ CeOx/Al2O3 at 573 K in 5% NO in He.
Figure 13. Palladium K edge EXAFS for Pd/CeOx/Al2O3 as a function of atmosphere at 573 K.
Figure 16. Fourier-filtered EXAFS function of the Pd-O shell and resulting curve fit used to calculate the structural parameters for Pd/ CeOx/Al2O3 at 573 K in 5% NO in He.
TABLE 3: Palladium EXAFS Fitting Results for Pd-O and Pd-Pd in Pd/CeOx/Al2O3 Catalyst
Figure 14. Radial structure function (not corrected for phase shifts) derived from the Fourier transform of the Pd K edge EXAFS of Pd/ CeOx/Al2O3 as a function of atmosphere at 573 K and PdO at room temperature.
+ CO reaction for rhodium,6,7 however, the observed effect was not as large (approximately 50%). The enhancement in observed rates is either due to a modification of the metal particles or creation of new sites at the metal/promoter interface. The maximum temperature in Figure 1 is approximately 600 K. Therefore, ceria provides a significant promotional effect even at temperatures below the steady operating temperature of a catalytic converter. To further investigate the synergistic effect between the ceria and the palladium, we used X-ray absorption spectroscopy to study the oxidation state and structural parameters of both Pd and Ce under in situ reaction conditions. The 4+ and 3+ oxidation states of the cerium atom are easily distinguished in the XAS spectrum at the LIII edge. The near edge structure shows two peaks in the Ce4+ compound having
treatment (at 573 K)
absorber
backscatterer
Na
Rb (Å)
∆σ2 c (Å2)
∆Eo d (eV)
He 5%NO 5%CO 5%NO/5%CO 5%NO 5%NO/5%CO
Pd Pd Pd Pd Pd Pd
Pd Pd Pd Pd O O
8.7 6.7 8.8 8.1 3.7 3.4
2.73 2.73 2.74 2.74 2.02 2.05
0.0047 0.0043 0.0047 0.0051 0.0070 0.0087
3 6 5 2 2 -2
a Coordination number. b Interatomic distance. c Change in DebyeWaller factor relative to bulk Pd or PdO at 303 K. d Shift in edge energy required to optimize curve fit.
an energy difference of about 7 eV. The final electronic states associated with these features have been attributed to 2p5-4f25d* for the first peak and 2p5-4f0-5d* for the second.11,19 The near edge structure of the Ce3+ compound has a single peak shifted to lower energy than those of Ce4+ by about 2 eV, with the final state attributed to 2p5-4f1-5d*. Therefore, in addition to the expected shifts in energy associated with changes in oxidation state, cerium also reveals a substantially modified near edge structure. These significant alterations in the edge structure allow subtle changes in oxidation state to be observed that are beyond the energy resolution of the monochromator. The XANES of our samples show that cerium changes its local environment as a function of surrounding atmosphere. When exposed to oxidizing conditions (NO in He at 673 K), the ceria component was in a substantially oxidized (Ce4+) state. Interestingly, even under these strong oxidizing conditions, the
Pd/CeOx/Al2O3 Catalysts for NOx Reduction sample was not as completely oxidized as it was initially at room temperature in He. After exposure to reducing conditions (CO in He at 673 K), the ceria component was partially reduced to Ce3+. The difference plots suggest that approximately onethird of cerium atoms change oxidation state when the gaseous environment is changed from NO to CO. The average oxidation state of Ce after exposure to an equimolar mixture of NO + CO was found to be between that of supported ceria exposed to either gas. Thus, under stoichiometric reaction conditions typical of our catalytic experiments, the ceria component appeared to be in a mixed formal oxidation state between 3+ and 4+. The in situ XANES results suggest that supported ceria has the ability to easily adjust its average oxidation state in response to changes in gaseous environment at 673 K without forming stoichiometric 3+ or 4+ oxides. Similar changes in cerium oxidation state were also detected for the Pd/CeOx/Al2O3 sample. The EXAFS associated with the Ce LIII edge provided additional information on the local microstructure of the ceria promoter. X-ray diffraction on a sample at room temperature indicated that the supported ceria was essentially in the CeO2 crystal structure. However, the XRD peaks were small and broad suggesting a very high dispersion of ceria on the support. Based on the crystal structure of CeO2,15 there are 8 nearest neighbor oxygen atoms at a distance of 2.34 Å from a central cerium atom. There are also 12 next nearest neighbor cerium atoms at a distance of 3.83 Å. These values are consistent with the interatomic distances derived from the EXAFS analysis of the sample. The Ce-O and Ce-Ce distances in Table 2 are 2.33 and 3.81 Å, respectively, at 303 K in He. The low Ce-Ce coordination number of 6.7 is also consistent with small ceria crystallites. For Ce-Ce fitting, the change in Debye-Waller (∆σ2) relative to the standard was set to zero for the sample at 303 K since allowing ∆σ2, N, and R to float resulted in physically unrealistic numbers due to strong correlation between N and ∆σ2. The coordination number determined at 303 K was then fixed at that value for all other cases since the Ce-Ce coordination number is not be expected to change as a function of temperature or oxidation state of the sample. Upon heating to 673 K in He, the Ce-O distance remained about the same at 2.32 Å with no change in coordination number. The Ce-Ce distance remained unchanged at 3.81 Å. Based on the similarity of these parameters at 673 K with the values at 303 K, the ceria structure appeared to still be the same as CeO2. However, the near edge data appear to contradict this by indicating that the average cerium oxidation state decreased to a value below 4+. Thus, we conclude that supported ceria remained in the CeO2 crystal structure upon partial loss of oxygen atoms due to thermal treatment. Changing the gaseous environment also had very little effect on interatomic distances and coordination numbers, and therefore likely did not change the ceria crystal structure. It does not appear that the change in oxidation state resulted from conversion to the Ce2O3 crystal structure upon heating. In the Ce2O3 structure, the Ce-Ce interatomic distance is 3.77 Å with a coordination number of 6 and the three shortest Ce-O distances are 2.12, 2.88, and 3.48 Å with a coordination number of 3 for each.20 The EXAFS results did not indicate a shift to shorter and/or multiple Ce-O interatomic distances together with a decrease in the coordination number. Since the Ce-Ce distance in the two different crystal structures is very similar at 3.81 Å in CeO2 and 3.77 Å for Ce2O3, it cannot be used to unambiguously assign crystal structure. However, the absence of any change in the Ce-Ce distance between 303 and 673 K is consistent with no phase
J. Phys. Chem. B, Vol. 104, No. 41, 2000 9659 change upon heating. From the Ce-O and Ce-Ce structural parameters, we conclude that a phase transformation from CeO2 to Ce2O3 cannot explain the observed change in near edge structure resulting from heating to 673 K in He. Beck et al. found that after heating a sample of aluminasupported ceria and rhodium to 773 K in hydrogen, a significant fraction of cerium atoms were reduced to 3+ as detected by X-ray photoelectron spectroscopy (XPS). They postulated that the reduced cerium formed Ce2O3. Our XANES results are also consistent with the presence of some Ce3+ in a sample heated in CO at 673 K. However, on the basis of the additional structural data from EXAFS analysis, we postulate that the 3+ cerium atoms remain in an oxygen-deficient CeO2 structure. The XANES at the Pd K edge indicated that the average oxidation state of Pd changed as a function of surrounding atmosphere. The Pd was partially oxidized when exposed to NO in He, and almost completely reduced when exposed to CO in He. An intermediate average oxidation state was observed for a sample exposed to a stoichiometric mixture of NO and CO at 673 K. The partial oxidation of Pd is most likely due to the formation of an oxide at the surface of the metal particle. Since about half of the Pd atoms are exposed to the surface, this surface oxidation can dramatically alter the average Pd oxidation state observed by XANES. The local microstructure of the metal particles is further defined using Pd K edge EXAFS. Bulk Pd metal has a Pd-Pd distance of 2.75 Å and a coordination number of 12. This agrees well with the results in Table 3 for our sample with a Pd-Pd distance of 2.73 for the sample in He at 573 K. The apparent shortening of the bond length probably results from anharmonic surface atom motions on the small metal particles instead of a true shortening of the bond length.21,22 Exposure of the sample to different gaseous environments did not affect the interatomic Pd-Pd distances. The Pd-Pd coordination number of approximately 9 for the sample in He at 573 K corresponds to an average Pd particle diameter of 1.8 nm which is in good agreement with our hydrogen chemisorption results.16,23 The parameters associated with oxygen in the samples were based on a Pd-O distance of 2.01 Å and coordination number of 4 for bulk PdO. The Pd-O distances in the catalysts were 2.02 Å after exposure to NO and 2.05 Å after exposure to NO + CO. The slight decrease in the in the Pd-O coordination number from 3.7 to 3.4 after introduction of a stoichiometric mixture of NO/CO was consistent with the reducing nature of the CO component. No structural feature associated with Pd-O was observed for the sample in He or CO at 573 K, indicating the presence of very little PdO under these conditions. The combination of XANES and EXAFS results can be used to understand the effect of gaseous environment on the supported Pd particles. The XANES results indicated that Pd was oxidized and reduced as conditions changed, although exposure to NO did not completely oxidize the particles to PdO at 573 K. Similarly, Pd-O coordination numbers were affected by the gaseous environment. The Pd-Pd interatomic distance was virtually the same as bulk Pd regardless of environment, which is surprising since the Pd-Pd distance in PdO is 3.02 Å with a coordination number of 4.18,24 Therefore, the influence of the gaseous environment on the XANES and EXAFS is consistent with formation and removal of surface oxygen on the supported Pd particles. It is difficult to differentiate between chemisorbed oxygen atoms and surface palladium oxide since the Pd-O distance is 2.03 Å for an oxygen atom chemisorbed in a Pd 3-fold fcc hollow site,25 which compares to 2.01 Å in PdO. In either case, oxidation of the Pd surface appears to be completely
9660 J. Phys. Chem. B, Vol. 104, No. 41, 2000 reversible since exposure to CO returns the catalyst to its original reduced state. Formation of an oxide layer on supported Pd catalysts by exposure to air has also been observed by McCaulley with the oxide layer being removed by reaction with hydrogen.16 Exposure to the NO in He environment may result in adsorbed O, N, and NO on the Pd surface. Adsorbed CO may also be present as a result of exposure to a stoichiometric NO + CO mixture. However, the expected Pd-O interatomic distance of about 2.02 Å is more consistent with our results in Table 3 than calculated values of 1.94 Å for Pd-N, 2.06 and 2.12 Å for Pd-NO (depending on site), and 2.07 and 2.12 Å for PdCO (depending on site).26,27 The Pd-O coordination numbers reported in Table 3 for the oxidizing and stoichiometric atmospheres approach the value for bulk PdO which is much higher than anticipated for a surface oxide layer. In both cases, these peaks in the transforms are small and noise in both the reference and sample spectrum may result in the slightly inflated values. Ali et al. found that small clusters of Pd on nonacidic supports were transformed into PdO when exposed to a CH4 + NO + O2 reaction mixture.13 The formation of an oxidized Pd species was also observed with infrared spectroscopy during NO reduction by CO (without the presence of O2) on unpromoted Pd.5 Conclusions In situ X-ray absorption spectroscopy has shown that supported cerium oxide readily changes oxidation state upon exposure to various mixtures of NO and CO. For stoichiometric mixtures of NO + CO typical of catalytic studies, the average oxidation state of cerium was between 3+ and 4+. Analysis of structural parameters from EXAFS indicated that ceria is in the CeO2 structure at 673 K regardless of gas composition. Apparently, supported CeO2 did not transform to Ce2O3 upon partial reduction at 673 K, but may instead form an oxygendeficient CeO2 crystallite. The average oxidation state of Pd was also affected by gas composition. In a stoichiometric mixture of NO and CO, the average oxidation state was between 0 and 2+, whereas exposure of Pd particles to NO-containing environments resulted in the formation of chemisorbed oxygen and/or a surface oxide layer which can be completely removed by exposure to CO.
Holles and Davis Acknowledgment. This work was supported by the Division of Chemical Sciences, Office of Basic Energy Sciences, Office of Energy Research, U.S. Department of Energy. Also, research was carried out in part at the National Synchrotron Light Source, Brookhaven National Laboratory, which is supported by the U.S. Department of Energy. We thank Professor Daniel Resasco of Oklahoma University for providing the X-ray absorption spectrum of PdO. References and Notes (1) Taylor, K. C. Catal. ReV.-Sci. Eng. 1993, 35, 457. (2) Rainer, D. R.; Vesecky, S. M.; Koranne, M.; Oh, W. S.; Goodman, D. W. J. Catal. 1997, 167, 234. (3) Rainer, D. R.; Koranne, M.; Vesecky, S. M.; Goodman, D. W. J. Phys. Chem. B 1997, 101, 10769. (4) Almusaiteer, K.; Chuang, S. S. C. J. Catal. 1998, 180, 161. (5) Almusaiteer, K.; Chuang, S. S. C. J. Catal. 1999, 184, 189. (6) Holles, J. H.; Switzer, M. S.; Davis, R. J. J. Catal. 2000, 190, 247. (7) Oh, S. H. J. Catal. 1990, 124, 477. (8) Le Normand, F.; Hilaire, L.; Kili, K.; Krill, G.; Maire, G. J. Phys. Chem. 1988, 92, 2561. (9) Yao, Y. F. Y. J. Catal. 1987, 87, 152. (10) Beck, D. D.; Capehart, T. W.; Hoffman, R. W. Chem. Phys. Lett. 1989, 159, 207. (11) El Fallah, J.; Boujana, S.; Dexpert, H.; Kiennemann, A.; Majerus, J.; Touret, O.; Villain, F.; Le Normand, F. J. Phys. Chem. 1994, 98, 5522. (12) Matsumoto, H.; Tanabe, S. J. Phys. Chem. 1995, 99, 6951. (13) Ali, A.; Alvarez, W.; Loughran, C. J.; Resasco, D. E. Appl. Catal. B 1997, 14, 13. (14) Antonio, M. R.; Brazdil, J. F.; Glaeser, L. C.; Mehicic, M.; Teller, R. G. J. Phys. Chem. 1988, 92, 2338. (15) Berry, L. G., Ed. Powder Diffraction File. Vol. 34. 1974, Joint Committee on Powder Diffraction Standards: Philadelphia. (16) McCaulley, J. A. J. Phys. Chem. 1993, 97, 10372. (17) Davis, R. J.; Landry, S. M.; Horsley, J. A.; Boudart, M. Phys. ReV. B: Condens. Matter 1989, 39, 10580. (18) Moore, W. J., Jr.; Pauling, L. J. Am. Chem. Soc. 1941, 63, 1392. (19) Prieto, C.; Lagarde, P.; Dexpert, H.; Briois, V.; Villain, F.; Verdaguer, M. J. Phys. Chem. Solids 1992, 53, 233. (20) Zachariasen, W. Z. Phys. Chem. 1926, 123, 134. (21) Hansen, L. B.; Stoltze, P.; Nørskov; Clausen, B. S.; Niemann, W. Phys. ReV. Lett. 1990, 64, 3155. (22) Clausen, B. S.; Topsøe, H.; Hansen, L. B.; Stoltze, P.; Nørskov, J. K. Jpn. J. Appl. Phys. 1993, 32, 95. (23) Kip, B. J.; Duivenvoorden, F. B. M.; Koningsberger, D. C.; Prins, R. J. Catal. 1987, 105, 26. (24) Waser, J.; Levy, H. A.; Peterson, Acta Crystallogr. 1953, 6, 661. (25) Van Santen, R. A.; Neurock, M. Catal. ReV.-Sci. Eng. 1995, 37, 557. (26) Loffreda, D.; Simon, D.; Sautet, P. J. Chem. Phys. 1998, 108, 6447. (27) Loffreda, D.; Simon, D.; Sautet, P. Surf. Sci. 1999, 425, 68.
