Microstructural evolution of nickel nanoparticle ... - Springer Link

Journal of ELECTRONIC MATERIALS, Vol. 35, No. 5, 2006

Special Issue Paper

Microstructural Evolution of Nickel Nanoparticle Catalysts Supported on Gadolinium-Doped Ceria during Autothermal Reforming of Iso-Octane VELMURUGAN PALANIYANDI,1 MOHAMMAD SHAMSUZZOHA,1,2 EARL T. ADA,2 GIOVANNI ZANGARI,1,3 and RAMANA G. REDDY1,4 1.—Department of Metallurgical and Materials Engineering, The University of Alabama, Tuscaloosa, AL 35487. 2.—Central Analytical Facilities, School of Mines and Energy Development, The University of Alabama. 3.—Present address: Department of Materials Science and Engineering, University of Virginia, Charlottesville, VA 22904. 4.—E-mail: [email protected]

The microstructure and composition of a nanoparticle Ni catalyst supported on gadolinium-doped ceria (Ce1"xGdxO(4"x)/2) were studied using transmission electron microscopy (TEM), x-ray diffraction (XRD), and x-ray photoelectron spectroscopy (XPS). The support of the fresh catalyst exhibits a homogenous aggregation of crystalline grains, with sizes ranging between 20 nm and 50 nm. The crystalline structure of the fresh catalyst support is of the CeO2 phase, in which gadolinium atoms exist in a solid solution of CeO2. Nickel in the fresh catalyst is highly dispersed and forms granular crystals that are 5–30 nm in size on the surface of the ceria support. The support of the used catalyst exhibits a bimodal distribution of grains in which smaller grains have similar structure and morphology as those in the fresh catalyst, while the larger sized grains appear dull and exhibit nonfaceted crystal morphology resulting either from the sintering of a number of CeO2 grains or by the occupation of highly defective crystals of Ce2O3 and CeO phases. A thin amorphous layer of carbon also covers most of the larger grains in the used catalyst. The Ni particles could not be imaged by TEM in the used catalyst, but energy dispersive x-ray spectroscopy (EDX) detected their presence. The XPS analysis of the catalyst samples suggests the participation of lattice O atoms from the ceria support in the catalytic reaction. The XPS data also show the presence of carbonate species and a higher hydrocarbon concentration in the used catalyst. Key words: Autothermal reforming, ceria, nickel catalyst, gadolinium

INTRODUCTION The most promising method of supplying hydrogen to polymer electrolyte fuel cells (PEFCs) for automotive applications is the onboard processing of liquid fuels. Hydrogen can be produced from liquid fuels by (1) direct decomposition, (2) steam reforming, or (3) partial oxidation of the fuel.1 For methane fuel, these reactions can be written as follows: CH4 ! C 1 2H2

DH 5 74:8 kJ=mol

CH4 1 H2 O ! CO 1 3H2

(1)

DH 5 225:4 kJ=mol (2)

CH4 1 1=2O2 ! CO 1 2H2 DH ¼ "22:2 kJ=mol (3) (Received May 16, 2005; accepted August 23, 2005)

814

Analogous reactions can be written for long chain hydrocarbons. Carbon monoxide, CO, is a byproduct of reactions (2) and (3). In order to avoid Pt deactivation at the PEFC anode, CO concentration in the fuel processor must be below 10 ppm, and this is partially achieved via the water gas shift reaction (4): CO 1 H2 O ! CO2 1 H2

(4)

The direct decomposition reaction (1) is endothermic and requires an external supply of energy; furthermore, it results in undesirable coke formation. Steam reforming (2) is also strongly endothermic and, as above, requires heat transfer to the reactor, making this reaction less attractive in an application that requires rapid start and fast dynamic response.1 Partial oxidation (3) is exothermic, and can raise the gas

Microstructural Evolution of Nickel Nanoparticle Catalysts Supported on Gadolinium-Doped Ceria during Autothermal Reforming of Iso-Octane

temperature to values where it becomes feasible to steam reform additional fuel. Hence, autothermal reforming—the combination of steam reforming and partial oxidation—of hydrocarbons, whereby the heat generated by (3) is absorbed by reaction (2), is the preferred route for hydrogen production.1 The group VIII metals supported on metal or rare-earth oxides are good catalysts for reforming reactions, but one major problem is catalyst deactivation due to carbon formation, especially for Nibased catalysts.2 Although noble metals (Pt, Pd, Rh) suffer less coking problems than Ni, their high cost and limited supply render their use impractical. It is therefore desirable to develop Ni catalysts on supported oxides with good resistance to coking. In general, a low metal loading favors a high degree of metal dispersion on the support, which reduces coke formation.2,3 Among the oxide-supported Ni catalysts, Ni/g-Al2O3 and Ni/CeO2-Al2O3 have recently been the subject of several investigations. Ceria additions, in particular, have been found to improve the catalytic properties of Ni catalysts.4–6 It is believed that the CeO2 support increases the catalytic activity of Ni by increasing the diffusivity of lattice oxygen.1 Doping the ceria lattice with Gd (characterized by a 13 oxidation state) generates oxygen vacancies due to the necessity of maintaining charge neutrality in the solid.7 These oxygen vacancies can further increase the diffusivity of lattice oxygen potentially increasing catalytic activity even further. It has also been reported that the addition of alkaline and lanthanide oxides greatly enhances the catalyst’s stability and its ability to resist carbon deposition.8 The abundant availability of gasoline, liquefied petroleum gas, and diesel makes them ideal fuels for hydrogen production, but little research has been conducted on the direct decomposition of higher hydrocarbons because of the increased possibility of coke formation. Recently, Jenkins and Shutt9 reported the conversion of higher hydrocarbons to hydrogen by using precious metal catalysts. In this case, however, coke formation over the catalyst was observed. In our present study, we focus on microstructural investigations of Ni/Ce0.8Gd0.2O1.9 catalysts with very low metal loading (1 mol.%, or 0.33 wt.%), for use in autothermal reforming of isooctane C8H18. A combined investigation of the catalyst microstructure, the crystallographic structure of the catalyst components, and the chemical states of the elements prior to and after the catalytic reaction can give important information regarding the mechanism responsible for the catalytic activity and long-term stability of the catalyst. EXPERIMENTAL PROCEDURE Catalyst Preparation and Testing The catalyst was prepared at the Argonne National Laboratory (ANL), Chemical Technology Division. The catalyst powder (Ce0.8Gd0.2O1.9, sur-

815

face area 38.24 m2 /g) was acquired from Praxair (Woodinville, WA). This was mixed with 5 wt.% stearic acid, 0.5 wt.% microcrystalline cellulose, and water to form an extrusion paste, which was successively mixed with a solution of Ni(NO3)2 calculated to yield a total metal loading in the final catalyst of 1 mol.% (0.33 wt.%Ni). The paste was then extruded using a syringe to obtain pellets with a diameter of 0.8 mm. The pellets were dried at 100°C for 4 h, heated to 383°C for 2 h to allow the stearic acid to decompose, and then sintered at 900°C for 2 h.10 Studies of catalytic activity were conducted at the Chemical Technology Division, ANL. The Ni catalyst powder (3.4 g) was tested for autothermal reforming of 2,2,4-trimethylpentane (or iso-octane) in a stainless steel reactor with a diameter of 1.25 cm. Liquid iso-octane (0.08 mL/min.) and water (0.08 mL/min.) were vaporized in a heated silicone oil bath maintained at 250°C. A small flow of N2 (20 mL/min.) was used to purge the vaporized isooctane/water mixture, which was further mixed with oxygen (43.5 mL/min.). Based on the feed rates, the O2 to C ratio was 0.46 and the H2O to C ratio was 1.14. The catalyst was tested over a temperature range of 500–800°C for 40 h and the exit gases were analyzed using gas chromatography. The hydrogen concentration of the exit gases ranged from 55% to 60% on a dry N2 basis. The conversion of iso-octane ranged from 80% at 500°C to 100% at temperatures above 650°C. No detectable decrease in the catalytic activity was observed over the duration of the experiment. Two types of catalyst samples were characterized in this study: a fresh catalyst sample that was examined directly after preparation and a used catalyst sample that was analyzed after exposure in the reactor for 40 h. Analytical and Microscopy Methods X-ray diffraction (XRD) data were collected using a Rigaku D-max 2200 x-ray diffractometer (The Woodlands, TX) with Ni-filtered Cu Ka radiation. The copper tube in the x-ray generator was operated at 40 kV and 10 mA. Specimens for transmission electron microscopy (TEM) investigations were obtained by preparing a colloidal suspension of crushed catalyst powder in ethanol. The suspended particles were then scooped and placed on a carbon-coated TEM grid. The catalyst samples thus obtained were thin enough to be electron transparent and could consequently be used for TEM studies. The specimens were examined with a 200 keV Hitachi H-8000 instrument (Pleasanton, CA) equipped with energy dispersive x-ray spectroscopy (EDX) capability. Samples of catalyst powders were pressed onto double-sided conductive tape for x-ray photoelectron spectroscopy (XPS) analysis. Core level XPS spectra were obtained using a Kratos Axis 165 spectrometer (Chestnut Ridge, NY) equipped with a monochromatic Al Ka excitation source (1486.6 eV) operated at constant pass energy of 20 eV. A built-in charge

816

Palaniyandi, Shamsuzzoha, Ada, Zangari, and Reddy

neutralizer was used to compensate for sample charge buildup during data acquisition. The binding energies of the XPS core lines were referenced to the Ce41 u’’’ peak fixed at 916.7 eV.11 RESULTS X-Ray Diffractometry of the Catalysts The XRD spectra from the two catalysts (Fig. 1) showed only diffraction peaks characteristic of the CeO2 face-centered-cubic (fcc) fluorite-type structure.12 The absence of peaks from the hexagonal Gd2O3 phase suggests that all the Gd dopant atoms substitute for Ce in the CeO2 lattice. The absence of Ni peaks further indicates that the Ni is negligibly small in the volume fraction of the investigated x-ray sample and did not yield any detectable diffraction maxima. The Ni present in the sample is highly dispersed on the surface of the support and is in the form of nanocrystalline particles, as shown by TEM analysis (refer to the following section). The high-angle diffraction peaks for the used catalyst are shifted toward lower angles with respect to corresponding peaks for the fresh sample, indicating an expansion of the lattice parameter a of the fcc ceria phase, as suggested elsewhere.13 The lattice parameters determined by graphical extrapolation ˚ and 5.442 A ˚ for the fresh and used cataare 5.432 A lyst support, respectively. These values are larger ˚ ) for highthan the lattice parameter (5.41134 A 12 purity CeO2 phase. This may indicate that the ceria phase in both samples contains a significant amount of solute atoms of ionic radius larger than that of Ce14, thus allowing the lattice parameter to expand. A general decrease in the x-ray scattering intensity was observed from the used catalyst,

Fig. 1. X-ray diffraction patterns of fresh and used Ni/Ce0.8Gd0.2O1.9 catalyst.

which is probably due to carbon deposition on the catalyst surface. However, no changes in the width of the diffraction peaks were observed, suggesting limited, if any, grain growth of the crystallites. TEM: Crystal Morphology and Structural Studies of the Catalysts The TEM bright-field images at low magnification of the fresh catalyst (Fig. 2a) show that the catalyst consists of homogenous aggregates of constituent grains with size ranging between 20 nm and 50 nm. A selected area diffraction (SAD) pattern of a single grain (Fig. 2b) shows an fcc structure with a 5 ˚ . Taking into account possible errors due to 5.44 A the finite width of the diffraction spots, this value is in fair agreement with the values found from the XRD spectrum of the fresh catalyst. The distribution of the elements was probed with a spatial resolution of about 10 nm by collecting EDX spectra using an electron beam of a few nanometers in diameter. Characteristic peaks due to Ce, Gd, and O were observed from all the grains, and their composition was homogeneous within the sensitivity of the EDX detector. No evidence of Gd oxide grains was observed, confirming the hypothesis that the catalyst support consists indeed of a mixed oxide. The EDX spectra collected over areas with diameters ;50 mm frequently showed Ni peaks, confirming the presence of Ni in a relevant fraction of the support grains. In order to determine the precise locations of the Ni nanoparticles in the catalyst, selected area EDX spectra were taken with a 10-nm-diameter electron beam at numerous locations. Several regions where the Ni signal was much higher than the background noise were found. Figure 3 presents the bright-field image and the SAD pattern from one such region. In the bright-field image of Fig. 3a, a rectangular-shaped Ni particle with dimensions of about 30 nm (dark in the image) can be seen to be in contact with a lightly contrasted oxide particle. Figure 3b shows diffraction spots characteristic of metallic Ni (indicated by the arrows and indexed) superposed with diffraction spots due to the mixed oxide. No lattice correspondence between low index planes of Ni and the oxide is observed, demonstrating that no epitaxial relationship develops between the Ni particle and the supporting oxide. Microstructural TEM investigations of the used catalyst revealed a different morphology of the support, with a bimodal grain size distribution. In particular, most of the grains were clearly faceted and identical in size to those in the fresh catalyst, while a few nonfaceted grains of larger sizes were also present. Good TEM image contrast could be obtained from the faceted grains, indicative of a well-developed crystalline structure. The structure and composition of these grains were identical to those of the support in the fresh catalyst, as demonstrated by SAD patterns in conjunction with EDX spectra from these regions. In a few cases, a small number of such


817

Fig. 2. (a) Bright-field TEM micrograph of the typical microstructure of the fresh Ni catalyst. (b) The nano-probe electron diffraction pattern taken with an electron beam parallel to [110] of the crystal constituting a typical grain in (a).

grains were sintered together and covered by amorphous material, indicated by an arrow in Fig. 4a. The SAD patterns (Fig. 4b) from this sintered grain reveal the splitting of diffraction spots (for example, the spot indicated by an arrow in Fig. 4b) due to the presence of small angle grain boundaries. Analysis of these diffraction patterns revealed that the fcc cell parameter increased by about 0.2% with respect to the fresh catalyst support in agreement with the XRD data. It should be noted that, due to the error introduced by the width of the diffraction spots, the measured increase of the cell parameter in the used catalyst by TEM is expected to be qualitative at best. The larger size grains constitute a small fraction of the catalyst and are present as isolated regions in the interconnected matrix of nonfaceted grains. These grains exhibit different crystalline structures. Though they all contain Ce, Gd, and O, their

O content is lower than that of the mixed oxide support in the fresh catalyst. The SAD patterns from these transformed grains, in general, yielded broad diffraction spots typical of poorly crystalline, highly defective structures. Two distinct crystalline structures were identified based upon the analyses of the diffraction patterns of single grains along various zone axes as well as of the selected area EDX spectra. The first structure (Fig. 5a and b) is hexagonal ˚ and c 5 6.65 A ˚. with lattice parameters a 5 3.95 A The lattice parameters and crystal system of this phase agree well with those reported for Ce2O3.14 Grains belonging to this crystal structure were found to contain a substantial number of Gd atoms in solid solution, as determined by EDX. The second structure (Fig. 6a) yielded relatively sharper electron diffraction patterns (Fig. 6b), has a cubic crystal system, and has a primitive Bravais lattice with

Fig. 3. (a) Bright-field TEM micrograph showing a 30-nm sized nickel particle on the surface of a ceria grain present in the fresh Ni catalyst. (b) Selected area electron diffraction pattern taken from the rectangular particle in (a). Diffraction spots belonging to cerium oxide crystal are indexed. Diffraction spots marked by arrows are due to Ni crystal.

818


Fig. 4. (a) Bright-field TEM micrograph showing a sintered aggregation of grains found in the microstructure of the used Ni catalyst. (b) Selected area diffraction taken with an electron beam parallel to [110] of the crystals constituting the sintered grain in (a). The splitting of diffraction spots is due to small-angle misorientation (shown by an arrow) between various grains of the aggregate.

˚ , which agrees well with the reported a 5 5.11 A structure of the CeO phase.15 Selected area EDX data also confirm that some Gd atoms are present in solid solution in this phase. Poor contrast in the TEM images from the larger grains of used catalyst, probably due to the extensive coverage of the support particles by an amorphous film (shown by an arrow in Fig. 6a), hindered the direct observation of Ni nanoparticles. The Ni peaks were observed in the EDX spectra taken from the grains of mainly hexagonal form of oxide phase, but it was not possible to precisely identify whether these peaks were originated from Ni/Ni oxide crystals or from grains of the mixed oxide support with Ni as solute atoms. In some Ni-rich regions, the resulting diffraction patterns could be indexed according to a fcc structure with a lattice parameter ˚ , consistent with that of the NiO structure. of 4.44 A It is not clear, however, to what extent the metallic Ni observed in the fresh catalyst was oxidized to NiO during the catalytic reaction.

XPS: Surface Spectroscopy of the Catalysts Table I shows the surface atomic composition of the fresh and used catalyst, as determined by quantitative XPS analysis. Due to low Ni loading, the Ni content is below the detection limit of XPS. The carbon content is much higher in the used catalyst, demonstrating extensive coke formation on the catalyst surface, which causes an overall decrease of the signal from the other elements. The increase in the total surface O concentration, however, is attributed to an increase in CO3"2/OH" groups on the catalyst surface, as discussed subsequently. Figure 7 shows high-resolution scans of the C1s region. The peak observed at 285 eV is assigned to the hydrocarbon CHx component, while the high binding energy peak at 288 eV is assigned to the carbonate, CO3"2 species. The relative intensity of the CO3"2 peak is higher in the used catalyst than in the fresh sample. The O1s region in Fig. 8 shows that contributions from the oxygen in the oxide lattice

Fig. 5 (a) Bright-field TEM micrograph of the used Ni catalyst, showing a new phase appearing as a lightly contrasted mottled grain marked by an arrow at the center of the micrograph. (b) Selected area diffraction pattern taken with an electron beam parallel to [0001] of the crystals constituting the marked grain in (a).


819

Fig. 6. (a) Bright-field TEM micrograph of the used Ni catalyst showing the grain, indicated by an arrow, of a new phase. (b) Selected area diffraction pattern taken with an electron beam parallel to (001) of the crystal constituting the marked grain in Fig. 6a.

(;529 eV), carbonates and/or hydroxides (;532 eV), and adsorbed water (;534 eV)16 can be assigned. A significantly stronger contribution from CO3"2/ OH" components in the O1s spectrum of the used catalyst is observed. The high-resolution XPS scan of the Ce3d region is reported in Fig. 9. The peaks labeled u(n) and v(n) are associated, respectively, with the 3d5/2 and 3d3/2 spin orbit pairs.11 In particular, the peaks labeled v, v’’, and v’’’ (for Ce3d5/2) and the peaks labeled u, u’’, and u’’’ (for Ce3d3/2) correspond to the Ce41 oxidation state, while the peaks v’ and u’ correspond to contributions from the Ce31 oxidation state. A relatively higher intensity of the v’ and u’ (Ce31 ) components observed in the used catalyst indicates the partial reduction of Ce as a consequence of the catalytic reaction. Calculation of the relative Ce31 state contributions in the two samples yielded values of 5% and 26% for the fresh and used samples, respectively. DISCUSSION Ni/CeO2 catalysts have been observed in the past to exhibit high activity and selectivity for the partial oxidation and/or reforming of light hydrocarbons.6,16 The activity of these catalysts however rapidly decreases as a consequence of carbon deposition, which can be avoided by low loading and high dispersion of the Ni.13 The low metal loading of the catalyst investigated here induces a very good dispersion of the metal to such an extent that it was very difficult to identify Ni particles in the fresh catalyst, and it was nearly impossible in the used

catalyst. As a consequence of the high degree of dispersion, the activity of the catalyst did not decrease with time for the duration of the experiment (40 h). Both XRD and TEM data seem to indicate that most of the mixed oxide support not only retained the structure of fcc ceria phase under the reforming conditions but also showed an increase in lattice parameter. In the crystal structure of CeO2, each Ce41 ion keeps an eightfold cation to anion coordi˚ .17 On nation and assumes an ionic radius of 0.97 A 31 the other hand, Gd ion has a comparatively large ˚ ) and is also known to possess ionic radius (1.05 A only an eightfold cation to anion coordination in any crystal structure.17 Since no significant amount of elements other than Ce, Gd, and O were found to be present in the structure of CeO2, only a substitutional solid solution of Gd31 in the Ce41 lattices can account for the observed expansion in the lattice parameter of fresh oxide support. Such substitutions are likely to produce some charge inequality between cation and anion in the structure. This can lead to release of some oxygen atoms from the structure of the oxide in order to preserve its charge neutrality thereby producing oxygen vacancies. An increase of 0.2% in the lattice parameter of the oxide support in the used catalyst can be attributed to a similar process caused by the catalytic reactions that occur during the reforming process. The higher percentage of Ce31 features observed in the Ce3d XPS lines of the used catalyst indicates a significant transformation of Ce41 to Ce31 ions that have a compara˚ ). The resulting tively larger ionic radius (1.14 A

Table I. Surface Atomic Composition of Fresh and Used Ni Catalysts from XPS Analysis Samples Fresh catalyst Used catalyst ND—not detected.

Ni (at.%)

Gd (at.%)

Ce (at.%)

O (at.%)

C (at.%)

ND ND

2.7 2.2

24.0 15.2

58.9 62.1

14.4 20.5

820


Fig. 7. C1s XP spectra of fresh and used Ni catalyst.

increase in the concentration of larger Ce31 ions in the structure of oxide support can account for the increase in the lattice parameter. The increment of Ce31 ions in the fcc lattices of CeO2 is likely to produce a further charge inequality between cations and anions, which can lead to oxygen vacancies. Structural investigation by TEM also revealed that some coarsening of grains as well as phase transformation has taken place in a small volume fraction of the oxide support. Three distinct crystalline structures, exhibiting highly defective structures, have been observed. All these structures can be identified as known Ce oxide phases—in all cases doped with Gd—of lower valance than the initial CeO2 structure. The oxide support structures found in the used catalyst are Ce0.8Gd0.2O2"x, Ce1.6Gd0.4O3"y, and Ce0.8Gd00.2O1"z, where x, y, and z are the occupation of oxygen vacancies in the lattices CeO2, Ce2O3, and CeO structures, respectively. This is based on the assumption that the ratio of Ce/Gd is maintained during the transformation and that the presence of Gd within the structure results in oxygen vacancies. Existence of the poorly crystalline and semicrystalline grains of the defective Ce2O3 phase in the reformed oxide support provides some opportunities to discuss the related catalytic reaction taking place in the reforming process. In the Ce1.6Gd0.4O3"y phase, most of the cation lattices are occupied by Ce31 ions, whereas the parent phase Ce0.8Gd0.2O2"w (where w is the occupancy of oxygen

Fig. 8. O 1s XP spectra of fresh and used Ni catalyst.

Fig. 9. Ce 3d XP spectra of fresh and used Ni catalyst.

vacancy in the fresh oxide support) contains mostly Ce41 cations. These structural features suggest that the catalytic reactions responsible for the formation of Ce1.6Gd0.4O3"y may have resulted in the transformation of a large number of Ce41 to Ce31 cations, which were observed by our XPS analysis. The present observation that some sintered grains of ceria having a relatively large lattice parameter compared to that present in the fresh oxide lends further support to the proposed mechanism of Ce41 to Ce31 phase transition. This implies that the suggested cationic transformation induced by catalytic reactions initially results in sintering grains. These sintered grains upon progressive enrichment of Ce31 ions as well as oxygen vacancies assume a state at which the fcc crystal structure of CeO2 no longer remains stable but transforms to a more stable hexagonal structure of Ce2O3. Such a phase transition is likely to result in large displacement of atoms in the lattices and to result in poor crystalline and semicrystalline grains for the transformed phase. This conforms well with the experimental TEM observations of such grains in the reformed oxide support. The CsCl type Ce0.8Gd0.2O1"z phase found in the used catalyst precludes a similar cationic transformation solely for its development. All of the Ce atoms present in this phase exist only in Ce21 ions, and the phase has a structure with an eightfold cation to anion coordination. It seems that this phase could be the by-product of CeO2 to Ce2O3 phase transformation. During this phase transformation, Ce31 ions from the progressively Ce31 enriched parent Ce0.8Gd0.2O2"w phase construct the lattices of the hexagonal Ce1.6Gd0.4O3"y phase. The Ce41 ions along with the O2" ions form associated eightfold anion to cation polyhedral in the parent phase and adopt the CsCl-type structure for the Ce0.8Gd0.2O1"z phase. The Ce ions in the newly formed phase can then assume Ce21 ionic state in order to produce appropriate charge equality between anion and cation. It is reported18 that cerium oxide acts as a promoter for the oxidation of CO and light hydrocarbons by


fostering the transfer of oxygen from the ceria support to the metal (Ni in this case), where oxidation of any carbon-containing species can easily proceed. The suggested charge inequality between cation and anion due to substitution of either Gd31 or Ce31 in the Ce41 lattices of the oxide support is also capable of ionizing Ni catalyst that is affixed to the surface of the support cerium oxide crystal. The Ni catalysts ionized by this process can readily accept oxygen atoms released from cerium oxide and thereby foster the catalytic process. This transfer of oxygen is usually reversible under controlled conditions with CeO2 losing oxygen in a reducing environment and acquiring it back in an oxidizing environment. The current study shows, however, that oxygen transfer is not reversible under the reaction conditions investigated. On the contrary, a small fraction of the grains comprising the supporting oxide lose oxygen atoms to such an extent that the initial structure becomes unstable and transforms to other Ce(Gd) oxide structures with a smaller cation to anion occupancy ratio. Such transformation has proceeded only to a small extent in the 40 h time frame of this experiment. In fact, no changes in the structure of the catalyst can be detected by XRD. However, continued use of the catalyst should eventually induce total transformation of the CeO2 structure to the other observed structures, with concurrent incapability of the support to perform its function due to the ceased capacity of the Ce atoms to reversibly transform between the 41 4 31 states. CONCLUSIONS Spatially resolved microstructural, structural, and analytical studies of 0.33 wt.%Ni/Ce0.8(Gd)0.2O1.9 were performed prior to and after the catalytic autothermal reforming of iso-octane. A high degree of metal dispersion is observed in the fresh catalyst, which contributes to a stable conversion during a reaction carried out for 40 h. After exposure in the reactor, a small fraction of the support is subject to phase transformation as a consequence of the irreversible loss of oxygen. The structures observed are

821

characterized by a lower cation/anion ratio. It is argued that, while a 40 h experiment is insufficient to detect any decrease in the activity of the catalyst, prolonged use would cause an irreversible deactivation associated with phase transformations in the oxide support. ACKNOWLEDGEMENTS The authors are thankful to ANL (catalyst samples were prepared by Laura Miller and tested by Rolf Wilkenhoener, both from ANL). This work was partially funded by a grant from the Fuel Cell Program at ANL, which is supported by the U.S. Department of Energy’s Office of Advanced Automotive Technologies. REFERENCES 1. S. Ahmed and M. Krumpelt, Int. J. Hydrogen Energy 26, 291 (2001). 2. C.H. Bartholomew, Catal. Rev.–Sci. Eng. 24, 67 (1982). 3. M. Ozawa and C.K. Loong, Catal. Today 50, 329 (1999). 4. J.M. Herrmann, E. Ramaroson, J.F. Tempere, and M.F. Guilleux, Appl. Catal. 53, 117 (1989). 5. S. Wang and G.Q. (Max) Lu, Appl. Catal., B 19, 267 (1998). 6. M.A. Pena, J.P. Gomez, and J.L.G. Fierro, Appl. Catal. A 144, 7 (1996). 7. P. Albers, K. Deller, B.M. Despeyroux, A, Schafer, and K. Seibold, J. Catal. 133, 467 (1992). 8. F.L. Normand, L. Hilaire, K. Kili, G. Krilland, and G. Maire, J. Phys. Chem. 131, 74 (1994). 9. J.W. Jenkins and E. Shutt, Plat. Met. Rev. 33, 118 (1989). 10. M. Krumpelt, S. Ahmed, R. Kumar, and R. Doshi, U.S. patent 6,110,861 (2001). 11. P. Burroughs, A. Hamnett, A.F. Orchard, and G. Thornton, J. Chem. Soc., Dalton Trans. 1686 (1976). 12. International Center for Diffraction Data, Powder Diffraction file 1-43, 34-394. 13. B.K. Cho, B.M. Sanks, and J.E. Bailey, J. Catal. 115, 236 (1989). 14. International Center for Diffraction Data, Powder Diffraction file 1-43, 23-1048. 15. International Center for Diffraction Data, Powder Diffraction file 1-43, 33-334. 16. Q. Miao, G. Xiong, S. Sheng, W. Cui, L. Xu, and X. Guo, Appl. Catal., A 154, 17 (1997). 17. D.R. Lide, CRC Handbook of Chemistry and Physics, 74th ed. (Boca Raton, FL: CRC Press, Inc.), pp. 12-8 and 9. 18. J. Barrault, A. Alouche, V. Paul-Boncour, B L. Hilaire, and A. Percheron-Guegan. Appl. Catal. 46, 269 (1989).