Report

Control of Metal Nanocrystal Size Reveals Metal-Support Interface Role for Ceria Catalysts

See allHide authors and affiliations

Science  16 Aug 2013:
Vol. 341, Issue 6147, pp. 771-773
DOI: 10.1126/science.1240148

A Measure of Metal-Oxide Interfaces

The rate of a catalytic reaction can sometimes be enhanced by using a different metal oxide as the support for adsorbed metal nanoparticles. Such enhancement is often attributed to more active sites at the metal-oxide interface, but it can be difficult to quantify this effect. Cargnello et al. (p. 771, published online 18 July) synthesized monodisperse nanoparticles of nickel, platinum, and palladium and dispersed them on high-surface-area ceria or alumina supports. High-resolution transmission electron microscopy enabled a detailed analysis of interfacial site structure, which showed that the rate of CO oxidation on ceria was indeed enhanced greatly at interface sites.

Abstract

Interactions between ceria (CeO2) and supported metals greatly enhance rates for a number of important reactions. However, direct relationships between structure and function in these catalysts have been difficult to extract because the samples studied either were heterogeneous or were model systems dissimilar to working catalysts. We report rate measurements on samples in which the length of the ceria-metal interface was tailored by the use of monodisperse nickel, palladium, and platinum nanocrystals. We found that carbon monoxide oxidation in ceria-based catalysts is greatly enhanced at the ceria-metal interface sites for a range of group VIII metal catalysts, clarifying the pivotal role played by the support.

The properties of heterogeneous catalysts are often determined by the synergy between support (typically metal oxides) and supported phases (typically metal nanoparticles). Ceria (CeO2) is an example of an “active support” that can greatly increase rates for reactions involving redox steps, such as CO oxidation and the water-gas shift (WGS) reaction (1, 2), by comparison to “inert,” nonreducible supports such as alumina (3, 4). The observed enhancement is assumed to result from active sites at the metal-ceria interface, because rates can be much greater than the sum of the rates over ceria and the metal individually (1). Evidence that the oxygen atoms migrate from the support to the metal particles has come only from model systems (5, 6) not operating under industrially relevant reaction conditions. Understanding size-activity relations for ceria-based catalysts is important for improving catalyst performance. The turnover rate for CO oxidation is thought to be independent of metal particle size (7, 8), so this reaction is an ideal probe for studying the role that the metal-support interface plays by measuring changes in rates upon varying the concentration of interfacial sites. Chemical methods for preparing metal nanoparticles on high-surface-area supports typically lead to large or asymmetric metal particle size distributions, which prevent definitive correlation between particle size and activity (for example, subsets of particles could be completely inactive or disproportionately more active). Monodisperse metal particles tested under realistic reaction conditions are critical for understanding the relation between catalytic activity and specific particle size (9, 10).

Here, we used monodisperse, size-tunable metal nanocrystals (NCs) of Ni, Pd, and Pt to demonstrate the role of the metal-support interface in ceria-based systems. The relative fraction of interfacial sites was varied for both ceria and alumina supports, and the role of ceria in enhancing CO oxidation rates under realistic conditions was revealed (scheme S1) (11).

We prepared monodisperse Ni, Pd, and Pt NCs by thermally decomposing metal(II) acetylacetonates in a benzyl ether solution in the presence of oleylamine (OLAM), trioctylphosphine (TOP), and, in some samples, oleic acid (OLAC) (table S1) (11). By varying the surfactant concentration and reaction temperature, various NC sizes for each metal (small, medium, and large) were obtained. Figure S1 shows transmission electron microscopy (TEM) images of Ni (4 to 12 nm), Pd (2.5 to 6.3 nm), and Pt NCs (1.6 to 2.9 nm) that were quantitatively obtained with particle size distributions below 6%, without any postsynthetic size-selective precipitation processes. The uniformity in size of the NCs was confirmed by small-angle x-ray scattering (fig. S2) (11) and by the fact that they formed large areas of three-dimensional (3D) hexagonal close-packed superlattices with single domains exceeding several micrometers (12, 13) (Fig. 1). High-resolution TEM (HRTEM) studies (Fig. 1, G to I) provided evidence of the overall crystallinity of the samples, although x-ray diffraction patterns showed that the presence of defects (visible in, e.g., Fig. 1G) broadened the diffraction peaks (fig. S2) (11), in agreement with previous studies (14).

Fig. 1 As-prepared nanocrystals.

(A to C) TEM images of hexagonal close-packed 3D assemblies of (A) Ni NCs, (B) Pd NCs, and (C) Pt NCs. (D to F) Magnified images. (G to I) High-resolution TEM images showing distinct NCs.

The monodisperse Ni, Pd, and Pt NCs were adsorbed from toluene solutions onto both alumina and ceria supports, and heating the materials in air at 300°C completely removed the organic capping agents. Low metal loadings (0.5 weight percent) and high-surface-area supports (~100 m2 g−1 for alumina and ~60 m2 g−1 for ceria) mimicked real catalyst formulations and ensured that the NCs were well separated and resistant to particle sintering. We examined these samples by means of TEM and CO chemisorption. Because the high electron density of ceria makes the determination of size distributions by TEM particularly difficult (especially in the case of Ni and Pd) (15), the particle sizes and distributions were initially determined by analyzing the alumina-based systems (fig. S3). We confirmed that particle sizes and distributions obtained on the alumina samples were also representative of the ceria-based counterparts through TEM analysis of the Pt/CeO2 samples (fig. S3, L to N). Strong Z-contrast between Pt and ceria in high-angle annular dark-field scanning TEM (HAADF-STEM) images let us confirm that the particle size and shape did not change upon deposition and calcinations of the particles on this support. The Z-contrast for the Ni/CeO2 and Pd/CeO2 samples was less strong, and thus we also used electron energy loss spectroscopy (EELS) to map the individual Ni and Pd NCs (fig. S4) (11) to measure size distributions. For Pd and Pt catalysts, there was little change in the NC sizes after deposition and calcination (fig. S5) (11). For Ni, we observed the formation of hollow spheres, likely caused by the Kirkendall effect (fig. S3, A to C) (11, 16). Nonetheless, even in this case, the very narrow size dispersion was maintained and there was no Ni metal loss during this process (fig. S5) (11). Furthermore, it is apparent from the particle size distribution measurements [histograms in fig. S5 (11)] that the particles supported on ceria had a slightly wider size distribution: The distributions for the small, medium, and large sizes were completely separated for the Pt/Al2O3, whereas there was a slight overlap in the distributions for the Pt/CeO2. There was no overlap in the particle size distributions for the Pd and Ni samples. We also conducted environmental TEM (ETEM) experiments by heating the samples in situ in air to 300°C under conditions otherwise similar to the calcination process (fig. S6) (11). The images show restructuring at the metal-ceria interface with the particles adhering to the ceria surface, but this neither changes the overall particle size nor the particle shape used for the calculation of the fraction of particular sites (see below). We also took into account the slight variability of the metal-ceria interface in our calculations by performing careful HRTEM studies (fig. S7) (11). The CO chemisorption experiments provided information on the total population of accessible metal sites (table S2) (11) and confirmed that the trends in metal particle size are retained after deposition and calcination. All of the above data confirm that particle sizes and distributions are maintained in the final catalysts (Fig. 2). HRTEM analysis indicates that the larger particles maintained their original cuboctahedral morphology, and suggests that smaller particles spread over both supports into shapes that resemble a cubo-octahedron truncated along the {100} direction.

Fig. 2 Heat-treated nanocrystals.

(A to C) HRTEM images of Pd/CeO2 catalysts after calcination at 300°C and reduction at 150°C: small (A), medium (B), and large (C) samples. (D) Physical models prepared to describe the particles. Blue, orange, and gray colors indicate corner, perimeter, and surface atoms, respectively; red and white are oxygen and cerium atoms of the ceria support.

The data obtained by conventional and aberration-corrected TEM and by CO chemisorption were used to prepare a physical model of the particles. In the case of Pd, the modeled particles are shown in Fig. 2D. The models were used to quantify the number of atoms with particular coordination environments, such as corner and perimeter sites at the interface with CeO2 and surface atoms that are not in direct contact with the support. Previous reports made use of the entire size distribution to develop a physical model of the particles (17, 18), but in our case, extremely narrow distribution of NC sizes and shapes allowed the use of average values.

The different particle sizes and shapes differ in terms of length of the metal-support interface, so we could directly analyze the impact of this parameter for CO oxidation. For the catalytic tests, the metallic phase of the NCs was ensured by a mild prereduction (fig. S8) (11), and TEM characterization of the catalysts after reaction did not show any change in size or shape of the particles. To ensure that neither mass nor thermal diffusion limitations affected the results, we used high space velocities and diluted each catalyst with inert support materials (8, 11, 19) (fig. S9). Reaction orders for CO and O2 were measured on the series of small Pt samples that gave the highest volumetric CO oxidation rates (fig. S10) (11). On the alumina-supported Pt, in excess CO, the reaction orders were ~ –1 in CO and ~1 in O2, in agreement with the previous values (3); hence, O2 activation is inhibited by adsorbed CO on the metal particle surface (8). In the case of the ceria sample, the reaction orders were ~0 in CO and slightly positive in O2, implying that a second reaction mechanism must be active (20). The results of CO oxidation on ceria- and alumina-supported metals under lean conditions (excess oxygen; see supplementary materials) are reported as kinetic plots in Fig. 3 and light-off curves in fig. S11 (11).

Fig. 3 Kinetic data.

(A and B) Arrhenius-type plots for CO oxidation over (A) Al2O3 and (B) CeO2 samples, where a difference in the 1000/T scale should be noted.

The metals deposited on ceria had higher catalytic rates than their alumina-supported counterparts, as evidenced by the much lower temperatures needed to completely oxidize CO (fig. S11) (11). The apparent activation energies (Ea) for the ceria-based catalysts (fig. S12) (11) were in the range of 40 to 70 kJ mol−1. Alumina-supported samples showed higher apparent Ea values of 50 to 150 kJ mol−1. These values are in agreement with results from other studies (3). Notably, we found similar activation energies for all ceria-supported catalysts, implying that a similar mechanism must be operative regardless of the metal.

The alumina-based catalysts exhibited rates that were independent of metal particle size when normalized to the metal surface area (Fig. 3A), as determined by CO chemisorption (8). However, the ceria-based catalysts displayed a strong size-dependent activity, with normalized reaction rates decreasing with increasing NCs size for all three metals studied (Fig. 3B) (11). The alumina-supported NCs were essentially saturated by CO, which likely limited any effects of these intrinsically different sites on the alumina-supported metals. By contrast, the zero-order rate in CO observed for ceria-based catalysts is a result of reaction between CO adsorbed on the metal and O2 provided by the ceria, so that the CO on the metal is unable to suppress the rate of O2 adsorption onto ceria (21, 22). Despite the large number of elegant theoretical and experimental studies of oxygen spillover from ceria to Pt on model single crystals under low pressures (6), no reports have overcome the so-called “material and pressure gap” (23, 24) by experimentally demonstrating—under realistic working conditions—the involvement of ceria lattice oxygen in the oxidation of CO, and thus addressing the role of the metal-ceria interface.

HRTEM allows us to build a model of the shape of the supported NCs as they form faceted solids on the support interface. To determine whether the important variable that is altered by particle size is the surface-to-volume ratio or the perimeter-to-surface ratio in the metal NCs, we used the model described above to calculate the fraction of atoms located at the various surface sites (surface atoms not in contact with ceria, and perimeter or corner atoms at the metal-support interface; see Fig. 2D) (17, 18, 25, 26). We analyzed the scaling relation in this framework with the use of the particle shapes obtained by HRTEM (Fig. 2). For any regular solid other than a sphere, the number of surface sites per volume is proportional to the diameter (d) as ~d−1, that of the edge sites to ~d−2, and that of the vertices to ~d−3. For this reason, the model was robust, in that the relations did not drastically change when particles of slightly different geometries were used. We then plotted in the same graph the fraction of sites with a particular position as a function of particle size for all nine ceria-based samples (Fig. 4). The slight scatter in the graph arose because we compared metals with dissimilar lattice constants and slightly different shapes. These results showed a scaling of d–0.9 ± 0.1 for the surface atoms, and of d–1.9 ± 0.2 and d–2.6 ± 0.1 for perimeter and corner atoms in direct contact with the support, respectively. We then collected the turnover frequency (TOF) values of the CeO2-based catalysts at 80°C (a convenient temperature to test all the catalysts under kinetic conditions) and plotted the data on the same graph. The TOFs for all nine samples showed a dependence of the diameter as d–2.3 ± 0.2, implying that the metal atoms at the nexus of the metal, support, and atmosphere were the active sites for this reaction and that the larger surface-to-volume ratio of small particles translates to an increased boundary length and higher activity. The value of the slope implies that the corner atoms were the most active sites overall, most likely because of their lower coordination number compared to the other perimeter atoms, as observed in other systems (17, 18). The slight deviation from the expected trend for the small and medium Pt samples might be an effect from some very small particles that were detected by high-resolution STEM (fig. S3) (11) that contributed more to the observed reactivity. Nonetheless, we conclude that the perimeter atoms were the active sites for CO oxidation on ceria-based catalysts. It may be fortuitous, but the TOF for our Pt small sample (~0.2 s−1 at 80°C) is similar to that reported for single-site Pt/FeOx catalysts (0.3 s−1 at the same temperature) (27). Despite the different nature of the systems, this further corroborates the validity of our approach.

Fig. 4 Model analysis.

Calculated number of sites with a particular geometry (surface and perimeter or corner atoms in contact with the support) as a function of diameter and TOF at 80°C of the nine ceria-based samples.

The trend in catalytic activity (size dependence) was not influenced by the reaction environment. Similar results were obtained from stoichiometric, lean (excess of oxygen), or rich (excess of CO) conditions (figs. S13 and S14) (11). The apparent activation energies are in the range of 40 to 70 kJ mol−1 for all the samples and conditions (fig. S15) (11). This experiment conclusively shows that CO oxidation by group VIII metals deposited on CeO2 is size-dependent, with a direct participation in the reaction of metal atoms at the perimeter and ceria surface oxygen, and that Ni in contact with ceria exhibits rates similar to those of Pd or Pt. Our results demonstrate a robust method to explore the role of interfacial sites in catalysis, and demonstrate that the use of size-selected nanoparticles can successfully identify catalytically active sites.

Supplementary Materials

www.sciencemag.org/cgi/content/full/science.1240148/DC1

Materials and Methods

Scheme S1

Figs. S1 to S15

Tables S1 and S2

References and Notes

  1. See supplementary materials on Science Online.
  2. Acknowledgments: We thank M. Graziani, T. Montini (University of Trieste), B. Diroll, and K. Bakhmutsky (University of Pennsylvania) for discussions and help. Supported by University of Trieste through FRA project and COST Action CM1104 (M.C. and P.F.); the U.S. Department of Energy’s Advanced Research Projects Agency, Energy (ARPA-E) grant DE-AR0000123 (V.V.T.D.-N.); NSF through the Nano/Bio Interface Center at the University of Pennsylvania, grant DMR08-32802 (T.R.G.); Air Force Office of Scientific Research Multidisciplinary University Initiative grant FA9550-08-1-0309 (R.J.G.); and a Richard Perry University Professorship (C.B.M.). Aberration-corrected EM (R.E.D. and E.A.S.) was carried out at the Center for Functional Nanomaterials, Brookhaven National Laboratory, which is supported by the U.S. Department of Energy, Office of Basic Energy Sciences, under contract DE-AC02-98CH10886. M.C. conceived the idea for the study. M.C. and V.V.T.D.-N. synthesized the metal NCs. M.C. prepared the catalysts and collected the catalytic data. R.E.D. performed TEM, STEM, and ETEM characterization with help from V.V.T.D.-N. and T.R.G. E.A.S. coordinated all TEM studies. T.R.G. prepared the physical model of the NCs. R.J.G., P.F., and C.B.M. supervised the project. M.C. wrote the draft and all authors commented on the data and the manuscript.
View Abstract

Navigate This Article