Ceria Maintains Smaller Metal Catalyst Particles by Strong Metal-Support Bonding

See allHide authors and affiliations

Science  20 Aug 2010:
Vol. 329, Issue 5994, pp. 933-936
DOI: 10.1126/science.1191778


The energies of silver (Ag) atoms in Ag nanoparticles supported on different cerium and magnesium oxide surfaces, determined from previous calorimetric measurements of metal adsorption energies, were analyzed with respect to particle size. Their stability was found to increase with particle size below 5000 atoms per particle. Silver nanoparticles of any given size below 1000 atoms had much higher stability (30 to 70 kilojoules per mole of silver atoms) on reduced CeO2(111) than on MgO(100). This effect is the result of the very large adhesion energy (~2.3 joules per square meter) of Ag nanoparticles to reduced CeO2(111), which we found to be a result of strong bonding to both defects and CeO2(111) terraces, apparently localized by lattice strain. These results explain the unusual sinter resistance of late transition metal catalysts when supported on ceria.

Nanoparticles of late transition metals adsorbed on oxide surfaces form the basis for many catalysts important in energy technology, pollution prevention, and environmental cleanup. The catalytic activity per surface metal atom and selectivity can depend strongly on the particle size below 6 nm (15), the choice of oxide support (1, 2, 621), and the extent of oxide reduction (2226). Furthermore, the metal nanoparticles often sinter—they form fewer, larger particles—under catalytic reaction conditions and even during catalyst preparation. Sintering results in loss of activity or selectivity, mainly through a decrease in the number of exposed metal atoms but also through the loss of the smallest particles, which may have electronic properties that make them especially reactive. Late transition metal catalysts have also been reported to sinter more slowly or maintain smaller particles when supported on CeO2 relative to other supports (16, 19, 21). To fundamentally understand such structure-reactivity relations in catalytic phenomena, it is important to know how the energy of transition metal atoms in catalyst nanoparticles depends on the particle size and the nature of the oxide support surface to which they bind.

Here, we analyzed our previous calorimetric measurements of the adsorption energies of Ag vapor onto different oxide surfaces, on which Ag nanoparticles grow with very similar size and number density upon adsorption. These data allow for a quantitative comparison of variations of the energy of Ag atoms in Ag nanoparticles of different sizes, We next compared how the stability of a metal particle of a given size depends on the nature of the oxide surface to which it is attached. This analysis provides a quantitative estimate of the underlying thermodynamic reasons why late transition metal catalysts maintain smaller particle sizes and better resist sintering when on ceria supports relative to other oxide materials. By comparing ceria surfaces with different extents of reduction, we could assess qualitatively the stabilizing effect of surface oxygen vacancies on the adhesion energy of Ag nanoparticles to ceria. These results help to explain why ceria often enhances the performance of late transition metal catalysts—especially their unusual sinter resistance relative to other supports—in a wide variety of reactions in energy and environmental technology (8, 11, 13, 1521). The results also show that metal-oxide adhesion is locally stronger around oxide defects.

Our calorimetric measurements of the adsorption energies of Ag gas atoms onto clean and ordered surfaces of CeO2–x(111) films (with x = 0.1 and 0.2) are described in detail elsewhere (27). The CeO2–x(111) films were grown on Pt(111), and results were compared for film thicknesses of 1, 2, 3, and 4 nm (27). The Ce(3d) region in the X-ray photoelectron spectroscopy (XPS) was used to measure the Ce3+/Ce4+ concentration ratio within the probe depth of XPS (~1 nm), as described in (28), which was found to be 1:4 for the as-grown films. We describe these surfaces as CeO2–x(111), with x = 0.1, or as CeO1.9(111), which denotes CeO2(111) with 5% oxygen vacancies (29). The total density of steps and kinks was estimated from scanning tunneling microscopy (STM) images and low-energy electron diffraction spot widths to be ~6% of the total surface atoms, and most of the vacancies were probably localized at these step or kink defects (27). Measurements of the Ce3+/Ce4+ ratio were also performed on such 4-nm films after further thermal reduction by heating in ultrahigh vacuum, for which x was shown to increase to 0.2; we refer to the resulting films as CeO1.8(111) (27).

The measurements of the heat of adsorption for Ag on MgO(100) films that we analyze here have also been reported previously (30). The MgO(100) was 4 nm thick and grown on Mo(100) (30), and later proved to have ~5% step and kink sites as the dominant defects (31). For all four surfaces, the heat of adsorption of Ag was measured in ultrahigh vacuum as a detailed function of the amount of adsorbed Ag per unit area (i.e., Ag coverage). We found that the heat started at some low value depending on the surface, and increased to the heat of sublimation of bulk Ag (285 kJ/mol) as Ag coverage increased (27, 30).

Our previous quantitative analysis of spectral intensities in Auger electron spectroscopy and ion scattering spectroscopy for the different elements as a function of Ag coverage allowed us to ascertain the growth morphology of the Ag films on these MgO(100) and CeO2–x(111) surfaces (27, 30). The data were well fitted on all four surfaces by assuming that the Ag grows as three-dimensional (3D) Ag particles with the shape of a hemispherical cap and a fixed number per unit area, N, independent of Ag coverage. The data are not sensitive to the exact shape of the 3D particles but are very sensitive to their aspect ratio (ratio of average thickness to effective diameter) or thickness/area ratio, so that any shape with the same aspect ratio as hemispherical caps would fit the data equally well if using the same N. Our finding that N was independent of coverage (above ~0.03 monolayer) is consistent with the usual finding for such systems that form 3D particles—that is, their number density during nucleation quickly reaches a saturation value and thereafter stays nearly constant (32). The value of N was similar for all four surfaces, equal to 2.5 × 1012 particles/cm2 for the MgO(100) surface (30) and 4 × 1012 particles/cm2 for all three CeO2–x(111) surfaces (27). Dividing the Ag coverage (atoms/cm2) by N (particles/cm2) gives the average number of Ag atoms per particle at any given Ag coverage, which can be combined with the bulk density of Ag (5.9 × 1022 atoms/cm3) to give the average particle volume. This volume then corresponds to an average Ag particle diameter at each coverage. Because diameter varies only as the inverse cube root of N, the factor of ~2 maximum error on N (27) makes only a 25% error in diameter, to which the conclusions below are insensitive.

Figure 1 shows the measured heat of adsorption for these four surfaces plotted versus the average Ag particle diameter to which it adds, using the above approach to convert from Ag coverage to Ag particle diameter. The heat of adsorption versus coverage data were taken from (27, 30). The data in this format show large differences between the different surfaces in the stability of metal atoms adding to particles of the same size, and also show how stability depends on particle size for a given surface. For example, Ag atoms bind more strongly by ~75 kJ/mol to a 1-nm Ag particle on thermally reduced CeO1.8(111) than to a particle of the same size on MgO(100), whereas they bind more strongly by ~90 kJ/mol to a 5-nm particle than to a 1-nm particle when both particles are on MgO(100).

Fig. 1

Measured heat of Ag atom adsorption versus the Ag particle diameter to which it adds, for Ag adsorption onto four different surfaces: two 4-nm CeO2(111) films with different extents of surface reduction (x = ~0.1 and 0.2 in CeO2–x), grown on Pt(111); a 1-nm CeO1.9(111) film grown on Pt(111); and a 4-nm MgO(100) film grown on Mo(100). The inset shows structural models for perfect CeO2(111) and MgO(100), with their unit cells in black lines.

Another way to view these same data is to plot the energy of a metal atom after it adds to a particle (relative to its energy in bulk Ag) versus the number of metal atoms in the particle (Fig. 2). Silver atoms are 30 to 70 kJ/mol more stable in Ag nanoparticles for any given size (up to ~1000 atoms) when those particles are attached to CeO2–x(111) surfaces than when the particles are attached to MgO(100) surfaces. These differences are present for 4-nm films, well beyond the range where oxide film thickness has a measureable effect (i.e., below 2 nm). The difference decreases for larger particles and essentially disappears by ~5000 atoms per particle, where the energy of the added metal atom reaches the stability of the bulk metal even on MgO(100). Given that in reality there is a distribution of metal particle sizes for a given metal coverage, which broadens out the curves in Figs. 1 and 2, these differences between substrates at a given average particle size may be even more striking than shown here.

Fig. 2

Measured energy of a Ag atom, relative to its energy in bulk Ag(solid), versus the Ag particle size to which it adds, for Ag particles on the same four surfaces as in Fig. 1. For comparison, previously published results for Pb adsorption on this same type of MgO(100) film are also shown (33, 41).

Also shown in Fig. 2 are the only other data of this type ever reported, for Pb particles at a density of 8 × 1011 particles/cm2 on this same type of 4-nm MgO(100) film (33). The Pb energy is plotted relative to its energy in bulk Pb(solid). Note that the Pb and Ag data for MgO(100) fall very close to each other and are distinctly different from the Ag data for any of the CeO2–x(111) surfaces. The similarity in these curves for Ag and Pb on MgO(100) confirms our earlier explanation of this Pb curve as being dominated by the effect of particle size on the number of metal-metal bonds per atom (33). This Ag-MgO curve is above the Pb-MgO curve, as expected because Ag-Ag bonds are stronger than Pb-Pb bonds.

These curves in Fig. 2 directly reflect the thermodynamic driving force for nanoparticle sintering. If a metal atom is less stable on a certain particle size than in the bulk, it prefers to move to a larger particle where it is more stable [i.e., move to lower energy (to the right on each of these curves) until it reaches the minimum, or zero]. This driving force for sintering drops below 10 kJ/mol for ~400-atom (3-nm) particles on all three CeO2–x(111) surfaces, but this requires ~3000-atom (6-nm) particles on MgO(100). Sintering will stop at much smaller particles on CeO2 surfaces. Because Au nanoparticles are very active for several catalytic reactions when 3 nm in diameter but almost completely inactive above 6 nm (13), this particular size range is very important. We previously derived kinetic rate equations showing that this thermodynamic driving force is exponentially reflected in the sintering rate as if it were a negative activation energy (33, 34), which predicts that the sintering rates for particles below 3 nm will be much slower on CeO2 surfaces than on MgO(100) (by a factor of at least 30 at 500°C), consistent with observations that ceria offers a more sinter-resistant support for late transition metals than do other oxides (16, 19, 21).

Figures 1 and 2 also show that the stability of Ag atoms in small Ag nanoparticles (<1.5 nm or 30 atoms) on CeO2–x(111) is ~15 kJ/mol greater when the oxide surface is thermally reduced to CeO1.8(111). This increase in stability with vacancy concentration is consistent with observations that Au nanoparticles nucleate at oxygen vacancies and vacancy clusters on CeO2–x(111) (35), and with calculations that Ag adatoms bind >100 kJ/mol more strongly to vacancy sites on CeO2–x(111) than to stoichiometric sites (36). Silver atoms in nanoparticles smaller than 1.5 nm (30 atoms) on CeO2–x(111) films are 20 to 50 kJ/mol more stable when the film is only 1 nm thick (versus 4 nm). This difference may be caused by long-range electronic interactions with the underlying Pt(111) substrate. Other related effects have been observed experimentally and theoretically in this oxide thickness range, with the extent of charging of the adsorbed metal even being affected by thickness (3740). Although we did not test the effect of MgO(100) film thickness on Ag adsorption, we found that the heat of adsorption of Li was higher by ~50 kJ/mol on 1-nm MgO(100) films than on 4-nm films (39).

The effect of the different support materials on the energy diagram for the sintering of Ag nanoparticle catalysts is summarized in Fig. 3. The reaction is much more exothermic on MgO(100) (~45 kJ/mol) than on the different reduced CeO2(111) surfaces, where a 400-atom Ag particle has very little energy release to drive the sintering reaction [<10 kJ/mol on the CeO2–x(111) surfaces studied here]. This effect increases with the extent of reduction of the CeO2 surface, and also increases when the CeO2 film is so thin (1 nm) that the underlying Pt(111) support can exert a stabilizing influence on the Ag particles.

Fig. 3

Schematic representation of the differences in the energetics for the sintering of Ag nanoparticles on the different support surfaces studied here. The thermodynamic driving force for sintering of 400-atom Ag particles (diameter of 3 nm) is much greater on MgO(100) than on the slightly reduced CeO2(111) surfaces. The particles are shown attached to step edges on the basis of prior STM studies of related systems (see text). The red squares represent oxygen vacancies at these steps, and their number reflects the density of these oxygen vacancies. The stability of the Ag atoms approaches their stability in bulk Ag(solid) (the zero energy reference here) by the time they grow to 5000 atoms on all these surfaces, so the structure at the right is meant to represent all four support surfaces. Although not measured here, the activation energies for sintering shown here were estimated based on Brønsted relations: The transition state energy changes proportional to the reaction energy change with a slope of 0.5 (i.e., β = 0.5). The activation energies change even more than shown, according to our kinetic model (33, 34).

The origin of these differences in the stability of small metal nanoparticles on different oxide surfaces is associated with the extra stability offered by the strength of chemical bonding of the metal nanoparticle to the underlying oxide surface (and possibly its metal support when the oxide is thin enough). We have integrated the measured heat versus coverage curves (up to the coverage where the heat reaches the bulk sublimation energy) for each of these three CeO2–x(111) surfaces, and extracted from that integral the adhesion energies for Ag nanoparticles to these oxide surfaces, by exactly the same procedure outlined in (30), which gave an adhesion energy of 0.3 J/m2 for Ag on MgO(100). In all cases, we are assuming a hemispherical shape for the Ag nanoparticles. The results are summarized in Table 1.

Table 1

Calorimetrically measured adhesion energies of Ag nanoparticles to MgO(100) and reduced CeO2–x(111) surfaces, and average Ag particle size and coverage used for these measurements. The adhesion energy for Ag on Ag is given for comparison. Also shown is the initial heat of Ag adsorption (∆Hads) for the first pulse (~0.03 monolayer) of Ag gas.

View this table:

These results show that the adhesion energy of Ag nanoparticles to CeO2(111) is much larger than to MgO(100), and that this adhesion energy increases with the extent of reduction of the CeO2 and when the oxide thickness is only 1 nm (thicknesses of 2, 3, and 4 nm gave the same result). Note that these adhesion energies correlate with the initial adsorption energy of the first Ag gas pulse (Table 1), which corresponds to making clusters of ~4 atoms. Both values reflect the strength of Ag-oxide bonding, but the adsorption energy also includes substantial Ag-Ag bonding. Because it is not possible to create oxygen vacancies in a similar controlled way on this MgO(100) surface (because of their instability), and because we could not make CeO2–x(111) with x < 0.1, we could not test whether this difference in Ag adsorption and adhesion strength between CeO2(111) and MgO(100) is mainly caused by the residual oxygen vacancies on CeO1.9(111) or is attributable to some intrinsic difference between CeO2(111) and MgO(100). However, our measurements do reflect faithfully the situation in real catalytic materials under reaction conditions, where MgO exists with very few oxygen vacancies but CeO2 typically has many vacancies.

Because this large energy difference for Ag particles between CeO2 and MgO in Fig. 2 extends to 3D particles as large as 1000 atoms, the Ag atoms in such particles are not mainly bonded to oxygen vacancies. Although almost every such Ag particle is likely attached to a few oxygen vacancies at a step edge, a much larger number of Ag atoms are stabilized by interactions with CeO2 (relative to MgO) than those few Ag atoms that are directly bound to oxygen vacancies on CeO2. Thus, this difference must also reflect stronger bonding of Ag to CeO2(111) terraces than to MgO(100) terraces. The adhesion energies of Ag nanoparticles to these CeO2–x(111) surfaces (~2.3 to 2.6 J/m2) approach and even exceed the adhesion energy of Ag to itself (twice the surface energy of Ag, 2.44 J/m2) (30). For an adhesion energy equal to or larger than the Ag-Ag adhesion energy, one would generally not expect the Ag to cluster into 3D islands, but instead to wet the surface and form a continuous film. However, these results are for a Ag particle binding locally to some part of the ceria surface where there is likely a much greater defect (step, kink, or vacancy) concentration than elsewhere, and thus the local adhesion energy is likely much less on the stoichiometric CeO2 terraces. Furthermore, the lattice mismatch between Ag(111) and the underlying CeO2(111) will cause the CeO2 lattice under the Ag island to contract or expand to gain interfacial bonding stability. These effects would force the ceria lattice immediately adjacent to the island to strain in the opposite direction and also destabilize Ag bonding to the oxide next to the particle.

This lattice strain effect, which is supported by the combined evidence of these adhesion energies and particle morphologies, effectively creates a repulsive interaction between neighboring Ag nanoparticles. Such a repulsion should act like an activation barrier to prevent two Ag nanoparticles from diffusing together and agglomerating, and thus should inhibit catalyst sintering. It also explains the unexpected result that the number density of Ag particles is not larger on the surfaces to which Ag bonds more strongly. These local bonding and local strain effects must be very important in determining the stability of late transition metal nanoparticle catalysts and in understanding metal film morphology, nucleation, and growth during metal deposition onto oxide surfaces.

References and Notes

  1. To our knowledge, it is not possible to grow CeO2–x(111) films in this thickness range (1 to 4 nm) with x < 0.1 on Pt(111).
  2. Supported by the U.S. Department of Energy, Office of Basic Energy Sciences, Chemical Sciences Division, grant DE-FG02-96ER14630, and by NSF Integrative Graduate Education and Research Traineeship DGE-0504573 from the Center for Nanotechnology, University of Washington (J.A.F.).
View Abstract

Stay Connected to Science

Navigate This Article