Communication: Suppression of sintering of size-selected Pd clusters under realistic reaction conditions for catalysis

Size-selected metal clusters present new and attractive model systems for the analysis and optimization of catalytic effects, notably the dependence on particle size and morphology of activity, selectivity, and sintering.^1–7 Here we report an investigation, using scanning tunneling microscopy (STM), of the sintering of size-selected clusters under realistic conditions for heterogeneous catalysis. The conditions employed may be contrasted with surface science studies conducted in ultrahigh vacuum.⁸ We focus on catalysis of the oxidation of methane, at low concentration and mild temperatures by size-selected Pd clusters, Pd₅₅⁺ and Pd₃₀₉⁺, supported on graphite. Methane is a major greenhouse gases but also a clean energy source, with the lowest CO₂ yield per unit of energy produced.^9,10 Thus efforts to utilize methane released in coal mine ventilation (concentration <1%) may offer significant energy, economic, and environmental benefits. To minimize unwanted NO_x formation, low reaction temperatures are needed and Pd-based catalytic oxidation is promising.^10–16 The reaction is structure sensitive; a decrease in the temperature of complete oxidation and reduced inhibition by water are found when small Pd catalyst particles are used.^10,11,17,18 A key problem with the reaction is sintering,^1,11,19 leading to loss of catalytic activity, precisely the issue we address here in the case of size-selected cluster catalysts. We show that pinning of the Pd clusters to the graphite surface heavily suppresses sintering even under realistic reaction conditions.

The Pd clusters were produced with a home-built radio frequency magnetron-sputtering, gas-condensation cluster-beam source, and lateral time-of-flight mass filter,^20–25 and deposited on freshly cleaved graphite (HOPG) in high vacuum at room temperature. Two cluster sizes were generated for comparison, Pd₅₅⁺ and Pd₃₀₉⁺. To compare the sintering of pinned and unpinned clusters, we exploited an approximately Gaussian cluster beam profile to generate films of varying coverage, from multilayer aggregates in the center to submonolayers at the edge (average coverage 35 000 clusters/μm²). The samples were characterized by STM (Nanoscope IIIa, Veeco) before and after methane oxidation, using mechanically cut Pt/Ir (90/10) wire tip at room temperature in air.

The methane oxidation reaction was carried out in an Al reaction cell (20 cm³ volume) using certified gas mixtures (Linde Gas LLC). The flow rates of CH₄, O_2, and He were controlled by mass flow controllers (Brooks 5850S) to yield 0.8% CH₄ and 16% O₂ concentration in He (total flow rate 30 sccm). The pressure inside the cell was 800 Torr. The sample temperature was controlled with a ceramic heater (Momentive Performance Materials) and programmable controller (DigiSense), and measured with a K-type thermocouple attached to the heater surface. Online analysis of the reactant/product mixture was performed by sampling with a calibrated quadrupole mass spectrometer (Pfeiffer Prisma 300). The continuous scan mode was used to monitor up to 80 amu. A blank graphite substrate was used for background signal correction.

Figure 1(a) is a typical constant current STM image of a high coverage area (cluster beam center) of graphite on which Pd₅₅⁺ clusters were deposited at 2 keV (36 eV/atom). The surface is fully covered by particles with diameters varying from 2.8 to 6.8 nm, Fig. 1(c), and height from 0.6 to 3 nm. These are aggregates of Pd clusters piled up on the graphite surface in this multilayer coverage regime. This morphology contrasts strongly with that observed in the low coverage regime near the cluster beam edge. The STM image obtained in this low coverage area, Fig. 1(b), shows individual and randomly distributed pinned clusters. We did not observe accumulation of clusters at steps, consistent with cluster pinning to the impact site on the graphite surface.^20–29 The height of the clusters varies from ∼0.2 to ∼0.5 nm (corresponding to one or two atomic layers). The diameter distribution, Fig. 1(d) exhibits a mean diameter of 2.8 nm and full-width half maximum (FWHM) of 0.8 nm. The pinned clusters observed in the low coverage regime are as expected very much more monodispersed than the aggregates of clusters observed in the high coverage regime.

Figure 2 shows the catalysis results from the size-selected Pd₅₅⁺ cluster film on graphite. The following features are significant. (1) Above 100 °C we see products corresponding to the total oxidation of CH₄ to CO₂ and H₂O; (2) a slight deactivation is found at temperatures approaching 180 °C; (3) the level of activation increases again above 200 °C; and (4) the formation of H₂ is also evident above 180 °C, and no significant CO production is observed. The temperature range examined covered the onset of the reaction rather than the optimal conditions reported for Pd-based catalysts (see, e.g., Ref. 16). A possible pathway for hydrogen formation could be the reaction of CO adsorbed on Pd with H₂O via the low temperature water gas shift route reported for Pd-, Pt-, and Au-based catalysts at similar temperatures.^30–32 However, we shall not focus on the mechanistic reaction details here, since the cluster coverage varies across the sample from isolated clusters to the aggregated film, as explained. Instead, our focus is on the sintering behavior of the cluster films under the realistic reaction conditions employed.

Figure 3(a) is a constant current STM image of the high coverage region of the Pd₅₅⁺/graphite sample obtained after the reaction. The unpinned Pd particles piled up in this multilayer regime are seen to be highly unstable against the oxidation reaction. The particle diameter after reaction, Fig. 3(c), now varies from 2.8 to 14 nm [previously 2.8– 6.8 nm, Fig. 1(c)] and the height varies from 0.6 to 5 nm (previously 0.6–3 nm). It is evident that many of the particles produced initially by cluster aggregation in the multilayer film have coalesced together in Fig. 3(a).

The behavior of the pinned clusters in the Pd₅₅⁺/graphite system is quite different. Comparing the STM image of Fig. 3(b) with Fig. 1(b), taken before reaction, we see that the reaction conditions have rather little effect on the pinned clusters in the low coverage area of the sample. The small increase in the average diameter from 2.8 to 3.4 nm, Fig. 3(d), may partly be due to tip convolution effects. The FWHM of the diameter distribution increases only marginally, from 0.8 to 0.9 nm. Clusters, deposited above the pinning energy in the low coverage, submonolayer regime, thus show considerable stability under the reaction conditions; sintering is heavily suppressed with respect to the multilayer regime (unpinned particles). This conclusion is confirmed by the observation of small, stable clusters between the particles sintered in the high coverage regime, see Fig. 3(a). These clusters (diameter ∼3 nm) are presumably first layer clusters pinned by direct impact on the graphite surface; they are stable under the reaction, even while the cluster aggregates formed above them are sintering.

The degree of sintering of the pinned, size-selected clusters which does occur under the reaction conditions, seems to arise from the release of clusters from their pinning sites (depinning). This mechanism is suggested by the appearance in Fig. 3(b) of several pinholes on the surface. We propose that the release of a cluster from its pinning site exposes the surface defect created upon cluster–surface impact (i.e., on pinning), which can then be oxidized to form the monolayer deep etch pits observed in Fig. 3(b).^33,34 Our analysis shows that about 2% of the Pd₅₅⁺ clusters lift off their pinning sites leading to pinhole creation.

Previous studies^{20–23,25–28,34} show that cluster pinning occurs when one or more carbon atoms in the graphite surface are displaced from their lattice sites upon cluster impact above the pinning threshold energy (720 eV or 13 eV/atom for Pd₅₅⁺ clusters²⁶). These defects substantially increase the bonding between cluster and surface.^{20–23,25–29} More surface carbon atoms can be removed from their lattice sites, and deeper graphite layers affected, as the deposition energy increases above threshold.^20,34 For the Pd₅₅⁺ clusters, the impact energy of 2 keV (36 eV/atom) is well above the pinning threshold. Comparative experiments with Pd₃₀₉⁺ clusters deposited at 3 keV [pinning threshold <2.2 keV (Ref. 26)] show broadly similar catalytic and coverage-dependent sintering behavior, except that the proportion of clusters depinned in the low coverage regime is now increased to ∼9%. This can be explained by a weaker cluster–surface bond, since the pinning threshold is only just exceeded. Note that the orientations of individual clusters just before collision with the surface vary, so the precise atomic architectures of the defects created also vary from cluster to cluster.

Our results demonstrate that cluster beam pinning in the submonolayer coverage regime generates rather robust films of size-selected clusters which generally survive the realistic reaction conditions for methane oxidation employed here. This behavior is in strong contrast to the sintering of unpinned particles generated by cluster aggregation in the multilayer coverage regime. Moreover the degree of stability (pinning) may be controlled by the choice of cluster impact energy with respect to the pinning threshold. In future it will be interesting to explore, in a quantitative fashion, the consequence of this control over model catalyst structure and sintering on product yields and selectivities under similarly realistic reaction conditions. It is to be expected that such investigations will contribute to the development of improved catalysts, including those for other, more complex chemical reactions with larger molecules. In this context, the suppression of sintering by cluster pinning should contribute substantially to the durability of these new model catalysts.

We are grateful to a number of bodies for their financial support of this work. The NPRL work was funded by the Engineering and Physical Sciences Research Council (EPSRC) and the University of Birmingham. The work at Argonne National Laboratory (S.L. and S.V.) was supported by the US Department of Energy, BES-Chemical Sciences, BES-Materials Sciences, and BES-Scientific User Facilities under Contract No. DE-AC-02–06CH11357 with UChicago Argonne, LLC, Operator of Argonne National Laboratory. F.Y., A.A. and R.E.P. are grateful for the use of the facilities of the Center for Nanoscale Materials and Advanced Photon Source supported by the same contract.