Size-dependent catalytic activity and geometries of size-selected Pt clusters on TiO₂(110) surfaces

Abstract

Scanning tunneling microscope observations show a geometrical transition from a planar structure to a 3D structure at n = 8. This geometrical transition resulted in a significant decrease in the activation energy of the CO oxidation reaction. The upper-layer Pt atoms of the 3D cluster structure that starts to form at n = 8 are low-coordinated Pt atoms, and they may play an important role in the CO oxidation reaction.

---

1. Introduction

There is an urgent need to decrease the usage of platinum metal in automotive exhaust catalysts and fuel cells. Heterogeneous catalysts in automotive exhaust systems and fuel cells such as polymer electrolytes and direct methanol fuel cells consist of platinum metal clusters. Both systems consume large amounts of platinum metal. In order to minimize the usage of platinum metal in these systems, we should extract the highest possible activity per metal atom.

The catalytic activity of metal clusters is highly dependent on cluster size. Several studies have been conducted on the catalytic properties of gas-phase mass-selected clusters.^1–3 Owing to the interaction between the clusters and the substrate,^4–8 the clusters on the substrate surface can provide further specificity. An ideal way to ensure maximum activation of platinum metal particles is to arrange a specific number of platinum metal atoms in single-sized clusters in order to maximize the catalytic activity resulting from the interactions between the cluster atoms and the substrate atoms.

Heiz et al.⁹ showed that the catalytic activity of size-selected Pt_n (n = 5–20) clusters on MgO(100) films in the oxidation of CO increased abruptly during the transition from Pt₈ to Pt₁₅. A similar investigation of Au_n (n = 1–7) clusters on TiO₂(110) by Lee et al.¹⁰ showed that the catalytic activity increased substantially for Au₆ and Au₇. The abovementioned two studies show strong dependence of catalytic activity on the deposited cluster size, the geometries of the metal clusters on the surfaces were not measured directly in either case. Tong et al.¹¹ investigated the shape and size of size-selected Au_n (n = 1–8) clusters on TiO₂(110) by scanning tunneling microscopy (STM). The alignment of gold atoms in the clusters is not identified clearly in the reported STM images. On the other hand, Piednoir et al.¹² showed atomic-resolution images of palladium clusters on surfaces but the clusters were not size preselected. In either case, it is difficult to discuss the origin of the strong size dependence of the catalytic activity of clusters on surfaces.

First, we investigated Pt clusters supported on a TiO₂ substrate. The well-known Pt/TiO₂ system has strong metal–support interactions (SMSI).¹³ We have attempted to highlight the cluster–substrate interactions and their knock-on effects. Accordingly, we deposited size-selected Pt_n (n = 4, 7–10, 15) clusters on a TiO₂(110)–(1 × 1) surface under soft-landing conditions, and we investigated their structure, energy, and catalytic behavior.

2. Experimental

The experiments were performed in an ultra-high vacuum (UHV) chamber with a size-selected cluster-ion beam deposition system, X-ray photoelectron spectrometer (XPS), and a homemade high-pressure reaction cell.¹⁴

The samples for the experiments were prepared by Pt_n⁺ deposition on rutile TiO₂(110) single crystals. Prior to cluster deposition, the TiO₂(110) surfaces were cleaned by repeated cycles of Ar⁺ sputtering (1 keV, 10 min) and annealed at 980 K under vacuum, until a well-defined (1 × 1) low energy diffraction (LEED) pattern was observed and no impurities were detected viaAuger electron spectroscopy (AES). This treatment also creates bulk oxygen vacancies, making the sample surfaces conductive enough for ion deposition with minimal charging. Pt cluster ions were produced by a dc magnetron-sputter cluster-ion source.¹⁵ Size-selected Pt cluster ions were deposited under soft-landing conditions (<2 eV per atom) at room temperature. Energetic metal atoms sputtered from the target were cooled by He gas, resulting in the nucleation of clusters. After expansion through a nozzle, the ionized clusters could be accelerated and focused by an ion funnel. Then, the cluster beam was mass-selected using an Extrel quadrupole mass filter (mass range, 1–4000 amu) and finally deflected to the substrate. Pt cluster ions were deflected by 90° using a quadrupole deflector in order to remove neutral clusters that were not mass-selected. The distribution of the Pt clusters was checked and the mass-selected cluster was clearly separated based on the number of atoms as reported in the previous work.¹⁴ The Pt dose was monitored continuously during deposition by computer integration of the Pt_n⁺ current on the sample. The Pt coverage for STM observation was 5 × 10¹²–2 × 10¹³ atoms cm⁻², corresponding to approximately 0.5–2% of a close-packed Pt monolayer. The Pt coverage for XPS and activity measurements was 5 × 10¹³–1 × 10¹⁴ atoms cm⁻², corresponding to approximately 5–10% of a close-packed Pt monolayer. The ambient pressure was less than 1 × 10⁻⁷ Pa during deposition.

The high-pressure reaction cell was designed for high-pressure studies of catalytic activity using small-area samples and a retractable internal isolation cell with quartz lining, which constitutes a micro batch reactor in the ∼20 kPa pressure range under practical conditions.¹⁴ The test temperature was increased step by step (572 K, 598 K, 625 K). At each temperature the reaction rate of CO oxidation was constant. As the temperature increased, the reaction rate also increased. These results indicate that the CO oxidation reactions in this reactor were observed accurately as previously reported.¹⁴

STM observations were performed using a low-temperature scanning tunneling microscope (LT-STM) attached to a surface analysis chamber. Then, the samples were imaged using a LT-STM with a carbon nanotube (CNT) tip. STM images of the surface were acquired at 80 K in a constant current mode. The typical operating parameters were a sample bias in the range of +1 to +3 V and a tunneling current of 0.05–0.1 nA. The CNT tip for the STM was fabricated by manual attachment of a CNT to the tip apex viaelectron-beam-induced deposition of amorphous carbon under a scanning electron microscope as previously reported.¹⁶

Then, the samples were characterized viaXPS using non-monochromatized Mg Kα (1253.6 eV) radiation from a dual-anode X-ray source, along with a hemispherical energy analyzer (Omicron EA125HR) and a seven-channel detector. XPS spectra were taken at an electron take-off angle of 40° with a pass energy of 20 eV. The spectra were calibrated so that the Ti 2p_3/2 peak appeared at 459.0 eV of the binding energy expected for bulk TiO₂.¹⁷

3. Results and discussion

Fig. 1 shows atomic-resolution images of the size-selected Pt_n⁺ clusters. The images were obtained using a scanning tunneling microscope with a CNT tip, which enabled the identification of cluster atoms with atomic resolution based on the high aspect ratio of the CNT.

Fig. 1 STM images of Pt_n/TiO₂(110) (n = 4, 7–10, 15). For each pair of images, the left-hand side image is a 20 × 20 nm² view, whereas the right-hand side image is a 3.5 × 3.5 nm² view of a single Pt cluster with a schematic of the cluster geometry (U = 2 V and I = 0.1 nA). (Reproduced with permission of American Institute of Physics in the format Journal via Copyright Clearance Center.)

From these images, we can identify the atomic alignment of the clusters.¹⁸ Note that the bright and dark lines visible in Fig. 1 were assigned to fivefold coordinated titanium atom rows and bridging oxygen atom rows, respectively. They were separated by approximately 0.65 nm, which is in agreement with the reported value.¹⁹ The bright spots between the titanium rows correspond to vacant sites in the bridging O rows.²⁰

Height distributions and the average heights of deposited Pt_n (n = 4, 7–10, 15) on TiO₂(110)–(1 × 1) surfaces are shown in Fig. 2 and 3, respectively.

Fig. 2 Cluster height distributions of deposited Pt_n (n = 4, 7–10, 15) on TiO₂(110)–(1 × 1) surfaces. (Reproduced with permission of American Institute of Physics in the format Journal via Copyright Clearance Center.)

Fig. 3 Average cluster heights for Pt_n (n = 4, 7–10, 15) on TiO₂(110)–(1 × 1) surfaces. The two heights in Pt₈ and Pt₉ correspond to two separate peaks in the height distributions. The long-dashed line represents the height of the invisible bridging oxygen atom relative to the visible titanium atom at zero. The short-dashed line indicates heights expected for various platinum layers in the cluster. (Reproduced with permission of American Institute of Physics in the format Journal via Copyright Clearance Center.)

Height distributions of the Pt₄ and Pt₇ clusters on the surfaces were narrow [Fig. 2(a) and (b)], suggesting that both clusters had only a planar structure and lay flat on the surfaces. The height distributions of the Pt₈ and Pt₉ clusters were separated into two peaks [Fig. 2(c) and (d)]. The average heights of each peak were approximately 0.35 and 0.53 nm in either size cluster (Fig. 3), indicating two types of geometric structures.

It is suggested that the shorter clusters lay flat on the surface with a planar structure and the taller ones had a three dimensional (3D) structure with two atomic layers. Fig. 2(e) and (f) show height distributions for the Pt₁₀ and Pt₁₅ clusters, respectively. In contrast to Pt₈ and Pt₉, the distributions of Pt₁₀ and Pt₁₅ clusters had only one peak, suggesting that both size clusters had only 3D structures with two atomic layers.

At n < 7, the clusters lie flat on the surface in a planar structure. At n = 8, the clusters begin the transition from planar geometry to a 3D form. At n = 8 and 9, two types of geometric structures are evident, depending on the number of atoms in the cluster. At n = 15, the clusters have a geometric structure consisting of 3 atoms in the upper layer and 12 atoms in the lower layer. We found that size-selected Pt clusters deposited on TiO₂(110)–(1 × 1) form structures whose geometry is highly dependent on the number of atoms in the deposited cluster. Clusters smaller than Pt₇ have planar structures, whereas those larger than Pt₈ have 3D structures. The cluster geometry may thus be affected by the relationship between the strength of the Pt–Pt bonds in the cluster and the Pt–O(–Ti) bonds between the Pt cluster and the TiO₂ surface.

Fig. 4 shows the Pt 4f core-level binding energy for the clusters, obtained viaXPS. Fig. 5 shows the peak position of Pt 4f_7/2 peaks for mass-selected Pt_n/TiO₂(110) (n = 4, 5, 7, 8, 10, 15) as a function of the cluster size.

Fig. 4 Size dependence of Pt 4f core-level shifts of Pt_n/TiO₂(110). (Reproduced with permission of American Institute of Physics in the format Journal via Copyright Clearance Center.)

Fig. 5 Pt 4f_7/2 peak position for mass-selected Pt_n/TiO₂(110) (n = 4, 5, 7, 8, 10, 15) as a function of the cluster size. (Reproduced with permission of American Institute of Physics in the format Journal via Copyright Clearance Center.)

The Pt 4f_7/2XPS peaks show a distinct shift to higher binding energies relative to the bulk metal state. Although the shifts may be due to the interactions of the Pt clusters with oxygen from the TiO₂ surface, they clearly depend on cluster size. At n < 7, the binding energy decreases steeply; at n > 8 it decreases much more gradually.²⁰Cluster size presumably affects binding energy because of its effect on cluster geometry, which in turn affects cluster–surface interactions. Note that such shifts result from a combination of initial- and final-state effects,²¹ but based on photoemission results alone, we cannot unambiguously differentiate them. Nevertheless, artifacts such as charging effects must be excluded. A problem with such samples (i.e. small particles on insulating surfaces) might be differential charging.²² This is excluded here because the surfaces would be conductive as a result of sputtering and annealing. The peak widths are small and show no broadening, which is a clue to the absence of charging effects. At n = 15, the binding energy is minimum (71.6 eV); however, it is higher than the value for the bulk metal (71.2 eV).²³

At n = 8, the cluster size around which the cluster geometry transforms from planar to 3D is in good agreement with the inflection point in the size dependence of the core-level shifts. Typically the core-level shift for various Pt oxides and hydroxides is 3 eV for stoichiometric compounds.²³ As compared to the core-level shifts of metals and stoichiometric oxides, those of the Pt clusters on TiO₂ are intermediate shifts because of the formation of Pt–O bonds. At n < 7, each atom of the cluster is in contact with an oxygen atom of the TiO₂ surface. At n > 8, atoms of the cluster begin to populate either the upper or the lower layer; hence, the contact with the TiO₂ surface increases gradually up to at least n = 10, as estimated from the cluster geometries.¹⁷ The gradual increase in surface contact indicates a gradual change in the binding energy at n > 8. In addition, the configuration or coordination of the Pt atoms may bring about gradual changes in the binding energy. Thus, the cluster-size dependence of core-level shifts can be attributed to the influence of cluster geometry on cluster–surface interactions. Density functional theory (DFT) calculations of free clusters showed that the binding energies of the Pt clusters increased with the cluster size.²⁴ It is assumed that strong interactions with the TiO₂ surfaces are responsible for the planar structures. The geometries of clusters on the surfaces could be affected by the relationship between the strength of Pt–Pt bonds in the clusters and the interactions with the surface.

The catalytic oxidation of CO on size-selected Pt clusters on the TiO₂(110) surface was investigated using a high-pressure reaction cell. As shown in Fig. 6, the reaction rates of CO oxidation on size-selected Pt clusters deposited on TiO₂(110) were measured at each cluster size. The normalized production rates of CO₂, relative to the number of Pt atoms in the sample, are dependent on the cluster size. The reactivity increases up to Pt₈ at each temperature of 572 K and 598 K, and it is maximum when n = 7 at 625 K.

Fig. 6 CO oxidation activity on Pt_n/TiO₂(110) (n = 4–8, 10, 15) with a mixture of 5.8 kPa CO, 2.2 kPa O₂ and 12 kPa Ar (for calibration).

We would like to emphasize that there is no difference between the Pt 4f spectra before and after the activity measurements.

Fig. 7 shows Pt 4f spectra of Pt₁₀/TiO₂(110) before and after the activity measurements. There might be observed a shift in the binding energy if the clusters aggregate and the cluster size increases or decreases, but no shift was observed. This result means that the cluster size did not change before and after the activity measurements under these experimental conditions.

Fig. 7 Pt 4f spectra for mass-selected Pt₁₀/TiO₂(110) before (blue line) and after (red line) the activity measurement at 625 K.

For CO oxidation, CO + 1/2O₂ → CO₂, the reaction rate r of CO oxidation is expressed as

r = k[CO]^a[O₂]^b.

The square brackets denote the concentrations of CO and O₂.

The rate constant k is given by the Arrhenius equation,

k = Ae^−E_a/RT

where T denotes the temperature in Kelvin; R = 8.31 J K⁻¹ mol⁻¹ is the gas constant; E_a is the activation energy, i.e., the minimum energy needed for the reaction to occur; and A is the frequency factor, i.e., the pre-exponential factor (frequency of collisions and their orientation) given by A = p·Z, where Z is the collision rate and p is a steric factor. Z is weakly dependent on temperature.

Thus, the frequency factor is a constant specific to each reaction.

The orders of reaction for CO oxidation were ∼1.0 for O₂ and ∼−1.0 for CO, corresponding to the orders reported previously.²⁵ To analyze the data, the activation energy of the CO oxidation reaction was estimated using the Arrhenius plot. The plot of ln k versus 1/T is linear with a negative slope; the slope is equal to the negative activation energy divided by the gas constant R.

Fig. 8 shows the activation energy of Pt_n (n = 4–8, 10, 15)/TiO₂(110) as a function of cluster size.

Fig. 8 Activation energy (E_a) of Pt_n (n = 4–8, 10, 15)/TiO₂(110) as a function of cluster size.

In the temperature range of our investigation, there should be no bond formation between the Pt and Ti cations or among the Ti atoms, as reported previously.¹³ A low activation energy indicates a low barrier height for CO oxidation. For n = 4–7, the activation energy is constant; however, it steeply decreases at n = 8, indicating that the cluster structure changes to a more active state.

The upper-layer Pt atoms of the 3D cluster structure that starts to form at n = 8 are low-coordinated Pt atoms, and they may play an important role in the CO oxidation reaction. In the temperature range under consideration, the rate-limiting step in CO oxidation is O₂ dissociation.

While we have no direct evidence of the active sites, a likely possibility is that at n = 8, the upper-layer Pt atoms of the 3D cluster adsorb O₂, and they behave essentially like PtO₂. Then, CO is easily adsorbed onto these upper-layer Pt atoms, decreasing the barrier height for CO oxidation to ∼0.3 eV (29 kJ mol⁻¹) as reported by density functional theory study,²⁶ which is consistent with the significant decrease in the barrier height (Fig. 3); this facilitated the overall reaction.²⁷

In addition, the role of the geometrical factor in CO oxidation on a platinum surface was reported.²⁸CO on terrace sites is more reactive than CO on step sites. On the other hand, chemisorbed oxygen on step sites is more reactive than chemisorbed oxygen on terrace sites. Thus, the most favorable reaction geometry is chemisorbed CO on terrace sites interacting with chemisorbed O on step sites.²⁹ In the upper-layer Pt atoms of the Pt₈ clusters, the two factors stated above cause the activation energy of CO oxidation to decline abruptly.

Fig. 9 shows the pre-exponential factor per cluster of Pt_n (n = 4–8, 10, 15)/TiO₂(110) as a function of cluster size. The pre-exponential factor increases up to n = 7, and it is maximum at n = 7.

Fig. 9 Pre-exponential factor (A) of Pt_n (n = 4–8, 10, 15)/TiO₂(110) as a function of cluster size.

It steeply decreases at n = 8, indicating an abrupt decline in the pre-exponential factor.

The E_a drop by 30% and abrupt decrease of the pre-exponential factor between n = 7 and n = 8 seem to be unacceptable. It might mean the limitation of the apparent kinetic analysis. But it is apparent that the reactivity increases up to Pt₈ at each temperature of 572 K and 598 K, and it is maximum when n = 7 at 625 K from Fig. 6. We might suggest the following speculation. In this reaction, two adjacent 3-fold hollow sites for oxygen dissociation and one on-top site for carbon monoxide constitute the minimum requirement for the CO oxidation reaction. Thus, four close-packed atoms constitute the minimum set of reaction sites, as shown in Fig. 10. Up to n = 7, the Pt clusters lie flat on the surface in a planar structure. The number of the reaction sites increases up to 5 times the number of reaction sites in the case of the Pt₄ cluster.

Fig. 10 Schematic of the reaction site of Pt_n (n = 4, 7, 8)/TiO₂(110).

At n = 8, the number of reaction sites declines significantly with the geometry transformation from planar to 3D. This significant decline is attributed to site blocking of the upper-layer Pt atoms. The number of reaction sites would be strongly but indirectly associated with the pre-exponential factor. The above discussion might explain the behaviour of the pre-exponential factor shown in Fig. 9. This explanation is only one possibility and too much straightforward. Further studies at the deep end are needed to give a clear explanation.

4. Conclusions

Using STM images taken with a CNT tip, we have shown that size-selected Pt clusters deposited on TiO₂(110)–(1 × 1) form structures whose geometry is highly dependent on the number of atoms in the deposited cluster. A geometrical transition from a planar structure to a 3D structure was observed when the cluster size increased to 8 Pt atoms. The catalytic activity and activation energy shifted to a more active state at the same cluster size at which planar-to-3D transition occurred. Therefore, the structural transition also shifted to a more active state, presumably under Pt–Pt and Pt–surface interactions. Significantly, we found that the binding energies and catalytic activation energy of Pt clusters show corresponding transitions at the same cluster size. The presence of a second layer of atoms in a Pt cluster on TiO₂(110) may be associated with a significant decrease in the activation energy of CO oxidation. The second layer Pt atoms are low-coordinated atoms, and they interact with the TiO₂ surface indirectly. In contrast, the upper-layer atoms may play an important role in blocking the reaction sites, thereby decreasing the frequency of the CO oxidation reaction. These two important factors mainly influence the reaction rate. Our findings provide novel insights into the fabrication of efficient precious-metal catalysts. The electronic states of small clusters would be easily affected by cluster–surface interactions. These interactions hinder the aggregation of small clusters on the surface. This is an important practical advantage. We can enhance the catalytic activity by controlling not only the size of the nanoclusters but also the interactions between the clusters and the substrate. Such an enhancement can potentially bring about a drastic decrease in the usage of precious metals.

Acknowledgements

We thank M. Yoshimura of the Toyota Technological Institute for providing CNT tips for the STM experiments.

Notes and references

1. P. Fayet and L. Wöste, Surf. Sci., 1984, 156, 134.
2. M. J. Manard, P. R. Kemper and M. T. Bowers, J. Am. Chem. Soc., 2005, 127(28), 9994–9995.
3. O. P. Balaj, I. Balteanu, T. T. J. Roßteuscher, M. K. Beyer and V. E. Bondybey, Angew. Chem., Int. Ed., 2004, 43, 6519–6522.
4. C. Fan, T. Wu, W. E. Kaden and S. L. Anderson, Surf. Sci., 2006, 600(2), 461–467.
5. S. Kunz, F. F. Schweinberger, V. Habibpour, M. Rttgen, C. Harding, M. Arenz and U. Heiz, J. Phys. Chem. C, 2010, 114(3), 1651–1654.
6. X. Tong, L. Benz, P. Kemper, H. Metiu, M. T. Bowers and S. K. Buratto, J. Am. Chem. Soc., 2005, 127(39), 13516–13518.
7. S. Vajda, M. J. Pellin, J. P. Greeley, C. L. Marshall, L. A. Curtiss, G. A. Ballentine, J. W. Elam, S. Catillon-Mucherie, P. C. Redfern, F. Mehmood and P. Zapol, Nat. Mater., 2009, 8, 213–216.
8. W. E. Kaden, T. Wu, W. A. Kunkel and S. L. Anderson, Science, 2009, 326(5954), 826–829.
9. U. Heiz, A. Sanchez, S. Abbet and W.-D. Schneider, J. Am. Chem. Soc., 1999, 121, 3214.
10. S. Lee, C. Fan, T. Wu and S. Anderson, J. Chem. Phys., 2005, 123, 124710.
11. X. Tong, L. Benz, P. Kemper, H. Metiu, M. T. Bowers and S. K. Buratto, J. Am. Chem. Soc., 2005, 127, 13516.
12. A. Piednoir, E. Perrot, S. Granjeaud, A. Humbert, C. Chapon and C. R. Henry, Surf. Sci., 1997, 391, 19.
13. S. J. Tauster, S. C. Fung and R. L. Garten, J. Am. Chem. Soc., 1978, 100, 170–175.
14. Y. Watanabe and N. Isomura, J. Vac. Sci. Technol., A, 2009, 27, 1153–1158.
15. H. Haberland, M. Karrais, M. Mall and Y. Thurner, J. Vac. Sci. Technol., A, 1992, 10, 3266.
16. M. Yu, M. J. Dyer, G. D. Skidmore, H. W. Rohrs, X. Lu, K. D. Ausman, J. R. Von Ehr and R. S. Ruoff, Nanotechnology, 1999, 10, 244–252.
17. A. F. Carley, P. R. Chalker, J. C. Riviere and M. W. Roberts, J. Chem. Soc., Faraday Trans. 1, 1987, 83, 351–370.
18. N. Isomura, X. Wu and Y. Watanabe, J. Chem. Phys., 2009, 131, 164707–164710.
19. U. Diebold, Surf. Sci. Rep., 2003, 48, 53.
20. S. Wendt, R. Schaub, J. Matthiesen, E. K. Vestergaard, E. Wahlström, M. D. Rasmussen, P. Thostrup, L. M. Molina, E. Lægsgaard, I. Stensgaard, B. Hammer and F. Besenbacher, Surf. Sci., 2005, 598, 226.
21. N. Isomura, X. Wu, H. Hirata and Y. Watanabe, J. Vac. Sci. Technol., A, 2010, 28, 1141.
22. W. F. Egelhoff Jr and G. G. Tibbetts, Phys. Rev. B, 1979, 19, 5028.
23. T. L. Barr, J. Vac. Sci. Technol., A, 1989, 7, 1677.
24. M. Peuckert and H. P. Bonzela, Surf. Sci., 1984, 145, 239–259.
25. T. Engel and G. Ertl, Adv. Catal., 1979, 28, 1.
26. J. Y. Park, C. Aliaga, J. R. Renzas, L. Hyunjoo and G. A. Somorjai, Catal. Lett., 2009, 129, 1–6.
27. T. M. Pedersen, L. W. Xue and B. Hammer, Phys. Chem. Chem. Phys., 2006, 8(13), 1566–1574.
28. L. Xiao and L. Wang, J. Phys. Chem. A, 2004, 108(41), 8605–8614.
29. J. Xu and J. T. Yates Jr, J. Chem. Phys., 1993, 99, 725–732.

---