The lack of distribution network and storage problems limits the use of stored hydrogen in fuel cell applications. Thus, producing pure hydrogen from hydrocarbons, like LPG, on site, either for mobile or immobile uses, is of great scientific interest [1-3]. The device called "fuel processor" converts hydrocarbon feeds to pure hydrogen. In a fuel processor, there are three catalytic reactions, namely reforming (either steam reforming (SR) or autothermal reforming (ATR), water-gas shift (WGS), and preferential oxidation (PROX), are conducted in series. From those reactions, reforming produces hydrogen from hydrocarbons, WGS increases hydrogen concentration and, at the same time, decreases CO concentration in the stream and PROX eliminates CO impurity from the resultant hydrogen rich stream. Since the Pt anode of a commercial fuel cell can operate stability only at CO levels below 4 ppm, PROX should decrease CO concentration down below that level in the produced hydrogen stream.

The PROX reaction on transition metal surfaces generally accepted to follow the Langmuir-Hinshelwood mechanism. Here, there are three steps: (i) both reactants, CO and O, adsorb on a solid surface; (ii) they diffuse toward each other or one to the other along the surface and react to form a product; and (iii) the product desorbs from the surface. Recent studies in heterogeneous catalysis have shown that bimetallic alloy surfaces have high catalytic activities in PROX during purification of hydrogen rich effluent of a fuel processor, especially when compared to monometallic catalysts like platinum. Experimental and theoretical calculations showed this enhancement in the activity of bimetallic alloys is through a combination of different factors [4,5]. First of all, the existence of a second metal like Sn or Ru at the surface creates empty sites for the adsorption of oxygen in the first step of the reaction mechanism and thus eliminates bonding competition between the reactants (bifunctional effect). Secondly, it decreases the adsorption strength of CO on the surface Pt atoms (electronic effect) and thus diffusion of the reactants towards each other becomes easier in the second step of the reaction mechanism.

Computational techniques using density functional theory (DFT) calculations are useful tools in examining the electronic properties of the metal surface atoms during the CO oxidation reaction. However, it is known that one common obstacle in the electronic structure calculations using DFT is that functionals, such as LDA or GGA, have a tendency to make a wrong estimation of site preference for CO adsorption on transition metal surfaces. On (111) surface termination of Pt and also Pt-alloys like Pt3Sn, DFT calculations result in a preference of hollow sites over atop sites, whereas experimentally it was proven that atop sites are the most favorable sites for CO adsorption, followed by occupation of bridge sites [6,7]. The chemical explanation of this error is that non-hybrid DFT calculations underestimate metal band gaps of Pt and HOMO-LUMO gaps for CO molecule. During adsorption, the backdonation of electron density from the substrate to the LUMO of CO molecule increases with the coordination number of the adsorption site and when the position of the LUMO orbital is estimated lower than it should have been during the calculations, the interaction with the metal electrons are falsely predicted higher on especially hollow sites. The reason of this error in theoretical calculations is neither convergence related or based on the choice of variables such as exchange-correlation functional and pseudopotential representation of the electron-nucleus interaction. Recent studies of Orita et al [8], on the other hand, solved this problem for CO-Pt(111) system using all electron scalar relativistic (AER) calculations. These give a deeper Fermi level for the metal surface which provides the effect for decreasing the LUMO of the CO with the metal substrate and thus leads to the correct site preference. In order to analyze the catalytic activity of a metal catalyst towards CO oxidation, it is crucial to have a detailed understanding of the behaviour of the reactants in the initial two steps of the reaction process, i.e. adsorption of the reactants on the surface and the reaction pathway.

In this study, we present our results of the adsorption properties of Pt3Sn(111) alloy using all-electron relativistic DFT calculations. AER calculations make a correct estimation of the site preference of adsorbed CO on the metal surface, occupation of atop sites followed by occupation of bridge sites, in correspondence with experimental findings in the literature. The decreased activation barrier of the reaction on the alloy surface describes the reasons of the increased activity of Pt-alloy compared to monometallic Pt. The results of CO and O adsorption, CO+O coadsorption properties of Pt3Sn(111) alloy along with an analysis of CO oxidation reaction pathway on this surface on using AER calculations are presented. Pt3Sn is chosen as a model considering its high activity and selectivity, as well as stability in PROX reaction. Whatsmore, it has one of the highest catalytic activities for CO electrooxidation recorded to date and thus an important catalyst in the low temperature fuel cell anodes. We found that the most favorable site, according to the results of AER calculations, is indeed atop site, followed by occupation of bridge sites, in correlation with the experimental findings in the literature. The most stable site for O adsorption is 3-fold fcc Pt2Sn site and it was shown that the binding of O onto the surface is much more (approximately 3 times) stronger than the binding of CO. The oxidation reaction occurs when initially CO is adsorbed atop Pt and O is adsorbed at a nearby fcc hollow site. The interaction between the adsorbed CO and O is short range and there is not a pronounced change in the binding strength of one reactant when the second reactant adsorbs on the surface. The reaction pathway is that CO atom adsorbed atop-Pt moves towards the adsorbed O atom and at the same time the adsorbed O diffuses from the 3-fold site to bridge PtSn site. The activation barrier of CO oxidation reaction in this way is lower on Pt3Sn(111) than it is on Pt(111) which describes the reasons of the increased activity of bimetallic Pt alloy towards CO oxidation reaction..

References

[1] Avci, A. K., Onsan, Z. I., Trimm, D. L., Topics in Catalysis, 22 (2003) 359-367 [2] Avci, A. K., Onsan, Z. I., Trimm, D. L., Applied Catalysis A-General, 216 (2001) 243-256 [3] Avci, A. K., Trimm, D. L., Onsan, Z. I., Chemical Engineering Journal, 90 (2002) 77-87 [4] Paffett, M.T., Gebhard, S.C., Windham, R.G., Koel, B.E., J. Phys. Chem. 94 (1990) 6831–6839 [5] Koper, T. M., Surface Science, 548 (2004) 1-3 [6] Gülmen, M. A., Sümer, A., Aksoylu, A. E., Surface Science, 600 (2006) 4909-4921 [7] Sümer, A., Gülmen, M. A., Aksoylu, A. E., Surface Science, 600(2006) 2026-2039 [8] Orita, H., Itoh, N., Ineda, Y., Chemical Physics Letters, 384 (2004) 271-276