CO Adsorption on Supported Gold Nanoparticle Catalysts: Application of the Temkin Model

The adsorption of CO on the supported gold nanoparticle catalysts Au/TiO2, Au/Fe2O3, and Au/ZrO2 was examined using infrared transmission spectroscopy to quantify the isobaric CO coverage as a function of temperature. The Temkin adsorbate interaction model was then applied to account for the adsorption behavior. To test the general applicability of the Temkin model, this treatment was also applied to three data sets from the literature. This included another real-world catalyst and two model catalysts. All data sets were accurately represented by the Temkin adsorbate interaction model. The resulting thermodynamic metrics are consistent with previous determinations and reflect a particle size-dependence. In particular, the intrinsic adsorption enthalpy at zero CO coverage varies almost linearly with Au particle size, and this trend appears to be correlated with the abundance of low-coordinate Au sites (cf., CN = 6 and 7 for corners and edges, respectively). For very small particles with mostly CN = 6 corner sites, the enthalpy reflects strong binding (cf., −ΔH0 ≈ 78 kJ/mol), while for large particles with mostly CN = 7 edge sites, the enthalpy reflects weaker binding (cf., −ΔH0 ≈ 63 kJ/mol). The results also suggest that these sites are coupled. This study demonstrates that the Temkin adsorbate interaction model accurately represents adsorption data, yields meaningful metrics that are useful for characterizing nanoparticle catalysts, and should be applicable to other adsorption data sets. ■ INTRODUCTION Of all of the reactions examined on supported metal nanoparticle catalysts, the oxidation of CO has received extensive experimental and theoretical attention. While the reaction may appear to be rather straightforward, an understanding of the mechanism has been surprisingly difficult to achieve. Not only is the oxidation of CO a model system to study, but the adsorption of CO on gold, a fundamental step in the mechanism, has also become a model interaction to investigate. Multiple studies indicate that CO is only weakly chemisorbed on gold. The adsorption is believed to involve low-coordinate sites with coordination number CN = 6 (i.e., corners or kinks) and CN = 7 (i.e., edges or steps). Adsorption to terrace sites with CN = 8 or 9 does not occur. CO adsorption studies also display a number of coverage-dependent results. For example, quantification of the coverage with CO pressure (or with temperature) reveals coverage-dependent, non-Langmuir behavior. The adsorption enthalpy shows a coveragedependence, decreasing with increasing coverage. Also, the infrared studies reveal a common CO peak that typically red shifts with increasing coverage. To account for the physicochemical, coverage-dependent behavior of CO adsorption on gold, we developed a treatment of the Temkin adsorption model. This thermodynamic model is an extension of the Langmuir adsorption model that incorporates a linear variation in binding energy with coverage, and has three cases that are similar but not equivalent. The adsorbate interaction case takes into account direct adsorbate− adsorbate interactions or indirect adsorbate−substrate interactions. This case assumes that these interactions produce a linear variation of adsorption enthalpy with adsorbate coverage. In contrast, the heterogeneous surface case assumes a uniform distribution of heterogeneous binding sites. It is assumed that the adsorption enthalpy varies linearly over these different binding sites. The third case involves a common approximation for midrange adsorbate coverage. According to this approximation, the expressions for the adsorbate interaction and the heterogeneous surface cases are simplified, producing a new common expression. This common expression is the familiar Temkin isothermal result; coverage varies with the logarithm of pressure. Previously, we demonstrated that this treatment of the Temkin adsorption model provides meaningful thermodynamic metrics for enthalpy and entropy, which can be used to characterize and explain differences between various catalysts. The model is straightforward and applicable for fitting both isothermal and isobaric data sets. Received: March 8, 2012 Revised: May 1, 2012 Published: May 1, 2012 Article