Abstract
Bare, low-index periodic surface models are typically used to examine metal-catalyzed reactions in density functional theory (DFT) studies, and these most closely resemble low-pressure surface science reactions and catalyzed reactions that occur on large terraces that prevail on large (>5 nm) supported nanoparticles. Many catalytic reactions, however, occur near conditions at which catalytic surfaces are saturated by one or more adsorbed intermediates, leading to strong coadsorbate interactions and surface reconstruction leading to increased curvature. Alkane hydrogenolysis is such a reaction and has been extensively studied using DFT-often on bare metal surfaces with the assumption that omitted coadsorbed hydrogen atoms (H*) do not significantly alter the relative activation barriers and with ad hoc assumptions about the site requirements for relevant reactions. Here, we use ethane hydrogenolysis on H*-covered Ir catalysts (using a periodic surface model and a nanoparticle model) as a probe reaction to examine coadsorbate interactions and to demonstrate the rigorous determination of site requirements. The kinetically relevant transition state [*CH-CH*](double dagger) is larger than the 0-3 coadsorbed H* atoms it replaces, such that the reaction has a positive activation area (a concept analogous to activation volume in homogeneous reactions) and thus repels coadsorbed H* atoms when fewer than four H* vacancies are created. This induced strain cannot be relieved on the periodic surface models, resulting in large effective free energy barriers and predictions that four vacant sites are required (gamma = 4). These barriers and site requirements lead to turnover rates that are 4 orders of magnitude lower than measured rates and incorrect H-2-pressure dependencies. Furthermore, varying the unit cell size of the Ir(111) surface dramatically alters the calculated reaction energetics, indicating that relevant transition states destabilize one another over long distances through the H* adlayer. Curved H*-covered Ir hemispherical particle models (119 atoms), however, stabilize transition states at a lower number of vacant sites (gamma = 2) through lateral relaxation of the adlayer, resulting in correct predictions of H-2-pressure dependencies and quantitative agreement between calculated and measured rates.