Neural-network analysis of CO adlayer structures and lateral interactions on transition metal surfaces

Under realistic operating conditions, heterogeneously catalyzed reaction events occur on surfaces containing many adsorbates. Lateral interactions between co-adsorbates affect the stability of the adlayer as well as the chemical activity by altering the potential energy surface. Elucidating the structure of the adlayer experimentally is difficult due to the large number of adsorption sites that may be occupied and the complex electronic interaction between the adlayer and the surface. Theoretical modeling can resolve these assignments, but is computationally limited due to the large number of adlayer configurations that have to be evaluated. To overcome this limitation, a neural-network-based method is utilized in this work to study the interaction and aggregation behavior of CO on a plethora of metallic surfaces as commonly encountered in catalytic applications. It is found that the configuration of the CO molecules on the surface is largely controlled by steric interactions, resulting in the formation of three distinct operating regimes (low-, medium- and high-coverage) for all systems. The exact transition points and site occupations for these regimes were found not to follow simple correlations, instead depending on a complex balance between the surface-adsorbate and lateral interactions. Data-driven methodologies make these lateral interaction potentials accessible for use in kinetic studies without compromising accuracy.