Machine Learning Molecular Dynamics Shows Anomalous Entropic Effect on Catalysis through Surface Pre-melting of Nanoclusters

Due to the superior catalytic activity andefficient utilization of noble metals, nanocatalysts areextensively used in the modern industrial production ofchemicals. The surface structures of these materials aresignificantly influenced by reactive adsorbates, leadingto dynamic behavior under experimental conditions.The dynamic nature poses significant challenges instudying the structure–activity relations of catalysts.Herein, we unveil an anomalous entropic effect oncatalysis via surface pre-melting of nanoclusters throughmachine learning accelerated molecular dynamics andfree energy calculation. We find that due to the pre-melting of shell atoms, there exists a non-linear variationin the catalytic activity of the nanoclusters with temper-ature. Consequently, two notable changes in catalystactivity occur at the respective temperatures of meltingfor the shell and core atoms. We further study thenanoclusters with surface point defects, i.e. vacancy andad-atom, and observe significant decrease in the surfacemelting temperatures of the nanoclusters, enabling thereaction to take place under more favorable and milderconditions. These findings not only provide novel in-sights into dynamic catalysis of nanoclusters but alsooffer new understanding of the role of point defects incatalytic processes.