Towards an Ab-initio Description of Adsorbate Surface Dynamics

The advent of machine learning potentials (MLPs) provides a unique opportunity to access simulation timescales and to directly compute physiochemical properties that are typically intractable using density functional theory (DFT). In this study, we use an active learning curriculum to train a generalizable MLP using the DeepMD-kit architecture. The resulting model, which provides DFT-level accuracy, is used to investigate the diffusion of key surface-bound adsorbates on a Ag(111) facet using sufficiently long MLP-based molecular dynamics (MD) simulations. The MLP/MD-calculated diffusivities, obtained at three different temperatures, demonstrates the potential shortcomings of DFT-based nudged elastic band calculations that are widely employed to study diffusion. While this work has focused on a few simple adsorbates, the resulting model is transferable and can be fine-tuned to study other adsorbates that were not included in the initial training dataset.