A cluster expansion model for rate constants of surface diffusion processes on Ag, Al, Cu, Ni, Pd and Pt(100) surfaces

Rate constants of surface diffusion processes in metals often depend on the local atomic environment. Accurate models that describe this local environment dependence for specific material systems are not easily available in the literature despite the need for such information in kinetic models. In this work, we find the prefactor and activation barrier associated with adatom hop processes for a large number of local environments and use this information to systematically train cluster expansion (CE) models for Ag, Al, Cu, Ni, Pd and Pt. We demonstrate that although the local environments used for training is a small subset of all possible local environments, the CE model can predict the rate constants for many local environments that have not been used while training the model with reasonable accuracy. It is found that up to fourth nearest neighbor positions can be relevant to the activation barrier. Furthermore, many-body effects play an important role. We also find that the CE model depends on the direction of the hop.