Cu/Ag EAM potential optimized for heteroepitaxial diffusion from ab initio data

A binary embedded-atom method (EAM) potential is optimized for Cu on Ag(1 1 1) by fitting to ab initio data. The fitting database consists of OFF calculations of Cu monomers and dinners on Ag(1 1 1), specifically their relative energies, adatom heights, and dimer separations. We start from the Mishin Cu-Ag EAM potential and first modify the Cu-Ag pair potential to match the FCC/HCP site energy difference then include Cu-Cu pair potential optimization for the entire database. The potential generated from this optimization method gives better agreement to DFT calculations of Cu monomers, dimers, and trimers than previous EAMs as well as a SEAM optimized potential. In trimer calculations, the optimized potential produces the DFT relative energy between FCC and HCP trimers, though a different ground state is predicted. We use the optimized potential to calculate diffusion barriers for Cu monomers, dinners, and trimers. The predicted monomer barrier is the same as DFT, while experimental barriers for monomers and dinners are lower than predicted here. We attribute the difference with experiment to the overestimation of surface adsorption energies by DFT and a simple correction is presented. Our results show that this optimization method is suitable for other heteroepitaxial systems; and that the optimized Cu-Ag EAM can be applied in the study of larger Cu islands on Ag(1 1 1). Published by Elsevier B.V.