Predicting large area surface reconstructions using molecular dynamics methods

In this paper we discuss a new simulation method that can be used to predict preferred surface reconstructions of model systems by Molecular Dynamics (MD). The method overcomes the limitations imposed by periodic boundary conditions for finite boundary MD simulations which can normally prevent reconstructions. By simulating only the reconstructed surface layer and by removing the periodic boundary effects and the free energy barriers to reconstruction, the method allows surfaces to reconstruct to a preferred structure. We test the method on three types of surfaces: (i) the Au(100) and Pt(100) hexagonally reconstructed surface, (ii) the Au(111) herringbone surfaces, and (iii) the triangularly reconstructed Ag surface layer on a Pt(111) substrate and find the method readily finds lower surface energy reconstructions as preferred by the potential.