An energy localization principle and its application to fast kinetic Monte Carlo simulation of heteroepitaxial growth

Simulation of heteroepitaxial growth using kinetic Monte Carlo (KMC) is often based on rates determined by differences in elastic energy between two configurations. This is computationally challenging due to the long range nature of elastic interactions. A new method is introduced in which the elastic field is updated using a local approximation technique. This involves an iterative method that is applied in a sequence of nested domains until a convergence criteria is satisfied. These localized calculations yield energy differences that are highly accurate despite the fact that the energies themselves are far less accurate: an effect referred to as the principle of energy localization. This is explained using the continuum analogue of the discrete model and error estimates are found. In addition, a rejection algorithm that relies on a computationally inexpensive estimate of hopping rates is used to avoid a substantial fraction of the elastic updates. These techniques are applied to 1 + 1-dimensional KMC simulations in physically interesting regimes. (C) 2008 Elsevier Ltd. All rights reserved.