A genetic algorithm for generating grain boundaries

Generation of realistic grain boundaries (GBs) in bicrystals for atomistic simulations requires a search in a multidimensional space for a configuration with the atoms at the GB organized in such a way that the energy of the whole system is minimized. This paper presents a genetic algorithm that allows one to find low-energy GB configurations by optimizing three main criteria: the local arrangement of the atoms, the relative translation between the two grains that compose the GB and an overall expansion/contraction of the system. It is designed to make a wider and more effective search through the energy landscape compared with other traditional methods, giving access to more configurations and increasing the possibility of finding the global minimum in energy.