The prediction of inorganic crystal structures using a genetic algorithm and energy minimisation

Abstract

A genetic algorithm has been used to generate plausible crystal structures from the knowledge of only the unit cell dimensions and constituent elements. We successfully generate 38 known binary oxides and various known ternary oxides with the Perovskite, Pyrochlore and Spinel structures, from starting configurations which include no knowledge of the atomic arrangement in the unit cell. The quality of the structures is initially assessed using a cost function which is based on the bond valence model with a number of refinements. The lattice energy, based on the Born model of a solid, is minimised using a local optimiser for the more plausible candidate structures. The method has been implemented within the computational package GULP. An extensive collection of Buckingham potential parameters for use in such simulations on metal oxides is also tabulated.