Grouping of structures for cluster expansion of multicomponent systems with controlled accuracy

Abstract

Control of errors over the whole range of structures is essential when we combine a large set of density-functional theory calculations and the cluster expansion method for predicting the ground-state structures and configurational thermodynamics of multicomponent systems. Minority structures that are far from a random structure are important for such a prediction. In this paper, we propose a procedure based on the cluster analysis of the structure population, which can adequately take into account the errors of minority structures as well as those of random structures. The usefulness of the procedure is demonstrated by applying it to configurational behaviors of MgAl_{2}O_{4} spinel.