PNNL

Abstract

Information-based materials discovery offers a structured method to evolve materials signatures based upon their physical properties, and to direct searches using performance-based criteria. In this current paper, we focus on the crystal structure aspects of an optical material and construct an information-based model to determine the proclivity of a particular AB composition to exhibit multiple crystal system behavior. Exploratory data methods used both supervised (support-vector machines) and unsupervised (disorder-reduction and principal-component) classification methods for structural signature development; revealing complementary valid signatures. Examination of the relative contributions of the materials chemistry descriptors within these signatures indicates a strong role for Mendeleev number chemistry which must be balanced against the cationic/anionic radius ratio and electronegativity differences of constituents within the unit cell.