Computational prediction of nanostructured alloys with enhanced thermoelectric properties

The Materials Genome Initiative calls for a dramatic increase in the rate of materials discovery and development. High-throughput (HT) calculations can advance this goal by efficiently screening a large search space for candidate materials to study in more depth. Thermoelectric materials (TEs) are prime candidates for such HT calculations: The properties required to achieve good performance are known, but systematic ways of improving these properties are scarce. Furthermore, known HT methods for TEs only address bulk crystals - screening realistic multicomponent alloys for their TE properties has yet to be accomplished. In this paper, we use a density functional theory driven HT screening-and-sorting procedure to search for new multicomponent bulk-nanostructured thermoelectric materials. We make maximum use of minimal calculations to obtain eight descriptors of the thermodynamics and TE performance of five-element semiconductor alloy systems from combinations of ternary additions in binary compounds. We use these descriptors to reduce a search space of 29 700 five-element systems to a set of 130 candidates. We screen these candidates using TE descriptors to identify several existing high-performance thermoelectrics as well as promising new material systems awaiting further experimental verification.