The Materials genome : rapid materials screening for renewable energy using high-throughput density functional theory

Abstract

This thesis relates to the emerging field of high-throughput density functional theory (DFT) computation for materials design and optimization. Although highthroughput DFT is a promising new method for materials discovery, its practical implementation can be difficult. This thesis describes in detail a software infrastructure used to perform over 80,000 DFT computations. Accurately calculating total energies of diverse chemistries is an ongoing effort in the electronic structure community. We describe a method of mixing total energy calculations from different energy functionals (e.g., GGA and GGA+U) so that highthroughput calculations can be more accurately applied over a wide chemical space. Having described methods to perform accurate and rapid DFT calculations, we move next to applications. A first application relates to finding sorbents for Hg gas removal for Integrated Gas Combined Cycle (IGCC) power plants. We demonstrate that rapid computations of amalgamation and oxidation energies can identify the most promising metal sorbents from a candidate list. In the future, more extensive candidate lists might be tested. A second application relates to the design and understanding of Li ion battery cathodes. We compute some properties of about 15,000 virtual cathode materials to identify a new cathode chemistry, Li₉V₃(P₂O₇)₃(PO₄)₂ . This mixed diphosphate-phosphate material was recently synthesized by both our research group and by an outside group. We perform an in-depth computational study of Li₉V₃(P₂O₇)₃(PO₄)₂ and suggest Mo doping as an avenue for its improvement. A major concern for Li ion battery cathodes is safety with respect to 02 release. By examining our large data set of computations on cathode materials, we show that i) safety roughly decreases with increasing voltage and ii) for a given redox couple, polyanion groups reduce safety. These results suggest important limitations for researchers designing high-voltage cathodes. Finally, this thesis describes the beginnings of a highly collaborative 'Materials Genome' web resource to share our calculated results with the general materials community. Through the Materials Genome, we expect that the work presented in this thesis will not only contribute to the applications discussed herein, but help make high-throughput computations accessible to the broader materials community.