UC Berkeley

Atomic scale imperfections, know as point defects, dictate the performance and efficiency of many modern energy materials. The challenges of climate change require the continued improvement and identification of materials which can have point defects engineered in a favorable fashion for the application of interest. With the rise of computer-aided materials design, the possibility of performing high-throughput, first-principles computation on point defects remains an attractive direction for improved screening of new energy materials. Yet major barriers have prevented the large scale implementation of point defect calculations in non-metals - namely, errors arising from the finite size of the computation cell, compounded with errors associated with the underestimation of the band gap. Moreover, the lack of an organized computational framework for storing and analyzing such calculations, as well as the lack of a reliable benchmark for understanding the quantities which can be reliably computed, have prevented high-throughput point defect calculations from being performed in a practical context. In this dissertation, the notion of performing first principles calculations of point defects in semiconductors and insulators in a high-throughput format is investigated. This begins with an overview of the theoretical requirements for performing first principles computation of defects, as well as the presentation of a set of open source command line tools for doing the same. Then three different application areas in the energy space - thermoelectrics for waste heat recovery, solid state electrolyte batteries, and solar cells - are explored with first principles calculations of point defects. Finally, scaling of the previously presented command line tools to a fully automated framework is demonstrated and used for a large benchmark study of fully-automated point defect calculations with semi-local functionals as compared to a set of previously published point defect calculations with hybrid-functionals. This benchmark work outlines the strengths and weaknesses associated with such an automation framework, and advocates for the use of such a framework for qualitative screening of doping limits and general carrier type for high-throughput computational discovery and design of new energy materials.