Descriptor-Based Approach for the Prediction of Cation Vacancy Formation Energies and Transition Levels

Point defects largely determine the observed optical and electrical properties of a given material, yet the characterization and identification of defects has remained a slow and tedious process, both experimentally and theoretically. We demonstrate a computationally-cheap model that can reliably predict the formation energies of cation vacancies as well as the location of their electronic states in a large set of II-VI and III-V materials using only parameters obtained from the bulk primitive unit cell (2-4 atoms). We apply our model to ordered alloys within the CdZnSeTe, CdZnS, and ZnMgO systems and predict the positions of cation vacancy charge-state transition levels with a mean absolute error of < 0.2 eV compared to the explicitly calculated values, showing useful accuracy without the need for the expensive and large-scale calculations typically required. This suggests the properties of other point defects may also be predicted with useful accuracy from only bulk-derived properties.