Machine learned defect and impurity levels in perovskite halides and CdTe*

Electronic levels introduced by impurities and intrinsic defects in the band gap of semiconductors affect optoelectronic performance. Predictions of these defect levels are possible, but expensive, using first principles density functional theory (DFT), and chemical trends are often not easily available. In this talk, we will describe using machine learning (ML) trained on DFT data for defect levels in hybrid perovskite halides [1] and CdTe [2] photovoltaic materials. Relevant descriptors, relative performance of different ML approaches, and insight from resultant ML models will be discussed.

[1] A. Mannodi-Kanakkithodi, et. al., “Comprehensive Computational Study of Partial Lead Substitution in Methylammonium Lead Bromide,” Chemistry of Materials 31, 3599 (2019).
[2] A. Mannodi-Kanakkithodi, et. al., “Machine-learned impurity level prediction for semiconductors: the example of Cd-based chalcogenides,” arXiv:1906.02244.