THE GOAL OF THIS PROPOSED RESEARCH IS TO ACCELERATE THE PREDICTION AND UNDERSTANDING OF DEFECT PROPERTIES IN CDTE AND RELATED MATERIALS USING A COMBINATION OF HIGH-THROUGHPUT DENSITY FUNCTIONAL THEORY (HT-DFT) COMPUTATIONS AND MACHINE LEARNING (ML) MODELS BASED ON GRAPH NEURAL NETWORKS (GNN). THE PRIMARY OBJECTIVE IS TO GENERATE A SUBSTANTIAL DFT DATASET, BUILDING UPON EXISTING DATA FROM THE PI’S WORK, OF FORMATION ENERGIES OF SINGLE AND COMPLEX POINT DEFECTS, DOPANTS, AND EXTENDED DEFECTS, IN A VARIETY OF CD/ZN-TE/SE/S COMPOSITIONS/STRUCTURES, AND TRAIN GNN MODELS TO PREDICT STABILITY OF ANY DEFECTIVE STRUCTURE, EVENTUALLY LEADING TO A LIBRARY OF LOW ENERGY DEFECT CONFIGURATIONS AND LIKELY DEFECT LEVELS THAT EXPLAIN EXPERIMENTAL SPECTROSCOPIC AND OPTICAL OBSERVATIONS IN CDTE PHOTOVOLTAIC (PV) MODULES.