Unified Graph-based Interatomic Potential for Perovskite Structure Optimization

Halide perovskites hold immense potential for applications such as optoelectronics and catalysis. Their vast compositional space spanning bulk alloys, defects, impurities, surfaces, and surface defects poses significant challenges for efficient exploration and optimization. To address this, we trained a unified graph-based deep learning framework capable of optimizing and predicting energetics across these diverse structural motifs. Using a comprehensive density functional theory dataset of HaP structures containing bulk alloys, native and impurity defects, and surface slabs, we rigorously trained and benchmarked the M3GNet-based machine learning interatomic potential. The M3GNet-IAP framework, trained on DFT-calculated energies, forces, and stresses, enables gradient-based optimization and efficient exploration of the potential energy surface. Our models showed robust generalizability across diverse structural domains and were able to accurately predict the crystal formation energy, perovskite decomposition energy, defect formation energy, and surface energy. Our unified surrogate model provides a holistic approach to geometry optimization across different structural variations in halide perovskites and will be transformative for the discovery of promising new compositions, important defects and dopants, and surface properties.