Machine Learning for Halide Perovskite Materials ABX3 (B = Pb, X = I, Br, Cl) Assessment of Structural Properties and Band Gap Engineering for Solar Energy

The exact control of material properties essential for solar applications has been made possible because of perovskites' compositional engineering. However, tackling efficiency, stability, and toxicity at the same time is still a difficulty. Mixed lead-free and inorganic perovskites have lately shown promise in addressing these problems, but their composition space is vast, making it challenging to find good candidates even with high-throughput approaches. We investigated two groups of halide perovskite compound data with the ABX(3) formula to investigate the formation energy data for 81 compounds. The structural stability was analyzed over 63 compounds. For these perovskites, we used new library data extracted from a calculation using generalized-gradient approximation within the Perdew-Burke-Ernzerhof (PBE) functional established on density functional theory. As a second step, we built machine learning models, based on a kernel-based naive Bayes algorithm that anticipate a variety of target characteristics, including the mixing enthalpy, different octahedral distortions, and band gap calculations. In addition to laying the groundwork for observing new perovskites that go beyond currently available technical uses, this work creates a framework for finding and optimizing perovskites in a photovoltaic application.

Automated summary

The control of material properties is crucial for solar applications, and perovskites' compositional engineering has made it possible. However, addressing efficiency, stability, and toxicity simultaneously remains challenging. Mixed lead-free and inorganic perovskites have shown promise in addressing these issues, but finding suitable compounds within their vast composition space is difficult. This study investigates the formation energy and structural stability of 81 compounds with the ABX(3) formula, using new library data extracted from calculations based on density functional theory. Machine learning models were then built to predict various target characteristics, providing a framework for finding and optimizing perovskites in photovoltaic applications.