Data-Driven Design of Halide Perovskites for Efficient Green Light-Emitting Diodes via Machine Learning and DFT

To accelerate the application of inorganic halide perovskite materials in green-light-emitting diodes (LEDs), it is crucial to develop novel perovskite materials with suitable band gaps. However, traditional experimental screening methods and density functional theory (DFT) calculations are time-consuming and costly. Therefore, we propose an innovative screening strategy that combines machine learning with DFT calculations to predict the band gaps of 30 inorganic halide perovskite materials. This study establishes a database based on 1193 inorganic perovskite materials and utilizes five machine learning models: random forest regression (RFR), gradient boosting regression (GBR), support vector regression (SVR), extreme gradient boosting regression (XGBR), and decision tree regression (DT). Using this approach, four promising inorganic perovskite candidates, Cs2KTlCl6, Cs2RbTlCl6, Cs2TlBiCl6, and Cs2KInBr6, were successfully identified. Moreover, detailed DFT calculations were conducted to study the band gaps, density of states, effective mass, and exciton binding energy, thereby identifying the most promising green LED perovskite candidates. Among these, Cs2TlBiCl6 not only possesses an ideal band gap of 2.23 eV for green LEDs but also demonstrates excellent performance in terms of effective mass, exciton binding energy, and stability. By combining machine learning with DFT calculations, we significantly enhanced the efficiency and accuracy of material screening, providing an efficient and innovative solution for the development of green LEDs.