Double perovskites are regarded as promising candidates for the next generation of luminescent materials. However, a significant number of double perovskite luminescent materials (DPLMs) remain unexplored. This study integrates the atomistic line graph neural network model with the interpretable extreme gradient boosting model to explore DPLMs, thereby proposing a framework for the screening of high-quality DPLMs. The electronic band structures and the transition dipole moments (TDMs) of 2592 double perovskites are calculated using the Perdew-Burke-Ernzerhof functional to establish a database. The TDM at the direct band gap, experimentally validated, and the exciton binding energy ($E_b$) serve as the criteria for assessing the luminescent capabilities of materials, and the refractive indices serve as the criterion for assessing the application prospects of materials. Based on these conditions, this study identifies five materials with direct band gaps exhibiting luminescent properties and selects two materials with promising high-quality application potential for further validation. Furthermore, this study reveals that the average deviation of number of unfilled p orbital states of the constituent elements in a compound can serve as a preliminary criterion for assessing the luminescent capabilities of materials. This study leverages machine learning to reveal the link between material structure and luminescence, providing a foundation for designing double perovskite light-emitting diodes.

Close

PrevNext

View figure in article

• Band gap
• Luminescence
• Optoelectronics
• Perovskites
• Artificial neural networks
• Density functional calculations
• High-throughput calculations
• Machine learning