Photoluminescence Color Prediction for Eu³⁺-Doped Perovskite Red Phosphors Using Machine Learning

Currently, data-driven approaches for exploring novel materials are garnering significant attention with the expectation of accelerating material development cycles and understanding materials from various aspects. This short article presents a supervised prediction model for the emission intensity ratio of D-5(0)-F-7(2) and D-5(0)-F-7(1) transition of Eu3+ ions, called an "asymmetry ratio", which determines the color purity of the red region of Eu3+ photoluminescence in perovskite phosphors. The model is developed using a dataset of 296 samples and 203 descriptors for Eu3+-doped perovskite. The accuracy of the prediction model trained by the dataset is statistically evaluated, which validates its sufficiently high prediction performance. Furthermore, the prediction model's performance is properly assessed by synthesizing a Eu3+-doped NaLaInNbO6 compound, which is unknown as a red phosphor, and by comparing the experimental asymmetry ratio for this compound with that predicted by the predictor, which exhibits satisfactory agreement.