Prediction and exploration of emission wavelength (or energy) of luminescent materials based on machine learning

Abstract

In the optical field of materials science, it is important to predict the emission wavelength (or energy) of luminescent materials, especially when different dopant ions are involved, which makes the investigation even more complex. The selection of doped ions directly determines the optical properties of luminescent materials, so the accurate prediction of the emission wavelength (or energy) of doped luminescent materials has become a key challenge in scientific research. Traditional theoretical calculation methods often fail to fully consider the complexity of the interactions between ions in different material systems, but machine learning models provide an efficient solution for the research in this field. In this study, we collected a large amount of data of light-emitting materials doped with different ions, combined with their structural feature descriptors, and used a variety of machine learning models to predict the emission wavelength. On the basis of this model we give a prediction of the emission wavelength of the actually synthesized luminous materials in our research group, which are more accurate in the quality of luminous materials doped with Eu^3+, Sm³⁺ plus some Tb³⁺ ions. In the further analysis of the factors affecting the emission wavelength (or energy) of the luminescent materials, we find that the mean first ionization potential, the mean electron affinity and the mean Pauling electronegativity are the key factors. This study shows that machine learning methods have great application potential in wavelength (or energy) prediction of luminous materials and provide an effective tool for material screening and performance optimization in the future.