Prediction of Emission Level Energy of Mn4+ in Oxides Based on the Local Structure Using First-Principles Calculations and Machine Learning

Abstract

The white LED is an important next-generation light source. Most of the high color rendering white LEDs that have become popular in recent years consist of blue LEDs combined with green and red phosphors. For example, Eu2+:SrxCa1-xAlSiN3 (SCASN) and Mn4+:K2SiF6 (KSF) are used as red phosphors. However, considering the production cost and the stability against high temperature and high humidity, novel red phosphors obtained by adding Mn4+ to oxide crystals are still desired. To realize a Mn-doped oxide red phosphor for white LEDs, it is indispensable to clarify the relationship between the multiplet energies and the local structures. Semi-empirical methods based on experimental data are generally used to analyze luminescent materials, but these methods cannot be applied to unknown or hypothetical materials. In this work, we analyzed the relationship between the local structure and the energy levels of hypothetical Mn4+-doped oxides by performing first-principles electronic structure calculations using the discrete-variational multi-electron (DVME) method. The local structures around Mn4+ were systematically changed by changing the bond distances and angles and the multiplet energies corresponding to each local structure were calculated non-empirically. Then the relationship between the energy and the local structure was visually represented by creating energy-structure maps. Using these systematic theoretical values as the training data, a machine learning model for the prediction of the first-principles calculation results based only on the structural parameters was also created. The results of this study would provide guidelines for the development of novel phosphors based on oxide crystals doped with Mn4+.