Prediction of Emission Energy of d3 Ions in Oxides Based on Local Structures Using First-Principles Calculations and Machine Learning

Abstract

Listen

Translate article

Red phosphors activated with Mn4+ are attracting attention as potential phosphor materials that can improve the color rendering properties of white LEDs. However, since Mn4+: KSF which is currently in practical use is unstable and toxic, a special surface treatment is required in actual usage. Therefore, in order to theoretically design stable and inexpensive red phosphors, we focused on oxide phosphors activated with Mn4+ and other d3 ions. Since emission energies of Mn4+ in oxides are always lower than the desired value, it is important to explore local structures which realize higher emission energies. For this purpose, we have recently created emission energy maps based on local structures using systematic first-principles calculations for CrO6 and MnO6 clusters with D4h and C4v symmetries. In this work, in order to investigate the influence of the ionic species, we created the emission energy maps of the other d3 ions such as V2+ and Fe5+. We constructed VO6 clusters and FeO6 clusters with gradually changed local structures and performed systematic first-principles calculations using the DV-Xα method and the DVME method. The emission energy maps of VO6 and FeO6 based on local structures were created and compared to those of CrO6 and MnO6. By using the obtained multiplet energies as the training data, we also created predictive models of emission energies of d3 ions in oxides by machine learning.