Presentation Information
[10a-PA3-7]Generative Models for Multicomponent Crystals and Their Application to Phosphor Materials
〇(M2)Manato Omori1, Hayato Higuchi2, Masaya Miyagawa2, Hiromitu Takaba2 (1.Sch. Adv. Eng. Kogakuin Univ., 2.Chem., Kogakuin Univ.)
Keywords:
WLED,Generative Model,Machine learning
In multicomponent phosphors such as YAG:Ce3+ and β-SiAlON:Eu2+, which are widely used in LEDs and displays, the constituent elements of the host structure influence the electronic states of the luminescent centers, resulting in substantial variations in luminescence properties. However, conventional trial-and-error-based materials exploration requires considerable time and cost. In this study, we investigated DiffCSP, a composition-conditioned crystal structure generative model, by evaluating the structural similarity between generated and experimentally reported crystal structures. Furthermore, we predicted the luminescence properties of the generated structures and explored novel host crystals for Eu2+-activated phosphors.