Presentation Information

[9p-S301-5]Exploration of Solar Cell Materials Based on Bandgap Correction via Symbolic Regression

〇Takahiro Kono1, Souta Miyamoto1, Taichi Masuda1, Katsuaki Tanabe1 (1.Kyoto Univ.)

Keywords:

semiconductor,bandgap,machine learning

Using symbolic regression, an equation relating experimentally measured bandgaps to those calculated with densityfunctional theory was derived. Correction values from this equation were employed to construct a machine-learning model that uses physicochemical properties of constituent elements as features. Cu- and Zn-containing ternary compounds were broadly evaluated to narrow down candidates with bandgaps suitable for solar-cell applications. Finally, formation enthalpy was predicted via machine learning to identify materials that are both highly efficient and thermodynamically stable.