Scale-invariant machine-learning model accelerates the discovery of quaternary chalcogenides with ultralow lattice thermal conductivity

Abstract We design an advanced machine-learning (ML) model based on crystal graph convolutional neural network that is insensitive to volumes (i.e., scale) of the input structures discover novel quaternary chalcogenides, AMM′Q 3 (A/M/M ' = alkali, alkaline earth, post-transition metals, lanthanides, and Q chalcogens). These compounds are shown possess ultralow lattice thermal conductivity ( κ l ), a desired requirement for thermal-barrier coatings thermoelectrics. Upon screening thermodynamic stability ~1 million using ML iteratively performing density-functional theory (DFT) calculations small fraction compounds, we 99 validated be stable in DFT. Taking several DFT-stable calculate their Peierls–Boltzmann transport equation, which reveals (<2 Wm −1 K at room temperature) due soft elasticity strong phonon anharmonicity. Our work demonstrates high efficiency scale-invariant predicting presents experimental-research opportunities with these new compounds.