Charting Lattice Thermal Conductivity of Inorganic Crystals

Abstract

Thermal conductivity is a fundamental material property but challenging to predict, with less than 5% out of about $10^5$ synthesized inorganic materials being documented. In this work, we extract the structural chemistry that governs lattice thermal conductivity, by combining graph neural networks and random forest approaches. We show that both mean and variation of unit-cell configurational properties, such as atomic volume and bond length, are the most important features, followed by mass and elemental electronegativity. We chart the structural chemistry of lattice thermal conductivity into extended van-Arkel triangles, and predict the thermal conductivity of all known inorganic materials in the Inorganic Crystal Structure Database. For the latter, we develop a transfer learning framework extendable for other applications.

References in corpus (5)

• Crystal Graph Convolutional Neural Networks for an Accurate and Interpretable Prediction of Material Properties
• A General-Purpose Machine Learning Framework for Predicting Properties of Inorganic Materials
• Distribution of phonon lifetime in Brillouin zone
• Discovery of low thermal conductivity compounds with first-principles anharmonic lattice dynamics calculations and Bayesian optimization
• Inverse Design of Ultralow Lattice Thermal Conductivity Materials Via Lone Pair Cation Coordination Environment

Cited by in corpus (2)

• Lattice Thermal Conductivity Prediction using Symbolic Regression and Machine Learning
• Screening and understanding Li adsorption on 2-dimensional metallic materials by learning physics