Crystal Graph Convolutional Neural Networks for an Accurate and Interpretable Prediction of Material Properties

TLDR

Machine learning for crystalline material design often relies on manually engineered features or complex coordinate transformations, limiting applicability to certain crystal types and hindering chemical insight. The authors propose a crystal graph convolutional neural network framework that learns material properties directly from atomic connectivity, offering a universal and interpretable representation. The framework extracts contributions from local chemical environments to global properties, enabling interpretability, and is trained on 10⁴ data points to predict eight DFT-derived properties across diverse crystal structures. The model achieves high accuracy on these properties and, as demonstrated on perovskites, can reveal empirical design rules by analyzing local environment contributions.

Abstract

The use of machine learning methods for accelerating the design of crystalline materials usually requires manually constructed feature vectors or complex transformation of atom coordinates to input the crystal structure, which either constrains the model to certain crystal types or makes it difficult to provide chemical insights. Here, we develop a crystal graph convolutional neural networks framework to directly learn material properties from the connection of atoms in the crystal, providing a universal and interpretable representation of crystalline materials. Our method provides a highly accurate prediction of density functional theory calculated properties for eight different properties of crystals with various structure types and compositions after being trained with $1{0}^{4}$ data points. Further, our framework is interpretable because one can extract the contributions from local chemical environments to global properties. Using an example of perovskites, we show how this information can be utilized to discover empirical rules for materials design.