Materials Discovery for High-Temperature Clean-Energy Applications (and Beyond) via Defect Graph Neural Networks

We present a graph neural network (GNN) approach that fully automates the prediction of defect formation enthalpies for any crystallographic site from the ideal crystal structure, without the need to create defected atomic structure models as input [1]. Using density functional theory (DFT) reference data for vacancy defects in oxides, we trained a defect GNN (dGNN) model that replaces the DFT supercell relaxations otherwise required for each symmetrically unique crystal site. Interfaced with thermodynamic calculations of reduction entropies and associated free energies, the dGNN model is applied to the screening of oxides in the Materials Project database, connecting the zero-Kelvin defect enthalpies to high-temperature process conditions relevant for solar thermochemical hydrogen production and other energy applications. The dGNN approach is applicable to arbitrary structures with an accuracy limited principally by the amount and diversity of the training data, and it is generalizable to other defect types and advanced graph convolution architectures. It will aid tackling future materials discovery problems in clean energy and beyond.<br/><br/>[1] Witman, M. et al. <i>Nature Computational Science</i>, <b>3</b>, 675–686 (2023)