We present a graph neural network 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. Here we used density functional theory reference data for vacancy defects in oxides, to train a defect graph neural network (dGNN) model that replaces the density functional theory 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. In conclusion, it will help to tackle future materials discovery problems in clean energy and beyond.
Witman, Matthew D., et al. "Defect graph neural networks for materials discovery in high-temperature clean-energy applications." Nature Computational Science, vol. 3, no. 8, Aug. 2023. https://doi.org/10.1038/s43588-023-00495-2
Witman, Matthew D., Goyal, Anuj, Ogitsu, Tadashi, McDaniel, Anthony H., & Lany, Stephan (2023). Defect graph neural networks for materials discovery in high-temperature clean-energy applications. Nature Computational Science, 3(8). https://doi.org/10.1038/s43588-023-00495-2
@article{osti_1995808,
author = {Witman, Matthew D. and Goyal, Anuj and Ogitsu, Tadashi and McDaniel, Anthony H. and Lany, Stephan},
title = {Defect graph neural networks for materials discovery in high-temperature clean-energy applications},
annote = {We present a graph neural network 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. Here we used density functional theory reference data for vacancy defects in oxides, to train a defect graph neural network (dGNN) model that replaces the density functional theory 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. In conclusion, it will help to tackle future materials discovery problems in clean energy and beyond.},
doi = {10.1038/s43588-023-00495-2},
url = {https://www.osti.gov/biblio/1995808},
journal = {Nature Computational Science},
issn = {ISSN 2662-8457},
number = {8},
volume = {3},
place = {United States},
publisher = {Springer Nature},
year = {2023},
month = {08}}
National Renewable Energy Laboratory (NREL), Golden, CO (United States); Lawrence Livermore National Laboratory (LLNL), Livermore, CA (United States); Sandia National Laboratories (SNL-CA), Livermore, CA (United States)
Sponsoring Organization:
USDOE Office of Energy Efficiency and Renewable Energy (EERE), Office of Sustainable Transportation. Hydrogen Fuel Cell Technologies Office (HFTO); USDOE National Nuclear Security Administration (NNSA); USDOE Laboratory Directed Research and Development (LDRD) Program