Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • Article
  • Published:

Defect graph neural networks for materials discovery in high-temperature clean-energy applications

Nature Computational Science volume 3, pages 675–686 (2023)Cite this article

Subjects

A preprint version of the article is available at ChemRxiv.

Abstract

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. It will help to tackle future materials discovery problems in clean energy and beyond.

This is a preview of subscription content, access via your institution

Access options

Buy this article

  • Purchase on Springer Link
  • Instant access to full article PDF
Buy now

Prices may be subject to local taxes which are calculated during checkout

Fig. 1: Defect training data acquisition.
Fig. 2: Model CV performance.
Fig. 3: Model validation on known water splitters.
Fig. 4: High-throughput materials discovery for STCH.
Fig. 5: Materials discovery for CO2 conversion and SOFC.
Fig. 6: Finite temperature defect calculations.

Similar content being viewed by others

Data availability

The pre-trained models, datasets generated during and/or analyzed during the current study, and scripts to post-process them and reproduce the manuscript figures, are available in the Zenodo repository: ‘A database of vacancy formation enthalpies for materials discovery’ at https://doi.org/10.5281/zenodo.8087871 (ref. 65). Source data are provided with this paper.

Code availability

The open-source dGNN code for training models the models presented in this article has been distributed on the Paper1 branch at the following GitHub repository (https://github.com/mwitman1/cgcnndefect/tree/Paper1)66, which is a modified fork of the original open-source CGCNN code (https://github.com/txie-93/cgcnn).

References

  1. Muhich, C. L. et al. A review and perspective of efficient hydrogen generation via solar thermal water splitting. WIREs Energy Environ. 5, 261–287 (2016).

    Article  Google Scholar 

  2. Komsa, H. P. et al. Two-dimensional transition metal dichalcogenides under electron irradiation: defect production and doping. Phys. Rev. Lett. 109, 1–5 (2012).

    Article  Google Scholar 

  3. Jeen, H. et al. Reversible redox reactions in an epitaxially stabilized SrCoOx oxygen sponge. Nat. Mater. 12, 1057–1063 (2013).

    Article  Google Scholar 

  4. Menéndez, C., Chu, D. & Cazorla, C. Oxygen-vacancy induced magnetic phase transitions in multiferroic thin films. npj Comput. Mater. 6, 1–9 (2020).

    Article  Google Scholar 

  5. Ong, S. P. et al. Python Materials Genomics (pymatgen): a robust, open-source Python library for materials analysis. Comput. Mater. Sci. 68, 314–319 (2013).

    Article  Google Scholar 

  6. Jain, A. et al. The Materials Project: a materials genome approach to accelerating materials innovation. APL Mater. 1, 011002 (2013).

    Article  Google Scholar 

  7. Stevanović, V., Lany, S., Zhang, X. & Zunger, A. Correcting density functional theory for accurate predictions of compound enthalpies of formation: fitted elemental-phase reference energies. Phys. Rev. B 85, 115104 (2012).

    Article  Google Scholar 

  8. Choudhary, K. et al. The joint automated repository for various integrated simulations (JARVIS) for data-driven materials design. npj Comput. Mater. 6, 173 (2020).

    Article  Google Scholar 

  9. Bertoldo, F., Ali, S., Manti, S. & Thygesen, K. S. Quantum point defects in 2D materials - the QPOD database. npj Comput. Mater. 8, 56 (2022).

    Article  Google Scholar 

  10. Saal, J. E., Kirklin, S., Aykol, M., Meredig, B. & Wolverton, C. Materials design and discovery with high-throughput density functional theory: the Open Quantum Materials Database (OQMD). JOM 65, 1501–1509 (2013).

    Article  Google Scholar 

  11. Deml, A. M., Holder, A. M., O’Hayre, R. P., Musgrave, C. B. & Stevanović, V. Intrinsic material properties dictating oxygen vacancy formation energetics in metal oxides. J. Phys. Chem. Lett. 6, 1948–1953 (2015).

    Article  Google Scholar 

  12. Varley, J. B., Samanta, A. & Lordi, V. Descriptor-based approach for the prediction of cation vacancy formation energies and transition levels. J. Phys. Chem. Lett. 8, 5059–5063 (2017).

    Article  Google Scholar 

  13. Goyal, A. et al. On the dopability of semiconductors and governing material properties. Chem. Mater. 32, 4467–4480 (2020).

    Article  Google Scholar 

  14. Frey, N. C., Akinwande, D., Jariwala, D. & Shenoy, V. B. Machine learning-enabled design of point defects in 2D materials for quantum and neuromorphic information processing. ACS Nano 14, 13406–13417 (2020).

    Article  Google Scholar 

  15. Wan, Z., Wang, Q. D., Liu, D. & Liang, J. Data-driven machine learning model for the prediction of oxygen vacancy formation energy of metal oxide materials. Phys. Chem. Chem. Phys. 23, 15675–15684 (2021).

    Article  Google Scholar 

  16. Wexler, R. B., Gautam, G. S., Stechel, E. B. & Carter, E. A. Factors governing oxygen vacancy formation in oxide perovskites. J. Am. Chem. Soc. 143, 13212–13227 (2021).

    Article  Google Scholar 

  17. Mannodi-Kanakkithodi, A. et al. Universal machine learning framework for defect predictions in zinc blende semiconductors. Patterns 3, 100450 (2022).

    Article  Google Scholar 

  18. Xie, T. & Grossman, J. C. Crystal graph convolutional neural networks for an accurate and interpretable prediction of material properties. Phys. Rev. Lett. 120, 145301 (2018).

    Article  Google Scholar 

  19. Park, C. W. & Wolverton, C. Developing an improved crystal graph convolutional neural network framework for accelerated materials discovery. Phys. Rev. Mater. 4, 063801 (2020).

    Article  Google Scholar 

  20. Schütt, K., Unke, O. & Gastegger, M. Equivariant message passing for the prediction of tensorial properties and molecular spectra. In Proc. 38th International Conference on Machine Learning Vol. 139, 9377–9388 (2021).

  21. Unke, O. T. et al. SpookyNet: learning force fields with electronic degrees of freedom and nonlocal effects. Nat. Commun. 12, 7273 (2021).

    Article  Google Scholar 

  22. Fung, V., Zhang, J., Juarez, E. & Sumpter, B. G. Benchmarking graph neural networks for materials chemistry. npj Comput. Mater. 7, 1–8 (2021).

    Article  Google Scholar 

  23. McDaniel, A. H. et al. Sr- and Mn-doped LaAlO3−δ for solar thermochemical H2 and CO production. Energy Environ. Sci. 6, 2424 (2013).

    Article  Google Scholar 

  24. Gorman, B. T., Lanzarini-Lopes, M., Johnson, N. G., Miller, J. E. & Stechel, E. B. Techno-economic analysis of a concentrating solar power plant using redox-active metal oxides as heat transfer fluid and storage media. Front. Energy Res. https://doi.org/10.3389/fenrg.2021.734288 (2021).

  25. Lany, S. The electronic entropy of charged defect formation and its impact on thermochemical redox cycles. J. Chem. Phys. 148, 071101 (2018).

    Article  Google Scholar 

  26. Park, J. E. et al. Computationally accelerated discovery and experimental demonstration of Gd0.5La0.5Co0.5Fe0.5O3 for solar thermochemical hydrogen production. Front. Energy Res. https://doi.org/10.3389/fenrg.2021.750600 (2021).

  27. Sai Gautam, G., Stechel, E. B. & Carter, E. A. A first-principles-based sub-lattice formalism for predicting off-stoichiometry in materials for solar thermochemical applications: the example of ceria. Adv. Theory Simul. 3, 2000112 (2020).

    Article  Google Scholar 

  28. Millican, S. L., Clary, J. M., Musgrave, C. B. & Lany, S. Redox defect thermochemistry of FeAl2O4 hercynite in water splitting from first-principles methods. Chem. Mater. 34, 519–528 (2022).

    Article  Google Scholar 

  29. Vieten, J. et al. Materials design of perovskite solid solutions for thermochemical applications. Energy Environ. Sci. 12, 1369–1384 (2019).

    Article  Google Scholar 

  30. Maiti, D. et al. Earth abundant perovskite oxides for low temperature CO2 conversion. Energy Environ. Sci. 11, 648–659 (2018).

    Article  Google Scholar 

  31. Tezsevin, I., van de Sanden, M. C. M. & Er, S. High-throughput computational screening of cubic perovskites for solid oxide fuel cell cathodes. J. Phys. Chem. Lett. 12, 4160–4165 (2021).

    Article  Google Scholar 

  32. Belsky, A., Hellenbrandt, M., Karen, V. L. & Luksch, P. New developments in the Inorganic Crystal Structure Database (ICSD): accessibility in support of materials research and design. Acta Crystallogr. Sect. B 58, 364–369 (2002).

    Article  Google Scholar 

  33. Goyal, A., Gorai, P., Peng, H., Lany, S. & Stevanović, V. A computational framework for automation of point defect calculations. Comput. Mater. Sci. 130, 1–9 (2017).

    Article  Google Scholar 

  34. Blöchl, P. E. Projector augmented-wave method. Phys. Rev. B 50, 17953–17979 (1994).

    Article  Google Scholar 

  35. Kresse, G. From ultrasoft pseudopotentials to the projector augmented-wave method. Phys. Rev. B 59, 1758–1775 (1999).

    Article  Google Scholar 

  36. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation made simple. Phys. Rev. Lett. 77, 3865–3868 (1996).

    Article  Google Scholar 

  37. Dudarev, S. L., Botton, G. A., Savrasov, S. Y., Humphreys, C. J. & Sutton, A. P. Electron-energy-loss spectra and the structural stability of nickel oxide: an LSDA+U study. Phys. Rev. B 57, 1505–1509 (1998).

    Article  Google Scholar 

  38. Sharan, A. & Lany, S. Computational discovery of stable and metastable ternary oxynitrides. J. Chem. Phys. 154, 234706 (2021).

    Article  Google Scholar 

  39. Pavone, M., Munoz-Garcia, A. B., Ritzmann, A. M. & Carter, E. A. First-principles study of lanthanum strontium manganite: Insights into electronic structure and oxygen vacancy formation. J. Phys. Chem. C 118, 13346–13356 (2014).

    Article  Google Scholar 

  40. Santana, J. A., Krogel, J. T., Kent, P. R. & Reboredo, F. A. Diffusion quantum Monte Carlo calculations of SrFeO3 and LaFeO3. J. Chem. Phys. 147, 034701 (2017).

    Article  Google Scholar 

  41. Peng, H. et al. Convergence of density and hybrid functional defect calculations for compound semiconductors. Phys. Rev. B 88, 115201 (2013).

    Article  Google Scholar 

  42. Lany, S. Semiconductor thermochemistry in density functional calculations. Phys. Rev. B 78, 245207 (2008).

    Article  Google Scholar 

  43. Emery, A. A., Saal, J. E., Kirklin, S., Hegde, V. I. & Wolverton, C. High-throughput computational screening of perovskites for thermochemical water splitting applications. Chem. Mater. 28, 5621–5634 (2016).

    Article  Google Scholar 

  44. Cheng, Y. et al. Vacancy formation energy and its connection with bonding environment in solid: a high-throughput calculation and machine learning study. Comput. Mater. Sci. 183, 109803 (2020).

    Article  Google Scholar 

  45. Choudhary, K. & Sumpter, B. G. A deep-learning model for fast prediction of vacancy formation in diverse materials. Preprint at arXiv https://doi.org/10.48550/arXiv.2205.08366 (2022).

  46. Witman, M. D., Goyal, A., Ogitsu, T., McDaniel, A. H. & Lany, S. Graph neural network modeling of vacancy formation enthalpy for materials discovery and its application in solar thermochemical water splitting. Preprint at ChemRxiv https://doi.org/10.26434/chemrxiv-2022-frcns (2022).

  47. Meredig, B. & Wolverton, C. First-principles thermodynamic framework for the evaluation of thermochemical H2O- or CO2-splitting materials. Phys. Rev. B 80, 245119 (2009).

    Article  Google Scholar 

  48. Barcellos, D. R., Sanders, M. D., Tong, J., McDaniel, A. H. & O’Hayre, R. P. BaCe0.25Mn0.75O3−δ - a promising perovskite-type oxide for solar thermochemical hydrogen production. Energy Environ. Sci. 11, 3256–3265 (2018).

    Article  Google Scholar 

  49. Bergeson-Keller, A. M., Sanders, M. D. & O’Hayre, R. P. Reduction thermodynamics of Sr1−xCexMnO3 and CexSr2−xMnO4 perovskites for solar thermochemical hydrogen production. Energy Technol. 10, 2100515 (2022).

    Article  Google Scholar 

  50. Zunger, A., Wei, S.-H., Ferreira, L. G. & Bernard, J. E. Special quasirandom structures. Phys. Rev. Lett. 65, 353–356 (1990).

    Article  Google Scholar 

  51. Titov, Y. et al. Effect of isovalent substitution on the crystal structure and properties of two-slab indates BaLa2−xSmxIn2O7. Open Chem. 18, 1294–1303 (2020).

    Article  Google Scholar 

  52. Tirmali, P. M. et al. Structural, magnetic and dielectric relaxation behaviour study of La2MnCoO6 and fully substituted b-site La2FeCoO6. J. Chin. Adv. Mater. Soc. 6, 207–221 (2018).

    Article  Google Scholar 

  53. Zhang, C. et al. Ferromagnetic Y2CoMnO6: spin-glass-like behavior and dielectric relaxation. J. Electron. Mater. 43, 1071–1075 (2014).

    Article  Google Scholar 

  54. Mevs, H. & Müller-Buschbaum, H. Neue verbindungen mit Ba6Ln2M43+O15-TYP: Ba6Nd2Fe4O15, Ba5SrLa2Fe4O15 und Ba5SrNd2Fe4O15. J. Less Common Met. 158, 147–152 (1990).

    Article  Google Scholar 

  55. Suescun, L., Chmaissem, O., Mais, J., Dabrowski, B. & Jorgensen, J. D. Crystal structures, charge and oxygen-vacancy ordering in oxygen deficient perovskites SrMnOx (x < 2.7). J. Solid State Chem. 180, 1698–1707 (2007).

    Article  Google Scholar 

  56. Nair, M. M. & Abanades, S. Experimental screening of perovskite oxides as efficient redox materials for solar thermochemical CO2 conversion. Sustain. Energy Fuels 2, 843–854 (2018).

    Article  Google Scholar 

  57. Fuks, D., Mastrikov, Y., Kotomin, E. & Maier, J. Ab initio thermodynamic study of (Ba,Sr)(Co,Fe)O3 perovskite solid solutions for fuel cell applications. J. Mater. Chem. A 1, 14320 (2013).

    Article  Google Scholar 

  58. Kim, Y.-M. et al. Probing oxygen vacancy concentration and homogeneity in solid-oxide fuel-cell cathode materials on the subunit-cell level. Nat. Mater. 11, 888–894 (2012).

    Article  Google Scholar 

  59. Lee, Y.-L., Kleis, J., Rossmeisl, J., Shao-Horn, Y. & Morgan, D. Prediction of solid oxide fuel cell cathode activity with first-principles descriptors. Energy Environ. Sci. 4, 3966–3970 (2011).

    Article  Google Scholar 

  60. Gokon, N., Yawata, T., Bellan, S., Kodama, T. & Cho, H.-S. Thermochemical behavior of perovskite oxides based on LaxSr1−x-(Mn, Fe, Co)-O3−δ and BaySr1−yCoO3−δ redox system for thermochemical energy storage at high temperatures. Energy 171, 971–980 (2019).

    Article  Google Scholar 

  61. Xiang, D., Gu, C., Xu, H. & Xiao, G. Self-assembled structure evolution of Mn-Fe oxides for high temperature thermochemical energy storage. Small 17, 2101524 (2021).

    Article  Google Scholar 

  62. Naghavi, S. S. et al. Giant onsite electronic entropy enhances the performance of ceria for water splitting. Nat. Commun. 8, 285 (2017).

    Article  Google Scholar 

  63. Chueh, W. C. et al. High-flux solar-driven thermochemical dissociation of CO2 and H2O using nonstoichiometric ceria. Science 330, 1797–1801 (2010).

    Article  Google Scholar 

  64. McDaniel, A. H. Renewable energy carriers derived from concentrating solar power and nonstoichiometric oxides. Curr. Opin. Green Sustainable Chem. 4, 37–43 (2017).

    Article  Google Scholar 

  65. Witman, M., Goyal, A., Ogitsu, T., McDaniel, A. H. & Lany, S. A database of vacancy formation enthalpies for materials discovery (0.0.1) [data set]. Zenodo https://doi.org/10.5281/zenodo.8087871 (2023).

  66. Witman, M. mwitman1/cgcnndefect: release for paper1 (v0.0.0). Zenodo https://doi.org/10.5281/zenodo.8051401 (2023).

  67. Paszke, A. et al. PyTorch: an imperative style, high-performance deep learning library. In Adv. Neural Information Processing Systems Vol. 32, 8024–8035 (2019).

  68. Zhang, C., Bengio, S., Hardt, M., Recht, B. & Vinyals, O. Understanding deep learning (still) requires rethinking generalization. Commun. ACM 64, 107–115 (2021).

    Article  Google Scholar 

  69. Bengio, Y. & Grandvalet, Y. No unbiased estimator of the variance of K-fold cross-validation. J. Mach. Learn. Res. 5, 1089–1105 (2004).

  70. Aykol, M., Dwaraknath, S. S., Sun, W. & Persson, K. A. Thermodynamic limit for synthesis of metastable inorganic materials. Sci. Adv. 4, eaaq0148 (2018).

    Article  Google Scholar 

Download references

Acknowledgements

This material is based on work supported by the US Department of Energy (DOE), Office of Energy Efficiency and Renewable Energy (EERE), specifically the Hydrogen and Fuel Cell Technologies Office. Sandia National Laboratories is a multi-mission laboratory managed and operated by National Technology and Engineering Solutions of Sandia, LLC (NTESS), a wholly owned subsidiary of Honeywell International Inc., for the US Department of Energy’s National Nuclear Security Administration (DOE/NNSA) under contract DE-NA0003525. This written work is authored by an employee of NTESS. The employee, not NTESS, owns the right, title and interest in and to the written work and is responsible for its contents. Part of this work was supported by the Sandia Laboratory Directed Research and Development program. Part of the work was performed under the auspices of the US Department of Energy by Lawrence Livermore National Laboratory under contract DE-AC52-07NA27344. The National Renewable Energy Laboratory (NREL) is operated by the Alliance for Sustainable Energy, LLC, for the DOE under contract DE-AC36-08GO28308. This work used high-performance computing resources at NREL, sponsored by DOE-EERE. Any subjective views or opinions that might be expressed in the written work do not necessarily represent the views of the US government. The publisher acknowledges that the US government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this written work or allow others to do so, for US government purposes. The DOE will provide public access to results of federally sponsored research in accordance with the DOE Public Access Plan.

Author information

Author notes

  1. These authors contributed equally: Matthew D. Witman, Anuj Goyal.

Authors and Affiliations

  1. Sandia National Laboratories, Livermore, CA, USA

    Matthew D. Witman & Anthony H. McDaniel

  2. National Renewable Energy Laboratory, Golden, CO, USA

    Anuj Goyal & Stephan Lany

  3. Indian Institute of Technology Hyderabad, Kandi, Telangana, India

    Anuj Goyal

  4. Lawrence Livermore National Laboratory, Livermore, CA, USA

    Tadashi Ogitsu

Authors
  1. Matthew D. Witman
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Anuj Goyal
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Tadashi Ogitsu
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Anthony H. McDaniel
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Stephan Lany
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

M.D.W., A.G., T.O., A.H.M. and S.L. conceptualized the study. Methodology, software and validation were the responsibility of M.D.W., A.G. and S.L.; M.D.W., A.G. and S.L. completed formal analyses and investigations. M.D.W., A.G. and S.L. curated the data. A.H.M. and S.L. acquired the funding. M.W., A.G. and S.L. wrote the original draft. All authors reviewed and edited the paper.

Corresponding authors

Correspondence to Matthew D. Witman, Anthony H. McDaniel or Stephan Lany.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Computational Science thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editor: Kaitlin McCardle, in collaboration with the Nature Computational Science team. Peer reviewer reports are available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Supplementary Sections 1–10, Tables 1–5 and Figs. 1–7.

Peer Review File

Supplementary Data 1

Summary of training compounds. Additional summary information of the compounds investigated with DFT and their supercell properties used to compute vacancy formation energies, as summarized in Supplementary Section 1.

Supplementary Data 2

Final summary data for MP screened oxides. Promising STCH candidates with summary information as described in Supplementary Section 10.

Source data

Source data for Fig. 2

(x, y) data of the various subplots.

Source data for Fig. 3

(x, y) data of the various subplots.

Source data for Fig. 4

(x, y) data of the various subplots.

Source data for Fig. 5

(x, y) data of the various subplots.

Source data for Fig. 6

(x, y) data of the various subplots.

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Witman, M.D., Goyal, A., Ogitsu, T. et al. Defect graph neural networks for materials discovery in high-temperature clean-energy applications. Nat Comput Sci 3, 675–686 (2023). https://doi.org/10.1038/s43588-023-00495-2

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/s43588-023-00495-2

Search

Advanced search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing