Machine-learning correction to density-functional crystal structure optimization

Hussein R, Schmidt J, Barros T, Marques MAL, Botti S (2022)

Publication Type: Journal article

Publication year: 2022

Journal

Mrs Bulletin Cambridge University Press (CUP)

Book Volume: 47

Pages Range: 765-771

Journal Issue: 8

DOI: 10.1557/s43577-022-00310-9

Abstract

Abstract: Density functional theory is routinely applied to predict crystal structures. The most common exchange-correlation functionals used to this end are the Perdew–Burke–Ernzerhof (PBE) approximation and its variant PBEsol. We investigate the performance of these functionals for the prediction of lattice parameters and show how to enhance their accuracy using machine learning. Our data set is constituted by experimental crystal structures of the Inorganic Crystal Structure Database matched with PBE-optimized structures stored in the materials project database. We complement these data with PBEsol calculations. We demonstrate that the accuracy and precision of PBE/PBEsol volume predictions can be noticeably improved a posteriori by employing simple, explainable machine learning models. These models can improve PBE unit cell volumes to match the accuracy of PBEsol calculations, and reduce the error of the latter with respect to experiment by 35 percent. Further, the error of PBE lattice constants is reduced by a factor of 3–5. A further benefit of our approach is the implicit correction of finite temperature effects without performing phonon calculations. Impact statement: Knowledge about the crystal structure of solids is essential for describing their elastic and electronic properties. In particular, their accurate prediction is essential to predict the electronic properties of not-yet-synthesized materials. Lattice parameters are most commonly calculated by density functional theory using the Perdew–Burke–Ernzerhof (PBE) approximation and its variant PBEsol as exchange-correlation functional. They are successful in describing materials properties but do, however, not always achieve the desired accuracy in comparison with experiments. We propose a computationally efficient scheme based on interpretable machine learning to optimize crystal structures. We demonstrate that the accuracy of PBE- and PBEsol-structures can be, therewith, enhanced noticeably. In particular, the PBE unit cells, available in materials databases, can be improved to the level of the more accurate PBEsol calculations and the error of the latter with respect to the experiment can be reduced by 35 percent. An additional advantage of our scheme is the implicit inclusion of finite temperature corrections, which makes expensive phonon calculations unnecessary. Graphical abstract: [Figure not available: see fulltext.]

Involved external institutions

European Theoretical Spectroscopy Facility (ETSF) BE Belgium (BE) Friedrich-Schiller-Universität Jena DE Germany (DE)

How to cite

APA:

Hussein, R., Schmidt, J., Barros, T., Marques, M.A.L., & Botti, S. (2022). Machine-learning correction to density-functional crystal structure optimization. Mrs Bulletin, 47(8), 765-771. https://doi.org/10.1557/s43577-022-00310-9

MLA:

Hussein, Robert, et al. "Machine-learning correction to density-functional crystal structure optimization." Mrs Bulletin 47.8 (2022): 765-771.

BibTeX: Download