Oxygen Vacancy Formation Energy in Metal Oxides: High Throughput Computational Studies and Machine Learning Predictions

TL;DR

This study combines high-throughput DFT calculations and machine learning to predict oxygen vacancy formation energies in over 1000 oxides, enabling efficient materials discovery for energy applications.

Contribution

It provides the largest dataset of computed vacancy energies and develops accurate ML models incorporating site-specific features for predicting these energies.

Findings

Achieved a mean absolute error of ~0.3 eV/O in predictions.

Identified over 250 new double perovskite candidates for water-splitting.

Demonstrated the effectiveness of site-specific features in ML models.

Abstract

The oxygen vacancy formation energy ($\Delta E_{vf}$) governs defect dynamics and is a useful metric to perform materials selection for a variety of applications. However, density functional theory (DFT) calculations of $\Delta E_{vf}$ come at a greater computational cost than the typical bulk calculations available in materials databases due to the involvement of multiple vacancy-containing supercells. As a result, available repositories of direct calculations of $\Delta E_{vf}$ remain relatively scarce, and the development of machine learning models capable of delivering accurate predictions is of interest. In the present, work we address both such points. We first report the results of new high-throughput DFT calculations of oxygen vacancy formation energies of the different unique oxygen sites in over 1000 different oxide materials, which together form the largest dataset of directly computed oxygen vacancy formation energies to date, to our knowledge. We then utilize the resulting dataset of $\sim$2500 $\Delta E_{vf}$ values to train random forest models with different sets of features, examining both novel features introduced in this work and ones previously employed in the literature. We demonstrate the benefits of including features that contain information specific to the vacancy site and account for both cation identity and oxidation state, and achieve a mean absolute error upon prediction of $\sim$0.3 eV/O, which is comparable to the accuracy observed upon comparison of DFT computations of oxygen vacancy formation energy and experimental results. Finally, we demonstrate the predictive power of the developed models in the search for new compounds for solar-thermochemical water-splitting applications, finding over 250 new AA$^{\prime}$BB$^{\prime}$O$_6$ double perovskite candidates.