Abstract:Density-functional theory with extended Hubbard functionals (DFT+$U$+$V$) provides a robust framework to accurately describe complex materials containing transition-metal or rare-earth elements. It does so by mitigating self-interaction errors inherent to semi-local functionals which are particularly pronounced in systems with partially-filled $d$ and $f$ electronic states. However, achieving accuracy in this approach hinges upon the accurate determination of the on-site $U$ and inter-site $V$ Hubbard parameters. In practice, these are obtained either by semi-empirical tuning, requiring prior knowledge, or, more correctly, by using predictive but expensive first-principles calculations. Here, we present a machine learning model based on equivariant neural networks which uses atomic occupation matrices as descriptors, directly capturing the electronic structure, local chemical environment, and oxidation states of the system at hand. We target here the prediction of Hubbard parameters computed self-consistently with iterative linear-response calculations, as implemented in density-functional perturbation theory (DFPT), and structural relaxations. Remarkably, when trained on data from 11 materials spanning various crystal structures and compositions, our model achieves mean absolute relative errors of 3% and 5% for Hubbard $U$ and $V$ parameters, respectively. By circumventing computationally expensive DFT or DFPT self-consistent protocols, our model significantly expedites the prediction of Hubbard parameters with negligible computational overhead, while approaching the accuracy of DFPT. Moreover, owing to its robust transferability, the model facilitates accelerated materials discovery and design via high-throughput calculations, with relevance for various technological applications.

What problem does this paper attempt to address?

This paper proposes a solution to accurately calculate the Hubbard parameters (U and V) for transition metal or lanthanide materials in complex materials. Within the framework of density functional theory (DFT), the extended Hubbard function (DFT+U+V) can effectively describe these materials, but it requires the precise determination of the Hubbard parameters to reduce the self-interaction errors of the semi-local functional. Currently, these parameters are usually determined through semi-empirical tuning or expensive first-principles calculations. In the paper, the researchers propose a machine learning model based on equivariant neural networks, which uses the atomic occupancy matrix as the descriptor to directly capture the electronic structure, local chemical environment, and oxidation states of the system. The goal is to predict the Hubbard parameters obtained through self-consistent calculations using density functional perturbation theory (DFPT), and perform structure optimization. After training the model on 11 different crystal structures and compositions of materials, the average absolute relative errors for Hubbard U and V parameters are 3% and 5%, respectively. The paper emphasizes that this model significantly speeds up the prediction of Hubbard parameters, with minimal computational cost and accuracy close to DFPT. Furthermore, due to its good transferability, the model helps accelerate the discovery and design of new materials and is suitable for various technological applications.

Machine learning Hubbard parameters with equivariant neural networks

Machine learning the Hubbard U parameter in DFT+U using Bayesian optimization

Predicting structure-dependent Hubbard U parameters via machine learning

Machine Learning Model for the Prediction of Hubbard U Parameters and Its Application to Fe-O Systems.

BMach: a Bayesian machine for optimizing Hubbard U parameters in DFT+U with machine learning

Machine learning density functional theory for the Hubbard model

High-throughput determination of Hubbard and Hund values for transition metal oxides via the linear response formalism

Self-consistent on-site and inter-site Hubbard parameters within DFT+U+V for UO$_2$ using density-functional perturbation theory

Universal Machine Learning Kohn-Sham Hamiltonian for Materials

Magnons from time-dependent density-functional perturbation theory and the noncollinear Hubbard formulation

Efficient Modelling of Anharmonicity and Quantum Effects in PdCuH$_2$ with Machine Learning Potentials

Machine-learned approximations to Density Functional Theory Hamiltonians

Bridging electronic and classical density-functional theory using universal machine-learned functional approximations

On-site and intersite Hubbard corrections in magnetic monolayers: The case of and

Effect of Hubbard U corrections on the electronic and magnetic properties of 2D materials: A high-throughput study

Pure density functional for strong correlations and the thermodynamic limit from machine learning

Standard model of the rare-earths, analyzed from the Hubbard I approximation

Orbital-resolved DFT+U for molecules and solids

Noncollinear DFT+$U$ and Hubbard parameters with fully-relativistic ultrasoft pseudopotentials