Automated workflows predict electronic correlations for high-throughput materials design with HPC-enabled DFPT calculations.
Application sectors: Energy storage, Materials design, Computational chemistry.
Keywords: DFT+U+V, Hubbard parameters, DFPT, Li-ion cathodes, high-throughput.
Modeling strongly correlated electrons in transition-metal compounds is critical for designing batteries and functional materials. Standard DFT often fails for d and f orbitals due to self-interaction errors, affecting electronic, magnetic, and redox predictions. This work introduces aiida-hubbard, an automated framework for self-consistently calculating on-site (U) and intersite (V) Hubbard parameters across Li-bearing compounds. Screening 115 materials, the study finds U can vary up to 6 eV depending on oxidation state and local environment, while V varies 0.2–1.6 eV. High workflow success (91%) demonstrates its reliability for high-throughput predictions of correlated electronic behaviour, surpassing fixed, empirical parameter approaches.
The framework uses density-functional perturbation theory (DFPT) with primitive cells, reducing computational cost compared with supercells. HPC resources enable parallel evaluation over multiple atoms and q-points, cutting wall-clock times dramatically. The Quantum ESPRESSO code, integrated through the AiiDA plugin, manages workflow automation and ensures reproducibility. Technological innovations include the HubbardStructureData type for unified data management and automatic detection of intersite interactions, removing manual parameter tuning.
Self-consistent U and V parameters are essential for accurate electronic property predictions in battery materials and other correlated systems. Aiida-hubbard provides a scalable, robust workflow for high-throughput screening, accelerating the discovery of next-generation energy materials and catalysts.
Code availability:
The code is open source and made available on GitHub (https://github.com/aiidateam/aiida-hubbard). It is also distributed as an installable package through the Python Package Index (https://pypi.org/project/aiida-hubbard/). The base code is open to external contributions for improvements through the GitHub pull request system. The full documentation with tutorials can be found at https://aiida-hubbard.readthedocs.io/en/latest/.
Reference paper
First-principles Hubbard parameters with automated and reproducible workflows.