Density Functional Theory Plus Dynamical Mean Field Theory within the Framework of Linear Combination of Numerical Atomic Orbitals: Formulation and Benchmarks

Xin Qu, Peng Xu, Rusong Li, Gang Li, Lixin He, Xinguo RenJournal of Chemical Theory and Computation 2022 · DOI

TL;DR

This paper presents a new, flexible framework for combining Density Functional Theory (DFT) with Dynamical Mean-Field Theory (DMFT) using a Linear Combination of Numerical Atomic Orbitals (NAO) basis set. This approach makes advanced DFT+DMFT calculations more accessible and robust, enabling accurate predictions for a wide range of strongly correlated materials.

What Problem Does This Paper Address?

Strongly correlated materials, where electron-electron interactions are dominant, exhibit fascinating phenomena like magnetism and unconventional conductivity. While standard DFT methods struggle to accurately describe these materials, the combination of DFT with DMFT offers a powerful first-principles approach. However, implementing DFT+DMFT can be computationally expensive and technically challenging, especially within the versatile NAO basis set framework, which is widely used in modern DFT codes. This paper addresses the need for a robust and user-friendly implementation of DFT+DMFT within the NAO basis, bridging the gap between advanced electronic structure theory and practical materials science applications.

Methods

The core contribution of this work is the development of a formalism to seamlessly integrate NAO-based DFT codes with DMFT impurity solvers. This involves defining a projection scheme that effectively maps the electronic states between the full DFT Hilbert space and the localized subspace relevant for DMFT. The authors utilize a "projector method" where the KS wavefunctions are projected onto a set of localized NAO basis functions, which are then orthonormalized. This allows for the construction of the local Green function, a key quantity for the DMFT self-consistency loop. The implementation is tested by interfacing two popular NAO-based DFT codes (FHI-aims and ABACUS) with three different DMFT impurity solvers (PACS, iQIST, and Rutgers).

Datasets

The paper does not introduce new datasets. Instead, it uses established experimental data and theoretical results for benchmarking. The materials studied include:

  • 3d transition metal compounds: SrVO3, MnO, NiO
  • 4f lanthanides: Ce metal, Ce2O3
  • 5f actinides: Pu2O3, PuO2

These materials represent a diverse range of strongly correlated systems, from metals to insulators, and are well-studied in the literature, providing a solid basis for validation.

Key Insights and Ideas

  • Bridging the gap: The development of a general formalism for DFT+DMFT within the NAO basis set is a significant step towards making these advanced calculations more accessible and widely applicable.
  • Robustness across solvers and codes: The remarkable agreement in results obtained using different DFT codes and DMFT impurity solvers for many systems highlights the robustness of the developed framework. This suggests that the underlying physics is well captured, despite variations in implementation details.
  • Importance of local orbitals: The choice of localized d or f-type NAOs as the correlated subspace is physically intuitive and technologically straightforward within the NAO framework, simplifying the setup of DFT+DMFT calculations.
  • Beyond standard DFT: The paper demonstrates the necessity of DFT+DMFT for accurately describing the electronic properties of strongly correlated materials, where standard DFT methods fail to capture crucial phenomena like Mott insulating gaps.

Key Results

The authors successfully implemented and benchmarked their DFT+DMFT formalism on a variety of strongly correlated materials. Key findings include:

  • SrVO3: The calculations accurately reproduce the characteristic three-peak structure of the spectral function, including the quasi-particle band and the lower and upper Hubbard bands, which are missed by standard DFT. The results show good agreement with previous theoretical and experimental data.
  • MnO and NiO: The DFT+DMFT approach successfully opens the insulating gaps in these materials, which are incorrectly predicted as metallic by standard DFT. The spectral functions show good agreement with experimental observations, although some dependence on the specific DFT code and impurity solver was noted for NiO.
  • Ce metal and Ce2O3: For Ce metal, the characteristic three-peak structure of correlated metals is reproduced, and the transition between different phases is captured. For Ce2O3, a Mott insulator, the calculated band gap agrees well with experimental and previous theoretical results.
  • Pu2O3 and PuO2: The DFT+DMFT calculations predict band gaps for these plutonium oxides that are in good agreement with experimental values. The analysis also provides insights into the localization of Pu 5f electrons, which is sensitive to the chemical environment.
  • Robustness: For SrVO3 and MnO, the results are nearly identical across different DFT codes and impurity solvers, demonstrating the robustness of the method.

Limitations and Future Work

The authors acknowledge that their scheme is still in its early stages and that further improvements are needed for more reliable descriptions of charge-transfer-type Mott insulators and complex lanthanide and actinide systems. They suggest that the developed infrastructure can serve as a platform for combining DFT+DMFT with more advanced beyond-DFT methods, such as hybrid functionals or GW approximations, which are becoming available in NAO-based DFT codes. Future work could focus on addressing the quantitative discrepancies observed in some cases, particularly for f-electron systems, and exploring the impact of different projector definitions and impurity solver implementations.

Summary updated

Abstract

The combination of density functional theory with dynamical mean-field theory (DFT+DMFT) has become a powerful first-principles approach to tackle strongly correlated materials in condensed matter physics. The wide use of this approach relies on robust and easy-to-use implementations, and its implementation in various numerical frameworks will increase its applicability on the one hand and help crosscheck the validity of the obtained results on the other. In this work, we develop a formalism within the linear combination of numerical atomic orbital (NAO) basis set framework, which allows for merging of NAO-based DFT codes with DMFT quantum impurity solvers. The formalism is implemented by interfacing two NAO-based DFT codes with three DMFT impurity solvers, and its validity is testified by benchmark calculations for a wide range of strongly correlated materials, including 3d transition metal compounds, lanthanides, and actinides. Our work not only enables DFT+DMFT calculations using popular and rapidly developing NAO-based DFT code packages but also facilitates the combination of more advanced beyond-DFT methodologies available in these codes with the DMFT machinery.
BibTeX
@article{qu2022density,
  author = {Xin Qu and Peng Xu and Rusong Li and Gang Li and Lixin He and Xinguo Ren},
  title = {{Density Functional Theory Plus Dynamical Mean Field Theory within the Framework of Linear Combination of Numerical Atomic Orbitals: Formulation and Benchmarks}},
  journal = {Journal of Chemical Theory and Computation},
  year = {2022},
  doi = {10.1021/acs.jctc.2c00472},
}
Figure