Dynamical Hubbard approach to correlated materials: the case of transition-metal monoxides

arXiv:2503.10893 · cond-mat.str-el, cond-mat.mtrl-sci · Submitted 2025-03-13 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.

Kai: Today's paper: "Dynamical Hubbard approach to correlated materials".

Mira: Electronic correlations beyond static mean-field theories are fundamental to describing materials like transition-metal oxides, and this work introduces a novel dynamical Hubbard functional to capture these correlations in MnO,

Kai: First, who's behind it and why it matters.

Paper summary: Kai: So we've discussed how this paper tackles electronic correlations in MnO, FeO, CoO, and NiO using a new dynamical formulation. The core idea is that they are introducing a dynamical Hubbard functional designed to capture those complex correlation effects in transition-metal oxides. Mira, can you summarize the main thesis and what this work claims it achieves?

Mira: The paper's central claim is that they have successfully introduced a simple and transparent dynamical formulation, derived from generalizing DFT+U with a local frequency-dependent screened interaction. This approach allows them to capture all the key correlated-electron signatures—specifically gap opening, band renormalization, and spectral weight transfer—in excellent agreement with experimental PES/IPES data and state-of-the-art DFT+DMFT results (<ref:2503.10893#pg1>).

Lev: What matters here is that they’ve managed to do this without resorting to the massive computational costs associated with traditional DMFT impurity solutions, which makes it more accessible for applying these concepts to a wider range of materials.

Kai: Right, Lev, and I think that accessibility is huge because it means we can get results on systems like FeO that are notoriously difficult for simpler methods to predict accurately. It seems the authors have found a way to create a functional that bridges the gap between static approximations and full dynamical simulations.

Mira: They achieved this by solving the Dyson equation using the algorithmic-inversion method on sum over poles, which they call AIM-SOP, applied to their dynamical Hubbard functional (dynH) thirty <ref:2503.10893#pg1,the algorithmic-inversion method on sum over poles>. The functional derivative of dynH with respect to the local Green’s function yields a self-energy localized on the correlated manifold.

Lev: That specific technical mechanism, using AIM-SOP instead of imaginary Matsubara frequencies common in standard impurity solutions, seems like a significant improvement for accuracy because it avoids the analytic continuation problems those methods often face <ref:2503.10893#pg2>.

Kai: It’s about making the calculation more direct and less prone to numerical errors when trying to extract the spectral properties of these correlated systems. They are essentially providing a cleaner way to compute what we need from first principles calculations.

Mira: And they’ve extended this beyond single sites, generalizing it to treat magnetic systems with multiple TM metal sites by spinpolarizing the momentum-bands Green’s function and the RPA screened interaction <ref:2503.10893#pg2>. This generalization is what allows them to model more realistic material structures.

Lev: Modeling those multi-site magnetic interactions correctly is a big test, because getting the spin-dependent exchange and correlation self-energy right across different sites is where many approximations usually break down in these complex oxides.

Kai: So, the paper shows that this generalized dynamical Hubbard functional is a viable path toward obtaining reliable results for the spectral properties of these transition-metal monoxides, which is exactly what we need to understand their electronic structure better.

Mira: And when you look at the specific findings for FeO, DFT+dynH predicts a small gap of circa two eV, which aligns well with experimental IPES data <ref:2503.10893#pg1>, which contrasts with other theoretical predictions that suggest wider gaps for all four monoxides <ref:2503.10893#pg1>.

Lev: That comparison to experimental IPES data is critical; if the model can nail that specific gap size, it validates the functional's ability to capture the subtle physics in this material.

Kai: It’s a strong validation point for this approach; it suggests that we might finally have a theoretical tool capable of providing quantitative agreement on these subtle electronic features in these important transition-metal oxides.

Mira: And they also noted that for NiO, the framework predicts electronic bands present in experimental ARPES data that seem to disappear in the strong correlation regime of the DMFT impurity solutions <ref:2503.10893#pg1>, which is another strong piece of evidence.

Lev: That means this method isn't just reproducing known results; it’s accessing physics that other methods are missing, which gives us a clearer picture of where we need to focus our experimental efforts next.

Kai: It really sounds like the whole point is that this paper delivers a result that is quantitatively consistent across different theoretical benchmarks and experimental data, which builds confidence in the method itself.

Conclusion: Kai: So we’ve walked through how this paper presents the "Dynamical Hubbard approach to correlated materials: the case of transition-metal monoxides." We need to wrap up by discussing what this title and the authors mean, and what the broader implications are in plain terms. Mira, can you summarize why this work is significant in a broader context?

Mira: This work is significant because it successfully extends existing DFT+U methods by incorporating a local frequency-dependent screened interaction into a dynamical framework <ref:2503.10893#pg2>. The authors show that this extension provides access to dynamic correlation signatures, which are vital for understanding the physics of materials like transition-metal oxides in detail.

Lev: In simple terms, what does this mean for the broader field? Does it suggest a new direction for theoretical chemistry or condensed matter physics?

Kai: It suggests a new direction where we can move beyond static pictures and get quantitative agreement with both experiment and high-level theory on complex electronic structures. The authors have provided a framework that is more robust for studying materials like MnO, FeO, CoO, and NiO than previous methods.

Mira: They’ve shown that this dynamical Hubbard functional can reliably capture essential dynamic features like spectral weight transfer and band renormalization <ref:2503.10893#pg1>, which are crucial for describing how these materials behave when they are interacting with their environment.

Lev: For experimentalists, it means we have a more reliable theoretical map to guide our measurements of these materials because we know what dynamic features to look for when probing them.

Kai: Essentially, the impact is providing a better predictive tool that helps us design materials with tailored electronic properties by allowing us to see the subtle but important correlation effects that determine their functionality in things like catalysis or magnetism.

Mira: Yes, this paper shows that a more transparent dynamical formulation can yield results comparable to state-of-the-art DFT+DMFT simulations while being more tractable for material studies <ref:2503.10893#pg1>. It’s about making high-level correlation physics more accessible through a clearer theoretical structure.

Lev: If we translate this success into practical terms, it means we can potentially run calculations that are closer to the complexity of real experimental setups without losing the fundamental physical insights they offer.

Kai: It's about moving from just calculating energies to understanding the underlying dynamics, and this paper provides a pathway for achieving that understanding in these important correlated systems.

Theory and Simulation of Materials (THEOS) · National Centre for Computational Design and Discovery of Novel Materials (MARVEL) · Department of Applied Physics and Materials Science, California Institute of Technology · Colleege Champittet · European Theoretical Spectroscopy Facility (ETSF) · PSI Center for Scientific Computing, Theory and Data

cond-mat.str-el, cond-mat.mtrl-sci

Submitted: 2025-03-13

Updated: 2025-03-13

Comments: The first two authors are equally contributing

DOI: 10.1103/y7gb-q22g

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 90/100

The gist: Electronic correlations beyond static mean-field theories are fundamental to describing materials like transition-metal oxides, and this work introduces a novel dynamical Hubbard functional to

Key concepts

Dynamical Hubbard Functional (dynH)
This is a new mathematical tool derived from solving the Dyson equation using the algorithmic-inversion method on sum over poles. It allows researchers to calculate the self-energy of correlated electrons directly from a functional, providing a more transparent way to model local frequency-dependent interactions.
DFT+U vs. DFT+DMFT
Standard DFT methods like DFT+U are static mean-field approximations that struggle with dynamic features like spectral weight transfer. In contrast, DFT+DMFT models local dynamical correlations but is computationally expensive. This paper proposes a more efficient, transparent framework (dynH) that bridges the gap between these two approaches.
Spectral Weight Transfer
This refers to how the energy distribution of electrons changes when interactions are turned on. In correlated materials, spectral weight transfer is a key signature indicating that electron states are being reorganized due to strong electronic correlations, which static methods cannot accurately predict.

Terminology

Summary

Electronic correlations beyond static mean-field theories are fundamental to describing materials like transition-metal oxides, and this work introduces a novel dynamical Hubbard functional to capture these correlations in MnO, FeO, CoO, and NiO. The key finding is that this formulation successfully captures all correlated-electron signatures—including gap opening, band renormalization, and spectral weight transfer—in excellent agreement with experimental PES/IPES data and state-of-the-art DFT+DMFT results.

The gist

This letter shows that a simple and transparent dynamical formulation, obtained from the generalization of DFT+U to host a local frequency-dependent screened interaction, is able to capture all the correlated-electrons signatures of gap opening, bands renormalization, and spectral weight transfer in excellent agreement with experimental PES/IPES data and state-of-the-art DFT+DMFT result.

Motivation and Context

Binary transition-metal oxides like MnO, FeO, CoO, and NiO are ideal test cases because they exhibit complex electronic structures categorized as Mott or charge-transfer insulators. Standard DFT with semi-local functionals often fails to predict the insulating state of CoO and FeO by underestimating the electronic gap. While DFT+U and hybrid functionals recover the insulating states, they are inherently static mean-field solutions that hinder the predictive first-principles nature of the methods, as they struggle to describe k-resolved spectral weight transfer and bands renormalization. DFT+DMFT is a tool that models local dynamical electronic correlations but suffers from high computational costs. This paper addresses this limitation by proposing a more efficient, transparent dynamical framework.

Methodology: The Dynamical Hubbard Functional (dynH)

The core of the approach is the introduction of the dynamical Hubbard functional (dynH) [30], which is derived from solving the Dyson equation with the algorithmic-inversion method on sum over poles (AIM-SOP) for a dynamical Hubbard functional. The functional derivative of dynH with respect to the local Green’s function provides the dynH self-energy localized on the correlated manifold.

  1. The local spin Green’s function is defined as G = P†GP, where projectors P transform from momentum-band (k, n) indexes to the (m) index, representing magnetic quantum numbers of atomic-like orbitals centered on each transition-metal site I in the cell.

  2. The local spin Green’s function is solved via the Dyson equation: G−1 = G0−1 − Σ, where Σ is the self-energy derived from dynH.

  3. The calculation of spectral functions and density of states is achieved using AIM-SOP, which avoids the use of imaginary Matsubara frequencies common in impurity solutions used in DMFT, thereby hindering accuracy due to analytic continuation.

Generalization to Magnetic Systems

The functional is generalized from a single site to treat magnetic systems with multiple TM metal sites. This generalization involves:

  1. Spinpolarization of the momentum-bands Green’s function and of the RPA screened interaction.

  2. Convolution of these two elements to arrive at a local spin-dependent exchange and correlation self-energy of the form:

Σσ,I dynH(ω) = 2πi ∂ΦdynH[Gσ,I] / ∂Gσ,I = − Z dω′UI (ω′)Gσ,I(ω + ω′) + U∞,I2.

The local screened Coulomb interaction U(ω) is calculated from the RPA spin-DFT polarization function and is fitted to a sum over poles. The zero-frequency limit of the real part of this interaction is related to the entity of the scissor effect involved in gap opening.

Results and Comparison with State-of-the-Art

The framework was applied to MnO, FeO, CoO, and NiO in their ground states (AFM phase). The results show:

  1. For FeO, DFT+dynH predicts a small gap of circa 2 eV, which is in agreement with experimental IPES data [3] and contrasts with DFT+DMFT and QSQW predictions that suggest wide gaps for all four monoxides.

  2. The approach predicts electronic bands present in the experimental ARPES data for NiO that disappear in the strong correlation regime of the DMFT impurity solution [7, 8].

  3. For MnO, CoO, and NiO, the overall bandwidths are consistent with experimental data for all four monoxides.

  4. The framework provides access to particle lifetimes, satellites, spectral weight transfer and bands renormalization, which are fundamental signatures of electron correlations unavailable to static methods like DFT+U or QSGW.

Conclusion

In conclusion, the work successfully generalized the dynamical Hubbard functional to magnetic systems and applied it to the prototypical Mott-Hubbard/charge-transfer monoxide series.

Improvements for AI systems

As a fastidious and diligent researcher, I have analyzed this scientific paper, Dynamical Hubbard approach to correlated materials: the case of transition-metal monoxides, which introduces a novel dynamical Hubbard functional (dynH) for multi-site magnetic systems.

The primary improvement lies in enhancing the accuracy of electronic structure calculations for complex, strongly correlated materials that are notoriously difficult for standard Density Functional Theory (DFT) and static many-body perturbation theory (MBPT).

Here are the specific improvements to AI systems derived from this research:

  1. A new class of machine learning interatomic potentials/descriptors trained on the output of the DFT+dynH framework.

  2. The ability for AI systems to predict ground-state electronic properties (like band gaps, magnetic ordering, and charge localization) with near-experimental accuracy for transition metal oxides without requiring computationally prohibitive GW or DMFT calculations.

Specific capabilities of the improved AI system:

  1. The improved system can accurately predict the magnetic ground state (e.g., AFM vs. FM) and structural stability of transition-metal oxide materials using only DFT+dynH, surpassing the limitations of standard DFT+U methods which often fail to capture insulating states correctly.

  2. It can perform high-fidelity prediction of spectral properties—specifically the Density of States (DOS), spectral functions, band renormalization, and spectral weight transfer—for materials like MnO, FeO, CoO, and NiO. This allows for the identification of subtle correlation effects such as charge-transfer vs. Mott-Hubbard character in real-time during material design.

  3. The AI system can be trained to distinguish between different types of electronic correlations (e.g., the difference between a small gap predicted by DFT+dynH for FeO versus the wider gap predicted by DFT+DMFT or QSGW). This enables predictive modeling for materials exhibiting complex phase transitions driven by correlation.

  4. It can predict particle lifetimes and satellite features in excitation spectra, which are fundamentally unavailable to static methods, providing a richer description of excited states relevant for catalysis and electronic device applications.

In essence, the improvement moves AI from merely predicting geometry or simple band structures to predicting the complex, frequency-dependent electronic signatures that govern the functional behavior of real-world correlated materials.

Abstract

Electronic correlations beyond static mean-field theories are of fundamental importance in describing the properties of complex materials - such as transition-metal oxides - where the low-energy physics is driven by localized d or f electrons. Here, we show that it is possible to capture these correlations with a local and dynamical self energy, extending to the spin-polarized and multi-site case our recently introduced dynamical Hubbard functional formulation. We apply this formalism to the prototypical transition-metal monoxide series of MnO, FeO, CoO, and NiO in their ground state, finding excellent agreement with experiments for the spectral properties. The results are comparable or improved with respect to state-of-the-art theories, both for the densities of states and for the spectral functions - including band renormalization and spectral weight transfer - in a numerically efficient and physically transparent treatment of correlations amenable to the study of realistic, complex materials.

Sources

Related papers