🔬 materials science

Implementation and application of a DFT $+U$ $+V$ approach within the all-electron FLAPW method

This paper presents a first-principles implementation of the DFT $+U$ $+V$ formalism within the all-electron FLAPW method using the FLEUR code, demonstrating improved accuracy for diverse materials ranging from covalent semiconductors to charge-transfer insulators by incorporating intersite Coulomb interactions derived via the constrained random-phase approximation.

Original authors: Wejdan Beida, Gustav Bihlmayer, Christoph Friedrich, Gregor Michalicek, Daniel Wortmann, Stefan Blügel

Published 2026-03-02

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Wejdan Beida, Gustav Bihlmayer, Christoph Friedrich, Gregor Michalicek, Daniel Wortmann, Stefan Blügel

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to predict how a city will behave. You have a map (the standard computer model called DFT) that tells you where the buildings are and how people generally move. This map is great for most cities. But, if you try to use it for a city where people are extremely clingy, constantly grabbing onto each other's hands, or where neighbors are constantly arguing over shared resources, your map starts to fail. It thinks everyone is too spread out and too independent.

This is the problem with strongly correlated materials (like certain magnets, superconductors, or semiconductors). The electrons in these materials don't act like independent citizens; they act like a tightly knit, chaotic crowd.

This paper presents a new, upgraded map called DFT+U+V. Here is how it works, broken down into simple concepts:

1. The Old Map: DFT (The "Average" View)

Standard physics models (DFT) treat electrons like a smooth, flowing river. They assume electrons are spread out evenly.

The Problem: In materials like Graphene (a super-thin carbon sheet) or Nickel Oxide (a magnetic rock), electrons are actually "clumping" together or fighting over space. The old map says, "Everyone is fine," but the reality is, "There's a traffic jam!" This leads to wrong predictions about how the material conducts electricity or how strong its magnetic pull is.

2. The First Upgrade: DFT+U (The "Selfish Neighbor" Fix)

Scientists realized that sometimes an electron gets stuck on one specific atom and refuses to leave. It's like a neighbor who is so possessive of their own garden that they won't let anyone else near it.

The Fix (U): The "U" term adds a penalty for this possessiveness. It forces the model to acknowledge that electrons on the same atom repel each other strongly. This helped fix some problems, but it was still incomplete. It only looked at what was happening inside one house, ignoring the fence between neighbors.

3. The New Upgrade: DFT+U+V (The "Neighborhood Watch")

The authors of this paper added a new layer: V.

The Concept: Electrons don't just care about their own atom; they care about their neighbors. If an electron on Atom A moves, it affects the electron on Atom B right next to it. This is like a neighborhood where if one house throws a party, the whole block gets noisy.
The "V" Term: This accounts for the interaction between sites (neighbors). It captures how electrons "talk" to each other across the crystal structure. This is crucial for materials where the chemistry involves sharing electrons between different types of atoms (like Oxygen and Nickel).

4. The Tool: The "All-Electron" Telescope

To build this new map, the researchers used a very precise tool called FLAPW (Full-Potential Linearized Augmented Plane-Wave).

Analogy: Imagine trying to measure the shape of a cloud. Some tools (like pseudopotentials) just guess the cloud's shape based on a few points. The FLAPW method is like a high-resolution 3D scanner that measures every single drop of water in the cloud. It's computationally heavy but incredibly accurate. The authors successfully installed their new "Neighborhood Watch" (DFT+U+V) onto this high-precision scanner.

5. How They Got the Numbers: The "Census Taker" (cRPA)

To make the map work, they needed to know exactly how strong the "U" (selfishness) and "V" (neighborly) forces were. They didn't just guess; they calculated them from scratch using a method called cRPA (constrained Random Phase Approximation).

The Analogy: Think of this as a census taker going into the city and asking, "If I move this electron here, how much does it bother the electron there?" They did this using two different ways of looking at the data:
1. Muffin-Tin (MTF): Looking at the atom as a hard sphere (like a billiard ball).
2. Wannier Functions: Looking at the electron clouds as fuzzy, overlapping shapes.
  They compared both methods to ensure their results were solid.

6. The Results: Fixing the City

They tested their new map on three very different "cities":

Graphene (The 2D City):
- The Issue: Standard models predicted electrons moved too slowly.
- The Fix: Adding the "V" term (neighbor interactions) corrected the speed of the electrons, matching real-world experiments perfectly. It's like realizing the traffic flow was faster because the cars were actually helping each other move, not just driving alone.
Silicon & Germanium (The Covalent City):
- The Issue: These materials are held together by shared electron bonds. Standard models got the size of the crystal and the energy gap (how hard it is to make electricity flow) wrong.
- The Fix: The "V" term helped tighten the bonds, making the crystal size and energy gaps match reality much better.
Nickel Oxide (The Magnetic Rock):
- The Issue: This is a classic "Mott insulator." Standard models thought it was a conductor (like a metal) when it's actually an insulator (like a rock). They also got the magnetic spin wrong.
- The Fix: By including the interaction between Nickel and Oxygen atoms (the "V" term), the model finally realized the material is an insulator and correctly predicted its magnetic strength. It fixed the "charge transfer" problem—understanding how electrons jump between the metal and the oxygen.

The Big Picture

This paper is a success story of precision engineering in physics. The authors took a powerful, high-precision microscope (FLAPW) and added a new lens (DFT+U+V) that allows us to see the subtle, complex arguments and friendships between electrons.

Why does this matter?
If we want to design better batteries, faster computer chips, or new magnetic storage devices, we need to understand how electrons behave in these tricky materials. This new method gives scientists a much more reliable "GPS" to navigate the complex world of quantum materials, helping us build the technology of the future.

1. Problem Statement

Standard Density Functional Theory (DFT) often fails to accurately describe materials with strong electronic correlations, particularly those involving localized $d$ - and $f$ -electrons. While the DFT+U method addresses onsite Coulomb repulsion ( $U$ ) to correct self-interaction errors and localize electrons, it remains a single-site correction. It neglects intersite Coulomb interactions ( $V$ ), which are crucial for describing:

Charge redistribution and orbital hybridization between neighboring sites.
Charge disproportionation and charge ordering.
Charge transfer processes (e.g., between metal and ligand orbitals).
Non-local correlation effects in covalently bonded systems (e.g., graphene) and charge-transfer insulators (e.g., NiO).

Existing implementations of DFT+U+V often rely on pseudopotentials, which may lack the precision required for all-electron descriptions or struggle with complex basis sets. There is a need for a rigorous, all-electron implementation that can accurately handle both onsite and intersite correlations across diverse material classes.

2. Methodology

The authors present a first-principles implementation of the DFT+U+V formalism within the Full-Potential Linearized Augmented Plane-Wave (FLAPW) method, utilizing the FLEUR code.

Formalism Extension:
- The standard DFT+U energy functional is extended to include an intersite term ( $E_V$ ).
- The correction involves a generalized occupation matrix ( $n_{IJ}$ ) that projects Kohn-Sham wavefunctions onto atomic basis functions of two distinct atoms ( $I$ and $J$ ).
- The implementation uses Muffin-Tin Functions (MTFs) as projectors, consistent with the FLAPW basis set. The formalism accounts for mixed terms (A-B and B-A contributions) arising from the linearization of the radial functions ( $u$ and $\dot{u}$ ), which are essential for accurate intersite coupling.
- The double-counting correction is applied in the mean-field limit.
Parameter Calculation (cRPA):
- Interaction parameters ( $U$ and $V$ ) are derived from first principles using the Constrained Random-Phase Approximation (cRPA) via the SPEX code.
- The screened Coulomb interaction is projected onto two different localized basis representations to compare their efficacy:
  1. Maximally Localized Wannier Functions (MLWFs): Provides a bridge to low-energy model Hamiltonians and captures delocalization/hybridization effects.
  2. Muffin-Tin Functions (MTFs): Ensures strict consistency with the FLAPW basis set but is dependent on the muffin-tin radius.
- The study compares parameters derived from both bases (denoted as w-cRPA and u-cRPA).
Test Systems:
The methodology is validated on a diverse set of materials representing different bonding and correlation regimes:
1. Graphene: A 2D covalently bonded system with delocalized $\pi$ -electrons.
2. Silicon (Si) and Germanium (Ge): Bulk elemental semiconductors with strong $sp^3$ covalent bonding.
3. Nickel Oxide (NiO): A prototypical Mott-charge-transfer insulator with strong $d$ - $p$ hybridization.

3. Key Contributions

All-Electron Implementation: First successful integration of DFT+U+V into the all-electron FLAPW framework (FLEUR), enabling high-precision calculations without pseudopotential approximations.
Basis Set Comparison: A systematic comparison of MTF vs. Wannier function projectors for cRPA parameter extraction. The study demonstrates that while MTF parameters are generally larger due to tighter localization, they converge toward Wannier-based values as the muffin-tin radius increases.
Unified Framework: Demonstrates that a single framework can effectively describe systems ranging from weakly correlated 2D materials (graphene) to strongly correlated charge-transfer insulators (NiO).
Benchmarking: Extensive validation against experimental data, previous DFT+U+V pseudopotential studies, and high-level GW (Green's function) calculations.

4. Key Results

A. Graphene (2D Covalent System)

Fermi Velocity ( $v_F$ ): Standard DFT significantly underestimates $v_F$ . DFT+U improves it slightly, but DFT+U+V yields values (10.74–11.11 $\times 10^5$ m/s) in excellent agreement with experiment (11.00 $\times 10^5$ m/s).
Band Structure: The inclusion of $V$ corrects the slope of the Dirac cone. Both u-cRPA and w-cRPA parameters produce similar results despite significant differences in the raw $U$ and $V$ values, suggesting robustness.
Insight: Non-local interactions are critical for capturing the correct quasiparticle dispersion in 2D covalent systems.

B. Bulk Si and Ge (Covalent Semiconductors)

Lattice Constants & Moduli: Standard DFT slightly overestimates lattice constants. Adding $U$ alone causes lattice expansion and softening (worsening agreement). Adding $V$ counteracts this, restoring lattice constants and bulk moduli closer to experimental values.
Band Gaps: DFT severely underestimates the gap. DFT+U opens the gap but often at the cost of structural accuracy. DFT+U+V successfully opens the gap while simultaneously improving structural parameters.
Parameter Sensitivity: The study notes that for Si and Ge, fitted $V$ values comparable to $U$ are required to match GW results, suggesting that nearest-neighbor interactions alone may be insufficient and next-nearest-neighbor terms might be necessary for $sp^3$ systems.

C. NiO (Charge-Transfer Insulator)

Electronic Structure: Standard DFT fails to reproduce the correct orbital character at the valence band top (predicting Ni $d$ -dominance instead of O $p$ -dominance). DFT+U improves the gap but struggles with the correct spectral weight distribution.
DFT+U+V Success: The inclusion of intersite $V$ (between Ni and O) correctly reproduces the O $p$ -character at the valence band maximum and the Ni $d$ -character at the conduction band minimum. It also captures the satellite peak at -7 eV, consistent with experimental XPS data.
Magnetic Properties: The method accurately predicts the magnetic moment and stabilizes the AFM-2 ground state. Increasing $V$ strengthens superexchange interactions, further stabilizing the antiferromagnetic phase while slightly reducing magnetic moments due to enhanced hybridization.
Comparison: DFT+U+V outperforms DFT+U and matches or exceeds the accuracy of various hybrid functionals (PBE0, HSE06) for structural and electronic properties.

5. Significance and Outlook

Accuracy vs. Cost: The DFT+U+V approach offers a superior compromise between the high accuracy of GW calculations and the computational efficiency of DFT, making it scalable for complex materials.
Physical Insight: The work confirms that intersite Coulomb interactions are not merely a correction for Mott insulators but are essential for describing charge transfer, hybridization, and non-local correlations in covalent and mixed-bonding systems.
Future Directions: The authors plan to implement self-consistent evaluation of Hubbard parameters and explore frequency-dependent (dynamical) $U$ and $V$ to improve descriptions over large energy ranges.

In conclusion, this paper establishes a robust, all-electron framework for DFT+U+V that significantly enhances the predictive power of electronic structure calculations for a wide variety of correlated materials, bridging the gap between standard DFT and high-level many-body perturbation theory.