---
title: 'DFT+U+V: Extended Hubbard-Corrected DFT'
url: https://www.emergentmind.com/topics/dft-u-v
type: topic
---

# DFT+U+V: Extended Hubbard-Corrected DFT

The DFT+U+V approach is a formally extended Hubbard-corrected density-functional theory (DFT) framework designed to simultaneously address local and nonlocal electronic correlations in materials where both electron localization and intersite hybridization are significant. Building directly on the original DFT+U formalism, DFT+U+V incorporates ab initio, self-consistent determination of both onsite ($U$), Hund's coupling ($J$), and intersite ($V$) screened Coulomb interaction parameters, generalizing and unifying previous pseudo-hybrid DFT+U approaches. Parameter-free by construction, and applicable both in all-electron and pseudopotential methodologies, DFT+U+V establishes a static mean-field limit of the extended Hubbard model, and serves as an efficient, systematically improvable alternative to conventional hybrid functionals or higher-level many-body techniques for a wide materials class spanning from $sp$ semiconductors to strongly correlated oxides and low-dimensional electronic systems [1911.10813], [2511.08002].

## 1. Formalism and Total Energy Functional

In the DFT+U+V scheme, the total energy functional augments the standard Kohn–Sham (KS) DFT energy $E_\mathrm{DFT}[\rho]$ with explicit onsite and intersite Hubbard correction terms evaluated over a localized orbital subspace:
\[
E_{\text{tot}} = E_{\text{DFT}}[\rho] + E_U + E_V
\]

The on-site correction $E_U$—in the rotationally invariant form—reads:
\[
E_U = \sum_I \frac{U_I^\text{eff}}{2} \sum_{m,\sigma} \left[ n_{mm,\sigma}^I - \sum_{m'} n_{mm',\sigma}^I n_{m'm,\sigma}^I \right]
\]
where $U_I^\text{eff} = U_I - J_I$, $I$ labels sites, $m$ orbital indices, and $n_{mm',\sigma}^I$ is the occupation matrix.

The intersite correction $E_V$ (extended Hubbard term) is given by:
\[
E_V = -\sum_{I \neq J} \frac{V_{IJ}}{2} \sum_{m, m', \sigma} n_{mm',\sigma}^{IJ} n_{m'm,\sigma}^{JI}
\]
where $V_{IJ}$ is the (screened) intersite Coulomb interaction and $n^{IJ}$ are generalized intersite occupation matrices [1911.10813], [2511.08002].

When both terms are combined:
\[
E_{U+V} = \sum_I \frac{U_I-J_I}{2} ...\ -\sum_{I \neq J} \frac{V_{IJ}}{2} ...
\]

This energy expression is valid for nonorthogonal or orthogonalized Hubbard manifolds and is applicable in both pseudopotential and all-electron contexts.

## 2. Self-Consistent Determination of U, J, and V

Rather than empirical tuning, the DFT+U+V methodology computes $U$, $J$, and $V$ ab initio and self-consistently at each electronic self-consistent-field (SCF) cycle.

### Coulomb Integral Derivation (ACBN0/eACBN0)

- **Onsite $U$**: Extracted using a Hartree–Fock-like average of screened four-center Coulomb integrals within the localized orbital subspace, weighted by occupation matrices. In the ACBN0 (pseudo-hybrid) scheme, this involves:
  \[
  U_I = \left[ \sum ... ( \bar{n}^I_{...} \bar{n}^I_{...} - ... ) \right] / [\text{normalization}]
  \]
- **Intersite $V$**: Similarly obtained by evaluating the intersite Coulomb integral contracted with occupation matrices corresponding to different atomic sites.
- **Hund's $J$**: Determined from exchange-like integrals, included where relevant.

Alternative approaches use constrained random-phase approximation (cRPA) or density-functional perturbation theory (DFPT) linear-response, where Hubbard $U$ and $V$ parameters are computed via differences in the interacting and noninteracting susceptibilities:
\[
U^I = (\chi_0^{-1} - \chi^{-1})_{II},\quad V^{IJ} = (\chi_0^{-1} - \chi^{-1})_{IJ}
\]
[2511.08002], [2306.06266], [2106.00520].

Parameters become dynamically "screened" by, and flow with, the evolving electron density, enabling ab initio, parameter-free functionals applicable to diverse correlated and hybridizing systems.

## 3. Physical Interpretation and Relation to Hybrid Functionals

DFT+U+V constitutes the static, mean-field limit of the extended Hubbard Hamiltonian:
\[
H_\text{eHub} = -\sum_{ij,\sigma} t_{ij} c_{i\sigma}^\dagger c_{j\sigma} + \sum_i U n_{i\uparrow} n_{i\downarrow} + \frac{1}{2}\sum_{i \neq j,\sigma,\sigma'} V_{ij} n_{i\sigma} n_{j\sigma'}
\]
This treats both local (U) and nonlocal (V) electron–electron interactions at the mean-field level.

The DFT+U+V correction delivers a "pseudo-hybrid" character in the sense that only a physically motivated, ab initio–determined subset of Coulomb integrals enter the exchange–correlation energy, without empirical mixing parameters as in conventional hybrid functionals. Calculations demonstrate that DFT+U+V mimics the effect of certain exact-exchange portions found in hybrid schemes, but applies them selectively—improving bandgaps, magnetic moments, and correlated electronic structure at a fraction of the computational cost.

## 4. Computational Workflow and Basis Set Considerations

DFT+U+V implementations require:

- **Definition of correlated subspace and projectors**: Ideally with Löwdin-orthogonalized atomic orbitals, muffin-tin functions, or maximally localized Wannier functions. The choice and orthogonality of projectors is critical for the accurate determination and application of U and V, particularly in systems with strong covalency [2106.00520], [2511.08002].
- **Self-consistent loop**:
  1. Perform a standard DFT calculation.
  2. Project onto the correlated subspace to construct occupation matrices.
  3. Compute U, V (either via ACBN0-like pseudo-hybrid integrals, cRPA, or DFPT response).
  4. Update the Kohn–Sham Hamiltonian with the derived corrective Hubbard potential.
  5. Iterate until convergence in both U, V and electronic structure.

This can be implemented in plane-wave, all-electron FLAPW, and localized basis codes, and remains computationally efficient compared to GW or fully hybrid functionals [2511.08002], [1911.10813].

## 5. Benchmarks, Applications, and Performance

DFT+U+V has been systematically benchmarked against experiment, hybrids, and $GW$ for a broad range of materials:

| System / Property                | DFT/PBE | DFT+U | DFT+U+V | Hybrids/GW | Experiment     |
|----------------------------------|---------|--------|---------|------------|---------------|
| Band gap in Si                   | ~0.6 eV | ~0.2 eV| ~1.0 eV | ~1.2 eV    | ~1.1 eV       |
| Band gap in NiO                  | 1.35 eV | 3.05 eV| 3.64 eV | 4.33 eV    | 4.0–4.3 eV    |
| Intercalation voltage, LiFePO₄   | 2.72 V  | 3.46 V | 3.47 V  | 3.64 V     | 3.43 V        |
| Band gap/Nodal-line ZrSiSe       | PBE: too small | +U+V ≈ hybrid | ≈ exact-exchange hybrid |        |

Key findings:
- DFT+U+V systematically improves lattice parameters, band gaps, crystal field splittings, and spectroscopic features (e.g., satellite splitting in correlated oxides, Fermi velocities in Dirac materials) [2511.08002], [1911.10813], [2203.15732].
- In low-dimensional or ligand-hybridized systems (e.g., NiS₂, pentahexoctite), V is indispensable to correctly place band edges, reproduce charge-transfer gaps, and capture correlated magnetic or topological behaviors [2308.02737], [2405.18683].
- For charge/orbital ordering, charge-density-wave (CDW) physics, and mixed-valence states (e.g., KV₃Sb₅, olivine LiₓMPO₄), only the inclusion of V enables correct stabilization of observed phases and reproduces oxidation-state jumps matching experiment [2504.17995], [2203.15732].

## 6. Practical Prescriptions, Methodological Aspects, and Limitations

- **Best practices**:
  - U and V should be computed ab initio, and self-consistently with structural relaxations and magnetic configurations, not chosen empirically [2306.06266].
  - The definition of the correlated manifold and projectors (NAO, OAO, Wannier) must be consistent between the calculation of parameters and the application of the correction [2106.00520].
  - Hund's coupling J can be included, and for systems with spin–orbit or noncollinear order, noncollinear generalizations are available [2604.23565].
- **Technical implementation**:
  - For pseudopotential frameworks, DFT+U+V correction appears as an additional potential constructed in the projector subspace and added within the self-consistent cycle.
  - In all-electron (FLAPW) implementations, explicit projections onto muffin-tin or Wannier (correlated) basis are performed, and V is usually restricted to the first or first few coordination shells [2511.08002].
- **Limitations**:
  - Static, frequency-independent U/V; for frequency-dependent screening and dynamical correlation see DFT+U($\omega$)+V extensions [2309.12144].
  - Double-counting corrections must be handled consistently. Sensitivity to subspace/projector choice can affect quantitative results and should be documented.
  - Higher corrections (three-center terms, beyond nearest neighbors, explicit frequency dependence) may be required for certain quantitative applications [2511.08002], [2309.12144].

## 7. Broader Implications and Outlook

DFT+U+V functionals bridge the divide between efficient mean-field DFT and highly correlated $GW$/DMFT techniques, allowing accurate characterization of electronic, structural, and spectroscopic properties in correlated and itinerant electron systems at modest computational expense. The ab initio, parameter-free character enables predictive calculations for new materials without recourse to empirical fitting. The same framework provides a route to generalize parameter-free hybrid functionals, and is extensible to time-dependent phenomena (TDDFT+U+V), phonons, and real-space analysis of bond strengths or defect responses [1911.10813], [2409.01795].

DFT+U+V thus constitutes a versatile and systematically improvable approach for electronic structure studies in strongly correlated, mixed-valence, or low-dimensional systems, and is compatible with both efficient periodic implementations and all-electron methods [2511.08002], [1911.10813].

Source: https://www.emergentmind.com/topics/dft-u-v