---
title: LDA+U Method for Correlated Electron Systems
url: https://www.emergentmind.com/topics/lda-u-method
type: topic
---

# LDA+U Method for Correlated Electron Systems

The LDA+U method is an extension of density functional theory (DFT) that introduces explicit on-site Coulomb corrections for localized electron states—most often transition-metal 3d and 4f/5f electrons—enabling first-principles calculations to capture Mott insulators, orbital polarization, and other physics beyond reach of conventional local (LDA) or semilocal (GGA) functionals. By augmenting the Kohn–Sham DFT total energy with a Hartree–Fock–like Hubbard U correction and subtracting a double-counting term, LDA+U corrects the self-interaction and delocalization errors that plague DFT in correlated systems. The method has been systematically developed and benchmarked across a wide range of compounds, providing a computationally tractable route to improved gaps, magnetic moments, orbital order, multiplet splitting, and excitation spectra in correlated materials.

## 1. Theoretical Principles and Formalism

LDA+U modifies the total-energy functional of DFT to explicitly penalize partial occupancy of localized atomic orbitals (typically $d$ or $f$ shells) via an on-site Hubbard correction. The general LDA+U functional is
\[
E_{\rm LDA+U}[\{\psi\},\{\phi_m^I\}] = E_{\rm LDA}[\rho] + E_{\rm Hub}[\{n_{mm'}^{I\sigma}\}] - E_{\rm dc}[\{n^I\}]
\]
where $n_{mm'}^{I\sigma}$ is the occupation matrix of correlated orbitals on site $I$. Multiple rigorous forms are in use:
- **Rotationally invariant (Liechtenstein et al.):**
  \[
  E_{\rm Hub} = \frac{1}{2} \sum_{I\sigma mm'k} \langle m, m''|V_{ee}| m', m'''\rangle\ n_{mm'}^{I\sigma} n_{m''m'''}^{I-\sigma} + (\mathrm{exchange\ term})
  \]
  The interaction tensor $V_{ee}$ is parameterized by Slater integrals $F^k$.
- **Simplified (Dudarev et al.):**
  \[
  E_U = \sum_{I,\sigma} \frac{U_{\rm eff}}{2}\, \mathrm{Tr}\left[ n^{I\sigma}(1-n^{I\sigma}) \right]
  \]
  where $U_{\rm eff} = U - J$ is the difference of averaged Hubbard and Hund couplings.

The **double-counting** term is necessary to subtract the contribution of on-site interactions already present in the LDA or GGA functional. The Fully Localized Limit (“FLL”) expression is
\[
E_{\rm dc}^{\rm FLL} = \sum_I \left[ \frac{U}{2} n^I(n^I-1) - \frac{J}{2} \sum_\sigma n^{I\sigma}(n^{I\sigma}-1) \right]
\]
with $n^I = \sum_{m,\sigma} n_{mm}^{I\sigma}$. Variants such as the Around Mean Field (AMF) scheme exist but are less commonly applied for Mott insulators [1309.3355, 2310.18994, 1002.1101, 1810.10393].

## 2. Implementation and Parameterization

LDA+U is implemented in various electronic structure codes, including VASP, Elk, OpenMX, and TB-LMTO-ASA, often using either atomic-like, pseudo-atomic, or Wannier-type projectors for defining the correlated subspace [1004.4741, 2310.18994, 1901.00647]. Parameters $U$ and $J$ can be:
- **Empirical**, chosen to reproduce experimental observables (gap, moment, XMLD, etc.) [2110.10274, 1201.6263, 1401.2277].
- **Ab-initio derived**, using methods such as constrained DFT, linear-response supercell techniques, or constrained Random Phase Approximation (cRPA) [1004.1321, 1312.4272, 1309.3355].

A representative computational workflow proceeds as:
1. Perform LDA/GGA calculation.
2. Project occupations onto correlated orbitals.
3. Add the +U correction and subtract double-counting in the Kohn–Sham equations.
4. Iterate self-consistently to convergence for charge and spin densities.
5. Optionally, adjust $U, J$ to match experiment or by self-consistency (cf. cRPA-LDA+U) [1004.1321].

For systems with strong spin–orbit coupling or multipolar order, explicit inclusion of SOC and multipolar tensor expansion of the local density matrix are essential [1111.4137, 1308.1488].

## 3. Physical Interpretations and Effects

LDA+U shifts partially occupied correlated bands away from the Fermi level, favoring integer occupancy, and thus stabilizes insulating states in Mott and charge-transfer systems. For sufficiently strong $U$:
- The correlated subspace splits into *lower* and *upper Hubbard bands*, opening a gap [1004.4741, 2110.10274].
- Magnetic moments and crystal-field splittings are more accurately captured, as shown for NiO, Gd, and 4f/5f oxides [1004.1321, 1111.4137].
- Orbital selectivity emerges in materials with multiple correlated orbitals, as interorbital hybridization and local crystal fields cause $U$ and $J$ to act differentially (see FeSe vs FeTe) [2310.18994].
- In hybrid-functional and DFT+U combinations, the $U$ parameter can be tuned to restore experimentally measured gaps and d-level binding energies, yielding improved energetics and ground-state structure [1312.4272].

## 4. Applications Across Correlated Materials

LDA+U has enabled:
- **Restoring correct insulating Fermiology and anisotropy**: In Sr$_2$VFeAsO$_3$, large $U$ applied on V $d$ drives the blocking layer Mott insulating, yielding conventional Fe-based superconductor Fermi surfaces and strong transport anisotropy [1004.4741].
- **Spectroscopic accuracy and limitations**: For Co$_2$FeSi, LDA+U opens a minority gap and recovers total moment, but the crystal-field splitting and minority-d bandwidth are overestimated, degrading agreement with x-ray magnetic linear dichroism. Fixed-spin-moment GGA provides better XMLD agreement [1201.6263].
- **Phonon spectra and electron–lattice coupling**: In La$_2$CuO$_4$, LDA+U with optimal $U$ (set by optical gap and local moment) reproduces Cu–O half- and full-breathing phonon frequencies in line with neutron data—standard LDA/GGA severely underestimates mode energies [2110.10274].
- **Multipolar exchange and spin–orbit-driven order**: Extensions allow explicit calculation of dipole– and quadrupole–quadrupole interactions, e.g. in UO$_2$ via a pair-flip mapping over the LDA+U+SOC energy landscape [1308.1488].
- **Self-interaction and crystal-field accuracy in 4f/5f dioxides**: Self-interaction–corrected LDA+U combined with on-site multibody fits recovers experimental crystal-field splittings within ~10–20 meV [1111.4137].
- **Exploration of metastable states**: Random density matrix control provides a systematic way to traverse the local minima in the multi-dimensional LDA+U energy landscape, ensuring comprehensive magnetic and orbital state identification [1810.10393].
- **Cluster Mott and $J_{\mathrm{eff}}$ physics**: For lacunar spinels, charge-only LDA+U formalism (vs. spin-polarized) stabilizes well-defined $J_{\mathrm{eff}}$=1/2 or 3/2 bands, with quantitative metrics for band purity [1901.00647].

## 5. Parameter Choice, Double Counting, and Best Practices

Key choices in LDA+U calculations include:
- **Projectors:** Choice determines the correlated subspace and directly affects occupation matrices, $U$, and physical predictions [1309.3355, 1004.1321].
- **U and J estimation:** Empirical procedures may target the gap or moment, while self-consistent cRPA or linear-response DFT provide first-principles $U$ values [1004.1321, 1312.4272, 1901.00647, 2310.18994].
- **Double counting:** FLL is standard for Mott and charge-transfer insulators, AMF for itinerant metals. Uncertainty in the form is a pervasive open issue [1309.3355].
- **Spin treatment:** For systems where Hund's $J$ and exchange dominate, charge-only DFT+U is recommended to avoid unphysical enhancement of Stoner exchange [1901.00647, 1810.10393].
- **SOC and multipolarity:** Manifest necessity in systems with strong spin–orbit coupling or multipolar order (especially 4f/5f compounds), requiring generalization to tensorial occupation matrices [1111.4137, 1308.1488].
- **Hybrid and combined approaches:** LDA+U can be combined with hybrid functionals (e.g. HSE+U) to further alleviate delocalization and self-interaction errors, often improving structure and gap predictions [1312.4272].

## 6. Strengths, Limitations, and Future Directions

LDA+U:
- **Strengths:** Delivers computational efficiency, systematic correction for DFT errors in correlated subspaces, orbital selectivity, and has a well-understood effect on Fermiology and gapping. LDA+U is essential for ground-state structural, magnetic, and spectroscopic calculations of Mott and charge-transfer insulators, multipolar ordered systems, and $J_{\mathrm{eff}}$ magnets [1309.3355].
- **Limitations:** Remains static (frequency-independent $U$), neglects dynamical correlation and fluctuations (important for quantitative ARPES, low-energy collective modes), and is sensitive to projector and double-counting ambiguity. LDA+U+SOC fails to describe broad multiplet spectra without explicit energy-dependent kernels [1002.1101]. Overcorrection and poor orbital selectivity can arise if $U$ is applied indiscriminately, as evidenced by XMLD failure in Co$_2$FeSi [1201.6263].
- **Development directions:** Self-consistent, frequency-dependent $U$ (dynamical mean field theory, cRPA-LDA+U), generalization to LDA+U+V (inter-site), tensorial forms with full orbital-resolved $U_{mm'm''m'''}$, and integration with TDDFT or GW for excited states [1004.1321, 1309.3355].

## 7. Summary Table: Representative LDA+U Applications

| Material/System         | LDA+U Role                                     | Key Outcome(s)                                          | Reference        |
|------------------------|------------------------------------------------|---------------------------------------------------------|------------------|
| Sr$_2$VFeAsO$_3$       | Layer selectivity, $U$ on V                     | Mott insulating perovskite, textbook Fe-pnictide FS     | [1004.4741]      |
| Co$_2$FeSi             | FLL scheme, $U$ on Co/Fe, XMLD comparison      | Over-broadened XMLD, FSM GGA preferable                 | [1201.6263]      |
| La$_2$CuO$_4$          | Dudarev $U$ on Cu                              | Gap and breathing phonons matched to experiment          | [2110.10274]     |
| FeSe, FeTe             | TB-LMTO-ASA, $U,J$ explicit, FLL DC            | Orbital/nematic correlations, orbital-selective effects | [2310.18994]     |
| UO$_2$                 | LDA+U+SOC, tensor expansion, pair-flip         | Antiferro-dipolar, ferro-quadrupolar exchange           | [1308.1488]      |
| II-VI chalcogenides    | HSE+U hybrid                                   | Optical gaps, formation energies improved                | [1312.4272]      |
| 4f/5f dioxides         | LDA+U+SIC, many-body mapping                   | Crystal field splittings within 10–20 meV of experiment | [1111.4137]      |
| GaM$_4$Se$_8$          | Charge-only LDA+U (FLL), $J_{\mathrm{eff}}$    | Molecular $J_{\mathrm{eff}}$ bands, controlled Hund's J | [1901.00647]     |
| FeAs, TM oxides        | Dudarev LDA/GGA+U, empirical $U$ range         | Limited moment correction, structure vs. magnetism trade-off | [1401.2277], [1810.10393] |

LDA+U stands as a foundational, physically motivated, and widely adopted augmentation to standard density functional approximations, providing robust access to the rich phenomenology of correlated electron systems within an efficient mean-field framework. Continued methodological innovation addresses its key limitations through self-consistent $U$, dynamical extensions, and integration with many-body and hybrid functionals.

Source: https://www.emergentmind.com/topics/lda-u-method