LDA+DMFT: Modeling Correlated Electron Materials

• LDA+DMFT is a hybrid approach that combines local density approximations with dynamical mean-field theory to accurately capture both itinerant band-structure effects and strong local electron correlations.
• It employs nested self-consistency loops to update impurity self-energy and electron density, ensuring precise modeling of spectral properties and total energies.
• The framework addresses challenges like double-counting corrections and optimal projector choices, advancing first-principles simulations of Mott insulators, itinerant magnets, and mixed-valence compounds.

The LDA+DMFT (Local Density Approximation plus Dynamical Mean-Field Theory) approach is a leading first-principles framework for the quantitative study of correlated electron materials. By combining density-functional theory in the local density approximation (LDA) with dynamical mean-field theory (DMFT), LDA+DMFT enables ab initio modeling of complex materials where strong, local electronic interactions and itinerant band-structure effects are both essential. The formalism was developed to overcome the limitations of LDA and related density-functional techniques, which underestimate electronic correlations in $d$- and $f$-electron systems, and to move beyond purely model-based DMFT by supplying realistic, material-specific one-particle Hamiltonians.

1. Theoretical Principles of LDA+DMFT

LDA+DMFT starts from a Kohn–Sham (KS) band Hamiltonian $H_{\mathrm{KS}}$ derived from conventional LDA or generalized gradient approximation calculations. This Hamiltonian is represented in a localized basis (often Wannier or linear muffin-tin orbital functions), allowing for the identification of a subset of "correlated" orbitals (typically $3d$, $4f$, or $5f$ shells) on each atomic site $R$ (Grånäs et al., 2011, Pavarini, 2014). To $H_{\mathrm{KS}}$ is added a local interaction term of the Hubbard–Hund type: $$H_{\mathrm{int}} = \frac{1}{2} \sum_{R,\,\xi\xi',\,\sigma\sigma'} U_{\xi\xi'\sigma\sigma'} c^\dagger_{R\xi\sigma} c^\dagger_{R\xi'\sigma'} c_{R\xi'\sigma'} c_{R\xi\sigma}$$ together with a double-counting correction $H^{DC}_R$, subtracting the LDA-included fraction of the local interaction energy (Grånäs et al., 2011, Pavarini, 2014, Lee et al., 2014).

The effective many-body problem thus reads: $f$0 In DMFT, the static mean-field $f$1 is replaced by a frequency-dependent, orbital-local self-energy $f$2 restricted to the correlated subspace. The lattice Green's function is computed as: $f$3 The DMFT loop enforces self-consistency between the lattice and an auxiliary impurity problem, ensuring that the impurity Green's function matches the on-site projected lattice Green's function (Grånäs et al., 2011, Kunes et al., 2010, Biermann, 2014).

2. Computational Workflow and Charge Self-Consistency

The LDA+DMFT computational cycle comprises two nested self-consistency loops: (i) the inner DMFT loop finding a self-consistent impurity self-energy, and (ii) the outer loop updating the total electron density to achieve full charge self-consistency (CSC) (Grånäs et al., 2011, Zhao et al., 2011, Pavarini, 2014).

• Initialization: An initial charge density $f$4 is constructed and the KS band structure $f$5 is generated.
• Correlated Subspace Construction: Projectors $f$6 define the local correlated orbitals from the electronic structure basis.
• DMFT Loop: For each inequivalent correlated site, the local Green's function is projected and the Weiss field (noninteracting bath Green's function) $f$7 is obtained. The impurity problem is solved (typically by CT-HYB quantum Monte Carlo, SPTF, or exact diagonalization), yielding a new impurity self-energy $f$8.
• Self-Consistency Condition: Impose $f$9 and $H_{\mathrm{KS}}$0.
• Charge-Density Update: Reconstruct the density matrix in the Bloch basis,

$H_{\mathrm{KS}}$1

and recover the new electron density

$H_{\mathrm{KS}}$2

(Grånäs et al., 2011, Zhao et al., 2011).

• Total Energy: The total LDA+DMFT energy is evaluated as

$H_{\mathrm{KS}}$3

(Zhao et al., 2011).

• Convergence: The process continues until both the electron density and the impurity self-energy are converged with respect to a tolerance.

Charge self-consistency is crucial to ensure the feedback of the DMFT local self-energy to the entire electronic structure, affecting both total energies and spectral properties (Grånäs et al., 2011, Zhao et al., 2011).

3. Double-Counting and Projected Subspace Strategies

A central issue in LDA+DMFT is the identification of the optimal double-counting correction, which must subtract those static correlation effects already accounted for in LDA (Lee et al., 2014, Nekrasov et al., 2012). Frequently used forms include the fully-localized limit (FLL): $H_{\mathrm{KS}}$4 and "around mean field" (AMF), based on a uniform occupancy assumption.

The recent "exact double counting" procedure (Lee et al., 2014) recasts the correlation functionals such that the intersection (double-counting) term is given by the LDA correlation evaluated on the local density. For solids, screening is incorporated via a screened interaction $H_{\mathrm{KS}}$5, and the double counting is defined with the correlation energy per particle of a screened homogeneous electron gas.

Projector definitions for the correlated subspace can be constructed via maximally-localized Wannier functions, muffin-tin orbitals, or atomic-like pseudo-orbitals, with the projection formalism determining the degree of overlap between DFT and DMFT treatments (Zhao et al., 2011, 0801.4353). The flexible choice of projector and basis set allows adaptation to both all-electron and plane-wave pseudopotential codes.

4. Materials Applications and Physical Properties

LDA+DMFT has been applied to a wide range of correlated materials—Mott insulators, itinerant magnets, mixed-valence compounds, rare-earths, and actinide oxides—with both spectral and thermodynamic observables accurately computed (Grånäs et al., 2011, Skornyakov et al., 2010, Pavarini, 2014). Specific examples include:

• NiO (Charge-Transfer Insulator): Full CSC LDA+DMFT corrects the 3d occupancy, enlarges the Mott gap to $H_{\mathrm{KS}}$64 eV, and reproduces experimental spin and orbital moments per Ni (e.g., $H_{\mathrm{KS}}$7, $H_{\mathrm{KS}}$8) (Grånäs et al., 2011).
• bcc Fe (Itinerant Ferromagnet): Spectral functions and mass enhancement are quantitatively described, with spin ($H_{\mathrm{KS}}$9) and orbital ($3d$0) moments in good agreement with experiment (Grånäs et al., 2011).
• SmCo$3d$1 (Permanent Magnet): Simultaneous treatment of Sm $3d$2 and Co $3d$3 electrons (via different impurity solvers) reveals the necessity of CSC for correct multiplet positions and total magnetization ($3d$4/f.u.) (Grånäs et al., 2011).
• Fe-pnictides (e.g., LaFePO): LDA+DMFT accounts for moderate correlation-induced mass enhancements ($3d$5–2.2) and reproduces ARPES band renormalization (Skornyakov et al., 2010).

The DMFT description yields dynamic local self-energies that redistribute spectral weight between quasiparticle bands and Hubbard satellites, capture temperature-driven Mott metal-insulator transitions, and enable quantitative modeling of valence and core-level spectroscopies, including X-ray photoemission and RIXS (Hariki et al., 2017, Hariki et al., 2019, Ghiasi et al., 2018).

5. Extensions: Core-Level Spectroscopies and Dynamical Screening

LDA+DMFT has been extended to model spectroscopies beyond one-particle spectral functions. For core-level XPS and RIXS, an Anderson impurity model is constructed with explicit core-shells (e.g., $3d$6) and multiplet coupling to the valence shell, using a hybridization bath derived from DMFT (Hariki et al., 2017, Hariki et al., 2019, Winder et al., 2020). Such methods quantitatively reproduce satellite features, nonlocal screening effects, and fluorescence-like continua in experimental spectra, which are not accessible to traditional cluster models.

Dynamical screening effects—frequency dependence of the Hubbard $3d$7—are integrated by promoting $3d$8 and including the retarded interaction either via direct coupling to bosonic modes or within the "Bose factor ansatz" (Biermann, 2014). This enables the description of plasmon satellites and the reduction of the low-energy mass renormalization by a bandwidth renormalization factor $3d$9.

6. Methodological Challenges and Perspectives

Open methodological questions remain regarding the optimal formulation of double-counting corrections, the unambiguous construction of correlated projectors, and the systematic inclusion of nonlocal correlations (Nekrasov et al., 2012, Pavarini, 2014). Full charge self-consistency is critical for obtaining accurate forces and energetics, and has been demonstrated in all-electron LMTO, LAPW, and plane-wave pseudopotential codes (Grånäs et al., 2011, Zhao et al., 2011). Efficient quantum impurity solvers (CT-HYB, SPTF, ED) are essential for practical calculations, dictating the accessible temperatures and number of correlated orbitals.

Extensions to GW+DMFT and inclusion of cluster or diagrammatic corrections beyond single-site DMFT are being actively developed to capture nonlocal fluctuations and screening. The parameter-free framework defined by the "exact double counting" scheme (Lee et al., 2014) addresses one of the longest-standing ambiguities in the field.

LDA+DMFT thus constitutes a predictive and material-specific theory for strongly correlated electrons, capable of addressing both spectroscopy and thermodynamics in real materials, and serves as a foundation for future developments in ab initio many-body electronic structure theory.