Fact-checked by Grok 2 weeks ago

Dynamical mean-field theory

Dynamical mean-field theory (DMFT) is a non-perturbative approximation method for investigating strongly correlated quantum many-body systems, particularly lattice models of interacting electrons, by mapping the full lattice problem onto a self-consistent single-site quantum impurity model embedded in an effective bath that captures local dynamical correlations while neglecting spatial fluctuations beyond the local scale.^[1] This approach becomes exact in the limit of infinite spatial dimensions or coordination number, where the self-energy is purely local and momentum-independent, allowing for a systematic treatment of strong electron-electron interactions across all energy scales, densities, and temperatures.^[1] Developed in the late 1980s and early 1990s, DMFT builds on earlier mean-field ideas for correlated systems, such as the Hubbard approximation, but incorporates full quantum dynamics by solving the impurity problem using techniques like quantum Monte Carlo, exact diagonalization, or numerical renormalization group.^[1] The foundational insight, formalized by Metzner and Vollhardt in 1989, leverages the scaling properties of interactions in high dimensions to justify the local approximation.^[1] A seminal review by Georges, Kotliar, Krauth, and Rozenberg in 1996 established DMFT as a cornerstone for studying phenomena like the Mott metal-insulator transition in the Hubbard model.^[1] At its core, DMFT operates through a self-consistency loop: the lattice Green's function is approximated using the local impurity Green's function, which determines the effective bath hybridization function, and this process iterates until convergence.^[2] This framework excels in capturing local quantum fluctuations and Mott physics but requires extensions, such as dynamical cluster approximation or dual fermions, to include short-range correlations in finite dimensions.^[2] Impurity solvers must handle the Anderson impurity model with frequency-dependent baths, enabling computations of spectral functions, susceptibilities, and transport properties.^[1] DMFT has profoundly impacted the study of strongly correlated materials, including transition-metal oxides, heavy-fermion compounds, and high-temperature superconductors, often combined with density functional theory (DFT+DMFT) for realistic ab initio simulations of electronic structure and phase diagrams.^[2] Key applications include elucidating the physics of the half-filled Hubbard model, where DMFT predicts a coexistence region between metallic and insulating phases near the Mott transition, and exploring unconventional superconductivity driven by correlations.^[1] Recent extensions address non-equilibrium dynamics, topological phases, and heterostructures, including implementations on quantum computers as of 2025, maintaining DMFT's relevance for advancing materials design and quantum simulation.^[2]^[3]

Foundations and Relation to Mean-Field Theory

Relation to classical mean-field theory

Classical mean-field theory (MFT) approximates interacting many-body systems by decoupling interactions through the replacement of operators with their average values, effectively reducing the problem to non-interacting quasiparticles in a static mean field.^[4] In the context of fermionic lattice models, such as the Hubbard model—a prototypical system for strongly correlated electrons—this corresponds to the Hartree-Fock approximation, where the self-energy is treated as a constant shift, \Sigma_\sigma = (U/2)(n_{-\sigma}), with U the on-site repulsion and n_{-\sigma} the average density of opposite spin electrons.^[4] However, static MFT fails to capture dynamical correlations essential in strongly correlated systems, as it neglects temporal fluctuations and assumes a frequency-independent self-energy, leading to inaccuracies near quantum phase transitions like the Mott transition and overestimation of susceptibilities, such as in the Brinkman-Rice approximation.^[4] For instance, it misses high-energy incoherent features, like Hubbard bands, and low-energy phenomena, such as the Kondo resonance, which arise from frequency-dependent interactions.^[4] Dynamical mean-field theory (DMFT) extends classical MFT by incorporating local dynamical fluctuations exactly while approximating non-local ones, generalizing the static Weiss mean field to a frequency-dependent form that treats quantum and temporal dynamics.^[4] This is achieved through a self-consistent mapping to an impurity problem, where the self-energy \Sigma(i\omega_n) becomes fully frequency-dependent, as opposed to the constant \Sigma(\omega) \approx constant in static MFT; for example, in the Fermi-liquid regime, \operatorname{Re} \Sigma(\nu) = U/2 + (1 - 1/Z)\nu with Z the quasiparticle weight.^[4] DMFT becomes exact in the infinite-dimensional limit, where local correlations dominate, enabling accurate descriptions of both metallic and insulating phases in correlated materials.^[4]

Historical development and motivations

The development of dynamical mean-field theory (DMFT) emerged in the late 1980s as an extension of mean-field approaches to address the limitations of static approximations in describing strongly correlated electron systems. Building on earlier work in classical mean-field theory, which treats interactions in an average field but neglects dynamical fluctuations, researchers recognized the need for a framework that incorporates frequency-dependent effects to model phenomena like the Mott metal-insulator transition in materials such as transition metal oxides. A foundational step was the analysis of the Hubbard model in the limit of infinite spatial dimensions by Metzner and Vollhardt in 1989, which demonstrated that diagrammatic expansions simplify dramatically in this regime, allowing for exact treatments of local correlations while nonlocal contributions become negligible.^[5] This infinite-dimensional limit provided the basis for DMFT's formalization in 1992 by Georges and Kotliar, who established an exact mapping of the infinite-dimensional Hubbard model onto a single-site quantum impurity problem with a self-consistent bath, enabling the inclusion of dynamical interactions beyond static mean-field methods like the Gutzwiller approximation.^[6] The primary motivations were to resolve longstanding challenges in strongly correlated systems, including the Mott transition—where DMFT qualitatively captures the coexistence of metallic and insulating phases, unlike the Gutzwiller approach which fails to describe the full dynamical crossover—and phenomena such as heavy-fermion behavior and unconventional superconductivity in materials like cuprates and f-electron compounds.^[1] By treating the self-energy as local and frequency-dependent, DMFT addressed the inadequacy of static approximations in capturing spectral weight transfer and quasiparticle renormalization essential for these systems.^[1] DMFT gained widespread acceptance following the seminal 1996 review by Georges, Kotliar, Krauth, and Rozenberg, which solidified it as a standard tool for correlated electron studies and highlighted its success in resolving the Mott transition with qualitative accuracy.^[1] Subsequent extensions, particularly the integration of DMFT with density functional theory (DFT) starting in the early 2000s, enabled ab initio calculations on realistic materials, such as transition metal oxides, by combining DFT's band structure with DMFT's treatment of strong correlations.^[7] As of 2025, this DFT+DMFT framework has become routine for predicting properties of real correlated materials, including spectral functions and thermodynamic responses in compounds exhibiting valence instabilities.^[8]

Core Formalism for the Hubbard Model

Single-orbital Hubbard model

The single-orbital Hubbard model serves as the foundational lattice model for studying strongly correlated electron systems, capturing the essential physics of itinerant electrons subject to local Coulomb repulsion. Introduced by Hubbard in 1963, it describes electrons on a lattice where each site hosts a single orbital that can accommodate up to two electrons of opposite spins.^[9] The model's Hamiltonian, expressed in the language of second quantization, is given by

H = -t \sum_{\langle i,j \rangle, \sigma} \left( c_{i\sigma}^\dagger c_{j\sigma} + \mathrm{h.c.} \right) + U \sum_i n_{i\uparrow} n_{i\downarrow} - \mu \sum_{i,\sigma} n_{i\sigma},

where c_{i\sigma}^\dagger (c_{i\sigma}) creates (annihilates) a fermionic electron of spin \sigma at lattice site i, n_{i\sigma} = c_{i\sigma}^\dagger c_{i\sigma} is the corresponding number operator, t > 0 is the nearest-neighbor hopping amplitude, U > 0 is the on-site Coulomb repulsion strength, and \mu is the chemical potential that controls the average electron density.^[9]^[4] The hopping term is non-local, facilitating electron delocalization across the lattice, while the interaction term is strictly local to each site, enforcing a penalty for double occupancy by opposite-spin electrons.^[4] Physically, the model embodies the competition between kinetic energy, which favors delocalization and arises from the hopping with an associated bandwidth scaling as z t (where z is the lattice coordination number), and potential energy from the local repulsion U, which promotes localization.^[4] This interplay is central to understanding correlated phenomena in materials like transition metal oxides.^[4] For analytical tractability in high-dimensional limits, the Bethe lattice—characterized by a tree-like structure without loops—is often employed as the canonical geometry, particularly as z \to \infty, yielding a semicircular non-interacting density of states with bandwidth $4t.^[4]

Mapping to the Anderson impurity model

The foundational principle of dynamical mean-field theory (DMFT) relies on a theorem established in the limit of infinite spatial dimensions (d → ∞) or infinite coordination number (z → ∞), where the electronic self-energy becomes local, independent of momentum: Σ(k, ω) ≈ Σ(ω). This locality arises because inter-site correlations scale in a way that suppresses non-local contributions, allowing the lattice problem to be mapped onto an effective single-site problem without loss of exactness in this limit.^[1] Consequently, the lattice Green's function simplifies to G(k, ω) = [ω + μ - ε_k - Σ(ω)]^{-1}, where ε_k is the dispersion, μ the chemical potential, and ω the frequency, enabling the local dynamics at each site to be captured by an impurity-like model.^[1] The auxiliary problem in DMFT is the single-impurity Anderson model (SIAM), which describes a correlated impurity orbital hybridized with a non-interacting bath of electrons. The Hamiltonian for this model is given by \begin{align*} H_\text{imp} &= \sum_{k,\sigma} \epsilon_k c_{k\sigma}^\dagger c_{k\sigma} + U n_{\uparrow} n_{\downarrow} + \sum_{k,\sigma} V_k (d_\sigma^\dagger c_{k\sigma} + \text{h.c.}), \end{align*} where d_σ (c_{kσ}) annihilates an electron on the impurity (bath) site with spin σ, U is the on-site Coulomb repulsion, and V_k the hybridization amplitudes.^[1] The bath is characterized by the hybridization function Δ(ω) = \sum_k |V_k|^2 / (ω - \epsilon_k + i0^+), which encodes the effective environment seen by the impurity and must be tuned self-consistently to reproduce the lattice physics.^[1] Conceptually, the mapping derives from viewing the local dynamics of a lattice site as those of an impurity embedded in a "cavity" field representing the rest of the lattice, excluding the site itself. By adjusting the bath hybridization Δ(ω) to match this cavity field—effectively the lattice Green's function with the site's contribution removed—the SIAM exactly reproduces the local lattice propagator G_loc(ω) in the infinite-dimensional limit.^[1] To compute this, the SIAM is solved numerically using impurity solvers such as quantum Monte Carlo (QMC) methods, which handle finite-temperature dynamics via stochastic sampling, or the numerical renormalization group (NRG), which provides high-precision spectral functions at low energies.^[1] The resulting impurity Green's function G_imp(ω) is then set equal to the local lattice Green's function G_lat(ω), closing the conceptual bridge between the full lattice and the single-site problem.^[1]

Self-consistency equations

The self-consistency equations in dynamical mean-field theory (DMFT) establish the crucial link between the lattice Green's function and the effective single-site impurity model, ensuring that the local properties of the lattice are reproduced by the impurity solution. The local Green's function G(\omega) on the lattice is obtained by integrating over the Brillouin zone, given by

G(\omega) = \int \frac{d\mathbf{k}}{(2\pi)^d} \frac{1}{\omega + \mu - \epsilon_{\mathbf{k}} - \Sigma(\omega)},

where \omega is the frequency, \mu is the chemical potential, \epsilon_{\mathbf{k}} is the dispersion relation, and \Sigma(\omega) is the local self-energy assumed to be momentum-independent in the DMFT approximation. This integral can be recast in terms of the non-interacting density of states \rho(\epsilon):

G(\omega) = \int d\epsilon \, \rho(\epsilon) \frac{1}{\omega + \mu - \epsilon - \Sigma(\omega)}.

For the Bethe lattice in the infinite-dimensional limit, which features a semicircular density of states, the expression simplifies analytically, with \rho(\epsilon) = \frac{2}{\pi D^2} \sqrt{D^2 - \epsilon^2} for |\epsilon| < D, where D = 2t is the half-bandwidth and t is the hopping parameter (corresponding to a full bandwidth of $4t). In this case, the self-consistency condition relates the lattice Green's function directly to the hybridization function of the impurity bath. The hybridization function \Delta(\omega), which describes the coupling of the impurity to its effective bath, is determined self-consistently to match the lattice local Green's function to that of the impurity model G_{\text{imp}}(\omega). It is given by

\Delta(\omega) = \omega + \mu - \Sigma(\omega) - \frac{1}{G(\omega)},

ensuring G_{\text{imp}}(\omega) = G(\omega). For the Bethe lattice, this takes the explicit form \Delta(\omega) = t^2 G(\omega), where the quadratic relation arises from the tree-like structure and infinite connectivity of the lattice. The self-energy is then extracted from the impurity solution via Dyson's equation:

\Sigma(\omega) = \omega + \mu - \Delta(\omega) - G_{\text{imp}}^{-1}(\omega).

This equation closes the DMFT mapping by feeding the impurity-derived self-energy back into the lattice problem. The effective bath for the impurity is parameterized by the non-interacting impurity Green's function, or Weiss field, G_0(\omega), whose inverse is

G_0^{-1}(\omega) = \omega + \mu - \Delta(\omega).

This form represents the bath as seen by the impurity site, excluding the local self-energy effects, and is used as input to solvers for the single-impurity Anderson model. In the Bethe lattice case, substituting the hybridization yields G_0^{-1}(\omega) = \omega + \mu - t^2 G(\omega), providing an analytically tractable condition that simplifies numerical implementations while capturing the essential physics of infinite-dimensional systems.

The DMFT iteration loop

The dynamical mean-field theory (DMFT) iteration loop implements the self-consistency equations through a numerical procedure that alternates between solving the effective impurity model and updating lattice quantities until convergence is achieved.^[4] This loop is essential for practical computations, as the DMFT mapping does not yield closed-form solutions but requires iterative refinement of the self-energy \Sigma(i\omega_n) and Green's functions on the Matsubara frequency axis.^[4] The procedure begins with an initial guess for the self-energy, typically \Sigma(i\omega_n) = 0 (non-interacting limit) or a Hartree-Fock approximation incorporating the static mean-field interaction.^[4] From this, the lattice Green's function G_{\text{loc}}(i\omega_n) is computed via the Dyson equation integrated over the Brillouin zone:

G_{\text{loc}}(i\omega_n) = \int \frac{d^dk}{(2\pi)^d} \frac{1}{i\omega_n + \mu - \epsilon_{\mathbf{k}} - \Sigma(i\omega_n)},

where \mu is the chemical potential, \epsilon_{\mathbf{k}} is the dispersion, and the integral yields the local component for the chosen lattice (e.g., Bethe lattice approximation).^[4] Next, the hybridization function \Delta(i\omega_n) for the single-impurity Anderson model (SIAM) is updated using the self-consistency condition:

\Delta(i\omega_n) = i\omega_n + \mu - \Sigma(i\omega_n) - G_{\text{loc}}^{-1}(i\omega_n).

This defines the non-interacting impurity Green's function \mathcal{G}_0^{-1}(i\omega_n) = i\omega_n + \mu - \Delta(i\omega_n).^[4] The SIAM is then solved numerically—using methods such as quantum Monte Carlo, exact diagonalization, or numerical renormalization group—to obtain the impurity Green's function G_{\text{imp}}(i\omega_n) and a new self-energy \Sigma(i\omega_n) = \mathcal{G}_0^{-1}(i\omega_n) - G_{\text{imp}}^{-1}(i\omega_n).^[4] These steps are repeated, with the updated \Sigma(i\omega_n) fed back into the lattice computation, until the difference |\Sigma^{\text{new}}(i\omega_n) - \Sigma^{\text{old}}(i\omega_n)| < \epsilon (typically \epsilon \sim 10^{-4} to $10^{-5}) across a sufficient range of Matsubara frequencies.^[4] For interaction strengths U below the bandwidth, convergence is usually reached in 10-50 iterations, though this can increase near critical points like the Mott transition.^[10]^[3] To accelerate convergence and stabilize the loop, especially in cases of oscillatory behavior, techniques such as linear mixing or Anderson (Pulay) mixing are applied to the self-energy updates, combining current and historical iterates to extrapolate toward the fixed point.^[11] These methods reduce the number of required iterations by promoting faster damping of errors across frequency components.^[11] A major computational challenge in the loop arises from the need for real-frequency properties, as impurity solvers often produce data on the imaginary Matsubara axis; analytic continuation to real frequencies \omega is performed post-convergence using methods like maximum entropy or Padé approximation to obtain spectral functions A(\omega) = -\frac{1}{\pi} \Im G(\omega + i0^+).^[4] This step is ill-posed and sensitive to noise, particularly in quantum Monte Carlo data, requiring careful regularization to avoid artifacts.^[4]

Key Approximations and Limitations

Locality approximation for the self-energy

The locality approximation for the self-energy constitutes the central assumption in dynamical mean-field theory (DMFT), positing that the self-energy \Sigma(\mathbf{k}, i\omega_n) is momentum-independent and thus local in real space, \Sigma_{ij}(\tau) \approx \delta_{ij} \Sigma(\tau). This approximation emerges rigorously in the limit of infinite spatial dimensions d \to \infty, where the hopping amplitude between nearest neighbors scales as t^* / \sqrt{z} with coordination number z \propto 2d, rendering non-local contributions to the self-energy of order $1/\sqrt{z} \to 0. Consequently, \Sigma(\mathbf{k}, \omega) - \Sigma_{\rm loc}(\omega) \sim 1/\sqrt{z} \to 0, justifying \Sigma(\mathbf{k}, \omega) \approx \Sigma(\omega), a site-local quantity that depends only on frequency. Physically, this reflects the dominance of local interactions, such as the on-site Hubbard repulsion U, over spatial fluctuations in high dimensions, where quantum fluctuations are effectively frozen on intersite scales, allowing the lattice problem to be mapped onto a single-site quantum impurity embedded in a self-consistent bath. In this regime, short-time, real-space correlations prevail, enabling DMFT to capture dynamical local effects like spectral weight transfer without needing full spatial resolution. Corrections for finite dimensions, where non-local effects become relevant, are addressed through extensions such as cluster methods. The derivation proceeds from the functional-integral formalism via the Luttinger-Ward \Phi-functional, which generates the thermodynamic potential and ensures conservation laws through its two-particle-irreducible skeleton diagrams. In DMFT, \Phi[G] is approximated by its local form \Phi_{\rm loc}[G] = \sum_i f[G_{ii}], summing only diagrams that collapse to single-site contributions in d \to \infty, as non-local skeleton diagrams vanish due to the central-limit theorem for the density of states. The self-energy then follows as the functional derivative \Sigma_{ij}(i\omega_n) = \frac{\delta \Phi_{\rm loc}[G]}{\delta G_{ji}(i\omega_n)} \approx \delta_{ij} \Sigma(i\omega_n), with self-consistency enforced by equating the local lattice Green's function to the impurity one:

G_{\rm loc}(i\omega_n) = \int d\epsilon \, \rho(\epsilon) \frac{1}{i\omega_n + \mu - \epsilon - \Sigma(i\omega_n)},

where \rho(\epsilon) is the non-interacting density of states. This local \Phi-approximation maintains thermodynamic consistency while simplifying computations. Despite its strengths, the locality approximation has notable limitations, particularly in finite dimensions where it neglects short-range non-local correlations, such as those underlying antiferromagnetic order or stripe phases, leading to unphysical results near ordering instabilities. For instance, in low dimensions, momentum-dependent self-energy components become significant, requiring beyond-DMFT techniques to restore accuracy.

Infinite-dimensional limit and coordination number effects

In the infinite-dimensional limit, dynamical mean-field theory (DMFT) becomes exact for lattice models of correlated electrons, such as the Hubbard model. This limit is achieved by considering either the spatial dimension d \to \infty or the coordination number z \to \infty, where z represents the number of nearest neighbors per site. To ensure finite kinetic energy and a well-defined density of states (DOS) as z \to \infty, the hopping parameters are scaled such that the off-diagonal elements satisfy t_{ij} \sim t^*/\sqrt{z} for i \neq j, with t^* fixed. Consequently, the bandwidth W scales as W \sim 2\sqrt{z} t^*, diverging as z \to \infty, while the variance of the density of states \langle \epsilon^2 \rangle = (t^*)^2 remains fixed.^[1] This scaling, originally proposed for maintaining nontrivial correlations in high dimensions, leads to a Gaussian DOS for the noninteracting system:

D(\varepsilon) = \frac{1}{\sqrt{2\pi (t^*)^2}} \exp\left(-\frac{\varepsilon^2}{2(t^*)^2}\right).

The exactness of DMFT in this limit arises from the self-averaging of local quantities and the suppression of nonlocal correlations. Momentum integrals over the Brillouin zone become self-averaging due to the central limit theorem, as contributions from a large number of neighbors (z \to \infty) dominate, making site-to-site fluctuations negligible. Additionally, vertex corrections in diagrammatic perturbation theory vanish proportionally to $1/z, ensuring that the self-energy \Sigma_{ij}(i\omega_n) is strictly local: \Sigma_{ij}(i\omega_n) = \delta_{ij} \Sigma(i\omega_n). This locality approximation, a direct consequence of the infinite-dimensional scaling, reduces the lattice problem to a self-consistent single-site impurity model without loss of accuracy. For finite coordination numbers, DMFT provides a controlled approximation via a $1/z expansion. The momentum dependence of the self-energy emerges as higher-order corrections: \Sigma(\mathbf{k}, \omega) = \Sigma(\omega) + (1/z) \Sigma^{(2)}(\mathbf{k}, \omega) + \mathcal{O}(1/z^2), where \Sigma^{(2)}(\mathbf{k}, \omega) captures initial nonlocal effects. These corrections diminish rapidly with increasing z, justifying DMFT's use even for moderate values. For instance, on the infinite-dimensional hypercubic lattice (where z = 2d), DMFT is exact as d \to \infty; in practice, it yields accurate results for three-dimensional simple cubic lattices with z = 6, as the $1/z terms remain small compared to the leading local contribution.

Applications to Correlated Electron Systems

Mott metal-insulator transition

In the single-orbital Hubbard model at half-filling within dynamical mean-field theory (DMFT), the Mott metal-insulator transition occurs as the on-site Coulomb repulsion U is increased relative to the bandwidth W. For U < W, the system remains metallic, characterized by a finite density of states at the Fermi level. As U approaches a critical value, a first-order transition emerges at finite temperatures, leading to a coexistence region between metallic and insulating solutions for U_{c1} < U < U_{c2}, where U_{c1} marks the instability of the metallic phase and U_{c2} the instability of the insulating phase. At zero temperature, the transition becomes continuous at U_{c2}, with the metallic phase stable for U < U_{c2} and insulating for U > U_{c2}. Numerical solutions using quantum Monte Carlo (QMC) impurity solvers on the Bethe lattice yield U_{c1} \approx 2.4W and U_{c2} \approx 3.0W, where W is the half-bandwidth. The spectral function A(\omega) provides a clear signature of the transition. In the metallic phase, a sharp quasiparticle peak appears at the Fermi energy \omega = 0, reflecting coherent fermionic excitations with a renormalized bandwidth ZW, where Z is the quasiparticle residue. Incoherent contributions form broad Hubbard bands centered around \pm U/2. Upon entering the insulating phase beyond U_{c2}, the quasiparticle peak vanishes, and a charge gap opens between the lower and upper Hubbard bands at \pm U/2, suppressing the density of states at the Fermi level. This evolution is captured through the DMFT self-consistency loop, where the local Green's function is iteratively solved for the effective Anderson impurity model. The Brinkman-Rice picture, derived from the Gutzwiller variational approach, elucidates the mechanism of the transition in DMFT by focusing on the double occupancy suppression. The effective mass diverges as m^*/m \sim 1/(1 - (U/U_{c2})^2), indicating the breakdown of Fermi liquid behavior at U_{c2}. This is tied to the quasiparticle weight Z = \left[1 - \left. \frac{\partial \operatorname{Re} \Sigma(\omega)}{\partial \omega} \right|_{\omega=0} \right]^{-1}, which approaches zero at the transition, quantifying the loss of coherence. DMFT refines this picture by incorporating dynamical correlations, confirming the divergence while predicting a finite local susceptibility due to exchange processes.

Magnetic and superconducting phases

In dynamical mean-field theory (DMFT) applied to the single-orbital Hubbard model, itinerant antiferromagnetism at half-filling is captured by breaking translational symmetry through a two-sublattice approach, where the unit cell is doubled to account for the Néel state with opposite spins on adjacent sites. This is implemented by introducing a staggered magnetic field h_{\text{stag}} that couples to the spin density with wave vector \mathbf{Q} = (\pi, \pi), opening an antiferromagnetic gap in the electronic spectrum. For lattices with perfect nesting of the Fermi surface, such as the Bethe lattice or bipartite structures in the infinite-dimensional limit, the antiferromagnetic transition occurs at a critical interaction strength U_c \approx 1.5 W, where W is the bandwidth, marking the onset of long-range magnetic order from a paramagnetic metal. The locality approximation inherent to single-site DMFT limits its ability to fully describe long-range order without such symmetry breaking, but it accurately predicts the enhancement of antiferromagnetism due to nesting, with the Néel temperature scaling as T_N \propto t^2 / U at large U. For superconducting phases, DMFT combined with cluster extensions and flexible impurity solvers, such as the non-crossing approximation, reveals d-wave pairing in the doped Hubbard model away from half-filling.^[12] In the intermediate-to-strong coupling regime (U > W), the theory predicts a superconducting dome as a function of doping \delta, with the critical temperature T_c determined via Eliashberg-like equations that incorporate the anomalous self-energy from the Nambu-Gorkov formalism.^[12] The d-wave symmetry arises from repulsive interactions mediated by antiferromagnetic fluctuations, yielding a maximum T_c \approx 0.05 t (or ~150 K for realistic parameters) near optimal doping \delta \approx 0.2, consistent with the phase diagram of high-temperature cuprates. In two-dimensional realizations of the doped Hubbard model treated with cluster DMFT, a pseudogap emerges above T_c in the underdoped regime, attributed to the formation of preformed electron pairs without long-range coherence. This pseudogap manifests as a suppression of spectral weight near the antinodal Fermi surface points, driven by short-range d-wave pairing correlations within the cluster, and persists up to temperatures higher than T_c due to phase fluctuations. Recent extensions of DMFT to cuprate models, incorporating multi-orbital effects and cluster treatments, have demonstrated the capture of stripe order—intertwined charge and spin density waves—in hole-doped systems.^[13] For instance, a 2023 study on infinite-layer nickelates using DFT+DMFT revealed stripe-like charge order with a (1/3, 0, 0) periodicity, stabilized by charge transfer and strong correlations at low doping.^[13] In La-based cuprates, DMFT approaches have explored related competing orders such as charge nematicity.^[14] These results highlight DMFT's role in elucidating the interplay between magnetic order and unconventional pairing in real materials.

Extensions and Modern Developments

Multi-orbital and realistic material implementations

Dynamical mean-field theory (DMFT) has been generalized to multi-orbital systems to capture the rich physics arising from orbital degrees of freedom in transition metal compounds, where multiple d-orbitals are partially filled and strongly interacting.^[15] In this framework, the local Hamiltonian includes crystal-field splitting and multi-orbital interactions, extending beyond the single-orbital Hubbard model. The non-interacting part incorporates orbital-dependent onsite energies, \sum_{m m' \sigma} \varepsilon_{m m'} n_{m \sigma}, while the interaction term is

H_{\text{int}} = U \sum_m n_{m\uparrow} n_{m\downarrow} + (U' - J) \sum_{m < m'} n_m n_{m'} + J \sum_{m \neq m'} (c^\dagger_{m\uparrow} c_{m\downarrow} c^\dagger_{m'\downarrow} c_{m'\uparrow} + \text{H.c.}).

where U is the intra-orbital Coulomb repulsion, U' the inter-orbital repulsion (typically U - 2J), and J the Hund's exchange coupling that favors high-spin states.^[15] These terms, often parameterized by rotationally invariant Slater integrals F^0, F^2, F^4 to ensure full rotational symmetry in the orbital space, allow DMFT to model phenomena like orbital-selective Mott transitions and Hund's metal behavior.^[16] To apply multi-orbital DMFT to realistic materials, the LDA+DMFT approach integrates density functional theory (DFT) in the local density approximation (LDA) with DMFT, embedding a correlated subspace—such as the d-orbitals of transition metals—into a non-interacting bath derived from LDA band structures.^[17] The LDA provides the one-body Hamiltonian and projects the full electronic structure onto localized Wannier orbitals for the correlated subspace, downfolding high-energy degrees of freedom to form an effective multi-orbital Anderson impurity model.^[17] Double-counting corrections subtract the mean-field treatment of interactions already included in LDA, and the DMFT self-energy, which is local and frequency-dependent, is computed orbitally resolved as \Sigma_{m m'}(\omega) to update the lattice Green's function self-consistently.^[18] This hybrid method captures both band formation from delocalized electrons and strong local correlations in the embedded subspace.^[17] In applications to real materials, LDA+DMFT has elucidated itinerant magnetism in Sr_2RuO_4, where multi-orbital correlations enhance spin fluctuations in the Ru 4d bands, stabilizing ferromagnetic tendencies near the Fermi level without long-range order at low temperatures.^[19] For iron pnictide superconductors like LaFeAsO, five-orbital DMFT models reveal multi-orbital pairing mechanisms, with s_{\pm}-wave superconductivity emerging from antiferromagnetic fluctuations and orbital-selective renormalization factors that suppress pairing in certain d-orbitals, yielding critical interactions around U_c \approx 1.6 eV.^[20] Recent advances as of 2025 have improved the efficiency of ab initio multi-orbital DMFT through natural orbital bases, which diagonalize the one-particle density matrix to reduce the impurity solver's Hilbert space dimensionality.^[21] The natural orbitals renormalization group (NORG) solver, integrated with DFT+DMFT, employs occupancy restrictions and zero-temperature formulations to accelerate computations for d-electron materials like SrVO_3 and MnO, achieving accurate spectral functions with reduced sign problems compared to traditional continuous-time quantum Monte Carlo methods.^[21] This approach enhances scalability for realistic multi-orbital implementations, enabling studies of complex oxides with full orbital resolution.^[21]

Cluster and diagrammatic extensions

Cluster extensions of dynamical mean-field theory (DMFT) address the limitations of the single-site approximation by incorporating short-range spatial correlations through finite clusters of lattice sites, thereby improving the treatment of phenomena like antiferromagnetism in low dimensions. In cluster DMFT (CDMFT), the lattice is divided into real-space clusters, such as a 2x2 plaquette for the two-dimensional Hubbard model, where the self-energy becomes cluster-dependent, denoted as \Sigma_R(\omega) for cluster sites R, allowing the capture of intra-cluster dynamics beyond the local approximation. This approach starts from the locality of the self-energy but extends it to include short-range non-local effects within the cluster, solved via an effective impurity problem for the cluster embedded in a self-consistent bath. Seminal work by Kotliar et al. introduced CDMFT as a natural generalization of single-site DMFT, demonstrating its ability to resolve pseudogaps and stripe orders in correlated systems.^[22] A variant, cellular DMFT (C-DMFT), emphasizes the cellular structure for realistic materials, focusing on supercells to handle multi-orbital physics, while the dynamical cluster approximation (DCA) coarse-grains the Brillouin zone into patches to preserve translational invariance. To reconstruct lattice quantities from cluster results, periodization schemes are essential; the cumulant periodization, which periodizes the cumulant M(\omega) = [\omega + \mu - \epsilon_k - \Sigma(\omega)]^{-1}, often converges faster and better preserves causality compared to Green's function or self-energy periodization, particularly for larger clusters in frustrated lattices like the Kagome. Studies show that cumulant periodization yields improved spectral functions with reduced artifacts in one-dimensional chains and two-dimensional Hubbard models. Notably, CDMFT with a 2x2 plaquette captures two-dimensional antiferromagnetism in the half-filled Hubbard model, which single-site DMFT misses due to the absence of short-range order in the infinite-dimensional limit.^[23]^[24] Diagrammatic extensions complement cluster methods by perturbatively including non-local correlations beyond short-range clusters, often using the DMFT solution as a starting point to add $1/z (long-wavelength) diagrams without explicit cluster periodization. The dual fermion approach maps the lattice problem onto a weakly interacting gas of dual fermions, incorporating non-local self-energy corrections \Sigma'(\omega, k) via a diagrammatic expansion around the DMFT impurity model, efficiently capturing antiferromagnetic fluctuations and Fermi arcs in the two-dimensional Hubbard model at moderate computational cost. Similarly, the TRILEX (Triply Irreducible Local EXpansion) method uses a local two-particle irreducible vertex from the impurity solver to resum bosonic fluctuation diagrams, enabling the description of d-wave superconductivity in the Hubbard model with a single-site cluster, unlike CDMFT which requires multi-site setups for such pairing. Recent 2024 reviews highlight how these extensions, particularly cluster methods, yield improved spectral functions in cuprates, reproducing waterfall-like features and charge nematicity observed in experiments for hole-doped systems.^[25]^[26]^[27]

Non-equilibrium and quantum computing applications

Non-equilibrium dynamical mean-field theory (NE-DMFT) generalizes the equilibrium DMFT framework to describe time-dependent phenomena in strongly correlated systems, employing the Keldysh contour to handle real-time evolution of contour-ordered Green's functions.^[28] In this approach, the lattice problem is mapped onto a single-impurity Anderson model with a time-dependent hybridization function \Delta(t, t'), which captures the dynamical coupling between the impurity and the effective bath; the contour-ordered Green's function G(t, t') satisfies a Dyson equation involving the self-energy \Sigma(t, t') and \Delta(t, t').^[29] This formulation enables the simulation of non-equilibrium steady states and transients, such as those induced by external fields or sudden quenches, by iteratively solving the impurity problem on the closed-time-path contour.^[30] Key applications of NE-DMFT include modeling pump-probe spectroscopy, where ultrafast laser pulses excite the system, and the subsequent relaxation dynamics are probed to reveal correlation effects like photo-induced phase transitions in Mott insulators.^[31] For solving the time-dependent impurity models, methods such as the auxiliary master equation approach embed the impurity in a finite auxiliary bath and evolve the density matrix using a Lindblad master equation, allowing efficient computation of steady-state properties without the need for full real-time propagation.^[32] Matrix product states have also been employed as impurity solvers, enabling larger bath discretizations and accurate treatment of long-time dynamics in one-dimensional-like impurity problems.^[33] A notable prediction from NE-DMFT is the emergence of Floquet phases in the periodically driven Hubbard model, where coherent driving stabilizes non-equilibrium steady states with modified magnetic ordering, such as ultra-fast control of antiferromagnetic relaxation. These phases arise from the interplay of driving frequency and interaction strength, offering insights into light-induced topological states inaccessible in equilibrium.^[34] In quantum computing applications, DMFT leverages variational quantum eigensolvers (VQE) and quantum phase estimation (QPE) to solve the single-impurity Anderson model, particularly for ground-state properties in regimes where classical methods like quantum Monte Carlo suffer from the fermion sign problem at high interaction strengths U.^[35] VQE optimizes trial wavefunctions on noisy intermediate-scale quantum (NISQ) devices to approximate the impurity Green's function, providing a hybrid quantum-classical loop that demonstrates potential quantum advantage for strongly correlated impurities by reducing exponential scaling in bath orbitals.^[36] Recent implementations have extended this to full self-consistent DMFT cycles for real materials, such as Ca₂CuO₂Cl₂, where NISQ simulations capture correlation effects with quasi-particle weights consistent with experiments, showing promise in high-U equilibrium regimes by avoiding the fermion sign problem in ground-state calculations.^[37]^[3] However, as of 2025, these NISQ implementations are limited to small bath sizes due to noise and decoherence, with full scalability pending fault-tolerant quantum computers. These approaches highlight DMFT's potential on quantum hardware for tackling equilibrium dynamics in multi-orbital systems, with demonstrations achieving chemical accuracy for impurity spectra on current platforms.^[3]

References

[1]
Dynamical mean-field theory of strongly correlated fermion systems ...
Jan 1, 1996 · We review the dynamical mean-field theory of strongly correlated electron systems which is based on a mapping of lattice models onto quantum impurity models.
[2]
https://doi.org/10.7566/JPSCP.30.011001
[3]
[PDF] Dynamical mean-field theory of strongly correlated fermion systems ...
We review the dynamical mean-field theory of strongly correlated electron systems which is based on a mapping of lattice models onto quantum impurity models ...
[4]
Correlated Lattice Fermions in Dimensions | Phys. Rev. Lett.
Jan 16, 1989 · We show that even in d = ∞ dimensions the Hubbard model, when scaled properly, describes nontrivial correlations among fermions.Missing: paper | Show results with:paper
[5]
Hubbard model in infinite dimensions | Phys. Rev. B
Mar 15, 1992 · We present an exact mapping of the Hubbard model in infinite dimensions onto a single-impurity Anderson (or Wolff) model supplemented by a self-consistency ...Missing: history | Show results with:history
[6]
Electronic structure calculations with dynamical mean-field theory
Aug 14, 2006 · Dynamical mean-field theory ( Georges et al., 1996 ) is the simplest approach which captures these features; it has been extensively developed ...
[7]
Applications of DFT + DMFT in Materials Science - Annual Reviews
Mar 8, 2019 · The DFT + DMFT approach is now well developed and tested enough to have predictive capabilities and is used extensively on materials science ...Missing: integration | Show results with:integration<|control11|><|separator|>
[8]
Electron correlations in narrow energy bands - Journals
A simple, approximate model for the interaction of electrons in narrow energy bands is introduced. The results of applying the Hartree-Fock approximation to ...
[9]
DMFT_W2K Tutorial 1: MnO
Within 10 or so dmft iterations (around 120 charge iterations), the total energy and the free energy are converged within 1meV (using 216 cores for impurity ...
[10]
Dynamical mean field theory for real materials on a quantum computer
Oct 31, 2025 · Introduction. Technological advances heavily rely on the design of innovative functional materials, a task chiefly driven by understanding ...
[11]
Convergence acceleration and stabilization of dynamical mean-field ...
Sep 25, 2009 · In simple linear mixing there is no exchange of information between different energies (“crosstalk”), thus it takes many DMFT iterations for ...Missing: accelerators | Show results with:accelerators
[12]
[cond-mat/0002352] d-wave Superconductivity in the Hubbard Model
Feb 23, 2000 · Access Paper: View a PDF of the paper titled d-wave Superconductivity in the Hubbard Model, by Th. Maier and 3 other authors. View PDF · TeX ...Missing: DMFT seminal Scalapino
[13]
An electronic origin of charge order in infinite-layer nickelates - Nature
Sep 6, 2023 · In this work, we first use density functional theory plus dynamical mean field theory (DFT+DMFT) calculations to reveal a special charge- ...
[14]
Cluster dynamical mean-field study of intra-unit-cell charge ... - PNAS
Hole-doped superconducting cuprates can display intra-unit-cell charge nematic order that breaks symmetry under rotation by ninety degrees.
[15]
[PDF] LDA+DMFT: Multi-Orbital Hubbard Models - cond-mat.de
This story is nicely told in a review article by Florin Diacu [2], who concludes. Did this mean the end of the N-body problem? Was this old question ...
[16]
A charge self-consistent LDA+DMFT study of the spectral properties ...
Sep 25, 2014 · The fully rotationally invariant U-matrix has been constructed from the Slater parameters F0, F2 and F4. These have been reduced to only two ...
[17]
[PDF] The LDA+DMFT approach to strongly correlated materials - JuSER
During the last few years conventional band-structure calculations in the local density ap- proximation (LDA) have been merged with a modern many-body approach ...
[18]
404 Not Found
Insufficient relevant content.
[19]
Orbital ordering and magnetism in layered Perovskite Ruthenate Sr ...
Apr 27, 2020 · We study the basic electronic and magnetic properties of Sr 2 RuO 4 including the magnetism, orbital ordering, correlation, pressure, and strain effect
[20]
Dynamical mean-field study on the superconductivity mediated by ...
In summary, we have investigated the electronic state and the superconductivity in the 5-orbital. Hubbard model for iron pnictides by using the DMFT+ED method ...
[21]
Ab initio dynamical mean field theory with natural orbitals ...
Mar 31, 2025 · In this study, we introduce a novel implementation of density functional theory integrated with single-site dynamical mean-field theory to ...Missing: review | Show results with:review
[22]
Cellular Dynamical Mean Field Approach to Strongly Correlated ...
Oct 15, 2001 · We propose a cellular version of dynamical mean field theory which gives a natural generalization of its original single-site construction.
[23]
[PDF] Electronic structure calculations with dynamical mean-field theory
Aug 14, 2006 · dynamical mean-field theory 共DMFT兲共Georges et al.,. 1996; Kotliar and Vollhardt, 2004兲 with electronic struc- ture techniques 关for ...
[24]
Cluster-size dependence in cellular dynamical mean-field theory
Dec 14, 2011 · We examine the cluster-size dependence of the cellular dynamical mean-field theory (CDMFT) applied to the two-dimensional Hubbard model.
[25]
Dual fermion approach to the two-dimensional Hubbard model - arXiv
Oct 21, 2008 · We present an efficient diagrammatic method to describe nonlocal correlation effects in lattice fermion Hubbard-like models.
[26]
[1705.08332] TRILEX and $GW$+EDMFT approach to $d - arXiv
May 23, 2017 · We show that, in the two-dimensional Hubbard model, at strong coupling, TRILEX yields a d-wave superconducting dome as a function of doping.
[27]
Restoring translational symmetry in periodic all-orbital dynamical ...
Jul 30, 2024 · Dynamical mean-field theory (DMFT) and its cluster extensions provide an efficient Green's function formalism to simulate spectral properties of ...
[28]
Characterizing the Non-equilibrium Dynamics of Field-Driven ...
We review non-equilibrium dynamical mean-field theory studies of both the transient and the steady states of a DC field-driven correlated quantum system.
[29]
Nonequilibrium Dynamical Mean Field Theory: an auxiliary ... - arXiv
Oct 15, 2012 · We introduce a versatile method to compute electronic steady state properties of strongly correlated extended quantum systems out of equilibrium.Missing: solvers CHESTNUT
[30]
An Auxiliary Quantum Master Equation Approach | Phys. Rev. Lett.
Feb 20, 2013 · The approach is based on dynamical mean-field theory (DMFT), in which the original system is mapped onto an auxiliary nonequilibrium impurity ...Missing: CHESTNUT | Show results with:CHESTNUT
[31]
Nonequilibrium electron dynamics in pump-probe spectroscopy
Dec 7, 2018 · This method corresponds to the nonequilibrium DMFT [34, 35] with a low-order perturbation theory as an impurity solver where both methods work ...
[32]
Auxiliary master equation approach to nonequilibrium correlated ...
Apr 7, 2014 · We present a numerical method for the study of correlated quantum impurity problems out of equilibrium, which is particularly suited to address steady-state ...Missing: CHESTNUT | Show results with:CHESTNUT
[33]
Solving nonequilibrium dynamical mean-field theory using matrix ...
Oct 13, 2014 · We solve the nonequilibrium dynamical mean-field theory (DMFT) using matrix product states (MPS). This allows us to treat much larger bath sizes.Missing: CHESTNUT | Show results with:CHESTNUT
[34]
[PDF] Ultra-fast control of magnetic relaxation in a periodically driven ...
Oct 25, 2017 · Our focus is on the Hubbard model for a Bethe lat- tice in infinite dimensions where non-equilibrium dy- namical mean-field theory (NE-DMFT) can ...
[35]
Self-Consistent Determination of Single-Impurity Anderson Model ...
We realize a self-consistent determination of the single impurity Anderson model through a feedback loop between quantum and classical computations.
[36]
Variational quantum eigensolver for dynamic correlation functions
In this work, we show how the calculation of zero-temperature dynamic correlation functions defining the linear response characteristics of quantum systems can ...
[37]
[2508.00118] Dynamical mean field theory with quantum computing
Jul 31, 2025 · Dynamical mean field theory, on the other hand, maps strongly-correlated lattice models like the Hubbard model onto simpler, yet still many-body ...Missing: VQE QPE SIAM