Kubo-Greenwood Linear Response Framework

• The Kubo-Greenwood framework is a quantum method that evaluates electrical and optical conductivity via eigenstate-based sum-over-states expressions and current correlation functions.
• It employs approximations such as the frozen-ion and independent-particle models with ad hoc broadening to simplify microscopic evaluations in ab initio simulations.
• Its practical applications span systems from crystalline solids to warm dense matter, offering actionable insights into transport phenomena, localization, and multiband effects.

The Kubo-Greenwood linear response framework provides the central methodology for evaluating the electrical conductivity and related transport properties of quantum many-body systems using microscopic quantum mechanics. Stemming from the general Green-Kubo approach to linear response, the Kubo-Greenwood (KG) formalism yields a practical, eigenstate-based sum-over-states expression for the optical and DC conductivity that is widely adopted in ab initio electronic structure codes, particularly for complex materials, systems under extreme conditions, and disordered or multiband systems. Extensions of this framework underpin contemporary approaches to spintronics, localization, high-temperature matter, and quantum kinetic theory.

1. Theoretical Foundations and Derivation

The KG formalism originates from the quantum Kubo linear response, which expresses the frequency-dependent conductivity via current-current (or momentum-momentum) correlation functions in thermal equilibrium: $$\sigma(\omega)\;=\;\frac{1}{\omega}\left[1-e^{-\beta\hbar\omega}\right]\, \Re\int_0^\infty dt\, e^{i\omega t}\, \psi(t)$$ with

$$\psi(t)\;=\;\frac{1}{3V}\left\langle\widehat{\mathbf J}(t)\cdot\widehat{\mathbf J}(0)\right\rangle, \quad \widehat{\mathbf J} = e\sum_{i=1}^{N_e} \frac{\widehat{\mathbf p}_i}{m_e}$$

where the brackets denote the grand canonical average and $V$ is the system volume (Dufty et al., 2017, Cytter et al., 2019, Zhang et al., 2010).

By inserting a complete set of many-electron eigenstates and using the Lehmann representation, the current correlation reduces to sums over matrix elements and Boltzmann factors. For a non-interacting or effective independent-particle (e.g., Kohn-Sham) system, it becomes

$$\sigma(\omega) = \frac{2\pi e^2\hbar}{3m^2V\omega}\sum_{i,j} [f_i - f_j] |\langle \psi_i | \widehat{p} | \psi_j \rangle|^2 \delta(\epsilon_j - \epsilon_i - \hbar\omega)$$

where $f_i$ are Fermi-Dirac occupations (Cytter et al., 2019, Dufty et al., 2017).

The KG formula naturally decomposes into intraband (Drude-type) and interband (quantum) contributions relevant for conduction band and interband transitions, respectively. For multiband and disordered systems, explicit expressions for each tensor component and their symmetry constraints are derivable (Valet et al., 2024, Wimmer et al., 2016, Huhtinen et al., 2022).

2. Approximations and Physical Assumptions

Transitioning from the full Green-Kubo formula to the KG expression rests on several physical approximations:

• Frozen-ion approximation: Ion positions are fixed (as in ab initio molecular dynamics “snapshots”); electron–ion dynamics is projected onto a static lattice. Valid if electron relaxation is much faster than ionic motion (Dufty et al., 2017).
• Independent-particle approximation: Electron–electron dynamical correlations are neglected beyond static mean-field (e.g., Kohn-Sham DFT), omitting explicit electron-electron scattering and dynamical exchange-correlation (Dufty et al., 2017, Zhang et al., 2010).
• Ad hoc broadening: The spectrum in a finite simulation cell is discrete; Dirac delta functions in transition energies are replaced by broadened forms (Lorentzian or Gaussian) with a small width $\eta$. The width must exceed the single-particle spacing but not wash out physical features (Dufty et al., 2017, Bulanchuk, 2019).
• Homogeneity and uniform carrier density: Uniformity enables replacement of the full current operator with the kinetic-velocity component, justifying the use of the KG formula in the weak gradient limit (Zhang et al., 2010).

These approximations define the range of physical applicability, especially regarding the DC limit, localization, and the role of disorder.

3. Practical Implementation and Computational Methods

The KG formalism is implemented in electronic structure codes to compute conductivity spectra for crystalline, amorphous, or disordered systems using finite-temperature Kohn-Sham DFT or tight-binding models (Dufty et al., 2017, Cytter et al., 2019). The practical workflow involves:

1. Generating an ensemble of ionic configurations (e.g., via AIMD).
2. For each snapshot, diagonalizing the effective one-particle Hamiltonian to obtain eigenvalues $\epsilon_n$ and orbitals $\psi_n$.
3. Evaluating Fermi occupations, momentum or velocity operator matrix elements, and constructing the double sum of transitions with appropriate broadening.
4. Sampling over $k$-points in the Brillouin zone for periodic boundary conditions; careful convergence in the number of bands, $k$-points, plane-wave cutoffs, and pseudopotential effects is mandatory (Dufty et al., 2017).

For large systems or high temperatures, the cubic scaling with system size and temperature of standard deterministic eigensolver-based approaches renders computations expensive (Cytter et al., 2019). Stochastic methods—such as stochastic DFT with Chebyshev polynomial expansion of spectral functions—enable linear scaling, as they bypass explicit diagonalization and use random wavefunction sampling to evaluate traces and time-propagators (Cytter et al., 2019, Fan et al., 2013).

The table below summarizes standard broadening schemes used in practical KG implementations:

Broadening Type	Functional Form	Typical Use
Lorentzian	$$\delta_\eta(x) = \frac{1}{\pi}\frac{\eta}{x^2+\eta^2}$$	Default, ad hoc width $\eta$
Gaussian	$$\delta_\eta(x) = \frac{1}{\sqrt{\pi}\eta} e^{-x^2/\eta^2}$$	Alternative; smoother decay
Sinc	$$\delta_\eta(x) = \frac{1}{2\pi} \sin(2x/\eta)/(x/\eta)$$	Finite time integrations (Bulanchuk, 2019)

Extrapolation schemes—using Drude-based functional forms—improve the accuracy of DC conductivity extraction in finite cells by fitting the $\eta$-dependent "pseudo-DC" conductivity and extrapolating to the double limit $\lim_{\eta \to 0}\lim_{V \to \infty}$ (Bulanchuk, 2019).

4. Extensions: Multiband, Disorder, and Relativistic Effects

For weakly disordered or multiband materials, the KG framework is underpinned microscopically by quantum kinetic (Keldysh) theory and diagrammatic perturbation. The total current response splits uniquely into intraband (Drude) and interband (“quantum-coherent”) density-matrix sectors. The latter yields the KG term via non-Abelian Berry connections and includes both intrinsic (Berry curvature) and extrinsic (vertex/ladder corrections) physics. Mesoscopic real-space gradients introduce new boundary-driven interband source terms not present in bulk KG treatments (Valet et al., 2024).

The fully relativistic generalization, as in the Kubo-Bastin formalism, treats spin-orbit physics, magnetic point group symmetries, and disorder via multiple-scattering Korringa-Kohn-Rostoker (KKR) plus the coherent potential approximation (CPA) (Wimmer et al., 2016, Ebert et al., 2011). In this framework, conductivity, spin conductivity, and related response tensors (e.g., torkance, Gilbert damping) are computed as traces over combinations of velocity, torque, or spin-current operators with Green’s function spectral projectors.

5. Relation to Other Transport Approaches and Regime-Specific Features

The KG formalism is equivalent to the Green-Kubo time-correlation approach and to wavefunction-based microscopic response methods in the uniform density limit (Zhang et al., 2010). In the diffusive regime, both the Green-Kubo velocity autocorrelation and the Einstein mean-square displacement forms yield identical DC conductivities and facilitate efficient real-time linear-scaling algorithms (Fan et al., 2013, Uppstu et al., 2013).

In the ballistic and localization regimes, special care is required:

• Ballistic: Conductivity diverges; conductance is regularized using an emergent propagation length from wavepacket dynamics (Fan et al., 2013, Uppstu et al., 2013).
• Localization: Mean-square displacement saturates at long times; extraction of the localization length is possible either by fits to conductance decay or directly from propagation length saturation, with the KG approach agreeing quantitatively with recursive Green’s function (Landauer–Büttiker) calculations up to several localization lengths (Uppstu et al., 2013).
• Disordered finite systems: The KG method with appropriate $\eta$-broadening and Drude-extrapolation recovers macroscopic DC conductivity values consistent with Landauer approaches, circumventing ambiguities inherent to finite-size spectra (Bulanchuk, 2019).

6. Critique, Limitations, and Pathologies

Applications of the KG formalism require careful attention to the following limitations:

• Missed many-body and vertex corrections: The standard KG formula omits dynamical electron-electron scattering, many-body correlations, and higher-order disorder effects, all of which may be significant in strongly correlated or highly disordered systems. Vertex corrections are needed for certain magneto-transport and nonlinear responses (Zhang et al., 2010, Wimmer et al., 2016, Valet et al., 2024).
• Flat-band anomalies: In perfectly flat bands, the DC (longitudinal) conductivity strictly vanishes in the clean limit. Spurious finite results can arise from inappropriate use of $\frac{\partial f}{\partial \epsilon}$ approximations or careless application of the Kubo-Streda formula, leading to incorrect predictions of DC conductivity proportional to the quantum metric. Only the antisymmetric (Berry curvature) term correctly survives, contributing to anomalous Hall effects (Huhtinen et al., 2022).
• Ad hoc broadening: The choice of $\eta$ is not physically controlled and affects fine structure, especially at low temperatures and for small systems.
• Thermodynamic limit and finite-size effects: Extrapolation with respect to both system volume and broadening is required; improper order of limits can yield unphysical results (Bulanchuk, 2019, Huhtinen et al., 2022).
• Strong disorder or inhomogeneity: The KG framework is valid for weak to moderate disorder and uniform density; for strong localization, percolation, or non-uniform systems, more sophisticated methods may be required (Uppstu et al., 2013, Zhang et al., 2010).

7. Applications and Impact

The KG formalism is universally adopted for calculation of optical and DC conductivity, dielectric functions, localization properties, and magnetization dissipation parameters in:

• Bulk metals, semiconductors, and insulators within DFT (Dufty et al., 2017, Cytter et al., 2019).
• Warm dense matter and extreme conditions; transition to metallization is characterized by KG-computed conductivity (Cytter et al., 2019).
• Quantum transport in 2D materials (e.g., graphene), with optimizations to run efficiently on GPU hardware for million-site models (Fan et al., 2013).
• Disordered alloys, spin-torque computations, and Gilbert damping, leveraging relativistic, multiple-scattering, and CPA techniques (Wimmer et al., 2016, Ebert et al., 2011).
• Extraction of the Anderson localization length and diffusive-to-localized regime crossovers (Uppstu et al., 2013, Fan et al., 2013).

The KG formalism bridges microscopic quantum theory and experimentally accessible transport quantities, underpinning modern computational materials physics and serving as a foundation for advanced quantum kinetic and mesoscopic transport approaches (Valet et al., 2024).