Virtual Spectral Conductivity Model

• Virtual spectral conductivity models are approaches that infer bulk transport coefficients by contracting an intermediate, energy-resolved spectral object, enhancing prediction accuracy.
• They employ forward transport maps such as Kubo sums, Sommerfeld–Bethe integrals, and Stieltjes transforms to translate microscopic data into observable responses.
• This approach bridges different physical regimes—from electronic and phononic transport to holographic and effective media implementations—improving modeling versatility.

Searching arXiv for the cited papers to ground the article in current records. Searching for "(Narikiyo, 2017)". Searching for "(Hirosawa et al., 2022)". Searching for "(Murphy et al., 2024)". Searching for "(Kikuchi et al., 13 May 2026)". The expression “Virtual Spectral Conductivity Model” (Editor’s term) can be used as an umbrella label for transport frameworks in which conductivity is reconstructed from an intermediate resolved object rather than treated ab initio as a single bulk coefficient. In the literature synthesized here, that resolved object may be a dressed electron spectral function, an energy-dependent spectral conductivity, a mode-resolved spectral energy density, a spectral measure of an effective-medium operator, or a thermally weighted moment of optical conductivity. Together, these works suggest a common pattern: specify or infer the relevant spectral object, couple it to a forward transport map such as a Kubo formula, a Sommerfeld–Bethe integral, a Stieltjes transform, or a mode sum, and then recover conductivity, Hall response, diffusivity, susceptibility, or thermoelectric coefficients from that representation (Narikiyo, 2017, Qiu et al., 2011, Hirosawa et al., 2022, Murphy et al., 2024).

1. Conceptual scope and common architecture

The broadest shared feature of these models is that conductivity is represented indirectly. Instead of beginning with a single relaxation time or a single Drude parameter, the model begins with a resolved transport descriptor and only afterward performs the contraction to an observable coefficient. This suggests a common architecture: a spectral object, a transport kernel, and a projection to the measurable response.

Spectral object	Forward map	Representative sources
$\rho_{\mathbf p}(\varepsilon)$, $G^{R/A}(\mathbf p,\varepsilon)$	Brillouin-zone Kubo sums for $\sigma_{xx}$, $\sigma_{xy}$	(Narikiyo, 2017)
$\sigma(E)$	Sommerfeld–Bethe integrals for $L_{11}$, $L_{12}$, $L_{22}$	(Hirosawa et al., 2022)
$\rho(\omega)$ from Euclidean correlators	$G(\tau)=\int d\omega\,K(\tau,\omega)\rho(\omega)$, then Kubo limit	(Andratschke et al., 19 Mar 2026)
$G^{R/A}(\mathbf p,\varepsilon)$0 and $G^{R/A}(\mathbf p,\varepsilon)$1	$G^{R/A}(\mathbf p,\varepsilon)$2	(Qiu et al., 2011)
Spectral measure $G^{R/A}(\mathbf p,\varepsilon)$3	Stieltjes representation of $G^{R/A}(\mathbf p,\varepsilon)$4	(Murphy et al., 2024)
Optical-current moment ratio $G^{R/A}(\mathbf p,\varepsilon)$5	Weighted spectral moments or midpoint-curvature ratio	(Chowdhury, 10 Dec 2025)

Within this architecture, the adjective “virtual” does not denote a single ontology. In some papers it refers to explicit virtual carriers, as in ballistic graphene near the Dirac point; in others it refers to a reduced or surrogate conductivity representation inferred from transport or spectroscopic data; in still others it describes an effective spectral object that stands in for a full microscopic transport solution. This suggests that the phrase is methodological rather than taxonomic.

A second shared feature is that transport is often governed by information away from the narrow Fermi-surface shell. Narikiyo’s PCCO analysis emphasizes full-Brillouin-zone integration in an incoherent metal (Narikiyo, 2017). The thermoelectric reconstruction literature treats $G^{R/A}(\mathbf p,\varepsilon)$6 as a finite-energy transport spectrum convolved by the derivative of the Fermi function (Hirosawa et al., 2022). The Euclidean spectral-reconstruction literature, by contrast, emphasizes that the low-frequency behavior relevant for conductivity is strongly smeared by the thermal kernel and therefore difficult to recover pointwise (Andratschke et al., 19 Mar 2026). The effective-medium spectral theory of polycrystals replaces direct field solving by a spectral measure of a self-adjoint operator, again shifting emphasis from a single coefficient to a latent spectral representation (Murphy et al., 2024).

2. Electronic spectral-function and spectral-conductivity formulations

A canonical electronic realization appears in Narikiyo’s model for the Hall conductivity of the normal metallic state of PCCO. The starting point is the exact retarded and advanced Green functions,

$G^{R/A}(\mathbf p,\varepsilon)$7

which define the Lorentzian spectral function

$G^{R/A}(\mathbf p,\varepsilon)$8

Using $G^{R/A}(\mathbf p,\varepsilon)$9 data for a degenerate Fermi system and neglecting current vertex correction, the conductivities are written as

$\sigma_{xx}$0

with $\sigma_{xx}$1 and $\sigma_{xx}$2 given by Brillouin-zone sums over the dressed propagators and velocity factors. The Hall kernel

$\sigma_{xx}$3

changes sign across momentum space, so a momentum-selective lifetime

$\sigma_{xx}$4

redistributes weight between positive and negative Hall-contributing regions. In the PCCO parametrization, this yields a non-monotonic $\sigma_{xx}$5, and the paper states that the $\sigma_{xx}$6 result is qualitatively similar to experiment (Narikiyo, 2017).

The thermoelectric literature uses a formally different but structurally similar object, the energy-resolved spectral conductivity

$\sigma_{xx}$7

which enters the linear-response coefficients

$\sigma_{xx}$8

and

$\sigma_{xx}$9

The reconstruction method of “Data-driven reconstruction of spectral conductivity and chemical potential from thermoelectric transport data” treats $\sigma_{xy}$0 as an effective experimental transport spectrum, not as a purely band-structure object. In $\sigma_{xy}$1, the model is fit directly to measured $\sigma_{xy}$2 and $\sigma_{xy}$3, with sample-dependent $\sigma_{xy}$4 and a shared or corrected $\sigma_{xy}$5, and is then used to estimate $\sigma_{xy}$6 and the electronic upper bound on $\sigma_{xy}$7 beyond the Wiedemann–Franz law (Hirosawa et al., 2022).

A third electronic realization appears in spin-1 chiral fermions with disorder. There the spectral conductivity itself develops an impurity-induced asymmetric cusp, and the paper introduces a minimal piecewise model,

$\sigma_{xy}$8

with $\sigma_{xy}$9. The central conclusion is that thermoelectric enhancement is strongest when the cusp is sharp, strongly asymmetric, and small at the cusp energy. Increasing the curvature of the trivial band strengthens this enhancement even when the density of states becomes smoother, which isolates spectral conductivity rather than density of states as the controlling quantity (Kikuchi et al., 13 May 2026).

3. Inverse reconstruction, spectral moments, and first-principles implementation

In many applications the spectral object is not known directly and must be reconstructed. For Euclidean lattice data, the forward relation is the Fredholm equation of the first kind

$\sigma(E)$0

with thermal kernel

$\sigma(E)$1

The conductivity is then extracted from the low-frequency limit of the vector spectral function,

$\sigma(E)$2

“Spectral reconstruction techniques, their shortcomings and relevance to the electric conductivity coefficient” compares MEM, Backus–Gilbert, Gaussian-process methods, an unsupervised neural-network method, and a new multipoint estimator for

$\sigma(E)$3

Its methodological conclusion is that low-frequency transport information may be more stable than the detailed global spectrum, but remains intrinsically regularization dependent. On quenched lattice data in nonzero magnetic field, the longitudinal conductivity $\sigma(E)$4 increases with field strength, while detailed values remain method dependent (Andratschke et al., 19 Mar 2026).

A complementary development is the use of model-independent moment constraints on optical conductivity. “Planckian Bounds via Spectral Moments of Optical Conductivity” defines

$\sigma(E)$5

where

$\sigma(E)$6

and proves

$\sigma(E)$7

with $\sigma(E)$8 as a conservative estimate and $\sigma(E)$9 as a tighter value. The same quantity is accessible from imaginary-time data through

$L_{11}$0

This replaces model-dependent width fitting by a bound on a thermally weighted spectral moment, and therefore functions as a constraint on admissible virtual spectral conductivity models rather than as a particular spectral ansatz (Chowdhury, 10 Dec 2025).

At the first-principles level, “Kubo-Greenwood Electrical Conductivity Formulation and Implementation for Projector Augmented Wave Datasets” provides the implementation backbone for a frequency-dependent conductivity tensor,

$L_{11}$1

built from band energies, occupations, and gradient matrix elements. The paper distinguishes the original KG formula from the popular approximation in which $L_{11}$2 is replaced by $L_{11}$3 on the support of the delta function, and shows that only the exact Dirac-delta form with Lorentzian broadening reproduces the exact low-frequency behavior for finite width. It also gives an explicit decomposition into intra-band, degenerate-state, and non-degenerate inter-band contributions, and a PAW formulation in which the all-electron gradient matrix element is the sum of a plane-wave pseudo contribution and an augmentation correction. In that sense, the KG formalism turns a discrete transition spectrum into a virtual conductivity spectrum by controlled broadening and tensorial summation (Calderin et al., 2017).

4. Microscopic realizations beyond standard electron-band transport

A mode-resolved phononic realization is provided by equilibrium molecular dynamics plus spectral energy density analysis in graphene. The normal-mode amplitude is modeled as

$L_{11}$4

which yields a Lorentzian SED peak and the linewidth–lifetime relation

$L_{11}$5

Thermal conductivity is then reconstructed as

$L_{11}$6

This produces a branch-resolved spectral conductivity representation of suspended and supported graphene. The paper reports that the ZA branch contributes around $L_{11}$7 of the thermal conductivity in suspended single-layer graphene at room temperature, while support on amorphous $L_{11}$8 reduces that contribution to around $L_{11}$9. At $L_{12}$0 K before quantum correction, the SED value $L_{12}$1 is in close agreement with the Green–Kubo value $L_{12}$2 (Qiu et al., 2011).

A compact many-body realization appears in the Dynamic Hubbard dimer. There the essential spectral ingredients are the overlap

$L_{12}$3

and the local pseudo-spin excitation energy

$L_{12}$4

The one-particle spectral function for $L_{12}$5 occupancy changes splits into a coherent and an incoherent part,

$L_{12}$6

while the optical conductivity develops sidebands at $L_{12}$7 and, in the three-electron sector, at $L_{12}$8. The paper’s main conclusion is that occupancy-dependent local dressing shifts optical and spectral weight from low-energy coherent motion to high-energy incoherent channels, producing explicit electron-hole asymmetry and the result that holes are less mobile than electrons (Bach et al., 2012).

The phrase “virtual” becomes literal in the weak-value model of graphene. There, virtual particle–hole excitations allowed by uncertainty relations are converted into real carriers by an electric field, with the transition kinematics governed by a weak value of the group velocity. The model estimates a quasi-Ohmic minimal conductivity

$L_{12}$9

per channel and predicts a crossover at

$L_{22}$0

to a Schwinger-like regime in which $L_{22}$1. In this case the “virtual spectral conductivity model” terminology is not merely metaphorical: the current is assembled from momentum-resolved virtual transition channels (Yokota et al., 2014).

5. Effective media, holography, and surrogate deployment

The effective-medium version is particularly explicit in the discrete spectral theory of uniaxial polycrystals. There the local conductivity is decomposed as

$L_{22}$2

and the effective conductivity tensor admits the Stieltjes representation

$L_{22}$3

The discrete spectral measure is built from the eigenvalues and orthonormal eigenvectors of the real-symmetric matrix $L_{22}$4. A projection method reduces the problem to smaller matrices, improving numerical efficiency by a factor of $L_{22}$5. This is one of the clearest examples of a virtual spectral conductivity model in the strict sense: the spectral measure functions as a latent transport descriptor, and the Stieltjes transform decodes it into $L_{22}$6 and $L_{22}$7 (Murphy et al., 2024).

Holographic models provide another route from a latent spectral problem to conductivity. In “A Simple Holographic Model of a Charged Lattice,” a zero-density striped scalar background with source $L_{22}$8 breaks translation invariance along $L_{22}$9. Fourier-mode mixing then yields anisotropic optical conductivity: the transverse channel remains non-Drude-like, while the longitudinal channel develops a Drude-like peak together with a delta function of negative weight, which the paper interprets as spectral weight transfer. The low-energy mechanism is direct coupling between the homogeneous current mode and the finite-momentum scalar-phase fluctuation (Aprile et al., 2014).

In V-QCD with matter, the same bulk fluctuation problem generates spectral functions, quasi-normal modes, diffusion, susceptibility, and conductivity. The flavor conductivity follows from the horizon formula

$\rho(\omega)$0

while diffusion satisfies

$\rho(\omega)$1

The vector and axial flavor conductivities rise sharply in the transition region, and vector and axial channels merge in the chirally restored phase. This is a holographic realization in which the virtual spectral object is a bulk gauge-field fluctuation with infalling boundary conditions, and conductivity is its low-frequency boundary flux (Iatrakis et al., 2014).

A deployment-oriented surrogate realization appears in the self-driving-lab literature on doped conjugated polymers. There optical spectra are converted into conductivity predictors by GA-selected area-under-the-curve features, second-derivative features, processing metadata, and SHAP-guided feature selection. The final hybrid QSPR model, combining expert and data-driven features, reports test $\rho(\omega)$2, RMSE $\rho(\omega)$3, and MAE $\rho(\omega)$4, while reducing direct conductivity measurements by about $\rho(\omega)$5. Here the “virtual” model is a compact spectroscopic surrogate for electrical conductivity, rather than a first-principles transport solver (Mishra et al., 6 Sep 2025).

6. Limitations, controversies, and boundary cases

The literature does not present a single accepted formalism under the name “Virtual Spectral Conductivity Model.” The phrase is best treated as an editorial umbrella. This is important because the underlying models make very different assumptions about what the spectral object is and how it is obtained.

The most explicit conceptual controversy concerns the role of current vertex corrections in incoherent electronic transport. Narikiyo’s PCCO work neglects CVC throughout and argues that the Hall anomaly is explained without CVC if the spectral function is correct, while criticizing both standard Fermi-liquid Hall theory and FLEX on spectral-function grounds. The same paper also states explicitly that deriving the model spectral function from a microscopic Hamiltonian is “another task” not addressed there. The resulting framework is therefore intentionally phenomenological even while making a sharp methodological claim (Narikiyo, 2017).

Inverse formulations face an independent difficulty: the relevant spectra are often only weakly identifiable. The Euclidean reconstruction literature emphasizes that the transport observable lives precisely in the low-frequency regime where the kernel most strongly smears information, that different methods can fit the same correlator while giving different $\rho(\omega)$6, and that midpoint, multipoint, MEM, Gaussian-process, and neural approaches all introduce different regularizations, priors, or analyticity assumptions (Andratschke et al., 19 Mar 2026). The thermoelectric reconstruction literature makes the same point in a different language: the matrices generated by the discretized Sommerfeld–Bethe kernels are nearly singular, single-sample data do not uniquely determine both $\rho(\omega)$7 and $\rho(\omega)$8, and moving-average smoothing, positivity penalties, bounded chemical-potential parametrizations, and multi-sample constraints are all parts of the inverse model, not optional decorations (Hirosawa et al., 2022).

Even when the underlying theory is explicit, numerical representation choices matter. In the Kubo–Greenwood setting, the paper argues that the commonly used approximated $\rho(\omega)$9 Dirac-delta form is not equivalent to the exact KG expression at finite broadening and has pathological low-frequency behavior, whereas Lorentzian broadening of the exact Dirac-delta form reproduces the original KG formula and its dc limit (Calderin et al., 2017). This makes broadening part of the physical model.

There are also boundary cases that are relevant but not spectral in the strict energy- or frequency-resolved sense. The SRFS-MS model for rectangular interconnects is described explicitly as “not a spectral conductivity model in the strict energy/frequency sense,” but as a physically grounded virtual conductivity model that produces a spatial field $G(\tau)=\int d\omega\,K(\tau,\omega)\rho(\omega)$0 from surface scattering, grain-boundary scattering, geometry, and temperature. Its compact surrogate, SRFS-MS-C3, reproduces the full model with an average error of less than $G(\tau)=\int d\omega\,K(\tau,\omega)\rho(\omega)$1. This marks the outer edge of the concept: virtual conductivity models can be spectral, spatial, or hybrid, but those categories should not be conflated (Chen et al., 17 May 2025).

Taken together, these limitations suggest a restrained conclusion. A virtual spectral conductivity model is most useful when the chosen spectral object is closely aligned with the observable of interest, when the forward transport map is explicit, and when regularization or phenomenology is treated as part of the model definition rather than hidden beneath a nominally microscopic label. In that restricted but productive sense, the concept spans incoherent Hall transport, thermoelectric inversion, phonon mode summation, operator-valued effective media, holographic conductivity, and spectroscopic surrogates without requiring a single canonical formalism.