Gradient-Expansion Formalism

• Gradient-expansion formalism is a systematic approximation method that organizes observables by powers of spatial or temporal derivatives to capture local and nonlocal effects.
• It uses a separation of scales by identifying a small parameter, allowing controlled inclusion of higher order corrections in systems like inhomogeneous electron gases and cosmological perturbations.
• Practical implementations include refining exchange-correlation functionals in density functional theory, modeling superhorizon perturbations in cosmology, and improving transport equations in nonequilibrium field theory.

Searching arXiv for the cited paper and closely related gradient-expansion literature. Gradient-expansion formalism denotes a class of systematic approximation schemes in which the dynamics are organized by powers of spatial or spacetime derivatives of slowly varying fields, densities, or correlators. In the inhomogeneous electron gas it appears as the gradient expansion approximation derived from linear response, with the $q^2$ term in the proper-polarization function controlling the leading gradient correction to the exchange-correlation functional (Benites et al., 30 Jan 2026). In cosmology it is the long-wavelength expansion in $\epsilon\sim (HL)^{-1}\sim k/(aH)$ for superhorizon perturbations (Takamizu, 2018). In nonequilibrium field theory it is the expansion of Wigner-space equations in derivatives with respect to the macroscopic coordinate $X$ (Millington et al., 2013). Closely related constructions underlie second-order relativistic hydrodynamics, quantum spin-torque calculations, inflationary magnetogenesis, and gauge-field production during axion inflation (Lahiri, 2019, Shitade, 2017, Sobol et al., 2020, Gorbar et al., 2021). The common aim is to reduce a fully inhomogeneous problem to an ordered hierarchy in which leading behavior is local or homogeneous and higher-order terms encode controlled nonlocal corrections.

1. Scope and organizing principle

The formalism is defined by a separation of scales. One identifies a small parameter that suppresses gradients relative to the dominant microscopic or background scale, and then expands observables, constitutive relations, or evolution equations in powers of that parameter. The specific parameter depends on the application.

Context	Ordering variable	Representative formulation
Inhomogeneous electron gas	small momentum $q$	$\Pi^*(q,0)\simeq \Pi^*(0,0)+b_{xc}q^2+O(q^4)$ (Benites et al., 30 Jan 2026)
Superhorizon cosmology	$\epsilon\equiv(HL)^{-1}\sim k/(aH)$	$\partial_i\sim O(\epsilon)$, $\partial_t\sim O(\epsilon^0)$ (Naruko et al., 2012)
Nonequilibrium QFT	derivatives $\partial_X$ in Wigner space	expansion of the Moyal product (Millington et al., 2013)
Inflationary gauge-field production	$\varepsilon\sim |\nabla|/k_{\rm UV}\ll1$ or curl number $\epsilon\sim (HL)^{-1}\sim k/(aH)$0	hierarchy of bilinear moments (Eckardstein, 14 Oct 2025)

At leading order, the dynamics are typically homogeneous, local, or algebraic. In superhorizon cosmology, the $\epsilon\sim (HL)^{-1}\sim k/(aH)$1 system reproduces the separate-universe picture and the $\epsilon\sim (HL)^{-1}\sim k/(aH)$2 formalism (Naruko et al., 2012). In nonequilibrium kinetic theory, the lowest-order Wigner-space equations yield the usual drift and collision terms of transport theory (Garny et al., 2010). In the inhomogeneous electron gas, the small-$\epsilon\sim (HL)^{-1}\sim k/(aH)$3 limit of $\epsilon\sim (HL)^{-1}\sim k/(aH)$4 determines the slowly varying density limit of the exchange-correlation functional (Benites et al., 30 Jan 2026).

Higher orders then encode curvature, nonlocality, dissipation, screening, memory, or mode-coupling effects. The formalism is therefore not a single universal algorithm; it is a general method of derivative ordering whose concrete realization depends on the variables used to parametrize the problem.

2. Mathematical realizations

Several mathematical architectures recur across the literature. In linear-response density functional theory, one expands the static proper polarization and relates its coefficients to gradient corrections in the energy functional. In ADM-based cosmology, one decomposes the metric into lapse, shift, spatial metric, and extrinsic curvature, then counts spatial derivatives as higher order in $\epsilon\sim (HL)^{-1}\sim k/(aH)$5 (Takamizu, 2018). In Wigner-space quantum field theory, one introduces relative and center coordinates and expands the noncommutative star product in derivatives of the macroscopic coordinate (Canevarolo et al., 2022). In Keldysh-based spintronics, the same Wigner machinery is promoted to a gauge-covariant form with Wilson lines and field strengths (Shitade, 2017).

A canonical Wigner transform is

$\epsilon\sim (HL)^{-1}\sim k/(aH)$6

after which convolutions become Moyal products. The resulting derivative expansion organizes corrections to transport, spectral, and effective-action equations (Millington et al., 2013). In gauge-field production during inflation, the Fourier-mode description is traded for bilinear moments such as

$\epsilon\sim (HL)^{-1}\sim k/(aH)$7

which are vacuum expectation values of electric and magnetic fields with an arbitrary number of spatial curls. These obey an infinite coupled chain of ordinary differential equations, closed by a truncation prescription and supplemented by boundary terms that account for newly amplified modes (Sobol et al., 2020).

The formalism therefore alternates between two complementary operations: derivative counting and closure. Derivative counting determines which terms are retained at a given order. Closure determines how an infinite hierarchy, nonlocal kernel, or regulator-dependent intermediate quantity is rendered calculable.

3. Inhomogeneous electron gas and the gradient expansion approximation

In the inhomogeneous electron gas, the formalism begins from

$\epsilon\sim (HL)^{-1}\sim k/(aH)$8

with

$\epsilon\sim (HL)^{-1}\sim k/(aH)$9

For a homogeneous reference state $X$0, linearization to second order in $X$1 gives

$X$2

and

$X$3

The slowly varying limit is controlled by

$X$4

where $X$5 determines the leading gradient correction and, equivalently, the $X$6 term in the gradient expansion approximation with $X$7 (Benites et al., 30 Jan 2026).

Up to $X$8, $X$9 is built by summing exchange diagrams and the five correlation ring diagrams in RPA. The decomposition

$q$0

suggests separate exchange and correlation coefficients. The central result of "Gradient-expansion of the inhomogeneous electron-gas revisited" is that this separation is not meaningful once the long-range Coulomb singularity is handled correctly (Benites et al., 30 Jan 2026).

The regularized interaction is introduced as

$q$1

or more generally as the family

$q$2

With $q$3 taken only after the integrals are performed, the exchange contribution becomes

$q$4

while the Ma-Brueckner correlation piece is

$q$5

Two further correlation pieces are independent of $q$6, but the full $q$7 remains $q$8-dependent. Numerically different authors had effectively used different $q$9, or no regulator, and thereby obtained conflicting values for $\Pi^*(q,0)\simeq \Pi^*(0,0)+b_{xc}q^2+O(q^4)$0 and $\Pi^*(q,0)\simeq \Pi^*(0,0)+b_{xc}q^2+O(q^4)$1 (Benites et al., 30 Jan 2026).

By contrast, the sum is unique:

$\Pi^*(q,0)\simeq \Pi^*(0,0)+b_{xc}q^2+O(q^4)$2

independent of $\Pi^*(q,0)\simeq \Pi^*(0,0)+b_{xc}q^2+O(q^4)$3. In atomic units this gives

$\Pi^*(q,0)\simeq \Pi^*(0,0)+b_{xc}q^2+O(q^4)$4

The immediate consequence is methodological rather than merely numerical. Most popular GGA functionals, including PBE and PBEsol, have been built by imposing separate small-$\Pi^*(q,0)\simeq \Pi^*(0,0)+b_{xc}q^2+O(q^4)$5, small-$\Pi^*(q,0)\simeq \Pi^*(0,0)+b_{xc}q^2+O(q^4)$6 constraints on $\Pi^*(q,0)\simeq \Pi^*(0,0)+b_{xc}q^2+O(q^4)$7 and $\Pi^*(q,0)\simeq \Pi^*(0,0)+b_{xc}q^2+O(q^4)$8. The revisited calculation concludes that such separate constraints are fundamentally invalid, because only the sum $\Pi^*(q,0)\simeq \Pi^*(0,0)+b_{xc}q^2+O(q^4)$9 is regulator-scheme independent (Benites et al., 30 Jan 2026). This directly addresses a long-standing misconception: the issue is not which separate exchange or correlation coefficient is correct, but that the separation itself is ill-defined.

4. Long-wavelength cosmological perturbations

In cosmology, the gradient expansion is the long-wavelength expansion for superhorizon perturbations. The bookkeeping parameter is typically

$\epsilon\equiv(HL)^{-1}\sim k/(aH)$0

with spatial derivatives counted as $\epsilon\equiv(HL)^{-1}\sim k/(aH)$1 and time derivatives as $\epsilon\equiv(HL)^{-1}\sim k/(aH)$2. In ADM form one writes

$\epsilon\equiv(HL)^{-1}\sim k/(aH)$3

decomposes the extrinsic curvature into trace and traceless parts, and expands all quantities order by order in $\epsilon\equiv(HL)^{-1}\sim k/(aH)$4 (Takamizu, 2018). At $\epsilon\equiv(HL)^{-1}\sim k/(aH)$5 one recovers a local FLRW universe in each patch, which is the nonlinear content of the $\epsilon\equiv(HL)^{-1}\sim k/(aH)$6 formalism. At $\epsilon\equiv(HL)^{-1}\sim k/(aH)$7 spatial curvature, shear, and matter inhomogeneities backreact on the large-scale evolution (Naruko et al., 2012).

This framework was extended beyond single-field slow roll in several directions. For a multi-component scalar field with a general kinetic term and general potential, a fully nonlinear gauge transformation was constructed through $\epsilon\equiv(HL)^{-1}\sim k/(aH)$8, allowing one to move from gauges in which the equations are simple to gauges in which the curvature perturbation is naturally defined (Naruko et al., 2012). For shift-symmetric G-inflation, a general solution was obtained up to second order in synchronous gauge, and a curvature perturbation conserved up to first order was defined (Frusciante et al., 2013). For general scalar-tensor theories, the formalism yields nonlinear evolution equations for both scalar and tensor superhorizon modes and extends beyond the constant mode captured by $\epsilon\equiv(HL)^{-1}\sim k/(aH)$9 (Takamizu, 2018).

The newer $\partial_i\sim O(\epsilon)$0 formulation with gradient interactions makes the departure from the separate-universe approximation explicit. Defining $\partial_i\sim O(\epsilon)$1, one expands the comoving curvature perturbation recursively as

$\partial_i\sim O(\epsilon)$2

with

$\partial_i\sim O(\epsilon)$3

To reproduce this structure in a $\partial_i\sim O(\epsilon)$4 language, one replaces the background Klein-Gordon equation by

$\partial_i\sim O(\epsilon)$5

The source $\partial_i\sim O(\epsilon)$6 restores the missing $\partial_i\sim O(\epsilon)$7 effects and captures the non-conservation of the comoving curvature perturbation in cases such as transitions to ultra-slow roll. In that setting the gradient-corrected formalism reproduces the higher-order super-Hubble evolution and modifies equilateral non-Gaussianity $\partial_i\sim O(\epsilon)$8 in ways missed by standard $\partial_i\sim O(\epsilon)$9 (Ahmadi et al., 31 Jan 2026).

5. Nonequilibrium field theory, transport, and effective action

In nonequilibrium thermal field theory, the gradient expansion arises after rewriting two-point functions in mixed coordinates. For the contour-ordered propagator one introduces relative and central coordinates,

$\partial_t\sim O(\epsilon^0)$0

and expands the Wigner-space Dyson-Schwinger equations in powers of derivatives with respect to $\partial_t\sim O(\epsilon^0)$1. Millington and Pilaftsis carried this construction to all orders in the gradient expansion by working with non-homogeneous free propagators and time-dependent vertices. The finite-time energy kernel

$\partial_t\sim O(\epsilon^0)$2

regulates the would-be pinch singularities at finite time, while the full evolution exhibits memory effects, non-Markovian behavior, and early-time energy-violating processes (Millington et al., 2013).

The first-order gradient correction can also alter kinetic source terms. In leptogenesis, a closed-time-path derivation shows that an additional CP-violating source term arises from the gradient expansion and is non-zero even when all species are in local thermal equilibrium. In a standard cosmological background it is proportional to the expansion rate and vanishes in static equilibrium, in accordance with the Sakharov conditions. Numerically it is small for standard thermal leptogenesis, but it can become the dominant source in the limit of ultra-strong washout (Garny et al., 2010).

A related but distinct application is the derivative expansion of the quantum effective action. Using midpoint coordinates and a Wigner-space expansion consistent with the propagator equations of motion and the transposition symmetry, the effective action was derived up to second order in gradients and up to two-loop order for interacting scalar field theory. At one loop, the second-order gradient correction vanishes in the single-field case but can be significant in the multi-field case. At two loops, the 1PI expansion becomes nonrenormalizable once spacetime-dependent resummed masses are used, and renormalizability is restored by adding 2PI counterterms. This identifies the 2PI formalism as the appropriate renormalization framework for that resummed gradient expansion (Canevarolo et al., 2022).

6. Hydrodynamics, spin torques, and inflationary gauge fields

In relativistic hydrodynamics, the gradient expansion is the organizing principle behind constitutive theory. In the Eckart frame one writes

$\partial_t\sim O(\epsilon^0)$3

and then classifies all allowed scalar, vector, and tensor structures by derivative order. At second order, one may build 10 independent scalars for $\partial_t\sim O(\epsilon^0)$4, 11 independent vectors for $\partial_t\sim O(\epsilon^0)$5, and 15 independent tensors for $\partial_t\sim O(\epsilon^0)$6, including explicit curvature couplings such as $\partial_t\sim O(\epsilon^0)$7, $\partial_t\sim O(\epsilon^0)$8, and $\partial_t\sim O(\epsilon^0)$9. Recasting the algebraic constitutive relations into Müller-Israel-Stewart-type relaxation equations restores causal, stable hydrodynamics (Lahiri, 2019). A complementary structural result is that, in a non-conformal fluid, the ordering of transverse derivatives is irrelevant in the gradient expansion, while in a conformal fluid the longitudinal projection of the Weyl-covariant derivative can be eliminated for hydrodynamic fields, though not for curvature tensors (Diles, 2020).

In condensed-matter transport, a gauge-covariant Keldysh-Wigner gradient expansion provides a quantum-mechanical formalism for generic spin torques. Applied to a three-dimensional ferromagnetic metal with nonmagnetic and magnetic impurities in the self-consistent Born approximation, it generates spin renormalization and Gilbert damping at first order in gradients, and spin-transfer torque together with the $\partial_X$0-term at mixed second order in gradients of magnetization and electromagnetic fields. The method avoids both the small-amplitude assumption and the need for a local SU(2) gauge rotation (Shitade, 2017).

Inflationary gauge-field production has produced another extensive branch of the formalism. In magnetogenesis and axion inflation, one introduces homogeneous bilinear correlators of the electric and magnetic fields with arbitrary numbers of curls,

$\partial_X$1

which satisfy an infinite chain of coupled ordinary differential equations. Boundary terms account for the continual addition of newly superhorizon or tachyonically amplified modes, and the hierarchy is truncated using UV-dominated asymptotics. For kinetic-coupling magnetogenesis, truncation at $\partial_X$2--12 already yields electric and magnetic energy densities with a few-percent accuracy during inflation (Sobol et al., 2020). For axion inflation, the same strategy allows one to include both gauge-field backreaction and Schwinger conductivity self-consistently (Gorbar et al., 2021). In the homogeneous-gradient limit of pure axion inflation, a comprehensive scan found that parameter points yielding detectable gravitational waves also violate the upper bound on $\partial_X$3 (Eckardstein et al., 1 Aug 2025). When fermionic Schwinger damping is included, the gauge-field production is attenuated, a fermion-tempered backreaction regime appears, and the gravitational-wave signal can fall into the sensitivity reach of LISA and ET without violating the upper limit on $\partial_X$4 (Eckardstein et al., 29 Sep 2025). The formalism has also been extended to massive vector production during inflation, including longitudinal polarization through additional bilinear hierarchies (Lysenko et al., 29 Sep 2025). The GEFF Python package systematizes these inflationary implementations by providing ready-to-use models for pure and fermionic axion inflation, an integrated error estimator, a self-correction algorithm, and tools for sourced primordial gravitational waves (Eckardstein, 14 Oct 2025).

Across these domains, the gradient-expansion formalism functions as a controlled derivative ordering rather than as a single specialized technique. Its technical content is set by the choice of variables, the definition of the small parameter, and the consistency of the closure or regularization procedure. The revisited inhomogeneous electron-gas calculation shows that even apparently natural decompositions can be scheme-dependent and physically ill-defined (Benites et al., 30 Jan 2026). The broader literature reaches the same methodological lesson in different language: the usefulness of a gradient expansion depends not only on truncation order, but also on identifying which composite quantities remain invariant, causal, renormalizable, or regulator independent under the approximation.