Fluctuation–Dissipation Theorem Overview

• FDT is a principle linking spontaneous microscopic fluctuations to a system’s linear response, defining how equilibrium systems relax after disturbances.
• It has been generalized for nonequilibrium states by incorporating flux, memory effects, and non-Gaussian noise to capture persistent currents.
• FDT underpins various applications from modeling Brownian motion and quantum dissipation to informing thermodynamic resource theories.

The Fluctuation–Dissipation Theorem (FDT) establishes a fundamental connection between spontaneous fluctuations in physical observables and the linear response of those observables to external perturbations in both classical and quantum systems. In its simplest form, the FDT implies that the extent of equilibrium fluctuations determines how a system relaxes back to equilibrium after a small disturbance, and vice versa. While the classical FDT is exact for systems in thermal equilibrium obeying detailed balance, a series of theoretical, computational, and experimental investigations have extended and generalized the FDT to nonequilibrium steady states (NESS), non-Gaussian and correlated noise, quantum systems, growth phenomena, and high-dimensional learning models.

1. Classical and Quantum Formulations of the FDT

In equilibrium statistical mechanics, the FDT relates the correlation function of an observable to its linear response to an applied field. The classical Langevin equation for a Brownian particle of mass $m$ in a fluid, subjected to friction $\gamma$ and random force $f(t)$,

$m\,\dot v(t) = -m\gamma\,v(t) + f(t),$

yields the FDT in the form

$$\langle f(t) f(t') \rangle = 2 m \gamma k_B T \delta(t-t')$$

and establishes the Einstein relation $D = k_B T /(m\gamma)$ between the diffusion constant and friction coefficient (Gomes-Filho et al., 2023, Gomes-Filho et al., 2021). At the linear level, the response function $R_A(t - t')$, quantifying the change of observable $A$ to a time-localized perturbing field $h(t')$, is linked to equilibrium correlation as

$$R_{A}(t-t') = \frac{1}{k_B T} \frac{d}{dt'} C_{AA}(t-t')$$

where $C_{AA}(t-t') = \langle A(t) A(t') \rangle$ (Colombani et al., 2017, Feng et al., 2011).

For quantum systems, the FDT relates symmetrized correlation functions and commutators to the linear response, as expressed in the Kubo–Callen–Welton relation: $$S_{AB}^{\mathrm{sym}}(\mathbf{k},\omega) = \hbar \chi_{AB}''(\mathbf{k},\omega) \coth \left( \frac{\hbar\omega}{2T} \right)$$ where $S_{AB}^{\mathrm{sym}}$ is the symmetrized correlation and $\chi_{AB}''$ is the imaginary (dissipative) part of the response function (Kirkpatrick et al., 2023, Kubo et al., 2018).

2. Generalized FDT in Nonequilibrium Steady States

In systems maintained far from equilibrium, detailed balance is broken and steady-state currents persist. Jiangfeng, Qian, and Wang (Feng et al., 2011) decomposed the force into a gradient component (potential landscape) and a divergence-free curl (steady current): $F(x) = -D \nabla \phi(x) + F_{\mathrm{curl}}(x),$ with probability current $J(x) = F(x) P_{ss}(x) - D \nabla P_{ss}(x)$ and steady-state distribution $P_{ss}(x)$. The generalized linear response function for observables $A(x)$, $B(x)$ reads

$$R_{AB}(t) = \langle A(t) \nabla B(0) \rangle + \langle A(t) F_{\mathrm{curl}}(x(0)) B(0) \rangle.$$

The first term is the spontaneous relaxation familiar from equilibrium, while the second term captures the correlation with the non-gradient curl current, reflecting persistent irreversibility and gauge curvature in the system. This flux term is physically associated with housekeeping entropy production required to maintain NESS (Feng et al., 2011).

Experiments with spatially extended mechanical systems driven by heat flux (e.g., microcantilevers under intense local heating) directly confirm this picture: the amplitude of fluctuations is not set by the spatially averaged temperature, but by a weighted average where the local dissipation serves as the weight (Fontana et al., 2021, Fontana et al., 2022).

3. Extensions: Memory, Noise Structure, and Hidden FDTs

Classical FDT assumes Markovian, white-noise forcing and immediate return to equilibrium. Extended scenarios include:

• Non-Markovian GLE: The generalized Langevin equation (GLE) introduces a memory kernel $K(t)$ and fluctuation force $f(t)$:

$$\frac{d}{dt} u(t) = - \int_0^t K(t-s) u(s) ds + \frac{1}{m} f(t) + \frac{1}{m} g(t).$$

In equilibrium, the second FDT relates the memory and noise correlations as $K(t) = \langle f(0) f(t) \rangle / (k_B T)$ (Zhu et al., 2021).

• Non-Gaussian / Lévy Noise: For systems subject to $\alpha$-stable (Lévy) noises, a generalized FDT holds if one uses as the conjugate observable $X = -\partial_\lambda \log \rho_{ss}(x; \lambda)$, with the steady-state $\rho_{ss}$ depending on the control parameter $\lambda$ (Dybiec et al., 2012, Gomes-Filho et al., 2023).
• Surface Growth (Hidden FDT): In 1+1D models (Edwards–Wilkinson, KPZ), the saturated width $w_s$ relates the noise $D$ and dissipation $\nu$ via $w_s^2 = cD/\nu\cdot L$, yielding a "hidden" FDT:

$$\langle \xi(x,t)\xi(x',t') \rangle = 2 b \nu w_s^2 \delta(x-x')\delta(t-t')$$

(Gomes-Filho et al., 2021, Gomes-Filho et al., 2023).

The FDT may break down in systems lacking mixing/ergodicity, exhibiting long memory, or anomalous diffusion (ballistic or glassy dynamics) (Gomes-Filho et al., 2023).

4. Quantum Generalizations, Nonequilibrium Quantum FDT, and Violations

The equilibrium quantum FDT rests on the Kubo–Martin–Schwinger (KMS) relation between correlation functions. Under the eigenstate thermalization hypothesis, the KMS relation (and thus FDT) holds in the thermodynamic limit for isolated quantum systems, with finite-size corrections scaling as $L^{-d}$ (Noh et al., 2020).

Nonequilibrium quantum generalizations introduce extra terms capturing steady-state currents (curl flux) that vanish at equilibrium. The response function includes a flux-driven contribution, and quantum master-equation formalism reveals that the violation of detailed balance (quantum curl) governs the nonequilibrium correction (Zhang et al., 2021, Mehboudi et al., 2017). For quantum fluids under steady gradients, Kirkpatrick and Belitz (Kirkpatrick et al., 2023) demonstrated that the commutator spectrum is related not to linear response, but to bilinear response functions of pairs of observables to the square of the driving field—these encode long-range correlations and generalized rigidity.

Genuine quantum violations of the equilibrium FDT can occur in the antisymmetric (Hall-like) response of macroscopic quantum systems, dramatically deviating from classical predictions under low temperatures and high field, as established for the 2D Hall conductivity (Kubo et al., 2018). The violation is controlled by the spectral centroid exceeding $k_B T /\hbar$.

5. Practical Computation and Experimental Probes

The data-driven application of FDT in high-dimensional, non-normal turbulent systems (e.g., Rayleigh–Bénard convection at high Rayleigh number) is enabled by projecting the FDT onto a physically aligned low-dimensional basis, such as Dynamic Mode Decomposition (DMD) modes. This procedure robustly yields reduced-order linear response models capable of accurate mean-state predictions in fully turbulent regimes, outperforming standard Proper Orthogonal Decomposition (POD)-based approaches and matching expensive Green's function calculations (Khodkar et al., 2018).

In glassy or ageing soft-matter systems, simultaneous measurement of mobility and diffusivity in the same sample avoids ambiguities in sample conditions and allows direct determination of effective temperature via generalized Einstein relations. For certain fast observables, the FDT can remain valid even as the system structurally ages, while slow observables integrated over sluggish degrees of freedom may display "effective temperatures" greatly in excess of the bath (Colombani et al., 2017).

6. FDT and Thermodynamic Resource Theory

Within resource-theoretic quantum thermodynamics, the FDT takes the form of quantitative relations between the dissipated free energy and the statistical variance of the initial state's free energy in optimal distillation protocols. The minimal dissipated work (up to second-order $N^{-1/2}$ corrections) scales with the standard deviation of the initial nonequilibrium free energy, fixing the performance limits of work extraction, information erasure, and thermodynamically-free communication tasks (Biswas et al., 2021). The FDT here plays the role of a second-order Kubo–Green relation connecting operational dissipation to fluctuations in out-of-equilibrium quantum ensembles.

7. Structural Overview: Scope, Validity, and Breakdown

The validity of the classical FDT is confined to systems near equilibrium, with linear response, detailed balance, and ergodicity. Breakdown occurs under strong driving, memory effects, spatial/temporal inhomogeneity, or lack of mixing. Nonequilibrium generalizations introduce extra terms (frenesy, flux, or conjugate variables specific to the steady state) that capture time-antisymmetric, persistent currents and break the simple proportionality between fluctuations and response (Feng et al., 2011, Gomes-Filho et al., 2023). In quantum systems, true macroscopic violations arise in dissipationless parts when relevant frequencies exceed thermal energies. Across physical, chemical, biological, and even information-theoretic systems, the FDT and its generalizations remain a principal tool for connecting stochastic dynamics and macroscopic transport (Gomes-Filho et al., 2023).