7.3 Fluctuation-dissipation theorem

Foundations of the Fluctuation-Dissipation Theorem

The fluctuation-dissipation theorem (FDT) establishes a precise quantitative link between the spontaneous fluctuations a system exhibits at equilibrium and the way that system dissipates energy when driven slightly out of equilibrium. In practical terms, if you can measure the noise a system produces on its own, you can predict how it will respond to a small external push.

This connection is remarkably general. It underlies results ranging from Einstein's relation for Brownian motion to the thermal noise floor in electronic circuits. For non-equilibrium statistical mechanics, the FDT is the central bridge between equilibrium correlation functions and macroscopic transport coefficients.

Equilibrium vs. Non-Equilibrium Systems

Equilibrium systems have macroscopic properties (temperature, pressure, density) that remain constant in time, even though microscopic variables fluctuate continuously. Non-equilibrium systems show time-dependent macroscopic behavior because some driving force or gradient is present.

The FDT sits right at the boundary: it applies to systems that are near equilibrium, subject to perturbations small enough that the response stays proportional to the applied force. The key insight is that the same microscopic interactions responsible for spontaneous equilibrium fluctuations also govern how the system dissipates energy when perturbed.

Linear Response Theory

Linear response theory provides the formal framework behind the FDT. The core assumption is that for a sufficiently weak external perturbation $F(t)$, the average response $\langle A(t) \rangle$ of some observable $A$ is linear in $F$:

$\langle \delta A(t) \rangle = \int_{-\infty}^{t} \chi(t - t') \, F(t') \, dt'$

Here $\chi(t - t')$ is the response function (or generalized susceptibility). It encodes everything about how the system reacts. The FDT then tells you that $\chi$ is not independent of the equilibrium fluctuations; it is directly determined by them.

This framework applies broadly: electrical conductivity relates current response to an applied field, magnetic susceptibility relates magnetization to an applied magnetic field, and so on.

Time Correlation Functions

A time correlation function measures how a fluctuating quantity at one time is statistically related to its value at a later time:

$C_{AA}(t) = \langle \delta A(0) \, \delta A(t) \rangle_{\text{eq}}$

where $\delta A = A - \langle A \rangle$. These functions capture the "memory" of the system. A correlation function that decays quickly means fluctuations are short-lived; a slow decay signals long relaxation times.

The FDT's power comes from relating these equilibrium correlation functions directly to the response function $\chi$. The Green-Kubo formulas (discussed below) are the most common way this relationship is expressed for transport coefficients.

Key Concepts and Principles

Fluctuations in Equilibrium Systems

Even at thermal equilibrium, microscopic variables never sit perfectly still. Particle velocities, local densities, and energy fluctuate constantly due to thermal agitation. These fluctuations are:

• Spontaneous: they require no external driving force
• Statistical: their magnitudes follow well-defined probability distributions (typically Gaussian for large systems, by the central limit theorem)
• Size-dependent: relative fluctuations scale as $1/\sqrt{N}$, so they become more prominent in smaller systems

The variance of fluctuations is tied to thermodynamic susceptibilities. For example, energy fluctuations in the canonical ensemble satisfy $\langle (\delta E)^2 \rangle = k_B T^2 C_V$, directly linking fluctuation amplitude to heat capacity.

Dissipation in Non-Equilibrium Systems

Dissipation refers to the irreversible conversion of directed energy into heat. When you push a system out of equilibrium, friction, viscosity, and electrical resistance all act to degrade the input energy into thermal motion. This increases entropy and drives the system back toward equilibrium.

The rate of energy dissipation depends on the same microscopic interactions (collisions, scattering events) that produce equilibrium fluctuations. That is exactly why the two are connected.

The Core Statement of the FDT

The FDT can be stated in several equivalent ways. One common frequency-domain form is:

$\text{Im}[\chi(\omega)] = \frac{\omega}{2k_BT} \, \tilde{C}_{AA}(\omega)$

where $\text{Im}[\chi(\omega)]$ is the dissipative (imaginary) part of the susceptibility and $\tilde{C}_{AA}(\omega)$ is the Fourier transform of the equilibrium correlation function. The dissipative part of the response is proportional to the spectral density of equilibrium fluctuations, with temperature setting the scale.

This means you can measure equilibrium noise, compute its power spectrum, and from that predict how the system will absorb energy from an external perturbation.

Mathematical Formulation

Green-Kubo Relations

The Green-Kubo relations express macroscopic transport coefficients as time integrals of equilibrium correlation functions. They are a direct consequence of the FDT combined with linear response theory.

Diffusion coefficient (velocity autocorrelation):

$D = \frac{1}{3} \int_0^\infty \langle \mathbf{v}(0) \cdot \mathbf{v}(t) \rangle \, dt$

Shear viscosity (stress autocorrelation):

$\eta = \frac{V}{k_BT} \int_0^\infty \langle \sigma_{xy}(0) \, \sigma_{xy}(t) \rangle \, dt$

Thermal conductivity (heat current autocorrelation):

$\kappa = \frac{V}{k_BT^2} \int_0^\infty \langle \mathbf{J}_Q(0) \cdot \mathbf{J}_Q(t) \rangle \, dt$

The pattern is the same in every case: integrate the equilibrium time-correlation function of the relevant microscopic flux to obtain the corresponding macroscopic transport coefficient. This is powerful because it lets you extract transport properties from equilibrium molecular dynamics simulations without ever applying an external field.

Onsager Reciprocal Relations

When multiple thermodynamic forces drive multiple fluxes simultaneously, the linear transport equations take the form:

$J_i = \sum_j L_{ij} X_j$

where $J_i$ are fluxes, $X_j$ are thermodynamic forces, and $L_{ij}$ are kinetic coefficients. Onsager showed, using microscopic reversibility (time-reversal symmetry of the microscopic equations of motion), that the off-diagonal coefficients are symmetric:

$L_{ij} = L_{ji}$

This has concrete consequences. In thermoelectricity, for instance, the coefficient governing the Seebeck effect (voltage from a temperature gradient) equals the coefficient governing the Peltier effect (heat flow from an applied current). These reciprocal relations reduce the number of independent transport coefficients you need to measure.

Nyquist Theorem

The Nyquist theorem is the FDT applied specifically to electrical circuits. A resistor at temperature $T$ generates a voltage noise with power spectral density:

$S_V(f) = 4k_BTR$

• $S_V(f)$: voltage noise power spectral density ($\text{V}^2/\text{Hz}$)
• $k_B$: Boltzmann constant
• $T$: absolute temperature
• $R$: resistance

This result is frequency-independent (white noise) in the classical regime where $hf \ll k_BT$. The resistance $R$ is the dissipative element; the thermal noise is the fluctuation. The Nyquist theorem connects them exactly as the FDT predicts.

Applications in Statistical Mechanics

Brownian Motion

Brownian motion is perhaps the most intuitive illustration of the FDT. A colloidal particle suspended in a fluid experiences two effects from the surrounding molecules:

• Random kicks (fluctuations) that cause erratic motion
• Viscous drag (dissipation) that resists directed motion

Both originate from collisions with solvent molecules. The Einstein relation connects them:

$D = \frac{k_BT}{\gamma}$

where $D$ is the diffusion coefficient and $\gamma$ is the friction (drag) coefficient. Higher friction means slower diffusion, but the ratio is fixed by temperature. This was one of the earliest results embodying the FDT, and Einstein's 1905 derivation provided key evidence for the atomic hypothesis.

Johnson-Nyquist Noise

Thermal noise in resistors is the electronic analog of Brownian motion. Random thermal motion of charge carriers produces a fluctuating voltage across any resistor, even with no applied current. The mean-square noise voltage over a bandwidth $\Delta f$ is:

$\langle V^2 \rangle = 4k_BTR\,\Delta f$

This sets a fundamental lower bound on the noise floor of any electronic measurement. You cannot eliminate it without cooling the resistor. The result was predicted by Nyquist (1928) and confirmed experimentally by Johnson, providing one of the cleanest verifications of the FDT.

Electrical Conductivity

The Drude model gives the DC conductivity of a metal as:

$\sigma = \frac{ne^2\tau}{m}$

• $n$: charge carrier density
• $e$: elementary charge
• $\tau$: mean relaxation time between collisions
• $m$: effective carrier mass

From the FDT perspective, $\sigma$ can also be obtained via a Green-Kubo relation involving the current-current autocorrelation function. The relaxation time $\tau$ that governs dissipation (resistivity) is the same timescale over which current fluctuations decay at equilibrium.

Experimental Verification

Measurement Techniques

• Noise spectroscopy in electrical circuits directly measures Johnson-Nyquist noise and compares it to resistance values
• Microrheology tracks the Brownian motion of probe particles in complex fluids, extracting viscoelastic moduli via the FDT
• Atomic force microscopy (AFM) measures thermal fluctuations of a cantilever to calibrate its spring constant
• Dynamic light scattering observes intensity fluctuations in scattered light to determine diffusion coefficients
• NMR relaxation measurements connect spin fluctuation correlation times to relaxation rates, consistent with FDT predictions

Limitations and Breakdown

The FDT has well-defined boundaries of applicability:

• Strong perturbations: When the system is driven far from equilibrium, the linear response assumption fails and the standard FDT no longer holds.
• Non-ergodic systems: Glasses, spin glasses, and other systems that fail to explore their full phase space can violate the FDT. An "effective temperature" extracted from the FDT ratio may differ from the thermodynamic temperature.
• Quantum regime: At low temperatures where $k_BT \lesssim \hbar\omega$, quantum fluctuations (including zero-point motion) become important, and the classical FDT must be replaced by its quantum generalization.
• Timescale limitations: Measuring correlation functions on very short (sub-picosecond) timescales remains experimentally challenging.

Extensions and Generalizations

Quantum Fluctuation-Dissipation Theorem

The quantum FDT replaces the classical thermal factor with one that accounts for quantum statistics. In the frequency domain:

$\tilde{C}_{AA}(\omega) = \frac{2\hbar \, \text{Im}[\chi(\omega)]}{1 - e^{-\hbar\omega / k_BT}}$

At high temperatures ($k_BT \gg \hbar\omega$), this reduces to the classical result. At low temperatures, it captures zero-point fluctuations and the asymmetry between absorption and emission. The quantum FDT is essential for understanding phenomena like spontaneous emission, the Casimir effect, and low-temperature noise in mesoscopic conductors.

Fluctuation Theorems

Fluctuation theorems generalize the FDT to systems driven arbitrarily far from equilibrium. Two landmark results are:

• Jarzynski equality: $\langle e^{-W/k_BT} \rangle = e^{-\Delta F/k_BT}$, which relates the free energy difference $\Delta F$ between two equilibrium states to an exponential average of the work $W$ done over many non-equilibrium realizations of the process.
• Crooks fluctuation theorem: Relates the probability of observing a given amount of work in a forward process to the probability of observing the negative of that work in the time-reversed process.

These results are exact, not limited to near-equilibrium conditions. They recover the second law of thermodynamics as a statistical statement and have been verified experimentally in single-molecule pulling experiments and colloidal particle manipulations.

Non-Linear Response

For perturbations too strong for linear response, the response function is expanded in a Volterra series with higher-order susceptibilities $\chi^{(2)}, \chi^{(3)}, \ldots$. Generalized FDT-like relations connect these higher-order susceptibilities to higher-order equilibrium correlation functions, though the expressions become considerably more complex. This framework is used in nonlinear optics, nonlinear rheology, and turbulence theory.

Importance in Modern Physics

Nanoscale Systems

At the nanoscale, thermal fluctuations are comparable in magnitude to the signals of interest. The FDT is essential for:

• Predicting noise floors in nanoelectromechanical systems (NEMS)
• Understanding energy dissipation in molecular motors and nanoscale engines
• Calibrating AFM cantilevers and optical traps using thermal noise
• Designing energy harvesting devices that operate near the thermodynamic limits set by fluctuations

Biological Systems

Living systems operate far from equilibrium, but the FDT still provides a useful reference point. Deviations from FDT predictions in biological systems can reveal the presence of active, energy-consuming processes. Applications include:

• Characterizing the mechanics of the cytoskeleton using microrheology
• Studying conformational fluctuations in proteins and their connection to function
• Analyzing ion channel gating as a stochastic process
• Quantifying how much "activity" (ATP-driven motion) a cell generates beyond thermal fluctuations

Non-Equilibrium Thermodynamics

The FDT and its generalizations form the backbone of modern non-equilibrium thermodynamics. They provide the tools to:

• Derive transport coefficients from first principles via Green-Kubo relations
• Extend free energy calculations to non-equilibrium protocols via Jarzynski and Crooks relations
• Study active matter systems (self-propelled particles, bacterial suspensions) by measuring violations of the equilibrium FDT
• Analyze entropy production rates in driven systems using fluctuation theorems