The Fluctuation-Dissipation Theorem

1. Fluctuations and Linear Response

At thermal equilibrium, every microscopic quantity fluctuates around its average — a particle's velocity, a system's energy, the voltage across a resistor. These fluctuations are not noise to be filtered out; they carry deep information. When you perturb the same system with a small external force, its response — how quickly it relaxes back to equilibrium — is governed by the exact same microscopic dynamics that produced the spontaneous fluctuations.

2. Brownian Motion and the Einstein Relation

The oldest and clearest example: a colloidal particle suspended in fluid is kicked by molecular collisions (fluctuation) and slowed by viscous drag (dissipation). Einstein's 1905 analysis connected the two:

The same friction coefficient γ that damps a directed push also sets the size of the random jitter — a stiffer drag means both a slower drift response and smaller thermal fluctuations. You cannot have strong dissipation without correspondingly strong fluctuations at the same temperature; the two are locked together.

3. The General FDT (Callen-Welton)

Einstein's relation is a special case of a much more general 1951 result by Callen and Welton, which connects the power spectrum of spontaneous fluctuations of any observable to the imaginary (dissipative) part of the system's linear response function at the same frequency:

4. The Langevin Equation and Thermal Noise

The Langevin equation is the FDT written as an equation of motion: a deterministic drag force plus a random thermal force, with the two forced to have compatible magnitudes:

This coupling is what makes Langevin dynamics a correct thermostat for molecular simulations: scaling up the friction to control temperature more aggressively also requires scaling up the injected random kicks by exactly the matching amount, or the simulated temperature will drift.

5. Johnson-Nyquist Electrical Noise

Every resistor at finite temperature generates a random open-circuit voltage across its terminals — thermal (Johnson-Nyquist) noise — purely because its conduction electrons are the same electrons responsible for its resistive (dissipative) response:

Doubling the resistance doubles the dissipative response and doubles the noise power — exactly the FDT relationship, and the fundamental noise floor that limits every sensitive electronic amplifier and radio receiver.

6. Green-Kubo Relations

Green-Kubo relations extend the same idea to macroscopic transport coefficients — viscosity, thermal conductivity, electrical conductivity — expressing each as the time integral of an equilibrium fluctuation correlation function:

This is exactly how molecular dynamics codes measure macroscopic material properties: run an equilibrium simulation (no external field, no imposed gradient), record the autocorrelation of the microscopic flux, integrate it over time, and out comes the transport coefficient — no need to actually apply a shear or a temperature gradient at all.

7. JavaScript Simulation

// Langevin thermostat integrator — velocity form, obeys the FDT exactly
function langevinStep(v, m, gamma, kT, dt, randn) {
  // Deterministic drag
  const drag = -gamma * v / m;
  // Thermal noise amplitude set by the fluctuation-dissipation constraint
  const noiseAmp = Math.sqrt(2 * gamma * kT / dt) / m;
  const noise = noiseAmp * randn();  // randn() = standard normal sample
  return v + (drag + noise) * dt;
}

// Box-Muller for a standard normal random number
function randn() {
  const u1 = Math.random() || 1e-12, u2 = Math.random();
  return Math.sqrt(-2 * Math.log(u1)) * Math.cos(2 * Math.PI * u2);
}

// Verify the FDT numerically: measured MSD should match D = k_BT/gamma (Einstein relation)
function measureDiffusion(gamma, kT, m, dt, steps) {
  let x = 0, v = 0;
  const xs = [];
  for (let i = 0; i < steps; i++) {
    v = langevinStep(v, m, gamma, kT, dt, randn);
    x += v * dt;
    xs.push(x);
  }
  const msd = xs[xs.length - 1] ** 2;
  const D_measured = msd / (2 * steps * dt);
  const D_theory = kT / gamma;
  return { D_measured, D_theory };  // converge as steps → ∞
}

8. Applications and Limits

• Molecular dynamics thermostats: Langevin and Nosé-Hoover thermostats both rely on the FDT to keep injected random forces and friction consistent with the target temperature.
• Electronics: Johnson-Nyquist noise sets the absolute noise floor of every resistor-limited amplifier — no clever circuit design can beat it at a given temperature and bandwidth.
• Single-molecule biophysics: optical-tweezer measurements of a trapped bead's thermal fluctuations are routinely used to calibrate the trap's own stiffness via the equipartition and fluctuation-dissipation relations.
• Limits of the theorem: the FDT in this simple form assumes linear response near equilibrium; systems driven far from equilibrium (active matter, glasses below the glass transition) can violate it, which is itself an active research area.
• Climate science: the fluctuation-dissipation approach is used to estimate how a climate system's response to forcing (dissipation) relates to its natural internal variability (fluctuations).