FLUCTUATIONS AND DISSIPATION

At Bell Telephone Laboratories in 1927, was chasing a problem that would not go away. Every amplifier he built, every circuit he measured, carried a faint random voltage he could not eliminate. He cooled the components, used the purest materials, removed every external source of interference — and still the hiss remained. Worse, it got louder when he warmed the resistor and louder still when he raised its resistance, in a way that had nothing to do with the current flowing through it. This was not a flaw in his apparatus. It was the thermal motion of the electrons themselves, jostling at temperature $T$, made audible.

Johnson took the puzzle to his theorist colleague , and within a year, in 1928, Nyquist had explained it — not with a model of electrons, but from pure thermodynamics. The noise, he showed, was the electrical signature of the same molecular agitation that had used to explain Brownian motion two decades earlier. The jiggle that moved Brown's pollen grains and the hiss in Johnson's resistor are the same randomness, seen in two different instruments. This topic is about that randomness: how big it is, why it is usually invisible, and the deep theorem that ties it to friction.

In the canonical ensemble, a system in contact with a heat bath does not hold a fixed energy — energy flows back and forth, and the system's energy fluctuates around its mean. Far from being a vague qualitative idea, the size of that fluctuation is fixed exactly by a derivative we already know, the heat capacity:

$\langle (\Delta E)^2 \rangle = k_B T^2 C_v$

The mean-square energy fluctuation equals $k_B T^2$ times the heat capacity. This is a startling statement on its own: the spontaneous wobble of a system at equilibrium (the left side) is set by how strongly it responds to being heated (the right side). A system that soaks up a lot of heat for a small temperature rise is also one that fluctuates wildly on its own. That linkage — fluctuation tied to response — is the seed of everything that follows.

If fluctuations are always present, why does thermodynamics get away with treating energy, pressure and volume as sharp, definite numbers? The answer is the most important number in statistical mechanics. The relative size of a fluctuation — its standard deviation divided by the mean — scales as

$\frac{\sigma_E}{\langle E\rangle} = \sqrt{\frac{2}{3N}} \;\propto\; \frac{1}{\sqrt{N}}$

the inverse square root of the number of particles. For a macroscopic sample with $N \approx 10^{23}$, this is about one part in $10^{12}$ — a wobble so small that no instrument could ever resolve it, which is exactly why the gas laws look exact. But the law cuts both ways. Shrink the system and the fluctuations grow: for a protein, a quantum dot, or a hundred-atom nanostructure, $N$ is small enough that the relative fluctuations reach percent levels and dominate the behavior. The same $1/\sqrt{N}$ that makes a litre of gas perfectly predictable makes a nanomachine fundamentally noisy.

Nyquist's 1928 result put a precise number on Johnson's hiss. The mean-square open-circuit voltage across a resistor $R$ at temperature $T$, measured over a frequency bandwidth $\Delta f$, is

$\langle V^2 \rangle = 4 k_B T R \, \Delta f$

Read it term by term: the noise power rises linearly with temperature (hotter electrons jiggle harder), linearly with resistance, and linearly with the bandwidth you listen over. What is absent is as telling as what is present — there is no dependence on the resistor's material, shape, or how it was made. Carbon, metal film, a column of salt water: at the same $R$ and $T$ they hiss identically. That universality is the fingerprint of a thermodynamic result, and it is why Nyquist could derive it without ever modelling an electron.

This noise sets a hard floor on every measurement. No amplifier can be quieter than the thermal hiss of its own input resistor, which is why sensitive detectors — radio telescopes, gravitational-wave interferometers, the front end of an MRI — are cooled to cryogenic temperatures: lowering $T$ is the only way to lower the floor.

By 1951, Herbert Callen and Theodore Welton had recognized that Einstein's relation and Nyquist's formula were two instances of one law, and in 1957 gave it its general and final form: the fluctuation–dissipation theorem. Its content is that the two things engineers usually keep in separate columns — the random fluctuations a system shows on its own, and the dissipation (friction, resistance, drag) it exhibits when you push on it — are not independent. They are rigidly proportional, with $k_B T$ as the constant of proportionality.

It could not be otherwise. The very same molecular collisions that knock a particle around at random are the ones that resist its motion when it is dragged through the medium. The kicks (fluctuation) and the drag (dissipation) come from one mechanism, so their magnitudes are locked together. A resistor that dissipates electrical energy must generate Johnson noise; a viscous fluid that damps a particle must also jiggle it. You cannot have dissipation without fluctuation, or fluctuation without dissipation — they are two readings of a single underlying agitation.

The theorem's first instance was already complete in 1905. In his Brownian-motion paper, connected the diffusion of a suspended particle — a pure fluctuation — to its mobility under an applied force — a pure dissipation — through the relation

$D = \mu \, k_B T$

where $D$ is the diffusion constant and $\mu$ the mobility (velocity per unit force). The left side measures how vigorously the particle wanders on its own; the right side measures how easily a force pushes it through the fluid's friction. That a single $k_B T$ joins them is the fluctuation–dissipation theorem half a century before it had a name. Around the same time reached the same result by a different route, and the two together turned Brownian motion from a curiosity into the most direct evidence that atoms are real — the thread we followed in Brownian motion.

Usually fluctuations are the small print of thermodynamics, suppressed by $1/\sqrt{N}$. But there is a place where they escape that suppression and seize control: a critical point. As a fluid approaches the critical temperature where the distinction between liquid and gas vanishes, its compressibility — its response to pressure — diverges, and by EQ.01's logic the density fluctuations diverge with it. Smoluchowski and Einstein showed that these runaway fluctuations grow until they reach the wavelength of light and scatter it, turning a clear fluid milky-white. This is critical opalescence, and it is fluctuation made directly visible to the eye.

Critical opalescence is the doorway to critical phenomena and universality, where the way these fluctuations diverge is captured by a handful of Critical exponent numbers that, astonishingly, are the same for fluids, magnets, and alloys alike. The randomness that began as Johnson's unwanted hiss becomes, at the critical point, the entire story — and connects back to the partition function, whose second derivative it was all along.