All lessons

Module 6: Thermodynamics and statistical physics · Lesson 6.2 · about 50 min

Fluctuations, diffusion and Brownian motion

Fluctuations for CSIR NET: energy and number fluctuations in the canonical and grand canonical ensembles, random walks and the diffusion equation, the Einstein and Stokes-Einstein relations, the Langevin equation and the fluctuation-dissipation theorem, Johnson, shot and kT/C noise, sedimentation, polymers and optical traps.

Key points, meanings and formulas on one page: choose "Save as PDF" in the print box.

Module 6 · about 50 minutes. Thermodynamics deals in averages, but every average comes with a spread, and the spread carries information. Energy fluctuations measure the heat capacity; number fluctuations measure the compressibility; the jiggling of a pollen grain measured Avogadro’s number; and the hiss in an amplifier measures temperature. This lesson covers the fluctuation results NET asks for, then random walks, diffusion and the Langevin equation. The ensembles themselves are in ensembles and partition functions; equipartition, used throughout, is in the classical gas.

The same molecular collisions that make a system dissipate (friction, resistance, viscosity) also make it fluctuate. The size of the fluctuation is always k_BT times the response: energy spread from heat capacity, number spread from compressibility, voltage noise from resistance, and diffusion from mobility. That is the fluctuation-dissipation theorem, and it runs through every section below.

1. Fluctuations in the ensembles

Energy (canonical). With Z=∑e−βE, ⟨E⟩=−∂βln⁡Z and

⟨(ΔE)2⟩=∂2ln⁡Z∂β2=kBT2CV,ΔErms⟨E⟩∼1N

For a monatomic ideal gas, ⟨E⟩=32NkBT and CV=32NkB, so ΔE/E=2/3N: for a mole, about 10−12. This is why the canonical and microcanonical ensembles give the same thermodynamics. A positive CV is also forced: a variance cannot be negative.

Particle number (grand canonical).

⟨(ΔN)2⟩=kBT(∂⟨N⟩∂μ)T,V=⟨N⟩2kBTκTV

For a classical ideal gas κT=1/P, so ⟨(ΔN)2⟩=⟨N⟩: Poisson statistics. Quantum gases differ: bosons fluctuate more (⟨n⟩(1+⟨n⟩) per mode, bunching) and fermions less (⟨n⟩(1−⟨n⟩), antibunching). Near a liquid-gas critical point κT→∞: density fluctuations grow to the wavelength of light and scatter it strongly, the milky critical opalescence.

Einstein’s formula. The probability of a fluctuation away from equilibrium is P∝eΔStot/kB. For a small system in contact with a reservoir it gives, for example, ⟨(ΔT)2⟩=kBT2/CV and ⟨(ΔV)2⟩=kBTVκT.

⟨(ΔX)²⟩
variance, ⟨X²⟩ − ⟨X⟩²; its square root is the rms fluctuation
D
diffusion coefficient: ⟨x²⟩ = 2Dt along each axis
γ
friction coefficient, force per unit velocity; for a sphere in a fluid 6πηa
μ
mobility, drift velocity per unit force, 1/γ
Physical meaning
Fluctuations are not noise to be averaged away. Their size, kBT times a response, is a measurement of the response itself, and Brownian motion was the first direct evidence that atoms exist.

2. Random walks and diffusion

A walker takes N independent steps of length l, each left or right with probability ½. The displacement x=l∑si has

⟨x⟩=0,⟨x2⟩=Nl2,P(m)=N!(N+m2)!(N−m2)!2−N

for net displacement ml. The cross terms ⟨sisj⟩ vanish for i≠j, which is the whole reason for the N. For large N the binomial becomes a Gaussian, and with steps every τ, so N=t/τ,

P(x,t)=14πDte−x2/4Dt,∂P∂t=D∂2P∂x2,D=l22τ

In d dimensions, ⟨r2⟩=2dDt. With unequal step probabilities p and q, the mean drifts as N(p−q)l and the variance is 4Npql2.

Three hundred walkers take unit steps in random directions. The cloud's rms radius grows as √N (circle), and the measured ⟨r²⟩ follows the line N l². In two dimensions ⟨r²⟩ = 4Dt with D = l²/4τ.

Einstein relation. In a weak force F, particles drift at v=μF. In equilibrium in a potential, the drift current nμF must cancel the diffusion current −Ddn/dx, and the density must be Boltzmann, n∝e−U/kBT. Together:

D=μkBT,D=kBT6πηa  (Stokes-Einstein, sphere of radius a)

For electrons in a semiconductor, D/μe=kBT/e (about 26 mV at room temperature). Perrin used Stokes-Einstein in 1908: measure ⟨x2⟩, η, a and T, and out comes kB, hence Avogadro’s number.

Sedimentation. Spheres of density ρp in a fluid of density ρf settle into n(z)∝e−z/h with h=kBT/meffg and meff=43πa3(ρp−ρf). For 0.5 μm radius latex (Δρ=50 kg/m³) at 300 K, h=16 μm, easy to see in a microscope.

3. The Langevin equation

Newton’s law for a particle of mass m in a fluid, with friction and random kicks:

mdvdt=−γv+F(t),⟨F(t)⟩=0,⟨F(t)F(t′)⟩=2γkBTδ(t−t′)

The strength of the random force is not free: it is fixed so that the particle reaches ⟨12mv2⟩=12kBT. Results, with τm=m/γ:

  • velocity correlation ⟨v(0)v(t)⟩=kBTme−t/τm; integrating it gives D=∫0∞⟨v(0)v(t)⟩dt=kBT/γ (a Green-Kubo formula);
  • mean square displacement
⟨x2(t)⟩=2D[t−τm(1−e−t/τm)]⟶{kBTmt2t≪τm  (ballistic)2Dtt≫τm  (diffusive)

Worked example: when is a bead ballistic?

The bead. Silica, radius 0.5 μm, density 2000 kg/m³, in water (η = 1.0 mPa s) at 293 K. Mass m = (4/3)π(0.5 × 10⁻⁶)³ × 2000 = 1.05 × 10⁻¹⁵ kg.

Friction. γ = 6πηa = 9.42 × 10⁻⁹ kg/s, so τm = m/γ = 1.1 × 10⁻⁷ s.

Diffusion. D = kBT/γ = 4.3 × 10⁻¹³ m²/s: in 1 s the rms step along one axis is √(2D) = 0.93 μm.

Ballistic regime. Below about 0.1 μs the bead moves in straight lines at vrms = √(kBT/m) = 2 mm/s, covering only about 0.2 nm before its velocity is randomised. Einstein thought this regime unmeasurable; optical tweezers resolved it in 2010.

Optical trap. Add a spring −κx. Equipartition gives κ⟨x2⟩=kBT at once, which is how trap stiffness is calibrated: a 2 nm rms jitter at 300 K means κ=4.14×10−21/(2×10−9)2≈1 mN/m. Positions relax with time constant γ/κ.

Escape over a barrier. A Brownian particle in a well of depth Eb≫kBT escapes at a rate ∝e−Eb/kBT (Kramers), the Arrhenius law. Raising T from 300 K to 310 K speeds a 0.5 eV process by a factor e(0.5/kB)(1/300−1/310)=1.9.

4. Noise in circuits

The fluctuation-dissipation theorem applied to charges in a resistor gives Johnson-Nyquist noise:

⟨V2⟩=4kBTRΔf,⟨I2⟩=4kBTΔfR

white up to f∼kBT/h≈6 THz at room temperature. A 1 MΩ resistor at 300 K over a 10 kHz bandwidth produces 4kBTRΔf=12.9 μV rms. Maximum noise power delivered to a matched load is kBTΔf, independent of R, which defines an amplifier’s noise temperature.

  • kT/C noise. A capacitor C in contact with any resistor at T has 12C⟨V2⟩=12kBT: Vrms=kBT/C, 64 μV for 1 pF at 300 K, whatever R is. It sets the floor of camera pixels.
  • Shot noise. Current carried by discrete independent charges fluctuates as ⟨I2⟩=2eIΔf (Poisson arrivals). It is not thermal: it exists at zero temperature. For 1 mA over 1 MHz, Irms=18 nA.

5. Polymers as random walks

A freely jointed chain of N links of length b is a three-dimensional random walk: Rrms=Nb. The number of configurations with end-to-end vector R→ is Gaussian, so the entropy is S=S0−3kBR2/2Nb2. Stretching it needs a force

F=−T∂S∂R=3kBTNb2R

an entropic spring, stiffer when hotter. A stretched rubber band warms up when stretched quickly and contracts when heated, the opposite of a metal wire.

“Fluctuations are always negligible in macroscopic systems.” Relative fluctuations of extensive quantities fall as 1/N, but two important exceptions exist. Near a critical point the response functions (CV, κT, the susceptibility) diverge, and so do the fluctuations, over all length scales: you can see them as critical opalescence. And in small systems (a single protein, a nanoscale capacitor, a single trapped bead), kBT fluctuations are the main physics, not a correction.

  • ⟨(ΔE)2⟩=kBT2CV; ⟨(ΔN)2⟩=kBT∂N/∂μ=N2kBTκT/V; classical ideal gas Poisson; diverge at the critical point.
  • Random walk ⟨x2⟩=Nl2; diffusion ⟨x2⟩=2Dt per axis, ⟨r2⟩=2dDt; D=l2/2τ in 1D.
  • Einstein D=μkBT; Stokes-Einstein D=kBT/6πηa; sedimentation height kBT/meffg.
  • Langevin: ⟨FF⟩=2γkBTδ; τm=m/γ; ⟨x2⟩ ballistic kBTt2/m then diffusive 2Dt; D=∫⟨v(0)v(t)⟩dt.
  • Trap κ⟨x2⟩=kBT; Kramers/Arrhenius e−Eb/kBT.
  • Johnson 4kBTRΔf; kT/C Vrms=kBT/C; shot 2eIΔf.
  • Polymer Rrms=Nb, entropic spring F=3kBTR/Nb2.

Practise: exam questions on this lesson

NET Part C style, each with a full solution. More on the CSIR NET Physics practice page.

Next: the Ising model and Landau theory, where fluctuations grow without limit at a critical point. Molecular motion behind diffusion is in kinetic theory.

Check yourself

1In the canonical ensemble, the mean square energy fluctuation equals what?
2A Brownian particle's mean square displacement grows as t² at very short times and as t at long times. What sets the crossover?
3What links the random force in the Langevin equation to the friction coefficient γ?

Sources

  • F. Reif, Fundamentals of Statistical and Thermal Physics (1965), chapters 1 and 15
  • R. K. Pathria and P. D. Beale, Statistical Mechanics, 3rd ed. (2011), chapter 15
  • A. Einstein, "Über die von der molekularkinetischen Theorie der Wärme geforderte Bewegung von in ruhenden Flüssigkeiten suspendierten Teilchen", Annalen der Physik 17, 549 (1905)
  • H. Nyquist, "Thermal agitation of electric charge in conductors", Physical Review 32, 110 (1928)
  • CSIR-UGC NET Physical Sciences syllabus, Thermodynamic and Statistical Physics unit, published by the National Testing Agency

Your ticks stay in this browser. Save them as a file on the course home.

Updated 2026-11-22. Not affiliated with CSIR, UGC or the National Testing Agency. Check the official CSIR NET website for the current pattern. All questions are original.