11.02.03 · stat-mech-physics / kinetic-theory

The Boltzmann transport equation: H-theorem, the approach to equilibrium, and the BGK approximation

shipped3 tiersLean: none

Anchor (Master): Lifshitz & Pitaevskii, Physical Kinetics (1981), Ch. 1–2; Cercignani, Theory and Application of the Boltzmann Equation (1975)

Intuition Beginner

Imagine a gas whose molecules are not uniformly distributed — say, there is a hot spot on the left and a cold spot on the right. Molecules carry energy, momentum, and particles as they move and collide. The Boltzmann equation describes how the distribution of molecular velocities $f (r, v, t)$ evolves in time as molecules stream through space and collide with each other.

The equation has two parts. The streaming term accounts for molecules moving freely: a molecule at position $r$ with velocity $v$ will be at position $r + v d t$ a moment later. The collision term accounts for velocity changes due to binary collisions: when two molecules collide, they exchange momentum and energy, changing their velocities.

Boltzmann's great insight was the H-theorem: he defined a quantity H based on the distribution function f (weighted by its own logarithm, averaged over all velocities) and showed that H always decreases over time. The distribution f always evolves toward the equilibrium Maxwell-Boltzmann distribution, which minimises H. This was the first microscopic derivation of the second law of thermodynamics — and it sparked a debate about irreversibility that lasted decades.

Visual Beginner

Picture a cloud of gas molecules with a velocity distribution that is initially non-equilibrium (e.g., two streams of molecules moving in opposite directions). As time passes, collisions between molecules randomise the velocities, and the distribution relaxes toward the Maxwell-Boltzmann form (a Gaussian in each velocity component). The H-function decreases monotonically, like a ball rolling down a hill toward the equilibrium valley.

Worked example Beginner

The BGK (Bhatnagar-Gross-Krook) model replaces the full collision operator with a simple relaxation term: the distribution $f$ relaxes toward the local Maxwell-Boltzmann distribution $f_{0}$ at a rate $1/ τ$ :

\frac{df}{d t} = - \frac{f - f _{0}}{τ},

where $τ$ is the relaxation time. In the absence of spatial gradients (homogeneous gas), the solution is

f (t) = f_{0} + (f (0) - f_{0}) e^{- t / τ} .

The distribution relaxes exponentially to equilibrium with time constant $τ = l / \overset{v}{ˉ}$ (the mean free time). This is the simplest model of approach to equilibrium and captures the essential physics: collisions drive the system toward equilibrium at a rate set by the collision frequency.

Check your understanding Beginner

Formal definition Intermediate+

The Boltzmann transport equation for a dilute gas of hard spheres is

\frac{\partial f}{\partial t} + v \cdot \nabla_{r} f + \frac{F}{m} \cdot \nabla_{v} f = C [f],

where $f (r, v, t)$ is the single-particle distribution function, $F$ is the external force, and $C [f]$ is the collision operator:

C [f] = \int d^{3} v_{1} \int d Ω ∣ v - v_{1} ∣ σ (Ω) [f^{'} f_{1}^{'} - f f_{1}] .

Here $f^{'} = f (v^{'})$ , $f_{1} = f (v_{1})$ , $f_{1}^{'} = f (v_{1}^{'})$ , and the primed velocities are the post-collision velocities. The gain term $f^{'} f_{1}^{'}$ counts collisions that increase $f$ at velocity $v$ ; the loss term $f f_{1}$ counts collisions that decrease it.

Key result: The H-theorem

Theorem (Boltzmann). Define $H [f] = \int f ln f d^{3} v$ . Then $d H / d t \leq 0$ , with equality if and only if $f$ is the Maxwell-Boltzmann distribution.

The proof uses the identity $ln (f^{'} / f_{1}^{'}) - ln (f / f_{1}) \leq 0$ (from detailed balance) and shows that the collision integral is non-positive. The equilibrium distribution satisfies detailed balance $f^{'} f_{1}^{'} = f f_{1}$ , which gives $f_{0} \propto e^{- m v^{2} / (2 k T)}$ .

BGK model

The BGK approximation replaces $C [f]$ with $- ν (f - f_{0})$ where $ν = 1/ τ$ is the collision frequency and $f_{0}$ is the local Maxwellian with the same density, mean velocity, and temperature as $f$ . This preserves the conservation laws (mass, momentum, energy) while simplifying the collision operator to a linear relaxation term.

Bridge. The Boltzmann equation builds toward the fluctuation-dissipation theorem 11.02.04, which provides an alternative derivation of transport coefficients from equilibrium time-correlation functions. This is exactly the Green-Kubo approach: the transport coefficients from the Chapman-Enskog solution of the Boltzmann equation agree with those from the Green-Kubo relations. The foundational reason the Boltzmann equation predicts irreversibility (the H-theorem) despite time-reversible microscopic dynamics is the molecular chaos assumption (Stosszahlansatz): the velocities of two colliding particles are assumed to be uncorrelated before the collision. This is the bridge between microscopic reversibility and macroscopic irreversibility. Putting these together, the H-theorem is the kinetic-theory derivation of the second law, and the BGK model captures its essential content in a simplified form.

Exercises Intermediate+

Exercise 4 (hard, symbolic).

The Chapman-Enskog expansion writes $f = f_{0} + ϵ f_{1} + \dots$ where $ϵ$ is the Knudsen number. At first order, show that $f_{1} = - τ (\partial f_{0} / \partial t + v \cdot \nabla f_{0})$ and that substituting into the moment equations recovers the Navier-Stokes equations with the transport coefficients $η = nk T τ$ , $κ = (5/2) nk T τ / m$ .

ϵ^{0}

C [f_{0}] = 0

(equilibrium). At order

ϵ^{1}

\partial f_{0} / \partial t + v \cdot \nabla f_{0} = C^{'} [f_{0}] \cdot f_{1}

where

C^{'}

is the linearised collision operator. For the BGK model:

C^{'} [f_{0}] \cdot f_{1} = - ν f_{1}

, so

f_{1} = - (1/ ν) (\partial f_{0} / \partial t + v \cdot \nabla f_{0}) = - τ (\partial f_{0} / \partial t + v \cdot \nabla f_{0})

The stress tensor from $f_{1}$ : $σ_{ij} = \int m (v_{i} - u_{i}) (v_{j} - u_{j}) f_{1} d^{3} v$ . Evaluating with the Maxwellian gradients gives $σ_{ij} = - η (\partial u_{i} / \partial x_{j} + \partial u_{j} / \partial x_{i} - \frac{2}{3} δ_{ij} \nabla \cdot u)$ with $η = nk T τ$ .

The heat flux: $q_{i} = \int \frac{1}{2} m ∣ v - u ∣^{2} (v_{i} - u_{i}) f_{1} d^{3} v = - κ \partial T / \partial x_{i}$ with $κ = (5/2) nk T τ / m$ (for a monatomic gas, the Prandtl number from BGK is Pr = 1, compared to the exact hard-sphere value Pr = 2/3).

Exercise 5 (hard, symbolic).

Discuss the Loschmidt paradox: if the microscopic dynamics are time-reversible, how can the Boltzmann equation predict irreversible behaviour? What is the role of the molecular chaos (Stosszahlansatz) assumption?

t

would make the system retrace its path, so

H

would increase — contradicting the H-theorem.

Resolution: the molecular chaos assumption assumes that the velocities of two particles about to collide are uncorrelated: $f_{2} (v, v_{1}) = f (v) f (v_{1})$ . This is true at $t = 0$ (by hypothesis) but is not maintained by the exact dynamics — post-collision velocities are correlated. The assumption is justified in the dilute-gas limit ( $n λ^{3} ≪ 1$ ) where three-body collisions are rare and correlations decay between successive binary collisions.

The Boltzmann equation is therefore an approximate description that becomes exact in the Boltzmann-Grad limit ( $N \to \infty$ , $a \to 0$ , $N a^{2} = const$ ). The irreversibility is not a property of the microscopic dynamics but of the approximation that discards correlations. This is the content of Lanford's theorem (1975): the Boltzmann equation is valid for a time $t < t_{0} / (5 λ)$ where $t_{0}$ is the mean free time, i.e., for a finite but physically relevant time window.

Exercise 6 (hard, symbolic).

The linearised Boltzmann equation for small perturbations $δ f = f - f_{0}$ in a homogeneous gas is $\partial δ f / \partial t = L [δ f]$ where $L$ is the linearised collision operator. Show that $L$ is self-adjoint in the weighted inner product $⟨ g, h ⟩ = \int (g h / f_{0}) d^{3} v$ and has non-positive eigenvalues.

C [f_{0} + δ f] = C^{'} [f_{0}] \cdot δ f + O (δ f^{2})

. From the H-theorem proof,

L

can be written in symmetric form:

$L [g] = - \frac{1}{2} \int\int\int f_{0} f_{01} W [g^{'} / f_{0}^{'} + g_{1}^{'} / f_{01}^{'} - g / f_{0} - g_{1} / f_{01}] d^{3} v_{1} d Ω$

where $g = δ f$ . The inner product $⟨ g, L [g]⟩ = \int (g L [g] / f_{0}) d^{3} v$ can be shown (by the same symmetrisation used in the H-theorem) to be non-positive:

$⟨ g, L [g]⟩ = - \frac{1}{4} \int\int\int f_{0} f_{01} W [g^{'} / f_{0}^{'} + g_{1}^{'} / f_{01}^{'} - g / f_{0} - g_{1} / f_{01}]^{2} d^{3} v d^{3} v_{1} d Ω \leq 0$ .

The eigenvalues of $L$ are non-positive, and the zero eigenvalues correspond to the collision invariants ( $1, v, v^{2}$ ). This is the spectral property that guarantees relaxation to equilibrium.

Lean formalization Intermediate+

The Boltzmann equation is an integro-differential equation for the distribution function $f (r, v, t)$ . Formalising it requires: (1) the distribution function as a measurable function on $R^{3} \times R^{3} \times R$ ; (2) the collision operator as a bilinear integral operator satisfying detailed balance; (3) the H-functional and its monotonicity. The H-theorem is a functional inequality of the form $(x - y) ln (x / y) \geq 0$ , which is provable in Mathlib. The main gap is the collision-operator framework.

Advanced results Master

Lanford's theorem

Lanford (1975) proved that the Boltzmann equation is the exact limit of the Liouville equation in the Boltzmann-Grad limit ( $N \to \infty$ , particle diameter $a \to 0$ , $N a^{2} = const$ ). The derivation is valid for a finite time $0 \leq t \leq C / λ$ (a fraction of the mean free time), during which most particles undergo at most one collision. This is the rigorous justification of the molecular chaos assumption.

Hydrodynamic limits

The Boltzmann equation reduces to the Euler equations (inviscid fluid) in the limit $ϵ \to 0$ (Knudsen number going to zero) at order $ϵ^{0}$ , and to the Navier-Stokes equations at order $ϵ^{1}$ (Chapman-Enskog expansion). This establishes the connection between kinetic theory and continuum fluid mechanics.

Entropy and irreversibility

The H-theorem provides the first microscopic derivation of the second law. The entropy production rate $\overset{σ}{˙} = - k_{B} d H / d t$ is non-negative, vanishing only at equilibrium. This connects the Boltzmann equation to the theory of irreversibility 11.02.04 and the fluctuation-dissipation theorem, where the entropy production rate is proportional to the product of fluxes and forces.

Synthesis. The Boltzmann equation builds toward the fluctuation-dissipation theorem 11.02.04 and the Green-Kubo relations, which provide an alternative route to transport coefficients. The central insight is that the H-theorem is the kinetic-theory embodiment of the second law: entropy increases because collisions randomise velocities, and the molecular chaos assumption is the link between time-reversible dynamics and irreversible macroscopic behaviour. This is dual to the information-theoretic interpretation of entropy: $H = - S / k_{B}$ is the information content of the distribution, and the H-theorem says information is lost (entropy increases) because collisions destroy correlations. The foundational reason the Boltzmann equation works is that in a dilute gas, most collisions are binary and uncorrelated; putting these together with the Chapman-Enskog expansion, the microscopic collision dynamics produce the macroscopic Navier-Stokes equations, generalising the mean-free-path estimates 11.02.02 to exact numerical prefactors.

Full proof set Master

Proposition. The equilibrium solution of the Boltzmann equation is the Maxwell-Boltzmann distribution $f_{0} = n (m / (2 π k T))^{3/2} exp (- m ∣ v - u ∣^{2} / (2 k T))$ .

Proof. The collision operator vanishes when detailed balance holds: $f (v^{'}) f (v_{1}^{'}) = f (v) f (v_{1})$ for all collisions. Taking logarithms: $ln f (v^{'}) + ln f (v_{1}^{'}) = ln f (v) + ln f (v_{1})$ . This is a conservation law: $ln f$ is a linear combination of the collision invariants:

ln f (v) = A + B \cdot v + C ∣ v ∣^{2} .

The constants $A, B, C$ are determined by the conditions: $\int f d^{3} v = n$ (density), $\int v f d^{3} v = n u$ (mean velocity), $\int ∣ v - u ∣^{2} f d^{3} v = 3 nk T / m$ (temperature). Solving gives $A = ln [n (m / (2 π k T))^{3/2}] - m u^{2} / (2 k T)$ , $B = m u / (k T)$ , $C = - m / (2 k T)$ . Substituting recovers the Maxwell-Boltzmann distribution. $□$

Connections Master

11.02.01 The Maxwell-Boltzmann distribution is the equilibrium solution of the Boltzmann equation.
11.02.02 Transport coefficients from the mean-free-path method are the first approximation to the Chapman-Enskog solution of the Boltzmann equation.
11.02.04 The fluctuation-dissipation theorem provides an alternative derivation of transport coefficients from equilibrium correlations.
11.04.01 The canonical ensemble gives the Maxwell-Boltzmann distribution as the most probable distribution.
02.16.01 The Boltzmann equation is a PDE on $R^{3} \times R^{3} \times R$ ; its functional-analytic treatment requires Sobolev space methods.

Historical and philosophical context Master

Boltzmann introduced his equation in 1872, provoking immediate objections. Loschmidt (1876) noted that time-reversal of the microscopic dynamics would reverse the sign of $d H / d t$ — the reversibility paradox. Zermelo (1896) noted that Poincare recurrence implies any system returns arbitrarily close to its initial state — the recurrence paradox. Both paradoxes were resolved by recognising that the H-theorem holds with probability approaching 1 for macroscopic systems (the thermodynamic limit), not with absolute certainty. The molecular chaos assumption selects the overwhelmingly probable direction of evolution.

Bibliography Master

@article{boltzmann1872,
  author = {Boltzmann, Ludwig},
  title = {Weitere Studien \"uber das W\"armegleichgewicht unter Gasmolek\"ulen},
  journal = {Sitzungsber. Akad. Wiss. Wien},
  volume = {66},
  pages = {275--370},
  year = {1872}
}

@book{kardar2007,
  author = {Kardar, Mehran},
  title = {Statistical Physics of Particles},
  publisher = {Cambridge University Press},
  year = {2007}
}

@book{chapman-cowling1970,
  author = {Chapman, Sydney and Cowling, T. G.},
  title = {The Mathematical Theory of Non-Uniform Gases},
  edition = {3rd},
  publisher = {Cambridge University Press},
  year = {1970}
}

@book{reif1965,
  author = {Reif, Frederick},
  title = {Fundamentals of Statistical and Thermal Physics},
  publisher = {McGraw-Hill},
  year = {1965}
}

}

Prerequisites

11.02.01
11.02.02

Used in

11.02.04

Tier anchors

beginner: Reif, Fundamentals of Statistical and Thermal Physics (1965), Ch. 13
intermediate: Kardar, Statistical Physics of Particles (2007), Ch. 5; Chapman & Cowling, Mathematical Theory of Non-Uniform Gases (1970), Ch. 3–4
master: Lifshitz & Pitaevskii, Physical Kinetics (1981), Ch. 1–2; Cercignani, Theory and Application of the Boltzmann Equation (1975)

References

Boltzmann 1872 — Weitere Studien uber das Warmegleichgewicht unter Gasmolekulen · Sitzungsber. Akad. Wiss. Wien 66, 275–370. The Boltzmann equation and H-theorem. Boltzmann defined H = integral of f ln f d^3v and showed dH/dt <= 0, with equality only for equilibrium.
Kardar — Statistical Physics of Particles (Cambridge, 2007) · Ch. 5 Kinetic theory. The Boltzmann equation, collision operator, H-theorem, and the Chapman-Enskog expansion.
Chapman & Cowling — The Mathematical Theory of Non-Uniform Gases, 3rd ed. (Cambridge, 1970) · Ch. 3 The Boltzmann equation and its properties; Ch. 4 The Maxwell-Boltzmann distribution as the equilibrium solution.
Cercignani — Theory and Application of the Boltzmann Equation (Scottish Academic Press, 1975) · Ch. 1–2 Existence and uniqueness theorems, the H-theorem, and the BGK model equation.

Estimated time

beginner: 15 min
intermediate: 40 min
master: 75 min