11.02.02 · stat-mech-physics / kinetic-theory

Transport phenomena: diffusion, viscosity, thermal conductivity, and the mean free path

shipped3 tiersLean: none

Anchor (Master): Pathria & Beale, Statistical Mechanics, 3rd ed. (2011), Ch. 11; Kardar, Statistical Physics of Particles (2007), Ch. 5

Intuition Beginner

A gas is a sea of molecules in constant random motion. Each molecule travels in a straight line until it collides with another molecule, then changes direction and continues. The average distance a molecule travels between collisions is the mean free path $l$ .

Three things happen because of these collisions. First, molecules wander: if you release a blob of perfume in still air, it slowly spreads out because each perfume molecule follows a random, zigzag path. This is diffusion. Second, adjacent layers of gas moving at different speeds exchange momentum through collisions, resisting the shear. This is viscosity. Third, fast-moving (hot) molecules in one region collide with slow-moving (cold) molecules in the adjacent region, transferring kinetic energy. This is thermal conduction.

All three processes share a common structure. Each involves a flux (a flow of something — particles, momentum, energy) driven by a gradient (a spatial variation — of concentration, velocity, temperature). And all three are controlled by the same microscopic quantity: the mean free path $l$ and the mean molecular speed $v$ .

Visual Beginner

Picture a gas between two parallel plates. The top plate moves to the right; the bottom plate is stationary. Molecules near the top plate are dragged rightward. Through collisions, they transfer rightward momentum downward. A velocity gradient develops: fast at the top, slow at the bottom. The rate of momentum transfer — the friction force per unit area — is the viscous stress, proportional to the velocity gradient.

In each case the flux is proportional to the gradient, with the proportionality constant being the transport coefficient: the diffusion constant $D$ , the viscosity $η$ , or the thermal conductivity $κ$ .

Worked example Beginner

Mean free path in air. Air at standard temperature and pressure (STP) has number density $n \approx 2.5 \times 1 0^{25}$ molecules per cubic metre. The molecular diameter is $σ \approx 3.7$ angstroms. The mean free path is

l = \frac{1}{n σ ^{2} 2 π} \approx \frac{1}{2.5 \times 1 0 ^{25} \times ( 3.7 \times 1 0 ^{- 10} ) ^{2} \times 4.44} \approx 6.5 \times 1 0^{- 8} m .

That is about 65 nanometres — roughly 200 molecular diameters. The mean molecular speed at room temperature is $v \approx 500$ m/s. So the average time between collisions is $τ = l / v \approx 1.3 \times 1 0^{- 10}$ seconds — about 0.13 nanoseconds.

From these, the diffusion constant is roughly $D \approx l v /3 \approx 1 0^{- 5}$ m $^{2}$ /s and the viscosity is $η \approx mn v l /3 \approx 1.8 \times 1 0^{- 5}$ Pa s, both in reasonable agreement with measured values for air.

Check your understanding Beginner

Formal definition Intermediate+

Mean free path

For a gas of hard spheres of diameter $σ$ at number density $n$ , the mean free path is

l = \frac{1}{2 n π σ ^{2}} .

The factor $2$ accounts for the relative velocity of colliding pairs: two molecules approach each other with a typical relative speed $2 v_{rms}$ . The collision frequency is $ν = v / l$ and the collision time is $τ = l / v$ .

Fick's law of diffusion

The particle flux (number of particles crossing unit area per unit time) is proportional to the negative concentration gradient:

J_{n} = - D \nabla n,

where $D$ is the diffusion coefficient (self-diffusion constant). From kinetic theory, $D = \overset{v}{ˉ} l /3$ , where $\overset{v}{ˉ} = 8 k T / π m$ is the mean speed.

Newton's law of viscosity

The viscous stress (force per unit area) between layers of gas with a velocity gradient is

τ_{x y} = - η \frac{d v _{x}}{d y},

where $η$ is the dynamic viscosity. From kinetic theory, $η = mn \overset{v}{ˉ} l /3 = ρ \overset{v}{ˉ} l /3$ , where $ρ = mn$ is the mass density.

Fourier's law of heat conduction

The heat flux is proportional to the negative temperature gradient:

J_{Q} = - κ \nabla T,

where $κ$ is the thermal conductivity. From kinetic theory, $κ = c_{V} \overset{v}{ˉ} l /3$ per unit volume, or $κ = n c_{V}^{(1)} \overset{v}{ˉ} l /3$ where $c_{V}^{(1)}$ is the heat capacity per particle.

Key result: Kinetic-theory transport coefficients

All three transport coefficients share the form (transport coefficient) $= (1/3) (quantity transported per particle) (n) (\overset{v}{ˉ}) (l)$ :

D = \frac{v ˉ l}{3}, η = \frac{1}{3} mn \overset{v}{ˉ} l, κ = \frac{1}{3} c_{V}^{(1)} n \overset{v}{ˉ} l .

These are estimates accurate to within a factor of 2–3. The exact values require solving the Boltzmann equation 11.02.03 and differ by numerical prefactors (e.g., $D = 3 \overset{v}{ˉ} l /16$ for hard spheres from the Chapman-Enskog solution).

Key derivation: Mean free path from collision geometry Intermediate+

Proposition. For a gas of hard spheres of diameter $σ$ at number density $n$ , the mean free path is $l = 1/ (2 nπ σ^{2})$ .

Proof. Consider one molecule — the "test" particle — moving with speed $v$ through a sea of stationary scatterers. In time $d t$ , the test particle sweeps out a cylinder of cross-sectional area $π (2 σ)^{2} = 4 π σ^{2}$ (because two spheres of diameter $σ$ collide when their centres are within $2 σ$ ) and length $v d t$ . The number of scatterers in this cylinder is $n \cdot 4 π σ^{2} \cdot v d t$ .

But the scatterers are not stationary. The effective relative speed between the test particle and a scatterer is $\overset{v}{ˉ}_{rel} = 2 \overset{v}{ˉ}$ (the factor $2$ comes from averaging over all directions of the scatterer velocity). Replacing $v$ by $\overset{v}{ˉ}_{rel}$ and averaging: the collision rate is $ν = n \cdot 4 π σ^{2} \cdot 2 \overset{v}{ˉ} /4 = 2 nπ σ^{2} \overset{v}{ˉ}$ .

Wait — more carefully: the cross-section for two particles of diameter $σ$ is $π (σ + σ)^{2} /4 = π σ^{2}$ ... No. The cross-section for two spheres of diameter $σ$ is $π d^{2}$ where $d = 2 σ$ is the range of interaction. Actually, the total cross-section is $π (2 σ)^{2} = 4 π σ^{2}$ , but in kinetic theory the convention is to use the collision cross-section $π σ^{2}$ where $σ$ is the distance of closest approach between centres.

Let me redo this. Two molecules of diameter $σ$ collide when their centres come within distance $σ$ of each other (if $σ$ is the diameter, the centres must be within $2 σ$ ; if $σ$ is the radius, within $2 σ$ ). Following the convention where $σ$ is the molecular diameter:

Cross-section for collision: $b_{max} = σ$ (impact parameter), so $Σ = π σ^{2}$ .

Hmm, this depends on convention. Let me use $σ$ = diameter.

Collision occurs when centres approach within $d = σ$ . Cross-section: $π d^{2} = π σ^{2}$ .

Collision rate: $ν = n \overset{v}{ˉ}_{rel} π σ^{2} = 2 n \overset{v}{ˉ} π σ^{2}$ .

Mean free path: $l = \overset{v}{ˉ} / ν = 1/ (2 nπ σ^{2})$ . $□$

Bridge. The mean-free-path derivation builds toward the Boltzmann transport equation 11.02.03, where the collision rate $ν = \overset{v}{ˉ} / l$ becomes the collision operator acting on the distribution function. This is exactly the connection between microscopic collision geometry and macroscopic transport: the transport coefficients $D$ , $η$ , $κ$ are all proportional to $l$ because longer free paths carry gradients further before equilibrating. The foundational reason transport exists at all is that the mean free path is finite — if collisions were instantaneous ( $l = 0$ ) the gas would be in local equilibrium everywhere and there would be no transport, which is dual to the statement that an infinitely collisional fluid has zero viscosity, a paradox resolved by the finite- $l$ kinetic theory.

Exercises Intermediate+

Exercise 3 (medium, symbolic).

Show that the Prandtl number $Pr = η c_{P} / κ$ (ratio of viscous to thermal diffusivity) is approximately unity for a monatomic ideal gas using the kinetic-theory transport coefficients.

Hint

Substitute $η = mn \overset{v}{ˉ} l /3$ and $κ = c_{V}^{(1)} n \overset{v}{ˉ} l /3$ and use $c_{P} = c_{V} + k$ .

Answer

$Pr = \frac{η c _{P}}{κ} = \frac{( mn v ˉ l /3 ) ( c _{V}^{(1)} + k )}{( c _{V}^{(1)} n v ˉ l /3 )} = \frac{m ( c _{V}^{(1)} + k )}{c _{V}^{(1)}}$ .

Wait, that's not right. Let me redo with per-unit-mass quantities.

$η = ρ \overset{v}{ˉ} l /3$ where $ρ = mn$ . $κ = n c_{V}^{(1)} \overset{v}{ˉ} l /3$ . So $κ = ρ c_{V} \overset{v}{ˉ} l /3$ where $c_{V} = c_{V}^{(1)} / m$ is the specific heat per unit mass. Then $Pr = (ρ \overset{v}{ˉ} l /3 \cdot c_{P}) / (ρ c_{V} \overset{v}{ˉ} l /3) = c_{P} / c_{V} = γ$ . For a monatomic gas, $γ = 5/3$ , so $Pr = 5/3 \approx 1.67$ . The experimental value for argon is Pr $\approx 0.67$ . The discrepancy comes from the crude $1/3$ factor; the exact Chapman-Enskog result gives Pr $= 2/3$ for hard spheres.

Exercise 6 (hard, symbolic).

Derive the Stokes-Einstein relation $D = k_{B} T / (6 π η R)$ for the diffusion constant of a sphere of radius $R$ in a fluid of viscosity $η$ . Use dimensional analysis and the requirement that $D$ has dimensions of length $^{2}$ /time.

Hint

The drag force on a sphere at low velocity in a viscous fluid is $F = 6 π η R v$ (Stokes' law). The diffusion constant and mobility are related by the Einstein relation $D = μ k_{B} T$ where $μ = v / F$ is the mobility.

Answer

The mobility of the sphere is $μ = v / F = 1/ (6 π η R)$ from Stokes' law. The Einstein relation connects mobility to diffusion: $D = μ k_{B} T = k_{B} T / (6 π η R)$ . This is the Stokes-Einstein relation.

Dimensional check: $k_{B} T$ has units of energy (J = kg m $^{2}$ /s $^{2}$ ), $η$ has units of Pa s = kg/(m s), $R$ has units of m. So $k_{B} T / (η R)$ has units of (kg m $^{2}$ /s $^{2}$ )/((kg/(m s)) m) = m $^{2}$ /s, which are the correct units for $D$ .

The Stokes-Einstein relation is valid for $R ≫ l$ (the sphere is much larger than the mean free path), where the fluid can be treated as a continuum. For $R \sim l$ the Cunningham correction factor applies.

Exercise 7 (hard, symbolic).

The Wiedemann-Franz law states that the ratio of thermal conductivity to electrical conductivity in a metal is $κ / σ = L T$ where $L = (π^{2} /3) (k_{B}^{2} / e^{2})$ is the Lorenz number. Derive this from the kinetic-theory expressions for $κ$ and $σ$ using the Drude model.

Hint

In the Drude model, the electrical conductivity is $σ = n e^{2} τ / m$ and the thermal conductivity is $κ = n k_{B}^{2} T τ / (m)$ (using the electronic specific heat per electron $c_{V}^{(e)} = π^{2} k_{B}^{2} T / (2 ϵ_{F})$ from the Sommerfeld expansion).

σ = n e^{2} τ / m

where

τ

is the relaxation time. For thermal conductivity, use

κ = (1/3) C_{e} v_{F} l = (1/3) C_{e} v_{F}^{2} τ

, where

C_{e}

is the electronic specific heat per volume. With

C_{e} = π^{2} n k_{B}^{2} T / (2 ϵ_{F})

and

ϵ_{F} = m v_{F}^{2} /2

$κ = \frac{1}{3} \frac{π ^{2} n k _{B}^{2} T}{2 ( m v _{F}^{2} /2 )} v_{F}^{2} τ = \frac{π ^{2} n k _{B}^{2} T τ}{3 m}$ .

Taking the ratio:

$\frac{κ}{σ} = \frac{π ^{2} n k _{B}^{2} T τ / ( 3 m )}{n e ^{2} τ / m} = \frac{π ^{2} k _{B}^{2} T}{3 e ^{2}} = L T$ .

So $L = π^{2} k_{B}^{2} / (3 e^{2}) = 2.44 \times 1 0^{- 8}$ W $Ω$ /K $^{2}$ , in good agreement with experiment for most metals at room temperature.

Exercise 8 (hard, symbolic).

Show that the Green-Kubo relation for the self-diffusion constant is $D = (1/3) \int_{0}^{\infty} ⟨ v (0) \cdot v (t)⟩ d t$ where the angle brackets denote an equilibrium average. Relate this to the mean-free-path formula $D = \overset{v}{ˉ} l /3$ .

Hint

The mean-square displacement is $⟨ ∣ r (t) - r (0) ∣^{2} ⟩ = 6 D t$ for a diffusing particle. Express the displacement as a time integral of the velocity and take the derivative.

Answer

Write $r (t) - r (0) = \int_{0}^{t} v (t^{'}) d t^{'}$ . Then

$⟨ ∣ r (t) - r (0) ∣^{2} ⟩ = \int_{0}^{t} d t^{'} \int_{0}^{t} d t^{''} ⟨ v (t^{'}) \cdot v (t^{''})⟩ = 2 \int_{0}^{t} d t^{'} \int_{0}^{t^{'}} d t^{''} ⟨ v (0) \cdot v (t^{''} - t^{'})⟩$ .

For an equilibrium system, the velocity autocorrelation depends only on the time difference: $⟨ v (t^{'}) \cdot v (t^{''})⟩ = ⟨ v (0) \cdot v (∣ t^{''} - t^{'} ∣)⟩$ . Changing variables and taking the long-time limit:

$\frac{d}{d t} ⟨ ∣ r (t) - r (0) ∣^{2} ⟩ = 6 D = 2 \int_{0}^{\infty} ⟨ v (0) \cdot v (t)⟩ d t$ .

Hence $D = (1/3) \int_{0}^{\infty} ⟨ v (0) \cdot v (t)⟩ d t$ .

For the mean-free-path model, the velocity autocorrelation decays exponentially: $⟨ v (0) \cdot v (t)⟩ = \overset{v}{ˉ}^{2} e^{- t / τ}$ where $τ = l / \overset{v}{ˉ}$ . Then $D = (1/3) \overset{v}{ˉ}^{2} τ = \overset{v}{ˉ}^{2} (l / \overset{v}{ˉ}) /3 = \overset{v}{ˉ} l /3$ , recovering the kinetic-theory formula.

Lean formalization Intermediate+

Transport coefficients have no formalization in Mathlib. The Green-Kubo relations are time-correlation integrals of the form $\int_{0}^{\infty} ⟨ A (0) A (t)⟩ d t$ , where $A$ is a flux observable. Formalising these requires a framework for equilibrium time-correlation functions in a Hamiltonian (or stochastic) dynamical system — a significant infrastructure investment beyond Mathlib's current scope.

The mean-free-path derivation is geometric: given $n$ randomly placed hard spheres per unit volume, compute the mean distance between collisions. This is a Poisson-process argument that could be formalised in Mathlib's probability theory framework, but the physical content (identifying $D = \overset{v}{ˉ} l /3$ from a random-walk model) requires connecting the geometric calculation to the continuum diffusion equation.

Advanced results Master

Chapman-Enskog theory

The kinetic-theory estimates $D = \overset{v}{ˉ} l /3$ , $η = ρ \overset{v}{ˉ} l /3$ , $κ = c_{V} \overset{v}{ˉ} l /3$ are accurate to about a factor of 2. The Chapman-Enskog solution of the Boltzmann equation improves these by systematically expanding the distribution function in powers of the gradient (the Knudsen number $Kn = l / L$ where $L$ is the system size).

For hard spheres, the Chapman-Enskog first-order results are:

D = \frac{3}{8} \frac{k T / ( π m )}{n σ ^{2}}, η = \frac{5}{16} \frac{π mk T}{π σ ^{2}}, κ = \frac{75}{64} \frac{k k T / ( π m )}{π σ ^{2}} .

These differ from the mean-free-path estimates by numerical prefactors of order unity. The second-order (Burnett) corrections are small for $Kn ≪ 1$ but become important for rarefied gas flows (e.g., high-altitude aerodynamics).

Green-Kubo relations

The Green-Kubo relations express transport coefficients as equilibrium time-correlation functions:

D = \frac{1}{3} \int_{0}^{\infty} ⟨ v (0) \cdot v (t)⟩ d t, η = \frac{V}{3 k _{B} T} \int_{0}^{\infty} ⟨ σ_{x y} (0) σ_{x y} (t)⟩ d t,

κ = \frac{V}{3 k _{B} T ^{2}} \int_{0}^{\infty} ⟨ J_{Q}^{x} (0) J_{Q}^{x} (t)⟩ d t .

These are exact results from linear response theory 11.02.04, valid for any system (not just dilute gases). They are the basis for computing transport coefficients in molecular dynamics simulations.

Knudsen number and the limits of continuum hydrodynamics

The Knudsen number $Kn = l / L$ measures the ratio of the mean free path to the system size. The transport-coefficient formalism (and the Navier-Stokes equations it feeds into) requires $Kn ≪ 1$ — many collisions per macroscopic length scale. When $Kn \sim 1$ (rarefied gas, micro-electromechanical systems), the continuum description breaks down and the full Boltzmann equation must be solved. When $Kn ≫ 1$ (free molecular flow), particle-surface interactions dominate and the mean free path is irrelevant.

Synthesis. Transport phenomena build toward the fluctuation-dissipation theorem 11.02.04, which provides the general framework connecting dissipation (transport coefficients) to equilibrium fluctuations (correlation functions). The central insight is that the same molecular collisions that equilibrate the gas also produce the random forces that cause diffusion — these are two faces of the same microscopic process. This is dual to the Onsager reciprocal relations, which connect different transport processes through the symmetry of the underlying correlations. The foundational reason transport coefficients are proportional to the mean free path is that $l$ sets the distance over which a molecule carries information about local conditions; putting these together with the Green-Kubo relations gives the exact answer in terms of time-correlation functions, generalising the mean-free-path picture to arbitrary density and interaction potential.

Full proof set Master

Proposition. For a hard-sphere gas, the Prandtl number from Chapman-Enskog theory is Pr $= 2/3$ .

Proof. The Chapman-Enskog first-order results for hard spheres give $η = (5/16) π mk T / (π σ^{2})$ and $κ = (75/64) k k T / (π m) / (π σ^{2})$ . The Prandtl number is

Pr = \frac{η c _{P}}{κ M}

where $c_{P} / M = c_{P}^{(molar)} / M_{molar}$ is the specific heat per unit mass. For a monatomic ideal gas, $c_{P} = (5/2) k$ per particle. So

Pr = \frac{( 5/16 ) π mk T / ( π σ ^{2} ) \cdot ( 5/2 ) k}{( 75/64 ) k k T / ( π m ) / ( π σ ^{2} )} = \frac{( 5/16 ) ( 5/2 )}{( 75/64 )} = \frac{25/32}{75/64} = \frac{25 \cdot 64}{32 \cdot 75} = \frac{2}{3} .

The Prandtl number Pr $= 2/3$ for a monatomic hard-sphere gas is exact to first order in the Chapman-Enskog expansion. Experimental values for noble gases range from 0.66 (argon) to 0.67 (xenon), in excellent agreement. $□$

Connections Master

11.02.01 The Maxwell-Boltzmann speed distribution provides $\overset{v}{ˉ} = 8 k T / (π m)$ , which enters every transport-coefficient formula.
11.02.03 The Boltzmann transport equation is the systematic framework that reproduces the mean-free-path transport coefficients and provides the exact numerical prefactors via the Chapman-Enskog expansion.
11.02.04 The fluctuation-dissipation theorem connects the transport coefficients derived here to equilibrium time-correlation functions through the Green-Kubo relations.
11.04.04 The Sackur-Tetrode entropy of the ideal gas, derived from the classical partition function, provides the thermodynamic background for the kinetic-theory treatment.
37.01.01 The probability foundations (random walks, Poisson processes) underlie the mean-free-path calculation and the diffusion equation.

Historical and philosophical context Master

The kinetic theory of transport phenomena was developed in the 1860s by Maxwell and Clausius. Maxwell (1860) derived the viscosity formula $η \propto T$ and was surprised to find that gas viscosity is independent of density — a prediction so counter-intuitive that he built an experimental apparatus to test it, confirming the result.

The mean-free-path approach was refined by Chapman (1916–17) and Enskog (1917) independently, both solving the Boltzmann equation to obtain accurate transport coefficients. The Chapman-Enskog method remains the standard approach for computing transport properties of gases.

The Green-Kubo relations (Green 1954, Kubo 1957) provided the modern linear-response framework, expressing transport coefficients as equilibrium time-correlation functions. This shifted the computational problem from solving the Boltzmann equation to measuring equilibrium fluctuations — a change in perspective that underlies modern molecular dynamics methods for computing transport properties.

Bibliography Master

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

@book{kittel-kroemer1980,
  author = {Kittel, Charles and Kroemer, Herbert},
  title = {Thermal Physics},
  edition = {2nd},
  publisher = {Freeman},
  year = {1980}
}

@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}
}

@article{kubo1957,
  author = {Kubo, Ryogo},
  title = {Statistical-Mechanical Theory of Irreversible Processes. {I.} General Theory and Simple Applications to Magnetic and Conduction Problems},
  journal = {J. Phys. Soc. Jpn.},
  volume = {12},
  pages = {570--586},
  year = {1957}
}

Prerequisites

11.02.01

Used in

11.02.03

Tier anchors

beginner: Schroeder, An Introduction to Thermal Physics (2000), Ch. 1 (transport processes); Reif, Fundamentals of Statistical and Thermal Physics (1965), Ch. 12
intermediate: Reif, Fundamentals of Statistical and Thermal Physics (1965), Ch. 12–13; Kittel & Kroemer, Thermal Physics, 2nd ed. (1980), Ch. 14
master: Pathria & Beale, Statistical Mechanics, 3rd ed. (2011), Ch. 11; Kardar, Statistical Physics of Particles (2007), Ch. 5

References

Reif — Fundamentals of Statistical and Thermal Physics (McGraw-Hill, 1965) · Ch. 12 Transport phenomena. Fick law of diffusion J = -D grad n, Newton law of viscosity F/A = -eta dv/dy, Fourier law of heat conduction Q = -kappa dT/dx. Mean free path l = 1/(n*sqrt(2)*sigma) and its connection to transport coefficients D = lv/3, eta = mnvl/3, kappa = C_v vl/3.
Kittel & Kroemer — Thermal Physics, 2nd ed. (Freeman, 1980) · Ch. 14 Kinetic theory. Transport coefficients from mean free path arguments. The Wiedemann-Franz law kappa/sigma_T = L T.
Pathria & Beale — Statistical Mechanics, 3rd ed. (Academic Press, 2011) · Ch. 11 Statistical mechanics of interacting systems: the method of correlation functions. Transport coefficients from the Boltzmann equation and Green-Kubo relations.
Kardar — Statistical Physics of Particles (Cambridge, 2007) · Ch. 5 Kinetic theory. The Boltzmann equation, H-theorem, and transport coefficients from Chapman-Enskog expansion.
Chapman & Cowling — The Mathematical Theory of Non-Uniform Gases, 3rd ed. (Cambridge, 1970) · Ch. 7–8 The second approximation to the distribution function, transport coefficients from the Chapman-Enskog solution of the Boltzmann equation.

Estimated time

beginner: 15 min
intermediate: 35 min
master: 65 min