Transport Properties of Gases: General Framework

Rub your hands together and they warm up. Leave a drop of perfume in a room and the scent spreads. Stir smoke into still air and the vortex slowly dies. These are three faces of one idea: gradients drive flows. Whenever a gas has a nonuniform property, temperature, composition, or velocity, molecular motion carries that quantity from where it is abundant to where it is scarce. This transport happens because molecules fly freely between collisions, carrying momentum, mass, or energy from their last collision site into a new region. Kinetic theory converts this molecular picture into precise laws relating a flux to a gradient. The proportionality constants in those laws, viscosity, diffusion coefficient, thermal conductivity, are called transport coefficients. This article builds the single unifying framework beneath all gas transport phenomena. You will see how flux-gradient relations arise from mean free path arguments. Then you will derive a master formula for a generic transport coefficient. Finally you will see how this one formula specializes into viscosity, diffusion, and thermal conductivity, the subjects of the next three articles.

Key takeaways

Gradients in velocity, concentration, or temperature drive fluxes of momentum, matter, or energy respectively.
Each flux obeys a linear law: flux equals minus a transport coefficient times the gradient.
All three coefficients share one kinetic-theory skeleton built from mean speed, mean free path, and number density.
The simple mean-free-path derivation gives the correct functional form but coefficients off by numerical factors of order unity.
Collision cross-section and mean free path, covered earlier, are the microscopic inputs that set every transport coefficient's magnitude.
Key idea: transport properties measure how efficiently random molecular motion smooths out gradients in velocity, density, or temperature.

The core insight

Every gas transport law has the same architecture: a flux proportional to a gradient, with the proportionality constant built from the same three microscopic ingredients, mean molecular speed, mean free path, and number density, because molecules ferry whatever property they carry across one mean free path before colliding and equilibrating.

The physical picture

Consider a gas that is not in equilibrium: it has a gradient in some macroscopic property along one direction, say $z$ . That property might be the flow velocity $u_x(z)$ (giving viscosity), the number density of a labeled species $n(z)$ (giving diffusion), or the temperature $T(z)$ (giving thermal conductivity). In each case, molecules crossing an imaginary plane at height $z$ carry with them the value of that property characteristic of the location where they last collided, roughly one mean free path away. Picture a plane at $z = 0$ . Molecules arrive from below, from $z = -\lambda$ , carrying the property value appropriate to that layer. Molecules arrive from above, from $z = +\lambda$ , carrying the property value there. Because the flux is symmetric in number, equal numbers cross upward and downward, but the property they carry differs. If the gradient is positive, molecules coming from above carry more of the property than those coming from below, so there is a net downward transport of that property, opposite to the gradient. This one argument, repeated three times with three different "properties," generates viscosity, diffusion, and thermal conductivity as three shadows of a single mechanism. The mean free path $\lambda$ and collision cross-section $\sigma$ , established in the kinetic theory of collisions, set the length scale over which a molecule "remembers" its point of origin. A short mean free path (high pressure, large molecules) means transport is sluggish and local; a long mean free path (low pressure, small molecules) means transport is efficient over the accessible distance, though fewer molecules exist to carry it. The competition between these effects is what makes viscosity pressure-independent, a famous and easily misunderstood prediction.

Figure

Three gradients, one mechanism: flux carries a property down the gradient

Figure

Generic transport coefficient built from three ingredients

The core relations

J_z(X) = -\Gamma \, \frac{dX}{dz}

where: $J_z(X)$ is the flux of property $X$ in the $z$ -direction, $\Gamma$ is the relevant transport coefficient, and $dX/dz$ is the gradient driving the flow.

\Gamma \approx \frac{1}{3}\, n \, \bar{c} \, \lambda \times (\text{property carried per molecule, per unit gradient})

where: $n$ is number density, $\bar{c}$ is the mean molecular speed, and $\lambda$ is the mean free path.

\bar{c} = \sqrt{\frac{8RT}{\pi M}}, \qquad \lambda = \frac{1}{\sqrt{2}\, n \sigma}

where: $M$ is molar mass, $R$ the gas constant, and $\sigma$ the collision cross-section.

J_z(\text{momentum}) = -\eta \frac{du_x}{dz}, \quad J_z(\text{matter}) = -D \frac{dn}{dz}, \quad J_z(\text{energy}) = -\kappa \frac{dT}{dz}

where: $\eta$ is viscosity, $D$ the diffusion coefficient, and $\kappa$ the thermal conductivity, each a special case of $\Gamma$ .

Derivation AThe mean free path flux argument: a master formula for Γ

Goal. Derive a general expression for the transport coefficient $\Gamma$ associated with the flux of any molecular property $X(z)$ down a gradient $dX/dz$ , using the simple mean-free-path model.

Step 1: Set up the flux-counting geometry. Consider a plane at height $z_0$ perpendicular to the transport direction, and consider molecules crossing it. In equilibrium, the number of molecules crossing per unit area per unit time in one direction (say upward) is one-quarter of the total molecular flux from kinetic theory,

\Phi = \frac{1}{4} n \bar{c}.

This factor of one-quarter comes from averaging $\cos\theta$ over the hemisphere of velocity directions crossing the plane and is a standard result carried over from the kinetic theory of effusion and collision rates.

Step 2: Assign an origin to each crossing molecule. In the crude but instructive mean-free-path model, assume every molecule crossing the plane last collided exactly one mean free path away, at $z_0 - \lambda$ if moving upward, or $z_0 + \lambda$ if moving downward. At that last collision, the molecule equilibrated to the local value of the property, $X(z_0 - \lambda)$ or $X(z_0 + \lambda)$ respectively. This is the key assumption: it treats the mean free path as the memory length over which a molecule "forgets" where it came from.

Step 3: Write the two one-way fluxes of property $X$ . The upward flux of $X$ carried through the plane is the crossing rate times the property value at the origin of those molecules,

J_z^{\uparrow}(X) = \frac{1}{4} n \bar{c}\, X(z_0 - \lambda).

The downward flux is

J_z^{\downarrow}(X) = \frac{1}{4} n \bar{c}\, X(z_0 + \lambda).

Step 4: Combine into a net flux. Net transport in the $+z$ direction is upward minus downward,

J_z(X) = \frac{1}{4} n \bar{c} \Big[ X(z_0 - \lambda) - X(z_0 + \lambda) \Big].

Step 5: Linearize for a slowly varying gradient. Since $\lambda$ is microscopically small compared to the macroscopic length scale over which $X$ varies, Taylor-expand,

X(z_0 \pm \lambda) \approx X(z_0) \pm \lambda \frac{dX}{dz}\bigg|_{z_0}.

Substituting into Step 4,

J_z(X) = \frac{1}{4} n \bar{c} \left[ -2\lambda \frac{dX}{dz} \right] = -\frac{1}{2} n \bar{c} \lambda \frac{dX}{dz}.

Step 6: Correct for the crude "one mean free path exactly" assumption. Real molecules do not all travel exactly $\lambda$ before their last collision; the free-path lengths are exponentially distributed, and the geometry of crossing at an angle modifies the effective distance. A more careful kinetic-theory treatment, averaging properly over the distribution of free paths and crossing angles, replaces the numerical prefactor $1/2$ with $1/3$ . This is the standard correction quoted in Chapman-Enskog-level treatments and reproduced in most textbook derivations, giving

\boxed{J_z(X) = -\frac{1}{3} n \bar{c} \lambda \frac{dX}{dz}}

Step 7: Read off the transport coefficient. Comparing to the generic law $J_z(X) = -\Gamma\, dX/dz$ identifies

\Gamma = \frac{1}{3} n \bar{c} \lambda

as the universal skeleton. To get viscosity, $X$ is replaced by momentum per molecule, $m u_x(z)$ ; to get diffusion, $X$ is the number density itself (with an extra factor since $D$ multiplies $dn/dz$ directly rather than $d(nX)/dz$ ); to get thermal conductivity, $X$ is replaced by molecular energy $c_V T / N_A$ . Each substitution is carried out in detail in the dedicated articles that follow.

Worked examples

Example 1 — Estimating the master transport prefactor for argon at 298 K, 1 bar. Argon has molar mass $M = 39.95 \times 10^{-3}\ \text{kg mol}^{-1}$ and collision cross-section $\sigma \approx 3.6 \times 10^{-19}\ \text{m}^2$ (diameter $\approx 340$ pm). Step 1: mean speed.

\bar{c} = \sqrt{\frac{8RT}{\pi M}} = \sqrt{\frac{8 (8.314\ \text{J mol}^{-1}\text{K}^{-1})(298\ \text{K})}{\pi (39.95 \times 10^{-3}\ \text{kg mol}^{-1})}}

= \sqrt{\frac{19{,}821\ \text{J mol}^{-1}}{0.1255\ \text{kg mol}^{-1}}} = \sqrt{157{,}900\ \text{m}^2\text{s}^{-2}} \approx 397\ \text{m s}^{-1}.

Step 2: number density at 1 bar.

n = \frac{P}{k_B T} = \frac{1.013 \times 10^5\ \text{Pa}}{(1.381 \times 10^{-23}\ \text{J K}^{-1})(298\ \text{K})} \approx 2.46 \times 10^{25}\ \text{m}^{-3}.

Step 3: mean free path.

\lambda = \frac{1}{\sqrt{2}\, n \sigma} = \frac{1}{\sqrt{2}\,(2.46\times 10^{25}\ \text{m}^{-3})(3.6\times 10^{-19}\ \text{m}^2)}

= \frac{1}{1.251 \times 10^{7}\ \text{m}^{-1}} \approx 8.0 \times 10^{-8}\ \text{m} = 80\ \text{nm}.

Step 4: master prefactor.

\Gamma_0 \equiv \frac{1}{3} n \bar{c} \lambda = \frac{1}{3}(2.46 \times 10^{25}\ \text{m}^{-3})(397\ \text{m s}^{-1})(8.0 \times 10^{-8}\ \text{m})

= \frac{1}{3}(7.82 \times 10^{20}\ \text{m}^{-2}\text{s}^{-1}) \approx 2.6 \times 10^{20}\ \text{m}^{-2}\text{s}^{-1}.

Sanity check. This quantity, $n\bar{c}\lambda/3$ , has units of (number density)(speed)(length) = number flux per unit gradient (m $^{-2}$ s $^{-1}$ ), exactly what is needed to multiply a gradient in a dimensionless or per-molecule property to yield a flux. It is this single number, of order $10^{20}$ m $^{-2}$ s $^{-1}$ for a typical gas at room conditions, that seeds the numerical value of $\eta$ , $D$ , and $\kappa$ once the specific carried property is attached in the follow-on articles. Note also that $n\lambda = 1/(\sqrt2\,\sigma) \approx 1.96\times10^{24}\ \text{m}^{-2}$ is pressure-independent, foreshadowing why $\eta$ turns out independent of pressure.

Example 2 — Comparing the transport-relevant length and time scales for He and Xe at the same T and P, and predicting which diffuses faster. Take $T = 300$ K, $P = 1$ atm $= 1.013 \times 10^5$ Pa for both gases. Use $\sigma_{\text{He}} = 0.21 \times 10^{-19}\ \text{m}^2$ and $\sigma_{\text{Xe}} = 6.0 \times 10^{-19}\ \text{m}^2$ (approximate kinetic cross-sections), $M_{\text{He}} = 4.00 \times 10^{-3}$ kg mol $^{-1}$ , $M_{\text{Xe}} = 131.3 \times 10^{-3}$ kg mol $^{-1}$ . Number density (same for both, since same $T,P$ ):

n = \frac{P}{k_B T} = \frac{1.013\times10^5}{(1.381\times10^{-23})(300)} \approx 2.44 \times 10^{25}\ \text{m}^{-3}.

Mean speeds:

\bar{c}_{\text{He}} = \sqrt{\frac{8(8.314)(300)}{\pi(4.00\times10^{-3})}} = \sqrt{1.588 \times 10^{6}} \approx 1260\ \text{m s}^{-1}.

\bar{c}_{\text{Xe}} = \sqrt{\frac{8(8.314)(300)}{\pi(131.3\times10^{-3})}} = \sqrt{4.836\times10^{4}} \approx 220\ \text{m s}^{-1}.

Mean free paths:

\lambda_{\text{He}} = \frac{1}{\sqrt{2}\,(2.44\times10^{25})(0.21\times10^{-19})} \approx \frac{1}{7.24\times10^{5}} \approx 1.38\times10^{-6}\ \text{m}.

\lambda_{\text{Xe}} = \frac{1}{\sqrt{2}\,(2.44\times10^{25})(6.0\times10^{-19})} \approx \frac{1}{2.07\times10^{7}} \approx 4.8\times10^{-8}\ \text{m}.

Master prefactors:

\Gamma_{0,\text{He}} = \frac{1}{3} n \bar{c}_{\text{He}} \lambda_{\text{He}} = \frac{1}{3}(2.44\times10^{25})(1260)(1.38\times10^{-6})

\approx \frac{1}{3}(4.24\times10^{22}) \approx 1.41 \times 10^{22}\ \text{m}^{-2}\text{s}^{-1}.

\Gamma_{0,\text{Xe}} = \frac{1}{3}(2.44\times10^{25})(220)(4.8\times10^{-8}) \approx \frac{1}{3}(2.58\times10^{20}) \approx 8.6\times10^{19}\ \text{m}^{-2}\text{s}^{-1}.

Interpretation. $\Gamma_{0,\text{He}} / \Gamma_{0,\text{Xe}} \approx 164$ . Helium's small size (tiny $\sigma$ ) and light mass (large $\bar c$ ) both push in the same direction, giving it a vastly larger transport prefactor. This correctly predicts that helium diffuses, conducts heat, and transmits momentum far more readily than xenon under identical conditions, consistent with helium's well-known use as a fast-diffusing tracer gas and superior thermal conductor among noble gases.

Limiting cases & common traps

Situation	What happens to Γ	Why
Pressure increases at fixed T	$n$ up, $\lambda$ down ( $\lambda \propto 1/n$ )	For $\eta$ and $\kappa$ , these cancel exactly in this simple model: pressure-independent. For $D$ , $n$ does not cancel, so $D \propto 1/n$ .
Temperature increases at fixed $n$	$\bar c$ up, $\lambda$ unchanged	$\eta$ and $\kappa$ increase roughly as $\sqrt T$ (more, in real gases, due to $\sigma(T)$ effects); $D$ increases similarly at fixed $n$ .
Larger collision cross-section $\sigma$	$\lambda$ down	Bulkier or more attractive molecules transport less efficiently per collision-free flight, though they also collide, and transfer property, more often.
Near a solid boundary or at very low pressure	Mean-free-path picture breaks down	When $\lambda$ approaches the container dimension (Knudsen regime), transport becomes ballistic, not diffusive, and these bulk laws no longer apply.

Common trapThe pressure-independence of viscosity is not a universal law of all transport coefficients

Students often overgeneralize that "transport coefficients don't depend on pressure." This cancellation of $n$ against $\lambda \propto 1/n$ occurs specifically for $\eta$ and $\kappa$ , because their generic factor $n\bar c\lambda$ multiplies a property (momentum or energy per molecule) that is itself pressure-independent. Diffusion is different: $D$ is conventionally defined so that it equals $\bar c \lambda /3$ times a numerical factor, without an extra $n$ , precisely so that Fick's law $J = -D\, dn/dz$ correctly carries the pressure dependence in the gradient itself; consequently $D \propto 1/n \propto 1/P$ at fixed temperature. Keep the three coefficients' pressure dependences straight rather than assuming one universal rule.

The chemist behind it

The mean free path approach to transport coefficients traces back to the kinetic theorists of the mid-nineteenth century, but it was James Clerk Maxwell who first derived, in 1860, the striking and counterintuitive prediction that gas viscosity is independent of pressure, a direct consequence of the $n\lambda$ cancellation explored above. Maxwell tested this experimentally with rotating-disk apparatus and confirmed it, a result so contrary to liquid intuition (where thicker, denser fluids are more viscous) that it became one of the most celebrated triumphs of early kinetic theory. Rudolf Clausius had already introduced the mean free path concept a few years earlier, and Ludwig Boltzmann later placed the whole framework on a rigorous statistical foundation with his transport equation, refining the crude factor-of-one-third estimates used here into the precise Chapman-Enskog coefficients used in modern gas-property tables.

Origins · 1 minJames Clerk Maxwell and how this was discovered

James Clerk Maxwell

1831-1879 · derived gas viscosity's pressure independence from kinetic theory and verified it experimentally

Ask the tutor: Want to see this same $n\bar c\lambda/3$ skeleton turned into the exact formula for gas viscosity, including the momentum-transfer bookkeeping?

Why do viscosity, diffusion, and thermal conductivity all get discussed together as one framework?

Because all three arise from the same underlying mechanism: molecules travel roughly one mean free path between collisions, carrying whatever property, momentum, particle identity, or kinetic energy, characterized their last collision site. The flux of that property across any plane is then proportional to the property's gradient, with a coefficient built from number density, mean speed, and mean free path. Only the identity of the transported property differs between the three cases. Studying them as one framework first, then specializing, avoids re-deriving the same mean-free-path argument three separate times.

Why does the simple mean-free-path derivation give a numerical prefactor that is not exactly right?

The derivation assumes every crossing molecule traveled exactly one mean free path, arriving perpendicular to the plane, before its last collision. In reality, free path lengths are exponentially distributed around the mean, and molecules cross the plane at all angles, not just head-on. A rigorous treatment, solving the Boltzmann transport equation with the full velocity distribution, corrects the naive factor of one-half from a basic Taylor argument to about one-third, and introduces further species-dependent correction factors of order unity in the Chapman-Enskog theory. The functional dependence on $n$ , $\bar c$ , and $\lambda$ survives; only the prefactor changes.

How does this framework connect to collision cross-section and mean free path from earlier material?

Directly: $\lambda = 1/(\sqrt2\, n\sigma)$ is the mean free path derived from the collision cross-section $\sigma$ , and it appears as one of the three multiplicative ingredients in every transport coefficient here. A larger cross-section shortens the mean free path, which generally reduces transport efficiency per unit gradient, though it simultaneously increases the collision frequency that enables property exchange in the first place. All the quantitative work from the collision-theory topics feeds straight into the $\Gamma = n\bar c\lambda/3$ skeleton used throughout this article.

Does this simple framework work for liquids or only for gases?

This particular derivation is specific to dilute gases, where molecules travel in straight lines between well-separated, infrequent collisions and the mean free path is much larger than a molecular diameter. In liquids, molecules are in near-constant contact, there is no well-defined free flight, and transport instead proceeds by activated hopping between neighboring sites or by collective density fluctuations. Liquid viscosity typically decreases with temperature, opposite to the gas-phase trend, precisely because the transport mechanisms are fundamentally different.