Effusion and Molecular Flux

Punch a hole in the wall of a gas container — small enough that molecules leave one at a time without colliding in the aperture — and molecules stream out at a rate you can compute exactly from the velocity distribution. This is effusion, and the rate is the molecular flux $\Phi=\tfrac14 n\bar v$ : one quarter of the number density times the mean speed. The factor $\tfrac14$ follows from integrating the outward normal velocity component over the Maxwell-Boltzmann distribution of molecules moving toward the aperture. Effusion is deceptively rich. Because the molecular flux is proportional to the normal velocity, faster molecules are preferentially transmitted through the aperture. Consequently, the mean translational kinetic energy of the effusing beam is $2k_BT$ , compared with the equilibrium value $\tfrac32k_BT$ in the reservoir. Its speed distribution is $\propto v^3 e^{-mv^2/2k_BT}$ rather than the bulk $\propto v^2 e^{-mv^2/2k_BT}$ . This velocity bias is a real, measurable effect and the reason molecular beams are used as sources of energetic, well-collimated atoms. The rate's dependence on mass — $\Phi\propto1/\sqrt m$ at fixed temperature and pressure — is Graham's law of effusion: light molecules escape faster. This tiny difference, $\sqrt{m_2/m_1}$ per stage, was the physical basis of the gaseous-diffusion enrichment of uranium in the Manhattan Project, where $^{235}\mathrm{UF_6}$ and $^{238}\mathrm{UF_6}$ differ in rate by only $\sqrt{352/349}\approx1.0043$ — requiring thousands of cascaded stages. The same flux formula, rewritten with the ideal-gas law $P=nk_BT$ , becomes the Hertz-Knudsen equation $\Phi=P/\sqrt{2\pi m k_BT}$ , the workhorse of vacuum science, thin-film deposition, and the kinetic theory of evaporation and condensation.

Key takeaways

Effusion is the flow of gas through an aperture small compared to the mean free path, so molecules leave without colliding in the hole.
The molecular flux is $\Phi=\tfrac14 n\bar v$ , with the $\tfrac14$ arising from integrating $v_z$ over the outgoing hemisphere of the velocity distribution.
The escaping beam is velocity-weighted, with speed distribution proportional to $v^3e^{-mv^2/2k_BT}$ , giving a mean translational kinetic energy of $2k_BT$ instead of the equilibrium value $\tfrac32k_BT$ .
Graham's law: at fixed $T,P$ the rate scales as $1/\sqrt m$ , so light molecules effuse faster — the basis of isotope separation by gaseous diffusion.
In terms of pressure, $\Phi=P/\sqrt{2\pi m k_BT}$ (Hertz-Knudsen), the standard formula for vacuum leaks and deposition rates.
Key idea: because the flux through a hole weights molecules by their outward speed, effusion samples a $v$ -enhanced distribution — giving $\Phi=\tfrac14 n\bar v$ , a beam mean energy of $2k_BT$ , and Graham's $1/\sqrt m$ law.

The core insight

The number of molecules crossing a hole of area $A$ per unit time is $\Phi A=A\!\int_{v_z>0} v_z\,n f(\mathbf v)\,\mathrm d^3v=\tfrac14 n\bar v A$ . The extra factor of $v_z$ (relative to just counting molecules) both produces the $\tfrac14$ and biases the beam toward fast molecules, raising its mean energy to $2k_BT$ .

Where this lives today

The Hertz-Knudsen flux $\Phi=P/\sqrt{2\pi m k_BT}$ is the everyday equation of vacuum technology: it sets how fast a chamber contaminates through a leak, the monolayer formation time on a clean surface, and the deposition rate in molecular-beam epitaxy, where effusion cells ("Knudsen cells") deliver precisely controlled atomic fluxes to grow semiconductor layers one atomic plane at a time. The velocity-selective nature of effusive beams is exploited in atomic and molecular beam experiments and in loading magneto-optical traps. Graham's law also underlies gaseous isotope separation and contributes to Jeans escape, in which lighter gases such as hydrogen and helium escape planetary atmospheres more readily than heavier species. The same kinetic-theory framework also describes evaporation and sublimation rates, while the cosine emission law appears in thermal-radiation transport and sputtering geometries.

The physical picture

Imagine the molecules near the hole. In any short time $\mathrm dt$ , a molecule with outward velocity component $v_z>0$ will escape if it lies within a slab of thickness $v_z\,\mathrm dt$ in front of the hole. So the escape rate for a given velocity is proportional to $v_z$ itself — faster-outward molecules sweep out a bigger slab. Summing over the whole velocity distribution, only the outgoing half ( $v_z>0$ ) contributes, and the average of $v_z$ over that half turns out to be $\bar v/4$ . Hence $\Phi=\tfrac14 n\bar v$ . That same $v_z$ weighting is why the beam is not a fair sample of the gas. A slow molecule rarely reaches the hole in a given interval; a fast one reaches it often. The escaping population is therefore skewed toward high speed — the beam is "hotter" than the box it came from, carrying $2k_BT$ per molecule instead of $\tfrac32 k_BT$ . The residual gas, having lost its fastest molecules, is left very slightly cooler: effusion is a (tiny) cooling mechanism.

Figure

Molecules with outward velocity v_z escape if within a slab of thickness v_z dt of the hole. Faster-outward molecules sweep a bigger slab, so the flux weights each molecule by v_z — producing Φ = ¼ n v̄ and biasing the beam toward fast molecules.

The payoff figure is the beam distribution itself. The bulk gas has speed distribution $\propto v^2 e^{-mv^2/2k_BT}$ , peaking at the most probable speed $v_p$ . The beam speed distribution acquires an additional factor of $v$ , becoming proportional to $v^3e^{-mv^2/2k_BT}$ ; compared with the equilibrium distribution, it peaks at a higher speed and contains a larger fraction of fast molecules. Overlaying the two curves shows the beam shifted bodily to the right — a direct visual of "effusion samples the fast molecules."

Figure

Payoff: the bulk speed distribution ∝ v² e^(−mv²/2kT) versus the effusing-beam distribution ∝ v³ e^(−mv²/2kT). The extra factor of v shifts the beam to higher speed — the escaping molecules are systematically faster than the gas they left.

The core equations

The molecular flux is the number of molecules crossing unit area per unit time in the outward direction. It is built by weighting the number density $n f(\mathbf v)$ by the outward velocity component $v_z$ and integrating over the escaping hemisphere.

\Phi=\int_{v_z>0} v_z\,n\,f(\mathbf v)\,\mathrm d^3v=\frac14 n\bar v,\qquad \bar v=\sqrt{\frac{8k_BT}{\pi m}} \qquad (1)

where: $\Phi$ — molecular flux (molecules per unit area per unit time); $n$ — number density; $f(\mathbf v)$ — normalized Maxwell-Boltzmann velocity distribution; $\bar v$ — mean speed; the integral runs only over outward velocities $v_z>0$ . The factor $v_z$ both selects molecules moving toward the aperture and weights their contribution to the flux, leading to the factor $\tfrac14$ after integration over the Maxwell--Boltzmann distribution. Using the ideal-gas relation $P=nk_BT$ expresses the molecular flux in terms of the measurable pressure and temperature, yielding the Hertz--Knudsen form commonly used in vacuum science.

\Phi=\frac{P}{\sqrt{2\pi m k_BT}} \qquad (2)

where: $P$ — gas pressure; this is the Hertz-Knudsen equation. Reading: higher pressure and lower mass or temperature give a larger flux; it sets the impingement rate on any surface exposed to a gas. The escaping beam is velocity-biased. Weighting the bulk speed distribution by the extra factor $v$ gives the beam distribution and a raised mean energy.

f_{\rm beam}(v)\propto v^3\,e^{-mv^2/2k_BT},\qquad \langle E\rangle_{\rm beam}=2k_BT \qquad (3)

where: $f_{\rm beam}$ — normalized speed distribution of molecules in the effusing beam; $\langle E\rangle_{\rm beam}=2k_BT$ — beam mean kinetic energy, above the bulk $\tfrac32 k_BT$ . Reading: effusion preferentially samples fast molecules, so the beam is energetic relative to its source.

Derivation AThe flux Φ = ¼ n v̄ from the hemisphere integral

The goal is to derive the effusion flux by integrating the outward velocity component over the Maxwell-Boltzmann distribution across the escaping hemisphere, and to show cleanly where the characteristic factor of $\tfrac14$ comes from. This is the calculation that turns "gas leaks out of a hole" into a precise, mass- and temperature-dependent rate, and it is the foundation for Graham's law and the Hertz-Knudsen equation.

Step 1. A molecule with velocity $\mathbf v$ reaches the aperture (taken to lie in the $xy$ -plane) only if $v_z>0$ . During a time interval $\mathrm dt$ , the contributing molecules occupy a volume $A\,v_z\,\mathrm dt$ , giving a differential flux $v_z\,n f(\mathbf v)\,\mathrm d^3v$ .

Step 2. Factorize the velocity distribution, $f(\mathbf v)=\phi(v_x)\phi(v_y)\phi(v_z)$ with $\phi(v_i)=\sqrt{\tfrac{m}{2\pi k_BT}}e^{-mv_i^2/2k_BT}$ . The $v_x,v_y$ integrals run over all values and each gives $1$ (they are normalized), so only the $v_z$ integral over $v_z>0$ survives: $\Phi=n\displaystyle\int_0^\infty v_z\,\phi(v_z)\,\mathrm dv_z$ .

Step 3. Evaluate: $\displaystyle\int_0^\infty v_z\sqrt{\tfrac{m}{2\pi k_BT}}e^{-mv_z^2/2k_BT}\mathrm dv_z=\sqrt{\tfrac{m}{2\pi k_BT}}\cdot\tfrac{k_BT}{m}=\sqrt{\tfrac{k_BT}{2\pi m}}$ , using $\int_0^\infty x e^{-ax^2}\mathrm dx=1/(2a)$ with $a=m/2k_BT$ . So $\Phi=n\sqrt{k_BT/2\pi m}$ .

Step 4. Express the result in terms of the mean speed $\bar v=\sqrt{8k_BT/\pi m}$ . Using $\sqrt{k_BT/(2\pi m)}=\bar v/4$ gives $\Phi=\tfrac14n\bar v$ . The factor $\tfrac14$ reflects both the restriction to molecules with $v_z>0$ and the averaging over the normal velocity component.

Step 5. Substitute $P=nk_BT$ : $\Phi=\tfrac14\,(P/k_BT)\bar v=\tfrac14(P/k_BT)\sqrt{8k_BT/\pi m}=P/\sqrt{2\pi m k_BT}$ , the Hertz-Knudsen equation (2).

Go deeper · Derivation B · 1 minThe beam is hotter: mean energy 2kT and Graham's law

Derivation BThe beam is hotter: mean energy 2kT and Graham's law

The goal is to show that the effusing beam has a mean translational kinetic energy of $2k_BT$ , exceeding the equilibrium value $\tfrac32k_BT$ because the molecular flux is weighted by an additional factor of speed. The same velocity weighting also leads directly to Graham's law, $\Phi\propto1/\sqrt{m}$ .

Step 1. The equilibrium speed distribution is $f(v)\propto v^2e^{-\alpha v^2}$ . Because the molecular flux is proportional to speed, the beam distribution acquires an additional factor of $v$ , giving $f_{\rm beam}(v)\propto v^3e^{-\alpha v^2}$ .

Step 2. Compute the beam mean energy. With $\alpha=m/2k_BT$ , $\langle E\rangle_{\rm beam}=\dfrac{\int_0^\infty \tfrac12 mv^2\cdot v^3 e^{-\alpha v^2}\,\mathrm dv}{\int_0^\infty v^3 e^{-\alpha v^2}\,\mathrm dv}=\dfrac{\tfrac12 m\int v^5 e^{-\alpha v^2}}{\int v^3 e^{-\alpha v^2}}$ .

Step 3. Use $\int_0^\infty v^5 e^{-\alpha v^2}\mathrm dv=1/\alpha^3$ and $\int_0^\infty v^3 e^{-\alpha v^2}\mathrm dv=1/(2\alpha^2)$ , so the ratio is $2/\alpha$ , and $\langle E\rangle_{\rm beam}=\tfrac12 m\cdot\dfrac{2}{\alpha}=\tfrac12 m\cdot\dfrac{2\cdot2k_BT}{m}=2k_BT$ . Why hotter than the bulk: the bulk mean is $\tfrac32 k_BT$ ; the extra factor of $v$ shifts probability weight to higher speed, adding exactly $\tfrac12 k_BT$ to the mean.

Step 4. Graham's law: at fixed $T$ and $P$ , equation (2) gives $\Phi\propto1/\sqrt m$ , so two gases effuse at rates in the ratio $\Phi_1/\Phi_2=\sqrt{m_2/m_1}$ . For $^{235}\mathrm{UF_6}$ ( $M=349$ ) and $^{238}\mathrm{UF_6}$ ( $M=352$ ), one stage separates them by $\sqrt{352/349}=1.00429$ — a $0.43\%$ enrichment per pass. Why so many stages are needed: to raise $^{235}$ enrichment from natural $0.7\%$ to weapons grade requires the tiny per-stage factor compounded over thousands of cascaded effusion cells.

Limiting cases

Case	Result	Meaning
Molecular (effusion) regime	hole $\ll$ mean free path	Molecules leave without colliding in the aperture; $\Phi=\tfrac14 n\bar v$ applies.
Hydrodynamic regime	hole $\gg$ mean free path	Bulk flow; Bernoulli/nozzle physics, not effusion.
Beam energy	$\langle E\rangle_{\rm beam}=2k_BT$	Fast-biased; hotter than bulk $\tfrac32 k_BT$ .
Two isotopes	$\Phi_1/\Phi_2=\sqrt{m_2/m_1}$	Graham's law; basis of gaseous-diffusion enrichment.
Pressure form	$\Phi=P/\sqrt{2\pi m k_BT}$	Hertz-Knudsen; impingement rate on any surface.

Worked examples

Example 1 — Effusion rate of helium from a spacecraft micro-leak (enumerate, apply, get a number). A tank holds helium ( $m=6.64\times10^{-27}\,\mathrm{kg}$ ) at $P=1.0\times10^5\,\mathrm{Pa}$ , $T=300\,\mathrm K$ , and develops a pinhole of area $A=1.0\times10^{-12}\,\mathrm{m^2}$ ( $1\,\mu\mathrm m^2$ ).

Step 1. Enumerate: the escape rate is $\dot N=\Phi A$ with $\Phi=P/\sqrt{2\pi m k_BT}$ from equation (2).

Step 2. Apply: $2\pi m k_BT=2\pi(6.64\times10^{-27})(1.381\times10^{-23})(300)=1.729\times10^{-46}$ , so $\sqrt{\cdot}=1.315\times10^{-23}$ . Then $\Phi=(1.0\times10^5)/(1.315\times10^{-23})=7.61\times10^{27}\,\mathrm{m^{-2}s^{-1}}$ .

Step 3. Boundary number: $\dot N=\Phi A=(7.61\times10^{27})(1.0\times10^{-12})=\mathbf{7.6\times10^{15}\ \text{atoms/s}}$ . That sounds enormous, but it is only $\sim1.3\times10^{-8}$ mol/s — a slow leak in molar terms, yet enough to matter over a mission lifetime, and helium (lightest, fastest) leaks fastest of all.

Example 2 — Uranium isotope separation by gaseous diffusion. $^{235}\mathrm{UF_6}$ has molar mass $349\,\mathrm{g/mol}$ ; $^{238}\mathrm{UF_6}$ has $352\,\mathrm{g/mol}$ . Natural uranium is $0.72\%$ $^{235}$ .

Step 1. Per-stage separation factor (Graham's law): $\alpha=\Phi_{235}/\Phi_{238}=\sqrt{m_{238}/m_{235}}=\sqrt{352/349}=\mathbf{1.00429}$ . Each effusion stage enriches $^{235}$ by only $0.43\%$ .

Step 2. Number of stages to reach a target. To go from $0.72\%$ to, say, $90\%$ weapons grade, the enrichment ratio in the $^{235}/^{238}$ abundance must rise by a factor $\sim(90/10)/(0.72/99.28)\approx1240$ . Each stage multiplies the abundance ratio by $\alpha=1.00429$ , so the number of ideal stages is $\ln(1240)/\ln(1.00429)=7.12/0.00428\approx\mathbf{1660}$ stages.

Step 3. Boundary check: real cascades used several thousand stages (extra stages compensate for imperfect cuts and back-mixing). The physics of the entire enterprise reduces to the $\sqrt{352/349}$ factor from this one flux formula — a striking case of a $0.4\%$ molecular effect scaled to industrial and geopolitical consequence. A tiny kinetic bias, cascaded, does enormous work.

Origins · 1 minknudsen number and how this was discovered

knudsen number

kinetic-theory concept · Knudsen 1909; Graham 1846; Hertz 1882

The Knudsen number $\mathrm{Kn}=\ell/L$ — the ratio of the mean free path $\ell$ to a characteristic system size $L$ — decides whether a gas flow is molecular ( $\mathrm{Kn}\gg1$ , effusion) or hydrodynamic ( $\mathrm{Kn}\ll1$ , continuum). Martin Knudsen established the molecular-flow regime experimentally around 1909, verifying the $\tfrac14 n\bar v$ flux and the cosine emission law for effusing beams; the regime and the number now bear his name. The rate law's mass dependence was found much earlier by Thomas Graham (1846), whose law of effusion $\Phi\propto1/\sqrt m$ underlies isotope separation, and the pressure form was developed by Heinrich Hertz (1882) and later Knudsen for evaporation and condensation — the Hertz-Knudsen equation of vacuum science.

History

1846

— Graham states his law of effusion: rate is inversely proportional to the square root of molar mass.

1882

— Hertz measures evaporation rates, formulating the maximum flux from a surface.

1909

— Knudsen establishes molecular (free-molecular) flow and the cosine emission law experimentally.

1920

— Stern uses effusive beams to measure the Maxwell speed distribution directly.

1945

— Gaseous-diffusion cascades enrich uranium using Graham's law at industrial scale.

Going further

References. Pathria & Beale, Statistical Mechanics, 3rd ed., §6.4 (kinetic considerations, effusion) · Kardar, Statistical Physics of Particles, Ch. 3 (kinetic theory) · Reif, Fundamentals of Statistical and Thermal Physics, §7.10–7.13 · Sethna, Statistical Mechanics: Entropy, Order Parameters, and Complexity, Ch. 3 · Landau & Lifshitz, Statistical Physics Part 1, §39. All formulas are standard kinetic-theory results. The key relations are: $\Phi=\tfrac14 n\bar v=n\sqrt{k_BT/2\pi m}$ (from integrating $v_z f(\mathbf v)$ over the outgoing hemisphere); $\bar v=\sqrt{8k_BT/\pi m}$ ; beam distribution $\propto v^3e^{-mv^2/2k_BT}$ ; $\langle E\rangle_{\rm beam}=2k_BT$ (via the $v^3$ -weighted energy average); Hertz-Knudsen $\Phi=P/\sqrt{2\pi m k_BT}$ (using $P=nk_BT$ ); and the $^{235}\mathrm{UF_6}$ / $^{238}\mathrm{UF_6}$ enrichment factor $\sqrt{352/349}\approx1.0043$ from molar masses $352$ and $349\,\mathrm{g/mol}$ . Boltzmann constant: $k_B=1.380649\times10^{-23}\,\mathrm{J/K}$ .

Builds on: Maxwell-Boltzmann Velocity Distribution · Maxwell-Boltzmann Speed Distribution · Characteristic Speeds: Mean, RMS, and Most Probable · Classical Ideal Gas Leads to: Equipartition Theorem · Sackur-Tetrode Equation · Thermal de Broglie Wavelength

Why is the flux ¼ n v̄ and not ½ n v̄, since half the molecules move toward the hole?

Because the flux is not just a count of outward-moving molecules — it weights each by its outward speed $v_z$ . Only half the molecules have $v_z>0$ , which is one factor of $\tfrac12$ ; but the average of $v_z$ over that outgoing half is $\bar v/2$ , not $\bar v$ , contributing another factor of $\tfrac12$ . Combining, $\Phi=\tfrac12\cdot(\tfrac12\bar v)\cdot n=\tfrac14 n\bar v$ . The extra factor comes from the fact that fast molecules cross more often than slow ones, so the flux is a $v_z$ -weighted average, not a simple number count.

Why does the effusing beam carry more energy than the gas it came from?

Effusion samples molecules in proportion to how fast they cross the hole, so it over-samples the fast ones. The beam's speed distribution therefore has an extra factor of $v$ relative to the bulk ( $\propto v^3$ instead of $v^2$ ), shifting it toward higher speeds. Working out the mean kinetic energy of this $v^3$ -weighted distribution gives $2k_BT$ , compared with the bulk $\tfrac32 k_BT$ . The residual gas, having lost its fastest molecules, is left infinitesimally cooler — effusion is, in principle, a cooling process, and evaporative cooling is exactly this effect applied to a liquid surface.

How does Graham's law separate uranium isotopes if the rates differ by less than half a percent?

A single effusion stage separates $^{235}\mathrm{UF_6}$ from $^{238}\mathrm{UF_6}$ by only $\sqrt{352/349}=1.0043$ — a $0.43\%$ enrichment. The trick is to cascade thousands of stages: the enriched output of one stage feeds the next, and the tiny per-stage factor compounds geometrically. Going from natural $0.7\%$ to reactor- or weapons-grade requires on the order of a thousand or more ideal stages (more in practice, to correct for imperfect separation and back-mixing). The gaseous-diffusion plants of the Manhattan Project and the Cold War were physically enormous precisely because each stage does so little.

What is the mean free path, and why must the hole be smaller than it?

The mean free path $\ell$ is the average distance a molecule travels between collisions. For effusion, the aperture must be smaller than $\ell$ so that molecules pass through the hole individually, without colliding with each other in the opening — each escaping molecule's trajectory is set by the velocity distribution inside the gas, not by collective flow. If the hole is much larger than $\ell$ , collisions in the aperture organize the gas into a bulk stream (hydrodynamic flow through a nozzle), and the effusion formulas $\tfrac14 n\bar v$ and the $2k_BT$ beam energy no longer apply.

Does effusion change the pressure and composition of the source gas over time?

Yes. As gas effuses out, the density (and hence pressure) of the source falls, and because lighter molecules leave faster (Graham's law), the composition of the residual gas shifts toward the heavier species. For a container with a small hole, the number decays exponentially, $N(t)=N_0 e^{-t/\tau}$ with $\tau=4V/(A\bar v)$ set by the volume $V$ , hole area $A$ , and mean speed. A mixture will fractionate: the residual gas becomes enriched in the heavier component while the escaped gas is enriched in the lighter one — the working principle of isotope separation.

Is the Hertz-Knudsen flux the same as the rate molecules hit a wall?

Yes — the impingement rate on any surface exposed to a gas is exactly $\Phi=\tfrac14 n\bar v=P/\sqrt{2\pi m k_BT}$ , whether or not there is a hole there. "Effusion through a hole" and "molecules striking a patch of wall" are the same integral; the hole simply lets the striking molecules through instead of bouncing back. This is why the same formula gives the monolayer formation time in vacuum systems (how fast a clean surface gets coated by residual gas) and the maximum evaporation or condensation rate at a liquid or solid surface.