Presets
Playback
Experiment
Conditions
View
Camera
Layers
Quality
Model check
The science
What you are looking at
Two gases share a box with a partition in the middle. In mixing mode the light gas (cyan) starts on the left, the heavy gas (orange) on the right, and lifting the gate lets them spread into each other. In effusion mode both gases start on the left and can only reach the right chamber through five small round pores; the light gas leaks out faster. One molecule of each gas leaves a trace so you can see the random walk.
Equations used
- Maxwell-Boltzmann speeds:
f(v) = 4π (m / 2πkT)3/2 v² exp(-m v² / 2kT), mean speed〈v〉 = sqrt(8kT / πm), rms speedsqrt(3kT/m). Every velocity component is drawn from a Gaussian of widthsqrt(kT/m). - Mean free path
λ = kT / (√2 π d² p)with the kinetic diameterdof each gas (rounded literature values) and the pressurep. Mean time between collisionsτ = λ / 〈v〉. - Collisions with the background gas are a Poisson process: each step a molecule has probability
1 - exp(-dt/τ)to get a fresh Maxwellian velocity. This gives exactlyD = kT τ / m = (π/8) λ 〈v〉. Hard-sphere kinetic theory gives a coefficient of the same size (about 1/3 to 0.6 depending on the level of approximation), so treat D as a scaling law, not a measured value. Heavier molecules are slower, soD ∝ T3/2 / (p √M). - Random walk:
〈r²〉 = 6 D tin three dimensions. The diffusion equation∂c/∂t = D ∂²c/∂x²with reflecting walls and a step initial profile (Fick's second law) has the solutionc(x,t) = 1/2 + Σk odd (2 / kπ) (-1)(k-1)/2 cos(kπx/L) exp(-(kπ/L)² D t); the dashed curves in the graphs are this series. The mixing index is1 - Σ 8 / (kπ)² exp(-(kπ/L)² D t). - Effusion (Graham's law): a hole much smaller than the free path passes
n〈v〉A/4molecules per unit time, so the rate is proportional to1/√Mandratelight / rateheavy = √(Mheavy / Mlight). For two equal chambers the fraction that has gone through is(1 - exp(-2kt)) / 2withk = 10π r² 〈v〉 / Lfor five pores of radiusr(in units of L).
Honest simplifications
- Molecules do not collide with each other; the other gas acts as a background that only re-randomises the velocity. So each gas diffuses on its own with its own D, whereas a real binary mixture has one shared inter-diffusion coefficient D12 and the lighter gas drags the heavy one along.
- The real free path at normal pressure is far below a micrometre, so a micrometre-sized box would need thousands of collisions per crossing. The picture therefore draws a free path of at least 0.03 chamber lengths (mixing) or 1 chamber length (effusion), marked with * in the readouts; the diffusion coefficient shown uses the drawn value. The graph "D versus temperature" uses the real free path.
- Time is slowed by about a billion. Spheres are drawn larger than the kinetic diameter for visibility; walls are specular and isothermal; no gravity, no chemical reaction, no bulk flow from a pressure difference.
What to try
- Mixing, H2 and CO2: the cyan gas spreads much faster; both mixing curves follow their dashed theory.
- Raise the temperature to 600 K: mean speeds and D rise; the thermometer and heater glow follow.
- Raise the pressure: the real free path shrinks, so D in the last graph falls roughly as 1/p.
- Effusion, He and N2: compare the fraction through the pores with the dashed law and read the rate ratio against sqrt(28.0/4.0) = 2.65.
Numerical checks run in this page
(1) sampled speeds against Maxwell-Boltzmann, (2) mean square displacement of free molecules against 6Dt, (3) the simulated step profile against the cosine-series Fick solution, (4) the effusion rate ratio against Graham's law. They use fresh particle sets, independent of the live run, and re-run when the gases or temperature change.
Models and 3D assets: the chamber, gate, five-pore membrane, heater plinth, pressure gauge and thermometer were modelled and texture-baked in Blender and are driven by the equations above in Three.js r160.