Beveridge.
Interactive lab · experiment 1

Hydrogen permeation through a metal membrane

Hydrogen enters one face of a thin metal sheet and is detected leaving the other. How fast it gets through depends on the lattice, on traps that hold it, and on grain boundaries that act as fast paths. Change the conditions and rerun the experiment; every curve is checked against its exact solution.

Experiment settings

Defaults: polycrystalline nickel, typical literature values.

Lattice only + trapping + trapping + grain boundaries exact solution
ModelEffective D (m²/s)Time lag, exactTime lag, simulatedMax curve errorCheck

Status. Each simulated curve is checked against the exact solution of the same equations (pass: largest difference below 1% of the steady flux). These curves have not yet been compared with the figures in the papers; that comparison is added once the papers' exact settings are entered.

What is being computed

1 · Lattice diffusion

Fick's second law across the membrane, ∂c/∂t = D ∂²c/∂x², with a fixed concentration on the entry face and zero on the exit face. Lattice diffusivity D_L = D₀ exp(−Q/RT); for nickel D₀ = 6.9×10⁻⁷ m²/s, Q = 39.5 kJ/mol.

2 · Trapping

Traps in local equilibrium with the lattice slow the hydrogen down: D_eff = D_L / (1 + (N_T/N_L) exp(E_b/RT)). Example: N_T/N_L = 10⁻⁵ and E_b = 30 kJ/mol at 298 K give a factor 1 + 1.8 ≈ 2.8 slower.

3 · Grain-boundary fast paths

A share f of the material is grain boundary, f = 3δ/d with boundary width δ = 0.5 nm and grain size d. The effective diffusivity mixes both paths: D = (1 − f) D_eff + f D_GB.

Flux leaving the membrane, as a share of its final value: J/J_ss = 1 + 2 Σ (−1)ⁿ exp(−n²π² D t / L²). Time lag (the classic measurement): t_lag = L² / 6D. Example: L = 0.5 mm and D = 1×10⁻¹³ m²/s give t_lag = (5×10⁻⁴)² / (6×10⁻¹³) ≈ 4.2×10⁵ s ≈ 116 h. The simulation solves the same equation step by step (finite differences, 120 points across the membrane) and must reproduce the exact curve.

Open the simulation code