Natural Hydrogen: From Serpentinization to a Membrane-Purified Gas Well

Hydrogen made by water-rock reactions fills a gas reservoir; a Darcy-law well and a polymer membrane skid turn the mixed gas into hydrogen-rich product.

Switch to 2D

Presets

Playback

Parameters

Camera

Layers

Quality

Model check

The science

What you see. A cutaway of a natural-hydrogen field (the underground part is schematic, not to scale) and its surface plant. In the basement, water reacts with iron-rich olivine in the rock (serpentinization) and hydrogen rises along the fault (bubbles) into a sandstone reservoir under a cap rock. A well brings up the mixed gas (H2 with nitrogen and methane); it is compressed and sent through glass-walled membrane vessels. The coloured discs in the vessels show the hydrogen fraction of the feed gas along the membrane (blue = rich, orange = depleted). Pure-ish hydrogen goes to the tank, the rest leaves through the stack.

Source. The hydrogen-forming reaction is 3 Fe2SiO4 + 2 H2O → 2 Fe3O4 + 3 SiO2 + 2 H2: 2 mol of H2 per 3 mol of fayalite, i.e. about 6.6 kg of hydrogen per tonne of that mineral. The model uses the generation rate as a slider and reports how much rock it implies.

Reservoir and well. The reservoir is a tank of gas: p = n z R T / V. The well follows the steady radial Darcy law for a real gas, q = π k h (p² − pwf²) / (z R T μ ln(re/rw)) in mol/s, with the mixture viscosity from a mole-weighted average. Because only hydrogen is generated, the composition changes in time; the reservoir pressure rises if the source exceeds the production and falls otherwise.

Membrane. A polymer membrane transports each gas by solution–diffusion, Ji = Πi (pf xi − pp yi), where Πi is the permeance (H2: 100 GPU; N2: 100/α; CH4: 100/(1.6 α)), pf and pp the feed and permeate pressures, and x, y the feed-side and local permeate mole fractions. The module is a cross-flow plug-flow model: the local permeate composition is solved at every point and the feed flows are integrated along the area with fourth-order Runge–Kutta. The compressor power is adiabatic with efficiency 0.75.

Honest simplifications. The reservoir is a single homogeneous tank (no layers, water drive, fault flow resistance, adsorption or gas dissolved in water); the reservoir volume, permeability and the generation rates are illustrative values, not measurements of any field; the generation rate is simply set, not computed from rock chemistry, temperature or water supply; the gas is treated with a constant compressibility factor and a rough viscosity mixing rule; the well has no tubing hydraulics and water or condensate is ignored; the membrane permeances are constant (real polymers show pressure, temperature and water effects, plasticization and ageing) and the modules are idealised cross-flow with no pressure drop or concentration polarisation; the compressor is a simple adiabatic estimate; only H2, N2 and CH4 are present. Use it for the qualitative trade-offs (area against purity and recovery, selectivity, drawdown against depletion), not for resource estimates. Educational physics only.

Checks. (1) the membrane solver reproduces the closed-form binary local permeate composition, including the vacuum limit; (2) the Runge–Kutta integration converges at fourth order and the integrated flux equals the gas lost from the feed side; (3) shooting the radial pressure ODE reproduces the closed-form Darcy rate; (4) the time-integrated reservoir relaxes to the analytic steady state pressure, and the serpentinization reaction balances atoms with the quoted hydrogen yield.

Loading model...
Loading Blender model...
Hydrogen fraction along the membrane
Reservoir pressure and composition history
Purity and recovery against membrane area
Made with 🟠 Blender · Three.js