H2sim v0.1.0 — development release

H2sim v0.1.0

First development release of H2sim. This pre-release provides a fixed source version for evaluation, collaboration and reproducible studies. APIs and numerical behavior may change before a stable 1.0 release.

Included workflows

• Hydrogen storage in saline aquifers with black-oil dissolution and vaporization.
• Compositional storage with microbial reactions and optional sequential PHREEQC coupling.
• Hydrogen–brine thermodynamics and fluid-property tabulation.
• Core-only startupH2sim; explicitly enable chemistry with startupH2sim('phreeqc', true).
• Reviewed EOS trace-component floor correction for reproducible physical PHREEQC restarts.
• Local unit tests, frozen numerical references and complete-cycle/restart regressions; reorganized examples and documentation.

Install this version

git clone --branch v0.1.0 --recurse-submodules https://github.com/sintefmath/h2-sim.git
cd h2-sim

In MATLAB, from the repository root:

startupH2sim

Use Git cloning: GitHub's automatic source archives do not contain dependency submodule contents. PHREEQC native libraries require the platform-specific setup described in the installation guide.

H2sim remains dependent on pinned MRST source and maintains its own H2-store and biochemistry implementations. Load them through startupH2sim, rather than loading MRST's separate h2-biochem module.

Recorded verification

The 29 October 2026 local verification used Windows MATLAB R2026a Update 5 launched from WSL, pinned MRST 8e218b0786cd4651814553b0f9ebd6844eaabcb2, pinned PhreeqcMatlab 20662bf8fab5fe2a9bcaa7b60bc7f1ae5337733f, and IPhreeqc 3.8.6:

• 126 biochemistry tests and 217 H2-store tests passed.
• Both complete-cycle/reference regressions passed, including serialized restarts with reconstructed models and fresh PHREEQC sessions.
• 13 Python tooling tests, the strict documentation build and local documentation link/image checks passed.
• Core startup was checked without a PhreeqcMatlab checkout; real chemistry was checked after explicit initialization.

This is the existing verification record for the source merged into this release, not a new test run on publication day. MATLAB regressions run locally; GitHub checks documentation and Python tooling.

Known limitations

• Full native Linux MATLAB regression validation is not yet recorded. Linux PHREEQC support is implemented; macOS sequential PHREEQC coupling is unsupported.
• The hybrid phreeqcPE diagnostic remains engine-history-sensitive and is reported separately from physical restart comparisons.
• Small regression cases do not replace full paper-duration reruns, exhaustive phase-transition checks or independent physical validation.
• The stated core minimum is MATLAB R2021a; the recorded full verification used R2026a Update 5.

See the versioned verification record, contributing guide, and citation metadata. Record the version and dependency revisions when reporting results. H2sim is GPL-3.0-or-later; inherited notices and dataset attribution remain applicable.

Release source: 41710b87a3b390facc0a23a474e202f7cef9a5a5.

Simulation animations

These supplementary previews were generated from development examples added after the tagged v0.1.0 source. The v0.1.0 tag and its installation snapshot are unchanged. Use the linked example source to run these demonstrations; they are not part of the original tagged checkout.

Hydrogen–methane mixing

Download MP4 · Source and parameters · Provenance

A 40-by-40 closed vertical section adapted from Okoroafor et al. (2023), Problem 2. Colors show hydrogen mole fraction; frame timestamps show uneven simulated time intervals. Both component masses were conserved to better than 7e-10 relative in the displayed run. This is a coarse-grid adaptation, not an external benchmark validation.

Cyclic hydrogen storage beneath an anticline

Download MP4 · Example source · Provenance

A small four-layer Gaussian reservoir adaptation of Problem 4 of the same published benchmark. This extended preview uses 730 days of cushion injection at 200,000 surface m³/day, then 90 days shut-in and ten withdrawal/reinjection cycles: 960 simulated days. At a 5% cell gas-saturation threshold, the resolved end-of-buildup plume radius grows from approximately 429 m to 596 m. Playback is slowed to six frames per second. The original example default remains 365 days of cushion injection. Fluid properties are explicit approximations, and no claim of benchmark reproduction is made. See the attached provenance for the tested source, parameters and numerical checks.

Animation attribution

These previews carry the H2sim logo and existing SINTEF software copyright/GPL attribution. The scientific frames and playback timing are preserved. Branding/export source · Branding provenance.