Computational screening pipeline for earth-abundant acid OER catalysts - targeting an iridium-free green hydrogen future.
Proton-exchange membrane (PEM) electrolysers split water into green hydrogen and oxygen. The oxygen evolution reaction (OER) at the anode is the bottleneck - it requires a catalyst that is active, stable in acid, and earth-abundant (IrO₂, the commercial standard, uses iridium at ~$50,000/kg).
This project uses a three-gate ML screening pipeline to identify Ca–Mn–W oxide compositions that could replace IrO₂:
| Gate | Question | Method |
|---|---|---|
| Gate 1 - Synthesis | Will the target phase form without decomposing? | XGBoost phase stability model, 216 temperature/pH conditions |
| Gate 2 - eg Tuning | Is the eg filling in the Sabatier optimal zone (0.45–0.59)? | Volcano-curve regression, Bayesian composition optimiser |
| Gate 3 - Lifetime | Will it survive >50,000 h of operation? | Dissolution kinetics model, pulsed vs continuous operation |
Primary candidate: Ca(0.11)Mn(0.55)W(0.34) - passes all three gates. Predicted η₁₀ ≈ 278 mV (IrO₂: 250 mV), P50 lifetime ≈ 143,506 h pulsed (IrO₂: ~50,000 h).
Gate Status Board - all three screening gates pass for the Ca–Mn–W system
Interactive sliders for Ca/Mn/W/Ti fractions. Real-time gate evaluation with volcano curve overlay.
P50 lifetime (median time to 2× dissolution rate) under pulsed and continuous operation. Benchmarked against IrO₂.
Side-by-side comparison of the best reported earth-abundant acid OER catalysts (α-MnO₂, δ-MnO₂, MnCoP, IrO₂ benchmark) against our prediction.
| Gate | Result | Detail |
|---|---|---|
| Gate 1 - Synthesis Sweep | GO ✅ | 216/216 conditions pass; f_CaWO₄ = 22.1% at optimal T/pH |
| Gate 2 - eg Tuning | GO ✅ | 3/3 compositions; best eg = 0.520 (target 0.45–0.59), η₁₀ = 245 mV |
| Gate 3 - Lifetime | GO ✅ | 3/4 compositions; Ca(0.11)Mn(0.55)W(0.34) P50 pulsed = 143,506 h |
%%{init: {'theme': 'dark', 'themeVariables': {'primaryColor': '#0f172a', 'primaryTextColor': '#e2e8f0', 'primaryBorderColor': '#22c55e', 'lineColor': '#22c55e'}}}%%
flowchart TD
classDef comp fill:#0c1a2e,stroke:#38bdf8,color:#7dd3fc,font-weight:bold
classDef gate1 fill:#0f1f3d,stroke:#38bdf8,color:#93c5fd
classDef gate2 fill:#1c1400,stroke:#f59e0b,color:#fcd34d
classDef gate3 fill:#0a2520,stroke:#10b981,color:#6ee7b7
classDef result fill:#0f2d2a,stroke:#22c55e,color:#86efac,font-weight:bold
classDef reject fill:#2d0f0f,stroke:#ef4444,color:#fca5a5
Comp(["Ca–Mn–W–Ti\nComposition Input"]):::comp
Gate1["Gate 1 Synthesis Sweep\nXGBoost phase stability\n216 temperature/pH conditions"]:::gate1
Pass1{"GO ✅\n216/216 pass\nf_CaWO₄ = 22.1%"}:::gate1
Gate2["Gate 2 eg Tuning\nVolcano-curve regression\nSabatier optimal zone 0.45–0.59"]:::gate2
Pass2{"GO ✅\neg = 0.520\nη₁₀ = 245 mV"}:::gate2
Gate3["Gate 3 Lifetime\nDissolution kinetics model\npulsed vs continuous operation"]:::gate3
Pass3{"GO ✅\nP50 = 143,506 h pulsed\nvs IrO₂ ~50,000 h"}:::gate3
Result(["🏆 Primary Candidate\nCa(0.11)Mn(0.55)W(0.34)\nAll three gates pass"]):::result
Reject(["❌ Eliminated"]):::reject
Comp --> Gate1 --> Pass1
Pass1 -->|pass| Gate2 --> Pass2
Pass2 -->|pass| Gate3 --> Pass3
Pass3 -->|pass| Result
Pass1 & Pass2 & Pass3 -->|fail| Reject
code/
dashboard.py # Streamlit dashboard (entry point)
gate1_phase_predictor.py # Gate 1: XGBoost phase stability
gate2_eg_tuner.py # Gate 2: volcano-curve eg optimiser
gate3_lifetime_projector.py # Gate 3: dissolution lifetime model
results_gate1_synthesis.csv
results_gate2_optimization.csv
results_gate3_projection.csv
results_acid_oer_pareto.csv
results_ca_mnw_pareto.csv
docs/
research/ # 20 background research documents
screenshots/ # README screenshots
tests/e2e/ # Playwright E2E suite (125 passing)
git clone https://github.com/m4cd4r4/green-h2-catalyst-research.git
cd green-h2-catalyst-research
pip install -r requirements.txt
cd code
streamlit run dashboard.pycd code
python gate1_phase_predictor.py # → results_gate1_synthesis.csv
python gate2_eg_tuner.py # → results_gate2_optimization.csv
python gate3_lifetime_projector.py # → results_gate3_projection.csvXGBoost classifier trained on DFT-derived formation energies and experimental phase diagrams. Features: composition vector, synthesis temperature (50–200 °C), electrolyte pH (5–11). Target: binary phase stability flag.
Sabatier volcano principle: OER activity peaks at eg ≈ 0.5 (half-filled eg orbital). Regression model maps Ca/Mn/W/Ti fractions → eg filling. Bayesian optimiser minimises |eg − 0.50|.
Tafel-law dissolution kinetics: D_ss = D₀ · exp(η / b). P50 = time to reach 2× initial dissolution rate under Monte Carlo parameter sampling. Pulsed operation modelled with accelerated degradation factor α = 1.8.
- IrO₂: η₁₀ = 250 mV, D_ss = 0.01 µg/cm²/h, P50 ≈ 50,000 h
- Target: η₁₀ < 300 mV, D_ss < 2 µg/cm²/h, P50 > 50,000 h
- Man, I. C. et al. Universality in Oxygen Evolution Electrocatalysis on Oxide Surfaces. ChemCatChem 3, 1159–1165 (2011). [eg volcano principle]
- Seitz, L. C. et al. A highly active and stable IrOₓ/SrIrO₃ catalyst for the oxygen evolution reaction. Science 353, 1011–1014 (2016).
- Hücker, S. M. et al. Dissolution of IrO₂ in acid electrolytes. J. Electrochem. Soc. 168, 044502 (2021).
- Frydendal, R. et al. Benchmarking the stability of oxygen evolution reaction catalysts. ChemElectroChem 1, 2075–2081 (2014).
- Nong, H. N. et al. A unique oxygen ligand environment facilitates water oxidation in hole-doped IrNiOₓ. Nat. Catal. 1, 841–851 (2018).
@software{green_h2_catalyst_2026,
author = {Ó Murchú, Macdara},
title = {Green H2 Catalyst Research Dashboard},
year = {2026},
url = {https://github.com/m4cd4r4/green-h2-catalyst-research},
note = {Computational screening dashboard for earth-abundant acid OER catalysts}
}See also CITATION.cff.
MIT - see LICENSE.