Green hydrogen with seven times less iridium
The electrolysers that split water into hydrogen using solar and wind power need iridium, one of the rarest metals on Earth. This team grew the iridium catalyst a few atoms at a time, straight onto the electrode, and got close to commercial performance with a fraction of the metal.
Why it matters
Hydrogen made from water with solar and wind power could replace fossil fuels both as a fuel and as a raw material for industry, but the proton-exchange-membrane electrolysers that make it need iridium on their anodes, and the whole world mines only about 7 tonnes of it a year. Growing the iridium one atomic layer at a time, this team reached about 83 % of a commercial anode’s current at the same 1.8 V while using 7× less iridium (0.21 vs 1.4 mg/cm²), and it held up through a test equivalent to at least 600 hours of stop-start solar or wind operation.
The paper
Ionomer-free Iridium Anodes with Low Loading for Zero-gap Electrolyzers via Direct Atomic Layer Deposition
Peter M. Piechulla, Iris Blommestein, Anne van Gerven, Henri Pelzer, Ezgi Demiröz, Jonas Englhard, Mingliang Chen, Julien Bachmann & J. Ruud van Ommen (TU Delft and FAU Erlangen-Nürnberg). arXiv:2610.15798 [cond-mat.mtrl-sci], submitted 11 Oct 2026. Read on arXiv · PDF
In plain language: a PEM water electrolyser is a sandwich. Water goes in at the anode, where an iridium catalyst pulls oxygen out of it, and hydrogen comes out at the other side. Normally the iridium is mixed into an ink with a plastic binder (the “ionomer”) and sprayed onto the membrane at more than 1 mg per cm². Here the team used atomic layer deposition (ALD), which builds a coating one layer of atoms per cycle, to grow iridium directly onto a porous titanium felt, sometimes topped with a finer layer of sintered titanium nanoparticles (a “microporous layer”, MPL). No ink and no binder are needed.
With ozone-based thermal ALD (75 cycles on the MPL) the anode held only 0.21 mg/cm² of iridium yet drew 1.24 A/cm² at 1.8 V, against 1.49 A/cm² for a commercial catalyst-coated membrane with 1.4 mg/cm². Halving the ALD cycles from 150 to 75 cut the iridium by 74 % with no loss in current, showing that only the iridium within about 2.5 µm of the membrane does useful work. A plasma-based ALD process cut the iridium to 0.04 mg/cm², but it made a less active, crystalline IrO₂ and wore out faster on the MPL. Because there is no binder, the same kind of anode also worked in a CO₂ electrolyser with a different, alkaline membrane.
Try it
Every measured number on this page comes from the paper (Table 1 and Sections 2.4, 3.1–3.2). The iridium-per-kilowatt figures are our arithmetic on Table 1. Panel 2 uses the paper’s Tafel equation (Eq. 3) plus the series-resistance term as a simple two-loss model. It leaves out mass transport, so treat its what-if answers as rough guides, not predictions (see “The maths”).
1 · How much iridium does each kilowatt need?
Anode iridium only, at 1.8 V and about 70 % efficiency, with every gram mined in a year going to anodes and no recycling. A real plant also needs catalyst for the cathode, spares and stack hardware, so these are lower bounds. World supply: about 7 t in 2025 (International Platinum Group Metals Association iridium factsheet, 2026).
2 · Where do the volts go?
3 · Does it last? (paper Section 2.4)
| Anode | Drift over 12–24 h at 1 A/cm² | Tafel-slope change after the test | What the paper saw |
|---|---|---|---|
| ThALD 75c MPL | +16 µV/h | +1 mV/dec | Only minor changes to the polarisation curve |
| PEALD 450c MPL | +583 µV/h | +10 mV/dec | Severe loss of activity |
| PEALD 450c bare felt | −23 µV/h | −11 mV/dec | Got better; part of the IrO₂ turned into a more active IrOx |
The paper calls the stress test equivalent to at least 600 h of intermittent operation, for example on wind or solar power. A straight-line extrapolation of a 12-hour drift rate is only a size guide; real electrolysers age in more complicated ways.
The maths in this page
Tafel analysis (paper Eq. 3). The authors fit the low-current part of each polarisation curve (10⁻³ to 10⁻² A/cm²) with
where b is the Tafel slope (mV per decade of current) and RΩ the series resistance from impedance fitting. Our simplification: panel 2 turns this around into a two-loss model, V(j) = Erev + b·log₁₀(j⁄j₀) + j·RΩ. The paper does not report j₀ (it calls it too uncertain), so we pick j₀ for each anode so that the curve passes through its measured point (j at 1.8 V, Table 1, with the Table 1 RΩ). The model leaves out mass-transport losses.
Low-current b vs effective beff. By integrating impedance data (Eqs. 1–2, Vcell(j) ≈ ∫ LFR(j) dj), the authors found that the low-current slopes understate the real catalyst losses at high current. Fitting the charge-transfer loss gave beff = 84 mV/dec for ThALD 75c MPL (vs 44–45 from the low-current fit) and 76 mV/dec for the CCM (vs 43). With beff our model gives about +61 % more current at 1.8 V for a membrane 60 mΩ·cm² thinner, close to the paper’s +59 % for ThALD 75c MPL. With the low-current b it gives about +69 %, which is too optimistic.
| Anode (paper Table 1) | Ir loading, mg/cm² | RΩ, mΩ·cm² | b, mV/dec | j at 1.8 V, A/cm² | g Ir/kW (ours) |
|---|
Our additions (not from the paper). Iridium per kilowatt = loading ⁄ (j × Vcell). Since mg/cm² ÷ W/cm² = mg/W = g/kW, this is the paper’s “power-specific loading” at 1.8 V. Efficiency = 1.253 V ⁄ Vcell, where 1.253 V is hydrogen’s lower heating value (241.8 kJ/mol) divided by 2F. This gives 70 % at 1.8 V, matching the paper’s “≈ 70 %LHV”. Hydrogen output follows Faraday’s law: ṁ = j ⁄ (2F) × 2.016 g/mol, which is about 0.376 kg per hour per m² for each A/cm². In panel 1 the hollow thin-membrane dots come from the same model with RΩ − 60 mΩ·cm². The paper instead shifted its measured curves, so expect differences.
Teaching demo built from the paper’s equation and published numbers. It is not the authors’ code or data.
How to use it
Pick an anode at the top; every panel follows it. In panel 1, find your anode on the map: dots low down use little iridium per kilowatt, and dots to the right push more current, and so more hydrogen, through the same size of cell. The green band is the 0.1–0.4 g/kW goal from electrolyser roadmaps. Drag “Electrolysers to build” to see how many tonnes of iridium a hydrogen roll-out would need with each anode, compared with the roughly 7 tonnes the world mines each year. Tick the thinner-membrane box to see the dots slide right. In panel 2, slide the operating current to see the cell voltage split into the unavoidable 1.168 V, the catalyst’s activation loss and the resistance loss. Lower the series resistance to mimic a thinner membrane, and switch to the effective slope beff (ThALD 75c MPL and CCM only) to see why the simple Tafel picture flatters the ALD anode at high current. In panel 3, drag the hours to compare how fast each anode’s voltage creeps up.