A Perovskite 'Leaf' Pairs Solar Hydrogen From Seawater With Hydrazine Cleanup

Researchers at Nanyang Technological University (NTU) have built a bias-free photoelectrochemical cell that makes hydrogen from alkaline seawater and destroys hydrazine, a toxic industrial pollutant, at the same time. According to the team's paper in Nature Communications, the device delivers 25 mA cm⁻² under 1-sun illumination and stays stable for three days. The device runs without any external electrical input. It is a proof of concept with real limits, and the most important one is built into the design.

01Replacing the oxygen reaction

In conventional water electrolysis, the anode evolves oxygen, which is the energy-hungry half of the process and usually yields a product nobody sells. The paper's abstract identifies two penalties in seawater: the large potential that oxygen evolution requires, and the unwanted oxidation of chloride. The chloride route produces corrosive, toxic chlorine compounds that attack electrodes.

The NTU group sidesteps both by giving the anode a different job. Hydrazine oxidises at a much lower potential than water, so a bias-free device can run at high performance and avoid chloride oxidation. The anode catalyst breaks hydrazine down into hydrogen and nitrogen, so hydrogen appears at both electrodes under illumination. The hydrazine is consumed as the fuel for the anode, so the pollutant is removed as a side effect.

The numbers are consistent with that picture. A current of 25 mA/cm² corresponds, by Faraday's law and assuming two electrons per H₂ molecule, to about 466 µmol of hydrogen per square centimetre per hour. That is the reported rate. This is my own arithmetic, not a figure from the paper. It suggests the headline production figure reflects essentially the cathodic reaction at near-full Faradaic efficiency, with the anode's extra hydrogen as a bonus rather than part of that number.

02Architecture: a protected perovskite and a noble-metal-free anode

The device is a two-electrode photoelectrochemical cell. It pairs a perovskite photocathode with a noble-metal-free spinel oxide anode, with no external bias. The anode catalyst contains iron, cobalt and chromium. NTU says it is corrosion-resistant, its properties can be tuned, and it sharply cuts chlorine-compound formation.

The cathode is the harder part. It is a lead-halide perovskite, a semiconductor that absorbs light and converts it to electricity. Because these materials are vulnerable to moisture and harsh chemistry, the team encapsulated it with a conductive epoxy loaded with silver and copper particles, plus titanium foil. The foil blocks water but lets charge through. A thin platinum layer on the titanium then catalyses hydrogen evolution. So "noble-metal-free" applies to the anode only. The cathode still uses platinum.

The semiconductor therefore sits behind a conductive barrier rather than touching the electrolyte. That is a common approach for perovskite photoelectrodes. It places the stability burden on the seal and on the epoxy and foil interfaces, not on the perovskite itself.

03What was measured

The cell ran for more than 72 hours and produced hydrogen at 466 µmol cm⁻² h⁻¹. The tests used simulated seawater and real seawater, both alkaline and with hydrazine added, under simulated sunlight. For the cleanup test the team scaled the photocathode from 0.56 to 6.25 cm². In about 30 hours, hydrazine fell from 0.5 M (about 1.6% by weight) to 0.5 ppb, against the EPA limit of 10 ppb that NTU cites. The abstract is a little more conservative than the press material: it says "below 1 ppb" under zero-bias operation. The team also ran the leaf outdoors in natural sunlight and saw hydrogen and nitrogen bubbles. That test was qualitative.

NTU describes the photocurrent as among the highest reported for lead-based perovskite cathodes. That is the relevant benchmark, because a bias-free device can only drive as much current as its light absorber can supply.

Context for the stability result comes from the wider field. A 2025 paper on a bismuth vanadate–silicon artificial leaf, which reported 8.4% efficiency, notes that a monolithic silicon–perovskite tandem reached 20.8% solar-to-hydrogen efficiency, but its photocurrent fell to 60% after 102 hours. Against that, 72 hours without significant decline is respectable for a perovskite in seawater. It is still far short of the years that a commercial photoelectrode would need.

The sources I reviewed do not give a solar-to-hydrogen efficiency for the NTU device. That is the standard metric for comparing solar-fuel devices. Because the anode reaction is not water oxidation, a comparison with oxygen-evolving systems would not be like for like anyway.

04The limits that matter

The central limitation is that hydrazine is not an optional extra. Intelligent Living's reading of the paper says the cell produced essentially no current without hydrazine, because the perovskite's operating range and the oxygen-evolution curve do not overlap, so the bias-free current falls to zero. The cell works precisely because hydrazine oxidation sits at a potential the perovskite can reach unassisted. Water oxidation does not. This is therefore not a general route to seawater hydrogen. It is a hydrazine treatment system that makes hydrogen as a by-product. One outlet summarised it as a clever pairing rather than a universal seawater solution.

Several other constraints follow:

• Feedstock. The demonstrations used hydrazine-spiked, alkaline electrolytes. Real hydrazine wastewater streams will vary in concentration, pH and co-contaminants. Those streams have not been tested.
• Throughput. Hydrazine removal is limited by photocurrent per area. The 0.5 M to ppb result came from a 6.25 cm² photocathode and took about 30 hours. As a rough estimate, and assuming the small-cell current density held, that area could treat only tens of millilitres at 0.5 M in that time. This is my estimate, not a reported figure. The final dilute stage is also usually limited by mass transport, not by light.
• Durability. Three days is not a lifetime. The silver–copper epoxy, titanium foil and platinum layers must survive alkaline, chloride-rich water, thermal cycling and years of illumination. The abstract's stability claim relates to the 3-day test.
• Material concerns. The absorber is lead-based. Any real deployment in treated water would need to show that encapsulation failure cannot release lead.
• Fuel value. Hydrazine oxidation lowers the voltage requirement, but it consumes a chemical that already has industrial value. Hydrogen yield from the anode side depends on the exact reaction pathway. The source material does not quantify that yield.

Oxford's James Durrant, who was not involved, described the work as a practical route toward multifunctional photoelectrochemical systems. That framing seems right. The claim is about combining functions, not about beating electrolysers on cost.

05What comes next

NTU says it is now developing catalysts that extend the platform beyond hydrazine, including to other pollutants and to converting waste materials into useful products. That direction matters most. The architecture, a protected perovskite cathode driving an oxidation reaction chosen to match its voltage, is potentially general. The anodic reaction is a swap-in module, provided its onset potential stays within what the light absorber can supply.

The next questions are measurable. These include solar-to-hydrogen efficiency, Faradaic efficiency at both electrodes, and stability over hundreds or thousands of hours. They also include performance in real industrial effluent and a materials-cost and lead-containment assessment at module scale. The existing results show a well-engineered integration at centimetre scale. They do not yet show a technology that can be scaled.