The central idea is deceptively simple. Zinc metal can be oxidized in an alkaline electrolyte, and in that oxidized state it stores chemical energy that can later be converted into hydrogen gas. The problem that has always plagued this concept is solubility. When zinc oxidizes in alkaline media, the resulting Zn(II) ions tend to react with hydroxide and precipitate as insoluble zinc oxide or hydroxide, clogging the system and choking off the reaction long before the full capacity of the metal can be exploited. In effect, the carrier saturates almost immediately, which is why zinc-mediated hydrogen delivery has never been taken seriously as an infrastructure-scale option.

The Tsinghua team’s breakthrough came from attacking that solubility limit directly, and they did so with an unusual ally: machine learning. Rather than testing candidate additives one by one in the laboratory, the researchers used a machine learning-guided screening workflow to search for ligands capable of outcompeting hydroxide for coordination to Zn(II) ions in strongly alkaline solution. The search identified lactate, the anion of lactic acid, as a standout performer. When lactate is present, it binds zinc ions so effectively that the soluble Zn(II) concentration doubles, reaching 1.12 molar. That figure matters enormously, because the amount of dissolved zinc species in the electrolyte directly determines how much hydrogen the system can carry and deliver per unit volume.

With the solubility barrier lifted, the researchers built what they call an energy vector-mediated pathway, or EVP. In the charging phase, renewable electricity drives the reduction of dissolved zinc species back to metallic zinc, effectively packing energy into the metal. In the delivery phase, that metallic zinc is oxidized again, and the electrochemical steps liberate hydrogen gas at close to theoretical efficiency. Crucially, the discharge step is self-sustained and circuit-free: once initiated, the chemistry proceeds on its own, releasing hydrogen on demand without needing to be wired to a power source. This decoupling of storage from generation is what makes the scheme attractive for decentralized use, since the energy-dense zinc can be transported as a benign solid or slurry and converted to hydrogen at the point of use.

The performance numbers reported in the paper are the reason the work is drawing attention. The system operates across a temperature window from 0 to 80 degrees Celsius, meaning it needs no exotic thermal management, and it delivers hydrogen with approximately 99 percent Faradaic efficiency, an indication that almost every electron involved in the chemistry goes into making hydrogen rather than wasteful side reactions. More striking still is the volumetric capacity: the team reports 137.9 to 218.4 kilograms of hydrogen per cubic metre of carrier. For comparison, even liquid hydrogen, which must be kept at a frigid minus 253 degrees Celsius, holds about 71 kilograms per cubic metre, and typical 700-bar compressed gas tanks fall far short of that. The authors state that, to the best of their knowledge, these values exceed many established and state-of-the-art hydrogen carriers.

Safety is another pillar of the argument. Compressed hydrogen at hundreds of atmospheres carries inherent risks, and the paper’s reference list includes recent work on the safety of natural hydrogen wells, a reminder that the industry’s incident record is far from spotless. A zinc-based carrier, by contrast, stores hydrogen in chemical form at ambient pressure. There is no high-pressure vessel to rupture, no cryogenic container to bleed away, and the carrier itself is a cheap, abundant, non-toxic metal that is already produced at industrial scale for galvanizing and batteries. For remote communities, mobile platforms and off-grid industrial sites, where the economics of pipelines collapse entirely, a spill-proof, pressure-free carrier could change the calculus of adopting hydrogen altogether.

The scientific detective work behind the ligand discovery deserves particular attention, because it illustrates how coordination chemistry can be engineered rather than merely observed. In conventional alkaline zinc systems, hydroxide ions dominate the coordination sphere of Zn(II), forming zincate complexes that eventually precipitate. Lactate’s carboxylate and hydroxyl groups give it a binding affinity strong enough to wrest zinc ions away from hydroxide, keeping them dissolved as stable lactate complexes even at high concentrations. The team’s electrochemical analyses, presented in the paper’s figures on the role of lactate, show how this ligand reshapes the reaction landscape, suppressing parasitic precipitation while facilitating the electron-transfer steps that produce hydrogen. It is a textbook example of molecular recognition being harnessed for energy technology.

Machine learning played the role of matchmaker in this process. Screening the vast space of possible ligands experimentally would have taken years, but by training models on the relationship between ligand structure, binding energy and performance in alkaline media, the researchers could prioritize the most promising candidates before ever picking up a pipette. The successful validation of lactate, a cheap biomolecule produced worldwide by fermentation, suggests the same workflow could be rerun to find even better ligands, or to adapt the chemistry to other metals. The authors also draw on their prior work in zinc battery electrochemistry, including studies of dendrite management and dual-electrolyte zinc-air systems, indicating that a coherent research program on reversible zinc chemistry underpins this hydrogen result.

The device demonstration rounds out the story. The team built a modular, circuit-free EVP device whose performance is documented in the paper, showing that the concept works not just in a beaker but in an engineered unit that could in principle be scaled and replicated. Modularity is key to the decentralized vision: instead of a continent-spanning pipeline network, one can imagine standardized zinc cartridges charged at renewable energy hubs and shipped by ordinary logistics to fuel cells, industrial furnaces or backup power installations, where they release hydrogen on demand and are returned for recharging. The zinc is not consumed in the overall cycle; it shuttles between oxidation states, functioning as a genuine energy vector in the same sense that lithium shuttles between electrodes in a battery.

Significant hurdles remain between laboratory demonstration and real-world deployment, as they always do. The economics will depend on the round-trip efficiency of charging and discharging, the longevity of the electrolyte, the management of any accumulated by-products over many cycles, and the cost of lactate at scale. The paper does not claim to have solved the hydrogen economy, and the authors declare no competing interests while acknowledging support from Chinese national and regional science foundations. But the core demonstration, that a solubility-limited, century-old zinc chemistry can be transformed by a machine learning-discovered ligand into a carrier with volumetric capacities exceeding liquid hydrogen, is the kind of result that reframes what is possible. If zinc can indeed ride trucks instead of pipelines, the last mile of the hydrogen economy may look far more like a delivery route than an engineering megaproject.

Subject of Research: Reversible zinc oxidation chemistry for on-demand, decentralized hydrogen storage and delivery

Article Title: Reversible zinc oxidation as a route to on-demand hydrogen delivery

Article References: Liu, Z., Zhu, Y., Yao, W., Song, Y., & Zhou, G. (2026). Reversible zinc oxidation as a route to on-demand hydrogen delivery. Nature Synthesis. https://doi.org/10.1038/s44160-026-01190-6

Image Credits: AI Generated

DOI: 10.1038/s44160-026-01190-6

Keywords: hydrogen, zinc, energy storage, machine learning, lactate, electrochemistry, green hydrogen, hydrogen carriers, decentralized energy, Faradaic efficiency, Nature Synthesis, renewable energy