Lila Sciences’ AI Lab Uncovers Palladium Catalysts for Green Hydrogen

Lila Sciences said on September 25, 2026, that its AI-directed laboratory proposed, synthesized and screened 2,942 catalysts for green hydrogen in three months, identifying six palladium-based material families for the acidic oxygen evolution reaction at the heart of hydrogen production.

The research team detailed the work in a preprint submitted to arXiv on September 24, 2026: “AI-guided high-throughput discovery of iridium- and ruthenium-free palladium-oxide catalysts for durable acidic oxygen evolution,” by Ken J. Jenewein and 20 other authors.

Producing hydrogen from water requires two half-reactions: the hydrogen evolution reaction at the cathode and the oxygen evolution reaction, or OER, at the anode. OER runs at high potentials in strongly acidic environments that dissolve most metals. In today’s commercial electrolyzers, iridium oxide is the catalyst best able to survive those conditions while remaining active; ruthenium is a less stable option. Both are among the rarest elements on Earth, and nearly all iridium is produced as a byproduct of platinum mining, at only a few tonnes per year worldwide.

The Palladium Result

While screening 2,942 oxides across 53 material systems and 26 elements, the platform began synthesizing and testing palladium compositions. According to the company, palladium had been considered a dead end for OER because it was assumed to underperform in acidic conditions. The model’s candidates were modified palladium compositions rather than the pure metal, with small additions of other elements transforming the metal’s behavior. During the second and subsequent campaigns, the program identified six palladium-based material families on or near the Pareto front for OER, meaning an optimal balance of activity and stability.

After repeated activity tests and more than 1,000 hours of stability testing, Lila said the top metal composition performed comparably to ruthenium, an industry-standard OER catalyst. John Gregoire, the company’s chief autonomous science officer, said his former group at Caltech explored catalysts for this reaction for 12 years and that he would have advised a student proposing this combination of elements to pursue a more promising direction. “This combination defies all traditional wisdom,” he said in the company’s account of the work. Rafael Gómez-Bombarelli, Lila’s chief scientific officer of the physical sciences and the Paul M. Cook associate professor in materials science and engineering at MIT, said that without lab confirmation, the team would have dismissed the result as a hallucination.

The preprint reports iridium- and ruthenium-free complex oxides such as InMnPdOx and NiTaPdOx, with lead catalysts advanced to long-term validation in 1 M sulfuric acid at 10 milliamps per square centimeter. The authors report that NiTaPdOx operated at a lower overpotential than PdOx in long-term testing, though both eventually exceeded 0.5 V: PdOx at about 200 hours and NiTaPdOx at about 470 hours. InMnPdOx showed a similar overpotential improvement and what the authors describe as a dramatic increase in operational stability, retaining an overpotential below 0.5 V over 1,000 hours of operation.

The authors state that the additive elements promote the formation of a nanostructure associated with catalytic activity while stabilizing palladium against corrosion, and that palladium’s greater availability relative to iridium and ruthenium offers a near-term option to ease supply constraints on scaled electrochemical hydrogen generation.

Inside the AI Science Factory

Starting in late 2024, the team built what Lila calls an AI Science Factory, or AISF, to conduct electrocatalysis experiments autonomously. The workflow ran as a closed loop through four blocks: synthesis, pre-test characterization, testing and post-test characterization, with data from each block flowing back into the AI model to inform future decisions. Lila’s in-house reasoning model pairs Bayesian models, which reason well under uncertainty, with language models carrying broad context about how the world works; Gómez-Bombarelli said the interplay between the two balances uncertainty and information gain.

Each cycle began with the AI analyzing the chemical space and proposing recipes for new materials. Electrochemist Ken Jenewein initially reviewed the suggestions and overruled a few before phasing out his review; eventually, a scientist only checked that proposed materials were safe before synthesis. Once approved, the AISF synthesized 96 catalysts in parallel through physical vapor deposition, quality-checked each composition, ran accelerated OER screening in acid measuring activity and stability, then returned the materials for post-mortem characterization.

Integration with a lab information management system made every step and data point traceable across campaigns. The first campaign, run while the infrastructure was still being developed, lasted four months; the second took four weeks.

The company reported that its pipeline screened catalysts 17 times faster than a standard laboratory, running around 240 samples per week with a maximum capacity of 100 samples per day and saving more than 90% of human time per sample. The preprint describes the platform as exceeding 90% automation, integrating combinatorial sputter synthesis, high-throughput screening, machine-learning composition-property models, adaptive multi-objective optimization and context-aware large-language-model reasoning; in the authors’ retrospective benchmarking, their sequential learning agent advanced the activity-stability frontier faster than fixed-policy Bayesian optimization or in-context language-model selection.

Limitations and Next Steps

Lila states the result is a materials science discovery rather than a scaled-up commercial product, and that any OER catalyst needs to work at industrial scale under intense operating conditions. Gómez-Bombarelli said the team continues long-term durability testing and is evaluating, with AI guidance and lab equipment, how the material behaves in a form factor that resembles industrial-scale use. The workflow is also not yet completely autonomous: for this campaign, humans transferred samples between machines while the company develops robotics for fully autonomous operation.

Lila Sciences, a Cambridge-based AI startup, publicly launched in March 2025 with the stated bet that AI could run the full scientific method — hypothesis, experiment, interpretation and iteration — in autonomous, AI-controlled laboratories. Senior scientist Fae Habib Zadeh said the end-to-end workflow is generalizable across other electrocatalysis and electrosynthesis reactions, such as the creation of green ammonia or plastics precursors, and that the team is already running the next campaigns.