AI-driven lab finds iridium-free palladium catalyst InMnPdOx stable for 1,000 hours
CatAstro_Piyush · x · 2026-09-26
Lila's new preprint (arXiv 2609.30133) reports an AI-guided closed-loop platform discovering iridium- and ruthenium-free palladium-oxide catalysts for acidic oxygen evolution.
- Platform is >90% automated, combining combinatorial sputter synthesis, high-throughput screening, ML composition-property models, adaptive multi-objective optimization, and LLM reasoning
- Synthesized and screened 2,942 compositions across 53 material systems and 26 elements
- Lead candidate InMnPdOx held <500 mV overpotential for 1,000 hours at 10 mA/cm² in 1 M H₂SO₄; unmodified PdOx crossed that threshold at 200 hours
- Counterintuitive: electrochemists wouldn't have prioritized palladium, but adding indium and manganese transformed activity and stability—a combination neither conventional design logic nor off-the-shelf LLMs predicted
- Industrial electrolyzer validation remains ahead
Blog write-up accompanies the paper.
More from Research
- Claude computes nine-loop scattering amplitude, breaking Lance Dixon's eight-loop record — kevinweil · 2026-09-26
- Stanford's Matryoshka Attribution tops mech interp benchmark, removes refusal by resetting 1% of weights — stanfordnlp · 2026-09-26
- Dual Covariance 3DGS SLAM: two covariances per Gaussian, robust tracking at 60 FPS — kwangmoo_yi · 2026-09-26
- Dual Covariance Gaussian Splatting SLAM decouples rendering and registration — kwangmoo_yi · 2026-09-26
- ImageJevBench: image decision benchmark ranks top models, full eval costs $0.02 — airesearch12 · 2026-09-26
- Biopharma Bench: agents complete only 8 of 71 real biopharma tasks, GPT-6 Astra leads — AllThingsApx · 2026-09-26