Beyond Structure Generation: Toward Reaction-Aware Generative AI for Catalyst Discovery
Abstract Generative artificial intelligence (AI) has expanded catalyst discovery from predefined libraries to ultralarge surface–adsorbate spaces. We contend that as generative workflows reach the million-structure scale, the central challenge is shifting from increasing generation scale to identifying catalytically relevant regions of chemical space. This calls for reaction-aware discovery: candidates must remain viable across coupled intermediates, competing pathways, and operating conditions, while predictions must be reliable enough to guide the next decision. This requires generative models to be coupled with hierarchical evaluation, mechanistic and kinetic modeling, and experimental feedback. Crucially, we distinguish automation from autonomy: autonomous discovery begins when new evidence changes what is generated, calculated, or tested next. Progress should be judged not by the number of generated structures, but by how effectively a discovery system converts chemical-space exploration into experimentally credible catalytic decisions.
Authors
- Yi Gao (ORCID: https://orcid.org/0000-0001-6015-5694)
- Beien Zhu (ORCID: https://orcid.org/0000-0002-0126-0854)
- Rui Qi (ORCID: https://orcid.org/0000-0002-2727-1770)
- Ruili Li
Institutions
- Chinese Academy of Sciences (CN)
- Shanghai Advanced Research Institute (CN)
- Shanghai Institute of Applied Physics (CN)
- University of Chinese Academy of Sciences (CN)
Publication Details
- Journal
- ACS Applied Catalysis
- Published
- 2026-11-12
- DOI
- https://doi.org/10.1021/acsappcat.6c00160
- Primary Topic
- Machine Learning in Materials Science
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Natural Science Foundation of China
- Science and Technology Commission of Shanghai Municipality
- National Key Research and Development Program of China
- Shanghai Advanced Research Institute, Chinese Academy of Sciences