How catalytic interfaces evolve, and how selectivity emerges — reasoned through artificial intelligence.
ISAAC (Integrated Scientific Agentic AI for Catalysis) unites complementary measurements from across the nation's scientific facilities with theory and the published record, in one continuous line of reasoning toward the grand challenges of the basic energy sciences.
The active state of a catalytic system is a moving target.
A working catalyst is governed at once by the catalyst, the interface, and the electrolyte, each obeying different physics. How that active state forms, and how competing pathways are steered toward one product, stands among the grand challenges the basic energy sciences have set for the field: a question of matter held away from equilibrium.
The complementary strengths of the nation's facilities, drawn into a single body of evidence.
DOE's light sources, neutron sources, and computational facilities were built to be complementary, each revealing what the others cannot. A result recorded at one facility and a measurement made at another are no longer separate findings. Set against a common schema, each becomes another piece of evidence in the same line of reasoning, the moment it is recorded.
The same approach reaches beyond any single system. Wherever complementary measurements can be drawn into common evidence, the reasoning applies, across catalysis and, in time, the wider basic energy sciences.
Reasoning that proposes, tests, and weighs scientific explanation.
ISAAC plans a line of inquiry and calls on specialized tools to pursue it: spectral analysis, first-principles calculation, and the digitized literature. It frames competing, falsifiable explanations and weighs each against the evidence those tools return. Its conclusions stay traceable to the evidence that earned them, and it can point to the next experiment the evidence calls for.
A scientist works alongside it, proposing a hypothesis, challenging a verdict, or steering the inquiry elsewhere. What the scientist brings is weighed on the same evidence as anything else, while the direction of the work, and the judgement of what holds, remain with the scientist. ISAAC extends the reach of that reasoning.
An explanation is held to the evidence over time.
Each verdict, whether from a measurement, a calculation, or the literature, is cited, and moves a confidence computed from the evidence. Where the evidence is thin, the reasoning can call for a new measurement and fold the result back in, closing the loop between hypothesis and experiment.
Conclusions held to a discipline of evidence.
An effort of the national laboratories, within the Genesis Mission.
As an AI Pathfinder project, ISAAC carries a charge that reaches beyond catalysis: to show, in working practice, what agentic AI can contribute to scientific discovery in the basic energy sciences.
Access to the ISAAC Repository, where the shared evidence is held, and the ISAAC Discovery Platform, where its agents reason, is held by the ISAAC team across the participating laboratories, and extends to selected DOE-funded efforts as the work scales. Researchers already supported by the Office of Basic Energy Sciences, or the Department of Energy more broadly, who wish to take part at this early stage are welcome to be in touch. Identity is verified through ORCID.
Conducted across the national laboratory complex.
Supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, through the AI Pathfinder initiative within the Genesis Mission.
Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information disclosed herein.