Engineering Papers⌕ Search
Home/Archive/DOE OSTI

DOE OSTI · 3017592

Agentic artificial intelligence for multistage physics experiments at a large-scale user facility particle accelerator

Hellert, Thorsten [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)] (ORCID:0000000227970926)·Bertwistle, Drew [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)] (ORCID:0000000267044484)·Leemann, Simon C. [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)] (ORCID:0000000160354634)·Sulc, Antonin [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)] (ORCID:000000017767778X)·Venturini, Marco [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)] (ORCID:0009000388836813)
View original source ↗Read source document ↗Export metadata ↓

Abstract

We present a language-model-driven agentic artificial intelligence (AI) system to autonomously execute multistage physics experiments on a production synchrotron light source. Implemented at the Advanced Light Source particle accelerator, the system translates natural language user prompts into structured execution plans that combine archive data retrieval, control-system channel resolution, automated script generation, controlled machine interaction, and analysis. In a representative machine physics task, we show that preparation time was reduced by 2 orders of magnitude relative to manual scripting even for a system expert, while operator-standard safety constraints were strictly upheld. Core architectural features, plan-first orchestration, bounded tool access, and dynamic capability selection, enable transparent, auditable execution with fully reproducible artifacts. These results establish a blueprint for the safe integration of agentic AI into accelerator experiments and demanding machine physics studies, as well as routine operations, with direct portability across accelerators worldwide and, more broadly, to other large-scale scientific infrastructures.

Explore related subjects

Accelerator/storage ring control systemsArtificial intelligenceMachine learningNatural language processingParticle accelerator facilitiesParticle acceleratorsReliability & operabilitySynchrotron radiation & free-electron lasers

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Hellert, Thorsten [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)] (ORCID:0000000227970926), Bertwistle, Drew [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)] (ORCID:0000000267044484), Leemann, Simon C. [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)] (ORCID:0000000160354634), Sulc, Antonin [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)] (ORCID:000000017767778X), Venturini, Marco [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)] (ORCID:0009000388836813). 2026-01-16. Agentic artificial intelligence for multistage physics experiments at a large-scale user facility particle accelerator. https://doi.org/10.1103/jtqy-9jz1

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related reports

DOE OSTI2025

Demonstration of a code coupling framework for modeling beam-collimator impacts in the advanced photon source

The high-brightness beams being produced in current and future accelerators present new machine protection concerns with the potential for high-energy-density (HED) conditions ( >100 J/mm 3 ) in beam-intercepting components. Simulating HED conditions in accelerators requires utilizing a suite of physics codes for particle dynamics, particle-matter interactions, and hydrodynamics. This paper describes a method of coupling the codes elegant, fluka, and flash to simulate the effects of a rapid beam loss in the advanced photon source storage ring and the resulting interaction of the beam and collimators. This paper expands previous work [J. Dooling et al., Collimator irradiation studies at the advanced photon source, in Proceedings of the IBIC-2023 (2023), pp. 245–249] by introducing a definition of the evolving geometry of the collimator surface as well as providing methods for simulating the absorption of synchrotron radiation and tracking shower particles produced during beam strikes. We demonstrate this framework by simulating machine conditions of the APS ring before and after its recent upgrade. Simulation results are compared with observed damage to collimators and test samples taken from the APS ring.

Accelerator/storage ring control systems↗