Enhancing Radioactive Beam Transport at CARIBU through AI-driven Automation

The Californium Rare Isotope Breeder Upgrade (CARIBU) at Argonne National Laboratory is a cornerstone facility for advancing nuclear physics research, providing radioactive ion beams (RIBs) derived from the spontaneous fission of Californium-252. These beams enable groundbreaking studies of rare and unstable atomic nuclei, contributing to our understanding of nuclear structure, astrophysical processes, and national security applications. Traditionally, the extraction and transport of these beams have depended on expert-driven manual tuning, a process that involves optimizing hundreds of parameters and is inherently time-consuming, limiting operational efficiency and scientific throughput.

To overcome these challenges, we have developed an innovative system that leverages Artificial Intelligence (AI) to automate the beam delivery process at CARIBU. By employing machine-learning techniques, specifically Bayesian Optimization, our system autonomously tunes each section of the beamline, achieving transport efficiency and delivery times comparable to those of experienced experts. Early implementations of this AI-driven approach have demonstrated significant improvements in operational efficiency, paving the way for enhanced scientific output.

This presentation will provide an update on the CARIBU-matic project, showcasing results from live tests and exploring future directions for integrating AI-driven optimization into other beamline sections and instruments. By implementing these methodologies, we aim to advance the field of nuclear physics, accelerating research and fostering autonomous scientific discovery.