Bayesian optimization algorithms for ISOL ion source optimal tuning and characterization

ISOL@MYRRHA is the Isotope Separation On Line facility to be constructed in the first phase of the MYRRHA project. It will be capable to produce a large variety of radioactive isotopes for applications in the field of nuclear physics, condensed-matter physics, biology, nuclear medicine and others.

The quality (purity) and quantity (intensity) of the supplied radioactive ion beam (RIB) depend heavily on the proper tuning of the underlying process steps and their mutual interaction. From feedback of running ISOL facilities (ISOLDE/TRIUMF) it is known that the operation of an ISOL system needs constant intervention of an experienced operator/user. ISOL systems are complicated with many tens of parameters to tune beforehand and require experience and training to be able to do it optimally.

Due to component degradation, such as in the ion source or target, performance can change over time. Furthermore, operation may be suboptimal because operators typically don't explore the full tuning parameter space.

State-of-the-art algorithms such as the Gaussian process-based Bayesian optimization have shown high potential tuning radioactive ion beam lines and is currently being the focus of attention in facilities like TRIUMF (Canada), CERN (Switzerland), the Argonne National Laboratory (USA) and the future ISOL@MYRRHA, in SCK CEN in Belgium. Previous work done in a collaboration between SCK CEN and U. Antwerp resulted in a Gaussian Process based Basian optimization (GPBO) algorithms implemented in both SCK CEN ISOL@MYRRHA offline separator and CERN’s Offline 2 facilities for beam line tuning. Additionally, the work developed for ISOL@MYRRHA, in collaboration with CERN was briefly extended into tuning and characterizing electron impact ion sources. This PhD project will further deepen the application of GPBO algorithms into electron impact ion sources as well as tackle changes over time in ISOL systems, to ensure an optimal tuning at any point in time during operation.