An experimental setup and Bayesian optimization for Target and Ion Source development

by Erwan Levillain (2^nd year PhD student)

In the ISOL (Isotope Separation On Line) method, RIBs (Radioactive Ion Beams) are produced using a beam of stable ions impinged on a thick target. Once stopped within the target material, radioactive atoms diffuse out and effuse toward a source where they are ionized before being accelerated and mass-separated. The resulting low-energy RIBs intensity depends on the TISS (Target and Ion Source System) release efficiency, which is governed by the competition between lifetime and release time of radionuclides. Maintaining high and homogeneous temperatures can speed up their diffusion from the target and enhance their surface desorption, minimizing the decay losses. To optimize the TISS temperature, thermal simulations can be a powerful tool, but their reliability depends on the accuracy of the temperature-dependent material properties involved. In this seminar, I will briefly introduce a renovated experimental setup [1] designed to measure the total hemispherical emissivity ε(T) and the electrical resistivity ρ(T) of material strips. I will mainly focus on the objective of the second series of measurements performed to capture temperature profiles along these strips: to infer the thermal conductivity k(T) using a parametric ‘numerical twin’ developed via PyMAPDL [2], the Python interface of Ansys. Bayesian Optimization [3] was employed to train a ‘surrogate model’ of the PyMAPDL instance, enabling the global exploration of the parameter space at a negligible computational cost.

[1] Venkateswarlu, K., Development of an innovative ISOL system for the production of short lived neutron-deficient ions, Thèse de doctorat, Université de Caen Normandie (2018).
[2] PyAnsys, Welcome to PyMAPDL, PyMAPDL documentation, [Online]. Available: https://mapdl.docs.pyansys.com/version/stable/index.html.
[3] R. Garnett, Bayesian Optimization. Cambridge: Cambridge University Press (2023). doi: 10.1017/ 9781108348973