Exploitation of Crystal Shadowing via Multi-Crystal Array, Optimisers and Reinforcement Learning

The CERN Super Proton Synchrotron (SPS) routinely delivers proton and heavy ion beams to the North experimental Area (NA) in the form of 4.8 s spills. To produce such a long flux of particles, resonant third integer slow extraction is used, which, by design, foresees primary beam lost on the electrostatic septum wires to separate circulating from extracted beam. Shadowing with thin bent crystal has been proposed and successfully tested in the SPS, as detailed in *. In 2021, a thin crystal was used for physics production showing results compatible with what measured during early testing. In this paper, the results from the 2021 physics run are presented also comparing particle losses at extraction with previous operational years. The setting up of the crystal using numerical optimisers is detailed, with possible implementation of reinforcement learning (RL) agents to improve the setting up time. Finally, the full exploitation of crystal shadowing via multi-array crystals is discussed, together with the performance reach in the SPS.