Possibilities to apply Machine Learning to electron beam phase space diagnostics at SwissFEL

by Philipp Dijkstal

Wednesday 14 Feb 2024, 11:30 → 13:00 Europe/Zurich

OHSA/E13

https://psich.zoom.us/j/63203692847
Meeting-ID: 632 0369 2847

Abstract:

SwissFEL is the X-ray Free Electron Laser at PSI, one of only six such facilities worldwide. Electron beams are accelerated up to 6 GeV beam energy and generate intense (tens of gigawatts) and short (down to sub-femtosecond) X-ray laser pulses, which enable cutting-edge research in a wide variety of applications. So far the potential of ML techniques to benefit facility operation and development is not exploited. The focus of this presentation is on the existing electron beam phase space diagnostics at SwissFEL, and how ML techniques could be used to improve their resolution and provide more detailed information. Three specific phase space diagnostics applications are discussed, all of which are based on the the analysis of 2D camera images of the transverse beam profile on a scintillating screen. First and simplest is the transverse emittance measurement, for which an existing ML algorithm [R. Roussel et al., https://doi.org/10.1103/PhysRevLett.130.145001] can be implemented. Second the measurement of the 5D (possibly 6D) phase space with the recently developed Polarix X-band transverse deflecting rf structure. This project requires some development but should be relatively straight-forward conceptually. The third and most challenging is the reconstruction of the 2D longitudinal phase space with a passive wakefield streaker, a method pioneered at PSI. The currently implemented iterative data analysis algorithms have limitations. Some of them could possibly be overcome with ML techniques, although the best approach is not clear.