Snapshots of ML-optimized Plasma Wakefield Accelerators

Plasma Wakefield Acceleration (PWFA) is a promising technology aimed at producing compact and cost-efficient particle accelerators for high energy physics as well as medicine, material science, nuclear physics, and defense applications. Research at the FACET-II facility at SLAC is underway to study how to preserve the quality of the accelerated beam. Sextupole magnets, while capable of precise beam shaping, are difficult to tune given the large number of degrees of freedom and the sensitivity of PWFA performance to small changes. Our first goal is to develop a beam synchronized Machine Learning (ML) algorithm to optimize over the sextupole parameter space, while working around accelerator phase jitter that would compromise a simple optimizer. Our second goal is to develop a dark-shadowgraphy plasma diagnostic. When using a probe laser to probe a plasma, placing a mask near the Fourier plane can remove noise by O(106), allowing us to image the fine features of a plasma wave. Importantly, we aim to directly observe how PWFA can be transformed into a wakeless wave appropriate for coherent light sources. Thus, while ML allows us to optimize PWFA performance, dark-shadowgraphy enables us to directly observe internal plasma dynamics, resulting in a rich and interdisciplinary scientific program.