Graduate Colloquium: Anthony Tran "Towards a Higher-Fidelity Phase Space Reconstruction in Accelerators"

Modern accelerator facilities frequently rely on simple RMS-based diagnostics and Gaussian approximations to characterize beam distributions. However, as accelerators push toward higher beam power and increased brightness, these simplified models become insufficient. Characterizing and controlling the beam at the level of individual particle distributions is now essential for maximizing power and preventing beam loss on sensitive components. This thesis addresses this challenge by developing novel algorithms for efficient phase-space reconstruction and implementing them experimentally at the Argonne Tandem Linac Accelerator System (ATLAS). The work evaluates three distinct approaches to achieve high-fidelity beam characterization. First, leveraging Machine Learning (ML), we developed models to optimize beam transmission and provide virtual diagnostics. Second, we created a robust analysis code for a pepper-pot detector, enabling accurate, "one-shot" direct characterization of the phase space. Finally, we investigated maximum entropy tomography as an indirect method, showing it provides higher-fidelity reconstruction compared to traditional quadrupole scans.