Generalize tomography sampling requirements to 6D reconstruction from 2D images

Determine how the conventional tomography rule that evenly spaced angular rotations over 360 degrees are needed for accurate 2-dimensional phase-space reconstruction from 1-dimensional projections translates to reconstructing 6-dimensional electron-beam phase-space distributions from 2-dimensional screen images, including scenarios that employ nonlinear beamline manipulations such as a transverse deflecting cavity and a dipole spectrometer.

Background

The paper introduces a generative phase space reconstruction (GPSR) method that combines a generative neural network with differentiable beam dynamics to recover detailed 6-dimensional (6D) electron-beam phase-space distributions from a small number of 2-dimensional (2D) screen images gathered via quadrupole scans and longitudinal phase space diagnostics (transverse deflecting cavity and dipole spectrometer).

While classical tomography provides a rule-of-thumb for 2D reconstructions from 1D projections (even angular sampling over 360 degrees), the authors highlight that it is not clear how analogous sampling requirements extend to 6D reconstructions from 2D images, particularly when nonlinear beam manipulations are involved. Clarifying these requirements would enable principled uncertainty quantification and guide the design of minimal yet sufficient measurement sets for accurate 6D reconstruction.