Evaluation of direct inversion of proton radiographs in the context of cylindrical implosions

Direct inversion of deflectometry data, such as proton radiographs and shadowgraphs, is a well-posed problem with a unique solution for the transverse deflection of each particle or ray if their trajectories do not cross. When trajectories cross, there exists an infinite set of solutions. In proton radiography direct inversion determines the line-integrated transverse Lorentz force. We have tested five publicly available direct inversion routines with a view to analyzing proton radiographs of cylindrical implosions on the OMEGA laser; four Monge-Ampère solvers [github.com/flash-center/PRaLine, github.com/flash-center/PROBLEM, github.com/mfkasim1/invert-shadowgraphy/tree/fast-inverse, github.com/OxfordHED/proton-radiography-no-source], and a power-diagram method [github.com/mfkasim1/invert-shadowgraphy]. Test problems were generated using four field profiles, three cylindrical and one spherical, with varying field amplitudes in proton-tracing routines. Two Monge-Ampère solvers did not run, the other two failed to reproduce radiographs when trajectories crossed, although for one field profile the solutions only diverged from the original at the boundary. The power-diagram method was successful even when proton trajectories crossed, giving a solution that minimized proton deflection, but failed for profiles that produced a single, sufficiently sharp peak. For cases where trajectories do not cross, the Monge-Ampère solvers have the advantage of being considerably faster than the power-diagram routine, up to $1000$ times in our tests. The test problems are provided in the supplementary information in pradformat [github.com/phyzicist/pradformat].