Using machine learning to measure ultrahigh-flux multi-MeV gamma rays from a laser accelerator

In this thesis we present a novel computational method capable of measuring the energy distribution of ultrahigh-flux and high-energy photons ranging from 1-300MeV produced via a Thomson Backscatter process at the University of Texas at Austin Petawatt Laser facility. Due to the large and complex particle showers these kinds of photons produce when interacting with matter, energy measurements of these kinds of sources is notoriously difficult. In our method however, we make use of the complex particle showers these sources produce to extract information about the energy profile by interacting the photons with a compact inorganic scintillator. Then, using predictive simulations in Geant4 and regression analysis techniques, we analyze the raw scintillator response resulting from the incident photon shower, and compute the most likely photon energy spectrum with confidence intervals. In the following thesis, we will cover the methodology of this analysis as well as look at how it performs when applied to a recent experimental shot. Finally, we will compare the result to theoretical predictions in order to gauge the feasibility of this diagnostic method