Machine-learning based X-ray spectrometer for High Repetition Rate Analysis of Betatron Radiation

Betatron radiation produced from a laser-wakefield accelerator is a broadband, hard X-ray (> 1 keV) source that has widely been used in a variety of applications including high resolution imaging and ultrafast spectroscopy. The characterization of betatron radiation is typically performed using X-ray filter packs (XFP), which consist of filters made of multiple materials and thicknesses. The standard analysis procedure utilizes a minimization algorithm that can require several seconds to reconstruct a single spectrum, and is therefore not feasible for real-time analysis of betatron X-ray sources at high repetition rate (> 1 Hz). We present the development deep learning algorithm for the analysis of an XFP spectrometer. The algorithm was developed using PyTorch with a training set of >10000 synthetic XFP images consisting of copper and aluminum filters. We discuss our progress towards fielding the deep learning algorithm for on-line source characterization at the Institut National de la Recherche Scientifique’s Advanced Laser Light Source.