Mitigating Polarization Leakage in Gas Pixel Detectors through Hybrid Machine Learning and Analytic Event Reconstruction

Abstract

Spatially resolved polarization measurements of extended X-ray sources are expanding our understanding of the emission mechanisms and magnetic field properties involved. Such measurements have been possible in the past few years thanks to the Imaging X-ray Polarimetry Explorer (IXPE). However, the analysis of extended sources suffers a systematic effect known as polarization leakage, which artificially affects the measured polarization signal. To address this issue, we built a hybrid reconstruction algorithm, which combines machine learning and analytic techniques to improve the reconstruction of photoelectron tracks in the Gas Pixel Detector and to significantly mitigate polarization leakage. This work presents the first application of this hybrid method to experimental data, including both calibration lab measurements and IXPE observational data. We confirmed the reliable performance of the hybrid method for both cases. Additionally, we demonstrated the algorithm's effectiveness in reducing the polarization leakage effect through the analysis of the IXPE observation of the supernova remnant G21.5-0.9. By enabling more reliable polarization measurements, this method can potentially yield deeper insights into the magnetic field structures, particle acceleration processes, and emission mechanisms at work within extended X-ray sources.