Speaker
Description
The origin of ultra-high-energy cosmic rays (UHECRs) is one of the greatest mysteries in astroparticle physics. Identifying their sources requires a reliable understanding of the UHECR mass composition. Since UHECRs cannot be measured directly, the estimation of the UHECR mass is non-trivial, necessitating the use of mass-sensitive observables, such as the atmospheric depth of the shower maximum and the number of secondary muons. However, these observables encode the mass information only stochastically, making high-statistics measurements essential. The Pierre Auger Observatory is the largest UHECR detector in the world. The Surface Detector (SD) of the Observatory comprises three sub-arrays of multi-detector stations achieving an uptime of nearly 100%. Arranged in differently spaced triangular grids, the arrays measure the time-dependent particle distribution of extensive air showers. In this contribution, we show how neural networks are utilized in the Pierre Auger Collaboration to analyze the complex spatio-temporal information contained in shower footprints. Using SD simulations, we demonstrate how to extract the muon signal fraction from individual triggered stations and how to combine the information of multiple stations to reconstruct various mass-sensitive observables on an event-by-event basis. Moreover, using the depth of the shower maximum, we exemplify how to apply the methodology to Phase-I SD measurements, demonstrate the excellent agreement with independent measurements from the Fluorescence Detector (FD) of the Observatory, and show how the method extends the measurement to higher energies than those probed by the FD.