Study of the Cosmic Ray Composition Sensitivity of AugerPrime

The AugerPrime upgrade of the Pierre Auger Observatory aims at enhancing the precision of primary particle composition measurements made by the Surface Detector. This is achieved by placing Surface-Scintilator-Detectors on top of the existing Water-Cherenkov-Detectors and comparing their differing responses to the electromagnetic and muonic components of extensive air showers as the ratio of these components is strongly related to the mass of the primary cosmic ray. While the deployment of the upgrade is ongoing, the composition sensitivity of AugerPrime can be probed using current machine learning techniques on simulations containing a mixed composition of protons, helium, oxygen, and iron.

In this presentation a deep learning approach is used to reconstruct the depth of shower maximum, X_max, an indicator of the primary mass. A convolutional neural network is developed and shown to be able to extract composition information from the difference in the signal pulses of the two different detector types on an event-by-event basis. The estimated bias and resolution of the reconstruction will be shown. The improvement in X_max resolution with AugerPrime is studied by comparing two networks trained with and without the additional detector information.