Extraction of the muon signals recorded with the surface detector of the Pierre Auger Observatory using recurrent neural networks

Abstract

The Pierre Auger Observatory, at present the largest cosmic-ray observatory ever built, is instrumented with a ground array of 1600 water-Cherenkov detectors, known as the Surface Detector (SD). The SD samples the secondary particle content (mostly photons, electrons, positrons and muons) of extensive air showers initiated by cosmic rays with energies ranging from 1017eV up to more than 1020eV. Measuring the independent contribution of the muon component to the total registered signal is crucial to enhance the capability of the Observatory to estimate the mass of the cosmic rays on an event-by-event basis. However, with the current design of the SD, it is difficult to straightforwardly separate the contributions of muons to the SD time traces from those of photons, electrons and positrons. In this paper, we present a method aimed at extracting the muon component of the time traces registered with each individual detector of the SD using Recurrent Neural Networks. We derive the performances of the method by training the neural network on simulations, in which the muon and the electromagnetic components of the traces are known. We conclude this work showing the performance of this method on experimental data of the Pierre Auger Observatory. We find that our predictions agree with the parameterizations obtained by the AGASA collaboration to describe the lateral distributions of the electromagnetic and muonic components of extensive air showers.

Ultra-High-Energy Cosmic Rays: The Intersection of the Cosmic and Energy Frontiers

“…Conversely, the size of shower-to-shower fluctuations in the muon number as measured by the Observatory does agree with model predictions, indicating that the mismatch in the average cannot be due only to a major mis-modeling of extreme-energy interactions at the top of the shower, but must be due to a small effect compounding throughout the shower development, including in lower-energy interactions close to the ground [55]. The X max and muon content of showers can also be estimated from SD data using machine learning techniques [56,57], and the new AugerPrime detectors are going to further reduce statistical and systematic uncertainties on the UHECR mass composition, shed more light on hadronic interactions at extreme energies, and allow us to compile proton-enhanced samples of events for anisotropy studies.…”

“…Furthermore, improved analysis techniques based on machine learning approaches [57] are expected to exclude, or strongly constrain models of muon production in hadronic interactions throughout the next decade. This combination of new but well understood experimental methods and new analysis techniques, will lead to very precise measurements of the muon component in air showers and subsequently to an understanding of the origin of the Muon Puzzle.…”

“…This enables improved composition studies at the highest energies compared to those possible with classic SD analyses, for example the interpretation of the signal rise time [172]. The second method [57] aims to directly extract the muon signals recorded by each WCD using recurrent neural networks as the total number of muons produced in a shower N µ is strongly correlated with primary mass and is subject to lower shower-to-shower fluctuations than X max . In simulations it was found that the network was able to estimate the fraction of the total WCD signal contributed by muons with a bias of less than 2 % and a resolution better than 11 %.…”

Deep learning method for identifying mass composition of ultra-high-energy cosmic rays

“…Similarly to [38], we use a small fully connected neural network for setting the initial state of the recurrent cells analyzing the signal in the signal encoder. The input data for this network includes 6 parameters: x, y, and z coordinates of the detector, reconstructed time of the plane front arrival, time difference between starting point of detectors' recordings and the time of plane front arrival, and reconstructed zenith angle of the primary particle.…”

“…The first one is to identify a more reliable hadronic model by comparing some observable quantity in Monte-Carlo simulations and real data. In the ongoing researches in this direction one compares the simulated and measured profiles of extensive air showers [24,28] or the recorded muon component of the signal [23,27,38].…”