Study of mass composition of cosmic rays with IceTop and IceCube

Abstract

The IceCube Neutrino Observatory is a multi-component detector at the South Pole which detects high-energy particles emerging from astrophysical events. These particles provide us with insights into the fundamental properties and behaviour of their sources. Besides its principal usage and merits in neutrino astronomy, using IceCube in conjunction with its surface array, IceTop, also makes it a unique three-dimensional cosmic-ray detector. This distinctive feature helps facilitate detailed cosmic-ray analysis in the transition region from galactic to extragalactic sources. We will present the progress made on multiple fronts to establish a framework for mass-estimation of primary cosmic rays. The first technique relies on a likelihood-based analysis of the surface signal distribution and improves upon the standard reconstruction technique. The second uses advanced methods in graph neural networks to use the full in-ice shower footprint, in addition to global shower-footprint features from IceTop. A comparison between the two methods for composition analysis as well as a possible extension of the analysis techniques for sub-PeV cosmic-ray airshowers will also be discussed.

Graph Neural Networks and Application for Cosmic-Ray Analysis

“…It was discussed how GNNs bridge the gap between MLPs (Multilayer Perceptrons) and Convolutional Neural Networks (CNNs), benefiting from the best principles and methods in both domains. Preliminary results for Cosmic-Ray Composition analysis at IceCube Observatory show promising results and more detailed results of this ongoing work are discussed in [34]. In the future, a full cosmic-ray energy spectrum and composition analysis at IceCube, in the transition region from galactic to extragalactic sources, using GNNs is envisioned.…”

“…As described in earlier sections, GNNs surely hold promise in their capability to allay the limits of MLPs and CNNs. IceCube already has few ongoing analysis using GNNs in cosmic-ray physics [34,35] and neutrino physics [36,37]. Results of one such preliminary test using GNNs for CR primary-mass prediction at IceCube will be presented here.…”

“…Also, only the in-ice DOMs are mapped as a graph. To capture information from IT as well as from global features associated to each graph, another MLP based sub-component is used (details in [34]).…”

“…Preliminary tests show a an improvement in mass resolution and preciseness in prediction of mass. More details of the work are presented in [34].…”

“…Figure5: KDE of the natural logarithm of the cosmic-ray primary mass as predicted by the GNN-based method, trained on equal fraction of H and Fe simulations (vertical dashed lines: true primary mass). Plot taken from[34].…”