Expectation on probing the origin of the cosmic ray knee with the LHAASO experiment

Abstract

The cosmic-ray (CR) knee and the compositions contain abundant information for probing the CR's origin, acceleration and propagation mechanisms, as well as the frontier of the fundamental physics. Realizing that major proposals toward the knee's shape can be divided into two categories: the rigidity-dependent (also Z-dependent) knee and the mass-dependent (also A-dependent) knee, where the former one relates to the acceleration or the propagation mechanisms, and the other one is often associated with the new physics, it is essential to precisely measure the individual compositions. Benefit from the high altitude and hybrid detection methods, the LHAASO experiment has the ability in determining the individual component and brings us an opportunity in discriminating these two models. We test this expected ability of LHAASO from 100 TeV to 10 PeV with 3-year observation. And find the dominant component at the knee is essential to this issue, while much heavier nuclei occupying the knee leads to higher significance. In the analysis, the He-dominant knee under the A-dependent case can be recognized at the significance about 6.6 σ, while the P-dominant knee under the Z-dependent case will be classified with 2 σ significance.

Classifying the Cosmic-Ray Proton and Light Groups on the LHAASO-KM2A Experiment with the Graph Neural Network

“…We illustrate the expected observation with 1-day operation of the KM2A on the proton-and light-group spectra in Fig. 8, where the rigidity-and mass-dependent knee models are adopted from [21,47]. We further construct a simple CNN model for comparing with the GNN model.…”

“…The triangular markers represent the spectra predicted by one of the rigidity-dependent knee models (Z) [47], and the square markers represent the spectra predicted by one of the mass-dependent knee models (A) [21] ditional machine learning method such as the Gradient Boosted Decision Tree (GBDT) [28]. In comparison with the hybrid method, we set the models in this work at the roughly same condition and compare their signal purity and the array apertures.…”

“…Precise localization of the component knee is the key issue for exploring the hidden physics. Current explanations of the CR knee can be classified into tow categories with different mechanisms, including the mass-dependent knee and rigidity-dependent knee models [21], where models with the rigidity-dependent knee are often considered originating from the acceleration limit and the galactic leakage mechanism, and many of the models with the massdependent knee are associated with the new physics. Although much efforts have been paid aiming at resolving this issue, those experimental measurements still make large discrepancy with each other [22][23][24][25].…”

“…In this work, we lever-age the GNN to improve the CR-component classification performance on the LHAASO-KM2A experiment, where the detector activated by the event is formed as the graph-structured data. Our previous work [21] manifested that the science issue requires high accuracy in classifying the CR Proton (P task) and Light-component (L task) from the background. Hence, we focus on these two tasks in this work.…”

Classifying cosmic-ray proton and light groups in LHAASO-KM2A experiment with graph neural network *

“…Precise measurement of the cosmic-ray (CR) spectrum and its components at the PeV scale is essential to probe the CR origin, acceleration, and propagation mechanisms, as well as to explore new physics. A spectral break at ~ 4 PeV, referred to as the CR knee, was discovered 60 years ago [20]; however, its origin remains a mystery. Precise localization of the knees of different chemical compositions is key to explore the hidden physics.…”