A non-parametric approach to infer the energy spectrum and the mass composition of cosmic rays

Abstract

The experiment KASCADE observes simultaneously the electron-photon, muon, and hadron components of high-energy extensive air showers (EAS). The analysis of EAS observables for an estimate of energy and mass of the primary particle invokes extensive Monte Carlo simulations of the EAS development for preparing reference patterns. The present studies utilize the air shower simulation code CORSIKA with the hadronic interaction models VENUS, QGSJet and Sibyll, including simulations of the detector response and efficiency. By applying non-parametric techniques the measured data have been analyzed in an event-by-event mode and the mass and energy of the EAS inducing particles are reconstructed. Special emphasis is given to methodical limitations and the dependence of the results on the hadronic interaction model used. The results obtained from KASCADE data reproduce the knee in the primary spectrum, but reveal a strong model dependence. Owing to the systematic uncertainties introduced by the hadronic interaction models no strong change of chemical composition can be claimed in the energy range around the knee.

Discrepant Hardening Observed in Cosmic-Ray Elemental Spectra

“…The data from subsequent flights are still being analyzed. These direct measurements bridge the energy gap between lower-energy direct measurements and the abundant indirect measurements at higher energies from the ground (e.g., Antoni et al 2002).…”

Multifractal nature of extensive air showers

“…The method of multifractal moment analysis has been applied to simulated data as observed by KASKADE experiment to distinguish EAS lateral distributions due to various types of primaries [12]. The fractal moments are part of the sample of observables used for KASKADE data anlysis [13].…”

On the possibility of primary identification of individual cosmic ray showers

“…This is usually done comparing the experimental data with simulated data. For this purpose non-parametric techniques like Bayes classifiers are very useful tools [25,26]. From the experimental data or even from simulations, several observable parameters, chosen such that they are very sensitive to the chemical composition of the primary, can be obtained.…”

“…In this way, 20 different values for protons and 20 for iron nuclei of the number of events correctly classified are obtained (see Eqs. (25,26,27)): n 0 pr (s) and n 0 f e (s) with s = 1, . .…”

“…Several multi-parametric techniques for the mass identification of individual showers have been studied and discussed in the literature. The most popular include non-parametric density estimates [25,26,27] and neural networks [25,28]. In particular, in this work we use a non-parametric density estimate technique of Gaussian kernel superposition, improved by adaptive choice of the smoothing parameters, in order to estimate the identification probability of individual nuclei, under the assumption of a binary proton and iron sample.…”