A composition-informed search for large-scale anisotropy with the Pierre Auger Observatory

TL;DR

This paper investigates how the large-scale anisotropy of ultra-high-energy cosmic rays varies with their composition, using data-driven estimators and simulations to identify potential differences in dipole amplitudes between lighter and heavier elements.

Contribution

It introduces a method to analyze composition signatures in cosmic ray anisotropy by dividing data into lighter and heavier element subsets using novel estimators and simulation-based evaluation.

Findings

Potential to measure dipole amplitude differences between composition groups

Use of two composition estimators: universality-based and deep learning-based

Simulation results support the feasibility of composition-dependent anisotropy analysis

Abstract

The large-scale dipolar structure in the arrival directions of ultra-high-energy cosmic rays with energies above $8\,$EeV observed by the Pierre Auger Collaboration is a well-established anisotropy measurement. This anisotropy is understood to be of extragalactic origin, as the maximum of the dipolar component is located ${\sim}115^\circ$ away from the Galactic Center. Cosmic rays interact with background radiation and magnetized regions on their path from their sources to Earth. These interactions, which depend on the cosmic-ray energy, charge and mass composition, give rise to different horizons and deflections that are expected to lead to different anisotropies in the arrival directions of cosmic rays at Earth. The Auger Collaboration has determined that the mass composition of cosmic rays at ultra-high energies is mixed, becoming increasingly heavier as a function of energy. Thus, different dipole amplitudes are expected to be measured at a given energy when separating the data into composition-distinct subsets of lighter and heavier elements. In this contribution, we investigate the composition signature on the large-scale anisotropy taking advantage of composition estimators obtained from the data gathered with the surface detector. A way of probing for composition signatures in anisotropy patterns is then to divide the data into subsets of ``lighter'' and ``heavier'' elements per energy bin. In a simulation library, we evaluate the possibility of measuring a separation in total dipole amplitude between two such populations of the measured dataset under a source-agnostic model. We present the results using two different composition estimators, one based on air-shower universality and one inferred with deep learning.