Stereograph: Stereoscopic event reconstruction using graph neural networks applied to CTAO

TL;DR

This paper introduces a novel graph neural network approach for stereoscopic event reconstruction in the CTAO gamma-ray observatory, demonstrating improved accuracy over traditional methods using simulated data.

Contribution

It proposes using Graph Neural Networks to enhance event reconstruction by effectively combining observations from multiple telescopes in CTAO.

Findings

GNNs outperform random forests in energy and angular resolution

GNNs improve gamma-proton separation accuracy

Enhanced stereoscopic reconstruction quality

Abstract

The CTAO (Cherenkov Telescope Array Observatory) is an international observatory currently under construction. With more than sixty telescopes, it will eventually be the largest and most sensitive ground-based gamma-ray observatory. CTAO studies the high-energy universe by observing gamma rays emitted by violent phenomena (supernovae, black hole environments, etc.). These gamma rays produce an atmospheric shower when entering the atmosphere, which emits faint blue light, observed by CTAO's highly sensitive cameras. The event reconstruction consists of analyzing the images produced by the telescopes to retrieve the physical properties of the incident particle (mainly direction, energy, and type). A standard method for performing this reconstruction consists of combining traditional image parameter calculations with machine learning algorithms, such as random forests, to estimate the particle's energy and class probability for each telescope. A second step, called stereoscopy, combines these monoscopic reconstructions into a global one using weighted averages. In this work, we explore the possibility of using Graph Neural Networks (GNNs) as a suitable solution for combining information from each telescope. The "graph" approach aims to link observations from different telescopes, allowing analysis of the shower from multiple angles and producing a stereoscopic reconstruction of the events. We apply GNNs to CTAO-simulated data from the Northern Hemisphere and show that they are a very promising approach to improving event reconstruction, providing a more performant stereoscopic reconstruction. In particular, we observe better energy and angular resolutions(before event selection) and better separation between gamma photons and protons compared to the Random Forest method.