Bayesian Deep-stacking for High-energy Neutrino Searches

TL;DR

This paper introduces a Bayesian deep-stacking method for high-energy neutrino searches, enhancing detection sensitivity and property estimation of faint sources by combining data from many distant sources.

Contribution

It develops a Bayesian framework for deep-stacking in neutrino astronomy, improving detection and characterization of faint sources over traditional methods.

Findings

Bayesian approach outperforms maximum likelihood in sensitivity.

Method effectively reconstructs source properties.

Framework applicable to large populations of faint sources.

Abstract

Following the discovery of the brightest high-energy neutrino sources in the sky, the further detection of fainter sources is more challenging. A natural solution is to combine fainter source candidates, and instead of individual detections, aim to identify and learn about the properties of a larger population. Due to the discreteness of high-energy neutrinos, they can be detected from distant very faint sources as well, making a statistical search benefit from the combination of a large number of distant sources, a called deep-stacking. Here we show that a Bayesian framework is well-suited to carry out such statistical probes, both in terms of detection and property reconstruction. After presenting an introductory explanation to the relevant Bayesian methodology, we demonstrate its utility in parameter reconstruction in a simplified case, and in delivering superior sensitivity compared to a maximum likelihood search in a realistic simulation.