Improving Neutrino Point Source Sensitivity with Source-Informed Event Selection

Abstract

Neutrino telescopes employ multi-level reconstruction chains, where computationally expensive high-quality reconstructions are applied only to events that survive initial quality cuts based on fast, coarse directional estimates. Currently, event selection between reconstruction levels is source-agnostic, giving no priority to events from directions of known neutrino source candidates. We propose a simple modification to inter-level event selection: preferentially retain events whose early-level reconstruction places them within an angular tolerance of pre-specified candidate source directions from established multi-messenger catalogs, while continuing to subsample remaining events at the baseline rate. Using a realistic two-level detector model with energy-dependent angular resolution, we show that this source-informed selection can improve median point source sensitivity by factors of $\sim 2$--$3$ compared to uniform subsampling, with the improvement depending on the baseline selection efficiency, angular tolerance, and correlation between reconstruction qualities at different levels. For catalogs of $\mathcal{O}(100)$ sources, the additional computational overhead is modest ($\sim 7$--$14\%$). This approach offers a path to substantially enhance the discovery potential of current and future neutrino telescopes without requiring new detector capabilities.