Data-driven core collapse supernova multilateration with first neutrino events

TL;DR

The article addresses rapid localization of a Galactic core-collapse supernova using the relative arrival times of the first neutrino events across detectors. It develops a data-driven multilateration framework that uses the first-event times and a reference lightcurve to estimate time differences, corrects for bias due to detector yield differences with a closed-form correction, and quantifies directional uncertainty directly from data. Through a simple Monte Carlo framework with four detectors (SK, JUNO, LVD, SNO+) and IBD lightcurves, it demonstrates that the residual bias becomes negligible (well below a millisecond for most pairings) and that the resulting probability skymaps yield reasonably calibrated confidence intervals, typically spanning a few thousand square degrees. This method offers a fast, model-light initial pointing for SNEWS2.0, complementary to more precise direction estimates from later neutrino channels and elastic scattering, and supports rapid multi-messenger follow-up of a Galactic CCSN.

Abstract

A Galactic core-collapse supernova (CCSN) is likely to be observed in neutrino detectors around the world minutes to hours before the electromagnetic radiation arrives. The SNEWS2.0 network of neutrino and dark matter detectors aims to use the relative arrival times of the neutrinos at the different experiments to point back to the supernova so as to facilitate follow-up observation. One of the simplest methods to estimate the CCSN direction is to use the first neutrino events detected through the inverse beta decay (IBD) process, $\overlineν_e p\rightarrow e^+n$. We will consider neutrino detectors sensitive to IBD interactions with low backgrounds. The difference in signal arrival times between a large and a small detector will be biased, however, with the first event at the smaller detector, on average, arriving later than that at the larger detector. This bias can be mitigated by using these first events in a data-driven approach without recourse to simulations or models. The resulting method requires, at minimum, only the times of the first events at most detectors, along with a longer time series of events from one larger detector to act as a reference lightcurve. In this article, we demonstrate this method and its uncertainty estimate using pairs of detectors of different sizes and with different supernova distances. Finally, we use this method to calculate probability skymaps using four detectors currently in operation (Super-Kamiokande, JUNO, LVD, and SNO+) and show that the calculated probabilities yield appropriate confidence intervals for all supernova directions. The area of the 68\% confidence interval varies by distance and direction, but is expected to be a few thousand square degrees. The resulting skymaps should be useful for the multi-messenger community as a rapid, initial pointing to follow up on the SNEWS2.0 Galactic CCSN neutrino alert.

Figures (17)

• Figure 1: Before bias correction: distributions of first-event time differences (observed lags) less true lags, $t_1^A-t_1^B-\tau_{AB}$, over 100,000 trials with the $27M_\odot$ IBD lightcurve shape at 10 kpc. Pairwise differences are always computed such that detector $A$ is larger.
• Figure 2: Top: mean uncorrected bias as a function of relative detector yield for Super-Kamiokande compared with detector locations at JUNO (red/pink), LVD (green/lime green), and SNO+ (blue/cyan). For each value of $\langle N_B \rangle/\langle N_A \rangle$, 100,000 Monte Carlo trials have been run with IBD lightcurves for two progenitor masses ($27M_\odot$ and $11.2M_\odot$). In each case, the average Super-Kamiokande yield is held constant at the value in Table \ref{['tab:yields']}. Bottom: mean observed RMS of the uncorrected distributions.
• Figure 3: Top: mean uncorrected bias for Super-Kamiokande compared with other detectors at different distances, using Table \ref{['tab:yields']} for the yields at 10 kpc. Bottom: mean observed RMS of the uncorrected distributions.
• Figure 4: After bias correction: residual bias (corrected lag less true lag, $Z_{AB}-\tau_{AB}$) distributions over 100,000 trials with the $27M_\odot$ IBD lightcurve shape at 10 kpc.
• Figure 5: Top: mean residual bias as a function of relative detector yield, after correcting for the first-event bias, for Super-Kamiokande compared with detector locations at JUNO, LVD, and SNO+. Bottom: mean observed RMS after first-event bias correction.