2026/01/20 by N. Parmiggiani, A. Bulgarelli, G. Panebianco +17 · 1 voice
Physics and Astronomy · #Gamma-ray bursts and supernovae #Planetary Science and Exploration #Radiation Detection and Scintillator Technologies
paper · doi:10.3847/1538-4357/ae25fc
openalex publication_date 2026/01/20 · openalex created_date 2026/01/21 · openalex updated_date 2026/06/11
Abstract The Compton Spectrometer and Imager (COSI) is a NASA satellite mission under development designed to survey the entire sky at 0.2–5 MeV with a wide-field gamma-ray telescope. Its main instrument is a germanium detector array surrounded on the sides and bottom by bismuth germanium oxide scintillator active shields (the “Anticoincidence Subsystem” (ACS)) to reduce and monitor background and for detecting transients. COSI will have an onboard trigger algorithm to detect gamma-ray bursts (GRBs) in the ACS and send data to the ground for further analysis. In this paper, we present three localization methods that we evaluated for the localization of short GRBs (sGRBs) using the ACS light curves. The first method is the χ 2 fit already used by the Fermi Gamma-ray Burst Monitor, which calculates the best fit between look-up tables and the GRB data. The second method is a maximum likelihood estimation fit implemented in bc-tools for the BurstCube mission that performs a fit between the instrument response function and the GRB data. The last method is based on deep learning techniques and consists of a neural network developed for the COSI mission and trained to perform a regression of the sGRB position, taking as input the count rates of each ACS panel. The localization errors obtained by analyzing simulated sGRBs with the three methods are consistent. Despite the theoretical similarity between the approaches, their consistency in results is noteworthy, as they differ substantially in their implementations and optimization processes. The bc-tools obtain the best localization accuracy.