Pre-merger sky localization and parameter estimation of binary neutron star inspirals using normalizing flows

Abstract

In the coming years, the LIGO and Virgo gravitational wave interferometers are planned to undergo a number of detector upgrades which will improve the detector sen- sitivity, and together with the construction of next-generation detectors like Einstein Telescope and Cosmic Explorer this should lead to more frequent and louder detec- tions of binary neutron star inspirals. To increase the information gained from these observations, we want to study the inspiral and merger directly with electromagnetic telescopes, which requires us to detect and localize these signals before their merger. Current state-of-the-art localization algorithms rely on matched filtering frameworks, which can introduce biases and only provide point estimates of intrinsic parameters. In this work, we present a normalizing flows based framework that can provide pre-merger sky location, in addition to estimating other parameters relevant for follow-up study, such as the component masses, luminosity distance and inclination angle. We train net- works for different maximum frequencies, corresponding to a different time-to-merger. The parameter estimations are better constrained when a larger part of the signal is ob- served and when the signal is louder, but the networks can also produce well-constrained localizations for smaller parts and quieter signals. The sky localizations produced by the networks are regularly accurate enough to enable follow-up studies.