Towards inference of overlapping gravitational wave signals

TL;DR

This paper demonstrates that current gravitational wave data analysis methods can accurately infer overlapping signals from binary mergers if their coalescence times are separated by more than 1-2 seconds, highlighting challenges for closely overlapping events.

Contribution

It shows that using coalescence time priors, existing infrastructure can effectively disentangle overlapping gravitational wave signals, identifying the limits of current methods.

Findings

Parameter inference is robust for signals separated by more than 1-2 seconds.

Signals within 0.5 seconds of each other cause significant biases.

Current methods need new strategies for closely overlapping signals.

Abstract

Merger rates of binary black holes, binary neutron stars, and neutron star-black hole binaries in the local Universe (i.e., redshift $z=0$), inferred from the Laser Interferometer Gravitational Wave Observatory (LIGO) and Virgo, are $16-130\, \mathrm{Gpc}^{-3}\,\mathrm{yr}^{-1}$, $13-1900\, \mathrm{Gpc}^{-3}\,\mathrm{yr}^{-1}$, and $7.4-320\,\mathrm{Gpc}^{-3}\,\mathrm{yr}^{-1}$, respectively. These rates suggest that there is a significant chance that two or more of these signals will overlap with each other during their lifetime in the sensitivity-band of future gravitational-wave detectors such as the Cosmic Explorer and Einstein Telescope. The detection pipelines provide the coalescence time of each signal with an accuracy $\mathcal{O}(10\,\rm ms)$. We show that using a prior on the coalescence time from a detection pipeline, it is possible to correctly infer the properties of these overlapping signals with the current data-analysis infrastructure. We study different configurations of two overlapping signals created by non-spinning binaries, varying their time and phase at coalescence, as well as their signal-to-noise ratios. We conclude that, for the scenarios considered in this work, parameter inference is robust provided that their coalescence times in the detector frame are more than $\sim 1-2 \,\mathrm{s}$. Signals whose coalescence epochs lie within $\sim 0.5\,\mathrm{s}$ of each other suffer from significant biases in parameter inference, and new strategies and algorithms would be required to overcome such biases.