Prompt sky localization of compact binary sources using a meshfree approximation

TL;DR

This paper introduces a rapid, mesh-free likelihood evaluation method for gravitational wave data analysis, enabling quick sky localization and parameter estimation of compact binary sources for multi-messenger astronomy.

Contribution

It extends previous mesh-free approximation algorithms to a ten-dimensional parameter space, improving speed and accuracy of gravitational wave source characterization.

Findings

Estimates BNS source properties in approximately 2.4-2.7 minutes.

Achieves rapid likelihood evaluation using 64 CPU cores.

Enhances posterior distribution accuracy through optimized node placement.

Abstract

The number of gravitational wave signals from the merger of compact binary systems detected in the network of advanced LIGO and Virgo detectors is expected to increase considerably in the upcoming science runs. Once a confident detection is made, it is crucial to reconstruct the source's properties rapidly, particularly the sky position and chirp mass, to follow up on these transient sources with telescopes operating at different electromagnetic bands for multi-messenger astronomy. In this context, we present a rapid parameter estimation (PE) method aided by mesh-free approximations to accurately reconstruct properties of compact binary sources from data gathered by a network of gravitational wave detectors. This approach builds upon our previous algorithm [L. Pathak et al., Fast likelihood evaluation using meshfree approximations for reconstructing compact binary sources, https://journals.aps.org/prd/abstract/10.1103/PhysRevD.108.064055, Phys. Rev. D 108, 064055 (2023)] to expedite the evaluation of the likelihood function and extend it to enable coherent network PE in a ten-dimensional parameter space, including sky position and polarization angle. Additionally, we propose an optimized interpolation node placement strategy during the start-up stage to enhance the accuracy of the marginalized posterior distributions. With this updated method, we can estimate the properties of binary neutron star (BNS) sources in approximately 2.4~(2.7) min for the \TaylorF~(\texttt{IMRPhenomD}) signal model by utilizing 64 CPU cores on a shared memory architecture. Furthermore, our approach can be integrated into existing parameter estimation pipelines, providing a valuable tool for the broader scientific community. We also highlight some areas for improvements to this algorithm in the future, which includes overcoming the limitations due to narrow prior bounds.