Rapid Parameter Estimation of Compact Binary Coalescences with Gravitational Waves

Abstract

In the age of multi-messenger astrophysics, fast, reliable information about gravitational-wave candidates is crucial for electromagnetic follow-up observations. While sky localization tells astronomers where to observe an event, source classification estimates the probability that the event might have an electromagnetic counterpart. Furthermore, astronomers need to have enough time to point their telescopes towards the fading light. Rapid PE is a low-latency parameter estimation scheme which parallelizes Bayesian inference by fixing the intrinsic parameters to a grid, and marginalizing over the extrinsic parameters at each grid point via Monte Carlo sampling. The gravitational-wave search pipelines identify the highest signal-to-noise ratio (SNR) template, which determines the initial grid region of Rapid PE. Adaptive mesh refinement (AMR) overcomes search biases placing additional grid points in areas with the highest likelihood calculations. This method provides posterior samples such as the masses and spins (intrinsic parameters) and sky location, distance, inclination, polarization, and phase (extrinsic parameters) in just under one minute per grid point with GPUs. From these Rapid PE results, we are able to provide accurate source classification estimates to astronomers in low latency, facilitating the detection of electromagnetic counterparts to gravitational waves.