An efficient pipeline for joint gravitational wave searches from individual binaries and a gravitational wave background with Hamiltonian sampling

TL;DR

This paper introduces a Hamiltonian Monte Carlo-based pipeline for joint gravitational wave searches from individual supermassive black hole binaries and a stochastic background, improving efficiency and accuracy in parameter estimation and sky mapping.

Contribution

The novel contribution is the development of an efficient Hamiltonian sampling pipeline that jointly analyzes signals from binaries and backgrounds, reducing computational costs and enhancing parameter exploration.

Findings

Accurate parameter estimation for simulated datasets at different frequencies.

Efficient generation of sky maps for binary sources.

Reduced number of analyses needed for real data from 72 to 1.

Abstract

The pulsar timing array community has recently reported the first evidence of a low-frequency stochastic gravitational wave background. With longer observational timespans we expect to be able to resolve individual gravitational wave sources in our data alongside the background signal. The statistical modeling and Bayesian searches for such individual signals is a computationally taxing task that is the focus of many different avenues of methods development. We present a pipeline for performing efficient joint searches for gravitational waves originating from individual supermassive black hole binaries as well as a gravitational wave background using a Hamiltonian Monte Carlo sampling scheme. Hamiltonian sampling proposes samples based on the gradients of the model likelihood, and can both converge faster to more complicated and high-dimensional distributions as well as efficiently explore highly covariant parameter spaces such as the joint gravitational wave background and individual binary model. We show the effectiveness of our scheme by demonstrating accurate parameter estimation for simulated datasets containing low- (6 nHz) or high- (60 nHz) frequency binary sources. Additionally we show that our method is capable at more efficiently generating skymaps for individual binary sources -- maps displaying the upper limits on the gravitational wave strain of the source, $h_{0}$, as a function of sky location -- by sampling over larger portions of the full sky. Comparing against results for the NANOGrav 12.5-year dataset, we find similar reconstructed upper limits on the gravitational wave strain while simultaneously reducing the number of required analyses from 72 independent binned searches down to a single run.