Searching For Anisotropic Gravitational-wave Backgrounds Using Pulsar Timing Arrays

TL;DR

This paper develops a Bayesian method to detect and characterize anisotropy in gravitational-wave backgrounds using pulsar timing arrays, enhancing the understanding of GW source distribution and local hotspots.

Contribution

It introduces a Bayesian search pipeline capable of identifying and quantifying anisotropic features in the gravitational-wave background affecting pulsar signals.

Findings

The pipeline can recover injected anisotropy in simulated data.

Assuming isotropy becomes invalid at high signal-to-noise ratios.

The method successfully constrains the nature of anisotropy in simulated backgrounds.

Abstract

We present the results of simulated injections testing the first Bayesian search-pipeline capable of investigating the angular-structure of a gravitational-wave (GW) background influencing pulsar signals. A stochastic background of GWs from the incoherent superposition of many inspiraling supermassive black hole binaries at nHz frequencies is likely to be the dominant GW signal detectable by pulsar timing arrays (PTAs). Even though one might expect a background composed of a high-redshift cosmological population of sources to be fairly isotropic, deviations from isotropy may be indicative of local GW hotspots or some form of continuous anisotropy in the angular-distribution of GW-power. A GWB induces time-of-arrival deviations in pulsar signals which are correlated between separated pulsars. In an isotropic background this cross-correlation follows a distinctive relationship, known as the Hellings and Downs curve, that depends only on the angular separation of the pulsars. If the background is anisotropic, the cross-correlation is different, but predictable, and also depends on the absolute position of the pulsars. By simulating datasets containing GWBs with various anisotropic configurations, we have explored the prospects for constraining anisotropy using near future data. We find that at moderate to high signal to noise ratio the assumption of isotropy is no longer an appropriate description of the simulated background. Furthermore, we can recover the nature of the injected anisotropy in a Bayesian parameter-estimation search, and propose a prior on the anisotropy search-space motivated by the physicality of the implied distribution of sources.