Disentangling Multiple Stochastic Gravitational Wave Background Sources in PTA Datasets

TL;DR

This paper evaluates the ability of current PTA analysis methods to distinguish and characterize multiple overlapping gravitational wave backgrounds from different astrophysical sources using simulated datasets.

Contribution

It demonstrates the separability of multiple GWBs in PTA data and provides a protocol for future searches of multiple backgrounds.

Findings

Bayesian analysis can recover weaker backgrounds with 20 years of data.

The Bayes' factor for a second process exceeds unity at 17 years.

Spectral index and amplitude of weaker GWB constrained with 64% and 110% uncertainty.

Abstract

With strong evidence of a common-spectrum stochastic process in the most recent datasets from the NANOGrav Collaboration, the European Pulsar Timing Array (PTA), Parkes PTA, and the International PTA, it is crucial to assess the effects of the several astrophysical and cosmological sources that could contribute to the stochastic gravitational wave background (GWB). Using the same dataset creation and injection techniques as in Pol et al. (2021), we assess the separability of multiple GWBs by creating single and multiple GWB source datasets. We search for these injected sources using Bayesian PTA analysis techniques to assess recovery and separability of multiple astrophysical and cosmological backgrounds. For a GWB due to supermassive black hole binaries and an underlying weaker background due to primordial gravitational waves with a GW energy density ratio of $\Omega_{\mathrm{PGW}}/\Omega_{\mathrm{SMBHB}} = 0.5$, the Bayes' factor for a second process exceeds unity at 17 years, and increases with additional data. At 20 years of data, we are able to constrain the spectral index and amplitude of the weaker GWB at this density ratio to a fractional uncertainty of 64% and 110%, respectively, using current PTA methods and techniques. Using these methods and findings, we outline a basic protocol to search for multiple backgrounds in future PTA datasets.