From new physics to a running power law and back again: Minimal refitting techniques for the reconstruction of the gravitational-wave background signal in pulsar timing array data

TL;DR

This paper presents innovative refitting techniques for pulsar timing array data analysis that enable fast, accurate Bayesian inference of gravitational-wave background models, improving model comparison and interpretation of recent PTA signals.

Contribution

The authors introduce a novel matched-filtering approach to construct maps from GWB spectral models to a running-power-law reference, enhancing spectral model refitting in PTA data analysis.

Findings

Validated techniques with cosmic string GW models

Applied methods to scalar-induced GWs

Demonstrated improved Bayesian inference accuracy

Abstract

Pulsar timing array (PTA) collaborations recently presented evidence for a gravitational-wave background (GWB) signal at nanohertz frequencies. In this paper, we introduce new refitting techniques for PTA data analysis that elevate related techniques in the literature to a more rigorous level and thus provide the basis for fast and accurate Bayesian inference and physically intuitive model comparisons. The key idea behind our approach is to construct maps \Phi from GWB spectral models to a running-power-law (RPL) reference model, such that the pullback \Phi^* P_RPL of the RPL posterior density P_RPL induces a likelihood on the GWB model parameter space; in other words, we refit spectral models to the RPL posterior density. In order to construct \Phi, we introduce a matched-filtering approach in which \Phi follows from a \chi^2 minimization that accounts for the frequency dependence of PTA sensitivity curves. We validate and illustrate our techniques by three concrete examples: GWs from stable cosmic strings, GWs from metastable strings, and scalar-induced GWs.