Beyond Gaps and Bumps: Spectral Siren Cosmology with Non-Parametric Population Models

TL;DR

This paper introduces a non-parametric spectral siren method using gravitational wave data to constrain cosmology, specifically the Hubble constant, without relying on electromagnetic counterparts, and demonstrates its effectiveness with GWTC-3 data.

Contribution

It develops a flexible, non-parametric model for black hole mass distribution to extract cosmological information solely from gravitational wave data, improving Hubble constant estimates.

Findings

Measured H_0=73.0^{+13.3}_{-7.7} km/s/Mpc with GWTC-3 data

Achieved a 1.4 times improvement over GW170817 alone

Validated the method with simulations and benchmarking

Abstract

Gravitational wave standard sirens typically require electromagnetic (EM) data to obtain redshift information to constrain cosmology. Difficult to find EM counterparts for bright sirens and galaxy survey systematics for dark sirens make cosmological constraints with spectral sirens, a gravitational wave data-only approach, extremely appealing. In this work, we use the GWTC-3 BBH detections as spectral sirens to constrain the BBH population and the underlying cosmological expansion with a flexible model for the black hole mass spectrum. We use a binned Gaussian process to model the BBH mass distribution in the source frame without any astrophysical assumptions on the shape and or inclusion (or lack of) features that drive the cosmological constraints as the redshifted detector frame masses become consistent with the underlying astrophysical mass distribution features. For GWTC-3 we find a measurement on the Hubble constant of $H_0=73.0^{+13.3}_{-7.7} \ {\rm{km \ s^{-1} \ Mpc^{-1}}}$ at $68\%$ C.L. when combined with that obtained from the bright standard siren analysis with GW170817 and its associated host galaxy NGC 4993. We find an improved estimate for the Hubble constant of around a factor of 1.4 times better than the GW170817 measurement alone. We validate our nonparametric spectral siren approach with simulations and benchmark its scalability and constraining performance when compared with parametric methods.