Towards a few percent measurement of the Hubble constant with the current network of gravitational wave detectors without using electromagnetic information

TL;DR

The paper addresses how to infer the Hubble constant $H_0$ using gravitational waves without electromagnetic information by leveraging spectral sirens and a population model for BBH masses. It compares a fiducial PLG model with a more complex three-power-law (3PL) mass model and finds that incorporating structure in the BBH mass distribution significantly improves $H_0$ constraints, e.g., from O3 data giving $H_0 = 84^{+35}_{-25}$ with 3PL. The authors forecast strong future gains for O4 and O5, predicting $H_0$ precisions of about 20% at O4 and as tight as 2.7% at O5 (with ~1800–1900 events) when using 3PL, along with partial breaking of degeneracies with $\\Omega_{m,0}$. Overall, the work demonstrates that GW-only cosmology could contribute substantially to resolving the Hubble tension within a decade, provided the BBH population features are robustly modeled and detector sensitivity continues to improve.

Abstract

Gravitational waves provide a novel and independent measurement of cosmological parameters, offering a promising avenue to address the Hubble tension alongside traditional electromagnetic observations. In the absence of electromagnetic counterparts or complete host galaxy catalogs, current measurements rely on population-based methods that statistically combine black hole merger events. Building on recent models that incorporate additional structure in the primary black hole mass distribution, using public data from the LIGO-Virgo-KAGRA (LVK) collaboration third observing run (O3), we obtain a 30% accuracy improvement on the measurement of the Hubble constant with respect to the result reported by LVK with the third GW transient catalog (GWTC-3). Employing a realistic simulation that includes full Bayesian single-event inference, we present forecasts for the upcoming LVK observational runs, O4 and O5. Using a three power-law mass model, we project a measurement of the Hubble constant with 20% accuracy at O4 sensitivity, improving to 2.7% accuracy at O5 sensitivity. Our findings demonstrate the potential for gravitational waves to provide a substantial contribution to solving the Hubble tension within the next decade of observations.

Figures (4)

• Figure 1: Marginalized posteriors on the Hubble constant from O3 real data analyses. Results using the PLG (3PL) model are shown in yellow (blue). The vertical lines in the reconstructed $H_0$ posterior distribution denote the $68\%$ C.I. The dashed posterior distribution shows PLG result from LVK LIGOScientific:2021aug. The purple and red shaded areas identify the 68% C.I. constraints on $H_0$ inferred from the CMB anisotropies by Planck Planck:2018vyg and in the local Universe by SH0ES Riess:2021jrx. (Colors available online)
• Figure 2: Marginalized posteriors on the Hubble constant from O4 (left) and O5 (right) simulations. Results using the PLG (3PL) model are shown in yellow (blue). The thin vertical lines in the reconstructed $H_0$ posterior denote the $68\%$ C.I. The purple and red shaded areas identify the $68\%$ C.I. constraints on $H_0$ inferred by Planck Planck:2018vyg and SH0ES Riess:2021jrx, respectively. For comparison, the posterior distribution from GW170817 LIGOScientific:2017adf is also reported in dashed grey. (Colors available online)
• Figure 3: Marginalized posterior distributions on the cosmological parameters from O4 (dashed) and O5 (solid) simulated data, obtained with the PLG (yellow) and 3PL (blue) population models. Contours correspond to $90\%$ credible regions. The injected values are indicated by the solid black line. The inset displays a magnified view of the parameter space, highlighting the O5 3PL contour. (Colors available online)
• Figure 4: Source frame primary mass distributions injected in the simulations. The population parameters correspond to the maximum likelihood values found in O3 analyses with fixed cosmology Gennari:2025nho. The fact that the 3PL model describes a more intricate population with additional features compared to the PLG model, as well as the sharpness with which it captures the first feature at $\sim 10~\rm{M_\odot}$ is clear in this plot. (Colors available online)