Cosmology with gravitational-wave dark sirens: Estimating H0 with EM-informed redshift priors from biased dark matter distributions

Abstract

Dark-siren gravitational-wave cosmology infers the Hubble constant by statistically associating events with the large-scale structure that hosts them, connecting H0 to the clustering and bias of the chosen tracer. This thesis presents an end-to-end forward-modelling framework, where synthetic compact-binary catalogues are generated from non-linear matter overdensities on a past lightcone, projected onto a terrestrial detector network with realistic selection thresholds, and analysed with a hierarchical Bayesian framework. We recover the injected cosmology without bias, with H0 constrained to a ∼6% statistical precision. Linearly biased tracers differ only in how strongly they cluster relative to the underlying matter: galaxies, GW hosts, and Hi each rescale the amplitude of the host field, but all preserve the spatial pattern carrying the cosmological information. The dominant systematic is instead the angular resolution of the host map, while extreme cluster bias breaks the linear model itself.