Physics-informed inferences of galaxy clustering with Bayesian forward modelling

Abstract: In this thesis, we showcase four novel approaches to constraining the relationship between cosmological observables and the large-scale structure. The majority of the energy content of the Universe in the concordance cosmological model remains largely unknown. Galaxy clustering can constrain the physical mechanism that links galaxy observations to structure formation. Traditional approaches to constraining this relation either rely on summary statistics of galaxy clustering, which lead to information loss or on ad-hoc assumptions, which can bias cosmological conclusions if accounted for improperly. The frameworks presented here ensure the self-consistent propagation of modelled observational uncertainties and extraction of high-order statistics from galaxy clustering. In Paper I, we investigate the potential of supernovae as large-scale structure probes, complementary to galaxy clustering for multi-tracer cosmology. If supernovae are biased relative to galaxies, their combination can improve the mapping of the large-scale structure. We find that supernovae cluster similarly to galaxies at 3.9 Mpc through cross-correlation of supernova locations with the gravitational tidal shear. In the second study, we generate supernova simulations informed about galaxy clustering, galaxy formation and evolution. We model supernova rates from the simulated star-formation histories and estimate the supernova bias with respect to the galaxy density field. We find that supernovae are less clustered than galaxies. The relative biasing signal was obscured in Paper I due to shot noise. In Paper II, we constrain galaxy intrinsic alignment -- a systematic effect in weak gravitational lensing and a probe of galaxy formation and evolution -- accounting for all high-order statistics and nonlinear structure growth down to 15.6 Mpc/h. We report a 4σ detection of galaxy intrinsic alignment, constant with luminosity, color and redshift at 20 Mpc/h. In Paper III, we present a framework to jointly constrain the large-scale structure and photometric galaxy clustering. Our approach is the first to infer the entire structure formation history and the filamentary pattern of the dark matter distribution, despite the large redshift uncertainties. Our approach guarantees the improvement of arbitrarily large photometric redshift uncertainties through galaxy clustering.