Forward modeling in the era of cosmological surveys

Abstract

The distribution of matter on small scales is highly dependent on the exact composition of the Universe and as such, holds the key to the nature of many of its enigmatic ingredients such as dark energy, neutrinos, and dark matter. The ultimate constraints on these unknowns will likely come from multi-scale, multi-tracer approaches that encompass a wide range of probes and redshifts and are accurate at the subpercent level. This thesis offers a pioneering approach to using hydrodynamical (hydro) simulations to the advantage of large-scale structure surveys. While historically, hydro simulations have been insightful in unveiling the poorly understood processes of galaxy formation and evolution, only recently have they reached volumes that are relevant for cosmological analysis. We employ the state-of-the-art hydro simulations IllustrisTNG and the larger-yet MillenniumTNG to uncover discrepancies in commonly used galaxy-halo models, which if overlooked, would lead to serious biases in our cosmological inference. Putting these models to the test, we show that their current level of accuracy is inadequate in the context of the expected measurement precision of current and upcoming cosmological surveys. Augmenting them with an additional dependence on physical properties of the dark matter field alleviates the most major tensions. Since the volume and parameter variation of hydro simulations is still limited, before applying our knowledge to observations, we need to graft these models onto N-body (dark-matter-only) simulations. As part of the development team of the AbacusSummit N-body simulation suite, we create publicly available products such as a novel halo finding algorithm and halo light cone catalogs, designed expressly for analyzing large galaxy surveys such as the Dark Energy Spectroscopic Instrument (DESI). The final step of this process is placing the simulation analysis in the domain of observational analysis. We present applications of two novel numerical methods to photometric data in the final two chapters. The first is a hybrid effective field theory (HEFT) model, which combines a more traditional analytic approach with the exactness of N-body simulations, while the second offers an analytic marginalization over the uncertainties inherent to photometric redshifts. To conclude, we discuss future applications of the methods developed in this thesis to galaxy surveys such as DESI and Euclid and cosmic microwave background (CMB) experiments such as the Atacama Cosmology Telescope (ACT) and Planck.