Analyzing cosmological-hydrodynamical simulations in light of future optical surveys.

Abstract

State-of-the-art numerical simulation suites are powerful theoretical tools with which to address systematics in methods that are used for analysis in cosmological surveys. Future optical and near-infrared surveys will provide us with detailed 3D maps of galaxy positions, weak lensing and cluster abundances - all of which will be used to quantify the large-scale structure. It is therefore crucial that the methods we use to measure the clustering of galaxies and matter be robust to systematics in order to accurately extract cosmological information. In this body of work, we use cosmological-hydrodynamical simulations to better understand complex physical processes that inhibit our ability to interpret clustering statistics. We first investigate how feedback can impact the total matter distribution in and around halos and use the Cosmology and Astrophysics for MachinE Learning Simulations (CAMELS) to train a machine learning algorithm on baryon content and halo abundance to predict the suppression of total matter clustering. We further address galaxy assembly bias as a systematic in the use of galaxy clustering as a cosmological probe. Utilizing the IllustrisTNG simulation suite in conjunction with machine learning tools, we provide alternative models of the galaxy-halo connection which incorporate global halo properties. Lastly, we address the intrinsic alignment of galaxies as a systematic in weak lensing signals. We measure the projected correlations of galaxy shapes with respect to different tracers of large-scale structure in the MillenniumTNG simulation and report strong alignment signals with dependence on redshift and morphology.