Towards better interpretation of the small-scale structure of the Universe

Abstract

There is a trove of cosmological information in the structure of the Universe on small scales (1-20Mpc). Revealing this information is of increasing interest as we exhaust the constraining power of the large scale (~100Mpc). With current and upcoming surveys such as DESI, DES, and Rubin Observatory resolving cosmic structure down to 1-halo scales, analysis of the small scales has become not just a curiosity, but also a necessity. However, the small scales is intrinsically complex, mixing nonlinear structural growth with the baryonic physics of galaxy formation. Thus, it is imperative that we develop a robust small-scale model that disentangles the two and enables robust cosmological constraints.

In this thesis, we first improve small-scale modeling by developing a robust galaxy-halo connection model that incorporates an array of physically motivated generalizations. We test this generalized model against the observed small-scale clustering of red luminous galaxies (LRGs) and find that our model reproduces the observations significantly better than the baseline galaxy-halo connection model that is commonly used in the literature. We find that a specific set of extensions, known as secondary galaxies or assembly biases, are particularly powerful in improving the model prediction. Moreover, these secondary biases can also account for at least 30% of the well-known ``lensing is low'' tension.

We also highlight the importance of more informative summary statistics in constraining the galaxy-halo connection model and cosmology from small scales. In our studies, we leverage the extra constraining power of the velocity space clustering, in addition to the commonly used projected galaxy clustering, to derive significantly stronger constraints on the galaxy-halo connection model. We also propose the squeezed 3-point correlation function, a novel statistic that is both richly informative and efficient to compute, to complement the popular 2-point correlation function (2PCF). Finally, we propose a regularized model of the 2PCF covariance matrix, which enables cheaper and more accurate estimation of observable uncertainties.