Void Clustering as a Probe of Cosmology, feat. the new CAMELS-SAM simulation suite for machine learning

The distribution of galaxies traces the structure of underlying dark matter, and carries signatures of both the cosmology that evolved the universe as well as details of how galaxies interact with their environment and each other. One uncommon statistic to measure clustering is the Void Probability Function (VPF): it simply asks, how likely is a circle/sphere of a given size to be empty in your galaxy sample? Simple and efficient to calculate, the VPF is tied to all higher order volume-averaged correlation functions as the 0th moment of count-in-cells, and encodes information from higher order clustering that the robust two-point correlation function cannot always capture. We probe the ability of the VPF to constrain cosmology and parameters for astrophysical feedback when compared against and working alongside other statistics, using neural networks and the new CAMELS-SAM simulation suite for machine learning. CAMELS-SAM encompasses one thousand dark-matter only simulations of (100 h-1 cMpc)³ with different cosmological parameters and run through the Santa Cruz semi-analytic model for galaxy formation over a broad range of astrophysical parameters. Additionally, we explore what constraints the VPF can give for the pacing and process of reionization using simulated Lyman-Alpha Emitters, in the contexts of the LAGER narrowband z=6.9 survey and future grism z>7.2 surveys with the Roman Space Telescope.