Brainerd, TereasaCurtis, Olivia M.2025-09-162025-09-162024https://hdl.handle.net/2144/512262024Galaxies and their dark matter halos form the cosmic web, an interconnected network of walls and filaments that stretch for tens of megaparsecs (Mpc), between which are vast regions of near nothingness. These underdense regions, known as cosmic voids, are among the largest structures in the universe, reaching tens of Mpc in diameter. Voids are an excellent laboratory for constraining the standard model of structure formation: a cold dark matter model in which the energy density at the present day is dominated by dark energy. The underdense nature of voids means that they are dominated by dark energy, and their internal gravitational interactions are well-modeled by linear perturbation theory. Furthermore, the study of matter within voids can inform theories of galaxy evolution, as differences in local matter densities between void and non-void environments might cause void galaxies to follow distinct evolutionary tracts. However, the extent to which the properties of void and non-void field galaxies vary is currently poorly constrained. This dissertation is thus motivated by these broad questions: [1] Does luminous matter trace dark matter within voids across cosmic time, and [2] Do the properties of galaxies within voids differ from those in denser regions of space? I investigate these questions by first analyzing the probability distributions of voids within dark-matter-only N-body simulations. Here, I use a generative adversarial network to double the sample size of number density contrast maps used to study void statistics and radial profiles. I also test whether deep learning algorithms can accurately produce the same void statistics as those from N-body simulations. Then, I employ a void-finder algorithm to search for voids in the redshift z = 0 snapshot of the IllustrisTNG300 simulation. Within this snapshot, I test whether dark matter traces the luminous matter of voids and compare void galaxies to their non-void field galaxy counterparts. Next, I discuss how the properties of voids and void galaxies evolve from the redshift z = 3 snapshot to the redshift z = 0 snapshot. Finally, I compare these theoretical results to observations of voids and void galaxies in the local universe using publicly available catalogs.en-USAttribution 4.0 Internationalhttp://creativecommons.org/licenses/by/4.0/AstrophysicsAstronomyGalaxy evolutionGenerative adversarial networksLarge-scale structure of the universeMagnetohydrodynamical methodsVoidsThesis/Dissertation2025-09-150000-0002-0212-4563