Cosmological simulations are reaching the resolution necessary to study ultra-faint dwarf galaxies. Observations indicate that in small populations, the stellar initial mass function (IMF) is not fully populated; rather, stars are sampled in a way that can be approximated as coming from an underlying probability density function. To ensure the accuracy of cosmological simulations in the ultra-faint regime, we present an improved treatment of the IMF. We implement a self-consistent, stochastically populated IMF in cosmological hydrodynamic simulations. We test our method using high-resolution simulations of a Milky Way halo, run to $z=6$, yielding a sample of nearly 100 galaxies. We also use an isolated dwarf galaxy to investigate the resulting systematic differences in galaxy properties. We find that a stochastic IMF in simulations makes feedback burstier, strengthening feedback, and quenching star formation earlier in small dwarf galaxies. For galaxies in halos with mass $\lesssim10^{8.5}$ M$_\odot$, a stochastic IMF typically leads to lower stellar mass compared to a continuous IMF, sometimes by more than an order of magnitude. We show that existing methods of ensuring discrete supernovae incorrectly determine the mass of the star particle and its associated feedback. This leads to overcooling of surrounding gas, with at least ${\sim}$10 per cent higher star formation and ${\sim}$30 per cent higher cold gas content. Going forward, to accurately model dwarf galaxies and compare to observations, it will be necessary to incorporate a stochastically populated IMF that samples the full spectrum of stellar masses.

The existence of ultra-faint dwarf (UFD) galaxies highlights the need to push our theoretical understanding of galaxies to extremely low mass. We examine the formation of UFDs by twice running a fully cosmological simulations of dwarf galaxies, but varying star formation. One run uses a temperature-density threshold for star formation, while the other uses an H$_{2}$-based sub-grid star formation model. The total number of dwarf galaxies that forms is different by a factor of 2 between the two runs, but most of these are satellites, leading to a factor of 5 difference in the number of luminous UFD companions around more massive, isolated dwarfs. The first run yields a 47\% chance of finding a satellite around a M$_{halo}$ $\sim 10^{10}$ M$_{\odot}$ host, while the H$_2$ run predicts only a 16\% chance. Metallicity is the primary physical parameter that creates this difference. As metallicity decreases, the formation of H$_2$ is slowed and relegated to higher-density material. Thus, our H$_2$ run is unable to form many (and often, any) stars before reionization removes gas. These results emphasize that predictions for UFD properties made using hydrodynamic simulations, in particular regarding the frequency of satellites around dwarf galaxies, the slope of the stellar mass function at low masses, as well as the properties of ultra-faint galaxies occupying the smallest halos, are extremely sensitive to the subgrid physics of star formation contained within the simulation. However, upcoming discoveries of ultra-faint dwarfs will provide invaluable constraining power on the physics of the first star formation.

We predict the stellar mass-halo mass (SMHM) relationship for dwarf galaxies and their satellites residing in halos down to M$_{halo} =$ 10$^7$ M$_{\odot}$ with 10$^4$ M$_{\odot} <$ M$_{star}$($z=0$) $< 10^8$ M$_{\odot}$, and quantify the predicted scatter in the relation at the low mass end, using cosmological simulations. The galaxies were drawn from a cosmological simulation of dwarf galaxies, run with the N-body + SPH code, ChaNGA, at a high resolution of 60 pc. For M$_{halo} > 10^9$ M$_{\odot}$, the simulated SMHM relationship agrees with literature determinations, including exhibiting a small scatter. However, the scatter in the SMHM relation increases dramatically for lower-mass halos. We find that some of this scatter is due to {\em dark dwarfs}, halos devoid of stars. However, even when only considering well-resolved halos that contain a stellar population, the scatter in stellar mass reaches nearly 1 dex for M$_{halo}$($z=0$) 10$^7$ M$_{\odot}$. Much of this scatter is due to including satellites of the dwarf galaxies that have had their halo masses reduced through tidal stripping. The fraction of dark dwarfs (those that contain no stars) increases substantially with decreasing halo mass. When these dark halos are considered, the true scatter in the SMHM at low masses is even larger. At the faintest end of the SMHM relation probed by our simulations, a galaxy cannot be assigned a unique halo mass based solely on its luminosity. We provide a formula to stochastically populate low-mass halos following our results. Our predicted large scatter at low halo masses increases the slope of the resulting stellar mass function on the ultra-faint dwarf galaxy scales currently being probed by such surveys as the Dark Energy Survey or the Hyper-Suprime Cam Subaru Strategic Program, and in the future by the Large Synoptic Survey Telescope.