Segue 1 is the current best candidate for a "first galaxy", a system which experienced only a single short burst of star formation and has since remained unchanged. Here we present possible star formation scenarios which can explain its unique metallicity distribution. While the majority of stars in all other ultra-faint dwarfs (UFDs) are within 0.5 dex of the mean [Fe/H] for the galaxy, 5 of the 7 stars in Segue 1 have a spread of $\Delta$[Fe/H] $>0.8$ dex. We show that this distribution of metallicities canot be explained by a gradual build-up of stars, but instead requires clustered star formation. Chemical tagging allows the separate unresolved delta functions in abundance space to be associated with discrete events in space and time. This provides an opportunity to put the enrichment events into a time sequence and unravel the history of the system. We investigate two possible scenarios for the star formation history of Segue 1 using Fyris Alpha simulations of gas in a $10^7$ M$_\odot$ dark matter halo. The lack of stars with intermediate metallicities $-3<$ [Fe/H] $<-2$ can be explained either by a pause in star formation caused by supernova feedback, or by the spread of metallicities resulting from one or two supernovae in a low-mass dark matter halo. Either possibility can reproduce the metallicity distribution function (MDF), as well as the other observed elemental abundances. The unusual MDF and the low luminosity of Segue 1 can be explained by it being a first galaxy that originated with $M_{\rm{vir}}\sim10^7$~M$_\odot$ at $z\sim10$.

New observations indicate that ultrafaint dwarf galaxies (UFD) -- the least luminous systems bound by dark matter halos (<10^5 Lsun) -- may have formed before reionization. The extrapolated virial masses today are uncertain with estimates ranging from 10^8 Msun to 10^9 Msun. We show that the progenitor halo masses of UFDs can be as low as Mvir = 10^7 Msun. Under the right conditions, such a halo can survive the energy input of a supernova and its radiative progenitor. A clumpy medium is much less susceptible to both internal and external injections of energy. It is less prone to SN sweeping because the coupling efficiency of the explosive energy is much lower than for a diffuse ISM. With the aid of the 3D hydro/ionization code Fyris, we show that sufficient baryons are retained to form stars following a single supernova event in dark matter halos down to Mvir ~ 10^7 Msun with radiative cooling. The gas survives the SN explosion, is enriched with the abundance yields of the discrete events, and reaches surface densities where low mass stars can form. Our highest resolution simulations reveal why cooling is so effective in retaining gas compared to any other factor. In the early stages, the super-hot metal-enriched SN ejecta exhibit strong cooling, leading to much of the explosive energy being lost. Consistent with earlier work, the baryons do *not* survive in smooth or adiabatic models in the event of a supernova. The smallest galaxies carry signatures of the earliest epochs of star formation, which may distinguish a small primordial galaxy from one that was stripped down to its present size through tidal interaction. We discuss these results in the context of local UFDs and damped Ly-alpha systems (z~2) at very low metallicity ([Fe/H] ~ -3). We show that both classes of objects are consistent with primordial low-mass systems that have experienced only a few enrichment events.

In earlier work we showed that a dark matter halo with a virial mass of $10^7$ M$_\odot$ can survive feedback from its own massive stars and form stars for $\gtrsim100$ Myr. We also found that our modelled systems were consistent with observations of ultrafaint dwarfs (UFDs), the least massive known galaxies. Very metal-poor damped Lyman-$\alpha$ systems (DLAs) recently identified at $z\sim2$ may represent the gas that formed at least some of the observed stars in UFDs. We compare projected sightlines from our simulations to the observed metal-poor DLAs and find that our models can reach the densities of the observed sightlines; however the metallicities are inconsistent with the single supernova simulations, suggesting enrichment by multiple supernovae. We model two scenarios for the history of these systems. The first explains the gas abundances in DLAs by a single burst of star formation. This model can produce the observed DLA abundances, but does not provide an explanation as to why the DLAs show suppressed [$\alpha$/Fe] compared to the stellar population of UFDs. The second scenario splits the DLAs into a population which is enriched by a single burst, and a population that is enriched by a second burst after the accretion of metal-poor gas. In this scenario, the suppressed average [$\alpha$/Fe] in DLAs compared to UFDs results from enrichment of second-burst systems by Type Ia supernovae.

We model the chemical evolution of six UFDs: Bootes I, Canes Venatici II, Coma Berenices, Hercules, Leo IV and Ursa Major I, based on their recently determined star formation histories. We show that two single-age bursts cannot explain the observed [$\alpha$/Fe] vs [Fe/H] distribution in these galaxies and that some self-enrichment is required within the first burst. An alternative scenario is modelled, in which star formation is continuous except for short interruptions when one or more supernovae temporarily blow the dense gas out from the centre of the system. This model allows for self-enrichment and can reproduce the chemical abundances of the UFDs in which the second burst is only a trace population. We conclude that the most likely star formation history is one or two extended periods of star formation, with the first burst lasting for at least 100 Myr. As found in earlier work, the observed properties of UFDs can be explained by formation at a low mass ($M_{\rm{vir}}\sim10^7$ M$_\odot$), rather than being stripped remnants of much larger systems.

In earlier work we showed that a dark matter halo with a virial mass of $10^7$ M$_\odot$ can retain a large percentage of its baryons in the face of the pre-ionization phase and supernova explosion from a 25$M_\odot$ star. Here we expand on the results of that work, investigating the star formation and chemical evolution of the system beyond the first supernova. In a galaxy with a mass $M_{vir} = 10^7$M$_\odot$, sufficient gas is retained by the potential for a second period of star formation to occur. The impact of a central explosion is found to be much stronger than that of an off-centred explosion both in blowing out the gas and in enriching it, as in the off-centered case most of the supernova energy and metals escape into the IGM. We model the star formation and metallicity given the assumption that stars form for 100, 200, 400 and 600 Myr and discuss the results in the context of recent observations of very low mass galaxies. We show that we can account for most features of the observed relationship between [$\alpha$/Fe] and [Fe/H] in ultra-faint dwarf galaxies with the assumption that the systems formed at a low mass, rather than being remnants of much larger systems.