The gold standard for determining star formation histories based on resolved stellar populations is Hubble Space Telescope (HST) photometry, because of its superior photometric accuracy and stability relative to ground-based data. One challenge facing such work for UFDs is small-number statistics. Obtaining strong constraints on the star formation history of an old stellar population requires a sample of at least ∼ 200−300 stars near the main sequence turnoff (Brown et al., 2014). The lowest-luminosity dwarfs simply do not contain enough stars to meet this criterion even if every star in the galaxy is observed. Accurate star formation histories can be obtained for systems with absolute magnitudes brighter than MV ≈ −3, although doing so may require a number of HST pointings in order to include as many stars as possible. HST-based star formation histories have been published for 6 UFDs.
The first analysis of deep HST imaging of UFDs was carried out by Brown et al. (2012), studying Hercules, Leo IV, and UMa I. They concluded that the three galaxies have similar ages, and are each as old or older than the prototypical ancient globular cluster M92. Brown et al. (2014) expanded the sample to six UFDs, adding Boo I, Canes Venatici II (CVn II), and Coma Berenices (Com Ber) to the previous three. By incorporating improved spectroscopic MDFs and updated isochrones matched to observed dwarf galaxy chemical abundance patterns, Brown et al. (2014) determined that all of the galaxies except UMa I had formed more than 75% of their stars by z ∼ 10. Using a star formation model consisting of two bursts, the best fit for UMa I has approximately half of its stars forming at z ∼ 3. A large majority of the stars in all six dwarfs had formed by the end of reionization at z ∼ 6, consistent with the idea that gas heating by reionization ended star formation in such objects (e.g., Bullock, Kravtsov & Weinberg, 2000, Somerville, 2002, Benson et al., 2002). Note, however, that quenching by reionization does not necessarily mean that star formation ends precisely at the redshift of reionization, since sufficiently high-density molecular gas can survive somewhat beyond reionization even in low-mass halos (e.g., Oñorbe et al., 2015). Star formation histories have also been derived for Hercules, Leo IV, and CVn II by Weisz et al. (2014) from shallower WFPC2 data. Weisz et al. found that > 90% of the stars in Hercules and Leo IV are older than 11 Gyr, consistent with the Brown et al. (2014) results. In CVn II, however, Weisz et al. concluded that star formation continued until ∼ 8 Gyr ago, in conflict with Brown et al. The reason for this discrepancy is not clear. Age estimates based on deep ground-based imaging are generally consistent with the HST results, although the constraints are not as tight (e.g., Sand et al., 2010, Okamoto et al., 2012).
Based on the available data, it appears likely that UFDs are uniformly ancient, with all or nearly all of their stars forming in the early universe. While most or all UFDs exhibit a blue plume of stars brighter than the main sequence turnoff, this population is best interpreted as blue stragglers rather than young stars (Santana et al., 2013). These objects can thus be considered pristine fossils from the era of reionization (e.g., Bovill & Ricotti, 2009, 2011, Salvadori & Ferrara, 2009). Improved age measurements to reveal how synchronized the star formation in such galaxies was would be very interesting. Conversely, a clear detection of younger stars in very low-luminosity dwarfs would have important implications for star formation in low-mass dark matter halos and perhaps for cosmology as well (e.g., Bozek et al., 2018).
Their low metallicities make UFDs some of the most extreme environments in which star formation is known to have occurred. They therefore present an promising opportunity to investigate how the stellar initial mass function (IMF) depends on galactic environment. Dwarf galaxies also offer the advantage that their low stellar densities mean that no dynamical evolution has occurred, unlike in globular clusters, so the present-day mass function can be assumed to match the initial one below the main sequence turnoff. Geha et al. (2013) measured the IMF in two ultra-faint dwarfs, Hercules and Leo IV, using star counts from the HST photometry of Brown et al. (2012). Over the mass range from ∼ 0.5−0.8 M⊙, they found that the best fitting power law had a slope of α ≈ 1.2, much shallower than the Salpeter (1955) value of α = 2.35. While the uncertainties for Leo IV are quite large, in Hercules the slope disagrees with a Salpeter IMF at 5.8σ.
Intriguingly, such a bottom-light IMF in the least massive, lowest metallicity galaxies known suggests the possibility of a monotonic trend in IMF slope with galaxy properties. The largest elliptical galaxies have bottom-heavy IMFs (e.g., van Dokkum & Conroy, 2010, Spiniello et al., 2012), and dwarf galaxies in the Local Group appear to exhibit increasingly shallower IMF slopes toward lower masses (Geha et al., 2013).
More recent analyses have complicated this picture. Gennaro et al. (2018b) measured the IMFs for the full sample of 6 UFDs from Brown et al. (2014), confirming that each galaxy has an IMF slope shallower than Salpeter when fit with a power law. However, when describing the IMF as a log-normal function (Chabrier, 2003), the parameters for the UFDs are consistent with the Milky Way IMF. Gennaro et al. (2018a) used deeper, near-infrared imaging of Com Ber with HST to probe the IMF down to masses of ∼ 0.2 M⊙, comparable to the characteristic mass in the log-normal description of the Milky Way IMF. The results are consistent both with the shallower optical observations of Com Ber and the Chabrier (2003) Galactic IMF. These findings suggest that there may be significant IMF variations even within the class of UFDs, with some galaxies having shallow IMFs and others that resemble the Milky Way despite their low metallicities (Gennaro et al., 2018a).
If the IMF is indeed bottom-light in UFDs, there would be important implications for SN rates, feedback, chemical enrichment, and gas loss in such systems. A bottom-light IMF extrapolated to higher masses is top-heavy, which would produce larger numbers of SN explosions for a given mass of stars (of course, the validity of such an extrapolation is only an assumption, since no stars heavier than ∼ 0.8 M⊙ exist in UFDs today). This effect can be dramatic; Frebel, Simon & Kirby (2014) estimated that for Segue 1 (with a present-day stellar mass of ∼ 500 M⊙), the galaxy would have hosted ∼ 15 core-collapse SNe for a Salpeter IMF compared to ≳ 250 SNe for the Geha et al. (2013) IMF. Until the behavior of the IMF in the ultra-faint dwarf regime is better understood, the number of SNe expected to have occurred in such systems will be highly uncertain.
Among the dwarfs discovered since the beginning of SDSS, only Leo T (which we do not consider a UFD; see Section 1.2) contains any neutral gas (Irwin et al., 2007, Ryan-Weber et al., 2008). Stringent upper limits have been placed on the H i content of many of the UFDs using archival data or deep pointed observations with large single-dish telescopes (Grcevich & Putman, 2009, Spekkens et al., 2014, Westmeier et al., 2015). For the most nearby dwarfs these limits can be as small as ∼ 100 M⊙, while for objects at distances of ∼ 100 kpc typical limits are ∼ 1000 M⊙. No ionized gas associated with UFDs has been detected either, but searches for low surface brightness Hα emission similar to what has been detected for high-velocity clouds (e.g., Putman et al., 2003, Barger et al., 2012) could be of interest.
The lack of gas in these tiny galaxies is not a surprise, but the mechanism by which they lost their gas is not clear. Plausible hypotheses for gas removal include reionization, supernova feedback, and ram-pressure stripping. Because nearly all currently known UFDs are close to massive galaxies that are likely surrounded by hot gaseous halos, ram-pressure stripping cannot be ruled out. Studies of isolated UFDs, which should be discovered with LSST, may shed light on this issue.