1. INTRODUCTION

The search for faint dwarf galaxies has been a nearly continuous endeavor since the serendipitous discovery of the first such system, Sculptor, by Shapley (1938a). As significantly deeper survey data became available, systematic searches for more dwarfs slowly revealed what are now known as the classical dwarf spheroidal (dSph) satellites of the Milky Way (Shapley, 1938b, Harrington & Wilson, 1950, Wilson, 1955, Cannon, Hawarden & Tritton, 1977). However, after the identification of Sextans by Irwin et al. (1990), the push to ever lower luminosities and surface brightnesses stalled for more than a decade. New efforts to find faint, low surface brightness Milky Way dwarf galaxies continued fruitlessly in this period (Kleyna et al., 1997, Simon & Blitz, 2002, Willman et al., 2002, Hopp, Schulte-Ladbeck & Kerp, 2003, Whiting et al., 2007). Notably, though, there were strong theoretical reasons to expect that dwarfs with substantially lower luminosities and surface brightnesses should exist (Benson et al., 2002).

This prediction proved resoundingly correct in 2005, when the first such objects were discovered in Sloan Digital Sky Survey (SDSS) imaging by Willman et al. (2005a, b). These results opened the floodgates, and within two years the known population of Milky Way satellite galaxies more than doubled (Zucker et al., 2006b, Zucker et al., 2006a, Belokurov et al., 2006, Belokurov et al., 2007, Sakamoto & Hasegawa, 2006, Irwin et al., 2007, Walsh, Jerjen & Willman, 2007). Over the following decade, new discoveries continued at a rapid pace in SDSS and other surveys (e.g., Belokurov et al., 2008, 2009, 2010, Bechtol et al., 2015, Koposov et al., 2015a, 2018, Drlica-Wagner et al., 2015, Drlica-Wagner et al., 2016, Martin et al., 2015, Kim et al., 2015a, Kim & Jerjen, 2015, Laevens et al., 2015b, 2015a, Torrealba et al., 2016b, 2018, Homma et al., 2016, 2018), such that the Milky Way satellite census has now doubled yet again (Figure 1). Thanks to significant investments of telescope time in deep imaging and spectroscopy of the newly discovered objects, along with accompanying theoretical modeling, we now have a general understanding of the properties of these systems and their place in galaxy evolution and cosmology.

Figure 1. Census of Milky Way satellite galaxies as a function of time. The objects shown here include all spectroscopically confirmed dwarf galaxies as well as those suspected to be dwarfs based on less conclusive spectroscopic and photometric measurements. The major discovery impact of SDSS (from 2005-2010) and DES/Pan-STARRS (2015), each of which approximately doubled the previously known satellite population, stands out in this historical perspective.

In this review we summarize the progress that has been made in characterizing the least luminous galaxies since their discovery. We begin by motivating the study of the least luminous galaxies and by offering a definition of the term “ultra-faint dwarf,” which has been in common usage since the initial discoveries. In Section 2 we discuss the stellar kinematics and mass modeling of UFDs, and the corresponding evidence that they are galaxies rather than diffuse star clusters. In Section 3 we describe the metallicities and chemical abundance patterns of stars in UFDs, including the mass-metallicity relation, the chemical evolution of the smallest dwarfs, and their role in establishing the site of r-process nucleosynthesis. We briefly summarize the structural properties of the UFD population in Section 4. In Section 5 we introduce the star formation histories and initial mass functions of UFDs, and in Section 6 we examine constraints on the luminosity function of the faintest galaxies. We consider the origin and evolution of these systems based on theoretical work and measurements of their orbits around the Milky Way in Section 7. We provide a brief overview of the manifold ways in which UFDs may be used to constrain the behavior of dark matter in Section 8. In Section 9 we introduce the study of ultra-faint dwarfs outside the immediate neighborhood of the Milky Way and the connection between faint dwarfs in the Local Group and the high-redshift universe. In Section 10 we summarize the current state of the field and suggest future paths for research.

1.1. The Cosmological Significance of the Lowest Luminosity Dwarf Galaxies

A reasonable astronomer might ask how the smallest, most inconspicuous galaxies ever formed could have broad importance to the field of astrophysics. However, several aspects of the UFDs make them critical objects to understand, with potentially wide-ranging implications. First, UFDs reside in the smallest dark matter halos yet found. While only the mass at the very center of the halo is currently measurable, the extrapolated virial masses of UFDs are ∼ 10⁹ M_⊙ (e.g., Strigari et al., 2008), and the halo masses at the time when the stars formed may have been ∼ 10⁸ M_⊙ (e.g., Bovill & Ricotti, 2009, Safarzadeh et al., 2018). UFDs are also the most dark matter-dominated systems known. This combination of small halo mass and negligible baryonic mass makes UFDs extremely valuable laboratories for constraining the nature of dark matter. Simply counting the number of such objects around the Milky Way places a limit on the mass of the dark matter particle (e.g., Jethwa, Erkal & Belokurov, 2018). The census and observed mass function of low-mass halos will point the direction toward solving the long-standing and highly contentious missing satellite problem (e.g., Klypin et al., 1999, Moore et al., 1999, Simon & Geha, 2007, Brooks et al., 2013, Kim, Peter & Hargis, 2017). The measured central densities, and perhaps eventually the density profiles, of UFDs provide significant clues to the behavior of dark matter on small scales (e.g., Calabrese & Spergel, 2016, Bozek et al., 2018, Errani, Peñarrubia & Walker, 2018).

Second, UFDs represent the extreme limit of the galaxy formation process. They have the lowest metallicities, oldest ages, smallest sizes, smallest stellar masses, and simplest assembly histories of all galaxies. Both observations and theoretical models indicate that UFDs formed at very high redshift, probably before the epoch of reionization. Unlike essentially all larger systems, they underwent little to no further evolution after that time, and have survived to the present day as pristine relics from the early universe (e.g., Bovill & Ricotti, 2009, Bovill & Ricotti, 2011, Wheeler et al., 2015). These objects therefore present us with a unique window into the conditions prevalent at the time when the first galaxies were forming.

To our knowledge, no previous reviews have focused primarily or exclusively on the properties of the faintest dwarf galaxies. Willman (2010) presented the first summary of searches for UFDs. There have been many reviews on the broader population of dwarfs (e.g., Mateo, 1998, Tolstoy, Hill & Tosi, 2009, McConnachie, 2012, the latter two of which also discuss UFDs), and various aspects of UFDs have been featured in recent reviews on dark matter (e.g., Bullock & Boylan-Kolchin, 2017, Strigari, 2018) and metal-poor stars (e.g., Frebel & Norris, 2015). Given the rapid maturation of the study of the very lowest luminosity galaxies over the last decade, here we aim to provide a comprehensive discussion of the current state of knowledge of these systems. After first results from LSST become available, some of this material may need to be revisited.

1.2. Defining “Ultra-Faint Dwarf”

The dwarf galaxies known prior to 2005 have absolute magnitudes brighter than M_V = −8.7, corresponding to V-band luminosities larger than 2.5 × 10⁵ L_⊙. Their Plummer (half-light) radii are ≳ 200 pc, and with the exception of Sextans and Ursa Minor, their central surface brightnesses are < 26 mag arcsec⁻². In contrast, the dwarfs discovered in SDSS and other modern surveys are up to a factor of ∼ 1000 less luminous, with half-light radii as small as ∼ 20 pc and surface brightnesses that can be ∼ 2−3 mag arcsec⁻² fainter than that of Sextans.

As was evident even from the titles of some of the first SDSS discovery papers — e.g., “A New Milky Way Companion: Unusual Globular Cluster or Extreme Dwarf Satellite?” ¹ (Willman et al., 2005a) and “A Curious Milky Way Satellite in Ursa Major” ² (Zucker et al., 2006a) — the nature of these new satellites was not immediately clear. Over the next several years, spectroscopy of stars in these objects pointed strongly to the conclusion that they were dwarf galaxies rather than globular clusters (Kleyna et al., 2005, Muñoz et al., 2006, Martin et al., 2007, Simon & Geha, 2007). Given the clear differences in global properties relative to previously known dwarf galaxies, the community rapidly began referring to these objects as “ultra-faint” dwarfs, a term first used by Willman et al. (2005a). However, no formal definition of such a class was ever offered in the literature, and the usage of it has not been entirely consistent. In particular, Canes Venatici I (CVn I) is often referred to as a UFD because it was discovered in SDSS data around the same time as many fainter dwarfs (Zucker et al., 2006b), but its size and luminosity are nearly identical to those of Ursa Minor, which was identified more than 50 years earlier thanks to its location ∼ 3× closer to the Milky Way.

Despite this new nomenclature, it is now obvious that ultra-faint dwarfs continuously extend the properties of more luminous dwarfs in stellar mass, surface brightness, size, dynamical mass, and metallicity (see Figure 2 and Sections 2-4). They are not a physically distinct class of objects. Nevertheless, there are several reasons why it may be useful to refer to them via a separate label. In particular, UFDs represent the extreme end (we presume) of the distribution of galaxy properties, orders of magnitude beyond the previously-known dwarfs in some respects. Moreover, while classical dSphs can already be identified and studied in other nearby groups of galaxies, the UFDs are special in that only the very brightest examples of such systems will be detectable beyond the Local Group in the foreseeable future. Because of their lack of bright stars, detailed spectroscopic characterization of ultra-faint dwarfs will likely remain limited to satellites of the Milky Way. Finally, it is tempting to suggest that UFDs might differ from classical dSphs in that their star formation was shut off by reionization at z ≳ 6 instead of continuing to lower redshift. While this hypothesis is consistent with the available data, the sample of M_V ≳ −9 dwarf galaxies with precision star formation histories is too small to draw firm conclusions yet. If this idea turns out to be correct, it would provide a physically-motivated division between ultra-faint and classical dwarfs.

Figure 2. Distribution of Milky Way satellites in absolute magnitude (MV) and half-light radius. Confirmed dwarf galaxies are displayed as dark blue filled circles, and objects suspected to be dwarf galaxies but for which the available data are not conclusive are shown as cyan filled circles. Dwarf galaxy candidates without any published classification (usually because of the lack of spectroscopy) are shown as open gray circles. The faint candidates with R1/2 ≳ 50 pc are almost certainly dwarf galaxies, but we do not include them in the confirmed category here given the currently available observations. The dwarf galaxy/candidate data included in this plot are listed in Table 1. The black diamonds represent the Milky Way's globular clusters (Harris, 1996). Lines of constant central surface brightness (at 25, 27, 29, and 31 mag arcsec−2 in V band) are plotted in pink. For stellar systems brighter than MV ≈ −5 there is no ambiguity in classification: globular clusters have R1/2 ≲ 20 pc and dwarf galaxies have R1/2 ≳ 100 pc. At fainter magnitudes the size distributions begin to impinge upon each other and classification based purely on photometric parameters may not always be possible. Whether the two populations actually occupy non-overlapping portions of this parameter space remains to be determined from spectroscopy of the faintest stellar systems with half-light radii between 10 and 40 pc.

Based on the above discussion, we suggest that dwarf galaxies with absolute magnitudes fainter than M_V = −7.7 (L = 10⁵ L_⊙) be considered UFDs. This definition matches the naming convention adopted by Bullock & Boylan-Kolchin (2017). Among the post-2005 discoveries, only four galaxies are within 1 magnitude of this boundary: CVn I (M_V = −8.7), Crater II (M_V = −8.2), Leo T (M_V = −8.0), and Eridanus II (M_V = −7.2). The first three of these systems stand out from the fainter population in obvious ways: CVn I is substantially more luminous, larger, and more metal-rich (e.g., Martin et al., 2007, Martin, de Jong & Rix, 2008, Simon & Geha, 2007, Muñoz et al., 2018), Crater II is a factor of ∼ 4 more extended than any fainter dwarf (Torrealba et al., 2016a), and Leo T hosts neutral gas and recent star formation (Ryan-Weber et al., 2008, de Jong et al., 2008). These objects more closely resemble the previously-known dSphs and dSph/dIrrs in the Local Group. Eridanus II, on the other hand, is distinct from other UFDs only in that it contains a star cluster (Crnojević et al., 2016b). Setting the dividing line such that it lands between Eridanus II and Leo T is therefore sensible, and minimizes the likelihood that future revisions to the absolute magnitudes of any of these systems will blur the boundary.

¹ Indeed, the classification of Willman 1 is still not entirely secure, although the metallicities of its brightest member stars suggest that it is, or was, a dwarf galaxy (Willman et al., 2011). Back.

² Although Zucker et al. (2006a) argued that Ursa Major II is a dwarf galaxy, the same system was identified independently by Grillmair (2006), who described it as “a new globular cluster or dwarf spheroidal.” Back.