Our understanding of galaxy formation and evolution has dramatically evolved in the past fifty years. The first and simplest idea for the formation scenario of our own Milky Way (MW) Galaxy was put forward by Eggen et al (1962), who proposed the bulk of a stellar halo to be formed in a rapid collapse of gas in the protogalaxy. This scenario, often referred to as “monolithic” collapse, is a dissipative process and takes place on dynamical timescales of the order of ∼ 108 yr. This process gives birth to a metal-poor stellar component in the halo outer regions, while the inner regions ends up being more metal-rich due to the reprocessing of the gas as it collapses deeper into the protogalaxy potential well. This idea was later challenged by an alternative explanation, based on the observation that globular clusters (GCs) at different Galactocentric distances have a wide range of metallicities. In this scenario, the halo is formed on longer timescales (∼ 109 yr) and, instead of being a self-contained system, it comes together as the product of several protogalactic fragments (Searle and Zinn 1978). These fragments can be pre-enriched before they are accreted. While both scenarios are capable of explaining many observed quantities of the Galactic halo, they cannot individually give a comprehensive picture (Norris and Ryan 1991, Chiba and Beers 2000), which has led to the development of hybrid “two-phase” models. In the latter, the inner Galaxy regions are formed in a first phase as a result of a monolithic-like process, while the outer halo regions are built up over the Galaxy's lifetime through dissipationless accretion events (Freeman and Bland-Hawthorn 2002).
In the past couple of decades, the most widely accepted paradigm of the hierarchical Lambda-Cold Dark Matter (ΛCDM) structure formation model has prevailed, favouring the predominance of merger and accretion events in the build-up of galactic haloes (White and Frenk 1991, Bullock and Johnston 2005, Springel et al 2006, Johnston et al 2008). These models predict the ubiquitous presence of haloes, which are characterized by old and metal-poor populations and often shows signs of recent interactions, in contrast with the smooth haloes predicted by dissipative models (Bullock and Johnston 2005, Abadi et al 2006, Font et al 2006). The interaction events provide a mine of information on the assembly of haloes: dynamical timescales become relatively long (up to several Gyr) in the outer regions of a galaxy, and thus accretion/merger events that occurred a long time ago are often still visible as coherent structures like disrupting galaxies or streams, which readily testify the past assembly history of their host. The assembly itself depends on a variety of factors, such as number, mass, stellar content and structural properties of the accreted satellites, as well as orbital properties, timing and energy of the accretion event. Even when the progenitor is completely dissolved in the host's halo (which is particularly true in the inner halo regions where dynamical timescales are relatively short), its stripped stellar content still retains a characteristic coherence in velocity space as well as in metallicity content, thus giving important clues about the progenitor's properties. Observing the stellar “fossils” that populate galaxy haloes thus offers a unique opportunity to reconstruct the modes, timing, and statistics of the halo formation process.
Besides being taletellers of their host system's merger history, the shape and size of haloes also hold vital clues to the process of galaxy formation. In particular, they can teach us about the primordial power spectrum of density fluctuations at the smallest scales; about the reionization process, that shall lead to faint and concentrated haloes for an early suppression of star formation in low-mass dark matter (DM) subhaloes; or about the triaxiality of DM haloes, which are predicted to be more flattened for dissipationless formation scenarios (Abadi et al 2006). Despite only accounting for a mere ∼ 1% of a galaxy's total mass (e.g., Morrison 1993), extended haloes are clearly extremely valuable to test and refine theoretical predictions on the halo assembly process. Due to their extreme faintness, however, haloes have not been as fully exploited as they should have been as key tests of galaxy formation models: they are not easily detected above the sky level, i.e., surface brightness values of µV ∼ 25 mag arcsec−2, posing a serious observing challenge to their investigation. Cosmological simulations predict the majority of past and ongoing accretion events to have surface brightness values well below this value (e.g., Bullock and Johnston 2005). According to some models, reaching a surface brightness of µV ∼ 29 mag arcsec−2 should allow the detection of at least one stream per observed galaxy (Johnston et al 2008, Cooper et al 2010). How is it then possible to extract the information locked in the faint outskirts of galaxies?
1.1. Resolved Stellar Populations
The best method to study faint haloes and their substructure in nearby galaxies is to resolve individual stars. Even when sparse and faint, resolved stars can be individually counted, and a stellar number density can easily be converted into a surface brightness value. When the Galactic extinction presents a high degree of spatial inhomogeneity (possibly mimicking faint irregular substructures), and the sky level is higher than the integrated light signal coming from extremely faint sources, resolved populations provide a very powerful means to trace them. This method is not free from complications: there will always be contamination coming both from foreground Galactic stars as well as from background unresolved galaxies. This can be accounted for statistically, by observing “field” regions away from the main target and quantifying the contaminants, while a direct confirmation of a star's membership requires spectroscopy. At the same time, resolving individual stars poses constraints on the inherent nature and on the distance of the putative targets: for systems where the stellar density is so high that stars fall on top of each other on the sky, the “crowding” prevents the resolution of individual objects. This can of course occur also in the case of a relatively sparse galaxy which has a large line-of-sight distance, so that the stars are packed in a small region of the sky. Distance is also the principal enemy of depth: the larger the distance, the brighter the detection limit, i.e., the absolute magnitude/surface brightness that we can reach for a fixed apparent magnitude. Nonetheless, resolved stellar populations are able to deliver powerful information for galaxies located within ∼ 10 Mpc, i.e., within the so-called Local Volume.
The discovery of the Sagittarius dwarf galaxy by Ibata et al (1994) from the identification of a comoving group of stars opened the door to the era of halo studies and their substructure: a galaxy resembling the properties of classical dwarf spheroidals was clearly in the process of being disrupted by its giant host, our own MW. This evidence was the first to support theoretical predictions for the hierarchical assembly models and the existence of observable accretion events. Soon thereafter, stellar density maps allowed the discovery of a prominent low surface brightness stream around the MW's closest giant spiral Andromeda (M31), the so-called Giant Stellar Stream (Ibata et al 2001). This feature, invisible to the naked eye, is a clear example of the elusive nature of haloes and their substructure: the surface brightness of the Giant Stellar Stream is µV ∼ 30 mag arcsec−2, which is prohibitive for integrated light images.
As challenging as it is, the mere detection of haloes and their substructures is not enough to provide quantitative constraints on models of galaxy evolution. From the stars' photometry and thus position in the colour-magnitude diagram (CMD), i.e., the observational counterpart of the Hertzsprung-Russel diagram, it is possible to characterize the properties of the considered stellar system. First and foremost, in contrast to integrated light, accurate distance measurements can be obtained from CMD features that act as standard candles, e.g., the luminosity of the tip of the red giant branch (TRGB) or of the horizontal branch (HB). Another key advantage of resolved populations is the possibility to constrain ages and metallicities more tightly than with integrated light alone. The CMD is used to quantify the star formation rate as a function of lookback time, and thus derive the star formation history (SFH) of a composite stellar population (e.g., Gallart et al 2005, and references therein). Spectroscopy of individual stars is the ultimate method to constrain their metallicity content and kinematical properties, such as radial velocity and proper motion, which allows for the full six-dimensional phase space to be investigated. The latter cannot, for the moment, be achieved beyond the LG limits, and still only occasionally for M31.
Besides giving precious insights into galaxy haloes and their accretion histories, resolved stellar populations can help us characterizing the “surviving” low-mass galaxies that have not been accreted to date and reside in the outskirts of giant hosts.
1.2. The Low-mass End of the Galaxy Luminosity Function
The low-mass end of the galaxy luminosity function (LF) is of no less interest than haloes themselves. Besides the MW and M31, the Local Group (LG) contains tens of smaller galaxies which can be studied in detail due to their proximity (see Tolstoy et al 2009 for a review). While the ΛCDM cosmological model has provided a convincing match to the large-scale structures observed in the high-redshift Universe, it falls short at the smallest, galactic scales, indicating an incomplete understanding of the physics involved in the evolution of galaxies: for example, the “missing-satellite problem” has been highlighted for the first time by Moore et al (1999) and Klypin et al (1999). Briefly, the number of DM subhaloes predicted in simulations exceeds the observed number of MW satellites by almost two orders of magnitude. The shape of the DM profile in the innermost regions of dwarf galaxies is also a matter of debate (the “cusp-core” problem; Walker and Peñarrubia 2011). In addition, the more massive among the MW satellites are less dense than what is expected from simulations, which is puzzling because they should be affected by fewer observational biases than their smaller, sparser siblings (the “too-big-to-fail” problem; Boylan-Kolchin et al 2011). In addition, the fact that many of the MW and M31 satellites are distributed along planes does not have a straightforward explanation in ΛCDM models (e.g., Pawlowski et al 2014).
From the theoretical point of view, the inclusion of baryonic physics in DM-only simulations is key to reconcile predictions with observations of the smallest galaxies. In particular, effects such as supernova feedback, stellar winds, cosmic reionisation, and tidal/ram pressure stripping all concur to reduce star formation efficiency in the least massive DM haloes. Tremendous progress is being made on this front, taking into account realistic physics as well as increasing the resolution of simulations (e.g., Stinson et al 2009, Brooks et al 2013, Sawala et al 2016, Wetzel et al 2016). At the same time, new observational discoveries keep offering intriguing challenges at the smallest galactic scales, as further described in Sect. 2.1.3 and 2.2.3.