|Annu. Rev. Astron. Astrophys. 2009. 47:
Copyright © 2009 by Annual Reviews. All rights reserved
Stellar abundances and kinematics have been shown to be excellent tools for disentangling the properties of complex stellar systems like our own Galaxy (e.g., Eggen, Lynden-Bell & Sandage 1962). This approach is the only means we have to separate the diverse stellar populations in the solar neighbourhood. These stars can be split up into disk and halo components on the basis of their 3D velocities, and these subsets can then be studied independently. This concept has subsequently been expanded and renamed "Chemical Tagging" (Freeman & Bland-Hawthorn 2002). As large samples of stellar velocities and metallicities have become available for other galaxies this approach remains the only way to obtain a detailed understanding of a multi-component stellar system.
The kinematics and metallicities of early (dSph/dE) and late (dI) type dwarf galaxies in the Local Group have almost always been measured using different tracers. This is due to the different distances and stellar densities which are typical for the two types of systems. It is also because dIs contain an easily observable ISM in the form of HI gas and dSphs do not. Because early-type galaxies usually do not contain any (observable) gas nor any young star forming regions, most of what we know of their internal properties comes from studies of their evolved stellar populations (e.g., RGB stars). Late-type galaxies are typically further away (the SMC being a clear exception), which can make the accurate study of individual RGB stars more challenging, and they contain HI gas and several HII regions. Thus, most of what we know about the kinematics and metallicity of dIs comes from gas and massive (young) star abundances. RGB stars have the advantage that they are all old (> 1 Gyr) and their properties are most likely to trace the gravitational potential and chemical evolution throughout the entire galaxy up to the epoch when they formed, and not the most recent star formation processes and the final metallicity. It is only with detailed studies of the same tracers that kinematics and metallicities in these different dwarf galaxies types can be accurately compared to make confident global statements about the differences and similarities between early and late-type galaxies.
3.1. Early-Type Dwarfs
Dwarf Spheroidal galaxies are the closest early-type galaxies that contain sufficient numbers of well distributed RGB stars to provide useful kinematic and metallicity probes. Moreover, dSphs are considered to be interesting places to search for dark matter, since there is so little luminous matter to contribute to the gravitational potential. The velocity dispersion of individual stars can be used to determine the mass of the galaxy. It is also possible to determine metallicities for the same stars. This allows a more careful distinction of the global properties based on structural, kinematic and metallicity information (e.g., Battaglia 2007).
3.1.1 GALACTIC dSPHS: It was originally thought that the luminosity profiles of Galactic dSphs resembled globular clusters and showed systems truncated by the gravitational field of the MW (e.g., Hodge 1971). As measurements improved and the discussion focused on the possible presence of dark matter, Faber & Lin (1983) showed that the profiles are exponential more like those of galaxy disks and thus predicted that Galactic dSphs had a much higher mass-to-light ratio (M/L), 30, than had previously been thought. In parallel Aaronson (1983) found observational evidence for this in radial velocity studies of individual stars in the Draco dSph. Thus it became clear that dSph are small galaxies, related to late-type disk and irregular systems, and not globular clusters.
For a given M/L the central velocity dispersion of a self-gravitating spheroidal system in equilibrium may scale with the characteristic radial scale length and the central surface brightness (Richstone & Tremaine 1986). Given that globular clusters typically have a central velocity dispersion ~ 2 - 8 kms-1 it was expected that dwarf galaxies, which have scale lengths at least 10 times bigger and surface brightnesses about 1000 times smaller (see Fig. 1), should have central velocity dispersions < 2 km s-1. This has been consistently shown not to be the case; all galaxies have stellar velocity dispersions which are typically larger than those of globular clusters (~ 8 - 15 km s-1). This was first shown by Aaronson (1983) for a sample of 3 stars in the Draco dSph. This early tentative (and brave!) conclusion has been verified and strengthened significantly over the past decades, with modern samples containing measurements for many hundreds of individual stars in Draco (e.g., Muñoz et al. 2005, Wilkinson et al. 2004) and in all other Galactic dSphs. If it can be assumed that this velocity dispersion is not caused by tidal processes, then this is evidence that dSph galaxies contain a significant amount of unseen (dark) matter (e.g., Mateo 1994, Olszewski 1998, Gilmore et al. 2007), or that we do not understand gravity in these regimes (e.g., MOND applies). There has been some uncertainty coming from the possible presence of binary stars but a number of studies have carried out observations over multiple epochs and this effect has been found to be minimal (e.g., Battaglia et al. 2008b and references therein).
As instrumentation and telescopes improved, a significant amount of work on the kinematic properties of dSphs has become possible. This field has benefited particularly from wide-field multi-fibre spectrographs on 6-8m-class telescopes (e.g., VLT/FLAMES and Magellan/MIKE), but also WYFOS on the WHT and AAOmega on the AAT. These facilities have allowed samples of hundreds of stars out to the tidal radii (e.g., Wilkinson et al. 2004, Tolstoy et al. 2004, Muñoz et al. 2005, Kleyna et al. 2005, Walker et al. 2006b, Walker et al. 2006a, Battaglia et al. 2006, Battaglia 2007, Battaglia et al. 2008a). These velocity measurements often have sufficient signal-to-noise to also obtain metallicities from the Ca II triplet, the Mg B index or a combination of weak lines (e.g., Suntzeff et al. 1993, Tolstoy et al. 2004, Battaglia et al. 2006, Muñoz et al. 2006b, Koch et al. 2006, Kirby, Guhathakurta & Sneden 2008, Battaglia et al. 2008b, Shetrone et al. 2009). This approach resulted in the discovery of surprising complexity in the "simple" stellar populations in dSphs. It was found that RGB stars of different metallicity range (and hence presumably age range) in dSphs can have noticeably different kinematic properties (e.g. Tolstoy et al. 2004, Battaglia et al. 2006). This has implications for understanding the formation and evolution of the different components in these small galaxies. It is also important for correctly determining the overall potential of the system. The presence of multiple components allows more accurate modelling of the overall potential of the system and so better constraints on the the underlying dark matter profile and the mass content (Battaglia 2007, Battaglia et al. 2008a), see Fig. 9.
Figure 9. From the DART survey, these are VLT/FLAMES line-of-sight velocity measurements for individual RGB stars in the Scl dSph (Battaglia 2007, Battaglia et al. 2008a). In the top panel elliptical radii are plotted against velocity for each star. The limits for membership are given by dotted lines about vhel = +110.6 km s-1, the heliocentric velocity, which is shown as a dashed line. It is apparent that the velocity dispersion and central concentration of the metal-rich (MR) stars, in red, can clearly distinguish them from the metal-poor (MP) stars, in blue, which have a larger velocity dispersion and are more uniformly distributed over the galaxy. In the bottom 2 panels the line of sight velocity dispersion profiles, with rotation removed, for the MR (red) and MP (blue) stars are shown along with the best fitting pseudo-isothermal sphere (solid line) and NFW model (dashed line), see Battaglia et al. (2008a) for details.
The VLT/FLAMES DART survey (Tolstoy et al. 2006) determined kinematics and metallicities for large samples of individual stars in nearby dSphs. There have also been similar surveys by other teams on VLT and Magellan telescopes (e.g., Gilmore et al. 2007, Walker, Mateo & Olszewski 2009). Traditionally the mass distribution of stellar systems has been obtained from a Jeans analysis of the line-of-sight velocity dispersion (e.g., Mateo 1994), assuming a single stellar component embedded in a dark matter halo. This analysis suffers from a degeneracy between the mass distribution and the orbital motions presumed for the individual stars, the mass-anisotropy degeneracy. From DART the mass of the Scl dSph was determined taking advantage of the presence of the two separate components distinguished by metallicity, spatial extent and kinematics (Battaglia 2007, Battaglia et al. 2008a), see Fig. 9. Here it was shown that it is possible to partially break the mass-anisotropy degeneracy when there are two components embedded in the same dark matter halo. The new dynamical mass of the Scl dSph is Mdyn = 3 × 108 M, within 1.8 kpc, which results in an M/L ~ 160. This is a factor ~ 10 higher than the previous value obtained from a much smaller and more centrally concentrated sample of stars (Queloz, Dubath & Pasquini 1995). This corresponds to a dark matter density within 600 pc, for the best fitting model of 0.22 M pc-3. This result is largely independent of the exact distribution of dark matter in the central region of the Scl dSph, see Fig. 9. This same study also found evidence for a velocity gradient, of 7.6-2.2+3.0 km s-1 deg-1, in Scl, which has been interpreted as a signature of intrinsic rotation. This is the first time that rotation has been detected in a nearby dSph, and it was a faint signal that required a large data set going out to the tidal radius.
Another aspect of these surveys has been the determination of metallicity distribution functions (MDFs), typically using the Ca II triplet metallicity indicator (Battaglia et al. 2006, Helmi et al. 2006). This uses the empirical relation between the equivalent width of the Ca II triplet lines and [Fe/H], and its accuracy was also tested (Battaglia et al. 2008b). The Ca II triplet method will start to fail at low metallicities, [Fe/H] < -2.5, but this is starting to be better understood on physical (e.g., Starkenburg et al. 2008) as well as empirical grounds from following up stars with Ca II triplet metallicities, [Fe/H] < -2.5. This means that the effect can be corrected for, and so far there is no significant change to the MDFs in Helmi et al. (2006). This is because the fraction of the stellar samples that may be affected by this uncertainty is very small (~ 1-2%), and then only a fraction of these actually need to be corrected. Thus it seems likely that there are very few, if any, extremely metal poor stars ([Fe/H] < -4) in most classical dSph. But this result has to be verified by careful follow up of Ca II triplet measurements with [Fe/H] < -2.5.
These dSph MDFs were compared to the Galactic halo MDF from the Hamburg-ESO survey (HES, Beers & Christlieb 2005), and found to be significantly different (Helmi et al. 2006), see Fig. 10 for an update. The Galactic halo MDF has recently been revised (Schoerck et al. 2008, submitted), and this revision is also shown in Fig. 10. It can be seen that the difference between the dSphs and the Galactic halo MDFs remains. However, it is obviously of critical importance that the different degrees of incompleteness are well understood, and this is particularly complicated in the halo. What is shown as the Galactic halo MDF in Fig. may still change, but it is most likely that the two halo MDFs shown represent a reasonable range of possibilities. This difference provides a challenge to models where all of the Galactic halo builds up from the early merging of dwarf galaxies, because it begs the question: where have all the most metal-poor stars in the Galactic halo come from? Was there a pre-enrichment of dwarf galaxies, perhaps by the most metal-poor stars which we appear to find only in the halo (Salvadori, Ferrara & Schneider 2008)? This mismatch applies equally to any merging scenario that extends over a significant fraction of a Hubble time, as it means that dSph and dI galaxies could not have merged to form the halo except at very select moments in the past (see Section 4). Fig. 10 does not include uFds, and there is evidence that they may include more metal poor stars than are to be found in dSph (e.g., Kirby et al. 2008, Frebel et al. 2009).
Figure 10. Comparison of the cumulative metallicity distribution functions (MDFs) of the stars in the mean bootstrapped Hamburg-ESO survey sample as a solid black line, and the new bias corrected Galactic halo MDF from Schoerck et al. (2008) as a solid blue line. These are compared to the MDFs for 4 dSphs from the DART survey (Helmi et al. 2006). The halo and the dSph MDFs have been normalised at [Fe/H] = -2.5, which assumes that the completeness of the halo MDF is well understood at this metallicity and below. Note that at present the Fornax dSph lacks sufficient stars at [Fe/H] < -2.5 to be properly present on this plot.
3.1.2 MORE DISTANT dSPHS: There are also more isolated dSphs, within the Local Group but not obviously associated to the MW or M 31, for example Antlia, Phoenix, Cetus and Tucana. They have also benefited from spectroscopic studies (e.g., Tolstoy & Irwin 2000, Gallart et al. 2001, Irwin & Tolstoy 2002, Lewis et al. 2007, Fraternali et al. 2009). Antlia, Tucana and Phoenix have HI gas in their vicinity, but after careful study, only for Phoenix and Antlia has the association been confirmed. These spectroscopic studies have shown evidence for rotation in Cetus and Tucana at a similar magnitude to Scl (Lewis et al. 2007, Fraternali et al. 2009). This rotation is consistent with the flattening of the galaxy. In Phoenix the kinematics and morphology of HI gas compared to the stellar component suggests that the HI is being blown out by a recent star formation episode (e.g., Young et al. 2007). This supports the theoretical predictions of this effect (e.g., Larson 1974, Mac Low & Ferrara 1999). All these more distant dSphs are far enough away from the MW that any strong tidal influence is likely to have been several Gyr in the past.
3.1.3 DWARF GALAXIES AROUND M 31 There are also diffuse dEs and dSphs around M 31 where spectra have been taken of RGB stars. From a sample of 725 radial velocity measurements in NGC 205, it was found to be rotating at 11 ± 5 km s-1 (Geha et al. 2006). A careful study of the structural properties of the dSphs around M 31 shows that on average the scale radii of the dSphs around M 31 are about a factor 2 larger than those of dSphs around the MW at all luminosities (McConnachie & Irwin 2006). This could either be due to small number statistics or it might suggest that the tidal field of M 31 is weaker than that of the MW or the environment in the halo of M 31 is different from that of the MW halo.
3.2. Late-Type Dwarfs
Most of what we know about the kinematics of dI galaxies comes from observations of their HI gas (e.g., Lo, Sargent & Young 1993, Young et al. 2003), which is strongly influenced by recent events in the systems. For example, the velocity dispersion measured in the HI gas is predominantly influenced by on-going star formation processes. Thus the HI velocity dispersion is almost always ~ 10 km s-1 in any system, from the smallest dIs to the largest spiral galaxies, regardless of the mass or rotation velocity of the HI. This makes it difficult to compare the kinematic properties of dI and dSph galaxies.
Likewise, most of the metallicity information comes from HII region spectroscopy (e.g., Pagel & Edmunds 1981, Hunter & Gallagher 1986, Skillman, Kennicutt & Hodge 1989, Izotov & Thuan 1999, Kunth & östlin 2000, Hunter & Elmegreen 2004) or spectroscopy of (young) massive stars (e.g., Venn et al. 2001, 2004b). For a few BCDs the FUSE satellite has also provided abundances for the neutral gas (e.g., Thuan, Lecavelier des Etangs & Izotov 2002, Aloisi et al. 2003, Lebouteiller et al. 2004). Thus the metallicity measures come from sources which are only a few million years old and the product of the entire history of star formation in a galaxy. By contrast, in dSphs the abundances are typically measured for stars older than ~ 1 Gyr, and the value quoted is some form of a mean of the values measured over the entire SFH. This makes it difficult to accurately compare the properties of early and late-type dwarf galaxies, as there are few common measurements that can be directly compared. Such comparisons have been attempted by Skillman, Kennicutt & Hodge (e.g., 1989), Grebel, Gallagher & Harbeck (e.g., 2003).
From studies of the HI gas in these systems it has been found that for the smallest and faintest dIs if rotation is detected at all, it is at or below the velocity dispersion (e.g., Lo, Sargent & Young 1993, Young et al. 2003). Despite this, in all dIs the HII regions in a single galaxy, even those widely spaced, appear to have, within the margins of error of the observations, identical [O/H] abundances. So it appears that either the enrichment process progresses uniformly galaxy wide, or the oxygen abundance within an HII region is affected by some internal, self-pollution process which results in uniform [O/H] values (e.g., Olive et al. 1995). However, the clear gradients in HII region abundances seen in spiral galaxies argue against this explanation.
One dwarf galaxy which has both HI and stellar kinematic information is the faint transition type dwarf LGS 3 (MV = -9.9). The stellar component looks like a dSph, dominated by (old) RGB stars (see Fig. 5). There are no HII regions (Hodge & Miller 1995), and the youngest stars are around 100 Myr old (Miller et al. 2001). However, the galaxy also contains 2× 105 M of HI (Lo, Sargent & Young 1993, Young & Lo 1997), which is more extended than the optical galaxy, and with no convincing evidence of rotation. Cook et al. (1999) measured the radial velocities of 4 RGB stars in LGS 3 at the same systemic velocity as for the HI, confirming the association. They found the stellar velocity dispersion of these 4 stars to be 7.9-2.9+5.3 kms-1. This leads to a high M/L (> 11, perhaps as high as 95), similar to other dSphs. However this sample of radial velocities is hardly sufficient.
LGS 3 used to be the lowest luminosity galaxy with HI, but that was before the recent discovery of Leo T (Irwin et al. 2007). Leo T contains 2.8 × 105 M of HI gas (Ryan-Weber et al. 2008), and it does not contain HII regions or young stars. The total dynamical mass was determined to be 8.2 ± 3.6 × 106 M, with M/L ~ 140. Leo T seems to be a particularly faint dwarf (MV = -8), at a distance of ~ 420 kpc (see Table 1). It is about 2 magnitudes fainter than the other transition type systems like LGS 3 and Phoenix, and 0.6 magnitudes fainter than the faintest dwarf spheroidal system, Draco.
Leo T is another of the very few systems for which kinematics have been derived from HI and velocities of individual stars. Simon & Geha (2007) measured the radial velocities and metallicities of 19 RGB stars, and found the average metallicity to be [Fe/H] ~ -2.3 with a range of ± 0.35. They found a central optical velocity of ~ +38 ± 2 km s-1, a velocity dispersion = 7.5 ± 1.6 km s-1 and no obvious sign of rotation. This is comparable to the HI value, HI = 6.9 km s-1, also with no sign of rotation (Ryan-Weber et al. 2008). This is the smallest and lowest luminosity galaxy with fairly recent star formation known. The inferred past SFR of 1.5-2 × 10-5 M/year might be sufficiently low that gas is neither heated nor blown out in this system, thus allowing it to survive (de Jong et al. 2008a).
3.3. Ultra-Faint Dwarfs
The stellar kinematics and metallicities of individual stars play an important role in determining what kind of systems uFds are. These measurements can attempt to quantify the degree of disruption uFds may have undergone and if they should be considered faint galaxies or some kind of diffuse globular clusters, such as are seen around M 31 (e.g., Mackey et al. 2006). These systems are so embedded in the foreground of our Galaxy, both in position and in velocity, and the total number of their stars is so often so small (many have MV -4) that studies can easily get different results for even the most fundamental properties, like their size and their dark matter content depending upon membership selection (e.g., Ibata et al. 2006, Simon & Geha 2007, Siegel, Shetrone & Irwin 2008, Geha et al. 2009).
The basic kinematic properties of these systems (e.g., Simon & Geha 2007) show evidence that they are much more dark matter dominated than previously known systems, with M/L ~ 140 - 1700, although this study did not try to correct for tidal effects which are almost certainly present. Many of these galaxies also have very small numbers of stars, and thus test particles, and so the properties of the dark matter content are often extrapolated from a small central region.
Simon & Geha (2007) also determined the average stellar metallicities ([Fe/H] -2) of uFds and found them to be lower than in most globular clusters, and with a scatter that is not expected in globular clusters. The average metallicities are also lower than in other more luminous dwarf galaxies (see also, Kirby et al. 2008), and the lowest metallicity stars appear to be more metal-poor than the most metal poor stars found in the brighter "classical" dSphs (Norris et al. 2008, Frebel et al. 2009). Norris et al. (2008) also found evidence for carbon-rich metal-poor stars in Boo I, which suggests that the metal-poor stars in uFds maybe more similar to those found in the MW halo, where a large fraction of stars more metal poor than [Fe/H] < -4 are carbon rich.
CVn I, at a distance of 220 kpc, at MV ~ -7.9 (Zucker et al. 2006b), is one of the brighter examples of uFds, and bears much similarity to classical dSphs, both in structural properties, kinematics and SFH. From a CMD analysis (Martin et al. 2008a) found that the galaxy is dominated by an ancient population (> 10 Gyr old), with about 5% of its stars in a young blue plume ~ 1.4-2 Gyr old. It has well populated, broad, RGB and HB which have been studied spectroscopically (Ibata et al. 2006, Simon & Geha 2007). With a sample of 44 stars Ibata et al. (2006) detected the presence of two components with different metallicities and velocities. However Simon & Geha (2007) with a much larger sample of 212 stars were not able to reproduce this result.
A recently discovered uFd is Leo V (Belokurov et al. 2008), at a distance of 180 kpc with MV = -4.3. Fig. 6 illustrates how hard it can be to quantify the structural properties and distinctness of these small diffuse systems several of which may be embedded in Galactic scale streams. Leo V may be related to Leo IV, to which it is very close to both spatially and in velocity. They may both be the remnants of the same tidal interaction, but given how metal poor their stars appear to be they would have to be the outer envelopes of dwarf galaxies (which are known to have metallicity gradients), like Sgr, and not disrupted globular clusters.
Thus the nature of some of the uFds still remains a mystery, and there is likely to be a range of origins for these systems. The brighter uFds (MV < -5) are relatively easy to study and do appear to be a low mass tail to dSphs (e.g., CVn I) and dI/transition systems (e.g., Leo T). This reduces the sizes of objects in which stars can form in the early universe (e.g., Bovill & Ricotti 2009, Salvadori & Ferrara 2009), and it follows the trend that these smaller systems could barely enriched themselves. This could be either due to efficient winds, or inefficient star formation, both of which could be the result of a low galactic mass.
It remains a matter of conjecture what are the fainter systems such as Coma and UMa II (MV ~ -4). From its highly irregular stellar distribution UMa II is clearly a disrupted system that sits behind high velocity cloud complex A (Zucker et al. 2006a, Belokurov et al. 2007a). Coma has very similar properties to UMa II, and it also has an irregular extended shape (Belokurov et al. 2007b). UMa II is one of the few objects which lie in the gap between globular clusters and dwarf galaxies in the left hand plot of Fig. 1. This is a region which has so far not been populated by either galaxies or globular clusters but if the large dark matter masses of the uFds are correct then they are more likely to be an extension of the galaxy class than of the globular cluster class.