Studies of the astrophysics of dwarf galaxies were relatively difficult in the era before sensitive multi-wavelength detectors came into routine use (Hodge 1971). The low surface brightness (LSB) nature of dwarfs make them difficult to detect at optical wavelengths, a problem which continues and is only slowly being corrected as more sensitive optical sky surveys have become available (Impey et al. 1988; Davies et al. 1989; Dalcanton et al. 1997). Fortunately, gas-rich dwarfs are readily detected in the H I 21 cm emission line, and new programs to systematically search the sky in the H I 21 cm line should yield complete samples of late-type dwarfs (Henning 1995; Briggs 1997a,b). Despite these observational obstacles, dwarf galaxies yield important perspectives on galaxy evolution, including the chemical evolution (Skillman 1998), star formation processes (e.g. Hunter & Gallagher 1986), and the characteristics of dark matter (e.g., Gallagher & Wyse 1994; Burkert & Silk 1997; Mateo 1998). Furthermore, in some theories of galaxy formation (such as cold dark matter), present-day dwarfs can be viewed as left-over building blocks of galaxy formation, and potentially offer us insights into this key astrophysical process (e.g. Burkert 1995).
The past 15 years have seen a dramatic increase in our knowledge of the common but faint dwarf galaxies. This is well-illustrated by the pair of landmark conferences ``Star Formation in Dwarf Galaxies and Related Objects'' (Kunth et al. 1985a) and ``ESO/OHP Workshop on Dwarf Galaxies'' (Meylan & Prugniel 1994) with many important discussions and meetings in between (and since!). This review primarily considers the structures of the stellar bodies of dwarf galaxies, with an emphasis on the evolutionary histories of their stellar populations. In this context it is also interesting to explore whether both Magellanic Clouds should be considered to be part of the extended family of dwarf galaxies; I believe this is a correct decision. Obviously the visible structures of galaxies do not tell the full story, and the variety of papers in this conference provide an excellent overview of the wider range of astrophysical issues.
Morphological definitions of dwarf galaxies are complicated by the existence of the two main structural classes: the disk and spheroidal systems (see van den Bergh 1998). However, both types share a common characteristic of low stellar densities, leading to their LSB natures, and low degrees of central light concentrations. Furthermore, the dE/dSph differ from giant Es in the way that surface brightness varies with luminosity; in giant Es surface brightness slowly declines with increasing L, while in dwarfs surface brightness and luminosity track together (Wirth & Gallagher 1984; Kormendy 1985).
The situation is more complicated in disky dwarfs, where star formation plays a key role. In small disk galaxies the surface brightness levels and profiles depend critically on the recent star formation history (SFH). BCDGs have high current star formation rates (SFRs) and as a result can be optically bright, while in other dwarfs bits of star formation distributed somewhat randomly produce the trademark irregular optical structures (see Fig. 1). However, the transition Sdm-Sm galaxies, which morphologically lie between the normal spirals and true irregulars, can be relatively symmetric, although in barred systems large-scale asymmetries are frequently introduced by the presence of off-center bars and associated lopsided disks (de Vaucouleurs & Freeman 1972; Odewahn 1994; Fig. 2).
Dwarf galaxies also differ from giants in that composite disk-bulge-spheroid structures seem to be rare. While Schombert et al. (1995) suggested bulges exist in some low luminosity Sd-Sm spirals, a closer examination suggests that many of these features are brighter inner regions of the flattened disks, and not spheroidal bulges (Matthews & Gallagher 1997; cf. Wyse et al. 1997). Modern CCD observations are especially sensitive to the presence of compact bulges of moderate-to-low optical surface brightnesses, features that are not seen in deep measurements of edge-on low luminosity disk galaxies (see Gallagher & Hudson 1976; Goad & Roberts 1981; Matthews et al. 1998). The question of diffuse stellar spheroids or analogs to the Galactic metal-poor old halo is more difficult to resolve. Minniti & Zijlstra (1996) have found some evidence for such structures, and they could be relatively common but difficult to detect.
Of course, the dEs and dSphs are pure spheroidal galaxies whose stellar halos extend to surprisingly large radii and amazingly low stellar volume densities (Caldwell et al. 1992). Among the least luminous dI galaxies the separation between disk and spheroidal systems is ill-defined. In these galaxies rotation velocities are less than or equal to typical H I gas one-dimensional random motions of σ = 5-10 km s-1 (Lo et al. 1993; Young & Lo 1997). At the bottom of the dwarf sequence we find non-rotating (at least in the gas) dI galaxies along with non-rotating (in the stars) dSph galaxies. Paradoxically extreme dwarf galaxies (L ≤ 10-3 Lsun) with their snail-speed internal velocities must be dynamically `warm' since they have vrot / σ ≤ 1. Thus the spheroidal and disk dwarfs must morphologically merge at the faint end of the galaxy luminosity distribution; all very faint galaxies become spheroidal. This behavior is closely related to the apparent existence of a minimum velocity dispersion of about 5 km s-1 in any galaxy, which in turn leads to the inferred presence of large amounts of dark matter in some dSph (e.g. Gallagher & Wyse 1994; see also Kroupa 1997), but none in some dIs (e.g. Young & Lo 1997). We still do not know if dark matter is a universal feature of small galaxies, or whether dwarf galaxies exist both with and without dark matter.
The Pegasus dI system has MV = -11.8, and provides a nearby example of a faint dI galaxy. It supports a modest amount of recent star formation; it has been making stars for at least several Gyr, possibly at a higher rate in the past (Gallagher et al. 1998). Pegasus contains a small mass of H I gas, concentrated within the optically visible galaxy (Lo et al. 1993). Irregulars with slightly more active star formation and lower central concentrations of luminous matter, however, have morphologies that clearly live up to their names. Such galaxies often have very low surface brightnesses and can optically appear to be limited to a set of almost disconnected luminous patches unless very deep images are obtained. Many LSB Im systems are gas rich, with their H I masses comfortably exceeding the inferred stellar masses (e.g. van Zee, Haynes, & Giovanelli 1995). The most extreme cases, such as the ``H I-galaxy'' A1225+01 contain at least 80% of their visible baryonic mass in the form of gas (Giovanelli & Haynes 1989). Thus these types of Im/dI systems are the least evolved nearby galaxies, with high gas mass fractions, blue optical colors of dominant young stellar populations, and low chemical abundances (Salzer et al. 1991; Giovanelli et al. 1991; van Zee et al. 1997). The tantalizing possibility also exists that some of these galaxies are really young in the sense of having come into existence as star-forming systems within the last few Gyr (Thuan et al. 1997; Tolstoy et al. 1998).
The irregular galaxy structural family extends upwards in luminosity; Im galaxies with LB > 0.1 L*B are relatively common (Hunter & Gallagher 1986). These tend to have higher optical surface brightnesses than faint irregulars, and frequently are Magellanic spirals, with off-center bars and rudimentary spiral structures. Such galaxies often have a a clustering of H II regions along one of the bar, giving them a distinctive asymmetric appearance (e.g., NGC 6822). The most luminous Sm galaxies like NGC 4449 have very high optical surface brightnesses in combination with very blue colors, and therefore stand out in UV-selected samples. Galaxies with these characteristics are preferentially chosen in samples of high redshift galaxies; they have small or even negative k-corrections and their high surface brightnesses partially offset the cosmological dimming of intensities with increasing redshifts. The Sm/Im galaxies are UV-bright because they combine moderately high SFRs per unit area with reduced amounts of extinction due to their low metallicities and disk densities. While at the present epoch these traits are limited to under-evolved ``dwarfs'', they may have been more common at higher redshifts when giant galaxy disks were just beginning to form. UV-bright irregulars are useful analogs to objects seen at high redshift, but this does not necessarily mean that high-z compact blue galaxies are related to present-day irregulars.
Patterns of star formation within irregular galaxies are revealed by observations of H II regions, e.g., through narrow-band imaging (Hunter et al. 1993). H II regions have lifetimes of < 5 Myr and thus give snapshots of star formation activity. Aside from correlations with bars, star formation in irregulars tends to be random, but with the mean level of star-forming activity decreasing radially roughly in proportion to the integrated star light. This suggests that the disks of these galaxies are in rough equilibrium in that the entire disk is evolving approximately as a single unit. Global SFRs can also be determined from Halpha observations, and these show a huge range in irregulars (e.g., Hunter 1997). The range in time scales for star formation to exhaust all of the H I within Im systems varies from a few Gyr to several Hubble times. However, it is not immediately obvious that all of the gas in these galaxies is accessible to star formation. For example, the outer H I disks, which do not support detectable levels of star formation in most irregular galaxies, may not mix into the actively star forming inner regions of these galaxies (see Gallagher & Hunter 1984; van Zee et al. 1997; Hunter et al. 1998).
Im galaxies have modest rotation speeds, typically of < 100 km s-1, and their rotation curves are nearly linear throughout the optical body of the galaxy. Im and Sm galaxies are nearly rigid-body rotators. This form of the rotation curve indicates a low degree of central mass concentration.
Galaxies that kinematically resemble Im systems can sometimes support intense bursts of star formation. When such events occur, the morphology of the galaxy is dominated by what is usually a centrally-concentrated star forming region with very high surface brightness. These BCDGs have been extensively investigated over the past two decades (e.g., Thuan & Martin 1981; Marlowe et al. 1997). When intense star formation occurs in an irregular galaxy, even on a more modest scale than in BCDGs, the star formation products often include super star clusters (Meurer et al. 1992, 1995). The NGC 2070 star cluster in 30 Doradus within the LMC is a beautiful nearby example of super star cluster. Such clusters typically produce LV > 106 Lsun from within a half light radius of <10 pc; they resemble young versions of globular star clusters (O'Connell et al. 1994). While the presence of dense super clusters in low density galaxies remains an intriguing puzzle, they may partially explain the huge range in SFRs seen in Im and related systems. The birth of a single super star cluster requires the production of ≥105 Msun of stars in less than 1 Myr, a huge jump for galaxies where normal SFRs are often <0.01 Msun yr-1.
The LMC displays the defining pattern of an off-center bar, but its stellar disk seems not to have suffered substantial damage from interactions with the SMC or Milky Way. This is beautifully displayed in Fig. 2 that shows the projected number densities of resolved stars brighter than J ≅ 22 from the U.S. Naval Observatory-A1.0 catalogue (Monet et al. 1996). This LMC map is not affected by reflection nebulae and gives a clear view of the basic structure of the LMC as a disk galaxy observed at a moderate inclination angle.
The LMC is dwarf-like in having an asymmetric bar, and also in having a roughly linear rotation curve over much of its optical body (Feitzinger 1980; Kim 1998). It also lacks a bulge or nucleus and its globular clusters are in a thick rotating disk rather than a spheroid. However, its luminosity is that of a small giant galaxy, and its outer H I disk has been strongly disturbed by its presence with the SMC in orbit around the Milky Way. It also shares with dwarfs the lack of a clear radial gradient in H II region oxygen abundances, and an admixture of stars covering a wide range in ages within the disk. An excellent summary of properties of the Magellanic Clouds is in Westerlund (1997).
We should feel secure in electing the LMC as an honorary dwarf galaxy for this meeting.
The astrophysical characteristics of low luminosity dSph galaxies are still mainly derived from observations of the Milky Way's satellites. These objects are in some ways regular; e.g. in terms of their structures, but other features show wide variations (Gallagher & Wyse 1994). For example, the stellar population age mixes range from old, globular cluster-like in the Ursa Minor dSph, to a mixture of intermediate ages in Carina, to dominate intermediate age in the Leo I system (Mateo 1998). Our hope that stellar population ages and galaxy kinematics would be tightly correlated, as in Baade's original two stellar populations model, breaks down in the Galactic retinue of dSph galaxies.
Many dSph/dE galaxies contain significant populations of very old stars, which give rise to horizontal branches in their color-magnitude diagrams (e.g. Gallagher & Wyse 1994; Han et al. 1997; Mateo 1998). The idea that spheroidal galaxies formed early-on still seems to have wide, although maybe not universal, validity within the Local Group. However, following formation, many spheroidal galaxies continued to actively form stars over extended time periods; in some cases star formation continued until only a few Gyr ago. Local Group spheroidal dwarfs have experienced rapid evolution in the past 3-5 Gyr, from being galaxies where star formation was common, to the current near stellar fossils with negligible stellar birth rates (see Da Costa 1998). If this picture is correct, then we need to identify the astrophysical `clocks' which determine the varying SFHs among otherwise rather similar galaxies.
Could these effects be due to special circumstances near the Milky Way, such as the ballet with the Magellanic Clouds? This seems unlikely. Star formation also potters along in the luminous M31 dE companions NGC 205 and NGC 185, while the currently inactive galaxy NGC 147 contains a substantial intermediate age stellar component (Han et al. 1997; see Fig. 3). Complex SFH appear to be the rule rather than the exception in the closest spheroidal dwarfs. On the other hand the idea of environmental meddling within the Local Group is attractive because the color-luminosity relationships are very well-defined for dSph/dEs in the Coma cluster (Secker et al. 1997). Maybe these urbanized dSph/dE members of clusters had their star formation truncated long enough ago that their optical colors are driven by metallicity effects? This would require that star formation ceased in most cluster dSph/dE galaxies more than ≅ 3-5 Gyr before the present, and would imply different SFHs for field and cluster dSph/dEs.
The range in ages and metallicity in Local Group dSph/dE is inconsistent with models where most of the intermediate age stars were created by a single violent starburst operating on an internal galactic dynamical time scale of <0.1-0.2 Gyr. This type of starburst appears to be needed if dEs were once the `faint blue galaxies' (Babul & Ferguson 1996). Understanding the transition of present-day dwarf spheroidals from star-forming to stellar fossil systems is still a major problem (Skillman & Bender 1995).
This origins mystery is further complicated by the kinematics of dSph/dE galaxies. These are non-rotating systems, i.e. they all have V/σ < 1. Their flattened shapes therefore should result from triaxial density distributions, which are most readily produced by collisionless processes, such as mergers or interactions. In addition, despite their low optical densities, the dSph/dEs are a robust family of galaxies; they routinely survive near giant spirals and in the cores of dense clusters of galaxies, where they are by far the most common type of galaxy. One promising model for the presence of fleets of faint spheroidal dwarfs in regular clusters is the `harassment' and subsquent heating of field disk galaxies by time varying gravitational forces as they fall into clusters (Moore et al. 1998). It remains to be seen whether this model reproduces the regularities of cluster spheroidal dwarf galaxy populations.
The idea that dwarf galaxies have distinctive structural features follows from their major observed properties:
The origins of disk dwarf galaxies should be easiest to understand, since their visible components must have initially consisted mainly of gas that underwent dissipation and settled into rotationally supported disks. The basic features of irregulars are associated with their low degrees of central mass concentration. Why might this preferentially occur in small disk galaxies? Part of the answer may lie in the way gas was accreted. If material is added slowly and with moderate angular momentum, then an initial build-up of high mass material in the central regions might be avoided. In addition, if the effective viscosity of the resulting disk is low, as might be expected in rigidly, slowly rotating irregulars, this would further inhibit the flow of gas towards the center. In particular, it would seem to be necessary for Im galaxies not to capture other comparably-sized dwarfs in order to avoid driving too much material into their central regions.
A second factor could be the relatively high dark matter mass fractions in some, but not all Im galaxies. If the dark matter mass overwhelms that of the baryonic disk, then a dark matter halo may be relatively rigid; i.e. its response to the luminous matter will be limited. As a result the dark matter can sustain its initial rigid-body rotation curve associated with the predicted flat central density profile (Burkert 1995), even when luminous matter is present. This means that gas with rather low specific angular momentum will orbit at comparatively large radii, an effect which further reduces the concentration of mass to the center of the galaxy; a flat rotation curve extending to the center of a galaxy maximizes the concentration of low angular momentum material to small radii. The present best candidate model for the origins of the irregular family is disk galaxies which failed to form central mass concentrations as a result of a combination of initial conditions and the nature of subsequent capture of gas.
The resulting low density disks in irregulars should be inefficient in making stars. While a global theory of star formation is not yet in hand, it is clear that denser regions form stars more rapidly than those with low densities. Similarly, the large cores of dark-matter dominated Im and Sm galaxies provide an opportunity for off-center structures to form, as suggested by Levine & Sparke (1998). They consider the case of an external perturbation and demonstrate that this could lead to an off-center bar moving along a retrograde orbit within a dark matter core. This behavior is also consistent with Odewahn's (1994) observations that many SBm galaxies with off-center bars have close companions, although this is not always the case (e.g., NGC 6822).
While it appears that we might be gaining an understanding of irregulars, the situation is less optimistic for the dEs. Standard models in which dEs form due to homogeneous collapse can reproduce many but not all of the features of this class of galaxy. These models are the opposite to those of the irregulars in that an initial collapse on a dynamical time scale yields an over-dense galaxy which experiences a giant starburst, leading to extensive gas loss via a galactic superwind. The galaxy reacts to mass loss by expanding; this converts an initially relatively dense galaxy into a low density dwarf (see Dekel & Silk 1986; Fukunaga-Nakamura & Tosa 1989); dE/dSph systems become low density dwarfs by a quirk of evolution. The basic model is supported by low stellar metallicities in spheroidal dwarfs (e.g. Vader 1986, 1987), which in combination with little or no gas argues that large fractions, in some cases >90%, of the baryonic mass was lost. When multi-phase versions of the ISM are included in the models, gas ejection from the initial starburst in a dE is not complete and extended star formation can occur (Prenzel 1993; Loxen 1996; Hensler et al. 1998; see also Spaans & Norman 1997).
Unfortunately, initial starburst models for spheroidal dwarfs seem to fit neither the observed stellar metallicity distributions, nor the prominent intermediate age components in some dSph (see Da Costa 1998; Mateo 1998). They also fail to provide a physical basis for the lack of rotation in dSph/dE systems; it is important to understand the bimodal distribution of specific angular momentum in the bodies of intermediate luminosity dwarfs (cf. Dalcanton et al. 1997b), and to study the few cases of objects which seem to be intermediate between the dSph/dE and dI structural families (e.g. Patterson & Thuan 1992). For these reasons attention continues to be given to the possibility that the spheroidal dwarfs are systems which formed as disk galaxies and then were converted into spheroids by collisions or mergers. This approach also faces a number of difficulties, including how to match surface brightness - luminosity correlations and how to get rid of most of the interstellar gas without overly enriching the remaining stars in the process (e.g. Bothun et al. 1986; Binggeli 1994; Skillman & Bender 1995). Simple fading clearly does not make a dI into a dE; a much more drastic structural change is required for such a transformation. We should also keep in mind the suggestion Silk et al. (1987) that evolution could go from non-rotating dSph/dE systems to dI galaxies if early-type dwarfs capture gas in galaxy groups, thereby converting stellar fossil galaxies into systems which support star formation.
The present understanding of dwarf galaxies is that a limited set of structural forms contain considerable ranges in stellar populations, presumably reflecting variations in individual SFHs. Dwarfs may be the simplest class of galaxy, and in this case they are telling us that all galaxies are complex. Despite these difficulties, a large selection of dwarfs are known in and near the Local Group, and more no doubt wait to be discovered. With such a rich sample of objects to work with, we have an excellent opportunity to disentangle the various internal and environmental astrophysical effects to gain a basic understanding of this small but beautiful type of galaxies.