Theories for the formation and evolution of dE galaxies are reviewed in Sect. 7. However, to motivate the discussion of the observations, it is useful first to describe briefly the theoretical milieu.
While they appear to be the natural low-luminosity extension of the E galaxy family, the structural differences between giant and dwarf E's suggest that different physical processes govern the evolution of each type. The standard picture is that dwarf galaxies, like giants, formed from the gravitational collapse of primordial density fluctuations. In hierarchical models, once the cosmological parameters and power spectrum are specified, the evolution of dissipationless dark-matter halos can be calculated (e.g. following Press and Schecter 1974). However, the emergence of galaxies as separate entities in the clustering hierarchy depends on the ability of the baryons in a given overdense region to cool and form stars. The requirement that the cooling time be less than the dynamical time sets a natural upper limit to the galaxy mass function (Binney 1977; Rees and Ostriker 1977; Silk 1977; White and Rees 1978). Once the baryons cool and start to form stars, feedback of energy into the interstellar medium is likely to regulate (or even halt) any subsequent star formation, and mass loss may modify the galaxies' structure. Originally proposed as a mechanism for clearing gas from giant elliptical galaxies (Burke 1968; Matthews and Baker 1971), winds from supernovae were soon recognized as potentially important in governing both the structure and stellar populations of low-luminosity ellipticals (Larson 1974). Models that invoke the cessation of star formation by supernova-driven winds provide a plausible explanation for the variation of density (surface brightness) and metallicity (color) with luminosity (Larson 1974; Saito 1979; Vader 1986; Dekel and Silk 1986; Arimoto and Yoshii 1987). The status of the observed correlations between color, metallicity, and luminosity are reviewed in Sect. 4. Such models also predict high values of M / L for low-luminosity galaxies; constraints from Local Group dE's will be discussed in Sect. 3.
Hierarchical models predict a power law for the low-mass tail of galaxy halos. Simple scaling relations between mass and star-formation efficiency typically also predict a power law, with a different exponent, for the faint end of the galaxy luminosity function (White and Rees 1978; White and Frenk 1991). Because dwarf galaxies condense from smaller perturbations than giants, models based on Gaussian random-phase fluctuations also predict that dwarfs will be less clustered than giants (Dekel and Silk 1986; White et al. 1987). The luminosity function and clustering properties of dwarf galaxies are thus of great interest for testing cosmological models, and will be reviewed in Sect. 5 and Sect. 6.
While models involving supernova-driven winds have been the most thoroughly developed, other processes may be important - or even dominant - in governing dE evolution. Other ways to remove gas include stripping by a nearby galaxy corona (Einasto et al. 1974; Lin and Faber 1983; Kormendy 1986), or sweeping by an intracluster medium. Such mechanisms predict an environmental dependence of dwarf galaxy properties that would not be expected if internal feedback were always the dominant regulator of star formation. It is also possible that star-formation is triggered in dwarf galaxies by tidal interaction with a neighbor (Lacey and Silk 1991), shocks caused by interaction with intergalactic gas in groups or clusters (Silk et al. 1987), or variations in the ionizing UV radiation field (Babul and Rees 1992; Efstathiou 1992). Triggered star formation at late epochs (0.3 < z < 1.0) could help account for the apparent excess of low-luminosity galaxies in deep redshift surveys. In Sect. 4 we will consider the constraints set on such models from observations of the stellar populations in nearby dwarfs.
Alternatively, or perhaps in addition to these other effects, it is possible that star-formation in dwarfs is a cyclical process as gas repeatedly cools to form stars, then gets reheated (but not completely ejected) by OB stars and supernovae (Gerola et al. 1980; Lin and Murray 1992). Finally, there are suggestions that some dwarfs form as debris from either the explosion (during an initial burst of star formation) or collision of galaxies (Gerola et al. 1983; Barnes and Hernquist 1992; Mirabel et al. 1992). Such models are clearly radically different from the collapse onto primordial density perturbations, and expectations for the scaling relations between color, luminosity, surface brightness, and velocity dispersion remain to be worked out. Nevertheless, formation of at least some dwarfs as debris from collisions may help explain the large number of dwarfs in clusters (Sect. 6.2) and the scatter in the color-magnitude relation (Sect. 4).