Annu. Rev. Astron. Astrophys. 1989. 27: 235-277
Copyright © 1989 by . All rights reserved

Next Contents Previous


In old stellar systems, colors are a complicated measure of metallicity (Burstein et al. 1984, Aragon et al. 1987) and age (O'Connell 1986). Colors, color gradients, and their correlations with other galaxy properties can therefore be used to test theories of galaxy formation. A few of the many reviews of this subject include Faber (1977), Pagel & Edmunds (1981), Burstein (1985), Norman et al. (1986), and Thomsen & Baum (1988).

The correlation between color and luminosity for early-type galaxies is well known (Sandage 1972, Visvanathan & Sandage 1977, Sandage & Visvanathan 1978a, b, Strom et al. 1976, 1978, Frogel et al. 1978): More luminous galaxies are redder and thus more metal rich. For example, Sandage & Visvanathan find that log L = 4.1 (u - V) + constant. Also, the centers of most early-type galaxies are redder than their envelopes; typical gradients are Delta(b - V) appeq - 0.03 and Delta(u - V) appeq - 0.10 magnitudes per decade in radius. Color gradients in bulges of spirals are generally stronger than those in ellipticals (Wirth 1981, Wirth & Shaw 1983). The same effects are seen in spectroscopic metallicity indicators (Faber 1973, Terlevich et al. 1981, Tonry & Davis 1981). From population synthesis models, Tinsley (1978) derives the mass-metallicity relation, Z propto M0.25, where Z is the logarithm of the metallicity. Metallicity variations can partly, but not entirely, explain the observed dependence of M/L on luminosity (Smith & Tinsley 1976):

These results can be understood within the framework of dissipative galaxy formation (Larson 1974, 1975, Silk & Norman 1981, Carlberg 1984a, b, Arimoto & Yoshii 1987, Yoshii & Arimoto 1987, Matteucci & Tornambe 1987). Carlberg's models in particular avoid some technical limitations of Larson's pioneering work and make more detailed predictions. In these models, the removal of enriched gas by supernova-driven galactic winds is more efficient for less massive galaxies (see also: Section 8.1). In this spirit, the color-luminosity relation is recast as a metallicity-escape velocity relation by Vigroux et al. (1981): Z propto Ve0.9. Similarly, colors and metallicities correlate better with central velocity dispersions than with luminosities. Carlberg (1984b) also predicts the existence of a second parameter in the Faber-Jackson and mass-metallicity relations. Larson and Carlberg both predict that color and metallicity gradients should-be stronger in more massive galaxies. Finally, they make testable predictions about the relative shapes of isophotes and isochromes (i.e. isometallicity contours). In Larson's models, isochromes are considerably flatter than isophotes. However, Larson's ellipticals are supported by rotation, which we now know is incorrect (e.g. Illingworth 1981, Davies et al. 1983). Carlberg's models are generally supported by velocity anisotropy; then isochromes are only slightly flatter than isophotes.

CCD photometry has provided high-quality measurements of color gradients for large numbers of galaxies. Boroson et al. (1983), Davis et al. (1985), Cohen (1986), Boroson & Thompson (1987), and Bender & Möllenhoff (1987) present data on relatively small samples of ellipticals, mostly in the Virgo cluster. They conclude that color gradients are common in ellipticals. In the absence of nonthermal emission or recent star formation, colors always get redder toward the center. Interestingly, isophotes and isochromes generally have the same shape. In fact, isochromes are occasionally rounder than isophotes (Boroson et al. 1983). This is consistent with Carlberg's but not Larson's models.

The interpretation of color gradients in terms of stellar population gradients has been discussed recently by Efstathiou & Gorgas (1985), Gorgas & Efstathiou (1987), Davies & Sadler (1987), Couture & Hardy (1988), and references therein. They present extensive evidence for gradients in Mg2 indices. Assuming the somewhat uncertain conversions between Mg2 index and metallicity (Terlevich et al. 1981) and between color and metallicity (Strom et al. 1976, 1978, Tinsley 1978), they find that the Mg2 and color gradients are mutually consistent and imply typical changes of Delta[Fe / H] ~ - 0.2 per decade in radius. In excellent papers, Baum et al. (1986) and Thomsen & Baum (1987, 1988) derive metallicity gradients from narrowband surface photometry. They also find that isochromes are not flatter than isophotes, in agreement with spectroscopic results. Similar photometric measurements of the Mg2 index are reported by Vigroux et al. (1988). Also, Vader et al. (1988) find that Mg2 gradients correlate well with broadband color gradients. Further constraints are obtained by Peletier and coworkers (Peletier et al. 1987, 1988a, b, c, Peletier & Valentijn 1988, Peletier 1988). They show that observed optical and near-infrared (JHK) color gradients are mutually consistent (i.e. one can be derived from the other using the separate optical and infrared color-luminosity relations). All this suggests that the same change in stellar population produces both the color-luminosity relation and the color gradients. Using the new Yale isochrones (Green et al. 1987), Peletier and coworkers conclude that most color variations and gradients are due to changes in metallicity. In typical ellipticals, these do not exceed a factor of 10 inside re. However, age gradients may be present as well; the fraction of young stars may increase at larger radii.

Large data sets are needed to investigate correlations of color gradients with other galaxy properties. The measurements are difficult because color gradients are weak and because differential magnitude measurements are sensitive to systematic errors. Nevertheless, important data for early-type galaxies have been obtained by Jedrzejewski (1987b), Vigroux et al. (1988), Franx (1988), Franx et al. (1988), and Peletier and collaborators (see above).

Vader et al. (1988) analyze data from Vigroux et al. (1988) and obtain several interesting results. Whereas inward reddening is the rule in elliptical galaxies and bulges, they find that dSph galaxies tend to become bluer toward the center. Particularly interesting is the observation that color gradients are correlated with the rotation parameter (V / sigma)*: Anisotropic, pressure-supported ellipticals have smaller color gradients. We find the same effect, although with more scatter, in the Franx and Peletier et al. data (Figure 4).

The bright ellipticals in the Franx (1988) sample show weak correlations of color gradients with luminosity, velocity dispersion, integrated color, and Mg2 index: Weaker gradients are seen in brighter, hotter, redder, and more metal-rich galaxies. Gorgas & Efstathiou (1987) also find a marginally significant anticorrelation between Mg2 gradients and velocity dispersion. However, Peletier et al. (1988a) find no significant correlations with the above quantities; When we combine the Vader et al., Franx, and Peletier et al. samples (Figure 4), color gradients in E and SO galaxies are weak or absent at low luminosities (MB > - 20) and largest near the peak of the luminosity function (MB ~ - 20). The scatter exceeds the measurement errors at all luminosities.

We also find a marginal correlation of color gradients with isophote shapes (Figure 4). Color gradients in boxy ellipticals get smaller as boxyness increases [i.e. as a(4) / a decreases further below 0]. There are not enough data to look for variations of color gradients in disky ellipticals.

Figure 4

Figure 4. Correlations of color gradients with other galaxy properties. The gradients are defined as Delta(Color) / Delta(log r), in magnitudes per decade in radius; positive values indicate reddening toward the center. The data are from Vader et al. (1988; circles), Franx (1988; squares), and Peletier et al. (1988a; triangles). The left panel shows the dependence of color gradients on luminosity for dSph galaxies (open circles) and for ellipticals and SOs (solid symbols). Ellipticals become redder toward the center, but most dSph galaxies have inverse gradients. The top-right panel shows the relation between color gradients and the level of rotational support; (V / sigma)* = 1 for an isotropic oblate rotator. Anisotropic galaxies tend to have smaller gradients. The bottom-right panel shows the correlation with isophote shape (measured by B + 88). More boxy galaxies [a(4) / a < 0] tend to have smaller color gradients. Similar trends are obtained using (U - R) color measurements.

These results are very preliminary. However, the trends are probably real: All of the parameters are measured independently, and measurement errors only diminish the correlations. If confirmed, these correlations will provide important new information about galaxy formation.

The absence of a strong correlation between the strengths of color gradients and L or sigma is contrary to the predictions of the Larson and Carlberg models. However, these do not include the effects of postcollapse mergers, which are important at least for bright ellipticals. The observations suggest that the properties of early-type galaxies are determined by dissipative collapse and then modified by mergers (Vader et al. 1988). Dissipative formation could produce a mass-metallicity relation and metallicity gradients, which are then gradually erased by mergers. The trend with luminosity at MB gtapprox - 21 would then be a fossil of the initial correlation predicted by Larson and Carlberg. Mergers may also produce correlated changes in (V / sigma)* and a(4) / a (cf. Figure 3). Normal color gradients are not produced in diffuse dwarfs because of galactic winds (Vader 1986a, Dekel & Silk 1986); their inverse color gradients could be due to recent star formation.

Finally, we discuss color gradients in cooling-flow galaxies. The existence of cooling flows in ellipticals and cluster cores is reasonably well established (e.g. Fabian 1988). Mass flow rates are uncertain, but Mdot ~ 1 - 1000 Msun yr-1 are believed to be deposited inside a few re, with a strong dependence on radius. The most plausible fate of the gas is star formation. Unless the initial mass function (IMF) is strongly biased toward low-mass stars, observable color gradients should result (Sarazin & O'Connell 1983, Silk et al. 1986, O'Connell 1988). Blueing toward the centers of some high-Mdot cooling-flow galaxies is seen and interpreted as evidence for age gradients (e.g. Wirth et al. 1983, Romanishin 1986a, 1987, Maccagni et al. 1988). However, the extreme gradients reported by Valentijn (1983) and Valentijn & Moorwood (1985) are not confirmed by subsequent work. A detailed comparison of color gradients in normal and cooling-flow galaxies could provide strong constraints on the fate of gas in cooling flows.

Next Contents Previous