Annu. Rev. Astron. Astrophys. 1991. 29: 543-79 Copyright © 1991 by Annual Reviews. All rights reserved

6. COLOR GRADIENTS AND METALLICITIES

The chemical compositions, and composition gradients, within the halos of galaxies trace out their early enrichment histories. The observable record (though not always an unambiguous one) lies in the integrated color indices and spectral features of the halo light and of the light from globular clusters. This type of approach to GCS properties was launched with the landmark papers of Strom et al. (192) and Forte et al. (64) on four Virgo galaxies including M87. Their UBR photographic photometry of these systems indicated both that the clusters were bluer on the average than the galaxy itself at any location and that the GCSs exhibited color gradients (bluer mean color at larger galactocentric radius R_gc). These results were interpreted to mean that the clusters had systematically lower metallicity than the field stars and became more metal-poor with increasing R_gc. In turn, such results supported the idea that the clusters were distinctly older than the bulk of the halo, and that the halo formed by collapse with dissipation and progressive metal enrichment. A calculation by De Young et al. (51) suggested that the clusters, by forming first, could have seeded the enrichment of the halo gas which only later condensed into stars, thus creating the observed metallicity offset between clusters and halo.

Subsequent observational work strengthened the first of these two results. Mean color differences at the same level, namely (B - V)(halo - GCS) 0.2-0.3 mag, also held in the dwarf ellipticals (92), and later CCD multicolor photometry for GCSs in several giant ellipticals has verified both the sense and the amount of the effect (18, 38, 39, 40, 88). The interpretation of color difference as primarily, or entirely, a metallicity difference was at first more of a reasonable hypothesis than a necessity. [See, for example, the understandably skeptical view by Burstein (24).] However, spectroscopic indices obtained for clusters in both dwarf and giant galaxies (21, 45, 81, 119, 151, 152) now more directly support this interpretation. The mean spectroscopic cluster metallicities equal those predicted from the integrated colors (e.g. 39). In addition, they show that the clusters are generally more metal-poor by [Fe/H] -0.5 than the spheroid stars at the same R_gc - just the right amount to produce the color differential. The same result has now been extended to include the Fornax dwarf (43, 55, 140, 176), M31 (20, 54), and the Milky Way spheroid itself (226). An important implication of the halo/GCS metallicity offset is that the spheroids of galaxies did not build up principally from dissolved globular clusters, but rather belong to a distinct stellar population with a different chemical enrichment history.

An alternate interpretation, though with much less observational backing, is to make the clusters bluer at the same metallicity by requiring them to be younger than the halo. Models of integrated cluster color against age (e.g. 19, 33, 37) show that enormous age differences ( ~ 5-10) Gy are needed to generate integrated color differences as large as (B - V) ~ 0.2-0.3 mag in an old stellar population. One piece of more direct evidence which might favor lower age is the presence of line-strength anomalies found in some M31 halo clusters (20, 25, 198), though other interpretations involving composition or stellar population variations are possible. The ultraviolet fluxes of the M31 clusters do not appear to be unusual (42), and spectral indices from a much larger new sample (21) suggest a more normal picture except for apparent nitrogen enhancements that may be primordial in origin.

The second principal suggestion of the Strom et al. work - that of metallicity gradients in the GCSs - has since been thrown into considerable confusion. Later CCD photometry in both the BVI and gri bands (38, 39) showed that the clusters in M87 exhibit no trace of a color gradient, and that the Strom et al. result was probably a residual artifact of photographic calibration across a nonuniform background. In NGC 5128 also, little net change in mean cluster metallicity with R_gc is evident (85, 88). However, recent BV photometry in three other large ellipticals, [NGC 1399, 4649, and M49 (18, 40)] has revealed GCS color gradients amounting to (B - V) 0.1 mag in all three over R_gc from 3 to 10 kpc. Figure 7 presents these schematically. Also, in both the Milky Way (101, 225) and M31 (14, 20, 54, 188, 189), slight gradients in mean cluster metallicity seem to exist throughout the halo even after the inner metal-rich disk globulars are excluded, suggesting that some small amount of systematic dissipation and enrichment took place during formation. In summary, the metallicity gradients in GCSs do not seem to correlate unambiguously with any other obvious systemic features. One might speculate that their presence or absence may simply be a residual of the whole set of events that occurred during galaxy formation (dissipative collapse, chaotic merger of gaseous fragments, later mergers or accretion of formed galaxies), which must have had different relative importances from place to place (123).

Figure 7. Color gradients in the GCSs of elliptical galaxies. The mean (B - V) color of the globular clusters around each of four giant E galaxies is plotted in radial bins. Error bars on each mean point are typically (B - V) ~ ± 0.03 mag. In M87 (N4486), no significant change in mean cluster color (metallicity) with galactocentric distance Rgc is seen. In the other three, a systematic decrease of (B - V) 0.1 mag over r ~ 10 kpc is seen, corresponding to a decrease of about -0.5 dex in [Fe/H]. Note that this graph is indicative of only relative changes in color within each system; the offsets between the different galaxies are not significant because of zero-point uncertainties in the photometry, which can amount to ± 0.05 mag per field (see the references cited in the text).

A major result now well established is that both the mean cluster metallicity and the cluster-to-cluster scatter in metallicity increase with the host galaxy size. The dEs hold uniformly metal-poor clusters (45). In the Milky Way (5, 225, 226) and M31 (20, 54), the halo clusters fall generally in the range -2 [Fe/H] -1, but a more metal-rich subgroup appears within R_gc 4 kpc. In the giant ellipticals, the systemic mean is consistently near [Fe/H] -1, but the total range is very large, with clusters extending up to solar metallicity and perhaps higher (71, 85, 151, 152). Figure 8 illustrates these trends. The correlation of mean [Fe/H] with M_V^T(galaxy) has a slope of 0.15, corresponding to a scaling of mean heavy-element abundance Z with galaxy luminosity as Z ~ L^0.4. Furthermore, it is parallel to the trend for the metallicities of the galaxies themselves, but simply offset to a lower [Fe/H], suggesting that the enrichment processes were similar in both types of old stellar populations. The clear implication is that the enrichment yield during cluster formation went much farther toward completion in the giant galaxies (see 148a).

Figure 8. Metallicity distributions for globular cluster systems. In the left panel, the mean [Fe/H] for the globular clusters in the galaxy (from Table 2) is plotted against galaxy luminosity MVT; filled symbols denote elliptical galaxies, and open symbols are spirals or irregulars. The solid line is the mean [Fe/H] vs. luminosity relation for the galaxies themselves (cf. 21); the GCSs fall along much the same slope, but displaced ~ 0.5 dex downward. If the spiral galaxies (open circles) were to be plotted by only their spheroidal luminosities excluding the disk, they would lie ~1-2 magnitudes further to the left and would fall much closer to the mean relation defined by the E galaxies. In the right panel, histograms of cluster metallicities are shown for a giant elliptical, NGC 5128; the Milky Way; and the four dwarf E galaxies in the Local Group. In the larger galaxies, the globular clusters are distributed more widely in composition and have a higher mean metallicity. For NGC 5128, the relative number of high-[Fe/H] clusters may in fact be underestimated, since no clusters from its central regions are included (cf. 85).

These observations are consistent with a picture in which protoclusters formed within dwarflike units with their own pre-enrichment histories (e.g. 148a, 186), or else over a time period long enough for the surrounding halo gas to have enriched significantly while clusters were condensing out. New direct evidence for an age range of perhaps ~ 2 Gy among the Milky Way halo globulars (e.g. 13, 130, 200) may also support such interpretations. In addition, the near-constancy of mean cluster luminosity in all types of galaxies (Figure 1) and the strong dependence of their metallicity on galaxy size (Figure 8) imply in tandem that the process of cluster formation ran almost independently of composition.

Another observational constraint on a smaller scale is that globular clusters are individually quite homogeneous in composition (with outstanding exceptions such as Cen), whereas in big galaxies we find a large cluster-to-cluster spread in metallicity. Whether or not a protoglobular gas cloud can self-enrich while still remaining bound and homogeneous then becomes an important issue for theoretical modeling (126, 127, 148, 153, 154, 155). Whatever pre-enrichment process went on to produce higher yield in the more massive galaxies, the individual protoglobular clouds seem to have remained well mixed while still gaseous.

The metallicity distributions of clusters provide additional evidence that the giant galaxies did not simply accumulate by mergers of already-formed smaller ones. However, the idea of mergers of disk systems in which globulars were rarer to begin with is an attractive hypothesis (Section 4) for explaining the very low specific frequencies in some field ellipticals. The clusters in such galaxies should have lower mean metallicities than those in high-S_N ellipticals of the same size, and direct measurements of their abundances should provide an interesting test.