Annu. Rev. Astron. Astrophys. 1984. 22: 185-222 Copyright © 1984 by Annual Reviews. All rights reserved

5.1 Galaxy Color as a Function of Look-Back Time

Butcher & Oemler (1978a) pioneered the study of the color evolution of average luminosity (L) galaxies by obtaining V and r images with an ISIT vidicon of two distant clusters: C10024 + 1654 (z = 0.39; Humason & Sandage 1957) and the cluster containing the luminous radio galaxy 3C 295 (z = 0.46; Minkowski 1960). These clusters are rich and centrally concentrated like the Coma cluster. However, Butcher & Oemler found that unlike Coma, whose population is ~ 95% red E and S0 galaxies, these distant clusters contained many galaxies with V - r colors (a color between U - B and B - V in the rest frame at z = 0.4) bluer than S0s or ellipticals. Specifically, in a square box about 1 Mpc on a side, only a few percent of the galaxies in the Coma cluster are significantly bluer in B - V than the average elliptical or S0 galaxy, but in a comparable area of the distant clusters, the fraction ranges from 30% for the 3C 295 cluster to 50% for C10024+1654.

From this comparison of the distribution of galaxy colors in two distant clusters with that of the nearby Coma cluster, Butcher & Oemler concluded that a very striking evolution in the galaxy content of rich clusters had occurred in only the last one third of a Hubble time. This ``excess of blue galaxies'' in distant clusters compared with nearby clusters is commonly referred to as ``the Butcher-Oemler effect.''

Butcher & Oemler, along with others in the field, were understandably surprised that galaxy evolution appeared to be so dramatic over such a small fraction of the age of the Universe. Rapid evolution of normal galaxies had been considered unlikely, perhaps because colors and luminosities of early-type galaxies for z 1 are consistent with an initial, single burst of star formation lasting ~ 10⁹ yr (see, for example, Bruzual 1981, and references therein). In fact, there is no convincing evidence for galaxy evolution from spectrophotometry of either field or cluster galaxies (see Kron (1982) and Ellis (1983) for reviews]. On the other hand, finding a large population of blue galaxies in clusters at z ~ 0.5 would imply that many of the early-type galaxies seen in present-epoch clusters, especially S0s, were actively forming stars until quite recently, but that over a relatively short period of time this activity has ceased.

It is not surprising, then, that the Butcher-Oemler effect has been greeted with a degree of skepticism and caution. DeGioia-Eastwood & Grasdalen (1980) suggested that the range of ultraviolet colors of early-type galaxies, the difference in K-corrections for red and blue galaxies, and the rather large photometric errors associated with these very faint objects all might conspire to produce a spread in V -r color at z 0.4 as large as that seen by Butcher & Oemler. Koo (1981) published photometry of C10016 + 16, an even more distant cluster (z = 0.54) that appeared to have no excess of blue galaxies. Mathieu & Spinrad (1981) obtained new photometry from photographic plates of the 3C 295 cluster and concluded that the contamination of the cluster counts by field galaxies was underestimated by Butcher & Oemler. This was later verified by Dressler & Gunn (1982), who obtained spectra for a sample of the blue galaxies in the Butcher-Oemler field and found 11 out of 20 to be either foreground or background galaxies.

On the other hand, subsequent studies have confirmed the original observations of Butcher & Oemler for larger samples. Couch & Newell (1984a, b; for an initial discussion, see Couch 1981) obtained photographic photometry in J and F for 14 rich clusters in the range 0.18 < z < 0.39 and concluded that the clusters with z > 0.26 have 2-5 times as many blue galaxies as do nearby clusters of similar type. A recent compilation by Butcher & Oemler (1984a), which includes data for 12 nearby clusters (z < 0.1), 16 intermediate-distance clusters (0.1 < z < 0.3; data primarily from Butcher et al. 1983), and 5 distant clusters, shows the blue fraction rising continuously for z > 0.1, although there is considerable scatter at any epoch. In this latest study, Butcher & Oemler define the blue fraction f_B as those galaxies that are (a) of magnitude M_v < -20, (b) within the radius that contains 30% of the cluster galaxies, and (c) bluer by more than 0.^m2 in (rest-frame) B - V than an elliptical galaxy of the same absolute magnitude. According to this definition, the nearby clusters in their sample with high central concentrations have f_B < 0.05, and clusters in the intermediate distance range have 0.00 < f_B <0.20, with a median of about 0.10. The four distant clusters have f_B values of 0.02 (C10016+ 16), 0.16, 0.21, and 0.22, although one distant irregular cluster has a value near 0.37. (These percentages of blue galaxies are lower by a factor of 1.5-2.0 than the numbers quoted in earlier studies, reflecting a change in the accounting system discussed in the next section.) The scatter from any mean line f_B z is so large that it is just as consistent at this time to interpret the relationship as an envelope diagram where the upper limit to the blue galaxy fraction rises with redshift, although Butcher & Oemler argue that most of the scatter from a mean line is accounted for by sampling statistics.

It seems, then, that samples from several sources confirm the Butcher-Oemler effect for z < 0.5. (Koo's study of C10016+16 at z = 0.54 provides the most deviant point.) Is it safe, therefore, to conclude from photometry alone that significant evolution has occurred since z = 0.5 in the populations of rich clusters? Unfortunately, the answer is probably no. The difficulty lies in selection effects that may have been present in the identification of clusters at high redshift.

First and foremost, the problem of subtracting the superposed field galaxies, most of which are blue, is a serious one for these types of studies. The contrast of distant clusters over the field is not large, which means that an underestimation of the field contribution by a factor of ~ 2 could account for the entire Butcher-Oemler effect. The photometry referenced above is of high quality, so that errors this large in the field determination should not arise from improper treatment of the data. Furthermore, if the fields that contained the clusters were randomly selected, fluctuations this large in the field counts would be rare. The problem is that these fields were not randomly selected, but chosen because of their contrast against the field. This means that if, by chance, the field counts were relatively low around a distant cluster, or relatively high in the cluster region, the probability that the cluster would be noticed and included in a catalog might be greatly increased. Therefore, such a sample of distant clusters might include a disproportionate number of cases with large fluctuations in the field counts. From Shectman & Dressler's (1984) redshift study of 14 of the 55 clusters chosen by Dressler (1980a), it is clear that even in a sample of nearby clusters, where contrast with the field is high, ~ 30% of the clusters contain a significant contamination (of order one third) by a superposed cluster or group. Since these alignments should be less common than this, it follows that the increase in richness caused by the contamination has biased Dressler to include them in his catalog. This type of problem will be much more severe in distant samples, and although it is possible to estimate the expected fluctuations of the field counts and the probability of superposition, it is nearly impossible to determine how much the biases discussed above increase the chances of inclusion in a catalog of distant clusters.

The problem is also complicated by the fact that rich clusters are apparently embedded in superclusters, which increases the chance of cluster superposition and may influence the field counts. For example, a prolate supercluster with one or more embedded clusters might have a very high probability of inclusion in a catalog if it were viewed pole-on. This could result in the superposition of many blue galaxies with nearly the cluster redshift that are nevertheless quite distant from the primary cluster.

These are examples of how photometry alone provides suggestive, but not yet compelling, evidence that greater percentages of blue galaxies inhabited the central regions of rich, concentrated clusters only one third of a Hubble time ago. The definitive evidence must come from knowledge of the redshifts of the galaxies in question.