As we discussed in Section 1, one of the most exciting new datasets that arrived in the mid-1990's was the first set of resolved images of galaxies at significant look-back times from HST. Much of the early work was conducted in rich clusters ([Couch et al 1994, Dressler et al 1994, Couch et al 1998, Dressler et al 1998]) where the well-known `Butcher-Oemler' effect ([Butcher & Oemler 1978]) - a surprisingly recent increase in the fraction of blue cluster members - was found to be due to a dramatic shift in the morphology-density relation (Figure 8). As recently as z 0.3-0.4 (3-4 Gyr ago), cluster S0s were noticeably fewer in proportion, their place apparently taken by spirals, many of which showed signs of recent disturbances, such a distorted arms and tidal tails.
Figure 8. Evolution in the morphology density relation from the `Morphs' survey of Dressler et al (1999). (Left) The fraction of E/S0/Spirals as a function of projected galaxian surface density for Dressler's 55 local cluster sample. (Right) As left, for all distant HST clusters with 0.3 < z < 0.55. The comparison refers to the same cluster core radius (< 0.6 Mpc, h = 0.5) and includes galaxies to the same rest-frame V luminosity (MV = -20.0). Note the dramatic decline in the S0 population and the marked increase in the spiral fraction for environments with high projected density.
The physical origin of this transformation from spirals to S0s remains unclear and is currently being explored by detailed spectroscopy of representative cluster members ([Barger et al 1996, Abraham et al 1996b, Poggianti et al 1999]). A key diagnostic here is the interplay between the changing morphologies, the presence of nebular emission lines (such as [O II] 3727 Å - H is generally redshifted out of the accessible range) and Balmer absorption lines (such as H 4101 Å ). The latter lines are prominent in main sequence A stars which linger for 1 Gyr after any enhanced starburst activity.
Barger et al (1996) proposed a simple cycle where an unsuspecting galaxy undergoes some perturbation, perhaps due to a merger or its first encounter with the intracluster gas, subsequently becomes morphologically-distorted and spectrally-active before subsiding to a regular spheroidal with a decaying Balmer absorption line. Whereas such a cycle can explain the proportion of unusual objects, it has difficulty matching their luminosities. A galaxy should be rendered more luminous during a burst and thus blue examples cannot easily be the precursors of the equally-luminous red post-burst cases. A controversy has since arisen over the fractions of objects seen in the various spectrally-active classes ([Balogh et al 1999]) suggesting much work is needed in this area, both in quantifying cluster-cluster variations and also radial variations in the responsible processes.
Although the cluster work discussed above represents something of a digression in our overall theme, the realisation that galaxies can so easily be transformed morphologically has profound implications for our understanding of galaxy formation. Much of the early work explaining the Hubble sequence ([Tinsley 1977]) assumed galaxies evolve as isolated systems, however the abundance of morphologically-peculiar and interacting galaxies in early HST images ([Griffiths et al 1994]) has been used to emphasize the important role that galaxy mergers must play in shaping the present Hubble sequence ([Toomre & Toomre 1972, Barnes & Hernquist 1992]). Merger-induced transformations of this kind are a natural consequence of hierarchical models ([Baugh, Cole & Frenk 1996]). Early disk systems are prone to merge during epochs when the cosmic density is high and the peculiar velocity field is cold, forming bulge-dominated and spheroidal systems which may then later accrete disks.
The possibility that galaxies transform from one class to another is a hard hypothesis to verify observationally since, as we have seen, traditionally observers have searched for redshift-dependent trends with subsets of the population chosen via an observed property (color, morphology, spectral characteristics) which could be transient. Moreover, experience ought to teach us that the outcome of tests of galaxy formation rarely come down simply to either Theory A or Theory B; usually it is some complicated mixture or the question was naive in the first place! Fortunately, the late formation of massive regular galaxies in the hierarchical picture (Figure 9) seems a particularly robust prediction and one in stark contrast to the classical `monolithic collapse' picture ([Tinsley 1977, Sandage 1983]). The distinction is greatest for ellipticals presumed to form at high redshift with minimum dissipation (their central density reflecting that of the epoch of formation). Studying the evolutionary history of massive ellipticals is thus an obvious place to start.
Figure 9. The important role of late merging in a typical CDM semi-analytical model (Baugh et al 1996). The panels show the redshift-dependent growth for the stellar mass (left) and that of all baryonic material (right), as indicated by the thickness of the black area at a given epoch, for two present-day massive galaxies. The top system grows gradually and is thought to represent a present-day spiral. The bottom system suffers a late equal-mass merger thought to produce a present-day elliptical. Note the remarkably late assembly; most of stellar mass in both cases assembles in the interval 0 < z < 1.
An oft-quoted result in support of old ellipticals is the remarkable homogeneity of their optical colors ([Sandage & Visvanathan 1978, Bower et al 1992]). The idea is simple: the intrinsic population scatter in a color sensitive to recent star formation, such as U - B, places a constraint either on how synchronous the previous star formation history must have been across the population or, if galaxies form independently, the mean age of their stellar populations. By combining cluster data at low redshift ([Bower et al 1992]) with HST-selected samples at intermediate redshift ([Ellis et al 1998]), the bulk of the cluster elliptical population was deduced to have formed its stars before z 2, in apparent conflict with hierarchical models. Similar conclusions have been drawn from evolution of the mass/light ratio deduced from the fundamental plane ([Ziegler & Bender 1997, van Dokkum et al 1998]).
Unfortunately, one cannot generalize from the results found in distant clusters. In hierarchical models, clusters represent early peaks in the density fluctuations and thus evolution is likely accelerated in these environments ([Kauffmann 1995]) plus, of course, there may be processes peculiar to these environments involving the intracluster gas. It is also important to distinguish between the history of mass assembly and that of the stars. Recent evidence for widespread merging of ellipticals in clusters ([van Dokkum et al 1999]) lends support to the idea that the stars in dense regions were formed at high redshift, in lower mass systems which later merged.
For these reasons, attention has recently switched to tracking the evolution of field ellipticals. The term field elliptical is something of a misnomer here since a high fraction of ellipticals actually reside in clusters. What is really meant in this case is that we prefer to select ellipticals systematically in flux-limited samples rather than concentrate on those found in the cores of dense clusters (3).
The study of evolution in field ellipticals is currently very active and I cannot possibly do justice, in the space available, to the many complex issues being discussed. Instead let me summarise what I think are the most interesting results.
Figure 10. Evidence for a clustered population of red objects in a 1000 arcmin2 area of the ongoing Las Campanas Infrared Survey ([McCarthy et al 2000]). The angular correlation function for objects with I - H > 3.5 (top set of data points) is substantially above that for all H-selected galaxies (bottom set) and consistent with a high fraction of the red objects being clustered ellipticals with 1.0-1.5.
Figure 11. Color inhomogeneities in HDF field ellipticals suggest continued star formation is occurring, possibly as a result of hierarchical assembly. Each row displays the I band HST image, a V - I color image and the Keck LRIS spectrum. The top set refers to a z = 0.92 elliptical with a blue core; its spectrum shows features indicative of active star formation ([O II] emission and deep Balmer absorption lines). The bottom set refers to a quiescent example at z = 0.966 whose spectrum is consistent with an old stellar population. The amount of blue light can be combined with the depth of the spectral features to statistically estimate the amount and timescale of recent star formation.
What evolution is found in the properties of other kinds of galaxy? Brinchmann et al (1998) secured HST images for a sizeable and statistically-complete subset of CFRS and LDSS redshift survey galaxies and found the abundance of spirals to I = 22 - a flux limit which samples 0.3 < z < 0.8 - is comparable to that expected on the basis of their local abundance if their disks were somewhat brighter and bluer in the past as evidenced from surface photometry ([Lilly et al 1999]). In practice, however, the detectability of spiral disks is affected by a number of possible selection effects ([Simard et al 1999, Bouwens & Silk 2000]) and it may be some time before a self-consistent picture emerges.
A less controversial result from Brinchmann et al (1998) claimed in earlier analyses without redshift data ([Glazebrook et al 1995, Driver et al 1995]) is the remarkably high abundance of morphologically-peculiar galaxies in faint HST data. Brinchmann et al quantified this in terms of the luminosity density arguing that a substantial fraction of the claimed decline in the blue luminosity density since z 1 (c.f. Figure 6) arises from the demise of this population (Figure 12).
Figure 12. The morphological dependence of the blue luminosity density from that subset of the CFRS/LDSS redshift survey imaged with HST (Brinchmann et al 1998). The marked decline in the luminosity density of galaxies with peculiar morphology over 0 < z < 1 is the primary cause for steep slope in the blue faint galaxy counts.
Given our earlier concerns with over-interpreting the cosmic star formation history, should we be cautious in drawing conclusions from Figure 12? Although Brinchmann et al's redshift sample is small, the basic result is consistent with the HST morphological number counts where much larger samples are involved. Whereas early skeptics argued that morphologically-peculiar galaxies represent regular systems viewed at unfamiliar ultraviolet wavelengths, recent NICMOS imaging (see Dickinson's lectures) suggests such `morphological bandshifting' is only of minor consequence. In quantitative detail, as before, uncertain corrections must be made for the effects of the flux limited sample and of course extinction is a major uncertainty. However, it seems inescapable that the bulk of the decline in blue light (the so-called faint blue galaxy problem, [Ellis 1997]) arises from the demise of a population of late-type and morphologically-peculiar systems. A key question therefore is what happened to this population? We will address this problem in the next section.
3 A major concern in all the work relating to the evolution of galaxies in clusters is precisely how the clusters were located. Back.