The key to detecting observable evidence for galaxy evolution, that is, to actually see morphological differences that are likely attributable to evolution, is to observe galaxies at high redshift with sufficient resolution to reveal significant details of morphology. Butcher & Oemler (1978) had already found strong evidence for morphological evolution in the excess number of blue galaxies in very distant (z = 0.4-0.5) rich galaxy clusters. These authors suggested that the blue galaxies are spiral galaxies, and that by the present epoch, these are the galaxies that become the S0s that dominate nearby rich, relaxed clusters (like the Coma Cluster, Abell 1656). An excellent summary of the issues connected with high redshift morphological studies is provided by van den Bergh (1998).
Progress on galaxy morphology at high redshift could only be achieved with the resolution and depth of the Hubble Space Telescope. The Hubble Deep Field North (HDF-N, Williams et al. 1996), South (HDFS, Volonteri et al. 2000), and Ultra-Deep Field (HUDF, Beckwith et al. 2006), and the GOODS (Great Observatories Origins Deep Survey; Giavalisco et al. 2004), GEMS (Galaxy Evolution from Morphology and SEDS; Rix et al. 2004), COSMOS (Cosmic Evolution Survey; Scoville et al. 2007), and other surveys (e.g., Cowie et al. 1995), have provided a large body of information to work with. For example, studies of galaxies in the redshift range 0.3 z 0.9 show that the proportion of irregular-shaped galaxies dramatically increases (e.g., Abraham et al. 1996). This means that the Hubble sequence as we know it did not always exist but was built up over time via mergers or secular evolution or both. Observations of submillimeter sources (Chapman et al. 2003) suggest some of these irregulars are extended major mergers.
Interpretation of evolution in the various deep surveys depends on knowledge of redshifts, which can be difficult to measure spectroscopically. A very useful technique for isolating galaxies in high redshift ranges is the UV-drop out method (Steidel & Hamilton 1992). Galaxies are compared in different filters, such as B435, V606, i775, and z850. If the redshift is high enough to move the Lyman limit at 0.0912µm out of any of the first three filters, there will be a significant drop in flux owing to absorption by hydrogen, and the galaxy is said to "drop out." Galaxies found in this way are called "Lyman break" galaxies because the effect is partly caused by the spectral characteristics of hot stars, which show such a break. (Another cause is UV self-absorption.)
Steidel et al. (1996) used the UV drop-out approach to identify high z galaxies in the HDF-N, and also used direct spectroscopy to confirm that the method works. Beckwith et al. (2006) utilized the method to identify galaxies in the HUDF having z from 3.5 to 7. If a galaxy is seen to drop out of a U-band filter such as F300W and seen in a B-band filter such as F450W (both used for the HDF-N; Williams et al. 1996), then the redshift range selected is z = 2.4 to 3.4. van den Bergh (1998) argues that most Lyman break galaxies are young ellipticals or bulges.
Van den Bergh et al. (2000) describe the issues connected with morphological classifications of galaxies to redshifts of z1. Resolution, band-shifting, and selection effects due to the magnitude-limited nature of surveys all enter into the interpretation of intermediate to high redshift galaxy morphology. Resolution is important because, typically, a nearby galaxy will have 100 times or more pixels in the image than a high z galaxy will have for classification. This is fewer pixels than sky survey images of nearby galaxies would have. The bandpass effect is important because the B-band, the standard wavelength for historical galaxy classification, is not sampling the same part of the spectrum as it would for nearby galaxies. For example, at z = 1, a B-band filter samples mid-ultraviolet light ( 0.22µm) and would be much more sensitive to young star-forming regions than it would be for nearby galaxies. Ideally, then, for comparison with nearby galaxies we would like to choose a redshifted band as close as possible to the rest-frame B-band. Even accounting for all these effects, significant differences between nearby and distant morphologies do exist. For example, van den Bergh et al. (2000) discuss the paucity of grand-design spirals and barred galaxies in the HDF-N, and use artifically redshifted images of nearby galaxies to demonstrate that the deficiencies are likely to be real.
Figure 46 shows several of the different categories of intermediate and high redshift galaxy morphologies, based on V and i-band images from the GEMS and Hubble UDF. The redshifts range from 0.42 to 3.35 and provide a wide range of look-back times. First, in such a range, some galaxies look relatively normal, as shown by the spiral and elliptical galaxies in the two upper left frames of Figure 46. The z = 0.59 spiral is classifiable as type SA(s)bc and the elliptical as type E3. The z = 0.99 spiral shown in the middle right frame of Figure 46 has larger clumps, no clear central object, and more asymmetry than the z = 0.59 spiral, but is still recognizable as a spiral. However, other less familiar categories are found. In general, high redshift galaxies are smaller than nearby galaxies on average (e.g., Elmegreen et al. 2007a = EEFM07).
"Chain galaxies" were first identified by Cowie et al. (1995) and are linear structures with superposed bright knots that have sizes and blue colors similar to normal late-type galaxies and relatively flat major axis luminosity profiles. They have the shapes of edge-on disk galaxies but lack clear bulges or nuclei. A recent study by Elmegreen, Elmegreen, & Sheets (2004 = EES04) of faint galaxy morphologies (redshifts 0.5-2) in the Advanced Camera for Surveys (ACS) "Tadpole" galaxy (UGC 10214) field showed that chain galaxies are the most common linear morphology at magnitudes fainter than I = 22, accounting for more than 40% of the sample. Their dominance (also found by Cowie et al. 1995) is interpreted by EES04 as a selection effect because relatively optically thin edge-on galaxies are more favored to be seen near the limit owing to a higher projected surface brightness than for face-on versions of the same galaxies. EES04 suggest that chain galaxies are edge-on irregular galaxies that will evolve to late-type disk galaxies. Chains are the most flattened linear morphology at faint magnitudes.
"Clump clusters" (EES04) are somewhat irregular collections of blue knots or clumps with very faint emission between clumps. The clumps have sizes of 500 pc and masses of 108-109 M. Both of the examples shown in Figure 46 have z > 1. Elmegreen, Elmegreen, & Hirst (2004 = EEH04) identified clump clusters as the face-on counterparts of the linear chain galaxies, based on the similarities of the properties of the clumps with those seen in chain galaxies, and on the distribution of axis ratios of the systems as compared with normal disk galaxies. The lack of a bulge clump is also consistent with this conclusion. Nevertheless, analysis of NICMOS IR images in the HUDF led Elmegreen et al. (2009a) to conclude that 30% of clump clusters and 50% of chain galaxies show evidence of young bulges, implying that at least half of these galaxies might be genuinely bulgeless. In a related study, Elmegreen et al. (2009b) show that the best local analogues of clump clusters are dwarf irregular galaxies like Ho II, scaled up by a factor of 10-100 in mass. This study also brought attention to clump clusters with faint red background disks, as opposed to blue clump clusters which lack such a feature. Elmegreen et al. argue that the red background clump clusters are part of an evolutionary sequence leading from the blue clump clusters to spirals with a "classical" bulge (e.g., KK04). The clumps, formed by gravitational instabilities in a turbulent disk, are large and few in number, and thus will eventually coalesce near the galaxy center if they survive the effects of supernova explosions (e. g., Elmegreen, Bournaud, & Elmegreen 2008).
"Tadpoles" (van den Bergh et al. 1996) are asymmetrically-shaped "head-tail" morphologies with a bright off-centered nucleus and a tail, like a tadpole. A rare local example is NGC 3991. Tadpoles were recognized in 3% of the galaxies in HDF-N and were found to be very blue in color. Usually both the head and the tail are blue, but van den Bergh et al. (1996) show one example where the head is red and the tail is blue. EEH04 showed that tadpoles have neither exponential major axis profiles nor clear bulges, and in their sample of linear objects, tadpoles are the least frequently seen.
The bottom frames of Figure 46 show bent chains and rings or partial rings (Elmegreen & Elmegreen 2006 = EE06). The rings and partial rings are thought to be mostly collisional in nature (i.e., like the conventional ring galaxies shown in Figure 27) and show the different morphologies expected when small companions plunge through a larger disk galaxy in different ways (Appleton & Struck-Marcell 1996). Although bent chains resemble the partial rings, they lack offset nuclei and any evidence of a background, more face-on disk. EE06 suggest that bent chains are simply warped versions of the more common linear chains that have suffered an interaction. The ages of the bent chain clumps are younger than those found in rings and partial rings, and EE06 argue that relative separations and sizes of the clumps indicate they form by gravitational instabilities.
Elmegreen et al. (2005) show that approximately 1/3 of the ellipticals catalogued in the HUDF have prominent blue clumps in their centers (see also Menanteau et al. 2001, 2004). They argued that these clumps probably imply accretion events based on comparison of their magnitudes and colors with local field objects. Menanteau et al. (2001) were able to reproduce the color distributions with a model having a starburst superposed on a pre-existing older stellar population.
Galaxy morphology at intermediate and high redshifts also includes obvious interacting cases as well as possible merger morphologies. Bridges, tidal tails, plumes, and even M51 analogues are seen as in nearby galaxies, but are smaller in scale than for nearby objects (EEFM07). The middle frame of the bottom row shows a possible merger in progress of two bent chains (or, alternatively, two interacting spirals), called an "assembly galaxy" by EEFM07 because they appear to be assembling from smaller objects. EEH04 and EEFM07 also discuss the double systems, considered another category of the linear systems. The double systems like the z = 3.35 one shown in Figure 46 are probably merging ellipticals. EEFM07 also describe "shrimp galaxies", which appear to be interacting galaxies with a single curved arm or tail, curling at one end into a "body."
Other studies of high redshift galaxy morphology have focussed on the specific redshift ranges that are selected by the UV drop-out technique. Conselice & Arnold (2009) examined the morphologies of galaxies in the z = 4-6 range from the HUDF, and measured quantitatve parameters such as the concentration-asymmetry-clumpiness (star formation) parameters (CAS; Conselice 2003) and other related parameters that are useful for distinguishing mergers from normal galaxies. The CAS system is based on simple global parameters that are easily derived automatically for large numbers of galaxies. Conselice (2003) tied the C parameter to the past evolutionary history of galaxies while parameters A and S measure more active evolution from mergers and star formation. Conselice & Arnold found that half of the HUDF drop-out galaxies they studied have significant asymmetries and may be undergoing merging, while the other half is mainly smooth symmetric systems that may have collapsed quickly into a temporary, quiescent state.
Other quantitative approaches to these issues include the Sersic n index that characterizes radial luminosity profiles (Ravindranath et al. 2006; Elmegreen et al. 2007b) and the Gini coefficient (Abraham et al. 2003; Lotz et al. 2006). The Gini coefficient provides a way of quantifying high redshift morphology that does not depend on galaxy shape or the existence of a well-defined center, and is well-suited to the kinds of objects shown in Figure 46. Lotz et al. (2006) found in a sample of 82 Lyman break galaxies that 10-25% are likely mergers, 30% are relatively undisturbed spheroids, and the remainder are disks, minor mergers, or post-mergers.
Given the rise in peculiar and irregular-shaped galaxies with increasing
redshift, the question naturally arises:
when did the Hubble sequence and all its accompanying details fall into
place? This question is considered by
Conselice et al. (2004),
who quantitatively analyzed a well-defined high
redshift sample using the CAS system. Conselice et al. identify
"luminous diffuse objects" (LDOs) as galaxies having C less than