Observational evidence that interactions and mergers were more frequent in the past has trickled in since the late 1970s and has grown more rapidly since the late-1993 repair of the Hubble Space Telescope. In general this evidence agrees with expectations based on numerical simulations of hierarchical clustering in an expanding universe dominated by dissipationless dark matter. However, quantitative observations of high-z interactions remain difficult to obtain. As we study objects from z 0.3 to ~ 1.2 morphological details and kinematic signatures fade, and we are reduced to judging gross morphologies from a few pixels or simply counting galaxy pairs.
Quasars yielded some of the earliest evidence for interactions at higher redshifts (z 0.2). When near enough for details to be visible, they are seen to often occur in host galaxies that either have close companions or are involved in major mergers (Stockton 1990; Bahcall et al. 1997). The quasar OX 169, for example, features at least one tidal tail (probably a pair) and shows a variable H emission-line profile indicative of two active nuclei (Stockton & Farnham 1991).
Of special interest is the emerging connection between quasars and infrared-luminous galaxies. At bolometric luminosities above 1012 L the so-called ultraluminous infrared galaxies (ULIG) become the dominant population in the local universe and are 1.5-2 times as numerous as optically selected quasars. When ordered by increasing far-infrared color temperature, ULIGs and quasars seem to form an evolutionary sequence: ULIGs with low color temperature are starbursting disk mergers with well separated components, warm ULIGs appear to be just completing their merging into one object, and the `hot', optically visible quasars shine in peculiar ellipticals that resemble nearby merger remnants (Sanders et al. 1988, 1999). Nuclear separations and merger velocities indicate that the ULIG phase lasts about 200-400 Myr. Hence, extreme starbursts occurring while the nuclei merge and nuclear feeding frenzies climaxing in a quasar phase appear to be natural byproducts of elliptical formation through mergers. The peak quasar activity observed around z 2 may, then, mark the culmination of major mergers and elliptical formation.
Beyond z 2 we have precious little direct evidence of interactions and merging. The radio galaxy MCR 0406-244 at z = 2.44 may be one of the highest redshift mergers for which there is some detailed structural information. Deep optical Hubble Space Telescope images show a double nucleus and a 30 kpc-size pair of continuum-emitting `tails' suggestive of a tidal origin, while infrared images show two emission-line bubbles indicative of a strong bipolar wind (Rush et al. 1997; McCarthy 1999). Hence, at least some mergers at this high redshift may have been similar to local ones and involved pairs of already sizeable disks.
The important role played by interactions and mergers is also becoming apparent in galaxy clusters of increasingly high redshifts. Despite a widely held prejudice that mergers cannot happen in clusters because of high galaxy-velocity dispersions, both theory and observations show unmistakably that strong interactions and mergers do occur there. In some local clusters ongoing interactions and mergers are obvious. In Hercules at least five major interactions and mergers are visible in the central region alone (fig. 5), and even in relaxed-looking Coma `The Mice' (NGC 4676) provide an example of a major merger occurring on the outskirts. In z 0.2-0.5 clusters a fair fraction of the blue galaxies causing the Butcher-Oemler effect (Butcher & Oemler 1978, 1984) have been found to be interacting or merging (Lavery & Henry 1994), while a majority appear to be disturbed gas-rich disks shaken either by high-velocity encounters or minor mergers (Dressler et al. 1994; Barger et al. 1996; Oemler et al. 1997). Most impressive are new Hubble Space Telescope images of the rich cluster MS 1054-03 at z = 0.83. Fully 17% of its 81 spectroscopically confirmed members are ongoing mergers, all with luminosities similar to, or higher than, that of a L* galaxy (van Dokkum et al. 1999). These mergers occur preferentially in the cluster outskirts, probably in small infalling clumps, and present `direct evidence against the formation of ellipticals in a single monolithic collapse at high redshift'.
In order to quantitatively assess the impact of mergers on galaxy evolution, one needs to determine the merger rate (i.e. the number of mergers per unit time and comoving unit volume) as a function of redshift. We can only hope to do this for major mergers, since minor mergers are undetectable at z 0.5 and accretions are known only in the Local Group. There are many estimates of the merger rate based on counts of binary galaxies as a function of redshift, and on the assumption that most such binaries will merge in a short time. This assumption is a bit unrealistic, given that even for a much studied interacting pair like M 51 we do not know whether the presumed merger will occur within 2, 5, or 10 Gyr. Nevertheless, taken at face value several recent estimates based on binary counts suggest a merger rate approximately proportional to (1 + z)3±1 (e.g. Abraham 1999), implying an order-of-magnitude increase in mergers at z 1 compared to the local rate.
Figure 5. Galaxy interactions and mergers in Hercules cluster. Strongly interacting pair near lower left corner shows giant diffuse tails. Photograph courtesy of Alan Dressler.
Two estimates of numbers of mergers are relatively reliable and bracket the range of likely rates. First, given that there are ~ 11 ongoing disk mergers among the 4000+ galaxies of the New General Catalog (NGC) and their median `age' is ~ 0.5 Gyr, there should be about 250 remnants of similar mergers among NGC galaxies if the rate has remained constant since high redshifts, and about 750 remnants if - more realistically - the rate declined with time like t-5/3 (Toomre 1977). Thus, nearly 20% of all NGC galaxies may be remnants of major mergers, a fraction that agrees remarkably well with the observed number of elliptical and S0 galaxies. Second, if all gas collapsed into disks and all spheroids are due to mergers, then the fractional amount of mass in spheroids - about 50% when estimated from bulge-to-disk ratios of a complete sample of nearby galaxies - provides an upper limit to the integrated effect of all mergers (Schechter & Dressler 1987). This upper limit emphasizes that at least major mergers cannot have been too frequent, or else they would have destroyed all disks. Especially in late-type, nearly pure-disk galaxies (e.g. M 33 and M 101) most of the assembly must have been gaseous, dissipative, and - after perhaps some initial collapse phase - involving mere accretions.