Galaxy collisions drive galaxy evolution, but how much compared to other processes like the passive conversion of gas to stars in isolated galaxy disks or compared to other dynamical processes like ram pressure stripping? Also, how does the role of collisions change with cosmological epoch?
To begin, we note that it has become quite clear that at least some massive galaxies and massive disk galaxies, in particular, formed very early, and had already attained a respectable age by redshifts of 1-2 (see review of Spinrad (2004)). From a practical point of view, this means that observations must push to very high redshifts to see big evolutionary changes. We will come back to what has been seen in a moment. This fact has also been taken as evidence that at least some galaxies formed in a rapid monolithic collapse, rather than building up steadily in a prolonged sequence of mergers.
4.1. Models of Structure Buildup
What do theory and cosmological structure formation simulations lead us to expect? Currently, hierarchical build-up, CDM models (i.e., models with cold dark matter plus "" dark energy) in the "concordance cosmology" are the dominant paradigm. This picture suggests the occurrence of many mergers of small building blocks in the earliest stages, and continuing mergers thereafter. Moreover, recent analysis shows that it is possible to form some massive galaxies, including disk galaxies, at early times in these models Nagamine et al. (2005).
Thus, part of the solution to the paradox of early massive galaxy formation is that fully nonlinear CDM models do not yield exactly the same results as simple, analytic hierarchical structure formation models. Early massive disk galaxies may be a roughly 2 outcome of the simulations, but that may be sufficient to account for the observations.
From another point of view, the early formation of massive disks allows for the possibility that major mergers form elliptical galaxies at early times, i.e., accounts for ellipticals containing only old stellar populations within the merger theory (see discussion in Schweizer (2005)).
To return to the CDM paradigm, another thing the simulations show us is that when substantial entities merge, not all their substructure is erased. In fact, the absence of hundreds of dwarf satellites around the Milky Way has been cited as a problem for this kind of model (e.g., Klypin et al. (1999)). However, RPS in the hot halo, tidal disruption, and collisions with the galactic disk may have destroyed many of the leftover building blocks. Indeed, digesting the leftovers may be an important secondary evolutionary process, operating alongside the primary hierarchical merging process.
Thus, the picture of sequential buildup of galaxies via successive major mergers in the simplest hierarchical models is not the only one in which collisions and mergers are crucial. In more realistic models minor mergers and the accretion of numerous small companions play important roles. Such lesser collision events are probably very common in groups and clusters. As in solar system formation, "core accretion" may be as important as monolithic collapse and hierarchical buildup.
4.2. Observations of Evolution
Let us return to observation. As in the case of induced SF discussed above, there are two approaches - study of the statistical properties of large samples, or study of individual objects in detail. In the last decade there have been a great many surveys to provide data for the first type of analysis (see the overview of Irion (2004)). These include Hubble Space Telescope projects like the Medium Deep Survey, the Hubble Deep Fields North and South, and most recently GOODS (the Great Observatories Origins Deep Survey), carried out in collaboration with the Chandra Observatory and the Spitzer Space Telescope (see special issue of the Astrophysical Journal Letters, Jan, 10, 2004), and GEMS (Galaxy Evolution from Morphology and Spectral energy distributions, e.g., Bell et al. (2005)).
The science and observing techniques of deep field survey and high redshift studies in general are well beyond the scope of this review. This is also not the place to consider the many different classes of high redshift galaxies in any detail. These topics have become the subject of wide interest and a burgeoning literature. However, specific products of these studies, like the merger rates, the cosmic star formation rate and mean morphological statistics as a function of redshift, can provide information on the history of galaxy collisions.
4.2.1. Merger Rate versus Redshift
The differing predictions of the different models of galaxy formation, and the interest in the role of major mergers/ULIRGs, have motivated a continuing interest in the merger rate as a function of redshift. For example, the CNOC cluster galaxy redshift project has reported quite modest evolution in the merger rate at redshifts less than 1 (see Patton et al. (2002) and references therein). Specifically, Patton et al. examined a sample of 4184 galaxies, found 88 galaxies in close pairs, and derived a merger rate of (1 + z)2.3±0.7.
This result of low (and not rapidly changing) merger rate is confirmed by several other recent studies, including the Caltech Faint Galaxy Redshift Survey Carlberg et al. (2000). Moreover, Lin et al. (2004) report initial results of the DEEP2 survey, in which they find a merger rate of (1 + z)m with exponent of m = 0.51 × 0.28 assuming mild luminosity evolution, or m = 1.6 × 0.29 assuming no luminosity evolution, since z = 1.2. They note that this implies only 9% of L* galaxies have undergone major mergers over this redshift interval. Using deep infrared observations from the Subaru telescope Bundy et al. (2004) found that the fraction of close pairs (which usually define the merger rate in these studies) increases "modestly" to only about 7 × 6% at z 1. This is less than that found by typical optical studies, and they note that the optical studies may be "inflated" by unrepresentative "bright star-forming regions."
Going in the other direction, Lavery et al. (2004) find a very rapid increase in the number of colliding ring galaxies with redshift. Head-on ring galaxy collisions generally result in merger, so if the rings represent a small randomly chosen fraction of all mergers, this would imply a very rapidly evolving merger rate. On the other hand, if the number of ring galaxies increases much more rapidly with redshift than other types of merger, one wonders why?
For reference, we note that Xu, Sun, & He (2004) recently used data from the 2MASS near infrared survey to estimate the local merger rate; they found the fraction of close major merger pairs to be 1.70 × 0.32%. For completeness, we note that at the time writing, mergers rates based on the Sloan Digital Sky Survey or the 2dF survey have not been published, though an initial atlas of SDSS merger pairs has Allam et al. (2004). In the coming years it will be very interesting to see statistically significant estimates of the merger rate extended to well beyond z = 1.
4.2.2. Cosmic Star Formation
Estimates of the mean SFR as a function of redshift are usually based on color or emission line indicators (as opposed to the morphology used to estimate merger rates in the pair surveys). In recent years there have been surveys in a variety of wavebands. Cram (1998) carried out a novel radio continuum survey and found a local SFR of about twice the optical H value - 0.025 M yr-1 Mpc-3. He found a value about 12 times greater at z 1. An analysis based on the 2dF survey also found an strong increase ( (1 + z)b, with b < 5) back to z 1, and a more moderate increase at redshifts of 1-5 Baldry et al. (2002). Analyses based on SDSS data come to similar conclusions Glazebrook et al. (2003), Brinchman et al.(2004). The HST STIS Parallel Survey found an SFR at z 1 of 0.043 × 0.014 M yr-1 Mpc-3, based on [OII] emission Teplitz et al. (2003), which is lower than most of the previous results.
The CADIS survey found SFR decreased by about a factor of 20 between redshift 1.2 and the present, and the authors note the agreement of their extinction corrected results with far infrared results (Hippelein et al. (2003), also see the Herschel Telescope survey of Glazebrook et al. (2004)). Results from the Gemini Deep Deep Survey indicate that the SFR was about 6 times higher at z = 2 than at present Juneau et al. (2005). One of the most dramatic changes in SFR was the factor of 30 found in a GALEX (Galaxy Evolution Explorer satellite) ultraviolet survey between z = 1 and the present Schiminovich et al. (2005). Ultraviolet luminous galaxies may not very representative of the cosmic SFR, but they could be related to colliding galaxies.
The newest and deepest surveys indicate that SFR declines from a peak at moderate redshifts to lower values at the highest redshifts. Bundy et al. (2004) identify and study 54 galaxies in the Hubble Ultra Deep field and conclude that the SFR at z = 6 is about 6 times less than at z 3. Heavens et al. (2004) come to similar conclusions about the general history of SF on the basis of an analysis of SDSS and other surveys. Juneau et al. (2005) describe the situation as a cosmic starburst at z 2.
Very recently, survey results have revealed the phenomena of "downsizing," wherein the most massive galaxies form first, and most of the SF takes place in progressively smaller galaxies as time goes on (e.g., Poggianti et al. (2004), Bouche & Lowenthal (2005), Juneau et al. (2005), Le Borgne et al.(2005), Shapley et al. (2005)). Bundy, Ellis & Conselice (2005) argue (based in part on GOODS data) that downsizing also proceeds from early to late Hubble types, and that merging plays a key role. The implication is that there is a mass dependence in the merger rate at any given epoch.
These new cosmic SFR results provide very interesting inputs to the story of galaxy evolution. However, the relation of these results to interactions and mergers remains to be clarified. Actually, the same is true of the merger rate results, which are not sensitive to many minor mergers or other interaction phenomena.
4.2.3. Morphology versus Redshift
For an outsider the phenomenology of high redshift galaxies, which is much constrained by detection techniques, is a daunting jungle of jargon. Moreover, the relation between increasingly elaborate simulations of structure formation and the observations is complex. With rapid advances on both fronts, and increased efforts in analysis and synthesis, we can expect much more clarity in the coming decade (see review of Spinrad (2004) for a lucid current discussion). For the present we focus on a few simple questions. Are we directly observing galaxy evolution, i.e., the changing appearance (build-up) of galaxies with redshift? Are collisions and mergers an important part of this evolution?
There is much new evidence in favor of an affirmative answer to both questions (see commentary of Conselice (2004), with further details on GOODS data in Conselice et al. (2004)). To say it a bit more emphatically, these papers and those referenced within them suggest that we may be beginning to acquire the observations that directly show the buildup of typical Hubble sequence galaxies (see Figure 9).
Figure 9. Conselice et al. (2004) sequence of "low density objects" at varying redshifts, illustrating the development of Hubble type galaxies. (Courtesy C. Conselice)
There is much information to be found in the literature on the properties of individual high redshift objects (individual galaxies and clusters). We cannot review this literature here, and refer the reader to the review of Spinrad (2004). Instead, let us return to the subject of massive, or at least luminous, galaxies at high redshift, and in particular, the interesting classes of extremely red galaxies and submillimeter galaxies (or SCUBA galaxies, after the detector on the James C. Maxwell Telescope). The latter are very infrared luminous, high redshift objects (e.g., Conselice, Chapman, & Windhorst (2003), Genzel et al. (2004), Swinbank et al. (2004), Pope et al. (2005)). Until recently, only a few were known, but recent deep searches are beginning to detect a substantial number (Greve et al. (2004), Smail et al. (2004), Wang, Cowie, & Barger (2004)). It appears that most of these objects are either dust obscured quasars or high redshift LIRGs or ULIRGs, with perhaps about 2/3 being the latter (Conselice, Chapman, & Windhorst (2003), Neri et al.(2003), Smail et al.(2004), Swinbank et al. (2004)). As with their local counterparts, the LIRGs and ULIRGs are generally mergers or mergers-in-progress (see Georgakakis et al. (2005)).
It appears that the submillimeter LIRGs are much more common than present day LIRGS, and that they generate a substantial fraction of the IR background (e.g., Genzel et al. (2004), Wang, Cowie, & Barger (2004)). They have typical redshifts of 2-3, and thus, coincide with the peak of the cosmic SFR. They may be well represented among the most luminous galaxies in that peak, but the evidence is preliminary. These results are beginning to provide direct evidence that major mergers, if not hierarchical buildup, were major contributors to galaxy evolution and the cosmic star formation rate at these redshifts.
The submillimeter galaxies may be related to another high redshift class, the Lyman Break Galaxies Shu, Mao, & Mo (2001). However, few of the latter are detected as the former (Chapman et al. (2000), but note the outstanding exception Westphal-MMD 11 discussed in Chapman et al. (2002)). The deep ISO (Infrared Space Observatory) ELAIS survey also found a number of ULIRGs at z < 1 Rowan-Robinson et al. (2004). These objects may bridge the gap between local ULIRGs and SCUBA galaxies. Recent observations with the Spitzer Space Telescope show that SCUBA galaxies are generally detectable at 24 microns, but with a wide range of mid-infrared colors Frayer et al. (2004). Spitzer observations also promise to delineate active nuclei from starburst powered submillimeter galaxies Ivison et al. (2004). All of this work should contribute substantially to our understanding of the "ULIRG rate" as a function of redshift (see the review of the cosmic evolution of luminous infrared galaxies by Sanders (2004)).
Before submillimeter galaxies were discovered, observers were already very interested in "extremely red objects" at high redshift. Naively, one might expect to find more and more blue galaxies at high redshift, and as described above, this is generally the case. In this context, finding very red galaxies is surprising. On the other hand, with a knowledge of dust-enshrouded starbursts in ULIRGs, maybe this is not so surprising, but are EROs ULIRGs? Recent studies suggest not, but rather many of them may be the already (at typical redshifts of > 1-2) old, red progenitors of present-day early type galaxies (e.g., Franx et al. (2003), Förster Schreiber et al. (2004), Yan & Thompson(2003), Yan et al.(2004), Yan, Thompson, & Soifer (2004), Bell et al. (2005)).
Redness and age are relative terms. The typical age of the stellar populations in these galaxies may be about 1 Gyr, which locally would not be described as an old, red population. However, at the high redshifts where these galaxies are found, the age of the universe when the light was emitted was only a few Gyr or less.
Nonetheless, a fraction of the extremely red objects may be ULIRGs. Yan & Thompson (2003) in an HST study of the morphology of a sample at redshifts of about 1-2, estimate that about 17 × 4% of the objects are mergers or interactions. However, for the majority dominated by older stellar populations we will have to seek merging and interacting progenitors at still higher redshifts.
In conclusion, the above paragraphs describe the great advances that have been made in recent years in studies of galaxy evolution at high redshift. This work is impressive, but it is still hampered by resolution, sensitivity and statistical limitations. There are hints that mergers and interactions are important at all stages, but there is a great deal more work to do before we understand the details.