1.1. Early Days
The study of galaxy collisions is not an ancient one; Erik Holmberg, Fritz Zwicky, and a few others did quite a bit of work relevant to colliding galaxies before the 1950s, but that decade opened and closed with two landmark papers. Thus, we can justify taking it as the first decade of general interest in the subject, and view the earlier work as pioneering. The first of the two papers, by Spitzer & Baade (1951), revived Zwicky's suggestion that collisions would be frequent within dense galaxy clusters, and considered what would happen in direct collisions between two galaxies. Specifically, they correctly argued that the stellar distribution might be only moderately disturbed, while strong shock waves could push the interstellar gas out of the galaxies. Their primary conclusion was that this process could account for the scarcity of late-type spiral galaxies with substantial ongoing star formation in clusters. For the first time galaxy collisions were seen to have an important role in galaxy evolution.
The second landmark paper was Zwicky's review of his extensive imagery of morphologically peculiar galaxies, together with arguments that many of these peculiarities were caused by tidal forces in collisions Zwicky (1959). At the time, Zwicky's theory seemed doubtful, since collisions between the widely separated "island universes" were deemed improbable. Moreover, his arguments were generally semi-quantitative, and so, not compelling.
In the middle of this first decade Baade & Minkowski (1954) suggested that one of the most prominent members of the newly discovered class of radio galaxies, Cygnus A, was in collision. Thus, we already have the first hints of many of the most important themes in this field, including: the generation of unique tidal morphologies, induced nuclear activity, induced star formation, the important role of collisions in galaxy evolution, and the dependence of these effects on the clustering environment. In this review I will focus on the last few items - induced star formation, galaxy evolution, and environmental effects - and say relatively little about the first two and many other related topics.
In the second decade, much of the work was of a more detailed, and sometimes indirect nature, which I cannot review here (see e.g., Struck (1999)). The major exception to this generalization was publication of Arp's pictoral atlas of more than 300 peculiar galaxies Arp (1966). Arp derived the atlas objects from the Palomar sky survey. In the following decades this atlas became the standard 'field guide' for workers in this field. The Arp galaxies were arranged in categories somewhat like Zwicky's, but with many more examples, and excellent photographic images. One psychological effect of so many images may well have been to make the peculiar galaxies seem less like freakish rarities, and more like zoological families in need of explanation.
1.2. The 1970s
Toomre & Toomre (1972) took a giant step toward these explanations. Their numerical models were not the first, and were simple by modern standards, but they were more extensive than previous efforts. They were able to account for many of the Arp atlas forms in detail, thereby making a strong case for the collisional origins theory. They also made a number of important predictions for observation, such as that strong tidal waves would lead to enhanced star formation and gas transfer to nuclear regions, which could fuel nuclear activity. These would become dominant themes in subsequent work.
However, Toomre and Toomre's models did not directly account for the 'messiest' objects in the Arp atlas (see examples in Figure 1). Alar Toomre returned to these objects in his contribution to the seminal Yale galaxies conference Toomre (1977). He pointed out that the earlier models had not included the effect of Chandrasekhar's dynamical friction Chandrasekhar (1943), and showed that the effect would draw the colliding galaxies into a merger. He further considered how merger remnants would evolve, and how they would appear observationally. This lead him to some radical conclusions for that time, that mergers between comparable large spiral galaxies could lead to the formation of elliptical galaxies, and that reasonable extrapolation of the statistics of such collisions suggested that a large fraction of ellipticals could be formed this way. The debate still continues on many aspects of this scenario, but it immediately had an important effect. Toomre had opened the door to the possibility that collisions were the dominant factor in the evolution of an important class of galaxies. Collisions were more than just a means of accounting for rare freaks, or a specialized process peculiar to the environment of dense clusters.
Figure 1. Multiwaveband images of several well-known merger remnants (courtesy D. B. Sanders and I. F. Mirabel): a) NGC 4038/39 (Arp 244, "The Antennae"), b) NGC 7252 (Arp 226, "Atoms for Peace"), c) IRAS 19254-7245 ("The Super-Antennae"), d) IC 4553/54 (Arp 220). An optical image shown in greyscale, HI (21 cm) surface intensity shown by contours, and K band (2.2 mm) shown in insets. Scale-bar represents 20 kpc in each case; see Sanders & Mirabel (1996) for details.
Other important developments in the 1970s included the work of Larson & Tinsley (1978), who suggested that Arp atlas galaxies had a wider range of optical colors and star formation rates (SFRs) than the more normal Hubble atlas Sandage (1961) galaxies. Extension of that work suggested that infrared colors would provide even more sensitive indications of varying SFRs Struck-Marcell & Tinsley (1978). Ever increasing evidence that galaxies (and groups and clusters) possessed massive dark halos (see Sofue & Rubin (2001) for a history of rotation curve studies) completely changed our understanding of what a galaxy is. The ten-fold increase of galaxy masses and sizes in the new picture provided an explanation of why collisions could be common, despite the great separations of the visible parts of galaxies. Their cross section were much larger than previously thought, and collision partners were born bound together in larger entities.
1.3. The 1980s and Early 1990s
This period saw expansion of the field into many new directions, with a number of major developments that defined the current epoch. One of the highest points was the discovery of ultraluminous far-infrared galaxies (ULIRGs) with the observations of the IRAS satellite (see reviews of Soifer, Houck, & Neugebauer (1987) and Sanders & Mirabel (1996)). This discovery set off a gold rush of studies of these objects, as illustrated by the papers of the 1986 Pasadena meeting Lonsdale Persson (1986) and the 1989 Alabama meeting Sulentic, Keel, & Telesco (1989), and which continues to some degree up to the present. A primary focus of most ULIRG papers has been the relative role of nuclear starbursts versus active nuclei in generating the huge emissions. This is a difficult question to answer because both are usually buried deeply in the gas and dust of the merger remnant; most observational techniques give only indirect clues. While elucidating the connection between starburst and nuclear activity is very important, the ULIRGs and their somewhat less luminous cousins, the LIRGs, offer a wealth of other information on questions of galaxy evolution.
A second focus of ULIRG studies was the determination of what sort of remnant would ultimately emerge from a major merger. ULIRGs could be seen as the missing link in Toomre's theory of elliptical formation from major mergers. They are recent mergers with prodigious amounts of star formation, which might eventually either consume or heat and disperse the gas, as required by the theory. The fact that the old star surface brightness profile approximated the de Vaucouleurs profile characteristic of ellipticals in the inner regions of some ULIRGs, despite the presence of tidal distortions in the outer parts, gave further support to the theory.
This was generally a period of rapid development of numerical models. It began with the publication of the first fully self-consistent three-dimensional models of galaxy collisions followed through the merger (see review of Barnes & Hernquist (1992)). In these models the galaxies were of comparable size and consisted of a single spheroidal component, i.e. like two elliptical galaxies without dark haloes. They showed that mergers occurred much more quickly than expected, as orbital energy was efficiently channeled into internal collective modes. They also revealed the rapid appearance of a de Vaucouleurs surface density profile in some major merger remnants. This profile can be viewed as a kind of meta-stable state, resulting from the prompt relaxation of collective modes. Its appearance in ULIRGs indicated agreement between observations and models, and provided more support for the ellipticals from mergers theory.
By the end of this period the state of the numerical art had advanced to self-consistent merger models of galaxies with stellar disk, gas disk, and dark halo components Barnes & Hernquist (1992). These models showed that different galaxy components behaved somewhat differently during the (major) merger process, with dynamically hot halo components generally merging more quickly than the disk components. Even more exciting from the point of view of ULIRG studies, the models showed that a fraction of the gas carried much of the angular momentum out into extended tidal structures, while the rest of the gas fell into a small volume in the remnant center. This mass of highly compressed gas could readily fuel ULIRG superstarbursts.
This period did not see many models of mini or micro mergers, in part because ULIRGs and major mergers were the focus, but also because adequate numerical resolution of small companions was difficult. Another lacuna of modeling in this period was realistic gas dynamics; most models used either an isothermal equation of state for the gas or 'sticky particle' algorithms with phenomenological collision rules between particles representing gas clouds. Cooling, heating and stellar feedback processes were not generally included, (but see e.g., Appleton & Struck-Marcell (1987b)).
Alongside the major thrusts of merger studies several quiet revolutions occurred in this period. One of these was based on the sensitive mapping of atomic hydrogen in galaxies generally, as well as collisional systems, by many observers using the Westerbork array, and later the VLA (Very Large Array of the National Radio Astronomy Observatory). These observations first made clear that the gas disks of typical disk galaxies were much larger than the stellar, and then as one might have expected, that these extended gas disks were more strongly affected by collisional encounters than the inner stellar disks. It soon became clear that such observations were essential for determing the full extent of tidal tails and bridges. HI mapping also provides a map of the line-of-sight velocities of the gas. Kinematic maps provide us a view in a third dimension of the six dimensional position-velocity space, and this information is usually crucial to the success of models of individual systems, thereby to detailed tests of collision theory. The accomplishments of the VLA were summarized at a recent symposium Hibbard, Rupen, & van Gorkom (2001), and a valuable legacy of that meeting was the creation of the HI Rogue's Gallery website of colliding galaxy HI maps by J. Hibbard (www.nrao.edu/astrores/HIrogues/webGallery/webGallery.html).
Another discovery that can be described as revolutionary is that tidal interactions can induce the formation of a bar component out of disk material. This was shown by the numerical models of Noguchi (1987), and studied in detail by Athanassoula (see review of Athanassoula (2004) and references therein). This result is important because bars transfer angular momentum outward in the disk, and so can drive gas into the central regions before merger. The bar can also drive spiral density waves. Both the increased central gas concentration and the bar/spiral waves can induce star formation.
We will examine the question of SF induced before merging in more detail below. However, we should note here that Keel, Kennicutt and collaborators carried out an extensive program of H imagery and spectra of both collisional systems and of a control sample (Keel et al. (1985), Kennicutt et al. (1987)). They found indications of enhanced SF in the collision sample, and particularly of SF enhancements in galaxy cores which were kinematically disturbed. On larger scales, Schombert, Wallin, & Struck-Marcell (1990) observed the broad band colors of a sample of tidal bridges, plumes and tails, and found that while SF in these structures was not especially strong, it did continue after their formation. This is somewhat surprising given the great extent of many of these structures, which would seem to imply diminished gas densities and SF.
In his continuing studies of putative merger remnant-to-elliptical systems, Schweizer also discovered large, young star clusters or dwarf galaxies formed in tidal tails, most notably in the "Antennae" system Schweizer (1983). These discoveries would inspire a great deal of new work in the 1990s and the present decade. More generally, Schweizer's detailed, multi-waveband studies of specific merger remnants, whose appearance suggested that they were on the road to becoming ellipticals, advanced Toomre's merger theory (see Schweizer (1998) and references therein).
As a final example of quiet revolutions of the 1980s I would include the extensions to dynamical friction theory by Tremaine & Weinberg (1984), and the application of the new theory to the evolution of galactic bars Weinberg (1985). The classical Chandrasekhar (1943) theory was too idealized to account for the frictional effects in major mergers, and even more so in the case of a "sinking satellite" orbiting outside of, but interacting with the disk of the primary galaxy. The Tremaine and Weinberg theory included the collective effects not accounted for in the classical theory, and is able to account for the rapidity of major mergers seen in numerical models.
Even beyond these revolutionary examples the tapestry of colliding galaxy studies also grew with the addition of more new threads in this period. These included studies of many specific types of collisional system, such as: colliding ring galaxies (see review of Appleton & Struck-Marcell (1987a)), polar rings (see review of Sparke (2002)), ocular ovals Elmegreen et al. (1991), and shell galaxies (e.g., Hernquist & Quinn (1988)). Numerical modeling demonstrated how these distinctive morphologies could be produced in collisions, and thus confirmed earlier conjectures on the broad scope of collision theory. In addition, distinctive morphologies were generally found to be the result of a relatively narrow set of collision parameters. Examples in each class can be viewed as a set of related natural experiments, seen at different times and with slightly different initial parameter values, which have the potential to provide much insight into difficult or obscure collision processes (e.g., hydrodynamic or SF processes).
1.4. Key Issues Up to the Present
The 1990s saw continued rapid expansion of the field, driven in part by new ground and satellite-based instrumentation, and by rapidly increasing computer power. It is very difficult to summarize the accomplishments of that decade briefly. Queries to NASA's Astrophysical Data System show that the number of literature papers with abstracts containing the words "galaxy" and "collision" grew very rapidly with each decade: 27 (1950s), 75 (1960s), 326 (1970s), 826 (1980s), 1413 (1990s). Similar increases in the number of studies in the related fields of galaxy formation and galaxy evolution at high redshift make the task even more difficult. In this review we will focus our attention on key issues relating to star formation and galaxy evolution.
It is clear that over the second half of the 20th century this field has gone from bare beginnings to a considerable maturity, providing answers to some of its most important questions and early paradoxes. Yet many questions remain, including some that have been common threads through the whole history of the subject, and which are connected to the deepest questions in astrophysics. For reference in the rest of this review, I list here some of the most important ones.
1. How do collisions and interactions affect galaxy evolution overall? More precisely, what are the relative roles of major and minor mergers in building galaxies? This question is related to that of how galaxies form, since major mergers are very important in hierarchical build-up models, and negligible in monolithic collapse models of galaxy formation.
2. How does the answer to the previous question depend on environment? How do collisions differ in cluster, group, or nearly isolated environments? Some partial answers to these questions have been known for a long time. For example, collisions between field galaxies are very different from those between cluster galaxies because the latter have typical relative velocities of thousands of km/s versus velocities of hundreds of km/s in the former case. High velocity collisions can remove interstellar gas and produce moderate tidal distortions, but are unlikely to result in merger, while mergers are generally inevitable in the lower velocity collisions in groups. Research over the last few decades has provided a great deal of information on these questions, and it has become clear that environment plays a very large role in determining the nature of collisions that can occur, and the relative importance of galaxy collisions versus other evolutionary processes (like gas sweeping in dense clusters).
3. How do the large-scale dynamics of collisions and interactions orchestrate star formation (SF) and nuclear activity, which are inherently small scale processes? The clear answer from the 1980s is that activity is induced by dumping a great deal of gas into the central regions of major merger remnants. Major mergers may be the way to make most of the stars in a significant fraction of early type galaxies, but they are a rare event in the world of galaxy collisions, and the question remains for other types of collision. Related questions include: when do galactic winds and fountains result from interaction induced SF, and what feedback role do they play in the subsequent SF?
4. To what degree do mergers trigger long-term (more than 1.0 Gyr) secular processes? Examples include the long-term effects of collisionally induced bar components, and the fallback of large scale tidal structures.
5. How far can we push the archaeology of individual systems? Do enough clues remain to determine the morphology of the precursor galaxies, and decipher the details of the interaction up to the present?
In the remainder of this review we will consider how developments in the last decade and the near future help to answer these questions. The first three sets of questions include the key questions of this review. The last two push beyond its scope, and I will not treat them in any detail, despite their intrinsic interest.