GALAXIES, BINARY AND MULTIPLE INTERACTIONS WILLIAM C. KEEL The large size of many galaxies in comparison to their separation means that tidal effects during close passages can significantly affect the galaxies' morphology, star formation, and gas content. An important fraction of luminous galaxies is found in binary systems or small groups, the properties of which can be used to probe the masses and dynamical history of the constituent galaxies. Close encounters are highly inelastic; galaxies can merge with one another in relatively short times. This process has been associated with the formation of elliptical galaxies from collisions of disk systems and with the formation of luminous galaxies from smaller precursors. There is direct observational support for links between interactions and such energetic and short-lived phenomena as infrared-bright galaxies, starbursts, and active nuclei of galaxies. BINARY GALAXIES Early studies of the projected distribution of galaxies on the sky showed a clear excess of close pairs above the number to be expected for a population randomly distributed in space. Erik Holmberg at Lund Observatory in Sweden compiled the first extensive catalog of such pairs, and noted correlations between several properties of galaxies in pairs. Such correlations suggest evolutionary connections between the galaxies, resulting either from common environment during galaxy formation or from mutual influences during the history of the galaxies. Specifically, Holmberg found that galaxies in a pair tend to have similar morphological (Hubble) types and, to a degree larger than expected from the type correlation, to have similar optical colors (the Holmberg effect). Later analysis of the binary-galaxy population, incorporating radial-velocity information, has confirmed that these are bound systems, and not simply chance encounters of unrelated galaxies. There are too few pairs with evidence of physical association and the large velocity differences to be expected from chance encounters of unbound systems; these flyby systems are found mainly in clusters, where the high galaxy density and velocity dispersion both favor chance encounters. The dynamical properties of binary galaxies can be used to trace the extended mass distribution in galaxy halos to very large radii, much as is done with rotation curves. Attempts to do so in detail have been hampered by the difficulty of isolating pure samples of physical pairs of galaxies. The most important source of interlopers in statistically selected catalogs is not foreground or background systems, which can be weeded out using redshift data. Rather, it is the large number of galaxies in the vicinity of a given galaxy (on the scale of a group, for example), which can appear, in projection, artificially close to it. Statistics of the clustering of galaxies (such as the two-point correlation function) do not clearly distinguish pairs from the general clustering trend among galaxies; so there is no physical scale at which pairs can be distinguished from two members of a group that happen to lie (or appear to lie) unusually close together. High-reliability subsets of the binary-galaxy population can be identified by using very stringent selection criteria or requiring evidence of tidal disturbance, but these techniques bias the sample toward various subsets of the whole population of binaries and are not satisfactory for overall statistical use. INTERACTIONS AND MERGERS Because the mass in galaxies is widely distributed, some of a pair's orbital energy can be converted into the energy of orbits of stars and gas clouds within the galaxies. This leads to changes in the morphology of galaxies during and after a close passage due to tidal disturbances (see Fig. 1), and to decay of the mutual orbit followed by actual merger of the galaxies. Morphological disturbances are most pronounced in disk galaxies; on the other hand, the larger random stellar velocities at each point in an elliptical smear tidal features into broad fan-like structures. The dynamically cold situation in disks, with much smaller velocity dispersion, allows production of coherent structures on very large scales, in some cases greater than the initial size of the disk. These are in turn most pronounced after direct (prograde), rather than retrograde, encounters, because the companion galaxy moves slowly in the rotating reference frame of the disk and so builds up a large total tidal impulse in part of the affected disk. Detailed numerical simulation of individual interacting systems, matching the overall forms and radial-velocity fields of the galaxies, has proven quite successful in understanding the spectacular bridges and tails produced in close tidal encounters. The interacting systems so modeled include ellipticals as well as disk systems; in these cases the recent orbital history of each pair is well understood. In principle, comparisons of observations and simulations could yield estimates of the amount of mass in an extensive unseen halo; this test is not yet sensitive enough to make it a useful probe of halo matter. There has been more success in exploiting the differing dynamics of stars in disks, and in elliptical galaxies or bulges in tracing relatively faint remnants of accreted companion galaxies that once had disks. These remnants are often seen as sets of ripples superimposed on the smooth background light of the more massive, relatively undisturbed system. Tidal debris around disk galaxies is best seen in polar rings, which represent material accreted from a companion that remains in near-polar orbit over the existing disk. Such orbits decay very slowly compared to those nearer the disk plane, and the material is not spread widely by the differential precession that would affect matter in an annulus close to the disk plane. These structures are best seen in SO systems, in which there is little interstellar material to interfere with the motion of the accreted matter. The transfer of energy responsible for tidal distortion in a galaxy's stellar distribution can also have a dramatic impact on the orbital evolution of a galaxy pair. In addition, an effect known as dynamical friction can accelerate the decay of an orbit when one galaxy approaches within the stellar distribution of the other. This is the statistical result of gravitational scattering of stars by the intruding galaxy. The whole aggregate of collisions, even if they are individually elastic in the intruder's reference frame, saps the intruder's orbital energy, because of the asymmetry in stellar distribution produced in the wake of the intruder. This loss of energy is faster for larger intruder masses and at smaller distances from the larger galaxy's core. For an initially grazing encounter of two marginally bound galaxies, a merger can result in only a few initial orbital periods. The past history of the interacting-galaxy population is therefore very difficult to ascertain, because there should have been many more systems that have already merged than we now see as pairs for any plausible primordial distribution of orbital parameters. The identity of the merged remnants is important in understanding the dynamical history and formation of galaxies in general. Broad statistical considerations suggest that most present luminous galaxies have undergone at least one merger of comparable systems at some time. Mergers with companions of much lower mass may not leave strong traces, but a merger of near equals should produce a dynamically distinct result. A substantial body of work suggests that the product of a merger of disk systems would look much like an elliptical galaxy; the burst of star formation often triggered during such an event can sweep most of the interstellar matter out of the remnant. Some ellipticals in fact show evidence, such as a core in retrograde rotation of having swallowed another system, but it remains unclear whether most ellipticals grew in this way. Protracted, piece-by-piece formation of bright galaxies through accretion and mergers of smaller predecessors may fit well with the high star formation rates inferred in some high-redshift galaxies (such as the most luminous radio galaxies in the 3C [Third Cambridge] catalog) while preserving the range of epochs for galaxy formation required by cosmologies dominated by cold dark matter in the amounts implied by galactic rotation curves. INTERACTIONS AND STAR FORMATION Considerable evidence links tidal disturbances of galaxies to increases in their rates of star formation. The fraction of strongly interacting galaxies with starburst nuclei, for example, is much higher than that of similar noninteracting galaxies with such nuclei. Statistical examination of large samples of systems indicates that both nuclear and disk-wide star formation rates are higher in the interacting systems, though the median enhancement is rather small and the most spectacular effects are confined to a small percentage of the systems. It must be stressed that most interacting galaxies do not show strong starburst responses to interactions, but that those that do have such high (temporary) luminosities that they are found very efficiently by flux-limited surveys such as the Infrared Astronomical Satellite (IRAS) far-infrared survey or the Markarian ultraviolet-excess survey. The enhancement of star formation in some interacting systems is widely thought to result from the behavior of interstellar clouds during cloud-cloud collisions, the rate of which is greatly increased during interactions even when clouds from one galaxy never contact those from the other. More detailed understanding is limited by lack of direct checks on just how a molecular cloud responds to a collision. The existence of many disk systems with strong tidal disturbances but very low star-forming rates implies that several parameters are involved. The situation is clearly more complicated than a simple triggering by even strong tidal disturbance. Interaction-induced star formation is not dominant at the present epoch; only a few percent of the population of massive stars in nearby galaxies is due to such induced processes. Extrapolation back in time is important to see whether interactions have ever been important in driving the evolution of galaxies in general (as opposed to just the few with strong bursts), which is not yet possible with certainty. If large numbers of today's bright galaxies are merger remnants, much of the total star formation in them might have been induced. On the other extreme, if the merger rate has been rather low, there might not have been many more close interactions in the past than are seen now. Considerable recent work has focused on the nature of a population of galaxies largely found as far-infrared sources during the IRAS survey which have bolometric luminosities up to 10** times that of the Sun, most of which is radiated in the 25-100-** range. It is not clear in many instances what the original source of energy is; grains might be heated by starbursts or dust-shrouded active nuclei, and in fact optical and near-infrared spectroscopy finds each of these in some systems. Many of these systems are in strongly interacting or merging systems, with fractions estimated from 50-100% in various surveys. An evolutionary scheme has been introduced involving mergers inducing starbursts which then feed a central massive object via mass loss and disruption of red-super-giant envelopes, thus linking all these phenomena in a causal sense. However, many physical details remain to be clarified; star formation in violent environments is not understood. A competing picture, of energy release directly from plasma processes in cloud collisions during direct impacts and mergers, has had some success in accounting for the occurrence and luminosity of these systems. A mix of physical processes may well be operating in these infrared-bright systems. TRIGGERING OF ACTIVE GALACTIC NUCLEI Several lines of evidence point to a connection between galaxy interactions and nuclear activity, defined here as phenomena in addition to those produced by stars or stellar remnants and thus including quasistellar objects (QSOs), Seyfert nuclei, radio galaxies, and their relatives. Indeed, the earliest round of extragalactic radiosource identifications showed a suspicious number of interacting or colliding systems, but the original interaction model for such sources went out of vogue once interferometric maps showed the sources to be symmetric about individual galaxies. Some studies show that optically selected Seyfert nuclei occur preferentially in galaxies with companions. However, the exact significance of this result is difficult to assess, since other studies with a somewhat different way of selecting a control sample show only a marginal excess of companions around Seyferts. This perhaps emphasizes the crucial role played by comparison samples in observational studies of interactions. A stronger result is shown by the converse experiment, in which large, statistically selected samples of interacting systems are observed spectroscopically and the number of Seyferts found in this way is compared with control samples. There is a clear excess of Seyfert nuclei in pairs with small projected separation but little tidal distortion, and there is an equally clear deficit of Seyfert nuclei among strongly distorted systems. These conflicting trends may imply a temporal sequence, in which case the activity must be short-lived compared to the lifetime of tidal features. Studies of the structure and environment of radio galaxies also indicate that most high-luminosity radio galaxies show structural evidence of recent interactions or mergers; the dust lane of Centaurus A prefigured this conclusion long ago. Signs of interactions frequently seen in these objects include close companion galaxies, tidal tails, "shells," and kinematic irregularities. These symptoms are largely confined to the systems with the most powerful radio emission, suggesting that there may be a distinction in cause between low-and high-luminosity radio galaxies. This distinction parallels the structural difference reflected in the Fanaroff-Riley classification of the radio sources themselves: More powerful sources are more likely to have undergone a recent interaction, and the associated extended radio lobes generally have prominent hot spots and well-defined edges (expected for supersonic jets). The situation for QSOs is more difficult to assess, because most are so distant that structural information on surrounding galaxies or companion systems is not yet available. For objects at redshifts below about 0.4, there are strong indications that more companions or instances of tidal distortion in the host galaxies are present than might otherwise be expected. Such a trend appears to continue that found for Seyfert nuclei, but the fact that many of the Seyfert nuclei are of rather low luminosity, and the apparent occurrence of some QSOs in very disturbed host galaxies, may argue for two different mechanisms of production. The existence of triggering of active nuclei via interactions must probe both the mechanism of fueling for the central engines of these objects and the transport of mass and angular momentum across large radial distances during interactions. Unraveling the connection is both one of the most promising and one of the least tractable problems in studying active nuclei. Several physical mechanisms likely to be operating have been explored in some detail. The nucleus may be fueled either by interstellar gas, or by gas liberated during tidal disruption of stars venturing too close to a massive core object. If the fueling is in the form of diffuse gas, it most likely originates within the host galaxy, but has been moved from farther out in the disk; this form of fueling must somehow overcome a large angular-momentum barrier in order to reach the proximity needed to be accreted by the central source, or captured into an accretion disk. Fueling via disruption of stars appears to require either that massive binary black holes are common, or that substantial bursts of star formation produce many red supergiants with loosely bound atmospheres. This is the extreme case of the general problem of "feeding the monster." Regardless of just how active galactic nuclei are fueled during interactions, the fact that such fueling can be recognized in large samples of systems has several interesting consequences. Individual encounters (or mergers, for that matter) last a very short time compared to a galaxy's lifetime. Thus, any episodes of nuclear activity triggered by encounters must be comparably short-lived, or we would not observe a significant excess of active nuclei in visibly interacting systems. Simple estimates of this lifetime suggest that individual episodes of activity last on the order of 10**yr. This implies that the luminosity function of these objects is to be interpreted as a statistical property of a population of short episodes, not as the distribution of individual objects that might change with cosmic epoch. Furthermore, the existence of triggering suggests that most luminous galaxies have the prerequisites for nuclear activity (such as a central massive object) and need only the proper conditions for its expression. Thus, most bright galaxies would have spent a few percent of their history with active nuclei, so that fossil quasars, for example, should be rather common. DENSE GROUPS OF GALAXIES The rapid time scales for merging of close neighbors indicate that strong interactions should mostly be two-galaxy processes; one would seldom expect additional objects to become involved during the relatively brief interaction between two systems. There is, however, a substantial population of groups that are so dense that their very existence may pose a challenge to current understanding of the origin and dynamical evolution of groups of galaxies. In some cases tidal interaction can be observed among four or five galaxies in a single group. If we observe these groups at a typical time (that is, if they are not looser groups accidentally viewed in angle or time so as to appear unusually dense), the time scale for many of these groups to merge into single objects is much less than a Hubble time. The presence of tidal features in at least some of these suggest that the galaxies are about as close together as they appear. There is a two-sided puzzle posed by the expected mergers of these systems: What are the remnants of the historical population of these groups, and why are we just now seeing the dregs of an originally much larger population of these groups? Additional Reading Arp, H.C.(1966). Atlas of Peculiar Galaxies. California Institute California Institute of Technology, Pasadena. (Also appeared in Astrophys. J. Suppl. Ser. 14 1). Arp, H.C. and Madore, B.F.(1987). A Catalogue of Southern Peculiar Galaxies and Associations. Cambridge University Press, Cambridge. Binney, J. and Tremaine, S.(1987). Galactic Dynamics. Princeton University Press, Princeton, Chap. 7. Keel, W.C.(1989). Crashing galaxies, cosmic fireworks. Sky and Telescope 77 18. Schweizer, F.(1986). Colliding and merging galaxies. Science 231 227. Toomre, A. and Toomre, J.(1973). Violent tides between galaxies. Scientific American 229 (No. 6) 39. [Reprinted in P. Hodge, ed. (1984). The Universe of Galaxies. Freeman, San Francisco, P. 55.] See also Active Galaxies, Seyfert Type; Active Galaxies and Quasistellar Objects, Central Engine; Galaxies, Infrared Emission; Galaxies, Properties in Relation to Environment.