Galaxies: Interactions and Mergers C Mihos

eaa.iop.org DOI: 10.1888/0333750888/2623

From Encyclopedia of Astronomy & Astrophysics P. Murdin

ISBN: 0333750888

Institute of Physics Publishing Bristol and Philadelphia

Downloaded on Tue Feb 07 18:44:10 GMT 2006 [131.215.103.76]

Terms and Conditions Galaxies: Interactions and Mergers E NCYCLOPEDIA OF A STRONOMY AND A STROPHYSICS

Galaxies: Interactions and Mergers For much of the 20th century—once the vast distances to galaxies were known—galaxies were thought to be ‘island universes’, forming and evolving in isolation with no contact between one another. In this picture, the processes which shape the galaxies we see today are uniquely determined by the initial conditions under which galaxies form and processes completely internal to the galaxies themselves. As galaxy catalogs began to grow, however, more and more examples of paired galaxies were found, as well as many PECULIAR GALAXIES with long, luminous ‘plumes’ and ‘tails’ emanating from their bodies. These galaxies were found to have anomalously blue colors, arguing that their star-forming properties were quite different from normal galaxies. Examples were found of galaxy pairs so strongly perturbed that they were suggested as possible examples Figure 1. A montage of interacting and merging systems. of actual merging encounters. Meanwhile, astronomers Clockwise from upper left: Arp 295; NGC 4676 (The Mice); NGC began to use computer simulation to study the effects of 520; NGC 7252; NGC 3921. Image from Hibbard and van nearby companions on galaxies and found that many of Gorkom 1996 Astron. J. 111 655. the properties of these peculiar galaxies could be explained through gravitational interactions and mergers of galaxies. From these studies, an alternative picture of GALAXY To understand the development of tidal tails, recall EVOLUTION began to grow. Rather than evolving in isolation, that tidal forces—the differential gravitational forces from galaxies are found in clusters and groups and can interact a nearby mass—act to stretch an object radially. In this manner, our own Moon raises bulges on the surface of the quite strongly with their nearby companions (see GALAXY Earth which give rise to our oceanic TIDES. Similarly, when CLUSTERS). These interactions can have a profound impact on the properties of galaxies, resulting in intense bursts of two galaxies experience a close encounter, their tidal fields stretch one another radially. This stretching, combined STAR FORMATION, the onset of QUASAR-like activity in galactic nuclei and perhaps even the complete transformation of with the galaxies’ rotation, causes the stars and gas in the outskirts of each galaxy to ‘shear off’ from their parent SPIRALGALAXIES into ELLIPTICALGALAXIES. Studies of galaxies in galaxies. Material on the far side of each disk—away the early universe show a significant fraction of interacting from the companion galaxy—is ejected into long, thin tidal and merging systems, and theories of cosmological tails, while material on the near side is drawn towards the structure formation indicate that most galaxies have had companion. Depending on the encounter geometry, the some form of strong interaction during their lifetime. nearside material may actually form a physical ‘bridge’ Rather than being rare events, galaxy interactions may be between the galaxies, along which material may flow from the dominant process shaping the evolution of the galaxy one galaxy to another. Because of the coupled effects of population in general. gravitational tides and galactic rotation, the development of tidal features depends strongly on resonances between Dynamics of Interacting Galaxies the rotational and orbital motions of the galaxies. Prograde The evolution of interacting galaxies is governed largely encounters, in which the galaxies’ sense of rotation and by gravitational effects. That these interactions can have a orbital motion are matched, are most effective at tail- profound effect on the participating galaxies is clear from building. The lack of spin–orbit resonances in retrograde the observational record. Morphologically, interacting encounters, on the other hand, acts to suppress the galaxies are found to sport long bridges and tails, stellar formation of tidal tails. bars and/or enhanced spiral structure and often severely The early computer models of Toomre and Toomre distorted main bodies (see figure 1). At first, many demonstrated these basics of building tidal tails and astronomers believed the long streamers of stars and opened the door to the use of computer simulation to gas seen emanating from some interacting galaxies to be study the evolution of interacting galaxies. Because shaped by magnetic fields or nuclear jets—the thinness of the long timescales involved in galaxy interactions— and linearity of these features made a gravitational origin hundreds of millions or billions of years—observations of seem unlikely. However, computer models of interacting interacting systems show only individual ‘snapshots’ of galaxies in the early 1970s by Alar and Juri Toomre a complex evolutionary process. Piecing together these showed convincingly that these streamers were the simple varying snapshots into a coherent sequence is a task consequence of gravitational tides acting on rotating disk made difficult by the unknown initial conditions of the galaxies, and these streamers were dubbed ‘tidal tails’. different interactions. However, computer simulations

Copyright © Nature Publishing Group 2001 Brunel Road, Houndmills, Basingstoke, Hampshire, RG21 6XS, UK Registered No. 785998 and Institute of Physics Publishing 2001 Dirac House, Temple Back, Bristol, BS1 6BE, UK 1 Galaxies: Interactions and Mergers E NCYCLOPEDIA OF A STRONOMY AND A STROPHYSICS can be used to study a wide variety of well-defined would be quite rare, unless the collision happened to be galaxy interactions, and by studying the evolution of these extremely penetrating. models the observational snapshots can be more easily In fact the luminous portions of galaxies represent placed into a evolutionary context. only a small fraction of the total galactic mass—galaxies Computer simulations employ a Lagrangian tech- are embedded in massive, extended ‘dark matter halos’. nique known as N-body modeling, wherein galaxies are These dark matter halos extend to many tens or even represented by N discrete particles whose initial positions hundreds of kiloparsecs, such that encounters where the and velocities sample the phase space distribution of a nor- luminous galaxies seem to pass by one another may in fact mal disk galaxy. Current simulations employ 106–107 par- be penetrating encounters for the dark halos. When this ticles, so that each particle represents 104–105 stars (sim- happens, dynamical friction ensures rapid orbital decay. ulating galaxies star by star is still well beyond current As the galaxies pass through each other’s dark halos, computational abilities). The N-body model is then ad- they set up a trailing wake in the halo mass distributions, vanced forward in discrete time steps by calculating the creating a gravitational drag on the galaxies’ relative net gravitational acceleration acting on each particle from motion. As a result, energy and angular momentum are the other particles and then advancing each particle for- transferred from the binary orbit to the internal motions ward in time given its position, velocity and acceleration of the dark halo—the orbit decays and the halos are spun (for the interstellar gas, hydrodynamic forces must also be up. It is this ability for dark matter halos to absorb considered; these are discussed in a later section). orbital energy and angular momentum that makes galaxy One example of such a computer model of colliding mergers possible; without halos, mergers would be rare galaxies is shown in figure 2. In this model, two equal- indeed (see DARK MATTER IN GALAXIES). mass spiral galaxies are placed on an initially parabolic orbit with a closest approach of 2.5h, where h is the Starbursts and active nuclei in interacting galaxies exponential scale length of the disk stars. One disk is With the launch of the infrared astronomy satellite (IRAS) ◦ perfectly prograde while the second is inclined by 71 to in 1983, a new population of galaxies was identified which the orbital plane. As the galaxies first collide, the tidal have extremely high, quasar-like luminosities (≥1012 L) forces act to distort each disk, launching the tidal tails. in the infrared (8–1000 µm). This emission is believed to Moving at parabolic velocity, the galaxies quickly pass one come from hot (T ∼ 50–60 K) dust reradiating energy another and move apart, reaching maximum separation at from a central starburst or active nucleus. The exact T ∼ 40 (unit time is roughly 10 million years). After this details of the energy generation—starburst versus AGN— point, the galaxies reverse their motions and fall back on remain controversial, and many ultraluminous infrared one another, merging together to form a ellipsoidal merger galaxies (ULIRGs) show evidence for both starburst and remnant around T ∼ 60. AGN activity. The difficulty of course lies in the fact that After first passage, in addition to forming the the nuclei of ULIRGs are, almost by definition, extremely extended tidal tails, the galaxies themselves develop a dusty environments, and seeing into the central regions strong internal response to the interaction. The passage is a very difficult task. Evidence suggests that in some of a companion seeds an m = 2 gravitational perturbation cases the nuclear dust is optically thick even at x-ray in the disk of each galaxy; these perturbations can then wavelengths; finding unequivocal evidence of an AGN in be amplified into strong spiral arms or even dramatic such systems will be very hard indeed. central bars by the self-gravity of the disk. The detailed While the details of the energy source in ULIRGs response of the galaxies depends crucially on their remain elusive, the causal connection between ULIRG internal structure: if the stellar disk dominates the mass activity and galaxy interactions is clear. Nearly all ULIRGs distribution in the inner portion of the galaxy, the disk self- show morphological evidence for strong interactions, gravity can easily amplify the perturbation into a strong through double nuclei, severely distorted isophotes or bar. On the other hand, if the galaxy is bulge or dark strong tidal features. Yet the converse is not true; that is, matter dominated in its interior, the disk is more stable not all strongly interacting galaxies show ultraluminous against growing m = 2 bar modes, and spiral structure is activity. In fact, the most dramatic nearby mergers, the more likely outcome. such as the Antennae (NGC 4038/39) or NGC 7252, What causes the rapid merging of the two galaxies? have experienced violent collisions, yet have infrared Certainly there is a transfer of orbital energy to internal luminosities an order of magnitude less than the extreme motions of the stars in each galaxy—it is this energy ULIRGs. Clearly the detailed triggering mechanism for transfer that provides the energy necessary to launch ultraluminous activity must depend on a variety of factors. the tidal tails. The amount of energy carried away by How do mergers trigger ultraluminous activity? the tidal tails amounts to only a few per cent of the Regardless of whether the central engine is a starburst or total orbital energy, however; if this was the sole energy an AGN, the first prerequisite is a large supply of fuel in sink, interacting galaxies would slowly spiral together the form of nuclear gas. If galaxy collisions can drive over many orbital periods, rather than exhibit the rapid gas inwards from the galaxies’ disks into their nuclei, merging seen in figure 2. Indeed, early investigations into this nuclear inflow can feed the central engine. Once the the collisions of galaxies argued that subsequent mergers gas reaches the inner kiloparsec, it can either fragment

Figure 2. An N-body model of two equal-mass merging disk galaxies. Time is shown in the upper left of each panel, where unit time is ∼10 million years. After ﬁrst collision at T ∼ 24, the galaxies pass by one another before turning around on their orbits and merging to form an elliptical-like object in the last panel. From Mihos and Hernquist 1996 Astrophys. J. 464 662.

− and form stars (at a rate of ∼100M yr 1 to provide the as gas is driven further inwards to the nucleus owing to observed luminosity) or continue to flow inwards and fuel the strong hydrodynamic and gravitation torques at work. an AGN. Because the inflow is tied to the (largely internal) The first step, then, is to drive gas inwards from gravitational response of the galaxies to the interaction, the disk to the nucleus. Computer simulations have the triggering of inflow and nuclear activity depends demonstrated the efficacy of interactions and mergers on a variety of factors, such as the internal structure of at driving these nuclear gas flows. These simulations the galaxies involved and the orbital geometry of the must also describe the hydrodynamic evolution of the collision. Prograde collisions drive inflow and activity interstellar gas, and do so using a Langrangian technique more rapidly than retrograde collisions, owing to the known as ‘smoothed particle hydrodynamics’, in which spin–orbital coupling of the encounter. If the encounter the gas is represented by discrete fluid elements (particles) involves DISK GALAXIES with massive central bulges, the which carry the local thermodynamic and hydrodynamic inflow should be delayed until the galaxies merge, owing properties of the fluid. These properties are updated to the disk stability provided by the central bulges. according to hydrodynamic conservation laws, and Conversely, without bulges, disk galaxies should be more artificial viscosity is included to model shocks. This prone to early bar formation, inflow and central activity. technique is ideal for simulating gas flows in interacting Similarly, galaxies in which the disk contributes little to the galaxies, which lack the symmetry and geometry needed dynamical mass (e.g. low surface brightness disk galaxies) for efficient grid-based hydrodynamical algorithms. will also be more resistive to inflow, as the self-gravity of As an example, figure 3 shows the evolution of the the disk is much weaker. Of course, in any case the galaxies interstellar gas in the merger model of figure 2. Shortly themselves must have sufficient disk gas to drive inwards after the galaxies first collide, and the strong bars and to begin with; collisions of gas-poor spirals will be less spiral features form from the self-gravitating response of effective at fueling ultraluminous levels of activity. the disks, gas shocks and crowds along the leading edge of Once triggered, the ultraluminous infrared phase these features. Because of the offset between the stellar and probably does not last long—less than ∼108 yr. A variety gaseous density peaks, the gas feels a strong gravitational of arguments support this claim. Stellar population torque from the stars, losing angular momentum and synthesis models successfully explain the optical and flowing inwards towards the nucleus. How far inwards infrared spectra of ULIRGs with a burst of star formation it can flow depends on the detailed response of the disk. ∼107–108 years old. Similar numbers come from gas If the disk develops a strong bar, the gas can flow into depletion arguments: if the luminous activity arises from τ ∼ the central kiloparsec on a dynamical timescale; however, star formation, the gas depletion time is given by gas M /M ∼ 7 8 if the disk is stable against bar formation, the gas tends to gas ˙ 10 –10 yr. Of course, if the luminosity comes ‘hang up’ in the inner few kiloparsecs. In this case, once the from accretion onto an AGN, the same amount of gas can galaxies ultimately merge a second phase of inflow occurs, sustain ultraluminous activity for a much longer period

Figure 3. The evolution of the gas in the N-body model shown in figure 2. Time is shown in the upper left of each panel, where unit time is ∼10 million years. After the first collision, gas shocks along spiral arms and flows inwards into the inner few kiloparsecs (T ∼ 48) and then finally into the nucleus itself when the galaxies ultimately merge. From Mihos and Hernquist 1996 Astrophys. J. 464 662. of time. However, the fact that nearly all ULIRG systems This ‘merger hypothesis’ for the formation of elliptical show strong dynamical evidence that they are in the late galaxies idea also had observational support from studies stages of a merger—evidence which fades rapidly once of the peculiar galaxy NGC 7252 (see figure 4) by Fran¸cois the merger is complete—argues that the ultraluminous Schweizer. While this galaxy possesses two gas-rich tidal phase cannot last much longer than a half-mass dynamical tails (indicating a merger of two late-type spirals), it also timescale, ∼108 years. In the latter case of AGN accretion, has the surface brightness profile expected for an elliptical the end of the ultraluminous IR phase may be marked by galaxy. Schweizer argued that we were catching the either an end to AGN fueling or else the destruction of elliptical formation process in the act and that NGC 7252 the ‘dust shroud’ by the intense UV radiation field of the would evolve into a normal elliptical galaxy given time. AGN. In this second scenario, rather than running out of While the merger hypothesis could explain many of gas, the ULIRG may simply evolve into a bona fide optical the qualitative properties of ellipticals, several detailed quasar. The detailed evolution of ULIRGs is a subject of objections were raised. The central phase space density intense current study. of spirals is lower than that of ellipticals; since violent relaxation preserves phase space density,merger remnants Merger remnants and elliptical galaxies should have phase space densities too low to compare Once a merger is complete, and the ULIRG phase (if any) is well with elliptical galaxies. Also, the specific frequency over, what is left behind? In 1977, Alar Toomre suggested of GLOBULAR CLUSTERS (the number of globular clusters per that the remnants of disk mergers could account for the unit luminosity) is much larger for ellipticals than spirals. population of elliptical galaxies in the universe. Because of Finally, because of the high relative velocity of galaxies violent relaxation, the merging process would effectively in clusters, mergers should be less common in these ‘scramble’ the stellar disks, giving the remnant the environments, yet that is where the elliptical fraction is surface brightness profiles and large velocity dispersions highest. How then could mergers of disk galaxies produce characteristic of elliptical galaxies. Toomre argued that the present-day elliptical galaxy population? if the merging timescale is ∼5 × 108 yr, extrapolating the Recent developments have relieved some of these number of nearby on-going mergers (∼10) over the total concerns. While it is true that stellar phase space density is age of the universe (assuming the merger rate scales like preserved during mergers, this constraint does not apply the t 5/3 expected from a flat distribution of binding energy) to the interstellar gas, which can dissipate energy and yields a total of ∼750 merger remnants, similar to the flow inwards. Subsequent star formation can raise the number of elliptical galaxies found in the nearby field. central phase space density in the merger and in principle Furthermore, Toomre argued, if the remnants of these produce an elliptical-like nucleus. Interaction-induced spiral mergers did not constitute the present-day elliptical star formation may also account for the differing globular galaxy population, where are they now? cluster specific frequencies of spirals and ellipticals.

Figure 4. The merger remnant NGC 7252. The image shows the optical light from NGC 7252, while the contours show the distribution of neutral hydrogen gas. Note the gas-rich tails, the loops of starlight surrounding the main body and the smooth, elliptical-like appearance of the inner regions. From Hibbard et al 1994 Astron. J. 107 67.

Recent Hubble Space Telescope observations indicate that survive for many billions of years, indicating a strong galaxy collisions can result in the formation of a large collision in the galaxy’s past. Observationally, a significant number of young, compact star clusters. If these clusters fraction of elliptical galaxies show kinematically distinct can survive the dynamical environment of tidal shocks cores, residual gas and dust, and large-scale tidal features. and mass loss, perhaps they are the progenitors of the old Through studies such as these, it may be possible to globular clusters observed in elliptical galaxies. Finally, constrain the fraction of ellipticals formed through disk the problem of the high relative velocity of galaxies in mergers. clusters can be surmounted by realizing that structure Of course, in a universe where structure forms forms hierarchically. Galaxy clusters are the last objects hierarchically, all objects form from the coalescence of to form; galaxies first assemble into smaller groups which smaller objects. From this point of view, it may be wiser fall together to form a larger galaxy cluster. The galaxy not to ask ‘did mergers form ellipticals?’. but rather ‘what mergers which produce the current cluster ellipticals merged to form ellipticals?’. Observational studies have may have occurred long ago in small groups (with low shown that cluster ellipticals must have formed very early, velocity dispersions), before the cluster formed as a single perhaps even before massive disk galaxies had formed. If virialized structure. so, the merging objects were probably very different from The merger hypothesis thus remains a viable the types of galaxies we see involved in nearby mergers. mechanism for the formation of elliptical galaxies. We Because the collapse time for structure is faster in high- can now ask whether the merging process leaves behind any signatures which can be used to constrain the density environments, cluster and field ellipticals have fraction of ellipticals which formed in this manner. The a different formation timescale, and (under the merger nuclear starbursts should result in the formation of a hypothesis) should have different progenitors. Cluster central population of stars which is distinct in a number ellipticals may form from the rapid assembly of many of ways from the older stars in the original galaxies. small progenitors, while field ellipticals may form through This population may manifest itself through features the more classical merger hypothesis picture of two (‘seams’) in the surface brightness profiles of ellipticals, merging spiral galaxies. These different histories should through strong age and metallicity gradients, and through manifest themselves through differences in the structural, kinematically distinct nuclear kinematics. On larger chemical and stellar population mix between cluster and scales, unless the starburst completely processes the field samples. The observational data, however, do not interstellar gas into stars, a small but significant amount of show as strong an effect as that predicted by models of cold gas may remain in the remnant. In the outer portions hierarchical structure formation. The question of whether of the remnant, the relaxation timescale is long and low most ellipticals were formed through major merger events surface brightness tidal feature—‘shells’ and ‘loops’—can remains open.

Minor mergers and satellite accretion As the satellite is stripped on its orbit, this stripped So far we have focused on so-called ‘major mergers’— material can be incorporated into the luminous halo of the mergers of two large, roughly equal-mass disk galaxies. main galaxy. Over several orbits, this stripped material While these represent some of the most spectacular will gradually spread out along the satellite’s orbital path, collisions we observe, they are far from the most common. making a kinematically distinct ‘tidal stream’ in the halo. Because of the shape of the galaxy luminosity function, In 1978, Leonard Searle and Robert Zinn proposed that, which rises at fainter luminosities, encounters more rather than forming in a single monolithic event, the halo commonly involve a galaxy interacting with a small of the Milky Way formed through continual accretion satellite companion. However, it turns out that even of material much like the infall of satellite galaxies. To these ‘minor mergers’ can have a dramatic impact on the reproduce the observed halo luminosity of the Milky Way L ∼ 9L evolution of galaxies. ( B 10 ), the Milky Way must have accreted on That minor mergers occur frequently can be seen from the order of a few hundred small satellites during its even a casual inspection of the environment of our own lifetime to build the halo in this manner. Whether or not such an evolutionary picture is consistent with the Milky MILKY WAY. The Milky Way is surrounded by 14 known satellite galaxies, the largest and most massive being the Way’s dynamically cold disk (see below) or with the long dynamical friction timescales involved remains unclear, LARGE and SMALL MAGELLANIC CLOUDS. At a galactocentric but a variety of observations argue that some fraction of the distance of 50 kpc, the Magellanic Clouds are believed to Milky Way’s halo—and presumably those of other spirals orbit our Galaxy once every few billion years. The effect as well—formed through satellite mergers. Observations of this interaction is clear: tidal forces from the Milky Way of halo stars show the presence of substructure, moving have torn a long stream of gas—the MAGELLANIC STREAM— groups and possible streams of globular clusters and dwarf from these companions, and it has been argued that some galaxies in the halo. The presence of young A stars in the of the younger Milky Way globular clusters may have been otherwise-old halo also argues for recent accretion events, tidally stripped from the Magellanic Clouds. Recently, the while tidally stripped material from the Sagittarius dwarf discovery of the SAGITTARIUS DWARF GALAXY plunging through ◦ extending out more than 30 from the core demonstrates the galactic disk on the far side of the Galaxy has given explicitly how satellite accretion can feed the Galactic halo. us an up-close view of the tidal destruction of a merging If the satellite is sufficiently dense and massive, its dwarf galaxy. A similar system of companions surrounds accretion can do more than simply build up the stellar halo the nearby Andromeda Galaxy, including the bright dwarf of a galaxy. Such satellites more readily survive into the elliptical galaxies NGC 205 and M32. Indeed, if models inner galaxy; once there, they can strongly influence the of structure formation in the universe are to be believed, dynamical evolution of the disk. An N-body simulation all bright galaxies should have a large number of satellite of such an accretion event is shown in figure 5, where a companions. prograde satellite with mass 10% that of the disk merges Unlike major mergers, where dynamical friction is with a larger disk galaxy. As this massive satellite plunges so efficient that the galaxies merge after only a few through the disk, it scatters disk stars off their orbits, > perigalactic passages, the extreme mass ratio ( 10:1) heating and thickening the stellar disk. As the satellite of minor mergers ensures that the orbital decay of the galaxy falls to the plane of the disk through dynamical satellite’s orbit is slow. For example, it is estimated that friction, conservation of total angular momentum dictates the orbit of the Magellanic Clouds will decay in another that the disk warp and tilt in response. The resulting ∼ ∼ 10 Gyr, giving the Clouds a total ‘survival time’ of 10 warped, thickened disk is qualitatively similar to those orbital periods. Rather than the sudden violence of a major seen in some nearby galaxies (such as NGC 3628) and merger, these minor mergers are more properly a relatively inferred in our own Milky Way. slow accretion event, and the dynamical evolution of these As a massive satellite spirals into the inner disk, systems is markedly different. In particular, the tidal it can also induce strong spiral arms and bars through field of the host galaxy will act to gradually strip material a resonant response in the disk. Interstellar gas can from the orbiting companion, such that the merger may be compressed along these features and flow inwards not play to completion—the companion may be totally towards the nucleus owing to gravitational torquing from destroyed before reaching the inner portion of the host the bar and arms, similar to, but at a more modest level galaxy. Since the tidal radius of a orbiting companion than, the major-merger-induced nuclear inflows. Thus / is roughly proportional to (ρs /ρG)1 3 where ρs and ρG these minor mergers can also drive increased disk and are the average mass density of the satellite and host nuclear star formation rates or activate a central engine. galaxy, respectively, it is the low-density, diffuse satellite When the merger is complete, a significant fraction of companions which are preferentially destroyed as they the disk gas may have been swept from the disk and orbit their host. Low-density DWARF IRREGULAR and DWARF into the nucleus. The disk itself has been kinematically SPHEROIDAL GALAXIES will certainly be tidally disrupted at heated, and spiral structure destroyed, by the passage of relatively large distances from their hosts, while compact the companion through the inner disk. The induced bar companions (such as Andromeda’s M32) may survive the may buckle in three-dimensional space, scattering stars accretion process well into the host. above and below the disk plane, in a pseudo-bulge. The

Figure 5. An N-body model of a large disk galaxy accreting a small satellite companion. The companion is initially on a circular orbit ◦ around the primary, inclined 30 from the disk plane. Top row shows a face-on view of the accretion; bottom row shows an edge-on view. As the satellite falls into the inner regions of the host, it excites strong spiral arms and a central bar, and thickens and warps the disk. Note also the stripping of material from the companion as it is accreted. From Hernquist and Mihos 1995 Astrophys. J. 448 41. morphological, kinematic and star-forming properties of Sanders D B and Mirabel I F 1996 Ann. Rev. Astron. the disk have all been radically altered by this interaction, Astrophys. 34 749 suggesting that even minor mergers can drive significant evolution in disk galaxies. The theoretical motivation behind the merger hypothesis for elliptical galaxy formation can be found in Bibliography Toomre A 1977 The Evolution of Galaxies and Stellar The subject of galaxy interactions covers such a wide range Populations ed B M Tinsley and R B Larson (New of topics that it is impossible to give a comprehensive Haven, CT: Yale Observatory Press) p 401 literature review. Instead, I have attempted to outline some of the original material in the field, while also giving while some of the first observational groundwork is in more recent reviews on the broader subjects. Schweizer F 1982 Astrophys. J. 252 455 Recent reviews on the subject of galaxy interactions include The accretion model for the galactic halo was originally proposed by Barnes J and Hernquist L 1992 Ann. Rev. Astron. Astrophys. 30 705 Searle L and Zinn R 1978 Astrophys. J. 225 357 Kennicutt R C, Schweizer F and Barnes J E 1998 Galaxies: Interactions and Induced Star Formation while a recent review of observational and theoretical (Berlin: Springer) work on the structure and formation of the halo is given in Some of the first detailed N-body calculations of interacting galaxies were done by Majewski S R 1996 Ann. Rev. Astron. Astrophys. 34 749

Toomre A and Toomre J 1972 Astrophys. J. 178 623 Models of satellite infall can be found in while more modern calculations can be found in Hernquist L and Mihos J C 1995 Astrophys. J. 448 41 Johnston K V,Spergel D N and Hernquist L 1995 Astrophys. Barnes J and Hernquist L 1996 Astrophys. J. 471 115 J. 451 598 Mihos J C and Hernquist L 1996 Astrophys. J. 464 662 Walker I R, Mihos J C and Hernquist L 1996 Astrophys. J. 460 121 The detailed properties of ULIRG galaxies, and the connection between mergers, ULIRGs, and quasars was C Mihos ﬁrst described in

Sanders D B et al 1988 Astrophys. J. 325 74 and have been recently reviewed in