Despite much initial skepticism, the general notion that mergers are a major driving force in galaxy evolution has become widely, almost universally, accepted. In fact, it has become the dominant theme in the field in the last decade. Not only have there been numerous simulational and observational papers published, but there have been a number of major review articles published as well. For example, the interacting galaxies review of Barnes and Hernquist (1992) focusses heavily on mergers and the numerical techniques used to simulate them. Schweizer's (1983, 1986, 1990) reviews cover the observational study of merger remnants. Hernquist (1993) reviews the increasingly detailed efforts to compare models to possible elliptical merger remnants. Barnes (1995, also Barnes and Hernquist 1996) summarizes simulational studies of gas dynamics in mergers. Recent updates can be found in the reviews of Barnes (1996), Bender (1996) and the articles in the Saas Fee lectures edited by Friedli et al. (1998).
With so many excellent reviews available a detailed examination of this area does not seem necessary. Thus, the discussion here will be brief. (Similar comments apply to the following chapters on induced SF and nuclear activity.) As in the previous chapters I will try to focus on the important physical processes. Moreover, merging is in large part a repetition and elaboration of the processes described above, so we do not start from scratch.
We will concentrate on a few key issues, including the following questions. How fast are mergers? What are their immediate effects on the galaxies involved? How do these effects differ from those experienced in nonmerging collisions? What kind of remnants do they leave, and are there specific observable signatures of a past merger? How common are mergers?
6.1. Overview and Historical Highlights
Toomre's (1977) paper, discussed in section 1, would have been a landmark even without the mergers-make-ellipticals conjecture presented there. The extension of his earlier work to the conclusion that most collisions end in merger, because of the operation of dynamical friction, was pivotal in itself. The N-body and N-ring simulations he presented showed two galaxies combining with truly shocking rapidity, though "N" was very small by modern standards. This evidently led Toomre to search the galaxy atlases for examples, and good candidates were not hard to find. As discussed in section 1.3 the merger elliptical idea initially seemed too speculative for many astronomers, and there was similar skepticism about the idea of rapid merging. The field of N-body modeling was still quite young, but within a few years after Toomre's paper quite a number of simulational studies had been published.
In this first epoch of N-body merger simulations the galaxies were generally modeled by a single (spheroidal or halo) component, with typically a few thousand particles. Listings and brief descriptions of the original papers can be found in the reviews of Barnes and Hernquist (1992a, Hernquist 1993), and in the introduction to their recent paper (Barnes and Hernquist 1996). In most cases the collision partners were assumed to be of equal or near-equal mass. The primary result of these models was confirmation that galaxy collisions from initially nearly bound orbits do indeed generally result in rapid merger. This was hard to understand at first because two-body relaxation times are very long, and two-body collisions are the basis of Chandrasekhar's dynamical friction equation. But these models revealed that the transfer of orbital energy to internal motions was a global, not a local, phenomenon (see sec. 5.1). Moreover, though the fraction of mass lost was found to be small, this material carries away significant energy and angular momentum. This is also true of material which is not lost, but flung out to large radii in tidal tails.
These models also showed that the merger remnant relaxed rapidly to a quasi-steady state, which was at least qualitatively similar to galactic bulges or elliptical galaxies. This was interpreted to be the result of "violent relaxation" described below. The early simulation papers contain detailed discussions of mass loss, radial mixing, and rotational and structural properties of the remnants as a function of initial parameters. However, it was hard to know how far the comparison of these single component models to observation could be taken.
The second epoch of simulational studies can be defined by the publication of two component (disk-halo) N-body models (e.g., Gerhard 1981, 1983a, b, Farouki and Shapiro 1982, Barnes 1988, 1992), and the first attempts to simulate gaseous dissipation in collisions and mergers (e.g., Negroponte and White 1983, Barnes and Hernquist 1991). Since model galaxies with a dynamically cold disk are generally assumed to be stabilized against the bar instability by a hot halo, two components are needed in simulations with disks. Such simulations were required, in turn, to test the theory that an elliptical could be formed from the merger of two disk galaxies. Much larger numbers of particles are required for such simulations, and so, their development has been closely coupled to improvements in computer hardware and software. Because the technology developed so rapidly during this time, N-body simulations now contain four to five orders of magnitude more particles than in the early 1980s.
Second epoch models revealed more complexity in both the merging process and in merger remnants. A first example is in the orbital deceleration. As described in section 5.3 above, massive halos merge first, and the merging of the denser disks and bulges is facilitated by the binding of the large halo. As expected, the dynamical heating in this process forges remnants of much earlier Hubble type than the progenitors, i.e., elliptical-like. However, as will be discussed below, the memory of the initial structure and orbit are not immediately erased by the heating and relaxation processes. The gas dynamical models of this period produced another important result, that a large fraction of the interstellar gas can be funneled deep into the core of the merger remnant. Bar-driven flows of the type described in sec. 5.6 are often an important part of this process.
There is no clear boundary between second epoch and current models, though we can arbitrarily mark it by the appearance of three component (bulge-disk-halo), gas-plus-star models (e.g., Barnes and Hernquist, 1996). While "third epoch" models have not produced as many major new results as earlier ones, many new avenues are being explored. These include, for example, more extensive studies of collisional bars and their interaction with other components of the merging galaxies (see sec. 5.6), and more sophisticated attempts to model the feedback effects of starbursts (see chapter 7). Large particle numbers have also just simply allowed the production of better models for specific mergers, like Barnes' (1998) model for the Antennae system.
In conjunction with these developments in models and theory, there has been great progress in the observational studies of the merger process. These were pioneered by F. Schweizer, foremost among the observers who took up Toomre's challenge to find the evidence for mergers. Schweizer's (1983 also 1998) review, already referred to in section 5.5, summarizes observational indicators of merging and early work on merger remnants. This paper begins with a discussion of cD galaxies, the supergiant galaxies that are obvious merger candidates. This is because they have huge stellar envelopes, and generally reside in clusters, so the chance of a cluster galaxy being captured via dynamical friction are good (Ostriker and Tremaine 1975). In fact, cD galaxies are probably made by a succession of such mergers ("galactic cannibalism"). Schweizer cites two other "promising pieces of observational evidence of mergers in cD(s)" - the presence of multiple cores and asymmetric envelopes. Both are nonequilibrium features and would be expected to disappear on timescales much less than the age of the galaxies. Moreover, neither feature is the exclusive property of cD galaxies.
Schweizer not only provides examples of galaxies with multiple cores, but also of galaxies with gas disks that do not rotate at the same rate as the coextensive stars (indicating a relative tilt), and of galaxies apparently containing two gas disks. Again these are not equilibrium structures. Since that time many more multiple core, and "counter-rotating" disk systems have been found. Schweizer also argued that the polar ring galaxies (discussed below) might be the result of accretion or merger. Finally, and most importantly, there is the presence of large scale "tidal" features: ripples, shells, tails, etc., around many merger candidates (see Figure 22). In many of the systems Schweizer reviewed there is no obvious alternative to a galaxy collision and merger origin for these morphologies.
Figure 22. Multiwavelength observations of four famous merger remnants: 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). Contours show the surface density of neutral hydrogen gas superposed on optical images (greyscale). The insets show K band (2.2 micron) images of the central regions as greyscale, with white contours representing molecular gas (CO) intensities. The scale-bar represents 20 kpc in each case. See Sanders and Mirabel (1996) for details.
While Schweizer pursued detailed studies of individual systems, statistical studies were beginning to provide evidence for enhanced "activity" in colliding and merging galaxies, in the form of star formation and nuclear activity, especially in optical colors and radio continuum emission (also see references in the following chapters). However, the IRAS (Infrared Astronomical Satellite) mission in 1984 brought observational studies of merger remnants into a new era. IRAS's whole sky survey in the far-infrared (with passbands centered at 60 and 100 microns) revealed numerous ultra-luminous infrared galaxies (ULIRGs or ULIGs (1)), many of which were soon determined to be merger remnants or other types of collisional galaxy (e.g., Aaronson and Olszewski 1984, Houck et al. 1984, 1985, Lonsdale, Persson and Mathews 1984, Wright, Joseph and Meikle 1984, Soifer et al. 1984a, b and Joseph and Wright 1985). Given the results on central activity in merger remnants, and the first inklings of the process of gas funneling from simulations, discovering hot dust was not a surprise. However, the magnitude and the extent of the phenomenon, i.e., that this class contains the brightest galaxies known, was a shock.
These discoveries enormously energized the field, generated many new observational and theoretical studies, and eventually suggested the possibility that the activity generated by mergers might be responsible for most of the star formation (and thus metal production), and nuclear (quasar) activity in the universe (e.g., Sanders et al. 1988, and the reviews of Soifer, Houck, and Neugebauer 1987, Sanders and Mirabel 1996, and the conference proceedings edited by Persson 1987, and Sulentic, Keel and Telesco 1990). This proposition remains speculative since, even if most ordinary galaxies experienced "major" mergers in the past, the traces are hard to find now. However, the more limited proposition that - almost all of the most infrared luminous objects (with LIR > 1012 solar luminosities) are active merger remnants has been quite firmly established in the time since IRAS. This story is told in detail in the recent review of Sanders and Mirabel (1996). Reaching this point has been an arduous process of: optical identification of IRAS sources, obtaining redshifts to determine distances, and obtaining observations in many other wavebands to understand the nature of the sources. Yet, by now many lines of evidence support the idea that these objects are generally,
"advanced mergers powered by a mixture of circumnuclear starburst(s) and active galactic (nuclei)...fueled by an enormous concentration of molecular gas that has been funneled into the merger nucleus." (Sanders and Mirabel 1996).
New high-resolution observational studies, like Scoville, Yun, and Bryant's (1997) recent study of molecular gas in Arp 220, are indicative of the type of advances we can expect in coming years.
While much observational effort in the last decade has focussed on merger ULIRGs, optical and radio studies of nearby (non-ULIRG) merger systems have continued to advance. This includes the work of Schweizer and Seitzer (1992), who used high quality optical images of faint tidal features to estimate the relative ages of merger remnants in a large sample (see section 2.2). Schweizer and his collaborators also presented an important series of papers detailing the star formation properties of several well-known merger remnants, using both ground-based and Hubble Space Telescope imaging (Whitmore et al. 1993, Schweizer and Seitzer 1993, Whitmore and Schweizer 1995, and Schweizer 1996). This work will be described in the following sections.
At the same time, Hibbard was studying an evolutionary sequence of five merger remnants for his thesis work (Hibbard 1995, also see Hibbard et al. 1994, Scoville et al. 1994, Schimonivich et al. 1995, Hibbard and van Gorkom 1996). The youngest two systems in this sample are yet not fully merged (Arp 295 and NGC 4676, vs. the older systems NGC 520, NGC 3921 and NGC 7252). He used the VLA to map the HI distribution, and also carried out H-alpha and R band imaging in all five systems. With these data he carried out the most complete analysis to date of the evolution the gas component in the merger process. In the systems he studied a significant fraction of the gas remained in an atomic phase in long (but bound) tidal arms or tails through the merger process. Hibbard makes the interesting point that these systems and others like them are generally not ULIRGs. Thus, it appears that the ULIRG phase is either short-lived and under-represented among nearby mergers, or there are at least two classes of merger (see the contrasting examples in Hibbard 1997). With the combined work of Hibbard and Schweizer (1996 and references therein), a great deal of high-resolution multi-waveband data is now published for the NGC 3921 and NGC 7252 systems, which provides much evidence in support of the Toomre's conjecture that both systems are well on their way to becoming elliptical galaxies.
Another parallel strand of merger research is the study of shape and kinematic profiles of elliptical galaxies, especially their cores, and comparison of these to merger simulations in the hope of finding unique and long-lived signatures of the merger origin of ellipticals. "Boxy" isophotes are a first example. These are defined as rather square-shaped surface brightness contours, that contrast with the usual spheroidal profiles. Bender (1990 and references therein) has provided evidence that boxy ellipticals are intermediate age merger remnants. A second example is the presence of a "kinematically decoupled core" in an elliptical. That is, the spin axis of the core is at a large angle relative to the spin axis of the bulk of the galaxy. Counter-rotating cores are extreme cases. Bender (1996 and references therein) states that "more than 50% of all luminous ellipticals contain kinematically decoupled cores, and these galaxies are found in all environments. In general, the arguments that these cores result from some kind of merger are strong. It also seems that they can have a long lifetime, and so they may provide the best long-time signature of a merger yet discovered. More subtle kinematic indicators, like the misalignment between the core spin axis and the minor axis of a flattened (but slowly rotating) elliptical may also be useful in future (Barnes 1992, 1998, Hernquist 1993). Ultimately, the reason for the existence of such structures is the fact that violent relaxation, and radial mixing in particular, do not go to completion in the merger process.
This concludes our general history and overview of mergers. In the following sections we will look at a few specific aspects of the process in more detail.
6.2. Major Merger Dynamics
In this section we will consider some of the physical processes that characterize "major mergers". This term, major merger, has been adopted in recent years to describe mergers between nearly equal progenitors, which have a major effect on both galaxies. "Minor mergers" involve a significantly smaller companion, and so the primary galaxy is not highly disrupted. Note that major mergers do not necessarily involve large or massive galaxies, nor do minor mergers involve only small galaxies.
6.2.1. Violent Relaxation
Violent relaxation is a wonderful oxymoron! The basic idea is that large amplitude fluctuations in the gravitational field, as in galaxy formation or collisions, drive a relaxation process that is much faster than the two-body relaxation time due to star-star encounters with a galaxy. The statistical formulation and a derivation of the equilibrium stellar phase space distribution was originally given by Lynden-Bell (1967, also see BT sec. 4.7). Specifically, he demonstrated that violent relaxation could lead to an equilibrium (e.g., Maxwell-Boltzmann) distribution function on a relatively short timescale. Tremaine, Hénon, and Lynden-Bell (1986, also see Kull, Treumann, and Béhringer 1997) later investigated constraints on mixing and the relaxation process using H-functions. They begin by noting that the equilibrium state that violent relaxation drives stellar systems towards, has infinite mass. Since galaxies do not, violent relaxation cannot go to completion, and "potential variations die away before relaxation is complete". They find that the remnants of this process can only resemble real elliptical galaxies if the initial state is cold or clumpy, e.g., far enough from equilibrium in the interaction environment to allow for considerable relaxation. This turns some of the old objections to forming ellipticals from dynamically cold spirals on their head.
While "incomplete violent relaxation" is a bit difficult to visualize, the theory is very helpful for understanding the merger simulations and observational properties discussed in the previous section. Specifically, it helps explain why even early simulations found rapid relaxation to an elliptical-like surface density profile, in conjunction with modest radial mixing. On the observational side, it explains why stellar surface density profiles typical of quiescent ellipticals (e.g., King or de Vaucouleurs profiles, see BT) are found in the presence of multiple cores and tidal structures. The discovery that dynamical friction is the result of a global response, and can be quite impulsive (see section 5.1), suggests that it is closely related to violent relaxation. This in turn helps understand the short timescale of merging.
Another aspect of the conjoint global response of violent relaxation and dynamical friction is a result emphasized by Barnes (1992), and which dates to Farouki and Shapiro (1982). This is that orbit shapes, and orbital decay are primarily determined by one parameter - the ratio of "pericentric separation" to the galactic half-mass radius. Note, however, that this result applies to mergers between equal-mass galaxies on initially parabolic orbits. Physically it is probably a result of the fact that in these cases halos merge promptly, and constrain subsequent evolution. In minor mergers with small companions other parameters are also important.
To approach the topic from aslightly different angle, consider the recent paper of Chavanis, Sommeria, and Robert (1996), which revives an old analogy between the equations of two dimensional turbulence (Euler eqs.) and stellar dynamics (Vlasov eq.), and compares them in detail. The authors reproduce Lynden-Bell's equilibrium distribution function using a mixing entropy function and the Principle of Maximum Entropy Production. They also derive a Fokker-Planck equation for the coarse-grained distribution function to describe violent relaxation. In both turbulence and stellar relaxation there is a tendency to "develop finer and finer filaments", though in the stellar case these filaments lie in the six dimensional phase space. In the turbulence case the relaxation also proceeds in two stages, the first is rapid, like violent relaxation, and leads to the formation of large coherent structures. The final viscous relaxation to the true equilibrium takes much longer, like two-body relaxation.
6.2.2. Gas Funneling
Because the gas disk is usually more extensive than the stellar disk in disk galaxies, and thus has higher mean specific angular momentum, it is reasonable to expect that more would be lost or swung out to large radii in tidal collisions. Thus, it is surprising how much gas can be funneled to the center of some merger remnants, and there provide fuel for high levels of activity. On the other hand, interstellar gas is highly dissipative. Large-scale shocks are not only good at radiating energy, but also can transport angular momentum efficiently. In several sections above we have described the mechanisms that simulations have shown affect this transport and the gas funneling. Foremost among these are collisionally induced bars (section 5.6) and tidal spirals, including both internal waves and extended tidal tails (sections 3.5 and 5.4).
While the numerical simulations demonstrate the effectiveness of these processes (see references in the previous section), to date no simple physical or analytical models have been produced which are capable of predicting the amount of funneled gas, or how it scales with collision parameters. This is understandable since very high-resolution, multi-component simulations with both gas and stars are required to study the phenomenon, and be confident of the accuracy of the results. Moreover, an extensive exploration of the high-dimensional parameter space will be required.
Before leaving this topic we should note that the existence of the funneling process resolves an important problem in merger theory. That is, how to explain the very dense cores found in some ellipticals (see e.g., Faber et al. 1997). As Hernquist (1993) points out, mergers between purely stellar disk galaxies do not generally produce such dense cores, unless they have dense bulges to begin with. However, the dissipative gas is not bound by the same fundamental phase space constraints as collisionless stellar systems. Thus, as observed, funneling can build large central densities, and starbursts can convert the gas to a dense stellar core, though realistic simulations of this process have not yet been carried out.
6.3. Minor Mergers: Disk Heating and Aging
Although, as discussed in section 5.2, there has long been interest in the "sinking satellite" problem, there has been much less simulation work than on mergers between equal mass galaxies. The observational signatures of minor mergers are much weaker than those of major mergers, so observational comparisons are harder. Moreover, there are technical difficulties in the simulations, especially in adequately resolving a companion that is much smaller than the primary.
On the other hand, minor mergers are likely to be much more common than major ones. Hernquist and Mihos (1995) summarize the evidence for this conclusion, and provide a listing of earlier works. At first this conclusion seems like a relatively simple matter. Galaxies have a wide range of masses (over 4-5 orders of magnitude), so it would seem unlikely that collision partners have nearly the same mass. However, the companion to primary mass ratio probably has to about 0.1 or less to make the merger "minor", and more than about 0.01 to have substantial consequences for the primary. We can make a simple estimate of the relative number of companions in these mass ranges based on the field galaxy luminosity function, but even this is dangerous. Collision partners are not selected randomly from the luminosity function. They are generally members of bound groups or of groups interacting with other groups or clusters (see chapter 9). Thus, hierarchical clustering biases the statistics of collision partners.
The work of Zaritsky et al. (1993b), discussed above in connection with the Holmberg Effect, provides especially useful input on this subject. In this study of satellites around late-type spirals the authors find mean of 1.5 satellites per primary, and of course, there may have been more in the past, which have already merged. The mean mass of these satellites is estimated to be about 10%, exactly the right magnitude for minor mergers. The luminosity distribution of all the satellites is well fit by a Schechter form, like the field luminosity function, which shows that at least they are not an unusual population. The number of objects in the survey (69 satellites) is not large enough to definitively answer the questions above, but it does support the opinion that they are common. So too does the increasingly common discoveries of mild signatures of old collisions or mergers in otherwise normal galaxies.
Despite the difficulties, several simulational studies of minor mergers have been published recently. These papers (Mihos and Hernquist 1994a, Hernquist and Mihos 1995, and Walker, Mihos and Hernquist 1996), present the results of fully self-consistent, high particle number, multiple component simulations. These simulations use about a 10% mass companion, on a prograde orbit of modest inclination (about 30°), with a companion density comparable to the mean of the primary disk. Thus, these models are a continuation and update of the sinking satellite studies described above. As in the earlier studies, the response is global, and the merger is prompt. Somewhat surprisingly, the results of these simulations are qualitatively similar to those of major merger models.
The Hernquist and Mihos (1995) simulations, in particular, show strong gas funneling. Up to almost half of the primary gas mass can be deposited in a dense core, plausibly inducing starbursts and other nuclear activity. This is comparable to the funneling in major mergers, which is surprising because the collisional distortions are far less, and no strong bar is formed. Instead, another now familiar mechanism seems to be the cause. Strong spiral arms are induced by the interaction, and Hernquist and Mihos (1995) argue that gravitational torqueing by these waves drives the radial gas flows. There is an interesting wrinkle however. The authors suggest that strong shocks form in the waves, and that the dissipation in these shocks yields a positional offset between the gas and the stars in the wave. This offset provides a lever arm with which the stellar wave can torque the gas. Thus, the offset is seen as the key to the strong radial flow. For the present these new results should be viewed with caution, since they are based on an isothermal equation of state. For example, heating by star formation in the waves could stir the gas, yielding a smaller offset, and less torque.
Hernquist and Mihos (1995 and references therein) also confirm a result hinted at in earlier works, the radial gas flows can be delayed or inhibited by the presence of a compact bulge in the primary. They were not able to derive a specific mechanism for this effect, but they conjecture that it may be related to the presence of inner Lindblad resonances. On the basis of the studies of induced bars described in the previous chapter, this seems very reasonable.
The paper of Walker et al. (1996) presents the most detailed study to date of another important consequence of sinking satellite minor mergers, dynamical heating of the stellar disk. The parameters of the 500,000 particle simulation presented there are essentially the same as those of Hernquist and Mihos. Over the 1.0 Gyr. course of the merger the stellar disk thickens by 60%. Velocity dispersions are increased in all three directions. The net result is that the merger remnant is a disk galaxy of substantially earlier Hubble type.
This result supports Schweizer's (1986, 1990) general conjecture, that minor mergers can push late-type progenitors along the Hubble sequence towards earlier types. This idea is a corollary of Toomre's mergers-make-ellipticals conjecture, and many of the same observational techniques are useful for investigating it. For example, in terms of tidal remnants, it has been clear from early on that many shell galaxies (sec. 5.5) were S0, rather than elliptical galaxies. Similarly, many polar ring galaxies are of relatively early type, except for the ring (see Schweizer 1986, and the following section). Counter-rotating disks or cores are also found in objects of type S0 and Sa. The probable merger or accretion event which produced these had to be minor enough to preserve the disks.
Another general point about minor mergers is that, since the outcome depends on both the structural and orbital parameters of the companion, they have the potential to amplify the diversity of galaxies. This raises the question, if minor mergers are common, then why haven't they disrupted structural relationships among the galaxies? This complex question may in fact have a fairly simple answer. Most of the increased diversity may be represented by relatively faint fossil collisional structures around the primary, while the effects of global or deeply penetrating perturbations rapidly relax to more generic forms. Processes like violent relaxation and dynamical friction evidently constrain evolution to approximate the fundamental relationships.
In sum, though minor mergers have been somewhat neglected relative to their "major" siblings, there are many motivations for future study, and recent work demonstrates the feasibility of such studies.
6.4. New Disks
We have seen that major mergers can destroy galaxy disks, converting the progenitors into ellipticals, and that minor mergers can heat and age disks. But there is another side to this subject, minor mergers with companion disruption can lead to the formation of new, or reinvigorated disks. Since we have already considered both companion disruption and minor mergers, the goal of this section is simply to add a few missing pieces to the picture.
6.4.1. Disks in Ellipticals
Until about the mid-1970s there was little evidence for a cool gas, disk component in ellipticals, but with more sensitive instruments this situation has changed greatly (see Bregman, Hogg, and Roberts 1992 and Macchetto et al. 1996 for survey statistics). Moreover, as noted above, kinematically decoupled cores have been found now in a high fraction of ellipticals. These core disks are discovered from their H emission or from dust obscuration (see the review of de Zeeuw 1994). However, they are not usually very blue, but rather normal photometrically relative to the cores of all early-type galaxies (Carollo 1997, Carollo et al. 1997). There are other possible sources for cool gas in ellipticals, e.g., stellar mass loss or cooling flows out of hot cluster gas. However, disks resulting from these sources would be expected to align with the kinematic axis of the galaxy. On the other hand, material from accretion or (minor) mergers is thought to settle promptly into a disk whose orientation depends on the spin of the progenitor and the orbital parameters. (See Barnes 1998 on the formation of inner and outer disks in major mergers, and Thakar and Ryden 1996, Struck 1997 on mass transfer.) These disks then evolve to a preferred plane of the host, but on a longer timescale (see de Zeeuw 1994 and references therein). Nonetheless, though the argument seems straightforward, it is still based on very circumstantial evidence in most cases.
An HST study of the brightness profiles in the cores of a sample of 61 early-type galaxies finds that they can be divided into two distinct types: "core" galaxies with steep "power-law" profiles that break to a much flatter form within a core radius, and shallower power-laws with no resolvable changes into the center (Lauer et al. 1997, Faber et al. 1997). The latter type is most easily explained as the result of a merger of a small, gas-rich companion. However, disks were not detectable in most of the power-law objects in this sample, but many new searches for HI and molecular gas in elliptical galaxies are underway (van Gorkom and Schimonivich 1997, Rupen 1997).
While kinematically decoupled cores are common in large, luminous ellipticals, a related phenomenon is at least as common in low luminosity ellipticals. This is the presence of flattened, pointed or "disky" isophotes. Galaxies containing them are called "disky ellipticals", and are usually also rotationally flattened, independent of the disky part. the basis for this statement is the fact that photometric and kinematic decompositions of disk and bulge parts, like those derived for disk galaxies, seem to work well on these objects (see Scorza and Bender 1996, Bender 1997). The derived disk-to-bulge luminosity ratios (D/B) overlap and extend the low end of the range for S0 galaxies, indicating continuity across the types. The disky ellipticals also extend the disk galaxy trends in plots of bulge or disk luminosity versus D/B. Thus, there are several indications that these galaxies represent an extension of the Hubble disk sequence.
Do the disks and the high rotation rates of these galaxies result from mergers? Scorza and Bender point out that merger simulations with gas do lead to changes in stellar orbit families in such a way as to make a more oblate remnant, which is encouraging for the merger theory. However, the discovery and study of these objects is very recent, and there is far too little data for firm conclusions.
Ellipticals with extensive cool gas disks, extending beyond the optical disk, are still very rare (van Gorkom 1992). Thus, if mergers of spirals make ellipticals, then they are indeed efficient about consuming or heating and dispersing the cool gas of their progenitors. However, the nearby giant elliptical, NGC 5128 = Cen A, with its large gas and dust disk, and evidence for a recent merger (Ebneter and Balick 1983), stands as an apparent counter-example. More likely, however, the elliptical was formed long before the recent merger. (For other such examples see the discussion of section 4.1, and the papers of Appleton 1983 and de Mello et al. 1995, 1996.) The work of Whitmore et al. (1997) on what may be dynamically young ellipticals provides especially interesting "archaeological" support for these ideas.
6.4.2. Counter-rotating Disks in S0 and Sa Galaxies
There are also numerous examples of counter-rotating disks among early-type spirals. E.g., Thakar and Ryden 1996 list a number of recent references for S0s. Bertola, Buson, and Zeilinger (1992) estimated that about 40% of S0 galaxies contain ionized gas of external origin on the basis of a small survey. Very recently, Lovelace, Jore, and Haynes (1997) have provided a list of (primarily) Sa type galaxies with extensive counter-rotating disks. For a review, history of the subject, and summary of several individual systems see Rubin (1994b). As in the case of the ellipticals, the most likely explanation for their existence is accretion or merger.
What is most remarkable about these disks is that, in contrast to counter-rotating cores in luminous ellipticals, they are extensive, sometimes as large as the parent disk. They can also contain significant mass, up to 20-50% of the disk. This makes it difficult to account for them in any way except as the result of a merger. NGC 4550 in Virgo is the most famous S0 example, see Rubin (1994b). A later type (Sab) example is the so-called "Black-Eye" or "Evil-Eye" galaxy, NGC 4826, whose gas disk is nearly co-planar with the stars, but has very complicated kinematics. It rotates with the stars in the inner regions, but reverses direction in the outer disk (see Rubin 1994a, Walterbos, Braun, and Kennicutt 1994). The prominant dust lane (the black-eye) lies in the transition region.
Another example, is NGC 4138, an Sa galaxy studied at multiple wavelengths by Jore, Broeils and Haynes (1996). In this latter system 20% of the disk stars and all of the gas rotate counter to the majority of the stars. The counter-rotating HI gas extends to a radius of 2.5 times the outer radius of the stars. The smooth distribution of the components and Balmer absorption lines indicate that the SF stopped about 108 yrs. ago, giving some constraints on the age of the counter-disk.
This raises the question of how stable and long-lived are these disks? Rubin (1994b) notes that this question has been considered for several decades, i.e., before the existence of oppositely rotating disks was demonstrated! The answers she records are quite varied. The latest input is the work of Lovelace et al. (1997), who summarize recent work and suggest that the oppositely rotating disks can drive density waves, especially the one-armed mode. These density waves increase the dissipation and effective viscosity in the gas leading to accretion.
The other obvious question - how do these extensive counterotating disks form - is addressed by Thakar and Ryden (1996). They modeled counter-disk formation as a result of both continuous and episodic infall. The former could be infall out of a stream, like the Magellanic stream around the Milky Way. The latter could result from the merger of a gas-rich disk, for example. The key result of these simulations is that the infall rate must not be too rapid, or the sudden introduction of a large amount of mass leads to excessive heating of the pre-existing disk. This constraint becomes a real difficulty in producing massive counter-disks. Thus, the authors argue that it is unlikely that such disks are produced by the merger of a gas-rich dwarf. This is a surprising result, worthy of further study.
6.4.3. Polar Ring Galaxies
In the present context, these objects are viewed as another kind of accreted disk. However, as their name implies they are generally oriented perpendicular to the main disk, rather than being contained within the same plane, e.g., Figure 23. Moreover, it believed that they are usually annular disks, with empty middles, though in many cases this is hard to determine observationally, see Figure 23 and the schematic examples of Figure 24. This may be because they are polar and often contain gas; gas cloud collisions with clouds in the primary disk would remove the inner part of a complete polar disk. Moreover, if they do indeed result from accretion in most cases, the material may have a relatively large specific angular momentum.
Figure 23. A well-known polar ring galaxy, NGC 4650A. (Image courtesy European Southern Observatory.)
While they are the third type of "ring" galaxy we have considered they are clearly unrelated to either collisional ring waves or resonance rings within disks. The primary or host galaxy typically has the characteristics of an S0 type, with relatively little gas or dust (again in contrast to the other types of ring). This seems to confirm the idea that polar rings generally orbit through the poles of an oblate galaxy, rather than around the equator of a prolate one. The polar ring is usually gas-rich and bluer than the host, with little or no old star component. Whitmore et al. (1990) have assembled a catalog with more than 100 examples. Review articles can be found in the book of Casertano, Sackett and Briggs (1991), with a very recent one by Cox and Sparke (1996). Because this latter paper is so current, we will limit the discussion in this section.
Figure 24. Schematic of a polar ring galaxy seen from a variety of different orientations from Whitmore et al. (1990). According to the authors, less than half of these views are readily identifiable as a polar ring galaxy.
Polar rings probably form out of the total or partial disruption of a gas-rich companion (see e.g., Richter, Sackett and Sparke 1994). The process is like that for forming disks in ellipticals described above. Specifically, the accreted gas settles promptly into a plane determined by the collision parameters, and evolves to a preferred plane on a timescale much longer than the orbital or azimuthal smearing time. The initial plane is likely the orbital plane of the companion if it is promptly destroyed (Katz and Rix 1992, Thakar and Ryden 1996). If the ring forms out of a bridge with gas from both galaxies, formed in a direct collision, its orbital plane may be perpendicular to the bridge (Struck 1997).
Recently, a couple of observational studies of forming polar rings have appeared. Cox et al. (1996) present HI mapping of the II Zw 70 / 71 system. Reshetnikov et al. (1996) present optical photometry of the Arp 87 and Arp 293 systems. In all three cases the two galaxies are still attached by a bridge, which may include gas from both galaxies as in the Struck (1997) models. Moreover, the polar ring is either the smaller galaxy (Arp 87), or the two galaxies are of comparable size. This is not consonant with the fact that most "mature" polar rings are isolated, so, if these systems are future polar rings, a great deal of evolution remains, including the merger of the two galaxies.
Since the discovery of most of these objects in the 1970s and 80s much effort has focussed on studies of ring dynamics, with the goal of understanding why the rings are generally oriented around the poles, rather than at arbitrary angle. Specifically, the secondary evolution to a preferred plane is thought to be the result of differential precession. Orbital precession depends on both the inclination and the mean orbital radius of a gas cloud. Thus, annular rings at different mean radii precess at different rates (see the formula in Cox and Sparke 1996). If these rings are elliptical rather than perfectly circular, which seems likely initially, differential precession will lead to ring intersections, gas cloud collisions, and dissipation. The most likely result is settling into one of the preferred planes, i.e., either the equatorial or a polar plane.
One problem with this scenario is that there are some polar rings, which do not appear to be young, and in fact have red colors, but are still significantly inclined. Katz and Rix (1992) have argued that this could be the result of viscous couplings that are strong enough to prevent the independent precession of different rings. Thus, the whole polar disk would be viscously connected, and precess at some mean rate. However, another effect of the viscosity would be relatively rapid radial spreading in the disk, and presumably decreased lifetime. Nonetheless, Katz and Rix's simulations show that a quite long-lived and warped disk can develop. Recent HI (Richter, Sackett and Sparke 1994) and CO (Galletta, Sage and Sparke 1997) studies of polar rings have shown that many contain as much gas as a typical late-type galaxy, rather than a dwarf. In this case the self-gravity of the gas can play a significant role in holding the polar-ring together, again suggesting that they can be much longer-lived than originally thought.
Because of their large radii and orthogonal orientation, polar rings can offer a unique probe of the dark halos of galaxies. Cox and Sparke (1996) describe several techniques that have been used to analyse the halos of more than half a dozen galaxies. The particular advantage offered by the polar disk is information on the flattening of the host halo. Not many systems have been studied yet, and there are ambiguities or significant uncertainties in all the methods. However, the derived flattenings range from round to quite flat, though the majority are quite round. Different methods also give conflicting results, which is the case with NGC 4650A in particular, whose halo is either rounder than an E3 galaxy, or as flat as an E6-7, see Cox and Sparke. Nonetheless, these are early days in the use of this technique, and ultimately the information obtained should be of great importance.
6.5. Multiple Mergers
The topic of the simultaneous merger of more than two galaxies, or a rapid sequence of mergers, brings us to the edge of a nearly unexplored frontier. At first we might question the importance of this subject, since the universe is a low density gas of galaxies. However, a glance at the entwined galaxies in the compact groups illustrated in the Arp (1966) and Arp and Madore (1987) atlases, and the equivalent Vorontsov-Velyaminov (1959, 1977) galaxy "chains", assures us that multiple collisions do occur. Schneider and Gunn (1982) presented a photometric and spectroscopic study of a "nightmarish" example in the cluster V Zw 311, where the central object has at least 9 "nuclei" wrapped in a common envelope. Each nucleus is roughly as bright as other cluster galaxies. As a virtual "poster-child" for mergers in progress it was adopted as the frontispiece for the book of Tinsley and Larson (1977). While this system illustrates the extremes of multiple merging, Ramella et al. (1994) present reasons why the process may be common.
These authors studied the 38 compact galaxy groups from the catalog of Hickson (1982, 1993), and which also were included in published redshift surveys. The redshifts confirm group membership and eliminate most chance superpositions. Ramella et al. find that 29 of these compact groups are embedded in larger, but looser systems. They report that these larger associations are similar to groups discovered in the Center for Astrophysics redshift survey, and that the latter also often contain compact subgroups. In addition they use N-body simulations to show that compact galaxy groups "form continually during the collapse of rich loose groups".
In the introduction (sec. 1.4) we reviewed how initial beliefs that galaxy collisions must be rare were overcome when it was realized that most collisions are between galaxies bound in groups. The environment makes a collision much more likely, if not inevitable. The results of the previous paragraph extend this conclusion to another level of complexity. Multiple mergers are also not accidental. They are apparently the result of the formation of rich substructures in the (hierarchical) evolution of rich groups.
Unfortunately, this insight does not make multiple mergers easier to model. On the contrary, there are many difficulties, including the fact that there are a myriad of initial conditions. The need for good spatial and particle resolution in each component of each galaxy strains computer capabilities. Nonetheless, some impressive simulational studies have been published recently. They include the paper of Governato, Tozzi, and Cavaliere (1996), who studied the formation of Hickson compact groups in the collapse of loose groups.
Weil and Hernquist (1996, also 1994) followed the evolution of compact groups of 6 identical galaxies until all were merged, and they then studied the properties of the final remnant. Each galaxy in their half-dozen N-body simulations has bulge, disk and halo components. Some of their basic results are the same as in pairwise mergers, which is not surprising since the basic physics is the same. For example, the halos merge first, and the remnant is a dynamically hot elliptical-like object. On the other hand, there are a number of important differences in both morphological and kinematic properties. The multiple merger remnants are more commonly oblate spheroids that "appear round from many viewing angles." A large fraction (about half) of the substantial orbital angular momentum of the initial group is incorporated in the remnant. The spin of all components increases. However, the ratio vr / of the rotational velocity to the random velocity dispersion is small in the central regions, but increases to unity in the outer regions. At a given ellipticity, multiple mergers have somewhat higher (averaged) values of vr / , in agreement with the observations. Several other properties of the multiple merger models were also found to agree better with observation than pairwise merger remnants.
This pioneering study has many limitations - only a few initial conditions are used, all galaxies in the group are identical, and gas is not included. However, the models are sufficient to support the idea that multiple mergers of early-type spirals might be the best way to make ellipticals. Support also comes from Statler, Smecker-Hane and Cecil's (1996) spectroscopic study of the elliptical NGC 1700, which found a variety of evidence for more than one merger, and also an oblate form and increasing rotation at large radii in accord with Weil and Hernquist's models. Presumably, the early-type spirals would themselves be made in earlier mergers. This suggests a modification of Toomre's ellipticals-from-mergers theory, to a multi-step or hierarchical buildup.
1 Also called FIRGs, far-infrared galaxies,
ELFs, extremely luminous galaxies, and LIGs, luminous infrared galaxies,
though this latter also includes somewhat less luminous objects.
1 Also called FIRGs, far-infrared galaxies, ELFs, extremely luminous galaxies, and LIGs, luminous infrared galaxies, though this latter also includes somewhat less luminous objects. Back.