Dept of Physics and Astronomy, Iowa State University, Ames, IA 50010 USA
9. Environments and Redshift Dependences
In previous decades most research on galaxy collisions has focussed on questions like: how common are mergers, what fraction of galaxies are merger remnants and what traces remain of the merger process? These questions can be viewed as basic and fundamental, while environmental dependences may be seen as details to be filled in later. Similarly, in the field of high-redshift studies the primary question has been how does the merger and interaction rate vary with time? Questions about the time dependences of environmental effects have been viewed as secondary.
On the other hand, if the nature and frequency of interactions have strong environmental dependences, we won't be able to answer the basic questions without an understanding of some of the "details". This situation is almost certainly qualitatively true in the present universe. For example, one important point is based on the galaxy density-morphology relation (e.g., Dressler 1980, and more recently Whitmore and Gilmore 1991, Whitmore et al. 1993), which tells us that ellipticals and early-type spirals are preferentially found in dense environments (e.g., clusters), and late-type spirals and irregulars are found mostly in low density (field) environments. Since the former types are generally gas poor, we would expect gas dynamics, induced star formation, and perhaps, nuclear activity to be less important in collisions and mergers in the densest environments. Another factor of comparable importance is that the relative velocities of galaxies in dense environments are much higher than those in small groups. This means that interactions are generally more rapid, and generate smaller disturbances in the galaxies involved. The complete story of interactions within galaxy clusters is, of course, much more complicated, as we will see below.
Currently we are entering a period of vigorous interest in the environmental dependences of collisions. On the theoretical side, large-scale N-body simulations of galaxy and cluster formation provide insights on not only the merger rate, but also on environmental and temporal dependences. Semianalytic hierarchical clustering models, based on the assumption that galaxies and larger structures are built up by hierarchical mergers from small initial units, e.g., dwarf protogalaxies, have also become very popular recently. These models are a useful quantitative tool for analysing and describing the detailed numerical simulations, and are also useful aids to in interpreting high redshift observations. The discovery of large numbers of "Lyman-break galaxies" at redshifts z > 3 provides one example. Hierarchical models can readily account for the numbers and SF histories of these objects (Baugh et al. 1998). Representative "merger trees" derived from these models show that many mergers occur at early times, and that mergers continue to be orchestrated by the clustering environment down to the present time, much as envisioned by Toomre in the 1970s. We shall discuss these studies further in section 9.3, after considering some specific environments in sections 9.1 and 9.2.
9.1. Groups and Compact Groups
There are many approaches to this subject, we will begin with the straightforward one of asking how many neighbors do collisional galaxies have? Unfortunately, there is not an equally straightforward answer, for several reasons. The first is that no very large-scale study of interacting systems has yet been undertaken. Optical searches of many systems for companions, down to low brightness levels, are very arduous. This should be clear from the discussion of the Holmberg effect in earlier chapters. On the other hand, anecdotal evidence suggests that such searches might be profitable, especially HI searches, where discovering even small (gas-rich) dwarfs is relatively easy. For example, the HI study of the VII Zw 466 ring galaxy system by Appleton et al. (1996) discovered two gas-rich, dwarf galaxies at some distance from the compact central group, but with redshifts suggesting that they are part of the group. In fact, numerous dwarfs have been found in the vicinity of many interacting systems, see for example, the recent catalog of 12 systems Deeg et al. (1998). However, the goal of this study and others cited above is the discovery of dwarfs produced in the collision, a phenomenon which could greatly complicate the present question of neighbor numbers. Thus, the present question is best addressed in Y-type collisions, where there has not been time for the formation of new dwarfs. We also must keep in mind the more general context, illuminated by the work of Ramella et al. (1994, 1995, also see Palumbo et al. 1995) that compact groups, like the Hickson groups, are generally imbedded in a larger loose groups which continually feed the compact center.
Besides anecdotes and studies of groups that are difficult to interpret, there are also some studies of the neighbor statistics of specific types of interacting galaxies. First of all, on the bright end of the spectrum we have the merging or merged ULIRGs. After reviewing a number of individual cases Sanders and Mirabel (1996) conclude that most of these objects - "appear to involve strong interactions/mergers of molecular gas-rich spirals where the pairs are either isolated or part of small groups" - and are generally not in dense clusters. While this conclusion is not based on a rigorous statistical study, there are many examples.
Few and Madore (1986) carried out a more formal study of the neighbors within two ring diameters of 69 ring galaxies, using southern sky survey plates. Rings are a readily identifiable morphology, and since collisional rings are a young Y morphology, the collision companion, at least, should be nearby. However, the disadvantage is that most rings are not collisional systems. Indeed, the goal of this study was to obtain some idea of what fraction might be collisional. The ring sample did indeed show a small excess of near neighbors over a control sample, though the most common number of neighbors was one in both cases. The authors took the analysis a step farther by classifying each ring as either P or O type, where the former have a knotty ring structure and usually a displaced nucleus or other morphological peculiarity, while O-types have a central nucleus, and a smooth ring. Thus, there is direct evidence for a collisional disturbance in the P-types, but little in the O-types. Most of the companion excess was found in the P subsample. We conclude that this special class of collisional system is generally found in small groups, with from one to a few companions that are not much fainter than the primary.
Another very interesting statistical study with a narrow focus is Odewahn's (1994) study of 170 Magellanic spirals. These are very late-type (beyond Sc, Sd) spirals, that are typically gas-rich. Odewahn first applied an arm strength classification to each galaxy, and then noticed that virtually all of the 75 objects with "well classified asymmetric arms" had a companion within a few diameters. Moreover, the companion distances peaked at small separations, indicating that most are probably not chance alignments, but physical associations, and in most cases, interactions. Thus, a very large fraction of one particular type of galaxy exists in small collisional groups, and most of the members of this class have unbalanced spiral waves or bars.
The study of Fried (1988) constitutes one of the most general attempts to answer the question above. He assigned each of 517 nearby galaxies to an interaction class ranging from 0 (no evidence of disturbance) to 3 (strongly disturbed), using sky survery prints. His discussions of the difficulties in carrying out such a task are illuminating and explain why few others have attempted it. He found very few field galaxies were disturbed, i.e., outside of class 0. In groups, on the other hand, he found nearly 30% of the galaxies were disturbed. Most of these are spirals. Fried notes difficulties in detecting faint shells and other collisional debris around ellipticals in the survey prints, and in deriving this figure adopts the fraction of disturbed ellipticals from the more sensitive study of ellipticals by Sadler and Gerhard (1985).
Compact galaxy groups have received a great deal of attention in recent years, especially the 100 groups catalogued by Hickson (1982, 1994, and Hickson et al. 1992, see Fig. 27). The nature and frequency of interactions in such provides valuable information on a very specific, but increasingly well-studied environment. Moreover, some important properties about this environment are reasonably well-established now, at least in the case of the Hickson groups. The first is that the groups are generally true physical associations, with a high density of galaxies and relatively low velocity dispersions, though they frequently contain one or more superposed background objects. This conclusion is based both on member redshifts (see e.g., Hickson et al. 1992 Ramella et al. 1994, 1995), and more indirect evidence. The latter includes the finding of Palumbo et al. (1995, also see Sulentic and Raba?a 1994) that the spiral fraction of compact groups differs from that of field galaxies. the conclusion even seems to hold for "poor" groups with fewer than 6 memebers (Zabludoff and Mulchaey 1998). The very overdense environments of these groups are presumably very favorable for interactions.
Another, more controversial property of these systems is that they may be commonly embedded in larger, loose groups, which continue to feed them galaxies. Specifically, Ramella et al. (1994, 1995) clearly find these extended groups in redshift space. However, in a study of sky survey prints, Palumbo et al. (1995) find that only about 18% of the groups have significant concentrations of galaxies outside the core group. Sulentic and Rabaça argue from lack of compact group merger remnants that the groups must simply have a much longer lifetime than would be expected from merger simulations (see, e.g., Mamon 1986, 1987, Barnes 1989, BH92a, Governato et al. 1996, Weil and Hernquist 1996). Zabludoff and Mulchaey (1998) find that a large fraction of the dark matter in poor groups is located in a large common halo, rather than in the individual galaxy halos. Thus, as in large clusters, galaxy virial velocities within the common halos may be high enough to delay merging, without any need for continued feeding. However, these authors also find evidence for continuing accretion onto their groups, suggesting both phenomena play a role.
The question of the interaction frequency in compact groups has been addressed with observational searches for enhanced activity in a variety of wavebands. Perhaps surprisingly, in a study of the optical luminosity function of compact groups, Sulentic and Rabaça (1994), found little evidence for an interaction-induced luminosity enhancement relative to control samples. This result does not necessarily contradict the conclusion that interactions are common in these systems, which is not only evident in the images, but confirmed by more detailed studies. For example, Rubin, Hunter and Ford (1991) carried out detailed imaging and spectroscopic observations of 21 Hickson groups. In addition to the obvious morphological disturbances, they found a variety of kinematic peculiarities, including disk galaxies with asymmetric rotation curves and velocity patterns "too peculiar to form rotation curves." Specifically, they found that 10 out of 12 elliptical or S0 galaxies had ionized nitrogen emission from their nuclei, with resolved ionized gas disks in about half of them. They speculated that these gas disks are recent acquisitions.
Pildis, Bregman, and Schombert (1995) undertook very deep photometry of about a dozen Hickson groups. With this high quality data they modeled the mean luminosity profiles of the early-type galaxies and subtracted them out. In many cases shell systems remained, providing still more evidence for mergers and accretion events. Mendes de Oliveira and Hickson (1994) also used extensive surface photometry for a study of 202 galaxies in the Hickson groups. They found that 43% of the galaxies in their sample had disturbances indicative of interactions, and this was true of 75% of the galaxies in a subsample of groups with published kinematical data. Thus, the conclusion that interactions are common in compact groups remains secure. The unenhanced luminosity functions may be the result of a failure of interactions to induce star formation in these particular environments.
We know from the discussion of the work of Hunsberger et al. (1996) in sections 5.4 and 7.3 that SF does occur in tidal structures in some compact groups. Unfortunately, it appears that no large statistical study of the colors of galaxies in compact groups has been published yet (though many B-R colors are given in Hickson (1994).
Recently, Menon (1995 and references therein) presented the results of a radio continuum survey of 133 spirals in 68 Hickson groups. In this sample 56 galaxies were detected, and the continuum emission was generally found in "slightly extended nuclear regions suggestive of starburst activity." On the other hand, the total continuum emission from Hickson spirals was found to be less on average than that of a comparison sample of isolated spirals. It appears that the nuclear emission in Hickson spirals is more than offset by an emission deficit in the other parts of the disks.
Williams and Rood (1987) carried out an HI survey of 51 of the Hickson groups. They found that the amount of HI, the basic fuel of SF, in these groups was about half that in a control sample of loose groups with similar morphological and dynamical processes. Preliminary results of a survey of southern compact groups are similar (Oosterloo and Iovino 1997). HI has been found outside the galaxies, and indeed, throughout the group in a few groups (e.g., Williams and van Gorkom 1988, Williams et al. 1991), but this does not add a significant amount of mass to the overall average. So where did the gas go?
The far-infrared and CO study of Verdes-Montenegro, et al. (1998) finds that the molecular content of 80 galaxies in Hickson groups is very similar to that of a control sample of spirals. The far-infrared data show also show little enhancement of SF or nuclear activity. These results are basically confirmed in the study of Leon, Combes, and Menon (1998), which included 70 galaxies in 45 Hickson groups. The latter study did find some evidence of enhanced molecular and dust masses, which together with normal far-infrared fluxes may imply a lower than average SF efficiency. Nonetheless, it appears that no great mass of gas has gone into the molecular phase.
In some cases it appears that a significant part of it has gone into a hot halo enveloping the group, which has been observed with the ROSAT X-ray satellite (e.g., Ebeling, Voges, and Bohringer, 1994, Saracco and Ciliegi 1995). Dynamical models suggest that it is plausible that this gas has been stripped from individual galaxies and heated in the collapse and subsequent evolution of the group, though there are difficulties in fitting the temperatures and emission profiles of the X-ray gas (Diaferio, Geller, and Ramella 1995, Pildis, Evrard, and Bregman 1996).
Finally, in a detailed optical imaging and spectroscopic study of ellipticals in compact groups, Zepf and Whitmore (1993) find evidence for disturbances, but little evidence for young stellar populations. Thus, they conclude that these ellipticals are not the result of recent mergers between (gas-rich) spiral precursors. On the contrary, most of the stars in them must have been formed at a much earlier time.
Thus, several lines of evidence suggest that compact groups of galaxies come together (collapse) rather late in the history of the universe, after forming most of their stars in the individual galaxies. The subsequent merger times can be increased by varying the galaxy mass distribution (Governato, Bhatia, and Chincarini 1991), or the fraction of dark matter distributed generally through the group (Bode, Cohn, and Lugger 1993). However, the continued input of galaxies from a larger, looser environment seems like the surest solution to the problem of the relatively high frequency, or longevity of compact groups. The special environments of these groups might be viewed as interaction generators.
The overall conclusion of this section is that most vigorous galaxy interactions occur in groups containing a modest number of galaxies. Compact groups are special, and unusually well-studied, cases. Unfortunately, the statistical studies of groups are generally quite narrowly focused, and the collections of anecdotal results suffer even stronger selection effects. Thus, while the above conclusion appears to be quite firm, it is difficult to pursue further analyses of causes and connections at present.
9.2. Dense Clusters
Dense clusters of galaxies, the urban environment of the galaxy world (see Fig. 28), do not seem to be very conducive places for strong interactions. First of all, in the previous section we concluded that most strongly interacting systems are found in modest sized groups. A glance at the Arp or Arp and Madore atlases confirms that most of the systems there do not lie within a dense cluster. Fried's (1988) study of interactions as a function of environment (discussed in the previous section), included the Virgo cluster, where he found signs of interaction in only 16% of the galaxies. This lack of obvious signs of galaxy collisions does not encourage studies of interactions in clusters, and until recently, relatively few studies, pursuing restricted subjects, have been published. However, the current evidence suggests that there is no lack of interaction in clusters, but rather that the interactions in this environment are qualitatively different from other environs.
Statistically, the best evidence for such differences comes from the study of SF in 15,749 (!) galaxies from the Las Campanas redshift survey undertaken by Hashimoto et al. (1998). They found that SF was reduced in all types of spirals in large clusters, and conjectured that this is the result of gas removal processes. On the other hand, they find a "prevalence of starbursts in intermediate density environments," such as groups and poor clusters, and they conclude that this prevalence is the result of interactions.
9.1.1. cD Galaxies
The notion that strong interactions are rare in cluster environments is historically ironic, because the buildup of central cluster galaxies and the giant cD galaxies, in particular, by mergers and "galactic cannibalism" was an early area of study (e.g., Ostriker 1977, Schneider and Gunn 1982). A defining characteristic of cD galaxies is that while in their inner regions they have surface brightness profiles like those of ellipticals, in the outer regions (which can be truly gigantic) they decline much more slowly. That is, cDs have giant luminous halos or envelopes, which can contain as much light as the rest of the galaxy. Schombert's (1987, 1988, and references therein) studies of the surface brightness distributions of 342 bright cluster ellipticals substantiate and quantify these general statements. According to the galactic cannibalism theory, while the cD precursor may well have been the largest galaxy to form in the center of the cluster (or subcluster, see Merritt 1984), the halo developed by the disruption and merger of numerous smaller galaxies.
There is much numerical and observational support for this theory. First of all, the general structure of cD envelopes can be accounted for by homology calculations (Hausman and Ostriker 1978), Monte Carlo calculations of the effects of collisions in clusters (Richstone 1975, 1976), and N-body simulations (e.g., Farouki, Shapiro and Duncan 1983, Barnes 1989). The accreted material ends up in an extended halo because it is dynamically hotter than the stars formed within the galaxy. The situation is like the case of the shell galaxies discussed in section 5.5; the extended envelopes are essentially multiple, relaxed shell systems. (Interestingly, there has been relatively little modeling work on this subject in recent years, with much more effort going into details of the formation of giant ellipticals from major mergers, see above.) A second type of evidence is provided by velocity studies that establish that some cDs are surrounded by a population of galaxies that are bound to the cD itself, rather than just the general cluster potential (Bothun and Schombert 1988, 1990). The effects of dynamical friction on these galaxies must be strong, and they should merge on a timescale much smaller than the age of the universe.
However, the most convincing evidence is the discovery of examples of dense groups in which the galaxies are surrounded by a common envelope, all are contained within a region whose size is no more than a few diameters of the largest galaxy. Examples include V Zw 311 (Gunn and Schneider 1982), discussed above, and Hickson 94 (Pildis 1995). Cavaliere, Colafrancesco, and Menci (1991) use an analytic formalism to argue for the existence of a merging runaway in dense groups. Interestingly, these nascent cDs also seem to confirm Merritt's suggestion that cDs form early, e.g., in subclusters, rather than at the core of large clusters of galaxies. On the other hand, Bothun and Schombert's work shows that their formation is a continuing process. In any case, the making of cDs is one example of how interactions in clusters, while not as spectacular as those in the field, can yield extreme products.
9.1.2. Collisions and Harassment in Clusters
Why should the effects of collisions between comparable galaxies be weaker in clusters, and what are some of the qualitative differences? In fact, there are a number of ways in which the cluster environment can modify the effects of galaxy collisions. The first and most important of these effects results from the fact that the random velocities of the galaxies in large clusters are generally greater than the internal velocities of the galaxies. Thus, a typical collision between cluster members occurs at much higher relative velocities than we have generally considered, and so, the time the two galaxies are close together is correspondingly shorter. The chance that dynamical friction will leave them in a bound orbit is also much reduced. We have considered two estimates of collisional perturbation strength earlier in this article, and both contain terms that illustrate this effect. In equation (1) there is an inverse dependence on relative velocity, while in equation (10) there is a direct dependence on the duration of the collision.
Richstone's (1975, 1976) early Monte-Carlo studies of hyperbolic collisions between spheroidal galaxies in clusters showed that while the galaxy cores were relatively unaffected, the envelopes could be changed signifcantly by tidal stripping and impulsive energy injection. Richstone also analysed how tidal cutoff radii of the galaxies evolve, and noted that halos can be stripped quite promptly. (However, Allen and Richstone (1988) revised the estimates of tidal radius evolution.) The result on halos has been confirmed in more recent simulations, and can be viewed as an aspect of what is called the "overmerging" problem, i.e., the inability of cluster simulations to retain substructure. This problem, however, may be the result of limited numerical resolution and other technical details (Moore, Katz, and Lake 1996, Frenk, Evrard, White, and Summers 1996).
Recently, the modest disturbances resulting from high-velocity encounters, which are frequent in the cluster environment, have been termed "galaxy harassment" by Moore et al. (1996), who present simulations of their effects. These effects appear qualitatively similar to the tidal disturbances of disk galaxies discussed in earlier chapters, but they are on the weak end of the disturbance continuum. Cluster ellipticals and other early types are relatively unaffected. They point out that the fraction of disturbed spirals is generally high in clusters, and the harassment explanation is very plausible. Icke's (1985) models of distant hyperbolic encounters produced similar results, and he emphasized that the effects on the gas would be greater. Moore et al. also suggest that the process plays a role in fueling quasars, and might account for the large fraction of blue galaxies in clusters at high redshift (the Butcher-Oemler effect). Oemler, Dressler and Butcher (1997) find no evidence that the blue galaxies are merger remnants, and favor the harrassment explanation.
Joseph's (1996) review of the Moore et al. harassment paper is cautious, pointing out that in the simulations the haloes of the harrassing galaxies were assumed to be simple rigid, softened point-masses. This again raises the issues of how much halo cluster galaxies retain, and the structure of those haloes. These are complex issues, both observationally and theoretically, and we cannot digress into a proper consideration of them. We will have a little more to say about the results of cluster formation simulations, which provide some insights, below. Observationally, it is very difficult to use traditional means, like HI rotation curves, because outer HI disks have often been stripped off. Some rotation curve studies have been carried out, including a recent series of papers by Amram et al. (1993, 1994, 1995, 1996). These studies find generally flat rotation curves, in contrast to some earlier reports. However, in most case they do not extend far enough in galaxy radius to detect much halo dark matter, if it is present. Natarajan et al.'s (1998) HST study of gravitational lensing in a cluster with a redshift of z=0.31 provides evidence that cluster galaxies have smaller and less massive halos than their counterparts in the field. It will be exciting to see additional lensing studies in the coming years.
9.1.3. The Cluster Environment: Stripping and Cluster Tidal Effects.
The first inkling of gas removal from cluster galaxies is Spitzer and Baade's (1951) suggestion that high velocity collisions in clusters could remove their gas and deposit it in an intracluster medium. Two decades later Gunn and Gott (1972) proposed an alternate mechanism that didn't depend on such collisions, which, by that time, didn't seem sufficiently frequent. Like Spitzer and Baade they reasoned that high galaxy random velocities would lead to the shock heating of any cool gas, so intracluster gas must be hot, millions of degrees Kelvin. Because of this they concluded that it must be smoothly distributed (a routine observation of X-ray telescopes now). The cluster galaxies would be constantly plowing through this medium, and experiencing ram pressure. Gunn and Gott estimated that this pressure could entirely strip the diffuse gas from a spiral. Later, Valluri and Jog (1990) described some modifications that would result from the multi-phase structure of the interstellar gas, and specifically, concluded that it would be difficult to strip dense molecular clouds. More recently, detailed numerical simulations indicate that only the outer parts of gas disks will be stripped (Kundic, Hernquist and Gunn 1992).
This latter conclusion finds some support in observation, especially in the case of the Virgo cluster, the nearest and best studied cluster. To begin with, Giovanelli and Haynes (1983, Haynes, Giovanelli and Chincarini 1984, Haynes and Giovanelli 1986) obtained HI observations of more than 160 Virgo Cluster spiral galaxies, and hundreds of field galaxies which could be used as a control sample. Using several different measures, they found a significant HI deficiency for a subpopulation of Virgo spirals, though the remaining subpopulation was found to normal HI contents. They attributed the observed deficiencies to ram pressure stripping, and noted that spirals within 5¡ of the Virgo center had lost 90% of their neutral hydrogen gas mass. They also noted that among the deficient galaxies it appeared that SF had been "quenched".
Following up on this work, Kenney and Young (1986, 1988, 1989) surveyed the molecular gas masses and distributions of Virgo cluster galaxies, using observations of the CO molecule. They found that the molecular gas contents of the bright Virgo spirals are not greatly deficient, or otherwise unusual, and that "the total gas deficiency is manifested largely by a lack of HI in the outer disk" (Kenney and Young 1989). Inner disk regions appeared relatively uneffected. More recently, Koopman and Kenney (1993, 1994, 1995, 1996) have used Ha and R band imagery to study the star formation properties of a large sample of Virgo galaxies and an isolated control sample. They find a wide range in the Ha surface brightnesses in the inner disks of their sample, some are consistent with "fading disk" models, e.g., with SF truncated by stripping. However, SF enhancements are evident in others, some of which are involved in an interaction, others may be accreting gas. Among their more interesting findings is a class of "peculiar early-type spirals", which are "small bulge, gas-deficient galaxies with featureless outer disks, but strong circumnuclear star formation." They note the characteristics of these galaxies are what one might expect in stripped Sc type galaxies. From a broader point of view, these detailed results on SF in Virgo galaxies, provide an interesting complement to the global picture of the SF reduction in dense clusters provided by the Las Campanas survey cited above.
The HI properties of Virgo cluster galaxies were surveyed again with higher resolution and sensitivity by Cayatte et al. (1990, 1994), and also Warmels 1988a,b. The higher resolution of the radial profiles of the galaxies they observed allowed Cayatte et al. (1994) to draw somewhat more detailed conclusions than previously possible. E.g., they find that in - "some galaxies ram-pressure stripping has done serious damage to the HI disks, while in other galaxies turbulent viscous stripping and thermal conductivity have caused a mild, but global HI deficiency..." Phookun and Mundy (1995) have also revived the idea of ram pressure "pushing", rather than stripping, as a cause of the HI disk asymmetry in the Virgo galaxy NGC 4654, and possibly others.
In sum, a number of different authors find strong evidence that stripping has played an important role in shaping Virgo cluster disk galaxies. It appears that the data are becoming good enough to allow the study of details of the stripping process, and related processes. Thus, the prospects for learning more about these processes in Virgo appear very good. But what about other clusters, is Virgo representative?
HI surveys of several other clusters have been reviewed recently by Van Gorkom (1996). Like Virgo, "shrunken HI disks" seem to be common in cores of nearby large clusters, suggesting that stripping and related processes are equally common. The Ursa Major cluster studied by Verheijen (1996) provides a counterexample, since the HI disks of the spirals in the center of that cluster appear normal, not HI deficient. However, the lack of X-ray gas in this cluster suggests that it may be much younger than other nearby clusters like Virgo, Coma and Abell 1367.
The two clusters that have been mapped to date at somewhat higher redshift each show individual peculiarities (see Van Gorkom 1996). The first of these is the Hydra cluster at z = 0.035. It has a bimodal velocity distribution function, no HI deficit, and other morphological peculiarities all of which suggest that the spirals in the cluster core are either currently falling into the cluster, or are a chance superposition. In Abell 2670, at z = 0.077, there is evidence for three distinct subsystems. It is not surprising that no general evolutionary trends have yet emerged from such studies, since the look-back times to these clusters are not a very large fraction of the age of the universe. Given optical observations of peculiarities in higher redshift clusters it will be interesting to see how such studies develop in the future.
Another effect of the cluster environment, tidal disturbances on cluster galaxies due to the overall cluster potential, has been studied recently by Henriksen and Byrd (1996, also Byrd and Valtonen, 1990) . They conclude that under a fairly wide range of conditions, cluster tides can disturb galaxy disks enough to excite waves, global compressions, and enhanced star formation. They suggest that ram pressure stripping will later remove the gas in galaxies, guaranteeing a limited timescale for the SF enhancement, as in the observational Butcher-Oemler effect. The end results of the tidal disturbances seem very similar to those of the galaxy harrassment described above, although the processes are distinct.
9.2.4. Cluster-Cluster Collisions
The HI observations described in the previous section provide evidence that substantial groups of galaxies continue to fall into large clusters. These observations can be viewed as a subset of a larger set of kinematic and morphological observations that suggest that unrelaxed substructure is common in galaxy clusters (see e.g., the review of Fitchett, 1988). A clumpy or asymmetric distribution of the hot X-ray gas in clusters also provides evidence of incomplete relaxation in nearby clusters, see e.g., the reviews of Jones and Forman (1991), and Henry and Briel (1993). More recent X-ray satellite (ROSAT) results are described in Henry and Briel (1996), and Knopp, Henry and Briel (1996). Mahdavi et al. (1996), Mohr, Geller and Wegner (1996) and Mohr et al. (1996) have used optical redshifts, surface photometry and X-ray observations to demonstrate the existence of substructure and probable cluster-cluster mergers in several more clusters.
While relaxation times are long in clusters, e.g., crossing times are of order 109 yrs., these times are still much shorter than the age of the universe. Thus, they should have had time to relax since their formation. Moreover, the large fraction of highly evolved galaxies found in large clusters suggests that they were not formed recently, but in fact, probably formed quite early. This is in contrast to the poor clusters recently sampled by Ledlow et al. (1996), who find that the morphology, dynamics, and environments of these systems "are indicative of young, dynamically evolving clusters."
The solution to the paradox of highly evolved member galaxies observed in conjunction with nonequilibrium cluster dynamics favored by most of these authors is that these phenomena are the result of continuing hierarchical growth of large-scale structure. Clusters continue to merge with clusters, and substantial "clouds" of galaxies continue to fall into large clusters, especially out of the sheets and filaments that make up the adjacent superclusters. The latter case (supercluster feeding) is very reminiscent of the discussion above of the continuing evolution of compact groups, and in a hierarchical growth picture we would expect such processes to co-exist on multiple scales. In the case of collisions between two large clusters of comparable mass, there can be dramatic and observable effects on the X-ray gas of the clusters, as recently shown in detail by the models of Roettiger, Burns, and Loken (1996).
Research on the topics of cluster-cluster collisions and continuing infall are relatively new, and we will conclude this discussion by noting that collisions between larger entities may orchestrate some unique collisional processes for the galaxies contained within them. First, we note that since clusters have massive halos like galaxies, and since in collisions these halos will exert a strong dynamical friction force, we expect colliding clusters like galaxies to merge within a few crossing times. (A timescale we cannot really call short!) However, during this time the galaxies involved will feel a strongly fluctuating cluster gravitational field. This field should significantly affect the orbits of dwarf satellite galaxies in the halos of large galaxies, and similarly affect the individual galaxies orbiting the center of mass in bound groups. In both cases, tidal torqueing may unbind some objects, releasing them into the general cluster field Antunes, Wallin and Struck, in preparation). It may also remove orbital angular momentum from other objects, putting them on more radial orbits, and enhancing the chance of collision with the primary galaxy or other group members. This process is not the same as the galaxy harrassment described above, but if infall into clusters continues at a steady rate, it may be an important process in shaping cluster galaxies.
9.3. High Redshift Collisions
The direct study of galaxies and galaxy collisions at high redshifts (and long lookback times) has become possible in this decade, especially because of the high sensitivity and resolution of instruments like the Keck telescopes and the Hubble Space Telescope. A large fraction of the evolutionary history of galaxies is now becoming directly observable. The amount of data obtained on high redshift objects, and our understanding of evolutionary processes will certainly grow rapidly in the coming decade as the exploration of this new frontier accelerates. We can expect new input on some of the oldest and most fundamental questions in this field, such as how much higher was the collision and merger rate in the past, and what is the average rate of change with redshift or lookback time?
The time dependence of the merger rate has been a subject of increasing attention in the last decade after the discovery of numerous, faint "blue galaxies" at a high redshift (e.g., Broadhurst, Ellis, and Shanks 1988, and more recently Driver, Windhorst, and Griffiths 1995, and references therein), which may be the result of more frequent interactions and mergers at earlier times. The merger rate of galaxies is often parametrized as a power-law function of redshift, MR µ (1+z) m, and thus, it can be conveniently discussed in terms of the value of the exponent m. However, to date it has not proven easy to determine the value of this exponent. With current resolutions and sensitivities it is still difficult to confidently recognize ongoing mergers and merger remnants, and eliminate chance superpositions of galaxies. Many of the tidal structures discussed in earlier chapters are too faint to observe at high redshifts.
Thus, it is perhaps not surprising that values of m reported in the literature range from 0 (no evolution in the merger rate) to m > 3 (high rates of evolution). A sampling of values reported in the recent literature are given in Table 3. It seems premature to make any strong conclusions, but it does appear that most studies find some evolution, and that the more extensive recent studies yield intermediate values of m. A recent study of Keel and Wu (1995) provides an estimate of the current merger rate, i.e., the constant in the power-law. Interestingly, they also find a large difference between the rate of spirals in pairs (4.2 per Hubble time), and the average for all spirals (0.33 per Hubble time). This suggests the possibility that the merger rate may depend sensitively on environment and the evolution of different environments with time.
Recently, there been a number of simulational studies that are relevant to the merger rate. Firstly, there are studies of effective merging cross sections and merger times, e.g., the recent work of Makino and Hut (1997). Secondly, there have been a number of recent papers on the merger rate in specific environments. These include studies of merging in compact groups (e.g., Athanassoula, Makino and Bosma, 1997, and the references of section 9.1), and in clusters and other large scale structures (e.g., Menci and Caldarini 1994, Frenk, Evrard, White, and Summers 1996), and in cosmological structure formation simulations (e.g., Lacey and Cole 1994, Navarro, Frenk, and White 1995, Baugh, Cole, and Frenk 1996, Baugh et al. 1998). Some of the latter calculations have been able to derive merger rates, as well as giving a qualitative feel for what merger histories are typically like. Some of them give encouraging agreement with merger rates derived from analytical, hierarchical clustering models, and both would favor the larger values of m.
Another fundamental question that will be clarified by high redshift observations is whether the nature of galaxy collisions was different when the universe was appreciably younger? We have seen in previous sections that there are differences in collisions occuring in different environments. Since these environments have themselves evolved considerably, we might expect differences at high redshifts. Average galaxy properties have also evolved continuously. For example, in the distant past galaxies were more gas-rich, and collisions and mergers between them may have generated spectacular star formation and nuclear activity more often than at present.
One phenomena that is very relevant to this discussion is the Butcher-Oemler effect, the presence of unusually high numbers of blue galaxies in clusters at moderately high redshifts (Butcher and Oemler 1978, 1984, Couch et al. 1994, Dressler et al. 1994, Barger et al. 1996). Spectral line analyses led to the discovery that many of the blue Butcher and Oemler galaxies are so-called E+A galaxies (see Gunn and Dressler 1988). These are galaxies with a relatively blue color, but strong hydrogen absorption lines rather than emission lines, plus absorption lines indicative of an old "elliptical-like" stellar population. These objects are now known to consist of two dominant stellar populations, an old one, together with an aging starburst, of order 1-2 x 109 yrs. old. Strong Balmer absorption line spectra (A-type star) are now used as essentially the defining characteristic of a "post-starburst" population, as these intermediate age populations are called. (Interestingly, Liu and Kennicutt (1995) also find E+A spectral features in a "considerable fraction" of the objects in their recent study of merging galaxies. However, it is not clear that there is any connection between these objects and the high redshift E+A galaxies.)
There is as yet no consensus on the cause of this effect. Explanations range from galaxy harrassment or mergers at times when cluster galaxies had more gas, to higher rates of spiral galaxy infall into younger clusters. Caldwell and Rose's (1997) recent study of galaxies with Butcher-Oemler type spectral characteristics in nearby clusters provides some support for the latter explanation. They find evidence of substructure in the clusters containing these galaxies, and suggest that the infall of a subcluster about 10 9 yrs. ago, could have triggered the star formation in the galaxies that presently have post-starburst spectra. It also seems very reasonable to expect that interactions induce more vigorous star formation and other effects in young, gas-rich galaxies. There is already some evidence for stronger SF responses in interactions at high redshifts (e.g., Butcher & Oemler 1978, 1984, Lavery & Henry 1986, Lavery, Pierce, & McClure 1992, Burkey et al. 1994, Driver et al. 1995a,b, Koo et al. 1996).
The "faint blue excess galaxies" seen at redshifts z < 1, are a related observational phenomenon. These objects are not found in clusters, and may include many dwarfs and low surface brightness galaxies. This has led to much discussion of the idea of two epochs of "galaxy formation", one associated with these objects, and the second forming giant galaxies at larger redshifts (e.g., at z > 2.0, see Babul & Rees 1992, Driver, Windhorst & Griffiths 1995, Lilly et al. 1995, Driver et al. 1996, Gwyn & Hartwick 1996). The high redshift objects are becoming increasingly accessible to observation, e.g., in the Hubble Deep Field (e.g., Gwyn and Hartwick 1996, Lowenthal et al. 1997, Madau et al. 1996, Steidel et al. 1996, van den Bergh 1996). Thus, we expect rapid progress in this area in the next few years, and hopefully, much clarification of these issues.