In previous sections I have mentioned the possibility that some E and cD galaxies formed by mergers of smaller galaxies. The probable importance of mergers (Toomre 1974, 1977b) has now been demonstrated by calculations of dynamical friction time scales, and by n-body simultations, which show that galaxy collisions are very sticky (see Tremaine 1981; White 1982 for reviews). I want in this section to discuss observational evidence that mergers do in fact make objects which look like ellipticals. The main controversy now is over what fraction of ellipticals originate in this way. This issue is still unsettled. However, observations reviewed here indicate that mergers are neither negligible nor the main process forming elliptical galaxies.
The first task is to see whether merger remnants look like ellipticals. One indication has already been provided by brightest cluster galaxies. Their profiles and characteristic parameters have properties like those of ellipticals, while the frequency of multiple nuclei suggests that many of them formed by mergers (section 3.3.4). More conclusive results require that we study individual galaxies in which merging has gone nearly to completion, but recently enough so that the identification of a merger remnant is reasonably secure. Criteria for recognizing such objects are discussed by Toomre (1977b) and especially by Schweizer (1978b, 1982). (1) The presence of two long tails in a single object is suggestive that there were originally two galaxies involved. Recall, for example, Toomre and Toomre's (1972) study of interacting doubles which form bridges and tails. (2) If the nucleus is single, then any merger is complete enough to be interesting. Younger mergers which still contain two nuclei should exist in small numbers, but these tell us less about the nature of merger remnants. (3) Candidates should be isolated enough so that there is no suspicion of a more prosaic tidal encounter. (4) Relative to the nucleus the two tails should be found to move in opposite directions, since tides are symmetric. This is important because it is practically the only non-trivial prediction that we can make. (5) Motions may still be complicated in the main body. In particular, the detection of two distinct velocity systems is a telling, although not quite conclusive, observation (e.g., NGC 2685, Schechter and Gunn 1978; Shane 1980). Additional, weaker evidence for a violent event such as a merger is provided (6) by the detection of a large population of young stars and gas, and (7) by the presence of luminous arcs or "shells" at large radii (Malin and Carter 1980, see below). Basically, these criteria exploit the fact that the central structure will settle into a final state long before the signature of a merger phase-mixes away at large radii.
The best candidate known for an object in the late stages of a merger is NGC 7252, the "Atoms-for-Peace" galaxy (Schweizer 1978b, 1982). It satisfies the above criteria very well (Fig. 18). It has a single body and one nucleus, but also two tails reminiscent of the tidal tails discussed in Toomre and Toomre (1972). These tails move in opposite directions with respect to the nucleus, at relative velocities of -92 and +65 km s-1. Combining these velocities with the lengths of the tails gives a kinematic expansion age of ~ 1 × 109 yr. The main body of the galaxy has an A-type absorption spectrum, emission lines and blue colors (U - B = 0.19; B - V = 0.67). Larson and Tinsley's (1978) stellar population models applied to these colors give an age of 1 ± 0.5 × 109 yr, consistent with the kinematic age of the tails. The main body of the galaxy also shows luminous ripples, loops and other transient features. Of special interest is the fact that the H II kinematics show two distinct systems of motion. Spectra taken at ten slit position angles through the nucleus show that there is a central disk of gas at r < 8" which has a very regular rotation field. The rotation rate at r ~ 2.4" is 80-100 km s-1; the kinematic minor axis is at position angle 26.4° ± 1.4°. In contrast, two measurements with long slits show that at r > 8" the gas rotates about a very different axis and with a more complicated velocity field (Fig. 18). Thus the morphology, the kinematics and the consistency of the kinematic and stellar-population ages all support the interpretation of this object as an almost-completed merger of two galaxies. Will it evolve into an elliptical? Figure 18 shows that the azimuthally averaged brightness profile is already close to an r1/4 law. There is too little gas left to make a prominent disk. It seems likely that NGC 7252 will turn into an object which is at least close in type to normal ellipticals.
Figure 18. Structure of NGC 7252, a probable merger in progress, from Schweizer (1982). The four photographs of increasingly deep exposure show the single nucleus, the main body full of loops and arcs (between arrows) and the two tails. Three of the photographs have the same scale, indicated by the 20"-long bar. The lower-right photograph gives the velocities, relative to systemic, of the H II regions at the ends of the tails (in km s-1). Also shown is the rotation axis of the gas disk at radii r < 8". The graphs at bottom left show emission-line velocities in km s-1 along the above rotation axis (upper) and perpendicular to it (lower). The symbols are a coding of the measuring accuracy (see Schweizer 1982). The rapid rotation about the axis illustrated gives way at r > 10" to rotation about a quite different axis. At bottom right is the azimuthally averaged brightness profile, in V mag arcsec-2. As indicated by the least-square fit (straight line), the profile is close to an r1/4 law.
Another possible remnant of a recent merger is the well known peculiar galaxy NGC 1316 (the radio source Fornax A). A beautiful series of photographs by Schweizer (1980) shows that the galaxy image is covered with luminous ripples and patches of absorption. At larger radii there are several loops, and at very large radii a faint tail, which may be tidal. In the main body of the galaxy there is a gas disk rotating about position angle 127° - 142°; however, the stars rotate around position angle 60°. Schweizer concludes that NGC 1316 has recently eaten one or more gas-rich galaxies. This is plausible, but further consideration should be given to possible interactions with the nearby companion NGC 1317. As in NGC 7252, the brightness profile of NGC 1316 is a very good r1/4 law over an unusually large range in surface brightness.
An interesting class of possibly related objects are elliptical galaxies with giant arcs or "shells" in their brightness distributions. For example, Malin (1979) has shown that NGC 4552 has a broad "jet" at large radii, and several luminous arcs or ripples out to µ ~ 27 B mag arcsec-2, r ~ 84 kpc (H0 = 50 km s-1 Mpc-1). The inner parts of the galaxy have the brightness distribution of a normal elliptical (King 1978; Kormendy 1977c). Still more impressive are the "giant shells" discovered in NGC 1344 and 3923 by Malin and Carter (1980). These reach radii of 56 and 180 kpc, respectively. In a survey of 12 isolated normal ellipticals, Malin and Carter find such features in 4 galaxies. The composition of the shells is unknown, but color measurements suggest that a likely possibility is stars (Carter, Allen and Malin 1982). A possible explanation is that a merger of an elliptical and a disk galaxy has left behind some disk material which is thoroughly sheared by differential rotation, but which is still cold enough to show sharp edges in projection (Quinn 1982, see Gilmore 1982; Toomre 1982; cf. Lynds and Toomre 1976).
The foregoing conclusions are still very preliminary. However, the observations strongly suggest that some mergers of disk galaxies are taking place, and that the result is very like an elliptical. Apparently (i) violent relaxation easily sets up an r1/4-law brightness distribution irrespective of whether merging galaxies or a collapsing protogalaxy were involved, or (ii) nearly all ellipticals form by mergers. Toomre (1977b) has argued for the second possibility, based on the high apparent frequency of mergers in progress. The more recent observation (Hoessel 1980) that multiple nuclei are very common in first-ranked cluster ellipticals is consistent with this suggestion. However, a number of observations suggest that mergers are not the main process forming ellipticals. (1) Ellipticals are most common in rich clusters, where the high velocity dispersion does not (currently) allow mergers to take place (Ostriker 1980; Tremaine 1981). (2) Ellipticals satisfy color-luminosity (e.g., Visvanathan and Sandage 1977), metallicity-luminosity (Faber 1977) and metallicity-instrinsic shape (Terlevich et al. 1981) correlations which are difficult to produce via mergers (Ostriker 1980; Tremaine 1981; White 1982). (3) Dwarf ellipticals cannot form by mergers because plausible encounter velocities are too high. But the properties of dwarf ellipticals are smooth extensions toward low luminosity of the properties of giant ellipticals (Tremaine 1981). (4) Ellipticals are more compact than spirals; merger remnants tend to be less compact than their progenitors (Ostriker 1980). (5) Ellipticals have a mean specific globular cluster frequency <S> ~ 4 - 7 per unit MV = -15 of galaxy. This is an order of magnitude larger than <S> ~ 0.2 for Sb spirals (Harris 1981; van den Bergh 1982, and references therein). In fact, the single Sa in the above papers has S = 1.4 ± 0.3 (NGC 4594), and several S0s have S ~ 4 - 7, intermediate between values for Es and Sbs. (6) Furthermore, the stellar populations and shapes of globular clusters are a function of galaxy type and luminosity. The Galaxy, for example, could not easily be made of many "Magellanic clouds" because they would leave behind too many young and highly elliptical globulars (van den Bergh 1975c, 1980b, 1982). Finally, (7) if collisions of comparably massive galaxies produce the majority of ellipticals, mergers of massive disks and much less massive galaxies might produce an embarrassingly large number of half-destroyed, thick disks. Many of these observations can be accommodated by merger theories (e.g., White 1982), but only at the cost of some theoretical squirming.
It is difficult to believe that the fraction of ellipticals produced by mergers is nearly negligible. It is equally difficult to believe that this fraction is nearly 100%. Further work is required to determine where between these extremes the truth lies.