Annu. Rev. Astron. Astrophys. 1985. 23: 147-168 Copyright © 1985 by Annual Reviews. All rights reserved

5.2. Theories of Shell Formation

The presence of ripples in several merger candidates was ascribed by Schweizer (112) to the merging process itself, presumably as a consequence of the involvement of a low-velocity-dispersion system. To explain the shells and ripples in normal ellipticals, Quinn (87, 88) invokes the capture and subsequent disruption of a small disk system by a big elliptical galaxy 10 to 100 times more massive than the disk. Unless otherwise stated, the results discussed here are from the simulations in (87) and (88). In (87) the small disk has been modeled self-consistently, but shells are essentially a test particle, phenomenon, as was demonstrated by the calculations of Toomre (114, and private communication), Quinn (88), and Dupraz & Combes (37). High-orbital-momentum encounters result in a spatial wrapping of the companion that produces encircling crossing structures unlike the shells observed. The fact that we do not observe them is an argument for the low frequency of such encounters. Low-angular-momentum encounters result in a phase wrapping of the companion in the potential well of the elliptical, which produces structures very reminiscent of those observed (87, 88, 114). The process can be understood as an analog of the "pig-trough" dynamics (69). The particles will oscillate in the potential well of the elliptical and form a sharp crest at their radii of turnaround. A spread in the periods will result from the initial spread in energy, so that multiple shells will form between a maximum and a minimum radius, defined by the maximum and minimum energies of the particles in the disk. The shell spacings should decrease with decreasing radius, as confirmed by observations, and their number should increase with time. Of course, in numerical simulations this is limited by the total. number of particles used, which explains why cases with as many as 20 shells have never been modeled. Thus it was shown that shells are density waves or ridges moving outward with time with a small phase velocity. The interleaving of shells at opposite sides of the nucleus, the facts that the shells are aligned, concentric, and not all encircling, and the consistency of the shell colors with those of disk stars are all natural consequences of this model.

Promising matches to observed galaxies can be simulated (114). The number of shells, assuming that all the shells are due to the. same infall, can be used to date the encounter, while their distribution gives us information on the potential well of the elliptical. Quinn (87, 88) thus finds evidence that the central potential wells in several galaxies are less concentrated than predicted by an r^l/4 law with constant mass-to-light ratio, but this result depends crucially on finding all the shells, which is not a trivial task. Indeed, the CCD data permit the digital analog of the unsharp masking technique, and new shell sections have been found, even in some of the galaxies previously studied by Quinn (cf. 42).

Several points remain to be clarified in this picture (88). The innermost shell has been found in many cases to sit quite deep inside the potential well of the elliptical, contrary to what one would predict from the model. This could be because the models neglect dynamical friction and energy transfer from orbital motion to internal velocity dispersions. Thus a full; self-consistent treatment of both galaxies might be needed for the correct modeling of the inner parts. Sufficiently high velocity dispersions in the companion can be expected to smear out the shells. Indeed, velocity dispersions enhance the phase width and thus lower the maximum possible number of shells. Therefore, accretion of a small elliptical would result in no more than a half dozen shells (37, 88). The apparent alignment with the major axis in flattened ellipticals points to a nonspherical potential. The numerical simulations of Dupraz & Combes (37) give spherical shells, in projection enclosing like parentheses the galaxy's major axis, when the underlying potential is a prolate elliptical. Around an oblate potential, however, it is more difficult to create convincing shells, and when these do occur they are encircling in the symmetry plane but of a small extent in azimuth.

The shells should have traceable kinematics; the clearest signatures are velocity "glitches" of opposite sign at two shell positions at opposite sides of the galaxy. Hints of this were found for NGC 3923 (87), but these features are at best 2 results. Small excess stellar radial velocities at the position of the innermost ripple in NGC 1316 have been reported (16), but more data would be welcome.

Several other theories have been proposed for the shell formation, mostly referring to the existence of galactic winds in elliptical galaxies and subsequent star formation under certain conditions. Fabian et al. (40) suggested that the shells are regions of star formation in a shocked galactic wind. The color data obtained so far (24, 42) rule out the possibility of the shells being made up of recently formed stars with a normal initial mass function. Williams & Christiansen (154) explain the shells in terms of a wind-driven blast wave associated with an active nucleus phase early in the history of the galaxy. The blast wave sweeps the initial interstellar medium out of the galaxy into an expanding shell, which radiatively cools behind its 'leading shock front. If the nuclear activity ceases, the shell can cool sufficiently to allow a brief episode of star formation that is terminated when the shell is heated by supernovae. The model thus relies on the fact that these stars would spend most of their time near apogalacticum of their radial bound orbits, thus giving the appearance of shells. Yet shells are sharp features, so that one would have to assume that the stars would still continue to move in phase after 20 or more crossing times!