Unlike stars, gas responds to pressure forces as well as gravity; moreover, gas flows develop shocks whereas streams of stars freely interpenetrate. Even without the complications of star formation, the dynamics of gas in interacting galaxies is a difficult problem. But dissipative dynamical systems generally possess attractors; in the long run, most trajectories are captured by one attractor or another. Consequently, gas in interacting galaxies tends to end up in a few stereotypical structures.
The thermodynamic history of the gas is probably the factor which determines its fate. To date, most simulations treat gas thermodynamics rather crudely; the cooling function is cut off at Tc = 104 K to prevent the gas from ``curdling'', and the resulting behavior is basically that of an isothermal fluid with T = Tc (Barnes & Hernquist 1996, hereafter BH96). Improving on the present treatments may require including star formation and feedback; one possible approach to this difficult problem is described in Section 4. The rest of this section reviews results obtained with and without cooling in an attempt to anticipate the results of more realistic experiments.
Work by several investigators confirms that tidal perturbations of gas-rich disk galaxies result in rapid gas inflows (Icke 1985, Noguchi 1988, Hernquist 1989, Barnes & Hernquist 1991). The immediate physical cause of these rapid inflows is a systematic transfer of angular momentum from the gas to the disk stars; tidally perturbed disks develop bars (or other non-axisymmetric structures) which exert gravitational torques on the gas (Combes et al. 1990, Barnes & Hernquist 1991). Such inflows require strong shocks and rapid cooling, which work together to drive the gas irreversibly toward the center of the potential. As a perturbed disk settles down the gas often converges onto kpc-scale closed orbits aligned with the stellar bar (BH96).
Dissipative mergers between disk galaxies lead to further inflows, with large amounts of gas collecting in ~ 0.2 kpc-scale clouds (Negroponte & White 1983, Barnes & Hernquist 1991). These nuclear clouds contain material driven in toward the centers of galaxies by earlier tidal interactions; during the final merger, the gas again loses angular momentum to the surrounding material. It seems likely that the same physical mechanism lies behind the inflows in perturbed disks and in merger remnants; in both cases the entropy of the system grows as gas falls inwards.
A different fate awaits the gas which does not suffer strong shocks and subsequent cooling in the early stages of an encounter. This material does not participate in rapid inflows, and retains much of its initial angular momentum. Consequently, it tends to collect in an extended, rotationally supported rings or disks; one such example has already been presented in Figure 1. In merger remnants, such disks may be strongly warped by gas falling back from tidal tails (BH96). Early-type galaxies with warped gas disks include NGC 4753 (Steiman-Cameron et al. 1992) and NGC 5128 (van Gorkom et al. 1990); though these disks are usually attributed to accretions of gas-rich satellite galaxies, some may actually result from major mergers.
The two outcomes just described - nuclear clouds or extended disks -
seem to be the only real attractors available to dissipative gas in
merger simulations. However, if the gas fails to cool then another
outcome is likely - a pressure-supported atmosphere about as extended
as the stellar distribution
(BH96).
Though most phases of the ISM
cool efficiently, initially hot gas (T 
105 K,
n 
10-3 cm-3) could be shock-heated during a merger and
might produce envelopes of X-ray gas like those found around some
elliptical galaxies. On the other hand, X-ray observations of the
Antennae
(Read et al. 1995)
and Arp 220
(Heckman et al. 1996)
reveal apparent outflows of up to 109
M
of hot
gas. The properties of these outflows are
inconsistent with shock-heating and seem to require significant
injections of mass and energy from merger-induced starbursts.