As demonstrated by Toomre & Toomre (1972) (and re-affirmed many times since, e.g. Barnes 1998), tidal features develop kinematically. As a result, they have a simple kinematic structure, with energy and angular momentum increasing monotonically with distance along the tail (Hibbard & Mihos 1995). Because of this simple kinematic structure, H I observations provide a uniquely useful constraint on N-body simulations of gas-rich mergers (e.g. Combes 1978, Combes et al. 1988, Hibbard & Mihos 1995, Yun 1997, Barnes 1998). While the primary parameters that are fit in this exercise are the physically uninteresting angles describing the orientations of the disks and the viewing perspective, the model matching gives us the confidence to explore the evolutionary history of mergers beyond the best-fit time. By running the simulations forward in time, we can explore the late-stage merger evolution for clues on the expected morphology of the remnants and the distribution of material at large radii in the halos around the remnants.
Because much of the tidal material remains bound to the remnant, it will eventually reach an apocenter, turn around, and move back inwards in the potential. There will therefore be a constant rain of tidal material back onto the remnant. Material which falls back while the potential is still violently relaxing will scatter and be mixed throughout the remnant body. Material which returns after the potential has relaxed will wrap coherently, forming shells, loops and other "fine structures" (Hernquist & Spergel 1992). Because of its high energy and angular momentum, the material which falls back later will fall back to larger and larger radii, forming loops rather than shells. At late times, the material outside of the loops will have a low density and may be ionized by the intergalactic UV field (Corbelli & Salpeter 1993, Maloney 1993) or the remnant itself (Hibbard, Vacca & Yun 1999). We would therefore expect evolved disk merger remnants to exhibit partial rings of H I with a rotational signature (since the loops correspond to turning points where the radial velocity goes to zero), lying outside the remnant body. This is exactly what has been found around a number of shell galaxies (Schiminovich et al. 1995; Fig. 2b-d).
Meanwhile, the loosely bound tidal material in the outer regions continues to travel outward. This material has radial periods ~ many Gyr and azimuthal periods even longer than this (Hibbard 1995). As a result, the tidal material will not give rise to a smooth, spherical halo of material; instead there will be specific regions of higher column density material with a low filling factor extending to very large radii. At late times, the atomic gas will be too diffuse to be detected in emission, and may anyways be largely ionized by the intergalactic UV field. Therefore the tidal features mapped in H I are likely the denser neutral peaks of a more extended distribution. This material should be detectable in absorption against background sources (Carilli & van Gorkom 1992).