Ever since a dissipational component was first included in numerical simulation of interactions it has been clear that large amounts of gas can be efficiently driven into the central regions [39, 40, 41]. In this section I examine individual models for clues as to what parameters play the largest role in driving gas to high densities. This is not intended to be a review of all the simulations (for that see ), but rather an assessment of the results which might have the most relevance for understanding the observations discussed above.
Progenitors: A very interesting result to emerge from the numerical studies is that the progenitor structure ought to play a dominant role in the star formation history (SFH) [43, 44]. This is illustrated in Figure 4. In the upper sequence the presence of a bulge stabilizes the disks against bar formation, lowering the pre-merger star formation levels, but leaving more gas in the progenitors to fuel a strong starburst when they finally merge. Because of the peaked appearance, it is tempting to associated this SFH with the ULIR systems, but it is important to realize that the vertical axis is a relative SFR, and the absolute SFR between the two curves could differ due to other factors (see below). Indeed, the fact that Arp 299 is so luminous while the nuclei are still well separated suggests that it follows the lower SFH in Fig. 4. This interpretation is supported by the H I observations, which suggests two late type progenitors . NGC 4038/9 on the other hand appears to have one late type (NGC 4038) and one earlier type (NGC 4039) progenitor , and may have a SFH more like the upper curve in Fig. 4, with its most active star forming period still to come. These considerations suggest that while a dense bulge may push certain systems into the ULIR regime, it need not be a strict requirement. They also point to the need for better constraints on the past and future star forming histories to facilitate comparisons with models.
Figure 4. Time history of two interaction induced starbursts. In the bottom sequence, the lack of a dense bulge allows strong bar-induced inflows to develop early in the encounter. The relatively mild starburst which ensues consumes much of the available gas supply, leading to a less extreme burst at the final merger. Adapted from Mihos & Hernquist .
Gas content: There is over three times as much molecular gas in Arp 299 than NGC 4038/9, although the relative gas contents are similar (MH2 / LK = 0.5, 0.4). Could this be responsible for the different levels of activity? Clearly the level of star formation must be closely connected to the amount of fuel available, but plots of SFE or LIR vs. MH2 show notoriously large amounts of scatter (e.g. Fig. 2a). So while the gas content may help shift the SFR upward for some systems (see Gao et al. these proceedings), high gas mass alone is not a sufficient condition for fueling an ULIR phase. Olson & Kwan  conducted one of the few simulations to explore the effects of different gas masses and distributions. Using a code in which the SFR is proportional to the cloud collision rate, they find that doubling the gas content in one of the disks in a disk-disk merger doubles the SFR, therefore keeping the SFE (= SFR / Mgas) constant, while splitting the same amount of gas between two galaxies leads to a 30% higher SFR and SFE than putting it all into one of the system. This suggests that how the gas is distributed and how it collides is as important as how much gas is present. These studies should be continued with a wider range of encounter parameters.
Spin geometries: The two systems under consideration here differ in their spin geometries, with Arp 299 undergoing a prograde-retrograde encounter [45, 48] and NGC 4038/9 undergoing a prograde-prograde encounter . Since retrograde encounters fail to raise strong tails , there will be more gas left in the inner regions at the late stages of merging where it will be violently perturbed as the encounter progresses. The simulations of Barnes & Hernquist  show that the fraction of gas at high densities depends on encounter geometry, with retrograde encounters leading to higher quantities of dense gas than prograde encounters. In Arp 299, the highest gas column densities are found in the nucleus of the disk experiencing the retrograde encounter (IC 694). Mihos & Hernquist [43, 44] also evaluated the effects of spin geometry, and while they showed that it had a much less dramatic effect on the SFRs than progenitor structure, geometry still made a difference of about a factor of two in the relative SFRs. If there is either a threshold to the onset of very efficient star formation activity, or if SFE is a function of the fraction of the gas at very high densities, it is possible that spin geometry plays a role beyond that attributed to it in these studies.
The reality is that all of the factors are probably playing a part, although in what combination and order of importance is not clear. Continued parameter studies are needed in concert with detailed observations of individual systems in order to discriminate between the different processes. It is also important to compare the actual distributions and dynamics of the star forming regions and dense gas with the predictions of the simulations in order to discriminate between different numerical formalisms for star formation and gas dynamics. Only by doing this will we know how far to trust the models or how to better conduct our observations.