To understand how this efficient mode of star formation is triggered, it is instructive to compare two nearby mergers at apparently similar stages of merging but at two quite different levels of star formation activity. The mergers I will examine are "The Antennae" (NGC 4038/9, Arp 244; Vo = 1630 km s-1) and Arp 299 (NGC 3690/IC 694; Vo = 3080 km s-1). Both the tidal structure and inner disks of these systems are shown in Figure 3 along with distribution of the cold gas components.
Figure 3. Two on-going mergers. (a)&(c) NGC 4038/9 and (b)&(d) Arp 299. In the upper two panels, a deep R-band image of the entire system is shown with VLA H I contours overlaid, while in the lower two panels a B-band image of the inner regions is shown with OVRO 12CO(1-0) contours overlain. The optical data were obtained by the author; the VLA data are from Hibbard & Yun 1996  (Arp 299) and Hibbard & van der Hulst in preparation (NGC 4038/9); the OVRO data are from Aalto et al. 1997  (Arp 299) and Stanford et al. 1990  (NGC 4038/9). These systems appear to be at similar stages of merging, with their disks in contact and their nuclei still separate.
The evidence that these systems are at a similar stage of merging is the following: both have a long tidal tails (130 kpc for NGC 4038/9 assuming a distance of 25 Mpc; 180 kpc for Arp 299 assuming a distance of 48 Mpc), suggesting several hundred Myr since first orbital periapse ( 500 Myr for NGC 4038/9; 700 Myr for Arp 299); the disks of their progenitors are highly distorted and in physical contact, yet still distinct; and their nuclei are still well separated (8 kpc and 4 kpc, respectively). Both systems have similar CO distributions (Fig. 3c-d), with concentrations of cold molecular gas near both nuclei and a significant concentration at the region of disk overlap. Despite these similarities, Arp 299 is almost an order of magnitude more luminous in the infrared than NGC 4038/9, indicating a much higher MSFR (see Table 1).
|LIR a||MH2 b||LIR / LK||MSFR||H2||SFE|
|( × 1010 L)||( × 109 M)||(M yr-1)||(M pc-2)||(L M-1)|
|a LIR is split between components using 15 µm ISOCAM measurements for NGC 4038/9 , and 32 µm measurement for Arp 299 .|
|b From OVRO CO observations of NGC 4038/9  and Arp 299 .|
In Table 1 we use NIR flux measurements [27, 28] to divide the IR luminosity among the various components. This comparisons shows that a major difference between the systems is in the amount of nuclear star forming activity: in NGC 4038/9, the overlap region is the most active star forming region, both in terms of the MSFR and the SFE. In Arp 299, on the other hand, the overlap region has similar properties as that in NGC 4038/9, but both nuclei outshine this region by large amounts. We note that observations argue against an energetically dominant AGN in any of the nuclei [13, 27, 28, 29] (but see ).
Similar differences are seen in the column densities of molecular gas: there is five times as much gas in the central kpc of the nuclei of Arp 299 as in NGC 4038/9. As a result, the peak column densities are over an order of magnitude higher in the nuclei of Arp 299 as in the nuclei of NGC 4038/9 or the overlap regions. The tight correlation between H2 and LIR shown in Fig. 2 shows this to be a general result for mergers over a broad range of IR luminosity . While it is still possible that MH2 has been overestimated [25, 26] a similarly tight correlation exists between LIR and LHCN, as well as between SFE and LHCN / LCO , showing that the fraction of gas at very high densities (n > 105 cm-3) is greatly increased in ULIR systems. As these fractions are much higher than those found in normal spirals, this indicates that MSFR is intricately linked to the amount of gaseous dissipation [19, 26, 36].
It therefore seems that in order to understand how such efficient periods of star formation are induced, we need to understand how so much gas attains such high gas densities. Most phenomenological scenarios of star formation predict that the MSFR and the SFE are proportional to the gas density to some power (e.g. [37, 38]), although they do not address in detail when and where such densities are attained. For this, we turn now to results from numerical simulations.