Even after ~ 13 Gyr of evolution, many present-day galaxies still have significant gas supplies available for star formation during interactions and mergers. The median gas fraction of neutral hydrogen alone, expressed relative to the total baryonic mass, is 15%, 10%, and 4% for dIrr, Sc, and Sa galaxies, respectively (Roberts & Haynes 1994). Even more impressive is the median fraction of all gas relative to the dynamical mass, MH I + H2 / Mdyn 25%, 15%, and 3% for the same three types of galaxies (Young & Scoville 1991). Since even at high redshifts no galaxy can be more than 100% gaseous, the relatively high gas fractions of local Sc and later-type galaxies tell us that there is less than one order-of-magnitude difference between the fractional gas contents of many local disk galaxies and their high-z counterparts. Hence, the often-heard objection that mergers at z 2-5 were completely different from local mergers is weak, and studying local mergers can, in fact, help us understand high-z mergers.
Molecular gas masses observed in local mergers and distant quasars support this point of view. Locally, MH2 ranges from 0.6 × 1010 M for an aging merger remnant like NGC 7252 through 1.5 × 1010 M for an ongoing merger like The Antennae to ~ 3 × 1010 M for extreme ULIRGs, while MH2 2 × 1010 M in a QSO at z = 2.56 (Solomon et al. 2003) and also in one at z = 6.42 (Walter et al. 2003).
With gas amply available for star formation during mergers at both low and high redshifts, what are the dynamical triggers for merger-induced starbursts?
During disk-galaxy interactions gravitational torques arise between the bars induced in the gas and among the stars. Because the gaseous bar leads, the gas experiences braking, which in turn leads to its infall. The resulting pressure increase in the gas has long been understood to be the root cause of interaction-induced starbursts (Noguchi 1988; Hernquist 1989; Barnes & Hernquist 1991).
Although the vehemence of this pressure increase clearly depends on such factors as the presence or absence of a central bulge (Mihos & Hernquist 1996) and the encounter geometry (Barnes & Hernquist 1996), the small-scale details of the dynamical triggers have been less clear until recently.
It now appears that shocks play a very major role, both in affecting the spatial distribution of star formation (Barnes 2004) and in squeezing giant molecular clouds (hereafter GMCs) into rapid star and cluster formation (Jog & Solomon 1992; Elmegreen & Efremov 1997). Questions such as whether cloud-cloud collisions are important and what role magnetic fields play are just beginning to be addressed by observers, as detailed below.
Merger-induced shocks can be fierce. In a simulation of two merging equal-mass disks (Barnes & Hernquist 1996), massive rings of dense gas form around the center of each galaxy and collide, during the third passage, head-on with a relative velocity of ~ 500 km s-1! This extreme final smash is made possible by the rapid ~ 90% loss of orbital angular momentum that the two gas rings experience within ~ 1/4 disk-rotation period.
Even during milder encounters shock-induced star formation may dominate. Because of the ubiquity of shocks in mergers involving gas, Barnes (2004) proposes a new star formation recipe that, in addition to the local gas density gas, includes the local rate of mechanical heating due to shocks and PdV work:
Assuming that energy dissipation balances the heating rate, setting m > 0 and n = 1 yields purely shock-induced star formation, while setting m = 0 and n > 1 yields density-dependent star formation (with Schmidt's law as a special case). Barnes compares simulations run for these two limit cases of star formation with observations of The Mice (see Fig. 1 here, and color figs. 3 & 4 in his paper) and shows convincingly that shock-induced star formation is spatially more extended and occurs earlier during the merger, which is in significantly better accord with the observations.
Figure 1. Simulations of star formation in The Mice for (left) density-dependent and (middle) shock-induced star formation recipes; halftones mark old stars, points mark star formation. (Right) Star formation rate vs. time for density-dependent (solid line) and various shock-induced (dashed & dotted) star formation recipes (from Barnes 2004).
One long-standing question has been whether high-velocity cloud-cloud collisions (50-100 km s-1) contribute significantly to the triggering of starbursts (Kumai, Basu, & Fujimoto 1993) or not. To address this issue, Whitmore et al. (in prep.) measured H velocities of the gas associated with young massive clusters in The Antennae, using HST/STIS and positioning the 52" long slit of STIS along different groups of clusters. From many clusters in 7 regions the measured cluster-to-cluster velocity dispersion is < 10-12 km s-1, which argues against high-velocity cloud-cloud collisions as a major trigger of starbursts. Instead, the squeezing of GMCs by the general pressure increase in the ISM (Jog & Solomon 1992) appears favored.
The role of magnetic fields in triggering starbursts in mergers remains unclear at present, but is beginning to be studied observationally. Chyzy & Beck (2004) have used the VLA to produce detailed maps of radio total power and polarization in NGC 4038/39 (see Chyzy's poster paper). The derived mean total magnetic field of ~ 20 µG is twice as strong as in normal spirals and appears tangled in regions of enhanced star formation. The field peaks at ~ 30 µG in the southern part of the Overlap Region, suggesting strong compression where the star formation rate is highest. The crucial question to address over the coming years is whether the enhanced magnetic field observed in mergers merely traces compression, or whether it contributes to the triggering of starbursts.