6. WHAT ENVIRONMENTS LEAD TO EXTREME STAR FORMATION?

In Figure 2 is shown the nearby M81-M82-NGC 3077 interacting group of galaxies (Yun et al. 1994). On the left is the Palomar Sky survey image, showing the stellar luminosity; on the right, is the VLA mosaic of the group in the 21 cm line of HI. The ties that bind this group are obvious in the 21 cm line emission. M82 is one of the best-known starburst galaxies, with L_IR ~ 6 × 10¹⁰ L, and an estimated 10⁴ - 10⁵ O stars (N_Lyc ~ 8 × 10⁵⁴ s^-1). NGC 3077 also has a modest starburst, of L_IR ~ 3 × 10⁸ L (N_Lyc ~ 2 × 10⁵² s^-1: Sanders et al. 2003), and M81 has a mildly active nucleus. Clearly the conditions for "extreme" star formation are favorable in this group. The atomic hydrogen of this interacting group has its own history, which is different from that of the stars within the galaxies; there is evidence that some of the starburst activity is caused by a delayed "raining down" of orbiting gas onto the galaxies several Myr after their closest encounters (Meier et al. 2001).

Figure 2. The M81-M82-NGC 3077 group. (Left) Palomar Sky Survey Image. (Right) VLA image of the 21cm line of HI. Yun et al. (1994). Credit: NRAO/AUI/NSF.

The most luminous infrared galaxies in the universe, with L_IR > 10¹¹ L, are merging and interacting systems, and these tend to be systems dominated by star formation. Our knowledge of the stellar distributions is greater than our knowledge of the gas: potentially many groups of galaxies have the connected appearance of the M81 group with tidal loops in atomic hydrogen, and perhaps even in molecular gas, since starburst galaxies are especially rich in molecular gas (Mirabel and Sanders 1989). Arrays with wide-field capability, such as the Allen Telescope Array, the Green Bank Telescope, and the SKA, are well-suited to the mapping of large HI fields in nearby galaxy groups.

Star formation efficiency (SFE) is an important characteristic of the star formation process, since it is a measure of how efficiently molecular clouds are turned into stars. SFE is closely tied to the "infant mortality" of SSCs described in the next section. There are numerous measures of the efficiency of star formation, broadly defined as the proportion of star formation per unit gas. The Schmidt law, in which stars follow a power law correlation with density (Schmidt 1959), or the corresponding Kennicutt law, in terms of gas surface density (Kennicutt 1998, 1998b) show that star formation on global scales in galaxies is correlated with gas density. The observable L_IR / M(H₂) is also used as an indication of SFE.

While from a global perspective the total gas content, H I + H₂, of galaxies appears to be well correlated with star formation tracers (Wong and Blitz 2002, Crosthwaite and Turner 2007), observations of star-forming regions in the Galaxy indicate that stars form from molecular gas clouds, rather than atomic. CO observations show a good correlation of L_CO with L_IR (Young et al. 1986, Sanders et al. 1986, Young and Scoville 1991, Young et al. 1996).

The atomic and molecular gas distribution in the spiral galaxy M83 is shown in Figure 3 along with an overlay of the GALEX ultraviolet image. This figure illustrates the general result that while the atomic hydrogen disk can far exceed the visible disk of a spiral galaxy, the optical portion of a spiral galaxy is primarily molecular gas. Gas that forms stars is molecular. While a good correlation of star formation tracers is seen with CO emission, an even tighter correlation is seen between star formation and the dense (n > 10^5-6 cm^-3) gas tracer HCN (Gao and Solomon 2004). This should not be surprising, since denser gas is more likely to form stars.

What complicates the study of molecular gas in star-forming regions is the necessity of using proxies, generally CO, to map out the distribution of H₂. A conversion factor between CO line intensity and H₂ mass seems to work well in the Galaxy, but will it do as well in ESF environments?

Figure 3. (Left) The barred spiral galaxy M83. Red is a VLA image of HI 21 cm line emission, Tilanus and Allen (1993), and yellow is CO emission mapped with the NRAO 12 Meter Telescope, Crosthwaite et al. (2002). (Right) Neutral gas with GALEX image overlay; GALEX image, Thilker et al. (2005), Gil de Paz et al. (2007).

Because of high energies of its first excited states, H₂ tends to be in the ground state for temperatures less than 100K. Most Galactic giant molecular clouds (GMCs) have temperatures of 7 - 12K (Sanders et al. 1985, Scoville et al. 1987), although clouds in starbursts can be warmer. By contrast, CO is relatively abundant, easily excited and thermalized, with a lowest energy level equivalent temperature E/k ~ 5.5K. For these reasons, CO lines are generally very optically thick. Yet CO is observed to be a good tracer of mass (Solomon et al. 1987). This is because Galactic disk GMCs appear to be turbulently supported against gravity, and near virial equilibrium (Myers 1983). GMCs, which consist of optically thick clumps with turbulent motions larger than systematic motions, have line profiles are Gaussian in spite of high optical depths, and "Large Velocity Gradient" (Sobolev approximation) radiative transfer holds (Wolfire et al. 1993). The empirically-determined Galactic conversion factor, X_CO = N_H2 / I_CO, is thus a dynamical mass tracer (Dickman et al. 1986, Solomon et al. 1987), in effect a Tully-Fisher relation for molecular clouds. Gamma ray observations indicate that a conversion factor of X_CO = 1.9 × 10²⁰ cm^-2 (K km s^-1)^-1 predicts H₂ mass to within a factor of two within the Galaxy, with some radial variation (Strong et al. 1988, 2004). As an indicator of dynamical mass, X_CO may actually be more robust than optically thin gas tracers in extreme environments, since mass estimates based on optically thin tracers depend upon temperature and relative abundance (Maloney and Black 1988, Dame et al. 2001).

While the CO conversion factor seems to work well in the Galaxy, and as a dynamical mass tracer may be more robust than tracers that are abundance-dependent, the association of CO and H₂ has not been extensively tested in extreme environments. There are clearly some situations in which X_CO fails to work well. The Galactic value of X_CO does not give good masses for the gas-rich centers of ultraluminous galaxies. In Arp 220, it overestimates the mass by a factor of ~ 5 due to a gas-rich nucleus, consisting of two counterrotating disks (Sakamoto et al. 1999), which are dominated by a warm, pervasive molecular gas in which systematic motions dominate over turbulence (Downes et al. 1993, Solomon et al. 1997, Downes and Solomon 1998). X_CO also appears to overpredict H₂ masses in the centers of local gas-rich spiral galaxies, including our own (Dahmen et al. 1998) by factors of 3 - 4. CO appears to systematically misrepresent H₂ mass in spiral galaxies when the internal cloud dynamics may be different from Galactic disk clouds, such as in the nuclear regions where tidal shear visibly elongates clouds, causing systematic cloud motions to dominate (Meier and Turner 2004, Meier et al. 2008). It may also fail where cloud structure may fundamentally differ from Galactic clouds, as in the LMC (Israel et al. 1986), where magnetic fields may be dynamically less important (Bot et al. 2007).

For an understanding of the links between star formation and molecular gas in ESF regions we require improved confidence in molecular gas masses in environments that are atypical of the Galaxy. Systematic studies of molecular clouds in different tracers of molecular gas, including dust, in ESF galaxies at high resolution in the millimeter and submillimeter with ALMA, CARMA, Plateau de Bure, and SMA will shed light on when we can confidently use CO as a tracer of molecular gas mass, and under what conditions it ceases to be a good tracer.

Environmental factors other than gas mass are also important in the fostering of star formation, but these are not as yet well understood. Tacconi and Young (1990) concluded that the efficiency of massive star formation is higher in spiral arms than between the arms, consistent with the "strings of pearls along the spiral arms" description of nebulae by (Baade (1957)). Clearly star formation is enhanced by spiral arms, but how their large-scale influence trickles down in a turbulent GMC to a parsec-scale core is not at all clear (Padoan et al. 2007). Starburst rings also appear to facilitate star formation, particularly young SSCs (Barth et al. 1995, Maoz et al. 1996, Maoz et al. 2001). SFE appears to steadily increase with the ferocity of the star formation. While L_IR / M_H2 ~ 4 L / M for the Galaxy, it is ~ 10 - 20 L / M in regions of active star formation, ~ 20 - 100 in ULIRGs (Sanders et al. 1986). These studies rely on L_IR, which may have contributions from older stellar populations. Studies of star formation efficiency using tracers of recent star formation are ongoing.

Extreme star formation should require extreme amounts of molecular gas, and luminous infrared galaxies have plenty of it (Sanders and Mirabel 1985, 1996 Young et al. 1986, Downes et al. 1993). The Antennae interacting system, with its thousands of young SSCs, contains exceptional amounts of gas. A CO image made with the Owens Valley Millimeter Array by Wilson et al. (2000, 2003) is shown on the HST image of Whitmore et al. (1999) in Figure 4. The greatest concentration of molecular gas is in the dusty, obscured region between the two galaxies where the brightest infrared emission is found (Mirabel et al. 1998). There is an estimated 10⁹ M of molecular gas in the Antennae, supporting a current star formation rate of N_Lyc ~ 10⁵⁴ s^-1 (Stanford et al. 1990, for D = 13 Mpc, and X_CO ~ 2 × 10²⁰ cm^-2 K km s^-1). Based on the total mass of K-band selected clusters, the SFE is about 3 - 6% (Mengel et al. 2005), slightly higher than Galactic efficiencies on these scales (Lada et al. 1984) but not by much. H₂ emission observed using Spitzer in the dusty collision region suggests that "pre-starburst" shocks may trigger the star formation there (Haas et al. 2005) as seen in other interacting ULIRGS (Higdon et al. 2006). This is a good illustration of the "Super Giant Molecular Clouds" posited for interacting systems by Harris and Pudritz (1994), Harris (2002). Like the stellar cluster mass function, which is power law with ~ -2, the mass function of giant molecular clouds in the Antennae is also power law, but with a slightly different slope, ~ -1.4 (Wilson et al. 2003).

Figure 4. CO in the Antennae. Contours are emission in the J = 1 - 0 line of CO at 3mm, imaged with the Owens Valley Millimeter Array. In color is the HST image. Wilson et al. (2000), Whitmore et al. (1999). Credit: C. Wilson.

A counterexample to the "lots of gas, lots of stars" theory is the case of NGC 5253. In this starburst dwarf galaxy, the star formation efficiency appears to be extremely high, SFE ~ 75% on 100 pc scales, with H₂ masses based on both CO (Turner et al. 1997, Meier et al. 2002) and dust emission (Turner et al. 2008, in prep.) This SFE is nearly two orders of magnitude higher than generally seen on GMC spatial scales in the Galaxy. NGC 5253 has several hundred young clusters, including several SSCs (Gorjian 1996, Calzetti et al. 1997). Why is this galaxy so parsimonious in its usage of gas compared to the Antennae? One difference between this starburst and that of the Antennae is that NGC 5253 is a dwarf galaxy, with an estimated mass of ~ 10⁹ M, which was probably originally a gas-poor dwarf spheroidal galaxy that has accreted some gas (Caldwell and Phillips 1989). Unlike the Antennae, which are in the midst of a full-blown major merger, NGC 5253 is relatively isolated, although part of the Cen A - M83 group (Karachentsev et al. 2007). The prominent dust lane entering the minor axis is the probable cause of the starburst. Molecular gas is present in the dust lane, and this gas is observed to be falling into the galaxy near the current starburst (Meier et al. 2002, Kobulnicky and Skillman 2008). Radio recombination line emission imaged at high resolution with the VLA of the central "supernebula" shows a velocity gradient in the same direction as that of the infalling streamer (Rodríguez-Rico et al. 2007). High star efficiency is a necessary condition that the SSCs can evolve into globular clusters (Goodwin 1997). NGC 5253 may be the best case yet for a galaxy in which the clusters might survive to become globular clusters.

The high resolution and sensitivity of ALMA, the new CARMA array, Plateau de Bure and SMA will allow many more such starburst systems to be imaged in molecular lines for study of the variation in SFE with environment. Star formation efficiency is a key parameter in the formation of long-lived clusters.