The baryon fraction in stars may be understood if approximately half the baryons that initially cooled within the virial radius have been expelled. Supernova feedback seems incapable of driving such a wind in the early gas-rich phase of galaxy formation because of the significance of cooling. However there are indications that vigorous winds are seen in the Lyman break galaxies, as inferred from the inverse P Cygni line profiles in template spectra and from the Lyman alpha forest suppression found within a Mpc or so of the Lyman break galaxies. In nearby galaxies where so-called superwinds are found, the observed mass outflow rate is comparable to the star formation rate, e.g. in NGC 253 [Strickland et al.2002]. Observations suggest that strong winds can be driven from typical galaxies, yet numerical simulations find that winds are stalled for the more massive halos.
What physics has hitherto been omitted from the simulations? The simulations are gravity-driven. Star formation may be, in extreme situations, pressure-driven. This could result in strong positive feedback via for example shock-triggered massive star formation and generate enhanced rates of correlated supernovae. that would in turn drive a transient wind.
Starbursts are known to drive winds. One mechanism appeals to bar formation. A merger creates a transient bar. The spin-down and dissolution of this bar torques and accelerates the gas which responds by an enhanced rate of dissipation and angular momentum loss, for example via cloud-cloud collisions. Bars are known to drive luminous starbursts, and minor mergers may drive ultraluminous starbursts, the ultimate sources of superwinds.
The gas-rich environment provides a plausible site for supermassive black hole formation. This phenomenon appears to occur contemporaneously with spheroid formation, as inferred from the remarkable correlation between central black hole mass and spheroid velocity dispersion. SMBH-driven outflows, manifest as a SMBH-energized component of the overall luminosity, may be important in early phases of galaxy formation. At least one explanation of the observed correlation appeals to outflows as the self-regulatory mechanism.
While the formation of structure in halos building up from clouds of weakly interacting dark matter is reasonably well understood, the formation of the luminous components of galaxies is founded on simplistic assumptions. A more realistic treatment of star formation in forming galaxies must ultimately modify many of the predictions that are failing to adequately confront the observations.
Feedback from supernovae is one solution that is being widely discussed. Supernova remnants deposit substantial energy into the interstellar medium, and are responsible for the observed multiphase structure of the interstellar gas. Some nearby low mass star-forming galaxies show a porous structure in the interstellar gas, where many overlapping shells and bubbles are seen that demonstrate that supernova input is playing an important role in pressurizing the interstellar gas. Some display an anti-correlation between X-ray and H emission, indicative of supernova-driven bubbles. The microphysics of the interface between the supernova-heated gas and the cold interstellar medium is inadequately resolved by the simulations. Plausibly, there are surprises in store.
Feedback reduces the efficiency of star formation. This is important for understanding the longevity of disk star formation. However attempts to use feedback in the phase of disk formation have had mixed success. A basic problem is that feedback delays star formation: if this works too well, all stars are young, in contrast to what is found in the outer parts of nearby disks such as M31. Numerical simulations that incorporate feedback find that disks are still too small, although possibly by only a factor of 2 [Sommer-Larsen, Gotz and Portinari2002].
If efficiency of star formation is to account for the observed colour-magnitude relation, then one needs a systematic reduction in efficiency of star formation with decreasing galaxy mass. Some evidence for this is found in the correlations between stellar surface brightness and total stellar mass in the SDSS study of 80,000 early and late-type galaxies [Kauffmann et al.2002]. In ellipticals, such an effect would help to simultaneously account for the trend of increasing [ / Fe] with galaxy mass, if the most massive galaxies are the oldest and formed stars most rapidly. However no detailed implementation has been made of the relevant physics, in large part because one needs to incorporate a multiphase interstellar medium.