2005ApJ...621..227M

Measurements of interstellar Na I λλ5890, 5896 absorption lines in 18 ultraluminous infrared galaxies (ULIGs) have been combined with published Na I data, to reassess the dependence of galactic outflow speeds on starburst luminosity and galactic mass. The Doppler shifts reveal outflows of relatively cool gas in 15 of 18 ULIGs with an average outflow speed at the line center of 330±100 km/s. The relation between outflow speed and star formation rate (SFR), defined by the distribution's upper envelope over 4 orders of magnitude in SFR, demonstrates that winds from more luminous starbursts accelerate interstellar gas to higher speeds roughly as v∝SFR^0.35. This result is surprising since, in the traditional model for starburst-driven winds, these relatively cool gas clouds are accelerated by the ram pressure of a hot, supernova-heated wind that exhibits weak (if any) X-ray temperature variation with increasing galactic mass. The lack of evidence for much hotter winds is partly a sensitivity issue, but the Na I velocities in ultraluminous starbursts actually are consistent with acceleration by the tepid wind, indicating that a hotter component is unlikely to dominate the momentum flux. The Na I velocities in the dwarf starburst winds do not reach the terminal velocity of a hot wind at the measured temperature of kT∼0.73 keV, a result that could be interpreted simply as evidence that the hot superbubbles are too confined in dwarf starbursts to generate a free-flowing wind. A dynamically motivated scenario, however, is that the dwarf starburst winds simply lack enough momentum to accelerate the clouds to the velocity of the hot wind. Among the subsample of starbursts with well-constrained dynamical masses, the terminal outflow velocities are always found to approach the galactic escape velocity. Motivated by a similar scaling relation for stellar winds, the galactic Eddington luminosity for dusty starbursts is shown to be within the range measured for ULIGs. If radiation pressure on dust grains, coupled to the cool wind, is indeed important for galactic wind dynamics, then feedback is stronger in massive galaxies than previously thought, helping shape the high-mass end of the galaxy luminosity function. Regardless of the nature of the acceleration mechanism in ULIGs, the mass flux of cool gas estimated from these data demonstrates that starburst-driven winds transport significant gas during the assembly stage of field elliptical galaxies, a factor that helps explain the rapid decline in SFR in these systems inferred from elemental abundance ratios.