We study large-scale winds driven from uniformly bright self-gravitating disks radiating near the Eddington limit .
We show that the ratio of the radiation pressure force to the gravitational force increases with height to a maximum of twice its value at the disk surface .
Thus , uniformly bright self-gravitating disks radiating at the Eddington limit are fundamentally unstable to driving large-scale winds .
These results contrast with the spherically symmetric case , where super-Eddington luminosities are required for wind formation .
We apply this theory to galactic winds from starburst galaxies that approach the Eddington limit for dust .
For hydrodynamically coupled gas and dust , we find that the asymptotic velocity of the wind is v _ { \infty } \simeq 3 \langle v _ { rot } \rangle and that v _ { \infty } \propto { SFR } ^ { 0.36 } , where \langle v _ { rot } \rangle is the mean disk rotation velocity and SFR is the star formation rate , both of which are in agreement with observations .
However , these results of the model neglect the gravitational potential of the surrounding dark matter halo and a ( potentially massive ) old stellar bulge , which both act to decrease v _ { \infty } .
A more realistic treatment shows that the flow can either be unbound , or bound , forming a ‘ ‘ fountain flow ’ ’ with a typical turning timescale of t _ { turn } \sim 0.1 - 1 Gyr , depending on the ratio of the mass and radius of the starburst disk relative to the total mass and break ( or scale ) radius of the dark matter halo or bulge .
We provide quantitative criteria and scaling relations for assessing whether or not a starburst of given properties can drive unbound flows via the mechanism described in this paper .
Importantly , we note that because t _ { turn } is longer than the star formation timescale ( gas mass/star formation rate ) in the starbursts and ultra-luminous infrared galaxies for which our theory is most applicable , if starbursts are selected as such , they may be observed to have strong outflows along the line of sight with a maximum velocity v _ { max } comparable to \sim 3 \langle v _ { rot } \rangle , even though their winds are in fact bound on large scales .