The driving force behind starburst-driven winds is the mechanical energy from stellar winds and supernova events (e.g., Chevalier & Clegg 1985). This mechanical energy is quickly thermalized to produce a hot cavity with a temperature ~ 108 -1 K, where = Mtotal / Mejecta 1 is the mass-loading term. This over-pressured cavity expands through the ambient medium, sweeping this material up in the process to produce a bubble-like structure. The complex interaction between the wind and the ISM of the host galaxy has been the subject of several numerical simulations (e.g., MacLow & McCray 1988; Suchkov et al. 1994, 1996; MacLow & Ferrara 1999; D'Ercole & Brighenti 1999; Strickland & Stevens 2000; Silich & Tenorio-Tagle 2001). If radiative energy losses are negligible (probably a good assumption in some objects; e.g., Heckman et al. 2001), the bubble expands adiabatically through the galaxy ISM with a velocity ~ 100 n0-0.2 420.2 t7-0.4 km s-1, where n0 is the ambient nucleon density in cm-3, 42 is the rate of deposition of mechanical energy in 1042 erg s-1, and t7 is the age of the bubble in 107 years (e.g., Weaver et al. 1977).
A powerful starburst may inject enough energy to produce a cavity of hot gas that can burst out of the disk ISM, at which point the dense walls of the bubble start accelerating outward, become Rayleigh-Taylor unstable, and break up into cloudlets and filaments. If halo drag is negligible (probably not a good assumption in general), the wind fluid may reach terminal velocities as high as ~ 3000 -1 km s-1, well in excess of the escape velocity of the host galaxy. In contrast, the terminal velocities of clouds accelerated by the wind are more modest, of order ~ 600 340.5 w-0.5 r0, kpc Ncloud, 21-0.5, where 34 is the wind momentum flux in 1034 dynes, W is the solid angle of the wind in steradians, r0, kpc is the initial position of the cloud in kpc, and Ncloud, 21 is the column density of the cloud in 1021 cm-2 (Strel'nitskii & Sunyaev 1973; Heckman et al. 2000).
A critical quantity in all of these calculations is the thermalization efficiency, or the percentage of the mechanical energy from the starburst that goes into heating the gas. Unfortunately, this quantity is poorly constrained observationally. Most simulations assume a thermalization efficiency of 100%, i.e. none of the energy injected by the starburst is radiated away. In reality, this efficiency depends critically on the environment, and is likely to be significantly less than 100% in the high-density environment of powerful nuclear starbursts (e.g., Thornton et al. 1998; Strickland & Stevens 2000; Silich, Tenorio-Tagle, & Muñoz-Tuñón 2003). Galactic magnetic fields may also "cushion" the effects of the starburst on the ISM, and reduce the impact of the galactic wind on the host galaxy and its environment (e.g., Tomisaka 1990; Ferrière et al. 1991; Slavin & Cox 1992; Mineshinge et al. 1993; Ferrière 1998).