Already many years ago galactic winds were suggested as a possible gas transfer mechanism (De Young 1978). Many supernova explosions provide large amounts of thermal energy, which can drive an outflow from a galaxy (see reviews by Heckman 2003 and Veilleux et al. 2005). A correlation between starburst galaxies and wind is well established through the finding of hot gas around starburst galaxies (e.g. Dahlem et al. 1998).
Spectacular examples of such winds are seen in the galaxies M 82 (Lynds & Sandage 1963) and NGC 253 (Demoulin & Burbidge 1970).
The outflows consist of a complex multi-phase medium of cool, warm and hot gas (see e.g. the Chandra observation of NGC 4631, Wang et al. 2001, Fig. 3). The morphologies of the optical emission-line gas and the X-ray emission as observed with Chandra have been found to be quite similar (Strickland et al. 2002, Cecil et al. 2002). Such correlations can be used to understand the interaction between the gas in the bubbles and the interstellar medium (ISM). It was found that the accelerated ISM can reach high velocities of several hundred km s-1 (Heckman et al. 2000, Rupke et al. 2002).
Figure 3. Chandra observation of the edge-on spiral galaxy NGC 4631. It shows the presence of a giant diffuse X-ray emitting corona. The corona has a temperature of (2-7) × 106 K and extends as far as 8 kpc away from the galactic plane (from Wang et al. 2001).
Martin (1999) gives an often used recipe for simulations: the mass outflow rate is proportional to the star formation rate SFR:
with being typically in the range of 1-3. By comparing different techniques Heckman (2003) also finds that the outflow rate is of the order of the star formation rate. The SFR can be estimated from observations, e.g. from far-infrared luminosities LFIR
(Kennicutt 1998). Another way to estimate the SFR is to use the tight relation between the SFR and the surface density of the gas gas
(Schmidt 1959) with SFR being the surface density of the SFR and the index N having measured values between 1 and 2. Only at densities below a critical threshold value the SFR is almost completely suppressed (Kennicutt 1989). Alternatively the dynamical time t* can be included
with the dynamical time t* being the local orbital timescale of the disk (Kennicutt 1998). For hydrodynamic simulations this has been extended to include the fraction of stars lost by supernova explosions by Springel & Hernquist (2003)
with * being the density of stars, c being the cold gas density in the disk and being the fraction of stars lost by supernova explosions. For a typical initial mass function and a mass threshold of 8 M for the supernovae a = 0.1 is used. These are of course only statistical estimates.
Other attempts to quantify the outflow rate take into account physical parameters like those describing the galaxy's gravitational potential and the effect of cosmic rays (Breitschwerdt et al. 1991). Using the Bernoulli equation, Kronberger et al. (2008b) recently derived an analytic approximation for the mass loss due to thermally driven galactic winds. The mass loss per unit area at a given position of the galactic disc reads
with 0 being the gas mass density, u0 the bulk velocity of the gas, 0 the gravitational potential, c0 the sound speed (all four quantities at the given position), vesc the escape velocity, and the adiabatic index of the thermal plasma. Hydrodynamic simulations of outflows have also been performed (Tenorio-Tagle & Munoz-Tunon 1998, Strickland & Stevens 2000).
Starbursts with subsequent winds can also be caused by cluster mergers (Ferrari et al. 2003, Ferrari et al. 2005, Ferrari et al. 2006), because in such mergers the gas is compressed and shock waves and cold fronts, which trigger star formation, are produced (Evrard 1991, Caldwell et al. 1993, Wang et al. 1997, Owen et al. 1999, Moss & Whittle 2000, Bekki & Couch 2003).