3.3 Outflows and Inflows
When massive stars are about to end their lives they explode as a supernovae (SNe). The energy output from a SN is over a short period, comparable to that of a whole galaxy. In a galaxy with a high local star formation rate, the collective action of supernovae may lead to a galactic superwind, which may cause loss of gas. Stellar winds can also contribute to the energetics of the ISM at the very early stage of a starburst (Leitherer et al. 1992). The relative importance of winds compared to SNe increases with metallicity.
A continuous wind proportional to the star formation rate has been applied in models predicting the evolution of starburst galaxies. But since different elements are produced on different timescales, it has been proposed that only certain elements are lost or in different proportions hence reducing the effective net yield of those metals as compared to a simple chemical evolution model (Matteucci and Chiosi 1983, Edmunds 1990). The SNe involved in such a wind are likely to be of type II because type Ia SNe explode in isolation and will less likely trigger chimneys from which metals can be ejected out of the plane of the galaxy. In this framework O and part of Fe are lost while He and N (largely produced by intermediate stars) are not. This would result in a cosmic dispersion in element ratios such as N/O between galaxies that have experienced mass loss and those that have not.
In a dwarf galaxy which has a weaker gravitational potential, these effects may result in gas loss from the galaxy. Recently galactic winds have been observationally investigated in dwarf galaxies (e.g. Israel and van Driel 1990, Meurer et al. 1992; Marlowe et al. 1995; Martin 1996, 1998). In VII Zw403 Papaderos et al. (1994) detected extended X-ray plumes which they interpreted as the result of outflows of hot gas. Lequeux et al. (1995) and Kunth et al. (1998) have shown that the escape of the Ly photons in star-forming galaxies strongly depends on the dynamical properties of their interstellar medium. The Lyman alpha profile in the BCG Haro2 indicates a superwind of at least 200 km/s, carrying a mass of ~ 107 M, which can be independently traced from the H component (Legrand et al. 1997a). However, high speed winds do not necessary carry a lot of mass. Martin (1996) argues that a bubble seen in IZw18 (see also Petrosian et al. 1997) will ultimately blow-out together with its hot gas component. Although little is known about the interactions between the evolving supernova remnants, massive stellar bubbles and the ISM it is possible that an outflow takes the fresh metals with it, in some cases leaves a galaxy totally cleaned of gas.
Will the gas leave a galaxy or simply stay around in the halo? Tenorio-Tagle et al. (1999) point out that superbubbles may initially expand with speeds that well exceed the local escape velocity of the galaxy but their motion into the gaseous halo causes a continuous deceleration lowering the velocity to values well below the escape speed. In such a case, ejecta condense into a cold phase, forming droplets that fall back and settle down onto the disc of the galaxy hence changing the composition of the ISM (the ``Galactic fountain'' model). Similarly in chemodynamical models (Hensler and Rieschick 1998, and references therein), the gas cools and falls back. Modelling the effect of SNe feedback on the ISM, De Young and Heckman (1994) suggested that the smallest dwarfs could have their entire ISM removed by a superwind. However, using models including dark matter, MacLow and Ferrara (1999) and Ferrara and Tolstoy (1999) conclude that winds are not very efficient in ejecting the ISM. Outflows are in most cases confined to the galaxy and ``blow-away'' occurs only for the smallest (luminous mass Mlum < 107 M) galaxies considered, while in other cases the mass loss is very modest. Winds may however still be efficient in ejecting fresh metals. However, Ferrara and Tolstoy (1999) nevertheless argue that outflows are not likely to be more metal rich than the average ISM value. Moreover, in their model, the SFR is a function of mass density which results in a mass-metallicity relation. Since the least massive dwarfs loose their entire ISM after the first star formation event, this results in a minimum expected ISM metallicity for a gas rich dwarf of 12 + log(O/H) = 7.2, i.e. the abundance of IZw18. Many assumptions go into these calculations, which must be further examined. Murakami and Babul (1999) showed that in high density environments the IGM pressure could confine outflows to the parent galaxies, inhibiting mass loss (cf. Babul and Rees 1992) .
It is clear that the role of galactic winds in regulating the chemical evolution is not a settled issue yet. If the metallicity-luminosity relation (cf. Sect. 4.1, 4.2 and 7.1) holds from gas-rich to gas-poor systems then the loss of metals due to galactic winds should be a second order effect (Skillman 1997).
It is also possible that a galaxy is subject to infall of gas, although evidence for this is scarce. Infalling gas may come from external galaxies, stolen in the process of interaction or from an external origin, perhaps via isolated pristine H I clouds (if such clouds exist). There is evidence that blue compact galaxies and low surface brightness galaxies (LSBGs) are sometimes associated with H I clouds (Taylor 1997), which in general have optical counterparts (Taylor, private communication). A third possibility is that gas expelled in a previous superwind falls back on its host galaxy. There are indications that infall of metal-poor gas, perhaps in the form of gassy dwarf galaxies, may have had major impact on the chemical evolution of the disc of our Galaxy (Edvardsson et al. 1993). However one should recall that the existence of infalling gas on the Galactic disc is better established, as inferred from high velocity clouds (Mirabel 1989). Infall of unpolluted gas could act as to lower the ISM abundance. Generally, models of Galactic chemical evolution including infall assume pristine gas, but results with pre-enriched matter do not differ as long as the metallicity does not exceed 0.1 Z (Tosi 1988).