Annu. Rev. Astron. Astrophys. 2005. 43: 769-826 Copyright © 2005 by Annual Reviews. All rights reserved

7. THE SIGNIFICANCE OF GALACTIC WINDS

There is growing evidence that GWs have inhibited early star formation and have ejected a significant fraction of the baryons once found in galaxies. The latter may explain why few baryons are in stars (_* / _b ~ 0.1; Fukugita, Hogan, & Peebles 1998) and why galaxies like the Milky Way contain fewer than expected from hydrodynamical simulations (Silk 2003). We review in this section the impact of winds on galaxies and on their environment.

7.1. Influence of Winds on Galactic Scales

7.1.1. GALAXY LUMINOSITY FUNCTION    GWs have modified substantially the shape of the galaxy luminosity function, flattening its faint-end slope compared to that of the halo mass function (e.g., Dekel & Silk 1986; Somerville & Primack 1999; Benson et al. 2003; Dekel & Woo 2003). The shallow potential of dwarf galaxies makes them vulnerable to photoevaporation (if v_c 10 - 15 km s^-1; Barkana & Loeb 1999, mechanical feedback (e.g., de Young & Heckman 1994; MacLow & Ferrara 1999; Ferrara & Tolstoy 2000), and ablation by GWs from nearby galaxies (Scannapieco, Thacker, & Davis 2001). Significant feedback also appears necessary to avoid the `cooling catastrophe' at high redshift that would otherwise overproduce massive luminous galaxies (e.g., Cole et al. 2000). Energies of a few × 10⁴⁹ ergs per solar mass of stars formed can explain the sharp cutoff at the bright end of the luminosity function (Benson et al. 2003). Starburst-driven winds are too feeble by a factor of several to fully account for the cutoff. Benson et al. (2003) therefore argue that feedback from BH accretion is the only way to expel winds hot enough to prevent subsequent gas recapture by group halos. The kinetic power supplied by jets in radio-loud AGN, Q ~ 0.1 L_Edd (see Section 5.4), may indeed suffice to account for the paucity of high-mass systems. Feedback from starburst- and AGN-driven winds may help set up the bi-modality observed in galaxy properties (see below; also Dekel & Birnboim 2005). AGN feedback may be particularly effective in clustered environment where the infalling gas is heated by a virial shock and thus more dilute.

7.1.2. CHEMICAL EVOLUTION    In the GW scenario, massive galaxies with deep gravitational potentials are expected to retain more of their SN ejecta than dwarf galaxies (e.g., Larson 1974; Wyse & Silk 1985; Dekel & Silk 1986; Vader 1986). Several authors have provided observational support for this picture, often using luminosity as a surrogate for mass (e.g., Bender, Burstein, & Faber 1993; Zaritsky et al. 1994; Jablonka, Martin, & Arimoto 1996; Trager et al. 1998; Kobulnicky & Zaritsky 1999; Salzer et al. 2005). An analysis of the Sloan Digital Sky Survey (SDSS) database by Tremonti et al. (2004) has shown that the gas-phase metallicity of local star-forming galaxies increases steeply with stellar mass from 10^8.5 to 10^10.5 M h₇₀^-2, but flattens above 10^10.5 M h₇₀^-2. Similar trends are seen when internal velocity or surface brightness is considered instead of stellar mass (Kauffmann et al. 2003). The stellar mass scale of this flattening coincides roughly with the dynamical mass scale of metal retention derived by Garnett (2002). In that paper, Garnett used a simple closed-box chemical model to infer that the effective yield increases with galaxy mass up to the stellar yield obtained at v_c 125 km s^-1. These results suggest that the chemical evolution of galaxies with v_c 125 km s^-1 is unaffected by GWs, whereas galaxies below this threshold tend to lose a large fraction of their SN ejecta. This is consistent with estimates based on X-ray temperatures in wind galaxies (Sections 4.4, 4.6, and 4.9).

7.1.3. DISK SIZE AND DARK MATTER CONCENTRATION    Hydrodynamical simulations reveal that dynamical friction from the inner dark halo acting on baryonic clumps overcompresses the galactic disk by a factor of five in radius compared what is observed (e.g., Steinmetz & Navarro 1999; Bullock et al. 2001). The suppression by stellar or AGN feedback of early gas cooling solves only part of this angular momentum problem (e.g., Sommer-Larsen, Götz, & Portinari 2003; Abadi et al. 2003). Entrainment and removal of material with low specific angular momentum by starburst- or AGN-driven GWs (e.g., Binney, Gerhard, & Silk 2001; Maller & Dekel 2002; Read & Gilmore 2005Kormendy & Kennicutt 2004). If many of the central baryons are thus ejected, CDM halo profiles become less cuspy (e.g., Navarro, Eke, & Frenk 1996) and more consistent with mass distributions observed in dwarfs and low-surface brightness galaxies (e.g., van den Bosch & Swaters 2001; de Blok, McGaugh, & Rubin 2001; Gentile et al. 2004; see Ricotti & Wilkinson 2004 for an alternative viewpoint).

It is interesting to speculate which of the structural properties observed in galaxies are due to the action of winds in the early universe. For example, it is well known that the average baryon density observed in a galaxy decreases with declining galaxy luminosity or mass (e.g. Kormendy 1985). As we have mentioned, it is probable that powerful central winds in the early galaxy can reshape the baryon density distribution in galaxies (e.g. van den Bosch 2001). This idea has been explored in many papers which incorporate centrally driven feedback processes in order to patch up the shortcomings of CDM simulations. Similarly, there is a decreasing baryon/dark matter fraction observed in galaxies with declining galaxy luminosity or mass (e.g. Persic, Salucci & Stel 1996; Mateo 1998). In a seminal paper, Dekel & Silk (1986) anticipated this now well-established trend after considering the impact of SN-driven outflows in protodwarfs (see also Nulsen & Fabian 1997).

But there are reasons for believing that most of the feedback prescriptions imposed in CDM hydro simulations to date have little or no relevance to real galaxies (see Springel & Hernquist 2003 for a recent summary); many of the prescriptions manifestly do not conserve energy or entropy in the flow. When one considers the evolution of GWs in different environments, the trends (i.e. those arising from the action of winds) with total galaxy mass may be far less marked. For example, gas accretion onto a galaxy can staunch the developing outflow for any galaxy mass (e.g. Fujita et al. 2004; Springel & Hernquist 2003). Therefore, it is not obvious which of the structural properties can be ascribed to the action of winds at the present time.

7.1.4. POROSITY OF HOST ISM    The relative contribution of AGN and starbursts to the inferred ionizing background depends critically on f_esc, the fraction of ionizing photons that escape from each object. The column densities of disks imply 10⁴ - 10⁷ at the Lyman edge, so leakage of ionizing radiation must be set by the topology of the ISM. GWs should play a key role in clearing a path for the escaping radiation (e.g., Dove, Shull, & Ferrara 2000), but this has not yet been confirmed observationally from constraints on f_esc. H measurements of high-velocity clouds above the disk of our Galaxy indicate that the escape fraction normal to the disk is 6% (f_esc 1% - 2% averaged over 4 sr; Bland-Hawthorn & Maloney 1999, 2002). Estimates of the escape fraction in local, UV-bright starburst galaxies yield f_esc 6% (Heckman et al. 2001b and references therein), and similar values are inferred for bright blue galaxies at z = 1.1 - 1.4 (Malkan, Webb, & Konopacky 2003). Star-forming galaxies thus contribute little (< 15%) to the ionizing background at z 1.5. The situation may be different at z 3, where the comoving number density of QSOs declines rapidly. Steidel, Pettini, & Adelberger (2001) infer f_esc 50 - 100% for < z > = 3.4 LBGs, but these results have been questioned by Giallongo et al. (2002), Fernández-Soto, Lanzetta, & Chen (2003), and Inoue et al. (2005). The dark cores of the saturated interstellar absorption lines in LBGs (Shapley et al. 2003) also appear inconsistent with large f_esc, unless we see little of the escaping ionizing radiation. The large value of f_esc inferred by Steidel et al. (2001), if confirmed, may be due to powerful GWs in these objects (Section 6.1).

7.1.5. SPHEROID - BLACK HOLE CONNECTION    The masses of the central BHs in early-type galaxies and bulges correlate well with the velocity dispersions of the spheroidal component: M_BH = 1.3 × 10⁸ ₂₀₀⁴ M (Ferrarese & Merrit 2000; Gebhardt et al. 2000; Tremaine et al. 2002). This correlation is remarkably similar to the Faber-Jackson relation (Bernardi et al. 2003), and suggests a causal connection between galaxy formation and BH growth by means of a GW that regulates BH fueling (e.g., Silk & Rees 1998; Haehnelt et al. 1998; Fabian 1999; King 2003; Murray, Quataert, & Thompson 2005; Begelman & Nath 2005). The wind may be produced by the starburst that accompanied the formation of the spheroid or by the BH itself. An Eddington-like luminosity is derived for the starburst or the BH, above which the growth of both spheroid and BH is stopped by the wind. In the case of a dominant BH wind, the Salpeter timescale, i.e. the timescale for M_BH to double, must be similar to the star formation timescale so that sufficient stars are formed before the BH wind blows away the ambient gas and stops star formation (Murray et al. 2005). The massive winds detected in nearby ULIRGs (Section 4.5) may be local examples of what might have occurred as spheroids formed (e.g., high-z LBGs and submm galaxies; Section 6.1).

7.2. Influence of Winds on Intergalactic Scales

7.2.1. INTRACLUSTER MEDIUM    Galaxy clusters are excellent laboratories to study the impact of GWs on the environment because the hot, metal-enriched material ejected from SNe is retained by the cluster gravitational potential. Most metals in clusters are in the ~ 0.3 Z ICM, not in galaxies. Several lines of evidence suggest that GWs, not ram-pressure stripping, has dominated the transfer of metals from galaxies to ICM (see review by Renzini 2004). One is that ejection of hot gas from proto-galaxy GWs can create the `entropy floor' (Kaiser 1991; Evrard & Henry 1991) necessary to explain the steep X-ray luminosity - temperature relation for nearby groups and clusters (e.g., Arnaud & Evrard 1999; Helsdon & Ponman 2000), the lack of cluster evolution out to z ~ 1 (e.g., Mushotzky & Scharf 1997), and the shallow density profiles of cooler groups (e.g., Horner et al. 1999). Heating of ~ 1 keV per gas particle would reproduce these results. Type II and Type Ia SNe (e.g., Lloyd-Davies, Ponman, & Cannon 2000), AGN (e.g., Cavaliere, Lapi, & Menci 2002), and Type II SNe from very massive, metal-poor progenitors (i.e. Population III stars; e.g., Loewenstein 2001) may all contribute to the heating.

Analyses of ICM abundances provide some constraints on the relative importance of these energy sources. Early reports of large -element abundances in bright clusters by Mushotzky et al. (1996) first showed that Type II SNe enrich (and thus heat) some of the ICM. ASCA and XMM results now suggest that iron-rich Type Ia ejecta dominate in the centers of rich clusters, whereas the -rich products of Type II SNe are distributed more evenly (e.g., Finoguenov et al. 2002, Tamura et al. 2004 and references therein). The iron mass scales with the optical light from the early-type galaxies and the cluster X-ray luminosity (e.g., Arnaud et al. 1992; de Grandi et al. 2004), suggesting iron enrichment by Type Ia SNe from these galaxies. A contribution from Population III stars may be needed to explain the inhomogeneity of -elements in the ICM (Baumgartner et al. 2005). In-situ enrichment by intracluster stars may also be significant (Zaritsky, Gonzalez, & Zabludoff 2004).

AGN winds help enrich the ICM with metals, and the ubiquity of large "cavities" in the X-ray surface brightness of clusters with radio galaxies (e.g., Böhringer et al. 1993; Fabian et al. 2000; McNamara et al. 2000, 2001; Heinz et al. 2002; Mathews & Brighenti 2003) confirms that they modify the thermodynamics of the ICM. The hot, relativistic gas injected into the ICM by the AGN reduces, and perhaps even quenches, the mass accretion of cooling flows. The exact mechanism by which energy in the radio bubbles turns into heat is still debated, but the absence of strong shocks along cavity walls, and the discovery of low-amplitude, semi-periodic ripples in the Perseus cluster (Fabian et al. 2003) suggest that viscous dissipation of sound waves may heat much of the inner ICM (see also Ruszkowski, Brüggen, & Begelman 2004a, 2004b; Reynolds et al. 2005). Other possible heaters include thermal conduction and turbulent mixing (e.g., Narayan & Medvedev 2001; Ruzkowski & Begelman 2002; Kim & Narayan 2003a, 2003b).

7.2.2. INTERGALACTIC MEDIUM    The sphere of influence of GWs appears to extend to the low-density environment of the Ly forest [N(H I) 10¹⁷ cm^-2]. Here, metallicities of 0.1% - 1% solar have been measured, with a possible excess of -rich SN II products in the denser clouds (e.g., Rauch, Haehnelt, & Steinmetz 1997; Songaila 1997; Hellsten et al. 1997; Davé et al. 1998; Carswell, Shaye, & Kim 2002). The detection of metals in the IGM seems to favor momentum- over energy-driven winds (Section 2.3), or scenarios where the winds emerge along paths of least resistance without disturbing the filaments responsible for the Ly forest (e.g., Theuns et al. 2002). Remarkably, both the column density distribution of C IV absorbers and its integral (_C IV) are invariant over 2 z 5 (Songaila 2001; Pettini et al. 2003). One possible explanation is that most of the IGM metals are already in place by z ~ 5, perhaps from SN-driven outflows from low-mass subgalactic systems (e.g., Qian & Wasserburg 2005). Such systems may also be responsible for reionizing the IGM (Loeb & Barkana 2001 and references therein). However, this scenario does not completely explain why _C IV remains constant over this redshift range despite variations in the intensity and spectrum of the ionizing background (Section 7.1.4). Alternatively, the C IV systems are associated directly with GWs from LBGs at z 5, and the constancy of _C IV arises instead from the flatness of the SFR density over z 1.5 - 4 (Adelberger et al. 2003). A critical discriminator between these two scenarios is to measure the metallicity in truly intergalactic clouds with N(H I) 10¹⁴ cm^-2 (Cen, Nagamine, & Ostriker 2005). This is a portion of the Ly forest that has not yet been explored in detail (although see Ellison et al. 2000).