Observational and theoretical lines of evidence suggest that accreting black holes in active galaxies release large amounts of kinetic energy to their environments. Observationally, a minority of active galaxies release the bulk of their power in the form of radio jets (Rees et al. 1982; Begelman et al. 1984). But more numerous radio quiet AGN may also generate much of their power in the form of fast winds, as indicated by recent spectral analyses of broad absorption line (BAL) QSOs in the optical/UV (Arav et al. 2001; de Kool et al. 2001) and X-ray (Reeves et al. 2003) bands. These studies suggest that the kinetic energy in BAL outflows - which are thought to occur in most radio quiet QSOs - is larger than previously thought and can approach the radiation output.
Theoretical arguments also point to the importance of the kinetic energy channel. Numerical simulations suggest that accretion disks, which transfer angular momentum and dissipate binding energy via magnetorotational instability, may inevitably produce magnetically active coronae (Miller & Stone 2000). These are likely to generate outflows that are further boosted by centrifugal force (Blandford & Payne 1982). Unless radiation removes at least 2/3 of the liberated binding energy, very general theoretical arguments indicate that rotating accretion flows must lose mass. The physical reason is that viscous stresses transport energy outward, in addition to angular momentum. If radiation does not remove most of this energy, then a substantial portion of the gas in the flow will gain enough energy to become unbound (Narayan & Yi 1995; Blandford & Begelman 1999, 2004). Blandford (this volume) discusses this effect and its consequences in more detail. While it may sometimes be possible to tune the system so that the gas circulates without escaping, any excess dissipation (i.e., increase of entropy) near a free surface of the flow will lead to outflow. There are several possible sources of such dissipation, including magnetic reconnection, shocks, radiative transport, and the magnetocentrifugal couple mentioned above. If radiative losses are very inefficient, outflows can remove all but a small fraction of the matter supplied at large radii.
While it is probably safe to say that some substantial release of kinetic energy always accompanies accretion, we are not yet able to predict its magnitude, nor how it compares with radiative losses. During the growth of a supermassive black hole to mass MBH, the integrated kinetic energy output could be comparable to the total release of binding energy, ~ 0.1 MBH c2. It is unlikely to be much smaller than a tenth that amount. Even with a kinetic energy output of ~ 0.01 MBH c2, the effects on the environment can be dramatic. At a kinetic energy conversion efficiency of KE c2 per unit of accreted mass, an accreting black hole liberates 1019(KE/0.01) ergs per gram. In a galactic bulge with a velocity dispersion of 200200 km s-1, the accretion of one gram liberates enough energy to accelerate 2 × 104(KE / 0.01) 2002 grams to escape speed -- provided that most of the energy goes into acceleration. Since the typical ratio of black-hole mass to galactic bulge mass is > 10-3 (Häring & Rix 2004), feedback from a supermassive black hole growing toward its final mass could easily exceed the binding energy of its host galaxy's bulge. Such feedback has been invoked in various models to explain the MBH - Mbulge and MBH - correlations (e.g., Silk & Rees 1998; Blandford 1999; Fabian 1999).
The effects of injecting a certain amount of kinetic energy into the environment of a supermassive black hole depend not only on the amount of energy injected, but also on the temporal and spatial characteristics of the injection process and the structure of the ambient medium. Explosive injection of a large amount of energy in a short time - which might lead to intense, centrally concentrated heating at a shock - will have very different consequences from gradual, intermittent, or spatially distributed injection, which would allow the surrounding medium to adjust. Whether mechanical heating is partially offset by radiative cooling is an important factor, as is the presence of absence of small-scale density inhomogeneities. The speed of a shock or sound wave propagating through a medium with a `cloudy' thermal phase structure will be highest in the phase with the lowest density (the intercloud medium). Dense regions will be overrun and left behind by the front, as first pointed out by McKee & Ostriker (1977) in connection with supernova blast waves propagating into the interstellar medium. Consequently, most of the energy goes into the gas which has the lowest density (and is the hottest) to begin with. The global geometric structure of the ambient gas is important as well. Since a wind or hot bubble emanating from an AGN will tend to follow the path of least resistance, a disk-like structure can lead to a `blowout' of hot gas along the axis, leaving the disk intact. Anisotropic injection of the kinetic energy, e.g., in a pair of jets or equatorial, can likewise affect certain regions while sparing others.
Given these complications, it is especially desirable to find a `laboratory' where one can study the details of mechanical energy injections by active galaxies in a specific context. Clusters of galaxies can serve this role well. In the remainder of this paper we will discuss how recent X-ray observations of the intracluster medium (ICM) suggest a particular mode of heating by AGNs, which appears to be susceptible to theoretical modeling. While mechanical heating is likely to be the only important form of AGN feedback into the hot, highly ionized atmospheres of galaxy clusters, we stress that other forms of energy injection, particularly radiative heating, must occur as well. These effects are can be particularly important on smaller scales (i.e., the interstellar medium of the host galaxy) and in less highly ionized environments; they are discussed by Ostriker (this volume).