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

5. LOCAL CENSUS OF LARGE-SCALE OUTFLOWS IN ACTIVE GALAXIES

The same methods detect outflows in nearby active galaxies and starburst-driven winds, so the same selection effects and observational limitations discussed in Section 4.1 apply. However, contrary to starburst-driven winds, winds associated with AGN need not be perpendicular to the galactic disk. Outflows directed close to the galactic disk are likely to be made very luminous by the high ambient densities. They are more easily observed in near face-on galaxies (e.g., M51: Cecil 1988; NGC 1068: Cecil, Bland, & Tully 1990). In a few systems, the orientation of an inclined AGN disk can be determined independently from maser spots.

5.1. Detection Rate

As we discuss below, there is ample evidence for outflows in AGN from sub-kpc to galactic scale. But one must be aware of processes that may bias the interpretation of bipolar emission. Several active galaxies display kiloparsec-scale "ionization cones" that align with the radio axes of AGN rather than with the principle axes of the host galaxy (e.g., Wilson & Tsvetanov 1994; Kinney et al. 2000). They arise when ISM is illuminated by hard radiation from the AGN. Gas within the cones is more highly ionized than outside. Because the gas kinematics are unaffected by ionization, one should search for kinematic signatures to evaluate the frequency of occurrence of AGN outflows. Often, both outflows and ionization cones are present, e.g. NGC 1068 (Pogge 1988; Cecil et al. 1990), NGC 1365 (Veilleux et al. 2003), and NGC 4388 (Veilleux et al. 1999), emphasizing the need for high-quality kinematic data.

Circumnuclear starbursts in active galaxies are another complication. Half of all nearby, optically-selected Seyfert 2 galaxies also host a nuclear starburst (see, e.g., Cid Fernandes et al. 2001 and Veilleux 2001 and references therein); the fraction is larger in IR-selected systems. The sustaining conditions for nuclear activity (e.g., deep central potential with a reservoir of gas) also favor starburst activity and perhaps trigger starburst-driven GWs. This symbiotic relation between starbursts and AGN therefore complicates interpretation of GWs in AGN/starburst composites. For example, bipolar ionization cones can also arise in pure starburst galaxies (M82: Fig. 15 of Shopbell & Bland-Hawthorn 1998).

With these caveats, we revisit the detection of winds in active galaxies. Zensus (1997) reviewed the evidence for outflows in powerful radio-loud systems. Here we focus on Seyfert galaxies. Linear radio features suggestive of jet-like ejections on sub-kpc scales have long been known in 20 - 35% of Seyferts (e.g., Ulvestad & Wilson 1989; Morganti et al. 1999 and references therein); this is a lower limit because of limitations in spatial resolution (~ 1"), sensitivity, and de-projection. New high-resolution images with the Very Large Baseline Array (VLBA) have indeed revealed jet-like outflows in previously unresolved sources and in low-luminosity AGN (Ho & Ulvestad 2001). As we describe below, optical studies of several Seyferts with linear radio structure show signs of jet-ISM interaction on a scale of tens of pc. Measurements of proper motion in a few bright sources confirm outwardly moving radio knots (e.g., Middelberg et al. 2004 and references therein). A jet interpretation is sometimes favored even when the radio emission is unresolved at the VLBI scale (e.g., Anderson, Ulvestad, & Ho 2004). Low-power, sub-kpc jets may exist in most Seyfert galaxies. Any thermal wind component cannot be established from the radio observations, yet there are speculations that it can be large (Bicknell et al. 1998), (Section 5.3).

Many active galaxies also show signs of large-scale, galactic outflows. In an optical study of a distance-limited sample of 22 edge-on Seyfert galaxies, Colbert et al. (1996a) found that > 25% have kinematic signatures of outflows or extraplanar line emission out to 1 kpc. The existence of extraplanar radio emission in 60% (6/10) supports this claim (Colbert et al. 1996b). Morphology and orientation suggest that this emission comes mainly from AGN-driven outflows (see Section 5.2). This detection rate is a lower limit because of sensitivity and selection effects. For instance, Baum et al. (1993) detected kiloparsec-scale extranuclear radio emission in ~ 90% (12/13) of their objects, a sample of slightly more powerful, face-on Seyferts. The difference in detection rate may be due to small number statistics, differences in sample selection, or may indicate that some of this emission is associated with starburst-driven winds. Contamination by starburst-driven winds may also explain the large fraction (11/12 90%) of starburst/Seyfert 2 composites with extended X-ray emission (Levenson, Weaver, & Heckman 2001a, 2001b). There is now irrefutable evidence that powerful AGN-driven outflows sometimes coexist with starburst-driven winds (e.g., NGC 3079: Cecil et al. 2001).

5.2. Morphology

Contrary to what we know of starburst-driven winds, AGN outflows are oriented randomly relative to the major axis of the host galaxy (e.g., Ulvestad & Wilson 1984; Kinney et al. 2000). Several well-known active galaxies, including NGC 1068 (Cecil et al. 1990), harbor a wide-angle outflow whose axis is not perpendicular to the galaxy disks and therefore interacts strongly with the disk ISM. Outflows near the disk plane are often well collimated (e.g., NGC 4258: Cecil et al 2000; ESO 428-G14Falcke, Wilson, & Simpson 1998), until they collide with dense disk clouds or until they rise into the halo (e.g., NGC 4258: Wilson, Yang, & Cecil 2001). When the jet is drilling through the disk, detailed correspondence between radio and optical emission-line features indicates strong jet/ISM interactions over tens of pc (e.g., Falcke et al. 1998; Schmitt et al. 2003a, 2003b and references therein). The radio jets compress and shock ambient gas, enhancing line emission that may dominate the morphology of the NLR. The well-known correlation between radio and NLR luminosities supports a dynamical connection between the two (e.g., de Bruyn & Wilson 1978; Wilson & Willis 1980). A few jet deflections by ISM clouds are also seen (e.g., NGC 1068: Gallimore, Baum, & O'Dea 1996b; NGC 4258: Cecil et al. 2000; NGC 4151: Mundell et al. 2003). Jet/ISM interaction has also been mapped in X rays thanks to CXO (e.g., Young, Wilson, & Shopbell 2001; Wilson et al. 2001; Yang, Wilson, & Ferruit 2001).

In many edge-on systems, the radio structure has a linear or elongated morphology on subkiloparsec scale, but beyond the disk becomes more diffuse and wide-angled. The change may arise from the vertical pressure gradient in the surrounding ISM or from momentum loss within the sub-kpc NLR. The morphologies of the warm and hot ionized extraplanar gas are often correlated. They are distributed in a broad cone near the base of the outflow (e.g., NGC 2992: Colbert et al. 1998; Allen et al. 1999; Veilleux, Shopbell, & Miller 2001; Circinus: Veilleux & Bland-Hawthorn 1997; Smith & Wilson 2001) but become filamentary above the disk (e.g., NGC 4388: Veilleux et al. 1999, Fig. 5c; NGC 5506: Wilson, Baldwin, & Ulvestad 1985). Spectacular bow shocks and finger-like structures are seen optically in the Circinus galaxy, perhaps due to the abnormally high gas content of this object (Fig. 5d; Veilleux & Bland-Hawthorn 1997). The correspondence between the extraplanar radio plasma and optical line-emitting material is often not as tight as that seen in the disk. Outflows with optical conical geometries may have very different radio morphologies: e.g., edge-brightened radio bubbles in NGC 2992 (Wehrle & Morris 1988) and Circinus (Elmouttie et al. 1998) but lumpy and filamentary radio structures in NGC 4388 and NGC 5506 (Colbert et al. 1996b). The extraplanar emission at radio wavelengths (where the foreground disk barely influences the emission) is sometimes lop-sided (e.g., NGC 4388) from asymmetric energy injection at the source or from asymmetric ISM on small scales.

5.3. Kinematics

Spatially resolved outflows are found on all scales in AGN. Relativistic outflows on parsec scales in powerful radio-loud sources are well established (Zensus 1997; Worrall & Birkinshaw 2004 and references therein). Recently, proper motions of radio components have been measured with VLBI in a few Seyfert galaxies (e.g., Middelberg et al. 2004 and references therein). These studies show that the outward motion of the radio components in these objects is non-relativistic ( 0.25c) on pc scales. There is now unambiguous kinematic evidence that these radio jets transfer momentum and energy to the ambient gas and drive some of the large-scale outflows seen in radio-quiet and radio-loud objects.

Gas in the NLR (~ 10 pc - 1 kpc) and extended NLR (ENLR; 1 kpc) is an excellent tracer of this jet/ISM interaction. The good match between nuclear emission-line widths and bulge gravitational velocities suggests that the gravitational component dominates in most Seyferts (Whittle 1985; Wilson & Heckman 1985; Veilleux 1991; Whittle 1992b; Nelson & Whittle 1996). But Seyferts with linear radio structures have long been known to have emission lines with complex profiles (e.g., Veilleux 1991) and supervirial velocity widths (e.g., Whittle 1992a) that implicate an additional source of KE in the NLR. Detailed long-slit spectra from the ground and from HST have presented evidence for a dynamical connection between the NLR and radio components in many of these galaxies (e.g., Whittle & Wilson 2004 and references therein). The complete spatio-kinematic coverage afforded by Fabry-Perot and integral-field spectrometers has constrained efficiently the intrinsic, 3D velocity fields of the outflowing ionized gas (e.g., Ferruit et al. 2004; Veilleux et al. 2002 a and references therein). The signatures of jet-driven kinematics seen in some Seyfert galaxies are also detected in several powerful radio sources, particularly in compact steep-spectrum radio galaxies and quasars (e.g., Baum, Heckman, & van Breugel 1992; Gelderman & Whittle 1994; McCarthy, Baum, & Spinrad 1996; Best, Röttgering, & Longair 2000; Solórzano-Iñarrea, Tadhunter, & Axon 2001; O'Dea et al. 2002).

This large data set indicates that expanding radio lobes (Pedlar, Dyson, & Unger 1985) or bow shocks/cocoons driven by radio jets (Taylor, Dyson, & Axon 1992; Ferruit et al. 1997; Steffen et al. 1997a, 1997b) accelerate some of the line-emitting gas to ~ 100 - 1000 km s^-1. The fate of the gas clouds - whether undisturbed, destroyed, or accelerated - depends on factors such as the cloud mass, jet energy flux, and interaction geometry. Dense, molecular clouds in Seyfert galaxies can deflect radio jets by a significant angle without experiencing significant damage (e.g., NGC 1068: Gallimore et al. 1996a, 1996b; NGC 4258: Cecil et al. 2000). Jet-cloud interactions can be used to deduce key properties of the jet. In their analysis of the jet-molecular cloud interaction in the NLR of NGC 1068, Bicknell et al. (1998) argued that this jet (and perhaps those in other Seyferts) is heavily loaded with thermal gas and has low bulk velocities (~ 0.06c), contrary to jets in powerful radio galaxies. Not surprisingly, the jets in Seyfert galaxies often deposit an important fraction of their KE well within ~ 1 kpc of the nucleus. The radial velocities of the emission-line knots in the NLRs of Seyfert galaxies show a slight tendency to increase out to ~ 100 pc from the nucleus then decrease beyond, whereas the line widths of the knots decrease monotonically with increasing distance from the nucleus (e.g., Crenshaw & Kraemer 2000; Ruiz et al. 2005). Deceleration beyond ~ 100 pc is likely due to drag from ambient gas.

Ram pressure from radio jets/lobes may not always dominate the acceleration of line-emitting gas in Seyfert galaxies. High-resolution studies with HST sometimes fail to find a one-to-one correspondence between the NLR cloud kinematics and the positions of the radio knots. Ram pressure from a diffuse, highly ionized wind or radiation pressure by the AGN radiation field has been suggested as the possible culprit in these cases (Kaiser et al. 2000; Ruiz et al. 2001). The high-velocity (~ 3000 km s^-1) line-emitting knots detected in NGC 1068 (Cecil et al. 2002b; Groves et al. 2004) may be explained by radiation pressure acting on dust grains in the clouds (Dopita et al. 2002). These knots may correspond to the well-known absorbers seen projected on the UV continua of some AGN (Crenshaw et al. 2003). Bright nuclear emission-line knots detected in several other nearby Seyfert galaxies may also contribute to the population of intrinsic UV absorbers (Crenshaw & Kraemer 2005; Ruiz et al. 2005).

Wind ram pressure or radiation pressure may also be responsible for blueshifted (~ 100 - 1000 km s^-1) neutral material detected in several AGN-dominated ULIRGs (Rupke, Veilleux, & Sanders 2005c), because few of them show jet-like radio structures. There is some evidence for higher average and maximum velocities in Seyfert-2 ULIRGs than in starburst-dominated ULIRGs, although the evidence for a strong influence of the AGN on these outflows is inconclusive. The situation is quite different among Seyfert-1 ULIRGs, where the outflows are driven mostly or solely by the AGN (Rupke, Veilleux, & Sanders 2005c; Fig. 6a). Similarly, nuclear activity is almost certainly responsible for the broad, blueshifted H I absorption wings detected in a growing number of compact radio galaxies. But here, jet-driven acceleration is implied (e.g., Morganti et al. 2003).

In Seyferts where the energy injection rate of the jet or wind suffices to eject radio/thermal plasma from the host galaxy, warm line-emitting gas is entrained into wide-angle outflows. The intrinsic, 3D velocity field of the line-emitting gas over 1 kpc indicates roughly conical radial outflow with opening angles 2 60°- 135° and hollow (NGC 1365: Hjelm & Lindblad 1996; NGC 2992: Veilleux et al. 2001), partially filled (NGC 5506: Wilson et al. 1985), or filamentary geometry (Circinus: Veilleux & Bland-Hawthorn 1997; NGC 4388: Veilleux et al. 1999). Deprojected velocities are ~ 100 - 500 km s^-1 in Seyferts, sometimes larger in radio galaxies. The geometry of the kinematic structures on large scales depends not only on the energy source but also on the galactic and intergalactic environment (e.g., the abnormally large gas content of Circinus may explain the peculiar filamentary morphology of its outflow). Rotation is sometimes detected in the velocity field of the outflowing gas (e.g., Circinus: Veilleux & Bland-Hawthorn 1997; NGC 2992: Veilleux et al. 2001), confirming that most of the line-emitting material comes from the disk ISM.

5.4. Mass Outflow Rates & Energetics

Kinematic deprojection is an essential prerequisite to meaningful dynamical analysis of AGN-driven winds. Data with complete spatio-kinematic coverage should be used to derive their masses and KE. We therefore discuss nearby (z 0.1) objects for which spatially resolved kinematic data are available. The broad range of scales and velocities discussed in Sections 5.2 and 5.3 implies dynamical times for the entrained line-emitting material of ~ 10⁴ - 10⁶ years in the NLR (including the 3000 km s^-1 knots in NGC 1068) and ~ 10⁶ - 10⁷ years in the ENLR.

Deriving ionized masses for AGN outflows follows the same steps as for starburst-driven winds, and therefore again favors high-density material (Section 4.5). Given uncertain volume filling factors, the warm ionized gas masses derived from optical data on Seyferts, ~ 10⁵ - 10⁷ M, are probably accurate to no better than about ± 0.5 dex. The corresponding mass outflow rates, M / t_dyn 0.1 - 10 M yr^-1, generally exceed tenfold the mass accretion rates necessary to fuel the AGN, indicating strong mass loading of the outflow by the galaxy ISM.

The KE of the warm ionized component in Seyfert galaxies is ~ 10⁵³ - 10⁵⁶ erg, including both bulk and "turbulent" (spatially unresolved) motions. This range is similar to that in starburst-driven winds. Dynamical timescales yield KE outflow rates of E_kin / t_dyn 10⁴⁰ - 10⁴³ erg s^-1. The power radiated by Seyferts in line emission is 10 times larger and ~ 10² - 10⁴ times larger than in the radio (e.g., Fig. 1 of Wilson, Ward, & Haniff 1988). Simple jet-driven models (Blandford & Königl 1979a, 1979b; Wilson 1981) reproduce these derived values using reasonable efficiency factors to relate jet and radio powers.

If the efficiency of energy/momentum transfer to the ambient material is nearly constant, both the entrained mass and KE scale with AGN luminosity (e.g., Baum & McCarthy 2000). Powerful quasars and radio galaxies with L_bol ~ 10⁴⁵ - 10⁴⁷ erg s^-1 (not discussed so far) seem to have KE outflow rates that exceed those of Seyferts by several orders of magnitude. The KE of some outflows is comparable to the gravitational binding energy of gas in the host (10⁵⁸ - 10⁶⁰ erg). Comparable energy is stored in the large radio lobes of radio-loud objects: / ( - 1) P V ~ 10⁵⁸ - 10⁶¹ erg, where P and V are the pressure and volume of the radio lobes, respectively, and = 4/3 is the mean adiabatic index of the relativistic fluid. Much of the energy dissipated by the AGN accretion disk appears to find its way into jet mechanical luminosity: powerful radio sources are Eddington-tuned engines with jet powers Q ~ 0.1 L_Edd (Willott et al. 1999; see also Falcke, Malkan, & Biermann 1995).

Outflowing neutral gas detected in absorption in H I and Na I D in many AGN may add significant energy. Unfortunately, the mass of entrained neutral gas in these objects is poorly constrained because its location is unknown. The mass-conserving wind model of Rupke et al. (2005b; R_outer = 5 kpc) applied to the Na I D absorption data on Seyfert 2 ULIRGs yields neutral mass outflow rates similar to those in starburst-driven winds, i.e. ~ 10 - 1000 M yr^-1. With this assumption, neutral-gas KEs and rates of ~ 10⁵⁶ - 10⁶⁰ erg and E_kin / t_dyn 10⁴¹ - 10⁴⁵ erg s^-1 are derived for Seyfert-2 ULIRGs, suggesting that this material may be a very important dynamical component of AGN-driven outflows.

5.5. Energy Injection Zone

An upper limit to the size of the energy injection zone of AGN-driven winds can be derived from the geometry of the optical outflow. Injection zones are often small enough to exclude a starburst origin for the wind. A good example is the Circinus galaxy (Fig. 5d), where the base of the outflow cone extends over 20 pc, 10% of the size of the starburst (e.g., Smith & Wilson 2001 and references therein). The same technique has been used to measure the size of the primary energy source for the outflows in NGC 1068 ( 50 pc; e.g., Cecil et al. 1990; Arribas, Mediavilla, & Garcia-Lorenzo 1996; Garcia-Lorenzo, Mediavilla, & Arribas 1999), NGC 2992 ( 100 pc; Veilleux et al. 2001; Garcia-Lorenzo, Arribas, & Mediavilla 2001), and in several other jet-induced outflows (e.g., Whittle & Wilson 2004).

More accurate assessment of the energy source in AGN-driven winds often relies on detailed VLA, MERLIN, and VLBA radio maps of the central regions of AGN, down to scales below tens of parsecs. Radio emission generally resolves into a few compact sources, occasionally accompanied by elongated, jet-like features. On VLA scales, approximately half of all Seyfert galaxies contain at least one component with a flat or inverted radio spectrum (i.e. spectral slope -0.20; e.g., Ulvestad & Ho 2001). Radio sizes are uncorrelated with Seyfert type, as expected from the unified scheme of AGN. On VLBI scales, 80% of sources are flat-spectrum (Middelberg et al. 2004; Anderson, Ulvestad, & Ho 2004), are often unresolved or only slightly resolved (~ 1 pc), and have brightness temperatures > 10⁷ K. Synchrotron self-absorption from the base of the jet is the usual explanation for compact and flat- or inverted-spectrum radio emission from the central engines of radio galaxies. The same scenario is likely in some Seyfert galaxies and low-luminosity AGN. Free-free absorption by the nuclear torus or NLR may also occur. In NGC 1068, the flat-spectrum source has modest brightness temperature (~ 10⁶ K) and extends over ~ 1 pc perpendicular to the jet axis; this emission is believed to be optically thin, free-free emission from the ionized inner edge of the torus (Gallimore, Baum, & O'Dea 1997). Regardless of the exact origin of the emission, the compact flat-spectrum component is thought to be the main energy source of the outflow, because of frequent alignment of the various radio components and from direct measurements of proper motions in a few objects (Middelberg et al. 2004 and references therein; see Section 5.3 above).

5.6. Source of Ionization

It is now generally agreed that the gas in Seyferts is photoionized by the nucleus, although whether spatial variations in line ratios result from a "classical" range of ionization parameters (e.g. Davidson & Netzer 1979), a range in the numbers of illuminated ionization- and matter-bounded clouds (Binette, Wilson, & Storchi-Bergmann 1996), or a dusty plasma at high ionization parameter (Dopita et al. 2002) are still discussed. The detection of X-ray cones coincident with the emission-line ionization cones in several Seyferts rules out photoionizing shocks as the primary source of ionization (e.g., Kinkhabwala et al. 2002; Iwasawa et al. 2003; Yang et al. 2001), although the relative importance of shocks almost certainly increases with jet power and the degree of interaction with the ambient medium (e.g., compact radio galaxies; Tadhunter 2002).

5.7. Dust

Dust is present in AGN outflows. Its reddening is evident in the narrow-line spectrum of radio-quiet and radio-loud galaxies (e.g., Osterbrock 1989 and references therein) and in the broad absorption line clouds of BAL QSOs (e.g., Crenshaw et al. 2003 and references therein). Dust has often been invoked to explain the extended blue wings in the narrow-line profiles (e.g, Veilleux 1991 and references therein), and in the IR continuum emission of several AGN (Barvainis 1987; Sanders et al. 1989; Pier & Krolik 1992. But more relevant to outflowing dust is that the correlation between color excess, E(B - V), and the strength of the Na I D line noted for starburst galaxies is even stronger for AGN (Veilleux et al. 1995b). Dust outflow rates of ~ 0.1 - 10 M yr^-1 are inferred from the neutral-gas mass outflow rates discussed in Section 5.4, assuming the Galactic gas-to-dust ratio. These are comparable to the dust outflow rates in starburst galaxies (Section 4.10).