Compactness, high luminosity, time variability, spectral energy distribution and spectral features of active galactic nuclei (AGN) are best explained as consequences of disk accretion onto a supermassive black hole (SMBH). However, we do not have any direct observational evidence that gas indeed accretes onto a SMBH. Instead we quite commonly observe the opposite, i.e., gas outflows in AGN.
Broad absorption lines (BALs) in ultraviolet (UV) spectra of quasars are the most spectacular manifestation for such outflows. They are almost always blueshifted relative to the emission-line rest frame, indicating the presence of winds from the active nucleus, with velocities as large as 0.2 c (e.g., Turnshek 1998; for a few examples of BAL quasars with both blueshifted and redshifted absorption see Hall et. at. 2002). BALs are observed not only in the UV but also in other wavelengths. For example, Chartas, Brandt & Gallagher (2003) discovered a very broad absorption line in the X-ray spectrum of PG 1115+80. There are also a few examples of BALs in optical spectra of quasars (Hutchings et al. 2002; Aoki et al. 2006; Hall 2006). Other evidence for AGN winds include narrow absorption lines (NALs). UV spectra of some quasars show NALs which are blueshifted by as much as ~ 50000 km s-1 (Hamann et al. 1997). NALs are found much more commonly in the UV spectra of Seyfert galaxies than in spectra of quasars, but in Seyfert galaxies the lines are blueshifted only by several 100 km s-1 (Crenshaw et al. 1999). As BALs, NALs are observed not only in the UV but also in the X-rays. For example, Kaastra et al. (2000) observed NALs due to highly ionized species in a high-resolution X-ray observation of the Seyfert galaxy NGC 5548 obtained by Chandra. The prominent broad emission lines (BEL) in the UV from H I, O VI, N V, C IV, and Si IV are the defining feature of quasars (Blandford et al. 1990; Osterbrock 1989), and they may also be associated with a high velocity wind (Murray et al. 1995, hereafter MCGV).
Many observations show that AGN winds are very complex flows, and neither spherical nor axial symmetry is applicable. For example, observations taken with HST show bright emission-line knots in the narrow-line regions (NLRs) of some Seyfert galaxies which demonstrate a lack of any symmetry (Ruiz et al. 2005). Detailed analysis of these knots indicates that they may be related to the NLR clouds outflowing from the nucleus, like the UV absorbers (Crenshaw & Kraemer 2005). Complex flow geometry is likely responsible for several distinct components of absorption (e.g., NGC 3783, see Gabel et al. 2003; Mrk 279, see Scott et al. 2004) and for transverse motion of absorbing gas seen in one of the flow components of NGC 3783 (Crenshaw et al. 2003). Additionally, Arav et al. (2005, and references therein) argued that to explain the absorption in the outflow observed in Mrk 279, the absorber should be inhomogeneous.
Generally, we would like to know the source of the wind and how it is powered. Additionally, we would like to know the wind geometry, energetics, content, and ionization state and what controls them. Results obtained from data interpretation are very important but they also show that without a physical model, data interpretation provides limited insight into the wind origin and properties. To make progress in understanding AGN winds, we need a physical multidimensional model which will be capable of capturing the complex wind geometry and dynamics and ultimately fit observations.
As mentioned above, many observational aspects of AGN can be explained as consequences of disk accretion onto a SMBH. Therefore, I will focus here on models which assume that winds are also consequences of disk accretion. Specifically, I will review models where an accretion disk is the source of the wind and its energy. A much broader review and discussion of other possibilities can be found for example in Krolik (1999) and Crenshaw, Kraemer, & George (2002).