5.1. Narrow-line profiles
AGNs are unique among galaxies in the property that defines them, namely having `wide' line profiles with a larger velocity spread than in gaseous nebulae and stars. The only interpretation that seems physically possible is that the profiles result from the internal velocity fields in the objects. Since the nuclei are mostly unresolved, the information the profiles bear are integrated and not straightforward to interpret, but they are the best information we have. Much of this work has been summarized by Osterbrock and Mathews (1986). The highest spectral resolution data, including those of Whittle (1985a, b, c) with resolution 1.1 Å 65 km s-1, Vrtilek and Carleton (1985) ith 0.4 Å 25 km s-1, and Veilleux (1989) with 0.17 Å 10 km s-1 have been especially important. Correlations between line-profile parameters and extinction have been studied in detail by Dahari and De Robertis (1988b).
In most respects the narrow lines in Seyfert 1 and Seyfert 2 galaxies have similar properties, and they may be discussed together. [O III] 4959, 5007 are the strongest unblended narrow lines and therefore the most studied. In Seyfert 1 and 2 galaxies their full widths at half maximum (FWHMs) have a broad distribution, centered perhaps at 350 to 400 km s-1, and extending from about 200 to about 900 km s-1. NGC 1068, the very well studied Seyfert 2 galaxy, has unusually wide `narrow' lines, with FWHM of [O III] 1200 km s-1. The [O III] FWHMs are correlated (with considerable scatter) with [O III] luminosity and with the radio-frequency luminosity of the nucleus, indicating that the velocities in the NLR are driven by or associated with the energy input to this region. There are also weaker, but significant, correlations of the narrow-line profiles with the morphological type of the host galaxy, and possibly with the luminosity of its bulge. This indicates that the gravitational effects of the galaxy affect the velocity field in the NLR (Whittle 1989, Veilleux 1991a).
In many Seyfert 1s and 2s the narrow-line profiles can be observed to be asymmetric. Almost invariably this is in the sense that the profile has a longer tail or wing extending to the shorter wavelength side and a more abrupt fall-off toward longer wavelength. No purely rotational velocity field will produce such profiles. The simplest possible interpretation is that the asymmetry results from radial flow combined with extinction by dust. Various combinations are possible; the sense of the flow is not definite because the physical structure of the emitting region is not certain. If the flow is radially outward, dust distributed throughout the AGN will, on the average, cut down the emission lines from the far (receding) side more than from the near (approaching) side, producing the observed asymmetry. On the other hand, if the emission comes from small, optically thick (to ionizing radiation) clouds which contain dust, the emission will come primarily from the sides of the clouds which face the ionizing source. The radiation from clouds on the far side of the source will then come directly to the observer without as much extinction as radiation from clouds on the near side, which must escape through the cloud. Hence, in this picture, if the clouds are falling radially inward, the line profiles will have the observed sense of asymmetry.
The most physically attractive picture of the NLR is that the flow is outward, with a high-temperature, low-density, invisible plasma wind carrying the low-temperature, high-density, emitting clouds within it. Such models, among others, fit the observed profiles well (Vrtilek 1985). All these models have dust distributed in the same regions as the gas. However there are some observational indications that distributed dust is not the only factor responsible for the line asymmetry. Considerable scatter occurs in the relation between asymmetry arid extinction (measured by the H / H intensity radio). Cylindrical rather than spherical symmetry is indicated by many observed features of NLRs, and an occulting or optically thick central disk, cutting off radiation from the far side of the NLR, may be the cause of the asymmetry in at least some galaxies (Veilleux 1991b).
In a fair number of bright Seyfert 1 and 2 galaxies, profiles of many narrow emission lines of different ions have been measured. In each NLR, the profiles have the same general form (degree of asymmetry). In many objects there is a correlation between the FWHM and critical electron density for de-excitation, in the sense that the lines that can be emitted at higher densities have larger FWHMs. An example is Mrk 1066, in which the FWHM ranges from 300 km s-1 for [S II] with Nec 103.5 cm-3 to 400 km s-1 for [O III] with Nec 105.8 cm-3. In 70% of the AGNs for which [O III] 4363, with Nec 107.5 cm-3, profiles have been measured, they are broader than those for [O III] 4959, 5007, with Nec 103.8 cm-3. In many NLRs there is also a correlation between line width and ionization potential, in the sense that the ions of highest ionization potential have the largest FWHMs. Since the velocity decreases outward from the central ionizing source in practically every model, these results indicate that the ionization decreases outward in typical NLRs, and that the mean electron density also decreases outward, if fitted by a power law approximately as Ne r-n with 0 < n < 2. [O I] 6300, a low-ionization line with a relatively high critical density (Nec 2 x 106 cm-3), typically has relatively broad wings, indicating that the emitting clouds are optically thick to ionizing radiation and contain, in their `transition zone', some O0 atoms.
De Robertis and Shaw (1990) have attempted to resolve observationally the question of whether the velocity field is primarily radially outward or inward in the NLR, by studying the asymmetry as a function of ionization and critical electron density for collisional de-excitation. Their general result, obtained by comparing high-quality observational data with simplified, spherically symmetric kinematic models, is that the dust extinction is within the emitting clouds and the direction of flow is therefore inward. However, as many other possible types of models, specifically cylindrically rather than spherically symmetric, were not investigated, the conclusion is still open to question.
In addition the narrow line profiles often show some substructure in some cases strong enough to show `shoulders' or secondary peaks. This is especially clear with high resolution data (Veilleux 1989, 1991c). No doubt these effects result from non-homogeneous gas distribution. In some of the nearest AGNs, NGC 1068 (Baldwin et al 1987) and NGC 4151 (Schulz 1987) for example, such inhomogeneities can be directly resolved. Several more are seen in other AGNs in the spectra of Veilleux (1989, 1991c). The NLR is certainly a complicated structure. Radial motions, probably outflow, modified by extinction and/or obscuration, photoionization from a central source, and inhomogeneities all form parts of its description. We will return to a more nearly complete discussion of the velocity field in the NLR in section 5.3.
5.2. Extended emission-line region
Long-slit spectra provided the first convincing evidence for the ionization cones described in section 4.3. Unger et al (1987) took spectra of seven Seyfert galaxies, known to have jets or linear radio structure, with the slit aligned parallel and perpendicular to the jets. Those with the slit parallel to the linear structure showed the gas photoionized by the hard spectrum extending far from the nucleus, as much as 20 kpc in some cases, but with very little extent in the perpendicular direction. Unger et al called the ionized gas far from the nucleus the extended narrow-line region (ENLR). Though the interference-filter images map the entire cone, the spectra give radial-velocity information. These spectra show that the velocity field in the ENLR is chiefly the smooth rotation field of the galaxy in which they lie. The lines are much narrower than in the NLR, typically with FW0I 50 km s-1 rather than around 300 km s-1. Likewise Baldwin et al (1987) mapped the velocity field in NGC 1068 in great detail, with long-slit spectra taken at several different positions. Again, they show that outside a radius of about 10" ( 750 pc) the gas photoionized by the nucleus has the smooth galactic rotation curve characteristic of the disk. Inside this distance the field is more complicated but appears to correspond to rotation with a smoothly varying line of nodes. One possible interpretation is that in this region the midplane of the gas is warping between the plane of the galaxy and that of the NLR (perpendicular to the axis of the radio jet). In addition, in the outer region a few isolated clumps have velocities significantly different from the rotational velocity field. Their line ratios show they are all photoionized by the AGN. All of the clumps that are seen projected against dusty regions have redshifts with respect to the rotation curve, that is are falling toward the disk. Near the nucleus several clumps are seen with a range of velocities; these are the condensations associated with the jet. An even more detailed and complete analysis of the velocity field in the NLR region of NGC 1068 has been given by Cecil et al (1990) with similar general conclusions. Likewise in the Seyfert 1 galaxy NGC 4151 Schulz (1988) found similar patterns of the photoionization by the AGN along the extended axis of the jet.
5.3. Overall NLR velocity field
The observational data previously summarized clearly show that the velocity field in the inner part of the NLR is not rotational, but radial (if spherical) or in the and z direction (if cylindrically symmetric). Its spatial structure is clearly more nearly cylindrically symmetric and associated with the axis of the jet instead of the axis of the main body of the host galaxy. It is not clear whether this velocity field is directed outward or inward. Some observational evidence favours inward motion, but it is highly model dependent. Theoretical expectations favour outward motion. The motion of the radio plasma in the jet is clearly outward, and since the NLR is so closely associated with it, this again would seem to favour outward motion for it. The extension of the NLR is probably continuous, through the ENLR to the gas in the disk of the galaxy. Its velocity field is rotational. The spatial structure warps from perpendicular to the axis of the jet to perpendicular to the axis of the galaxy.
The ionizing radiation from the nucleus is anisotropic, perhaps collimated or `beamed', and these ionization cones in many Seyfert 2 galaxies include segments of the main body of the disk. The evidence for anisotropy of the AGN radiation is not very marked for Seyfert 1s. They may in fact have thin disks and `cones' with opening angles of nearly 180°.
Far from the nucleus and from the extended axis of the jet, isolated clouds or elements of gas seem to be falling toward the disk in NGC 1068, the one best studied Seyfert 2. Although it cannot be (or has not been) observed, probably the mainly rotational velocity field in the innermost parts of the ENLR contain a slight inward component, which probably also continues in the central plane of the NLR, bringing material in to the accretion disk. This overall pattern is schematically represented in figure 7, which is probably the best present working hypothesis for the velocity field in the NLR and ENLR.
Figure 7. Schematic representation of the velocity field in the inner (r 50 pc) and outer (r 50 pc) regions of the NLR. The line-emitting gas in zones 1, 2, 3 and 4 corresponds to the wind-accelerated, jet-associated, and inner and outer gravity-dominated components, respectively. Note that the rotation axis of the outer, gravity-dominated component (which merges into the interstellar matter in the galaxy) is not necessarily perpendicular to the axis of the double cone associated with the jet. Clouds in zone 5 probably contain substantial amounts of neutral gas. (Veilleux 1989)
5.4. BLR velocity field
The BLR is completely unresolved for all observed AGNs, as expected from the calculated sizes. It is expected to remain even so with the much better angular resolution of the Hubble Space Telescope (~ 0.1"). This prediction of course should, however, be checked observationally. Thus the only observed information on the velocity fields in the BLRs is in the form of the broad-line profiles. An atlas containing H, H, He I 5876 and He II in 19 Seyfert 1 galaxies has been published by Osterbrock and Shuder (1982). As previously stated, the H I lines have a wide range of FWHMs, from approximately 500 km s-1 in Mrk 359, NGC 4051 and Akn 564 up to approximately 7000 km s-1 in Mrk 279, Mrk 876 and Mrk 926. The FW0Is, as well as they can be defined, range from approximately 4 x 103 km s-1 in Mrk 359 to approximately 3 x 104 km s-1 in Mrk 876. Most Seyfert 1s thus have much wider lines than Seyfert 2s, though a small group of `narrow-line Seyfert 1s', of which Mrk 359 is the prototype, have only slightly larger H I FWHMs than these.
There is no common pattern for the broad-line profiles, as there is for the narrow-line profiles. Approximately half the well observed objects have essentially symmetric broad-line profiles; the remainder with asymmetric profiles in some cases extend further to the blue, in others, to the red. Thus radial flow combined with extinction is not the main velocity field in the BLR.
There are real differences among the broad-line profiles in a single Seyfert 1 galaxy. Usually H is somewhat broader than H (1.16 ± 0.05 average deviation is the ratio in the sample mentioned), He I still broader (1.36 ± 0.05), and He II probably still broader, though its profile is generally badly blended with neighbouring Fe II features and is less well defined. All calculations of the H I spectrum under BLR conditions show that the H / H intensity ratio increases with an increasing Ne and with increasing ionization parameter . Thus, from the observed profiles, the velocity dispersion in the BLR must increase with either one or other of these parameters. Thus if for instance Ne = constant, the velocity dispersion must increase inward with , which increases as R-2; or if = constant, the velocity dispersion must increase inward with Ne, which increases as R-2 for constant . Likewise, calculations show that the intensity ratio He I 5876 / H increases with Ne and , and the same argument applies again. Thus the profiles show that the velocity dispersion increases inward in the BLR. In addition the Fe II emission lines have FWHMs typically somewhat smaller than the broad H I lines. The exact value is difficult to quantify, because all the Fe II multiplets are badly blended, but ratios of FWHMs of Fe II to H between 0.8 and 1.0 will satisfy the observational data. This again indicates that the velocities decrease outward in the BLR.
A rotational velocity field about a central black hole has this property and it is natural to assume that it is responsible for the observed profiles. Calculations have been made which show that simple models combining rotation with outward directed flow, perpendicular to the disk, will produce profiles that match the best-quality observed ones. Likewise, however, assumed velocity fields with only the outward flow confined to a cone whose axis is normal to the disk, assumed density distributions, cutoffs in the flow velocity, or cone opening angles will fit almost all of the observed profiles, if the asymmetries are ignored. These are kinematical models only, but they indicate at least some of the directions to explore for complete physical models (Osterbrock and Mathews 1986).
Much the same is true of the broad emission-line profiles in QSOs. Different broad lines tend to have similar, but not identical profiles. The lower-ionization lines probably have somewhat smaller FWHMs than the higher-ionization lines, as in Seyfert 1 nuclei. L, the strongest emission line, does not have a particularly asymmetric profile, as would be expected for such an optically thick resonance line, if the velocity field was primarily radial inflow or outflow. The FWHMs of C IV 1549, a strong and quite unblended line, in a large sample of quasars (primarily) and QSOs range from about 2000 to 7000 km s-1, with a mean of 4500 km s-1, somewhat higher than in Seyfert 1 galaxies, in which the mean is more like 3000 km s-1 (Wilkes 1984, 1986).
The earliest physical models of the velocity field in the BLR considered radiation-pressure-driven outflow. This predicts a logarithmic form for the line profile, which approximately fits many of the observed profiles. Additional acceleration by a hot wind helps to increase the stability of the clouds, and essentially preserves the same form of line profile. However, as previously mentioned, essentially all such wind media that would lead to stability have been ruled out on the grounds that they should be observable in one spectral region or another, but have not been observed (Mathews and Ferland 1987, Mathews and Doane 1990).
If the observed broad-line profiles represent outflow, the rate of mass loss ranges from 0.1 to 10 M y-1. This is of the order of the mass accretion rate. If we imagine that an AGN's lifetime is of the order of a few percent of the Hubble time, corresponding to the fraction of spirals that have AGNs, the integrated mass-loss rate is comparable with the mass in the black hole, which does not seem unreasonable.
If the velocity field is largely rotational flow, the mass-loss rate is much smaller. In addition, the magnitudes of the velocities deduced from the line widths comparable with those expected under gravitational forces. This has been particularly emphasized by Wandel (1989) who found that the velocity dispersions from the FW0Is lead, through the relation v2 = GM / R, to a dynamical mass for the black hole that agrees fairly well with the value for the mass found from fitting the ultraviolet continuum with an accretion disk model. However, there are many only poorly-known parameters and the specific numerical values are uncertain. The most recent and complete data on dynamical masses found in this way are in Padovani et al (1990).
The best information on the nature of the BLR velocity fields should come from the temporal-variation measurements summarized in section 2.7. By determining the time-lag between variations in the continuum anti various parts of the profile, for instance the red side or the blue side, it is in principle possible to localize the motions to some extent. Gaskell (1988), using IUE spectra of NGC 4151, found that the variations in the blue wing of CIV 1549 lagged the red one by 3.4 ± 3.4 day, and the same for Mg II 2798 by 4.5 ± 3.1 day. He interpreted this as evidence for inflow (the nearer side of the object has a redshift with respect to the center), but the effect is too small to be convincing. Much better ground-based data for this same object give a lag of only 1 day for the blue wing of H, considerably smaller than the probable error. Purely radial motion, either inflow or outflow, is ruled out by these data. Combinations of chaotic and orbital motion are allowed (Maoz et al 1991).
On the other hand for the considerably more luminous (and thus larger) QSO Fairall 9, Koratkar and Gaskell (1989), again using IUE spectra taken over several years, found R = 124 ± 39 light day as the time lag for CIV 1549 with respect to the continuum, and the direction of motion preferentially inward. The chance of purely random or chaotic motion was given as only 10%. Mg II 2798 gave a somewhat larger mean size, 167 ± 61 light day, and a chaotic (which includes orbital in all directions) velocity field.
Finally, using the IUE spectra of six objects taken from 1978 to 1988, Koratkar and Gaskell (1991) found that both pure outflow and inflow were excluded at the 99% level. Purely chaotic or circular motion is not ruled out, though these authors prefer the disk or isotropic orbital interpretations. However, much observational research is going on in this field at present, and none of these results should be regarded as final. The recent large study of regular, frequently spaced spectra of NGC 5548 should give much more definite information but the results for the velocity field were not available at the time of writing (Clavel et al 1991, Peterson et al 1991).
The BLR, close to the nucleus, is the most interesting but least well understood part of active galactic nuclei. The velocity field is particularly uncertain: taken to be outflow in most theories, demanding a confining medium which does not seem to be there, and with the observational evidence favouring chaotic or orbital motions, but still highly uncertain.
One very interesting idea, recently put forward or actually revivified, is that the `clouds' in the BLR are actually the atmospheres of stars, held together by gravity rather than by an external medium. One version has them as red-giant stars, another version as main-sequence stars 'bloated' by the intense radiation field near the central source. In any case they are photoionized and heated by the central source; their surfaces are continuously flowing off as a wind and being ablated. The calculations, all quite recent and rather simplified, suggest that the observed emission-line spectrum and profiles can be produced by such models. One problem lies in replenishing the supply of stars, for their rate of mass loss is high (Voit and Shull 1988, Norman and Scoville 1988, Penston 1988, Kazanas 1989).
5.5. Possible magnetic fields
There is certainly ionized gas in AGNs, and almost certainly rotation. Under these circumstances, magnetic fields may naturally arise. The observed radio emission is synchrotron radiation, the result of the acceleration of relativistic electrons in a magnetic field. At the center of our own Galaxy, long filaments of radio emitting plasma are observed, clearly associated with a poloidal type magnetic field. Thus very probably magnetic fields are prevalent in AGNs, and especially in the BLRs close to their centers. Some of their effects on the radiation and structure have been discussed (Rees 1984). Otherwise they have not been very much considered in discussions of the velocity fields, but it is clear that they may be important in some situations (Greyber 1989).