4.1. Cylindrical symmetry
There are many reasons for supposing the structure of AGNs to be more nearly cylindrically symmetric about an axis than spherically symmetric about a point. One is the theoretical idea of an accretion disk, the other is the observation in many objects of radio-frequency jets, which appear to be axes of the structure (Shields 1978, Osterbrock 1978). There is no strong observational reason to suppose that the axis (angular momentum vector) of the AGN coincides in direction with the axis of the galaxy in which it is, and comparison of the directions of jets with images of their host galaxies suggest they are often not in the same direction (Tohline and Osterbrock 1982, Ulvestad and Wilson 1984b). Thus the highly schematic AGN of figure 1, in which the BLR is disk-shaped, should be understood as having its axis in an arbitrary direction. In this picture ionizing photons can escape into the NLR along the axis, but not along the equatorial plane of the BLR where the optical depth is greater. There are many indications that the lower-density NLR merges continuously into gas in the disk of the host galaxy. Hence a schematic picture like figure 5 might be more appropriate, in which the BLR is tipped with respect to the NLR, also shown as a disk, but with the density condensations represented as spherical clouds. In reality no doubt there is a continuous transition from BLR to NLR both in density and in axis or plane; thus the NLR would be warped, as is observed for gas near the center of our galaxy and in many other spirals in which the interstellar H I distribution has been mapped.
Figure 5. Schematic AGN model, showing tipped BLR disk in larger NLR disk. Most of the ionizing photons escape from the BLR disk in the cone about its axis, but a few `tunnels' may exist in other directions. Highest degree of ionization, indicated by close cross-hatching, occurs on faces of clouds closest to the central accretion disk. Actually the BLR probably merges and warps continuously into the NLR. (Osterbrock 1978b)
There are very few nearly edge-on Seyfert galaxies, as detected by their optical emission lines. (NGC 4388 is one of the rare known exceptions.) No doubt this is due to extinction by dust in the planes of their host galaxies (Keel 1980). However, hard x-rays can escape freely, and indeed there are Seyfert 1.8 and 1.9 galaxies with luminosities in this band comparable to Seyfert 1 galaxies, but with much weaker broad H emission. Some of them are nearly edge-on, but others among them are not, suggesting that considerable extinction also occurs in the plane of the BLR, and that these objects are seen with their BLR edge-on to us. These and other correlations of x-ray luminosity, path length along the line of sight to the nucleus as shown by soft x-ray absorption, and narrow-line luminosity led to the suggestion that many if not all Seyfert 2 galaxies contain BLR regions, which however are obscured by the dust surrounding them, more or less in the neutral regions marked BLG0 on figure 1 (Lawrence and Elvis 1982).
4.2. Seyfert 2s with `hidden' BLRs
The earlier interpretation of the differences between Seyfert 1, 1.5 and 2 galaxies was in the amount of dense gas in the BLR, and the relative numbers of ionizing photons absorbed in it and in the NLR. However, that this is not correct in all cases was clearly shown by spectropolarimetric measurements of the Seyfert 2 AGN, NGC 1068 (Antonucci and Miller 1985, Miller 1988). These data show that the forbidden lines are only slightly polarized, by approximately 1%. However, after removal of the galaxy absorption-line continuum, the featureless continuum (presumably radiated by the nucleus or accretion disk itself) is much more strongly plane-polarized. The degree of polarization is approximately 16%, independent of wavelength and in a constant direction. The direction of polarization is with the E vector approximately perpendicular to the axis of the radio jet in this AGN. Furthermore, in the polarized light spectrum, weak broad emission lines of H I and Fe II, characteristic of a Seyfert 1 spectrum, can be seen. Their FW0I are about 7500 km s-1, much wider than in the Seyfert 2 spectrum seen in natural light. More recent spectropolarimetric measurements of eight other Seyfert 2 nuclei with high polarization show that at least four of them have similar `hidden Seyfert 1' spectra, broad H I and in some cases Fe II emission lines, visible only in the plane-polarized spectrum (Miller and Goodrich 1990).
A general property of these and still more Seyfert 2 nuclei observed with broad-band polarimeters is that their continua are plane-polarized perpendicular to the jet or other radio structure. On the other hand, Seyfert 1 nuclei have much weaker polarization, and in the opposite sense, namely parallel to the radio jet or linear structure (Antonucci 1983). This same direction also occurs in quasars and QSOs. In all the objects for which there are spectropolarimetric measurements, the degree of polarization is independent of wavelength over the observed range, suggesting that electron scattering is the mechanism. The suggested interpretation is that in Seyfert 1 nuclei a thin disk is seen directly, more or less face on, and the polarization is weak but parallel to the axis because it is averaged over the disk. On the other hand in the Seyfert 2s the disk is not seen directly, but radiation from the central source and the BLR which escapes along the axis (of the cone marked NLG+ in figure 1) is seen only if it is scattered toward us. Hence it is polarized on the average perpendicular to the axis of the disk. A schematic thick-disk model of this type is shown in figure 6. From the widths of the plane-polarized H I emission lines in NGC 1068, an upper limit to the temperature of the scattering electrons is T < 106 K. In NGC 1068 Miller (1988) has also made spectropolarimetric measurements of regions near but outside the nucleus, namely a strip more or less along the direction of the jet, and a bright knot about 5 in ( 400 pc) northeast of the nucleus. These areas are also strongly polarized and show the broad H I and Fe II emission lines in polarized light, in each case with plane perpendicular to the direction of the nucleus. In the knot the H I lines are not as broad as in the nuclear polarized-light spectrum, indicating that dust scattering rather than electron scattering is dominant in the knot. This interpretation is borne out by the differing wavelength dependence of the polarized light fluxes from the knot and the nucleus. From the temperature limit previously stated, the scattering electrons appear to be part of an outward flowing wind, Compton heated by the direct ultraviolet and x-ray radiation of the central source, and cooled by expansion. The total optical depth is 0.1. A predicted signature of this temperature, density and radiation field is Fe K x-ray line and continuum emission (Krolik and Kallman 1987, Krolik and Begelman 1988). This feature has been observed, with a strength that agrees with the hidden Seyfert 1 picture of NGC 1068 (Koyama 1989).
Figure 6. Schematic drawing of central source of AGN, surrounded by thick accretion torus, which channels radio plasma and ionizing radiation along the axis. Directions from which the object's spectrum is Seyfert 1, 1.5 and 2 are marked (Pogge 1988c).
NGC 1068 is a unique Seyfert 2 in several respects, as previously mentioned. Hence the observations of `hidden BLRs' by the same spectropolarimetric technique used on the other Seyfert 2 AGNs, as mentioned before, is important in establishing that it occurs in more than this one object. Whether all Seyferts have this same structure is another question. Half the eight Seyfert 2 galaxies observed spectropolarimetrically to date show the presence of a `hidden' Seyfert 1 AGN, but this sample of eight was selected as objects of known high polarization. If all Seyfert galaxies were physically the same type of objects, seen from different orientations, many more highly polarized Seyfert 2s would be expected than are observed (Miller and Goodrich 1990). As described in the previous section, the infrared and radio properties of Seyfert 1 and 2 galaxies are now understood to be relatively similar. Thus the picture is an attractive one, but far from proven at present.
4.3. Ionization cones
The nearest Seyfert galaxy NLRS can be resolved. One of the most studied is again the bright Seyfert 2 NGC 1068. As previously mentioned, it is far from typical of this class, but its nearness makes it a rewarding object for study. Very complete long-slit data are available (Baldwin et al 1987), and also Fabry-Perot imaging data (Cecil et al 1990). These two papers are the most recent of a continuing series of improvements in detector technology, going back to the photographic spectra of Walker (1968), who first mapped the velocity field and homogeneities in this object.
The long-slit spectra cover the central 10 kpc x 2 kpc (140" x 30") with good wavelength and spatial resolution. They show that in the outer part of the galaxy (r > 1000 pc, based on H = 75 km s-1 Mpc-1) the gas lies in a fairly normal disk, with a flat rotation curve. Within this distance the velocity field is more complicated, containing several condensations or high-velocity clouds with velocities up to several hundred km s-1. In this inner region these optical condensations tend to lie along the axis of the radio jet, as recognized by Wilson and Ulvestad (1983).
The velocity field will be discussed in the next section, but the emission-line spectra themselves show a mixture of two disk components projected on top of each other. They can be separated from each other by the diagrams of figures 2, 3 and 4. One component consists of normal H II regions, photoionized by hot stars. They are particularly strong in a ring of about 15" (1 kpc) radius, and within it. The other component consists of gas photoionized by the central source within the AGN, as shown by the fact that measured line-ratios for it cluster close to the AGN lines in the diagnostic diagrams. This component is located in space along the axis of the jet and its extension. It can best be seen in ionization maps based on continuum-subtracted images taken with interference filters centered on [O III] 5007 and on H + [N II] 6548, 6583 in the rest system of the Seyfert galaxy. H alone would be preferable, but the somewhat wider filters which include the [N II] lines are also easier to obtain and use. Dividing the [O III] image by the H + [N II] image, after first cutting out the lowest levels of each (which are dominated by noise), and then plotting the contours only of high [O III] / (H +[N II]), essentially the same as high [O III] / H, produces a qualitative `ionization map'. The regions it shows are those photoionized by the hard spectrum of the AGN nucleus, while the H II regions photoionized by hot stars drop out (Pogge 1988a). In NGC 1068 the region photoionized by the nuclear continuum is triangular-shaped, (evidently a cone in projection) centered on the axis of the jet and its extension. It disappears where it crosses the ring of H II regions (which are brighter than it), and then reappears further out. It can be followed to a distance of about 25" (nearly 2000 pc). The observed structure is quite similar to that predicted for one side of the ionized `narrow-line gas' in models such as those shown in figures 2, 3 and 4. Evidently the other side of the cone lies below the central plane of this galaxy, rich with dust, and is therefore not seen.
Similar ionization maps for approximately 20 other nearby, non-interacting Seyfert 1 and Seyfert 2 galaxies have been obtained (Pogge 1989a, 1989b). Of the eleven Seyfert 2 galaxies, eight show extended ionized gas, with strong [O III] 4959, 5007, and emission-line ratios consistent with photoionization by a hard source. The [O III] / (H + [N II]) maps reveal distinctly conical high-ionization regions, with the nucleus at the apex, in four of these eight Seyfert 2 galaxies. In the objects for which high-quality radio continuum maps are available, the cones are aligned with the axes of the cones. Although the number of objects is still small, this is strong evidence for the ionization-cone picture. The opening angles of the cones range from 30° to 90°, with a mean of perhaps 50°. Only three of the nine Seyfert 1 galaxies show extended high-ionization emission. In none of them does the ionization map show either a cone, or a halo (as a cone seen approximately end on, from a direction within it might appear). These data thus do not show the difference between the presumed `bare nucleus' Seyfert 1s and `hidden nucleus' Seyfert 2s that would be expected from this picture. None of these Seyfert 1s appear to be `pole-on' versions of the Seyfert 2s with extended conical photoionized regions. Although there is much good evidence that some if not all Seyfert 2 galaxies have a hidden BLR, the statement that Seyfert 1 nuclei have the same structures seen face-on is not explicitly confirmed.
In an independent study, ten nearby Seyfert galaxies with known linear radio structure were imaged in [O III] and H (Haniff et al 1988). Most of the objects are Seyfert 2s, but two, Mrk 6 and Mrk 79 are Seyfert 1.5s. In all ten cases the [O III] emission outside the nucleus is aligned with the radio structure to the accuracy of the data (about 5°). There is also good agreement between the H and radio structure, but not as much as with the [O III] images, indicating the relatively greater contribution of H II regions to the H image. Furthermore, these same authors have found evidence for anisotropic emission of the ionizing radiation from the nucleus in at least two of these Seyfert 2 galaxies, Mrk 3 and Mrk 78 (Wilson et al 1988). In each object they estimated the available number of ionizing photons from the nucleus from its optical, ultraviolet or x-ray brightness, a plausible extrapolation of its spectrum to h > h 0 = 13.6 eV by a power law, and an estimate of the solid angle subtended at the nucleus by the photoionized gas. The number of photoionizations required is given directly by the integrated flux of the extended emission in H or H. For both Mrk 3 and Mrk 78 the discrepancy is at least a factor of ten, probably indicating that the ionizing flux which escapes to the extended emission in the direction of the axis of the jet is significantly greater than expected from the continuum flux which escapes in our direction. The connection with the torus model is clear; it is sometimes referred to as the `beaming' of ionizing radiation. However, it is also possible that the ionizing flux is larger than the extrapolation of the power law would suggest.