Though classification is an empirical process, we always hope that it will isolate physically significant types of objects, and will lead to understanding them, as in the case of stars. From many studies we understand the main features of Seyfert galaxy spectra as resulting from photoionization by a spectrum that extends to high-energy photons. The gas is in clouds with a small filling factor, the dense, broad-line gas closest to the source, and the less-dense, narrow-line gas on the average at larger distances. The diagnostic line ratios of the narrow-line spectrum, and the equivalent widths of the broad-plus narrow-line spectra agree with this picture. (For a recent review of this material see e.g. Osterbrock 1984.)
An observational model that fits many of these data and others connected with the differences in optical spectra between Seyfert and radio galaxies is shown schematically in Figure 2 (Osterbrock 1978). The dense (BLG) and low-density (NLG) gas components are shown in cylindrical and spherical distributions, respectively, about the central photoionizing source, presumably connected with an accretion disk and black hole. Ionized regions are indicated by cross hatching; most highly ionized regions closest to the source, where the ionization parameter is largest, for a fixed density. In this model, both the ionizing photons and the radio plasma can escape through the BLG most easily along the axis of the cylinder, where the amount of material encountered is the smallest, and correspondingly the ionized NLG has a roughly conical distribution about this axis. In this picture pure Seyfert 1 galaxies are those in which the amount of BLG is large and few ionizing photons reach the NLG; pure Seyfert 2 galaxies are objects in which the amount of BLG is small or nonexistent and nearly all the ionizing photons reach the NLG. Intermediate objects in the classification scheme are those which have intermediate amounts of BLG, according to this model.
Figure 2. Schematic observational model of an AGN as described in text. Ionized gas is shown in various cross-hatchings as indicated, neutral gas is shown as plain white.
Furthermore, there are many indications that the active galactic nucleus may be tipped with respect to the main body of the galaxy in which it occurs. This is suggested by the lack of correlation between the FWHMs of the broad-emission lines and inclination of the parent galaxy, that would be expected if the velocity field were rotational, or indeed had any other type of cylindrical symmetry (Tohline and Osterbrock 1982). Also, there are many observational indications that the structures of the NLG and the radio plasma are closely connected, and high-resolution radio-frequency maps show that the jets or axes of the nuclear structure are often misaligned with the parent galaxy (Wilson and Ulvestad 1982; Ulvestad and Wilson 1984; Wilson, Baldwin and Ulvestad 1985).
A schematic drawing of such a model is indicated in Figure 3 (Osterbrock 1983). The small cylindrically symmetric. BLR, at the center, is tilted with respect to the larger cylindrically symmetric NLR. (As with Figure 2, these models cannot be drawn to scale correctly, for the ratio of sizes of the two regions is ~ 103.) The core about the axis of the BLR, in which most of the ionizing photons that get to the NLR escape, is shown by the dashed lines, and the units of the structure implied by the small filling factor are schematically indicated as spherical, homogeneous, uniform-density clouds. The various shadings and hatchings indicate that the ionization is highest on the faces of the clouds closest to the ionizing source (under the simplifying assumptions stated), and drops to nearly zero at the back (or side away from the ionizing source) of the optically thick clouds. The picture is simplified; the BLR must have an inhomogeneous structure also, that can be mimicked by similar but denser clouds on a much smaller scale. They cannot be drawn in the figure, but coincidental "tunnels" between them may allow ionizing photons to escape in some few directions outside the primary cone, as shown by the ionized clouds in the NLR, near the projection of the equator of the BLR.
Figure 3. Schematic, slightly more sophisticated observational model of an AGN as described in text. Tilt of BLR with respect to NLR, and ionization structure of clouds in NLR are indicated.
In addition, there are many observational indications from line-profile studies that the division into two regions, BLR and NLR, is a gross oversimplification. In reality there is a wide range of density, extending more or less continuously from what we call the "BLR" through the "NLR" (De Robertis and Osterbrock 1984). Correspondingly, the ionized gas in the AGN probably has a smoothly warped distribution, with the greatest misalignment at the center, in the densest, smallest part of what is conventionally called the BLR, and merging into the plane of the galaxy at the outer part of the "NLR". The density distribution everywhere must be highly inhomogeneous, or "cloud"-like, so the picture is a complicated one indeed.
In testing whether or not there is a correlation between emission-line width and inclination of the parent galaxy, Keel (1980) discovered that there is a paucity of nearby edge-on Seyfert 1s. This indicates that there exist a fairly large number of galaxies, with the physical properties of Seyfert 1s, not detected or known because of extinction by the dust outside their nuclei, but in their principal planes. The most nearly edge-on nearby Seyfert 1 galaxy, NGC 4235, indeed has an unusual spectrum, in which both the broad and narrow emission lines show the effects of strong reddening, as does the continuum (Abell, Eastmond and Jenner 1978; Osterbrock 1984).
In addition to the conventional NLR, direct images and spectra to very faint limits show the presence of gas photoionized by hard photons from the central source and at distances of several kpc from it, in a few very well studied active galaxies, as will be reported, for instance, by Pogge at this Symposium. The distribution of this ionization appears to be more or less conical about an axis that can be associated with the central source (Unger et at. 1987; Pogge 1988). This is a strong argument in favor of models of the general type represented in Figures 2 and 3.
In Seyfert 1.8 and 1.9 galaxies, the ratio of the intensities of the broad emission-line the components H / H is often large, typically perhaps 8-10. Though this ratio can have a range us, of values under BLR conditions, depending on density (collisional excitation and deexcitation), size (line optical depths) and velocity fields (ditto), the fact that it is so large in Seyfert 1.8 and 1.9s suggests that dust extinction may be especially significant in them (Osterbrock 1981). Likewise, there are X-ray galaxies with very weak broad H components, so weak that they an would not be classified Seyfert 1.9 on spectra of normal signal-to-noise ratios; they may be objects in which the dust extinction is even stronger (Veron et at. 1980; Shuder 1980). This idea has been carried still further by the suggestion that all Seyfert galaxies have approximately the same structure, and the same relative amounts of BLG and NLG, but have different amounts of extinction as seen from different directions. According to this hypothesis, the Seyfert 2s are objects in which the BLR is completely hidden by extinction, from the direction in which we see it (Lawrence and Elvis 1982).
A very great step was the recognition of the "hidden Seyfert 1" spectrum seen in the polarized flux from NGC 1068 (a Seyfert 2 in the total flux) by Antonucci and Miller (1985). From this observation they developed a picture of the BLR within a tipped, cylindrically symmetric torus, containing dust, that is optically very thick in its equatorial plane, but much thinner or transparent along its axis. It has many similarities to the earlier models described above, but goes far beyond them. As these observations and ideas will be described by Miller himself at this Symposium, I will go into no more detail here. One most interesting aspect of this model is the question of whether it applies to all, or most Seyfert 2 galaxies, that is, if they all have "hidden Seyfert 1" broad-line regions within them. Miller and his collaborators have made spectropolarimetric observations of several more Seyfert 2s, and has found similar broad emission-line features in the polarized light of some, but not all of these, as he explains in his paper. Krolik and Begelman (1986) have proposed a picture according to which all Seyfert 2s are of this type, and Krolik will describe further progress along these lines in his paper at this Symposium.
Certainly there is a wide range of properties among observed Seyfert 2s, and among Seyfert 1s also. One interesting group is the narrow-line Seyfert 1 galaxies, as described by Osterbrock and Pogge (1985). We have found more of these objects among the AGN candidates isolated by the IRAS survey (Osterbrock and De Robertis 1985) and the PG survey (Osterbrock and Pogge 1987). The latter provides fairly good statistics in the relative number of galaxies of this type; if we adopt the definition that a Seyfert 1 or 1.5 galaxy with broad line FWHM < 2000 km s-1 is to be called a narrow line Seyfert 1, approximately 15% of the Seyfert 1 and 1.5 galaxies belong to this group. I Zw 1 is one member, Mrk 507, recently described by Halpern and Oke (1987), is another. One immediate possibility is that the narrow-line Seyfert 1s may be the most nearly "pole-on" members of the Seyfert 1 class, objects in which the component of the velocity field (if it is rotational, or has some other origin but is largely in a plane) along the line of sight is small. Observational aspects of this possible working hypothesis will be discussed by Goodrich at this Symposium.
These examples, and others, show observationally that all Seyfert galaxies do not fit into a single, one-parameter sequence, or even into a two-parameter grid. Observationally, the parameters include luminosity, relative strengths of the BLR and NLR, line widths of the broad lines, line widths of the narrow lines, level of ionization, and relative strength of the Fe II emission lines. (Their relative strength seems to be partly but not completely anticorrelated with high ionization.)
On the other hand the physical parameters, in addition to the overall physical structure, include mean ionization parameter , mean electron density Ne, the variation or both of these with distance (in the simplest spherical models), or with , z (for the simplest cylindrically symmetric models), or with r, , z in general, plus the fluctuations or distributions around these mean values (the "filling factor" is one of the simplest possible descriptions of these), and the internal velocity field. Our aim is to understand AGNs by recognizing the connections between the observational and physical parameters.