The fundamental question that we still have not yet addressed is what is the nature of the BLR? What is the origin of the gas that gives rise to the emission lines, and how is it related to the accretion flow, if at all? A number of different scenarios have been proposed, and we discuss some of these briefly.
5.1. "Cloud" Models
Our first notions of what the BLR might look like was based on observations of Galactic nebulae, especially the Crab Nebula system of "clouds" or "filaments," partly because of the obvious merits of such an interpretation (e.g., we noted earlier that supersonic motions of BLR gas argue for such a system or some kind of large-scale flow), but probably also because the first astrophysicists to work on the problem had previously worked on nebular physics in Galactic sources.
If we suppose that the BLR is comprised of some large number Nc of identical line-emitting clouds of radius Rc, the BLR covering factor (i.e., fraction of the sky covered by BLR clouds, as seen by the central source) will be proportional to Nc Rc2. The covering factor is determined by estimating the fraction of ionizing continuum photons absorbed by BLR clouds and re-processed into emission lines, and is constrained to be of order ∼ 10% by the equivalent widths of the emission lines, notably Lyα. An independent constraint is given by the total line luminosity, which is proportional to Nc Rc3, i.e., the total volume of line-emitting material. By combining these two independent constraints on Nc Rc2 and Nc Rc3, we can independently solve for Nc and Rc. For a typical Seyfert galaxy like NGC 5548, we find Nc ≈ 107 and Rc ≈ 1013 cm. Furthermore, we can combine the size of the cloud with the putative particle density of ne ≈ 1010 cm-3, to get a cloud column density of NH ≈ 1023 cm-2. Interestingly and ultimately coincidentally, this is the same order of magnitude as the first measurements of the column densities of "warm absorbers" detected in the X-ray spectra of AGNs, and it was natural to ascribe this absorption to BLR clouds along the line of sight to the very small X-ray continuum source. Finally, given the number, size, and density of the clouds, it is trivial to compute a mass for the entire BLR, which works out to be ∼ 1 M⊙.
Early views on cloud dynamics were also influenced by observations of supernova remnants, the only other astrophysical environment in which line-emitting gas moves at velocities higher than 1000 km s-1. Either ballastic outflow (as in supernovae) or radiation pressure driven outflow can produce logarithmic line profiles. A preference for outflow models was at least in some part driven by concerns that a gravitationally bound BLR seemed quite implausible. Prior to the advent of reverberation mapping, there was no way to determine the size of the BLR other than photoionization equilibrium modeling. Sizes predicted by single-zone (i.e., all lines produced in approximately the observed ratios by a single representative cloud) models were about an order of magnitude too large, thus leading to mass estimates that were far too large to be consistent with observed nuclear stellar kinematics.
The argument on the number of emission-line clouds is, however, flawed. The fact that the emission lines vary in response to continuum variations argues that the clouds are optically thick, in which case the emitting volume of a cloud is proportional to Rc2 RS where RS is the depth of the ionized layer of the cloud (i.e., the Stroömgren depth), rather than Rc3, which removes the independent constraints on Nc and Rc. However, there is a second argument that also argues for a large number of clouds, namely the smoothness of the emission-line profiles, as shown in Fig. 10. Here the argument is that a collection of clouds with a velocity dispersion of a few thousand km s-1, but each emitting lines of thermal widths of 10 km s-1, ought to produce a rather "grainy" composite line profile on account of statistical fluctuations in the number of clouds at different line-of-sight velocities. The lack of grainy structure at high spectral resolution and high signal-to-noise ratios argues either that the number of clouds must be very large or the BLR gas is some kind of continuous flow rather than in the form of discrete clouds. Even an extreme case, NGC 4395, the least luminous known Seyfert 1, has characteristically smooth line profiles. But the BLR in this source is expected to be so small that the number of individual clouds could not exceed a few thousand. This essentially leaves us only with models that involve some kind of supersonic flow, unless there is another significant source of microturbulence within the broad-line clouds.
Figure 10. Keck high-resolution spectra of the Hα and Hβ emission lines in NGC 4151. Some contaminating narrow-line features are labeled. While there is structure in the line profiles on scales of hundreds of km s-1, on thermal scales the profiles are very smooth. From Arav et al., 1998, MNRAS, 297, 990.
Another interesting consequence of variability observations is the realization that there must be a large reservoir of gas throughout the BLR, but at any particular time we are detecting primarily the gas that is emitting most efficiently . Thus the total mass of gas in the BLR is much larger than computed above. Estimates of the BLR mass run as high as 103–104 M⊙.
5.2. "Bloated Stars" Model
An early suggestion for the origin of the broad emission lines was gas provided by stars in the nucleus, e.g., . This solves a number of problems, such as fuel supply and cloud stability, but encounters other problems, notably that, except in the case of giant stars, which are relatively rare, the stellar surface gravity is too high to easily remove gas. Another problem is whether or not one can fit an adequate number of stars into the BLR, i.e., a variation on the "number of clouds" problem referred to above.
5.3. Double-Peaked Emission Lines
A relatively small subset of AGNs have very broad double-peaked Balmer line profiles, as shown in Fig. 11. Double-peaked profiles are characteristic of rotating disks; such profiles are commonly observed in accretion disks in Galactic binaries. Sometimes double-peaked profiles appear in the variable part of an AGN spectrum, i.e., the line profile that appears in difference spectra (i.e., high-state spectrum minus low-state spectrum) or in rms spectra.
Extensive efforts to model such disks indicate that they must be fairly complex. Profile variations show that the disks are clearly not axisymmetric, and they sometimes have large-scale structural or thermal irregularities. Sources with double-peaked emission lines are certainly important to our understanding of AGNs, as these are the only sources where we can identify a plausible disk structure. Whether or not such disk structures are present and merely hidden or disguised in other AGNs remains an open question. Clearly, however, disk-like structures cannot explain everything about the BLR.
5.4. Disk Winds
There is increasing evidence for the existence of disk winds in AGNs, e.g., [6, 10, 19]. Disk winds are observed in both Galactic binaries and in young stellar objects, and may be a feature common to accretion disks on all scales. There has been much theoretical investigation in this area, but it is not clear whether AGN disk winds ought to be driven radiatively or hydromagnetically, or perhaps by some hybrid.
It may well be that what we think of as the BLR is itself a composite. Photoionization equilibrium modeling has suggested that there must be two distinct regions, one highly ionized and of moderate optical depth and another of moderate ionization and greater optical depth, but of similar physical scale and distance from the central source, e.g., . The latter, which is largely responsible for the Balmer-line emission, has been often identified with the disk itself; and indeed the double-peaked Balmer lines in some sources support this. The higher ionization lines, however, may represent a disk wind component. Observational evidence for this includes:
clear blueward asymmetries in the higher-ionization lines in narrow-line Seyfert 1 galaxies, e.g., ;
the peaks of high ionization lines tend to be blueshifted relative to systemic, and the maximum observed blueshift increases with source luminosity, e.g., ;
in radio-loud quasars, the width of the bases of the C IVlines are larger in edge-on sources than in face-on sources, implying that the wind has a strong radial component to it, e.g., .
The disk wind model also affords a possible connection to the outflows detected in AGNs of all luminosity. Strong absorption features, generally blueshifted relative the systemic velocities, are ubiquitous features of the X-ray and UV spectra of AGNs, e.g., . In high-luminosity quasars, the outflows are manifest as "broad absorption-line (BAL)" quasars. In lower-luminosity Seyfert galaxies, the features are weaker, but may still cover a fairly large velocity range, but with multiple discrete components as opposed to more-or-less continuous absorption troughs as in BAL quasars. In Seyfert galaxies, the amount of kinetic energy in the absorbing region can easily be of the same order as the radiative energy; in any given case, this calculation is subject to uncertainties due to unknown covering factors, which are quite reasonably assumed to be high because strong absorption features are so common in these sources.