Next Contents Previous


The relative strengths of the various emission lines in AGN spectra are very similar to those emitted by a wide variety of astrophysical plasmas: to some low order of approximation, the emission lines of planetary nebulae, H II regions, supernova remnants, and AGN emission-line regions are all quite similar. The reason is simple: photoionization equilibrium of all these gases is achieved at the same temperature, T ≈ 104 K, e.g., [20, 21, 23, 28].

Photoionization equilibrium is attained when the rate of photoionization is balanced by the rate of recombination. The conditions where this occurs can be expressed in terms of an "ionization parameter"

Equation 1


where Qion(H) is the number of H-ionizing photons produced per second by the central source, i.e.,

Equation 2


and the integral is over ionizing photon energies and Lν is the specific luminosity of the ionizing source. The ionization parameter is thus the ratio of the density of ionizing photons Qion / 4 π r2 c, reflecting ionization rate, divided by the electron density ne, reflecting recombination rate, at the face of a cloud exposed to the radiation. Over a rather broad range in U, photoionization equilibrium is realized at electron temperatures in the range 10,000 to 20,000 K. As a result, the line spectra emitted by such gases are very similar, since variations in elemental abundances are also relatively minor. There are, of course, more subtle spectral differences among these various 104 K gases which are attributable to differences in gas density ne, the input ionizing spectrum, the gas dynamics, and elemental abundances (which appear to be approximately solar or slightly enhanced in AGN emission-line regions) [14].

Since at least the work of Khachikian and Weedman in the mid-1970s, it has been recognized that AGN emission-line spectra are kinematically composite. The "narrow components" have Doppler widths usually less than around 500 km s-1; these emission lines arise in relatively low-density (ne ≈ 103 cm-3) gas that is spatially extended – indeed, the narrow-line region (NLR) is at least partially resolved in some of the nearest AGNs. In contrast, the "broad components" have Doppler widths in the range ∼ 1000 to 25,000 km s-1 and arise in gas of fairly high density by nebular standards (i.e., ne > 109 cm-3), as determined from the weakness of certain metastable and forbidden lines that are relatively more prominent in lower-density gases.

Indeed, the gas density leads us to an important distinction that is not obvious based on line widths alone: the widest narrow lines and the narrowest broad lines have similar Doppler widths, around 1000 km -1 (though not in the same object: in general, there is a correlation between the widths of the narrow and broad lines in AGN spectra). However, even the subclass of broad-line objects known as narrow-line Seyfert 1 (NLS1) galaxies still have distinct "broader" and "narrower" features, with the difference between them being the absence of the "broader" component in the forbidden or semiforbidden lines, i.e., from the high-density gas in which these transitions are collisionally suppressed.

Temperatures of order 104 K correspond to thermal line widths of order 10 km s-1, so it is clear that in both the NLR and the BLR, the gas moves supersonically. This in turn suggests some kind of organized flow, or a system of discrete "clouds." The especially large Doppler widths of the broad lines immediately suggest that they may arise in a deep gravitational potential, which makes the broad lines especially valuable as probes of the central source.

Compared to the NLR, the actual amount of emission-line gas required to produce the broad emission lines can be quite modest as line emission is very efficient in high-density gases (the emissivity per unit volume is proportional to ne2). We also observe that the emission-line fluxes vary with the continuum flux, but with a short time delay. From this we infer that the broad-line emitting gas is photoionized and optically thick to ionizing continuum radiation. Moreover, we conclude that the BLR must be fairly small from light-travel time arguments. Conversely, the narrow lines do not vary on short time scales; the region is spatially extended, geometrically diluting the variable continuum signal over a large volume, and the recombination timescale is very long, further smearing out any temporal variations.

The kinematics of the line-emitting gas remains problematic. In the case of the NLR, the narrow-emission line widths are correlated with the stellar bulge velocity dispersions, albeit with considerable scatter in the correlation. Nevertheless, this suggests that gravitation provides an important component. However, there is also considerable evidence for interaction between the narrow lines and radio-emitting plasma that is being ejected from the nucleus.

How the BLR gas is moving, whether in infall, outflow, or orbital motion, remains unknown. Emission-line profiles alone only weakly constrain the possibilities because there are a wide variety of profoundly different kinematic models that yield similar line profiles. Broad-line profiles are distinctly non-Gaussian, and have sometimes been described as "logarithmic," i.e., the flux at some displacement Δλ from line center is proportional to -ln Δλ,for Δλ not too close to line center. In many cases, however, the line profiles have some structure, variously described as "bumps" or "humps" or, in other cases as "asymmetric wings" or "shelves." Such features can be either prominent or subtle, and they can change over long time scales. Indeed, the persistence of features in profiles over longer than dynamical time scales τdynR / ΔV, where R is the size of the region and ΔV represents a typical velocity, strongly suggests that the BLR has some kind of regular structure or symmetry, though variability studies, as discussed below, show that the gas is not in simple spherical outflow (like a supernovae explosion) or infall (i.e., Bondi accretion).

The average line spectra of AGNs are very similar over a wide range of luminosity, as shown in Fig. 1. This suggests that in addition to temperature, particle densities and ionization parameters are quite similar. An important exception to this statement is the behavior of the C IV λ1549 emission line; relative to the continuum, C IVis weaker in more luminous objects (i.e., its equivalent width decreases with luminosity), a well-known anticorrelation known as the Baldwin Effect. There are other obvious differences among AGNs, and many of the more subtle differences are correlated with one another; these correlated properties stand out prominently in "principal component analysis." The strongest set of correlations, known as "Eigenvector 1," reveals that strong optical Fe II emission is negatively correlated with the width of the Hβ broad component and the strength of the narrow [O III] λ5007 line, e.g., [5]. It has been speculated by many investigators that Eigenvector 1 measures the Eddington ratio dot{M} / dot{M}Edd, i.e., the accretion rate relative to the rate required to produce the Eddington luminosity; in terms of the observables, this is essentially the luminosity to mass ratio.

Figure 1

Figure 1. Composite spectra from the Sloan Digital Sky Survey, binned by luminosity (Mi). Note the similarity of the spectra, with the exception of the "Baldwin Effect" in C IV λ1549; i.e., relative to the continuum and the other lines, C IV is stronger in lower-luminosity objects. From Vanden Berk et al., 2004, in "AGN Physics with the Sloan Digital Sky Survey" ed. G.T. Richards and P.B. Hall (San Francisco: Astronomical Society of the Pacific)

Next Contents Previous