Annu. Rev. Astron. Astrophys. 1997. 35: 445-502 Copyright © 1997 by Annual Reviews. All rights reserved

5.1. Results of Variability: The Stratification of the Broad Line Region in Velocity and Degree of Ionization

OBSERVED STRATIFICATION     Among the HIL, the more highly ionized lines have shorter delays and larger amplitudes. The He II lines at 1640,4686 and N V1240 vary the fastest, then C IV and Ly (Figure 4). Moreover, the wings vary faster than the core (NGC 5548: Krolik et al 1991, Dietrich et al 1993, Kollatschny & Dietrich 1996, Wamsteker et al 1990; NGC 3783: Stirpe et al 1994; NGC 4151: Ulrich et al 1984, Clavel et al 1987, Ulrich & Horne 1996; F9: Recondo-González et al 1997).

Figure 4.Top: The optical spectrum of the nucleus of NGC 3783 on February 8 and March 23, 1992. Bottom: The difference spectrum showing the strong decrease and change of spectral shape of the continuum and the varying component of the main broad emission lines. Note the variations of the He II4686. [From Stirpe et al (1994).] Ordinates in 10-15 erg s-1 cm-2 A-1.

The H; and H lines have, in general, a longer time response and smaller amplitude than the C IV line. In NGC 5548, we see some stratification in the Balmer line emission region as well: The wings vary faster than the core, and the higher Balmer lines vary with larger amplitudes and shorter time delays than the lower ones (Dietrich et al 1993, Wamsteker et al 1990, Kollatschny & Dietrich 1996, van Groningen 1987). Representative values of the time delay and of r_max, the ratio of maximum to minimum emission during a given fluctuation are given in Table 1. The correlation coefficient is usually larger than 0.7.

PHYSICAL INTERPRETATION     The variations of the HIL intensity ratios are generally consistent with the photoionization of ionization-bounded clouds, but there is also evidence for matter-bounded clouds: (a) a progressively weaker response of the C IV line to increases of the continuum flux above a certain level and (b) a flattening or decrease of the ratio C IV/Ly at high-ionizing flux levels (NGC 3516: Ulrich & Boisson 1983; F9: Wamsteker & Colina 1986; NGC 5548: Dietrich & Kollatschny 1995; see also Binette et al 1989, Sparke 1993, Shields et al 1995). The BLR appears to be a mix of optically thin and optically thick gas clouds.

Some large variations of the Balmer decrement are associated with variations of the spectral shape of the optical continuum on time scales of 5-10 years and are entirely consistent with transient, strong, and variable dust extinction, possibly caused by clouds torn from the molecular torus (Goodrich 1989, 1995, Tran 1995, Villar-Martin 1996).

A CAVEAT     The BLR can respond only to continuum variations that last long enough to penetrate its volume significantly, and the amplitude of the continuum variations must also be large enough to alter the gas clouds' emissivity. That is, the BLR filters out continuum variations that are too fast or too small.

For example, in NGC 4151 the continuum variations occurring in ~ 1 day (Figure 2, December 1993) did not result in any detectable variations of the C IV line intensity (Crenshaw et al 1996), although their amplitude, by a factor 1.3, was sufficient to produce line intensity variations in slower conditions.

Because the velocity and line emissivity vary with the radial distance ("stratification"), the line intensity and profile variations differ according to the duration and the amplitude of the continuum event (Netzer & Maoz 1990). Care should be exercised when comparing delays of line responses during different episodes or in different AGN. Only comparisons between events with similar continuum amplitudes are valid.

THE TREND WITH ABSOLUTE LUMINOSITY     Although the data are few, spectroscopic observations of high-luminosity AGN show that time delays of lines with respect to the continuum appear to increase with intrinsic luminosity (Zheng et al 1987, Pérez et al 1989, Gondhalekar 1990, Hooimeyer et al 1992). The recent measurement of the H; time delay in two high-luminosity PG quasars, combined with data for low-luminosity AGN (Table 1), suggests the BLR radius scales as L^-0.5 (Kaspi et al 1996b). This derived scaling is subject to the caveat above, but as the Kaspi et al (1996b) data have amplitudes in the range 1.4-4, it is probably valid.

3C 273 is the highest luminosity AGN (Table 1) for which long-term spectroscopic monitoring has been organized (regularly observed with IUE since 1978, and every other week since 1985). While the continuum flux has varied several times by a factor of 2 on time scales of ~ 2 years, no intensity variations of Ly + N V above 10% have been detected (Ulrich et al 1993 and references therein). Therefore, only a very small fraction of the Ly + N V emitting gas can be within 2 light-years of the continuum source. [The small amplitude Ly + N V variations were judged by Ulrich et al (1993) not to be enough above the measurements errors to produce a reliable value for the time delay. With the same data set, time delays of 74 ± 33 days and 118 ± 57 days were found by Koratkar & Gaskell (1991), O'Brien & Harries (1991)). Whatever the robustness of these values of the delay, they apply to only a minute fraction of the Ly + N V region. The most important result on 3C 273 is that the variations of the Ly + N V line are, at most, of very small amplitude.]

This contrasts with lower luminosity AGN where a continuum flux variation by a factor of 2 always produces a response of the lines of comparable amplitude (e.g. Ulrich et al 1993, Figures 1 and 2). This implies the quasi-absence in 3C 273 of Ly-emitting gas at a distance less than ct from the continuum source, t being the characteristic time scale of the continuum variations, defined here as the time separating two maxima.