Structure and Central Masses in Active Nuclei

AGNs have been observed to vary everywhere from the X-ray to radio regimes. This variability is associated with the continuum (core) sources and the BLR, with NLR variability mild (as would be inferred from its large size). Early work used broad-band photographic data (from archival plates, for example, giving spotty coverage over spans as large as 90 years) or photoelectric observations (accurate but time-consuming). These are sensitive mostly to the continuum (at low redshifts contamination by lines is usually of order 15%). The variations may be episodic (turning on and off, frustrating planning of major campaigns). They are of different kinds - flares, dips, with multiple timescales present. Timescales range from days to decades in the optical. Most QSOs are only slightly variable - 5% or so over a couple of years. Many of these results are summarized in Variability of Active Galactic Nuclei, ed. Miller and Wiita (Cambridge, 1991).

In the continuum, one might look for the spectral shape of the variable component - if nothing else, there should be a nonvariable componant from the surrounding galaxy, and possible from any other large-size continuum components. In Seyferts and QSOs, the bluer wavelengths are generally most variable, and there is little or no variability in the IR. A few violently variable (OVV) QSOs and the Bl Lac objects often show "grey" variations, with no wavelength dependence. Note that these classes show polarization variations; intensity, polarization, and its direction may all vary independently. Some objects (especially in the radio) show components marching redward after a flare, as predicted by synchrotron aging theory. Much interpretation has been thwarted by the bewildering variety of forms it takes - there are no general patterns that could provide detailed physical clues beyond the light-time arguments on the general size of the emitting region.

More useful is reverberation mapping, in which the response of emission lines to continuum variations is used to study the radial distributions of line emissivity and radial velocity. Recombination is quite fast at these densities compared to the light-travel times, so that we can think of a volume of BLR gas as responding instantaneously to a change in ionizing intensity. In a first approximation, the delay time between our observation of a continuum pulse (assuming we measure a continuum that comes from the same place as the ionizing continuum) and the pulse of reprocessed line emission gives one measure of the radius of the BLR. In more detail, one determines a transfer function Y such that

where L is the line intensity and C is the continuum intensity. Various sizes and shapes of BLR give various forms for Y. It is straightforward to show from geometrical definitions that the surface probed by any particular t-t is a paraboloid (to be pedantic, an ellipsoid) with the continuum source at the nucleus; this is another ingredient of Y. The same mathematics applies to study of the light echoes from novae and supernovae (Couderc 1939 Ann. d'astrophys. 2, 271). In practice, one may try to determine a single moment of Y, say via cross-correlation analysis, or do a full-blown and very noisy measurement of Y for each emission line and look for ionization structure within the BLR. The first general formulation for this was presented by Blandford and McKee 1982 (ApJ 255, 419); more recently a maximum-entropy solution has been described by Horne and Welsh (in Miller and Wiita p. 332).

The cross-correlation technique is simple and in principle robust, but difficult to apply in the usual case of uneven temporal sampling, compounded by gaps due to solar and lunar interference (to make it worse, these are periodic) plus weather. One generally has to interpolate in time (for one or both members of the correlation; Gaskell and Sparke 1986 ApJ 305, 175; Koratkar and Gaskell 1991 ApJSuppl 75, 719) or in time delay (the discrete correlation function of Edelson and Krolik 1988 (ApJ 333, 646). These methods average over time in different ways; the second suffers when the source sometimes varies and sometimes not. Note also the problem of statistical dependence of simultaneous measurement of line and continuum flux from the same spectrum - any flux-calibration error will introduce a spurious peak at DT =0 in the correlation. These techniques may be applied to the wings of emission lines as well, which could in principle specify the dominant kind of motion in the BLR (inward, outward, circular) and lead to a mass measurement for the central object if the motions are ballistic rather than related to some kind of hydrodyanamic driving.

The first systematic spectroscopic studies were done on NGC 4151, in the optical (Antonucci and Cohen 1983 ApJ 271, 564) and UV (Ulrich et al. 1984 MNRAS 206, 221 and 209, 474). From these, it was found that

  • the near-UV bump arises where Hb and Hg do.
  • the BLR is a light-week or less in radius.
  • more emitting gas is present out to a light-year.
  • the Ha profile is constant over several crossing times.
  • the C IV region is much smaller than the Mg II region.
  • variation is in the order "thermal" continuum, power-law, then lines.
  • there may be a slight outflow excess.

    More extensive IUE monitoring has extended these conclusions (see Clavel in Miller and Wiita p. 301). Much of the apparent size of the Mg II zone may appear there as a dilution of short-time variability; some Mg II comes from the C IV region.

    The state of the art in variability programs was represented by the campaign to monitor NGC 5548, including 60 regularly-spaced sets of IUE spectra and intensive ground-based observations (Clavel et 56 al. 1991 ApJ 366, 64; Peterson et al. 1991 ApJ 368, 119). The main advances here were in experiment design - regular spacing of data to circumvent the need for interpolation in analysis, and space-based data to decrease the chance of a break in the series. The galaxy also cooperated, showing several strong peaks and dips in brightness over this 8-month period (December 1988 - August 1989). IUE results are shown below for various continum bands and the strong lines, with data from the International AGN Watch.

    Detailed analysis suggests that different lines come from regions with different shapes, with higher-ionization features in an inner, nearly spherical configuration, surrounded by a flattened region from which we observe C III] and Mg II. The limits here are set by how helpful the nucleus was in its variability pattern, since quasiperiodic variations can mimic the effects of discrete structures. Followon studies have combined IUE and HST data with X-ray monitoring, trying to sample wide ranges of of luminosity and BLR radius. The NGC 5548 HST campaign yielded time lags (i.e. peak in the response function Y slightly under 2 days for the strong UV lines, giving a radial scale of the BLR of 2 light-days. NGC 3783 in another IUE campaign gave a range of delays from 0-5 days for assorted emission lines, haviong shown a more complex variability pattern than the observing strategy could reconstruct. The high-luminosity Seyfert Fairall 9 shows lags 14-20 days, consistent with a broad theoretical suggestion that the BLR should scale roughly as L². NGC 7469 gave 2-3 days for emission-line delays. A campaign on the BLRG 3C 390.3 gave 35-70 day lags for the UV lines, larger than for other objects (Seyfert nuclei) of similar UV luminosity; it may be relevant that the BLR profile has a prominent and higholy variable double component in this object. In none of these objects is there a strong signature of either inflow or outflow in the BLR, which would make either the red or blue line wings vary first. References for this series are given in the AGN Watch bibliography.

    It is nontrivial that all the UV-optical continuum regimes (including the EUV ionizing continuum) vary closely in step. In a simple accretion-disk model, the outer, cooler parts of the disk take longer to respond to disturbances in the inner parts (on a viscosity timescale). It now appears that transmission must be by photons, perhaps scattered by a hot corona inferred from X-ray line measurements. Usually, the deep-UV continuum varies first, with the X-rays and longer wavelengths following in hours to days.

    The most important result has been that the BLR is smaller than simple ionization-parameter arguments suggested, mostly since the densities are higher and not well-determined spectroscopically. This may mean that cloud models must be rethought.

    X-ray variability is known in both intensity and spectral shape. There is evidence of changes in the soft X-ray cutoff produced by H photoelectric absorption (Barr et al 1977 MNRAS 181, 438), probably as clouds drift across the line of sight to the tiny central X-ray source; the variations observed would be Poisson fluctuations. There would be typically 100 clouds in front of the nucleus, with column densities typically 5 × 1022 cm-2 and velocities 1000 km/s. This may be direct detection of BLR clouds.

    Strong, rapid, and continual X-ray variability has proven to be a hallmark of a subset of AGN with the slightly oxymoronic designation of narrow-line Seyfert 1 galaxies (NLSy1). Their optical spectra show all the usual high-density lines, but with widths of 1000 km/s or so instead of the 5000 which is more usual (I Zw 1 is a prime example). This may indicate that these objects have simultaneously lower central masses and higher accretion rates relative to the Eddington luminosity than more usual Seyferts (the combination being likely from observational selection effects), as noted by Leighly 1999 (ApJS 125, 297). NLSy1 thus pose an interesting puzzle in that they seem to be trying to remedy their previous violation of the usual bulge/black hole mass relation.

    Rapid X-ray variability has been seen on timescales as small as 100 seconds (where they disappear into the noise) in numerous AGN. No characteristic timescale has been found, with much of the variability appearing as 1/f "noise"; the variations are strongest at higher energies (maybe a surrounding softer disk component dilutes the variable part?). This behavior has been reported for Mkn 421, Mkn 1310, MCG -6-30-15, NGC 5506, Mkn 335, NGC 6814 - about every bright AGN observed.

    This implies that the hard X-rays come from a VERY small region; a black hole of 108 solar masses has Schwarzschild radius of 980 light-seconds. Beaming of relativistically-moving blobs might relax this problem somewhat. This also makes it easier to understand how some AGN (especially BL Lac objects) can generate enormously powerful gamma-ray flares.

    Still unknown: where does the radio continuum come from? Just what is the role of jets? They might be an important mode of energy transport from the core to the surroundings, rather than radiation being the whole story.

    Evidence for accretion disks

    Since most AGN models agree that the action is controlled by some kind of accretion disk around a massive black hole, it would be highly desirable to see some direct evidence of such a disk. This has so far been equivocal (at best). There are several different conceptual matters here - do such disks exist, do are they important in channeling the energetics of the nuclei, and can we see the disks directly in some way? Direct evidence of accretion disks in AGN is in short supply. Some of the approaches that have been reported include a quasi-thermal excess in the UV, polarized emission scattered off a disklike structure, two-peaked disk-like emission line profiles, and a search for the expected Lyman limit absorption due to a disk sufficient to produce the UV bump. The results indicate either disks on the wrong scale, inconclusive evidence, or no such signature as predicted.

    The UV excess (or big blue bump, to distinguish it from the little blue bump mostly due to Fe II and blended Balmer lines) appears as an excess above the broader-band power-law continuum at frequencies approaching the Lyman limit, and perhaps as a soft X-ray excess on the other side of the hydrogen absorption curtain. It is especially prominent in an energy per wavelength n Fn plot, and may be responsible for the excess ionizing radiation indicated by Zanstra arguments. The characteristic temperature of the emission peak would be of order 105 K. An extensive analysis was given by Malkan and Sargent (1982 ApJ 254,22) and Malkan 1983 (ApJ 268, 582). The bump may be fit with a fat blackbody (or multiple blackbodies) and has been considered as direct evidence for radiation from the accretion disk. This kind of spectral decomposition is illustrated by Fig. 2 of Malkan 1983 (courtesy of the AAS):

    These results are rather sensitive to model details; Sun and Malkan have more recently (1989 ApJ 346, 68) examined the effect of relativistic light propagation and disk inclination on the observed spectrum.

    The Lyman-limit test described by Antonucci et al. 1989 (ApJ 342, 64) throws a disk interpretation of the UV bump into doubt. For any reasonable disk structure, stellar atmosphere theory may be applied and predicts a substantial drop in flux at the Lyman edge - these are basically funny shaped supergiant stars. Even smeared out by mildly relativistic Doppler motions, such an edge should be detectable - and it isn't in high-redshift QSOs.

    There have been claims that certain multicomponent broad-line profiles are well fit by a central (NLR) peak plus the characteristic velocity signature of a rotating disk, complete with Doppler boosting of the blue component. The type example is Arp 102B, as shown by Halpern and Filippenko 1988 (Nature 331, 46). A fit to a detailed model was given by Chen et al. 1989 (ApJ 339, 742) in their Fig. 2 (ADS, by permission of the AAS):

    There are several objects with profiles of this general kind: also included are 3C 309.3 and 3C 382 (Osterbrock et al 1976 ApJ 206, 898), and 3C 332 (Halpern 1990 ApJLett 354, L1). Even the LINER NGC 1097 and the LINER and low-level Seyfert 1 nucleus in M81 show this behavior at least occasionally, especially in the UV lines. However, a disk interpretation for these profiles has both immediate and philosophical problems. There exists a large variety of BLR profiles, so that a few of the thousands of objects known will fit almost any theory. Why should only a few oddball objects show this signature of what should be a general phenomenon? More immediately, the specific properties that are regarded as fitting the model (such as higher "Doppler-boosted" blue peak) are not constant over episodes of variability, and the two peaks may represent independent kinematic entities (Miller and Peterson 1990 ApJ 361, 98). Sulentic et al (1990 APJLett 355, L15) show that double-horned profiles do not show the expected systematic redshift relative to indicators of systemic velocity that would be expected for relativistic accretion disks. As a judge in Scotland might say, the verdict is "not proven".

    Some kind of disk structure, though far too large to be the much-sought active accretion disk (that is, the one which is hot enough to be radiuating energy and shedding angular momentum), has been detected in many cases by spectropolarimetry. The classic case is NGC 1068, in which Antonucci and Miller (1985 ApJ 298, 935) found that a Seyfert 2 nucleus becomes a Seufert 1 in polarized intensity. Something is scattering light from a nucleus that we don't see. Fragmentary statistics suggest that the disks are geometrically thin in Type 1 Seyferts and thick for type 2s, as expected for the direct visibility or invisibility of the BLR (Antonucci 1983 Nature 303, 158; 1984, ApJ 276, 499). The argument rests on polarization parallel to the smallest-scale radio structure in Sy 1 and most radio galaxies, and perpendicular for Sy 2. This disk must be much larger than the BLR, of scale perhaps a parsec, and is thus not the accretion disk around the central object. Its existence is the key to so-called unified schemes for joining broad-line and narrow-line objects into a single physical class.

    Many active nuclei show dust or emission-line disks at high resolution, on a scale of parsecs. These are much too large to be the radiating disks discussed earlier, but they are systematically aligned perpendicular to radio jets, and may provide evidence of the expected angular momentum as well as mass transport. The best example is NGC 4261 (see Ferrarese et al. 1996 ApJ 470, 444). These turn out to be reasonably common in nearby radio galaxies, and even borderline Seyferts like M51. The disk in NGC 4261 is shown below, from the HST PC image and with my attempted 3D reconstruction.


    Unified AGN Models

    The variety of classifications of AGN has led to several efforts to bring order to the field by asking whether any of these distinctions are only artifacts of how we happen to view particular objects. Two such schemes have been strongly supported by observations - the unified scheme for QSOs, radio galaxies, and BL Lac objects, and an orientation pictures for Seyfert nuclei of types 1 and 2.Both kinds of unified models are discussed at length by Antonucci (1993, ARA&A 31, 473).

    In the "unified scheme", broadly traceable to Blandford & Königl (1979 ApJ 232, 34), beaming and obscuration make QSOs look different from different directions. If there is an equatorial torus which is optically thick to UV and visible radiation, we won't see the central source and BLR if it's in the way, while, since so many radio-loud objects have prominent jets, Doppler beaming will make them much brighter when seen near the jet axis. This would reduce the number of different kinds of AGN by 2 or 3 (since then NLRGs, radio-loud QSOs, BL Lac objects, and optically-violently-variable or OVV QSOs are all the same thing - and maybe NLRGs too). If the orientation differences are due solely to Doppler favoritism, there is a connection between the cone angle for beaming and its amplitude, which translates to a relation among luminosity functions for sideways, beamed, and highly beamed populations. The scheme predicts that radio galaxies will have larger projected sizes (for the double radio sources) than QSOs, which will in turn look larger than BL Lac objects, all of which are apparently true (Barthel 1989 ApJ 336, 606), and that the double-lobe luminosities should be comparable for all, since the lobes aren't expanding relativistically and therefore are nearly isotropic radiators. One also expects the incidence of apparent superlumninal motion, an indicator of relativistic bulk motion close to the line of sight, to be a strong function of orientation angle. Possibly related is the Laing effect (Nature 331, 149, 1988), in which Faraday depolarization shows that the sides of double radio sources with brighter jets are indeed the ones closer to us (indicating Doppler boosting on kiloparsec scales). The observational status of this issue seems to be "Yes, but..." At this point, it is certainly tenable that this unified scheme holds for many of these objects (recognizing that variation among individual objects will introduce additional effects on statistical properties). Lots of details still beckon - is Doppler boosting the whole story, and at what g over what region? Are there further influences on the covering factor of any obscuring matter, such as is often seen in radio galaxies?

    For Seyfert galaxies, much evidence indicates that most nuclei are surrounded by dusty, massive tori, optically thick to the nuclear radiation until one gets to the far-infrared and much larger than the BLR. Pole-on configurations are seen as type 1, edge-on as type 2. Strong support comes from spectropolarimetry of many Sy 2 nuclei, whch show in polarized (scattered) light the broad lines and strong Fe II emission of type 1 objects; this must be present in integrated intensity as well but is swamped by the direct components. Also, the NLR configuration is often quite elongated, typically with the highest-ionization material in a biconical configuration centered at the nucleus. This is interpreted as the illumination pattern of radiation escaping the torus.

    The tests for both hypotheses amount to either comparing isotropic quantities for two allegedly different classes, usually using some such quantity to scale for luminosity, or comparing isotropic and anisotropic quantities versus some proxy for orientation angle. For radio-loud objects, the core dominance R is often used (defined as the core-to-lobes flux ratio). By now, perhaps the biggest puzzle is what distinguished radio-loud and radio-quiet objects.

    The lack of strong evidence for luminous active accretion disks has strengthened the case for advection-dominated accretion flows (ADAFs), which basically means accretion at such low rates that a steady disk doesn't form. There is some hope that the SIM interferometer can resolve the scales of the active disk in such objects as M87 in which we clearly expect to see one.

    Evidence for central supermassive objects

    More or less direct mass determinations of the central objects in AGN have been used to support the notion of accretion onto (super)massive black holes. The techniques have included

  • Estimates of the BLR radius combined with its velocity extent. This can be taken as a virial-like estimator of the total mass. The velocity scale is straightforward, once you decide which moment of the profile represents a characteristics velocity at what radius. The distance from the center can be deduced from ionization-parameter arguments (which turns out to have very poor error behavior at high densities) or from variability studies. A famous example was the NGC 4151 paper from a long IUE monitoring campaign, by Ulrich et al. 1984 (MNRAS 206, 221), most notable for a typo in the value of G they used. An erratum (MNRAS 209, 479) includes, "It has been drawn to our attention by Dr. N.Y. Chen of Sulzer Brothers Limited, Winterthur, Switzerland, that we made an arithmetical error of a factor 10 in our estimate of the mass of the central object... There is no longer such an interesting coincidence..." Always check your numbers even if you're really, really sure of them. Anyway, the notion should still hold, although there are real uncertainties in how one should connect parts of a featureless (more or less logarithmic) line profile to specific radii.
  • Dynamics of surrounding gas and stars. Examples for gas, easy to trace but not unambiguous as a gravitational probe, include:
  • The maser ring in NGC 4258, which is at high confidence in Keplerian motion at distances 0.14-0.28 pc from a central mass of 4 × 107 solar masses (Herrnstein et al. 1999 Nature 400, 539). The "Keplerian" part is important in showing that the mass is concentrated well inside the orbits; that much mass within less than 0.001 cubic parsec is very difficult to hide in any other form. In this case, the expected accelerations for masers in circular orbits are in fact seen, giving confirmation of the general picture.
  • Similar (but sparser) features in NGC 1068, interpreted by Gallimore et al. (1996 ApJ 462, 740) as in Keplerian motion 1-2 pc from a similar mass of 4.4 × 107 solar masses.
  • The gaseous disk of M87. For stars, M87 has long been a strong candidate, and several other radio galaxies have been added (I hasten to add, so have an equiva;ent number of non-active galaxies) to the list with stellar dynamics indicating massive central objects. It turns out to be cooperative enough to have an emission-line disk perpendicular to the jet axis, too small to be easily distinguished from the surrounding complex gas structure using ground-based data. HST spectroscopy analyzed by Harms et al. (1994 ApJL 435, L35) and Macchetto et al. (1997 ApJ 489, 579) showed that its kinematics are well described by a Keplerian disk surrounding a mass 2.5-3.2 × 109 solar masses, a value which harks back to the Young et al. velocity-dispersion study in 1978.

  • Stellar kinematics are hard to derive for Seyfert nuclei, much less quasars, because the starlight is swamped by the nonstellar light from the nucleus. It's easier for radio galaxies (like, for example, M87). On the theory that the compact radio source Sagittarius A* is some kind of weak AGN, studies of kinematics near the Milky Way's nuclear become relevant. A real breathrough has been the newly acquired ability to measure proper motions of stars within the central parsec over a few years. Ghez et al. (2000 Nature 407, 349) show arcs for these motions which suggest a central mass of 2.6 × 106 solar masses. The great thing about proper-motion studies is that the results only improve with time. Current data show substantial parts of stellar orbits here, making it clear thet they are Keplerian about a mass smaller than about 10 AU in radius, and making this by far the best-attested massive black hole. (Link to animated GIF of stellar orbits)

  • Gravitational redshift of X-ray spectral features. Over the years there have been limits or possible detections of gravittaional redshift from the broad-line region. If one has a secure Doppler plus cosmological redshift for the whole system (from surrounding stars or gas at large distances), one could search for excess redshift from material close to the central object. This effect has been measured on Earth (the Pound-Rebka experiment, and these days every use of a GPS receiver), in the solar spectrum, and for the surfaces of a couple of white dwarfs in binary systems (needed to get the purely Doppler component). See, for example, section 38 of Gravitation by Misner, Thorne, and Wheeler for details. In the weak-field case (appropriate for gas at significant distances from the core if not the central object itself), the gravitational redshift becomes z = GM/Rc². Multiple sources of redshift combine by multiplication of the relevant (1+z) values.

    Iwasawa et al. (1996 MNRAS 282, 1038) and Lee et al. (1999 MNRAS 310, 973) report that the Fe K line in MCG -6-30-15 shows a significant excess redshift compared to other features produced at lower temperatures, and that this component remains and is most prominent when the overall X-ray emission is weak. They suggest that this is gravitational redshift from the extremely hot and dense material, extending inwards to something like 7 Schwarzschild radii, with general accretion-rate arguments suggesting a central mass near 107 solar masses and thus the X-ray emitting material extends inward to something like 1 AU. This has the interesting consequence that the X-rays vary more rapidly than this light-crossing time so they would have to come from a region even smaller than this inner part of a presumed accretion disk.


    « Emission-line spectra | Accretion pictures, BAL QSOs and outflow »
    Course Home | Bill Keel's Home Page | Image Usage and Copyright Info | UA Astronomy

    keel@bildad.astr.ua.edu
    Last changes: 10/2006	  © 2000-6