Relativistic Jets in Active Galactic Nuclei

Relativistic Jets in Active Galactic Nuclei Abstract. This review discusses the basic characteristics of the jets in active galactic nuclei (AGN). It concentrates on the highly relativistic variety found in blazars, since these are the closest cousins to the jets seen in some X-ray binaries (microquasars) and the collimated, ultra-fast flows inferred to exist in gamma ray bursts. The fastest AGN jets are gamma-ray bright, with flow Lorentz factors that range up to at least 45, although such high values must be rare unless the radiative efficiency is usually very low. If jets are magnetically launched, observations indicate that the acceleration and collimation occur over an extended region and the Poynting flux is converted into both flow energy and turbulence. Particle domination of the pressure occurs during outbursts, as might be possible in shocks. The cores of AGN jets may have different origins depending on frequency: points where the Doppler factor becomes high, conical recollimation shocks, or the optically thick/thin transition region. Superluminal knots and outbursts are well explained by shocks, but the polarization properties require than in many cases the shock fronts are oblique to the jet axis. In two radio galaxies, appearances of superluminal knots in the jet follow low-hard X-ray states, reminiscent of the behavior of microquasars. Models that relate the accretion state to the geometry of the magnetic field must explain how the transition back to a radiative inner disk releases an ultra-fast moving energetic disturbance into the jet. INTRODUCTION Ultra-relativistic jets are features that the more extreme AGN and X-ray binary systems (XRBs or “microquasars”) and perhaps all γ -ray bursts (GRBs) have in common. This gives us some hope that by studying, comparing, and contrasting the different properties of these classes, we can gain powerful insights into the processes that govern the creation, physics, and behavior of relativistic jets. In an attempt to facilitate this endeavor, this short, partial review concentrates on the aspects of AGN jets that appear to relate most closely to those in the other types of jet-producing systems. One of the main advantages of studying AGN jets is that the temporal and angular scales match fairly well those that are observationally available. Time scales of variability of flux and structure are in the minutes to years range, and the angular scales of the radio emitting sections are those that can be explored with very long baseline interferometry (VLBI). You do need to be somewhat patient, since cycles of activity tend to take several years, but it usually requires no longer than a year or two for a given AGN to do something interesting. As a result, we have accumulated vast vats of data exhibiting a wide range of variability behavior. Here I will try to distill the contents of the vats in an attempt to draw out the vital essence of the phenomenon. Figure 1 sketches slightly more than half of an AGN; the other half is presumed 1 FIGURE 1. Sketch of the various sites of emission in an active galactic nucleus with a relativistic jet. The density of dots signifies in a qualitative way the intensity of the emission. The radiation produced in the jet is relativistically beamed, while the emission from outside the jet is not. It is not clear whether the emission from the ambient jet between the black hole (small black circle near the base of the jet) and the core is visible. The length of the arrows indicates the Lorentz factor of the flow. Note the logarithmic scale of approximate distance from the black hole, measured in Schwarzschild radii. (Adapted from Marscher, 2005) to be essentially identical. According to our standard paradigm, the central engine consists of an accreting supermassive black hole (BH) weighing in at millions to several billions of solar masses. We do not know the geometry of the accretion flow within a few tens of gravitational radii, i.e., whether it is a thin disk that continues down to the innermost stable circular orbit or if the flow becomes closer to spherical near the BH, perhaps advecting energy along with the mass as opposed to radiating it as the material plunges toward the BH. The accretion disk emits thermal ultraviolet, optical, and infrared radiation. There must also be some sort of corona containing hot electrons that Compton scatter some of these photons to X-ray energies. Unlike an XRB, the accretion disk is not hot enough (at least according to the standard models; see Malkan & Sargent, 1982). to emit X-rays. Instead, the central engine’s X-rays are thought to come from Compton upscattering in the corona, which could be the atmosphere of the disk or the base of either a wind or jet emanating from the disk. The current picture of jet formation, as summarized at the conference by J. McKinney, invokes differential rotation of the poloidal component of the magnetic field either in the inner disk or, for a rotating BH, in the ergosphere. Theorists have had difficulty getting a jet propelled from the disk to reach ultrarelativistic flow speeds, whereas jets launched from the ergosphere can accelerate to high Lorentz factors by starting with a high initial energy density in Poynting flux relative to rest mass owing to a low density of matter 2 (see, e.g., Meier, Koide, & Uchida, 2000; McKinney, 2006). The jet propagates from the vicinity of the black hole out to parsec, kiloparsec and, in some objects, megaparsec distances. Here I concentrate on the “inner jet” on scales of parsecs or less, since this region lends itself to comparison with XRBs and GRBs. The inner jet is accessible to VLBI and its emission is highly time-variable. Hence, the combination of VLBI observations and monitoring of flux and polarization provides the means to explore the structure and behavior of AGN jets. OBSERVED PROPERTIES OF THE INNER JETS OF BLAZARS Besides being the cousins of XRB and GRB jets, blazar jets have the advantage that they tend to dominate the emission across essentially the entire electromagnetic spectrum. That is, the emission diluting that coming from the jet is relatively unimportant. The disadvantage is that all the emission is affected by relativistic beaming, the effects of which we need to quantify in order to examine the underlying physical processes. Structure of an AGN Jet The classical structure of a compact jet in a radio-loud AGN consists of an unresolved or barely resolved “core” from which a jet protrudes. Figures 2-5 present some examples of images obtained with the Very Long Baseline Array (VLBA). The jets can appear long or short and stubby, nearly straight or sharply curved, and relatively smooth or dominated by knots. The majority of knots move at apparent superluminal velocities, but some are stationary or move at subluminal speeds (see Jorstad et al., 2001; Kellermann et al., 2004; Jorstad et al., 2005, for many examples.) We know that the jets need to be pointing within several degrees of the line of sight to display high superluminal apparent speeds. Projection effects then foreshorten the jet so that it appears fatter, shorter, and more sharply bent (if it has a modest intrinsic degree of curvature) than it would if observed side-on. Images of most Fanaroff-Riley class II (FR II, high radio luminosity) radio galaxies—thought to be the parent population of quasars viewed at large angles to the jet axis—feature straight jets, so it seems reasonable to assume that the jets of blazars are actually bent by less than a few degrees in three dimensions. Speeds of Jet Flows How fast can a jet flow? Fortunately, in the case of blazars, we can measure apparent superluminal motion, which we are convinced is an illusion caused by relativistic motion directed almost right along the line of sight. The apparent speed in units of c, βapp , cannot exceed the actual Lorentz factor Γ of the pattern we are observing. If the pattern is a shock, it can move at a higher Lorentz factor than does the energized plasma after the shock front overtakes it. But for the highest velocities, the difference is modest. Even if the plasma has an ultrarelativistic equation of state, the velocity of the plasma relative to 3 FIGURE 2. Sample VLBA images of radio-loud AGN with straight jets. Left: the quasar 3C 279 in January 2002 (Marscher et al., in preparation); right: the radio galaxy 3C 111 in December 1999 (data published by Jorstad et al., 2005) FIGURE 3. Examples of VLBA images showing short, broad jets. Left: the quasar CTA 26 in July 2005 (Marscher et al., in preparation); right: the quasar 3C 454.3 in February 1999 (data published by Jorstad et al., 2005). Here, and in subsequent images, linear polarization is indicated by sticks whose length is proportional to the polarized intensity and direction is parallel to the electric vector. Note that the jets do not actually end where the emission is no longer seen; rather, they propagate out to kiloparsec scales. the shock front is c/3, so that the Lorentz factor of the plasma in the observer’s frame is Γplasma = 1.06Γshock [(1 − βshock /3)]. The speed of the centroid of the emission region— the most easily measured representation of the apparent velocity—is roughly the mean of the shock and plasma velocities. For example, if Γshock = 20 then Γplasma ≈ 14. (Γplasma is closer to Γshock if the protons are not highly relativistic so that the sound speed is 4 FIGURE 4. Examples of VLBA images displaying bent jets. Left: the quasar PKS 0528+134 in December 1999; right: BL Lac in July 2000 (data for both published by Jorstad et al., 2005). The actual (in 3-D) bends are probably very gentle. √ significantly less than c/ 3). The centroid of the knot moves at a Lorentz factor between these two values. So, at least for the high-Γ jets, the Lorentz factor of the centroid of a knot is only a slight overestimation of the bulk Lorentz factor of the emitting plasma. The fastest AGN jets are found in γ -ray bright blazars (Jorstad et al., 2001; Kellermann et al., 2004; Piner, 2006). In a VLBA survey of about 60% of the EGRET sources identified as blazars, Jorstad et al. (2001) reported apparent velocities as high as ∼ 45c, and follow-up measurements by Jorstad et al. (2005) (discussed by S. Jorstad at the conference) yielded 46c for one of the objects (PKS 1510−089). Figure 6 provides a histogram of the apparent speeds of γ -ray blazars measured thus far. Jorstad et al. (2005) have determined Doppler beaming factors by comparing timescales of flux declines with light-crossing times of resolved knots, with the highest value exceeding 60. These numbers are comparable to the most conservative estimates for GRBs. But a typical radio-loud AGN is not so fast. High-Γ blazars overpopulate any centimeter-wave survey of bright, flat-spectrum sources because of beaming bias. A volume-limited sample of radio-loud AGN would be dominated by objects with more mundane jets. In a population synthesis study, Lister & Marscher (1997) used realistic luminosity functions to show that the majority of compact extragalactic radio sources have jets with low Lorentz factors. If the Lorentz factor distribution of the population is a power law, N(Γ) ∝ Γ−a , a value of a between 1.5 and 1.75 provides a good match between the synthetic and observed distribution of apparent velocities. If the highest Lorentz factor of the population is 45, then the allowed range for a means that between 83 and 88% of compact jets have Lorentz factors lower than 10. 5 FIGURE 5. Example of a jet with both stationary and superluminally moving knots, in the BL Lac object OJ 287. The bright knot about 0.3 mas from the core (which is on the eastern end) and the one that lies about 1 mas to the west-southwest do not propagate downstream, while a superluminally moving knot can be seen between the two during the last three epochs. (From data published by Jorstad et al., 2005) Jorstad et al. (2005) (and S. Jorstad’s talk at the conference) have shown that the trajectories of individual superluminal knots tend to be curved and that their apparent speeds often change, usually increasing with time. The apparent acceleration might only be in direction, as expected if the jet in extreme blazars points closer to the line of sight than an angle θ = Γ−1 . In this case, nearly all directions of bending will be away from the line of sight and closer to the angle that maximizes apparent motion. Opening Angles The opening angles of the jets in FR 2 radio galaxies viewed more or less side-on are of order a few degrees. One might then expect the fastest blazar jets, which must be pointing to within a few degrees of the line of sight, to be extremely broad. Yet some of the very fast jets (e.g., 3C 279, in which knots appeared to move at more than 20c in 2002-2003; Jorstad & Marscher, 2005) usually has quite a narrow jet (see Fig. 2). This 6 FIGURE 6. Apparent speeds of 41 γ -ray bright blazars. For each object with measured proper motion, the fastest well-established value is adopted since that should represent a lower limit to the Lorentz factor of the jet flow given the generally unknown angle of the jet axis to the line of sight. (From data reported by Jorstad et al., 2005; Kellermann et al., 2004; Piner, 2006; Gabuzda & Cawthorne, 1996, and at the MOJAVE website maintained by M. Lister, URL http://www.physics.purdue.edu/astro/MOJAVE/.) puzzle was solved by Jorstad et al. (2005), who found that the intrinsic opening angle is inversely proportional to the Lorentz factor of the flow (see Fig. 7). A likely physical cause of such a relationship is readily found in models for extended acceleration of the jet flow, either by gas dynamics (Blandford & Rees, 1974; Daly & Marscher, 1988) or gradients in magnetic field tied to the central engine (Vlahakis & Königl, 2004). In the former case, the opening angle is inversely proportional to the Mach number as it adjusts to balance its pressure with that of the external medium, and the Mach number is proportional to Γ. In the case of magnetic acceleration, the jet becomes more cylindrical while its flow velocity increases with distance down the jet, as the hoop stress of the toroidal component of the field pinches it toward the axis. Changes in Direction of the Jet with Time If the jet is really anchored by a supermassive black hole, then we should not expect it to change direction as a function of time within a few parsecs of the central engine 7 FIGURE 7. Opening angles of blazar jets as a function of Lorentz factor for 8 radio-loud quasars, 5 BL Lac objects, and 2 radio galaxies with blazar-like radio emission. (adapted from Jorstad et al., 2005). The line corresponds to a dependence θ1/2 ∝ Γ−1 . over intervals of only a few years. Yet we do in a number of blazars, as is illustrated in Figure 8. Projection effects amplify the magnitude of the variations: the observed change in direction is of order 10◦ , which translates to less than ∼ 1◦ when deprojected. Nevertheless, this involves a substantial change in momentum given the relativistic flow velocities. The timescale is set by the movement of some section of the jet, perhaps the base. It is therefore unaffected by Doppler blueshifting, since the location of the change in direction presumably does not physically move relativistically. One possible cause of swings in the jet direction is precession, which would imply that two black holes orbit each other at the center of the system. If this is the case, then the jet direction should oscillate smoothly. Sterling et al. (2003) have claimed to observe the expected periodic changes in the direction of the jet as well as in the position angle of the polarization near the core of BL Lacertae, although this conclusion has been called into question by Mutel & Denn (2005). Jorstad et al. (2005), on the other hand, have found that the jets of several other blazars change direction more erratically than would be consistent with precession. The implication is that either the axis of the jet launching mechanism wobbles or an instability between the base of the jet and the core causes a jitter in the direction of the 8 FIGURE 8. Example of a jet whose direction of ejection of knots from the core changes with time: the BL Lac object OJ287. Shown are two images separated by 3 years; the change in direction is striking. (Data published by Jorstad et al., 2001, 2005) jet flow (see the paper by P. Hardee in this volume for a discussion of instabilities in relativistic jets). Another possibility is that the jet is an irregular spray that usually emits only one stream at a time. A variation of this idea is that only some of the flow lines in the jet emit radiation at any given time, presumably because particle acceleration is sporadic. If the stream that we can see changes, it will appear as a sudden shift in jet direction. Energy and Matter Content There are basically two methods for determining the energy contained in relativistic particles in extragalactic jets. The first is to assume equipartition between the magnetic and particle energy densities, which is similar to calculating the minimum energy required to produce the synchrotron emission (Burbidge, 1959). Readhead (1994) found that semi-compact jets (between parsec and kiloparsec scales, observed at frequencies less than 1 GHz) have brightness temperatures that are very close to those corresponding to equipartition. He suggested that perhaps parsec-scale jets are in equipartition as well. The second method is to analyze compact knots that are synchrotron self-absorbed, in which case one can determine the energy density in radiating electrons and magnetic field separately (see, e.g., Marscher, 1983). Readhead (1994) showed that the ratio of the relativistic electron to magnetic energy density ure /umag ∝ Tm0 8.5 Fm−0.5 , where Tm0 is the rest-frame brightness temperature and Fm is the flux density at the self-absorption 9 turnover frequency νm . (There is also a weak dependence on the upper and lower cutoff frequencies of the synchrotron radiation that is generally ignored.) Use of this method requires accurate measurements of the angular size, νm , and Fm of the core or individual knots in jets, a difficult undertaking requiring multifrequency VLBI. Only a relatively small number of well-observed knots have the requisite properties to allow reliable estimates of the physical parameters in this way. Also, sufficiently accurate values of the Doppler factor to make the necessary relativistic corrections are difficult to obtain in individual objects. At sufficiently low frequencies—perhaps below 20 GHz—what is seen as the core on VLBI images should be the section of the jet where the optical depth is roughly unity at the frequency of observation. In that case, the measured brightness temperature will indicate the ratio of relativistic electron to magnetic energy density, following the formula above. ? have analyzed the VLBA visibility data of a large sample of sources at 15 GHz with measured apparent speeds (used to estimate the Doppler factors needed to transform observed brightness temperatures to instrinsic ones) to determine the brightness temperatures of many of the cores. These cores are partially resolved both in the direction along and transverse to the jet axis. Homan et al. (2006) have analyzed this dataset, finding that the cores are typically near equipartition (Tm0 ∼ 3 × 1010 K), but during outbursts the relativistic electrons dominate over the magnetic energy density. We must keep in mind that the “core” at 15 GHz is not the base of the jet, which would only be revealed by VLBI observations at frequencies above the ultimate turnover at short millimeter or even submillimeter wavelengths (see Impey & Neugebauer, 1988). It is therefore still possible that the jet is magnetically dominated within the main acceleration and collimation zone. This presents a challenge to explain how it could become particle dominated a relatively short distance downstream. Perhaps velocity shear generates turbulence that eventually brings the initially magnetically dominated plasma into rough energy equipartition with the particles. Shocks, with their strong particle acceleration, could then convert kinetic energy of the flow into particle energy and create an outburst of flux in the process. It is unclear whether the particles in jets are “normal matter”—electrons and protons—pair plasmas, or a combination of the two. This is a crucial issue, since the composition of a jet is intimately related to the physical processes that create and energize it (e.g., Blandford & Levinson, 1995; Sikora & Madejski, 2000, and the discussion below of the theory of jet production). At this point, however, the evidence is conflicting. Upon detecting circular polarization in the core of the quasar 3C 279, Wardle et al. (1998) concluded that about 5/6 of the positively-charged particles are positrons. This is based on their application of conversion of linear to circular polarization in an optically thick region and the need to avoid excessive Faraday rotation that would reduce the observed level of circular polarization. However, Ruszkowski & Begelman (2002) subsequently found that a suitably turbulent magnetic field in a 100% normalmatter plasma could produce the observed effect. Various other attempts to determine the answer by indirect means have also led to contradictory conclusions (e.g., Cellotti & Fabian, 1993; Reynolds et al., 1996; Sikora & Madejski, 2000; Cellotti, 2003). Definitive evidence for a pair plasma would be detection of the 511 keV annihilation line. In the jet itself, the line would be broadened beyond recognition even if the 10 density of pairs were high enough to compensate for the diminished annihilation crosssection when the pairs are relativistic. If a pair-loaded jet were to mix with the interstellar medium of the host galaxy, however, they could thermalize, which would both increase the cross-section and produce a narrow line as the positrons annihilate with the interstellar electrons. Such mixing of the jet with clouds seems to be occurring in the radio galaxy 3C 120 (Hua, 1988; Axon et al., 1989; Gómez et al., 2000). An attempt by the author and collaborators (Marscher et al., in preparation) to detect the redshifted 511 keV line in 3C 120 with INTEGRAL failed, however. Either the positrons compose less than about 10% of the positive particles in the jet (a preliminary upper limit) or the minimum energy of the positrons is at least 5 MeV. If the latter, we need to understand why the jet does not accelerate from conversion of internal energy to flow energy (see Blandford & Rees, 1974), since the rest mass would be too low to make the flow ballistic. THEORETICAL MODELS OF RELATIVISTIC JETS IN AGN As outlined in the above sections, the jets of AGN can flow at Lorentz factors up to almost 50 and carry a highly energized plasma out of the nucleus with a kinetic power greater than 1046 erg s−1 . No one would have dared to predict such a phenomenon in a system that relies on accretion onto the ultimate cosmic abyss—a supermassive BH. Yet that is what theorists now are called upon to explain. The general consensus at this point is that the production of jets is tied to the magnetic field in the ergosphere of a rotating BH or, perhaps for the not so highly relativistic jets, in the inner accretion disk (see, e.g., Meier, Koide, & Uchida, 2000). Here I briefly summarize the currently most prominent theoretical scenarios so that they can be related to observations of jets, which are restricted to regions that are a large number of gravitational radii away from the central engine. Launch, Acceleration, and Collimation of the Jet As was presented by J. McKinney, E. Blackman, and D. Meier at the conference (see also McKinney, 2006; Meier, Koide, & Uchida, 2000; Camenzind, 2005), theorists are now concentrating on production of jets via differentially twisting magnetic fields tied to the ergosphere of the central BH. One of the primary difficulties is to get enough particles across the field lines so that the jet is more than just Poynting flux. However, if at least some particles make it into the flow, this difficulty could be an advantage, since a high ratio of magnetic to rest-mass energy density at the base is needed to propel the jet to a high Lorentz factor farther out (Vlahakis & Königl, 2004; Camenzind, 2005). Pair creation might be the dominant process that injects particles into the base of the jet. As envisaged by McKinney (2006), photons from the accretion disk and corona can enter the jet and pair produce off the strong magnetic field. One basic question is whether the jet is accelerated to a high bulk Lorentz factor close to the central engine or over a distance of hundreds or thousands of gravitational radii. Initial calculations by Phinney (1987) led to the conclusion that only jet flows with Γ < 10 or so could escape the immediate environs of the central engine without 11 being slowed down by Compton drag imparted by scattering of the ambient radiation field. Subsequent analyses (e.g., Sikora et al., 1996) found that this does not apply to electron-proton or Poynting flux dominated jets. Nevertheless, distributed acceleration and collimation has some advantages with respect to explaining the inverse relation between opening angle and Lorentz factor, the SED, and the substantial offset of the bright core from the position of the black hole (see below). In gas dynamical models, the acceleration and focusing of the jet occur over an extended distance (Blandford & Rees, 1974; Marscher, 1980; Melia, Liu, & Fatuzzo, 2002). The same is true in the magnetic acceleration and collimation model of Vlahakis & Königl (2004). In the author’s opinion, the acceleration region in the Vlahakis & Königl calculation extends too far from the black hole, but the results are scalable to smaller dimensions. Based on some of the ideas mentioned above, a popular picture of AGN jets envisages an ultra-fast “spine,” perhaps composed of electron-positrons pairs, surrounded by a slower (but still relativistic) “sheath” containing normal matter (e.g., Sol, Pelletier, & Asséo, 1989; Punsley, 1996). There is considerably observational justification for this conjecture on kiloparsec scales in FR I (low luminosity) radio galaxies (Laing et al., 1999), but this is thought to result from entrainment of the external medium (Bicknell, 1994; Laing, 1996). Ghisellini, Tavecchio, & Chiaberge (2005) find that lateral velocity structure in compact jets would allow less extreme physical parameters to be used for explaining the rapid high-energy variability of TeV BL Lac objects. Pushkarev et al. (2005) see evidence for sheaths with polarization electric vector roughly perpendicular (magnetic field parallel) to the jet axis in the polarized intensity VLBA images of four compact jets, while closer to the axis the polarization is nearly parallel to the axis. A major challenge for the ultra-fast spine picture is the small fraction of blazars jets that have Lorentz factors exceeding 10 (roughly 15%; see above). This implies that most, but not all, of the spines would need to be radiatively inefficient in order for them to remain unobservable. While possible, this seems contrived, so the burden is on the proponents of such a scenario to develop a model that can be tested, e.g., through source statistics. One strong prediction of the magnetic launching models is that the geometry of the field should correspond to a helix. Although there have been some indications that helical fields might be consistent with polarized intensity VLBI images (e.g., Gabuzda, Murray, & Cronin, 2004; Lyutikov, Pariev, & Gabuzda, 2005), models incorporating shocks propagating through a turbulent plasma with disordered field explain a wider range of phenomena associated with jets (e.g., Hughes, Aller, & Aller, 1985, 1989; Marscher, Gear & Travis , 1992; Lister & Homan, 2005; Cawthorne, 2006). In addition, Sikora et al. (2005) consider various observational constraints and conclude that the region of helical magnetic field is restricted to distances from the central engine considerably less than that of the structure resolved on VLBI images. Models for Knots in Jets and Variability in Flux and Polarization If you were to ask 20 compact jet experts what causes the superluminal knots in AGN jets, probably about 18 would say “shocks.” Our own neighborhood, the Solar System, 12 contains an outflow—the solar wind—through which shocks sweep following coronal mass ejections, so it is a familiar concept. It is more difficult, though, to generate a shock when the flow is near the speed of light, since there needs to be a relative fluid √ velocity exceeding the local sound speed. If that is the fully relativistic value, c/ 3, then the Lorentz factor of the faster flow must be at least a factor ∼ 2 higher than that of the slower flow that it overtakes. Superluminal knots formed by shocks must therefore arise from major disturbances at the injection point of the jet. After the original proposal (Blandford & Königl, 1979) that the moving knots might be shocks, Hughes, Aller, & Aller (1985), Hughes, Aller, & Aller (1989), and Marscher & Gear (1985) were able to reproduce the time variability of the flux, polarization, and continuum spectra of blazars quite well with models involving a limited number of free parameters. In the (Marscher & Gear, 1985) description, electrons are accelerated at the shock front, after which they lose energy to radiative and expansion losses as the flow in which they are imbedded lags behind the front. This results in energy stratification, with the highest energy electrons confined to a thin layer behind the shock front and progressively lower energy electrons occupying larger volumes. Since very little synchrotron radiation occurs above the critical frequency, which is proportional to the square of the energy, there is strong frequency stratification behind the shock. When applied to an outburst or a flare, this leads to time lags (higher frequency variations leading those at lower frequencies), a steepening of the spectrum (spectral index increases by 0.5), and a rather flat dependence of peak flux on frequency after the initial rise and before a final adiabatic decay. Lower amplitude flares and flickering can be explained as the consequence of a shock moving through turbulence (Marscher, Gear & Travis , 1992). Türler, Courvoisier, & Paltani (1999) and Türler, Courvoisier, & Paltani (2000) have used the (Marscher & Gear, 1985) model as a template for fitting multiple outbursts in the quasar 3C 273, with satisfactory results. The shocks that these models invoke are transverse square waves, which should compress the magnetic field component that lies parallel to the shock front. In most jets showing superluminal motion, the orientation of the shock front is aberrated to appear nearly transverse to the jet axis in the observer’s frame. Subsequent VLBI observations of a large sample by Lister & Homan (2005) revealed polarization directions in quasars that are inconsistent with this assumption, although BL Lac objects and the quasars modeled in the Hughes et al. papers have polarizations in agreement with transversely oriented shocks. In the quasar case, the magnetic field seems to be mostly turbulent but with some fraction being uniform and aligned parallel to the jet axis. In general, if this model for knot creation is valid, most shocks must be oblique to the jet axis (see, e.g., Aller, Aller, & Hughes, 2003). The other prominent model for knots gives them the affectionate appellation of “blobs.” This seems to mean that the knot is an entity that is distinct from the ambient jet flow as opposed to a feature that is fully integrated into the flow, as is the case for a shock. This allows a blob to have its own independent magnetic field and for its destruction to require either expansion or diffusion into oblivion, or perhaps some catastrophe such as running into a cloud. The main difference for models is that the blob can expand in three dimensions, whereas steady jets only expand laterally, and the particles and magnetic field do not need to be directly related to that of the surrounding jet plasma. The author does not favor the idea, since his attempt to model an outburst of 3C 273 as an 13 expanding blob (Marscher & Gear, 1985) led to the nonsensical requirement—despite (or maybe because of) the increased number of free parameters—that the magnetic field decreased while the particle density increased to cause the flare. One of the most striking characteristics of AGN jets is their flat spectra from radio to millimeter or, in extreme cases, submillimeter wavelengths. This has been dubbed a “cosmic conspiracy” by Cotton et al. (1980). The gradients in magnetic field and electron energy distribution must have specific values that cause the superposed spectra of the core plus many knots to produce a flat composite spectrum. As noted by Königl (1981), this occurs if the magnetic field falls off as z−1 while the normalization to the electron energy distribution (usually designated as No or K) declines with distance z down the jet as z−2 . The former can occur if the component of the magnetic field that lies transverse to the axis of the jet dominates on parsec scales. But the low polarization of jet cores implies that the field is turbulent, in which case one would expect a steeper fall-off than z−1 . Only the transverse component of the turbulent field should be amplified in transverse shock waves, leading to an observed mean alignment perpendicular to the jet axis. However, only the polarization of knots in BL Lac objects tends to follow this trend. In most quasars, the polarization is oblique when the effects of relativistic aberration are removed (Lister & Homan, 2005). The z−2 dependence of No requires particle acceleration that negates expansion cooling as the flow diverges. This means that some mechanism—turbulence or shocks, perhaps—prevents draining of the internal energy of the plasma as the flow spreads out and accelerates. It is a bit discomforting that such fine tuning is required to produce the ubiquitous flat centimeter to millimeter-wave spectra of AGN jets. The Marscher & Gear (1985) model provides another way to solve the “cosmic conspiracy” mystery. Their shock model for the evolution of the spectral energy distributions of outbursts in relativistic jets contains a stage at millimeter and submillimeter wavelengths when the flux at the synchrotron self-absorption turnover remains nearly constant as the turnover moves to lower frequencies, owing to synchrotron losses becoming weaker with time. (The adiabatic evolution of outbursts causes them to have higher peak flux densities at higher frequencies as the flaring region expands down the jet. Synchrotron losses suppress the emission at higher frequencies.) This naturally leads to a flat overall spectrum, although it is probably only part of the solution. The cores themselves have either flat spectra or flux densities that gently rise with frequency; read on for more on this topic. Models for the Core As mentioned above, at sufficiently low radio frequencies, what is seen as the core on VLBI images is the transition region between optically thick and thin emission (see Königl, 1981). In this case, the position of the core should move toward the central engine at higher frequencies. This effect is apparent in some (compact) extragalactic jets (e.g., Lobanov, 1998), but not in others (e.g., Mittal et al., 2006). The implication of the latter negative result is that the jet is not just a smooth flow punctuated by superluminal knots, but also contains multiple bright stationary features. 14 As the author has pointed out previously (Marscher, 1995), while the jet seems roughly self-similar on parsec scales, it cannot remain so all the way down to the launching site adjacent to the central engine. The point where the change must occur is indicated by the frequency νm at which the spectral energy distribution steepens, signaling that the jet emission is optically thin. According to the observations of Impey & Neugebauer (1988), this is in the range of tens to > 1000 GHz. The angular size of the core at this point can be estimated by the value at 43 GHz, ∼ 50 µ arcsec times 43 GHz/νm . For a quasar with redshift of order 0.5, this translates to ∼ 0.3(43 GHz/νm ) pc, which is small (a few 1016 cm for νm ∼ 1000 GHz) but still hundreds of gravitational radii even for MBH ∼ 109 M . This is just the width of the jet: the distance from the black hole is larger still since the jet is at least somewhat focused within this point. What can cause the jet to have its brightest spot well downstream of its launching pad? One possibility, mentioned previously, is that the acceleration of the flow is gradual, so that the greatest Doppler beaming occurs some distance from the central engine, as in the recent magnetic launching model of Vlahakis & Königl (2004). This explanation for the detached core works under certain conditions connected with the relationship between the increase in bulk Lorentz factor and the lateral expansion of the jet (see Marscher, 1980). A prime prediction of this model is that the cores of radio galaxies whose jets are viewed at an angle  Γ−1 should lie at the base of the jet if electrons are accelerated there. This implies that there should be a very short gap between the start of the jet and the beginning of the counterjet, except for a limited section obscured by free-free opacity in the accretion disk. Furthermore, the flow should start out fairly broad and become more collimated as it accelerates downstream. VLBA observations of the radio galaxy M87, whose very massive black hole (∼ 3 × 109 M ; Ford et al., 1994; Harms et al., 1994; Macchetto et al., 1997) and local distance allow a linear resolution of tens of gravitational radii, indicate that the flow is indeed quite broad near the central engine (Junor, Biretta, & Livio, 1999). Another possibility is that the core is a conical “recollimation” shock that accelerates particles and amplifies the component of the magnetic field that is parallel to the shock front without decelerating the flow too severely (Daly & Marscher, 1988; Gómez et al., 1995; Bogovalov & Tsinganos, 2005). This idea received support at the conference in a paper by T. Cawthorne, who compared quite favorably the polarization structure near the core in some blazars observed with the VLBA at 43 GHz by Jorstad et al. (2005) with the expectations of a turbulent (i.e., with random magnetic field direction) plasma passing through conical shocks in jets viewed at small angles to the line of sight (see Cawthorne, 2006). In some cases, the polarization electric vectors are aligned radially from the centroid of the core. The variation of percent polarization across the core region can be matched as well. But why should the core at 43 GHz be something other than the position where the optical depth is of order unity? The answer could be that the conical shock seen at this frequency is one of several that form a sequence in the jet, as occurs in gas dynamical simulations when the jet is circularly symmetric (Gómez et al., 1995). The most compact feature in the jet if it could be imaged at ν = νm would then be the first conical shock, while at lower frequencies the secondary standing shocks would be seen. Indeed, VLBI images at various frequencies have revealed that stationary features downstream of the core are common (Jorstad et al., 2001; Kellermann et al., 2004; Jorstad et al., 2005). In addition, the non-universality of the frequency-dependent 15 core position mentioned above implies that the core might be a feature of the flow rather than an optical depth transition point in many jets. In some objects the core could be a point where the jet bends such that it is more aligned with the line of sight, thereby corresponding to an increase in the Doppler beaming factor relative to points farther upstream. The core could well be different phenomena at different frequencies: the optical depth transition point at low frequencies, standing conical recollimation shocks at various frequencies, and the point where the Doppler factor reaches its asymptotic value in an accelerating or bending jet at the spectral turnover frequency νm . RELATIONSHIP BETWEEN THE JET AND THE CENTRAL ENGINE One of the great breakthroughs in the study of relativistic jets was the discovery of a connection between accretion state and events in the jet in XRBs (Mirabel & Rodríguez, 1998). As was reviewed by R. Fender at the conference (see also Fender & Belloni, 2004), in the best studied of the microquasars, GRS 1915+105, a weak jet with a high self-absorption turnover frequency persists during low-hard X-ray emission states. During the transition to the high-soft X-ray state, a pulse of energy is sent down the jet—subsequently seen in the form of a superluminal knot—following which event the jet shuts down, perhaps completely. Since in XRBs the soft X-rays come from the inner accretion disk, there seems to be an intimate relationship between the accretion state and conditions in the jet. In particular, when the innermost disk no longer radiates Xrays profusely, the jet turns on for an extended period. Then the radiative conditions return to the inner disk, accompanied by a whack to the jet that excites it just before it shuts down. Since Fender and Miller will discuss the various states of XRBs in their papers in this volume, this section will focus on the question of whether a similar connection between drops in X-ray flux and appearance of superluminal knots in the jets of AGN exists. The author and collaborators have been observing the radio galaxies 3C 120 and 3C 111 in an attempt to find out, since these objects have X-ray emission mostly from the central engine (Eracleous, Sambruna, & Mushotzky, 2000) as well as relativistic jets featuring apparent superluminal motion (Gómez et al., 2000; Jorstad et al., 2005). As reported in Marscher et al. (2002), superluminal “ejections” observed in sequences of VLBA images in 3C 120 follow marked dips in X-ray flux accompanied by hardening of the X-ray spectrum. There is a time delay of 60 days (updated value) that is partly caused by the offset of the core from the central engine—more than 0.3 pc or 200,000 gravitational radii for a black-hole mass of 3 × 107 M (Wandel, Peterson, & Malkan, 1999; Marshall et al., 2004)—and partly by the core shift between the observation frequency of 43 GHz and the self-absorption turnover frequency νm ≈ 100 GHz (Bloom et al., 1999). More intense monitoring over the past four years (Marscher, 2006) confirms this trend, with the quantitative addition of a formal, if weak, anticorrelation between X-ray flux and high-frequency radio flux. The X-ray dips in 3C 120 are not nearly as dramatic as are the transitions between X-ray states in GRS 1915+105, but they are not minor—typically of order a factor of 16 2 change in flux. If the time scales in 3C 120 were shortened by a factor of 2 million according to the ratio of the black-hole masses, the dips would last for ∼ 1 s. There are in fact variations of about this amplitude and time scale in GRS 1915+105 (Fender & Belloni, 2004), but no connection of these to the energy flow in the jet is apparent. Perhaps it is too much to expect direct scaling over so many orders of magnitude in size scale. Nevertheless, we can declare that the radiative state of the accretion disk plus corona system has a direct affect on events in the jet of this AGN. We have only been monitoring the FR 2 radio galaxy 3C 111 since March 2004. As is shown in Figures 9 and 10, the results thus far strongly suggest a similar relation between low X-ray states and the production of a bright superluminal knot in the jet. The first and third knots, whose appearances are marked by upward arrows in Figure 9, were initially brighter than the core. Of particular note is the prolonged low-hard X-ray state in 2005. The superluminal knot that emerged a few months after the first minimum/hardening during this state was trailed by a string of bright emission (see Fig. 10) commensurate with the extended duration of the low X-ray state. Despite its rather unspectacular appearance in the light curve, the spectrally hard drop in X-ray flux in early 2005 was about a factor of 2 lower than the maximum level of the high-soft state. The corresponding superluminal knot was fairly bright, but not as prominent as the other two; the 15 GHz flux was obviously elevated by the stronger two events. [Note: although the X-ray flux rises contemporaneously with the 15 GHz (λ 2 cm) light curve, the author considers this to be a coincidence, with the time delay between the X-ray event in the central engine and the core of the jet at νm happening to match the opacity-related time delay between the short millimeter-wave and 2 cm outbursts. This latter time lag is corroborated by unpublished 1.3 mm light curves from the SubMillimeter Array calibration database that can be viewed at website http://sma1.sma.hawaii.edu/callist.html.] The physical cause of the connection between events in the central engine and the jet is a matter of considerable speculation. If the jet is magnetically launched, then there must be a link between the magnetic state at the base of the jet and the accretion state in the inner disk. One scenario, proposed by Livio, Pringle, & King (2003), involves a change in the magnetic field configuration in the inner disk from turbulent in the high-soft state to mainly poloidal in the low-hard state. The turbulence is needed for viscous heating, which in the XRB case leads directly to bright X-ray emission with a soft spectrum. In an AGN it causes strong ultraviolet emission, which is Compton scattered in the corona to a hard X-ray spectrum. If the field switches from chaotic to mainly poloidal—Livio, Pringle, & King (2003) suggest that this can occur by random near-alignment of the field in the relatively small number of turbulent cells in the inner disk—then the radiation in the inner disk will be quenched at the same time as energy flow into the jet is promoted. The transition back to the turbulent, radiative inner disk would need to be violent in order to send a shock wave (see Miller-Jones et al., 2005) down the jet. Perhaps global magnetic reconnection could cause this, but no detailed MHD model has been published to date. At the conference, J. Miller raised the possibility that the “corona,” where the X-ray emission seen in AGN supposedly arises from Compton upscattering of softer accretiondisk photons, might be the base of the jet or of a wind blown off the disk (see his paper in this volume for the observational justification). If it is in fact the innermost section of the jet in the picture where the jet starts out relatively slow and speeds up with distance 17 FIGURE 9. X-ray light curve, obtained from measurements with the Rossi X-ray Timing Explorer (RXTE), along with X-ray spectral index vs. time and 15 GHz light curve from the University of Michigan Radio Astronomy Observatory. The upward arrows in the top panel indicate times when new, bright superluminal knots were coincident with the core as seen in sequences of VLBA images; horizontal bars indicate the uncertainties in the corresponding dates. Data from the author and collaborators (Marscher et al., in preparation). from the black hole, then the X-ray flux will be related to the number of electrons residing there. When the jet flow is faster, we should expect the electron density to decrease according to the continuity equation. So, a disturbance that increases the speed of the jet might decrease the scattered X-ray flux before it steepens into a shock farther downstream and becomes a superluminal knot. The reason for the increased velocity is unclear, however, as would be the reason why it occurs as the radiative inner accretion disk re-establishes itself in the XRB GRS 1915+105. SUMMARY The above review reflects the state of our knowledge of AGN jets as seen through the reading glasses of just one astronomer who has been examining the phenomenon over the past 30 years. The author hopes that it gives a flavor of where the observations have led us as well as of the controversies that remain. The primary goal has been to characterize those properties of AGN jets that can be related to XRBs and GRBs. The jets of AGN flow with Lorentz factors ranging from close to one to about 45, with the higher values found in a small fraction of the population. A connection exists between events in the central engine and the subsequent appearance of superluminal knots downstream, as is the case for microquasars. There is considerable observational 18 FIGURE 10. VLBA images of the FR 2 radio galaxy 3C 111 at 43 GHz. The core is the bright feature on the rightmost (western) end of the jet. Left: Image from May 2004 showing the typical appearance of the jet when there are no strong superluminal knots present. Right: Image from March 2006 displaying a bright superluminal (3-4c) knot with a strong bridge of emission between the knot and the core. support—e.g., opening angles that are inversely proportional to the Lorentz factors— for jets that accelerate and collimate over some distance as they propagate, with the first prominent synchrotron emission occurring hundreds or thousands of gravitational radii from the black hole. Magnetic launching models are popular, but the weak linear polarization observed in VLBI images indicates that, in the majority of blazars, the ordered field gives way to a turbulent field before the flow reaches the core. Rough equipartition between the field and particle energy densities is the norm in the radioemitting portion of the jet, with the particles dominating during outbursts. We can sketch a rough physical picture for how this change in energy states is accomplished, with velocity-shear inducing turbulence featuring magnetosonic waves that convert Poynting flux to electron energy in turbulent regions, and with shocks that tap the flow energy provided by the initially strong field to accelerate particles further as they repeatedly cross the shock front. Depending on the frequency of observation, the core seen in VLBI images can be the location of the optically thick/thin transition, a standing conical or non-axiallysymmetric oblique shock, or the point at the end of the acceleration zone or site of a bend where the Doppler factor reaches its highest value. “Internal” shock models for the moving knots boast substantial explaining power, but the polarization indicates that either the shocks are mostly oblique or another description is needed. We do not yet 19 know whether the jet is composed mainly of protons and electrons or pairs, although individual opinions about this tend to be very definite; the polarization of the community on this point is much stronger than the polarization of the jet cores! AGN jets tend to bend slightly and wobble near the core. The latter motion seems too irregular in most—but not all—cases to be explained by precession. As P. Hardee reviewed at the conference and in this volume, jets are subject to various instabilities, which can impart pinches, asymmetries, transverse gradients, ribbon-like structure, and other irregularities. These can have striking affects on the patterns of emission that we observe, in large part because variations in Doppler beaming are present across and down the jet. This makes deciphering our images, polarization, and light curves quite challenging. But it also guarantees that there will be plenty to do even after we find solutions to the basic physics that govern the formation, acceleration, collimation, propagation, energization, and radiation of relativistic jets. 20