You are on page 1of 5

Proc. Natl. Acad. Sci. USA Vol. 92, pp.

11442-11446, December 1995 Colloquium Paper

This paper was presented at a colloquium entitled "Quasars and Active Galactic Nuclei: High Resolution Radio Imaging, " organized by a committee chaired by Marshall Cohen and Kenneth Kellermann, held March 24 and 25, 1995, at the National Academy of Sciences Beckman Center, Irvine, CA.

The acceleration and collimation of jets


(magnetic rields/magnetohydrodynamics/shock waves/turbulence)

MITCHELL C. BEGELMAN
Joint Institute for Laboratory Astrophysics and Department of Astrophysical, Planetary, and Atmospheric Sciences, University of Colorado, Boulder, CO 80309-0440

I will discuss several issues related to the ABSTRACT acceleration, collimation, and propagation of jets from active galactic nuclei. Hydromagnetic stresses provide the best bet for both accelerating relativistic flows and providing a certain amount of initial collimation. However, there are limits to how much "self-collimation" can be achieved without the help of an external pressurized medium. Moreover, existing models, which postulate highly organized poloidal flux near the base of the flow, are probably unrealistic. Instead, a large fraction of the magnetic energy may reside in highly disorganized "chaotic" fields. Such a field can also accelerate the flow to relativistic speeds, in some cases with greater efficiency than highly organized fields, but at the expense of self-collimation. The observational interpretation of jet physics is still hampered by a dearth of unambiguous diagnostics. Propagating disturbances in flows, such as the oblique shocks that may constitute the kiloparsec-scale "knots" in the M87 jet, may provide a wide range of untapped diagnostics for jet properties.

The hydromagnetic paradigm for the acceleration and collimation of jets has risen to dominance, in part, through a process of elimination. Models involving acceleration by radiation pressure would have to be stretched to implausible extremes (even if the jets were largely made of pairs; see ref. 1; M. Sikora, H. Sol, M. C. Begelman, and G. Madejski, unpublished results) to account for the Lorentz factors (!10) that are needed to explain extreme cases of superluminal motion, variable gigaelectronvolt y-rays from blazars, and intraday radio variability. Losses due to cooling would place even more severe constraints on acceleration by gas pressure. Moreover, hydromagnetic propulsion has other attractive features besides the ability to produce the high speeds indicated by observations. Chief among these is the tendency of magnetic tension to collimate the flows automatically due to the development of a predominantly toroidal magnetic field. Thus, it is argued, one might not need to invoke a "funnel" or external confining medium to explain the collimation of jets. The fact that magnetic propulsion is the favored mechanism for protostellar jets (where the photon flux is clearly inadequate for radiative acceleration) adds the cachet of universality to the mechanism.
The Hydromagnetic Paradigm: Strengths and Shortcomings

As a theoretical problem, hydromagnetic propulsion of jets has proven to be particularly tricky. Nevertheless, concerted attacks by several groups over the past few years have yielded
The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. 1734 solely to indicate this fact.

some important advances. Most recently, Ustyugova et al. (3) and Meier et al. * have produced two-dimensional, timedependent simulations of hydromagnetic outflows from accretion disks. As expected, the rotation of the disk quickly winds up the toroidal field. To release its pressure, the field expands along the rotation axis, yielding rapid acceleration, while pinching the flow laterally, yielding collimation. All of this occurs quite close in, which is encouraging given recent very-long-baseline radio interferometry (VLBI) observations, which show that the M87 jet is already well-collimated within 100 Schwarzschild radii of the central black hole (4). But the prospects for hydromagnetic "self-collimation" of jets on large scales are less clear. The Ustyugova et al. (3) simulations show most of the collimation occurring on flux surfaces that are inside the Alfven surface and well inside the fast magnetosonic surface (see, for example, their figure la). Flux surfaces outside the Alfven surface seem to pinch a bit but then flare outward (see their figure 2a), although the general trend is for an increasing number of flux surfaces to gather inside the Alfven surface with time. It is not surprising that collimation is more effective inside the Alfven surface than outside. Aligned magnetic field-even if it is wrapped toroidally and thus exerts a tension force-tries to expand to occupy as large a volume as possible. Magnetic self-collimation is, at some level, a misnomer-what the field really does is to transmit the collimating force from one place to another. In addition the field provides a "lever arm" so that a weak pressure applied over a large area can be concentrated into a stronger pressure applied over a smaller area. Where a field line is still inside the Alfven surface, the collimating force can be provided easily by the accretion disk at the base of the flow, largely by torsional stresses. But such forces are transmitted at the Alfven speed and, therefore, cannot be transmitted beyond the Alfven surface. And, once the fast magnetosonic point (which always lies outside the Alfven point) is crossed, there can be no causal connection at all between the disk and the flow. Collimation thereafter can be provided only by an external medium, pressing on the flux surfaces from the side. The question, then, is whether the one can arrange for most of the flux surfaces never to cross the Alfven surface. This seems implausible to me. The Alfven surface lies inside the light cylinder radius c/fl, where fl is the angular velocity at a foot of a given flux surface. For the subset of flux surfaces that presumably would carry most of the energy in a powerful relativistic jet-those from the inner accretion disk and/or black-hole ergosphere-this would amount to c 100 Schwarzschild radii! We know from observation that jets rapidly become broader than this, even when they are very
Abbreviations: AGN, active galactic nucleus; VLBI, very-longbaseline radio interferometry. *Meier, D. L., Payne, D. G. & Lind, K. R., 184th Meeting of the American Astronomical Society, 1994 (poster).

11442

Colloquium Paper: Begelman


well-collimated, so it would seem that some form of lateral confinement is necessary. This does not mean that jets must propagate into some stationary external medium. In the hydromagnetic wind from a differentially rotating disk, the Alfven surface for the outer parts of the wind can lie much further away than that for the inner parts. Thus, it is possible for the inner jet to be collimated by the pressure of the outer wind, which in turn is anchored by the inertia of the accretion disk. However, if collimation is to continue to large distances, then it is most likely to be provided by an ambient medium. We might envisage the collimation process as shown in Fig. 1. A similar scheme has been proposed by Blandford (5). What is required for an external medium to collimate a jet? The pressure in a radial flow with constant velocity and dominated by a toroidal (transverse) magnetic field drops as r-2. The external pressure must decrease with distance more slowly than this to collimate the flow (6). For significant collimation to take place, the rate of pressure decrease must be very gradual indeed, compared to what one might expect in the central regions of an active galaxy. If the pressure in the external medium drops more rapidly than r-2, the flux surfaces will expand to fill the space. Zhi-yun Li and I (7) studied the asymptotic structure of relativistic, unconfined flows at distances well beyond the light cylinder. The relativistic case is somewhat more complicated to analyze than the nonrelativistic case because the electric force given by the v x B/c electric field times the associated space charge density (both familiar from pulsar theories) is comparable to the magnetic force; indeed, the two forces tend to cancel to within a factor of order r-2, where F is the jet Lorentz factor (8). As a result, the flux surfaces in a free wind tend to focus toward the rotation axis, but only very slowly; the opening angle decreases with distance as (F2In r)-1 Another troubling aspect of relativistic hydromagnetic winds is the difficulty they have in converting electromagnetic energy into kinetic energy. The conversion of Poynting flux into kinetic energy is almost entirely governed by the rate at which the poloidal flux surfaces diverge. Even where the flow crosses the fast magnetosonic point, the Poynting flux exceeds the kinetic energy by a factor _F2/3. In an unconfined wind, this ratio can decrease with distance until most of the energy is in kinetic form but only at a logarithmic rate (7). The

Proc. Natl. Acad. Sci. USA 92 (1995)

11443

conversion is so slow because the net tension force almost exactly cancels the pressure gradient forces that try to push apart the toroidal loops of magnetic field, accelerating the flow at the expense of magnetic energy (9). If the flow is externally confined, the rate of conversion tends to be even lower. Even in the nonrelativistic case (10), the best one can hope for probably is a rough equipartition between electromagnetic and kinetic energy. The means that the flows never become highly supermagnetosonic, which makes it difficult for strong shocks to form in the flow (as might be needed, for example, to

explain radio knots). Most models for hydromagnetic winds require the central region of the accretion flow to be threaded by a well-organized and substantial poloidal magnetic field. To drive a relativistic jet with power Ej, regions of net poloidal flux D -~(cAj/W2)1/2 must be supported by the disk (9). No model has successfully accounted for the presence of such a large flux. Suppose a disk advects highly organized poloidal flux from the interstellar medium of the host galaxy (say, at a radius of 1 kiloparsec) all the way in to a region about 10 gravitational radii from a 108 MO black hole. To power a 1046 ergs- 1 relativistic jet from this region then requires a net magnetic flux of 1032 G cm2. If the accretion rate supports a similar luminosity, with a conservative inflow speed of O.Olc, then the mass found within 10 gravitational radii is Md 10-2 MO. Now, in the presence of perfect flux-freezing, CF a Md (or, equivalently, the poloidal field strength is proportional to the surface density). At 1014 cm, we have (D/M 10. Let us suppose that the catchment area for the accretion flow, at 1 kiloparsec, contains 108 MG) and has unidirectional poloidal field with the generous strength of 10-3 G. We therefore have CF/M 0.05. Although I have made assumptions and chosen numbers that support a high magnetic flux near the black hole, the numbers still fail by a factor of more than 100. Realistically, flux-freezing of poloidal field over a large range of disk radii is likely to be far from perfect. At radii of more than a few parsecs, the accretion disks around even the most luminous active galactic nuclei (AGNs) are likely to be cold and molecular. Poloidal field will slip due to ambipolar diffusion and possibly from turbulent resistivity associated with cloud-cloud collisions. These effects would further reduce the amount of flux that can be advected inward by the disk in real time. However, flux dragged inward by the accretion

DISK

BH

DISK

FIG. 1. Possible configuration of a fast jet core surrounded by a slower wind. Between the source of the flow and Alfven surface (region I), the jet core can be collimated by magnetic stresses transmitted from the disk. In region II, magnetic pinching of the outer wind collimates the jet core. In region III, the flow is collimated by the pressure in the ambient medium. The filled black circle represents the central black hole (BH), the rotation of which could power the fast inner jet.

11444

Colloquium Paper: Begelman

Proc. Natl. Acad. Sci. USA 92

(1995)

flow over a period of time might accumulate in the central regions, perhaps leading to a jet powered by black-hole spin a la Blandford and Znajek (11). This flux cannot accumulate forever, since the pressure of the confined field would eventually stop the flow. But it is not clear that a steady-state solution in which outward slippage balances inward advection, as has been advocated by Lovelace et al. (12) and Konigl (13), can concentrate the flux to a large enough degree to drive powerful jets. Moreover, it is not even clear that a steady state can be established, in practice. Instead, one might imagine a highly unsteady mode of field evolution, characterized by episodic reconnection of large regions of flux. It is tempting to identify the large-scale reconnection events with the production of superluminal knots in AGN jets (9).

Chaotically Magnetized Jets


Regardless of the amount of ordered poloidal field that reaches the central regions of AGNs, the "tangled" field is likely to be substantial. A turbulent field with zero net flux could be generated through instabilities inside the disk such as that described by Balbus and Hawley (14), reaching levels close to equipartition with the energy density of the gas. Thus, the net magnetic flux advected inward by the disk is likely to contribute only a small fraction of the total magnetic energy present close to the hole. The energy content of the dynamo field could be adequate to power jets, but its means of conversion into directed kinetic energy would differ considerably from what occurs in the magnetically ordered flows discussed above. For one thing, the tendency of a chaotically magnetized flow to self-collimate would be considerably weaker than in the case of an ordered toroidal field. External confinement of the flow would be needed to achieve collimation. (A trade-off would be that the large-scale tension and electric forces would not hamper the f low acceleration as much as in highly ordered relativistic flows.) Without a confining medium, the emergence of a self-collimated jet from the messy dynamo field above the disk would require the reconnection and dissipation of most of the dynamo field; one might imagine that the energy liberated by reconnection could be used to launch the jet, by literally flinging blobs of gas onto open field lines. It seems at least as likely that the random field would continue to dominate, forcing us to consider a rather different mode for jet acceleration. There are good reasons to believe that the magnetic fields in jets observed on large scales are not well-ordered. Most powerful jets show a region in which the mean field is aligned with the flow. Yet, if this field represented net flux and were extrapolated into the nucleus under the assumption of fluxfreezing, the magnetic pressure would be orders of magnitude too large to be contained (15). The emission of highly polarized synchrotron radiation by a source does not necessarily imply a highly ordered magnetic field (16, 17). In fact, the field can be completely chaotic in the sense that the mean field components (Bi) all vanish under suitable averaging, provided that the mean-square field components (B2) are anisotropic. Gas dominated by a highly disorganized magnetic field is very unlikely to be in static equilibrium, especially if field reversals occur on small scales. Small-scale field reversals mean large, localized magnetic pressure gradients and tension forces, which will drive turbulent motions. Presumably, some sort of equipartition will be established between the rms magnetic energy density and the turbulent kinetic energy density. Equipartition is predicted to occur in weak hydromagnetic turbulence, where magnetic fluctuations can be treated as perturbations to a mean field (18). In the case I wish to discuss the "mean field" might be negligible and the fluctuations might be highly nonlinear. A Simple Model for Anisotropic Chaotic Fields. To illustrate some intriguing possible behaviors of such a system, imagine

a volume of fluid containing completely chaotic magnetic field (in the sense described above). Let us assume that the ordinary gas pressure is negligible but that there is a velocity field consisting of an ordered flow, u, plus a chaotic velocity field, V, driven by the magnetic fluctuations. We will average over time with the assumptions (Bi) = (Vi) - (ujVj) = 0 for all i, j; and (BIBj) = (ViVj) = 0 for i ]j. We then define the magnetic energy densities associated with each field component by Ui (B;/87r) and the corresponding kinetic energy densities by K, = (pV22), where p is the mass density. We make no assumptions about the anisotropy of the turbulence-in other words the Ui and Ki can be different in different directions. One can think of this as a model for the sheared or compressed chaotic fields considered by Laing (16, 17). By using the continuity equation and the condition V * B = 0, along with the assumption that (Vj((p/at)) = 0, we can write the components of the time-averaged momentum equation in the following form:

-lX, [(Uj+ Uk) + 2Ki- Ui], [1] for each permutation of coordinate indices i, j, k, where the
at
T

p(

+p(uVU)i *

left-hand side contains the usual advective derivative and p now represents the time-averaged density. On the right-hand side, we find pressure gradient forces due to the transverse components of the field and the turbulent ram pressure (pV12). The final term comes from the magnetic tension in the chaotic field, which tries to force closed loops of field to contract. In the time-averaged equation, this gives rise to a force that behaves like the gradient of a negative pressure. Implicitly in our averaging process, we have neglected temporal correlations among the components of B and V. This is equivalent to assuming that the field reversals occur on arbitrarily small scales and explains why Eq. 1 contains no turbulent transport terms. We next consider the magnetic induction (flux-freezing) equation. Multiplying the ith component of this equation by B, and averaging as before (under the assumption here that terms like (Bj[V x (V x B)]i) vanish, we obtain
a~t

auUi +(u*VU,)=-2U,

V*U-

au

[2]

for each index i. Our neglect of turbulent transport is reflected here in the absence of dynamo terms. Instead of coupling among the various components, the rms field strengths behave as if the field components were independently frozen into the flow. Thus, in a narrow cylindrical jet with width d(z) and moving at (subrelativistic) speed v(z), the rms transverse field would scale as v-d-I while the rms longitudinal field would scale as d 2-i.e., no differently than if the field were well organized. Note that we have neglected the effects of resistive diffusion in the induction equation, which would tend to dissipate chaotic field. To make further progress, we need some prescription for relating the kinetic energy densities to the magnetic energy densities. Because of the highly anisotropic manner in which magnetic fields couple to fluid, it is not obvious how to make this connection. Hopefully, computer simulations of magnetohydrodynamic turbulence might shed some light on this problem within the next few years. In lieu of such simulations, one might adopt a parametric prescription intended to cover a wide range of possibilities, all subject to the physically motivated constraint that overall equipartition of energy hold. (It is not certain that even the assumption of overall equipartition applies under conditions of nonlinear turbulence.) An instructive but admittedly ad hoc model would be to assume that L,Ki = YiUi and also to adopt the Ansatz that Ui couples to Kj and Kk linearly through the combination Kj + Kk. The

Colloquium Paper: Begelman


latter assumption reflects the fact that magnetic fields only
exert forces in directions normal to the field line. We obtain the one-parameter family of models
1

Proc. Natl. Acad. Sci. USA 92 (1995)

11445

U,
which
can

(1 - )(Kj + Kk) + Wi,

[3]

be inverted to give

(3k- 1)Ki = (1 + ()Ui

(1

()(Uj + Uk).

[4]

On the strength of Eqs. 3 and 4 alone, we cannot restrict the value of the parameter (, and we therefore consider the entire range - o < ( < x, with the proviso that all of the Ui or Ki must be positive definite. This places a weak limit on the permissible level of anisotropy for certain values of (. To illustrate the range of behaviors modeled by Eqs. 3 and
4,
we

note the

following special

cases.

In the limits

( -+

00,

we

have K, = (U, + Uj + Uk)/3 (i.e., isotropic fluid turbulence for arbitrary magnetic anisotropy). For ( = 1/3, we have the converse: the magnetic energy densities must be isotropic regardless of the anisotropy of the fluid turbulence. (The latter does not seem plausible physically.) For ( = -1, we have Ki = (Uj + Uk)!2, implying that the turbulent velocity field couples only to the transverse magnetic field. Finally, for the case ( = 1 we have Ki = Ui, and the coupling is entirely separable along each coordinate. With these closures, one can use Eqs. 1 and 2 to study the dynamical properties of chaotically magnetized
systems.

anisotropy. Could the dissipation associated with this anisotropy cause the distributed emission from jets? Further Implications of Anisotropy. The existence of a saturated level of anisotropy maintained by instability could have important implications for the dynamics of jets. We can use Eqs. 1-4 to analyze the behavior of such a jet as it propagates down a background pressure gradient. In the quasi-one-dimensional approximation, we assume transverse pressure balance with the prescribed external pressure, U, + Uck + 2Kr - Ur = p(z). Note once again that we must include the magnetic "surface tension" term. The individual "components" of the flux-freezing equations (Eq. 2) are no longer valid under the action of small-scale instability, which continuously rearranges the directional distribution of rms field. Only the sum of the components holds, since the instability presumably does not change the total magnetic energy density. Integrating the summed flux-freezing equation across the cross-sectional area of the jet, A(z), and adopting the assumed form for the anisotropy, we obtain dlog Uz/dlog A = - 2(a + 1)/ (2a + 1). For a narrow jet in pressure balance with its surroundings we put Uz x p(z). The opening angle of the jet then scales as When the toroidal field dominates (a > 1), the effect of the instability is to make the jet easier for external pressure to collimate. Ifp(z) z ", the opening angle of the supersonic jet scales according to

0(z) A1/2/z.

Does Magnetic Anisotropy Drive Dissipation in Jets? A fluid permeated by chaotic but anisotropic magnetic field as described above is not in a minimum energy state. To show this, consider a cube of unit dimension (x = y = z = 1), containing initial magnetic energy densities Uzo, Uxo = axUzo, Uyo = ayUz0. Now imagine that we stretch and compress the sides of the cube subject to the constraints of flux-freezing and constant volume. The total magnetic energy is then given by
EB
= U O(

dlog 0 _ (2a + 1)n 1 dlog z -4(a + 1)-1[6

[6]

+ AX +

2y

[5]

In the limit a -> oc we recover the collimation law for pure toroidal field, requiring n < 2 for collimation. For a = 1 (isotropic field), the requirement is n < 8/3, corresponding to an adiabatic equation of state with y = 4/3. Note that the shear amplification leading to a dominant longitudinal field would make a jet harder to collimate than a jet with purely toroidal field, but this would not be a problem if the longitudinal field were confined to a thin shear layer on the surface of the disk as suggested by Laing (19).

where xyz = 1. One can easily show that EB is minimized when the system is isotropic, with Ux = Uy = Uz = (axay)'/3Uzo. The existence of a unique minimum energy state suggests that an anisotropic system should be unstable. One is tempted to guess that the system would relax to isotropy on a dynamical time scale. But when the we perturb Eqs. 1-4 (assuming wavelengths much larger than the field reversal scale), the system seems to be stable. The reason is that Eqs. 3 and 4, which enforce equipartition between magnetic and kinetic energies, do not permit the system to lower its energy. In other words, the minimum energy state is not accessible. For the system to approach isotropy, there must be some means of storing energy which does not participate in the equipartition, such as a form of dissipation that channels energy into thermal pressure or relativistic particles. In fact, an anisotropic system should be secularly unstable, with the relaxation rate determined by the rate of dissipation and the degree of anisotropy. This idea is directly applicable to jets in AGNs. A chaotically magnetized jet, continuously driven toward anisotropy by stretching or shearing, should approach a steady state in which the rate of increase in EB associated with anisotropy is balanced by dissipation. The degree of anisotropy should saturate at a fixed value. For simplicity, let us assume that the jet is axisymmetric with Ur = U,q, = acUz in cylindrical polar coordinates. Flux-freezing alone would then lead to > 1 (15). However, if shearing effects (which we neglected in Eq. 2) are strong enough, then U, could dominate with a > 1. The effect to either a of the secular instability would be to limit maximum or a minimum value, depending on the source of the
a a

Shocks as Jet Diagnostics


It is difficult to determine the nature of a jet by observing its steady emission. Chaotic but anisotropic magnetic fields can produce high enough polarization as readily as inhomogeneous but highly organized large-scale fields (16). There is generally too little thermal material interior to a jet to decide from Faraday rotation whether the field is organized or chaotic or even whether the jet plasma is composed of normal plasma or

electron-positron pairs. While not yet providing a panacea for these unknowns, observations of disturbances in the flow, such as shocks, can provide a great deal of information about the structures of jets. On scales observed with the Very Large Array, the potential value of shocks as diagnostics can be appreciated by considering the bright knots in the M87 jet, the proper motions of which have recently been measured by Biretta et al. (20). Knots B, D, and especially A are readily interpretable as shock fronts, propagating down the jet at about 0.6c (and observed to have proper motions of about 0.5c). As Biretta has pointed out, the morphology of knot A makes it appear edge-on (if it is assumed to be a planar feature) and tilted with respect to the jet axis by 180. Taking into account relativistic aberration and the likely orientation of the M87 jet relative to the plane of the sky [to explain the jet-counterjet ratio and the orientation of the gaseous disk observed with the Hubble Space Telescope (21)], Geoffrey Bicknell and I have deduced that the normal to the plane of the knot A shock would have to be highly inclined with respect to the jet axis. In other words, it would have to be a highly oblique shock. The other knots, and even the

11446

Colloquium Paper: Begelman

Proc. Natl. Acad. Sci. USA 92

(1995)

apparent "helical filaments," could all be oblique shocks seen in various projections. Oblique shocks have been discussed in the literature mainly in the context of supersonic jets propagating through pressure gradients (and thereby developing over- or under-pressures with respect to their surroundings) or glancing off dense

regions of the interstellar medium. Such shocks would be slowly moving relative to the jet's surroundings. The M87 knots, however, are moving at a significant fraction of the jet speed and presumably arise from some other cause. Rees (22) suggested that they arise as a result of fluctuations in the jet speed at the source. While this cannot be ruled out, it explains neither the quasi-periodicity of the knot spacing nor why the shocks are so oblique. An alternative explanation is that the shocks arise during the nonlinear development of a helical (m = 1) Kelvin-Helmholtz instability. G. V. Bicknell and I (unpublished results) have developed this idea quantitatively and find that we can self-consistently identify the knot speed and spacing with the group velocity and wavelength, respectively, of the most unstable Kelvin-Helmholtz modes. To obtain a high enough group velocity, the jet must be propagating through a tenuous background medium, with a density not more than 10-100 times that of the jet and about three orders of magnitude less than the density of the observable interstellar gas near the center of M87. If this interpretation is correct, it means that M87 has inflated a near-vacuum bubble around itself! What we do not yet know is the shock strength attainable before the instability saturates. Probably this will have to wait for three-dimensional relativistic hydrodynamic simulations. Bicknell and I have also analyzed the other properties that might characterize oblique shocks associated with the M87 knots. Observational clues include the brightness jump across the shock front, suggesting that the pressure jump is not too small; the maintenance of a large jet-to-counterjet brightness ratio (assuming that the counterjet undergoes similar shocks), implying that the deceleration across the shock front is not too great; and the relative straightness of the jet, implying an upper limit to the deflection angle as the flow crosses the oblique shock front. We find a range of consistent solutions with pressure jumps of a few, yet relatively weak decelerations and deflections of less than 10. In typical models, the upstream flow enters the shock at an angle of -50-60 with respect to the shock normal (in the stationary shock frame). We infer the bulk Lorentz factor of the jet to be in the range -3-7 and the rest mass energy density of "cold" (nonrelativistic) plasma to be at least of the order of the energy density in relativistic particles and electromagnetic fields. Many of the arguments made above for M87 would carry over straightforwardly to the superluminal knots in VLBI sources, if the latter can be identified with oblique shocks. The relativistic Kelvin-Helmholtz instability should be operative on VLBI scales if the jet is propagating through (and is confined by) a slower background medium. The association of knots with apparent curvature in VLBI jets fits in with their interpretation in terms of oblique shocks. In addition to the properties discussed above, one could also learn something about the jet orientation, magnetic field anisotropy, and degree of organization of the field by mapping polarization on VLBI scales [as is already being done for VLBI knots modeled as normal shocks by Wardle and collaborators (23, 24)]. Spatially resolved VLBI spectroscopy could yield information about synchrotron cooling rates in the flow and indirectly about the flow field downstream of the shock (2).
Discussion

more or less, descendants of that remarkable 1982 paper on self-similar magnetohydrodynamic flows by Blandford and Payne (10)] contain a lot of very elegant physics, but it is not clear how far they can be carried into the real world, where magnetic fields are likely not to be so well organized. One might be forced to sacrifice most of the magnetic selfcollimation, which made for much of the early appeal of hydromagnetic models. On the other hand, in doing so one might get more efficient acceleration, since large-scale "hoop" stresses will no longer hold back the flow. Magnetic stresses do seem to provide the most promising way of accelerating flows to the high bulk Lorentz factors indicated by observations. This seems to be the case whether the jet's magnetic field is highly organized, or completely chaotic. The study of physical conditions in AGN jets continues to be bedeviled by a lack of unambiguous observational diagnostics. Bright knots in jets, on scales ranging from parsecs to kiloparsecs, can plausibly be identified with shocks. Some of these knots, particularly those with high proper motions (such as the superluminal VLBI features and the knots in the kiloparsecscale M87 jet) are likely to be highly oblique shocks representing internal disturbances within the jet. If this interpretation holds, then careful study of these features may yield a great deal of new information about the physical makeup of jets. On VLBI scales, it will require the mapping of knot structures with high resolution, high dynamic range, and high precision polarization ... capabilities that are just now routinely within our grasp!
My research on jets is supported in part by National Science Foundation Grant AST91-20599.
1. Phinney, E. S. (1987) in Superluminal Radio Sources, eds. Zensus, J. A. & Pearson, T. J. (Cambridge Univ. Press, Cambridge, U.K.), p. 301. 2. Coleman, C. S. & Bicknell, G. V. (1988) Mon. Not. R. Astron. Soc. 230, 497-509. 3. Ustyugova, G. V., Koldoba, A. V., Romanova, M. M., Chechetkin, V. M. & Lovelace, R. V. E. (1995) Astrophys. J. 439, L39L42. 4. Junor, W. & Biretta, J. A. (1995) Astron. J. 109, 500. 5. Blandford, R. D. (1993) in Astrophysical Jets, STScI Symposium Series, eds. Burgarella, D., Livio, M. & O'Dea, C. (Cambridge Univ. Press, Cambridge, U.K.), Vol. 6, pp. 15-33. 6. Eichler, D. (1993) Astrophys. J. 419, 111-116. 7. Begelman, M. C. & Li, Z. (1994) Astrophys. J. 426, 269-278. 8. Tomimatsu, A. (1994) Pub. Astron. Soc. Jpn. 46, 123-130. 9. Begelman, M. C. (1994) in The Nature of Compact Objects in Active Galactic Nuclei, eds. Robinson, A. & Terlevich, R. (Cambridge Univ. Press, Cambridge, U.K.), pp. 361-367. 10. Blandford, R. D. & Payne, D. G. (1982) Mon. Not. R. Astron. Soc. 199, 883-903. 11. Blandford, R. D. & Znajek, R. L. (1977) Mon. Not. R. Astron. Soc.

179, 433.
12. Lovelace, R. V. E., Wang, J. C. L. & Sulkanen, M. E. (1987)

Astrophys. J. 315, 504-535.


13. Konigl, A. (1989) Astrophys. J. 342, 208-223. 14. Balbus, S. A. & Hawley, J. F. (1991) Astrophys. J. 376, 214-222. 15. Begelman, M. C., Blandford, R. D. & Rees, M. J. (1984) Rev. Mod. Phys. 56, 255-351. 16. Liang, R. A. (1980) Mon. Not. R. Astron. Soc. 193, 439-449. 17. Laing, R. A. (1981) Astrophys. J. 248, 87. 18. Zweibel, E. G. & McKee, C. F. (1995)Astrophys. J. 439,779-792. 19. Laing, R. A. (1995) Proc. Natl. Acad. Sci. USA 92, 11413-11416. 20. Biretta, J. A., Zhou, F. & Owen, F. N. (1995) Astrophys. J. 447, 582-596. 21. Ford, H. C., Harms, R. J., Tsvetanov, Z. I., Hartig, G. F., Dressel, L. L., Kriss, G. A., Bohlin, R. C., Davidsen, A. F., Morgan, B. & Kochhar, A. K. (1994) Astrophys. J. 435, L27-L30. 22. Rees, M. J. (1978) Mon. Not. R. Astron. Soc. 184, 61P-65P. 23. Wardle, J. F. C., Cawthorne, T. V., Roberts, D. H. & Brown, L. F. (1994) Astrophys. J. 437, 122-135. 24. Brown, L. F., Roberts, D. H. & Wardle, J. F. C. (1994)Astrophys. J. 437, 108-121.

The hydromagnetic paradigm for jet acceleration and collimation is alive and well, but it is very far from being fully worked out. The simple models of highly organized flows [all,

You might also like