THE ASTROPHYSICAL JOURNAL, 483 : L45L48, 1997 July 1

1997. The American Astronomical Society. All rights reserved. Printed in U.S. A.

THREE-DIMENSIONAL MAGNETOHYDRODYNAMIC SIMULATIONS OF RELATIVISTIC JETS INJECTED ALONG A MAGNETIC FIELD

ROBERT L. MUTEL4

ABSTRACT We present the rst numerical simulations of moderately hot, supersonic jets propagating initially along the eld lines of a denser magnetized background medium with Lorentz factor W 4.56 and evolving in a four-dimensional spacetime. Compared with previous simulations in two spatial dimensions, the resulting structure and kinematics differ noticeably: the density of the Mach disk is lower, and the head speed is smaller. This is because the impacted ambient uid and its embedded magnetic eld make efcient use of the third spatial dimension as they are deected circularly off of the head of the jet. As a result, a signicant magnetic eld component normal to the jet is created near the head. If the eld is strong, backow and eld reversals are strongly suppressed; upstream, the eld closes back on the surface of the beam and assists the collimation of the jet. If the eld is weak, backow and eld reversals are more pronounced, although still not as extended as in the corresponding plane-parallel case. In all studied cases, the high-pressure region is localized near the jet head irrespective of the presence strength of the magnetic eld, and the head decelerates efciently by transferring momentum to the background uid that recedes along a thin bow shock in all directions. Furthermore, two oppositely directed currents circle near the surface of the cylindrical beam, and a third current circles on the bow shock. These preliminary results underline the importance of performing fully three-dimensional simulations to investigate the morphology and propagation of relativistic extragalactic jets. Subject headings: galaxies: jets magnetic elds methods: numerical MHD relativity jets constitutes an exercise in two-dimensional relativistic magnetohydrodynamics (RMHD)and this is neglecting three-dimensional effects as well as radiation processes that give rise to the observed synchrotron emission (but see Gomez et al. 1995, Komissarov & Falle 1996a, and Georganopoulos & Marscher 1996 for radiative models). The inclusion of relativistic effects, in particular, is a recent endeavor in jet hydrodynamics. Following the older, brave effort of Yokosawa, Ikeuchi, & Sakashita (1982), two-dimensional, special relativistic simulations have been carried out over the past 4 years (van Putten 1993; Duncan & Hughes 1994; Mart Muller, & , Ibaez 1994; Mart et al. 1995, 1996; Koide, Nishikawa, & n Mutel 1996a; Komissarov & Falle 1996b). Furthermore, the study of frozen-in magnetic elds and of the third spatial dimension in the relativistic domain is just beginning (Koide et al. 1996a, van Putten 1996, and this Letter, respectively). These effects have already been included in nonrelativistic jet simulations and have led to successful comparisons between numerically predicted and actually observed characteristics (see, e.g., Balsara & Norman 1992, Hardee & Clarke 1992, 1995, and Hardee, Clarke, & Howell 1995 for threedimensional hydrodynamic simulations and Clarke, Norman, & Burns 1986, Stone & Norman 1992, Clarke 1993, Hardee & Clarke 1995, Hardee 1996, and Koide et al. 1996c for two- and three-dimensional magnetohydrodynamic [MHD] simulations). Reviews covering these topics have been written by Burns, Norman, & Clarke (1991) and by Norman (1993, 1996). Despite complications and changes in the details of jet propagation, no major surprises emerged from the consideration of active or passive magnetic elds. The most interesting results obtained from three-dimensional MHD simulations are the higher degree of beam collimation effected by dynamically evolving, strong toroidal magnetic elds as well as the creation of compact nose cones in magnetically conned jets and L45

1. INTRODUCTION

There is now ample evidence that extragalactic jets at scales of 10 100 pc ow out of active galactic nuclei at speeds comparable to the speed of light (v 2 c) and that the outows maintain signicant relativistic speeds (v 1 0.6c 0.8c) out to scales of at least 1 kpc. The evidence comes from modeling VLBI observations of superluminal motions in many ( 40) quasars and BL Lacertae objects (Gabuzda, Wardle, & Roberts 1989; Mutel et al. 1990; Cawthorne 1991; Davis, Unwin, & Muxlow 1991; Gabuzda et al. 1992; Hummel et al. 1992a, 1992b; Ghisellini et al. 1993; Biretta 1993; Biretta, Zhou, & Owen 1995; Wardle & Aaron 1996) and from the detection of intraday variability in more than 25% of the known compact radio sources (see Qian et al. 1991, 1996 and Krichbaum, Quirrenbach, & Witzel 1992 in conjunction with the theoretical estimates of Begelman, Rees, & Sikora 1994). These observations indicate that the bulk of the outowing material must be moving at Lorentz factors of W 1 10 (v 0.995c) or faster (W 1 30 100 and v 0.999c, according to Begelman et al. 1994). Polarization measurements and modeling also reveal that the jets are magnetized and that the magnetic elds should be dynamically important in some particular FR II radio sources and in BL Lacertae objects (Clarke 1993; Laing 1996; Koide et al. 1996c and references therein). Including the above effects in simulations of cylindrical uid
1 Department of Physics and Astronomy, Louisiana State University, Baton Rouge, LA 70803; [email protected], [email protected]. 2 Laboratory of Plasma Physics and Fusion Science, Faculty of Engineering, Toyama University, Gofuku, Toyama 930, Japan; [email protected], [email protected]. 3 DARC, Observatoire de Paris-Section de Meudon, Place Jules Janssen, F-92195 Meudon Cedex, France; [email protected]. 4 Department of Physics and Astronomy, University of Iowa, 203 Van Allen Hall, Iowa City, IA 52242; [email protected].

L46

NISHIKAWA ET AL.

Vol. 483

their shedding to the side before they grow too large (Clarke 1993). On the other hand, purely hydrodynamic three-dimensional simulations revealed some rather surprising structural changes: jets in three dimensions are not intrinsically as stable as in a two-dimensional geometry because their beams are deected, splatter, and create two hot spots instead of a Mach disk; their bow shocks are distorted accordingly; and the backow in their asymmetric cocoons is more turbulent, with vortical eddies maintaining fully three-dimensional structures (e.g., Norman 1993). In this Letter, we discuss new results obtained from fully three-dimensional RMHD jet simulations carried out for a brief period of time and at a modest grid resolution of 1013 zones. In 2, we demonstrate the effects of the included, third spatial dimension through comparisons with two-dimensional results that we obtained in a plane-parallel geometry (Koide et al. 1996a). A summary of our conclusions is presented in 3.
2. INITIAL CONDITIONS AND NUMERICAL RESULTS

The relevant RMHD equations and the employed, threedimensional computer program have been briey introduced by Koide et al. (1996a, hereafter KNM). A detailed description of the numerical techniques and related test problems in the relativistic domain are given in Koide, Nishikawa, & Mutel (1996b). For the present simulations, we employ a Cartesian XYZ grid with 101 equally spaced zones in each direction. The computational box is a cube that occupies the region 0 X 20, 10 Y, Z 10. This box is lled uniformly with uid of rest mass density a 1, pressure Pa 0.6, and specic energy a 0.9; a uniform magnetic eld B is also embedded and points initially in the X-direction. At times, t 0, jet material of rest mass density j 0.3 ( 0.3), and j a specic energy j 3 is injected in the X-direction from an orice at X 0, Y 2 Z 2 1. The jets proper pressure is initially equal to the thermal pressure of the ambient medium. No symmetry is assumed across any of the boundaries during evolution. In addition, xed-inow and radiative boundary conditions are implemented at the X 0 surface and at all the other surfaces, respectively. Thus, the propagation of the jet can be followed only for a relatively brief period of time spanning 20 initial jet radii. With this setup and resolution, we do not hope to follow the long-term evolution of the jets that might lead to pronounced asymmetries and head uctuations, and we concentrate on the structure and kinematics that develop near and around the surfaces of the beams. Computations are currently performed on the Power Challenge of the National Center for Supercomputing Applications with over 1 Gbyte dynamic memory and require 2.5 minutes of CPU time per time step. The present simulations were designed for a direct comparison with the corresponding twodimensional simulations presented by KNM. For this reason, we express velocities in units of the sound speed vs in the ambient medium, and we measure time in units of s, the transit time of a sound wave over unit distance. In addition, we choose an adiabatic index of 5 3 for the uid and an injection speed vj that leads to a proper Mach number of Mj vj vs 4 and a Lorentz factor of Wj (1 v2 c2 ) 1 2 4.56 j (vj c 0.9756 for c 4.1vs ). So far, we have performed three runs characterized by various strengths of the initial magnetic eld: in run A, the energy density of the eld, B2 2, was equal to 5 9 of the internal energy density, a a, of the ambient medium (Alfven

B 1 2 vs ); in run B, the energy density of the speed vA eld was reduced by a factor of 16 (vA 0.25vs ); and in run C, there was no magnetic eld. Because the results from runs B and C are similar, we describe here only the rst two runs with a strong and a weak initial magnetic eld, respectively. Figures 1a and 1b and Figures 1c and 1d (Plate L4) show various MHD variables on the Y 0 plane at t 8.5 s for runs A and B, respectively. Figures 1a and 1c depict the rest mass density and the velocity, while Figures 1b and 1d depict the thermal pressure and the magnetic eld. (Horizontal [XY] cross sections at Z 0 reveal very much the same morphology, while YZ cross sections show the overall axisymmetry of the jets.) In both cases, the Mach disks (Figs. 1a and 1c) are almost at the same location, X 2 17, because the average propagation speeds of the relativistic jet heads are almost the same: vh 2.06vs. (In more detail, the heads slow down to the following average speeds: vh 2.17vs for 0 t 3.0 s, 2.10vs for 3.0 s t 6.0 s, and 1.88vs for 6.0 s t 8.5 s.) Although higher than the one-dimensional, nonrelativistic estimate of 1 2 1 2 vh (1 ) 1vj 1.42vs, this average speed is signicantly smaller than both the two-dimensional value of vh 2.67vs found by KNM and the corresponding one-dimensional, relativistic, nonmagnetic estimate (e.g., Mart et al. 1996) of 1 2 1 2 1 vh 2.92vs, where the relativistic inertia R (1 R ) vj ratio R (hj ha )W 2 7.42 and h 1 ( P ) c2 dej notes the relativistic specic enthalpy. The observed head speed is in between the above relativistic and nonrelativistic estimates. This outcome is clearly due to a combination of relativistic and three-dimensional effects: Lorentz contraction increases the effective mass density of the jet, and hence its thrust and average speed, but the head of the beam is decelerated by the ambient medium more strongly in three dimensions than in two dimensions; hence, an average propagation efciency of only 2.06 2.92 70% is achieved because of more efcient momentum transfer to the impacted, three-dimensional uid. Comparing the two cases, the transverse propagation of the bow shock is faster in run A (Figs. 1a and 1b) because the fast magnetosonic waves are faster in the strongly magnetized medium. Also, the width of the bow shock in run A is smaller because the shocked ambient ow is not affected substantially by backow as the magnetic eld is deected by the head less than in run B (see Fig. 2 below). In relation to the twodimensional case, however, the widths of both three-dimensional bow shocks are smaller, and their tips are closer to the beams (see Figs. 1 and 2 in KNM). The density of the Mach disk and the head pressure are both higher by 110% in the weak-eld run B. In both runs, however, the high-pressure region is strongly localized around the head (Figs. 1b and 1d). This feature is also due to three-dimensional effects because it is not observed in the two-dimensional simulations of KNM. The localization occurs because there is no preferred direction for momentum transfer in the three-dimensional runs, leading to more efcient matter transport away from the head. Figure 1c shows wings of pronounced backow forming beside the jet in run B. With the exception of a thin surface layer behind the head, no such feature develops in the strong-eld case. The absence of backow in Figure 1a is similar to the result obtained by Mart et al. (1996) from a nonmagnetic, two-dimensional simulation of a considerably hotter jet. It seems then that either a strong, aligned magnetic eld in the ambient medium or a high, internal, specic energy

No. 1, 1997

MHD SIMULATIONS OF RELATIVISTIC JETS

L47

in the jet can independently prevent backow and cocoon turbulence. One of the best ways to illustrate structural details on the head and the surface of the jet is to trace the magnetic eld lines in three-dimensional space. This is shown in Figure 2 (Plate L00) for runs A and B at t 8.5 s. The observer is located in front and to the side of the approaching beam. The following features are readily apparent: (1) bending of the eld lines (rotational discontinuity) at the location of the bow shock; (2) a ared region behind the head, inside of which the ambient eld has been canceled by a current induced around the beam; and (3) conning, straight eld lines upstream along the surface of the beam in run A and pronounced eld-line reversals in run B (red curves). In run A, the strong magnetic eld that is pushed away by the propagating beam is deected almost circularly; as a result, it develops signicant normal components (on the Y-Z plane), comparable to the X-components. The eld closes back on the beam upstream and contributes to the connement of the exceptionally thin and internally featureless jet (Figs. 1a and 2a). Although deected in a similar fashion, the eld in run B is too weak to play the same role. Thus, the weak magnetic eld in run B does not inuence the propagation and collimation of the jet that evolves very much as in the nonmagnetic run C and results in the formation of a signicantly less conned beam (Figs. 1c and 2b). In addition, Figure 2b shows that the reversal of eld lines in run B is limited near the surface of the beam. By comparison, Figure 2a shows only a hint of very small eld reversals in run A. In the corresponding two-dimensional cases (KNM), both eld-line reversals and backowing wings are more pronounced and more extended. The regions where the magnetic eld is near zero (Figs. 1b and 1d) indicate the location of the induced current whose density distribution J B is quite complex, as shown in Figure 3 (Plate L5): two distinct, concentric, roughly cylindrical surfaces with toroidal current form near the surface of the beam and approach each other near the head, and another current circulates on the bow shock. The currents on the bow shock and the inner cylindrical surface have the same polarity, which is opposite to that of the current on the outer cylindrical surface. All three currents are generated from the interaction of the jet with the ambient plasma in places where the magnetic eld lines are bent or canceled, as described above. We nally note that a deviation of the beam from its original path has been observed during these simulations, and longer time integrations are needed in order to nd out whether this effect is the beginning of nonlinear head oscillations analogous to those seen in nonmagnetic, nonrelativistic, three-dimensional jets (Norman 1993). At the endpoint of our runs, the

heads of the jets are displaced by a few grid cells relative to the original direction of propagation. A displacement of this magnitude is not sufcient to produce pronounced asymmetries in the leading section of the beam. Its presence, however, is revealed on the bow shock in Figure 3, where the circulating current appears to be off-centered.
3. CONCLUSIONS

Fully three-dimensional RMHD jet simulations of modest grid resolution (1013 zones) and duration (up to t 8.5 s ) conrm the general results obtained from previous twodimensional simulations performed with an initially aligned magnetic eld or in the absence of a eld: both the relativistic terms in the RMHD equations and the presence of strong magnetic elds efciently assist the collimation of propagating jets that appear to be thinner when both effects are included. The new simulations indicate that, in addition, consideration of all three spatial dimensions is necessary in order to capture correctly the morphology and kinematics of the jets. In contrast to the two-dimensional case, we have found that moderately hot, three-dimensional relativistic beams decelerate more efciently, develop no internal structures, and maintain circular currents at parts of their surfaces and small localized heads of lower density and pressure, and that the associated bow shocks are considerably thinner and stay closer to the beams. Furthermore, the jets are better collimated and show no strong backows or signicant eld reversals in the presence of dynamically important ambient magnetic elds. A strong magnetic eld is temporarily pushed away by the head almost axisymmetrically but returns to conne the beam farther upstream. In the cases of weak or no magnetic eld, backowing wings and pronounced eld reversals do develop, but they are not as strong as in the corresponding twodimensional plane-parallel cases studied by KNM; hence, the lesser but signicant connement of the jet is entirely due to relativistic effects. We note that the magnetic eld in this study is initially aligned with the jet and that this conguration does not resist the propagation of the beam. The study of three-dimensional RMHD jets injected into an oblique ambient magnetic eld is currently in progress. We thank an anonymous referee for constructive comments and suggestions. This work was supported in part by the US-Japan cooperative science program JSPS and by NSF (INT 92-17650); by NSF grants ATM 91-19814, ATM 91-21116, and AST 95-28424; by NASA grant NAG5-2777; and by supercomputing grants at NCSA, PSC, and SDSC.

REFERENCES Balsara, D. S., & Norman, M. L. 1992, ApJ, 393, 631 Begelman, M. C., Rees, M. J., & Sikora, M. 1994, ApJ, 429, L57 Biretta, J. 1993, in Astrophysical Jets, ed. D. Burgarella, M. Livio, & C. ODea (Cambridge: Cambridge Univ. Press), 263 Biretta, J. A., Zhou, F., & Owen, F. N. 1995, ApJ, 447, 582 Burns, J. O., Norman, M. L., & Clarke, D. A. 1991, Science, 253, 522 Cawthorne, T. V. 1991, in Beams and Jets in Astrophysics, ed. P. A. Hughes (Cambridge: Cambridge Univ. Press), 187 Clarke, D. A. 1993, in Jets in Extragalactic Radio Sources, ed. H.-J. Roser & K. Meisenheimer (Berlin: Springer), 243 Clarke, D. A., Norman, M. L., & Burns, J. O. 1986, ApJ, 311, L63 Davis, R. J., Unwin, S. C., & Muxlow, T. W. B. 1991, Nature, 354, 374 Duncan, G. C., & Hughes, P. A. 1994, ApJ, 436, L119 Gabuzda, D. C., Cawthorne, T. V., Roberts, D. H., & Wardle, J. F. C. 1992, ApJ, 388, 40 Gabuzda, D. C., Wardle, J. F. C., & Roberts, D. H. 1989, ApJ, 336, L59 Georganopoulos, M., & Marscher, A. P. 1996, in ASP Conf. Ser. 100, Energy Transport in Radio Galaxies and Quasars, ed. P. E. Hardee, A. H. Bridle, & J. A. Zensus (San Francisco: ASP), 67 Ghisellini, G., Padovani, P., Celotti, A., & Maraschi, L. 1993, ApJ, 407, 65 Gomez, J. L., Mart J. Ma., Marscher, A. P., Ibaez, J. Ma., & Marcaide, J. M. , n 1995, ApJ, 449, L19 Hardee, P. E. 1996, in ASP Conf. Ser. 100, Energy Transport in Radio Galaxies and Quasars, ed. P. E. Hardee, A. H. Bridle, & J. A. Zensus (San Francisco: ASP), 273 Hardee, P. E., & Clarke, D. A. 1992, ApJ, 400, L9 . 1995, ApJ, 451, L25 Hardee, P. E., Clarke, D. A., & Howell, D. A. 1995, ApJ, 441, 644 Hummel, C. A., Muxlow, T. W. B., Krichbaum, T. P., Quirrenbach, A., Schalinski, C. J., Witzel, A., & Johnston, K. J. 1992a, A&A, 266, 93 Hummel, C. A., et al. 1992b, A&A, 257, 489 Koide, S., Nishikawa, K.-I., & Mutel, R. L. 1996a, ApJ, 463, L71 (KNM)

L48

NISHIKAWA ET AL.
Mutel, R. L., Phillips, R. B., Su, B., & Bucciferro, R. R. 1990, ApJ, 352, 81 Norman, M. L. 1993, in Astrophysical Jets, ed. D. Burgarella, M. Livio, & C. ODea (Cambridge: Cambridge Univ. Press), 211 . 1996, in ASP Conf. Ser. 100, Energy Transport in Radio Galaxies and Quasars, ed. P. E. Hardee, A. H. Bridle, & J. A. Zensus (San Francisco: ASP), 319 Qian, S. J., Quirrenbach, A., Witzel, A., Krichbaum, T. P., Hummel, C. A., & Zensus, J. A. 1991, A&A, 241, 15 Qian, S. J., Witzel, A., Kraus, A., Krichbaum, T. P., & Britzen, S. 1996, in ASP Conf. Ser. 100, Energy Transport in Radio Galaxies and Quasars, ed. P. E. Hardee, A. H. Bridle, & J. A. Zensus (San Francisco: ASP), 55 Stone, J. M., & Norman, M. L. 1992, ApJ, 389, 297 van Putten, M. H. P. M. 1993, ApJ, 408, L21 . 1996, ApJ, 467, L57 Wardle, J. F. C., & Aaron, S. E. 1996, in ASP Conf. Ser. 100, Energy Transport in Radio Galaxies and Quasars, ed. P. E. Hardee, A. H. Bridle, & J. A. Zensus (San Francisco: ASP), 123 Yokosawa, M., Ikeuchi, S., & Sakashita, S. 1982, PASJ, 34, 461

Koide, S., Nishikawa, K.-I., & Mutel, R. L. 1996b, J. Comput. Phys., submitted Koide, S., Sakai, J., Nishikawa, K.-I., & Mutel, R. L. 1996c, ApJ, 464, 724 Komissarov, S. S., & Falle, S. A. E. G. 1996a, in ASP Conf. Ser. 100, Energy Transport in Radio Galaxies and Quasars, ed. P. E. Hardee, A. H. Bridle, & J. A. Zensus (San Francisco: ASP), 165 . 1996b, in ASP Conf. Ser. 100, Energy Transport in Radio Galaxies and Quasars, ed. P. E. Hardee, A. H. Bridle, & J. A. Zensus (San Francisco: ASP), 173 Krichbaum, T. P., Quirrenbach, A., & Witzel, A. 1992, in Variability of Blazars, ed. E. Valtaoja & M. Valtonen (Cambridge: Cambridge Univ. Press), 331 Laing, R. A. 1996, in ASP Conf. Ser. 100, Energy Transport in Radio Galaxies and Quasars, ed. P. E. Hardee, A. H. Bridle, & J. A. Zensus (San Francisco: ASP), 241 Mart J. Ma., Font, J. A., Ibaez, J. Ma., & Muller, E. 1996, in ASP Conf. Ser. , n 100, Energy Transport in Radio Galaxies and Quasars, ed. P. E. Hardee, A. H. Bridle, & J. A. Zensus (San Francisco: ASP), 149 Mart J. Ma., Muller, E., Font, J. A., & Ibaez, J. Ma. 1995, ApJ, 448, L108 , n Mart J. Ma., Muller, E., & Ibaez, J. Ma. 1994, A&A, 281, L9 , n

FIG. 1.(a, b) Run A and (c, d) run B with strong and weak magnetic eld, respectively: XZ cross sections through Y 0 at t 8.5 s. (a, c) Contours of rest mass density and velocity vectors; (b, d) contours of thermal pressure and magnetic eld vectors. The maximum values represented by white color are (a) max 1.94; (b) Pmax 3.29; (c) max 2.17; and (d) Pmax 3.64. NISHIKAWA et al. (see 483, L46)

PLATE L4

PLATE L5
FIG. 2
8.5
s

FIG. 3
from a viewpoint in front and to the side of the jet. The red, blue, and green lines are traced starting at the front

FIG. 2.(a) Run A and (b) run B: eld lines in three-dimensional space at t side (X 19.8) from a radius of (Y 2 Z 2 )1 2 1, 3, and 5, respectively.

FIG. 3.Horizontal (XY) cross section through Z 0 at t 8.5 s for the strongly magnetized run showing the distribution of three different electric currents (JZ ). The currents shown in red blue color recede approach the observer. Two currents circle near the surface of the cylindrical beam, and a third current circles around the bow shock.

NISHIKAWA et al. (see 483, L47)