Magnetic Field Amplification and Saturation in Turbulence Behind A Relativistic Shock

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Mon. Not. R. Astron. Soc.

000, 114 (2013)

Printed 29 January 2014

A (MN L TEX style le v2.2)

Magnetic Field Amplication and Saturation in Turbulence Behind a Relativistic Shock

arXiv:1401.7080v1 [astro-ph.HE] 28 Jan 2014

Yosuke Mizuno1 , Martin Pohl2,3, Jacek Niemiec4 , Bing Zhang5 , Ken-Ichi Nishikawa6 , and Philip E. Hardee7
of Astronomy, National Tsing-Hua University, Hsinchu 30013, Taiwan, Republic of China of Physics and Astronomy, University of Potsdam, 14476 Potsdam, Germany 3 DESY, Platanenallee 6, 15738 Zeuthen, Germany 4 Institute of Nuclear Physics PAN, ul. Radzikowskiego 152, 31-342 Krak ow, Poland 5 Department of Physics and Astronomy, University of Nevada, Las Vegas, NV 89154, USA 6 Department of Physics, University of Alabama in Huntsville, Huntsville, AL 35805, USA 7 Department of Physics and Astronomy, The University of Alabama, Tuscaloosa, AL 35487, USA
2 Institute 1 Institute

Accepted 2014 January 27. Received 2013 December 2; in original form 2013 December 2

ABSTRACT

We have investigated via two-dimensional relativistic MHD simulations the long-term evolution of turbulence created by a relativistic shock propagating through an inhomogeneous medium. In the postshock region, magnetic eld is strongly amplied by turbulent motions triggered by preshock density inhomogeneities. Using a long-simulation box we have followed the magnetic-eld amplication until it is fully developed and saturated. The turbulent velocity is sub-relativistic even for a strong shock. Magnetic-eld amplication is controled by the turbulent motion and saturation occurs when the magnetic energy is comparable to the turbulent kinetic energy. Magnetic-eld amplication and saturation depend on the initial strength and direction of the magnetic eld in the preshock medium, and on the shock strength. If the initial magnetic eld is perpendicular to the shock normal, the magnetic eld is rst compressed at the shock and then can be amplied by turbulent motion in the postshock region. Saturation occurs when the magnetic energy becomes comparable to the turbulent kinetic energy in the postshock region. If the initial magnetic eld in the preshock medium is strong, the postshock region becomes turbulent but signicant eld amplication does not occur. If the magnetic energy after shock compression is larger than the turbulent kinetic energy in the postshock region, signicant eld amplication does not occur. We discuss possible applications of our results to gamma-ray bursts and active galactic nuclei. Key words: (stars:) gamma-ray burst: general - (magnetohydrodynamics) MHD - methods: numerical - relativistic processes - shock waves - turbulence

1 INTRODUCTION Nonthermal emission is observed from many astrophysical sources harbouring relativistic shocks. In general, the composition of the plasma, the Lorentz factor of the shock, and the structure and strength of the preshock magnetic eld are unknown. Radiation modeling of gamma-ray bursts (GRBs) suggests that the magnetic energy density in the emission region constitutes a substantial fraction B 103 101 of the internal energy density (e.g., Panaitescu & Kumar 2002, Yost et al. 2003; Panaitescu 2005, Piran 2005; M esz aros 2006; Santana, Barniol Duran, & Kumar 2013). However, such a high magnetization cannot be attained solely by a simple compressional amplication of the weak magnetic eld

E-mail:[email protected]

pre-existing in the upstream plasma (Gruzinov 2001; Barniol Duran 2013). Magnetic-eld amplication beyond shock compression also seems necessary for emission modeling of young supernova remnants (SNRs), for which magnetic elds as strong as 1 mG have recently been inferred from observations of the thin X-ray rims in several young SNRs (Bamba et al. 2003, 2005a, 2005b; Vink & Laming 2003, but see also Pohl et al. 2005) along with rapid time variation of the synchrotron X-ray emission in RX J1713.7-3946 (Uchiyama et al. 2007, but see also Bykov et al. 2008). Magnetic elds in GRB afterglow shocks can be generated through Weibel and lamentation instabilities (e.g., Medvedev & Loeb 1999), as was demonstrated with Particle-in-cell (PIC) simulations of relativistic collisionless shocks (e.g., Nishikawa et al. 2005, 2009; Spitkovsky 2008). Recent studies have also proved that Weibel-type instabilities operate in sub-relativistic shocks (Kato &

c 2013 RAS

Y. Mizuno et al.
homogeneous interstellar medium (ISM) have been also performed by Inoue et al. (2009, 2012). The simulations indicated strong magnetic-eld amplication in the postshock medium. The peak magnetic-eld strength was found to be more than a hundred times larger than the preshock eld strength. Similar results have been also obtained in recent hybrid (kinetic ions and uid electrons) simulations by Caprioli & Spitkovsky (2013). Fraschetti (2013) has investigated magnetic eld amplication by turbulence generated downstream of a two-dimensional rippled hydromagnetic shock analytically in non-relativistic regime. These results strengthen the case for turbulence being an important contributor to magnetic-eld amplication and emission variability. In an earlier paper (Mizuno et al. 2011b), we demonstrated that the magnetic eld is amplied by the turbulence that develops in the post-shock region behind a relativistic shock propagating through an inhomogeneous medium. Inoue et al. (2011) performed 3D relativistic MHD simulations of a propagating relativistic shock and obtained results similar to ours. However, the growth of the magnetic eld had not saturated in the relatively short time covered by this previous work. In this paper, we continue our investigation and present results from two-dimensional (2.5D) relativistic MHD simulations using a much longer grid. This longer grid permits us to investigate the long-term evolution of turbulence and the saturation of magnetic-eld amplication. We also extend our investigation to non-relativistic in addition to relativistic shock speeds and to a range of magnetizations of the upstream medium. This paper is organized as follows: We describe the numerical method and setup used for our simulations in 2, present our results in 3, and discuss their astrophysical implications in 4. 2 NUMERICAL METHOD AND SETUP We solve the 3D RMHD equations for a mildly relativistic shock propagating in an inhomogeneous medium in two-dimensional Cartesian geometry (x y plane), but follow all three components of the velocity and magnetic eld vectors (so-called 2.5D or 2D3V model) using the 3D GRMHD code RAISHIN (Mizuno et al. 2006, 2011a). For the simulations described here, we have introduced a fth-order weighted essentially non-oscillatory (WENO) scheme. The setup of the code for shock simulations was outlined in Mizuno et al. (2011b). However, in these simulations we use a computational box that is 4 times longer, namely (x, y ) = (8L, L). The numerical resolution is as before with N/L = 256. At x = xmax , the uid which is initially moving with velocity vx = v0 in the positive x-direction is stopped by setting vx = 0 and thermalized 1 . As the pressure increases at x = xmax , a shock forms and propagates in the x-direction. The downstream plasma velocity is thus zero on average. To produce different shock strengths, we choose three different ow speeds, v0 = 0.2c, 0.5c and 0.9c, where c is the speed of light. As in our previous work (Mizuno et al. 2011b), simulations are initialized with an inhomogeneous plasma with mean rest-mass
1 This condition for the boundary is different from a typical reecting boundary, at which vx = vx (e.g., Spitkovsky 2008). The advantage of our approach is that the conversion of kinetic to thermal energy at the boundary closely mimics the uid behaviour behind a shock and in front of the contact discontinuity. Additionally, the created shock propagates slower and we can follow it for a longer time.

Takabe 2008; Niemiec et al. 2012). It is a matter of debate whether magnetic elds thus generated will persist at sufcient strength over the entire emission region, that, for example, in GRBs is estimated to extend over some 106 plasma skin depths downstream of the shock. On larger scales, magnetic elds can be amplied through nonresonant cosmic-ray streaming instabilities in the precursor of nonrelativistic (e.g., Bell 2004; Niemiec et al. 2008; Riquelme & Spitkovsky 2009, 2010; Stroman et al. 2009) and relativistic (e.g., Milosavljevi c & Nakar 2006; Niemiec et al. 2010) shocks, which in the nonlinear phase induce density uctuations. Upstream density uctuations of any origin, e.g., those found in the wind zone of the GRB progenitor (e.g., Ramirez-Ruiz et al. 2005; Sironi & Goodman 2007) or arising from cosmic-ray streaming instabilities (Stroman et al. 2009), can trigger a Richtmyer-Meshkov-type instability that leads to turbulent dynamo processes in the postshock region. Interaction of the shock front with such density uctuations generates a signicant vorticity at the shock. This turbulent plasma motion stretches and deforms magnetic eld lines leading to eld amplication (Sironi & Goodman 2007; Goodman & MacFadyen 2008; Palma et al. 2008). The existence of relativistic turbulence in GRBs has been invoked to explain the observation of large variations in the prompt GRB -ray luminosity as well as intraburst variability in the afterglows. Narayan & Kumar (2009) and Lazar et al. (2009) proposed a relativistic turbulence model instead of the well-known internal shock model to interpret the variable GRB light curves. However, the applicability of these models has recently been challenged by the results of MHD simulations performed by Inoue et al. (2011), who showed that relativistic turbulence decays much faster than the rate of magnetic eld amplication. Zhang & Yan (2011) proposed a new GRB prompt-emission model in the highly magnetized regime, which invokes internal-collision-induced magnetic reconnection and turbulence. Within this model, turbulence could be sustained by continuous reconnection in the energy dissipation region. The short-time variability spikes in GRB lightcurves can be attributed to turbulent reconnection in the magnetic dissipation region, while the long-time variability stems from the activity of the central engine. The blazar zone the innermost part of the relativistic jets in Active Galactic Nuclei (AGN) is probed through multiwavelength observations. Aharonian et al. (2003) and Krawczynski et al. (2004) reported correlated X-ray/TeV -ray ares with timescales from 15 minutes (for Mrk 421) to a few hours (for Mrk 501 and 1ES 1959+650). In the TeV band alone, ux doubling has been observed on timescales down to 2 minutes (Aharonian et al. 2007; Albert et al. 2007; Arlen et al. 2013). Huge Doppler factors D 50 appear required to provide opacities 1 and permit emission regions larger than the Schwarzschild radius of the central black holes. A scenario of fast-moving needles within a slower jet or of a jet-within-a-jet (Levinson 2007; Begelman et al. 2008; Ghisellini & Tavecchio 2008; Giannios et al. 2009) has been invoked to explain the fast variability of blazars. The short-term uctuations can be also understood as a consequence of a turbulent ambient-jet-plasma that passes through shocks in the jet ow (Marscher & Jorstad 2010; Marscher et al. 1992). The potential importance of turbulence to magnetic eld amplication and variability led Giacalone & Jokipii (2007) and Guo et al. (2012) to perform 2D non-relativistic MHD shock simulations involving upstream density and magnetic eld uctuations with a Kolmogorov power spectrum. Two- and three-dimensional MHD simulations of non-relativistic shocks propagating in a cloudy in-

c 2013 RAS, MNRAS 000, 114

Magnetic Field Amplication in a Relativistic Shock


density 0 = 1 containing uctuations established across the entire simulation domain. Following Giacalone & Jokipii (1999, 2007), density uctuations are created by superposing 50 discrete wave modes with wavelengths between min = 0.025L and max = 0.5L. The wave amplitudes are chosen to mimic a twodimensional Kolmogorov-like power-law spectrum given by Pk 1 , 1 + (kL)8/3 (1)
Table 1. Simulation Parameters Case A1 A2 A3 B1 B2 B3 C1 C2 C3 D1 D2 D3 E1 E2 E3 F1 F2 F3 v0 0.5 0.5 0.5 0.5 0.5 0.5 0.9 0.9 0.9 0.9 0.9 0.9 0.2 0.2 0.2 0.2 0.2 0.2 0.0001 0.001 0.01 0.0001 0.001 0.01 0.0001 0.001 0.01 0.0001 0.001 0.01 0.0001 0.001 0.01 0.0001 0.001 0.01 B0 0.01 0.032 0.1 0.0115 0.0365 0.115 0.01 0.032 0.1 0.023 0.0725 0.23 0.01 0.032 0.1 0.01 0.032 0.1 Bn 0 0 0 90 90 90 0 0 0 90 90 90 0 0 0 90 90 90 c s /c 0.0408 0.0408 0.0408 0.0408 0.0408 0.0408 0.0408 0.0408 0.0408 0.0408 0.0408 0.0408 0.0408 0.0408 0.0408 0.0408 0.0408 0.0408
/c vA

where L is the turbulence coherence length. The uctuation variance is 2 = 0.0120 . Note, that in contrast to the method used by Giacalone & Jokipii (2007), our initial preshock turbulence does not include any uctuating magnetic eld (see also Mizuno et al. 2011b). The gas pressure of preshock medium is a constant with p = 0.0010 c2 , which is an order of magnitude lower than the value used in Mizuno et al. (2011b). The shock waves that form may thus have a larger sonic Mach number. An equation of state (EoS) relates the enthalpy h to the gas-pressure and density, and here we use the so-called TM EoS proposed by Mignone et al.(2005): h= 5 + 2 9 2 + 1, 4
2

0.0115 0.0365 0.114 0.01 0.0315 0.1 0.01 0.0315 0.1 0.01 0.0315 0.1 0.0102 0.0322 0.101 0.01 0.0316 0.1

(2)

where p/(c ). The TM EoS is a simple algebraic function of and a good approximation to Synges EoS (Synge 1971) that describes single-component perfect gases in the relativistic regime. The TM EoS corresponds to a lower bound of Taubs fundamental inequality (Taub 1948), i.e., (h )(h 4) = 1. The TM EoS reproduces the correct asymptotic values for the equivalent adiabatic index eq = (h 1)/(h 1 ), i.e. eq 5/3 for nonrelativistic temperatures and eq 4/3 in the ultra-relativistic limit. In our simulations the pre-shock plasma carries a constant mean magnetic eld. To investigate the effect of the initial magnetic eld strength, we choose three different magnetizations in the pre-shock medium, b2 /c2 = 0.0001, 0.001, and 0.01. Here is the density and b is the magnetic eld in the comoving (pre-shock medium) frame. Note that b2 = B2 / 2 + (v B)2 where B is the magnetic eld seen in the simulation frame (Komissarov 1997; Del Zanna et al. 2007). We also consider two different magnetic-eld orientations with respect to the shock normal, a parallel (to the shock normal, Bx , Bn = 0 ) and a perpendicular (By , Bn = 90 ) eld conguration. The sound speed, cs /c = [(5h 8)/3h(h )]1/2 , and the Alfv en speed, vA /c = [b2 /(h + b2 )]1/2 , in the different simulations are calculated using the mean plasma density (0 ) measured in the comoving (pre-shock medium) frame , and are listed in Table 1 along with the ow speeds v0 .

A1-A3). In the following, we refer to the three different plasma magnetizations studied in this paper, = 0.0001, 0.001, and 0.01, as the low-, medium-, and high- cases, respectively. 3.1.1 Global Structure Figures 1 and 2 show 2-D images of the density (Fig. 1) and the total magnetic-eld strength (Fig. 2) at ts = 42, where ts is in units of L/c with c = 1, for the three magnetization parameters studied in case A. As described in Mizuno et al. (2011b), the shock front develops ripples when the inhomogeneous-density preshock plasma encounters the shock. In all cases these ripples lead to strong, random transverse ow behind the shock, thus introducing rotation and vorticity in the postshock region through a process similar to the Richtmyer-Meshkov instability (e.g., Brouillette 2002; Sano et al. 2012; Inoue 2012). The turbulent plasma motions produce the velocity shears along magnetic eld lines that lead to the magneticeld amplication. Note that because our simulations start from pre-existing nite-amplitude density uctuations in the preshock medium, the ow pattern in the postshock region is initially highly nonlinear and comparison with linear Richtmyer-Meshkov instability analysis is not useful. In the low- case ( =0.0001, run A1), the preshock magnetic eld energy density is much less than the postshock turbulent energy density. Thus, the turbulent velocity eld can easily stretch and deform the frozen-in magnetic eld, resulting in eld-amplication. Near the shock front, the vorticity scale size is small, but farther downstream the scale of the vortices increases through an inverse cascade of turbulent eddies, and the magnetic eld is strongly amplied. The turbulent density structure is nearly isotropic because the magnetic eld is weak. The amplied magnetic eld develops a lamentary structure. In the region far behind the shock front at x 6 8, the magnetic-eld strength decreases relative to that at x 3 5, where the highest amplication is observed. This indicates that magnetic-eld amplication via turbulent motion saturates and then tends to decay farther behind the shock (see Fig. 4 for the time evolution of the volume-averaged magnetic eld). In the medium- case ( = 0.001, run A2), the preexisting magnetic eld energy density is still less than the postshock turbu-

3 RESULTS 3.1 Dependence on the Initial Magnetic Field Strength In this section we describe the main characteristics of the turbulence generated when a shock propagates into an inhomogeneous medium, and the dependence of the properties of the system on the preshock plasma magnetization. As a representative example of this dependence (studied here for all the cases considered, see Table 1), we choose the case with the mildly relativistic ow speed of v0 = 0.5c and a parallel conguration for the mean preshock magnetic eld with respect to the shock normal, Bn = 0 (cases
c 2013 RAS, MNRAS 000, 114

Y. Mizuno et al.

Figure 1. Two-dimensional images of density at ts = 42 for magnetization parameters = 0.0001 (upper panel), 0.001 (middle), and 0.01 (lower panel) for case A, with magnetic eld parallel to the shock normal and v0 = 0.5c. White arrows indicate the ow direction in the postshock region.

lent energy density. As in the low- case, the downstream magnetic eld is amplied in the turbulent velocity eld. The magnetic eld is structured in thicker laments than are seen in the low- case. The magnetic laments are aligned along the initial magnetic-eld direction, Bx , after saturation, most likely because magnetic-eld tension resists motion perpendicular to the mean-eld direction. In the high- case ( = 0.01, run A3), magnetic-eld amplication through the turbulent dynamo process is not efcient, even though a turbulent velocity eld develops in the postshock region. The magnetic eld is amplied only by a factor of about 2 relative to the initial magnetic eld. Magnetic lamentary structures seen in the lower- cases are signicantly suppressed at this higher value. Figure 3 shows 1-D cuts along the x-axis at y/L = 0.5 and ts = 42 of the density, the in-plane transverse velocity vy , and the total magnetic eld strength for parallel magnetic eld with magnetization parameters = 0.0001 (black solid), 0.001 (red dotted), and 0.01 (blue dashed) and the mildly relativistic ow velocity v0 = 0.5c (runs A1-A3). The shock front is located at x 0.8L. The left and right sides of the shock front are upstream and downstream regions, respectively. The measured shock propagation speed is about vsh 0.17c in the contact discontinuity frame. Analytic calculations using an ideal gas EoS (see the Appendix in Mizuno et al. 2011b) give a shock velocity in the contactdiscontinuity frame of 0.18c for = 5/3 and 0.09c for = 4/3. The measured shock velocity is thus in good agreement with that expected for = 5/3. Conventionally, the shock velocity and Mach number are given in the upstream rest frame. We

parametrize the shock strength by the relativistic sonic Mach number


Ms sh vsh /s cs , s

(3)

where with the shock Lorentz factor measured in the upstream rest frame. A trivial Lorentz transformation converts the shock propagation speed measured in the simulation to the standard upstream-frame shock speed,
vsh =

2 2 1/2 (1 c is the Lorentz factor associated s /c ) sound speed and vsh and sh are the shock speed and the

vsh + v0 . 1 + vsh v0 /c2

(4)

The shock propagation speed of vsh = 0.17c obtained from the simulations thus corresponds to vsh 0.82c in the upstream ow frame. This leads to Ms 19 for the sound speed in the preshock region where c s 0.04c. In all cases, the density jumps by about a factor of 4, which is close to the strong-shock limit in the Newtonian approach. The transverse-velocity is strongly uctuating. The maximum transverse velocity is about 0.04c, and in all cases the average root mean square turbulent velocity of 0.02c is subsonic in the postshock region ( cs 0.35c). The total-magnetic-eld also shows strong variation. The magnetic eld is not compressed at the shock because the direction of the initial magnetic eld is parallel to the shock normal. In the low- case ( = 0.0001), the local magnetic eld reaches nearly 8 times the amplitude of the initial eld. When the initial magnetic eld is larger, magnetic-eld amplication is reduced. In the medium- case ( = 0.001), the maxic 2013 RAS, MNRAS 000, 114

Magnetic Field Amplication in a Relativistic Shock

Figure 2. Two-dimensional images of the total magnetic eld normalized to the initial magnetic eld strength, B0 , at ts = 42.0 for magnetization parameters = 0.0001 (upper panel), 0.001 (middle), and 0.01 (lower panel). As in Fig. 1 the magnetic eld is parallel to the shock normal and v0 = 0.5c (case A).

mum amplitude of the amplied magnetic eld is Btot /B0 4, whereas in the high- case ( = 0.01), the local magnetic eld in the postshock region reaches only 2 times the initial magnetic eld strength. The Alfv en velocity in the postshock region depends on the initial mean magnetic eld strength. In the low- case, the Alfv en velocity in the postshock region uctuates strongly and the average Alfv en velocity is vA 0.01c. Turbulence is superAlfv enic in most of the postshock region in the low- case. This result is consistent with earlier non-relativistic studies (e.g., Giacalone & Jokipii 2007; Inoue et al. 2009; Guo et al. 2012). When the mean magnetic eld is larger, the Alfv en velocity in the postshock region is larger. In the middle- and high- cases, the average Alfv en velocities in the postshock region are vA 0.03c and 0.06c respectively. Postshock turbulence in the medium- and high cases is sub-Alfv enic.

3.1.2 Magnetic Field Amplication & Saturation Previous results showed that magnetic-eld amplication via the turbulent dynamo process depends on the initial magnetic-eld strength. Figure 4 shows the time evolution of the volume-averaged total magnetic eld (Fig. 4a) and the maximum total magnetic-eld strength (Fig. 4b) in the postshock region for cases A1-A3. We continuously check the shock position for each y -coordinate and average the absolute magnetic-eld strength from the y-dependent shock position to x = xmax . The region over which the average is taken thus explicitly depends on time and the case in study.
c 2013 RAS, MNRAS 000, 114

All cases are normalized by the initial magnetic-eld strength, B0 , which varies with the assumed . In the low- case ( = 0.0001), the average postshock magnetic eld gradually increases with time, saturates at ts 20, and then decreases slightly to assume a constant value for ts 30. The postshock magnetic eld is amplied by about a factor of 2.5 at saturation. The peak eld strength is much larger than the mean eld amplitude and about 16 times larger than the initial magnetic-eld strength. In the medium- case ( = 0.001), the mean postshock eld also gradually increases with time, but saturates at ts 10, sooner than in the low- case, and is amplied by only a factor of 1.3. The peak eld strength is about 7 times larger than the initial magnetic-eld strength. In the high- case ( = 0.01), the mean magnetic eld does not become stronger with time, thus the average magnetic eld is not amplied in this case. The peak eld in the postshock region is about 2 times larger than the initial magnetic eld, suggesting some localized eld amplication. These results show that the efciency of magnetic-eld amplication declines as the magnetization ( ) increases.

3.1.3 The Kinetic to Magnetic Energy Density Ratio Figure 5 shows 2-D images of the ratio of the kinetic to the magnetic energy at ts = 42 for cases A1-A3. The kinetic and magnetic energy densities are dened as Ekin = ( 1)c2 and Emag = B 2 + [v 2 B 2 (v B )2 ]/2, respectively. In the low- case, the kinetic energy dominates near the shock front (x/L 2). In an intermediate region farther downstream from the shock at

Y. Mizuno et al.

Figure 4. Time evolution of (a) the volume-averaged total magnetic eld and (b) the maximum total magnetic-eld strength in the postshock region normalized by B0 for the case of mildly relativistic ow velocity (v0 = 0.5c) and parallel mean magnetic eld conguration (runs A1-A3). Different lines are for different initial magnetic eld strength: = 0.0001 (solid lines), = 0.001 (dotted lines), and = 0.01 (dashed lines).

Figure 3. One-dimensional cuts along the x-direction of (a) the normalized density (/0 ), (b) the transverse velocity (vy ), (c) the root-mean-square velocity averaged in the y direction (vrms ) and (d) the total normalized magnetic eld strength (Btot /B0 ), at y/L = 0.5 and ts = 42.0 for parallel magnetic eld cases with = 0.0001 (black solid lines), = 0.001 (red dotted lines), and = 0.01 (blue dashed lines) and mildly relativistic ow velocity v0 = 0.5c (cases A1-A3). The shock front is located at x/L 0.8. The left and the right sides of the shock front are upstream and downstream regions, respectively.

2 x/L 5, the magnetic energy becomes dominant as the magnetic eld is amplied. Far behind the shock front (x/L 5), where the eld amplitude has saturated, the magnetic energy dominates in most locations. Saturation occurs when the magnetic energy density becomes comparable to the turbulent kinetic energy density, in agreement with previous MHD studies (e.g., Schekochihin & Cowley 2007; Cho et al. 2009; Zhang et al. 2009; Inoue et al. 2011). For cases A1-A3 with mildly relativistic ow speed, v0 = 0.5c, and parallel magnetic eld conguration, the root mean square turbulent veloc2 ity is vrms 0.02c (see Fig. 3c). The average kinetic vturb 2 energy density can be estimated from Ekin d vrms /2 4 8 10 , where d 4 is the average density in the post-

shock (downstream) region. Here we use the Newtonian approximation because the turbulence is not relativistic. If the magnetic eld is amplied to the limit, i.e., the magnetic energy density becomes comparable to the turbulent kinetic energy density, then the estimated magnetic eld strength at saturation in the laboratory (contact-discontinuity) frame is Bsat,est 2Ekin 0.04. In the low- case A1 the average magnetic eld strength in the saturation region is Bsat,sim 0.035. This simulation result is in good agreement with the estimate. A similar result is found for the medium- case A2, and at saturation the magnetic energy density is comparable to the turbulent kinetic energy density in this case as well. In the high- case A3, the estimated magnetic-eld saturation level is lower than the initial eld strength, even for the maximum turbulent velocity. Therefore, signicant eld amplication does not occur in this case. subsubsectionTurbulent Magnetic and Kinetic Energy Power Spectra The statistical properties of turbulent uctuations in the postshock region can be determined from their power spectra. Figure 6 shows spherically-integrated kinetic and magnetic energy spectra for cases A1-A3. As found by Mizuno et al. (2011b), the kinetic-energy spectra are only slightly atter than Kolmogorov, i.e., Ekin (k) k(5/3)(D 1) with D = 2 in two-dimensional systems. The kinetic-energy power spectra do not change significantly with time and are almost the same for all -cases. A Kolmogorov-like kinetic energy power spectrum seems to be an inherent property of the postshock turbulence produced by the interaction of the shock front with the upstream density inhomogeneities, since this power spectrum is observed both in studies that assume a Kolmogorov (e.g., Mizuno et al. 2011b) and also a non-Kolmogorov (e.g., Inoue et al. 2009) power spectrum in the upstream uctuations. In all cases, the magnetic energy power spectrum amplitude rapidly increases at early simulation times and the shape remains almost constant at later times, implying that magnetic-eld amplication has reached saturation. The largest enhancement in the magnetic energy power spectrum occurs for the low- case and this reects a larger amplication than for the higher -cases. Consistent with Mizuno et al. (2011b), the magnetic energy spectra at large scales (k 50) are almost at and strongly deviate from the Kolmogorov spectrum. Such spectra are typical of the small-scale dynamo process (Kazantsev 1968). Flat magnetic-energy spectra are produced in turbulent-dynamo simulations (e.g., Schekochihin
c 2013 RAS, MNRAS 000, 114

Magnetic Field Amplication in a Relativistic Shock

Figure 5. Two-dimensional images of the kinetic to magnetic energy density ratio at ts = 42.0 for magnetization parameters, = 0.0001 (upper), 0.001 (middle), and 0.01 (lower) with mean magnetic eld parallel to the shock normal and v0 = 0.5c (cases A1-A3).

et al. 2004; Brandenburg & Subramanian 2005). The same properties are also observed in simulations of driven super-Alfv enic turbulence (e.g., Cho & Lazarian 2003) and in relativistic MHD turbulence simulations (e.g., Zhang et al. 2009; Inoue et al. 2011). Note that the magnetic energy spectrum in the low- and medium cases is at over a broader region than the high- case. This is again an indication of larger magnetic-eld amplication for lower magnetizations.

magnetic eld. This difference is a result of the shock compression of the magnetic eld that does not occur for a parallel eld conguration, and the stronger postshock elds inuence on the turbulent motion. In the medium- case, after the initial shock compression, the magnetic eld is only moderately amplied in the postshock region. It appears that the compressed magnetic eld is too strong to allow a signicant deformation of the eld lines by the turbulent velocity eld. For the same reason, in the high- case basically only shock compression is observed. One-dimensional cuts along the x-axis at y/L = 0.5 and ts = 42 (ts = 38 for the = 0.01 case) for cases B1-B3 are shown in Figure 8. The shock front is located at x/L = 0.7 0.9. For the perpendicular magnetic eld conguration, the shock propagation speed depends slightly on the pre-shock plasma magnetization, and as becomes larger, the shock velocity increases. Thus for the low and medium the shock propagation speed is vsh 0.17c in the contact discontinuity frame, the same as for the parallel initial eld case. In the high- case, the shock velocity is slightly faster, vsh 0.18c. The density jumps by about a factor of 4 in all cases. The transverse velocity proles show strong velocity uctuations with a similar root mean square turbulent velocity of vrms 0.02c. This is comparable to the root mean square turbulent velocity obtained for a parallel preshock eld. The level of magnetic eld shock-compression depends on the magnetization. For low , the magnetic eld is amplied by about factor of 4 but the amplication is lower when the magnetization is larger. For high the amplication factor is only 3.5. The amplitude of the magnetic eld uctuations becomes smaller as becomes larger. Figure 9 shows the time evolution of the volume-averaged to-

3.2 Dependence on Initial Magnetic Field Direction Figure 7 shows 2-D images of the total magnetic-eld strength for the three different magnetization parameters and a mildly relativistic ow velocity v0 = 0.5c but now for mean magnetic eld perpendicular to the shock normal, Bn = 90 (cases B1-B3). The low- and medium- runs are shown at ts = 42, whereas the high- case is shown at ts = 38. When the initial eld direction is perpendicular to the shock normal, the magnetic eld in the postshock region is shock-compressed by about a factor of 4 (see the 1D plot in Fig. 8). In all cases, turbulence develops in the postshock region as the relativistic shock passes through the inhomogeneous preshock medium. In the low- case, after shock-compression, the magnetic eld is still weak enough to be twisted up by the turbulent motion. Therefore, efcient magnetic-eld amplication via the turbulent dynamo occurs and the magnetic eld develops lamentary structure similar to that seen for the case with parallel mean magnetic eld. However, the laments are thicker in this case and aligned in the y -direction (the initial magnetic-eld direction). Turbulent structure is not isotropic as it was in the low- case with parallel
c 2013 RAS, MNRAS 000, 114

Y. Mizuno et al.

Figure 6. Spherically-integrated power spectra of (a, c, e) the kinetic energy, (b, d, f) the magnetic energy in the postshock region for parallel magnetic eld with = 0.0001 (upper panel), = 0.001 (middle), and = 0.01 (lower panel) and mildly relativistic ow v0 = 0.5c (cases A1-A3). Different lines denote the spectra generated at different times: ts = 3 (solid), 8 (dashed), 13 (dotted), and 17 (dash-dotted). A short dotted line, representing the 2D Kolmogorov power law E (k ) k 8/3 and a short dashed line following k 1/4 are shown for comparison in all the panels.

c 2013 RAS, MNRAS 000, 114

Magnetic Field Amplication in a Relativistic Shock

Figure 7. The same as Fig. 2 but for magnetic eld perpendicular to the shock propagation direction (cases B1-B3) at simulation time ts = 42 for = 0.0001 (upper panel) and = 0.001 (middle panel), and ts = 38 for = 0.01 (lower panel).

tal magnetic eld and the peak total magnetic-eld strength in the postshock region for cases B1-B3. The volume-averaged strength of the magnetic eld increases at the shock by about a factor of 4 for both the low and medium cases and by 3.5 for the high- case. The magnetic eld is subsequently amplied by turbulence. In the low- case, the volume-averaged magnetic eld increases up to Btot /B0 5 and saturates at about ts 6. This saturation value is larger than in the low- case with parallel mean magnetic eld. For the medium case, magnetic-eld amplication by turbulent motion is not signicant. The volume-averaged eld strength increases to Btot /B0 4.1 in the initial shock compression, and maintains this value farther behind the shock. In the high- case, no turbulent amplication of magnetic eld is seen. The behaviour in the medium- and high- cases (B2-B3) is similar to the behaviour seen in the strong parallel magnetic eld cases (A2-A3). The peak value of the magnetic eld is highly variable with time and reaches about 20 times the initial magnetic-eld strength for the low case B1. For comparison, the peak value of the magnetic eld was 15 in the low- case A1 with parallel magnetic eld. In the medium- case B2, the peak value is Btot,max /B0 10. This is larger than the maximum of about 6 times the preshock magnetic eld strength seen in the medium- parallel magnetic eld case A2. For the high case B3, the local maximum value
c 2013 RAS, MNRAS 000, 114

Btot,max /B0 5 is only slightly larger than the volume-averaged strength, again reecting the fact that magnetic-eld amplication is weak in this case (see Fig. 7c). In Figure 10 we plot the time evolution of the root mean square uctuation amplitudes of Bx and By , normalized to the mean magnetic-eld strength in the postshock region (after shock compression for perpendicular cases) to investigate the level of magnetic-eld amplication in the turbulent dynamo process for the cases with mildly relativistic ow v0 = 0.5c and both magnetic eld orientations (cases A and B). In all cases, the uctuation amplitudes increase and saturate. For parallel shocks (Fig. 10a), By initially increases much faster than Bx , but at saturation Bx is larger than By . For perpendicular shocks (Fig. 10b), the uctuations Bx and By grow similarly and after saturation have about the same strength. These results indicate that the orientation of the upstream magnetic eld relative to the shock normal signicantly affects magnetic-eld amplication. In the perpendicular case, the magnetic eld is enhanced rst by shock compression and then by turbulence in the postshock region. The total magnetic-eld gain is thus larger in all perpendicular cases than in parallel magnetic eld cases. These results are consistent with our earlier study Mizuno et al. (2011b).

10

Y. Mizuno et al.

Figure 10. Time evolution of the volume-averaged root mean square (rms) uctuation amplitudes of Bx (solid lines) and By (dashed lines) in the postshock region, normalized to the mean magnetic-eld strength in the postshock region. Panel (a) is for magnetic eld parallel, and panel (b) for magnetic eld perpendicular to the shock propagation direction with = 0.0001 (black lines), = 0.001 (red lines), and = 0.01 (blue lines).

3.3 Dependence on Shock Strength Figures 11 and 12 show 1-D cuts along the x-axis at y/L = 0.5 for simulations with parallel magnetic eld and different ow speeds. In particular, Figure 11 shows results for a highly relativistic ow (v0 = 0.9c) at ts = 24 (cases C1-C3), and Figure 12 shows results for a sub-relativistic ow velocity (v0 = 0.2c) at ts = 98 (cases E1-E3). For highly relativistic inow, the shock propagation speed in the contact discontinuity frame, vsh 0.3c, is considerably faster than for the mildly relativistic speed, v0 = 0.5c, cases. For high inow velocity, v0 = 0.9c, Eq. 4 gives vsh 0.945c. The sound speed in the preshock region is c 0.04c s and the relativistic Mach number of the shock is Ms 70 with sh 3 (see Eq. 3). The density jump in the contact discontinuity frame, observed to be a factor 9 in this simulation (Fig. 11b), is close to that expected for a strong relativistic shock (Blandford & McKee 1976), nd = 4 0 9 , (5) nu where 0 is the relative Lorentz factor between the upstream and downstream frames. The continuity condition mandates
vsh sh nu = vsh sh nd ,

Figure 8. The same as in Fig. 3, but for magnetic eld perpendicular to the shock propagation direction (cases B1-B3) with = 0.0001 (black solid lines), = 0.001 (red dotted lines), and = 0.01 (blue dashed lines).

(6)

Figure 9. The same as in Fig. 4, but for magnetic eld perpendicular to the shock propagation direction (cases B1-B3) with = 0.0001 (solid lines), = 0.001 (dotted lines), and = 0.01 (dashed lines).

which is fullled for vsh 0.3c. For cases C1-C3, the transverse velocity uctuates strongly with vrms 0.05c (Fig. 11b). These velocity uctuations in the postshock region are sub-relativistic, even though the shock is very strong. The magnetic eld is strongly amplied locally when the initial magnetic eld is weak (low ). In this case the eld amplitude achieves more than 15 times the initial magnetic eld strength. Magnetic-eld amplication is reduced when the preshock eld is stronger (higher ). This trend is similar to that observed for the mildly relativistic inow cases A1-A3. For sub-relativistic ows, the shock propagation speed in the simulation frame is vsh 0.07c, corresponding to vsh 0.27c, and leads to a shock Mach number Ms 6.8. The density jumps by only a factor of 3.8, because we are not in the strong-shock limit. In the downstream region, the transverse velocity strongly uctuates but the maximum is less than 0.02c. The magnetic eld in the postshock region is amplied by a factor of 5 above the initial eld strength. Magnetic eld amplication is weaker than found in the mildly relativistic ow cases. Figure 13 shows the time evolution of the volume-averaged total magnetic eld and the peak total magnetic-eld strength in
c 2013 RAS, MNRAS 000, 114

Magnetic Field Amplication in a Relativistic Shock

11

Figure 11. The same as in Fig. 3 but for highly relativistic ow (cases C1C3) at time ts = 24.

Figure 12. The same as in Fig. 3 but for sub-relativistic ow (cases E1-E3) at time ts = 98.

the postshock region for parallel and perpendicular initial magnetic eld congurations, with highly relativistic ow, v0 = 0.9c, for the three different initial magnetizations (cases C and D). Magnetic-eld amplication via turbulent motion happens only when the initial magnetic eld is weak. The magnetic energy at saturation is comparable to the turbulent kinetic energy, and magnetic-eld amplication in the highly relativistic ow cases is stronger than in the mildly relativistic ow cases (see Figs. 4 and 9) because the stronger shock leads to higher turbulent velocity in the postshock region. For perpendicular shocks with all magnetizations considered, shock compression of the eld is more efcient than the turbulent amplication downstream. For sub-relativistic ows with v0 = 0.2c (cases E and F), time evolution of the volume-averaged (mean) total magnetic eld and the maximum total magnetic-eld strength in the postshock region is shown in Figure 14. Weaker turbulence in the postshock region provides less efcient magnetic-eld amplication in these cases. For parallel magnetic eld cases, the maximum amplication factor, achieved with a low , is at most Btot /B0 1.5. For perpendicular magc 2013 RAS, MNRAS 000, 114

netic eld cases we see little amplication beyond the effects of the shock compression, and even shock compression is reduced when is larger. The local maximum magnetic-eld strength reaches about 10 times the initial eld strength for the low case. Figure 15 shows the time evolution of the root mean square uctuation amplitudes of Bx and By , normalized to the mean magnetic eld in the postshock region, for three different magneticeld strengths in the highly relativistic (Fig. 15 a-b) and subrelativistic (Fig. 15 c-d) ow cases. As for mildly relativistic ows (Fig. 10), the uctuation amplitude rst increases and then saturates in all cases. In parallel shocks, By builds up earlier than Bx , but at the time of saturation, Bx is larger than By . In perpendicular shocks, the two magnetic-eld components grow and saturate similarly. In general, the relative amplitude of eld uctuations increases with shock strength. In summary, as expected our results show that magnetic eld amplication strongly depends on the shock strength. A stronger shock leads to a larger density jump, higher turbulent velocity in the postshock region, and a stronger magnetic eld at saturation.

12

Y. Mizuno et al.

Figure 13. Time evolution of (panels a and c) the volume-averaged total magnetic eld and (panels b and d) the maximum total magnetic-eld strength in the postshock region for a highly relativistic ow velocity, v0 = 0.9c, with parallel (upper) and perpendicular (lower) magnetic eld congurations (cases C and D, respectively). Lines indicate different magnetic eld strengths: = 0.0001 (solid lines), = 0.001 (dotted lines), and = 0.01 (dashed lines).

Figure 15. Time evolution of the volume-averaged root mean square (rms) uctuation amplitudes of Bx (solid lines) and By (dashed lines) in the postshock region, normalized to the mean postshock magnetic eld, for magnetic eld parallel (a, c) and perpendicular (b, d) to the shock propagation direction with = 0.0001 (black lines), = 0.001 (red lines), and = 0.01 (blue lines) for highly relativistic ow velocity (upper panels) and sub-relativistic ow velocity (lower panels).

4 SUMMARY AND DISCUSSION Using a simulation domain of unprecedented length in the ow direction, we have performed two-dimensional relativistic MHD simulations of a relativistic shock propagating through an inhomogeneous medium. Consistent with previous relativistic (Mizuno et al 2011b; Inoue et al. 2011) and non-relativistic studies (Giacalone & Jokipii 2007; Inoue et al. 2009; Inoue & Inutsuka 2012; Guo et al. 2012), the postshock magnetic eld is amplied through turbulent motions. The amplied magnetic eld assumes lamentary structures, and its power spectrum is atter than Kolmogorov, which is typical for a turbulent dynamo process. We nd that the saturation level of turbulent eld amplication depends on the initial magnetic-eld strength. If the initial eld is strong, the postshock region becomes turbulent but eld amplication does not occur. At a perpendicular shock, the magnetic eld is rst compressed at the shock and then amplied by turbulent motion in the postshock region. The total eld enhancement is larger than at parallel shocks, as observed by Mizuno et al. (2011b). Generally, saturation occurs when the magnetic energy becomes comparable to the turbulent kinetic energy in the postshock region. This implies that magnetic-eld amplication via turbulent motion happens only if the magnetic energy after shock compression is smaller than the turbulent kinetic energy in the postshock region. In our simulations the turbulent velocity in the postshock region is sub-relativistic and subsonic, even for a strong relativistic shock. All our simulations employ comparable small-amplitude ( 2 /0 = 0.012) density uctuations in the pre-shock medium. We note that analytical work by Sironi & Goodman (2007) (see also Goodman & MacFadyen 2008) indicates that the energy density of vortical motions, generated in an interaction bec 2013 RAS, MNRAS 000, 114

Figure 14. Time evolution of (panels a and c) the volume-averaged total magnetic eld and (panels b and d) the maximum total magnetic-eld strength in the postshock region for sub-relativistic ow velocity, v0 = 0.2c, with parallel (upper) and perpendicular (lower) magnetic eld congurations (cases E and F, respectively). Lines indicate different magnetic eld strengths: = 0.0001 (solid lines), = 0.001 (dotted lines), and = 0.01 (dashed lines).

Magnetic Field Amplication in a Relativistic Shock


tween an ultra-relativistic shock and a small-amplitude density uctuation, increases with the square of the density contrast (i.e., the postshock turbulent velocity grows linearly with the density uctuation amplitude). Thus, larger density perturbations lead to greater deformations of the shock front and produce stronger vorticity in the postshock region. As a result, the magnetic eld can be more efciently amplied. This analytical prediction has been recently conrmed in MHD relativistic shock simulations by Inoue et al. (2011) (see also Guo et al. 2012 for the nonrelativistic case), who showed that the rate of the initial exponential growth of postshock magnetic turbulence is proportional to the amplitude of the velocity uctuations. They also found that the shock-induced velocity dispersion can approach the postshock sound speed if the shock propagates in a medium with large-amplitude density inhomogeneities. However, it is not clear whether supersonic relativistic turbulence can be produced and maintained downstream of the shock. For example, in simulations initialized with turbulence by Inoue et al. (2011), the kinetic energy in relativistic turbulence decayed much faster than kinetic energy in transonic turbulence due to dissipation of relativistic turbulence into internal energy via shocklets. It remains to be veried that the same result applies to the case of turbulence driven by ultra-relativistic shock propagation through an upstream medium containing strong density contrasts. In this paper we have performed simulations in twodimensional geometry to take long simulation region in shock propagation direction in order to follow the saturation of magneticeld amplication by turbulence in postshock region which did not achieved previous paper (Mizuno et al. 2011b). In general the turbulence structure, for example the slope of the Kolmogorov-like power spectrum, is different in two and three dimensions. Although we expect that the possible difference is not signicant (see Inoue et al. 2009), to more realistically analyze three-dimensional phenomena we will extend the current investigation to three-dimensional simulations in future work. In GRBs, afterglow modeling suggest that the ratio of the magnetic energy density to the internal energy density, B , has a broad range of values in the emission region. Our simulation results imply that the broad range suggested by the observations is indicative of different properties in the circumburst medium, e.g., different amplitude of density uctuations, and magnetic-eld strength and direction, notwithstanding differences in shock strength. The strong variability in the prompt emission phase requires maintenance of strong relativistic turbulence in the emission region. If prompt emission were due to jet-medium interaction, then this would require a weak medium magnetic eld strength. More likely, GRB prompt emission is of an internal origin, which may invoke collisions of highly magnetized shells (Zhang & Yan 2011). Our simulations in this paper do not apply to this regime, and future work is needed to test whether substantial reconnection-driven turbulence can be generated and maintained in a magnetically dominated ow. In AGN blazars, multi-waveband monitoring often nds rapidly variable gamma-ray ares that typically accompany lesser variability and/or increase in the emission at other wavebands. Multi-wavelength light curves of gamma-ray bright blazars reveal strong correlations across wavebands, yet striking dissimilarities in the details. The linear polarization tends to be highly variable in both degree and position angle, which implies that the magnetic eld is turbulent (Marscher 2013). In general, the emission characteristics based on a turbulent plasma crossing a standing conical shock associated with the millimetre-wave core in VLBA images of blazars agrees with the characteristics of multi-waveband light curves and polarization variations (Marscher 2011). In particular,
c 2013 RAS, MNRAS 000, 114

13

multi-waveband monitoring of AO 0235+164 and OJ 287 indicates gamma-ray ares associated with enhanced emission from the 43 GHz radio core that is consistent with this turbulent cell model for variability (Marscher & Jorstad 2010; Agudo et al. 2011b, 2011c). In these sources the radio core is located at considerable distance (parsecs) from the central engine, AO 0235+164 (Agudo et al. 2012) and OJ 287 (Agudo et al., 2011a). Our simulation results suggest that strong density inhomogeneities in the preshock relativistic jet uid would lead to the development of strong turbulence in the post recollimation shock region from which the observed emission originates. The highest photon energies would come from the smallest regions containing the highest magnetic elds and exhibit the most rapid time variability. Structure in the preshock magnetic eld would inuence the ordering of the postshock turbulent eld depending on the magnetization of the preshock jet uid

ACKNOWLEDGMENTS This work has been supported by NSF awards AST-0908010, and AST-0908040 to UA and UAH, AST-0908362 to UNLV and NASA awards NNX08AG83G and NNX12AH06G to UAH. Y.M. acknowledges support from Taiwan National Science Council award NSC 100-2112-M-007-022-MY3. The work of J.N. has been supported by the Polish National Science Centre through projects DEC-2011/01/B/ST9/03183 and DEC-2012/04/A/ST9/00083. M.P. acknowledges support by the Helmholtz Alliance for Astroparticle Physics, HAP, funded by the Initiative and Networking Fund of the Helmholtz Association. The simulations were performed on the Columbia and Pleiades Supercomputer at the NAS Division of the NASA Ames Research Center, the SR16000 at YITP in Kyoto University, and the Nautilus at the National Institute for Computational Sciences in the XSEDE project supported by National Science Foundation.

REFERENCES Agudo I. et al., 2011a, ApJ, 726, L13 Agudo I. et al., 2011b, in Proc. 2011 Fermi Symposium, eConf C110509, arXiv:1110.6463 Agudo I. et al., 2011c, ApJet al., 735, L10 Agudo I. et al., 2012, Int. J. Mod. Phys. Conf. Ser., 08, 271 Aharonian F. et al., 2003, A&A, 410, 813 Aharonian F. et al., 2007, ApJ, 664, L71 Albert J. et al., 2007, ApJ, 669, 862 Arlen T. et al., 2013, ApJ, 762, 92 Bamba A., Yamazaki R., Ueno M., & Koyama K., 2003, ApJ, 589, 827 Bamba A., Yamazaki R., Yoshida T., Terasawa T., & Koyama K., 2005a, ApJ, 621, 793 Bamba A., Yamazaki R., & Hiraga J.S., 2005b, ApJ, 632, 294 Barniol Duran R., 2013, arXiv: 1311.1216 Begelman M.C., Blandford R.D., & Rees M.J., 2008, MNRAS, 384, L19 Bell A.R., 2004, MNRAS, 353, 550 Blandford R.D., & McKee C.F., 1976, Physics of Fluids, 19, 1130 Brandenburg A. & Subramanian K., 2005, Phys. Rep., 417, 1 Brouillette M., 2002, Annu. Rev. Fluid Mech., 34, 445 Bykov A.M., Uvarov Y.A., & Ellison D.C., 2008, ApJ, 689, L133 Caprioli D., & Spitkovsky A., 2013, ApJ, 765, L20 Cho J. & Lazarian A., 2003, MNRAS, 345, 325

14

Y. Mizuno et al.
Riquelme M.A., & Spitkovsky A., 2010, ApJ, 717, 1054 Riquelme M.A., & Spitkovsky A., 2009, ApJ, 694, 626 Sano T., Nishihara K., Matsuoka C., & Inoue T., 2012, ApJ, 758, 126 Santana R., Barniol Duran R., & Kumar P., 2013, arXiv:1309.3277 Schekochihin A., & Cowley S., 2007, in Magnetohydrodynamics - Historical Evolution and Trends, ed S. Molokov, R. Moreau, & H. Moffatt (Berlin: Springer) 85 Schekochihin A.A., Cowley S.C., Taylor S.F., Maron J.L., & McWilliams J.C., 2004, ApJ, 612, 276 Sironi L., & Goodman J., 2007, ApJ, 671, 1858 Spitkovsky A., 2008, ApJ, 682, L5 Stroman T., Pohl M., & Niemiec J., 2009, ApJ, 706, 38 Synge J.L., 1957, The Relativistic Gas (Amsterdam: NorthHolland) Taub A.H., 1948, Phys. Rev., 74, 328 Uchiyama Y., et al., 2007, Nature, 449, 576 Vink J., & Laming J.M., 2003, ApJ, 584, 758 Yost S.A., Harrison F.A., Sari R., & Frail D.A., 2003, ApJ, 597, 459 Zhang W., MacFadyen A., & Wang P., 2009, ApJ, 690, L40 Zhang B., & Yan H., 2011, ApJ, 726, 90
A This paper has been typeset from a TEX/ L TEX le prepared by the author.

Cho J., Vishniac E.T., Beresnyak A., Lazarian A., & Ryu D., 2009, ApJ, 693, 1449 Del Zanna L., Zanotti O., Bucciantini N., & Londrillo P., 2007, A&A, 473, 11 Fraschetti F. 2013,, ApJ, 770, 84 Ghisellini G., & Tavecchio F., 2008, MNRAS, 386, L28 Giacalone J., & Jokipii J.R., 1999, ApJ, 520, 204 Giacalone J., & Jokipii,J.R., 2007, ApJ, 663, L41 Giannios D., Uzdensky D.A., & Begelman M.C., 2009, ApJ, 395, L29 Goodman J. & MacFadyen A., 2008, J. Fluid. Mech., 604, 325 Gruzinov A., 2001, ApJ, 563, L15 Guo F., Li S., Giacalone J., Jokipii J.R., & Li D., 2012, ApJ, 747, 98 Inoue T., 2012, ApJ, 760, 43 Inoue T., & Inutsuka S., 2012, ApJ, 759, 35 Inoue T., Yamazaki R., & Inutsuka S., 2009, ApJ, 695, 825 Inoue T., Asano K., & Ioka K., 2011, ApJ, 731, 77 Inoue T., Yamazaki R., Inutsuka S., & Fukui Y., 2012, ApJ, 744, 71 Kato T.N., & Takabe H., 2008, ApJ, 681, L93 Kazantsev A.P., 1968, Sov. Phys.-JTEP Lett., 26, 1031 Komissarov S.S., 1997, Phys. Lett. A, 232, 435 Krawczynski H. et al., 2004, ApJ, 601, 151 Lazar A., Nakar E., & Piran T., 2009, ApJ, 695, L10 Levinson A., 2007, ApJ, 671, L29 Marscher A.P., 2011, in Proc. Fermi Symposium, eConf C110509, arXiv:1110.6463 Marscher A.P., 2013, AAS, 221, 339.53 Marscher A.P., & Jorstad S.G., 2010, in Proc. Fermi meets Jansky - AGN in Radio and Gamma-ray, ed. T. Savolainen et al. 171, arXiv:1005.5551 Marscher A.P., Gear W.K., & Travis J.P., 1992, in Blazar Variability, ed. E. Valtaoja & M. Valtonen (Cambridge Univ. Press), 85 Marscher A.P., et al., 2010, ApJ, 630, L5 Medvedev M.V. ,& Loeb A., 1999, ApJ, 526, 697 M esz aros P., 2006, Rep. Prog. Phys. 69, 2259 Mignone A., Plewa T., & Bodo G., 2005, ApJS, 160, 199 Milosavljevi c M., & Nakar E., 2006, ApJ, 651, 979 Mizuno Y., Nishikawa K.-I., Koide S., Hardee P., & Fishman G.J., 2006, arXiv astro-ph, 0609004 Mizuno Y., Hardee P.E., & Nishikawa K.-I., 2011a, ApJ, 734, 19 Mizuno Y., Pohl M., Niemiec J., Zhang B., Nishikawa K.-I., & Hardee P.E. 2011b, ApJ, 726, 62 Narayan R., & Kumar P., 2009, MNRAS, 394, L117 Niemiec J., Pohl M., Stroman T., & Nishikawa K.-I., 2008, ApJ, 684, 1174 Niemiec J., Pohl M., Bret A., & Stroman T., 2010, ApJ, 709, 1148 Niemiec J., Pohl M., Bret A., & Wieland V., 2012, ApJ, 759, 73 Nishikawa K.-I., Hardee P., Richardson G., Preece R., Sol H., & Fishman G.J., 2005, ApJ, 622, 927. Nishikawa K.-I., et al., 2009, ApJ, 698, L10 Palma G., Mignone A., Vietri M., & Del Zanna L., 2008, ApJ, 686, 1103 Panaitescu A., 2005, MNRAS, 363, 1409 Panaitescu A., & Kumar P., 2002, ApJ, 571, 779 Parker E.N., 1971, ApJ, 163, 255 Piran T., 2005, Rev. Mod. Phys., 76, 114 Pohl M., Yan H., & Lazarian A., 2005, ApJ 626, L101 Ramirez-Ruiz E., Garc a-Segura G., Salmonson J.D., & P erezRend on B., 2005, ApJ, 631, 435

c 2013 RAS, MNRAS 000, 114

You might also like