1: \documentclass[12pt,preprint]{aastex}
2:
3: %\slugcomment{Version \today}
4:
5: \shorttitle{Nonresonant instability}
6: \shortauthors{Ohira et al}
7:
8: \begin{document}
9:
10: \title{Two-Dimensional particle-in-cell simulations of the nonresonant, cosmic-ray driven instability in SNR shocks}
11:
12: \author{Yutaka Ohira\altaffilmark{1}, Brian Reville\altaffilmark{2},
13: John G. Kirk\altaffilmark{2} and Fumio Takahara\altaffilmark{1}}
14:
15: \begin{abstract}
16:
17: In supernova remnants,
18: the nonlinear amplification of magnetic fields upstream of
19: collisionless shocks is essential for the acceleration
20: of cosmic rays to the energy of the \lq\lq knee\rq\rq\ at
21: $10^{15.5}\,$eV. A nonresonant instability driven by the cosmic ray
22: current is thought to be responsible for this effect. We perform
23: two-dimensional, particle-in-cell simulations of this instability. We
24: observe an initial growth of circularly polarized non-propagating
25: magnetic waves as predicted in linear theory.
26: It is demonstrated that in some cases
27: the magnetic energy density in the growing waves,
28: can grow to at least 10 times its initial value. We find no
29: evidence of competing modes, nor
30: of significant modification by thermal effects. At late times we observe saturation
31: of the instability in the simulation, but the mechanism responsible is an
32: artefact of the periodic boundary conditions and has no
33: counterpart in the supernova-shock scenario.
34: \end{abstract}
35:
36:
37: \keywords{supernova remnants -- shock waves -- plasmas --
38: cosmic rays}
39:
40: \altaffiltext{1}{Department of Earth and Space Science,
41: Graduate School of Science, Osaka University, 1-1 Machikaneyama-cho,
42: Toyonaka, Osaka 560-0043, Japan; yutaka@vega.ess.sci.osaka-u.ac.jp}
43: \altaffiltext{2}{Max-Planck-Institut f{\"u}r Kernphysik, Heidelberg 69029, Germany}
44:
45:
46: \section{Introduction}
47:
48: Diffusive shock acceleration (DSA) at supernova remnant shocks
49: is widely considered to be the primary source of galactic cosmic rays
50: \citep[for a recent review see][]{Hillas05}. A crucial aspect of
51: DSA is the self-excitation of hydromagnetic waves
52: due to currents produced by the streaming energetic ions.
53: These provide the pitch-angle scattering necessary to allow spatial
54: diffusion of the relativistic particles.
55: The maximum
56: acceleration rate
57: that can be obtained corresponds to the smallest possible spatial diffusion
58: coefficient. In this \lq Bohm limit\rq\
59: the maximum particle energy is
60: determined either by the age of the remnant \citep{lag83},
61: or by its geometry \citep{Berezhko}. Even adopting optimistic
62: values for the upstream parameters of a supernova remnant shock,
63: particles are limited to energies below the knee of the
64: cosmic-ray spectrum at $\sim 10^{15.5}$~eV. However,
65: nonlinear amplification of the upstream magnetic field due
66: to cosmic-ray currents may provide a possible solution to
67: this problem \citep{Quest, luc00}.
68:
69: There is increasing observational evidence that the magnetic fields
70: immediately downstream of several young supernova remnant shocks are
71: much stronger than would be obtained by compressing the ambient
72: interstellar field at an MHD shock front \citep{Vink,ber03,bam05,uch07}.
73: These observations give additional motivation to the
74: development of theoretical models of nonlinear field generation by
75: both plasma instabilities in the precursor~\citep{luc00,bel04} and
76: fluid-type instabilities in the downstream plasma~\citep{gia07}.
77: These two mechanisms can, in principle, operate
78: simultaneously \citep{zir08}. However, to reach the highest energies, particles
79: must be speedily returned to the shock from excursions into both the
80: upstream and downstream plasmas. The generation of an
81: amplified magnetic field
82: in the precursor that is subsequently advected into the downstream
83: region is sufficient to ensure this, whereas amplification of only the
84: downstream field is not.
85:
86: Using a hybrid kinetic-MHD analysis, it was demonstrated by \cite{bel04}
87: that efficient cosmic-ray acceleration can, in the linear regime,
88: drive a short wavelength, almost aperiodic, instability with wave-vector
89: parallel to the ambient field. The growth of this mode
90: can be considerably more rapid than that of the resonantly
91: driven modes previously considered \citep{mcken,acht83}.
92: The instability is driven by the reaction of the thermal plasma, as
93: it attempts to compensate the current produced by the streaming cosmic rays.
94: Numerical investigations of this instability have
95: been performed using both MHD \citep{bel04, bel05, rev08, zpv08}
96: and Particle-In-Cell (PIC) simulations \citep{nie08, riq08}.
97: While the results of MHD simulations show that the perturbed
98: magnetic field $\delta B$ becomes much larger than the initial
99: uniform magnetic field $B_0$, the PIC simulations of
100: \cite{nie08} did not observe the
101: parallel mode predicted in the linear analysis, and only moderate
102: amplification of the magnetic field was observed.
103: On the other hand, the recent results reported by
104: \citet{riq08} do identify the parallel mode and propose
105: an explanation of its saturation in the non-linear stage.
106:
107: In this paper we report new PIC simulations of this instability.
108: We extend the range of parameters beyond that
109: investigated by \cite{nie08}, and select values that more
110: closely represent the environment in an SNR precursor. Although both 2D and 3D
111: simulations in a similar parameter range have been performed
112: \citep{riq08}, we restrict our work to 2D. Our results
113: --- obtained independently and with some technical differences in the
114: numerical treatment --- can be regarded as complementary, and our
115: physical interpretation, at least of the early stages of the non-linear
116: development, are congruent. In section~2, we recall the
117: linear kinetic analysis
118: of the nonresonant instability, in order
119: to elucidate the conditions
120: that must be satisfied in the PIC simulations.
121: The details of our simulation
122: parameters are described in section~3.
123: We present the results of the simulations in
124: section~4 followed by a discussion of the
125: saturation and its implications in section~5.
126: We conclude with a summary of the results.
127:
128: \section{Linear analysis}
129: \label{lineartheory}
130:
131: We review the linear kinetic analysis of an
132: electron-ion plasma
133: upstream of a non-relativistic shock with a power-law
134: distribution of streaming cosmic rays. We focus our
135: attention on parallel shocks, with the streaming velocity directed along the
136: zeroth order magnetic field. The kinetic linear dispersion relation
137: for circularly polarized hydromagnetic waves propagating
138: parallel to the mean magnetic field in such a plasma was investigated
139: by \cite{acht83}, subject to the conditions of zero net charge and current.
140: In the upstream plasma frame the cosmic ray streaming velocity is
141: approximately that of the shock velocity \citep{McClements96}.
142: Using the diffusion approximation it can be shown
143: that, for a power-law momentum distribution of streaming
144: cosmic rays with a spectral index $s>3$ $(f_{\rm cr}(p)\propto p^{-s})$, the dispersion
145: relation can be written as \citep{rev07}:
146: \begin{eqnarray}
147: \label{disp}
148: {\omega}^2
149: +\epsilon \left(\frac{k^2V^2_{\rm ti}}{\Omega_{\rm ci}}\right)
150: {\omega}
151: -v_{\rm A}^2 k^2-
152: \epsilon \zeta v_{\rm s}^2 \frac{k}{r_{\rm g}}
153: \left(\sigma(kr_{\rm g})-1\right)=0,
154: \end{eqnarray}
155: where $\omega=\omega_r+{\rm i}\gamma$ is the frequency of the waves in the upstream
156: ion rest frame, $k>0$ the wavenumber, $v_{\rm A}$ the Alfv\'en velocity,
157: $V_{\rm ti}$ and $\omega_{\rm ci}$ the ion thermal velocity
158: and cyclotron frequency, $v_{\rm s}$ the speed of the shock and
159: $r_g=p_{\rm min}c/eB_0$
160: the gyroradius of the lowest energy cosmic ray. The polarization
161: of the waves is determined by $\epsilon = +1(-1)$ for right (left) handed
162: modes ($\omega_r > 0$). This result agrees with that found by \citet{bel04} using an MHD approach,
163: and extends it by including an additional term representing thermal ion damping.
164: The growth rate of the unstable modes is determined by the driving
165: parameter
166: \begin{equation}
167: \zeta =
168: \frac{n_{\rm cr} p_{\rm min}}{n_{\rm i} m_{\rm i} v_{\rm s}},
169: \nonumber
170: \end{equation}
171: where $n_{\rm cr,i}$ are the densities of cosmic rays and ions.
172: Finally $\sigma$ is a complex function
173: describing the electric fields produced by the cosmic ray current. It
174: can be shown that $\sigma$ is a decreasing function of
175: wavenumber for $kr_{\rm g}>1$ \citep{rev07}.
176: In particular, for wavelengths much less than
177: the gyroradius of the cosmic rays, $\sigma$ can be neglected with respect to
178: unity, and one finds that there exists a purely growing mode with
179: growth rate
180: \begin{equation}
181: \label{gmax}
182: \gamma_{\rm NR} = \frac{\zeta}{2}\frac{v_{\rm s}}{v_{\rm A}}
183: \frac{v_{\rm s}}{r_{\rm g}}
184: \approx\frac{1}{2}\frac{v_{\rm s}}{v_{\rm A}}
185: \frac{n_{\rm cr}}{n_{\rm i}}\Omega_{\rm ci}.
186: \end{equation}
187: An important constraint in deriving Eq.~(\ref{disp}), is the
188: magnetization condition, $|\omega| \ll \Omega_{\rm ci}$.
189: On length scales much shorter than their gyroradius,
190: the cosmic rays are essentially unmagnetized, i.e., their trajectories
191: are rectilinear. The fact that the background plasma is magnetized leads
192: to an asymmetry in the system, resulting in an uncompensated perpendicular
193: current. This is what drives the growth or decay of all
194: elliptical modes with ${\bf k \cdot B_0} \neq 0$ \citep{bel05, melrose05}.
195: If the background ions are unmagnetized, there is no inertia to
196: support the growing waves, and the ions quickly act to compensate the
197: cosmic-ray current. Significant heating of the plasma causes the
198: ions to behave as if they were unmagnetized, thereby
199: reducing the growth-rate. The same effect arises
200: in the relativistic analysis of this instability
201: \citep{rev06}.
202:
203: In order to maintain a steady galactic cosmic-ray energy density,
204: supernova remnants must channel approximately 10~percent of the
205: kinetic energy entering the shock front into cosmic rays. Assuming a
206: shock velocity of $v_{\rm s}=0.01c$ with a cosmic-ray distribution
207: $f_{\rm cr}(E)\propto E^{-2} $ from $10^7$~eV to $10^{15}$~eV,
208: this implies a cosmic-ray number density of
209: $n_{\rm cr}=10^{-5}\,{\rm cm}^{-3}$. Taking fiducial interstellar medium values for the
210: ion number density and magnetic field: $n_{\rm i}=1\,{\rm cm}^{-3}$
211: and $B=3~\mu {\rm G}$, the maximum growth rate of the nonresonant
212: instability is
213: $\gamma_{\rm NR}\sim 2.5\times 10^{-3} {\Omega_{\rm ci}}$.
214: In the upstream rest frame, the cosmic rays drift with
215: respect to the thermal ions at the shock velocity $v_{\rm s}$ . To
216: compensate this current, there is relative velocity between upstream
217: ions and electrons, $V_{\rm d}=v_{\rm s}n_{\rm cr}/n_{\rm
218: e}=10^{-7}c$. This value is much smaller than the ion and
219: electron thermal velocity. Hence an MHD description of the plasma is
220: appropriate for the parameters in the precursor of a supernova
221: remnant. P.I.C. simulations of this essentially MHD instability are
222: computationally very intensive. However, they are the only way of
223: quantifying the relative importance of high frequency waves and other
224: kinetic effects on the nonlinear evolution of the instability.
225:
226: \section{Simulation}
227: \label{simulation}
228:
229: We use a 2D fully relativistic electromagnetic PIC code
230: with a
231: fast algorithm that solves for the current density and conserves
232: charge \citep{ume03}. The
233: simulations are carried out in the upstream rest frame. While the
234: simulations of \cite{nie08} used a full kinetic treatment of the
235: cosmic rays, during the linear stage of growth there was no
236: significant reduction in the streaming energy of the cosmic rays. In
237: order to identify the growth and evolution of the instability, we
238: maintain a constant, uniform external cosmic-ray current throughout
239: the simulations, as has been used in previous MHD
240: simulations \citep{bel04, rev08}.
241:
242: Similar to previous simulations, we investigate the role
243: played by streaming cosmic rays of uniform density in a homogeneous
244: electron-ion plasma. This is intended to represent a small region
245: in the precursor of a supernova remnant shock. However, when, during
246: the course of the simulation, the magnetic field fluctuations
247: associated with the nonresonant instability become strong, they begin
248: to damp the relative drift velocities of the cosmic rays and the gas.
249: In a shock precursor, a change in relative drift velocity is
250: automatically associated with an increase in the gas and cosmic ray
251: densities, as the shock is approached. The simulation box, however,
252: has periodic boundary conditions and, therefore, simulates a spatially
253: uniform plasma. These boundary conditions therefore fail to reproduce
254: the behavior expected in the shock precursor, as soon as the relative
255: drift velocity changes significantly. For this reason, the saturation
256: of the instability observed in the simulation, which occurs when the
257: drift speeds become equal, does not correspond to the saturation
258: mechanism expected in the shock precursor. However, the initial linear
259: phase of the instability and a substantial part of its nonlinear
260: evolution are accurately modelled.
261:
262: \subsection{Setting}
263:
264: We define the zeroth order magnetic field to be along the
265: negative $x$-direction, and the
266: cosmic rays stream in the positive $x$-direction. Periodic boundary
267: conditions are used in both the $x$ and $y$ directions.
268: The electrons, ions and cosmic rays are initialized
269: such that the overall
270: charge and current densities vanish. Thus,
271: $n_{\rm e} = n_{\rm i} + n_{\rm cr}$
272: and the electrons have a
273: net drift velocity in the $x$-direction of $V_{\rm d,e}=V_{\rm d,cr}
274: n_{\rm cr}/n_{\rm e}$. (Subscripts e, i and cr represent electrons,
275: ions and cosmic rays, respectively.) At the beginning of the simulation,
276: each population is uniformly distributed in the $x-y$ plane with a
277: Maxwellian momentum distribution at equal temperatures
278: ${ T=T_{\rm e}=T_{\rm i}}=1.3{\rm~keV}$.
279: The length of each cell $\Delta x = \Delta y$ and time step $\Delta t$
280: of the simulation are twice the Debye length and
281: $0.0714\omega_{\rm pe}^{-1}$, respectively.
282: Initially, each cell contains 49 positively and 49 negatively
283: charged macroparticles. In the presence of a uniform cosmic ray
284: density, the charge of each species of macroparticle
285: must be chosen such that the overall plasma is
286: initially charge neutral. The charge to mass ratio of the negatively
287: charged macroparticles corresponds to that of an electron.
288:
289: The values for $\Omega_{\rm ce}/ \omega_{\rm pe}$, $n_{\rm cr}/n_{\rm i}$,
290: $V_{\rm d}/c$ and the number of cells for each simulation run are
291: given in Table~\ref{table1}, where $\Omega_{\rm ce}$ and $\omega_{\rm pe}$
292: are the electron cyclotron frequency and electron plasma frequency,
293: respectively. We also specify the mass ratio
294: $m_{\rm i}/m_{\rm e}$ for each run. To make the problem more tractable,
295: we are forced to use unrealistically high values for the cosmic-ray
296: number densities. This is done purely to reduce the computation time,
297: and, as we show, the essential physical mechanisms are still well-captured.
298:
299: The size of the simulation box in the $x$ and $y$-directions for run A
300: is taken to be $L_x=L_y=6.66\lambda_{\rm NR} $,
301: where $\lambda_{\rm NR}=2\pi v_{\rm A}/\gamma_{\rm NR}^{\rm max}$ is
302: the wavelength of the most unstable mode of the nonresonant instability.
303: The simulation is followed until 40$\tau_{\rm grow}$,
304: where $\tau_{\rm grow}=\gamma_{\rm NR}^{-1}$.
305: The parameters used in run B are chosen such that the magnetization condition
306: $\gamma_{\rm NR} < \Omega_{\rm ci}$ is not satisfied. As discussed in section
307: \ref{lineartheory}, it is not expected that the nonresonant current-driven
308: instability will operate in such a situation. Similar parameters
309: have been used by \cite{nie08}, and we perform this simulation in order to
310: facilitate comparisons with that work.
311: The size of the simulation box for run-B is $L_x=L_y=9.65\lambda_{\rm NR}$.
312:
313: We also perform two additional simulations, runs C and D, with higher
314: cosmic-ray drift velocity and mass ratio, respectively. The box size is the
315: same as that used in run A, $L_x=L_y=6.66\lambda_{\rm NR} $.
316: While the conditions
317: in run C are less appropriate for a supernova shock precursor than those of
318: run A, the increased cosmic-ray drift speed provides a greater amount of free
319: energy in the system. Likewise, the magnetization condition in run A is satisfied
320: better than in run D, however, the larger mass ratio allows us to investigate the effect
321: that increased inertia has on the thermal ions.
322:
323:
324: \section{Results}
325: \label{simresults}
326:
327: The evolution of the spatially averaged energy density of the magnetic
328: and electric field components associated with the growing waves,
329: for run-A is shown in Figure 1.
330: The linear growth rate of the spatially averaged transverse component
331: of the magnetic field is approximately $\sim0.5\gamma_{\rm NR}$. The
332: characteristic wave length is almost the same as that predicted by the
333: linear analysis. Figure 2a shows the spatial distribution of the
334: $z-$component of magnetic field during the linear phase of growth
335: ($t=14\tau_{\rm grow}$)\footnote{Note $t>\tau_{\rm grow}$ does not
336: mean nonlinear phase because the initial perturbation is very
337: small.} , where the $k$-vector is clearly directed along the zeroth
338: order field. The $y$-component of the magnetic field has a similar
339: structure with the phase of the wave pattern shifted by $\pi/2$.
340:
341: For comparison, we show in Figure 2b, the $z-$component of the magnetic
342: field during the linear phase of growth for run-B. The ions in this
343: case are not magnetized, and the nonresonant mode is
344: not observed. The wave is aperiodic and its wavevector is
345: almost perpendicular
346: to the beam. This is typical of Weibel-type instabilities. We discuss this
347: further in the next section.
348:
349: The linear development of the fluid quantities in the simulations is similar to
350: that of previous MHD simulations \cite{bel04, bel05, rev08, zpv08}.
351: The initial stage sees the development of the aperiodic instability, with
352: uniform density. As the fastest growing mode emerges from the initial noise,
353: the net ${\bf j \times B}$ force begins to push the plasma in the direction
354: transverse that of the cosmic ray drift, generating low
355: density regions between filaments of compressed plasma.
356: The filaments are uncorrelated in the direction of
357: the cosmic ray drift, similar to what is observed in MHD simulations
358: \citep[e.g.][]{bel04, rev08}. The growth of the cavities
359: is eventually limited due to the expansion of neighbouring cavities,
360: at which point they appear to have a radius of $\lambda_{\rm NR}$.
361:
362: The late-time nonlinear evolution, however, differs from that of previous MHD
363: simulations. While both see the eventual disruption of the filamentary structures,
364: the particle in cell simulations see an acceleration of the background plasma in
365: the direction of the cosmic-ray drift.
366: This effect is not observed in simulations that
367: represent the system as a single MHD fluid driven by a fixed external
368: current, because the charge of the streaming particles, and
369: hence, the direction in which they drift, is undefined.
370: Figures 3(a) and 3(b) show the distribution of the total magnetic field energy and upstream
371: proton density at the end of the simulation ($t=40\tau_{\rm grow}$).
372: Locally, the amplification factor $\delta B/ B_0$ is about~16 but the
373: global, spatially averaged value is about~4.
374: These images differ quite considerably from the nonlinear results presented
375: in previous MHD simulations \citep{bel04, zpv08}, not only because of the net drift,
376: but also due to the anti-correlation between the magnetic field and the plasma density
377: (Figure 3(b)).
378:
379: At the point at which the magnetic field ceases to grow,
380: the bulk plasma is drifting with a constant
381: velocity $v_{d,i}=0.9v_{d,cr}=0.09c$. This is almost the shock velocity.
382: The development of run~C and run~D are similar to that of run~A. However, the
383: saturated field values are slightly larger, since the initial amount of free energy
384: available is larger in both cases. The growth of the different
385: field components is plotted for run C and run~D in Figures 4 and 5, respectively.
386: We discuss the physical nature of the field saturation in the next section.
387:
388: \section{Discussion}
389: \label{discussion}
390:
391: We find that the nonresonant cosmic-ray driven instability develops in
392: all cases where the background plasma is magnetized.
393: From Figure \ref{fig2}(a) and \ref{fig2}(b), it can be seen that there are
394: considerable differences between the results of run~A and run~B although the observed wavelength is similar to $\lambda_{\rm NR}$.
395: The most unstable mode in run~B is the Weibel type
396: instability, occurring between the counterstreaming electrons and
397: ions. Neglecting thermal effects and
398: the magnetic field, the growth rate $\gamma_{\rm WI}^{\rm max}$ and the
399: wavelength $\lambda_{\rm WI}$ of the Weibel instability are
400: \begin{equation}
401: \gamma_{\rm WI}^{\rm max}=\frac{n_{\rm cr}}{n_{\rm e}}\frac{v_{\rm s}}{c}\omega_{\rm pi}, \ \ \lambda_{\rm WI}=2\pi \frac{c}{\omega_{\rm pe}},
402: \end{equation}
403: where the direction of the wave vector is perpendicular to the drift direction.
404: Comparing the two theoretical growth rates and the wavelengths we see that
405: \begin{equation}
406: \frac{\gamma_{\rm WI}^{\rm max}}{\gamma_{\rm NR}^{\rm max}} = \frac{n_{\rm e}}{n_{\rm i}} \simeq 1, \ \ \frac{\lambda_{\rm WI}}{\lambda_{NR}}=\frac{1}{2}\frac{n_{\rm cr}}{n_{\rm i}}\frac{V_{\rm d,cr}}{c}\frac{\omega_{\rm pe}}{\Omega_{ce}} \simeq 1.
407: \end{equation}
408: For the choice of parameters used in run~B, the wavelengths of
409: the different instabilities are quite similar,
410: even though the mechanism is quite different.
411: This is consistent with the results of \citet{nie08},
412: and we also find the fastest growing mode to be slightly oblique.
413: Hence we may be observing a mixed-mode between the Weibel
414: and the nonresonant current driven instability or some other type
415: of mixed-mode instability similar to those discussed in \cite{Bret}.
416:
417: In the late stages of the simulations we see the emergence of a
418: non-thermal power law tail in the ion distribution.
419: The use of a constant external cosmic-ray current
420: prevents the development of the Buneman instability, a source
421: of ion heating in the linear development. This dramatically
422: reduces the heating in the initial stages, that was previously observed
423: in \cite{nie08}. Figure 6 shows
424: the final electron and ion energy distribution for run~A, C and
425: D. The distributions are essentially isotropic, and since the
426: bulk drift velocity is small in comparision with the thermal velocities,
427: we can safely calculate the spectrum in the box frame.
428: The electron energy distribution can be well fitted with a
429: Maxwellian distribution with temperature $10$ keV for run~A and $40$
430: keV for run~C and D. A non-thermal tail appears in the distribution of
431: the ions with a very soft power-law index of $\sim 7.3$ for Run-A,
432: but a relatively hard $\sim2$ for run~C and D. The
433: mechanism responsible for the formation of this
434: power-law distribution is not clear, and although our simulations
435: have a relatively large number of particles per cell, we cannot rule out a
436: numerical artefact. As
437: regions of oppositely polarized magnetic field collide each other, an
438: anti-parallel configuration of the magnetic fieldlines appears. It is,
439: therefore,
440: possible that collisionless magnetic dissipation may also play an
441: important role in plasma heating mechanism and the saturation of the
442: field growth as well as the reduction in the drift velocity between
443: the electrons and ions.
444: This is an interesting process in its own right,
445: and further investigation is warranted.
446: However, in the current context, it is not clear what
447: role these mechanisms play, since the late-stage nonlinear behavior
448: has no counterpart in the supernova-shock scenario.
449:
450:
451: In agreement with \cite{nie08} and \cite{riq08},
452: we observe in our simulations that the
453: magnetic field growth ceases when the plasma bulk velocity becomes
454: comparable to the cosmic-ray drift velocity. Although we have fixed
455: the cosmic ray current in our simulations, the observed saturation is
456: clearly due to the reduction in the relative speeds of plasma and cosmic rays.
457: Indeed, if the cosmic rays are isotropic in the plasma frame, no streaming
458: instability can operate.
459: The simulations are performed in a box with periodic boundary
460: conditions in space. The plasma speed is initially zero,
461: and the cosmic ray streaming speed is maintained
462: constant in time and space. It represents the speed with which the shock
463: front approaches the simulation box.
464: Within this picture, the instability can saturate only when the entire
465: plasma moves with the drift speed of the cosmic rays.
466:
467: In the precursor of a shock front, conditions are different.
468: There, the interaction
469: between plasma and cosmic rays defines the diffusion length scale of the
470: cosmic rays. This is the scale on which the cosmic ray current, or,
471: equivalently, pressure, decays with distance ahead of the shock front.
472: In the case of efficient acceleration, it is also the length scale on
473: which the precursor plasma is compressed.
474: Simulations can model this situation provided the box size is small
475: compared to the diffusion length scale, and the time scales are short
476: compared to the time on which the cosmic ray current and
477: plasma density change because of the approaching shock front.
478: However, this restriction means that the simulations no longer model
479: a precursor when they are followed until the plasma speed approaches the
480: shock speed (which equals the cosmic-ray drift speed).
481:
482: This can be seen explicitly from the equation of
483: conservation of mass in a stationary precursor,
484: $\rho(x) u(x) = \mbox{constant}$, which dictates that any change in the flow
485: velocity must be associated with a change in the density of the fluid.
486: In the present context,
487: the simulation box is in the upstream plasma frame,
488: approaching the shock with velocity $v_{\rm s}$.
489: Defining the density and bulk velocity inside the box to be $\rho_b$ and
490: $u_{\rm b}$ respectively, we see that
491: \begin{equation}
492: \rho_{\rm b}= \frac{v_{\rm s}\rho_0}{v_{\rm s}-u_{\rm b}}
493: \end{equation}
494: where $\rho_0$ is the initial density of the thermal plasma. Thus, as
495: the flow speed inside the box increases, the density should also
496: increase. Since the total density in the box is fixed throughout the
497: simulation, the observed growth rate is only accurate provided the
498: bulk velocity inside the box is small compared to the shock velocity,
499: or, equivalently, the streaming speed of the cosmic rays, i.e.,
500: $u_{\rm b} \ll v_{\rm s}$. This corresponds to $\tau_{\rm
501: grow}< 20$ for run A and $\tau_{\rm grow}< 23$ for runs C and D.
502: The amplification of the energy density in
503: the total field in run~A is not yet substantial, but in runs~B and C
504: roughly 10 times the initial energy density is reached. The final
505: saturation levels are much higher, but our simulations do not
506: necessarily imply that these can be reached in a precursor scenario,
507: and a PIC simulation of this
508: case is currently out of reach.
509:
510:
511: \section{Summary}
512:
513: We have performed 2D PIC simulation to investigate the nonlinear
514: physics of the nonresonant current instability. We have demonstrated
515: that the energy in the growing waves can substantially exceed that of
516: the initial seed field. However, limitations inherent in the simulation
517: method lead to an artificial saturation level.
518: Although we cannot compute a realisitic saturated
519: value of the magnetic field, substantial amplification is seen
520: in some runs before they lose validity,
521: and the possibility remains open that the field would
522: continue to grow, transferring the energy to
523: longer lengthscales, in accordance with theoretical predictions.
524: Thus, as predicted, magnetic field amplification is triggered by
525: the nonresonant streaming instability and is likely
526: to play an important role in the
527: acceleration of cosmic rays to the knee and beyond in supernova remnants.
528:
529:
530:
531: \acknowledgments Y.O. is grateful to S. Matsukiyo and T. Umeda for
532: discussions on the nonresonant instability and PIC
533: simulations. B.R. thanks A. Spitkovsky for many useful discussions.
534: Y.O. is supported by a Grant-in-Aid for
535: JSPS Research Fellowships for Young Scientists. BR gratefully
536: acknowledges support from the Alexander von Humboldt foundation.
537: Numerical computations were carried out on the Cray XT4 at Center for
538: computational Astrophysics, CfCA, of the National Astronomical Observatory
539: of Japan and on the Blue Gene/P at the Rechenzentrum Garching of the Max
540: Planck Society.
541:
542: \begin{thebibliography}{}
543:
544: \bibitem[{{Achterberg}(1983)}]{acht83}{Achterberg}, A. 1983, \aap, 119, 274
545:
546: \bibitem[Bamba et al.(2005)]{bam05} Bamba, A., Yamazaki R., Yoshida T., Terasawa T., and Koyama, K., 2005, \apj, 621, 793
547:
548: \bibitem[Bell(2004)]{bel04} Bell, A. R. 2004, \mnras, 353, 550
549:
550: \bibitem[Bell(2005)]{bel05} Bell, A. R. 2005, \mnras, 358, 181
551:
552: \bibitem[{{Berezhko}(1996)}]{Berezhko} Berezhko, E. G.,Astroparticle Physics, 5, 367
553:
554: \bibitem[Berezhko et al.(2003)]{ber03} Berezhko, E. G., Ksenofontov, L. T., \& V{\"o}lk, H. J., \aap, 412, L11
555:
556: \bibitem[Bret et al.(2005)]{Bret}{{Bret}, A., {Firpo}, M.-C. \& {Deutsch}, C.} 2005 \pre, 72, 6403
557:
558:
559: \bibitem[Giacalone \& Jokipii(2007)]{gia07} Giacalone, J., \& Jokipii, J.R., 2007, \apjl, 663, L41
560:
561: \bibitem[{Hillas}(2005)]{Hillas05}
562: {Hillas}, A.~M. 2005, Journal of Physics G Nuclear Physics, 31, 95
563:
564: \bibitem[{{Lagage} \& {Cesarsky}(1983)}]{lag83}
565: {Lagage}, P.~O. \& {Cesarsky}, C.~J. 1983, \aap, 125, 249
566:
567: \bibitem[Lucek \& Bell(2000)]{luc00} Lucek, S. G. \& Bell, A. R. 2000, \mnras, 314, 65
568:
569: \bibitem[McClements et al.(1996)]{McClements96}
570: {McClements}, K.~G., {Dendy}, R.~O., {Drury}, L.~O., {Duffy}, P., 1996, \mnras, 280, 219
571:
572: \bibitem[McKenzie \& V\"olk (1982)]{mcken}
573: J.F. McKenzie, H.J. V\"olk, 1982 \aap, 116, 191
574:
575: \bibitem[Melrose(2005)]{melrose05} Melrose D., 2005, AIPC, 781, 135
576:
577: \bibitem[Niemiec et al(2008)]{nie08} Niemiec, J., Pohl, M., Stroman, T. \& Nishikawa, K., 2008, \apj, 684, 1174
578:
579: \bibitem[Reville et al.(2006)]{rev06} Reville, B., Kirk, J. G., \& Duffy, P., 2006, Plasma Physics and Controlled Fusion, 48, 1741
580:
581: \bibitem[Reville et al.(2007)]{rev07} Reville, B., Kirk, J. G., Duffy, P. \& O'Sullivan, S., 2007, \aap, 475, 435
582:
583: \bibitem[Reville et al.(2008)]{rev08} Reville, B., O'Sullivan, S., Duffy, P. \& Kirk, J. G., 2008, \mnras, 386, 509
584:
585: \bibitem[Riquelme \& Spitkovsky(2008)]{riq08} Riquelme, M. A. \& Spitkovsky, A., 2008, astro-ph/0810.4565
586:
587: \bibitem[{Shapiro et al.}(1998)]{Quest}
588: {{Shapiro}, V.~D., {Quest}, K.~B. \& {Okolicsanyi}, M.}, 1998, \grl 25, 845
589:
590: \bibitem[Uchiyama et al.(2007)]{uch07} Uchiyama, Y., Aharonian, F A., Tanaka, T., Takahashi, T., \& Maeda, Y., 2007, \nat, 449, 576
591:
592: \bibitem[Umeda et al. (2003)]{ume03} Umeda,T., Omura, Y., Tominaga, T., \&Matsumoto, H. 2003, Comp. Phys. Comm., 156, 73
593:
594: \bibitem[{{Vink} \& {Laming}(2003)}]{Vink}{Vink}, J. \& {Laming}, J.~M. 2003, \apj, 584, 758
595:
596: \bibitem[{{Zirakashvili} \& {Ptuskin}(2008)}]{zir08}
597: {Zirakashvili}, V.~N., \& {Ptuskin}, V.~S. 2008, \apj, 678, 939
598:
599: \bibitem[{{Zirakashvili et al.}(2008)}]{zpv08}
600: {Zirakashvili}, V.~N., {Ptuskin}, V.~S. \& {V\"olk}, H.~J. 2008, \apj, 678, 255
601: \end{thebibliography}
602:
603: \clearpage
604:
605: \begin{figure}
606: \plottwo{f1a.eps}{f1b.eps}
607: \caption{Time development of the spatially averaged root mean square of (a) magnetic field, (b) electric field for Run-A. Solid, dashed and dotted curves represent the $x$, $y$ and $z$-components, respectively.
608: \label{fig1}}
609: \end{figure}
610:
611: \clearpage
612:
613: \begin{figure}
614: \plottwo{f2a.eps}{f2b.eps}
615: \caption{$z$-component of magnetic field at (a) $t=14\tau_{\rm grow}$ in run-A, (b) $t=8 \tau_{\rm grow}$ in run-B.
616: \label{fig2}}
617: \end{figure}
618:
619: \clearpage
620:
621: \begin{figure}
622: \plottwo{f3a.eps}{f3b.eps}
623: \caption{(a) The magnetic field energy density, and (b) density of ions at $t=40\tau_{\rm grow}$ in run-A \label{fig3}}
624: \end{figure}
625:
626: \clearpage
627:
628: \begin{figure}
629: \plottwo{f4a.eps}{f4b.eps}
630: \caption{Time development of the spatially averaged root mean square of (a) magnetic field, and (b) electric field for Run-C. Solid, dashed and dotted curves represent the $x$, $y$ and $z$-components, respectively.
631: \label{fig4}}
632: \end{figure}
633:
634: \clearpage
635:
636: \begin{figure}
637: \plottwo{f5a.eps}{f5b.eps}
638: \caption{Time development of the spatially averaged root mean square of (a) magnetic field, and (b) electric field for Run-D. Solid, dashed and dotted curves represent the $x$, $y$ and $z$-components, respectively.
639: \label{fig5}}
640: \end{figure}
641:
642: \clearpage
643:
644: \begin{figure}
645: \plotone{f6.eps}
646: \caption{Energy distribution at the final stage in run-A, C and D \label{fig6}}
647: \end{figure}
648:
649:
650: \clearpage
651: \begin{deluxetable}{lccccc}
652: \tablecaption{Simulation Parameters}
653: \tablewidth{0pt}
654: \tablehead{
655: Run & $\Omega_{\rm ce}/\omega_{\rm pe}$ & $m_{\rm i}/m_{\rm e}$ & $n_{\rm cr}/n_{\rm i}$ & $V_{\rm d,cr}/c$ & cell}
656: \startdata
657: A & $3.26\times 10^{-2}$ & 10 & 1/20 & 0.1 & $4096\times 4096$\\
658: B & $4.52\times 10^{-2}$ & 10 & 1/3 & 0.3 & $512 \times 512 $\\
659: C & $3.26\times 10^{-2}$ & 10 & 1/20 & 0.2 & $2048\times 2048$\\
660: D & $3.26\times 10^{-2}$ & 40 & 1/20 & 0.1 & $4096\times 4096$\\
661: \enddata
662: \tablecomments{Parameters of the simulation runs described in this paper. Listed are:
663: the ratio of electron cyclotron frequency and electron plasma frequency $\Omega_{\rm ce}/\omega_{\rm pe}$,
664: ion-electron mass ratio $m_{\rm i}/m_{\rm e}$,
665: the density ratio of ambient ions and cosmic rays $n_{\rm cr}/n_{\rm i}$,
666: the velocity ratio of cosmic-ray drift and light speed $V_{\rm d,cr}/c$, the cell number}
667: \label{table1}
668: \end{deluxetable}
669:
670: \end{document}
671:
672: