1: \documentclass{emulateapj}
2: %\documentclass[12pt,preprint]{aastex}
3: %\voffset=-0.5cm
4: \shorttitle{Wave Decay in MHD Turbulence}
5: \begin{document}
6:
7: \title{Wave Decay in MHD Turbulence}
8: \author{Andrey Beresnyak and Alex Lazarian}
9: \email{andrey, lazarian@astro.wisc.edu}
10: \affil{University of Wisconsin-Madison, Dept. of Astronomy}
11: \journalinfo{The Astrophysical Journal, 678:961-???, 2008 May ??
12: \hfil\break\copyright\ 2008. The American Astronomical Society. All rights reserved.}
13:
14:
15: \begin{abstract}
16: We present a model
17: for nonlinear decay of the weak wave in three-dimensional
18: incompressible magnetohydrodynamic (MHD) turbulence.
19: We show that the decay rate is different for parallel
20: and perpendicular waves.
21: We provide a general formula
22: for arbitrarily directed waves and discuss particular limiting
23: cases known in the literature.
24: We test our predictions with direct numerical
25: simulations of wave decay in three-dimensional MHD turbulence,
26: and discuss the influence of turbulent damping
27: on the development of linear instabilities in the interstellar
28: medium and on other important astrophysical processes.
29: \end{abstract}
30:
31: \keywords{turbulence, MHD}
32:
33: \section{Introduction}
34: The interstellar medium (ISM) is turbulent on scales ranging from AUs
35: to kpc (see Armstrong et al. 1995, Elmegreen \& Scalo 2004) with an
36: embedded magnetic field that influences almost all of its properties.
37: Turbulence is essential for understanding the ISM, the
38: intracluster medium (ICM), the Earth's magnetosphere,
39: solar wind, accretion disks, etc.
40:
41: The literature on astrophysical turbulence and its applications is
42: vast (see Biskamp~2003, McKee \& Ostriker 2007 and references
43: therein). The concepts of waves and eddies are both used to describe
44: magnetized turbulence. It is accepted that weak Alfv\'enic turbulence
45: consists of waves, while for strong turbulence wave-eddy
46: dualism is important (Goldreich \& Sridhar 1995, henceforth GS95).
47: In strong Alfv\'enic turbulence, nonlinear interactions damp the
48: energy of wave-eddies over approximately one wave period.
49: However, there are quite a few sources which exist in the turbulent
50: environment that emit waves with different degrees
51: of monochromaticity. These waves propagate in
52: the turbulent medium while being modified by the medium.
53: The evolution of such waves is of considerable astrophysical interest
54: and has been addressed in the literature on numerous occasions.
55: For instance, collisional and collisionless damping of fast
56: modes in a turbulent
57: medium was shown to differ considerably from the case when waves
58: propagate along laminar magnetic field lines, with important
59: consequences for cosmic ray propagation (Yan \& Lazarian 2004). The
60: latter, however, is the linear damping of waves, while the focus of this
61: paper is the nonlinear interaction of Alfv\'en waves with turbulence.
62:
63: Since we are addressing the subtle issue of nonlinear damping of Alfv\'en
64: waves by
65: surrounding strongly nonlinearly damped Alfv\'enic eddies, we feel that a
66: discussion of the properties of MHD turbulence is due.
67: It is well known that magnetohydrodynamics (MHD)
68: describes the dynamics of a highly conducting fluid with magnetic field.
69: Highly disordered quasi-stochastic flow, usually called turbulence, is
70: generally observed in settings when the Reynolds number, the ratio of
71: the inertial force to the viscous force, is large. When conductivity of the
72: fluid is very low, it can be described rather well by the Navier-Stokes
73: equations, which are, basically, the MHD equations without magnetic
74: field. If the conductivity, however, is high, the magnetic field can
75: no longer be ignored. In this case, the {\it magnetic Reynolds number} $Re_m$, or the ratio of the fluid velocity to the magnetic diffusion velocity, is
76: large. In fact, in the turbulent flow of a conductive
77: fluid there is a mechanism called dynamo, which increases total
78: magnetic energy, making it more dynamically important over time (see,
79: e.g., Vishniac et al, 2003). It is no wonder that in the
80: astrophysical environment, where both Reynolds and magnetic Reynolds
81: numbers are high, magnetic fields are always present and always dynamically
82: important.
83:
84: The paradoxical nature of turbulence is that even though the dissipation
85: parameters (e.g. viscosity and magnetic diffusivity) tend to zero,
86: the dissipation rates stay pretty much the same.
87: The general explanation for this, as proposed by Kolmogorov (1941),
88: is that the energy is carried from large scales through many
89: intermediate scales down to small dissipation scales.
90: This property is shared by turbulence in magnetized fluids.
91: Much of the progress in understanding turbulence comes from
92: so-called Kolmogorov models which assume a cascade of a
93: conserved quantity, such as energy, down-scale. Once the functional
94: dependence of the cascade decay rate, local in k-space,
95: is known, we roughly know the spectrum of the conserved quantity
96: (normally energy, but sometimes other quantities, such as the number of quasiparticles,
97: enstrophy, etc.)\footnote{The Kolmogorov-type models do not necessarily
98: give $-5/3$ spectrum, which is the result specific for the incompressible
99: hydro turbulence. For example, so-called Langmuir turbulence has the cascade
100: direction inverted and the flat spectrum of energy.}.% (...)}.
101: These models do not attempt to describe
102: all properties of turbulence, but some important ones, such as
103: the spectrum of energy. They essentially assume
104: self-similarity, so they don't describe deviations
105: from self-similarity, known as intermittency. The level
106: of sophistication of these so-called mean field models is enough
107: for many applications.
108:
109: While the spectrum of hydrodynamic incompressible turbulence
110: is well-established, strong MHD turbulence
111: is an area of active research (e.g. Montgomery \& Turner (1981),
112: Shebalin, Matthaeus \& Montgomery (1983), Higdon (1984)).
113: Numerical research (Cho \& Vishniac 2000, Maron \& Goldreich 2001,
114: Cho, Lazarian \& Vishniac 2002, Cho \& Lazarian, 2002, 2003, M\"uller, Biskamp \& Grappin 2003)
115: is roughly\footnote{The particular value of the spectral index and the index of
116: scale-dependent anisotropy is still an issue of debate (Galtier, Pouquet \& Mangeney 2005, Boldyrev 2005, 2006, Beresnyak \& Lazarian 2006, Gogoberidze 2007).}
117: consistent with a particular
118: model of strong turbulence, called Goldreich-Sridhar model (Goldreich \&
119: Sridhar, 1995). This mean-field model predicts structural anisotropy
120: with respect to the magnetic field according to so-called critical balance,
121: $k_\perp \delta v \sim k_\|v_A$, as well as a $k^{-5/3}$
122: energy spectrum.
123:
124: The wave dissipation rate has two major applications. First, if it is
125: directly measured, either in experiment, observation, or
126: numerical simulation, it can be used as a test of turbulence
127: itself, and, in certain mean-field models of turbulence, since
128: the latter rely on the phenomenologically estimated
129: value of the mean dissipation rate for every particular scale.
130:
131: The other application is turbulent damping
132: of linear instabilities. If the instability growth rate is smaller
133: than the turbulent damping rate, the instability does not develop.
134: This can have numerous consequences. For instance, it has been customary
135: to compare the instability rate to the viscous or magnetic dissipation
136: rates. Because of this, almost any ``large scale'' perturbation
137: with positive imaginary part of the frequency obtained from linear analysis
138: has been thought to develop a growing unstable mode. It is possible,
139: however, that this instability is slow enough to be damped
140: by the turbulent field.
141: This way the unstable configuration is stabilized, provided it is not
142: disrupted by turbulence itself. In the case when turbulence itself is
143: generated by an instability, the turbulent dissipation
144: rate will provide a powerful generic mechanism for nonlinear
145: saturation of the instability. Such a generic saturation mechanism
146: is highly desirable, since typically nonlinear saturation models
147: are very specific to the instability in question
148: and contain quite a number of assumptions.
149:
150: One example of the application of turbulent damping rates
151: to astrophysics is the study of the escape of high energy Cosmic Rays
152: (CRs) from our Galaxy (Yan \& Lazarian 2002, 2004, Farmer \& Goldreich, 2004).
153: Lazarian and Beresnyak (2006) developed a model in which
154: the low-energy CR mean free path is effectively reduced in the
155: presence of compressive motions due to the CR-Alfv\'en
156: anisotropic instability.
157:
158: Previous works that considered the turbulent wave dissipation
159: include Hollweg (1984), Similon \& Sudan (1989),
160: Kleva \& Drake (1992), Farmer \& Goldreich (2004),
161: Bian \& Tsiklauri (2007) and others.
162:
163: In our paper we develop a model of turbulent dissipation
164: which is purely nonlinear (does not depend on either $Re$ or $Re_m$) and has
165: self-similar power-law dependencies on the wavenumber, which is
166: characteristic of turbulence. We also explain why parallel waves
167: (with wavevector parallel to the field) decay in a different way
168: than perpendicular waves. We give a general formula for the decay
169: rate for arbitrarily directed\footnote{We call Alfvenic or pseudo-Alfvenic waves
170: parallel/perpendicular/arbitrarily {\it directed} instead of {\it propagating},
171: since these waves always propagate along the local magnetic field, regardless
172: of the orientation of their wavevector.} waves. The study of the turbulent decay
173: of parallel directed waves by Farmer \& Goldreich (2004) was motivated
174: by the fact that such waves have the highest instability grows rate
175: for CR-plasma instabilities such as the streaming
176: and the anisotropy instabilities (e.g., Kulsrud 2005).
177:
178: In what follows we describe the model of turbulent dissipation and
179: derive the formula for the dissipation rate in \S 2, provide numerical evidence
180: for turbulent wave dissipation in \S 3, discuss astrophysical implications
181: in \S 4, mention competing mechanisms for wave damping in \S 5,
182: discuss previous work in \S 6 and summarize our results in \S 7.
183:
184: \section{Wave dissipation}
185:
186: The equations of incompressible ideal MHD,
187:
188: \begin{equation}
189: \partial_t{\bf w^+}+({\bf w^-}\cdot\nabla){\bf w^+}=-\nabla P, %\eqno (1)
190: \end{equation}
191: \begin{equation}
192: \partial_t{\bf w^-}+({\bf w^+}\cdot\nabla){\bf w^-}=-\nabla P, %\eqno (2)
193: \end{equation}
194: \begin{equation}
195: \nabla\cdot{\bf w^+}=\nabla\cdot{\bf w^-}=0, %\eqno (3)
196: \end{equation}
197:
198: written in terms of Elsasser variables ${\bf w^+=v+b}$ and ${\bf w^-=v-b}$, where
199: ${\bf v}$ is the velocity and ${\bf b}$ is the magnetic field in velocity units
200: ${\bf b=B}/(4\pi \rho)^{1/2}$, bear close resemblance to the Navier-Stokes equations.
201: This resemblance is somewhat superficial, since ${\bf w^+}$ and ${\bf w^-}$ do not transform
202: the same way velocity transforms under Galilean transformations. Due to this fact,
203: there is usually some residual, or local mean magnetic field, that cannot
204: be excluded by the choice of frame of reference (Kraichnan, 1965).
205: If we go to sufficiently
206: small scales the decaying perturbation fields will be much smaller
207: than the local mean field. This case is called sub-Alfv\'enic turbulence.
208: The important conservation laws in non-dissipative MHD is the conservation
209: of the integrals $\int |w^+|^2d^3x$ and $\int |w^-|^2d^3x$, which are the analogs
210: of conservation of energy in the Navier-Stokes equations.
211: This allows us to build phenomenology of energy transfer in the spirit of
212: Kolmogorov (1941) for each of the ${\bf w^+}$ and ${\bf w^-}$ variables with
213: two different dissipation rates\footnote{In this paper we
214: consider only the symmetric case in which the kinetic viscosity equals magnetic
215: diffusivity ($Pr=1$), otherwise some exchange of energy between
216: Elsasser fields is possible at the dissipation or viscous scale. High Prandtl
217: number turbulence, which is often relevant for astrophysical processes, was studied
218: in Schekochihin et al 2002, Lazarian, Vishniac \& Cho 2004, etc.}.
219: The setup when those dissipation rates are equal is called balanced
220: turbulence, while the opposite case is called imbalanced
221: or cross-helical turbulence.
222: % (see, e.g., Grappin et al 1983,
223: %Cho et al 2002, Biskamp 2003, Beresnyak \& Lazarian 2007).
224:
225: The major difference between the incompressible Navier-Stokes and MHD equations
226: is that the characteristics of the latter, due to the presence of local
227: magnetic field, are always wave-like, with the speed of the perturbation being $v_A$.
228: While it is not possible to have waves in incompressible hydrodynamics,
229: it is possible to have waves or wave-like perturbations in incompressible
230: MHD. According to the GS95 model, most perturbations produced
231: by turbulence cannot be called well-defined waves. But there are other perturbations
232: generated, for example, by instabilities, which, as we will demonstrate in this paper,
233: have frequency much larger than decay rate and, therefore, can be called waves.
234:
235: This paper is focused on the study of the decay of a small perturbation
236: in stationary forced turbulence. Unlike decaying turbulence, where
237: the energy decays self-similarly, or as a power-law with time, the decay
238: of a small perturbation in stationary turbulence will be exponential.
239: This statement is based on the general concept of the flow of energy
240: through scales, which underlies all Kolmogorov-type models.
241: Indeed, in Kolmogorov models, the decay rate is some function
242: of the perturbation strength and the length scale: $\gamma_{\rm turb}(\delta v, \lambda)$.
243: For example, in strong hydrodynamical turbulence, $\gamma_{\rm turb}=\delta v/\lambda$.
244: Since the wave amplitude is small, it can not significantly affect $\delta v$. Therefore
245: the dissipation rate will be constant, meaning that the wave will decay exponentially.
246:
247: In this paper we only consider the Alfv\'en and pseudo-Alfv\'en waves
248: that exist in incompressible turbulence. The study of the decay
249: of all three wave modes in generic compressible turbulence
250: is a task that will be addressed elsewhere.
251:
252: Now we will provide a phenomenological
253: description of wave decay and the formula for the decay rate
254: for arbitrarily directed waves. To highlight the general problem
255: we first consider parallel waves.
256: According to the second-order perturbation theory, such waves
257: are not cascaded (e.g., Galtier et al 2002). Does
258: this property necessarily apply to strong turbulence?
259: Farmer \& Goldreich (2004) argued that this is not the case,
260: addressing the issue of the wave propagation parallel to the local
261: magnetic field when the effective perpendicular wavenumber
262: is obtained by so-called field wandering.
263: Let us imagine a plane wave with very large perpendicular
264: correlation length, corresponding to $k_\perp=0$ and finite
265: parallel correlation length $\Lambda$, corresponding
266: to $k_\|=1/\Lambda$. This wave, moving a distance of $\Lambda$,
267: will get disrupted by field wandering, as various pieces of this
268: wave propagate along its local field direction, which is slightly
269: different along the transverse spread of the wave. Note that only
270: field-wandering from the counter-propagating waves will contribute
271: to this process, as the co-propagating waves will move exactly the
272: same distance, being stationary in the frame of the wave.
273: The ``effective'' perpendicular wavenumber $k_{\perp, {\rm eff}}$ is determined by the condition
274: $k_{\perp, {\rm eff}}=(\delta B/B) k_\|$, where we use $\delta B$ of counter-wave
275: on scale $1/k_{\perp, {\rm eff}}$.
276:
277: The arguments above can be generalized for the case of an arbitrarily
278: directed wave in the following way.
279: While the wave is being decorrelated by turbulence, it obtains
280: a new ``effective'' $k_{\perp, {\rm eff}}$ from the vector sum of the original $k_\perp$ and the one
281: produced by turbulent field wandering. The effective $k_{\perp, {\rm eff}}$ in the
282: RMS sense will be the square mean of the two.
283: If the original wave has a
284: pitch angle $\theta$ with respect to the field and a
285: wavenumber magnitude of $k$, its effective perpendicular wavenumber
286: will be determined by the equation
287:
288: \begin{equation}
289: k_{\perp, {\rm eff}}=k(\sin^2\theta+(\delta B/B)^2\cos^2\theta)^{1/2}, %\eqno (4)
290: \end{equation}
291:
292: or, if we assume the GS95 scaling $\delta B/B \sim k_\perp^{-1/3}$,
293:
294: \begin{equation}
295: k_{\perp, {\rm eff}}=k(\sin^2\theta+(k_{\perp, {\rm eff}} L)^{-2/3}\cos^2\theta)^{1/2}. %\eqno (5)
296: \end{equation}
297:
298: According to the arguments above,
299: we should include only the part of $\delta B/B$ which corresponds
300: to the counter-wave, since co-propagating perturbations do not influence
301: the structure of the wave. This is due to the exact solution
302: of the incompressible MHD which involves only one wave species of
303: arbitrary amplitude. In eq. (5), for the sake of simplicity, we omitted
304: the coefficient $1/2$ before
305: $\delta B/B$ which would indicate that only half of the perturbations
306: contribute to the decorrelation effect.
307:
308: It is easy to verify from eq. (5), that for perpendicular modes
309: $k_{\perp, {\rm eff}}=k\sin\theta=k$ (as if field wandering is unimportant),
310: while for parallel modes, $k_{\perp, {\rm eff}}\sim k^{3/4}$ (as in Farmer \& Goldreich 2004).
311:
312: The decay rate is determined by the $k_{\perp, {\rm eff}}$
313: we calculated from (5):
314: \begin{equation}
315: \gamma=v_A k_{\perp, {\rm eff}}^{2/3}L^{-1/3}. %\eqno (6).
316: \end{equation}
317: Eq. (6) deserves a more detailed explanation. While cascading
318: of perturbations in {\it weak turbulence} have resonant
319: character, that is, they require counter-waves with equal
320: and oppositely directed ${\bf k_\|}$ (e.g., Ng \& Bhattachargee 1996,
321: Goldreich \& Sridhar 1997, Galtier et al 2002),
322: {\it strong turbulence} involves nonlinear
323: perturbations where strict resonance condition is not required.
324: This can be demonstrated if we consider a cascaded perturbation
325: as a passively advected vector field. This approach is valid
326: as long as the amplitude of this sample or test perturbation
327: is small enough to avoid any back-reaction to fully developed,
328: balanced strong turbulence. Now, according to critical balance of GS95,
329: the transverse structure of this passively advected field
330: is strongly perturbed by counter-waves with the same $k_\perp$
331: after traveling a distance of $1/k_\|=k_\perp^{-2/3}L^{1/3}$.
332: On the other hand, according to GS95, this is the correlation
333: length of the counter wavepacket, so the ``next''
334: perturbation will be applied to our test field incoherently.
335: Therefore, these perturbations of the transverse structure
336: of the test field are irreversible and can be called cascading
337: in perpendicular direction. Our main point here is that
338: the lateral structure of the {\it test field} is irrelevant
339: for its cascading (unless the lateral structure produces
340: the transverse structure by field wandering), as long as cascading
341: happens in strong GS95 turbulence\footnote{This way the test-field
342: approach is different from the dynamic model of GS95 itself,
343: where the lateral structure of the counter-waves is essential for
344: back-reaction on co-waves and creating its own critical balance.}.
345:
346: The formulae (5) and (6) give the general dependence of the
347: decay rate $\gamma$ on arbitrarily directed (oblique) waves
348: with wavevector $k$ and pitch-angle $\theta$.
349: This dependence is, in general, not a power-law of $k$.
350: There are two limiting cases of
351: the perpendicular directed and the parallel directed waves which give
352: power-laws, namely
353:
354: \begin{equation}
355: \gamma_\perp=v_A k^{2/3} L^{-1/3}, \ \ \ \gamma_\|=v_A k^{1/2} L^{-1/2}. %\eqno (7)
356: \end{equation}
357:
358: Fig. 1 shows the dependence of the decay rate $\gamma$ on the wavevector $k$
359: for a variety of pitch angles $\theta$ with respect to the magnetic field.
360:
361: \begin{figure}
362: \plotone{f1.eps}
363: \caption{The relative decay rate versus wavenumber for different angles
364: as obtained from Eqs. 5 and 6.}
365: \label{gamma_theor}
366: \end{figure}
367:
368: \section{Numerical Results}
369: We have conducted a series of 71 incompressible three-dimensional $320^3$
370: MHD simulations with stochastic turbulent driving. We used a pseudo-spectral
371: code to solve Eqs. 1-3 and introduced linear dissipation as an energy sink.
372: We used relatively low-order (6th order) hyperdiffusion to extend the
373: inertial range while avoiding a strong bottleneck effect. The turbulent
374: driving supplied energy between $k=2$ and $3.5$ while most of the
375: dissipation happened around $k=70$. We regarded the interval $k=5$ to $50$
376: as an inertial interval of turbulence. Simulations had an
377: external magnetic field and the driving was tuned to produce
378: strong turbulence which was nearly isotropic around the driving scale.
379: More details of the code and the driving can be found in
380: Cho and Vishniac (2000).
381:
382: In this series one of the simulations was without wave driving,
383: and all others were driven with a plane wave with a particular
384: wave vector ${\bf k}$. The turbulent driving and the initial conditions
385: in all series were exactly the same. The initial conditions were taken
386: from another turbulent simulation with driving that ran for around 20
387: Alfv\'en times. We waited for about 0.3 Alfv\'en times for the wave
388: to reach saturation. We chose the wave driving amplitude so that
389: the saturated wave does not incur any significant back-reaction
390: to turbulence. After this the wave driving was switched off but
391: the turbulent driving continued just like in the simulation
392: without wave driving.
393:
394: As expected, after the wave reached saturation,
395: it was no longer plane-parallel and
396: its energy distribution in k-space was affected by both field line
397: wandering and strong turbulent cascading.
398: It is more appropriate to say that the simulation box was filled with a collection
399: of wave packets which had the dispersion in the direction of ${\bf k}$ {\it and}
400: the dispersion in $k_\|$ and $k_\perp$ that comes from turbulent decorrelation
401: and the corresponding uncertainty relation. In Fig. 2 we presented a visual
402: representation of the distribution of wave energy that was initially driven
403: as the parallel wave with $k_\|=17$ (in cube size units) and reached saturation.
404:
405: \begin{figure}
406: \plotone{f2.eps}
407: \caption{Decorrelation of the parallel-propagating wave by turbulent
408: field wandering. Darker areas indicate larger wave energy. The wave
409: is decorrelated and cascaded primarily in the direction perpendicular
410: to the magnetic field.}
411: \label{decorrel}
412: \end{figure}
413:
414: The wave energy was obtained by subtracting the results of the simulation with
415: and without wave driving. We summed the result of subtraction
416: only over the relevant regions in $k$-space, to reduce errors.
417: We note again, that driving, although being turbulent-like, was predetermined,
418: and within our short simulation times (about 0.6 Alfv\'en times) chaos
419: in motions on large scales was not yet able to develop. This was confirmed by the fact
420: that the result of subtraction was mostly positive and significantly differed
421: from zero only around the region where we pumped the wave. One notable exception
422: from this rule was the velocity at k=0, or the mean velocity. It increased,
423: which is interpreted as a transfer of momentum from the wave to the fluid.
424: We have carried out a total of 70 $320^3$ simulations with different values
425: of the wave ${\bf k}$ vector which have covered the inertial range
426: of our turbulent simulation from $k=10$ to $k=40$ with different angles
427: with respect to the mean magnetic field.
428:
429: After the wave driving was switched off, the energy of the wave started to decay exponentially
430: (see Fig. 3). We measured the decay rate by fitting a decay curve.
431: Fig. 4 shows the dependence of the decay rates on the $k$ value and on the angle
432: of original inclination of wave to the magnetic field (the angle $\theta$ between ${\bf k}$ and ${\bf B}$).
433: As we see, for waves parallel to the external field the decay rate follows
434: a $k^{1/2}$ law (see also Farmer \& Goldreich 2004).
435: The theoretical curves for the different inclinations that were fitted into
436: numerical values of $\gamma$, unexpectedly, have slightly different
437: values for the outer scale $L$. We believe that this could be either due
438: to the fact that the wavelengths that we considered were relatively close
439: to the driving scale, or the fact that we simulated trans-Alfv\'enic
440: turbulence.
441:
442: \begin{figure}
443: \plotone{f3.eps}
444: \caption{The evolution of the oblique wave energy during the driving and decay stages.
445: The wave driving was at $(k_\|, k_\perp)=(12,12)$ in code units, approximately
446: $(4, 4) L^{-1}$.}
447: \label{wave_decay}
448: \end{figure}
449:
450: \begin{figure}
451: \epsscale{.6}
452: \plotone{f4.eps}
453: \epsscale{1}
454: \caption{The dependence of the decay rate on the magnitude of the wavevector for
455: different angles between ${\bf k}$ and ${\bf B_0}$ in a direct 3D numerical simulation (points),
456: fitted with curves obtained from eqs. 5 and 6 (lines).}
457: \label{gamma_simul}
458: \end{figure}
459:
460:
461: \section{Applications of Turbulent Damping}
462: The damping of Alfven waves with ${\bf k}$ parallel to the
463: magnetic field was invoked by Farmer \& Goldreich (2004) in the context
464: of the damping of the streaming instability by ambient turbulence
465: (first mentioned in Yan \& Lazarian 2002). They estimated that with the
466: modest level of turbulence in our Galaxy, the streaming instability
467: will be suppressed for cosmic rays with energies higher than
468: about 100 GeV; therefore such cosmic rays can freely propagate
469: along the magnetic field lines.
470: Lazarian \& Beresnyak (2006) have used turbulent damping
471: as one of the mechanisms that limits the CR anisotropy instability
472: which occurs in a magnetized fluid in
473: the presence of compressible turbulence. This instability greatly
474: decreases the CR's mean free path. However, due to turbulent
475: damping, there is a high energy limit of around 1000 GeV
476: (for our Galaxy) for this mechanism. Particles with energy
477: higher than this limit will be unaffected by the instability
478: and scattered primarily through other mechanisms, such as direct
479: interaction with MHD modes (Chandran 2000, Yan \& Lazarian 2002, 2004, 2007).
480:
481: The influence of the turbulent decay to the development of
482: instabilities can lead to the various interesting astrophysical
483: effects. Let us consider one example. In a recent paper, Everett et al
484: (2007) (henceforth E07) considered a model of launching the Galactic
485: wind by CR pressure (see also Breitschwerdt et al, 1991). They
486: argued that CR pressure is necessary to launch the wind, since
487: thermal pressure is too small. One of the critical assumptions of
488: this model is that the streaming instability operates effectively and
489: allows CRs to exchange momentum with the plasma. Let us now
490: assume that the streaming instability is damped by turbulence.
491: According to Farmer \& Goldreich (2004), if the amplitude of turbulence is
492: similar to that of our Galaxy, the damping of the streaming instability
493: will affect CRs with energies higher than 100 GeV, making them
494: effectively disconnected from the background plasma and able to
495: escape. If most of the energy (and the momentum) of CRs are in the lowest
496: energy CRs, as E07 originally assumed, this would present no problem
497: for their model. It is believed, that our Galaxy, on average, has a
498: relatively steep spectrum of CRs, between the slopes of $-2.6$ and
499: $-3.1$, justifying this assumption. However, the winds are most
500: likely to be launched from the sites of intensive
501: CR production, such as regions with many supernova shocks. Some modern
502: theories of shock acceleration with precursor favor a rather shallow
503: spectrum of the accelerated particles, such as shallower than $-2$ (Malkov
504: \& Drury 2001) \footnote{The average galactic value of $-2.6$ is then
505: obtained by a variety of mechanisms such as re-acceleration,
506: escape, etc.}. CRs with such a spectrum have most of their energy and momentum
507: in higher energy particles that can freely escape because of damping
508: of the streaming instability and, therefore, CR pressure is too weak to launch the
509: wind. The existence of the Galactic wind would therefore suggest that it is
510: always being launched by a CR population with a steep spectrum, or that these winds
511: have very low level of turbulence. Since we cannot directly observe
512: the spectrum of the CRs near acceleration sites such information about the slope
513: being shallower or steeper than $-2$ can be used to limit theories of
514: CR acceleration in shocks.
515:
516: The other application of energy dissipation of Alfv\'en
517: waves is the problem of coronal heating. It has been long known
518: that due to the very small values of viscous and magnetic dissipation
519: coefficients, ohmic and viscous heating are unlikely to dissipate
520: Alfv\'en waves significantly. On the other hand, we know that
521: a relatively small fraction of energy of waves actually escapes the
522: coronal region. Turbulent dissipation can deal with this problem,
523: since it provides much higher dissipation rates than ohmic
524: and viscous heating, and, unlike the latter, does not depend
525: on the Reynolds numbers $Re$ and $Re_m$. Parker (1991) noted,
526: however, that in the corona, the stochastic component of the field
527: mostly consists of waves propagating in one direction,
528: outwards from the Sun. In our terminology this is a strongly imbalanced
529: case. While there is certainly a strong imbalance close to the photosphere
530: near the source of the waves, the situation in the upper
531: solar corona could be closer to balanced turbulence due to various mechanisms
532: that allow waves to reflect back, such as the geometry of the guiding
533: magnetic field, the parametric instability, etc. Despite coronal heating still
534: being a poorly understood process there are few things that we
535: would like to note with regards to our model. First of all, our model
536: does not consider dissipation in imbalanced turbulence, due
537: to the fact that models of strong imbalanced turbulence
538: are being developed (see Beresnyak \& Lazarian 2007 and references therein, see Discussion).
539: Secondly, we only considered waves with small amplitudes
540: that do not produce strong back-reaction to turbulence.
541: We hope, however, that the present work can provide
542: insight into a more complicated picture that will include
543: imbalanced turbulence, reflection of waves, etc.
544:
545: There is also the problem of launching winds, such as a stellar
546: wind, by momentum from waves. It is similar to the previous
547: problem, in that we need mechanisms of wave reflection, in order
548: to generate turbulence, otherwise waves will freely escape.
549: It is well known that the solar wind is strongly imbalanced
550: close to the Sun, but becomes less imbalanced at larger distances, almost reaching
551: equipartition between waves near the Earth's orbit.
552: Roberts et al (1987) suggested that this was due to the
553: dominance of the kinetic energy on the outer scale.
554:
555: \section{Competing Mechanisms of Damping}
556:
557: While the ohmic and viscous dissipation rates decrease very rapidly
558: with increasing wavelength, as $k^2$, there are a number
559: of mechanisms of Alfv\'en wave dissipation that, in principle,
560: can compete with turbulent dissipation. The dissipation mechanisms
561: can be subdivided into linear and nonlinear mechanisms. We call it
562: a linear mechanism when the decay rate does not depend on the amplitude
563: of the wave. Such mechanisms include viscous damping, turbulent damping (even
564: though the turbulence itself is a nonlinear process), collisionless
565: (Landau) damping, and the resonance damping. Nonlinear damping
566: mechanisms have a decay rate that depends on the wave amplitude.
567: These include nonlinear Landau damping, wave steepening, etc.
568:
569: Nonlinear Landau damping (e.g., Kulsrud 2005) requires
570: two waves with close frequencies and a population
571: of ions that is in resonance with the beat wave of those
572: two waves. While, generically, the decrement of this damping
573: is proportional to the frequency, it has a nonlinear
574: $(\delta B/B_0)^2$ factor that limits damping to high
575: amplitude waves. Note that the turbulent damping mechanism that
576: we consider in this paper has a substantial limitation in that the
577: wave amplitude has to be small enough to not back-react on the
578: turbulence itself. Thus, nonlinear damping mechanisms and the
579: linear (with respect to the wave amplitude) turbulent
580: damping are, in a sense, complimentary to each other (see fig. 5).
581: Another nonlinear mechanism of damping
582: is wave steepening (e.g., Cohen \& Kulsrud, 1974).
583: This mechanism is also supposed to have $\gamma \sim kv_A(\delta B/B_0)^2$.
584:
585: \begin{figure}
586: \plotone{f5.eps}
587: \caption{Dominating dissipation mechanisms for various wave lengths and
588: wave amplitudes in the ICM or hot ISM. Linear collisionless (Landau) damping acts only on the
589: pseudo-Alfven mode, while nonlinear Landau damping and turbulent damping
590: operate on both modes.}
591: \label{nl_landau}
592: \end{figure}
593:
594: The prominent linear mechanism of damping is so-called resonance
595: damping (e.g., Hollweg, 1984). Unlike previously mentioned
596: mechanisms that are considered in a homogeneous medium and usually
597: work equally well in any geometry, this linear damping mechanism appears
598: when, due to the specific boundary conditions or inhomegeneity of the flow,
599: the so-called Alfvenic resonance appears. This resonance corresponds
600: to the flow speed being equal to the phase speed of the Alfvenic
601: wave. The resonance is typically the $1/x$ type divergence for the magnetic
602: field and velocity. Notably, the dissipation provided by this resonance
603: does not depend on viscosity or magnetic diffusivity (Kappraff \& Tataronis 1977).
604: The dissipation rate can be calculated from the ideal MHD equations with $\nu=\eta=0$
605: using contour integrals (e.g., Livshitz \& Pitaevskii 1981,
606: Pariev \& Istomin 1996). The Alfven resonance of the flow has been proposed as a prominent
607: candidate for an in-situ acceleration mechanism in jets (Beresnyak, Istomin \& Pariev 2003).
608: Resonance damping usually provides damping rates that are a fraction
609: of the wave frequency.
610: It is considered to be important when the wave length is of order
611: the scale at which boundary conditions are set. In jets we expect
612: turbulent damping to be more important than resonance damping
613: (if the latter is present) for wavelengths that are much smaller
614: than the jet's diameter, assuming that for these wavelengths the medium
615: can be treated as homogeneous.
616:
617:
618:
619: \section{Discussion}
620: This paper deals with wave dissipation in a {\it balanced}
621: MHD turbulence. The more general imbalanced case was not considered
622: for a variety of reasons. First, the theory of strong imbalanced MHD
623: turbulence is still being developed (Lithwick, Goldreich \& Sridhar 2007,
624: Beresnyak \& Lazarian 2007). On the other hand, if we adopt the model
625: of imbalanced turbulence presented in Beresnyak \& Lazarian 2007 (furthermore BL07)
626: a number of questions arise. Most importantly, we will be unable to reproduce the
627: statement of \S 2 that the cascading of the wave is irrelevant to its lateral
628: structure. This is due to the fact that neither the strong nor the weak wave from BL07
629: has a critical balance with itself (i.e. $k_\|v_A=k_\perp\delta v$ does not
630: hold if all quantities refer to the same wave, instead BL07 provides a different
631: expression for the critical balance of each wave). Therefore there is
632: no guarantee that we can use the strong cascading formula, at least when the
633: wave has a large $k_\|$.
634:
635: An attempt to include irregularities of magnetic field lines due
636: to turbulence to increase predicted dissipation in solar corona was
637: made in Silimon \& Sudan (1989). Their model relies on a description
638: of turbulent magnetic field lines as divergent with a specific
639: Lyapunov constant, or exponentiation rate $\lambda$. We argue that
640: this is not a proper description of turbulent fields in developed
641: turbulence because such fields are approximately self-similar and do not have any
642: designated scale in its inertial interval. In developed turbulence
643: every scale has its own exponentiation rate. Silimon \& Sudan (1989)
644: give a formula for dissipation length which is inversely proportional
645: to the Lyapunov constant (which is not specified, but estimated) and
646: only weakly (logarithmically) proportional to the wavenumber. It also
647: depends on magnetic Reynolds number.
648: This paper, in contrast, advocates purely nonlinear dissipation
649: rates that do not depend on $Re_m$ or $Re$.
650: Parker (1991) criticized Silimon \& Sudan's approach of using stochastic field
651: lines to increase dissipation by noting that
652: the stochastic component of the field is itself part of the
653: waves propagating outwards along the field. We commented on this
654: controversy previously by noting that one has to have a mechanism
655: to reflect the waves back, otherwise they escape freely.
656:
657: Kleva \& Drake (1992) have studied nonlinear dissipation of waves
658: in the presence of a stochastic field on an outer scale by numerical
659: methods. They confirmed that the dissipation of large scale waves
660: does not depend on dissipation coefficients, while for small scale
661: waves it approached the viscous and resistive limit. However, their
662: stochastic field was not turbulent, but rather a predetermined field
663: on the outer scale. Also, the damping rate they measured did not depend
664: on wavevector as a power-law.
665:
666: Bian \& Tsiklauri (2007) have considered mixing of Alfv\'en wavepackets
667: in chaotic magnetic fields. They obtained an analytical
668: expression for the advection-diffusion of wavepackets using
669: a WKB approximation and claimed that the stationary wave energy
670: spectrum is $k^{-1}$. We feel that the WKB method is not the appropriate
671: tool to derive turbulent dissipation, as it requires
672: that the wave should have wavelengths that are much smaller than the
673: underlying perturbations, while in turbulence the most effective
674: nonlinear dissipation comes from scales comparable with the
675: wavelength.
676:
677: Hollweg (1984) considered dissipation of Alfv\'en waves
678: in coronal loops, primarily due to resonance mechanism,
679: but also speculated on turbulent heating. He took the Kolmogorov
680: dissipation rate (which is a decay that happens on
681: a kinetic timescale $\lambda/v$) with the perturbation
682: length scale as wavelength and the RMS wave velocity
683: as perturbation velocity. This way he obtained
684: dissipation which is compatible with heating in
685: coronal loops as well as coronal holes. We have to note,
686: however, that propagating waves do not necessarily
687: generate turbulence and even if there is a flux
688: of waves in both directions, there are certain
689: requirements for MHD turbulence to be strong
690: turbulence and have the Kolmogorov dissipation rate.
691: For example, in the context of coronal loops, with a strong
692: guide field, it is easy to imagine a situation where
693: turbulence is not strong on the outer scale, therefore
694: it has dissipation rate lower than Kolmogorov.
695:
696: \section{Summary}
697:
698: In this paper we demonstrated the following:
699:
700: 1. Parallel and perpendicular directed waves
701: are dissipated in different ways. Perpendicular
702: waves are cascaded naturally by counter-waves,
703: while parallel waves are first decorrelated
704: by the field wandering and then cascaded.
705:
706: 2. In the inertial interval of sub-Alfv\'enic turbulence,
707: parallel waves are damped more slowly than perpendicular
708: waves with the same wavenumber. The decay rate for
709: the parallel wave is much smaller than its frequency,
710: it is not eddy but a well-defined wave.
711:
712: 3. In a homogeneous medium with open boundaries turbulent
713: dissipation is more effective than the other dissipation
714: mechanisms, provided that the wave amplitude is sufficiently
715: small and the wavelength is in the inertial interval
716: of turbulence.
717:
718: \acknowledgments
719: We thank Jungyeon Cho for providing the initial version of the code
720: and for some ideas on wave propagation in turbulence.
721: AB thanks John Everett and Vladimir Pariev for fruitful discussions.
722: AB thanks IceCube project for support of his research.
723: AL acknowledges the NSF grant AST-0307869 and the support from
724: the Center for Magnetic Self-Organization in Laboratory and Astrophysical
725: Plasmas.
726:
727:
728: \begin{thebibliography}{}
729:
730: \bibitem{BerIstPar}
731: Beresnyak, A. R., Istomin, Ya. N., Pariev, V. I. 2003 A\&A, 403, 793
732:
733: \bibitem{polar}
734: Beresnyak, A., Lazarian, A. 2006, ApJ, 640, L175
735:
736: \bibitem{imbal}
737: Beresnyak, A., Lazarian, A. 2007, astro-ph/0709.0554
738:
739: \bibitem{Bian}
740: Bian, N., Tsiklauri, D. 2007, astro-ph/0709.0260
741:
742: \bibitem{Bold}
743: Boldyrev, S. 2005, ApJ, 626, L37
744:
745: \bibitem{Bold2}
746: Boldyrev, S. 2006, Phys. Rev. Lett., 96, 115002
747:
748: \bibitem{Breit}
749: Breitschwerdt, D., McKenzie, J. F., \& V\"olk, H. J. 1991, A\&A, 245, 79
750:
751: \bibitem{Chandr}
752: Chandran B., 2000, Phys. Rev. Lett., 85, 4656
753:
754: \bibitem{CL02}
755: Cho, J., \& Lazarian, A. 2002, Phys. Rev. Lett. 88, 24, 245001
756:
757: \bibitem{CL03}
758: Cho, J., \& Lazarian, A. 2003, MNRAS, 345, 325
759:
760: %\bibitem{CL05}
761: %Cho, J., \& Lazarian, A. 2005, Theoret. Comput. Fluid Dynamics, 19, 127
762:
763: \bibitem{CLV02}
764: Cho, J., Lazarian, A., \& Vishniac 2002, ApJ, 564, 291
765:
766: %\bibitem{CL02b}
767: %Cho, J., Lazarian, A., \& Vishniac, E. 2002b, ApJ, 566, L49
768:
769: \bibitem{CV00}
770: Cho, J. \& Vishniac, E. 2000, ApJ, 539, 273
771:
772: \bibitem{Cohen}
773: Cohen, R. H., Kulsrud, R. M. 1974 Phys. Fluids, 17, 2215
774:
775: \bibitem{Everett}
776: Everett, J.E., Zweibel E.G., Benjamin, R.A., McCammon, D., Rocks, L., Gallagher, J. S. III 2006, ApJ, in press
777:
778: \bibitem{Farm}
779: Farmer, A.J., \& Goldreich, P., 2004, ApJ, 604, 671
780:
781: \bibitem{Galt2005}
782: Galtier, S., Pouquet, A., Mangeney, A. 2005, Phys. Plasmas, 12, 092310
783:
784: \bibitem{Galt}
785: Galtier, S., Nazarenko, S.V., Newell, A.C., \& Pouquet, A. 2002, ApJ, 564, L49
786:
787: \bibitem{Giga}
788: Gogoberidze, G. 2007, Phys. Plasmas, 14, 022304
789:
790: \bibitem{GolS95}
791: Goldreich, P., \& Sridhar, S., 1995, ApJ, 438, 763
792:
793: \bibitem{GolS97}
794: Goldreich, P., \& Sridhar, S., 1997, ApJ, 485, 680
795:
796: \bibitem{Grapp}
797: Grappin, R., Pouquet, A., \& L\'eorat, J. 1983 A\&A, 126, 51
798:
799: \bibitem{Hig84}
800: Higdon J. C., 1984, ApJ, 285, 109
801:
802: \bibitem{Hollw}
803: Hollweg, J. 1984, ApJ, 277, 392
804:
805: \bibitem{Kappr}
806: Kappraff, J. M., \& Tataronis, J. A. 1977, J. Plasma Phys., 18, 209
807:
808: \bibitem{Kleva}
809: Kleva, R. G. \& Drake, J. F. 1992, ApJ, 395, 697
810:
811: \bibitem{Kol41}
812: Kolmogorov, A. 1941, Dokl.~Akad.~Nauk SSSR, 31, 538
813:
814: \bibitem{Kraich}
815: Kraichnan, R.H., Phys. Fluids 8, 1385 (1965)
816:
817: \bibitem{Kuls}
818: Kulsrud R., 2004, Plasma Physics for Astrophysics,
819: Princeton, NJ, Princeton University Press
820:
821: \bibitem{LB06}
822: Lazarian, A., Beresnyak, A. 2006, MNRAS, 373, 1195
823:
824: \bibitem{kiny}
825: Lifshitz, E. M. \& Pitaevskii, L. P. 1981, Physical kinetics, New York: Pergamon Press
826:
827: \bibitem{LitG07}
828: Lithwick, Y., Goldreich, P., \& Sridhar, S. 2007, ApJ, 655, 269
829:
830:
831: \bibitem{Malk}
832: Malkov, M. A., Drury, L.O'C. 2001, Rep. Prog. Phys, 64, 429
833:
834: \bibitem{Mar}
835: Maron, J., \& Goldreich, P. 2001, ApJ, 554, 1175
836:
837: \bibitem{Mckee}
838: McKee, C. F. Ostriker, E. C. 2007 Ann. Rev. of Astron. \& Astrophys., 45, 565
839:
840: \bibitem{Mont}
841: Montgomery, D.C., \& Turner, L. 1981, Phys. Fluids, 24, 825
842:
843: \bibitem{Mull0}
844: M\"uller W.-C., Biskamp, D., \& Grappin, R 2003, Phys. Rev. E, 67, 066302
845:
846: \bibitem{Ng}
847: Ng, C.S., Bhattacharjee, A. 1996, ApJ, 465, 845
848:
849: \bibitem{Par}
850: Pariev V. I., Istomin Ya. N., 1996, MNRAS, 281, 1
851:
852: \bibitem{Parker}
853: Parker, E. N. 1991, ApJ, 372, 719
854:
855: %\bibitem{LV99} Lazarian, A. \& Vishniac, E. 1999, ApJ, 517, 700
856:
857: %\bibitem{LitG01}
858: % Lithwick, Y., \& Goldreich, P., 2001, ApJ, 562, 279
859:
860: \bibitem{Robe}
861: Roberts, D. A., Goldstein, M. L., Klein, L. W., Matthaeus, W. H. 1987, J. of Geophys. Res., 92, 12023
862:
863: \bibitem{Sheb}
864: Shebalin, J.V., Matthaeus, W.H., \& Montgomery, D. 1983, J. of Plasma Phys., 29, 525
865:
866: \bibitem{Sheko}
867: Schekochihin, A. A., Maron, J. L., Cowley, S. C., McWilliams, J. C. 2002, ApJ, 576, 806
868:
869: \bibitem{Simil}
870: Similon, P., \& Sudan, R. 1989, ApJ, 336, 442-453
871:
872: \bibitem{VishnDyn}
873: Vishniac, E. T., Lazarian, A., \& Cho, J. 2003, in Turbulence and Magnetic Fields in Astrophysics Ed. by E. Falgarone, T. Passot., Lecture Notes in Physics, 614, 376
874:
875: \bibitem{YL02}
876: Yan H., Lazarian A., 2002, Phys. Rev. Lett., 89, 281102
877:
878: \bibitem{YL04}
879: Yan H., Lazarian A., 2004, ApJ, 614, 757
880:
881: \bibitem{YL07}
882: Yan H., Lazarian A., 2007, astro-ph/0710.2617
883:
884:
885: \end{thebibliography}
886:
887: \end{document}
888: