1: \documentclass[preprint2]{aastex}
2: %\documentstyle[12pt,aaspp4]{article}
3: %ApJ 50880
4: %\documentstyle[11pt,emulateapj]{article}
5: \newcommand{\be}{\begin{equation}}
6: \newcommand{\ee}{\end{equation}}
7: \newcommand{\bea}{\begin{eqnarray}}
8: \newcommand{\eea}{\end{eqnarray}}
9: \begin{document}
10: \title{The Anisotropy of MHD Alfv\'{e}nic Turbulence}
11: \author{Jungyeon Cho \altaffilmark{1}}
12: \affil{Department of Astronomy, University of Texas, Austin, TX
13: 78712; cho@astro.as.utexas.edu}
14:
15: \and
16:
17: \author{Ethan T. Vishniac}
18: \affil{Department of Physics and Astronomy, Johns Hopkins University,
19: Baltimore, MD 21218; ethan@pha.jhu.edu}
20: \altaffiltext{1}{Department of Physics and Astronomy, Johns Hopkins University,
21: Baltimore, MD 21218}
22:
23: %ApJ, in press (Vol 539; Aug 10, 2000?)
24:
25: \begin{abstract}
26: We perform direct 3-dimensional numerical simulations for
27: magnetohydrodynamic (MHD) turbulence in a periodic box
28: of size $2\pi$ threaded by strong uniform
29: magnetic fields.
30: We use a pseudo-spectral code with hyperviscosity and hyperdiffusivity
31: to solve the incompressible MHD equations.
32: We analyze the structure of the eddies
33: as a function of scale.
34: A straightforward calculation of anisotropy in wavevector space
35: shows that the anisotropy is scale-{\it independent}.
36: We discuss why this is {\it not} the true scaling law and how the curvature of
37: large-scale magnetic fields affects the power spectrum
38: and leads to the wrong conclusion.
39: When we correct for this effect, we find
40: that the anisotropy of eddies
41: depends on their size: smaller eddies are
42: more elongated than larger ones along {\it local} magnetic field lines.
43: The results are consistent with the scaling law
44: $\tilde{k}_{\parallel} \sim \tilde{k}_{\perp}^{2/3}$ proposed by
45: Goldreich and Sridhar (1995, 1997).
46: Here $\tilde{k}_{\|}$ (and $\tilde{k}_{\perp}$) are wavenumbers measured
47: relative to the local magnetic field direction.
48: However, we see some
49: systematic deviations which may be a sign of limitations to
50: the model, or our inability to
51: fully resolve the inertial range of turbulence in our
52: simulations.
53: \end{abstract}
54: \keywords{ISM:general-MHD-turbulence}
55:
56: \section{Introduction}
57: Many astrophysical plasmas, including the interstellar medium
58: and the solar wind,
59: often show magnetic fields whose energy density is
60: greater than or equal to the local
61: kinetic energy density. In these plasmas the magnetic fields
62: play a dominant dynamical role, mediated
63: by magnetohydrodynamic (MHD) waves.
64: In the incompressible limit, there are only two types of linear modes:
65: shear Alfv\'{e}n waves and pseudo Alfv\'{e}n waves.
66: While these two modes have different polarization directions,
67: they have the same dispersion relation and propagate along
68: the magnetic field lines at the Alfv\'{e}n speed.
69: Therefore, the nonlinear interactions of wave packets
70: moving along the magnetic field lines at the Alfv\'{e}n speed
71: determine the dynamics of incompressible magnetized plasmas with a
72: strong background field.
73: In this paper, we study the anisotropy of
74: the MHD turbulence in this regime. We will refer to
75: this turbulence as incompressible Alfv\'{e}nic turbulence.
76:
77: Nonlinear processes and the corresponding energy spectrum of
78: incompressible Alfv\'{e}nic turbulence are still among
79: the most controversial problems in MHD.
80: Since the pioneering works of Iroshnikov (1963) and Kraichnan (1965),
81: the Iroshnikov-Kraichnan (IK) theory has been widely accepted as a
82: model for incompressible, highly conducting MHD turbulence.
83: The IK theory predicts
84: $E_M(k) \sim E_K(k) \sim k^{-3/2}$ from a Kolmogorov-like dimensional
85: analysis. Here, $E_M(k)$ and $E_K(k)$
86: are the magnetic and kinetic energy spectra respectively.
87: In this framework, two counter-traveling eddies
88: (i.e. Alfv\'{e}n wave packets)
89: interact and transfer energy to smaller spatial scales only when they
90: collide, as they move in opposite directions along
91: the magnetic field lines.
92: Since the duration of such a collision is shorter than the
93: conventional eddy turnover
94: time by a factor of $t_v(l)/t_A(l)$, this collisional process is
95: inefficient and the spectral energy
96: transfer time as a function of scale $l$ ($=t_{cas}(l)$) increases
97: by the same factor compared to
98: the eddy turnover time ($l/v_l$) in ordinary hydrodynamic turbulence.
99: Here $t_v(l)=l/v_l$ and $t_A(l)=l/V_A$ are
100: eddy turnover time and Alfv\'{e}n time respectively,
101: $V_A \equiv B/\sqrt{4\pi \rho}$, and
102: $B$ is the rms magnetic field strength. When the
103: external field is strong, as assumed in IK theory, this
104: quantity is usually set to $B_0$, the strength of
105: the uniform background field.
106: If the spectral energy cascade rate
107: \be
108: \epsilon \sim \frac{ v_l^2 }{ t_{cas}(l) } \sim
109: \frac{ v_l^3 }{l} \frac{ t_A(l) }{ t_v(l) }
110: \ee
111: is scale-independent and $E_M(k) \approx E_K(k)$, then we obtain
112: the IK energy spectra.
113:
114: The IK theory assumes an isotropic distribution of energy in
115: ${\bf k}$-space.
116: However, many researchers have argued that anisotropy
117: is an important characteristic in MHD turbulence
118: (for example, Shebalin et al 1983, Montgomery and Matthaeus 1995).
119: This anisotropy results from the resonant conditions for 3-wave
120: interactions (or 4-wave
121: interactions, when 3-wave interactions are null).
122: The resonant conditions for the 3-wave interactions are
123: \begin{eqnarray}
124: {\bf k}_1 + {\bf k}_2 & = & {\bf k}_3, \\
125: \omega_1 + \omega_2 & = & \omega_3,
126: \end{eqnarray}
127: where ${\bf k}$'s are wavevectors and $\omega$'s are wave frequencies.
128: The first condition can be viewed as momentum conservation
129: and the second as energy conservation.
130: Alfv\'{e}n waves satisfy the dispersion relation: $\omega = V_A |k_{\|}|$,
131: where $k_{\|}$ is the component of wavevector parallel to the background
132: magnetic
133: field.
134: Since only opposite-traveling wave packets interact, ${\bf k}_1$ and
135: ${\bf k}_2$ must have opposite signs.
136: Then from equations (2) and (3),
137: either $k_{\|,1}$ or $k_{\|,2}$ must be equal to 0.
138: That is, zero frequency modes are essential
139: for energy transfer. If $k_{\|,2}=0$, we have
140: \begin{eqnarray}
141: k_{\|,1} &=& k_{\|,3},\\
142: k_{\|,2} &=& 0
143: \end{eqnarray}
144: (Shebalin et al 1983).
145: Therefore, in the wavevector space,
146: 3-wave interactions make energy cascade
147: in directions perpendicular to the mean magnetic field.
148: Since the energy cascade is strictly perpendicular to the
149: mean magnetic field, the actives modes in wavevector space have a
150: slab-like geometry with a constant width. The implication is
151: that the nonlinear cascade of energy works against isotropy in
152: ${\bf k}$ space. Furthermore, it is important to note that
153: equations (2) and (3) are true only when
154: wave amplitudes are constant.
155: In reality, nonlinear interactions provide a natural broadening
156: mechanism, following the uncertainty relation,
157: $\Delta t \cdot \Delta \omega \sim 1$.
158: In particular, if a wave has a frequency less than or comparable to the
159: nonlinear interaction rate, it is effectively a zero frequency mode.
160:
161: Goldreich and Sridhar (1995, 1997)
162: showed that in the strong incompressible shear
163: Alfv\'{e}nic turbulence regime (i.e. $\tau_{NL}^{-1} \sim \omega$),
164: these arguments lead
165: to a new scaling law with a scale-dependent anisotropy.
166: In this model smaller eddies are more elongated.
167: Their arguments are based on the assumption of
168: a {\it critically balanced} cascade, $k_{\parallel}V_A \sim k_{\perp}v_l$,
169: where $k_{\perp}$ and $k_{\|}$ are wave numbers perpendicular
170: and parallel to the external dc field\footnote{Later in this paper, we
171: will use $\tilde{k}_{\perp}$ and $\tilde{k}_{\|}$,
172: instead of $k_{\perp}$ and $k_{\|}$, to represent
173: their scaling relation in a slightly different,
174: yet we believe equivalent, viewpoint.
175: Here, $\tilde{k}_{\perp}$ and $ \tilde{k}_{\|}$
176: are wavenumbers with respect to the direction of the local magnetic field,
177: not the external field.}.
178: The argument given
179: above for 3-wave interactions makes it clear that $k_{\perp}$
180: will tend to increase until it becomes important in the
181: plasma dynamics. The assumption of strong nonlinearity
182: implies that wave packets lose their identity after they
183: travel one wavelength along the field lines. Consequently
184: the eddy turnover time
185: ($(k_{\perp}v_l)^{-1}$) is actually the same as Alfv\'{e}nic time
186: ($(k_{\parallel}V_A)^{-1}$).
187: In this model, the cascade time, $t_{cas}(l)$ can be determined
188: without ambiguity:
189: $t_{cas} \approx (k_{\perp}v_l)^{-1} \approx (k_{\parallel}V_A)^{-1}$.
190: Since the cascade time is comparable to the period of the Alfv\'{e}n wave,
191: the 3-wave resonant condition can be violated according to the
192: uncertainty relation
193: $\Delta \omega \cdot \Delta t \sim V_A k_{\|} \cdot t_{cas}
194: \sim 1$.
195: The quantity $k_{\|}$ is the width of the active region in
196: wavevector space.
197: {}Finally the assumption of a scale-independent cascade rate
198: $\epsilon \sim v_l^2/t_{cas}(l)\sim E_{waves}V_A/L$ gives
199: \bea
200: k_{\parallel} \sim k_{\perp}^{2/3} L^{-1/3}
201: \left({E_{waves}\over V_A^2}\right)^{1/3}, \\
202: v_l \sim V_A(k_{\perp}L)^{-1/3}
203: \left({E_{waves}\over V_A^2}\right)^{1/3},
204: \eea
205: where $E_{waves}$ is the wave energy per mass. These
206: formulae assume that all scales, from $L$ on down, are
207: within the inertial range of MHD turbulence.
208: Here the typical $k_{\|}$ should be interpreted as the
209: size of the range of parallel wavevectors, corresponding to a given
210: $k_{\perp}$, that contain significant energy.
211:
212: Matthaeus et al. (1998) recently tested this model
213: numerically, and showed that the anisotropy of low frequency
214: MHD turbulence scales linearly with the ratio of
215: perturbed and total magnetic field strength ($b/B$),
216: a result which seems inconsistent with Goldreich and
217: Sridhar's model. To explain this scaling relation, they suggested
218: that the region of Fourier space where the energy transfer
219: takes place actively is given by
220: \be
221: |{\bf k}\cdot\frac{{\bf B}_0}{\sqrt{4\pi\rho}} |< \frac{1}{ \tau_{NL} },
222: \ee
223: where $\tau_{NL}$ is
224: the eddy turnover time of the energy containing length $L$.
225: Consequently, ``the region of the wave number
226: space where spectral transfer is most vigorous'' has a slab-like
227: geometry with a constant width proportional to $1/(\tau_{NL}B_0)$.
228:
229: All these theories (except the IK theory) share a common prediction for
230: anisotropy: anisotropy should be more pronounced on smaller scales.
231: Oughton et al. (1994) and Ghosh and Goldstein (1997) already reported
232: this scale-dependent anisotropy through numerical simulations.
233: The former extended Shebalin et al. (1983)'s 2-D calculations
234: to 3-D cases. To measure anisotropy,
235: they introduced the Shebalin angles, $\theta_Q$, defined by
236: $
237: \tan^2{\theta_Q}=( \sum k_{\perp}^2 |{\bf Q}({\bf k},t)|^2 )/
238: ( \sum k_{\|}^2 |{\bf Q}({\bf k},t)|^2 ),
239: $
240: where ${\bf Q}$ is vector potential {\bf A}, magnetic field {\bf B},
241: or current {\bf J}, etc.
242: Greater $\theta_Q$ means greater anisotropy.
243: They found that $\theta_A < \theta_B < \theta_J$.
244: If the energy spectrum of {\bf B} scales as $E_M(k)\propto k^{-s}$,
245: then the spectra of vector potential and current scale as
246: $E_A(k)\propto k^{-s-2}$
247: and $E_J(k)\propto k^{-s+2}$, respectively.
248: The spectra of vector potential has the steepest slope among the three spectra.
249: This means that
250: the vector potential is least strongly dependent on small scales (and
251: the current is most strongly dependent on small scales).
252: Therefore they concluded that anisotropies are more pronounced
253: at smaller scales\footnote{
254: The width of the active region in Fourier space, $k_{\|}$, is a function
255: of $k_{\perp}$. If
256: $$ \left[\frac{ k_{\parallel} }{ k_{\perp} } \right]_{\mbox{small $k_{\perp}$}}
257: >\left[\frac{ k_{\parallel} }{ k_{\perp} } \right]_{\mbox{large $k_{\perp}$}},$$
258: then we will have the ordering of Shebalin angles as observed
259: by Shebalin et al.\ and Oughton et al.
260: However, if $k_{\|} \propto k_{\perp}$, then we do not expect any ordering
261: among the angles. In \S3, we will find that,
262: no matter what the true functional form $k_{\|} = k_{\|}(k_{\perp})$ is,
263: Fourier transformation smooths out the {\it true} relation and
264: leads to a {\it fake} linear relationship between $k_{\|}$ and $k_{\perp}$.
265: This suggests that their results are in contradiction to our discussion in \S3.
266: However, the apparent linear relationship between
267: $k_{\|}$ and $k_{\perp}$ in \S3 is not perfectly linear.
268: Instead, we expect $k_{\parallel}=c_1k_{\perp}+c_2$, where $c_1$
269: is a decreasing function of $B_0$
270: and $c_2$ depends on the $k_{\|}$ of forcing terms
271: (or initial values of $k_{\|}$ for decaying turbulence).
272: The presence of $c_2$ does not seem to be important
273: in our Fig. 3 in \S3.
274: However, it does affect the calculation of Shebalin angles.
275: That is, because of $c_2$,
276: the ratio $k_{\|}/k_{\perp}=c_1 + c_2/k_{\perp}$ becomes a function of
277: $k_{\perp}$.
278: At the largest energy containing eddy scale, $k_{\perp} \sim c_2$ and
279: hence $k_{\|}/k_{\perp}\sim c_1 +O(1)\sim O(1)$.
280: But, at small scales, the ratio can be much smaller than
281: unity. Therefore, because of $c_2$, we can obtain
282: a scale-dependent anisotropy:
283: $(k_{\|}/k_{\perp})$ at small $k_{\perp}$ is greater than that at large
284: $k_{\perp}$.
285: Note that this is a result of the initial conditions, rather than a true
286: scaling relation.
287: This will lead the ordering of the angles as
288: observed by Shebalin et al.\ and Oughton et al..
289: This interpretation is qualitatively consistant with their results.
290: For example. the ratio $k_{\|}/k_{\perp}=c_1 + c_2/k_{\perp}^{peak}$,
291: therefore $(\tan{\theta_Q})^{-1}$,
292: approaches to a constant value $c_2/k_{\perp}^{peak}$
293: as $B_0$ becomes strong.
294: Here $k_{\perp}^{peak}$ is the
295: wavenumber of the peak of the energy spectrum.
296: It is also possible to explain the increased anisotropy
297: at high magnetic Reynolds
298: numbers.
299: If the magnetic Reynolds number increases, then $k_{\perp}^{peak}$ increases
300: and,
301: therefore, the ratio decreases. }.
302: They also found a similar ordering for velocity field (and vorticity, etc).
303: On the other hand, Ghosh and Goldstein (1997) calculated the Shebalin angles
304: as a function of wavenumber bin.
305: They found that the Hall-MHD\footnote{Hall MHD includes
306: the Hall term, which is important at ion-cyclotron
307: scales. In this paper, we consider only the standard MHD equations.}
308: simulations show
309: increased anisotropy at small scales (i.e. greater Shebalin angles
310: at smaller scales). However their standard MHD simulations
311: do not show increased anisotropies at small scales.
312: We refrain from comparing their work with ours because they used different
313: physics (the Hall effect) and their
314: simulations are $2\slantfrac{1}{2}$ dimensional.
315: The simulations given in this paper are 3-D standard MHD simulations.
316: We note that none of the previous papers quantitatively compared their
317: results with particular theories of anisotropy.
318:
319: In this paper, we examine the scaling law for Alfv\'{e}nic MHD
320: turbulence numerically and resolve the controversy concerning
321: the anisotropic structure of the turbulence.
322: Our results are consistent with Goldreich and Sridhar's
323: model. We describe our numerical methods in \S 2.
324: In \S 3, we describe our results for anisotropy in wavevector space.
325: In this section, we demonstrate that none of the scaling laws mentioned
326: above agrees with our data, and explain why a straightforward
327: evaluation of the distribution of spectral power does not
328: correspond to a physically meaningful set of scaling laws.
329: We attribute this to the systematic effects of large scale
330: curvature of the magnetic fields.
331: In \S 4, we determine the shape of individual eddies, avoiding the
332: systematic error described above.
333: We compare our results to Goldreich and Sridhar's model
334: and give our conclusions in \S 5.
335:
336:
337: \section{Numerical Methods and General Results}
338: We have employed a pseudospectral code to solve the
339: incompressible MHD equations in a periodic box of size $2\pi$:
340: \begin{equation}
341: \frac{\partial {\bf V} }{\partial t} = (\nabla \times {\bf V}) \times {\bf V}
342: -V_{A0}^2 (\nabla \times {\bf B})
343: \times {\bf B} + \nu \nabla^{2} {\bf V} + {\bf f} + \nabla P' ,
344: \label{veq}
345: \end{equation}
346: \begin{equation}
347: \frac{\partial {\bf B}}{\partial t}= {\bf B} \cdot \nabla {\bf V}
348: - {\bf V} \cdot \nabla {\bf B} + \eta \nabla^{2} {\bf B} ,
349: \label{beq}
350: \end{equation}
351: where $\bf{f}$ is random driving force
352: and $P'\equiv P + {\bf V}\cdot {\bf V}/2$.
353: Other variables have their usual meaning.
354: $V_{A0}=B_0/(4\pi\rho)^{1/2}$ is
355: the Alfv\'{e}n velocity of
356: the background field, which is set to be order unity in our simulations.
357: In pseudo spectral methods, we calculate the temporal evolution of
358: the equations (\ref{veq}) and (\ref{beq}) in Fourier space.
359: To obtain the Fourier components of nonlinear terms, we first calculate
360: them in real space, and transform back into Fourier space.
361: We use 21 forcing components with $2\leq k \leq \sqrt{12}$.
362: Each forcing component has correlation time of one.
363: The peak of energy injection occurs at $k\approx 2.5 $.
364: The amplitudes of the forcing components are tuned to ensure $V \approx 1$.
365: Since we expect the perturbed magnetic field strength $b$ to be
366: comparable to or less than $V$ (i.e. $b \leq 1$),
367: the resulting turbulence can be regarded as strong MHD turbulence
368: (i.e. $V \sim B_0 \sim B$).
369: %Each forcing component consists of two parts: linearly polarized component
370: %and small circularly polarized component with preferred helicity.
371: %The latter is necessary to provide non-zero helicity injection to
372: %the turbulence.
373: The average helicity in these simulations is not zero. However,
374: our results are insensitive to the value of helicity, possibly
375: because the box size is only slightly bigger than the energy
376: injection scale.
377: We use an appropriate projection operator to calculate
378: $\nabla P'$ term in
379: {}Fourier space and also to enforce divergence-free condition
380: ($\nabla \cdot {\bf V} =\nabla \cdot {\bf B}= 0$).
381: We use up to $256^3$ collocation points.
382: We use integration factor technique for kinetic and magnetic dissipation terms
383: and leap-frog method for nonlinear terms.
384: We eliminate $2\Delta t$ oscillation of the leap-frog method by using
385: an appropriate average.
386: At $t=0$, magnetic field has only uniform component and velocity has a support
387: between $2\leq k \leq 4$ in wavevector space.
388:
389:
390:
391: We use hyperviscosity and hyperdiffusivity for dissipation terms
392: to maximize the inertial range.
393: The only exception is Run 256P-$B_0$1, where we use physical viscosity and diffusivity.
394: The power of hyperviscosity
395: is set to 8, so that the dissipation term in the above equation
396: is replaced with
397: \be
398: -\nu_8 (\nabla^2)^8 {\bf V},
399: \ee
400: where we set the value of $\nu_8$ from the condition $\nu_h (N/2)^{2h} \Delta t
401: \approx 0.5$ (see Borue and Orszag 1996).
402: Here $\Delta t$ is the time step and $N$ is the number
403: of grids in each direction.
404: We use exactly same expression for the magnetic dissipation term.
405: We list parameters used for the simulations in Table 1.
406: We use the notation 256X-Y, where X = H or P refers to hyper- or physical viscosity;
407: Y = $B_0$1 or $B_0$0.5 refers to the strength of the external magnetic fields.
408:
409: {}Fig. 1 shows the evolution of $V^2$ and $B^2$ from
410: Runs 256H-$B_0$1 and REF2.
411: Results of 256H-$B_0$1 are plotted as solid lines, and results of REF2 are plotted
412: as dotted lines. Both runs have similar results for the overlapped time,
413: which means that the values of $V^2$ and $B^2$ are not sensitive to the spatial resolution.
414: The magnetic energy density grows fast at the beginning of the simulations,
415: as a result
416: of the stretching of external field lines by turbulent motion.
417: Since the external field is strong, magnetic forces
418: soon become strong enough to balance the stretching effect.
419: This balance happens at $t \sim 1$ and, after a transient stage ($1\leq t\leq 5$),
420: the turbulence reaches the saturation stage ($t\geq 5$).
421: In this case the saturation stage begins quite early,
422: since the external uniform magnetic field
423: is strong and the magnetic diffusivity is effectively 0.
424: In general, when the external field is weaker and/or magnetic
425: diffusivity is larger, the saturation stage occurs later.
426: At the saturation stage, $B^2 \sim 1.45$ and $V^2 \sim 0.6$.
427: Since $b^2 = B^2 - B_0^2 (\equiv 1) \approx 0.45$,
428: there is a rough energy equipartition between $b$ and $V$.
429: Note that, since $V \sim B$, the condition for strong MHD turbulence is
430: met.
431:
432: \section{Fourier Analysis: A False Scaling Law?}
433: \subsection{Fourier Analysis}
434:
435:
436:
437: In this subsection, we investigate the spectral
438: structure of Alfv\'{e}nic turbulence.
439: It is important to note that we lose some
440: information
441: about individual eddies when we perform a global transformation,
442: such as the Fourier transformation.
443: In particular, the scaling law given in this subsection may $not$ be
444: true for individual eddies (See \S 3.2 for details).
445:
446: We plot the 1-dimensional energy spectra $E_K(k)$ and $E_M(k)$ in Fig. 2.
447: Both spectra peak at the same $k$,
448: which is also the scale of the energy injection.
449: This reflects that there is a rough energy equipartition between $b$ and $V$ at
450: the largest energy containing scale.
451: In general, $E_K(k)$ always peaks at the energy injection scale.
452: However, $E_M(k)$ peaks at a larger k than $E_K(k)$ when
453: the external field is weak.
454: The inertial range exhibits of two different scaling ranges:
455: a small $k$ range where the slopes are steep
456: and a large $k$ range where the slopes are mild.
457: We believe that the former reflects a true inertial range,
458: whereas the latter is a result of the $bottleneck$ effect.
459: The $1/k$ bottleneck effect is a common feature in
460: numerical hydrodynamic simulations with hyperviscosity
461: (see Borue and Orzag 1996, 1995).
462: Interestingly enough, the slope of the bottleneck is actually steeper than
463: $-1$.
464: This might mean that the bottleneck effect is less serious in the
465: MHD case.
466: The kinetic and magnetic energy spectra have slightly different
467: powers in the inertial range.
468: Although the power indexes of both spectra are close to
469: $-5/3$ in the (true) inertial range,
470: the data are also compatible with IK theory, where the power index is $-3/2$.
471:
472: We plot normalized 3-dimensional energy spectra of Run 256H-$B_0$1 in Fig. 3.
473: In the figure, we plot contours of same
474: \be
475: E_3(k_{\perp}, k_{\parallel}) / E_3(k_{\perp}, 0)
476: \ee
477: in the $ (k_{\perp},k_{\parallel})$ plane.
478: As the energy cascades
479: in the directions perpendicular to the mean field, this ratio drops
480: as we move away from the $k_{\perp}$ axis.
481: The figure implies most of the energy is confined in a region around the
482: $k_{\perp}$ axis.
483: The thickness of the region depends on the strength
484: of the external magnetic fields (Fig. 4).
485: We use the contours of $ E_3(k_{\perp}, k_{\parallel}) / E_3(k_{\perp}, 0)=0.5$
486: to measure the thickness. The angle $\Theta_{0.5}$ is the angle
487: formed by the contours and the horizontal axis.
488: We can see that $\tan \Theta_{0.5}$ ($= k_{\|}/k_{\perp}$)
489: is proportional to $b/B_0$.
490: This result confirms the scaling relation found by Matthaeus et al. (1998).
491: Both velocity and magnetic fields have very similar structures.
492: {}From the fact that $k_{\|} < k_{\perp}$ in the active region (see Fig. 3),
493: one can conclude that eddies do have anisotropic structure: eddies
494: are stretched along the direction of the mean field.
495: Apparently, Fig. 3 suggests that $k_{\|} \propto k_{\perp}$ and, hence, that
496: anisotropy is scale-independent.
497: No theory mentioned in \S1 agrees with our result.
498: However, it is not clear from the figure whether or not
499: the true anisotropy is a function of scale:
500: although the contours show a linear relationship between $k_{\|}$ and $k_{\perp}$,
501: the relationship doesn't mean that
502: all individual eddies have the same major axis to minor axis ratio.
503: As explained in next subsection, the
504: $rotation$ of eddies by large-scale waves in the magnetic fields
505: can distort the actual scaling relation
506: and lead to the linear relationship shown in the figure.
507:
508: {}Fig. 5 shows $t_{phase}$ as a function of wavenumber $k$.
509: This time scale is defined as the average correlation time $\Delta t$
510: such that Fourier components at ${\bf k} = k$
511: have a phase shift of $60^o$ with respect to the original phase.
512: We plot the result for zero frequency modes (i.e. $k_{\parallel}$=0
513: modes). We see that $t_{phase} \sim k^{-1}$ (that is, $k_{\perp}^{-1}$).
514: How can we interpret this result? If a turbulent structure,
515: characterized by a wavenumber $k$, moves a distance $l$
516: then the phase is shifted by roughly $kl$, even if the eddy
517: is unaffected by the motion. This implies that a large
518: scale velocity $V$ will change the phase at a rate $kV$, so
519: that $t_{phase}\propto k^{-1}$. This implies that our
520: calculation of $t_{phase}$ is dominated by large scale
521: motions rather than by the local cascade of energy, as long
522: as the nonlinear cascade rate is proportional to $k$ to some
523: power less than one, which is generally the case. This is
524: an example of how large-scale fluctuations can complicate attempts
525: to find physically meaningful scaling relations.
526:
527: \subsection{Rotation Effect}
528: In the previous subsection, we showed that anisotropy appears to be
529: scale-independent.
530: In this subsection, we will show that this apparent scaling relation
531: is an artifact caused by the Fourier transformation.
532: %Apparently, neither Goldreich and Sridhar (1995) nor
533: %Matthaeus et al. (1998) satisfies both Fig. 3 and Fig. 4.
534: That is,
535: we will demonstrate that large-scale modes in the magnetic field can
536: distort the actual scaling law at smaller scales
537: when we perform a Fourier transformation.
538: In this manner, a straightforward evaluation of anisotropy in
539: Fourier space is strongly contaminated by the curvature of the large-scale
540: magnetic fields and does not reflect the actual local anisotropy
541: when anisotropy is more pronounced at smaller scales.
542: Consequently, figures 3 and 4 are compatible with
543: any scaling law that predicts that smaller eddies are more elongated.
544:
545: {}Fig. 3 implies that eddies have anisotropic shapes: on average,
546: eddies are stretched along the direction of ${\bf B}_0$.
547: However, not all eddies are aligned along the large scale field.
548: The elongation of an eddy is determined by its interaction
549: with the $local$ magnetic field, not the background
550: field.
551: Since the large-scale magnetic field lines wander with respect to
552: ${\bf B}_0$, all smaller scale eddies have similar angular
553: distributions around $ {\bf B}_0$.
554: We illustrate this effect in Fig. 6.
555:
556: In this way, we can explain the results of Goldreich and
557: Sridhar (1995) and Matthaeus et al. (1998) simultaneously.
558: Suppose that eddies are oriented according to the local
559: field lines (Fig. 6a).
560: Goldreich and Sridhar's result implies smaller eddies are
561: relatively more elongated.
562: When we perform a Fourier transformation and measure the ratio of
563: $k_{\|}/k_{\perp}$,
564: what we actually measure is not the ratio of the minor axis
565: to the major axis of individual eddies.
566: Instead, because the direction of the local magnetic field varies
567: according to location and Fourier transformation is none other than
568: a (weighted) averaging process,
569: we actually measure the ratio averaged over all possible orientation of
570: the eddies (Fig. 6b).
571: The Fourier transformation sees that $L_1$ (and $l_1$) is the minor axis
572: and $L_2$ (and $l_2$) is the major axis of an eddy.
573: The ratio of $L_1/L_2$ ($=l_1/l_2$)
574: is determined by the degree of
575: the {\it wandering} of the large scale magnetic field lines with respect to
576: ${\bf B}_0$ and, therefore, the
577: ratio is nearly constant for all eddies, regardless of their sizes and shapes.
578: In fact, the ratio will depend on the tangent of the angle
579: between ${\bf B}_0$ and ${\bf B}$.
580: Since $k_{\perp} \propto 1/L_1$ and $k_{\|} \propto 1/L_2$,
581: the {\it measured} value of
582: the ratio $ k_{\|}/k_{\perp}$ is nearly scale-independent.
583: We expect the angle $\theta$ ($\equiv 90^o-\theta_w$)
584: between ${\bf B}_0$ and ${\bf B}$
585: to be
586: \be
587: \tan{\theta} = b/B_0.
588: \ee
589: Therefore, we have
590: \be
591: \sin{\theta} = b/B =\cos{\theta_w},
592: \ee
593: which is exactly the scaling relation found by Matthaeus et al.
594:
595: \section{Measuring Eddy Shapes}
596:
597: In this section, we analyze the shape of eddies in real space.
598: As explained in the previous section, we assume that elongated eddies are
599: aligned in the direction of the {\it local} magnetic fields.
600: If we want to visualize the shape of eddies, we
601: need to first identify the direction of the {\it local} magnetic fields.
602: It is important to note that, although we use the term
603: `local magnetic fields' for simplicity, the fields are different from
604: ${\bf B}({\bf r})$.
605: Let us consider an eddy of size $l$.
606: The `local magnetic field' of the eddy must act as the `large-scale magnetic
607: field' for the eddy.
608: Therefore, the `local magnetic fields' for eddies of size $l$, must be
609: smoothly varying vector fields whose characteristic length of
610: variation is $>l$.
611: Note also that another eddy of size $l^{\prime}$ ($\neq l$) at the same
612: location can have a slightly different direction for the
613: `local magnetic field.'
614: The `local magnetic fields' are functions of position (${\bf r}$)
615: and eddy size ($l$).
616: Then, how can we define the direction of the {\it local} magnetic fields?
617: We implement
618: 2 independent numerical algorithms to calculate the
619: direction of the $local$ magnetic fields.
620:
621:
622: In the first method, the local magnetic fields are calculated by
623: \be
624: {\bf B}_l = ( {\bf B}({\bf r}_1) +{\bf B}({\bf r}_2) )/2
625: \ee
626: and the second order structure functions for $v$ and $b$ are given by
627: \bea
628: F_2^v (R,z)= < |{\bf V}({\bf r}_1) -{\bf V}({\bf r}_2)|^2 >,\\
629: F_2^b (R,z)= < |{\bf b}({\bf r}_1) -{\bf b}({\bf r}_2)|^2 >,
630: \eea
631: where $R=|\hat{ {\bf z} }\times ({\bf r}_2-{\bf r}_1 )|,$
632: $z=\hat{ {\bf z} }\cdot ({\bf r}_2-{\bf r}_1 )$ and
633: $\hat{ {\bf z} }= {\bf B}_l/|{\bf B}_l|.$
634: That is, $R$ and $z$ are coordinates in a cylindrical coordinate system
635: in which the z-axis is parallel to ${\bf B}_l$ (Fig. 7).
636: ${\bf B}({\bf r})$ and ${\bf b}({\bf r})$ are the total and
637: perturbed magnetic fields at a point ${\bf r}$. As usual, brackets
638: denotes a spatial average.
639:
640:
641:
642: In Fig. 8 we plot the second order structure functions in z-R plane.
643: The horizontal axis (z-axis) is parallel to ${\bf B}_l$.
644: The contours reflect
645: the shapes of the eddies. Eddies are clearly elongated
646: along the local field lines, and
647: smaller eddies are more elongated.
648: The velocity and magnetic fields show different structure at
649: large scales.
650: However, their small scale structure is quite similar.
651:
652: In Fig. 9, we plot R-intercepts and z-intercepts of the contours.
653: The R-intercept and z-intercept of a given contour can be regarded as
654: a measure of $ \tilde{k}_{\perp}$ and $ \tilde{k}_{\|}$
655: for the corresponding eddy scale\footnote{We interpret $ \tilde{k}_{\|}$
656: as the inverse of the major axis of eddies and
657: $ \tilde{k}_{\perp}$ as that of the minor
658: axis. Since the major axis is assumed to be parallel to the local
659: magnetic fields, $ \tilde{k}_{\|}$ is the parallel wavenumber
660: with respect to the {\it local} magnetic field direction.
661: On the other hand, in \S3.1, $k_{\|}$ is parallel to
662: ${\bf B}_0$. Therefore, $ \tilde{k}_{\|}$ and
663: $ \tilde{k}_{\perp}$ in this section
664: have different meanings from $k_{\|}$ and $k_{\perp}$ in \S3.1.}.
665: {}Fitting the results for velocity fields gives
666: \begin{equation}
667: v: \tilde{k}_{\parallel} \sim \left\{ \begin{array}{ll}
668: \tilde{k}_{\perp}^{0.69}, & (256H-B_00.5) \\
669: \tilde{k}_{\perp}^{0.70}, & (256H-B_01) \\
670: \tilde{k}_{\perp}^{0.73}, & (256P-B_01)
671: \end{array} \right.
672: \end{equation}
673: On the other hand, for the magnetic fields we find
674: \be
675: b: \tilde{k}_{\parallel} \sim \left\{ \begin{array}{ll}
676: \tilde{k}_{\perp}^{0.64}, & (256H-B_00.5) \\
677: \tilde{k}_{\perp}^{0.50}, & (256H-B_01) \\
678: \tilde{k}_{\perp}^{0.53}, & (256P-B_01)
679: \end{array} \right.
680: \ee
681: The velocity fields show good agreement with the scaling relation,
682: $ \tilde{k}_{\|} \sim \tilde{k}_{\perp}^{2/3}$,
683: proposed by Goldreich and Sridhar (1995).
684: The power indices are insensitive to the strength of $B_0$ or the form
685: of viscosity (and diffusivity).
686: However, the magnetic field shows different scaling behavior.
687: When the external field is moderately strong (256H-$B_00.5$),
688: the power index is very close to 2/3.
689: On the other hand, for stronger external fields (256H-$B_01$ and
690: 256P-$B_01$), the power indices are smaller than 2/3.
691: The results are insensitive to the choice of viscosity (and diffusivity).
692: It is not clear whether the existence of a separate scaling law for the
693: magnetic field represents a physical effect not included in Goldreich
694: and Sridhar's model, e.g. the first signs of small scale intermittency
695: in the magnetic field distribution, or merely the failure of the
696: numerical models used here to fully resolve the inertial range of
697: strong MHD turbulence.
698:
699: In the second method, we employ a completely different approach.
700: We obtain the local large scale magnetic fields by filtering
701: out the small scale magnetic fields:
702: \be
703: {\bf B}_{\sigma} ({\bf r}) =
704: \sum_{ {\bf r}^{\prime} }
705: {\bf B}( {\bf r}^{\prime} ) \phi( |{\bf r}-{\bf r}^{\prime} | ),
706: \ee
707: where $\phi(r) \propto \exp(-r^2/\sigma_r^2) $ is a gaussian function.
708: To determine the shape of small scale eddies (i.e. eddies smaller than
709: the filter size, $\sim \sigma_r$), we consider the following quantities:
710: \bea
711: {\bf v}_{\sigma}({\bf r}) = {\bf V}({\bf r}) - {\bf V}_{\sigma} ({\bf r}),\\
712: {\bf b}_{\sigma}({\bf r}) = {\bf B}({\bf r}) - {\bf B}_{\sigma} ({\bf r}),
713: \eea
714: where ${\bf V}_{\sigma}$ and ${\bf B}_{\sigma}$ are filtered fields
715: (cf. eq. (20)).
716: The fields ${\bf v}_{\sigma}({\bf r})$ and ${\bf b}_{\sigma}({\bf r})$
717: represent small scale fluctuation of velocity and magnetic fields, the shape
718: of which can be regarded as an adequate approximation of small scale eddies.
719: {}From these two fields we calculate
720: the structure functions
721: \bea
722: F^v(R,z) = | {\bf v}_{\sigma}({\bf r}_2)-{\bf v}_{\sigma}({\bf r}_1) |,\\
723: F^b(R,z) = | {\bf b}_{\sigma}({\bf r}_2)-{\bf b}_{\sigma}({\bf r}_1) |,
724: \eea
725: where $R$ and $z$ are similarly
726: defined as in the first method with
727: $\hat{ {\bf z} } \parallel {\bf B}_{\sigma} ({\bf r}_1)$ (Fig. 10).
728:
729: In Fig. 11, we plot the results of the second method.
730: We can clearly observe the flattening effect:
731: When the filter size is large, eddies are less anisotropic.
732: When filter size is small, eddies show highly anisotropic structure.
733: In the figure, we plot only the results for magnetic fields.
734: Velocity fields show similar trends.
735: The thick contours represent $F^b(R,z) = <|{\bf b}_{\sigma}|^2>^{1/2}$.
736: The values next to the thick contours are $ <|{\bf b}_{\sigma}|^2>^{1/2}$.
737:
738:
739: This second method is useful for visualizing small scale eddies,
740: but may not be as useful for quantitative analysis.
741: The difficulty comes from the fact that there is no well
742: defined eddy scale associated with filtered fields
743: ${\bf v}_{\sigma}({\bf r})$
744: and ${\bf b}_{\sigma}({\bf r})$.
745:
746: In this paper, we will not pursue quantitative analysis using the method 2.
747: Instead, we just wish to point out
748: that both methods describe the same scaling law: smaller eddies are
749: relatively more stretched along the local magnetic field lines.
750: In particular, the results from the first method are consistent with the relation
751: $ \tilde{k}_{\|} \sim \tilde{k}_{\perp}^{2/3}$
752: proposed by Goldreich and Sridhar (1995).
753:
754:
755:
756: \section{Discussion and Conclusions}
757:
758: Here we rederive the scaling law
759: $ \tilde{k}_{\parallel} \propto \tilde{k}_{\perp}^{2/3}$
760: in the framework of 3-wave interactions.
761: Except for the use of the uncertainty principle,
762: the work in this section is independent of Goldreich and Sridhar's
763: derivation.
764: As noted by Goldreich and Sridhar (1995), 3-wave interactions are an
765: adequate
766: proxy for wave-wave interactions of all orders in a strong MHD
767: turbulence.
768: As long as we assume the locality of interactions, it is pointless
769: to distinguish $k_{\|}$ and $k_{\perp}$ from
770: $\tilde{k}_{\|}$ and $\tilde{k}_{\perp}$.
771: Hence, for simplicity, we use $k_{\|}$ and $k_{\perp}$
772: instead of $\tilde{k}_{\|}$ and $\tilde{k}_{\perp}$ during the derivation.
773:
774: Suppose the 3-dimensional energy spectrum is given by
775: \begin{equation}
776: E_3(k_{\perp},k_{\parallel})\equiv |\hat{{\bf V}}({\bf k})|^2
777: \propto k_{\perp}^{-2\alpha} f,
778: \end{equation}
779: where $\hat{{\bf V}}({\bf k})$ is the amplitude of the mode whose wavevector
780: is ${\bf k}$ and $f(u)$ is a positive, symmetric function of $u$
781: (cf. equation (7) of Goldreich and Sridhar (1995)) which describes
782: the power distribution as a function of eddy shape.
783: If the $width$ (or, $thickness$) of the energy spectrum
784: in the direction of $k_{\parallel}$
785: is $k_{\perp}^{\beta}$, then we can write
786: \be
787: E_3(k_{\perp},k_{\parallel})
788: \propto k_{\perp}^{-2\alpha} f(k_{\parallel}/k_{\perp}^{\beta}).
789: \ee
790: If the $width$ is caused by the uncertainty principle
791: ($\Delta t \cdot \Delta \omega \approx 1$
792: with $\Delta t \propto t_{cas}(l)$ and $\Delta \omega \propto
793: k_{\parallel}$), then
794: \be
795: t_{cas}(l) \propto k_{\perp}^{-\beta}.
796: \ee
797:
798: Suppose the energy cascade rate
799: $\epsilon \sim v_l^2/t_{cas}(l)$, where $l = 2\pi/k_{\perp}$,
800: is scale-independent.
801: Because $v_l^2 \sim k_{\perp}^{-2\alpha} k_{\perp}^2 k_{\perp}^{\beta}$
802: ($\sim k_{\perp} E(k_{\perp})$, $E(k_{\perp})=$ 1-dimensional spectrum)
803: and $t_{cas}(l) \sim k_{\perp}^{-\beta}$, we have
804: \be
805: k_{\perp}^{-2\alpha+2+2\beta} = const.
806: \ee
807: Therefore,
808: \begin{equation}
809: 1-\alpha + \beta = 0.
810: \label{c1}
811: \end{equation}
812:
813: Now, let's pick up a mode at a wavevector ${\bf p}$ and consider nonlinear
814: interactions with
815: other wave modes. First, the strength of the interaction with another
816: mode at ${\bf q}$ is
817: $ \propto p |\hat{ {\bf V} }({\bf p})| |\hat{ {\bf V} }({\bf q})|$.
818: (This comes from the $(\nabla \times {\bf V}) \times {\bf V}$ term in equation (9))
819: Hereafter we assume $p\equiv |{\bf p}|\approx p_{\perp}$.
820: Second, the number of interactions is $\propto p^2 p^{\beta}$.
821: This is the number of modes $near$ ${\bf p}$. Here we use locality of 3-wave
822: interactions.
823: If the interactions are random,
824: the net change of amplitude per unit time will be the
825: strength of the interaction times the square root of
826: the number of interactions, or
827: \be
828: |\Delta \hat{ {\bf V} }({\bf p})| \propto
829: p |\hat{ {\bf V} }({\bf p})| |\hat{ {\bf V} }({\bf q})|\cdot
830: (p^2 p^{\beta})^{1/2}.
831: \ee
832: Therefore, we have
833: \be
834: t_{cas} \propto |\hat{ {\bf V} }({\bf p})|/|\Delta \hat{ {\bf V} }({\bf p})|
835: \propto p^{-2} p^{\alpha} p^{-\beta/2},
836: \ee
837: where we assumed $p\propto q$.
838: Equating this with $t_{cas}\propto p^{-\beta}$, we can write
839: \begin{equation}
840: \alpha - 2 = -\beta/2.
841: \label{c2}
842: \end{equation}
843: {}From equations (\ref{c1}) and (\ref{c2}), we have
844: \begin{equation}
845: \alpha=5/3, \beta=2/3,
846: \end{equation}
847: which is just the result of Goldreich and Sridhar (1995):
848: \be
849: k_{\parallel} \propto k_{\perp}^{2/3}.
850: \ee
851: As a consequence, the 3-D energy spectrum becomes
852: \be
853: E_3(k_{\perp},k_{\parallel})\propto k_{\perp}^{-10/3}
854: f(k_{\parallel}/k_{\perp}^{2/3})
855: \ee
856: and the corresponding 1-D spectrum is given by
857: \be
858: E(k) \propto k^{-5/3}.
859: \ee
860:
861:
862: In summary, we have shown that the
863: anisotropy of Alfv\'{e}nic turbulence depends on
864: the spatial scales of eddies. In particular, our results confirm
865: the claim by Goldreich and Sridhar (1995, 1997) that
866: smaller eddies are relatively more elongated along the direction of the
867: local magnetic field lines than larger ones.
868: Quantitative measurements of the anisotropy using the velocity fields
869: show good agreement with
870: their proposed scaling law, $k_{\parallel}\sim k_{\perp}^{2/3}$ as long
871: we interpret these wavenumbers as referring to the
872: {\it local} magnetic field direction.
873: However, when the external magnetic field is very strong,
874: magnetic fields scale somewhat differently, showing a slightly more
875: rapid increase in anisotropy at smaller scales.
876:
877: It is important to note that the correct scaling laws depend
878: on comparing the eddy shape to the {\it local} magnetic
879: field direction.
880:
881: As a final note, we wish to stress that our results are not in agreement
882: with the IK theory. The IK theory is based on the assumption of isotropy in
883: wavenumber space, which may be true when the external magnetic field is very
884: weak or zero. However, even in these cases, the turbulence is globally isotropic,
885: but locally very anisotropic. In this paper, we showed that eddies do show
886: anisotropy and that the anisotropy is scale-dependent when there is a strong
887: large scale field (which should apply to very small scales within any MHD
888: turbulence cascade). On the other hand, our results
889: are consistent with Goldreich and Sridhar's theory of strong MHD turbulence.
890: More precisely, if we consider the ratio of hydrodynamic to Alfv\'{e}nic
891: rates, that is $(k v_k/ \tilde{k}_{\|} V_A)$, we find from equations (18)
892: and (19) that
893: \begin{equation}
894: {k v_k\over \tilde{k}_{\|} V_A}\propto \tilde{k}_{\perp}^{0.3-0.5} v_k,
895: \end{equation}
896: where $k\approx \tilde{k}_{\perp}$.
897: {}From Fig. 2 we see that for the inertial range this implies a ratio
898: which is either constant or increasing with wavenumber. A constant
899: ratio is predicted by Goldreich and Sridhar's model. The IK model
900: predicts a slow decline.
901:
902: \acknowledgements
903: This work was partially supported by National Computational Science
904: Alliance under CTS980010N and utilized the NCSA SGI/CRAY Origin2000.
905:
906:
907: \begin{references}
908:
909: \reference{} Borue, V. and Orszag, S.A., 1995,
910: Europhys. Lett., 29(9), 687
911: \reference{} Borue, V. and Orszag, S.A., 1996, J. Fluid Mech., 306, 293
912: \reference{} Ghosh, S. and Goldstein, M.L., 1997, J. Plasma Physics, 57, 129
913: \reference{} Goldreich, P. and Sridhar, H., 1995, ApJ, 438, 763
914: \reference{} Goldreich, P. and Sridhar, H., 1997, ApJ, 485, 680
915: \reference{} Iroshnikov, P., 1963, Astron. Zh., 40, 742
916: [1963, Sov. Astron., 7, 566]
917: \reference{} Kraichnan, R., 1965,
918: Phys. Fluids, 8, 1385
919: \reference{} Matthaeus, W.H., Oughton, S., Ghosh, S. and Hossain, M.,
920: 1998, Phy. Rev. Lett. 81, 2056
921: \reference{} Montgomery, D.C. and Matthaeus, W.H., 1995, ApJ, 447, 706
922: \reference{} Oughton, S., Priest, E.R. \& Matthaeus, W.H. 1994,
923: J. Fluid Mech., 280, 95
924: \reference{} Shebalin J.V., Matthaeus, W.H. and Montgomery, D., 1983,
925: J. Plasma Phys., 29, 525
926: \end{references}
927:
928: \onecolumn
929: \clearpage
930:
931: \begin{deluxetable}{ccccl}
932: %\footnotesize
933: \tablecaption{Parameters}
934: %\tablewidth{0pt}
935: \tablehead{
936: \colhead{Run \tablenotemark{a}} & \colhead{$N^3$} & \colhead{$\nu$} & \colhead{$\eta$} &
937: \colhead{$B_0$}
938: }
939: \startdata
940: REF1 & $144^3$ & $3.20 \times 10^{-28}$ & $3.20 \times 10^{-28}$ & 0.5
941: \nl
942: REF2 & $144^3$ & $3.20 \times 10^{-28}$ & $3.20 \times 10^{-28}$ & 1
943: \nl
944: REF3 & $144^3$ & $3.20 \times 10^{-28}$ & $3.20 \times 10^{-28}$ & 2
945: \nl
946: REF4 & $144^3$ & $3.20 \times 10^{-28}$ & $3.20 \times 10^{-28}$ & 3
947: \nl
948: 256H-$B_0$0.5 & $256^3$ & $6.42 \times 10^{-32}$ & $6.42 \times 10^{-32}$ & 0.5
949: \nl
950: 256H-$B_0$1 & $256^3$ & $6.42 \times 10^{-32}$ & $6.42 \times 10^{-32}$ & 1
951: \nl
952: 256P-$B_0$1 & $256^3$ & 0.001 & 0.001 & 1
953: \nl
954: \enddata
955:
956: \tablenotetext{a}{ For $256^3$ grids, we use the notation 256X-Y,
957: where X = H or P refers to hyper- or physical viscosity;
958: Y = $B_0$1 or $B_0$0.5 refers to the strength of the external magnetic fields.}
959:
960: \end{deluxetable}
961:
962:
963: \clearpage
964: \figcaption{Time evolution of $V^2$ and $B^2$ (Solid lines, 256H-$B_0$1;
965: Dotted lines, REF2). $B_0= 1$ for both runs.}
966: \figcaption{256H-$B_0$1. Kinetic energy spectrum ($E_K(k)$)
967: and magnetic energy spectrum ($E_M(k)$). For $2\leq k\leq 20$, spectra
968: are compatible with $k^{-5/3}$. A 1/k bottleneck effect is observed
969: before the dissipation cutoff $k_d \sim 90$. }
970: \figcaption{256H-$B_0$1. 3D energy spectra normalized by the value on the
971: $k_{\perp}$ axis.}
972: \figcaption{$\tan \Theta_{0.5}$ ($=k_{\|}/k_{\perp}$) versus $b/B_0$.
973: These results are consistent with Matthaeus et al (1998) showing
974: that anisotropy ($k_{\|}/k_{\perp}$) scales linearly with $b/B_0$.}
975: \figcaption{256H-$B_0$1. Phase correlation time $t_{phase}$ as a
976: function of wavenumber, defined as the average time required for the
977: phase to shift by $60^o$. }
978: \figcaption{Rotation effect. (a) Anisotropic wave packets are moving
979: along magnetic field lines. We assume the wave packets are aligned
980: with the direction of local magnetic field lines.
981: (b) Large scale variations in the magnetic field direction causes
982: wave packets to point in different directions as a function of
983: location. We expect $L_1/L_2 = l_1/l_2$. In Fourier space,
984: $L_1$ (and $l_1$) become $\sim 1/k_{\perp}$ and $L_2$ (and $l_2$)
985: becomes
986: $\sim 1/k_{\|}$. Consequently, information about
987: eddy shapes is lost when we look at the power spectrum.}
988: \figcaption{Method 1. We adopt a cylindrical coordinate system in which
989: z-axis is parallel to
990: ${\bf B}_l = ( {\bf B}({\bf r}_1) +{\bf B}({\bf r}_2) )/2$.}
991: \figcaption{256H-$B_0$1. Visualization of eddies using method 1. Note that horizontal
992: axis is z-axis (see Fig. 7 for definition). Unit is grid spacing. Smaller eddies are more elongated.}
993: \figcaption{R-intercept (semi-minor axis; $\sim 1/ \tilde{k}_{\perp}$) versus
994: z-intercept (semi-major axis; $\sim 1/ \tilde{k}_{\|}$)
995: from Fig. 8. In 256H-$B_0$0.5, both velocity and magnetic fields follow
996: the relation $ \tilde{k}_{\|} \sim \tilde{k}_{\perp}^{2/3}$.
997: In 256H-$B_0$1 and 256P-$B_0$1, velocity fields follow the same scaling relation.
998: However, magnetic fields scale slightly differently.}
999: \figcaption{Method 2. We adopt a cylindrical coordinate system in which the
1000: z-axis is parallel to large-scale fields ${\bf B}_{\sigma}$.}
1001: \figcaption{256H-$B_0$1. Visualization of eddies using method 2. Note that the horizontal
1002: axis is the z-axis (see Fig. 10 for definition).
1003: Unit is grid spacing. Only magnetic fields are shown.
1004: Smaller eddies are more flattened.}
1005: \clearpage
1006: \begin{figure}
1007: \plotone{ani1c.eps}
1008: \end{figure}
1009: \clearpage
1010: \begin{figure}
1011: \plotone{ani2c.eps}
1012: \end{figure}
1013:
1014: \end{document}
1015: