1: %\documentclass[aps,prl,preprint,groupedaddress]{revtex4}
2: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb,floatfix]{revtex4}
3: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
4: % Some other (several out of many) possibilities
5: %\documentclass[preprint,aps]{revtex4}
6: %\documentclass[preprint,aps,draft]{revtex4}
7: %\documentclass[prb]{revtex4}% Physical Review B
8:
9: \usepackage{graphicx}% Include figure files
10: \usepackage{dcolumn}% Align table columns on decimal point
11: \usepackage{bm}% bold math
12: \usepackage{color}
13:
14: %\topmargin -3pt
15:
16: % COLOR %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
17: %\def\note#1{{\textcolor{red}{\bf [#1]}}} % note
18: %\def\opt#1{{\textcolor{green}{#1}}} % suggested deletions
19: %\def\add#1{{\textcolor{blue}{#1}}} % addition
20: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
21: % NO COLOR %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
22: \def\note#1{{{}{#1}}} % note
23: \def\opt#1{{{}{#1}}} % suggested deletions
24: \def\add#1{{{}{#1}}} % addition
25: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
26:
27:
28:
29: \begin{document}
30:
31: %\preprint{APS/123-QED}
32: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
33: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
34: \title{Shell to shell energy transfer in MHD,\\
35: Part I: steady state turbulence}% Force line breaks with \\
36: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
37: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
38: \author{Alexandros Alexakis}
39: \email{alexakis@ucar.edu}
40: \author{Pablo D. Mininni}%
41: \email{mininni@ucar.edu}
42: \author{Annick Pouquet}
43: \email{pouquet@ucar.edu}
44: \affiliation{National Center for Atmospheric Research,
45: P.O. Box 3000, Boulder, Colorado 80307}
46: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
47: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
48: \date{\today}
49: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
50:
51: \begin{abstract}
52: %\opt{The assumption of locality of transfer of energy among the different
53: %scales involved in a turbulent flow is one of the building blocks of
54: %Kolmogorov (1941) theory of turbulence . This assumption for
55: %hydrodynamic turbulence has been carried over to present models of
56: %magneto-hydrodynamical turbulence. However, if this assumption is true
57: %for conducting fluids in the presence of magnetic fields has to be
58: %tested.}
59: We investigate the transfer of energy from large
60: scales to small scales in
61: fully developed forced three-dimensional MHD-turbulence
62: by analyzing the results of direct numerical simulations
63: in the absence of an externally imposed
64: uniform magnetic field. Our results show that the transfer of
65: kinetic energy from the large scales to kinetic energy at smaller scales,
66: and the transfer of magnetic energy from the large scales to magnetic
67: energy at smaller scales, are local, as is also found in the case of neutral fluids,
68: and in a way that is compatible with Kolmogorov (1941) theory of turbulence.
69: However, the transfer of energy from
70: the velocity field to the magnetic field is a highly non-local process in Fourier space.
71: Energy from the velocity field at large scales can be transfered directly
72: into small scale magnetic fields without the participation of intermediate
73: scales. Some implications of our
74: results to MHD turbulence modeling are also discussed.
75: \end{abstract}
76:
77: \pacs{47.65.+a; 47.27.Gs; 95.30.Qd}% PACS, the Physics and Astronomy
78: % Classification Scheme.
79: %\keywords{Suggested keywords}%Use showkeys class option if keyword
80: %display desired
81: \maketitle
82:
83: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
84: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
85: \section{ \label{Intro} Introduction }
86:
87: Most astrophysical and planetary systems, e.g. solar/stellar winds,
88: accretion disks and interstellar medium, are in a turbulent state and
89: coupled to magnetic fields. Understanding and quantifying
90: the statistical properties of magnetohydrodynamical (MHD) turbulence
91: is crucial to explain many physical processes
92: in the cosmos, and in industrial flows as well \cite{Davidson}.
93: Although the
94: phenomenology of hydrodynamical (HD) turbulence is understood to some
95: extent, and the theory has been able to make predictions
96: like Kolmogorov's 4/5th law and the functional form of the energy spectrum
97: in the inertial range,
98: that have been well verified in experiments and numerical simulations, a similar
99: statement cannot be made for MHD turbulence at the same level. In MHD
100: flows, the two fields (velocity and magnetic) and two associated energies involved in the
101: dynamical processes allow for many possibilities for the energy to transfer between
102: smaller or larger scales, making the dynamics more complex
103: to address in both theory and modeling.
104:
105: We briefly describe some phenomenological aspects of HD turbulence to
106: point out some of the difficulties usually encountered when the
107: formulation of HD turbulence is applied in the MHD case. To follow
108: Kolmogorov (1941) theory \cite{K41} (hereafter, K41), we need to assume
109: a statistically isotropic and homogeneous flow in steady state in which
110: the energy is cascading from eddies of scale $l$ to smaller eddies, and
111: so on until energy reaches the dissipation scales. Since we are
112: considering a statistically steady state,
113: the flux of energy to smaller scales has to be constant. We can further assume
114: that the flux at some scale can depend only on the scale $l$ and the
115: amplitude of the velocity field $|{\bf u}_l|$ at this scale. This assumption
116: is justified by the argument that larger eddies will only advect smaller
117: eddies without significantly altering their scale
118: and only when eddies of similar size interact do they produce a cascade.
119: Therefore, only
120: ``local" interactions among the different scales control the cascade.
121: Here we use the term ``local" in terms of the different scales
122: involved (i.e. scales of similar size) and not as locality in physical space.
123: With these assumptions we obtain that the energy $|{\bf u}_l|^2$ at the
124: scale $l$ will cascade to smaller scales in a time $l/|{\bf u}_l|$, and
125: since the energy cascade rate $\epsilon$ is constant, we obtain
126: $\epsilon \sim |{\bf u}_l|^3/l$ that implies $|{\bf u}_l|\sim l^{1/3}$, which
127: finally leads to the well verified K41 spectrum $dE/dk \sim k^{-5/3}$
128: to within small intermittency corrections.
129:
130: The assumptions of the HD theory of turbulence have been tested in the
131: literature. Ref. \cite{Domaradzki88} first tested the assumption of
132: locality using direct numerical simulations (DNS) of $64^3$ grid points.
133: Their work has been followed by a number of authors with higher resolution
134: simulations
135: \cite{Domaradzki90,Ohkitani92,Zhou93L,Zhou93,Yeung95,Zhou96,Kishida99}.
136: Refs. \cite{Yeung91,Brasseur94,Yeung95} have also investigated the effect of
137: long-range interactions and anisotropy induced by an anisotropic large
138: scale flow. Although some issues still remain regarding the effect of
139: long range interactions, the locality of the energy transfer has been
140: confirmed.
141:
142: However, there are two important assumptions used in the HD
143: case that are not necessarily true for the MHD case.
144: First, the assumption of isotropy breaks down if an imposed uniform magnetic
145: field is considered. We will not investigate such effects in
146: the present work and will only consider flows with
147: $\int {\bf b} \, d{\bf x}^3=0$.
148: The second assumption, that of locality
149: of interactions among the different scales is what motivates our work. Unlike the HD case where
150: the effect of larger eddies on smaller ones is the advection of the later
151: ones (an effect that can be taken away by a Galilean transformation), in
152: MHD the effect of a large scale fluctuation of the magnetic field cannot be so
153: eliminated. Therefore, in MHD it is possible for small scales to
154: interact directly with the large scales. If this is the case, we can not consider
155: a ``contiguous" transfer of energy in wave number space and cannot {\it a priori} follow
156: the same arguments Kolmogorov used for HD turbulence.
157: Therefore knowledge of the energy transfer among different scales
158: is important for the construction of any phenomenological model of
159: turbulence.
160:
161: Present phenomenological models follow Kolmogorov like arguments
162: that take into account the effect of the magnetic field.
163: Iroshnikov \cite{Iroshnikov}
164: and Kraichnan \cite{Kraichnan} proposed the first models to describe
165: isotropic-MHD turbulence, predicting a spectrum of $k^{-3/2}$
166: (hereafter, IK).
167: Goldreich and Shridar \cite{Goldreich} proposed a new model
168: for anisotropic MHD turbulence that takes into account the anisotropy
169: introduced by a uniform magnetic field ${\bf B}_0$, predicting a spectrum of
170: $k_\perp^{-5/3}$, where $k_{\perp}$ refers to the direction perpendicular to ${\bf B}_0$.
171: Several models have been proposed that combine the
172: two spectra (see e.g. \cite{Matthaeus,Boldyrev,Galtier05}), suggesting that the
173: index of the energy spectrum is sensitive to the presence and intensity of ${\bf B}_0$.
174: %%
175: Some aspects of non-locality of interactions are taken into account
176: in the afore-mentioned models by considering that large scale fluctuations of
177: the magnetic field act as a uniform magnetic field to the smaller scales,
178: and as a result they speed up or slow down the rate at which the energy is
179: cascading. However, in these models, although non-local interactions
180: are taken into account, the energy is transfered
181: locally from one scale to a slightly smaller scale, like in Kolmogorov's
182: HD turbulence model.
183: %%
184:
185: The locality of the interactions and the energy transfer in MHD turbulence
186: has been investigated through various closure models. The energy transfer
187: has been studied within the EDQNM closure model by \cite{Pouquet}
188: and more recently by \cite{Schilling} where non-local interactions have been noted.
189: Using field theoretical calculations the transfer of energy has been estimated by
190: \cite{Verma01,Verma03,Verma05}.
191: As far as we know, the locality of the energy transfer in MHD
192: has been investigated through three dimensional direct numerical simulations (DNS) only very
193: recently \cite{Debliquy} (see also \cite{Dar01} for the two-dimensional case).
194: These authors measured the transfer of energy between different scales and fields
195: using free decaying MHD turbulence simulations with $512^3$ grid points.
196: Their results showed that there is local transfer of energy between
197: the same fields, while the transfers involving the two different fields
198: showed a less local behavior, in the sense that a wider range of scales
199: was involved in the interactions.
200:
201: In our work we use the results of DNS of mechanically forced MHD turbulence
202: (\add{unlike the free decaying case studied in \cite{Verma05}})
203: to study the locality of the energy transfer between different scales
204: and fields.
205: %
206: In all the cases studied we consider a mechanic external forcing that
207: generates a well defined large scale flow and small scale turbulent
208: fluctuations. This is a regime of interest for several astrophysical and geophysical flows
209: where
210: magnetic fields are believed to be sustained against Ohmic dissipation by a dynamo process
211: \cite{Moffatt}, and the only external source of energy driving the system
212: is mechanical (e.g. convection and rotation).
213: %
214: There is an important difference between the case studied in \cite{Debliquy}
215: and the case considered in our work. In our case energy is forced through
216: the velocity field and the system reaches a steady state with equipartition
217: between the two fields. For this to happen there must be a non-zero
218: flux for all times from the velocity field
219: to the magnetic field. This is not necessarily true for the case of decaying turbulence
220: and as our results show this significantly modifies the energy transfers from the velocity
221: field to the magnetic field.
222:
223: In Sec. \ref{Theory}
224: we introduce the definitions of the transfer terms for MHD, and in
225: Sec. \ref{Results} we present the code we use for the numerical
226: simulations as well as the results of the analysis. Finally, in
227: Sec. \ref{Concs} we summarize the main results of our work.
228:
229: \section{ \label{Theory} Theory and definitions}
230:
231: The equations that describe the dynamics of an
232: incompressible conducting fluid coupled to a magnetic field in the MHD approximation are given by:
233: %
234: \begin{equation}
235: \partial_t {\bf u} + {\bf u}\cdot \nabla {\bf u} = - \nabla p +
236: {\bf b}\cdot \nabla {\bf b} + \nu \nabla^2 {\bf u} +{\bf f} ,
237: \label{eq:momentum}
238: \end{equation}
239: %
240: \begin{equation}
241: \partial_t {\bf b} + {\bf u}\cdot \nabla {\bf b} = {\bf b}\cdot \nabla {\bf u}
242: + \eta \nabla^2 {\bf b} ,
243: \label{eq:Induction}
244: \end{equation}
245: %
246: \begin{equation}
247: \nabla \cdot {\bf u} =0, \,\,\, \nabla \cdot {\bf b} =0 ,
248: \label{eq:Incompresible}
249: \end{equation}
250: where ${\bf u}$ is the velocity field and ${\bf b}$ is the magnetic field.
251: $p$ is the (total) pressure and $\nu$ and $\eta$ are the viscosity and the
252: magnetic diffusivity respectively.
253: Here, ${\bf f}$ is the external force that drives the turbulence and the
254: dynamo.
255: The largest wavenumber of the Fourier transform of $f$ is going to be denoted
256: as ${\bf k}_F$ and we are going to refer to ${|\bf k}_F|^{-1}$ as the forced scale.
257: We are also going to define the viscous dissipation scale as
258: $k_\nu^{-1}=(\epsilon/\nu^3)^{-1/4}$ and resistive dissipation scale as
259: $k_\eta^{-1}=(\epsilon/\eta^3)^{-1/4}$ where $\epsilon$ is the energy dissipation rate.
260: A large separation between the two
261: scales (${|\bf k}_F|^{-1} \gg \max \{ \{k_\nu^{-1},k_\eta^{-1} \} $) is required
262: for the flow to reach a turbulent state.
263:
264: To investigate the transfer of energy among different
265: scales of turbulence we use the Fourier transforms
266: of the fields:
267: \[ {\bf {u(x)}} = \sum_{\bf k} {\bf \tilde{u}(k)}e^{i{\bf kx}}
268: \,\,\, {\mathrm{,} } \,\,\,
269: {\bf \tilde{u}(k)} = \frac{1}{({2\pi})^3}\int {\bf {u}(x)}e^{-{\bf ikx}}
270: d{\bf x}^3 \,\,\, \]
271: and
272: \[ {\bf {b(x)}} = \sum_{\bf k} {\bf \tilde{b}(k)}e^{i{\bf kx}}
273: \,\,\, {\mathrm{,} } \,\,\,
274: {\bf \tilde{b}(k)} = \frac{1}{({2\pi})^3}\int {\bf {b}(x)}e^{-i{\bf kx}}
275: d{\bf x}^3 \,\,\, , \]
276: where the domain is taken to be a triply periodic cube of size $L=2\pi$.
277: We can now introduce the shell filter decomposition:
278: \[ {\bf u(x)} = \sum_K {\bf u}_K(x) , \,\, {\bf b(x)} = \sum_K {\bf b}_K(x) \]
279: where
280: \[{\bf u}_K(x)= \sum_{K<|{\bf k}|\le K+1} {\bf \tilde{u}(k)}e^{i{\bf kx}} , \]
281: and similar for the field ${\bf b}$
282: \[{\bf b}_K(x)= \sum_{K<|{\bf k}|\le K+1} {\bf \tilde{b}(k)}e^{i{\bf kx}} . \]
283: The fields ${\bf u}_K$ and ${\bf b}_K$ are therefore defined as the part
284: of the velocity
285: and magnetic field respectively, whose Fourier transform contains only
286: wave numbers in the shell $(K, K+1]$ (hereafter called the shell $K$)
287: and represent ``eddies" of scale $K^{-1}$. The evolution of the kinetic energy in a shell $K$,
288: $E_u(K)=\int {\bf u}_K^2/2 \, dx^3$ is given by:
289: \begin{eqnarray}
290: \partial_t E_u(K) &=&
291: \int \sum_Q \left[- {\bf u}_K {\bf \cdot (u \cdot \nabla) \cdot u}_Q
292: + {\bf u}_K {\bf \cdot (b \cdot \nabla) \cdot b}_Q \right]
293: \nonumber \\
294: { } && - \nu |\nabla {\bf u}_K|^2
295: + {\bf f \cdot u}_K \,\, d{\bf x}^3 \, ,
296: \end{eqnarray}
297: and for the magnetic energy $E_b(K)=\int {\bf b}_K^2/2 \, dx^3$
298: we obtain
299: \begin{eqnarray}
300: \partial_t E_b(K) &=&
301: \int \sum_Q \left[- {\bf b}_K {\bf \cdot (u \cdot \nabla) \cdot b}_Q
302: + {\bf b}_K {\bf \cdot (b \cdot \nabla) \cdot u}_Q\right]
303: \nonumber \\
304: { } && - \eta |\nabla {\bf b_K}|^2 \,\, d{\bf x}^3 \, .
305: \end{eqnarray}
306: %
307: The above equations can be written in the more compact form:
308: \begin{equation}
309: \partial_t { E}_{u}(K) =
310: \sum_Q [{\mathcal T}_{uu}(Q,K)+{\mathcal T}_{bu}(Q,K)] - \nu {\mathcal D}_u(K)
311: + {\mathcal F}(K) ,
312: \label{eq:Eu}
313: \end{equation}
314: %
315: \begin{equation}
316: \partial_t { E}_{b}(K) =
317: \sum_Q [{\mathcal T}_{ub}(Q,K)+{\mathcal T}_{bb}(Q,K)] -
318: \eta {\mathcal D}_b(K).
319: \label{eq:Eb}
320: \end{equation}
321: %
322: Here we have introduced the functions ${\mathcal T}_{uu}(Q,K)$,
323: ${\mathcal T}_{ub}(Q,K)$, ${\mathcal T}_{bb}(Q,K)$, and
324: ${\mathcal T}_{bu}(Q,K)$
325: that express the energy transfer between different fields and shells.
326: %
327:
328: ${\mathcal T}_{uu}(Q,K)$ expresses the transfer rate of kinetic energy
329: lying in the shell $Q$ to kinetic energy lying in the shell $K$
330: through the velocity advection term and is defined as:
331: \begin{equation}
332: {\mathcal T}_{uu}(Q,K) \equiv
333: -\int {\bf {u}_K (u \cdot \nabla) {u}_Q } d{\bf x}^3 .
334: \label{eq:Tuu}
335: \end{equation}
336: We similarly define
337: \begin{equation}
338: {\mathcal T}_{bb}(Q,K) \equiv
339: -\int {\bf {b}_K (u \cdot \nabla) {b}_Q } d{\bf x}^3,
340: \label{eq:Tbb}
341: \end{equation}
342: which expresses the rate of energy transfer of magnetic energy lying in
343: the shell Q to magnetic energy lying in the shell $K$ through the
344: magnetic advection term.
345: %
346: The Lorentz force is responsible for the transfer
347: of energy from the magnetic field to the velocity field. The resulting
348: transfer rate is defined as:
349: %
350: \begin{equation}
351: {\mathcal T}_{bu}(Q,K) \equiv
352: \int {\bf {u}_K (b \cdot \nabla) {b}_Q } d{\bf x}^3.
353: \label{eq:Tub}
354: \end{equation}
355: Finally the term responsible for the stretching of the magnetic field lines
356: results in the transfer from kinetic energy to magnetic energy, given by:
357: \begin{equation}
358: {\mathcal T}_{ub}(Q,K) \equiv
359: \int {\bf {b}_K (b \cdot \nabla) {u}_Q } d{\bf x}^3 .
360: \label{eq:Tbu}
361: \end{equation}
362: In summary, the functions ${\mathcal T}_{vw}(Q,K)$ (for arbitrary fields
363: ${\bf v}$ and ${\bf w}$) represent the rate of transfer of energy
364: from the field ${\bf v}$ (first index) in the shell $Q$ (first argument),
365: into energy of the field ${\bf w}$ (second index) in the shell $K$ (second argument).
366: If ${\mathcal T}_{vw}(Q,K)>0$, then a positive amount of $v-$energy is transfered
367: from the shell $Q$ to $w-$energy in the shell $K$.
368: If ${\mathcal T}_{vw}(Q,K) < 0$, then
369: a negative amount of $v-$energy is transfered from the shell $Q$ to $w-$energy in the shell
370: $K$, or in other words, energy is transfered backwards from the shell $K$ to the shell $Q$.
371:
372: In eqs. (\ref{eq:Eu}-\ref{eq:Eb}) we have also introduced two
373: dissipation functions: the kinetic energy dissipation rate
374: \begin{equation}
375: \nu {\mathcal D}_{u}(K) \equiv \nu \int |{\bf \nabla u}_K|^2 d{\bf x}^3 ,
376: \end{equation}
377: and the magnetic energy dissipation rate
378: \begin{equation}
379: \eta {\mathcal D}_{b}(K) \equiv \eta \int |{\bf \nabla b}_K|^2 d{\bf x}^3 .
380: \end{equation}
381: Finally,
382: \begin{equation}
383: {\mathcal F}(K) \equiv \int { \bf f} \cdot {\bf u}_K \, d{\bf x}^3
384: \end{equation}
385: is the energy injection rate to the velocity field through the forcing term.
386: %%
387: %%
388: %%
389: %%
390:
391: Before presenting the results from numerical simulations, let us discuss
392: some of the properties of the transfer functions. If ${\mathcal T}_{vw}(Q,K)$
393: (where $v,w$ can be either $u$ or $b$) is expressing the rate of energy
394: transfer from the field ${\bf v}$ in the shell $Q$ to the field ${\bf w}$
395: in the shell $K$, then the following identity should hold
396: \begin{equation}
397: \label{tran_id}
398: {\mathcal T}_{vw}(Q,K)=-{\mathcal T}_{wv}(K,Q) .
399: \end{equation}
400: The interpretation of eq.(\ref{tran_id}) is
401: that the rate at which the shell $Q$ is giving energy to the shell
402: $K$ must be equal to the rate the shell $K$ is receiving energy from the
403: shell $Q$. Eq. (\ref{tran_id}) can be easily shown to hold for all the
404: transfer functions we defined (eqs. [\ref{eq:Tuu}-\ref{eq:Tbu}]). It is
405: this property that allows us to interpret the functions ${\mathcal T}_{uu}$,
406: ${\mathcal T}_{bu}$, ${\mathcal T}_{ub}$, and ${\mathcal T}_{bb}$ as the
407: energy transfer between different scales and fields.
408:
409: For a turbulent flow in a statistically steady state,
410: equations (\ref{eq:Eu}) and (\ref{eq:Eb}) imply that:
411: \begin{equation}
412: \sum_Q\langle{\mathcal T}_{uu}(Q,K)+{\mathcal T}_{bu}(Q,K) \rangle =
413: \langle {\mathcal D}_u(K) \rangle - \langle {\mathcal F}(K)\rangle ,
414: \end{equation}
415: and
416: \begin{equation}
417: \sum_Q\langle{\mathcal T}_{ub}(Q,K)+{\mathcal T}_{bb}(Q,K) \rangle=
418: \langle {\mathcal D}_b(K) \rangle ,
419: \end{equation}
420: where $\langle \cdot \rangle$ stands for a time average or an ensemble
421: average. For fixed $K$ outside the forcing band, and in the limit of
422: $\nu,\eta \to 0$, we have that
423: \begin{equation}
424: \sum_Q\langle{\mathcal T}_{uu}(Q,K)+{\mathcal T}_{bu}(Q,K) \rangle = 0
425: \end{equation}
426: and
427: \begin{equation}
428: \sum_Q\langle{\mathcal T}_{ub}(Q,K)+{\mathcal T}_{bb}(Q,K) \rangle= 0 .
429: \end{equation}
430: However, limited resolution will allow us to be in the regime
431: where these last two equations hold only for a small range of wavenumbers.
432:
433: Finally we need to comment on the definitions of the various transfer
434: functions we are using in this paper and the connection to the triad of
435: wave numbers (${\bf k,p,q}$) that satisfy the relation ${\bf k+p+q=0}$
436: (because of the convolution term resulting from the quadratic nonlinearities
437: of the primitive equations); such triad is the basis for mode to mode interactions
438: (see e.g. \cite{Kraichnan2}).
439: Our approach is equivalent to
440: considering all triad interactions with the one wavenumber ${\bf k}\in K$
441: and ${\bf q} \in Q$ and summing over all ${\bf p}$ satisfying
442: ${\bf k+p+q=0}$ in all shells, where ${\bf p}$ is the wave-number of
443: the advecting field, and ${\bf k}$ and ${\bf q}$
444: are the wavenumbers of the modes energy is transfered to and from.
445: Although the approach we are using gives us information
446: on whether the energy is transfered locally or not,
447: it cannot give definite conclusions on whether the interactions themselves
448: are local. For example, even if energy is transfered locally from a wavenumber
449: ${\bf k}$ to a wavenumber ${\bf q\sim k}$,
450: the wavenumber ${\bf p}$ that is responsible for the transfer
451: is not necessarily of the same order of magnitude as $|{\bf k}|$ and $|{\bf q}|$.
452: Ideally, one would investigate transfer terms of the form:
453: ${\bf T}_{uu}(K|P|Q)\equiv \int {\bf u}_K({\bf u}_P \cdot \nabla) {\bf u}_Q d{\bf x}^3$
454: that contain information about the third wave number involved in the
455: interactions taking place. However the difficulty of manipulating data
456: from high resolution runs and the difficulty of interpreting the results
457: of transfer functions that depend on three arguments restricts us for the present time, to
458: examine just the locality of the energy transfer.
459:
460: \section{ \label{Results} Results}
461:
462: To study the transfer of energy in MHD turbulence we use the
463: turbulent steady state of several mechanically forced three dimensional
464: MHD direct numerical simulations. The simulations and details of the
465: code can be found in \cite{MininniApJ,MininniPRE}. The runs were
466: performed in a triply periodic domain with a resolution of $256^3$ grid
467: points, using a pseudo-spectral scheme with the $2/3$-rule for dealiasing.
468: The equations were evolved in time using a second order Runge-Kutta method.
469:
470: Turbulence was generated by two different types of forcing. In the first
471: case a non-helical Taylor-Green force (hereafter referred as TG) was used
472: ${\bf f}_{\rm TG}(k_0)=(\, \sin(k_0 x) \cos(k_0 y) \cos(k_0 z),
473: -\cos(k_0 x) \sin(k_0 y) \cos(k_0 z),
474: 0)$
475: with $k_0=2$
476: \cite{MininniApJ}. In the second case a helical ABC force was used
477: ${\bf f}_{ABC}(k_0) = (\,B\cos(k_0 y)+C\sin(k_0 z),\,
478: C\cos(k_0 z)+A\sin(k_0 x),\,
479: A\cos(k_0 x)+B\sin(k_0 y)\,)$
480: with $k_0=2$
481: \cite{MininniPRE}. All simulations were done with constant in time external
482: force. First a hydrodynamic simulation was carried using each force, to reach
483: a turbulent steady state. Both external forces generate a well defined
484: large scale flow at $|K_F|\sim3$, and small scale turbulent fluctuations
485: following to a good approximation a 5/3 Kolmogorov law.
486: Then MHD simulations were carried, and a
487: small magnetic field was amplified and sustained to equipartition by a
488: dynamo process. The results in this paper are based on the saturated
489: stage of the dynamo, which we will refer in the following as the MHD
490: turbulent steady state.
491:
492: The transfers were calculated based on the definitions
493: (\ref{eq:Tuu}--\ref{eq:Tbu}).
494: The transfer of energy during the early
495: stages of the MHD simulations, when the magnetic energy is small and the
496: velocity field is not modified by the Lorentz force (often referred to as
497: the kinematic dynamo regime) are examined in a companion paper \cite{us}
498: (hereafter referred as Paper II). Table \ref{runs} gives several relevant parameters for
499: each run, and figure \ref{fig_01} shows the resulting energy spectra.
500:
501: Both simulations display a large scale magnetic field, although the
502: spectrum of magnetic energy in the ABC simulation shows a stronger peak
503: at $k=1$. This peak is related with the dynamo $\alpha$-effect
504: and the inverse cascade of magnetic helicity.
505: Details of this process will be discussed
506: in Paper II. However, it is important to note that in the ABC simulation
507: the large scale magnetic field is strongly helical, while in the TG
508: simulation the magnetic helicity is negligible. This large scale magnetic
509: field is self-sustained by the turbulence. In both simulations, the
510: net cross helicity (correlation between the velocity and the magnetic field)
511: is small and can be neglected.
512:
513: %
514: %
515: \begin{table}
516: \caption{\label{runs}Simulations. $L$ is the integral length-scale of the
517: flow, defined as $L=\int{E_u(k)\, dk}/\int{E_u(k) k^{-1} \, dk}$; $\nu$
518: is the kinematic viscosity, and $\eta$ the magnetic diffusivity.
519: The kinetic and magnetic Reynolds numbers $Re$ and R$_M$ are based on $L$
520: and the rms velocity, while the ratio of magnetic to kinetic energy
521: $E_b /E_u$ is the average in the turbulent steady state.}
522: \begin{ruledtabular}
523: \begin{tabular}{lcccccc}
524: Forcing & $L$ & $\nu$ & $\eta$ & Re & R$_M$&$E_b/E_u$\\
525: \hline
526: ABC &$1.64$&$2\times10^{-3}$ & $2\times10^{-3}$ & 820 & 820 & $0.84$ \\
527: TG &$1.35$&$2\times10^{-3}$ & $5\times10^{-3}$ & 675 & 270 & $0.72$ \\
528: \end{tabular}
529: \end{ruledtabular}
530: \end{table}
531: %%%%%%%%%%%%%%%%%%%%%%
532: %%%%%%%%FIGURE 1%%%%%%
533: \begin{figure}
534: \includegraphics[width=8cm]{fig_01}
535: \caption{\label{fig_01}Spectra of kinetic energy (solid line)
536: and magnetic energy (dashed line) of the ABC and Taylor Green runs,
537: where the Taylor Green spectra have been shifted down by a factor of 20
538: for clarity. The Kolmogorov slope is showed as a reference.
539: Note that the magnetic Prandtl number $P_M \equiv \nu/\eta$ differs for the two runs.}
540: \end{figure}
541: %%%%%%%%%%%%%%%%%%%%%%
542:
543: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
544: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
545: \subsection{ \label{Hydro} Hydrodynamic Turbulence }
546:
547: Locality of interactions in hydrodynamic turbulence have been investigated
548: before in the literature
549: \cite{Domaradzki90,Ohkitani92,Zhou93L,Zhou93,Yeung95,Zhou96,Kishida99}.
550: Although some open issues still remain
551: %about the persistence of some long range interactions
552: \cite{Yeung91,Brasseur94,Yeung95}
553: it has been shown that energy is transfered mostly locally.
554: Here, for reasons of comparison we show the transfer
555: ${\mathcal T}_{uu}(Q,K)$ from hydrodynamical simulations using the same
556: external forces and parameters used in the MHD simulations. The results
557: are in good agreement with previous works.
558:
559: In figure \ref{fig_02} we show the energy transfer for a few modes for
560: the TG flow and in figure \ref{fig_03} the energy transfer
561: for the ABC flow.
562: %
563: \begin{figure}
564: \includegraphics[width=8cm]{fig_02}
565: \caption{\label{fig_02} The transfer of energy ${\mathcal T}_{uu}(Q,K)$
566: for the Taylor-Green run. The figure shows the rate that energy is
567: transfered from the modes $Q=3,10,20,30$ to all the other modes $K$.}
568: \end{figure}
569: \begin{figure}
570: \includegraphics[width=8cm]{fig_03}
571: \caption{\label{fig_03} The transfer of energy ${\mathcal T}_{uu}(Q,K)$
572: for the ABC run. The figure shows the rate that energy is
573: transfered from the modes $Q=3,10,20,30$ to all the other modes $K$.}
574: \end{figure}
575: %
576: In both cases the transfer of energy is direct and local:
577: all the curves (with the exception of the forced mode $Q=3$)
578: %
579: are negative for $K$ smaller than $Q$
580: %
581: and positive for $K$ larger than $Q$.
582: As a result, all inertial range modes
583: receive energy from modes with slightly smaller wavenumbers
584: (negative ${\mathcal T}_{uu}$)
585: and give energy to modes with slightly
586: larger wave number (positive ${\mathcal T}_{uu}$).
587: %
588: The locality of
589: the transfer is expressed from the fact that the transfer of
590: energy from the modes in the shell $Q$ to modes in shells $K$ with
591: $K \ll Q$ or $K \gg Q$ is very small, and decreases fast with the
592: separation of the two wave numbers.
593: Finally, as the shell wavenumber $K$ and $Q$ is increased, there is
594: a drop in the amplitude of the transfer. If the transfer functions were
595: self-similar then an increase of the wave numbers $K$ and $Q$ to
596: $\lambda K$ and $\lambda Q$ would imply ${\mathcal T}_{uu}(\lambda Q,\lambda K)=
597: \lambda^{-2}{\mathcal T}_{uu}(Q , K)$ \cite{Kraichnan2}.
598: This scaling could explain this drop of amplitude. However
599: the inertial range in our DNS is too small to test self-similarity and
600: a large part of the drop is due to the presence of viscosity.
601:
602: The forced mode has a slightly different behavior. The transfer
603: rate from the forced wave number to its nearby shells has a
604: considerably larger amplitude. Also, for both flows there
605: is some backscattering from the forced wave number to shells with
606: smaller wavenumber. This is clearer in the helical (ABC) flow.
607:
608: \subsection{ \label{MHD} Magneto-Hydrodynamic Turbulence }
609:
610: We are now ready to examine results from the energy transfer for MHD
611: turbulence. First we examine the transfer of kinetic energy from large
612: scales to kinetic energy in small scales through the term ${\mathcal T}_{uu}(Q,K)$, and
613: magnetic energy from large scales to magnetic energy in small scales
614: through the term ${\mathcal T}_{bb}(Q,K)$. These two transfer functions bare
615: some significant similarities with the hydrodynamic case.
616:
617: In figures \ref{fig_04} and \ref{fig_05} we show ${\mathcal T}_{uu}$
618: (top panel) and ${\mathcal T}_{bb}$ (bottom panel) for the non-helical
619: TG flow and the helical ABC flow.
620: %%%% FIGURE 4 %%%%%
621: \begin{figure}
622: \includegraphics[width=8cm]{fig_04}
623: \caption{\label{fig_04} Top panel:The transfer of energy
624: ${\mathcal T}_{uu}(Q,K)$ for the Taylor-Green run.The figure shows
625: the rate that kinetic energy is transfered from the modes $Q=3,10,20,30$
626: to kinetic energy to all the other modes $K$.
627: Bottom panel: The transfer of energy ${\mathcal T}_{bb}(Q,K)$ for the same flow. The
628: figure shows the rate that magnetic energy is transfered from the
629: modes $Q=3,10,20,30$ to magnetic energy to all the other modes $K$.}
630: \end{figure}
631: %%%%%%%%%%%%%%%%%%%
632: %%%% FIGURE 5 %%%%%
633: \begin{figure}
634: \includegraphics[width=8cm]{fig_05}
635: \caption{\label{fig_05} Same as figure 4 for the ABC run}
636: %\caption{\label{fig_05} Top panel:The transfer of energy
637: %${\mathcal T}_{uu}(Q,K)$ for the ABC run.The figure shows the
638: %rate that kinetic energy is transfered from the modes $Q=3,10,20,30$
639: %to kinetic energy to all the other modes $K$.
640: %Bottom panel: The transfer of energy ${\mathcal T}_{bb}(Q,K)$. The
641: %figure shows the rate that magnetic energy is transfered from the modes
642: %$Q=3,10,20,30$ to magnetic energy to all the other modes $K$}
643: \end{figure}
644: %%%%%%%%%%%%%%%%%%%
645: The velocity to velocity transfer has not changed drastically
646: (other than a decrease in amplitude) from the pure hydrodynamic case.
647: As in Sec. \ref{Hydro}, the transfer implies a local direct cascade.
648: All the curves are negative for $K$ smaller than $Q$,
649: and positive for $K$ larger than $Q$.
650: Each mode is therefore receiving energy
651: from the larger scales (negative transfer) and giving energy to the
652: smaller scales (positive transfer). The decrease in amplitude
653: (when compared with the
654: hydrodynamic case) is partly because the magnitude of the velocity
655: field is decreased when magnetic field comes to equipartition,
656: % with the velocity field,
657: and partly because now there is a net transfer
658: of energy from the velocity field to the magnetic field, making the
659: available energy to cascade to small velocity scales smaller.
660: %Note also a small backscattering from the small scales to
661: %$K$-shells larger than the forcing band, not present in the non-helical
662: %case. \note{I dont understand}
663:
664:
665: The transfer of magnetic energy to magnetic energy ${\mathcal T}_{bb}(Q,K)$
666: seems to follow the same behavior as the velocity field transfer. The
667: results show a direct cascade with local transfer of energy from large
668: scales to small scales. We note that for the helical case the transfer
669: of magnetic energy is larger than the transfer of kinetic energy.
670: The likely reason for this behavior is that in the ABC flow the magnetic energy
671: at large scales and intermediate scales saturates at higher values than
672: in the TG flow, due to the presence of helicity or the dynamo
673: $\alpha$-effect. This process will be discussed in more detail
674: in Paper II.
675:
676: Next we investigate the transfer of energy from one field to the other,
677: by examining the terms ${\mathcal T}_{ub}$ and ${\mathcal T}_{bu}$.
678: Because of the anti-symmetric property
679: ${\mathcal T}_{ub}(Q,K)=-{\mathcal T}_{bu}(K,Q)$, it is sufficient to
680: just study the transfer of energy from the velocity field to the magnetic
681: field. However, we need to remark that
682: unlike the ${\mathcal T}_{uu}(Q,K),{\mathcal T}_{bb}(Q,K)$
683: terms that their dependence on $K$ and $Q$ is the same up to a minus sign,
684: the behavior of
685: ${\mathcal T}_{ub}(Q,K)$ as we vary $K$ is not the same as if we vary
686: $Q$. Therefore the two behaviors need to be studied separately (i.e.,
687: the transfer of energy from a velocity mode to two different magnetic
688: modes is different from the transfer of energy from
689: two different velocity modes to a magnetic mode).
690: In figure \ref{fig_06} (TG), and \ref{fig_07} (ABC), we show the
691: transfer of kinetic energy from the velocity modes $Q=3,4,5,15,17$, and 20
692: to all the examined magnetic modes $K$.
693: %
694: %%%% FIGURE 6 %%%%%
695: \begin{figure}
696: \includegraphics[width=8cm]{fig_06}
697: \caption{\label{fig_06} The transfer of kinetic energy to magnetic energy
698: ${\mathcal T}_{uu}(Q,K)$
699: for the Taylor Green run.The figure shows the rate that kinetic energy is
700: transfered from the modes $Q=3,4,5$ (inset modes $Q=15,17,20$)
701: to magnetic energy in the modes $K$. }
702: \end{figure}
703: %%%%%%%%%%%%%%%%%%%
704: %%%% FIGURE 7 %%%%%
705: \begin{figure}
706: \includegraphics[width=8cm]{fig_07}
707: \caption{\label{fig_07} The transfer of kinetic energy to magnetic energy
708: ${\mathcal T}_{ub}(Q,K)$ for the ABC run.
709: The figure shows the rate that kinetic energy is
710: transfered from the modes $Q=3,4,5$ (inset modes $Q=15,17,20$)
711: to magnetic energy in the modes $K$.}
712: \end{figure}
713: %%%%%%%%%%%%%%%%%%%
714:
715: A few things should be noted. First, in both runs (ABC and TG) the
716: modes associated with the large scale flow ($Q=3$) seem to play a
717: dominant role in the transfer of energy from the velocity field to
718: the magnetic field. Note also that there is a wider range of magnetic
719: field modes into which the forced velocity field modes input energy.
720:
721: This is more apparent for the helical flow, that seems better at
722: stretching and folding the magnetic field. The cascade in the modes
723: inside the inertial range is direct in both cases but with a small difference.
724: In both cases the large scale velocity field is
725: transferring energy to smaller scale magnetic field and receiving energy
726: from larger scale magnetic field. However, for the Taylor-Green case
727: there is very small transfer from one field to the other in the same shell.
728: On the other hand, in the ABC flow the peak of the transfer from the magnetic
729: field to the velocity
730: field (the negative peaks in figure \ref{fig_07}) is for the same shell.
731: Note also that for the $K$-shells
732: larger than $Q$, the transfer for all $Q$ follows the same curve.
733: This implies that all the small scale velocity modes give energy
734: to the magnetic field modes at the same rate.
735: This is clearer when we examine the dependence with $Q$.
736:
737: %%%% FIGURE 8 %%%%%
738: \begin{figure}
739: \includegraphics[width=8cm]{fig_08}
740: \caption{\label{fig_08}The transfer of kinetic energy to magnetic energy
741: ${\mathcal T}_{ub}(Q,K,)$ for the Taylor Green run.
742: The figure shows the rate that kinetic energy is
743: transfered from the modes $Q$ (x-axis)
744: to magnetic energy in the modes $Q=10$ (top panel), $Q=20$ (middle panel),
745: $Q=30$ (bottom panel).}
746: \end{figure}
747: %%%%%%%%%%%%%%%%%%%
748: %%%% FIGURE 9 %%%%%
749: \begin{figure}
750: \includegraphics[width=8cm]{fig_09}
751: \caption{\label{fig_09} The transfer of kinetic energy to magnetic energy
752: ${\mathcal T}_{ub}(Q,K,)$ for the ABC run.
753: The figure shows the rate that kinetic energy is
754: transfered from the modes $Q$ (x-axis)
755: to magnetic energy in the modes $Q=10$ (top panel), $Q=20$ (middle panel),
756: $Q=30$ (bottom panel).}
757: \end{figure}
758: %%%%%%%%%%%%%%%%%%%
759: %
760: In figures \ref{fig_08} and \ref{fig_09} we show the same transfer function
761: ${\mathcal T}_{ub}(Q,K)$ for three values of $K=10,20,30$.
762: The energy cascade is also direct (energy going from large scales to small scales),
763: however it is clear
764: from these figures that the transfer from the velocity field to the
765: magnetic field is a highly non-local process. Each magnetic field mode $Q$
766: is receiving energy (positive ${\mathcal T}_{ub}$) from all the velocity
767: modes with wave number $K$ smaller than $Q$, with the same rate!
768: The only exception is the mechanically sustained large scale velocity field that gives
769: even more energy (observe the peak at $k=3$). In fact,
770: most of the energy that is transfered from the velocity field
771: to the magnetic field
772: originates from the velocity field modes at $Q=3$ (around $60\%$ for the
773: TG run and $75\%$ for the ABC run.)
774: This energy turns into magnetic energy at several wavenumbers $K$
775: which locally cascades to smaller scales through the ${\mathcal T}_{bb}$
776: term. This bigger contribution of the large scale flow to
777: ${\mathcal T}_{ub}$ (compared with the contribution of the turbulent
778: components) is in good agreement with the suppression of small scale
779: velocity fluctuations by the large scale magnetic field, as observed
780: in \cite{MininniApJ}.
781: However, we need to note that as the scale of the magnetic field becomes smaller
782: there is more energy input from the turbulent components of the
783: velocity field than from the large scale (forced) flow. This
784: just follows from the fact that for $K$ large enough, the area below
785: the curve with constant ${\mathcal T}_{ub}$ is larger than the peak
786: at $Q=3$. It is possible therefore that in the limit of large
787: inertial range the effect of the forced velocity scales in the
788: small magnetic scales will not be as strong.
789: Finally we note that this mechanism described above
790: is different in a kinematic dynamo regime, as
791: is shown in Paper II.
792:
793: In summary, the existence of the long plateau with constant
794: ${\mathcal T}_{ub}(Q,K)$ at each fixed value of $K$, and the fact
795: that all the magnetic wavenumbers $K$ receive energy from the large
796: scale flow at $Q=3$ points that interactions between the velocity
797: field and the magnetic field are non-local in Fourier space.
798:
799: This non-local behavior of energy transfer from the velocity
800: field to the magnetic field seems to be absent from the decaying
801: MHD turbulence case studied by \cite{Debliquy}. In that case although
802: the ${\mathcal T}_{ub}$ and ${\mathcal T}_{bu}$ were more non-local
803: than the ${\mathcal T}_{bb}$ and ${\mathcal T}_{uu}$ terms
804: (since energy was transfered from the former ones in a wider range of shells
805: than the later ones), eventually at large separation of wave numbers
806: the transfer goes to zero. This is very different from the
807: plateau behavior we observe in the forced turbulence runs.
808: We suspect that this difference is due to the fact
809: that in the mechanically forced turbulence there is a net flux of
810: energy from the velocity field to the magnetic field that is responsible
811: for the formation of the plateau which does not exist in the decaying
812: turbulence case.
813:
814:
815: \subsection{A comparison between the transfers}
816:
817: In the previous section we showed that the transfer of energy from
818: the velocity field to velocity field and from the magnetic field to
819: magnetic field exhibit a local behavior similar to the transfer
820: in hydrodynamic turbulence, and the transfer from one field to the other
821: is exhibiting a non-local behavior.
822: In order to draw conclusions we need to compare the magnitude of these transfers.
823: Figures \ref{fig_10} and \ref{fig_11} show a comparison of the transfers
824: ${\mathcal T}_{uu}(Q,K),{\mathcal T}_{bb}(Q,K),{\mathcal T}_{ub}(Q,K)$ and
825: ${\mathcal T}_{bu}(Q,K)$ with $Q=15$ for the TG and ABC runs respectively.
826: %%%% FIGURE 10 %%%%
827: \begin{figure}
828: \includegraphics[width=8cm]{fig_10}
829: \caption{\label{fig_10} A comparison of the transfers
830: ${\mathcal T}_{uu}(Q,K),{\mathcal T}_{bb},(Q,K){\mathcal T}_{ub}(Q,K)$ and ${\mathcal T}_{bu}(Q,K)$
831: for $Q=15$ for the Taylor Green flow.}
832: \end{figure}
833: %%%%%%%%%%%%%%%%%%%
834: %%%% FIGURE 11 %%%%
835: \begin{figure}
836: \includegraphics[width=8cm]{fig_11}
837: \caption{\label{fig_11} A comparison of the transfers
838: ${\mathcal T}_{uu}(Q,K),{\mathcal T}_{bb},(Q,K),{\mathcal T}_{ub}(Q,K)$ and ${\mathcal T}_{bu}(Q,K)$
839: for $Q=15$ for the ABC flow.}
840: \end{figure}
841: %%%%%%%%%%%%%%%%%%%
842: The local transfers $u$ to $u$ and $b$ to $b$ appear to be of larger
843: magnitude than the non-local transfers $u$ to $b$ and $b$ to $u$.
844: In the case of the ABC flow the magnitude of the $b$ to $b$ transfer
845: seems to be twice the magnitude of the $u$ to $u$ transfer. This is
846: due to the fact that the magnetic energy in this run is larger than
847: in the TG run at large and intermediate scales.
848:
849: Figure \ref{fig_12} illustrates the transfer functions
850: ${\mathcal T}_{uu}(Q,K),{\mathcal T}_{bb},(Q,K)$ and ${\mathcal T}_{ub},(Q,K)$
851: as in figure \ref{fig_10} (TG flow), but we focus here on the large $K$
852: tail of the transfer and we consider $Q$=10.
853: %The fastest drop is for the transfer
854: %${\mathcal T}_{bu}(Q,K)$ which is the one exhibiting constant
855: %transfer with respect to $K$ for $K<Q$.
856: The fastest drop is for the transfer
857: ${\mathcal T}_{uu}(Q,K)$ making it the most 'local' one, next come the
858: ${\mathcal T}_{bb}(Q,K)$ transfer, and finally ${\mathcal T}_{ub}(Q,K)$
859: has the slowest drop. The same result was obtained for the ABC flow
860: (not shown here).
861:
862:
863: \begin{figure}
864: \includegraphics[width=8cm]{fig_12}
865: \caption{\label{fig_12} A comparison of the large $K$ tails
866: of transfers ${\mathcal T}_{uu}(Q,K),{\mathcal T}_{bb}(Q,K)$ and
867: ${\mathcal T}_{ub}(Q,K)$ in a log-linear plot
868: for $Q=10$ for the Taylor Green flow flow.}
869: \end{figure}
870:
871:
872: Figures \ref{fig_10}, \ref{fig_11}, and \ref{fig_12}
873: (as well as a comparison of the nonlocal transfers shown in figures
874: \ref{fig_08} and \ref{fig_09} with the local transfers in figures
875: \ref{fig_04} and \ref{fig_05} respectively) show that local interactions
876: between the same fields are much stronger than nonlocal interactions
877: between different fields.
878: However, nonlocal interactions spread over several shells,
879: and the magnetic field at a given scale $K$ can receive (give) energy
880: from (to) several velocity field $Q$ wavenumbers (instead of mostly
881: the nearest neighbors as is the case for local interactions).
882: Figure \ref{fig_13} shows the ratio
883: \begin{equation}
884: \frac{NL}{L}(K) = \sum_{Q=1}^{K}{{\mathcal T}_{ub}(Q,K)} \bigg/
885: \sum_{Q=1}^{K}{{\mathcal T}_{bb}(Q,K)} .
886: \end{equation}
887: This is the ratio of the total energy that the magnetic field at the shell $K$ receives
888: from the velocity field
889: through non-local transfer, to the total magnetic
890: energy received at the same scale through the local direct cascade of
891: (magnetic) energy.
892: Although in individual shells the local interactions are one
893: order of magnitude larger than the non-local transfer, the net amount
894: of energy received at a given scale $K$ by the two processes is
895: comparable (this ratio is different in a kinematic dynamo regime, as
896: will be shown in Paper II). At small scales, the ratio seems to
897: settle to a value close to 0.2, indicating that 20\% of the energy
898: received by these scales is through the non-local transfer
899: ${\mathcal T}_{ub}$.
900: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
901: \begin{figure}
902: \includegraphics[width=8cm]{fig_13}
903: \caption{\label{fig_13} The ratio $NL/L$
904: of energy received through the non-local transfer ${\mathcal T}_{ub}$
905: to local ${\mathcal T}_{bb}$, for the ABC and TG simulations.
906: The small scales receive 20\% of their energy through the non-local
907: transfer ${\mathcal T}_{ub}$.}
908: \end{figure}
909:
910: \section{ \label{Concs} Conclusions }
911:
912: In this paper we examined the transfer of energy in forced MHD turbulence
913: between the different scales and fields involved using the results from
914: numerical simulations in a turbulent steady state sustained by a
915: mechanical external force.
916: No qualitative differences in the transfer of kinetic to kinetic
917: (or magnetic to magnetic) energy has been observed, when compared against
918: the transfer of energy in a hydrodynamic simulation.
919: These transfers were found to be
920: always local and direct. However, all kinetic
921: energy modes have been observed to give energy to magnetic modes non-locally,
922: in the sense that a small scale magnetic field receives
923: the same amount of energy from all larger scales of the velocity
924: field in the inertial range.
925: Also each magnetic mode was found to receive a significant
926: amount of energy from the large scale flow at $|k_F| \sim 3$
927: (the scale of the forcing), an effect that seems to become smaller as we move
928: to smaller scales in the inertial range. We note that it is the
929: non-local interactions that actually sustain the magnetic
930: field against Ohmic dissipation.
931: A summary of our results is sketched in figure \ref{fig_14}.
932:
933: We have already noted that a different behavior for the
934: non-local transfers ${\mathcal T}_{ub}$ and ${\mathcal T}_{bu}$
935: was obtained for the mechanically forced turbulence investigated
936: in this work, when compared with the decaying turbulence case
937: studied by \cite{Debliquy}. Compared with incompressible hydrodynamic
938: turbulence, involving only one field and one transfer function, MHD
939: turbulence is richer and more complex. It involves two interacting fields,
940: several transfer functions, and as a result the energy injected at large
941: scales can travel to small scales through several channels. Also the number
942: of quadratic ideal invariants is larger, and inverse cascades
943: (not present in three dimensional hydrodynamics) can take place.
944: This suggests that in MHD flows the particular way the system is
945: set-up (e.g. mechanically or magnetically forced, free decaying cases
946: without external forces), or even the scale at which the energy is
947: injected (compared with the length of the box), might have a direct
948: effect in the evolution of the flow and lead to different transfers.
949:
950: \begin{figure}
951: \includegraphics[width=8cm]{fig_14}
952: \caption{\label{fig_14} A sketch of the energy transfer between different
953: scales and different fields. The thickness of the lines is an indication of the
954: magnitude of the transfers.
955: The figure illustrates how energy is transfered
956: to magnetic modes with wavenumber $k=q$ in the inertial range.
957: The transfers between same fields is always local and direct.
958: Each magnetic mode receives energy from all larger in scale velocity modes
959: and gives to slightly smaller in scale velocity modes. }
960: \end{figure}
961: %
962:
963: We would also like to comment on the implications of our results to
964: the different models of magneto-hydrodynamic turbulence. In the
965: present phenomenological models of MHD-turbulence
966: \cite{Iroshnikov,Kraichnan,Goldreich,Matthaeus,Boldyrev},
967: locality of the energy transfer is assumed.
968: That is to say, these models are derived assuming that
969: scales of different magnitude do not strongly interact. While this
970: assumption seems to be valid for HD-turbulence, this is not necessarily
971: true for MHD. As we have shown non-local interactions are present in MHD
972: turbulence and control the $u$ to $b$ transfers of energy.
973: %%%% FIGURE 16%%%%%%%%%
974: \begin{figure}
975: \includegraphics[width=8cm]{fig_15}
976: \caption{\label{fig_15} A comparison of the Els\"asser energy transfer
977: ${\mathcal T}_{z^+z^+}(Q,K)$
978: with the transfers from the local
979: ${\mathcal T}_{uu}+{\mathcal T}_{bb}$
980: and the non-local
981: ${\mathcal T}_{ub}+{\mathcal T}_{bu}$
982: contributions.
983: The inset is showing a blow up of the small $K$ tail where
984: the non-local interactions are more dominant.}
985: \end{figure}
986: %%%%%%%%%%%%%%%%%%%%%%
987: However, these non-local interactions are smaller in amplitude
988: and most of the input of energy to the magnetic field comes from
989: the large scale flow and then cascades to smaller scales making the
990: assumption of locality justified to some extend.
991: However non-local $u-to-b$ need to be considered to have a proper description
992: of the energy cascade.
993: To illustrate this we show in figure \ref{fig_15} the energy transfer
994: in terms of the Els\"asser variables ${\bf z}^\pm={\bf u}\pm {\bf b}$, often used in
995: turbulence models and we compare it with the contributions from
996: $u$ to $u$, $b$ to $b$, $b$ to $u$, and $u$ to $b$. In the figure we plot
997: ${\mathcal T}_{z^+z^+}\equiv -\int {\bf z^+}_K{\bf z^-\nabla z^+}_Q d{\bf x}$
998: and compare it with the energy transfer due to the local transfer terms
999: ${\mathcal T}_{uu}+{\mathcal T}_{bb}$ and the non-local transfer terms
1000: ${\mathcal T}_{ub}+{\mathcal T}_{bu}$. The local transfer terms appear
1001: to be dominant, except in the tails where the transfer of Els\"asser
1002: variables is dominated by the non-local transfers between the magnetic
1003: and kinetic energies.
1004: This tails, although with small amplitude, cannot
1005: be completely neglected, as shown by the $NL/L$ ratio of figure 13. The non-local tail
1006: in the transfer gives a net contribution of energy at magnetic small
1007: scales of roughly 1/5 when compared with the local transfer.
1008:
1009: Finally, we would like to say that our results were based on numerical
1010: simulations of moderate Reynolds number much smaller than what is
1011: observed in most physical phenomena. We already noted that due to the small
1012: inertial range we cannot test self similarity that would require to
1013: compare the transfers (i.e. ${\mathcal T}_{uu}(Q,K_1)$,
1014: ${\mathcal T}_{uu}(Q,K_2)$ )
1015: to wave numbers that are both significantly away from each other
1016: ($K_1 \ll K_2$)
1017: and away from the forced and dissipative scales ($K_F \ll K_1$ and
1018: $K_2 \ll k_{\eta}$).
1019:
1020: Finally the transfer of
1021: magnetic helicity and cross-helicity and their effect on the
1022: turbulence dynamics is also worth studying, but we leave these issues
1023: however for our future work.
1024:
1025:
1026: %However, we expect that the qualitative features of our results should
1027: %hold in higher Reynolds numbers $Re,R_M$.
1028:
1029:
1030: \begin{acknowledgments}
1031: The authors are grateful to J. Herring for valuable discussions
1032: and his careful reading of this document.
1033: Computer time was provided by NCAR. The NSF grant CMG-0327888
1034: at NCAR supported this work in part and is gratefully acknowledged.
1035: \end{acknowledgments}
1036:
1037: \bibliography{ms}% Produces the bibliography via BibTeX.
1038:
1039: \end{document}
1040: %
1041: % ****** End of file apssamp.tex ******
1042: