physics0505189/ms.tex
1: %\documentclass[aps,prl,preprint,groupedaddress]{revtex4}
2: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
3: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
4: 
5: \usepackage{graphicx}% Include figure files
6: \usepackage{dcolumn}% Align table columns on decimal point
7: \usepackage{bm}% bold math
8: \usepackage{color}
9: 
10: %\topmargin -3pt
11: 
12: % COLOR %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
13: %\def\note#1{{\textcolor{red}{\bf [#1]}}}  % note
14: %\def\opt#1{{\textcolor{green}{#1}}}        % optional
15: %\def\add#1{{\textcolor{blue}{#1}}}          % addition
16: \def\note#1{{{}{#1}}}  % note
17: \def\opt#1{{{}{#1}}}        % optional
18: \def\add#1{{{}{#1}}}          % addition
19: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
20: 
21: \begin{document}
22: 
23: \title{Shell to shell energy transfer in MHD,\\
24:        Part II: Kinematic dynamo}
25: 
26: \author{Pablo D. Mininni}
27:     \email{mininni@ucar.edu}
28: \author{Alexandros Alexakis}
29:     \email{alexakis@ucar.edu}
30: \author{Annick Pouquet}
31: \affiliation{National Center for Atmospheric Research,
32:              P.O. Box 3000, Boulder, Colorado 80307}
33: 
34: \date{\today}
35: 
36: \begin{abstract}
37: 
38: We study the transfer of energy between different scales for forced 
39: three-dimensional MHD turbulent flows in the kinematic dynamo regime.
40: Two different forces are examined:  a non-helical Taylor Green flow with
41: magnetic Prandtl number $P_M=0.4$, and a helical ABC flow with $P_M=1$.
42: This analysis allows us to examine which scales of the velocity flow
43: are responsible for dynamo action, and \add{identify} which scales of the magnetic field
44: receive energy directly from the velocity field and which scales receive
45: magnetic energy through the cascade of the magnetic field from large to
46: small scales. Our results show that the turbulent velocity fluctuations   
47: are responsible for the magnetic field amplification in the small scales 
48: (small scale dynamo) while the large scale field is amplified mostly due
49: to the large scale flow. A direct cascade of the magnetic field energy from
50: large to small scales is also present and is a complementary mechanism for the
51: increase of the magnetic field in the small scales.
52: Input of energy from the velocity field in the small magnetic scales
53: dominates over the energy that is cascaded down from the large scales
54: until the large-scale peak of the magnetic energy spectrum is reached.
55: At even smaller scales, most of the magnetic energy input is from the cascading process.
56: 
57: \end{abstract}
58: 
59: \pacs{47.65.+a; 47.27.Gs; 95.30.Qd}
60: \maketitle
61: 
62: %\noindent 
63: %\note{Red: notes}\\
64: %\opt{Green: optional additions}\\
65: %\add{Blue: important additions}
66: 
67: \section{ \label{Intro} Introduction }
68: 
69: Dynamo action is often invoked to explain the generation and
70: sustainment of magnetic fields in astronomical objects. In the
71: magnetohydrodynamic (MHD) dynamo, an initially small magnetic field
72: is amplified by currents induced solely by the motion of a
73: conducting fluid \cite{Moffatt}. In typical astrophysical
74: situations where amplified magnetic fields are met, the velocity
75: field is composed of a large scale flow (e.g. rotation and/or
76: meridional flows) together with turbulent fluctuations in smaller
77: scales. As an example, in the sun both large and small scale magnetic fields
78: are observed. The large scale components \add{of the magnetic field} 
79: are generated by a large-scale flow \cite{Dikpati99}. 
80: Simulations also show that the small
81: scale magnetic fields can be generated by turbulent fluctuations in
82: the convective region \cite{Cattaneo99}.
83: Understanding the generation of magnetic fields under these conditions and
84: the role played by the two components of the flow
85: (large scale and turbulent) is today a crucial aspect of dynamo theory. 
86: 
87: Dynamos are often classified as small-scale and large-scale
88: dynamos, depending on the properties of the amplified magnetic field
89: \cite{Vainshtein72}. In large-scale dynamos, the focus is on whether a
90: flow can amplify and sustain magnetic fields at
91: scales larger than the velocity integral scale. This interest is motivated by
92: astrophysical problems where large scale magnetic fields are
93: actually observed, such as the dipolar component in stars and
94: planets. The amplification of the magnetic field in these scales 
95: is usually explained by invoking a turbulent $\alpha$-effect and/or 
96: amplification due to a large scale flow. The linear (or kinematic) regime 
97: of large-scale dynamo action has been studied with the use of mean field 
98: theory \cite{Steenbeck66,Krause}, MHD closures \cite{Pouquet76}, and with 
99: the aid of numerous direct numerical simulations (DNS) (see {\it e.g.} 
100: \cite{Meneguzzi81,Brandenburg01,Gomez04}).
101: In theoretical investigations of large-scale dynamo 
102: action, helical flows are generally considered, that are thought of as better
103: candidates for amplifying the magnetic field at larger scales.
104: However the presence of helicity is not necessarily needed to
105: generate large scale magnetic fields \cite{Gilbert88}; 
106: they can also be amplified in non-helical flows if
107: anisotropy \cite{Nore97}, or other mean field effects 
108: \cite{Urpin02,Geppert02} are present. 
109: 
110: % \bibitem{Gilbert88}
111: %Gilbert, A., U. Frisch, \& A. Pouquet, 1988: Helicity is unnecessary for
112: %dynamo action but it helps. {\it Geophys. Astrophys. Fluid
113: %Dynamics}, {\bf 42}, 151--161 (1988).
114: 
115: Small-scale dynamos on the other hand amplify magnetic fields on scales
116: smaller than the energy containing scales of the turbulence
117: \citep{Kazantsev68,Zeldovich,Schekochihin02,Haugen03,Schekochihin04b}.
118: Theoretical investigations usually involve assumptions of non-helical
119: velocity fields, $\delta$-correlated in time (as a simplifying
120: approximation to a turbulent flow), and often the limit of
121: large magnetic Prandtl number $P_M>1$ is considered. 
122: Numerically small-scale dynamos have been investigated in Refs. 
123: \cite{Meneguzzi81,Haugen03,Schekochihin04b}. 
124: Here we note that an argument due to Batchelor
125: \cite{Batchelor50} suggests that this dynamo can only operate if
126: $P_M>1$. However, there are reasons to believe the small-scale
127: dynamo can work even when $P_M<1$ if the magnetic Reynolds number
128: $R_M$ is large enough \cite{Schekochihin04a,Ponty05}.
129: 
130: However, this separation between large-scale and small-scale dynamos
131: is in some cases artificial and may be misleading. Most astronomical objects display
132: a large scale flow with turbulent fluctuations at smaller scales, 
133: and both large and small scales magnetic fields are observed. 
134: The transition between the two magnetic fields is often smooth 
135: and a clear distinction between the two cannot be made. 
136: This has led some authors to develop models trying to unify the
137: two regimes \cite{Subramanian99}. 
138: Furthermore, the two amplification mechanisms
139: in the small and large scales are coupled in many cases 
140: and cannot be considered independently.
141: According to mean field theory \cite{Krause,Zeldovich}, 
142: the large scale magnetic field in
143: a turbulent dynamo results from the small scale (helical) velocity fields.
144: Moreover, concerning the amplification of small scale magnetic fields, 
145: it has been argued that when a large scale magnetic field is present, 
146: small scales can be generated
147: by the distortion of large scale field lines (see e.g. \cite{Moffatt}), 
148: even in the absence of self-excitation (small-scale dynamo action).
149: This is a common assumption in mean field dynamos, where it is 
150: often considered that the needed small scale magnetic fields are only 
151: fed by the large scale field through a nonlinear cascade process.
152: 
153: In the presence of both a large scale flow and turbulent
154: fluctuations, the role played by the different scales involved in
155: the amplification process is thus of crucial importance and 
156: is not well understood. When magnetic fields are present at scales 
157: both smaller and larger than the energy containing scales of the 
158: velocity field, it is not clear what portion of the small scale magnetic 
159: field is generated by direct cascade of magnetic energy from the large scales, 
160: and what from self-excitation. 
161: Furthermore it is not well understood what portion of the 
162: amplification of the large or small scale dynamo is due to the 
163: forced \add{component of the} flow and what part is due to the turbulent fluctuations which emerge through nonlinear interactions at high Reynolds number.
164: To answer these questions, a detailed study of the energy transfer
165: from the different velocity scales to the different magnetic scales 
166: is required. This kind of approach naturally raises the question of 
167: the locality (in Fourier space) of the interactions that are taking 
168: place in a turbulent dynamo.
169: 
170: In a companion paper \cite{Alexakis05} (hereafter referred to as Paper I),
171: the transfer of energy between the velocity and magnetic
172: field at different scales was studied for mechanically forced MHD turbulence
173: in a steady state where both fields are in quasi equipartition,
174: by introducing the energy transfer functions
175: between different shells of wavenumbers in Fourier space.
176: %different transfer functions for MHD turbulence were introduced, and
177: %the transfer of energy was studied for mechanically forced MHD turbulence
178: %in a steady state. 
179: In this paper, we present shell-to-shell
180: energy transfers during the kinematic regime of two different MHD dynamos.
181: Our main interest is to identify which velocity field scales are 
182: responsible for the amplification of the large and small scale 
183: magnetic field, which scales of the magnetic field
184: receive most of the energy, and how the magnetic energy cascades 
185: among the different scales.
186: 
187: In Sec. \ref{Transfer} we present a brief review of the equations
188: and definition of transfer functions needed to study this problem, and
189: in Sec. \ref{Results} we give the results from simulations; we also
190: discuss in this section some details of the nonlinear saturation of the
191: dynamo. Finally, in Sec. \ref{Discussion} we present the conclusions of
192: our work.
193: 
194: \section{ \label{Transfer} The transfer functions}
195: 
196: We will consider the incompressible MHD equations,
197: \begin{equation}
198: \partial_t {\bf u} + {\bf u}\cdot \nabla {\bf u} = - \nabla p +
199:     {\bf b}\cdot \nabla {\bf b} + \nu \nabla^2 {\bf u} +{\bf f}
200: \label{eq:momentum}
201: \end{equation}
202: %
203: \begin{equation}
204: \partial_t {\bf b} + {\bf u}\cdot \nabla {\bf b} =
205:     {\bf b}\cdot \nabla {\bf u} + \eta \nabla^2 {\bf b} ,
206: \label{eq:induc}
207: \end{equation}
208: where ${\bf u}$ is the velocity field, ${\bf b}$ is the magnetic
209: field, $\nu$ is the kinematic viscosity, $\eta$ is the magnetic
210: diffusivity, $p$ is the total pressure and ${\bf f}$ a constant external force. This equations
211: are accompanied by the conditions
212: $\nabla \cdot {\bf u} = 0 = \nabla \cdot {\bf b}$. Equations
213: (\ref{eq:momentum}) and (\ref{eq:induc}) are solved in a periodic
214: domain using a pseudospectral method with the $2/3$ dealiasing
215: rule and second order Runge-Kutta to advance in time.
216: 
217: We are interested in the kinematic regime of the dynamo, where a
218: small magnetic seed is amplified exponentially without modifying the
219: velocity field (i.e., the effect of the Lorentz force on the
220: velocity field is negligible). 
221: 
222: To this end, we made two numerical
223: simulations using a grid of $256^3$ points under the following
224: procedure. First, a hydrodynamic simulation was performed to obtain
225: a turbulent steady state. 
226: Then, a random small magnetic field was
227: introduced and the simulation was carried to observe exponential
228: amplification of the magnetic energy. The data were analyzed
229: during this stage and as the systems approached saturation.
230: 
231: Two expressions for the external force were used:
232: Taylor-Green (hereafter called TG), and ABC. The TG forcing
233: is non-helical (${\bf f \cdot \nabla \times f}=0$ pointwise), while 
234: the ABC forcing is of maximum helicity and the resulting flow has 
235: non-negligible helicity (for a description of the resulting flows see 
236: e.g. \cite{Mininni05a,Mininni05b}). 
237: In both simulations, the amplitude of the external force was set to
238: obtain a unity r.m.s. velocity, and the characteristic wavenumber of
239: the force was chosen to obtain a large scale flow at $k_F \sim 3$.
240: The TG simulation had $\nu = 2 \times 10^{-3}$ and $\eta = 5 \times
241: 10^{-3}$ (the magnetic Prandtl number in this simulation was $P_M =
242: \nu/\eta = 0.4$). In the ABC run, $\nu = \eta = 2 \times 10^{-3}$
243: ($P_M = 1$). \add{The mechanical Reynolds numbers reached by the two flows 
244: are $Re=675$ for the Taylor Green flow and $Re=820$ for the ABC}
245: \cite{Mininni05a,Mininni05b}.
246: 
247: As we stated in the introduction, we are interested in
248: quantifying the rate of energy transfer from the different
249: scales of the velocity field to the different scales of the magnetic
250: field. To rigorously  define the velocity and magnetic field at
251: different scales we introduce the shell filtered
252: velocity and magnetic field components ${\bf u}_K({\bf x})$ and 
253: ${\bf b}_K({\bf x})$,
254: where the subscript $K$ indicates that the field has been filtered to
255: keep only the modes in the Fourier shell $[K,K+1)$ (hereafter called
256: the shell $K$). Clearly the sum of all the $K$ components gives back
257: the original field. We are interested therefore in the rate that 
258: energy from the velocity or magnetic field at a given shell $Q$ 
259: is transferred into kinetic or magnetic energy at another shell $K$. 
260: From the MHD equations, 
261: by doting Eq. (\ref{eq:induc}) with ${\bf b}_K$ and integrating over 
262: space, we obtain the evolution of the magnetic 
263: energy $E_b(K)=\int b_K^2/2 \, dx^3$ in the shell $K$:
264: \begin{equation}
265: \partial_t { E}_{b}(K) = 
266: \sum_Q [{\mathcal T}_{ub}(Q,K)+{\mathcal T}_{bb}(Q,K)] - 
267: \eta {\mathcal D}_b(K),
268: \label{eq:Eb}
269: \end{equation}
270: where we have introduced the two transfer functions ${\mathcal T}_{ub}(Q,K)$ 
271: and ${\mathcal T}_{bb}(Q,K)$ as defined below.
272: The transfer rate of kinetic energy 
273: at the shell $Q$ into magnetic energy at the shell $K$ is defined as:
274: \begin{equation}
275: {\mathcal T}_{ub}(Q,K) = \int{ {\bf b}_K ({\bf b}\cdot\nabla){\bf
276: u}_Q
277:     d{\bf x}^3} , \label{eq:Tub}
278: \end{equation}
279: and the transfer rate of magnetic energy
280: from the shell $Q$ into the shell $K$ is defined as:
281: \begin{equation}
282: {\mathcal T}_{bb}(Q,K) = -\int{ {\bf b}_K ({\bf u}\cdot\nabla){\bf b}_Q
283:     d{\bf x}^3} . \label{eq:Tbb}
284: \end{equation}
285: %
286: The transfer ${\mathcal T}_{ub}(Q,K)$ is due to the stretching of magnetic 
287: field lines by the velocity field gradients and leads to 
288: energy input in the magnetic field.  This term is responsible for dynamo 
289: action, i.e. conversion of kinetic energy into magnetic energy.
290: The function ${\mathcal T}_{bb}(Q,K)$ is due to the advection of
291: magnetic field \add{vector components} by the velocity field and it does not amplify
292: the total magnetic energy. Instead, it is responsible for the
293: redistribution of magnetic energy among the different shells
294: and it is related with the cascade of magnetic energy from larger to 
295: smaller scales. Finally, we introduced the dissipation rate 
296: ${\mathcal D}_b(K)$ in the shell $K$ defined as:
297: \[{\mathcal D}_b(K)=-\int |\nabla \times {\bf b}_K|^2 d{\bf x}. \]
298: More  detailed definitions of these transfer terms and their general 
299: properties can be found in Paper I.
300: 
301: We measured the transfer functions based on Eqs. 
302: (\ref{eq:Tub}) and (\ref{eq:Tbb}) using ten different outputs 
303: for each run during the kinematic regime. 
304: The transfers were normalized using the total magnetic energy,
305: and were then averaged. As the system was approaching saturation
306: and was deviating from the exponential growth, single time
307: outputs were used and the transfer functions were normalized
308: using the total magnetic energy but were not averaged, since in this
309: stage the normalized magnetic energy spectrum is changing with time.  
310: From here on, we will use the notations ${\mathcal T}_{ub}(Q,K)$ and
311: ${\mathcal T}_{bb}(Q,K)$ for the normalized transfer functions 
312: ${\mathcal T}_{ub}(Q,K)/\sum_{K'}E_b(K')$ and
313: ${\mathcal T}_{bb}(Q,K)/\sum_{K'}E_b(K')$, unless otherwise noted.
314: 
315: \section{ \label{Results} Results}
316: 
317: \subsection{ \label{kinematic} The kinematic regime}
318: 
319: \begin{figure}
320: \includegraphics[width=8cm]{fig_01}
321: \caption{\label{fig_01} Spectra of kinetic energy (thick solid line)
322: and magnetic energy (thin solid line) \add{scaled up by a factor $10^6$}
323: for the Taylor Green runs during 
324: the kinematic dynamo regime. The dashed line indicates the Kolmogorov 
325: slope as a reference. Note that during this stage, all the magnetic 
326: modes grow with approximately the same rate.}
327: \end{figure}
328: \begin{figure}
329: \includegraphics[width=8cm]{fig_02}
330: \caption{\label{fig_02} Spectra of kinetic energy (thick solid line)
331: and magnetic energy (thin solid line) \add{scaled up by a factor $10^2$}
332: for the ABC runs. The dashed 
333: line indicates the Kolmogorov slope as a reference.}
334: \end{figure}
335: %
336: We begin by describing the general properties of the two dynamos
337: investigated in this work. Figure \ref{fig_01} shows the kinetic
338: and magnetic energy spectra for a TG simulation in the kinematic
339: dynamo regime. In Figure \ref{fig_02} we show the same spectra for the
340: ABC run. Note that the kinetic energy spectrum peaks in both cases
341: at $k_F \sim 3 $, where a well-defined large scale flow is present.
342: For larger wavenumbers the spectrum presents a short inertial range,
343: with Kolmogorov scaling.
344: During the kinematic regime, 
345: the magnetic energy spectrum peaks at small scales ($k \sim 9$ for TG 
346: and $k \sim 12$ for ABC) and all the modes grow
347: exponentially with the same rate. As a result, all the spectra (and
348: transfer functions) preserve their dependence with wavenumber 
349: \add{(up to an amplitude normalization)} as time evolves.
350: 
351: \begin{figure*}
352: \includegraphics[width=16cm,height=6.5cm]{fig_03}
353: \caption{\label{fig_03}Transfer functions (a) ${\mathcal T}_{ub}$, 
354: and (b) ${\mathcal T}_{bb}$ (see Eqs. (4) and (5)), as a function of $Q$ and $K$ in the TG run 
355: during the kinematic regime. ${\mathcal T}_{ub}$ is positive for all 
356: values of $Q$ and $K$ shown. \add{ Each point (Q,K) in panel (a) represents the rate of transfer
357: of energy from the velocity mode $Q$ to the magnetic mode $K$.
358: Each point (Q,K) in panel (b) represents the rate of transfer
359: of energy from the magnetic mode $Q$ to the magnetic mode $K$.}
360: The dashed line indicates the diagonal $K=Q$.}
361: \end{figure*}
362: %
363: To study the kinematic regime, the MHD simulations
364: were started with a random magnetic field with values of magnetic
365: energy as low as $E_M/E_K = 5 \times 10^{-9}$, 
366: to ensure that the Lorentz force was negligible at all wavenumbers 
367: even with the magnetic energy spectrum peaking at small scales. 
368: 
369: We first start with some general properties of the two transfer functions. 
370: Contour plots of the transfers ${\mathcal T}_{ub}(Q,K)$
371: and ${\mathcal T}_{bb}(Q,K)$ during the kinematic regime of the 
372: TG run are shown in Figure \ref{fig_03}. The gray scale indicates 
373: magnitude of the transfer, with `dark' being positive and `bright' 
374: negative. The figure should be
375: interpreted as follows: At a given point $(Q,K)$ on Fig. \ref{fig_03}(a),
376: where the transfer is positive (negative), energy is given
377: (received) by the velocity field at the scale $Q$ to \add{(from)} the magnetic field
378: at scale $K$. Similarly at a given point $(Q,K)$ on Fig. \ref{fig_03}(b),
379: where the transfer is positive (negative), energy is given
380: (received) by the magnetic field at the scale $Q$ to the magnetic field
381: at scale $K$. 
382: 
383: %\add{the grey-scale is not very visible on my printer at least ...}
384: 
385: Note that ${\mathcal T}_{bb}$ is by definition anti-symmetric along the 
386: diagonal $K=Q$ and is mostly concentrated in the surroundings of the 
387: diagonal. The ${\mathcal T}_{ub}$ transfer is concentrated on a triangle 
388: below the diagonal and is positive everywhere. The fact that 
389: ${\mathcal T}_{bb}$ is concentrated along the diagonal implies as we 
390: will show later locality of interactions, while the ``triangular" shape 
391: of ${\mathcal T}_{ub}$ implies long range interactions in Fourier space.
392: 
393: To draw conclusions from the functional form of the transfers we need to
394: examine their behavior for different fixed values of $K$ or $Q$. 
395: \begin{figure}
396: \includegraphics[width=8cm]{fig_04}
397: \caption{\label{fig_04}Transfer function ${\mathcal T}_{ub}(Q,K)$ 
398: (from the kinetic energy at $Q$ to the magnetic energy at $K$) for fixed 
399: values of $Q$ during the kinematic regime of the TG run.}
400: \end{figure}
401: %
402: \begin{figure}
403: \includegraphics[width=8cm]{fig_05}
404: \caption{\label{fig_05}Transfer function ${\mathcal T}_{ub}(Q,K)$ 
405: for fixed values of $Q$ during the kinematic regime of the ABC run.}
406: \end{figure}
407: %
408: Figures \ref{fig_04} and \ref{fig_05} show the ${\mathcal T}_{ub}(Q,K)$
409: function at constant values of $Q$, for the TG and ABC simulations
410: respectively. The transfer is always positive, implying that kinetic energy 
411: is transfered from all the velocity wave numbers $Q$ 
412: to magnetic energy at different $K$-shells. 
413: The transfer is maximum for wavenumbers close to $Q$, and then slowly decays. 
414: Note that in the ABC run the flow at $Q=3$ gives more energy than the
415: turbulent fluctuations ($Q=5,10,20,30$) when compared with the TG
416: simulation. This is related with the fact that in the ABC run the
417: $Q=3$ shell contains most of the kinetic helicity of the flow, an
418: ingredient known to be relevant for dynamo action \cite{Moffatt,Krause}.
419: We note here that since the transfers are normalized by the total
420: magnetic energy and the two runs have different magnetic energy spectra,
421: a direct comparison of the values of the transfers between the two runs
422: cannot be made.
423: 
424: \begin{figure}
425: \includegraphics[width=8cm]{fig_06}
426: \caption{\label{fig_06}Transfer function ${\mathcal T}_{ub}(Q,K)$ 
427: for fixed values of $K$ during the kinematic regime of the TG run.}
428: \end{figure}
429: %
430: \begin{figure}
431: \includegraphics[width=8cm]{fig_07}
432: \caption{\label{fig_07}Transfer function ${\mathcal T}_{ub}(Q,K)$ 
433: for fixed values of $K$ during the kinematic regime of the ABC run.}
434: \end{figure}
435: %
436: In Figures \ref{fig_06} and \ref{fig_07} we show the same transfer
437: function ${\mathcal T}_{ub}$ but now for constant values of $K$. 
438: The transfer is positive at all scales, pointing to the fact 
439: that all velocity
440: shells are giving energy to the magnetic field (compare this result
441: with the turbulent steady state in Paper I, where energy is being
442: transfered from the magnetic field to the velocity field at small
443: scales). A peak at $Q=3$ can be identified at all wavenumbers $K$, 
444: indicating that the large scale flow gives energy non-locally to all magnetic
445: shells. For wavenumbers $Q>3$ also a plateau can be identified, where
446: ${\mathcal T}_{ub}$ as a function of $Q$ is approximately constant.
447: The plateau drops at $K \gtrsim Q$. This region of constant
448: ${\mathcal T}_{ub}$ corresponds to all kinetic energy shells at
449: $3 < Q \lesssim K$ (the turbulent fluctuations) transferring the same
450: amount of energy to the magnetic field at the shell $K$. In the
451: ABC simulations, the role played by the turbulent fluctuations is
452: again observed to be smaller than in the TG runs when compared with
453: the large scale flow at $Q \sim 3$. 
454: 
455: \begin{figure}
456: \includegraphics[width=8cm]{fig_08}
457: \caption{\label{fig_08}Transfer function ${\mathcal T}_{bb}(Q,K)$ 
458: (from magnetic energy at shell $Q$ to magnetic energy at shell $K$) 
459: at fixed values of $Q$, during the kinematic regime of the TG and ABC 
460: runs.}
461: \end{figure}
462: %
463: The transfer of magnetic energy between different scales is shown
464: in Figure \ref{fig_08}. As previously mentioned, this transfer
465: is associated with the cascade of energy to smaller scales.
466: Each shell $Q$ is giving energy to a slightly
467: larger wavenumber $K$ (the positive peak of the curves) and receiving
468: energy from a slightly smaller wavenumber $K'$ (the negative peak
469: of the curves). There is an increase of the amplitude of the transfer
470: as the wavenumber $Q$ is increased up until the peak of the spectrum is
471: reached and then it drops again.
472: This transfer function drops fast for wavenumbers
473: $K$ and $Q$ far apart and therefore 
474: \add{indicates a local transfer of energy}.
475: 
476: 
477: We are ready now to answer some of the questions posed in the introduction.
478: First we want to consider if it is the large scale flow that drives
479: the dynamo or the turbulent fluctuations. On average the contribution
480: to the injection of magnetic energy from the large scale flow is $16\%$ 
481: for the TG flow, and $25\%$ for the ABC flow. Note that this fraction is 
482: much smaller than what is obtained in the saturated regime ($60\%$ for TG, 
483: and $75\%$ for ABC in \cite{Alexakis05}). Furthermore, the influence of the 
484: large scale flow becomes smaller as we are deeper in the inertial range. In 
485: Figure \ref{fig_09} we show the ratio:
486: \[{\mathcal R}_{LS}(K)=\sum_{Q=2,3,4}{\mathcal T}_{ub}(Q,K) / 
487:     \sum_Q{\mathcal T}_{ub}(Q,K) \]
488: that expresses the fraction of energy a magnetic shell $K$ receives only 
489: from the the large scale flow (the peak at $Q=2,3,4$ in Figs. 
490: \ref{fig_06} and \ref{fig_07}), to the total energy received by the 
491: same shell from the velocity field at all scales. For both flows the 
492: energy input from large scales becomes smaller as the wavenumber $K$ is 
493: increased and the large scale flow \add{only} dominates the injection of magnetic 
494: energy over a small range $K_F<K<K_{LS}$, with $K_{LS}\simeq 5$.
495: 
496: 
497: %%%ELAPISO
498: 
499: Another question we posed in the introduction is whether the small scale 
500: magnetic fluctuations are the result of a cascade of energy from the 
501: large scale magnetic field, or from a direct input of energy 
502: (amplification) from the velocity field. To answer this 
503: question, in Figure \ref{fig_09} we also plot the ratio:
504: \[{\mathcal R}_{C}(K)=\sum_{Q=0}^K{\mathcal T}_{bb}(Q,K) / 
505:     \sum_Q{\mathcal T}_{ub}(Q,K) \]
506: that expresses the fraction of energy a magnetic shell $K$ receives from 
507: the cascade of energy from larger magnetic scales to the total energy 
508: received in the same shell directly from the velocity field. The cascading 
509: term appears to be smaller up to a wavenumber
510: $K_C\simeq 12$ close to the peak of the magnetic energy spectrum. For 
511: $K>K_C$ there is more energy input from the cascade than the input from the 
512: velocity field. Between these two processes, a range of wavenumbers 
513: $K_{LS}<K<K_C$ exists where the amplification of the magnetic field
514: is purely dominated by injection from the turbulent velocity scales.
515: \begin{figure}
516: \includegraphics[width=8cm]{fig_09}
517: \caption{\label{fig_09} Ratio ${\mathcal R}_{LS}$  of energy received by the magnetic field 
518:     at wave number $K$ from the forced wavenumbers against all the wavenumbers, 
519:     and ratio ${\mathcal R}_{C}$ of energy received by the magnetic 
520:     field at wave number $K$ from the magnetic field at larger scales 
521:     \add{through a cascade process}
522:     against 
523:     energy received from the velocity field. The 
524:     solid lines correspond to the TG run while the dashed lines correspond 
525:     to ABC, both in the kinematic regime.}
526: \end{figure}
527: 
528: \begin{figure}
529: \includegraphics[width=8cm]{fig_10}
530: \caption{\label{fig_10}Energy received by the magnetic field at 
531:     scales larger than the forcing band ($K=1$ and 2) from the velocity 
532:     field at wavenumbers $Q$.}
533: \end{figure}
534: We also investigate the growth rate of large scale magnetic fields restricted 
535: to the shells $K=1,2$. In order to 
536: obtain the highest possible Reynolds numbers in the simulations, the scale 
537: separation between the forcing band and the large scale magnetic field was chosen to be small 
538:  \add{and therefore an investigation of the alpha dynamo effect is not possible in the present study.
539: Here we just limit ourselves to investigate which scale of the velocity
540: field is responsible for the input of energy in the large scales $K=1$ and $K=2$
541: of the magnetic field.} 
542: In figure \ref{fig_10} we show the transfer of energy from
543: the velocity field to these large scale modes. Although there is a 
544: contribution from the turbulent fluctuations, the bulk of the energy 
545: originates from the forced modes. A similar result was obtained in 
546: Ref. \cite{Brandenburg01}, in simulations with larger scale separation 
547: ($k_F \simeq 5$) but lower Reynolds numbers.
548: 
549: \begin{figure}
550: \includegraphics[width=8cm]{fig_11}
551: \caption{\label{fig_11} Terms \add{determining} the growth rate of magnetic 
552:     energy [see Eq. (\ref{eq:growth})] as a function of wavenumber $K$ 
553:     for the TG run during the kinematic regime. The dashed line is 
554:     the energy received by the magnetic field from the velocity field, 
555:     and the dash-dotted line is the cascade of magnetic energy. The 
556:     solid line is the total energy received by the magnetic field, while 
557:     the dotted line is the Ohmic dissipation. The difference between the 
558:     \add{last} two curves gives the growth rate.} 
559: \end{figure}
560: %
561: \begin{figure}
562: \includegraphics[width=8cm]{fig_12}
563: \caption{\label{fig_12}The budget of magnetic energy giving rise to the 
564:     growth rate for the ABC run during the kinematic regime. Labels are 
565:     as in Fig. \ref{fig_11}.}
566: \end{figure}
567: %
568: Finally, in both simulations, all wavenumbers are observed to grow with
569: the same growth rate during the kinematic regime. 
570: To investigate this
571: we can write the energy budget using the induction equation
572: (\ref{eq:induc}) in Fourier space. Taking the dot product with the
573: magnetic field ${\bf b}_K$ at the shell $K$, and dividing by the
574: magnetic energy $E_b(K)$ in that shell, we finally obtain
575: \begin{eqnarray}
576: \label{growth}
577: \frac{1}{E_b(K)} \frac{\partial}{\partial t} E_b(K) &=&
578: \frac{1}{E_b(K)} \sum_Q [{\mathcal T}_{ub}(Q,K)+{\mathcal T}_{bb}(Q,K)]
579:     \nonumber \\
580: - \eta K^2, \label{eq:growth}
581: \end{eqnarray}
582: where the simplification ${\mathcal D}(K)\simeq K^2E_b(K)$ was used.
583: The left-hand side of equation (\ref{growth}) gives the growth rate $\sigma$. 
584: The first two terms in the right hand side are 
585: the energy received by the magnetic field at the shell $K$ from
586: the velocity field and from the magnetic field at all scales.
587: The last term is the Ohmic dissipation.
588: In figures \ref{fig_11} and \ref{fig_12} we show
589: each term of this budget as a function of the wavenumber $K$ for the
590: TG and ABC runs. The difference between the solid line and the dotted line 
591: is the
592: growth rate. In spite of the fluctuations, the growth rate seems to
593: be constant in a wide range of wavenumbers. This is more clearly
594: observed in the ABC run because of the larger growth rate in this
595: simulation. 
596: The constant growth rate over all scales therefore is the result
597: of a balance between the energy received by the magnetic field at each
598: shell $K$ locally (from the direct cascade), non-locally (from the
599: stretching of field lines), and of the Ohmic dissipation. Note that 
600: when integrated over all $Q$, the direct cascade ${\mathcal T}_{bb}(Q,K)$ 
601: gives a negative contribution (up to $k\simeq20$ in the TG run, and larger 
602: wavenumbers for the ABC case), indicating that each magnetic shell $K$ 
603: gives locally more energy to smaller scales than what it receives from 
604: the larger scales. This is compensated by the energy injected by the 
605: velocity field through the transfer ${\mathcal T}_{ub}$.
606: 
607: \subsection{ \label{saturation} The saturation of the dynamo }
608: 
609: \begin{figure}
610: \includegraphics[width=8cm]{fig_13}
611: \caption{\label{fig_13} Transfer function from the kinetic energy 
612: at $Q$ to the magnetic energy at $K=20$ 
613: \add{for three different times as the magnetic 
614: field approaches saturation in the TG run.}}
615: \end{figure}
616: %
617: \begin{figure}
618: \includegraphics[width=8cm]{fig_14}
619: \caption{\label{fig_14}Transfer function from the kinetic energy 
620: at $Q=16$ to the magnetic energy at $K$
621: \add{for three different times as the magnetic
622: field approaches saturation in the TG run.}} 
623: \end{figure}
624: %
625: \begin{figure}
626: \includegraphics[width=8cm]{fig_15}
627: \caption{\label{fig_15}Transfer of magnetic energy at shell $Q$ 
628: to shell $K$ (for $K=10$ and $K=20$)  
629: \add{for three different times as the magnetic
630: field approaches saturation in the TG run.}}
631: \end{figure}
632: %
633: In this section we discuss the evolution of the transfer function for the TG run
634: as the dynamo approaches the nonlinear saturation. The ABC run shows
635: similar features except for a slow growth of the magnetic field at $K=1$
636: that finally dominates the magnetic energy. The transfer of magnetic
637: energy at the large scales in this case has been studied in
638: \cite{Brandenburg01}. For details of the transfer in the final state
639: reached by the two simulations, we refer the reader to Paper I.
640: 
641: Figure \ref{fig_13} shows the ${\mathcal T}_{ub}$ transfer at $K=20$
642: as a function of time, as the nonlinear saturation takes place. Each
643: transfer has been normalized by the total magnetic energy at that time.
644: The transfer at $t=13.5$ corresponds to the kinematic regime. At $t=73.5$,
645: the small scale magnetic field saturates and stops growing (see
646: \cite{Mininni05a}). The velocity field turbulent fluctuations are
647: \add{partially} quenched, and the kinetic energy at small scales is reduced. This
648: suppression of turbulence by the magnetic field has been previously
649: observed \cite{Brandenburg01,Mininni05a}, 
650: and as a result the transfer of energy 
651: to the magnetic field at wavenumber $K$
652: from the velocity field between $3<Q \lesssim K$
653: is also strongly reduced. However, the large scale velocity field at
654: $Q=3$ keeps transferring energy to the magnetic field. In this stage,
655: the large scale magnetic field keeps growing until the large scales
656: are dominated by magnetic energy, and suppress even more the turbulent
657: fluctuations. At $t=178.5$, the system has finally reached the steady
658: state. The magnetic field at each shell $K$ is sustained by both the
659: large scale flow and the turbulent fluctuations, but now the amplitude
660: at all scales has been \add{reduced due to}  the Lorentz force. Note also that
661: the transfer at scales $Q \gtrsim K$ is now negative, pointing out to the fact that
662: the magnetic energy is feeding the velocity field at small scales.
663: 
664: In figure \ref{fig_14}, we show the transfer ${\mathcal T}_{ub}(Q,K)$
665: for a fixed velocity wave number
666: $Q=16$ for the same times as in figure \ref{fig_13}. Again,
667: the transfers have been normalized by the total magnetic energy at
668: each time. In the kinematic regime, the velocity field at $Q=16$ gives
669: energy (positive transfer) to the magnetic field at all scales. When
670: the small scale magnetic field saturates ($t=73.5$) the velocity field
671: at $Q=16$ gives energy only to magnetic shells with $K \ge Q$, and modes
672: with $K < Q$ receive almost no energy. This regime corresponds to the
673: case where the magnetic field at small scales has saturated and its energy
674: is sustained by the dynamo without further amplification. The large
675: scale field keeps growing, mostly fed by the large scale flow
676: at $Q=3$ as previously discussed. Finally, in the saturated regime
677: ($t=178.5$) the magnetic field at $Q=16$ receives energy from all the
678: kinetic shells $Q$ with $K \ge Q$, while it gives energy (negative transfer)
679: to all kinetic shells with $K < Q$.
680: 
681: Finally, in
682: Figure \ref{fig_15} we show the transfer of magnetic to magnetic
683: energy ${\mathcal T}_{bb}$ as a function of time. The local transfer
684: between magnetic shells is not changing as much as ${\mathcal T}_{ub}$
685: as the dynamo saturates, except for a change in the amplitude. 
686: \add{The amplitude is decreasing as saturation is approached,
687: an effect that appears first in the small scales.
688: We note that the role of the local direct cascade ${\mathcal T}_{bb}$
689: becomes more dominant when compared with the ${\mathcal T}_{bb}$ term
690: as saturation is approached \cite{Alexakis05}.}
691: 
692: \section{ \label{Discussion} Discussion }
693: 
694: In this work, the transfer of energy from the different scales of the velocity field
695: to the different scales of the magnetic field has been studied.
696: We now give a brief summary of our most important results, and discuss
697: the implications for dynamo theory.
698: 
699: It has been shown that the magnetic field grows as the result of a complex interaction
700: between large and small scales. Both large and small scales of the flow give a
701: contribution to the dynamo. 
702: The amplification of the large scale magnetic field 
703: during the kinematic regime is due to
704: the large scale flow (large scale dynamo).
705: At smaller scales, most of the
706: injection of energy from the velocity field 
707: during the kinematic regime is due to the turbulent
708: fluctuations of the velocity field (small scale dynamo). 
709: A competing mechanism for the amplification of  
710: the magnetic field at the small scales is the cascade of magnetic field energy
711: from the large scales to the small scales that is also transferring energy
712: to the small scales. 
713: The rate that energy is transfered to the small scales through the
714: cascading process is smaller than the rate that the velocity field is 
715: injecting energy at the small scales for a finite range of wave numbers.  
716: For sufficiently small scales 
717: (close to the scale at which the peak of the magnetic energy spectrum is reached), 
718: the cascading term becomes larger than the small scale dynamo term.
719:  
720: The results in this paper and the formalism used
721: help us understand and classify the dynamo processes involved
722: in the amplification of the magnetic field. In this formalism,
723: we were able to measure and compare each component in the dynamo process that
724: is involved in the amplification of the magnetic field in both small 
725: and large scales. Therefore, 
726: %\note{from now on and forever after} 
727: we can distinguish
728: between different dynamos based on whether the cascading terms dominate over
729: the injection terms in the small scales and on whether the turbulent fluctuations
730: are more dominant for the generation of the magnetic field when compared with the
731: input from the large scale flow. Of course, this
732: is not the only possible distinction that can be made between dynamos; 
733: however, it is an important step 
734: %\note{in clarifying this messy field}
735: towards classifying dynamos in the presence of both a large scale
736: flow and turbulent fluctuations.
737: 
738: 
739: Finally we would like to note that the investigation of the growth of large scale magnetic
740: field with enough scale separation between  the forced scale and the domain
741: size is required to study \add{processes such as the dynamo $\alpha$-effect 
742: and the inverse cascade of magnetic helicity}.
743: A similar analysis \add{will be performed in this context in future works.}
744: 
745: %We rock! 
746: 
747: \begin{acknowledgments}
748: The authors are grateful to J. Herring for valuable discussions
749: and his careful reading of this document.
750: Computer time was provided by NCAR. The NSF grant CMG-0327888
751: at NCAR supported this work in part and is gratefully acknowledged.
752: \end{acknowledgments}
753: 
754: \bibliography{ms}% Produces the bibliography via BibTeX.
755: 
756: \end{document}
757: