1: \documentclass[prl,twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
2: %\documentclass[prl,preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
3: \usepackage{graphicx}
4: \usepackage{dcolumn}
5: \usepackage{color}
6:
7: \def\NOTE#1{{\textcolor{red}{\bf [#1]}}} % note
8: \def\ADD#1{{\textcolor{blue}{#1}}} % question
9: \def\AD#1{{\textcolor{magenta}{#1}}} % plug a value, a ref, ...
10: \def\DEL#1{{\textcolor{green}{ #1}}} % suggested deletion in text
11: \def\BB#1{{\textcolor{blue}{\bf #1}}}
12:
13: % table definitions
14: \def\hfq{\hfill\quad}
15: \def\cc#1{\hfq#1\hfq}
16: \def\tvi{\vrule height 12pt depth 5pt width 0pt}
17: \def\traithorizontal{\noalign{\hrule}}
18: \def\tv{\tvi\vrule}
19: %
20: \newcommand{\be}{\begin{equation}}
21: \newcommand{\ee}{\end{equation}}
22: \newcommand{\uh}{\hat{u}}
23: \def\eg{{\it e.g.}\ }
24: \def\etal{{\it et al.}}
25: \def\ie{{\it i.e.}\ }
26: \def\lhs{{\it l.h.s.}\ }
27: \def\op{{\it op. cit.}\ }
28: \def\resp{{\it resp.}\ }
29: \def\rhs{{\it r.h.s.}\ }
30: \def\rms{{\it r.m.s.}\ }
31: \def\viz{{\it viz.}\ }
32: \def\vs{{\it vs.}\ }
33: \def\al{Alfv\'en\ }
34: \def\els{Els\"asser variables\ }
35: \def\kol{Kolmogorov\ }
36: \def\nse{Navier-Stokes equations\ }
37: \def\u{{\bf u}} \def\v{{\bf v}} \def\x{{\bf x}}
38: \def\dv{\delta {\bf v}}
39: %
40: \newcommand{\curlv} {\nabla \times {\bf v}}
41: \newcommand{\ba}{\mathbf{a}} \newcommand{\bb}{\mathbf{b}}
42: \newcommand{\bA}{\mathbf{A}} \newcommand{\bB}{\mathbf{B}}
43: \newcommand{\Asz}{A_{s_z}}
44: \newcommand{\alp}{\alpha} \newcommand{\alpm}{\alpha^{-1}}
45: \newcommand{\alpmm}{\alpha^{-1}_m} \newcommand{\alpmv}{\alpha^{-1}_v}
46: \newcommand{\bc}{\mathbf{c}} \newcommand{\bd}{\mathbf{d}}
47: \newcommand{\bj}{\mathbf{j}} \newcommand{\bk}{\mathbf{k}}
48: \newcommand{\bom}{\mbox{\boldmath $\omega$}}
49: \newcommand{\bomp}{\mbox{\boldmath $\omega^+$} }
50: \newcommand{\bomm}{\mbox{\boldmath $\omega^-$} }
51: \newcommand{\bompm}{\mbox{\boldmath $\omega^{\pm}$} }
52: \newcommand{\br}{\mathbf{r}}
53: \newcommand{\bu}{\mathbf{u}} \newcommand{\bv}{\mathbf{v}}
54: \newcommand{\bAs}{\mathbf{A_s}} \newcommand{\bjs}{\mathbf{j_s}}
55: \newcommand{\bus}{\mathbf{u_s}} \newcommand{\bBs}{\mathbf{B_s}}
56: \newcommand{\boms}{\mbox{\boldmath $\omega_s$}} \newcommand{\bw}{\mathbf{w}}
57: \newcommand{\bx}{\mathbf{x}} \newcommand{\bxp}{\mathbf{x^{\prime}}}
58: \newcommand{\bzp}{\mathbf{z^{+}}} \newcommand{\bzm}{\mathbf{z^{-}}}
59: \newcommand{\bzpm}{\mathbf{z^{\pm}}} \newcommand{\bzmp}{\mathbf{z^{\mp}}}
60: \newcommand{\ca}{{\rm a}} \newcommand{\caa}{{\rm aa}}
61: \newcommand{\cab}{{\rm a} b} \newcommand{\vba}{vb{\rm a}}
62: \newcommand{\K}{{\cal K}} \newcommand{\ud}{{\langle{u}^2 \rangle}}
63: \newcommand{\li}{\ell_{I}} \newcommand{\Rla}{R_{\lambda}}
64: \newcommand{\up}{{{\bf u}({\bf x})}} \newcommand{\R}{{\cal R}}
65: \newcommand{\dr}{{\partial_r}} \newcommand{\dt}{{\partial_t}}
66: \newcommand{\vg}{{{\bf v(x)} \cdot \nabla}}
67: %
68: %\topmargin -3pt
69:
70: \begin{document}
71: \title{The imprint of large-scale flows on turbulence}
72:
73: \author{A. Alexakis, P.D. Mininni and A. Pouquet}
74: \affiliation{NCAR, P.O. Box 3000, Boulder, Colorado 80307-3000, U.S.A.
75: }
76: \date{\today}
77:
78: \begin{abstract}
79:
80: We investigate the locality of interactions in hydrodynamic turbulence using
81: data from a direct numerical simulation on a grid of $1024^3$ points; the
82: flow is forced with the Taylor-Green vortex. An inertial range for the energy
83: is obtained in which the flux is constant and the spectrum follows an
84: approximate Kolmogorov law. Nonlinear triadic interactions are dominated by
85: their non-local components, involving widely separated scales. The resulting
86: nonlinear transfer itself is local at each scale but the step in the energy
87: cascade is independent of that scale and directly related to the integral
88: scale of the flow. Interactions with large scales represent 20\% of the total energy flux.
89: %The energy flux resulting from interactions with the
90: %large scale flow is 20\% of the total flux.
91: Possible explanations for the
92: deviation from self-similar models, the link between these findings and
93: intermittency, and their consequences for modeling of turbulent flows
94: are briefly discussed.
95: \end{abstract}
96: \pacs{47.27.Eq,47.27.Ak,47.65.+a}
97: \maketitle
98:
99: Flows in nature are often in a turbulent state driven by large scale forcing
100: (\eg novae explosions in the interstellar medium) or by instabilities
101: (\eg convection in the sun). Such flows involve a huge number of coupled
102: modes
103: %of different scales
104: leading to great complexity both in their temporal
105: dynamics and in the physical structures that emerge. Many scales are excited,
106: for example from the planetary scale to the kilometer
107: for convective clouds in the atmosphere, and much smaller scales when considering
108: microprocesses such as droplet formation. The question then arises
109: concerning the nature of the interactions between such scales:
110: are they predominantly local, involving only eddies of similar size, or
111: are they as well non-local? It is usually assumed that the dominant mode
112: of interaction is the former, and this hypothesis is classically viewed as
113: underlying the Kolmogorov phenomenology that leads to the prediction of a
114: $E(k)\sim k^{-5/3}$ energy spectrum; such a spectrum has been observed in
115: a variety of contexts although there may be small corrections to this
116: power-law due to the presence in the small scales of strong localized
117: structures, such as vortex filaments \cite{kaneda}.
118:
119: Several studies have been devoted to assess the degree of locality of
120: nonlinear interactions, either through modeling of turbulent flows, as
121: is the case with rapid distortion theory (RDT) \cite{LDN01} or Large Eddy
122: Simulations (LES) \cite{Zhou}, or through the analysis of direct numerical
123: simulations (DNS) of the Navier-Stokes equations (see \eg
124: \cite{Zhou,AD90,BW94}), and more recently through rigorous bounds
125: \cite{Eyink}. The spatial resolution in the numerical investigations
126: was moderate, without a clearly defined inertial range and the
127: differentiation between local and non-local interactions was somewhat
128: limited. Thus, a renewed analysis at substantially higher Reynolds
129: numbers in the absence of any modeling is in order; we address this issue
130: by analyzing data stemming from a newly performed DNS on a grid of $1024^3$
131: points using periodic boundary conditions.
132:
133: The governing Navier-Stokes equation for an incompressible velocity field
134: $\bv$, with ${\cal P}$ the pressure,
135: ${\bf F}$ a forcing term and $\nu=3\times10^{-4}$ the viscosity, reads:
136: \begin{equation}
137: \frac{\partial {\bf v}}{\partial t}+ {\bf v} \cdot \nabla {\bf v} =
138: -\nabla {\cal P} +\nu \nabla^2 \bv + {\bf F} \label{NS}
139: \end{equation}
140: together with ${\bf \nabla} \cdot {\bf v} =0$. Specifically, we consider
141: the swirling flow resulting from the Taylor-Green vortex~\cite{meb}:
142: \begin{equation}
143: {{\bf F}_{\rm TG}(k_0)}= { 2F } \, \left[
144: \begin{array}{c}
145: \sin(k_0~x) \cos(k_0~y) \cos(k_0~z) \\
146: - \cos(k_0~x) \sin(k_0~y) \cos(k_0~z)\\ 0
147: \end{array} \right] \ ,
148: \label{eq:Ftg}
149: \end{equation}
150: with $k_0=2$.
151: This forcing generates cells that have locally differential rotation and
152: helicity, although its net helicity is zero. The resulting flow models
153: the fluid between two counter-rotating cylinders \cite{meb} and has been
154: used widely to study turbulence, including studies in the context of the
155: generation of magnetic fields through dynamo instability
156: \cite{GydroSpecialIssue}. The Reynolds number based on the integral
157: scale $L \equiv 2\pi \int E(k)k^{-1} d k/E \approx 1.2 $ (where $E$ is the
158: total energy), is $R_e \equiv U L/\nu \approx 4000$, where $U$ is the
159: r.m.s velocity. The Reynolds number based on the Taylor scale
160: $\lambda \equiv 2\pi (E/\int k^2E(k)d k)^{1/2} \approx 0.24 $, is
161: $R_{\lambda} \approx 800$.
162:
163: The code uses a dealiased pseudo-spectral method, with maximum wavenumber
164: $k_{max}=341$ and $k_{max}\eta=1.15$, where
165: $2\pi \eta=2\pi (\nu^3/\epsilon)^{1/4}$ is the dissipation scale and
166: $\epsilon$ is the energy injection rate:
167: %This indicates that
168: the flow is
169: sufficiently resolved since $1/\eta$ is within the boundaries of the
170: wavenumbers handled explicitly in the computation.
171:
172: \begin{figure}
173: \centerline{\includegraphics[width=8.3cm]{FIG_1}}
174: \caption{Compensated energy spectrum and (inset) absolute value of the
175: energy flux $\Pi(k)$ in the stationary regime.}
176: \label{E(K)}
177: \end{figure}
178:
179: Details of the flow dynamics will be reported elsewhere; suffice it to say
180: that the flow reproduces classical features of isotropic turbulence
181: \cite{Frisch_book}: the energy spectrum is well-developed (see Fig.
182: \ref{E(K)}) with a constant energy flux for $k\in [5,20]$ and maximally
183: helical vortex tubes are found, as predicted in
184: \cite{moffatt} and shown in \cite{moffatt_tsinober,farge}. Finally,
185: the anomalous exponents of longitudinal structure functions are in
186: excellent agreement with previous studies \cite{kaneda} up to order
187: $p=8$ (see Table I), including analysis without using the extended
188: self-similarity (ESS) hypothesis \cite{ESS}.
189:
190: \begin{table}[h!]
191: \caption{\label{tab1} Order $p$ and anomalous exponents $\zeta_p$ computed
192: on two snapshots of the velocity field using the interval of scales with
193: constant energy flux; the anomalous exponents $\zeta_p^{ESS}$ are computed
194: using the ESS hypothesis.}
195: \begin{ruledtabular}
196: \begin{tabular}{ccccccccc}
197: p & 1 & 2 & 3 & 4 & 5 & 6 & 7 & 8\\ \hline
198: $\zeta_p$ &0.366 & 0.704 & 1.005 &1.271 & 1.502 & 1.703 & 1.878 & 2.029 \\
199: $\zeta_p^{ESS}$ &0.364 & 0.699 & 1 &1.264 &1.495 & 1.695 & 1.869 & 2.020
200: \end{tabular}
201: \end{ruledtabular}
202: \end{table}
203:
204: \begin{figure}
205: \centerline{\includegraphics[width=8.3cm]{FIG_2}}
206: \caption{Normalized energy transfer from the shell Q to the shell K with
207: $Q \in [10,80]$. The width of the lobes is independent of K and all the
208: peaks are at $K-Q \sim k_0$.}
209: \label{trans_fig}
210: \end{figure}
211:
212: To investigate the interactions between different scales we split the
213: velocity field into spherical shells in Fourier space of unit width,
214: \ie $\bv=\sum_K {\bf v}_K$ where ${\bf v}_K$ is the filtered velocity
215: field with $K\le|k|<K+1$ (from now on called shell $K$) \cite{note1}.
216: Usually, octave bands are used to define the shells (\ie shells
217: of width $\Delta K^n$ are used, where $\Delta K$ is a constant often set
218: to 2). This choice is based on the usual hypothesis that interactions
219: are mostly local and self-similar in Fourier space, \ie the nonlinear
220: term in eq. (\ref{NS}) couples triads of modes $(k,p,q)$ in Fourier
221: space with $k \sim p \sim q$. Since we want to verify these hypotheses,
222: we choose to use a linear step for the shells. This election does not
223: imply any loss of generality, and if interactions are indeed local our
224: results should be compatible with results using octave bands.
225:
226: From equation (\ref{NS}), the rate of energy transfer $T_3(K,P,Q)$ (a
227: third-order correlator) from energy in shell $Q$ to energy in
228: shell $K$ due to the interaction with the velocity field in shell $P$ is
229: defined as usual \cite{kraichnan,Alexakis05} as:
230: \be
231: T_3(K,P,Q) = -\int \bv_K \cdot (\bv_P \cdot \nabla) \bv_Q d{\bf x}^3 \ .
232: \label{triad_eq}
233: \ee
234: If we sum over the middle wave number $P$ we obtain the total energy
235: transfer $T_2(K,Q)$ from shell $Q$ to shell $K$:
236: %
237: \be
238: T_2(K,Q) = \sum_P T_3(K,P,Q) = -\int \bv_K \cdot (\bv \cdot \nabla)
239: \bv_Q d{\bf x}^3 \ .
240: \label{trans_eq}
241: \ee
242: Positive transfer implies that energy is transfered from shell $Q$ to $K$,
243: and negative from $K$ to $Q$; thus, both $T_3$ and $T_2$ are antisymmetric
244: in their $(K,Q)$ arguments. $T_2(K,Q)$ gives information on the
245: shell-to-shell energy transfer between $K$ and $Q$, but not about the
246: locality or non-locality of the triadic interactions themselves.
247: The energy flux plotted in Fig. \ref{E(K)}is reobtained from these
248: transfer functions as
249: $\Pi(k) = -\sum_{K=0}^k T_1(K) = -\sum_{K=0}^k \sum_Q T_2(K,Q)$. Note
250: that the transfer terms defined in Eqs. (\ref{triad_eq},\ref{trans_eq})
251: are integrated over all volume in real space. Since in periodic boundary
252: conditions there is no net flux of energy through the walls, this is
253: enough to ensure that contributions due to advection (which do not lead
254: to cascade of energy to smaller scales) cancel out (see \eg \cite{Eyink}).
255:
256: Figure \ref{trans_fig} shows the energy transfer $T_2(K,Q)$ plotted as a
257: function of $K-Q$ for 70 different values of $Q$ varying from 10 to 80.
258: For each value of $Q$, the $x$ axis shows the different $K$ shells
259: giving or receiving energy from that shell $Q$. All curves collapse
260: to a single one: the energy in shell $K$ is received locally from shells
261: with wavenumber $K-\Delta_K$ and deposited mostly in the vicinity of
262: $K+\Delta_K$, with $\Delta_K\sim k_0$ for all values in the inertial range.
263: In other words, the integral scale of the flow, related to the forcing
264: scale $k_0^{-1}$, plays a determinant role in the process of energy
265: transfer. As a result, the transfer is not self-similar, and the
266: integral length scale is remembered even deep inside the constant-flux
267: inertial range.
268:
269: This break down of self-similarity indicates that dominant triadic
270: interactions can be non-local. To examine further this point, we need
271: to investigate individual triadic interactions between Fourier
272: shells by considering the tensorial transfer $T_3(K,P,Q)$. We will
273: study three values of $Q$, ($Q=10$, $20$, and $40$); for each case, $P$
274: will run from 1 to 80, and $K$ from $Q-12$ to $Q+12$.
275:
276: \begin{figure}
277: \centerline{\includegraphics[width=8.3cm]{FIG_3}}
278: \caption{Contour levels of the transfer function $T_3(K,P,Q)$ for
279: $Q=40$. Solid lines correspond to positive transfer, and dotted lines
280: to negative transfer.}
281: \label{triad_fig1}
282: \end{figure}
283:
284: In figure \ref{triad_fig1} we show contour levels of the transfer
285: $T_3(K,P,Q)$ for $Q=40$. This figure represents energy going from a
286: shell $Q$ to a shell $K$ through interactions with modes in the shell
287: $P$. As in Fig. \ref{trans_fig}, positive transfer means the shell $K$
288: receives energy from the shell $Q$, while negative transfer implies the
289: shell $K$ gives energy to $Q$. The strongest interactions occur with $P \sim k_0$,
290: and therefore the large scale flow is involved in most of the $T_2$
291: transfer of energy from small scales to smaller scales. Note that the
292: individual triadic interactions with $P \sim k_0$ and $K \sim Q \pm k_0$
293: are two orders of magnitude larger than local triadic interactions.
294:
295: %In figure \ref{triad_fig1} we show the transfer $T_3(K,P,Q)$ for
296: %$K=Q+3$, a value chosen in the positive peak of ${\bar T}_2(K,Q)$
297: %(see fig. \ref{trans_fig}). Most interactions responsible for the
298: %bulk of the energy transfer occur with $P \sim k_0$ and therefore
299: %the large scale flow is involved in most of the transfer of energy
300: %from small scales to smaller scales. Note that the amplitude of the
301: %nonlocal interactions with the shell $P=3$ is more than ten times
302: %larger than those close to $P=10$ or higher. Even when we compare
303: %the transfer due to the wave number $P=3$ with the sum of all
304: %interactions involving shells with $P>3$, the $P=3$ transfer is more
305: %dominant by a factor of 1.5. However, the largest contribution to the
306: %transfer when $k_0$ is removed is not coming from shells with
307: %$P \sim K \sim Q$, but from shells with $P$ slightly larger than $k_0$.
308:
309: \begin{figure}
310: \centerline{\includegraphics[width=8cm]{FIG_4}}
311: \caption{Comparison of the transfer functions $T_2(K,Q)$ (solid line),
312: $T_2^{Loc}(K,Q)$ (dotted line), and $T_3(K,P=3,Q)$ (dashed line),
313: for three values of Q.}
314: \label{triad_fig2}
315: \end{figure}
316:
317: When $T_3(K,P,Q)$ in Fig. \ref{triad_fig2} is summed over all
318: values of $P$, the transfer function $T_2(K,Q)$ is recovered. This
319: allow us to define the transfer rate due to interactions with the
320: large scale flow, and due to local interactions, summing $P$ over
321: different ranges. Indeed, to further illustrate the dominance of the
322: large scale flow in the involved interactions, we compare in Fig.
323: \ref{triad_fig2} the total transfer function $T_2(K,Q)$
324: with the transfer due to the large scale flow $T_3(K,P=3,Q)$, and
325: with the transfer due to local interactions in octave bands
326: $T_2^{Loc}(K,Q)=\sum_{P=Q/2}^{2Q}T_3(K,P,Q)$. The figure indicates that
327: the transfer due to the local interactions ($Q/2 < P < 2Q$) is smaller
328: than the transfer due to the integral length scale velocity field, and
329: this behavior appears to be stronger as the value of $Q$ is increased.
330: The remaining transfer comes from interactions with $P$-shells with
331: wavenumbers between 1 and $Q/2$ (excluding $P=3$), which are also
332: non-local in nature. Therefore, as $K$ and $Q$ get larger (as we
333: go further down in the inertial range), the dominant triads
334: $(K,P,Q)$ become more and more elongated, corresponding to more nonlocal
335: interactions. As a result, detailed interactions between triads of modes
336: are nonlocal, while the transfer of energy $T_2(K,Q)$ takes place
337: between neighboring shells: local energy transfer occurs through
338: non-local interactions. These results support previous claims at
339: smaller resolution \cite{AD90,BW94,Zhou} that a significant
340: role in the cascade of energy in the inertial range is played by
341: the large scale components of the velocity field.
342:
343: However, when computing the energy flux through a shell $k$, \ie
344: integrating $T_2(K,Q)$ over all values of $Q$, and $K$ from $0$ to
345: $k$, these non-local interactions give $\sim 20\%$ of the total flux,
346: since many more local triads contribute in the global summation. We note
347: that this fraction (20\%) is independent of $k$, provided that $k$ is
348: large enough and in the inertial range.
349:
350: %Also, \ADD{if the $T_2$ transfer}
351: %is summed over octaves of wavenumbers, one finds an approximate
352: %equipartition between the local and non-local contributions to the
353: %energy flux.
354:
355: We are left therefore with two puzzles. First, why is the large scale
356: flow more effective (at the level of individual triadic interactions)
357: in ``destroying'' small size eddies than similar size eddies, when
358: phenomenological arguments in the Kolmogorov spirit suggest otherwise?
359: And secondly, why is the energy spectrum so close to $k^{-5/3}$ in the
360: constant flux region, when just advection by the large scale flow would
361: suggest a shallower spectrum $\sim k^{-1}$? (see \eg \cite{LDN01}). In
362: what follows, we give a brief review of possible answers as well as a
363: simple model that shows how a $k^{-5/3}$ energy spectrum can be obtained
364: by advection and stretching of the small scales just by the large scale
365: flow.
366:
367: A possible answer to explain the strong non-local triadic interactions
368: is that the Reynolds number in the present simulation is not high enough
369: to observe dominance of local triads, and the decrease in amplitude of
370: the small scale fields due to viscosity makes this interactions (when
371: compared to the large scale flow) smaller.
372:
373: Another possible answer would be that the wavenumber bands defining
374: the local interactions (\ie the range of values in $P$ used to define
375: $T_2^{Loc}$), that were arbitrarily taken here to have a width of $2^n$,
376: could be as wide as $10^n$ as some authors suggest \cite{BW94}. If this
377: is the case, a DNS with an inertial range that spans at least three
378: orders of magnitude in wavenumbers would be required to actually observe
379: strong local interactions.
380:
381: However, neither of these answers address the second question concerning
382: why a Kolmogorov energy spectrum is observed at moderate values of
383: the Reynolds number. If we look at phenomenological scaling arguments,
384: we see that there is one major assumption that may not be satisfied.
385: Current models assume that the energy is distributed in a hierarchy of
386: vortices of size $L,L/\alpha,L/\alpha^2,...$ (with $\alpha>1$), with no
387: specific geometry. However, experiments as well as numerical simulations
388: have shown that enstrophy is distributed in vortex tubes, where two
389: distinct length scales can be identified: one is the width of the tube
390: $l$ that is typically small and varies, and one is its length $L$,
391: typically of the order of the integral scale. It is not clear therefore
392: when two such structures interact, which length scale is responsible for
393: determining the time scale of the cascade.
394:
395: From the analysis presented here, a simple model for turbulent flows
396: consistent with several features observed in simulations and
397: experiments can emerge (see below). First, recall that Ref.
398: \cite{farge} found that helical vortex tubes capture 99\% of the
399: energy, give a $k^{-5/3}$ spectrum, and are responsible for the strong
400: wings in the PDF of velocity gradients. Furthermore, it was shown
401: in \cite{LDN01} that, when decomposing the velocity field in a
402: large scale component $U$ and a small scale one $u$,
403: artificially dropping local interactions in a simulation
404: (an operation akin to RDT) gives enhanced intermittency (in the
405: sense that a stronger departure from linear scaling of anomalous
406: exponents is observed), while when non-local interactions are dropped
407: the intermittency of the flow decreases \cite{note_MHD}.
408: %only keeping the uU interactions in the evolution equation for the
409: %small-scale velocity (an operation akin to Rapid Distortion Theory, or
410: %RDT), then intermittency develops, stronger than for turbulent
411: %flows in the sense that a stronger departure from linear scaling of
412: %anomalous exponents is observed
413:
414: The data analyzed in the present paper implies that, at low order of
415: correlators, \ie when considering the energy flux, the interactions
416: are mostly local. But when going to third-order individual triadic
417: interactions (such as with $T_3$), the non-local components are dominant
418: and involve the integral scale. We note that this is consistent with the
419: fact that departures from a linear scaling by anomalous exponents with
420: the order of structure function is stronger as the order is increased,
421: since it involves more non-local interactions linked to the geometrical
422: structure of vortex tubes. This leads to a model of small-scale
423: interactions involving three small scales that are substantially weakened
424: and gaussian, thus in agreement with the findings in \cite{LDN01} that
425: such $uu$-like terms weaken intermittency as well when included in the full dynamics.
426:
427: As a result, if we take into account the vortex tube structure of a
428: turbulent flow, the picture of the classical Richardson cascade may
429: change: a possible model to explain the aforementioned results is to
430: take the time scale of the cascade as given by the geometric average
431: of the length scales involved, based on the cubic root of the volume
432: of the vortex tube. If this is the case, the energy dissipation
433: rate of vortex tubes with velocity $u_l$ due to the large scale flow
434: $U_L$ is given by $\epsilon \sim u_l^2 \cdot U_L/(l^2L)^{1/3}$. This
435: implies that, for constant flux,
436: $u_l \sim l^{1/3} \sqrt{\epsilon L^{1/3}/U_L}$; this scaling recovers
437: the Kolmogorov spectrum, although in a different spirit \cite{note_beltrami}.
438: Note that the spirit of this derivation is close to multifractal models
439: used to explain intermittency corrections \cite{Frisch_book}.
440:
441: Finally, we would like to point out that the success of models involving as
442: an essential agent of nonlinear transfer the distortion of turbulent eddies
443: by a large-scale flow -- as in RDT and its variants \cite{LDN01} or as in
444: the alpha model \cite{holm} where the flow is interacting with a smooth
445: velocity field (see also \cite{montgo}) -- may be in part explained by the
446: present findings that confirm that nonlinear triadic interactions are mostly
447: nonlocal and involve the integral scale. Similar results have already been
448: obtained for flows coupled to a magnetic field \cite{Alexakis05}, where the
449: weakening of nonlinear interactions may occur in different fashions, \eg
450: Alfv\'enization or force-free fields, and where the second order transfer
451: $T_2$ between velocity and magnetic field in the induction equation is
452: itself non-local.
453:
454: %{\bf Acknowledgements}%\acknowledgments
455:
456: {\it NSF grant CMG--0327888 is gratefully acknowledged. Computer time was provided by NCAR.}
457:
458: \begin{thebibliography}{aa}
459:
460: \bibitem{kaneda}
461: Z.S. She and E. L\'ev\^eque, \prl {\bf 72}, 336 (1994); Kaneda, T. \etal,
462: %Ishihara, M. Yokokawa, K. Itakura, and A. Uno,
463: Phys. Fluids {\bf 15},
464: L21 (2003); K. Yoshida, T. Ishihara, and Y. Kaneda, Phys. Fluids {\bf 15},
465: 2385 (2003).
466:
467: \bibitem{LDN01}
468: J.-P. Laval, B. Dubrulle and S. Nazarenko, Phys. Fluids {\bf 13}, 1995 (2001);
469: B. Dubrulle, J.-P. Laval, S. Nazarenko, and O. Zaboronski, J. Fluid Mech.
470: {\bf 520}, 1 (2004).
471:
472: \bibitem{Zhou}
473: Y. Zhou, Phys. Fluids A {\bf 5}, 1092 (1993); {\bf 5}, 2511 (1993).
474:
475: \bibitem{AD90}
476: J.A. Domaradzki and R.S. Rogallo, Phys. Fluids {\bf A2}, 413 (1990);
477: K. Ohkitani and S. Kida, Phys. Fluids {\bf A4}, 794 (1992).
478:
479: \bibitem{BW94}
480: J.G. Brasseur and C-H. Wei, Phys. Fluids {\bf6}, 842 (1994); P.K. Yeung
481: J.G. Brasseur and Q. Wang, J. Fluid Mech. {\bf 283}, 43 (1995).
482:
483: \bibitem{Eyink} G.L. Eyink, {\it Physica D} {\bf 207}, 91 (2005).
484:
485: \bibitem{meb}
486: Brachet, M., C. R. Acad. Sci. Paris {\bf 311}, 775 (1990); Fluid Dyn.
487: Res. {\bf 8}, 1 (1991).
488:
489: \bibitem{GydroSpecialIssue}
490: ``MHD dynamo experiments'', special issue of {\it Magnetohydodynamics},
491: {\bf 38} (2002); Y. Ponty \etal, \prl {\bf 94}, 164502 (2005).
492:
493: \bibitem{Frisch_book}
494: U. Frisch, {\it Turbulence : the Legacy of A.N. Kolmogorov} (Cambridge,
495: Cambridge University Press, 1995).
496:
497: \bibitem{moffatt}
498: H.K. Moffatt, J. Fluid Mech. {\bf 150}, 359 (1985).
499:
500: \bibitem{moffatt_tsinober}
501: A. Tsinober and E. Levich, Phys. Lett. {\bf 99A}, 321 (1983); E. Levich,
502: Phys. Rep. {\bf 151}, 129 (1987).
503:
504: \bibitem{farge}
505: M. Farge, G. Pellegrino, and K. Schneider, \prl {\bf 87}, 054501 (2001).
506:
507: \bibitem{ESS}
508: R. Benzi \etal, Europhys. Lett. {\bf 24}, 275 (1993); R. Benzi \etal, Phys.
509: Rev. E {\bf 48}, R29 (1993).
510:
511: \bibitem{note1} With this definition, the force is acting on shell $K=3$.
512: Note also we are using a sharp filter in Fourier space, and as a result
513: the bounds presented in \cite{Eyink} may not apply.
514:
515: \bibitem{kraichnan}
516: R.H. Kraichnan, J. Fluid Mech. {\bf 47}, 525 (1971); M. Lesieur,
517: {\it Turbulence in fluids} (Dordrecht, Kluwer Acad. Press, 1997).
518:
519: \bibitem{Alexakis05}
520: M. Verma, Phys. Rep. {\bf 401} 229 (2004); A. Alexakis, P.D. Mininni,
521: and A. Pouquet, arXiv:physics/0505183 (2005); P.D. Mininni,
522: A. Alexakis, and A. Pouquet, arXiv:physics/0505189 (2005).
523:
524: \bibitem{note_MHD}
525: Note that in agreement with these arguments, MHD, which is more
526: non-local than neutral fluids \cite{Alexakis05} is also more intermittent
527: as shown \eg in H. Politano, A. Pouquet, and V. Carbone, Europhys. Lett.
528: {\bf 43} 516 (1998).
529:
530: \bibitem{note_beltrami}
531: Note that this is consistent with a Beltramization of the flow quenching
532: small-scale interactions. Using the Batchelor analogy between vorticity
533: and induction, and the properties of the transfer in conducting fluids
534: \cite{Alexakis05}, we conjecture that the vorticity transfer itself will
535: be non-local;
536: % \ie involving the large-scale flow
537: this allows for the existence of a non-Beltrami component of nonlinear
538: interactions which in turn could lead to the possible development of
539: singularities.
540:
541: \bibitem{holm}
542: S.Y. Chen \etal, Phys. Fluids {\bf 11}, 2343 (1999);
543: S.Y. Chen \etal, Physica D {\bf 133}, 66 (1999).
544:
545: \bibitem{montgo}
546: D. Montgomery and A. Pouquet, Phys. Fluids {\bf 14}, 3365 (2002).
547:
548: \end{thebibliography}
549: \end{document}
550: