1: \documentclass[aps,pre,groupedaddress,twocolumn,showpacs,floatfix]{revtex4}

2: \usepackage{epsfig} 

3: 

4: \begin{document}

5: 

6: \title{

7: Statistics of active vs. passive advections in magnetohydrodynamic

8: turbulence}   

9: \author{Thomas Gilbert}

10: \email[]{thomas.gilbert@inln.cnrs.fr}

11: \affiliation{Institut Non-Lin\'eaire de Nice, CNRS, Universit\'e de Nice,

12: 1361 Route des Lucioles, 06560 Valbonne, France}

13: \author{Dhrubaditya Mitra}

14: \email[]{dhruba@physics.iisc.ernet.in}

15: \affiliation{Centre for Condensed Matter Theory, Department of Physics,

16:               Indian Institute of Science, Bangalore 560 012, India}

17: \date{\today}

18: 

19: \begin{abstract}

20: Active turbulent advection is considered in the context of

21: magneto-hydrodynamics. In this case, an auxiliary passive field bears no

22: apparent connection to the active field. The scaling  

23: properties of the two fields are different. In the framework of a shell

24: model, we show that the two-point structure function of the passive field

25: has a unique zero mode, characterizing 

26: the scaling of this field only. In other words, the existence of

27: statistical invariants for the decaying passive field carries no information

28: on the scaling properties of the active field.

29: \end{abstract}

30: \pacs{47.27.-i,47.10.+g}

31: \maketitle

32: 

33: In the context of turbulent advection, the understanding of fluid turbulence 

34: has greatly improved in the recent years \cite{fgv01}. The anomalous scaling 

35: has been shown to be universal and connected to the existence of statistical

36: integrals of motion \cite{cv01}. In \cite{abcpv01}, it was shown that the 

37: statistically conserved structures of decaying passive turbulence 

38: dominate the statistics of forced turbulence, thus offering a rather general

39: framework for understanding the universality of anomalous scaling in forced

40: turbulence.

41: 

42: Let $\phi$ be a decaying field transported by a stationary turbulent flow. The

43: linearity of the advection implies the following relation for the correlation

44: functions~:

45: \begin{eqnarray}

46: \label{propcorf}

47: \lefteqn{\langle \phi(\vec{r}_1, t) \dots \phi(\vec{r}_N, t)\rangle =}&&\nonumber\\

48: &&\int d\underline{\vec{q}} \mathcal{P}^{(N)}(\underline{\vec{r}},t 

49: |\underline{\vec{q}},t_0) 

50: \langle \phi(\vec{q}_1, t_0) \dots \phi(\vec{q}_N, t_0)\rangle \ ,

51: \label{tevolcf}

52: \end{eqnarray}

53: where we used the compact notation $\underline{\vec{r}}\equiv

54: \vec{r}_1,\dots,\vec{r}_N$ to denote a collection of $N$ position vectors.

55: Equation (\ref{tevolcf}) tells us there exists a linear operator 

56: $\mathcal{P}^{(N)}$ that propagates the 

57: $n$th order correlation function from time $t_0$ to time $t$. Without fresh 

58: input, that is in the absence of forcing, the correlation functions of $\phi$

59: decay due to dissipative effects. Nevertheless, as conjectured in 

60: \cite{abcpv01}, there exist special functions $Z^{(N)}$ that are left 

61: eigenfunctions of eigenvalue $1$ of the operator $\mathcal{P}^{(N)}$,

62: \begin{equation}

63: Z^{(N)}(\underline{\vec{r}}) = \int

64: d\underline{\vec{q}}\mathcal{P}^{(N)}(\underline{\vec{q}},t 

65: |\underline{\vec{r}},t_0) Z^{(N)}(\underline{\vec{q}}),

66: \label{leftev}

67: \end{equation}

68: such that

69: \begin{equation}

70: I^{(N)}(t) = \int d\underline{\vec{r}}

71: Z^{(N)}(\underline{\vec{r}}) \langle \phi(\vec{q}_1,t) \dots 

72: \phi(\vec{q}_N,t)\rangle

73: \label{prestr}

74: \end{equation}

75: is preserved in time. $I^{(N)}$  and $Z^{(N)}$ are respectively called 

76: a statistical integral of motion and a statistically

77: preserved structure of order $N$, also referred to as zero modes

78: \footnote{Note that Eq. (\ref{leftev}) is not stricto sensu an eigenvalue

79:   problem; $\mathcal{P}^{(N)}$ is not defined on a compact space. Only in the

80:   case of a scaling advecting field, is it possible to write down a proper

81:   operator and identify the zero modes as left eigenmodes with unit

82:   eigenvalues \cite{bgk98}.}. 

83: 

84: Now, consider the same passive advection problem with an external forcing,

85: such that the system reaches a

86: stationary state. Define,

87: the correlation function of $\phi$ in that stationary state to be,

88: \begin{equation}

89: F^{(N)}(\underline{\vec{r}}) = 

90: \langle \phi(\vec{r}_1, t) \dots \phi(\vec{r}_N, t)\rangle_f,

91: \end{equation}

92:  where the symbol $ \langle \cdot \rangle_f$ denotes averaging over the

93: statistical stationary state.   

94: It was conjectured in Ref~\cite{abcpv01}, that the anomalous part of 

95: $F^{(N)}(\underline{\vec{r}})$  is dominated by the leading 

96: zero modes of the decaying problem, 

97: i.~e. $Z^{(N)} \sim F^{(N)}$. 

98: The conjecture was

99: verified in the context of a shell model for passive scalar advection.

100: 

101: In subsequent studies, it was discovered that the existence of statistical 

102: invariants of the motion for passive turbulence may help understand the 

103: statistics of active turbulence, a case where the advected quantity affects 

104: the dynamics of the advecting field. 

105: In \cite{cmmv02} and \cite{ccgp02,ccgp03}, the case of 

106: thermal convection in the Boussinesq approximation was studied. There it

107: was shown 

108: that the scaling of the active field is also dominated by the statistically 

109: preserved structures of auxiliary passive fields. It is yet unclear how 

110: general this connection between the statistics of active and auxiliary 

111: passive fields is. The case of 2-dimensional magneto-hydrodynamics is 

112: revealing. Indeed, in this case active and passive fields have very different 

113: scaling behaviors. As shown in \cite{ccmv02}, the transported fields

114: cascade in different directions. It is therefore unexpected  

115: that the statistics of the auxiliary passive field holds information on the 

116: statistics of the active field. The two fields have different scaling 

117: properties.  In \cite{ccgp03}, despite this difference, the claim was made 

118: that the analogy does hold in the sense that there exist sub-leading 

119: zero modes of the propagator of the correlation functions of the auxiliary 

120: passive field with the scaling of the correlation functions of the active 

121: field.

122: 

123: The purpose of this note is to show that this is actually not the case. To

124: this end, we will limit our investigation to the case of the second

125: order structure  

126: functions of a shell model of 2-dimensional magneto-hydrodynamic turbulence. 

127: The case of the second order structure function is the simplest

128: one. Because of the absence of geometry, its scaling is

129: non-anomalous \cite{cv01,gz98,bgk98}, which in the language of zero

130: modes implies non-degeneracy, that is there is a unique conserved structure

131: associated to the two-point statistical invariant. The same holds in the

132: language of shell models, where the only two point function of a passive

133: scalar field $\theta_n$ associated to the scale $k_n$ is

134: $\langle|\theta_n|^2\rangle$, in contrast to higher order structure

135: functions, e.~g. the fourth order for which we have contributions from 

136: $\langle|\theta_n|^4\rangle$,

137: $\langle|\theta_{n-1}|^2|\theta_{n+1}|^2\rangle$, etc. 

138: Using methods similar to those used in \cite{cgp02}, we will construct the 

139: operator propagating the second order structure functions, and will demonstrate

140: that the auxiliary passive field has the same statistical integral of motion 

141: as other shell models of passive advection. It will be inferred that there

142: is no sub-leading zero-mode with the scaling of the (active) magnetic field.

143:   

144: 

145: In analogy to other models \cite{basu98,frick98,g99}, the following two

146: sets of equations  

147: generalize the usual Sabra shell model \cite{lpppv98} for the turbulent 

148: velocity field to magneto-hydrodynamic turbulence

149: (we omit dissipative terms)~:

150: \begin{eqnarray}

151: \frac{d u_n}{dt} &=& i [k_{n+1} (u_{n+1}^*u_{n+2}-b_{n+1}^*b_{n+2})

152: \nonumber\\&&

153: - (\epsilon +1) k_n (u_{n-1}^*u_{n+1} - b_{n-1}^*b_{n+1})\label{mhdsabrau}

154: \\

155: &&- \epsilon k_{n-1}(u_{n-2}u_{n-1} - b_{n-2}b_{n-1})] + f_n ,\nonumber

156: \\

157: \frac{d b_n}{dt}&=& i [-(\epsilon+\delta) k_{n+1}

158: (u_{n+1}^*b_{n+2}-b_{n+1}^*u_{n+2}) \nonumber\\&&

159: + \delta k_n (u_{n-1}^*b_{n+1} - b_{n-1}^*u_{n+1})\label{mhdsabrab}

160: \\

161: &&+ (\delta -1) k_{n-1}(u_{n-2}b_{n-1} - b_{n-2}u_{n-1})] + f'_n .

162: \nonumber

163: \end{eqnarray}

164: Here $u_n$ (the velocity field) and $b_n$ (the magnetic field) are complex

165: variables defined on a discrete set of shells indexed by the integer $n$ whose

166: associated wave-number $k_n = k_0 \lambda^n$, $\lambda>1$ (hereafter taken to 

167: be  $\lambda=2$). $f_n$ and $f'_n$ are two forcing terms which are taken to be

168: stochastic white noises with identical statistics and concentrated on a limited

169: number of neighboring shells ($n=5,6,7$ in our numerical experiments). The

170: model's parameters $\epsilon$ and $\delta$ are conveniently parametrized 

171: in the following way. The three-dimensional model for which $\epsilon < 0$ 

172: reads ($\alpha>0$)

173: \begin{equation}

174: \begin{array}{l}

175: \epsilon = - \lambda^{-\alpha}\ ,\\

176: \delta = (1 + \lambda^{\alpha})^{-1}\ .

177: \end{array}

178: \quad

179: \mbox{(3D model)}

180: \label{3dparam}

181: \end{equation}

182: The two-dimensional problem on the other hand has $\epsilon>0$ and reads

183: \begin{equation}

184: \begin{array}{l}

185: \epsilon = \lambda^{-\alpha}\ ,\\

186: \delta = -(\lambda^{\alpha}-1)^{-1}\ .

187: \end{array}

188: \quad

189: \mbox{(2D model)}

190: \label{2dparam}

191: \end{equation}

192: Correspondingly, we have the following quadratic dynamical invariants

193: (i.~e. time-invariant in the limit of zero viscosity and zero external

194: forcing).  

195: \begin{eqnarray}

196: E &=& \sum_n (|u_n|^2 + |b_n|^2)\quad\mbox{(total energy)}, \label{toten}\\

197: K &=& \sum_n \Re(u_n^* b_n)\quad\mbox{(cross helicity)},\label{crosshe}\\

198: H &=& \sum_n \mathrm{sign}(\delta)^n k_n^{-\alpha} |b_n|^2 

199: \quad\mbox{(magnetic helicity)}.

200: \label{magnhe}

201: \end{eqnarray}

202: Thus, in both two- and three-dimensional models, the equations 

203: (\ref{mhdsabrau}, \ref{mhdsabrab}) have one single free parameter, 

204: $\alpha>0$.  The two- and three-dimensional models actually have very 

205: different dynamical behaviors, see \cite{g99}. 

206: Only the two-dimensional model can sustain a stationary state and we will

207: limit ourselves to this case.

208: As dimensional analysis shows \cite{ccgp03}, the conservation of the first two

209: invariants implies that $u_n$ and $b_n$ must both have Kolmogorov scalings,

210: $\langle |u_n|^2\rangle, \langle |b_n|^2\rangle \sim k_n^{-2/3}$,

211: for which the corresponding fluxes are constant. This 

212: is indeed what has been measured for similar models \cite{g99}, where both 

213: fields appear to display the same anomalies. Further, as shown in 

214: \cite{ccgp03}, the conservation of the third invariant allows for another 

215: scaling, $\langle |b_n|^2\rangle \sim k_n^{\alpha-2/3}$, for which the 

216: magnetic helicity flux is constant. But since this scaling is incompatible 

217: with the conservation of the two other invariants, it is not relevant to the 

218: statistics of the magnetic field.

219: 

220: However if one considers a passive auxiliary field obeying an equation 

221: identical to Eq. (\ref{mhdsabrab}) for the 2-dimensional case,

222: \begin{eqnarray}

223: \frac{d a_n}{dt}&=& i [\lambda^{-\alpha}(\lambda^\alpha-1)^{-1} k_{n+1}

224: (u_{n+1}^*a_{n+2}-a_{n+1}^*u_{n+2}) \nonumber\\

225: &&- (\lambda^\alpha-1)^{-1} k_n (u_{n-1}^*a_{n+1} - a_{n-1}^*u_{n+1})\nonumber\\

226: && - \lambda^\alpha(\lambda^\alpha-1)^{-1} k_{n-1}(u_{n-2}a_{n-1} 

227: - a_{n-2}u_{n-1})]\ ,\nonumber\\

228: \label{mhdsabraa}

229: \end{eqnarray}

230: the only relevant invariant is the equivalent of the magnetic helicity 

231: Eq. (\ref{magnhe}), to which a constant flux is associated. Thus the

232: dimensional scaling $\langle|a_n|^2\rangle \sim k_n^{\alpha-2/3}$

233: is expected to be observed. And that is indeed what was found in 

234: \cite{ccgp03}. 

235: 

236: Notice though that the linearity of Eq. (\ref{mhdsabraa}) allows for the

237: substitution $\psi_n = \lambda^{-\alpha n/2} a_n$. 

238: With this new variable, Eq. (\ref{mhdsabraa}) takes the form,

239: \begin{eqnarray}

240: \frac{d\psi_n}{dt} &=& i[A(k_{+1} u^*_{n+1}\psi_{n+2} +

241: k_{n-1}\psi_{n-2}u_{n-1}) \nonumber\\

242: &&+ B(k_{n+1}\psi^*_{n+1}u_{n+2} - 

243: k_n \psi^*_{n-1}u_{n+1}) \nonumber\\

244: &&+ C(k_n u^*_{n-1}\psi_{n+1} +

245: k_{n-1}u_{n-2}\psi_{n-1})],

246: \label{passcalar}

247: \end{eqnarray}

248: which describes the advection of a scalar for which the

249: quadratic invariant is $\sum_n |\psi_n|^2$, which is similar to the 

250: shell model with only nearest neighbor interaction considered in 

251: \cite{abcpv01,cgp02}.  The coefficients

252: in Eq. (\ref{mhdsabraa}) correspond to the choice

253: $A = (\lambda^\alpha-1)^{-1}$,

254: $B = - \lambda^{-\alpha/2}(\lambda^\alpha-1)^{-1}$, and

255: $C = - \lambda^{\alpha/2}(\lambda^\alpha-1)^{-1}$.

256: This shows that as far as shell models are concerned, the difference

257: between a passively advected vector and passively advected scalar

258: is just a numerical factor. Henceforth

259: we do not make a distinction between the two and use the name "passive

260: field" for both.   

261: 

262: Consider now the equivalent of Eq. (\ref{propcorf}) for the propagation 

263: of the second order structure functions $\langle |\theta_n|^2 \rangle$ in the 

264: decaying problem -- including dissipative terms on the RHS of Eq. 

265: (\ref{passcalar}).

266: Following notations similar to those used in \cite{cgp02}, we can write the 

267: equation of motion for $\psi_n$ under the form

268: \begin{equation}

269: \frac{d \psi_n}{dt} = \mathcal{L}_{n,m} \psi_m\ ,

270: \end{equation}

271: with the solution 

272: \begin{eqnarray}

273: \psi_n(t) &=& \mathsf{T}^+ 

274: \left\{\exp\left[\int_{t_0}^t ds \mathcal{L}(s)\right]\right\}_{n,m}

275: \psi_m(t_0), \nonumber\\

276: &\equiv& \mathcal{R}_{n,m}(t|t_0) \psi_m(t_0),

277: \end{eqnarray}

278: ($\mathsf{T}^+$ denotes the time ordering operator). Letting

279: \begin{equation}

280: \mathcal{P}^{(2)}_{n,m}(t|t_0) \equiv \langle 

281: \mathcal{R}_{n,m}(t|t_0)\mathcal{R}^*_{n,m}(t|t_0) \rangle\ ,

282: \end{equation}

283: the propagation of second order structure functions obeys the following 

284: equation~:

285: \begin{equation}

286: \langle|\psi_n(t)|^2\rangle = \sum_m \mathcal{P}^{(2)}_{n,m}(t|t_0)

287: \langle|\psi_m(t_0)|^2\rangle\ .

288: \label{propa2}

289: \end{equation}

290: The form of the operator $\mathcal{P}^{(2)}$ was discussed in \cite{cgp02}. 

291: It is a matrix whose elements can be obtained by propagating an initial 

292: condition concentrated at a given shell. Similar considerations hold for the 

293: models considered here. In Fig. \ref{fig.P2} we show these elements for 

294: successive times, starting from an initial conditions at shell $20$. The model

295: we used is Eq. (\ref{passcalar}) for the parameters corresponding to Eq. 

296: (\ref{mhdsabraa}) and advected by the magneto-hydrodynamic fields Eqs. 

297: (\ref{mhdsabrau}, \ref{mhdsabrab}). The parameters of the simulation are given 

298: in the figure caption.

299: \begin{figure}

300: \centerline{\psfig{figure=fig.P2.eps,width=.7 \hsize,angle=-90}}

301: \caption{The elements of $\mathcal{P}^{(2)}_{n,20}$ for the scalar field

302: Eq. (\ref{passcalar}) advected by the MHD field Eqs. (\ref{mhdsabrau}, 

303: \ref{mhdsabrab}), where the parameter was chosen to be $\alpha=2$. 

304: The times displayed are $.4$, $.55$, $.7$, $.85$ and $1$ respectively

305: (measured in the natural time units of the model). The horizontal axis

306: corresponds to the shell numbers. The units on the vertical scale are

307: arbitrary. 

308: The simulation was done using a total of $35$ shells, with

309: the first shell wave-number $k_0 = 1/16$ and a shell spacing of $\lambda=2$.

310: All the fields were dissipated on the small scales with a term $\nu k_n^2$,

311: with $\nu = 10^{-12}$. The advecting fields were forced on shells $5-7$ with 

312: white delta correlated noise of amplitudes $1/\sqrt{2}$, $1/2\sqrt{2}$ and

313: $1/4$ respectively. Moreover the phase of the forcing on shell $7$ was taken 

314: to be equal to the sum of the phases of the forcings on shells $5$ and $6$.}

315: \label{fig.P2}

316: \end{figure}

317: 

318: In analogy to \cite{abcpv01}, the statistical invariant $I^{(2)}$ for 

319: the passive scalar field is 

320: \begin{equation}

321: I^{(2)}(t) = \sum_n Z_n^{(2)} \langle |\psi_n (t)|^2 \rangle

322: \label{I2}

323: \end{equation}

324: where $Z_n^{(2)}$ is a left-eigenfunctions of the operator 

325: $\mathcal{P}^{(2)}$, with the scaling of the second order structure 

326: function of the forced problem. The invariance of $I^{(2)}$ is most 

327: easily demonstrated by re-scaling the decaying second order objects 

328: according to Eq. (\ref{I2}). The curves indeed collapse 

329: if the ordinate is shifted with the appropriate time dependence. This is shown

330: in Fig. \ref{fig.coll} and is analogous to Fig. 4 in \cite{cgp02}.

331: \begin{figure}

332: \centerline{\psfig{figure=fig.coll.eps,width=.7 \hsize,angle=-90}}

333: \caption{The curves of Fig. \ref{fig.P2} collapsed according to Eq. 

334: (\ref{I2}).}

335: \label{fig.coll}

336: \end{figure}

337: Thus in the language of the passive magneto-hydrodynamic model Eq. 

338: (\ref{mhdsabraa}), we have that 

339: \begin{equation}

340: \sum_n \left\langle |a_n|^2\right\rangle_f/k_n^\alpha

341: \left\langle|a_n|^2(t)\right\rangle/k_n^\alpha

342: \label{mhdI2}

343: \end{equation}

344: is a statistical integral of motion. 

345: To claim that the scaling of the magnetic field structure function 

346: $\langle |b_n|^2\rangle$ is a sub-leading zero mode of $\mathcal{P}^{(2)}$

347: is equivalent to claiming that

348: \begin{equation}

349: \sum_n \left\langle |b_n|^2\right\rangle/k_n^\alpha

350: \left\langle |a_n|^2(t)\right\rangle/k_n^\alpha

351: \label{mhdnoI2}

352: \end{equation}

353: is a statistical integral of motion. Clearly, in view of Fig. \ref{fig.coll} 

354: this cannot be the case and the collapse will not occur should the scaling 

355: exponent of $Z^{(2)}$ be replaced by another one

356: \footnote{Our claim is that there is only one unique scaling exponent $\beta_2$

357: (which is the scaling exponent of $Z^{(2)}$) which will make 

358: the quantity $\sum_n k_n^{\beta_2}/k_n^\alpha

359: \left\langle |a_n|^2(t)\right\rangle/k_n^\alpha $ a statistical integral of

360: motion. This can be proved if the velocity driving the passive field

361: is Kraichnan (i.~e. Gaussian delta-correlated in time), but for a generic

362: velocity field we know no way to  

363: rigorously prove this. One can however argue that the statistical integral

364: of motion should depend on the geometry of the correlation function and not 

365: on the details of the statistics of the advecting field.}.

366: In this line of thought, 

367: it is perhaps worthwhile pointing out that the collapse as seen in 

368: Fig. \ref{fig.coll} would not be possible should there be zero-modes with 

369: distinct scaling exponents. Indeed, as seen from Eq. (\ref{I2}), the collapse 

370: occurs provided $\langle |\psi_n(t)|^2\rangle$ ``falls'' precisely on the 

371: right eigenmode with a scaling identical to $Z_n^{(2)}$.

372: 

373: To conclude we emphasize that the linearity of the passive advection models 

374: assigns them to a narrow class of equivalence. The 

375: passive magneto-hydrodynamic model is in fact equivalent to a scalar advection 

376: model for which the statistical invariants have already been investigated in 

377: some detail \cite{cgp02}. In magneto-hydrodynamics, active and passive fields

378: have different scaling properties. The arguments that were used in the 

379: framework of thermal convection to account for the anomalous scaling of the

380: active field in terms of a passive auxiliary one do not carry over to 

381: magneto-hydrodynamics. The claim that one can nevertheless account for the 

382: scaling of the (active) magnetic field by considering sub-leading 

383: zero-modes of the operators propagating the decaying correlation functions was

384: proven wrong. In view of the form of the second order propagator, it is clear

385: that there are no zero mode but the one whose scaling is that of the

386: passive field. 

387: %Finally let us note that our work applies of course only in the 

388: %context of $2$nd order propagator.  Higher order propagators indeed

389: %possess more than one zero mode.

390: 

391: 

392: \acknowledgments

393: The authors gratefully acknowledge discussions with A. Celani, U. Frisch

394: and R. Pandit. T.~G. also wishes to thank Y. Cohen, I. Procaccia and

395: A. Pumir. D.~M. thanks D. Vincenzi. This work was partially done while 

396: D.~M. was visiting the Observatoire de la C\^ote d'Azur in Nice. 

397: The support of CEFIPRA under project number 2404-2 is acknowledged.  

398: T.~G. acknowledges financial support from the European Union under contract 

399: numbers HPRN-CT-2000-00162 and HPRN-CT-2002-00300.  D.~M. also wishes to

400: thank CSIR India for financial support.

401: 

402: \begin{thebibliography}{99}

403: 

404: \bibitem{fgv01} G. Falkovich, K. Gawedzki, and M. Vergassola,

405: {\em Particles and fields in fluid turbulence}, Rev. Mod. Phys. {\bf 73}, 

406: 913 (2001).

407: 

408: \bibitem{cv01} A. Celani and M. Vergassola, 

409: {\em Statistical Geometry in Scalar Turbulence}, 

410: Phys. Rev. Lett. {\bf 86}, 424 (2001).

411: 

412: \bibitem{abcpv01} I. Arad, L. Biferale, A. Celani, I. Procaccia, and M. 

413: Vergassola, {\em Statistical Conservation Laws in Turbulent Transport},

414: Phys. Rev. Lett. {\bf 87}, 164502 (2001).

415: 

416: \bibitem{cmmv02} A. Celani, T. Matsumoto, A. Mazzino, and M. Vergassola,

417: {\em Scaling and Universality in Turbulent Convection}, 

418: Phys. Rev. Lett. {\bf 88}, 054503 (2002).

419: 

420: \bibitem{ccgp02} E. S. C. Ching, Y. Cohen, T. Gilbert, and I. Procaccia, 

421: {\em Statistically Preserved Structures and Anomalous

422: Scaling in Turbulent Active Scalar Advection},

423: Europhys. Lett. {\bf 60}, 369 (2002).

424: 

425: \bibitem{ccgp03} E. S. C. Ching, Y. Cohen, T. Gilbert, and I. Procaccia, 

426: {\em Active and passive fields in turbulent transport: the role

427: of statistically preserved structures}, 

428: Phys. Rev. E {\bf 67}, 016304 (2002). 

429: 

430: \bibitem{ccmv02} A. Celani, M. Cencini, A. Mazzino, and M. Vergassola

431: {\em Active versus Passive Scalar Turbulence},

432: Phys. Rev. Lett. {\bf 89}, 234502 (2002)

433: 

434: \bibitem{gz98} O. Gat and R. Zeitak, {\em Multiscaling in passive scalar

435:   advection as stochastic shape dynamics}, Phys. Rev. E {\bf 57}, 5511 (1998).

436: 

437: \bibitem{bgk98} D. Bernard, K. Gawedzki, and A. Kupiainen, {\em Slow modes

438:     in passive scalar advection}, J. Stat. Phys. {\bf 90}, 519 (1998).

439: 

440: \bibitem{cgp02}  Y. Cohen, T. Gilbert, and I. Procaccia, 

441: {\em Statistically preserved structures in shell models of passive scalar

442: advection}, Phys. Rev. E {\bf 65}, 026314 (2002).

443: 

444: \bibitem{g99} P. Giuliani, {\em Shell models for magnetohydrodynamic 

445: Turbulence}, in {\em Nonlinear MHD Waves and Turbulence}, Eds. T. Passot 

446: and P.~L. Sulem, Lecture Notes in Physics, Springer, (1999). 

447: 

448: \bibitem{basu98} A. Basu, A. Sain, S.~K. Dhar, and R. Pandit

449: {\it Phys. Rev. Lett.} {\bf 81}, 2687 (1998).  

450: 

451: \bibitem{frick98} P. Frick and D. Sokoloff {\it Phys. Rev. E.} {\bf 57},

452: 4155 (1998)

453: 

454: \bibitem{lpppv98} V. S. L'vov, E. Podivilov, A. Pomyalov, I. Procaccia, 

455: and D. Vandembroucq, {\em Improved shell model of turbulence},

456: Phys. Rev. E {\bf 58}, 1811 (1998).

457: 

458: \end{thebibliography}

459: 

460: \end{document}

461: