1: \documentclass{jpp}

2: \usepackage{subeqn}

3: \usepackage{upmath}

4: \usepackage{amssymb}

5: \usepackage{amsbsy}

6: 

7: \usepackage{graphicx}% Include figure files

8: \usepackage{color}   % Color

9: 

10: % BIBLIOGRAPHY %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

11: \usepackage{natbib}

12: \bibliographystyle{jpp}

13: \def\newblock{\hskip .11em plus .33em minus .07em} 

14: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

15: 

16: % COLOR %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

17: \def\note#1{{\textcolor{red}{\bf [#1]}}}  % note

18: \def\opt#1{{\textcolor{green}{#1}}}       % optional

19: \def\add#1{{\textcolor{blue}{#1}}}        % addition

20: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

21: 

22: \title[Energy transfer in Hall-MHD turbulence]

23:       {Energy transfer in Hall-MHD turbulence: \\ 

24:        cascades, backscatter, and dynamo action}

25: \author[P.\,D. Mininni, A. Alexakis and A. Pouquet]%

26:        {P\ls A\ls B\ls L\ls O\ns D.\ns M\ls I\ls N\ls I\ls N\ls N\ls I%

27:        \thanks{e-mail: mininni@ucar.edu},\\

28:        A\ls L\ls E\ls X\ls A\ls N\ls D\ls R\ls O\ls S\ns 

29:        A\ls L\ls E\ls X\ls A\ls K\ls I\ls S%

30:        \thanks{e-mail: alexakis@ucar.edu}\ns \and \ns 

31:        A\ls N\ls N\ls I\ls C\ls K\ns P\ls O\ls U\ls Q\ls U\ls E\ls T%

32:        \thanks{e-mail: pouquet@ucar.edu}}

33: \affiliation{National Center for Atmospheric Research, 

34:              P.O. Box 3000, Boulder, Colorado 80307, USA}

35: 

36: \pubyear{2005}

37: \volume{0}

38: \part{0}

39: \pagerange{1--25}

40: \date{\today}

41: \setcounter{page}{1}

42: 

43: \begin{document}

44: 

45: \maketitle

46: 

47: \begin{abstract}

48: Scale interactions in Hall MHD are studied using both the mean field theory 

49: derivation of transport coefficients, and direct numerical simulations in 

50: three space dimensions. In the magnetically dominated regime, the eddy 

51: resistivity is found to be negative definite, leading to large scale 

52: instabilities. A direct cascade of the total energy is observed, although 

53: as the amplitude of the Hall effect is increased, backscatter of magnetic 

54: energy to large scales is found, a feature not present in MHD flows. 

55: The coupling between the magnetic and velocity fields is different 

56: than in the MHD case, and backscatter of energy from small scale magnetic 

57: fields to large scale flows is also observed. For the magnetic helicity, a 

58: strong quenching of its transfer is found. We also discuss non-helical 

59: magnetically forced Hall-MHD simulations where growth of a large scale 

60: magnetic field is observed. 

61: \end{abstract}

62: 

63: \section{ \label{Intro} Introduction }

64: 

65: The relevance of two fluid effects has recently been pointed out in 

66: several studies of astrophysical and laboratory plasmas 

67: \citep{Balbus01,Sano02,Mirnov03,Ding04}. The effect of adding the Hall 

68: current to the dynamics of the flow was studied in several scenarios, 

69: particularly dynamo action 

70: \citep{Helmis68,Galanti95,Mininni02,Mininni03b,Mininni05a} and 

71: reconnection \citep{Birn01,Shay01,Wang01,Morales05}. Several of these 

72: works showed that the Hall currents increase the reconnection rate 

73: of magnetic field lines. However, most of the studies of magnetic 

74: reconnection were done for particular configurations of current sheets. 

75: It was shown in particular by \citet{Smith04} that when a turbulent 

76: background is present the reconnection rate is dominated by the 

77: amplitude of the turbulent fluctuations. The process of 

78: magnetic reconnection is relevant in several astrophysical and 

79: geophysical scenarios, such as the magnetopause, the magnetotail, the 

80: solar atmosphere, or the interplanetary and interstellar medium. 

81: Reconnection can also play a role in the generation of large 

82: scale magnetic fields by dynamo action \cite{Zeldovich}.

83: 

84: Some of the works in Hall-magnetohydrodynamics (Hall-MHD) present 

85: conflicting results, indicating in some cases that the Hall effect can 

86: help the growth of a large scale magnetic field \citep{Mininni05a} or a 

87: large scale self-organization process \citep{Mahajan98,Numata04,Ohsaki05}, 

88: while in other cases the Hall currents were observed to generate small 

89: scales and filamentation \citep{Laveder02a,Laveder02b,Rheinhardt02}. 

90: 

91: As a result, it becomes of interest to study the physical processes 

92: leading to cascades and transfer of ideal invariants in three-dimensional 

93: Hall-MHD turbulence. Phenomena observed in the laboratory and space 

94: plasmas tend to show an intermittent or impulsive behavior 

95: \citep{Bhattacharjee99} characteristic of turbulent flows. The relevance 

96: of Hall-MHD turbulence in the solar wind was shown by \citet{Ghosh96}. 

97: Also, Hall-MHD turbulence can play a crucial role in the transfer of 

98: matter in the magnetopause as was pointed by \citet{Rezeau01}.

99: 

100: In this work, we study both analytically and numerically three dimensional 

101: Hall-MHD turbulence as the result of a dynamo process, and from a purely 

102: electromotive forcing. Detailed studies of shell-to-shell energy transfer 

103: from direct numerical simulations (DNS) have been done for hydrodynamic 

104: \citep{Domaradzki90,Ohkitani92,Zhou93,Yeung95,Alexakis05b}

105: and magnetohydrodynamic flows \citep{Debliquy05,Alexakis05a,Mininni05b}. 

106: To the best of our knowledge, the energy transfer in Hall-MHD 

107: turbulence has not been studied before.

108: 

109: We show evidence of non-locality of the transfer in Fourier space, 

110: and that the Hall effect can increase both the transfer of magnetic 

111: energy to smaller scales (locally), as well as give a novel non-local 

112: backscatter of magnetic energy to large scales. These results become 

113: clear when examining the modification to the turbulent magnetic 

114: diffusivity due to the Hall term. Also, we observe that the Hall 

115: currents impact on the coupling between the magnetic and velocity fields. 

116: The transfer of energy between these two fields is different than in the 

117: MHD case. The Hall-MHD equations also display a backscatter of energy 

118: from small scale magnetic fluctuations to the large scale flows themselves. 

119: The transfer of helicity is briefly discussed as well and observed to be 

120: quenched by the Hall effect.

121: 

122: The structure of the paper is as follows. In Sec. \ref{Equations} we 

123: introduce the Hall-MHD equations and we define the various transfer 

124: terms. In Sec. \ref{Transport} we derive turbulent transport 

125: coefficients for the Hall-MHD induction equation. In Sec. 

126: \ref{Simulations} we briefly discuss the code and details of the 

127: mechanically forced numerical simulations for completeness. 

128: Section \ref{Transfers} presents the 

129: transfer terms in Hall-MHD as obtained from the numerical simulations. 

130: Section \ref{Upscaling} shows backscatter of magnetic energy in 

131: non-helical Hall-MHD magnetically forced simulations. Finally, Sec. 

132: \ref{Discussion} summarizes the results and discusses implications of 

133: our work for the understanding of turbulence, dynamo action, and 

134: reconnection in Hall-MHD.

135: 

136: \section{ \label{Equations} The Hall-MHD equations and transfer terms }

137: 

138: In dimensionless Alfv\'enic units, the Hall-MHD equations are

139: \begin{equation}

140: \partial_t {\bf U} + {\bf U}\cdot \nabla {\bf U} = - \nabla {\cal P} + 

141:     {\bf B}\cdot \nabla {\bf B} + \nu \nabla^2 {\bf U} +{\bf f} ,

142: \label{eq:momentum}

143: \end{equation}

144: %

145: \begin{equation}

146: \partial_t {\bf B} = \nabla \times \left [ \left( {\bf U} - \epsilon 

147:     {\bf J} \right) \times {\bf B} \right] + \eta \nabla^2 {\bf B} ,

148: \label{eq:induction1}

149: \end{equation}

150: %

151: where ${\bf U}$ is the bulk velocity field, ${\bf B}$ is the magnetic 

152: field, ${\bf J} = \nabla \times {\bf B}$ is the current density, 

153: ${\cal P}$ is the pressure, $\nu$ is the kinematic viscosity, and 

154: $\eta$ is the magnetic diffusivity. From the Maxwell equations and 

155: incompressibility of the flow, 

156: $\nabla \cdot {\bf U} = \nabla \cdot {\bf B} = 0$. 

157: 

158: The Hall term $\epsilon {\bf J} \times {\bf B}$ in Eq. (\ref{eq:induction1}) 

159: measures the velocity difference between species, where the electron 

160: velocity is ${\bf U}^e = {\bf U} - \epsilon {\bf J}$. Here, $ \epsilon $ 

161: measures the relative strength of the Hall effect, with the Hall term being 

162: dominant for wavenumbers larger than $k_{Hall} \sim 1/\epsilon$ if 

163: equipartition between the fields is assumed. The measure of strength 

164: of the Hall effect can be written as $\epsilon = L_{Hall}/L_0$ where 

165: $L_0$ is a characteristic length (we will use $L_0=2\pi$, the size of 

166: the box in our simulations). In terms of physical parameters, and for 

167: a fully ionized plasma, the Hall length is $L_{Hall}=cU_A/(\omega_{pi}U_0)$, 

168: where $U_A$ is the Alfv\'enic speed, $U_0$ is a characteristic speed, 

169: $c$ is the speed of light, and $\omega_{pi}$ is the ion plasma 

170: frequency (when $U_0=U_A$, $L_{Hall}$ reduces to the ion skin depth). 

171: In a partially ionized plasma, expressions for $L_{Hall}$ can be found 

172: in \citet{Sano02} and \citet{Mininni03b}.

173: 

174: Of special interest is the ratio between the integral length $L$, the 

175: Hall length $L_{Hall}$, and the Ohmic dissipation length $L_\eta$. For 

176: $L_{Hall} \ll L_\eta$ ($\epsilon \to 0$), the Hall-MHD equations reduce 

177: to the well known MHD case. In several astrophysical problems, such as 

178: accretion disks, protoplanetary disks, or the magnetopause 

179: \citep[see e.g.][]{Birn01,Balbus01,Sano02}, the Hall scale is larger than 

180: Ohmic scales although smaller than the integral scale $L$ of the flow. 

181: We will be interested in this regime in this work, although we remark 

182: that the separation between these scales in astrophysical or geophysical 

183: problems is far from what can be achieved in numerical simulations.

184: 

185: The Hall-MHD equations have three ideal invariants \citep{Turner86}. In 

186: this work we will focus on two invariants, the total energy

187: \begin{equation}

188: E = \frac{1}{2} \int (U^2 + B^2) \, d{\bf x}^3 ,

189: \end{equation}

190: and the magnetic helicity

191: \begin{equation}

192: H = \frac{1}{2} \int {\bf A} \cdot {\bf B} \, d{\bf x}^3 ,

193: \end{equation}

194: where {\bf A} is the vector potential, $\nabla \times {\bf A} =  {\bf B}$. 

195: These quantities are also ideal invariants of the MHD equations 

196: ($\epsilon=0$). The third MHD invariant, the cross helicity, is replaced 

197: in Hall-MHD by the hybrid helicity \citep{Turner86} 

198: and is small in the simulations we will discuss.

199: 

200: The expressions we will use for the shell-to-shell energy transfers have

201: been derived for the MHD case by \citet{Verma04,Debliquy05}; and 

202: \citet{Alexakis05a}. Here we present the derivation of 

203: the transfer terms for the Hall-MHD equations. Equation 

204: (\ref{eq:induction1}) can be rewritten as

205: \begin{equation}

206: \partial_t {\bf B} + {\bf U}\cdot \nabla {\bf B} = {\bf B}\cdot \nabla 

207:     {\bf U} - \epsilon \nabla \times \left( {\bf J} \times {\bf B} \right) 

208:     + \eta \nabla^2 {\bf B} .

209: \label{eq:induction}

210: \end{equation}

211: We introduce a filter in shells in Fourier space, such as ${\bf F}_K$ which

212: denotes the components of the field with wavenumbers between $K$ and 

213: $K+1$ [i.e. 

214: $ {\bf F}_K ({\bf x})= \sum_{k=K}^{K+1} \hat{\bf F}({\bf k}) e^{i {\bf k} 

215: \cdot {\bf x}} $], from Eqs. (\ref{eq:momentum}) and (\ref{eq:induction}) 

216: we can write detailed balance equations for the energy,

217: {\setlength\arraycolsep{2pt}

218: \begin{eqnarray}

219: \partial_t E_U(K) &=& \int \left\{ \sum_Q \left[ - {\bf U}_K 

220:    {\bf \cdot (U \cdot \nabla) \cdot U}_Q + {\bf U}_K 

221:    {\bf  \cdot (B \cdot \nabla) \cdot B}_Q \right] - \right. {} 

222:  \nonumber \\

223: &&{} - \nu \nabla^2 {\bf U}_K + {\bf f \cdot U}_K \Bigg\} \; d{\bf x}^3  \, ,

224: \end{eqnarray}}

225: %

226: {\setlength\arraycolsep{2pt}

227: \begin{eqnarray}

228: \partial_t E_B(K) &=& \int \left\{ \sum_Q \left[ - {\bf B}_K 

229:    {\bf \cdot (U \cdot \nabla) \cdot B}_Q + {\bf B}_K 

230:    {\bf \cdot (B \cdot \nabla) \cdot U}_Q + {} \right. \right.

231:  \nonumber \\

232: &&{} + \epsilon {\bf J}_K \cdot ({\bf B} \times {\bf J}_Q) \Big] 

233:    - \eta \nabla^2 {\bf B_K} \Bigg\} \; d{\bf x}^3 \, .

234: \end{eqnarray}}

235: %

236: Here, $E_U(K)$ and $E_B(K)$ denote respectively the kinetic and magnetic 

237: energy in the shell $K$. The above equations can be written in the more 

238: compact form:

239: \begin{equation}

240: \partial_t E_U(K) = \sum_Q [{\mathcal T}_{UU}(K,Q)+{\mathcal T}_{BU}(K,Q)] 

241:     - \nu {\mathcal D}_U(K) + {\mathcal F}(K) ,

242: \label{eq:Eu}

243: \end{equation}

244: %

245: \begin{equation}

246: \partial_t E_B(K) = 

247: \sum_Q [{\mathcal T}_{UB}(K,Q)+{\mathcal T}_{BB}(K,Q)] - 

248: \eta {\mathcal D}_B(K).

249: \label{eq:Eb}

250: \end{equation}

251: %

252: The functions ${\mathcal T}_{UU}(K,Q)$, 

253: ${\mathcal T}_{UB}(K,Q)$, ${\mathcal T}_{BB}(K,Q)$, and 

254: ${\mathcal T}_{BU}(K,Q)$ 

255: express the energy transfer between different fields and shells,

256: \begin{equation}

257: {\mathcal T}_{UU}(K,Q) \equiv 

258:     -\int {{\bf U}_K ({\bf U} \cdot \nabla) {\bf U}_Q} \; d{\bf x}^3 ,

259: \label{eq:TUU}

260: \end{equation}

261: %

262: \begin{equation}

263: {\mathcal T}_{UB}(K,Q) \equiv

264: \int {{\bf U}_K ({\bf B} \cdot \nabla) {\bf B}_Q} \; d{\bf x}^3 ,

265: \label{eq:TUB}

266: \end{equation}

267: %

268: \begin{equation}

269: {\mathcal T}_{BU}(K,Q) \equiv

270: \int {{\bf B}_K ({\bf B} \cdot \nabla) {\bf U}_Q} \; d{\bf x}^3 .

271: \label{eq:TBU}

272: \end{equation}

273: %

274: In general, for positive transfer, the first subindex denotes the 

275: field that receives energy, the second subindex the field that gives 

276: energy. The first wavenumber corresponds to the field receiving 

277: energy, and the second wavenumber to the field giving energy. As 

278: an example, positive ${\mathcal T}_{UU}(K,Q)$ represents energy 

279: transfered from the velocity field at the shell $K$ to velocity 

280: field at the shell $Q$. In the same way, positive $T_{UB}(K,Q)$ 

281: represents energy transfered from the magnetic field at wavenumbers 

282: $K$ to the velocity field at wavenumbers $Q$.

283: 

284: The transfer of magnetic to magnetic energy ${\mathcal T}_{BB}(K,Q)$ 

285: in Hall-MHD consists of two terms

286: \begin{equation}

287: {\mathcal T}_{BB}(K,Q) = {\mathcal T}_{BB}^\textrm{MHD}(K,Q) + 

288:     {\mathcal T}_{BB}^\textrm{Hall}(K,Q)

289: \end{equation}

290: where

291: \begin{equation}

292: {\mathcal T}_{BB}^\textrm{MHD}(K,Q) \equiv

293:     -\int {{\bf B}_K ({\bf U} \cdot \nabla) {\bf B}_Q  } \; d{\bf x}^3, 

294: \label{eq:TBBMHD}

295: \end{equation}

296: is the usual MHD transfer of magnetic energy through advection by the 

297: bulk velocity field, and 

298: \begin{equation}

299: {\mathcal T}_{BB}^\textrm{Hall}(K,Q) \equiv

300:     \epsilon \int {{\bf J}_K \cdot ({\bf B} \times {\bf J}_Q)} \; d{\bf x}^3, 

301: \label{eq:TBBHall}

302: \end{equation}

303: is the transfer of magnetic energy due to the Hall current. Note that 

304: the definition of the transfer terms corresponds to the MHD case in 

305: \citet{Alexakis05a}, except for the new term 

306: ${\mathcal T}_{BB}^\textrm{Hall}(K,Q)$. However, as will be shown later, 

307: the behavior of the rest of the transfer terms in Hall-MHD will also be 

308: indirectly modified by the presence of the Hall effect.

309: 

310: All these transfer functions satisfy the identity

311: \begin{equation}

312: \label{tran_id}

313: {\mathcal T}_{vw}(K,Q)=-{\mathcal T}_{wv}(K,Q) ,

314: \end{equation}

315: where $v,w$ can be either $U$ or $B$. This detailed conservation is 

316: what allows us to define the terms as transfers of energy between shells. 

317: Note that other groupings of the nonlinear terms in the Hall-MHD equations 

318: would not satisfy this symmetry condition.

319: 

320: In Eqs. (\ref{eq:Eu}-\ref{eq:Eb}) we also have two dissipation functions 

321: and the energy injection rate

322: \begin{equation}

323: \nu {\mathcal D}_{U}(K) \equiv  \nu   \int |{\bf \nabla U}_K|^2  d{\bf x}^3 ,

324: \end{equation}

325: \begin{equation}

326: \eta {\mathcal D}_{B}(K) \equiv  \eta \int |{\bf \nabla B}_K|^2 d{\bf x}^3 ,

327: \end{equation}

328: \begin{equation}

329: {\mathcal F}(K) \equiv \int { \bf f} \cdot {\bf U}_K  \, d{\bf x}^3 .

330: \end{equation}

331: 

332: Finally, we can also define the transfer of magnetic helicity. From Eq. 

333: (\ref{eq:induction1}) we have

334: \begin{equation}

335: \partial_t H(K) = \sum_Q {\mathcal T}_H (K,Q) - \eta {\mathcal D}_H (K),

336: \end{equation}

337: where the transfer of magnetic helicity from the wavenumber $K$ to the 

338: wavenumber $Q$ is given by

339: \begin{equation}

340: {\mathcal T}_H (K,Q) \equiv \int{ {\bf B}_K \cdot \left[\left({\bf U} 

341:     - \epsilon {\bf J}\right) \times {\bf B}_Q \right] d{\bf x}^3}.

342: \label{eq:TH}

343: \end{equation}

344: This transfer function satisfies the relation 

345: ${\mathcal T}_H (K,Q)=-{\mathcal T}_H (K,Q)$. 

346: The first term in Eq. (\ref{eq:TH}) proportional to 

347: ${\bf U}$ is the usual transfer of $H_M$ in MHD, while the second term 

348: proportional to $\epsilon {\bf J}$ is the contribution due to the Hall 

349: effect. Note that as a whole, magnetic helicity is transfered between 

350: the shells $K$ and $Q$ interacting with the electron velocity field 

351: ${\bf U} - \epsilon {\bf J}$. This is in agreement with the fact that 

352: in the ideal limit the magnetic field in the Hall-MHD system is frozen 

353: to the electron velocity field, instead of the bulk velocity field of 

354: the plasma as in MHD.

355: 

356: The dissipation rate of magnetic helicity at the wavenumber $K$ is given by

357: \begin{equation}

358: \eta {\mathcal D}_H (K) \equiv \eta \int{{\bf B}_K \cdot {\bf J}_K d{\bf x}^3}.

359: \end{equation}

360: It is also worth noting that, since the magnetic helicity is not a positive 

361: defined quantity contrary to the energy, the interpretation of its transfer 

362: is more difficult. We will not attempt here a separation of its different sign 

363: components \citep[see e.g.][]{Waleffe91,Chen03a,Chen03b} for the case of 

364: kinetic helicity in hydrodynamic turbulence).

365: 

366: \section{ \label{Transport} Transport coefficients }

367: 

368: In \citet{Mininni02}, the expression of the $\alpha$ dynamo coefficient 

369: was derived for Hall-MHD. Although an expression of the Hall-MHD turbulent 

370: diffusivity was derived by \citet{Mininni03a}, the closure was 

371: only valid for specific solutions of the Hall-MHD equations. To 

372: interpret the results from the energy transfer, it will be useful to 

373: have expressions for all the turbulent transport coefficients in the 

374: induction equation. To this end, and for the sake of simplicity, we 

375: will use mean field theory (MFT) \citep{Steenbeck66,Krause} and the 

376: reduced smooth approximation (RSA) \citep{Blackman99}. RSA was introduced 

377: to solve some ambiguities present in MFT when the magnetic field is 

378: strong enough to affect the velocity field trough the Lorentz force. 

379: Although there are still assumptions in MFT not completely justified, 

380: at least a qualitative agreement has been observed with simulations 

381: in the MHD case \citep{Brandenburg01} and the Hall-MHD case 

382: \citep{Mininni03b,Mininni05a}. The transport coefficients can also be 

383: derived using more elaborate closures, such as the Lagrangian History 

384: Direct Interaction Approximation (LHDIA) or the Eddy Damped Quasi Normal 

385: Markovian (EDQNM) closures \citep[see e.g.][]{Lesieur}. It is worth noting 

386: that the analysis that follows in Sections \ref{Simulations} and 

387: \ref{Transfers} is of general validity and independent of the assumptions 

388: we will use here to derive the turbulent transport coefficients.

389: 

390: We split the fields into

391: \begin{eqnarray}

392: {\bf U} &=& \overline{\bf U} + {\bf u} + {\bf u}_0 , \\

393: {\bf B} &=& \overline{\bf B} + {\bf b} + {\bf b}_0 ,

394: \end{eqnarray}

395: where ${\bf u}_0$ and ${\bf b}_0$ are isotropic and homogeneous 

396: solutions of Eqs. (\ref{eq:momentum}) and (\ref{eq:induction}) in 

397: the absence of the mean fields $\overline{\bf U}$ and $\overline{\bf B}$. 

398: The fields with overbars are large scale fields, and ${\bf u}$ and ${\bf b}$ 

399: are small scale corrections to the isotropic and homogeneous solutions 

400: due to the presence of the large scale fields. The fluctuating fields 

401: satisfy 

402: $\left< {\bf u} \right> = \left< {\bf u}_0 \right> = 

403:     \left< {\bf b} \right> = \left< {\bf b}_0 \right> = 0 $, 

404: where the brackets denote an average that satisfies Taylor's 

405: hypothesis \citep{Krause}. Replacing in Eq. (\ref{eq:induction}), using 

406: the equations for the ${\bf u}_0$ and ${\bf b}_0$ fields, dropping terms 

407: quadratic in the fluctuating fields ${\bf u}$ and ${\bf b}$, and averaging leads to

408: \begin{equation}

409: \partial_t \overline{\bf B} = \nabla \times \left[ \left( \overline{\bf U} 

410:     - \epsilon \overline{\bf J} \right) \times \overline{\bf B} 

411:     + \varepsilon \right] + \eta \nabla^2 \overline{\bf B} ,

412: \label{eq:largeB}

413: \end{equation}

414: where $\varepsilon$ is the mean field electromotive force

415: \begin{equation}

416: \varepsilon = \left< {\bf u}_0^e \times {\bf b} + {\bf u}^e \times 

417:     {\bf b}_0 \right> .

418: \label{eq:emfepsilon}

419: \end{equation}

420: Our main aim in this section is to close Eq. (\ref{eq:largeB}) and 

421: write $\varepsilon$ only as a function of averages of the fields 

422: ${\bf u}_0$, ${\bf b}_0$, and spatial derivatives of $\overline{\bf B}$. 

423: A simple argument of symmetry shows that in the approximately isotropic 

424: case

425: \begin{equation}

426: \varepsilon = \alpha \overline{\bf B} - \beta \nabla \times \overline{\bf B} 

427:     + \gamma \nabla \times \nabla \times \overline{\bf B} .

428: \label{eq:emf}

429: \end{equation}

430: 

431: From Eqs. (\ref{eq:momentum}) and (\ref{eq:induction}), and subtracting 

432: the equations for the mean flows, we can also write equations for the 

433: evolution of the turbulent fluctuations ${\bf u}$ and ${\bf b}$. We drop 

434: terms quadratic in ${\bf u}$ and ${\bf b}$, and keep only terms to 

435: zeroth and linear order in $\overline{\bf B}$,

436: \begin{equation}

437: \partial_t {\bf b} = \nabla \times \left( \overline{\bf U} \times {\bf b}_0 

438:     + {\bf u}_0^e \times \overline{\bf B} + \epsilon {\bf b}_0^e \times 

439:     \overline{\bf J} + {\bf u}^e \times {\bf b}_0 + 

440:     {\bf u}^e_0 \times {\bf b} - \varepsilon \right) + 

441:     \eta \nabla^2 {\bf b} .

442:     \label{eq:smallb}

443: \end{equation}

444: In this equation, $\varepsilon$ involves averaged quantities, and 

445: from the Taylor's hypothesis it gives no contribution to the mean 

446: electromotive force. The fourth and fifth terms on the {\it r.h.s.} can be 

447: dropped using RSA, namely that $|{\bf u}| , |{\bf b}| \ll |\overline{\bf B}|$ 

448: (note that this condition is less stringent than the usual assumptions 

449: in MFT, since the amplitude of the fields ${\bf u}_0$, ${\bf b}_0$ can 

450: be much larger than the amplitude of the mean magnetic field). We will 

451: assume the viscosity and diffusivity are small, and as a result we will 

452: also drop the last term on the right hand side of Eq. (\ref{eq:smallb}). 

453: Terms proportional to $\overline{\bf U}$ can be removed in the proper 

454: frame of reference. Finally we obtain

455: \begin{equation}

456: \partial_t {\bf b} \approx \nabla \times \left( {\bf u}_0^e \times 

457:     \overline{\bf B} + \epsilon {\bf b}_0^e \times \overline{\bf J} \right) .

458:     \label{eq:RSAb}

459: \end{equation}

460: The second term on the right hand side of Eq. (\ref{eq:RSAb}) involves 

461: only spatial derivatives of $\overline{\bf B}$ and gives no contribution 

462: to the $\alpha$ coefficient, but is retained here since it will give 

463: contributions to $\beta$ and $\gamma$.

464: 

465: Following the same steps, we can also write an equation for the evolution 

466: of ${\bf u}$,

467: \begin{equation}

468: \partial_t {\bf u} \approx \overline{\bf B} \cdot \nabla {\bf b}_0 

469:     + {\bf b}_0 \cdot \nabla \overline{\bf B} - \nabla p .

470:     \label{eq:smallu}

471: \end{equation}

472: 

473: To obtain the mean field electromotive force we replace time derivatives 

474: in Eqs. (\ref{eq:smallb}) and (\ref{eq:smallu}) by the inverse of a 

475: correlation time $\tau$. This step, common in MFT, assumes the existence 

476: of a finite correlation time. At present there is no evidence of its 

477: validity in general

478: \citep[see e.g.][]{Gruzinov95,Blackman02,Brandenburg05}. Since we are 

479: introducing a correlation time to close these equations, the expressions 

480: obtained for the turbulent transport coefficients will be considered as 

481: symbolic expressions.

482: 

483: Before replacing the expression for ${\bf u}$ in 

484: Eq. (\ref{eq:emf}), Eq. (\ref{eq:smallu}) has to be solved for the small 

485: scale pressure $p$. We will use a technique developed by \citet{Gruzinov95} 

486: \citep[see also][]{Blackman02}. The $\alpha$ effect is linear in 

487: $\overline{\bf B}$ and therefore the correct result can be obtained 

488: assuming $\overline{\bf B}$ uniform. Then 

489: $\partial_t {\bf u} \approx \overline{\bf B} \cdot \nabla {\bf b}_0 $. 

490: Replacing the time derivative by $\tau^{-1}$ and replacing the expression 

491: in Eq. (\ref{eq:emf}), we obtain in the weak isotropic case

492: \begin{equation}

493: \alpha = \frac{\tau}{3} \left< -{\bf u}_0^e \cdot \nabla \times 

494:     {\bf u}_0^e + {\bf b}_0 \cdot \nabla \times {\bf b}_0 - 

495:     \epsilon {\bf b}_0 \cdot \nabla \times \nabla \times  

496:     {\bf u}_0^e \right> .

497: \end{equation}

498: 

499: To compute $\beta$ and $\gamma$ we have to keep spatial derivatives of 

500: $\overline{\bf B}$ in Eq. (\ref{eq:smallu}), and therefore we have to 

501: solve for the pressure. This was done by \citet{Gruzinov95} transforming 

502: Eq. (\ref{eq:smallu}) to Fourier space, and doing a Taylor expansion of 

503: the projector operator for incompressible ${\bf u}$ assuming a large 

504: scale separation between the mean and fluctuating fields. In three spatial 

505: dimensions, it was shown that the pressure and 

506: ${\bf b}_0 \cdot \nabla \overline{\bf B}$ terms give no contribution to 

507: $\varepsilon$ in Eq. (\ref{eq:emfepsilon}). As a result, we are 

508: only left with the terms proportional to spatial derivatives of 

509: $\overline{\bf B}$ when Eq. (\ref{eq:RSAb}) and the first term on the 

510: r.h.s. of Eq. (\ref{eq:smallu}) are replaced on Eq. (\ref{eq:emfepsilon}). 

511: Again, assuming weak isotropy, we obtain the expressions 

512: for the remaining turbulent transport coefficients

513: {\setlength\arraycolsep{2pt}

514: \begin{eqnarray}

515: \beta &=& \frac{\tau}{3} \left< {{\bf u}_0^e}^2 + \epsilon \left( {\bf u}_0 

516:     \cdot \nabla \times {\bf b}_0^e + {\bf b}_0 

517:     \cdot \nabla \times {\bf u}_0^e \right) \right> , \\

518: \gamma &=& - \frac{\tau \epsilon}{3} \left< {\bf b}_0 \cdot {\bf u}_0^e 

519:     \right> .

520: \end{eqnarray}}

521: The two last terms in $\alpha$, and the third term in $\beta$ come from 

522: the small scale momentum equation and are related with the backreaction 

523: of the magnetic field into the velocity field. In the kinematic regime 

524: of a dynamo, \

525: $\alpha = - \tau/3\left<{\bf u}_0^e\cdot\nabla\times {\bf u}_0^e\right>$, 

526: which for $\epsilon=0$ reduces to the MHD case \citep{Krause}. The general 

527: expression for $\epsilon=0$ reduces to the MHD expression first found 

528: using the EDQNM closure by \citet{Pouquet76}, 

529: $\alpha = \tau/3\left<-{\bf u}_0^e\cdot\nabla\times {\bf u}_0^e 

530:     + {\bf b}_0\cdot\nabla\times {\bf b}_0 \right>$. Note also that in 

531: the MHD case in three dimensions, the turbulent diffusivity 

532: $\beta=-\tau \left< {{\bf u}_0}^2 \right>/3 $ is not changed during the 

533: nonlinear saturation \citep{Gruzinov95}.

534: 

535: The turbulent diffusivity $\beta$ in Hall-MHD is not positive definite, 

536: contrary to the pure MHD case \citep[note that negative effective 

537: diffusivities can be found in MHD if the assumption of homogeneity 

538: is dropped, see e.g.][]{Lanotte99}. A negative value of $\beta$ represents 

539: non-local transfer of energy from the small scale turbulent fields to the 

540: large scale magnetic field. This result will be of interest in the 

541: following sections.

542: 

543: It is worth studying the values of $\beta$ for particular cases. If 

544: $\epsilon$ is large enough and the system is magnetically dominated 

545: ($E_B \gg E_U$), then 

546: $\beta \approx - \tau \epsilon^2 \left< j_0^2 \right>/3$, where 

547: ${\bf j}_0 = \nabla \times {\bf b}_0$ and we assumed the average is 

548: a spatial average. In this case, $\beta$ is always negative

549: implying transfer of energy from the small scales to the large.

550: 

551: The normal modes of the Hall-MHD equations are circular polarized 

552: ($\nabla \times {\bf u}_0^e = \pm k {\bf u}_0^e$, 

553: $\nabla \times {\bf b}_0 = \pm k {\bf b}_0$) and dispersive, and in the 

554: limit $k\gg1$ they satisfy dispersion relations 

555: $\omega \sim \epsilon k^2 \overline{B}$ (whistlers, right-handed polarized) 

556: and $\omega \sim \overline{B}/ \epsilon$ (ion-cyclotron waves, left-handed 

557: polarized). Also, for these waves the fields are related by 

558: ${\bf b}_0 = -k \overline{B}/\omega {\bf u}_0^e$ \citep{Mininni05c}. 

559: If we assume a background of waves, 

560: $\beta \sim \tau/3 \left< {u_0^e}^2 \right> (1 \pm 2 \epsilon k^2 

561:     \overline{B} / \omega)$, which for $k$ large enough can give positive 

562: or negative turbulent diffusivity according to the orientation of the wave. 

563: Note that from the dispersion relations, at small scales whistlers give a 

564: finite contribution to the turbulent diffusivity, while ion-cyclotron waves 

565: give a much larger turbulent diffusivity that grows as $k^2$.

566: 

567: \section{ \label{Simulations} Simulations }

568: 

569: In this section we summarize the simulations that will be used to 

570: compute the energy and helicity transfer functions defined in 

571: Sec. \ref{Equations}. We performed three simulations in three 

572: dimensions with periodic boundary conditions, using a pseudospectral 

573: Hall-MHD code as described in \citet{Mininni03b,Mininni05a}. Runge-Kutta 

574: of second order is used to evolve the system of Eqs. (\ref{eq:momentum}) 

575: and (\ref{eq:induction1}). To ensure the divergence-free condition for the 

576: magnetic field, a curl is removed from Eq. (\ref{eq:induction1}) and the 

577: equation for the vector potential is instead solved, with the Couloumb's 

578: gauge $\nabla \cdot {\bf A} = 0$. The three simulations are done with a 

579: spatial resolution of $N^3=256^3$ grid points. The $2/3$ dealiasing rule is 

580: used, and as a result the maximum wavenumber resolved by the code is 

581: $k_{max} = N/3 \approx 85$. The kinematic viscosity and magnetic 

582: diffusivity are set to $\nu = \eta = 2 \times 10^{-3}$, and all the 

583: simulations are well resolved, in the sense that the kinetic 

584: [$k_\nu = (\left<\omega^2\right>/\nu^2)^{1/4}$] and magnetic 

585: [$k_\eta = (\left<J^2\right>/\eta^2)^{1/4}$] dissipation wavenumbers 

586: are smaller than $k_{max}$ at all times.

587: 

588: \begin{figure}

589: \begin{center} \includegraphics[width=9cm]{JPPmininni01} \end{center}

590: \caption{\label{fig:ener} Kinetic (thick curves) and magnetic 

591:     energy (thin curves) as a function of time, for runs 

592:     with $\epsilon = 0$ (solid lines), $\epsilon = 0.05$ (dotted lines), 

593:     and $\epsilon = 0.1$ (dashed lines).}

594: \end{figure}

595: 

596: In Hall-MHD, the Courant-Friedrichs-Levy (CFL) condition is more 

597: stringent than for MHD for which, with equipartition of kinetic and 

598: magnetic energy, the CFL condition for explicit time-stepping imposes 

599: an upper boundary on the time step $\Delta t \lesssim \Delta x/U_A$ 

600: where $\Delta x$ is the spatial step. In Hall-MHD, the dispersive 

601: nature of the whistlers impose 

602: $\Delta t \lesssim {\Delta x}^2/(\epsilon U_A)$. As a result, smaller 

603: time steps will be needed as $\epsilon$ is increased. Also, since the 

604: time step decreases quadratically as the spatial resolution is linearly 

605: increased, we cannot achieve now spatial resolutions higher than 

606: $256^3$ because of these constraints.

607: 

608: A helical forcing at $k_0=2$ given by an ABC flow 

609: {\setlength\arraycolsep{2pt}

610: \begin{eqnarray}

611: {\bf f} &=& \left[C\sin(k_0z)+B\cos(k_0y)\right] \hat{x} + 

612:     \left[A\sin(k_0x)+C\cos(k_0z)\right] \hat{y} + {} 

613:  \nonumber \\

614: &&{} + \left[B\sin(k_0y)+A\cos(k_0x)\right] \hat{z} ,

615: \end{eqnarray}}

616: with $A=0.9$, $B=1$, and $C=1.1$ was applied in the momentum equation. 

617: This election of the amplitude coefficients was done to ensure breaking 

618: the symmetries of the ABC flow and ensuring a faster development of 

619: turbulence \citep{Archontis03}. After a first hydrodynamic run made 

620: to reach a turbulent steady state, a random and small magnetic field was 

621: introduced at small scales. Initially the ratio of kinetic to magnetic 

622: energy was $E_u/E_b \sim 10^{-3}$. 

623: 

624: \begin{figure}

625: \begin{center} \includegraphics[width=9cm]{JPPmininni02} \end{center}

626: \caption{\label{fig:spec} Spectra of kinetic energy (thick curves) 

627:     and magnetic energy (thin curves) as a function of time, for 

628:     $t=5$ (solid), 15 (dash-dotted), 30 (dotted), and 45 (dashed). (a) 

629:     $\epsilon = 0.05$, and (b) $\epsilon = 0.1$.}

630: \end{figure}

631: 

632: The simulation was continued to see exponential growth of the magnetic 

633: energy (in the following, we will refer to this stage as the kinematic 

634: regime), and finally nonlinear saturation of the small scale magnetic 

635: field (in the following, Hall-MHD turbulence). Three simulations were 

636: done, with $\epsilon = 0$ (MHD), $\epsilon = 0.05$ (which corresponds to 

637: $k_{Hall} \approx 20$), and $\epsilon = 0.1$ ($k_{Hall} \approx 10$). 

638: Figure \ref{fig:ener} shows the time history of the kinetic and magnetic 

639: energies for these three runs. 

640: 

641: After $t \approx 20$, the small scale magnetic fields have reached 

642: saturation for all values of $\epsilon$, while the large scale magnetic 

643: field keeps growing slowly. As $\epsilon$ is increased, the magnetic 

644: energy reached by the system after the non-linear saturation of the 

645: small scales increases. However, this behavior is not monotonical in 

646: $\epsilon$ as shown by \citet{Mininni03b}.

647: 

648: \begin{figure}

649: \begin{center} \includegraphics[width=9cm]{JPPmininni03} \end{center}

650: \caption{\label{fig:heli} Absolute value of the magnetic helicity as a 

651:     function of time. Labels are as in Fig. \ref{fig:ener}. For 

652:     $\epsilon=0.1$ the magnetic helicity changes sign from negative 

653:     to positive at $t\approx24$.}

654: \end{figure}

655: 

656: The saturation of the large scale magnetic field takes place in a longer 

657: time \citep{Brandenburg01,Mininni05a}. Note that one of the biggest 

658: challenges for DNS is to attain scale separation between the

659: different dynamical ranges that must be resolved. Reynolds numbers in 

660: simulations are much smaller than the values observed in astrophysics 

661: and geophysics. Moreover, compared with hydrodynamics and MHD, the extra 

662: characteristic length scale in Hall-MHD (the Hall scale) makes it even 

663: harder to achieve a proper separation between all these scales. As a 

664: result, we will focus in this work in the energy transfer at scales 

665: smaller than $k_0$ (the energy injection band), and the late time 

666: large-scale evolution of these runs will not be discussed here 

667: \citep[more details can be found e.g. in][]{Mininni05a}.

668: 

669: Figure \ref{fig:spec} shows the time evolution of the magnetic and 

670: kinetic energy spectra, for the runs with $\epsilon = 0.05$ and 

671: $\epsilon = 0.1$. As previously mentioned ($t \approx 40$) the 

672: spectrum of energy at scales smaller than $k_0$ has saturated and 

673: reached a steady state, while the magnetic energy at $k=1$ keeps 

674: growing slowly. Note that the ratio of kinetic to magnetic energy at 

675: small scales in the saturated state depends on the value of $\epsilon$.

676: 

677: Another quantity that will be of interest in the next section is the 

678: magnetic helicity. Figure \ref{fig:heli} shows the time history of 

679: the absolute value of magnetic helicity for the three runs. Note that 

680: while in the MHD run ($\epsilon = 0$) the magnetic helicity grows 

681: monotonically with time, in the Hall-MHD runs the time evolution is 

682: strongly modified. For $\epsilon = 0.05$ the magnetic helicity grows 

683: slower than in the MHD case, and for $\epsilon = 0.1$ it changes sign 

684: at $t \approx 24$. As was observed by \citet{Mininni03b}, the Hall 

685: effect inhibits the generation of net magnetic helicity at large scales 

686: by the helical dynamo process. This inhibition grows monotonically with 

687: the amplitude of Hall term, and for values of $\epsilon$ large enough 

688: the magnetic helicity fluctuates around zero. The reason for this 

689: behavior will be discussed in the next section.

690: 

691: \section{ \label{Transfers} Transfers }

692: 

693: In this section we discuss the energy transfer terms defined in 

694: Sec. \ref{Equations} as obtained from the three DNS discussed in 

695: the previous section.

696: 

697: \subsection{The run with $\epsilon = 0.1$}

698: 

699: \begin{figure}

700: \begin{center} \includegraphics[width=9cm]{JPPmininni04} \end{center}

701: \caption{\label{fig:T1uu} Transfer of kinetic energy from $Q=20$ to $K$, 

702:     ${\mathcal T}_{UU}(K,Q=20)$, in the kinematic regime (solid line), 

703:     at $t=26$ (dotted line), and at $t=45$ (dashed line) in the run with 

704:     $\epsilon = 0.1$.}

705: \end{figure}

706: 

707: We start discussing in detail the transfer in the Hall-MHD run with 

708: $\epsilon =0.1$. At late times in this simulation, when the system 

709: is close to equipartition ($U_A \approx U$), the Hall wavenumber is 

710: $k_{Hall} \approx 10$. Since we consider transfer functions at different 

711: times, for the sake of comparison and unless explicitly said, all 

712: transfers in this subsection will be normalized using the {\it  r.m.s.} 

713: velocity and magnetic field according to their expressions 

714: [Eqs. (\ref{eq:TUU}-\ref{eq:TBBHall})]. Note that since $\epsilon {\bf J}$ 

715: has units of velocity (and ${\bf U}-\epsilon {\bf J}$ is the electron 

716: velocity), the transfer function $T_{BB}^\textrm{Hall}$ is normalized 

717: using $\left<B^2 |{\bf U}|\right>$. This election also allows for a 

718: direct comparison of this term against $T_{BB}^\textrm{MHD}$ [see Eq. 

719: (\ref{eq:TBBMHD})].

720: 

721: Figure \ref{fig:T1uu} shows the transfer of kinetic energy from 

722: the shell $Q=20$ to kinetic energy in shells $K$ for three different 

723: times. As previously mentioned, positive transfer denotes energy 

724: given by the shell $Q$, while negative transfer corresponds to 

725: energy received by this shell. In this case, kinetic energy 

726: in the shell $Q=20$ is mostly received from the shell $K=18$ 

727: (negative peak), and given to $K=22$ (positive peak). This 

728: function represents the local and direct transfer of kinetic energy 

729: to small scales. There are no noticeable differences in this transfer 

730: between the Hall-MHD runs ($\epsilon = 0.05$ and $0.1$), and the MHD 

731: run ($\epsilon = 0$).

732: 

733: The curve for early times (kinematic regime) corresponds to the 

734: initial exponential growth of magnetic energy, and is a time average 

735: properly normalized. As time evolves and the magnetic energy grows, 

736: the amount of kinetic energy transfered to small scales diminishes, 

737: since a larger amount of kinetic energy at large scales is turned 

738: into magnetic energy. This effect was previously observed in MHD runs 

739: \citep{Mininni05b}.

740: 

741: \begin{figure}

742: \begin{center} \includegraphics[width=9cm]{JPPmininni05} \end{center}

743: \caption{\label{fig:T1bbM} ${\mathcal T}_{BB}^\textrm{MHD}(K,Q=20)$ 

744:     in the kinematic regime (solid line), at $t=26$ (dotted line), and at 

745:     $t=45$ (dashed line) in the run with $\epsilon = 0.1$.}

746: \end{figure}

747: 

748: Figure \ref{fig:T1bbM} shows 

749: ${\mathcal T}_{BB}^\textrm{MHD}(K,Q=20)$, the transfer of magnetic 

750: energy at the shell $Q=20$ to magnetic energy in shells $K$ due to the 

751: advection by the bulk velocity field. As in the case of 

752: ${\mathcal T}_{UU}$, the transfer is local and the shell $Q=20$ 

753: receives most of the energy from $K=18$ (negative peak) and gives 

754: energy to the shell $K=22$ (positive peak). Again, no significant 

755: differences are observed between the three runs with different values 

756: of $\epsilon$, except that this transfer, in amplitude, gets 

757: substantially stronger as $\epsilon$ (and ${\bf B}$) increases.

758: 

759: \begin{figure}

760: \begin{center} \includegraphics[width=9cm]{JPPmininni06} \end{center}

761: \caption{\label{fig:T1bbh} ${\mathcal T}_{BB}^\textrm{Hall}(K,Q=20)$ in 

762:     the kinematic regime (solid line), at $t=26$ (dotted line), and at 

763:     $t=45$ (dashed line) in the run with $\epsilon = 0.1$.}

764: \end{figure}

765: 

766: \begin{figure}

767: \begin{center} \includegraphics[width=9cm]{JPPmininni07} \end{center}

768: \caption{\label{fig:T1buk} ${\mathcal T}_{BU}(K=20,Q)$ in the kinematic 

769:     regime (solid line), at $t=26$ (dotted line), and at $t=45$ (dashed 

770:     line) in the run with $\epsilon = 0.1$.

771:     The inset shows a blow up of the last transfer.}

772: \end{figure}

773: 

774: The total shell-to-shell transfer of magnetic energy is given by 

775: ${\mathcal T}_{BB}^\textrm{MHD}$ plus 

776: ${\mathcal T}_{BB}^\textrm{Hall}$. Figure \ref{fig:T1bbh} shows the 

777: ${\mathcal T}_{BB}^\textrm{Hall}(K,Q)$ transfer at $Q=20$. As in the 

778: previous cases, positive transfer denotes energy is given from the 

779: shell $Q=20$ to shells $K$, while negative transfer indicates the 

780: shell $Q$ receives energy from $K$. The 

781: ${\mathcal T}_{BB}^\textrm{Hall}$ transfer is small during the kinematic 

782: regime, but grows as the small scales reach nonlinear saturation. 

783: Although this transfer is noisier than the previous terms studied, 

784: two regions can be identified at late times. Around $Q=K=20$, the 

785: transfer is local and direct: positive and negative peaks can be 

786: observed at $K \approx 25$ and $K \approx 15$, indicating 

787: energy is received and given respectively by the shell $Q$ from 

788: and to these wavenumbers. On the other hand, at large scales 

789: (up to $K \approx 10$) a region with positive transfer can also be 

790: identified. This region indicates a non-local and inverse transfer 

791: of energy: the shells with $K$ between 1 and 10 receive magnetic 

792: energy from the shell $Q=20$. This combination of a local direct 

793: transfer of energy and a non-local inverse transfer is characteristic 

794: of the Hall term, and is in qualitative agreement with the turbulent 

795: dissipation derived in Sec. \ref{Transport} where it was shown that it

796: can take negative values.

797: 

798: \begin{figure}

799: \begin{center} \includegraphics[width=9cm]{JPPmininni08} \end{center}

800: \caption{\label{fig:T1buq} ${\mathcal T}_{BU}(K,Q=20)$ at three 

801:     different times: the kinematic regime (solid line), $t=26$ (dotted 

802:     line), and  $t=45$ (dashed line) for the run with $\epsilon = 0.1$.}

803: \end{figure}

804: 

805: The remaining transfer term is ${\mathcal T}_{BU}(K,Q)$, which 

806: when positive represents transfer of kinetic energy from the shell 

807: $Q$ to magnetic energy in the shell $K$.  Although the expression 

808: of this transfer function is equal for MHD and Hall-MHD, the 

809: transfer is modified by the Hall currents. The reason for this can 

810: be explained in two ways. On the one hand, the expression of the 

811: $\alpha$-effect in Sec. \ref{Transport} is modified by the Hall term, 

812: and this term represents transfer of energy from the turbulent 

813: velocity field to the mean magnetic field. On the other hand, 

814: waves are expected to give non-local coupling between the velocity 

815: and magnetic fields \citep[see e.g.][in MHD]{Iroshnikov63,Kraichnan65}. 

816: In Hall-MHD, the non-dispersive Alfv\'en waves of MHD are replaced by 

817: dispersive circularly polarized waves and as a result, the coupling 

818: between the two fields should also be modified.

819: 

820: Figure \ref{fig:T1buk} shows ${\mathcal T}_{BU}(K=20,Q)$, the 

821: energy transfered to the magnetic field at $K=20$ from the velocity 

822: field at shells $Q$. In the kinematic regime this transfer is non-local 

823: and similar to the MHD transfer \citep{Alexakis05a,Mininni05b}: the 

824: magnetic field at $K=20$ receives energy from the large scale flow at 

825: $Q=3$ and from all turbulent scales up to $Q \approx 20$. However, at 

826: late times the transfer is strongly modified. The magnetic field at 

827: $K=20$ still receives energy from a broad range of wavenumbers $Q$ 

828: smaller than $K$ \citep[as was found in][]{Alexakis05a}, 

829: but it also receives energy from larger wavenumbers 

830: ($Q \approx 22$), and gives energy to the velocity field at slightly 

831: smaller wavenumbers ($Q \approx 18$). Note that this indicates that in 

832: Hall-MHD a magnetic field at a given scale can give rise to velocity 

833: fluctuations at larger scales, a process studied by \citet{Mahajan05b} 

834: and referred there as the {\it reverse dynamo}.

835: 

836: \begin{figure*}

837: \begin{center} \includegraphics[width=12cm]{JPPmininni09} \end{center}

838: \caption{\label{fig:THall} ${\mathcal T}_{BB}^\textrm{Hall}(K,Q)$ 

839:     (left column), and ${\mathcal T}_{BU}(K,Q)$ (right column) at three 

840:     different times: the kinematic regime (top), $t=26$ (middle), and 

841:     $t=45$ (bottom) for the run with $\epsilon = 0.1$. In all figures, 

842:     $K$ is on the $x$ axis and $Q$ in the $y$ axis. Shading go from dark 

843:     (${\mathcal T}<0$) to light (${\mathcal T}>0$).}

844: \end{figure*}

845: 

846: Figure \ref{fig:T1buq} shows ${\mathcal T}_{BU}(K,Q=20)$, the 

847: energy received by the magnetic field at all wavenumbers $K$ 

848: from the velocity field in the shell $Q=20$.  During the kinematic regime, 

849: the velocity field in this shell gives energy to all magnetic shells, 

850: although the transfer peaks at wavenumbers larger than $Q$. But in 

851: the saturated regime, the transfer changes drastically again. The magnetic 

852: field at wavenumbers $K$ smaller than $Q \approx 16$, and in shells 

853: between 20 and 23 gives energy to the velocity field (negative transfer), 

854: while the magnetic field in shells between $K \approx 16 $ to 20 and for 

855: $K \gtrsim 23$ receives energy from the velocity field (positive transfer). 

856: This is just the counterpart of ${\mathcal T}_{BU}(K,Q)$ for constant $K$, 

857: and again shows that in Hall-MHD a small scale magnetic field can create 

858: large scale flows.

859: 

860: Figure \ref{fig:THall} shows shaded plots of 

861: ${\mathcal T}_{BB}^\textrm{Hall}(K,Q)$ and ${\mathcal T}_{BU}(K,Q)$ 

862: at different times. These are the two transfers that are strongly modified 

863: by the Hall currents, and the figures allow for a study of the terms for all 

864: values of $K$ and $Q$. 

865: Although noisy, a characteristic pattern can be recognized in 

866: ${\mathcal T}_{BB}^\textrm{Hall}$. As time evolves and the magnetic 

867: energy grows, the relative importance of this term grows. For 

868: wavenumbers $K,Q \gtrsim k_{Hall}\sim 10$, the function is positive 

869: (light) near and below the diagonal $K=Q$, and negative (dark) near and 

870: above this diagonal. This region close to the diagonal represents local 

871: and direct transfer of energy: a cut at constant $Q$ shows that close to 

872: the diagonal the shell $Q$ receives energy from neighboring shells with 

873: $K \lesssim Q$ (negative ${\mathcal T}_{BB}^\textrm{Hall}$) and gives 

874: energy to neighbor shells with $K \gtrsim Q$ (positive 

875: ${\mathcal T}_{BB}^\textrm{Hall}$). As we move far from this diagonal, 

876: the sign of the regions above and below the diagonal changes. This 

877: indicates a non-local and inverse transfer of magnetic energy, from 

878: small to large scales, in agreement with the expression for the 

879: turbulent magnetic diffusivity obtained in Sec. \ref{Transport}.

880: 

881: The ${\mathcal T}_{BU}(K,Q)$ also shows an interesting behavior as a 

882: function of time. During the kinematic regime, ${\mathcal T}_{BU}$ is 

883: positive in a triangle defined by $K \gtrsim Q$. This indicates that the 

884: velocity field in a given shell amplifies the magnetic field in that shell 

885: and all the shells with larger wavenumber (smaller scales). Also a strong 

886: band around $Q = 3$ is observed, indicating that the velocity field in the 

887: energy injection band gives a lot of energy to the magnetic field. These 

888: results are similar to the kinematic MHD dynamo \citep[see][]{Mininni05b}. 

889: However, at late times an inverse process can be identified close to the 

890: diagonal $K=Q$. Above it, ${\mathcal T}_{BU}$ is positive, while below 

891: it, it is negative. This represents transfer of magnetic energy from a shell 

892: $K$ to kinetic energy in slightly smaller wavenumbers $Q$.

893: 

894: \begin{figure}

895: \begin{center} \includegraphics[width=9cm]{JPPmininni10} \end{center}

896: \caption{\label{fig:T1flu} Energy fluxes in the run with $\epsilon = 0.1$ at 

897:     $t=45$: $\Pi_{BB}^\textrm{Hall}(k)$ (solid line), 

898:     $\Pi_{BB}^\textrm{MHD}(k)$ (dash-dotted line), $\Pi_{UU}(k)$ (dashed), 

899:     and $\Pi_{BU}(k)$ (dotted). The thick line is the total flux 

900:     in the simulation.}

901: \end{figure}

902: 

903: Since ${\mathcal T}_{BB}^\textrm{Hall}$ and ${\mathcal T}_{BU}$ give both 

904: direct and inverse transfers of energy (locally or non-locally), it is of 

905: interest to quantify which direction wins when all the contributions to the 

906: transfer are added. To this end, we computed the contribution of each 

907: transfer term to the energy flux. The total energy flux at a wavenumber 

908: $k$ is given by

909: \begin{equation}

910: \Pi(k) = \sum_{K=0}^k \sum_Q {\mathcal T}(K,Q) ,

911: \end{equation}

912: where ${\mathcal T} = {\mathcal T}_{UU} + {\mathcal T}_{BB} + 

913:     {\mathcal T}_{UB} + {\mathcal T}_{BU}$ is the total energy transfer. We 

914: can split this flux into the energy flux due solely to the transfer of kinetic 

915: energy

916: \begin{equation}

917: \Pi_{UU}(k) = \sum_{K=0}^k \sum_Q {\mathcal T}_{UU}(K,Q) ,

918: \end{equation}

919: the flux due to the transfer of magnetic energy 

920: $\Pi_{BB} = \Pi_{BB}^\textrm{MHD} + \Pi_{BB}^\textrm{Hall}$, where

921: \begin{equation}

922: \Pi_{BB}^\textrm{MHD}(k) = \sum_{K=0}^k \sum_Q 

923:     {\mathcal T}_{BB}^\textrm{MHD}(K,Q) ,

924: \end{equation}

925: \begin{equation}

926: \Pi_{BB}^\textrm{Hall}(k) = \sum_{K=0}^k \sum_Q 

927:     {\mathcal T}_{BB}^\textrm{Hall}(K,Q) ,

928: \end{equation}

929: and the hybrid flux due to interactions between the velocity and magnetic 

930: fields

931: \begin{equation}

932: \Pi_{BU}(k) = \sum_{K=0}^k \sum_Q \left[ {\mathcal T}_{BU}(K,Q) + 

933:     {\mathcal T}_{UB}(K,Q) \right] .

934: \end{equation}

935: To compute the fluxes, the transfer functions are not normalized.

936: 

937: Figure \ref{fig:T1flu} shows the partial energy fluxes at $t=45$. Since 

938: all transfer functions were computed up to $K,Q=40$, the partial fluxes 

939: go to zero artificially at this wavenumber, although in the simulation 

940: the total energy flux goes to zero only at the maximum resolved wavenumber 

941: $k_{max}$. 

942: 

943: The total flux is positive at wavenumbers larger than $k_0$ (the energy 

944: injection band), indicating a direct cascade of the total energy. At 

945: wavenumbers smaller than $k_0$ the total flux is negative, an evidence 

946: of large scale dynamo action. A substantial portion of the total flux is due 

947: to the transfer of energy from the kinetic to the magnetic reservoirs 

948: ($\Pi_{BU}$), and this contribution to the flux is positive at all 

949: wavenumbers (larger than the forced ones) indicating a net direct 

950: transfer of the energy. We note that this flux is due to the non-local 

951: $T_{UB}$ and $T_{BU}$ transfer terms. The flux due to the transfer of 

952: kinetic energy $\Pi_{UU}$ is also positive at all wavenumbers. But the 

953: flux due to the transfer of magnetic energy $\Pi_{BB}^{Hall}$ is only 

954: positive at wavenumbers larger than $k_{Hall}$. For wavenumbers smaller 

955: than $k \approx 10\sim k_{Hall}$, $\Pi_{BB}^\textrm{Hall}$ changes sign, 

956: giving as a result a net inverse transfer of magnetic energy, from small 

957: to large scales. This indicates that a magnetically dominated Hall-MHD 

958: system could display backscatter of the magnetic energy. Magnetic 

959: fluctuations at small scales could give rise to large scale magnetic 

960: fields, as is also implied by the expression of $\beta$ found in Sec. 

961: \ref{Transport}.

962: 

963: Note that although in MHD the inverse cascade of magnetic helicity 

964: can give a similar result, the backscatter predicted in 

965: Hall-MHD by the turbulent diffusivity is novel, since it can take place even 

966: in the absence of helicity in the fields. To illustrate this we show 

967: results of non-helical magnetically dominated simulations in Sec. 

968: \ref{Upscaling}. It is worth noting that at wavenumbers smaller than 

969: $k_{Hall}$, the Hall term increases the flux of magnetic energy to 

970: smaller scales, thus also in agreement with results showing the Hall 

971: currents increase the amount of small scale perturbations 

972: \citep{Birn01,Laveder02a,Laveder02b,Morales05}.

973: 

974: \subsection{Dependence with $\epsilon$}

975: 

976: Now we discuss in detail the dependence of the results as $\epsilon$ (or 

977: the Hall scale) is varied. To this end, we consider the runs with 

978: $\epsilon = 0.1$, $0.05$, and 0 (MHD). As previously mentioned, the 

979: transfer terms ${\mathcal T}_{UU}(K,Q)$ and ${\mathcal T}_{BB}(K,Q)$ do not 

980: show a dependence with the amplitude of the Hall effect. These terms give 

981: direct and local transfer of energy to small scales, as in MHD 

982: \citep{Alexakis05a}. As a result, we will discuss the change in the 

983: remaining transfer terms as $\epsilon$ is varied.

984: 

985: Since the Hall effect is more relevant when the magnetic field is stronger, 

986: we will consider the transfer terms at $t=45$, when a large scale magnetic 

987: field is present and the small scales have reached saturation. Since we 

988: will examine runs with different values of $\epsilon$ at the same time, 

989: in this subsection the transfers are not normalized using the energies. 

990: 

991: The transfer of magnetic energy due to the Hall term 

992: ${\mathcal T}_{BB}^\textrm{Hall}(K,Q=20)$ is of course zero in the MHD 

993: case ($\epsilon=0$). As $\epsilon$ is increased, except for an increase 

994: in its amplitude (not shown), no significant differences are observed 

995: and its behavior is similar to the one examined in Sec. \ref{Transfers}.

996: 

997: \begin{figure}

998: \begin{center} \includegraphics[width=9cm]{JPPmininni11} \end{center}

999: \caption{\label{fig:Tebuq} ${\mathcal T}_{BU}(K=20,Q)$ at $t=45$ for 

1000:     $\epsilon = 0$ (solid line), $0.05$ (dotted line), and $0.1$ 

1001:     (dashed line).} 

1002: \end{figure}

1003: 

1004: Figure \ref{fig:Tebuq} shows the behavior of ${\mathcal T}_{BU}(K=20,Q)$ 

1005: (the transfer of kinetic energy in the shells $Q$ to magnetic energy 

1006: in the shell $K=20$) as $\epsilon$ is varied. The strong peak at $Q=3$ is 

1007: associated with the injection band. This transfer is non-local in the three 

1008: runs, as is evidenced by the positive plateau from $Q\approx3$ 

1009: to $Q\approx 16$. As a result, the velocity field in all these shells 

1010: gives energy to the magnetic field at $K=20$. As $\epsilon$ is increased, 

1011: a local transfer grows in the neighborhood of $Q=20$. The velocity field 

1012: at wavenumbers $K$ slightly larger give energy to the magnetic field at 

1013: $K=20$, while the magnetic field gives energy to the velocity field at 

1014: wavenumbers slightly smaller ($Q \approx 18$).

1015: 

1016: \begin{figure}

1017: \begin{center} \includegraphics[width=9cm]{JPPmininni12} \end{center}

1018: \caption{\label{fig:fluxbb} $\Pi_{BB}(k)$ at $t=45$ for $\epsilon = 0$ 

1019:     (solid line), $0.05$ (dotted line), and $0.1$ (dashed line).}

1020: \end{figure}

1021: 

1022: We can compute the energy flux as $\epsilon$ increases. Since the transfer 

1023: of kinetic energy is not changed, we will focus on two contributions to 

1024: the total flux: the flux of magnetic energy $\Pi_{BB}(k)$, and the 

1025: hybrid flux $\Pi_{BU}(k)$ due to the terms turning kinetic into magnetic 

1026: energy and vice-versa. The magnetic energy flux 

1027: $\Pi_{BB} = \Pi_{BB}^\textrm{MHD} + \Pi_{BB}^\textrm{Hall}$ is shown in 

1028: Fig. \ref{fig:fluxbb}. At scales larger than $k_{Hall}$, negative flux 

1029: of magnetic energy is observed, giving backscatter of magnetic energy 

1030: to large scales. As $\epsilon$ is increased, the amplitude of the 

1031: backscatter grows, and the wavenumber where the flux changes sign moves 

1032: to larger $k$.

1033: 

1034: \begin{figure}

1035: \begin{center} \includegraphics[width=9cm]{JPPmininni13} \end{center}

1036: \caption{\label{fig:fluxbu} $\Pi_{BU}(k)$ at $t=45$ for $\epsilon = 0$ 

1037:     (solid line), $0.05$ (dotted line), and $0.1$ (dashed line).}

1038: \end{figure}

1039: 

1040: Figure \ref{fig:fluxbu} shows the flux $\Pi_{BU}(k)$. At wavenumbers 

1041: smaller than the forcing wavenumber ($k=3$), the flux is negative. This 

1042: is a signature of large scale dynamo action: the magnetic field at 

1043: large scales is fed by the small scale velocity field. Remarkably, as 

1044: $\epsilon$ is increased, the amplitude of the negative flux at large 

1045: scales increases. This is in good agreement with dynamo simulations 

1046: where the large scale magnetic field was observed to grow faster in the 

1047: presence of Hall currents \citep{Mininni03b,Mininni05a}.

1048: 

1049: \subsection{Transfer of magnetic helicity}

1050: 

1051: We discuss briefly the transfer of magnetic helicity in the saturated 

1052: case $t=45$. To study the transfer associated with the inverse cascade 

1053: of magnetic helicity at scales larger than the forcing scale, a large 

1054: separation between this scale and the largest scale in the box is needed. 

1055: At a fixed spatial resolution, this reduces the Reynolds numbers, and 

1056: as a result reduces also the separation between the Ohmic scale and the 

1057: Hall scale. This study is beyond the aim of this work. But we want to 

1058: point out a remarkable feature observed in the transfer of magnetic 

1059: helicity at scales smaller than the forcing scale.

1060: 

1061: \begin{figure}

1062: \begin{center} \includegraphics[width=9cm]{JPPmininni14} \end{center}

1063: \caption{\label{fig:Thelicity} ${\mathcal T}_H(K,Q=20)$ normalized by the 

1064:     magnetic helicity at the shell $Q=20$, at $t=45$ for $\epsilon = 0$ 

1065:     (solid line), $0.05$ (dotted line), and $0.1$ (dashed line).}

1066: \end{figure}

1067: 

1068: Figure \ref{fig:Thelicity} shows the transfer ${\mathcal T}_H(K,Q=20)$ 

1069: normalized by the magnetic helicity in the shell $Q=20$ at $t=45$. 

1070: The transfer is mostly local in the three simulations, peaking at 

1071: wavenumbers $K$ slightly smaller and larger than $Q$. However, as 

1072: $\epsilon$ is increased the transfer rate of magnetic helicity 

1073: is strongly quenched. This slow down in the transfer in Hall-MHD 

1074: explains the behavior observed in Fig. \ref{fig:heli}. In MHD and 

1075: Hall-MHD dynamos, the external mechanical forcing generates equal 

1076: amounts of magnetic helicity of opposite sign at scales smaller 

1077: and larger than the forcing band \citep{Seehafer96,Brandenburg01,Mininni03b}. 

1078: Since the transfer of magnetic helicity between different shells in 

1079: the Hall-MHD runs is almost stopped, it takes more time for the 

1080: magnetic helicity at scales smaller than the forcing scale to reach 

1081: the dissipative scale where it can be destroyed. As a result, both 

1082: signs of magnetic helicity piles up close to the forcing band, 

1083: decreasing the growth rate of net magnetic helicity at scales larger 

1084: than the forcing scale, and allowing also for 

1085: the possibility for a sign change of the net magnetic helicity.

1086: 

1087: \section{ \label{Upscaling} Backscatter of magnetic energy in Hall-MHD}

1088: 

1089: The mechanically forced runs discussed in the previous section show 

1090: negative flux of magnetic energy at large scales due to the Hall effect, 

1091: in agreement with negative values of the turbulent diffusivity. This 

1092: indicates that in a magnetically dominated simulation, backscatter of 

1093: magnetic energy could be observed if the Hall term is strong enough. 

1094: Note that here we are using the word {\it backscatter} to refer to this 

1095: transfer of magnetic energy from the small to the large scales. This is 

1096: done in opposition to the usual terminology of inverse cascades, since 

1097: we have been unable to identify any ideal invariant of the Hall-MHD 

1098: equations cascading inversely with constant flux to the large scales.

1099: 

1100: To study this scenario, we did three simulations with $\epsilon = 0$, 

1101: $0.2$, and $0.5$. The kinematic viscosity and magnetic diffusivity 

1102: were $\nu = \eta = 3.5 \times 10^{-2}$, and the spatial resolution was 

1103: $N^3 = 128^3$. The initial condition was ${\bf U} = {\bf B} = 0$. The 

1104: system was forced with a non-helical and random electromotive force given 

1105: by a superposition of harmonic modes at wavenumbers $k=9$ and 10. The 

1106: phases of the force were changed with a correlation time of 

1107: $\tau = 1.25 \times 10^{-2}$, and the time step was set to 

1108: $\Delta t = 1.5 \times 10^{-3}$. Note that in the absence of magnetic 

1109: helicity, no inverse cascade is expected in the MHD case 

1110: \citep[see however][for cases where a large scale shear is present]{Lanotte99}.

1111: 

1112: \begin{figure}

1113: \begin{center} \includegraphics[width=9cm]{JPPmininni15} \end{center}

1114: \caption{\label{fig:invsp1} (a) Kinetic energy spectrum $E_U(k)$ and 

1115:     (b) magnetic energy spectrum $E_B(k)$ at $t=3$, for runs with 

1116:     $\epsilon = 0$ (solid line), $0.05$ (dotted line), and $0.1$ 

1117:     (dashed line).}

1118: \end{figure}

1119: 

1120: \begin{figure}

1121: \begin{center} \includegraphics[width=9cm]{JPPmininni16} \end{center}

1122: \caption{\label{fig:invsp2} (a) Kinetic energy spectrum and (b) 

1123:     magnetic energy spectrum at $t=75$ for runs with $\epsilon = 0$ 

1124:     (solid line), $0.05$ (dotted line), and $0.1$ (dashed line).}

1125: \end{figure}

1126: 

1127: The system was run until reaching a turbulent steady state. All the 

1128: quadratic invariants (with the exception of the total energy) were 

1129: verified to be small: the magnetic helicity fluctuates in the three 

1130: runs around zero, both the global quantity as well as its spectral 

1131: density at each individual Fourier shell. Figure \ref{fig:invsp1} 

1132: shows the kinetic and magnetic energy spectrum at early times ($t=3$). 

1133: The shell of wavenumbers associated with the external magnetic force 

1134: is easily recognized in the peak in Fig. \ref{fig:invsp1}(b).

1135: 

1136: As time evolves, an increase in the magnetic energy at wavenumbers 

1137: smaller than the forcing wavenumber is observed. Figure \ref{fig:invsp2} 

1138: shows the kinetic and magnetic energy spectrum at $t=75$, when the 

1139: system has reached a steady state. The three runs are dominated by 

1140: the magnetic energy (note that the peak in the magnetic energy spectrum 

1141: around the forcing band gives the largest contribution to the energy). 

1142: The spectrum of kinetic energy is similar for the three runs, and 

1143: large scale perturbations are observed because of the injection of 

1144: kinetic energy by the Lorentz force. However, the magnetic energy 

1145: spectrum is strongly modified as $\epsilon$ is increased. While the 

1146: spectra of the three simulations peak in the energy injection band, 

1147: the magnetic energy at $K=1$ in the run with $\epsilon=0.5$ is 

1148: three orders of magnitude larger than in the MHD run. The magnetic 

1149: energy in all wavenumbers smaller than the forcing wavenumbers 

1150: increases as $\epsilon$ is increased. 

1151: 

1152: The backscatter of magnetic energy is in good agreement with the 

1153: negative flux $\Pi_{BB}$ observed in the previous section, and the 

1154: negative turbulent transport coefficients derived for the magnetically 

1155: dominated case. Note that an increase in the level of the small 

1156: scale magnetic fluctuations (for wavenumbers smaller than the energy 

1157: injection wavenumbers) is also observed in Fig. \ref{fig:invsp2}(b).

1158: 

1159: \section{ \label{Discussion} Discussion }

1160: 

1161: In this work we presented energy transfer in Hall-MHD turbulence as 

1162: obtained from numerical simulations. The properties of the spectral 

1163: transfer is one of the building blocks of turbulence theories, and to 

1164: the best of our knowledge no attempt to study transfer and cascades of 

1165: ideal invariants in this system of equations had been attempted before.

1166: 

1167: Before proceeding with the discussion of our results, we have to 

1168: warn the reader about a clear limitation of the numerical results 

1169: presented. As previously mentioned, an astrophysics-like scale 

1170: separation between the box size, the energy injection scale, the Hall scale, 

1171: and the Ohmic dissipation scale is well beyond today computing resources. 

1172: We tested the dependence of our results as $\epsilon$ was varied, but 

1173: no attempt was made to change the Reynolds numbers in our simulations. 

1174: This being said, we believe that even under this limitation, an 

1175: understanding of the transfer of energy between different scales is of 

1176: uttermost importance for the development of a theory of turbulence for 

1177: Hall-MHD or other extensions of magnetohydrodynamic to take into account 

1178: kinetic plasma effects.

1179: 

1180: Direct evidence of nonlocality of the energy transfer was observed. While 

1181: the total energy displays a direct cascade to small scales, in the 

1182: individual transfer terms, both directions (toward small and large scales) 

1183: were identified. Coupling between the magnetic and velocity fields is 

1184: strongly modified by the Hall effect, and a local backscatter of energy 

1185: from the magnetic field to the velocity field at slightly larger scales 

1186: was observed. This behavior can be expected since the Hall term changes 

1187: the nature of the nondispersive MHD Alfv\'en waves, into dispersive and 

1188: circularly polarized waves. As a result, the nonlinear coupling between 

1189: the two fields is also changed.

1190: 

1191: Also a nonlocal backscatter of magnetic energy was observed at scales 

1192: larger than the Hall scale. This backscatter was verified in non-helical 

1193: magnetically forced simulations, where the amplitude of the magnetic 

1194: field at scales larger than the forcing scale was observed to grow in 

1195: the Hall-MHD simulations, but not in the MHD run. In some sense, the 

1196: magnetic field in Hall-MHD being frozen in the ideal case to the 

1197: electron velocity field, couples non-locally both small scales 

1198: (the current) and large scales (the bulk velocity field).

1199: 

1200: All these processes can be partially explained considering transport 

1201: turbulent coefficients estimated from MFT. 

1202: Unlike MHD, the turbulent diffusivity in Hall-MHD 

1203: is not positive definite. In particular, its expression shows that 

1204: ion-cyclotron waves are more likely to produce large values of negative 

1205: (backscatter) or positive (reconnection) turbulent diffusivity than 

1206: the whistler mode.

1207: 

1208: The transfer of magnetic helicity at small scales was also observed 

1209: to be quenched by the Hall effect. While the mechanisms generating 

1210: magnetic helicity in the Hall-MHD dynamo are the same as in MHD 

1211: \citep{Mininni03b}, the transport of helicity is expected to be changed 

1212: by the Hall currents \citep{Ji99}. As a result of the slow down in the 

1213: transfer rate of magnetic helicity by the Hall effect, the late time 

1214: evolution of the system is not characterized by a maximally helical 

1215: large scale magnetic field as in the MHD case 

1216: \citep{Pouquet76,Meneguzzi81,Brandenburg01}.

1217: 

1218: The Hall term gives a direct transfer of magnetic energy at scales 

1219: smaller than the Hall scale, and an inverse transfer at scales larger 

1220: than the Hall scale. This finding sheds light into the conflicting 

1221: results reported in the literature, where the Hall effect was observed 

1222: to increase the amount of small scales and magnetic dissipation in some 

1223: cases, and to help large scale reorganization processes in other cases, 

1224: as mentioned in the introduction.

1225: 

1226: As a result of this dual direction of the Hall transfer, a change in 

1227: the power law followed by the total energy spectrum can be expected close 

1228: to the Hall wavenumber. Steepening of the energy spectrum for wavenumbers 

1229: smaller than $k_{Hall}$ was observed in 2.5D simulations with strong 

1230: magnetic fields imposed, when the cross-correlation between the velocity 

1231: and magnetic fields was significant \citep{Ghosh96}. In three dimensional 

1232: dynamo simulations where the cross correlation is in general small, 

1233: no change was observed \citep{Mininni05a}, although a faster 

1234: growth of the large scale magnetic field was found. Given the nonlocal 

1235: nature of the transfer in Hall-MHD, and the scale separation needed 

1236: to observe a clear change in the energy spectrum, probably a huge 

1237: increase in the spatial resolution is needed to confirm it.

1238: 

1239: \begin{acknowledgments}

1240: Computer time was provided by NCAR. The NSF grant CMG-0327888

1241: at NCAR supported this work in part and is gratefully acknowledged.

1242: \end{acknowledgments}

1243: 

1244: \bibliography{ms}

1245: 

1246: \end{document}

1247: