1: \documentstyle[aps,epsfig,twocolumn,citesort,psfig19,floats]{revtex}

2: %\documentstyle[12pt,prl,aps,epsfig,citesort,psfig19,floats]{revtex}

3: \renewcommand {\baselinestretch}{1.}

4: 

5: 

6: 

7: 

8: %\renewcommand\floatpagefraction{0.9}

9: 

10: 

11: 

12: \newcommand{\aleq}{\mbox{\ 

13: \raisebox{-.9ex}{$\stackrel{\textstyle<}{\sim}$}\ }}

14: \newcommand{\ageq}{\mbox{\

15: \raisebox{-.9ex}{$\stackrel{\textstyle >}{\sim}$}\ }}

16: 

17: 

18: 

19: % should go away for length estimates

20: \topmargin -1cm

21: %\newcommand{\marg}[1]{\marginpar{\raggedright\scriptsize#1}}

22: %\parskip=1pt

23: %\setlength{\textwidth}{6.in}

24: %\setlength{\textheight}{8.5in}

25: %\setlength{\evensidemargin}{0.25in}

26: %\setlength{\oddsidemargin}{0.25in}

27: 

28: 

29: 

30: 

31: \def\x{{\mbox{\boldmath$x$}}}

32: \def\u{{\mbox{\boldmath$u$}}}

33: \def\r{{\mbox{\boldmath$r$}}}

34: \def\f{{\mbox{\boldmath$f$}}}

35: \def\q{{\mbox{\boldmath$q$}}}

36: \def\p{{\mbox{\boldmath$p$}}}

37: \def\k{{\mbox{\boldmath$k$}}}

38: \def\n{{\mbox{\boldmath$n$}}}

39: \def\e{{\mbox{\boldmath$e$}}}

40: \def\v{{\mbox{\boldmath$v$}}}

41: 

42: 

43: 

44: \def\cinf{{c_{\epsilon , \infty}}}

45: \def\la{{\langle}}

46: \def\ra{{\rangle}}

47: \def\om{{\omega}}

48: \def\nab{{\bf \nabla}}

49: \def\eps{{\epsilon}}

50: \def\pb{{\bar p }}

51: \def\pt{\tilde{p}}

52: \def\rel{{Re_\lambda}}

53: \def\fns{{\footnotesize}}

54: \def\begineq{\begin{equation}}

55: \def\endeq{\end{equation}}

56: \def\be{\begin{equation}}

57: \def\ee{\end{equation}}

58: 

59: \newcommand{\marg}[1]{\marginpar{\raggedright\scriptsize#1}}

60: 

61: 

62: \begin{document}

63: \bibliographystyle{prsty}

64: %\psdraft

65: 

66: \title{

67: Periodically kicked turbulence

68: }

69: \author{

70: Detlef Lohse 

71: }

72: \address{

73: Department of Applied Physics  and J.\ M.\ Burgers Centre for

74: Fluid Dynamics, 

75: University of Twente, 7500 AE Enschede, 

76: Netherlands}

77: 

78:   

79: 

80: \date{\today}

81: 

82: \maketitle

83: \begin{abstract}

84: Periodically kicked turbulence is theoretically analyzed within a

85: mean field theory.

86: For large enough kicking strength $A$ and kicking frequency $f$

87: the Reynolds number 

88: grows exponentially and then runs into some saturation. The saturation

89: level 

90: $Re^{sat}$ 

91: can be calculated

92: analytically;

93: different regimes can be observed. 

94: For large enough $Re$ we find  $Re^{sat} \propto 

95: Af$, but intermittency can modify this scaling law. 

96: We suggest an experimental realization of periodically kicked turbulence to

97: study the different regimes we theoretically predict and thus to better

98: understand the effect of forcing on fully developed turbulence. 

99: \end{abstract}

100: 

101: 

102: %----------------------------------------------------------------------

103: 

104: \vspace{1cm}

105: 

106: Periodically driven flow is ubiquitous. Faraday's experiment

107: \cite{far31} is an idealized version thereof, more relevant examples are the

108: earth's atmosphere, driven by periodical heating of the sun, or 

109: the blood flow in veins, driven by the beating heart. Another example

110: is the gas flow inside a sonoluminescing bubble

111: which is periodically kicked by the collapsing bubble 

112: wall

113: \cite{cru94}.

114: Another example of periodically kicked turbulence is 

115: the numerical realization of homogeneous shear flow 

116: \cite{pum96} where periodical remeshing is necessary.

117: 

118: 

119: In this paper we set up a mean field theory for periodically kicked flow, based

120: on the mean field theory for decaying turbulence 

121: \cite{loh94a} which was able to

122: describe the experimentally measured energy decay in turbulent liquid

123: helium flow with fixed external length scale \cite{smi93a}.

124: The goal of this paper is 

125: to theoretically understand the different flow regimes

126: which are to be expected, to explore the effect

127: of intermittency corrections on these regimes,

128:  and to ultimately initiate experiments.

129: 

130: Another motivation for the paper is to study the effect of a specified

131: type of forcing on turbulence. In most theoretical studies on turbulence a

132: Gaussian random noise, acting on the largest length scales, is assumed.

133: Only recently experimentalists started to systematically vary the type

134: of forcing \cite{lab96,cam97b}. 

135: This

136: work is a further step towards

137: the analysis of a more specific type of forcing. 

138: 

139: To define the model, we have to (i) calculate the energy input during the kick

140: and (ii) know how the energy is dissipated in the time

141: $\Delta t$ between successive kicks.

142: 

143: {\underline{(i) Kick:}}

144: As an illustration, consider plane shear flow in the 1-direction; the flow is

145: sheared in the 3-direction.

146: The width of the channel is $L$, the velocity of the upper plate is $U$, the

147: lower one is at rest. The average energy dissipation rate can be

148: calculated to be

149: (see e.g.\ \cite{geb95}) 

150: \be

151: \eps = -{U\over L} \left( \la u_3 u_1 \ra_A -\nu \partial_3 \la u_1 \ra_A

152: \right),

153: \label{eq1}

154: \ee

155: where $\la\ra_A$ denotes the average over the x-y-plane and $\nu $ is the

156: viscosity.

157: For laminar flow the first term in the bracket is zero and

158: the second one is $-\nu U/L$. In turbulent flow

159: in the middle of the channel 

160: the second term on the rhs will hardly

161:  contribute for large enough Reynolds numbers. 

162: The first term

163:  represents the total turbulent flow energy $E$, order of magnitude wise.

164: Therefore, in general,

165: \be

166: \dot E(t) = -\eps (t) \sim {U\over L} E(t) + {U^2 \over L^2} \nu .

167: \label{eq2}

168: \ee

169: Imagine now a short intense kick of time $\Delta t_{kick} \ll \Delta t$

170: on the flow by rapidly moving the upper plate with $U$.

171: After this kick the initial energy $E_0$ increases according to (\ref{eq2})

172: \be

173: E_1 = E_0 +\left( E_0 + { U\nu \over L}\right) {U\over L} \Delta t_{kick}, 

174: \label{new1}

175: \ee

176: where we have assumed $\Delta t_{kick} \ll L/U$. 

177: Assume isotropic turbulence in the flow center and define a Reynolds number

178: {\footnote{Note that this is not the standard definition and

179: gives lower values for the laminar-turbulent transition than what one

180: is used to.}}

181: \be

182: Re(t) = {L u_{1,rms}(t) \over \nu} = \sqrt{2\over 3}{L \sqrt{ E(t)}\over \nu}.

183: \label{eq3}

184: \ee

185: Then eq.\ (\ref{new1}) translates to

186: \be

187: Re_1 = Re_0 \sqrt{ 1+ 2A + {Re_{lam}^2 \over Re_0^2}}

188: \label{eq4}

189: \ee

190: with the dimensionless kicking strength 

191: $A = {1\over 2} \Delta t_{kick} U/L

192:  \ll 1 $ 

193: and the ``laminar'' Reynolds number $Re_{lam} = {2\over 3} \Delta t_{kick}

194: U^2/\nu $. We choose this name because for very small $Re_0$ we have

195: $Re_1 = Re_{lam}$. For very large $Re_0 \gg Re_{lam}$ we have

196: $Re_1 = (1+A) Re_0$.

197: Shear flow and the

198: ``derivation'' of

199: eq.\ (\ref{eq4}) are only thought of as a motivation;

200: there will be many other experimental situations where the energy

201: input roughly corresponds to a law of type (\ref{eq4}).

202: 

203: {\underline{(ii) Decay:}}

204: In ref.\ \cite{loh94a} we calculated how the turbulent activity decays within a

205: time $ t$ for flow with fixed external length scale $L$. The

206: calculation was based on Effinger and Grossmann's variable range mean field

207: theory of turbulence \cite{eff87} in which viscous subrange and inertial

208: subrange can be treated equally well. The result of ref.\ \cite{loh94a} is that

209: for given initial Reynolds number $Re_i$ the time dependence of $Re( t)$

210: (defined as in eq.\ (\ref{eq3}))

211: is determined by the inverse function of

212: \be

213: { t (Re) \over \tau } = {3\over \cinf} \left[ F(Re) - F(Re_i)\right]

214: \label{eq5}

215: \ee

216: where $\tau = L^2 /\nu $ and $F(Re)$ is given by

217: \begin{eqnarray}

218: F(Re) &=& {1\over 2Re^2} \left\{ -\gamma + \sqrt{\gamma^2 + Re^2} \right\}

219: \nonumber \\

220: &+& {1\over 2\gamma} \log

221: \left\{ {\gamma + \sqrt{\gamma^2 + Re^2} \over Re} \right\}.

222: \label{eq6}

223: \end{eqnarray}

224: $\cinf = (6/b)^{3/2}$ (in this theory) is the dimensionless energy

225: dissipation in the large Reynolds number limit,

226: $\gamma = 9/\cinf$, and $b$ is the Kolmogorov constant \cite{my75} which is the

227: only free parameter in the theory of ref.\ \cite{loh94a}.

228: From experiment

229: \cite{sre95}

230: $b=6.0$, a value which we take in all calculations here. Consequently,

231: $\cinf = 1.0$ and $\gamma = 9.0$. 

232: Rather than the Reynolds number eq.\

233: (\ref{eq3}) one could also give the Taylor-Reynolds number $\rel$

234: \cite{loh94a}, 

235: \be

236: \rel = \sqrt {15 Re^2 \over \cinf \left( \gamma + \sqrt{\gamma^2

237: + Re^2}\right)}.

238: \label{eq7}

239: \ee

240: 

241: 

242: Eqs.\ (\ref{eq4}) - (\ref{eq6}) with $t=\Delta t$

243: (the time between successive kicks) 

244: fully define the present model. The two main

245: physical

246: parameters in the model are the kicking strength $A$ and the kicking frequency

247: $f=1/\Delta t$. The third physical parameter is $Re_{lam}$, the minimal

248: Reynolds number after a kick. We pick $Re_{lam } = 1$ throughout.

249: 

250: 

251: 

252: 

253: %caption1

254: \begin{figure}[htb]

255: \setlength{\unitlength}{1.0cm}

256: \begin{picture}(6,8.5)

257: \put(-0.0,1.0)

258: {\psfig{figure=fig1.ps,width=9cm,angle=-90}}

259: \end{picture}

260: \caption[]{

261: Time series $Re(t)$ for fixed $A=0.1$ and three kicking frequencies

262: $f=10/\tau$,

263: $f=100/\tau$, and $f=500/\tau$, bottom to top. 

264: The

265: corresponding (upper) saturation levels are

266: $Re_u^{sat} = 1.3 $,

267: $Re_u^{sat} = 17.3$, and

268: $Re_u^{sat} = 136$, respectively. 

269: }

270: \label{reoft}

271: \end{figure}

272: 

273: 

274: Figure \ref{reoft} shows $Re(t)$ for fixed $A$ and three different kicking

275: frequencies $f$ for the initial Reynolds number $Re_0 = Re_{lam} =1$. 

276: During each

277: cycle there is a kick $Re_0 \to  Re_1$ 

278: according to eq.\ (\ref{eq4}) 

279: and a subsequent decay according to

280: eq.\ (\ref{eq5}).

281: Overall, there is

282: growth up to some saturation level $Re^{sat}(A,f)$, achieved after

283: $t_{sat} (A, f)$. In this saturation state energy input and loss through decay

284: in $\Delta t$ balance and the degree of excitation fluctuates between

285: a lower saturation level

286: $Re_l^{sat}$ and an upper saturation level

287: \be

288: Re_u^{sat} = Re_l^{sat} \sqrt{1+2A +(Re_{lam}/Re_l^{sat})^2}.

289: \label{upper}

290: \ee

291: 

292: The (lower) level of saturation

293: $Re_l^{sat}(A,f)$ is given through the implicit 

294: equation

295: \be

296: {1\over \tau f} = {3\over \cinf} \left[ 

297: F\left( {Re_l^{sat} }\right)

298: - F\left(

299:  Re_l^{sat} \sqrt{1+2A +\left({Re_{lam}\over Re_l^{sat}}\right)^2}

300: \right)\right].

301: \label{eq8}

302: \ee

303: For large

304:  $Re_u^{sat}$, $Re_l^{sat} \gg \gamma$ one has the explicit result

305: \be

306: Re_u^{sat} = {3\tau f \over \cinf } (\sqrt{1+2A} -1) 

307: \approx {3\tau \over \cinf } Af.

308: \label{eq9}

309: \ee

310: In figure \ref{resat} we show a log-log plot of $Re_{l,u}^{sat}$ as a function

311: of $f$ for different kicking strength $A$. Two  regimes are

312: seen:

313: (i) For small $Re_{l,u}^{sat} \aleq \gamma$

314: a laminar regime 

315: in which after the decay

316: the kick always brings back the level of excitation to the

317: laminar value $Re_{lam }=1$.

318: (ii) 

319: For large $Re_{l,u}^{sat}\ageq \gamma$

320: we have  a turbulent scaling regime with 

321: $Re_u^{sat} \propto f$. 

322: The transition from the laminar to the turbulence regime

323: takes place around

324: \be

325: f_{trans} = {3\over \tau A}

326: \label{transition}

327: \ee

328:  Similarily, 

329: these two regimes are also seen in 

330: fig. \ref{resat_of_a} where we plotted 

331: $Re_{l,u}^{sat}$ vs $A$. 

332: In the turbulent regime 

333:  for large $Re_{l,u}^{sat}$ it is 

334: $Re_{l,u}^{sat} \propto A$. 

335: 

336: 

337: 

338: %caption1

339: \begin{figure}[htb]

340: \setlength{\unitlength}{1.0cm}

341: \begin{picture}(6,8.0)

342: \put(-0.,1.0)

343: {\psfig{figure=fig2.ps,width=9cm,angle=-90}}

344: \end{picture}

345: \caption[]{

346: Saturation level $Re_l^{sat}$ (lower curve of pair)

347: and $Re_u^{sat}$ (upper curve of pair) 

348: as a function of $f$ for three

349: different kicking strengths

350: $A=0.01$ (solid),

351: $A=0.1$ (dashed), and 

352: $A=0.5$ (dashed-dotted), 

353: bottom to top. Below $Re_u^{sat} \sim \gamma =9.0$ the

354: excited state is laminar, above $Re_u^{sat} \sim 9$ it is turbulent. 

355: }

356: \label{resat}

357: \end{figure}

358: 

359: %caption1

360: \begin{figure}[htb]

361: %\begin{figure}[p]

362: \setlength{\unitlength}{1.0cm}

363: \begin{picture}(6,7.5)

364: \put(0.0,1.0)

365: {\psfig{figure=fig3.ps,width=9cm,angle=-90}}

366: \end{picture}

367: \caption[]{

368: $Re_u^{sat}$ and 

369: $Re_l^{sat}$ 

370: for $f=10$, $10^2$, $10^3$, $10^4$, bottom to top. 

371: The increasing difference between

372: $Re_u^{sat}$ and 

373: $Re_l^{sat}$ at the right edge of the figure has its origin in

374: the breakdown of the requirement $A\ll 1$. 

375: }

376: \label{resat_of_a}

377: \end{figure}

378: 

379: 

380: 

381: We now come to the important question of 

382: how 

383: intermittency effects \cite{fri95} change the exponents calculated within

384: this mean field theory. 

385: In refs.\ \cite{gro95}

386: intermittency effects have been included into the mean field theory

387: of ref.\ \cite{loh94a} on a phenomenological basis. 

388: One possibility for their effect is  that the dimensionless

389: energy dissipation rate $c_\eps$

390: becomes slightly Reynolds number dependent even

391: in the large Reynolds number limit \cite{gro95},

392: $c_\eps \propto Re^{-\kappa}$,

393: with $\kappa = (9/8) \delta \zeta_2 /(1+3\delta\zeta_2/8)$. Here,

394: $\delta\zeta_2 \approx 0.03$ is the experimentally found deviation from

395: the classical scaling

396: exponent $\zeta_2 = 2/3$ of the second order velocity structure function.

397: The consequences of this small ($\kappa \approx 0.03$) scaling correction

398: can straightforwardly be embodied in the present mean field approach

399: to periodically kicked turbulence. The  result is that in the turbulent

400: regime the saturation level now obeys

401: \be

402: Re_{l,u}^{sat} \propto (Af)^{1/(1-\kappa)}

403: \label{interm}

404: \ee

405: rather than eq.\ (\ref{eq9}).

406: Equation (\ref{interm}) may

407: offer a

408: new and independent way to experimentally determine

409: intermittency exponents. 

410: 

411: Another effect related to intermittency

412: is the following: 

413: We expect the total energy to build up 

414: again and again over time scales larger than $\Delta t$

415: and then to suddenly drop because a energy pulse is traveling downscale.

416: Such behavior has  been observed in numerical simulations 

417: of periodically remeshed homogeneous shear flow \cite{pum96}

418: and in simulations of periodically kicked shell modells of turbulence

419: [results to be publihsed]. 

420: As based on a mean field theory the model of this paper

421: is only applicable to the

422: mean energy and not to these fluctuations. 

423: 

424: 

425: 

426: 

427: We now come back to an experimental realization.

428: It will be easier to 

429: perform experiments in a closed system 

430: rather than in a channel flow.

431: A particular suited experimental setup for periodically kicked

432: turbulence would be the flow in a cylinder between two counterrotating

433: disks \cite{lab96,zoc94}. Also, Rayleigh-Benard convection

434: may be well suited. Here, as an example we take Taylor-Couette flow 

435: \cite{lew99}: 

436: If the radii of the inner and outer cylinder are

437: similar, the energy input will still roughly follow eq.\ (\ref{eq4}).

438: Take water as fluid which has $\nu = 10^{-6}m^2/s$ and take $L=1cm$.

439: Then $\tau =100s$. Realistically achievable kick-strengths would be

440: $A= \Delta t_{kick} U /(2L) \sim  0.1s \cdot 0.1 m/s / (2\cdot 1cm) = 0.5$,

441: e.g., the whole range $A<1$ in which the theory is applicable. 

442: For this value the 

443: scaling regime sets in at 

444: $f_{trans} = 0.06Hz$. 

445: Roughly two  decades of scaling are necessary to explore

446: intermittency corrections

447: and to distinguish between eqs.\ (\ref{eq9}) and (\ref{interm}). 

448: I.e., one has to go up 

449: to frequencies of around $6Hz$ which should be achievable. 

450: At these relatively large kicking frequencies measurements

451: can only reveal {\it instantaneous} values. To get statements

452: on the averaged quantities dealt with in this paper, ensemble

453: averaging is necessary. This is best done by repeatedly probing

454: the flow at some fixed phase after the respective kick. Averaging

455: over the results at phase $0^+$ will give $Re_u^{sat}$, averaging 

456: over the results at phase $\Delta t^-$ will give $Re_l^{sat}$, etc.

457: 

458: 

459: 

460: 

461: We suggest to perform a periodically kicked turbulence 

462: experiment and to measure $Re^{sat}$ as a function of both $A$ and $f$.

463: To our knowledge it  would be one of the first

464: ones with some active control on the type of forcing.

465: Immediate questions to ask

466: are:

467: Does the level of saturation for large $Re$ indeed only depend on the

468: product $Af$ as suggested by eqs.\ (\ref{eq9}) and (\ref{interm})

469: or do boundary effects carry on into the strongly turbulent central regime and

470: cause a more subtle relation? If so, an application of the so commonly used

471: volume forcing for turbulence becomes more questionable. 

472: Do the (scaling) relations between the quantities

473: introduced in this paper, e.g., $Re_{l,u}^{sat} (f, A)$ 

474: offer a

475: new way to measure intermittency effects?

476: What modifications arise if forcing and decay do not decouple

477: as assumed in this simple model?

478: 

479: 

480: \vspace{0.5cm}

481: 

482: 

483: 

484: \noindent

485: {\bf Acknowledgements:}

486: The author thanks L. Biferale, M. Brenner, B. Eckhardt, 

487: and F. Toschi for

488: discussions. 

489: The work is part of the research  program of the Stichting voor 

490: Fundamenteel Onderzoek der Materie (FOM), which is financially supported 

491: by the Nederlandse  Organisatie voor Wetenschappelijk Onderzoek (NWO).

492: This research was also supported in part 

493: by the European Union under contract HPRN-CT-2000-00162 and 

494: by the National Science

495: Foundation under Grant No. PHY94-07194 and we thank the Institute

496: of Theoretical Physics in Santa Barabara for its hospitality.

497: 

498: 

499: \vspace{0.5cm}

500: 

501: \noindent

502:  e-mail address:

503: lohse@tn.utwente.nl

504: 

505: 

506: 

507: %\bibliography{literatur,sl_literatur}

508: 

509: \begin{thebibliography}{10}

510: 

511: \bibitem{far31}

512: M. Faraday, Phil. Trans. Roy. Soc. London {\bf 52},  319  (1831).

513: 

514: \bibitem{cru94}

515: L.~A. Crum, Physics Today {\bf 47},  22  (1994);

516: B.~P. Barber {\it et~al.}, Phys. Rep. {\bf 281},  65  (1997). 

517: 

518: \bibitem{pum96}

519: A. Pumir, Phys. Fluids {\bf 8}, 3112 (1996). 

520: 

521: 

522: \bibitem{loh94a}

523: D. Lohse, Phys. Rev. Lett. {\bf 73},  3223  (1994).

524: 

525: \bibitem{smi93a}

526: M. Smith, R.~J. Donelly, N. Goldenfeld, and W.~F. Vinen, Phys. Rev. Lett. {\bf

527:   71},  2583  (1993).

528: 

529: \bibitem{lab96}

530: R. Labbe, J.~F. Pinton, and S. Fauve, J. Phys. II France {\bf 6},  1099

531:   (1996).

532: 

533: \bibitem{cam97b}

534: R. Camussi, S. Ciliberto, and C. Baudet, Phys. Rev. E {\bf 56},  6181  (1997).

535: 

536: \bibitem{geb95}

537: {Th.} Gebhardt, S. Grossmann, M. Holthaus, and M. L\"ohden, Phys. Rev. E {\bf

538:   51},  360  (1995).

539: 

540: \bibitem{eff87}

541: H. Effinger and S. Grossmann, Z. Phys. B {\bf 66},  289  (1987).

542: 

543: \bibitem{my75}

544: A.~S. Monin and A.~M. Yaglom, {\em Statistical Fluid Mechanics} (The MIT Press,

545:   Cambridge, Massachusetts, 1975).

546: 

547: \bibitem{sre95}

548: K.~R. Sreenivasan, Phys. Fluids {\bf 7},  2778  (1995).

549: 

550: \bibitem{fri95}

551: U. Frisch, {\em Turbulence} (Cambridge University Press, Cambridge, 1995).

552: 

553: \bibitem{gro95}

554: S. Grossmann,

555: Phys. Rev. E {\bf 51},  6275  (1995);

556: G. Stolovitzky and K. R. Sreenivasan,

557: Phys. Rev. E {\bf 52},  3242  (1995).

558: 

559: 

560: \bibitem{zoc94}

561: G. Zocchi, P. Tabeling, J. Maurer, and H. Willaime, Phys. Rev. E {\bf 50},

562:   3693  (1994); for a review on this type of turbulence see 

563: P.~J. Zandbergen and D. Dijkstra, Ann. Rev. Fluid. Mech. {\bf 26},  137

564:   (1987).

565: 

566: \bibitem{lew99}

567: G. S. Lewis and H. L. Swinney, Phys. Rev. E {\bf 59}, 5457 (1999).

568: 

569: 

570: 

571: \end{thebibliography}

572: 

573: 

574: 

575: 

576: \end{document}

577: 

578: