1: %\documentclass{elsart}

2: \documentclass[doublespacing]{elsart}

3: \usepackage{epsfig}

4: 

5: \begin{document}

6: 

7: % ------------------------------

8: \begin{frontmatter}

9: 

10: \title{Multiscale SOC in turbulent convection}

11: \author[label1]{K.R. Sreenivasan\corauthref{col1}}

12: \author[label1,label2]{A. Bershadskii}

13: \author[label1]{J.J. Niemela}

14: \address[label1]{The Abdus Salam International Centre for Theoretical

15: Physics, Strada Costiera 11, I-34100 Trieste, Italy}

16: \address[label2]{ICAR, P.O. Box 31155, Jerusalem 91000, Israel}

17: \corauth[col1]{Corresponding author: K.R. Sreenivasan, E-mail:

18: krs@ictp.trieste.it.}

19: 

20: \begin{abstract}

21: Using data obtained in a laboratory thermal convection experiment

22: at high Rayleigh numbers, it is shown that the multiscaling

23: properties of the observed mean wind reversals are quantitatively

24: consistent with analogous multiscaling properties of the

25: Bak-Tang-Wiesenfeld prototype model of self-organized criticality

26: in two dimensions.

27: 

28: \end{abstract}

29: 

30: \begin{keyword}

31: convection \sep multiscaling \sep SOC \sep turbulence \PACS

32: 47.27.Te \sep 05.65.+b \sep 47.27.Jv

33: \end{keyword}

34: 

35: \end{frontmatter}

36: %\newpage

37: % ----------------------------

38: 

39: % main text

40: \section{Introduction}

41: \label{1} The existence of well-organized large-scale motion in

42: turbulent thermal convection is an intriguing recent discovery. It

43: has been under active investigation for some time, both

44: theoretically and experimentally (see \cite{kh}-\cite{vc} and

45: references therein). At least in a convection apparatus whose

46: aspect ratio (= the ratio of the diameter to the height) is unity,

47: the large-scale motion is in the form of a persistent circulation

48: (on the average), its physical extent being of the order of the

49: size of the apparatus itself. This is often called the ``mean

50: wind". The mean wind evolves with the Rayleigh number (which can

51: be regarded as an external tunable parameter in the flow), and

52: seems to asymptote to a well-defined shape only beyond Rayleigh

53: numbers of the order of $10^{10}$. However, beginning more or less

54: at this Rayleigh number, at least for three or so further decades,

55: the mean wind undergoes abrupt and apparently irregular reversals

56: of direction. These reversals are the object of the study here.

57: 

58: The measurements to be analyzed were obtained at a Rayleigh number

59: of $1.5 \times 10^{11}$ in thermal convection occurring in a

60: closed container of circular cross-section of 50 cm diameter and

61: aspect ratio unity. The working fluid was cryogenic helium gas.

62: The same data have been analyzed in the past \cite{sbn}, where the

63: fluid dynamical origin of reversals was proposed to be the

64: imbalance between buoyancy and friction (with inertia playing a

65: secondary role). It was further suggested that a stochastic

66: ``change of stability" occurs between two metastable states

67: corresponding to the opposite direction of the wind. The main

68: physical quantity analyzed in \cite{sbn} was the interval $\tau$

69: (the life-time of the metastable states, or the interval between

70: reversals). The time of actual switching between the two states

71: was short on the scale of some average measure of $\tau$, so the

72: wind reversals could be regarded as abrupt. The last property

73: (perhaps also the precise details of reversals themselves) is most

74: likely dependent on the specific boundary conditions of the

75: experiment, but it is thought that the statistical properties of

76: the duration times $\tau$ are insensitive to these details. The

77: analysis of \cite{sbn} indicated a general dynamic mechanism

78: similar to self-organized criticality (SOC) \cite{bak}-\cite{ms}.

79: The particular emphasis of this paper is the elaboration of this

80: analogy.

81: 

82: Self-organized criticality occurs through a nonlinear feedback

83: mechanism. There could be many possible scenarios of SOC in our

84: system where numerous plumes and jets are developed as a result of

85: boundary layer and thermal instabilities, all of which are

86: embedded in a background of strong turbulent fluctuations

87: prevalent in the core. It was suggested recently for plasma

88: turbulence (see Refs. \cite{dh},\cite{cnld}) that instabilities

89: governed by a threshold may lead to self-organized criticality by

90: producing transport events at all scales (avalanches). These

91: avalanches are due to local accumulation of energy, leading to an

92: increasing gradient. Once the gradient exceeds an appropriate

93: threshold, a burst of activity, which expels the accumulated

94: energy, ensues. This process can be renewed, much like a domino

95: effect, leading to a large transport event. Specific conditions of

96: the thermal convection in the container make these states

97: metastable and produce random reversals.

98: 

99: 

100: \section{Multiscaling properties of the mean wind and SOC}

101: \label{2} The probability density function (PDF) of the life-times

102: of the metastable states, $\tau$, observed in the laboratory data

103: on the wind, exhibit the characteristic power-law

104: $$

105: p(\tau) \sim \tau^{-1},  \eqno{(1)}

106: $$

107: as shown in figure 1.

108: 

109: \begin{figure}

110: \epsfig{file=fig1.eps,width=4.5in}

111: \vspace{-6cm}\caption{\footnotesize The probability density

112: function of the life-times of the metastable states of the wind

113: (in log-log scales). The dashed line indicates the power law

114: (1).}\vspace{1cm}

115: 

116: \end{figure}

117: Different physical mechanisms can lead to this power law. It could

118: be the ``exchange of stability" under turbulent noise effect

119: \cite{sbn}, or it could be the SOC as mentioned in the

120: Introduction. These mechanisms differ substantially. The

121: differences can be ascribed in part, almost trivially, to the lack

122: of long time correlations for the avalanches in the ``monoscale"

123: SOC models, leading to the inability of SOC to incorporate the

124: far-from-equilibrium characteristics of hydrodynamic systems such

125: as turbulent convection. This trivial distinction between the two

126: mechanisms disappears if we consider the two-dimensional

127: Bak-Tang-Wiesenfeld (BTW) model \cite{btw},\cite{dhar}, which

128: obeys a specific form of multiscaling for the PDF of several

129: avalanche measures \cite{tms},\cite{ms1},\cite{ms}. This model has

130: bursts with long-time correlation, and other turbulence-like

131: intermittent properties, if studied on the time scales of its

132: waves (see below). The question now is this: Can one still

133: distinguish between turbulence and the multiscaling SOC in this

134: situation? Or, are there still significant differences between the

135: model and the specific physical phenomenon? We shall address this

136: issue briefly in this paper, using as example the life-time $\tau$

137: of the metastable states of the mean wind.

138: 

139: The BTW prototype model of self-organized criticality is defined

140: on a square lattice \cite{dhar}. The number of ``grains" stacked

141: on a given site $i$ is denoted by $z_i=0,1,2,\ldots$ If $z_i<4$,

142: $\forall i$, the configuration is stable by construction. A random

143: site $k$ is selected and a grain is added to it, thus increasing

144: $z_k$ to $z_k+1$. If, in the process, $z_k \ge 4$, immediately

145: $z_k$ is reduced to $z_k-4$. This is called ``toppling". The

146: expelled four grains are received one each by the four nearest

147: neighbor site of $k$. The grains disappear if $k$ is a boundary

148: site. It is clear that the toppling at site $k$ may lead to

149: topplings at the next microscopic time step, and so on. Thus, an

150: avalanche made by a total number $s\geq 0$ of topplings occurs

151: before a new stable configuration is reached, and a new grain is

152: added to a random site. After many additions the system reaches a

153: stationary critical state in which the properties of the

154: avalanches are sampled.

155: 

156: A key notion of the BTW model is that of wave decomposition of

157: avalanches \cite{ikp}. The first wave is obtained as the set of

158: all topplings which can take place as long as the site of addition

159: is prevented from a possible second toppling. The second wave is

160: constituted by the topplings occurring after the second toppling

161: of the additional site takes place and before a third one is

162: allowed, and so on. The total number of topplings in an avalanche

163: is the sum of those of all its waves.  The wave size, s, has the

164: power-law PDF

165: $$

166: p(s)\sim s^{-1}.   \eqno{(2)}

167: $$

168: 

169: Returning now to convection, as mentioned in the Introduction, the

170: times taken to switch from one metastable state to the other are

171: very short in comparison with the interval between switchings; in

172: addition, the magnitude of the mean thermal wind velocity is

173: approximately {\it constant} during the metastable events.

174: Therefore, the size of the events in the experiment (in SOC terms)

175: is proportional to their duration; that is, $s \sim \tau$ (compare

176: equations (1) and (2), \cite{klgp}).

177: 

178: \begin{figure}

179: \epsfig{file=fig2.eps,width=4.5in}

180: \vspace{-5cm}\caption{\footnotesize Moments of the ``local

181: dissipation rate" of the metastable states duration times (or, in

182: this case, their sizes). The straight lines (the best fit) are

183: drawn in order to indicate (in log-log scales) the multiscaling

184: given by (4).}\vspace{0.6cm}

185: 

186: \end{figure}

187: 

188: \begin{figure}

189: \epsfig{file=fig3.eps,width=4.5in}

190: \vspace{-6cm}\caption{\footnotesize Multiscaling exponents from

191: (4) for the mean wind (circles) and for the two-dimensional BTW

192: prototype model of SOC (triangles).}\vspace{1cm}

193: 

194: \end{figure}

195: 

196: With this identification, one can take the analogy a step further.

197: A characteristic similar to local dissipation rate in turbulence,

198: $$

199: \epsilon_t=\sum_{k=1}^t (s_{k+1}-s_k)^2 /t,  \eqno{(3)}

200: $$

201: was introduced in the Ref. \cite{ms} in order to characterize

202: multiscaling properties of SOC. The multiscaling (if it exists)

203: has the form

204: $$

205: \frac{\langle \epsilon_t^p \rangle}{\langle \epsilon_t \rangle^p}

206: \sim t^{-\mu_p}.      \eqno{(4)}

207: $$

208: Such multiscaling was observed in Ref. \cite{ms} for the

209: two-dimensional BTW prototype model of the SOC. Since $s \sim

210: \tau$ in our case (see above and \cite{klgp}), we calculated the

211: analogous ``local dissipation rate" for the duration $\tau$

212: obtained in our experiment. The multiscaling and corresponding

213: scaling exponents $\mu_p$, given by (4), are shown in figures 2

214: and 3, respectively. Circles in figure 3 correspond to the

215: exponents computed for our data whereas the triangles correspond

216: to computations performed in \cite{ms} for the two-dimensional BTW

217: model of SOC.

218: 

219: \section{Concluding remarks}

220: \label{3}

221: 

222: The fluid flow here has many features such as jets, boundary

223: layers (both attached and detached), plumes, and so forth. These

224: are hydrodynamical entities which, in the end, may have to be

225: explained through their hydrodynamic origin. Amidst this

226: complexity, however, the comparison of the multiscaling exponents

227: shown in figure 3 suggests that the multiscale SOC is a possible

228: model for the observed reversal of the thermal mean wind. We might

229: ask: Is the mechanism in the flow indeed the same as that of

230: two-dimensional BTW, or does the agreement of multiscaling

231: exponents $\mu_p$ shown in figure 3 characterize some class of the

232: multiscale SOC universality? In some sense, the first question

233: looks for more detailed dynamical understanding, while the latter

234: looks for a statistical analogy. We cannot answer the first

235: question with any certainty yet, but the answer to the second

236: question seems to be in the affirmative.

237: 

238: \noindent{\bf Acknowledgment}

239: 

240: For sometime soon after SOC was invented, Per Bak was actively

241: proposing SOC as the leading mechanism to explain turbulence. By

242: the time he wrote his book \cite{bak}, however, Per seemed to have

243: withdrawn that emphasis. One imagines that this may have resulted

244: at least partly from the cautionary notes he constantly received

245: from the likes of one of us (KRS). It is therefore especially

246: fitting to point out a quantitative, albeit small, feature of

247: turbulence that bears similarity to SOC, even if the implications

248: of this observation are not yet fully clear. We appreciate the

249: efforts of the organizers in holding the meeting and arranging

250: this publication honoring the ebullient character of Per Bak. We

251: are pleased to be a part of it, and are sad at the same time that

252: his life was terminated prematurely.

253: 

254: \begin{thebibliography}{00}

255: 

256: \bibitem{kh} R. Krishnamurti and L.N. Howard, Proc. Natl. Acad. Sci. U.S.A.

257: {\bf 78} (1981) 1981.

258: \bibitem{swl} M. Sano, X.-Z. Wu, and A. Libchaber,

259: Phys. Rev. A {\bf 40} (1989) 6421.

260: \bibitem{cas}B. Castaing, G. Gunaratne, F. Heslot, L. Kadanoff,

261: A. Libchaber, S. Thomae, X.-Z. Wu, S. Zaleski and G. Zanetti, J.

262: Fluid Mech. {\bf 204} (1989) 1.

263: \bibitem{ccl} S. Ciliberto, S. Cioni and C. Laroche, Phys. Rev.

264: E {\bf 54} (1996) R5901.

265: \bibitem{ccs} S. Cioni, S. Ciliberto and J. Sommeria, Dyn. Atmos. Oceans,

266: {\bf 24} (1996) 117.

267: \bibitem{nature} J.J. Niemela, L. Skrbek and R.J. Donnelly {\it Nature} {\bf 404}, (2000)

268: 837.

269: \bibitem{sl} S. Grossmann and D. Lohse, J. Fluid Mech. {\bf 407} (2000) 27.

270: \bibitem{qyt} X.-L. Qiu, S.-H. Yao and P. Tong, Phys. Rev. E {\bf 61} (2000) R6075.

271: \bibitem{qt} X.-L. Qui and P. Tong, Phys. Rev. Lett., {\bf 87}

272: (2001) 094501.

273: \bibitem{kad} L. Kadanoff, Phys. Today {\bf 54} (2001) 34.

274: \bibitem{nssd} J.J. Niemela, L. Skrbek, K.R. Sreenivasan and R.J.

275: Donnelly, J. Fluid Mech. {\bf 449} (2001) 169.

276: \bibitem{sbn} K.R. Sreenivasan, A. Bershadskii and J.J. Niemela,

277: Phys. Rev. E {\bf 65} (2002) 056306.

278: \bibitem{ns1} J.J. Neimela and K.R. Sreenivasan, Physica A {\bf 315} (2002) 203.

279: \bibitem{ns2} J.J. Niemela and K.R. Sreenivasan, Europhys. Lett.

280: {\bf 62} (2003) 829.

281: \bibitem{vc} R. Verzicco and R. Camussi, J. Fluid Mech. {\bf 447} (2003) 19.

282: \bibitem{bak} P. Bak, How Nature Works (Copernicus, New York,

283: 1996).

284: \bibitem{jen} H.J. Jensen, Self-organized Criticality (Cambridge University

285: Press, Cambridge, England, 1998).

286: \bibitem{btw} P. Bak, C. Tang and K. Wiesefeld, Phys. Rev. Lett.

287: {\bf 59} (1987) 381.

288: \bibitem{dhar} D. Dhar, Physica A {\bf 263} (1999) 4.

289: \bibitem{tms} C. Tebaldi, M. De Menech and A.L. Stella, Phys. Rev.

290: Lett. {\bf 83} (1999) 3952.

291: \bibitem{psv} R. Pastor-Satorras and A. Vespignani, Phys. Rev. E

292: {\bf 61} (2000) 4854.

293: \bibitem{ms1} M. De Menech and A.L. Stella, Phys. Rev. E {\bf 62} (2000) R4528.

294: \bibitem{ms} M. De Menech and A.L. Stella, Physica A {\bf 309}

295: (2002) 289 (see also cond-mat/0103601).

296: \bibitem{dh} P.H. Diamond and T.S. Hahm, Phys. Plasmas {\bf 2} (1995) 3640.

297: \bibitem{cnld} B.A. Carreras, D. Newman, V.E. Lynch and P.H. Diamond, Phys.

298: Plasmas {\bf 3} (1996) 2903.

299: \bibitem{ikp} E.V. Ivashkevich, D.V. Ktitarev, and V.B. Priezzhev,

300: Physica A {\bf 209} (1994) 347.

301: \bibitem{klgp} D.V. Ktitarev, S. L\"{u}beck, P. Grassberger and

302: V.B. Priezzhev, Phys. Rev. E {\bf 61} (2000) 81.

303: 

304: 

305: 

306: \end{thebibliography}

307: 

308: 

309: \end{document}

310: