1: \documentclass[prl,aps,twocolumn]{revtex4}
2: \usepackage{epsfig}
3: \usepackage{bm}
4: \newcommand{\B}[1]{{\bm{#1}}}
5: \newcommand{\C}[1]{{\mathcal{#1}}}
6: \renewcommand{\it}[1]{\textit{#1}}
7: \newcommand{\Onecol} {\begin{widetext} \onecolumngrid} %% 2 -> 1
8: \newcommand{\Twocol} {\end{widetext} \twocolumngrid} %% 1 -> 2
9: \newcommand{\ds}{\displaystyle}
10: \newcommand{\be}{\begin{equation}}
11: \newcommand{\ba}{\begin{array}}
12: \newcommand{\bea}{\begin{eqnarray}}
13: \newcommand{\bfi}{\begin{figure}}
14: \newcommand{\ee}{\end{equation}}
15: \newcommand{\ea}{\end{array}}
16: \newcommand{\eea}{\end{eqnarray}}
17: \newcommand{\efi}{\end{figure}}
18: \newcommand{\dV}{\delta u}
19: \newcommand{\T}{\theta}
20: \newcommand{\lp}{\left(}
21: \newcommand{\rp}{\right)}
22: \newcommand{\ra}{\right\rangle}
23: \newcommand{\la}{\left\langle}
24: \begin{document}
25: \title{Theory of concentration dependence
26: in drag reduction by polymers and of the MDR asymptote}
27: \author{Roberto Benzi$^{1}$, Emily S.C. Ching$^2$,
28: Nizan Horesh$^{2,3}$ and Itamar
29: Procaccia$^{2,3}$}
30: \affiliation{$^1$ Dip. di Fisica and INFM, Universit\`a ``Tor
31: Vergata", Via della Ricerca Scientifica 1, I-00133 Roma, Italy\\
32: $^2$ Dept. of Physics, The Chinese University of Hong Kong,
33: Shatin, Hong Kong\\ $^3$ Dept. of Chemical Physics, The Weizmann
34: Institute of Science, Rehovot, 76100 Israel}
35: %\pacs{47.27-i, 47.27.Nz, 47.27.Ak}
36: \begin{abstract}
37: A simple model of the effect of polymer concentration on the amount of drag reduction in turbulence
38: is presented, simulated and analyzed. The qualitative phase diagram of drag coefficient
39: vs. Reynolds number (Re) is recaptured in this model, including the theoretically elusive onset
40: of drag reduction and the Maximum
41: Drag Reduction (MDR) asymptote. The Re-dependent drag and the MDR are analytically
42: explained, and the dependence of the amount of drag on material parameters is rationalized.
43: \vskip 0.2cm
44: \end{abstract}
45: \maketitle
46: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
47: ``Drag reduction" refers to the intriguing phenomenon when the
48: addition of few tens of parts per million (by weight) of
49: long-chain polymers to turbulent fluids can bring about a
50: reduction of the friction drag by up to 80\%
51: \cite{75Vir,97VSW,00SW}. The phenomenon is well documented since
52: Toms discovered it accidentally in 1946 while studying the
53: degradation of polymers. The pioneering work of Virk
54: \cite{75Vir,97VSW} had systematized and organized a huge amount
55: of experimental information, but the fundamtental mechanism for
56: the phenomenon has remained under debate for a long time
57: \cite{69Lu,90Ge,00SW}. All the {\em experimental} and many of the
58: {\em numerical} \cite{97THKN,98DSB,00ACP,02ACLPP} investigations
59: of drag reduction focused on channel and pipe geometries;
60: recently however it had been discovered by {\em numerical
61: simulations} \cite{03ACBP} of model equations of viscoelastic
62: flows (like the FENE-P model) that drag reduction appears also in
63: homogeneous and isotropic turbulence when seeded with polymers.
64: This brought about a new focus to the search for the mechanism
65: for drag reduction, since the analysis of model equations without
66: wall effects should suffice to uncover a mechanism. Indeed, in a
67: recent paper \cite{03BDGP} the FENE-P equations were simplified
68: further to a shell model of viscoelastic flow which was shown to
69: exhibit drag reduction whose mechanism could be fully explored
70: analytically. In this Letter we present additional crucial
71: progress where we demonstrate and explain two of the most
72: prominent (and least understood) characteristics of drag
73: reduction, i.e. the onset [as a function of Reynolds number (Re)]
74: and the Maximal Drag Reduction (MDR) asymptote.
75: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
76: \begin{figure}
77: \centering
78: \includegraphics[width=0.50\textwidth]{MDRFig.1.eps}
79: \caption{Drag reduction in Prandtl-Karman coordinates
80: \cite{97VSW}. As a function of Re the drag exhibits a
81: (concentration independent) transition to drag reduction. The
82: amount of drag reduction depends on the concentration until the
83: asymptote denoted by MDR is reached. The Prandtl-Karman law is
84: the Re-dependent drag of the neat fluid. The numbers indicate
85: concentrations of the polymer additive in wppm.}
86: \label{cartoonnear}
87: \end{figure}
88: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
89:
90: To set up the issues we reproduce in Fig.1 a typical experimental
91: figure from Ref. \cite{97VSW} which refers to the dependence of
92: the friction (or drag) coefficient in pipe flows on Re. For a pipe
93: of radius $R$ and length $L$, with $\Delta p$, $\rho$ and $U$
94: being the pressure drop across $L$, the fluid density and the mean
95: velocity over a section, the drag coefficient $f$ reads
96: \begin{equation}
97: f=\frac{\Delta p}{\rho U^2}\frac{R}{L}\quad \text{(Pipe flow)}\ . \label{deff}
98: \end{equation}
99: ``Drag reduction" is tantamount to, say, an increase in the
100: throughput $U$ for a fixed pressure drop $\Delta p$ when polymer
101: is added to the working fluid. In Fig. 1 one sees that for low Re
102: there is no drag reduction: the drag coefficient of pure water is
103: unchanged by the addition of small concentration of polymers.
104: Then there is a sharp onset of drag reduction at a value of Re
105: that does not depend on the concentration. From this point on the
106: amount of drag reduction depends both on the concentration of the
107: polymer and on Re. It was shown by Virk however that the amount
108: of drag reduction asymptotes to an apparently universal curve that
109: cannot be exceeded by increasing the concentration further. This
110: asymptote is referred to as the MDR, and was claimed to be
111: insensitive to the nature of the polymer used in the experiments.
112: In spite of the ample experimental evidence, both the onset and
113: the existence of the MDR have not been theoretically understood.
114: In this Letter we wish to close this gap.
115:
116: Our strategy is to explore simulationally and analytically
117: simplified models of viscoelastic flows which in spite of the
118: simplification still represent the robust properties that we are
119: after. As is well known, viscoelastic flows are represented well
120: by hydrodynamic equations in which the effect of the polymer
121: enters in the form of a ``conformation tensor" $\B R(\B r,t)$
122: which stems from the ensemble average of the diadic product of
123: the end-to-end distance of the polymer chains. Flexibility and
124: finite extendability of the polymer chains are reflected by the
125: relaxation time $\tau$ and the Peterlin function $P(\B r,t)$
126: which appear in the equation of motion for $\B R$:
127: \begin{eqnarray}
128: \frac{\partial R_{\alpha\beta}}{\partial t}+(\B u\cdot \B \nabla)
129: R_{\alpha\beta}
130: &&=\frac{\partial u_\alpha}{\partial r_\gamma}R_{\gamma\beta}
131: +R_{\alpha\gamma}\frac{\partial u_\beta}{\partial r_\gamma}\nonumber\\
132: &&-\frac{1}{\tau}\left[ P(\B r,t) R_{\alpha\beta} -\rho_0^2
133: \delta_{\alpha\beta} \right]\label{EqR}\\
134: P(\B r,t)&&=(\rho_m^2-\rho_0^2)/(\rho_m^2 -R_{\gamma\gamma})
135: \end{eqnarray}
136: In these equations $\rho^2_m$ and $\rho^2_0$ refer to the maximal
137: and the equilibrium values of the trace $R_{\gamma\gamma}$. Since
138: in most applications $\rho_m\gg \rho_0$ the Peterlin function can
139: be also written approximately as $P(\B r,t)\approx 1/(1 -\alpha
140: R_{\gamma\gamma})$ where $\alpha=\rho_m^{-2}$. In its turn the
141: conformation tensor appears in the equations for fluid velocity
142: $\B u(\B r,t)$ as an additional stress tensor:
143: \begin{eqnarray}
144: &&\frac{\partial \B u}{\partial t}+(\B u\cdot \B \nabla) \B u=-\B \nabla p
145: +\nu_s \nabla^2 \B u +\B \nabla \cdot \B {\C T}+\B F\ , \label{Equ}\\
146: &&\B {\C T}(\B r,t) = \frac{\nu_p}{\tau}\left[\frac{P(\B r,t)}{\rho_0^2} \B
147: R(\B r,t) -\B 1 \right] \ .
148: \end{eqnarray}
149: Here $\nu_s$ is the viscosity of the neat fluid, $\B F$ is the
150: forcing and $\nu_p$ is a viscosity parameter which is related to
151: the concentration of the polymer, i.e. $\nu_p/\nu_s\sim c$ where
152: $c$ is the volume fraction of the polymer. Note that the tensor
153: field can be rescaled to get rid of the parameter $\alpha$ in the
154: Peterlin function, $\tilde R_{\alpha\beta}=\alpha
155: R_{\alpha\beta}$ with the only consequence of rescaling the
156: parameter $\rho_0$ accordingly. These equations were simulated on
157: the computer in a channel or pipe geometry, reproducing
158: faithfully the characteristics of drag reduction in experiments
159: \cite{97THKN,98DSB,00ACP}. It should be pointed out however that
160: even for present day computers simulating these equations is
161: quite tasking. We therefore simplify the model further.
162:
163: In developing a simple model we are led by the following ideas.
164: First, it should be pointed out that all the nonlinear terms
165: involving the tensor field $\B R(\B r,t)$ can be reproduced by
166: writing an equation of motion for a vector field $\B B(\B r,t)$,
167: and interpreting $R_{\alpha\beta}$ as the diadic product
168: $B_\alpha B_\beta$. The relaxation terms with the Peterlin
169: function are not automatically reproduced this way, and we need
170: to add them by hand. Second, we should keep in mind that the
171: above equations exhibit a generalized energy which is the sum of
172: the fluid kinetic energy and the polymer free energy. Led by
173: these consideration we write the following shell model
174: \cite{03BDGP}, which we refer to as the SabraP model:
175: \begin{eqnarray}
176: \frac{d u_n}{d t} &=& \frac{i}{3} \Phi_n (u,u) - \frac{i}{3}
177: \frac{\nu_p}{\tau} P(B)
178: \Phi_n (B,B) - \nu_s k^{2}_n u_n + F_n , \nonumber\\
179: \frac{d B_n}{d t} &=& \frac{i}{3} \Phi_n (u,B) - \frac{i}{3}
180: \Phi_n (B,u) - {1 \over \tau} P(B) B_n - \nu_B k_n^2 B_n,\nonumber\\
181: P(B) &=& {1 \over 1 - \sum_n B_n^* B_{n} } \ . \label{SP}
182: \end{eqnarray}
183: In these equations $u_n$ and $B_n$ stand for the Fourier amplitudes $u(k_n)$ and $B(k_n)$ of the two
184: respective vector fields, but as usual in shell model we take $n=0,1,2,\dots$ and the
185: wavevectors are limited to the set $k_n=2^n$. The nonlinear interaction terms take
186: the explicit form
187: \begin{eqnarray}
188: &&\Phi_n(u,B) = k_n\Big[(1-b) u_{n+2} B^*_{n+1} +(2+b) u^*_{n+1}
189: B_{n+2}\Big] \nonumber\\&&+ k_{n-1}\Big[(2b+1) u^*_{n-1}B_{n+1}-(1-b)u_{n+1}B^*_{n-1} \Big]\nonumber\\
190: &&+k_{n-2}\Big[(2+b)u_{n-1}B_{n-2}+(2b+1)u_{n-2}B_{n-1}\Big]\ ,
191: \end{eqnarray}
192: with an obvious simplification for $\Phi_n(u,u)$ and
193: $\Phi_n(B,B)$. Here $b$ is a parameter taken below to be $-0.2$.
194: In accordance with the generalized energy of the FENE-P model,
195: also our shell model has the total energy
196: \begin{equation}
197: E \equiv {1 \over 2} \sum_n |u_n|^2 - {1 \over 2} {\nu_p \over \tau}
198: \ln \left(1-\sum_n |B_n|^2\right)\ .
199: \end{equation}
200: The second term in the generalized energy contributes to the
201: dissipation a positive definite term of the form
202: $(\nu_p/\tau^2)P^2(B) \sum_n|B_n|^2$. With $\nu_p=0$ the first of
203: Eqs. \ref{SP} reduces to the well-studied Sabra model of
204: Newtonian turbulence \cite{98LPPPV}. As in the FENE-P case we
205: consider $\nu_p/\nu_s$ to be $c$. All the simulations below are
206: performed with a constant rate of energy input, choosing
207: $F_n=\phi/u^*_n$ for $n=0,1$ and zero otherwise.
208: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
209: \begin{figure}
210: \centering
211: \includegraphics[width=.5\textwidth]{MDRFig.2.eps}
212: \caption{Power spectra of the SabraP model (line) and the Sabra
213: model (dashed line) for $\phi=0.001$, $\nu_s=10^{-6}$ and
214: $\tau=0.4$. The dashed line with symbols represents the power
215: spectrum of the $B_n$ field.} \label{logspectrum}
216: \end{figure}
217: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
218:
219: In \cite{03BDGP} it was shown that this shell model exhibits drag
220: reduction, and the mechanism for the phenomenon was elucidated.
221: The basic phenomenon is exhibited well by the spectra of the
222: $u_n$ and $B_n$ fields which are presented at one value of the
223: parameters in Fig. \ref{logspectrum}. The spectra for the Sabra
224: model (dashed line) and the SabraP model (line) are compared for
225: the same amount of power input per unit time. The discussion
226: \cite{03BDGP} of the spectra revolves around the typical Lumley
227: scale $k_c$ which is determined by the condition \cite{69Lu}
228: \begin{equation}
229: u(k_c) k_c\approx \tau^{-1} \ . \label{defkc}
230: \end{equation}
231: For $k_n\gg k_c$ the decay time $\tau$ becomes irrelevant for the
232: dynamics of $B_n$. The nonlinear interaction between $u_n$ and
233: $B_n$ at these scales results in both of them having the same
234: spectral exponent which is also the same as that of the Sabra
235: model. The amplitude of the $u_n$ spectrum is however smaller in
236: the SabraP model compared to the Sabra case, since the $B_n$
237: field adds to the dissipation. On the other hand, for $k_n\ll
238: k_c$ the $B_n$ field is exponentially suppressed by its decay due
239: to $\tau$, and the spectral exponent of $u_n$ is again as in the
240: Sabra model. Drag reduction comes about due to the interactions at
241: length scales of the order of $k_c$ which force a strong tilt in
242: the $u_n$ spectrum there, causing it to cross the Sabra spectrum,
243: leading to an increase in the amplitude of the energy containing
244: scale. This is why the kinetic energy is increasing for the same
245: amount of power input, and hence drag reduction. Note that a very
246: similar spectral cross-over had been documented also for the
247: FENE-P model in channel flow simulations \cite{02ACLPP}
248:
249: The qualitative phenomena that we are about to explain in this Letter are
250: demonstrated in the simulational results presented in Fig. \ref{c-Re}.
251: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
252: \begin{figure}
253: \centering
254: \includegraphics[width=.5\textwidth]{MDRFig.3a.eps}
255: \includegraphics[width=.5\textwidth]{MDRFig.3b.eps}
256: \caption{Upper panel: Drag as a function of $\log_{10}$(Re)
257: including the laminar and the turbulent regimes. Both regimes
258: agree with the Eqs.(\ref{fsmallRe})-(\ref{flargeRe}). Lower panel:
259: Blow up of the turbulent regime. In both panels the upper
260: straight line indicates the drag of the neat fluid, whereas the
261: MDR is seen as the convergence of the drag data for large
262: concentrations. } \label{c-Re}
263: \end{figure}
264: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
265: Here we show the drag coefficient $f$ as a function of Re for the
266: Sabra and for the SabraP models for various values of the
267: concentration. The drag coefficient is computed in analogy to Eq.
268: (\ref{deff}) as
269: \begin{equation}
270: f\equiv \frac{\sum_n F_n u^*_n}{\Big(\sum_n |u_n|^2\Big)^{3/2}k_0} \ . \quad \text{(Our model)}
271: \end{equation}
272: We observe all the phenomena discovered by Virk: (i) For the
273: model of the neat fluid the drag has a laminar branch and a
274: turbulent branch, with a sharp transition between them. (ii) For
275: the model of the viscoelastic flow in the laminar region there is
276: no drag reduction; the laminar branch is not changed by the
277: addition of polymer with any concentration. (iii) Drag reduction
278: has an onset that is independent of the concentration of the
279: polymer. (iv) As the concentration increases the amount of drag
280: reduction increases, but (v) there exists an asymptote which is
281: not exceeded when the concentration is increased. In other words,
282: our simple model appears to reproduce extremely well the
283: phenomena that were uncovered in so many experiments as
284: summarized by Virk.
285:
286: Next we explain all these observations. First we rationalize the
287: Re-dependence of the friction factor in the Sabra model of the
288: neat fluid. For low Re the nonlinear terms $\Phi_n(u,u)$ are
289: negligible compared to the viscous term. Forcing only on the
290: largest scale $k_0$ we can evaluate,
291: \begin{equation}
292: \nu_s k_0^2 u_0 \approx F_0 \quad \rightarrow~f\sim
293: \frac{\nu_sk_0}{|u_0|}= {\rm Re}^{-1} \quad \text{Re small} \ .
294: \label{fsmallRe}
295: \end{equation}
296: For large Re we have the exact result \cite{98LPPPV} that the
297: third order correlation function $S_n^{(3)}\equiv \Im \langle
298: u_{n-1} u_n u^*_{n+1}\rangle =C \bar \epsilon /k_n$ with $\bar
299: \epsilon$ being the mean energy flux and $C$ a known constant
300: (the analog of the 4/5th law for Navier-Stokes turbulence). We
301: therefore expect the friction factor to tend to a constant value
302: for large Re (up to terms $\sim \log$Re),
303: \begin{equation}
304: f\sim Re^{0} \quad \text{Re large} \ . \label{flargeRe}
305: \end{equation}
306: Eqs. (\ref{fsmallRe}) and (\ref{flargeRe}) (which are the analogs
307: of the Prandtl-Karman law for pipe flows) are well borne out by
308: the data in Fig. \ref{c-Re} for the model of the neat fluid. The
309: laminar branch, which is exponential in these coordinates ($f \sim
310: \exp\{-\log({\rm Re})\})$, is unaffected by changing the
311: concentration $c$. The transition between the two branches is
312: expected when turbulence sets in, i.e. for Re such that the
313: dissipative terms just begin to be overwhelmed by the nonlinear
314: interactions. Thus point (i) is understood. Note that similar
315: arguments will hold for the FENE-P equations in homogeneous
316: flows. Points (ii) and (iii) are explained as follows; we said
317: above that drag reduction comes about due to the interaction
318: between the two dynamical fields at scale of the order of $k_c$.
319: Clearly, as long as $k_c$ exceeds the dissipative scale $k_d$ of
320: the the $u_n$ field, no interaction between the two field can be
321: of any significance. Since $k_d$ is of the order of $k_d\sim k_0
322: Re^{3/4}$, we can expect a concentration independent onset of
323: drag reduction when $k_c\approx k_d$. Using $u_n\sim
324: u_0(k_n/k_0)^{-1/3}$, $k_c$ can be estimated as $k_c\sim k_0
325: (\tau u_0k_0)^{-3/2}$, and we end up with a prediction for the
326: onset of drag reduction when
327: \begin{equation}
328: Re \approx (\tau u_0k_0)^{-2} \quad \text {Onset of drag reduction} \ .
329: \end{equation}
330: This prediction is well borne out by our simulations (due to the
331: space constraint we do not display simulations at different values
332: of $\tau$ and $k_0$). Again we point out that similar arguments
333: can be presented for the FENE-P model as well.
334:
335: Point (iv) is obvious - when the concentration increases the
336: mechanism discovered in \cite{03BDGP} comes into play. What
337: remains to explain is the asymptotic MDR. This also follows
338: directly from the analysis of the equations. Consider Eqs.
339: (\ref{SP}) for two values of the parameter $\nu_p$,
340: $\nu_p^{(1)}\ll \nu_p^{(2)}$, with $y^2=\nu_p^{(1)}/\nu_p^{(2)}$.
341: Rescaling $B_n$ according to $B_n=y\tilde B_n$, we see that the
342: Peterlin function tends to unity when $y\to 0$,
343: \begin{equation}
344: P(\tilde B)=\frac{1}{1- y^2\sum_n|\tilde B_n|^2} \to 1\ , \quad
345: \text {when}~y\to 0 \ .
346: \end{equation}
347: When $P(\tilde B)\approx 1$ the dynamical equation for $\tilde
348: B_n$ is independent of of $y$ due to its linearity, whereas the
349: $u_n$ equation remains independent due to the rescaling:
350: \begin{equation}
351: \frac{i}{3} \frac{\nu_p^{(2)}}{\tau} P(B) \Phi_n (B,B)\to
352: \frac{i}{3}\frac{\nu_p^{(1)}}{\tau} P(\tilde B) \Phi_n(\tilde
353: B_n, \tilde B_n)\ ,
354: \end{equation}
355: Thus increasing the concentration
356: brings to the dynamical equations to an asymptotic concentration invariant form and therefore
357: to an asymptotic MDR. For the last time we remark that similar rescalings are also
358: available in the FENE-P equations, making the points discussed here quite general for
359: any sensible model of viscoelastic flow.
360:
361: In summary, we have presented a simple model of drag reduction
362: for which the observed characteristics can be explained on the
363: basis of the equations of motion. It remains to go back to
364: channel and pipe simulations of the FENE-P equations to
365: demonstrate that the discussion presented above includes the main
366: phenomena observed also there.
367: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
368: \acknowledgments \vskip 0.5 cm This work was supported in part by
369: the European Commission under a TMR grant, the Minerva
370: Foundation, Munich, Germany, and the Naftali and Anna
371: Backenroth-Bronicki Fund for Research in Chaos and Complexity.
372: ESCC acknowledges the Hong Kong Research Grants Council for
373: support (CUHK 4046/02p).
374: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
375: \begin{thebibliography}{99}
376:
377: \bibitem{75Vir}
378: P.S. Virk, AIChE J. {\bf 21}, 625 (1975); Nature {\bf 253}, 109 (1975).
379: \bibitem{97VSW}
380: P.S. Virk, D.C. Sherma and D.L. Wagger, AIChE Journal, {\bf 43}, 3257 (1997).
381: \bibitem{00SW}
382: K. R. Sreenivasan and C. M. White, J. Fluid Mech. {\bf 409}, 149 (2000).
383: \bibitem{69Lu}
384: J. L. Lumley, Ann. Rev. Fluid Mech. {\bf 1}, 367 (1969).
385:
386: \bibitem{90Ge}
387: P.-G. de Gennes {\em Introduction to Polymer Dynamics}, (Cambridge, 1990).
388: \bibitem{97THKN}
389: J.M.J de Toonder, M.A. Hulsen, G.D.C. Kuiken and F.T.M Nieuwstadt, J. Fluid.
390: Mech {\bf 337}, 193 (1997).
391: \bibitem{98DSB}
392: C.D. Dimitropoulos, R. Sureshdumar and A.N. Beris, J. Non-Newtonian Fluid
393: Mech. {\bf 79}, 433 (1998).
394: \bibitem{00ACP}
395: E. de Angelis, C.M. Casciola and R. Piva, CFD Journal, {\bf 9}, 1 (2000).
396:
397:
398: \bibitem{02ACLPP}
399: E. De Angelis, C.M. Casciola, V.S. L'vov, R. Piva and I. Procaccia,
400: ``Drag reduction by polymers in turbulent channel flows: Energy redistribution between invariant
401: empirical modes", Phys. Rev. E, in press.
402:
403:
404: \bibitem{03ACBP}
405: E. de Angelis, C. Casciola, R. Benzi, and R. Piva, Phys. of
406: Fluids, submitted.
407:
408: \bibitem{03BDGP}
409: R. Benzi, E. De Angelis, R. Govindarajan and I. Procaccia, ``Shell Model for Drag Reduction with Polymer
410: Additive in Homogeneous Turbulence", Phys. Rev. E, submitted.
411:
412:
413: \bibitem{98LPPPV}
414: V.S. L'vov, E. Podivilov, A. Pomyalov, I. Procaccia and D.
415: Vandembroucq., Phys. Rev. E {\bf 58} 1811 (1998).
416:
417:
418: \end{thebibliography}
419: \end{document}
420:
421: