1: \documentclass[aps,showpacs,onecolumn,nofootinbib,byrevtex]{revtex4}
2: \usepackage{mathrsfs,amsmath,graphicx,natbib,bm}
3: \preprint{adahsdkh}
4: \begin{document}
5: \title{Single polymer dynamics in elongational flow and the confluent Heun equation}
6: \author{D. Vincenzi}
7: \email{Dario.Vincenzi@ds.mpg.de}
8: \affiliation{Max-Planck-Institut f\"ur Dynamik und
9: Selbstorganisation, 37077 G\"ottingen, Germany}
10: \affiliation{School of Mechanical \& Aerospace Engineering
11: and LASSP, Cornell University, Ithaca, NY 14853, USA}
12: \author{E. Bodenschatz}
13: \affiliation{Max-Planck-Institut f\"ur Dynamik und
14: Selbstorganisation, 37077 G\"ottingen, Germany}
15: \affiliation{School of Mechanical \& Aerospace Engineering
16: and LASSP, Cornell University, Ithaca, NY 14853, USA}
17: \pacs{05.10.Gg, 47.57.Ng, 02.30.Hq}
18: \begin{abstract}
19: We investigate the
20: non-equilibrium
21: dynamics of an isolated polymer in a stationary elongational flow.
22: We compute the relaxation time to the
23: steady-state configuration as a function of
24: the Weissenberg number. A strong increase of the relaxation time is
25: found around the coil--stretch transition, which is attributed to the large
26: number of polymer configurations. The relaxation dynamics of the
27: polymer is solved analytically in terms of a
28: central two-point connection problem for the singly confluent Heun equation.
29: \end{abstract}
30: \maketitle
31:
32: \centerline{\textit{J. Phys. A: Math. Gen.} \textbf{39}, 10691--10701 (2006)}
33: \centerline{\fontfamily{cmr}\selectfont
34: \tt{statcks.iop.org/JPhysA/39/10691}}
35:
36: \section{Introduction}
37: \label{sec:introduction}
38: Dilute polymer solutions exhibit physical behaviours that
39: distinguish them from ordinary Newtonian fluids. Even a small
40: polymer concentration can considerably change the
41: large-scale behaviour of the flow by enhancing viscosity and
42: reducing the turbulent drag. A comprehensive understanding
43: of hydrodynamical properties of polymer solutions is still
44: lacking despite the large number of
45: industrial applications (e.g. Sreenivasan and White 2000).
46: A starting point for the theoretical description of dilute polymer
47: solutions is the dynamics of isolated polymer molecules. The knowledge of how a single molecule
48: is deformed by the velocity gradient allows the development of
49: constitutive models that in turn can be used to calculate the large scale flow.
50: The stationary dynamics of a single, isolated polymer
51: has received much attention to date (see Larson 2005, Shaqfeh 2005, for a review).
52: On the contrary, less is known about non-equilibrium dynamics of isolated polymers in flow.
53: Since Rouse (1953) and Zimm's (1956) seminal works,
54: experimental, numerical, and theoretical studies focused on the internal
55: relaxation dynamics of a polymer floating in a solvent
56: under the influence of Brownian fluctuations.
57: The two model situations considered were:
58: (a)~a polymer suspended in solution and pulled at
59: the ends (Quake \textit{et al.} 1997, Hatfield and Quake 1999);
60: (b)~a tethered polymer submitted to a uniform flow and freely relaxing
61: after cessation of the flow (Perkins \textit{et al.} 1994, Brochard-Wyart 1995,
62: Manneville \textit{et al.} 1996, Rzehak and Zimmermann 2002).
63: The above studies did not take into account the interaction
64: between the polymer and an external flow. Hern\'andez Cifre and Garc\'{\i}a de la Torre (1999)
65: note that Rouse and Zimm's theories
66: may not provide the adequate time scale for coil--stretch processes
67: in flow on the basis of Brownian Dynamics simulations.
68: Here, we investigate how an elongational flow influences
69: polymer relaxation dynamics. We determine the time scale associated with polymer deformation
70: \textit{in the flow} and show a significant deviation from Rouse and Zimm's predictions in the vicinity
71: of the coil--stretch transition.
72: Our analysis is based on an eigenvalue problem for the confluent Heun equation,
73: and constitutes a new physical application of this latter equation.
74:
75: Polymer dynamics is extremely rich already in simple deterministic flows,
76: such as elongational flows~(Perkins \textit{et al.} 1997, Smith and Chu 1998),
77: shear flows~(Smith \textit{et al.} 1999, Celani \textit{et al.} 2005b, Puliafito and Turitsyn 2005,
78: Gerashchenko and Steinberg 2006)
79: or combinations of the two~(Hur \textit{et al.} 2002, Babcock \textit{et al.} 2003).
80: The velocity gradient of a non-uniform flow stretches the polymer, while
81: entropic forces attempt
82: to restore the polymer into the coiled equilibrium shape.
83: In other words, the configuration of the polymer results from the
84: counterbalance between
85: the entropic forces and the hydrodynamic drag.
86: Here we consider an elongational flow, which
87: is defined by a constant velocity gradient~$\lambda$.
88: This flow is particularly effective in stretching polymers
89: far from their equilibrium configuration.
90: A transition to the stretched state of the polymer occurs as
91: the velocity gradient exceeds the critical
92: value~$\lambda_{c}=1/(2\tau)$, where~$\tau$
93: is the time associated with the slowest relaxation mode of the polymer
94: in thermal equilibrium with the surrounding medium.
95: For~$\lambda<\lambda_{c}$
96: polymers stay in the coiled equilibrium configuration;
97: for~$\lambda>\lambda_{c}$ they become fully extended.
98: This phenomenon is known as the coil--stretch transition~(de Gennes 1974).
99: Rouse (1953) computed~$\tau$ as a function of the number of monomers
100: in the polymer assuming the polymer could be described as a beads-and-springs
101: chain with Hookean interactions. Zimm (1956) refined Rouse's prediction by
102: taking into account hydrodynamic interactions between segments of the chain.
103:
104: The study of polymer dynamics has benefited from a class of mesoscopic models
105: that are based on a coarse-grained treatment of the polymer molecule. One
106: of the simplest is
107: the elastic dumbbell model, which only takes into account the slowest oscillation mode of the molecule (Bird \textit{et al.} 1987).
108: Notwithstanding this crude simplification, the dumbbell model captures the main
109: aspects of polymer dynamics in elongational flows, such as the
110: coil--stretch transition (Perkins \textit{et al.} 1997)
111: and finite-time conformation hysteresis (Schroeder \textit{et al.} 2003, Schroeder
112: \textit{et al.} 2004).
113: We examine how the probability density function (PDF) of the
114: extension of a dumbbell-like
115: molecule approaches its stationary form. We compute
116: the typical time it takes for the initial PDF to attain its steady-state form
117: and show that, in the proximity of the
118: coil--stretch transition, this time is exceedingly long compared to~$\tau$.
119: A similar behaviour is encountered in
120: white-in-time isotropic random
121: flows (Celani \textit{et al.} 2005a, Martins Afonso and Vincenzi 2005);
122: however, for an elongational flow
123: the amplification of the transient relaxation time is much stronger.
124:
125: The problem is solved within the framework of the Fokker--Planck equation
126: for the PDF of the extension of the polymer.
127: The computation of the transient relaxation time is recast as a central
128: two-point connection problem~(CTCP) for a singly confluent Heun
129: equation.
130: The Heun equation is the general Fuchsian differential equation with
131: four regular singularities.
132: Its singly confluent form, also known as the generalised
133: spheroidal wave equation, results from the merging of
134: two regular singularities into one irregular singularity of
135: Poincar\'e rank~$1$
136: (see Slavyanov 1995, for a review). A CTCP is an eigenvalue problem
137: for an ordinary differential equation where: (a)~at
138: both endpoints of the
139: interval of definition is located a singularity of the equation;
140: (b)~the solutions are required to have a specified asymptotic behaviour
141: while approaching the two singularities from inside.
142: Contrary to the hypergeometric equation,
143: an explicit formula for the CTCP is not known for the Heun family
144: of equations. The eigenvalues can be determined only as solutions of
145: transcendental equations involving continued fractions.
146: Well-known applications of the confluent Heun equation
147: are the electronic spectra of
148: the hydrogen-molecule ion in the
149: Born-Oppenheimer approximation (Jaff\'e 1934) and
150: the Teukolsky equation describing small perturbations to the Kerr geometry
151: in black hole theory (Leaver 1985).
152: For a comprehensive introduction to physical applications of the
153: Heun equations we refer the reader to the book by
154: Slavyanov and Lay (2000).
155:
156: The rest of this paper is organised as follows. The
157: dumbbell model is introduced in
158: section~\ref{sec:dumbbell}. The computation of the
159: transient relaxation time and the related CTCP for the confluent
160: Heun equation are presented
161: in section~\ref{sec:stationary}.
162: Section~\ref{sec:conclusions} is devoted to conclusions.
163:
164:
165: \section{Elastic dumbbell model}
166: \label{sec:dumbbell}
167: When one is interested in the statistics of polymer extension
168: the dynamics of a polymer mole\-cule can be described,
169: to a first approximation, in terms of its slowest oscillation mode.
170: In this case, the molecule can be modelled as an elastic dumbbell,
171: i.e. as two beads connected by a
172: spring.
173: The beads represent the ends of the molecule and
174: the spring models entropic forces. The separation between the beads
175: measures the extension of the polymer.
176: When introduced into a
177: non-uniform flow the molecule experiences collisions with
178: fluid particles and becomes stretched under the action of the velocity
179: gradient. In the simplest case, the drag force
180: is assumed to be proportional to the velocity of the polymer relative
181: to the fluid and thermal agitation is modelled by Brownian motion.
182: In most applications the extension of the polymer remains smaller than the
183: dissipative scale of the carrier flow, and therefore the velocity field can be
184: assumed to be linear%
185: \footnote{Davoudi and Schumacher (2006) recently investigated the situation
186: where the extension of the polymer can reach the inertial range of
187: turbulence.}.
188: The separation vector of the ends of the polymer, $\bm{Q}$,
189: is then a stochastic process evolving according to
190: the three-dimensional stochastic ordinary differential equation%
191: \footnote{There is no Ito--Stratonovich ambiguity in this case
192: because the coefficient of Brownian motion does not depend
193: on $\bm{Q}$.} (e.g. Bird \textit{et al.} 1987)
194: \begin{equation}
195: \label{eq:dumbbell}
196: d\bm{Q}=\bm{Q}\cdot\nabla\bm{v}(s)\, ds-
197: f(Q)\,\frac{\bm{Q}}{2\tau}\, ds
198: +\sqrt{\frac{Q_0^2}{\tau}}\, d\bm{B}(s),
199: \end{equation}
200: where $Q=|\bm{Q}|$, $Q_0^2$ is the equilibrium mean-square separation
201: of the ends of the molecule, and~$\bm{B}(s)$
202: denotes the three-dimensional standard Brownian motion.
203: The function~$f$ determines the entropic force. We consider the
204: finitely extensible nonlinear elastic (FENE) dumbbell model with
205: \begin{displaymath}
206: f(Q)=\frac{1}{1-Q^2/L^2}, \qquad
207: 0\le Q< L,
208: %0\leq\left\Arrowvert\bm{Q}\right\Arrowvert\leq L.
209: \end{displaymath}
210: where~$L$ is the maximum extension of the molecule.
211: The force diverges as the molecular extension
212: approaches~$L$; thus the extension of the molecule will be finite and smaller than~$L$.
213: The FENE model is appropriate for synthetic
214: polymers such as polyacrylamide and polyethyleneoxide;
215: biological macromolecules such as DNA and polyeptides
216: are better described by the worm-like chain model (e.g. Larson 2005).
217: In elongational-flow experiments, the flexibility parameter
218: $b=L^2/Q_0^2$ usually ranges from~$10^2$ to~$10^4$
219: (Larson 2005, Shaqfeh 2005).
220:
221: The FENE model~\eqref{eq:dumbbell}
222: neglects hydrodynamic interactions between the segments of the
223: polymer. Inclusion of hydrodynamic interactions makes the coil--stretch
224: transition sharper, but do not change the transition qualitatively
225: (e.g. Wiest \textit{et al.} 1989, Hern\'andez Cifre and Garc\'{\i}a de la Torre 1998, 1999).
226: Furthermore,
227: the dumbbell model without hydrodynamic interactions accurately reproduces
228: the extension--strain curve observed in experiments (Perkins \textit{et al.} 1997).
229: The accuracy of the model can be ascribed to the
230: cancellation between the effect of the
231: distribution of drag forces along the chain
232: and the effect of the increase in effective drag coefficient with polymer
233: extension (Larson \textit{et al.} 1997). Neglecting hydrodynamic interactions
234: has the benefits of providing analytical results.
235: %Hydrodynamic interactions would of course play an essential role in
236: %conformation hysteresis.
237:
238: \section{Relaxation dynamics}
239: \label{sec:stationary}
240: A steady
241: planar elongational flow is characterised by one direction of stretch, one direction of
242: compression, and one neutral direction. The velocity gradient is
243: constant along the directions both of the extensional and compressional axes:
244: $\nabla_jv_i=\lambda(\delta_{i1}\delta_{j1}-\delta_{i2}\delta_{j2})$,
245: $\lambda>0$.
246: %\begin{equation}
247: %\label{eq:flow}
248: %\nabla\bm{v}%\left(\begin{array}{ccc}
249: %\lambda & 0 & 0\\
250: %0 & -\lambda & 0\\
251: %0 & 0 & 0
252: %\end{array}\right) \qquad \lambda>0.
253: %\end{equation}
254: When a polymer is immersed in that flow the
255: first component of the separation vector rapidly becomes much greater than the
256: other components: $Q_1\gg Q_2$ and~$Q_1\gg Q_3$.
257: The extension of the molecule, therefore, is approximatively
258: $Q\simeq Q_1$, and $f(Q)$ can be replaced by
259: $f(Q_1)$.\footnote{%
260: The above approximation holds true for a steady uniaxial extensional flow as well:
261: $\nabla_jv_i=\lambda(\delta_{i1}\delta_{j1}-\delta_{i2}\delta_{j2}/2
262: -\delta_{i3}\delta_{j3}/2)$.
263: %\begin{displaymath}
264: %\nabla\bm{v}%\left(\begin{array}{ccc}
265: %\lambda & 0 & 0\\
266: %0 & -\lambda/2 & 0\\
267: %0 & 0 & -\lambda/2
268: %\end{array}\right) \qquad \lambda>0.
269: %\end{displaymath}
270: Therefore, the results henceforth presented are
271: unchanged for this flow.}
272: According to this approximation, the first component of~\eqref{eq:dumbbell}
273: gives a one-dimensional stochastic differential
274: equation for the extension of the dumbbell:
275: \begin{equation}
276: \label{eq:extension}
277: dQ=\lambda Q\,d s-
278: f(Q)\,\frac{Q}{2\tau}\, d s
279: +\sqrt{\frac{Q_0^2}{\tau}}\, dB(s),
280: \end{equation}
281: where $B(s)$ is the one-dimensional Brownian motion.
282: For the sake of notational simplicity, we introduce the rescaled
283: separation vector~$\bm{q}=\bm{Q}/L$. The PDF of~$q=\vert\bm{q}\vert$, $\psi(q;t)$ with
284: $q\in[0,1]$, satisfies the one-dimensional Fokker--Planck equation associated
285: with~\eqref{eq:extension} (e.g. Stratonovich 1963)
286: \begin{equation}
287: \label{eq:FPE}
288: \begin{array}{l}
289: \partial_t \psi=\mathscr{L}\psi,\\
290: \displaystyle
291: \mathscr{L}\psi:=-\frac{\partial}{\partial q}
292: \bigg[\bigg(\mathit{Wi}-\frac{\hat{f}(q)}{2}\bigg)q\psi\bigg]+
293: \frac{1}{2b}\,\frac{\partial^2 \psi}{\partial q^2},
294: \end{array}
295: \end{equation}
296: where $t=s/\tau$, $\hat{f}(q)=f(Lq)$, and $\textit{Wi}=\lambda\tau$.
297: The dimensionless number~\textit{Wi} is known as the Weissenberg number and
298: measures the level of polymer stretching. For~$\textit{Wi}<1/2$
299: polymers are in the coiled state; for~$\textit{Wi}>1/2$ polymers are fully
300: extended. The critical value
301: $\textit{Wi}_{c}=\lambda_{c}\tau=1/2$ marks the
302: coil--stretch transition in elongational flows (de Gennes 1974).
303: Strictly speaking, the approximation leading
304: to~\eqref{eq:extension} and~\eqref{eq:FPE} holds true when the polymer is
305: sufficiently stretched, i.e. only for~$\textit{Wi}\gtrsim\textit{Wi}_c$.
306: Equation~\eqref{eq:FPE}, indeed, does not yield a good approximation of~$\psi(q;t)$
307: at low~$\mathit{Wi}$.
308: However, we shall see at the end of this section that
309: it is appropriate to investigate
310: polymer relaxation dynamics in terms of~\eqref{eq:extension} and~\eqref{eq:FPE}
311: also below the coil-stretch transition.
312:
313: The Fokker-Planck equation~\eqref{eq:FPE}
314: is solved with reflecting boundary conditions, that is
315: the probability current
316: \begin{displaymath}
317: j(q;t)=\textit{Wi}\: q\psi(q;t)-\frac{\hat{f}(q)}{2}q\psi(q;t)-\frac{1}{2b}
318: \left.\frac{\partial \psi}{\partial q}\right|_{q,t}
319: \end{displaymath}
320: is assumed to vanish at the endpoints of the interval of definition:
321: $j(0;t)=j(1;t)=0$ $\forall\,t>0$ (Stratonovich 1963).
322: Under these conditions,
323: the stationary distribution of the extension, $\psi_0(q)=\lim_{t\to\infty}
324: \psi(q;t)$, can be derived by simple integration (Bird \textit{et al.} 1987):
325: \begin{equation}
326: \label{eq:stationary}
327: \psi_0(q)=N e^{b\mathit{Wi}\, q^2}(1-q^2)^\frac{b}{2},
328: \qquad 0\leq q\leq 1,
329: \end{equation}
330: where
331: \begin{displaymath}
332: N=2\Gamma((b+3)/2)\,/\,[\sqrt{\pi}\,
333: \Gamma(b/2+1)\,\times\, _1F_1(1/2;(b+3)/2;b\textit{Wi})]
334: \end{displaymath}
335: and~$\Gamma$ and~$_1F_1$ denote the Euler Gamma function and the
336: confluent hypergeometric function, respectively (e.g. Erd\'elyi \textit{et al.} 1953).
337: Large extensions become more and more probable with
338: increasing~\textit{Wi} in accordance with experiments (Perkins \textit{et al.} 1997).
339: Note however that, at low~\textit{Wi},
340: equation~\eqref{eq:stationary} does not constitutes a
341: good description of the PDF of small extensions due to the
342: approximation behind~\eqref{eq:FPE}.
343:
344: The time behaviour of the system depends on the form of the spectrum of the
345: operator~$\mathscr{L}$ with reflecting boundary conditions:
346: \begin{eqnarray}
347: \label{eq:spectrum}
348: \mathscr{L}\psi_\mu=-\mu \psi_\mu,\\
349: \label{eq:bc}
350: j_\mu(0)=j_\mu(1)=0,
351: \end{eqnarray}
352: where $j_\mu$ denotes the probability
353: current associated with the eigenfunction~$\psi_\mu$.
354: Under conditions~\eqref{eq:bc}, $\mathscr{L}$
355: is non-positive defined and symmetric with respect to the scalar product
356: with weighting function~$1/\psi_0$. Its eigenvalues~$\mu$ are therefore
357: real and non-negative, $\mu=0$
358: being associated with the long-time solution~$\psi_0$ (Stratonovich 1963).
359: If~$\mathscr{L}$ has a countable spectrum~$\{0,\mu_1,\mu_2,\dots\}$
360: with~$\mu_i<\mu_{i+1}$, then~\mbox{$T_{\mathrm{rel}}=\tau/\mu_1$}
361: is the characteristic time needed for~$\psi(q;t)$
362: to attain its long-time form~$\psi_0(q)$ when the initial
363: condition~$\psi(q;0)$ is taken far from equilibrium (Schenzle and Brand 1979).
364: We note that while~$\tau$ is the relaxation time
365: of the polymer \textit{in the absence
366: of flow}, $T_{\mathrm{rel}}$ characterises polymer relaxation
367: dynamics \textit{in the flow}.
368:
369: To compute~$T_{\mathrm{rel}}$, we need to solve the eigenvalue
370: problem~\eqref{eq:spectrum}, \eqref{eq:bc}.
371: By making the substitutions~$z=q^2$,
372: $\psi_\mu(z)=(1-z)^{\frac{b}{2}}w_\mu(z)$, we can
373: rewrite~(\ref{eq:spectrum}) in the form
374: \begin{equation}
375: \label{eq:CHE}
376: w''_\mu+\bigg(\beta+\frac{\gamma}{z}+\frac{\delta}{z-1}\bigg)w'_\mu
377: +\frac{\alpha\beta z-\nu}{z(z-1)}\,w_\mu=0,
378: \end{equation}
379: where $w'_\mu:=dw_\mu/dz$, $\gamma=1/2$, $\delta=b/2$,
380: $\alpha=(1+b-\mu/\mathit{Wi})/2$, $\beta=-b\mathit{Wi}$, and
381: $\nu=b(\mu-\mathit{Wi})/2$.
382: The above equation is a singly confluent Heun equation in
383: the non-symmetrical canonical form
384: (Decarreau \textit{et al.} 1978a, Decarreau \textit{et al.} 1978b, Slavyanov 1995)%
385: \footnote{%
386: The worm-like chain model, $\hat{f}(q)=2/3-1/(6q)+1/[6q(1-q)^2]$,
387: would lead to a second-order linear differential equation with two
388: irregular singularities of Poincar\'e rank~1 and~3, respectively.
389: The FENE model in a $\delta$-correlated random flow can be solved in terms of a general Heun equation
390: (Martins Afonso and Vincenzi 2005).}.
391:
392:
393: Equation~(\ref{eq:CHE}) has two regular singularities at~$z=0$ and $z=1$
394: and an irregular singularity of Poincar\'e rank 1 at~$z=\infty$.
395: The characteristic exponents
396: at~$z=0$ are~0 and~$1-\gamma$; at~$z=1$, they are~0 and~$1-\delta$.
397: We neglect the case of~$b$ integer and odd since it does not have
398: physical relevance.
399: Consequently, there are no logarithmic solutions and,
400: for $|z|<1$, $w_\mu(z)$ can be written as $w_\mu(z)
401: =a_0\varphi_0(z)+d_0z^{1/2}\chi_0(z)$ with $a_0$, $d_0$ complex constants and
402: $\varphi_0$, $\chi_0$ analytic functions of~$z$ (equivalently of~$q^2$)
403: such that~$\varphi_0(0)\ne 0$ and~$\chi_0(0)\ne 0$.
404: Similarly,
405: for $|z-1|<1$, $w_\mu(z)$ takes the form~$w_\mu(z)=a_1\varphi_1(z-1)+d_1
406: (1-z)^{1-b/2}\chi_1(z-1)$ with $a_1$, $d_1$ complex constants and $\varphi_1$,
407: $\chi_1$ analytic functions of~$z$ such
408: that~$\varphi_1(0)\neq 0$ and~$\chi_1(0)\neq 0$.
409:
410: It is easily seen that boundary conditions~(\ref{eq:bc})
411: can be matched only by those solutions
412: of~(\ref{eq:CHE}) that belong to the exponent~0 both
413: in $z=0$ and $z=1$ (that is both $d_0$ and $d_1$ have to be zero)\footnote{
414: We leave aside the case $b=2$
415: since it does not have physical relevance.}. The eigenvalue problem defined
416: by~(\ref{eq:spectrum}) and (\ref{eq:bc}) is therefore mapped into a
417: CTCP on $[0,1]$ for the confluent Heun equation~(\ref{eq:CHE}).
418: To compute the eigenvalues~$\mu$, we adapt the procedure exploited
419: by Svartholm (1939) and Erd\'elyi (1942, 1944)
420: to~(\ref{eq:CHE}) (see also Slavyanov 1995, Slavyanov and Lay 2000).
421: We then expand~$w_\mu$ in series of
422: Jacobi polynomials having the required characteristic exponent
423: at~$z=0$ and~$z=1$:
424: \begin{equation}
425: \label{eq:expansion}
426: w_\mu(z)=\sum_{n=0}^{\infty}c_n(\mu) u_n(z)
427: \end{equation}
428: with
429: \begin{displaymath}
430: u_n(z)=\,
431: _2F_1(-n,\omega+n;\gamma;z)=(-1)^n\frac{n!\sqrt{\pi}}{\Gamma(n+\gamma)}
432: \,P_n^{(\omega-\gamma,\gamma-1)}(2z-1)
433: \end{displaymath}
434: and $\omega=\gamma+\delta-1$.
435: The function~$_2F_1$ denotes the Gauss hypergeometric function;
436: $P_n^{(\omega-\gamma,\gamma-1)}$ are the Jacobi polynomials of
437: parameters~$\omega-\gamma$ and~$\gamma-1$ and degree~$n$
438: (e.g. Erd\'elyi \textit{et al.} 1953).
439: %Jacobi polynomials~$P_n^{(\omega-\gamma,\gamma-1)}(2z-1)$
440: %form an orthogonal basis on~$L^2[0,1]$ with respect to the weighting
441: %function~$z^{\omega-1}$.
442: The conditions for the convergence
443: of~\eqref{eq:expansion} will determine the spectrum of~$\mathscr{L}$.
444:
445: By defining the operators
446: \begin{eqnarray*}
447: \mathscr D_1w_\mu:=z(z-1)
448: \bigg[w''_\mu+\bigg(\frac{\gamma}{z}+\frac{\delta}{z-1}\bigg)w'_\mu\bigg],\\
449: \mathscr D_2w_\mu:=-z(z-1)w'_\mu,
450: \end{eqnarray*}
451: we can rewrite~(\ref{eq:CHE}) in the form
452: \begin{equation}
453: \label{eq:CHE_2}
454: \mathscr D_1w_\mu+\varepsilon \mathscr D_2w_\mu+(\rho z+\sigma)w_\mu=0,
455: \end{equation}
456: where $\varepsilon=b\mathit{Wi}$, $\rho=b[\mu-(1+b)\mathit{Wi}]/2$,
457: $\sigma=b(\mathit{Wi}-\mu)/2$. The polynomials~$u_n$ satisfy the
458: differential relations (Erd\'elyi \textit{et al.} 1953)
459: \begin{equation}
460: \label{eq:pol_1}
461: \mathscr{D}_1u_n=n\left(\omega+n\right)u_n
462: \end{equation}
463: and
464: \begin{equation}
465: \label{eq:pol_2}
466: \mathscr{D}_2u_n=\widetilde{A}_nu_{n+1}+\widetilde{B}_nu_{n}+
467: \widetilde{C}_nu_{n-1}
468: \end{equation}
469: with
470: \begin{eqnarray*}
471: \widetilde{A}_n=\frac{n(n+\omega)(n+\gamma)}{(2n+\omega)(2n+\omega+1)},
472: \quad
473: \widetilde{B}_n=\frac{n(n+\omega)(\delta-\gamma)}{(2n+\omega-1)(2n+\omega+1)},
474: \\[0.2cm]
475: \widetilde{C}_n=-\frac{n(n+\omega)(n+\delta-1)}{(2n+\omega)(2n+\omega-1)}\,.
476: \end{eqnarray*}
477: In addition, the following recurrence relation holds (Erd\'elyi \textit{et al.} 1953):
478: \begin{equation}
479: \label{eq:pol_3}
480: zu_n=\widehat{A}_nu_{n+1}+\widehat{B}_nu_n+\widehat{C}_nu_{n-1}
481: \end{equation}
482: with coefficients
483: \begin{eqnarray*}
484: \widehat{A}_n=-\frac{(n+\gamma)(n+\omega)}{(2n+\omega)(2n+\omega+1)},
485: \quad
486: \widehat{B}_n=\frac{2n(n+\omega)+\gamma(\omega-1)}{(2n+\omega-1)(2n+\omega+1)},\\[0.2cm]
487: \widehat{C}_n=-\frac{n(n+\delta-1)}{(2n+\omega-1)(2n+\omega)}\,.
488: \end{eqnarray*}
489: Inserting expansion~(\ref{eq:expansion}) in~(\ref{eq:CHE_2}) and
490: exploiting relations~(\ref{eq:pol_1}), (\ref{eq:pol_2}), (\ref{eq:pol_3})
491: yield the following three-term recurrence relation for~$c_n(\mu)$:
492: \begin{eqnarray}
493: \label{eq:iniz_1}
494: g_{-1}c_0=0&\\
495: \label{eq:iniz_2}
496: g_0c_1+h_0c_0=0&\\
497: \label{eq:3}
498: g_nc_{n+1}+h_nc_n+k_nc_{n-1}=0& \quad \mathrm{for}\ n\ge 2,
499: \end{eqnarray}
500: where the coefficients
501: \begin{eqnarray*}
502: g_n=\varepsilon\widetilde{C}_{n+1}+\rho\widehat{C}_{n+1},\qquad
503: h_n=n(\omega+n)+\varepsilon\widetilde{B}_n+\rho\widehat{B}_n+\sigma,
504: \\
505: k_n=\varepsilon\widetilde{A}_{n-1}+\rho\widehat{A}_{n-1}
506: \end{eqnarray*}
507: depend on~$\mu$ through~$\rho$ and~$\sigma$.
508: The asymptotic behaviour of~$h_n/g_n$ and~$k_n/g_n$ is given by
509: \begin{equation}
510: \label{eq:asympt}
511: \frac{h_n}{g_n}\sim
512: -\frac{4n}{\varepsilon}\qquad\mathrm{and}\qquad
513: \frac{k_n}{g_n}\sim -1\qquad (n\to\infty).
514: \end{equation}
515: According to the Perron-Kreuser theorem (e.g. Gautschi 1967, Wimp 1984),
516: \eqref{eq:asympt}
517: implies that the recurrence rela\-tion~(\ref{eq:3}) has two linearly independent
518: solutions~$(c_{n}^+)_n$ and~$(c_{n}^-)_n$ such that
519: \begin{equation}\label{eq:minimal}
520: \frac{c^+_{n+1}}{c^+_{n}}\sim\frac{4n}{\varepsilon}
521: \qquad \mathrm{and} \qquad
522: \frac{c^-_{n+1}}{c^-_{n}}\sim-\frac{\varepsilon}{4n}
523: \qquad (n\to\infty).
524: \end{equation}
525: The above sequences have the property: $\lim_{n\to\infty}c^-_n/c^+_n=0$;
526: hence $(c^-_n)_n$ is what is called a minimal solution of~\eqref{eq:3}.
527: Every other solution non proportional to~$(c^-_n)_n$
528: is asymptotically proportional to~$(c^+_n)_n$, and
529: therefore diverges with increasing~$n$. Therefore, we can already say
530: that~\eqref{eq:expansion} can converge only if the
531: sequence~$(c_n)_n$ is a minimal solution of~\eqref{eq:3}.
532:
533: The second limit in~\eqref{eq:minimal} implies that
534: $\lim_{n\to\infty}\sqrt[n]{|c^-_n|}=0$. This condition ensures
535: that, if the coefficients~$c_n$ form a minimal solution of~\eqref{eq:3}, then
536: expansion~\eqref{eq:expansion}
537: convergences absolutely to an analytic
538: function in the whole complex plane (Szeg\"o 1939, p.~252).
539:
540: Nonetheless,
541: it is also required that~$c_n=0$ for~$n<0$, that is the
542: sequence~$(c_n)_n$ is subject to initial conditions~\eqref{eq:iniz_1}
543: and~\eqref{eq:iniz_2}.
544: Equation~\eqref{eq:iniz_1} is trivially satisfied since~$g_{-1}=0$, and so
545: does~\eqref{eq:iniz_2} for~$\mu=0$. (This is in accordance with the fact
546: that~$\psi_0$ exists for all~$\textit{Wi}$ and~$b$.)
547: However, for~$\mu\ne 0$, equation~\eqref{eq:iniz_2} fixes the
548: ratio~$c_1/c_0=-(b+3)/2$. This latter requirement is not satisfied by
549: whatever~$\mu$, and therefore select the spectrum of~$\mathscr{L}$.
550:
551: To summarise, the spectrum of~$\mathscr{L}$ is the set
552: of those~$\mu$ such that~$(c_n(\mu))_{n\ge 0}$ is a minimal
553: solution of~\eqref{eq:3} satisfying~$c_1(\mu)/c_0(\mu)=-(b+3)/2$.
554: From Pincherle's theorem (e.g. Gautschi 1967, p.~31), this conclusion is equivalent
555: to stating that the eigenvalues~$\mu$ are the solutions of
556: \[\displaystyle
557: \frac{k_1/g_1}{h_1/g_1-
558: \displaystyle\frac{k_2/g_2}{h_2/g_2-\displaystyle\frac{k_3/g_3}{h_3/g_3-
559: \dots}}}
560: =\frac{b+3}{2}.
561: \]
562: The latter represents a transcendental equation
563: for the relaxation spectrum associated with the Fokker-Planck
564: equation~\eqref{eq:FPE}. The reciprocal of the smallest
565: nonzero solution is~$T_{\mathrm{rel}}/\tau$.
566:
567: We calculated~$T_{\mathrm{rel}}$ numerically by means of
568: the modified Miller algorithm (Wimp 1984, pp.~82--85).
569: The behaviour of~$T_{\mathrm{rel}}$ as a function of~\textit{Wi}
570: is reported in fig.~\ref{fig}.
571: \begin{figure}
572: \begin{center}
573: \includegraphics[width=0.6\textwidth]{T_Wi_24.5.eps}
574: \end{center}
575: \caption{\label{fig}Rescaled relaxation time~$T_{\mathrm{rel}}/\tau$
576: as a function of the Weissenberg
577: number~\textit{Wi} for~$b=600.25$; the inset shows the comparison
578: with the small-\textit{Wi} behaviour $1/(1-2\mathit{Wi})$ (dashed line).
579: }
580: \end{figure}
581: For very small~$\textit{Wi}$ the influence of the external flow is negligible
582: and~$T_{\mathrm{rel}}$ is approximately equal to~$\tau$.
583: With increasing~$\textit{Wi}$ the ratio~$T_{\mathrm{rel}}/\tau$ starts
584: growing as~$1/(1-2\textit{Wi})$.
585: This behaviour can be deduced by
586: replacing the spring force with a Hookean force, $\hat{f}(q)=1$,
587: and computing the relaxation of the moments of~$q$.
588: For a fixed~$b$, the rescaled relaxation time reaches a
589: sharp maximum~$T_{\mathrm{max}}$ in the neighbourhood of the coil--stretch
590: transition ($\textit{Wi}_{c}=1/2$);
591: the value of~$T_{\mathrm{max}}$ increases
592: linearly with~$\sqrt{b}$ (fig.~\ref{fig:2}). With increasing~$b$,
593: $T_{\mathrm{rel}}$ attains its maximum value~$T_{\mathrm{max}}$
594: closer and closer
595: to~$\textit{Wi}_{c}$ and the width of the peak decreases
596: (fig.~\ref{fig:2}).
597: \begin{figure}
598: \begin{center}
599: \includegraphics[width=0.325\textwidth]{T_rel.eps}\hspace{0.2cm}
600: \includegraphics[width=0.35\textwidth]{wi_max.eps}\\[0.4cm]
601: \includegraphics[width=0.62\textwidth]{width.eps}
602: \end{center}
603: \caption{(\textit{a}) Maximum rescaled relaxation
604: time~$T_{\mathrm{max}}/\tau$ vs. the square root
605: of the flexibility parameter~$b=L^2/Q_0^2$: $T_{\mathrm{max}}/\tau$
606: increases as $\theta_1\sqrt{b}+\theta_2$
607: with $\theta_1\approx 0.69$ and $\theta_2\approx -1.18$; (\textit{b})
608: Behaviour of~$\mathit{Wi}_{\mathrm{max}}$ as a function of~$\sqrt{b}$,
609: where~$\mathit{Wi}_{\mathrm{max}}$ is the Weissenberg
610: number at which the relaxation time attains its maximum
611: value~$T_{\mathrm{max}}$;
612: (\textit{c})~Relaxation time vs.~\textit{Wi} for different~$b$,
613: rescaled to make the maximum values coincide.}
614: \label{fig:2}
615: \end{figure}
616: For large~$\textit{Wi}$ the relaxation time is fixed by the time scale
617: of the flow~$\lambda^{-1}$, and therefore~$T_{\mathrm{rel}}/\tau$ decreases
618: as~$\textit{Wi}^{-1}$.
619:
620:
621: The computation of~$T_{\mathrm{rel}}$ shows that,
622: near the coil--stretch transition, the
623: typical time scale involved in polymer relaxation
624: dynamics is strongly different from~$\tau$. The time associated with the
625: fundamental oscillation mode
626: is not representative of polymer temporal dynamics in flow.
627: These results confirm the behaviour
628: encountered in isotropic $\delta$-correlated Gaussian flows
629: (Celani \textit{et al.} 2005a, Martins Afonso and Vincenzi 2005),
630: but here the
631: enhancement of~$T_{\mathrm{rel}}$ is stronger since the elongational flow
632: is more efficient in deforming polymers.
633:
634: In Brownian Dynamics simulations and experiments the relaxation time of~$\psi(q;t)$ can be measured
635: from the decay of the moments of the extension to their steady-state value.
636: To observe~$T_{\mathrm{rel}}$,
637: one has to make sure that the initial condition~$\psi(q;0)$ is not orthogonal to the first
638: eigenfunction of~$\mathscr{L}$, that is $\psi_1(q)$ in~\eqref{eq:spectrum}, with respect to the scalar
639: product with weighting function~$\psi_0(q)$.
640: Figure~\ref{fig:eigen} shows the
641: shape of~$\psi_1(q)$ for different~$\textit{Wi}$.
642: An initial PDF concentrated either at a coiled or a stretched extension~$\bar{q}$
643: (formally $\psi(q;0)=\delta(q-\bar{q})$) has projection onto~$\psi_1(q)$ equal
644: to~$\psi_1(\bar{q})/\psi_0(\bar{q})$ by construction (e.g. Stratonovich 1963).
645: The initial condition with all the polymers having the same extension~$\bar{q}$ is therefore not orthogonal
646: to~$\psi_1(q)$ if~$\psi_1(\bar{q})\neq 0$;
647: all the moments will thus tend to their stationary value with a typical time scale~$T_{\mathrm{rel}}$.
648: \begin{figure}[t]
649: \includegraphics[width=0.47\textwidth]{staz.eps}
650: \hfill
651: \includegraphics[width=0.458\textwidth]{first.eps}
652: \caption{Stationary PDF of the extension and eigenfunction of the Fokker--Planck operator~$\mathscr{L}$
653: associated with the smallest nonzero eigenvalue~$\mu_1$ ($b=600.25$).
654: The eigenfunctions have been computed numerically by means of the iteration--variation method
655: (Morse and Feshbach 1953).}
656: \label{fig:eigen}
657: \end{figure}
658:
659:
660: We now show that it is accurate to compute~$T_{\mathrm{rel}}$ by means of~\eqref{eq:FPE}
661: also for~$\textit{Wi}<\textit{Wi}_c$.
662: At low~$\textit{Wi}$, the entropic force can be modelled as
663: a Hookean force, i.e.~$f(Q)=1$ in~\eqref{eq:dumbbell}.
664: The vector equation~\eqref{eq:dumbbell} therefore reduces to a set of three decoupled stochastic
665: differential equations for $Q_1$, $Q_2$, $Q_3$. Hence we
666: have~$\Psi(\bm{q};t)=\psi^{(1)}(q_1;t)\psi^{(2)}(q_2;t)\psi^{(3)}(q_3;t)$,
667: where~$\Psi(\bm{q};t)$ is the PDF of the vector~$\bm{q}$
668: and~$\psi^{(i)}(q_i;t)$ is the PDF of the $i$-th component of~$\bm{q}$.
669: The relaxation time of~$\Psi(\bm{q};t)$ is thus the longest of the
670: relaxation times of~$\psi^{(1)}(q_1;t)$, $\psi^{(2)}(q_2;t)$,
671: and~$\psi^{(3)}(q_3;t)$ given that all the~$\psi^{(i)}(q_i;t)$ have a non-trivial
672: long-time limit.
673: It is easy to check that at low~\textit{Wi} the relaxation time
674: of~$\psi^{(1)}(q_1;t)$ is~$T_{\mathrm{rel}}^{(1)}=\tau/(1-2\mathit{Wi})$,
675: while~$T_{\mathrm{rel}}^{(2)}=\tau/(1+2\mathit{Wi})$ and~$T_{\mathrm{rel}}^{(3)}=\tau$ (with obvious
676: notation). The function $\Psi(\bm{q};t)$ therefore has the same relaxation time
677: as~$\psi^{(1)}(q_1;t)$.
678: Now we note that~$\psi(q;t)$ and~$\Psi(\bm{q};t)$ have the same long-time behaviour
679: since~$\psi(q;t)=q^2\int_0^{\pi}\int_0^{2\pi} \Psi(\bm{q}(q,\vartheta,\phi);t) \sin{\vartheta}\,d\vartheta d\phi$. We thus come to the following conclusion:
680: the relaxation times of~$\psi(q;t)$ and~$\psi^{(1)}(q_1;t)$
681: coincide at low~\textit{Wi} and $\psi^{(1)}(q_1;t)$ satisfies~\eqref{eq:FPE}.
682: This fact explains why~\eqref{eq:FPE} yields an accurate computation of~$T_{\mathrm{rel}}$ also below the
683: coil--stretch transition.
684:
685: \section{Summary and conclusions}
686: \label{sec:conclusions}
687: We have investigated the relaxation dynamics of an isolated polymer in an
688: external flow. Previous studies focused on the relaxation of
689: a polymer in a solvent driven only by Brownian
690: motion. Those studies
691: determined the time~$\tau$ associated with the slowest oscillation mode
692: of the molecule in the absence of external flow.
693: We have considered an elongational flow and derived a transcendental equation for the relaxation
694: spectrum associated with the time evolution of the PDF of polymer extension.
695: The problem of computing this spectrum has been recast as
696: a central two-point connection problem for a confluent Heun equation.
697: The Heun equation results from the separation of the time variable
698: in the Fokker--Planck equation for the PDF of the extension.
699:
700: Our analysis shows that an external elongational flow strongly influences
701: polymer relaxation dynamics.
702: The longest relaxation time associated with the evolution of the PDF of the
703: extension provides an estimation of the time scale of polymer deformation.
704: Near the coil--stretch transition,
705: this time is significantly greater than~$\tau$
706: (already one order of magnitude greater
707: for short molecules).
708: The physical reason is the large number of polymer configurations close to the coil--stretch transition. At intermediate
709: Weissenberg numbers polymer conformation results from a critical
710: competition between the entropic force and the velocity gradient.
711: Therefore, around the transition,
712: coiled and stretched polymer coexist in the flow, and this makes
713: the relaxation to the equilibrium PDF particularly long.
714: It is worth noticing that the amplification of the relaxation
715: time is stronger when hydrodynamic interactions are taken into account;
716: this behaviour is intimately related to the
717: finite-time conformation hysteresis observed in elongational
718: flows (Celani \textit{et al.} 2006).
719:
720: \acknowledgments
721: The authors are grateful to A.~Celani, C.~Doering, M.~Martins Afonso,
722: A.~Puliafito, and B.~Rajaratnam for useful
723: suggestions. This work has been partially supported by the European Union
724: under the contract HPRN-CT-2002-0030
725: ``Fluid Mechanical Stirring and Mixing: the Lagrangian Approach''.
726:
727: \begin{itemize}
728: \addtolength{\itemsep}{-0.25cm}
729: \item[] Babcock~H~P, Teixeira~R~E, Hur~J~S, Shaqfeh~E~S and Chu~S
730: 2003 {\it Macromolecules} {\bf 36} 4544--8
731: \item[] Bird R B, Hassager O, Armstrong R C and Curtiss C F 1987
732: {\it Dynamics of Polymeric Liquids, Fluid Mechanics} vol~2
733: (New York: Wiley)
734: \item[] Brochard-Wyart F 1995 \textit{Europhys. Lett.} \textbf{30} 387--92
735: \item[] Celani A, Musacchio S and Vincenzi D 2005a
736: {\it J. Stat. Phys.} {\bf 118} 531--54
737: %\item[] Celani A, Musacchio S and Vincenzi D 2005b Single polymer
738: %dynamics in random flow In: Frontiers of Nonlinear Physics, Proceedings of the
739: %2nd International Conference (Nizhny Novgorod - St. Petersburg, Russia,
740: %5-12 July, 2004), ed A. Litvak (Nizhny Novgorod, Russia, Institute of Applied
741: %Physics RAS, 2005), pp. 264-267
742: \item[] Celani A, Puliafito A and Turitsyn K 2005b {\it Europhys. Lett.}
743: {\bf 70} 464--70
744: \item[] Celani A, Puliafito A and Vincenzi D 2006 \textit{Phys. Rev. Lett.}
745: \textbf{97} 118301
746: \item[] Davoudi J and Schumacher J 2006 {\it Phys. Fluids} \textbf{18} 025103
747: \item[] Decarreau~A, Dumont-Lepage~M-Cl, Maroni~P, Robert~A and Ronveaux~A
748: 1978a {\it Ann. Soc. Sci. Bruxelles} {\bf 92} 53--78
749: \item[] Decarreau A, Maroni~P, Robert~A
750: 1978b {\it Ann. Soc. Sci. Bruxelles} {\bf 92} 151--89
751: \item[] Erd\'elyi A 1942 {\it Duke Math. J.} {\bf 9} 48--58
752: \item[] Erd\'elyi A 1944 {\it Q. J. Math.} {\bf 15} 62--9
753: \item[] Erd\'elyi, Magnus W, Oberhettinger F and Tricomi F G 1953
754: \textit{Higher transcendental functions}
755: (New York: McGraw-Hill)
756: \item[] Gautschi W 1967 {\it SIAM Rev.} {\bf 9} 24--82
757: \item[] de Gennes P G 1974 {\it J. Chem. Phys.} {\bf 60} 5030--42
758: \item[] Gerashchenko~S and Steinberg~V 2006 \textit{Phys. Rev. Lett.}
759: \textbf{96} 038304
760: \item[] Hatfield J W and Quake S R 1999 \textit{Phys. Rev. Lett.}
761: \textbf{82} 3548--51
762: \item[] Hern\'andez Cifre J G and Garc\'{\i}a de la Torre J 1998
763: \textit{J. Non-Cryst. Solids} \textbf{235-237} 717--22
764: \item[] Hern\'andez Cifre J G and Garc\'{\i}a de la Torre J 1999 {\it J. Rheol.}
765: \textbf{43} 339--58
766: \item[] Hur~J~S, Shaqfeh~E~S, Babcock~H~P and Chu~S 2002 {\it Phys. Rev.} E
767: {\bf 66} 011915
768: \item[] Jaff\'e G 1933 {\it Z. Phys.} {\bf 87} 535--44
769: \item[] Larson R G, Perkins T T, Smith D E and Chu S \textit{Phys. Rev.} E
770: \textbf{55} 1794--7
771: \item[] Larson R G 2005 {\it J. Rheol.} {\bf 49} 1--70
772: \item[] Leaver~E~W 1985 {\it Proc. R. Soc. London} A {\bf 402} 285--98
773: \item[] Manneville S, Cluzel Ph, Viovy J-L, Chatenay D and Caron F 1996
774: \textit{Europhys. Lett.} \textbf{36} 413--8
775: \item[] Martins Afonso M and Vincenzi D 2005 {\it J. Fluid Mech.} {\bf 540}
776: 99--108
777: \item[] Morse P M and Feshbach H 1953 {\it Methods of Theoretical Physics}
778: (New York: McGraw--Hill)
779: \item[] Perkins T T, Quake S R, Smith D E and Chu S 1994 \textit{Science}
780: \textbf{264} 822
781: \item[] Perkins T T, Smith D E and Chu S 1997 {\it Science} {\bf 276} 2016--21
782: \item[] Puliafito A and Turitsyn K 2005 {\it Physica}~D {\bf 211} 9--22
783: \item[] Quake S R, Babcock H and Chu S 1997 \textit{Nature} \textbf{388} 151--4
784: \item[] Rouse P E 1953 \textit{J. Chem. Phys.} \textbf{21} 1272--80
785: \item[] Rzehak R and Zimmermann W 2002 \textit{Europhys. Lett.} \textbf{59}
786: 779--85
787: \item[] Schenzle A and Brand H 1979 {\it Phys. Rev.} A {\bf 20} 1628--47
788: \item[] Schroeder C M, Babcock H P, Shaqfeh E S G and Chu S 2003
789: {\it Science} {\bf 301} 1515--9
790: \item[] Schroeder C M, Shaqfeh E S G and Chu S 2004 {\it Macromolecules}
791: {\bf 37} 9242--56
792: \item[] Shaqfeh E S G 2005 \textit{J. Non-Newton. Fluid Mech.} \textbf{130}
793: 1--28
794: \item[] Slavyanov S Yu 1995 {\it Confluent Heun equation}, in {\it Heun's Differential Equations} ed A. Ronveaux (New York: Oxford University Press)
795: \item[] Slavyanov S Yu and Lay W 2000
796: {\it Special Functions, A Unified Theory Based on Singularities}
797: (New York: Oxford University Press)
798: \item[] Smith~D~E and Chu~S 1998 {\it Science} {\bf 281}, 1335--40
799: \item[] Smith D E, Babcock~H~P and Chu~S 1999 {\it Science} {\bf 283} 1724--7
800: \item[] Sreenivasan K R and White C M 2000 {\it J. Fluid Mech.} \textbf{409}
801: 149--64
802: \item[] Stratonovich R L 1963 {\it Topics in the Theory of Random Noise}
803: (New York: Gordon and Breach Science Publishers)
804: \item[] Svartholm N 1939 {\it Math. Ann.} {\bf 116} 413--21
805: \item[] Szeg\"o G 1939 {\it Orthogonal Polynomials}
806: (Providence, RI: American Mathematical Society)
807: \item[] Wiest J L, Wedgewood L and Bird R B 1989 \textit{J. Chem. Phys.} \textbf{90} 587--94
808: \item[] Wimp~J 1984 {\it Computation with Recurrence Relations}
809: (Pitman Publishing Inc.)
810: \item[] Zimm~B~H 1956 {\it J. Chem. Phys.} \textbf{24} 269--78
811: \end{itemize}
812: \end{document}
813: