1:
2: \documentclass[pre,twocolumn,showpacs,preprintnumbers,floatfix,superscriptaddress,showpacs]{revtex4}
3: \usepackage{graphicx,citesort,amsmath,amssymb}
4: \usepackage[dvips]{color}
5:
6: \renewcommand{\vec}[1]{\pmb{#1}}
7: \newcommand{\avg}[1]{\left\langle #1\right\rangle}
8: \newcommand{\spd}[2]{\left( #1|#2\right) }
9: \newcommand{\abs}[1]{\left\lvert #1\right\rvert}
10: \newcommand{\mint}[4]{\int_{#2}^{#3}\!\!#1\,#4}
11: \newcommand{\rem}[1]{\textbf{#1}}
12: \newcommand{\kil}[1]{{\tiny #1}}
13: \newcommand{\Ord}[1]{{\cal O}(#1)}
14: \newcommand{\lOrd}[1]{{\cal o}(#1)}
15: \newcommand{\ord}[1]{{\text{ord}}(#1)}
16: \newcommand{\pe}{\perp}
17: \newcommand{\pa}{\parallel}
18: \newcommand{\rpe}{r_\perp}
19: \newcommand{\rpa}{r_\parallel}
20: \newcommand{\dpe}{\avg{\delta\rpe^2}}
21: \newcommand{\dpa}{\avg{\delta R_\pa^2}}
22: \newcommand{\epe}{\hat{\vec e}_\pe}
23: \newcommand{\epa}{\hat{\vec e}_\pa}
24: \newcommand{\cpe}{\chi_\pe}
25: \newcommand{\tv}{\vec {\mathfrak t}}
26: \newcommand{\nv}{\vec {\mathfrak n}}
27: \newcommand{\bv}{\vec {\mathfrak b}}
28:
29: \newcommand{\kr}{\chi}
30: \newcommand{\ts}{\tau}
31:
32: \newcommand{\fopt}{f^{\text{OPT}}}
33:
34: \newcommand{\fe}{\mathfrak f \,}
35: \newcommand{\efe}{\hat{\vec e}_\fe}
36: \newcommand{\fel}{\vec{f}_{\text{el}}\,}
37: \newcommand{\gex}{\mathfrak g \,}
38: \newcommand{\gext}{\mathfrak g_0}
39: \newcommand{\gfl}{\vec{ g}_{\text{fr}}}
40: \newcommand{\gel}{\vec{ g}_{\text{el}}}
41:
42: \newcommand{\Hext}{{\cal H}_{\text{ext}}}
43: \newcommand{\Hb}{{\cal H}_{\text{WLC}}}
44: \newcommand{\Ht}{{\cal H}_{\text{tors}}}
45: \newcommand{\Hc}{{\cal H}_{\text{comp}}}
46: \newcommand{\Htot}{{\cal H}}
47:
48: \newcommand{\msd}{\text{MSD}}
49: \newcommand{\msdl}{\text{msd}}
50:
51: \newcommand{\fc}{c}
52: \newcommand{\ludo}{\ell_\text{SWN}}
53: \newcommand{\lralf}{\ell_\text{EJAM}}
54: \newcommand{\lpierre}{\ell_\text{BBG}}
55: \newcommand{\lflower}{\ell_\text{Flower}}
56: \newcommand{\lp}{\ell_p}
57: \newcommand{\lpe}{\ell_\perp}
58: \newcommand{\lpa}{\ell_\|}
59: \newcommand{\tfin}{t_L^\|}
60: \newcommand{\tf}{t_\fe}
61: \renewcommand{\sf}{s_\fe}
62: \newcommand{\tot}{\tau_{\text{tot}}}
63: \newcommand{\totpe}{t_L^\pe}
64: \newcommand{\totpa}{t_L^\pa}
65:
66: \newcommand{\eom}{equations of motion}
67: \newcommand{\leom}{longitudinal equation of motion}
68: \newcommand{\teom}{transverse equation of motion}
69: \newcommand{\rhs}{right hand side}
70:
71: \newcommand{\crm}[1]{\color{red}{#1 }\color{black}}
72:
73: \newcounter{fnt}
74:
75: \newcommand{\pull}{\text{\emph{Pulling}}}
76: \newcommand{\epull}{\text{\emph{Towing}}}
77: \newcommand{\lelo}{\text{\emph{Release}}}
78: \newcommand{\tjump}{\text{$\lp$--\emph{Quench}}}
79: \newcommand{\push}{\text{\emph{Pushing}}}
80:
81:
82: \DeclareMathOperator{\sgn}{sgn}
83: \DeclareMathOperator{\erf}{erf}
84:
85:
86: \sloppy
87:
88: \preprint{LMU-ASC 33/06}
89:
90: \begin{document}
91: \bibliographystyle{apsrev}
92:
93:
94: \title{Tension dynamics in semiflexible polymers. \\Part~I:
95: Coarse-grained equations of motion}
96:
97:
98: \author{Oskar Hallatschek} \email{ohallats@fas.harvard.edu}
99: \affiliation{Lyman Laboratory of Physics, %
100: Harvard University, Cambridge, Massachusetts 02138, USA}
101:
102: \author{Erwin Frey} \affiliation{Arnold Sommerfeld Center for Theoretical Physics
103: and Center for NanoScience, %
104: LMU M\"unchen, Theresienstr.~37, 800333 M\"unchen,
105: Germany}
106:
107: \author{Klaus Kroy}
108: \affiliation{Institut f\"ur Theoretische Physik,%
109: Universit\"at Leipzig, Augustusplatz 10/11, 04109 Leipzig, Germany}
110:
111:
112: \date{\today}
113:
114:
115: \begin{abstract}
116: Based on the wormlike chain model, a coarse-grained description of the nonlinear
117: dynamics of a weakly bending semiflexible polymer is developed. By means of a
118: multiple scale perturbation analysis, a length-scale separation inherent to the
119: weakly-bending limit is exploited to reveal the deterministic nature of the
120: spatio-temporal relaxation of the backbone tension and to deduce the corresponding
121: coarse-grained equation of motion. From this partial integro-differential
122: equation, some detailed analytical predictions for the non-linear response of a
123: weakly bending polymer are derived in an accompanying paper
124: (Part~II~\cite{hallatschek-part2:2006}).
125: \end{abstract}
126:
127:
128: \pacs{ 87.15.He, 87.15.Aa, 87.16.Ka, 83.10.-y}
129:
130:
131: \maketitle
132:
133:
134:
135:
136: \section{Introduction}
137: \label{sec:intro}
138:
139:
140: The laws of Brownian motion have played the role of a mediator between the apparently
141: smooth deterministic dynamics on a macroscopic scale and the microscopic molecular
142: chaos since their discovery a century ago~\cite{FreyK05}. They are pivotal to our
143: understanding of a broad class of animate and inanimate soft condensed matter systems
144: that owe their characteristic softness to low-dimensional and strongly fluctuating
145: meso-scale structures such as polymeric networks and membranes~\cite{GardelSMMMW04}.
146: Conversely, these systems are well suited to study how complex deterministic dynamics
147: on a macro- or meso-scale emerges from the underlying stochastic differential
148: equations~\cite{baschnagel:00}.
149:
150: Take, for example, a stiff polymer like actin that is suddenly stretched by strong
151: forces applied at its ends. Or, conversely, consider a polymer that is held in a
152: virtually straight conformation and suppose that the forces at its ends are suddenly
153: released. These two paradigmatic experimental setups, which we call \pull\/ and
154: \lelo, are illustrated in Fig.~\ref{fig:pull-release}. How will the end-to-end
155: distance of the polymer relax to its new equilibrium value? Given the manifestly
156: stochastic underlying dynamics, which for \lelo\/ is exclusively driven by thermal
157: forces, it is not immediately obvious that the initial contraction or stretching
158: dynamics should obey a deterministic law, as tacitly assumed by several heuristic
159: derivations~\cite{seifert-wintz-nelson:96,ajdari-juelicher-Maggs:97,morse:98II,everaers-Maggs:99,brochard-buguin-de_gennes:99}.
160: Indeed, these studies, which predicted a variety of interesting new dynamic scaling
161: regimes, arrived at partially contradicting
162: results~\cite{seifert-wintz-nelson:96,brochard-buguin-de_gennes:99}. Despite
163: considerable experimental, theoretical and numerical work, both relaxation laws (for
164: \pull\/ and \lelo) remained controversial for some time.
165:
166: It therefore appears worthwhile to undertake a detailed mathematical derivation of
167: the meso-scale dynamic equations that govern the nonlinear dynamics of semiflexible
168: polymers ``from first principles''~\cite{baschnagel:00}; i.e. from the underlying
169: stochastic differential equations of motion. In a recent Letter, we have outlined
170: such a systematic approach that resolves the aforementioned theoretical problems,
171: together with some of its consequences~\cite{Hallatschek:F:K::94:p077804:2005}. The
172: present contribution offers a more comprehensive discussion. In Part~I, an effective
173: coarse-grained meso-scale description of the dynamics of a semiflexible polymer is
174: developed by means of a multiple scale theory from the stochastic differential
175: equations of motion. Our detailed analysis also reveals the limits of validity of the
176: deterministic mesoscopic description and shows how to deal with subtle end effects
177: that may in some cases mask the non-trivial predictions for certain observables.
178: Building on this general framework, the theory is elaborated for the specific
179: problems of {\pull} and {\lelo} in Part~II~\cite{hallatschek-part2:2006}, which provides a template for the future
180: analysis of a variety of related problems with somewhat different boundary/initial
181: conditions~\cite{obermayer-hallatschek-frey-kroy:tbp,obermayer-hallatschek2:tbp,obermayer-kroy-frey-hallatschek3:tbp}.
182: Thereby, we corroborate the importance of a regime of homogeneous tension relaxation,
183: which generally occurs in {\lelo}-experiments, and establish the complete crossover
184: scenario between the various intermediate scaling regimes. Additionally, in Part~II
185: some consequences for common observables are worked out in detail to facilitate
186: experimental verification of the theory.
187:
188:
189: \begin{figure}
190: \centerline{\includegraphics[width=\columnwidth] {pull-release-psfraged.eps}}
191: \caption{Two basic examples of dynamic force-extension
192: experiments. {\pull}: a weakly bending polymer in equilibrium is
193: suddenly pulled longitudinally by two external forces
194: $\fe$. {\lelo}: a pre-stretched
195: polymer is suddenly released.}
196: \label{fig:pull-release}
197: \end{figure}
198:
199:
200: Before entering a detailed quantitative analysis, it seems useful to summarize the
201: main ideas on a qualitative level in order to make the remainder more easily
202: accessible. A characteristic property of semiflexible polymers and many other
203: fluctuating meso-scale structures in soft condensed matter, is their reduced
204: dimensionality or slender shape. It entails the presence of thermally excited
205: transverse fluctuations of an essentially inextensible backbone. Returning to the
206: above example, an actin filament is much more susceptible to bending undulations than
207: to stretching or compressing its backbone. An analogous statement holds holds for
208: other biopolymers or two-dimensional locally flat objects like membranes and
209: surfaces. This suggests to idealize these structures as \emph{undulating
210: inextensible manifolds}~\cite{2002nelson}. To be specific, we focus for the
211: following on the case of a single semiflexible polymer in solution, which seems to be
212: the simplest paradigmatic example of the more general soft-matter meso-structures
213: alluded to above. Its equilibrium mechanical properties and conformational statistics
214: have by now been thoroughly studied theoretically and
215: experimentally~\cite{frey-kroy-wilhelm-sackmann:97,yamakawa97}. Both are well
216: understood in terms of the self-affine roughness of the equilibrium contour
217: fluctuations within the so-called \emph{wormlike chain model}, which idealizes the
218: polymer as an inextensible space curve with an energetic cost for
219: bending~\cite{saito-takahashi-yunoki:67}. Here, we are primarily interested in the
220: transient non-equilibrium stretching and contraction dynamics of such a wormlike
221: chain, i.e., in the much less studied problem of how a semiflexible polymer relaxes
222: to equilibrium after a sudden drastic change in its boundary conditions. This
223: question is of considerable fundamental and practical interest alike, e.g.\ for
224: single-molecule
225: manipulations~\cite{perkins-smith-chu:97,quake-babcock-chu:97,CollinRJSTB05} and for
226: understanding and controlling the dynamic response of polymer solutions and networks
227: such as those constituting the cytoskeleton of biological
228: cells~\cite{GaborForgacs06012004,JanmeyW04}.
229:
230:
231:
232:
233: Because of the inextensible backbone, the stretching or contraction dynamics of a
234: wormlike chain is entirely due to a spatio-temporal re-organization of the contour
235: length stored in the transverse thermal wrinkles. It is governed by the dynamic
236: \emph{backbone tension} $f(s,t)$, which is the force that holds the backbone
237: together. The dynamics is always assumed to be strongly overdamped by solvent
238: friction, which can be decomposed into transverse and longitudinal components with
239: respect to the local tangent, in view of the locally rod-like conformation of the
240: polymer. At first sight, one might suppose that for small enough undulations one can
241: resort to a formulation of the dynamics solely in terms of transverse modes for which
242: only transverse friction matters. Actually, for the equilibrium dynamics in the
243: absence of external forces, the conclusions based on such an assumption are in accord
244: with a series of experimental
245: data~\cite{caspi-etal:98,legoff-hallatschek-frey:02,Shusterman:A:G:K::92:p048303:2004,hohenadl-etal:99}.
246: At a second thought, considering the constraint of inextensibility, it is far from
247: obvious how the interplay between transverse and longitudinal friction limits the
248: relaxation after a sudden application or release of external forces as e.g.\ in
249: \lelo\/ and \pull. It is the major objective of the present study to resolve this
250: puzzle for the case of longitudinal forces, while the somewhat more complex issue of
251: the nonlinear transverse response will be addressed in a separate
252: contribution~\cite{obermayer-hallatschek2:tbp}.
253:
254:
255: As a cornerstone of our derivation, we establish a dynamic scale separation between
256: the scales where transverse and longitudinal friction reign, namely the transverse
257: and longitudinal \emph{dynamic correlation lengths} $\lpe(t)$ and $\lpa(t)$,
258: respectively. We demonstrate that $\lpe(t)\ll \lpa(t)$ holds at any time in the limit
259: of a \emph{weakly bending rod}. As a central result, we obtain that the tension
260: varies to leading order only on the large scale $\lpa(t)$, over which the
261: short-wavelength transverse undulations that dominate the dissipation on the scale
262: $\lpe(t)$ are self-averaging. The short-scale transverse fluctuations may thus be
263: said to provide an effective local backbone compliance for the large-scale
264: longitudinal dynamics. In this way, transverse and longitudinal dynamics are
265: effectively decoupled, and the microscopic stochastic differential equations can be
266: reduced, in a controlled way, to a deterministic (integro-differential) equation for
267: the coarse-grained meso-scale dynamics.
268:
269:
270: With some care, most of the predictions that emanate from this reduced description
271: can qualitatively be obtained from a relatively simple scaling analysis, to which we
272: will occasionally resort in order to promote the intuitive physical understanding of
273: our analysis. In fact, not only physical insights but also several interesting formal
274: predictions were originally obtained on the basis of simple scaling arguments.
275: However, due to some subtleties and pitfalls, the results available in the literature
276: remained somewhat contradictory. The following development, in particular Part~I of
277: this article series, aims at settling the corresponding issues conclusively by
278: employing a controlled perturbation theory instead of evoking scaling assumptions.
279:
280: As suggested by the introductory examples of dynamic force-extension experiments,
281: exemplary realizations of the meso-scale dynamics occur in response to highly
282: localized forces, which we generally (though not invariably) assume to be applied
283: abruptly at the boundaries. They initiate universal self-similar relaxation
284: processes that spread through the polymer, resulting in characteristic power-law
285: signatures, so-called intermediate asymptotics~\cite{barenblatt:96}, in various
286: observables. These will be derived analytically in Part~II~\cite{hallatschek-part2:2006} of this article, which is
287: devoted to solving the effective deterministic meso-scale equations for the tension,
288: obtained below. There, we will also consider the consequences of the tension
289: dynamics on pertinent observables like the (projected) end-to-end distance and the
290: novel experimental perspectives brought up by our analysis. For the impatient reader,
291: our Letter~\cite{Hallatschek:F:K::94:p077804:2005} may serve as a quick guide to our
292: main arguments and results.
293:
294:
295: The outline of the present part (Part~I), is as follows. We develop
296: the systematic coarse grained description of tension dynamics in
297: stiff, respectively, stretched semiflexible polymers that
298: emerges from an appropriate small gradient expansion of the dynamical
299: wormlike chain model (Sec.~\ref{sec:dwmc}). An ordinary perturbation
300: expansion leading to linear first order equations of motion
301: (Sec.~\ref{sec:linear-dynamics}) turns out to be restricted to times
302: when the tension has already relaxed to its equilibrium value
303: (Sec.~\ref{sec:intro-tensprop-sfp}). The actual tension dynamics on
304: shorter times is resolved by a stochastic multiple scale perturbation
305: theory (Sec.~\ref{sec:multi-scale}), which is based on a dynamical
306: length scale separation. As a major result, we obtain a rigorous
307: deterministic partial integro-differential equation (PIDE) that
308: describes the long wave-length (-all time-) dynamics of the tension.
309:
310:
311:
312: \section{Dynamical wormlike chain model}
313: \label{sec:dwmc}
314:
315:
316: At low Reynolds numbers, the dynamics of a polymer is determined by the balance of
317: elastic forces, friction, and stochastic forces. We will briefly motivate how these
318: forces are modeled in the usual stochastic description of the Brownian motion of a
319: semiflexible polymer, leading to Eq.~(\ref{eq:eom-sfp}) below, which is the basis of
320: our subsequent analysis.
321:
322: The natural model for the description of the elastic properties of a semiflexible
323: polymer arises from the idea to regard the polymer as a thin cylinder consisting of a
324: homogeneous elastic material. In the slender limit, where the ratio of thickness
325: over total length approaches zero, the deformation modes of the cylinder become much
326: stiffer in the axial direction (``phonons'') than in directions transverse to the
327: cylinder axis (``bending modes'')~\cite{landau-lifshitz-7}. Consequently, such a
328: slender rod or ``thread'' subject to external forces merely allows for bending
329: deformations, while its contour length is to a good approximation locally conserved.
330:
331: These features are idealized in the wormlike chain model, where the
332: polymer is at any time $t$ represented as an inextensible space curve
333: $\vec r(s,t)$ parameterized by the arc length $s$, i.e.\ the tangents
334: have to obey the constraint
335: \begin{equation}
336: \label{eq:inextensibility}
337: \vec r'^2=1 \;.
338: \end{equation}
339: Here, we have introduced the shorthand notation $\vec r'\equiv
340: \partial \vec r(s,t)/\partial s$. The effective free energy $\Hb$ of a
341: particular conformation is only due to bending (curvature) and is
342: given by
343: \begin{equation}
344: \label{eq:bending-energy}
345: \Hb= \frac\kappa 2\mint{ds}{0}{L} \vec r''^2 \;,
346: \end{equation}
347: where $\kappa$ is the bending stiffness. To assure that the integrand is the square
348: of the local curvature, the inextensibility constraint,
349: Eq.~(\ref{eq:inextensibility}), has to be enforced as a rigid constraint.
350:
351: The elastic force (per arclength) $\gel$ derives from Eq.~(\ref{eq:bending-energy})
352: by a functional derivative,
353: \begin{equation}
354: \label{eq:elastic-force-general}
355: \gel(s,t)=-\left. \frac{\delta \Hb}{\delta \vec r }\right|_{\vec r'^2=1} \;.
356: \end{equation}
357: As indicated, the contour variations $\delta \vec r$ used to determine the functional
358: derivative on the right-hand-side have to respect the local inextensibility
359: constraint, Eq.~(\ref{eq:inextensibility}). The variational calculation, detailed in
360: App.~\ref{sec:elastic-forces}, yields
361: \begin{equation}
362: \label{eq:beq}
363: \vec{r}' \times \left( \kappa\, \vec{r}'''+\fel
364: \right)= 0 \;,
365: \end{equation}
366: where $\fel(s)\equiv \mint{d\tilde{s}}{0}{s} \gel(\tilde{s})$ is the
367: spatial integral over the elastic force density defined in
368: Eq.~(\ref{eq:elastic-force-general}).
369:
370: According to the implicit equation for $\fel$, Eq.~(\ref{eq:beq}), the force
371: $\kappa\, \vec{r}'''+\fel$ has a vanishing component transverse to the local tangent
372: $\vec{r}'$. Equivalently, we may require that both vectors are proportional to each
373: other,
374: \begin{equation}
375: \label{eq:beq-2}
376: \kappa\, \vec{r}''' +\fel = f\, \vec{r}' \;,
377: \end{equation}
378: where the proportionality factor $f(s)$ has dimensions of a force and, in general,
379: depends on the arc length. A spatial derivative then yields a direct expression for
380: the elastic force density,
381: \begin{equation}
382: \label{eq:beq-3}
383: \gel = -\kappa\, \vec{r}'''' + (f \vec{r}')' \;.
384: \end{equation}
385: The apparent simplification with respect to Eq.~(\ref{eq:beq}) is somewhat deceiving,
386: since Eq.~(\ref{eq:beq-3}) still contains the unknown function $f(s)$, which has to
387: be fixed by the local arc-length constraint, Eq.~(\ref{eq:inextensibility}). The new
388: force field $f(s)$ has however a very direct and intuitive physical interpretation.
389: It is the \emph{local line tension}~\cite{JMDeutsch05131988} that ``ties the polymer
390: together'' and thereby enforces the inextensibility condition.
391:
392: Much of the following deals with the dynamics of this local line tension, which turns
393: out to be closely related to the dynamics of the local excess length stored in
394: undulations, defined in Eq.~(\ref{eq:t-t-corr-fct}) below, to which we will refer to
395: as \emph{stored length}. Time-dependence is introduced into the description by
396: requiring the elastic force density $\gel$ to be balanced by the dynamic friction
397: (per arc length) with the solvent,
398: \begin{equation}
399: \label{eq:gex-local-friction}
400: \gfl(s,t)=-\vec \zeta(\vec r,t) \partial_t \vec r(s,t) \qquad \text{(free
401: draining)}\;.
402: \end{equation}
403: At any arclength $s$ and time $t$, the friction matrix $\vec \zeta$
404: with elements $\zeta_{ij}$ can be decomposed into its transverse and
405: longitudinal components with respect to the local tangent $\vec
406: r'(s,t)$,
407: \begin{equation}
408: \label{eq:friction-tensor}
409: \vec \zeta=\left[\zeta_\perp \left(1-\vec r'\otimes \vec r'
410: \right)+\zeta_\pa \;\vec r' \otimes \vec r' \right] \;.
411: \end{equation}
412: The constants $\zeta_\perp$ and $\zeta_\pa$ can to a first
413: approximation be estimated by the friction coefficients (per length)
414: for transverse respectively longitudinal motion of a \emph{rigid}
415: slender rod in a solvent of viscosity $\eta$~\cite{doi-edwards:86},
416: \begin{equation}
417: \label{eq:tfriction}
418: \zeta_\pe=2 \zeta_\pa \simeq 4\pi\eta \;.
419: \end{equation}
420: A more sophisticated analysis would consider logarithmic
421: corrections~\cite{batchelor:70,cox:70a,cox:70b} to account for the dynamic coupling
422: of distant chain segments $s$ and $s'$ via long-ranged hydrodynamic
423: interactions~\cite{degennes:79,doi-edwards:86}. While such logarithmic factors may
424: sometimes be crucial in a quantitative comparison with some
425: experiments~\cite{hohenadl-etal:99,legoff-hallatschek-frey:02}, their (feasible) implementation is not of
426: primary interest to our present discussion, so that we chose to dismiss them for
427: greater clarity of the presentation.
428:
429: On the same level of approximation, the force balance of elastic and frictional
430: forces can be extended by adding thermal white noise fully characterized in terms of
431: its mean and variance,
432: \begin{subequations}
433: \begin{eqnarray}
434: \label{eq:zero-mean}
435: \avg{\xi_i(s,t)}&=&0\;,\\
436: \label{eq:noise-corrs}
437: \avg{ \xi_i(s,t) \xi_j(s',t')}&=&2 k_B T \zeta_{ij} \delta(s-s')
438: \delta(t-t')\;
439: \end{eqnarray}
440: \end{subequations}
441: with the angular brackets indicating an ensemble average. The strength of the noise
442: correlations in Eq.~(\ref{eq:noise-corrs}) is dictated by the
443: Fluctuation--Dissipation theorem~\cite{kubo83bookI,kubo91bookII}, which assures that
444: the steady state of the stochastic equations of motion correspond to thermal
445: equilibrium (for complications in simulations of discrete bead-rod chains, see
446: Ref.~\cite{Fixman78,Hinch94,MorseD04}).
447:
448:
449:
450: Upon using Eqs.~(\ref{eq:beq-3},~\ref{eq:gex-local-friction}), the balance of elastic,
451: friction and stochastic forces, $0=\gel+\gfl+\vec \xi$, takes the form of an equation
452: of motion,
453: \begin{equation}
454: \label{eq:eom-sfp}
455: \vec \zeta \partial_t \vec r=- \kappa\, \vec r'''' + (f \vec
456: r')'+\vec \xi\;.
457: \end{equation}
458: The partial differential equation Eq.~(\ref{eq:eom-sfp}), the arc length constraint
459: Eq.~(\ref{eq:inextensibility}) and the Gaussian noise-correlation comprise a
460: complete stochastic description~\footnote{Since the thermal forces have a physical
461: origin, Eq.~(\ref{eq:eom-sfp}) has to be interpreted according to
462: Stratonovi$\hat{c}$~\cite{vanKampen:book}.} of the Brownian dynamics in the free
463: draining limit.
464:
465: \section{Linearized stochastic dynamics}
466: \label{sec:linear-dynamics}
467: The nonlinear stochastic dynamics of a wormlike chain, represented by
468: Eq.~(\ref{eq:eom-sfp}), in combination with the inextensibility constraint,
469: Eq.~(\ref{eq:inextensibility}), is hard to analyze in general. The difficulties are
470: largely due to the calculation of the line tension $f(s,t)$, which has to enforce the
471: local inextensibility constraint, Eq.~(\ref{eq:inextensibility}). There have been
472: attempts to relax the local constraint and merely enforce local inextensibility on
473: average~\cite{ha-thirumalai:97a,lee-thirumalai:04,harnau-winkler-reineker:95}. This
474: corresponds to replacing the field $f(s,t)$ by a (spatially averaged) mean field.
475: However those approaches fail to describe semiflexible polymers on local scales,
476: where tension fluctuations (in time and space) are
477: important~\cite{seifert-wintz-nelson:96}.
478:
479:
480:
481: \begin{figure}
482: \centerline{\includegraphics[width=\columnwidth]
483: {flucting-line-psfraged.eps}}
484: \caption{Typical conformations of a freely fluctuating stiff polymer with
485: suppressed rotations. In order to generate these conformations, which serve
486: illustrative purposes only, we have represented each conformation as a linear
487: superposition of the first $10$ modes that solve the linearized equation of
488: motion. For each realization, the mode amplitudes were then drawn randomly
489: according to their Gaussian distribution in equilibrium. }
490: \label{fig:flucting-line}
491: \end{figure}
492:
493:
494: Our analytical approach is instead based on the weakly bending limit, in which the
495: polymer conformation is appropriately described by its deviations from a straight
496: line. Accordingly, we parameterize the polymer's space curve by
497: \[\vec r=(\vec r_\pe,s-r_\pa)^T \qquad\text{(cf. Fig.~\ref{fig:flucting-line}),}\] where $\vec r_\pe(s)$
498: and $r_\pa(s)$ are the transverse and longitudinal displacements at arc length $s$.
499: Such parameterization is only useful if the gradients of the transverse displacements
500: are small everywhere along the contour,
501: \begin{equation}
502: \label{eq:wbl-sfp}
503: \vec r_\pe'^2=\Ord{\epsilon}\ll 1 \;,\qquad s\in (0,L)\;.
504: \end{equation}
505: The weakly bending limit, as defined in Eq.~(\ref{eq:wbl-sfp}),
506: assumes that there is a small parameter $\epsilon$ that controls the
507: polymer roughness \emph{uniformly} along the contour. In the case of a
508: stiff polymer, on which we focus if otherwise stated, the small
509: parameter is provided by the ratio
510: \begin{equation}
511: \label{eq:wbl-sfp-2}
512: \epsilon\equiv \frac L{\lp}=\frac{L k_B T}{\kappa}\ll1 \qquad
513: \text{(stiff polymer)}
514: \end{equation}
515: of length $L$ over persistence length $\ell_p$. Alternatively, a weakly bending
516: conformation may be realized by applying a static external stretching force $\fe$
517: larger than the internal characteristic force scale $k_B T/\lp$. In this case, $\vec
518: r_\pe'^2\sim\sqrt{k_B T/\lp\fe}$~\cite{marko-siggia:95} and we may identify
519: \begin{equation}
520: \label{eq:stretched-p}
521: \epsilon\equiv \sqrt{k_B T/\lp\fe}\qquad \text{(stretched polymer).}
522: \end{equation}
523:
524:
525:
526:
527: Eq.~(\ref{eq:wbl-sfp}) allows us to expand the dynamical equations of motion in terms
528: of small gradients. We start with the inextensibility constraint, which ``slaves''
529: the higher order longitudinal displacements to the transverse ones. After resolving
530: Eq.~(\ref{eq:inextensibility}) for $r_\pa'$ and expanding the square root, the local
531: constraint takes the simple form
532: \begin{equation}
533: \label{eq:constraint-wbl-sfp}
534: \rpa'=\frac1 2 \vec r_\pe'^2+\Ord{\epsilon^2}=\Ord{\epsilon}\;.
535: \end{equation}
536: This entails that the parameter $\epsilon$ is a measure for the reduction of the
537: longitudinal extension
538: \begin{equation}
539: \label{eq:rpa-def}
540: R_\pa\equiv L-r_\pa(L)+r_\pa(0)
541: \end{equation}
542: of the polymer due to the presence of thermal undulations: An arclength integral of
543: Eq.~(\ref{eq:constraint-wbl-sfp}) shows that $R_\pa$ is smaller then the total
544: contour length $L$ by an amount of the order
545: \begin{equation}
546: \label{eq:contraction}
547: L-R_\pa=r_\pa(L)-r_\pa(0)=\Ord{\epsilon L}\;.
548: \end{equation}
549: We may think of the length difference in Eq.~(\ref{eq:contraction}) as being
550: ``stored'' in undulations. The distribution of this excess length along the filament
551: is, according to Eq.~(\ref{eq:constraint-wbl-sfp}), described by the function
552: \begin{equation}
553: \label{eq:t-t-corr-fct}
554: \varrho(s,t)\equiv \frac12 \vec{ r}_\pe'^2(s,t)= \Ord{\epsilon}\;,
555: \end{equation}
556: which will have central importance in our analysis. It is the fraction of the contour
557: length that is at arclength $s$ and time $t$ \emph{locally} stored in undulations.
558:
559:
560: After taking a ``spatial'' (i.e.\ arc-length) derivative of its longitudinal part, we
561: expand the equation of motion, Eq.~(\ref{eq:eom-sfp}), to order
562: $\Ord{r_\pe'^2}=\Ord{\epsilon}$ and obtain
563: \begin{subequations}
564: \label{eq:sfeom}
565: \begin{eqnarray}
566: \label{eq:perp-eq-motion-sfp}
567: \partial_t \vec r_\pe&=&-\vec r_\pe''''+ (f \vec
568: r_\pe')'+\vec{\gex}_\pe+\boldsymbol \xi_\pe \\
569: \hat \zeta \partial_t r'_\pa&=&(\hat \zeta-1)(\vec r_\pe'
570: \partial_t\vec r_\pe)'- r_\pa'''''-f''+ (f r_\pa')''\nonumber\\
571: &&-\gex_\pa'-\xi_\pa'
572: \label{eq:parallel-eq-motion-sfp}
573: \;.
574: \end{eqnarray}
575: \end{subequations}
576: Here, we have neglected terms of order $\Ord{\epsilon^{3/2}}$ and made the following
577: choice of units: time and tension, respectively, are rescaled according to
578: \begin{eqnarray}
579: \label{eq:rescaling-time-tension}
580: t&\to&\zeta_\perp t/\kappa \\
581: f&\to& \kappa f\;.
582: \end{eqnarray}
583: This corresponds to setting $\kappa\equiv\zeta_\pe\equiv 1$ and $\hat \zeta\equiv
584: 1/2=\zeta_\pa$. As a consequence all variables represent powers of length, e.g., $t$
585: and $f$ are a length$^4$ and a length$^{-2}$, respectively. In Eq.~(\ref{eq:sfeom})
586: we have further allowed for an external~\footnote{Throughout, we reserve Gothics for
587: \emph{external} forces or force fields like $\vec \gex$.} force density $\vec \gex=(\vec \gex_\pe,\gex_\pa)$, which is a
588: length$^{-3}$, that may for instance represent the effect of a solvent flow, an
589: optical/magnetical tweezer or an electrical field.
590:
591: As long as one is taking the limit $\epsilon\to 0$ with $t,s,L$ fixed, the polymer is
592: correctly described by the linearized version of Eq.~(\ref{eq:sfeom}),
593: \begin{subequations}
594: \label{eq:eom-low}
595: \begin{align}
596: \label{eq:eompe-low}
597: \partial_t \vec r_\pe&=-\vec r_\pe''''+ (f \vec
598: r_\pe')'+\vec{\gex}_\pe+\vec \xi_\pe \\
599: \label{eq:eompa-low}
600: f''&=-\gex_\pa' \;,
601: \end{align}
602: \end{subequations}
603: From the magnitude of the noise-correlations, given by Eq.~(\ref{eq:noise-corrs}), it
604: may be inferred that $\vec \xi_\pe$ to leading order obeys
605: \begin{equation}
606: \label{eq:noise-corrs-low}
607: \avg{\vec \xi_\pe(s,t) \vec \xi_\pe(s',t')}=2 (\bold I/\lp)\delta(s-s')
608: \delta(t-t')\;,
609: \end{equation}
610: where we used $k_B T \zeta_\pe\to\lp^{-1}$ in our choice of units and $\bold
611: I_{ij}=\delta_{ij}$ is the identity matrix.
612:
613: From Eq.~(\ref{eq:eompa-low}), it is seen that the tension profile of
614: a polymer that is forced at the ends is linear; slope and offset are
615: fixed by the boundary conditions. The (higher order) longitudinal
616: displacements are slaved to the transverse ones by the arc length
617: constraint, Eq.~(\ref{eq:constraint-wbl-sfp}). At the present level of
618: approximation the exact equations of motion have reduced to a linear
619: equation for the transverse displacements alone.
620:
621: \subsection{Generalized transverse response}
622: \label{cha:LT}
623:
624: In many practical cases, for instance if the polymer is symmetrically
625: pulled apart by a (possibly time-dependent) force, the tension
626: $f=f(t)$ is to lowest order spatially homogeneous such that the
627: equation of motion, Eq.~(\ref{eq:eompe-low}), reduces to
628: \begin{equation}\label{eq:eom1}
629: \partial_t \vec r_\pe = -\vec r_\pe''''+f \vec r_\pe'' + \vec\xi_\pe
630: \;.
631: \end{equation}
632: Let us anticipate at this point that we will identify an inherent length scale
633: separation, Eq.~(\ref{eq:lscale-sep}) in Sec.~\ref{sec:multi-scale} below, according
634: to which the tension can be considered as slowly varying in space, such that
635: Eq.~(\ref{eq:eom1}) describes the polymer dynamics \emph{locally} (and even globally
636: at late times). As a consequence, the solution of Eq.~(\ref{eq:eom1}) for a given
637: spatially homogeneous tension history $f(t)$ becomes an important ingredient of the
638: nonlinear theory and shall be analyzed in the following.
639:
640:
641: Formally, the linear Langevin equation Eq.~(\ref{eq:faltung}) is solved by
642: \begin{equation}
643: \label{eq:faltung}
644: \vec r_\pe(s,t)=\mint{ds'}{0}{L}\mint{dt'}{-\infty}{\infty}\chi(s,s';t,t')
645: \,\vec \xi_\pe(s',t') \;.
646: \end{equation}
647: The Green's function $\chi(s,s';t,t')$ satisfies
648: \begin{eqnarray}
649: \label{eq:scb-bc}
650: \partial_t\chi(s,s';t,t')&=&-\partial_s^4\chi(s,s';t,t') + f(t)
651: \partial_s^2 \chi(s,s';t,t')\nonumber \\ &&+\delta(s-s')
652: \delta(t-t')
653: \end{eqnarray}
654: and appropriate boundary conditions. It may be interpreted as a causal response
655: function that describes the spreading and the decay of contour undulations induced by
656: a transverse force impulse at location $s'$ and (elapsed) time $t'$.
657: Eq.~(\ref{eq:faltung}) therefore can be said to represent the conformation at time
658: $t$ in terms of the accumulated response to the transverse noise history $\vec
659: \xi_\pe(s',t')$ along the contour.
660:
661: In the general case of a time-dependent tension, it can be quite difficult to
662: determine the Green's function $\chi$ that obeys the prescribed boundary conditions,
663: because eigenmodes and eigenvalues of the linear operator
664: $(f(t)+\partial_s^2)\partial_s^2$ depend on the value of $f(t)$ and thus become
665: time-dependent. In terms of Fourier modes, on the other hand, a translationally
666: invariant Green's function $\overline{\chi}(s-s';t,t')$ can easily be found, below.
667: As will be detailed in Sec.~\ref{sec:intro-tensprop-sfp}, this function describes the
668: universal bulk dynamics far away from the ends, while a correction term that
669: manifestly breaks translational invariance has to be added ``close'' to the ends to
670: correct for the actual boundary effects.
671:
672: To formalize this decomposition into bulk and boundaries, the full
673: response function $\chi$ may be written as a superposition
674: \begin{equation}
675: \label{eq:superpose}
676: \chi(s,s';t,t')=\overline{\chi}(s-s';t,t')+\chi_{bc}(s,s';t,t') \;,
677: \end{equation}
678: where $\overline{\chi}(s-s';t,t')$ and $\chi_{bc}(s,s';t,t')$ represent a
679: translationally invariant part and the boundary correction, respectively. The former
680: is taken to satisfy
681: \begin{equation}
682: \label{eq:trans-inv-eom}
683: \partial_t\overline{\chi}(s;t,t')=[-\partial_s^2+f(t)]\partial_s^2\overline{\chi}(s;t,t') +\delta(s) \delta(t-t')
684: \end{equation}
685: on an infinite arc length interval. With help of Fourier modes,
686: \begin{equation}
687: \label{eq:FT}
688: \overline{\chi}(q;t,t')\equiv \mint{ds}{-\infty}{\infty}
689: \overline{\chi}(s;t,t') e^{-i q s} \;,
690: \end{equation}
691: Eq.~(\ref{eq:trans-inv-eom}) reads
692: \begin{equation}
693: \label{eq:scb}
694: \partial_t\overline{\chi}(q;t,t')+\lambda(q,t)\overline{\chi}(q;t,t') = \delta(t-t') \;,
695: \end{equation}
696: where $\lambda(q,t)$ is the dispersion relation~\footnote{The dispersion relation
697: Eq.~(\ref{eq:disp-rel}) is easily extended to accommodate a general time dependent
698: transverse harmonic confinement potential represented by an additional spring
699: constant on the RHS of Eq.~(\ref{eq:disp-rel}).}
700: \begin{equation}
701: \label{eq:disp-rel}
702: \lambda(q,t)=q^4+f(t)q^2 \;.
703: \end{equation}
704: By the method of integrated factors, the solution to Eq.~(\ref{eq:scb}) is found to
705: be~\footnote{We note aside, that $\overline{\chi}^2(q;t,0)$ is identical to the
706: ``amplification factor'' introduced in Ref.~\cite{hallatschek-frey-kroy:04} to
707: describe the growth of the squared mode amplitudes.}
708: \begin{equation}
709: \label{eq:chi-q}
710: \overline{\chi}(q;t,t')=\Theta(t-t') \exp\left[-\mint{d\hat t}{t'}{t}\lambda(q,\hat t)\right]\;,
711: \end{equation}
712: which may be checked by direct substitution. The real space susceptibility is
713: given by the inverse Fourier transform of Eq.~(\ref{eq:chi-q}),
714: \begin{equation}
715: \label{eq:chi-real}
716: \overline{\chi}( s;t,t')=\mint{\frac{dq}{\pi}}{0}{\infty}\overline{\chi}(q;t,t')\cos(q s) \;,
717: \end{equation}
718: where it has been used that $\overline{\chi}(q;t,t')$ is even in $q$.
719:
720:
721: The part $\chi_{bc}(s,s';t,t')$ of the susceptibility, which is not translationally
722: invariant, satisfies the homogeneous differential equation
723: \begin{equation}
724: \label{eq:homo-lin-pde}
725: \partial_t\chi_{bc}=[-\partial_s^2+f(t)]\partial_s^2 \chi_{bc} \;,
726: \end{equation}
727: and has boundary conditions that have to compensate for the generally inappropriate
728: behavior of $\overline{\chi}(s-s';t,t')$ at the boundaries. The difficulty of
729: solving Eq.~(\ref{eq:scb-bc}) has been shifted to $\chi_{bc}(s,s';t,t')$. However,
730: for the time-dependent quantities to be studied below, this boundary term represents
731: a relevant contribution only within a characteristic length $\lpe(t)$ (defined below)
732: close to the boundaries. For times small enough, such that $\lpe(t)\ll L$, one may
733: use $\chi_{bc}$ derived on a semi-infinite polymer to approximate the situation near
734: one boundary, say the one at $s=0$. For simplicity, we will mostly refer to the
735: model boundary conditions of ``hinged'' (h) or ``clamped'' (c)
736: ends~\cite{WigginsROG98}, for which the full susceptibility on a semi-infinite
737: arclength interval $s\in (0,\infty)$ is given by a symmetric and antisymmetric
738: combination of the bulk susceptibility $\overline{\chi}$,
739: \begin{equation}
740: \label{eq:susceptibility-hinged-clamped}
741: \chi_{\frac hc}(s,s';t,t')= \overline{\chi}(s-s';t,t')\mp\overline{\chi}(s+s';t,t')\;.
742: \end{equation}
743: Evidently, $\chi_{\frac hc}$ satisfies Eq.~(\ref{eq:scb-bc}) on $s\in (0,\infty)$, as
744: well as the boundary conditions
745: \begin{eqnarray}
746: \label{eq:bcs-h-c}
747: \chi_h(0,s';t,t')=&0&=\partial_s^2\chi_h(0,s';t,t')\qquad
748: \text{(hinged)}\nonumber \\
749: \partial_s\chi_c(0,s';t,t')=&0&=\partial_s^3\chi_c(0,s';t,t')\qquad
750: \text{(clamped)}\;.\nonumber
751: \end{eqnarray}
752: A hinged end has vanishing transverse displacement and is torque free
753: (vanishing second derivative), whereas a (``gliding'') clamped
754: end has a vanishing slope and is force free (vanishing third
755: derivative).
756:
757:
758:
759:
760:
761: \subsection{Local response of the stored excess length }
762: \label{sec:stored-length}
763:
764: As a basis for our subsequent systematic analysis of tension dynamics and as an
765: application of the above results, we wish to determine the longitudinal motion
766: implied by the linearized transverse stochastic dynamics, Eq.~(\ref{eq:eompe-low}).
767: To this end, consider a stiff polymer equilibrated under a constant tension $f_<$ at
768: time zero, on which a \emph{spatially constant} tension $f(t)$ is imposed that varies
769: \emph{deterministically} for $t>0$. We ask, how the density of stored excess length
770: $\varrho(s,t)$, defined in Eq.~(\ref{eq:t-t-corr-fct}), changes in time by
771: considering the ensemble average of the increase (during the time interval $t$) of
772: the stored length
773: \begin{equation}
774: \label{eq:change-in-slength}
775: \Delta \varrho(s,t)\equiv\varrho(s,t)-\varrho(s,0) \;.
776: \end{equation}
777: This is an important observable, since it governs the leading order contribution to
778: the change \[\Delta R_\pa(t)\equiv R_\pa(t)-R_\pa(0)\] in the projected end-to-end
779: distance $R_\pa(t)$, defined in Eq.~(\ref{eq:rpa-def}), which is can be directly
780: measured in dynamic single polymer experiments~\cite{legoff-hallatschek-frey:02}.
781: Here, the (average) end-to-end axis of the polymer is assumed to be controlled by
782: external means; e.g.\ by an external force field, flow field, boundary conditions,
783: etc. To the relevant order, the precise measures taken to orient the polymer
784: (strictly or on average) do not matter, and we find
785: \begin{equation}
786: \label{eq:end-end-length}
787: \avg{\Delta R_\pa} (t)= -\mint{ds}{0}{L}\avg{\Delta\varrho}(s,t)+o(\epsilon)\;.
788: \end{equation}
789: Here, the notation with the arguments $s$ and $t$ outside the brackets of
790: $\avg{\Delta\varrho}$ was introduced to emphasize that, even after averaging, these
791: dependencies generally persist.
792:
793:
794:
795: For an explicit calculation of the stored length, we insert Eq.~(\ref{eq:faltung})
796: into Eq.~(\ref{eq:t-t-corr-fct}) and perform an ensemble average upon employing
797: Eq.~(\ref{eq:noise-corrs-low})
798: \begin{eqnarray}
799: \avg{\varrho}(s,t) &= & \mint{dt'}{-\infty}{t}\mint{ds'}{0}{L}\mint{d\tilde
800: t'}{-\infty}{t}\mint{d\tilde s'}{0}{L}\partial_s \chi(s,s';t,t')\nonumber\\
801: &&\partial_s\chi(s,\tilde s';t,\tilde
802: t')\avg{\vec \xi_\pe(s',t') \cdot \vec \xi_\pe(\tilde s',\tilde t')}/2\nonumber \\
803: &=& 2 \mint{\frac{dt'}{\lp(t')}}{-\infty}{t}\mint{ds'}{0}{L}\partial_s \chi(s,s';t,t')^2
804: \;. \label{eq:varrho-full}
805: \end{eqnarray}
806: For later convenience, we have allowed for a \emph{time-dependent} persistence length
807: $\lp(t)$ and an optional prestress $f_<\gg L^2$. The latter is also technically
808: advantageous, since it acts as a physical regularization to suppress modes with
809: wavelength larger than the total length. It enables us to take the total length to
810: infinity and to discuss the stored length $\avg{\varrho}(s,t)$ on a semi-infinite
811: arclength interval. For our ultimate goal of calculating $\avg{\Delta \varrho}(s,t)$
812: an intrinsic regularization renders modes with wave length beyond a characteristic
813: length scale $\lpe(t)$ irrelevant, so that $f_<$ can eventually be set to zero if
814: required.
815:
816: Inserting Eq.~(\ref{eq:susceptibility-hinged-clamped}), valid for hinged/clamped
817: boundary conditions, into Eq.~(\ref{eq:varrho-full}) (with $L\to\infty$) yields
818: \begin{eqnarray}
819: \avg{\Delta\varrho_{\frac hc}}&=&\mint{\frac{dq}{\pi}}{0}{\infty}
820: \left[ \hat\varrho(q,t)-\hat\varrho(q,0) \right] \left[
821: 1\pm \cos(2 q s) \right] \nonumber\\
822: \hat\varrho(q,t) &=&2q^2 \mint{d\tilde t}{-\infty}{t}
823: \overline{\chi}^2(q;t,\tilde t)/\lp(\tilde t)\;. \label{eq:general-stolength-b}
824: \end{eqnarray}
825: The general expression Eq.~(\ref{eq:general-stolength-b}) is now specialized to the
826: scenario
827: \begin{equation}
828: \label{eq:force-history}
829: \left.
830: \begin{aligned}&\quad \text{tension } & \quad\text{persistence length} \\
831: t<0&:\quad f_< = \text{const.} & \lp= \text{const.} \qquad \\ t>0&:\quad f(t) & \lp/\theta = \text{const.} \qquad
832: \end{aligned}
833: \right.
834: \end{equation}
835: As preparation for a more general discussion, we have included the possibility of a
836: sudden change in persistence length by a factor $1/\theta>0$ at $t=0$.
837:
838: With a constant tension at negative times, the time-integral in
839: Eq.~(\ref{eq:general-stolength-b}) can be evaluated from $\tilde t=-\infty$ to
840: $\tilde t=0$, after which we obtain
841: \begin{eqnarray}
842: \label{eq:slength-modes-final}
843: \hat \varrho(q,t)\,\lp&=&\frac{\overline{\chi}^2(q;t,0)}{q^2+f_<}+
844: 2 \theta q^2\mint{d\tilde t}{0}{t} \overline{\chi}^2(q;t,\tilde t)\;.
845: \end{eqnarray}
846: Therewith, the full expression for the average change in stored excess length density
847: becomes
848: \begin{eqnarray}
849: \avg{\Delta\varrho_{\frac hc}}(s,t)&= &\mint{\frac{dq}{\pi\lp}}{0}{\infty}
850: \left\{\frac{1}{q^2+f_<}\left(e^{-2q^2[q^2 t+ F(t)]}-1\right)\right. \nonumber\\
851: & &+\left. 2 \theta
852: q^2\mint{d\tilde t}{0}{t}e^{-2q^2\left[q^2(t-\tilde
853: t)+F(t)-F(\tilde t)\right]}\right\}\nonumber\\
854: & &\left[ 1\pm\cos(2 q s) \right]\;, \label{eq:delta-rho-mega-expression}
855: \end{eqnarray}
856: where $F(t)$ is the integral of the tension over positive times,
857: \begin{equation}
858: \label{eq:F}
859: F(t)=\mint{d\hat t}{0}{t} f(\hat t) \;.
860: \end{equation}
861: One observes that the integral in Eq.~(\ref{eq:delta-rho-mega-expression}) is well
862: defined, even for $f_<=0$, because the integrand vanishes for $q\to 0, \infty$. It is
863: dominated by wave numbers for which the exponents become of order one, i.e., the
864: dominant modes have wave numbers $q$ for which the characteristic relaxation time
865: \begin{equation}
866: \label{eq:relaxation-times}
867: \tau_q=(q^4 +q^2 F/t)^{-1}
868: \end{equation}
869: is of the order of $t$. This suggests to define a characteristic length $\lpe(t)$ as
870: the wave length for which the relaxation time is just $t$,
871: \begin{equation}
872: \label{eq:transverse-corr-length}
873: 1=\lpe^{-2}\left[\lpe^{-2} t +F(t)\right] \;.
874: \end{equation}
875: Asymptotically $\lpe(t)$ is given by
876: \begin{equation}
877: \label{eq:lpe-asymptotically}
878: \lpe(t)\sim\left\{ {t^{1/4}\;, \text{ for }t\ll (F/t)^{-2}
879: \atop F^{1/2} \;, \text{ for }t\gg
880: (F/t)^{-2}}\right. \;.
881: \end{equation}
882:
883:
884: \begin{table}
885: \caption{The transverse equilibration length
886: $\lpe(t)$ and the tension propagation length
887: $\lpa(t)$ both exhibit a crossover at a time $t_\fe\equiv\fe^{-2}$, which depends on
888: the external force $\fe$
889: (here, for the {\pull} problem with $\fe\gg L^{-2}$).}
890: \label{tab:pulling-growth-laws}
891: \begin{center}
892: \begin{tabular}{r|cc|c} &
893: $\lpe(t)$ & & $\lpa(t)$ \qquad \quad \mbox{}\\ \hline
894: $t\ll t_\fe$ & $t^{1/4}$ & & $t^{1/8} (\lp/ \zeta )^{1/2}$
895: \hfill~\cite{morse:98II,everaers-Maggs:99}\\
896: $t\gg t_\fe$ & $t^{1/2}\fe^{1/2}$ & & $t^{1/4}\fe^{1/4}
897: (\lp/\zeta)^{1/2}$ \quad ~\cite{seifert-wintz-nelson:96}
898: \end{tabular}
899: \end{center}
900: \end{table}
901:
902:
903: Due to the competition between bending forces ($\propto r_\perp\lpe^{-4}$) and
904: tension ($\propto r_\perp (F/t) \lpe^{-2}$), the growth of $\lpe$ thus exhibits a
905: dynamic crossover from free relaxation ($\lpe\simeq t^{1/4}$) to relaxation under
906: tension ($\lpe\simeq \sqrt{F}$) at a characteristic time $t_\fe\equiv (F/t)^{-2}$
907: (Table~\ref{tab:pulling-growth-laws}/left for a constant tension equal to the
908: external force $\fe$).
909:
910:
911:
912: As we will explicitely demonstrate in the next section, the change $\avg{\Delta
913: \varrho_{\frac hc}}$ in stored length saturates at a constant value for distances
914: to the boundaries much larger than the characteristic length $\lpe(t)$. This ``bulk''
915: value is given by~\footnote{This equation for the stored length at a spatially
916: constant but time-dependent tension has recently independently been derived in a
917: related context~\cite{bhobot-wiggins-granek:04}.}
918: \begin{eqnarray}
919: \avg{\Delta\overline{\varrho}}(t)&\equiv& \mint{\frac{dq}{\pi\lp}}{0}{\infty}
920: \left\{\frac{1}{q^2+f_<}\left(e^{-2q^2[q^2 t+ F(t)]}-1\right) \right.\nonumber\\
921: &&+\left.2 \theta
922: q^2\mint{d\tilde t}{0}{t}e^{-2q^2\left[q^2(t-\tilde
923: t)+F(t)-F(\tilde t)\right]}\right\}. \label{eq:change-stored-length}
924: \end{eqnarray}
925: The quantity $\avg{\Delta\overline{\varrho}}(t)$ will be central in our systematic
926: analysis of tension propagation and relaxation, because it turns out to determine the
927: local curvature of the tension profile. Each of the two terms inside the curly
928: brackets of Eq.~(\ref{eq:change-stored-length}) have a direct physical
929: interpretation. Since the parameter $\theta$ tunes the strength of the thermal
930: kicks, it is seen that the $\theta$-independent first term represents the
931: deterministic change in the excess length that is stored in mode $q$ in absence of
932: any stochastic force. For pulling forces $F>0$ its sign is always negative, since
933: both the internal elastic and the external driving forces act to straighten the
934: filament. On the contrary, thermal kicks represented by the (strictly positive)
935: second term are favoring undulations.
936:
937:
938: \section{Breakdown of ordinary perturbation theory}
939: \label{sec:intro-tensprop-sfp}
940: The previous sections employed ``ordinary'' perturbation theory (OPT) in the small
941: parameter $\epsilon$, leading to a linear equation of motion to lowest order. As
942: detailed below, the use of OPT is, however, limited to long times even for
943: $\epsilon\ll1$. The predictions derived above for the longitudinal segment motion
944: turn out to be incompatible with the longitudinal force balance on short times. In
945: particular, Eq.~(\ref{eq:change-stored-length}) reveals an infinite longitudinal
946: friction for $t\to0$. This section extends a heuristic argument of
947: Ref.~\cite{everaers-Maggs:99} to resolve this problem. The following, furthermore,
948: elucidates a very general feature of the non-linear response, namely, a crossover
949: from ``weak-'' to ``strong-force'' behavior. Finally, it reveals the crucial
950: length-scale separation underlying our subsequent systematic analysis.
951:
952: The breakdown of OPT becomes evident when we try to use
953: Eq.~(\ref{eq:delta-rho-mega-expression}) to evaluate the longitudinal segment motion
954: in a non-equilibrium situation. For definiteness and as a telling example, let us
955: consider an initially equilibrated polymer that is suddenly pulled longitudinally by
956: a constant force $\fe$ at both ends, i.e., at time $t=0$, the tension at the ends is
957: suddenly increased from $0$ to a given positive value $\fe$,\[\fe=f(0,t>0)=f(L,t>0)
958: \quad\text{({\pull}-scenario).}\] As a consequence of Eq.~(\ref{eq:eompa-low}), the
959: tension $f=\fe \Theta(t)$ is to lowest order spatially uniform and fixed by the
960: driving force $\fe$ at the boundaries. Our above result for the change in stored
961: length due to a given tension history, Eq.~(\ref{eq:delta-rho-mega-expression}), thus
962: applies and evaluates to
963: \begin{eqnarray}
964: \label{eq:drho-pull-1}
965: \avg{\Delta\varrho_{\frac hc}}(s,t)&=&\mint{\frac{dq}{\pi\lp}}{0}{\infty}
966: \frac{\fe}{q^2(q^2+\fe)}\left[ e^{-2q^2(q^2+\fe)t} -1 \right]\nonumber\\
967: &&\times\left[1\pm\cos(2 q s)\right] \;.
968: \end{eqnarray}
969: After the variable substitutions $k\equiv q\fe^{-1/2}$, $\sigma\equiv s \fe^{1/2}$ and
970: $\tau\equiv \fe^2 t=t/\tf$, Eq.~(\ref{eq:drho-pull-1}) takes the form
971: \begin{equation}
972: \label{eq:scform-1}
973: \avg{\Delta
974: \varrho_{\frac hc}}(s,t)=\lp^{-1}\fe^{-1/2}\Sigma_{\frac hc}(\sigma,\tau)
975: \end{equation}
976: with the
977: scaling function
978: \begin{eqnarray}
979: \label{eq:drho-pull-2}
980: \Sigma_{\frac hc}(\sigma,\tau)&=&\mint{\frac{dk}{\pi}}{0}{\infty}
981: \frac{1}{k^2(k^2+1)}\left[ e^{-2k^2(k^2+1)\tau} -1 \right]\nonumber \\
982: & &\times\left[1\pm\cos(2 k \sigma)\right] \;.
983: \end{eqnarray}
984: Another variable substitution $\tilde k=k \tau^{1/4}$ generates factors $\tilde
985: k^4+\tilde k^2\tau^{1/2}$ in exponent and denominator, which can be replaced by
986: $\tilde k^4$ in the asymptotic limit $\tau\ll 1$. Just as for the dispersion relation
987: Eq.~(\ref{eq:relaxation-times}), tensile forces $\propto k^2$ may be neglected as
988: compared to the dominant bending forces $\propto k^4$ for times smaller than the
989: crossover time $\tf$. In the opposite limit $\tau\gg 1$, the reverse approximation
990: applies. After a variable substitution $\tilde k=k \tau^{1/2}$, factors $\tilde k^4
991: \tau^{-1}+\tilde k^2$ appear, which may be replaced by $\tilde k^2$. We thus find
992: that the two-parameter scaling form Eq.~(\ref{eq:drho-pull-2}) collapses onto
993: one-parameter scaling forms for small and large times,
994: \begin{eqnarray}
995: \label{eq:scf-stimes}
996: \Sigma_{\frac hc}&\sim& -\tau^{3/4}\Sigma_{\frac hc}^<(\sigma
997: \tau^{-1/4})\;,\qquad \tau\to 0 \nonumber\\
998: \label{eq:scf-ltimes}
999: \Sigma_{\frac hc}&\sim& -\tau^{1/2}\Sigma_{\frac hc}^>(\sigma
1000: \tau^{-1/2})\;,\qquad \tau\to \infty \nonumber
1001: \end{eqnarray}
1002: with scaling functions given by
1003: \begin{eqnarray}
1004: \label{eq:drho-pull-3}
1005: \Sigma_{\frac hc}^<(\xi)&=&\mint{\frac{d\tilde k}{\pi}}{0}{\infty}\frac{1-e^{-2 \tilde
1006: k^4}}{\tilde k^4}\left[ 1\pm\cos(2 \tilde k \xi) \right]\;,\\
1007: \label{eq:drho-pull-4}
1008: \Sigma_{\frac hc}^>(\xi)&=&\mint{\frac{d\tilde k}{\pi}}{0}{\infty}\frac{1-e^{-2 \tilde
1009: k^2}}{\tilde k^2}\left[ 1\pm\cos(2 \tilde k \xi) \right]\;.
1010: \end{eqnarray}
1011: Note that the spatial part of these scaling functions depicted in
1012: Fig.~\ref{fig:blayer-pull-opt} decays to zero within several rescaled time units. As
1013: an important consequence, we note that the part of $\avg{\Delta \varrho_\frac hc}$
1014: that depends on the boundary conditions really only matters close to the boundaries,
1015: i.e., up to a distance for which the scaling variable becomes of order one. In fact,
1016: this distance can be identified with $\lpe(t)$, as defined in
1017: Eq.~(\ref{eq:transverse-corr-length}), because the scaling variable of $\Sigma$ is
1018: given by $\sigma \tau^{-1/4}=s/t^{1/4}=s/\lpe(t)$ for $\tau\ll1$ and $\sigma
1019: \tau^{-1/2}=s/\lpe(t)$ for $\tau\gg1$, respectively.
1020:
1021: \begin{figure}
1022: \includegraphics[width=\columnwidth]
1023: {boundary-layers-psfraged.eps}
1024: \caption{The absolute value $\abs{\Delta\Sigma^{\gtrless}_{\frac
1025: hc}}$ of the spatially varying part $\Delta\Sigma^{\gtrless}_{\frac
1026: hc}\equiv\Sigma^{\gtrless}_{\frac hc}(\xi)-\Sigma^{\gtrless}_{\frac hc}
1027: (\infty)$ of the scaling functions $\Sigma_{\frac hc}^\gtrless$ defined in
1028: Eqs.~(\ref{eq:drho-pull-3}, \ref{eq:drho-pull-4}), which happens
1029: to be the same for hinged/clamped boundary conditions.}
1030: \label{fig:blayer-pull-opt}
1031: \end{figure}
1032:
1033: The bulk of the polymer stores length according to the universal part
1034: $\avg{\Delta\overline{\varrho}}(t)$ (independent of the boundary condition), which
1035: asymptotically takes the form
1036: \begin{equation}
1037: \label{eq:DR-opt}
1038: \avg{\Delta \overline\varrho}(t)\sim \left\{ { \Gamma(7/4)^{-1} \fe
1039: \lp^{-1} (t/2)^{3/4}
1040: \text{~\cite{granek:97}} \;,\quad \text{for } t\ll \tf \atop
1041: \lp^{-1} (2\fe t/\pi)^{1/2} \;, \quad \text{for } t\gg \tf}
1042: \right. \;.
1043: \end{equation}
1044: For alternative derivations of the short-time linear response law $\sim t^{3/4}$, see
1045: Refs.~\cite{granek:97,gittes-mackintosh:98_pub,morse:98II}. The crossover time
1046: $\tf\equiv \fe^{-2}$ is the time where the external force $\fe$ equals the Euler
1047: buckling force $\lpe^{-2}(t)$ corresponding to the correlation length $\lpe(t)$.
1048:
1049: These results comprise the predictions of ordinary perturbation theory (OPT) to
1050: leading order. As evident from Eq.~(\ref{eq:eom1}), longitudinal friction forces have
1051: thereby been completely neglected, because they are of higher order in $\epsilon$.
1052: However, on the semi-infinite arclength interval considered here, a spatially
1053: constant change in stored length - no matter how small - implies via
1054: Eq.~(\ref{eq:end-end-length}) an infinitely fast change $\partial_t \avg{\Delta
1055: R_\pa}$ of the longitudinal extension and thus friction. Hence, OPT must fail on
1056: sufficiently large length scales. Let us define $L_\star(t)$ as the length scale
1057: beyond which OPT breaks down for a given time $t$. In the present case of a suddenly
1058: applied pulling force, this critical length can be estimated from the physical
1059: requirement that the total longitudinal friction is not only finite but at most equal
1060: to the driving force $\fe$, i.e.,
1061: \begin{equation}
1062: \label{eq:condition}
1063: \fe\stackrel{!}{\simeq} \hat \zeta L \partial_t
1064: \avg{\Delta R_\pa}=-\hat \zeta L\mint{ds}{0}{L}\partial_t\avg{\Delta\varrho}(s,t)\;.
1065: \end{equation}
1066: For an average stored length given by the spatially
1067: constant value Eq.~(\ref{eq:DR-opt}), this condition is met if the polymer length is
1068: smaller than
1069: \begin{eqnarray}
1070: \label{eq:critical-length}
1071: L_\star(t) &\equiv&\sqrt{\fe/(\hat \zeta\partial_t\avg{\Delta\overline{\varrho}})} \nonumber\\
1072: &\simeq&\left\{{ \sqrt{\lp/\hat \zeta}\;t^{1/8} \;,\qquad \text{for } t\ll \tf \atop
1073: \sqrt{\lp /\hat \zeta}\;(\fe t)^{1/4} \;, \qquad \text{for } t\gg \tf}\right.\;.
1074: \end{eqnarray}
1075: OPT can thus only be valid on length scales (much) smaller than $L_\star(t)$. The
1076: condition
1077: \begin{equation}
1078: \label{eq:tstar}
1079: L_\star(t_\star)=L \;,
1080: \end{equation}
1081: implicitly defines the critical time scale $t_\star$ above which OPT applies to the
1082: whole polymer, and below which OPT is limited to subsections shorter than
1083: $L_\star(t)$.
1084:
1085: In summary, the omission of tension propagation has been identified as the reason for
1086: the breakdown of the OPT predictions. On a heuristic level~\cite{everaers-Maggs:99},
1087: the problem can thus be resolved by requiring that the applied tension is not
1088: immediately perceptible everywhere in the filament, but propagates a finite distance
1089: $\lpa(t)$ during time $t$ from the ends towards the bulk of the filament. Hence, only
1090: segments up to a distance $\lpa(t)$ from the ends are set into longitudinal motion.
1091: If the length $\lpa(t)$ over which the tension varies is smaller than the critical
1092: length $L_\star(t)$, it is ensured that the longitudinal friction does not exceed the
1093: driving force. In heuristic discussions, it was generally assumed that both lengths
1094: can be identified up to numerical factors of order one,
1095: \begin{equation}
1096: \label{eq:heuristic-argument}
1097: \lpa(t)\simeq L_\star(t)\;,\;t\ll t_\star\;,\quad \text{(heuristic hypothesis)}
1098: \end{equation}
1099: as summarized in Table \ref{tab:pulling-growth-laws}/right. The scaling assumption
1100: in Eq.~(\ref{eq:heuristic-argument}) turns out to reproduce the correct scaling for most
1101: cases (with, however, interesting exceptions elaborated in Part~II~\cite{hallatschek-part2:2006}, as well as, in
1102: Ref.~\cite{obermayer-kroy-frey-hallatschek3:tbp}). The ``weak- and strong- force''
1103: limits $\lpa\propto t^{1/8}$ and $\lpa\propto t^{1/4}$ of
1104: Refs.~\cite{everaers-Maggs:99,seifert-wintz-nelson:96} are with
1105: Eqs.~(\ref{eq:critical-length},~\ref{eq:heuristic-argument}) recovered as ``short- and
1106: long-time'' asymptotics. The crossover at time $\tf$ signals the change from ``free''
1107: to ``forced'' relaxation and is inherited from the one of the scaling function
1108: $\Sigma$, defined in Eq.~(\ref{eq:drho-pull-2}).
1109:
1110: Finally, we note that the tension may be considered as ``slowly''
1111: varying in space because $\lpa(t) \propto \Ord{\epsilon^{-1/2}}$ is
1112: for $\epsilon\to0$ larger than any length that does not dependent on
1113: the small parameter (for $t,$ $\fe,$ $L$ fixed). As it turns
1114: out, the most important length of the latter type is the dynamic
1115: correlation length $\lpe(t)$ for transverse
1116: displacements. Namely, the \emph{scale separation}
1117: \begin{equation}
1118: \label{eq:lscale-sep}
1119: \lpa/\lpe=\Ord{\epsilon^{-1/2}} \gg 1\;
1120: \end{equation}
1121: indicates that the tension is nearly constant on the equilibration length scale
1122: $\lpe(t)$ for transverse displacements. Intuitively, it should be clear that this
1123: simplifies the further analysis considerably, because it allows to apply (certain)
1124: results locally that are derived for spatially constant tension. Formally,
1125: Eq.~(\ref{eq:lscale-sep}) lends itself as a starting point for a \emph{multiple-scale}
1126: calculus, which separates the physics on different scales to obtain an improved
1127: perturbation expansion that is regular in the limit $t\to0$ while $\epsilon\ll1$ is
1128: fixed~\footnote{It may be remarked that the encountered contradiction does not
1129: appear, if one takes the limit $\epsilon\to0$ while the parameters $L$ and $t$ are
1130: held fixed. This corresponds to a lower temperature, respectively, a stiffer
1131: polymer. Then, the physically motivated requirement of the external force
1132: exceeding the total longitudinal friction force is satisfied for small enough
1133: $\epsilon\ll1$. The same conclusion may be drawn from the cross-over length scales
1134: $L_\star$ in Eq.~(\ref{eq:critical-length}) upon eliminating $\lp$ in favor of the
1135: small parameter $\epsilon$, because $L_\star(\epsilon,t) >L$ for given $L$ and $t$
1136: and small enough $\epsilon$. In a mathematical sense, the expansion generated by
1137: ordinary perturbation is pointwise asymptotic in $t$, but not
1138: uniformly~\cite{hinch:91}: the smaller $t$ the smaller $\epsilon$ has to be for the
1139: expansion to be asymptotic.}. The procedure, detailed in the next section, is
1140: similar in spirit to the procedure for athermal rod
1141: dynamics~\cite{hallatschek-frey-kroy:04}, but some complications related to the
1142: stochastic nature of the equations of motion have to be faced. The final result will
1143: be an effective deterministic description of the tension on the macro-scale
1144: $\lpa(t)$, where the stochasticity on the micro-scale $\lpe(t)$ has been integrated
1145: out.
1146:
1147:
1148:
1149: \section{Multiple scale analysis}
1150: \label{sec:multi-scale}
1151: We introduce a rapidly and a slowly varying arc length coordinate, $x\equiv s$ and
1152: $y\equiv s \epsilon^{\gamma}$, respectively, where the exponent $\gamma>0$ will be
1153: fixed later. The dynamic functions $\vec r_\pe$ and $f$ are now considered to depend
1154: on both variables $\{f,\vec r_\pe\}\to \{f(x,y),\vec r_\pe(x,y)\}$, where $x$ and $y$
1155: are treated as independent. The original arc length derivative of those functions
1156: then becomes
1157: \begin{equation}
1158: \label{eq:spatio-derivative}
1159: \partial_s \equiv \partial_x|_{y} + \epsilon^\gamma
1160: \partial_y|_{x} \;.
1161: \end{equation}
1162: The noise $\vec \xi=\Ord{\epsilon^{1/2}}$ being the source of any
1163: transverse displacements suggests an expansions of the dynamic
1164: variables $\vec r_\perp$ and $f$ in powers of $\epsilon^{1/2}$,
1165: \begin{eqnarray}
1166: \label{eq:power-series}
1167: \vec r_\perp(x,y)&=&\epsilon^{1/2} \vec h_1(x,y)+ o\left(\epsilon^{1/2}\right)
1168: \;, \nonumber \\
1169: f(x,y) &=& f_0(x,y)+\epsilon^{1/2} f_1(x,y)+\epsilon f_2(x,y)\nonumber \\
1170: & &+o(\epsilon) \;.
1171: \end{eqnarray}
1172: In the case of isotropic friction (i.e.~$\zeta_\pe=\zeta_\pa$), the stochastic forces
1173: have no intrinsic scale. Hence, they can only depend on the microscopic variable,
1174: $\vec \xi=\epsilon^{1/2}\vec \xi_1(x)$. In the anisotropic case, the friction forces
1175: and, hence, the stochastic forces, are coupled to the orientation of the filament, so
1176: that one has to assume a power expansion
1177: \begin{equation}
1178: \label{eq:noise-expansion}
1179: \begin{split}
1180: \boldsymbol \xi_\pe(x,y)&=\epsilon^{1/2}\boldsymbol
1181: \xi_{\pe,1}(x)+o\left(\epsilon^{1/2}\right) \\
1182: \xi_\pa(x,y)&=\epsilon^{1/2} \xi_{\pa,1}(x)+\epsilon \xi_{\pa,2}(x,y)+o(\epsilon)
1183: \;.
1184: \end{split}
1185: \end{equation}
1186: The $y$-arguments in Eq.~(\ref{eq:noise-expansion}) are inherited from the
1187: $y$-arguments of $\vec r'$ entering the noise correlations in
1188: Eq.~(\ref{eq:noise-corrs}) via the friction matrix, defined in
1189: Eq.~(\ref{eq:friction-tensor}). The $y$-dependence would disappear for isotropic
1190: friction. Note, that the leading order $\vec \xi_1(x)$ still depends on the
1191: microscopic variable $x$ only, because the anisotropy merely enters the higher orders
1192: (a formal argument is given in App.~\ref{sec:noise-mspt}).
1193:
1194: In the following, it is crucial to require that the expansion coefficients in each
1195: order are bounded, so that we obtain a uniformly valid power
1196: expansion~\cite{holmes:95} in terms of the small parameter $\epsilon$. Inserting all
1197: expansions, Eqs.~(\ref{eq:noise-expansion},~\ref{eq:power-series}), into the
1198: equations of motion, Eq.~(\ref{eq:perp-eq-motion-sfp},
1199: \ref{eq:parallel-eq-motion-sfp}), yields
1200: \begin{subequations}
1201: \begin{align}
1202: 0&= \epsilon^{1/2}\left[\partial_t \vec h_1+\partial_x^4\vec
1203: h_1-\partial_x(f_0 \partial_x \vec
1204: h_1)-\boldsymbol \xi_{\pe,1}\right] \nonumber \\
1205: &\quad + o(\epsilon^{1/2}) \label{eq:multiscale-1} \\
1206: 0&=\partial_x^2 f_0 +\epsilon^\gamma 2 \partial_x\partial_y f_0
1207: +\epsilon^{1/2} \left[\partial_x^2
1208: f_1+\xi_{\pa,1}'\right]\nonumber\\
1209: &\quad +\epsilon^{2\gamma}\partial_y^2 f_0
1210: +\epsilon\left[\partial_x^2 f_2+X_2(x,y)+\xi_{\pa,2}'\right]
1211: \nonumber \\
1212: &\quad +\epsilon^{\gamma+1/2}\partial_y\partial_x f_1+
1213: o(\epsilon;~\epsilon^{2\gamma}) \;.
1214: \label{eq:multiscale-2}
1215: \end{align}
1216: \end{subequations}
1217: In order to arrive at Eq.~(\ref{eq:multiscale-2}) we used the local arc length
1218: constraint, Eq.~(\ref{eq:constraint-wbl-sfp}). By
1219: \begin{eqnarray}
1220: \label{eq:nl}
1221: X_2(x,y)&=&\frac{\hat \zeta}{2}\partial_t(\partial_x \vec h_1)^2+(1-\hat
1222: \zeta)\partial_x\left[(\partial_x \vec h_1) (\partial_t\vec
1223: h_1)\right]\nonumber\\
1224: &&+\frac{1}{2}\partial_x^4\left(\partial_x \vec
1225: h_1\right)^2-\frac{1}{2}\partial_x^2\left[f_0\left(\partial_x\vec
1226: h_1\right)^2 \right]
1227: \end{eqnarray}
1228: we have summarized terms nonlinear in $\vec h_1$. The first term in $X_2$, which is
1229: proportional to $ \hat \zeta$, accounts for the longitudinal friction and is thus
1230: responsible for the short-time divergence encountered in the heuristic discussion of
1231: Sec.~\ref{sec:intro-tensprop-sfp}.
1232:
1233: The $\Ord{1}$ part of Eq.~(\ref{eq:multiscale-2}), $\partial_x^2
1234: f_0=0$, together with the requirement of $f_0$ being bounded for large
1235: $x$, implies that
1236: \begin{equation}
1237: \label{eq:x-indep-tension}
1238: f_0(x,y)=\hat f_0(y)
1239: \end{equation}
1240: is independent of $x$. Hence both the $\Ord{1}$ and the $\Ord{\epsilon^\gamma}$ term
1241: of Eq.~(\ref{eq:multiscale-2}) vanish. Requiring the $\Ord{\epsilon^{1/2}}$
1242: coefficient to be zero fixes $f_1$ up to an integration constant. The value of $f_1$
1243: does not affect the evolution of $\vec h_1$ because it does not enter the
1244: $\Ord{\epsilon}$ coefficient in Eq.~(\ref{eq:multiscale-1}). Thus, the precise value
1245: of $f_1$ does not change the pathological behavior of longitudinal friction $\propto
1246: \hat \zeta \partial_t (\partial_x \vec h_1)^2$, which is why we shift the discussion
1247: of $f_1$ to App.~\ref{sec:f1}. From the latter we merely need that
1248: $\partial_y\partial_x f_1(x,y)=0$ which renders the
1249: $\Ord{\epsilon^{\gamma+1/2}}$-term in Eq.~(\ref{eq:multiscale-2}) zero. Then the next
1250: higher order is either $\Ord{\epsilon^{2\gamma}}$ or $\Ord{\epsilon}$ depending on
1251: the value of $\gamma$. With Eq.~(\ref{eq:x-indep-tension}) we can solve the
1252: $\Ord{\epsilon^{1/2}}$ part of Eq.~(\ref{eq:multiscale-1}) for $\vec h_1(x,y)$ along
1253: the lines of Sec.~\ref{cha:LT} and use the result to evaluate $X_2(x,y)$. It then
1254: turns out that the first term in $X_2$ (the longitudinal friction) would require
1255: $f_2$ to grow without bound with increasing system size to render the
1256: $\Ord{\epsilon}$-expansion coefficient in Eq.~(\ref{eq:multiscale-2}) finite, if they
1257: were required to balance each other. This represents the same unphysical divergence
1258: that is responsible for the breakdown of ordinary perturbation theory, discussed in
1259: Sec.~\ref{sec:intro-tensprop-sfp}.
1260:
1261: In order to obtain an improved perturbation theory, we attempt to balance the
1262: nonlinear term by the $\Ord{\epsilon^{2\gamma}}$ term after choosing $\gamma=1/2$;
1263: i.e., the exponent $\gamma$ is fixed such that the expansion coefficient $f_2$
1264: remains finite in the semi-infinite system considered here~\footnote{We note aside
1265: that the small parameter $\epsilon^\gamma=\epsilon^{1/2}$ appearing here is the
1266: same as in the length scale separation, Eq.~(\ref{eq:lscale-sep}), observed in
1267: Sec.~\ref{sec:intro-tensprop-sfp}.}. The equation fixing $f_2$ thus reads
1268: \begin{eqnarray}
1269: \label{eq:phi1}
1270: \partial_x^2f_2(x,y)&=&-\partial_y^2\hat
1271: f_0(y)-X_2(x,y)-\xi_{\pa,2}'\;.
1272: \end{eqnarray}
1273: Given $h_1$ and $f_0$ the last equation can be solved for $f_2$
1274: \begin{eqnarray}
1275: \label{eq:fi-2}
1276: f_2(x,y)&=&\mint{d\tilde x}{0}{x}\mint{d\hat x}{0}{\tilde
1277: x}\left\{ -\partial_y^2\hat
1278: f_0(y)\right.\nonumber\\
1279: &&\left.-X_2(\hat x,y)-\xi_{\pa,2}'(\hat x) \right\} \,.
1280: \end{eqnarray}
1281: For $f_2$ to be bounded for large system sizes, we have to require
1282: \begin{equation}
1283: \label{eq:fixing-f0}
1284: \partial_y^2 \hat
1285: f_0(y)=\overline{-X_2-\xi_{\pa,2}'\,}^x(y) \;,
1286: \end{equation}
1287: where the over-line denotes the spatial average over the rapidly
1288: varying coordinate $x$,
1289: \begin{equation}
1290: \label{eq:xavg} \overline{g}^x(y)\equiv \lim_{l\to\infty}
1291: \mint{\frac{dx}{l}}{0}{l}g(x,y)
1292: \end{equation}
1293: for a function $g(x,y)$. The expansion coefficient $f_2$ would show a divergence
1294: quadratic in the system size if Eq.~(\ref{eq:fixing-f0}) was not satisfied. Hence,
1295: the $y$-dependence of $\hat f_0(y)$ must be fixed such that the expansion coefficient
1296: $f_2$ remains finite. For a finite polymer, the limit $l\to\infty$ is not to be taken
1297: literally, though. Rather, the average in Eq.~(\ref{eq:xavg}) is required to become
1298: independent of $l$ to leading order in $\epsilon$ for $l$ much smaller than the
1299: system size $L$.
1300:
1301:
1302: As it turns out, the only quantity in Eq.~(\ref{eq:fixing-f0}) that
1303: does not vanish upon $x$-averaging is the first term in
1304: $X_2$, the longitudinal friction. This is easily
1305: seen for all other terms in $X_2$ and the $f_1$-term. They are total
1306: derivatives with respect to $x$ of products of expansion coefficients
1307: that are (required to remain) bounded (non-secular~\cite{holmes:95})
1308: by definition. Hence, the $x$-integrals of those total derivatives
1309: are bounded and the $x$-averages vanish upon formally taking the
1310: coarse-graining length $l\to\infty$ in Eq.~(\ref{eq:xavg}). The
1311: noise term also represents a total derivative with respect to $x$. The
1312: average of that term represents a stochastic variable with an
1313: amplitude that scales as $1/l$ and, hence, also vanishes in the limit
1314: $l\to\infty$.
1315:
1316:
1317: Dropping all terms that vanish under coarse-graining and integrating
1318: over time, Eq.~(\ref{eq:fixing-f0}) takes the form
1319: \begin{eqnarray}
1320: \label{eq:cg-eom1}
1321: \frac1{\hat\zeta}\partial_y^2\hat F_0(y)&=&\frac12 \left
1322: [\overline{(\partial_x \vec h_1)^2}^x(y,0)-\overline{(\partial_x
1323: \vec h_1)^2}^x(y,t)\right] \nonumber \\
1324: &=&-\epsilon^{-1}\overline{\Delta\varrho}^s(t)\;,
1325: \end{eqnarray}
1326: where $\Delta \varrho(s,t)$ is the change in stored length, as defined in
1327: Eq.~(\ref{eq:change-in-slength}), of a semi-infinite polymer for the tension history
1328: \begin{equation}
1329: \label{eq:tension-history-msp}
1330: F(t)=\hat F_0(y,t) \;.
1331: \end{equation}
1332: Note that the dependence on the slowly varying arclength coordinate $y$ enters the
1333: tension history in Eq.~(\ref{eq:tension-history-msp}) only parametrically. The same
1334: holds true for the calculation of the right hand side of Eq.~(\ref{eq:cg-eom1}).
1335:
1336:
1337: The closed set of equations Eqs.~(\ref{eq:cg-eom1},~\ref{eq:multiscale-1}) represents
1338: the lowest order of the multiple scale perturbation expansion. It incorporates the
1339: feedback mechanism already found in the heuristic discussion above. The evolution
1340: of transverse displacements implies longitudinal motion via the arc length
1341: constraint, and, according to Eq.~(\ref{eq:cg-eom1}), the corresponding longitudinal
1342: friction sets the polymer under tension. This, in turn, feeds back onto the evolution
1343: of transverse displacements, Eq.~(\ref{eq:multiscale-1}), typically acting as to
1344: reduce the longitudinal friction.
1345:
1346: For solving Eqs.~(\ref{eq:cg-eom1},~\ref{eq:multiscale-1}) self-consistently, it
1347: would be handy to perform an ensemble average $\avg{\dots}$ on the right-hand-side
1348: of Eq.~(\ref{eq:cg-eom1}), because we could then apply expression
1349: Eq.~(\ref{eq:delta-rho-mega-expression}) for $\avg{\Delta\varrho}$. We argue that
1350: such an ensemble average is indeed justified, because $\overline{\Delta\varrho}^s(t)$
1351: is, in fact, a deterministic quantity as a consequence of the central limit theorem.
1352: Recall that the quantity $\avg{\Delta{\varrho}}(s,t)$ is dominated by transverse
1353: modes of the wave length $\lpe(t)$ given by Eq.~(\ref{eq:lpe-asymptotically}). Since
1354: wavelength much larger than $\lpe(t)$ are not relevant in the mode sum, the dynamic
1355: length $\lpe(t)$ can be interpreted as the correlation length of the change in stored
1356: length, i.e., the length over which the correlation function
1357: $C_{s,t}(z)\equiv\avg{\Delta\varrho(s,t)\Delta\varrho(s+z,t)}$ varies. Hence, the
1358: integral $X_l\equiv \mint{ds}{0}{l}\Delta\varrho(s,t)$ may be understood as the sum
1359: of $l/\lpe(t)$ weakly correlated random variables. For large $l$, the distribution
1360: function of $X_l$ thus becomes Gaussian, and the variance of $X_l$ grows linearly
1361: with the number of independent random variables, $\avg{(X_l-\avg{X_l})^2}\propto l$.
1362: As a consequence, the distribution of the average $X_l/l\to
1363: \overline{\Delta\varrho}^s$ approaches a delta function as $l\to\infty$.
1364:
1365: An ``additional'' ensemble average therefore does not change the value of the right
1366: hand side of Eq.~(\ref{eq:cg-eom1}). Evaluating the spatial average
1367: after the ensemble average levels out the boundary term of expression
1368: Eq.~(\ref{eq:delta-rho-mega-expression}) for $\avg{\Delta \varrho_{\frac hc}}$ and
1369: reduces it to its \emph{bulk} value $\avg{\Delta\overline{\varrho}}(t)$, defined in
1370: Eq.~(\ref{eq:change-stored-length}), which is completely independent of the boundary
1371: conditions. We thus have the important relation
1372: \begin{equation}
1373: \label{eq:self-averaging}
1374: \overline{\Delta\varrho_{\frac hc}}^s(t)=\overline{\avg{\Delta\varrho_{\frac
1375: hc}}}^s(t)=\avg{\Delta\overline{\varrho}}(t) \;.
1376: \end{equation}
1377: We would like to emphasize, that our argument for replacing the spatial by an
1378: ensemble average requires a finite driving force, such that
1379: $\lim_{l\to\infty}\overline{\Delta \varrho}^s$ approaches a finite value. This
1380: specifically excludes the linear response limit, i.e., the limit of vanishing
1381: external force $\fe\to0$ while $\epsilon\ll1$ is fixed. In this case, the tension
1382: dynamics has to be described by Eq.~(\ref{eq:cg-eom1}), which is \emph{stochastic}
1383: even to leading order (we will come back to this point in Part~II).
1384:
1385:
1386:
1387: Given the external driving is finite such that Eq.~(\ref{eq:self-averaging}) may be
1388: applied, Eq.~(\ref{eq:cg-eom1}) takes the form
1389: \begin{equation}
1390: \label{eq:cg-eom}
1391: \partial_s^2 F(s,t)=-\hat \zeta \avg{\Delta \overline{\varrho}}\left[F(s,\tilde t
1392: \leq t),t\right] \;,
1393: \end{equation}
1394: where we introduced $s=y\epsilon^{1/2}=x$ again and made the
1395: parametric dependence of $\avg{\Delta\overline{\varrho}}$ on the
1396: tension history explicit. The \emph{deterministic} tension dynamics,
1397: as described by Eq.~(\ref{eq:cg-eom}), provides the sought-after
1398: rigorous local generalization of the heuristic argumentation of
1399: Sec.~\ref{sec:stored-length}: local longitudinal motion is driven by
1400: tension gradients (like in a thread pulled through a viscous medium).
1401:
1402: Upon inserting our result in Eq.~(\ref{eq:change-stored-length}) for the
1403: right-hand-side, we obtain a nonlinear, partial integro-differential equation (PIDE)
1404: for the tension history $F(s,t)$,
1405: \begin{eqnarray}
1406: \label{eq:pide}
1407: &\partial_s^2 F(s,t)= \hat \zeta
1408: \mint{\frac{dq}{\pi\lp}}{0}{\infty}
1409: \left\{\frac{1}{q^2+f_<}\left(1-e^{-2q^2[q^2 t+ F(s,t)]}\right)
1410: \right.& \nonumber \\
1411: &\left. -2 \theta q^2\mint{d\tilde
1412: t}{0}{t}e^{-2q^2\left[q^2(t-\tilde t)+F(s,t)-F(s,\tilde
1413: t)\right]}\right\}\;.&
1414: \end{eqnarray}
1415: We have arrived at a closed description of the polymer dynamics to lowest
1416: order in MSPT that consists of two parts. On a length scale $\lpe(t)$, Brownian
1417: motion gives rise to fluctuations of transverse displacements that are described by
1418: the linear Eq.~(\ref{eq:multiscale-1}). This \emph{stochastic} differential
1419: equation, in turn, adiabatically depends on a tension profile that varies on a much
1420: larger scale $\lpa(t)$ and satisfies a \emph{deterministic} nonlinear equation of
1421: motion, Eq.~(\ref{eq:pide}).
1422:
1423:
1424: \begin{figure}
1425: \centerline{\includegraphics[width=\columnwidth] {tension-vs-friction-psfraged.eps}}
1426: \caption{Tensile forces acting on a polymer subsection of size $\Delta
1427: s$. Balancing these forces with the drag that arises from the longitudinal velocity
1428: of magnitude $\partial_t r_\pa(s,t)$, one estimates $\Delta s \zeta_\pa \partial_t
1429: r_\pa(s,t)\approx f(s-\Delta s/2)-f(s+\Delta s/2)$ in the weakly bending limit.
1430: For infinitesimal $\Delta s$, this becomes $\zeta_\pa \partial_t
1431: r_\pa(s,t)\approx-\partial_s f(s,t)$. A further spatial derivative yields
1432: $\zeta_\pa \partial_t\partial_s r_\pa(s,t)\approx-\partial_s^2 f(s,t)$, which is
1433: the time derivative of Eq.~(\ref{eq:cg-eom}) (in original units) up to a spatial
1434: and ensemble average on the left-hand side, which correspond to an adiabatic,
1435: respectively, equilibrium approximation. }
1436: \label{fig:tension-vs-friction}
1437: \end{figure}
1438:
1439: \section{Conclusion}
1440: \label{sec:conclusion1}
1441:
1442: We would like to conclude the present general discussion of tension dynamics with a
1443: simple physical interpretation of the outcome, Eq.~(\ref{eq:cg-eom},~\ref{eq:pide}),
1444: of our multiple scale analysis. Effectively, our MSPT analysis is a rigorous
1445: justification of certain approximations that can be made to analyze the tension
1446: dynamics in the small $\epsilon$ limit. According to Eq.~(\ref{eq:cg-eom}), the
1447: curvature of the integrated tension is (up to a constant) given by the ensemble
1448: average of the local stored length release. As shown in
1449: Fig.~\ref{fig:tension-vs-friction}, it may be conceived as (1) a force balance
1450: equation between the locally acting tensile and longitudinal friction force, in which
1451: (2) the latter may be computed from the equations of motion for the bulk of an
1452: equilibrated polymer under a spatially constant, though time-dependent, tension. The
1453: longitudinal friction force is obtained via taking a time-derivative of this
1454: coarse-grained \emph{dynamical} force extension relation,
1455: Eq.~(\ref{eq:change-stored-length}). The first order MSPT equation of motion neglects
1456: longitudinally acting bending forces, employs an adiabatic approximation and assumes
1457: local equilibrium. The latter fully retains memory effects and therefore must not be
1458: mistaken as an approximation of quasi-stationary dynamics, which, in
1459: Part~II~\cite{hallatschek-part2:2006}, will turn out to be a valid approximation only
1460: for specific driving forces in a particular time regime.
1461:
1462:
1463:
1464:
1465:
1466:
1467:
1468:
1469: \section{Acknowledgments}
1470: \label{sec:ack}
1471:
1472: It is a pleasure to acknowledge helpful conversations with Benedikt Obermayer. This
1473: research was supported by the German Academic Research Foundation (DAAD) through a
1474: fellowship within the Postdoc-Program (OH) and by the Deutsche
1475: Forschungsgemeinschaft through grant no.~Ha 5163/1 (OH) and SFB 486 (EF).
1476:
1477:
1478: \appendix
1479:
1480: \begin{table}[b]
1481: \caption{Some important notations}
1482: \label{tab:common-notation}
1483: \begin{ruledtabular}
1484: \begin{tabular}{c|l}
1485: Symbol(s) & General Meaning \\ \hline
1486: $L$ & total length of the worm-like chain \\
1487: $\kappa$ & bending stiffness \\
1488: $\epsilon$ & small parameter, defined such that $\vec r_\pe'^2=\Ord{\epsilon}$ \\
1489: $\vec r_\pe(s,t)$ & transverse displacement; $\vec r_\pe=\Ord{\epsilon^{1/2}}$ \\
1490: $r_\pa(s,t)$ & longitudinal displacement; $\vec r_\pa=\Ord{\epsilon}$ \\
1491: $\varrho(s,t)$ & stored-length density; $\varrho=\vec
1492: r_\pe'^2/2+\Ord{\epsilon^2}=\Ord{\epsilon}$ \\
1493: $R$ & end-to-end distance \\
1494: $R_\pa$ & end-to-end vector projected onto the long.~axis
1495: \\
1496: $\simeq$ & equal up to numerical factors of
1497: order $1$ \\
1498: $\propto$ & proportional to \\
1499: $\sim$ & asymptotically equal \\
1500: $\lp$ & persistence length \\
1501: $\theta$ & $\lp$ changes by a factor $1/\theta>0$ at $t=0$. \\
1502: $k_B T$ & thermal energy \\
1503: $f(s,t)$ & line tension \\
1504: $\lpe(t)$ & equilibration scale for transverse bending modes
1505: \\
1506: $\lpa(t)$ & scale of tension variations \\
1507: $\vec \fe$ & external force \\
1508: $\vec \gex$ & external force per arc length \\
1509: $\vec \xi(s,t)$ & thermal force per arc length
1510: \end{tabular}
1511: \end{ruledtabular}
1512: \end{table}
1513:
1514:
1515: \section{Elastic forces}
1516: \label{sec:elastic-forces}
1517:
1518:
1519:
1520:
1521: From the effective free energy of a worm-like chain, Eq.~(\ref{eq:bending-energy}),
1522: we seek to determine the elastic force $\gel(s)$ per arc length that a polymer of a
1523: given conformation $\bar{\vec r}(s)$ exerts at arc length $s$ onto its surroundings.
1524: To this end, let us assume the polymer was subject to a constant external force field
1525: equal to $-\gel(s)$, the negative of the local elastic forces. Then, the
1526: conformation $\bar{\vec r}(s)$ is in balance with the external force and minimizes
1527: the total free energy
1528: \begin{equation}
1529: \label{eq:Htot}
1530: \Htot[\vec r]=\Hb[\vec r]+\Hext[\vec r]
1531: \end{equation}
1532: with
1533: \begin{equation}
1534: \label{eq:Hext}
1535: \Hext[\vec r]\equiv \mint{ds}{0}{L}\gel(s) \,\vec r(s) \;,
1536: \end{equation}
1537: under all possible paths that obey the local inextensibility,
1538: Eq.~(\ref{eq:inextensibility}). Hence, requiring a vanishing free energy change
1539: $\delta \Htot$ for an infinitesimal (permitted) change $\delta \vec r$ in the space
1540: curve leads to the sought-after relation between the elastic forces and the contour.
1541: The central question of the minimization problem is how to deal with the local
1542: constraint.
1543:
1544:
1545: Here, we show that the common~\cite{goldstein-langer:95} introduction
1546: of a Lagrange multiplier function ensuring the local inextensibility
1547: constraint is not necessary. We will instead present a minimization
1548: procedure that considers only variations that obey the inextensibility
1549: constraint to leading order.
1550:
1551: Consider a test contour
1552: \begin{equation}
1553: \label{eq:tcontour}
1554: \vec r(s)=\bar{ \vec{r}}(s)+\delta\vec r(s)\;,
1555: \end{equation}
1556: which is infinitesimally displaced by $\delta \vec r(s)$ from the equilibrium contour
1557: $\bar {\vec{r}}(s)$. The inextensibility constraint, Eq.~(\ref{eq:inextensibility}), is
1558: fulfilled to $O(\delta \vec r)$ if we only consider displacements that are
1559: constructed from another infinitesimal vector field $\vec \epsilon(s)$ according to
1560: \begin{equation}
1561: \label{eq:transvers-variation}
1562: \delta\vec r'(s)=\vec \epsilon(s) \times \bar{\vec r}'(s)\;.
1563: \end{equation}
1564: Those displacements are transverse to the local tangent vector, so
1565: that $\delta(\vec r'^2)$ is quadratic in $\delta \vec r$. They
1566: correspond to local rotations of the tangents.
1567:
1568:
1569: The variation of the contour induces a variation $\delta {\cal H}$
1570: of the total elastic energy ${\cal H}=\Hb+\Hext$ of the form
1571: \begin{subequations}
1572: \begin{eqnarray}
1573: \delta {\cal H}&=& \mint{ds}{0}{L} \left(\kappa\, \bar {\vec{r}}''
1574: \delta\vec r''+\gel \delta \vec r \right)
1575: \label{eq:venergy}\\
1576: &=& -\mint{ds}{0}{L} \left(
1577: \kappa\, \bar {\vec{r}}'''+\mint{d\tilde{s}}{0}{s} \gel \right) \delta \vec r'+\text{b.t.}
1578: \label{eq:venergy-2}\\
1579: &=&- \mint{ds}{0}{L} \left( \kappa\, \bar{ \vec{r}}'''+\fel \right) \left(\vec
1580: \epsilon
1581: \times \bar{\vec{r}}'\right)+\text{b.t.} \label{eq:venergy-3} \\
1582: &=&- \mint{ds}{0}{L} \vec \epsilon \left[ \bar{ \vec{r}}' \times \left( \kappa\,
1583: \bar{\vec{r}}'''+ \fel \right) \right]+\text{b.t.}
1584: \label{eq:venergy-4}
1585: \end{eqnarray}
1586: \end{subequations}
1587: Here, we performed a partial integration to obtain Eq.~(\ref{eq:venergy-2}), which
1588: introduces some boundary terms abbreviated by ``b.t.''. In the subsequent step we
1589: inserted Eq.~(\ref{eq:transvers-variation}) for $\delta \vec r'$ and introduced the
1590: arc length dependent force
1591: \begin{equation}
1592: \label{eq:g-integrated}
1593: \fel(s)\equiv \mint{d\tilde{s}}{0}{s} \gel(\tilde{s}) \;.
1594: \end{equation}
1595: Finally, we used the property $\vec A\cdot(\vec B\times\vec C)=\vec
1596: C\cdot(\vec A\times \vec B)$ of the triple scalar product to obtain
1597: Eq.~(\ref{eq:venergy-4}).
1598:
1599:
1600: If $\bar{ \vec{r}}(s)$ is indeed the equilibrium contour then $\delta {\cal H}$ has
1601: to vanish for all variations parameterized by $\vec \epsilon(s)$ and $\delta \vec
1602: r(L)$. Therefore, the term in the square brackets of the integrand in
1603: Eq.~(\ref{eq:venergy-4}) has to vanish,
1604: \begin{equation}
1605: \label{eq:beq-app}
1606: \bar{ \vec{r}}' \times \left( \kappa\, \bar {\vec{r}}'''+\fel
1607: \right)= 0 \;,
1608: \end{equation}
1609: so that we recover Eq.~(\ref{eq:beq}), used in Sec.~\ref{sec:dwmc}. In addition, the
1610: boundary terms
1611: \begin{subequations}
1612: \begin{eqnarray}
1613: \text{b.t.}&=& \kappa\left(\bar{ \vec{r}}'' \delta \vec r'\right)\vert^L_0+
1614: \delta \vec r(L)\mint{d\tilde{s}}{0}{L} \gel \nonumber\\
1615: &=&\kappa \left[\bar {\vec{r}}'' \left(\vec \epsilon \times
1616: \bar{\vec{r}}'\right)\right]\vert^L_0 +\left[\fel\delta\vec r\right]\vert^L
1617: \nonumber \;.
1618: \label{eq:venergy-5}
1619: \end{eqnarray}
1620: \end{subequations}
1621: have to cancel, implying the requirements
1622: \begin{eqnarray}
1623: \fel(L)&=&0 \label{eq:tot-force} \\
1624: \left( \bar {\vec{r}}' \times \bar{ \vec{r}}''\right)\vert_{0,L}&=&0 \;.
1625: \label{eq:bc-free-:(}
1626: \end{eqnarray}
1627: The condition expressed by Eq.~(\ref{eq:tot-force}) simply states that a force
1628: balance can only exist if the external forces sum up to zero. Using the
1629: inextensibility constraint, Eq.~(\ref{eq:inextensibility}), it is seen that $\bar
1630: {\vec{r}}''\cdot \bar{ \vec{r}}'=0$, so that Eq.~(\ref{eq:bc-free-:(}) can be
1631: rewritten as the boundary condition
1632: \begin{equation}
1633: \label{eq:bc-free}
1634: \bar{ \vec{r}}''\vert_{0,L}=0\;.
1635: \end{equation}
1636: The curvature, which is proportional to the local torque, has to vanish at the (free)
1637: ends.
1638:
1639:
1640: Note that the over-bar in $\bar{\vec{r}}(s)$ to denote the equilibrium contour is
1641: dropped in the main text, for simplicity.
1642:
1643:
1644:
1645:
1646:
1647:
1648: \section{Multiple scale analysis (details)}
1649: \label{sec:app-mspt}
1650:
1651:
1652: \subsection{Thermal forces}
1653: \label{sec:noise-mspt}
1654:
1655:
1656: In the multiple scale perturbation theory, presented in Sec.~\ref{sec:multi-scale},
1657: one should, in principle, assume an expansion $\vec \xi(x,y;t)=\epsilon^{1/2}\vec
1658: \xi_1(x,y,t)+o(\epsilon)$ where the leading order noise $\vec \xi_1$ is a function of
1659: the coarse-grained variable $y$. However, from the fundamental correlations obeyed
1660: by the original noise function, Eq.~(\ref{eq:noise-corrs}), we have to require to
1661: leading order
1662: \begin{equation}
1663: \label{eq:mspt-noise-bc}
1664: \begin{split}
1665: &\avg{\vec \xi_{1}(s,\epsilon s;t) \vec \xi_{1}(s',\epsilon s';t)}=2 (\bold I/L)
1666: \delta(s-s')\delta(t-t')
1667: \end{split}
1668: \end{equation}
1669: for \emph{all} $\epsilon\equiv L/\lp$. Apparently, the right hand side of
1670: Eq.~(\ref{eq:mspt-noise-bc}) does not depend on $\epsilon$. As a consequence, the
1671: left hand side cannot depend on $\epsilon$ either, in particular not on $\epsilon s'$
1672: or $\epsilon s$. Hence, the two-point correlations are independent of the slowly
1673: varying arc length coordinate. Using Wick's theorem, we may argue in the same way
1674: for all higher-order correlation functions as well and conclude that the leading
1675: order stochastic force $\vec \xi_1$ is itself independent of $y$.
1676:
1677: We note, that the inverse length appearing on the right hand side of
1678: Eq.~(\ref{eq:mspt-noise-bc}) is due to the definition $\epsilon\equiv L/\lp$ of the
1679: small parameter for a stiff polymer. In the case of a strongly pre-stretched
1680: polymer, the small parameter is defined as $\epsilon\equiv (\lp\fe)^{-1/2}$, see
1681: Eq.~(\ref{eq:wbl-sfp-2}) and the subsequent paragraph, so that $L^{-1}$ in
1682: Eq.~(\ref{eq:mspt-noise-bc}) has to be replaced by $\fe^{-1/2}$.
1683:
1684: \subsection{Next to leading order tension}
1685: \label{sec:f1}
1686:
1687: Requiring the $\Ord{\epsilon^{1/2}}$ coefficient in
1688: Eq.~(\ref{eq:multiscale-2}) yields a noisy first order tension
1689: \begin{equation}
1690: \label{eq:fi-1}
1691: f_1(x,y)=\mint{d\tilde
1692: x}{0}{x}\left[\xi_{\pa,1}(0)-\xi_{\pa,1}(\tilde x)\right] +b_1(y) x+a_1(y) \,.
1693: \end{equation}
1694: For $f_1$ to be bounded in $x$, the term $b_1(y)x$ has to cancel the
1695: linearly growing the noise-term on the right hand side,
1696: \begin{equation}
1697: \label{eq:fix-c-1}
1698: b_1=\lim_{x\to\infty}\frac1x \mint{d\tilde
1699: x}{0}{x}\left[\xi_{\pa,1}(0)-\xi_{\pa,1}(\tilde x)\right]\;.
1700: \end{equation}
1701: However, important for the multiple scale analysis in
1702: Sec.~\ref{sec:multi-scale} is merely that $b_1$ is independent of $y$,
1703: so that $\partial_y\partial_x f_1(x,y)=0$.
1704:
1705:
1706:
1707:
1708:
1709:
1710:
1711:
1712:
1713: \bibliographystyle{apsrev}
1714:
1715: \bibliography{bibis/tdpre06,bibis/elastic-rod-dynamics,bibis/journals,bibis/sfpdynamics-tcited,bibis/unpub,bibis/actin-viscoelastic,bibis/semiflexibleA04,bibis/sf,bibis/klaussf,bibis/mysf,bibis/sfnet,bibis/unpub,bibis/mysfnet}
1716:
1717: \end{document}
1718: