1: \documentclass[prb,twocolumn,tightenlines,floatfix,showpacs]{revtex4}
2: \usepackage{graphicx}
3: \usepackage{amssymb}
4: \newcommand{\beq}{\begin{equation}}
5: \newcommand{\eeq}{\end{equation}}
6: \newcommand{\bea}{\begin{eqnarray}}
7: \newcommand{\eea}{\end{eqnarray}}
8:
9: \begin{document}
10:
11:
12: \title{Persistence length of a polyelectrolyte in salty water:
13: a Monte-Carlo study}
14:
15: \author{T. T. Nguyen and B. I. Shklovskii}
16:
17: \affiliation{Theoretical Physics Institute, University of Minnesota, 116
18: Church Street Southeast, Minneapolis, Minnesota 55455}
19:
20: \begin{abstract}
21: We address the long standing problem of the dependence of
22: the electrostatic
23: persistence length $l_e$ of a flexible polyelectrolyte (PE)
24: on the screening length $r_s$ of the solution
25: within the linear Debye-H\"{u}ckel theory.
26: The standard Odijk, Skolnick and Fixman (OSF)
27: theory suggests $l_e \propto r_s^2$, while
28: some variational theories and computer simulations
29: suggest $l_e \propto r_s$.
30: In this paper, we use Monte-Carlo simulations
31: to study the conformation of a simple polyelectrolyte.
32: Using four times longer PEs than in previous simulations
33: and refined methods for the treatment of
34: the simulation data, we show that the results are consistent
35: with the OSF dependence $l_e \propto r_s^2$.
36: The linear charge density of the PE which enters
37: in the coefficient of this dependence
38: is properly renormalized to take into account local
39: fluctuations.
40: \end{abstract}
41:
42: \pacs{61.25.Hq, 87.15.Bb, 36.20.Ey, 87.15.Aa}
43:
44:
45: \maketitle
46:
47:
48: \section{Introduction}
49: Despite numerous theoretical studies of
50: polyelectrolyte (PE),
51: due to the long range nature
52: of the Coulomb interaction, the description
53: of their conformation is still not as satisfactory as
54: that of neutral polymers.
55: One of the longest standing problem is related to
56: the electrostatic effect on the rigidity of
57: a PE. In a water solution with monovalent ions,
58: within the Debye-H\"{u}ckel linear screening theory,
59: the electrostatic interaction between PE charged monomers
60: has the form:
61: %
62: \beq
63: V(r)=\frac{e^2}{Dr}\exp\left(-\frac{r}{r_s}\right)~,
64: \label{VDH}
65: \eeq
66: %
67: where $r$ is the distance
68: between monomers,
69: $D$ is the dielectric
70: constant of water, $e$ is the elementary charge,
71: and $r_s$ is the Debye-H\"{u}ckel screening length,
72: which is related to
73: the ionic strength $I$ of the solution by $r_s^2=4\pi l_BI$.
74: ($l_B=e^2/Dk_BT$ is the Bjerrum length,
75: $T$ is the temperature of the solution).
76:
77: The rigidity of a polymer is usually characterized by
78: one parameter, the so called persistence length
79: $l_p$.
80: For a polyelectrolyte chain, besides the intrinsic
81: persistence length $l_0$ which results from the specific
82: chemical structure of the monomers and bonds
83: between them, the total
84: persistence length also includes an ``electrostatic"
85: contribution $l_e$ which results from the
86: screened Coulomb interactions between monomers:
87: %
88: \beq
89: l_p=l_0+l_e~.
90: \eeq
91: %
92:
93: Because the interaction (\ref{VDH}) is exponentially
94: screened at distances larger than $r_s$,
95: early works concerning the structure of the PE
96: assumed that $l_e$ is of the order of $r_s$.
97: However, this simple assumption
98: was challenged by the pioneering
99: works of Odijk\cite{Odijk} and
100: Skolnick and Fixman\cite{SF} (OSF),
101: who showed that Debye-H\"{u}ckel interaction
102: can induce a rod-like conformation at length scales much larger
103: than $r_s$. Their calculation gives
104: %
105: \beq
106: l_e=l_{OSF}
107: =\frac{\eta_0^2}{4Dk_BT}r_s^2~,
108: \label{eq:lOSF}
109: \eeq
110: %
111: where
112: $\eta_0$ is the linear
113: charge density of the PE.
114: Because $l_e \propto r_s^2$, it can
115: be much larger than $r_s$ at weak screening (large $r_s$).
116:
117: Although the idea that electrostatic interaction enhances
118: the stiffness of a PE is qualitatively accepted and confirmed
119: in many experiments,
120: the quadratic dependence of $l_e$ on
121: the screening length $r_s$ is still the subject of many
122: discussions. In the work of OSF, the bond angle deflection
123: was assumed to be small everywhere along the chain,
124: what is valid for large $l_0$. They suggested that
125: if $l_0$ is not small but $l_e$ is large enough (week screening),
126: their assumption is still valid.
127: Ref. \onlinecite{BJ}, however, has questioned this assumption
128: especially when $l_0$ is so small
129: that the bond angle deflection is large before
130: electrostatics comes into play and rigidifies the
131: chain.
132:
133:
134: A significant progress was made by
135: Khokhlov and Khachaturian (KK) who proposed a
136: generalized OSF theory\cite{KK}
137: for the case of flexible polyelectrolyte (small $l_0$).
138: It is known that in the absence of screening
139: ($r_s \rightarrow \infty$), the structure of
140: a polyelectrolyte can be conveniently described
141: by introducing the concept of electrostatic blobs.
142: A blob is a chain subunit within which
143: the electrostatic interaction is only a weak perturbation.
144: The blob
145: size $\xi$ is related to the number of Kuhn segments $g$
146: within one blob as $\xi=l_0g^{1/2}$. The condition
147: of weak Coulomb interaction suggest that
148: the electrostatic self energy of a blob,
149: $(\eta_0gl_0)^2/D\xi$ is of the order of $k_BT$.
150: This leads to $\xi \simeq (Dk_BTl_0^2/\eta_0^2)^{1/3}$.
151: At length
152: scale greater than $\xi$, Coulomb interaction plays
153: important role and the string of blobs assumes
154: a rod-like conformation, with the end-to-end distance
155: proportional to the number of blobs.
156:
157: Using this blob picture,
158: KK proposed that OSF theory is still applicable for
159: a flexible PE provided
160: one deals with the chain of blobs instead of the original
161: chain of monomers. This means, in Eq. (\ref{eq:lOSF}),
162: one replaces the bare linear charge density $\eta_0$
163: by that of the blob chain $\eta = \eta_0gl_0/\xi$.
164: The intrinsic persistence length $l_0$ should also be
165: replaced by $\xi$.
166: As a result, the total persistence length of the flexible PE reads:
167: %
168: \beq
169: l_{p,KK}=\xi+\frac{\eta^2}{4Dk_BT}r_s^2~.
170: \label{eq:lpKK}
171: \eeq
172: %
173: Thus, in KK theory, despite the flexibility of the PE,
174: its electrostatic persistence length remains
175: quadratic in $r_s$. Small $l_0$ only
176: renormalizes the linear charge density
177: from $\eta_0$ to $\eta$.
178:
179: Note that $r_s$ is implicitly
180: assumed to be larger than the blob size $\xi$ in KK theory
181: (weak screening). For strong screening $r_s < \xi$,
182: there are no electrostatic rigidity and the chain behaves
183: as flexible chain with the Debye-H\"{u}ckel
184: short range interaction playing the role of
185: an additional excluded volume interaction.
186:
187: A number of variational
188: calculations have also been proposed to describe
189: more quantitatively the
190: structure of flexible chain. These
191: calculations, although based on different ansatz, have the
192: same basic idea of describing the flexible charged chain
193: by some model of noninteracting semiflexible chain and
194: variationally optimizing the persistence length of
195: the noninteracting system.
196: Surprisingly, while some of these calculations
197: support the OSF-KK dependence $l_e \propto r_s^2$
198: such as Refs. \onlinecite{Witten,Netz,Thirumalai2},
199: other calculations found that $l_e$
200: scales linearly with $r_s$ instead\cite{BJ,Dawson,Thirumalai}.
201: However, because variational
202: calculation results depend strongly on the
203: variational model Hamiltonian,
204: none of these results can be considered conclusive.
205:
206: Computer simulations\cite{BB,Reeds,Seidel,Ullner2,Ullner3,Kremer} also
207: have been used to determine
208: the dependence $l_e$ on $r_s$ and to verify
209: OSF or variational theories. Some of these papers
210: claim to support the linear dependence of $l_p$ on
211: $r_s$. The simulation of
212: Ref. \onlinecite{Kremer} concludes that the
213: dependence of $l_p$ on $r_s$ is sublinear.
214: Thus, the problem of the dependence $l_e(r_s)$, despite
215: being very clearly stated, still remains unsolved for a
216: flexible PE. More details about the present status
217: of this problem can be found
218: in Ref. \onlinecite{Ullner4}.
219:
220:
221: In this paper, we again use computer simulations
222: to study the dependence of $l_e$ on $r_s$.
223: The longest polyelectrolyte
224: simulated in our paper contains 4096 charged monomers,
225: four times
226: more than those studied in previous simulations.
227: This allows for better studying of size effect on
228: the simulation result. Furthermore, we use
229: a more refined analysis of the simulation result,
230: which takes into account local fluctuations
231: in the chain at short distance scale.
232: Our results show that OSF formula quantitatively
233: describes the structure of a polyelectrolyte.
234:
235: The paper is organized as follows.
236: The procedure of Monte-Carlo simulation of
237: a polyelectrolyte using the primitive freely jointed
238: beads is described in
239: the next section. The data for the end-to-end distance
240: $R_{ee}$ is given. In Sec. III, we analyze this
241: data using the scaling argument to show that
242: it is consistent with OSF theory. In Sec. \ref{sec:worm},
243: we analyze the data for the case of large $r_s$,
244: where excluded volume effect is not important,
245: in order to extract $l_e$ and again show that
246: it obeys OSF theory in this limit.
247: In Sec. V, we use the bond angle correlation function
248: to calculate $l_e$ and to confirm the result
249: of Sec. IV. The good agreement between $l_e$ calculated
250: using different methods further suggests that OSF theory is
251: correct in describing a polyelectrolyte structure.
252: We conclude in Sec. \ref{sec:concl}.
253:
254: Several days after the submission of our paper
255: to the Los Alamos preprint
256: archive\cite{Nguyen}, another paper\cite{Everaers} with
257: Monte-Carlo simulations for PE molecules in the
258: same range of lengths appears in the same archive.
259: Results of this paper are
260: in good agreement with our Sec. III.
261:
262:
263:
264:
265: \section{Monte-Carlo simulation}
266: The polyelectrolyte is modeled as a chain of $N$ freely jointed
267: hard spherical beads
268: each with charge $e$.
269: The bond length of the PE is fixed and equal to $l_B$,
270: where $l_B=e^2/Dk_BT$ is
271: the Bjerrum length which is about 7\AA\ at room temperature in
272: water solution. Thus the bare linear charge density of
273: our polyelectrolyte is $\eta_0=e/l_B$.
274: Because we are concerned about the electrostatic
275: persistence length only, the bead radius is set to zero so that
276: all excluded volume of monomers is provided by the screened
277: Coulomb interaction between them only. For convenience, the middle
278: bead is fixed in space.
279:
280: To relax the PE configuration globally,
281: the pivot algorithm\cite{pivot} is used.
282: In this algorithm, in an attempted move,
283: a part of the chain from
284: a randomly chosen monomer to one end of the chain is rotated
285: by a random angle about a random axis. This algorithm is known
286: to be very efficient. A new
287: independent sample can be produced in a computer time of the order
288: of $N$, or in other words, uncorrelated samples are obtained every
289: few Monte Carlo (MC) steps (one MC step is defined as the number
290: of elementary moves such that, on average, every particle
291: attempts to move once).
292: To relax the PE configuration locally, the flip
293: algorithm is used. In this algorithm, a randomly chosen monomer is
294: rotated by a random angle about the axis connecting its two neighbor
295: (if it is one of the end monomers, its new position is chosen
296: randomly on the surface of a sphere with radius $l_B$ centered at
297: its neighbor.) In a simulation, the number of pivot moves is about 30\%
298: of the total number of moves.
299: The usual Metropolis algorithm is used to accept
300: or reject the move. About $1\div 2\times 10^4$ MC steps are
301: run for each set of parameters ($N$, $r_s$), of which
302: 512 initial MC steps are discarded and the rest
303: is used for statistical average (due to time constrain, for $N=4096$,
304: only 2000 MC steps are used).
305: Two different initial configurations,
306: a Gaussian coil and a straight rod, were used to ensure
307: that final states are indistinguishable and the
308: systems reaches equilibrium.
309:
310: %
311: \begin{figure}
312: \resizebox{7.5cm}{!}{\includegraphics{Ree.eps}}
313: \caption{The square of the end-to-end distance of a
314: polyelectrolyte $R_{ee}^2$ as a function of the screening
315: length $r_s$ for chains with
316: different number of monomers $N$:
317: 64($\diamond$), 128($+$), 256 ($\square$), 512 ($\times$),
318: 1024 ($\triangle$), 2048($\ast$), and 4096 ($\blacklozenge$).
319: The arrows on the right side show
320: $R_{ee}^2$ obtained
321: using unscreened Coulomb potential $V(r)=e/r$.}
322: \label{fig:ree}
323: \end{figure}
324: %
325: The simulation result for the end-to-end distance $R_{ee}$
326: of a polyelectrolyte for different $N$
327: is plotted in Fig. \ref{fig:ree} as a function of the
328: screening radius $r_s$ of the solution. At very small $r_s$,
329: Coulomb interactions between monomers are strongly screened
330: and the chain behaves as a neutral Gaussian
331: chain with $R_{ee}=l_B\sqrt{N-1}$.
332: At very large $r_s \gg N$, Coulomb interactions between
333: the monomers
334: are not screened and $R_{ee}$ is saturated and equal to that
335: of an unscreened PE with the same number of monomers
336: (see the arrows in Fig. \ref{fig:ree}).
337:
338: Three different methods are used to verify the validity
339: of OSF theory for flexible PE:
340: i) study of the scaling dependence of $R_{ee}$ on $r_s$
341: in whole range of $r_s$, ii) extraction
342: of $l_e$ in the large $r_s$ limit and
343: iii) analysis of the bond correlation function. In the next three
344: sections, we discuss these methods in details together with
345: their limitations. Comparison with
346: previous simulations is also made to explain
347: their results which so far have not
348: supported either of the theories.
349:
350: \section{Scaling dependence of $R_{ee}$ on $r_s$}
351:
352: Let us first describe theoretically
353: how the chain size should behave as a function
354: of the screening radius $r_s$ when $r_s$ increases
355: from 0 to $\infty$.
356:
357: When $r_s \ll l_B$, the Coulomb interaction is strongly screened.
358: Because there are no other interaction present
359: in our chain model of freely jointed beads,
360: the chain statistic is Gaussian. Its
361: end-to-end distance $R_{ee}$ is proportional to the
362: the square root of the number of bonds and independent on $r_s$:
363: %
364: \beq
365: R_{ee}^2=l_B^2 N~.
366: \label{eq:scale1}
367: \eeq
368: %
369:
370: When $r_s \gg l_B$, the chain persistence length
371: is dominated by the Coulomb contribution $l_p \simeq l_e$.
372: If $N$ is very large such that the chain contour length
373: $Nl_B$ is much larger than $l_e$ then
374: the chain behaves as a linear chain with $Nl_B/l_e$ segments
375: of length $l_e$ each and thickness $r_s$.
376: The excluded volume between segments is
377: $v\simeq l_e^2r_s$, and the end-to-end distance\cite{KK}:
378: %
379: \beq
380: R_{ee}^2=l_e^2
381: \left(\frac{v}{l_e^3}\right)^{2/5}
382: \left(\frac{Nl_B}{l_e}\right)^{6/5}
383: \propto
384: \left\{
385: \begin{array}{ll}
386: r_s^{4/5} & \mbox{if\ } l_e\propto r_s \\
387: r_s^{6/5} & \mbox{if\ } l_e\propto r_s^2
388: \end{array}
389: \right.~.
390: \label{eq:scale2}
391: \eeq
392: %
393:
394: At larger $r_s$ where $l_e$ becomes comparable to the PE contour
395: length, the excluded volume effect
396: is not important. In this case, the chain statistics is again
397: Gaussian and
398: %
399: \beq
400: R_{ee}^2 \simeq l_e^2 %\sqrt{
401: \frac{Nl_B}{l_e}
402: \propto
403: \left\{
404: \begin{array}{ll}
405: r_s & \mbox{if\ } l_e\propto r_s \\
406: r_s^2 & \mbox{if\ } l_e\propto r_s^2
407: \end{array}
408: \right. ~~.
409: \label{eq:scale3}
410: \eeq
411: %
412:
413: Finally, at even larger $r_s$ when $l_p$ is greater than
414: $Nl_B$, the chain becomes a straight rod with length
415: independent on $r_s$:
416: %
417: \beq
418: R_{ee}^2\simeq l_B^2N^2~.
419: \label{eq:scale4}
420: \eeq
421: %
422:
423: If $l_e \propto r_s^2$, the transition from
424: the scaling range of Eq. (\ref{eq:scale2}) to
425: Eq. (\ref{eq:scale3})
426: happens at $r_s \simeq l_BN^{1/4}$, while
427: the transition from
428: the scaling range of Eq. (\ref{eq:scale3}) to
429: Eq. (\ref{eq:scale4}) happens at
430: $r_s \simeq l_BN^{1/2}$.
431: On the other hand,
432: if $l_e \propto r_s$, both transitions from
433: the scaling range of Eq. (\ref{eq:scale2}) to
434: Eq. (\ref{eq:scale3}) and from
435: the scaling range of Eq. (\ref{eq:scale3}) to
436: Eq. (\ref{eq:scale4}) happen at $r_s \simeq l_B N$.
437: This means, there is no scaling range
438: of Eq. (\ref{eq:scale3}) in this theory.
439:
440: Thus, one can distinguish between the OSF result,
441: $l_p \propto r_s^2$, and the variational result,
442: $l_p \propto r_s$ by plotting the exponent
443: $\alpha=\partial \ln [R_{ee}^2]/\partial\ln r_s$
444: as a function of $\ln r_s$. The schematic figure of this
445: plot is shown in Fig. \ref{fig:alpha}.
446: OSF theory gives plateaus
447: at $\alpha=6/5$ and 2, and when $r_s > l_BN^{1/2}$,
448: $\alpha$ drops back to 0.
449: Variational theories, on the other hand, would
450: suggest one large plateau at $\alpha=4/5$
451: up to $r_s \simeq l_BN$.
452: %
453: \begin{figure}
454: \resizebox{8.5cm}{!}{\includegraphics{alpha.eps}}
455: \caption{Schematic plot of
456: $\alpha$
457: as a function of $r_s$ for the OSF theory $l_p\propto r_s^2$
458: (solid line) and for variational theories $l_p \propto r_s$
459: (dashed line).}
460: \label{fig:alpha}
461: \end{figure}
462: %
463: %
464: \begin{figure}
465: \resizebox{7.5cm}{!}{\includegraphics{Ree2.eps}}
466: \caption{Simulation result for
467: $\alpha$
468: as a function of $r_s$ for different $N$:
469: 64($\diamond$), 128($+$), 256 ($\square$), 512 ($\times$),
470: 1024 ($\triangle$), 2048($\ast$), and 4096 ($\blacklozenge$).
471: They agree reasonably well
472: with the solid curve of Fig. \ref{fig:alpha},
473: suggesting that OSF theory is correct.}
474: \label{fig:ree2}
475: \end{figure}
476: %
477:
478: The simulation results for $\alpha$
479: are shown in Fig. \ref{fig:ree2} for different $N$.
480: One can see that as $N$ increases, the agreement with
481: OSF theory becomes more visible.
482: Note that
483: the plateaus in Fig. \ref{fig:alpha} are scaling ranges,
484: and relatively sharp transitions between plateaus
485: are valid only for $N\rightarrow \infty$.
486: For a finite $N$, the plateaus
487: may be too narrow to be observed and can be masked in
488: the transition regions. This explains why one cannot
489: see the plateau at $\alpha=2$ in our results. Nevertheless,
490: the tendencies of $\alpha$ to develop a plateau
491: at $\alpha=6/5$, then to grow higher toward $\alpha=2$
492: at larger $r_s$ and
493: finally to collapse to zero when approaching
494: relatively small $r_s=l_B\sqrt{N}$
495: are clearly seen for large $N$.
496: Thus, generally speaking,
497: the curves agree with OSF theory much better than
498: with variational theories (where $\alpha$ is supposed
499: to be about 4/5 and to decrease to zero only when
500: $r_s \rightarrow l_BN$, i.e. at much larger $r_s$
501: than what observed).
502:
503: Fig. \ref{fig:ree2} also shows one reason
504: why similar simulations done
505: by other groups do not support OSF theory. All
506: of these simulations are limited to 512 charges.
507: As one can see from Fig. \ref{fig:ree2}, the curves
508: for $N \le 512$ do not permit to discriminate
509: between the two theories as clearly as
510: the case $N=2048$ or 4096. Only when $N$ becomes very
511: large can scaling ranges with $\alpha > 1$
512: show up and one observes better agreement
513: with OSF result.
514:
515:
516:
517: \section{Large $r_s$ limit}
518:
519: \label{sec:worm}
520:
521: In this section, we attempt to extract directly
522: from the simulation data the
523: persistence length in order to compare with OSF theory.
524: To do this, one notices
525: that
526: even a chain with excluded volume interaction
527: behaves as a Gaussian chain when
528: its contour length is very short such that it contains only a few
529: Kuhn segments. In this case, one can use
530: the Bresler-Frenkel formula\cite{Landau} to describe the
531: relationship between the end-to-end distance $R_{ee}$ and
532: the chain persistence length $l_p$:
533: %
534: \beq
535: R_{ee}^2=2Ll_p-2l_p^2[1-\exp(-L/l_p)]~,
536: \label{eq:master}
537: \eeq
538: %
539: where $L$ is the contour length of the chain.
540:
541: For our polyelectrolyte, this formula can be used
542: for large $r_s$ when
543: the persistence length is of the order of $R_{ee}$
544: or larger. However,
545: one cannot use the bare contour length $L_0=(N-1)l_B$ in
546: the Eq. (\ref{eq:master}) because the chain where OSF theory
547: is supposed to be applicable is not
548: the bare chain but an effective chain which takes into account
549: local fluctuations. The contour length $L$ of this
550: effective chain is
551: %
552: \beq
553: L = Ne/\eta
554: \label{eq:L}
555: \eeq
556: %
557: where $\eta$ is the renormalized linear charge
558: density of the PE.
559:
560: In KK theory, the effective chain is
561: the chain of electrostatic blobs,
562: and the normalized charge density is $\eta=\eta_0gl_0/\xi$.
563: However, the standard blob picture
564: can only be used to describe flexible
565: weakly charged chains where the fraction of
566: charged monomers is small so that the number of monomers, $g$,
567: within one blob is large and Gaussian statistics can be used to
568: relate its size and molecular weight.
569: Because, for a given number
570: of charged monomers, Monte-Carlo simulation for weakly
571: charged polyelectrolyte is extremely time consuming,
572: all monomers of our simulated
573: polyelectrolyte are charged.
574: In this case, the neighbor-neighbor monomers
575: interaction equals $k_BT$. This makes $g\simeq 1$ and the standard
576: picture of Gaussian blobs does not apply. Thus,
577: in order to treat our data, we assume that
578: both $l_p$ and $\eta$ are unknown quantities.
579:
580: To proceed further, one needs an equation
581: relating $\eta$ and $l_p$, and
582: in order to verify OSF theory,
583: we could use their formula
584: %
585: \beq
586: l_p=\eta^2r_s^2/4Dk_BT~,
587: \label{eq:lOSF11}
588: \eeq
589: %
590: for this purpose.
591: Thus, we could substitute Eq. (\ref{eq:L}) and
592: (\ref{eq:lOSF11}) into Eq. (\ref{eq:master}),
593: and solve for $\eta$ using $R_{ee}^2$
594: obtained from simulation.
595: If OSF theory is valid,
596: the obtained values of $\eta$ should be a very slow
597: changing function of $r_s$.
598: In addition,
599: in the limit $N\rightarrow\infty$, they should
600: also be independent on $N$.
601:
602: The OSF equation (\ref{eq:lOSF}), however, was
603: derived for the case $r_s\ll L$ while
604: in our simulation, the
605: ratio $r_s/L$ is not always small.
606: Therefore, instead of Eq. (\ref{eq:lOSF11}),
607: we use the more general Odijk's finite size
608: formula\cite{Odijk}
609: %
610: \beq
611: l_p=\frac{\eta^2r_s^2}{12Dk_BT}\left[
612: 3-\frac{8r_s}{L}+\left(5+\frac{L}{r_s}+
613: \frac{8r_s}{L}
614: \right)e^{-L/r_s}
615: \right]~.
616: \label{eq:lOSF2}
617: \eeq
618: %
619: for the persistence length $l_p$.
620: When $L\gg r_s$, the term in the square brackets
621: is equal to 3 and the standard OSF result is recovered.
622: On the other hand, when $r_s \gg L $, the
623: persistence length $l_p$ saturates
624: at $\eta^2L^2/72Dk_BT$.
625:
626: Below, we treat our Monte-Carlo simulation data
627: with the help of Eq. (\ref{eq:master})
628: using Eq. (\ref{eq:L})
629: and (\ref{eq:lOSF2}) for $L$ and
630: $l_p$.
631: The results for $\eta$ are plotted in
632: Fig. \ref{fig:eta} for different PE sizes $N$.
633: %
634: \begin{figure}
635: \resizebox{7.5cm}{!}{\includegraphics{eta.eps}}
636: \caption{The linear charge density $\eta$ as
637: a function of the screening length for different
638: $N$:
639: 64($\diamond$), 128($+$), 256 ($\square$), 512 ($\times$),
640: 1024 ($\triangle$) and 2048($\ast$). The thick solid line
641: is the theoretical estimate which is the numerical
642: solution to Eq. (\ref{eq:F}), (\ref{eq:cost})
643: and (\ref{eq:etat}).}
644: \label{fig:eta}
645: \end{figure}
646: %
647: As one can see, at large $r_s$,
648: $\eta$ changes very slowly with $r_s$,
649: and as $N$ increases, tends to saturate at
650: an $N$ independent value.
651:
652: It should be noted that the lines $\eta(r_s)$
653: in Fig. \ref{fig:eta}
654: unphysically start to drop below
655: certain values of $r_s$. This is because at smaller
656: $r_s$, the electrostatics-induced excluded volume
657: interactions
658: between monomers become so strong that the right
659: hand side of Eq. (\ref{eq:master}) (which is
660: derived for a Gaussian worm like chain) strongly
661: underestimates $R_{ee}$.
662:
663:
664: Even though the picture of Gaussian blobs
665: does not work for our chain,
666: $\eta$ can still be calculated analytically
667: in the limit $L \gg r_s$ ($N\rightarrow \infty$).
668: Indeed, let us assume that
669: the effective chain is straight at length scale smaller than $r_s$
670: (which is a reasonable assumption because all the analytical
671: theories so far suggested that the PE persistence length
672: scales as $r_s$ or $r_s^2$). Thus, the self energy of the chain
673: can be written as $E=L\eta^2\ln(r_s/l_B)/D$.
674: At length scale smaller than $r_s$,
675: the polyelectrolyte behaves as a neutral chain under
676: an uniform tension
677: %
678: \beq
679: F=\partial E/\partial L=\eta^2\ln(r_s/l_B)/D~.
680: \label{eq:F}
681: \eeq
682: %
683: The average
684: angle a bond vector makes with respect to the
685: axis of the chain, therefore, is:
686: %
687: \bea
688: \left<\cos\theta\right>&=&\frac{\int_0^{\pi}\exp(Fl_B\cos\theta/k_BT)
689: \cos\theta\sin\theta d\theta}
690: {\int_0^{\pi}\exp(Fl_B\cos\theta/k_BT)\sin\theta d\theta}\nonumber\\
691: &=&\coth\frac{Fl_B}{k_BT}-\frac{k_BT}{Fl_B}~.
692: \label{eq:cost}
693: \eea
694: %
695: The charge density $\eta$ can be calculated as
696: %
697: \beq
698: \eta=\eta_0/\left<\cos\theta\right>~.
699: \label{eq:etat}
700: \eeq
701: %
702: At weak screening $r_s \gg l_B$, it can be estimated
703: analytically:
704: %
705: \beq
706: \eta
707: \simeq\eta_0\left[1+\frac{1}{\ln(r_s/l_B)}+...\right]~,
708: \eeq
709: %
710: where the expansion terms of the order of $1/\ln^2(r_s/l_B)$ and
711: higher were neglected.
712:
713: The more accurate numerical solution of Eq. (\ref{eq:F}),
714: (\ref{eq:cost}) and (\ref{eq:etat}) for $\eta$ is
715: plotted in Fig. \ref{fig:eta} by the thick solid line.
716: One can see that the values $\eta(r_s)$
717: calculated experimentally using OSF theory
718: with growing $N$ converge well to the theoretical
719: curve for $N=\infty$.
720: Remarkably, the theoretical estimate for $\eta$
721: does not use any fitting
722: parameters. This, once again, strongly
723: suggests the OSF theory is valid for flexible
724: PE as well.
725:
726: The Bresler-Frenkel formula, Eq. (\ref{eq:master}),
727: is also used
728: to extract the persistence length
729: in Ref. \onlinecite{Kremer}
730: where the authors concluded that the dependence
731: of $l_e$ on $r_s$ is sublinear.
732: The authors, however, used in Eq. (\ref{eq:master})
733: the bare contour
734: length $L$, or in other words $\eta=\eta_0$,
735: for the calculation of $l_e$. As one
736: can see from Fig. \ref{fig:eta}, this leads
737: to 20-30\% overestimation of the contour length
738: of the effective chain where OSF theory is supposed
739: to apply. To show that this overestimation is
740: crucial, let us treat our data
741: similarly to Ref. \onlinecite{Kremer} using $\eta=\eta_0$.
742: We plot the resulting dependence of $l_e/r_s$ on $r_s$
743: (similarly to Fig. 4 of Ref. \onlinecite{Kremer})
744: and compare it with our own results using corrected $\eta$.
745: %
746: \begin{figure}
747: \resizebox{7.5cm}{!}{\includegraphics{lp.eps}}
748: \caption{Plots of $l_e/r_s$ as a function of $r_s$
749: calculated with the help of Eq. (\ref{eq:master}),
750: (\ref{eq:L}) and (\ref{eq:lOSF2})
751: using our data for $R_{ee}^2$
752: ($+$) and
753: using unperturbed $\eta=\eta_0$ as in
754: Ref. \onlinecite{Kremer}
755: ($\diamond$). The chain with $N=1024$ is used.
756: }
757: \label{fig:lpKremer}
758: \end{figure}
759: %
760: The case $N=1024$ is shown in Fig.
761: \ref{fig:lpKremer}. Obviously, the two
762: results are different
763: qualitatively. While the upper curve
764: follows Eq. (\ref{eq:lOSF2}) with
765: slightly decreasing $\eta$,
766: the lower curve shows sublinear growth
767: of $l_e$ with $r_s$ ($l_e/r_s$
768: is a decreasing function of $r_s$). This
769: sublinear dependence observed in
770: Ref. \onlinecite{Kremer} is clearly
771: a manifestation of their overestimation of
772: the PE length $L$ which should be used in Eq.
773: (\ref{eq:master}).
774:
775: Note that the true $l_e/r_s$ curve should
776: also eventually decrease to zero because $l_e$
777: saturates to the constant value $\eta^2L^2/72Dk_BT$
778: when $r_s \gg L$ [See Eq. (\ref{eq:lOSF2})].
779: But, according to Eq. (\ref{eq:lOSF2}), this decay
780: starts only at
781: very large $r_s$ where $r_s/L\simeq 0.25$.
782: The deviation from $l_e\propto r_s^2$ at large $r_s$
783: seen in Fig. \ref{fig:lpKremer}
784: is due to both the violation of the inequality
785: $r_s \ll L$ and
786: to the slight decrease of $\eta$ with $r_s$.
787:
788:
789: \section{Bond angle correlation function}
790:
791: Another standard procedure used in literature is
792: to calculate the persistence length
793: of a polyelectrolyte
794: as the typical decay length of the bond angle
795: correlation function (BACF) along the
796: contour of the chain, assuming the later is exponential.
797: %
798: \beq
799: f(|s^\prime-s|)=\left<
800: \cos[\angle({\mathbf b}_s,{\mathbf b}_{s^\prime})]
801: \right>
802: \propto \exp\left(-\frac{|s^\prime-s|}{l_p}\right)~.
803: \label{eq:bac}
804: \eeq
805: %
806: Here ${\mathbf b}_s$ and ${\mathbf b}_{s^\prime}$
807: are the bond numbered $s$ and $s^\prime$ respectively
808: and $\angle({\mathbf b}_s,{\mathbf b}_{s^\prime})$
809: is the angle between them. The symbol $\left<...\right>$
810: denotes the averaging over different chain conformations.
811: To improve averaging,
812: the pair $s$ and $s^\prime$ are also allowed to move
813: along the chain keeping $|s^\prime-s|$ constant.
814:
815: We argue in this section that this method
816: of determining persistence length actually
817: has a very limited range of applicability.
818: At either small or large $r_s$, the results
819: of persistence length obtained from BACF
820: are not reliable.
821: In the range where this method is supposed
822: to be applicable, we show that the
823: obtained $l_p$ are close to those obtained
824: in Sec. \ref{sec:worm} above.
825:
826:
827: For small $r_s$, excluded volume plays important
828: role and, strictly speaking, it is not
829: clear whether BACF is exponential, and if yes,
830: how one should eliminate excluded volume
831: effect and extract $l_p$ from the decay length.
832: According to Ref. \onlinecite{Ullner3}, the
833: decay is not exponential in this regime.
834:
835: The procedure of determining the persistence length
836: using BACF becomes unreliable
837: at large $r_s$ as well. To elaborate this point,
838: in Fig. \ref{fig:decay}a,
839: we plot the logarithm of
840: the bond angle correlation function $f(x)$ along the
841: PE contour length for a $N=512$ and $r_s=50l_B$,
842: typical values of $N$ and $r_s$ where
843: the excluded volume due to Coulomb interactions is small.
844: There are three regions in this plot.
845: In region A at very small distance along the PE contour
846: length, monomers are within one electrostatic blobs
847: from each other and the effects of Coulomb interaction are small.
848: The bond angle correlation in this region decays
849: over one bond length $l_B$. At larger distance along
850: the PE contour length, the region B, the decay is
851: exponential and a constant decay length seems well-defined.
852: Finally, at distance comparable to the chain's contour length,
853: one again observes a fast drop of the BACF
854: (region C). This end effect is due to the fact that the
855: stress at the end of the chain goes to zero and the end bonds
856: become uncorrelated.
857: The persistence length of interest
858: can be defined as the decay length in region B.
859: %
860: \begin{figure}
861: \resizebox{7cm}{!}{\includegraphics{b.eps}}
862: \caption{The logarithm of the bond correlation
863: function $f(x)$ as a function of the distance $x$
864: (in units of $l_B$)
865: along the chain for the case $N=512$, $r_s=50l_B$.
866: There are three regions A, B and C. The dotted
867: line, $-0.47-x/1083$, is a linear fit of region B
868: suggesting that the persistence length for this
869: case is $l_p=1083l_B$.}
870: \label{fig:decay}
871: \end{figure}
872: %
873:
874: Problem arises, however, at large enough $r_s$ when
875: the region C (the end effect) becomes so large that
876: region B is not well defined. In this case the obtained
877: decay length underestimates the correct persistence
878: length. As one can
879: see from Fig. \ref{fig:decay}, region C can be quite
880: large. It occupies 40\% of the available range
881: of $x$, even though the screening length is only 10\% of
882: the contour length in this case.
883:
884: There is an even more strict condition on how large $r_s$
885: is when
886: the method of BACF loses its reliability.
887: If $l_e$ is larger than $L$, the decrease of
888: $\ln f(x)$ in region B is less than unity.
889: When this happens, an exponential decay is ambiguous.
890:
891: \begin{table}
892: \caption{\label{tab:le}Comparison between $l_{BACF}$ calculated
893: using BACF method and $\eta l_e/\eta_0$ calculated in
894: Sec. \ref{sec:worm}. All lengths are measured in units of $l_B$.}
895: \begin{ruledtabular}
896: \begin{tabular}{cccc}
897: $N$ & $r_s$ & $l_{BACF}$ & $\eta l_e/\eta_0$\\ \hline
898: 2048 & 100 & 4590 & 3682 \\ \cline{2-4}
899: & 150 & 9000 & 7484 \\ \hline
900: 1024 & 80 & 2535 & 2180 \\ \cline{2-4}
901: & 100 & 3733 & 3111 \\ \hline
902: 512 & 50 & 1083 & 809
903: \end{tabular}
904: \end{ruledtabular}
905: \end{table}
906:
907: Because of all these limitations,
908: in this section we
909: use BACF to calculate $l_e$ only
910: in the very limited range of $r_s$ where
911: excluded volume is not important and $l_e$ is not
912: much larger than $L$ (the decrease in region B is greater
913: than 0.1). The obtained $l_{BACF}$, which is measured along
914: the chain contour, is compared to $\eta l_e/\eta_0$ obtained
915: using the Bresler-Frankel formula
916: in the previous subsection. (The factor $\eta/\eta_0$ is
917: needed because $l_{BACF}$ is measured along
918: the real PE contour while $l_e$ is measured
919: along the renormalized PE contour.) The results are
920: shown in the Table \ref{tab:le}. The two persistence
921: lengths are within 20-25\% of each other. This reasonably
922: good agreement between two different methods
923: shows that our calculations are consistent.
924: It further strengthens the conclusion of two previous
925: sections that OSF theory is correct in describing
926: flexible polyelectrolytes.
927:
928:
929:
930: \section{Conclusion}
931:
932: \label{sec:concl}
933:
934: In this paper, we use extensive Monte Carlo simulation to
935: study the dependence of the electrostatic persistence
936: length of a polyelectrolyte on the screening
937: radius of the solution. Not only did we
938: simulate a much longer polyelectrolyte than
939: those studied in previous simulations in order to
940: show the scaling ranges, we also used a refined
941: analyses which take into account local
942: fluctuations to calculate the persistence length.
943: These improvements result in
944: a good support for OSF theory. They also
945: help to explain why previous simulations
946: failed to support OSF theory.
947:
948:
949:
950: In order to describe our numerical data we used
951: a modified OSF theory in the framework of ideas of KK.
952: Linear charge density $\eta$
953: was corrected to allow for short range fluctuations.
954: In our case this is a relatively small
955: correction to $\eta_0$ because
956: we deal with a strongly charged PE. When one crosses
957: over to sufficiently weakly charged PE linear charge density becomes
958: strongly renormalized and matches KK expressions.
959: We confirmed that corrections of $\eta$ do not affect $r_{s}^2$
960: dependence of persistence length which was predicted by OSF
961: for $l_0 \ll r_s \ll L$.
962: In other words, we confirm KK idea that at large
963: $r_s$ all effects of flexibility
964: of PE are limited to a renormalization of $\eta$.
965: At $r_s$ comparable to contour length $L$ we
966: found a good agreement of the numerical
967: data with OSF formula modified for
968: this case [Eq. (\ref{eq:lOSF2})],
969: which is derived in Ref. \onlinecite{Odijk}.
970: Again all effects of local flexibility
971: are isolated in the small correction to the linear
972: charge density $\eta$.
973:
974: \begin{acknowledgments}
975: The authors are grateful to A. Yu. Grosberg
976: M. Rubinstein, M. Ullner and R. Netz for useful discussions
977: and comments.
978: This work is supported by NSF No. DMR-9985785. T.T.N. is also
979: supported by the Doctoral Dissertation Fellowship of the
980: University of Minnesota.
981: \end{acknowledgments}
982:
983:
984: \begin{thebibliography}{99}
985:
986: \bibitem{Odijk}T. Odijk, J. Polym. Sci., Polym. Phys. Ed. {\bf 15}
987: (1977) 477.
988: \bibitem{SF}J. Skolnick, and M. Fixman, Macromolecules {\bf 10}
989: (1977) 944.
990: \bibitem{BJ}J.-L. Barrat, and J.-F. Joanny, Europhys. Lett.
991: {\bf 24} (1993) 333; J.-L. Barrat, and J.-F. Joanny,
992: Adv. Chem. Phys., {\bf 94} (1996) 1.
993: \bibitem{KK}A. R. Khokhlov, and K. A. Khachaturian, Polymer {\bf 23}
994: (1982) 1793.
995: \bibitem{Witten} Hao Li, and T. A. Witten, Macromolecules
996: {\bf 28} (1995) 5921.
997: \bibitem{Netz}R. R. Netz, and H. Orland, Eur. Phys. J. B
998: {\bf 8}, 81 (1999).
999: \bibitem{Thirumalai2} B. -Y. Ha, and D. Thirumalai, J. Chem.
1000: Phys. {\bf 110}, 7533 (1999).
1001: \bibitem{Dawson}D. Bratko, and K. A. Dawson, J. Chem. Phys.
1002: {\bf 24} (1993) 5352.
1003: \bibitem{Thirumalai}B. -Y. Ha, and D. Thirumalai,
1004: Macromolecules {\bf 28} (1995) 577.
1005: \bibitem{BB}J. L. Barrat, and D. Boyer, J. Phys. II France {\bf 3}
1006: (1993) 343.
1007: \bibitem{Reeds}C. E. Reed, and W. F. Reed, J. Chem. Phys.
1008: {\bf 94}, 8479 (1991).
1009: \bibitem{Seidel}C. Seidel, Ber. Bunsen-Ges. Phys. Chem.
1010: {\bf 100}, 757 (1996).
1011: \bibitem{Ullner2}M. Ullner, B. J\"{o}nsson, C. Peterson,
1012: O. Sommelius, and B. S\"{o}derberg, J. Chem. Phys.
1013: {\bf 107}, 1279 (1997);
1014: \bibitem{Ullner3}M. Ullner, and C. E. Woodward,
1015: Macromolecules {\bf 35}, 1437 (2002).
1016: \bibitem{Kremer}U. Micka, and K. Kremer, Phys. Rev. E {\bf 54} (1996) 2653.
1017: \bibitem{Ullner4}M. Ullner, in ``Handbook of Polyelectrolytes
1018: and Their Applications", J. Kumar S.
1019: Tripathy and H. S. Nalwa, eds, American Scientific Publishers, Los
1020: Angeles (2002).
1021: \bibitem{Nguyen}T. T. Nguyen, and B. I. Shklovskii,
1022: cond-mat/0202168.
1023: \bibitem{Everaers}R. Everaers, A. Milchev, and V. Yamakov,
1024: cond-mat/0202199.
1025: \bibitem{pivot}M. Lal, Mol. Phys. {\bf 17}, 57 (1969);
1026: N. Madras and A. D. Sokal, J. Stat. Phys. {\bf 50}, 109 (1988);
1027: B. J\"{o}nsson, C. Peterson, and B. S\"{o}derberg,
1028: J. Phys. Chem. {\bf 99}, 1251 (1995); M. Ullner, B. J\"{o}nsson,
1029: B. S\"{o}derberg, and C. Peterson, J. Chem. Phys. {\bf 104},
1030: 3048 (1996).
1031: %T. Wallin, P. Linse, Langmuir {\bf 12} (1996) 305.
1032: \bibitem{Landau}L. D. Landau, and E. M. Lifshitz,
1033: {\it Statistical Physics}, Butterworth and Heinemann, Oxford, 1996.
1034:
1035: \end{thebibliography}
1036: \end{document}
1037: