1: \documentclass[epj]{svjour}
2: \usepackage{epsfig}
3:
4: \newcommand {\Rsq} {\left< R^{2} \right>}
5: \newcommand {\rsq} {\left< r^{2} \right>}
6: \newcommand {\Zst} {{\cal Z}}
7: \newcommand {\Ham} {{\cal H}}
8: \newcommand {\ij} {{\left<i,j<i\right>}}
9: \newcommand{\PA}{PA}
10: %\newcommand{\PA}{polyampholyte}
11: \newcommand{\PE}{PE}
12: \renewcommand{\DH}{Debye--H\"uckel\ }
13: \newcommand{\OSFKK}{OSFKK\ }
14: \newcommand{\KMBJ}{KMBJ\ }
15:
16:
17:
18: %\newcommand{\PE}{polyelectrolyte}
19:
20: \newcommand{\dft}{{\tilde{\delta\! f}}}
21: \newcommand{\df}{{\delta\! f}}
22: \newcommand{\Nt} {\tilde{N}}
23: \newcommand{\ct} {\tilde{c}}
24: \newcommand{\Psit} {\tilde{\Psi}}
25: \renewcommand{\textfraction}{0.0}
26:
27: \begin{document}
28:
29: \title{The Electrostatic Persistence Length of Polymers
30: beyond the OSF Limit}
31: \titlerunning{The Electrostatic Persistence Length of Polymers
32: beyond the OSF Limit}
33: \author{Ralf Everaers\inst{1}, Andrey Milchev\inst{1,2} and Vesselin Yamakov\inst{1,2}
34: \thanks{present address: Argonne National Laboratory,
35: Materials Science Division, Build. 212,
36: 9700 S. Cass Avenue, Argonne, IL--60439, USA}
37: }
38: \authorrunning{R. Everaers, A. Milchev and V. Yamakov}
39:
40: \institute{
41: Max--Planck--Institut f\"ur Polymerforschung,
42: Postfach 3148, D--55021 Mainz, Germany \and
43: Institute for Physical Chemistry, Bulgarian Academy
44: of Sciences, G. Bonchev Street, Block 11, 1113 Sofia, Bulgaria}
45:
46: \abstract{
47: We use large scale Monte Carlo simulations to test scaling theories
48: for the electrostatic persistence length $l_e$ of isolated,
49: uniformly charged polymers with \DH intrachain interactions in the
50: limit where the screening length $\kappa^{-1}$ exceeds the intrinsic
51: persistence length of the chains. Our simulations cover a
52: significantly larger part of the parameter space than previous
53: studies. We observe no significant deviations from the prediction
54: $l_e\propto\kappa^{-2}$ by Khokhlov and Khachaturian which is based
55: on applying the Odijk-Skolnick-Fixman theory to the stretched de
56: Gennes-Pincus-Velasco-Brochard polyelectrolyte blob chain. A linear
57: or sublinear dependence of the persistence length on the screening
58: length can be ruled out. We argue that previous numerical results
59: pointing into this direction are probably due to a combination of
60: excluded volume and finite chain length effects. The paper
61: emphasizes the role of scaling arguments in the development of
62: useful representations for experimental and simulation data.
63: }
64: \PACS{64.60.-i \and 36.20.-r \and 87.15.By}
65:
66: \maketitle
67: \section{Introduction}
68: The theoretical understanding of macromolecules carrying ionizable
69: groups is far from complete~\cite{BarratJoanny_acp_96,Ullner_handbook_02}. In spite of
70: the long range of the interactions, the systems are often discussed
71: using analogies to neutral polymers. A prominent example is the
72: concept of an electrostatic persistence length, which was introduced
73: more than 20 years ago by Odijk~\cite{Odijk_jpspp_77} and by Skolnick
74: and Fixman~\cite{SkolnickFixman_mm_77} (OSF). They considered a
75: semiflexible polymer or wormlike chain (WLC) with intrinsic
76: persistence length $l_0$ and Debye-H\"uckel (DH) screened
77: electrostatic interactions $U_{DH}/k_BT = (q^2 l_B/r) \exp(-\kappa r)$
78: between charges $q/e$ spaced at regular intervals $A$ along the chain.
79: The Bjerrum length $\l_B$ characterizes the strength of the
80: electrostatic interactions and is defined as the distance where
81: the Coulomb energy of two unit charges $e$ is equal to $k_BT$. Due to
82: the presence of mobile ions the bare Coulomb interaction is cut off
83: beyond the screening length $\kappa^{-1}$. OSF were interested in
84: bending fluctuations and considered the resulting increase of the
85: electrostatic energy relative to the straight ground state. In the
86: (``OSF'') limit $\kappa^{-1}\ll l_0$ where the screening length is
87: smaller than the intrinsic persistence length of the chain and to
88: lowest order in the local curvature, the \DH interaction makes an
89: additive contribution to the bending rigidity. As a consequence, a
90: WLC with DH interactions (DHWLC) behaves in this limit on large length
91: scales like an ordinary WLC with renormalized persistence length
92:
93: \begin{eqnarray}
94: l_p = l_0 + l_{OSF}\\
95: l_{OSF} = \frac{q^2 l_B}{4 A^2 \kappa^2}
96: \label{eq:OSF}
97: \end{eqnarray}
98: Ever since, there has been a lively debate on how to extend the theory
99: to parameter ranges beyond the OSF limit $\kappa^{-1}\ll l_0$. Barrat
100: and Joanny (BJ)\cite{BarratJoanny_epl_93} have shown that the original
101: OSF derivation breaks down, if the chains start to bend significantly
102: on length scales comparable to the screening length. As a
103: consequence, Eq.~(\ref{eq:OSF}) cannot simply remain valid beyond the
104: OSF limit as was sometimes speculated~\cite{Odijk_jpspp_78}. Two main
105: scenarios, which we denote by ``\OSFKK'' and ``\KMBJ'' after the
106: initials of the main authors, have been discussed in the literature:
107: \begin{description}
108: \item[\OSFKK] According to Khokhlov and Khachaturian
109: (KK)~\cite{KhokhlovKhachaturian_pol_82} the OSF theory can be
110: applied to a ``stretched chain of polyelectrolyte blobs'', a concept
111: introduced by de Gennes et al.~\cite{GPVB_jp_76} to describe the
112: behavior of weakly charged flexible polyelectrolytes in the absence
113: of screening. The persistence length of the blob chain is then
114: calculated from Eq.~(\ref{eq:OSF}) using suitably renormalized
115: parameters.
116: \item[\KMBJ] Refinements by
117: Muthukumar~\cite{Muthukumar_jcp_87,Muthukumar_jcp_96,Muthukumar_jcp_01}
118: of the original theory of Katchalsky~\cite{Katchalsky_ahca_48} treat
119: electrostatic interactions in strict analogy to short-range excluded
120: volume interactions. Quite interestingly, the results are consistent
121: with the scaling picture of de Gennes et al.~\cite{GPVB_jp_76} in
122: the two limits of strong and vanishing screening. Moreover, they are
123: supported by recent calculations by BJ and
124: others~\cite{BarratJoanny_epl_93,Ha_mm_95} who determined the
125: persistence length of the blob chain in a variational procedure and
126: found $l_e\sim\kappa^{-1}$.
127: \end{description}
128: In addition, there is a number of recent theories which fall into
129: neither of the two classes outlined
130: above~\cite{VilgisWilder_ctps_98,WilderVilgis_pre_98,LiverpoolStapper_epl_97,LiverpoolStapper_epje_01,HansenPodgornik_jcp_01}.
131: While there is a growing consensus among theoreticians that the \OSFKK
132: result is {\em asymptotically}
133: correct~\cite{LiWitten_mm_95,Ha_jcp_99,NetzOrland_epjb_99},
134: experiments~\cite{Tricot_mm_84,Reed_mm_91,Foerster_jpc_92,FoersterSchmidt_aps_95,Beer_mm_97}
135: as well as computer
136: simulations~\cite{Micka_pre_96,Ullner_jcp_97,Ullner_mm_02} have
137: consistently provided evidence for a comparatively weak
138: $\kappa$-dependence of the electrostatic presistence length.
139: %a non-perturbative $1/d$-expansion~\cite{HansenPodgornik_jcp_01}
140: %yielding $l_e\sim\kappa^{-7/6}$, a self-consistent variational
141: %computation of an effective field
142: %theory~\cite{VilgisWilder_ctps_98,WilderVilgis_pre_98}, and a
143: %field-theoretic renormalization group
144: %analysis~\cite{LiverpoolStapper_epl_97,LiverpoolStapper_epje_01}.
145:
146:
147: The purpose of the present paper is to shed some new light on this
148: problem by combining a scaling analysis
149: with large scale Monte Carlo simulations. We reexamine the \KMBJ and
150: the \OSFKK theory in order to extract guidance for the data analysis
151: and the choice of simulation parameters. As a result we are able (i)
152: to rule out the \KMBJ theory, (ii) to provide benchmark results for
153: analytical solutions of the DHWLC model as well as (iii) for a
154: comparison to experiments in order to clarify if the systems under
155: consideration are actually described by the DHWLC model or if
156: additional effects such as solvent quality or counter-ion condensation
157: need to be taken into account as well.
158:
159: The paper is organized as follows: In
160: section~\ref{sec:choice_of_model} we review the predictions of the
161: \KMBJ and \OSFKK theories, followed by a discussion in
162: section~\ref{sec:consequences} of how experiments and simulations
163: should be set up and analyzed in order to discriminate between the two
164: scaling pictures. Details of our Monte Carlo simulations can be found
165: in in section~\ref{sec:method}. We present our results in
166: section~\ref{sec:results} and close with a brief discussion.
167:
168:
169: \section{Scaling theories of intrinsically flexible \DH chains}
170: \label{sec:choice_of_model}
171: The DHWLC is characterized by the following set of parameters: $q$,
172: $l_B$, $\kappa$, $A$, $f$, $l_0$, and $L_{tot}$, where $L_{tot}$
173: denotes the total chain length and $f\le1$ the fraction of ionized
174: charged groups which needs to be determined independently for
175: experimental systems. Setting $f=l_B/A$ is a rough way of accounting
176: for Manning condensation~\cite{BarratJoanny_acp_96} of counter ions in
177: cases where $l_B<A$. Correlation functions can be calculated for chain
178: segments with arbitrary contour length $L<L_{tot}$. We focus on the
179: non-OSF limit $l_B,A,l_0\ll \kappa^{-1} \ll \sqrt{\rsq(L_{tot})}$
180: where the screening length is larger than all microscopic length
181: scales of the polymer model but smaller than the size of the entire
182: chain.
183:
184: For an understanding of the physics, some of these length scales and
185: parameters are less relevant than others. For example, the actual
186: distribution of the charges on the chain should be unimportant as long
187: as $A\ll\kappa^{-1}$. In our simulations we use discrete charges
188: spaced by a distance equal to the intrinsic
189: persistence length, while the scaling arguments assume a continuous
190: charge distribution. Similarly, in the non-OSF limit with
191: $l_0\ll\kappa^{-1}$ the physics should not depend on the details of
192: the WLC crossover from rigid rod to random coil behavior for $L\approx
193: l_0$. In our simulations we therefore use freely jointed chains (FJC)
194: whose (Kuhn) bond length $b$ corresponds (up to a henceforth
195: neglected factor of two) to the persistence length of a WLC. Finally,
196: it is convenient to consider the limit of infinite total chain length
197: in order to eliminate the $L_{tot}$ dependence.
198: Again the practical limitation to $N_{tot}=L_{tot}/b=4096$ segments
199: should be unimportant, since our chains always fulfill the condition
200: $\kappa^{-1} \ll \sqrt{\rsq(L_{tot})}$.
201:
202: %For our present purposes, we are left with three independent
203: %dimensionless parameters: the coupling constant $u=q^2 l_B/(A/f)$
204: %describing the strength of the electrostatic interactions between
205: %adjacent charges, the screening constant $(\kappa (A/f))^{-1}$ and the
206: %segment length $N=L/b$ for which we want to calculate correlation
207: %functions such as end-to-end distances.
208:
209: The remaining independent parameters (the line charge density $f q/A$,
210: $l_B$, $\kappa$, $l_0$ and $L$) can be reduced further using the
211: notion of a ``polyelectrolyte blob'' which was introduced by de Gennes
212: et al.~\cite{GPVB_jp_76} to describe the crossover from locally
213: unchanged chain statistics to stretching on long length scales.
214:
215: Consider first weakly charged flexible polyelectrolytes, where the
216: electrostatic interactions are irrelevant on the length scales
217: comparable to the intrinsic persistence length $l_0$. On larger
218: length scales an undisturbed WLC with a contour length $L$ has a
219: spatial extension $\rsq=2 L l_0$. Neglecting prefactors, the
220: electrostatic energy of such a chain is given by
221: $U_e/k_BT \simeq q^2 (f L/A)^2 l_B/\sqrt{\rsq}$.
222: Electrostatic interactions become
223: relevant for $U_e/k_BT\ge1$ or chain lengths $L$ exceeding
224:
225: \begin{eqnarray}
226: l_g & = & l_0^{1/3} \left(\frac {A^2}{f^2 q^2 l_B}\right)^{2/3}
227: \label{eq:ge}
228: \end{eqnarray}
229: and whose spatial extension is given by
230: \begin{eqnarray}
231: \xi &=& l_0^{2/3} \left(\frac {A^2}{f^2 q^2 l_B}\right)^{1/3} .
232: \label{eq:xe}
233: \end{eqnarray}
234: However, this derivation breaks down for strongly interacting systems
235: where the contour length per blob becomes smaller than the intrinsic
236: persistence length. In this case, a similar argument can be made for
237: a WLC with $L<l_0$ and $\rsq= L^2$ yielding
238:
239: \begin{equation}
240: l_g = \xi = \frac {A^2}{q^2 f^2 l_B}.
241: \label{eq:gexe}
242: \end{equation}
243: Both definitions match for $l_g=l_0$, hiding a subtle
244: crossover~\cite{Ha_jcp_99} behind a crudely renormalized system of
245: units. Throughout the paper all quantities will be expressed using
246: these natural units of contour length and spatial distance. On a
247: scaling level, they become a function of only {\em two} dimensionless
248: parameters: the reduced chain or segment length
249:
250: \begin{equation}\label{eq:X}
251: X = L/l_g
252: \end{equation}
253: and the reduced screening length
254:
255: \begin{equation}\label{eq:Y}
256: Y = (\kappa \xi)^{-1} \ .
257: \end{equation}
258: %While our extension Eq.~(\ref{eq:gexe}) of the blob scaling
259: %to strongly charged polyelectrolytes is plausible, its
260: %verification of the validity of this approach
261:
262: \begin{figure}[t]
263: \begin{center}
264: \epsfig{file=BJKKMap.eps,width=8cm}
265: \caption{ \label{fig:BJKKMap}
266: Conformations of DHWLC beyond the OSF limit as a function of
267: reduced chain length and reduced screening length.
268: Electrostatic interactions are relevant in the upper right part of
269: the map which is limited by solid black lines.
270: % The solid black lines indicate the chain length beyond which
271: % electrostatic interactions become relevant.
272: The dashed and dotted lines correspond to crossover lines in the
273: \OSFKK and \KMBJ scaling theories respectively. We have also
274: included the scaling predictions Eqs.~(\protect\ref{eq:r2_blob})
275: to (\protect\ref{eq:r2_SAW_KK}) for the chain radii. For
276: details see the main text.}
277: \end{center}
278: \end{figure}
279:
280:
281:
282:
283: In the model under consideration the electrostatic interactions are
284: purely repulsive. Therefore the chains are always extended relative
285: to the neutral case. For $\kappa l_B\ll 1$ the details of the process
286: are quite involved and difficult to treat from first principles. The
287: approximation schemes used are often based on mechanical analogies
288: such as stretching, bending and swelling due to short-range excluded
289: volume interactions and have been reviewed in
290: Ref.~\cite{BarratJoanny_acp_96,Ullner_handbook_02}. We have
291: summarized the various predictions in terms of a schematic map of the
292: $XY$ parameter space (Figure~\ref{fig:BJKKMap}). Before we discuss
293: the controversial parts, we first present those aspects which seem
294: well understood:
295:
296: \begin{itemize}
297: \item The \DH interaction is irrelevant inside the
298: electrostatic blob, i.e. for weakly charged chains
299:
300: \begin{equation}\label{eq:r2_blob}
301: \rsq/\xi^2 \simeq X \ \ \ \ \ \mbox{for $X<1$.}
302: \end{equation}
303: For strongly charged chains, this regime does not exist.
304: \item In the absence of screening, when the monomers interact via an
305: {\em infinite range} Coulomb potential, the chains are stretched
306: into a ``blob pole'':
307:
308: \begin{equation}\label{eq:r2_blobpole}
309: \rsq/\xi^2 \simeq X^2 \ \ \ \ \ \mbox{for $1<X<\infty$ and $Y\rightarrow\infty$}
310: \end{equation}
311: In Figure~\ref{fig:BJKKMap} we have marked the line dividing these
312: two regimes as ``GPVB'' after the initials of the authors of
313: Ref.~\cite{GPVB_jp_76} where the notion of the electrostatic blob was
314: introduced.
315:
316: \item For sufficiently long chains, the \DH interaction becomes effectively
317: short--ranged, leading to self-avoiding walk (SAW) behavior
318:
319: \begin{equation}\label{eq:r2_SAWgeneral}
320: \rsq/\xi^2 \sim X^{2\nu} \ \ \ \ \ \mbox{for $0<Y<\infty$ and $X\rightarrow\infty$}
321: \end{equation}
322: where $\nu \approx 3/5$ is the usual Flory exponent.
323:
324: \item For strong screening with $q^2 l_B<\kappa^{-1}<A$ the \DH interaction
325: reduces to an ordinary excluded volume potential with a second virial
326: coefficient $v\simeq q^2 l_B \kappa^{-2}$ between charges.
327: Using a conventional Flory argument to balance the two-body repulsion
328: $v (f L/A)^2/R_F^3$ with the entropic elasticity of a Gaussian chain
329: $R_F^2/(L l_0)$, one obtains:
330:
331: \begin{equation}\label{eq:r2_SAW_Ysmall}
332: \rsq/\xi^2 \simeq \left\{\begin{array}{ll}
333: X & \mbox{for $X<Y^{-4}$}\\
334: Y^{4/5} X^{6/5} & \mbox{for $X>Y^{-4}$}\\
335: \end{array}\right.
336: \ \ \mbox{for $Y\ll1$}
337: \end{equation}
338: In Figure~\ref{fig:BJKKMap} the corresponding line,
339: beyond which the short range excluded volume interaction becomes
340: relevant, is marked as ``F''.
341: \end{itemize}
342:
343: The controversial parts of the phase diagram concern the crossover
344: from the blob pole to the self-avoiding walk regime. The problem is often
345: treated in analogy to a simple WLC.
346: %the thermal undulations of worm-like chains
347: %with a certain bending rigidity $E$, which are stiff with
348: %$\rsq = L^2$ on length scales below the persistence length
349: %$l_p = E/k_BT$ and flexible with $\rsq\simeq L l_p$ for $L\gg l_p$.
350: With an Onsager virial coefficient $v\simeq l_p^2 d$ between rigid segments
351: of length $l_p$ and diameter $d$, excluded volume effects become relevant
352: beyond a ``Flory length''
353: $l_F=l_p^3/d^2$ leading to a Flory radius of
354: $R_F^2 \simeq d^{2/5} l_p^{8/5} (L/l_p)^{6/5}$.
355: In the case of the blob chain, the diameter of the electrostatically
356: excluded volume is given by $d\simeq\kappa^{-1}$
357: \cite{Odijk_jpspp_78,FixmanSkolnick_mm_78}. However, there
358: is disagreement with respect to the $\kappa$--dependence
359: of the electrostatic persistence length $l_e$.
360:
361: \begin{itemize}
362: \item Variational approaches such as the theory of Barrat and Joanny
363: (BJ) often predict $l_e \simeq d \simeq \kappa^{-1}$. As a
364: consequence, $l_F = l_p =\kappa^{-1}$ so that there is a direct
365: crossover from the stiff blob chain to a SAW regime when the contour
366: length $\xi \frac Ng$ of the blob chain reaches the screening length
367: $\kappa^{-1}$. Using dimensionless units this corresponds to
368: $X= N/g=l_F/\xi = l_p/\xi =1/(\kappa\xi)=Y$ (the dotted line
369: in Figure~\ref{fig:BJKKMap} marked ``BJ''). The result
370:
371: \begin{equation}\label{eq:r2_SAW_BJ}
372: \rsq/\xi^2 \simeq Y^{4/5} X^{6/5} \ \ \ \ \ \mbox{for $X>Y$ and $Y\gg1$}
373: \end{equation}
374: is identical to Eq.~(\ref{eq:r2_SAW_Ysmall}). On a scaling level,
375: the predictions of the BJ theory coincide with those of the
376: excluded volume theories of Katchalsky~\cite{Katchalsky_ahca_48} and
377: Muthukumar~\cite{Muthukumar_jcp_87,Muthukumar_jcp_96,Muthukumar_jcp_01}.
378: \item
379: Most theories favour the relation $l_e\simeq \xi^{-1}\kappa^{-2}$ first
380: obtained by Khokhlov and Khachaturian (KK). KK argued that the OSF
381: result Eq.~(\ref{eq:OSF}) should also apply to a stretched
382: chain of blobs with line charge density $f q l_g/(A\xi)$
383: so that $l_e/\xi = 1/(\kappa \xi)^2=Y^2$.
384: Since $l_e\gg d$, the resulting phase diagram is considerably
385: more complicated. The blob chain starts to bend for
386: reduced segment lengths $X$ exceeding $l_p/\xi = l_e/\xi = Y^2$
387: (the ``$\mathrm{KK}_1$'' line in Figure~\ref{fig:BJKKMap}),
388: while excluded volume effects become
389: relevant beyond
390: $l_F/\xi= 1/(\kappa \xi)^4 =Y^4$ (``$\mathrm{KK}_2$'').
391:
392: \begin{equation}\label{eq:r2_SAW_KK}
393: \rsq/\xi^2 \simeq \left\{\begin{array}{ll}
394: Y^2 X & \mbox{for $Y^{2}<X<Y^{4}$}\\
395: Y^{6/5} X^{6/5} & \mbox{for $X>Y^{4}$}\\
396: \end{array}\right.
397: \ \ \mbox{for $Y\gg1$}
398: \end{equation}
399: Finally, and in contrast to the \KMBJ theory, the \OSFKK approach
400: implies another crossover (``$\mathrm{KK}_3$'') within the SAW regime at $Y=1$
401: from Eq.~(\ref{eq:r2_SAW_KK}) to Eq.~(\ref{eq:r2_SAW_Ysmall}).
402: \end{itemize}
403:
404:
405:
406: \section{Implications for Data Production and Analysis}
407: \label{sec:consequences}
408: In general, scaling theories make two kinds of predictions: (i) about
409: the existence of characteristic length scales or crossover lines in
410: conformation space and (ii) about the {\em asymptotic} behavior of
411: observables in the areas between these crossover lines. In principle,
412: attempts at refutation can aim at either type of prediction.
413: However, in the present case the identification of asymptotic
414: exponents turns out to be particularly difficult. Apart from
415: numerical prefactors and logarithmic corrections~\footnote{For
416: example, in the absence of screening the long-range Coulomb
417: interactions along the blob pole create a tension which grows
418: logarithmically with the chain length $N$. As a consequence, the
419: blob diameter is reduced and the chains grow with $R_g \sim N
420: \log^{1/3}(N)$. In particular, a $N$ monomer segment of a longer
421: chain will always be more extended than a $N$
422: monomer chain whereby the deformation is strongest for segments
423: located near the center of the longer
424: polymer~\cite{GPVB_jp_76,CastelnovoSensJoanny_epje_00}. In the
425: presence of screening, this effect leads on the one hand to an
426: increase of the contour length of the blob chain. On the other
427: hand, the correspondingly reduced line charge density results in a
428: reduction of the \OSFKK persistence length. Similarly, but neglecting
429: the stretching, there are logarithmic corrections to the
430: electrostatically excluded volume around the blob
431: chain~\cite{FixmanSkolnick_mm_78}. Moreover, although fairly
432: robust, the Flory argument used to estimate the excluded volume
433: effects is far from exact. In addition to the aforementioned
434: crossovers, a complete theory will have to account for all of these
435: effects.} one is faced with four problems:
436: \begin{itemize}
437: \item Although the various regimes predicted by the \KMBJ and the \OSFKK
438: theory are characterized by different combinations
439: of powers of $X$ and $Y$, the exponents are often similar and the
440: {\em absolute} differences between the predicted chain extensions
441: relatively small.
442: \item At least on a scaling level all crossover lines meet at $X=Y=1$
443: for chain and screening length of the order of the diameter of the
444: polyelectrolyte blob. In the case of the \OSFKK theory some of the
445: predicted regimes are extremely narrow in the sense that chain
446: lengths of $X=10^4$ blobs are required for a width of one order of
447: magnitude in $Y$-direction. This validity range would seem to be the
448: absolute minimum for identifying power law behavior.
449: \item The crossover lines are neither parallel to each other nor to
450: the ``natural'' $X$ and $Y$ directions of variations of chain
451: and screening length respectively.
452: \item In particular, results will be influenced by the finite total
453: length $L_{tot}$ of polyelectrolyte chains studied in experiments or
454: simulations. The importance of these effects varies with the ratio
455: of the screening length $Y$ and the contour length $X$ of the blob
456: chain. As a consequence, they risk to mask the asymptotic
457: $Y$-dependence of observables such as the electrostatic persistence
458: length, if they are evaluated for chains with {\em fixed} $L_{tot}$.
459: \end{itemize}
460: Quite obviously, the discrimination between the two scaling pictures
461: requires the investigation of chains whose {\em effective} length $X$ is as
462: large as possible. In addition one should rely on those observables and
463: data representations which are {\em most} sensitive to the differences
464: between the theories and {\em least} sensitive to the omitted constants,
465: corrections and crossovers. In the following we discuss the analysis
466: of data for internal distances and for the tangent correlation function.
467: In particular, we will argue that it is relatively easy to discard the \KMBJ
468: picture using simple scaling plots, while the verification of
469: some of the predictions by the \OSFKK
470: theory requires astronomical chain lengths.
471:
472:
473: \begin{figure}[t]
474: \begin{center}
475: \epsfig{file=SimExpXY.eps,width=8cm}
476: \caption{ \label{fig:SimExpXYMap}
477: Areas of the conformation diagram Fig.~\protect\ref{fig:BJKKMap}
478: for which there are experimental and simulation data available.
479: In (a) the green lines indicate the parameter ranges
480: investigated in previous numerical studies: Barrat and
481: Boyer~\protect\cite{BarratBoyer_jp2_93} (--- ---),
482: Seidel~\protect\cite{Seidel_bunsen_96} (------), J\"onsson et
483: al.~\protect\cite{Jonsson_jpc_95,Ullner_jcp_97}
484: ($\cdots\cdots$), Micka and Kremer~\protect\cite{Micka_pre_96}
485: (-- -- --). Experiments (shown in red) have access to longer
486: chains, but the reduced screening length $Y$ are typically
487: smaller than ten: Reed et al.~\protect\cite{Reed_mm_91} (-- --
488: --), Beer et al.~\protect\cite{Beer_mm_97} (------). The
489: colored grids in (b) denote the parameter ranges covered by
490: different sets of our MC simulations. Note that the predictions
491: of the \KMBJ and the \OSFKK theory differ most strongly along the
492: $\mathrm{KK}_1$ line and are identical outside of the the gray
493: shaded area. }
494: \end{center}
495: \end{figure}
496:
497: Consider first the scaling predictions Eqs.~(\protect\ref{eq:r2_blob})
498: to (\protect\ref{eq:r2_SAW_KK}) for the mean square internal distances
499: at reduced screening lengths $Y>1$. Describing the GPVB crossover to
500: the blob pole within the chain-under-tension
501: model~\cite{GPVB_jp_76,BarratBoyer_jp2_93}, Eqs.~(\ref{eq:r2_blob}) and
502: (\ref{eq:r2_blobpole}) can be combined as $\rsq/\xi^2=X+X^2$. Taking
503: this into account, the \KMBJ theory suggests that {\em all} data points
504: should collapse when plotted in the following manner as a function of
505: the \KMBJ persistence length:
506:
507: \begin{equation}\label{eq:r2BJScaling}
508: \frac{\langle r^2(Z=X/Y)\rangle/\xi^2-X}{X^2}
509: \simeq
510: \left\{\begin{array}{ll}
511: 1 & \mbox{for $Z<1$}\\
512: Z^{-4/5} & \mbox{for $Z>1$}\\
513: \end{array}\right.
514: \end{equation}
515: In contrast, the \OSFKK theory predicts data collapse, if the segment lengths
516: are rescaled with the \OSFKK persistence length $Y^2$, and {\em a breakdown
517: of scaling} for segment lengths approaching
518: the Flory length $Y^4$:
519:
520: \begin{equation}\label{eq:r2KKScaling}
521: \frac{\langle r^2(Z=X/Y^2)\rangle/\xi^2-X}{X^2}
522: \simeq
523: \left\{\begin{array}{ll}
524: 1 & \mbox{for $Z<1$}\\
525: Z^{-1} & \mbox{for $1<Z<Y^2$}\\
526: Y^{-2/5}Z^{-4/5} & \mbox{for $Y^2<Z$}\\
527: \end{array}\right.
528: \end{equation}
529: The predictions of the two scaling theories differ
530: most strongly for chain radii along the
531: $Y=X^{1/2}$ $\mathrm{KK}_1$ line for segment lengths equal to the \OSFKK
532: persistence length. In the \KMBJ theory, this line is already deep in
533: the SAW regime. Eqs.~(\ref{eq:r2_SAW_BJ}) and (\ref{eq:r2_SAW_KK}) imply
534:
535: \begin{equation}\label{eq:r2_XeqY2}
536: \frac {\langle r^2(X,Y=X^{1/2})\rangle}{X\xi^2} \simeq
537: \left\{\begin{array}{ll}
538: X^{3/5} & \mbox{(\KMBJ)}\\
539: X & \mbox{(\OSFKK)}\\
540: \end{array}\right.
541: \end{equation}
542: Thus the ratio
543: $\rsq_{\OSFKK}/\rsq_{\KMBJ}=X^{2/5}$ is fairly small and
544: $X=10^{5/2}$ blobs are required for this ratio to become of order
545: 10.
546: For comparision, both theories predict full extension along the \KMBJ line
547:
548: \begin{equation}\label{eq:r2_XeqY}
549: \frac {\langle r^2(X,Y=X)\rangle}{X\xi^2} \simeq
550: \left\{\begin{array}{ll}
551: X & \mbox{(\KMBJ)}\\
552: X & \mbox{(\OSFKK)}\\
553: \end{array}\right.
554: \end{equation}
555: and differ only by a factor of $\rsq_{\OSFKK}/\rsq_{\KMBJ}=X^{1/10}$
556: along the $\mathrm{KK}_2$ line
557:
558: \begin{equation}\label{eq:r2_XeqY4}
559: \frac {\langle r^2(X,Y=X^{1/4})\rangle}{X\xi^2} \simeq
560: \left\{\begin{array}{ll}
561: X^{2/5} & \mbox{(\KMBJ)}\\
562: X^{1/2} & \mbox{(\OSFKK)}\\
563: \end{array}\right.
564: \end{equation}
565:
566: While segment lengths of the order of $X=10^3$ blobs are thus
567: sufficient to discriminate between the \KMBJ and the \OSFKK proposals
568: for the electrostatic persistence length, the requirements
569: for resolving the additional crossovers predicted by the \OSFKK
570: theory are much higher.
571: %Compared to a data analysis which traces properties {\em along} the
572: %predicted crossover lines, crossover scaling {\em across}
573: %these lines will run into even more severe problems.
574: Consider again the $\mathrm{KK}_2$ line where excluded volume
575: effects are expected to become relevant for the undulating blob chain.
576: Eq.~(\ref{eq:r2_SAW_KK}) can be rewritten in the form
577:
578:
579: \begin{equation}\label{eq:r2_SAW_KK_Z}
580: \frac{\langle r^2(Z=Y^{-4/5}X^{1/5})\rangle}{Y^2 X\xi^2} \simeq
581: \left\{\begin{array}{ll}
582: 1 & \mbox{for $Z<1$}\\
583: Z & \mbox{for $Z\ge 1$}\\
584: \end{array}\right.
585: \end{equation}
586: where $Z$ is the variable measuring the effective
587: distance from the crossover line at $Z=1$. In order to identify the
588: asymptotic behavior one needs at least data covering the interval
589: $Z\in [0.1,10]$. Since the validity range of
590: Eq.~(\ref{eq:r2_SAW_KK_Z}) is limited by the $\mathrm{KK}_1$ and
591: $\mathrm{KK}_3$ lines (so that $1<Y<X^{1/2}$ or
592: $X^{1/5}>Z>X^{-1/5}$), this implies a minimum segment length of $X=10^5$
593: blobs for establishing the $\mathrm{KK}_2$ crossover. Similarly the
594: $\mathrm{KK}_3$ crossover between Eqs.~(\ref{eq:r2_SAW_Ysmall})
595: and (\ref{eq:r2_SAW_KK}) at $Y=1$ becomes relevant for chains of
596: $X=10^{10}$ blobs!
597:
598: The mean-square internal distances and the tangent correlation
599: function (TCF) obey a Green-Kubo like relation:
600:
601: \begin{equation}\label{eq:BACF_GreenKubo}
602: \langle {\vec b}_N \cdot {\vec b}_0 \rangle =\frac12
603: \frac {d^2}{d\, N^2} \langle r^2(N) \rangle =\frac12
604: \frac {\xi^2}{g^2} \frac {d^2}{d\, X^2} \frac {\langle r^2(X) \rangle}{\xi^2}
605: \end{equation}
606: For a WLC the TCF is simply given by
607:
608: \begin{equation}\label{eq:bacf}
609: \langle {\vec b}(s) \cdot {\vec b}(0) \rangle = b^2 \exp(-s/l_p)
610: \end{equation}
611: so that the persistence length can be read off directly from a
612: semi-logarithmic plot. Numerical studies of
613: polyelectrolytes~\cite{Micka_pre_96,Ullner_jcp_97,Ullner_mm_02} have
614: therefore often focused on this quantity in spite of two intrinsic
615: problems: (i) the TCF is considerably more difficult to measure with
616: the same relative precision than internal distances and (ii) the TCF
617: is particularly sensitive to finite chain length effects (a
618: characteristic sign is the {\em faster} than exponential decay of the
619: TCF on length scale approaching the chain length). In contrast to the
620: case of ordinary SAWs~\cite{Schaefer_jp_99}, nothing is known about
621: the functional form of the corrections. In the following discussion we
622: will focus on a third aspect: the sensitivity of the TCF to the
623: neighborhood of crossover lines.
624:
625: On a scaling level, the behavior of the TCF can be obtained by
626: applying Eq.~(\ref{eq:BACF_GreenKubo}) to
627: Eqs.~(\protect\ref{eq:r2_blob}) to (\protect\ref{eq:r2_SAW_KK}).
628: For $Y>1$ the \KMBJ theory predicts
629:
630: \begin{equation}\label{eq:bacfBJ}
631: \frac {g^2}{\xi^2} \langle {\vec b}(Z=X/Y) \cdot {\vec b}(0) \rangle \simeq
632: \left\{\begin{array}{ll}
633: 1 & \mbox{for $Z<1$}\\
634: Z^{-4/5} & \mbox{for $Z>1$}\\
635: \end{array}\right.
636: \end{equation}
637: Within the \OSFKK theory, simple predictions can only be made
638: for segment lengths below the persistence length and beyond
639: the Flory length:
640:
641: \begin{equation}\label{eq:bacfKK}
642: \frac {g^2}{\xi^2} \langle {\vec b(Z=X/Y^2)} \cdot {\vec b}(0) \rangle \simeq
643: \left\{\begin{array}{ll}
644: 1 & \mbox{for $Z<1$}\\
645: Y^{-2/5} Z^{-4/5} & \mbox{for $Y^2<Z$}\\
646: \end{array}\right.
647: \end{equation}
648: However, since both theories are based on the analogy to a mechanical
649: WLC, they are often associated with the much more detailed prediction
650:
651: \begin{eqnarray}\label{eq:bacfWLC}
652: \lefteqn{\frac {g^2}{\xi^2} \langle {\vec b}(X) \cdot {\vec b}(0) \rangle \simeq }\\
653: &&
654: \left\{\begin{array}{ll}
655: \exp(-X/Y) & \mbox{for $X<Y$\ \ \ (\KMBJ)}\\
656: \exp(-X/Y^2) & \mbox{for $X<Y^2$\ \ \ (\OSFKK)}\\
657: \end{array}\right.
658: \nonumber
659: \end{eqnarray}
660: for the functional form of the decay of the tangent correlations.
661: Measuring this quantity for DHWLC therefore seems to be the most
662: direct way of justifying or refuting this analogy and its
663: exploitation. In particular, numerical
664: work~\cite{Micka_pre_96,Ullner_jcp_97,Ullner_mm_02} has concentrated on (i)
665: establishing the existence of a single exponential decay of the
666: TCF over a certain range of length scales and
667: (ii) extracting the $\kappa$-dependence of the measured decay length.
668: In the following we will reexamine this approach by taking a closer
669: look at Eqs.~(\ref{eq:bacfBJ}) and (\ref{eq:bacfKK}), since they
670: contain additional crossovers neglected in Eq.~(\ref{eq:bacfWLC}).
671:
672: The situation should be uncritical for the GPVB crossover where the
673: chain-under-tension models~\cite{GPVB_jp_76,BarratBoyer_jp2_93}
674: suggests that Eq.~(\ref{eq:bacfWLC}) remains valid for $X<1$. In
675: contrast, nothing is known in detail about the way the TCF crosses
676: over to the slow power law decay characteristic for the SAW behavior on
677: large length scales. However, matching Eq.~(\ref{eq:bacfWLC}) (which
678: only accounts for the local bending rigidity) with the asymptotic
679: behavior in Eqs.~(\ref{eq:bacfBJ}) and (\ref{eq:bacfKK}) shows that
680: the tangent-correlation function is much more sensitive to excluded
681: volume effects than the chain radii. This is most obvious for the \OSFKK
682: theory where the two limits match close to the \OSFKK persistence length
683: $X=Y^2$ instead of the Flory length $X=Y^4$. While the scaling of the
684: TCF with the \OSFKK persistence length should start to break down around
685: $X/Y^2\approx 1$, one can nevertheless expect Eq.~(\ref{eq:bacfWLC})
686: to hold up to this point. In the case of the \KMBJ theory the situation
687: is quite different, since the persistence length and the Flory length
688: coincide. As a consequence, Eq.~(\ref{eq:bacfWLC}) effectively breaks
689: down as soon as the tangent-correlation function starts to deviate
690: from one. On the other hand, in the absence of other relevant length
691: scales the TCF should scale with the \KMBJ persistence length for {\em
692: arbitrary} segment length!
693:
694: In our opinion, these arguments shed some doubts on attempts to
695: identify the electrostatic presistence length which are based
696: too closely on Eq.~(\ref{eq:bacfWLC}). Scaling plots testing
697: Eqs.~(\ref{eq:bacfBJ}) and (\ref{eq:bacfKK}) may offer a simpler
698: {\em and} safer alternative.
699:
700:
701:
702: \section{Simulation Model, Method and Parameters}
703: \label{sec:method}
704: \begin{figure*}[t]
705: \begin{center}
706: \epsfig{file=r2XY.eps,width=16cm}
707: \caption{ \label{fig:r2OverXPlot}
708: Comparison of measured internal distances $\langle
709: r^2(X,Y)\rangle/\xi^2$ to the predictions of the \KMBJ ((a) and
710: (c)) and \OSFKK ((b) and (d)) scaling theories. In the top row we
711: show log-log-log representations where all distances are
712: normalized to the undisturbed random walk. The colored areas
713: were generated by interpolation between the results of all
714: simulations for a given coupling strength. The supporting grids
715: and the crossover lines show the two sets of scaling predictions
716: Eqs.~(\protect\ref{eq:r2_blob}) to (\protect\ref{eq:r2_SAW_KK})
717: as an extensions of Fig.~(\protect\ref{fig:BJKKMap}) to three
718: dimensions. In the plots of the bottom row the colors indicate
719: the ratios $\langle r^2(X,Y)\rangle/\rsq_{\OSFKK}$ and $\langle
720: r^2(X,Y)\rangle/\rsq_{\KMBJ}$ respectively. }
721: \end{center}
722: \end{figure*}
723: %\cleardoublepage
724:
725:
726: As already mentioned in section~\ref{sec:choice_of_model},
727: we model the polymers as freely jointed chains (FJC) with unit
728: charges $q=1$ at each joint.
729: Lengths were measured in units of the bond length $b$.
730: We varied the the Bjerrum length
731: $l_B=4^2,1,1/4^2,1/16^2,1/100^2b$
732: and the screening length
733: $\kappa^{-1}= 1,2,4,8,16,32,64,128\ b$.
734: %All monomers in the simulation were equally charged
735: %with $q=4,1,1/4,1/16,1/100$, which can be interpreted
736: %as distances $A/b=1/4,1,4,16,100$ between unit charges $q_A=1$.
737: As in our previous study on
738: poly\-ampholytes~\cite{Yamakov_prl_00}, the chains have a length of
739: up to $N=4096$ monomers.
740: \begin{itemize}
741: \item Since we study the conformations of isolated chains, we employ
742: the efficient technique of pivot rotations due to Sokal et
743: al.\cite{MadrasSokal_jstatphys_88,Jonsson_jpc_95}. We use two types
744: of pivot moves: Either we rotate the part between the free end of
745: the chain and a randomly selected monomer around an axis, defined by
746: the bond between this monomer and its nearest neighbor; or we rotate
747: a segment between two randomly selected monomers around an axis
748: joining them. The latter provides better efficiency in the case of a
749: stretched chain with large excess charge. One MC step consists of
750: $N$ attempted rotations at random positions along the chain. Chain
751: conformations are stored at intervals of 8-32 MC steps. For each
752: parameter set we simulate 8 independent Markov chains in parallel.
753: We typically store $8\times60$ conformations representing a total of
754: $1.5\times 10^7$ attempted rotations for our longest chains.
755: \item Instead of the slower procedure of Stellman and
756: Gans\cite{Stellman_mm_72} which corrects the accumulating numeric error in
757: off-lattice implementations of the pivot algorithm, we regularly
758: reconstruct the chains with the correct bond length.
759: \item For calculating the long-range electrostatic interactions we
760: use a direct summation whereby the energy of the system is obtained
761: by direct counting of all the pair energies of the beads. This
762: method is still efficient for macromolecules of up to few thousand
763: monomers. The DH potential is tabulated in two
764: arrays for short and long distances respectively.
765: \item For better efficiency starting configurations of the chains are
766: created by means of the configurational biased\cite{FrenkelSmit_96}
767: MC method although one should keep in mind that due to the long
768: range interactions the first part of the newly grown chain does not
769: experience the cumulative field of the rest of the chain and a
770: number of rotational moves are still needed before the chain is well
771: equilibrated. Measurements are performed and conformations stored
772: only after the chain end-to-end distances are well equilibrated.
773: \item Since statistics is gathered both with respect to chain
774: conformations as well as to different Bjerrum length $l_B$ and
775: screening length $\kappa^{-1}$, we use a simple parallelization
776: where different processors of a CrayT3E supercomputer perform
777: independent simulation of single chains. The total CPU time used
778: for this project is of the order of $1.5 10^5$ single processor hours.
779: \end{itemize}
780: The simulation parameters translate into
781: our blob units as
782:
783: \begin{eqnarray}
784: g & = &\left(\frac b{l_B}\right)^{2/3} \ \ \ \ \ \ (g>1)
785: \label{eq:ge_large}\\
786: \xi &=& b \left(\frac b{l_B}\right)^{1/3}\ .
787: \label{eq:xe_large}
788: \end{eqnarray}
789: and
790:
791: \begin{eqnarray}
792: g & = &\frac b{l_B} \ \ \ \ \ \ (g\le1)
793: \label{eq:ge_small}\\
794: \xi &=& b \frac b{l_B}\ .
795: \label{eq:xe_small}
796: \end{eqnarray}
797: where we now use the number $g$ of monomers per blob instead
798: of the corresponding contour length $l_g=b g$.
799:
800: Fig. (\ref{fig:SimExpXYMap}) shows where our own data are located within the
801: $XY$-conformation space.
802: %We actually keep the grid representation of our
803: %data (typically spaced by powers of 2 in $X$- and $Y$-direction)
804: %throughout much of the data analysis, since the grid lines represent
805: %two typical experimental situations: variation of chain lengths under
806: %constant conditions and variation of the screening length for a given
807: %polymer sample.
808: The effective chain and screening lengths studied cover a range of
809: seven and five orders of magnitude respectively. The reduced mean
810: square internal distances vary over ten orders of magnitude. Along the
811: $\mathrm{KK}_1$ line our data extend {\em on a logarithmic scale}
812: about a factor of two further into the asymptotic regime than previous
813: studies. While this allows us to discriminate between the \KMBJ and the
814: \OSFKK predictions for the electrostatic presistence length, our chains
815: are still too short to resolve the different RW and SAW regimes
816: predicted by the \OSFKK theory.
817:
818:
819: Note, that only by studying strongly stretched chains
820: we are able to push the {\em effective} chain
821: length $X$ close to $10^5$ and that our unified description
822: of strongly and weakly charged flexible polyelectrolytes
823: needs to be confirmed by the data analysis. To facilitate the
824: comparison, all figures make use of the same color code to indicate
825: data obtained for a particular coupling strength $l_B/b$ ranging from
826: blue for $g=10000^{2/3}\approx470$ over different shades of violet
827: for $g=256^{2/3}\approx40$ and $g=16^{2/3}\approx6.4$ to red for
828: $g=1$ and orange for $g=1/16$. The first three systems can safely
829: be regarded as Gaussian chains, while the last two are at and beyond
830: the crossover to the strong stretching regime.
831:
832: \section{Results}
833: \label{sec:results}
834:
835: \begin{figure}[t]
836: \begin{center}
837: \epsfig{file=r2XeqY2.eps,width=7cm}
838: \caption{ \label{fig:r2XeqY2}
839: Extension of chain segments along the BJ, $\mathrm{KK}_1$, and
840: $\mathrm{KK}_2$ crossover lines in comparison to the predictions
841: of the \OSFKK (solid line) and the \KMBJ (dashed line) scaling
842: theories (see Eqs.~(\protect\ref{eq:r2_XeqY}),
843: (\protect\ref{eq:r2_XeqY2}), and (\protect\ref{eq:r2_XeqY4});
844: note that we have not used additional prefactors for this
845: comparison). Results for different coupling constants are
846: shifted by factors of $\sqrt{1000}$. }
847: \end{center}
848: \end{figure}
849:
850:
851:
852: \begin{figure}
853: \begin{center}
854: \epsfig{file=R2CrossOver.eps,width=8cm}
855: \caption{ \label{fig:LpCrossover}
856: Crossover scaling for internal distances versus segment length.
857: In the top (bottom) row segment lengths $X$ are normalized to
858: the \KMBJ (\OSFKK) persistence length $Y$ ($Y^2$) respectively.
859: Figures (a) and (c) on the left-hand side show internal
860: distances $\langle r^2(X,Y)\rangle$ normalized to the
861: mean-square extension $\xi^2 X^2$ of the blob pole. The grid is
862: the same as in Fig.~\ref{fig:SimExpXYMap}(b) and shows $const-X$ and
863: $const-Y$ lines. Figures (b) and (d) on the right-hand side are
864: inspired by the chain-under-tension model for Gaussian chains
865: Eqs.~(\protect\ref{eq:r2BJScaling}) and
866: (\protect\ref{eq:r2KKScaling}). Only $const-X$ lines are shown.
867: Data points falling into the range $1<Y<X<Y^2$ are marked using
868: the color code indicating the coupling strength. Results for
869: different coupling constants are shifted by factors of ten. }
870: \end{center}
871: \end{figure}
872:
873:
874: In the data analysis we mainly concentrate on identifying the scaling
875: behavior: (i) Do different data sets overlap when rescaled according
876: to our extension Eq.~(\ref{eq:gexe}) of
877: the definition of the electrostatic blob? (ii) How do the results of
878: our simulations compare to the predictions of the \KMBJ and \OSFKK scaling
879: theories? In terms of observables we start by presenting data for
880: internal distances averaged along our chains of total length $N=4096$.
881: In the second part, we discuss results for the tangent correlation
882: function. While the TCF was also averaged along the chain, we only
883: take into account distances up to half the chain length in order
884: to reduce finite-chain length effects~\cite{Schaefer_jp_99}. Except
885: for the most weakly charged system, the chain extensions are much
886: larger than the screening length, so that we do not expect finite
887: chain length effects to be very important. At the end, we briefly
888: present results for shorter chain lengths.
889:
890:
891: Figs.~\ref{fig:r2OverXPlot}(a) and (b) are
892: three-dimensional log-log-log plots giving an overview of all data.
893: We show reduced internal distances $\langle
894: r^2(X,Y)\rangle/((N/g)\xi^2)$ normalized to the size of
895: the undisturbed random walk as a function of the reduced chain and
896: screening lengths Eqs.~(\ref{eq:X}) and (\ref{eq:Y}) respectively.
897: Results for different coupling constants are combined into colored
898: surfaces, while the supporting grids show the two sets of scaling
899: predictions Eqs.~(\ref{eq:r2_blob}) to (\ref{eq:r2_SAW_KK}) as an
900: extensions of Fig.~\ref{fig:BJKKMap} to three dimensions.
901:
902: The complementary
903: Figs.~\ref{fig:r2OverXPlot}(c) and (d)
904: show the ratios $\langle r^2(X,Y)\rangle/\langle r^2(X,Y)\rangle_{\KMBJ}$
905: and $\langle r^2(X,Y)\rangle/\langle r^2(X,Y)\rangle_{\OSFKK}$ of the
906: interpolated simulation results to the scaling predictions in
907: a color coding where green, red and blue indicate agreement,
908: under- and overestimation by a factor of three or more respectively.
909: The advantage of this representation is the localization of
910: deviations in our schematic map of the parameter space.
911:
912: Qualitatively, the interpretation of Fig.~\ref{fig:r2OverXPlot}
913: seems clear. There are neither indications for a failure
914: of the blob scaling nor for significant deviations from the predictions
915: of the \OSFKK theory. (We emphasize again that we have neglected all numerical
916: prefactors and that Eqs.~(\ref{eq:r2_blob}) to (\ref{eq:r2_SAW_KK})
917: treat crossovers in the crudest manner).
918: In particular, there is no evidence that
919: the chains start to bend on length scales comparable to the
920: screening length as predicted by \KMBJ. In the relevant part of
921: conformation space, the \KMBJ theory systematically underestimates the
922: observed chain extensions.
923: %This conclusion is supported by all of the following more
924: %detailed comparisons.
925:
926: Nevertheless, Fig.~\ref{fig:r2OverXPlot} could be misleading, since
927: the rejection of the \KMBJ theory is mainly based on data falling into
928: the strong-stretching regime, while the theory is meant to apply to
929: weakly stretched Gaussian chains. Thus so far our conclusions rest on
930: the assumption that the extension Eq.~(\ref{eq:gexe}) of the blob
931: scaling to strongly charged chains can be used to {\em extrapolate}
932: the behavior of weakly charged systems to segment lengths inaccessible
933: by simulation. How well this assumption is fulfilled is hard to judge
934: from Fig.~\ref{fig:r2OverXPlot}. Definite conclusions require a more
935: detailed analysis.
936:
937: Fig.~\ref{fig:r2XeqY2} presents chain radii measured along the BJ,
938: $\mathrm{KK}_1$ and $\mathrm{KK}_2$ crossover lines. The first point
939: to note is that in all three cases we observe almost perfect scaling
940: of data obtained for different coupling constants.
941: Eqs.~(\ref{eq:r2_XeqY2}) to (\ref{eq:r2_XeqY4}) predict simple
942: crossovers at the blob size around $X=1$. In agreement with both
943: scaling pictures, we observe stretched blob chains along the BJ line.
944: The most important set of data are the radii measured along the
945: $\mathrm{KK}_1$ line for chains with a contour length $X=Y^2$ equal to
946: the \OSFKK persistence length. In agreement with
947: the \OSFKK theory we find a simple crossover around $X=1$ to stretched
948: blob chains. Contrary to the predictions of the \KMBJ theory the radii
949: are essentially identical to those observed along the BJ line and do
950: not show SAW behavior. In particular, the asymptotic slope predicted
951: by the \OSFKK theory is already observable for weakly charged chains to
952: which the \KMBJ theory can be applied directly. The last set of data is
953: taken along the $\mathrm{KK}_2$ line which marks the onset of excluded
954: volume effects in the \OSFKK theory. Here our results are consistent with
955: the predictions of both theories. This observation is in agreement
956: with the estimate of a minimum segment length of $X=10^{10}$ blobs for the
957: difference to become relevant (see Eq.~(\ref{eq:r2_XeqY4})).
958: %We show these results in order to illustrate that the method of
959: %plotting data along crossover lines derived from a scaling theory has
960: %some intrinsic limits due to the fact that the location of the
961: %crossovers is only known up to a prefactor.
962:
963: \begin{figure}[t]
964: \begin{center}
965: \epsfig{file=BACF.eps,width=8cm}
966: \caption{ \label{fig:BACF}
967: Scaled tangent correlation functions. Results for different
968: coupling constants are shifted by factors of ten. The colored
969: lines mark the results of our fits to a simple exponential decay
970: in the range $Y<X<Y^2$. (a) log-log plot using \KMBJ scaling, (b)
971: log-log plot using \OSFKK scaling, (c) semi-log plot using \OSFKK scaling. }
972: \end{center}
973: \end{figure}
974: The screening length dependence of the effective bending rigidity of
975: the blob chain can also be determined from crossover scaling of
976: internal distances normalized to the size of the stretched blob pole
977: (Fig.~\ref{fig:LpCrossover}). The disadvantage of plotting $\langle
978: r^2(X,Y)\rangle/(\xi^2 X^2)$ directly (Figs.~\ref{fig:LpCrossover} (a)
979: and (c)) is the occurence of a $1/X$ divergence of results for segment
980: lengths smaller than the blob size. Correcting for this in the manner
981: suggested by the chain-under-tension model
982: (Eqs.~(\ref{eq:r2BJScaling}) and (\ref{eq:r2KKScaling})) as in
983: Figs.~\ref{fig:LpCrossover} (b) and (d) largely eliminates effects due
984: to the GPVB crossover, but introduces some artifacts for small $N$
985: where $\langle r^2(N=1)\rangle -b^2 \equiv 0$ for a FJC. In agreement
986: with our previous results and independently of the coupling constant
987: we observe extremely poor scaling when the data are plotted as a
988: function of the ratio $X/Y$ of chain length over \KMBJ persistence
989: length. In constrast, the data superimpose considerably better, if the
990: \OSFKK scaling is used as in the corresponding
991: Figs.~\ref{fig:LpCrossover}(c) and (d). In particular,
992: Fig.~\ref{fig:LpCrossover} eliminates the possibility of an
993: electrostatic persistence length scaling like $\kappa^{-1}$ but with
994: an unusually large prefactor.
995:
996:
997: \begin{figure}
998: \begin{center}
999: \epsfig{file=FittedLp.eps,width=8cm}
1000: \caption{ \label{fig:FittedLp}
1001: Location of apparent electrostatic persistence lengths
1002: $l_{e,app}$ in our schematic map of the $XY$ parameter space. We
1003: show results for the reduced crossover distance
1004: $X_{cd}=l_{cd}/\xi$ (a) and the reduced orientational
1005: correlation length $X_{oc}=l_{oc}\xi$ (b) for chains of
1006: length $N_{tot}=4096$ (solid lines) and $N_{tot}=256$ (dashed
1007: lines). In the first case, we compare $\rsq(X,Y)$ to the size of
1008: the stretched blob pole Eq.~\protect\ref{eq:r2_blobpole} and
1009: define $X_{cd}=l_{cd}/\xi$ implicitely via $\rsq(X_{cd},Y)\equiv
1010: \xi^2 X_{cd}^2/3$. The results for $X_{oc}$ presented in (b) are
1011: decay lengths extracted from fits of TCFs to simple
1012: exponentials. For the fits we used data from the segment length
1013: interval $X<Y,\frac12 N_{tot}/g$. However, for $Y<10$ the decay of the
1014: TCFs ceases to be well described by a simple exponential (see
1015: also Fig.~\protect\ref{fig:BACF}). Being strongly depend
1016: on the data range selected for the fits (data not shown, the
1017: values presented in (b) for $Y<10$ thus have to be taken with a
1018: grain of salt. We note that the
1019: results for datasets with different coupling constant scale
1020: quite well and that there is good qualitative agreement between
1021: the two methods. The results nicely follow the \OSFKK
1022: prediction for reduced screening lengths $Y>10$, but are
1023: strongly influenced by excluded volume effects for smaller
1024: values of $Y$. In particular, the extracted persistence lengths
1025: systematically {\em exceed} the \OSFKK estimate. With respect
1026: to finite chain length effects the first method turns out to be
1027: more robust than the second. }
1028: \end{center}
1029: \end{figure}
1030:
1031: Similar conclusions can be drawn from an analysis of the
1032: tangent-correlation function (Fig.~\ref{fig:BACF}).
1033: In fact, our discussion in section~\ref{sec:consequences} shows that
1034: Figs.~\ref{fig:LpCrossover} (b,d) and the log-log plots in
1035: Figs.~\ref{fig:BACF} (a,b) are directly comparable.
1036: Fig.~\ref{fig:BACF} (c) shows the same data in the semi-logarithmic
1037: representation commonly used to identify a simple exponential decay of
1038: the correlation function. Clearly, the \OSFKK scaling does not work
1039: perfectly up to the \OSFKK persistence length, but, at least qualitatively,
1040: we observe the expected slow-down of the decay of the correlations.
1041:
1042: The discrepancy between our conclusion, $l_e\sim \kappa^{-y}$ with
1043: $y=2$, and the results of previous numerical and experimental
1044: investigations, $y\ll 2$,
1045: can be traced back to the definition of the electrostatic persistence
1046: length. So far we have used an indirect method (scaling plots of
1047: segment radii and TCFs).
1048: More direct methods usually proceed by (i) defining an {\em apparant}
1049: electrostatic persistence length $l_{e,app}$,
1050: (ii) calculating $l_{e,app}$ for numerical or experimental data,
1051: (iii) plotting $l_{e,app}/\xi$ as a function of the reduced screening length
1052: $Y$ (Fig.~\ref{fig:FittedLp}), and (tentatively)
1053: (iv) extracting effective values for $y$.
1054: Suitable definitions for $l_{e,app}$ were recently reviewed by
1055: Ullner and Woodward~\cite{Ullner_mm_02}.
1056: Fig.~\ref{fig:FittedLp} locates our results
1057: for the ``orientational correlation length'' $l_{oc}$ and
1058: the ``crossover distance'' $l_{cd}$ in our map of conformation space.
1059: The first length is defined as the decay length of a simple exponential
1060: fitted to the TCF while the second tries to identify the crossover
1061: from the blob pole Eq.~(\ref{eq:r2_blobpole}) to an undulating
1062: blob chain Eqs.~(\ref{eq:r2_SAW_BJ}) or (\ref{eq:r2_SAW_KK}).
1063:
1064:
1065: Clearly, the extracted length scales can only be identified with
1066: the electrostatic persistence length $l_e$ as long as
1067: $l_e$ is well separated from other relevant length scales,
1068: i.e. for sufficiently large chain and screening lengths.
1069: Fig.~\ref{fig:FittedLp} shows that our results closely follow
1070: the $\mathrm{KK}_1$ line for $N=4096$ and $Y>10$.
1071: However, the violation of either of the two conditions leads
1072: to deviations. Relative to the \OSFKK prediction the
1073: extracted persistence lengths
1074: \begin{description}
1075: \item[increase] for small reduced screening lengths due to excluded
1076: volume effects and
1077: \item[decrease] in the opposite limit
1078: due to finite chain length effects which are particularly strong
1079: for the TCF and quantities related to it.
1080: \end{description}
1081: Depending on the definition of $l_{e,app}$ and the range of chain and
1082: screening lengths studied, the combination of these two effects can
1083: lead to the observation of effective exponent $l_{e,app}\sim
1084: \kappa^{-y}$ which are much smaller than $y=2$. However, since the
1085: weak $\kappa$ dependence of $l_{e,app}$ is an artifact of the
1086: definition of the quantity, there seems to be no contradition to the
1087: \OSFKK theory. Attempts along these
1088: lines~\cite{Micka_pre_96,Ullner_jcp_97,Ullner_mm_02} therefore risk
1089: to create more confusion than insight as long as $l_{e,app}$ is not
1090: defined within a theoretical framework which explicitly accounts for
1091: excluded volume effects~\cite{Reed_mm_91}.
1092:
1093:
1094:
1095: \section{Discussion}
1096:
1097: In this paper we have combined a scaling analysis
1098: of the conformational properties of intrinsically flexible
1099: polyelectrolytes with Debye-H\"uckel interactions with extensive
1100: Monte Carlo simulations of isolated chains. Our study was
1101: focused on the controversial case of polyelectrolytes beyond
1102: the OSF limit, i.e. to the case where the electrostatic
1103: screening length $\kappa^{-1}$ exceeds the bare persistence
1104: length of the polymers in the absence of electrostatic interactions.
1105:
1106: Our main result is the refutation of
1107: theories~\cite{BarratJoanny_epl_93,Ha_mm_95} predicting an
1108: electrostatic persistence length scaling as $\kappa^{-1}$. In
1109: contrast, we have observed no significant deviations from the scenario
1110: proposed by Khokhlov and
1111: Khachaturian~\cite{KhokhlovKhachaturian_pol_82} who combined the idea
1112: by de Gennes et al.~\cite{GPVB_jp_76} of a stretched chain of
1113: polyelectrolyte blobs with the Odijk-Skolnick-Fixman theory of the
1114: electrostatic persistence
1115: length~\cite{Odijk_jpspp_77,SkolnickFixman_mm_77} and the
1116: electrostatically excluded
1117: volume~\cite{Odijk_jpspp_78,FixmanSkolnick_mm_78} between chain
1118: segments.
1119: Our results suggest that it is indeed possible to understand
1120: DHWLC by considering a hierarchy of effects due to interactions
1121: between different classes of monomer pairs:
1122: \begin{description}
1123: \item[Stretching] due to the (effectively unscreened) Coulomb
1124: repulsion between neighboring monomers into a chain
1125: of blobs which has a finite
1126: \item[Bending rigidity] due to the screening of interactions
1127: between monomers with a distance larger than $g/(\kappa\xi)$
1128: along the chain. As a consequence, the blob chain remains straight up
1129: the electrostatic persistence length $l_e=\kappa^{-2}/\xi$.
1130: Beyond $l_e$ the chain behaves like a random walk, before
1131: \item[Swelling] due the electrostatically excluded volume between
1132: chain segments with a distance larger than $l_e$ becomes
1133: relevant beyond the Flory length $l_F=\kappa^{-4}/\xi^3$
1134: \end{description}
1135:
1136: An interesting side result of our work is the extension of
1137: the polyelectrolyte blob scaling to the case of strongly
1138: interacting, almost fully stretched chains for which the
1139: OSF theory is known to work~\cite{Odijk_jpspp_78}.
1140: This extension was done in the logic of the \OSFKK theory but is
1141: incompatible with the ansatz of BJ. Its success provides strong
1142: evidence for the irrelevance of
1143: longitudinal and transverse fluctuations within the blob
1144: chain~\cite{LiWitten_mm_95}
1145: and proves the KK idea almost by itself.
1146:
1147: Clearly, scaling arguments cannot do justice to the full complexity of
1148: the problem. Omitting all numerical prefactors, the ubiquituous
1149: logarithmic corrections, finite chain length effects and, in our
1150: opinion most importantly, a refined description of the crossovers
1151: between narrow neighboring regimes, they cannot hope (and should not
1152: be expected) to describe numerical or experimental data in detail.
1153: Quite obviously, these features call for a {\em quantitative}
1154: explanation. While our numerical results can serve as benchmarks for
1155: the development of theories, it is a sobering thought that the {\em
1156: simplest} model of a {\em single, isolated} polyelectrolyte chain is
1157: still unsolved. Compared to the much better understood neutral
1158: polymers, the \OSFKK theory represents the equivalent of the standard
1159: Flory argument for the excluded volume effect. Nevertheless, we
1160: believe to have shown that the \OSFKK theory provides the indispensible
1161: ``big picture'' needed for the design and analysis of experiments and
1162: computer simulations.
1163:
1164:
1165: \section{Acknowledgements}
1166: We gratefully acknowledge helpful discussions with B. D\"unweg,
1167: J.-F. Joanny, K. Kremer and H. Schiessel. RE is supported by
1168: an Emmy-Noether fellowship of the DFG.
1169:
1170: \bibliographystyle{unsrt}
1171: %present address: Argonne National Laboratory, Materials Science Division, Build. 212, 9700 S. Cass Avenue, Argonne, IL--60439, USA
1172: \bibliography{../PE}
1173:
1174: \end{document}
1175:
1176: If we compare our results to those presented in previous papers
1177: devoted to numerical studies of the same problem, it seems to us that
1178: the information content of the older works is at the same time lower
1179: and higher: Lower in the sense that the results obscure the essential
1180: feature of an electrostatic persistence length which does not vary
1181: with $\kappa^{-1}$ but rather with $\kappa^{-2}$; Higher in the sense
1182: that the imperfect scaling due to the proximity of additional
1183: crossover lines and finite chain lengths is an important feature of
1184: the model (and of experiments described by it) which call for a {\em
1185: quantitative} theoretical explanation.
1186:
1187:
1188: As a last point, we discuss the tangent-correlation function
1189: Eqs.~(\ref{eq:bacf}) to (\ref{eq:bacfKK}) which, at least in
1190: principle, provides the most direct evidence for the existence of an
1191: electrostatic persistence length. Measuring this quantity, Micka and
1192: Kremer (MK) concluded that the apparent electrostatic persistence
1193: length varies with $\kappa^{-y}$ where $0.2\le y\le0.8$ is a parameter
1194: dependent exponent. Fig.~( \ref{fig:BACF}) shows a semi-logarithmic
1195: scaling plot of the tangent-correlation function for data points
1196: corresponding to the $Y=const$-lines in
1197: Fig.~(\ref{fig:LpCrossoverKK}). The poor scaling seems to contradict
1198: our earlier conclusions and qualitatively supports the observations of
1199: MK. While the significant curvature of the curves sheds some doubts on
1200: the relevance of persistence lengths extracted from fits to a simple
1201: exponential decay,
1202:
1203:
1204: \begin{figure}[t]
1205: \begin{center}
1206: \epsfig{file=FittedLp.eps,width=8cm}
1207: \caption{ \label{fig:FittedLp}
1208: {\bf preliminary plot} The figure represents length scales
1209: extracted from fits of tangent correlation functions to a simple
1210: exponential decay for our longest chains of length $N=4096$ (see
1211: also Fig.~\protect\ref{fig:BACF}). The symbols indicate the
1212: apparent electrostatic persistence lengths $l_{e,app}$. To
1213: illustrate how the results are affected by excluded volume
1214: effects, the fits were carried out in two different ways: (i) By
1215: preselecting data points to be included into the fitting
1216: procedure from specific regions of conformation space
1217: ($1<Y<X<Y^2,\frac12 N_{tot}/g$ ($\circ$) and
1218: $X<\max(1,Y^2),\frac12 N_{tot}/g$ ($\times$)) and (ii) by
1219: selecting by eye a contour length interval (indicated by the
1220: solid lines) from the raw data ($\Box$). Since the TCFs are
1221: fitted to an incorrect model, the results have to be taken with
1222: a grain of salt and can only be discussed qualitatively. We
1223: note that the results for datasets with different coupling
1224: constant scale quite well, but that the persistence length can
1225: only be extracted with some reliability for $Y>10$. In these
1226: cases the results agree with the \OSFKK scaling prediction. For
1227: $Y<10$ excluded volume effects strongly affect the TCF. We note
1228: in particular, that the extracted persistence lengths
1229: systematically {\em exceed} the \OSFKK estimate. }
1230: \end{center}
1231: \end{figure}
1232: \begin{figure}[t]
1233: \begin{center}
1234: \epsfig{file=FittedLpEndEffects.eps,width=8cm,angle=-90}
1235: \caption{ \label{fig:FittedLpEndEffects}
1236: {\bf preliminary plot} Chain length dependence of the apparent
1237: electrostatic persistence length $l_{e,app}$. In general a
1238: reduction of the chain length leads to a faster decay of the TCF
1239: so that the fitted persistence lengths {\em shrink} with the
1240: chain length. For chain lengths approaching the screening
1241: length, the extracted persistence lengths become systematically
1242: {\em smaller} than the \OSFKK estimate in the critical region with
1243: $Y>10$, where the asymptotic behavior is no longer hidden by
1244: excluded volume effects (Fig.~\protect\ref{fig:FittedLp}). All
1245: results are for $q=1$ (i.e. the ``red system'') and were
1246: obtained by choosing the fit range by eye. }
1247: \end{center}
1248: \end{figure}
1249: