1: \documentstyle[aps,pra,multicol,psfig]{revtex}
2: \begin{document}
3: \draft
4:
5: \title{A simple model of DNA denaturation and mutually
6: avoiding walks statistics}
7: \author{Marco Baiesi$^{1}$,
8: Enrico Carlon$^{2}$,
9: Enzo Orlandini$^{1}$
10: and Attilio L. Stella$^{1}$}
11: \address{$^{1}$INFM - Dipartimento di Fisica, Universit\`a di Padova,
12: I-35131 Padova, Italy \\
13: $^{2}$Theoretische Physik, Universit\"at des Saarlandes,
14: D-66041 Saarbr\"ucken, Germany
15: }
16:
17: \date{\today}
18: \maketitle
19:
20: \begin{abstract}
21: Recently Garel, Monthus and Orland ({\it Europhys. Lett.} {\bf 55},
22: 132 (2001)) considered a model of DNA denaturation in which excluded
23: volume effects within each strand are neglected, while mutual avoidance
24: is included.
25: Using an approximate scheme they found a first order denaturation.
26: We show that a first order transition for this model follows from
27: exact results for the statistics of two mutually avoiding random walks, whose
28: reunion exponent is $c > 2$, both in two and three dimensions. Analytical
29: estimates of $c$ due to the interactions with other denaturated loops, as well
30: as numerical calculations, indicate that the transition is even
31: sharper than in models where excluded volume effects are fully incorporated.
32: The probability distribution of distances between homologous base pairs
33: decays as a power law at the transition.
34: \end{abstract}
35:
36: \pacs{PACS numbers: 87.14.Gg 05.70.Jk 05.70.Fh 87.15.Aa}
37:
38: \begin{multicols}{2} \narrowtext
39: \section{Introduction}
40: \label{sec:intro}
41:
42: Simple models of DNA thermal denaturation have attracted a lot of attention
43: for a long time \cite{Pola66,Fish66,Fish84,Peyr89}.
44: Although it was pointed out rather early that excluded volume effects
45: may be crucial in determining the order of the denaturation transition
46: \cite{Fish66}, it was only very recently that such effects were
47: taken into account both at analytical and numerical level
48: \cite{Caus00,Kafr00,Carl02,Baie02}.
49:
50: Even once accepted excluded volume effects as a key factor in determining the
51: nature of the transition, one can still be interested in understanding how the
52: relaxation of self-avoidance constraints influences the overall behavior of the
53: system. As this can be done in different ways, learning in detail
54: about the effects of different geometrical exclusion constraints can help in
55: conceiving better models or more efficient approximations.
56:
57: The main general aim of this paper is that of producing a contribution to
58: the debate on these general issues. In particular we analyze the model of
59: DNA denaturation recently introduced and studied by Garel, Monthus and
60: Orland (GMO) \cite{Gare01}. Among other things, we show that this model
61: is a rather ideal context in which a recently proposed representation
62: of DNA in terms of block copolymer networks \cite{Baie02} can be applied.
63:
64: The Hamiltonian for the GMO model is \cite{Gare01}:
65: \begin{eqnarray}
66: H &=& \frac{1}{2} \sum_{i=1,2} \int_0^N ds
67: \left( \frac{d {\vec r}_i (s)}{d s} \right)^2
68: + \int_0^N ds \ V({\vec r}_{12} (s)) \nonumber \\
69: &+& g \int_0^N ds \int_0^N ds' \ \delta ({\vec r}_1 (s) - {\vec r}_2 (s'))
70: \label{hamin}
71: \end{eqnarray}
72: where ${\vec r}_1 (s)$, ${\vec r}_2(s)$ describe the conformations of the two
73: strands as functions of the curvilinear coordinate $s$ and
74: $V({\vec r}_{12}(s))$ is the attractive interaction between homologous
75: base pairs (${\vec r}_{12}(s) \equiv {\vec r}_1 (s) - {\vec r}_2 (s)$).
76: The last term on the r.h.s. of Eq. (\ref{hamin}) is due to excluded
77: volume interactions, which act only between the two chains and not within
78: a single chain. Thus the self-avoidance constraint is relaxed and each
79: isolated strand follows a simple random walk (RW) statistics.
80:
81: The authors of Ref. \cite{Gare01} simplified further the Hamiltonian
82: (\ref{hamin}) by approximating the excluded volume term by a long range
83: interaction between the strands, which allows for an analytical solution of
84: the problem. Within such approximation, the denaturation transition turns
85: out to be of first order type \cite{Gare01},
86: as in models where excluded volume effects are fully incorporated
87: \cite{Caus00,Kafr00,Carl02,Baie02}.
88:
89: Our purpose here is to clarify further this point. We start from a description
90: {\it \`a la} Poland - Sheraga (PS) \cite{Pola66} of the GMO model and use
91: a series of exact results on reunion exponents of mutually avoiding walks,
92: which show indeed that the denaturation transition is first order. More
93: surprisingly, the transition turns out to be even sharper than that occurring
94: in models where excluded volume effects are fully taken into account. This
95: conclusion is corroborated by approximate analytical treatments of excluded
96: volume effects due to loop-loop interactions and by a series of numerical
97: results obtained for polymer lattice models in $d=2$ and $d=3$.
98:
99: \begin{figure}[b]
100: \centerline{\psfig{file=fig01.eps,height=1.4cm}}
101: \vskip 0.2truecm
102: \caption{In the Poland - Sheraga model the full partition function of the
103: DNA chain is factorized in terms of products of partition functions $S_k$
104: of bound segments of length $k$ and of partition functions $L_l$ of
105: denaturated loops of length $2 l$. Thick lines denote double stranded
106: segments.
107: }
108: \label{FIG01}
109: \end{figure}
110:
111: In the Poland - Sheraga (PS) model \cite{Pola66} the DNA chain is treated
112: approximately as a sequence of non-interacting bound double segments and
113: denaturated loops. Within this scheme the full partition function $Z_N$ for
114: a chain whose constituent strands have lengths $N$, is factorized in terms of
115: elementary partition functions for loops and segments (see Fig. \ref{FIG01}).
116: In the grand canonical ensemble, where a fugacity $z$ per lattice step is
117: assigned, the total partition function reads \cite{Fish84}:
118: \begin{equation}
119: Z(z) = \sum_N z^N Z_N = \frac{V_0(z)}{1 - L(z) S(z)}
120: \end{equation}
121: where $L(z)$ and $S(z)$ are the grand canonical partition functions for loops
122: and double segments, respectively. The form of the numerator $V_0(z)$ depends
123: on boundary conditions and is not interesting for our purposes. The crucial
124: quantity is the loop partition function, which in the most general form
125: reads \cite{Fish84}:
126: \begin{equation}
127: L(z) = \sum_l \frac{\left( \mu^2 z \right)^l}{(2 l)^c}
128: \label{loop}
129: \end{equation}
130: where $\mu$ is the (non-universal) effective connectivity constant, while
131: $c$ is a universal exponent. The order of the transition is related to the
132: value of $c$ \cite{Fish84}: At the transition temperature (which in the PS
133: model corresponds to $\mu^2 z \to 1$) loops are power-law distributed in size
134: with probability $P(l) \sim l^{-c}$, thus their average size $\langle l
135: \rangle = \sum_l l P(l)$ is finite for $c > 2$, implying a first order
136: denaturation, while it diverges for $1 < c \leq 2$ and the transition becomes
137: continuous.
138:
139: For self- and mutually avoiding strands $c$ was estimated analytically in
140: an extended PS scheme \cite{Kafr00}, where excluded volume effects between a
141: self-avoiding loop and the rest of the chain are approximately taken into
142: account. This exponent was also obtained directly by Monte Carlo simulations
143: \cite{Carl02,Baie02} for polymer lattice models. The agreement between the
144: analytical estimates and simulation results for $c$ is very good both in $d=2$
145: and $d=3$ \cite{Kafr00,Carl02,Baie02}.
146:
147: In the next Sections we will present analytical and numerical estimates of
148: the exponent $c$ for the GMO model. The former are based on the general
149: theory of copolymer networks, which is briefly reviewed in the Appendix.
150:
151: \begin{figure}[b]
152: \centerline{\psfig{file=fig02.eps,height=3.0cm}}
153: \vskip 0.2truecm
154: \caption{
155: Example of loop in the GMO model. The two strands indicated as dashed and
156: solid lines are allowed to overlap themselves, but are mutually avoiding.
157: Equation (\ref{cisolated}) gives the associated reunion exponent.
158: }
159: \label{FIG02}
160: \end{figure}
161:
162: \section{Mutually avoiding walks}
163: \label{sec:maw}
164:
165: A denaturated loop in the GMO model is
166: formed by two mutually avoiding random walks of equal length $l$, with common
167: origin and endpoint (see example of Fig. \ref{FIG02}).
168: The probability that two random walks, with common origin, meet for the
169: first time after $l$ steps decays asymptotically as $P(l) \sim l^{-c}$
170: with \cite{Dupl88,Dupl88b}:
171: \begin{equation}
172: c = d \nu - 2 \nu \eta_{\rm 11}^{\rm (G)}.
173: \label{cisolated}
174: \end{equation}
175: Here $d$ is the dimensionality of the embedding space, $\nu = 1/2$ for RWs
176: and $\eta_{\rm 11}^{\rm (G)}$ denotes the contribution of a vertex from
177: which the two walks originate (see Fig. \ref{FIG02}).
178: We follow the notation of Ref. \cite{vonF97} and indicate with
179: $\eta_{f_1 f_2}^{\rm (G)}$ the exponent associated to a vertex from
180: which $f_1$ solid and $f_2$ dashed lines depart, at the gaussian fixed
181: point, which is the relevant one for our purposes. At this fixed point
182: solid and dashed lines do not interact with themselves, but they avoid
183: the lines of the other group.
184:
185: Figure \ref{FIG02}, as well as Figs. \ref{FIG03}-\ref{FIG05} below,
186: are examples of so called {\it copolymer networks},
187: i.e. networks whose constituents are polymers of different species.
188: The relevance of these block copolymer networks for a description of
189: denaturating DNA has been already stressed in the context of models
190: with full excluded volume effects \cite{Baie02}.
191: A general network of this type may be made, for instance, by an arbitrary
192: mixture of random and self-avoiding walks (SAWs). Any pair of these walks may
193: either avoid, or be allowed to cross each other.
194: As for the homopolymer case, the general entropic exponent for a network with
195: arbitrary topology follows from those of the constituents vertex exponents
196: (see Refs. \cite{Dupl88,Dupl86} for details). The general rule to calculate
197: the entropic exponent of a network is simple and it is reviewed in the
198: Appendix.
199: Restricting to the case in which all constituent polymers have the
200: same $\nu$ exponent, one associates to each independent loop a factor
201: $d \nu$ and to each vertex a factor $-\nu \eta$, with $\eta$ the appropriate
202: exponent which depends on the number of outgoing legs, and on their type.
203: The generalization to an arbitrary mixture of polymer segments with distinct
204: $\nu$'s, as for instance RWs and SAWs, is also possible.
205:
206: Coming back to Eq. (\ref{cisolated}), the vertex exponent for
207: one solid and one dashed line is known exactly in $d=2$ \cite{Dupl99}.
208: Its value is $\eta_{\rm 11}^{\rm (G)} = -5/4$, which leads to $c=2 + 1/4 >2$.
209: In higher dimensions one has to resort to $\varepsilon$ expansion
210: renormalization group results. In $d=3$ resummation techniques yield
211: $\eta_{\rm 11}^{\rm (G)} \approx 0.57$ \cite{vonF97}, which implies
212: $c \approx 2.07$.
213: Thus, both in $d = 2$ and $3$ the reunion exponent is larger than the threshold
214: value ($c > 2$), which immediately implies a first order denaturation for the
215: GMO model, within the PS picture of non-interacting loops and bound segments.
216:
217: As a comparison we recall that for an isolated loop made by two self- and
218: mutually avoiding strands the entropic exponent is \cite{deGe79}:
219: \begin{equation}
220: c = d \nu.
221: \end{equation}
222: For a SAW one has $\nu = 3/4$ in $d=2$, which implies $c = 3/2$, while $\nu
223: \approx 0.588$ in $d=3$, implying $c \approx 1.76$. Thus, within the PS
224: picture, denaturation is a continuous transition ($c < 2$) when full self-
225: avoidance within a single loop is included \cite{Fish66,Fish84,nota}.
226:
227: It is also possible to go beyond the PS approximation in the GMO model,
228: but it is already quite clear at this point that excluded volume effects
229: between a single loop and the rest of the chain "localize" even more the
230: loops and $c$ can only increase.
231: This can be shown explicitly for some simple geometries in which the loop
232: is embedded. The calculations follow closely those of Ref. \cite{Kafr00}
233: for the model in which self - avoidance is fully included and are also
234: briefly reviewed in the Appendix. We give here only the final results.
235:
236: \begin{figure}[b]
237: \centerline{\psfig{file=fig03.eps,height=3.1cm}}
238: \vskip 0.2truecm
239: \caption{
240: A small loop (between points A and B) embedded between two much larger
241: loops is characterized by an exponent given by Eq. (\ref{cinteracting}).
242: }
243: \label{FIG03}
244: \end{figure}
245:
246: We consider the geometry shown in Fig. \ref{FIG03} in which a loop of
247: length $2l$ is connected to two loops of length $N-l$ each. In the limit $l
248: \ll N$ the partition functions of the inner loop is still of the type of
249: Eq. (\ref{loop}) and factorizes with the partition function of the rest of
250: the chain \cite{Kafr00}. As explained in Appendix, the exponent $c$ is however
251: modified:
252: \begin{equation}
253: c = d \nu - \nu \eta_{\rm 22}^{\rm (G)}.
254: \label{cinteracting}
255: \end{equation}
256:
257: \begin{figure}[b]
258: \centerline{\psfig{file=fig04.eps,height=2.7cm}}
259: \vskip 0.2truecm
260: \caption{In $d=2$ there are two distinct vertices formed by two dashed and
261: two solid lines (see text).
262: }
263: \label{FIG04}
264: \end{figure}
265:
266: In $d=2$ some care has to be taken in considering vertices, as the order of
267: walks does matter. This is illustrated in Fig. \ref{FIG04} where the two
268: possible distinct types of vertices with four outgoing lines, two solid and
269: two dashed, are shown.
270: For a vertex of type (a) more configurations are possible as neighboring
271: lines of the same type are allowed to overlap. In (b), instead, the
272: distinction between solid and dashed lines becomes irrelevant as
273: neighboring lines avoid each other.
274: In the latter case the lines form a vertex of four mutually avoiding walks
275: (MAWs) \cite{vonF02}. A vertex with $L$ outgoing MAWs, has an associated
276: entropic exponent \cite{Dupl88,Dupl99}:
277: \begin{equation}
278: \eta_L^{\rm (MAW)} = - \frac{4 L^2 - 1}{12}
279: \end{equation}
280: Hence, for loops embedded in a geometry as in Fig. \ref{FIG03}, and formed by
281: vertices of type (b) one finds $c=d \nu - \nu \eta_4^{\rm (MAW)}
282: = 3 + 5/8 = 3.625$.
283: Let us consider now vertices of type (a) for which we keep the notation
284: $\eta_{\rm 22}^{\rm (G)}$ to indicate the associated exponent. Using the
285: prescriptions of Ref. \cite{Dupl99} one has $\eta_{\rm 22}^{\rm (G)} = 35/12$
286: and thus from Eq. (\ref{cinteracting}) $c = 2 + 11/24 \approx 2.46$.
287: Notice that stronger excluded volume effects (as for vertices of type (b))
288: imply a higher value for $c$, with a quite large difference between the
289: two estimated values.
290: We point out that, obviously, $\eta_{\rm 11}^{\rm (G)} = \eta_2^{\rm (MAW)}$
291: in all dimensions.
292:
293: In $d=3$ the above subtleties do not occur.
294: The vertex exponent $\eta_{\rm 22}^{\rm (G)} \approx -1.81$ was estimated in
295: Ref. \cite{vonF97} using $\varepsilon$ expansions and resummation techniques.
296: Thus, from Eq. (\ref{cinteracting}), one obtains $c \approx 2.40$.
297:
298: We consider now another geometry, where a loop is connected to two long double
299: stranded segments, as illustrated in Fig. \ref{FIG05}.
300: Using the same method of Ref. \cite{Kafr00} we find (see Appendix):
301: \begin{equation}
302: c = d \nu - 2 \nu \overline{\eta}_{1,2}
303: \label{c2sigma3}
304: \end{equation}
305: where with $\overline{\eta}_{1,2}$ we indicate the exponent associated to a
306: vertex where one double, and two single stranded segments join.
307: Notice that the double stranded segment is actually a self avoiding walk
308: since, if it would overlap itself in some part, the mutual avoidance condition
309: between the two constituent strands would be violated.
310:
311: \begin{figure}[b]
312: \centerline{\psfig{file=fig05.eps,height=3.1cm}}
313: \vskip 0.2truecm
314: \caption{
315: A loop embedded between two long double stranded segments. The latter
316: follow the self-avoiding walks statistics.
317: }
318: \label{FIG05}
319: \end{figure}
320:
321: Again in $d=2$ one can use the techniques of Ref. \cite{Dupl99} to find
322: $\overline{\eta}_{1,2} = -2 - 7/16$ which leads to $c \approx 3.44$. To our
323: knowledge, there exists no field theoretical analysis for this type of
324: vertices in higher dimensions, thus we are not able to make any predictions
325: for $c$ in $d=3$, at present.
326:
327: Table \ref{TABLE01} summarizes the exponents found in the GMO
328: model for an isolated loop ($c^{(\rm IL)}$), for a loop interacting with
329: other loops ($c^{(\rm LL)}$), as in the geometry of Fig. \ref{FIG03}, and
330: for a loop interacting with double stranded segments ($c^{(\rm LS)}$), as in
331: Fig. \ref{FIG05}. For the two dimensional case we reported both values of
332: $c^{(\rm LL)}$ corresponding to the loops formed by the two types of vertices
333: of Fig. \ref{FIG04}.
334: We notice that, as anticipated above, the interaction with other parts of the
335: chain has the effect of increasing the loop exponent $c^{(\rm LL)},
336: c^{(\rm LS)} > c^{(\rm IL)}$ as is the case for all other models studied so
337: far \cite{Kafr00,Carl02}. The first order character of the transition in the
338: GMO model is therefore strengthened.
339:
340: It is interesting to compare the exponents obtained in this paper with the
341: corresponding ones for the model in which strands are fully self-avoiding
342: (FSA). The latter, taken from Ref. \cite{Kafr00}, are given in Table
343: \ref{TABLE01}. The comparison between the two models indicates that the
344: transition in the GMO model is generally sharper (higher $c$).
345: This is always true with the exception of the exponent $c^{\rm (LL)}$ in $d=2$
346: for vertices of the type of Fig. \ref{FIG04}(a) for which we find
347: $c^{\rm (LL)} \approx 2.46$, for the GMO model, while $c^{\rm (LL)} \approx
348: 2.69$ in the FSA case. This is due to the fact that in the GMO model for
349: the type of vertex considered excluded volume effects are not so pronounced
350: as walks have considerable freedom to overlap.
351:
352: %============================================================================
353: \vbox{
354: \begin{table}[tb]
355: \caption{
356: Summary of the exponents for the GMO model for an isolated loop (IL) and for
357: a loop interacting with neighboring loops (LL) or segments (LS). The results
358: for fully self-avoiding (FSA) strands (from Ref. \protect\cite{Kafr00}) are
359: also given. The last column shows the numerical estimates of $c$ (those for
360: the FSA are taken from Ref. \protect\cite{Baie02}).
361: }
362: \label{TABLE01}
363: \vskip 0.2truecm
364: \begin{tabular}{c|cccc}
365: & $c^{\rm (IL)}$ & $c^{\rm (LL)}$ & $c^{\rm (LS)}$ & c\\
366: \hline
367: GMO (d=2) & 2.25 & 2.46-3.62 & 3.44 & 2.95(5)\\
368: FSA (d=2) & 1.50 & 2.69 & 2.41 & 2.46(9) \\
369: GMO (d=3) & 2.07 & 2.40 & - & 2.55(5)\\
370: FSA (d=3) & 1.76 & 2.22 & 2.11 & 2.18(6)
371: \end{tabular}
372: \end{table}
373: }
374: %=========================================================================
375:
376: \section{Numerical Results}
377: \label{sec:num}
378:
379: We performed a series of
380: numerical calculations for a lattice version of the GMO model both in $d=3$
381: and $d=2$ (cubic and square lattices).
382: We considered two random walks of length $N$, described by the vectors
383: $\vec{r}_1 (i)$ and $\vec{r}_2 (i)$ ($i = 0, 1, 2 \ldots N$) which identify
384: the positions of the lattice monomers. The walks have common origin
385: ($\vec{r}_1 (0)= \vec{r}_2 (0)$), and no specific boundary condition is
386: imposed on the other ends.
387: While each strand can overlap itself, mutual
388: avoidance requires that $\vec{r}_1 (i) \neq \vec{r}_2 (j)$ for $i \neq j$.
389: Overlaps at homologous points are however allowed ($\vec{r}_1 (i) =
390: \vec{r}_2 (i)$); these correspond to a bound state of complementary base
391: pairs, to which we assign an energy $\varepsilon = -1$. At sufficiently low
392: temperature ($T$) the walks are fully bound and form an object
393: obeying self-avoiding walk statistics.
394:
395: We used the PERM \cite{Gras97} algorithm and computed $P(l)$ the probability
396: distribution function (pdf) to find a loop of length $2l$ in the chain.
397: While at low temperatures loops are exponentially distributed in size, at the
398: critical point the pdf has an algebraic decay $P(l) \sim l^{-c}$, from which
399: we can extract the exponent $c$ \cite{Carl02,Baie02}.
400: Figure \ref{FIG06} shows a log-log plot $P(l)$ vs. $l$ in $d=3$ at the
401: estimated critical point ($T_c =0.5181(3)$), for chains of lengths up to
402: $N = 1280$. A linear fit of the data yields $c=2.55(5)$, not far from the
403: $c^{\rm (LL)}$ given by Eq. (\ref{cinteracting}) (see Table \ref{TABLE01}).
404: The result confirms
405: that the transition in the GMO model is sharper than when self-avoidance is
406: fully included. Indeed, in the latter case the most accurate numerical estimate
407: of the loop exponent is $c = 2.18(6)$ \cite{Baie02}.
408:
409: \begin{figure}[b]
410: \centerline{\psfig{file=fig06.eps,height=6.25cm}}
411: \vskip 0.2truecm
412: \caption{Plot of $\log_{\rm 10} P$ vs $\log_{\rm 10} l$ at the estimated
413: critical point in $d=3$ and up to $N=1280$. A linear fit of the data yields
414: $c = 2.55(5)$.}
415: \label{FIG06}
416: \end{figure}
417:
418: We consider next $P' (r)$, the probability that the distance between homologous
419: points in the strands $|\vec{r}_1 (k) - \vec{r}_2 (k)|$ equals $r$. This
420: quantity was found to decay as a stretched exponential for $r \to \infty$
421: \cite{Gare01}, from an analysis of the Hamiltonian of Eq. (\ref{hamin}).
422:
423: This conclusion is at odd with the algebraic decay of the loop pdf found here
424: for the GMO model, and also in other similar models with a first order
425: transition \cite{Carl02}.
426: For the numerical calculation of $P'(r)$ we registered all distances between
427: homologous pairs along the chains $|\vec{r}_1 (k) - \vec{r}_2 (k)|$, starting
428: from $k=1$ and up to the highest $k$ for which the two strands are in contact
429: ($\vec{r}_1 (k) = \vec{r}_2 (k)$).
430: Figure \ref{FIG07} shows a plot of $\log_{\rm 10} P' (r)$ vs. $\log_{\rm 10} r$
431: at the transition temperature in $d = 3$. We cannot fit the data with a
432: stretched exponential decay, but by increasing the system size the data
433: approach a power law decay $P' (r) \sim r^{-k}$ with $k = 2.1(1)$.
434: The fact that $k > 1$ implies that the probability is normalizable.
435: This is a confirmation that
436: the transition is of first order type (see Ref. \cite{deGe79}).
437: Notice however, that the algebraic nature of $P'(r)$ implies
438: infinite moments $\langle r^n \rangle$ starting from $n=2$.
439: We conclude that the exponential decay found in Ref. \cite{Gare01} is an
440: artifact of the approximation introduced.
441:
442: We performed a series of calculations also in the two dimensional case.
443: At the estimated critical temperature ($T_c = 0.5811(7)$) we find, from the
444: decay of the probability distributions $P(l)$ and $P'(r)$, the estimates
445: $c = 2.95(5)$ and $k = 3.0(1)$, confirming the first order nature of the
446: transition. Again denaturation happens to be
447: sharper than in the case where self-avoidance is fully incorporated, for
448: which $c \approx 2.46$ \cite{Baie02}. In the PERM algorithm both types
449: of vertices of Fig. \ref{FIG04} are generated, thus $c$ cannot be directly
450: compared with any of the value obtained analytically in the previous section.
451:
452: \begin{figure}[b]
453: \centerline{\psfig{file=fig07.eps,height=6.25cm}}
454: \vskip 0.2truecm
455: \caption{
456: Plot of $\log_{\rm 10} P'$ vs. $\log_{\rm 10} r$. The sets are the same as in
457: Fig. \protect\ref{FIG06}. A linear fit yields $k =2.1(1)$.
458: }
459: \label{FIG07}
460: \end{figure}
461:
462: \section{Conclusion}
463: \label{sec:concl}
464:
465: In this paper we studied the model for DNA denaturation recently introduced
466: by Garel, Monthus and Orland \cite{Gare01}. Our analysis is complementary to
467: that of Ref. \cite{Gare01} and it is based on the calculation of the exponent
468: $c$ describing the algebraic decay of the probability of denaturated loop
469: lengths at the critical point. Using exact results from mutually avoiding walks
470: statistics we find that $c > 2$ already for isolated loops, implying that the
471: transition is of first order type.
472: This conclusion is in agreement with the results of Ref. \cite{Gare01}, but
473: it is not based on the approximate treatments \cite{Batt02}.
474:
475: For some specific simple geometries in which a loop is embedded in the chain
476: we provide analytical estimates of $c$, following ideas from the theory of
477: polymer networks \cite{Kafr00,Dupl86}. These calculations show explicitly
478: that $c$ increases with respect to the isolated loop value, i.e. the transition
479: robustly maintains its first order character.
480: More surprising is, at first sight, the fact that the denaturation in the GMO
481: model is sharper than that occurring in models where excluded volume
482: constraints are fully incorporated (higher $c$), as indicated by our analytical
483: and numerical estimates of $c$ both in $d=2$ and $3$. This behavior can be
484: explained by taking into account that the self-avoidance of the single strands
485: somehow contrasts the excluded volume effects between each loop and the rest
486: of the molecule. The result is that, when the single strand self-avoidance is
487: relaxed, the loop sizes are further reduced, consistently with an increase of
488: $c$.
489:
490: It is somehow remarkable how results of the theory of block copolymer networks,
491: which are relevant to issues like that of establishing a link between field
492: theory and multifractal measures (see e.g. Ref. \cite{vonF97}), allow to draw
493: reliable predictions on models of the denaturation transition of DNA.
494:
495: There is another possibility of relaxing the self-avoidance which is somehow
496: complementary to that of the GMO model. One can indeed neglect the mutual
497: avoidance between the strands, while keeping self-avoidance within each stand.
498: This possibility was studied in $d=3$ in Ref. \cite{Carl02} and the
499: denaturation transition turns out to be continuous ($c < 2$).
500: Thus, relaxing the self-avoidance constraint in two different ways
501: within the single loop, results
502: in a softening (as in the latter example) or a sharpening (as in the GMO
503: model) of the transition.
504:
505: Finally, we also considered the probability distribution of the distances
506: between homologous monomers of the two strands. We found that this quantity
507: decays as a power law at the transition point, in disagreement with the
508: stretched exponential behavior found in Ref. \cite{Gare01}, which is probably
509: an artifact of the approximation introduced there.
510: The algebraic decay of various quantities at denaturation, even in the case
511: of first order transitions, is a common feature of other similar models.
512: Finding the appropriate treatment of the Hamiltonian (\ref{hamin}), which
513: provides power-laws decay at the transition point is still an open question.
514:
515: \section*{acknowledgements}
516:
517: Useful discussions with David Mukamel are gratefully aknowledged.
518:
519: \begin{figure}[b]
520: \centerline{\psfig{file=fig08.eps,height=6.5cm}}
521: \vskip 0.2truecm
522: \caption{Examples of polymer networks with (a) ${\cal L}=2$, $n_1 = 5$,
523: $n_3 = 2$ and $n_4 = 2$, (b) ${\cal L}=1$, $n_1 = 2$, $n_3 = 2$ and
524: (c) ${\cal L}=3$, $n_4 = 2$.}
525: \label{FIG08}
526: \end{figure}
527:
528: \section*{Appendix}
529:
530: We review here some known results for entropic exponents of networks of
531: arbitrary topology. We focus mainly on networks whose constituent polymers
532: are all of the same type, and just briefly mention copolymer networks.
533:
534: Let us consider a network made of self-avoiding walks, all avoiding each
535: other, as the one shown in Fig. \ref{FIG08}(a). Let $N = \sum_i s_i$ be the
536: total length of the network, where each segment has length $s_i$.
537: In the asymptotic limit $N \to \infty$, $s_i \to \infty$, the total number
538: of configurations for the network scales as \cite{Dupl86}:
539: \begin{equation}
540: \Gamma \sim \mu^N N^{\gamma_G - 1} f\left( \frac{s_1}{N}, \frac{s_2}{N}
541: \ldots \right)
542: \label{partfunc}
543: \end{equation}
544: where the factor $\mu$ is the effective connectivity constant for the walks
545: (see e.g. Ref \cite{Vand98}) and $f$ is a scaling function.
546: The universal exponent $\gamma_G$ depends only on the number of independent
547: loops, ${\cal L}$, and the number of vertices with $k$ outgoing segments $n_k$
548: as \cite{Dupl86}
549: \begin{equation}
550: \gamma_G = 1 - {\cal L} d \nu + \sum_k \nu \eta_k n_k
551: \label{gammaG}
552: \end{equation}
553: where $\eta_k$ are exponents associated to a vertex with $k$ outgoing legs.
554: Such exponents are known exactly in $d=2$ from conformal invariance
555: \cite{Dupl86}, while they have been obtained from renormalization group
556: and resummation techniques in $d=3$ \cite{Scha92}.
557: We notice that it is also customary to use the definition $\sigma_k \equiv
558: \nu \eta_k$.
559:
560: This theory was recently applied to the denaturation problem \cite{Kafr00}.
561: In this case one considers network geometries as that of Fig. \ref{FIG08}(b),
562: in which a loop of total length $2l$ is connected to two polymers of lengths
563: $N/2$ each. Following the above notation one has ${\cal L}=1$, $n_1=2$ and
564: $n_3=2$ and thus the total number of configurations scale as:
565: \begin{equation}
566: \Gamma \sim \mu^N N^{\tilde \gamma - 1} f\left(\frac{l}{N} \right)
567: \label{eq11}
568: \end{equation}
569: with ${\tilde \gamma} = 1 -d \nu + 2 \nu ( \eta_1 + \eta_3 )$.
570: Now in the limit $l \ll N$ one should recover the partition function for a
571: self-avoiding walk of total length $N$ which has an entropic exponent
572: $\gamma = 1 + 2 \nu \eta_1$. This requires that:
573: \begin{equation}
574: f(x) \sim x^{2 \nu \eta_3 - d \nu}
575: \end{equation}
576: in the limit $x \to 0$. Thus, loops of length $l$ embedded within a long chain
577: of length $N$ ($ \gg l$) are distributed according to the probability $P(l)
578: \sim l^{-c}$, with \cite{Kafr00}:
579: \begin{equation}
580: c = d \nu - 2 \nu \eta_3.
581: \label{sigma3}
582: \end{equation}
583:
584: As second example we consider a loop of length $2l$ embedded between two long
585: loops of total length $2N$, as depicted in Fig. \ref{FIG08}(b). The total
586: number of configurations is still given by Eq. (\ref{eq11}) with ${\tilde
587: \gamma} = 1 - 3 d \nu + 2 \nu \eta_4$, as the network has now ${\cal L} = 3$
588: independent loops and two vertices with four outgoing legs, i.e. $n_4 = 2$.
589: In the limit $l \ll N$ one should find the partition function of two loops
590: joined at a single vertex.
591: Matching the scaling function in this limit yields
592: \begin{equation}
593: f(x) \sim x^{\nu \eta_4 - d \nu}
594: \end{equation}
595: for $x \to 0$. One finds thus that loops are distributed according to the
596: probability $P(l) \sim l^{-c}$, with \cite{Kafr00}:
597: \begin{equation}
598: c = d \nu - \nu \eta_4.
599: \label{sigma4}
600: \end{equation}
601:
602: Finally, this theory can be generalized to the case of networks with more
603: than one type of polymer, as for {\it copolymer} networks. Eq. (\ref{gammaG})
604: can be generalized easily in the case that the constituting polymers have all
605: the same metric exponents (for instance in the case of mutually avoiding
606: random walks). The only difference is that now vertex exponents will not
607: depend only on the number of outgoing legs, but also on the type of polymers
608: forming the vertex and, in two dimensions, also on the order.
609: For instance, for a network made of two types of polymers, which are random
610: walks and mutually avoiding with a geometry of Fig. \ref{FIG03} one arrives
611: to a $c$ exponent in the limit of an inner loop much shorter than the rest of
612: the chain as given by Eq. (\ref{cinteracting}), which is the copolymer network
613: counterpart of Eq. (\ref{sigma4}).
614: Notice that while for homopolymers $\eta_{\rm 2} = 0$, in general a vertex with
615: two outgoing polymers of different type will have a non-vanishing exponent (as
616: is the case of the $\eta_{\rm 11}^{\rm (G)}$ introduced in Fig. \ref{FIG02}).
617:
618: Finally if polymers constituting the inhomogeneous network will have different
619: $\nu$ exponent, as for networks formed by a mixture of random and self-avoiding
620: walks than Eq. (\ref{partfunc}) is still valid, but in a form involving the
621: radii of gyrations $R_i$ of the constituent polymers, where
622: $R_i \sim s_i^{\nu}$, for a segment of length $s_i$.
623:
624: \begin{references}
625: \bibitem{Pola66} D. Poland and H. A. Sheraga, J. Chem. Phys. {\bf 45},
626: 1456 (1966); {\bf 45}, 1464 (1966).
627: \bibitem{Fish66} M. E. Fisher, J. Chem. Phys. {\bf 45}, 1469 (1966).
628: \bibitem{Fish84} M. E. Fisher, J. Stat. Phys. {\bf 34}, 667 (1984).
629: \bibitem{Peyr89} M. Peyrard and A. R. Bishop, \prl {\bf 62}, 2755 (1989).
630: \bibitem{Caus00} M. S. Causo, B. Coluzzi, and P. Grassberger,
631: \pre {\bf 62}, 3958 (2000).
632: \bibitem{Kafr00} Y. Kafri, D. Mukamel, and L. Peliti, \prl {\bf 85}, 4988
633: (2000); Y. Kafri, D. Mukamel, and L. Peliti, Eur. Phys. J. B {\bf 27}, 132
634: (2002).
635: \bibitem{Carl02} E. Carlon, E. Orlandini, and A. L. Stella,
636: \prl {\bf 88}, 198101 (2002).
637: \bibitem{Baie02} M. Baiesi, E. Carlon, and A. L. Stella, \pre, in press
638: ({\tt cond-mat/0205125}).
639: \bibitem{Gare01} T. Garel, C. Monthus, and H. Orland, Europhys. Lett.
640: {\bf 55}, 132 (2001).
641: \bibitem{Dupl88} B. Duplantier and K.-H. Kwon, \prl {\bf 61}, 2514 (1988).
642: \bibitem{Dupl88b} B. Duplantier, Commun. Math. Phys. {\bf 117}, 279 (1988).
643: \bibitem{vonF97} C. von Ferber and Yu. Holovatch, Europhys. Lett. {\bf 39},
644: 31 (1997); C. von Ferber and Yu. Holovatch, \pre {\bf 56}, 6370 (1997).
645: \bibitem{Dupl86} B. Duplantier, \prl {\bf 57}, 941 (1986).
646: \bibitem{nota} Notice that denaturation becomes first order for $d > 4$,
647: since $\nu =1/2$ and thus $c = d \nu = d/2 > 2$.
648: \bibitem{deGe79} P. G. de Gennes, {\it Scaling concepts in Polymer Physics}
649: (Cornell University Press, Ithaca, NY, 1979).
650: \bibitem{vonF02} C. von Ferber and Yu. Holovatch, \pre {\bf 65}, 042801 (2002).
651: \bibitem{Dupl99} B. Duplantier, \prl {\bf 82}, 880 (1999).
652: \bibitem{Gras97} P. Grassberger, \pre {56}, 3682 (1997).
653: \bibitem{Batt02} The approximate treatment of the excluded volume
654: interaction as done in Ref. \cite{Gare01} has been the subject of recent
655: debate. See: S. M. Battacharjee, Europhys. Lett. {\bf 57}, 772 (2002);
656: T. Garel, C. Monthus, and H. Orland, Europhys. Lett. {\bf 57}, 774 (2002).
657: \bibitem{Vand98} C. Vanderzande, {\it Lattice Models of Polymers} (Cambridge
658: University Press, Cambridge, United Kingdom, 1998).
659: \bibitem{Scha92} L. Sch\"afer, C. von Ferber, U. Lehr, and B. Duplantier,
660: Nucl. Phys. {\bf B374}, 473 (1992).
661: \end{references}
662:
663: \end{multicols}
664:
665: \end{document}
666: