1: \documentstyle[aps,floats,epsf,color,axodraw]{revtex}
2: \addtolength{\topmargin}{2.0cm}
3:
4:
5:
6: \begin{document}
7: \baselineskip=14pt
8: \def\be{\begin{equation}}
9: \def\ee{\end{equation}}
10: \def\bea{\begin{eqnarray}}
11: \def\eea{\end{eqnarray}}
12: \def\E{{\rm e}}
13: \def\bearst{\begin{eqnarray*}}
14: \def\eearst{\end{eqnarray*}}
15: \def\peleven{\parbox{11cm}}
16: \def\peffec{\peight{\bearst\eearst}\hfill\peleven}
17: \def\pspace{\peight{\bearst\eearst}\hfill}
18: \def\ptwelve{\parbox{12cm}}
19: \def\peight{\parbox{8mm}}
20: %\twocolumn[\hsize\textwidth\columnwidth\hsize\csname@twocolumnfalse\endcsname]
21:
22:
23: \title
24: {Explicit Renormalization Group for D=2 random bond Ising model with
25: long-range correlated disorder}
26:
27: \author
28: {M. A. Rajabpour \footnote{e-mail: rajabpour@physics.sharif.edu},
29: R. Sepehrinia \footnote{e-mail: sepehrinia@physics.sharif.edu}}
30:
31: %\vskip 1cm
32:
33: \address
34: {\it {Department of Physics, Sharif University of Technology, Tehran
35: ,Iran 11365-9161\\}}
36: \maketitle
37: %\date{00/07/2000}
38:
39: \begin{abstract}\baselineskip=12pt
40: We investigate the explicit renormalization group for fermionic
41: field theoretic representation of two-dimensional random bond Ising
42: model with long-range correlated disorder. We show that a new fixed
43: point appears by introducing a long-range correlated disorder. Such
44: as the one has been observed in previous works for the bosonic
45: ($\varphi^4$) description. We have calculated the correlation length
46: exponent and the anomalous scaling dimension of fermionic fields at
47: this fixed point. Our results are in agreement with the extended
48: Harris criterion derived by Weinrib and Halperin.
49: \end{abstract}
50:
51: \section{introduction}
52:
53: Critical properties of systems with short-range and long-range
54: correlated randomness have been studied
55: extensively\cite{BY,KR,C,br,KLS,h,WH,Elka}. One important question
56: to address is whether the introduction of weak randomness changes
57: the universality class of transition. That is, in the
58: renormalization group (RG) language, whether the disorder is
59: relevant at the critical point of pure system or not. According to
60: the well known Harris criterion \cite{C,h}, disorder is irrelevant
61: if $d \nu>2$, where $d$ is the dimensionality and $\nu$ is the
62: correlation length exponent of pure system. This criterion should be
63: modified in the presence of long-range correlations in the disorder.
64: A special type of such a disorder has been considered by Weinrib and
65: Halperin. They showed that the disorder with power law correlation
66: $\sim x^{2\rho-d}$ ( for large separations $x$ ) is irrelevant if
67: $[(d-2\rho)\nu-2]>0$ for $\rho>0$, whereas the usual Harris
68: criterion recovers for $\rho<0$ \cite{WH}. Therefore the existence
69: of long-range correlations in the disorder would have significant
70: effect in the sense that they can change the universality class of
71: phase transitions.
72:
73: The most useful method to study disordered systems in RG language is
74: the Replica method. By employing this method, one can average out
75: the free energy using a trick based on the equation $\ln Z = \lim_{n
76: \rightarrow 0}\frac{Z^n-1}{n}$. The idea is then to average $Z^n$.
77: However, RG analysis can be implemented explicitly without averaging
78: on disorder\cite{M}. Using this method, one can avoid some of the
79: mathematical problems e.g. the $n \rightarrow 0$ limit in the
80: replica method. Moreover, application of the explicit method in some
81: cases, such as the one studied in this letter, is more
82: straightforward.
83:
84:
85: In this paper we consider random-bond Ising model with fermionic
86: action and long-range correlated disorder. The effect of short-range
87: correlated disorder has been studied in previous works
88: \cite{HL,GL,dd,L,DPP,M}. Also the $\varphi^{4}$ version of this
89: model is investigated through the double expansion near four
90: dimensions with short-range and long-rang disorder\cite{WH,Elka}. We
91: found a new long-range fixed point and we calculate both correlation
92: length exponent and scaling dimension of fermionic field at this
93: fixed point.
94:
95:
96:
97: \section{Explicit Renormalization Group}
98:
99: Two dimensional Ising model near its critical point can be described
100: by the massive free fermionic action
101: \begin{equation}
102: S=\frac{1}{2}\sum_{x}\bar{\psi}(x)(\hat{\partial}+m)\psi(x),
103: \end{equation}
104: $\psi=\tiny{\left(\begin{array}{c}\psi_1 \\ \psi_2
105: \end{array} \right)}$ is a two component Grassmannian field
106: $(\psi_i^{\dag}=\psi_i)$ and
107: \begin{equation}
108: \bar{\partial}=\sigma_3\partial_1+\sigma_1\partial_2, \hspace{0.5cm}
109: \bar{\psi}=\psi^{T}i\sigma_2,
110: \end{equation}
111: where $\sigma_i$ are Pauli matrices. The two point function of the
112: Grassmannian fields of the model in the momentum space is
113: \begin{equation}
114: G_0(p)(2\pi)^{2}\delta(p+q)=<{\psi}(p){\psi}(q)>
115: =\frac{-i\hat{p}+m}{p^{2}+m^{2}}(2\pi)^{2}\delta(p+q),
116: \end{equation}
117: where ${\psi}(p)$ is the Fourier transform of $\psi(x)$
118: \begin{equation}
119: {\psi}(p)=\sum_{x}\psi(x)exp(-ip\cdot x)
120: \end{equation}
121: and
122: \begin{equation}
123: \psi(x)=\int_{-\pi}^{\pi}\int_{-\pi}^{\pi}\frac{d^{2}p}{(2\pi)^{2}}{\psi}(p)exp(ip\cdot
124: x).
125: \end{equation}
126: The randomness can be inserted into the action in the following way
127: \begin{equation}
128: S=\frac{1}{2}\sum_{x}\bar{\psi}(x)(\hat{\partial}+m+c(x))\psi(x),
129: \end{equation}
130: where c(x) is a random variable with the following correlation
131: function in the momentum space
132: \begin{eqnarray}\label{correlation}
133: &<&c(p)>=0 \nonumber\\&<&c(p)c(q)>=(D_0+D_\rho|p|^
134: {-2\rho})\delta(p+q).
135: \end{eqnarray}
136:
137:
138: Here $D_0$ and $D_\rho$ are short-range and long-range disorder
139: strengths respectively. Positivity of two point function imposes
140: some restrictions on $D_{0}$ and $D_{\rho}$ for example $D_{\rho}$
141: should be positive for $\rho>0$ and $D_{0}$ should be positive for
142: $\rho<0$. Here we consider the case with $0<\rho<1$.\\
143:
144:
145: We can write the action in the momentum space as
146: \begin{eqnarray}
147: S=\frac{1}{2}\int\int_{-\pi}^{\pi}\frac{d^{2}p}{(2\pi)^{2}}\bar{\psi}(-p)
148: (i\hat{p}+m){\psi}(p)
149: +\frac{1}{2}\int\int\int\int_{-\pi}^{\pi}\frac{d^{2}p_{1}}{(2\pi)^{2}}
150: \frac{d^{2}p_2}{(2\pi)^2}\bar{\psi}(p_1)c(-p_{1}-p_{2})\psi(p_{2}),
151: \end{eqnarray}
152: where
153: \begin{eqnarray*}
154: c(p)=\sum_{x}c(x)exp(-ip.x).
155: \end{eqnarray*}
156:
157: It is convenient to introduce a diagrammatic representation. We have
158: the following vertex
159: \begin{center}
160: \unitlength 1mm
161: \begin{picture} (80,15)(0,0)
162: \linethickness {0.2mm}
163: \Photon(10,10)(35,10){1}{6}
164: \Line(35,10)(60,10)
165: \ZigZag(35,10)(35,25){1.3}{7}
166: \put(5,0){\small$p_1$}\put(17,0){\small$p_2$}
167: \put(8,10){\small{$p_1+p_2$}}
168: \put(23,3){\small$:=\bar{\psi}(p_1)c(-p_1-p_2)\psi(p_2)$.}
169: \end{picture}
170: \end{center}
171: Wavy line stands for {\small$\bar{\psi}$}, solid line for
172: {\small$\psi$} and zigzag represents the $c$ insertion. Momentum
173: conservation should be regarded at each vertex. So we have the
174: following graphs for vertex renormalization
175:
176:
177: \begin{center}
178: \unitlength 1mm
179: \begin{picture} (120,15)(0,0)
180: \linethickness {0.2mm}
181: \Photon(10,10)(30,10){1}{6}
182: \DashLine(30,10)(50,10){3}
183: \ZigZag(30,10)(30,25){1.3}{7}
184: \ZigZag(50,10)(50,25){1.3}{7}
185: \Line(50,10)(70,10)
186: \put(30,0){\Photon(10,10)(30,10){1}{6}
187: \DashLine(30,10)(50,10){3}
188: \ZigZag(30,10)(30,25){1.3}{7}
189: \ZigZag(50,10)(50,25){1.3}{7}
190: \DashLine(50,10)(70,10){3}\ZigZag(70,10)(70,25){1.3}{7}
191: \Line(70,10)(90,10)}
192: \put(70,0){\Photon(10,10)(30,10){1}{6}
193: \DashLine(30,10)(50,10){3}
194: \ZigZag(30,10)(30,25){1.3}{7}
195: \ZigZag(50,10)(50,25){1.3}{7}
196: \DashLine(50,10)(70,10){3}\ZigZag(70,10)(70,25){1.3}{7}
197: \DashLine(70,10)(90,10){3}\ZigZag(90,10)(90,25){1.3}{7}\Line(90,10)(110,10)}
198: \put(0,3){$2$}\put(27,3){$+6$} \put(66,3){$+24$}
199: \put(116,3){$+\cdots$}
200: %\put(8,10){\small{$p_1+p_2$}}
201: %\put(23,3){\small$:=\bar{\psi}(p_1)c(-p_1-p_2)\psi(p_2)$}
202: \end{picture}
203: \end{center}
204: where the dashed line represents {\small$\bar{\psi}\psi$}
205: propagator. Symmetry factors will be canceled with $n!$ of
206: perturbation expansion in each order. The key equation will be
207: \cite{M}
208: \begin{eqnarray}\label{key}
209: c'(r_1,r_2)=c(r_1+r_2)-\int_q
210: c(r_1-q)G_0(q)c(q+r_2)+\int_{q_1}\int_{q_2}
211: c(r_1-q_1)G_0(q_1)c(q_1-q_2)G_0(q_2)c(q_2+r_2)\nonumber \\ -
212: \int_{q_1}\int_{q_2}\int_{q_3}c(r_1-q_1)G_0(q_1)c(q_1-q_2)G_0(q_2)c(q_2-q_3)G_0(q_3)c(q_3+r_2).
213: \end{eqnarray}
214:
215: The random function in the original action has zero mean and we
216: should keep the mean value fixed after RG transformation. So the
217: mean value of new function should be extracted and in fact it will
218: renormalize the mass and the kinetic term. Then we need to introduce
219: a field renormalization constant $Z$ to keep the coefficient of
220: kinetic term equal to $\frac{1}{2}$, just as in the original action.
221: After rescaling momentums ($r\rightarrow\lambda r$) and fields
222: ($\psi\rightarrow\frac{\lambda^{3/2}}{\sqrt{Z}} \psi$), the
223: renormalized random function and mass will be
224: \begin{eqnarray}
225: c_r(p_1,p_2)=\frac{1}{\lambda
226: Z}[c'(p_1/\lambda,p_2/\lambda)-<c'(p_1/\lambda,p_2/\lambda)>].
227: \end{eqnarray}
228: \begin{eqnarray}
229: m_r=\frac{\lambda}{Z}(m + \langle c'\rangle_{r=0})
230: \end{eqnarray}
231:
232: A simple expansion of the second term of (\ref{key}) in powers of
233: $r$, with $m=0$, leads to the following expression for field
234: renormalization constant to first order in disorder strength
235:
236:
237: \begin{eqnarray}
238: Z = 1 + \frac{D_\rho}{4\pi^{1+2\rho}}(\lambda^{2\rho}-1) + O(D^2).
239: \end{eqnarray}
240: The short-range correlated disorder does not contribute up to this
241: order.
242:
243: So the renormalized mass is
244: \begin{eqnarray}
245: m_r=\lambda m\left\{1-\frac{1}{2\pi}(\ln
246: \lambda)D_0-\frac{1+\rho}{4\rho\pi^{1+2\rho}}(\lambda^{2\rho}-1)D_\rho+((\frac{\ln
247: \lambda}{2\pi})^{2}-I_{1})D_{0}^{2}
248: -(I_{2}+I_{3}+I_{5})D_{0}D_{\rho}-(I_{4}+I_{6})D_{\rho}^{2}+\cdots
249: \right\},
250: \end{eqnarray}
251: where the $I_{j}$'s are
252: \begin{eqnarray}
253: I_{1}&=&\int_{q_1}\int_{q_2}\frac{-q_{2}.(q_{2}-q_{1})+m^{2}}{(q_{1}^{2}+m^{2})(q_{2}^{2}+m^{2})((q_{2}-q_{1})^{2}+m^{2})}
254: \hspace{1cm}|q_{1}-q_{2}|\geq \frac{\pi}{\lambda},\nonumber\\
255: I_{2}&=&\int_{q_1}\int_{q_2}\frac{(-q_{1}.(q_{2}-q_{1})+m^{2})|q_{1}|^{-2\rho}}{(q_{1}^{2}+m^{2})(q_{2}^{2}+m^{2})((q_{2}-q_{1})^{2}+m^{2})} \hspace{1cm}|q_{1}-q_{2}|\geq \frac{\pi}{\lambda},\nonumber\\
256: I_{3}&=&\int_{q_1}\int_{q_2}\frac{(-q_{1}.q_{2}+m^{2})|q_{2}-q_{1}|^{-2\rho}}{(q_{1}^{2}+m^{2})(q_{2}^{2}+m^{2})((q_{2}-q_{1})^{2}+m^{2})} \hspace{1cm}|q_{1}-q_{2}|\geq \frac{\pi}{\lambda},\nonumber\\
257: I_{4}&=&\int_{q_1}\int_{q_2}\frac{(-q_{1}.(q_{2}+q_{1})+m^{2})|q_{1}|^{-2\rho}|q_{2}-q_{1}|^{-2\rho}}{(q_{1}^{2}+m^{2})(q_{2}^{2}+m^{2})((q_{2}-q_{1})^{2}+m^{2})}\hspace{7mm}|q_{1}-q_{2}|\geq \frac{\pi}{\lambda},\nonumber\\
258: I_{5}&=&\int_{q_1}\int_{q_2}\frac{(-q_{1}.(q_{1}+2q_{2})+m^{2})|q_{2}-q_{1}|^{-2\rho}}{(q_{1}^{2}+m^{2})^2(q_{2}^{2}+m^{2})},\nonumber\\
259: I_{6}&=&\int_{q_1}\int_{q_2}\frac{(-q_{1}.(q_{1}+2q_{2})+m^{2})|q_{2}-q_{1}|^{-2\rho}|q_{1}|^{-2\rho}}{(q_{1}^{2}+m^{2})^2(q_{2}^{2}+m^{2})}.\nonumber
260: \end{eqnarray}
261: we did not include the contribution of $Z$ in the second order.
262:
263: The renormalized values of disorder strengths can also be obtained
264: by calculating the correlations of renormalized random function
265: $c_r(p_1,p_2)$. Up to second order of bare strengthes, the
266: renormalized strengthes are
267: \begin{eqnarray}
268: D_{0r}&=&D_{0}(1-\frac{\ln \lambda}{2\pi}D_{0}-\frac{1+\rho}{2\rho
269: \pi^{1+2\rho}}(\lambda^{2\rho}-1)D_{\rho}+\cdots),\\
270: D_{\rho r}&=&\lambda^{2\rho}D_{\rho}(1-\frac{\ln
271: \lambda}{2\pi}D_{0}-\frac{1+\rho}{2\rho
272: \pi^{1+2\rho}}(\lambda^{2\rho}-1)D_{\rho}+\cdots).
273: \end{eqnarray}
274: By differentiating these equations with respect to $\ln \lambda$,
275: and then replacing the bare parameters in terms of renormalized
276: ones, one can obtain the following Wilson's functions
277: \begin{eqnarray}
278: \frac{dm_r}{d\ln \lambda}&=&m_{r}(1-\frac{1}{2\pi}D_{0r}-\frac{1+\rho}{2\pi^{1+2\rho}}D_{\rho r}+\cdot\cdot\cdot),\\
279: \frac{dD_{0r}}{d\ln
280: \lambda}&=&-D_{0r}(\frac{1}{\pi}D_{0r}+\frac{1+\rho}{\pi^{1+2\rho}}D_{\rho
281: r}+\cdot\cdot\cdot),\\ \frac{dD_{\rho r}}{d\ln \lambda}&=& D_{\rho
282: r}(2\rho-\frac{1}{\pi}D_{0r}-\frac{1+\rho}{\pi^{1+2\rho}}D_{\rho
283: r}+\cdot\cdot\cdot).
284: \end{eqnarray}
285:
286: It is clear from the equations that at the Gaussian fixed point
287: ($D_{0}^{*}=D^{*}_{\rho}=0$), $D_{0}$ is marginally irrelevant and
288: $D_{\rho}$ is relevant. So, by introduction of small amount of
289: long-range correlated disorder it becomes unstable in $D_{\rho}$
290: direction. Apart from the trivial Gaussian fixed point, we see that
291: there is a nontrivial fixed point at $D_{0}^{*}=0$ and
292: ${D^{*}_{\rho}}=\frac{2\rho}{1+\rho}\pi^{1+2\rho}$. The new fixed
293: point is attractive in all directions in the $D_{0}$ and $D_{\rho}$
294: plane. The RG flows starting in the vicinity of Gaussian fixed
295: point, end up at the nontrivial fixed point. One of the most
296: important critical exponents is the correlation length exponent,
297: which turns out to be different at two fixed points, $\nu=1$ at the
298: Gaussian fixed point and $\nu=\frac{1}{1-\rho}$ at the nontrivial
299: fixed point. The result is in agreement with \cite{WH}. A good
300: numerical confirmation of this relation for the special choice of
301: $\rho=\frac{1}{2}$ can be found in \cite{Igloi}.
302:
303: Also the field renormalization would acquire an anomalous dimension
304: which is defined through the asymptotic behavior of vertex function
305: in the long wavelength limit as
306:
307: \begin{equation}
308: \Gamma^{(2)}(p)\sim p^{1-\eta}
309: \end{equation}
310: so at the nontrivial fixed point we have
311: \begin{equation}
312: \eta=\frac{dZ}{d\ln\lambda}=\frac{\rho}{2\pi^{1+2\rho}}D^*_{\rho
313: r}=\frac{\rho^2}{1+\rho}
314: \end{equation}
315:
316:
317: Now we want to show that, some less singular terms in the
318: correlation functions of $c_{r}$, which were omitted in the equation
319: (\ref{correlation}), are irrelevant. We start with a general form of
320: the correlation function of disorder
321: \begin{eqnarray}
322: &<&c(p)>=0,
323: \nonumber\\&<&c(p)c(q)>=(D_0+\sum_{i=1}^{n}D_{\rho_{i}}|p|^
324: {-2\rho_{i}})\delta(p+q),
325: \end{eqnarray}
326: and then find the Wilson's functions for the $D_0$ and
327: $D_{\rho_{i}}$
328: \begin{eqnarray}
329: \frac{dD_{0r}}{d\ln\lambda}&=&-D_{0r}(\frac{1}{\pi}D_{0r}+\sum_{i=1}^{n}\frac{1+\rho_i}{\pi^{1+2\rho_{i}}}D_{\rho_{i}
330: r}+\cdot\cdot\cdot),\\
331: \frac{dD_{\rho_{i} r}}{d\ln \lambda}&=& D_{\rho_{i}
332: r}(2\rho_i-\frac{1}{\pi}D_{0r}-\sum_{j=1}^{n}\frac{1+\rho_j}{\pi^{1+2\rho_{j}}}D_{\rho_{j}
333: r}+\cdot\cdot\cdot).
334: \end{eqnarray}
335:
336: Here we have $n$ nontrivial fixed points at $D_{0}^{*}=0$ and all
337: $D^{*}_{\rho_{i}}=0$ except one of them, say ${D^{*}_{\rho_{l} }}$
338: which is $\frac{2\rho_{l}}{1+\rho_l}\pi^{1+2\rho_{l}}$. At these
339: fixed points the RG eigenvalues are $-2\rho_{l}$ and
340: $2(\rho_{i}-\rho_{l})$ which means that there is just one stable
341: fixed point (with $\rho_{l}>\rho_{i}$ for all $i$). Other fixed
342: points are unstable at least in one direction.
343:
344:
345:
346: Finally we want to compare the above results with the replica
347: results. In the case of short-range disorder, as pointed out by
348: Murthy\cite{M}, the results of the replica method and the explicit
349: method are in complete agreement. The replica action for the model
350: considered here can be easily obtained,
351: \begin{eqnarray}
352: S=\frac{1}{2}\sum_{x,\alpha}\bar{\psi}(x)(\hat{\partial}+m)\psi(x)-\frac{D_{0}}{8}
353: \sum_{x,\alpha,\beta}\bar{\psi}_{\alpha}(x)\psi_{\alpha}(x)\bar{\psi}_{\beta}(x)\psi_{\beta}(x)-\nonumber\\
354: \frac{D_{\rho}}{8}\sum_{x,y,\alpha,\beta}\bar{\psi}_{\alpha}(x)\psi_{\alpha}(x)\frac{1}{|x-y|^{2-2\rho}}\bar{\psi}_{\beta}(y)\psi_{\beta}(y).
355: \end{eqnarray}
356:
357: The third term in this action is nonlocal, and the computation of
358: the $\beta$ functions are not easily tractable. This shows the
359: advantage of explicit calculations with which one can avoid such a
360: nonlocal action. The other advantage of this method is that there is
361: no restriction on the distribution function of disorder while it
362: should have a Gaussian distribution in the replica method.
363:
364: \section{Acknowledgment}
365: We thank Professor G. Murthy for helpful discussions and Professor
366: J. Cardy for useful comments. R. S. would like to thank E. Khatami
367: and A. A. Saberi for critical reading of manuscript.
368:
369: \begin{thebibliography}{99}
370: \bibitem{BY} K. Binder and A.P. Young, Rev. Mod. Phys. 58, 801 (1986)
371: “\textit{Spin-glasses - experimental facts, theoretical concepts,
372: and open questions}”
373: \bibitem{KR} N. Kawashima, and H. Rieger, \textit{in Frustrated Spin Systems edited by H.T.
374: Diep, World Scientific, “Recent Progress in Spin Glasses}”
375: (Singapore, 2004)
376: \bibitem{C} J.Cardy \textit{Scaling and Renormalization in Statistical Physics} (Cambridge Univ.Press 1996)
377: \bibitem{br} K. Binder, J. D. Regir, Adv.Phys. 41, 547(1992)
378: \bibitem{KLS}A. L. Korzhenevskii, A. A. Luzhkov and W. Schirmacher, Phys. Rev. B50, 3661 (1994).
379: \bibitem{h} A. B. Harris, J. Phys. C7, 1671 (1974).
380: \bibitem{WH} A. Weinrib and B. I. Halperin, Phys. Rev. B27, 413 (1983).
381: \bibitem{Elka} Elka Korutcheva and F. Javier de la Rubia, Phys. Rev.
382: B 58, 5153-5156(1998).
383: \bibitem{HL} Harris and T.C.Lubensky, Phys.Rev.Lett. 33, 1540(1974)
384: \bibitem{GL} G. Grinstein and A. Luther, Phys. Rev. B13, 1329 (1976).
385: \bibitem{dd}V.S.Dotsenko, Vl.S.Dotsenko, Adv. Phys. 32, 129(1983)
386:
387: \bibitem{L} A. W. W. Ludwig, Nucl. Phys. B330, 639 (1990).
388: \bibitem{DPP} V. Dotsenko, M. Picco, P. Pujol, Nucl.Phys. B455 (1995) 701-723
389: \bibitem{M} G. Murthy, Phys. Rev. B36, 766 (1987)
390: \bibitem{extended} V.V. Prudnikov, P.V. Prudnikov and A.A. Fedorenko. J.
391: Phys. A 32, L399 (1999); V.V. Prudnikov, P.V. Prudnikov, and A.A.
392: Fedorenko, J. Phys. A 32, 8587 (1999),V. Blavats’ka, Ch. von Ferber,
393: and Yu. Holovatch, J. Mol. Liq. 92, 77 (2001); Phys. Rev. E 64,
394: 041102 (2001).
395: \bibitem{Igloi} F. A. Bagamery, L. Turban, and F. Igloi,
396: Phys. Rev. B 72, 094202 (2005).
397: \end{thebibliography}
398:
399:
400:
401: \end{document}
402: