1: \documentclass[twocolumn, showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
2:
3: %\usepackage{dcolumn}% Align table columns on decimal point
4: \usepackage{bm}% bold math
5: \makeatletter
6: \newenvironment{figurehere}
7: {\def\@captype{figure}}
8: {}
9: \makeatother
10: \usepackage{graphicx}
11: %\usepackage{dcolumn}
12: \usepackage{bm}
13: \usepackage{epsfig}
14:
15: \begin{document}
16:
17: \title{Topological Disorder in Spin Models on Hierarchical Lattices.}
18:
19: \author{N. S. Ananikian, K. G.~Sargsyan\\[1mm]
20: {\small \textsl{Department of Theoretical Physics, Yerevan Physics
21: Institute,}}\\{\small Alikhanian Br. 2, 375036, Yerevan, Armenia
22: }\\[1mm]}
23:
24: \begin{abstract}
25: A general approach for the description of spin systems on
26: hierarchial lattices with coordination number $q$ as a dynamical
27: variable is proposed. The ferromagnetic Ising model on the Bethe
28: lattice was studied as a simple example demonstrating our method.
29: The annealed and partly annealed versions of disorder concerned with
30: the lattice coordination number are invented and discussed.
31: Recurrent relations are obtained for the evaluation of
32: magnetization. The magnetization is calculated for the particular
33: disorder choices, $q=2,3$ and $q=3,4$. A nontrivial localization of
34: critical point is revealed.
35: \end{abstract}
36: \pacs{05.50.+q, 05.45.-a, 89.65.-s.} \maketitle
37: \section{Introduction}
38: The hierarchical models in statistical physics can be used to
39: exactly describe the critical phenomena. For these purposes only
40: approximate or numerical methods exist in most statistical
41: mechanical models. Therefore complicated models representing
42: nontrivial critical behavior and demonstrating analytical solution
43: may be recognized as very important. A good example of such systems
44: are spin models on hierarchical lattices. Besides the general
45: interest related to the critical phenomena, spin models on
46: hierarchical lattices are also formulated for description of the
47: concrete physical systems. The interesting applications are Bethe
48: approximation in polymer physics, and many others
49: \cite{Pretti}-\cite{esk}.
50:
51: The main purpose of this work is a study of spin models on hierarchical lattices
52: with so called topological disorder. The problem of disorder for the
53: hierarchical models is known and has been discussed in series of
54: works \cite{dis1}-\cite{dis4}. However, disorders concerned with
55: couplings and magnetic fields are considered mainly. So called
56: topological disorder, describing by the changing lattice
57: coordination number, is investigated less. In our definition
58: topological disorder represents non-homogeneity of the lattice
59: structure and may be even non-connected to the spins dynamics. A
60: more precise definition of the topological disorder in context of
61: the hierarchical models is given below within a concrete example.
62:
63: Although all results in our article are related to the physics and
64: have direct meaning in a physical sense, it is interesting to
65: discuss their possible applications in the context of social
66: dynamics. As is known, in the recent decades there is a rapidly
67: growing interest to the application of the statistical physics to
68: interdisciplinary fields (as biology, computer science, social
69: dynamics and e.t.c). Considering the social dynamics we deal not
70: with physical particles but with humans or crowds. The human
71: behavior is not rather well understood and depends on many hidden
72: factors. Nevertheless, one may suppose that in the statistical limit
73: the new statistical laws would be obtained, which are less dependent
74: on the individuals. As it was shown in recent decade, this
75: conjecture is correct. The transitions from disorder to order are
76: discovered describing the social dynamics from the statistical
77: physics point of view. We may bring as the examples of those
78: transitions spontaneous formation of a common language, culture or
79: the emergence of consensus about a specific issue \cite{soc}. There
80: are also examples of scaling and universality.
81:
82: As an illustration of our approach to spin systems on hierarchical
83: lattices with changing coordination number $q$ we consider
84: ferromagnetic Ising model on the Bethe lattice as a simplest
85: possible example.
86:
87: The article is organized as follows. In section \ref{second} a
88: simple ferromagnetic Ising model with topological disorder on
89: hierarchical lattice is formulated. A more precise and clear
90: definition of the topological disorder in context of the
91: hierarchical lattices is given. The connection between proposed
92: model and description of social dynamics is discussed in detail.
93:
94: In section \ref{3} annealed and partly annealed topological
95: disorders are defined. A general method for those mathematical
96: description is proposed. The annealed version of the topological
97: disorder is studied for the particular realizations of the lattice
98: structure. The critical temperature is found and analyzed.
99:
100: Partly annealed and quenched topological disorders are described
101: within ferromagnetic Ising model in section \ref{4}. The same
102: particular realizations of the topological disorder as for the
103: annealed one are discussed. The critical point is obtained.
104:
105: Main observations are presented in conclusion.
106: \section{Ising Model on Bethe Lattice with topological
107: disorder}\label{second}
108: The one of the simplest recursive lattices is a Cayley tree. Constructing Cayley
109: tree one starts with a central site \emph{O}. After that we connect
110: the central site \emph{O} by links with $q$ new other sites. This
111: procedure of $q$ new sites connection by links to the each site
112: originated on the previous step repeats $n$ times. Following this
113: procedure one gets the recursive connected graph, containing no
114: cycles and called the Cayley tree with coordination number $q$ and
115: $n$ generations.
116:
117: The Bethe lattice may be considered as the interior of the Cayley
118: tree. Only sites lying deep inside Cayley tree constitute the Bethe
119: lattice. Dealing with the Bethe lattice we suppose the respective
120: Cayley tree is large enough to achieve thermodynamical limit and
121: neglect boundary effects. The general structure and topology of the
122: Bethe lattice is the same as for Calyley tree. Here we consider
123: changing along the lattice coordination number $q_{i}$, which is
124: different in general for the different $i$ sites, even from the same
125: generation. The Betthe lattice is not uniform now and topological
126: disorder related to it can be discussed.
127:
128: The Hamiltonian of the Ising model on Bethe lattice reads
129: \begin{equation}
130: H=-J\sum_{i,j}s_{i}s_{j}-h\sum_{i}s_{i},\label{hamilton}
131: \end{equation}
132: where $J$ and $h$ do not depend on the indexes $i,j$, for the sake
133: of simplicity. The summation is over all neighbor $s_{i}$ and
134: $s_{j}$ spins along the lattice and spin variables are $s_{i}=\pm1$.
135: The interpretation of this model in context of social dynamics is
136: following. The common task of social dynamics is the understanding
137: of the transition from an initial disordered state to a
138: configuration that displays order. The one approach is an analogy to
139: physics of the Ising model for ferromagnetic. The Ising model may be
140: considered as the simplest description of the opinion dynamics. Each
141: spin on the Bethe lattice represents a human opinion on some subject
142: (agree/disagree units)locating in corresponding lattice site. We
143: describe the situation in the simplest sense. The coupling $J$ is
144: positive reflecting the fact that the people communicating with each
145: other long or even not a very long time usually tend to have the
146: same opinion. In other words, humans or other social units like
147: communicate with people (social units) having usually a good
148: agreement with them. However, it is not always true. The discussion
149: of the anti-ferromagnetic $J$ is also meaningful. The Bethe lattice
150: in our interpretation is not a spatial distribution of humans but a
151: construction in relation space and represents possible "stable"
152: communications (relations) between "humans" $s_{i}$. Our proposition
153: is to consider "humans" $s_{i}$ communicating with (acting on) some
154: finite number of neighbors in relation space. This number, which
155: coincides with the coordination number of lattice $q_{i}$, may be
156: influenced by some external noise and changes in time in general.
157: Here we consider $q_{i}$ individually for the each site $i$ and the
158: coordination number is not the same for the all sites of the same
159: step of the lattice generation.
160:
161: An external magnetic field may be treated as an external
162: informational factor, like TV or something similar. As we use
163: methods of the statistical physics, the temperature is introduced
164: and represents the noise in human opinion ( measure of uncertainty
165: in taking stable decision ). The main weakness of the model is an
166: absence of the small cycles (loops) in relation space. Some effects
167: induced by relation loops may be demonstrated on more complicated
168: hierarchical lattices (as Husimi lattice).
169: \section{Annealed topological disorder}\label{3}
170: First of all let us recall some general aspects related to the spin systems with disorder.
171: Let the Hamiltonian may be written as $H[J,s]$, where $J$ represents
172: disorder degree of freedom and $s$ describes the spin variables
173: correspondingly. Suppose that the spin and disorder degrees of
174: freedom are not thermally equilibrated and disorder degrees of
175: freedom have their own temperature $\overline{T}$. The partition
176: function reads:
177: \begin{eqnarray}
178: &\textbf{Z}&=\int DJ P(J) \exp(-\overline{\beta}F(J))\nonumber
179: \\&=&\int DJ P(J)\exp(\frac{\overline{\beta}}{\beta}\ln
180: Z(J))\nonumber \\&=&\int DJ P(J)(Z(J))^{n},\label{part}
181: \end{eqnarray}
182: where $Z(J)$ and $F(J)$ are respectively partition function and free
183: energy with given disorder $J$. Here we propose to describe the
184: disorder by the probability distribution $P[J]$. We may consider the
185: number $n=\frac{\overline{T}}{T}$ as a number of replicas within the
186: replica formalism \cite{Dotsenko}. The total free energy would be
187: \begin{equation}
188: \textbf{F}=-\frac{T}{n}\ln(<Z(J)^{n}>_{dis}), \label{total}
189: \end{equation}
190: where the average over. Using $\textbf{F}$ one would obtain the free
191: energy describing quenched disorder after taking the limit
192: $n\rightarrow0$. The $n=1$ value corresponds to annealed disorder,
193: when both temperatures $T$, $\overline{T}$ are equal. The value
194: $1>n>0$ may be treated here as a "degree of quenchness". The same
195: heuristics arises from the non-equilibrium two temperature system
196: viewpoint in \cite{Allahv}.
197:
198: Now let us consider topological disorder given by the set of
199: $\{q_{i}\}$ variables. We describe only the case of slow changing
200: $q_{i}$ compared with the rate of the spin flips. This assumption is
201: in agreement with the ferromagnetic hypothesis of social opinion
202: dynamics. It is natural to suppose that the stable relations
203: dispose to agreement between neighbors to a greater extent. So we do
204: not analyze a more fast rate of $q_{i}$ than it is for the annealed
205: disorder. Also we propose the ability of the probability
206: factorization $P(\{ q_{0},...,q_{M}\})=P(q_{0})...P(q_{M})$.
207:
208: The one of the best methods developed for the models on hierarchical
209: lattices is a so-called dynamical approach
210: \cite{din0}-\cite{Monroe}. Here we propose to modify this formalism
211: for our purposes. The main quantity to be calculated is
212: $<Z(\{q_{i}\})^{n}>_{dis}$. Let us start from the description of the
213: annealed topological disorder $n=1$. The main advantage of the
214: dynamical approach is that for the models formulated on recursive
215: trees the exact recursion relations can be derived. When the tree is
216: cut apart at the central site $O$, it separates into $q_{0}$
217: identical branches. According to it for the Hamiltonian
218: (\ref{hamilton}) one can write
219: \begin{eqnarray}
220: \textbf{Z}=<Z(\{q_{i}\})>_{dis}=\sum_{q_{0}}P(q_{0})\sum_{s_{0}}\exp(\beta
221: h s_{0})\times \nonumber
222: \\<g_{k}^{(1)}(q_{0},s_{0})...g_{k}^{(q_{0})}(q_{0},s_{0})>_{dis},\label{dyn}
223: \end{eqnarray}
224: where independent replicas $g_{k}^{(i)}(q_{0},s_{0})$ are invented.
225: The Eq. (\ref{dyn}) also may be rewritten as
226: \begin{eqnarray}
227: <Z(\{q_{i}\})>_{dis}=\sum_{q_{0}}P(q_{0})\sum_{s_{0}}\exp(\beta
228: h s_{0})\times \nonumber
229: \\<g_{k}(s_{0})>_{dis}^{q_{0}},\label{dyn1}
230: \end{eqnarray}
231: where
232: \begin{eqnarray}
233: <g_{k}(s_{0})>_{dis}=\sum_{q_{0}}P(q_{0})\sum_{s_{1}}\exp(\beta J
234: s_{0}s_{1}+\beta h s_{1})\times \nonumber
235: \\<g_{k-1}(s_{1})>^{q_{1}-1}_{dis}.
236: \end{eqnarray}
237: After denoting $x_{k}=<g_{k}(+)>$ and $y_{k}=<g_{k}(-)>$ we obtain
238: nonlinear two dimensional mapping
239: \begin{eqnarray}
240: x_{k}=\sum_{q} P(q)\left[e^{\beta J+\beta h}x^{q-1}_{k-1}+e^{-\beta
241: J-\beta h}y^{q-1}_{k-1}\right]\nonumber\\
242: y_{k}=\sum_{q} P(q)\left[e^{-\beta J+\beta h}x^{q-1}_{k-1}+e^{\beta
243: J-\beta h}y^{q-1}_{k-1}\right],\label{mapp}
244: \end{eqnarray}
245: where $x_{k}$ and $y_{k}$ must be positive.
246:
247: As a possible order parameter one may consider magnetization
248: averaged over the disorder. This quantity represents averaged
249: opinion in terms of agree/desagree units. Using Eq. (\ref{part}) we
250: get magnetization of the central cite \emph{O} averaged over
251: topological disorder as
252: \begin{equation}
253: m_{dis}=\frac{\sum_{q_{0}}P(q_{0})\sum_{s_{0}}s_{0}e^{\beta h
254: s_{0}}<g_{k}(s_{0})>_{dis}^{q_{0}}}{\sum_{q_{0}}P(q_{0})\sum_{s_{0}}e^{\beta
255: h s_{0}}<g_{k}(s_{0})>_{dis}^{q_{0}}}.\label{magn}
256: \end{equation}
257: Getting fixed points $(\overline{x},\overline{y})$ of mapping
258: (\ref{mapp}) and following to Eq. (\ref{magn}) one would obtain the
259: order parameter $m_{dis}$
260: \begin{equation}
261: m_{dis}=\frac{\sum_{q_{0}}P(q_{0})\left[\exp(\beta h
262: )\overline{x}^{q_{0}}-\exp(-\beta h
263: )\overline{y}^{q_{0}}\right]}{\sum_{q_{0}}P(q_{0})\left[\exp(\beta h
264: )\overline{x}^{q_{0}}+\exp(-\beta h
265: )\overline{y}^{q_{0}}\right]}.\label{magn1}
266: \end{equation}
267: For the more concreteness we regarded the Bethe lattice where
268: $q_{i}$ takes values $q=3,4$ with equal probability on the each step
269: of the lattice generation and for the each site. Corresponding
270: magnetization $m_{dis}$ with respect to external field $\beta h$ at
271: $\beta J=0.5$ point is given in Fig. (\ref{magni}).
272: \begin{figure}
273: \begin{center}
274: \includegraphics[width=7.5cm]{magn.ps}
275: \caption{\label{magni} Magnetization curves with $\beta J=0.5$ in presence of topological disorder, where
276: $q_{k}=3,4$ with equal probabilities.}
277: \end{center}
278: \end{figure}
279: To obtain the magnetization we use the following numerical
280: procedure. At each point $\beta h$, starting from $\beta h=0$ with
281: step $\Delta(\beta h)=0.001$, we get fixed points of the mapping
282: (\ref{mapp}). In our simple example, evaluating fixed points one
283: have to replace ${\{x_{k},y_{k}\}}$ and ${\{x_{k-1},y_{k-1}\}}$ in
284: equations (\ref{mapp}) on ${\{x,y\}}$. Solution of the obtained
285: system of equations with respect to ${\{x,y\}}$ gets the set of the
286: searched fixed points. In presence of the magnetic field there are
287: only one fixed point holding conditions $x_{k}>0$ and $y_{k}>0$. In
288: case of the zero magnetic field one have two stable fixed points and
289: one nonstable, or otherwise, depending on temperature. For checking
290: stability of the fixed points of (\ref{mapp}), one can calculate
291: Jacobian of (\ref{mapp}) as
292: \begin{equation}
293: J=\left|
294: \begin{array}{ll}
295: e^{J+h}x+\frac{3}{2}e^{J+h}x^2 & e^{-J-h}y+\frac{3}{2}e^{-J-h}y^2 \\
296: e^{-J+h}x+\frac{3}{2}e^{-J+h}x^2 & e^{J-h}y+\frac{3}{2}e^{J-h}y^2
297: \end{array}
298: \right|. \label{jakob}
299: \end{equation}
300: The condition of stability now sounds as $|J|<1$.
301:
302: It is known that the Ising model on Bethe lattice with constant $q$
303: has a critical point defined by condition $\beta_{c}
304: J=\frac{1}{2}\ln \frac{q}{q-2}$ \cite{Baxter}. According to
305: \cite{Baxter}, it is possible to write one dimensional mapping for
306: the partition function of the model with fixed $q$. This mapping has
307: three fixed points in zero magnetic field. It is easy to check that
308: the temperature corresponding to the case of the three non stable
309: fixed points defines by the same condition as the critical
310: temperature. When temperature is less then the critical temperature,
311: one obtains two stable fixed points (spontaneous magnetization) and
312: one non stable $x=1$. For the temperature greater then critical one,
313: we get two nonstable and one stable $x=1$ points. This
314: interpretation of known regimes \cite{Baxter} around the critical
315: point gives a possibility for the numerical investigation in
316: presence of the topological disorder. For the evaluation of critical
317: temperature of our example $q=3,4$ we propose the following
318: numerical procedure. Starting from point $\beta J=0$ and holding
319: $h=0$, with step $\Delta (\beta J)=0.005$, we check stability of
320: obtained fixed points and looking for $\beta J$ at which two stable
321: points become unstable and the third one becomes stable. The
322: obtained result is not an exact value of the critical temperature,
323: but an interval of possible values. The size of this interval
324: depends on the precision $\Delta (\beta J)$. As a result we found
325: $\beta_{c} J=0.725$, which is greater then the corresponding values
326: for the Ising models on Bethe lattice with fixed $q=3$ and $q=4$.
327:
328: Another interesting choice of topological disorder is $q=2,3$, where $q$
329: takes corresponding values with probability $P(q)$. Here we consider
330: only uniform probability distribution $P(2)=P(3)=0.5$. The general
331: study is a subject of further research. The coordination number
332: $q=2$ corresponds to the one dimensional Ising model having zero
333: critical temperature. The existence and localization of the critical
334: point in this mix of one-dimensional Ising model and Ising model on
335: Bethe lattice is a subject of special interest. The main idea is an
336: evaluation of the fixed points of corresponding mapping, as it was
337: done above for $q=2,3$. However, during this procedure one obtains
338: no fixed points holding condition $x>0,y>0$. For any positive
339: initial seed of mapping, after some finite number of iterations
340: (after $15-20$ iterations) $x_{k},y_{k}$ become infinitely big. To
341: deal with these big numbers we define new variable
342: $z_{k}=x_{k}/y_{k}$. The mapping (\ref{mapp}) may be represented now
343: as
344: \begin{equation}
345: z_{k}=\frac{P(3)(e^{\beta J+\beta h}z^{2}_{k-1}+e^{-\beta J-\beta
346: h})+o(n)}{P(3)(e^{-\beta J+\beta h}z^{2}_{k-1}+e^{\beta J-\beta
347: h})+o(n)}.\label{mappa}
348: \end{equation}
349: The infinitesimals $x_{k}/y^{2}_{k},1/y_{k}$ may be neglected and we
350: obtain that the critical temperature is the same as for Ising model
351: on Bethe lattice with constant $q=3$.
352: \section{Quenched topological
353: disorder}\label{4} In case of partly annealed topological disorder
354: $0<n<1$ we use the following heuristics. $\textbf{Z}$ quantity may
355: be represented as
356: \begin{eqnarray}
357: \textbf{Z}=<Z(\{q_{i}\})^n>_{dis}=\sum_{\{q_{i}\}}P(\{q_{i}\})[\exp(\beta
358: h )g_{\{q_{i}\}}^{q_{0}}(+)\nonumber\\+\exp(-\beta h
359: )g_{\{q_{i}\}}^{q_{0}}(-)]^n,\label{dyn2}
360: \end{eqnarray}
361: where $g_{\{q_{i}\}}(s_{0})$ depends on the particular realization
362: of disorder. One may obtain recursion relations from the previous
363: expression as
364: \begin{eqnarray}
365: x_{k}=e^{\beta J+\beta h}x^{q_{k}-1}_{k-1}+e^{-\beta
366: J-\beta h}y^{q_{k}-1}_{k-1}\nonumber\\
367: y_{k}=e^{-\beta J+\beta h}x^{q_{k}-1}_{k-1}+e^{\beta
368: J-\beta h}y^{q_{k}-1}_{k-1},\label{mapp2}
369: \end{eqnarray}
370: denoting $x_{k}=g_{\{q_{i}\}}(+)$ and $y_{k}=g_{\{q_{i}\}}(-)$. It
371: is assumed that on the each step of recursion the power $q_{k}$
372: takes random values with given probability distribution $P(q)$. The
373: main idea is to consider that although the mapping is stochastic in
374: nature, after some huge number of iterations one would obtain an
375: attractor set of possible pairs $(\overline{x},\overline{y})$ with
376: corresponding stationary probability distribution $P(x,y)$. To find
377: the attractor numerically one may study the mappings (\ref{mapp2})
378: with fixed non stochastic values of $q_{k}=q$. The fixed points of
379: those mappings should be a part of searching attractor. One may
380: start stochastic iteration (\ref{mapp2}) from those fixed points and
381: get the set of possible pairs $(\overline{x},\overline{y})$. There
382: is a possibility of a disconnected attractor (and broken
383: ergodicity), therefore the final set should be based on the
384: summation of sets derived from iterations of all possible fixed
385: points. During the iteration process one gets not only an attractor
386: set but also it is possible to evaluate $\textbf{Z}$. The method of
387: iterations takes a chance to calculate value of $<[\exp(\beta h
388: )x_{k}^{q_{0}}+\exp(-\beta h)y_{k}^{q_{0}}]^n>_{dis}$. Although the
389: numerical and approximate nature of such evaluation we suppose a
390: consistent result in the limit of large number of iterations. The
391: magnetization in presence of partly annealed disorder may be written
392: as
393: \begin{widetext}
394: \begin{eqnarray}
395: m_{dis}&=&\frac{\int DJ P(J)m(J)\exp(-\overline{\beta}F(J))}{\int DJ
396: P(J)\exp(-\overline{\beta}F(J))}=
397: \frac{\int DJ P(J)\frac{\partial Z(J)}{\partial \beta h}(Z(J))^{n-1}}{\int DJ P(J)(Z(J))^{n}}\nonumber\\
398: &=&\frac{<[\exp(\beta h )x_{k}^{q_{0}}-\exp(-\beta
399: h)y_{k}^{q_{0}}]\times[\exp(\beta h )x_{k}^{q_{0}}+\exp(-\beta
400: h)y_{k}^{q_{0}}]^{(n-1)}>_{dis}}{<\exp(\beta h
401: )x_{k}^{q_{0}}+\exp(-\beta h)y_{k}^{q_{0}}]^n>_{dis}}.\label{magn3}
402: \end{eqnarray}
403: \end{widetext}
404: However, the analysis of mapping (\ref{mapp2}) for the simple
405: example where $q_{k}=3,4$ with equal probabilities demonstrates some
406: difficulties. After few steps of iteration the values of
407: $x_{k},y_{k}$ approach to zero or infinity. To overcome it we
408: propose a mathematical trick, which should give correct values of
409: magnetization for $n\rightarrow0$ and exact magnetization in
410: presence of quenched topological disorder $n=0$. Instead of $x_{k}$
411: and $y_{k}$ one may consider new variable $z_{k}=x_{k}/y_{k}$ and
412: write a one dimensional mapping
413: \begin{equation}
414: z_{k}=\frac{e^{\beta J+\beta h}z^{q_{k}-1}_{k-1}+e^{-\beta J-\beta
415: h}}{e^{-\beta J+\beta h}z^{q_{k}-1}_{k-1}+e^{\beta J-\beta
416: h}}.\label{onemap}
417: \end{equation}
418: The mathematical trick lies in taking out the means over the
419: disorder $y_{k}$ in Eq. (\ref{magn3}), like it is possible to do
420: with constant parameter (real number). It is supposed that if
421: $y_{k}$ is infinite quantity or infinitesimal one, than this
422: approach is reliable. The magnetization may be rewritten now as
423: \begin{widetext}
424: \begin{equation}
425: m_{dis}\approx \frac{<[\exp(\beta h )z_{k}^{q_{0}}-\exp(-\beta
426: h)]\times[\exp(\beta h )z_{k}^{q_{0}}+\exp(-\beta
427: h)]^{(n-1)}>_{dis}}{<\exp(\beta h )z_{k}^{q_{0}}+\exp(-\beta
428: h)]^n>_{dis}}.\label{mag4}
429: \end{equation}
430: \end{widetext}
431: Previous relation Eq. (\ref{mag4}) becomes exact in presence of
432: quenched topological disorder as it follows directly from Eq.
433: ({\ref{magn3}}). Using this trick is easy to get magnetization for
434: our example with $q=3,4$. To obtain magnetization we change the
435: values $\beta h$ like it was done for the annealed version of
436: disorder. We start evaluation of important means after $10^4$-th
437: iteration and calculating them using the same number of iterations.
438: The result does not change if this value is greater, therefore we
439: suppose that it is stable. The values of $z_{k}$ are distributed
440: around the fixed points of the mappings (\ref{onemap}) with fixed
441: values of $q_{k}$. The dependence of magnetization on external field
442: $h$ in presence of quenched topological disorder is given in Fig
443: (\ref{magni}).
444: \begin{figure}
445: \begin{center}
446: \includegraphics[width=7.5cm]{critic.ps}
447: \caption{\label{crit} Spontaneous magnetization in presence of quenched topological disorder, where
448: $q_{k}=3,4$ with equal probabilities.}
449: \end{center}
450: \end{figure}
451:
452: To get the position of the critical temperature for our example
453: $q=3,4$ we use the basic knowledge about spontaneous magnetization.
454: The dependence of the spontaneous magnetization on temperature
455: provides an opportunity to single out the critical point. Iteration
456: process and evaluation of the magnetization under the zero magnetic
457: field, as it was described above, give us dependence of the
458: spontaneous magnetization on $\beta J$ (Fig. (\ref{crit})). The
459: critical temperature lies between the critical temperatures
460: corresponding to the same models on Bethe lattices with constant
461: $q=3$ and $q=4$ coordination numbers.
462:
463: The same analysis may be done for the quenched topological disorder
464: with $q=2,3$. Our study of the spontaneous magnetization shows that
465: the critical point exists and the critical temperature differs from
466: zero even when the probability of $q=2$ is greater than the
467: probability of $q=3$ ($P(3)=0.01$, $2<\beta_{c} J<3$). It means that
468: the little possibility of branching in one dimensional Ising model
469: (rare branches) lead to the critical behavior with non zero critical
470: temperature in the thermodynamical limit.
471:
472: As it is easy to see, we mentioned above some propositions which
473: are very general (statements about attractor and other ) and they
474: need to be studied with mathematical rigor. However, this statements
475: are useful and numerical research gives an evidence of their
476: correctness.
477: \section{Conclusion}\label{5}
478: Summarizing written above, we developed an approach to deal with
479: topological disorder for the spin models on hierarchical lattices.
480: Despite its demonstration only on the simple Bethe lattice, a
481: generalization on more complicated models is straightforward. A wide
482: spectra of problems solving within approach of hierarchical models
483: and having well physical description within it may be revisited now
484: in presence of fluctuating coordination number of lattice. A
485: complicated disorder may lead to the nontrivial solutions of the
486: recursive relations Eq. (\ref{mapp}). The distributions of the
487: Yang-Lee and Fisher zeros in presence of disorder is also an
488: interesting problem. Speaking about social dynamics we see that the
489: hierarchical models take a possibility to simplify the problem
490: within Ising paradigm without significant loss of verity and
491: developed approach makes it possible to get the major interesting
492: quantities.
493: \section{Acknowledges}
494: Authors are grateful to E. Mamasakhlisov and A. Allahverdyan for discussions
495: . K.S. would like to thank K. Mazmanian for useful remarks.
496: This work was supported by ANSEF grant PS-condmatth-521.
497: \begin{thebibliography}{99}
498: \bibitem{Pretti}
499: M. Pretti, Phys. Rev. E \textbf{74},051803(2006).
500: \bibitem{Gujrati}
501: P. D. Gujrati, Phys. Rev. Lett. \textbf{74}, 809 (1995).
502: \bibitem{Onsager}
503: M.-S. Chen, L. Onsager, J. Bonner, and J. Nagle, J. Chem. Phys. \textbf{60}, 405 (1974).
504: \bibitem{Akheyan}
505: A. Z. Akheyan, N. S. Ananikian, J. Phys. A \textbf{29}, 4, 721
506: (1996).
507: \bibitem{Izmail}
508: N. S. Ananikian, A. R. Avakian, N. Sh. Izmailian, Physica A
509: \textbf{172}, 391 (1991).
510: \bibitem{sherb}
511: N. S. Ananikian, R. R. Shcherbakov, J. Phys. A \textbf{27}, L887
512: (1994).
513: \bibitem{zp}
514: J. Vannimenus, Z. Phys. B 43, 141 (1981).
515: \bibitem{Keskin}
516: E. Albayrak and M. Keskin, Eur. Phys. J. B \textbf{24}, 505 (2001).
517: \bibitem{esk}
518: M. Eckstein, M. Kollar, K. Byczuk, and D. Vollhardt, Phys. Rev. B \textbf{71}, 235119 (2005).
519: \bibitem{dis1}
520: D. J. Thouless, Phys. Rev. Lett. \textbf{56}, 1082 (1986).
521: \bibitem{dis2}
522: J. M. Carlson, J. T. Chayes, L. Chayes, J. P. Sethna, D. J. Thouless, Europhys. Lett. \textbf{5} 355 (1988).
523: \bibitem{dis3}
524: T. Morita, Phisica A \textbf{98}, 566 (1979).
525: \bibitem{dis4}
526: M. J. de Oliveira, S. R. Salinas, Phys. Rev. B \textbf{35}, 2005 (1987).
527: \bibitem{soc}
528: C. Castellano, S. Fortunato, V. Loreto, physics.soc-ph: arXiv:0710.3256v1.
529: \bibitem{Dotsenko}
530: V. Dotsenko, Introduction to the Replica Theory of Disordered
531: statistical systems, Cambridge ; New York : Cambridge University
532: Press, 2001.
533: \bibitem{Allahv}
534: A. E. Allahverdyan, Th. M. Nieuwenhuizen, Phys. Rev. E \textbf{62},
535: 845(2000); A. E. Allahverdyan, K. G. Petrosyan Europhys. Lett.
536: \textbf{75}, 6, 908 (2006).
537: \bibitem{din0}
538: V. R. Ohanyan, L. N. Ananikyan, N. S. Ananikian, Physica A \textbf{377}, 501-513 (2007).
539: \bibitem{din1}
540: R.G. Ghulghazaryan, N.S. Ananikian. J.Phys. A., \textbf{36}, 6297 (2003).
541: \bibitem{din2}
542: R.G. Ghulghazaryan, N.S. Ananikian, P.M.A.Sloot. Phys. Rev. E, \textbf{66},
543: 046110 (2002).
544: \bibitem{din3}
545: N.S. Ananikian, R.G. Ghulghazaryan. J. Comp.
546: Methods in Sciences and Engineering, 2, 75. (2002).
547: \bibitem{din4}
548: A.E. Alakhverdian., N.S. Ananikian, S.K. Dallakian. Phys.Rev.E, \textbf{57},
549: 2452 (1998).
550: \bibitem{Monroe}
551: L. Monroe. Phys. Lett. A, 188, 80 (1994).
552: \bibitem{Baxter}
553: R. J. Baxter, Exactly Solaed Models in Statistical Mechanics
554: (Academic, London, 1982).
555: \end{thebibliography}
556: \end{document}
557: