1: \documentclass[twocolumn,showpacs,a4paper,pre]{revtex4}
2: \usepackage{graphicx}
3: \usepackage{amsmath}
4: \usepackage{amssymb}
5:
6: \begin{document}
7:
8: \title{Star-graph expansions for bond-diluted Potts models}
9: \date{\today}
10:
11: \author{Meik Hellmund}
12: \email{Meik.Hellmund@itp.uni-leipzig.de}
13: \homepage{http://www.physik.uni-leipzig.de/~hellmund}
14: \author{Wolfhard Janke}
15: \email{Wolfhard.Janke@itp.uni-leipzig.de}
16: \affiliation{Institut f{\"u}r Theoretische Physik, Universit{\"a}t Leipzig,
17: Augustusplatz 10/11, D-04109 Leipzig, Germany}
18:
19: \begin{abstract}
20: We derive high-temperature series expansions for the free energy and the
21: susceptibility of random-bond $q$-state Potts models on hypercubic lattices
22: using a star-graph expansion technique. This method enables the exact
23: calculation of quenched disorder averages for arbitrary uncorrelated coupling
24: distributions. Moreover, we can keep the disorder strength $p$ as well as the
25: dimension $d$ as symbolic parameters. By applying several series analysis
26: techniques to the new series expansions,
27: one can scan large regions of
28: the $(p,d)$ parameter space for any value of $q$. For the bond-diluted
29: 4-state
30: Potts model in three dimensions, which exhibits a rather
31: strong first-order phase transition in the undiluted case, we present
32: results for the transition
33: temperature and the effective critical exponent $\gamma$
34: as a function of $p$ as obtained from the analysis of
35: susceptibility series up to order 18. A comparison with recent Monte Carlo
36: data (Chatelain {\em et al.}, Phys.~Rev.~E64, 036120(2001))
37: shows signals for the softening to a second-order transition
38: at finite disorder strength.
39: \end{abstract}
40: \pacs{\\
41: 05.50.+q Lattice theory and statistics (Ising, Potts, etc.) \\
42: 64.60.Cn Order-disorder transformations; statistical mechanics
43: of model systems\\
44: 64.60.Fr Equilibrium properties near critical points, critical exponents
45: }
46:
47:
48: \maketitle
49:
50: \section{Introduction}
51:
52: Systematic series expansions for statistical models defined on a lattice are a
53: well-known method to study phase transitions and critical phenomena
54: \cite{domb3}. They provide an useful complement to large-scale numerical
55: simulations, in particular for quenched, disordered systems where the
56: average over many different disorder realizations is numerically very
57: time consuming and only some points in the vast parameter space of the systems
58: can be sampled with realistic effort.
59:
60: Using high-temperature series expansions, on the other hand, one can obtain
61: for many quantities results which are
62: exact up to a certain order in the inverse
63: temperature. Here the infinite-volume limit can be taken without problems and
64: the quenched disorder is treated exactly.
65: Moreover, one can keep the disorder strength $p$ as well as the
66: dimension $d$ as symbolic parameters and therefore analyse much larger regions
67: of the parameter space of disordered systems.
68: To this end we developed further the method of
69: ``star-graph expansion'' which allows us to take the disorder average on
70: the level of individual graphs exactly and apply it to $q$-state
71: Potts models with a
72: bimodal quenched distribution of ferromagnetic couplings.
73: Using optimized cluster algorithms for the exact calculation of spin-spin
74: correlators on graphs with arbitrary inhomogeneous couplings,
75: we obtained series up to order 18 in the inverse temperature
76: for the susceptibility and the free energy of bond-diluted
77: Potts models in two, three and four dimensions.
78:
79: Depending on the dimension $d$ and the number of states $q$, pure Potts
80: models show first- or second-order phase transitions. According to the Harris
81: criterion~\cite{harris}
82: one expects in the second-order case either the appearance of a new
83: random fixed point ($d=2$, $q=3,4$ and $d=3$, $q=2$) or
84: logarithmic corrections
85: to the pure fixed point ($d=2$, $q=2$).
86: At first-order transitions, randomness softens the transitions. For $d=2$ even
87: infinitesimal disorder induces a continuous transition~\cite{aiz}, whereas for
88: $d=3$, $q>2$ a tricritical point at a finite disorder strength is
89: expected~\cite{card97}.
90: This softening to a second-order phase transition
91: beyond a tricritical point at some finite disorder strength has recently been
92: verified in Monte Carlo (MC) simulations
93: of the three-dimensional {\em site\/}-diluted 3-state~\cite{balles00}
94: and {\em bond\/}-diluted 4-state~\cite{chat01a} Potts model.
95:
96: The critical part of the series expansion methods
97: lies in the extrapolation techniques which are used
98: in order to obtain information on the critical singularity from
99: a finite number of known coefficients of the high-temperature series.
100: One can question the use of these extrapolation techniques in disordered
101: systems, where the complete singularity structure of the function may be very
102: complicated, involving Griffiths-type singularities or logarithmic
103: corrections~\cite{card99}.
104:
105: Anyhow, we are able to determine the transition temperature
106: for the bond-diluted 4-state Potts model in three dimensions
107: reliably up to the vicinity of the geometrical percolation point and in good
108: agreement with analytic estimates~\cite{turb} and MC
109: results~\cite{chat01a}.
110:
111: The critical exponent $\gamma$ extracted from our analysis appears to be dependent
112: on the disorder strength which is caused by crossover effects and the
113: complicated singularity structure. Using sophisticated analysis methods,
114: we find a range of the disorder strength where
115: $\gamma_{\rm eff}=1$, indicating critical behavior
116: governed by a tricritical point.
117:
118: The rest of the paper is organized as follows.
119: In Sect.~\ref{sec:m} we briefly recall the model.
120: In Sect.~\ref{sec:s} we describe the
121: methods used for generating the series, and Sect.~\ref{sec:a}
122: is devoted to a representation of the analysis techniques used
123: and their application to the study of the bond-diluted 4-state Potts model
124: in three dimensions.
125:
126: \section{Model}
127: \label{sec:m}
128: The $q$-state Potts model on the hypercubic lattice $\mathbb{Z}^d$, or more
129: generally on any
130: graph $G$ with arbitrary coupling
131: constants $J_{ij}$ assigned to the links $\langle ij\rangle$
132: of $G$, is defined by its partition function
133: %
134: \begin{equation}
135: \label{eq:1}
136: Z = \sum_{\{S_i\}} \exp \left(\beta \sum_{\langle ij\rangle } J_{ij} \delta(S_i, S_j)\right),
137: \end{equation}
138: %
139: where $\beta=1/k_B T$ is the inverse temperature, $S_i=1,\ldots,q$ and
140: $\delta(.,.)$ is the Kronecker symbol.
141: In our series expansion the combination
142: \begin{equation}
143: \label{eq:v2}
144: v_{ij} = \frac{e^{\beta J_{ij}}-1}{e^{\beta J_{ij}}-1+q}
145: \end{equation}
146: will be the relevant expansion parameter.
147: In the symmetric high-temperature phase, the
148: susceptibility corresponding to the coupling to an external field $h_i$,
149: $\sum_i h_i \left(\frac{q \delta(S_i,1) -1}{q-1}\right)$,
150: is given for a graph with $N$ spins by
151: %
152: \begin{equation}
153: \label{eq:s1}
154: \chi = \frac{1}{N} \sum_i \sum_j \left[ \left\langle \frac{q \delta(S_i, S_j)
155: -1}{q-1}\right\rangle \right].
156: \end{equation}
157: %
158: Quenched disorder averages $[\ldots]$ are taken
159: over an uncorrelated bimodal distribution of the form
160: %
161: \begin{equation}
162: \label{eq:bi}
163: P(J_{ij}) = (1-p) \delta(J_{ij}-J_0) + p \delta(J_{ij}- RJ_0),
164: \end{equation}
165: %
166: which can include spin glasses ($R=-1, p=1/2$),
167: random-bond ferromagnets ($0<R<1$) and
168: bond dilution ($R=0$)
169: as special cases. Other distributions can,
170: in principle, also be considered with our method.
171:
172:
173: \section{Series generation}
174: \label{sec:s}
175:
176: \subsection{Basic notations from graph theory}
177: A graph of order $E$ consists of $E$ links connecting $N$ vertices.
178: We consider only connected, undirected graphs that are simple:
179: no link starts and ends at the
180: same vertex and two vertices are never connected by more than one link.
181: Subgraphs are defined by the deletion of links. In this process, isolated
182: vertices can be dropped. A graph of order $E$ has $2^E$
183: (not necessarily non-isomorphic) subgraphs
184: since each link may be present or absent.
185: These subgraphs may consist of several connected components
186: and are called clusters.
187:
188: An articulation point is a vertex the deletion of which renders the graph
189: disconnected. A graph without articulation points is called ``star graph''.
190:
191: A graph is bipartite if the vertices can be separated into red and black
192: vertices so that no link connects two vertices of the same
193: color. Equivalently, all closed paths in the graph consist of an even number
194: of links.
195:
196: \subsection{Star-graph expansion method}
197:
198: There are two well-established methods \cite{domb3} for the systematic
199: generation of high-temperature series expansions,
200: the linked cluster and the star-graph
201: method. The longest known series (up to order $\beta^{25}$)
202: for classical spin models without disorder
203: are produced by linked cluster expansions \cite{butera02}. This technique
204: allows one to obtain series for observables (such as the second moment of the
205: spin-spin correlation function) which have no star-graph
206: expansion. Furthermore, it works with free embeddings of graphs into the
207: lattice which can be counted orders of magnitude faster than the weak
208: embedding numbers used by the star-graph technique. Nonetheless, this method
209: has not yet been applied to problems with quenched disorder.
210:
211: The star-graph method can be adopted to systems involving quenched disorder
212: \cite{rap1,singh87} since it allows one to take the disorder average
213: on the level of individual graphs.
214: The basic idea is to assemble the value of some extensive thermodynamic
215: quantity $F$ on a large or even infinite graph from its values on subgraphs:
216: Graphs constitute a partially ordered set under the ``subgraph'' relation.
217: Therefore, for every function $F(G)$ defined on the set of graphs
218: exists another function $W_F(G)$ such that for all graphs $G$
219: \begin{equation}
220: \label{eq:2}
221: F(G) = \sum_{g \subseteq G} W_F(g),
222: \end{equation}
223: and this function can be calculated recursively via
224: \begin{equation}
225: \label{eq:3}
226: W_F(G) = F(G) - \sum_{g \subset G} W_F(g).
227: \end{equation}
228: This gives for an infinite (e.g. hypercubic) lattice
229: \begin{equation}
230: \label{eq:4}
231: F(\mathbb{Z}^d) = \sum_G (G:\mathbb{Z}^d)\, W_F(G),
232: \end{equation}
233: where $(G:\mathbb{Z}^d)$ denotes the weak
234: embedding number of the graph $G$ in the given
235: lattice structure~\cite{martin74}.
236:
237: The following observation makes this a useful method:
238: Let $G$ be a graph with an articulation vertex
239: where two star subgraphs $G_{1,2}$ are glued together.
240: Then $W_F(G)$ vanishes if
241: \begin{equation}
242: \label{eq:sg}
243: F(G) = F(G_1) + F(G_2).
244: \end{equation}
245: An observable $F$ for which Eq.~(\ref{eq:sg}) is true on arbitrary graphs
246: with articulation points allows a star-graph expansion. All non-star graphs
247: have zero weight $W_F$ in the sum Eq.~(\ref{eq:4}).
248:
249: It is easy to see that the (properly normalized) free energy $\log Z$
250: has this property and it can be proved~\cite{singh87} that the inverse
251: susceptibility $1/\chi$ has it, too, even for arbitrary inhomogeneous couplings
252: $J_{ij}$. This restricts the sum in Eq.~(\ref{eq:4}) to a sum over star graphs.
253: The linearity of Eqs.~(\ref{eq:2})-(\ref{eq:4}) enables the calculation of
254: quenched averages over the coupling distribution on the level of individual
255: graphs.
256:
257: The resulting recipe for the susceptibility series is:
258:
259: \begin{itemize}
260: \item Graph generation and embedding number counting.
261: \item Calculation of $Z(G)$ and the correlation matrix\\
262: $M_{nm}(G) = \mathrm{Tr}\, (q\delta(S_n,S_m)-1) e^{-\beta H(\{J_{ij}\})}$\\
263: for all graphs as polynomials in $E$ variables $v_{ij}$.
264: \item Inversion of the $Z$ polynomial as a series up to the desired order.
265: \item Averaging over quenched disorder,\\
266: $N_{nm}(G) = \left[ M_{nm}/Z \right]_{P(J)},$\\
267: resulting in a matrix of polynomials in $(p,v)$.
268: \item Inversion of the matrix $N_{nm}$ and subgraph subtraction,\\
269: $W_\chi (G) = \sum_{n,m} (N^{-1})_{nm} - \sum_{g\subset G} W_\chi(g)$.
270: \item Collecting the results from all graphs,\\
271: $1/\chi = \sum_G (G: \mathbb{Z}^d)\; W_\chi(G)$.
272: \end{itemize}
273:
274:
275:
276: \subsection{Generation of star graphs and calculation of embedding numbers}
277:
278: The most complicated part in every attempt to
279: generate lists of graphs by recursively adding nodes and edges
280: to a smaller list
281: is the isomorphism test, i.e., the decision whether two
282: given adjacency lists or adjacency matrices describe the same graph modulo
283: relabelling and reordering of edges and nodes. We used the \texttt{nauty}
284: package by
285: McKay \cite{mckay81} which makes very fast isomorphism tests by
286: calculating a canonical representation of the
287: automorphism group of the graphs.
288:
289: {\squeezetable
290: \begin{table}
291: \caption{\label{tab:1}Number of star graphs with $E$ links and
292: non-vanishing embedding numbers on $\mathbb{Z}^d$.}
293: \begin{ruledtabular}
294: \begin{tabular}{l|*{15}{r|}r}
295: order $E$&1 &4&6&7&8&9&10&11&12&13&14&15&16&17&18&19\\
296: \hline
297: $\#$ &1&1&1&1&2&3&8&9&29&51&142&330&951&2561&7688&23078\\
298: \end{tabular}
299: \end{ruledtabular}
300: \end{table}
301: }
302:
303: Since we are only interested in star graphs with non-vanishing
304: weak embedding numbers in $\mathbb{Z}^d$, the following simple observations are
305: helpful:
306: \begin{itemize}
307: \item Only bipartite graphs occur since $\mathbb{Z}^d$ is bipartite.
308: \item A generic $k$-dimensional embedding (i.e. one which really needs all
309: $k$ dimensions) contributes in $d>k$
310: dimensions with degeneracy $\binom{d}{k}$.
311: \item A biconnected graph of odd order $E=2n+1$ has generic embeddings only up
312: to dimension $n$ since it must have at least two edges in each dimension.
313: \item The only biconnected graph of even order $E=2n$ which has generic
314: embeddings of dimension $n$ is the cycle of length $2n$. All the other
315: graphs will use at most $n-1$ dimensions.
316: \end{itemize}
317:
318: For the embedding count we implemented
319: a refined version of the backtracing algorithm by
320: Martin \cite{martin74}. We did extensive tests to find the optimal algorithm
321: for the ``innermost'' loop, the test for collisions in the embedding, and ended
322: up using optimized hash tables.
323:
324: \begin{figure}
325: \unitlength6mm
326: \begin{center}
327: \begin{picture}(8,1)
328: \thicklines
329: \put(0,0){\line(1,0){8}}
330: \put(0,0){\line(0,1){1}}
331: \put(0,1){\line(1,0){8}}
332: \put(8,0){\line(0,1){1}}
333: \put(3,0){\line(0,1){1}}
334: \put(0,0){\circle*{.24}}
335: \put(1,0){\circle*{.24}}
336: \put(2,0){\circle*{.24}}
337: \put(3,0){\circle*{.24}}
338: \put(4,0){\circle*{.24}}
339: \put(5,0){\circle*{.24}}
340: \put(6,0){\circle*{.24}}
341: \put(7,0){\circle*{.24}}
342: \put(8,0){\circle*{.24}}
343: \put(0,1){\circle*{.24}}
344: \put(1,1){\circle*{.24}}
345: \put(2,1){\circle*{.24}}
346: \put(3,1){\circle*{.24}}
347: \put(4,1){\circle*{.24}}
348: \put(5,1){\circle*{.24}}
349: \put(6,1){\circle*{.24}}
350: \put(7,1){\circle*{.24}}
351: \put(8,1){\circle*{.24}}
352: \end{picture}
353: \end{center}
354: \small
355: \begin{eqnarray*}
356: & 7620 \binom{d}{2} + 76851600 \binom{d}{3}+ 14650620864 \binom{d}{4}\\
357: &+\; 404500471680\binom{d}{5}+ 3355519311360 \binom{d}{6}
358: \end{eqnarray*}
359:
360:
361: \begin{center}
362: \begin{picture}(6,3)
363: \thicklines
364: \put(0,0){\line(1,0){3.9}}
365: \put(1,1){\line(1,0){3.9}}
366: \put(0,0){\line(1,1){1}}
367: \put(3.9,0){\line(1,1){1}}
368: \put(4.9,1){\line(0,1){1.3}}
369: \put(2.6,0){\line(0,1){1.3}}
370: \put(1.3,1.3){\line(1,0){2.6}}
371: \put(2.3,2.3){\line(1,0){2.6}}
372: \put(1.3,1.3){\line(1,1){1}}
373: \put(2.6,1.3){\line(1,1){1}}
374: \put(3.9,1.3){\line(1,1){1}}
375: \put(0,0){\circle*{.24}}
376: \put(1,1){\circle*{.24}}
377: \put(1.3,0){\circle*{.24}}
378: \put(2.6,0){\circle*{.24}}
379: \put(3.9,0){\circle*{.24}}
380: \put(2.3,1){\circle*{.24}}
381: \put(3.6,1){\circle*{.24}}
382: \put(4.9,1){\circle*{.24}}
383: \put(1.3,1.3){\circle*{.24}}
384: \put(2.6,1.3){\circle*{.24}}
385: \put(3.9,1.3){\circle*{.24}}
386: \put(2.3,2.3){\circle*{.24}}
387: \put(3.6,2.3){\circle*{.24}}
388: \put(4.9,2.3){\circle*{.24}}
389: \end{picture}
390: \end{center}
391: \small
392: \begin{eqnarray*}
393: &12048 \binom{d}{3}+ 396672\binom{d}{4} + 2127360\binom{d}{5}+
394: 2488320\binom{d}{6}
395: \end{eqnarray*}
396:
397:
398: \caption{Two star graphs of order 17 and 19
399: and their weak embedding numbers up to 6 dimensions.}
400: \label{fig:emb}
401: \end{figure}
402:
403: By this means, we
404: classified for the first time all star graphs up to order 19 which can be
405: embedded in hypercubic lattices (see Table~\ref{tab:1}) and
406: calculated their (weak) embedding numbers for $d$-dimensional hypercubic
407: lattices (up to order 17 for arbitrary $d$, order 18 and 19 for
408: dimensions $\leq4$), see Fig.~\ref{fig:emb} for typical results.
409:
410: \subsection{Cluster representation}
411:
412: The partition function and the matrix of
413: correlations $M_{nm}$ for each graph are calculated
414: with arbitrary symbolic couplings $J_{ij}$ using the cluster representation
415: %
416: \begin{eqnarray}
417: \label{eq:cl1}
418: Z &=& q^{N-E} \prod_{\langle ij\rangle} (e^{\beta J_{ij}}-1+q) \; {\mathcal Z}, \\
419: {\mathcal Z} &=& q^{-N} \mathrm{Tr} \prod_{\langle ij\rangle}\left[1-v_{ij} +v_{ij} q \delta(S_i,S_j)
420: \right]
421: \label {eq:cl3}\\
422: &=&\!\sum_{C} q^{e+c-N}\!\! \left(\prod_{\langle ij\rangle \in C}
423: v_{ij}\!\right)\!\!\left( \prod_{\langle ij\rangle \notin C} (1-v_{ij})\! \right)\!.\label{eq:cl2}
424: \end{eqnarray}
425: %
426: Here the sum goes
427: over all clusters $C \subseteq G$, $E$ is the number of links ($=$ order) of the
428: graph $G$, $e$ is the number of links of
429: the cluster and $c$ the number of connected components of $C$.
430: ${\mathcal Z}$ is normalized such that $\log \mathcal Z$ has a star-graph
431: expansion. This essentially reduces the partition sum from a sum over $q^N$
432: states to a sum over $2^E$ clusters.
433: In the Ising case $q=2$ another huge
434: simplification takes place
435: since only clusters where all vertices are of even degree contribute
436: to the cluster sum.
437:
438:
439: The $2^E$ clusters belonging to a graph
440: are enumerated by Gray codes~\cite{numrec}
441: such that two consecutive clusters in the sum (\ref{eq:cl2}) differ by exactly
442: one (added or deleted) link. Gray codes are a reordering of the binary
443: representation of numbers such that the difference to the successor is in
444: exactly one bit position. For example, for $E=4$ the sequence is
445: 0000, 0001, 0011, 0010, 0110, 0111, 0101, 0100, 1100, 1101,
446: 1111, 1110, 1010, 1011, 1001, 1000 where zeros denote the deleted links.
447: This allows to speed up the calculation
448: considerably by re-using every term in the sum for the calculation of the next
449: one.
450:
451: The calculation of the susceptibility involves the matrix of correlations
452: $M_{nm}$. The effect of inserting
453: $\frac{q \delta(S_i, S_j) - 1}{q-1}$ into the trace of
454: Eq.~(\ref{eq:cl3}) can easily be seen: we get one if the vertices $n$ and $m$
455: belong to the same connected component of the cluster and zero otherwise.
456: Therefore,
457: \begin{equation}
458: \label{eq:clu}
459: M_{nm} \propto \sum_{C_{nm}} q^{e+c-N} \left( \prod_{\langle ij\rangle \in C}
460: v_{ij} \right)\left( \prod_{\langle ij\rangle \notin C} (1-v_{ij})\right),
461: \end{equation}
462: where the sum is restricted to all clusters $C_{nm}\subseteq G$ in
463: which the vertices $n$ and $m$ are connected.
464:
465:
466: For the symbolic calculations
467: we developed a C\raise2pt\hbox{\tiny++}
468: template library using an expanded degree-sparse
469: representation of polynomials and series in many variables.
470: The open source library GMP is used for the
471: arbitrary-precision arithmetics.
472:
473: Our longest series, up to order 18, are obtained for the case of bond dilution
474: where (\ref{eq:bi}) simplifies to
475: \begin{equation}
476: \label{eq:dil}
477: P(J_{ij}) = (1-p) \delta(J_{ij}-J_0) + p \delta(J_{ij}),
478: \end{equation}
479: since in this case the disorder average for a series is most easily done
480: via
481: \begin{equation}
482: \label{eq:bd}
483: [v_1^{n_1}\ldots v_k^{n_k}]_{P(J)} = (1-p)^k v_0^{n_1+\ldots+n_k}.
484: \end{equation}
485:
486:
487:
488:
489: \section{Series analysis: techniques and results}
490: \label{sec:a}
491:
492: In the following we shall illustrate the analysis using the
493: bond-diluted 4-state Potts model
494: in three dimensions as our primary example.
495: This model
496: exhibits in the pure case a strong
497: first-order transition which is expected to stay first order up to some
498: finite disorder strength, before it gets softened to a second-order transition
499: governed by a disordered fixed point.
500:
501: In the latter case we
502: are interested in locating power-law divergences in the susceptibility
503: series of the form
504: \begin{equation}
505: \label{eq:div}
506: \chi (v) = A (v_c-v)^{-\gamma} + \ldots
507: \end{equation}
508: For such a critical behavior
509: many different series analysis techniques have been
510: discussed in the literature
511: which all have their merits and drawbacks~\cite{guttmann89}.
512:
513: To localize a first-order transition point, however,
514: a high-temperature series alone is not sufficient since there
515: the correlation length remains finite and no critical singularity occurs.
516: In analysing series by ratio, Pad{\'e} or differential approximants, the
517: approximant will
518: provide an analytic continuation of the thermodynamic quantities beyond the
519: transition point into a metastable region on a pseudo-spinodal line
520: with a singularity $T^*_c < T_c$ and effective ``critical exponents''
521: at $T^*_c$.
522:
523: Employing the techniques described above
524: we obtained the high-temperature series expansions for the susceptibility up
525: to order~18
526: with coefficients given as polynomials in the disorder strength $p$,
527: as listed in the Appendix for general dimensions $d\leq4$. For
528: such a series in two variables, the method of partial differential
529: approximants~\cite{pda80}
530: should be well suited. Up to date, however, the only application
531: of this method to a tricritical point \cite{adler97} used a test series of
532: order 50 generated from an exactly solvable model. In our case, it was unable
533: to give conclusive results. Therefore, we confined ourselves to the analysis of
534: single-parameter series for selected values of $p$.
535:
536:
537: %
538: \begin{figure}
539: \includegraphics[scale=.33,angle=-90]{fig2.ps}
540: \caption{Ratio approximants for different dilutions $p$ vs.\ $1/n$.}
541: \label{fig:1}
542: \end{figure}
543: %
544: The ratio method is the least sophisticated method of series analysis,
545: but usually it is quite robust and
546: gives a good first estimate of the series behavior.
547: It assumes that the expected singularity of the form~(\ref{eq:div})
548: is the closest to the origin. Then the
549: consecutive ratios of series coefficients behave asymptotically as
550: %
551: \begin{equation}
552: \label{eq:rat}
553: r_n = \frac{a_n}{a_{n-1}} = v_c^{-1} \left(1+\frac{\gamma-1}{n}\right).
554: \end{equation}
555: %
556: Figure~\ref{fig:1} shows these ratios for different values of $p$.
557: In order to
558: make them visually comparable, they are normalized by their respective
559: critical couplings $v_c$.
560: For small $p$ they show the typical oscillations related to the existence
561: of an antiferromagnetic singularity at $-v_c$. Near the percolation threshold
562: at $p=0.751\,188$~\cite{lorenz98}
563: (where $T_c$ goes to 0) the series is clearly ill-behaved,
564: related to the $\exp(1/T)$ singularity expected there. Besides that,
565: we observe that the slope
566: ($\propto \gamma-1$) is increasing with $p$, changing from $\gamma<1$ to
567: $\gamma>1$ around $p=0.5$.
568:
569: The widely used DLog-Pad{\'e} method consists in calculating Pad{\'e}
570: approximants to the logarithmic derivative of $\chi(v)$. The smallest real
571: pole of the approximant is an estimation of $v_c$ and its residue gives
572: $\gamma$.
573: %
574: \begin{figure}
575: \includegraphics[scale=.33,angle=-90]{a.ps}
576: \caption{Critical temperature for different dilution $p$ as obtained from
577: MC simulations \cite{chat01a} and DLog-Pad{\'e} series analyses.
578: The inset shows the difference between the two estimates.}
579: \label{fig:2}
580: \end{figure}
581: %
582: Figure~\ref{fig:2} compares the critical temperature, estimated from an
583: average of 25-30 Pad{\'e} approximants for each value of $p$~\footnote{%
584: Notice that ``$p$'' in the present notation corresponds to ``$1-p$'' in
585: Ref.~\cite{chat01a}.},
586: with the results of recent MC simulations \cite{chat01a}.
587: For small $p$, in the first-order region, the series underestimates the
588: critical temperature. As explained above, this is an estimate not of $T_c$
589: but of $T^*_c$. Between $p=0.3$ and $p=0.5$, the estimates confirm,
590: within errors, the MC results, indicating that now both methods see
591: the same second-order transition. Beyond $p=0.5$, the scatter of different
592: Pad{\'e} approximants increases rapidly, related to the crossover to the
593: percolation point.
594:
595: \begin{figure}
596: \includegraphics[scale=.33,angle=-90]{figq4p04.ps}
597: \caption{Scattering of different Pad{\'e} approximants at a dilution $p=0.4$:
598: critical exponent $\gamma$ against critical coupling $v_c$.}
599: \label{fig:3}
600: \end{figure}
601:
602:
603:
604: The situation is more complicated with respect to the critical exponent $\gamma$.
605: A DLog-Pad{\'e} analysis gives inconclusive results due to a large scattering
606: between different Pad{\'e} approximants, as shown in Fig.~\ref{fig:3}. One
607: possible reason for this failure is the existence of confluent singularities:
608: The dots in Eq.~(\ref{eq:div}) indicate correction terms which can be
609: parametrized as follows:
610: \begin{equation}
611: \label{eq:div2}
612: \chi (v) = A (v_c-v)^{-\gamma} [1 + A_1(v_c-v)^{\Delta_1} + A_2 (v_c-v)^{\Delta_2} + \ldots],
613: \end{equation}
614: where $\Delta_i$ are the confluent correction exponents.
615: We used different
616: more sophisticated analysis methods,
617: such as inhomogeneous differential approximants~\cite{guttmann89}
618: and the methods M1 and M2~\cite{adler91},
619: especially tailored to deal with
620: confluent singularities. In the case at hand,
621: the Baker-Hunter method~\cite{bakerhunter} appeared to be quite successful,
622: giving consistent results at larger dilutions $p>0.35$ where the DLog-Pad{\'e}
623: analysis failed.
624: Assume the function under investigation has confluent singularities
625: \begin{equation}
626: \label{eq:b1}
627: F(z) = \sum_{i=1}^N A_i \left(1-\frac{z}{z_c}\right)^{-\lambda_i} = \sum_{n=0} a_n z^n .
628: \end{equation}
629: This can be transformed into an auxiliary function $g(t)$ which is
630: meromorphic and therefore suitable for Pad{\'e} approximation. After the
631: substitution $z = z_c(1-e^{-t})$ we expand $F(z(t)) = \sum_n c_n t^n $
632: and construct the new series
633: \begin{equation}
634: \label{eq:b2}
635: g(t) = \sum_{n=0} n!\;c_n\; t^n = \sum_{i=1}^N \frac{A_i}{1-\lambda_it}.
636: \end{equation}
637: We see that Pad{\'e} approximants to $g(t)$ have poles at $t=1/\lambda_i$, with
638: residues at the poles of $-A_i/\lambda_i$.
639: This method is applied by plotting these poles and residues for different
640: Pad{\'e} approximants to $g(t)$ as functions of $z_c$. The optimal set of values
641: for the parameters is determined visually from the best clustering of
642: different Pad{\'e} approximants, as demonstrated in Fig.~\ref{fig:4}.
643:
644: \begin{figure}
645: \includegraphics[scale=.33,angle=-90]{bh1.ps}
646: \includegraphics[scale=.33,angle=-90]{bh2.ps}
647:
648:
649: \caption{Values for the critical exponent $\gamma$ and amplitude $A$
650: at $p=0.4$ as function of trial $v_c$ estimates from the Baker-Hunter analysis.
651: From the clustering of different Pad{\'e} approximants in both pictures
652: we estimate $v_c=0.3217,$ $\gamma=0.966,$ and $A = 1.21$.}
653: \label{fig:4}
654: \end{figure}
655:
656: Using this method, our results for the critical exponent $\gamma$ are plotted in
657: Fig.~\ref{fig:6}. They show an effective exponent monotonically increasing
658: with $p$ but reaching a
659: plateau at $\gamma=1$ for dilutions between
660: $p=0.42$ and $p=0.46$. The following sharp increase is to be interpreted as
661: due to the crossover to the percolation fixed point $p_c=0.751\,188$, $T_c=0$,
662: where a $\chi \sim \exp(1/T)$ behavior is expected.
663:
664:
665: \begin{figure}
666: \includegraphics[scale=.33,angle=-90]{gam_p.ps}
667:
668: \caption{Effective critical exponent $\gamma$ as function of the dilution $p$
669: from Baker-Hunter analyses.}
670: \label{fig:6}
671: \end{figure}
672:
673: It is well known (see, e.g., Ref.~\cite{barma85})
674: that series analysis in crossover
675: situations is extremely difficult. If the parameter $p$ interpolates between
676: regions governed by different fixed points, the exponent obtained from a
677: finite number of terms of a series expansion must cross somehow
678: between its universal values, and does this usually quite slowly.
679: Therefore it does not come as a
680: surprise that the MC simulations quoted above see
681: the onset of a second order phase transition already for smaller values of
682: the disorder strength $p$.
683: The mere existence of a plateau in $\gamma_{\rm eff}(p)$, however,
684: is an indication that here truly critical behavior is seen.
685: It is governed by a
686: fixed point for which we obtain $\gamma=1.00(3)$.
687: Here, as always in series analyses, the error estimates the
688: scattering of different approximants.
689:
690: \section{Conclusions}
691: \label{sec:c}
692:
693: We have implemented a comprehensive toolbox for generating and enumerating
694: star graphs as required for high-temperature series expansions of quenched,
695: disordered systems.
696: Monte Carlo simulations of systems with quenched disorder require an enormous
697: amount of computing time because many realizations have to be simulated for
698: the quenched average. For this reason it is hardly possible to scan a whole
699: parameter range. Using high-temperature series expansions, on the other hand,
700: one can obtain this average exactly. Since the relevant parameters (degree
701: of disorder $p$, spatial dimension $d$, number of states $q$, etc.)
702: can be kept
703: as symbolic variables, the number of potential applications is very large.
704:
705: Here we presented an analysis of the three-dimensional bond-diluted 4-state
706: Potts model. The phase diagram confirms recent Monte Carlo results and, by
707: comparing with the numerical data, we also see signals for the onset of a
708: second-order transition at a finite disorder strength.
709:
710: \begin{acknowledgments}
711: Support by DFG grant No.~JA 483/17-1 and partial support from the
712: German-Israel-Foundation under
713: grant No.~I-653-181.14/1999 is gratefully acknowledged.
714: \end{acknowledgments}
715:
716:
717: \newpage
718: \onecolumngrid
719:
720:
721: \appendix*
722: \section{}
723: %\begin{widetext}
724: As an example, we publish here the inverse
725: susceptibility for the bond-diluted 4-state Potts model in $d\leq4 $
726: dimensions up to order $v^{18}$ ($P=1-p$): \\
727:
728: \small
729:
730: $\chi^{-1}(P,v,d) =
731: 1 - 2\,P\,v\,{d} + 2\,P^2\,v^2\,{d} -
732: 2\,P^3\,v^3\,{d} +
733: \left[ 2\,P^4\,{d} - 16\,P^4\,{d\choose 2} \right]\,v^4 +
734: \left[ -2\,P^5\,{d} + \left(24\,P^4 +
735: 72\,P^5\right){d\choose 2} \right]\,v^5 \\[.2cm] +
736: \left[ 2\,P^6\,{d} + \left(24\,P^4 -
737: 96\,P^5 - 248\,P^6\right){d\choose 2} - 768\,P^6\,{d\choose 3}
738: \right]\,v^6 \\[.2cm] +
739: \left[ -2\,P^7\,{d} + \left( 24\,P^4 -
740: 96\,P^5 + 264\,P^6 +
741: 640\,P^7\right) {d\choose 2} + \left(576\,P^6 +
742: 3264\,P^7\right){d\choose 3} \right]\,v^7 \\[.2cm] +
743: \left[ 2\,P^8\,{d} + \left(72\,P^4 -
744: 96\,P^5 + 264\,P^6-
745: 216\,P^7 - 1384\,P^8\right) {d\choose 2} +
746: \left(576\,P^6 - 144\,P^7 -
747: 22704\,P^8\right) {d\choose 3}
748: - 62208\,P^8\,{d\choose 4} \right]\,v^8 \\[.2cm] +
749: \left[ -2\,P^9\,{d} + \left(-72\,P^4 -
750: 288\,P^5 + 264\,P^6 -
751: 312\,P^7 - 1416\,P^8 +
752: 1888\,P^9\right) {d\choose 2} \right.\\ \left.\;\;\; + \left(576\,P^6 -
753: 720\,P^7 + 720\,P^8 +
754: 66944\,P^9\right) {d\choose 3} + \left(31104\,P^8 +
755: 221312\,P^9\right) {d\choose 4} \right]\,v^9 \\[.2cm] +
756: \left[ 2\,P^{10}\,{d} +\left(- 72\,P^4 +
757: 288\,P^5 + 648\,P^6 +
758: 384\,P^7 - 144\,P^8 +
759: 9336\,P^9 - 296\,P^{10}\right) {d\choose 2} \right.\\ \left.\;\;\;+
760: \left(576\,P^6 + 3456\,P^7 +
761: 19296\,P^8 + 75456\,P^9 -
762: 387168\,P^{10}\right){d\choose 3} + \left(31104\,P^8 +
763: 109440\,P^9 - 4000512\,P^{10}\right){d\choose 4} \right]\,v^{10}
764: \\[.2cm] +
765: \left[ -2\,P^{11}\,{d} +\left(- 72\,P^4 +
766: 288\,P^5 - 504\,P^6 -
767: 1008\,P^7 - 3024\,P^8 +
768: 3144\,P^9 - 33336\,P^{10} -
769: 9616\,P^{11}\right){d\choose 2} \right.\\ \left. \;\;\; + \left(576\,P^6 -
770: 1440\,P^7 - 8352\,P^8 -
771: 31248\,P^9 - 309744\,P^{10} +
772: 781824\,P^{11}\right){d\choose 3} \right.\\ \left.\;\;\;
773: + \left( 31104\,P^8 +
774: 11520\,P^9 + 635520\,P^{10} +
775: 10415872\,P^{11}\right){d\choose 4} \right] \, v^{11} \\[.2cm] +
776: \left[ 2\,P^{12}\,{d} +\left(- 216\,P^4 +
777: 288\,P^5 - 216\,P^6 -
778: 4272\,P^7 + 240\,P^8 +
779: 11856\,P^9 - 4968\,P^{10} +
780: 81744\,P^{11} + 37320\,P^{12}\right){d\choose 2} \right.\\ \left.\;\;\; +
781: \left(2880\,P^6 - 31392\,P^7 +
782: 14112\,P^8 + 169200\,P^9 +
783: 489024\,P^{10}
784: + 2692800\,P^{11} -
785: 5811664\,P^{12}\right){d\choose 3} \right.\\ \left. \;\;\;
786: + \left(31104\,P^8 +
787: 273024\,P^9 + 3204864\,P^{10} +
788: 16037760\,P^{11} - 179275648\,P^{12}\right){d\choose 4}
789: \right] \, v^{12}\\[.2cm] +
790: \left[ -2\,P^{13}\,{d} +
791: \left(216\,P^4 + 864\,P^5 -
792: 792\,P^6 - 3912\,P^7 +
793: 29736\,P^8 + 5952\,P^9 -
794: 20736\,P^{10} + 23088\,P^{11} \right.\right. \\ \left.\left.\;\;\; -
795: 144624\,P^{12} - 96160\,P^{13}\right){d\choose 2} \right. \\ \left. \;\;\; +
796: \left(-1728\,P^6 - 31536\,P^7 +
797: 162288\,P^8 + 15408\,P^9 -
798: 223344\,P^{10} - 113760\,P^{11} -
799: 8412192\,P^{12} + 5990784\,P^{13}\right){d\choose 3} \right. \\
800: \left. \;\;\; +
801: \left(31104\,P^8 + 67968\,P^9 +
802: 1022976\,P^{10} - 693504\,P^{11} -
803: 16255872\,P^{12} + 304010112\,P^{13}\right){d\choose 4}
804: \right]\,v^{13} \\[.2cm] +
805: \left[ 2\,P^{14}\,{d} +
806: \left(216\,P^4 - 864\,P^5 -
807: 1944\,P^6 - 8616\,P^7 +
808: 18360\,P^8 - 99600\,P^9 -
809: 65544\,P^{10} + 33936\,P^{11} \right.\right. \\ \left. \left. \;\;\; -
810: 86952\,P^{12} + 73704\,P^{13} +
811: 169400\,P^{14}\right){d\choose 2} \right. \\ \left. \;\;\;
812: + \left( - 1728\,P^6 -
813: 55152\,P^7 + 67248\,P^8 -
814: 894240\,P^9 - 918000\,P^{10} +
815: 2799648\,P^{11} + 8589744\,P^{12} \right.\right. \\ \left.\left.\;\;\; +
816: 58983984\,P^{13} - 98045424\,P^{14}\right){d\choose 3}
817: \right.\\ \left.\;\;\; +
818: \left(31104\,P^8 - 1057536\,P^9 +
819: 17280\,P^{10} + 24870528\,P^{11} +
820: 179980416\,P^{12} + 1095494784\,P^{13} -
821: 7487817088\,P^{14}\right) {d\choose 4} \right]\, v^{14}\\[.2cm] +
822: \left[ -2\,P^{15}\,{d} + \left(216\,P^4 -
823: 864\,P^5 + 1512\,P^6 -
824: 4536\,P^7 + 44568\,P^8 -
825: 55200\,P^9 + 168480\,P^{10} +
826: 363072\,P^{11} - 11832\,P^{12} \right.\right. \\ \left. \left. \;\;\;
827: + 530040\,P^{13} + 501600\,P^{14} -
828: 145632\,P^{15}\right){d\choose 2} \right.\\ \left.\;\;\;
829: + \left(- 1728\,P^6 -
830: 41040\,P^7 + 209232\,P^8 -
831: 609984\,P^9 + 1319328\,P^{10} +
832: 7874208\,P^{11} + 5670048\,P^{12} +
833: 24319296\,P^{13} \right.\right. \\ \left. \left. \;\;\;
834: - 141840288\,P^{14} -
835: 14817536\,P^{15}\right) {d\choose 3} \right.\\ \left.\;\;\; +
836: \left(31104\,P^8 -
837: 740736\,P^9 + 131328\,P^{10} +
838: 22334976\,P^{11} + 66366720\,P^{12} +
839: 188319744\,P^{13} \right.\right. \\ \left. \left. \;\;\;
840: - 1467511296\,P^{14} +
841: 5362518016\,P^{15}\right){d\choose 4} \right]\, v^{15}\\[.2cm] +
842: \left[ 2\,P^{16}\,{d} + \left(648\,P^4 -
843: 864\,P^5 + 1512\,P^6 +
844: 24336\,P^7 + 38496\,P^8 -
845: 148008\,P^9 + 1656\,P^{10} +
846: 117024\,P^{11} - 1325376\,P^{12} \right.\right. \\ \left.
847: \left. \;\;\; +
848: 7200\,P^{13} - 1644000\,P^{14} -
849: 2926176\,P^{15} - 373984\,P^{16}\right) {d\choose 2}
850: \right.\\ \left.\;\;\; + \left(
851: - 1728\,P^6 + 163296\,P^7 +
852: 105984\,P^8 - 2305728\,P^9 -
853: 1857888\,P^{10} - 6620544\,P^{11} -
854: 48148992\,P^{12} + 18163728\,P^{13} \right.\right. \\ \left.
855: \left. \;\;\;
856: + 118520640\,P^{14} + 1144225008\,P^{15} -
857: 1918717248\,P^{16}\right){d\choose 3} \right.\\ \left.\;\;\;
858: + \left(217728\,P^8 -
859: 3438720\,P^9 - 7119360\,P^{10} -
860: 64137600\,P^{11} - 149601024\,P^{12} +
861: 1152714240\,P^{13} + 8368094208\,P^{14} \right.\right. \\ \left.
862: \left. \;\;\;+
863: 58294742400\,P^{15} - 317165909504\,P^{16}\right) {d\choose 4}
864: \right] \, v^{16} \\[.2cm] +
865: \left[ -2\,P^{17}\,{d} + \left(-
866: 648\,P^4 - 2592\,P^5 +
867: 1512\,P^6 + 15408\,P^7 -
868: 161712\,P^8 - 123384\,P^9 + 66792\,P^{10}
869: + 39264\,P^{11} \right.\right. \\ \left.\left.\;\;\; -
870: 1976760\,P^{12} + 3413424\,P^{13} +
871: 848256\,P^{14} + 4241568\,P^{15} +
872: 8541960\,P^{16} + 2398960\,P^{17}\right){d\choose 2} \right. \\
873: \left. \;\;\; + \left(
874: - 1728\,P^6 + 109728\,P^7 -
875: 984096\,P^8 - 1584432\,P^9 +
876: 413424\,P^{10} - 11287296\,P^{11} -
877: 32069376\,P^{12} + 206240976\,P^{13} \right.\right. \\ \left.\left.\;\;\; +
878: 380730960\,P^{14} + 1087235856\,P^{15} -
879: 1859056704\,P^{16} - 2643006384\,P^{17}\right){d\choose 3}
880: \right. \\ \left. \;\;\; + \left(
881: - 93312\,P^8 - 2522880\,P^9 -
882: 4468608\,P^{10} - 63930240\,P^{11} -
883: 127255680\,P^{12} + 1693209600\,P^{13} +
884: 6161021568\,P^{14} \right.\right. \\ \left.\left.\;\;\; + 23385824256\,P^{15} -
885: 50368269312\,P^{16} - 105383991680\,P^{17}\right) {d\choose 4}
886: \right]\, v^{17}\\[.2cm] +
887: \left[ 2\,P^{18}\,{d} + \left(- 648\,P^4 +
888: 2592\,P^5 + 4104\,P^6 +
889: 83520\,P^7 - 82080\,P^8 +
890: 465984\,P^9 + 586080\,P^{10} +
891: 605064\,P^{11} - 166248\,P^{12} \right.\right. \\ \left.\left.\;\;\; +
892: 8121312\,P^{13} - 4714536\,P^{14} -
893: 2886168\,P^{15} - 3604536\,P^{16} -
894: 20651832\,P^{17}- 7297424\,P^{18}\right){d\choose 2}
895: \right.\\ \left.\;\;\; + \left(
896: -8640\,P^6 + 518400\,P^7 -
897: 492480\,P^8 + 3752592\,P^9 +
898: 12513744\,P^{10} - 23522256\,P^{11} -
899: 66704640\,P^{12} - 53106912\,P^{13} \right.\right. \\ \left.\left.\;\;\;
900: - 884626272\,P^{14} - 122168448\,P^{15} +
901: 2205877392\,P^{16} + 22700601216\,P^{17} -
902: 42014019168\,P^{18}\right){d\choose 3} \right.\\ \left.\;\;\;
903: + \left(- 93312\,P^8 +
904: 657024\,P^9 + 23180544\,P^{10} -
905: 169350912\,P^{11} - 762268032\,P^{12} -
906: 3977024256\,P^{13} - 10126195200\,P^{14} \right.\right. \\
907: \left.\left.\;\;\;
908: + 47139877632\,P^{15} +
909: 379559824128\,P^{16} +
910: 2866361546496\,P^{17} +
911: 3747410465664\,P^{18}\right){d\choose 4} \right]\, v^{18}
912: $
913: %\end{widetext}
914: \twocolumngrid
915: %\bibliography{star}
916: %\bibliographystyle{apsrev}
917: \begin{thebibliography}{21}
918: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
919: \expandafter\ifx\csname bibnamefont\endcsname\relax
920: \def\bibnamefont#1{#1}\fi
921: \expandafter\ifx\csname bibfnamefont\endcsname\relax
922: \def\bibfnamefont#1{#1}\fi
923: \expandafter\ifx\csname citenamefont\endcsname\relax
924: \def\citenamefont#1{#1}\fi
925: \expandafter\ifx\csname url\endcsname\relax
926: \def\url#1{\texttt{#1}}\fi
927: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
928: \providecommand{\bibinfo}[2]{#2}
929: \providecommand{\eprint}[2][]{\url{#2}}
930:
931: \bibitem[{\citenamefont{Domb and Green}(1974)}]{domb3}
932: \bibinfo{editor}{\bibfnamefont{C.}~\bibnamefont{Domb}} \bibnamefont{and}
933: \bibinfo{editor}{\bibfnamefont{M.~S.} \bibnamefont{Green}}, eds.,
934: \emph{\bibinfo{title}{Series Expansions for Lattice Models}},
935: vol.~\bibinfo{volume}{3} of \emph{\bibinfo{series}{Phase Transitions and
936: Critical Phenomena}} (\bibinfo{publisher}{Academic Press},
937: \bibinfo{address}{New York}, \bibinfo{year}{1974}).
938:
939: \bibitem[{\citenamefont{Harris}(1974)}]{harris}
940: \bibinfo{author}{\bibfnamefont{A.~B.} \bibnamefont{Harris}},
941: \bibinfo{journal}{Journal of Physics C: Solid State Physics}
942: \textbf{\bibinfo{volume}{7}}, \bibinfo{pages}{1671} (\bibinfo{year}{1974}).
943:
944: \bibitem[{\citenamefont{Aizenman and Wehr}(1989)}]{aiz}
945: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Aizenman}} \bibnamefont{and}
946: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Wehr}},
947: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{62}},
948: \bibinfo{pages}{2503} (\bibinfo{year}{1989}).
949:
950: \bibitem[{\citenamefont{Cardy and Jacobsen}(1997)}]{card97}
951: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Cardy}} \bibnamefont{and}
952: \bibinfo{author}{\bibfnamefont{J.~L.} \bibnamefont{Jacobsen}},
953: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{79}},
954: \bibinfo{pages}{4063} (\bibinfo{year}{1997}).
955:
956: \bibitem[{\citenamefont{Ballesteros et~al.}(2000)\citenamefont{Ballesteros,
957: Fern{\'a}ndez, Mart{\'\i}n-Mayor, Mu{\~n}oz~Sudupe, Parisi, and
958: Ruiz-Lorenzo}}]{balles00}
959: \bibinfo{author}{\bibfnamefont{H.~G.} \bibnamefont{Ballesteros}},
960: \bibinfo{author}{\bibfnamefont{L.~A.} \bibnamefont{Fern{\'a}ndez}},
961: \bibinfo{author}{\bibfnamefont{V.}~\bibnamefont{Mart{\'\i}n-Mayor}},
962: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Mu{\~n}oz~Sudupe}},
963: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Parisi}}, \bibnamefont{and}
964: \bibinfo{author}{\bibfnamefont{J.~J.} \bibnamefont{Ruiz-Lorenzo}},
965: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{61}},
966: \bibinfo{pages}{3215} (\bibinfo{year}{2000}).
967:
968: \bibitem[{\citenamefont{Chatelain et~al.}(2001)\citenamefont{Chatelain, Berche,
969: Janke, and Berche}}]{chat01a}
970: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Chatelain}},
971: \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Berche}},
972: \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Janke}}, \bibnamefont{and}
973: \bibinfo{author}{\bibfnamefont{P.~E.} \bibnamefont{Berche}},
974: \bibinfo{journal}{Phys. Rev. E} \textbf{\bibinfo{volume}{64}},
975: \bibinfo{pages}{036120} (\bibinfo{year}{2001}).
976:
977: \bibitem[{\citenamefont{Cardy}(1999)}]{card99}
978: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Cardy}},
979: preprint \eprint{cond-mat/9911024}.
980:
981: \bibitem[{\citenamefont{Turban}(1980)}]{turb}
982: \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{Turban}},
983: \bibinfo{journal}{Journal of Physics C: Solid State Physics}
984: \textbf{\bibinfo{volume}{13}}, \bibinfo{pages}{L13} (\bibinfo{year}{1980}).
985:
986: \bibitem[{\citenamefont{Butera and Comi}()}]{butera02}
987: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Butera}} \bibnamefont{and}
988: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Comi}},
989: preprint \eprint{hep-lat/0204007}.
990:
991: \bibitem[{\citenamefont{Singh and Chakravarty}(1987)}]{singh87}
992: \bibinfo{author}{\bibfnamefont{R.~R.~P.} \bibnamefont{Singh}} \bibnamefont{and}
993: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Chakravarty}},
994: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{36}},
995: \bibinfo{pages}{546} (\bibinfo{year}{1987}).
996:
997: \bibitem[{\citenamefont{Rapaport}(1972{\natexlab{a}})}]{rap1}
998: \bibinfo{author}{\bibfnamefont{D.~C.} \bibnamefont{Rapaport}},
999: \bibinfo{journal}{Journal of Physics C: Solid State Physics}
1000: \textbf{\bibinfo{volume}{5}}, \bibinfo{pages}{1830, 2813}
1001: (\bibinfo{year}{1972}{\natexlab{a}}).
1002:
1003: \bibitem[{\citenamefont{Martin}(1974)}]{martin74}
1004: \bibinfo{author}{\bibfnamefont{J.~L.} \bibnamefont{Martin}}, in \cite{domb3},
1005: pp. \bibinfo{pages}{97--112}.
1006:
1007: \bibitem[{\citenamefont{McKay}(1981)}]{mckay81}
1008: \bibinfo{author}{\bibfnamefont{B.~D.} \bibnamefont{McKay}},
1009: \bibinfo{journal}{Congressus Numerantium} \textbf{\bibinfo{volume}{30}},
1010: \bibinfo{pages}{45} (\bibinfo{year}{1981}),
1011: \bibinfo{note}{http://cs.anu.edu.au/{$\sim$}bdm/nauty/}.
1012:
1013: \bibitem[{\citenamefont{Press et~al.}(1992)\citenamefont{Press, Teukolsky,
1014: Vetterling, and Flannery}}]{numrec}
1015: \bibinfo{author}{\bibfnamefont{W.~H.} \bibnamefont{Press}},
1016: \bibinfo{author}{\bibfnamefont{S.~A.} \bibnamefont{Teukolsky}},
1017: \bibinfo{author}{\bibfnamefont{W.~T.} \bibnamefont{Vetterling}},
1018: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{B.~P.}
1019: \bibnamefont{Flannery}}, \emph{\bibinfo{title}{Numerical Recipes in C}}
1020: (\bibinfo{publisher}{Cambridge University Press},
1021: \bibinfo{address}{Cambridge},
1022: \bibinfo{year}{1992}).
1023:
1024: \bibitem[{\citenamefont{Guttmann}(1989)}]{guttmann89}
1025: \bibinfo{author}{\bibfnamefont{A.~J.} \bibnamefont{Guttmann}}, in
1026: vol.~\bibinfo{volume}{13} of
1027: \emph{\bibinfo{series}{Phase Transitions and Critical Phenomena}},
1028: edited by
1029: \bibinfo{editor}{\bibfnamefont{C.}~\bibnamefont{Domb}} \bibnamefont{and}
1030: \bibinfo{editor}{\bibfnamefont{J.~L.} \bibnamefont{Lebowitz}},
1031: (\bibinfo{publisher}{Academic Press}, \bibinfo{address}{New York},
1032: \bibinfo{year}{1989}), pp. \bibinfo{pages}{1--234}.
1033:
1034: \bibitem[{\citenamefont{Fisher and Chen}(1982)}]{pda80}
1035: \bibinfo{author}{\bibfnamefont{M.~E.} \bibnamefont{Fisher}} \bibnamefont{and}
1036: \bibinfo{author}{\bibfnamefont{J.-H.} \bibnamefont{Chen}}, in
1037: \emph{\bibinfo{booktitle}{Phase Transitions: Carg{\`e}se 1980}}, edited by
1038: \bibinfo{editor}{\bibfnamefont{M.}~\bibnamefont{L{\'e}vy}},
1039: \bibinfo{editor}{\bibfnamefont{J.~C.} \bibnamefont{Le~Guillou}},
1040: \bibnamefont{and}
1041: \bibinfo{editor}{\bibfnamefont{J.}~\bibnamefont{Zinn-Justin}}
1042: (\bibinfo{publisher}{Plenum}, \bibinfo{address}{New York},
1043: \bibinfo{year}{1982}), pp. \bibinfo{pages}{169--216}.
1044:
1045: \bibitem[{\citenamefont{Salman and Adler}(1997)}]{adler97}
1046: \bibinfo{author}{\bibfnamefont{Z.}~\bibnamefont{Salman}} \bibnamefont{and}
1047: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Adler}},
1048: \bibinfo{journal}{Journal of Physics A: Mathematical and General}
1049: \textbf{\bibinfo{volume}{30}}, \bibinfo{pages}{1979} (\bibinfo{year}{1997}).
1050:
1051: \bibitem[{\citenamefont{Lorenz and Ziff}(1998)}]{lorenz98}
1052: \bibinfo{author}{\bibfnamefont{C.~D.} \bibnamefont{Lorenz}} \bibnamefont{and}
1053: \bibinfo{author}{\bibfnamefont{R.~M.} \bibnamefont{Ziff}},
1054: \bibinfo{journal}{Phys. Rev. E} \textbf{\bibinfo{volume}{57}},
1055: \bibinfo{pages}{230} (\bibinfo{year}{1998}).
1056:
1057: \bibitem[{\citenamefont{Adler et~al.}(1991)\citenamefont{Adler, Aharony,
1058: Harris, Klein, and Meir}}]{adler91}
1059: \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{Klein}},
1060: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Adler}},
1061: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Aharony}},
1062: \bibinfo{author}{\bibfnamefont{A.~B.} \bibnamefont{Harris}},
1063: \bibnamefont{and}
1064: \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Meir}},
1065: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{43}},
1066: \bibinfo{pages}{11249} (\bibinfo{year}{1991}).
1067:
1068: \bibitem[{\citenamefont{George A.~Baker and Hunter}(1973)}]{bakerhunter}
1069: \bibinfo{author}{\bibfnamefont{G.~A.}~\bibnamefont{Baker}}
1070: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{D.~L.}
1071: \bibnamefont{Hunter}}, \bibinfo{journal}{Phys. Rev. B}
1072: \textbf{\bibinfo{volume}{7}}, \bibinfo{pages}{3377} (\bibinfo{year}{1973}).
1073:
1074: \bibitem[{\citenamefont{Barma and Fisher}(1985)}]{barma85}
1075: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Barma}} \bibnamefont{and}
1076: \bibinfo{author}{\bibfnamefont{M.~E.} \bibnamefont{Fisher}},
1077: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{31}},
1078: \bibinfo{pages}{5954} (\bibinfo{year}{1985}).
1079:
1080: \end{thebibliography}
1081:
1082: \end{document}
1083: