1: \documentclass{elsart}
2: \usepackage{epsfig}
3: \usepackage{amssymb}
4: \begin{document}
5: \begin{frontmatter}
6:
7: \title{Novel order parameter to describe
8: the critical behavior of Ising spin glass models}
9:
10: \author{F. Rom\'a, F. Nieto, A. J. Ramirez-Pastor\corauthref{cor1}}
11: \address{Departamento de F\'{\i}sica, Universidad Nacional de San Luis - CONICET. Chacabuco 917, 5700 San Luis,
12: Argentina. \\
13: E-mail: froma@unsl.edu.ar, fnieto@unsl.edu.ar, antorami@unsl.edu.ar}
14:
15: \author{E.E. Vogel}
16:
17: \address{Departamento de F\'{\i}sica, Universidad de La
18: Frontera, Casilla 54-D, Temuco, Chile. \\ E-mail:
19: ee{\_}vogel@ufro.cl}
20:
21: \thanks[cor1]{Corresponding author. Fax +54-2652-430224, E-mail: antorami@unsl.edu.ar}
22:
23: \begin{abstract}
24:
25: A novel order parameter $\Phi$ for spin glasses is defined based
26: on topological criteria and with a clear physical interpretation.
27: $\Phi$ is first investigated for well known magnetic systems and
28: then applied to the Edwards-Anderson $\pm J$ model on a square
29: lattice, comparing its properties with the usual $q$ order
30: parameter. Finite size scaling procedures are performed. Results
31: and analyses based on $\Phi$ confirm a zero temperature phase
32: transition and allow to identify the low temperature phase. The
33: advantages of $\Phi$ are brought out and its physical meaning is
34: established.
35:
36: \end{abstract}
37:
38:
39: \begin{keyword}
40: Lattice theory and statistics \sep Spin-glass and other random
41: models \sep Phase transitions \PACS 05.50.+q \sep 75.10.Nr \sep
42: 05.70.Fh
43: \end{keyword}
44: \end{frontmatter}
45: \parskip 0cm
46:
47: \section{Introduction}
48:
49: Despite over three decades of intensive work, the nature of the
50: low temperature phase of two-dimensional Edwards-Anderson (EA)
51: \cite{EA} model for spin glasses remains controversial. It is
52: agreed that a phase transition occurs at zero temperature for a
53: Gaussian distribution of bonds (GD)
54: \cite{McMillan,Bray,Rieger,Hartmann}. Similarly, for a symmetric
55: $\pm J$ distribution or bimodal distribution (BD) of bonds, very
56: convincing numerical evidence has been found that there is no
57: transition at finite temperature
58: \cite{Hartmann,Bhatt,Kawashima,Houdayer,Hartmann2,Amoruso,Lukic}.
59: In most of these references, the authors do not use an order
60: parameter for characterizing the phase transition. On the other
61: hand, data arising from other contributions, which are based on
62: the behavior of a standard overlapping order parameter, support
63: the existence of a finite critical temperature
64: \cite{Shirakura,Shiomi,Matsubara}.
65:
66: In this context, the main purposes of this paper are the
67: following: a) To show that the disagreement pointed out in
68: previous paragraph is related to the non-zero overlap of
69: site-order parameters obtained for quite distinct energy valleys;
70: b) To overcome this situation by proposing here a novel order
71: parameter $\Phi$, which is quite drastic to characterize phases
72: but still is general enough to coincide with usual descriptions of
73: ferromagnetic (F) and antiferromagnetic (AF) systems; c) To apply
74: $\Phi$ to do a scaling analysis for two-dimensional EA systems
75: including Binder cumulant \cite{Binder01}; d) To confirm the
76: assumption of the zero-temperature phase transition for
77: two-dimensional BD, thus reinforcing this result obtained by
78: previously quoted authors; and e) To give a physical meaning to
79: this result by using the grounds on which $\Phi$ is based on.
80:
81: The present work is organized as it follows. In Section 2, we
82: introduce the model and define a novel order parameter, $\Phi$,
83: very useful for spin glasses and other frustrated systems. Results
84: of the simulation are presented in Section 3. Finally, our
85: conclusions are drawn in Section 4.
86:
87: \section{Model and basic definitions}
88:
89: Let us begin by very briefly introducing the system under study.
90: Ising spin $s_i$ occupies $i-th$ site of a two dimensional (square
91: for simplicity) lattice. The interaction with the spin at site $j$
92: is mediated by exchange interaction $J_{ij}$. In the absence of
93: magnetic field (which is the case for the scope of the present
94: paper) the Hamiltonian of such system can then be written as
95: \begin{equation}
96: H = \sum_{\langle i,j \rangle} J_{ij} s_is_j \; ,
97: \end{equation}
98: where interactions $\{J_{ij}\}$ are restricted to nearest neighbor
99: couplings. In the ferromagnetic (F) Ising model, $J_{ij}=-J$
100: $\forall$ $\langle i,j \rangle$. For the EA model, we will
101: consider half of the bonds F, while the other half will be
102: described by antiferromagnetic (AF) bonds of the same magnitude,
103: namely, $J_{ij}=+J$ ($J>0$). A sample is one of the possible
104: random distributions of these mixed bonds. For simplicity spins
105: take values $s_j=\pm 1$, which can be equally denoted by their
106: signs.
107:
108: %A state of the system is a collection of ordered spin orientations
109: %according to a prescribed enumeration order including the $N$
110: %spins of the lattice. Notice the trivial degeneracy that arises by
111: %the simultaneous inversion of all spins.
112:
113: Now, let us consider a configuration $\alpha$ defined by a
114: collection of ordered spin orientations $\{s^{\alpha}_j\}$. The
115: usual EA order parameter $q$ is built up by means of overlaps
116: between two configurations $\alpha$ and $\beta$ and takes the form
117: \begin{equation}
118: q_{\alpha\beta}=\frac{1}{N}\sum _{j=1} ^{N} s_j ^{\alpha}s_j
119: ^{\beta}, \label{eqqab}
120: \end{equation}
121: where $N$ ($\equiv L \times L$) is the total number of spins.
122:
123: For models in which the ground state is non degenerate after
124: breaking ergodicity, such as the pure F case, the distribution of
125: $q_{\alpha\beta}$ values for the ground manifold (T=0.0) is
126: trivial and it is given by delta functions at $q_{\alpha\beta} =
127: 1.0$ and $q_{\alpha\beta}=-1.0$. This also happens in general for
128: all systems with non-degenerate ground level. But this also
129: applies to GD, where local fields have all different values at
130: different sites, leading to a true minimum energy for just one
131: pair of opposite ground states. However, for the BD the local
132: field assumes a few discrete values only, which necessarily means
133: highly degenerate ground manifolds leading to $|q_{\alpha\beta}|
134: <1.0$, for a large number of possible pairs of ground states. This
135: distribution will have two broad symmetric maxima but it will not
136: vanish in the intermediate region \cite{Shiomi}.
137:
138: On the other hand, a more detailed description based on a
139: topological picture of the ground state of BD was presented
140: \cite{Vogel01,Vogel02}. This framework allows us to define a state
141: function with a clear physical meaning, which is a good candidate
142: to be a new order parameter for a phase transition. In fact, it
143: has been reported an important feature of the ground state,
144: namely, at $T=0$ there exist clusters of solidary spins (CSS)
145: preserving the magnetic memory of the system (solidary spins
146: maintain their relative orientation for all states of the ground
147: manifold)\cite{Barahona1982}. The main idea of this work is to
148: characterize the nature of the low temperature phase through the
149: CSS.
150:
151: Let us consider a particular sample of any given size $N$. We
152: denote by $\Gamma_{\kappa}$ any of the $n$ CSS of the sample
153: ($\kappa$ runs from $1$ to $n$). Calculations begin recognizing
154: all of the CSS of each sample belonging to a set of $2000$
155: randomly generated samples of each size. This process is closely
156: related to finding the so-called ``diluted lattice" that prevails
157: after removing all frustrated bonds \cite{Valdes98}, so the
158: algorithms designed for that purpose can also be used here.
159:
160: Let us first pick any arbitrary ground state configuration denoted
161: by an asterisk ($*$) fixing one of the two possible relative
162: orientations of the CSS, thus becoming a reference configuration.
163: Then a local overlap corresponding to the configuration $\alpha$
164: in the ${\kappa}$-th cluster, of size $N_{\kappa}$, can be defined
165: as
166:
167: \begin{equation}
168: \phi _{\kappa}^{\alpha} =\frac{1}{N_{\kappa}} \sum _{j \in \Gamma
169: _{\kappa}} s^{*}_{j} s_j^{\alpha}, \label{lop}
170: \end{equation}
171: where the sum runs over all spins in the cluster $\Gamma_{\kappa}$
172: only. Thus, $\vert \phi_{\kappa}^{\alpha} \vert =1$ indicates a
173: fully ordered cluster; otherwise $\vert \phi_{\kappa}^{\alpha}
174: \vert < 1$. The magnetic order of the sample is characterized by
175: the set of overlaps, namely, $\{ \phi_{\kappa}^{\alpha} \}$. Under
176: the occurrence of a phase transition, the new set $\{
177: \phi_{\kappa}^{\alpha} \}$ will determine uniquely the ergodic
178: component of the reached phase. This fact is a required
179: characteristic for a well behaved order parameter \cite{Enter84}.
180:
181: We are now ready to define the new order parameter introduced in
182: this paper. It is given by
183: \begin{equation}
184: \Phi_{\alpha} = \sum _{{\kappa}=1}^n f_{\kappa} \vert
185: \phi_{\kappa}^{\alpha} \vert, \label{op}
186: \end{equation}
187: where $f_{\kappa}=N_{\kappa}/N_I$, being $N_I=\sum_{{\kappa}=1}^n
188: N_{\kappa}$, ($N_I\le N$). From the definition it flows that for
189: $T \geq T_c$ the average value of $\Phi_{\alpha}$, namely, $\Phi$,
190: should be $0$. Similarly, for $T < T_c$ it should hold that
191: $0<\Phi \leq 1$, being $\Phi = 1$ for $T=0$ only. It is important
192: to emphasize that $\Phi_{\alpha}$ is a state function, which is an
193: advantage over $q_{\alpha\beta}$ defined in eq. (\ref{eqqab}) as
194: an overlap between two configurations of the system.
195:
196:
197:
198: The calculation of the new order parameter $\Phi_{\alpha}$
199: requires the previous determination of the set of CSS for each
200: considered sample. This procedure, which was performed by using
201: the numerical scheme introduced in Ref.\cite{Ramirez04}, is a
202: computational limitation for going to larger system sizes. Once
203: the ground manifold of each sample is completely characterized
204: after this procedure, the numerical calculations converge very
205: quickly by flipping the spins not present in the largest CSS only.
206: The second run on each sample takes much less time than the first
207: one that is needed to find all CSS.
208:
209: In the F Ising model there is a unique cluster of $N$ solidary
210: parallel spins at $T=0$. As it can be trivially demonstrated, eq.
211: (\ref{op}) leads to the magnetization per spin, which is the
212: natural order parameter of such system. Similarly, for the AF case
213: we get the well-known order parameter defined as the magnetization
214: difference between the two possible interpenetrating sublattices.
215: Finally, for GD, there also exists an unique CSS in the ground
216: state and the phase transition occurring in the system is
217: completely described by the new order parameter, eq.(\ref{op}), as
218: well. So the new parameter retains all the properties of the
219: well-known non-degenerate systems.
220:
221: Finally, the reduced fourth-order cumulant, introduced by Binder
222: \cite{Binder01}, can be calculated as
223: \begin{equation}
224: U_L=1-\frac{\left[ \langle m^4 \rangle \right]}{3\left[ \langle
225: m^2 \rangle \right] ^2}, \label{Bcumulant}
226: \end{equation}
227: where $m$ is a given order parameter, and $\langle \ldots \rangle$
228: and $[ \ldots ]$ mean the spin configuration (thermal) average and
229: the bond configuration (sample) average, respectively. In general,
230: the structure of the distribution of $m$ affects the behavior of
231: the fourth-order Binder cumulant. Thus, for a trivial
232: distribution, both $|m| \to \pm 1$ and $U_L \to 2/3$, as $T$ goes
233: to zero. On the other hand, if the distribution is nontrivial,
234: $|m|$ tends to a value $m^o$ lower than 1, while $U_L$ tends to a
235: value $U_L^o$ lower than $2/3$ upon decreasing temperature.
236:
237:
238:
239: \section{Results}
240:
241:
242: Distribution functions for $q_{\alpha \beta}$ and
243: $\Phi_{\alpha}$, were obtained for BD by using a standard
244: simulated-tempering procedure\footnote{It must be emphasized that
245: it is not necessary a simulated-tempering scheme for calculating
246: the new order parameter, $\Phi$.} \cite{marinari,kerler} along
247: with the well known Glauber's dynamics \cite{kawa72}. For
248: illustration purposes, we perform calculations on $1000$ samples
249: of size 64 $(8 \times 8)$ at different temperatures ranging from
250: $T=0.2$ to $T=1.0$. (throughout this paper, $k_B/J=1$ without any
251: loss of generality). The results corresponding to $T=0.31$,
252: $T=0.53$ and $T=0.69$ are presented in Fig. 1. As it is shown in
253: part (a), the distribution of the new order parameter,
254: $R(\Phi_{\alpha})$, exhibits a drastic behavior as $T$ decreases.
255: In part (b) it is shown how the corresponding curves for
256: $r(|q_{\alpha \beta}|)$ have a broad maximum over the plotted
257: range and $r(0)>0$. These undesired characteristics for this order
258: parameter remain even at low temperatures.
259:
260: In Fig. 2, $U_L(T)$, built up from $R(\Phi_{\alpha})$
261: distribution, is presented for different lattice sizes ranging
262: from $N=16$ to $N=144$ and each point was calculated by averaging
263: over a set of 2000 samples. With the help of the inset, it is
264: observed that the curves do not intersect each other as a direct
265: indication of the absence of a phase transition for finite
266: temperature, at least for the sizes considered here. Eventually we
267: are not free from finite size considerations yet as it has been
268: recently proposed that at least samples with $L=50$ should be
269: reached when conventional parameters are used \cite{Katzgraber05}.
270: However, using a more drastic parameter like the one proposed
271: here, a faster convergence towards large $L$ values is expected.
272: It is clear that all curves go to $2/3$ as $T \to 0$, which
273: reinforces the robustness of eq.(\ref{op}). On the other hand,
274: this property is not followed by cumulants obtained from other
275: overlapping order parameters. This is the case of Fig. 7 in
276: Ref.\cite{Shiomi}, where it is possible to think that the reported
277: crossing of the cumulants of $q$ arises from the dependence of
278: $U_L^o$ on size. In this contribution, the authors reported a
279: critical temperature different from zero, $T_c \approx 0.23$.
280:
281: Finite-size scaling \cite{Binder01} predicts that all curves in a
282: figure such as Fig. 2, should collapse onto a single one when
283: using $(T-T_c)L^{1/\nu}$ as independent variable, being $T_c$ the
284: critical temperature for the transition and $\nu$ an appropriate
285: critical exponent. Upon choosing $T_c=0$ and the exponent $\nu$ is
286: taken as $\nu=2.63 \pm 0.20$, the standard universal behavior for
287: $U_L(T)$ is obtained as shown in Fig. 3. This is an independent
288: confirmation of previously reported results
289: \cite{Bhatt,Young01,Cheung,McMillan01}.
290:
291:
292: The following two parameters were also measured as each
293: sample was solved exactly: (a) The mean fraction of spins, $P
294: \equiv \left[ N_I \right] /N$, belonging to any CSS; and (b) The
295: fraction of spins in the largest CSS $p\equiv \left[ N_{\ell}
296: \right] /N$, where $N_{\ell}$ is the number of spins in the
297: largest CSS. Fig. 4(a) shows that while $P$ remains rather
298: constant, $p$ clearly decreases with size and the stabilizing role
299: of the largest CSS is lost. The average number of CSS $[n]$ as
300: function of size was also measured, finding that $[n]$ grows
301: linearly with $N$, as it is shown in Fig. 4(b). For $N>49$, say
302: (when small size effects do not play an important role), the
303: following approximate law is obtained $[n]\approx 0.03N+0.60$.
304:
305:
306: The size dependence of a possible spin-glass phase can be
307: described in the following terms. For small sizes ($N<49$ say)
308: most of solidary spins are grouped in one large cluster
309: stabilizing a spin-glass phase. As size grows, the number of CSS
310: increases linearly with $N$ (or quadratically with $L$), while the
311: relative size of the largest CSS diminishes. This can be
312: visualized as if the original lattice would break into portions of
313: relatively smaller sizes, none of them large enough to stabilize a
314: spin-glass phase. This is the reason for the numeric result of
315: Fig. 2, showing no intersection of curves for different sizes.
316:
317: If the same procedure used here for the symmetric case is applied
318: to different concentrations of F and AF bonds, a stable phase is
319: found in the extremes of high and low concentrations of F bonds in
320: correspondence with results already reported in the literature
321: \cite{Mor01,Mor02}. As the relative concentration of F bonds
322: varies the behavior of $P$ and $[n]$ is very similar to that shown
323: in Fig.~4. However, $p$ tends to be constant for very asymmetric
324: distributions of $\pm J$ bonds. The last statement indicates the
325: presence of an infinite CSS in the thermodynamical limit, which is
326: associated to a stable phase. These results are not shown
327: graphically in the present paper.
328:
329: \section{Conclusions}
330:
331: A new order parameter $\Phi$ has been introduced and applied to
332: the study of magnetic systems. It proves to be particularly
333: essential for characterizing degenerate systems such as Ising-like
334: models with bimodal distribution. Parameter $\Phi$ is well
335: behaved, having all desired properties for a drastic order
336: parameter. This behavior is based on the properties of CSS. When
337: this order parameter is used for systems with BD, properties
338: similar to order parameters for non-degenerate systems are found.
339: Then, the characterization of magnetic phases after using the
340: scaling techniques of cumulants becomes unambiguous. In this way
341: it was shown that the two-dimensional Edwards-Anderson model
342: exhibits a phase transition at $T_c=0$, with a critical exponent
343: $\nu=2.63 \pm 0.20$.
344:
345: The identification of all CSS for each sample is the bottleneck in
346: the present computational scheme. This procedure is very time
347: consuming for large lattice sizes. The extra time needed for
348: finding all CSS is well paid by the better precision achieved in
349: the characterization of the phase, and the elimination of overlaps
350: in the new order parameter, thus making the identification of the
351: ergodic valley reliable.
352:
353:
354: The characteristics of this new order parameter make it also
355: useful for other frustrated systems, where large overlaps occur
356: due to the complex energy valley. The extension of the use of the
357: parameter $\Phi$ to other kind of problems is clearly foreseen.
358: For instance, it can be the key element $i)$ to describe the phase
359: diagram for the asymmetric distribution problem around the
360: critical concentration of ferromagnetic (or antiferromagnetic)
361: bonds of $\approx 0.1$ and $ii)$ to study the critical behavior of
362: 3D Ising spin glasses. This task requires serious improvements in
363: the numerical techniques used in order to get access to large
364: lattice sizes. Work along this line is in progress.
365:
366:
367:
368:
369: \ \
370:
371: \noindent{\bf Acknowledgments}
372:
373: \ \
374:
375: We thank Fondecyt (Chile) under projects 1020993 and 7020993. One
376: author (EEV) thanks Millennium Scientific Initiative (Chile) under
377: contract P-02-054-F for partial support. Three authors (FR, FN and
378: AJRP) thank CONICET (Argentina) and the Universidad Nacional de
379: San Luis (Argentina) under project 322000.
380:
381: \newpage
382:
383: \begin{thebibliography}{9}
384:
385: \bibitem{EA} S.F. Edwards and P.W. Anderson, J. Phys. F \bf 5, \rm 965 (1975); G. Toulouse, Commun.
386: Phys. \bf 2, \rm 115 (1977).
387: \bibitem{McMillan} W.L. McMillan, Phys. Rev. B \bf 30, \rm R476 (1984).
388: \bibitem{Bray} A.J. Bray and M.A. Moore, J. Phys. C \bf 17, \rm L463 (1984); A.J. Bray
389: and M.A. Moore, in \em Heidelberg Colloquium on Glassy Dynamics,
390: \rm edited by J.L. van Hemmen and I. Morgenstern (Springer-Verlag,
391: Heidelberg, 1987).
392: \bibitem{Rieger} H. Rieger, L. Santen, U. Blasum-U, M. Diehl, M. J\"unger, and G.
393: Rinaldi, J. Phys. A \bf 29, \rm 3939 (1996).
394: \bibitem{Hartmann} A. K. Hartmann and A. P. Young, Phys. Rev. B \bf 64, \rm 180404(R) (2001).
395: \bibitem{Bhatt} R. N. Bhatt and A. P. Young, Phys. Rev. B \bf 37, \rm 5606 (1988).
396: \bibitem{Kawashima} N. Kawashima and H. Rieger, Europhys. Lett. \bf 39, \rm 85 (1997).
397: \bibitem{Houdayer} J. Houdayer, Eur. Phys. J. B \bf 22, \rm 479 (2001).
398: \bibitem{Hartmann2} A. K. Hartmann, A. J. Bray, A. C. Carter, M. A. Moore, A. P. Young, Phys. Rev. B \bf 66 \rm 224401 (2002).
399: \bibitem{Amoruso} C. Amoruso, E. Marinari, O. C. Martin, A. Pagnani, Phys. Rev. Lett. \bf 91, \rm 087201 (2003).
400: \bibitem{Lukic} J. Lukic, A. Galluccio, E. Marinari, O. C. Martin, G. Rinaldi, Phys. Rev. Lett. \bf 92, \rm 117202 (2004).
401: \bibitem{Shirakura} T. Shirakura and F. Matsubara, Phys. Rev. Lett. \bf 79, \rm 2887 (1997).
402: \bibitem{Shiomi} M. Shiomi, F. Matsubara and T. Shirakura, J. Phys. Soc. Jpn. \bf 69, \rm 2798 (2000).
403: \bibitem{Matsubara} F. Matsubara, T. Shirakura and M. Shiomi, Phys. Rev. B \bf 58, \rm R11821 (1998).
404: \bibitem{Binder01} K. Binder, Rep. Prog. Phys. \bf 60, \rm 488 (1997).
405: \bibitem{Vogel01} E.E. Vogel, S. Contreras, M.A. Osorio, J. Cartes,
406: F. Nieto and A.J. Ramirez-Pastor, Phys. Rev. B \bf 58, \rm 8475 (1998).
407: \bibitem{Vogel02} E.E. Vogel, S. Contreras, F. Nieto and A.J. Ramirez-Pastor,
408: Physica A \bf 257, \rm 256 (1998).
409: \bibitem{Barahona1982} F. Barahona, R. Maynard, R. Rammal and J.P.
410: Uhry, J. Phys. A: Math. Gen. \bf 15, \rm 673 (1982).
411: \bibitem{Valdes98} J.F. Valdes, J. Cartes, E.E. Vogel, S. Kobe,
412: and T. Klotz, Physica A \bf 257, \rm 557 (1998).
413: \bibitem{Enter84} A.C.D. van Enter and J.L. van Hemmen, Phys. Rev. A {\bf 29}, 355 (1984).
414: \bibitem{Ramirez04} A.J. Ramirez-Pastor, F. Rom\'a, F. Nieto and E.E. Vogel, Phys. A {\bf 336}, 454 (2004).
415: \bibitem{marinari} E. Marinari and G. Parisi, Eurphys. Lett. {\bf
416: 19},
417: 451 (1992).
418: \bibitem{kerler} W. Kerler and P. Rehberg, Phys. Rev. E {\bf 50}, 4220 (1994).
419: \bibitem{kawa72} K. Kawasaki, in {\it Phase Transitions and Critical Phenomena},
420: edited by C. Domb and M. Green (Academic Press, London, 1972),
421: Vol. 2.
422: \bibitem{Katzgraber05} H.G. Katzgraber and L.W Lee, Phys. Rev. B \bf 71, \rm 134404 (2005).
423: \bibitem{Young01} A. P. Young, J. Phys. C \bf 17, \rm L517 (1984).
424: \bibitem{Cheung} H.F. Cheung and W.L. McMillan, J. Phys. C \bf 16, \rm 7027 (1983).
425: \bibitem{McMillan01} W.L. McMillan, Phys. Rev. B \bf 28, \rm 5216 (1983).
426: \bibitem{Mor01} I. Morgenstern and K. Binder, Phys. Rev. Lett. \bf 43, \rm 1615 (1979).
427: \bibitem{Mor02} I. Morgenstern and K. Binder, Phys. Rev. B \bf 22, \rm 288 (1980).
428: \end{thebibliography}
429:
430: \newpage
431:
432: \begin{center}
433: \section*{FIGURE CAPTIONS}
434: \end{center}
435:
436: \vspace{1cm}
437:
438: \noindent Fig. 1. Distributions of the order parameters (a)
439: $\Phi_{\alpha}$ and (b) $|q_{\alpha \beta}|$,
440: at 3 different temperatures as indicated. \\[18pt]
441:
442: \noindent Fig. 2. Cumulant $U_L(T)$ plotted versus $T$ for
443: various lattice sizes as indicated. The inset zooms
444: the area indicated by a dashed frame.\\[18pt]
445:
446: \noindent Fig. 3. Scaling plot of $U_L$ against
447: $(T-T_c)L^{1/\nu}$, with $T_c=0$ and $\nu=2.63 \pm 0.20$.
448: \\[18pt]
449:
450: \noindent Fig. 4. (a) $P$ and $p$ and (b) the growing of $[n]$ as
451: a function of the lattice size, respectively.
452:
453: \newpage
454: \begin{figure}[tbp]
455: \centering
456: \includegraphics [width=20cm,angle=0]{figure1}
457: \end{figure}
458:
459: \newpage
460: \begin{figure}[tbp]
461: \centering
462: \includegraphics [width=30cm,angle=0]{figure2}
463: \end{figure}
464:
465:
466: \newpage
467: \begin{figure}[tbp]
468: \centering
469: \includegraphics [width=30cm,angle=0]{figure3}
470: \end{figure}
471:
472: \newpage
473: \begin{figure}[tbp]
474: \centering
475: \includegraphics [width=20cm,angle=0]{figure4}
476: \end{figure}
477:
478:
479:
480:
481:
482: %\end{large}
483:
484: \end{document}
485: