1: \documentclass[12pt]{article}
2: \usepackage{graphicx}
3: \DeclareGraphicsRule{.JPG}{eps}{*}{`jpeg2ps #1}
4:
5: \begin{document}
6: \centerline{\bf Ising model $S=1$, 3/2 and $2$ on directed networks}
7:
8:
9: \bigskip
10: \centerline{F.W.S. Lima$^1$ and Edina M.S. Luz$^2$}
11:
12: \bigskip
13: \noindent
14: $^1$Departamento de F\'{\i}sica,
15: Universidade Federal do Piau\'{\i}, 64049-550, Teresina - PI, Brazil \\
16: $^2$Departamento de F\'{\i}sica,
17: Universidade Estadual do Piau\'{\i}, 64002-150, Teresina - PI, Brazil
18:
19: \medskip
20: e-mail: wel@ufpi.br, edina@uespi.br
21: \bigskip
22:
23: {\small Abstract: On {\it directed} Barab\'asi-Albert and Small-World networks the
24: Ising model with spin $S=1$, $3/2$ and $2$ is now studied through Monte Carlo simulations.
25: In this model, the order-disorder phase transition of the
26: order parameter
27: is well defined on Small-World networks for Ising model with spin $S=1$. We calculate the value of the critical
28: temperature $T_{c}$ for several values of rewiring probability $p$ of the
29: {\it directed} Small-World network. This model on {\it directed} Small-World networks we
30: obtained a second-order phase transition for $p=0.2$ and first-order
31: phase transition for $p=0.8$. The critical
32: exponentes $\beta/\nu$, $\gamma/\nu$ and
33: $1/\nu$ were calculated for $p=0.2$. On {\it directed} Barab\'asi-Albert we
34: show that no there is phase transition for Ising model with spin $S=1$, $3/2$
35: and $2$. }
36:
37: Keywords: Monte Carlo simulation, spins , networks, Ising.
38:
39: \bigskip
40:
41: {\bf Introduction}
42:
43:
44: This paper deals with Ising spin on {\it directed} Barab\'asi-Albert(BA) and
45: Small-World(SW) networks. Sumour and Shabat \cite{sumour,sumourss} investigated Ising models with
46: spin $S=1/2$ on {\it directed} BA networks \cite{ba} with
47: the usual Glauber dynamics. No spontaneous magnetisation was
48: found, in contrast to the case of {\it undirected} BA networks
49: \cite{alex,indekeu,bianconi} where a spontaneous magnetisation was
50: found below a critical temperature which increases logarithmically with
51: system size. In S=1/2 systems on {\it undirected}, scale-free hierarchical-lattice
52: SW networks \cite{nihat}, conventional and algebraic
53: (Berezinskii-Kosterlitz-Thouless) ordering, with finite transition
54: temperatures, have been found. Lima and Stauffer \cite{lima} simulated
55: {\it directed} square, cubic and hypercubic lattices in two to five dimensions
56: with heat bath dynamics in order to separate the network effects form
57: the effects of directedness. They also compared different spin flip
58: algorithms, including cluster flips \cite{wang}, for
59: Ising-BA networks. They found a freezing-in of the
60: magnetisation similar to \cite{sumour,sumourss}, following an Arrhenius
61: law at least in low dimensions. This lack of a spontaneous magnetisation
62: (in the usual sense)
63: is consistent with the fact
64: that if on a directed lattice a spin $S_j$ influences spin $S_i$, then
65: spin $S_i$ in turn does not influence $S_j$,
66: and there may be no well-defined total energy. Thus, they show that for
67: the same scale-free networks, different algorithms give different
68: results. The $q$-state Potts model has been studied in scale-free networks
69: by Igloi and Turban \cite{igloi} and depending on the value of $q$ and the
70: degree-exponent $\gamma$ first- and second-order phase transitions
71: are found, and also by Lima \cite{lima2} on {\it directed}
72: BA network, where only first-order phase transitions
73: have being obtained independent of values of $q$ for values of
74: connectivity $z=2$ and $z=7$ of the {\it directed} BA network.
75: More recently, Lima \cite{lima1} simulated the the Ising model
76: for spin $S=1$ on {\it directed}
77: BA network and different from the Ising model for
78: spin $S=1/2$, the {\it unsual} order-disorder phase transition of
79: order parameter was seen; this effect is re-evaluated in the light of the time
80: depemdence presented below. We study the Ising model for spin $1$, $3/2$ and $2$ on {\it directed}
81: BA network. The Ising model with spin $1$, $3/2$ and $2$ was seen not to show a {\it usual} spontaneous
82: magnetisation and this decay time for flipping of the magnetisation
83: followed an Arrhenius law for HeatBath algorithms that agree with the
84: results of the Ising model for
85: spin $S=1/2$ \cite{sumour,sumourss} on directed BA
86: network.
87: Edina and Lima \cite{lima3} calculate the exponentes
88: $\beta/\nu$, $\gamma/\nu$, and $1/\nu$ exponents for
89: majority-vote model on
90: {\it directed} SW networks of S\'anchez et al. \cite{sanches},
91: and on these networks the exponents are
92: different from the Ising model a two-dimensional and independ on the values of
93: rewiring probability $p$ of the
94: {\it directed} SW networks. Here we also study the Ising model for spin
95: $S=1$ on {\it directed} SW, we obtained a second-order phase transition for
96: $p=0.2$ and a first-order
97: phase transition for $p=0.8$. We also calculate the critical
98: exponents $\beta/\nu$, $\gamma/\nu$ and
99: $1/\nu$ for $p=0.2$.
100:
101: \bigskip
102:
103: \begin{figure}[hbt]
104: \begin{center}
105: \includegraphics [angle=-90,scale=0.5]{lima01.eps}
106: %\includegraphics[angle=-90,scale=0.46]{lima_stauffer1b.eps}
107: \end{center}
108: \caption{Reciprocal logarithm of the relaxation times on directed BA networks for
109: $S=1$ to $S=2$.}
110: \end{figure}
111:
112: \bigskip
113:
114: \begin{figure}[hbt]
115: \begin{center}
116: \includegraphics [angle=-90,scale=0.5]{lima02.eps}
117: %\includegraphics[angle=-90,scale=0.46]{lima_stauffer1b.eps}
118: \end{center}
119: \caption{Characteristic time for $M(\tau)=3/4$ using 7 million spins for
120: $m=2$ neighbours and $2$ million spins for $m=7$ neighbours (+). Ten,
121: hundred, and thousand times smaller systems are denoted by x, star, and squares.
122: We plot the median over nine samples in this log-log plot. The
123: two straight lines have negative slopes 8 (left) and 12 (right). From
124: \cite{sumourss}.
125: }
126: \end{figure}
127:
128: \bigskip
129:
130: \begin{figure}[hbt]
131: \begin{center}
132: \includegraphics [angle=-90,scale=0.5]{lima04.eps}
133: %\includegraphics [angle=-90,scale=0.46]{elfig2b.eps}
134: \end{center}
135: \caption{
136: Magnetisation as a function of the temperature $T$, for
137: $N=16384$ sites. A second-order phase transition for values of $p=0.2$ and a first-order phase transition for $p=0.8$ .}
138: \end{figure}
139:
140: \begin{figure}[hbt]
141: \begin{center}
142: \includegraphics[angle=-90,scale=0.29]{lima05a.eps}
143: \includegraphics[angle=-90,scale=0.29]{lima05b.eps}
144: \end{center}
145: \caption{
146: Binder's fourth-order cumulant as a function of $K$. We have $p=0.2$ (part a)
147: and $p=0.8$ (part b) for $L=8$, 16, 32, 64 and 128.}
148: \end{figure}
149:
150: \begin{figure}[hbt]
151: \begin{center}
152: \includegraphics[angle=-90,scale=0.50]{lima06.eps}
153: \end{center}
154: \caption{$2/3-B_{i,min}$ for $p=0.2$ (circles) and $0.8$ (squares).
155: }
156: \end{figure}
157:
158: \begin{figure}[hbt]
159: \begin{center}
160: \includegraphics[angle=-90,scale=0.50]{lima07.eps}
161: \end{center}
162: \caption{Binder's fourth-order cumulant of the magnetisation versus temperature for $p=0.2$. }
163: \end{figure}
164:
165: \bigskip
166:
167: \begin{figure}[hbt]
168: \begin{center}
169: \includegraphics[angle=-90,scale=0.60]{lima08.eps}
170: \end{center}
171: \caption{ln $M(T_{c})$ versus ln $L$ for $p=0.2$.}
172: \end{figure}
173: \bigskip
174:
175: \begin{figure}[hbt]
176: \begin{center}
177: \includegraphics[angle=-90,scale=0.50]{lima09.eps}
178: \end{center}
179: \caption{ Plot of ln $\chi^{max}(L)$ (square) and ln$ \chi(T_{c})$ (circle) versus ln $L$ for $p=0.2$.}
180: \end{figure}
181:
182: \begin{figure}[hbt]
183: \begin{center}
184: \includegraphics[angle=-90,scale=0.50]{lima010.eps}
185: \end{center}
186: \caption{ Reciprocal logarithm of the relaxation times on directed SW networks for versus $T$ for p=0.2(+) and $0.8$(x).}
187: \end{figure}
188:
189: \bigskip
190:
191: {\bf Model and Simulaton}
192:
193:
194:
195: {\bf Ising model on {\it directed}
196: BA Networks:}
197:
198: We consider the spin 1, 3/2 and 2 Ising models on {\it directed}
199: BA Networks, defined by a set of
200: spins variables ${S}$ taking the values $\pm 1$ and
201: $0$ for $S=1$, $\pm 3/2$ and $1/2$ for $S=3/2$, and $\pm 1$, $\pm 2$ and
202: $0$ for $2$, respectivaly, located
203: on every site of a directed
204: BA Networks with $N$ sites.
205:
206: The probability for spin $S_{i}$ change its state in this {\it directed} networks is
207: \begin{equation}
208: %\begin{center}
209: p_{i}= 1/1+\exp(2E_{i}/k_BT), \quad E_{i}=-J\sum_{k}S_{i}S_{k}
210: %\end{center}
211: \end{equation}
212: where $k$-sum runs over all nearest neighbors of $S_{i}$. In this network,
213: each new site
214: added to the network selects with connectivity $z=2$
215: already existing sites as neighbours influencing it; the newly
216: added spin does not influence these neighbours.
217:
218: To study the spin $1$, $3/2$ and $2$ Ising models we start with
219: all spins up, a number of spins equal to $N=500000$, and Monte
220: Carlo step (MCS) time up $2,000,000$, in our simulations, one MCS is
221: accomplished
222: after all the $N$ spins are updated, here, with HeatBath Monte Carlo
223: algorithm. Then we vary the temperature and study nine samples.
224: The temperature is measured in units of critical temperature of the
225: square-lattice Ising model. We determine the time $\tau$ after which the
226: magnetisation has flipped its sign for the first time, and then take the
227: median value of our nine samples. So we get different values $\tau_{1}$
228: for different temperatures.
229:
230: In the study the critical behavior this Ising model (with spins
231: $1$, $3/2$ and $2$) we define the variable
232: $m=\sum_{i=1}^{N}S_{i}/N$ as normalized magnetisation.
233:
234: Our BA simulations, using the HeatBath algorithm, indicate that the
235: spins $S=1$, $3/2$ and $2$ Ising model do not display a phase transition and
236: the plot of the time $1/\ln\tau$ versus temperature in Fig. 1 shows that our BA
237: results for all spins agree with the modified Arrhenius law for relaxation time,
238: defined as the first time when the sign of the magnetisation flips:$1/\ln(\tau)
239: \propto T +...$. This result agrees with the results of Sumour
240: et al. for spin $S=1/2$ \cite{sumourss} with $7$ million of sites, see Fig. 2(figure
241: retired of reference \cite{sumourss}) and our results are more reliable than
242: both Lima \cite{lima1} and Sumour et al. \cite{sumuor2007} because that this
243: are longer simulation times.
244: \bigskip
245:
246:
247: {\bf Ising model on {\it directed}
248: SW Networks:}
249:
250:
251: We consider the Ising model with spin $S=1$, on {\it directed}
252: SW Networks, defined by a set of
253: spins variables ${S}$ taking the values $-1$, $0$ and
254: $+1$, situated on every site of a {\it directed}
255: SW Networks with $N$ sites.
256:
257: The probability for spin $S_{i}$ to change its state in these {\it directed}
258: network is given eq. (1) where the
259: sum runs over all nearest neighbors of $S_{i}$. In this network, created for
260: S\'anchez et al. \cite{sanches}, we start from a two-dimensional square
261: lattice consisting of sites linked to their four nearest neighbors by both outgoing and
262: incoming links. Then, with probability $p$, we reconnect nearest-neighbor outgoing
263: links to a different site chosen at random. After repeating this process for every
264: link, we are left with a network with a density $p$ of SW {\it directed} links. Therefore,
265: with this procedure every site will have exactly four outgoing links and
266: varying (random) number of incoming links. To study the critical behavior of
267: the model we use the HeatBath algorithm and define the variable $e=E/N$, where
268: $E=2 \sum_{i}E_{i}$ and $m=\sum_{i=1}^{N}S_{i}/N.$ From the fluctuations of
269: the $e$ measurements we can compute: the average $e$,
270: the specific heat $C$ and the fourth-order cumulant of $e$,
271: %
272: \begin{equation}
273: u(K)=[<E>]_{av}/N,
274: \end{equation}
275: %
276: \begin{equation}
277: C(K)=K^{2}N[<e^{2}>-<e>^{2}]_{av},
278: \end{equation}
279: %
280: \begin{equation}
281: B_{i}(K)=[1-\frac{<e^{4}>}{3<e^{2}>^{2}}]_{av},
282: \end{equation}
283: where $K=J/k_BT$, with $J=1$, and $k_B$ is the Boltzmann constant.
284: Similarly, we can derive from the magnetization measurements
285: the average magnetization, the susceptibility, and the magnetic
286: cumulants,
287: %
288: \begin{equation}
289: m(K)=[<|m|>]_{av},
290: \end{equation}
291: %
292: \begin{equation}
293: \chi(K)=KN[<m^{2}>-<|m|>^{2}]_{av},
294: \end{equation}
295: %
296: \begin{equation}
297: U_{4}(K)=[1-\frac{<m^{4}>}{3<|m|>^{2}}]_{av}.
298: \end{equation}
299: where $<...>$ stands for a thermodynamic average and $[...]_{av}$ square brackets
300: for a averages over the 20 realizations.
301:
302: In order to verify the order of the transition this model, we apply finite-size scaling (FSS) \cite{fss}. Intially we search for the minima of $e$ fourth-order parameter of eq. (4). This quantity gives a qualitative as well as a quantitative description of the order of the transition \cite{mdk}. It is known \cite{janke} that this
303: parameter takes a minimum value $B_{i,min}$ at effective transition temperature
304: $T_{c}(N)$. One can show \cite{kb} that for a second-order transition $\lim_{N\to \infty}$
305: $(2/3-B_{i,min})=0$, even at $T_{c}$, while at a first-order transition the same limit measurng the same quantity is small and $(2/3-B_{i,min})\neq0$.
306:
307: A more quantitative analysis can be carried out through the FSS of the $C$ fluctuation
308: $C_{max}$, the susceptibility maxima $\chi_{max}$ and the minima of the Binder parameter $B_{i,min}$. If the hypothesis of a first-order phase transition is correct, we should then expect, for large systems sizes, an asymptotics FSS behavior of the form
309: \cite{wj,pbc},
310: \begin{equation}
311: C_{max}=a_{C} + b_{C}N +...
312: \end{equation}
313: %
314: \begin{equation}
315: \chi_{max}=a_{\chi} + b_{\chi}N +...
316: \end{equation}
317: \begin{equation}
318: B_{i,min}=a_{B_{i}} + b{B_{i}}N +...
319: \end{equation}
320:
321: Therefore, if the hypothesis of a second-order phase transition is correct, we should then expect, for large systems sizes, an asymptotics FSS behavior of the form
322: \begin{equation}
323: C=C_{reg}+L^{\alpha/\nu}f_{C}(x)[1+...],
324: \end{equation}
325: \begin{equation}
326: [<|m|>]_{av}=L^{-\beta/\nu}f_{m}(x)[1+...],
327: \end{equation}
328: \begin{equation}
329: \chi=L^{\gamma/\nu}f_{\chi}(x)[1+...],
330: \end{equation}
331: \begin{equation}
332: \frac{dU_{4}}{dT}=L^{1/\nu}f_{U}(x)[1+...],
333: \end{equation}
334: where $C_{reg}$ is a regular background term,
335: $\nu$, $\alpha$, $\beta$, and $\gamma$, are the usual critical
336: exponents, and $f_{i}(x)$ are FSS functions with
337: \begin{equation}
338: x=(K-K_{c})L^{1/\nu}
339: \end{equation}
340: being the scaling variable, and the brackets $[1+...]$ indicate
341: corretions-to-scaling terms. Therefore, from the size dependence of $M$ and $\chi$
342: we obtained the exponents $\beta/\nu$ and $\gamma/\nu$, respectively.
343: The maximum value of susceptibility also scales as $L^{\gamma/\nu}$. Moreover, the
344: value of $T$ for which $\chi$ has a maximum, $ T_{c}^{\chi_{max}}=T_{c}(L)$,
345: is expected to scale with the system size as
346: \begin{equation}
347: T_{c}(L)=T_{c}+bL^{-1/\nu},
348: \end{equation}
349: were the constant $b$ is close to unity. Therefore, the relations $(14)$ and $(16)$
350: are used to determine the exponent $1/\nu$.
351:
352: We have performed Monte Carlo simulation on {\it directed} SW network with
353: various values of probability $p$. For a given $p$, we used systems
354: of size $L=8$, 16, 32, 64, and 128. We waited $50,000$ Monte Carlo
355: steps (MCS) to make the system reach the steady state, and the time averages were
356: estimated from the next $ 50,000$ MCS. In our simulations, one MCS is accomplished
357: after all the $N$ spins are updated. For all sets of parameters, we have generated
358: $20$ distinct networks, and have simulated $20$
359: independent runs for each distinct network.
360:
361: \bigskip
362:
363: {\bf Results and Discussion}
364:
365: In Fig. 3 we show the dependence of the magnetisation $M$ on the temperature, obtained from simulations on {\it directed}
366: SW network with $L=128 \times 128$ sites and two values of probability $p$: $p=0.2$ for
367: second-order transition and $p=0.8$ for first-order transition. In Fig. 4 we plot
368: Binder's fourth-order cumulant of $e$ for different values of $L$ and two different values
369: of $p$: $p=0.2$ in part (a) and $p=0.8$ in part (b).
370: In Fig. 5 the difference $2/3-B_{i,min}$ is shown as a function of
371: parameter $1/N$ for probability $p=0.2$ (circles) and $=0.8$ (squares) obtained
372: from the data of Fig. 4. In Fig. 6 we show Binder's fourth-order cumulant of the magnetisation versus temperature for $p=0.2$. The temperature obtained is $T_{c}=1.890(6)$. Figs. 7 we plot the dependence of the magnetisation at $T=T_{c}$ versus the system
373: size $L$. The slopes of curves correspond to the exponent ratio $\beta/\nu$ of according
374: to Eq. (12).
375: The results show that the exponent ratio $\beta/\nu$ at $T_{c}$ is $0.379(21)$.
376:
377: In Fig. 8 we display the scalings for susceptibility at $T=T_{c}(L)$ , $\chi(T_{c}(L))$ (circles), and for its maximum amplitude, $\chi_{L}^{max}$ (squares), and the scalings for susceptibility at $T=T_{c}$ obtained from Binder's cumulant, $ \chi(T_{c})$ versus $L$ for probability $p=0.2$. The exponents ratio $\gamma/\nu$ are obtained from the slopes
378: of the straight lines. The exponents $\gamma/\nu$ of the two estimates agree (within errors). The values obtained are $\gamma/\nu=1.252(13)$ (circles) and $1.264(34)$ (squares), respectively. Therefore we can use the Eq. (16) , for $p=0.2$, obtain the critical exponent $1/\nu$, that is equal to $1.209(165)$ and $1.248(196)$ obtained of according
379: to Eq. (14).
380: To improve our results obtained above we start with all spins up, a number of spins equal to $N=640000$, and time up 2,000,000 (in units of Monte Carlo steps per spins). Then we vary temperature $T$ and at each $T$ study the time dependence for 9 samples. We determine the time $\tau$ after which the magnetisation has flipped its sign for first time, and then take the median values of our nine samples. So we get different values $\tau_{1}$ for different temperatures $T$. In Fig. 9 show that the decay time goes to infinity at some positive $T$ value.
381: This behavior ensures that there is a phase transition for Ising model spin
382: $S=1$ on {\it directed} SW network.
383:
384: \bigskip
385:
386: {\bf Conclusion}
387:
388: In conclusion, we have presented the Ising model spin spins $S=1$, $3/2$ and $2$ on
389: {\it directed} BA and {\it directed} SW network($S=1$). The Ising model does not display a phase transition on {\it directed} BA for spins $S=1/2$, $1$, $3/2$ and $2$. In the {\it directed} SW network \cite{sanches} this model presents a
390: first- and second-order phase transition which occurs with
391: probability $p=0.8$ and $0.2$, respectively. The exponents obtained for $p=0.2$ are different from the exponents the Ising model on square lattice, that it suggests that these exponents belong to one another class of universality.
392:
393: The authors thanks D. Stauffer for many suggestions and fruitful
394: discussions during the development this work and also for the revision of
395: this paper. We also acknowledge the Brazilian agency FAPEPI
396: (Teresina-Piau\'{\i}-Brasil) for its financial support. This work also was supported the
397: system SGI Altix 1350 the computational park CENAPAD.UNICAMP-USP, SP-BRAZIL.
398:
399: \begin{thebibliography}{99}
400:
401: \bibitem{sumour} M.A. Sumour and M.M. Shabat, Int. J. Mod. Phys. C 16,
402: 585 (2005) and cond-mat/0411055 at www.arXiv.org.
403:
404: \bibitem{sumourss} M.A. Sumour, M.M. Shabat and D. Stauffer,
405: Islamic University Journal (Gaza) 14, 209 (2006).
406:
407:
408: \bibitem{ba} R. Albert and A.L. Barab\'asi, Rev. Mod. Phys. 74, 47
409: (2002).
410:
411: \bibitem{alex} A. Aleksiejuk, J.A. Ho\l yst and D. Stauffer, Physica A
412: 310, 269 (2002).
413:
414: \bibitem{indekeu} J.O. Indekeu, Physica A 333, 461 (2004).
415:
416: \bibitem{bianconi} G. Bianconi, Phys. Lett. A 303, 166 (2002).
417:
418: \bibitem{nihat} M. Hinczewski and A.N. Berker, Phys. Rev. E 73,
419: 066126 (2006).
420:
421: \bibitem{lima} F.W.S. Lima and D. Stauffer, Physica A 359, 423 (2006).
422:
423: \bibitem{wang} J.S. Wang and R. H. Swendsen, Physica A 167, 565 (1990).
424:
425: \bibitem{igloi} F. Igloi and L. Turban, Phys. Rev. E66, 036140 (2002), cond-mat/0206522.
426:
427: \bibitem{lima2} F. W. S. Lima, Commun. Comput. Phys., in press cond-mat/0608137.
428:
429: \bibitem{lima1} F. W. S. Lima, Int. J. Mod. Phys. C. 17, 1267 (2006).
430:
431: \bibitem{lima3} E.M.S.Luz and F.W.S. Lima, Int. J. Mod. Phys. C 18, in press
432: (2007).
433:
434: \bibitem{sanches} Alejandro D. S\'anchez, Juan M. Lopes, and Miguel A. Rodriguez, Phys.
435: Rev. Lett. 88, 048701 (2002)
436:
437: \bibitem{fss} See Finite Size Scaling and Numerical Simulation of Statistical Systems, edited by V. Privman (World Scientific, Singapore, 1990).
438:
439: \bibitem{mdk} M.S.S. Challa, D. P. Landau, K. Binder, Phys. Rev. B, 34, 1841 (1986).
440:
441: \bibitem{janke} W. Janke, Phys. Rev. B 47, 14757 (1993).
442:
443: \bibitem{kb} K. Binder, D. J. Herrmann, in Monte-Carlo Simulation in Statistical
444: Phys., edited by P. Fulde (Springer-Verlage,Berlin, 1988), p. 61-62.
445:
446: \bibitem{wj} W. Janke, R. Villanova, Phys. Lett. A 209, 179 (1995).
447:
448: \bibitem{pbc} P. E. Berche, C. Chatelain, B. Berche, Phys. Rev. Lett. 80, 297 (1998).
449:
450: \bibitem{sumuor2007} M.A. Sumour, A.H. El-Astal, F.W.S. Lima, M.M. Shabat, H.M.
451: Khalil, Int. J. Mod. Phys. C 18, in press; cond-mat/0612189.
452:
453: \end{thebibliography}
454: \end{document}
455: