1: \documentclass{cpcauth}
2: \usepackage{psfig}
3: \newcommand{\D}{{\rm d}}
4: \begin{document}
5: \begin{frontmatter}
6: \title{Phase transitions at surfaces, edges, and corners}
7:
8: \author[ER,NA]{Michel Pleimling\thanksref{ca}}
9: \address[ER]{Institut f\"ur Theoretische Physik 1, Universit\"at Erlangen-N\"urnberg,
10: D--91058
11: Erlangen, Germany}
12: \address[NA]{Laboratoire de Physique des Mat\'eriaux$^{\, 2}$, Universit\'e Henri
13: Poincar\'e Nancy I, B.P. 239,
14: F--54506 Vand{\oe}uvre l\`es Nancy Cedex, France}
15: \thanks[ca]{
16: e-mail: pleim@theorie1.physik.uni-erlangen.de}
17: \thanks[an]{Laboratoire associ\'e au CNRS UMR 7556}
18:
19: \begin{abstract}
20: Results of large-scale Monte Carlo simulations of three-dimensional
21: Ising models with edges and corners are reviewed. At the ordinary transition,
22: angle dependent critical exponents are observed, whereas at the surface transition
23: edge and corner critical exponents are non-universal and depend on the details of the model.
24: The results obtained at the surface transition are compared to exact findings on critical
25: two-dimensional Ising models with different types of defect lines.
26: \end{abstract}
27: \end{frontmatter}
28: \section{Introduction}
29: Critical phenomena at perfect, flat surfaces of three-dimensional systems have been
30: the subject of intensive research during the last three decades \cite{Bin83,Die86,Dos92}.
31: For the three-dimensional semi-infinite Ising model two different ferromagnetic couplings
32: are usually introduced, depending on whether the neighbouring spins are both located
33: at the surface, $J_s > 0$, or not, $J_b > 0$. The resulting phase diagram is well
34: established. If the ratio of the surface coupling $J_s$ to the bulk
35: coupling $J_b$, $r=J_s/J_b$, is sufficiently small, the system undergoes at the bulk critical temperature
36: $T_{c}$ an ordinary transition, with the bulk and surface ordering occurring at the same temperature.
37: Beyond a critical ratio, $r > r_{sp} \approx 1.50$ \cite{Bin84,Rug93}, the surface orders at a higher
38: temperature $T_s > T_{c}$ at the surface transition, followed by the extraordinary transition
39: of the bulk at $T_{c}$. At the critical ratio $r_{sp}$, one encounters the multicritical
40: special transition point.
41:
42: The semi-infinite Ising model with a flat surface may be considered to be a special
43: case of a more complex wedge geometry where two planes meeting at an angle $\theta$
44: form an infinite edge. For $\theta= \pi$ the flat surface is recovered. Cardy \cite{Car83}
45: showed that at the ordinary transition edge critical exponents depending
46: continuously on $\theta$ arise on purely geometrical grounds. For a given opening angle
47: $\theta$, however, the values of the critical exponents are expected to be universal and independent
48: on microscopic details like the strengths of the coupling constants or the lattice type.
49: Whereas edge singularities at the ordinary transition have been studied intensively,
50: especially in two dimensions \cite{Igl93}, edge critical
51: behaviour at the surface transition of three-dimensional Ising models has been largely overlooked.
52: At the surface transition the critical fluctuations are essentially of two-dimensional
53: character. One may then argue that the edge should act like a
54: local perturbation in a two-dimensional system.
55: In analogy with exact results obtained for two-dimensional
56: Ising models with defect lines \cite{Bar79}, intriguing nonuniversal edge critical behaviour
57: is therefore expected.
58: Similarly, corner critical exponents depending on the details of the model should also be observed
59: at the surface transition.
60:
61: In this contribution, I discuss results on edge and corner critical behaviour obtained in large-scale
62: Monte Carlo simulations of three-dimensional Ising models \cite{Ple98,Ple99,Ple00}.
63: After presenting some details of the simulations in the next Section, I then discuss
64: edge and corner criticality at the ordinary (Section 3) and at the surface transition
65: (Section 4).
66: A brief summary concludes the paper.
67:
68: \section{Models with edges and corners}
69: To introduce edges in Ising magnets defined on simple cubic lattices,
70: periodic boundary conditions along one axis,
71: the $z$-axis, are applied. The remaining four free surfaces of the crystal may be
72: oriented in various ways leading to different opening angles $\theta$ at the edges.
73: As shown in Figure 1, pairs of (100) and (010) surfaces lead to four equivalent edges
74: with opening angles $\theta= \pi/2$. The intersection of (100) and (110) surfaces form
75: two pairs of edges with $\theta=\pi/4$ and $\theta= 3 \pi/4$, see Figure 2.
76: Besides the bulk coupling $J_b$ and the surface coupling $J_s$, further couplings may
77: be introduced \cite{Ple98,Ple99}: two neighbouring edge spins interact with the edge coupling
78: $J_e$, whereas an edge spin is coupled to its neighbouring surface spin by the edge-surface
79: interaction $J_{es}$.
80: The Hamiltonian is then given by
81: \begin{eqnarray}
82: H & = & -\sum\limits_{bulk} J_b S_{{\bf r}} S_{{\bf r}'}
83: -\sum\limits_{surface} J_s S_{{\bf r}} S_{{\bf r}'} \nonumber \\
84: & & -\sum\limits_{edge-surface} J_{es} S_{{\bf r}} S_{{\bf r}'}
85: -\sum\limits_{edge} J_e S_{{\bf r}} S_{{\bf r}'}
86: \end{eqnarray}
87: with spins $S_{{\bf r}}= \pm 1$ at sites ${\bf r}=(xyz)$.
88: For the study of corners we use free boundary conditions in all three directions.
89: The couplings between corner spins and neighbouring
90: edge spins are set equal to $J_e$, thus regarding corners as endpoints of edges.
91:
92: \begin{figure}
93: \centerline{\psfig{figure=figure1.eps,width=7.5cm,angle=270}}
94: \caption{Geometry of an Ising model with (100) and (010) surfaces forming edges
95: with opening angles $\theta = \pi/2$.}
96: \label{fig1} \end{figure}
97:
98: \begin{figure}
99: \centerline{\psfig{figure=figure2.eps,width=7.5cm,angle=270}}
100: \caption{Geometry of a model with edges with opening angles $\theta=\pi/4$ and
101: $\theta=3 \pi/4$. Shown is a cut through the crystal perpendicular to the edge direction. The full lines
102: represent the couplings between
103: neighbouring spins, whereas the (110) surfaces are indicated by the dashed lines.}
104: \label{fig2} \end{figure}
105:
106: The results presented in the following have been obtained with the
107: efficient one-cluster-flip algorithm \cite{Wol89}. Systems with $L \times L \times
108: N$ sites were considered, with $L \times L$ being the number of sites in the planes
109: perpendicular to the $z$-axis, and $N$ being the number of sites along that axis.
110: $L$ ranged from 10 to 80 and $N$ from 10 to 640. For the investigation of corner
111: critical behaviour Ising cubes with up to $80^3$ spins were simulated.
112:
113: As we aim at studying properties of systems with infinitely long edges, finite-size effects have
114: to be monitored closely. The Monte Carlo system should be large enough to reproduce
115: the thermodynamic values of the bulk and the surface magnetization
116: sufficiently far away from the edge. In general, at a given temperature,
117: edge quantities must be stable against enlargening the system size. The same holds for corner
118: quantities.
119:
120: An interesting quantity in systems without corners
121: is the magnetization per site for lines parallel to the $z$-axis:
122: \begin{equation}
123: m(x,y)= \langle | \sum\limits_z S_{xyz} | \rangle / N .
124: \end{equation}
125: The sum runs over the $N$ spins in a line, with $x$ and $y$ being fixed. The brackets denote
126: thermal averages. As usual, the absolute value is taken to avoid vanishing magnetizations
127: for finite systems.
128:
129: From the line magnetizations, various magnetizations of interest are obtained. The edge
130: magnetization $m_2$ is identical to
131: \begin{equation}
132: m_2 = \langle | \sum\limits_z S_{x_0y_0z} | \rangle / N
133: \end{equation}
134: where $(x_0y_0z)$ denotes an edge site. The surface magnetization results
135: from the line magnetization in the center of a surface, whereas the bulk magnetization
136: is given by the line magnetization in the middle of the crystal.
137:
138: The local magnetization $m_l(x,y,z)$ may be obtained by computing the
139: correlation function $\left< S_{xyz} \, S_{x'y'z'} \right>$ between topologically
140: equivalent sites ($xyz$) and ($x'y'z'$) with maximal separation distance.
141: In this way, the corner magnetization of Ising cubes with linear dimension $L$ is given by
142: \begin{equation}
143: m_3 = \sqrt{ \langle S_{000} \, S_{LLL} \rangle }.
144: \end{equation}
145: I refer to corner quantities by the subscript 3 (indicating that three planes meet at this
146: point), whereas edge quantities have the subscript 2.
147:
148: \section{Ordinary transition}
149: In \cite{Car83}, Cardy considered $d$-dimensional systems with $(d-1)$-dimensional hyperplanes
150: meeting at an angle $\theta$. At the bulk critical point
151: local critical exponents changing continuously with
152: $\theta$ are already obtained in mean-field approximation.
153: %The edge magnetization critical
154: %exponent is for example given by
155: %\begin{equation} \label{gl:4_1}
156: %\beta_2^{MF}= \frac{1}{2} + \frac{\pi}{2 \theta}.
157: %\end{equation}
158: Cardy showed that all edge critical exponents can be obtained by combining
159: bulk and surface critical exponents with a new angle dependent edge exponent.
160: He also computed the value of this
161: exponent in first order of an $\epsilon=d-4$ expansion. From the renormalization group point of view,
162: the angle dependence has its origin in
163: the invariance of the edge under rescaling. This makes the opening angle to a marginal variable,
164: which may then lead to angle dependent local critical exponents.
165:
166: Edge magnetizations computed by Monte Carlo simulations are plotted in Figure 3 for
167: four different opening angles: $\theta= \pi/4$, $\pi/2$,
168: $3\pi/4$, and $\pi$ (correponding to the free surface), with all coupling
169: constants set equal to $J_b$. Only data
170: presenting no noticeable finite-size dependences are displayed.
171: This is achieved by studying systems with a large number of edge sites, the
172: edge length being the crucial quantity \cite{Ple98}.
173: %The geometry
174: %of Figure 1 is used to obtain the data for $\theta= \pi/2$, whereas
175: %simulating the geometry of Figure 2 yields the data for
176: %$\theta= \pi/4$ and $\theta= 3\pi/4$.
177:
178: \begin{figure}
179: \centerline{\psfig{figure=figure3.eps,width=7.5cm,angle=270}}
180: \caption{Edge magnetization versus temperature
181: for various opening angles $\theta$: $\pi/4$ (circles), $\pi/2$ (squares), $3 \pi/4$ (diamonds),
182: and $\pi$ (triangles). Only data not affected by finite-size effects are
183: displayed. Error bars are smaller than the size of the symbols.}
184: %The largest system for the opening angle
185: %$\pi/2$ contained $60 \times 60 \times 640$ spins.}
186: \label{fig3} \end{figure}
187:
188: At the ordinary transition, the edge magnetizations vanish at the bulk
189: critical temperature $k_B \, T_c/J_b \approx 4.5115$ with a power law
190: behaviour $m_2 \sim (T_c - T)^{\beta_2}$. Here, $\beta_2$ is the critical exponent
191: of the edge magnetization. In order to estimate $\beta_2$, one may study \cite{Ple98a}
192: an effective temperature dependent exponent $\beta_{2,eff}(t)$ defined by
193: \begin{equation}
194: \beta_{2,eff}(t) = \mbox{d} \ln m_2 / \mbox{d} \ln t
195: \end{equation}
196: where $t= | T - T_c| /T_c$ is the reduced temperature. Certainly, on approach to $T_c$,
197: $t \longrightarrow 0$, $\beta_{2,eff}$ becomes the asymptotic exponent $\beta_2$.
198: %Because the simulations are performed at discrete
199: %temperatures, $t_i$, $\beta_{eff}$ is replaced by
200: %\begin{equation}
201: %\beta_{eff}(t)= \ln (m_2(t_i)/m_2(t_{i+1}))/\ln (t_i/t_{i+1})
202: %\end{equation}
203: %with $t=(t_i+t_{i+1})/2$ (alternatively $t=\sqrt{t_i \, t_{i+1}}$).
204: Very accurate Monte Carlo
205: data are required to get reliable estimates for the critical exponents.
206:
207: \begin{table}
208: \caption
209: {Predictions for the edge magnetization critical exponents
210: $\beta_2(\theta)$ from various methods. MF: mean-field
211: approximation, RNG: renormalization group theory, HTS: high temperature series expansions, MC:
212: Monte Carlo simulations.}
213: \vspace*{0.2cm}
214: \begin{tabular}{|c||c|c|c|c|}
215: \hline
216: & $\pi/4$ & $\pi/2$ & $3 \pi/4$ & $\pi$ \\
217: \hline\hline
218: MF \cite{Car83} & 2.50 & 1.50 & 1.17 & 1.00 \\
219: \hline
220: RNG \cite{Car83} & 2.48 & 1.39 & 1.02 & 0.84 \\
221: \hline
222: HTS \cite{Gut84} & 2.30 & 1.31 & 0.98 & 0.81 \\
223: \hline
224: MC & $2.30 \pm 0.10$ & $1.28 \pm 0.04$ &$0.96 \pm 0.02$ & $0.80 \pm 0.01$ \\
225: \hline
226: \end{tabular}
227: \end{table}
228: %\begin{figure}
229: %\centerline{\psfig{figure=figure4.eps,width=7.5cm,angle=270}}
230: %\caption{Edge magnetization effective exponent as function of the reduced temperature $t$
231: %for various opening angles. The meaning of the symbols is the same as in Figure 3.
232: %The dashed lines
233: %represent the results obtained in an $\epsilon$-expansion \cite{Car83}, the dot-dashed lines
234: %indicate the values computed using high temperature series expansions \cite{Gut84}.}
235: %\label{fig4} \end{figure}
236:
237:
238: %Figure 4 shows the edge effective exponents derived from the edge magnetizations
239: %displayed in Figure 3.
240: As the slope of $\beta_{2,eff}$ changes rather mildly at sufficiently
241: small reduced temperatures, meaningful estimates for the asymptotic exponent $\beta_2(\theta)$
242: are feasible. The values for $\beta_2(\theta)$ obtained in
243: this way \cite{Ple98} are listed in Table 1 and may be compared with results obtained
244: from mean-field theory \cite{Car83}, renormalizations group calculations to first order
245: in $\epsilon$ \cite{Car83}, or high temperature series expansions \cite{Gut84}.
246: The predictions of the renormalization group
247: seem to be systematically too large, as suggested both by high temperature series expansions
248: and our Monte Carlo simulations. The last two methods yield results which
249: are in close agreement with each other.
250:
251: Changing the values of the coupling constants or rotating the crystal about the $z$-axis
252: at fixed opening angle $\theta$ (the wedge is then formed by a different pair of surfaces)
253: does not alter the critical exponents \cite{Ple98}, in accordance with the expected
254: universality at the ordinary transition.
255:
256: A similar analysis of the corner magnetization effective exponent yields for an Ising cube the
257: corner magnetization critical exponent $\beta_3 = 1.77 \pm 0.05$
258: \cite{Ple00}, significantly lower than the mean-field value $\beta_3^{MF}=2$ \cite{Igl93}.
259:
260: \section{Surface transition}
261: When approaching the surface transition from high temperatures,
262: the surface of a semi-infinite system orders whereas the bulk
263: remains disordered. In a three-dimensional system the critical fluctuations are then
264: essentially two-dimensional, and the surface critical exponents reflect the reduced dimensionality
265: by taking the values of the corresponding two-dimensional bulk system. At this transition,
266: an edge may be viewed as a local perturbation, acting
267: presumably like a line defect in a two-dimensional system.
268:
269: Two-dimensional Ising models with defect lines have been shown to display non-universal
270: critical exponents \cite{Bar79}.
271: A chain defect is formed by a column of perturbed couplings with strength $J_{ch}$,
272: whereas at a ladder defect modified couplings of strength $J_l$ connect spins
273: belonging to two neighbouring columns. In both cases
274: local magnetization critical exponent depend on the strengths of these defect couplings \cite{Bar79}.
275: In analogy with these results, one expects at the surface transition, for a fixed opening
276: angle, intriguing non-universal edge and corner critical behaviour in three-dimensional
277: Ising models \cite{Ple99,Ple00}.
278:
279: In the following, I discuss the influence of edges with opening angle $\theta=\pi/2$, see Figure 1,
280: on the local critical behaviour at the surface transition. The ratio $J_s/J_b= 2 > r_{sp}$
281: is kept fixed,
282: whereas the edge-edge, $J_e$, and the edge-surface, $J_{es}$, couplings are varied.
283: To determine edge critical behaviour accurately, the critical temperature of the
284: surface transition, $T_s$,
285: needs to be determined accurately. Using standard finite-size analysis, one obtains
286: $k_B \, T_s/J_b=4.9575 \pm 0.0075$ for $J_s=2J_b$ \cite{Ple99}.
287: Because edges are one-dimensional, and all couplings
288: in the models are of short range, edge quantities become singular at $T_s$.
289:
290: The profile of the line magnetization at the surface
291: %, defined by (see Figure 1 and eq.\ (2))
292: %\begin{eqnarray}
293: %m_{s}(i)& = & \left( m(x,L) + m(x,1) \right. \nonumber \\
294: %& & \left. + m(1,y) + m(L,y) \right)/4
295: %\end{eqnarray}
296: %with $i=x=y=1, \ldots L$,
297: reflects the influence of bulk spins close to the surface and edges
298: as well as the strength of the couplings near the edge, $J_e$ and $J_{es}$. Typical profiles
299: are shown in Figure 4 for $J_e=J_{es}=J_s=2J_b$. The edge magnetization is $m(1,L)$.
300: At low temperatures, $m(x,L)$ increases monotonically with the distance from the edge.
301: The magnetization of the surface spins, which are connected directly to an ordered bulk spin,
302: is enhanced compared to the edge magnetization. In contrast, on approach to $T_s$, the ordering
303: of the spins falls off quickly by going from the surface to the bulk, and the surface
304: magnetization is pulled down below the edge magnetization by the coupling to the disordered
305: bulk spins. Roughly at $T_c$, a non-monotonic behaviour in the profile shows up, with
306: a maximum close to the edge. It may be explained by the fact that surface spins next to
307: but on different sides of an edge are more strongly connected to each other, through the same
308: neighbouring bulk spin, than spins on the flat part of the surface with the separation distance
309: of two \cite{Ple99}.
310:
311: \begin{figure}
312: \centerline{
313: \psfig{figure=figure4.eps,width=7.5cm,angle=270}}
314: \caption{Line magnetization profiles near an edge with opening angle $\theta=\pi/2$
315: at various temperatures. The couplings are $J_e=J_{es}=J_s=2J_b$. Systems with $40^3$ spins
316: have been simulated.}
317: \end{figure}
318:
319: %Near the surface transition,
320: %the edge acts like a defect line in an essentially two-dimensional bulk Ising model. The edge
321: %coupling $J_e$ then corresponds to a chain-like defect, the edge-surface coupling $J_{es}$
322: %to a ladder-type defect. The change in the topology at the edge compared to the surface amounts
323: %to a complicated ladder-type defect.
324:
325: The effect on the edge critical behaviour of weakening
326: the interactions between the edge and the surfaces, $J_{es}$, is demonstrated in Figure 5.
327: A plateau-like behaviour is observed close to $T_s$, which facilitates
328: a precise determination of the asymptotic value. Obviously, the value of the critical exponent
329: is non-universal, changing continuously as a function of the coupling strength $J_{es}$, see
330: Table 2. A similar dependency is found when $J_{es}$ is kept fixed whereas $J_e$ is changed
331: \cite{Ple99}. The
332: findings of Figure 5 and Table 2 may be related to the
333: reported results on two-dimensional Ising models with a ladder
334: defect \cite{Bar79}. The ladder column corresponds to the edge, and the ladder couplings $J_l$ reflect
335: not only the edge-surface interaction $J_{es}$ but also the reduced connectedness to bulk spins
336: at the edge compared to the surfaces. For $J_{es}=J_e=J_s$, the critical exponent of the edge
337: magnetization has the value $\beta_2= 0.095 \pm 0.005$, significantly lower than the
338: critical exponent of the perfect two-dimensional Ising model, $\beta = 1/8$.
339: Of course, the non-monotonic profiles of the line magnetization
340: suggest that a closer analogy between edge properties and descriptions by two-dimensional
341: Ising models with defect lines would require more complicated, extended ladder-type defects.
342:
343: \begin{figure}
344: \centerline{
345: \psfig{figure=figure5.eps,width=7.5cm,angle=270}
346: }
347: \caption
348: {Effective edge exponents as function of the reduced temperature
349: for various values of the coupling $J_{es}$ with $J_e=J_s=2J_b$.
350: Only data not affected by finite-size effects are displayed. The largest system simulated
351: contains $80^3$ spins. Error bars result from averaging over several realizations.}
352: \end{figure}
353:
354: Corner criticality in three-dimensional Ising models deserves to be analyzed at the surface
355: transition as well. As edges are local perturbations acting similar to defect lines in
356: two-dimensional models, the corners of a cube may be interpreted as intersection points of
357: three defect lines.
358:
359: As for the edge, the local magnetization profiles show a
360: non-monotonic behaviour along paths from the edges or corners towards the center of the
361: surface, reflecting the influence of bulk spins. At lower temperatures, the profil
362: is monotonic, with the smallest magnetization at the corners. Close to $T_s$, a monotonic
363: behaviour is also expected, with the highest magnetization at the corners. Due to finite-size
364: effects, this last behaviour is not observed in the simulated finite cubes \cite{Ple00}.
365:
366: The analysis of corner magnetization effective exponents yields again critical exponents
367: changing continuously with the strengths of the local couplings, see Table 2. Note that
368: corner criticality has not been studied for the case $J_{es}=J_s/4$. Changing corner critical
369: exponents are also observed when varying $J_e$ \cite{Ple00}. This non-universal corner criticality
370: is understood by the analogy to two-dimensional Ising models with star defects formed by three intersecting
371: ladder defects \cite{Hen88}.
372:
373: \begin{table}
374: \caption
375: {Edge and corner magnetization critical exponents $\beta_2$ und $\beta_3$
376: at the surface transition for systems with opening angles $\theta=\pi/2$ and
377: $J_e=J_s=2 J_b$.}
378: \vspace*{0.2cm}
379: \begin{tabular}{|c||c|c|c|}
380: \hline
381: & $J_{es}=J_s$ & $J_{es}=J_s/2$ & $J_{es}=J_s/4$ \\
382: \hline\hline
383: $\beta_2$ & $0.095 \pm 0.005$ & $0.176 \pm 0.005$ & $0.244 \pm 0.005$ \\
384: \hline
385: $\beta_3$ & $0.06 \pm 0.01$ & $0.26 \pm 0.02$ & \\
386: \hline
387: \end{tabular}
388: \end{table}
389:
390: \section{Conclusion}
391: Critical phenomena may occur not only in the bulk but also at surfaces, edges, and corners.
392: I have presented results of an extensive Monte Carlo study of edge and corner criticality in three-dimensional
393: Ising systems. At the ordinary transition, edge and corner critical exponents have been computed
394: and compared to analytical predictions. Whereas the critical exponents depend on the wedge angle,
395: they do not depend on microscopic details as the strengths of the local couplings. This is different at
396: the surface transition where coupling dependent edge and corner critical exponents are observed.
397: This non-universal behaviour is understood by analogy with exact results on two-dimensional Ising
398: models with defect lines.
399:
400: \ack
401: It is pleasure to thank Walter Selke for the fruitful and stimulating
402: collaboration on the edge and corner criticality.
403: \begin{thebibliography}{99}
404: \bibitem{Bin83} K.\ Binder,
405: in {\sl Phase Transitions and Critical Phenomena}, Volume 8 (C.\ Domb and
406: J.\ L.\ Lebowitz, eds.), Academic Press, London, and New York (1983).
407: \bibitem{Die86} H.\ W.\ Diehl, in {\sl Phase Transitions and Critical Phenomena}, Volume 10 (C.\ Domb and
408: J.\ L.\ Lebowitz, eds.), Academic Press, London, and New York (1986);
409: H.\ W.\ Diehl, Int.\ J.\ Mod.\ Phys.\ B {\bf 11}, 3503 (1997).
410: \bibitem{Dos92} H.\ Dosch, {\sl Critical Phanomena at Surfaces and Interfaces},
411: Springer, Berlin, Heidelberg, and New York (1992).
412: %\bibitem{Die97} H.\ W.\ Diehl, Int.\ J.\ Mod.\ Phys.\ B {\bf 11}, 3503 (1997).
413: \bibitem{Bin84} K.\ Binder and D.\ P.\ Landau, Phys.\ Rev.\ Lett. {\bf 52}, 318 (1984).
414: \bibitem{Rug93} C.\ Ruge, S.\ Dunkelmann, F.\ Wagner, and J.\ Wulf,
415: J.\ Stat.\ Phys.\ {\bf 73}, 293 (1993).
416: \bibitem{Car83} J.\ L.\ Cardy, J.\ Phys.\ A {\bf 16}, 3617 (1983).
417: \bibitem{Igl93} F.\ Igl\'{o}i, I.\ Peschel, and L.\ Turban, Adv.\ Phys.\ {\bf 42}, 683 (1993).
418: \bibitem{Bar79} R.\ Z.\ Bariev, Sov.\ Phys.\ JETP {\bf 50}, 613 (1979).
419: \bibitem{Ple98} M.\ Pleimling and W.\ Selke, Eur.\ Phys.\ J.\ B {\bf 5}, 805 (1998).
420: \bibitem{Ple99} M.\ Pleimling and W.\ Selke, Phys.\ Rev.\ B {\bf 59}, 65 (1999).
421: \bibitem{Ple00} M.\ Pleimling and W.\ Selke, Phys.\ Rev.\ E {\bf 61}, 933 (2000).
422: \bibitem{Wol89} U.\ Wolff, Phys.\ Rev.\ Lett.\ {\bf 62}, 361 (1989).
423: \bibitem{Ple98a} M.\ Pleimling and W.\ Selke, Eur.\ Phys.\ J.\ B {\bf 1}, 385 (1998).
424: \bibitem{Gut84} A.\ J.\ Guttmann and G.\ M.\ Torrie, J.\ Phys.\ A {\bf 17}, 3539 (1984).
425: \bibitem{Hen88} M.\ Henkel and A.\ Patk\'{o}s, J.\ Phys.\ A {\bf 21}, L231 (1988).
426: \end{thebibliography}
427: \end{document}
428: