1: \documentclass[twocolumn,showpacs]{revtex4}
2:
3: \newcommand{\A}{{\bf A}}
4: \newcommand{\D}{{\bf D}}
5: \renewcommand{\k}{{\bf k}}
6: \newcommand{\n}{{\bf n}}
7: \newcommand{\p}{{\bf p}}
8: \renewcommand{\P}{{\bf P}}
9: \newcommand{\rr}{{\bf r}}
10: \newcommand{\Emin}{E_{\rm min}}
11: \newcommand{\Eq}[1]{Eq.~(\ref{#1})}
12: \newcommand{\Fig}[1]{Fig.~\ref{#1}}
13: \newcommand{\Figure}[1]{Figure~\ref{#1}}
14: \newcommand{\Qmin}{Q_{\rm min}}
15: \newcommand{\xy}{\textsl{XY} }
16:
17: \usepackage{graphicx}
18:
19: \begin{document}
20:
21: \title{A Zero-Temperature Study of Vortex Mobility in\\
22: Two-Dimensional Vortex Glass Models}
23:
24:
25: \author{Petter \surname{Holme}}
26: \author{Peter \surname{Olsson}}
27:
28: \affiliation{Department of Theoretical Physics, Ume\aa\ University,
29: 901 87 Ume{\aa}, Sweden}
30:
31: \begin{abstract}
32: Three different vortex glass models are studied by examining the energy
33: barrier against vortex motion across the system. In the two-dimensional
34: gauge glass this energy barrier is found to increase logarithmically with
35: system size which is interpreted as evidence for a low-temperature phase
36: with zero resistivity. Associated with the large energy barriers is a
37: breaking of ergodicity which explains why the well established results
38: from equilibrium studies could fail. The behavior of the more realistic
39: random pinning model is however different with decreasing energy barriers
40: a no finite critical temperature.
41: \end{abstract}
42:
43: \pacs{64.60.Cn, 75.10.Nr, 74.60.Ge, 74.76.Bz }
44:
45: \maketitle
46:
47: The effect of disorder on the behavior of type-II superconductors in
48: magnetic fields is a subject of enormous interest both theoretically and
49: experimentally. Much discussion has the last few years focused on the
50: possibility of a stable vortex glass phase \cite{Fisher_Fisher_Huse} and
51: some recent papers find evidence from simulations for the existence of such
52: a phase in three dimensions
53: \cite{Olson_Young:00,Kawamura:00,Vestergren_Lidmar_Wallin}. The necessary
54: ingredients of a vortex glass model is, however, still an open question
55: \cite{Huse_Seung,Kawamura:00} and recent results actually suggest that two
56: of the popular models, the gauge glass \cite{Olson_Young:00} and the random
57: pinning model \cite{Vestergren_Lidmar_Wallin}, belong to different
58: universality classes.
59:
60: In two dimensions, which is the focus of the present Letter, the different
61: vortex glass models are commonly believed to behave similarly. One of these
62: is the gauge glass for which equilibrium
63: \cite{Nishimori:94,Fisher_Tokuyasu_Young,Gingras:92,Kosterlitz_Akino}
64: analyses show no transition. In the present paper we approach this model
65: by examining the low-temperature properties of a vortex hopping dynamics.
66: The possibility we explore is that ergodicity breaking might lead to
67: different conclusions when considering dynamics as compared to equilibrium,
68: and we actually find results that strongly suggest a finite-temperature
69: transition. Interestingly, the behavior of the random pinning model is
70: qualitatively different and this points to rich and unexpected behavior of
71: the vortex glass models even in two dimensions.
72:
73: Results in statistical physics often rely on an assumption of ergodicity.
74: Only if this assumption is valid it is permissible to draw conclusions
75: regarding the physically relevant time-averages from the ensemble averages
76: \cite{Binder_Young}. Generally speaking this assumption is expected to be
77: valid for models with a smooth energy landscape, whereas an energy
78: landscape with valleys separated by huge energy barriers, could give rise
79: to a breaking of ergodicity. The concept ergodicity breaking presupposes
80: some kind of dynamics and in the present Letter we examine a vortex hopping
81: dynamics where one vortex moves one lattice constant at a time. This is
82: the kind of dynamics that is typically used for the study of dynamical
83: properties in vortex glass models
84: \cite{Hyman_WFGY,Ying-Hong.Li:92,Kim:gauge-glass}.
85:
86: The possibility of ergodicity breaking means that the standard arguments
87: for the absence of a phase transition could fail and this applies both to
88: the rigorous analytical argument by Nishimori \cite{Nishimori:94} and the
89: vanishing of the zero-temperature domain wall energy with increasing system
90: size \cite{Fisher_Tokuyasu_Young,Gingras:92,Kosterlitz_Akino}. The
91: dynamical studies should, however, still be reliable, but they arrive at
92: conflicting conclusions. Whereas some have reported evidence for a zero
93: temperature transition \cite{Hyman_WFGY} others are strongly in favor of a
94: transition at a finite temperature \cite{Ying-Hong.Li:92,Kim:gauge-glass}.
95: The different conclusions seem to be due to different judgments regarding
96: the reliability of data at rather low temperatures.
97:
98: In this Letter we present a novel zero-$T$ approach designed to measure
99: the height of the energy barriers in vortex glass models. The quantity
100: in focus is the energy barrier against phase slips (defined more
101: precisely below). Beside probing the possibility of ergodicity
102: breaking this quantity is physically relevant since the resistivity
103: in dynamical simulations is inversely proportional to the time between
104: phase slips. A phase slip energy barrier that diverges as
105: $L\rightarrow \infty$ would suggest both breaking of ergodicity and a
106: vanishing resistivity (= immobile vortices) which in turn implies a
107: superconducting low-temperature phase \cite{Kim_Minnhagen_Olsson}.
108:
109: To determine the phase slip energy barrier we start in the ground state and
110: perform an exhaustive search of all relevant local moves, in the vortex
111: representation. Considering the huge phase space of these models this
112: could seem like an impossible task and the only reason why it is at all
113: feasible is that it is sufficient to focus on some relatively few states
114: with rather low energy. Our main result is that the phase slip energy,
115: $V_L$, in the gauge glass model grows logarithmically with system size, a
116: finding that we interpret as evidence for the existence of a
117: superconducting low-temperature phase.
118:
119: The Hamiltonian for the random gauge \xy model is
120: \begin{equation}
121: H = \sum_{\langle ij\rangle} U\left(\theta_i - \theta_j -
122: A_{ij} - \frac{1}{L}\rr_{ij}\cdot {\bf\Delta}\right),
123: \label{Hamiltonian}
124: \end{equation}
125: where $\theta_i$ is the phase at lattice point $i$ on a square lattice of
126: size $L\times L$ (where $L$ is in units of the lattice constant), the sum
127: is over nearest neighbors, and $A_{ij}$ is a randomly chosen vector
128: potential in the interval $[-r\pi,r\pi)$, where $r$ is the disorder
129: strength. The gauge glass model corresponds to the case $r=1$. In the above
130: Hamiltonian we also include the twist variable ${\bf\Delta} =
131: (\Delta_x,\Delta_y)$ \cite{FTBC} to allow for phase slips. The unit vector
132: $\rr_{ij}$ picks $\Delta_x$ or $\Delta_y$ depending on the direction of the
133: link $ij$.
134:
135: The spin interaction is often taken to be a cosines function, but to
136: be able to take advantage of the direct relation to the vortex
137: representation we have instead chosen $U(\phi) = J\phi^2/2$. With the
138: frustration given by
139: \begin{displaymath}
140: f_\rr = \frac{1}{2\pi}\nabla\times\A_\rr,
141: \end{displaymath}
142: and fluctuating twists \cite{FTBC} the vortex Hamiltonian becomes
143: \begin{equation}
144: H^{\rm v} = -\frac{1}{2}\sum_{\rr,\rr'} (q_\rr + f_\rr)
145: G(\rr - \rr') (q_{\rr'} + f_{\rr'})
146: \label{Hv}
147: \end{equation}
148: where the $q_\rr$ are unit charges, the factor $1/2$ compensates for the
149: double counting, and $G(\rr)$ is the lattice Green's function
150: \begin{displaymath}
151: G(\rr) = \left(\frac{2\pi}{L}\right)^2 \sum_{\k\neq0}
152: \frac{e^{i\k\cdot\rr} - 1}{2\cos k_x + 2\cos k_y - 4}.
153: \end{displaymath}
154:
155:
156: For the discussion of the phase slip energy barrier we start by
157: considering this quantity in the pure 2D XY model. The ground state is
158: trivially given by $q_\rr=0$ for all $\rr$. To consider vortex
159: transport we create a vortex pair with a negative vortex at the origin
160: and a positive at $(1,0)$, and then move the positive one to $(2,0)$,
161: $(3,0)$,\ldots $(L,0)$. The twist variable $\Delta_y$ will change
162: proportionally to make the current in the $y$-direction vanish. Since
163: $(L,0)$ in a system with periodic boundary conditions is the same as
164: the origin we are then back again in the ground state but with the
165: twist variable $\Delta_y = -2\pi$. (We use the term phase slip for
166: this kind of process even in the vortex representation.) By these
167: steps we have transported unit vorticity across the system. In this
168: simple case the energy for a configuration with the positive vortex at
169: $(x,0)$ is given by the lattice Green's function, $G(x,0)$, and the
170: phase slip energy barrier for the pure 2D XY model is the value of
171: this energy at the largest separation, $V_L = G(L/2,0)$.
172:
173: For the following we note that each elementary move may be described by
174: a position $\rr$ and a unit dipole vector $\p$:
175: \begin{displaymath}
176: \begin{array}{rcll}
177: q_\rr - 1 & \rightarrow & q_\rr, & \mbox{put a $-1$ vortex at $\rr$},\\
178: q_{\rr+\p} +1 & \rightarrow & q_{\rr+\p}, \quad & \mbox{put a $+1$ vortex
179: at $\rr+\p$}.
180: \end{array}
181: \end{displaymath}
182: To keep track of the transport of vorticity we introduce the
183: polarization $\P$. An elementary move always changes the polarization,
184: $\P + \p \rightarrow \P$. The vorticity transport discussed above
185: starts in the ground state and ends in the same configuration but with
186: $\Delta P=(L,0)$. The polarization therefore contains some memory of
187: the steps taken and may be used to examine the phase slip.
188:
189: For the gauge glass model the sequence of elementary moves that corresponds
190: to the lowest possible energy barrier is in general much more complex than
191: in the pure XY model. To illustrate this point we show in \Fig{path} such a
192: sequence of moves for a certain disorder realization in a $4\times 4$
193: system. In this case six steps are needed to transport vorticity across
194: the system and return back to the ground state. We now describe the method
195: used to find such complex paths.
196:
197: \begin{figure}
198: \includegraphics[width=8.2truecm]{path_bw.eps}
199: \caption{Illustration of a sequence of elementary moves for a
200: realization of a $4\times 4$-system, corresponding to transport
201: of vorticity across the system, $\Delta \P = (4,0)$. The upper
202: left corner is the ground state. Charge is symbolized by the
203: shading of the squares, black represents $-0.67$, white represents
204: $+0.72$, and intermediate charges are proportionally grey. The digits
205: show the polarization relative to the ground state. The charge
206: symbols represent the dipole excited in the next step. The bar next
207: to the square shows the energy of each configuration relative to
208: the ground state.}
209: \label{path}
210: \end{figure}
211:
212: We first search for the ground state by applying the standard spin quench
213: algorithm \cite{Walker_Walstedt} a large number of times. The lowest
214: energy state is taken to be the true ground state when it has been found
215: $N_{\rm rep} = 10$ times. No change was found in the results when we
216: instead used $N_{\rm rep} = 100$ for a set of $10^4$ disorder
217: configurations with $L=4$. We then use an algorithm that is analogous to
218: the filling of a (energy) landscape with a liquid rising from a source at
219: the lowest position. At each time step it is the lowest level accessible by
220: the liquid that is invaded. The disorder configurations give rise to
221: different periodic energy landscapes in polarization space, and the liquid
222: is made to rise until there is a connection between these periodic copies.
223: The landscape picture is, however, a great over-simplification. For each
224: value of the polarization there is a large number of possible
225: configurations with different energy. The true picture is therefore more of
226: a plumber's nightmare with a large network of pipes criss-crossing the
227: polarization space but the intuitive picture still holds. With an
228: algorithm that slowly increases the level one is guaranteed that the first
229: connection found is the lowest possible.
230:
231: In the computer, the algorithm consists of repeating the following
232: steps:
233: \begin{enumerate}
234: \item Generate $4L^2$ configurations by applying the $4L^2$ possible dipole
235: excitations to the current configuration.
236: \item Calculate the energy of each such configuration and put them in a sorted
237: list, lowest energy first, together with their polarization relative to the
238: ground state.
239: \item Take the first (lowest energy) configuration from this list to be the new
240: current configuration.
241: \item If this configuration has already been encountered, but with a
242: different polarization such that $\Delta\P = (\pm L,0)$ then we are
243: done. Otherwise, go to step~1.
244: \end{enumerate}
245:
246: To make this algorithm work one also needs a list of already used
247: configurations. In step~1 the current configuration is added to that list
248: and in step~4 each configuration is compared to the list. The main output
249: is the phase slip energy barrier which equals the highest energy used in
250: the algorithm. The data below are obtained by averaging the energy barrier
251: from 3000 to $2\times 10^5$ disorder configurations. Much as expected, the
252: number of iterations, $N$, of the above algorithm grows rapidly with $L$,
253: $N \approx 0.5 \exp L$. In our simulations the spin quench algorithm that
254: is used to find the ground state is however more time consuming than the
255: determination of the energy barrier.
256:
257:
258: \Figure{Vl-gg} shows $V_L$ versus $L$ for several different values of
259: $r$. We first focus on the results for $r=1$, the usual gauge glass
260: model, shown by solid diamonds in panel (a). By plotting the data
261: with log scale on the $x$ axis it is found that the points with
262: $L\geq3$ to an excellent approximation fall on a straight line with a
263: positive slope. We consider this to be strong evidence for an energy
264: barrier that diverges with increasing lattice size and thereby also
265: evidence for the existence of a low-temperature phase with immobile
266: vortices. As shown in the figure the scaling holds down to $L=3$.
267: This is the same size as in the examinations of the domain wall energy
268: in the gauge glass \cite{Kosterlitz_Akino} and it therefore seems likely
269: that the logarithmic increase in \Fig{Vl-gg} is the true behavior at
270: large $L$. As further support for this belief we note that the
271: logarithmic size-dependence is not a foreign behavior, but is rather
272: built into the model at the outset through the logarithmic
273: length-dependence in the vortex interaction.
274:
275: \begin{figure}
276: \includegraphics[width=8.5truecm]{ggeb.ps}
277: \includegraphics[width=8.5truecm]{pf.ps}
278: \caption{Size dependence of the phase slip energy $V_L$ for the
279: random gauge \xy model. The result for the gauge glass ($r=1$) are
280: shown by the solid diamonds in panel (a) whereas the results for the
281: ordinary 2D \xy model ($r=0$) are shown by crosses in panel (b). We
282: always find $V_L \sim \log L$ which suggests diverging energy barriers
283: against vortex motion and a superconducting low-temperature phase.
284: Panel (c) shows the slope of the lines in (b). There are error bars
285: corresponding to one standard deviation for all the data points, but
286: most of them are far too small to be visible. }
287: \label{Vl-gg}
288: \end{figure}
289:
290: For weaker disorder, $r<1$, the general feature is the same with $V_L \sim
291: \log L$. Considering the change in behavior as $r$ decreases we find that
292: $V_L$ initially changes very slowly. The values of $V_L$ for $r=0.9$ and
293: 0.8 (not shown) are almost identical to the ones for $r=1$. For an analysis
294: of the data for $r\leq 0.7$ we fit $V_L = a(r) + b(r)\ln L$, shown by solid
295: lines. When $r$ is reduced below $\approx0.8$, $a(r)$ slowly increases but
296: the slope $b(r)$ at first remains unchanged, \Fig{Vl-gg}c. The slope
297: actually to a good approximation remains constant down to $r=0.5$ where it
298: starts to increase and finally approaches $b(0)=2\pi$ of the pure 2D \xy
299: model. We believe that the behavior of the slope $b$ shown in panel (c) is
300: related to the finding \cite{Kosterlitz_Simkin,Maucourt_Grempel:97} of a
301: phase with quasi long range order (without ergodicity breaking) for
302: $r<r_c\approx 0.37$. \Figure{Vl-gg}c is also very similar to the phase
303: diagram for the same model recently obtained on the basis of Monte Carlo
304: simulations \cite{Holme_Kim_Minnhagen}.
305:
306:
307: Since the gauge glass model is an often used model of a disordered
308: superconductor in an applied magnetic field, the above result would
309: quite surprisingly seem to suggest the existence of a dissipation-free
310: low temperature phase in disordered thin superconducting films. To
311: examine this question in more detail we have also studied the phase
312: slip energy in the random pinning model introduced in
313: Ref.~\cite{Hyman_WFGY}, which is meant to be a more realistic model of
314: a disordered superconductor in an applied magnetic field. The
315: difference is that the random pinning model has a constant non-zero
316: magnetic field at each plaquette whereas the random gauge \xy model
317: instead is a superconductor with random fields that sum up to zero.
318: The Hamiltonian for the random pinning model is
319: \begin{equation}
320: H_{\rm rp} = -\frac{1}{2}\sum_{\rr,\rr'} (q_\rr - f) G(\rr - \rr')
321: (q_{\rr'} - f) -\sum_\rr v_\rr q_\rr^2.
322: \label{Hrp}
323: \end{equation}
324: The frustration (magnetic field) is here homogeneous and the disorder
325: is instead included through the pinning potential $v_\rr$. We follow
326: Ref.~\cite{Hyman_WFGY} and let $v_\rr$ be a random variable
327: uniformly distributed between $-\pi$ and $\pi$, and restrict the
328: possible values for the vorticity to $q_\rr = -1, 0, 1$.
329:
330: \begin{figure}
331: \includegraphics[width=8.2truecm]{cgpin.ps}
332: \caption{Size-dependence of $V_L$ for the random pinning model with
333: three different values for the frustration. The energy barrier
334: decreases with increasing $L$.}
335: \label{Vl-rp}
336: \end{figure}
337:
338: \Figure{Vl-rp} shows the size dependence of $V_L$ for the random pinning
339: model with $f=1/3$, $1/4$, and $1/9$. Due to the requirement that $fL^2$
340: has to be an integer it is only possible to get data for a few lattice
341: sizes for each value of $f$. The behavior is here markedly different from
342: what we found in the gauge glass; $V_L$ decreases with increasing lattice
343: size and we therefore conclude that the phase slip energy barrier vanishes
344: in the thermodynamic limit. The results are therefore very different from
345: Ref.~\cite{Hyman_WFGY} where it was concluded from a scaling analysis of
346: the $I$-$V$ characteristics that the gauge glass model and the random
347: pinning model do belong to the same universality class. Interestingly,
348: recent simulations strongly suggest that the three-dimensional versions of
349: these two models belong to different universality classes
350: \cite{Olson_Young:00,Vestergren_Lidmar_Wallin}.
351:
352: We also shortly mention our results for a generalized \xy spin glass,
353: which is given by the same Hamiltonian as the gauge glass but with
354: only two possible values for the vector potential, $A_{ij} = 0$,
355: $\pi$. Let $s$ denote the fraction of links with $A_{ij} =\pi$. Due to
356: the chiral symmetry every ground state has a chiral mirror image
357: $q_\rr \mapsto -q_\rr$ that is also a ground state. We may therefore
358: define a chiral energy barrier $V_L^c$, as the energy barrier that has
359: to be climbed to reach the chirally mirrored ground state.
360:
361: \begin{figure}
362: \includegraphics[width=8.5truecm]{xyeb.ps}
363: \caption{The energy barrier for $2\pi$ phase slip $V_L$ (a) and chiral
364: mirroring $V_L^c$ (b) for the generalized $2\pi$ \xy spin glass. For the
365: ordinary \xy model ($s=0$) $V_L$ is shown in \Fig{Vl-gg} while
366: $V_L^c$ would be identically zero.}
367: \label{xyspinglass}
368: \end{figure}
369:
370: The phase slip energy $V_L$ is shown in \Fig{xyspinglass}a. For $s\geq
371: 0$ $V_L$ is independent of size. For $s=0.05$ the data seems to go as
372: $\log L$ suggesting a critical $s_c>0.05$, but we cannot exclude the
373: possibility that $V_L$ saturates and approaches a constant for
374: $L\geq5$. The behavior of the chiral energy barrier in
375: \Fig{xyspinglass}b is different, suggesting $V_L^c \sim \log L$ for all
376: $s>0$ which means that the system at low temperatures and with only
377: local moves is trapped in the part of the configuration space with the
378: given chirality.
379:
380: To conclude, we have examined the existence of growing energy barriers
381: and thereby a possible breaking of ergodicity in three different 2D
382: vortex glass models. By examining the phase slip energy barrier $V_L$
383: we found different behaviors in these three models. In the gauge
384: glass $V_L$ increases logarithmically which we take to suggest the
385: existence of a low-temperature phase with zero resistivity. In the
386: more realistic random pinning model $V_L$ instead decreases in
387: accordance with a transition at zero temperature. Yet another behavior
388: is found in the \xy spin glass where $V_L$ is independent of $L$.
389:
390: We thank B. J. Kim, P.~Minnhagen, and S.~Teitel for valuable discussions.
391: This work was supported by the Swedish Natural Science Research Council
392: through Contract No.\ E 5106-1643/1999.
393:
394: %\bibliography{lett,../../o}
395: \begin{thebibliography}{19}
396: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
397: \expandafter\ifx\csname bibnamefont\endcsname\relax
398: \def\bibnamefont#1{#1}\fi
399: \expandafter\ifx\csname bibfnamefont\endcsname\relax
400: \def\bibfnamefont#1{#1}\fi
401: \expandafter\ifx\csname citenamefont\endcsname\relax
402: \def\citenamefont#1{#1}\fi
403: \expandafter\ifx\csname url\endcsname\relax
404: \def\url#1{\texttt{#1}}\fi
405: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
406: \providecommand{\bibinfo}[2]{#2}
407: \providecommand{\eprint}[2][]{\url{#2}}
408:
409: \bibitem[{\citenamefont{Fisher et~al.}(1991{\natexlab{a}})\citenamefont{Fisher,
410: Fisher, and Huse}}]{Fisher_Fisher_Huse}
411: \bibinfo{author}{\bibfnamefont{D.~S.} \bibnamefont{Fisher}},
412: \bibinfo{author}{\bibfnamefont{M.~P.~A.} \bibnamefont{Fisher}},
413: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{D.~A.} \bibnamefont{Huse}},
414: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{43}},
415: \bibinfo{pages}{130} (\bibinfo{year}{1991}{\natexlab{a}}).
416:
417: \bibitem[{\citenamefont{Kawamura}(2000)}]{Kawamura:00}
418: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Kawamura}},
419: \bibinfo{journal}{J. Phys.\ Soc.\ Jpn.} \textbf{\bibinfo{volume}{69}},
420: \bibinfo{pages}{29} (\bibinfo{year}{2000}).
421:
422: \bibitem[{\citenamefont{Olson and Young}(2000)}]{Olson_Young:00}
423: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Olson}} \bibnamefont{and}
424: \bibinfo{author}{\bibfnamefont{A.~P.} \bibnamefont{Young}},
425: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{61}},
426: \bibinfo{pages}{12467} (\bibinfo{year}{2000}).
427:
428: \bibitem[{\citenamefont{Vestergren et~al.}(2001)\citenamefont{Vestergren,
429: Lidmar, and Wallin}}]{Vestergren_Lidmar_Wallin}
430: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Vestergren}},
431: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Lidmar}}, \bibnamefont{and}
432: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Wallin}}
433: (\bibinfo{year}{2001}), \bibinfo{note}{unpublished}.
434:
435: \bibitem[{\citenamefont{Huse and Seung}(1190)}]{Huse_Seung}
436: \bibinfo{author}{\bibfnamefont{D.~A.} \bibnamefont{Huse}} \bibnamefont{and}
437: \bibinfo{author}{\bibfnamefont{H.~S.} \bibnamefont{Seung}},
438: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{42}},
439: \bibinfo{pages}{1059} (\bibinfo{year}{1190}).
440:
441: \bibitem[{\citenamefont{Nishimori}(1994)}]{Nishimori:94}
442: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Nishimori}},
443: \bibinfo{journal}{Physica A} \textbf{\bibinfo{volume}{205}},
444: \bibinfo{pages}{1} (\bibinfo{year}{1994}).
445:
446: \bibitem[{\citenamefont{Fisher et~al.}(1991{\natexlab{b}})\citenamefont{Fisher,
447: Tokuyasu, and Young}}]{Fisher_Tokuyasu_Young}
448: \bibinfo{author}{\bibfnamefont{M.~P.~A.} \bibnamefont{Fisher}},
449: \bibinfo{author}{\bibfnamefont{T.~A.} \bibnamefont{Tokuyasu}},
450: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{A.~P.} \bibnamefont{Young}},
451: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{66}},
452: \bibinfo{pages}{2931} (\bibinfo{year}{1991}{\natexlab{b}}).
453:
454: \bibitem[{\citenamefont{Gingras}(1992)}]{Gingras:92}
455: \bibinfo{author}{\bibfnamefont{M.~J.~P.} \bibnamefont{Gingras}},
456: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{45}},
457: \bibinfo{pages}{7547} (\bibinfo{year}{1992}).
458:
459: \bibitem[{\citenamefont{Kosterlitz and Akino}(1998)}]{Kosterlitz_Akino}
460: \bibinfo{author}{\bibfnamefont{J.~M.} \bibnamefont{Kosterlitz}}
461: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{N.}~\bibnamefont{Akino}},
462: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{81}},
463: \bibinfo{pages}{4672} (\bibinfo{year}{1998}).
464:
465: \bibitem[{\citenamefont{Binder and Young}(1986)}]{Binder_Young}
466: \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Binder}} \bibnamefont{and}
467: \bibinfo{author}{\bibfnamefont{A.~P.} \bibnamefont{Young}},
468: \bibinfo{journal}{Rev. Mod. Phys.} \textbf{\bibinfo{volume}{58}},
469: \bibinfo{pages}{801} (\bibinfo{year}{1986}).
470:
471: \bibitem[{\citenamefont{Hyman et~al.}(1995)\citenamefont{Hyman, Wallin, Fisher,
472: Girvin, and Young}}]{Hyman_WFGY}
473: \bibinfo{author}{\bibfnamefont{R.~A.} \bibnamefont{Hyman}},
474: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Wallin}},
475: \bibinfo{author}{\bibfnamefont{M.~P.~A.} \bibnamefont{Fisher}},
476: \bibinfo{author}{\bibfnamefont{S.~M.} \bibnamefont{Girvin}},
477: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{A.~P.} \bibnamefont{Young}},
478: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{51}},
479: \bibinfo{pages}{15304} (\bibinfo{year}{1995}).
480:
481: \bibitem[{\citenamefont{Li}(1992)}]{Ying-Hong.Li:92}
482: \bibinfo{author}{\bibfnamefont{Y.-H.} \bibnamefont{Li}},
483: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{69}},
484: \bibinfo{pages}{1819} (\bibinfo{year}{1992}).
485:
486: \bibitem[{\citenamefont{Kim}(2000)}]{Kim:gauge-glass}
487: \bibinfo{author}{\bibfnamefont{B.~J.} \bibnamefont{Kim}},
488: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{62}},
489: \bibinfo{pages}{644} (\bibinfo{year}{2000}).
490:
491: \bibitem[{\citenamefont{Kim et~al.}(1999)\citenamefont{Kim, Minnhagen, and
492: Olsson}}]{Kim_Minnhagen_Olsson}
493: \bibinfo{author}{\bibfnamefont{B.~J.} \bibnamefont{Kim}},
494: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Minnhagen}},
495: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Olsson}},
496: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{59}},
497: \bibinfo{pages}{11506} (\bibinfo{year}{1999}).
498:
499: \bibitem{FTBC}
500: A.~Vallat and H.~Beck, Phys.\ Rev.\ B {\bf 50}, 4015 (1994);
501: P.~Olsson, Phys.\ Rev.\ B {\bf 46}, 14598 (1992); P.~Olsson, Phys.\ Rev.\ B
502: {\bf 52}, 4511 (1995); B.~J.~Kim, P.~Minnhagen, and P.~Olsson, Phys.\ Rev.\ B
503: {\bf 59}, 11506 (1999).
504:
505: \bibitem[{\citenamefont{Walker and Walstedt}(1980)}]{Walker_Walstedt}
506: \bibinfo{author}{\bibfnamefont{L.~R.} \bibnamefont{Walker}} \bibnamefont{and}
507: \bibinfo{author}{\bibfnamefont{R.~E.} \bibnamefont{Walstedt}},
508: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{22}},
509: \bibinfo{pages}{3816} (\bibinfo{year}{1980}).
510:
511: \bibitem[{\citenamefont{Kosterlitz and Simkin}(1997)}]{Kosterlitz_Simkin}
512: \bibinfo{author}{\bibfnamefont{J.~M.} \bibnamefont{Kosterlitz}}
513: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{M.~V.}
514: \bibnamefont{Simkin}}, \bibinfo{journal}{Phys. Rev. Lett.}
515: \textbf{\bibinfo{volume}{79}}, \bibinfo{pages}{1098} (\bibinfo{year}{1997}).
516:
517: \bibitem[{\citenamefont{Maucourt and Grempel}(1997)}]{Maucourt_Grempel:97}
518: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Maucourt}} \bibnamefont{and}
519: \bibinfo{author}{\bibfnamefont{D.~R.} \bibnamefont{Grempel}},
520: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{56}},
521: \bibinfo{pages}{2572} (\bibinfo{year}{1997}).
522:
523: \bibitem[{\citenamefont{Holme et~al.}()\citenamefont{Holme, Kim, and
524: Minnhagen}}]{Holme_Kim_Minnhagen}
525: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Holme}},
526: \bibinfo{author}{\bibfnamefont{B.~J.} \bibnamefont{Kim}}, \bibnamefont{and}
527: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Minnhagen}}.
528:
529: \end{thebibliography}
530:
531: \end{document}
532: