1: %\documentstyle[prl,aps,multicol,epsf]{revtex}
2: \documentclass[aps,twocolumn,groupedaddress,floats,showpacs]{revtex4}
3: \usepackage{graphicx}
4: %\usepackage{dcolumn}
5: \usepackage{bm}
6: %\topmargin -5pt
7: \begin{document}
8:
9: \title{Supersolid phase of hardcore bosons on triangular lattice.}
10: \author{Massimo Boninsegni$^{1}$ and Nikolay Prokof'ev$^{2,3}$}
11: \affiliation{ ${^1}$Department of Physics, University of Alberta,
12: Edmonton, Alberta T6G 2J1\\
13: ${^2}$Department of Physics, University of
14: Massachusetts, Amherst, MA 01003 \\
15: ${^3}$Russian Research Center ``Kurchatov Institute'', 123182
16: Moscow }
17:
18: \begin{abstract}
19: We study properties of the supersolid phase observed for hardcore
20: bosons on the triangular lattice near half-integer filling factor,
21: and the phase diagram of the system at finite temperature. We find
22: that the solid order is always of the $(2m,-m',-m')$ with $m$
23: changing discontinuously from positive to negative values at
24: half-filling, in contrast with phases observed for Ising spins in
25: transverse magnetic field. At finite temperature we find two
26: intersecting second-order transition lines, one in the $3$-state
27: Potts universality class and the other of the Kosterlitz-Thouless
28: type.
29:
30: \end{abstract}
31:
32: \pacs{75.10.Jm, 05.30.Jp, 67.40.Kh, 74.25.Dw}
33: \maketitle
34:
35: Since the supersolid state of matter was introduced to physics
36: nearly half a century ago and its theoretical feasibility was
37: demonstrated,\cite{Penrose}
38: there was a long history of
39: experimental attempts to find it in Nature (mostly in $^4$He, see,
40: e.g., Ref. \onlinecite{Meisel}) along with numerical simulations and
41: theoretical predictions for models of interacting lattice bosons.
42: Recent years have seen a renewed interest in this topic. On the one
43: hand, lattice bosons are no longer in the realm of idealized
44: models and can be now studied in controlled experiments with
45: ultra-cold atoms in optical potentials \cite{Greiner}. On the
46: other hand, the non-classical moment of inertia observed for solid $^4$He
47: samples in the torsional oscillator experiments by Kim and Chan
48: \cite{Kim} remains largely a mystery.
49:
50: Hardcore bosons on triangular lattice with nearest-neighbor
51: repulsion $V>0$ and hopping $t>0$ represent one of the simplest
52: (and thus most promising from the experimental point of view)
53: models displaying a supersolid phase in an extended region of the
54: phase diagram. The model Hamiltonian is given by:
55: \begin{equation}
56: H=-t\sum_{\langle ij\rangle}(\hat b_i^{\dagger}\hat b_j{\;} + h.c) + V\sum_{\langle ij\rangle} \hat n_i\hat n_j
57: - \mu \sum_{i}\hat n_i\;. \label{H}
58: \end{equation}
59: Here $\hat b_i^{\dagger}$ is the bosonic creation operator,
60: $\hat n_i=\hat b_i^{\dagger}\hat b_i{\;}$, and $\mu$ is the chemical potential.
61: A triangular lattice of $N=L\times L$ sites, with periodic boundary conditions is assumed.
62: The alternative formulation of (\ref{H}) in terms of quantum spin-1/2 variables
63: $\hat s_i$, namely
64: \begin{equation}
65: H=-2t\sum_{
66: \langle ij\rangle}(\hat s_i^{x}\hat s_j^{x} +\hat s_i^{y}\hat s_j^{y} ) + V\sum_{\langle ij\rangle}
67: \hat s_i^{z}\hat s_j^{z} - (\mu-3V) \sum_{i}\hat s_i^{z}\ \label{Hs}
68: \end{equation}
69: provides a useful mapping to the XXZ-magnet. The superfluid state
70: of Eq.~(\ref{H}) for $t>>V$ corresponds to the XY-ferromagnetic
71: state of Eq.~(\ref{Hs}), while the solid state of bosons is
72: equivalent to magnetic order in the $\hat{z}$-direction. At
73: half-integer filling factor, $n(\mu=3V) =1/2$, the model has an
74: exact particle-hole symmetry.
75:
76: A robust confirmation of early mean-field predictions of a
77: supersolid phase in the ground state of (\ref{H}), \cite{Murthy}
78: was obtained by means of Green function Monte Carlo (GFMC)
79: simulations.\cite{Boninsegni} The supersolid phases identified in
80: that study for densities away from half-filling (i.e., for $\mu/V
81: >3$ and $\mu/V<3$), can be viewed as solids, with filling factors
82: $\nu=2/3$ and $\nu=1/3$, doped with holes and particles
83: respectively. In what follows, we denote them as supersolids
84: ${\cal A}$ and ${\cal B}$. Density correlations in ${\cal A}$ and
85: ${\cal B}$ have $\sqrt{3}\times \sqrt{3}$ ordering with the wave
86: vector ${\bf Q}=(4\pi /3 ,0)$. In ${\cal A}$ and ${\cal B}$ the
87: average occupation numbers on three consecutive sites along any of
88: the principal axes follow the sequence $(-2m,m',m')$ and
89: $(2m,-m',-m')$ respectively (it is conventional to count densities
90: from $1/2$ to make connection with the magnetization in the spin
91: language, $m_i=n_i-1/2$), see Fig.~\ref{fig1}.
92:
93: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
94: \begin{figure}[tbp]
95: \centerline{\includegraphics[bb=30 30 500 750, angle=-90, width=1.\columnwidth]{fig1}}
96: %bb=-50 50 750 380 ,width=3.5in,angle=-90]{fig1}}
97: \caption{Schematic phase diagram of Eq.~(\ref{H}) near half-integer
98: filling factor.}
99: \label{fig1}
100: \end{figure}
101: %%%%%%%%%%%%%%%%%%height=2.5in, width=3.in%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
102: %MB
103: The model (\ref{H}) has been investigated in a series of recent
104: papers, making use of advanced numerical
105: techniques.\cite{Kedar,Melko,Wessel} The proposed zero-temperature
106: phase diagram is similar to that of Ref. \cite{Boninsegni}, with
107: the notable addition of a quantum superfluid-supersolid phase
108: transition at $n$=1/2 and $t/V\approx 0.115$ and the stable
109: supersolid state persisting for smaller values of $t/V$. In Ref.
110: \cite{Boninsegni} the system was thought to remain a disordered
111: superfluid for arbitrary $t$/$V$. The discrepancy can be
112: attributed to known limitations of the GFMC method. \cite{Note}
113:
114: Based on field-theoretic, exact diagonalization, and other
115: arguments, Ref.~\cite{Melko} hints at the possibility of the
116: $(m,0,-m)$ density order in the ground state at $n=1/2$ (state
117: ${\cal C}$). These considerations involved, in particular, an
118: analogy between the properties of Eq.~(\ref{Hs}), and those of the
119: Ising antiferromagnet on the triangular lattice, in the presence
120: of a transverse magnetic field \cite{Isakov}. If true, there
121: should exist quantum ${\cal A}-{\cal C}$ and ${\cal C}-{\cal B}$
122: phase transitions away from half-filling and three
123: finite-temperature transitions of the Kosterlitz-Thouless (KT)
124: type.
125: % \cite{Melko,Burkov}.
126: Though Ref.~\cite{Kedar} finds that the ground state is of the
127: ${\cal A}$ or ${\cal B}$ type, it makes
128: similar predictions for the finite temperature phase diagram at
129: $n=1/2$ which follow from the assumption that spontaneous symmetry
130: breaking between ${\cal A}$, ${\cal B}$, and their lattice
131: translations is described by the six-clock model \cite{Jose}.
132:
133: In what follows, we provide strong evidence that the supersolid
134: state at half-filling is always of either the ${\cal A}$ or ${\cal
135: B}$ type. Our data suggest that there is a discontinuous
136: transition from ${\cal A}$ to ${\cal B}$ at $\mu=3V$ similar to
137: the I-order phase transition (driven by the large energy of the
138: ${\cal A-B}$ domain walls). What makes it special is the exact
139: particle-hole symmetry; structure factor, superfluid density, and
140: energy remain continuous functions of $\mu$ through the transition
141: line. For the supersolid ${\cal A}$ (or ${\cal B}$) with the
142: three-fold degenerate ground state, one expects to see the
143: normal-superfluid KT and the solid-liquid $3$-state Potts
144: transitions, as temperature is increased. Moreover, the KT and
145: Potts transitions are independent of each other and for $n \ne
146: 1/2$ intersect on the phase diagram. The failure of the mean-field
147: description and analogies with the transverse-field Ising model to
148: predict the supersolid structure at $n=1/2$ can be traced back to
149: the U(1)-symmetry of Eqs.~(\ref{H}) and (\ref{Hs}), as noticed in
150: \cite{Kedar}. For example, the $(1,0,-1)$ state can {\it not} be
151: the true groundstate at finite $t$ in the limit of $t/V \to 0$
152: simply because it does not respect the particle conservation law.
153:
154: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
155: \begin{figure}[tbp]
156: %\centerline{\includegraphics[bb=0 80 800 600 ,width=3.5in,angle=-90]
157: \centerline{\includegraphics[bb=30 30 570 790, angle=-90, width=1.\columnwidth]{fig2}}
158: \caption{ (Color online). Probability distributions $P(n^{+}) $ for different system sizes
159: and temperatures at $\mu/V=3$ and $t/V=0.1$.}
160: \label{fig2}
161: \end{figure}
162: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
163:
164: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
165: \begin{figure}[tbp]
166: %\centerline{\includegraphics[bb=0 80 800 600 ,width=3.5in,angle=-90]
167: \centerline{\includegraphics[bb=30 30 570 790, angle=-90, width=1.\columnwidth]{fig3}}
168: \caption{ (Color online). Probability distributions $P(n^+) $ for
169: different system sizes
170: and temperatures at $\mu/V=3$ and $t/V=0.05$. }
171: \label{fig3}
172: \end{figure}
173: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
174:
175: We use the worm-algorithm Monte Carlo scheme in the lattice path-integral
176: representation \cite{worm} to simulate Eq.~(\ref{H}).
177: Since the structure factor
178: \begin{equation}
179: S({\bf Q})= \biggl \langle \bigg| \sum_{k=1}^{N} \hat n_k\
180: e^{i{\bf Qr}_k} \bigg|^2 \biggr \rangle/N^2
181: \end{equation}
182: does not distinguish between supersolids ${\cal A,\ B,\ C}$,
183: we adopt the following strategy: for
184: each system configuration, we compute the distribution of
185: time-averaged occupation numbers,
186: $\bar{n}_k=\beta^{-1} \int_0^{\beta } d\tau\ \hat n_k(\tau )$,
187: and use it to determine the fraction of sites with $\bar{n}_k>1/2$
188: \begin{equation}
189: n^{+}= N^{-1} \sum_{k=1}^{N} \theta ( \bar{n}_k -1/2 )\;,
190: \label{np}
191: \end{equation}
192: where $\theta(x)$ is the Heviside function. ${\cal A,\ B,\ C}$
193: density structures correspond to $n^{+}_{A}=2/3$, $n^{+}_{C}=0$,
194: and $n^{+}_{B}=1/3$. Finite systems are characterized by broad
195: probability distributions $P(n^+)$, and the formation of different
196: solid orders can be seen as the development of sharp peaks, as the
197: thermodynamic limit is approached.
198:
199: In Fig.~\ref{fig2} we show the evolution of the $P(n^{+})$
200: distribution for the half-filled system at $V/t=10$, i.e., close
201: to the superfluid-supersolid transition point,
202: estimated\cite{Kedar,Melko,Wessel} at $V/t \approx$ 8.5. The
203: distribution is peaked at $n^{+}=0$ in the smallest system
204: considered ($L$=6), but, as the system size is increased, the
205: weight is shifted toward the wings of the distribution. For
206: $L$=18, there are already three peaks with comparable height.
207: Finally, in the $L$=24 system we observe only two peaks
208: corresponding to the supersolid phases ${\cal A}$ and ${\cal B}$.
209: Though the probability density between the peaks is still
210: measurable, the dynamics of the algorithm becomes very slow; it
211: typically takes millions of Monte Carlo sweeps, in order for the
212: system to make a transition from the ${\cal A}$ to the ${\cal B}$
213: structure and vice versa. We have explicitly verified that
214: configurations with $n^{+} \approx 2/3$ and $n^{+} \approx 1/3$
215: have density orders depicted as in Fig.~\ref{fig1}, with a large
216: contrast in density between sublattices. We have also checked that
217: the $V/t=10$, $L$=48, $T$ = $t/$50 system spontaneously develops
218: either ${\cal A}$ or ${\cal B}$ order, starting from the initial
219: configuration corresponding to the superfluid phase at $V/t$=5.
220:
221: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
222: \begin{figure}[tbp]
223: %\centerline{\includegraphics[bb=0 80 800 600 ,width=3.5in,angle=-90]
224: \centerline{\includegraphics[bb=30 30 570 790, angle=-90, width=1.\columnwidth]{fig4}}
225: \caption{(Color online). Superfluid density in the vicinity of the KT transition for $t/V=0.1$
226: and $\mu/V=2.74$. The solid line is the thermodynamic curve calculated
227: using Eq.~(\ref{R}) with $\kappa (T)$ deduced from the plot shown in the inset.
228: Inset: solutions of the Eq.~(\ref{RG}) for different pairs of system sizes:
229: $L_2=24$, $L_1=12$---filled circles, $L_2=48$, $L_1=12$---open circles
230: $L_2=48$, $L_1=24$---filled squares. The dashed line is the linear fit
231: $\kappa=1+1.03(T_c-T)/t$ with $T_c/t=0.50$.}
232: \label{fig4}
233: \end{figure}
234: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
235:
236: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
237: \begin{figure}[tbp]
238: %\centerline{\includegraphics[bb=0 80 800 600 ,width=3.5in,angle=-90]
239: \centerline{\includegraphics[bb=30 30 570 790, angle=-90, width=1.\columnwidth]{fig5}}
240: \caption{(Color online). Structure factor in the vicinity of the 3-state Potts transition for $t/V=0.1$
241: and $\mu/V=2.74$.
242: Inset: data collapse using exact critical exponents for the 3-state Potts model
243: \cite{Alexander} and $\delta =(T-T_c)/t$ with $T_c/t=1.035$.}
244: \label{fig5}
245: \end{figure}
246: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
247:
248: In Fig.~\ref{fig3} we show what happens at larger values of $V/t$.
249: Now, the central peak is already absent in relatively small $L$=12 and $L$=18 systems.
250: We thus conclude that the nature of the supersolid state at half-integer filling factor
251: is determined by the ${\cal A}$ and ${\cal B}$ structures,
252: for {\it all} values of $t/V$ for which a supersolid phase exists.
253:
254: If spontaneous symmetry breaking of the ground state degeneracies
255: is described by the six-clock model \cite{Jose}, one should
256: observe three finite-temperature transitions for systems near
257: half-filling, and a solid phase with algebraic correlations
258: ``sandwiched" between the solid and normal liquid phases. This
259: prediction was made in Ref.~\cite{Kedar} for $n=1/2$. Since the
260: ground state was found here to be only of the ${\cal A}$ or ${\cal
261: B}$ type, and we do not see why domain wall energies between
262: translated ${\cal A}$ states are the same as between ${\cal A}$
263: and ${\cal B}$ states (in fact, the Landau theory prediction
264: \cite{Melko,Kedar,Burkov} is that ${\cal A}$ and ${\cal B}$ states
265: phase separate and have different average densities even at
266: $\mu=3V$), the finite temperature phase diagram should instead
267: feature the normal-superfluid KT and the liquid-solid 3-state
268: Potts (for $n \ne 1/2$) transitions breaking U(1) and translation
269: symmetry respectively. At $n=1/2$ we expect only one liquid-solid
270: transition. An interesting question is whether transition lines
271: simply intersect, or there are bicritical and tricritical points
272: and I-order lines as observed for the similar model on the square
273: lattice \cite{square}. We performed simulations for two
274: representative cases, one for constant chemical potential
275: $\mu/V=2.74$ (or density $n\approx 0.44$), and the other for
276: constant $t/V=0.1$.
277:
278: In Fig.~\ref{fig4} we show typical data for the KT transition between the solid
279: and supersolid phases. The transition is smeared by logarithmic finite-size effects,
280: but the critical temperature can be still determined with good accuracy
281: by utilizing the well known renormalization flow and the universal jump of the
282: superfluid density, $\rho_s$, at $T_c$. The data analysis is as follows:\cite{weak2D}
283: we define $R=\pi \rho_s /2mT$ (where $m=1/3t$ is the effective mass for the
284: triangular lattice) and study the finite-size scaling of the data using KT renormalization
285: group equations in the integral form
286: \begin{equation}
287: 4 \ln (L_2/L_1) = \int_{R_2}^{R_1} {dt \over t^2(\ln (t) - \kappa )+t } \;.
288: \label{RG}
289: \end{equation}
290: The microscopic (system size independent) parameter $\kappa$ is an
291: analytic function of temperature, and the critical point corresponds to
292: $R=1$ at $\kappa=1$. For $T<T_c$, the thermodynamic curve is defined by the
293: equation
294: \begin{equation}
295: 1/R+{\rm ln} \ R=\kappa (T) \;,
296: \label{R}
297: \end{equation}
298: with $\kappa = 1 + \kappa ' (T_c-T)$. We use different pairs of system sizes
299: in Eq.~(\ref{RG}) to determine the $\kappa (T)$ curve, and obtain the location
300: of the critical point from $\kappa (T_c)=1$. The results are shown in the inset of
301: Fig.~\ref{fig4}. Data collapse and smooth analytic behavior of $\kappa (T)$ proves that
302: the transition is indeed of the KT type. We used the same protocol and system sizes
303: to determine other critical points.
304:
305: In Fig.~\ref{fig5}, we present our data for the transition into the state with the long-range
306: density order. For the three-fold degenerate ${\cal B}$ structure this transition
307: is expected to be in the 3-state Potts universality class. The critical exponents are
308: known exactly \cite{Alexander}: $\nu =5/6$, and $\beta = 1/9$. We thus perform the
309: data collapse using $L^{2\beta}S_{\bf Q} = f(\delta L^{1/\nu})$ where $\delta =(T-T_c)/t$ and
310: $T_c$ is the only fitting parameter. The result is shown in the inset of Fig.~\ref{fig5}.
311: This confirms the above-mentioned expectation, and establishes that there is only
312: one transition to the solid phase (there are no visible finite-size effects below $T_c$).
313:
314: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
315: \begin{figure}[tbp]
316: %\centerline{\includegraphics[bb=0 80 800 600 ,width=3.5in,angle=-90]
317: \centerline{\includegraphics[bb=30 30 570 790, angle=-90, width=1.\columnwidth]{fig6}}
318: \caption{(Color online). Finite temperature phase diagram for two representative cuts:
319: the left panel is for fixed $\mu/V=2.74$; the right panel is for fixed $t/V=0.1$.
320: The solid line indicates a degenerate I-order transition line between supersolids
321: ${\cal B} $ and ${\cal A} $. }
322: \label{fig6}
323: \end{figure}
324: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
325:
326: Finally, we compute the phase diagram in the $(T/t,t/V)$ (at
327: constant $\mu/V=2.74$) and $(T/t,\mu/V)$ (at constant $t/V=0.1$)
328: planes and observe that KT and Potts transition lines form a
329: simple cross for $n \ne 1/2$, i.e., the corresponding order
330: parameter fields are not strongly interacting, see
331: Fig.~\ref{fig6}. The transition temperature to the superfluid and
332: supersolid states in this part of the phase diagram is determined
333: by the hopping amplitude. Within the statistical uncertainties of our
334: calculation, KT and Potts transition temperatures cannot be distinguished
335: at $\mu/V$=3.
336:
337: %One interesting physical aspect of the finite-temperature
338: %phase diagram is the predicted possibility of observing a
339: %supersolid phase either by cooling a superfluid, which freezes
340: %into a supersolid phase (for $t$/$V$ $>$ 0.11), or by cooling a
341: %normal solid, which then features a KT transition to a supersolid.
342: We did not see evidence for the algebraic solid state at $\mu=3V$.
343: The finite-size scaling for the supersolid-solid transition at
344: $\mu=3V$ is consistent with the $3$-state Potts universality, though
345: the data collapse is not as impressive as in Fig.~\ref{fig4}
346: (the other alternative is the KT transition).
347:
348: It is instructive to understand why the $(m,0,-m)$ phase for the
349: Hamiltonian (\ref{H}) is not an obvious groundstate. At the
350: mean-field level, ${\cal C}$ has a better energy than ${\cal A}$
351: or ${\cal B}$. For the transverse-field Ising model \cite{Isakov}
352: the $(1,0,-1)$ spin arrangement is obtained by orienting the
353: middle spin along the magnetic field direction, i.e. putting it in
354: the equal-amplitude superposition of up- and down-states. In
355: bosonic language, it corresponds to the superposition of states
356: with one or zero particles on a given site. Such a state can not
357: be reconciled with the Hamiltonian (\ref{H}) which conserves the
358: particle number. Any non-integer average occupation number
359: necessarily involves hopping transitions to the nearest neighbor
360: sites. In the $(1,0,-1)$ structure the middle site is completely
361: surrounded by the fully occupied or empty sites and thus can not
362: be the ground state of the system even in the limit of $t/V \to
363: \infty$. The problem appears to be inherently quantum with no
364: obvious solution at the mean-field level.
365:
366: We are grateful M. Troyer, A. Paramekanti, K. Damle, and A. Burkov
367: for stimulating discussions. The research was supported by the
368: National Science Foundation under Grant No. PHY-0426881, by the
369: Petroleum Research Fund of the American Chemical Society under
370: research grant 36658-AC5, and by the Natural Science and
371: Engineering Research Council of Canada under research grant
372: G121210893.
373:
374: \begin{thebibliography}{99}
375:
376: \bibitem{Penrose} O. Penrose and L. Onsager,
377: Phys. Rev. {\bf 104}, 576 (1956);
378: A.F. Andreev and I.M. Lifshitz,
379: Sov. Phys. JETP {\bf 29}, 1107 (1960);
380: G. Chester, Phys. Rev. A {\bf 2}, 256 (1970);
381: A.J. Leggett, Phys. Rev. Lett. {\bf 25}, 1543 (1970).
382:
383: \bibitem{Meisel} M.W. Meisel, Physica B {\bf 178}, 121 (1992).
384:
385: \bibitem{Greiner} M. Greiner, C.A. Regal, and D.S. Jin, Nature {\bf 415} 39 (2002).
386:
387: \bibitem{Kim} E. Kim and M. H. W. Chan, Nature {\bf 427}, 225 (2004);
388: Science {\bf 305}, 1941 (2004).
389:
390: \bibitem{Murthy} G. Murthy, D. Arovas, and A. Auerbach,
391: Phys. Rev. B {\bf 55}, 3104 (1997).
392:
393: \bibitem{Boninsegni} M. Boninsegni, J. Low Temp. Phys. {\bf 132}, 39 (2003).
394:
395: \bibitem{Note} Working with a finite population of random walkers in GFMC,
396: has the effect of biasing estimates for the static structure
397: factor, required to establish the presence of diagonal long-range
398: order. Such a bias can in principle be eliminated by increasing
399: the size of the population, but this is often impractical. See,
400: for instance, M. Calandra Bonaura and S. Sorella, Phys. Rev. B {\bf 57},
401: 11446 (1998).
402:
403: \bibitem{Kedar} D. Heidarian and K. Damle, Phys. Rev. Lett. {\bf 97}, 127206
404: (2005).
405:
406: \bibitem{Melko} R.G. Melko, A. Paramekanti, A.A. Burkov, A. Vishwanath,
407: D.N. Sheng, and L. Balents, Phys. Rev. Lett. {\bf 97}, 127207
408: (2005).
409:
410: \bibitem{Wessel} S. Wessel and M. Troyer, Phys. Rev. Lett. {\bf 97}, 127205 (2005).
411:
412: \bibitem{Isakov} R. Moessner and S.L. Sondhi,
413: Phys. Rev. B {\bf 63}, 224401 (2001);
414: R. Moessner, S.L. Sondhi, and P. Chandra,
415: Phys. Rev. B {\bf 64}, 144416 (2001);
416: S. V. Isakov and R. Moessner, Phys. Rev. B {\bf 68}, 104409 (2003).
417:
418: \bibitem{Burkov} A.A. Burkov and L. Balents, cond-mat/0506457.
419:
420: \bibitem{Jose} J. Jose, L. Kadanoff, S. Kirkpatrick, and D. Nelson,
421: Phys. Rev. B {\bf 16}, 1217 (1977).
422:
423: \bibitem{worm} N.V. Prokof'ev, B.V. Svistunov, and I.S. Tupitsyn,
424: Phys. Lett. {\bf 238}, 253 (1998);
425: Sov. Phys. JETP {\bf 87}, 310 (1998).
426:
427: \bibitem{square}
428: G.G. Batrouni, and R.T.Scalettar, Phys. Rev. Lett. {\bf 84}, 1599 (2000);
429: F. H\'ebert, G.G. Batrouni, R.T. Scalettar, G. Schmid, M. Troyer, and
430: A. Dorneich, Phys. Rev. B {\bf 65}, 014513 (2001); G. Schmid, S. Todo,
431: M. Troyer, and A. Dorneich, Phys. Rev. Lett. {\bf 88}, 167208 (2002);
432: K. Bernardet, G.G. Batrouni, J.-L. Meunier, G. Schmid, M. Troyer, and
433: A. Dorneich, Phys. Rev. B {\bf 65}, 104519 (2002).
434:
435: \bibitem{weak2D} N. ProkofÕev and B. Svistunov, Phys. Rev. A {\bf 66}, 043608 (2002).
436:
437: \bibitem{Alexander} S. Alexander, Phys. Lett. A {\bf 54}, 353 (1975).
438:
439:
440: \end{thebibliography}
441:
442: \end{document}
443: