cond-mat0611343/jq.tex
1: \documentclass[10pt,prl,aps,twocolumn,floatfix,showpacs]{revtex4}
2: \usepackage{graphicx}
3: \usepackage{amssymb}
4: \usepackage{bm}% bold math
5: \newcommand{\dn}{\downarrow}
6: \newcommand{\up}{\uparrow}
7: 
8: \begin{document}
9: 
10: \title{Evidence for deconfined quantum criticality \\
11: in a two-dimensional Heisenberg model with four-spin interactions}
12: 
13: \author{Anders W. Sandvik} 
14: \affiliation{Department of Physics, Boston University, 
15: 590 Commonwealth Avenue, Boston, Massachusetts 02215}
16: 
17: \date{\today}
18: 
19: \begin{abstract}
20: Using ground-state projector quantum Monte Carlo simulations in the valence bond basis, 
21: it is demonstrated that non-frustrating four-spin interactions can destroy the N\'eel order 
22: of the two-dimensional $S=1/2$ Heisenberg antiferromagnet and drive it into a valence-bond 
23: solid (VBS) phase. Results for spin and dimer correlations are consistent with a single 
24: continuous transition, and all data exhibit finite-size scaling with a single set of exponents;
25: $z=1, \nu=0.78 \pm 0.03$, and $\eta=0.26 \pm 0.03$. The unusually large $\eta$ and an 
26: emergent $U(1)$ symmetry, detected using VBS order parameter histograms, provide strong 
27: evidence for a deconfined quantum critical point.
28: \end{abstract}
29: 
30: \pacs{75.10.-b, 75.10.Jm, 75.40.Mg, 75.40.Cx}
31: 
32: \maketitle
33: 
34: Since the discovery in 1986 of high-$T_{\rm c}$ superconductivity in layered cuprates, quantum 
35: phase transitions in two-dimensional (2D) antiferromagnets have formed a central topic in
36: condensed matter physics \cite{chn,sachdevbook}. While superconductivity is induced in the 
37: CuO$_{\rm 2}$ planes of the cuprates by doping with charge carriers, other mechanisms for destroying 
38: the N\'eel order and stabilizing different ground states have also been intensely investigated 
39: theoretically. Considerable efforts have been devoted to possible spin liquid 
40: ("RVB" \cite{and87}) and valence-bond solid (VBS) states driven by magnetic frustration 
41: \cite{rea89,j1j2old,mis05}. This work has been partially motivated by the hope that 
42: an understanding of generic features of quantum phase transitions in 2D antiferromagnets could shed 
43: light also on the mechanisms at work in the cuprates \cite{sac03}. Quantum fluctuation driven 
44: phase transitions are also of broader relevance in the context of strongly correlated 
45: systems \cite{sondhi97}. 
46: 
47: A quantum phase transition occurs as a function of some parameter at temperature $T=0$ and
48: corresponds to a $T>0$ transition in an effective classical system with an imaginary-time 
49: dimension---the path integral  \cite{herzandsuzuki}. 
50: The standard Landau-Ginzburg-Wilson framework for critical phenomena should thus be applicable, with the 
51: dimensionality $d \to d+z$, where the dynamic exponent $z$ depends on the way space and time correlations are 
52: related. In the paradigm prevailing until recently, the ``Landau rules" for the nature of the 
53: transition---continuous or first-order---were also assumed to remain valid for quantum phase transitions. 
54: A direct transition between two ordered phases should thus be generically first-order if two 
55: different symmetries are broken. This notion has recently been challenged by Senthil {\it et al.}, who 
56: argued that quantum phase transitions separating two ordered phases can be generically continuous, even 
57: when different symmetries are broken \cite{sen04}. This theory of ``deconfined" quantum critical 
58: points was first developed for the transition between an antiferromagnetic (AF) and a 
59: valence-bond-solid (VBS) state. Both these states have confined $S=1$ excitations---gapless 
60: magnons and gapped "triplons", respectively. The critical point is 
61: characterized by deconfined $S=1/2$ spinons coupled to an emergent $U(1)$ gauge field \cite{sen04}. 
62: In 2D the deconfined state is unstable and exists only at a point separating the two ordered phases. 
63: The AF and VBS order parameters arise as a consequence of spinon confinement. In this Letter, 
64: quantum Monte Carlo (QMC) results are presented which support this theory. 
65: 
66: Preceding the theory of deconfined quantum critical points, continuous transitions between two 
67: ordered quantum states had been suggested based on numerical simulations \cite{assaad,san02}. 
68: However, in more detailed studies following the theoretical developments it has 
69: proved difficult to confirm their existence. Instead, many studies have pointed to weakly 
70: first-order AF--VBS transitions \cite{kuk0405,kuk06,kagome,mel06,sirker} or other  
71: scenarios inconsistent with deconfined quantum criticality \cite{san06}. To date, large-scale QMC studies 
72: of potential deconfined quantum critical points have focused on spin (or hard-core bosonic) models with 
73: spin-anisotropic interactions \cite{kuk0405,kuk06,kagome,mel06}. Frustrated $SU(2)$ (Heisenberg) 
74: symmetric interactions, which cannot be studied using QMC simulations due to the infamous
75: "sign problem", have been considered in exact diagonalization studies \cite{poilblanc}. Because
76: of the limitations to very small lattices, it has not been possible to study phase transitions 
77: in detail, however. In fact, not even the nature of the VBS state has been completely settled in 
78: basic models such as the J$_{\rm 1}$-J$_{\rm 2}$ frustrated Heisenberg model \cite{j1j2new}. 
79: 
80: Here it will be shown that the AF order of the square-lattice Heisenberg model can be destroyed 
81: also by non-frustrated isotropic interactions accessible to QMC simulations. The following 
82: Hamiltonian will be discussed:
83: \begin{equation}
84: H = J\sum_{\langle ij\rangle} \mathbf{S}_i \cdot \mathbf{S}_j -
85: Q\sum_{\langle ijkl\rangle} (\mathbf{S}_i \cdot \mathbf{S}_j-\hbox{$\frac{1}{4}$})
86: (\mathbf{S}_k \cdot \mathbf{S}_l-\hbox{$\frac{1}{4}$}),
87: \label{jqham}
88: \end{equation}
89: where $\langle ij\rangle$ denotes nearest-neighbor sites and $\langle ijkl\rangle$ refers 
90: to the corners of a plaquette, such that $ij$ and $kl$ form two parallel adjacent horizontal or 
91: vertical links. This interaction contains a subset of the four-site ring-exchange, and with $Q>0$ 
92: there is no QMC sign problem. Note that the purpose here is not to model any specific material, but
93: simply to construct a model system in which an AF--VBS transition can be investigated. It will be 
94: shown below that the ground state of the J-Q model has AF order for $J/Q \agt 0.04$ and VBS order 
95: for $J/Q \alt 0.04$.
96: 
97: \begin{figure}
98: \includegraphics[width=6.15cm, clip]{fig1.eps}
99: \caption{(Color online) Finite size scaling of the squared spin ($M$) and dimer ($D$)
100: order parameters at $J/Q=0$ and $0.1$. The curves are cubic fits. Statistical errors are much 
101: smaller than the symbols.}
102: \label{fig1}
103: \vskip-5mm
104: \end{figure}
105: 
106: To study the ground state of the hamiltonian (\ref{jqham}), an approximation-free projector technique 
107: in the valence bond basis \cite{vbmc} is employed which is ideally suited for multi-spin interactions 
108: formed out of singlet projection operators $(\mathbf{S}_i \cdot \mathbf{S}_j-\hbox{$\frac{1}{4}$})$.
109: Here $L\times L$ lattices with $L$ up to $32$ are considered. Larger systems may be reachable 
110: using loop algorithms in the standard $S^z$ basis, which have been used for $U(1)$ models with 
111: four-site interactions \cite{jkmethod,errornote}. The valence bond basis has its advantages, 
112: however, including an improved estimator for the singlet--triplet gap.
113: 
114: Results will be presented for spin-spin ($s$) and dimer-dimer ($d$) correlation functions,
115: \begin{eqnarray}
116: C_s(\mathbf{r}) & = & \langle \mathbf{S}(\mathbf{0}) \cdot \mathbf{S}(\mathbf{r})\rangle, 
117: \label{cs} \\
118: C_d(\mathbf{r}) & = & \langle [\mathbf{S}(\mathbf{0}) \cdot \mathbf{S}(\hat \mathbf{x})]
119: [\mathbf{S}(\mathbf{r}) \cdot \mathbf{S}(\mathbf{r} + \hat \mathbf{x})]\rangle,
120: \label{cd}
121: \end{eqnarray}
122: where $\hat \mathbf{x}$ denotes a lattice unit vector in the $x$ direction. The AF order parameter
123: is the staggered magnetization, the square of which is calculated;
124: \begin{equation}
125: M^2 = \frac{1}{N}\sum_{\bf r} C_s(\textbf{r})(-1)^{r_x+r_y}.
126: \end{equation}
127: The VBS state can have either columnar or plaquette order, both of which break $Z_4$ symmetry. An 
128: important aspect of the theory is that these order parameters should both exhibit divergent fluctuations 
129: at the deconfined critical point. Only at some length-scale diverging as a power 
130: of the correlation length should one of them be singled out \cite{sen04}. This is analogous to the 
131: irrelevance of $Z_4$ anisotropy in the 3D XY model \cite{jose} and corresponds directly to the 
132: predicted emergent $U(1)$ symmetry. The $\mathbf{q}=(\pi,0)$ 
133: dimer order parameter,
134: \begin{equation}
135: D^2 = \frac{1}{N}\sum_{\bf r} C_d(\textbf{r})(-1)^{r_x},
136: \end{equation}
137: is divergent for both columnar and plaquette VBS order and will be studied here.
138: 
139: Extrapolations of the AF and VBS order parameters, shown in Fig.~\ref{fig1}, demonstrate that 
140: there is long-range VBS order but no AF order  at maximal four-spin interaction; $J/Q=0$ (note that there 
141: are still two-site interactions present when $J=0$; simulations for $J < 0$ are 
142: sign problematic). Also shown are results at $J/Q=0.1$, where the situation is the reverse;
143: there is AF order but the VBS order vanishes. Thus there is an AF--VBS transition
144: somewhere in the range $0 < J/Q < 0.1$, or there could be a region of AF/VBS coexistence (which would 
145: be analogous to a supersolid state). The nature of the VBS order---columnar or plaquette---is not
146: clear from these calculations. However, simulations of open-boundary rectangular lattices, in which 
147: a unique columnar or plaquette pattern can be stabilized \cite{san02}, indicate that columnar order is preferred. 
148: The extrapolated VBS correlation at $J/Q=0$ is $D^2 \approx 0.0024$.
149: 
150: \begin{figure}
151: \null\includegraphics[width=6.05cm, clip]{fig2.eps}
152: \caption{(Color online) Finite size scaling of the singlet-triplet excitation gap multiplied by 
153: the system size $L$. The behavior at $J/Q=0.04$ corresponds to $z=1$.}
154: \label{fig2}
155: \vskip-5mm
156: \end{figure}
157: 
158: The deconfined theory has dynamic exponent $z=1$ \cite{sen04}. 
159: This exponent can be directly accessed through the finite-size scaling of the singlet--triplet gap; 
160: $\Delta \sim L^{-z}$. To demonstrate consistency with $z=1$, the scaling of $L\Delta$ is shown in 
161: Fig.~\ref{fig2} for $J/Q=0$ and $0.1$, as well as for $J/Q=0.04$ which will be shown below to be close 
162: to criticality. Here $L\Delta$ extrapolates to a non-zero value, supporting $z=1$, and at $J/Q=0$ and 
163: $0.1$ the behaviors are what would be expected off criticality. The inset of Fig.~\ref{fig2} shows an 
164: infinite-size extrapolation of the gap at $J/Q=0$, giving $\Delta/Q\approx 0.07$.
165: 
166: \begin{figure}
167: \includegraphics[width=6.25cm, clip]{fig3.eps}
168: \caption{(Color online) Scaling with $\nu=0.78$ and $g_c=0.04$ of the correlation lengths (a) 
169: and the spin Binder ratio (b).} 
170: \label{fig3}
171: \vskip-5mm
172: \end{figure}
173: 
174: 
175: Correlation lengths $\xi_s$ and $\xi_d$ for spins and dimers are defined in the standard way as the 
176: square-roots of the second moments of the correlation functions (\ref{cs}) and (\ref{cd}). 
177: Also useful is the Binder cumulant, defined for the 
178: spin as $q_s=\langle M^4\rangle/\langle M^2\rangle^2$. Finite-size scaling of these quantities is used to extract 
179: the critical coupling and the correlation length exponent $\nu$. To achieve good data collapse, a subleading 
180: correction is also included. With $g=J/Q$, the scaling ansatz is,
181: \begin{equation}
182: A(g,L)=L^{\kappa}(1+aL^{-\omega})f[(g-g_c)L^{1/\nu}],
183: \label{scaling}
184: \end{equation}
185: where $A=\xi_s,\xi_d$, or $q_s$, and $\kappa=1$ for $\xi_s,\xi_d$ and $0$ for $q_s$. As seen 
186: in Fig.~\ref{fig3}, these quantities can be scaled using $g_c=0.040 \pm 0.003$ and a common $\nu=0.78\pm 0.03$. 
187: In all cases, the subleading exponent $\omega \approx 2$, and the scaling is nearly as good if 
188: $\omega=2$ is fixed throughout. Interestingly, the best prefactor $a$ is then almost equal for 
189: $\xi_s$ and $\xi_d$, $a\approx 8$, but this may be coincidental.
190: 
191: Next, the correlation functions $C_{s,d}(\mathbf{r})$ at the longest lattice distance, $\mathbf{r}=(L/2,L/2)$,
192: are analyzed to extract the correlation function exponent $\eta$. The expected scaling is as in 
193: Eq.~(\ref{scaling}) with $\kappa=-(1+\eta)$. Now $g_c$ and $\nu$ are kept fixed at the values determined 
194: above. As shown in Fig.~\ref{fig4}, a single exponent describes both the spin and dimer data, and in this 
195: case a subleading correction is not needed ($a=0$). The exponent, $\eta = 0.26 \pm 0.03$, is unusually 
196: large. In the 3D $O(3)$ universality class, describing transitions between the AF and a featureless 
197: gapped state \cite{chn,lingwang}, $\eta\approx 0.04$. A larger $\eta$ for deconfined quantum criticality 
198: was argued for on physical grounds by Senthil {\it et al.} \cite{sen04}. The universality class was argued
199: to be that of the hedgehog suppressed $O(3)$ transition, for which $\beta/\nu=(1+\eta)/2=0.80\pm 0.05$ 
200: was obtained in simulations of a classical model in \cite{mot04}. This is larger than $\beta/\nu=0.63 \pm 0.02$ 
201: found here, but on the other hand smaller lattices were used in \cite{mot04} and there may also be issues with 
202: how hedgehogs were suppressed. The direct study of an actual AF--VBS transition presented above can thus be 
203: expected to be more reliable.
204: 
205: 
206: \begin{figure}
207: \null~~\includegraphics[width=6.cm, clip]{fig4.eps}
208: \caption{Long-distance spin and dimer correlations scaled using $\nu=0.78$, $\eta=0.26$, 
209: and $g_c=0.04$.}
210: \label{fig4}
211: \vskip-5mm
212: \end{figure}
213: 
214: It is also interesting to study the probability distribution $P(D_x,D_y)$ of the dimer order 
215: parameter. In the VBS phase, one would expect this distribution to reflect the $Z_4$ symmetry, 
216: i.e., for a columnar VBS there should be peaks at $D_x=0,D_y=\pm D$ and $D_x=\pm D,D_y=0$
217: (whereas a plaquette state would give rise to peaks rotated by $45^\circ$). It should be noted, however, 
218: that $P(D_x,D_y)$ is a basis dependent function. In the valence bond simulations \cite{vbmc} the 
219: order parameters used to construct $P(D_x,D_y)$ are matrix elements
220: (with $\hat \mathbf{e} =  \hat \mathbf{x}, \hat \mathbf{y}$),
221: \begin{equation}
222: D_{e} = \frac{\langle \Psi_b |
223: \frac{1}{N}\sum_r \mathbf{S}(\mathbf{r}) \cdot \mathbf{S}(\mathbf{r}+\hat \mathbf{e})(-1)^{r_e}
224: |\Psi_a\rangle}{\langle \Psi_b |\Psi_a\rangle},
225: \end{equation}
226: where $|\Psi_a\rangle$, $|\Psi_b\rangle$ are valence bond states generated by operating with a 
227: high power $H^n$ on trial state (stochastically sampling valence bond evolutions). 
228: Although $P(D_x,D_y)$ is not a physically measurable quantity, any symmetry detected in
229: it should reflect an underlying symmetry of the projected state. Fig.~\ref{fig5} shows 
230: a color-coded $P(D_x,D_y)$ histogram generated at $J/Q=0$. The expected $Z_4$ symmetry 
231: of the VBS is not seen; instead the histogram is ring shaped, which indicates a $U(1)$ 
232: symmetric order parameter. Such an emergent $U(1)$ symmetry is in fact predicted \cite{sen04} by the 
233: deconfined theory in the VBS phase below a length scale $\Lambda$ which diverges faster than the VBS 
234: correlation length; $\Lambda \sim \xi_d^a$, with $a>1$. Thus, inside the VBS phase, if the system length 
235: $L \ll \Lambda$ one should expect to find an $U(1)$ symmetric order parameter, with the $Z_4$ becoming 
236: relevant only for larger sizes (and then seen as four peaks emerging in the histogram). Here, apparently, 
237: even at $J/Q=0$ the system is close enough to the critical point for the system length ($L=32$) to
238: be less than $\Lambda$ and, hence, $Z_4$ to be irrelevant. Recalling that the VBS gap is small, 
239: $\Delta/Q \approx 0.07$, and that $\Lambda \sim \xi_d^{a} \sim \Delta^{-a}$, this seems
240: reasonable. On moving closer to the critical point, $P(D_x,D_y)$ smoothly evolves into a single 
241: broad peak centered at $(0,0)$, as is expected for a continuous transition. Note that the finite-size 
242: extrapolation of the order parameter in Fig.~\ref{fig1} is not sensitive to the nature of the VBS 
243: state---plaquette or columnar---and should give the correct magnitude of the order parameter even 
244: though no $Z_4$ features are yet seen in the histogram for these system sizes. 
245: 
246: \begin{figure}
247: \null~~~\includegraphics[width=5.5cm, clip]{fig5.eps}
248: \caption{(Color online) Histogram of the dimer order parameter for an $L=32$ system at $J/Q=0$. 
249: The ring shape demonstrates an emergent $U(1)$ symmetry, i.e., irrelevance of the $Z_4$ 
250: anisotropy of the VBS order parameter.}
251: \label{fig5}
252: \vskip-5mm
253: \end{figure}
254: 
255: The above analysis points consistently to a deconfined quantum critical point as the most
256: likely scenario for the AF--VBS transition in the J-Q model. One set of exponents describes
257: both spin and dimer correlations, the value of $\eta$ is unusually large, and there is an emergent 
258: $U(1)$ symmetry in the VBS order parameter. In principle one cannot rule out a weakly 
259: first-order transition on the basis of finite-size data. However, although the lattice sizes used 
260: in this work are not extremely large, it would be hard to explain why a first-order transition should 
261: lead to the kind of scaling observed. A narrow region of AF/VBS coexistence is also unlikely, as there 
262: would then be two transitions and there is no reason to expect the spin and dimer critical exponents 
263: to be the same (in particular, the unusually large $\eta$). It is difficult to say anything more 
264: quantitative regarding a possible first-order transition or coexistence based on the calculations 
265: presented here.
266: 
267: An emergent $U(1)$ symmetry may also explain why it has been so difficult to determine the nature 
268: of the VBS state in the J$_{\rm 1}$-J$_{\rm 2}$ Heisenberg model \cite{j1j2new}. Even if the transition 
269: would be weakly first-order in this case \cite{sirker,kruger}, an emergent $U(1)$ symmetry could still 
270: affect small lattices, thus making it difficult to distinguish between columnar and plaquette VBS
271: patterns. Emergent $U(1)$ symmetry may be more common than deconfined quantum criticality and could 
272: hence affect many models with VBS states. The high density of low-lying singlets associated with 
273: $U(1)$ symmetry may also affect exact diagonalization studies \cite{mis05} of level spectra.
274: 
275: {\it Acknowledgments.---}
276: I would like to thank I. Affleck, L. Balents, K. Beach, M. P. A. Fisher, K. Harada, N. Kawashima, R. Melko, 
277: O. Motrunich, N. Prokof'ev, S. Sachdev, D. Scalapino, T. Senthil, B. Svistunov, and A. Vishwanath for 
278: stimulating discussions. This work was supported by the NSF under grant No.~DMR-0513930.
279: 
280: \null\vskip-8mm
281: \begin{thebibliography}{00}
282: 
283: \bibitem{chn} 
284: S. Chakravarty, B. I. Halperin, and D. R. Nelson, Phys. Rev. Lett. {\bf 60},
285: 1057 (1988); Phys. Rev. B {\bf 39}, 2344 (1989).
286: 
287: \bibitem{sachdevbook} 
288: S. Sachdev, {\it Quantum Phase Transitions} (Cambridge University Press,
289: Cambridge 1999).
290: 
291: \bibitem{and87} 
292: P. W. Anderson, Science \textbf{235}, 1196 (1987).
293: 
294: \bibitem{rea89} 
295: N. Read and S. Sachdev, Phys. Rev. Lett. \textbf{62}, 1694 (1989).
296: 
297: \bibitem{j1j2old} 
298: E. Dagotto and A. Moreo, Phys. Rev. Lett. \textbf{63}, 2148 (1989);
299: H. J. Schulz, T. Ziman, and D. Poilblanc, J. Phys. I \textbf{6}, 675 (1996).
300: 
301: \bibitem{mis05}
302: G. Misguich and C. Lhuillier, in {\it Frustrated Spin Systems}, 
303: edited by H. T. Diep (World-Scientific, 2005).
304: 
305: \bibitem{sac03} 
306: S. Sachdev, Rev. Mod. Phys. \textbf{75}, 913 (2003).
307: 
308: \bibitem{sondhi97} 
309: S. L. Sondhi, S. M. Girvin, J. P. Carini, and D. Shahar, 
310: Rev. Mod. Phys. \textbf{69}, 315 (1997).
311: 
312: \bibitem{herzandsuzuki}
313: J. A. Hertz, Phys. Rev. B \textbf{14}, 1165 (1976);
314: M. Suzuki, Prog. Theor. Phys. \textbf{56}, 1454 (1976).
315: 
316: \bibitem{sen04}
317: T. Senthil, A. Vishwanath, L. Balents, S. Sachdev, and M. P. A. Fisher,
318: Science \textbf{303}, 1490 (2004); T. Senthil, L. Balents, S. Sachdev, A. Vishwanath, 
319: and M. P. A. Fisher, Phys. Rev. B \textbf{70}, 144407 (2004).
320: 
321: \bibitem{assaad}
322: F. F. Assaad, M. Imada, and D. J. Scalapino, Phys. Rev. Lett. \textbf{77}, 4592 (1996).
323: 
324: \bibitem{san02}
325: A. W. Sandvik, S. Daul, R. R. P. Singh, and D. J. Scalapino,
326: Phys. Rev. Lett. \textbf{89}, 247201 (2002).
327: 
328: \bibitem{kuk0405}
329: A. Kuklov, N. Prokof'ev, and B. Svistunov, Phys. Rev. Lett. {\bf 93}, 230402 (2004); 
330: cond-mat/0501052.
331: 
332: \bibitem{kuk06}
333: A. Kuklov, N. Prokof'ev, B. Svistunov, and M. Troyer, Annals of Physics \textbf{321}, 1602 (2006).
334: 
335: \bibitem{kagome}
336: S. V. Isakov, S. Wessel, R. G. Melko, K. Sengupta, and Y. B. Kim, 
337: Phys. Rev. Lett. \textbf{97}, 147202 (2006); K. Damle and T. Senthil, 
338: {\it ibid.}, 067202 (2006).
339: 
340: \bibitem{mel06}
341: R. G. Melko, A. Del Maestro, and A. A. Burkov, Phys. Rev. B {\bf 74}, 214517 (2006).
342: 
343: \bibitem{sirker}
344: J. Sirker, Z. Weihong, O. P. Sushkov, and J. Oitmaa,
345: Phys. Rev. B \textbf{73}, 184420 (2006).
346: 
347: \bibitem{san06}
348: A. W. Sandvik and R. G. Melko, 
349: Annals of Physics {\bf 321}, 1651 (2006); cond-mat/0604451.
350: 
351: \bibitem{poilblanc}
352: D. Poilblanc, A. L\"auchli, M. Mambrini, and F. Mila, Phys. Rev. B \textbf{73},
353: 100403(R) (2006); M. Mambrini, A. L\"auchli, D. Poilblanc, and  F. Mila, {\it ibid.},
354: \textbf{74}, 144422 (2006).
355: 
356: \bibitem{j1j2new} 
357: M. E. Zhitomirsky and K. Ueda, Phys. Rev. B \textbf{54}, 9007 (1996);
358: O. P. Sushkov, J. Oitmaa, and Z. Weihong, {\it ibid.}, \textbf{63}, 104420 (2001);
359: L. Capriotti, F. Becca, A. Parola, and S. Sorella, {\it ibid.}, \textbf{67},
360: 212402 (2003).
361: 
362: \bibitem{vbmc} 
363: A. W. Sandvik, Phys. Rev. Lett \textbf{95}, 207203 (2005).
364: 
365: \bibitem{jkmethod} 
366: R. G. Melko and A. W. Sandvik, Phys. Rev. E {\bf 72}, 026702 (2005).
367: 
368: \bibitem{errornote} 
369: A claim in \cite{vbmc} of sign problems in the $S^z$ basis is wrong.
370: 
371: \bibitem{lingwang} 
372: L. Wang, K. S. D. Beach, and A. W. Sandvik, Phys. Rev. B \textbf{73}, 014431 (2006).
373: 
374: \bibitem{mot04} 
375: O. I. Motrunich and A. Vishwanath, Phys. Rev. B \textbf{70}, 075104 (2004).
376: 
377: \bibitem{jose} 
378: J. V. Jos\'e, L. P. Kadanoff, S. Kirkpatrick, and D. R. Nelson,
379: Phys. Rev. B \textbf{16}, 1217 (1977).
380: 
381: \bibitem{kruger} 
382: F. Kr\"uger and S. Scheidl, Europhys. Lett. \textbf{74}, 896 (2006).
383: 
384: \end{thebibliography}
385: 
386: \end{document}
387: 
388: 
389: