1: %\documentstyle[prb,eqsecnum,aps,epsf]{revtex}
2: %\documentstyle[aps,preprint,version2]{revtex}
3: %documentstyle[preprint,eqsecnum,aps]{revtex}
4: %\documentstyle[prl,twocolumn,aps,epsf]{revtex}
5: %\documentstyle[pra,aps,multicol,epsf]{revtex}
6: \documentstyle[prl,aps,multicol,epsfig]{revtex}
7: %\documentstyle[aps,prl,twocolumn]{revtex}
8: %\documentstyle[aps,preprint]{revtex}
9: %\documentstyle[11pt]{article}
10: %\documentstyle[prl,eqsecnum,aps,epsf,floats]{revtex}
11:
12: \begin{document}
13: %\draft
14:
15:
16: \title{Critical Point of a Weakly Interacting Two-Dimensional Bose Gas}
17:
18: \author{ Nikolay Prokof'ev$^{1}$, Oliver Ruebenacker$^{1}$, and
19: Boris Svistunov $^{2}$}
20:
21: \address{
22: $^1$ Department of Physics, University of Massachusetts,
23: Amherst, MA 01003, USA \\
24: $^2$ Russian Research Center ``Kurchatov Institute", 123182 Moscow,
25: Russia}
26:
27: \maketitle
28: \begin{abstract}
29: We study the Berezinskii--Kosterlitz--Thouless transition in a
30: weakly interacting 2D quantum Bose gas using the concept of
31: universality and numerical simulations of the classical
32: $|\psi|^4$-model on a lattice. The critical density and chemical
33: potential are given by relations $n_c=(mT/2\pi \hbar^2) \ln(\xi
34: \hbar^2/ mU)$ and $\mu_c=(mTU/\pi \hbar^2) \ln(\xi_{\mu} \hbar^2/
35: mU)$, where $T$ is the temperature, $m$ is the mass, and $U$ is
36: the effective interaction. The dimensionless constant $\xi= 380
37: \pm 3$ is very large and thus any quantitative analysis of the
38: experimental data crucially depends on its value. For $\xi_{\mu}$
39: our result is $\xi_{\mu} = 13.2 \pm 0.4 $. We also report the
40: study of the quasi-condensate correlations at the critical point.
41:
42: \end{abstract}
43:
44: \pacs{PACS numbers: 03.75.Fi, 05.30.Jp, 67.40.-w}
45: % 03.75.Fi Phase coherent atomic ensembles; quantum condensation phenomena
46: % 05.30.Jp Boson systems (for Bose-Einstein condensation, see 03.75.Fi)
47: % 67.40.-w Boson degeneracy and superfluidity of 4He
48:
49: \begin{multicols}{2}
50:
51: \narrowtext An accurate microscopic expression for the critical
52: temperature of the Berezinskii--Kosterlitz--Thouless (BKT)
53: transition \cite{BKT} has been a weak point of the theory of
54: weakly interacting two-dimensional Bose gas. The theory of Ref.~
55: \onlinecite{Popov} (see also \cite{KSS,Fisher} and analysis
56: below), suggests that the critical density of the BKT transition
57: in the weakly interacting system reads (we set $\hbar=1$)
58: \begin{equation}
59: n_c = {mT \over 2 \pi} \ln {\xi \over mU} \; .
60: \label{n_c_quantum}
61: \end{equation}
62: However, the value of $\xi$ cannot be obtained within standard
63: analytical treatments since $\xi$ is related to the system
64: behavior in the fluctuation region where the perturbative
65: expansion in powers of $U$ does not work. With unknown $\xi$, one
66: finds Eq.~(\ref{n_c_quantum}) rather inaccurate unless $mU$ is
67: exponentially small. Moreover, as we will find in this Letter, the
68: value of $\xi$ is very large: $\xi \approx 380$. This
69: means that for all experimentally available up to date (quasi-)2D
70: weakly interacting Bose gases \cite{Safonov,Ketterle}
71: the quantitative analysis of the data for the
72: critical ratio $n_c/T_c$ requires a precise value of $\xi$.
73: In the system of spin-polarized atomic hydrogen on helium film
74: \cite{Safonov}, the value of $mU$ is of order unity \cite{KSS};
75: in the recently created quasi-2D
76: system of sodium atoms \cite{Ketterle}, $mU$ is of order
77: $10^{-2}$, according to the formula of Ref.~\onlinecite{Petrov}.
78:
79: To quantitatively describe the BKT transition in a weakly
80: interacting Bose gas, it is sufficient to solve a classical-field
81: $|\psi|^4$-model with the effective long-wave Hamiltonian
82: \cite{Popov}
83: \begin{equation}
84: H[\psi]= \int \left\{ {1 \over 2m}|\nabla \psi|^2 + {U \over 2}
85: |\psi|^4 - \mu' |\psi|^2 \right\} \, d {\bf r} \; , \label{H}
86: \end{equation}
87: where $\mu' $ is the chemical potential, and $\psi$ is the classical
88: complex field.
89:
90: In this Letter, we first discuss the origin of the relation
91: (\ref{n_c_quantum}) in the limit of small $U$, and how quantum and
92: classical models relate to each other. Then we present our numeric
93: results (for the critical density, critical chemical potential,
94: and quasi-condensate correlations at the BKT point) obtained by
95: simulating the critical behavior of the 2D $|\psi|^4$-model on a
96: lattice using recently developed Worm algorithm \cite{Worm} for
97: classical statistical models. In particular, we show that the
98: quasi-condensate correlations are very strong at $T_c$, in
99: agreement with the experimental observation in the spin-polarized
100: atomic hydrogen \cite{Safonov} and quantum Monte Carlo simulations
101: \cite{KKKPS}.
102:
103: A simple dimensional analysis of the Hamiltonian (\ref{H}) allows
104: to write a generic formula for the critical point in a weakly
105: interacting 2D $|\psi|^4$-model. The routine itself is completely
106: analogous to that in the 3D case (see, e.g., \cite{Baym,KPS}), but
107: final results naturally reflect the specifics of the 2D case.
108:
109: We begin with introducing the mode-coupling momentum, $k_c$, that
110: characterizes the onset of strong non-linear coupling between the
111: long-wave harmonics of $\psi ({\bf r})$ (harmonics with $k \gg
112: k_c$ are almost free). This momentum is just the inverse of the
113: {\it healing} length, or vortex core radius, $r_c$\, \cite{BKT}.
114: We denote by $\tilde{n}$ the contribution to the total density due
115: to strongly coupled harmonics, and introduce the renormalized
116: chemical potential
117: \begin{equation}
118: \tilde{\mu} = \mu ' - 2 U \int_{k>k_c} n_k^{\rm (ideal)} d^2 k
119: /(2\pi)^2 \label{mu_k0}
120: \end{equation}
121: by subtracting the mean field contribution of non-interacting
122: high-momentum harmonics. Here
123: $n_k=\langle \, |\psi_{\bf k}|^2 \rangle$, and
124: $\langle \ldots \rangle$ stands for the statistical averaging.
125:
126: An estimate for $\tilde{n}$ follows from the Nelson-Kosterlitz
127: formula
128: \begin{equation}
129: n_s = {2 m T \over \pi} \; , \label{NK}
130: \end{equation}
131: since it is intuitively expected that $ \tilde{n} \sim n_s$. An
132: independent estimate of the parameters of the fluctuation region
133: is obtained by considering when all three terms in Eq.~(\ref{H})
134: are of the same order:
135: \begin{equation}
136: k_c^2/m \sim |\tilde{\mu}| \sim \tilde{n} U \; , \label{rel1}
137: \end{equation}
138: and relating $ \tilde{n} \sim \sum_{k<k_c} n_k \sim k_c^2 n_{k_c}$
139: to the renormalized chemical potential by using
140: $T/|\tilde{\mu}|$ in place of the occupation number $n_{k_c}$.
141: By definition, $k_c$ separates strongly coupled and free harmonics, and thus
142: $n_{k_c} \sim T/[ k_c^2/2m - \tilde{\mu}] \sim T/ |\tilde{\mu}|$.
143: The final order-of-magnitude estimates read (at $T=T_c$)
144: \begin{eqnarray}
145: \tilde{n} &\sim & m T \;, \label{n_tilde} \\
146: k_c &\sim & m (UT)^{1/2} \; , \label{k_c} \\
147: \tilde{\mu} &\sim & U m T \;. \label{mu_tilde}
148: \end{eqnarray}
149:
150: We are now in a position to derive Eq.~(\ref{n_c_quantum})
151: for the critical density. In 2D the main contribution
152: to the integral
153: \begin{equation}
154: n = \int n_k d^2 k/(2 \pi)^2 \; , \label{rel4}
155: \end{equation}
156: comes from large momenta between $k_c$ and
157: some ultra-violet scale $k_*$. The value and physical meaning of $k_*$
158: depend on the model. For classical lattice
159: models $k_*$ is given by the inverse lattice spacing;
160: in the continuous quantum system $k_* \sim \sqrt{mT}$
161: is the thermal momentum. At $k_c \ll k \ll k_*$
162: we have $n_k \approx 2mT/k^2$, and thus can write
163: \begin{equation}
164: n_c = {mT \over 2\pi} \ln (C k_*^2 /k_c^2 ) \; ,
165: \label{n_c}
166: \end{equation}
167: where $C$ is some constant. Critical density,
168: Eq.~(\ref{n_c_quantum}), for the quantum Bose gas is obtained by
169: substituting $C k_*^2/k_c^2 \equiv \xi mT/m^2UT = \xi/mU $.
170:
171: The dependence on the ultraviolet cutoff is associated with the
172: properties of {\it ideal} systems only, while the long-wave
173: behavior of all weakly-interacting $|\psi |^4$-theories is
174: universal. This fact allows one to relate results for different
175: models by adding and subtracting non-interacting contributions,
176: i.e., up to higher order corrections in $U$ the difference between
177: models ${\cal A}$ and ${\cal B}$ is given by $(n_c^{\cal (A)} -
178: n_c^{\cal (B)}) = \int [n_k^{\rm (ideal~{\cal A})} - n_k^{\rm
179: (ideal~{\cal B})}] \, d^2k/(2\pi)^2$. In what follows, the
180: reference system ${\cal A}$ will be the classical lattice model
181: with lattice spacing $a$, and our results are analyzed using
182: \begin{equation}
183: n_c^{\rm (lat)} = {m T \over 2 \pi} \ln {A \over m^2a^2UT} \; .
184: \label{n_c_lattice}
185: \end{equation}
186: The actual system of interest is the quantum Bose gas, so we add
187: and subtract the corresponding ideal-system contributions to get
188: \begin{equation}
189: \ln {A \over \xi ma^2T} = {1 \over 2\pi mT } \left(
190: \int_{BZ} { T ~d^2 k\over E({\bf k}) } -
191: \int {d^2 k \over e^{k^2/2mT} -1 } \right)
192: \; , \label{A_xi}
193: \end{equation}
194: where $BZ$ means that the first integral is over the Brillouin
195: zone, and $E({\bf k})$ is the dispersion law for the ideal lattice
196: model such that $ E({\bf k} \to 0) \to k^2/2m$. [The divergences
197: of the two integrals in Eq.~(\ref{A_xi}) at $k \to 0$ compensate
198: each other.]
199:
200: Our simulations were done for the simple square lattice
201: Hamiltonian
202: \begin{equation}
203: H = \sum_{{\bf k} \in BZ} [E({\bf k}) - \mu ] |\psi_k |^2 +
204: {U\over 2 } \sum_i |\psi_i|^4\; ,
205: \label{H_lat}
206: \end{equation}
207: where $\psi_k$ is the Fourier transform of the
208: complex lattice field $\psi_i$, and
209: \begin{equation}
210: E({\bf k}) = (1 / ma^2) [ 2- \cos (k_xa) - \cos (k_ya) ]
211: \label{E}
212: \end{equation}
213: is the tight-binding dispersion law.
214: With this dispersion relation the r.h.s.
215: in (\ref{A_xi}) can be evaluated analytically and we obtain
216: the ``conversion'' formula
217: \begin{equation}
218: \xi = A / 16 \; .
219: \label{xi_A}
220: \end{equation}
221: Since final results for dimensionless constants do not depend on
222: $m$, $T$, and $a$, in numerical simulations
223: we set $a=1$, $T=1$, and $m=1/2$ for convenience.
224:
225: The above consideration for the critical density can be readily
226: generalized to the critical chemical potential, with the result
227:
228: \begin{equation}
229: \mu_c ={m T U \over \pi} \ln {\xi_{\mu} \over mU } \; .
230: \label{mu_c}
231: \end{equation}
232: First, we notice that Eq.~(\ref{mu_c}) immediately follows form
233: Eqs.~(\ref{mu_tilde}) and Eq.~(\ref{mu_k0}) because the
234: mean-field term is proportional to $ - (mUT/\pi ) \ln (mU)$ (we
235: actually deal with exactly the same integral). Since the
236: renormalized value $\tilde{\mu }$ is universal, to account for the
237: difference between the classical and quantum models one has to add
238: and subtract mean-field contributions dominated by the ideal
239: behavior. Thus, if the classical model is analyzed using $\mu_c=(m
240: T U / \pi ) \ln [ A_{\mu} / m^2a^2 UT ] $, one has to apply
241: $\xi_{\mu} = A_{\mu} / 16$ to get the quantum result,
242: Eq.~(\ref{mu_c}).
243:
244: We now turn to our numerical procedure. To simulate the
245: grand-canonical Gibbs distribution corresponding to the
246: Hamiltonian (\ref{H_lat}), we employ the Worm algorithm (see
247: Ref.~\cite{Worm} for the description) that has demonstrated its
248: efficiency for the analogous problem in 3D \cite{KPS}. The formal
249: criterion of the critical point for the finite-size system is
250: based on the exact (Nelson--Kosterlitz) relation (\ref{NK}): We
251: say that the system of linear size $L$ is at the critical point,
252: if its superfluid density, $n_s(L)$, satisfies $n_s(L) = 2 m T /
253: \pi$. [The superfluid density has a direct estimator in the Worm
254: algorithm via the statistics of winding numbers \cite{Worm}, and
255: its autocorrelation time does not suffer from critical slowing
256: down.]
257:
258: The finite size scaling of $n_c(L)$ is well known from the
259: Kosterlitz-Thouless renormalization group theory \cite{BKT}
260: \begin{equation}
261: n_c(L) = n_c - \frac{A'\, m T}
262: {\ln^2 \big[ A''\, L m (UT)^{1/2} \big] } \; ,
263: \label{rel5}
264: \end{equation}
265: where $A'$ and $A''$ are dimensionless constants.
266: A similar relation applies also to the critical chemical
267: potential. Equation ~(\ref{rel5}) was used for
268: the finite-size scaling analysis. We found that
269: instead of extrapolating data for each value of $U$
270: to the $L \to \infty $ limit independently,
271: a much more efficient procedure is to perform a joint
272: finite-$L$ and finite-$U$ analysis.
273: To this end we heuristically introduce parameters accounting
274: for non-universal finite-$U$ corrections by adding linear in
275: $U$ terms to each of the three of the dimensionless constants:
276: $A \to A+BU$, $A' \to A'+B'U$, and $A'' \to A'' + B''U $.
277: We thus have six fitting parameters to describe
278: all our data points \cite{fit}. The data for $n_c(U,L)$ and
279: $\mu_c (U,L)$ are presented in Fig.~1. The fitting procedure
280: yields $A=(6.07 \pm 0.05)\cdot 10^{3}$, $A_{\mu}=(211 \pm 6)$,
281: which, according to Eq.~(\ref{xi_A}), means that
282: \begin{equation}
283: \xi=380 \pm 3 \; , ~~~~~ \xi_{\mu}=13.2 \pm 0.4 \; .
284: \label{xi}
285: \end{equation}
286: The fit is extremely good---20 points for the critical density at
287: $U\le 2.5$ and $L m (UT)^{1/2} >15 $, each calculated with
288: relative accuracy of order $10^{-4}$, are described with the
289: confidence level of $62$ \%.
290:
291:
292: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
293: \vspace*{-0.3cm}
294: \begin{figure}
295: \epsfxsize=0.48\textwidth
296: \hspace*{-0.5cm} \epsfbox{ncUL.ps}
297: \end{figure}
298: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
299: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
300: \vspace*{-5.5cm}
301: \begin{figure}
302: \epsfxsize=0.48\textwidth
303: \hspace*{-0.5cm} \epsfbox{mucULU.ps}
304: \vspace*{-3.8cm}
305: \caption{Critical density and chemical potential
306: for various coupling parameters
307: and system sizes.
308: Typical error bars are much smaller
309: than symbol sizes. The dotted line is the fitting function
310: described in the text.}
311: \label{fig:fig1}
312: \end{figure}
313: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
314:
315: Experiments on helium films often report that the ratio
316: $n_s(T_c)/n_s(0) = n_s(T_c)/n_c = 2mT_c/\pi n_c $ is close to
317: $0.75$ \cite{Reppy,Hallock}. Our simulation predicts that this
318: ratio is given by
319: \begin{equation}
320: n_s(T_c)/n_c = {4 \over 5.94 - \ln (mU) } \; ,
321: \label{ratio1}
322: \end{equation}
323: and $mU \approx 1.8 $ is required to describe helium films,
324: provided the small-$U$ approximation may be pushed that far
325: \cite{comment}. We are not aware of the published data on the
326: critical chemical potential. [For helium and hydrogen films on
327: substrates one has to shift $\mu_c$ by the value of the absorption
328: energy (for the delocalized atom, in the case of helium film),
329: $\mu_c \to \mu_c=E_0 + (m T U/\pi ) \ln ( \xi_{\mu}/ mU )$. In
330: thermal equilibrium this quantity can be readily measured through
331: the chemical potential of the bulk vapor.]
332:
333: In the absence of long-range order parameter, 2D systems below
334: $T_c$ are characterized by the local correlation properties of the
335: quasicondensate density, identical to those of a system with
336: genuine condensate \cite{KSS}. These properties reflect the
337: specific structure of the $\psi$-field:
338: \begin{eqnarray}
339: \psi ({\bf r}) & = & \psi_0 ({\bf r}) + \psi_1 ({\bf r}) \; ,
340: \label{psi} \\
341: \psi_0 ({\bf r})& \approx & \sqrt{n_0} \, e^{i \Phi ({\bf r})} \; ,
342: \label{psi_0}
343: \end{eqnarray}
344: where the quasicondensate density $n_0$ may be considered as a constant,
345: and $\psi_1$ is the Gaussian field independent of $\psi_0$. Both experiment
346: \cite{Safonov} and model Monte Carlo simulations \cite{KKKPS}
347: indicate that in 2D systems with $mU \sim 1$ the
348: quasicondensate correlations appear well above $T_c$
349: and are pronounced at $T_c$. Below we show that this is a generic feature
350: of weakly interacting $|\psi|^4$-models.
351:
352: It is convenient to characterize the quasi-condensate properties
353: by the correlator
354: \begin{equation}
355: Q = 2\langle \, |\psi|^2\rangle^2 - \langle \, |\psi|^4 \rangle
356: \; . \label{Q}
357: \end{equation}
358: The Gaussian component of the field obeys the Wick's theorem and
359: does not contribute to Eq.~(\ref{Q}). If, for a moment, by
360: $\psi_1$ we understand short-wave harmonics of $\psi$, we conclude
361: that only long-wave and strongly non-linear harmonics with the
362: momenta $k \sim k_c$ contribute to the correlator $Q$, i.e. $Q
363: \sim \tilde{n}^2$. Thus, we expect that all weakly interacting
364: $|\psi|^4$-models satisfy
365: \begin{equation}
366: Q = C_* m^2 T^2 ~~~~~~~(T=T_c) \label{C_*}
367: \end{equation}
368: in the limit of small $U$, where $C_*$ is a universal constant. By
369: definition, $n_0=\sqrt{Q}$.
370:
371: The finite-size and small-$U$ analysis of the data for $Q(U,L)$
372: was done in complete analogy with previously discussed cases of
373: $n_c(U,L)$ and $\mu_c(U,L)$ (see Ref.~\onlinecite{fit}). We found
374: that
375: \begin{equation}
376: C_* = 1.30 \pm 0.02
377: \label{C_*_result}
378: \end{equation}
379: The ratio between $n_0(T=T_c)$ and $n_c$ describes how pronounced
380: are the quasicondensate correlations in the Bose gas at the BKT
381: point:
382: \begin{equation}
383: {n_0^{(T=T_c)} \over n_c} = {2\pi \sqrt{C_*} \over \ln (\xi /
384: mU)} = {7.16 \over 5.94 + \ln (1/mU)} \; . \label{ratio}
385: \end{equation}
386: We see, that it is of order unity unless $mU$ is exponentially small.
387: Another interesting ratio is
388: \begin{equation}
389: {n_0 \over n_s} = {\pi \sqrt{C_*} \over 2} \approx 1.79 ~~~~~~~
390: (T=T_c) \;,
391: \label{ratio_2}
392: \end{equation}
393: which is interaction independent and shows that the superfluid density is
394: substantially smaller than the quasicondensate density at $T_c$.
395:
396:
397: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
398: \vspace*{-0.4cm}
399: \begin{figure}
400: \epsfxsize=0.48\textwidth \hspace*{-0.5cm} \epsfbox{q4025.ps}
401: \vspace*{-3.8cm} \caption{Quasicondensate correlations as a
402: function of system size. The dotted line is to guide the eye.}
403: \label{fig:fig3}
404: \end{figure}
405: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
406:
407:
408:
409: Finally, we would like to derive an accurate estimate for the
410: mode-coupling radius $r_c$. In an ideal system $Q \equiv 0$.
411: Hence, $Q(L)$ should decrease with decreasing $L$, and for system
412: sizes $L \sim r_c$ it has to drop significantly from its
413: thermodynamic value. We rather formally define $r_c$ from
414: $Q(L=r_c) \approx Q(L \to \infty) / 2$, and from Fig.~2 obtain
415: \begin{equation}
416: r_c \approx 2 / m (UT)^{1/2} \; .
417: \label{r_c}
418: \end{equation}
419:
420: We conclude by noting that Nelson-Kosterlitz formula (\ref{NK})
421: and Eqs.~(\ref{n_c_quantum}), (\ref{mu_c}), and (\ref{C_*})
422: constitute a complete set of equations which allow to fully
423: determine system parameters from measurements with independent
424: cross-checks. We are not aware of another study were dimensionless
425: constants $\xi$, $\xi_{\mu }$, and $C_*$ were determined with high
426: precision.
427:
428: We thank J. Machta and R. Hallock for valuable discussions.
429: This work was supported by the National Science Foundation under Grant
430: DMR-0071767. BVS acknowledges a support from Russian Foundation for
431: Basic Research under Grant 01-02-16508.
432:
433: \begin{thebibliography}{99}
434:
435: \bibitem{BKT} V.L.\ Berezinskii, Sov. Phys. JETP {\bf 32}, 493 (1971);
436: {\bf 34}, 610 (1972);
437: J.M.\ Kosterlitz and D.J.\ Thouless, J. Phys. C {\bf 6}, 1181 (1973);
438: J.M.\ Kosterlitz, J. Phys. C {\bf 7}, 1046 (1974).
439:
440: \bibitem {Popov} V.N.\ Popov, {\it Functional Integrals in Quantum Field
441: Theory and Statistical Phisics} (Reidel, Dordrecht, 1983).
442:
443: \bibitem{KSS} Yu.\ Kagan, B.V.\ Svistunov, and G.V.\ Shlyapnikov,
444: Sov. Phys. - JETP {\bf 66}, 314 (1987).
445:
446: \bibitem{Fisher} D.S.\ Fisher and P.C.\ Hohenberg,
447: Phys. Rev. B {\bf 37}, 4936 (1988).
448:
449: \bibitem{Safonov} A.I.\ Safonov, S.A.\ Vasilyev, I.V.\ Yasnikov,
450: I.I.\ Lukashevich, and S.\ Jaakkola, Phys. Rev. Lett. {\bf 81},
451: 4545 (1998).
452:
453: \bibitem{Ketterle} A.\ G\"{o}rlitz {\it et al}., cond-mat/0104549.
454:
455: \bibitem{Petrov} D.S.\ Petrov, M.\ Holzmann, and G.V.\
456: Shlyapnikov, Phys. Rev. Lett. {\bf 84}, 2551 (2000).
457:
458: \bibitem{Worm} N.V.\ Prokof'ev, and B.V.\ Svistunov,
459: cond-mat/0103146.
460:
461: \bibitem{KKKPS} Yu.\ Kagan, V.A.\ Kashurnikov, A.V.\ Krasavin, N.V.\ Prokof'ev,
462: and B.V.\ Svistunov, Phys. Rev. A {\bf 61}, 4360 (2000).
463:
464: \bibitem{Baym} G.~Baym, J.-P. Blaizot, M. Holzmann, F. Lalo\"{e},
465: and D. Vautherin, Phys.~Rev.~Lett. {\bf 83}, 1703 (1999).
466:
467: \bibitem{KPS} V.A.\ Kashurnikov, N.V.\ Prokof'ev, and B.V.\ Svistunov,
468: cond-mat/0103149.
469:
470: \bibitem{NK} D.R.\ Nelson and J.M.\ Kosterlitz,
471: Phys. Rev. Lett. {\bf 39}, 1201 (1977).
472:
473: \bibitem{fit} Fitting parameters are obtained from the
474: stochastic optimization procedure. The error bars are estimated
475: from fluctuations observed when some of the points are
476: added/removed from the optimization. We have also tried to look
477: for the non-universal finite-$U$ corrections of the form $U \ln
478: U$, but, within the error bars, obtained the same result for
479: $\xi$, and $\xi_{\mu}$. On another hand, $U \ln U$ corrections
480: were important in the fit for $Q(U,L)$.
481:
482: \bibitem{Reppy} G.\ Agnolet, D.F.\ McQueeney, and J.D.\ Reppy,
483: Phys. Rev. B {\bf 39}, 8934 (1989).
484:
485: \bibitem{Hallock} P.S. Ebey, P.T. Finley, and R.B. Hallock, J. Low
486: Temp. Phys. {\bf 110}, 635 (1998).
487:
488: \bibitem{comment} Actually, the interaction $mU \sim 2$ is too
489: large for a quantum Bose system to be accurately described as a
490: weakly interacting gas. From our estimate for the mode-coupling
491: radius, Eq.~(\ref{r_c}), we see that at such interactions $r_c$ is
492: already on the order of the interparticle distance. Also,
493: Eq.~(\ref{ratio}) makes no sense at all for $mU >1$.
494:
495: \end{thebibliography}
496: \end{multicols}
497:
498: \end{document}
499: