1: \documentstyle[aps,preprint,pra,epsfig]{revtex}
2: %\documentstyle[aps,multicol,pra,epsfig]{revtex}
3: %\documentclass{article}
4:
5: \begin{document}
6:
7: \draft
8:
9: \title{Finite-Temperature Renormalization Group Analysis of
10: Interaction Effects in $2D$ Lattices of Bose-Einstein Condensates}
11:
12: \author{A. Smerzi$^{1,2}$, P. Sodano$^{3}$, and A. Trombettoni$^{4}$}
13: \address{$^1$ Istituto Nazionale per la Fisica della Materia BEC-CRS and
14: Dipartimento di Fisica, Universita' di Trento, I-38050 Povo, Italy\\
15: $^2$ Theoretical Division, Los Alamos
16: National Laboratory, Los Alamos, NM 87545, USA\\
17: $^3$ Dipartimento di Fisica and Sezione I.N.F.N., Universit\`a di
18: Perugia, Via A. Pascoli, I-06123 Perugia, Italy\\
19: $^4$ Istituto Nazionale per la Fisica della Materia and
20: Dipartimento di Fisica, Universita' di Parma,
21: parco Area delle Scienze 7A, I-43100 Parma, Italy\\
22: }
23:
24: \date{\today}
25: \maketitle
26:
27: \begin{abstract}
28: By using a renormalization group analysis,
29: we study the effect of interparticle interactions on the
30: critical temperature $T_{BKT}$ at which the Berezinskii-Kosterlitz-Thouless
31: (BKT) transition occurs for Bose-Einstein condensates loaded at
32: finite temperature in a $2D$ optical lattice.
33: We find that $T_{BKT}$ decreases as the
34: interaction energy decreases; when $U/J=36/\pi$ one has $T_{BKT}=0$,
35: signaling the possibility of a quantum phase transition of BKT type.
36: \end{abstract}
37:
38: %\begin{multicols}{2}
39:
40: \section{Introduction}
41:
42: It has been recently suggested \cite{trombettoni03}
43: that a $2D$ optical lattice of Bose-Einstein
44: condensates \cite{greiner01} may allow for the observation
45: of a finite-temperature phase transition to a superfluid regime
46: where the phases of the single-well condensates are coherently aligned.
47: In fact, in an appropriate range of parameters,
48: the thermodynamical properties of the bosonic lattice at finite temperature
49: may be well described by the Hamiltonian of the $XY$ model
50: \cite{trombettoni03}, and, as it is well known, the $2D$ $XY$ model
51: exhibits the Berezinskii-Kosterlitz-Thouless (BKT)
52: transition \cite{BKT,minnaghen87,kadanoff00}.
53: The BKT phase transition occurs via an
54: unbinding of vortex defects: the low-temperature phase,
55: for $T$ below the BKT temperature $T_{BKT}$,
56: is characterized by the presence of bound vortex-antivortex pairs
57: and the spatial correlations exhibit a power-law
58: decay. For $T \sim T_{BKT}$ the pairs starts to unbind
59: (see e.g. \cite{tobochnik79,leoncini98}): in the high-temperature phase,
60: only free vortices are present, leading to an
61: effective randomization of the phases and to an exponential decay
62: of the correlation functions.
63:
64: The Hamiltonian describing the properties of bosons
65: in deep optical lattices
66: is the so called Bose-Hubbard Hamiltonian \cite{jaksch98},
67: in which two terms are present: a kinetic term
68: describing the hopping of the bosons with
69: tunneling rate $t$, and a potential term describing
70: the interaction between bosons in the wells of the array
71: with interaction energy $U$ (which is proportional to the
72: $s$-wave scattering length). The description of bosons in
73: deep lattices by means of the Bose-Hubbard Hamiltonian holds
74: also at finite temperature, provided that the temperature effects
75: do not induce the occupation of higher bands: for a $2D$ optical
76: lattice confined to a plane by a magnetic potential having frequency
77: $\omega_z$, this implies that the Bose-Hubbard Hamiltonian
78: description is
79: valid up temperatures $T$ such that $\hbar \omega_z \gtrsim k_B T$.
80: When the average number of bosons $N_0$ per site is high enough
81: and the interaction energy is larger than $t/N_0$, one may map -
82: {\em also at finite temperature} - the Bose-Hubbard Hamiltonian
83: in the quantum phase Hamiltonian \cite{trombettoni03}:
84: the conjugate variables are the phases and the particle numbers
85: of the condensates in the different sites of the lattice.
86: Thus, also at finite temperature, the phase diagram is
87: determined by the competition of two energies:
88: the Josephson energy $J \approx 2 t N_0$,
89: proportional to the tunneling rate between neighbouring sites of the $2D$
90: square optical lattice, and the interaction energy $U$. For $U=0$, the
91: quantum phase Hamiltonian reduces to the $XY$ Hamiltonian, exhibiting then
92: a BKT transition at $T_{BKT} \sim J/k_B$: for $T < T_{BKT}$ the system
93: as a whole behaves as a superfluid with phase coherence across the array.
94: Very accurate Monte Carlo simulations yield, in the thermodynamic limit,
95: $T_{BKT} = 0.898 J/k_B$ \cite{gupta88}.
96:
97: In this paper we determine the effect of the
98: interaction energy $U$ on the BKT transition.
99: When $U \ll J$, a BKT transition still occurs at a critical
100: temperature $T_{BKT}(U)$: we find that
101: $T_{BKT}(U)$ is smaller than $T_{BKT}$ and decreases with $U$
102: (see Fig. 1). Intuitively speaking, the quantum phase Hamiltonian,
103: as well as the Bose-Hubbard Hamiltonian, in the limit $U \gg J$
104: describes a Mott insulator, while in
105: the opposite limit, $J \gg U$, the array behaves as a
106: phase-coherent superfluid
107: (e.g., see \cite{sondhi97,fazio01}): thus, when $U$ increases,
108: the superfluid region in the phase diagram decreases.
109: Although in two dimensions there is not long-range order at finite temperature
110: due to the Mermin-Wagner theorem,
111: still the system exhibits superfluid behavior
112: for $T < T_{BKT}$ \cite{fazio01}: thus one expects
113: that $T_{BKT}(U)$ should decrease when $U$ increases.
114: We also find (see Section III) that, when $U/J$ is equal to the
115: critical value $(U/J)_{cr}=36/\pi$,
116: one has that $T_{BKT}(U) = 0$, signaling the possibility of a ($T=0$)
117: quantum phase transition.
118:
119: The quantum phase Hamiltonian describes also the behavior
120: of superconducting Josephson networks below the temperature
121: $T_{BCS}$ at which each junction becomes
122: superconducting \cite{fazio01,simanek94,martinoli00}.
123: The momenta conjugate to the phases of the macroscopic
124: wavefunctions of the superconducting grains are
125: the number of Cooper pairs, $J$ is the Josephson energy of the
126: superconducting Josephson junctions and $U$ is the charging energy
127: due to the Coulomb repulsion between Cooper pairs.
128: We observe that, in superconducting
129: Josephson arrays, the interaction term is written
130: in general as $\sum_{ij} U_{ij} N_i N_j$ where $N_i$ is the Cooper pairs
131: number at the site $i$: $U_{ij}$ is proportional to the inverse
132: of the capacitance matrix and may be also non-diagonal (i.e., $U_{ij}
133: \neq 0$ with $i \neq j$). At variance, for bosons in deep optical lattices,
134: the interaction term is $\sum_{i} U_{ii} N_i^2$, which corresponds,
135: in a suitable range of parameters, to a $diagonal$ quantum phase model.
136:
137: The analogy between superconducting Josephson networks and
138: atomic gas in deep optical lattices is then clear:
139: in each well of the periodic potential there is a condensate grain,
140: appearing at the Bose-Einstein condensation temperature
141: $T_{BEC}$. When $T_{BEC}$ is larger than all other energy scales,
142: the atoms in the well $i$ of the $2D$ optical lattice may be
143: described by a macroscopic wavefunction.
144: It becomes apparent, then, that an optical network
145: can be regarded as a network of {\em bosonic} Josephson junctions.
146: Furthermore, in both systems, when the interaction term is neglected
147: respect to the energy associated to the particle hopping, one has
148: that the whole $2D$ array becomes superfluid at the temperature
149: $T_{BKT}$ at which the BKT transition occurs, with $T_{BKT}$ smaller
150: than $T_{BCS}$ for superconducting Josephson
151: networks and smaller than $T_{BEC}$ for bosonic Josephson networks.
152:
153: The plan of the paper is the following: in Section II we introduce the
154: effective Hamiltonian describing bosons hopping on a deep optical lattice
155: and, by using a semiclassical approximation
156: \cite{jose84}, we compute the effective Josephson
157: energy in presence of the interaction energy $U$.
158: In Section III we evaluate the effect of the quantum fluctuations
159: on $T_{BKT}$ by putting in the renormalization group equations
160: the effective Josephson energy obtained in Section II;
161: a comparison with previous results is then carried out.
162: Details of the computations skipped in the text are in the
163: Appendix A. Section IV is devoted to some concluding remarks.
164:
165: \section{The renormalized Josephson energy}
166:
167: $2D$ optical lattices are created using two standing waves \cite{greiner01}:
168: when the polarization vectors of the two laser fields are orthogonal,
169: the periodic potential is
170: \begin{equation}
171: V_{opt}=V_0 [ \sin^2{(kx)} + \sin^2{(ky)} ],
172: \label{V_opt}
173: \end{equation}
174: where $k=2 \pi/\lambda$ is the wavevector of the lattice beams.
175: $V_0$ is usually expressed in units of $E_R=\hbar^2 k^2/2m$
176: (where $m$ is the atomic mass):
177: in \cite{greiner01} it is $\lambda=852 \, nm$ and
178: $E_R=h \cdot 3.14 kHz$.
179: Around the minima of the potential (\ref{V_opt}) one has
180: $V_{opt} \approx m\tilde{\omega}_r^2 (x^2 + y^2)/2$
181: with
182: \begin{equation}
183: \tilde{\omega}_r=\sqrt{\frac{2 V_0 k^2} {m}}.
184: \label{omega_tilde}
185: \end{equation}
186: When $\tilde{\omega}_r \gg \omega_z$ (with $\omega_z$
187: the frequency of the confining magnetic potential superimposed
188: to the optical potential and acting along $z$),
189: the system realizes a square array of tubes,
190: i.e. an array of harmonic traps elongated along the $z$-axis
191: \cite{greiner01}.
192:
193: When all the relevant energy scales are small compared to the excitation
194: energies, one can expand the field operator \cite{jaksch98} as
195: \begin{equation}
196: \hat{\Psi}(\vec{r},t)= \sum_i \hat{\psi}_i(t)
197: \Phi_i(\vec{r})
198: \label{expansion}
199: \end{equation}
200: with $\Phi_i(\vec{r})$ the Wannier wavefunction localized
201: in the $i$-th well
202: (normalized to $1$) and $\hat{N}_i=\hat{\psi}^{\dag}_i
203: \hat{\psi}_i$ the bosonic number operator.
204: Substituting the expansion of $\hat{\Psi}(\vec{r},t)$
205: in the full quantum Hamiltonian, one gets
206: the effective Hamiltonian describing the bosons hopping on
207: the deep optical lattice \cite{jaksch98,fisher89}
208: \begin{equation}
209: H=-t \sum_{<i, j>}
210: (\hat{\psi}^{\dag}_i \hat{\psi}_j+ h.c.)
211: + \frac{U}{2} \sum_i \hat{N}_i (\hat{N}_i-1).
212: \label{B-H}
213: \end{equation}
214: In Eq.(\ref{B-H}) $\sum_{<i, j>}$ denotes a sum over all the
215: distinct pairs of nearest neighbours;
216: $t$ and $U$ are respectively the tunneling rate and
217: the interaction energy, and are given by
218: \begin{equation}
219: t \simeq - \int d\vec{r} \, \Biggl[ \frac{\hbar^2}{2m}
220: \vec{\nabla} \Phi_i \cdot \vec{\nabla} \Phi_j +
221: \Phi_i V_{ext} \Phi_j \Biggr]
222: \label{t}
223: \end{equation}
224: and
225: \begin{equation}
226: U = \frac{4 \pi \hbar^2 a}{m} \int d\vec{r} \,
227: \Phi_i^4
228: \label{U}
229: \end{equation}
230: ($a$ is the $s$-wave scattering length).
231:
232: As discussed in the Introduction, one can show that,
233: when $V_0$ and $\omega_z$ are large enough,
234: the system is described by the Hamiltonian (\ref{B-H}) up to temperatures
235: $T \sim \hbar \omega_z / k_B$ \cite{trombettoni03}.
236: Upon defining $J \approx 2 t N_0$, the Hamiltonian (\ref{B-H}),
237: for $N_0 \gg 1 $ and $J/N_0^2 \ll U$ \cite{anglin01}, reduces to
238: \begin{equation}
239: \hat{H}=- J \sum_{<i,j>} \cos{(\theta_i-\theta_j)}
240: - {U \over 2} \sum_i \frac{\partial^2}{\partial \theta_i^2}.
241: \label{Q-P-M}
242: \end{equation}
243: For a $2D$ lattice with $V_0$ between $20E_R$ and $25E_R$ and
244: $N_0 \approx 170$ as in \cite{greiner01}, one sees
245: that the condition $\gg J/N_0^2$ is rather well
246: satisfied and that $J/k_B$ is of order of $20nK$.
247: For $U=0$, (\ref{Q-P-M}) is the $XY$ Hamiltonian:
248: \begin{equation}
249: \hat{H}=- J \sum_{<i,j>}
250: \cos{(\theta_i-\theta_j)}.
251: \label{X-Y}
252: \end{equation}
253:
254: The Hamiltonian (\ref{Q-P-M}) describes the so-called
255: quantum phase model \cite{fazio01,simanek94,martinoli00}.
256: There is an huge amount of literature on the properties of $2D$
257: superconducting Josephson arrays studied by means of the quantum phase model
258: \cite{fazio01,simanek94,martinoli00}. A (not exhaustive) list of
259: relevant papers includes mean-field and coarse-graining
260: approaches \cite{vanotterlo93,kopec00,grignani00},
261: Monte Carlo results
262: \cite{jacobs84,rojas96,capriotti03,alet03,alsaidi04},
263: renormalization group calculations \cite{jose84,rojas96}
264: and self-consistent harmonic approximations \cite{rojas96,kim97,cuccoli00}
265: (more references are in \cite{fazio01}). In the following we shall study
266: the renormalization-group equations in which it is used
267: an effective value of the Josephson energy computed within the
268: harmonic approximation \cite{jose84}.
269:
270: The starting point is the partition function $Z$ of the quantum phase model:
271: using the path integral formalism,
272: from (\ref{Q-P-M}), one has
273: \begin{equation}
274: Z=\int {\cal D} \theta e^{-\frac{1}{\hbar} S[\theta]},
275: \label{part_funct}
276: \end{equation}
277: where the action $S$ is given by
278: \begin{equation}
279: S[\theta]=
280: \int\limits_{0}^{\hbar \beta} {d\tau \Biggl[\frac{\hbar^2}{2U}
281: \sum\limits_j { \Biggl(\frac{\partial \theta_j}{\partial
282: \tau}\Biggr)^2}+ J \sum\limits_{<i, j>} {(1-
283: \cos\theta_{ij})} \Biggr]}
284: \label{azione}
285: \end{equation}
286: with $\beta=1/k_B T$ and
287: $\theta_{ij} \equiv \theta_{i}-\theta_{j}$.
288: Separating the phases as $\theta_i(\tau)=\varphi_i+
289: \delta_i(\tau)$, where $\varphi_i$ is a
290: static vortex configuration and $\delta_i(\tau)$ is a fluctuation
291: about $\varphi_i$, the path-integral partition function
292: (\ref{part_funct}) can be written as
293: \begin{equation}
294: Z=\int {\cal D} \varphi {\cal D} \delta e^{-\frac{1}{\hbar}
295: S[\varphi,\delta]}
296: \label{ZETA}
297: \end{equation}
298: where
299: \begin{equation}
300: S[\varphi,\delta]=\int\limits_{0}^{\hbar \beta}
301: {d\tau \Biggl[ \frac{\hbar^2}{2U} \sum\limits_{i}
302: {\dot{\delta}_i^2 (\tau)}+J \sum\limits_{<i,j>}
303: {(1- \cos\varphi_{ij} \, \cos\delta_{ij})} \Biggr]},
304: \label{azione_phi_delta}
305: \end{equation}
306: with $\varphi_{ij} \equiv \varphi_{i}-\varphi_{j}$ e
307: $\delta_{ij} \equiv \delta_{i}-\delta_{j}$.
308: Assuming that $\varphi_i$ and $\delta_i$ are slowly varying over the
309: size of the array \cite{jose84,simanek94}, i.e.
310: \begin{equation}
311: \cos\varphi_{ij} \approx 1-\varphi_{ij}^2/2
312: \label{APPR1}
313: \end{equation}
314: and
315: \begin{equation}
316: \cos\delta_{ij} \approx 1-\delta_{ij}^2/2,
317: \label{APPR2}
318: \end{equation}
319: one gets
320: \begin{equation}
321: Z=Z_0 \int{{\cal D}\varphi \exp\Bigl\{-{1 \over 2}
322: \beta \bar{J} \sum\limits_{<i,j>} {\varphi_{ij}^2}\Bigr\}},
323: \label{part_funct_phi_delta}
324: \end{equation}
325: where $Z_0 \equiv \int{{\cal D}\delta \, e^{-{1 \over \hbar} S_0[\delta]}}$
326: and
327: \begin{equation}
328: S_0[\delta]=\int\limits_{0}^{\hbar \beta}
329: {d\tau \Biggl[\frac{\hbar^2}{2U} \sum\limits_{i} {\dot{\delta}_i^2}+
330: {J \over 2} \sum\limits_{<i,j>} {\delta_{ij}^2}\Biggr]}.
331: \end{equation}
332: In Eq.(\ref{part_funct_phi_delta})
333: $\bar{J}$ is the renormalized Josephson energy, which is given by
334: \begin{equation}
335: \bar{J} \simeq J \Bigl(1-{1 \over 2} <\delta_{ij}^2>_0\Bigr)
336: \label{J_rin}
337: \end{equation}
338: with
339: \begin{equation}
340: <\delta_{ij}^2>_0 \equiv \frac{1}{Z_0}
341: \int{{\cal D}\delta \, e^{-{1 \over \hbar} S_0[\delta]}\delta_{ij}^2}.
342: \label{delta_ij}
343: \end{equation}
344: The evaluation of $<\delta_{ij}^2>_0$ can be carried out in a standard way
345: \cite{simanek94} and one has
346: \begin{equation}
347: <\delta_{ij}^2>_0=\sqrt{\frac{\pi U}{J}}
348: \frac{1}{\eta^3} \int\limits_{0}^{\eta} x^2 \coth{x} \, dx
349: \label{delta}
350: \end{equation}
351: where
352: \begin{equation}
353: \eta=\beta \sqrt{\pi U J}.
354: \label{eta}
355: \end{equation}
356:
357: \section{Renormalization group equations}
358:
359: In the renormalization group equations for the $2D$ $XY$ model
360: \cite{BKT,minnaghen87,kadanoff00} the scale-dependent screened charge
361: ${\cal K}$ depends on the dimensionless scaling parameter
362: $l=\log{(r/a)}$ (where $r$ is the vortex
363: distance and $a$ is the lattice spacing) and it is given by
364: ${\cal K}(l)=\beta J/\epsilon(l)$, where the
365: dielectric constant $\epsilon(l)$ expresses the screening of the
366: vortex-antivortex interaction due to the presence of other vortices
367: \cite{kadanoff00}. The renormalization group recursion equations
368: read \cite{kadanoff00}
369: \begin{equation}
370: \frac{d {\cal K}^{-1}(l)}{dl}=4 \pi^3 y^2(l)
371: \label{prima_rec}
372: \end{equation}
373: and
374: \begin{equation}
375: \frac{d y(l)}{dl}=[2- \pi {\cal K}(l)] y(l)
376: \label{seconda_rec}
377: \end{equation}
378: where $y \propto r^2 e^{-\beta V(r)/2} e^{-\beta \mu}$,
379: with $e^{-\beta \mu}$ is the fugacity for creating a vortex pair
380: and $V(r)$ correspond to the screened vortex-antivortex potential.
381: The study of the scaling equations (\ref{prima_rec})-(\ref{seconda_rec})
382: about the fixed point $y_f=0$, ${\cal K}_f=2/\pi$ shows that
383: the BKT transition occurs at $2-\pi {\cal K}(l=0) \approx 0$, where
384: ${\cal K}(l=0)=\beta J/\epsilon(l=0)=\beta J$ \cite{kadanoff00}.
385:
386: In presence of the quantum fluctuations, one has to replace the
387: scale-dependent charge ${\cal K}(l)$ by
388: $\bar{\cal K}(l)$ with $\bar{\cal K}(l=0)= \beta \bar{J}$ and
389: $\bar{J}$ given by Eqs.(\ref{J_rin}) and (\ref{delta}).
390: Denoting with $T_{BKT} (U)$
391: the BKT transition temperature for a given $U$, one finds
392: the following equation for $K \equiv J/ k_B T_{BKT}(U)$:
393: \begin{equation}
394: F(K)=2-\pi K \Biggl(1-\frac{1}{2 \pi^2 X_u K^3} \int_{0}^{\pi \sqrt{X_u} K}
395: x^2 \coth{x} \, dx \Biggr)=0
396: \label{radice}
397: \end{equation}
398: where $X_u=U/\pi J$.
399:
400: The root of Eq.(\ref{radice}) indicates the critical point
401: at which a BKT transition occurs. The eventual occurrence
402: of a double root for Eq.(\ref{radice}) might correspond to what is called
403: in literature a {\em reentrant} behavior. Indeed it has been
404: often argued (see Refs. in the review
405: \cite{fazio01}) that the quantum phase model
406: may undergo at low temperatures a reentrant transition
407: induced by the quantum fluctuations: namely, fixing $U/J$ and
408: lowering the temperature, one could switch from
409: an insulating phase to a superconducting one at $T_{BKT}(U)$ and then,
410: lowering further the temperature, one finds another critical temperature
411: $T^{(1)}(U)$ at which the system comes back ({\em reenters}) in the
412: insulating phase. Consistent with the reentrant scenario is the dramatic
413: decrease of the specific heat at very low temperatures
414: \cite{jacobs84} and, as discussed,
415: the presence of double roots for $T_{BKT}$ in the
416: renormalization group equations.
417:
418: The phenomenon of the reentrance - although not universal -
419: is a non-perturbative effect: in fact, opposite
420: results may be obtained by means of
421: different truncations for the series expansion of the function $F$.
422: For instance, if in Eq.(\ref{radice}) one expands the hyperbolic cotangent
423: as $\coth x \approx 1/x+x/3$, one gets an expression for $F(K)$
424: which gives two roots for $T_{BKT}$. If instead, for $\eta<\pi$,
425: one uses the expansion $\coth x=\frac{1}{x}+\sum\limits_{n=1}^{\infty}
426: \frac{2^{2n} B_{2n}}{(2n)!} x^{2n-1}$ where the $B_n$'s are the
427: Bernoulli numbers \cite{abramowitz72}, one obtains the equation:
428: \begin{equation}
429: F(K)=2-\pi K \Biggl(1-\frac{1}{4 K}-\sum\limits_{n=1}^{\infty}
430: \frac{6^n \, 2^{4n} \, B_{2n} \, x_u^n \, K^{2n-1}}
431: {2 \, (2n+2) \, (2n)!} \Biggr)=0
432: \label{bernoulli}
433: \end{equation}
434: with $x_u=\pi U/24 J$. Since $B_{4n}<0$ for $n=1,2, \ldots$, and
435: $B_{4n+2}>0$ for $n=0,1,2, \ldots$, if one truncates the sum
436: in Eq.(\ref{bernoulli}) to the $(2n+1)$-th order with
437: $n=0,1,2,\ldots$, one finds
438: $F(K) \to \infty$ for $K \to \infty$ and
439: then Eq.(\ref{bernoulli}) has two roots;
440: at variance, if one truncates the sum
441: to the $(2n)$-th order with
442: $n=1,2,\ldots$, one finds
443: $F(K) \to -\infty$ for $K \to \infty$ and
444: Eq.(\ref{bernoulli}) admits only one root. Thus one cannot truncate
445: the sum in Eq.(\ref{bernoulli}) to any finite order, even for $U/J \ll 1$;
446: the origin of this problem is that the expansion used
447: to get (\ref{bernoulli}) is a series with terms having alternating
448: signs. In addition, to find the roots
449: of the equation $F(K)=0$ one has to evaluate the integral
450: in (\ref{delta}) up to $\sqrt{UJ}/k_B T_{BKT}$, and, also
451: with $U$ small, $\eta$ may become large requiring to use all the orders
452: in Eq.(\ref{bernoulli}).
453:
454: A way to overcome the above mentioned
455: difficulties is to use in Eq.(\ref{radice})
456: the Mittag-Leffler expansion of
457: the hyperbolic cotangent, i.e.
458: $\pi \coth \pi x=1/x+2x\sum\limits_{n=1}^{\infty} (x^2+n^2)^{-1}$:
459: in this way Eq.(\ref{radice}) can be written as
460: \begin{equation}
461: F(K)=2-\pi K \Biggl[1-\frac{1}{4 K}-\frac{1}{X_u
462: K^3} \, g(K,X_u) \Biggr]=0
463: \label{radice_impr}
464: \end{equation}
465: where
466: \begin{equation}
467: g(K,X_u) \equiv \sum\limits_{n=1}^{\infty}
468: \Bigl[\frac{X_u K^2}{2}-\frac{n^2}{2}
469: \log \Bigl(1+ \frac{X_u K^2}{n^2}\Bigr)\Bigr].
470: \label{g}
471: \end{equation}
472: An analytic expression for $F(K)$ valid for $U/J \ll 1$ is given
473: by Eq.(\ref{approxx}) in the Appendix A.
474:
475: A detailed study of the function $F(K)$ is
476: in the Appendix A. Here we observe that, in the limit $U/J \to 0$,
477: one finds $2+\frac{\pi}{4}-\pi K=0$, from which
478: $K_0 \equiv K(U/J \to 0)=\frac{8+\pi}{4 \pi}$ and
479: $k_B T_{BKT} (0) \simeq 1.128 J$. Furthermore,
480: for $U/J<36/\pi$, one can show (see the Appendix A) that
481: $F(K) \to - \infty$ for $K \to \infty$ and that $F^\prime (K)<0$:
482: since $F$ tends to the positive value $2+\pi/4$ for $K \to 0$,
483: one can safely conclude
484: that, for $U/J<36/\pi$, Eq.(\ref{radice_impr}) has an {\it unique} solution.
485: At variance, one can show that, for $U/J>36/\pi$, Eq.(\ref{radice_impr})
486: does not admit any solution, while for
487: $U/J=36/\pi$ one has $K \to \infty$, i.e. $T_{BKT} (U) \to 0$.
488: A plot of $F(K)$ and $F^\prime(K)$ for three values of $U/J$
489: respectively smaller than, equal to and larger than $36/\pi$ is
490: given in Figs. 2 and 3.
491: One may infer that, at $T=0$, a phase transition is expected to occur
492: at the critical value $(U/J)_{cr}=36/\pi$: this
493: value is in reasonable agreement with the
494: mean-field estimates for the $T=0$ transition. In fact,
495: for the diagonal quantum phase model, one has that
496: the mean-field prediction
497: \cite{vanotterlo93,grignani00}
498: is $(U/J)_{cr} \approx 2z =8$ where $z=4$ is the number
499: of nearest neighbours.
500:
501: In Fig. 1 we plot as empty circles the values of $T_{BKT}(U)/T_{BKT}(0)$
502: as a function of $U/J$ from the numerical solution
503: of Eq.(\ref{radice_impr}).
504: If one uses the analytic expression (\ref{approxx})
505: for the function $F(K)$ one gets, for $U/J \lesssim 1/2$,
506: an error lesser than $1 \%$. A much better estimate may be obtained
507: by expanding the function
508: $F(K)$ near $K_0$: in this way Eq.(\ref{radice_impr}) reads
509: \begin{displaymath}
510: F(K) \approx F(K_0)+ (K-K_0) \cdot
511: F^{\prime}(K_0)=0,
512: \end{displaymath}
513: and then
514: \begin{equation}
515: K \approx K_0 - \frac{F(K_0)}
516: {F^{\prime}(K_0)}.
517: \label{sviluppo}
518: \end{equation}
519: In Fig. 1 Eq.(\ref{sviluppo}) is plotted as a solid line.
520:
521: We may conclude that Eq.(\ref{radice_impr}) admits, for small $U/J$,
522: a unique solution and this excludes the possibility of reentrant
523: behavior. Of course, our conclusion
524: relies on the approximations (\ref{APPR1})-(\ref{APPR2})
525: made in order to estimate $\bar{J}$ and $<\delta_{ij}^2>_0$.
526: A more careful treatment accounting for the
527: periodicity of the phases in Eq.(\ref{azione_phi_delta})
528: seems to be highly desirable to put on a more solid base
529: all the contrasting issues related to the reentrant
530: behavior of the systems described by the Hamiltonians (\ref{B-H}) and
531: (\ref{Q-P-M}). It is comforting to see that the experimental study
532: of weakly interacting bosons on $2D$ optical networks and
533: the investigations of their superfluidity properties at very low temperatures
534: could provide new insights also on this
535: intriguing and yet poorly understood problem
536: (see also the recent papers \cite{kleinert03}).
537:
538: \section{Concluding remarks}
539:
540: In this paper we studied, by means of a renormalization group analysis,
541: the effect of interparticle interactions on the
542: critical temperature at which the Berezinskii-Kosterlitz-Thouless
543: transition occurs for Bose-Einstein condensates loaded in a $2D$
544: optical lattice at finite temperature. We determined the shift
545: of the Berezinskii-Kosterlitz-Thouless transition temperature
546: induced by the interaction term and we compared our findings with
547: previously known results.
548:
549: \appendix
550:
551: \section{Properties of the renormalization group equation}
552:
553: In this Appendix we study the properties of the function
554: $F(K)$ defined in Eq.(\ref{radice_impr}).
555: We show in (1) that $F(0)>0$, in (2) that
556: $F(K) \to -\infty$ for $K \to \infty$ when $U/J<36/\pi$ and
557: in (3) that for $U/J<36/\pi$ one has $F^\prime(K)<0$:
558: one may then conclude that the equation $F(K)=0$ has only
559: one root for $U/J<36/\pi$. Finally, we derive an analytic expression
560: for $F(K)$ holding for small $U/J$.
561:
562: From Eq.(\ref{radice_impr}) one can see that:\\
563: (1) for $U/J \to 0$ one has $2+\frac{\pi}{4}-\pi K=0$,
564: from which $K=\frac{8+\pi}{4 \pi}$ and $K_B T_{BKT} \simeq 1.128 J$:
565: indeed, for small $U/J$, one has that
566: \begin{displaymath}
567: \frac{X_u K^2}{2}- \frac{n^2}{2} \log
568: \Bigl(1+\frac{X_u K^2}{n^2}\Bigr)=\frac{X_u K^2}{2}-
569: \frac{n^2}{2} \Bigl(\frac{X_u K^2}{n^2}-\frac{X_u^2 K^4}{2 n^4}+
570: \ldots\Bigr)=\frac{X_u^2 K^4}{4 n^2}-\ldots
571: \end{displaymath}
572: and thus
573: \begin{displaymath}
574: \frac{1}{X_u K^2} g(K,X_u) \approx \frac{1}{X_u K^2}
575: \sum\limits_{n=1}^{\infty} \frac{X_u^2 K^4}{4 n^2} \to 0.
576: \end{displaymath}
577: In the same way, $F(K) \to 2+\frac{\pi}{4}$ for $K \to 0$.\\
578:
579: (2) $F(K) \to -\infty$ per $K \to \infty$ for $x_u<3/2$
580: (where $x_u=\pi^2 X_u/24=\pi U/24J$): indeed for large $K$
581: \begin{displaymath}
582: F(K) \approx 2-\pi K \Bigl(1-\frac{1}{48 x_u K^3} \int\limits_{0}^{2 \sqrt{6}
583: \sqrt{x_u} K} dx x^2 \Bigr)=2-\pi K \Bigl(1-\frac{\sqrt{6}}{3}
584: \sqrt{x_u}\Bigr):
585: \end{displaymath}
586: then, for $K \to \infty$, $F(K) \to -\infty$ with $x_u<3/2$
587: and $F(K) \to \infty$ with $x_u>3/2$.\\
588:
589: (3) For $x_u<3/2$, it is $F^\prime (K)<0$ for $K>0$:
590: indeed
591: \begin{equation}
592: F^\prime(K)=-\pi-\frac{2 \pi}{X_u K^3} \, g(K,X_u)+\frac{\pi^2 X_u^{1/2}}
593: {2} \Bigl[\coth(\pi X_u^{1/2} K)- \frac{1}{\pi X_u^{1/2} K}\Bigr].
594: \label{21}
595: \end{equation}
596: Since $\lim_{x \to 0} \Bigl(\coth x-\frac{1}{x}\Bigr)=0$,
597: one gets
598: \begin{displaymath}
599: \lim_{K \to 0} F^\prime(K)=-\pi.
600: \end{displaymath}
601: For large $K$ one finds
602: \begin{displaymath}
603: F^\prime(K) \approx -\pi+\frac{\pi^2}{6} X_u^{1/2}.
604: \end{displaymath}
605: Therefore one has that $\lim_{K \to \infty} F^\prime(K)<0$ for
606: $X_u<36/\pi^2$, and $\lim_{K \to \infty} F^\prime(K)>0$ for
607: $X_u>36/\pi^2$.
608: Using similar arguments, one can show that, for $K>0$, it is
609: \begin{equation}
610: F^\prime(K)<-\pi+\frac{\pi^2}{6} X_u^{1/2}.
611: \label{21d}
612: \end{equation}
613: which implies that, for $X_u<36/\pi^2$, one has $F^\prime(K)<0$.
614: The behavior of $F(K)$ and $F^\prime(K)$ for three values of $U/J$
615: respectively smaller than, equal to and larger than $36/\pi$ is
616: plotted in Figs. 2 and 3.
617:
618: One may also obtain an analytic approximation for $F(K)$ holding
619: for $U/J \ll 1$ by putting
620: $z=\sqrt{X_u} K$ and using $\log (1+z)=\sum\limits_{j=1}^{\infty}
621: \frac{(-1)^{j+1} z^j}{j}$ for $\mid z \mid<1$: one has
622: \begin{equation}
623: g(K,X_u)=\sum\limits_{j=2}^{\infty} \frac{(-1)^j z^{2j}}{2j} \zeta(2j-2)
624: \label{appross}
625: \end{equation}
626: where $\zeta (j)=\sum\limits_{n=1}^{\infty} \frac{1}{n^j}$ is the
627: Riemann zeta-function. One finds
628: \begin{equation}
629: \sum\limits_{j=2}^{\infty} \frac{(-1)^j z^{2j}}{2j} \zeta(2j)=
630: \frac{1}{2} \Bigl\{\frac{\pi^2 z^2}{6}+\log \frac{\pi z}{\sinh \pi z}\Bigr\}.
631: \label{appross1}
632: \end{equation}
633: For large $n$ one has the recurrence relation \cite{abramowitz72}
634: \begin{displaymath}
635: \zeta(n+1) \simeq \frac{1}{2} [1+ \zeta(n)].
636: \end{displaymath}
637: Substituting back in Eq.(\ref{appross}) and using (\ref{appross1}),
638: one has
639: \begin{displaymath}
640: g(K,X_u) \approx X_u K^2 \Bigl(\frac{\pi^2}{3}-{3 \over 2}\Bigr)+
641: 2 \log \frac{\pi X_u^{1/2} K}{\sinh \pi X_u^{1/2} K}+
642: {3 \over 2} \log (1+X_u K^2)
643: \end{displaymath}
644: and finally
645: \begin{equation}
646: F \approx 2-\pi K
647: \Bigg[1-\frac{1}{4 K}-\frac{1}{X_u
648: K^3} \Bigg( X_u K^2 \bigg(\frac{\pi^2}{3}-{3 \over 2}\bigg)+
649: 2 \log \frac{\pi X_u^{1/2} K}{\sinh \pi X_u^{1/2} K}+
650: {3 \over 2} \log (1+X_u K^2) \Bigg) \Bigg].
651: \label{approxx}
652: \end{equation}
653:
654: \begin{thebibliography}{10}
655:
656: \bibitem{trombettoni03} A. Trombettoni, A. Smerzi, and P. Sodano,
657: New J. Phys. {\bf 7}, 57 (2005).
658:
659: \bibitem{greiner01}
660: M. Greiner {\em et al.}, Phys. Rev. Lett. {\bf 87}, 160405 (2001);
661: Appl. Phys. B {\bf 73}, 769 (2001).
662:
663: \bibitem{BKT} V. L. Berezinskii, Sov. Phys. JETP {\bf 32},
664: 493 (1971); J. M. Kosterlitz and D. J. Thouless,
665: J. Phys. C {\bf 6}, 1181 (1973); J. M. Kosterlitz,
666: {\em ibid.} {\bf 7}, 1046 (1974).
667:
668: \bibitem{minnaghen87} P. Minnaghen, Rev. Mod. Phys. {\bf 59},
669: 1001 (1987).
670:
671: \bibitem{kadanoff00} L. P. Kadanoff, {\em Statistical Physics: Statics,
672: Dynamics and Renormalization}, (Singapore, World Scientific, 2000),
673: Chapt.s 16-17 and reprints therein.
674:
675: \bibitem{tobochnik79} J. Tobochnik and G. V. Chester,
676: Phys. Rev. B {\bf 20}, 3761 (1979).
677:
678: \bibitem{leoncini98} X. Leoncini, A. D. Verga, and S. Ruffo,
679: Phys. Rev. E {\bf 57}, 6377 (1998).
680:
681: \bibitem{jaksch98} D. Jaksch, C. Bruder,
682: J. I. Cirac, C. W. Gardiner, and P. Zoller,
683: Phys. Rev. Lett. {\bf 81}, 3108 (1998).
684:
685: \bibitem{gupta88} R. Gupta, J. DeLapp,
686: G. G. Batrouni, G. C. Fox, C. F. Baillie, and J. Apostolakis,
687: Phys. Rev. Lett. {\bf 61}, 1996 (1988).
688:
689: \bibitem{sondhi97} S. L. Sondhi, S. M. Girvin,
690: J. P. Carini, and D. Shahar,
691: Rev. Mod. Phys. {\bf 69}, 315 (1997).
692:
693: \bibitem{fazio01} R. Fazio and H. van der Zant,
694: Phys. Rep. {\bf 355}, 235 (2001).
695:
696: \bibitem{simanek94} E. Sim\`anek,
697: {\em Inhomogeneous Superconductors},
698: Oxford University Press, New York, 1994.
699:
700: \bibitem{martinoli00} P. Martinoli and C. Leeman,
701: J. Low Temp. Phys. {\bf 118}, 699 (2000).
702:
703: \bibitem{jose84} J. V. Jos\'e,
704: Phys. Rev. B {\bf 29}, 2836 (1984).
705:
706: \bibitem{fisher89} M. P. A. Fisher,
707: P. B. Weichman, G. Grinstein, and D. S. Fisher,
708: Phys. Rev. B {\bf 40}, 546 (1989).
709:
710: \bibitem{anglin01} J. R. Anglin, P. Drummond, and A. Smerzi,
711: Phys. Rev. A {\bf 64}, 063605 (2001).
712:
713: \bibitem{vanotterlo93}
714: A. van Otterlo, K.-H. Wagenblast, R. Fazio and G. Sch\"{o}n,
715: Phys. Rev. B {\bf 48}, 3316 (1993).
716:
717: \bibitem{kopec00}
718: T. K. Kope\'c and J.V. Jos\'e,
719: Phys. Rev. Lett. {\bf 84}, 749 (2000).
720:
721: \bibitem{grignani00}
722: G. Grignani, A. Mattoni, P. Sodano, and A. Trombettoni,
723: Phys. Rev. B {\bf 61}, 11676 (2000).
724:
725: \bibitem{jacobs84} L. Jacobs, J. V. Jos\'e, and M. A. Novotny,
726: Phys. Rev. Lett. {\bf 53}, 2177 (1984).
727:
728: \bibitem{rojas96} C. Rojas and J. V. Jos\'e,
729: Phys. Rev. B {\bf 54}, 12361 (1996).
730:
731: \bibitem{capriotti03} L. Capriotti, A. Cuccoli, A. Fubini,
732: V. Tognetti, and R. Vaia, Phys. Rev. Lett. {\bf 91}, 247004 (2003).
733:
734: \bibitem{alet03} F. Alet and
735: E. S. S$\o$rensen, Phys. Rev. E {\bf 67}, 015701(R) (2003).
736:
737: \bibitem{alsaidi04} W. A. Al-Saidi and D. Stroud,
738: Physica C {\bf 402}, 216-222 (2004).
739:
740: \bibitem{kim97} B. J. Kim, J. Kim, S. Y. Park, and M. Y. Choi,
741: Phys. Rev. B {\bf 56}, 395 (1997).
742:
743: \bibitem{cuccoli00} A. Cuccoli, A. Fubini, V. Tognetti, and
744: R. Vaia,
745: Phys. Rev. B {\bf 61}, 11289 (2000).
746:
747: \bibitem{abramowitz72} M. Abramowitz and I. Stegun,
748: {\em Handbook of Mathematical Functions}, U.S. Government Printing Office,
749: Washington, DC, 1972.
750:
751: \bibitem{kleinert03} H. Kleinert, S. Schmidt, and A. Pelster,
752: Phys. Rev. Lett. {\bf 93}, 160402 (2004).
753:
754: \end{thebibliography}
755:
756: %\end{multicols}
757:
758: \begin{figure}
759: \centerline{\psfig{figure=fig1.ps,width=60mm,angle=270}}
760: \caption{The BKT critical temperature $T_{BKT}(U)$
761: (in units of $T_{BKT}(0)$) as a function
762: of $U/J$. Empty circles: numerical solution of Eq.(\ref{radice_impr});
763: solid line: Eq.(\ref{sviluppo}).}
764: \label{fig1}
765: \end{figure}
766:
767: \begin{figure}
768: \centerline{\psfig{figure=fig2.ps,width=60mm,angle=270}}
769: \caption{$F(K)$ for $U/J=1$ (solid line), $36/\pi$ (dashed line)
770: and $15$ (dotted line). The dot-dashed line is the analytic approximation
771: (\ref{approxx}) for $F$ holding for $U/J \ll 1$. For
772: $U/J=36/\pi$ $F$ asymptotically tends to $0$.}
773: \label{fig2}
774: \end{figure}
775:
776: \begin{figure}
777: \centerline{\psfig{figure=fig3.ps,width=60mm,angle=270}}
778: \caption{$F^\prime(K)$
779: for $U/J=1$ (solid line), $36/\pi$ (dashed line)
780: and $15$ (dotted line).}
781: \label{fig3}
782: \end{figure}
783:
784: \end{document}
785:
786:
787:
788:
789:
790:
791:
792:
793:
794:
795:
796:
797: