cond-mat0201158/1d.tex
1:  \documentstyle[prl,twocolumn,aps,epsfig]{revtex}
2: %\documentstyle[preprint,aps]{revtex}
3:  %\tighten
4: 
5: %prl format (two column):
6:  % \documentstyle[aps,prl,twocolumn]{revtex}
7: 
8:  %  paper format :
9:   %\documentstyle[aps]{revtex}
10: 
11:  \draft
12: 
13:  \begin{document}
14: 
15: 
16: \title{ Collective oscillations of a 1D trapped Bose gas}
17:  
18: \author{Chiara Menotti$^{1,2}$ and Sandro Stringari$^1$}
19: \address{$^1$ Dipartimento di Fisica, Universit\`{a} di Trento,}
20: \address{and Istituto Nazionale per la Fisica della Materia,}
21: \address{I-38050 Povo, Italy}
22: \address{$^2$ Universit\`{a} Cattolica del Sacro Cuore,}
23: \address{I-25121 Brescia, Italy}
24: 
25: 
26: 
27:   \date{\today}
28: 
29:   \maketitle
30: \begin{abstract}
31: \noindent
32: Starting from the hydrodynamic equations of superfluids,
33: we calculate the  frequencies of the collective oscillations of a harmonically 
34: trapped Bose gas for various 1D configurations. These include the
35: mean field regime described by  Gross-Pitaevskii theory and the beyond mean 
36: field regime at small  densities described by Lieb-Liniger theory.  
37: The relevant combinations of the physical parameters governing 
38: the transition between the different regimes are  discussed.
39: 
40: \end{abstract}
41: 
42: \pacs{PACS numbers: 03.75.Fi, 05.30.Jp, 32.80.Pj, 67.40.-w}
43:  
44:  \narrowtext
45: 
46: Recent experiments on trapped Bose gases at low temperature have pointed out
47: the occurrence of characteristic 1D features. These include deviations of the
48: aspect ratio and of the release energy \cite{mit01,salomon} from the 
49: 3D behaviour as well as the appearence of thermal fluctuations of the 
50: phase, peculiar of 1D configurations \cite{hannover}.
51: Interest in 1D interacting Bose gases arises from the occurrence of  
52: quantum features which are not encountered in 2D and 3D. For example  
53: in 1D the  fluctuations of the phase  of the order parameter
54: rule out the occurrence of long range order even at zero 
55: temperature \cite{PS}. 
56: Such systems cannot be in general described using traditional
57: mean field theories and require the development of a 
58: more advanced many-body approach. 
59: In the case of  1D Bose gases interacting with repulsive zero-range 
60: forces, this has  been implemented  by Lieb and Liniger \cite{lieb} who 
61: studied both the equation of state and the spectrum of elementary 
62: excitations of a uniform gas. In the presence of harmonic trapping,
63: 1D Bose gases exhibit new interesting features. The corresponding 
64: equilibrium properties have been already discussed  in a recent series 
65: of theoretical papers (see  \cite{gora,olshanii,gir} and references therein). 
66: In the present work we investigate the consequences of  harmonic trapping  
67: on the collective oscillations of an interacting 1D Bose gas at zero 
68: temperature.
69: We will consider various configurations, ranging from the mean field 
70: regime \cite{rmp} where the healing length is larger than the average 
71: interparticle distance, to  the Tonks-Girardeau limit \cite{tonks} where 
72: the gas acquires Fermi like properties.   
73: We will show that  the   frequency of the lowest compression mode provides
74:  a useful indicator of the different regimes.
75: 
76: We start our discussion from the hydrodynamic equations of superfluids in 1D
77: %
78: \begin{eqnarray}
79: \label{HD1}
80: &&\frac{\partial}{\partial t}\delta n_1 + 
81: \frac{\partial}{\partial z} \left(n_1 v \right) =0 \;, \\
82: &&\frac{\partial}{\partial t} v + 
83: \nabla_z \left(\mu_{\ell e}(n_1) +V_{ext}+\frac{1}{2}mv^2 \right) =0 \;,
84: \label{HD2}
85: \end{eqnarray}
86: %
87: which decribe  the dynamic behaviour of such systems at zero temperature.
88: In these equations $n_1(z,t)$ is the 1D  density of the gas, $v(z,t)$ is the 
89: velocity field, while $V_{ext}(z)$ is the external trapping potential 
90: which in the following will be assumed to be harmonic: 
91: $V_{ext}(z) = m\omega_z^2z^2/2$. 
92: The hydrodynamic equations of superfluids have been already
93: successfully employed to predict the collective frequencies
94: of 3D trapped Bose-Einstein condensates \cite{sandro96}.
95: A crucial ingredient of these equations is the local equilibrium
96: ($\ell e$) chemical potential $\mu_{\ell e}$, which should be 
97: evaluated for a uniform 1D gas 
98: ($V_{ext}=0$) at the density $n_1$. 
99: The applicability of the above equations 
100: requires the validity of the local density approximation along the 
101: $z$-th direction. This is expected to be accurate for 
102: sufficiently large systems.
103: Furthermore, Eqs.(\ref{HD1},\ref{HD2}) 
104: should be limited to the study of macroscopic 
105: phenomena where variations in space take place over distances
106: larger than the average distance between particles.
107: From Eq.(\ref{HD2}) one can easily calculate 
108: the ground state profile through the equation
109: %
110: \begin{equation}
111: \mu_{\ell e}(n_1(z)) + V_{ext}(z) = \mu \; .
112: \label{n0}
113: \end{equation}
114: %
115: The collective oscillations are instead 
116: determined by writing 
117: the density in the form $n_1(z,t) = n_1(z) + e^{-i\omega t}\delta n_1(z)$, 
118: with the function $\delta n_1(z)$ obeying the
119: linearized equation 
120: %
121: \begin{equation}
122: \omega^2 \delta n_1(z) = 
123: \frac{1}{m}\nabla_z \left[ n_1(z) \nabla_z \left( 
124: \frac{\partial \mu_{\ell e}}{\partial n_1} \delta n_1(z) \right)
125: \right] \;,
126: \label{deltan}
127: \end{equation}
128: %
129: which immediately follows from Eqs.(\ref{HD1},\ref{HD2}).
130: %
131:  In the case of  uniform systems ($V_{ext}=0$) one has 
132: plane wave solutions $\delta n_1 = \exp(iqz)$ with $\omega^2 = c^2q^2$,
133: where $q$ is the wave vector of the excitation and 
134: $c^2 =n_1 (\partial \mu_{\ell e}  /\partial n_1) /m$ 
135: is the square of the sound velocity. It is worth reminding that 
136: the applicability of the hydrodynamic equations is not limited 
137: to the mean field scenario. Actually in \cite{lieb} it has been proven
138: that in 1D  Bose gases the velocity of sound,
139: derived from the macroscopic compressibility, coincides with the one derived 
140: from the microscopic calculation of the phonon excitation spectrum
141: also in regimes far from mean field.
142: 
143: In the first part of the work we evaluate $\mu_{\ell e}(n_1)$ and the 
144: corresponding solutions for the collective oscillations in the framework 
145: of the mean field Gross-Pitaevskii theory.
146: Even  in this regime one can 
147: explore a  rich variety of situations ranging from the Thomas-Fermi regime 
148: in the radial direction to the one of tight confinement where the motion 
149: in the radial direction is frozen.
150: In the second part we extend the analysis to 
151: regimes beyond mean field, including the limit of the Tonks-Girardeau gas.
152: 
153: Let us  consider  a uniform system of length $L$ in the $z$-th direction 
154: and confined by a harmonic potential 
155: $V(r_{\perp})=m\omega_{\perp}^2 r_{\perp}^2/2$ in the radial direction. 
156: By writing the order parameter in the form 
157: $\Psi=\sqrt{n_1}f(\rho_{\perp})/a_{\perp}$, where $n_1 =N/L$ is the 1D 
158: density, $a_{\perp}=\sqrt{\hbar / m\omega_{\perp}}$ is the 
159: oscillator length in the radial direction and 
160: $\rho_{\perp}=r_{\perp}/a_{\perp}$ is the dimensionless radial 
161: coordinate, the 3D Gross-Pitaevskii equation yields the dimensionless 
162: equation 
163: %
164: \begin{equation}
165: \left(-\frac{1}{2} \frac{\partial^2}{\partial \rho_{\perp}^2}
166: -\frac{1}{2\rho_{\perp}}  \frac{\partial}{\partial \rho_{\perp}}
167: + \frac{1}{2} \rho_{\perp}^2 + 4 \pi a n_{1} f^2 \right) f = 
168: \frac{ \mu_{\ell e}}{\hbar \omega_{\perp}} f
169: \label{f}
170: \end{equation}
171: %
172: for the function $f$ obeying the normalization 
173: condition $2\pi \int |f(\rho_{\perp})|^2 \rho_{\perp} d \rho_{\perp} =1$. 
174: In Eq.(\ref{f}), 
175: $\mu_{\ell e} /\hbar \omega_{\perp}$ is the chemical 
176: potential in units of the radial quantum oscillator energy.
177: Eq. (\ref{f}) shows that the relevant dimensionless parameter of 
178: the problem  is $an_1$. 
179: It is worth considering two important limits. 
180: If $a n_1 \gg 1$, one  enters the  radial Thomas-Fermi regime 
181: (hereafter called 3D cigar), where many configurations
182: of the harmonic oscillator Hamiltonian are excited in the radial 
183: direction and the equation of state takes the analytic form 
184: %
185: \begin{equation}
186: {\mu_{\ell e} \over \hbar \omega_{\perp}} = 2 (a n_1)^{1/2}.
187: \label{mu3D}
188: \end{equation}
189: %
190: Notice that in this limit  the chemical potential is not linear 
191: in the density. This implies, in particular, that the sound velocity 
192: is related to the chemical potential by the law 
193: $c^2 =\mu_{\ell e}/2m$ \cite{zaremba} rather then by the
194: Bogoliubov relation $c^2 =\mu_{\ell e}/m$.
195: A second important case is the perturbative regime where $a n_1 \ll 1$ 
196: (hereafter called 1D mean field).
197: In this case the solution of (\ref{f}) approaches the Gaussian ground 
198: state of the radial harmonic oscillator  and one finds  the linear 
199: law
200: %
201: \begin{equation}
202: {\mu_{\ell e} \over \hbar \omega_{\perp}} = 1 + 2 a n_1
203: \label{mu1D}
204: \end{equation}
205: for the chemical potential.
206: 
207: If the gas is harmonically trapped also along the $z$-th direction, 
208: the density $n_1$ exhibits a $z$-th dependence which is worth calculating
209: as a function of the 
210: relevant parameters of the problem: the scattering length $a$, the 
211: number $N$ of atoms and the trapping frequencies 
212: $\omega_{\perp}$ and $\omega_z$. 
213: To this purpose one has to solve Eq.(\ref{n0}) by imposing the normalization
214: condition $\int n_1(z) dz=N$. 
215: A useful quantity is the Thomas-Fermi radius $Z$ defined by the value of $z$
216: at which the equilibrium  density $n_1(z)$ vanishes.
217: According to Eq.(\ref{n0}), one has 
218: $\mu-\mu_{\ell e}(an_1=0) = (1/2)m\omega_z^2Z^2$.
219: In terms of $Z$,  Eq.(\ref{n0}) can be rewritten as 
220: ${\tilde \mu}_{\ell e}(an_1(z)) = 
221: (m\omega_z^2 Z^2/2\hbar\omega_{\perp})(1-z^2/Z^2)$, 
222: where we have defined the dimensionless quantity
223: $ {\tilde \mu}_{\ell e}(an_1(z))=
224: \left[\mu_{\ell e}(an_1(z))- \mu_{\ell e}(an_1=0)\right]/\hbar \omega_{\perp}$.
225: This function is fixed by the solution of the Gross-Pitaevskii equation
226: (\ref{f}).
227: Its  inverse ${\tilde \mu}_{\ell e}^{-1}$ gives
228: the value of $a n_1$ as a a function of $z$, so that the normalization 
229: condition obeyed by the density can be written as
230: %
231: %
232: \begin{equation}
233: \frac{Za_{\perp}}{a_z^2} 
234: \int_{-1}^{1} {\tilde \mu}_{\ell e}^{-1} 
235: \left[ \frac{1}{2} 
236: \left( \frac{Za_{\perp}}{a_z^2} \right)^{2} (1-t^2) \right] dt=
237: \frac{N a a_{\perp}}{a_z^2} ,
238: \label{s5}
239: \end{equation}
240: %
241: with $t=z/Z$. 
242: Eq. (\ref{s5}) explicitly points out  the relevance of 
243: the dimensionless combination $Naa_{\perp}/a^2_z$ where 
244: $a_z =\sqrt{\hbar / m\omega_z}$ is the oscillator length in the
245: axial directions.
246: From Eq.(\ref{s5}) one can calculate, for a given choice of 
247: the parameters, the radius $Z$ and hence the $1D$ density profile. 
248: In the 3D cigar limit $Naa_{\perp}/a^2_z\gg 1$, one has
249: $Z =  (a^2_z/a_{\perp})(15 Naa_{\perp}/a^2_z)^{1/5}$ and
250: 
251: \begin{eqnarray}
252: n_1(z) = \frac{1}{16 a}
253: \left( \frac{15 Naa_{\perp}}{a^2_z}\right)^{4/5}
254: \left(1-\frac{z^2}{Z^2}\right)^2.
255: \end{eqnarray}
256: %
257: %
258: In the 1D mean field limit $Naa_{\perp}/a^2_z\ll 1$,
259: one instead finds \cite{olshanii}
260: $Z= (a^2_z/a_{\perp})(3Naa_{\perp}/a^2_z)^{1/3}$ and
261: 
262: \begin{eqnarray}
263: n_1(z) = \frac{1}{4a}
264: \left(\frac{3Naa_{\perp}}{a^2_z}\right)^{2/3} 
265: \left(1-\frac{z^2}{Z^2}\right).
266: \label{n1MF}
267: \end{eqnarray}
268: %
269: The density profiles are 
270: different in the two regimes, reflecting the different behaviour of 
271: the equation of state.
272: %
273: The conditions of applicability of the local density approximation 
274: employed above are determined by requiring that $Z \gg a_z$. In the 
275: 1D mean field regime  this implies the non trivial condition
276: $(a_z/a_{\perp}) (N a a_{\perp}/a_z^2)^{1/3} \gg 1$.
277: 
278: Let us now discuss the behaviour of the collective oscillations. 
279: The hydrodynamic equation (\ref{deltan}) has simple analytic solutions if 
280: the density derivative of the chemical potential is a power law function:
281: $\partial \mu_{\ell e} /\partial n_1 \propto n_1^{\gamma-1}$. 
282: The Thomas-Fermi  and the 1D mean field regimes belong to this class 
283: of solutions with $\gamma =1/2$ and $\gamma =1$ respectively. 
284: Also the Tonks-Girardeau limit (see Eq.(\ref{muT}) below) belongs to 
285: the same class with 
286: $\gamma = 2$. Hence in these three relevant limits the dipersion relation 
287: for the collective frequencies can be obtained analytically. By looking 
288: for solutions of the form 
289: $n_1^{\gamma -1}\delta n_1(z) = z^k + a z^{k-2} \dots$, where 
290: $k \le 1$ and only positive powers of $z$ are included in the
291: polynomial, one finds the result
292: %
293: \begin{equation}
294: \omega^2 = \omega^2_z {k \over 2} [2 +\gamma (k-1)].
295: \label{omegagamma}
296: \end{equation}
297: %
298: The case $k=1$ corresponds to the center of mass motion whose frequency 
299: is given by $\omega=\omega_z$ independent of the value of $\gamma$. 
300: The most interesting $k=2$ case (lowest compressional mode) is instead 
301: sensitive to the regime considered. 
302: One finds $\omega^2 = (5/2)\omega_z^2$, $\omega^2 = 3\omega_z^2$ and 
303: $\omega^2 = 4\omega_z^2$ for the 3D cigar, 1D mean field and Tonks-Girardeau 
304: regimes respectively. The result $\omega^2 = (5/2)\omega_z^2$
305: was first derived in \cite{sandro96} by solving the hydrodynamic 
306: equations for a trapped 3D system in the limit of a highly elongated
307: trap ($\omega_z \ll \omega_{\perp}$). 
308: This prediction has been confirmed experimentally with high precision 
309: \cite{mit96}. 
310: The result $\omega^2 = 3 \omega^2_z$ was derived in 
311: \cite{csordas,sandro98,ho}, 
312: while the result  $\omega^2 = 4\omega^2_z$ for the Tonks-Girardeau gas 
313: is simply understood by recalling that, in this limit, there is an exact 
314: mapping with the 1D ideal Fermi gas \cite{tonks} where the excitation 
315: spectrum, in the presence of harmonic confinement, is  $\omega = k \omega_z$.  
316: The same result has been recently derived in
317: \cite{Minguzzi} using the 
318: mean field equations of Kolomeisky et al. \cite{kol}.
319: 
320: 
321: \begin{center}
322: \begin{figure}
323: \includegraphics[width=0.95\linewidth]{figure1.eps}
324: \caption{Transition between the 1D mean field and the 3D cigar regimes: 
325: $\omega^2/\omega_z^2$ as a function of the parameter 
326: $Naa_{\perp}/a_z^2$ (full line). Transition to the Tonks-Girardeau regime for
327: $a_{\perp}/a=100$ (dashed line)
328:  and $10$ (dotted line).}
329: \label{fig1}
330: \end{figure}
331: \end{center}
332: %
333: 
334: In order to evaluate the collective frequencies in the intermediate 
335: regimes where 
336: the hydrodynamic equations are not  analytically  soluble, 
337: we have  developed a sum rule approach \cite{sandro96}, based on the 
338: evaluation of the ratio $\hbar^2 \omega^2 = m_1/m_{-1}$ between the 
339: energy weighted and inverse energy weighted sum rules. In the following 
340: we will limit the discussion to the lowest compression mode which is 
341: naturally excited by the operator $\sum_{i=1}^N z_i^2$. 
342: The energy weighted moment is given by 
343: $m_1= (1/2)\langle [\sum_{i=1}^N z_i^2,[H,\sum_{i=1}^N z_i^2]]\rangle=
344:  (2N\hbar^2/m)\langle z^2\rangle$ 
345: where $\langle z^2\rangle = \int n_1(z)z^2dz/N$ is the average square 
346: radius fixed by the ground state solution $n_1(z)$. 
347: The inverse energy weighted moment $m_{-1}$ is related to the static 
348: polarizability $\alpha$ by $m_{-1}= (1/2) \alpha$.
349: This  is evaluated by adding the pertubation $-\epsilon z^2$ 
350: to the Hamiltonian and calculating the corresponding changes 
351: $\delta \langle z^2\rangle $ of the expectation value of the
352: square radius:
353: $\alpha = N\delta\langle z^2\rangle /\epsilon$. 
354: Adding the perturbation  $-\epsilon z^2$ is equivalent 
355: to changing the frequency $\omega_z$ of the harmonic confinement, so 
356: that the result for the collective frequency takes the  compact
357: form \cite{note}
358: %
359: \begin{equation}
360: \omega^2 = -2 {\langle z^2\rangle \over 
361: d\langle z^2\rangle /d\omega^2_z}.
362: \label{omega1-1}
363: \end{equation}
364: %
365: We have calculated the ratio $\omega^2/\omega_z^2$ as a function of the 
366: dimensionless parameter $Naa_{\perp}/a^2_z$, thereby exploring the
367: transition between the 3D cigar and the 
368: 1D mean field  regimes.
369: The results are reported in Fig.~\ref{fig1}. For the experimental
370: conditions of \cite{mit01,salomon} where 
371: $Naa_{\perp}/a^2_z=0.24$ and $0.08$, we 
372: predict $\omega^2/\omega^2_z=2.85$ and $2.91$ respectively, 
373: confirming that those experiments are actually touching 
374: the transition between the two mean field regimes. 
375: 
376: It is worth noticing that result (\ref{omega1-1}) for the collective 
377: frequency, being based on general sum rule arguments, applies also
378: to  regimes beyond mean field where the density profile 
379: $n_1(z)$ and hence the value of $\langle z^2\rangle$ cannot be evaluated 
380: starting from the Gross-Pitaevskii equation of state.
381: Deviations from the mean field regime become important when the 
382: healing length $\xi =(8\pi an)^{-1/2}$ is comparable to the average 
383: distance $d$ between particles. 
384: In  the presence of tight radial confinement one can use the 
385: relationship  $n=n_1/\pi a_{\perp}^2$ between the 3D density evaluated 
386: at $r_{\perp}=0$ and the 1D density  
387: $n_1=\int n(r_{\perp}) d{\vec r}_{\perp}$. 
388: When $a$ becomes smaller than $d$, one can write  $d=1/n_1$.
389: One then obtains the result
390: %
391: $\xi / d =  \sqrt{a^2_{\perp}n_1/8a }$
392: %
393: which becomes smaller and smaller 
394: as the 1D density decreases, thereby suggesting the occurrence
395: of important deviations from
396: the mean field behaviour for very dilute 1D samples. This
397: should be contrasted with the   3D case 
398: where the mean field condition  ($\xi > d$) is better
399: and better satisfied as the density decreases. 
400: The combination $a_{\perp}^2n_1/a$ can then be used as an indicator 
401: of the applicability of the mean field approach. When its value 
402: becomes  of the order of $1$ or smaller, one enters a new regime 
403: characterized by important quantum correlations.
404: The corresponding  many-body problem was investigated by
405: Lieb and Liniger \cite{lieb} who considered 1D repulsive zero-range 
406: potentials of the form $g_{1D}\delta(z)$. 
407: The interaction parameter $g_{1D}$  can be also written as 
408: $g_{1D}=\hbar^2/(ma_{1D}) $  where
409:  $a_{1D}$   is the one-dimensional scattering length \cite{olshanii2}. 
410:  By averaging the 3D interaction $4\pi^2\hbar^2a \delta({\bf r})/m$ 
411: over the radial density profile, one obtains the simple  identification 
412: $a_{1D}=a^2_{\perp}/a$ \cite{noteolshanii}. 
413: The comparison with the expression for $\xi/d$ shows that the 
414: deviations from the mean field increase
415: by decreasing  $a_{1D}n_1$.
416: 
417: In the Lieb-Liniger scenario the 
418: energy per particle $\epsilon(n_1)$,  
419: when  expressed in  units of the energy $\hbar^2 /2ma^2_{1D}$,
420: turns out to be a  universal function of the dimensionless parameter 
421: $ a_{1D}n_1$. 
422: Important limits  are the high density limit $a_{1D}n_1\gg 1$, where  
423: one finds $\epsilon(n_1) = \hbar^2 n_1/ma_{1D}$ and hence 
424: the mean field result (\ref{mu1D}) for the chemical potential 
425: $\mu_{\ell e}= \partial (n_1 \epsilon(n_1))/\partial n_1$  
426: (a part from the constant term arising from the radial external
427: force).  
428: Note that under the assumption $a_{\perp} \ll a$, the two conditions 
429: $an_1 \ll 1$ and $a_{1D}n_1 \gg 1$ required to realise the 1D mean 
430: field regime can be simultaneously satisfied.
431: %
432: The other important limit is the low density 
433: Tonks-Girardeau limit $a_{1D}n_1\ll 1$,
434: where  the chemical potential takes the value
435: 
436: \begin{eqnarray}
437: \mu_{\ell e} = \pi^2 \hbar^2 n_1^2 /2m. 
438: \label{muT}
439: \end{eqnarray}
440: %
441: Here the chemical potential  no longer depends
442: on the interaction coupling constant and reveals  a typical 
443: Fermi-like behaviour \cite{tonks}. 
444: 
445: 
446: In the presence of axial harmonic
447: trapping   the ground state density profile has been evaluated 
448: in \cite{olshanii} using the local density approximation
449: (\ref{n0}).   In this case the normalization condition  
450: $\int n_1(z) dz=N$  takes the form
451: 
452: 
453: %
454: \begin{equation}
455: \frac{Za_{1D}}{a_z^2} 
456: \int_{-1}^{1} {\tilde \mu}_{\ell e}^{-1} 
457: \left[ \left( \frac{Za_{1D}}{a_z^2} \right)^{2} (1-t^2) \right] dt=
458: \frac{N a_{1D}^2}{a_z^2}.
459: \label{T5}
460: \end{equation}
461: %
462: Similarly to Eq.(\ref{s5}), 
463: we have introduced the  radius $Z$ at which the density vanishes and the
464: inverse of the function ${\tilde \mu}_{\ell e}(n_{1D}a_{1D})$,
465: where  ${\tilde \mu}_{\ell e}$ is now the chemical potential 
466: expressed in units of  $\hbar^2 /2ma^2_{1D}$.
467: Eq.(\ref{T5}) shows that the relevant combination of parameters 
468: is given by $N a_{1D}^2/a_z^2$. 
469: This differs by the factor $(a_{\perp}/a)^3$ from the combination 
470: $N a a_{\perp}^2/a_z^2$
471: characterizing the transition between the  
472: mean field regimes  discussed in the first part of this work. 
473: 
474: 
475: \begin{center}
476: \begin{figure}
477: \includegraphics[width=0.95\linewidth]{figure2.eps} 
478: \caption{ Transition between the Tonks-Girardeau and the 1D mean field  
479: regimes: $\omega^2/\omega_z^2$ as a function of the parameter 
480: $Na_{1D}^2/a_z^2$.}
481: \label{fig2}
482: \end{figure}
483: \end{center}
484: %
485: Analytic solutions are obtained in the two limits $N a_{1D}^2/a_z^2 \gg 1$ 
486: and $N a_{1D}^2/a_z^2 \ll 1$. In the first case, one recovers the mean field
487: result (\ref{n1MF}). In the second one, we find the  profile \cite{gir,kol}
488: 
489: \begin{eqnarray}
490: n_1(z) = \frac{\sqrt{2N}}{\pi a_z}
491: \left(1-\frac{z^2}{Z^2}\right)^{1/2},
492: \end{eqnarray}
493: % 
494: with $Z=\sqrt{2N}a_z$. 
495: In this case the  applicability of the local density approximation simply
496: requires $N \gg 1$.
497: 
498: 
499: By determining numerically the density profiles 
500: in the intermediate regimes we have calculated the frequency 
501: of the lowest compressional mode using the sum rule formula
502: (\ref{omega1-1}). 
503: The results are reported in Fig.~\ref{fig2} as a function of  
504: $Na_{1D}^2/a_z^2$. 
505: Fig.~\ref{fig1} shows the evolution of the collective frequency 
506: as a function of the parameter $Naa_{\perp}/a_z^2$ for two different 
507: choices of the ratio $a_{\perp}/a$. 
508: The corresponding curves  reveal the transition between 3D cigar,
509: 1D mean field and Tonks-Giradeau regimes. 
510: It is however worth pointing out that  if the condition
511: $a_{\perp} \gg a$ is not well satisfied, the 1D mean field regime 
512: is not clearly identifiable 
513:  (see for example the dotted line in Fig.~\ref{fig1}).
514: In this case the determination of the collective
515: frequencies   requires 
516: the simultaneous  inclusion of 3D and beyond mean field effects. 
517: 
518: 
519: This research is supported 
520: by the Ministero della Ricerca Scientifica e Tecnologica (MURST). 
521: 
522: \begin{references}
523: 
524: 
525: \bibitem{mit01} A.~G\"orlitz et al., Phys.~Rev.~Lett. {\bf 87},
526: 130402 (2001)
527: 
528: \bibitem{salomon} F.~Schreck, et al., Phys.~Rev.~Lett {\bf 87}, 080403 (2001)
529: 
530: \bibitem{hannover} S.~Dettmer et al, Phys.~Rev.~Lett {\bf 87}, 160406 (2001)
531: 
532: \bibitem{PS} M.~Schwartz, Phys.~Rev.~B {\bf 15}, 1399 (1977), 
533: F.D.M.~Haldane, Phys.~Rev.~Lett. {\bf 47}, 1840 (1981),
534: L.P.~Pitaevskii and S.~Stringari, J.~Low~Temp.~Phys. {\bf 85}, 377 (1991)
535: 
536: 
537: \bibitem{lieb} E.H.~Lieb and W.~Liniger, Phys.~Rev. {\bf 130},   
538: 1605 (1963); E.H.~Lieb, Phys.~Rev. {\bf 130}.
539: 
540: 
541: \bibitem{gora} D.S.~Petrov, G.V.~Shlyapnikov and J.T.M.~Walraven, 
542: Phys. Rev. Lett. {\bf 85}, 3745 (2000);
543: 
544: \bibitem{olshanii} V.~Dunjko, V.~Lorent, and M.~Olshanii,
545: Phys.~Rev.~Lett. {\bf 86}, 5413 (2001)
546: 
547: \bibitem{gir} \mbox{M.D.~Girardeau, 
548: E.M.~Wright,
549: and J.M.~Triscari,}
550: Phys.~Rev.~A {\bf 63}, 033601 (2001)
551: 
552: \bibitem{rmp} F.~Dalfovo, S.~Giorgini, L.P.~Pitaevskii and S.~Stringari,
553: Rev.~Mod.~Phys. {\bf 71}, 463 (1999).
554: 
555: \bibitem{tonks} L.~Tonks, Phys.~Rev. {\bf 50}, 955 (1936); M.~Girardeau,
556: J.~Math.~Phys. (N.Y.) {\bf 1}, 516 (1960)
557: 
558: \bibitem{sandro96} 
559: S.~Stringari, Phys. Rev. Lett. {\bf 77}, 2360 (1996);
560: 
561: \bibitem{zaremba} E.~Zaremba, Phys.~Rev.~A {\bf 57}, 518 (1998)
562: 
563: \bibitem{mit96} M.-O.~Mewes, et al.,
564: Phys.~Rev.~Lett. {\bf 77}, 988 (1996);
565: M.R.~Andrews, et al., Phys.~Rev.~Lett. 79, 553 (1997) 
566: 
567: \bibitem{csordas} A.~Csordas, R.~Graham,
568: Phys.~Rev.~A {\bf 59}, 1477 (1999)
569: 
570: \bibitem{sandro98} S.~Stringari, Phys.~Rev.~A {\bf 57}, 2385 (1998)
571: 
572: 
573: \bibitem{ho} T.-L.~Ho and M.~Ma, J. Low Temp. Phys. {\bf 115}, 61 (1999)
574: 
575: \bibitem{Minguzzi} A.~Minguzzi, P.~Vignolo, M.L.~Chiofalo, and M.P.~Tosi,
576: Phys.~Rev.~A {\bf 64}, 033605 (2001)
577: 
578: \bibitem{kol} E.B.~Kolomeisky, T.J.~Newman,
579: J.~P.~Straley, Phys. Rev. Lett. {\bf 85}, 1146 (2000)
580: 
581: \bibitem{note} Result (\ref{omega1-1}) provides in general an upper 
582: bound to the frequecy of the
583: lowest state excited by $z^2$. This bound coincides with 
584:  the exact frequency in the three relevant limits discussed
585: in the paper (3D cigar, 1D mean field and Tonks-Girardeau) 
586: 
587: 
588: \bibitem{olshanii2}  M.~Olshanii, Phys.~Rev.~Lett. {\bf 81}, 938 (1998)
589: 
590: \bibitem{noteolshanii} The averaging procedure and the consequent 
591: identification of the 1D coup
592: ling constant is justified only if 
593: $a_{\perp} \gg a$ \cite{olshanii2}.
594: 
595: 
596: \end{references}
597:   
598: 
599: 
600: \noindent
601: %FIGURE CAPTION:
602: 
603: \noindent
604: 
605: \end{document}
606: 
607: 
608: 
609: 
610: 
611: 
612: 
613: 
614: 
615: 
616: 
617: 
618: