1: %BPS.tex: Bosonization, pairing, and superconductivity of the
2: %fermionic Tonks-Girardeau Gas
3: %THIS FILE IS IN REVTEX4 FORMAT
4: %
5: \documentclass[prl,twocolumn,showpacs,superscriptaddress,amsfonts,amsmath,
6: floatfix]{revtex4}
7: \usepackage{graphicx}
8: \usepackage{psfrag}%To include LaTeX subscripts on figures
9: %
10: % Uncomment the following line if epsfig syntax is used for figures
11: %\usepackage{epsfig}
12: % Following line for graphicx figure syntax
13: %\usepackage[dvips]{graphicx}
14: % Note: Font size of figure axis numbers and labels can be changed
15: % by editing the .eps figure file. Such a file produced by gnuplot 3.7
16: % will contain a line like the following (without the prefixed percent sign):
17: % (Helvetica) findfont 140 scalefont setfont
18: % To increase size of font increase the 140. (For PRL and PRA figures try
19: % increasing 140 to 210.)
20: %
21: % Following line activates \bm{} command for boldface math symbols
22: %\usepackage{bm}
23: %
24: % A useful Journal macro
25: \newcommand{\Journal}[4]{#1 {\bf #2}, #3 (#4)}
26: %
27: % Journal names used in this paper
28: \newcommand{\PR}{Phys. Rev.}
29: \newcommand{\PRL}{Phys. Rev. Lett.}
30: \newcommand{\PRA}{Phys. Rev. A}
31: \newcommand{\PRB}{Phys. Rev. B}
32: \newcommand{\JMP}{J. Math. Phys.}
33: \newcommand{\EPJD}{Eur. Phys. J. D}
34: \newcommand{\EJDE}{Elec. J. Diff. Eqs.}
35: \newcommand{\Science}{Science}
36: \newcommand{\JCP}{J. Chem. Phys.}
37: \newcommand{\PLA}{Phys. Lett. A}
38: %
39: \begin{document}
40: \title {Bosonization, Pairing, and Superconductivity of the Fermionic Tonks-Girardeau Gas}
41: \author{M. D. Girardeau}
42: \email{girardeau@optics.arizona.edu}
43: \affiliation{College of Optical
44: Sciences, University of Arizona,
45: Tucson, AZ 85721, USA}
46: \author{A. Minguzzi}
47: \email{anna.minguzzi@grenoble.cnrs.fr}
48: \affiliation{Laboratoire de Physique et Mod\'elisation des Mileux Condens\'es,
49: C.N.R.S., B.P. 166, 38042 Grenoble, France}
50: \affiliation{Laboratoire de Physique Th\'{e}orique et Mod\`{e}les
51: Statistiques, Universit\'{e} Paris-Sud, B\^{a}t. 100, F-91405 Orsay, France}
52: %\date{\today}
53: %
54: \begin{abstract}
55: We determine some exact static and time-dependent properties of the fermionic
56: Tonks-Girardeau (FTG) gas, a spin-aligned one-dimensional Fermi gas
57: with infinitely strongly attractive zero-range odd-wave interactions. We show
58: that the two-particle reduced density matrix exhibits maximal superconductive
59: off-diagonal long-range order, and on a ring an FTG gas with an even number
60: of atoms has a highly degenerate ground state with quantization of
61: Coriolis rotational flux and high sensitivity to rotation and to external
62: fields and accelerations.
63: For a gas initially under harmonic confinement we show that during an expansion
64: the momentum distribution undergoes a ``dynamical bosonization'', approaching
65: that of an ideal Bose gas without violating the Pauli exclusion principle.
66: \end{abstract}
67: %
68: \pacs{03.75.-b,05.30.Jp}
69: \maketitle
70: %
71: If an ultracold atomic vapor is confined in a de Broglie wave guide with
72: transverse trapping so tight and temperature so low that the transverse
73: vibrational excitation quantum $\hbar\omega$ is larger than available
74: longitudinal zero point and thermal energies, the effective dynamics becomes
75: one-dimensional (1D) \cite{Ols98,PetShlWal00}, a regime currently under
76: intense experimental
77: investigation \cite{Tol04Mor04,Par04Kin04}. Confinement-induced 1D
78: Feshbach resonances (CIRs) reachable by tuning the 1D
79: coupling constant via 3D Feshbach scattering resonances
80: occur for both Bose gases \cite{Ols98} and spin-aligned Fermi gases
81: \cite{GraBlu04}. Near a CIR the 1D interaction is very
82: strong, leading to strong short-range correlations, breakdown of
83: effective-field theories, and emergence of highly-correlated $N$-body
84: ground states. In the bosonic case with very strong repulsion (1D hard-core
85: Bose gas with coupling constant $g_{1D}^{B}\to +\infty$, the
86: Tonks-Girardeau (TG) gas), the exact
87: $N$-body ground state was determined some 45 years ago by a
88: Fermi-Bose (FB) mapping to an ideal Fermi gas \cite{Gir60}, leading to
89: ``fermionization'' of many properties of this Bose system, as recently
90: confirmed experimentally \cite{Par04Kin04}.
91: The ``fermionic TG'' (FTG) gas \cite{GirOls03GirNguOls04}, a spin-aligned Fermi gas
92: with very strong \emph{attractive} 1D odd-wave interactions, can be realized by
93: 3D Feshbach resonance mediated tuning to the attractive side of the CIR with 1D
94: coupling constant $g_{1D}^{F}\to -\infty$. It has been pointed out
95: \cite{GraBlu04,GirOls03GirNguOls04} that the generalized FB mapping
96: \cite{CheShi98,GraBlu04,GirOls03GirNguOls04} can be exploited in the opposite
97: direction to map this system to the trapped \emph{ideal Bose} gas,
98: leading to determination of the exact $N$-body ground state and
99: ``bosonization'' of many properties of this Fermi system. We recently
100: examined the equilibrium one-body density matrix and
101: exact dynamics following sudden turnoff of the interactions by detuning from
102: the CIR \cite{GirWri05}. Here we determine some other exact properties of
103: the untrapped, ring-trapped, and harmonically trapped fermionic
104: TG gas, the most striking of which are pairing, superconductive off-diagonal
105: long-range order (ODLRO) of the two-body density matrix, a highly degenerate
106: ground state of an even number of atoms on a ring with quantization of
107: Coriolis rotational flux and high sensitivity to rotation and to external
108: fields and accelerations, and a ``dynamical
109: bosonization'' of the momentum distribution following sudden relaxation
110: of the trap frequency.
111: %
112:
113: {\it Untrapped FTG gas:} The Hamiltonian is
114: $\hat{H}=\sum_{j=1}^{N}
115: \left[-\frac{\hbar{^2}}{2m}\frac{\partial^2}{\partial x_{j}^{2}}\right]
116: +\sum_{1\le j<\ell\le N}v_\text{int}^{\text{F}}(x_{j}-x_{\ell})$
117: where $v_\text{int}^{\text{F}}$ is the two-body interaction. Since the spatial
118: wave function is antisymmetric due to spin polarization, there is no
119: s-wave interaction, but it has been shown
120: \cite{GraBlu04,GirOls03GirNguOls04} that a strong, attractive,
121: short-range odd-wave interaction (1D analog of 3D p-wave interactions)
122: occurs near the CIR. This can be modeled
123: by a narrow and deep square well of depth $V_0$ and width $2x_0$. The contact
124: condition at the edges of the well is \cite{GirOls03GirNguOls04}
125: $\psi_{F}(x_{j\ell}=x_{0})=-\psi_{F}(x_{j\ell}=-x_{0})
126: = -a_{1D}^{F}\psi_{F}^{'}(x_{j\ell}=\pm x_{0})$
127: where $a_{1D}^{F}$ is the 1D scattering length and the prime denotes
128: differentiation. Consider first the relative wave function $\psi_{F}(x)$
129: in the case $N=2$. The FTG limit is
130: $a_{1D}^{F}\to -\infty$, a zero-energy scattering resonance.
131: The exterior solution is $\psi_{F}(x)=\text{sgn}(x)=\pm 1$
132: ($+1$ for $x>0$ and $-1$ for $x<0$) and the
133: interior solution fitting smoothly onto this is $\sin(\kappa x)$ with
134: $\kappa=\sqrt{mV_{0}/\hbar^2}=\pi/2x_0$. In the zero-range limit
135: $x_{0}\to 0+$ the well area $2x_{0}V_{0}=(\pi\hbar)^{2}/2mx_{0}\to\infty$,
136: stronger than a negative delta function. In this limit the
137: wave function is discontinuous at contact $x_{0}=0\pm$, allowing an infinitely
138: strong zero-range interaction in spite of the
139: antisymmetry of $\psi_F$ \cite{CheShi98}. This generalizes immediately to
140: arbitrary $N$: the exact FTG gas ground state is
141: \begin{equation}
142: \label{eq:psiF}
143: \psi_{F}(x_{1},\cdots,x_{N})=
144: A(x_{1},\cdots,x_{N})\prod_{j=1}^{N}\phi_{0}(x_{j})
145: \end{equation}
146: with
147: $A(x_{1},\cdots,x_{N})=\prod_{1\le j<\ell\le
148: N}\text{sgn}(x_{\ell}-x_{j})$ the ``unit antisymmetric function''
149: employed in the original discovery of fermionization \cite{Gir60} and
150: $\phi_0=1/\sqrt{L}$ the ideal Bose gas ground orbital, $L$ being the
151: periodicity length. Its energy
152: is zero \cite{Note1} and it satisfies periodic boundary conditions for odd
153: $N$ and antiperiodic boundary conditions for even $N$ \cite{commentbc}.
154:
155: The exact single-particle density matrix
156: $\rho_{1}(x,x')=N\int\psi_{F}(x,x_{2},\cdots,x_{N})
157: \psi_{F}^{*}(x',x_{2},\cdots,x_{N})dx_{2}\cdots dx_N$ is
158: \cite{BenErkGra04,GirWri05}
159: $\rho_1(x,x')=N\phi_{0}(x)\phi_{0}^{*}(x')[F(x,x')]^{N-1}$ with
160: $F(x,x')=\int_{-L/2}^{L/2}\text{sgn}(x-y)\text{sgn}(x'-y)|\phi_{0}(y)|^{2}dy
161: =1-2|x-x'|/L$. In the thermodynamic limit
162: $N\to\infty$, $L\to\infty$, $N/L=n$ this gives an exponential decay
163: \cite{BenErkGra04}: $\rho_1(x,x')=ne^{-2n|x-x'|}$. Its Fourier transform
164: $n_k$, normalized to $\sum_{k}n_{k}=N$ (allowed momenta $\nu 2\pi/L$ with
165: $\nu=0,\pm 1,\pm 2,\cdots$), is the momentum distribution function
166: $n_{k}=[1+(k/2n)^2]^{-1}$. It satisfies the exclusion principle limitation
167: $n_{k}\le 1$, but nevertheless, for $n\to 0$ the
168: continuous momentum density $n(k)=(L/2\pi)n_k$ reduces to $N$
169: times a representation of the Dirac delta function, simulating the
170: ideal \emph{Bose} gas distribution:
171: $n(k)\longrightarrow_{_{\hspace{-.6cm}n\to 0}}N\delta(k)$
172: \cite{BenErkGra04}.
173:
174: The two-particle density matrix $\rho_{2}(x_{1},x_{2};x_{1}',x_{2}')
175: = N(N-1)\int \psi_{F}(x_{1},..,x_{N})
176: \psi_{F}^{*}(x_{1}',x_{2}',x_{3},..,x_{N}) dx_{3}...dx_{N}$
177: also has a simple closed form:
178: %
179: \begin{eqnarray}
180: \label{eq:rho2}
181: & &\rho_{2}(x_{1},x_{2};x_{1}',x_{2}')=
182: N(N-1)\text{sgn}(x_{1}-x_{2})\phi_{0}(x_{1})\phi_{0}(x_{2})\nonumber\\
183: &\times&\text{sgn}(x_{1}'-x_{2}')
184: \phi_{0}^{*}(x_{1}')\phi_{0}^{*}(x_{2}')
185: [G(x_{1},x_{2};x_{1}',x_{2}')]^{N-2}
186: \end{eqnarray}
187: %
188: where $[G(x_{1},x_{2};x_{1}',x_{2}')]^{N-2}
189: =[\int_{-L/2}^{L/2}\text{sgn}(x_{1}-x)\text{sgn}(x_{2}-x)
190: \text{sgn}(x_{1}'-x)\text{sgn}(x_{2}'-x)|\phi_{0}|^{2}(x)dx]^{N-2}
191: =e^{2n(y_{1}-y_{2}+y_{3}-y_{4})}$ in the thermodynamic limit and
192: $y_{1}\le y_{2}\le y_{3}\le y_{4}$ are the arguments
193: $(x_{1},x_{2};x_{1}',x_{2}')$ in ascending order.
194: $\rho_2$ is of order $n^2$ in the following cases: (a)
195: $|x_{1}-x_{1}'|\le \text{O}(1/n)$,\ $|x_{2}-x_{2}'|\le \text{O}(1/n)$; (b)
196: $|x_{1}-x_{2}'|\le \text{O}(1/n)$,\ $|x_{2}-x_{1}'|\le \text{O}(1/n)$; (c)
197: $|x_{1}-x_{2}|\le \text{O}(1/n)$,\ $|x_{1}'-x_{2}'|\le \text{O}(1/n)$.
198: These are just Yang's criteria \cite{Yan62} for superconductive ODLRO of
199: $\rho_2$ in the absence of ODLRO of $\rho_1$. In case (c) $\rho_2$ remains
200: of order $n^2$ for arbitrarily large separation of the centers of
201: mass $X=(x_{1}+x_{2})/2$ and $X'=(x_{1}'+x_{2}')/2$, the hallmark
202: of ODLRO. On the other hand, in cases (a) and (b) $\rho_2$
203: decays exponentially with $|X-X'|$. In the thermodynamic limit only
204: configurations (c) contribute to the largest eigenvalue of $\rho_2$,
205: and $\rho_2$ separates apart from negligible contributions (a) and (b)
206: \cite{Note2}:
207: %
208: \begin{eqnarray}
209: \rho_{2}(x_{1},x_{2};x_{1}',x_{2}')
210: =n^{2}\text{sgn}(x_{1}-x_{2})e^{-2n|x_{1}-x_{2}|}\nonumber\\
211: \times\ \text{sgn}(x_{1}'-x_{2}')e^{-2n|x_{1}'-x_{2}'|}
212: +\text{terms negligible for }\lambda_{1}.
213: \end{eqnarray}
214: %
215: By Yang's argument \cite{Yan62} the largest eigenvalue is $\lambda_{1}=N$,
216: and this is confirmed by comparison with the $\lambda_1$ contribution
217: $\lambda_{1}u_{1}(x_{1},x_{2})u_{1}(x_{1}',x_{2}')$ to the spectral
218: representation of $\rho_{2}$, implying that the corresponding
219: eigenfunction is $u_{1}(x_{1},x_{2})
220: =\mathcal{C}\ \text{sgn}(x_{1}-x_{2})e^{-2n|x_{1}-x_{2}|}$ with \cite{Note3}
221: $\mathcal{C}=\sqrt{n/L}$, confirming the value
222: $\lambda_{1}=n^{2}/{\mathcal{C}}^{2}=N$.
223: The range $1/2n$ of $u_1$ is in the region of onset
224: of a BEC-BCS crossover between tightly bound bosons and loosely bound Cooper
225: pairs. There is an upper bound \cite{Yan62}
226: $\lambda_{1}\le N$ on the largest eigenvalue, so
227: \emph{the untrapped FTG gas is maximally superconductive} in the
228: sense of Yang's ODLRO criterion.
229:
230: {\it FTG gas on a ring:} If the FTG gas is trapped on a
231: circular loop of radius $R$, with particle
232: coordinates $x_j$ measured around the circumference $L=2\pi R$, the
233: FTG gas must satisfy periodic boundary conditions for both odd and even
234: $N$ because of single-valuedness of its wave function. Since the
235: mapping function $A(x_{1},\cdots,x_{N})=\prod_{1\le j<\ell\le
236: N}\text{sgn}(x_{\ell}-x_{j})$ is periodic (antiperiodic) for odd (even)
237: $N$ as a result of its definition, it follows that the mapped
238: ideal Bose gas used to solve the FTG problem must satisfy periodic
239: (antiperiodic) boundary conditions
240: for odd (even) $N$. The ground state of a FTG
241: gas on a ring is then different depending on the particle number parity.
242: For odd $N$ the FTG ground state in Eq.~(\ref{eq:psiF}) is built from
243: the zero-momentum orbital $\phi_0=1/\sqrt{L}$ and corresponds to mapping the
244: FTG gas onto the ideal Bose gas ground state, the usual complete
245: Bose-Einstein condensate (BEC), and is nondegenerate.
246: On the other hand, for even $N$, which we henceforth assume, antiperiodicity
247: requires that the only plane-wave orbitals allowed are
248: $e^{ikx_j}/\sqrt{L}$ with $k=\pm\pi/L,\pm 3\pi/L,\cdots$. The
249: ground state of this fictitious ideal Bose gas, and hence that of the
250: mapped FTG gas, is then $(N+1)$-fold degenerate, with energy eigenvalue
251: $N(\hbar^2/2m)(\pi/L)^2$. These degenerate ground states
252: are fragmented BECs with $wN$ atoms in the orbital $e^{i\pi x_j/L}$
253: and $(1-w)N$ in $e^{-i\pi x_j/L}$ with $0\le w\le 1$, and are conveniently
254: labelled by a quantum number
255: $\ell_z=(w-\frac{1}{2})N=0,\pm 1,\pm 2,\cdots,\pm \frac{N}{2}$
256: related to the eigenvalue
257: $P$ of circumferencial linear momentum and that $L_z$ of angular momentum
258: $z$-component by $P=\ell_z \hbar/R$ and $L_z=\ell_z\hbar$. The angular
259: momentum per particle is half-integral due to
260: antiperiodicity of the orbitals, and the degenerate ground states
261: are in one-one correspondence with the eigenstates of \emph{spin}
262: angular momentum $z$-component of $N$ spin-$1/2$ fermions.
263:
264: The ground state degeneracy makes the FTG gas on a ring a good candidate
265: for detecting small external fields and linear accelerations. Suppose that
266: there is a potential gradient parallel to a diameter of the
267: ring, or an acceleration leading
268: to a gradient in the inertial potential arising from Einstein's
269: principle of equivalence, with the circumferential minimum
270: of this potential occurring at a point $x_0$. Then the degeneracy
271: is lifted and to lowest order in degenerate perturbation theory
272: all $N$ atoms occupy
273: the orbital $\phi_0(x)=\sqrt{2/L}\cos[\pi(x-x_0)/L]$, leading to an
274: observable asymmetric density profile $n(x)=2n\cos^2[\pi(x-x_0)/L]$.
275:
276: Due to its quantum coherence the FTG gas is also a good candidate
277: for a sensitive rotation detector. Suppose that the
278: ring trap is rotating with angular velocity $\vec{\omega}$
279: perpendicular to the plane of the ring. In the rotating
280: coordinate system each atom sees an effective Coriolis force
281: $\vec{F}_{\text{Cor}}=2m\vec{v}\mathbf{\times}\vec{\omega}$.
282: Comparing this with the usual magnetic force
283: $\vec{F}_{\text{mag}}=(e/c)\vec{v}\mathbf{\times}\vec{B}$, one sees that
284: the kinetic energy operators in the Hamiltonian in the rotating system
285: are $[\hat{p}_j-\frac{h}{L}\frac{\Phi}{\Phi_0}]^2/2m$ where
286: $\hat{p}_j=(\hbar/i)\partial/\partial x_j$, $\Phi=\pi R^2\omega$ is the
287: Coriolis flux through the loop, and $\Phi_0=h/2m$ is the Coriolis flux
288: quantum. The energy of each state $|\ell_z\rangle$ then becomes
289: $E=E_0(\Phi=0)+\frac{N\hbar^2}{2mR^2}[(\frac{\Phi}{\Phi_0})^2
290: -2\ell_z\frac{\Phi}{\Phi_0}]$ which is minimized when
291: $\ell_z=\frac{N}{2}$ if $\Phi>0$ and $\ell_z=-\frac{N}{2}$ if $\Phi<0$,
292: i.e., even a very small angular velocity leads to a nondegenerate ground state
293: with all $N$ atoms at either $k=\pi/L$ or $k=-\pi/L$. Generalizing to states
294: differing from the $\Phi=0$ ground states by displacement in $k$-space
295: by integral multiples of $2\pi/L$ one obtains the $\Phi$-dependent
296: ground state energy $E_0(\Phi)$ shown by the heavy line in Fig. \ref{fig1},
297: in which the lighter lines show the lowest energies for
298: $\ell_z=\pm\frac{N}{2},\pm\frac{3N}{2},\cdots.$ The ground
299: state energy is a periodic function of $\Phi$ with period
300: $\Phi_0$ in accord with a general theorem \cite{Yan62}, but unlike the usual
301: situation for a superconductor, (a) there is no smaller period $\Phi_0/2$,
302: and (b) for even $N$, $\Phi_0=0$ is a relative maximum of $E_0$ rather than a
303: minimum (as is the case of odd $N$),
304: the first minima occuring at $\Phi=\pm\Phi_0/2$.
305: %**********************************************************************
306: \begin{figure}
307: %[htbp]
308: \centering
309: \psfrag{N hbar^2/2mR^2}{$E$ $[N\hbar^2/2mR^2]$}
310: \psfrag{Phi/Phi0}{$\Phi/\Phi_0$}
311: \psfrag{lz=-5N/2}{$\frac{-5N\hbar}{2}$}
312: \psfrag{lz=-3N/2}{$\frac{-3N\hbar}{2}$}
313: \psfrag{lz=-N/2}{$\frac{-N\hbar}{2}$}
314: \psfrag{lz=N/2}{$\frac{N\hbar}{2}$}
315: \psfrag{lz=3N/2}{$\frac{3N\hbar}{2}$}
316: \psfrag{lz=5N/2}{$\frac{5N\hbar}{2}$}
317: \includegraphics[width=7.5cm,angle=0]{fig1.BPS.eps}
318: \caption{Dependence of energies $E$ on rotational flux $\Phi$.
319: Heavy line: Ground state energy $E_0(\Phi)$. Lighter lines: Lowest energy
320: for each value of total angular momentum.}
321: \label{fig1}
322: \end{figure}
323: %**********************************************************************
324: The barrier heights of the energy landscape in Fig.\ref{fig1} vanish
325: like $1/N$ for $N\to\infty$, so flux quantization
326: will not be observable for a macroscopic ring. However, it may be observable
327: for mesoscopic rings using BEC-on-a-chip technology. For example, assuming
328: a ring radius $R=$5 $\mu$m, one finds that for $^6\mbox{Li}$,
329: $\Delta E>k_B T$ for $T<50$ nK.
330:
331: {\it Flow properties on a nonrotating ring:}
332: According to the FB
333: mapping the excitation spectrum of the FTG gas is the same as that of
334: an ideal Bose gas, and hence it is sufficient to analyze the latter.
335: Since the excitation energy of the
336: ideal Bose gas is quadratic in the excitation momentum $\hbar q$,
337: the FTG gas does not satisfy the Landau-Bogoliubov criterion for
338: superfluidity. We investigate here the possibility
339: of flow metastability associated with barriers in the
340: excitation energy landscape as a function of the transferred momentum.
341: It was shown by F. Bloch
342: \cite{Blo73} that for the usual ideal Bose gas, which corresponds to the
343: case of odd $N$ in our treatment, no such barriers exist. In the case of even $N$ both the ground state and the excitation branches are $(N+1)$-fold
344: degenerate, but it is sufficient here to consider
345: the $\ell_z=0$ ground state and the excitations arising from it by promoting
346: atoms to higher $k$-values. Generalizing Bloch's analysis,
347: we note that for $0<\nu\le N/2$ the lowest branch corresponds to excitation
348: of $\nu$ atoms from $k=-\pi/L$ to $k=3\pi/L$, yielding a state with
349: angular momentum $z$-component $\ell_z\hbar$ with $\ell_z=2\nu$, and
350: with excitation energy $\epsilon(\ell_z)=\ell_z\hbar^2/mR^2$. At $\nu=N/2$
351: one has reached a state differing from the ground state by translation
352: of all atoms by an amount $2\pi/L$ in $k$-space, and one can repeat
353: this process, promoting atoms from $k=\pi/L$ to $5\pi/L$,
354: yielding another straight-line segment connecting the points $\ell_z=N$
355: and $\ell_z=2N$ on a parabolic curve $(\ell_z\hbar)^2/2NmR^2$, etc.
356: Together with symmetry $\epsilon(\ell_z)=\epsilon(-\ell_z)$ this yields an
357: excitation energy curve composed of straight-line segments as in the dashed curve of Bloch's Fig. 2 \cite{Blo73} with the notation
358: $P=\ell_z\hbar/R$. Hence for both odd and even $N$
359: there are no energy barriers, and
360: the FTG gas on a nonrotating ring does not exhibit flow metastability.
361:
362: {\it Expansion from a longitudinal harmonic trap:} We focus finally on
363: a 1D expansion, as could be achieved by keeping on the transverse confinement.
364: If the 1D interactions are suddenly turned off before the gas is let free
365: to expand from a longitudinal harmonic trap, the density profile at long times
366: reflects the initial momentum distribution \cite{GirWri05}. If instead the
367: interactions are kept on during the expansion we find that the density
368: profile expands self-similarly, while the momentum distribution
369: evolves from an initial overall Lorentzian shape \cite{BenErkGra04} to
370: that of an ideal Bose gas.
371: These properties can be demonstrated with the aid of an exact scaling
372: transformation as we outline below. Since the
373: FB mapping holds also for time-dependent phenomena induced by
374: one-body external fields \cite{time-dep}, the exact many-body wavefunction
375: $\psi_F(x_{1},\cdots,x_{N};t)=A(x_{1},\cdots,x_{N})\prod_{j=1}^{N}
376: \phi_{0}(x_{j};t)$
377: during the dynamics is fully determined by the solution of the single-particle
378: Schr\"{o}dinger equation for the orbital $\phi_{0}(x_{j};t)$. For the
379: case of an external potential $V_{ext}(x,t)=m\omega(t)^2 x^2/2$ with
380: $\omega(0)=\omega_0$ the solution is known \cite{perelomov} to be
381: $\phi_0(x;t)=\phi_0(x/b(t);0)e^{imx^2\dot b/ 2b\hbar-iE_0\tau(t)/\hbar}$
382: where $b(t)$ is the solution of the differential equation $\ddot
383: b+\omega^2(t) b=\omega^2_0/b^3$ with $b(0)=1$ and $\dot b(0)=0$,
384: $\tau(t)=\int_0^t dt'\,1/b^2$ and $E_0=\hbar \omega_0/2$. Since the unit
385: antisymmetric wavefunction $A$ is invariant under the scaling
386: transformation, we immediately obtain the expression for the many-body
387: wavefunction,
388: $\psi_F(x_{1},..,x_{N};t)=b^{-N/2}\psi_F(x_{1}/b,..,x_{N}/b;0)
389: e^{i(\dot b/b\omega_0) \sum_{j=1}^N x_j^2/2x_{osc}^2}
390: e^{-iNE_0 \tau(t)/\hbar}$,
391: and for the one-body density matrix,
392: $\rho_1(x,x';t)=\frac{1}{b}\rho_1\left(\frac{x}{b},\frac{x}{b};0\right) \exp\left[-i \frac{\dot b}{b}\frac{(x^2-x'^2)}{2x_{osc}^2}\right].$
393: This yields the momentum distribution as a function
394: of time. While the intermediate-time dynamics has to be determined
395: numerically, the stationary-phase method determines the
396: long-time evolution of the momentum distribution
397: in the same way as for the bosonic TG gas \cite{MinGan05}. For the case of a 1D
398: expansion the scaling parameter is $b(t)=\sqrt{1+\omega_0^2t^2}$ and the
399: momentum distribution tends to that of an ideal Bose gas under
400: harmonic confinement,
401: \begin{equation}
402: \label{eq:bos}
403: n(k,t\rightarrow\infty)\simeq |\omega_0/\dot b| n_B(k\omega_0/\dot b),
404: \end{equation}
405: where $n_B(k)=2\pi N |\tilde \phi_0(k)|^2$, with $\tilde \phi_0(k)=\pi^{-1/4}k_{osc}^{-1/2}e^{-k^{2}/2k_{osc}^2}$ and $k_{osc}=1/x_{osc}$.
406: This behavior is illustrated in Fig.~\ref{fig2}. Quite noticeably, the
407: ``bosonization'' time appears to be much longer than the ``fermionization''
408: time of the momentum distribution of the bosonic TG gas \cite{MinGan05}.
409: \begin{figure}
410: %[htbp]
411: \centering
412: \psfrag{p/p_ho}{$k/k_{osc}$}
413: \psfrag{t=0}{$t=0$ $\longrightarrow$}
414: \psfrag{t=0}{$t=0$ $\longrightarrow$}
415: \psfrag{t=5}{$t=5$ $\longrightarrow$}
416: \psfrag{t=10}{$t=10$ $\longrightarrow$}
417: \psfrag{t=20}{$t=20$ $\longrightarrow$}
418: \psfrag{t=30}{$t=30$ $\longrightarrow$}
419: \psfrag{t=40}{$t=40$ $\longrightarrow$}
420: \psfrag{ideal Bose gas}{ideal Bose gas $\longrightarrow$}
421: \psfrag{n(p,t) p_ho}{$n(k,t)\, k_{osc}$}
422: \includegraphics[width=7.5cm,angle=0]{fig2.BPS.eps}
423: \caption{Momentum distributions of a FTG gas (solid lines) with $N=9$ particles as
424: functions of the wavevector $k$ at subsequent times $t$ (in units
425: of $1/\omega_0$) during a 1D expansion, and asymptotic long-time
426: expression (\ref{eq:bos}) (dashed line).
427: }
428: \label{fig2}
429: \end{figure}
430: Note that the ``dynamical bosonization'' described above
431: does not violate the Pauli exclusion principle: by using the above scaling solution
432: for the one-body density matrix and fixing unit
433: normalization of the natural orbitals at all times it follows that the
434: eigenvalues $\alpha_j$ of $\rho_1(x,x';t)$ are invariant during the expansion
435: %\cite{Note3}
436: and hence always satisfy the condition $\alpha_j\le 1$.
437: %
438:
439: In conclusion, we have found that (a) the untrapped system exhibits
440: superconductive ODLRO of the two-body density matrix $\rho_2$
441: associated with its maximal eigenvalue $N$ and pair eigenfunction
442: $\mathcal{C}\ \text{sgn}(x_{1}-x_{2})e^{-2n|x_{1}-x_{2}|}$;
443: (b) on a ring it has a highly degenerate ground state for even atom number,
444: and it exhibits quantization of rotational Coriolis flux
445: and high sensitivity to rotation and to
446: accelerations, making it a good candidate for high-sensitivity detectors;
447: (c) the harmonically trapped system undergoes a ``dynamical bosonization''
448: of its momentum distribution during a 1D expansion.
449:
450: %
451: \begin{acknowledgments}
452: This work was initiated at the Aspen Center for Physics during the
453: summer 2005 workshop ``Ultracold Trapped Atomic Gases''. We are grateful to the
454: organizers, G~Baym, R.~Hulet,
455: E.~Mueller, and F.~Zhou, for the opportunity to participate, and to
456: S.~Giorgini, R.~Seiringer, F.~Zhou,
457: E.~Zaremba, and G.~Shlyapnikov for helpful comments.
458: The Aspen Center for Physics is supported by the
459: U.S. National Science Foundation, research of M.D.G. at the University of
460: Arizona by U.S. Office of Naval Research grant N00014-03-1-0427 through a
461: subcontract from the University of Southern California, and that of
462: A.M. by the Centre National de la
463: Recherche Scientifique (CNRS).
464: \end{acknowledgments}
465: %
466: \begin{thebibliography}{33}
467: %
468: \bibitem{Ols98} M. Olshanii, \Journal{\PRL}{81}{938}{1998}.
469: %
470: \bibitem{PetShlWal00} D.S. Petrov, G.V. Shlyapnikov, and J.T.M. Walraven,
471: \Journal{\PRL}{85}{3745}{2000}.
472: %
473: \bibitem{Tol04Mor04} B.L. Tolra {\it et al.}, \Journal{\PRL}{92}{190401}{2004};
474: T. St\"{o}ferle , H. Moritz, C. Schori, M. K\"{o}hl,
475: and T. Esslinger, {\it ibid.} {\bf 92}, 130403 (2004).
476: %
477: \bibitem {Par04Kin04} B. Paredes, {\it et al.},
478: Nature {\bf 429}, 277 (2004);
479: T. Kinoshita, T. Wenger, and D.S. Weiss, \Journal{\Science}
480: {305}{1125}{2004}.
481: %
482: \bibitem{GraBlu04} B.E. Granger and D. Blume, \Journal{\PRL}{92}{133202}{2004}.
483: %
484: \bibitem{Gir60} M. Girardeau, \Journal{\JMP}{1}{516}{1960}; M.D. Girardeau,
485: \Journal{\PR}{139}{B500}{1965}, Secs. 2,3, and 6.
486: %
487: \bibitem{GirOls03GirNguOls04} M.D. Girardeau and M. Olshanii, cond-mat/0309396;
488: M.D. Girardeau, Hieu Nguyen, and M. Olshanii,
489: Optics Communications {\bf 243}, 3 (2004).
490: %
491: \bibitem{CheShi98} T. Cheon and T. Shigehara, \Journal{\PLA}{243}{111}{1998}
492: and \Journal{\PRL}{82}{2536}{1999}.
493: %
494: \bibitem{GirWri05} M.D. Girardeau and E.M. Wright,
495: \Journal{\PRL}{95}{010406}{2005}.
496: %
497: \bibitem{Note1} The infinitely large negative interaction energy is exactly
498: cancelled by the infinitely large positive kinetic energy, as one sees
499: by passing to the zero-range limit $x_{0}\to 0+$ of the finite square well.
500: %
501: \bibitem{commentbc} This is the most convenient choice to pass to the the
502: thermodynamic limit
503: of an infinitely long linear trap containing an FTG gas of uniform density.
504: %
505: \bibitem{BenErkGra04} S.A. Bender, K.D. Erker, and B.E. Granger,
506: \Journal{\PRL}{95}{230404}{2005}.
507: %
508: \bibitem{Note2} The omitted terms determine the smaller eigenvalues, which
509: are of order unity and smaller, but make
510: the dominant contribution to $\text{Tr}\ \rho_{2}=N(N-1)$; see \cite{Yan62}.
511: They are important for the diagonal elements of $\rho_2$. From the
512: exact expression for $G$ one sees, on setting $y_{1}=x_1$, $y_{2}=x_{1}'$,
513: $y_{3}=x_2$, and $y_{4}=x_{2}'$, that
514: $\rho_{2}(x_{1},x_{2};x_{1},x_{2})=n^2$.
515: %, in agreement with
516: %the result from the FB mapping, valid for diagonal
517: %but not off-diagonal elements of $\rho_2$.
518: %
519: \bibitem{Yan62} C.N. Yang, Rev. Mod. Phys. {\bf 34}, 694 (1962),
520: Sec.~18 and Appendix A.
521: %
522: \bibitem{Note3} To correctly calculate $\mathcal{C}$ one must periodically
523: extend $\rho_2$ and $u_1$ in accordance with Sec. 37 of \cite{Yan62}.
524: %
525: %\bibitem{commentoddeven} For even $N$ the antiperiodic Bose gas is
526: % a
527: % mathematical convenience used to implement the FB mapping and is not
528: % realizable by a Bose gas on a ring.
529: %
530: \bibitem{Blo73} F. Bloch, \Journal{\PRA}{7}{2187}{1973}.
531: %
532: %\bibitem{Reg03} C.A. Regal, C. Ticknor, J.L. Bohn, and D.S. Jin,
533: %\Journal{\PRL}{90}{053201}{2003}.
534: %
535: %\bibitem{Sim01} R.W. Simmonds {\it et al.},
536: % A. Marchenkov, E. Hoskinson, J.C. Davis, and R.E. Packard,
537: %Nature {\bf 412}, 55 (2001).
538: %
539: %\bibitem{Alb05} M. Albiez {\it et al.}, \Journal{\PRL}{95}{010402}{2005}.
540: %
541: %\bibitem{BlaBut55} J.M. Blatt and S.T. Butler, \Journal{\PR}{100}{476}{1955}.
542: %
543: %\bibitem{NhoBlu05} K. Nho and D. Blume, cond-mat/0506156.
544: %
545: \bibitem{time-dep} M.D. Girardeau and E.M. Wright,
546: Phys.~Rev.~Lett. \textbf{84}, 5691 (2000).
547: %
548: \bibitem{perelomov} A.M. Perelomov and V.S. Popov, Sov. Phys. JETP
549: {\bf 30}, 910 (1970); A.M. Perelomov and Y.B. Zel'dovich, {\em Quantum
550: Mechanics}, (World Scientific, Singapore, 1998).
551: %
552: \bibitem{MinGan05} A. Minguzzi and D.M. Gangardt, \Journal{\PRL}{94}{240404}{2005}.
553: %
554: \end{thebibliography}
555: %
556: \end{document}
557: