1: \documentclass[prl,showpacs,twocolumn]{revtex4}
2: %\documentclass[prl,showpacs,preprint]{revtex4}
3: \usepackage{epsfig}
4: %\usepackage{refcheck}
5: \catcode`\"=\active \let"=\" \let\3=\ss
6: \begin{document}
7: \bibliographystyle{revtex}
8: \title{Charge and spin dynamics of interacting Fermions in a
9: one dimensional harmonic trap}
10: \author{Lars Kecke}
11: \affiliation{Physikalisches Institut, Universit\"at Freiburg,
12: Hermann-Herder-Str.~3, D-79104 Freiburg, Germany}
13: \author{Wolfgang H\"ausler}
14: \affiliation{Physikalisches Institut, Universit\"at Freiburg,
15: Hermann-Herder-Str.~3, D-79104 Freiburg, Germany}
16: \affiliation{Institut f\"ur Theoretische Physik,
17: Universit\"at Erlangen, Staudtstr.~7, D-91058 Erlangen,
18: Germany}
19: \author{Hermann Grabert}
20: \affiliation{Physikalisches Institut, Universit\"at Freiburg,
21: Hermann-Herder-Str.~3, D-79104 Freiburg, Germany}
22:
23: \begin{abstract}
24: We study an atomic Fermi gas interacting through repulsive contact
25: forces in a one dimensional harmonic trap. Bethe-Ansatz solutions
26: lead to an inhomogeneous Tomonaga-Luttinger model for the low energy
27: excitations. The equations of motion for charge and spin density
28: waves are analyzed both near the trap center and near the trap edges.
29: While the center shows conventional spin-charge separation the edges
30: cause a giant increase of the separation between these modes.
31: \end{abstract}
32: \pacs{73.22.Lp, 32.80.Pj}
33: \maketitle
34: Recent advances in cooling technology have allowed to reach the degenerate
35: regime of Fermionic quantum gases. Different hyperfine states effectively
36: correspond to spin polarizations of spin-$\frac{1}{2}$ particles
37: of respective densities $n_{\uparrow}$ and $n_{\downarrow}$ \cite{demarco}.
38: Furthermore, the spatial dimensionality of the gases can be reduced using
39: either a hollow beam setup \cite{bongs} or arrays of microtraps
40: \cite{reichel} so that one dimensional (1D) gases can be studied, where
41: interaction effects are known to be most pronounced. As a consequence of
42: pure $s$-wave scattering in 3D \cite{olshanii} the interaction is short
43: ranged \cite{morigi} and acts only between particles of opposite spins.
44: Also, the strength of the forces can, in principle, be varied over wide
45: ranges \cite{loftus} by tuning the Feshbach resonance \cite{feshbach}. An
46: optical lattice along the trap can further enhance correlations by
47: reducing the bandwidth and thus the kinetic energy
48: \cite{zwerger,esslinger}.
49:
50: As opposed to higher dimensions, total charge
51: ($\rho=n_{\uparrow}+n_{\downarrow}$) and relative spin
52: ($\sigma=n_{\uparrow}-n_{\downarrow}$) density waves comprise the only low
53: energy excitations in 1D; no Fermionic quasiparticles can be excited in
54: this regime. The Tomonaga-Luttinger liquid (TLL) theory
55: \cite{tomolut,haldane} describes the properties of homogeneous 1D systems
56: entailing that charge and spin modes move at different velocities when
57: interactions are present. This spin-charge separation is considered as a
58: hallmark of TLL behavior. A parabolic trap potential, however,
59: causes the particle density to vary along the trap. To treat such an
60: inhomogeneous gas cloud, new Boson representations of the Fermi operators
61: have been introduced \cite{wonneberger}. Here, we follow the other route
62: put forward recently by Recati et al.\ \cite{zwerger} and
63: consider an inhomogeneous TLL \cite{schulz} with $x$-dependent parameters, assuming
64: a trap potential which is slowly varying on the scale of the Fermi wavelength.
65:
66: In the bulk of the gas cloud, away from the trap edges,
67: interactions are sufficiently weak to justify perturbative
68: estimates to the (local) TLL model parameters. In this regime,
69: the picture of charge and spin waves moving at different velocities
70: is recovered \cite{jagla,zwerger}. Near the trap edges,
71: however, the gas density decreases and interactions become
72: very strong. In this regime we employ the Bethe-Ansatz solution for
73: spin-$\frac{1}{2}$ particles with contact forces
74: \cite{yang,liebwu,takahashi} to compute the TLL parameters
75: exactly \cite{shiba,coll}. To leading order in the inverse
76: interaction strength we can access the dynamics analytically.
77: Charge modes are found to propagate in a similar way as in
78: the bulk of the gas cloud. At the edges they are reflected and
79: thus keep oscillating in the trap until damping processes
80: become significant. Remarkably, spin density waves show an
81: exponential slowing down of their velocity and, ultimately,
82: accumulation at an edge without reflection. Thus, Fermionic
83: 1D quantum gases in shallow confinements should exhibit a giant
84: increase of spin-charge separation, much more pronounced than
85: electrons in quantum wires \cite{voit}. The effect of spin
86: accumulation should be detectable: one way would be to observe
87: (e.g.\ by fluorescence measurements) the time evolution of an initial
88: spin-up density peak containing charge and spin modes of equal
89: amounts, $\rho_\uparrow=(\rho+\sigma)/2$.
90:
91: Without parabolic confinement the quantum gas is described by
92: the Hamiltonian $(\hbar=1)$
93: \begin{equation}
94: H_{{\rm hom}}=\frac{1}{2m}\left[\sum_{i=1}^N-\frac{\partial^2}
95: {\partial x_i^2}+2c\sum_{i<j}\delta(x_i-x_j)\right]\,.
96: \label{Hhom}\end{equation}
97: Effectively, by virtue of the Pauli principle, only Fermions of
98: opposite spins are interacting. As long as the 3D
99: scattering length $a$ is much smaller than the transversal
100: width $d$ of the gas cloud, the interaction strength $c=2a/d^2$
101: \cite{olshanii}.
102:
103: For given particle density $n$, the ground state energy density
104: $\epsilon_{{\rm hom}}(n)$ is obtained exactly from the Hamiltonian
105: (\ref{Hhom}) via the Bethe-Ansatz \cite{liebwu}.
106: In presence of the longitudinal trap potential $V(x)$, we determine
107: the inhomogeneous profile of the particle density $n(x)$ in the spirit
108: of the local density approximation (LDA) \cite{zwerger} by minimizing
109: \begin{equation}
110: E[n]=\int dx\left[\epsilon_{{\rm hom}}(n(x))+
111: V(x)n(x)-\mu n(x)\right]\;.
112: \label{lda}\end{equation}
113: This is adequate for slow variations of $V$
114: compared to the Fermi wavelength and for large particle numbers,
115: $N\gg 1$, where we can ignore Friedel oscillations occurring on the scale
116: of the Fermi wavelength $\pi/2k_{{\rm F}}$ \cite{wonneberger}.
117: The chemical potential $\mu$ ensures that $N=\int dx\;n(x)$ and
118: determines the length $2R$ of the gas cloud through $V(R)=\mu$,
119: independent of $V(|x|<R)$ and the form of
120: $\epsilon_{{\rm hom}}(n)$, provided
121: $\partial_n\epsilon_{{\rm hom}}(n=0)=0$ \cite{astra}. Assuming now a
122: harmonic trap potential $V(x)=\frac{m}{2}\omega_{{\rm T}}^2x^2$ of
123: frequency $\omega_{{\rm T}}$, one finds from Eq.~(\ref{lda})
124: \begin{equation}
125: \partial_n\epsilon_{{\rm hom}}(n(x))+
126: \frac{m}{2}\omega_{{\rm T}}^2x^2-\mu=0\;.\label{ldan}\end{equation}
127: The resulting density profile $n(x)$ is depicted in Fig.~\ref{dichte}
128: \cite{olshanii01}.
129:
130: \begin{figure}
131: \centerline{\epsfig{file=dichteprofil.eps,width=8cm}}
132: \caption{LDA density profile of interacting Fermions in a trap
133: of length $2R$ for dimensionless interaction strength $u=0.3$,
134: using the full Bethe-Ansatz equations numerically (solid line),
135: Eq.~(\ref{ncent}) (dashed line),
136: and Eq.~(\ref{nedge}) (dotted line); $n_0=\frac{2}{\pi}
137: m\omega_{{\rm T}} R$ is the density of a non-interacting Fermi gas
138: at the same $\mu$.}
139: \label{dichte}
140: \end{figure}
141:
142: To proceed we now focus on two regions of particular interest.
143: Near the center of the trap, where the density $n$ is large enough
144: to validate perturbative expressions in the interaction strength,
145: one finds to leading order in $c/n$ \cite{mahan}
146: \begin{equation}
147: \epsilon_{{\rm pert}}(n)=\frac{\pi^2n^3}{24m}
148: \left(1+\frac{c}{\pi^2n}\right)\;. \label{epert}\end{equation}
149: Inserting (\ref{epert}) into (\ref{ldan}) gives a modified
150: Thomas-Fermi profile
151: \begin{equation}
152: n_{{\rm pert}}(\xi)=\frac{2m}{\pi}v_{{\rm F}0}
153: \left(\sqrt{1+u^2-\xi^2}-u\right)\;.
154: \label{ncent}\end{equation}
155: Here we have introduced the dimensionless
156: coordinate $\xi=x/R$ and interaction strength $u=c/m\pi v_{{\rm F}0}$
157: where $v_{{\rm F}0}=R\omega_{{\rm T}}$ is the Fermi velocity of a
158: homogeneous, non-interacting system at chemical potential $\mu$.
159: On the other hand, near the trap edges, $c\gg n$ so that we can
160: evaluate the Bethe-Ansatz solution to leading order in $n/c$,
161: yielding \cite{zwerger}
162: \begin{equation}
163: \epsilon_{{\rm edge}}(n)=\frac{\pi^2n^3}{6m}
164: \left(1-\frac{4n}{c}\ln 2\right)\;.
165: \label{eba}\end{equation}
166: Inserting (\ref{eba}) into (\ref{ldan}) gives
167: \begin{equation}
168: n_{{\rm edge}}(\xi)=\frac{2m}{\pi}\frac{v_{{\rm F}0}}{2}
169: \left(\sqrt{1-\xi^2} +\frac{8\ln 2}{3\pi^2u}(1-\xi^2)\right)\;.
170: \label{nedge}\end{equation}
171: As seen in Fig.~\ref{dichte}, the analytic forms (\ref{ncent})
172: and (\ref{nedge}) describe the trap
173: density accurately in their regions of validity.
174: We stress that $n_{{\rm edge}}(\xi)$ vanishes with infinite slope
175: at the edges $\xi\to\pm 1$, which is not seen from the
176: perturbative expression $n_{{\rm pert}}(\xi)$.
177:
178: The inhomogeneous TLL Hamiltonian is of the form
179: \begin{equation}\label{htl}
180: H_{{\rm TLL}}=\frac{1}{2}\sum_{\nu=\rho,\sigma}\int dx
181: \big[v_{{\rm J}\nu}(x)(\Theta_\nu'(x))^2+v_{{\rm N}\nu}(x)(\Phi_\nu'(x))^2
182: \big]\;,
183: \end{equation}
184: where primes denote spatial derivatives, and $\Theta_\nu$ and
185: $\Phi_\nu$ are the usual Bosonic phase fields in the charge $(\nu=\rho)$
186: and spin $(\nu=\sigma)$ sectors, obeying
187: $[\Phi_\nu(x),\Theta_{\nu'}'(x')]={\rm i}\delta_{\nu\nu'}\delta(x-x')$.
188: They define density excitations
189: $\nu=\langle\Phi_\nu'\rangle/\sqrt{\pi}$.
190: The velocity parameters $v_{{\rm N}\nu}(x)$ follow from exact
191: compressibility relations \cite{haldane,voit,whlk}
192: \begin{equation}
193: v_{{\rm N}\nu}=\frac{2}{\pi}\frac{\partial^2\epsilon_{{\rm hom}}}
194: {\partial\nu^2}\;,\label{vofe}
195: \end{equation}
196: while symmetries fix the values of velocity parameters $v_{{\rm J}\nu}$:
197: $v_{{\rm J}\rho}=\pi n/2m$ owing to Galilei invariance and
198: $v_{{\rm J}\sigma}=v_{{\rm N}\sigma}$ as a consequence of SU(2)
199: invariance in the spin sector. The model (\ref{htl}) can be justified if
200: $\nu\ll n$ which, evidently, fails too close to the trap edges where
201: $n\to 0$, see below.
202:
203: After separating off the time dependence $\sim{\rm e}^{{\rm i}\omega t}$
204: in the Heisenberg equations of motion, the eigenvalue equations for the
205: density excitations become
206: \begin{eqnarray}\label{dyn}
207: \mbox{}-v_{{\rm N}\nu}v_{{\rm J}\nu}\nu''
208: -\left(2v'_{{\rm N}\nu}v_{{\rm J}\nu}
209: +v_{{\rm N}\nu}v'_{{\rm J}\nu}\right)\nu'&&\\
210: \mbox{}-\left(v'_{{\rm N}\nu}v'_{{\rm J}\nu}
211: +v''_{{\rm N}\nu}v_{{\rm J}\nu}\right)\nu&=&
212: \omega^2\nu\;.\nonumber
213: \end{eqnarray}
214: As boundary conditions we require that
215: charge and spin currents vanish at the trap edges, $j_\nu(x=\pm
216: R)=0$. In view of (\ref{htl}) and the identity
217: $j_\nu=\dot\Theta_\nu/\sqrt{\pi}$, this boundary condition becomes
218: $v_{{\rm J}\nu}v_{{\rm N}\nu}'\nu+v_{{\rm J}\nu}v_{{\rm N}\nu}\nu'=0$
219: at $\xi=\pm 1$ which, together with (\ref{dyn}), governs the
220: dynamics of density wave packets. In the homogeneous case
221: $v'_{{\rm (N/J)}\nu}$ and $v''_{{\rm (N/J)}\nu}$ vanish and (\ref{dyn})
222: simplifies to a wave equation. In the inhomogeneous case, without
223: interactions, all four TLL velocity parameters are equal and given by
224: $v_{{\rm F}}(\xi)=v_{{\rm F}0}\sqrt{1-\xi^2}$ which follows from
225: Eqs.~(\ref{ldan}) and~(\ref{vofe}) for $\epsilon_{{\rm hom}}(n)=
226: \pi^2n^3/24m$. Then the solutions of (\ref{dyn}) \cite{minguzzi}
227: \begin{equation}
228: \nu_{{\rm nonint}}(x,\omega)=\frac{1}{\sqrt{R^2-x^2}}
229: \cos\left(\bar\omega\:{\rm arccos}\:\xi\right)\, ,\label{nonint}
230: \end{equation}
231: with integer $\bar\omega=\omega/\omega_{{\rm T}}$, constitute a complete
232: orthonormal set with regard to the measure $\frac{2R}{\pi\sqrt{R^2-x^2}}$
233: \cite{sturmliouville}.
234:
235: Near the trap center, we obtain from Eqs.~(\ref{ncent}) and~(\ref{vofe})
236: to leading order in $u$
237: \begin{eqnarray}\label{vpert1}
238: v_{{\rm J}\rho}&=&v_{{\rm F}}
239: =v_{{\rm F}0}\left(\sqrt{1+u^2-\xi^2}-u\right)\\
240: v_{{\rm N}\rho}&=&v_{{\rm F}}+u v_{{\rm F}0}
241: =v_{{\rm F}0}\sqrt{1+u^2-\xi^2}\label{vpert2}\\
242: v_\sigma&=&\sqrt{v_{{\rm F}}(v_{{\rm F}}-uv_{{\rm F}0})}
243: \approx v_{\sigma 0}\sqrt{1-x^2/R_\sigma^2}\;,
244: \label{vpert3}\end{eqnarray}
245: where $v_{\sigma 0}=v_{{\rm F}0}\sqrt{1-3u}$ and
246: $R_\sigma=R\sqrt{1-3u/2}$. The solutions of (\ref{dyn}) now become
247: \begin{eqnarray}\label{rhonx}
248: \rho_n(x)&=&\frac{n/2}{\sqrt{(1+u^2)R^2-x^2}}{\rm C}_n^{\left(h/2\right)}
249: \left(\frac{x}{R_2}\right)\\
250: \sigma_n(x)&=&\frac{1}{\sqrt{R_\sigma^2-x^2}}\cos
251: \left(n\:{\rm arccos}\:\frac{x}{R_\sigma}\right)\;,
252: \label{sigmanx}\end{eqnarray}
253: where the ${\rm C}_n^{(\alpha)}$ denote ultraspherical
254: polynomials \cite{AS}, $R_2=R\sqrt{1-u/2}$, $h=u/(2-u)$, and the
255: eigenfrequencies are $\omega_{\rho,n}=\omega_{{\rm T}}\sqrt{n(n+h)/(1+h)}$,
256: $\omega_{\sigma,n}=\omega_{{\rm T}} n\sqrt{(1-3u)(1-3u/2)}$. Thus, in
257: the charge sector the spectrum is no more
258: equidistant (see also Refs.~[\onlinecite{zwerger,astra}]),
259: although the lowest mode ($n=1$) remains unaffected by interactions
260: in accordance with Kohn's theorem \cite{kohn}.
261: Eqs.~(\ref{rhonx}) and~(\ref{sigmanx}) describe how weak
262: interactions smoothly deform the non-interacting solution
263: (\ref{nonint}); note that
264: ${\rm C}^{(0)}_n(\xi)=\frac{2}{n}\cos(n\:{\rm arccos}\:\xi)$. The
265: inequalities $R_2\omega_{\rho,n}>n v_{{\rm F}0}$ and
266: $R_\sigma\omega_{\sigma,n}<n v_{{\rm F}0}$ reflect the enhanced and
267: suppressed dynamics of charges and spins, respectively.
268:
269: Near the trap edges, in the non-perturbative regime where
270: $\sqrt{R^2-x^2}\ll uR$ (but still $R-|x|\gg k_{{\rm F}}^{-1}$),
271: interactions alter the above picture, even qualitatively. Using
272: Eqs.~(\ref{eba}) and~(\ref{nedge}), and employing the Bethe-Ansatz
273: result by Coll \cite{coll}, we get from (\ref{vofe}) to leading
274: order in $u^{-1}\sqrt{1-\xi^2}$
275: \begin{eqnarray}
276: v_{{\rm J}\rho}&=&v_{{\rm F}0}[\sqrt{1-\xi^2}/2+
277: \beta(1-\xi^2)/2]\label{vj}\\
278: v_{{\rm N}\rho}&=&v_{{\rm F}0}[2\sqrt{1-\xi^2}-
279: 4\beta(1-\xi^2)]\label{vn}\\
280: v_\sigma&=&\frac{4\pi m v_{{\rm F}}^2}{3c}=
281: \frac{v_{{\rm F}0}}{3u}(1-\xi^2)
282: \label{vs}
283: \end{eqnarray}
284: where $\beta=8\ln 2/(3\pi^2u)$. Disregarding for the moment
285: subleading terms $\propto\beta$ in Eqs.~(\ref{vj})
286: and (\ref{vn}), we see that the charge sector exponent
287: $K_{\rho}=\sqrt{v_{{\rm J}\rho}/v_{{\rm N}\rho}}$ approaches the
288: value $1/2$ close to the edges, which is consistent with the limiting
289: value of $K_{\rho}$ in the Hubbard model when the filling goes to
290: zero \cite{schulz90}. The plasmon velocity
291: $\sqrt{v_{{\rm N}\rho}v_{{\rm J}\rho}}$, on the other hand, takes
292: exactly the value of the non-interacting system [cf.\
293: Eqs.~(\ref{vpert1}) and (\ref{vpert2}) for $u\to 0$]. This amazing
294: property originates from the fact that the particle density
295: is reduced by a factor $1/2$ at infinite interactions, compared
296: to the non-interacting system.
297: Furthermore, since Eq.~(\ref{dyn}) only
298: contains products $v_{{\rm N}\nu}v_{{\rm J}\nu}$ and their spatial
299: derivatives, we conclude that even the whole charge dynamics
300: coincides exactly with the non-interacting dynamics (\ref{nonint})
301: sufficiently close to the trap edges.
302:
303: To leading non-trivial order in $\beta$, Eq.~(\ref{dyn}) can be
304: transformed into a damped Mathieu equation, that is solved by
305: \begin{equation}
306: \rho_n(x)=\frac{\exp(\frac{3}{2}\beta\sqrt{1-\xi^2})}{\sqrt{R^2-x^2}}
307: \left\{{se_{2n}(q,\frac{1}{2}\arcsin\xi)\,,n\,\mbox{odd}}
308: \atop{ce_{2n}(q,\frac{1}{2}\arcsin\xi)\,,n\,\mbox{even}}\right\}
309: \label{rhomat}\end{equation}
310: with the Mathieu functions $se_n$ and $ce_n$. The eigenenergies
311: $\omega_n$ follow from the characteristic values \cite{AS}
312: $a_{2n}(q_n)=4\omega_n^2-9\beta^2$ of the Mathieu functions while
313: $q_n=-\beta(2\bar\omega_n^2-1)$. When $\beta\sqrt{1-\xi^2}\to
314: 0$, the $\rho_n(x)$ turn into the non-interacting solutions
315: (\ref{nonint}), in agreement with the above observations for
316: the vicinity of the trap edges.
317:
318: In the spin sector the corresponding solutions,
319: \begin{equation}
320: \sigma(x,\omega)=\frac{R}{R^2-x^2}\left\{\sin\atop\cos\right\}
321: \left(\frac{3u\omega}{\omega_{{\rm T}}}\:{\rm Atanh}\:\xi\right)\;,
322: \end{equation}
323: differ qualitatively from the non-interacting solutions
324: (\ref{nonint}), as well as from the weakly interacting ones (\ref{sigmanx}).
325: They are rapidly oscillating near the trap edges which
326: manifests extremely slow spatial propagation of spin density waves,
327: as we discuss now.
328:
329: With the eigenfunctions for charge and spin densities at hand,
330: we are in the position to solve for the time evolution of an initial
331: wave packet $\nu_0(x)$ in the trap center, or near the edge.
332: After preparation, the packet splits into left-
333: and right moving parts, $\nu_+$ and $\nu_-$, and
334: the charge and spin constituents separate. Near the edge the time
335: evolution follows
336: \begin{eqnarray}
337: \rho_\pm(x,t)&\approx&\frac{1}{2}\sqrt{\frac{R^2-x'^2}{R^2-x^2}}
338: \rho_0(x')\nonumber\\
339: x'&=&R\cos\left({\rm arccos}\:\xi\pm\omega_{{\rm T}} t\right)\label{rhopm}\\
340: \sigma_\pm(x,t)&=&\frac{1}{2}\frac{R^2-x'^2}{R^2-x^2}\sigma_0(x')\nonumber\\
341: x'&=&R\tanh\left({\rm Atanh}\:\xi\pm\frac{\omega_{{\rm T}}}{3u}t\right)\;.
342: \label{sigmapm}\end{eqnarray}
343: Subleading corrections of order $\beta$ require a summation over
344: the eigenfunctions (\ref{rhomat}) and can no longer be expressed in
345: the simple form (\ref{rhopm}).
346:
347: \begin{figure}
348: \centerline{\epsfig{file=chargeprop.ps,angle=270,width=7cm}}
349: \centerline{\epsfig{file=spinprop.ps,angle=270,width=7cm}}
350: \caption{(Color online) Propagation of a charge density (a) and a
351: spin density (b) excitation for dimensionless interaction strength
352: $u=0.3$. Time steps are $\Delta t\omega_{{\rm T}}=0.25$ as
353: indicated. The charge wave is reflected at the edge of the trap,
354: similar to the non-interacting case, while the
355: spin wave exponentially slows down without reflection.}
356: \label{evfig}
357: \end{figure}
358:
359: It follows from Eq.~(\ref{rhopm}) that the charge density exhibits
360: a temporal evolution that roughly resembles that in the
361: absence of interactions: the right moving part of the initial
362: wave packet slows down somewhat due to the decreasing Fermi
363: velocity, and then is reflected at the edge (cf.\ Fig.~\ref{evfig}a).
364: We thus expect
365: charge density excitations to keep oscillating in the trap,
366: before damping mechanisms set in. On the contrary, the right
367: moving part of the spin density wave is slowed down even exponentially
368: fast, see Eq.~(\ref{sigmapm}), and is not reflected
369: (cf. Fig.~\ref{evfig}b). We mention that for long times
370: Eq.~(\ref{sigmapm}) gives $\sigma_\pm(x,t\to\infty)\to A_0
371: \delta(x\pm R)$ where $A_0=\int dx\:\sigma_0(x)$
372: is the conserved magnitude of the initial spin wave packet.
373: This follows from the finite slope of $v_\sigma$
374: at $\xi=\pm 1$, as opposed to the infinite slope $v'_{\rho{\rm (N/J)}}$ at
375: $\xi=\pm 1$ in the charge sector. However, our analysis is only valid
376: up to times of order $t_c\approx\frac{u}{\omega_{{\rm T}}}
377: \left[\ln\left(\frac{2mR\omega_{{\rm T}}}
378: {\pi\sigma_0(\xi_0)}\sqrt{1-\xi_0^2}\right)+3\ln\frac{2}{1+\xi_0}\right]$
379: since then $\sigma_\pm(x,t>t_c)$ exceeds $n(x)$. This time is large for small initial
380: displacements $\sigma_0$ so that a dramatic separation of the spin and
381: charge peaks will have occured well before effects on the length scale of
382: the Fermi wavelength, that have been disregarded here, become relevant.
383:
384: In conclusion, we have studied interacting Fermionic atoms of
385: two spin species in an effectively one dimensional harmonic trap
386: by exploiting solvability by the Bethe-Ansatz for contact forces.
387: We have evaluated the particle density profile within the local
388: density approximation. Near the trap center we confirm the
389: occurrence of spin-charge separation. Near the edges of the trap
390: interactions affect the dynamics in an unexpectedly drastic way.
391: While in the immediate vicinity of an edge charge density waves
392: move as if they were non-interacting with reflection at the
393: edge, spin density waves are not reflected and accumulate at the
394: edge. This establishes an even more pronounced spin-charge
395: separation than in homogeneous systems. Experimentally one
396: could confirm our predictions e.g.\ by selectively evaporating
397: parts of the `up' spin component $\rho_\uparrow$ in a small
398: region of space and observe the right and left moving spin and
399: charge density waves separating from one another (since
400: $2\rho_\uparrow=\rho+\sigma$), both in the bulk and near the
401: edge of the trap.
402:
403: We thank R. Egger, A. Komnik and W. Zwerger for valuable comments.
404: \begin{thebibliography}{99}
405: \bibitem{demarco}B. DeMarco and D.S. Jin, Science {\bf 285}, 1703 (1999);
406: S.R. Granade et al., Phys.\ Rev.\ Lett.\ {\bf 88}, 120405 (2002);
407: K.M. O'Hara et al., Science {\bf 298}, 2179 (2002).
408: \bibitem{bongs} K. Bongs et al., Phys.\ Rev.\ A {\bf 63},
409: 031602(R) (2001).
410: \bibitem{reichel} J. Reichel et al., Applied Physics B, {\bf 72}, 81 (2001);
411: J. Reichel and J.H. Thywissen, cond-mat/0310330.
412: \bibitem{olshanii} M. Olshanii, Phys.\ Rev. Lett.\ {\bf 81}, 938 (1998);
413: T. Bergeman, M.G. Moore, and M. Olshanii, Phys.\ Rev.\ Lett.\ {\bf 91},
414: 163201 (1998); cf.\ also C. Mora et al., Phys.\ Rev.\ Lett.\ {\bf 93},
415: 170403 (2004).
416: \bibitem{morigi} For long range interactions cf.\ G. Morigi and S. Fishman,
417: cond-mat/0407675.
418: \bibitem{loftus} T. Loftus et al., Phys.\ Rev.\ Lett.\ {\bf 88}, 173201 (2002).
419: \bibitem{feshbach} H. Feshbach, D.C. Peaslee, and V.F. Weisskopf,
420: Phys.\ Rev.\ {\bf 71}, 145 (1947);
421: E. Tiesinga, B.J. Verhaar, and H.T.C. Stoof, Phys.\ Rev.\ A {\bf 47},
422: 4114 (1993).
423: \bibitem{zwerger} A. Recati et al., Phys.\ Rev.\ Lett.\ {\bf 90},
424: 020401 (2003) and J.\ Optics B {\bf 5}, S55 (2003).
425: \bibitem{esslinger} H. Moritz et al., Phys.\ Rev.\ Lett.\ {\bf 91},
426: 250402 (2003);
427: T. Stoferle et al., Phys.\ Rev.\ Lett.\ {\bf 92}, 130403 (2004);
428: B. Peredes et al., Nature {\bf 429}, 277 (2004).
429: \bibitem{tomolut}
430: S. Tomonaga, Prog.\ Theo.\ Phys.\ {\bf 5}, 544 (1950);
431: J.M. Luttinger, J.\ Math.\ Phys.\ {\bf 4}, 1154 (1963).
432: \bibitem{haldane} F.D.M. Haldane, J.\ Phys.\ C {\bf 14}, 2585 (1981).
433: \bibitem{wonneberger} W. Wonneberger, Phys.\ Rev.\ A {\bf 63}, 063607 (2001);
434: G. Xianlong et al., Phys.\ Rev.\ A {\bf 67}, 023610 (2003).
435: \bibitem{schulz} I. Safi and H.J. Schulz, Phys.\ Rev.\ B {\bf 52},
436: R17040 (1995);
437: D.L. Maslov and M. Stone, Phys.\ Rev.\ B {\bf 52}, R5539 (1995).
438: \bibitem{jagla} E.A. Jagla, K. Hallberg, and C.A. Balseiro,
439: Phys.\ Rev.\ B {\bf 47}, 5849 (1993).
440: \bibitem{yang} C.N. Yang, Phys.\ Rev.\ Lett.\ {\bf 19}, 1312 (1967).
441: \bibitem{liebwu} E.H. Lieb and F.Y. Wu, Phys.\ Rev.\ Lett.\ {\bf 20},
442: 1445 (1968).
443: \bibitem{takahashi} M. Takahashi, Prog.\ Theo.\ Phys.\ {\bf 44}, 348 (1970).
444: \bibitem{shiba} H. Shiba, Phys.\ Rev.\ B {\bf 6}, 930 (1972).
445: \bibitem{coll} C.F. Coll, Phys.\ Rev.\ B {\bf 9}, 2150 (1974).
446: \bibitem{voit} J. Voit, Rep.\ Prog.\ Phys.\ {\bf 58}, 977 (1995).
447: \bibitem{astra} G.E. Astrakharchik et al., Phys.\ Rev.\ Lett.\
448: {\bf 92}, 030402 (2004) and cond-mat/0312538.
449: \bibitem{olshanii01} For Bosons a similar calculation based on
450: the Bethe-Ansatz has been carried out by V. Dunjko, V. Lorent,
451: and M. Olshanii, Phys.\ Rev.\ Lett.\ {\bf 86}, 5413 (2001).
452: \bibitem{mahan} G. D. Mahan, {\em Many Particle Physics},
453: Plenum Press, New York, 1991.
454: \bibitem{whlk} W. H\"ausler, L. Kecke, and A.H. MacDonald,
455: Phys.\ Rev.\ B {\bf 65}, 085104 (2002).
456: \bibitem{minguzzi} A. Minguzzi et al., Phys.\ Rev.\ A {\bf 64},
457: 033605 (2001).
458: \bibitem{sturmliouville} Note that the eigenvalue problem of
459: Eq.~(\ref{dyn}) is not of the Sturm-Liouville type.
460: \bibitem{AS} M. Abramowitz, I.A. Stegun (eds.), {\em Handbook of mathematical
461: functions}, Plenum Press, New York, 1981.
462: \bibitem{kohn} W. Kohn, Phys.\ Rev.\ {\bf 123}, 1242 (1961).
463: \bibitem{schulz90} H.J. Schulz, Phys.\ Rev.\ Lett.\ {\bf 64}, 2831 (1990).
464: \end{thebibliography}
465: \end{document}
466: