1: % -*- TeX:US -*-
2: \documentclass[aps,prl,twocolumn,superscriptaddress,showpacs]{revtex4}
3: %
4:
5: \usepackage{amsfonts}
6: \usepackage{graphicx}
7: %
8: \begin{document}
9: %
10: \title{Spin-$\frac{1}{2}$ fermions on spin-dependent optical lattices}
11: %
12: \author{M. ~A. Cazalilla}
13: \affiliation{Donostia Int'l Physics Center (DIPC), Manuel de Lardizabal,
14: 4. 20018-Donostia, Spain.}
15: %
16: \author{A.~F. Ho}
17: \affiliation{School of Physics and Astronomy, The University of Birmingham,
18: Edgbaston, Birmingham B15 2TT, UK.}
19: %
20: \author{T. Giamarchi}
21: \affiliation{University of Geneva, 24 Quai Enerst-Ansermet,
22: CH-1211 Geneva 4, Switzerland.}
23: %
24: \pacs{03.75.Ss,71.10.pm, 71.10.Fd, 05.30.Fk}
25: %
26: \begin{abstract}
27: We study the phase diagram of one dimensional spin-$\frac{1}{2}$
28: fermionic cold atoms. The two ``spin'' species can have different
29: hopping or mass. The phase diagram at equal densities of the species
30: is found to be very rich, containing Mott insulators and superfluids.
31: We also briefly discuss coupling 1D systems
32: together, and some experimental signatures of these phases. In
33: particular, we compute the spin structure factor for small momentum, which should allow the spin gap to be detected.
34: \end{abstract}
35: %
36: \date{\today}
37: \maketitle
38:
39: Quantum engineering~\cite{LWZ04,HCZ02}
40: of strongly correlated many-body systems has recently become
41: possible thanks to the spectacular advances in trapping ultracold
42: atoms in optical lattices~\cite{G02,M03,S04,K04} or in
43: microchip traps~\cite{Reichel}. This has led to the study of
44: models that would otherwise be hard to realize
45: in solids, which may shed light on basic issues in
46: quantum many body physics, including
47: the understanding of \emph{e.g.} the origin
48: of high-T$_c$ superconductivity in doped copper oxides.
49: In particular, correlated boson~\cite{G02,HCG04,DG04,S04},
50: Bose-Fermi~\cite{CH03},
51: and Fermi~\cite{FS04,FRZ04,GRJ03} systems have
52: received much experimental and
53: theoretical attention in recent times.
54:
55: In sharp contrast to electrons in solids, in cold atomic systems,
56: different types of atoms (different hyperfine states or different
57: atomic species) can be trapped and controlled \emph{independently},
58: such that the hopping, strength and sign of interactions
59: (inter- or intra-species) and densities can be \emph{continuously}
60: tuned. For example, Mandel \emph{et al.}~\cite{M03} managed to control
61: independently the periodic potential for each atom
62: type loaded in an optical lattice. A p-wave
63: Feshbach resonance~\cite{Sa04} can create a tunable
64: asymmetry in the interactions in a multi-species
65: Fermi gas. All this, of course, leads to a much
66: richer physics, which remains to be understood.
67:
68: In this paper, motivated by these recent developments and the
69: availability (now or soon) of fermions in elongated
70: traps~\cite{K04,Reichel}, we study the interesting effects of having
71: different Fermi velocities for two species of fermions in
72: one dimension (1D). With equal densities of the two species, this
73: system is different from the case of a two-leg
74: spinless ladder~\cite{LH00,G04} or the spin-$\frac{1}{2}$
75: electrons in a magnetic field~\cite{PS93,G04}. One main result
76: of this paper is the phase diagram
77: as a function of velocity difference, for equal densities
78: (Fig.~\ref{fig1}). With repulsive interactions, a finite velocity
79: difference breaks the SU(2) spin symmetry
80: and turns the gapless Tomonaga-Luttinger
81: liquid (TLL) into an Ising
82: spin-density wave with a spin gap. Demixing may occur if
83: one type of fermions has a very tiny velocity. With attractive
84: interactions, a singlet superfluid (SS) of bound pairs of fermions
85: of different types may give way to a charge density wave (CDW) of
86: pairs for sufficient velocity difference. We also briefly study the
87: effects of a small tunneling term coupling an array of 1D tubes
88: together. In particular, if there are different densities of
89: fermions in neighboring tubes, a triplet superfluid (TS) may become
90: stable for \emph{ repulsive} interactions. Finally, we describe how
91: to detect the spin gap in these phases by measuring the dynamical
92: spin structure factor, which we have computed. In contrast to
93: previous studies~\cite{FDL95,L04}, we have worked out
94: the phase diagram for equal number of spin up and down
95: fermions as a function of the (Fermi) velocity difference
96: and considered coupling the 1D systems together.
97:
98:
99: We study the following generalized Hubbard model:
100: %
101: \begin{equation}
102: H = -\sum_{\sigma, m}
103: t_{\sigma} \left( c^{\dag}_{\sigma m} c_{\sigma m+1} + {\rm H.c.} \right)
104: + U\sum_{m} n_{\uparrow m}
105: n_{\downarrow m}.
106: \label{Ham}
107: \end{equation}
108: %
109: This Hamiltonian describes a 1D Fermi gas with contact
110: interactions (related to $U$) \emph{prepared} with $N_{0\sigma}$ fermions
111: (i.e. we work in the \emph{canonical} emsemble) and loaded in a 1D
112: optical lattice with $m=1,\dots, M$
113: lattice sites; $n_{\sigma m} = c^{\dag}_{\sigma m} c_{\sigma m}$, and
114: $\sigma = \uparrow, \downarrow$ is the spin index that may refer
115: to two hyperfine states, or to {\it e.g.}
116: $^6$Li and $^{40}$K. Even though there may be no \emph{true} spin
117: symmetry, we will continue to use the spin (and magnetic) language
118: to describe this binary mixture.
119: We assume the number of fermions of
120: each spin species is \emph{separately} conserved, \emph{ i.e.} one
121: spin type cannot be converted to another. Motivated by the
122: experimental considerations above, we allow for different hopping
123: $t_{\sigma}$ for different spins. This Hamiltonian may be realized
124: in either a quasi-1D chip trap \cite{Reichel} or in a 2D optical
125: lattice, which is made up of an array of 1D gas
126: tubes~\cite{S04,HCG04} weakly coupled by a hopping $t_{\perp} \ll
127: \min\{t_{\uparrow}, t_{\downarrow} \}$. When the finite ``charging
128: energy'' (due to the finite length of each tube) exceeds the renormalized
129: hopping $E_{J} \propto t_{\perp}$~\cite{HCG04,unpub}, the tubes are
130: decoupled from one another and a set of independent 1D tubes is
131: recovered~\cite{HCG04}. Although we assume there is a (spin-dependent)
132: periodic potential parallel to the tubes such that (\ref{Ham}) applies,
133: much of what is discussed below also applies
134: in the absence of this potential when the two species have
135: different masses. More discussion on engineering Hamiltonians like
136: (\ref{Ham}) can be found in~\cite{LWZ04}.
137:
138: We first study the homogeneous 1D system
139: in the thermodynamic limit. Finite-size and trap effects
140: will be discussed below in connection with possible experiments.
141: %
142: \begin{figure}[t]
143: \begin{center}
144: \includegraphics[width=\columnwidth]{fig1.eps}
145: \caption{Schematic phase diagram for the model in Eq. (\ref{Ham})
146: with equal number of spin up and down fermions
147: away from half-filling (\emph{i.e.} $N_{\uparrow 0} = N_{\downarrow 0} \neq M/2$) .
148: The interaction strength is $U$ and
149: $z = |t_{\uparrow} - t_{\downarrow}|/
150: (t_{\uparrow}+t_{\downarrow})$. All phases (SDW: spin density
151: wave, CDW: charge density wave, SS singlet superfluid, TS triplet
152: superfluid) exhibit a spin gap $\Delta_s$
153: (however, $\Delta_s = 0$ for $U = 0$ and $z = 0$ with $U > 0$.
154: A cartoon of the type of order characterizing
155: each phase is also shown. In the area between dashed lines
156: the dominant order (either CDW or SS) depends on the lattice filling
157: (see text for more detailed explanations). In the SG phase,
158: spin up and down fermions are segregated (demixed).
159: \label{fig1}
160: }
161: \end{center}
162: \end{figure}
163: %
164:
165: The weak coupling limit $|U| \ll
166: \min\{t_{\uparrow},t_{\downarrow}\}$ can be solved by taking the
167: continuum limit of (\ref{Ham}) in the standard way \cite{G04} and
168: linearize the dispersion around the Fermi points $\pm k^{\sigma}_F =
169: \pi N_{0\sigma}/M a$, ($M$ is the number of lattice sites in the
170: tube). This leads to the so-called ``g-ology'' representation~\cite{G04} with a
171: finite number of coupling constants representing low energy
172: scattering processes. The coupling $g^{\sigma}_{2||}$
173: ($g_{2\perp}$) is the scattering amplitude for processes where a
174: small momentum $q$ is exchanged between fermions of equal (opposite)
175: spin at opposite Fermi points, for arbitrary values of $k^{\sigma}_{F}$.
176: On the other hand, $g_{1\perp}$ is the back-scattering
177: amplitude where two fermions of opposite spin exchange a momentum
178: $q \approx 2 k_F = 2 k^{\uparrow}_F = 2 k^{\downarrow}_{F}$, and is
179: relevant only when $N_{0\uparrow}=N_{0\downarrow}$; $g_{3\perp}$ is
180: the amplitude for \emph{umklapp} scattering ($q \approx
181: 2k^{\uparrow}_F + 2 k^{\downarrow}_F =\pi/a$) and is important only
182: at \emph{half-filling}: $N_{0\uparrow} + N_{0\downarrow} = M$. Thus
183: for generic fillings, $g_{1\perp}$ and $g_{3\perp}$ are irrelevant,
184: and the system is a TLL~\cite{G04}), which has a
185: completely gapless spectrum of two distinct branches of phonons.
186:
187: We focus here on the case $N_{0\uparrow} = N_{0 \downarrow} \neq M/2$.
188: The case of a half filled lattice $N_{0\uparrow} = N_{0\downarrow} = M/2$ is more
189: involved and will be reported elsewhere \cite{unpub}
190:
191:
192: The physical properties can be
193: established by analyzing the renormalization group (RG) flow of the
194: various scattering amplitudes upon the varying of a cutoff such as
195: the temperature $T$. To second order in the interaction
196: parameters, the RG flow is \cite{unpub}:
197: \begin{eqnarray}
198: \dot{y}^{\sigma}_{2||} &=& r_{-\sigma} y^{2}_{1\perp}, \quad \quad
199: \dot{y}_{2\perp} = -y^2_{1\perp},\label{RG1}\\
200: \dot{y}_{1\perp} &=& \left(r_{\uparrow} y^{\uparrow}_{2||} +
201: r_{\downarrow} y^{\downarrow}_{2||} - 2 y_{2\perp} \right) y_{1\perp},\label{RG2}
202: \end{eqnarray}
203: where $y_{\alpha} = g_{\alpha}/\pi \hbar v$ are dimensionless
204: couplings, $v = (v_{\uparrow} + v_{\downarrow})/2$ the mean
205: velocity and $r_{\sigma} = v/2v_{\sigma}$; $\dot{y}_{\alpha} =
206: dy_{\alpha}/d\ell$, with $\ell = \ln (\Lambda/T)$. Eqs.
207: (\ref{RG1},\ref{RG2}) can be mapped to the RG equations of the
208: Berezinskii-Kosterlitz- Thouless (BKT) transition in terms of
209: $y_{2s} = -\sum_{\sigma }r_{\sigma} y^{\sigma}_{2||}+ 2 y_{2\perp}$
210: and $y_{1\perp}$. The behavior of the BKT equations is entirely
211: determined \cite{G04} by the constant of motion ${\cal
212: C}=y^{2}_{1\perp} - y^2_{2s}/2(r_{\uparrow} r_{\downarrow} + 1) =
213: (Ua/\hbar v)^2 z^2/(2-z^2)$, where $z = |t_{\uparrow} - t_{\downarrow}|/(
214: t_{\uparrow} + t_{\downarrow})$ is the key velocity difference
215: parameter. For $z = 0$ we recover the well-known results for the
216: spin-symmetric Hubbard model~\cite{G04}. However, for $z\neq 0$
217: and $U\neq 0$, ${\cal C} > 0$, the scattering amplitude
218: $y_{1\perp}(\ell)$ diverges as the system is cooled down to its
219: ground state. This signals the formation of bound states, and the
220: opening of a gap in the spin sector (the charge excitations remain
221: gapless). For $z \ll 1$, the gap has thus the characteristic BKT
222: form $\Delta_s \sim \Lambda\: e^{-A/\sqrt{\cal C}} \simeq \Lambda \:
223: e^{-A'/|t_{\uparrow} - t_{\downarrow}|}$, where $\Lambda$ is of the
224: order of $t_{\uparrow} \approx t_{\downarrow}$ and $A, A'$ are
225: constants. Note that this gap is non-perturbative in $|t_{\uparrow}
226: - t_{\downarrow}|$.
227:
228: The properties of the spin-gapped phase depend on the sign of $U$.
229: Ground states of 1D systems are characterized by the dominant form
230: of order that they exhibit, which is typically quasi-long range in
231: character, true long-range order being only possible in 1D when a
232: \emph{discrete} symmetry is broken. For $U > 0$ and $z\neq 0$,
233: then $y_{1\perp}(\ell) \to \infty$, and a bosonization study \cite{unpub}
234: shows that the dominant order is a spin-density wave (SDW) and the
235: subdominant order is triplet superfluidity (TS). In the
236: attractive case ($U < 0$), as $z$ is increased, the dominant
237: order changes from a singlet superfluid phase (SS) to a
238: charge density wave (CDW), with CDW and SS being the
239: subdominant order in the former and latter case, respectively. We
240: wish to point out that our analysis takes fully into account the
241: marginal (in the RG sense) coupling between the gapless charge and
242: the gapped spin modes arising at $z\neq 0$, which leads to an often
243: substantial decrease in the value of the Luttinger-liquid parameter
244: $K_c$ (proportional to the charge compressibility). In particular,
245: for $U< 0$ we find that $K_c$ goes from $K_c>1$ to $K_c<1$ as $z$
246: is increased, which changes the character of the dominant
247: correlations from SS to CDW, as described above. A summary of the
248: phase diagram is shown in Fig.~\ref{fig1}.
249:
250: The weak-coupling regime smoothly crosses over to the strong
251: coupling regime $|U| \gg \max\{t_{\uparrow}, t_{\downarrow} \}$, as
252: confirmed by a strong coupling expansion of (\ref{Ham}). We only
253: give here the main steps, technical details can be found in
254: \cite{unpub}. It is simplest to first consider a half-filled lattice
255: with $N_{\uparrow 0} = N_{\downarrow 0} = M/2$. For the strongly
256: repulsive case $U \gg \max\{t_{\uparrow}, t_{\downarrow} \}$
257: fermions cannot hop around and there is a gap of order $U$ to charge
258: excitations. Degenerate perturbation theory~\cite{T77} then shows that in this limit
259: the Hamiltonian in (\ref{Ham}) maps to the Heisenberg-Ising (XXZ)
260: spin chain $H_{\rm XXZ} = J \sum_{m} \left[{\bf S}_{m}
261: \cdot {\bf S}_{m+1} + \gamma \, S^{z}_{m}
262: S^{z}_{m+1} \right]$, where the ${\bf S}_{m}$ denotes the spin operator
263: at site $m$, $J = 4 t_{\uparrow} t_{\downarrow}/U$, and the anisotropy $\gamma =
264: (t_{\uparrow} - t_{\downarrow})^2/2t_{\uparrow}t_{\downarrow}
265: \propto z^2$. Thus, for unequal hopping ($z > 0$), the
266: chain is in the Neel phase (SDW with true long-range order),
267: and has a spin gap which for small $z$ is
268: $\Delta_s \sim J\: e^{-A''/\sqrt{\gamma}} = J\:
269: e^{-A'''/|t_{\uparrow}-t_{\downarrow}|}$~\cite{FDL95}. Note the same
270: non-perturbative dependence on $t_{\uparrow}-t_{\downarrow}$ as for
271: the weak coupling regime. Away from half-filling, the system is
272: described by a $t$-$J$-like model with anisotropic spin interactions.
273: The charge gap is destroyed ($K_c = \frac{1}{2}$ close to
274: half-filling~\cite{G04}), but the spin gap remains and the dominant
275: order is still SDW. Physically, the finite velocity difference
276: breaks the SU$(2)$ spin symmetry to the lower Z$_2 \times$ U$(1)$.
277: Thus, the TLL becomes an Ising anti-ferromagnet in the spin sector.
278:
279: For $U\ll 0$,
280: degenerate perturbation theory shows that (\ref{Ham}) is equivalent
281: to a model of tightly bound fermion pairs (hard-core bosons)
282: annihilated by $b_{m} = c_{\uparrow m} c_{\downarrow m}$.
283: Their hopping amplitude is $J = 4t_{\uparrow}
284: t_{\downarrow}/|U|$, and they interact with strength $V = 2
285: (t^{2}_{\uparrow} + t^2_{\downarrow})/|U|$ when sitting at
286: nearest-neighbor sites. This model can be mapped to the above XXZ chain
287: via $b_{m} \to S^{-}_{m}$ and $\left(b^{\dag}_{m}
288: b_{m} - \frac{1}{2} \right) \to S^{z}_{m}$.
289: At half-filling, charge excitations are gapless for equal hopping
290: and SS is the dominant order~\cite{G04}. However, with unequal
291: hopping the spectrum of the tube is fully gapped, becoming a CDW
292: with true long-range order, a spin gap of order $|U|$ (energy
293: to break a pair), and a charge gap $\Delta_c \sim J \:
294: e^{-A/|t_{\uparrow} - t_{\downarrow}|}$. Away from half-filling, the
295: spin gap remains $\sim |U|$ but the bosons are able to hop
296: (\emph{i.e.} the charge gap disappears). It is worth pointing out
297: that very close to half-filling for $z\neq 0$, the dominant order is
298: CDW since $K_c \to \frac{1}{2}$, as can be inferred from the exact
299: solution of the XXZ chain~\cite{H80,G04}. However, as the filling
300: deviates more and more from half-filling, $K_c$ rises above one and
301: the system becomes a 1D superfluid (SS). This change in the
302: character of the dominant order also takes place at constant
303: filling, provided the system is sufficiently far from half-filling:
304: a SS ($K_c > 1$) can turn into a CDW ($K_c < 1$) as $|z|$ is varied
305: at strong coupling. This agrees with the above
306: weak coupling analysis. Note that at very low density
307: ($N_{0\sigma}/M\rightarrow 0$), and at least for not too different velocities,
308: only a SS phase is possible: in this limit, (\ref{Ham}) maps to a
309: continuum (Gaudin-Yang like) model of interacting fermions with
310: spin-dependent mass. For $U\to -\infty$, the fermions pair up to become a
311: 1D superfluid (SS) of tightly bound pairs with irrelevant residual
312: interactions between the pairs.
313:
314: Finally both the weak and strong coupling analysis described above
315: break down for sufficiently large $|t_{\uparrow} - t_{\downarrow}|$.
316: For weak coupling, linearization of the free fermion dispersion is
317: no longer justified when $t_{\uparrow} \gg t_{\downarrow}$ (or
318: viceversa), that is, for $z \to 1$. In the large $|U|$ limit
319: degenerate perturbation theory becomes quite subtle. Unfortunately,
320: rigorous results are available only for $t_{\uparrow} = 0$ or
321: $t_{\downarrow} = 0$ ($z = 1)$, which is the limit of the
322: Falicov-Kimball model. In 1D, Lemberger~\cite{L92} (see also
323: \cite{FLX}) has proved that spin up fermions segregates from spin
324: down ones for $U > U_c > 0$ at equal densities. There is no
325: segregation for $U<0$ at equal densities. As argued in
326: \cite{FLX,unpub}, it is quite likely that this segregated phase will
327: survive also when $|z|$ is not one but close to one.
328:
329: The predicted phase diagram of Fig.~\ref{fig1} for a single 1D tube
330: can be directly tested experimentally in cold atoms. However, it is
331: also interesting to analyze the case when
332: the tubes are weakly coupled by tunneling between the tubes.
333: We thus briefly describe the phase diagram for an array of
334: coupled 1D tubes in a 2D optical lattice geometry~\cite{S04,HCG04}. The
335: Hamiltonian for each tube at site ${\bf R}$ of the 2D lattice is
336: as in Eq.~(\ref{Ham}), with all fermion operators now carrying the $\bf R$
337: label. The hopping between the \emph{nearest} neighbor tubes
338: at ${\bf R}$ and ${\bf R}'$ is described
339: by $H_{\perp} = - t_{\perp} \sum_{\langle{\bf R}, {\bf R}'
340: \rangle} \sum_{m, \sigma} c^{\dag}_{\sigma {\bf R} m} c_{\sigma {\bf
341: R'} m}$, where $t_{\perp} \ll \min\{t_{\uparrow}, t_{\downarrow}
342: \}$, but such that fermions can now overcome the `charging energy' of the
343: finite-size tubes. In general, when the isolated tube has a gap
344: $\Delta_s \ll t_{\perp}$, $H_{\perp}$ is a relevant
345: perturbation (in the RG sense) that leads to coherent hopping of
346: fermions from tube to tube. Thus the ground state will most likely be a
347: very anisotropic 3D Fermi liquid, which in turn may become unstable
348: to 3D CDW/SDW formation or 3D BCS superfluidity under appropriate
349: conditions. This limit has been much studied in the past \emph{e.g.} in
350: connection with organic superconductors (see~\cite{chem_rev} for a review).
351: We shall not consider it here,
352: and instead we study
353: $t_{\uparrow} \neq t_{\downarrow}$ so that the gap $\Delta_s \gg t_{\perp}$.
354:
355: Since the tubes (or at least a large number of them near the center
356: of the trap due to inhomogeneity effects) can develop a sufficiently large
357: spin gap as described above, coherent hopping between tubes is now
358: suppressed. However, the term $H_{\perp}$ can generate,
359: through virtual transitions which are second order in $t_{\perp}$,
360: intertube interactions of two kinds~\cite{G04}: i) particle-hole
361: pair hopping generates spin-spin and density-density
362: interactions: $H_1= \sum_{m,\langle {\bf R}, {\bf R}'\rangle} \left[
363: J_{\perp} \: {\bf S}_{{\bf R} m} \cdot {\bf S}_{{\bf R}' m} +
364: V_{\perp} n_{m {\bf R}} n_{m {\bf R}'} \right]$, and ii) fermion pair hopping
365: yields $H_2 = J_{c \perp} \sum_{m, \langle {\bf R}, {\bf R}'\rangle}
366: b^{\dag}_{{\bf R} m} b_{{\bf R} m+j}$, where $V_{\perp},
367: J_{c\perp}\sim t^2_{\perp}/\Delta_s$,
368: $b_{{\bf R}m} = c^{\dag}_{\uparrow {\bf R} m} c_{\downarrow {\bf R} m}$
369: and $ja \lesssim \xi_s$ a distance smaller
370: than the spin correlation length, $\xi_s \propto \Delta^{-1}_s$.
371: The dominant term then drives a phase transition to a 3D ordered
372: phase~\cite{unpub}: for the $U < 0$ case, if the
373: tubes are in the SS phase, then the dominant process is
374: fermion pair tunneling, and the tubes develop 3D
375: long-range superfluid order. The low-temperature properties of this
376: system become identical to the superfluid of bosons studied in
377: \cite{HCG04}. However, if the tubes are in the CDW ($U < 0$ and
378: sufficiently large $z$) or in the SDW ($U>0$) phases, the dominant
379: interactions arise from hopping of particle-hole pairs and lead to
380: insulating phases that are either 3D CDW or SDW. The ordering
381: temperatures in all cases (at small $t_{\perp}$) are power-laws:
382: $T_c \propto \Delta_s (t_{\perp}/\Delta_s)^{\alpha}$, with $\alpha^{-1} =
383: 2(2-d)$ and $d$ the scaling dimension of the dominant
384: inter-tube interaction. Interestingly, the SDW or CDW ordering is
385: anisotropic: incommensurate (relative to the optical lattice) along
386: the tube, but commensurate perpendicular to the tube direction.
387:
388: Particle-hole hopping may drive a transition to a 3D insulating
389: state with density wave order \emph{only if} the density in
390: neighboring tubes are equal or very similar: for a particle and a
391: hole to hop coherently at low-temperatures, they must be extracted
392: from opposite Fermi points of one tube and must match the momenta in
393: the neighboring tube by momentum conservation. If the mismatch in
394: the density between tubes is sufficiently large, particle-hole hopping is
395: suppressed and only the hopping of fermion pairs (which carry zero net
396: momentum) is possible. The system will then order as a
397: superfluid. Interestingly, for $U > 0$, TS is the subdominant
398: order in the spin gapped phase of the tubes. Suppression of
399: particle-hole pair hopping may then lead to a 3D triplet superfluid.
400: We note that $T_c$ for these cases is also a power law of
401: $t_{\perp}/\Delta_s$.
402:
403: The phase diagram shown in Fig.~\ref{fig1} holds, strictly speaking,
404: in the thermodynamic limit. In real 2D optical lattices only a finite
405: number of fermions ($\sim 10^5$~\cite{K04}) can be loaded, but we expect all
406: predicted phases to appear. Due to the finite size, the phase
407: boundaries will not correspond to true phase transitions, but rather
408: to sharp crossovers. The trap can lead to phase coexistence and even to
409: suppression of quantum criticality~\cite{W04}, but we are concerned
410: here with the phases themselves and not with the quantum critical
411: points between them.
412:
413: The most important signature of the single tube phases that we
414: predict is the existence of a spin gap, $\Delta_s$. By measuring the absorption of a
415: laser that causes Raman transitions between the two hyperfine states
416: $\sigma =\uparrow, \downarrow$, it should be possible to measure~\cite{BZZ04} the small momentum limit of the dynamic structure factor $S_s({\bf q}, \omega)$, which is the Fourier transform of $S_s({\bf r},{\bf r}', t) =
417: \langle S^{+}({\bf r}, t) S^{-}({\bf r}', 0) \rangle$. Using the
418: so-called form factor approach~\cite{LZ01} ,
419: we find~\cite{unpub} that at $T \ll \Delta_s$, the structure factor rises from zero as
420: $\sqrt{(\hbar \omega)^2 - (2\Delta_s)^2}$ for $\hbar\omega \geq 2\Delta_s$.
421:
422: Concerning the coupled tubes, the most exotic phase is of course
423: the triplet superfluid (TS). To `engineer' it, we need to suppress
424: particle-hole hopping by making the number of fermions in neighboring
425: tubes sufficiently different. This could be achieved by imposing
426: a rapid spatial variation of the
427: trap potential, or better, by means of a biperiodic optical potential in the
428: direction perpendicular to the tubes. The coherence properties of
429: the 3D superfluid phases could be probed by exciting low frequency
430: collective modes in the transverse direction to the tubes. Coherent
431: oscillations should exist only in the superfluid phases and not in
432: the insulators.
433:
434: We acknowledge useful conversations with T. Esslinger and his group, M. Fabrizio,
435: J. Fjaerestad, A. Georges, M. Gunn, D. Haldane, B.
436: Marston, M. Long, A. Nersesyan, B. Normand and A. Tsvelik. M.A.C.
437: is supported by \emph{Gipuzkoako Foru Aldundia}, A.F.H. by EPSRC(UK)
438: and DIPC (Spain), and T.G. by the Swiss National Science Foundation
439: under MANEP and Division II.
440:
441: %
442: \begin{thebibliography}{30}
443: %
444: \bibitem{LWZ04}
445: W.~V. Liu, F.~Wilczek, and P.~Zoller, Phys.~Rev.~ A {\bf 70}, 033603 (2004), and
446: references therein.
447: %
448: \bibitem{HCZ02}
449: W. Hofstetter {\rm et al.}, Phys.~Rev.~Lett. {\bf 89}, 220407 (2002).
450: %
451: \bibitem{G02}
452: M. Greiner \emph{et al.}, Nature (London) {\bf 415}, 39 (2002).
453: %
454: \bibitem{S04}
455: T. St\"oferle \emph{et al.}, Phys.~Rev.~Lett. {\bf 92}, 13403 (2004)
456: %
457: \bibitem{M03}
458: O. Mandel \emph{el al.}, Phys.~Rev.~Lett.~{\bf 91}, 010407 (2003);
459: O. Mandel \emph{et al.}, Nature (London) {\bf 425}, 937 (2003).
460: %
461: \bibitem{K04}
462: M. K\"ohl \emph{et al.}, cond-mat/0410369 (2004).
463: %
464: \bibitem{Reichel} S. Aubin \emph{et al.}, cond-mat/0502196 (2005).
465: %
466: \bibitem{HCG04}
467: A.~F. Ho, M.~A. Cazalilla, and T. Giamarchi, Phys.~Rev.~Lett.~{\bf 92}, 130405 (2004).
468: %
469: \bibitem{DG04}
470: A. Kuklov, N. Prokof'ev, and B. Svistunov, Phys.~Rev.~Lett.~ {\bf 90},
471: 100401 (2003); D.~M. Gangardt \emph{et al.}, cond-mat/0408437 (2004).
472: %
473: \bibitem{CH03}
474: M.~A. Cazalilla and A.~F. Ho, Phys.~Rev.~Lett. {\bf 91} 150403 (2003);
475: M. Lewenstein \emph{et al.}, \emph{ibid} {\bf 92}, 050401 (2004);
476: L. Mathey \emph{et al.}, \emph{ibid} {\bf 93}, 120404 (2004).
477: %
478: \bibitem{GRJ03}
479: M.~Geiner, C.~A. Regal, and D.~S. Jin, Nature (London) {\bf 426}, 537 (2003).
480: %
481: \bibitem{FS04}
482: G.~M. Falco, H.~T.~C. Stoof, Phys.~Rev.~Lett.~{\bf 92}, 130401 (2004).
483: %
484: \bibitem{FRZ04}
485: J.~N. Fuchs, A. Recatti, and W. Zwerger, Phys.~Rev.~Lett. {\bf 93} 090408 (2004).
486: %
487: \bibitem{Sa04}
488: C.~A. Regal \emph{et al}., Phys.~Rev.~Lett.~{\bf 90}, 053201 (2003).
489: J.~Zhang \emph{et al.}, Phys.~Rev.~A {\bf 70}, 030702 (2004).
490: %
491: \bibitem{LH00}
492: A.~ Nersesyan, A. Luther, and F. Kusmartsev, Phys.~Lett.~B {176}, 363 (1993);
493: U. Ledermann and K. Le Hur, Phys.~Rev.~B {\bf 61}, 2497 (2000).
494: %
495: \bibitem{PS93}
496: K. Penc and J. Solyom, Phys.~Rev.~{47}, 6273 (1993).
497: %
498: \bibitem{unpub}
499: M.~A. Cazalilla, A.~F. Ho, and T. Giamarchi, in preparation.
500: %
501: \bibitem{T77}
502: M. Takahasi, J. Phys. C {\bf 01}, 1289 (1977).
503: %
504: \bibitem{L92}
505: P. Lemberger, J.~Phys. A: Math.~Gen.~ {\bf 25}, 715 (1992).
506: %
507: \bibitem{FLX}
508: J.~F. Freericks, E.~H. Lieb, and D. Ueltschi, Phys.~Rev.~Lett. {\bf 88}, 106401
509: (2002).
510: %
511: \bibitem{G04}
512: T. Giamarchi, \emph{Quantum Physics in One dimension}, Clarendon
513: Press (Oxford, UK, 2004); A.O. Gogolin, A.A. Nersesyan, A.M. Tsvelik,
514: \emph{ Bosonization and Strongly Correlated Systems}, Cambridge
515: University Press (Cambridge, UK, 1999); J. Voit,
516: Rep.~Prog.~Phys.~{\bf 57}, 997 (1994).
517: %
518: \bibitem{FDL95}
519: G.~F\'ath, Z.~Doma\'nski, R.~Lema\'nski, Phys.~Rev. {\bf B} 52, 13910 (1995), C.~D.~Batista, Phys.~Rev.~Lett. {\bf 89}, 166403 (2002).
520: %
521: \bibitem{L04}
522: C. Lee, Phys.~Rev.~Lett. {\bf 93}, 120406 (2004).
523: %
524: \bibitem{chem_rev}
525: T. Giamarchi, Chem.~Rev.~ {\bf 104}, 5037 (2004).
526: %
527: \bibitem{H80}
528: F.~D.~M. Haldane, Phys.~Rev.~Lett.~{\bf 45}, 1358 (1980).
529: %
530: \bibitem{W04}
531: S. Wessel \emph{et al.}, Phys. Rev. A {\bf 70}, 053615 (2004).
532: %
533: \bibitem{TZ99}
534: P. Torma and P. Zoller, Phys.~Rev.~Lett.~{\bf 85}, 487 (1999).
535: %
536: \bibitem{BZZ04}
537: H.~P. Buchler, W. Zwerger, and P. Zoller, Phys.~Rev.~Lett.~{\bf 93} 080401 (2004).
538: %
539: \bibitem{LZ01}
540: S. Lukyanov and A. Zamolodchikov, Nucl. Phys. B {\bf 607}, 437 (2001).
541: %
542: \end{thebibliography}
543: %
544: \end{document}
545: