1: \documentclass[aps,twocolumn]{revtex4}
2: \usepackage{graphicx}
3:
4: \begin{document}
5:
6:
7: \title{Ultracold fermions and the SU($N$) Hubbard model}
8: \author{Carsten Honerkamp$^{1}$ and Walter Hofstetter$^{1,2}$ }
9: \affiliation{$^{1}$Department of Physics, Massachusetts Institute of
10: Technology, Cambridge MA 02139, USA \\
11: $^2$Lyman Laboratory, Harvard University, Cambridge, MA 02138, USA}
12: \date{September 16, 2003}
13: \begin{abstract}
14: We investigate the fermionic SU($N$) Hubbard model
15: on the two-dimensional square lattice
16: for weak to moderate interaction strengths using one-loop
17: renormalization group and mean-field methods.
18: For the repulsive case $U>0$ at half filling and small $N$ the
19: dominant tendency is towards breaking of the SU($N$) symmetry.
20: For $N>6$ staggered flux order takes over as the dominant
21: instability,
22: in agreement with the large-$N$ limit.
23: Away from half filling for $N=3$
24: the system rearranges the particle densities such that two flavors
25: remain half filled by cannibalizing the third flavor. In the
26: attractive case and odd $N$ a full Fermi surface
27: coexists with a superconductor in the ground state.
28: These results may be relevant to
29: future experiments with cold fermionic atoms in optical lattices.
30: \end{abstract}
31:
32: \pacs{}
33: \maketitle
34: \vskip1pc
35:
36: \paragraph*{Introduction:}
37: After the celebrated observation of Bose-Einstein
38: condensation \cite{bec} ultracold atom systems receive growing
39: attention in the field of condensed matter physics.
40: Recently, also quantum degenerate Fermi gases have been realized
41: \cite{deMarco,truscott,fer3,fer4}, opening up the possibility to
42: study phenomena such as BCS superfluidity in a new context.
43: As a further important advance, optical lattices have been used
44: to realize the transition between a bosonic superfluid and a Mott
45: insulator \cite{greiner}. It has thus been demonstrated that cold atoms
46: systems can become a very flexible and clean laboratory for many
47: exciting phenomena from the purview of condensed matter or interacting
48: many particle systems. In particular, it has been suggested
49: \cite{hofstetter}
50: that cold fermions in optical lattices may help to
51: understand the notorious complexities of strongly correlated solid
52: state
53: systems such as the cuprate high--temperature superconductors.
54:
55: Besides the realization of phenomena that are known to exist in some
56: form in condensed matter systems, it is also interesting to ask
57: whether
58: the degrees of freedom offered by cold atoms could give rise to
59: states
60: of matter that do not have obvious counterparts in the physics of
61: interacting electrons. Typical electron systems, at least in the
62: first approximation, possess SU(2) spin rotational symmetry which can
63: be broken spontaneously, leading to magnetic phenomena such as ferro--
64: and anti--ferromagnetism. For alkali atoms, the nuclear spin $I$ and
65: electron spin $S$ are combined in a hyperfine state.
66: Its total angular momentum $F$ can be different from $1/2$,
67: and for each $F$ there are $2F+1$ hyperfine states differing by
68: their azimuthal quantum number $m_F$. E.g. for the fermionic
69: $^{40}$K,
70: the nuclear
71: spin is $I=4$ and the lowest hyperfine multiplet (at weak fields)
72: has $F=9/2$. In magnetic traps only a subset of these $2F+1$ states
73: (the \emph{low--field--seekers}) can be trapped \cite{deMarco},
74: but this constraint can be avoided by using all--optical traps
75: \cite{granade}.
76:
77: In fact, coexistence of the three hyperfine states
78: $|F=9/2, m_{F}=-5/2,-7/2,-9/2\rangle$ of $^{40}K$ in an optical trap
79: has already been realized,
80: with tunable interactions due to Feshbach resonances between
81: $m_{F}=-5/2/-9/2$ and $m_{F}=-7/2/-9/2$, respectively \cite{regal}.
82: A situation with strong attractive interaction between all three
83: components can be realized e.g. for the spin polarized states with
84: $m_{s}=1/2$
85: in $^{6}$Li where the triplet scattering length $a=-2160a_{0}$
86: is anomalously large \cite{abraham}.
87:
88: Optical lattices are created by a standing light wave leading
89: to a periodic potential for the atomic motion of the form
90: $V(x)=V_{0} \sum_{i} \cos^{2}(k x_{i})$
91: where $k$ is the wavevector of the laser, $i$ labels the spatial
92: coordinates and the lattice depth $V_{0}$ is usually measured in units of the
93: atomic recoil energy $E_{R}=\hbar^{2} k^{2}/2m$.
94: In the following we will consider the 2D case where
95: $i=1,2$. It has been shown \cite{boseth} that
96: the \emph{Hubbard model} with a local density--density interaction
97: provides an excellent description of the low--energy physics.
98: Here we are interested in a situation where fermionic atoms with
99: $N$ different spin states (``flavors'') $m$ are loaded into the
100: optical lattice.
101: We thus consider a Hubbard Hamiltonian
102: \begin{equation}
103: H = -t \sum_{m,\langle ij \rangle}
104: \left[c^\dagger_{i,m} c_{j,m} + c^\dagger_{j,m} c_{i,m} \right] + \frac{U}{2} \sum_i n_i^2 \, . \label{hubb}
105: \end{equation}
106: Here $n_i=\sum_{m} n_{i,m}$ is the total number
107: density of atoms on site $i$ which can be written in terms of
108: creation and annihilation operators according to
109: $n_{i,m} = c^\dagger_{i,m} c_{i,m}$.
110: The interaction (second term in Eq. \ref{hubb})
111: is invariant under local U($N$) rotations of the $N$
112: flavors with different $m$. The hopping term of the atoms between
113: nearest neighbors $\langle ij \rangle$
114: reduces the invariance of the complete
115: Hamiltonian to a global U($N$) symmetry. Stripping off
116: the overall U(1) phase factor, we arrive at the SU($N$) Hubbard model.
117: Note that in the optical lattice the effective Hubbard parameters are given by
118: $t=E_{R} \left(2/\sqrt{\pi}\right) \xi^{3} \exp\left(-2 \xi^{2}\right)$
119: and $U=E_{R} a_{s} k \sqrt{8/\pi} \xi^{3}$ where
120: $\xi=\left(V_{0}/E_{R}\right)^{1/4}$ and $a_{s}$ is the $s$--wave
121: atomic scattering length.
122:
123: The fermionic SU($N$) Hubbard model for
124: $U>0$ on the two-dimensional (2D) square lattice was studied
125: in the large-$N$ limit \cite{marston} in the early days of
126: high--$T_c$
127: superconductivity, mainly as a controllable limit connected to the
128: then
129: physically relevant case $N=2$. A generalized SU($N$) model could
130: describe
131: orbitally degenerate electronic states in crystals, but it is likely
132: that in these systems different overlaps between the orbitals
133: pointing in distinct lattice directions break the SU($N$)
134: invariance.
135:
136: In the following we focus on the density region near half band
137: filling with an average $N/2$ fermions per site. In the conventional
138: $N=2$ Hubbard model at half filling the ground state exhibits
139: spin-density wave (SDW) order, and when the filling is changed
140: $d$-wave superconductivity is very likely \cite{zanchi}. The SDW
141: state breaks the translational invariance and spin-up and spin-down
142: electrons (for staggered moment along the $z$-direction)
143: occupy the two sublattices differently (see Fig. \ref{orders}).
144: For large $N$ and small exchange interactions $J$, staggered flux order
145: is expected to dominate over the SU($N$)-breaking
146: states \cite{marston}.
147:
148: \paragraph*{One-loop renormalization group for the half-filled band
149: and general $N$ and $U>0$:}
150: First let us analyze the one-loop renormalization group (RG)
151: flow for the half filled band.
152: We apply the perturbative temperature-flow RG
153: method of Ref. \cite{tflow} that has proved to give good results for
154: $N=2$.
155: As initial condition one fixes the interaction at a high temperature
156: of the order of the bandwidth. Then the RG flow describes the change
157: of the interactions as the temperature is lowered and perturbative
158: corrections due to one-loop particle-hole and particle-particle
159: processes are taken into account.
160: The interaction is described by a coupling function $V(\vec{k}_1,\vec{k}_2,\vec{k}_3)$ \cite{salmhofer}, where the flavor indices $m_1$ and $m_3$ belonging to the first incoming particle with wavevector $\vec{k}_1$ and the first outgoing particle with $\vec{k}_3$ are the same.
161: Similarly $m_2=m_4$. The second outgoing wavevector $\vec{k}_4$ is fixed by momentum conservation on the lattice.
162: As in the $N=2$ case the RG flow goes to strong coupling. This
163: means, as we start the flow at high temperatures with a purely local
164: interaction $V(\vec{k}_1,\vec{k}_2,\vec{k}_3)=U$, some coupling
165: functions start to grow when the temperature is reduced and finally
166: leave the perturbative range.
167: At this temperature scale we stop the RG flow and analyze which class
168: of coupling constants grows most strongly towards low $T$. In close
169: analogy with the spin-1/2 case we consider couplings in the
170: charge channel
171: $ V_c(\vec{k},\vec{k}',\vec{q})= N
172: V(\vec{k}+\vec{q},\vec{k}',\vec{k}) - V(\vec{k}',
173: \vec{k}+\vec{q},\vec{k}) $
174: and in the SU($N$) symmetry breaking channel,
175: $ V_s(\vec{k},\vec{k}',\vec{q})= - V(\vec{k}',
176: \vec{k}+\vec{q},\vec{k})$,
177: which, if divergent, signals a singular response for a small external
178: field coupling to one of the $N^2-1$ generators of SU($N$).
179: We define averages over the Fermi surface,
180: $%\[
181: \bar{V}^{\ell}_{c/s}(\vec{Q}) = \oint_{FS} d\phi_k
182: \oint_{FS'} d\phi_{k'} \, g_\ell(k) g_\ell(k') V_{c/s}
183: (\vec{k},\vec{k}',\vec{q}) \,
184: $%\]
185: with $g_1(\vec{k})=1$ in the $s$-wave channel or $g_2(\vec{k})=
186: (\cos k_x - \cos k_y) /\sqrt{2}$ in the $d$-wave channel.
187: For $N=2$ the couplings $\bar{V}_s(\vec{Q})$ in the $s$-wave channel
188: with
189: $\vec{Q}=(\pi,\pi)$ diverge most strongly,
190: indicating the formation of an antiferromagnetic (AF) SDW state.
191: Note that for larger $N$ the charge couplings can become quite large
192: as they scale with $N$.
193:
194: Analyzing the RG flows to strong coupling, the picture is as follows:
195: for $N \le 6$ the $s$-wave coupling function $\bar{V}^s_s(\vec{Q})$
196: is the strongest divergent family of coupling constants, signalling a
197: dominant tendency towards breaking of the SU($N$) symmetry with
198: staggered two-sublattice real space dependence.
199: For $N>2$ (especially for odd $N$) this leads to the interesting
200: question how the $N/2$ particles per site will arrange themselves
201: on the bipartite square lattice (see Fig. \ref{orders}).
202: Below we describe what happens in a mean-field analysis.
203:
204:
205: \begin{figure}
206: \includegraphics[width=.44\textwidth]{orders.eps}
207: \caption{a): AF spin-density wave state for $N=2$. Spin-up and
208: spin-down particles occupy the two sublattices with different
209: probabilities (here idealized to 0 and 1). b) SU(3)-breaking flavor-density wave state
210: for $N=3$. Flavors 1 and 2 prefer one sublattice, flavor 3 the other.
211: c) Staggered flux state for $N> 6$: the arrows indicate the particle
212: currents.}
213: \label{orders}
214: \end{figure}
215:
216: \begin{figure}
217: \includegraphics[width=.49\textwidth]{s38plot.eps}
218:
219: \caption{
220: Left plots: RG results for SU(3) at half filling and $U=4t$. Upper left:
221: Flow of the coupling constants $\bar{V}_s^s$ in the SU(3)-breaking
222: channel (dashed line) and $\bar{V}_{c}^{d}$ in the staggered flux
223: (SF) channel, averaged around the Fermi surface (FS).
224: Lower left:
225: $V_s(\vec{k},\vec{k}',\vec{Q})$ (dashed line) and
226: $V_c(\vec{k},\vec{k}',\vec{Q})$ with $\vec{k}$ fixed on the FS at
227: $(\pi,0)$ and $\vec{k}'$ moving around the FS after termination of the flow.
228: Right plots: The same for SU(8) at half filling. }
229: \label{s38plot}
230: \end{figure}
231:
232: For $N>6$ the flow to strong coupling changes qualitatively. Now the
233: leading divergence is in the charge couplings $V_c
234: (\vec{k},\vec{k}',\vec{Q})$ with a $d_{x^2-y^2}$-wave dependence
235: on $\vec{k}$ and $\vec{k}'$. $\bar{V}_c^d$ diverges more strongly
236: than $\bar{V}_s^s$ (see Fig. \ref{s38plot}), albeit at lower
237: temperature $T \approx 0.014t$ for $U=4t$ and $N=7$.
238: This signals a tendency towards
239: staggered flux (SF) order with long range ordering of the expectation
240: value $\Phi_{SF} = \sum_{\vec{k},m} (\cos k_x -\cos k_y ) \langle
241: c_{\vec{k},m}^\dagger c_{\vec{k}+\vec{Q},m} \rangle$.
242: This result agrees with the large--$N$ limit
243: for small exchange interactions $J$ \cite{marston}.
244: The SF state has surfaced several times for the SU(2) case in
245: connection with the high-$T_c$ cuprates and related
246: models \cite{marston,sfrefs}, also as $d$-density wave state
247: (although the particle density is {\em not} modulated).
248: Its quasiparticles have a
249: wavevector-dependent energy gap that vanishes at $\vec{k}=(\pm
250: \pi/2,\pm \pi/2)$.
251: Nonzero $\Phi_{SF}$ breaks translational and time-reversal symmetry with
252: alternating particle currents around the plaquettes (see Fig. \ref{orders}).
253: If the particles were charged, their motion would give rise to alternating
254: magnetic moments pointing out of the plane, hence the name staggered
255: flux state.
256: Note that $\Phi_{SF}$ is SU($N$)-invariant and
257: no continuous symmetries are broken.
258: Correspondingly the SF state can order at finite temperatures in 2D.
259: For the same reason it may be possible that the staggered flux state
260: sets in for somewhat lower $N$ than the critical $N=6$ in our
261: one-loop RG study that neglects collective fluctuations.
262:
263: Away from half filling the flow of the dominant $(\pi,\pi)$-
264: instability gets cut off at some low energy scale that increases
265: with the distance to half filling.
266: Below that scale there is a tendency towards
267: $d_{x^2-y^2}$-wave Cooper pairing. However the energy scales for
268: pairing instabilities become very small with increasing $N$.
269: Below we discuss superfluid pairing for $N>2$ in the attractive
270: case $U<0$.
271:
272: \paragraph*{Ground state near half filling for $N=3$:}
273: Having established that SU($N$) symmetry breaking at wave vector
274: $(\pi,\pi)$ is the dominant instability of the repulsive model near half
275: filling with $N<6$, we now turn to a mean-field description of the
276: ground state for $N=3$.
277: We decouple the interaction terms in the particle-hole channel with
278: local mean-fields $\langle c_{\alpha,i}^\dagger c_{\beta,i}\rangle =
279: M_{\alpha \beta, i}$.
280: The hermitian local mean-field matrix $M_{\alpha\beta}$
281: can be decomposed into a traceful part $M_0 $ proportional to the
282: identity matrix $t_0$ and a traceless part $\sum_{a=1,\dots 8} M^a
283: t_a$ with the 8 generators $t_a$ of the fundamental representation of
284: SU(3). A finite value of one of the traceless components breaks the
285: SU(3) invariance. We will now restrict the analysis to commensurate
286: order, where only uniform and staggered components of a
287: commuting subset of the 9 $M^a$ acquire nonzero expectation values.
288: SU(3) has rank 2 and the two diagonal generators commute mutually and
289: with the identity matrix. We can choose these three degrees of
290: freedom to be contained in the three flavor density mean-fields
291: $\langle n_{\alpha} \rangle$.
292:
293: The results of $T=0$ mean-field solutions
294: are shown in Fig. \ref{kplot}. At half filling, $n=1.5$/site, the
295: SU(3)-breaking creates a flavor-density wave:
296: two flavors prefer one sublattice with equal density,
297: while the third flavor goes predominantly on the other sublattice with
298: a somewhat larger density modulation.
299: The staggered components
300: do not add up to zero. Thus there is a charge density wave
301: accompanying the SU(3) symmetry breaking.
302: For $U=3t$ the mean-field $T_c$ for this state is $\sim 0.45t$, but in the one-loop RG is it is reduced down to $\sim 0.12t$.
303:
304: Fig. \ref{kplot} also describes the results away from half filling.
305: For example at $U=1.6t$ and $1.42<n<1.48$/site,
306: two flavors order with opposite staggered densities on the two
307: sublattices, keeping their individual average density at half filling.
308: Since the total density is less than half filling,
309: the third flavor gets decimated with uniform density of $(n-1)$.
310: As can be seen from the right plot in Fig. \ref{kplot}, this state
311: only occurs above a critical
312: interaction strength $U_c$ that increases from zero with increasing
313: distance to $n=1.5$/site.
314: Note that the depletion of one flavor allows the system to preserve
315: the commensurate order away from commensurate band filling.
316: A similar pinning of a part of the system
317: to half filling is found in ladder systems \cite{rice}.
318: We add that for larger $U$ and close to half filling
319: ($1.46< n <1.5$/site for $U=4t$) we find a another regime where the
320: mean field equations converge slowly and microscopic phase separation
321: might occur.
322:
323: \begin{figure}
324: \includegraphics[width=.49\textwidth]{kplot.eps}
325:
326: \caption{Left: Uniform densities of the $N=3$ flavors (solid
327: line: flavor 1 and 2, crosses: flavor 3) and staggered densities (squares) vs. total
328: density for $U=1.6t$.
329: Right: Difference in uniform density between the two
330: majority flavors and the minority flavor versus interaction $U$ and total density $\langle n \rangle$ per site. The scalebar
331: indicates the density difference per site.}
332: \label{kplot}
333: \end{figure}
334:
335:
336: \paragraph*{Attractive SU($N$) Hubbard model:}
337: Next let us consider the attractive interactions $U<0$.
338: In the SU(2) case in 2D, there is a power-law
339: $s$-wave singlet superconductor/superfluid (SSC) below a
340: Kosterlitz-Thouless
341: transition away from half filling \cite{scalettar}. At half filling
342: the Kosterlitz-Thouless $T_c$ is zero, and SSC and
343: charge-density wave (CDW) mean field states are degenerate and in the
344: true ground state, both types of order coexist.
345: One-loop RG finds in this case that CDW and SSC susceptibilities are
346: perfectly degenerate and diverge together at low $T$. This symmetry
347: is destroyed for larger $N>2$ and the CDW susceptibility grows much
348: faster than the one for SSC. Correspondingly we expect the ground
349: state to have CDW long range order only. This is corroborated by a
350: mean-field calculation for the SU(3) case that shows that the CDW
351: order suppresses any kind of SSC admixture, and the CDW ground state
352: energy is lower than that of the SSC state.
353:
354: We now consider the generic case sufficiently far away from half
355: filling. Then the dominant instability is onsite
356: pairing. We decouple the interaction as
357: $ H_{U,\mathrm{mf}}= -\frac{1}{2} \sum_{\vec{k}, \alpha ,\beta}
358: c^\dagger_{\vec{k}\alpha} c^\dagger_{-\vec{k}\beta} \Delta_{\beta
359: \alpha} + h.c. $
360: with the local mean-fields $\Delta_{\alpha \beta} = -U \sum_{\vec{k}}
361: \langle
362: c_{\vec{k}\alpha} c_{-\vec{k}\beta} \rangle = - \Delta_{\beta
363: \alpha} $.
364: For $N>2$ these even parity gap functions
365: $\Delta_{\alpha \beta}$ transform non-trivially under SU($N$).
366: Depending on the global gauge,
367: $\Delta_{\alpha \beta} $ takes different values. This is unlike
368: the SU(2) case where even parity gap functions are singlets and
369: invariant under spin rotations \cite{fn1}.
370: For SU(2) the ground state is degenerate with respect to the
371: global phase of the gap function, and long-wavelength variations of
372: the latter are gapless in absence of long-range forces. In the
373: SU($N$) case we find a higher degeneracy and more gapless modes. It turns
374: out that for SU(3) all gap functions with
375: with the same $\Delta_0^2= \sum_{\alpha \beta} |\Delta_{\alpha
376: \beta}|^2$ are degenerate and have the same total density of states.
377: Apart from the global phase there are four additional gapless
378: modes,
379: two associated with the internal phases between $\Delta_{12}$,
380: $\Delta_{13}$ and $\Delta_{23}$, and two modes modulating
381: $|\Delta_{12}|$,
382: $|\Delta_{13}|$ and $|\Delta_{23}|$ with fixed $\Delta_0$.
383:
384: A particularly simple choice in the degenerate manifold is
385: $\Delta_{12}= \Delta_0$ and $\Delta_{13}=\Delta_{23}=0$. Then flavor
386: 3 remains completely unpaired and metallic. Since we can always
387: rotate into this gauge, all SU(3) $s$-wave superconducting mean-field
388: states are one-third (neutral) metals and two-thirds superfluids.
389: The gauge with only $\Delta_{12}\not= 0$ makes the symmetry breaking
390: pattern obvious. The original symmetry group of the problem SU(3)
391: $\otimes$ U(1) with nine generators gets broken down to an SU(2) in
392: flavor 1 and 2, leaving $\Delta_{12}$ invariant, and an additional
393: U(1) that acts on the phase of the unpaired flavor 3. This leaves 5
394: generators broken, yielding the collective modes described above.
395: For $3/8$ band filling and $U=4t$, the mean-field $T_c$ is $\sim 0.17t$.
396: The coexistence of a full Fermi surface with a superconductor should
397: have interesting consequences. For example the collective modes may
398: be subject to damping below twice the gap frequency and
399: could hence render the ungapped fermionic spectrum observable.
400: Experimentally, these Goldstone modes can be detected in the
401: spectrum of elementary excitations measured via Bragg scattering
402: off two non--collinear laser beams with frequency and momentum
403: difference $\omega$,$q$ \cite{stenger}.
404: By monitoring the number of scattered atoms as a function
405: of $\omega$, this technique yields the dynamical
406: structure factor $S(q,\omega)$.
407: The collective modes will then lead to peaks in the scattering cross section.
408:
409: Theoretically, an additional weak $p$-wave attraction could trigger
410: a superfluid transition of the unpaired flavor at much lower
411: temperatures, leading to a coexistence of even- and odd-parity
412: superfluidity.
413:
414: The SU(4) case is more complicated. There the degeneracy
415: of the ground state is subject to more constraints than just constant
416: $\Delta_0$. The mean-field solutions have
417: $|\Delta_{12}|=|\Delta_{34}|$,
418: $|\Delta_{13}|=|\Delta_{24}|$ and $|\Delta_{14}|=|\Delta_{23}|$.
419: The single particle spectrum is fully gapped.
420:
421:
422:
423: \paragraph*{Conclusions:}
424: The fermionic SU($N$) Hubbard model on the 2D square lattice can
425: possibly be realized with ultracold atoms in an optical lattice. Its
426: ground states may exhibit phenomena that do not occur right away
427: in traditional solid state systems. For $U>0$ we find a staggered
428: flux state for large $N>6$ at half band filling where the particles
429: run around the plaquettes of the lattice in an alternating
430: way. This state has a partially gapped excitation spectrum with nodes
431: along the Brillouin zone diagonals, which may be detectable via
432: the momentum distribution function.
433: Near half filling for
434: $N=3$ we find a redistribution of the particle densities where two of
435: the three flavors remain half filled and occupy different sublattices
436: while the third flavor becomes depleted. Finally, in the attractive
437: case $U<0$ we point out that the $s$-wave paired superfluid states
438: may exhibit new collective modes.
439: For $N=3$ a third of the particles remains
440: ungapped, leading to a full Fermi surface coexisting with the
441: superfluid. We expect this to be a general feature for odd $N$, also
442: in three dimensions or in absence of a lattice potential.
443:
444: Finally, a comment on the temperature scales
445: which we have given in terms of the hopping parameter $t$.
446: It has been shown \cite{hofstetter} that if the optical lattice
447: is switched on slowly after termination of evaporative cooling,
448: an additional \emph{adiabatic cooling} process takes place.
449: The final temperature is given by the identity
450: $T_{\rm initial}/T_{ F,\rm free} \approx T_{\rm final}/T_{ F,\rm lattice}$
451: where $T_{ F,\rm free (lattice)}$ denote the
452: Fermi temperature of the free atomic cloud and in the presence
453: of the lattice, respectively. In particular, in 2D at half filling
454: one has $T_{ F,\rm lattice} = 4 t$. As a result, the critical atomic
455: temperatures which have to be reached \emph{before}
456: the lattice is switched on can be obtained from our results via the substitution
457: $t \to T_{F, \rm free}/4$. For the $s$-wave superfluid phase ($U<0$)
458: and the flavor-density wave states ($U>0$)
459: we therefore find transition temperatures of order
460: $0.05 T_{F}$ which are within experimental reach.
461:
462: We thank P.A. Lee, W.V. Liu, T.M. Rice and C.H. Schunck for useful
463: discussions. E. Demler and M.D. Lukin are acknowledged for suggesting the topic
464: to one of us. C.H. and W.H. were supported by the German Research Foundation (DFG) and
465: W.H. by a Pappalardo fellowship.
466:
467: \begin{thebibliography}{99}
468: \bibitem{bec}For a review see special issue of Nature, {\bf 416}, 206
469: (2002).
470: \bibitem{deMarco}B. DeMarco and D.S. Jin, Science {\bf 285}, 1703
471: (1999).
472: \bibitem{truscott} A.G. Truscott {\em et al.}, Science {\bf 291},
473: 2570 (2001).
474: \bibitem{fer3} F. Schreck {\em et al.}, Phys. Rev. Lett. {\bf 87},
475: 80403 (2001).
476: \bibitem{fer4} Z. Hadzibabic {\em et al.}
477: Phys. Rev. Lett. {\bf 88}, 160401 (2002).
478: \bibitem{greiner} M. Greiner \emph{et al.}, Nature (London) {\bf
479: 415}, 39
480: (2002).
481: \bibitem{hofstetter} W. Hofstetter \emph{et al.}, Phys. Rev. Lett.
482: {\bf 89}, 220407 (2002).
483: \bibitem{granade} S.R. Granade \emph{et al.}, Phys. Rev. Lett. {\bf
484: 88},
485: 120405 (2002).
486: \bibitem{regal} C.A. Regal and D.S. Jin, cond-mat/0302246.
487: \bibitem{abraham}E.R.I. Abraham \emph{et al.}, Phys. Rev. A {\bf 55},
488: R3299 (1997).
489: \bibitem{boseth} D. Jaksch {\em et al.}, Phys. Rev. Lett. {\bf 81},
490: 3108 (1998).
491: \bibitem{marston} I. Affleck and J.B. Marston, Phys. Rev. B {\bf 37},
492: 11538 (1988); J.B. Marston and I. Affleck, Phys. Rev. B {\bf 39},
493: 11538 (1989).
494:
495: \bibitem{zanchi} D. Zanchi and H.J. Schulz, Europhys.Lett. {\bf 44},
496: 235 (1997); Phys.Rev. B {\bf 61}, 13609 (2000); C.J. Halboth and W.
497: Metzner, Phys.Rev. B {\bf 61}, 7364
498: (2000); Phys. Rev. Lett. {\bf 85}, 5162 (2000).
499: \bibitem{tflow} C. Honerkamp and M.
500: Salmhofer, Phys. Rev. B {\bf 64} 184516 (2001); Phys. Rev. Lett.
501: {\bf 87},
502: 187004 (2001).
503: \bibitem{salmhofer} M. Salmhofer and C. Honerkamp, Prog. Theor. Phys.
504: {\bf 105}, 1 (2001).
505: \bibitem{sfrefs} T. Hsu, J.B. Marston, and I. Affleck, Phys. Rev. B
506: {\bf 43}, 2866 (1991); E. Cappelluti and R. Zeyher,
507: Phys. Rev. B {\bf 59}, 6475 (1999); S. Chakravarty {\em et al.},
508: Phys. Rev. B {\bf 63}, 094503 (2001); U. Schollw\"ock {\em et al.}, Phys.
509: Rev. Lett. {\bf 90}, 186401 (2003).
510: \bibitem{rice} T. M. Rice {\em et al.}, Phys. Rev. B {\bf 56}, 14655
511: (1997).
512: \bibitem{scalettar} R.T. Scalettar {\em et al.}, Phys. Rev. Lett. {\bf 62},
513: 1407 (1989); R. Micnas, J. Ranninger, and S. Robaszkiewicz, Rev. Mod.
514: Phys. {\bf 62}, 113 (1990).
515: \bibitem{fn1} Note that for SU(2) the product of two fundamental
516: representations $2 \otimes 2= 1 \oplus 3$ contains a singlet, while
517: for SU(3) $3 \otimes 3 = \bar{3} \oplus 6$. Thus the even parity gaps span
518: a three-dimensional subspace. SU(3)-singlets can only be formed in
519: the particle-hole channel (cp. pions).
520: \bibitem{stenger}J. Stenger {\em et al.}, Phys. Rev. Lett. {\bf 82},
521: 4569 (1999).
522:
523: \end{thebibliography}
524:
525: \end{document}
526:
527:
528:
529:
530: