1: \documentclass[showpacs,twocolumn,prl,floatfix,amssymb,aps]{revtex4}
2: %\documentclass[preprint]{revtex4}
3: \usepackage{graphicx}
4:
5: \begin{document}
6:
7: \hsize\textwidth\columnwidth\hsize\csname@twocolumnfalse\endcsname
8:
9: \title{Low-energy excitations of the one-dimensional half-filled $SU(4)$ Hubbard model
10: with an attractive on-site interaction: Density-matrix renormalization-group calculations and
11: perturbation theory}
12: \author{Jize Zhao and Kazuo Ueda}
13: \affiliation{Institute for Solid State Physics, University of
14: Tokyo, Kashiwa, Chiba 277-8581, Japan}
15: \author{Xiaoqun Wang}
16: \affiliation{Department of Physics, Renmin University of China,
17: Beijing 100872, China}
18: \date{\today}
19: \begin{abstract}
20: We investigate low-energy excitations of the one-dimensional
21: half-filled $SU(4)$ Hubbard model with an attractive on-site
22: interaction $U<0$ using the density matrix renormalization group
23: method as well as a perturbation theory. We find that the ground
24: state is a charge density wave state with a long range order. The
25: ground state is completely incompressible since all the
26: excitations are gapful. The charge gap which is the same as the
27: four-particle excitation gap is a non-monotonic function of $U$,
28: while the spin gap and others increase with increasing $|U|$ and
29: have linear asymptotic behaviors.
30: \end{abstract}
31:
32: \pacs{71.10.Fd, 71.10.Pm, 71.30.+h, 05.10.Cc}
33: \maketitle
34:
35: The Hubbard model is one of the most classic models for strongly
36: correlated electronic systems and has attracted long-term interest
37: since the pioneering work in 1960s\cite{HUBB1}. By taking the
38: on-site Coulomb interaction into account, it well explains the
39: puzzle that some materials with half-filled band are insulators.
40: However, this model which consider only a single band, on-site
41: Coulomb interaction and the nearest-neighbor hopping is often
42: thought being oversimplified. To account for other features beyond
43: the Mott physics\cite{MOTT}, there are two kinds of natural
44: extensions: one is to incorporate the orbital degree of freedom
45: which may be called multi-band Hubbard model\cite{IMADA1}, and the
46: other is to consider the hopping and/or the Coulomb interaction
47: with longer ranges.
48:
49: In the last several years, ultra-cold atomic experiments evoke
50: systematic studies of correlation effects in the optical lattice
51: systems where interactions are tunable through Feshbach resonance.
52: The Hubbard model becomes again an appropriate one to envisage
53: some relevant issues with both positive and negative interactions.
54: Recently, fermionic atoms with higher spins are successfully
55: trapped into optical lattices \cite{MODUG1}. This calls for a
56: generalization of the $SU(2)$ Hubbard model into the $SU(N)$
57: case\cite{AFFLECK1,WU1}. In this paper, we study one-dimensional
58: $SU(4)$ Hubbard model which is represented as
59: \begin{equation}
60: \mathcal{H}=-t\sum_{i=1}^{L}\sum_{\sigma}(c^{\dagger}_{i\sigma}c_{i+1\sigma}+h.c.)+\\
61: \frac{U}{2}\sum_{i=1}^{L}\sum_{\sigma\ne\sigma^{'}}n_{i\sigma}n_{i\sigma^{'}}
62: \label{HAM}
63: \end{equation}
64: where $t>0$ is a hopping matrix element, $L$ the number of the
65: lattice sites, $\sigma$ and $\sigma^{'}$ the spin indices taking
66: $-\frac{3}{2}$, $-\frac{1}{2}$, $\frac{1}{2}$ and $\frac{3}{2}$.
67: $c^{\dagger}_{i\sigma}(c_{i\sigma})$ denotes
68: the creation(annihilation) operator of a particle with spin
69: $\sigma$ at the site $i$,
70: $n_{i\sigma}=c^{\dagger}_{i\sigma}c_{i\sigma}$ is the corresponding
71: number operator and $U$ is the on-site interaction.
72:
73: This model is not exactly solvable even in one dimension in
74: contrast to the $SU(2)$ case\cite{LIEB1,CHOY1}. Nevertheless,
75: some aspects of physical properties can be reliably explored by
76: some analytical approaches as well as numerical methods such as
77: density matrix renormalization group (DMRG)\cite{WHITE1, PESCH1}
78: and Quantum Monte Carlo simulations\cite{ASSARAF1, ASSARAF2}. Most recently,
79: an $SO(8)$ symmetry regime was proposed between $0<U<3t$ at
80: half-filling by Assaraf et al\cite{ASSARAF1} with using a
81: nonperturbative renormalization group method and quantum Monte
82: Carlo simulation. They found that the low-energy spectrums are
83: gapful in this regime. The similar results were also shown later by
84: Szirmai and S\'olyom for the other $N>2$ case\cite{SZIRMAI1}.
85: Those studies were concentrated on the repulsive case $U>0$,
86: but for the $U<0$ case few results are obtained so far. On the
87: other hand, it is well known that the one-dimensional attractive
88: half-filled $SU(2)$ Hubbard model is described by a Luther-Emery
89: liquid\cite{LUTHER1}, in which the charge excitation is gapless,
90: whereas the spin excitation is gapful. By the hidden $SU(2)$
91: transformation the $SU(2)$ Hubbard model with $U$ can be mapped to
92: the one with $-U$, while for the $SU(4)$ case such a mapping does not
93: exist so that one cannot obtain any insights into the
94: low-energy properties through the mapping. In this paper, we will
95: show that the $SU(4)$ Hubbard model at half-filling with the
96: attractive interaction belongs to a different universality class
97: from the $SU(2)$ one.
98:
99: Let us start with a perturbation theory for the strong coupling
100: regime. For this purpose, we rewrite the Hamiltonian (\ref{HAM})
101: as $\mathcal{H}=\mathcal{H}_{t}+\mathcal{H}_{u}$, where the
102: hopping term $\mathcal{H}_{t}=-t\sum_{i\sigma}
103: (c^{\dagger}_{i\sigma} c_{i+1\sigma}+h.c.)$ is regarded as a
104: perturbation and the on-site interaction
105: $\mathcal{H}_{u}=\frac{U}{2}\sum_{i\sigma\ne\sigma^{'}}
106: n_{i\sigma}n_{i\sigma^{'}}$ as the zeroth order Hamiltonian has
107: highly degenerate ground states in which each site is either fully
108: occupied by four particles forming a $SU(4)$ singlet, or empty. Up
109: to the second-order, the effective Hamiltonian is given by
110: \begin{equation}
111: \mathcal{H}^{(2)}_{eff}=\frac{2t^2}{3U}P\sum_{i}n_{i}P
112: -\frac{t^2}{6U}P\sum_{i}n_{i}n_{i+1}P,
113: \label{HEFF}
114: \end{equation}
115: where $P$ is a projection operator which projects a state onto the
116: subspace spanned by the ground states of $\mathcal{H}_{u}$, and
117: $n_{i}=\sum_{\sigma}n_{i\sigma}$ the number operator at the site
118: $i$. The hopping term $\mathcal{H}_{t}$
119: lifts the degeneracy of $\mathcal{H}_{u}$ and gives rise to both the
120: energy gain of $\frac{4t^{2}}{3U}$ per site at half-filling as denoted by the
121: first term of (\ref{HEFF}) and an effective repulsive interaction between
122: particles on the nearest-neighbor sites as denoted by the second
123: term of (\ref{HEFF}). The second term induces essentially a
124: charge-density-wave (CDW) ground state with a true long range
125: order such that every other site is fully occupied with an empty
126: site in between, which is consistent with the mean field result for the
127: weak coupling region\cite{WU1}. Moreover, ${\mathcal H}^{(2)}_{eff}$ has two-fold
128: degenerate ground states, each of which is a $SU(4)$ singlet.
129:
130: An important question concerning the CDW ground state is whether
131: it is metallic or insulating. To address this question, we need to
132: examine charge excitations. The charge gap $\Delta_{c}$ is the
133: energy difference of the lowest excitation in the spin singlet
134: channel from the ground state as defined by
135: $\Delta_{c}=E_1(L,2L,0)-E_{0}(L,2L,0)$, in which $E_{n}(L,N,S)$
136: stands for the the $n-$th excitation energy in a spin-$S$ channel
137: with $L$ sites and $N$ particles. And another interesting issue is
138: to explore the relevance of the four-particle excitation gap
139: $\Delta_4$ to $\Delta_c$ for the $SU(4)$ symmetry. $\Delta_{4}$
140: represents the energy cost of adding four particles or holes into
141: the systems such that $\Delta_{4}=\frac 1
142: 2\left[E_{0}(L,2L+4,0)\right.+E_{0}(L,2L-4,0)\left.-2E_0(L,2L,0)\right]$.
143: From ${\mathcal H}^{(2)}_{eff}$, one can easily find
144: $\Delta_{c}=-\frac{8t^{2}}{3U}$ based on the fact that the motion
145: of four particles from one of fully occupied sites to its neighbor
146: costs a minimal energy. Similarly, adding four particles to the
147: system would gain the energy $6U+2\times(-\frac{8t^2}{3U})$, while
148: adding four holes costs $-6U$, thus one has $\Delta_{4} =
149: -\frac{8t^{2}}{3U}$. It turns out that the four-particle
150: excitation gap is essentially the same as the charge gap and
151: finite.
152:
153: However, the situation is completely different for the $SU(2)$
154: case, where the ground state is metallic. To understand this, one
155: can write the effective Hamiltonian in the strong coupling limit
156: for the $SU(2)$ case as\cite{EMERY1}:
157: \begin{eqnarray}
158: \mathcal{H}^{(2)}_{eff,su(2)} & = &
159: \frac{2t^2}{U}P\sum_{i\sigma}n_{i\sigma}P \label{HEFFSU2}\\
160: & -& \frac{t^2}{U}P\sum_{\langle{ij}\rangle\sigma}
161: (n_{i\sigma}n_{j\sigma}-c^{+}_{i\sigma}c^{+}_{i\bar{\sigma}}
162: c_{j\bar{\sigma}}c_{j\sigma})P\nonumber.
163: \end{eqnarray}
164: This Hamiltonian distinguishes itself from ${\mathcal
165: H}^{(2)}_{eff}$ with its last additional term which allows a
166: ``pair hopping" process between the two neighbor sites. Although
167: the second repulsive term would be in favor of forming a CDW
168: ground state with gapful charge excitations, the extra ``pair
169: hopping" term eventually destabilizes this CDW long range order
170: resulting in gapless charge excitations. On the other hand for the $SU(4)$
171: case, similar hopping term can only appear after calculating to
172: higher order. Therefore, the
173: perturbation theory in the strong coupling regime shed light on
174: the different nature of the ground states between the $SU(2)$ and
175: $SU(4)$ cases.
176:
177: In order to go beyond the validity range of the above perturbation
178: theory for a full exploration of the low-energy properties, we
179: have performed systematic DMRG computations. There are actually
180: some difficulties inherent to this model, such as a large number
181: of degrees of freedom for each site and different edge states so
182: that some measures are taken necessarily to reach sufficient
183: accuracy in our computations. To calculate $\Delta_c$, we have to
184: use the periodic boundary condition (PBC) with the even number of
185: sites because of multi-edge excitations in the $SU(4)$ singlet
186: subspace when open boundary condition (OBC) is imposed. For other
187: gaps, we can efficiently expel the corresponding edge excitations
188: by an OBC algorithm. In particular, one lattice site is added at
189: each step and broken into two pseudo-sites. When the infinite
190: system algorithm is conducted with the size of the superblock up
191: to some odd number of sites $L$ which are preselected, the
192: sweeping procedure is performed for those $L$. The necessary
193: extrapolations for the thermodynamic limit are finally made
194: properly on the data with these preselected sites. In this case,
195: we have to redefine the gaps correspondingly, for instance
196: $\Delta_{4}$ $=E_0(L,2L+6,0) -E_0(L,2L+2,0)-6U$, where the
197: particle-hole symmetry is explicitly taken into account. In the
198: strong coupling region, PBC is often used to identify the bulk values of
199: a gap rather than edge-excitation energy obtained under OBC. In
200: our computations, $t$ is set to be unit and 2000 states are kept
201: for most cases and the maximal truncation error is the order of
202: $10^{-7}$.
203:
204: \begin{figure}[ht]
205: \includegraphics[width=7.0cm,angle=0]{gs-energycorrection.eps}
206: \caption{Ground state energy correction per site from
207: $\mathcal{H}^{(2)}_{eff}$ (solid line) versus $U$ in comparison
208: with DMRG data ($\bigcirc$).} \label{fig1}
209: \end{figure}
210: Figure \ref{fig1} displays the ground state energy correction per site arising
211: from the hoping term for the thermodynamic limit. Both the
212: numerical results and the perturbation theory show this energy
213: increases monotonically as $|U|$ increases and approaches to
214: zero asymptotically. When $-2<U<0$, the results of the perturbation
215: theory deviates from DMRG ones, and the deviation becomes
216: significant for small $|U|$. This is reasonable because in this
217: region the hopping term $\mathcal{H}_{t}$ is no longer
218: perturbative. However, for $U<-2$, the perturbation theory
219: provides very good results which accurately agrees with the DMRG
220: data. On the other hand, our DMRG calculations with both OBC and
221: PBC show that for a finite and even $L$, the ground state is
222: unique and a $SU(4)$ singlet which belongs to the irreducible
223: representation $[1^4]$\cite{YAMASHITA1}. In addition, slightly
224: above the ground state there is one accompanied $SU(4)$ singlet
225: excited state, whose energy difference from the ground state
226: diminishes as $L\rightarrow\infty$. Therefore, one obtains
227: two-fold degenerate ground states in the thermodynamic limit,
228: which are consistent with our analysis based on ${\mathcal
229: H}_{eff}^{(2)}$. These degenerate CDW ground states with the long
230: range order result from the translational symmetry breaking.
231:
232: In FIG. \ref{fig2}, we show the DMRG results on the charge gap
233: $\Delta_c$ and the four-particle excitation gap $\Delta_4$ for the
234: entire range of $U<0$. First of all, one can see that both
235: $\Delta_c$ and $\Delta_4$ are non-vanishing for all finite $U<0$
236: so that the ground state is insulating rather than metallic in
237: contrast to the $SU(2)$ case. Secondly, these two gaps behave
238: non-monotonically with $U$. The maximum shown around $U=-2$
239: indicates a crossover region between weak and strong interaction
240: regimes. Third, the perturbation theory for the strong coupling
241: regime provides correctly the asymptotic behavior for large $|U|$
242: limit and shows a qualitative agreement with the DMRG results in
243: the strong coupling regime. The visible deviation from the DMRG
244: results sets on at about $U\approx-5$ lower than that ($U\approx
245: -2$) for the ground state energy correction shown in Fig. 1. In
246: the weak coupling regime, $\Delta_c$ and $\Delta_4$ shown by DMRG
247: decrease with increasing $U$. Finally, while ${\mathcal
248: H}^{(2)}_{eff}$ can predict $\Delta_c=\Delta_4$ only in the strong
249: coupling limit, our DMRG calculations show that within the
250: numerical accuracy $\Delta_c$ remains equal to $\Delta_4$ beyond
251: the strong coupling regime. Although it is difficult from the DMRG
252: calculations to obtain sufficiently accurate $\Delta_c$ for $-1
253: <U<0$ yet, it is reasonable to conclude that $\Delta_4$ is equal
254: to $\Delta_c$ for all $U<0$ in the $SU(4)$ case.
255:
256: \begin{figure}
257: \includegraphics[width=7.0cm,angle=0]{four-charge-gap.eps}
258: \caption{DMRG results for charge gap (+) and four-particle gap
259: ($\bigcirc$) versus $U$ in comparison with perturbation ones
260: (solid line). Error bars are smaller than size of symbols except for
261: $\Delta_c$ at $U=-1$ as estimated with keeping different
262: states.} \label{fig2}
263: \end{figure}
264:
265: Now we turn to the other three types excitations: the first one is
266: the quasi-particle gap $\Delta_1$ for adding single particle or
267: single hole to the system, the second one spin gap $\Delta_s$
268: corresponding to the excitation energy in the spin triplet channel
269: from the ground state, and the last one two-particle gap
270: $\Delta_2$ defined as energy cost when two particles or two holes
271: are added to the system. While these three gaps together with
272: $\Delta_c$ and $\Delta_4$ essentially involve all kinds of
273: relevant excitations, they have significantly different behaviors.
274: Since the ground state is the CDW state with the long range order,
275: it is insightful to analyze those excitations in the Hartree-Fock
276: (HF) approximation. The CDW state in HF approximation can be
277: achieved by simply writing the on-site interaction as
278: $n_{i\sigma}n_{i\sigma^{'}}\approx \langle{n_{i\sigma}}\rangle
279: n_{i\sigma^{'}}+n_{i\sigma}\langle{
280: n_{i\sigma^{'}}}\rangle-\langle{n_{i\sigma}}\rangle\langle{
281: n_{i\sigma^{'}}}\rangle$ and assuming $\langle{n_{i\sigma}}\rangle
282: =n+(-1)^{i}\delta{n}$, where $n$ is the average number of
283: particles per site for each spin and $\delta{n}$ the corresponding
284: order parameter. At half-filling, one has $n=\frac{1}{2}$ and
285: $0\le{\delta{n}} \le\frac{1}{2}$. By further introducing
286: $a_{l\sigma}=c_{2l\sigma}$ and $b_{l\sigma}=c_{2l+1\sigma}$ for
287: each sublattice of the bipartite lattice, respectively, and taking
288: the Fourier transformation, then we can write down the
289: Hartree-Fock Hamiltonian as:
290: \begin{eqnarray}
291: \mathcal{H}^{HF} & = & -t\sum_{k\sigma}\left((1+e^{-ik}) a^{+}_{k\sigma}
292: b_{k\sigma}+h.c.\right) \nonumber \\
293: & & +3U\delta{n}\sum_{k\sigma}(a^{+}_{k\sigma}a_{k\sigma}
294: -b^{+}_{k\sigma}b_{k\sigma})+const.
295: \label{HF-K}
296: \end{eqnarray}
297: Diagonalizing this Hamiltonian, one can obtain two bands for each
298: spin species $\sigma$ with the quasi-particle dispersions
299: $w^{\pm}_{\sigma}=\pm{ \sqrt{\Delta_1^2+4t^{2}\cos^{2}
300: {\frac{k}{2}}}}$ where $\Delta_1=-3U\delta{n}$. Moreover, one has
301: $\Delta_c=\Delta_s=2\Delta_1$ in the HF approximation. These gaps
302: can be then evaluated after solving the following self-consistent
303: equation $2\delta{n}=\langle{a^{+}_{l\sigma}a_{l\sigma}} \rangle
304: -\langle{b^{+}_{l\sigma} b_{l\sigma}}\rangle$ for the order
305: parameter $\delta n$. In order to calculate the two-particle
306: excitation gap, however, it is necessary to account for the
307: particle-particle correlations into
308: %in the same spirit as
309: the HF approximation, which is nothing but the random phase
310: approximation (RPA). For this purpose, one first constructs the
311: basis with two-particle excitations from the ground state of
312: ${\mathcal H}^{HF}$ as follows:
313: \begin{eqnarray}
314: |\Psi_{pp'\sigma}\rangle=\alpha^{+}_{p\sigma}\alpha^{+}_{p'\bar{\sigma}}
315: |\Psi_{g}\rangle, ~~~~~~
316: |\Psi_g\rangle=\prod_{k\sigma}\beta^{+}_{k\sigma}|0\rangle
317: \label{basis})
318: \end{eqnarray}
319: where $\alpha^{+}_{p\sigma}$ is an operator creating one
320: quasi-particle with momentum $p$ and spin $\sigma$ in the $w^+$
321: band and in $|\Psi_{g}\rangle$ the band $w^-_{\sigma}$ is fully
322: filled up by quasi-particles $\beta^{+}_{k\sigma}$ with the
323: momenta $k$ and spin $\sigma$. Then $\Delta_{2}$ can be obtained
324: by diagonalizing ${\mathcal H}$ on the above basis (\ref{basis}).
325: \begin{figure}[h!]
326: \includegraphics[width=7.4cm,angle=0]{single-two-four-c-s-gap.eps}
327: \caption{DMRG results for one particle gap $\Delta_{1}$
328: ($\times$), two-particle gap $\Delta_{2}$($\Box$), spin gap
329: $\Delta_{s}$ ($\triangle$), four-particle excitation gap
330: $\Delta_{4}$ ($\small\bigcirc$) and charge gap $\Delta_{c}$($+$)
331: are shown as a function of $U$. The results of HF-RPA for
332: $\Delta_1$, $\Delta_2$ and $\Delta_s$ denoted by solid,
333: dot-dashed, and dashed lines, respectively. Inset shows
334: $\Delta/|U|$ versus $U$ for $\Delta_1$ ($\times$), $\Delta_2$ ($\Box$) and
335: $\Delta_s$ ($\triangle$).}
336: \label{fig3}
337: \end{figure}
338:
339: As compared to the DMRG results, we found that HF-RPA can provide
340: a qualitatively correct description for $\Delta_{1}$, $\Delta_2$,
341: and $\Delta_{s}$. Figure \ref{fig3} shows the DMRG data on all
342: five gaps as well as $\Delta_1,\Delta_2$, and $\Delta_s$ from
343: HF-RPA approximation for the region of $-2\leq U\leq 0$ and the
344: inset illustrates $\Delta_{1}$, $\Delta_2$, and $\Delta_{s}$ by
345: showing ratios for them over $|U|$ up to $|U|=10$. In contrast to
346: $\Delta_c$ and $\Delta_4$ as seen from Fig. 2, $\Delta_{1}$,
347: $\Delta_2$, and $\Delta_{s}$ increase with increasing $|U|$ and
348: become linear in large $|U|$ limit. It turns out that the relation
349: $\Delta_c=\Delta_s$, given by the HF approximation, is invalid for
350: general $U<0$. Moreover, it is unclear but beyond the present
351: approaches whether there is a symmetry enlargement similar to the
352: one proposed for the repulsive case\cite{ASSARAF1}. Nonetheless,
353: HF-RPA presents precise asymptotic behaviors for $\Delta_1$,
354: $\Delta_2$ and $\Delta_s$. In the weak coupling limit, the
355: exponential opening of these gaps can be well reproduced from the
356: solution to the self-consistent equation $\delta n\sim{-\frac
357: {2\pi t}{3U}e^{\frac{2\pi t}{3U}}}$ and with taking into account the
358: two-particle correlations. In the strong coupling limit, one has
359: $\Delta_1\sim -1.5U$ and $\Delta_s\sim-3U$ from $\delta
360: n\rightarrow0.5$, and $\Delta_2\sim-2U$ from the HF-RPA
361: calculations\cite{note}. The corresponding coefficients are in
362: good agreement with the DMRG results as can be seen from the
363: inset. On the other hand, the results of HF-RPA deviate from the
364: DMRG data apparently in the intermediate coupling regime, but this
365: is quite understandable since correlations involved in (\ref{HAM})
366: cannot be accurately handled in HF-RPA when $\mathcal{H}_{t}$ and
367: $\mathcal{H}_{u}$ become comparable, i.e. neither of them are
368: perturbative.
369:
370: In summary, we have studied the low energy properties of the
371: one-dimensional half-filled $SU(4)$ Hubbard model with the
372: attractive on-site interaction by using the DMRG method as well as
373: the perturbation theory. We found that the ground state is a CDW
374: insulating state with the long range order in which the
375: translational symmetry is broken and all kinds of excitations are
376: gapful for finite $U<0$. Within our numerical accuracy, we found
377: that the four-particle excitation gap is the same as the charge
378: gap. While the charge gap (the four particle excitation gap)
379: behaves non-monotonically, the others increase with increasing
380: $|U|$ and have a linear-U asymptotic behavior with different
381: coefficients. Therefore, we believe that the one-dimensional
382: attractive half-filled Hubbard model for the $SU(4)$ and $SU(2)$
383: cases belong to different universality classes. Moreover, we find
384: that the nature for the $SU(4)$ case can be further generalized to
385: the other $SU(N>2)$ cases\cite{ZHAO1}. At the end, it is
386: worthwhile to mention that since the four-particle excitation gap
387: as well as the charge gap are the smallest energy scale for the
388: $SU(4)$ case with $U<0$, it would be very interesting to detect
389: four-particle process (excitations) in an ultra-cold fermionic
390: atom system with the hyperfine spin-3/2.
391:
392: We would like to thank Y. Yamashita and Y.Z. Zhang for fruitful
393: discussions. J. Zhao acknowledges J. S\'olyom for the helpful
394: correspondence. X.Wang is supported under the Grants 2005CB32170X
395: and NSFC10425417.
396:
397: \begin{references}
398: \bibitem{HUBB1} J. Hubbard, Proc. Roy. Soc. London, Ser.A {\bf 276}, 238 (1963); {\bf 281}, 401 (1964)
399: \bibitem{MOTT} N.F. Mott, Proc. Phys. Soc. London, Ser.A {\bf 62}, 416 (1949); Can. J. Phys. {\bf 34}, 1356 (1956).
400: \bibitem{IMADA1} M. Imada, A. Fujimori and Y. Tokura, Rev. Mod. Phys. {\bf 70}, 1039 (1998).
401: \bibitem{MODUG1} G. Modugno, et al. Phys. Rev. A {\bf 68}, 011601(R) (2003).
402: \bibitem{AFFLECK1} I. Affleck and J.B. Marston, Phys. Rev. B {\bf 37}, 3774 (1988);
403: J.B. Marston and I. Affleck, Phys. Rev. B {\bf 39}, 11538 (1989).
404: \bibitem{WU1} C.J. Wu, J.P. Hu and S.C. Zhang, Phys. Rev. Lett. {\bf 91}, 186402 (2003).
405: \bibitem{LIEB1} E.H. Lieb and F.Y. Wu, Phys. Rev. Lett. {\bf 20}, 1445 (1968).
406: \bibitem{CHOY1} T.C. Choy, Phys. Lett. {\bf 80A}, 49 (1980); F.D.M. Haldane, Phys. Lett. A {\bf 80A}, 281 (1980);
407: T.C. Choy and F.D.M. Haldane, Phys. Lett. {\bf 90A}, 83 (1982).
408: \bibitem{WHITE1} S.R. White, Phys. Rev. B {\bf 48}, 10345 (1993).
409: \bibitem{PESCH1} I. Peschel, X. Wang, M. Kaulke and K. Hallberg, {\it Density Matrix Renormalization},
410: LNP{\bf 528}, Springer-Verlag, 1999.
411: \bibitem{ASSARAF2} R. Assaraf, et al. Phys. Rev. B {\bf 60}, 2299 (1999).
412: \bibitem{ASSARAF1} R. Assaraf, et al. Phys. Rev. Lett. {\bf 93}, 016407(2004).
413: \bibitem{SZIRMAI1} E. Szirmai and J. S\'olyom, Phys. Rev. B {\bf 71}, 205108 (2005).
414: \bibitem{LUTHER1} A. Luther and V.J. Emery, Phys. Rev. Lett. {\bf 33}, 589 (1974).
415: \bibitem{EMERY1} V.J. Emery, Phys. Rev. B {\bf 14}, 2989 (1976).
416: \bibitem{YAMASHITA1}e.g. Y. Yamashita, N. Shibata and K. Ueda, Phys. Rev. B {\bf 58}, 9114(1998)
417: \bibitem{note} The asymptotic behavior of $\Delta_1$, $\Delta_2$ and $\Delta_s$ can be obtained
418: alternatively from $H_u$. For instance, $\Delta_1$ is given by considering that
419: adding one hole costs energy $-3U$ and adding one particle gains no energy.
420: \bibitem{ZHAO1} J. Zhao, K. Ueda, and X. Wang, (unpublished).
421: \end{references}
422: \vfill
423: \end{document}
424: