cond-mat0401307/v3.tex
1: %
2: %       Phase Diagram of the t-t'-t"-J Model
3: 
4: %       First Draft: C. T. Shih (May 2, 2003)
5: 
6: %
7: %       Please notify your modification below:
8: 
9: %
10: %
11: %\documentclass[aps,prl,preprint,groupedaddress]{revtex4}
12: 
13: %\documentclass[aps,prl,preprint,superscriptaddress,showpacs]{revtex4}
14: 
15: %\documentclass[aps,prl,twocolumn,groupedaddress,showpacs]{revtex4}
16: 
17: \documentclass[aps,prl,twocolumn,showpacs]{revtex4}
18: 
19: %\usepackage{graphicx}% Include figure files
20: 
21: \usepackage{epsfig}
22: 
23: \usepackage{dcolumn}% Align table columns on decimal point
24: 
25: \usepackage{bm}% bold math
26: 
27: \newcommand{\bk}{{\bf k}}
28: 
29: \newcommand{\pd}{$P^{ave}_d$ }
30: 
31: \newcommand{\nq}{$n({\bf k}=(\pi/2,\pi/2))$ }
32: 
33: \newcommand{\npi}{$n({\bf k}=(\pi,0))$ }
34: 
35: 
36: % You should use BibTeX and apsrev.bst for references
37: 
38: % Choosing a journal automatically selects the correct APS
39: 
40: % BibTeX style file (bst file), so only uncomment the line
41: 
42: % below if necessary.
43: 
44: %\bibliographystyle{apsrev}
45: 
46: 
47: 
48: \begin{document}
49: 
50: 
51: 
52: % Use the \preprint command to place your local institutional report
53: 
54: % number in the upper righthand corner of the title page in preprint mode.
55: 
56: % Multiple \preprint commands are allowed.
57: 
58: % Use the 'preprintnumbers' class option to override journal defaults
59: 
60: % to display numbers if necessary
61: 
62: %\preprint{}
63: 
64: 
65: 
66: %Title of paper
67: 
68: \title{Enhancement of Pairing Correlation by $t'$ in the Two-Dimensional Extended $t-J$ Model}
69: 
70: % repeat the \author .. \affiliation  etc. as needed
71: 
72: % \email, \thanks, \homepage, \altaffiliation all apply to the current
73: 
74: % author. Explanatory text should go in the []'s, actual e-mail
75: 
76: % address or url should go in the {}'s for \email and \homepage.
77: 
78: % Please use the appropriate macro foreach each type of information
79: 
80: 
81: % \affiliation command applies to all authors since the last
82: 
83: % \affiliation command. The \affiliation command should follow the
84: 
85: % other information
86: 
87: % \affiliation can be followed by \email, \homepage, \thanks as well.
88: 
89: \author{C.~T. Shih$^{1}$, T.~K. Lee$^{2}$, R. Eder$^{3}$, C.-Y. Mou$^{4}$ and Y.~C. Chen$^{1}$}
90: 
91: %\email[]{Your e-mail address}
92: %\homepage[]{Your web page}
93: %\thanks{}
94: %\altaffiliation{}
95: \address{$^1$Department of Physics, Tunghai University, Taichung,
96: Taiwan\\
97: $^2$Institute of Physics, Academia Sinica, Nankang,
98: Taiwan\\
99: $^3$Institut f\"ur Festk\"orperphysik, Forschungszentrum
100: Karlsruhe, Germany\\
101: $^4$Department of Physics, National Tsing Hua University,
102: Hsinchu, Taiwan}
103: 
104: 
105: \date{\today}
106: 
107: \begin{abstract}
108: We investigate the effects of the next-nearest-neighbor ($t'$) and
109: the third-nearest-neighbor ($t''$) hopping terms on
110: superconductivity correlation in the 2D hole-doped extended
111: $t-J$ model based on the variational Monte-Carlo, mean-field
112: calculation, and exact diagonalization method. Despite
113: of the diversity of the methods employed, the results all point to
114: a consistent conclusion: While the $d-$wave SC correlation is
115: slightly suppressed by $t'$ and $t''$ in underdoped regions, it is
116: greatly enhanced in the optimal and overdoped regions. The optimal
117: $T_c$ is a result upon balance of these two opposite trends.
118: \end{abstract}
119: \pacs{74.20.-z}
120: 
121: \maketitle
122: 
123: Right after the discovery of high temperature superconductors, the
124: two-dimensional (2D) $t-J$ model has been proposed to provide the
125: mechanism of superconductivity (SC)\cite{anderson87}. This idea quickly gained
126: momentum when variational calculations showed that the doping
127: dependence of pairing correlation\cite{gros88,zhang88} and the
128: phase diagram of antiferromagnetic (AFM) phase and  SC
129: \cite{lee88} seem to agree with experimental results fairly well.
130: However, recently calculations\cite{shih98} beyond the variational
131: method have challenged the notion that pure 2D $t-J$ model without
132: including other interactions is enough to explain the high values
133: of $T_c$. Although this issue is yet to be
134: settled\cite{sorella02,lee02,sorella02b}, there are
135: results by band-structure calculations\cite{raimondi96,pavarini01}
136: and experimental analysis\cite{fujimori} that hopping beyond
137: nearest neighbors is essential to raise  $T_c$. In fact they found
138: highest $T_{c,max}$ for different monolayer cuprates strongly
139: correlates with $t'/t$, where $t'$ is the second nearest-neighbor
140: hopping amplitude. This contradicts with previous
141: results\cite{white99,martins01} of exact calculations
142:  that  for the hole doped systems, introducing $t'$
143: into the t-J model will actually reduce pairing. Although the last
144: results are studied for systems doped only with a few  number of
145: holes, all these conflicting results raised a very serious
146: challenge to the $t-J$ type models.
147: 
148: There are many experimental and theoretical results to support the
149: presence of $t'$ and possibly also $t''$, the third nearest
150: neighbor hopping, in cuprates. The topology of large Fermi surface
151: seen by ARPES\cite{damascelli03,ino02} and
152: the change of sign of Hall coefficient as a function of
153: doping\cite{clayhold89}
154:  can best be understood in the
155: presence of $t'$ and $t''$. The single hole dispersion observed by
156: ARPES and the difference between hole- and electron-doped
157: systems\cite{lee03} also support the presence of these terms.
158: %Further more a recent experiment\cite{balakirev03}
159: %has reported a possible change of Fermi surface topology near
160: %optimal doping.
161: 
162: In this letter, we will use variational approach supplemented by
163: slave-boson mean-field (MF) calculations and exact diagonalization (ED)
164:  method to show that the
165: presence of $t'$ is indeed important for enhancing pairing beyond the
166: underdoped regime. The largest values of pairing correlation
167: obtained are proportional to $t'/t$ up to $t'/t=-0.3\sim -0.4$. In
168: addition we will show that the decrease of pairing correlation  at
169: very large hole density is related to the change of Fermi surface
170: topology. The conflicting results between theories and experiments
171: discussed above are naturally resolved within the extended $t-J$
172: model.
173: 
174: The Hamiltonian of the extended $t-J$ model is
175: \begin{eqnarray}
176: &H&=H_t+H_J=\\
177: &-&\sum_{ij}t_{ij}(\tilde{c}^\dagger_{i,\sigma}\tilde{c}_{j,\sigma}
178: + H.C.)+J\sum_{<i,j>}({\bf{S_i}\cdot
179: S_j}+\frac{1}{4}n_in_j)\nonumber \label{e:tjm}
180: \end{eqnarray}
181: where $t_{ij}=t$, $t'$, and $t''$ for sites $i$ and $j$ are
182: nearest, next nearest, and the third nearest neighbors,
183: respectively, and $t_{ij}=0$ for longer distance.
184: $\tilde{c}_{i,\sigma}=(1-n_{i,-\sigma})c_{i,\sigma}$, satisfies
185: the no-double-occupancy constraint. In our notation for hole doped
186: materials, $t'/t$ is negative while $t''=-t'/2$ most of the time.
187: 
188: The trial wave function (TWF) used in this study is the
189: $d_{x^2-y^2}$ resonating-valence-bond wave function
190: \begin{eqnarray}
191: \mid \Psi\rangle =
192: P_G\prod_k(u_k+v_kc^\dagger_{k,\uparrow}c^\dagger_{-k,\downarrow})\mid
193: 0\rangle
194: \end{eqnarray}
195: with
196: $u_k/v_k=\Delta_k/(\epsilon_k+\sqrt{\epsilon_k^2+\Delta_k^2})$,
197: $\epsilon_k=-2t(cosk_x+cosk_y)-4t'_vcosk_xcosk_y-2t''_v(cos2k_x+cos2k_y)-\mu$,
198: and $\Delta_k=2\Delta(cosk_x-cosk_y)$. Here the projection
199: operator $P_G$ enforces the constraint of one electron per site.
200: In addition to $\Delta$ and $\mu$, we have
201: included two important variational parameters $t'_v$ and $t''_v$ which
202: determine
203: the Fermi surface topology.
204: 
205: The d-wave pair-pair correlation $P_d({\bf R})$ is defined as
206: $\frac{1}{N_s}\langle\sum_i\Delta^\dagger_{\bf R_i}\Delta_{\bf
207: R_i+R}\rangle$, where $\Delta_{\bf R_i}=c_{{\bf
208: R_i}\uparrow}(c_{{\bf R_i+\hat{x}}\downarrow}+c_{{\bf
209: R_i-\hat{x}}\downarrow}-c_{{\bf R_i+\hat{y}}\downarrow}-c_{{\bf
210: R_i-\hat{y}}\downarrow})$. The long range part of $P_d({\bf R})$
211: is a flat plateau for nonzero $\Delta$, and we define $P^{ave}_d$
212: as the averaged value of the $\mid{\bf R}\mid > 2$ part of
213: $P_d({\bf R})$ to estimate the strength of SC of the system.
214: 
215: \begin{figure}[here]
216: \rotatebox{-90}{\includegraphics[width=2.4in]{hole_144.eps}}
217: \caption{$P^{ave}_d$ for $J/t=0.3$ with several $t'$ and $t''$ for
218: a hole-doped $12\times 12$ lattice.} \label{f:phase_144v}
219: \end{figure}
220: 
221: Fig.\ref{f:phase_144v} shows the variational Monte Carlo (VMC)
222: results of $P^{ave}_d$ for several $t'$ and $t''$ for different
223: hole densities $\delta$ with $J/t=0.3$ in a $12\times 12$ lattice.
224: $P^{ave}_d$ for the $t-J$ model (open circles) has the
225: ``dome-like'' shape which is similar to the experimental results
226: of $T_{c}$ versus doping. However, this could be an artifact of
227: the variational study which, we believe, overestimates the order
228: parameters and will be largely suppressed when we go beyond
229: variational calculation as shown in our previous
230: study\cite{shih98}. The most suprising result is that when $t'$ is
231: included, $P^{ave}_d$ changes dramatically. For the overdoped
232: regime, $P^{ave}_d$ is greatly enhanced by almost one order of
233: magnitude for $t'/t=-0.3$ and $t''/t=0.15$. The SC region extends
234: to $\delta\sim 0.4$ and the peak of the superconducting dome is at
235: $\delta\sim 0.3$ and the magnitude of the maximal \pd is about 2.5
236: times larger than for $t'=t''=0$. In the underdoped region, \pd is
237: almost unchanged or very slightly suppressed. Another thing to
238: note is that beyond the value of $t'/t=-0.3\sim-0.4$ $P^{ave}_d$
239: is no longer enhanced.
240: 
241: \begin{figure}[here]
242: \rotatebox{-90}{\includegraphics[width=2.3in]{144t2.eps}}
243: \caption{Maximal \pd for different $t'$ with $t''=-t'/2$ for
244: $8\times 8$ (open circle) and $12\times 12$ (full circle)
245: lattices.} \label{f:144t2}
246: \end{figure}
247: 
248: Fig.\ref{f:144t2} plots the maximal possible value of \pd for all doping density
249: as a function of $t'$. The
250: maximal \pd is proportional to $t'$ in the range $0\ge t'\ge
251: -0.3\sim-0.4$. Beyond these values pairing is no longer enhanced.
252: Coincidentally these values are about the same  value of $t'/t$
253: for mercury cuprates as estimated by Pavarini {\em et al.}\cite{pavarini01}
254: but much larger than what was reported in Ref.\cite{raimondi96}.
255: Among all the cuprate series, mecury
256: cuprate maintains the record of having highest $T_c$ for almost a decade.
257: 
258: The above result resolves the discrepancy between previous
259: denisty-matrix-renormalization-group (DMRG)
260: studies \cite{white99,martins01} and the band structure
261: calculation \cite{pavarini01}. As DMRG studies were concerned with underdoped
262: region
263: while the highest  $T_{c,max}$ examined by the band structure calculation
264:  certainly depends on
265: optimal and overdoped regions.
266: 
267: \begin{figure}[here]
268: \rotatebox{-90}{\includegraphics[width=2.7in]{pdnk144}}
269: \caption{(a)\pd and (b) \npi vs. $t'$ for $\delta=0.31$ (full
270: circles) and $0.083$ (open circles), $J/t=0.3$, $t''=-t'/2$.}
271: \label{f:nk_pd}
272: \end{figure}
273: 
274: The different effects of $t'$ and $t''$ on \pd between
275: overdoped and underdoped regions are related to the
276:  shape of Fermi surface.
277:  Fig.\ref{f:nk_pd} shows the relations of \pd and
278: \npi versus $t'$ for $\delta=0.31$ and $0.083$. For the overdoped
279:  case ($\delta=0.31$), as $-t'$ increases up to $0.3$,  \npi increases
280: from less than $0.4$ to larger than $1$ and \pd also increases sharply from
281: less than $0.01$ to larger then $0.04$.
282:  Since $d-$wave SC gap
283: is largest at ${\bf k}=(\pi,0)$,
284:  occupation of this ${\bf k}$ states by electrons
285:  enhances \pd.
286:  On the other hand for the
287: underdoped case, \npi is almost unchanged because the
288: occupation \npi is already quite  large ($>0.9$) for $t'=t''=0$
289: and the effect of Fermi surface becomes unimportant.
290: The slight suppression of \pd
291: may be due to the destructive interference mechanism of the
292: pair-hopping as suggested by Martins {\em et al.}\cite{martins01}.
293: 
294: The decrease of \pd for $-t'\ge 0.4$ in the overdoped regime such
295: as $\delta=0.31$
296:  is also likely the consequence of the change of  the
297: Fermi surface. \npi is almost  saturated at $-t'=0.4$ and remains
298: unchanged for larger $-t'$. It is not difficult to recognize that as $-t'$
299: becomes much larger than $t$, electrons will occupy in separate
300: regions around $\bk=(\pm\pi,0)$ and $\bk=(0,\pm\pi)$. Hence the
301: Fermi surface becomes disjoint pieces. Although at $-t'/t=0.4$,
302: the Fermi surface is still connected but this tendency is already
303: observed. The density of states begins to decrease and this is
304: probably the reason for the suppression of pairing beyond
305: $-t'/t'\ge 0.4$.
306: 
307: \begin{figure}[here]
308: \rotatebox{-90}{\includegraphics[width=2.3in]{fs_144}}
309: \caption{Fermi surface of $\delta=$ (a)0.19, (b)0.31 (c)0.42 (d)
310: 0.49 for $12\times 12$ lattice. (e)optimal parameters $t'_v$
311: (squares) and $\Delta$ (circles).} \label{f:para144}
312: \end{figure}
313: 
314: Fig.\ref{f:para144} shows the Fermi surface and optimal parameters
315: $\Delta$ and $t'_v$ as a function of doping density for a
316: $12\times 12$ lattice with $J/t=0.3$, $t'=-0.3$ and $t''=0$. In
317: (a-d) the white region denotes $n({\bf k})\geq1.2$ and dark region
318: for  $n({\bf k})\leq0.5$. The density with maximal \pd is near
319: $\delta_{opt}=0.31$ as shown by the solid square in
320: Fig.\ref{f:phase_144v}. The shapes of the Fermi surfaces are very
321: different for $\delta>\delta_{opt}$ and $\delta<\delta_{opt}$
322: cases. Fig.\ref{f:para144}(e) shows that in the  region
323: $0.2<\delta<0.3$ although $\Delta$ becomes  smaller, $-t'_v$ is
324: still quite large and pairing is further enhanced. Doping beyond
325: $\delta>0.3$, $-t'_v$ begins to decrease quickly. This gives very
326: low electron occupation at  $(\pi,0)$ as shown in
327: Fig.\ref{f:para144}(c) and (d), then the pairing is reduced. This
328: result shows that the
329: enhancement of pairing by including $t'$ is not due to larger
330: $\Delta$ but from the deformation of the Fermi surface instead.
331: 
332: Although we have emphasized the
333: particular correlation between d-wave SC gap and electron
334: occupation at ${\bf k}=(\pi,0)$ as the reason for enhancement of
335: pairing, another familiar effect may also have played a role.
336: It is well known that
337: $t'$ will shift the van Hove singularity\cite{newns} in density of states,
338: but it is always around  ${\bf k}=(\pi,0)$. Hence it could be that the optimal
339: $t'_v$ are chosen to have the high density of states at Fermi surface. This
340: may be related to the observed extended region of flat band by
341: ARPES\cite{dagotto}.
342: %But these normal state properties are out of the scope of this paper.
343: 
344: The results of  slave-boson MF calculation\cite{ubben}
345:  in Fig.\ref{f:mft1} show
346: similar behavior for the overdoped regime that indeed $t'$
347: enhances $T_c$. Including $t''$ pushes the superconducting regime
348: to even larger doping density $\delta$ by occupying the momenta
349: around $(\pi,0)$. A similar effect of $t''$ can also be seen in
350: Fig.\ref{f:phase_144v} by comparing the $(t',t'')=(-0.3,0)$ (full
351: squares) and $(-0.3,0.15)$ (open triangles) curves. Since the
352: slave-boson method is not quite reliable quantitatively in the
353: underdoped regime, we did not show the values of $T_c$ in
354: Fig.\ref{f:mft1}. However, if we do take the values literally, the
355: values achieved for $T_{c,max}$ are not as greatly enhanced by
356: $t'$ as for the VMC result shown in Fig.\ref{f:144t2}.
357: %Also when $-t'$ is too large, $T_c$ is
358: %no longer enhanced.
359: Similar results are reported by
360: the interlayer tunneling model\cite{hqlin2003}.
361: 
362: \begin{figure}
363: \rotatebox{-90}{\includegraphics*[width=60mm]{mft1.eps}}
364: \caption{Mean-field results of $T_c$ for overdoped regime with
365: $(t',t'')=(0,0)$ (circles), $(-0.3,0)$ (squares) and $(-0.3,0.15)$
366: (diamonds).} \label{f:mft1}
367: \end{figure}
368: 
369: Fig.\ref{f:ed20} shows the pair-pair correlation\cite{pair_ed} for
370: the longest distance ${\bf R}=(1,3)$ for 20-site lattices obtained
371: by the ED method. Pairing correlations for 2 and 4 holes are
372: suppressed by $t'$ and $t''$, but enhanced for the overdoped 6-
373: and 8-hole cases. The non-monotonic behavior of the overdoped
374: cases  is due to the level crossing of this system.
375: If we focus on the $s-$like symmetry states,
376: \pd will vary monotonically in the region $0\ge t'\ge -0.3$. The
377: result of ED method is quite consistent with the variational and
378: MF results that the enhancement of \pd by $t'$ occurs for larger
379: hole densities.
380: 
381: \begin{figure}[here]
382: \rotatebox{-90}{\includegraphics[width=2.5in]{20pair}}
383: \caption{$P_d({\bf R}=(1,3))$ for versus $t'$ ($t''=-t'/2$) for
384: different 2 (open circle) , 4 (full circle), and 6 (open triangle)
385: holes in 20 sites. The dash line shows $P_d$ for the same symmetry
386: of the 8-hole case.} \label{f:ed20}
387: \end{figure}
388: 
389: In summary, in the optimal and overdoped regions, SC is greatly
390: enhanced because of the deformation of the Fermi surface at these
391: doping densities. \npi is enhanced by including $t'$ and $t''$.
392: The occupation of $(\pi,0)$ by electrons is important for the
393: enhancement. The maximum enhancement of pairing correlation seems
394: to be reached for $-t'/t=0.3\sim 0.4$. On the other hand, \pd is
395: not enhanced for the underdoped regime as $n({\bf k}=(\pi,0))$ is
396: hardly affected by including $t'$.
397: It is well accepted that the physics on the overdoped side
398: is apparently much simpler in
399: that
400: experiments on the cuprates and theory for the t-J model indicate
401: that the overdoped materials are very close to ordinary Fermi liquids, whence
402: `Fermi-surface-based' arguments like ours are much more reliable than
403: in the underdoped region.
404: Our result shows that the
405: extended $t-J$ model naturally predicts the strong
406: correlation\cite{pavarini01} between $t'/t$ and $T_{c,max}$
407: observed in experiments for monolayer cuprates. In addition it
408: also indicates that further increase of $t'$ beyond what mercury
409: cuprates have most likely will not enhance $T_{c,max}$.
410: 
411: 
412: Although we have consistent results from VMC, ED methods and
413: slave-boson MF calculations, the optimal doping density,
414: $\delta_{opt}$, in Fig.\ref{f:phase_144v} is around 0.3 instead of
415: 0.17 obtained in the experiments and in the VMC results without
416: including $t'$. But this is actually not a drawback. As argued in
417: References\cite{shih98, lee02}, VMC is expected to overestimate
418: the  values of the variational parameters $\Delta$ which is
419: related to the exhange energy $J$. Hence the pairing correlation
420: is definitely  much larger than that of a real ground state. From
421: our previous experiences the $\delta_{opt}$ seems to be always
422: shifted to a smaller value when we improve the variational wave
423: functions.  Thus in the future work going beyond VMC calculations,
424: we believe there is a better chance that we will have
425: $\delta_{opt}$  closer to the experimental value. We also expect
426: interlayer coupling will be important in getting the correct
427: $\delta_{opt}$. Now $t'$ is shown to be important in
428: enhancing pairing and it is also present in all high $T_c$
429: cuprates, the debate\cite{lee02,sorella02b} about pairing
430: robustness in the 2D $t-J$ model without $t'$ becomes somewhat irrelevant.
431: The agreement of $\delta_{opt}$ between $t-J$ VMC and experiments
432: looks fortuitous.
433: 
434: One of the consequences of our results is that the shape
435: of the Fermi surface plays an important role for high
436: temerature superconductors in the optimal and overdoped regions.
437: Fig.\ref{f:para144} (b) and (c) show
438: that the Fermi surface changes from hole-like to electron-like
439: once the maximum pairing is reached and the pairing is reduced as
440: doping increases. This is consistent with the ARPES results for
441: $La_{2-x}Sr_xCuO_4$\cite{ino02}. It may also be related to the
442: recent experiment\cite{balakirev03} which shows that the
443: low-temperature Hall coefficient  for the
444: $Bi_2Sr_{1.51}La_{0.49}CuO_{6+\delta}$ system exhibits a sharp
445: change at the optimal doping density. Clearly this issue deserves
446: more detailed study in the future.
447: 
448: The work is supported by the National Science Council in Taiwan
449: with Grant nos. NSC-91-2112-M-029-005, 006, and 068, and
450: NSC-92-2112-M-007-038. Part of the calculations are performed in
451: the IBM P690 in the National Center for High-performance Computing
452: in Taiwan, and the PC clusters of the Department of Physics and
453: Department of Computer Science and Engineering of Tunghai
454: University, Taiwan. We are grateful for their support.
455: 
456: % Create the reference section using BibTeX:
457: 
458: 
459: %\bibliography{version2}
460: 
461: \begin{references}
462: \bibitem{anderson87}P. W. Anderson, Science {\bf 235}, 1196
463: (1987).
464: 
465: \bibitem{gros88} C. Gros, Phys. Rev. B {\bf 38}, 1196 (1988).
466: 
467: \bibitem{zhang88} F. C. Zhang {\it et al.}, Supercon. Sci. Tech.
468: {\bf 1}, 36 (1988).
469: 
470: \bibitem{lee88} T. K. Lee {\it et al.}, Phys. Rev. B {\bf 38},
471: 11809 (1988).
472: 
473: \bibitem{shih98} C. T. Shih {\it et al.}, Phys. Rev. Lett. {\bf 81},
474: 1294 (1998).
475: 
476: \bibitem{sorella02} S. Sorella {\it et al.}, Phys. Rev. Lett. {\bf
477: 88}, 117002 (2002).
478: 
479: \bibitem{lee02} T. K. Lee {\it et al.}, Phys. Rev. Lett. {\bf 89},
480: 279702 (2002).
481: 
482: \bibitem{sorella02b} S. Sorella {\it et al.}, Phys. Rev. Lett.
483: {\bf 89}, 279703 (2002).
484: 
485: \bibitem{raimondi96} R. Raimondi {\it et al.}, Phys. Rev. B {\bf
486: 53}, 8774 (1996); L.F. Feiner {\it et al.}, Phys. Rev. Lett. {bf\ 76},
487: 4939 (1996).
488: 
489: \bibitem{pavarini01} E. Pavarini {\it et al.}, Phys. Rev. Lett.
490: {\bf 87}, 047003 (2001).
491: 
492: \bibitem{fujimori} K. Tanaka {\it et al.}, cond-mat/0312575.
493: 
494: \bibitem{white99} S. R. White {\it et al.}, Phys. Rev. B {\bf 60},
495: R753 (1999).
496: 
497: \bibitem{martins01} G. B. Martins {\it et al.}, Phys. Rev. B {\bf
498: 64}, 180513 (2001).
499: 
500: \bibitem{damascelli03} A. Damascelli {\it et al.}, Rev. Mod. Phys.
501: {\bf 75}, 473 (2003).
502: 
503: \bibitem{ino02}A. Ino {\it et al.}, Phys. Rev. B {\bf 69}, 094504
504: (2002).
505: 
506: \bibitem{clayhold89} J. Clayhold {\it et al.}, Phys. Rev. B {\bf
507: 39}, 7324 (1989).
508: 
509: \bibitem{lee03}T. K. Lee {\it et al.}, Phys. Rev. Lett. {\bf 90},
510: 067001 (2003).
511: 
512: 
513: %\bibitem{yoshida01} T. Yoshida {\it et al.}, Phys. Rev. B {\bf
514: %63}, 220501 (2001).
515: 
516: \bibitem{newns}D. M. Newns {\it et al.}, Phys. Rev. B {\bf 43},
517: 3075 (1991).
518: 
519: \bibitem{dagotto}Elbio Dagotto {\it et al.}, Phys. Rev. Lett. {\bf 73},
520: 728 (1994).
521: 
522: \bibitem{ubben}M.U. Ubbens and P.A. Lee, Phys. Rev. B {\bf 46},
523: 8434 (1992).
524: 
525: \bibitem{hqlin2003} X. J. Chen {\it et al.}, cond-mat/0306680.
526: 
527: \bibitem{pair_ed} $P_d({\bf R})$ is slightly modified in the ED calculation:
528: First, $P_d({\bf R})= \frac{1}{N_s} \langle \sum_i \Delta_{{\bf
529: R}_i} \Delta_{{\bf R}_i+{\bf R}}^\dagger \rangle$. Second,
530: $c_{{\bf R}_i,\uparrow}^\dagger \hat{c}_{{\bf
531: R}_j,\downarrow}^\dagger \rightarrow \frac{1}{\sqrt{2}}(c_{{\bf
532: R}_i,\uparrow}^\dagger \hat{c}_{{\bf R}_j,\downarrow}^\dagger-
533: c_{{\bf R}_i,\downarrow}^\dagger \hat{c}_{{\bf
534: R}_j,\uparrow}^\dagger)$.
535: 
536: \bibitem{balakirev03} F. F. Balakirev {\it et al.}, Nature {\bf
537: 424}. 912 (2003).
538: 
539: \end{references}
540: 
541: \end{document}
542: 
543: %
544: 
545: % ****** End of file template.aps ******
546: