1: \documentclass[article,showpacs]{revtex4}
2: \usepackage{graphicx}
3: \textwidth 17cm \textheight 25cm
4:
5: \begin{document}
6: \title{Bound states in the phase diagram of the extended Hubbard model}
7: \author{Maciej Bak}
8: \email[]{karen@delta.amu.edu.pl}
9: \affiliation{Institute of
10: Physics, A. Mickiewicz University, Umultowska 85, 61-614 Pozna\'n,
11: Poland}
12:
13: \begin{abstract}
14: The paper shows how the known, exact results for the two electron
15: bound states can modify the ground state phase diagram of extended
16: Hubbard model (EHM) for on-site attraction, intersite repulsion
17: and arbitrary electron density. The main result is suppression of
18: the superconducting state in favor of normal phase for small
19: charge densities.
20: \end{abstract}
21:
22: \pacs{71.10.Fd, 71.10.Hf, 71.27.+a}
23:
24: \maketitle
25:
26:
27: \section{Introduction}
28: The Hubbard model appears in almost all areas of solid state
29: physics. Its universality is connected with the fact that it
30: describes both band movement of charges as well as local and
31: nonlocal -- in the extended model -- correlations~\cite{hubbard}.
32: Treating its parameters as effective ones, the model has been used
33: in research of magnetism, superconductivity and especially high
34: temperature superconductivity (HTS), other various phenomena in
35: the solid, charge orderings, phase separation etc., in materials
36: like high temperature superconductors, bismuthates, Chevrel
37: phases, amorphous semiconductors, heavy fermion materials, systems
38: with alternating valence to name the few (for a review see, e.g.,
39: Ref.~\cite{review}).
40:
41: Unfortunately there are not many exact results concerning this
42: model. Usually the results are obtained in specific limits:
43: infinite dimensions, one dimension (the most numerous group),
44: infinite repulsion, half-filled band or other specific band
45: fillings. We can mention the exact solution in one dimension
46: ($d=1$) obtained by Bethe ansatz~\cite{betheA}, Lieb's
47: ferrimagnetism~\cite{lieb}, Nagaoka ferromagnetism in repulsive,
48: half-filled systems with one hole~\cite{nagaoka}, flat band
49: ferromagnetism~\cite{mielke}, some bounds on correlation
50: functions~\cite{koma,kubo,shen} and a finding of
51: Randeria~\cite{randeria}, according to which, existence of
52: two-electron bound states of s-wave symmetry is necessary and
53: sufficient condition for appearance of s-wave superconductivity in
54: $d=2$ systems with low electron density. Let's also note that
55: the mean-field BCS equations for superconductivity in the Hubbard
56: model with effective attractive interaction between electrons, in
57: the limit of vanishing electron density turn into Schrodinger
58: equations, which can also be solved exactly~\cite{nozieres}.
59:
60: A phase diagram in two dimensions for arbitrary $n$, a case of
61: special interest due to the possible connection with high
62: temperature superconductivity, is still under examination. The
63: results obtained in the mean-field approximation show competition
64: of phases: in half-filled band charge density waves (CDW) for
65: $W\ge 0$, superconductivity for $U<0$ and spin density waves for
66: $U>0$~\cite{review}. The calculations for $n\neq 1$ suggest
67: possibility of phase separation for $U<0$~\cite{rp,review}:
68: electron droplets for large enough $W<0$ and phase separation of
69: CDW (with $n=1$) with SS for $W>0$ (PS[CDW/SS]), for $n$ around
70: half-filled band competing with the pure SS state in low density
71: limit.
72:
73: This paper shows, how the known solutions for the bound states
74: (including exact solutions of Schrodinger equation) can be used
75: for modification of the ground state phase diagram of the extended
76: Hubbard model.
77:
78:
79: \section{Hamiltonian and the superconducting state}
80: We begin with the extended Hubbard Hamiltonian in standard
81: notation:
82: \begin{equation}\label{ham}
83: H=\sum_{ij\sigma}t_{ij}c^\dagger_{i\sigma}c_{j\sigma}+
84: U \sum_i n_{i\uparrow}n_{i\downarrow}+
85: \frac{1}{2}W \sum_{ij}n_{i\sigma}n_{j\sigma'}-
86: \mu \sum_i n_i \; ,
87: \end{equation}
88: where we sum over nearest-neighbors (nn) sites only. $U$ and $W$
89: are treated as effective parameters. We use broken-symmetry
90: Hartree-Fock approach (for details see Ref.~\cite{mrrt}). As we
91: are interested in the properties of the superconducting state, we
92: introduce averages of operators $c_{-k\downarrow}c_{k\uparrow}$ in
93: Wick's-type decoupling~\cite{mahan} of the four-operator terms in
94: the Hamiltonian. Non-zero average of such pair-creating operators
95: means phase-coherence among pairs, i.e. a presence of
96: superconducting state, and serves as an order parameter (see
97: Eq.~(\ref{dq})).
98: \begin{equation}\label{ham2}
99: H_0=\sum_{k\sigma}(\varepsilon_k-\overline\mu)c^\dagger_{k\sigma}c_{k\sigma}+
100: \sum_k (\Delta_k c^\dagger_{k\uparrow}c^\dagger_{-k\downarrow}+h.c.)+C \; ,
101: \end{equation}
102: where:
103: \begin{equation}\label{dq}
104: \Delta_{k_1}=\frac{1}{N}\sum_{k_2}(W_{k_2-k_1}+U)\langle c_{-k_2\downarrow}c_{k_2\uparrow}\rangle \; ,
105: \end{equation}
106: and $\overline\mu=\mu-(\frac{U}{2}+z W)n$, where $z$ is
107: coordination number of hypercubic lattice, $W_k=W\gamma_k$,
108: $\varepsilon_k=-t\gamma_k$, $\gamma_k=2\sum_{\alpha}^{d}\cos
109: k_\alpha$, $\alpha\in(x,y,z)$. After diagonalization of the
110: Hamiltonian Eq.~(\ref{ham2}) we obtain quasiparticle energy:
111: \begin{equation}\label{eq}
112: E_q=\sqrt{(\varepsilon_q-\overline\mu\ )^2+|\Delta_q|^2} \; .
113: \end{equation}
114: and a self-consistent equation for the gap:
115: \begin{equation}
116: \Delta_k=\frac{1}{N}\sum_q(W_{k-q}+U)\frac{\Delta_q}{2E_q}\tanh\frac{\beta E_q}{2} \; ,
117: \end{equation}
118: where $\beta=1/k_BT$, $T$ is temperature and $k_B$ Boltzmann
119: constant. The constant $C$ in the Hamiltonian Eq.~(\ref{ham2}) can
120: be expressed now as:
121: \begin{equation}
122: C=-\frac{1}{4}(U+2W z)n^2+\frac{1}{N}\sum_k\frac{|\Delta_k|^2}{2E_k}\tanh\frac{\beta E_k}{2} \; ,\\
123: \end{equation}
124: The pairing potential in the singlet channel (see Eq.~(\ref{dq}))
125: takes on the separable form for the square lattice and nn
126: interaction: $U+W_{k_1-k_2}=U+W\gamma_{k_1}\gamma_{k_2}/z$
127: (retaining only the terms of s-wave symmetry), and that makes
128: possible solving Eq.~(\ref{dq}) by an ansatz:
129: \begin{equation}\label{ansatz}
130: \Delta_k=\Delta_0+\Delta_{\gamma} \gamma_k \; ,
131: \end{equation}
132: what leads us to the set of self-consistent equations:
133: \begin{eqnarray}\label{delta0}
134: \Delta_0&=&-U \frac{1}{N}
135: \sum_q(\Delta_0+\gamma_q\Delta_{\gamma})F_q\;,\\\label{delta1}%\frac{1}{2E_q}\tanh\frac{\beta E_q}{2} \; ,\\\label{delta1}
136: \Delta_{\gamma}&=&-\frac{W}{z} \frac{1}{N}
137: \sum_q\gamma_q(\Delta_0+\gamma_q\Delta_{\gamma})F_q\;,\\\label{n}%\frac{1}{2E_q}\tanh\frac{\beta E_q}{2} \; ,\\\label{n}
138: n-1&=&-\frac{2}{N}\sum_q(\varepsilon_q-\overline\mu)F_q\;.%{2E_q}\tanh\frac{\beta E_q}{2} \; ,%\\
139: % p&=&-\frac{1}{N}\sum_k\frac{{\tilde\varepsilon}_k\gamma_k}{2E_k}\tanh\frac{\beta E_q}{2} \; ,
140: \end{eqnarray}
141: where $F_q=(\tanh\frac{\beta E_q}{2})/2E_q$. In the case of the
142: rectangular density of states (DOS) and pure on-site pairing we
143: can obtain analytical solutions~\cite{rp}; in the case of the
144: extended s-wave superconductivity (Eqs~(\ref{delta0}) --
145: (\ref{n})) analytical solutions exist in the limit of low electron
146: density. Introducing a new parameter: $\Delta_\gamma/\Delta_0$, we
147: can expand Eqs~(\ref{delta0}) -- (\ref{n}), treating $\Delta_0$ as
148: a small parameter. As a result we obtain a formula for critical
149: value for appearance of superconductivity for given $U$ and $n$ in
150: the ground state~\cite{bakmix}:
151: \begin{equation}\label{wcrmy}
152: W_{cr}=\frac{8t^2}{\mu(1-n)+8tI-2\mu^2/U} \; ,\mbox{\hspace{1cm}where\hspace{1cm}}
153: I=\int_{-1}^{\mu/D}x\rho(x)\;dx \; ,
154: \end{equation}
155: and $D=zt$ is half-bandwidth unit. In the case of rectangular DOS
156: this formula reduces to~\cite{bak2sol}:
157: \begin{equation}\label{wcrmysq}
158: W_{cr}(1+(n-1)^2(1+16t/U))=-4t \; .
159: \end{equation}
160: We can go a step further in our mean-field analysis and include
161: Fock term $ p=\frac{1}{N}\sum_{k\sigma}\gamma_k\langle
162: c^\dagger_{k\sigma}c_{k\sigma}\rangle$ into calculations. In
163: Eqs~(\ref{ham2}), (\ref{eq}), (\ref{delta0}) -- (\ref{n}),
164: $\varepsilon_k$ must be changed into
165: ${\tilde\varepsilon}_k=\varepsilon_k(1+pW/zt)$ then, and we have
166: to solve Eqs~(\ref{delta0}) -- (\ref{n}) self consistently with
167: the equation for the Fock term:
168: $p=-\frac{1}{N}\sum_k{\tilde\varepsilon}_k\gamma_k F_k$. Equations
169: (\ref{wcrmy}) -- (\ref{wcrmysq}) remain valid, with the change
170: $X\rightarrow X/(1+pW/zt)$ where $X=\mu$, $W_{cr}$ and $U$.
171:
172: \section{Low density limit} Going back to Hamiltonian
173: Eq.~(\ref{ham}) we can obtain exact results in the low density
174: limit. In the center-of-mass coordinate system we can expand the
175: wave function of the two-electron bound pair $\psi$ in the basis
176: of plane waves (i.e., eigenstates of the hopping part of the
177: Hamiltonian Eq.~(\ref{ham})). We can easily find the equations for
178: the coefficients of the expansion, what finally yields a set of
179: self-consistent equations for the wave function in the position
180: space, in terms of lattice Green functions~\cite{blaer,review}:
181: \begin{equation}\label{psiSelf}
182: \psi(\textbf{r})=\sum_{\textbf{r}'}G(E,\textbf{P},\textbf{r},\textbf{r}')g(\textbf{r}')\psi(\textbf{r}') \; ,
183: \end{equation}
184: where $G$ is lattice Green function defined by:
185: \begin{equation}
186: G(E,\textbf{P},\textbf{r},\textbf{r}')=\frac{1}{N}\sum_q\frac{\;e^{i\textbf{q}\cdot \textbf{r}}
187: e^{-i\textbf{q}\cdot \textbf{r}'}}{E-E_{\textbf{Pq}}} \; ,
188: \end{equation}
189: and $g(\textbf{r})$ is diagonal interaction matrix, consisting of
190: elements $U$ and $W$. Eigenenergy equation takes the form:
191: \begin{equation}\label{det}
192: det[{\cal G}-g^{-1}]=0 \; ,
193: \end{equation}
194: where $\cal{G}$ is a matrix with elements ${\cal
195: G}_{ij}=G(E,\textbf{P},\textbf{r}_i,\textbf{r}_j)$. This is an
196: analogue of Eqs~(\ref{delta0}) -- (\ref{delta1}). Let's note that
197: in the case of the two-electron bound pairs $\Delta=0$ and the
198: role of the binding energy is played by $\overline\mu/2$. For the
199: hypercubic lattices these equations were solved and it has been
200: found out that in one and two dimensions pairs for $W=0$ bind for
201: any negative $U$, while in three dimensions there is critical
202: value for $W$~\cite{mattis}. The formula for $W_{cr}$ in the case
203: of two-electron bound state reads~\cite{review}:
204: \begin{equation}\label{wcrex}
205: \frac{|W_{cr}|}{2t}=\left[1+\frac{2D}{U}\right]^{-1}+(\overline{C}-1)^{-1} \; ,
206: \end{equation}
207: where $\overline{C}=1/N\sum_k(1-\gamma_k/z)^{-1}$ is the Watson
208: integral. This is an exact result. Remembering that $\overline{C}$
209: is divergent for lattices of dimensions $d=1$ and $d=2$ we can
210: see, that for $n\rightarrow 0$ $\mu/D\rightarrow -1$,
211: $I\rightarrow 0$ (Eq.~(\ref{wcrmy})) and Eq.~(\ref{wcrex}) is a
212: limiting value of the formula Eqs~(\ref{wcrmy}) and
213: (\ref{wcrmysq}) for $d=1$ and $d=2$, as it should be. Not for
214: $d=3$ though; this case will be discussed later on.
215:
216: Let's note that Eqs~(\ref{wcrmy}), (\ref{wcrmysq}) and
217: (\ref{wcrex}), are valid for any combination of signs of $U$ and
218: $W$ and for large enough $U<0$ and $W<0$ there is a second branch
219: of solutions~\cite{friedberg,bak2sol}. The two branches are the
220: two solutions which realize in the two opposite limits:
221: $U=+\infty$ and $U=-\infty$ (or $W=\pm\infty$). The formal
222: equations and their solutions in both these limits are the same,
223: despite completely different physical situation. Nevertheless
224: these are the specific cases of two distinct solutions.
225:
226:
227: \section{Results and discussion}
228: In view of the Randeria's notice, described in the Introduction,
229: in the case of s-wave symmetry in two dimensions we can use
230: condition for existence of bound states as a condition for
231: existence of superconductivity. In Fig.~1 the boundaries expressed
232: by Eq.~(\ref{wcrmy}) for different lattice dimensionalities and
233: electron densities are shown. For parameters $U$, $W$ belonging to
234: the area above the plotted lines (mostly in the 1st quarter of
235: coordinate system) two-electron bound states, and what follows
236: s-wave superconductivity in two dimensions, can not exist. The
237: curves for $n=0$ in all dimensions are exact results.
238:
239: \begin{figure}[h]%fig1
240: \includegraphics[height=6cm]{fig1b.eps}
241: \raisebox{2.7cm}{\parbox{8cm}{\caption{Critical values for
242: existence of bound pairs and superconductivity for: $n=0$ for
243: $d=3$ (black circles - bound pairs only), $n=0$ and rectangular
244: DOS (full line), $n=0.25$, rectangular DOS and Fock term (dotted
245: line), $n=0.25$ and rectangular DOS (squares), $n=0.25$ for $d=2$
246: (triangles) and $n=0.25$ for $d=3$ (white circles -
247: superconductivity only). Line $n=0$ for $d=2$ is the same as for
248: the rectangular DOS. Half-bandwidth unit $D=4t$ for rectangular
249: DOS and for $d=2$ while $D=6t$ for $d=3$.}}}
250: \end{figure}
251:
252: As $n$ gets larger, the area of existence of bound pairs increases
253: for $W>0$ and $U<0$ and decreases in the part of the diagram with
254: $W<0$ and $U>0$. All curves except the one for $n=0$ in $d=3$ go
255: through the point with coordinates $(0,0)$. This illustrates the
256: fact that for $W=0$ infinitesimally small $U$ creates bound state
257: in $d=2$, while the threshold exists in $d=3$. Nevertheless we do
258: not have threshold in $d=3$ for $n\neq 0$ -- in agreement with
259: Randeria's notion about necessity of bound states for
260: superconductivity only in $d=2$. Let's note that for large $U$ and
261: $W$ the curves approach the asymptotes -- for curves crossing
262: through the axes origin the asymptotes are given by the formulas:
263: $U_{as}/t=-16(n-1)^2/(1+(n-1)^2)$ and $W_{as}/t=-4/(1+(n-1)^2)$.
264: This is connected with the fact that in the 3rd quarter of the
265: coordination system for $U<0$ and $W<0$ there exist second
266: branches of the solutions. As they have higher energy than the
267: solutions described in Fig.~1 they do not modify the ground state
268: phase diagram and are not shown here.
269:
270: The dotted line in Fig.~1 (and in Fig.~2) describes the results of
271: calculations with inclusion of the Fock term, using
272: Eq.~(\ref{wcrmysq}) modified by the $(1+pW/zt)$ term, as was
273: described in the end of Section~2. To simplify the calculations
274: the Fock term from the normal state was used: $p=n(2-n)$. For
275: $W>0$ this term broadens the band moving the system more into weak
276: coupling limit and enlarging the normal state area (opposite
277: behavior for $W<0$). The same effect can be seen in Fig.~2.
278:
279: In
280: Fig.~2 the boundaries of existence of bound states,
281: Eq.~(\ref{wcrmysq}) (black symbols), are plotted on a ground state
282: phase diagram of the extended Hubbard model for $U<0$ and $W>0$,
283: for arbitrary $n$ and rectangular DOS, together with the phase
284: boundary PS[CDW/SS]/SS taken from Ref.~\cite{rp} (white symbols).
285: Above the lines with black symbols s-wave superconductivity can
286: not exist in $d=2$. This way the superconducting state is
287: suppressed and normal state (NO) area is introduced into the phase
288: diagram. Let's note that also the phase separated state PS[CDW/SS]
289: is "reduced" to the NO phase and not to the CDW phase. This is due
290: to the fact that the CDW in the PS state is the CDW with $n=1$.
291: CDW with $n\neq 1$ is unstable, as it has negative
292: compressibility.
293:
294:
295: \begin{figure}[h]%fig2
296: \includegraphics[height=6cm]{fig2b.eps}
297: \raisebox{2.3cm}{\parbox{8cm}{\caption{Phase boundaries for
298: $U/D=-0.4$ (circles) and $U/D=-0.8$ (squares) calculated for the
299: rectangular DOS. Black symbols denote boundary of existence of
300: bound state, white symbols boundary between singlet
301: superconductivity (SS) and phase separated area PS[CDW/SS] (with
302: charge density wave CDW and SS). Symbols on dotted lines show the
303: results of calculations including Fock term.}}}
304: \end{figure}
305:
306: Another thing to note is the threshold for appearance of bound
307: states for $n=0$, which increases with increasing $|U|$. The phase
308: diagram is modified only for intermediate values of $|U|$ and
309: $|W|$, smaller from their asymptotic values $|U_{as}|$ and
310: $|W_{as}|$. For $|U|$ or $|W|$ larger than these values, bound
311: states exist for arbitrary value of the other parameter, in
312: agreement with Fig.~1.
313:
314: The calculations in Ref.~\cite{rp} consider only pure, on-site
315: s-wave pairing. Including $\Delta_\gamma$ (Eq.~(\ref{delta1}))
316: into calculations does not change much the described PS[CDW/SS]/SS
317: boundary -- $\Delta_\gamma$ is two orders of magnitude smaller
318: than $\Delta_0$ on this boundary. Inclusion of Fock term into the
319: calculations of the bound states, results in extending the area of
320: the normal phase, as was mentioned before. This effect increases
321: with increasing $|n|$ and $W$.
322:
323: In conclusion it was shown, how the analytical (and exact for
324: $n=0$) formulas for bound two-electron states can be used for the
325: modification of the $U<0$, $W>0$ part of the phase diagram of the
326: extended Hubbard model. The main result of this approach is
327: suppression of the superconducting and phase separated areas in
328: favor of the normal phase around half-filled band for intermediate
329: values of $|U|$ and $|W|$, larger than threshold values and
330: smaller than $|U_{as}|$ and $|W_{as}|$.
331: \vspace{-5mm}
332:
333: \begin{acknowledgements}
334: I acknowledge discussions with R. Micnas and S. Robaszkiewicz and
335: support from the Foundation for Polish Science. This work was also
336: supported by the Polish State Committee for Scientific Research
337: (KBN), Project No. 1 P03B 084 26.
338: \end{acknowledgements}
339:
340:
341: \begin{thebibliography}{10}
342: \bibitem[1]{hubbard}J.~Hubbard, Proc.~Roy.~Soc.~A {\bf 276}, 238
343: (1963).
344: \bibitem[2]{review}R.~Micnas, J.~Ranninger and S.~Robaszkiewicz,
345: Rev.~Mod.~Phys. {\bf 62}, 113 (1990).
346: \bibitem[3]{betheA}E. H. Lieb and F. Y. Wu, Phys. Rev. Lett. {\bf
347: 20}, 1445 (1968).
348: \bibitem[4]{lieb}E. H. Lieb, Phys. Rev. Lett. {\bf 62}, 1201
349: (1982), errata Phys. Rev. Lett. {\bf 62}, 1927 (1989).
350: \bibitem[5]{nagaoka}Y. Nagaoka, Phys. Rev. {\bf 147}, 392 (1966).
351: \bibitem[6]{mielke}A. Mielke, J. Phys. {\bf A 24}, 3311 (1991).
352: \bibitem[7]{koma}T. Koma and H. Tasaki, Phys. Rev. Lett. {\bf 68},
353: 3248 (1992).
354: \bibitem[8]{kubo}K. Kubo and K. Kishi, Phys. Rev. {\bf B 41}, 4866
355: (1990).
356: \bibitem[9]{shen}S. Q. Shen, Z. M. Qiu and G. S. Tian, Phys. Rev.
357: Lett. {\bf 72}, 1280 (1994).
358: \bibitem[10]{randeria}M.~Randeria, J.-M.~Duan and L.-Y.~Shieh,
359: Phys.~Rev.~Lett. {\bf 62}, 981 (1989).
360: \bibitem[11]{nozieres}P. Nozieres and S. Schmitt-Rink, J. Low Temp.
361: Phys. {\bf 59}, 195 (1985).
362: \bibitem[12]{rp}S.~Robaszkiewicz and G.~Pawlowski, Acta.~Phys.~Pol.~A {\bf 90}, 569 (1996).
363: \bibitem[13]{mrrt}R.~Micnas, J.~Ranninger, S.~Robaszkiewicz and
364: S.~Tabor, Phys.~Rev.~B {\bf 37}, 9410 (1998).
365: \bibitem[14]{mahan}G.~D.~Mahan, Many-Particle Physics, 2nd ed.
366: (Plenum Press, New York and London, 1993).
367: \bibitem[15]{bakmix}M.~Bak, cond-mat/0408692.
368: \bibitem[16]{bak2sol}M.~Bak, cond-mat/0406143.
369: \bibitem[17]{blaer}A. S. Blaer, H. C. Ren and O. Tchernyshyov,
370: Phys. Rev. {\bf B 55}, 6035 (1997).
371: \bibitem[18]{mattis}D.~C.~Mattis, Rev.~Mod.~Phys. {\bf 58}, 361
372: (1986).
373: \bibitem[19]{friedberg} R. Friedberg, T. D. Lee and H. C. Ren,
374: Phys. Rev. B {\bf 50}, 10190 (1994).
375: \end{thebibliography}
376:
377: \end{document}
378: