1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: %\documentstyle[preprint,aps]{revtex}
3: %\documentstyle[prl,aps,multicol]{revtex}
4: \documentstyle[prl,twocolumn,aps,graphics]{revtex}
5: \input{epsf}
6: \begin{document}
7: \draft
8: \twocolumn[\hsize\textwidth\columnwidth\hsize\csname @twocolumnfalse\endcsname
9: \title{Current-current correlations in a model $CuO_3$ system
10: }
11:
12: \author{Bhargavi Srinivasan and Marie-Bernadette Lepetit}
13:
14: \address {Laboratoire de Physique Quantique, UMR 5626 du CNRS,
15: Universit\'e Paul Sabatier, F-31062 Toulouse Cedex 4, France}
16:
17:
18: \date{\today}
19:
20: \maketitle
21:
22: \begin{abstract}
23: We study a 3-band extended Hubbard model on a $CuO_3$ system
24: by the density matrix renormalization group (DMRG) method.
25: Our system has geometry
26: similar to that of a section of the $CuO_2$ plane of the copper-oxide
27: superconductors.
28: We have studied the effects of nearest-neighbour
29: repulsions on the current-current corre\-lations of
30: two model $CuO_3$ systems.
31: We have calculated several types of staggered cur\-rent correlations
32: as a function of various
33: model parameters such as interaction strengths and filling.
34: We show that repulsive interactions
35: are clearly capable of increasing the magnitude of the
36: current-current correlations in both models studied.
37: The long-distance behavior of the current-current correlations
38: is qualitatively influenced by $V$.
39: We show that repulsive interactions are capable of
40: enhancing correlations involving at least one copper-oxygen
41: rung over distances comparable to the length of the
42: system. In some cases, $V$ decreases the exponent
43: corresponding to the power-law decay, leading to a
44: considerable slowing down of the long-distance decay
45: of the current-current correlations, compared to that
46: of the Hubbard model.
47: \end{abstract}
48: \pacs{PACS numbers: 71.10.Fd, 71.10Hf, 74.20Mn}
49: \vskip1pc]
50:
51: %\begin{multicols}{2}
52: \narrowtext
53:
54:
55:
56: \newpage
57: \section{Introduction}
58: The unusual normal-state phenomenology of the
59: high $T_c$ superconductors
60: constitutes an important theoretical challenge since
61: there is not yet a consensus in the community
62: about a consistent
63: microscopic theory that would reproduce all aspects of
64: the phase diagram
65: of the cuprates.
66: Angle-resolved photoemission spectroscopy (ARPES) experiments
67: have revealed several properties of the so-called
68: pseudogap phase in the underdoped part of the
69: phase diagram \cite{timusk}
70: and it is likely that this phase holds the key
71: to a general understanding of the phase diagram.
72: Among the scenarios proposed to explain the pseudogap
73: phase, we mention the 'circulating-current' order-parameter,
74: which breaks time-reversal symmetry and four-fold rotational
75: symmetry about the copper atoms, while preserving the translational
76: symmetry of the lattice\cite{varma}. This theory also gives rise
77: to non-Fermi liquid behaviour. The 'd-density wave' (DDW)
78: order-parameter
79: based on orbital currents \cite{ddw} has been proposed as
80: a hidden order parameter for the phase diagram of the
81: cuprates. In Fig. 1 we present a schematic picture of these
82: circulating currents in the $CuO_2$ plane
83: of the copper-oxides.
84:
85: Phases similar to those with DDW or circulating-currents order-parameters
86: have been discussed
87: previously in the literature under
88: different guises. The so-called orbital antiferromagnet
89: is a phase with broken translational and time-reversal symmetries,
90: described for the excitonic insulator \cite{halperin}
91: and in the two-dimensional (2D)
92: Hubbard model \cite{schulz}. 'Flux phases' were studied in
93: the context of the large n limit of the Heisenberg model \cite{affleck}.
94: Staggered currents were found to
95: be stabilized by attractive nearest-neighbour interactions
96: in the 2D Hubbard model \cite{stanescu}.
97: On the other hand, $t-J$ clusters \cite{lee,leung} were
98: shown variationally and numerically to have a staggered flux
99: order-parameter.
100: Recently, the doped two-leg $t-J$ ladder, studied by the
101: density matrix renormalization group (DMRG) method \cite{scalapino}
102: was found to have exponentially decaying
103: rung-rung current-current correlations.
104: However, the staggered flux phase was found to be stabilized in the
105: half-filled 2-leg ladder by abelian bosonization methods\cite{marston1}.
106: Further, inclusion of longer-range interactions in the
107: 2-leg ladder (extended $t-U-V-J$ model) appears to
108: stabilize the
109: staggered flux phase, as seen from DMRG calculations\cite{marston2}.
110:
111: \begin{figure}
112: \begin{center}
113: \resizebox{6.0cm}{6cm}{\includegraphics{currents_1.eps}}
114: \end{center}
115: \caption{
116: Schematic picture of the circulating-current phase in
117: the copper-oxides. The filled circles represent copper atoms
118: and the vertices oxygen atoms of the $CuO_2$ plane. The
119: arrows on each bond correspond to the direction of the current
120: along the bond.
121: }
122: \end{figure}
123:
124: With this collection of stimulating results on flux phases in
125: the Hubbard and $t-J$ models, it would be interesting to study a
126: system whose geometry more closely resembles that of the copper-oxides.
127: This is especially relevant in the light of the debate
128: regarding the suitability of the $t-J$
129: model to the copper-oxide system and in particular
130: the importance of an explicit treatment of the oxygen atoms
131: \cite{dagotto}.
132: There is now general agreement in the community that
133: the $CuO_2$ planes play a crucial role in the high $T_c$
134: superconductors\cite{dagotto} and that the geometry
135: of the copper-oxides is relevant to superconductivity.
136: Further, several models
137: with effective interactions have been proposed to describe the
138: $CuO_2$ plane\cite{varma3,emery3,dagotto}. These studies
139: have emphasized the role of strong electronic correlations. Mean-field
140: studies\cite{varma} have indicated that it is necessary to go
141: beyond on-site interactions and an explicit treatment of the
142: nearest-neighbour interactions is necessary to stabilize the
143: circulating-current phase. An accurate treatment of the
144: strongly correlated full 2D $CuO_2$ system, though desirable,
145: is not presently forthcoming. Therefore, we have opted for a
146: $CuO_3$ system that we believe to incorporate the crucial geometric
147: ingredients of the system, including an explicit treatment
148: of oxygen sites, which can then be studied by the
149: highly accurate DMRG method\cite{white}.
150:
151: \begin{figure}
152: \begin{center}
153: \epsfxsize=0.8\columnwidth
154: \epsfbox{cuoext_new.eps}
155: \end{center}
156: \caption{
157: (a) Single unit cell of the $CuO_3$ system studied, the copper site is marked
158: by a filled circle and all vertices correspond to oxygen sites.
159: The phase conventions of the transfer
160: integrals, corresponding to the symmetry of the $d$- and $p$-orbitals
161: are indicated (b) Model 1 : 6 unit cells of the
162: extended $CuO_3$ system with
163: $t_{\parallel} = 0$
164: (c) Model 2 : similar to (b) with
165: $t_{\parallel} \ne 0$.
166: }
167: \end{figure}
168:
169:
170:
171:
172:
173: In this paper we study currents in a 3-band extended Hubbard
174: Hamiltonian on a
175: $CuO_3$-type system (Fig. 2), which is believed to model the essential
176: aspects of the 2D system. It is possible to envisage several quasi-1D
177: sections of the 2D $CuO_2$ plane. It is necessary to choose geometries that
178: are capable of supporting circulating currents. The plaquette that
179: forms a single unit cell of the $CuO_2$ plane is shown in Fig. 2a.
180: The system shown in Fig. 2b, henceforth to be called Model 1, consists
181: of such plaquettes connected at a single point ($t_{\parallel}$ = 0).
182: A strip of the
183: $CuO_2$ plane is shown in Fig. 2c. In this system, henceforth known as
184: Model 2, links parametrized by $t_{\parallel}$ have been added
185: between adjacent plaquettes to model the 2D nature of the
186: system. This is expected to better correspond to scenario in the
187: $CuO_2$ planes where
188: the plaquettes are all coupled due to hopping along both axes.
189:
190: The essential difference between models 1 and 2 is that since the
191: plaquettes of Model 1 are connected at just a single point, it is
192: possible to globally reverse the signs of the currents in each unit-cell
193: completely independently of all the other unit-cells, without any energy
194: cost. Therefore, while each unit-cell is a-priori capable of sustaining
195: currents, long range order is not possible in Model 1.
196: Thus, it would be necessary to include links between these plaquettes
197: (present in the physical system), a scenario which corresponds to Model 2.
198: In this paper, we have examined both these models in detail in order to study the
199: effects of interactions locally in a single unit cell as well as the
200: long-range effects brought about by coupling plaquettes as in Model 2.
201:
202: For weak interactions, ladder-type systems can be treated using
203: a combination of bosonization and perturbative renormalization group
204: techniques, as in the case of the 2-leg ladder (with disorder)\cite{orignac}
205: or N-leg Hubbard ladder\cite{balents}. In fact, abelian bosonization
206: methods have been used to study staggered currents in the 2-leg
207: ladder\cite{marston1}. These powerful analytical tools have proved
208: useful in extracting a great deal of information.
209: However, these methods cannot fully take into account the
210: exact values of the
211: parameters as well as the precise geometry of the system, both
212: of which are crucial to the copper-oxides as discussed above.
213: While our system appears to be
214: a quasi-1D system, it has a highly non-trivial geometry and
215: up to $4^{th}$ neighbours are connected.
216: Given the large number of parameters in the 3-band Hubbard
217: model and the importance of the delicate balance between them,
218: we have chosen to use the highly accurate DMRG method which gives
219: information even in the strong-coupling regime
220: inaccessible to perturbative approaches.
221:
222: We have used a 3-band extended Hubbard model on the systems in
223: Fig. 2, described
224: in greater detail below.
225: The 3-band Hubbard model for clusters of the $CuO_2$ system
226: (up to $6 \times 6$ unit-cells) has
227: been studied by the constrained path quantum Monte Carlo method (CPMC)
228: \cite{gubernatis1,gubernatis2}.
229: However, the focus of
230: the CPMC studies
231: was pair-binding of holes as well as the pairing and other
232: correlation functions, but not currents.
233: Thus, a study of flux-type phases or currents in this model does not exist, to
234: our knowledge.
235:
236:
237: \section{Model and Method}
238:
239:
240: The 3-band Hamiltonian for our $CuO_3$ system in the hole picture is given by:
241: %\begin{equation}
242: %\begin{array}{lll}
243: \begin{eqnarray}
244: & & {} H = H_0 + H_1 + H_2 \\
245: & & {} H_0 = \Big[ \sum\limits_{i,\sigma} -t_{pd} d^{\dagger}_{i\sigma}
246: (p_{x,i+a/2,\sigma}-p_{x,i-a/2,\sigma} -
247: p_{y,i+a/2,\sigma} \nonumber \\
248: & & {} +p_{y,i-a/2,\sigma}) -
249: t_{pp}p^{\dagger}_{x,i+a/2,\sigma}
250: (-p_{y,i+a/2,\sigma}+p_{y,i-a/2,\sigma})- \nonumber \\
251: & & {} t_{pp}p^{\dagger}_{x,i-a/2,\sigma}
252: (p_{y,i+a/2,\sigma}-p_{y,i-a/2,\sigma}) \nonumber \\
253: & & {} +\sum\limits_{\langle ij \rangle \sigma}
254: t_{\parallel} (p^{\dagger}_{y,i+a/2,\sigma}
255: p_{y,j+a/2,\sigma} +
256: p^{\dagger}_{y,i-a/2,\sigma} p_{y,j-a/2,\sigma}) \Big] \nonumber \\
257: & & {} + h.c. +
258: \sum\limits_{i\mu\sigma} (\epsilon_p n^i_{p_{x(y),\mu}\sigma}+
259: \epsilon_d n^i_{d\sigma}) \nonumber \\
260: & & {} H_1 =
261: U_d\sum\limits_{i\sigma}n^i_{d\sigma}n^i_{d{\bar\sigma}}+
262: U_p\sum\limits_{i\sigma\mu}n^i_{p_{x(y),\mu}\sigma}n^i_{p_{x(y),\mu}{\bar\sigma}} \nonumber \\
263: & & {} H_2 = V_{pd} \sum\limits_{i \mu}
264: n^i_{d} (n^i_{p_{x,\mu}} + n^i_{p_{y,\mu}}) +
265: V_{pp} \sum\limits_{i \mu} n^i_{p_{x,\mu}}n^i_{p_{y,\mu}} \nonumber \\
266: & & {} +
267: V_{\parallel} \sum\limits_{\langle ij \rangle \mu} n^i_{p_{y,\mu}}
268: n^j_{p_{y,\mu}}
269: \nonumber
270: %\end{array}
271: \end{eqnarray}
272: %\end{equation}
273: where $d^{\dagger}_{i\sigma}$ creates a hole of spin $\sigma$
274: on the copper $d$-orbital at unit-cell $i$, $p_{x(y),i\pm a/2,\sigma}$
275: creates a hole of spin $\sigma$ at the oxygen $p_x$ ($p_y$) orbital at
276: sites $\mu = i \pm a/2$ in the $x$ and $y$ directions of unit-cell $i$.
277: Here $\langle ij \rangle $
278: corresponds to a sum over nearest-neighbour unit cells $i$ and $j$
279: and the summation over $ \mu$ corresponds to $i \pm a/2$.
280: Operator $n^i_{d\sigma}$ is the number operator for the copper atom
281: of unit cell $i$ and
282: $ n^i_{p_{x(y),\mu}\sigma}$ are the number operators for the
283: $p_x$ ($p_y$) orbitals at sites $\mu = i \pm a/2$ of unit cell $i$.
284: The parameters $\epsilon_d$ and $\epsilon_p$ are the site-energies
285: of the copper and oxygen sites ($\Delta = \epsilon_d - \epsilon_p$).
286: The copper-oxygen hopping is parametrized by $t_{pd}$ and
287: the oxygen-oxygen hopping by $t_{pp}$ while
288: $t_{\parallel}$ is the hopping between the $p_y$ orbitals
289: of neighbouring unit cells.
290: $U_d$ and $U_p$ are the on-site Hubbard repulsions at the
291: copper and oxygen sites, respectively.
292: The nearest-neighbour copper-oxygen repulsion
293: is parametrized by $V_{pd}$, the oxygen-oxygen repulsions within a
294: unit cell by $V_{pp}$ and the repulsions between $p_y$ orbitals
295: in neighbouring unit cells by $V_{\parallel}$.
296: Further, the phases of the hopping integrals $t$, corresponding to
297: the hybridization of $d$- and $p$-orbitals are represented
298: in Fig. 2.
299:
300:
301: We study this model by the DMRG method.
302: The complicated geometry (up to 4th neighbours are
303: connected) of this system required the implementation of a
304: 7-block scheme (Fig. 3). This choice of RG scheme
305: ensures that the relatively complicated geometry,
306: believed to be crucial
307: in the cuprates, is preserved at each step of the calculation.
308: In Fig. 3a, we see the breakup of the two unit-cell system
309: which is the starting point of our calculation.
310: The system size increases by one unit cell (4 sites) at each iteration.
311: The blocks marked '1' and '7' in the figure are renormalized at each
312: iteration and five new blocks (unrenormalized) of one site each
313: are added to the middle
314: of the system. The dimension of the Hilbert space of each of the added sites is 4,
315: corresponding to empty, one hole (up and down spins) and 2 holes.
316: In Fig. 3b, we represent the blocks and superblocks of
317: the three unit-cell system. The following iterations can be obtained
318: by replacing the superblocks '1' and '7' by the appropriate
319: system.
320: The DMRG method is a highly efficient way to study 1- and quasi 1-dimensional
321: strongly correlated fermionic systems. We are also not limited
322: to a choice of fillings that correspond to closed-shells, a
323: habitual constraint in quantum Monte Carlo simulations.
324:
325: \begin{figure}
326: \begin{center}
327: \epsfxsize=0.8\columnwidth
328: %\epsfysize=2.5in
329: \epsfbox{dmrg_1.eps}
330: \end{center}
331: \caption{The breakup of the extended system into blocks and superblocks used
332: in our 7-block DMRG scheme for (a) even iterations and (b) odd iterations. Copper
333: atoms are marked by filled circles and oxygen atoms by open circles.
334: The physical bonds are marked in dashed lines. The superblocks of each
335: iteration are marked in dot-dashed lines and also carry their
336: number (1 and 7). The five central blocks consist of one site each.
337: Superblock sizes increase
338: by two sites in each iteration and the superblock of the
339: following iteration is marked by solid lines. }
340: \end{figure}
341:
342: The parameters varied in our study are mainly the interaction strengths
343: ($V$'s), $\Delta$ and the filling.
344: We have studied this model with
345: $ t_{pp}$ = 0.6, $U_d$ = 10.0 and $U_p$ = 5.0 in units
346: where $t_{pd} = 1.0$. Further, we consider two different geometries
347: corresponding to values of $t_{\parallel} = 0$ and
348: $t_{\parallel} = 0.6$. The former corresponds to the
349: case of plaquettes that are coupled only at a point (Model 1)
350: and the
351: latter to the case of fully connected plaquettes (Model 2).
352: The choice of parameters for 3-band models of the copper-oxides
353: has been extensively discussed in the literature \cite{parameters}.
354: Our choice corresponds to quite a
355: realistic range of parameters for the copper oxides \cite
356: {gubernatis1,gubernatis2,parameters}.
357: In fact, one of the inherent difficulties associated with any
358: study of the copper-oxides is the
359: large number of parameters, giving rise to a
360: huge phase space. We have tried to vary the
361: different $V$'s in order to locate the interesting
362: part of the phase diagram. One of the difficulties encountered in
363: our calculations has been the inaccessibility of certain
364: ranges of $V$
365: due to convergence problems, as discussed in the end of Section IIIB.
366:
367:
368: We have studied up to 25 $CuO_3$ unit cells, corresponding to
369: 101 sites, at different dopings.
370: Half-filling corresponds to one hole per copper site
371: in the hole picture, where the system is known to be an antiferromagnetic
372: insulator for a wide range of parameter values. We have studied
373: systems with different numbers of holes, corresponding to
374: doping of the physical system.
375: We add two holes at every other iteration. Thus, the filling
376: is not the same at every step of the calculation. The cases
377: we discuss are systems with 20 - 25 unit cells.
378: We have studied systems of $N$ unit cells with
379: $n_h$ holes given by, $n_h = N + 3 + mod(N+1 , 2)$.
380: The filling $x$ is given by $ n_h / N - 1$. Thus, $x$ = 0
381: corresponds to the half-filled case.
382: We have studied fillings $x$ = 0.12, 0.13 and 0.19
383: ($N=25, 23, 21$) for odd numbers of unit cells and
384: $x$ = 0.17, 0.18 and 0.25 ($N = 24, 22, 20$) for
385: even numbers of unit cells. Thus, we are able to study a
386: range of doping values. In addition, we have studied
387: the case of $n_h = N + 1 + mod(N+1 , 2)$, for the model
388: with $t_{\parallel}$ = 0.
389:
390: We compared the ground state energies for several iterations
391: with the energies obtained from exact diagonalization calculations to
392: verify our RG scheme.
393: We varied the number of states $m$ on the superblocks (
394: $m$ = $125,150$ and $200$).
395: We have kept up to 200 states on the superblocks.
396: This correspond to a Fock space dimension $H_F = m^2 \times 4^5$ =
397: 40,960,000.
398: This provides
399: sufficient accuracy, estimated in several ways. The discarded
400: weight of the density matrix states is of order $10^{-6}$ with
401: 200 states. In certain cases, we have also calculated the
402: energies and properties of the non-interacting system (free-particle
403: model with all $U$'s and $V$'s set to zero), where exact results
404: are available. This provides a strong check on our data, since
405: convergence is known to be most difficult in this case for the
406: DMRG algorithm. Another property that we have used to check the
407: quality of our data is the magnetic moment $|\langle S^z_i \rangle | $,
408: at site $i$,
409: which ought to be zero in principle. Thus, only data which correspond to
410: relatively small magnitudes of $|\langle S^z_i \rangle | $ have been
411: used.
412:
413: \section{Results and Discussion}
414:
415: The physical property we have studied is the
416: current-current (CC) correlation function for links $k$ and $l$.
417: The current along a bond connecting sites $\mu$ and $\nu$ is defined as
418: $J_{\mu\nu} = it_{\mu\nu} \sum\limits_{\sigma} (c^{\dagger}_{\mu\sigma}
419: c_{\nu\sigma} - c^{\dagger}_{\nu\sigma}c_{\mu\sigma})$.
420: The mean-field analysis of the 3-band Hubbard
421: model indicated that the nearest-neighbour
422: repulsion $V$ is essential to stabilize the circulating-current
423: phase\cite{varma}.
424: The circulating-current phase shown in Fig. 1 would then
425: consist of currents represented by the
426: arrows in the figure. Since we work in the real representation
427: of the Hamiltonian, we use the current-current correlation
428: functions as a measure of this order.
429:
430:
431:
432:
433: We compared our data for the non-interacting model with
434: the exact data
435: available in this case to assure ourselves of
436: convergence.
437: The current for bond $l$ in unit-cell $i$
438: is denoted by $J^{i}(l)$, according to the conventions for
439: bond numbers shown in
440: Fig. 4. We refer to the vertical copper-oxygen bonds as rungs and
441: the horizontal copper-oxygen bonds as legs in analogy to the
442: terminology used for ladders.
443:
444: \begin{figure}
445: \begin{center}
446: \epsfxsize=\columnwidth
447: %\epsfysize=1.0in
448: \epsfbox{cuocc.eps}
449: %\epsfxsize=\columnwidth
450: %\epsfbox{ccceven.eps}
451: \end{center}
452: \caption{Conventions for current-current correlation functions studied for
453: systems with (a) odd and (b) even numbers of unit cells. Unit-cell
454: numbers $i$ are marked beneath and the numbers in the figures
455: correspond to the bond numbers. }
456: \end{figure}
457:
458:
459:
460: We have calculated the equal-time, zero-frequency current-current (CC)
461: correlation function, which is the ground state expectation value
462: $\langle J^{0}(k) J^{i}(l) \rangle$. These correlations are between
463: currents in the central
464: unit cell, denoted by $J^{0}(k)$ and currents on the other unit-cells
465: on one half of the system $J^{i}(l)$ ($i = 1, \ldots (N-1)/2$), for
466: systems with odd numbers of unit-cells.
467: Slightly different correlations are
468: calculated at even and odd iterations due to the DMRG scheme used
469: and these are pictured in Fig. 4.
470: For systems with even numbers of unit-cells,
471: we introduce the notation $\langle J^{0}(k) J^{i}(l) \rangle$,
472: and $\langle J^{0^{\prime}}(k) J^{i}(l) \rangle$
473: with $i = 1, \ldots (N-1)/2$, as shown in Fig. 4.
474: However, we continue to refer to both cases as $\langle J^{0}(k) J^{i}(l)
475: \rangle$ in the text, with the different notations
476: for even and odd unit-cells being implicit.
477: By the bilateral symmetry about the $O-Cu-O$ axis,
478: it is not necessary to take into account all the possible cases.
479: We therefore restrict ourselves to the case of $\langle J^{0}(k) J^{i}(l)
480: \rangle$, $k$,$l$ = 1, \ldots 3.
481: We analyse cases of
482: systems with 20-25 unit-cells, since these are sufficiently large
483: to study the decay of the current-current correlations.
484: In particular, we analyze the
485: behaviour of the different types of diagonal ($\langle J^{0}(k) J^{i}(k)
486: \rangle$ and off-diagonal $\langle J^{0}(k) J^{i}(l)\rangle$,
487: ($k \ne l$)
488: current-current correlations and try to classify them.
489:
490: \subsection{Analysis of Model 1}
491:
492: We first present our results for the CC correlations
493: on the $CuO_3$ system with Model 1 ($t_{\parallel} = 0$) which consists
494: of plaquettes coupled at a single point. We have performed calculations
495: with values of $t_{pd}$ and $t_{pp}$ equal to 1.0 and 0.6. We have varied
496: $\Delta$ and $V_{pd}$, while $V_{pp}$ was set to 0.
497: We have further considered two sets of fillings corresponding to
498: $n_h = N + 3 + mod(N+1 , 2)$ and $n_h = N + 1 + mod(N+1 , 2)$.
499: The second series then gives us fillings $x$ = 0.04, 0.043 and 0.048
500: ($N = 25, 23, 21$) and $x$ = 0.083, 0.09 and 0.1 ($N = 24, 22, 20$).
501: Thus, the second series $n_h = N + 1 + mod(N+1 , 2)$ has the
502: advantage of providing fillings that are practically constant
503: for different system sizes. This helps to verify that our conclusions
504: hold over different system sizes.
505:
506: \begin{figure}
507: \begin{center}
508: \epsfxsize=0.8\columnwidth
509: %\epsfysize=1.5in
510: \epsfbox{1a.eps}
511: \end{center}
512: \caption{ Current-current correlations $| \langle J^0 (k) J^i (l) \rangle |$
513: for a 25 unit-cell system with 28 holes, for Model 1,
514: in a log-log representation.
515: All systems have the parameters
516: $t_{pp} = 0.6$, $t_{\parallel} = 0$, $\Delta=1$
517: in units with
518: $t_{pd}$ = 1.
519: Circles represent the case with the above parameters and
520: all $V$'s = 0, squares are for $V_{pd} = 1$ and
521: diamonds are for $V_{pd} = 2$, all with $U_d = 10$ and $U_p = 5$,
522: obtained from DMRG calculations with $m = 200$ states.
523: Triangles represent the exact
524: values for the correlation for the non-interacting model
525: ($U_p = U_d = V_{pd} = V_{pp} = 0$), and inverted triangles
526: the values obtained for the same parameters
527: from $m = 200$ state DMRG calculations, presented for comparison.
528: Dashed lines are a guide to the eye while full
529: lines correspond to the power-law fits of the data.
530: (a) represents the $| \langle J^0 (2) J^i (1) \rangle |$
531: correlation with $\alpha_{NI}^* = -2.94$, $\alpha_{NI} = -3.12$,
532: $\alpha_0 = -3.09$, $\alpha_1 = -3.03$, $\alpha_2 = -3.09$
533: and (b) the $| \langle J^0 (2) J^i (3) \rangle |$ correlation
534: with
535: $\alpha_{NI}^* = -2.92$, $\alpha_{NI} = -3.11$,
536: $\alpha_0 = -3.16$, $\alpha_1 = -3.21$, $\alpha_2 = -3.37$.
537: See text for definitions of the $\alpha$s.
538: }
539: \end{figure}
540:
541:
542:
543: We have also carried out numerical fits on our data. For our results on
544: Model 1, we found our data well-fitted by power-laws. We define
545: $\alpha_{NI}$ as the slope corresponding to the non-interacting case
546: obtained from DMRG and $\alpha_{NI}^*$ as the slope of the exact data.
547: Likewise $\alpha_{k}, k = 0\ldots2$ correspond to the cases with
548: $V_{pd} = k$. These slopes give us the exponents of the
549: power-law.
550:
551: We have studied the effect of increasing the copper-oxygen repulsion
552: $V_{pd}$ on the current-current correlations. In Fig. 4a, we present
553: results for the dependence of the CC correlation function
554: $| \langle J^0 (2) J^i (1) \rangle |$ with distance $i$ for a 25 unit-cell
555: system with 28 holes. From Fig. 4, it can be seen that this is the
556: correlation of a rung of the central unit-cell with an oxygen-oxygen
557: bond ($l$ = 1) of the unit-cells away from the centre. We first
558: consider this correlation for the non-interacting case (all $U$'s and
559: $V$'s set to 0), where exact results are available. We have compared the
560: results obtained for $| \langle J^0 (2) J^i (1) \rangle |$ from
561: a DMRG calculation with $m = 200$ states with the exact results.
562: The convergence of the data is excellent, as seen in Fig. 5a.
563: The exponent of the DMRG data $\alpha_{NI} = -3.12$, compares
564: favorably with the exponent obtained for the exact data,
565: $\alpha_{NI}^* = -2.94$.
566: This is a strong check on our results, since the convergence
567: of the DMRG method is known to be harder for the free-particle
568: model than for the equivalent system with strong correlations.
569:
570: We now note from Fig. 5a that the curve which represents
571: current-current correlations for the
572: purely local Hubbard model with $U_d = 10, U_p = 5$ and $V$'s = 0,
573: lies above the non-interacting
574: curve. We then turn on nearest-neighbour interactions.
575: Increasing the copper-oxygen repulsion $V_{pd} = 1$
576: has the effect of increasing
577: the magnitudes of the current-current correlations. Further increasing
578: $V_{pd}$ to 2 increases the correlations, but by a smaller amount.
579: The decay of the correlations follows power-law behaviour,
580: as seen from the log-log plots.
581: Thus, we observe that
582: repulsive interactions are capable of increasing the
583: magnitude of the currents. It can be seen from the curves
584: that the decay at large-distances is hardly affected by the
585: $V$s, the curves are nearly parallel. This is borne out by the
586: slopes ($-3.09, -3.03$ and $-3.09$), which are practically unchanged.
587: This is because the plaquettes are coupled at a single point,
588: which prevents the formation of a state with long-range-order.
589:
590: In Fig. 5b, we present our results for the correlation
591: $| \langle J^0 (2) J^i (3) \rangle |$. This is a correlation between
592: a copper-oxygen rung of the central unit-cell, with a
593: copper-oxygen leg of the side unit-cells. Increasing the copper-oxygen
594: repulsion enhances the magnitude of the correlations compared to
595: the Hubbard model, while
596: their decay is power-law, again with similar exponents,
597: as is seen from the curves which are nearly parallel and from the
598: power-law fits.
599:
600: In Fig. 6a we present the behaviour of the correlation $| \langle J^0 (2) J^i
601: (2) \rangle |$, which is the copper-oxygen rung-rung correlation
602: function. The data presented are for the case of 25 unit-cells with
603: 26 holes, since the absolute magnitude of the correlations is
604: much larger than in the case of 28 holes. The qualitative behaviour remains
605: the same. We observe that the absolute values of the $V_{pd} = 2$
606: data are enhanced by almost an order of magnitude compared to the
607: $V_{pd} = 0$ curve. We have fitted our data with two series of power-laws
608: for even and odd sites in order to take into account the alternating behavior
609: of the curves. The slight differences between $\alpha_0$ and
610: $\alpha_1$ can be attributed to the rather large alternance
611: of the $V_{pd}$ = 0 data. This alternance is reminescent of the
612: generic behavior of correlation functions in quasi-1D systems,
613: due to the harmonics of $k_F$. We notice that the magnitude of the
614: oscillations decreases when $V$ is increased. In quasi-1D systems,
615: similar behavior of spin-spin and charge-charge correlations is
616: interpreted as an increase in the value of the charge-stiffness
617: ($K_{\rho}$)\cite{haldane}.
618:
619: \begin{figure}
620: \begin{center}
621: \epsfxsize=0.8\columnwidth
622: %\epsfysize=1.5in
623: \epsfbox{1b.eps}
624: \end{center}
625: \caption{ Current-current correlations for a 25 unit-cell system for
626: Model 1. Note the log-log presentation.
627: All systems have the parameters
628: ($t_{pp} = 0.6$, $t_{\parallel} = 0$, $\Delta=1$, $U_d = 10$, $U_p = 5$).
629: Circles represent the case with the above parameters and
630: all $V$'s = 0, squares are for $V_{pd} = 1$ and
631: diamonds are for $V_{pd} = 2$.
632: Dashed lines are a guide to the eye, while full
633: lines represent the power-law fits.
634: (a) $| \langle J^0 (2) J^i (2) \rangle |$
635: for a 25 unit-cell system with 26 holes with
636: ($\alpha_0 = -3.34$ (even) and $-3.94$ (odd));
637: ($\alpha_1 = -4.00$ (even) and $-4.18$ (odd)).
638: and (b) $| \langle J^0 (3) J^i (3) \rangle |$
639: for a 25 unit-cell system with 28 holes with
640: ($\alpha_0 = -2.40$, $\alpha_1 = -2.69$, $\alpha_2= -3.00$).
641: }
642: \end{figure}
643:
644: The behaviour of the correlations $| \langle J^0 (3) J^i (3) \rangle |$
645: can be seen in Fig. 6b. This correlation is a copper-oxygen leg-leg
646: correlation function, as seen from Fig. 4. Here, the behaviour
647: observed is different from the previous cases studied, since the
648: correlation is enhanced very slightly
649: by increasing repulsions for first neighbours.
650: However, the situation is reversed at very short distances
651: and increasing repulsions causes
652: the correlations to decay even faster, as can be noted from the
653: crossing of the curves in Fig. 6b. Their decay is power-law
654: with the exponents decreasing, as can be seen from the figure.
655:
656: We have also analysed the behaviour of correlation functions
657: of the type $| \langle J^0 (1) J^i (1) \rangle |$. These
658: are not presented for reasons of space. These correlations
659: are oxygen-oxygen correlations and decay as power-laws, with the magnitudes
660: falling as $V_{pd}$ is increased. This however, could be due to
661: the value of $V_{pp}$ = 0, chosen here.
662:
663:
664:
665: We now briefly discuss the case
666: of $x = 0.18$ (22 unit-cells with 26 holes), which
667: fits into the above analysis with an interesting
668: reversal. In this case, the rung-rung correlations
669: (indeed, correlations involving one rung) are
670: initially decreased when $V_{pd}$ is turned on. Increasing
671: $V_{pd}$ to 2 brings about a reversal of this behaviour, with
672: the magnitude of the correlations increasing compared to the
673: $V_{pd}$ = 0 case.
674: In Fig. 7 we present one example of this behaviour
675: with the correlation $| \langle J^{0^{\prime}} (2)
676: J^i (3) \rangle |$, which is a rung-leg correlation.
677: The reversal of the trend is clearly
678: seen from the $V_{pd}$ = 1 curve (lies below the $V_{pd}$ = 0 curve)
679: and the $V_{pd}$ = 2 data which is enhanced with
680: respect to the $V_{pd}$ = 0 curve. This indicates that there
681: could be a threshold value of nearest-neighbour interactions
682: $V$ for enhancing the magnitude of current-current
683: correlations.
684:
685: \begin{figure}
686: \begin{center}
687: \epsfxsize=0.8\columnwidth
688: %\epsfysize=1.5in
689: \epsfbox{2.eps}
690: \end{center}
691: \caption{ Current-current correlations
692: $| \langle J^{0^{\prime}} (2) J^i (3) \rangle |$
693: for a 22 unit-cell system with 24 holes. Note the
694: log-log representation. All systems have the parameters
695: ($t_{pp} = 0.6, t_{\parallel} = 0$, $\Delta=1$, $U_d = 10$, $U_p = 5$)
696: Circles represent the case with the above parameters and
697: all $V$'s = 0, squares are for $V_{pd} = 1$,
698: diamonds for $V_{pd} = 2$. Lines represent power-law fits
699: with exponents $\alpha_0 = -2.41$, $\alpha_1 = -2.69$ and
700: $\alpha_2 = -3.00$.
701: }
702: \end{figure}
703:
704: Thus, the above analyses show that the copper-oxygen rung-rung
705: correlation (or even correlations involving one rung)
706: are enhanced in magnitude by repulsive interactions,
707: compared to the purely local Hubbard model, probably requiring a minimum
708: threshold value of $V$.
709: The copper-oxygen leg-leg correlations show an initial
710: increase followed by a decay in magnitude and correlation
711: length, as $V_{pd}$ is increased. The oxygen-oxygen
712: correlation is diminished in magnitude and correlation
713: length, as $V_{pd}$ is increased. We have analysed the
714: cases of different fillings described above and have
715: found similar behaviour in all these cases.
716:
717: \subsection{Analysis of Model 2}
718:
719: We now turn to an analysis of Model 2 with $t_{\parallel} = 0.6$,
720: The behavior of this model is expected to be quite different from that
721: of the $t_{\parallel} = 0$ case. It is expected that the
722: long-distance behavior of the CC correlations is much more
723: sensitive to the nearest-neighbour repulsions $V$.
724: Here, we first study the purely local Hubbard model on our lattice
725: ($U_d = 10, U_p = 5$ and $V's = 0$). We note that this corresponds to
726: the system in Fig. 2c, with bonds connecting different plaquettes
727: ($t_{\parallel} = 0.6$). However, the $V$ on this bond is initially
728: set to zero, to give us a model of connected plaquettes
729: with purely local interactions. We the begin to turn on $V$s,
730: i.e we turn on copper-oxygen
731: and oxygen-oxygen nearest-neighbour interactions, along with
732: the interaction $V_{\parallel}$ ( $V_{pd} = V_{pp} = V_{\parallel} = 1.0$).
733: Thereafter, we further increase $V_{\parallel}$ to study the
734: effect of longer-range interactions on the current-current correlations.
735:
736:
737:
738: We shall first consider the case of 25 unit cells at a filling of 0.12
739: in some detail before analysing our other data.
740: In Fig. 8a - 8c we present diagonal correlations of the type
741: $ |\langle J^0 (k) J^i (k) \rangle |$,
742: for $k$ = 1, \ldots 3, on a log-log scale.
743: These correspond respectively to the oxygen-oxygen diagonal correlation, the
744: rung-rung correlation and the leg-leg correlation of bonds in the central
745: unit cell with bonds in the side unit cells.
746: In Fig. 8a, we present the decay of the oxygen-oxygen diagonal correlation
747: ($ |\langle J^0 (1) J^i (1) \rangle | $) with increasing
748: distance for increasing values of the nearest-neighbor repulsion $V$.
749: These are compared to the data
750: with all $V$'s = 0, i.e. the purely local Hubbard model.
751: From the data in Fig. 8a, we note the effect of repulsions on
752: the correlations
753: between currents on oxygen-oxygen bonds.
754: Turning on $V$'s, ($V_{pd} = V_{pp} = V_{\parallel} = 1$) causes the
755: correlations to decay in magnitude
756: compared to the Hubbard model. Further increasing
757: $V_{\parallel}$ to 3, causes the correlations to decay further. Data for
758: $V_{pd} = V_{pp} = 1, V_{\parallel} = 2$, fit smoothly into this
759: pattern and have not been shown, for clarity. The
760: correlations show exponential decay as can be seen from the
761: plots as well as the inset. We have fitted the Hubbard model
762: curve and the $V_{\parallel} = 3$ data (scaled by a factor of 10)
763: by expressions of the form
764: $ y = A\,exp(\beta x) $, shown in the inset. The clear straight
765: lines seen in the semilog plot are indicative of the quality
766: of the fit. We obtain
767: $A = -6.0065$ and $\beta = -0.6296$ for the Hubbard model ($V$s $= 0$) and
768: $A = -3.3280$ and $\beta = -0.7753$ for the $V_{\parallel} = 3$ data.
769:
770: \begin{figure}
771: \begin{center}
772: \resizebox{7.5cm}{10cm}{\includegraphics{251.eps}}
773: \end{center}
774: \caption{
775: Diagonal current current correlations
776: for a 25 unit-cell system with 28 holes
777: ($|\langle J^0 (k) J^i (k)\rangle |$)
778: in a log-log representation.
779: All systems have the parameters
780: ($t_{pp} = t_{\parallel} = 0.6$, $\Delta=1$, $U_d = 10$, $U_p = 5$)
781: in units with
782: $t_{pd}$ = 1. Circles represent the case with the above parameters and
783: all $V$'s = 0, squares are for $V_{pd} = V_{pp} = V_{\parallel} = 1$,
784: triangles for $V_{pd} = V_{pp} = 1, V_{\parallel} = 3$.
785: Triangles are in some cases filled, for clarity.
786: Lines in the main figures are meant to be a guide to the eye.
787: Lines in the insets correspond to fits which are discussed in greater
788: detail in the text.
789: From top to bottom, (a) diagonal oxygen-oxygen correlation function
790: $ |\langle J^0 (1) J^i (1) \rangle| $. Inset (semilog representation)
791: contains $V_{\parallel}$ = 3 data multiplied by a factor 10, for visibility.
792: (b) rung-rung correlation $ |\langle J^0 (2) J^i (2) \rangle| $. The inset
793: (log-log plot)
794: contains $V_{\parallel}$ = 3 data multiplied by a factor 100.
795: (c) leg-leg correlation $ |\langle J^0 (3) J^i (3) \rangle| $.
796: The inset (log-log plot)
797: contains $V_{\parallel}$ = 3 data multiplied by a factor 10.
798: }
799: \end{figure}
800:
801: In Fig. 8b we present
802: cor\-re\-lations of type
803: $ |\langle J^0 (2) J^i (2) \rangle | $, which correspond
804: to the copper-oxygen rung-rung correlation,
805: for different values of interaction strengths.
806: Here, we notice a sizeable enhancement of the
807: current-current correlations (by an order of magnitude)
808: over distances of a few lattice
809: spacings as interactions ($V$'s) are turned on, compared to the
810: Hubbard model.
811: Thereafter, interactions cause
812: the correlations to fall, with respect to the $V$'s = 0 case.
813: However, further increasing $V_{\parallel}$
814: causes the correlations to rise i.e. there is a
815: reversal of the trend.
816: It is reassuring that the DMRG data in these cases
817: are well converged, as seen from the small discarded weight and
818: the small
819: magnetic moments at the central sites, $ | S^z_i | \le 4 \times 10^{-4} $.
820: These data have been fitted by power-laws. The fits are good
821: as seen from the straight-lines in the log-log plots, both
822: in the main figure and in the inset (thick lines correspond to the
823: fits). The slope of the straight line
824: gives the exponent of the power-law and we obtain
825: $\alpha_0 = -4.4070$ and $\alpha_3 = -4.6830$.
826: We also note that the cross-rung-rung correlation, of type
827: $ |\langle J^0 (2) J^i (4) \rangle| $ are similar
828: qualitatively and quantitatively to the rung-rung correlations
829: and are not presented here.
830:
831: In Fig. 8c we present
832: correlations of type $ |\langle J^0 (3) J^i (3) \rangle | $, which correspond
833: to the copper-oxygen leg-leg correlation,
834: for different values of interaction strengths.
835: Here, we notice a slight enhancement of the
836: current-current correlations over distances of a few lattice
837: spacings as interactions ($V$'s) are turned on, compared to the
838: Hubbard model.
839: Thereafter, interactions cause
840: the correlations to fall, with respect to the $V$'s = 0 case,
841: upon increasing $V$s.
842: The long-distance behaviour in this case is interesting. The
843: $V = 0$ data (Hubbard model) are clearly exponential as seen
844: from the main figure and the inset
845: ($A = -3.9204$ and $\beta = -0.6254$). On the other
846: hand the $V_{\parallel} = 3$ data (triangles) show much more
847: complex behaviour corresponding to power-law behavior at short
848: distances
849: (up to 6 sites) followed by an exponential tail at long distances.
850: The power-law fit is indicated in the inset by the thin line and
851: the exponential tail by the thick line. We
852: obtain the power-law exponent $\alpha_3 = -2.6299$ from the slope
853: of the power-law part of the fit and
854: $A = -0.0143$ and $\beta = -0.8947$ for the exponential part
855: (thick line).
856:
857:
858:
859: We now turn to an analysis of the off-diagonal correlations.
860: Given the large number of correlations, we attempt to
861: set out the general trends by treating one off-diagonal
862: correlation of each type.
863: In Fig. 9a - Fig. 9c, we present off-diagonal CC correlations,
864: $ |\langle J^0 (1) J^i (2) \rangle |$,
865: $ |\langle J^0 (3) J^i (2) \rangle |$ and
866: $ |\langle J^0 (3) J^i (1) \rangle |$.
867: The first of these corresponds to the correlation of an
868: oxygen-oxygen bond in the central unit-cell with
869: copper-oxygen rungs of the side unit cells. We refer
870: to this henceforth as an "oxygen-rung" correlation.
871: As seen from the log-log plot of Fig. 9a, the correlation (1,2)
872: exhibits power-law decay both for the Hubbard model and upon increasing
873: $V$s. There is a substantial increase of this correlation
874: at short distances with increasing $V$. This trend persists
875: over long distances for the case of $V_{\parallel} = 3$ as seen
876: from the figure. The power-law fits to these data are presented
877: in the inset, with the $V=3$ data scaled by a factor 10.
878: We have fitted the Hubbard model data in
879: two series corresponding to the odd and even points to take care
880: of the alternance present in our data and obtain
881: $\alpha_0$s of $-3.9408$ (odd) and $-3.3321$ (even). The
882: $V_{\parallel} = 3$ data were fitted by a power-law with exponent
883: $\alpha_3 = -4.0923$. We note that the alternance of the
884: $V_{\parallel} = 3$ is strongly diminished with respect to
885: the $V = 0$ data.
886:
887: Similar behavior is observed in the leg-rung correlation function
888: $ |\langle J^0 (3) J^i (2) \rangle |$, presented in Fig. 9b.
889: The $V_{\parallel} = 3$ data are clearly enhanced with respect to the
890: Hubbard model data almost over all the length of the curve.
891: We followed a procedure similar to the previous case, fitting
892: the $V = 0$ data by power-laws in two series
893: to obtain $\alpha_0$s of $-3.0559$ (odd) and $-4.0292$ (even).
894: The power-law fits to the $V_{\parallel} = 3$ data gave us an
895: exponent $\alpha_3$ of $-4.1966$. Correlations quench the
896: even-odd alternation of the data.
897:
898: \begin{figure}
899: \begin{center}
900: \resizebox{7.5cm}{10cm}{\includegraphics{252.eps}}
901: \end{center}
902: \caption{
903: Off-diagonal current current correlations
904: $| \langle J^0 (k) J^i (l) \rangle |$ ($k \ne l$)
905: for a 25 unit-cell system with 28 holes in a log-log representation.
906: All systems have the parameters
907: ($t_{pp} = t_{\parallel} = 0.6$, $\Delta=1$, $U_d = 10$, $U_p = 5$)
908: in units with
909: $t_{pd}$ = 1. Circles represent the case with the above parameters and
910: all $V$'s = 0, squares are for $V_{\parallel} = V_{pp} = V_{\parallel} = 1$,
911: triangles for $V_{pd} = V_{pp} = 1, V_{\parallel} = 3$.
912: Symbols are in some cases filled, for clarity.
913: Lines in the main figures are meant to be a guide to the eye.
914: Lines in the insets correspond to fits which are discussed in greater
915: detail in the text.
916: From top to bottom, (a) "oxygen-rung" correlation
917: $ |\langle J^0 (1) J^i (2) \rangle| $. Inset (log-log)
918: contains $V_{\parallel}$ = 3 data multiplied by a factor 10, for visibility.
919: (b) leg-rung correlation $ |\langle J^0 (3) J^i (2) \rangle| $. The inset
920: (log-log plot)
921: contains $V_{\parallel}$ = 3 data multiplied by a factor 10.
922: (c) "leg-oxygen" correlation $ |\langle J^0 (3) J^i (1) \rangle| $.
923: The inset (semilog plot)
924: contains $V_{\parallel}$ = 3 data multiplied by a factor 10.
925: }
926: \end{figure}
927:
928:
929: In Fig. 9c, we present the correlation
930: $ |\langle J^0 (3) J^i (1) \rangle |$, which corresponds to the
931: correlation of the central oxygen-oxygen bond with the
932: copper-oxygen leg of the side unit cell. We refer to this
933: correlation henceforth as the "oxygen-leg" correlation. We
934: observe from the figure and from the fits in the inset that
935: these correlations have exponential behavior for
936: the Hubbard model as well as for the case with $V_{\parallel}=3$.
937: From the fits seen in the inset (presented in semilog form)
938: we obtain $A = -4.61476$ and $\beta = -0.6600$ for the
939: $V = 0$ case and $A = -1.849094$ and $\beta = -0.8081$ for the
940: $V_{\parallel} = 3$ case.
941:
942: \begin{figure}
943: \begin{center}
944: \epsfxsize=0.8\columnwidth
945: %\epsfbox{22rr6_final.eps}
946: \resizebox{7.0cm}{9cm}{\includegraphics{22rr_new.eps}}
947: \end{center}
948: \caption{ Current-current correlations
949: $| \langle J^{0^{\prime}} (2) J^i (k) \rangle |$
950: for a 22 unit-cell system with 26 holes. All systems have the parameters
951: $t_{pp} = t_{\parallel} = 0.6$, $\Delta=1$, $U_d = 10$, $U_p = 5$
952: in units with $t_{pd}$ = 1.
953: Circles represent the case with the above parameters and
954: all $V$'s = 0, squares are for $V_{pd} = V_{pp} = V_{\parallel} = 1$
955: and
956: triangles for $V_{pd} = V_{pp} = 1, V_{\parallel} = 3$.
957: Triangles are in some cases filled, for clarity.
958: Lines in the main figures are meant as a guide to the eye and lines
959: in the insets are the power-law fits to the data.
960: In both insets, $V_{\parallel} = 1$
961: data (squares) is multiplied by a factor of 5 and $V_{\parallel} = 3$
962: data (triangles) by a
963: factor of 25 for clarity. From top to bottom,
964: (a) rung-leg correlation, $| \langle J^{0^{\prime}} (2) J^i (3) \rangle |$.
965: From the fits, $\alpha_0 = -4.4157$, $\alpha_1 = - 4.3648$ and
966: $\alpha_2 = -3.5904$.
967: (b) another rung-leg correlation
968: $| \langle J^{0^{\prime}} (2) J^i (5) \rangle |$.
969: From the fits, $\alpha_0 = -4.9475$, $\alpha_1 = - 4.7718$ and
970: $\alpha_2 = -4.3446$.
971: }
972: \end{figure}
973:
974:
975: In addition to the data presented in
976: Figs.~8-9, for the 25 unit-cell system with filling $x = 0.12$,
977: we discuss some results for the case of 22 unit-cells with
978: $x$ = 0.18.
979: In Fig. 10, we present the evolution of
980: two types of rung-leg correlations
981: that are similar:
982: $ |\langle J^{0^{\prime}} (2) J^i (3) \rangle| $
983: and $| \langle J^{0^{\prime}} (2) J^i (5) \rangle |$, henceforth
984: referred to as the $(2^{\prime},3)$ and $(2^{\prime},5)$
985: correlations respectively.
986: In Fig. 10a, we see that the $V_{\parallel} = 1$ data of the
987: $(2^{\prime},3)$ correlation
988: is already enhanced with respect to
989: the Hubbard model data over intermediate distances. The
990: $V_{\parallel} = 3$ data is clearly enhanced with respect to the
991: Hubbard model data. We have fitted these data by means
992: of power-laws, as seen from the log-log scale of the curve.
993: The results of the fits are seen from the inset. We see that $\alpha_3$,
994: the exponent of the $V_{\parallel} = 2$ curve is visibly reduced to
995: $-3.59$ compared to $\alpha_0 = -4.94$ for the Hubbard model
996: and the $V_{\parallel} = 3$ data falls off much more slowly than the
997: Hubbard model data.
998: As seen from the inset, the number of points we were able to
999: include in the fit decreases with increasing $V$, as
1000: compared to the Hubbard model.
1001: It is accepted that
1002: convergence of the DMRG algorithm is best in the presence
1003: of strong, local interactions. Our Hubbard model case falls
1004: into this category with $U_d = 10$ and $U_p = 5$. Increasing
1005: $V$ clearly renders convergence much more difficult, especially
1006: when $V_{\parallel} = 3$ which is most probably the reason
1007: for this behavior.
1008:
1009:
1010:
1011: In Fig. 10b, we present in a manner analogous to Fig. 10a, data
1012: for the $(2^{\prime},5)$
1013: correlation functions. As in the previous case, we note that
1014: the $V_{\parallel} = 3$ data is stabilized compared to the
1015: $V = 0$ data and the $V_{\parallel} = 1$ data, with the exponents
1016: $\alpha_0 = -4.9475$, $\alpha_1 = -4.7718$ and $\alpha_3 = -4.3446$.
1017: Thus, we note that the decay of the $(2^{\prime},5)$ correlation
1018: is slower for $V_{\parallel} = 3$ than for the Hubbard model. We have
1019: dropped 3 points from the $V_{\parallel} = 3$ curve and 1 point from
1020: the Hubbard model curve to obtain these results.
1021:
1022:
1023: Thus, the analysis of the diagonal and off-diagonal
1024: CC correlation functions
1025: of Model 2 yields the following trends : the nearest-neighbor
1026: repulsive interaction $V$ is clearly capable of increasing the
1027: magnitude of the current-current correlations locally and
1028: of strongly modifying the long-distance behavior.
1029: The correlations fall into different categories.
1030:
1031: The rung-rung correlation
1032: function is enhanced by increasing $V$s and shows power-law decay.
1033: In certain cases, the exponent of the power-law decreases upon
1034: increasing $V$ and the correlations fall off slowly as
1035: compared to the Hubbard model.
1036: In fact, this extends to any correlation that includes at least
1037: one rung, diagonal or off-diagonal.
1038:
1039: The leg-leg correlations show exponential decay in the Hubbard model.
1040: However, increasing the values of $V$s and $V_{\parallel}$ in particular
1041: changes this behaviour - we now have a power-law behavior at
1042: short distances followed by an exponential tail. Thus, we see
1043: that $V$s are capable of locally enhancing the CC correlations and
1044: by connecting the plaquettes, the long-distance behaviour is
1045: greatly modified.
1046: We believe that the difference between the
1047: rung-rung and leg-leg correlations is due to our choice of model
1048: and that this difference should disappear in the 2D system.
1049:
1050: The oxygen-oxygen correlation function shows exponential decay
1051: with very slight enhancement of the absolute values for $V_{\parallel} = 3$.
1052: This is also true of the "oxygen-leg" type correlation functions.
1053: This could however be due to our choice of parameter values.
1054: Clearly, it would be interesting to study other parameter
1055: ranges for the Vs, especially for $V_{pd}$ and $V_{pp}$. This
1056: point is discussed further down in this section.
1057:
1058: We have studied various cases involving other parameter values. Notably,
1059: we tried to increase $V_{pd}$ and $V_{pp}$
1060: ($V_{pd}$ = 2.0, 3.0, $V_{pp}$ = 2.0, in units of $t_{pd}$),
1061: but encountered convergence problems at the level of the Lanczos
1062: diagonalization algorithm. It is quite likely that
1063: increased degeneracy of the ground state
1064: slows down convergence, as is known to occur in other cases
1065: such as the $J_1 - J_2$ spin chain, at parameter values close
1066: to the critical point.
1067: This is most probably due to the increased delocalization of the
1068: ground state, brought about by larger $V$'s.
1069: To overcome this problem, it would be necessary to increase
1070: the number of states $m$ that we keep on
1071: the superblocks.
1072: However, at our present
1073: value of $m$ = 200, we are already at the limit of our computational
1074: resources.
1075: The charge densities in our
1076: systems show that the charges tend to be concentrated
1077: on the copper atoms, which possibly competes with the
1078: stabilization of currents. Thus, increasing $V_{pd}$ and
1079: $V_{pp}$ is likely to bring about a rehomogenization
1080: of charge and this could be favourable to the current-current correlations.
1081:
1082: At this juncture, it is interesting to compare the models with
1083: $t_{\parallel} = 0$ and 0.6. We have observed enhancement in
1084: the magnitudes of the currents in both models. Thus, repulsive
1085: nearest-neighbour interactions are clearly capable of increasing
1086: the magnitudes of currents. The long-distance behaviour of
1087: correlations is different in the two models.
1088: Model 1 shows power-law decay of correlations and the interaction $V$ has
1089: a very small effect on the exponents.
1090: In Model 2, any correlation involving a rung
1091: shows power-law behaviour. We have very interesting indications
1092: that repulsive nearest-neighbour interactions
1093: can enhance currents over distances comparable to the length scale
1094: of the system. In certain cases, the exponent is
1095: considerably reduced and the current-current correlations
1096: stabilized compared to the Hubbard model.
1097:
1098: \section{Conclusions}
1099: In conclusion, we have studied a 3-band extended Hubbard
1100: Hamiltonian on a model $CuO_3$ system, with geometry
1101: similar to that of a section of the $CuO_2$ planes of the
1102: copper-oxides. We have studied two different cases, with
1103: the parameter $t_{\parallel} = 0$ and 0.6,
1104: which correspond respectively to Model 1 with plaquettes
1105: coupled at a single point and Model 2, with plaquettes
1106: coupled all along the length of the system.
1107: Our calculations show that
1108: nearest-neighbour repulsive interactions are clearly
1109: capable of enhancing current-current correlations
1110: in magnitude and of qualitatively changing the long-distance behavior
1111: of the system.
1112: Model 1 shows power-law decay of correlations
1113: and the interaction $V_{pd}$ has
1114: a very small effect on the exponents.
1115: Increasing the copper-oxygen repulsion $V_{pd}$ can
1116: increase the magnitude of the rung-rung current-current correlation
1117: (or correlations involving a single rung)
1118: by up to an order of magnitude.
1119: In Model 2,
1120: the effect of introducing
1121: nearest-neighbour interactions $V$ in the system is to produce a
1122: sizeable increase in the rung-rung correlations over distances
1123: of a few lattice spacings.
1124: $V_{\parallel}$ can enhance correlations involving at least one
1125: copper-oxygen rung over distances comparable to
1126: the length of the system. In certain cases, interactions bring about
1127: a sizeble decrease in the exponent, causing the correlations to
1128: decay slowly with respect to the Hubbard model.
1129: We have indications that threshold values of the nearest-neighbour
1130: repulsions are necessary to observe stabilization at long
1131: distances. Tuning the
1132: copper-oxygen and the oxygen-oxygen repulsions would
1133: redistribute charge more uniformly in the system which is likely to
1134: stabilize the current-current correlation functions in this
1135: system. We believe that the enhancement and stabilization of
1136: certain correlation functions seen in our model constitutes a
1137: trend that will hold for the 2D system.
1138:
1139:
1140: \noindent
1141: {\bf Acknowledgements}: We thank the IDRIS, Orsay for time on the NEC SX-5
1142: and the CICT, Toulouse for time on CALMIP.
1143: We thank C. J. Calzado, J-P. Malrieu and C. M. Varma
1144: for useful discussions.
1145:
1146:
1147: \vskip -0.5cm
1148: \begin{thebibliography}{99}
1149: \bibitem{timusk} for a recent experimental survey of the pseudogap
1150: phase, see, T. Timusk and B. Stratt, Rep. Prog. Phys.
1151: {\bf 62}, 61 (1999) and references therein.
1152: \bibitem{varma} C. M. Varma, Phys. Rev. B {\bf 55}, 14554 (1997);
1153: Phys. Rev. Lett. {\bf 83}, 3538 (1999); Phys. Rev. B {\bf 61},
1154: 3804 (2000).
1155: \bibitem{ddw} S. Chakravarty, R. B. Laughlin, D. K. Morr and
1156: C. Nayak, Phys. Rev. B {\bf 63}, 094503 (2001).
1157: \bibitem{halperin} B. I. Halperin and T. M. Rice, Solid State Phys.
1158: {\bf 21}, 116 (1968).
1159: \bibitem{schulz} H. J. Schulz, Phys. Rev. B {\bf 39}, 2940 (1989).
1160: \bibitem{affleck} I. Affleck and J. B. Marston, Phys. Rev. B
1161: {\bf 37}, 3774 (1991).
1162: \bibitem{stanescu} T. D. Stanescu and P. Phillips,
1163: Phys. Rev. B {\bf 64}, 220509 (2001).
1164: \bibitem{lee} D. A. Ivanov, P. A. Lee and X-G. Wen, Phys. Rev.
1165: Lett. {\bf 84}, 3958 (2000).
1166: \bibitem{leung} P. W. Leung,
1167: Phys. Rev. B {\bf 62}, R6112 (2000).
1168: \bibitem{scalapino} D. J. Scalapino, S. R. White and I. Affleck,
1169: Phys. Rev. B {\bf 64}, 100506 (2001).
1170: \bibitem{marston1} J. O. Fjaerestad and J. B. Marston,
1171: Phys. Rev. B {\bf 65} 125106 (2002).
1172: \bibitem{marston2} J. B. Marston, J. O. Fjaerestad and A. Sudbo,
1173: to be published in Phys. Rev. Lett.
1174: \bibitem{dagotto} E. Dagotto, Rev. Mod. Phys. {\bf 66}, 763 (1994).
1175: \bibitem{varma3} P. B. Littlewood, C. M. Varma and E. Abrahams,
1176: Phys. Rev. Lett. {\bf 60}, 379 (1987), C. M. Varma, S. Schmitt-Rink and
1177: E. Abrahams, Solid State Commun. {\bf 62}, 681 (1987).
1178: \bibitem{emery3} V. J. Emery, Phys. Rev. Lett. {\bf 58}, 2794 (1987).
1179: \bibitem{white} S. R. White, Phys. Rev. B {\bf 48}, 10345 (1993).
1180: \bibitem{orignac} E. Orignac and T. Giamarchi, Phys. Rev. B {\bf 56}, 7167 (1997).
1181: \bibitem{balents} H-H. Lin, L. Balents and M. P. A. Fisher,
1182: Phys. Rev. B {\bf 48}, 10345 (1993).
1183: \bibitem{gubernatis1} M. Guerrero, J. E. Gubernatis and
1184: S. Zhang, Phys. Rev. B {\bf 57}, 11980 (1998).
1185: \bibitem{gubernatis2} Z. B. Huang, H. Q. Lin and J. E. Gubernatis,
1186: Phys. Rev. B {\bf 63}, 115112 (2001).
1187: \bibitem{parameters} M. S. Hybertson, M. Schl\"uter and N. E. Christensen,
1188: Phys. Rev. B {\bf 39}, 9028 (1989); A. K. McMahan, J. F. Annett
1189: and R. L. Martin, Phys. Rev. B {\bf 42}, 6268 (1990);
1190: R. L. Martin, Phys. Rev. B {\bf 53}, 15501 (1996).
1191: \bibitem{haldane} F. D. M. Haldane, J. Phys. C {\bf 14}, 2585 (1996);
1192: T. Giamarchi and H. J. Schulz, Phys. Rev. B {\bf 39}, 4620 (1989).
1193: \end{thebibliography}
1194: \end{document}
1195: