1: \documentclass[prb,twocolumn,amsmath,amssymb,showpacs]{revtex4}
2: \usepackage{graphicx}
3:
4: \begin{document}
5:
6: \title{Effects of domain walls on hole motion in the
7: two-dimensional anisotropic $t$-$J$ model at finite
8: temperature}
9:
10: \author{Jos\'e A. Riera}
11: \affiliation{
12: Instituto de F\'{\i}sica Rosario, Consejo Nacional de
13: Investigaciones
14: Cient\'{\i}ficas y T\'ecnicas, y Departamento de F\'{\i}sica,\\
15: Universidad Nacional de Rosario, Avenida Pellegrini 250,
16: 2000-Rosario, Argentina \\
17: Institute for Materials Research, Tohoku University, Sendai 980-8577,
18: Japan }
19:
20: \date{\today}
21:
22: \begin{abstract}
23: The $t$-$J$ model on the square lattice, close to the $t$-$J_z$ limit,
24: is studied by quantum Monte Carlo techniques at finite temperature
25: and in the underdoped regime. A variant of the Hoshen-Koppelman
26: algorithm was implemented to identify the antiferromagnetic domains
27: on each Trotter slice. The results show that the model presents at
28: high enough
29: temperature finite antiferromagnetic (AF) domains which collapse at
30: lower temperatures into a single ordered AF state. While there are
31: domains, holes would tend to preferentially move along the domain
32: walls. In this case, there are indications of hole pairing starting
33: at a relatively high temperature. At lower temperatures, when the
34: whole system becomes essentially fully AF ordered, at least in finite
35: clusters, holes would likely tend
36: to move within phase separated regions. The crossover between both
37: states moves down in temperature as doping increases and/or as the
38: off-diagonal exchange increases. The possibility of hole motion
39: along AF domain walls at zero temperature in the fully isotropic
40: $t$-$J$ is discussed.
41: \end{abstract}
42:
43: \pacs{74.20.-z, 74.25.Dw, 74.25.Ha, 02.70.Uu}
44:
45: \maketitle
46:
47: The interplay between hole dynamics and magnetic background in
48: the Cu-O plane is considered the central issue in
49: high-Tc superconductivity. This interplay is quite likely
50: essential not only for the mechanism of hole pairing but also
51: for other important features present in the cuprates such as
52: the pseudogap phase\cite{Tallon} and various types of phase separated,
53: inhomogeneous, states.\cite{Savici,Pan,Lang,Fujita}
54:
55: These problems have been extensively studied by a number of
56: analytical and numerical techniques. One of the most important
57: approaches adopted is through the study of microscopic
58: Hamiltonians, such as the $t$-$J$ model in two dimensions (2D).
59: In particular, numerical techniques have provided a number of
60: highly reliable results for many of the relevant properties which are
61: experimentally measured.\cite{Dagotto} Many of these properties
62: are related to the problem of holes and binding of holes in an
63: antiferromagnet.\cite{poilblanc,white1,riera-dagotto}
64: These numerical results were
65: obtained on relatively small clusters as compared, for example, with
66: the antiferromagnetic correlation length in the doped region or
67: with the superconducting coherence length in the superconducting
68: phase. As in these numerical studies (see below), many theoretical
69: scenarios,
70: like the ``string"\cite{Trugman} and ``spin bag"\cite{schrieffer}
71: pictures, assume the presence of an homogeneous antiferromagnetic
72: background.
73:
74: However, a number of experimental results indicate the presence
75: of various kinds of inhomogeneities\cite{Savici,Pan,Lang}, in
76: addition to the most well-known and controversial ones,
77: stripes\cite{Fujita},
78: which can not easily be included in numerical calculations.
79: Although these inhomogeneities appear in some specific compounds,
80: there is another more universal type of inhomogeneity, in this case
81: of dynamical nature, which are the domain walls (DW) between
82: short range AF regions which appear when doping the AF half-filled
83: insulator. Inelastic neutron scattering studies\cite{Gooding} have
84: shown that the AF correlation length decreases as doping increases.
85: For example, the AF correlation length is between 10 and $20 \AA$
86: in the doping range $0.04 < x <0.15$. This finite
87: length of the AF ordered domains is missed in most numerical
88: calculations on small clusters but even in those cases where
89: the largest distance on the cluster studied is about or larger
90: than the AF correlation length, the numerical technique employed
91: should be able to determine these dynamical AF domains and to follow
92: the movement of holes within this array of domains. So far this kind
93: of analysis has not been performed.
94: The closest approach has been done by using density matrix
95: renormalization group (DMRG) techniques.\cite{white1,white2}
96: However in this approach, domain walls appear as static features
97: due to the use of open boundary conditions in rectangular clusters.
98:
99: The effect of domain walls on hole dynamics has been discussed
100: specially in the context of stripes.\cite{Emery} It has been realized
101: that stripes are essentially DWs with an almost one-dimensional pattern.
102: It is also apparent that holes acquire a larger mobility by moving
103: along these DWs since there is no cost in magnetic
104: energy.\cite{Chernyshev} It is then possible that this same physics
105: of holes moving along the boundaries of AF domains be present
106: in {\em all} doped cuprates, even in those where there is so far no clear
107: experimental evidence of stripes just due to the presence of finite
108: AF domains. It has been also suggested that stripes could be the
109: way in which a strongly correlated electron system avoids the tendency
110: to macroscopic phase separation.\cite{Emery} Although the presence
111: of phase separation is still controversial in the 2D $t$-$J$ model
112: from numerical studies,\cite{Boninsegni} it is likely that the
113: proximity to a phase separated state is at the origin of the various
114: kinds of inhomogeneities mentioned above.
115:
116: The purpose of the present study is then to determine the presence
117: of finite AF domains in the 2D $t$-$J$ model and the interrelation
118: between hole movement and domain walls. A finite temperature
119: conventional quantum Monte Carlo technique using the checkerboard
120: decomposition is used.\cite{RegerYoung} This technique allows the
121: study of larger clusters than the ones accessible by exact
122: diagonalization, and more importantly, although working at low
123: hole doping, the number of holes in the system is also much
124: larger.
125: Not only most numerical calculations have so far involved few holes
126: but also analytical approaches, like those mentioned above, are
127: essentially independent particle pictures. On the other hand,
128: it has been noticed that the system with many holes
129: shows a broad range of features, absent in the few holes problem,
130: covered by the concept of ``topological ordering".\cite{Emery1}
131: In fact, not only many holes destroy long range AF order but also
132: the experimental result that spin stripes order at lower temperature
133: than charge stripes is presumably a collective effect and not
134: due to single-hole physics.\cite{Riera}
135:
136: The model studied is the exchange anisotropic $t$-$J$ model:
137: \begin{eqnarray}
138: {\cal H} = &-& t \sum_{ \langle i j \rangle,\sigma }
139: ({\tilde c}^{\dagger}_{ i\sigma}
140: {\tilde c}_{ j\sigma} + h.c. ) \nonumber \\
141: &+& J \sum_{ \langle i j \rangle }
142: \left[ S^z_i S^z_j +\gamma (S^x_i S^x_j +S^y_i S^y_j )
143: - \frac{1}{4} n_{i} n_{j} \right]
144: \label{hamtj}
145: \end{eqnarray}
146: \noindent
147: in standard notation. $t=1$, $J=0.35$. Square clusters with side
148: length 8, 10, 12 and 16 with periodic boundary conditions (PBC)
149: and hole doping $0.03 < x < 0.10$ are considered.
150: Temperature is given in units of $t$.
151:
152: It is possible to argue that an enhanced exchange ZZ component
153: appears at an effective level due to inter plane exchange coupling.
154: For example, in La$_2$CuO$_{4.11}$ ($x=0.14$), superconducting
155: below 42~K, there is an AF correlation length in the
156: c-direction that couples a Cu-O$_2$ plane at least with its
157: two (one above, one below) adjacent planes.\cite{YSLee}
158: Although smaller in strength than other 2D interactions
159: (spin-orbit, Dzialoshinskii-Moriya, four-site ring exchange), the
160: inter plane coupling is the responsible for a finite temperature
161: transition at half-filling, below which the correlation length
162: diverges.
163: In any case, we are going to work at very small $\gamma$,
164: i.e. close to the Ising limit of the exchange interaction, or
165: $t$-$J_z$ model, which is not a realistic case
166: for the cuprates but is required to reduce the ``minus sign
167: problem" of QMC simulations (see below). In addition, at
168: half-filling, the isotropic 2D Heisenberg model does not have an AF
169: transition at finite temperature (Mermin and Wagner theorem) but the
170: 2D Ising model has a critical temperature of $2.269~J=0.794~t$.
171: In any case, as far as the interaction between holes and short-range
172: AF order is concerned, the isotropic $t$-$J$ and the $t$-$J_z$
173: models lead to similar results.\cite{ChLeung}
174:
175: The ``minus sign problem" is a very well-known aspect
176: of QMC simulations and it has been extensively analyzed.\cite{Loh}
177: We give here the outline of the sign calculation in order
178: to make our results fully reproducible. This calculation
179: consists of two parts. First, a local contribution which
180: takes into account fermion permutations inside each cube.
181: Then a global contribution which is due to fermion permutations
182: between different plaquettes at each time slice. The
183: permutations which appear when one hole moves from one side
184: of the cluster to the opposite side have not been included. This
185: is expensive numerically and it is just a boundary contribution.
186: It has been also checked that the influence of this contribution
187: is negligible, at least in the range of temperatures considered.
188: The results shown below have been obtained by averaging at least
189: over six independent runs and at one of the lowest temperatures,
190: $T=0.3 t$, each run consisted of 700,000 MC steps after
191: thermalization. Only the
192: results corresponding to an average sign larger than 0.01 were
193: considered in the present study. The relative
194: error of the measured quantities is around 0.01, i.e. about
195: the size of the symbols used in the plots, except at the lowest
196: temperatures reached, where the error bars are two times larger.
197:
198: In order to identify the AF domains on each Trotter slice on the
199: square lattice, a variant of the Hoshen-Koppelman
200: algorithm\cite{HoshKopp} was implemented. This algorithm consists
201: in sweeping each slice row by row. At each site, its state
202: ($z$-projection of the spin) has to be compared with the one of the
203: previous nearest neighbor (NN) site on the same row and with the state
204: of the NN site on the previous row. Two NN sites belong
205: to the same domain if their spins are antiparallel. Care should
206: be taken due to spatial PBC for the last
207: site on each row and for the last row on each slice. Once the
208: domains are identified the number of sites belonging to it is
209: counted. In addition, for all spins belonging to a given
210: domain, the number of neighbors belonging to the
211: same domain and the number of its neighbors belonging to
212: neighboring domains are computed.
213:
214: We start with a study of the magnetic ordering in the $t$-$J_z$
215: limit ($\gamma=0$).
216: In Fig.~\ref{fig1}(a) it is shown the dependence with hole
217: doping $x$ of
218: the volume fraction of the two largest domains, $v_1$ and $v_2$
219: obtained by dividing the number of spins belonging to
220: those domains, $V_1$ and $V_2$, by the number of cluster
221: sites. It can be seen that
222: already at a high temperature, $\rm T \approx t$, there is a large,
223: percolating AF domain occupying roughly half the volume of the
224: lattice. The second largest domain occupies on average roughly
225: one quarter of the lattice, and the relative volume of the third
226: largest domain is smaller than 0.1.
227: One should take into account that for randomly distributed
228: $\pm 1$ numbers on a lattice ($\rm T=\infty$ limit), undoped case,
229: the relative volumes are 0.31, 0.17 and 0.10 respectively.
230: As the temperature is lowered, at around $\rm T \approx 0.7 t$
231: (this value decreases as doping increases)
232: there is a rather sharp decrease of the volume of the second
233: largest domain with a consequent increase of the largest domain
234: i.e., the second largest domain collapses into the first one.
235: This change gets smoother as $x$ increases.
236: Below this crossover temperature region, the largest ordered domain
237: occupies 0.95 of the cluster for $x=0.031$, decreasing smoothly to
238: 0.81 for $x=0.12$.
239:
240: \begin{figure}
241: \begin{center}
242: \setlength{\unitlength}{1cm}
243: \includegraphics[width=6cm,angle=-90]{./clusg0.eps}
244: \caption{Results for the $16\times 16$ (open symbols),
245: $12\times 12$ (full symbols), and $10\times 10$ (stars) clusters
246: as a function of temperature, for $\gamma =0$ and various hole
247: densities as indicated in the plot.
248: (a) Volume fraction of the largest AF domain (solid lines)
249: and second largest domain (dashed lines).
250: (b) Spin-spin correlations at the maximum distance.}
251: \label{fig1}
252: \end{center}
253: \end{figure}
254:
255: In Fig.~\ref{fig1}(b) the resulting AF order as seen from $v_1$ and
256: $v_2$ is studied by using a more
257: conventional measure of magnetic order, i.e. the spin-spin
258: ($\langle S^z_i S^z_j \rangle$) correlation function at the
259: maximum distance, $S_{max}$. Again there is a steep increase of
260: $S_{max}$ as $T$ decreases presumably indicating a finite temperature
261: transition in the bulk limit. The temperature of this jump decreases
262: from $T \approx 0.7$, for $x=0.039$ ( $16\times 16$ cluster) to
263: $T \approx 0.6$, for $x=0.097$ ($12\times 12$ cluster).
264:
265: In Fig.~\ref{fig2}(a) it is shown the dependence of the volume
266: fraction of the two largest domains $v_1$ and $v_2$ with $\gamma$.
267: Although up to $\gamma=0.25$ the results do not depart
268: appreciably from those of the Ising limit, at $\gamma=0.5$ it can
269: be observed an important reduction of the volume of the largest
270: domain, and hence in the AF ordering of the system. This
271: behavior of the $\gamma$ dependence is consistent with the one
272: observed for other quantities as will be shown below, and it
273: suggests that the behavior found for small $\gamma$ at high
274: temperature is going to remain valid at lower temperatures
275: (and eventually zero temperature) as the isotropic limit is
276: reached. Fig.~\ref{fig2}(b) contains a study of the dependence
277: of these domain sizes with cluster size at $\gamma =0.5$.
278: The finite size effects are not very important for this anisotropy
279: value and these clusters sizes, although it may become more
280: important as the isotropic Heisenberg case is reached and for
281: smaller sizes.\cite{RegerYoung} Notice however that there is a
282: slightly larger value of the doping fraction as the linear size
283: is reduced in the data shown in this figure. Hence, if the
284: results were corrected at the same doping fraction, finite size
285: effects would be somewhat larger.
286:
287: \begin{figure}
288: \begin{center}
289: \setlength{\unitlength}{1cm}
290: \includegraphics[width=6cm,angle=-90]{./clusiz.eps}
291: \caption{ (a) Volume fraction of the largest AF domain (solid lines)
292: and second largest domain (dashed lines), on the $10\times 10$ cluster
293: and various values of the anisotropy $\gamma$ indicated in the plot.
294: (b) Same as (a) but for $\gamma=0.5$ and various cluster linear sizes
295: and hole densities, indicated in the plot.}
296: \label{fig2}
297: \end{center}
298: \end{figure}
299:
300: Now, the main point of this study is to detect the location of
301: holes in the presence of ordered AF domains as long as this
302: doped system is approximately described by a $t$-$J$ model close
303: to the Ising limit.
304: To get some insight about the microscopic interrelation between
305: holes and AF domains it is useful to look at ``snapshots" of
306: the system generated during the QMC simulations.
307: Fig.~\ref{snaps}(a) shows a picture at a relatively high temperature
308: ($T=0.8$) where there are still ordered domains with the opposite
309: sublattice magnetization (darker shades of gray) than the main
310: ordered domain (lightest shade of gray).
311: The salient feature in this picture is that holes are located
312: preferentially on the boundary of the largest AF domain.
313: There are some other situations that can be identified in this
314: picture, for example isolated holes inside the largest domain
315: with a ``string" of overturned spins attached to it.\cite{Trugman}
316:
317: \begin{figure}
318: \begin{center}
319: \setlength{\unitlength}{1cm}
320: \includegraphics[width=5cm,angle=0]{./conftj.eps}
321: \caption{Snapshot of the $16\times 16$ cluster with 16 holes
322: at $T=0.8$ (top panel) and at $T=0.4$ (bottom panel). The
323: two lightest shades of gray correspond to spins with the
324: same sublattice magnetization, opposite to that of the
325: two darkest shades of gray. Holes are indicated by circles.}
326: \label{snaps}
327: \end{center}
328: \end{figure}
329:
330: On the other hand, Fig.~\ref{snaps}(b) corresponds to a typical
331: lower temperature behavior. The system presents a single ordered
332: domain occupying most of the cluster and with just few disordered
333: sites. Some features can be identified, such as isolated holes
334: or pairs inside the percolating ordered domain. However, the most
335: important feature is that most holes seems to be located in a phase
336: separated region. In addition, in this and many other snapshots,
337: we observed that most of the spins
338: inside and around this region are ferromagnetically aligned, but
339: we have not studied this issue systematically.
340: Presumably, as AF domains with the opposite
341: magnetization to that of the main one shrink with decreasing
342: temperature, they would leave behind these highly doped regions
343: forming a ``mesoscopic" phase separated state or, more properly,
344: a charge inhomogeneous state. In this situation holes would now
345: gain kinetic energy by moving in a ferromagnetic background.
346:
347: Of course, it is necessary to determine at a more quantitative level
348: the presence of the features shown in Fig.~\ref{snaps}.
349: To this end, we adopted the criterion of looking only at the spins
350: and holes belonging to the bulk or to the surface of the largest
351: ordered AF domain, neglecting the contribution from smaller domains.
352: We believe that this criterion gives the right behavior of the
353: system. Then,
354: for any site belonging to the largest ordered AF domain
355: we computed the number of its nearest and next-nearest neighbor
356: sites belonging to the same domain. We considered a spin in the bulk
357: of the largest AF domain (inner spin) if the number of its neighbors
358: belonging to the same domain is seven or eight, and we consider it
359: at the boundary (boundary spin) if that number is between three and
360: five. We prefer to leave aside the intermediate case of six neighbors
361: and also the
362: case of less than three neighbors, in this case to eliminate some
363: sites which are loosely connected to the main AF domain. In any case,
364: we repeated the calculations taking 2-5 neighbors for the boundary sites
365: and 6-8 neighbors for the inner sites obtaining essentially the same
366: results. We are confident then that the results shown below are quite
367: robust, independent of the details of the classification of spins
368: and holes.
369:
370: The average fractions of spins located on the boundary ($n^s_b$) and
371: inside ($n^s_i$) the largest AF domain, are defined as the average
372: number of boundary spins divided by the total number of spins in
373: this cluster ($V_1$) and the average number of inner spins
374: divided by $V_1$ respectively.
375: In Fig.~\ref{spinnboun}(a), $n^s_b$ and $n^s_i$ are
376: shown as a function of temperature, at $\gamma=0$ and other parameters
377: as in Fig.~\ref{fig1}. At high temperature the fraction of boundary
378: spins is larger than the fraction of inner ones but this situation is
379: reversed as the temperature is reduced due to the overall ordering of
380: the system. As doping increases, the crossover temperature decreases,
381: consistent with the reduction of AF order with doping.
382: As it can be seen in Fig.~\ref{spinnboun}(b),
383: this crossover temperature is further reduced as the off-diagonal
384: exchange term ($\gamma$) increases. Again, one could expect that the
385: region over which there is an important fraction of boundary spins
386: extends down to zero temperatures at the isotropic limit.
387:
388: \begin{figure}
389: \begin{center}
390: \setlength{\unitlength}{1cm}
391: \includegraphics[width=6cm,angle=-90]{./spnei.eps}
392: \caption{Average fraction of spins located on the boundary (dashed lines)
393: and inside (solid lines) the largest AF domain as a function
394: of temperature. (a) Same clusters and hole filling as in
395: Fig.~\ref{fig1}, $\gamma=0$.
396: (b) $8\times 8$, 6 holes, $\gamma=0$ (filled triangles), 0.25
397: (open triangles) and 0.5 (crosses); $12\times 12$, 8 holes, $\gamma=0$
398: (filled squares), 0.25 (open squares) and 0.5 (pluses).}
399: \label{spinnboun}
400: \end{center}
401: \end{figure}
402:
403: To determine the location of holes with respect to these AF domains,
404: a similar procedure as above was extended to holes. That is,
405: for every hole the number of its nearest and next-nearest neighbor
406: spins belonging to the largest ordered AF domain was computed.
407: As before, we considered a hole located inside
408: the largest AF domain (inner hole) if the number of its
409: neighbors belonging to the same site is seven or eight, and we consider
410: it at the boundary (boundary hole) if that number is between three and
411: five. The average fraction of boundary holes ($n^h_i$) and the
412: average fraction of inner holes
413: ($n^h_i$) are obtained by dividing the number of boundary holes
414: and the number of inner holes respectively, by the total number of
415: holes.
416:
417: \begin{figure}
418: \begin{center}
419: \setlength{\unitlength}{1cm}
420: \includegraphics[width=6cm,angle=-90]{./holnei.eps}
421: \caption{Average fraction of holes located on the boundary (dashed lines)
422: and inside (solid lines) the largest AF domain as a function
423: of temperature. (a) Same clusters and hole filling as in
424: Fig.~\ref{fig1}, $\gamma=0$.
425: (b) Same clusters and parameters as in Fig.~\ref{spinnboun}(b).}
426: \label{hlinnboun}
427: \end{center}
428: \end{figure}
429:
430: Results of these calculation are depicted in Fig.~\ref{hlinnboun}.
431: The overall features are similar to those of Fig.~\ref{spinnboun}, i.e.
432: there is a dominance of boundary holes at higher temperatures
433: changing to inner holes at lower temperatures (Fig.~\ref{hlinnboun}(a)).
434: The fact that this crossover temperature at a given doping is lower than
435: the corresponding temperature for spins suggests that holes have a
436: preference to be located at the boundary, a possibility that will
437: be confirmed below. Fig.~\ref{hlinnboun}(b) also shows that this
438: crossover temperature is reduced by increasing $\gamma$ again
439: consistent with the results of Fig.~\ref{spinnboun}(b). Another
440: feature that can be observed is that at the lowest temperatures,
441: the reverse behavior occurs, i.e., there is an increase (decrease) of
442: the boundary (inner) hole fractions. It is easy to see that this
443: behavior is consistent with the formation of a phase separated
444: hole-rich region inside the main AF percolating domain. This
445: possibility is also examined below.
446:
447: The central problem of the interrelation between holes and AF domains
448: is addressed in Fig.~\ref{hol_loc}, where the ratio $r^h_b$ of the
449: fraction of boundary holes to the fraction of boundary spins is shown.
450: This quantity, obtained from two quantities computed in QMC
451: simulations, have then larger error bars than the ones of those
452: quantities. The main result is that the ratio $r^h_b$ is always
453: noticeably larger than one indicating a preference of the holes to
454: move along the boundary of AF domains. In Fig.~\ref{hol_loc}(a), it
455: is shown that $r^h_b$ increases as $T$ is reduced, and
456: decreases as doping increases. In Fig.~\ref{hol_loc}(b), it is also
457: shown that $r^h_b$ decreases as $\gamma$ increases. It is out of
458: question to perform an extrapolation to $\gamma=1$ but this
459: behavior suggests that one should be cautious about extending
460: the conclusions of the present study to the isotropic case.
461: The inset of Fig.~\ref{hol_loc}(b) shows the hole population of
462: the domain walls, $p_w$, defined as the number of holes on the
463: boundary of the largest ordered domain divided by the number of
464: spins (i.e., the ``length") of this boundary. It turns out that
465: $p_w$ is just $r^h_b$ multiplied by the hole density $x$ and
466: divided by $v_1$. To take a reference, on the stripes $p_w=0.5$,
467: which seems considerably higher than an extrapolation to zero
468: temperature of the present data, specially taking into account the
469: reduction with increasing $\gamma$.
470:
471: \begin{figure}
472: \begin{center}
473: \setlength{\unitlength}{1cm}
474: \includegraphics[width=6cm,angle=-90]{./hol_loc.eps}
475: \caption{Ratio of the average fraction of holes located on the
476: boundary to the average fraction of spins located on the
477: boundary as a function of temperature. (a) Same clusters and
478: hole filling as in Fig.~\ref{fig1}, $\gamma=0$.
479: (b) $8\times 8$, 6 holes, $\gamma=0$ (filled triangles), 0.25
480: (open triangles) and 0.5 (crosses); $12\times 12$, 8 holes,
481: $\gamma=0$ (filled squares), 0.25 (open squares) and 0.5 (pluses).
482: The inset shows hole population of DW for the same parameters as
483: in (b), $\gamma=0$ and 0.5.}
484: \label{hol_loc}
485: \end{center}
486: \end{figure}
487:
488: An elementary argument of minimization of the number of broken AF
489: bonds leads to the possibility of pairing of holes even when they
490: are constrained to move along the boundary between AF domains. In
491: order to confirm this possibility, and to get further indications
492: of the presence of a low temperature phase separation, we computed
493: the hole-hole correlations $C(r)$ as a function of temperature. In
494: Fig.~\ref{pairing}(a) these correlations are shown as a function
495: of distance and at various temperatures for a fixed value of
496: $x$ and $\gamma$. As $T$ decreases, the correlations at large
497: distances decrease while the correlations at short distances
498: increase indicating the onset of hole attraction. Quite
499: interesting, at intermediate temperatures, the largest
500: correlation corresponds to the distance of $\sqrt{2}$, as
501: was observed in many numerical studies for two holes at zero
502: temperature in the fully isotropic $t$-$J$
503: model\cite{white1,poilblanc,riera-dagotto}
504: and this is typical of d$_{x^2-y^2}$ pairing. At the lowest
505: temperature, the correlations are enhanced both at near and
506: at far distances and suppressed at intermediate distances
507: which is typical of a phase separated state.
508:
509: \begin{figure}
510: \begin{center}
511: \setlength{\unitlength}{1cm}
512: \includegraphics[width=6cm,angle=-90]{./pair.eps}
513: \caption{(a) Hole-hole correlations as a function of distance on the
514: $12\times 12$ cluster, $x=0.069$, $\gamma=0$, at various
515: temperatures indicated on the plot.
516: (b) Average value of the hole-hole correlations at short
517: distance (open symbols) and at large distance (filled symbols)
518: on the same cluster, $\gamma=0$, at various hole doping
519: levels indicated in the plot, as a function of temperature.
520: The arrows indicate approximately the crossover temperatures.
521: (c) Same as (b) but for $x=0.069$
522: and various values of $\gamma$, indicated in the plot.}
523: \label{pairing}
524: \end{center}
525: \end{figure}
526:
527: For better visualizing the attraction between holes
528: we averaged $C(r)$ for $r=1, \sqrt{2}$ and compared it with
529: the average over the three largest distances on the cluster.
530: Fig.~\ref{pairing}(b) shows these averages as a function of
531: temperature, on the $12\times 12$ cluster, $\gamma=0$ and
532: at various dopings. It can be seen that the
533: crossover temperature at which the near distance
534: correlations start to dominate over the long distance ones,
535: i.e. the probability of finding two holes becomes larger
536: at short distance than at long distance,
537: decreases slightly as a function of doping (of course, error
538: bars exclude a precise determination of this crossover).
539: This crossover temperature could be identified with $T^{MF}$ as
540: discussed in Ref.~\onlinecite{Emery2}.
541: Fig.~\ref{pairing}(c) seems also to indicate that the pairing
542: crossover temperature, for a given cluster size and doping,
543: decreases with increasing $\gamma$. One should stress the
544: fact that this onset of pairing takes place while there is
545: still a larger fraction of holes on the DW than inside the
546: largest ordered AF domain. For example, from
547: Fig.~\ref{pairing}(b),
548: the onset of pairing occurs approximately at $T=0.76$,
549: 0.70 and 0.66, at $x=0.056$, 0.069 and 0.083, respectively.
550: The corresponding crossover from DW hole fraction to
551: bulk hole fraction (Fig.~\ref{hlinnboun}(a))takes place at
552: $T=0.69$, 0.66 and 0.60 for the same fillings.
553:
554: It is also worth to mention that by looking at hole-hole correlations
555: in momentum space
556: we have not found indications of stripe formation. This hole
557: static structure factor is loosely distributed around $(0,0)$
558: with not definite pattern. Presumably, additional mechanisms
559: have to be invoked to stabilize a stripe state.\cite{Riera}
560: The peak of the magnetic structure factor is also always at
561: $(\pi,\pi)$. In addition,
562: by examining snapshots like the ones in Fig.~\ref{snaps},
563: and by computing a measure of compactness using the data of
564: Fig.~\ref{spinnboun}, it turns out that the AF domains are in close
565: contact between each other and are far from being ``round" objects
566: as implicitly assumed in Ref.~\onlinecite{Savici}. Hence, also in
567: this type of inhomogeneity an external agent should be present to
568: stabilize the reported static AF islands.
569:
570: Let us summarize and discuss the main results of the present study.
571: The 2D exchange anisotropic $t$-$J$ model, close to the $t$-$J_z$
572: limit, presents at high enough temperature finite AF domains
573: which, at least in finite clusters, collapse at
574: lower temperatures into a single ordered AF state. While there are
575: domains, and hence domain walls, holes would tend to preferentially
576: locate along these DW. In this case, there are indications of hole
577: pairing starting at a relatively high temperature.
578: One could speculate that this situation
579: extends down to zero temperature in the fully isotropic $t$-$J$
580: model as this model is supposed to describe, for the appropriate
581: hole doping, the superconducting phase of the cuprates.
582: On the other hand, the fully isotropic case may behave as the
583: $t$-$J_z$ model behaves at zero temperature, at least in finite
584: clusters, and in this case it could undergo a crossover to a charge
585: inhomogeneous state. As mentioned in the introductory paragraphs,
586: there are many experimental indications of various kinds of
587: inhomogeneities in the cuprates. In any case, the nature of
588: the possible charge inhomogeneous state detected in the
589: simulations should be further analyzed.\cite{newpap}
590:
591: A possible consequence of the reduction of the boundary length
592: as temperature is decreased, is that the holes
593: would become increasingly localized. This localization could
594: contribute to the increase of in-plane resistivity in lightly
595: doped La$_{2-x}$Sr$_x$CuO$_4$ and YBa$_2$CuO$_y$ as T is reduced
596: reported in recent experimental studies.\cite{Ando1,Ando2}.
597: An indirect support to this scenario comes from the
598: relation between resistivity and AF correlation length in
599: La$_{2-x}$Sr$_x$CuO$_4$. It can be seen (Fig.~3 of
600: Ref.~\onlinecite{Ando1}) that the inverse mobility and the AF
601: correlation length $\rho_{AF}$ have an identical behavior as a
602: function of doping. However, one should be cautious since the
603: mobility is measured at room temperature while $\rho_{AF}$ is the
604: zero temperature value.
605: Of course, in these compounds, stripes, which are a particular
606: kind of DW, could be stabilized. In addition, the kinetic energy,
607: computed in our simulations, is almost constant down to the lowest
608: temperatures attainable. Perhaps, the system responds to this
609: tendency to localization by going into a charge inhomogeneous
610: state. In any case, one could suggest recipes to avoid this
611: tendency to localization due to the shrinking of
612: AF domains, present in all cuprates. For example, DWs could be
613: enforced into the system by introducing magnetic
614: impurities outside CuO$_2$ layers in order to pin
615: domains with random sublattice magnetizations. The resulting
616: enlarged DW length would reduce hole localization and prevent
617: phase separation.
618:
619: Finally, let us compare the present results with the ones obtained
620: in some of the most recent and closest studies. In the first place,
621: in Ref. \onlinecite{Chernyshev}, DWs are essentially straight lines,
622: and essentially {\em static}, both in their analytical treatment
623: and in their DMRG calculations.
624: Hence their predictions are of limited applicability in the
625: finite temperature real physics of the $t$-$J_z$ model, as
626: revealed by the present simulations, where dynamical domains are
627: considerably non-compact regions. In particular, our main result,
628: holes moving essentially along DWs, suggests that ``transverse"
629: motion of holes is unlikely. Notice also that, by looking at the
630: present results on a larger
631: ``time scale", the AF domains could appear as more compact regions
632: separated by broadened disordered domain walls. In addition,
633: Ref. \onlinecite{Chernyshev} is a study at zero temperature, where
634: we obtain, with PBC, indications of clustering of holes. In this
635: sense, their zero temperature DW should be a consequence of
636: additional effects not included in the $t$-$J_z$ model. In our
637: case, finite AF domains appear as a competition between charge and
638: magnetic energy, at finite temperatures. The issue of domain
639: wall formation in our model has to be further investigated and
640: it would be useful to compare the wall energetics with the
641: predictions of Ref. \onlinecite{Chernyshev}.
642: Another of our most important results are the indications of
643: binding of holes. This is a delicate issue which has been
644: hotly debated in the context of stripes.\cite{white2,Emery,Riera}
645: In Ref. \onlinecite{Chernyshev} it was concluded that the presence
646: of a DW is mostly irrelevant to pairing. In studies where charge
647: stripes are forced into the system by an on-site potential added to
648: an anisotropic $t$-$J$ model\cite{Riera}, no indications of pairing
649: were found: stripes are metallic. In these studies, spin and hole
650: correlations along the stripe are very much similar to the ones
651: in isolated $t$-$J$ chains. One may suggest then, that the
652: behavior of dynamical domain walls studied in the present work
653: is considerable different than for stripes where charge is ordered
654: by an external mechanism.
655:
656: \begin{acknowledgments}
657: The author wishes to thank many useful discussions with S.
658: Maekawa, G. B. Martins, T. Tohyama and Y. J. Uemura.
659: The use of supercomputers and friendly technical assistance at the
660: Center for Computational Materials Science, IMR, Tohoku University,
661: is also gratefully acknowledged.
662: \end{acknowledgments}
663:
664: \begin{thebibliography}{}
665:
666: \bibitem{Tallon} J. L. Tallon and J. W. Loram, Physica C {\bf 349},
667: 53 (2001).
668:
669: \bibitem{Savici} A. T. Savici, Y. Fudamoto, I. M. Gat, T. Ito, M. I. Larkin,
670: Y. J. Uemura, G. M. Luke, K. M. Kojima, Y. S. Lee, M. A. Kastner, R. J.
671: Birgeneau, and K. Yamada, Phys. Rev. B {\bf 66}, 014524 (2002).
672:
673: \bibitem{Pan} S. H. Pan, J. P. ONeal, R. L. Badzey, C. Chamon, H.
674: Ding, J. R. Engelbrecht, Z. Wang, H. Eisaki, S. Uchida, A. K.
675: Gupta, K. W. Ng, E. W. Hudson, K. M. Lang, and J. C. Davis,
676: Nature {\bf 413}, 282 (2001).
677:
678: \bibitem{Lang} Lang K.M., Madhavan V., Hoffman J.E., Hudson E.W.,
679: Eisaki H.,
680: Uchida S., and J. C. Davis, Nature {\bf 415}, 412 (2002).
681:
682: \bibitem{Fujita} M. Fujita, H. Goka, K. Yamada, and M. Matsuda, Phys. Rev.
683: Lett. {\bf 88}, 167008 (2002), and references therein.
684:
685: \bibitem{Dagotto} E. Dagotto, Rev. Mod. Phys. {\bf 66}, 763 (1994).
686:
687: \bibitem{poilblanc} D. Poilblanc, Phys. Rev. B {\bf 49}, 1477 (1994).
688:
689: \bibitem{white1} S. R. White and D. J. Scalapino, Phys. Rev. B {\bf 55},
690: 6504 (1997).
691:
692: \bibitem{riera-dagotto} J. Riera and E. Dagotto, Phys. Rev. B {\bf 57},
693: 8609 (1998).
694:
695: \bibitem{Trugman} S. A. Trugman, Phys. Rev. B {\bf 37}, 1597 (1988).
696:
697: \bibitem{schrieffer} J. R. Schrieffer, X. G. Wen, and S. C. Zhang,
698: Phys. Rev. B {\bf 39}, 11663 (1989).
699:
700: \bibitem{Gooding} R. J. Gooding, N. M. Salem, R. J. Birgeneau,
701: and F. C. Chou, Phys. Rev. B {\bf 55}, 6360 (1997).
702:
703: \bibitem{white2} S. R. White and D. J. Scalapino, Phys. Rev. Lett
704: {\bf 80}, 1272 (1998); {\bf 81}, 3227 (1998).
705:
706: \bibitem{Emery} V. J. Emery, S. A. Kivelson, and O. Zachar, Phys. Rev. B
707: {\bf 56}, 6120 (1997).
708:
709: \bibitem{Chernyshev} A. L. Chernyshev, S. R. White, and A. H. Castro Neto,
710: Phys. Rev. B {\bf 65}, 214527 (2002).
711:
712: \bibitem{Boninsegni} M. Boninsegni, Phys. Rev. B {\bf 65}, 134403 (2002),
713: and references therein.
714:
715: \bibitem{RegerYoung} J. D. Reger and A. P. Young, Phys. Rev. B {\bf 37},
716: 5978 (1988).
717:
718: \bibitem{Emery1} V. Emery and S. A. Kivelson, Synth. Met. {\bf 80}, 151
719: (1997).
720:
721: \bibitem{Riera} J. A. Riera, Phys. Rev. B {\bf 64}, 104520
722: (2001); id. {\bf 65}, 064524 (2002), and references therein.
723:
724: \bibitem{YSLee} Y. S. Lee, R.J. Birgeneau, M.A. Kastner, Y. Endoh,
725: S. Wakimoto, K. Yamada, R.W. Erwin, S.H. Lee, and G. Shirane,
726: Phys. Rev. B {\bf 60}, 3643 (1999).
727:
728: \bibitem{ChLeung} A. L. Chernyshev and P. W. Leung, Phys. Rev. B
729: {\bf 60}, 1592 (1999), and references therein.
730:
731: \bibitem{Loh} E. Y. Loh Jr, J. E. Gubernatis, R. T. Scalettar, S. R.
732: White, D. J. Scalapino, and R. L. Sugar, Phys. Rev. B {\bf 41}, 9301
733: (1990).
734:
735: \bibitem{HoshKopp} J. Hoshen and R. Koppelman, Phys. Rev. B {\bf 14},
736: 3438 (1976).
737:
738: \bibitem{Emery2} V. Emery and S. A. Kivelson, Nature {\bf 374}, 434
739: (1994). The idea of a high temperature scale for pairing is
740: present in the early RVB papers, S. A. Kivelson, D. S. Rokhsar, and
741: J. P. Sethna, Phys. Rev. B {\bf 35}, 8865 (1987),
742: and in studies of preformed pairs,
743: M. Randeria, N. Trivedi, A. Moreo, and R. T. Scalettar,
744: Phys. Rev. Let. {\bf 69}, 2001 (1992).
745:
746: \bibitem{Ando1} Y. Ando, A. N. Lavrov, S. Komiya, K. Segawa, and X. F.
747: Sun, Phys. Rev. Let. {\bf 87}, 017001 (2001).
748:
749: \bibitem{Ando2} Y. Ando, K. Segawa, S. Komiya, and A. N. Lavrov,
750: Phys. Rev. Lett. {\bf 88}, 137005 (2002).
751:
752: \bibitem{newpap} After completion of this work, we became aware of a
753: related study (R. Eder and Y. Ohta, preprint conmat/0304554) in which
754: dynamical domain walls appear in a small cluster due, essentially,
755: to a frustration mechanism.
756:
757: \end{thebibliography}
758:
759: \end{document}
760: