1: %%
2: %%
3: %%
4:
5: %%
6: %%
7: %%
8: %%
9: %%
10: %%
11: %%
12: %%
13: %%
14: %%
15: %%
16: %%
17: %%
18:
19:
20: %%
21: %%
22: %%
23:
24:
25: %%
26:
27:
28: %%
29: %%
30: %%
31: \documentclass{epl}
32:
33: \usepackage{graphicx}
34: \usepackage{bm}
35: \usepackage{color}
36:
37: \def\beq{\begin{equation}}
38: \def\eeq{\end{equation}}
39: \def\beqn{\begin{eqnarray}}
40: \def\eeqn{\end{eqnarray}}
41: %%
42: \def\eqref#1{(\ref{#1})}
43:
44:
45: \def\vv#1{{\mathbf #1}}
46: \def\up{{\uparrow}}
47: \def\down{{\downarrow}}
48:
49: \newcommand{\de}{\partial}
50:
51: %%
52:
53: \def\pit{$(\pi/2,\pi/2)$}
54: \def\pio{$(\pi,0)$}
55:
56: \newcommand{\myeq}[4]{\ifx#1#2#3\else#4\fi}
57: \newcommand{\nosp}[1]{\myeq#1.#1{\myeq#1,#1{\myeq#1;#1{\myeq#1]#1{\myeq#1)#1%%
58: {\myeq#1:#1{\myeq#1-#1{ #1}}}}}}}}
59:
60:
61:
62: %%
63:
64: \def\pcite#1#2{\cite{#1}#2}
65:
66: %%
67:
68: %%
69:
70:
71:
72:
73: \title{
74: %%
75: %%
76: %%
77: %%
78: %%
79: %%
80: %%
81: %%
82: %%
83: %%
84: %%
85: %%
86: %%
87: %%
88: %%
89: %%
90: %%
91: %%
92: Weak phase separation and the pseudogap\\
93: %%
94: %%
95: in the electron-doped cuprates
96: }
97: \shorttitle{
98: Phase separation in electron-doped cupr.}
99:
100: \author{M. Aichhorn and E. Arrigoni}
101:
102: \institute{
103: Institut f\"ur Theoretische Physik - Computational Physics,
104: Technische Universit\"at Graz, Petersgasse 16, A-8010 Graz, Austria
105: }
106: \pacs{74.20.-z}{Theories and models of superconducting state}
107: \pacs{05.70.Jk}{Critical point phenomena}
108: \pacs{71.10.-w}{Theories and models of many-electron systems}
109:
110:
111: %%
112: \def\cha#1{{#1 }}
113:
114: \begin{document}
115:
116: \maketitle
117:
118: \begin{abstract}
119:
120: We study the quantum transition from an antiferromagnet to a
121: superconductor in a model for
122: electron- and hole-doped cuprates by means of a variational cluster
123: perturbation theory approach. In both cases, our results suggest a
124: tendency towards phase separation between a mixed
125: antiferromagnetic-superconducting phase at low doping and a pure
126: superconducting phase at larger doping. However, in the
127: electron-doped case the energy scale for phase separation is an order
128: of magnitude smaller than for hole doping.
129: We argue that this can explain the different pseudogap and
130: superconducting transition scales in hole- and electron-doped
131: materials.
132:
133:
134: \end{abstract}
135:
136: \section{Introduction}
137:
138: %%
139: High-temperature superconducting materials (HTSC) are characterized by
140: strong electronic correlations which are responsible for a number of
141: anomalous properties and competing phases.
142: The occurrence of these anomalous phases is probably related to the close
143: proximity of the Mott-insulating and antiferromagnetic (AF) phases at half-filling.
144: This is particularly true for hole- (p-)doped materials which,
145: besides the
146: AF and the superconducting (SC)
147: states, display a number of unconventional phases such as
148: stripes~\pcite{em.ki.99,ca.em.03}, checkerboard
149: structures~\pcite{ho.hu.02}, non-Fermi liquid phases, etc.
150: %%
151: It has been argued that
152: these inhomogeneous phases
153: originate
154: from
155: an instability of the AF phase towards phase
156: separation~\pcite{ca.em.03},
157: which is intrinsic in
158: models with short-range interactions such as Hubbard or
159: $t-J$ models.
160: On the other hand, long-range Coulomb repulsion enforces charge
161: homogeneity at long distances
162: so that stripes, other short-range charge inhomogeneities, or strong
163: charge fluctuations can originate as a
164: compromise between
165: the two competing effects~\cite{ca.em.03,lo.em.94,ar.ha.02}.
166: %%
167: %%
168: %%
169: %%
170: %%
171: %%
172: %%
173: %%
174: %%
175: %%
176: %%
177: %%
178: %%
179: \cha{
180: A number of theories support the notion that
181: fluctuations of these inhomogeneities, or of related order parameters, are
182: responsible for the pseudogap
183: %%
184: observed in p-doped materials~\cite{ca.em.03,ca.ca.99,sa.em.96}. In this case,
185: one would expect
186: the corresponding (low) pseudogap temperature scale
187: $T^*$ to be indirectly related
188: to the phase-separation energy scale.
189: }
190:
191: \cha{
192: The situation is not clear
193: in electron- (n-)doped cuprates (especially in the Nd- and Pr-CeCuO
194: compounds). These materials display a
195: more continuous transition from the AF to the SC
196: state~\cite{lu.le.90}. Stripes have not been observed so far, although
197: there are indications of electronic phase inhomogeneities
198: %%
199: %%
200: ~\cite{klauss,da.ka.05}.
201: In some mean-field calculations based upon the single-band Hubbard
202: model for n-doped materials, the AF gap
203: closes continuously leading to a
204: quantum-critical point (QCP) and no instabilities towards inhomogeneous
205: phases are detected~\cite{ku.ma.02,se.la.05}, suggesting that the pseudogap is
206: directly related to fluctuations of the AF gap.
207: }
208: On the other hand, calculations based upon the three-band Hubbard model
209: yield stripes - although with different properties - even in n-doped
210: cuprates~\cite{sa.gr.00}.
211:
212: %%
213: %%
214: %%
215: %%
216: %%
217: %%
218: It is an important issue
219: to understand
220: whether the closing of the AF gap in n-doped cuprates, and thus
221: the evolution from the AF to the SC state may be
222: related to an instability towards inhomogeneous phases,
223: in qualitative analogy to p-doped materials, or whether
224: this evolution is more continuous, as the phase diagram of n-doped
225: cuprates seems to suggest~\cite{na.he}.
226: %%
227: %%
228: %%
229: %%
230: %%
231: %%
232: %%
233: %%
234: %%
235: %%
236: %%
237: %%
238: %%
239: %%
240: %%
241: %%
242: %%
243: %%
244:
245: In this paper, we address the issue of the AF to SC transition
246: by an analysis of the
247: single-band Hubbard
248: model for parameters appropriate to n- and p-doped compounds
249: via {\em variational cluster perturbation theory}
250: (VCPT)~\pcite{po.ai.03,pott.03,da.ai.04}. VCPT is appropriate to deal with
251: strongly-correlated systems, since the short-range interaction part is
252: solved exactly within a small cluster.
253: A similar analysis has been recently carried out
254: by S{\'e}n{\'e}chal {\em et al.}~\pcite{se.la.05}. In that paper,
255: the authors show that the single-band Hubbard model is sufficient to
256: explain the different shape of the phase diagram between p- and n-doped
257: cuprates. Their results also suggest that the AF to SC transition
258: is continuous and associated with a quantum critical point.
259: %%
260: %%
261: %%
262: %%
263: %%
264: %%
265: {\em In contrast,} our results show
266: that the issue of phase separation for the AF to SC transition
267: in n-doped materials
268: {\em requires a high
269: accuracy} in the determination of energies, which
270: is possible within the present work only thanks to
271: two technical improvements of the VCPT calculations, which we are
272: going to describe in detail below.
273: \cha{
274: %%
275: The interplay between AF and SC was also studied previously via a
276: number of numerical techniques, see, e. g. Ref.~\cite{ga.ch.05}.
277: %%
278: }
279:
280: Our main results concerning the nature of the AF to SC transition
281: (see Fig.~\ref{vsmu})
282: %%
283: \begin{figure}[t]
284: \includegraphics[width=\textwidth]{vsmue}
285: \caption{
286: Difference between the $T=0$ grand-canonical potentials
287: $\Delta \Omega\equiv
288: \Omega_{SC}-\Omega_{AF+SC}$
289: of the SC and of the AF+SC
290: phases (solid line)
291: plotted as a function of chemical potential $\mu/t$.
292: $m$ gives the staggered magnetization (dashed),
293: %%
294: $x$ the doping (dotted, scaled by a factor $4$ for convenience) in the AF+SC phase.
295: Data are for
296: the Hubbard model
297: with $U=8 t $ and $t'/t=-0.3$ for hole and electron doping.
298: Right: blowup of the region around the transition for the n-doped
299: case: Here, the critical chemical potential $\mu_c$ at which
300: $\Delta\Omega =0$ is $\mu_c\approx 6.189$, and
301: %%
302: $\mu^*\approx 6.196$ (see text for its meaning).
303: %%
304: In both p-doped and n-doped system, $m$ is finite at the transition
305: point, and has a jump to $m=0$.
306: }
307: \label{vsmu}
308: \end{figure}
309: are the following:
310: In p-doped systems the situation is quite clear: the transition is
311: first order in $\mu$, i.e., there is an
312: instability towards
313: phase separation with a jump in doping $x$ and staggered
314: magnetisation $m$ at the transition ($\mu_c$ in Fig.~\ref{vsmu}).
315: %%
316: We stress again that by allowing for a more general
317: spatial dependences of the order parameter, this macroscopic phase
318: separation could be possibly replaced by other microscopically inhomogeneous
319: phases, such as, e.g., stripes. Certainly, this is expected to be the
320: case if long-range Coulomb interaction is taken into account~\pcite{ca.em.03}.
321: On the other hand, in contrast to previous theoretical calculations,
322: our results
323: {\em also
324: suggest phase separation in the
325: n-doped case}, although the corresponding energy scale
326: (see blowup in Fig.~\ref{vsmu})
327: is one order of
328: magnitude smaller than in p-doped compounds.
329:
330: %%
331: In agreement with Ref.~\cite{se.la.05}, we find that the AF phase
332: actually mixes with
333: a weak d-wave SC component at small doping.
334: %%
335: A similar coexistence
336: phase
337: is observed, for example, in PrCeCuO~\cite{da.ka.05}, and
338: was obtained in several
339: calculations~\pcite{li.ka.00, in.do.88}.
340: On the other hand, such a cohexistence obtained at the mean-field
341: level could be
342: just a signal for the presence of
343: strong pairing fluctuations within the
344: AF phase, as we discuss below.
345: \cha{
346: Upon increasing doping, there is a
347: transition to a pure d-wave SC phase.
348: %%
349: At larger doping,
350: some experiments and theories suggest a transition to a SC phase with
351: different pairing symmetries in n-doped cuprates, such as
352: $s$ or mixed $s$ and $d$ phases~\cite{st.kr.95,ch.jo.90,sk.ki.02}.
353: However, this high-doping regime in beyond the scope of the present paper.
354: }
355:
356: \section{Method}
357: We consider the two-dimensional single-band Hubbard model with nearest-neighbor ($t$)
358: and next-nearest-neighbor ($t'$) hoppings, and Hubbard repulsion $U$.
359: We take typical parameters valid for both hole- and electron-doped high-$T_c$
360: cuprates~\pcite{ki.wh.98}, namely $t'/t = -0.3$ and $U/t=8$.
361: \cha{
362: %%
363: Although
364: some slight difference have been suggested between the
365: parameter values
366: ($t'/t$ and $U/t$) of n- and p-doped compounds, we keep them equal in
367: order to {\em just focus on the differences due to hole- and electron-
368: doping only}.
369: We stress that
370: the aim of our calculation is a general
371: %%
372: description of the materials,
373: so that we don't aim at carrying out an accurate fitting of the parameters.
374: We have in fact checked that
375: different values for the parameters (e. g., a different $U/t$ or $t'/t$
376: within $30\%$) or
377: a third-neighbor hopping does not affect our
378: qualitative
379: conclusions.
380: %%
381: %%
382: }
383:
384: %%
385: Within the CPT approach~\pcite{gr.va.93,se.pe.00},
386: the lattice is partitioned into disconnected clusters. The
387: Hamiltonian $H'_{\rm CPT}$ of the disconnected lattice can be solved
388: by exact diagonalization. The intercluster hopping Hamiltonian $V_{\rm CPT}$
389: is then
390: treated perturbatively up to first order for the self-energy.
391: VCPT (or SFA)~\cite{pott.03,da.ai.04}
392: generalizes this approach
393: by decomposing
394: $H$
395: into a ``reference'' part $H'$ and a ``perturbation''
396: $V$, the latter containing single-particle terms only.
397: In general, one can take $H'=H'_{\rm CPT}+\Delta H$, i. e., add to
398: the cluster an {\em arbitrary
399: single-particle Hamiltonian}
400: $\Delta H$, as long as $H'$ is still exactly solvable numerically.
401: In order to conserve the total physical Hamiltonian
402: $H=H'_{\rm CPT}+V_{\rm CPT}$, $\Delta H$ must then be
403: subtracted from $V$~\pcite{da.ai.04}, which requires
404: $V=V_{\rm CPT}-\Delta H$.
405: This extension allows for the description of symmetry-broken phases by
406: introducing, via $\Delta H$, a corresponding symmetry-breaking field.
407: \cha{
408: %%
409: Notice that, due to this subtraction the field $\Delta H$ is only fictitious.
410: %%
411: However, due to the fact that the subtraction is only carried out
412: in a perturbative way, results in fact do depend on $\Delta H$, except for the
413: case where CPT becomes exact.
414: }
415: %%
416: %%
417: This apparent ``arbitrariness'' in the choice of $\Delta H$
418: is restricted by the requirement that the
419: %%
420: {\it self-energy functional~\cite{pott.03}}
421: (which corresponds to the CPT grand-canonical potential)
422: \beq
423: \label{omega}
424: \Omega = \Omega'+ {\rm Tr} \ln (-G_{\rm CPT}) - {\rm Tr} \ln (-G')
425: \eeq
426: has to be stationary with respect
427: to $\Delta H$.
428: In \eqref{omega}, $\Omega'$ and $G'$ are the grand-canonical potential
429: and the Green's function of the reference system
430: $H'$,
431: respectively, and $G_{\rm CPT}$ is
432: the Green's function of the physical system $H$ calculated
433: perturbatively in $V$~\cite{da.ai.04}.
434: Physically, one can consider VCPT as an extension of CPT in which the
435: intercluster perturbation is not carried out with respect to the
436: exact cluster ground state but with respect to an
437: ``optimized'' state parametrized by $\Delta H$.
438: %%
439: This is similar in
440: spirit to diagrammatic expansions
441: in which the ``bare'' Green's
442: functions are replaced by ``dressed'' mean-field Green's functions in
443: some symmetry-broken state such as AF or SC.
444:
445:
446:
447: Since we expect to describe both an AF and a d-wave SC phase,
448: in our calculation
449: $\Delta H$ contains both
450: a staggered ($h$) and a nearest-neighbor d-wave pairing field
451: ($\Delta$).
452: More specifically,
453: $\Delta H=\Delta_{AF}+\Delta_{SC} + \Delta_{\epsilon}$,
454: where
455: $
456: %%
457: \Delta_{AF} = h \sum_{\vv R} ( n_{\vv R ,\up} - n_{\vv R ,\down} ) e^{i \vv
458: Q \cdot \vv R}
459: $,
460: and
461: %%
462: $
463: \Delta_{SC}=
464: \frac{\Delta}{2} \sum_{\vv R,\vv R'} \eta(\vv R-\vv R') ( c_{\vv R ,\up} c_{\vv R'
465: ,\down} + h.c.)
466: %%
467: $.
468: Here, $c_{\vv R ,\sigma}$ destroys an electron on lattice site $\vv R$
469: with spin projection $\sigma$. $n_{\vv R ,\sigma}
470: \equiv c^{\dag}_{\vv R,\sigma} c_{\vv R ,\sigma}$ is the corresponding density
471: operator, $\vv Q = (\pi,\pi)$ is the antiferromagnetic wave vector,
472: and the $d$-wave factor
473: $\eta(\vv R-\vv R')$ is non vanishing for nearest-neighbor lattice sites
474: only and is equal to
475: $+1$ ($-1$) for $\vv R-\vv R'$ in $x$ ($y$) direction.
476: In the SC term, the sum is restricted to $\vv R$ and $ \vv R'$
477: belonging to the same cluster.
478:
479:
480: Away from half-filling, in addition to the symmetry-breaking terms,
481: it is necessary to add to the reference system
482: a ``fictitious'' on-site
483: energy, i. e.
484: \beq
485: \Delta_{\epsilon} = \epsilon \ \sum_{\vv R,\sigma } n_{\vv R ,\sigma}
486: \;,
487: \eeq
488: which
489: plays the role of a ``shift'' in the cluster chemical potential
490: with respect to the ``physical'' chemical potential $\mu$.
491: Without this term, which has been omitted in Ref.~\cite{se.la.05},
492: the mean particle density $n=1\mp x$ ($x$ is the doping, $\mp$
493: corresponds to p- and n-doping, respectively) cannot be unambiguously
494: determined, as different results would be obtained by evaluating it as
495: $n=-\de \Omega/\de \mu$ or as the usual trace over the Green's function.
496: Of course, a consistent treatment of the particle density is important
497: for an accurate analysis of phase transitions as a function of doping,
498: and of phase separation.
499: Unfortunately, the inclusion of $\epsilon$ considerably complicates
500: the numerical treatment, as now $\Omega$ is not simply a minimum with respect
501: to its parameters, but a saddle point (typically a maximum in the
502: $\epsilon$ direction).
503: Physically, one can regard $\epsilon$ as a Lagrange multiplier which
504: enforces an appropriate
505: constraint in the particle number~\pcite{details}.
506: Along this
507: constraint, $\Omega $ is then a minimum with respect to $h$ and $\Delta$.
508:
509:
510: For an appropriate characterization of the phase transition, it is necessary
511: to evaluate $\Omega$ with high accuracy. This is difficult to achieve
512: by simply numerically carrying out the frequency integrals
513: contained in the trace in \eqref{omega}.
514: It is rather
515: convenient to exploit the fact that these integrals can be replaced
516: (at zero temperature $T$)
517: by a sum over the negative poles
518: of
519: $G_{\rm CPT}(\vv q)$ and
520: of
521: $G'$ (see Ref.~\cite{pott.03.se} for details).
522: While the latter are directly given by the
523: single-particle and single-hole
524: excitation energies
525: of the cluster, the former
526: can be directly obtained as the eigenvalues
527: of the
528: matrix $h_{\alpha,\beta}(a,b)$ in Eq. (17) of Ref.~\cite{za.ed.02}~\pcite{details}.
529: Since this method requires the evaluation of
530: all cluster single-particle and single-hole states,
531: a complete diagonalization of the cluster Hamiltonian must
532: be carried out. In addition, the dimension of the matrix $h_{\alpha,\beta}(a,b)$
533: is equal to the number of single-particle and single-hole states.
534: This
535: restricts the maximum cluster sizes that can be
536: considered in our VCPT calculation.
537: For this reason, we only consider an infinite
538: lattice split into $2\times2$ clusters as a
539: reference system.
540: We stress again that this is necessary
541: in order to achieve an accuracy that
542: permits the resolution of the tiny phase-separation energy scale in
543: the n-doped case (see blowup in Fig.~\ref{vsmu}).
544: \cha{
545: %%
546: Of course one could argue that finite-size corrections may introduce
547: a even larger inaccuracy. However, while quantitative changes are
548: expected for larger clusters,
549: we observe that our result about phase separation is
550: quite robust and only weakly dependent on other parameters such as
551: $t'/t$ for $0.2<|t'/t|<0.4$, or $U/t$. In addition, our calculation {\em does} show
552: that {\it at least} such an accuracy is required
553: in order to address the issue of phase
554: separation in the n-doped case.
555: %%
556: %%
557: %%
558: }
559:
560: \section{Results}
561: Our calculation, thus, proceeds as follows. For a given value of the
562: variational parameters $\epsilon$, $h$, and $\Delta$ we carry out a
563: complete diagonalization of
564: the cluster Hamiltonian (due to the presence of the pairing field,
565: particle number is not conserved), evaluate $\Omega$
566: as described above,
567: and look for a stationary solution as a function of
568: the variational parameters.
569: Quite generally, different stationary solutions can be found,
570: corresponding, for example, to different phases. In this case, the
571: minimum $\Omega$ selects the most stable phase.
572: Near half-filling, the two most stable solutions are a coherently mixed AF+SC
573: and a pure SC phase, in agreement with Ref.~\cite{se.la.05}.
574: Results for these solutions are plotted in
575: Fig.~\ref{vsmu} for the p- and n-doped cases.
576: The AF order parameter $m$ is plotted as a dashed line
577: for both n- and p-doped systems as a function of $\mu$.
578: $m$ monotonously decreases upon going away
579: %%
580: from half-filling.
581: In Fig.~\ref{vsmu}, we also plot the doping $x$ and the difference
582: between the $T=0$ grand-canonical potentials
583: $\delta \Omega\equiv
584: \Omega_{SC}-\Omega_{AF+SC}$
585: of
586: %%
587: the SC and the AF+SC
588: solutions.
589: %%
590: While at lower doping the AF+SC phase is stable,
591: a crossing to the SC phase
592: occurs at a critical chemical potential $\mu_c$ for which $m$ is
593: still nonzero, { i. e. the transition is first order as a function of $\mu$
594: \em both for the p- as well as for the n-doped case}, as anticipated.
595: Although, from Fig.~\ref{vsmu}, this effect appears to be very small for the n-doped
596: case, we stress that our calculation was carried out with sufficient
597: accuracy to resolve this energy difference.
598: At this value of $\mu$, the doping $x_c$ of the AF+SC phase is
599: different from the one $x_c'$ of
600: the pure SC
601: phase so
602: that there is a jump $\Delta x \equiv x_c'-x_c>0$ in the doping at $\mu_c$ indicating
603: phase separation between a weakly doped AF+SC and a higher doped SC
604: phase.
605: Eventually, for some $\mu=\mu^*$ (further away from half filling than $\mu_c$)
606: the AF+SC solution ceases to exist.
607: %%
608: In principle, an unstable AF+SC solution continues to exist for
609: $x>x_c$. However, in this region $\mu$ is a decreasing function of the particle density,
610: which is an equivalent indication of phase separation.
611:
612: The difference $\Delta\mu \equiv|\mu^* -\mu_c|$ can be considered as a
613: measure of the characteristic
614: phase-separation energy per particle, i.e., it is proportional to the energy
615: barrier
616: between the two doping values $x_c$ and $x_c'$.
617: %%
618: As one can see from Fig.~\ref{vsmu},
619: in the p-doped case this energy scale
620: ($\Delta\mu/t \approx 5 \cdot \ 10^{-2}$)
621: is about an order of magnitude
622: larger than in the n-doped case
623: ($\Delta\mu/t \approx 7 \cdot \ 10^{-3}$).
624: Taking typical values for the energy unity $t$ ($t\approx 0.25 eV$),
625: this corresponds to a
626: temperature scale
627: of $\sim 125 K$ in p-doped,
628: and $\sim 18 K $ in electron- doped cuprates, which
629: is roughly of the order of the
630: corresponding (low-energy) pseudogap temperature scales observed in
631: these materials~\cite{al.kr.03}.
632: Also the discontinuity
633: $\Delta x$ (not shown)
634: is larger in the p-doped case ($\Delta x = 0.11 $, vs. $\Delta x =
635: 0.08 $ for n-doped), although
636: the difference is not significant.
637: At the same time, the doping at which the AF+SC solution is destroyed
638: is much
639: larger in n-doped ($x_c\approx 0.13\%$) than in p-doped systems
640: ($x_c\approx 0.03\%$), in qualitative agreement with
641: experiments~\cite{lu.le.90,na.he}.
642: Therefore, our results suggest that the AF+SC to SC transition as a
643: function of $\mu$ is clearly
644: first order in the p-doped case while weakly first order
645: %%
646: %%
647: for n-doping.
648: %%
649: %%
650: %%
651: %%
652: %%
653:
654: \section{Single-particle spectrum}
655: The different behavior between n- and p-doped systems can be understood
656: by considering the
657: doping evolution of the single-particle spectrum from the
658: AF insulator to a SC displayed in Fig.~\ref{spectr}.
659: \begin{figure}[t]
660: \includegraphics[width=\textwidth]{spectre}
661: \caption{
662: Single-particle spectrum for the hole-doped (a), and for the electron-doped
663: system (b) just before the AF+SC to SC transition, i.e., for
664: dopings
665: $x \approx 0.03$ and $x \approx 0.13$, respectively.
666: }
667: \label{spectr}
668: \end{figure}
669: In agreement with
670: experiments~\pcite{ar.ro.02,ar.lu.01}, in p-doped cuprates doped holes
671: first enter at \pit{}\ \pcite{da.hu.03}.
672: On the other hand, in n-doped materials doped electrons
673: initially form pockets around \pio~\pcite{ar.ro.02,ar.lu.01,ku.ma.02}.
674: At \pio the density of states is larger, and the fact that
675: $\mu$ lies within the SC gap apparently
676: stabilizes the AF solution for a larger doping range, allowing for
677: the AF gap to decrease more gradually in the n-doped case. The transition
678: %%
679: to a non-magnetic solution appears to occur as soon as the Fermi
680: energy touches the top of the band at \pit.
681: At this nodal point the SC gap is zero, so that doping into nodal
682: particles apparently destabilizes the AF solution.
683: This
684: %%
685: appears to be the
686: reason
687: why
688: in the p-doped case
689: the AF solution,
690: where hole are first doped near \pit (see also Ref.~\cite{ku.ma.02}),
691: is stable only in a smaller doping range.
692: %%cut here?
693: Numerically, we observe the following behavior:
694: as long as $\mu$ remains below (above for p-doped) the band around \pit, there is
695: only an absolute
696: minimum of $\Omega$
697: at a finite value $h_{opt}$ of the
698: staggered
699: field $h$ (we keep the other two fields $\Delta$ and $\epsilon$ at
700: their saddle point).
701: As soon as $\mu$ enters
702: the band at \pit, $\Omega(h)$ also develops a local minimum at
703: $h=0$,
704: which becomes rapidly lower in energy than the
705: minimum $h=h_{opt}$. Eventually, the local maximum lying between the two minima at
706: $h=0$ and $h=h_{opt}$ merges into the minimum at
707: $h=h_{opt}$ so that this latter minimum disappears.
708: %%
709: The observation that doping into \pit makes the AF phase unstable
710: suggests that the occurrence of phase separation is generic and quite
711: independent on specific parameters (unless, of course, one adds some
712: doping-dependent potential).
713: Finally,
714: let us mention that the Fermi-surface evolution of the
715: n-doped system as a function
716: of $x$ (not shown) qualitatively reproduces the ARPES
717: experiments~\cite{ar.lu.01}:
718: For small doping, we obtain electron pockets around \pio, while for
719: larger doping our results display an evolution to a large
720: Fermi surface centered around $(\pi,\pi)$.
721:
722: \section{Discussion and conclusions}
723: The VCPT method exactly treats fluctuations up to the range of the cluster
724: size, so that
725: %%
726: the question arises whether the SC solution we (and also
727: others~\cite{ch.jo.90,li.ka.00,se.la.05})
728: obtain within the AF phase is a
729: true long-range SC phase or whether it is only a signal of strong pairing fluctuations
730: within the AF phase leading to a SC pseudogap.
731: \cha{
732: %%
733: The latter hypothesis could be supported by the fact that
734: results obtained with different cluster
735: sizes~\pcite{se.la.05}
736: %%
737: seem to indicate a size dependence of the SC order parameter, and by
738: the fact that the SC order parameter is about a factor three smaller in the AF+SC
739: phase than in the pure SC one.
740: The presence or not of such microscopic cohexistence phase may depend
741: on material details.
742: Certainly, our results suggest that the
743: SC gap (or pseudogap in the case of fluctuations) $\Delta$
744: is important in order to stabilize the AF phase in n-doped materials.
745: %%
746: %%
747: %%
748: %%
749: %%
750: %%
751: %%
752: %%
753: %%
754: %%
755: %%
756: %%
757: %%
758: %%
759: On the other hand, $\Delta$ hardly has an effect in p-doped systems
760: since it does not produce any gap at the Fermi Surface,
761: as doped holes first enter near the nodal points.
762: }
763:
764: In conclusion, our calculations suggest that
765: the destruction of the AF phase in n-doped systems is associated with a
766: tendency towards phase separation, similarly to p-doped materials, although
767: with an energy scale which is one order of magnitude smaller.
768: In the presence of Coulomb interaction,
769: %%
770: this tendency is expected to produce a microscopically segregated phase,
771: such as stripes or similar inhomogeneous structures~\cite{sa.gr.00}
772: Following the idea that
773: the pseudogap temperature is related to the formation of such
774: segregated phases~\cite{ca.em.03}, our results
775: provide reasonable orders of magnitudes for the pseudogap temperature scale, and,
776: in particular, provide a possible explanation why
777: this is much smaller in n-doped than in p-doped
778: cuprates.
779: \cha{
780: %%
781: The results of the present paper
782: support the idea
783: that the asymmetry between hole- and
784: electron-doped materials can be understood within a simple {\em single-band
785: Hubbard model}, at least
786: concerning the portion of phase diagram we
787: studied here. Of course, it remains an open issue, whether
788: a three-band model may be necessary for a complete description.
789: In fact, alternative theories propose that Ce- doping
790: in n-doped compounds, such as NdCeCuO,
791: occurs in a total different way. The idea there is that doped Ce atoms do not add
792: electrons to the CuO$_2$-layers, but rather are responsible for a
793: quenching of Cu moments~\cite{bask.05u}.
794: }
795: %%
796: %%
797: %%
798: %%
799: %%
800: %%
801: %%
802: %%
803: %%
804: %%
805: %%
806: %%
807: %%
808: %%
809: %%
810: %%
811: %%
812: %%
813: %%
814:
815: %%
816: %%
817: %%
818: %%
819: %%
820: %%
821: %%
822: %%
823: %%
824: %%
825: %%
826: %%
827:
828: \acknowledgments
829: We thank M. Potthoff and L. Alff for discussions and
830: useful suggestions.
831: This work is partially supported by the Doctoral Scholarship
832: Program of the Austrian Academy of Sciences (M.A.) and by the
833: DFG {\it Forschergruppe} n. 538.
834:
835: %%
836: \bibliographystyle{prsty-etal}
837: %%
838: \bibliography{footnotes,references_database}
839:
840:
841: \end{document}
842:
843: %%
844: %%
845: %%
846: %%
847: %%
848: %%
849: %%
850: %%
851: %%
852: %%
853: %%
854: %%
855: %%
856: %%
857: %%
858: %%
859:
860:
861: