PHYSICAL REVIEW B 76, 184521 共2007兲

s-wave superconductivity phase diagram in the inhomogeneous two-dimensional attractive

Hubbard model
K. Aryanpour,1,2 T. Paiva,3 W. E. Pickett,1 and R. T. Scalettar1
1
Department of Physics, University of California, Davis, California 95616, USA
2
Department of Physics, SUNY at Buffalo, Buffalo, New York, 14260, USA
3Departamento de Fisica dos Sólidos, Instituto de Física, Universidade Federal do Rio de Janeiro, Cx.P. 68528, 21945-970,

Rio de Janeiro, RJ, Brazil

共Received 2 August 2007; published 30 November 2007兲
We study s-wave superconductivity in the two-dimensional square-lattice attractive Hubbard Hamiltonian
for various inhomogeneous patterns of interacting sites. Using the Bogoliubov–de Gennes mean-field approxi-
mation, we obtain the phase diagram for inhomogeneous patterns in which the on-site attractive interaction Ui
between the electrons takes on two values Ui = 0 and −U / 共1 − f兲 共with f the concentration of noninteracting
sites兲 as a function of average electron occupation per site, n, and study the evolution of the phase diagram as
f varies. In certain regions of the phase diagram, inhomogeneity results in a larger zero-temperature average
pairing amplitude 共order parameter兲 and also a higher superconducting critical temperature Tc, relative to a
uniform system with the same mean interaction strength 共Ui = −U on all sites兲. These effects are observed for
stripes, checkerboard, and even random patterns of the attractive centers, suggesting that the pattern of inho-
mogeneity is unimportant. The phase diagrams also include regions where superconductivity is obliterated due
to the formation of various charge-ordered phases. The enhancement of Tc due to inhomogeneity is robust as
long as the electron doping per site, n, is less than twice the fraction of interacting sites 关2共1 − f兲兴 regardless of
the pattern. We also show that for certain inhomogeneous patterns, when n = 2共1 − f兲, increasing temperature
can work against the stability of existing charge-ordered phases for large f and, as a result, enhance Tc.

DOI: 10.1103/PhysRevB.76.184521 PACS number共s兲: 74.81.⫺g, 71.10.Fd, 74.20.⫺z, 74.25.Dw

I. MOTIVATION A variety of physically relevant models such as the repul-

sive Hubbard and t-J Hamiltonians have been extensively
Fascination with inhomogeneous superconducting 共SC兲 studied to understand the interplay between spatial inhomo-
phases extends back several decades, with many conferences geneity, magnetism, and superconductivity.21–40 In the repul-
and monographs having been devoted to the subject.1–3 For sive Hubbard and t-J Hamiltonians in particular, inhomoge-
conventional superconductors, the inhomogeneities were ex- neity has been introduced either through the hopping
trinsic, arising from a granular nature of samples or due to amplitude t or magnetic coupling J or the local energy on the
the deliberate synthesis of disordered-built materials or films. lattice sites. For the 2D square lattice these two models are
The high-temperature superconductors 共HTSs兲 introduced known to display antiferromagnetism at half-filling and, al-
new aspects into this area of study. First, inhomogeneous though it is less certain, perhaps also d-wave superconduc-
states 共normal and SC兲 seem to be intrinsic to HTSs, at least tivity when doped. There is considerable evidence that they
in the underdoped regime, similar to quenched disorder in also might possess inhomogeneous stripe or checkerboard
the metal-insulator transition in two dimensions 共2D兲.4,5 Sec- ground states.24–28,30,31,33,34 Phenomenological d-wave BCS
ond, the inhomogeneity occurs on a smaller length scale of Hamiltonians with spatially inhomogeneous pairing
just a few lattice constants as evidenced by scanning tunnel- amplitude41–43 or lattice site energy43,44 have also been em-
ing spectroscopy 共STS兲 at the nanoscale.6–9 ployed mostly to reproduce the local density of states
In addition, the strong electronic correlation in HTS cu- 共LDOS兲 results obtained from scanning tunneling micros-
prates plays a major role in the elucidation of the inhomoge- copy 共STM兲.45 In addition, there have been theoretical stud-
neous SC state; indeed, the inhomogeneity is widely believed ies of the SC quantum phase fluctuations using the QED3
to arise from the strong intra-atomic interactions that tend to effective theory of the HTSs in relation with the inhomoge-
frustrate bandlike conduction, to induce local magnetic mo- neous pattern formation in cuprates from the STM experi-
ments, and to drive charge and spin order on a few lattice mental results.46,47
constant scale. The study of the doped Mott insulating phase While density matrix renormalization group 共DMRG兲
has been one of the most active areas of theoretical study of treatments24 provide detailed information on the real-space
HTSs, one that has proven to be unexpectedly complex and charge, spin, and pairing orders, the precise nature of the
rich and which is still under intense exploration and debate. interplay, and whether the different orders compete or coop-
Surprisingly, holes doped into the high-temperature super- erate, remains unclear. In addition, the enhancement of the
conductors tend to arrange themselves nonuniformly in the superconducting transition temperature Tc by local inhomo-
CuO2 planes in the form of stripes, checkerboard, or perhaps geneity has been demonstrated by Martin et al. in Ref. 48
even more exotic structures.10–14 Moreover, spatially varying and also in Ref. 49. Recently, the XY model Hamiltonian
density and spin structures have also been observed in the with certain types of inhomogeneous patterns for the
physics of the manganites15–18 and cobaltites.19,20 coupling constant between spins sitting on two nearest-

1098-0121/2007/76共18兲/184521共15兲 184521-1 ©2007 The American Physical Society

ARYANPOUR et al. PHYSICAL REVIEW B 76, 184521 共2007兲

neighboring sites has also been shown to enhance Tc by Loh H = − t 兺 共ci†␴c j␴ + c†j␴ci␴兲 − ␮ 兺 ci†␴ci␴ − 兺 兩Ui兩ni↑ni↓ ,
and Carlson in Ref. 50. 具ij典,␴ i␴ i
Many of the basic characteristics of this short-range-scale 共1兲
inhomogeneous superconducting state can be addressed with
a more tractable model, one which is well understood in the with t the hopping amplitude, ␮ the chemical potential, and
homogeneous limit. This model is the attractive Hubbard Ui the local attractive interaction between the fermions of
model, which has been applied previously to address some opposite spins residing on the same lattice site i. We will
aspects of the impact of inhomogeneity. Recently old sugges- study inhomogeneous patterns in the interaction Ui. The ori-
tions based on “negative-U” superconductivity have been gin of the attraction in this model can result from, for ex-
revived,49,51 which may provide additional applications for ample, integrating out a local phonon mode.55 The two-
the results we present in this paper. Tl-doped PbTe achieves dimensional uniform attractive Hubbard model is known to
a critical temperature up to 1.5 K, and more extensive heat yield degenerate superconductivity and charge-density-wave
capacity and transport data52,53 have led to an analysis in 共CDW兲 long-range order at half-filling and zero
terms of a “charge Kondo effect” that could be linked to the temperature.56–58 However, away from half-filling, the CDW
observed superconductivity.54 This system intrinsically in- pairing symmetry is broken and superconductivity is more
volves both negative-U centers and inhomogeneity. favorable, and the SC phase transition is at finite tempera-
This article extends previous work49 to a more general ture.
range of noninteracting site concentration f values. We show The BdG mean-field decomposition bilinearizes the
the presence of different conduction phases both in the phase Hamiltonian by replacing the local pairing amplitude and
diagram at T = 0 and in the density of states 共DOS兲. We also local density by their average values, ⌬i = 具ci↑ci↓典 and 具ni␴典
show the local occupation and SC order parameter for elec- = 具ci†␴ci␴典 and yields the quadratic effective Hamiltonian
trons on different lattice sites as the concentration f varies
for different inhomogeneity patterns. Finally, the Tc enhance- Hef f = − t 兺 共ci†␴c j␴ + c†j␴ci,␴兲 − 兺 ␮
˜ ici†␴ci␴ − 兺 兩Ui兩
ment conditions are also extracted based on the relationship 具ij典,␴ i␴ i
between the average doping of electrons, n, on the lattice and *
inhomogeneity concentration f. ⫻关⌬ici↑
† †
ci↓ + ⌬i ci↓ci↑兴, 共2兲
The paper is organized as follows: in the next section we
introduce our model and describe the method we have em- where ␮ ˜ i = ␮ + 兩Ui兩具ni典 / 2 includes a site-dependent Hartree
ployed. In Sec. III we present and discuss the phase diagram shift with 具ni典 = 兺␴具ni␴典. All energies will be referenced to t
at zero temperature. Section IV contains our finite- = 1.
temperature results, and Sec. V summarizes our findings. We adopt the criterion of comparing the tendency for su-
perconductivity in the homogeneous system with the same
attraction −U on all lattice sites, with cases when sites with
II. MODEL AND METHODOLOGY attraction are mixed with sites where the attraction is
absent—i.e., Ui = 0.48,59,60 Specifically, we have studied sys-
tems in which sites with attractive interaction are randomly
This article focuses on a general question: Under what
distributed59 or arranged in checkerboard and stripe patterns.
conditions is it more favorable to have an inhomogeneous
The last two regular patterns have been purposely chosen
pairing attraction, compared to the same average strength
due to their relevance to the experimentally observed pattern
spread homogeneously over the lattice? By “conditions” we
formation in the HTS cuprates.
refer to the average occupation number of electrons per site,
Figure 1, panel 共a兲, presents the patterns for the interact-
n; the average attraction strength per lattice site, Ū, which ing lattice sites with four different values for the fraction of
remains constant in any comparison between systems with noninteracting sites, f. The uniform pattern corresponds to
and without inhomogeneity; and the inhomogeneity concen-
f = 0 with interaction Ui = Ū = −U on all lattice sites. Check-
tration f. We address this question by comparing the average
¯ over the entire lattice erboard, stripes, and random patterns with f = 0.25 include
zero-temperature pairing amplitude ⌬
1 − f = 75% interacting sites with Ui = 34 Ū = − 34 U and f = 25%
and the SC transition temperature Tc for a system in the
noninteracting sites with Ui = 0. For f = 0.5, half of the sites
presence and absence of inhomogeneity.
are interacting with Ui = −2U and half noninteracting with
For the cuprate superconductors, for example, such a
Ui = 0. f = 0.75 corresponds to 25% interacting sites with Ui
question is complicated by the presence of other types of
= −4U and 75% noninteracting sites with Ui = 0, once again
order such as antiferromagnetism and exotic spin-gap phases
and by the nontrivial d-wave symmetry of the SC order pa- averaging to Ū = −U per site.
rameter. For these systems and phenomena, models like the For the random pattern we have averaged over typically
repulsive Hubbard or t-J Hamiltonians are essential.21–38 20 different disorder realizations. One may note that regard-
Nevertheless, it is yet beneficial to study the problem first by less of the pattern and the impurity fraction, the average
employing a more simple and phenomenological model. attraction per site—i.e., Ū = −U—remains constant. We adopt
Here we will present a solution of the inhomogeneous this criterion throughout the rest of this article for compari-
Bogoliubov–de Gennes 共BdG兲 equations for the attractive son between uniform and inhomogeneous lattices. This cri-
Hubbard Hamiltonian, terion is believed to be most appropriate for exhibiting the

184521-2
s-WAVE SUPERCONDUCTIVITY PHASE DIAGRAM IN… PHYSICAL REVIEW B 76, 184521 共2007兲

and four different types of sites at f = 0.5 and f = 0.25 or 0.75,

f = 0.0
respectively, as illustrated by different colors and numbers in
(uniform) panel 共b兲 of Fig. 1. Sites carrying the same color or number
are equivalent by the symmetry in the pattern geometry. For
the random pattern, due to the lack of both regularity and
periodicity, we average over all the interacting and all the
noninteracting sites separately 关black and red 共red shows
f = 0.25 dark gray in the grayscale version兲 regions in panel 共c兲 of
(checkerboard) (stripes) Fig. 1, respectively兴 before the configurational averaging
over all different impurity patterns is performed.
We self-consistently diagonalize the BdG mean-field
Hamiltonian in Eq. 共2兲 by assigning initial values to the local
occupation number ni and order parameter ⌬i and solving
f = 0.5 again for these parameters after diagonalizing the Hamil-
(a) (checkerboard) (stripes) tonian, until convergence is achieved at a desired tolerance.
For the checkerboard and striped patterns, sites with the
same color have the same density, due to symmetry, and do
not need to be averaged. For the random pattern we calculate
the average occupation number ncolor and order parameter
f = 0.75 ⌬color per site at the different types of sites by averaging
(checkerboard) (stripes)
1
ncolor = 兺
Ncolor i苸兵color其
ni ,

1
f = 0.5
2 1 2 ⌬color = 兺
Ncolor i苸兵color其
⌬i . 共3兲
1
1 2
(b)
For the checkerboard and striped patterns with f = 0.5 and
4 for the random pattern 共all f兲 we simply have nattrac = nblack
2 3
3 and nfree = nred 共red shows dark gray in the grayscale version兲.
f = 0.25 and 0.75 2 For the striped and checkerboard with f = 0.25 we have
1 4 1 nattrac = 共ngreen + 2nred兲 / 3 共green shows light gray in the gray-
scale version兲 and nfree = nblack, whereas for f = 0.75 we have
U 1 = (4/3)U U 1 = 2U U 1 = 4U nattrac = nblack and nfree = 共ngreen + 2nred兲 / 3. The same combina-
(c) random tions hold for ⌬attrac and ⌬free. The average occupation num-
ber n and order parameter ⌬ ¯
U2 = 0 U2 = 0 U2 = 0 inhom per site are defined
f = 0.25 f = 0.5 f = 0.75 n = 共1 − f兲nattrac + fnfree ,
FIG. 1. 共Color online兲 Panel 共a兲: regular patterns for the inter-
¯
⌬
acting sites in the attractive Hubbard model at different inhomoge- inhom = 共1 − f兲⌬attrac + f⌬free . 共4兲
neity concentration values. Stripes and checkerboard have been par-
The chemical potential ␮ in Eq. 共2兲 is self-consistently ad-
ticularly selected because of their relevance to the experimental
justed after every iteration in order to arrive at a desired total
observations in cuprates. Panel 共b兲: color coding and numbering
different types of sites for the checkerboard and stripes blocks as
average occupation per site, n, for the entire lattice. For the
presented in panel 共a兲 共two colors for f = 0.5 and four for f = 0.25 regular patterns—i.e., uniform, checkerboard, and stripes—
and 0.75兲. Sites carrying identical color code and number are due to their periodicity, by Fourier transforming the Hamil-
equivalent by the symmetry in the lattice geometry. Panel 共c兲: color tonian into momentum space, we significantly reduce the nu-
coding of the lattice sites for the random inhomogeneous pattern at merical cost of the calculations and at the same time can
different f values. Regions of interacting 共noninteracting兲 sites are increase the size of the lattice close to the thermodynamic
coded black 共red or dark gray in the grayscale version兲 with the limit to avoid finite-size artifacts in the results 共up to 1500
appropriate weight of 1 − f共f兲. ⫻ 1500 lattice sites in our calculations兲. For the random pat-
tern, however, such a simplification is not possible due to the
effect of inhomogeneity in particular in the systems having lack of periodicity. Hence, we are limited to the finite-size
the same strength on average for forming superconducting lattices of up to 24⫻ 24 sites. As a result, especially at small
Cooper pairs.
Panel 共b兲 depicts the color coding and numbering of the values for the average on-site interaction magnitude 兩Ū兩,
lattice sites for the checkerboard and stripes based on their finite-size effects are to be cautiously monitored. Our calcu-
value of f. Due to the regular geometry and periodicity of lations also include the DOS for the entire lattice. We study
their inhomogeneous patterns, lattices with the checkerboard simultaneous variations of the average on-site interaction
and stripe patterns can be classified into blocks including two magnitude 兩Ū兩, occupation number n, and also temperature

184521-3
ARYANPOUR et al. PHYSICAL REVIEW B 76, 184521 共2007兲

and their effects in the average order parameter ⌬ ¯ the lower limit of 兩Ū兩 = t. Beyond n = 1.5, the system becomes
inhom and
DOS. Our goal is to obtain the phase diagram for the effect metallic. For stripes as shown in panel 共b兲, similar features as
of inhomogeneity in superconductivity and discuss the con- in panel 共a兲 are observed. One exception is the lack of the
ditions under which inhomogeneity can result in enhance- charge-ordered insulating phase at n = 1. This can be the re-
ments in the average superconductivity order parameter or sult of further overlap between the Cooper pairs, since for
SC phase transition temperature Tc. the stripes, nearest-neighboring sites are interacting in one
It is further realized that our conventional mean-field ap- dimension. The charge-ordered insulating phase at n = 1.5
proach does not capture the Kosterlitz-Thouless nature of the
also forms for rather higher 兩Ū兩 values compared to the
phase transition in two dimensions. Nevertheless, this weak-
ness can be repaired61 upon regarding the local pairing am- checkerboard pattern. The random pattern in panel 共c兲 also
plitudes as complex variables and performing a finite- shows features similar to the stripes.
temperature Monte Carlo integration over the associated The second row 关panels 共d兲–共f兲兴 corresponds to f = 0.5
amplitude and phase degrees of freedom. Unlike BCS, this with rather similar features to the first row. As anticipated,
Monte Carlo mean-field 共MCMF兲 approach allows identifi- superconductivity gradually goes away above n = 1 = n* for
cation of the weak- and strong-coupling regimes via the all the patterns. For the checkerboard in panel 共d兲 and n ⬎ 1,
phase correlation function. In an earlier work49 this Monte superconductivity strictly goes away and the system turns
Carlo technique was employed as an independent examina- metallic. For the striped and random patterns, however, su-
tion for the validity of our results and the agreement between perconductivity persists slightly above n = 1 until it is totally
the two techniques was clearly confirmed. obliterated. At n = 1, all three inhomogeneous patterns exhibit
a charge-ordered insulating phase for large enough values of
III. PHASE DIAGRAM AT T = 0 兩Ū兩 共or the smallest value of 兩Ū兩 for the checkerboard兲. Nev-
ertheless, it can be readily seen that for f = 0.5 compared to
Figure 2 presents the phase diagram for the average inter- f = 0.25, the enhancement of the average order parameter due
action magnitude 兩Ū兩 and electron doping n per site at T = 0 to inhomogeneity is considerably larger as the ratio r in-
for three different inhomogeneous patterns of checkerboard, creases for small 兩Ū兩 values.
stripes, and random and for f values of 0.25, 0.5, and 0.75. The third row 关panels 共g兲–共i兲兴 for f = 0.75 shows ratios as
We show isocontours of r = ⌬ ¯
inhom / ⌬uniform—i.e., the ratio of large as r = 15 for small values of 兩Ū兩 and n values not much
the average inhomogeneous pattern order parameter over its larger than 0.5= n*. For f = 0.75, also, superconductivity
uniform pattern counterpart. Thus, whenever r ⬎ 1, inhomo- gradually dies away when n ⬎ 0.5= n* and a charge-ordered
geneity leads to a larger average order parameter at T = 0
phase sets in for large enough 兩Ū兩 values at n = 0.5= n* analo-
compared to a homogeneous system and therefore is more
gous to f = 0.25 and 0.5. The only difference is a slight rem-
favorable for superconductivity over a uniform pattern of the
nance of superconductivity for the checkerboard pattern at
interacting sites. We also adopt the lower limit of 兩Ū兩 = t, n ⬎ 0.5. Thus, by further diluting the interacting sites in the
since for smaller values of 兩Ū兩, r will be the ratio of two very lattice and keeping the attractive pairing energy constant at
small numbers and is subject to numerical uncertainty. The the same time, superconductivity is driven towards smaller n
first row of Fig. 2 关panels 共a兲–共c兲兴 corresponds to the con- values; on the other hand, the average order parameter be-
centration value of f = 0.25 for the noninteracting sites. At comes significantly more enhanced due to inhomogeneity.
first glance, one can observe that regardless of the geometry Generally, regardless of the pattern, for large enough 兩Ū兩 val-
for the inhomogeneity, above n = 1.5= 2共1 − f兲, inhomogene- ues, inhomogeneity weakens superconductivity for every
ity gradually 关or abruptly for the checkerboard pattern in value of n due to the localization and compression of the
panel 共a兲兴 results in the obliteration of superconductivity Cooper pairs in the interacting sites. For n ⬍ n*, ⌬ ¯
inhom in-
consistent with the findings of Litak and Györffy59 We can
understand how this obliteration takes place if we examine creases as a function of 兩Ū兩 and saturates for large 兩Ū兩 values.
For n 艌 n*, ⌬ ¯
the behavior of the system in strong coupling. When we start inhom reaches a maximum as a function of 兩Ū兩
with an empty system and add electrons they are placed on and eventually vanishes for large enough 兩Ū兩 values. How-
the attractive sites first due to the strong attractive interac- ever, ⌬uniform is a monotonically increasing function of 兩Ū兩
tions. It is useful to define n* = 2共1 − f兲, which for a given f and is symmetric with respect to n = 1. Therefore, for suffi-
corresponds to the density for which all attractive sites are ¯
ciently large 兩Ū兩, the ratio r = ⌬ inhom / ⌬uniform becomes less
doubly occupied and all free ones are empty. As we will see
than 1 while n ⬍ n and eventually 0 when n 艌 n* as illus-
*
below, this density corresponds to an insulating charge-
trated in Fig. 2.
ordered state. Above this density, superconductivity is re-
The obliteration of superconductivity is associated with
duced because the pairs cannot move within the attractive
the vanishing of the average order parameter ⌬ ¯
sublattice, since it is completely filled. inhom = 0.
For the checkerboard pattern in panel 共a兲, there are two Whether a non-SC state is a metal or insulator is determined
insulating regions within the phase diagram at n = 1 and 1.5 by the DOS results for that state. In Fig. 3, panel 共a兲 presents
共hatched orange lines兲, both corresponding to the formation the DOS results for the checkerboard pattern at 兩Ū兩 = 6t 共the
of charge-ordered phases of electrons in the interacting sites. largest in our calculations for f = 0.25兲 and four different val-
No superconductivity was observed for n = 1 and 1.5 down to ues for the average electron doping n. There is a gap in the

184521-4
s-WAVE SUPERCONDUCTIVITY PHASE DIAGRAM IN… PHYSICAL REVIEW B 76, 184521 共2007兲

6 6 6
1 0 SC 0 (a) 1 0.75 0.5 0 (b) 1 0.75 0.5 0 (c)
0.75
5 5 5 1.25 SC
SC SC
metallic
4 4 4 metallic
1.25 0.5 1.25 metallic

3 1.5 1.5

U
U

3 1.5 3
1.75
0.75 1.75
2 1.75 1 2 2
1
2
2 1.25
1 1.5 1 1
checkerboard, f = 0.25 charge ordered insulator stripes, f = 0.25 charge ordered insulator random, f = 0.25 charge ordered insulator
0 0 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
n n n
6 6 6
1 0.75 0.5 0 (d) 0.75 0.5 0 (e) 0 (f)
1 SC
5 charge ordered insulator 5 5 1 0.75 0.5 0.25
1.25 SC 1.25 SC charge ordered insulator charge ordered insulator
4 1.5 4 1.5 4 1.25
2 2 metallic metallic
metallic U 3 3 2 1.5

U
3
U

3 3
4
2 4 2 2 4 3

1 1 1
checkerboard, f = 0.5 stripes, f = 0.5 random, f = 0.5
0 0 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
n n n
6 6 6
0.5 0 (g) 0.5 0 (h) 0 (i)
5 SC charge ordered insulator 5 SC charge ordered insulator 5 SC charge ordered insulator
0.75 0.75
1 0.75 0.5
4 4 1 4 1
1.5
1.5 metallic metallic
metallic 1.5
3 2 3 2 3
U

U
U

2
5 5
2 10 2 10 2 10 5
15 15
1 1 1
checkerboard, f = 0.75 stripes, f = 0.75 random, f = 0.75
0 0 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
n n n

FIG. 2. 共Color online兲 Panel 共a兲: the contour plot phase diagram for the checkerboard pattern at f = 0.25 and T = 0. The horizontal axis
presents the average occupation of electrons per site n and the vertical axis refers to the average interaction magnitude between two electrons
per site 兩Ū兩. Lines with numbers next to them correspond to different ratios of r = ⌬ ¯
checkerboard / ⌬uniform. Solid lines at r = 1 determine the
enhancement boundary. Dashes along r = 0 lines indicate charge-ordered insulating phase behavior. Dotted lines at 兩Ū兩 = t are lower limits for
the interaction as for too small 兩Ū兩 values, r is an ill-defined quantity. Panel 共b兲: the same results for stripes with r = ⌬ ¯
stripes / ⌬uniform at f
= 0.25 and T = 0. r = 0 line for stripes is diverted towards larger n values at smaller 兩Ū兩 and does not run down to arbitrarily small 兩Ū兩 values
at n = 1.5. Dashes along the r = 0 line for stripes appear only at n = 1.5, and beyond that, r = 0 corresponds to a metallic phase. Panel 共c兲
presents results for the random pattern. Similar to the stripes, for the random pattern, the r = 0 line does not run down to arbitrarily small 兩Ū兩
values at n = 1.5 either. Panels 共d兲, 共e兲, and 共f兲 correspond to f = 0.5. Note that the charge ordered phases for the striped and random patterns
at f = 0.5 again occur only at n = 1 portion of the r = 0 line and beyond that, r = 0 yields a metal. Similarly in panels 共g兲, 共h兲, and 共i兲
corresponding to f = 0.75, all three different inhomogeneous patterns have a r = 0 line portion above n = 0.5 at which the systems become
metallic. Panel 共i兲 also lacks the r = 15 contour due to the finite-size-effect uncertainties at small 兩Ū兩 values.

DOS around the Fermi energy at ␻ = 0 at n = 0.4. This gap respond to insulators for n = 1.5. In the second row of Fig. 3
corresponds to a SC state as for n = 0.4, ⌬¯
inhom ⫽ 0. The gaps 关panels 共d兲–共f兲兴, for all the patterns at f = 0.5 and 兩Ū兩 = 5t 共the
at n = 1 and n = 1.5 both correspond to insulating states as for largest in our calculations for f = 0.5兲, the system is supercon-
both these n values ⌬ ¯
inhom = 0. At n = 1.8, we also find
ducting for n ⬍ 1, insulating for n = 1, and metallic for n ⬎ 1
¯ in conjunction with the results of the second row in Fig. 2.
⌬inhom = 0 according to Fig. 2, panel 共a兲. However, the DOS
at n = 1.8 has a finite value around the Fermi energy at ␻ The third row of Fig. 3 关panels 共g兲–共i兲兴, for f = 0.75 and 兩Ū兩
= 0 as shown in panel 共a兲 of Fig. 3, indicating a metallic = 4t 共the largest of our calculations for f = 0.75兲, confirms the
state. results presented in Fig. 2 关panels 共g兲–共i兲, respectively兴—
In panels 共b兲 and 共c兲 in Fig. 3 for striped and random namely, superconductivity for n ⬍ n*, insulator at n = 0.5
patterns, respectively, gaps at n = 1 do not correspond to in- = n* and large enough 兩Ū兩, and metal for n = 0.65⬎ n* for all
sulating phases as opposed to panel 共a兲 whereas they do cor- inhomogeneous patterns.

184521-5
ARYANPOUR et al. PHYSICAL REVIEW B 76, 184521 共2007兲

6 1.2 1.2
(a) f = 0.25 (b) f = 0.25 (c)
n = 0.4 n = 0.4 U = 6t (stripes) n = 0.4
5 n=1 1 n=1 1 n=1 U = 6t (random)
n = 1.5 f = 0.25 n = 1.5 n = 1.5
n = 1.8 U = 6t (checkerboard) n = 1.8 n = 1.8
4 0.8 0.8
DOS

DOS

DOS
3 0.6 0.6

2 0.4 0.4

1 0.2 0.2

0 0 0
-12 -10 -8 -6 -4 -2 0 2 4 6 8 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10
ω ω ω
9 0.8 0.8
(d) n = 0.5 f = 0.5 (e) n = 0.5 f = 0.5 (f)
8 n = 0.5 n=1 U = 5t (stripes) n=1 U = 5t (random)
n=1 n = 1.2 n = 1.2
7 n = 1.2 0.6 0.6
6
f = 0.5
5 U = 5t (checkerboard)
DOS

DOS

DOS
0.4 0.4
4
3
2 0.2 0.2

1
0 0 0
-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10
ω ω ω
5 0.8 1.2
(g) n = 0.25 f = 0.75 (h) n = 0.25 f = 0.75 (i)
n = 0.25 n = 0.5 U = 4t (stripes) n = 0.5 U = 4t (random)
n = 0.5 1 n = 0.65
4 n = 0.65 n = 0.65
0.6
0.8
3 f = 0.75
U = 4t (checkerboard)
DOS

DOS

DOS
0.4 0.6
2
0.4
0.2
1
0.2

0 0 0
-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12
ω ω ω

FIG. 3. 共Color online兲 Panel 共a兲: density of states 共DOS兲 for the checkerboard pattern at f = 0.25 and 兩Ū兩 = 6t 共the largest in our
calculations for f = 0.25兲 for different values of the average electron occupation per site n. Panels 共b兲 and 共c兲: the same results as in panel 共a兲
for striped and random patterns respectively. Panel 共d兲: results of panel 共a兲 at f = 0.5 and 兩Ū兩 = 5t 共the largest in our calculations for f = 0.5兲.
Panels 共e兲 and 共f兲: the same results as in panels 共b兲 and 共c兲 for striped and random patterns, respectively. Panels 共g兲, 共h兲, and 共i兲 correspond
to f = 0.75 and 兩Ū兩 = 4t 共the largest in our calculations for f = 0.75兲 for the checkerboard, striped, and random patterns, respectively. The
particular selection of colors is for better visibility in both the color and grayscale versions.

The insulating state for the phase diagram in Fig. 1 at all 关sites 2, 3, and 4 in red 共dark gray in the grayscale version兲,
values of f and n and all inhomogeneous patterns is always green 共light gray in the grayscale version兲, and red, respec-
associated with the formation of a charge-ordered state. In tively兴 for large 兩Ū兩 at n = 1 does not approach 2 while for the
Fig. 4, panel 共a兲, for the checkerboard at f = 0.25, for the noninteracting sites 共sites 1 in black兲 it approaches zero.
insulating phase at n = 1, as 兩Ū兩 increases, electrons form For the random pattern in panel 共c兲 of Fig. 4 at f = 0.25,
pairs in the interacting sites with higher symmetry 关sites 2 we plot n̄black and n̄red 共red shows dark gray in the grayscale
and 4 in red 共dark gray in the grayscale version兲兴, leaving the version兲 as defined in Eq. 共3兲. The fact that n̄black does not
noninteracting and lower-symmetry interacting sites 关sites 1
approach 2 as 兩Ū兩 increases 共no pair compression兲 is consis-
and 3 in black and green 共light gray in the grayscale version兲,
tent with the lack of an insulating phase at n = 1 for the ran-
respectively兴 essentially empty. For the insulating phase in
dom pattern at f = 0.25 关Fig. 1, panel 共c兲兴. However, at n
panel 共a兲 at n = 1.5, the lower-symmetry interacting site 关site
= 1.5, both striped and random patterns turn insulating as a
3 in green 共light gray in the grayscale version兲兴 also obtains
result of charge-ordered phase formation as shown in panels
a pair, leaving only the noninteracting site 共site 1 in black兲
共b兲 and 共c兲 in Fig. 4, where their occupation numbers on the
empty. In other words, charges rearrange themselves into or-
dered pair configurations forming an insulating phase. Ac- interacting sites all approach 2 as 兩Ū兩 increases. Charge-
cording to panel 共b兲 in Fig. 1 for stripes, n = 1 does not cor- ordered phase formation is more obvious in the second row
respond to an insulating phase. Panel 共b兲 in Fig. 4 confirms 关panels 共d兲–共f兲兴 and third row 关panels 共g兲–共i兲兴 of Fig. 4 for
this finding as the local occupation of the interacting sites f = 0.5 and f = 0.75, respectively, at large enough 兩Ū兩 values.

184521-6
s-WAVE SUPERCONDUCTIVITY PHASE DIAGRAM IN… PHYSICAL REVIEW B 76, 184521 共2007兲

2 2 2
(a) (b) (c)
U4 = (4/3)U
U3 = (4/3)U
1.5 (checkerboard, f = 0.25, T = 0) 1.5 U2 = (4/3)U 1.5
U1 = 0
U4 = (4/3)U U3 = (4/3)U
n1 (n = 1.0)
1 n1 (n = 1.5) 1 n1 (n = 1.0) 1 U1 = (4/3)U U2 = 0
n

n
U1 = 0

n
U2 = (4/3)U n2 (n = 1.0) n1 (n = 1.5)
n2 (n = 1.5) n2 (n = 1.0) n1 (n = 1.0)
n3 (n = 1.0) n2 (n = 1.5) n1 (n = 1.5)
0.5 n3 (n = 1.5) 0.5 n3 (n = 1.0) 0.5 n2 (n = 1.0)
U = -U n3 (n = 1.5) U = -U n2 (n = 1.5)
U = -U
(stripes, f = 0.25, T = 0) (random, f = 0.25, T = 0)
0 0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
U U U
2 2 2
(d) (e) (f)

1.5 1.5 (stripes, n = 1.0, f = 0.5, T = 0) 1.5 (random, n = 1.0, f = 0.5, T = 0)

U2 = 0 U1 = 2U

U1 = 2U U2 = 0 n 1 U2 = 0 1

n
1
n

U1 = 2U U1 = 2U U2 = 0

U = -U
0.5 (checkerboard, n = 1, f = 0.5, T = 0) 0.5 0.5
n1 n1 n1
n2 n2 U = -U n2 U = -U

0 0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
U U U
2 2 2
(g) (h) (i)

1.5 U2 = 0 U3 = 0 1.5 n1 1.5 n1

n2 n2
n1 U1 = 4U U4 = 0 n3 U2 = 0
n2 U3 = 0
1 1 1

n
n
n

n3 U4 = 0
U1 = 4U U2 = 0
U1 = 4U
(checkerboard, n = 0.5, f = 0.75, T = 0)
0.5 0.5 (stripes, n = 0.5, f = 0.75, T = 0) 0.5 (random, n = 0.5, f = 0.75, T = 0)

U = -U U = -U U = -U
0 0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
U U U

FIG. 4. 共Color online兲 Panel 共a兲: the evolution of the local electron occupation number ni on different lattice sites 关as color coded inside
the blocks in Fig. 1, panels 共b兲 and 共c兲兴 for the checkerboard pattern at f = 0.25 and n = 1 and n = 1.5 关referring to charge-ordered phases in
Fig. 2, panel 共a兲兴 as a function of 兩Ū兩. Panels 共b兲 and 共c兲: the same results as in panel 共a兲 for the striped and random patterns, respectively.
For the random pattern, data are taken by averaging the occupation number over the interacting and noninteracting sites. Panels 共d兲, 共e兲, and
共f兲: the same results at f = 0.5 and n = 1 关charge-ordered phases in Fig. 2, panels 共d兲, 共e兲, and 共f兲兴. Also panels 共g兲, 共h兲, and 共i兲 at f = 0.75 and
n = 0.5 关charge-ordered phases in Fig. 2, panels 共g兲, 共h兲, and 共i兲兴.

The proximity effect for the noninteracting sites neigh- plummet too far down with respect to its maximum as a
bored by the interacting sites plays a key role in the magni- function of 兩Ū兩.
tude of the average order parameter ⌬ ¯
inhom in the inhomoge- Panel 共b兲 shows the same behavior for stripes. In panel
neous lattice. In Fig. 5, panel 共a兲, for the checkerboard at f 共b兲, there is an intersection between ⌬2 and ⌬3 near 兩Ū兩
= 0.25 and n = 0.25, the local order parameter on all the in- ⬇ 3t, indicating that due to the particular symmetry of the
teracting sites 关2, 3, and 4 in red 共dark gray in the grayscale stripes, sites 2 and 3 behave very closely. In panel 共c兲, we
version兲, green 共light gray in the grayscale version兲, and red, have plotted ⌬black and ⌬red 共red shows dark gray in the gray-
respectively兴 increases as a function of 兩Ū兩. The noninteract- scale version兲 as defined in Eq. 共3兲 and it is clear that ⌬red
ing site 共1 in black兲 is also superconducting due to the prox- eventually falls off at large 兩Ū兩 values. In the second row of
imity effect of its neighboring sites. However, its local order Fig. 5 关panels 共d兲–共f兲兴 corresponding to f = 0.5 and n = 0.5,
there are only two different sites for each pattern and the
parameter has a maximum at a critical 兩Ūc兩 value beyond
lattice has a more dilute interacting pattern. As a result, com-
which it decreases as a result of the compression of the Coo-
pared to f = 0.25 results, ⌬¯
per pairs in the interacting sites and therefore their weaker inhom at f = 0.5 tends to saturate for

overlap around the noninteracting site. Thus, ⌬ ¯ large 兩Ū兩 values for all the patterns. In the third row of Fig. 5
inhom on all
these four sites will be larger than its uniform pattern coun- 关panels 共g兲–共i兲兴 for f = 0.75, the lattice is even more dilute in
terms of interacting energy. Therefore, ⌬ ¯
terpart due to this proximity effect as long as the noninter- inhom shows even
acting site local order parameter 共⌬1 in this case兲 does not faster saturation at smaller 兩Ū兩 values.

184521-7
ARYANPOUR et al. PHYSICAL REVIEW B 76, 184521 共2007兲

0.6 0.6 0.6

checkerboard (n = 0.5, f = 0.25, T = 0) (a) stripes (n = 0.5, f = 0.25, T = 0) (b) random (n = 0.5, f = 0.25, T = 0) (c)
0.5 U = (4/3)U U1 = 0 0.5 U4 = (4/3)U 0.5
2 U2 = 0
U3 = (4/3)U U1 = (4/3)U
0.4 0.4 U2 = (4/3)U 0.4

U3 = (4/3)U U4 = (4/3)U U1 = 0

∆
∆

∆
0.3 0.3 0.3
∆1
∆1 ∆2 = ∆4 ∆1
0.2 ∆2 = ∆4 U = -U 0.2 0.2 ∆2
∆3 U = -U U = -U
∆3 ∆stripes ∆random
0.1 0.1 0.1
∆check

0 0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
U U U
0.6 0.6 0.6
checkerboard (n = 0.5, f = 0.5, T = 0) (d) stripes (n = 0.5, f = 0.5, T = 0) (e) random (n = 0.5, f = 0.5, T = 0) (f)
0.5 0.5 0.5
U2 = 0 U1 = 2U
U2 = 0 U1 = 2U U2 = 0
0.4 0.4 U1 = 2U 0.4
U1 = 2U U2 = 0
∆
∆

∆
0.3 0.3 0.3
∆1
∆1
∆1 ∆2
0.2 0.2 ∆2 0.2
∆2 ∆random
U = -U ∆stripes U = -U
0.1 ∆check 0.1 0.1 U = -U

0 0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
U U U
0.6 0.6 0.6
checkerboard (n = 0.25, f = 0.75, T = 0) (g) stripes (n = 0.25, f = 0.75, T = 0) (h) random (n = 0.25, f = 0.75, T = 0) (i)
0.5 U = 0 U1 = 4U 0.5 U1 = 4U 0.5
2 U1 = 4U U2 = 0
U2 = 0
0.4 0.4 U3 = 0 0.4
U3 = 0 U4 = 0 U4 = 0
∆

∆
0.3 U = -U 0.3 0.3
∆1 ∆1 U = -U ∆1 U = -U
∆2 = ∆4 ∆2 = ∆4 ∆2
0.2 0.2 0.2
∆3 ∆3 ∆random
0.1 ∆check 0.1 ∆stripes 0.1

0 0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
U U U

FIG. 5. 共Color online兲 Panel 共a兲: the evolution of the local order parameter ⌬i on different lattice sites 关as color coded inside the blocks
in Fig. 1, panels 共b兲 and 共c兲兴 for the checkerboard pattern at f = 0.25 and n = 0.5 as a function of 兩Ū兩. The curve symbolized with stars refers
to the average order parameter ⌬ ¯
inhom. Panels 共b兲 and 共c兲: the same results as in panel 共a兲 for the striped and random patterns, respectively.
One notices that curves for ⌬2 and ⌬3 in panel 共b兲 intersect around 兩Ū兩 ⬇ 3t. Panels 共d兲, 共e兲, and 共f兲 present the same results at f = 0.5 and
n = 0.5 for the checkerboard, striped, and random patterns, respectively. Also panels 共g兲, 共h兲, and 共i兲 at f = 0.75 and n = 0.25.

As mentioned earlier in this section, for n ⬍ n*, ⌬ ¯ ¯

⌬ ¯ for 共n ⬍ n*兲.
inhom inhom共r = 1兲 ⬎ ⌬inhom共r ⬎ 1兲 共7兲
increases as a function of 兩Ū兩 and saturates for large enough
兩Ū兩 values while ⌬uniform is a consistently increasing function Therefore, knowing that r = 1 yields the largest magnitude of
¯
⌬
of 兩Ū兩 and is symmetric with respect to n = 1. Therefore, for inhom that is still enhanced compared to ⌬uniform when n
¯ ⬍ n*, the optimum effect due to inhomogeneity corresponds
n ⬍ n* as illustrated in Fig. 1 for r = ⌬ inhom / ⌬uniform at a given
to a value of n along the r = 1 contour in Fig. 1, for which
n,
¯
⌬ inhom is maximized. Figure 6 demonstrates the variation of
¯
⌬ inhom共r = 1兲 as a function of n for different f values. In panel
Ū2 ⬎ Ū1 ⇒ r共Ū2兲 ⬍ r共Ū1兲. 共5兲
共a兲 corresponding to f = 0.25, all three different inhomoge-
neous patterns yield the maximum ⌬ ¯
inhom共r = 1兲 within the
Now since
range of n = 0.5– 0.75. In panel 共b兲 for f = 0.5, the maxima are
closer to n = 0.5 while in panel 共c兲 for f = 0.75 they are around
¯
⌬ ¯ ¯
inhom共Ū2兲 ⬎ ⌬inhom共Ū1兲 ⇒ ⌬inhom„r共Ū2兲… n = 0.25. These results indicate that apparently the optimum
value for the doping of electrons in these inhomogeneous
¯
⬎⌬ for 共n ⬍ n*兲,
inhom„r共Ū1兲… 共6兲 systems is nopt ⬃ 1 − f = n* / 2. In strong coupling this density
corresponds to leaving the free sites empty and singly occu-
and as a result pying the attractive ones. By comparing this behavior with

184521-8
s-WAVE SUPERCONDUCTIVITY PHASE DIAGRAM IN… PHYSICAL REVIEW B 76, 184521 共2007兲

0.35 0.6
(a) 0.4 (a)
checkerboard 0.3 U = 6t
0.3 stripes 0.5
0.2

∆
random
0.25
0.4 0.1
0.2 f = 0.25 0
∆(r=1)

0.3 0 0.5 1 1.5

∆
T
0.15 uniform
0.2 checkerboard
0.1 stripes
random
0.1
0.05 (n = 1.35, f = 0.25, T = 0)
0
0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
n U
0.35 0.6
(b) 0.3
(b)
checkerboard
0.3 stripes 0.5 0.2 U = 2t

∆
random U = 4t
0.25 0.1
0.4
0.2 f = 0.5 0
∆(r=1)

0.3 0 0.1 0.2 0.3 0.4 0.5 0.6

∆
T
0.15 uniform
0.2 checkerboard
0.1 stripes (n = 1.5, f = 0.25, T = 0)
random
0.1
0.05
0
0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
n U
0.35
(c)
checkerboard FIG. 7. 共Color online兲 Panel 共a兲: 共main兲 the variation of the
0.3 stripes
averaged order parameter ⌬ ¯ for the uniform and three different
random
0.25 inhomogeneous patterns of checkerboard, stripes, and random at f
0.2
f = 0.75 = 0.25, n = 1.35⬍ n*, and T = 0 as a function of 兩Ū兩. The inset pre-
∆(r=1)

sents the evolution of these order parameters as a function of tem-

0.15 perature for 兩Ū兩 = 6t 共the largest at T = 0兲. Panel 共b兲: 共main兲 the same
results as in panel 共a兲 for n = 1.5= n* and T = 0. The inset shows the
0.1
evolution of order parameters against temperature for two different
0.05 values of 兩Ū兩 = 2t 共solid line with solid symbols兲 and 4t 共dashed line
with open symbols兲 as indicated in the T = 0 results by the dashed
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 lines.
n

FIG. 6. 共Color online兲 The magnitude of the averaged order

acting sites gradually vanishes above n = n* and for large
¯ for the inhomogeneous patterns along the enhancement
parameter ⌬ enough 兩Ū兩 values at n = n* we have an insulating phase as a
boundary 共r = 1兲 in Fig. 1 as a function of the average electron result of a charge-ordered phase formation. For larger f val-
ues, the enhancement ratio r = ⌬¯
occupation n for f = 0.25 关panel 共a兲兴, f = 0.5 关panel 共b兲兴, and f inhom / ⌬uniform will be larger
= 0.75 关panel 共c兲兴. for small values of 兩Ū兩 and n. However, the enhancement of
¯
⌬
the uniform system for which, due to particle-hole symmetry, inhom occurs at smaller values of n. We also find an opti-

nopt = 1 we can understand why nopt ⬃ n* / 2. Also, one ob- mum value of nopt ⬃ 1 − f = n* / 2 for the largest enhanced
¯
⌬
serves in Fig. 6 that by further diluting the interacting sites in inhom for a system in the presence of inhomogeneity.

a lattice, the magnitude of ⌬¯

inhom at nopt decreases.
We conclude in this section that by further diluting the IV. RESULTS AT FINITE T
density of interacting sites in a lattice while maintaining the
average pairing energy per site constant at T = 0, the average The SC transition temperature Tc of a lattice with an in-
order parameter ⌬ ¯ homogeneous pattern of interacting sites can also be larger
inhom may enhance. This enhancement re-
sults from the proximity effect in the noninteracting sites due compared to a uniform interaction distribution on the same
to their interacting neighbors leading to a larger average or- lattice. In this section, we investigate the conditions under
der parameter compared to the uniform lattice and in many which inhomogeneity in any form can lead to the enhance-
respects is independent of the particular inhomogeneous pat- ment of Tc as a function of 兩Ū兩 and n at different concentra-
tern. Superconductivity in an inhomogeneous lattice of inter- tion values f. In Fig. 7, panel 共a兲 presents the variation of the

184521-9
ARYANPOUR et al. PHYSICAL REVIEW B 76, 184521 共2007兲

average order parameter ⌬ ¯

inhom and ⌬uniform as functions of
0.6
0.4 (a)
兩Ū兩 for f = 0.25, n = 1.35⬍ 1.5= n*, and T = 0. We pick the 0.5 0.3 U = 4t
value of 兩Ū兩 = 6t, the largest in panel 共a兲, at T = 0 and plot both 0.2

∆
¯
⌬ inhom共T兲 and ⌬uniform共T兲 as functions of T in the inset inside
0.4 0.1
0
panel 共a兲. As seen in panel 共a兲, for 兩Ū兩 = 6t, ⌬uniform共T = 0兲 has 0.3
0 0.5 1 1.5 2

∆
uniform T
already exceeded all its inhomogeneous counterparts appre- checkerboard
ciably. Nevertheless, in the inset inside the same panel, Tc for 0.2 stripes
random
the inhomogeneous patterns are still larger than their uniform
0.1
pattern counterpart at 兩Ū兩 = 6t, indicating the strong enhance- (n = 0.9, f = 0.5, T = 0)
ment of Tc. 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
For the uniform pattern at all values of 兩Ū兩 and n, we find
U
Tc in good agreement with the BCS prediction, kbTc 0.6
⬇ 关⌬共0兲兩Ū兩兴 / 1.76, as expected from our mean-field ap- 0.3 U = 2.4t (b)

proach. In panel 共b兲 in Fig. 7 for n = 1.5= n*, however, for 0.5
0.2
U = 3.5t

∆
兩Ū兩 = 2t for which ⌬uniform共T = 0兲 is slightly larger than 0.4 0.1
¯
⌬ inhom共T = 0兲, we find that Tc for all inhomogeneous patterns 0
0 0.2 0.4 0.6 0.8 1
0.3

∆
关except the checkerboard whose ⌬inhom共T = 0兲 = 0 at 兩Ū兩 = 2t兴 uniform T
checkerboard
are also slightly larger than the uniform pattern Tc as shown 0.2 stripes (n = 1, f = 0.5, T = 0)
random
in the inset of the same panel. At 兩Ū兩 = 4t for which
¯ 0.1
⌬uniform共T = 0兲 becomes noticeably larger than ⌬ inhom共T = 0兲,
as presented in the inset inside panel 共b兲, Tc for the uniform 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
pattern also becomes larger than its inhomogeneous pattern
counterparts. In other words, the enhancement of Tc is rather U
0.6
weak when n 艌 n* compared to n ⬍ n* and ceases to persist 0.15 (c)
U = 1.5t
as 兩Ū兩 increases. In panel 共b兲, the checkerboard pattern has a 0.5 ∆
0.1
U = 1.8t
vanishing average order parameter at both T = 0 and finite T. 0.05
¯ 0.4
For stripes, ⌬ inhom starts with a finite value at T = 0 and
0
0 0.05 0.1 0.15 0.2
gradually vanishes as T increases. For the random pattern, T
0.3 uniform
∆

¯
⌬ checkerboard
inhom starts at a value very close to zero at T = 0. However, stripes
0.2
at 兩Ū兩 = 4t as T increases, there is a slight rise in the magni- random

tude of ⌬ ¯
inhom over some finite-temperature window before it 0.1 (n = 1.2, f = 0.5, T = 0)
completely vanishes at high enough T.
In Fig. 8, panel 共a兲 presents similar results for f = 0.5 at 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
n = 0.9⬍ n* and 兩Ū兩 = 4t. As illustrated in the inset of the same U
figure, even at 兩Ū兩 = 4t for which ⌬ ¯
inhom共T = 0兲 ⬍ ⌬uniform共T
= 0兲, all inhomogeneous patterns lead to larger Tc compared FIG. 8. 共Color online兲 Panels 共a兲, 共b兲, and 共c兲 refer to f = 0.5 and
n = 0.9⬍ n*, n = 1 = n*, and n = 1.2⬎ n*, respectively. Two different
to the uniform pattern. In panel 共b兲, for n = 1 = n*, at 兩Ū兩
¯ values of 兩Ū兩 have been chosen in panels 共b兲 and 共c兲 at finite tem-
= 2.4t, ⌬ inhom共T = 0兲 ⬍ ⌬uniform共T = 0兲. However, except for the perature as shown by the dashed lines in the T = 0 results. In panel
checkerboard, we still find an increase in Tc due to inhomo-
共b兲, for better visibility, results for stripes at 兩Ū兩 = 3.5t 共dashed line
geneity. Similar to Fig. 7 panel 共b兲, for both striped and
with open diamonds兲 are shown in magenta inside the inset.
random patterns, we also find a gradual increase in ⌬ ¯
inhom共T兲
as T increases before it totally vanishes at high enough T. largest value of 兩Ū兩 = 3t, inhomogeneity significantly in-
The enhancement of Tc continues to persist up to 兩Ū兩 = 3.5t, creases Tc as illustrated in the inset of the same panel. This
¯ occurs despite the fact that ⌬ ¯
although ⌬ inhom共T = 0兲 ⬇ 0 for all inhomogeneous patterns, inhom共T = 0兲 ⬍ ⌬uniform共T = 0兲. For
¯ *
n = 0.5= n , even at 兩Ū兩 = 1.86t where ⌬ ¯
due to the gradual increase of ⌬ inhom共T兲 with temperature. inhom共T = 0兲 ⬇ 0, for
Panel 共c兲 corresponds to n = 1.2⬎ n* and it can be observed both the striped and random patterns we again find an in-
that for 兩Ū兩 = 1.5t, for stripes only, Tc is increased. However, crease in Tc due to inhomogeneity. In this case, the gradual
increase in ⌬¯
at slightly larger 兩Ū兩 = 1.8t, Tc for the uniform pattern signifi- inhom共T兲 as a function of T is further pronounced
cantly dominates the inhomogeneous pattern ones and no until Tc totally vanishes for these two patterns at high enough
gradual increase in ⌬ ¯ T. Panel 共c兲 corresponds to n = 0.65⬎ n* and similar to f
inhom共T兲 occurs as T increases.
Figure 9共a兲 has f = 0.75 and n = 0.4⬍ n*. Similar to the = 0.25 and f = 0.5 cases when n ⬎ n*, for large enough 兩Ū兩
behavior observed for f = 0.25 and 0.5 when n ⬍ n*, up to the = 1.5t 共slightly larger than 兩Ū兩 = 1.25t in the inset兲 inhomoge-

184521-10
s-WAVE SUPERCONDUCTIVITY PHASE DIAGRAM IN… PHYSICAL REVIEW B 76, 184521 共2007兲

0.6 1.2
0.15 U = 2.5t (a) stripes (n = 0.5, f = 0.75, U = 1.86t) (a)
0.5 0.1 U = 3t 1

∆
T=0
0.05 T = 0.6t
0.4 0.8
T = 1.1t
0 T = 1.5t
0 0.5 1 1.5 2 2.5 3

DOS
0.3 0.6 T = 1.65t
∆

uniform T
checkerboard (n = 0.4, f = 0.75, T = 0)
0.2 stripes 0.4
random
0.1 0.2

0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1
U ω
0.6 0.6
0.08 (b) ∆1 ( U = 1.84t) U4 = 0 (b)
U = 1.84t ∆1 ( U = 1.86t) U3 = 0
0.5 0.06 0.5
U = 1.86t ∆2 ( U = 1.84t) U2 = 0
∆

0.04 ∆2 ( U = 1.86t) U1 = 4U
0.4 0.4
0.02 ∆3 ( U = 1.84t)
∆3 ( U = 1.86t)
0.3 0
∆

0.3

∆
0 0.5 1 1.5
uniform T
0.2 checkerboard 0.2
stripes
0.1 random
0.1 (stripes, n = 0.5, f = 0.75)
(n = 0.5, f = 0.75, T = 0)
0
0 0.5 1 1.5 2 2.5 3 3.5 4 0
0 0.25 0.5 0.75 1 1.25 1.5 1.75 2
U T
0.6 2
0.06 (c) (c)
1.75 (stripes, n = 0.5, f = 0.75)
0.5
0.04
1.5
∆

U = 1.25t
0.4 0.02 U = 1.5t n1 ( U = 1.84t)
1.25 n1 ( U = 1.86t)
0 n2 ( U = 1.84t)
0.3 0 0.25 0.5 0.75 1
∆

U4 = 0
n

uniform T n2 ( U = 1.86t)
checkerboard n3 ( U = 1.84t) U3 = 0
0.2 stripes 0.75 U2 = 0
n3 ( U = 1.86t)
random 0.5 U1 = 4U
0.1
(n = 0.65, f = 0.75, T = 0) 0.25
0
0 0.5 1 1.5 2 2.5 3 3.5 4 0
0 0.25 0.5 0.75 1 1.25 1.5 1.75 2
U T

FIG. 9. 共Color online兲 Panels 共a兲, 共b兲, and 共c兲 refer to f = 0.75 FIG. 10. 共Color online兲 Anomalous behavior of the averaged
and n = 0.4⬍ n*, n = 0.5= n*, and n = 0.65⬎ n*, respectively. Two order parameter ⌬ ¯ as a function of temperature T at f = 0.75 and n
different values of 兩Ū兩 have been chosen in all panels as shown by = 0.5 as presented in Fig. 9, panel 共b兲 共inset兲, for stripes. Panel 共a兲
the dashed lines in the T = 0 results. In panel 共c兲, results for check- illustrates the evolution of the DOS as a function of temperature for
erboard at 兩Ū兩 = 1.5t 共dashed line with open diamonds兲 are shown in 兩Ū兩 = 1.86t. Panel 共b兲 demonstrates how the local ⌬i on any of the
cyan inside the inset for better visibility. individual four sites inside the block shown in Fig. 1, panel 共b兲 vary
as a function of temperature T. Panel 共c兲: the evolution of the local
neity no longer yields larger Tc compared to the uniform. occupation number ni on any of the individual four sites inside the
共See inset.兲 block shown in Fig. 1, panel 共b兲, as a function of temperature T.
The anomalous increase of ⌬ ¯
inhom as a function of T at
*
n = n for f 艌 0.5 is an actual feature and is believed to be conductivity consistent with the behavior shown inside the
related to the gradual destruction of the charge-ordered phase inset in Fig. 9, panel 共b兲. By further increasing T, the gap
due to temperature, leading to an intermediate SC phase. In begins to shrink due to the destruction of superconductivity
Fig. 10 corresponding to Fig. 9, panel 共b兲 for n = 0.5, f by temperature until it entirely vanishes at T = 1.65t, in agree-
= 0.75, and 兩Ū兩 = 1.86t, panel 共a兲 presents the DOS results at ment with the results inside the inset in Fig. 9, panel 共b兲.
several T values within the temperature window of the inset Panel 共b兲 in Fig. 10 illustrates the variation of the local
in Fig. 9, panel 共b兲. At T = 0, the gap in the DOS is barely order parameter on all four different types of sites for stripes
nonzero at ␻ = 0 共Fermi energy兲, indicating a charge-ordered at f = 0.75. Only ⌬1 along the line of interacting sites varies
phase. By increasing T to 0.6t the gap widens towards super- appreciably as T increases. Panel 共c兲 of Fig. 10 shows the

184521-11
ARYANPOUR et al. PHYSICAL REVIEW B 76, 184521 共2007兲

0.6 stripes, temperature acts against the more localized charge-

random (n = 0.5, f = 0.75, U = 1.86t) (a) ordered phase, allowing for the Cooper pairs to move and
0.5 T=0 overlap more freely and consequently the enhancement of Tc.
T = 0.6t
T=t
Panel 共c兲 also confirms that the charge-ordered phase is in-
0.4
T = 1.5t deed destroyed by temperature, allowing the noninteracting
T = 1.6t
DOS

0.3
sites to become more occupied, resulting in an intermediate
SC phase. Metallic behavior is established at sufficiently
0.2 high temperature.
It is also noteworthy that this anomalous increase in
0.1 ¯
⌬ inhom was not observed for the checkerboard pattern for
which there is no superconductivity at any temperature when
0
-3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 n = n*. One possible explanation is that due to the particular
ω geometry for the checkerboard, localized Cooper pairs are
0.2 farther apart from one another compared to the striped and
∆1 ( U = 1.84t) (b)
∆1 ( U = 1.86t) random patterns. As mentioned earlier, when n = n* we have
∆2 ( U = 1.84t) two electrons per interacting site. For the checkerboard lat-
0.15 U1 = 4U U2 = 0
∆2 ( U = 1.86t) tice, this leads to a pair localized in the interacting sites with
(random, n = 0.5, f = 0.75) empty nearest-neighboring sites. Therefore, the effect of tem-
perature is minor in causing further overlap among the pairs
∆

0.1
before it totally destroys them 关especially at f = 0.5 and 0.75
as shown in Fig. 1, panel 共a兲兴.
0.05 A similar increase in Tc upon introducing a checkerboard
pattern is found in the MCMF calculations as well, arising
from the loss of long-range phase coherence. This is particu-
0 larly significant because the MCMF approach incorporates
0 0.25 0.5 0.75 1 1.25 1.5 1.75 2
T the subtle nature of the SC transition in 2D discussed earlier.
2
(c) We have also independently confirmed that our conclusions
1.75 (random, n = 0.5, f = 0.75) and arguments equally apply for a model with nearest-
1.5
neighbor attraction, leading to a d-wave SC close to half-
n1 ( U = 1.84t) filling, which reflects the cuprates’ phenomenology more
1.25 n1 ( U = 1.86t) truthfully.49,61
n2 ( U = 1.84t) We have found that upon introducing inhomogeneity into
1
n

n2 ( U = 1.86t)
the pattern of interacting sites on a lattice. Tc can be in-
0.75
U1 = 4U U2 = 0 creased over a wide range of 兩Ū兩 as long as n 艋 n* even if
0.5 ¯ *
⌬ inhom共T = 0兲 ⬍ ⌬uniform共T = 0兲. When n ⬎ n , however, for suf-
0.25 ficiently large 兩Ū兩, inhomogeneity acts against superconduc-
0 tivity and therefore Tc becomes smaller compared to the uni-
0 0.25 0.5 0.75 1 1.25 1.5 1.75 2
T form pattern of interacting sites. The case of n = n* is
anomalous. The charge-ordered phase established at large
FIG. 11. 共Color online兲 Same as Fig. 10, but for the random enough 兩Ū兩 values at T = 0 gradually vanishes as T increases.
pattern. ⌬inhom first increases with increasing T, leading to a SC phase
共at least for large enough f values and lower-symmetrical
local occupation numbers on all the four different types of inhomogeneity patterns兲, and then vanishes, indicating a me-
sites for stripes at f = 0.75 关Fig. 1, panel 共b兲兴. By increasing tallic state.
T, while the occupation of the interacting sites gradually
drops from n = 2 to n ⬇ 1, an indication of the destruction of
V. SUMMARY AND DISCUSSION
the charge-ordered phase, the occupation of the noninteract-
ing sites grows, leading to additional mobility and overlap of In summary, we have shown that for the attractive Hub-
the Cooper pairs and therefore enhancing Tc. bard model on a square lattice, there is a significant range of
Figure 11, panel 共a兲 shows the DOS results for the random electron doping and interaction strength over which the av-
pattern for n = 0.5, f = 0.75, and 兩Ū兩 = 1.86t depicted in Fig. 9, erage superconducting order parameter is larger for a lattice
panel 共b兲. Similar to stripes, over the window of temperature with inhomogeneous patterns of interacting sites than a uni-
studied in panel 共a兲, the gap in the DOS initially grows up to form distribution of these interacting sites at a constant in-
T ⬇ t and then gradually drops to zero slightly beyond T teraction strength per site. We have presented the phase dia-
= 1.6t to indicate the formation of a metallic phase as the grams for three different inhomogeneous patterns,
temperature destroys superconductivity. Panel 共b兲 in Fig. 11 checkerboard, stripes, and random, and also three different
also shows significant variation of the local order parameter values for the concentration of the noninteracting sites. Apart
only on the interacting sites on average. Hence, similar to from a few particular features, the overall physics illustrated

184521-12
s-WAVE SUPERCONDUCTIVITY PHASE DIAGRAM IN… PHYSICAL REVIEW B 76, 184521 共2007兲

in the phase diagrams is pattern independent. As we vary the plications to. However, for an inhomogeneous system being
mean interaction strength 兩Ū兩 and the doping level n at zero a mixture of different phases, defining an average SC gap is
temperature, we have verified the existence of at least three not trivial. Thus, the thermodynamic properties of inhomo-
different phases—namely, 共i兲 superconducting, 共ii兲 insulating geneous superconductors will not necessary exhibit the same
due to the charge order phases, and 共iii兲 metallic. Our find- behavior as their homogeneous counterparts. It has been
ings and claims are strongly supported by studying the be- shown that the rise of the specific heat in inhomogeneous
havior of a variety of quantities computed in this work all superconductors obeys a power law behavior as opposed to
consistently corroborating one another. exponential in homogeneous ones using the attractive Hub-
The enhancement of the average order parameter for the bard model with random interacting sites.62 Also, the super-
inhomogeneous interacting site patterns is due to the prox- fluid density and stiffness in general decrease due to the pres-
imity effect—i.e., the tunneling effect of the Cooper pairs ence of disorder.63 Nevertheless, lower superfluid density
from the interacting sites, leading to finite order parameter does not necessarily lead to lower Tc as according to the
values on neighboring sites. This conclusion is supported by Anderson theorem, a nonmagnetic impurity should not affect
the effect occurring at weak coupling, where the coherence the Tc and therefore thermodynamic properties of an s-wave
length is large, rather than in the strong-coupling regime of superconductor. We have presented clear evidence for the
preformed pairs. Agreement between the BdG results and the enhancement of Tc which does have a direct experimental
MCMF calculations justifies the application and conclusions implication. Thermodynamic properties of inhomogeneous
of the BdG approach within the small-兩Ū兩 regime. Our cal- superconductors are very rich in physics and a great deal of
culations also clearly confirm that an inhomogeneous inter- contributions and new ideas in this area are yet to appear.
action potential can lead to the increase in the phase transi- While the attractive Hubbard Hamiltonian obviously does
tion temperature Tc over a wide range of n and 兩Ū兩 for not incorporate many of the features of high-Tc supercon-
various f values. Counterintuitively, as long as n is less than ductors 共notably the symmetry of the pairing兲, the model has
or equal to twice the fraction of interacting sites, this in- been shown to provide useful insight into some of their
phenomenology—for example, the spin gap.64 It is therefore
crease in Tc continues even for values of 兩Ū兩 for which the
tempting to speculate that our results concerning inhomoge-
order parameter is larger for the uniform pattern than for
neity may have similar connections. Specifically, earlier
inhomogeneous patterns at T = 0.
angle-resolved photoemission spectroscopy 共ARPES兲 data65
One possible explanation takes into account that in this
suggest that the underdoped phase of LSCO 共La2−xSrxCuO4兲
weak-coupling parameter regime, Tc is a supralinearly in-
consists of SC clusters, embedded in the antiferromagnetic
creasing function of U. In such a case, it may be that in the
host. In such a system, inhomogeneous gaps appear naturally
inhomogeneous system the sites with larger U produce a
and our results here indicate that the SC transition is in fact
nonlinear enhancement relative to Tc of the uniform system
determined by the largest gap values rather than the much
and, through the proximity effect, drag the noninteracting
smaller gaps found at phase boundaries, as one might naively
sites along with them. This trend changes when n exceeds
think. This renders the SC phase more stable than it would
twice the number of interacting sites 共i.e., some electrons
otherwise be and also simplifies the description of these sys-
must occupy noninteracting sites兲, for which at large enough
tems.
兩Ū兩 values inhomogeneity fails to increase Tc over that of the It is worth emphasizing that in most situations, inhomo-
uniform pattern. The n = 2共1 − f兲 = n* case for sufficiently geneities reduce values of order parameters51 and critical
large f values and less symmetric inhomogeneous patterns temperatures, even when comparisons are made, as they are
共such as stripes and random as opposed to the checkerboard兲 in this article, to homogeneous systems with the same aver-
is anomalous as it shows the enhancement of ⌬ ¯
inhom共T兲 as age value of all parameters. This is true, for example, for
temperature increases. classical site-diluted Ising models, where the ferromagnetic J
It is even more surprising to find that a system which is is increased to compensate for absent sites, and quantum
nonsuperconducting 共charge-ordered insulator兲 at T = 0 can models like the boson Hubbard model where random chemi-
become superconducting upon increasing T for a finite win- cal potentials monotonically decrease and ultimately destroy
dow of temperature before turning metallic. This anomalous superfluidity.66,67 An exception is the increase of TNéel by
behavior was shown to be related to a crossover from a randomness reported in density mean-field theory studies of
charge-ordered insulating phase for n = n* at large enough 兩Ū兩 the repulsive model68 and recently, the SC Tc in the XY
values to an intermediate SC phase upon increasing T before model Hamiltonian with certain types of inhomogeneous pat-
entering the metallic phase at sufficiently large T. terns for the coupling constant between spins sitting on
We wish to emphasize that in this article we have focused nearest-neighboring sites.50
on the enhancement 共or not兲 of the pairing amplitude 共our ⌬
defined in Sec. II兲, rather than its product with the local
interaction strength-兩Ui兩 which is more directly related to the ACKNOWLEDGMENTS
gap but contains less information, and gives less insight, be-
cause it automatically vanishes on any site without interac- We acknowledge useful conversations with Jian-Xin Zhu
tion. Thermodynamic measurements would probe quantities and Wei Ku. This research was supported by National Sci-
which include the energy scale, such as the specific heat or ence Foundation Grants Nos. DMR-0421810 and DMR-
superfluid rigidity which our results may not have direct im- 0426826, U.S. ONR, CNPq-Brazil, and FAPERJ-Brazil.

184521-13
ARYANPOUR et al. PHYSICAL REVIEW B 76, 184521 共2007兲

1 Inhomogeneous Superconductors—1979, AIP Conf. Proc. No. 58, 100505共R兲 共2007兲.

edited by D. U. Gubser, T. L. Francavilla, J. R. Leibowitz, and 32
A. Garg, H. R. Krishnamurthy, and M. Randeria, Phys. Rev. Lett.
S. A. Wolf 共AIP, New York, 1979兲. 97, 046403 共2006兲.
2 Percolation, Localization, and Superconductivity, edited by A. M. 33 D. Zhang, arXiv:cond-mat/0607762 共unpublished兲.

Goldman and S. A. Wolf 共Plenum, New York, 1984兲. 34 J. A. Robertson, S. A. Kivelson, E. Fradkin, A. C. Fang, and A.
3
E. Dagotto, Rev. Mod. Phys. 66, 763 共1994兲. Kapitulnik, Phys. Rev. B 74, 134507 共2006兲.
4 P. A. Lee and T. V. Ramakrishnan, Rev. Mod. Phys. 57, 287 35 A. Ghosal, M. Randeria, and N. Trivedi, Phys. Rev. B 65, 014501

共1985兲. 共2001兲.
5
D. Belitz and T. R. Kirkpatrick, Rev. Mod. Phys. 66, 261 共1994兲. 36
V. Dobrosavljevic, T. R. Kirkpatrick, and G. Kotliar, Phys. Rev.
6 T. Cren, D. Roditchev, W. Sacks, J. Klein, J.-B. Moussy, C. Lett. 69, 1113 共1992兲.
Deville-Cavellin, and M. Laguës, Phys. Rev. Lett. 84, 147 37 V. Dobrosavljevik and G. Kotliar, Phys. Rev. B 50, 1430 共1994兲.

共2000兲. 38 V. Dobrosavljevic, A. A. Pastor, and B. K. Nikolic, Europhys.

7 S.-H. Pan, J. P. O’Neal, R. L. Badzey, C. Chamon, H. Ding, J. R. Lett. 62, 76 共2003兲.
Engelbrecht, Z. Wang, H. Eisaki, S. Uchida, A. K. Gupta, K.-W. 39 A. J. Coleman, E. P. Yukalova, and V. I. Yukalov, Physica C 243,

Ng, E. W. Hudson, K. M. Lang, and J. C. Davis, Nature 共Lon- 76 共1995兲.

don兲 413, 282 共2001兲. 40
V. I. Yukalov and E. P. Yukalova, Phys. Rev. B 70, 224516
8 C. Howald, P. Fournier, and A. Kapitulnik, Phys. Rev. B 64, 共2004兲.
100504共R兲 共2001兲. 41 D. Valdez-Balderas and D. Stroud, Phys. Rev. B 74, 174506
9 K. M. Lang, V. Madhavan, J. E. Hoffman, E. W. Hudson, H. 共2006兲.
Eisaki, S. Uchida, and J. C. Davis, Nature 共London兲 415, 412 42 K. Seo, H.-D. Chen, and J. Hu, Phys. Rev. B 76, 020511共R兲

共2002兲. 共2007兲.
10 43
K. McElroy, D.-H. Lee, J. E. Hoffman, K. M. Lang, J. Lee, E. W. B. M. Andersen, A. Melikyan, T. S. Nunner, and P. J. Hirschfeld,
Hudson, H. Eisaki, S. Uchida, and J. C. Davis, Phys. Rev. Lett. Phys. Rev. B 74, 060501共R兲 共2006兲.
94, 197005 共2005兲. 44 E. S. Caixeiro, E. V. L. de Mello, and A. Troper, arXiv:cond-mat/
11
T. Hanaguri, C. Lupien, Y. Kohsaka, D.-H. Lee, M. Azuma, M. 0508381 共unpublished兲.
Takano, H. Takagi, and J. C. Davis, Nature 共London兲 430, 1001 45 R. Jamei, J. Robertson, E.-A. Kim, A. Fang, A. Kapitulnik, and S.

共2004兲. A. Kivelson, Phys. Rev. B 74, 174521 共2006兲.

12 M. Vershinin, S. Misra, S. Ono, Y. Abe, Y. Ando, and A. Yazdani, 46 A. Melikyan and Z. Tešanović, Phys. Rev. B 71, 214511 共2005兲.

Science 303, 1995 共2004兲. 47

Z. Tešanović, Phys. Rev. Lett. 93, 217004 共2004兲.
13 H. A. Mook, P. Dai, and F. Dogan, Phys. Rev. Lett. 88, 097004 48 I. Martin, D. Podolsky, and S. A. Kivelson, Phys. Rev. B 72,

共2002兲. 060502共R兲 共2005兲.

14 49
J. M. Tranquada, J. D. Axe, N. Ichikawa, A. R. Moodenbaugh, Y. K. Aryanpour, E. R. Dagotto, M. Mayr, T. Paiva, W. E. Pickett,
Nakamura, and S. Uchida, Phys. Rev. Lett. 78, 338 共1997兲. and R. T. Scalettar, Phys. Rev. B 73, 104518 共2006兲.
15
Ch. Renner, G. Aeppli, B.-G. Kim, Y.-A. Soh, and S.-W. Cheong, 50
Y. L. Loh and E. W. Carlson, Phys. Rev. B 75, 132506 共2007兲.
Nature 共London兲 416, 518 共2002兲. 51 B. Chatterjee and A. Taraphder, arXiv:cond-mat/0702150 共unpub-
16 J. Burgy, A. Moreo, and E. Dagotto, Phys. Rev. Lett. 92, 097202
lished兲.
共2004兲. 52 Y. Matsushita, H. Bluhm, T. H. Geballe, and I. R. Fisher, Phys.
17
A. Moreo, S. Yunoki, and E. Dagotto, Science 283, 2034 共1999兲. Rev. Lett. 94, 157002 共2005兲.
18 E. Dagotto, T. Hotta, and A. Moreo, Phys. Rep. 344, 1 共2001兲. 53 Y. Matsushita, P. A. Wianecki, A. T. Sommer, T. H. Geballe, and
19 M. L. Foo, Y. Wang, S. Watauchi, H. W. Zandbergen, T. He, R. J.
I. R. Fisher, Phys. Rev. B 74, 134512 共2006兲.
Cava, and N. P. Ong, Phys. Rev. Lett. 92, 247001 共2004兲. 54 M. Dzero and J. Schmalian, Phys. Rev. Lett. 94, 157003 共2005兲.
20 K.-W. Lee, J. Kunes, P. Novak, and W. E. Pickett, Phys. Rev. Lett. 55 R. Micnas, J. Ranninger, and S. Robaskiewicz, Rev. Mod. Phys.

94, 026403 共2005兲. 62, 113 共1990兲 and references therein.

21 J. Zaanen and O. Gunnarsson, Phys. Rev. B 40, R7391 共1989兲. 56 S. Robaszkiewicz, R. Micnas, and K. A. Chao, Phys. Rev. B 23,
22
K. Machida, Physica C 158, 192 共1989兲. 1447 共1981兲.
23 M. Kato, K. Machida, H. Nakanishi, and M. Fujita, J. Phys. Soc. 57 H. Shiba, Prog. Theor. Phys. 48, 2171 共1972兲.

Jpn. 59, 1047 共1990兲. 58

V. J. Emery, Phys. Rev. B 14, 2989 共1972兲.
24 S. R. White and D. J. Scalapino, Phys. Rev. B 70, 220506共R兲 59 G. Litak and B. L. Györffy, Phys. Rev. B 62, 6629 共2000兲.

共2004兲. 60
D. Hurt, E. Odabashian, W. E. Pickett, R. T. Scalettar, F. Mon-
25 M. Vojta, Phys. Rev. B 66, 104505 共2002兲.
daini, T. Paiva, and R. R. dos Santos, Phys. Rev. B 72, 144513
26
G. Seibold, C. Castellani, C. Di Castro, and M. Grilli, Phys. Rev. 共2005兲.
B 58, 13506 共1998兲. 61
M. Mayr, G. Alvarez, C. Şen, and E. Dagotto, Phys. Rev. Lett.
27 S. A. Kivelson and E. Fradkin, arXiv:cond-mat/0507459 共unpub-
94, 217001 共2005兲.
lished兲. 62
G. Litak, J. Magn. Magn. Mater. 272-276, 966 共2004兲.
28 H.-X. Huang, Y. Q. Li, J.-Y. Gan, Y. Chen, and F.-C. Zhang, Phys. 63 A. Paramekanti, N. Trivedi, and Mohit Randeria, Phys. Rev. B

Rev. B 75, 184523 共2007兲. 57, 11639 共1998兲.

29 M. M. Maśka, Ż. Śledź, K. Czajka, and M. Mierzejewski, Phys. 64 M. Randeria, in Bose Einstein Condensation, edited by A. Griffin,

Rev. Lett. 99, 147006 共2007兲. D. Snoke, and S. Stringari 共Cambridge University Press, Cam-
30 W.-F. Tsai and S. A. Kivelson, Phys. Rev. B 73, 214510 共2006兲.
bridge, England, 1994兲, and references cited therein.
31 65
G. Seibold, J. Lorenzana, and M. Grilli, Phys. Rev. B 75, T. Yoshida, X. J. Zhou, T. Sasagawa, W. L. Yang, P. V. Bogdanov,

184521-14
s-WAVE SUPERCONDUCTIVITY PHASE DIAGRAM IN… PHYSICAL REVIEW B 76, 184521 共2007兲

A. Lanzara, Z. Hussain, T. Mizokawa, A. Fujimori, H. Eisaki, 67 R. T. Scalettar, G. G. Batrouni, and G. T. Zimanyi, Phys. Rev.
Z.-X. Shen, T. Kakeshita, and S. Uchida, Phys. Rev. Lett. 91, Lett. 66, 3144 共1991兲.
027001 共2003兲. 68
M. Ulmke, V. Janis, and D. Vollhardt, Phys. Rev. B 51, 10411
66
M. P. A. Fisher, P. B. Weichman, G. Grinstein, and D. S. Fisher, 共1995兲.
Phys. Rev. B 40, 546 共1989兲.

184521-15