Frustrated further-neighbor antiferromagnetic and electron-hopping interactions in the d = 3 t − J
model: Finite-temperature global phase diagrams from renormalization group theory
C. Nadir Kaplan,1,2,3A. Nihat Berker,4,5,6and Michael Hinczewski6,7
1Department of Physics, Istanbul Technical University, Maslak, 34469 Istanbul, Turkey 2Department of Physics, Koç University, Sarıyer, 34450 Istanbul, Turkey
3Martin Fisher School of Physics, Brandeis University, Waltham, Massachusetts 02454, USA 4Faculty of Engineering and Natural Sciences, Sabancı University, Orhanlı, Tuzla, 34956 Istanbul, Turkey
5Department of Physics, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA 6Feza Gürsey Research Institute, TÜBITAK–Bosphorus University, Çengelköy, 34684 Istanbul, Turkey
7Department of Physics, Technical University of Munich, 85748 Garching, Germany
共Received 30 September 2008; revised manuscript received 21 May 2009; published 23 December 2009兲 The renormalization-group theory of the d = 3 t − J model is extended to further-neighbor antiferromagnetic or electron-hopping interactions, including the ranges of frustration. The global phase diagram of each model is calculated for the entire ranges of temperatures, electron densities, further/first-neighbor interaction-strength ratios. With the inclusion of further-neighbor interactions, an extremely rich phase diagram structure is found and is explained by competing and frustrated interactions. In addition to thetJphase seen in earlier studies of the nearest-neighbor d = 3 t − J model, theHbphase seen before in the d = 3 Hubbard model appears both near and away from half filling.
DOI:10.1103/PhysRevB.80.214529 PACS number共s兲: 74.25.Dw, 71.10.Fd, 05.30.Fk, 64.60.De
I. INTRODUCTION
The simplest electron conduction model, including nearest-neighbor hopping on a lattice and on-site Coulomb repulsion, is the Hubbard model.1In the limit of very strong
on-site Coulomb repulsion, second-order perturbation theory on the Hubbard model yields the t − J model,2,3in which sites
doubly occupied by electrons do not exist. Studies of the Hubbard model4 and of the t − J model,5 including spatial anisotropy6 and quenched nonmagnetic impurities7 in good
agreement with experiments, have shown the effectiveness of renormalization-group theory, especially in calculating phase diagrams at finite temperatures for the entire range of elec-tron densities in d = 3. These calculations have revealed phases, dubbed thephases, which occur only in these elec-tronic conduction models under doping conditions. The tell-tale characteristics of thephases are, in contrast to all other phases of the systems, a nonzero electron-hopping probabil-ity at the largest length scales共at the renormalization-group thermodynamic-sink fixed points兲 and the divergence of the electron-hopping constant t under repeated rescalings. Furthermore, the phase diagram topologies, the doping ranges, and the contrasting quantitative and antiferromag-netic behaviors under quenched impurities7 have been in
agreement with experimental findings.8,9 A benchmark for
this renormalization-group approach has also been estab-lished by a detailed and successful comparison, with the ex-act numerical results of the quantum transfer matrix method,10,11 of the specific heat, charge susceptibility, and
magnetic susceptibility of the d = 1 Hubbard model calcu-lated with our method.12 Furthermore, results with this
method have indicated that no finite-temperature phase tran-sition occurs in the t − J model in d = 1. A phase separation at zero temperature has been found in d = 1 in Ref.13. Thus, the
d = 1 t − J model appears to have a first-order phase transition
at zero temperature that disappears as soon as temperature is
raised from zero, as in other d = 1 models such as the Ising and Blume-Capel models.14,15A phase separation16–18occurs
in d = 2 for low values of t/J, but not for t/J⬎0.24.5 In d
= 3, a narrow phase separation occurs, as seen in the density-temperature phase diagrams below. Two distinct phases have been found in the Hubbard model,4HbandtJ,
respec-tively occurring at weak and strong coupling. The calculated low-temperature behavior and critical exponent of the spe-cific heat4have pointed to BCS-like and BEC-like behaviors, respectively. Only thetJphase was found in the t − J model. The current work addresses the issue of whether both phases can be found in the t − J model, via the inclusion of further-neighbor antiferromagnetic 共J2兲 or further-neighbor electron hopping共t2兲 interactions. We find that, depending on the temperature and doping level, the further-neighbor inter-actions may compete with the further-neighbor effects of the nearest-neighbor interactions, namely, that frustration occurs as a function of temperature and doping level. This compe-tition共or reinforcement兲 between the interactions of succes-sive length scales underpins the calculated evolution of the phase diagrams. Global phase diagrams are obtained for the entire ranges of each type of further-neighbor interaction. With the inclusion of further-neighbor interactions, an tremely rich phase diagram structure is found and is ex-plained by competing and frustrated interactions. Both Hb and tJphases are indeed found to occur in the t − J model with the inclusion of these further-neighbor interactions. Fur-thermore, distinctive lamellar phase diagram structures of antiferromagnetism interestingly surround thephases in the doped regions.
II. t − J HAMILTONIAN
On a d-dimensional hypercubic lattice, the t − J model is defined by the Hamiltonian
−H = P
冋
− t兺
具ij典,共ci †c j+ c†jci兲 − J兺
具ij典 Si· Sj+ V兺
具ij典 ninj +˜兺
i n册
P, 共1兲where = 1/kBT and, with no loss of generality,5 tⱖ0 is used. Here ci† and ci are the creation and annihilation op-erators for an electron with spin =↑ or ↓ at lattice site i, obeying anticommutation rules, ni= ni↑+ ni↓ are the number operators where ni= ci
†
ci, and Si=兺⬘ci
†
s⬘ci⬘ is the single-site spin operator, with s being the vector of Pauli spin matrices. The projection operator P =兿i共1−ni↓ni↑兲 projects out all states with doubly-occupied sites. The interaction constants t, J, V, and ˜ correspond to electron hopping,
nearest-neighbor antiferromagnetic coupling共J⬎0兲, nearest-neighbor electron-electron interaction, and chemical poten-tial, respectively. From rewriting the t − J Hamiltonian as a sum of pair Hamiltonians −H共i, j兲, Eq. 共1兲 becomes
−H =
兺
具ij典 P冋
− t兺
共ci † cj+ cj † ci兲 − JSi· Sj+ Vninj +共ni+ nj兲册
P⬅兺
具ij典兵− H共i, j兲其, 共2兲where=˜/2d. The standard t−J Hamiltonian is a special
case of Eq. 共2兲 with V/J=1/4, which stems from
second-order perturbation theory on the Hubbard model.2,3
III. RENORMALIZATION-GROUP TRANSFORMATION A. d = 1 recursion relations
In d = 1, the Hamiltonian of Eq.共2兲 is
−H =
兺
i
兵−H共i,i + 1兲其. 共3兲
A decimation eliminates every other one of the successive degrees of freedom arrayed in a linear chain, with the parti-tion funcparti-tion being conserved, leading to a length rescaling factor b = 2. By neglecting the noncommutativity of the op-erators beyond three consecutive lattice sites, a trace over all states of even-numbered sites can be performed,19,20
Trevene−H= Trevenexp
冉
兺
i
兵−H共i,i + 1兲其
冊
= Trevenexp
冉
兺
i
even
兵−H共i − 1,i兲 −H共i,i + 1兲其
冊
⯝
兿
i even Trie兵−H共i−1,i兲−H共i,i+1兲其=兿
i even e−⬘H⬘共i−1,i+1兲 ⯝ exp共兺
i even 兵−⬘
H⬘
共i − 1,i + 1兲其兲 = e−⬘H⬘, 共4兲 where −⬘
H⬘
is the renormalized Hamiltonian. This ap-proach, where the two approximate steps labeled with⯝ are in opposite directions, has been successful in the detailed solutions of quantum spin19–25and electronic4–7systems. Theanticommutation rules are correctly accounted within the
three-site segments, at all successive length scales, in the iterations of the renormalization-group transformation.
The algebraic content of the decimation in Eq. 共4兲 is e−⬘H⬘共i,k兲= Trje−H共i,j兲−H共j,k兲, 共5兲 where i , j , k are three consecutive sites of the unrenormalized linear chain. The renormalized Hamiltonian is given by
−
⬘
H⬘
共i,k兲 = P冋
− t⬘
兺
共ci†c
k+ ck†ci兲 − J
⬘
Si· Sk+ V⬘
nink+
⬘
共ni+ nk兲 + G⬘
册
P, 共6兲where G
⬘
is the additive constant per bond, which is always generated in renormalization-group transformations, does not affect the flow of the other interaction constants, and is nec-essary in the calculation of expectation values. The values of the renormalized共primed兲 interaction constants appearing in −⬘
H⬘
are given by the recursion relations extracted from Eq. 共5兲, which will be given here in closed form, while theAppendix details the derivation of Eq.共7兲 from Eq. 共5兲: t
⬘
=1 2ln ␥4 ␥2 , J⬘
= ln␥6 ␥7 , V⬘
=1 4ln ␥1 4␥ 6␥7 3 ␥24␥44 , ⬘
=+1 2ln冉
␥2␥4 ␥1 2冊
, G⬘
= b d G + ln␥1, 共7兲 where ␥1= 1 + 2u3f冉
2冊
, ␥2= uf冉
− 2冊
+ 1 2u 2x2+ 3 2u 2vf冉
− J 8+ V 2 + 2冊
, ␥4= 1 + 3 2u 2v2+1 2u 2xf冉
3J 8 + V 2 + 2冊
, ␥6= 2v3x + xf冉
− 3J 8 − V 2 − 2冊
, ␥7= 2 3vx 3+4 3v 4+vf冉
J 8− V 2 − 2冊
, 共8兲 and v = exp共− J/8 + V/2 +/2兲, x = exp共3J/8 + V/2 +/2兲, u = exp共/2兲, f共A兲 = cosh冑
2t2+ A2+冑
A 2t2+ A2sinh冑
2t 2+ A2. 共9兲 B. d⬎1 recursion relationsThe Migdal-Kadanoff renormalization-group procedure generalizes our transformation to d⬎1 through a
bond-moving step.26,27Equation共7兲 can be expressed as a mapping
of interaction constants K =兵G,t,J,V,其 onto renormalized interaction constants, K
⬘
= R共K兲. The Migdal-Kadanoff pro-cedure strengthens by a factor of bd−1 the bonds of linear decimation, to account for a bond-moving effect.26,27 Theresulting recursion relations for d⬎1 are,
K
⬘
= bd−1R共K兲, 共10兲 which explicitly aret
⬘
=b d−1 2 ln ␥4 ␥2 , J⬘
= bd−1ln␥6 ␥7 , V⬘
=b d−1 4 ln ␥1 4␥ 6␥7 3 ␥2 4␥ 4 4 , ⬘
= bd−1+b d−1 2 ln冉
␥2␥4 ␥12冊
, G⬘
= bdG + bd−1ln␥1. 共11兲 This approach has been successfully employed in studies of a large variety of quantum mechanical and classical 共e.g., ref-erences in Ref.4兲 systems.C. Calculation of phase diagrams and expectation values The global flows of Eq. 共10兲, controlled by stable and
unstable fixed points, yield the phase diagrams in tempera-ture versus chemical potential:28 The basin of attraction of
each fixed point corresponds to a single thermodynamic phase or to a single type of phase transition, according to whether the fixed point is completely stable共a phase sink兲 or unstable. Eigenvalue analysis of the recursion matrix at an unstable fixed point determines the order and critical expo-nents of the phase transitions at the corresponding basin.
Table I gives the interaction constants t , J , V , at the
t − J model phase sinks. The tJandHbphases are the only regions where the electron-hopping term t does not renor-malize to zero at the phase sinks. On the contrary, in these phases, t→⬁ and t→−⬁, respectively.
To compute temperature versus electron density共doping兲 phase diagrams, thermodynamic densities are calculated by
summing along entire renormalization-group flow
trajectories.29A density, namely, the expectation value of an
operator in the Hamiltonian, is given by
M␣= 1
Nd
ln Z
K␣ , 共12兲
where K␣ is an element of K =兵K␣其, Z is the partition func-tion, and N is the number of lattice sites. The recursion rela-tions for densities are
M␣= b−d
兺
M
⬘
T␣, where T␣⬅K⬘
K␣. 共13兲
In terms of the density vector M =兵M␣其 and the recursion matrix T =兵T␣其, T =
冢
bd G⬘
t G⬘
J G⬘
V G⬘
0 t⬘
t t⬘
J t⬘
V t⬘
0 J⬘
t J⬘
J J⬘
V J⬘
0 V⬘
t V⬘
J V⬘
V V⬘
0 ⬘
t ⬘
J ⬘
V ⬘
冣
, 共14兲 Equation共13兲 simply is M = b−dM⬘
· T. 共15兲 At a fixed point, the density vector M␣= M␣⬘
⬅M␣ⴱ is the left eigenvector, with eigenvalue bd, of the fixed-point recursion matrix Tⴱ共TableII兲. For nonfixed-points, iterating Eq. 共15兲 ntimes,
M = b−ndM共n兲· T共n兲· T共n−1兲¯ T共1兲, 共16兲 where, for n large enough, the trajectory arrives as close as desired to a completely stable 共phase sink兲 fixed point and
M共n兲⯝Mⴱ. The latter density vector Mⴱis the left eigenvec-tor of the recursion matrix with eigenvalue bd. When two
K
K
2FIG. 1. Construction of the further-neighbor models. Part of a single plane of the three-dimensional model studied here is shown. TABLE I. Interaction constants at the phase sinks.
Phase
Interaction constants at sink
t J V d共dilute disordered兲 0 −⬁ 0 0 D共dense disordered兲 0 ⬁ 0 0 AF共antiferromagnetic兲 0 ⬁ −⬁ −⬁ V J→ 1 4 tJ共BEC-like superconductor兲 ⬁ ⬁ ⬁ −⬁ t →1 J→2 VJ→− 3 4 Hb共BCS-like superconductor兲 −⬁ ⬁ −⬁ −⬁ t →−1 J→−2 VJ→ 1 4
such density vectors exist, the two branches of the phase separation of a first-order phase transition are obtained,29,30
as illustrated with the phase separations found below. IV. FURTHER-NEIGHBOR INTERACTIONS, TEMPERATURE- AND DOPING-DEPENDENT FRUSTRATION, AND GLOBAL PHASE DIAGRAMS For the results presented below, we use the theoretically and experimentally dictated initial conditions of V/J=1/4 and t/J=2.25.
The details of the thermodynamic phases found in this work, listed in Tables I and II, have been discussed
previ-ously within context of the nearest-neighbor t − J 共Refs.5–7兲
and, for the Hb phase, Hubbard4 models. The Hb phase is
seen here in the t − J model with the inclusion of the further-neighbor antiferromagnetic or electron-hopping interaction. Suffice it to recall here that thephases are the only phases in which: 共1兲 the electron-hopping strength does not renor-malize to zero, but to infinity; 共2兲 the electron density does not renormalize to complete emptiness or complete filling, but to partial emptiness/filling, leaving room for electron/ hole conductivity; 共3兲 the nearest-neighbor electron occupa-tion probability does not renormalize to zero or unity, again leaving room for conductivity at the largest length scales;共4兲
0 0.1 0.2 0.3 0.4 0.5 d D −0.5 0 0.1 0.2 0.3 0.4 0.5 D d −0.5 0 0.1 0.2 0.3 0.4 Temperature 1 /t −0.25 d D 0 0.1 0.2 0.3 0.4 d D −0.1875 0 0.1 0.2 0.3 0.4 D d −0.1875 0 0.1 0.2 0.3 0.4 d D −0.25 −0.50 0 0.5 1 1.5 2 2.5 0.1 0.2 0.3 0.4 /J d d D −0.125 0.5 0.6 0.7 0.8 0.9 1 0.4 0 0.1 0.2 0.3 0.4 <ni> d D −0.125 d −0.25 d D AF τtJ τHb
FIG. 2. Global phase diagram of the further-neighbor t2model for t/J=2.25, in temperature vs chemical potential 共first column兲 and,
correspondingly, temperature versus electron density共second column兲. The relation t/J=2.25 is used for all renormalization-group trajectory initial conditions. The t2/t values are given in boxes. The dilute disordered 共d兲, dense disordered 共D兲, antiferromagnetic AF 共lightly colored兲,
tJ 共medium colored兲, and Hb 共darkly colored兲 phases are seen. Second-order phase transitions are drawn with full curves, first-order
transitions with dotted curves. Phase separation occurs between the dense共D兲 and dilute 共d兲 disordered phases, in the unmarked areas within the dotted curves in the electron density vs temperature diagrams. These areas are bounded, on the right and the left, by the two branches of phase separation densities, evaluated by renormalization-group theory as explained in Sec.III C. Note that these coexistence regions between dense共D兲 and dilute 共d兲 disordered phases are very narrow. Dashed curves are not phase transitions, but disorder lines between the dense and dilute disordered phases. As explained in the text, on each side of the thick full curves共not a phase boundary兲, the further-neighbor electron hopping affects the phases oppositely. On the dash-dotted curve 共also not a phase boundary; overlaps, for t2/t=0, with the thick full curve兲
the electron-hopping expectation value is nonzero at the larg-est length scales; 共5兲 the experimentally observed chemical potential shift as a function of doping occurs;6and共6兲 a low
level共⬃6%兲 of quenched nonmagnetic impurities causes to-tal disappearance, in contrast to the antiferromagnetic phase 共⬃40% for total disappearance兲,7 again as seen
experimen-tally. The low-temperature behavior and critical exponent of the specific heat4 have pointed to BCS-like and BEC-like
behaviors for theHbandtJphases, respectively.
The only approximations in obtaining the results below are the Suzuki-Takano and Migdal-Kadanoff procedures, ex-plained above in Secs.III AandIII B, respectively. There are no further assumptions in Secs. IV AandIV Bbelow.
A. t2model
The t2 model includes further-neighbor electron-hopping interaction, as shown in Fig. 1. The three-site Hamiltonian, between the lattice nodes at the lowest length scale, has the form: −H共i, j,k兲 = −H共i, j兲 −H共j,k兲 − t2
兺
共ci †c k+ ck†ci兲, 共17兲 where −H共i, j兲 is given in Eq. 共2兲, so that the first equationof Eq.共7兲 gets modified as t
⬘
=12ln ␥4
␥2
+ t2, 共18兲
only for the first renormalization. Thus, for d = 3, the first equation of Eq. 共11兲 gets modified as
t
⬘
= 2 ln␥4 ␥2+ 4t2, 共19兲
only for the first renormalization. Thus, the hopping strength
t2contributes to the first renormalization, but is not regener-ated by this first renormalization. Note that the quantitative memory of the further-neighbor interaction is kept in all sub-sequent renormalization-group steps, as the flows are modi-fied by the different values of the first-renormalized
interac-tions due to the effect of the further-neighbor interaction. The subsequent global renormalization-group flows are in the space of t , J , V ,, as is the case in position-space renormalization-group treatments31–33 using a prefacing
transformation. Which surfaces in this large 共four-dimensional兲 flow space of t,J,V, are accessed is con-trolled by the original further-neighbor interaction. Thus, the further-neighbor interaction t2shifts the value of t
⬘
obtained after the first renormalization-group transformation, as dic-tated by the physical model 共Fig.1兲. Since the value of thefirst-renormalized t
⬘
in the absence of t2already has a com-plicated dependence on the unrenormalized temperature and electron density, the variety of phase diagrams is obtained. Indeed, the effect of the further-neighbor interaction is de-pendent on the electron density, temperature, and other inter-actions in the system, due to the presence of the first term in Eq. 共19兲, which is the key to the resulting spectacularlydif-ferent phase diagrams as the further-neighbor interaction is varied.共1兲 If the two terms in Eq. 共18兲 are of the same sign,
the nearest-neighbor and further-neighbor electron-hopping terms of the original system reinforce each other and the phases are enhanced. 共2兲 If the two terms are of opposite signs, the nearest-neighbor and further-neighbor electron-hopping terms of the original system compete with each other and, with the introduction of further-neighbor electron hopping, the phases are initially suppressed, but enhanced as further-neighbor hopping becomes dominant. The two re-gimes共1兲 and 共2兲 are separated by the thick full lines in the phase diagrams in Figs. 2and3. In the case 共2兲 of opposite signs, when the two terms cancel out each other, the system is frustrated, in which case, after the first renormalization, there is no electron hopping in the system. Since this condi-tion is closed under renormalizacondi-tion, both on physical grounds and of course in our recursion relations关Eq. 共7兲兴, no
phase can occur in such a system. The dash-dotted curves in Figs. 2 and3 indeed show such systems. These competi-tion, reinforcement, and frustration effects are temperature and doping dependent. These, and all other physical effects, do not depend on the sign of nearest-neighbor t of the origi-nal unrenormalized system, due to the symmetry of hypercu-bic lattices5and as seen in Eq.共9兲.
Figures 2 and 3 give the global phase diagram of the t2 model, as a function of temperature, electron density, chemi-cal potential, and t2/t. The values of the hopping-strength ratios t2/t for the consecutive panels in these figures are cho-sen so that they sequentially produce the qualitatively differ-ent phase-diagram cross sections, thereby revealing the evo-lution in the global phase diagram. Second-order phase transitions are drawn with full curves, first-order transitions with dotted curves. Phase separation occurs between the dense共D兲 and dilute 共d兲 disordered phases, in the unmarked areas within the dotted curves in the electron density vs tem-perature diagrams. These areas are bounded, on the right and the left, by the two branches of phase separation densities, evaluated by renormalization-group theory as explained in Sec.III C. Note that these coexistence regions between dense 共D兲 and dilute 共d兲 disordered phases are very narrow.
The cross-section t2= 0 is the phase diagram obtained in previous work.5This diagram contains thetJphase between
33% and 37% hole doping away from half filling and at TABLE II. Expectation values at the phase sinks. The
expecta-tion values at a sink epitomize the expectaexpecta-tion values throughout its corresponding phase because, as explained in Sec.III C, the expec-tation values at the phase sink underpin the calculation of the ex-pectation values throughout the corresponding phase which is con-stituted from the basin of attraction of the sink.
Phase sinks
Expectation values at sink
兺具ci†cj+ cj†ci典 具ni典 具Si· Sj典 具ninj典 d 0 0 0 0 D 0 1 0 1 AF 0 1 14 1 tJ − 2 3 2 3 − 1 4 1 3 Hb 0.664 0.668 0.084 0.336
temperature 1/t⬍0.12. The thick full curve here gives the systems devoid of electron hopping due to the combined effects of temperature and doping on a nearest-neighbor-only interaction system. The first term of Eq. 共18兲 is positive on
the high density/chemical potential, low temperature side of the thick full curve and negative on the low chemical potential/density, high-temperature side of the thick full curve. Thus, the inclusion of t2⬎0 will create competition and frustration共respectively reducing and eliminating the phases兲 on the low chemical potential/density, high-temperature side of the curve discussed here, reinforcement 共enhancing the phases兲 on the high chemical potential/ density, low-temperature side of the same curve. The oppo-site occurs at t2⬍0. The thick full 共no hopping兲 curve of t2 = 0 is included, again as thick and full, in the t2⫽0 phase diagrams and the effects discussed here are seen in the evo-lution, in both directions, of the global phase diagram.
Pursuing the negative values of t2, we see at t2/t= −0.0625 that the tJphase, being below the thick full curve, is indeed reduced and bisected into two disconnected regions by the frustration 共dash-dotted兲 curve. At the more negative
value of t2/t=−0.125, only the higher doping region of the
tJ phase remains and is enhanced as explained after Eq. 共18兲, extending through a wider range to 45%–55% hole
doping. The antiferromagnetic and disordered phases take part in a complex lamellar structure, in a narrow band be-tween 35% and 45% hole doping at low temperatures. At the even more negative values of t2/t=−0.25 and −0.5, the tJ phase appears in a wide range of hole doping, between 35% and 55%. Besides the complex lamellar structure of antifer-romagnetic and disordered phases, we also see that the Hb phase participates in the lamellar phase structure and, sepa-rately, appears adjacently to the antiferromagnetic phase near half filling. Particularly near half filling, theHbphase which evolves adjacently to the antiferromagnetic phase reaches to the higher temperatures of 1/t⬃0.5. This topology is identi-cal to that obtained for the Hubbard model.4
For the positive values of t2/t, thephases are enhanced as explained after Eq.共18兲 and the topology quickly evolves
to that encountered in the Hubbard model. ThetJ phase is not bisected by the frustration 共dash-dotted兲 curve and ap-pears between 33%–37% hole doping as a continuation of
0 0.1 0.2 0.3 0.4 0.5 d D −0.0625 0 0.1 0.2 0.3 0.4 0.5 d D −0.0625 0 0.1 0.2 0.3 0.4 Temperature 1/t D d 0 0 0.1 0.2 0.3 0.4 D d 0 0 0.1 0.2 0.3 0.4 D d 0.125 0 0.1 0.2 0.3 0.4 d D 0.125 −0.50 0 0.5 1 1.5 2 2.5 0.1 0.2 0.3 0.4 /J d D 0.5 0.5 0.6 0.7 0.8 0.9 1 0.4 0 0.1 0.2 0.3 0.4 <ni> d D 0.5 0.5 0.125 d d D D AF τtJ τHb
the structure at t2= 0. The Hb phase occurs again in two distinct regions and the one which lies nearer to half filling again extends to high temperatures.
B. J2model
The J2model includes further-neighbor antiferromagnetic interaction, as shown in Fig. 1. The three-site Hamiltonian, between the lattice nodes at the lowest length scale, has the form:
−H共i, j,k兲 = −H共i, j兲 −H共j,k兲 − J2
兺
具ik典Si· Sk, 共20兲 where −H共i, j兲 is given in Eq. 共2兲, so that the second
equa-tion of Eq.共7兲 gets modified as J
⬘
= ln␥6 ␥7+ J2, 共21兲
only for the first renormalization. Thus, for d = 3, the second equation of Eq. 共11兲 gets modified as
0 0.1 0.2 0.3 0.4 0.5 d −0.5 D 0 0.1 0.2 0.3 0.4 0.5 d D −0.5 0 0.1 0.2 0.3 0.4 Temperature 1/t d D −0.25 0 0.1 0.2 0.3 0.4 d D −0.25 0 0.1 0.2 0.3 0.4 D d −0.125 0 0.1 0.2 0.3 0.4 d D −0.125 −0.50 0 0.5 1 1.5 2 2.5 0.1 0.2 0.3 0.4 /J D d 0 0.5 0.6 0.7 0.8 0.9 10 0.1 0.2 0.3 0.4 <ni> d 0 D AF τtJ τHb
FIG. 4. Global phase diagrams of the further-neighbor J2model for t/J=2.25, in temperature vs. chemical potential 共first column兲 and, correspondingly, temperature versus electron density共second column兲. The relation t/J=2.25 is used for all renormalization-group trajectory initial conditions. The J2/J values are given in boxes. The dilute disordered 共d兲, dense disordered 共D兲, antiferromagnetic AF 共lightly colored兲, tJ 共medium colored兲, and Hb 共darkly colored兲 phases are seen. Second-order phase transitions are drawn with full curves, first-order
transitions with dotted curves. Phase separation occurs between the dense共D兲 and dilute 共d兲 disordered phases, in the unmarked areas within the dotted curves in the electron density vs. temperature diagrams. These areas are bounded, on the right and the left, by the two branches of phase separation densities, evaluated by renormalization-group theory as explained in Sec. III C. Note that these coexistence regions between dense共D兲 and dilute 共d兲 disordered phases are very narrow. Dashed curves are not phase transitions, but disorder lines between the dense and dilute disordered phases. As explained in the text, on each side of the thick full curves共not a phase boundary兲, the further-neighbor interaction affects the antiferromagnetic phase oppositely. On the dash-dotted curve共also not a phase boundary; overlaps, for J2/J=0, with
J
⬘
= 4 ln␥6 ␥7+ 4J2, 共22兲
only for the first renormalization. Again, the interaction J2 contributes to the first renormalization, but is not regenerated by this first renormalization. Reinforcement or competition occurs when J2 is, respectively, of same or opposite sign as the first term in Eq.共22兲. These two regimes are again
sepa-rated by the thick full lines in the phase diagrams of Figs.3
and 4, while again frustration occurs on the dash-dotted lines. In the reinforcement regime, we expect a large extent of the antiferromagnetic phase. The Hb phase is also ex-pected to grow in the reinforced region, for it is found along the temperature extent of the antiferromagnetic phase.
Figures4 and5 show the global phase diagram of the J2 model, as a function of temperature, electron density, chemi-cal potential, and J2/J. Again, the values of the coupling-strength ratios J2/J for the consecutive panels in these fig-ures are chosen so that they sequentially produce the qualitatively different phase-diagram cross sections, thereby
revealing the evolution in the global phase diagram. Again, the phase separation regions of the first-order phase transi-tions are very narrow. For negative values of J2/J, the
TABLE III. The two-site basis states, with the corresponding particle number 共n兲, parity 共p兲, total spin 共s兲, and total spin z-component共ms兲 quantum numbers. The states 兩3典, 兩5典, and 兩8典 are obtained by spin reversal from兩2典, 兩4典, and 兩7典, respectively.
n p s ms Two-site eigenstates 0 + 0 0 兩1典=兩ⴰⴰ典 1 + 1/2 1/2 兩2典= 1 冑2兵兩↑ⴰ典+兩ⴰ↑典其 1 − 1/2 1/2 兩4典= 1 冑2兵兩↑ⴰ典−兩ⴰ↑典其 2 − 0 0 兩6典=冑12兵兩↑↓典−兩↓↑典其 2 + 1 1 兩7典=兩↑↑典 2 + 1 0 兩9典= 1 冑2兵兩↑↓典+兩↓↑典其 0 0.1 0.2 0.3 0.4 0.5 AF D d 0.125 0 0.1 0.2 0.3 0.4 0.5 D d 0.125 D 0 0.1 0.2 0.3 0.4 Temperature 1/t d D 0.25 0 0.1 0.2 0.3 0.4 d D 0.25 D 0 0.1 0.2 0.3 0.4 d D 0.3125 0 0.1 0.2 0.3 0.4 D d 0.3125 D −0.50 0 0.5 1 1.5 2 2.5 0.1 0.2 0.3 0.4 /J d D 0.5 0.5 0.6 0.7 0.8 0.9 10 0.1 0.2 0.3 0.4 <ni> d D D D 0.5 AF τtJ τHb
antiferromagnetic phase is enhanced, both near half filling by the mechanism explained after Eq. 共22兲 and, separately and
to a lesser extent, displacing thetJphase. The latter behav-ior is similar to that seen under the introduction of quenched
impurities, both experimentally34–36 and from
renormalization-group theory.7TheHbphase improves near
the large antiferromagnetic region near half filling. At J2/J = −2, theHbphase is found in a wide range of hole doping, namely between 15% and 30%. Another interesting result is that the tJ phase is depressed in temperature but remains stable in the interval of 33%–37% hole doping.
For positive values of J2/J, the antiferromagnetic phase is reduced in the region near half filling and enhanced in the region near the tJ phase, for reasons explained after Eq. 共22兲. The Hb phase grows adjacently to the enhanced anti-ferromagnetic region, being located above the tJ phase, causing a complex structure at higher hole dopings and low temperatures.
V. CONCLUSION
We have shown that the t − J model with further-neighbor antiferromagnetic 共J2兲 or further-neighbor electron hopping 共t2兲 interactions exhibits extremely rich global phase dia-grams. The phase separation regions of the first-order phase transitions are very narrow. Furthermore, these calculated phase diagrams are understood in terms of the competition and frustration of nearest- and further-neighbor interactions. We find that the two types ofphases, previously seen in the Hubbard model, occur in the t − J model with the inclusion of further-neighbor interactions.
ACKNOWLEDGMENTS
This research was supported by the Scientific and Tech-nological Research Council of Turkey 共TÜBİTAK兲 and by the Academy of Sciences of Turkey.
APPENDIX: DERIVATION OF THE DECIMATION RELATIONS
The derivation of Eq.共7兲, first done in Ref.5, is given in this Appendix. In Eq. 共5兲 the operators −
⬘
H⬘
共i,k兲 and−H共i, j兲−H共j,k兲 act on two-site and three-site states,
re-spectively, where at each site an electron may be either with spin=↑ or ↓, or may not exist 共0 state兲. In terms of matrix elements,
具uivk兩e−⬘H⬘共i,k兲兩u¯i¯vk典 =
兺
wj
具uiwjvk兩e−H共i,j兲−H共j,k兲兩u¯iwj¯vk典, 共A1兲 where ui, wj,vk, u¯i,¯vkare single-site state variables, so that the left-hand side reflects a 9⫻9 and the right-hand side a 27⫻27 matrix. Basis states that are simultaneous eigenstates of total particle number 共n兲, parity 共p兲, total spin magnitude 共s兲, and total spin z-component 共ms兲 block-diagonalize Eq. 共A1兲 and thereby make it manageable. These sets of nine
two-site and 27 three-site eigenstates, denoted by 兵兩p典其 and 兵兩q典其, respectively, are given in TablesIIIandIV. Equation 共A1兲 is thus rewritten as
TABLE IV. The three-site basis states, with the corresponding particle number 共n兲, parity 共p兲, total spin 共s兲, and total spin z-component 共ms兲 quantum numbers. The states 兩4–5典, 兩7典,
兩15–16典, 兩19典, 兩21典, 兩23典, and 兩26–27典 are obtained by spin
rever-sal from 兩2–3典, 兩6典, 兩11–12典, 兩17典, 兩20典, 兩22典, and 兩24–25典,
respectively. n p s ms Three-site eigenstates 0 + 0 0 兩1典=兩ⴰⴰⴰ典 1 + 1/2 1/2 兩2典=兩ⴰ↑ⴰ典, 兩3典=冑1 2兵兩↑ⴰⴰ典+兩ⴰⴰ↑典其 1 − 1/2 1/2 兩6典=冑1 2兵兩↑ⴰⴰ典−兩ⴰⴰ↑典其 2 + 0 0 兩8典= 1 2兵兩↑↓ⴰ典−兩↓↑ⴰ典−兩ⴰ↑↓典+兩ⴰ↓↑典其 2 − 0 0 兩9典= 1 2兵兩↑↓ⴰ典−兩↓↑ⴰ典+兩ⴰ↑↓典−兩ⴰ↓↑典其, 兩10典= 1 冑2兵兩↑ⴰ↓典−兩↓ⴰ↑典其 2 + 1 1 兩11典=兩↑ⴰ ↓典, 兩12典= 1 冑2兵兩↑↑ⴰ典+兩ⴰ↑↑典其 2 + 1 0 兩13典= 1 2兵兩↑↓ⴰ典+兩↓↑ⴰ典+兩ⴰ↑↓典+兩ⴰ↓↑典其, 兩14典= 1 冑2兵兩↑ⴰ↓典+兩↓ⴰ↑典其 2 − 1 1 兩17典= 1 冑2兵兩↑↑ⴰ典−兩ⴰ↑↑典其 2 − 1 0 兩18典= 1 2兵兩↑↓ⴰ典+兩↓↑ⴰ典−兩ⴰ↑↓典−兩ⴰ↓↑典其 3 + 1/2 1/2 兩20典=冑16兵2兩↑↓↑典−兩↑↑↓典−兩↓↑↑典其 3 − 1/2 1/2 兩22典=冑1 2兵兩↑↑↓典−兩↓↑↑典其 3 + 3/2 3/2 兩24典=兩↑↑↑典 3 + 3/2 1/2 兩25典= 1 冑3兵兩↑↓↑典+兩↑↑↓典+兩↓↑↑典其
TABLE V. Block-diagonal matrix of the renormalized two-site Hamiltonian −⬘H⬘共i,k兲. The Hamiltonian being invariant under spin reversal, the spin-flipped matrix elements are not shown.
φ1 φ2 φ4 φ6 φ7 φ9 φ1 G′ φ2 −t′+ µ′+ G′ 0 φ4 t′+µ′+G′ φ6 3 4J′+ V′+ 2µ′+ G′ φ7 0 −1 4J′+ V′+ 2µ′+ G′ φ9 −1 4J′+ V′+ 2µ′+ G′
具p兩e−⬘H⬘共i,k兲兩p ¯典 =
兺
u,v,u¯, v ¯,w兺
q,q¯ 具p兩uivk典具uiwjvk兩q典 ⫻具q兩e−H共i,j兲−H共j,k兲兩q ¯典具q¯兩u¯iwj¯vk典 ⫻具u¯i¯vk兩p¯典. 共A2兲There are five independent elements for具p兩e−⬘H⬘共i,k兲兩p
¯典 in
Eq. 共A2兲 共thereby leading to five renormalized interaction
constants兵t
⬘
, J⬘
, V⬘
,⬘
, G⬘
其兲, which we label␥p,␥p⬅ 具p兩e−⬘H⬘共i,k兲兩p典 for p = 1,2,4,6,7. 共A3兲 The diagonal matrix具p兩−
⬘
H⬘
共i,k兲兩p¯典 is given in TableV. The exponential of this matrix yields the five renormalized interaction constants in terms of ␥p, as given in Eq. 共7兲.Furthermore, according to Eq.共A2兲, each␥pis a linear com-bination of some具q兩e−H共i,j兲−H共j,k兲兩q¯典,
␥1=具1兩兩1典 + 具2兩兩2典 + 具4兩兩4典, ␥2=具3兩兩3典 + 1 2具8兩兩8典 + 具12兩兩12典 + 1 2具13兩兩13典, ␥4=具6兩兩6典 + 1 2具9兩兩9典 + 具17兩兩17典 + 1 2具18兩兩18典, ␥6=具10兩兩10典 + 2具22兩兩22典, ␥7=具11兩兩11典 + 2 3具20兩兩20典 + 4 3具24兩兩24典,
where具q兩兩q典⬅具q兩e−H共i,j兲−H共j,k兲兩q典. In order to calculate 具q兩e−H共i,j兲−H共j,k兲兩q
¯典 the matrix blocks in TableVIare nu-merically exponentiated.
TABLE VI. Diagonal matrix blocks of the unrenormalized three-site Hamiltonian −H共i, j兲−H共j,k兲. The Hamiltonian being invariant under spin-reversal, the spin-flipped matrix elements are not shown. ψ1 ψ1 0 ψ2 ψ3 ψ2 2µ −√2t ψ3 −√2t µ ψ6 ψ8 ψ6 µ 0 ψ8 0 34J + V + 3µ ψ9 ψ10 ψ9 34J + V + 3µ −√2t ψ10 −√2t 2µ ψ11 ψ12 ψ11 2µ −√2t ψ12 −√2t −14J + V + 3µ ψ13 ψ14 ψ13 −14J + V + 3µ −√2t ψ14 −√2t 2µ ψ17 ψ18 ψ17 −14J + V + 3µ 0 ψ18 0 −14J + V + 3µ ψ20 ψ20 J + 2V + 4µ ψ22 ψ22 2V + 4µ ψ24 ψ24 −12J + 2V + 4µ ψ25 ψ25 −12J + 2V + 4µ
1J. Hubbard, Proc. R. Soc. London, Ser. A 276, 238共1963兲; 277,
237共1964兲; 281, 401 共1964兲.
2P. W. Anderson, Science 235, 1196共1987兲.
3G. Baskaran, Z. Zhou, and P. W. Anderson, Solid State Commun. 63, 973共1987兲.
4M. Hinczewski and A. N. Berker, Eur. Phys. J. B 48, 1共2005兲. 5A. Falicov and A. N. Berker, Phys. Rev. B 51, 12458共1995兲. 6M. Hinczewski and A. N. Berker, Eur. Phys. J. B 51, 461
共2006兲.
7M. Hinczewski and A. N. Berker, Phys. Rev. B 78, 064507
共2008兲.
8A. V. Puchkov, P. Fournier, T. Timusk, and N. N. Kolesnikov,
Phys. Rev. Lett. 77, 1853共1996兲.
9C. Bernhard, J. L. Tallon, T. Blasius, A. Golnik, and C.
Nieder-mayer, Phys. Rev. Lett. 86, 1614共2001兲.
10G. Jüttner, A. Klümper, and J. Suzuki, Nucl. Phys. B 522, 471
共1998兲.
11G. Jüttner, A. Klümper, and J. Suzuki, Physica B 259-261, 1019
共1999兲.
12M. Hinczewski, Ph. D. thesis, Massachusetts Institute of
Tech-nology共2005兲.
13M. Ogata, M. U. Luchini, S. Sorella, and F. F. Assaad, Phys. Rev.
Lett. 66, 2388共1991兲.
14S. Krinsky and D. Furman, Phys. Rev. Lett. 32, 731共1974兲. 15S. Krinsky and D. Furman, Phys. Rev. B 11, 2602共1975兲. 16S. A. Kivelson, V. J. Emery, and H. Q. Lin, Phys. Rev. B 42,
6523共1990兲.
17W. O. Putikka, M. U. Luchini, and T. M. Rice, Phys. Rev. Lett. 68, 538共1992兲.
18E. Dagotto, Rev. Mod. Phys. 66, 763共1994兲.
19M. Suzuki and H. Takano, Phys. Lett. A 69, 426共1979兲. 20H. Takano and M. Suzuki, J. Stat. Phys. 26, 635共1981兲. 21P. Tomczak, Phys. Rev. B 53, R500共1996兲.
22P. Tomczak and J. Richter, Phys. Rev. B 54, 9004共1996兲. 23P. Tomczak and J. Richter, J. Phys. A 36, 5399共2003兲. 24C. N. Kaplan and A. N. Berker, Phys. Rev. Lett. 100, 027204
共2008兲.
25O. S. Sarıyer, A. N. Berker, and M. Hinczewski, Phys. Rev. B 77, 134413共2008兲.
26A. A. Migdal, Zh. Eksp. Teor. Fiz. 69, 1457共1975兲 关Sov. Phys.
JETP 42, 743共1976兲兴.
27L. P. Kadanoff, Ann. Phys.共N.Y.兲 100, 359 共1976兲. 28A. N. Berker and M. Wortis, Phys. Rev. B 14, 4946共1976兲. 29S. R. McKay and A. N. Berker, Phys. Rev. B 29, 1315共1984兲. 30M. E. Fisher and A. N. Berker, Phys. Rev. B 26, 2507共1982兲. 31A. N. Berker, S. Ostlund, and F. A. Putnam, Phys. Rev. B 17,
3650共1978兲.
32R. G. Caflisch and A. N. Berker, Phys. Rev. B 29, 1279共1984兲. 33R. G. Caflisch, A. N. Berker, and M. Kardar, Phys. Rev. B 31,
4527共1985兲.
34M.-H. Julien, T. Fehér, M. Horvatić, C. Berthier, O. N.
Ba-kharev, P. Ségransan, G. Collin, and J.-F. Marucco, Phys. Rev. Lett. 84, 3422共2000兲.
35I. Watanabe, T. Adachi, K. Takahashi, S. Yairi, Y. Koike, and K.
Nagamine, Phys. Rev. B 65, 180516共R兲 共2002兲.
36Y. Itoh, T. Machi, C. Kasai, S. Adachi, N. Watanabe, N.