Download as pdf
Download as pdf
You are on page 1of 4
Journal of the Physical Society of Japan Vol. 58, No. 1, January, 1989, pp. 32-35 Mean Field Theory of the Square Lat Letters e Antiferromagnetic Heisenberg Model Daijiro YosHioKa College of General Education, Kyushu University, Ropponmatsu, Fukuoka 810 (Received August 29, 1988) ‘A nearly half-filled wo-dimensional Hubbard model has been considered as a ‘model for a high transition temperature oxide superconductor. A new scheme to treat this model, the slave-fer ‘method, is proposed. The method is applied to the ‘model in the limit of large U at half-flled, where the model reduces to the antifer- romagnetic Heisenberg model. It is shown that the ground state energy becomes ‘much lower in the slave-fermion scheme than that in the slave-boson scheme. The ‘ground state shows antiferromag Since the discovery of the oxide high transi- tion temperature superconductors," inten- sive investigations on strongly correlated electron systems have been conducted. The common feature of the high-T. superconduc- tors is the presence of Cu-O two-dimensional planes, a factor widely considered to be responsible for the superconductivity. Ander- son” proposed to model this Cu-O plane by the square lattice Hubbard model. He claimed that the resonating valence bond (RVB) state is realized in this system, and that the RVB causes the superconductivity. Although numerical calculations seem to indicate that this model does not give rise to the supercon- ductivity, it is still quite important to in- vestigate this model analytically. The Hamiltonian for the Hubbard model is given as follows —tY) Dice hee. D+ UZ manus ch a) where ci,(cie) is the creation (annihilation) operator of an electron with spin o(: or 1) at site i, the first sum runs over the nearest neighbor bonds (i, j), and Mic=CirCe. In the large U limit this Hamiltonian is expanded as a series in 1/U.*" In the lowest order, the effective Hamiltonian is given by H=—-1Y) DIA n-0eeGo — ne Nj-0) long range order at T° 32 t+hel+S Dy (s-s, an Q) where J=4f°/U, S; is the spin operator at site i, m=ny+n, and 36" only consists of terms which involve three sites. At the half-filled and large U limit, each site is occupied by an elec- tron. Therefore the first and third terms of eq. (2) vanish, and the model becomes an antifer- romagnetic Heisenberg model. In this case, the RVB ground state is a superposition of dimer coverings of the whole lattice, where each dimer is a pair of sites whose spins form the singlet state." The elementary excitation from the RVB ground state is a spinon (un- paired spin), and in the case of less than half- filled, a holon (the site where there is no elec- tron). There have been a lot of investigations on variational wave function based on the con- cept of the RVB state."' Another well-studied method is the slave- boson formulation.!"° In this method, in order to explicitly describe the four possible states of a site, empty, occupied by up-spin electron, occupied by down-spin electron, and doubly occupied, four operators ¢;, sir, six and d, are introduced respectively with the aux- iliary condition: eJe+sisn+shsu+did; Here e,(d;) represent the empty (doubly oc- cupied) state, and si, represents the singly occupied state with spin 6. The electron operator ci, is expressed as Cla Lerters 3 =eslo+sgn (a)d!s\-0, and if we consider e, and d; as bose operators and Si as a fermi operator, anticommutation relations for cj, and Cy are satisfied. In the large U limit we can neglect the doubly occupied state. Then the Hamiltonian in this formulation is written by e, e/, sic, and sl In the mean field theory, Hartree-Fock- Gorkov decoupling of this Hamiltonian is done and the RVB state or superconductivity has been discussed. In so doing, the auxiliary condition for each site is usually replaced by a global condition. In this formulation, sig can be considered as spinon operator and ¢; as holon operator. The assignment that si, is a fermi operator and e; is a bose operator corresponds to Kivelson et al’s discussion that spinon is a fermion and holon is a boson. However, the anticommuta- tion relations for the electron operators cj, and ¢jy are also satisfied, even if we reverse the assignment of the statistics to si, and e, namely, we can consider sj, as a boson operator and @ as a (slave) fermion operator.” The mean field calculation for this case can also be performed. We will show below that this assignment gives a much lower ground state energy in the half-filled case. The case for less than half-filled will be discussed in a separate paper.” It should be noted that this assignment of the statistics is compatible with the latest theory of the spinon and holon in the RVB state where spinon (holon) is found to obey bose (fermi) statistics." In the following we only consider the half- filled case. Thus =0, and the Hamilto- nian is the same as the Heisenberg model, which is written only by the bose fields s}, and Sic. In this case sj, and sig are referred to as ‘Schwinger bosons. The mean field calculation for this formalism has already been done by Arovas and Auerbach®? and Auerbach and Arovas.”” However they only gave the results for the s-wave state and they could not obtain a solution at T=0. In this letter we show that there is the Bose-Einstein condensation at T=0. We also consider s+id wave state, which is known by variational calculations"? to give lower energy. It should be noted that their Schwinger-boson scheme only applies to the Heisenberg model, but the present scheme is applicable to the Hubbard model, and the Heisenberg limit treated here is an example to show how our scheme gives a lower energy than the ordinary slave-boson scheme. Now our Hamiltonian is given as -aS)-0Sia) J + H=F YD [oS}-oSSi-e-S 2 wont) “o +n DO Dshsie @) Here, chemical potential u is introduced to ensure the condition that the system is half- filled: KSinSio) @ 2 ‘We decouple this Hamiltonian introducing the following order parameters, A, and Ay: A=} (susie Si5i081>s © wherez=x or y, and i+2(i+5) indicates a site next to the site i in the x(y) direction. These order parameters give amplitude for the single: pairing in the neighboring sites. The lattice we are considering is a square one with N lattice points. We introduce Fourier transform of sie, Siz. Then the mean field Hamiltonian is given by 1 => IN+2QINU A + Ay] FD Asher ts eisai) + Yast as r + yhsers—e], © where A=p~—J, and Me=2iJ(A,sink,+A,sin ky). a This #™" is diagonalized by the Bogoliubov transformation, if 2>ly|. We introduce bose operators a4 and fi such that Ate AEN" ay (ac) etter) en ® 4 Lerrers F=VE=InP, and 0, is the phase of ys: 74= Iyxle™. It is evi- dent that we must have 42 1)!. However, special caution is necessary, if at some k(=K;) \ye! is equal to A. In this case x, and Br, are not defined, and we have Bose condensation at such K;. There we introduce operators ¢x, and nnx, such that 1 Tz Gat as (0) 1 Sm = Tee COMER Me)- CD) Then the component of the Hamiltonian at K; is given by A(SksSks FS" KaS—KA) + PRSKIS™ KA YESKIS—K = +, 1 =u [cite I |: 2 The operators Cx, and ‘x, satisfy the following commutation relations: (Cx) Ck) =lna, mk) =0, [x5 k]=lnx, CxJ=1. Thus Cx, and fx, commute with the Hai iltonian. Since A>0, the ground state must be also an eigenstate of Gx, with zero eigenvalue. The self-consistency equations are given as follows. First from eq. (4) 1 we ([2( coth $* ) -ss+mdue. (13) where Kj is the value of k at which |E|=. and m=. At T>0 this equation is satisfied with temperature dependent finite gap for Ex, and no is zero, i.e. there is no Bose- Einstein condensation. On the other hand, at T=Dit is possible that Ex becomes gapless and ‘no becomes finite. As for the order parameter, eq. (5) is given by J Iv asin k(Ay sin kick Ay sin ky) A= A Ex xl coth ==) (I~ dix) mod, (4) We first consider the transition temperature T.. At Te, 4;=A,=0, thus E,=2. Equation (13) gives coth (4/T)=2, and eq. (14) gives J=A, Therefore T-=J/in 3=0.91 J. This is almost a factor of 4 higher than Tavs by the slave-boson scheme." Next we calculate ()-Si+29=—(@/4)(IAeI? +14,) at T=0, which is related to the ground state energy. To calculate it we must specify the phase between A, and 4,. 4,=A,, iA,, —A, are respectively called s, s+id and d wave state. The energies are the same for s and d wave states, which are lower than s-+id wave state, In the s wave state there is Bose-Einstein condensation at two values of Ki: Ki= = (n/2, x/2). After numerical integration of eqs. (13) and (14) we obtain mo/N=0.3034, 4=2.316 =A=A/4 J=0.5790, and 0.5029. However, the current best esti- mate for (5/-S,+:> is —0.334.?” Although the presen: value is larger by a factor of about 1.5, it is better than that of the slave-boson scheme, which is —0.086 for the s+id wave state. In the framework of the mean field theory, the solution which should be chosen as the ground state is the one with the lowest energy. Thus the present slave-fermion scheme should be preferable to the slave- boson scheme. The discrepancy between the present result and that of the variational calculation”? comes partly from the neglect of fluctuation and from the fact that the condi- tion of one electron at each site is satisfied only on the average. As a manifestation of the latter fact by this factor without justification, we get 1, it is given by rte 3 8 Cun aN Namely, there is the long-range Néel order. It should be noted that the Bose-Einstein conden- sation here does not imply superconductivity. The size of the long-range order, (3/2) n/N?=0.138 is again 1.5 times larger than the result of the Monte Carlo” or the spin- wave theory.”” Such a relation is also noticed in ref. (26) for the spin correlation at finite tem- perature. The present formulation is similar to the or- dinary slave-boson formalism. We assumed singlet pairing between neighboring spins. However, we obtained a much lower energy by the present scheme. Moreover, the ground state shows long-range Neel order. This is remarkable since our order parameter has the same form as that used to discuss RVB state in the slave-boson scheme. It should be noted that there is a long-range order at low tempera- ture for a three- or higher dimensional antifer- romagnetic Heisenberg model. Our scheme can naturally describe such a state. This is another merit of the present slave-fermion scheme. Although we considered only the half-filled case in this letter, our scheme could be ex- tended to a less than half-filled case. Since the ground state energy changes continuously when we change the filling factor, we can ex- pect that even away from the half-filled, our scheme gives a lower ground state energy than the ordinary slave-boson mean-field theory. The calculation for such a situation will be given in a separate paper.”? KSiSink (16) Part of this work was done while the author stayed at the Aspen Center for Physics in the summer of 1988. Comments by N. Read were useful and the hospitality of the Center was much appreciated. References 1) J. G, Bednorz and K. A. Miller: Z. Phys. B64 (1986) 189. 2) -M.K. Wu, J. R, Ashburn, C. J. Torng, P. H. Hor, R.L. Meng, L. Gao, Z. L. Huang, Y. Q. Wang and C. W. Chu: Phys. Rev. Lett. $8 (1987) 908. 3) H. Maeda, Y. Tanaka, Y. Fukutomi and T. Asano: Jpn. J. Appl. Phys. 27 (1988) L209. 4) Z.Z, Sheng and A. M. Herman: Nature 332 (1988) 55. 5). W. Anderson: Science 235 (1987) 1196. 6) M. Imada: J. Phys. Soc. Jpn. 57 (1988) 42. 7) J.E. Hirsch, E. Loh, D. J. Scalapino and S. Tang: Physica C 153-155 (1988) 549, 8) JE, Hirsch: Phys. Rev. Lett. 54 (1985) 1317. 9) C. Gros, R. Joint and T. M. Rice: Phys. Rev. B36 (1987) 381. 10) A. H. MacDonald, S. M. Girvin and D. Yoshioka: Phys. Rev. B37 (1988) 9753. 11) G. Baskaran, Z. Zou and P. W. Anderson: Solid State Commun. 63 (1987) 973. 12) S.A. Kivelson, D. $. Rokhsar and J, P. Sethna: Phys. Rev. B35 (1987) 8865. 13) P.W. Anderson, G. Baskaran, Z. Zou and T. Hsu: Phys. Rev. Lett. $8 (1987) 2790. 14) C.Gros, D. Poilblane, T. M. Rice and F. C, Zhang: Physica C153-185 (1988) 543. 15) B. Sutherland: Phys. Rev. B37 (1988) 3786. 16) H. Yokoyama and H. Shiba: J. Phys. Soe. Jpn. $6 (1987) 3570. 17) Z.Zouand P. W. Anderson: Phys. Rev. B37 (1988) 27. 18) A.E, Ruckenstein, P. J. Hirschfeld and J. Appel: Phys. Rev. B36 (1987) 857. 19) Y. Suzumura, Y. Hasegawa and H. Fukuyama: J. Phys. Soc. Jpn. $7 (1988) 401 and preprint. 20) G. Kotliar and J. Liu: preprint. 21) C.L. Kane, P. A. Lee and N. Read (preprint) also ‘considered the slave-fermion scheme. However, the order parameter is different from the present scheme. 22). Yoshioka: in preparation. 23) _N. Read and B. Chakraborty: preprint. 24) F.D. M, Haldane and H. Levine: preprint, 25) D.P. Arovas and A. Auerbach: Phys. Rev. B38 (1988) 316. 26) A. Auerbach and D. P. Arovas: Phys. Rev. Lett. 61 (1988) 617. 27) S. Liang, B. Doucot and P. W. Anderson: Phys. Rev. Lett. 61 (1988) 365. 28) J.D. Reger and A. P. Young: Phys. Rev. B37 (1988) 5978. 29) P.W. Anderson: Phys. Rev. 86 (1952) 694.

You might also like