CHINESE JOURNAL OF PHYSICS

VOL. 31, NO. 5

OCTOBER 1993

Holes in a Two-Dimensional Quantum Antiferromagnet Lu Yu

International Centre for Th;;Ttical Physics, Trieste, Italy Institute of Theoretical Physics, Chinese Academy of Sciences, Beijing 100080, China

Zhao-Bin Su
Institute of Theoretical Physics, Chinese Academy of Sciences, Beijing 100080, China

Yan-Min Li
Deparlment of Physics, University of Warwick, Coventry, CV4 7AL, United Kingdom

(Received July 14, 1993)

A brief review is presented on the studies of the hole motion in a two-dimensional quantum antiferromagnet. An extended introduction is given to cover the background of the problem. The quantum Bogoliubov-de Gennes formalism which treats the local distortion of the spin configuration and the quantum renormalization process on an equal footing, is outlined. The latest development on the central issue, whether a hole can freely propagate on an antiferromagnetic background, is overviewed.

I. INTRODUCTION The discovery of high temperature oxide superconductors [l] has stimulated great interest in the studies of the interplay between magnetism and superconductivity. It has been established now that the reference compounds like La2Cu04 in the case of lanthanum family or YBazCusOs for the yttrium family are antiferromagnetic (AF) insulators [a]. Upon doping the AF long range order fades away and superconductivity occurs below certain critical temperature. A generic phase diagram is shown in Fig. 1. Whether the superconductivity is caused by magnetic fluctuations is a separate issue. It is, however, crucial to understand the behaviour of charge carriers on an AF background. This is not a trivial problem, because according to the single electron theory, the reference compound has a half-filled valence band, and hence should be a metal [3], whereas it is indeed a Mott insulator. Unlike the simple band insulators, the energy gap for charge excitations in this type of insulators originates from strong electron correlations. In the
579

@ 1993 THE PHYSICAL SOCIETY OF THE REPUBLIC OF CHINA

580

HOLES IN A TWO-DIMENSIONAL QUANTUM ANTIFERROMAGNET

VOL. 31

Amorphous semiconductor

Strongly correlated

metal

Paramagnetk metal

I
3
8

I I I

I I
I

I-I
Superconductor

Doping level

FIG. 1. The generic phase diagram of oxide superconductors. early days of the high-T, heat wave Anderson suggested that a doped Mott insulator described by the one-band Hubbard model should contain the basic physics of the high-T, superconductors [4]. This model, independently proposed in 1963 by Gutzwiller, Hubbard and Kanamori [5], to describe the competition between delocalization and correlations in narrow-band transition metals has a simple looking:

H = -t C C,foCj~ + U ~C~~C~~C~~C~~~ <nn>,o i

(1)

where < nn > indicates summation over nearest neighbours, the indices i,j run over N lattice sites, CjO(Cib) are the electron creation (annihilation) operators of spin cr at site i, and the two-body interaction is on-site and repulsive (U > 0). There is a vast literature devoted to the study of this model in the last 30 years, especially after 1987 [6]. In the large U limit, it is more convenient to use a canonical transformation to project out the doubly occupied sites costing energy U and to obtain an equivalent, the so-called t - J model [7]: (2) where restriction to the Hilbert space without double occupancy is assumed and Izi = c nig (J is the density operator at site i with n;, = Croci,. The summation in the second term

VOL.

31

LU YU,ZHAO-BIN SU, ANDYAN-MIN LI

581

is carried out only over In deriving (2) from (1) substantially the physics, spin density operator Si

nonrepeated bonds < ij > and under this convention J = 4t2/U. a three-site term has been neglected which should not modify especially when the deviation from half filling S is small. The is defined as: Si = C!;Ci, where u are the Pauli m a t r i c e s .

At half-filling (one electron per site) the hopping term is absent, so the t - J model (2) becomes the well-known Heisenberg model for spin l/2 in the fermion representation up to a constant. Of course, the t - J model (2) can be studied independently, even in the limit t < J, but the correspondence with the Hubbard model is established only in the limit t >> J, or U > t. In this article we consider a slightly generalized t - J model with anisotropic exchange interactions, i.e., H = -t c (C&, + kc.) + J c [S,zS; + cx(S;SJ

t S;S;)] ,

<i,j>,a

<i,j>

(3)

where o = 0 corresponds to the Ising limit while (3~ = 1 is the Heisenberg case. It is now well-established that the essential ingredient in cuprate superconductors is the two-dimensional CuO;! plane [8]. The Cu-0 two band model, proposed by several authors [9] would be more appropriate for the description of physical properties of this plane. However, as shown by Zhang and Rice [lo], the low-energy physics is determined by the singlet state formed by the additional hole on oxygen with the existing hole on copper. The hopping of this singlet is described by an effective t - J model which makes this model even more pertinent to high-T, superconductors. Of course, there are aspects of the twoband model, like the formation of charge-transfer polarons [ll], which cannot be described by the one-band t - J model. The standard BCS theory of superconductivity is based on the assumption that the Landau Fermi liquid theory can adequately describe the normal state of superconductors [12]. According to the Fermi liquid theory [13] the elementary excitations of an interacting fermion system can be mapped onto a noninteracting Fermi gas. These quasi-particles have spin l/2, charge e and are well defined, which means that the real part of the self-energy (of the order temperature T or external frequency U) is much greater than the imaginary part (proportional to the square of the corresponding small quantities). The propagator of these quasi-particles has a nonvanishing residue 2 at the pole wk, i.e., the spectral density can be written as A(k,w) = zk6(W - wk) -I- incoherent part, (4)

which means existence of a well-defined Fermi surface where the momentum distribution has a finite jump equal t0 zkF. This was first observed by Migdal [14] and later elaborated by Luttinger [15] who also showed that the volume enclosed by this Fermi surface should be the same as for noninteracting systems.

---_--.

--~~~.

582

HOLES IN A TWO-DIMENSIONAL QUANTUM ANTIFERROMAGNET

VOL. 31

It is not clear at all whether charge carriers on an AF background will share all these properties, because the starting AF insulating state exhibits strong electron correlations. The central issue is whether a single hole can freely propagate on an AF background, i.e., whether 2 is different from zero or not. Of course, there is a finite concentration of charge carriers in real superconductors, but we will limit ourselves in this article to the single hole issue with the hope that the solution of this simple looking problem will shed some light on the underlying physics of high-T, superconductors. The reader is referred to the relevant literature [8] for issues of finite doping concentration. In this brief review we will summarize our studies on the single hole problem with a rather detailed introduction and an overview of the recent developments. The discussion will be developed around the central issue mentioned above. The rest of the paper is organized as follows. Sec. II is devoted to the background material, while various approaches to the single hole problem are discussed in Sec. III. The quantum Bogoliubov-de Gennes (BdeG) approach developed by us to treat the distortion of the spin background and quantum fluctuations on an equal footing, is summarized in Sec. IV. The latest developments and perspectives are discussed in the final section.
II. BACKGROUND II-l. The reference state

The ground state of the large U Hubbard model at half-filling or its equivalent - the Heisenberg model, depends strongly on the dimensionality. In one dimension (1D) an exact solution for the spin l/2 Heisenberg chain was obtained by Bethe [16] using the famous Bethe Ansatz (a term coined by C. N. Yang and C. P. Yang [17]). An exact solution for the Hubbard model at arbitrary filling was obtained by Lieb and Wu [lS], using the same ansatz. The ground state at half-filling is an insulating state for a.rbitrary U without AF long range order, while the spin-spin correlation function shows a power law decay. There are no exa.ct solutions beyond 1D. However, some rigorous results are available. In particular, Marshall showed in 1955 [19] that the ground state for S = l/2 systems on a bipartite lattice (2-sublattice system with all nearest neighbours of a given site belonging to the second sublattice) should be a singlet state. This is known as the Marshall sign rule. It has been proved recently by Kennedy, Lieb and Shastry [20] that there is AF long range order in three dimensions (3D) for S = l/2 system at T = 0, which vanishes at certain critical temperature. For two dimensions (2D) the AF long range order in the ground state has been proved for a square lattice in the case S 2 1 or S = l/2, but 0 < l/cx 5 0.13 and l/a > 1.78 [21]. For 2D S = l/2 1-I eisenberg model, relevant for the high-T, problem, neither existence, nor nonesistence of AF long range order has been proved. Nevertheless, there is strong evidence from esact diagonalization of small clusters and quantum Monte

-. _. _

VOL.31

LU YU,ZHAO-BIN SU, AND YAN-MIN LI

583

Carlo simulations that such order does exist for a 2D square lattice at T = 0 [22]. However, the staggered magnetization is reduced to m z 0.31 f 0.02 due to quantum fluctuations. We will take this as the starting point of our discussion. Of course, there cannot be any long range order at finite temperatures for the 2D Heisenberg model, as follows from the Mermin- Wagner theorem [23]. Historically, the question of the ground state for spin l/2 systems has been under active discussion. Since such antiferromagnets were not found for a long time, Landau and Pomeranchuk suspected that there might be no AF long range order and the elementary excitations might be spin l/2 fermions for S = l/2 AF systems [24]. Anderson [25] proposed in 1973, the so-called Resonant-Valence-Bond (RVB) state as a possible candidate for the ground state of such systems, most probably for a frustrated lattice. The RVB sta.te is a condensate of singlet pairs without AF long range order. In 1987 he reformulated his proposal in terms of a BCS-like variational wave function [4]. Since then a large amount of literature has been devoted to this issue [6,8]. The general consensus now is: There is an AF long range order in the ground state of 2D spin l/2 Heisenberg model, at least for a square lattice. However, the generalized RVB state with long range pairs and expressed in terms of spin variables instead of fermions to satisfy the Marshall sign rule [19], has a very close energy and similar correlation functions as the AF state [26]. 11-2. Spin wave t h e o r y It is well known that the classical and the quantum mechanical description for a ferromagnetic ground state is essentially the same because the order parameter - the spontaneous magnetization - commutes with the Hamiltonian. This is not true for the AF ground state, where the order parameter - the staggered magnetization - does not commute with the Hamiltonian, serving as a typical case of nontrivial symmetry breaking. The classical ground state would be a Neel state with alternating spin alignment, which is true for the Ising model, but not for the Heisenberg case. For the latter one should take into account the quantum fluctuations (or spin zero-point motion). The spin operators satisfy the following commutation relations

(5)
w h e r e cy,/3,7 = x,~,z,.?& is the unit antisymmetric tensor. If we introduce the spin raising and spin lowering operators Sl = S; + iSky, S, = S; - iSky, for S = l/2 they satisfy the commutation relations

[s:,s;]+ = 1,

(6)

584

HOLES IN A TWO-DIMENSIONAL QUANTUhl

ANTIFERROMAGNET

VOL. 31

[S,t,S;j = 0,k #!

(7)

i.e., they behave like fermions on the same site, and as bosons on different sites. These are hard-core boso+ or Pauli operators. We will come back to this special property of the spin l/2 operator at the end of this subsection, but now we consider the case of arbitrary S. Holstein and Primakoff [27] have proposed a nonlinear transformation to convert the spin operators into bosons S = S - b+b, S- = t/%b+(l - b+b/2S)2,

(8)

and one can then expand them in terms of btb/2S. This is the so-called linear spin wave theory, if only the leading term is considered. Apparently, the expansion parameter here is 1/2S which is 1 for S = l/2. However, the actual expansion parameter turns out to be l/Sz where z is the coordination number [22]. With this understanding, the application of the linear spin wave theory to the 2D S = l/2 Heisenberg model is more or less justified. Anderson and Kubo [28] have applied this theory to the AF case. Since its presentation is available in standard textbooks [29], we will outline here only the main results for further reference. Using the Holstein-Primakoff transforma.tion S+ = &ETb;, S, = &C?bJ, 5; = S - Lib;, i ET sublattice (9)

SJ = -S + bibj, j ~1. s u b l a t t i c e ,

the anisotropic Heisenberg Hamiltonian becomes HJ = JS C (2blbi + a(b;bj + blbi)) - 7 <i,j> which, in turn, can be diagonalized using the Bogoliubov transformation
bk = UkPk + Vk,&k

(10)

with ]?L:] - ]?Jkl2 = 1

(11)

to obtain

where
wk =

JSz( 1 - a2$)lj2,

Yk = i $ ,ik.6

(13)

with b as the nearest neighbour lattice vector. The spin-wave vacuum 10 > can be expressed in terms of the classical N6el vacuum IN > as

-_

.,

VOL.31 IO >= with

Ak = -vk/uk,

LU YU,ZHAO-BIN SU, AND YAN-MIN LI

585

nu;l
k

e x p (-X,b:bt,) IN >

(14)

Uk = {;

[(l - 2y;)-r2

+ l]}li2)
112
- I]} .

(15)

vk = -s&+/k)

(1 - y2y;p2

The spin-wave spectrum for the Heisenberg case o = 1 is linear in L

.dk =

v]k]

(16)

with the spin-wave velocity v = 2&JS. This is the Goldstone mode originating from the breaking of O(3) continuous symmetry. One can easily calculate the reduction of the order parameter due to the quantum fluctuations < S >= S- < btbi > and the correction is approximately 0.1 for a 3D cubic lattice, 0.2 for a 2D square lattice and diverges for 1D. There is one serious difficulty with the spin wave theory, namely, the hard-core nature of the spin operator, or equivalently, the single occupancy constraint is not well preserved. In fact, one can calculate the expectation value [30] F e I < [S,S;]+ > -11 = 21 < [bfbi - (bib;)21 > I M 1 . 0 6 as temperature T --+ 0, which should vanish if the constraint is strictly obeyed. There have been many attempts to improve the linear spin wave theory. Dyson and Maleev [31] have introduced another transformation of spin operators into boson operators to consider the kinematic constraints. Arovas and Auerbach have used the Schwinger boson transformation to represent the spin operators and the path integral technique to impose the constraint [32]. Takahashi [33] h as applied the Dyson-Maleev transformation to quantum AF to impose the kinematic constraints. His results turned out to be essentially the same as those of the Schwinger boson approach. As a common weakness of these treatments, the single occupancy constraint is imposed only on average. An interesting attempt was made by Wang [34]w h o used the Jordan-Wigner transformation generalized by several authors [35] to 2D to represent the hard-core boson in terms of spinless fermions. In this representation the hard-core constraint is satisfied exactly and very good results can be

586

HOLES IN A TWO-DIMENSIONAL QUANTUM ANTIFERROMAGNET

VOL. 31

obtained even in the mean field approximation. Recently, this transformation has been extended to the case away from half-filling and a number of interesting results have been obtained [36]. 11-3. Nagoaka theorem and ferromagnetic polaron The phase diagram of the Hubbard model or the t - J model upon doping is a complicated problem. However, there is one rigorous theorem proven by Nagaoka [37] which can be stated as follows: For a bipartite lattice with one hole in the infinite U limit, the ferromagnetic state with maximum spin is the ground state. Instead of presenting its proof we give here some simple argument in favour of it. The main difficulty of the Hubbard model in the infinite U limit is the enormous spin degeneracy at exact half-filling. The presence of a single hole can lift the degeneracy. Intuitively, it is clear that the hole can gain maximum kinetic energy if all spins are aligned ferromagnetically. This maximum energy is zt with z as the coordination number. Using a set of states with spins permuted by single hole hopping around a closed loop, Nagaoka could prove this statement rigorously for a bipartite lattice where all nearest neighbours of a given site belong to the second sublattice and the sign of t is irrelevant. This theorem has nontrivial consequences. First of all it is not true for 1D where no closed loops are available. Secondly, a generalization to finite hole concentration in the thermodynamic limit seems to be natural, but has confounded all analytical attempts. The Pauli exclusion principle is irrelevant for a single hole, but it makes the problem messy even for two holes. The result depends strongly on the boundary conditions. For the periodic boundary condition a spiral phase is preferred, whereas the ferromagnetic state survives for a veraged boundary conditions [38]. Numerically, there is strong evidence that a ferromagnetic state with non-saturated magnetization is the ground state at small, but finite doping concentration as U -+ 00 [39]. In real systems U is finite and J is different from zero. A more pertinent question would be: How big is the ferromagnetic region around a hole as a result of balancing between the kinetic energy t favouring the ferromagnetic configuration and the exchange energy J favouring the AF alignment of neighbouring spins. A rough estimate for the 2D Ising square lattice gives the hole energy as [40] &h/t = -4 t 6.03(J/t)2,

(17)

where the radius of the f erromagnetic disc diverges as (J/t)-/. For a small cluster (4 x 4), the ferromagnetic state with maximum spin is materialized only for J/t < 0.075 [41]. We will call such a formation ferromagnetic polaron. In view of the Nagaoka theorem, the problem of single hole motion on an A F background is irrelevant in the limit U = 00, or J = 0. However, apart from very small J/t,

VOL.31

LU YU,ZHAO-BIN SU, AND YAN-MIN LI

587

when the energy of ferromagnetic polaron (17) is lower than the competing configuration of spin polaron (to be considered later), the single hole problem is an important issue and will be dealt with in the rest of this article. 11-4. One-dimensional case Although the exact solution of the 1D Hubbard model was obtained by Lieb and Wu 25 years ago [18], much of the rich physics contained in this solution has been appreciated only during the recent intensive studies on the high-T, problem. These results have no direct relevance for the single hole issue in 2D, but they are very instructive and inspiring by providing a profound example with a number of peculiar properties, in many ways contrary to our intuition. Therefore, we outline here some of these properties, which were emphasized by Anderson [42]. One of the distinguished features is the absence of quasi-particles in the standard sense, Using the Bethe Ansatz solution [18], one can calculate the wave function renormalization 2, defined as square of the overlap integral between the N-particle ground state wave function plus one free particle in state k with the N + l-particle ground state wave function, i.e. [42]

where the phase shift 6 = 4 in the large U limit. In the thermodynamic limit, N -+ co,Zk -+ 0, hence the quasi-particle does not exist as such. This is another example of the infrared catastrophe, suggested by Anderson earlier [43]. The main physical origin for a vanishing 2 is the spin-charge separation which can be explicitly seen in the asymptotic form of the Bethe Ansatz wavefunction in the large U limit, derived by Ogata and Shiba [44]
+h,...

ZN) = (-1P det[exP(ilcj~Q,)I~(yl,. . . YM),

(19)

where 58, are sites of spinless fermions (holons), while ?JM are the down-spin sites of a squeezed Heisenberg chain with all hole sites being removed. When a hole is added at kF, it decays immediately into a holon (charge e without spin at 2kF) and a spinon (spin l/2 and no charge). The vanishing of 2 reflects the absence of charge for excitations at kF. Moreover, one can calculate various correlation functions and the single-particle Greens function [44,45,42]. All of them show a power law decay with exponents depending on the interaction. It is remarkable that in spite of the vanishing of 2, hence the jump of the momentum distribution at the Fermi surface, there is still a singularity of the latter at kF (not 2kF), i.e.,
nk = - constlk - kFlasgn(k - kF),

588

HOLES IN A TWO-DIMENSIONAL QUANTUM ANTIFERROMAGNET

VOL. 31

where the exponent a = l/8 for U + oo. It turns out that the above described properties (2 = 0, spin-charge separation, power decay of the correlation functions, etc.) are rather generic for 1D interacting systems. Haldane [4G] first realized the existence of this large universality class and has coined a term Luttinger Liquids to contrast them with the Landau theory of Fermi liquids. Anderson suggested [42] that the 2D Hubbard model should share these properties and the experimental data on high-T, superconductors, especially the normal state properties seem to be consistent with this assumption. Of course, the question is still widely open. Clearly, the single hole problem cannot replace the issue of the charge carrier behaviour at finite doping concentration, but it is a pertinent question to consider in order to clarify the situation. III. VARIOUS APPROACHES FOR STUDYING THE SINGLE HOLE PROBLEM The question of charge carrier motion on an AF background and the possible formation of self-trapped states have been discussed in the literature long time back in connection with magnetic semiconductors [47,48]. In particular, de Gennes [48] has considered the formation of a small ferromagnetic region around the charge carrier which can propagate in an AF under certain conditions. There were extensive studies on this issue in Russia [49,50], summarized later in a review article and a book [51]. Unfortunately, these research works are not well known in the Western literature. We will, therefore, mainly discuss those studies which have shaped our current understanding of the problem. III-l. Strong coupled Ising case (the Brinkman-Rice limit) Consider first the Ising case. The hopping of a hole on a Neel background would create a string of wrongly aligned spins (see Fig. 2) costing energy of the order eJ, where e is the length of this string. To avoid this energy expense, Brinkman and Rice [52] have introduced the so-called retraceable path approximation under which the hole hops back exactly the same way as it left the starting point, so there is no difference in the initial and final spin configurations. In the atomic limit of the Hubbard model (U -+ 00 or J = 0), they have summed up all these retraceable paths (without closed loops) and have come to the following conclusions: (1) The hole cannot propagate freely, i.e., 2 = 0, and the Greens function is kindependent, i.e.

G(w) = u-l

{l__& [;- (i- (y)q12]}-1

(21)

VOL.31

LU YU,ZHAO-BIN SU, AND YAN-MIN LI

589

(al

T1T.LT.L 1

jl-

r- i-;, T
(b)

T -l-?-i-T .L

FIG. 2. A mismatched spin string is created when a hole moves from one site (a) to another site (b). with z as the coordination number and t as the hopping integral. (2) The density of states p(w) has a square-root singularity with a 25% narrowing compared with the original bare bandwidth, namely, Z@(U) = ; [(S - 9U2)/2/(1 - u)] )
u = W/A,

(3) The electric conduction is due to diffusive motion of holes and the mobility can be calculated. In spite of the great success of the Brinkman-Rice approach, there are two serious limitations. (i) They have ignored the closed loops which turned out to be very important. As noted by Trugman [56], a hole can travel around a square one and half times without disturbing the spin background and finds itself translated to the next-nearest-neighbour (see Fig. 3). This means that the hole can unwind the string and self-generate a nextnearest-neighbour hopping. Therefore, the complete localization of charge carriers in the Brinkman-Rice limit is an artifact of the appro,ximation. (ii) Their consideration for the spin dynamics is purely classical, i.e. the emission and reabsorption of spin excitations are not allowed. As we will see later, the renormalization effect of the hole motion due to these processes is essential even for the Ising case. If J/t is finite, but small, the string picture holds, and one can calculate the bound states for the hole in this confining potential which leads to an Airy equation [.50,53]. The

500

HOLES IN A TWO-DIMENSIONAL QUANTUM ANTIFERROMAGNET

VOL. 31

FIG. 3. Tht Trugman loop (see Ref. 56). semi-classical ground state energy in 2D is then given by E,/t = -2fi + 2.74(J/t)23,

(23)

where the band narrowing effect (2& instead of 4) of Brinkman and Rice has been taken into account. The energy difference between higher levels is also w (J/t)*j3. This string picture has been explored further by many authors to interpret various physical properties [54]. Vollhardt et al. [55] h ave found that the retraceable path approximation is exact in infinite dimensions and they have generated a systematic l/d expansion. 111-2. Quasi-classical approach (Nonlinear a-model) In Sec. II-2 we outlined the traditional spin wave theory for AF. An alternative way of describing the long-wavelength spin fluctuations for large S is to use the nonlinear a-model, assuming the existence of AF long range order. The order parameter-staggered magnetization R is a vector on a unit sphere, whereas the magnetization m is a small quantity. In the long wa.velength limit (a.t the scale la.rger tl1a.n the la.ttice constant) the fluctuations a.re described by the nonlinear g-model [57]: fIhLo = -

ai J d2r

[X-m 2 + p(V .

fl)] ,

where S is the magnetic susceptibility, while p is the spin stiffness. Their ratio p/X = c2 is the square of the spin-wave velocity. n and m satisfy the commutation relations [17P(.r), ?-r&J)] = i&%??(a:)G(X - y), [?72~(X),fP(?j)] = iEflT W(X)6(2 - ?j) (25)

VOL. 31

LU YU, ZHAO-BIN SU, AND YAN-MIN LI

591

and the constraints

f-l2 = 1,

m(r). R(r) = 0.

(26)

The spin current is defined as J, = R x

a,n,

(27)

satisfying the equation of motion dm/dt = -p&J,. This current is a vector in both spin and coordinate space. Shraiman and Siggira [583 have considered the coupling of the charge carrier to the spin configuration in the quasi-classical approximation t < J. The spin background is assumed to be classical but the distortion of this background due to the hole motion has been taken into account. In analogy with the backflow in helium they considered the dipole moment generated by the charge carrier

where $k is the spinor wave function of the hole, u the Pauli matrices. The k-summation is carried over the minimum energy states. (Cluster calculations, as discussed in Sec. 111-4, show they are at (&5,&t)). The coupling to the spin current is written as
-9 ~Pab-l). J&l).

atq

(29)

They found that in the continuum limit (30) which gives rise to a dipolar spin configuration (Fig. 4). Based on this result, they proposed that the commensurate AF phase is unstable with respect to an incommensurate spiral phase upon doping. The pitch of the spiral should be inversely proportional to the doping concentration [58]. Most of their resuits were derived under the assumption J >> t, but they have also used the Brinkman-Rice string picture to argue that a similar conclusion should also hold in the opposite limit J << t, since the symmetry properties are the major factor. This approach also has the limitation that the quantum fluctuations of spins have not been included, as in the Brinkman-Rice method. As we will see in the next subsection, inclusion of these fluctuations will modify some conclusions in a substantial way. Also, the spin distortion discussed by them is related to areas far away from the hole location, i.e. it

-._-

592

HOLES IN A TWO-DIMENSIONAL QUANTUM ANTIFERROMAGNET

VOL. 31

FIG. 4. Schematic diagram of the dipolar spin configurations (see Ref. 58).

is a long distance behavior. As shown in the next section, the near-field distortion is quite different. This approach has also been used by several other groups to consider the single hole problem [59] with similar results. In particular, a self-trapped polaron with local ferromagnetic distortion is found when the hole state is at the Brillouin zone center. The structure of polaron is much more complicated if the hole state is at the zone boundary. A semiclassical large S expansion has also been employed by Auerbach and Larson to study the single hole problem [60]. Using the spin-hole coherent-state path integral, they have generated a large S expansion of the t - J model on a square lattice. The singlehole classical energy is minimized by a small polaron solution. In the parameter range 1 < tI2J.S < 4.1, the polaron involves only five sites with one flipped spin at the center. The hole density is l/2 and l/8 at the center and the nearest neighbour, respectively. Contrary to Shraiman and Siggia, they have not found any long-distance distortion of the AF background. They have also found that the dynamical nearest neighbour hopping is suppressed and have recovered an effective polaron model with next-nearest-neighbour hopping only [61]. X11-3. Self-consistent Born approximation There are two effects which have not been properly considered in the above described approaches, namely, the quantum fluctuations of the Heisenberg spin and the renormalization of hole motion due to interactions with spin waves. These effects have been taken into account in the self-consistent Born approximation carried out by Schmitt-Rink, Varma and Ruckenstein (SVR, [62]) and IX ane, Lee and Read (KLR, [63]). The hopping of holes leads to disturbance of the spin background. Suppose there is a hole on spin-up sublattice site. After hopping to spin-down sublattice site, a down-spin

-_. --..--

VOL.

31

LU YU,ZHAO-BIN SU, AND YAN-XlIN LI

593

appears on a spin-up lattice site, corresponding to a local spin escitation. Likewise, a hole on spin-down sublattice would request the neighbouring up-spin to flip first in order to avoid frustration. Introducing hole operator hi to replace C,t, on the spin-up sublattice, while representing the corresponding operator Ci,--hiA1 as a composite operator (similarly for the spin-down sublattice), and then performing the Holstein-Primakoff transformation, one obtains the hopping Hamiltonian [62,63] Ht = C hfhj(bj + bi) <ii> (31)

in the linear spin-wave approximation. A well-argumented derivation of (31) has been given in Ref. 64. The form of (31) is similar to the electron-phonon interaction. It is natural then to compare the spin-polaron under discussion with the well-studied lattice polaron [65]. However, there are important differences between these two, as we will discuss in Sec. IV-2. Comparing the Heisenberg Hamiltonian (10) and the hopping term (31), one finds two different limits. When J >> t, one can use the perturbation theory, but this is a less interesting case. The alternative limit J << t, corresponding to the large U Hubbard model, cannot be treated perturbatively. One has to perform at least a partial resummation to obtain a meaningful result. This was done by SVR [62] and KLR [63]. They have ignored the Hartree diagram, but have taken into account the Fock term in the self-energy (Fig. 5). Neglecting the crossing diagrams (Fig. S), they can solve the corresponding Dyson equation. The self-energy part can be written as

.....

0
(a)

.......

.... 0

FIG. 5. Self-energy diagrams for a hole (a) Hartree (h) Fock terms and (c) the polxization diagram for the spin propagator.

HOLES IN A TWO-DIMENSIONAL QUANTUM ANTIFERROMAGNET

VOL. 31

FIG. 6. An example of crossing diagram.

C(k,w) = y &ir~+~ + sy#G(k - q,w - wq),

k

(32)

where the renormalized hole propagator

G(k,o) = (w - C(k,w) + ib)-,

.

(33)

while the definitions for Uk, ?Jk? Wk, Yk were given in (13)-( 15). One can then easily calculate the spectral function A(k,w) = -iImG(k,yl). A detailed numerical study of Eqs. (32)-(33) h as been carried out in Refs. 64, 66 and 67 with the following main conclusions: There is a well-defined quasi-particle propagating on an AF background. The spectral function shows several coherent peaks plus some incoherent states at higher energies. The separation between peaks as well as the band width scales as (J/t)2/3 for both Ising and Heisenberg cases. This has been checked on clusters of different size (4 x 4:8 x 8,16 x 16, etc.) to make sure that the above features are not due to finite size effects. The results for small J/t are in good agreement with those of exact dia.gonalization on small clusters. A similar approach using l/.z expansion with z as the coordination number has also been developed [68] with comparable results. There were several approximations involved in this approach: linear spin wave expansion and the neglect of crossing dia,grams. The former does not guarantee the hard-core boson nature of the spin excitations, which, in turn, may lead to violation of the single occupancy constraint or appearance of extra degrees of freedom. As discussed in Ref. 64, this will affect the short distance behaviour (as we will see later in Sec. IV), but may not change the properties for small k near the ba.nd bottom, especially for the Heisenberg case. The single occupa.ncy constraint has been implemented in the va.ria.tional calculation of Sachdev [69]. As for crossing diagrams, it has been shown in Ref. 67 tha.t vertex corrections due to two-loop diagrams vranish, while the effects of three-loop dia.grams are small as found from numerical calculations. On the other hand, the distortion of the spin background in the

VOL.31

LU YU,ZHAO-BIN SU,AND YAN-MIN LI

505

presence of the hole, neglected in this approach, may have important consequences, as we will see later, in Sec. IV. 111-4. Numerical approaches Apart from the analytic approaches discussed above, there have been extensive numerical studies on the single hole problem, using mainly two types of techniques: exact diagonalization of small clusters and quantum Monte Carlo simulations. The exact diagonalization using the Lanczos-type algorithms has been carried out by different groups for clusters of 8,10,16,18 sites [41,70-761. Recently, the number of sites has been increased up to 26 [77], for which the maximum power of the current generation of supercomputers has been fully exploited. (The dimension of the Hilbert space is of the order of lo7 and the calculation requires at least G G bites disc space.) The advantage of the exact diagonalization is the absence of the fermion minus sign problem inherent to the quantum Monte Carlo technique. On the other hand, the limitation to small clusters may give rise to artifacts due to finite sizes. Now the results seem to converge: There are well defined quasi-particles with energy minima at (A$,&;) on the boundary of the reduced Brillouin zone. The effective mass is strongly anisotropic: small along the zone boundary and big in the perpendicular direction. Apart from the ground state there are some excited states for certain momenta as seen from the spectral function. The higher excited states have intrinsic width due to the probability of decaying into lower states. We will discuss later how to interpret these levels. A finite size scaling study has been performed for clusters up to 26 sizes [77] (see also [78]), and the wave function renormalization factor 2 seems to remain finite in the thermodynamic limit. The scaling properties of excited states are less conclusive, especially near the zone boundary. Another related approach is to use a truncated Hilbert space [56] which allows one to study bigger clusters using less computing power. This technique has also been used in Refs. 79, 80 to study the hole bands and the spin distortions around the hole. In general, the results obtained using this technique are consistent with those of exact diagonalization and the self-consistent Born approximation, considered in the previous section. The quantum Monte Carlo technique has not been used extensively for studying the single hole issue mainly because of the fermion sign problem. Sorella [79] has, however, systematically studied this issue using the so-called projection Monte Carlo technique [S2]. He has proved a rigorous upper bond 2 5 l/2 for the t - J model and has generalized the Migdal theorem relating 2 to the jump of the momentum distribution at the Fermi surface An, = 2 for the 2D t - J model. Using this relation and choosing appropriate trial wave functions, he succeeded to minimize the sign problem and to carry out a finite size scaling

596

HOLES IN A TWO-DIMENSIONAL QUANTUM ANTIFERROMAGNET

VOL. 31

study for 2. It seems that 2 vanishes in the thermodynamics limit, to the contrary of the results from a similar study using the exact diagonalization, just described above. Whether this discrepancy is due to limitations of one of these techniques, is an open question for further studies. The Greens function Monte Carlo technique has also been used to study the single hole problem on a 4 x 4 cluster [83]. The energy seems to agree well with the exact result. Another alternative is to use the Gutzwiller-projected trial wave function to perform the variational Monte Carlo calculation [84]. The location of the energy minimum and the band width thus obtained is consistent with the known results. Before concluding this overview section, we would like to briefly mention the alternative itinerant approach. Starting from the Hubbard model and using the Hartree-Fock approximation, one obta.ins the spin-density-wave (SDW) ground state for half-filling. The one loop correction (random phase approximation) then gives rise to the collective excitations (spin waves). This approach has been used by Schrieffer, Wen and Zhang [85] to propose the spin bag model for high-?, superconductivity. These spin-bags are nothing but spin polarons we are considering in this paper. The spin wave velocity they have obtained is the same as what follows from the strong-coupling t - J model. This approach has been pursued by many authors [86]. The question on where the energy minimum is located, at (&S,&$) or (fr, 0), (0, *7r), depends on which spin fluctuations (longitudinal or transverse) are considered. It is interesting to point out that the hole-hole correlations calculated in this approach (see paper of Frenkel and Hanke in Ref. 86) are the same as follows from Shraiman-Siggias approach [58]. IV. QUANTUM BOGOLIUBOV-DE GENNES FORMALISM FOR THE S P I N POLARON So far we have discussed the background of the single hole problem and have overviewed various approaches in studying this problem. In this section we outline the quantum BdeG formalism developed by us for this purpose, summarize the main results obtained and discuss the open issues. IV-l. Questions to be answered Basically, there are two sets of questions to be answered: (1) How will the holes affect the spin background? What is the local and long-distance distortion around the hole? How would the spin excitations and spin correlations be modified in the presence of a dynamical hole? (2) What are the properties of the holes (energy spectrum, effective mass, etc.) which a.re embedded in this background? In view of the 6 _ 0 limit, one might think of foregoing the first set of questions, assuming the ground state of a. half-filled system. However, this is

-._

VOL.31

LU YU,ZHAO-BIN SU,AND YAN-MIN LI

597

not completely true, because, as we will see later, the hole states and the distortion of the background should be considered in a self-consistent way, which has not been materialized in analytical formalisms reviewed in the previous section. In fact, there are three major effects to be considered for the hole motion, namely, (1) the quantum fluctuations of the spin system; (2) the renormalization effect due to virtual process of emitting and reabsorbing the spin excitations; (3) the local distortion of the spin background. Factors (1) and (2) h ave been considered in the self-consistent Born approximation (Sec. III-3), but not factor (3). The quantum BdeG formalism we proposed [87] treats all these three factors on an equal footing and is aimed to answer both sets of questions. Of course, the treatment is not exact and we will discuss later its limitations and open problems. IV-2. Lattice relaxation approach and its generalization We have mentioned in Sec. III-3 the analogy with the lattice polarons. It is interesting to compare in more detail the properties of these two kinds of polarons. For the convenience of reference we write down here explicitly the model Hamiltonian for small lattice polarons [88]:

H = He + Hph t He+,
He = -to
He.+, =

C
<i,j>
1

CtCj,

Hph =

~~'-~q(@'q t ;,,
9

(35)

~c/c~ c
4

M,eiq"(bq t bt,),

where M, is the electron-phonon coupling, wq is the longitudinal optical phonon frequency, while the rest of notation is standard. In the anti-adiabatic limit fLwq >> to, one can use the Lang-Firsov transformation [89]

0=

c hwql,eiq?C,(6, - Q, 12 q,l

(36)

to decouple the electron and phonon degrees of freedom to obtain an effective Hamiltonian for the electron dressed by the phonon cloud. Comparing (35) with our t - J Hamiltonian (10) and (31), the exchange integral J is, to some extent, equivalent to the phonon energy tw, in the lattice polaron problem, and a large J means a stiff spin background. However, there is no bare hopping term to in the spin polaron case. As discussed in Sec. III-S, the hole hopping on AF background will certainly lead to frustration of the spin background. Therefore one cannot use the

VOL.31

LU YU,ZHAO-BIN SU, AND YAN-MIN LI

599

IV-3. Basic equations For the convenience of reference, we rewrite our working Hamiltonian as H = H.r+Ht
HJ = J C Iij b!bi $ G(b!bJ + bibj) Pip,, i,.i

(38)
(39)

Ht = t

C Iijhjhi(bi
i,j

+ bj),

where P; = 1 - hfhi, Iij = 1 for the nearest neighbours and zero otherwise. The presence of the projection operator Pi is essential for the calculation of the spin properties. Let us start from the case when t > J. As mentioned earlier, in this case we can use the Born-Oppenheim approximation, i.e., decompose the combined n-th state of the hole-spin system as a direct product

where In, & >h, is the n-th eigenstate of a hole centered at Re, whose local representation is given by pn(i - R,e) to be determined later, while Ifi, R,c >S, is the corresponding spin part. It is important to note that the spin configurations are different for different n. For the moment we fix Re at the origin. The hole states ]n,Rc >h and the corresponding spin distortions are determined by BdeG type equations. Unlike the case of quasi-1D systems, we have to go beyond the Hartree approximation. The generalization of this type of equations including the Fock and RPA corrections, has been worked out earlier [94], using the closed time-path Greens function formalism. We refer the reader to the review article [94] for technical details. The basic tool is to use the spectral representation and the Dyson equation for the Greens function. In particular, the retarded Greens function for the hole is defined as

Gr(j,; t - t> = -i~(t - t) < {hj(t),hl(t)} >,

with its spectral representation

(41)

(42)
where (P,,, and E, are m-th eigenstate and eigenvalue of the hole, respectively. In the Hartree-Fock approximation the Dyson equation can be written as G&j+) = G!o( i 7 j-w)
7

~G!o'(i,l;w)[C~([,I';w) t C~(I,I';w)]G,(I',j;w),
1,l

(43)

600

HOLES IN A TWO-DIMENSIONAL QUANTUM ANTIFERROMAGNET

VOL. 31

where

(45)

are the Hartree (Fig. 5a) and the Fock (Fig. 5b) self-energy parts, respectively, with
Pi = t

C N(Em)~m(l)~~(j)~lj[g,(i,I;w
j,l,m

0) + gi(i,j;W = 0)]

(46)

as spin distortion and Fl,lt(uq, us) =

IL: + ~~o~(e-~~+-j) + e iq.(l-i)

> + uT!eiq~(l-l- i+j),

(47)

Substituting (42) into (43) and remembering that G$(j,k;w) = --&, (48)

multiplying both sides of (43) by w + iv and let o + E,, we obtain the following BdeG equation

En~n(i) = c[C:(i,j; En) t C%,.i; En)]vn(j>

(49)

to determine the energy and wave function of the hole states. In Eq. (46), g,(j, C; w) and g:(j,k; w) are the Fourier transformations of the retarded spin Greens functions, defined as
gr(j, k; t -

t) = -iB(t - t) < [bj(i), Q(C)] >,

(5Oa) (5Ob)

g:(j, I;; t - t) = -iO(t - t) < [bj(t), bk(t')] >,

while Uk, Vk are given by (15) and N(E,) is the hole state occupation (0 or 1 at zero temperature). The spin distortion pi can be determined via a canonical transformation

VOL.31 leading to ,?,/bJ,/ = b.I - p.1, I

LU YU,ZHAO-BIN SU, AND YAN-MIN LI

601

Ubt,?,/t = b?* - p . t t

(52)

We can define a distorted Neel vacuum Ifi >= UJN > and new spin operators &; = UbiUt so that ,0; =< Nlbilfi >= - < NjLiIIV > .
.

(53)

(54)

(55)

The true ground state with quantum fluctuations can then be represented as > (56)

with &I, as Fourier transformation of 6; and Xk = -~k/uk. We note that the unitary transformation 21 and the Bogoliubov transformation (11) commute with each other, so all the above operations are well defined. Before proceeding further, a few comments regarding the BdeG equation (49) are in order. (i) The Born-Oppenheim approximation (40) used here is different from the standard static case, because the potential in the BdeG equation (49) is energy dependent, including some renormalization effects due to virtual spin excitations. Therefore, the retardation effects of spin-hole interactions have been incorporated to some extent. In fact, this is the reason why we can get qualitatively correct results even in the range oft comparable with J, when the usual adiabatic approximation would breakdown. (ii) Strictly speaking, Eq. (49) is not an eigenvalue problem due to the energy dependence of CF. Nevertheless, we can find self-consistent solutions. They are dynamical generalizations of states in the confining string potential discussed in Sec. III-l. (iii) The Brinkman-Rice results can be easily reproduced in the limit J/t - 0 within our formalism. In fact, in this limit pn(i) becomes an extended state, so the retarded Greens function turns out to be G,(w)Si,. In that case, the Hartree contribution to the self-energy vanishes, while the Fock part becomes bij Zt2G,(w - ZJ). One ca.n then find G?(W) and the spectral function in the Brinkman-Rice limit.

602

HOLES IN A TWO-DIMENSIONAL QUANTUM ANTIFERROMAGNET

VOL. 31

IV-4. Hole states with a fixed origin

The BdeG equation (49) has been solved by us numerically, first for 0.6 > (Y 2 0 [87], then for arbitrary 1 > o > 0 [95]. M oreover, the Schwinger boson formalism has been applied for the single hole problem in the Heisenberg limit o = 1 [96]. The results of all these calculations confirm each other and we summarize here some main features, which are compared with those obtained in other approaches. There is a series of self-trapped states emerging a.s solut.ions of (49). They are similar to bound states of a hydrogen atom. The lowest state is of s-symmetry, while the states immediately above it are doubly degenerate, of p, and p, symmetry. The higher states cl,, and dz2-y 2 are not degenerate, but the energy difference is small. The ground state wave function is rather localized in the Ising case cy = 0, and becomes more extended as cr increases (Fig. 7). The extent of the wave function is not sensitive to J/t within the range (0.2-2). The ground state energy Eu and the sepa.ration between this level and the second
0.000 0.255 0.000

0.255 0.860 0.255 0.000 0.255 0.000

(4
0.000 0.000 -0.001 0.000 -0.001 0.000 0.000 0.000 0.000 -0.007 -0.025 -0.007 0.000 0.000 -0.001 -0.007 -0.081 0.165 -0.081 -0.007 -0.001 0.000 -0.025 0.165 0.928 0.165 -0.025 0.000 -0.001 0.000 -0.007 0.000 -0.081 -0.007 0.165 -0.025 -0.081 -0.007 -0.007 0.000 -0.001 0.000 0.000 0.000 -0.001 0.000 -0.001 0.000 0.000

(b)
-0.001 0.000 0.000 0.007 0.003 -0.002 0.000 -0.007 -0.001 -0.002 0.000 -0.007 0.003 -0.002 0.000 0.007 -0.001 0.000 0.003 -0.002 -0.010 -0.002 -0.035 -0.002 -O.ClO -0.002 0.003
0.000 -0.001

-0.007 -0.002 -0.056 0.122 -0.056 -0.002 -0.007 0.000

-0.002 -0.035 0.122 0.960 0.122 -0.035 -0.002 -0.001

0.000 -0.007 -0.002 -0.056 0.122 -0.056 -0.002 -0.007 0.000

0.003 -0.002 -0.010 -0.002 -0.035 -0.002 -0.010 -0.002 0.003

0.000 -0.001 0.007 0.000 -0.002 0.003 -0.007 0.000 -0.002 -0.001 -0.007 0.000 -0.002 0.003 0.007 0.000 0.000 -0.001

Cc)
FIG. 7. Hole ground state wave functions for (a) (Y = 0, (b) a = 0.5 and (c) a = 0.9 at J/t = 0.1.

--___

VOL.31

LU YU,ZHAO-BIN SU, AND YAN-MIN LI

603

level Er as a function of J/t is plotted in Fig. 8 for the Ising case. Upon recovering the translational invariance these states will become propagating states to form bands, as we will see in the next section, but let us first understand the origin of coherent propagation even for the Ising case. Starting from the hopping Hamiltonian (38), one can derive an equation of motion in the Hartree approximation

Enhi = f$ = t CIij[< 6; > + < 6: >]hj* i I

(57)

Since the hole wave function is not confined to one site, the spin distortion < 6; > will not vanish on the neighbouring sites. So the coherent motion is due to a self-generated hopping, which is more efficient than the Trugman cycle discussed in Sec. III-l. This effect can be clearly seen by comparing Figs. 7a and Fig. lla and is one of the major consequences coming from the explicit account of spin distortion (the Hartree term, neglected in the self-consistent Born approximation). On the other hand, the Hartree term alone will not do the job. As seen from Fig. 8, the ground state energy goes like -t*/J N -U, which is a wrong energy scale. The inclusion of the Fock term reflects higher order virtual fluctuations which bring the energy scale back to t, and eventually to J.

0.0

Eo

A 0.4

0.0
1

2 ln(l+ J/t)

FIG. 8. The ground state energy Eo and its separation A from the second level as functions of
ln(1 + J/1) in units of 1 for the Ising case. They are extrapolated to the Brinkman-Rice limit for very small J/2. The dot-dashed curve is the ground state energy Eo with Hartree term only.

604

HOLES IN A TWO-DIMENSIONAL QUANTUM ANTIFERROMAGNET

VOL. 31

The presence of a set of self-trapped states is in agreement with the results of exact diagonalization on small clusters [41,70-781 and the self-consistent Born approximation [6264,66,67]. A detailed comparison of the J/t dependence requires some further studies. However, our formalism provides explicit information about the spin distortion. In fact, the expectation ,& (see Eq. (55)) is a real number in the ground state, so < S, > is finite, while < S, >= 0. This means that the spin configuration in the ground state is a canted state, rather than a spiral state (see Fig. 9), as suggested by Shraiman and Siggia [58] and further elaborated by other groups [97]. We believe that, as far as the near-by spin distortion is concerned, our result is correct. In fact, it agrees with an early consideration of de Gennes [48] using the s - d model for discussing the spin polaron problem. The canted state corresponds to a ferroma.gnetic component on an AF background. Therefore, to consider the spin polaron as a small region with fully polarized spins is an oversimplification. What happens in reality is the appearance of a small ferromagnetic component favouring the kinetic energy. A recent path integral treatment of small polarons by Auerbach and Larson [60] confirmed this result. They found a small ferroma.gnetic region of 5 sites with probability of hole occupancy very close to ours, as mentioned in Sec. 111-2. A dipolar distortion \vas found in the quasi-classical treatment [58] as a long distance behaviour. Strictly speaking, the spiral state is irrelevant for the single hole problem, since the pitch N l/6 would be infinity in this case. Whether the spiral state is a ground state for some finite doping, and how would the long distance distortion match with the short distance behavior, are all open questions. It is, however, worthwhile to mention that the dipolar distortion was not found in the path-integral calcula.tion [60] and the spiral phase is unstable due to strong local distortions not included in the mea.n field theory [98]. IV-5. Hole states as continuum bands So far we have discussed only hole states with a fixed origin. One can easily form Bloc11 states to recover the translational invariance, like small polarons in the lattice case. The Bloc11 state of a propagating hole can be written as

FIG. 9. Schematic diagrams for (a) canted and (h) spiral states.

VOL. 31

LU YU, ZHAO-BIN SU, AND YAN-hlIN LI

605

where ]$u(R,)> is the hole state in the local representation, while pk is the Fourier transform of the overlap integral p,,u =< $u(Rn)]+u(0) >. As defined by (40) ]&(Rn) > is a direct product of the hole part and the spin part. The energy of a propagating hole state is then given by
Ek =< 4klHl$k >= pk*

Cexp(ik. R,)E,,o < 0,

R,j6,0 >,

(59)

where the spin overlap integral < 0, R,(6,0 > can be written as e- with S as the HuangRhys factor, while En,0 is the hopping matrix element between hole states localized at 0 and R,, respectively. The energy dispersion for the polaron band formed by the ground state is presented in Fig. 10. The energy minimum is located at (5, $), in agreement with exact diagonalization [41,71-781 and the self-consistent Born approximation [64,66,67]. One can also calculate the effective mass. The results for the Ising case cr = 0 are given in Table I. From the numerical results we find that the wave function in this case is not sensitive to the ratio J/t within

-2L

FIG. 10. The hoie energy band (in units of 2) corresponding to the lowest bound state for J/t = 0.1
in the Ising case.

606

HOLES IN .4 TWO-DIMENSIONAL QUANTUM ANTIFERROMAGNET

VOL. 31

TABLE I. The dependence of hole effective mass on J/1 for o = 0

J/t
m*lm0

0.1 330

0.2 11.8

0.3 8.7

0.4 6.3

0.5 4.9

0.7 4.1

0.9 3.9

1.1 4.1

1.3 4.4

1.5 4.7

the range (0.2 - 2). By resealing the parameters /3i + /?iJ/t in the overlap integral we find that the band width N t/Jexp[-a(t/J)] where a is a positive constant. We note that the band width goes to zero in both limits t/J + 0 and t/J -+ 00. The energy separation A between the lowest level and the next one shows the same tendency, as seen from Fig. 8. The limit t/J + 0 is easy to understand, because the spin background is extremely stiff, so the hole cannot move. The opposite limit t/J -+ 03 is more interesting. The background is soft but the coupling is strong, so the hole carries along a huge spin excitation cloud which gives rise to a diverging Huang-Rhys factor and a full localization. Let us now compare the spin and the lattice polarons. If we consider t as an analogue of the electron-phonon coupling M, in (35), then the polaron effect diminishes in the lattice case if hciq/liiw, << 1, while in the spin case the polaron is completely self-trapped. The difference is due to the presence of a bare hopping term in the la.ttice case (35), as we mentioned in Sec. IV-2. On the other hand, the t/J + 00 limit is quite similar, where the polaron effect is pronounced. The functional dependence of the band width and the energy level separation found from the exact diagonalization and self-consistent Born approximation (- ( J/t)2/3 for small J/t, [64,66,67],(72-77]), is different from ours, except for Ref. 71, where the authors also reported an exponential dependence. The reason for this discrepancy has to be understood. One possibility is that these two functional forms are valid in different ranges of parameters. As for our calculations we still need to improve the self-consistency for the propagating state. The formation of the polaron band in terms of Bloch states (57) is well justified, if the localized state is determined by a Hartree type self-consistent equation. Here, however, the Fock term is crucial, so the self-consistency should be checked directly for a given k state. This problem is under consideration now. IV-6. Effects on the spin system The effect of single hole motion on the spin system is of the order of l/N which can be ignored in many cases. However, the presence of a. hole may have an important impact on local spin properties. We have alrea.dy seen in Sec. IV-4 the local spin distortion,

VOL.31

LU YU,ZHAO-BIN SU, AND YAN-hlIN LI

607

but there are other effects as well. The influence of a static hole (without t-term) has been studied by several authors [99], whereas the effects of a dynamical hole have been considered only in some numerical simulations and in a mean field treatment [loo]. Recently, the renormalization effect due to spin fluctuations on the hole motion has been discussed using a static approximation [loll, as a boson analogue of the well-known X-ray edge problem [102]. The authors of Ref. 101 found a zero frequency local mode which contributes substantially to the overlap integral e - between the state with a hole and the ground state without it. However, the resulting normalization factor X = 0.2 is quite modest. The quantum BdeG formalism developed by us to study the hole motion is a suitable tool to treat this problem. In fact, the hole motion and the spin background are two sides of the same coin. The formalism for this part was outlined in the second paper of Ref. 87, while the results have been reported recently [103]. The main task is to calculate the magnon propagator renormalized by the hole motion. In the lowest approximation the polarization operator II is given by the one-loop correction due to the hole (Fig. 5~). Using the spectral representation for the hole propagator, II can be expressed explicitly in terms of the solutions of the BdeG equation, described in Sec. IV-3. Omitting all technical details, we summarize here the main results [95,103]. Firstly, we have calculated the spin distortion. For the Ising case cy = 0 the.distortions are restricted to the nearest neighbours of the hole (Fig. lla). As a increases, they extend over a larger region (Fig. llb). It is important that they have the same s-symmetry as the hole state itself. As we mentioned earlier, this corresponds to a canted state with real parameters 0;. We have tried other possibilities like complex ,L?i, but the canted state has the lowest energy. Then, we have calculated the spin excitation spectrum in the presence of a hole. In all cases when a # 1, we have found several local modes. They can be classified according to symmetry as well. Again, one of the s-modes has the lowest energy. As cr increases, some of the local modes merge with the spin-wave continuum, becoming resonance modes. Furthermore, we have calculated various spin correlation functions around the hole. Bulut et al. [99] have found an amazing result for the Heisenberg model when the AF fluctuations are reduced near the hole in the static case. This is not true in the dynamical case. In fact, a confining structure is formed, i.e., the staggered magnetization is smaller near the hole than that far away from it, as conjectured in the spin bag model by Schrieffer et al. [85]. Even more interesting is the induced ferromagnetic < SFS, > correlations for the Ising case, which shows clearly the quantum nature of these fluctuations. Similarly, the AF correlations near the hole in the Heisenberg case is reduced, again due to this ferromagnetic tendency. All the above results agree with those of small cluster calculations

.-

608

HOLES IN A TWO-DIMENSIONAL QUANTUM ANTIFERROMAGNET

VOL. 31

0.0000

0.0000

0.0000

0.1845

0.0000

0.0000

63

0.1845

0.0000

(a)

0.0009

0.0000

0.0000

0.0000

-0.0422

0.0081

0.0000

0.0000

0.1503

-0.1386

0.0081

0.0000

63

0.1503

-0.0422

0.0009

b)
FIG. 11. Spin distortions around the hole for (a) cx = O.O,J/t = 0.3 and (b) a = 0.4, J/t = 0.1. In
view of the s-symmetry only one-quarter around the hole is shown.

Finally, we have used the corrected spin excitation spectrum to solve the BdeG equation for the hole states and have found no significant difference, which shows the self-

consistency of our procedure.

V. CONCLUDING REMARKS

We have made a short excursion to various theoretical approaches exploring the single hole problem in a quantum AF. Let us now briefly summarize what has been established and what are the open questions for further studies. (1) The first question was whether a hole can propagate in a quantum AF. The answer coming from the Brinkman-Rice treatment and Shraiman-Siggias quasi-classical approach

VOL. 31
was

LU YU,ZHAO-BIN SU, AND YAN-MIN LI

609

n o, i.e., the wave function renormalization factor 2 = 0. On the contrary, the self-consistent Born approximation and our quantum BdeG approach, as well as extensive numerical studies on clusters have shown a coherent propagation. What is the origin of this discrepancy? The problem has become especially acute in view of the controversy regarding the Fermi liquid vs non-Fermi liquid behaviour of charge carriers in high-T, superconductors. In particular, Anderson has used Shraiman-Siggias dipolar distortion of the spin background to argue the existence of an infrared catastrophe, leading to nonrenormalizability of the wave function and 2 = 0 [104]. It seems that the main origin of discrepancy is due to neglecting quantum fluctuations in the quasi classical approach. In fact, if these fluctuations are included in the retraceable path approximation, the catastrophe disappears [105]. T here are still long distance dipolar distortions, but the normalization integral converges. There is also some difference between the numerical results of exact diagonalization and quantum Monte Carlo simulations, as mentioned in Sec. 111-4. In view of the minus sign problem with Monte Carlo simulations and possible finite size effects in cluster calculations, this issue has to be further clarified. However, from a general physical point of view, it is hard to imagine that charge excitations cannot propagate in a magnetically long-range ordered medium. In all known examples, either the quantum fluctuations are strong so as to destroy the long range order, or they are sufficiently weak to allow long range order and well-defined quasi-particles [lOG]. However, it should be very clear that the free propagation of a single hole does not imply a Fermi liquid behaviour for a system with finite hole concentration. In fact, the collective behaviour of charge carriers is quite different from that of individual particles. For example, according to the single hole calculation, the Fermi surface would be small pockets around (&t,ltz). However, this is not true. A two-hole calculation in a small cluster, corresponding to 10% doping already exhibits a large electronic Fermi surface, compatible with the Luttinger theorem [107]. This implies that the hole-hole correlation and the decrease of the correlation length upon doping combine to make the single hole problem only of marginal relevance in the doping range where superconductivity occurs. (2) The existence of AF long-range order is the main factor leading to self-trapping (spin polaron), which can propagate coherently. The original suggestions of fully polarized ferromagnetic micro-region around the hole was a simplification of the real picture. It seems to us that a local distortion of ferromagnetic nature (canted state) is a quite good description. The size of this distortion corresponds, roughly, to the size of the confining string. Originally people thought that the behaviour of Ising and Heisenberg AF should be very different, since the spin distortion can be cured by the spin-flip process allowed for the latter. It turned out, however, that the string picture is valid to some extent even

610

HOLES IN A TWO-DIMENSIONAL QUANTUM ANTIFERROMAGNET

VOL. 31

in the Heisenberg case, at least for small J/t, as we mentioned in Sec. 111-3. One possible explanation could be that the reaction of the spin system (- l/J) is rather slow, so the string picture holds at a short time scale. However, there is only one parameter - the string length in the string picture which leads to a one-dimensional Airy equation. In the BdeG approach we have adopted, the self-trapping is taking place in 2D and we can classify the hole and the corresponding spin states by symmetry which should be helpful to gain more insight. (3) The spin polaron and the lattice polaron look rather similar, but there is a major difference. There exist three energy scales in the lattice polaron problem: the bare electron hopping, the phonon energy and the electron-phonon coupling strength, whereas in the spin problem these are only two: the magnon energy J and the coupling strength t. This is the basic reason why the spin polaron problem is more complicated: Some kind of renormalization procedure (infinite resummation of diagrams, summation of different paths in real space, etc.) is needed to generate the hole hopping. Then, what is the correct energy 7 scale for this hopping. J,t, or a new scale like t exp[-a(t/J)2] as we speculated? This is an open question. Probably, the answer depends on the parameter range. This feature of the spin polaron is the main origin of difficulty in analytical treatments. The self-consistent Born a.pproximation treats well the renormalization effect of a propagating state, but does not include the local distortion, which is essential over a quite large range of t/J. On the other hand, the BdeG formalism we are using is very good to guarantee the self-consistency of a localized state with spin distortion. However, the procedure is not fully self-consistent for a propagating state with fixed k. How to take advantages of these two approaches is a challenge for further research. As a general remark we would point out that the research on the spin polaron problem is still at a beginning stage compared with its lattice couterpart. Finally, we should mention that we have not considered in this article the implications for physical observable quantities, like optical conductivity, Raman spectra, photo-electron emission and so on. Some of these properties have been discussed in papers quoted (see, e.g. [41], [71-731, [107]), but t o summa.rize all the related research work is beyond the scope of this paper. Also, we have not touched at all on the hole-hole correlations and possible occurrence of superconducting fluctuations. There exists a vast literature on this issue [108], which we have to leave out to keep a reasonable size of the article. ACKNOWLEDGMENTS We would like to thank our collaborators W. Y. Lai, C. Q. Wu, and Q. F. Zhong. Some results of our joint research have been summarized in this article. We are grateful to many colleagues for helpful discussions shaping up our current understanding of this issue.

VOL. 31

LU YU, ZHAO-BIN SU, AND YAN-MIN LI

611

Among them we would especially mention P. A. Lee and E. Tosatti. Y. M. L. acknowledges financial support from SERC of the United Kingdom under grant No. GR/E/79798. This work is also supported from MUST/British Council under grant No. Ram/889/92/47. REFERENCES

[11 PI

J. G. Bednorz and I<. A. Miiller, Z. Phys. B64, 188 (1986); C. W. Chu et al., Phys. Rev. Lett. 58, 405 (1987). R. J. Birgeneau and G. Shirane, in Physics Properties of High Temperature Superconductors, Vol. 1, edited by D. M. Ginsberg (World Scientific, Singapore, 1989) p. 151; J. P. Rossat-Mignod, L. P. Regnault, J. M. Jarguens, P. Burlet, 3. Y. Henry, and G. Lapertot, in Dynamics of Magnetic Fluctuations in High Temperature Superconductors, G. Reiter, P. Horsch, and G. C. Psaltakis, eds. (Plenum, New York, 1991) p. 35. See, e.g., L. F. Mattheiss, Phys. Rev. Lett. 58, 1028 (1987). P. W. Anderson, Science 235, 1196 (1987); in Frontiers and Borderlines in ManyBody Physics, J. R. Schrieffer and R. A. Broglia, eds. (North-Holland, Amsterdam, 1987), p. 1. M. C. Gutzwiller, Phys. Rev. Lett. 10, 159 (1963); J. Hubbard, Proc. Roy. Sot. A276, 238 (1963); J. Kanamori, Progr. Theor. Phys. 30, 275 (1963). One can get some feeling about the recent developments from the book The Hubbard Model, edited by M. Rasetti (World Scientific, Singapore, 1991) and the series of Proceedings of ICTP Workshops on Strongly Correlated Electron Systems (I-III), aIso published by World Scientific. Many authors have carried out this transformation, see, e.g., K. A. Chao et al., J. Phys. ClO, L271 (1977); J. E. Hirsch, Phys. Rev. Lett. 54, 1371 (1985); A. H. MacDonald, S. M. Girvin, and D. Yoshioka, Phys. Rev. B37, 9753 (1988). See, e.g., Physical Properties of High Temperature Superconductors, Vols. 1-3, edited by D. M. Ginsberg (World Scientific, Singapore); High Temperature Superconductivity, K. S. Bedell et al., eds. (Addison-Wesley, Redwood City, 1990); Yu Lu in Recent Progress in Many-Body Theories, Vol. 3, T. L. Ainsworth, C. E. Campbell, B. E. Clements, and E. Krotschek, eds. (Plenum, New York, 1992) p. 157. C. M. Varma, S. Schmitt-Rink, and E. Abrahams, Solid State Commun. 62, 681 (1987); V. J. Emery, Phys. Rev. Lett. 58, 2794 (1987); J. E. Hirsch, Phys. Rev. Lett. 59, 228 (1987). F. C. Zhang and T. M. Rice, Phys. Rev. B37, 3759 (1988).

[31 141 PI 161

PI Dl

Dl [ 101

L.___

612

HOLES IN A TWO-DIMENSIONAL QUANTUM ANTIFERROMAGNET

VOL. 31

[ 111 J. G. Lorenzana and L. Yu, Phys. Rev. Lett. 67, 871 (1993). rie [ 121 J. R. Sc h ffer, Theory of Soperconductivity (Benjamin, 1964).
[ 13 ] See, e.g., P. Nozidres, Theory of Interactiny Fernzi systems (New York, 1964). [ 14 ] A. B. MigdaI, S ov. Phys. JETP, 5, 333 (1957). [ 15 ] J. M. Luttinger, Phys. Rev. 119, 1153 (1960). [ 16 ] H. A. Bethe, Z. Phys. 71, 205 (1931); S ee also R. L. Orbach, Phys. Rev. 112, 309

(1958).
[ 17 ] C. N. Yang and C. P. Yang, Phys. Rev. Lett. 20, 9 (1966). [ 18 ] E. H. Lieb and F. Y. Wu, Phys. Rev. Lett. 20, 1445 (1968). [ 19 ] W. Marshall, Proc. Roy. Sot., London, Ser. A232, 48 (1955). [ 20 ] T. Kennedy, E. H. Lieb, and B. S. Shastry, J. Stat. Phys. 53, 1019 (198s). . [ 21 ] See, K. Kubo, Phys. Rev. Lett. 61, 110 (19SS); K. Kubo and T. Kishi, Phys. Rev.

Lett. 61, 2585 (19S8).

[ 22 ] See, e.g., the review article, E. Manousakis, Revs. Mod. Phys. 63, 1 (1991). [ 23 ] N. D. Mermin and H. Wagner, Phys. Rev. Lett. 22, 1133 (1966). [ 24 ] See, e.g., I. E. Dzyaloshinskii, A. M. Polyakov, and P. B. Wiegman, Phys. Lett.

127A, 112 (19S8).

[ 25 ] P. W. Anderson, Mat. Res. Bull. 8, 153 (1973). [ 26 ] S. D. Liang, B. D oucot, and P. W. Anderson, Phys. Rev. Lett. 61, 365 (1988). [ 27 ] T. Holstein and H. Primakoff, Phys. Rev. 58, 109s (1940). [ 28 ] P. W. A n derson, Phys. Rev. 86, 694 (1952); R. Kubo, Phys. Rev. 87, 568 (1952).

[ 29 ] See, e.g., D. Mattis, Theory of A4agnetism (Harper and Row, New York, 1982).
[ 30 ] Y. R. Wang and M. J. Rice, Phys. Rev. B45, 5045 (1992). [ 311 F. Dyson, Phys. Rev. 102, 1217, 1230 (1956); S.V. Maleev, Zh. Eksp. Teor. Fiz.

30, 1010 (1957).

[ 32 ] D. P. Arovas a.nd A. Auerba.ch, Phys. Rev. B38, 316 (1988). [ 33 ] M. Takahashi, Phys. Rev. B40, 2494 (1989). [ 34 ] Y. R. Wang, Phys. Rev. B46, 151 (1992) and references therein. [ 35 ] See, e.g. E. J. Mele, Phys. Ser. T27, 82 (1989); E. Fradkin, Phys. Rev. Lett. 63,322

(1989); J. A bj rn and G. Semenoff, Phys. Lett. 226B, 107 (1989); G. Baskaran, in m o Two-Dimensional Strongly Correlated Electron Systems, Z. Z. Gan and Z. B. Su, eds. (Gordon and Breach, 198s) p. 47.

VOL.

31

LU YU,ZHAO-BIN SU,AND YAN-MIN LI

613

[ 36 ] S. P. Feng, 2. B. S u, and L. Yu, sumbitted to Phys. Rev. B.

[ 37 ] Y. Nagaoka, Phys. Rev. 147, 392 (1966).

[ 38 ] See, e.g., Y. Fang et al., Phys. Rev. B40, 7406 (1989); J. A. Riera and A. P. Young, ibid, 40, 5285 (1989); B. D oucot and X. G. Wen, ibid, 40, 2719 (1989). ra [ 39 ] X. Y. Zhang, E. Ab 1 rams, and G. Rotliar, Phys. Rev. Lett. 66, 1236 (1991).
[ 40 ] See, e.g., T. Barnes, E. Dagotto, A. Moreo, and E. S. Swanson, Phys. Rev. B40,

10977 (19S9).
[ 41 ] See, e.g., J. BonEa, P. PrelovSek, and I. Sega, Phys. Rev. B39, 7074 (1989). [ 42 ] P. W. Anderson, in Current Trends in Condensed Matter, Particle Physics and Cosmology, J. Pati, Q. Shafi, S. Wadia, and Yu Lu, eds. (World Scientific, Singapore,

1990) p. 323; P. W. Anderson and Y. Ren, in High Temperature Superconductivity, quoted in Ref. 8, and Ch. 5 in the book on RVB to be published by Princeton University Press.
[ 43 ] P. W. Anderson, Phys. Rev. Lett. 18, 1049 (1967). . [ 441 M. Ogata and H. Shiba, Phys. R.ev. B41, 2326 (1990). [ 45 ] A. Parola a.nd S. S orella, Phys. Rev. Lett. 64, 1831 (1990). [ 46 ] F. D. Haldane, Phys. Rev. Lett. 45, 1358 (19SO) ; Phys. Lett. SlA, 151 (1981); J.

Phys. C14, 2585 (1981).

[ 47 ] J. B. Goodenough, Phys. Rev. 100, 564 (1955). [ 48 ] P. G. de Gennes, Phys. Rev. 118, 141 (1960).

[ 49 ] See, e.g., E. L. Nagaev, Sov. Phys. JETP Lett. 6, 181 (1967); L. N. Bulaevskii and D. I. Khomskii, Sov. Phys. JETP 25, 1067 (1967).
[ 50 ] L. N. Bulaevskii, E. L. Nagaev, and D. I. Khomskii, Sov. Phys. JETP 27, 638

(1968).
[ 511 E. L. Nagaev, Phys. Stat. Sol. (b) 65, 11 (1979); E. L. Nagaev, Physics of Magnetic Semiconductors (MIR, Moscow, 1979). [ 52 ] W. F. Brinkman and T. M. Rice, Phys. Rev. B2, 1324 (1970). [ 53 ] B. I. Shraiman and E. D. Siggia, Phys. Rev. Lett. 60. 740 (1988).

[54] M. M. M olan and N. Kumar, J. Phys. C20, L.527 (1987); N. Kumar, Phys. Rev. 1 B40, 844 (19S9); M. M. Mohan, ibid, 46, 340.5 (1992) and references therein; B. D. Simons and J. M. F. Gunn, ibid, 41, 7019 (1990); J. L. Richard and V. Yu. Yushankhai, ibid, 47, 1103 (1993).
[ 55 ] W. Metzner, P. Schmit, and D. Vollhardt, Phys. Rev. B45, 2237 (1992); R. Strack

and D. Vollhardt, ibid, 46, 138.52 (1992).

614

HOLESINATWO-DIMENSIONALQUANTUMANTIFERROMAGNET

VOL.31

[ 56 ] S. A. Trugman, Phys. Rev. B37, 1597 (1988); ibid, 41, 892 (1990). [ 57 ] F. D. M. H aId ane, Phys. Lett. 93A, 464 (1983); I. Affleck, Nucl. Phys. B257, 397 (1985).
[ 58 ] B. I. Shraiman and E. D. Siggia, Phys. Rev. Lett. 61, 467 (1988); ibid, 62, 1564

(1989); Phys. Rev. B42, 2485 (1990); ibid, 46, 8305 (1992). [ 59 ] P. Wiegmann, Phys. Rev. Lett. 60, 821 (1988); X. G. Wen, Phys. Rev. B39, 7223 (1989); M. J. Godfrey and J. M. F. Gunn, J. Phys: Cond. Matt. Phys. 1, 5 8 2 1 (1989); Phys. Rev. B47, 8124 (1993) and references therein.
[ 60 ] A. Auerbach and B. E. Larson, Phys. Rev. Lett. 66, 2262 (1991); A. Auerbach,

Proceedings of 1991 Jerusalem Winter School Strongly Correlated Electrons, to be published.

[ 61 ] R. Shankar, Phys. Rev. Lett. 63, 203 (1989); P. A. Lee, Phys. Rev. Lett. 63, .

690

(1989); papers by P. Wiegmann and X. G. Wen in Ref. 59.

[ 62 ] S. Schmitt-Rink, C. M. Varma, and A. E. Ruckenstein, Phys. Rev. Lett. 60, 2793

(1988).
[ 63 ] C. L. Kane, P. A. Lee, and N. Read, Phys. Rev. B39, 6880 (1989). [ 64 ] G. Martinez and P. Horsch, Phys. Rev. B44, 317 (1991). [ 65 ] See, e.g., J. Appel in Solid State Physics, Vol. 21 (Academic Press, New York, 1968)

p. 193.
[ 66 ] F. Marsiglio, A. E. Ruckenstein, S. Schmitt-Rink, and C. M. Varma, Phys. Rev.

B43, 10882 (1991).

[ 67 ] Z. Liu and E. Manousakis, Phys. Rev. B44, 2414 (1991); ibid, 45, 2425 (1992). [ 68 ] C. Gros and M. D. Johnson, Phys. Rev. B40, 9423 (1989); M. D. Johnson, C. Gras,

and I<. J. von Szczepanski! ibid, 43, 11207 (1991). [ 69 ] S. Sachdev, Phys. Rev. B40,5024 (1989); S ee, also M. Boninsegui and E. Manousakis, ibid, 43, 10353 (1991).
[ 70 ] E. Kasiras and E. Manousakis, Phys. Rev. B37, 656; ibid, 38, 866 (1988). [ 711 C.-X. C lien and H. B. Schi;ttler, Phys. Rev. B40, 239 (1989); ibid, 41, 8702 (1990);

C.-X. Chen, H. B. Schiittler, and A. J. Fedro, ibid, 41, 2581 (1990).

[ 72 ] I<. J. von Szczepanski, P. Horsch, W. H. Stephan, and M. Ziegler, Phys. Rev. B41,

2107 (1990).
[ 731 E. Dagotto, R. Joynt, A. Moreo, S. Bacci, and E. Gaghano, Phys. Rev. B41, 2585,

9049 (1990) and references therein; for a review, see: E. Dagotto, Int. J. Mod. Phys. B5, 77 (1991); S ee, also E. Da.gotto a.nd J. R. Schrieffer, Phys. Rev. B43, 8705 (1991).

VOL.

31

LU YU,ZHAO-BIN SU,AND YAN-MIN LI

615

[ 74 ] A. Parola, S. Sorella, S. B aroni, M. Parrinello, a.nd E. Tosatti, Int. J. Mod. Phys. B3, 1865 (1989); A. Parola, S. Sorella, M. Parrinello, and E. Tosatti, Phys. Rev. B 43, 6190 (1991); G. Fano, F. Ortolani, and A. Parola, ibid, 46, 1048 (1992).
[ 75 ] V. Elser, D. Huse, B. Shraimann, and E. D. Siggia, Phys. Rev. B41, 6715 (1990). [ 76 ] J. A. Riera and A. P. Young, Phys. Rev. B39, 9697, 40, 5285 (1989); Y. Hasegawa

and D. Poilblanc, ibid, 40, 9035 (1989); K.-H. Luk and D. L. Cox, ibid, 41, 4456 (1990); N. Furu yama and M. Imada, J. Phys. Sot. Jpn. 59, 177 (1990); T. Xiang, Phys. Rev. B44, 2276 (1991).
[ 77 ] D. Poilbl ant, H. J. Schulz, and T. Ziman, Phys. Rev. B46, 6435 (1992); ibid, 47,

3268 (1993); D. Poilbla.nc, T. Ziman, H. J. Schulz, and E. Da.gotto, ibid, 47, 14267 (1993).
[ 78 ] J. Song and J. F. Annett, Europhys. Lett. 18, 549 (1992).

[ 79 ] J. M. F. Gunn and B. D. Simons, Phys. Rev. B42, 4370 (1990).

[ 80 ] J. Inoue and S. Maekawa, J. Phys. Sot. Jpn. 59, 2110 (1990). . [ 811 S. Sorella, Phys. Rev. B46, 11670 (1992). [ 82 ] J. E. Hirsch, Phys.

Rev. 31, 4403 (1985); S. S orella, S. Baroni, R. Car, and M. Parrinello, Europhys. Lett. 8, 663 (1989).

[ 83 ] M. Boninsegui and E. Manousakis, Phys. Rev. B46, 560 (1992). [ 84 ] T. Giamarchi and C. Lhillier, Phys. Rev. B47, 2775 (1993). [ 85 ] J. R. Schrieffer, X. G. Wen, and S. C. Zhang, Phys. Rev. Lett. 60, 944 (1988); Phys.

Rev. B39, 11663 (1989).

[ 86 ] W. P. Su and X. Y. Chen, Phys. Rev. B38, 8879 (1988); G. Vignale and M. R.

Hedayati, ibid, 42, 786 (1990); A. Sing11 and Z. Tesanovic, ibid, 41, 614, 11604(E) (1990); Z. Y. Weng, C. S. Ting, and T. K. Lee, ibid, 41, 1990 (1990); D. M. Frenkel and M. R. Hanke, ibid, 42, 786 (1990).
[ 87 ] Z. B. Su, Y. M. Li, W. Y. Lai, and L. Yu, Phys. Rev. Lett. 63, 1318 (1989); Int. J.

Mod. Phys. B3, 1913 (1989).

[ 88 ] T. Holstein, Ann. Phys. 8, 325, 343 (1989); S ee, also G. D. Mahan, Many-Particle

Physics (Plenum, New York, 1990) 2nd edition. [ 89 ] I. C. Lang and Yu. A. Firsov, Sov. Phys. JETP 16. 1301 (1963). [ 901 K. Huang and A. Rhys, Proc. Roy. Sot. A204, 406 (1950).
[ 911 Z. B. Su and L. Yu, Phys. Rev. B27, 5199 (1983); Commun. in Theor. Phys. 2,

1203, 1323, 1341 (1983).

616

HOLES IN A TWO-DIMENSIONAL QUANTUM ANTIFERROMAGNET

VOL. 31

[ 92 ] P. G. de Gennes, Superconductivity of n4etals and Alloys (Benjamin, New York, 1966). [ 93 ] See, e.g., Yu Lu, Solitons and Polarons in Conducting Polymers (World Scientific, Singapore, 1988). [ 94 ] K. C. Chou, Z. B. Su, B. L. Hao, and L. Yu, Phys. Rep. 118, 1 (1985). [ 95 ] Q. F. Zhong, W. Y. L ai, Y. M. Li, Z. B. Su, and L. Yu, to be published. [ 96 ] C. Q. Wu, Y. M. Li, Z. B. Su, and L. Yu, to be published. [ 97 ] See, e.g., C. L. Kane, P. A. Lee, T. K. Ng, B. Chakraborty, and N. Read, Phys. Rev. B41, 2653 (1990); B. Chakraborty, N. Read, C. L. Kane, and P. A. Lee, Phys. Rev. 42, 4819 (1990). [ 98 ] A. Auerbach and B. E. Larson, Phys. Rev. B43, 7800 (1991). . [ 99 ] N. Bulut, D. Hone, D. J. Scalapino, and E. Y. Loh, Phys. Rev. Lett. 62, 2192 (1989); N. Nagaosa, Y. Hatsugai, and M. Imada, J. Phys. Sot. Jpn. 58, 978 (1989). [loo] H. Y. Choi, Phys. Rev. B44, 2609 (1991). [loll A. G. M 1s huk ov and G. D. Maha.n, Phys. Rev. Lett. 68, 2200 (1992). a [102] See, e.g., G. D. Maha.ns book, Ref. 88.

[103] Y. M. Li, W. Y. Lai, Z. B. Su, and L. Yu, Phys. Rev. B47, 7938 (1993); See, also W. Y. Lai, Y. M. Li, Z. B. Su, and L. Yu, Pmgress in High Temperature Superconductivity, Vol. 22 (World Scientific, Singapore, 1990) p. 503. [104] P. W. Anderson, Cargese Lectures (1990). His argument has been reproduced in Refs. Sl and 105. [105] G. Reiter, to be published. [106] We thank E. Tosatti for this rema.rk. [107] W. Stephan and P. Horsch, Phys. Rev. Lett. 66, 2258 (1991). [lOS] See, e.g., P. PrelovSek and X. Zotos, Phys. Rev. B47, 5984 (1993); D. Poilblanc, preprint LPQTH 9301; D. Poilblanc, J. Riera, and E. Dagotto, preprint LPQTH 93/06 and references therein.