quantum magnets- a brief overview
DESCRIPTION
Review on quantum magnetsTRANSCRIPT
-
arX
iv:c
ond-
mat
/010
7399
v1 [
cond
-mat.
str-el
] 19
Jul 2
001
Quantum magnets: a brief overview
Indrani Bose
February 1, 2008
Department of Physis
Bose Institute
93/1, A. P. C. Road
Calutta-700009
Abstrat
Quantum magnetism is one of the most ative areas of researh in on-
densed matter physis. There is signiant researh interest speially in
low-dimensional quantum spin systems. Suh systems have a large num-
ber of experimental realizations and exhibit a variety of phenomena the
origin of whih an be attributed to quantum eets and low dimensions.
In this review, an overview of some aspets of quantum magnetism in
low dimensions is given. The emphasis is on key onepts, theorems and
rigorous results as well as models of spin hains, ladders and frustated
magneti systems.
1 Introdution
Quantum magnets are spin systems in whih the spins interat via the well-
known exhange interation. The interation is purely quantum mehanial in
nature and the form of the interation was derived simultaneously by Heisenberg
and Dira in 1926 [1. The most well-known model of interating spins in an
insulating solid is the Heisenberg model with the Hamiltonian
H =ij
JijSi.Sj (1)
Si is the spin operator loated at the lattie site i and Jij denotes the strength
of the exhange interation. The spin
Si an have a magnitude 1/2, 1, 3/2, 2,...et. The lattie, at the sites of whih the spins are loated, is d-dimensional.
Examples are a linear hain (d = 1), the square lattie (d = 2 ) and the ubi
lattie (d =3 ). Ladders have strutures interpolating between the hain and the
square lattie. An n-hain ladder onsists of n hains (n = 2, 3, 4,...et.) oupled
by rungs. Real magneti solids are three-dimensional (3d) but an be eetively
onsidered as low-dimensional systems if the exhange interations have dierent
1
-
strengths in dierent diretions. To give an example, a magneti solid may
onsist of hains of spins. The solid may be onsidered as a linear hain (d=1)
ompound if the intra-hain exhange interations are muh stronger than the
inter-hain ones. In a planar (d=2) magneti system, the dominant exhange
interations are intra-planar. Several examples of low-dimensional magneti
systems are given in [2.
The strength of the exhange interation Jij in Eq.(1) falls down rapidly
as the distane between interating spins inreases. For many solids, the sites
i and j are nearest-neighbours (n.n.s) on the lattie and Jij 's have the same
magnitude J for all the n.n. interations. The Hamiltonian in (1) then beomes
H = Jij
Si.Sj (2)
There are, however, examples of magneti systems in whih the strengths of the
exhange interations between su
essive pairs of spins are not the same. Also,
the interation Hamiltonian (1) may inlude n.n. as well as further-neighbour
interations. The well-known Majumdar-Ghosh hain [3 is desribed by the
Hamiltonian
HMG = JNi=1
Si.S i+1 +
J
2
Ni=1
Si.S i+2 (3)
and inludes n.n. as well as next-nearest-neighbour (n.n.n.) interations.The
Haldane-Shastry model [4 has a Hamiltonian of the form
H = Jij
1
|i j|2Si.Sj (4)
and inludes long-ranged interations. Real materials are haraterised by var-
ious types of anisotropy. The fully anisotropi n.n. Heisenberg Hamiltonian in
1d is given by
HXY Z =Ni=1
[JxSxi S
xi+1 + JyS
yi S
yi+1 + JzS
zi S
zi+1] (5)
The speial ases of this Hamiltonian are: the Ising (Jx = Jy = 0) , the XY(Jz = 0), the XXX or isotropi Heisenberg (Jx = Jy = Jz ) and the XXZ oranisotropi Heisenberg (Jx = Jy 6= Jz ) models. There is a huge literatureon these models some of whih are summarised in Refs. [1, 2, 5, 6, 7. Other
anisotropy terms may be present in the full spin Hamiltonian besides the basi
exhange interation terms.
Consider the isotropi Heisenberg Hamiltonian in (2) where ij denotesa n.n. pair of spins. The sign of the exhange interation determines the
favourable alignment of the n.n. spins. J > 0(J < 0) orresponds to anti-ferromagneti (ferromagneti) exhange interation. To see how exhange in-
teration leads to magneti order, treat the spins as lassial vetors. Eah n.n.
2
-
spin pair has an interation energy JS2cos where is the angle between n.n.
spin orientations. When J is < 0, the lowest energy is ahieved when = 0,i.e., the interating spins are parallel. The ferromagneti (FM) ground state
has all the spins parallel and the ground state energy Eg = J NzS22 where zis the oordination number of the lattie. When J is > 0, the lowest energy is
ahieved for = , i.e., the n.n. spins are antiparallel. The antiferromagneti(AFM) ground state is the Nel state in whih n.n. spins are antiparallel to
eah other. The ground state energy Eg = JNzS22 .Magnetism , however, is a purely quantum phenomenon and the Hamiltonian
(2) is to be treated quantum mehanially rather than lassially. For simpli-
ity, onsider a hain of spins of magnitude
12 . Periodi boundary ondition is
assumed, i.e.,
S N+1 =
S1 . The Hamiltonian (2) an be written as
H = J
Ni=1
[Szi Szi+1 +
1
2(S+i S
i+1 + S
i S
+i+1)] (6)
where
Si = Sxi iSyi (7)
are the raising and lowering operators. It is easy to hek that in the ase of a
FM, the lassial ground state is still the quantum mehanial ground state with
the same ground state energy. However, the lassial AFM ground state (the
Nel state) is not the quantum mehanial ground state. The determination of
the exat AFM ground state is a tough many body problem and the solution
an be obtained with the help of the Bethe Ansatz tehnique (Setion 2).
For a spin-1/2 system, the number of eigenstates of the Hamiltonian is 2N
where N is the number of spins. In a real solid N is 1023 and exat determi-nation of all the eigenvalues and the eigenfuntions of the system is impossible.
There are some lasses of AFM spin models for whih the ground state and in
a few ases the low-lying exitation spetrum are known exatly (Setion 2). In
the majority of ases, however, the ground state and the low-lying exited states
are determined in an approximate manner. Knowledge of the low-lying exita-
tion spetrum enables one to determine the low-temperature thermodynamis
and the response to weak external elds. The usual thermodynami quantities
of a magneti system are magnetisation, spei heat and suseptibility. Ex-
hange interation an give rise to magneti order below a ritial temperature.
However, for some spin systems, the ground state is disordered, i.e., there is no
magneti order even at T = 0. Long range order (LRO) of the Nel-type exists
in the magneti system if
limRS (0).
S (R )6= 0 (8)
where
R denotes the spatial loation of the spin. At T = 0, the expeta-
tion value is in the ground state and at T 6= 0 , the expetation value is theusual thermodynami average. The dynamial properties of a magneti system
3
-
are governed by the time-dependent pair orrelation funtions or their Fourier
transforms. Quantities of experimental interest inlude the dynamial orre-
lation funtions in neutron sattering experiments, the NMR spin-lattie re-
laxation rate, various relaxation funtions and assoiated lineshapes as well as
the dynamial response of the magneti system to various spetrosopi probes
[8. Knowledge of the ground and low-lying states and the orresponding energy
eigenvalues is essential to determine the thermodynami and dynami properties
of a magneti system.
The disovery of high-temperature uprate superondutors in 1987 has
given a tremendous boost to researh ativity in magnetism. The dominant
eletroni and magneti properties of the uprate systems are assoiated with
the opper-oxide (CuO2) planes. The Cu2+
ions arry spin-
12 and the spins
interat via the Heisenberg AFM exhange interation Hamiltonian. This fat
has given rise to a large number of studies on 2d antiferromagnets. The uprates
exhibit a variety of novel phenomena in their insulating, metalli and superon-
duting phases some of whih at least have links to quantum magnetism. The
subjet of magnetism has, as a result, expanded signiantly in sope and on-
tent. A rih interplay between theory and experiments has led to the disovery
of materials exhibiting hitherto unknown phenomena, formulation of new theo-
retial ideas, solution of old puzzles and opening up of new researh possibilities.
In this review, a brief overview of some of the important developments in quan-
tum magnetism will be given. The fous is on quantum antiferromagnets and
insulating solids.
2 Theorems and rigorous results
(i) Theorems :
A. Lieb-Mattis theorem [9
For general spin and for all dimensions and also for a bipartite lattie, the
entire eigenvalue spetrum satises the inequality
E0(S) E0(S + 1) (9)where E0(S) is the minimum energy orresponding to total spin S. The weakinequality beomes a strit inequality for a FM exhange oupling between spins
of the same sublattie. The theorem is valid for any range of exhange oupling
and the proof does not require PBC. The ground state of the S = 12 HeisenbergAFM with an even number N of spins is a singlet a
ording to the Lieb-Mattis
theorem.
B. Marshall's sign rule [10
The rule speies the struture of the ground state of a n.n. S = 12 Heisen-berg Hamiltonian dened on a bipartite lattie. The rule an be generalised to
spin S, n.n.n. FM interation but not to n.n.n. AFM interation. A bipartite
lattie is a lattie whih an be divided into two sublatties A and B suh that
the n.n. spins of a spin belonging to the A sublattie are loated in the B sub-
lattie and vie versa. Examples of suh latties are the linear hain, the square
4
-
and the ubi latties. A
ording to the sign rule, the ground state has the
form
| =
C | (10)
where | is an Ising basis state. The oeient C has the form
C = ()pa (11)with a real and 0 and p is the number of up-spins in the A sublattie.C. Lieb, Shultz and Mattis (LSM) theorem [11:
A half-integer S spin hain desribed by a reasonably loal Hamiltonian re-
speting translational and rotational symmetry either has gapless exitation
spetrum or has degenerate ground states, orresponding to spontaneously bro-
ken translational symmetry.
In the ase of a gapless exitation spetrum, there is at least one momentum
wave vetor for whih the exitation energy is zero. For a spetrum with gap,
the lowest exitation is separated from the ground state by an energy gap .The temperature dependene of thermodynami quantities is determined by
the nature of the exitation spetrum (with or without gap). The LSM theorem
does not hold true for integer spin hains. For suh hains, Haldane made a
onjeture that the spin exitation spetrum is gapped [12. This onjeture
has been veried both theoretially and experimentally [13.
D. Oshikawa, Yamanaka and Aek theorem [14
This theorem extends the LSM theorem to the ase of an applied magneti
eld. The ontent of the theorem is : translationally invariant spin hains in an
applied eld an have a gapped exitation spetrum, without breaking transla-
tional symmetry, only when the magnetization per site m (m = 1N
Ni=1 S
zi , N
is the total number of spins in the system ) obeys the relation
S m = integer (12)where S is the magnitude of the spin. The proof is an easy extension of that of
the LSM theorem. The gapped phases orrespond to magnetization plateaux in
the m vs. H urve at the quantized values of m whih satisfy (12). Whenever
there is a gap in the spin exitation spetrum, it is obvious that the magnetiza-
tion annot hange in hanging external eld. Frational quantization an also
o
ur, if a
ompanied by (expliit or spontaneous) breaking of the translational
symmetry. In this ase, the plateau ondition is given by
n(S m) = integer (13)where n is the period of the ground state. Hida [15 has onsidered a S = 12HAFM hain with period 3 exhange oupling. A plateau in the magnetization
urve o
urs at m = 16 (13 of full magnetization ). In this ase, n =3, S =
12 and
m = 16 and the quantization ondition in (13) is obeyed. Ref. [16 gives a reviewof magnetization plateaux in interating spin systems. Magnetization plateaux
5
-
have been observed in the magneti ompound NH4CuCl3 at m =14 and
34
[17. Possible extensions of the LSM theorem to higher dimensions have been
suggested [18. The ompound SrCu2(BO3)2 is the rst AFM ompound in 2din whih magnetization plateaux have been observed experimentally [19. Like
the quantum Hall eet, the phenomenon of magnetization plateaux is another
striking example of the quantization of a physially measurable quantity as a
funtion of the magneti eld.
E. Mermin-Wagner's theorem [20
There annot be any AFM LRO at nite T in dimensions d =1 and 2. The
LRO an, however, exist in the ground state of spin models in d =2. LRO exists
in the ground state of the 3d HAFM model for spin S 12 [21. At nite T, theLRO persists upto a ritial temperature Tc . For square [22 and hexagonal
[23 latties, LRO exists in the ground state for S 1 . The above results arebased on rigorous proofs. No suh proof exists as yet for S = 12 , d =2 (this
ase is of interest beause the CuO2 plane of the high-Tc uprate systems is a
S = 12 2d AFM).(ii) Exat Results :
A. the Bethe Ansatz [24
The Bethe Ansatz (BA) was formulated by Bethe in 1931 and desribes a
wave funtion with a partiular kind of struture. Bethe onsidered the spin 12Heisenberg linear hain in whih only n.n. spins interat. In the ase of the
FM hain, the exat ground state is simple with all spins aligned in the same
diretion, say, pointing up. An exitation is reated in the system by deviating
a spin from its ground state arrangement, i.e., replaing an up-spin by a down-
spin. Due to the exhange interation, the deviated spin does not stay loalised
at a partiular site but travels along the hain of spins. This exitation is the
so-alled spin wave or magnon. For the isotropi FM Heisenberg Hamiltonian,
the exat one-magnon eigenstate is given by
=
Nm=1
eikmSm | ....... (14)
where m denotes the site at whih the down-spin is loated and the summation
over m runs from the rst to the last site in the hain. The k's are the momenta
whih from periodi boundary onditions have N allowed values
k =2
N, = 0, 1, 2, ..., N 1 (15)
The exitation energy k, measured w.r.t. the ground state energy and in units
of J, is
k = (1 cosk) (16)In the ase of r spin deviations (magnons), the eigenfuntion an be written as
(r) =
m1
-
The amplitudes are given by the BA
a(m1,m2, ...,mr) =P
ei
jkPjmj+
12i
1,r
j
-
as in the FM ase but the sign of the exhange integral hanges from J to J(J > 0 ). The total spin of the AFM ground state is S = 0 a
ording to theLieb-Mattis theorem. In the ground state,
N2 spins are up and
N2 spins down
(r = N2 ). The ground stae is non-degenerate and there is a unique hoie of thei 's as
1 = 1, 2 = 3, 3 = 5, ...., N2= N 1 (23)
The ground state is also spin disordered, i.e., has no AFM LRO. The exat
ground state energy Eg is
Eg =NJ
4 JNln2 (24)
The low-lying exitation spetrum has been alulated by des Cloizeaux and
Pearson (dCP) [27 by making appropriate hanges in the distribution of i 's
in the ground state. The spetrum is given by
=
2|sink| , k (25)
for spin 1 states. The wave vetor k is measured w.r.t. that of the ground
state. A more rigorous alulation of the low-lying exitation spetrum has
been given by Faddeev and Takhtajan [28. There are S = 1 as well as S = 0states. We give a qualitative desription of the exitation spetrum, for details
Ref. [28 should be onsulted. The energy of the low-lying exited states an be
written as E(k1, k2) = (k1) + (k2) with (ki) =pi2 sinki and total momentum
k = k1 + k2. At a xed total momentum k, one gets a ontinuum of satteringstates. The lower boundary of the ontinuum is given by the dCP spetrum
(one of the kis = 0 ). The upper boundary is obtained for k1 = k2 =k2 and
Uk =
sink2 (26)
The energy-momentum relations suggest that the low-lying spetrum is atually
a ombination of two elementary exitations known as spinons. The energies
and the momenta of the spinons just add up, showing that they do not interat.
A spinon is a S = 12 objet, so on ombination they give rise to both S = 1 andS = 0 states. In the Heisenberg model, the spinons are only noninterating inthe thermodynami limit N . For an even number N of sites, the total spinis always an integer, so that the spins are always exited in pairs. The spinons
an be visualised as kinks in the AFM order parameter. Due to the exhange
interation, the individual spinons get deloalised into plane wave states. In-
elasti neutron sattering study of the linear hain S = 12 HAFM ompoundKCuF3 has onrmed the existene of unbound spinon pair exitations [29.
The Haldane-Shastry model [4 is another spin 12 model in 1d for whih theground state and low-lying exitation spetrum are known exatly. The ground
state has the same funtional form as the frational quantum Hall ground state
and is spin-disordered. The elementary exitations are spinons whih are non-
interating even away from the thermodynami limit, i.e., in nite systems.
8
-
The individual spinons behave as semions, i.e., have statistial properties inter-
mediate between fermions and bosons. In the ase of integer spin hains, the
spinons are bound and the exitation spetrum onsists of spin-wave-like modes
exhibiting the Haldane gap. The BA tehnique desribed in this Setion is the
one originally proposed by Bethe. There is an algebrai version of the BA whih
is in wide use and whih gives the same nal results as the earlier tehnique. For
an introdution to the algebrai BA method, see the Refs. [30, 31. A tutorial
review of the BA is given in Ref. [32. The BA was originally proposed for
the Heisenberg model in magnetism. Later, the method was applied to other
interating many body systems in 1d suh as the Fermi and Bose gas models
in whih partiles on a line interat through delta funtion potentials [33, the
Hubbard model in 1d [34, 1d plasma whih rystallizes as a Wigner solid [35,
the Lai-Sutherland model [36 whih inludes the Hubbard model and a dilute
magneti model as speial ases, the Kondo model in 1d [37, the single impurity
Anderson model in 1d [38, the supersymmetri t-J model (J = 2t) [39 et. In
the ase of quantum models, the BA method is appliable only to 1d models.
The BA method has also been applied to derive exat results for lassial lattie
statistial models in 2d.
B. The Majumdar-Ghosh hain [3, 7
The Hamiltonian is given in Eq. (3). The exat ground state of HMG is doubly
degenerate and the states are
1 [12][34]...[N 1N ], 2 [23][45]...[N1] (27)where [lm] denotes a singlet spin onguration for spins loated at the sitesl and m. Also, PBC is assumed. One nds that translational symmetry is
broken in the ground state. The proof that 1 and 2 are the exat ground
states an be obtained by the method of `divide and onquer'. One an verify
that 1 and 2 are exat eigenstates of HMG by applying the spin identityS n.(
S l +
S m)[lm] = 0 . Let E1 be the energy of 1 and 2 . Let Eg be the
exat ground state energy. Then Eg E1 . One divides the Hamiltonian H intosub-Hamiltonians , Hi 's, suh that H =
iHi . Hi an be exatly diagonalised
and let Ei0 be the ground state energy. Let g be the exat ground state wave
funtion. By variational theory,
Eg = g |H |g =i
g |Hi|g i
Ei0 (28)
One thus gets, i
Ei0 Eg E1 (29)
If one an show that
i Ei0 = E1 , then E1 is the exat ground state energy.
For the MG-hain, the sub-Hamiltonian Hi is
Hi =J
2(S i.
S i+1 +
S i+1.
S i+2 +
S i+2.
S i) (30)
9
-
There are N suh sub-Hamiltonians. One an easily verify that Ei0 = 3J8 andE1 = 3J4 N2 ( - 3J4 is the energy of a singlet and there are N2 VBs in 1 and2). From (29), one nds that the lower and upper bounds of Eg are equal and
hene 1 and 2 are the exat ground states with energy E1 = 3JN8 . Thereis no LRO in the two-spin orrelation funtion in the ground stae:
K2(i, j) =Szi S
zj
=
1
4ij 1
8|ij|,1 (31)
The four-spin orrelation funtion has o-diagonal LRO.
K4(ij, lm) =Sxi S
xj S
yl S
ym
= K2(ij)K2(lm)
+1
64|ij|,1|lm|,1exp(i(
i+ j
2 l +m
2)) (32)
Let T be the translation operator for unit displaement. Then
T1 = 2, T2 = 1 (33)
The states
+ =12(1 + 2),
=12(1 2) (34)
orrespond to momentum wave vetors k = 0 and k = . The exitationspetrum is not exatly known. Shastry and Sutherland [40 have derived the
exitation spetrum in the basis of `defet' states. A defet state has the wave
funtion
(p,m) = ...[2p3, 2p2]2p1[2p, 2p+1]...[2m2, 2m1]2m[2m+1, 2m+2]...(35)
where the defets (2p1 and 2m) separate two ground states. The two defetsare up-spins and the total spin of the state is 1. Similarly, the defet spins an
be in a singlet spin onguration so that the total spin of the state is 0. Beause
of PBC, the defets o
ur in pairs. A variational state an be onstruted
by taking a linear ombination of the defet states. The exitation spetrum
onsists of a ontinuum with a lower edge at J(52 2 |cosk|). A bound stateof the two defets an o
ur in a restrited region of momentum wave vetors.
The MG hain has been studied for general values J of the n.n.n. interation
[41. The ground state is known exatly only at the MG point = 12 . Theexitation spetrum is gapless for 0 < < cr( 0.2411). Generalizations ofthe MG model to two dimensions exist [42, 43. The Shastry-Sutherland model
[42 is dened on a square lattie and inludes diagonal interations as shown
in Figure 1. The n.n. and diagonal exhange interations are of strength J1and J2 respetively. For
J1J2
below a ritial value 0.7 , the exat ground state
onsists of singlets along the diagonals. At the ritial point, the ground state
10
-
hanges from the gapful disordered state to the AFM ordered gapless state. The
ompound SrCu2(BO3)2 is well-desribed by the Shastry-Sutherland model[19. Bose and Mitra [43 have onstruted a J1 J2 J3 J4 J5 spin- 12model on the square lattie. J1, J2, J3, J4 and J5 are the strengths of the n.n.,
diagonal, n.n.n., knight's-move-distane-away and further-neighbour-diagonal
exhange interations (Figure 2). The four olumnar dimer states (Figure 3)
have been found to be the exat eigenstates of the spin Hamiltonian for the
ratio of interation strengths
J1 : J2 : J3 : J4 : J5 = 1 : 1 :1
2:1
2:1
4(36)
It has not been possible as yet to prove that the four olumnar dimer states are
also the ground states. Using the method of `divide and onquer', one an only
prove that a single dimer state is the exat ground state with the dimer bonds
of strength 7J . The strengths of the other exhange interations are as speiedin (36). For a 4 4 lattie with PBC, one an trivially show that the four CDstates are the exat ground states.
C. The Aek-Kennedy-Lieb-Tasaki (AKLT) model [44
We have already disussed the LSM theorem, the proof of whih fails for
integer spin hains. Haldane [12, 13 in 1983 made the onjeture, based on
a mapping of the HAFM Hamiltonian, in the long wavelength limit, onto the
nonlinear model, that integer-spin HAFM hains have a gap in the exitation
spetrum. The onjeture has now been veried both theoretially and exper-
imentally [45. In 1987, AKLT onstruted a spin-1 model in 1d for whih the
ground state ould be determined exatly [44. Consider a 1d lattie, eah site
of whih is o
upied by a spin-1. Eah suh spin an be onsidered to be a
symmetri ombination of two spin-
12 's. Thus, one an write down
++ = |++ , Sz = +1 = | , Sz = 1+ =
12(|++ |+ , Sz = 0
+ = + (37)
where `+' (`') denotes an up (down) spin.AKLT onstruted a valene bond solid (VBS) state in the following man-
ner. In this state, eah spin-
12 omponent of a spin-1 forms a singlet (valene
bond) with a spin-
12 at a neighbouring site. Let
(, = + or ) be the
antisymmetri tensor:
++ = = 0, + = + = 1 (38)A singlet spin onguration an be expressed as
12 | , summation over
repeated indies being implied. The VBS wave funtion (with PBC) an be
written as
11
-
|V BS = 2N2 11122223 .....iiii+1NN N1 (39)
|V BS is a linear superposition of all ongurations in whih eah Sz = +1 isfollowed by a Sz = 1 with an arbitrary number of Sz = 0 spins in betweenand vie versa. If one leaves out the zero's, one gets a Nel-type of order. One
an dene a non-loal string operator
ij = Si exp(ij1l=i+1
Sl )Sj , ( = x, y, z) (40)
and the order parameter
Ostring = lim|ij|ij
(41)
The VBS state has no onventional LRO but is haraterised by a non-zero
value
49 of O
string. After onstruting the VBS state, AKLT determined the
Hamiltonian for whih the VBS state is the exat ground state. The Hamiltonian
is
HAKLT =i
P2(S i +
S i+1) (42)
where P2 is the projetion operator onto spin 2 for a pair of n.n. spins. The
presene of a VB between eah neighbouring pair implies that the total spin
of eah pair annot be 2 (after two of the S = 12 variables form a singlet, theremaining S = 12 's ould form either a triplet or a singlet). Thus, HAKLTating on |V BS gives zero. Sine HAKLT is a sum over projetion operators,the lowest possible eigenvalue is zero. Hene, |V BS is the ground state ofHAKLT with eigenvalue zero. The AKLT ground state (the VBS state) is spin-
disordered and the two-spin orrelation funtion has an exponential deay. The
total spin of two spin-1's is 2, 1, 0. The projetion operator onto spin j for a
pair of n.n. spins has the general form
Pj(S i +
S i+1) =
l 6=j
[l(l + 1)S 2
][l(l + 1) j(j + 1)] (43)
where
S =
S i +
S i+1 . For the AKLT model, j = 2 and l = 1, 0 . From (42)
and (43),
HAKLT =i
[1
2(S i.
S i+1) +
1
6(S i.
S i+1)
2 +1
3
](44)
The method of onstrution of the AKLT Hamiltonian an be extended to higher
spins and to dimensions d > 1. The MG Hamiltonian (apart from a numerial
prefator and a onstant term) an be written as
12
-
H =i
P 32(S i +
S i+1 +
S i+2) (45)
The S = 1 HAFM and the AKLT hains are in the same Haldane phase, har-aterised by a gap in the exitation spetrum. The physial piture provided
by the VBS ground state of the AKLT Hamiltonian holds true for real systems
[46. The exitation spetrum of HAKLT annot be determined exatly. Arovas
et al. [47 have proposed a trial wave funtion
|k = N 12Nj=1
eikjSj |V BS , = z,+, (46)
and obtained
(k) =k |HV BS | kk | k =
25 + 15cos(k)
27(47)
The gap in the exitation spetrum = 1027 at k = . Another equivalent wayof reating exitations is to replae a singlet by a triplet spin onguration [48.
3 Spin Ladders
A. Undoped ladders
In the last Setion, we disussed some exat results for interating spin sys-
tems. The powerful tehnique of BA was desribed. The BA annot provide
knowledge of orrelation funtions. There is another powerful tehnique for 1d
many body systems known as bosonization [49 whih enables one to alu-
late various orrelation funtions for 1d systems. After the disovery of high-
Tc uprate systems, the study of 2d AFMs aquired onsiderable importane.
There are, however, not many rigorous results available for 2d spin systems.
Ladder systems interpolate between a single hain (1d) and the square lattie
(2d) and are ideally suited for the study of the rossover from 1d to 2d. Consider
a two-hain spin ladder (Figure 4) desribed by the AFM Heisenberg exhange
interation Hamiltonian
HJJR =ij
JijS i.
S j (48)
The n.n. intra-hain and the rung exhange interations are of strength J and
JR respetively. When JR = 0, one obtains two deoupled AFM spin hains for
whih the exitation spetrum is known to be gapless. Dagotto et al. [50 derived
the interesting result that the lowest exitation spetrum is separated by an
energy gap from the ground state. The result is easy to understand in the simple
limit in whih the exhange oupling JR along the rungs is muh stronger than
the exhange oupling J along the hains. The intra-hain oupling may thus be
treated as perturbation. When J = 0 , the exat ground state onsists of singlets
13
-
along the rungs, eah singlet having the spin onguration
12[| |]. The
ground state energy is 3JRN4 , where N is the number of rungs in the ladder.In rst order perturbation theory, the orretion to the ground state energy is
zero. The ground state has total spin S = 0. A S = 1 exitation may be reatedby promoting one of the rung singlets to a S = 1 triplet. A triplet has the spin
onguration | (Sz = +1), 1
2| + (Sz = 0 ) and | (Sz = 1). A
triplet osts an exhange energy equal to JR . The weak oupling along the
hains gives rise to a band of propagating S = 1 magnons with the dispersionrelation
(k) = JR + Jcosk (49)
in rst order perturbation theory (k is the wave vetor). The spin gap, dened
as the minimum exitation energy is given by
SG = () (JR J) (50)The two-spin orrelations deay exponentially along the hains showing that the
ground state is a quantum spin liquid (QSL). As the rung exhange oupling JRdereases, one expets that the spin gap will also derease and ultimately be-
ome zero at a ritial value of JR. Barnes et al. [51, however, put forward the
onjeture that SG > 0 for allJRJ
> 0, inluding the isotropi limit JR = J . Avariety of numerial tehniques like exat diagonalization of nite-sized ladders
[50, Quantum Monte Carlo (QMC) simulations [52 and density-matrix renor-
malization group (DMRG) [53 have veried the onjeture. We now onsider
the ase of an n-hain ladder. A surprising fat emerging out of several theoret-
ial studies [54, 55 is: the exitation spetrum has spin gap (is gapless) when n
is even (odd). In the rst (seond) ase, the two-spin orrelation funtion has
an exponential (power-law) deay. For odd n, the ladder has properties similar
to those of a single hain. The strong oupling limit (JR J) again provides aphysial piture as to why this is true. When n is even, the S =
12 spins along
a rung ontinue to form a singlet ground state. Hene the reation of a S =1
exitation requires a nite amount of energy as in the ase of the two-hain
ladder. The gap should derease as n inreases so that the gapless square lat-
tie limit is reahed for large n. When n is odd, eah rung onsists of an odd
number of spins, eah of magnitude
12 . The inter-rung (intra-hain) oupling J
generates an eetive interation between the S = 12 rung states, whih beauseof rotational invariane, should be of the Heisenberg form with an eetive ou-
pling Jeff setting the energy sale. The equivalene of an odd-hain ladder to
the single Heisenberg hain leads to a gapless exitation spetrum. Rojo [56
has given a rigorous proof of the gaplessness of the exitation spetrum when
n is odd. Khveshhenko [57 has shown that for odd-hain ladders, a topolog-
ial term governing the dynamis at long wavelengths appears in the eetive
ation, whereas, it exatly anels for even-hain ladders. The topologial term
has similarity to the one that auses the dierene between integer and half-
odd integer spin hains. In the rst ase, the spin exitation spetrum has the
14
-
well-known Haldane gap. In the latter ase, the LSM theorem shows that the
exitation spetrum is gapless. Ghosh and Bose [58 have onstruted an n-hain
spin ladder model for whih the exat ground state an be determined for all
values of n. For n even (odd), the exitation spetrum has a gap (is gapless).
This is true even for large n, thus the square lattie limit annot be reahed in
the model. Thermodynami properties of the S = 12 two-hain ladder have beenrst studied by Troyer et al. [59. Using a quantum transfer matrix method,
they obtained reliable results down to temperature T 0.2J . The AFM or-relation length AFM has been found to be 3-4 lattie spaings. The magneti
suseptibility (T ) shows a rossover from a Curie-weiss form, (T ) = CT+at
high temperature to an exponential fall-o, (T ) eSGTT
as T 0. Thefall-o is a signature of a nite spin gap SG. Frishmuth et al. [60, using apowerful loop algorithm, have alulated the magneti suseptibility and found
evidene for the gapped (gapless) exitation spetrum in the ase of an even
(odd)-hain ladder.
A major interest in the study of ladder systems arises from the fat that
there is a large number of experimental realizations of ladder systems. A om-
prehensive review of major experimental systems is that by Dagotto [61. We
disuss here only a few interesting ladder systems. Hiroi et al. [62 were the
rst to synthesize the family of layer ompounds Srn1Cun+1O2n. Rie et al.[54 subsequently reognized that these ompounds ontained weakly-oupled
ladders of
n+12 hains. For n =3 and 5, respetively, one gets the two-hain and
three-hain ladder ompounds. Azuma et al. [63 have determined the tem-
perature dependene of the magneti suseptibility in these ladder ompounds
experimentally. A spin gap is indiated by the sharp fall of (T )for T < 300Kin the two-hain ladder ompound SrCu2O3 . The magnitude of the spin gap is
SG 420K. This is approximately in agreement with the theoretial result ofSG J2 , if an exhange oupling J 1200K is assumed. For the three-hainladder ompound Sr2Cu3O5 , Azuma et al. found that (T ) approahes a on-stant as T 0, , as expeted for the 1d Heisenberg AFM hain. Muon spinrelaxation measurements by Kojima et al. [64 shows the existene of a long
range ordered state with Nel temperature TN = 52 K, brought about by the
interlayer oupling. No sign of long range ordering was found in the two-hain
ladder ompound, onrming the dierene between odd and even hain ladders.
The ompound LaCuO2.5 is formed by an array of weakly interating two-hain
ladders [65. The evidene of spin-liquid formation at intermediate tempera-
tures (onrmed by the existene of a spin gap) and an ordered Nel state at
low temperatures, shows that the spin singlet state is in lose ompetition with
a Nel state. Spin ladders, belonging to the organi family of materials, have
also been synthesized. A reent example is the ompound (C5H12N)2CuBr4[66. This ompound is a good example of a strongly oupled (
JRJ 3.5 ) ladder
system. The phase diagram of the AFM spin ladder in the presene of an ex-
ternal magneti eld is partiularly interesting. In the absene of the magneti
eld and at T = 0, the ground state is a quantum spin liquid with a gap in theexitation spetrum. At a eld Hc1 , there is a transition to a gapless Luttinger
15
-
liquid phase (gBHc1 = SG , the spin gap, B is the Bohr magneton and g theLand splitting fator). There is another transition at an upper ritial eld Hc2to a fully polarised FM state. Both Hc1 and Hc2 are quantum ritial points.
The phase transitions that o
ur at these points are quantum phase transitions
as they o
ur at T = 0. At a quantum ritial point, the system swithes fromone ground state to another. The transition is brought about by hanging a
parameter (magneti eld in the present example) other than temperature. At
small temperatures, the behaviour of the system is determined by the rossover
between two types of ritial behaviour: quantum ritial behaviour at T = 0and lassial ritial behaviour at T 6= 0. Quantum eets are persistent inthe rossover region at small nite temperature and suh eets an be probed
experimentally. Refs. [67, 68, 69 give extensive reviews of quantum ritial
phenomena. In the ase of the ladder system (C5H12N)2CuBr4, the magneti-zation data, obtained experimentally, exhibit universal saling behaviour in the
viinity of the ritial elds, Hc1 and Hc2 . We remember that in the viinity of
ritial points, physial quantities of a system exhibit saling behaviour. Quan-
tum spin systems provide several examples of quantum phase transitions and
organi spin ladders are systems whih provide experimental testing grounds of
theories of suh transitions. For inorgani spin ladder systems, the value of Hc1is too high to be experimentally a
essible.
B. Frustrated spin ladders
Bose and Gayen [70 have studied a two-hain spin ladder model with frus-
trated diagonal ouplings (Figure 5, frustrated spin systems are dened in Se-
tion 4). The intra-hain and diagonal spin-spin interations are of equal strength
J . The exhange interations along the rungs are of strength JR . It is easy
to show that for JR 2J , the exat ground state onsists of singlets along therungs with the energy Eg = 3JRN4 where N is the number of rungs. Xian [71pointed out that the Hamiltonian of the frustrated ladder model an be written
as
H = JRi
S 1i.
S 2i + J
i
P i.
P i+1 (51)
where,
Pi =
S 1i +
S 2i , represents a omposite operator at the i-th rung
and `1' and `2' refer to the lower and upper hains respetively. Due to the
ommutativity of the rung interation part of the Hamiltonian with the seond
part, the eigenstates of the Hamiltonian an be desribed in terms of the total
spins of individual rungs. The energy eigenvalue for the state with singlets on all
the rungs is Esg = 3JRN4 . The seond term in the Hamiltonian (Eq.(51)) doesnot ontribute in this ase. If the two rung spins form a triplet, the seond term
is equivalent to the Hamiltonian of a spin-1 Heisenberg hain with a one-to-one
orrespondene between a rung of the ladder and a site of the S = 1 hain.Beause the two parts of H ommute, the eigenvalue, when the rung spins form
a triplet is
ETg = (Je0 +JR
4)N (52)
16
-
where e0 = 1.40148403897(4) is the ground state energy/site of the spin 1Heisenberg hain. Comparing the energy ETg with the energy E
sg of the rung
singlet state, one nds that as long as
JRJ> (JR
J)c = e0 , the latter state is the
exat ground state. At the ritial value (JRJ)c , there is a rst order transition
from the rung singlet state to the Haldane phase of the S = 1 hain. Thelowest spin exitation in the rung singlet state an be reated by replaing a
rung singlet (S = 0) by a triplet (S = 1). The triplet exitation spetrumhas no dynamis. In a more general parameter regime, i.e., when the intra-
hain exhange interation is not equal in strength to the diagonal exhange
interation, the ground and the exited states an no longer be determined
exatly. In this ase, one takes reourse to approximate analytial and numerial
methods. Kolezhuk and Mikeska [72 have onstruted a lass of generalised
S = 12 two-hain ladder models for whih the ground state an be determinedexatly. The Hamiltonian H is a sum over plaquette Hamiltonians and eah
plaquette Hamiltonian ontains various two-spin as well as four-spin interation
terms. They have further introdued a toy model, the Generalised Bose-Gayen
(GBG) model whih has a rih phase diagram in whih the phase boundaries
an be determined exatly. Reently, some integrable spin ladder models with
tunable interation parameters have been introdued [73, 74, 75. The integrable
models, in general, ontain multi-spin interation terms besides two-spin terms.
C. Doped spin ladders
A major reason for the strong researh interest in ladders is that doped
ladder models are toy models of strongly orrelated systems. The most well-
known examples of the latter are the high-Tc uprate systems. As already
mentioned in the Introdution, these systems exhibit a rih phase diagram as
a funtion of the dopant onentration. Doping eetively replaes the spin-
12
's assoiated with the Cu2+ ions in the CuO2 planes by holes. The holes are
mobile in a bakground of antiferromagnetially interating Cu spins. Also,
due to strong Coulomb orrelations, the double o
upany of a site by two
eletrons, one with spin up and the other with spin down, is prohibited. This
is a non-trivial many body problem beause it involves a ompetition between
two proesses: hole deloalization and exhange energy minimization. A hole
moving in an antiferromagnetially ordered spin bakground, say, the Nel state,
gives rise to parallel spin pairs whih raise the exhange interation energy of the
system. The questions of interest are: whether a oherent motion of the holes
is possible, whether two holes an form a bound state (in the superonduting
(SC) phase of the doped uprates, harge transport o
urs through the motion
of bound pairs of holes), the development of SC orrelations, the possibility
of phase separation of holes et. For the uprates, a full understanding of
many of these issues is as yet laking (see [76 for a reent review of high-Tcsuperondutivity). The doped ladders are simple model systems in whih the
onsequenes of strong orrelation an be studied with greater rigour than in
the ase of the struturally more omplex uprate systems. Reent experimental
evidene [61 suggests that some phenomena are ommon to ladder and uprate
systems. The study of ladder systems is expeted to provide insight on the
17
-
ommon origin of these phenomena. Some ladder ompounds an be doped
with holes. Muh exitement was reated in 1996 when the ladder ompound
Sr14xCaxCu24O41 was found to beome SC under pressure at x = 13.6. Thetransition temperature Tc 12K at a pressure of 3 GPa. As in the ase ofSC uprate systems, holes form bound pairs in the SC phase of ladder systems.
The possibility of binding of hole pairs in a two-hain ladder system was rst
pointed out by Dagotto et al. [50. The strongly orrelated doped ladder system
is desribed by the t-J Hamiltonian
HtJ = ij,
tij(C+iCj +H.C.) +
ij
Jij(S i.
S j 1
4ninj) (53)
The C+iand Ci are the eletron reation and annihilation operators whih at
in the redued Hilbert spae (no double o
upany of sites).
C+i = C+i(1 ni)
Ci = Ci(1 ni) (54)
is the spin index and ni , nj are the o
upation numbers of the i-th and j-th
sites respetively. The term proportional to ninj is often dropped. The rst
term in Eq.(53) desribes the motion of holes with hopping integrals tR and t for
motion along the rung and hain respetively. In the onventional t J laddermodel, i and j are n.n. sites. The seond term (minus the 14ninj term) isthe usual AFM Heisenberg exhange interation Hamiltonian. The t J modelthus desribes the motion of holes in a bakground of antiferromagnetially
interating spins. In the undoped limit, eah site of the ladder is o
upied by a
spin-
12 and the tJ Hamiltonian redues to the AFM Heisenberg Hamiltonian.
Removal of a spin reates an empty site, i.e., a hole. A large number of studies
have been arried out on t J ladder models. These are reviewed in Refs.[61, 77, 78. We desribe briey some of the major results. A hole-doped
single AFM hain is an example of a Luttinger Liquid (LL) whih is dierent
from a Fermi liquid. The latter desribes interating eletron systems in higher
dimensions and at low temperatures. A novel harateristi of a LL is spin-
harge separation due to whih the harge and spin parts of an eletron (or
hole) move with dierent veloities and thus beome separated in spae. The
undoped two-hain ladder has a spin gap. This gap remains nite but hanges
disontinuously on doping. This is beause there are now two distint triplet
exitations (remember that the spin gap is the dierene in energies of the lowest
triplet exitation and the ground state). One triplet exitation is obtained
by exiting a rung singlet to a rung triplet as in the undoped ase. A new
type of triplet exitation is obtained in the presene of two holes. A lear
physial piture is obtained in the limit JR J . In this ase, the ground statepredominantly onsists of singlets along the rungs. On the introdution of a
hole, a singlet spin pair is broken and the hole exists with a free spin-
12 . In the
presene of two holes on two separate rungs, the two free spin-
12 's ombine to
18
-
give rise to an exited triplet state. The ground state is a singlet and onsists
of a bound pair of holes. The binding of holes an be understood in a simple
manner. Two holes loated on two dierent rungs break two rung singlets and
the exhange interation energy assoiated with two rungs is lost in the proess.
If the holes are loated on the same rung, the exhange interation energy of only
one rung is lost. If JR is muh greater than the other parameters of the system,
the holes preferentially o
upy rungs in pairs. As JR dereases in strength, the
hole bound pair has a greater spatial extent. The lightly doped ladder system
is not in the LL phase, i.e., no spin-harge separation o
urs. The system is in
the so-alled Luther-Emery phase with gapless harge exitations and gapped
spin exitations. A variety of numerial studies show that the hole pairs and the
spin gap are present even in the isotropi limit JR = J . Also, the relative stateof hole pairs has approximate d-wave symmetry with the pairing amplitude
having opposite signs along the rungs and the hains. The d-wave symmetry
is a feature of strong orrelation and is onsidered to be the symmetry of the
pairing state in the ase of uprate systems.
Bose and Gayen [70, 79, 80 have onstruted a two-hain tJ ladder modelwith frustrated diagonal ouplings. The intra-hain n.n. and the diagonal hop-
ping integrals have the same strength t. The other parameters have been dened
earlier. The speial struture of the model enables one to determine the exat
ground and exited states in the ases of one and two holes. The most signif-
iant result is an exat, analyti solution of the eigenvalue problem assoiated
with two holes in the innite t J ladder. The binding of holes has been ex-pliitly demonstrated and the existene of the Luther-Emery phase established.
For onventional t J ladders (the diagonal bonds are missing), the only ex-at results that have been obtained are through numerial diagonalization of
nite-sized ladders. Derivation of exat, analytial results in this ase has not
been possible so far. The reason for this is that as a hole moves in the anti-
ferromagnetially interating spin bakground, spin exitations in the form of
parallel spin pairs are generated. Proliferation of states with spin exitations
makes the solution of the eigenvalue problem extremely diult. In the ase of
the frustrated t J ladder model, there is an exat anellation of the terms
ontaining parallel spin pairs [80. Thus the hole has a perfet oherent motion
through the spin bakground. Frahm and Kundu [81 has onstruted an inte-
grable tJ ladder model and obtained the phase diagram. The model ontainsterms desribing orrelated hole hopping in hains whih may not be realizable
in real systems. Several studies have been arried out on the two-hain Hubbard
ladder as well as on multi-hain Hubbard and tJ ladders. Referenes of someof the studies may be obtained from [61.
4 Frustrated spin models in 2d
In Setions 2 and 3 we have disussed quasi-1d interating spin systems, namely,
spin hains and ladders. As already mentioned in the Introdution, the CuO2plane of the undoped uprate systems is a 2d AFM. The undoped uprates
19
-
exhibit AFM LRO below a Nel temperature TN . On the introdution of a
few perent of holes, the AFM LRO is rapidly destroyed leaving behind spin-
disordered states in the CuO2 planes. This fat has triggered lots of interest in
the study of spin systems with spin disordered states as ground states. Frus-
trated spin models are ideal andidates for suh systems. To understand the
origin of frustration, onsider the AFM Ising model on the triangular lattie.
An elementary plaquette of the lattie is a triangle. The Ising spin variables
have two possible values, 1, orresponding to up and down spin orientations.An antiparallel spin pair has the lowest interation energy J . A parallel spinpair has the energy +J . In an elementary triangular plaquette, there are threeinterating spin pairs. Due to the topology of the plaquette, all the three pairs
annot be simultaneously antiparallel. There is bound to be at least one paral-
lel spin pair. The parallel spin pair may be loated along any one of the three
bonds in the plaquette and so the ground state is triply degenerate. The Ising
model on the full triangular lattie has a highly degenerate ground state suh
that the entropy/spin is a nite quantity. As a result, the system never orders
inluding at T = 0 . Frustration o
urs in the system sine all the spin pairinteration energies annot be simultaneously minimised. On the other hand,
onsider the AFM Ising model on the square lattie. All the four spin pairs
in an elementary square plaquette an be made antiparallel and so there is no
frustration. The system exhibits magneti order below a ritial temperature.
If one of the spin pair interations in eah elementary square plaquette is FM
and the rest AFM, frustration o
urs in the square lattie spin system. A spin
system with mixed FM and AFM interations is frustrated if the sign of the
produt of exhange interations around an elementary plaquette is negative.
In the ase of a purely AFM model, frustration o
urs if the number of bonds
in an elementary plaquette of the lattie is odd. Examples of suh latties in 2d
are the triangular and kagom latties. In 3d, the pyrohlore lattie, the elemen-
tary plaquette of whih is a tetrahedron provides an example. A spin system
is also frustrated due to the presene of both n.n. as well as further-neighbour
interations. Consider AFM n.n. as well as n.n.n. interations between a row
of three Ising spins. Again, all the three spin pairs annot simultaneously be
made antiparallel.
Let us now treat the spins as lassial vetors (S ). For AFM spin-spininteration, the lowest energy is ahieved for an antiparallel spin onguration.
In the lassial limit, the spins on a bipartite lattie are ordered in the AFM
Nel state. On a non-bipartite lattie, suh as the triangular lattie, the lassial
ground state represents a ompromise between ompeting requirements. In the
ground state, the spins form an ordered three-sublattie struture with 1200
between n.n. spins on dierent sublatties. The ground state of the lassial
Heisenberg model on the kagom lattie is, however, highly degenerate and
disordered. We now onsider the full quantum mehanial spin Hamiltonian and
ask the question how the lassial ground states are modied when quantum
utuations are taken into a
ount. In the ase of the triangular lattie, it is now
believed that the quantum mehanial ground state of the S = 12 HAFM modelhas AFM LRO of the Nel-type, i.e., quantum utuations do not destroy the
20
-
three-sublattie order of the lassial ground state. In the seond senario, when
the lassial ground state is highly degenerate and disordered, thermal/quantum
utuations selet a subset of states whih tend to inorporate some degree of
long range order. This is the phenomenon of `order from disorder' whih is
ounterintuitive sine order is brought about by utuations whih normally
have disordering eets. The lassial kagom lattie HAFM ground states
inlude both oplanar as well as nonoplanar spin arrangements and utuations
lead to the seletion of oplanar order. This kind of ordering is partiularly true
for large values of the spin S. As the magnitude of the spin is dereased towards
S = 12 , the quantum utuations inrease in strength. These utuations oftendestroy the ordered struture obtained for large S. The quantum mehanial
ground states of the S = 12 HAFM on the kagom and pyrohlore latties havebeen found to be spin disordered. Some reent referenes of frustrated magneti
systems are [82, 83. The triangular lattie S = 12 HAFM is the rst exampleof a spin model in whih frustration o
urs due to lattie topology [84. The
S = 12 HAFM model has also been studied on a partially frustrated pentagonallattie [? and a parameter region identied in whih the ground state has AFM
LRO of the Nel-type.
Two well-known examples of spin-disordered states are the quantum spin
liquid (QSL) and dimer or valene bond (VB) states. A QSL state is a spin
singlet with total spin S = 0 and has both spin rotational and translationalsymmetry. In a VB state, spin rotational symmetry is present but tanslational
symmetry is broken. In suh states pairs of spins form singlets whih are alled
VBs or dimers with the VBs being frozen in spae. A well-known example of a
QSL state is the resonating-valene-bond (RVB) state [84 whih is a oherent
linear superposition of VB states (Figure 6). The RVB state is the starting
point for the well-known RVB theory of high-Tc SC. Spin-disordered (no AFM
LRO as dened in Eq. (8)) states with novel order parameters are:
(a) Chiral states
In these states, the spins are arranged in ongurations haraterised by the
order parameter
i =S i.(
Si+x
Si+y
(55)
with the three spins belonging to one plaquette of the square lattie and x, y
denoting unit vetors in the x and y diretions respetively. The hiral state
breaks time reversal symmetry or a reetion about an axis (parity).
(b) Dimer states
These are the VB states in whih the VBs are frozen in spae. A well-known
example of suh states is the olumnar dimer (CD) states. In suh states,
the VBs are arranged in olumns. On the square lattie, four suh states are
possible. The order parameter of CD states is
Dl =(l)
S l.(
Sl+x
+ iSl+y
Slx i
Sly)
(56)
where the l-sites are even and (l) = +1(1) if both lx and ly are even (odd).The order parameter takes the values 1, i,1,i for the four CD states shown
21
-
in Figure 3.
() Twisted states:
At the lassial level (S ), the spins in the twisted state are arrangedin inommensurate strutures. These ongurations an be visualised as spins
lying in a plane and with a twist angle in some diretion. It is possible that
suh states survive the inlusion of quantum utuations. The order parameter
is vetorial in nature and is given by
Tl =S l (S l+x +
Sl+y
)
(57)
These states are alled spin nematis and are dierent from helimagnets in whih
both Tl and the spin-spin orrelation funtions show LRO.
(d) Strip or Collinear States
In a lassial piture, the spins are ferromagnetially ordered in the x diretion
and antiferromagnetially ordered in the y diretion. The onguration obtained
by rotating the previous one by
pi2 is also possible. The order parameter is given
by
Cl =S l.(
Sl+x
Sl+y
+Slx
Sly)
(58)
Cl takes the values 1,1 in the two dierent strip ongurations.The spin-disordered states deribed above are quantum-oherent states and
are haraterised by novel order parameters. The term `quantum paramagnet'
is often used to desribe suh states.
Examples of real frustrated systems are many [82, 83. The best studied ex-
perimental kagom system is the magnetoplumbite, SrCr8xGa4+xO19 . Thesystem onsists of dense kagom layers of S = 32 Cr ions, separated by dilute tri-angular layers of Cr. In a mean-eld theory of the HAFM, the high temperature
suseptibility is given by
=C
T + cw, T TN (59)
Here, the Curie onstant C =2Bp
2
3kB, where B is the Bohr magneton and,
p = g[s(s + 1)]12, g being the Land splitting fator governing the splitting of
the spin multiplet in a magneti eld. Also, cw is the Curie-Weiss temperature.
The Nel ordering temperature TN is dened experimentally using bulk probes.
One looks for singularities in either the spei heat C(T) or the temperature
derivative of the suseptibility (T ) . In the ase of non-frustrated systems,TN cw and in the seond ase, TN
-
Frustration orresponds to f > 1 . For the kagom AFM SrCr8Ga4O19 , f is ashigh as 150.
SrCrGaO displays unonventional low-T behaviour [86. One of these is
the insensitivity of the spei heat C(T ) to applied magneti elds H as largeas twie the temperature. Suseptibility measurements show the existene of a
gap SG in the triplet spin exitation spetrum. For T < SG, eSGkBT
.
The spei heat, however, does not derease exponentially for T < SG, i.e.,does not have a thermally ativated behaviour. It has a T 2 dependene. This
fat along with the experimental observation of insensitivity of C(T ) to externalmagneti eld have been explained by suggesting that a large number of singlet
exitations fall within the triplet gap [86 . Numerial evidene of suh exita-
tions has been obtained in the ase of the S = 12 HAFM on the kagom lattie[87. The number of suh exitations has been found to be (1.15)N whereN is the number of spins in the lattie. Mambrini and Mila [88 have reently
established that a subset of short-range RVB states aptures the spei low
energy physis of the kagom lattie HAFM and the number of singlet states
in the singlet-triplet gap is (1.15)N in agreement with the numerial results.The appearane of singlet states in the singlet-triplet gap ould be a generi
feature of strongly frustrated magnets. Other examples of suh systems are:
the S = 12 frustrated HAFM on the15 -depleted square lattie desribing the 2d
AFM ompound CaV4O9 [89, 90, the HAFM on the 3d pyrohlore lattie and
a 1d system of oupled tetrahedra [91. Bose and Ghosh [90 have onstruted a
frustrated S = 12 AFM model on the15depleted square lattie and have shown
that in dierent parameter regimes the plaquette RVB (PRVB) and the dimer
states are the exat ground states. In the PRVB state, the four-spin plaquettes
(Figure 7) are in a RVB spin onguration whih is a linear superposition of
two VB states. In one suh state, the VBs (spin singlets) are horizontal and
in the other state the VBs are vertial. In the dimer state, VBs or dimers
form along the bonds joining the four-spin plaquettes. Both the PRVB and
the dimer states are spin disordered states. The state intermediate between the
PRVB and the dimer states has AFM LRO. Both the PRVB and dimer phases
are haraterised by spin gaps in the exitation spetrum. For the unfrustrated
HAFM model on the
15 -depleted square lattie, Troyer et al. [92 have arried
out a detailed study of the quantum phase transition from an ordered to a dis-
ordered phase. In the ordered phase, the exitation spetrum is gapless. The
spin gap SG ontinuously goes to zero in a power-law fashion at the quan-tum ritial point separating a gapped disordered phase from a gapless ordered
phase. Chung et al. [93 have arried out an extensive study on the possible
paramagneti phases of the Shastry-Sutherland model [19. In addition to the
usual dimer phase, they nd the existene of a phase with plaquette order and
also a topologially ordered phase with deonned S = 12 spinons and helialspin orrelations. Takushima et al. [94 have studied the ground state phase
diagram of a frustrated S = 12 quantum spin model on the square lattie. Thismodel inludes both the Shastry-Sutherland model as well as the spin model on
the
15depleted square lattie as speial ases. The nature of quantum phase
23
-
transitions among the various spin gap phases and the magnetially ordered
phases has been laried.
Lieb and Shupp [95 have derived some exat results for the fully frustrated
HAFM on a pyrohlore hekerboard lattie in 2d. This lattie is a 2d version
of the 3d pyrohlore lattie. The ground states have been rigorously proved to
be singlets. The Lieb-Mattis theorem is appliable only for bipartite latties.
Hene, the proof for the pyrohlore hekerboard lattie is a new result. Lieb and
Shupp have further proved that the magnetization in zero external eld vanishes
separately for eah frustrated tetrahedral unit. Also, the upper bound on the
suseptibility is
18 in natural units for both T = 0 and T 6= 0 . Frustration an
also o
ur from a ompetition between exhange anisotropy and the transverse
eld terms as in the ase of the transverse Ising model. Moessner et al. [96
have shown that the transverse Ising model on the triangular (kagom) lattie
has an ordered (disordered) ground state.
Frustrated AFMs with short-range dimer or RVB states as ground states
have a gap in the spin exitation spetrum and the two-spin orrelation fun-
tion has an exponential deay as a funtion of the distane separating the spins.
A single branh desribes the S = 1 triplet exitation spetrum. In Setion 2we pointed out that in the ase of the S = 12 HAFM hain, a pair of spinons(eah spinon has spin S = 12 ) are the fundamental exitations. The lowestexitation spetrum is thus not a single branh of S = 1 magnon exitationsbut a ontinuum of sattering states with well-dened lower and upper bound-
aries. AFM ompounds in 2d, in general, have exitation spetra desribed
by S = 1 magnons. Anderson [84 suggested that a RVB state may supportpairs of spinons as exitations whih are deonned via a rearrangement of the
VBs. In this ase, an extended and highly dispersive ontinuum of exitations
is expeted. Reently, Coldea et al. [97 have investigated the ground state or-
dering and dynamis of the 2d S = 12 frustrated AFM Cs2CuCl4 using neutronsattering in high magneti elds. The dynami orrelations exhibit a highly
dispersive ontinuum of exited states whih are harateristi of the RVB state
and arise from pairs of S = 12 spinons. A reent paper `RVB Revisited' byP.W.Anderson [98 fouses on the relevane of the RVB state to desribe the
normal state of the CuO2 planes in the high-Tc uprate superondutors.
5 Conluding remarks
In Setions 1-4, a brief overview of low-dimensional quantum magnets, speially,
antiferromagnets has been given. The subjet of quantum magnetism has wit-
nessed an unpreedented growth in researh ativity in the last deade. This is
one of the few researh areas in whih rigorous theories an be worked out and
experimental realizations are not diult to nd. Coordination hemists and
material sientists have prepared novel materials and onstruted new applia-
tion devies. Experimentalists have employed experimental probes of all kinds
to rene old data and unover new phenomena. Theorists have taken reourse
to a variety of analytial and numerial tehniques to explain the experimen-
24
-
tally observed properties as well as to make new preditions. Powerful theorems
have been proposed and exat results obtained. This trend is still ontinuing
and will lead to further breakthroughs in materials, phenomena and tehniques
in the oming years. The study of doped magneti materials whih inlude
uprates, ladders and spin hains has aquired onsiderable importane in re-
ent times. The dopants an be magneti and nonmagneti impurities as well as
holes. To give one example, it has been possible to dope the spin-1 Haldane-gap
AFM ompound Y2BaNiO5 with holes. Neutron sattering experiments reveal
the existene of midgap states and an inommensurate double-peaked struture
fator [99. Several new experimental results on the Haldane-gap AFMs have
been obtained in the last few years [100 whih add to the rihness of phenom-
ena observed in magneti syatems. In this overview, only a few of the aspets
of quantum magnetism have been highlighted. Conventional 2d and 3d mag-
nets have not been disussed at all as a good understanding of these materials
already exists. We have not disussed reent advanes in material appliations
whih inlude materials exhibiting olossal magnetoresistane, moleular mag-
nets, nanorystalline magneti materials, magnetoeletroni devies (the study
of whih onstitutes the new subjet of spintronis) et. A report on some of
these developments as well as some reent issues in quantum magnetism may
be obtained from Ref. [101.
Referenes
[1 D. C. Mattis, The Theory of Magnetism I (Springer-Verlag 1981)
[2 L. J. de Jongh and A. R. Miedema, Adv. Phys. 23, 1 (1974)
[3 C. K. Majumdar and D. K. Ghosh, J. Math. Phys. 10, 1388 (1969)
[4 F. D. M. Haldane, Phys. Rev. Lett. 60, 635 (1988); B. S. Shastry, Phys. Rev.
Lett. 60, 639 (1988)
[5 I. Bose, S. Chatterjee and C. K. Majumdar, Phys. Rev. B 29, 2741 (1984)
[6 D. C. Mattis, The Theory of Magnetism II (Spriger-Verlag 1985)
[7 The Many-body Problem: An Enylopedia of Exatly Solved Models in
One Dimension ed. by D. C. Mattis ( World Sienti 1993 )
[8 M. Steiner, J. Villain and C. G. Windsor, Adv. Phys. 25, 87 (1976)
[9 E. H. Lieb and D. C. Mattis, J. Math. Phys. 3, 749 (1962)
[10 W. Marshall, Pro. R. So. London Ser. A 232, 48 (1955)
[11 E. H. Lieb, T. D. Shultz and D. C. Mattis, Ann. Phys. 16, 407 (1961)
[12 F. D. M. Haldane, Phys. Rev. Lett. 50, 1153 (1983); Phys. Lett. A 93, 464
(1983)
25
-
[13 I. Aek, J. Phys.: Condens. Matter 1, 3047 (1989)
[14 M. Oshikawa, M. Yamanaka and I. Aek, Phys. Rev. Lett. 78, 1984 (1997)
[15 K. Hida, J. Phys. So. Jpn. 63, 2359 (1994)
[16 D. C. Cabra, M. D. Grynberg, A. Honeker and P. Pujol, ond-mat/0010376
[17 W. Shiramura et al., J. Phys. So. Jpn. 67, 1548 (1998)
[18 I. Aek, Phys. Rev. B 37, 5186 (1988); G. Misguih and C. Lhuillier,
ond-mat/0002170
[19 H. Kageyama et al., Phys. Rev. Lett. 82, 3168 (1999); H. Kageyama et al.,
Phys. Rev. Lett. 84, 5876 (2000)
[20 N. D. Mermin and H. Wagner, Phys. Rev. Lett. 22, 1133 (1966)
[21 F. Dyson, E. H. Lieb and B. Simon, J. Stat. Phys. 18, 335 (1978); T.
Kennedy, E. H. Lieb and B. S. Shastry, J. Stat. Phys. 53, 1019 (1988)
[22 E. J. Neves and J. F. Perez, Phys. Lett. 114 A, 331 (1986)
[23 I. Aek, T. Kennedy, E. H. lieb and H. Tasaki, Commun. Math. Phys.
115, 447 (1988)
[24 H. Bethe, Z. Physik 71, 205 (1931); see also [7 for an English translation
of Bethe's paper
[25 J. B. Torrane and M. Tinkham, Phys. Rev. 187, 59 (1969)
[26 D. F. Nioli and M. Tinkham, Phys. Rev. B 9, 3126 (1974)
[27 J. des Cloizeaux and J. J. Pearson, Phys. Rev. 128, 2131 (1962)
[28 L. D. Faddeev and L. A. Takhtajan, Phys. Lett. 85A, 375 (1981); see also
C. K. Majumdar in Exatly Solvable Problems in Condensed Matter and Rel-
ativisti Field Theory ed. by B. S. Shastry, S. S. Jha and V. Singh (Springer-
Verlag, Berlin Heidelberg 1985)
[29 D. A. Tennant, T. G. Perring, R. A. Cowley and S. E. Nagler, Phys. Rev.
Lett. 70, 4003 (1993)
[30 L. A. Takhtajan in Exatly Solvable Problems in Condensed Matter and
Relativisti Field Theory ed. by B. S. Shastry, S. S. jha and V. Singh
(Springer-Verlag, Berlin Heidelberg 1985)
[31 I. Bose in Models and Tehniques of Statistial Physis ed. by S. M. Bhat-
taharjee (Narosa, New Delhi India 1997), p. 156
[32 I. Bose, ond-mat/0011262
26
-
[33 E. H. Lieb and W. Liniger, Phys. Rev. 130, 1605 (1963); E. H. Lieb, Phys.
Rev. 130, 1616 (1963); M. Gaudin, Phys. Letters 24A, 55 (1967); C. N. Yang,
Phys. Rev. Lett. 19, 1312 (1967); C. N. Yang and C. P. Yang, J. Math. Phys.
10, 1115 (1969)
[34 E. H. Lieb and F. Y. Wu, Phys. Rev. Lett. 20, 1445 (1968)
[35 B. Sutherland, Phys. Rev. Lett. 34, 1083 (1975); ibid 35, 185 (1975)
[36 B. Sutherland, Phys. Rev. B 12, 3795 (1975)
[37 N. Andrei, Phys. Rev. Lett. 45, 379 (1980)
[38 P. B. Wiegmann, J. Phys. C 14, 1463 (1981); V. M. Filyov, A. M. Tsvelik
and P. G. Wiegmann, Phys. Lett. 81 A, 175 (1981)
[39 P. A. Bares and G. Blatter, Phys. Rev. Lett. 64, 2567 (1990)
[40 B. S. Shastry and B. Sutherland, Phys. Rev. Lett. 47, 964 (1981)
[41 F. D. M. Haldane, Phys. Rev. B 25, 4925 (1982); K. Okamoto and K.
Nomura, Phys. Lett. A 169, 433 (1992)
[42 B. S. Shastry and B. Sutherland, Physia B 108, 1069 (1981)
[43 I. Bose and P. Mitra, Phys. Rev. B 44, 443 (1991); see also U. Bhaumik
and I. Bose, Phys. Rev. B 52, 12484 (1995)
[44 I. Aek, T. Kennedy, E. H. Lieb and H. Tasaki, Phys. Rev. Lett. 59, 799
(1987)
[45 J. P. Renard et al., Europhys. Lett. 3, 945 (1987); M. Date and K. Kindo,
Phys. Rev. Lett. 65, 1659 (1991)
[46 M. Hagiwara et al., Phys. Rev. Lett. 65, 3181 (1990)
[47 D. P. Arovas, A. Auerbah and F. D. M. Haldane, Phys. Rev. Lett. 60, 531
(1988)
[48 G. Fth and J. Slyom, J. Phys.: Condens. Matter 5, 8983 (1993)
[49 J. von Delft and H. Shoeller, Ann. der Physik 4, 225 (1998)
[50 E. Dagotto, J. Riera and D. J. salapino, Phys. Rev. B 45, 5744 (1992)
[51 T. Barnes, E. Dagotto, J. Riera and E. Swanson, Phys. Rev. B 47, 3196
(1993)
[52 M. Greven, R. J. Birgeneau and U.-J. Wiese, Phys. Rev. Lett. 77, 1865
(1996)
[53 S. R. white, R. M. Noak and D. J. Salapino, Phys. Rev. Lett. 73, 886
(1994)
27
-
[54 T. M. Rie, S. Gopalan and M. Sigrist, Europhys. Lett. 23, 445 (1993)
[55 S. Gopalan, T. M. Rie and M. Sigrist, Phys. Rev. B 49, 8901 (1994)
[56 A. Rojo, Phys. Rev. B 53, 9172 (1996)
[57 D. V. Khveshenko, Phys. Rev. B 50, 380 (1994)
[58 A. Ghosh and I. Bose, Phys. Rev. B 55, 3613 (1997)
[59 M. Troyer, H. Tsunetsugu and D. Wrtz, Phys. Rev. B 50, 13515 (1994)
[60 B. Frishmuth, B. Ammon and M. Troyer, Phys. Rev. B 54, R3714 (1996)
[61 E. Dagotto, Rep. Prog. Phys. 62, 1525 (1999)
[62 Z. Hiroi, M. Azuma, M. Takano and Y. Bando, J. Solid State Chem. 95,
230 (1991)
[63 M. Azuma, Z. Hiroi, M. Takano, K. Ishida and Y. Kitaoka, Phys. Rev.
Lett. 73, 3463 (1994)
[64 K. Kojima et al., Phys. Rev. Lett. 74, 2812 (1995)
[65 Z. Hiroi and M. Takano, Nature 377, 41 (1995); Z. Hiroi, J. Solid State
Chem. 123, 223 (1996)
[66 B. C. Watson et al., ond-mat/0011052
[67 S. L. Sondhi, S. M. Girvin, J. P. Carini and D. Shahar, Rev. Mod. Phys.
69, 315 (1997)
[68 A. Dutta and B. K. Chakrabarti, Indian J. Phys. 72B, 301 (1998)
[69 S. Sahdev, Quantum Phase Transitions (Cambridge Univ. Press, Cam-
bridge, 1999)
[70 I. Bose and S. Gayen, Phys. Rev. B 48, 10653 (1993)
[71 Y. Xian, Phys. Rev. B 52, 12485 (1995)
[72 A. K. Kolezhuk and H. J. Mikeska, Int. J. Mod. Phys. B 12, 2325 (1998)
[73 S. Albeverio, S.-M. Fei and Y. Wang, Europhys. Lett. 47, 364 (1999)
[74 Y. Wang, ond-mat/9901168
[75 M. T. Bathelor and M. Maslen, ond-mat/9907486
[76 J. Orenstein and A. J. Millis, Siene 288, 468 (2000)
[77 E. Dagotto and T. M. Rie, Siene 271, 618 (1996)
[78 T. M. Rie, Z. Phys. B 103, 165 (1997)
28
-
[79 I. Bose and S. Gayen, J. Phys.: Condens. Matter 6, L405 (1994)
[80 I. Bose and S. Gayen, J. Phys.: Condens. Matter 11, 6427 (1999)
[81 H. Frahm and A. Kundu, J. Phys.: Condens. Matter 11, L557 (1999)
[82 A. P. Ramirez, Ann. Rev. Mater. Si. 24, 453 (1994)
[83 R. Moessner, ond-mat/0010301
[84 P. W Anderson, Mater. Res. Bull. 8, 153 (1973); P. Fazekas and P. W.
Anderson, Phil. Mag. 30, 432 (1974)
[85 U. Bhaumik and I. Bose, Phys. Rev. B 58, 73 (1998)
[86 P. Sindzingre et al., ond-mat/9907220
[87 C. Waldtmann et al., Eur. Phys. J. B 2, 501 (1991)
[88 F. Mila, Phys. Rev. Lett. 81, 2356 (1998); M. Mambrini and F. Mila, ond-
mat/0003080
[89 M. Albreht, F. Mila and D. Poilblan, Phys. Rev. 54, 5856 (1996)
[90 I. Bose and A. Ghosh, Phys. Rev. B 56, 3154 (1997)
[91 M. Mambrini, J. Trbos and F. Mila, Phys. Rev. B 59, 13806 (1999); B.
Canals and C. Laroix, Phys. Rev. Lett. 80, 2933 (1998)
[92 M. Troyer and M. Imada, ond-mat/9703049
[93 C. H. Chung, J. B. Marston and S. Sahdev, ond-mat/0102222
[94 Y. Takushima, A. Koga and N. Kawakami, ond-mat/0103264
[95 E. H. Lieb and P. Shupp, math-ph/9908019
[96 R. Moessner, S. L. Sondhi and P. Chandra, Phys. Rev. Lett. 84, 4457 (2000)
[97 R. Coldea, D. A. Tennant, A. M. Tsvelik and Z. Tylzynski, ond-
mat/0007172
[98 P. W. Anderson, ond-mat/0104332
[99 G. Xu et al., Siene 289, 419 (2000); see also I. Bose and E. Chattopadhyay,
ond-mat/0012006
[100 A. Zheludev et al., ond-mat/0006350; see also I. Bose and E. Chattopad-
hyay, Phys. Rev. B 60, 12138 (1999); A. Zheludev et al., ond-mat/0104311
[101 R. R. P. singh, W. E. Pikett, D. W. Hone and D. J. Salapino, ond-
mat/0007086
29
-
Figure Captions
Figure 1. The Shastry-Sutherland model
Figure 2. Five types of interation in the J1 J2 J3 J4 J5 modelFigure 3. Four olumnar dimer states
Figure 4. A two-hain spin ladder
Figure 5. The two-hain frustrated spin ladder model
Figure 6. An example of a RVB state
Figure 7. The
15depleted square lattie.
30
-
ab
-
.......
+