HAL Id: jpa-00246490
https://hal.archives-ouvertes.fr/jpa-00246490
Submitted on 1 Jan 1992
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Self-consistent interaction of magnons in a frustrated quantum antiferromagnet
L. Bergomi, Th. Jolicœur
To cite this version:
L. Bergomi, Th. Jolicœur. Self-consistent interaction of magnons in a frustrated quantum antifer- romagnet. Journal de Physique I, EDP Sciences, 1992, 2 (4), pp.371-377. �10.1051/jp1:1992149�.
�jpa-00246490�
Classification Physics Abstracts 75.10
Short Communication
Self-consistent interaction of magnons in
aii~ustrated quantum antiferromagnet
L.
Bergomi
and Th.Jolicceur(*)
Service de Physique
Th60rique(**),
C-E- Saclay, F-91191 Gif-sur-Yvette Cedex, France(Recdved
19 December 1991, accepted 27 January1992)
Abstract. We study the magnon interactions by means of a self-consistent calculation in a frustrated quantum Heisenberg antiferromagnet with first and second neighbor exchange couplings. We have used the Hartree-Fock-Bogolyubov scheme at zero temperature. The effect of frustration on the shape of the spin-wave spectrum is
investigated:
we find a strong decrease of the spin-wave velocity. A clear deviation from spin wave theory is found when comparingwith finite lattice results.
The
discovery
of the remarkablemagnetic properties
ofhigh-Tc superconductors
has led to a reexamination ofantiferromagnetic
quantumspin
systemsmainly
in two dimensions.In the
insulating compounds La2Cu04
andYBa2Cu306+~
with ~ < 0A the copper atomsare endowed with local moments which order in the N6el pattern at low
enough
temperature.Inelastic neutron
scattering
studies have beenperformed
below and above the Ndel temperature TN for tridimensionalordering.
When T > TN it is verylikely
that the copperspins
realizethe square lattice quantum
Heisenberg antiferromagnet (QHAF)
withspins lI2
andexchange only
betweennearest-neighbors [I].
One of the crucial issues is the nature of the
ground
state.Magnetically
disorderedresonating-
valence-bond
(RVB)
states have beenproposed
[2] aspossible
candidates for theground
states of theHeisenberg antiferromagnets (HAF)
in two dimensions. However thespin lI2
HAFwith nearest
neighbor exchange
has been studiedby
severaltechniques
and there is now aconvincing
evidenceII,
3] that it hasan ordered
ground
state at leaston the square lattice.
The situation is far from
being
understood in the case of frustratedspin
systems. The addition of next-to-nearestneighbor antiferromagnetic couplings certainly
destabilizes N6el orderon a
square lattice and it has been
suggested
[4] thatbeyond
some critical values of thecouplings
it isreplaced by
a newmagnetically
disorderedphase.
There arepresently
several candidates for thisphase. Large
Ntechniques
[5, 6] suggest thatspin-Peierls
order can occur. This idea has(*) C-N-R-S- Research Fellow
(**)
Laboratoire de la Direction des Sdences de la Matibre du Commissariat hl'Energie
Atomique372 JOURNAL DE PHYSIQUE I N°4
received some support from numerical [7] and series [8] studies on a S
=
lI2
model with first and secondneighbor couplings (the
so-called Ji J2model).
Lancz6s studies [7, 9] have mea-sured a
susceptibility
suited to dimerized statesshowing
an enhancement nearJ2/Ji
"
lI2.
More exotic
proposals
have also been made such as chiral [10] orspin
nematic[Ill ground
states for
antiferromagnets
with strongenough
frustration.In this paper we
study
the Ndelphase
of the Ji J2 model in detailby
means of a self- consistentspin-wave
calculation at zero temperature. Thespin-wave velocity
determined in thisapproximation
can thus becompared
with the finite-sizescaling analysis
of small size clusters [9]. The data show a clear diference between the semi-classical resultsfavoring
a first-orderphase
transition [12] and the ab iniiio resultsshowing
a new behaviour forJ2IJI
> 0A. We stress that such a deviation can be best seenby looking
atquantities
different from the energy which stays very close to the semi-classical value as is clear fromprevious
works with similaranalytical
methods [13]. It should be clear that energy alone is notenough
to capture thephysics
of such a system: in the case of the unfrustrated lattice there are disorderedRVB-type
variational wavefunctions which are
extremely
close in energy to theground
state but with the wrong symmetry.Recent studies have revealed how close is the unfrustrated lattice from the semiclassical
spin-wave approximation [Ii.
Thisphenomenon
however does not survive with frustration: forJ2/Ji
> 0A there is evidence for an intermediatephase
that is not found in the S- c1o limit.
Let us focus on the so-called Ji J2 model:
H =
Ji ~j S; S;
+J2 ~ S;
Sk(I)
n-n- n-n-n-
The first sum runs over nearest
neighbor
links of a square lattice and the second over next-to- nearestneighbor
links. The S; arespin lI2
operators and thecouplings
Ji and J2 are bothpositive
I-e-antiferromagnetic.
It is convenient to introduce a=
J2/Ji
which is theunique
dimensionless parameter of this model. We first discuss the
properties
of the classical model related to(I)
when thespin
value S - c1o. For values of a smaller thanlI2
the classicalground
state is Ndel ordered I-e- theordering
wavevector is Q"
(x,x).
For a >lI2
there isa continuous
degeneracy
at the classical level which is liftedby
quantumfluctuations,
the so- called "order from disorder"phenomenon [14].
Theordering
selectedby
quantum fluctuations is definedby
the wavevectorQ
=(0, x)
or(x, 0).
It is thenpossible
tostudy
the model(I) by
aperturbative expansion
in powers of IIS starting
from the classicalground
states [4]: this is the standardspin-wave theory
d la Holstein-Primakoff. In the Ndelphase
a <lI2
the calculation atleading
order in the IIS expansion
leads to magnon modes withdispersion
relationgiven by:
Wk " Jl
/(I
tY
(1 7[))~ 7(, (1)
where 7k "
lI2 (cosk~
+ coskg)
and7[
=cosk~
coskg.
In the collinearphase
a >lI2
onefinds:
wk " Ji
~/(2a
+ cos kg )~(2a7[
+ cosk~)~, (2)
for the magnon
energies
around the(0,x)
state. In the classicaldescription,
there is nohysteresis
at the transition between the(x, x)
and(x, 0) (or (0, x)) phases.
The energy obtained for the JiJ2
model whenincluding zero-point
fluctuations of free magnons is discontinuous at a =1/2
and the derivatives are infinite on both sides of the transition. Thestaggered
magnetization
< S~ > whencomputed
at the same order ofapproximation
has adivergent
correction at a =
1/2.
This has beeninterpreted
as thesignature
of aspin-liquid ground
state[4].
However such behavior is
partly
an artefact of thesimple approximation
ofnon-interacting
magnons. At the classical transition
point
a =1/2
thedispersion
relation forspin
waves softens: entire lines of zero modes appear in the Brillouin zone. Since this is more than we expect from the Goldstonetheorem,
thisdegeneracy
is accidental and should be lifted when the interactions betweenspin-waves
are taken into account. Moreover, theexpansion
parameter of the naivespin-wave approximation
grows in theneighborhood
of a=
1/2.
Its relevance becomes thus doubtful.To
investigate
the effect of magnon interaction we use the self-consistent Hartree-Fock-Bogolyubov (HFB)
scheme at zero temperature. This isexactly equivalent
to the so-called modifiedspin-wave theory (MSW)
introducedby
Takahashi [15]. In his scheme one enforces theMermin-Wagner
constraint < S~ >= 0
by
the use of a chemicalpotential (Lagrange multiplier).
At zero temperature this chemicalpotential
for bosons goes to zero and one is leftwith an
approximation
which issimply
the HFB scheme. We have used the Holstein-Primakov(HP) representation
ofspin
operatorsleading
to a bosonic Hamiltonian. It is convenient to usea
quantization
axis for thespin
operators which isaligned
with the classical minimum at each site so that there isonly
one kind of boson in theproblem.
Theexpansion
of the square roots in the HPrepresentation
isperformed keeping
four-bosons terms in theresulting
Hamiltonian:H =
NS~J(Q)
+ ~~j Bka)ak Ah (aka-k
+a)at ~)
2
~
2
+
~ ~ (2J(k4) J(k4
+
Q) J(k4 Q))
6(ki
k2k3 k4) (a)a2a3a4
+h-c-)
i~ (2J(ki)
+
J(ki
+Q)
+J(ki Q))
6(ki
+ k2 k3k4) (a)a)a3a4
+h-c-)
~~~
~~
k,
+
j ~ (J(k~
k~
Q)
+J(k~
k~ +Q))
6(ki
+ k~ k~k~) all
a~a~, k,
where h. c. means hermitian
conjugate.
In thisequation
thespin-wave
operator ahsatisfy
the bosonic commutationrelations,
theordering
wavevector is denotedby Q
so thatequation (3)
is true for both N6el and collinear
phase.
There are no three-bosons interactions since we restricted ourselves to these two commensurate structures. The k sums run over the two- dimensional Brillouin zone andJ(Q)
is the Fourier transform of theexchange coupling.
Thequadratic
part contains the twoquantities:
Ah =
2J(k) J(k
+Q) J(k Q)
and Bk =2J(k)
+J(k
+Q)
+J(k Q) 4J(Q). (4)
We then search a variational vacuum (0) defined
by
thedensity
matrix elements pk #(0(a)
ah(0)and Kk
"
(0(aka-k(0).
These matrix elements areconveniently parametrized
in terms of anangle bkl
p~ =
(cash
R~i)
and ~~=
)sinh
R~.(5)
Minimization of the energy leads to a set of self-consistent
equations
for the unknownik
Direct iteration can then beimplemented numerically.
Thisreproduces
thefindings
of reference [13]using
the formalism of Takahashi. It isamusing
to note that in the zero temperature limit(as
inRef.
[13])
there is no need of such a formalism since it reduces to the well-known self-consistent scheme for bosons.The Ndel state is stabilized
beyond
the classicalinstability point.
Thisinteresting phe-
nomenon of quantum stabilization of a state is also found in the
Schwinger
bosonapproach
374 JOURNAL DE PHYSIQUE I N°4
to the J2 Ji model
[16, Ii].
It alsohappens
in the case of helicalordering [18].
Both N4el and collinearphase
have metastable domains of existence. Thecrossing
ofenergies happens
for ac m 0.62 and at this
point
there is nodivergence
of thestaggered magnetization.
The collinearphase disappears
withdiverging
< S~ >- -c1oright
at the classical transitionpoint
a =
1/2.
However thisphenomenon happens only
within the metastableregion
where the Ndelphase
is lower in energy. In the HFB scheme there is thus a first-order transition between the two ordered states. Therefindings
can bequalitatively
obtained from asimple
treatment of the HFBequations
[12]: one uses the zeroth order contractions of the Bose operators in thequartic
terms of
equation (4)
to find the correction to theBogolyubov angles
ik. In such a treatmentone expects to
get
rid of the accidentaldegeneracies
seen in the lowest order calculation. Our calculation confirms that the behavior found in this way isgeneric.
The finite size
scaling analysis
for thestaggered magnetization performed
in reference [9]shows a behavior
quite
different from that obtained in the HFBanalysis.
The value < S~ >goes to zero
already
at o m 0A well before the semiclassical value.We have thus studied the
shape
of the one-magnon excitation in the IIFB scheme. This self-consistent method allows to find non-trivial modifications of thespin-wave
spectrum wk.We have concentrated on the Ndel
phase
of the Ji J2 model where thediscrepancy
between small-cluster results and the semiclassical results isstronger.
Infigure I,
we haveplotted
thespin-wave velocity
C as a function of a up to the first-order transitionpoint
ac m 0.6. There is an almost linear decrease from the unfrustrated value to a nonzero value of C at ac. This isvery different from the finite-size
scaling
results which show an almost constant value of C up to a transitionpoint
located around a m 0A.18
16
/
~
~~
~ 14
/
~
~
~
~ ]~~
~
Q
j
© 10~
j
z, u
~
T
~~~
o~i
,
j
-0 O 0 2 0 4 0 6
J z Ii
Fig. I. The spin wave velocity as a function of J2 from the self-consistent calculation. The points
are the results of the finite size scaling of reference [9]
To
gain
furtherunderstanding
in the behavior of the model underincreasing
frustration we havecomputed
thedispersion
relation wk for various values of thecoupling
a(see Figs. 2,3).
At-2
2
Fig. 2. The dispersion relation of magnons from the self-consistent calculation when there is no frustration I-e- J2 " 0. The energy along the vertical axh is plotted throughout the Brillouin zone
[-x,
x]~ in the basal plane.J2 = 0
(Fig.2)
theonly
remarkablephenomenon
is the presence of the usual Goldstone modes of theantiferromagnet
andnothing peculiar happens
at the zone boundaries. However withincreasing J2
the(0, x)
and(x, 0)
"collinear" modesbegan
to lower their energy. This trend shows up even moreclearly
very close to thepoint
ac G# 0.6(Fig.3)
where the "collinear modes"have smallest energy in the Ndel
phase.
It isinteresting
to note that up to the transitionpoint
no
complete softening
appears. There is thus nodivergence
in thestaggered magnetization
from the Ndel side of the transition. This behavior of the
dispersion
relation is incomplete
agreement with the fact that the transition of the Ji J2 model is first-order.Let us
emphasize
that the limit S- c1o of the self-consistent IIFB calculation
(or MSW)
is identical to the naivespin-wave theory.
If oneignores
magnon interactions then HFBsimply
reduces to the standard linearspin-wave theory.
Forexample
the Ndelphase
has a metastable domain up to a critical value [12] o =1/2
+O(logs/S) (m
0.7 forspin 1/2).
In the limit S - c1o this metastable domaindisappears
as iscorrectly
found within theSchwinger
bosonapproach [16, Ii].
Thephase diagram given by
Xu andTing
[13] is on the contraryincqrrect.
The quantum stabilization of the N4el state is thus
intrinsically
afinite
Sphenomenon.
In
conclusion,
we have studied thedispersion
relation of magnons in the Ji J2 modelby
means of a self-consistent calculation. Thespin
wavevelocity
and the energy wk of theone-magnon states has been obtained in the Ndel
phase.
We find the first-order transitionquoted
inprevious
works [6,12, 13, 16, Ii].
Wedisplay explicitly
the effect ofincreasing
frustration on the spectrum. lvhen
compared
with finite-size results [9] a clear deviation is found in the behavior of thespin-wave velocity (Fig.I).
ForJ2/Ji
> 0A thecomparison
of the self-consistent results with Lancz6s data isstrongly suggestive
of an intermediatephase,
whose nature remains up to now unsettled.
376 JOURNAL DE PHYSIQUE I N°4
-2
o
Fig.
3. The dispersion relation at the boundary of the N6el stateJ2/Ji
= 0.6 near the first-order transition. There is partial softening at the zone boundary of the (o, K), (K,0)
modes.Acknowledgements.
We thank A. V.
Chubukov,
T. Garel for various discussions and II. J. Schulz forinforming
usabout reference [9] before
publication.
References
[I] For general reviews see: Barnes T., Int. J. Mod. Phys. C2
(1991)
659;Manousakis E., Rev. Mod. Phys. 63
(1991)
1.[2] Anderson P.W., Science 235
(1987)
1196;Fazekas P. and Anderson P.W., Philos. Mag. 30
(1974)
423.[3] Reger D. and Young A-P-, Phys. Rev. 837
(1988)
5978;Liang S., Phys. Rev. 842
(1990)
6555;Dagotto
E. and Moreo A., Phys. Rev. Lett. 63(1989)
2148;Ruse D. and Elser V., Phys. Rev. Lett. 60
(1988)
2531.[4] Iofle L-B- and Larkin A-I-, Int. J. Mod. Phys. 82
(1988)
203;Chandra P. and Doucot B., Phys. Rev. 838
(1988)
9335;Einarsson T. and Johannesson H., Phys. Rev. 843
(1991)
5867.[5] Read N. and Sachdev S., Nucl. Phys. 8316
(1989)
609, Phys. Rev. Lett. 59(1989)
799.[6] Read N. and Sachdev S., Phys. Rev. Left. 66
(1991)
17?3.[?] Dagotto E. and Morec A., Phys. Rev. Lett. 63
(1989)
2148;Singh R.R.P. and Narayan R., Phys. Rev. Lett. 65
(1990)
1072.For a review see Dagotto E., Int. J. Mod. Phys. 85
(1991)
907.[8] Gelfand M-P-, Singh R-R-P- and Ruse D.A., Phys. Rev. B40
(1989)
10801;Sachdev S. and Bhatt R.N., Phys. Rev. 841
(1990)
4502.[9] Schulz H.J. and Ziman T-A-L-, Finite-size scaling study of a quantum frustrated
antiferromagnet,
Orsay preprint, to appear in Europhys. Lett.;See also Poilblanc D. et al., Phys. Rev. 843
(1991)
10970.[10] Kalmeyer V. and Laughlin R.B., Phys. Rev. Lent. 59
(1987)
2095;Wen X.-G., Wilczek F. and Zee A., Phys. Rev. 839
(1989)
11413.[11] Chandra P., Coleman P. and Larkin A-I-, Phys. Rev. Lett. 64
(1990)
88;Chandra P. and Coleman P., Int. J. Mod. Phys. 83
(1989)
1729.[12] Chubukov A-V-, Phys. Rev. 844
(1991)
392.[13] Barabanov A.F, and Starykh O-A-, JETP Lett. 51
(1990)
311;Xu J.H. and Ting C.S., Phys. Rev. 842
(1990)
6861;Nishimori H. and Saika Y., J. Phys. Soc. Jpn 59
(1990)
4454.[14] Shender E., Sov. Phys. JETP 56
(1982)
178;Chandra P., Coleman P. and Larkin A.I., Phys. Rev. Lent. 64
(1990)
88; J. Phys. Cond. Matter 2(1990)
7933;Pimpinelli A., Rastelli E. and Tassi A., J. Phys. Cond. Matter1
(1989)
2131;Moreo A., Dagotto E., Jolicoeur Th. and Riera J., Phys. Rev. 842
(1990)
6283.[Is]
Takahashi M., Phys. Rev. B40(1989)
2494.[16] Mila F., Poilblanc D. and Bruder C., Phys. Rev. 843
(1991)
7891.[17] Bruder C. and Mila F., Europhys. Lent, to appear
(1991).
[18] Harris A-B- and Rastelli E., J. Appl. Phys. 63