Journal of Physics: Condensed Matter

You may also like

- Stripe domains reorientation in
Quantum theory of an antiferromagnet on a ferromagnetic films with perpendicular
magnetic anisotropy
triangular lattice in a magnetic field Louis-Charles Garnier, Massimiliano
Marangolo, Mahmoud Eddrief et al.

- Reorientation of pentacene molecules

To cite this article: A V Chubukov and D I Golosov 1991 J. Phys.: Condens. Matter 3 69 from flat-lying to standing manners on a
surface-modified amorphous SiO2
substrate investigated by molecular
dynamics simulations
Susumu Ikeda
View the article online for updates and enhancements. - Flying over uneven moving terrain based
on optic-flow cues without any need for
reference frames or accelerometers
Fabien Expert and Franck Ruffier

This content was downloaded from IP address 84.236.58.101 on 14/02/2023 at 19:34

1. Phys.: Condens. Matter3(1991)69-82. Printed inthe UK

Quantum theory of an antiferromagneton a triangular

lattice in a magnetic field

A V Chubukov and D I Golosov

Institute for Physical Problems. USSR Academy of Sciences. 117334 ul. Kosygina 2,
Moscow, USSR

Received 9 February 1990

Abstract. The reorientation process in a magnetic field in two-dimensional isotropic and XY

quantum Heisenberg antiferromagnets is shown to occur through the intermediate phase
with unbroken continuous symmetry and constant magnetization equal to one third of the
saturation value. The same reorientation process is also found in the more complicated
classical models.

1. Introduction

In the last few years there has been renewed interest in the study of the possibility for
the disordered ground state to occur in S = 1/2 two dimensional (ZD)antiferromagnets
onatriangular lattice ( A F M T tostrongzero-pointvibrations[
)~~~ 1-31, Strictlyspeaking,
the exact answer whether this is possible or not is not yet known. but most investigators
now believe that long-range order, though strongly suppressed by fluctuations, does
exist in AFMT even for sufficiently small values of the site spin [4,5].
InthispaperwealsostudymAFMTbut wecarry outourinvestigationsinthepresence
of an external magnetic field. Interest in this problem stems from the fact that at T = 0
switchingon the magnetic field does not change the degree of continuous degeneracy in
the classical Heisenberg model [6].To see how this happens let us consider first the
situation in three dimensions (3D) (i.e. in the so-called antiferromagnet on a stacked
triangular lattice). Here the spin arrangement in the presence of the field is well estab-
lished: it is of the 'umbrella' type and the order parameter space is reduced from SO(3)
+
in the zero-field case to SO(2) X Z 2 for H 0 (SO(2) corresponds to a rotational
symmetry around the magnetic-field axis and Z 2 distinguishes left- and right-twisted
120" helicoids formed by the components perpendicular to the field). Due to the non-
collinearity of the spin structure the excitation spectra contains three low-energy modes
-
[7]: a Goldstone mode with w1 I k I associated with the breakingof S 0 ( 2 ) , a mode with
w2(k= 0) = ZpH, representing the precession of the total magnetic moment around the
field direction, and the third mode with energy w 3 = q 2 p H , where q = (xl - X I I ) / X I Iis
the anisotropy of susceptibilities. In the classical limit (S- w) ?J can beexpressed in
terms of the microscopic parameters as follows [SI:
q = (1 t sJ/J")-' (1)
where J and J" are in-plane and interplane exchange integrals. One can immediately

095>8984/91/010069 + 14 $03.50 0 1991 IOP Publishing Ltd 69

IO A V Chubukov and D I Golosov

Figure 1. Reorientation process in the magnetic

held in ZDHeisenberg AN on a triangular lattice.
Zero-point fluctuations stabilize the collinear
phase in [he finite region H , < H C Hi in [he
He H,.,/3 H = H,., A H,,t/3r H e HM vicinity of H,J3.

see that in the purely ?D case ( J " = 0), q = 0, that is, in addition to the Goldstone
mode associated with the SO(2) breaking there exists an 'accidental' gapless branch of
excitations. This is of course a reflection of the fact that the Heisenberg interaction for
the triad of classical spins can be expressed only in terms of magnetization vector
S, S,. + S2 .S, + S3.S , = t M 2 - constant M = S , +S2+S3
without drawing any of the antiferromagnetic vectors. According to simple calculations,
w3 remains gapless in all fields up to the saturation value (Usmt = IUS). Moreover, in
non-zero fields this mode turns out to be quadratic in k (the state of a triad of classical
'unit vectors is specified to an accuracy of the rotation around the field axis by three
equations for five angles). As a result, many exotic configurations have classically the
same ground state energy as the umbrella-like one and, hence, the type of reorientation
at T = 0 in a real quantum Heisenberg model must be selected by quantum fluctuations.
The same is also true for easy-plane systems, and the XY model serves here as a good
example [9-121. The order parameter space must normalIy be reduced from SO(2) x Zz
in the zero-field case to Z , x Z3 in the presence of a magnetic field 191. However, for
classicalspins the G'oldstone mode, associated with SO(2) breakingin a zero field, does
not acquire a gap in all fields up to the saturation value (again, the ground state of a triad
oftwo-component unitvectorsiscompletelyspecifiedbytwoequationsforthreeangles).
One way to lift the 'accidental' degeneracy in the XYmodel was proposed in [9.12]. It
was shown that at non-zero temperature the 'lacking at T = 0' condition for the angles
arises thus fixing the mode of reorientation. For T + 0 it ocCurS in conformity with figure
1 and is accompanied by a phase transition at H = H,,/3 when the spins of the two
sublattices align parallel to each other. Moreover, numerical experiments [9] indicate
that at non-zero temperatures the collinear phase survives in the finite range of magnetic
fields. This is very natural since in spite of the fact that the order parameter space is the
same (Z, x Z3) in low and high-field phases, it is easy to determine the parameter
distinguishing between them. This is a chirality vector 31, which for each elementary
spin triangle is a measure of proximity to a 120" structure:
X = (2/3fi)(St X S2 + S 2 x S3+ S, x S,). (2)
Evidently, 31 = 0 in the high-field phase when the spins of two sublattices are parallel.
Less information is known about the isotropic system, where the degeneracy of the
ground state is much stronger. Kawamura and Miyashita proposed [6] that at T # 0 the
process of reorientation for classical spins occurs in the same way as in the XY system,
that is, firstly, in the presence of the field all the spins remain in the same plane and,
secondly, the reorientation occurs via the intermediate collinear phase with unbroken
continuous symmetry. Numerical calculations [6] seem to confirm this scheme.
The aim of the present paper is firstly to show that quantum fluctuations also remove
the 'accidental' degeneracy both in Heisenberg andXYmodels and select the same type
of reorientation as d o the temperature fluctuations. The isotropic Heisenberg model
 Quantum theory of an anhferromagnet in a magneticfield 71

will beconsideredin section2and the resultsfor t h e n m o d e l will be reportedinsection

3. Note that in contrast to the case of non-zero temperature, the analysis of the zero-
point motion effects is performed analytically on the basis of I/S expansion. We also
note that this is not the only case when quantum fluctuations lift the accidental degener-
acy existingon theclassical level. The same holds in anumber of systemswith competition
between nearest and next-nearest neighbours exchange interactions [13-151. Finally,
we would like to mention that here one meets the phenomenon which Villain er a1 [16]
termed 'order from disorder', since if there were no gap generation then the ordering in
the ground state would be destroyed by quantum fluctuations due to the existence of the
quadratic in k mode in the bare spectrum.
Evidently, the 'accidental' degeneracy is to a great extent a peculiarity of the Hei-
senberg model. The complicationof the model removes the degeneracyevenfor classical
systems. In the exchange approximation this is achieved (for S > 1) by adding the
biquadratic coupling [17]. An analogous effect occurs in the four-sublattice anti-
ferromagnet UO, [18].We discuss this in more detail in section 4. The important
point is that though the accidental degeneracy is removed by biquadratic coupling, the
umbrella-like configuration peculiar to 3D systems occurs only for one sign of this
coupling, while for the other sign the non-trivial mode of reorientation as in figure 1,
i.e. via the intermediate collinear phase, becomes energetically favourable. Hence, the
adding of biquadratic coupling may lead to the same consequences as the account of
quantum 0uctuations.
The other way to remove the degeneracy (also available only for S > 1) is to add a
single-ion anisotropy of the easy-axis type [19]. It is remarkable that this additional
interaction also favours a planar arrangement and reorientation via an intermediate
collinear phase; but now, due to the fact that, in the presence of the easy-axis anisotropy,
reorientation always (even in three dimensions) starts from the planar arrangement
(antiferromagnetic phase [ Z O ] ) , the peculiarity of the ZD case reveals in the formal
divergency of the spin-flop field value [8]
H:, = 16J"DS2(1 + SJ/J")(l + %J/J") (3)
that is, the antiferromagnetic (planar) phase survives in all fields. The role of easy-axis
anisotropy is also considered in more detail in section 4.
A brief comparison with experiment is given in section 5.
A short version of the paper was published in [21].

2. Isotropic case

We start with an isotropic Heisenberg antiferromagnetic on a triangular lattice:

X = J x S , S , + A- 2 p H x S f . (4)
CA I

We presume that the classical 120" structure in zero field is not destroyed by quantum
fluctuations and we will explore a traditional spin wave approach, based on the 1/S
expansiont. The type of lattice requires us to introduce three bosonic fields. Doing this

t Numerically, the reduction of the sublattice magnetization in the leading order in l/Sis(S)/S = 1 - 0.261s.
72 A V Chubukoo and D I Golosov

with the help of the Holstein-Primakoff transformation we obtain the hosonic version
of the spin Hamiltonian.
We shall not present the calculations step by step since they are absolutely standard
though very cumbersome. In all cases the problem actually was to find the canonical
transformation diagonalizing the quadratic form for three bosonic fields. The cor-
responding transformation at zero field is presented as an example in appendix 1, I n the
followingwe shall instead list only the results.
Our first aim is to establish the sign of the anisotropy of susceptibilities q since it
determines the type of arrangement (umbrella-like or planar) in low fields. Direct
calculations lead to the following result in the leading order in:s’,1

where
E’ -
k - Ekt2n/3 f k’ =fk?Z;r/3 Ek + 4fk)(l- 2fk))”2
fk =a(vk + v-k) +
v k = R{exp(ik,) exp(ik,) + exp[-i(k, + k y ) ] } .
Here and below the ‘prime’symhol at 2 indicates that the summation is performed over
thewholeBrillouinzone (EL 1= 3 E k 1= N,whereNis the totalnumberofspins), and
the X and Y axes are directed towards the nearest neighbours (n, nv = . -U.
Trans-
formation to the extended zone scheme is possible since at H = 0 the spectrum has no
gaps at the zone boundaries.
Numerical calculations give
q -O.OS/S. (6)
The minussign meansthat at least at low fieldsquantumfluctuationsselect the planar
arrangement (though the configuration inside the plane is not yet specified). In order to
exclude definitely the possibility of an umbrella-like ground state we calculated and
compared the first quantum corrections to the ground state energies for umbrella-like
andplanar(as in figure 1)configurationsat H = H,,/3 = 6JS(fortrivial reasonsquantum
fluctuations do not renormalize the saturation field value). The results is
Epimac - Eumbieiia= -NJs2Q (7)
where

/ l k = - 4 + 10[V,1* +2E31Yk12+4Ej-3[Y2 +(vf)’]

and E, are the (positive) classical frequencies of three branches of excitations above the
collinear state in figure 1. The values E , , and -e3 are the roots of the following
equation:
e 3 - c 2 - ~ ( 1 - I V ~ ~ ~ ) + ~ - - ~ ~ V ~ ~ ~ + Y ~ + ( U ~ (8)
) ~ =
The result of numerical calculation is
Q = 0.13/S
that is, a planar arrangement is energetically favourable also at H = H,,/3 and, hence
most likely at all other fields.
 Quantum theory of an antiferromagnet in a magnericfield 73

The next step is to find the real in-plane arrangement. At low fields, the degree of
freedom associated with the rotation of a triad of spins inside the plane remains massless
if we restrict ourselves to the leading order in H , i.e. with energies, A E , of the order of
H 2 .Thus we have to compare the energies in the next order in a magnetic field. This was
done for two configurations: those as in figure 1 and a second where one of the spins of
a triad is parallel to the field direction. The result is that the first configuration is
energetically favourable, the difference in energies being
A E a (J/S)(H/J)3. (9)
Thus we tentatively conclude that quantum fluctuations select the same mode of
reorientation as do the temperature fluctuations. Below we present another argument
confirming this conclusion.
We anticipate that with A E as in equation (9) the gap in the low-energy mode, fa3,
will be proportional to (I q [AE/xl,)1/2cc (J/S)(IY/J)~/*
for a purely isotropic system.
Now we examine the situation near H,,,/3. Classically, the collinear configuration
exists only at a single field value. Meanwhile, the transitions at H - t Hs,,/3 from above
and from below are of completely different nature and there are no reasons to expect
both of them to occur at the same field value. The calculation of the lability points of
low- and high-field phases as those where the renormalizedvaluesof the angles between
the two spins of the triad and the magnetic field tend to zero confirms this suspicion: the
low-field phase becomes unstable at

while the high-field phase loses stability at

(here and below h = 2pH/(6SJ) = 3H/Hs,,). As expected, h z > 12,. In the intermediate
region, h , < h < h,, the collinear configuration is stable. The absence of broken con-
tinuoussym‘metry then impliesthat all theexcitationshave afinitegap. Direct calculation
of the spectrum with quantum corrections involved leads to the following result:
wt/(6JS) E h - h, w2/(6JS) 5 h w3/(6JS) 5 h2 - h. (12)
Note that the mode associated with the precession of the magnetic moment is identically
equal to 2 p H also for a quantum system [22]. As found in equation (12), the collinear
phase is really stable for h , < h < h 2 ,and the gaps obtained for two would-be Goldstone
modes are proportional to 1/S. The AFMR frequencies versus magnetic field are shown
schematically in figure 2.
The absence of continuous degeneracy of the ground state probably implies that the
magnetization, M , , remains constant inside the collinear phase since the exchange
Hamiltonian commutes with the z-projection of the total spin*. Classically, M , equals
one third of the saturation value at h = 1.We calculated the first quantum correction to
M , andascertainedthatit equalszero, that isthemagnetizationretainsitsclassicalvalue,

* This is not an exact statement since the discrete R, degeneracy of the ground state remains and thus the
renormalization of magnetization of the orderoiexp(-S) is not excluded in principle, but not expected in
reality [7].
74 A V Chubitkov and D I Golosoo

Figure 2. The anticipated behaviour of AFMR fre- Figure 3. The anticipated behaviour of longi-
quencies versus magnetic field in ID Heisenberg tudinal magnetization in ?D Heisenbcrg AFM on a
AFM on a triangular lattice. The broken lines triangular lattice. The plateau on the magnet-
denote the branchcswhich wouldbe gaplessin the ization curve results from the stabilization of the
ciassical treatment. collinear phase in the finite region of magnetic
fields due to zero-point motion.

pS/3, also in the quantum case. We believe this result to be true in all orders of
perturbation in l/S. The anticipated behaviour of magnetization is shown in figure 3.
We conclude thissection with a brief discussion of the more realisticquasi-2D system.
For ferromagnetic interplane interaction (this is the case for a metamagnet) the mode
of reorientation will evidently remain unchanged independently of the strength of
interplane exchange. The case of antiferromagnetic interplane interaction, I”, is less
trivial. Firstly, since all the excitations in the intermediate phase with constant mag-
netizarion do have finitegaps, this phase cannot be destroyed by smallperturbations. In
contrast,thesituation inlow fieldsisverysensitive to theswitchingofthenewinteractions
and whatever small J” is the reorientation will always start with a different planar
configuration which is antiferromagnetic in the direction perpendicular to the plane
[U. 241. The transition to a ?D planar arrangement (which is ferromagnetic in the
direction perpendicular to the plane) will occur when the gain in energy AE ( J / S ) ( H / -
J)) favouring a metamagnetic configuration reaches the energy difference associated
- -
with the interplane exchange, A E J”, i.e. when H J((J”/J)S)”’. Secondly, the
increase in J” will evidently change the sign of q and, hence, for not very small J”
the reorientation will start from the umbrella-like configuration. We cannot definitely
answer the question what will happen in higher fieldssince it depends on the correlations
between unknown numerical parameters?.

3. Easy-plane systems

As was pointed out in the introduction, the accidental degeneracy in the classical AFMT
does not disappear when we switch on the anisotropy favouring the spin arrangement in
the basal plane (now singled out initially and not as a result of spontaneous breaking of
symmetry) and direct the field along the plane,

t The discussion of quasi-?o properties in section 3 in (211was based on thr proposal that q may be identically
equal to zero in ID AFMT The direct Calculation of7 (see equation ( 6 ) ) does not confirm this proposal.
 Quantum theory of an antiferromagnet in a mngneticfield 75

For simplicity, we shall mainly restrict ourselves to the case oftheXYmodel:

We shall first assume that the reorientation again occurs in direct compliance with figure
1and then verify this by calculating the spectrum.
The classical picture of excitations above the state in figure 1 is rather simple: in
addition to the ‘accidental’masslessbranch there are twoexcitations with finitegaps, one
of them softens at the transition point H = H,,,/3. Trivial calculations give (h = 2pH/
61s = 3 H / H , , )
= 3V%S[3 F h(h + 2)]’” uJ3 =O O<h<l
(14)
02.3 *
3JS[3 ( f i / 2 ) ( h 4 - 10h2 + 27)’/z]’/2 0, =o l<hC3.
The analogous expressions for the case of AFMT with small single-ion anisotropy,
closer to the experiment, are presented in appendix 2 .
Our aim is to investigate the effect of zero-point vibrations. To do this we calculated
in the first order in 1/S the lability points for the low- and high-field phases. The critical
fields, h , and hz respectively, are again different, favouring the collinear ground state to
occur in the finite region of the magnetic field:
1 ’ 2
h , = l - - 2NS
E ( -E~- 1 - . e k )

hz-h --E’
1 ( 1- - ~ ~ ) = -’ x - l)2(&, +
1 (EX 1)
> 0.
I-NS ck Ns k &k

Here
(obviously -
1 ’ =1
Ek = (1 + +Kk)’/2
Vk
Ns k
+
Numerically, h , = 1 - 0.59/S; h , = 1 O.OS/S. Note, that in the XY case quantum
fluctuations also renormalize h,,,: h,,,/3 = 1 - 0.14/S.
The same stability boundaries were determined by studing the AFMR modes above
the collinear state. They happened to be positive, and soften separately at h = h , and
hz:

W I 5 6JS(h - hI)’/’ ( 6 J S / f i ) ( h z - h)‘fi. (16)

Correspondingly the lowest gaps in the collinear phase are proportional to S-’Iz. Note
that at low fields the gap, pickingout the configuration as in figure 1,will be for obvious
reasons proportional to ( J / d S ) ( H / J ) 3 / z The
. field dependencies of the AFMR fre-
quencies are presented schematically in figure 4.
Classically, the magnetization in the collinear phase is constant and equals one third
of the saturation value. In contrast to the Heisenberg system this is no longer true for
tbe quantum XYcase, since the Hamiltonian of the XY model, equation (13), does not
commute with the Xprojection of the total spin.
16 A V Chubukov and D I Golosou

Concluding this section we note that the planar case allows a simple explanation of
the difference between phase transitions at h = h , and h = h2. In fact, in both low- and
high-field phases theorderparameterspaceisZ2 x Z3(Z,disappears within thecolliiear
phase), but the nature of the Z2 element is quite different: in the low-field phase 22
reflects the breakdown of chiral symmetry while in high fields Z, arises as a result of the
breakdown of the usual king symmetry. The difference in the realizations of the order
parameter space evidently means that there are no physical reasons for both transitions
to occur at the same field value.

4. Other spin models

In this section we will show that the 'accidental' degeneracy in a magnetic field is to a
great extent a peculiarity of the Heisenberg model and that the unusual mode of
reorientation may become energetically favourable at T = 0 even on the classical level,
but for more genera1 Hamiltonians. In order to demonstrate this we will at first restrict
ourselves to a purely exchange system (see equation (4)), but will add the biquadratic
nearest-neighbour interaction:

Being of exchange origin, this term will not moderate the 120" structure in a zero
field. Nevertheless, for a triad of spins (classical unit vectors) this term will give rise to
the coupling between the vectors of ferro- and antiferromagnetism and, hence, lead to
non-zero q :
q = 3B. (18)
As seen from equation (18). the umbrella-like configuration is to be chosen for B > 0
while for negative B the planar arrangement becomes energetically favourable.
Determination of the arrangement within the plane demands comparison of the
energies in the next to leading order in H. Doing this we ascertain that the (expected)
configuration shown in figure 1has a minimal energy for B < 0.
We therefore conclude that biquadratic coupling with negative B should lead to the
same reorientation process as d o quantum (or temperature [6,9-121) fluctuations, the
role of l/S(or T/(JS))being played by IBI. For completeness, we present below the
expressions for the critical fields, AFMR frequencies and longitudinal magnetization
under the assumption that biquadratic coupling is small compared to the Heisenberg
one ( B e 1). Qualitatively, these expressions repeat those from section 2.
(i) The lability fields for low- and high-field phases are, correspondingly,
h i = 1 - 6/51 h , = 1 + 2181 (19)
while h,,, = 3(1 - 21 El). Note that ( h , + h2)/2 = h,,/3.
(ii) The exchange nature of the interaction forces one of the AFMR frequencies to be
identically equal to 2 p H , while the other two are given by the following expressions:
 Quanfum theory of an antiferromagnet in a magneticfield I1

Figure 4. The same as figure 2 but for the XY FigureS. The fielddependenceofthe longitudinal
model. The broken lines denote the branches and transverse magnetizations in ZD classical Hei-
which would be gapless in the classical treatment. wnberg AFM with easy-axis anisotropy on the tri-
angular lattice (see equations (27) and (28)).

The field dependencies given by equation (20) are similar to those presented in figure 2.

[1 lm
(iii) The longitudinal magnetization is evidently constant and equal to 2pS/3 within
the collinear phase. Outside this region it behaves as follows:
Sh + + + 2) 0 c: h <h
MI, 2p 7 1 + (IB1/2)(h2- 5 ) h2 < h < hsat (21)
h > A,,
Note, that complementary to the analysis of quantum effects the engaging of the
however small easy-plane anisotropy immediately changes the dependence of AFMR
frequencies on [ E l :at extremely low fields w3becomes proportional to 1 BI'/'h3P instead
of IBlh3/lZas in equation (20).
The other way to lift a degeneracy is to engage the single-ionanisotropy of easy-axis
type
6 X = -D (ST)* D > 0. (22)
I

This term breaks the 120"structure in a zero field and normally must produce a spin-Aop
-
field H, (DJ)'/', that is, the reorientation starts from a planar configuration and then
there would be a transition into the umbrella-like structure [20].The peculiarity of the
ZD case, already mentioned in the introduction, is that since q reduces to zero, the
calculated value of this field formally tends to infinity (see equation ( 3 ) ) and as a result
the antiferromagnetic (planar) phase with the spin arrangement as in figure 1survives
in all fields up to the saturation value (191.The different nature of the transitions when
approaching the collinear configuration from above and from below (though in both
cases the order parameter space is isomorphic to SO(2) x Z , ) is again apparent in a
78 A VChubukov and D I Golosou

finite stability region of the intermediate phase with only discrete symmetry Z3broken
[19]:
hl=1-26 h2 = 1 + 66, (23)
while h,,, = 3(1 -ad). Here the role of l/S is played by the dimensionless ratio D =
D[1 - 1/(2S)]/(6J). We propose that b 1.
As usual, the AFMR frequencies are adapted to the non-zero h, - h,. For h < h , ,
w , = 0 reflecting the trivial invariance with respect to the rotation about the magnetic
field axis, while
+
w 2 = 6JS(hZ 3 d ( l + h)?)’P (244
and the thirdnon-zerofrequency, theanalogueofquantumgap, isgiven by the following
expression:
h + 3)j + 54D4 (246)
h’ + 3 d ( h + l)?
In the limiting cases equation (24b) becomes:
3fiDjh(l + ih/D2) h9D2

i
w3 = 6JS 3 ( h b ) ’ ”
b ( h + 3)’fi/hiI2
At zero field w3=D3f2.
6’Q h 9 D I P
h > D‘”.
An analogous result has been obtained in the opposite case
(244

ofquasi-lDAFMT[8,19,25]. In thezDcase the D3”dependence wasfound in [26]though

with different coefficient. The source of the discrepancy is not known to us.
Note that though the spin-flop field formally tends to infinity, the region pH -
S(DJ)@issingledout asthew,valuepassesthroughamaximum, w 3 6SJD3I4,atthese -
fields.
For h l < h < h , the AFMR frequencies are practically the same as in the isotropic
quantum case:

W I =6JS(h - h i ) w , = 6JSh wj 6JS(h, - h ) . (25)

They all have finite gaps since the order parameter has no transverse components.
At least for h,,, > h > h,, the analogue of the quantum gap is maintained in the w ,
mode, while the w3 excitation turns out to be massless:

Wi a 6JS6(h2a - h’)3fi/(2fihz) w 2 = 6JSh wj =o. (26)

Evidently, the longitudinal magnetization remains constant and equal to one third
of the saturation value in the intermediate region between h , and h2.Outside this region
it behaves as follows:
O<h<h,
(27)
h , < h < h,,,.
The peculiarity of the easy-axis case is that in addition to the external field there is a
non-zero internal anisotropy field. This field violates the condition sin LY = 2 sin fi for
 Quantum theory of an antiferromagnet in a magneticfield 79

the high-field phase in figure 1 and leads to a non-zero transverse magnetization for
h > h2:
M, 2pS(gd/4)[(h/h,)’ - 1]”2[1 - (h/h,,,)’I3”(h,,/(3h)’. (28)
Note that M , a (h - h,)‘”for h 3 h, and x(h,,, - h)3’2forh S h-!. The field depen-
dence of magnetization is presented in figure 5.
We mention that the difference of ‘susceptibilities’ with respect to the easy-axis
anisotropy may be regarded as a measure of the difference between low- and high-field
non-collinear phases.

5. Comparison with experiment

In this section we wish to discuss briefly the possibilities of observing the unusual
reorientation process discussed in this paper. A relatively large number of substances
are known which with high accuracy can be regarded as quasi-zr, AFMT [27,28]. Among
them two vanadium compounds, VCI2 (TN= 36 K) and VBr, (T,= 28.5 K), are the
most well known [29-331. Susceptibility measurements [30] have shown that the trans-
verse and longitudinal susceptibilities are practically equal to each other, that is, 7 is
actually very small. By experimental estimations, in both substances J” = 0.2 K [31]
while the in-plane exchange is 23 K in VCI, and 16 K in \.’Br2 [32]. The anisotropy
constant (measured through the anisotropy of the g-factor in the paramagnetic state)
was predicted to be negative (i.e. of easy-axis type) and of the order of 0.1 K 1311.
Unfortunately, large in-plane exchange (peculiar for vanadium compounds) forces
the saturation fields to be of the order of 102T,which makesobservation of the plateau
on the field dependence of longitudinal magnetization very problematical.
Nevertheless, useful information can be obtained from AFMR measurements in
relatively low fields. In both substances the low-energy mode at zero field was observed
at w = 9.1 GHzfor T = 1.7 K(VBr,)and T = 2.6 K(VC12)[30, 331. Meanwhile,accord-
ing to calculations with the given values of d and J (the latter is usually known to very
high accuracy from neutron measurements), the lowest mode had to be placed at w =
0.5 GHz. An attempt to resolve this discrepancy was undertaken in [26]: in addition to
theincreaseofDvalueit was proposed that asuperfineinteractionalsoplaysanessential
role. We propose to verify whether the low-lying mode isreallyw3and also todetermine
the anisotropy constant by studying the field dependence of this mode since, as found in
equation (246), this mode increases with the field at low fields and passes through a
-
maximum at p H 6JSD’n.
Fortunately, a quasi-,r, substance with a rather low saturation field is known. This is
the intercalation compound C,Eu [34,35]. This substance is a metamagnet (the
interplane exchange is ferromagnetic) and thus for our aims the smallness of the inter-
and in-plane exchange integral, though it exists in practice due to the difference in the
interatomic spacings, is not of great importance. Suematsu et a1 [35] have investigated
the magnetization versus magnetic field in fields up to 25 T for two mutually orthogonal
field orientations. For one of the orientations the magnetization increased more or less
monotonically with the magnetic field while for the other they found a plateau on
the magnetization curve which tended to be increasingly noticeable with decreasing
temperatures. For the convenience of readers the experimental curves of [35] are
80 A V Chubukou and D I Golosoo

Figure6.Magnefiz~fioncur;esofC,Eu: ( a ) H 1 c(forvarioustemperatures);(b)HJJc(T=
4.2Kf.Thehgure irtakenfromSuemalsuelu1[35].

presentedinfigure6. Thisisexactly thesituationintheeasy-planem(section3) with

the easy-plane anisotropy strong enough to destroy the plateau for the field directed
perpendicularly to the plane (strictly speaking, tracesof the plateau can be seen on the
experimental curves even in this case). We therefore conclude that C,Eu serves as an
experimental exampleOf2DAFMTWiththeunusual modeofreorientation discussed in this
article. The intermediate region of constant magnetization (which with good accuracy
equals one third of the saturation value 7 p ) was estimated in [35] as an interval
22 kG < H < 82 kG,while the saturation fields were 240 kG and 205 kG for H per-
pendicular and parallel to the plane, respectively [35].We cannot definitely answer what
mechanism (quantum or classical) is responsible for the plateau, but since the spin value
is rather high, S = $, the mode of reorientation through the intermediate phase seems
to be mainly singled out by classical effects: biquadratic (or possibly even more exotic
permitted for S = I ) exchange interaction. We hope that AF~Rexperimentsmay clarify
the situation.

6. Summary
We have found that spin reorientations at T = 0 in the external magnetic field in quasi-
2D antiferromagnets on a triangular lattice may occur in an unusual way via an inter-
mediate collinear phase with the constant longitudinal magnetization equal to one third
of the saturation value. For quasi-zr, Heisenberg and XY antifertomagnets this mode
of reorientation is singled out by quantum fluctuations, while for more complicated
Hamiltonians this may already be the case for classical spins. The reorientation via
intermediate phase with a plateau on the M,,(H)curve was detected experimentally.

Acknowledgments

The authors would like to thank M 1 Kaganov, R 2 Levitin, V I Marchenko and S E

Korshunov for helpful discussions.

Appendix 1

At zero field the bosonic version of the spin Hamiltonian for isotropic ZD AFM on the
triangular lattice is as follows:
 Quantum theory of an antiferromagnet in a magneticfield 81

X/6SJ = Eo + x{a:ak + bk+bk+ c l c h + +[(albk + c:ak + b,tck)v-k

k

+ ( a l c k + c:bh + b,+ah)vh]- 2[(alb!k + b;C!h + c:a?p)v+

+ (ahb-h + bhc-k + c ~ u - ~ )+~. ~. . ] } (‘41)
where the dots stand for the anharmonic terms. The canonical transformation diag-
onalizing the quadratic form of equation (Al) is:
ak = ( l / ~ ) ( p -h qk - iybrk)
bh =(l/fi)bh - exp[-(h/3)iYhlqk - iyk exP[(2n/3)iYhlrh} (-42)
ck = ( l / d ) ( P h - exp[(k/3)iyklqh - iyk exp[-(2x/3)iyhIrh)
where pk, qk and rh are new bosonic fields and Y h = sign(1m vh).Knowledge of the
transformation allows calculation of the first quantum corrections to the ground state
energies for umbrella-like and planar configurations, that is, determination of 7 (see
equation ( 5 ) ) .

Appendix 2

The classical AFMR frequencies for the ZD AFMT with single-ion easy-plane anisotropy.
D ( S f ) 2 ( D > 0), are the following [d = D ( l - 1/(2S))/(U), h = 2 p H / ( 6 1 S ) ] : for
O<h<l,
W I US[6(3 - 2h - !-z’)]’!~
w2 = U S [ 6 ( 3+ 2h + h 2 ) + h’]’” wj = O
while for 1 < h < 3 ( h = 3 is the saturation value) and d 4 1,
W I =o
~2 61S[h2 + ( d / 1 6 h 2 ) ( h 6- 3h4 + 35h2 + 63)]’/’
03 US[d(9 - h2)(h’ - l ) ( h z + 7 ) / ( 1 6 h 2 ) ] 1 / 2 .

References

[ 11 Fazekas P and Anderson P W 1974 Phi/. Mag. 30 423

[2] AndersonPW 1987ScienceW5 1196
[3] Dombre T and Read N 1988 Phys. Re”. B 38 7181
[4] Huse D A and Elser V 1988 Pltys. Reo. Len. 60 2531
[SI FujikiS 1987Can. J . Phys. 65489
[6] Kawamura Hand Miyashita S 1985 1. Phys. Soc. Japan 544530
[7] Andreev A F and Marchenko V I 1980 Usp.Fiz, Nauk IN 39
[S] Zaliznyak I A, Prozorova L A and Chubukov A V 1989 J . Phys.: Condew. Maner 14743
[9] Kawamura H 1984J. Phys. Soc. Japan 532452
[IO] Lee D H, JoannopoulosJ D, Negrle 1 W and Landau D P 1984 Pkys. Reo. Le!!. 52 433
[ l l ] Lee D H, Galfish R G , Joannopoulos J D and Wu F Y 1984 Phys. Reu. B 29 2680
[12] Korshunov S E 1986 J . Phys. C: Solid Slate Phys. 19 5927
1131 RastelliE,ReattoLandTassiA 19833. Phys. C:SofidSfatePhys. 16L331: 19841. App/. Phys. 551871
[I41 Shender E 1982Soo. Phys. JETP56 178
82 A V Chubukov and D I Golosov
11.51 Chandra P, Coleman P and Larkin A I Rurgers Preprinr Ru-89-20
[16] Villain J, Bidaux R, Curton I P and Conte R 1980I. Pkysique41 1263
[17] Marchenko V I 1989 private communication
[IS] Dzyaloshinsky 1 E and Kukharenko B G 1976 Zh. E h p . Teor. Fiz. 70 2360
[19] Miyashita S 1986J. Pkys. Soc. Japan 55 3605
[20] Zaliznyak I A. Marchenko V I, Petrov S V, Prozorova L A and Chubukov A V 1988 Zh. E h p . Teor.
Fir. Pis. Red. 47 172
[21] Golosov D I and Chubukov A V 1989 Z k . E h p . Teor. Fir. Pis.Red, 50 416
[22] Golosov D I and Chubukov A V 1988 Fir. Tuerd. Tela. 30 1542
[23] Chubukov A V 19881. Pkys. C: Solid Stare Phys. 21 LJ41
I241 Tanaka H. Kaahwa Y. HasegawaT. lgarashi M. Teraoka S,lio K and Nagata K 19891. Phys. Soc. Japan
58 2930
[ E ] Tanaka H. Teraoka S. Kakehashi E, lio K and Nagata K 1988I. Phys. Soc. Iopon 57 3979
[26] Suzuki T and Narsume Y 1987J. Phys. Soc. Jnpon 56 1577
I271 Gekht R S 1989 Usp. Fiz. Nouk. 159261
[28] Kawamura H 19881. Appl. Pkys. 63 3086
[291 Nishi M, Ito Y, Kadowaki Hand Hirakawa K 19843. Pkys. Soc. Jnpm 53 1214
1301 Hirakawa K, Ikeda H, Kadowaki H and Ubukoshi K 1983J. Pkys. Soc. Japan 522882
1311Kadowaki H, Ubukoshi K and Hirakawa K 1985I . Pkys. Soc. Jupan 54 363
[32] Niel M, Cros C, Le Rem G, Pouchard M and Hagenmuller P 1977 Physica B % 4 S 702
[33] Yamada I. Ubukoshi K and Hirakawa K 1984I. Phys. Soc. Japan 53 381
[34] Makrini M E , Guerard D.Lagrange P and Herold A 1980 Physiclr B 99 481
[351 Suematsu H , Ohmatsu K , Sugiyama K, Sakakibara T, Motokawa M and Date M 1981 Solid Smre
Commun. 40241