HAL Id: jpa-00208597
https://hal.archives-ouvertes.fr/jpa-00208597
Submitted on 1 Jan 1977
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, estdestinée 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.
A magnetic analogue of stereoisomerism : application to helimagnetism in two dimensions
J. Villain
To cite this version:
J. Villain. A magnetic analogue of stereoisomerism : application to helimagnetism in two dimensions.
Journal de Physique, 1977, 38 (4), pp.385-391. �10.1051/jphys:01977003804038500�. �jpa-00208597�
A
MAGNETIC ANALOGUE OF STEREOISOMERISM :
APPLICATION TO HELIMAGNETISM IN TWO DIMENSIONS
J. VILLAIN
Laboratoire de Diffraction
Neutronique, Département
de RechercheFondamentale,
Centre d’Etudes Nucléaires deGrenoble,
85X,
38041 GrenobleCedex,
France(Reçu
le13 juillet 1976,
revise le 8 novembre1976, accepté
le4 janvier 1977)
Résumé. 2014 Les systèmes magnétiques invariants par un groupe de
symétrie
continu (tels que les modèles de Heisenberg ou XY) peuvent avoir un état fondamental à dégénérescence discrète, c’est- à-dire que les états de plus basse énergie forment dans l’espace de configuration despoches
séparéespar des barrières de potentiel. Ce phénomène est étudié en détail dans le cas d’un hamiltonien de
spins isotrope sur un réseau de Bravais. Outre les verres de spin, dont nous ne
parlons
pas ici, une application de cette idée est le magnétisme à 2 dimensions : les systèmesmagnétiques
deHeisenberg
dont l’état fondamental présente une dégénérescence discrète ont un certain type d’ordre à grande
distance en dessous d’une certaine température de transition, alors que les ferromagnétiques bidi-
mensionnels considérés habituellement sont désordonnés à toute température.
Abstract. 2014 Magnetic systems invariant by a continuous symmetry group
(e.g.
the Heisenbergand XY models) may have a ground state which exhibits a discrete degeneracy, i.e. the lowest energy states form pockets separated by energy barriers in the phase space. This phenomenon is investigated
in detail for the case of an isotropic, bilinear spin Hamiltonian on a Bravais lattice. Apart from spin glasses, which are not considered here, this idea can be
applied
to magnetism in two-dimensional lattices : two-dimensional, Heisenberg magnets with a discrete ground-state degeneracy areexpected
to have some kind of
long
range order below some transition temperature, whereas two-dimensional,conventional ferromagnets are disordered at all finite temperatures.
Classification Physics Abstracts
7.480 - 8.514
1. Introduction. - Broken symmetry is a current
phenomenon
in Nature : a system describedby
aHamiltonian invariant under some symmetry
group
may have a
ground
state which is notinvariant,
butwhich is
degenerate
because anyoperation
of 9transforms it into another state of lowest energy.
In this paper, it is assumed that 19 is a continuous group : in this case, the
ground
state can be conti-nuously modified,
at least in classical systems, and it can be said that it has a continuousdegeneracy.
But it may or may not
happen
that all states of lowestenergy cannot be reached
by
continuous transforma- tions of theground
state; ifthey
cannot, i.e. if states of lowest energy formpockets separated
in thephase
space
by potential barriers,
it will be said that theground
state has a discretedegeneracy,
in addition toits continuous
degeneracy.
Well-knownexamples
in
chemistry
are the stereoisomers ofoptically
activemolecules.
However,
thisphenomenon
does not seemto have been much
investigated
inmagnetism, though
it may be of some
importance
forspin glasses [1].
It has also
important
consequences for two-dimen- sional magnets, as will be seen in section 4.A
simple magnetic example
isprovided by
a clusterof 4 classical
spins SA, SB, Sc, SD
describedby
aHeisenberg
Hamiltonian :with J 0.
If I J’IJ is sufficiently small,
an elemen- tary calculation shows that there are 2 stereoisomericground
states(Fig. 1)
which cannot coincideby
rotation.
In the present
work,
theground
statedegeneracy
is
investigated
for aperiodic
array of Nclassical,
n-dimensional
spins Si
of modulusS2
=1,
sub-FIG. 1. - A finite magnetic system with a discrete ground state degeneracy. The C direction points into the paper and the C’
direction points out.
Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:01977003804038500
386
mitted to an
isotropic Hamiltonian,
for instance abilinear Hamiltonian :
We shall assume n > 2
throughout
this paper.The cases n = 2
(XY model)
and n = 3(Heisenberg model)
are ofspecial
interest for obviousphysical
reasons. The case n = 1
(Ising model)
will not beconsidered.
Sections 2 and 3 are devoted to the
ground
statedegeneracy,
and in section4,
this idea isapplied
totwo-dimensional magnets at finite temperatures T =A 0.
2. Bravais lattices with bilinear Hamiltonian. - 2. 1 GENERAL FEATURES OF THE GROUND STATE. -
If the
spins
are situated at the sitesRi
of a Bravaislattice,
and if the Hamiltonian(1) acting
on them hasthe
corresponding
translation invariance property,states of lowest energy are known
[2]
to begiven,
forn > 2, by:
where u and v are 2
orthogonal,
unit vectors andQ
is one of the absolute maxima of :
Two values of
Q
areequivalent
if their difference is areciprocal
lattice vector. The set ofnon-equivalent
values of
Q
in the first Brillouin zone constitute the socalled star ofQ ;
the star ofQ
is a discrete setinvariant under the symmetry
operations
of thepoint
group
[3].
2.2 FIRST CASE : THE STAR OF
Q
CONSISTS OF ONLY ONE ELEMENT. - This case is the most common in Nature and is of no interest for the presentwork;
it includes :
i) ferromagnets (when Q
=0) ; ii)
aclass of
antiferromagnets (when
2Q
is areciprocal
lattice
vector) ;
classicalantiferromagnets
of thisclass can be transformed into
ferromagnets by
thetransformation :
In this case, all states of lowest energy can be reached from one of them
by
a continuous rotation of the whole system ofspins.
2.3 SECOND CASE : : THE STAR OF
Q
CONSISTS OF MORE THAN ONE ELEMENT. - This case is less common, but notexceptional
in Nature[2, 4, 5].
Forgiven
values of u and v,
Q
can assume a finite number of discretevalues,
but theground
state has notnecessarily
a discrete
degeneracy,
because of twopossible
mecha-nisms :
i)
The reversal ofQ
in(2)
isequivalent
to thereversal of v, and the latter can be done
continuously,
except for n =
2; therefore,
theground
state has nodiscrete
degeneracy
for n >, 3 if the star ofQ
isreduced to
Q
and -Q.
ii) Equation (2) always
describes lowest energy states, but notnecessarily
all ofthem; examples
willbe
given
at the end of this section.It is shown in the
appendix
that all lowest energy states are describedby (2),
if n = 2 or3,
except ifQ
has certain
special positions
in thereciprocal
space.When all lowest energy states are
given by (2),
theground
state has a discretedegeneracy : i)
for n = 3if the star of
Q
is not reduced toQ
and -Q ; ii)
forn = 2 if the star has more than one element.
An exhaustive list of the cases, where all lowest energy states are not described
by
eq.(2),
will not begiven here,
and we shall be contented with a fewexamples.
a) Spin dimensionality n >
4. Thesimplest
way to obtain a lowest energy state, which is notgiven by (2),
is the addition of 2 states(2)
with different structure vectorsQ, Q’, namely :
where 0 is an
arbitrary phase,
u, v, w, m are 4 unit vectors, which must beorthogonal
tosatisfy
thecondition I Si I’
= 1.Q
can becontinuously changed
into
Q’ by varying
0 from 0 ton/2.
b)
If2 Q
= i is areciprocal
lattice vector, sinQ . Ri
= sinQ’ . Ri
=0,
so that v and mdisappear
from
equation (4),
andonly
2orthogonal,
unit vec-tors u and w are
required ; equation (4)
now reads :These states exist for all values of n >
2 ; again
thetransition from
Q
toQ’
can be doneby
a continuoustransformation.
c)
If 4Q
= r is areciprocal
lattice vector, and is 2Q
is not, this isagain possible
for n >,2, though
it is not obvious from
(2).
This can beexplained
asfollows : let F be the set of
spins parallel
to u in thestate
(2),
and let G be the set ofspins parallel
to v :the
spins
F see no fieldproduced by
thespins
G(and
vice
versa) ;
this isobvious,
because otherwise thespins
wouldalign parallel
to the same direction.Therefore,
it ispossible
to tilt thespins
G withoutenergy
change :
In
particular,
thespins
can bealigned (qi
=0),
andthen it is
possible
to add two states with different values ofQ, obtaining
a state describedby :
Again Q
can bechanged
intoQ’ by
the continuous transformation of 0 from 0 ton/2.
Remarks :
i)
Cases(b)
and(c)
involve colinear structures(in
addition to cantedstructures)
which can beconsidered as starred
antiferromagnets.
Colinear struc-tures can be obtained
only
forQ = 0 or 2 Q = r
or 4
Q
= i, where r is areciprocal
lattice vector;Cases 3
Q
= r or 6Q
= r, forinstance,
have nospecial properties.
ii)
Cases(a), (b), (c)
are not theonly examples
of structures not described
by (2) ;
anotherexample
ofminor
experimental
interest isprovided by
the casewhen 2
Q
lies on a Brillouin zoneboundary,
forcertain
symmetries.
iii)
The valuesQ
=0, i/2
andi/4
are favouredby
symmetry
[18],
as well asQ
=i/3 or r/6,
forcertain
lattices. Other commensurable values are not favour- ed for
purely isotropic
interactions and canonly
occur
accidentally.
iv)
The absence of discreteground
statedegeneracy
in the above
examples (a), (b), (c)
is apathological
feature of the bilinear Hamiltonian
(1); biquadratic
terms, for
instance,
restore the discretedegeneracy,
as will be seen in section 3.
v)
When all lowest energy states aregiven by (2),
the discrete
degeneracy
of theground
state iseasily
seen to be 2 m for n =
2,
and m for n =3,
if 2 m is the number of elements of the starof Q.
2.4 ORDER PARAMETER. - When the lowest energy states
split
into several continuous setsseparated by potential barriers,
the various sets can be characterizedby
thecorresponding
values ofQ ; however,
it ispreferable,
in order to extend thetheory
to finite temperatures T =1=0,
to define alocal,
observable order parameter, as a function of the fieldSi.
For n =
2,
such an order parameter is thefollowing quantity,
which is a vector in the lattice space, with D components which arepseudo-scalars
in thespin-
space :
where a = x, y, z ;
0,,Si
is defined for the cubic latticeby :
where ax, ay and a., are the interatomic vectors
along
x,y, Z.
At T =
0,
thequantity Qi
definedby (7)
reduces to :and its
knowledge
isequivalent
to theknowledge of Q,
except ifQx
orQy or 6z
isequal
ton/2
a.For n >
3,
nopseudo-scalar
similar to(7)
can bederived from the two-dimensional field
(2),
but apossible
order parameter is thefollowing
D x Dtensor, whose components are scalars in the
spin-
space :
At T =
0,
theknowledge
ofqfY
isequivalent
to theknowledge
ofQ,
except for certainspecial positions of Q.
The difference between
Heisenberg
and XY modelsis related to the difference which was noticed in section 2. 3.
3. Bravais lattice with
biquadratic
interactions. -Biquadratic
interactions are thesimplest sophisti-
cation of the model
(1),
which does not violate theisotropic
character. The Hamiltonian becomes :There is no
general recipe
for theground
state.It can
happen, however,
that the classicalground
state is still
given by (2).
Inparticular, positive gij’s
stabilise colinear states encountered in section 2.3 : 0 = 0 or
n/2
in eq.(5)
and(6b), ip
= 0 in(6a).
Moregenerally, biquadratic
interactions can restore the discretedegeneracy
of theground
state.Results from sections 2 and 3 are summarised in table I.
TABLE I
Existence
(Yes)
or non-existence(No) of
a discreteground-state degeneracy for
Bravais latticeshaving
astate
of
lowest energy describedby equation (2).
Incases denoted
by
a star,biquadratic
terms are necessaryfor
a discreteground
statedegeneracy.
The list is notexhaustive.
388
An
interesting
case, for n =3,
is when theground
state is not described
by
eq.(2),
but instead contains Fourier componentsQ, Q’, Q"
which do notbelong
all to the same star; an
example
isprovided by Erbium,
whereQ’ = - Q
andQ" =
0 :The essential difference with the cases considered in section 2 is that there is a
discrete,
doubledegeneracy
for n = 3 even if the star of
Q
is reduced toQ
and-
Q,
since it isimpossible
to transformQ
into -Q
in
(10) by
a continuous transformation without cross-ing
an energy barrier. Apossible
orderparameter
is thefollowing
set of 2pseudo-scalar components :
where
where gii
= 1 for nearestneighbours,
gij = 0 other- wise.The structure
(10)
can beproduced by anisotropic forces,
but alsoby biquadratic, isotropic exchange.
Similar structures with Fourier components
Q, - Q
and 0 can also result from
bilinear, isotropic exchange
forces in non-Bravais lattices.
So
far, only
theground
stateproperties
have beenconsidered in this paper. In the next
sections,
theconsequences of a discrete
ground
statedegeneracy
on the
properties
of two-dimensional magnets at finite temperature areinvestigated.
4.
Application
tomagnetism
in two-dimensional lattices. - 4.1 THE MERMIN-WAGNER THEOREM AND THE SPIN-PAIR CORRELATION FUNCTION. - The theorem establishedby
Mermin andWagner [6]
for ferroma- gnets andantiferromagnets
caneasily
be extended to allmagnets
describedby
the Hamiltonian(1)
forn = 3 and short range
interactions,
on a two-dimen- sional lattice : theMermin-Wagner
theorem statesthat the
staggered magnetisation :
has a
vanishing
limit at all finite temperatures when theapplied staggered
field goes to zero.In the
ferromagnetic
case, it iscommonly
admittedthat the
spin pair
correlationfunction Si. S j )
has a limit when the distance rij goes to
infinity.
Thislimit is
obviously ( Si >’,
and theMermin-Wagner
theorem
implies
that it is zero for two-dimensional lattices if T i= 0 :We suggest that all
two-dimensional, isotropic
magnets with n >
2,
have the property(12),
whateverbe their structure
(ferromagnetic, antiferromagnetic, helimagnetic...).
For n =2,
the property(12)
resultsfrom any
approximation,
for instance the harmonicapproximation, proposed by Wegner [7]
for ferroma- gnets, and which caneasily
be extended to heli- magnets(i.e.
structuresgiven by
eq.(2)
when 2Q i= i) ;
the extension of the statement
(12)
tohigher
values of n isjustified by
the fact that the increase of n isexpected
to increase
fluctuations,
and therefore to decrease correlations.Eq. (12)
can beproved
for n = oo(spherical model).
4.2 PHASE TRANSITIONS IN TWO-DIMENSIONAL, HEI-
SENBERG OR XY MAGNETS. - We shall now argue that two-dimensional
magnetic
systems which have adiscrete
ground
statedegeneracy
do exhibit some kind oflong
range order(L.R.O.)
at low temperature, and therefore have a transition at some temperatureT,.
Moreprecisely,
the XYhelimagnet
consideredin section 2. 3
c)
satisfies for TT,,
the property :where
Q"
is definedby (7).
TheHeisenberg helimagnet
described in section 2.3
d)
satisfies for n = 3 the property :where
qaY
is definedby (8).
The system describedby (10)
at T = 0 has below
Tc
the property :where
T?
is definedby (11). Properties (13)
to(15)
do not contradict
equation (12)
nor the Mermin-Wagner
theorem.4.3 THE ARGUMENT. - For
properties (13)
to(15)
at low temperatures, is as follows. For
definiteness,
we shall argue on the XY
helimagnet (eq. (13))
butthe argument is
general.
First, if KB
T is much lower than thetypical exchange interaction I J 1, Qi
isnearly equal
to one of the maxima of3(k)
atnearly
all sites(1).
This can be seen froma standard
spin-wave
calculation in afinite
volume CO’containing
L 2 unit cells. This calculation will not beexplicitly given here,
but it isquite
similar to that ofBerezinskii and Blank
[8]
forferromagnets.
It startswith the
assumption
thatSi
isgiven by
the expres- sion(2), plus
small deviationsbSi; neglecting
thirdand
higher
order terms in65;
in theHamiltonian,
one(1) This argument is for a continuous model, when Qi is defined by (7) again, but ocxSiis an ordinary gradient ; in this case, Qi reduces
to Q at T = 0. The extension to a lattice only implies a change of vocabulary.
calculates ( ðS? ),
which is indeed found to be smallif L
exp J 1/ KB T,
theself-consistency
condition.The
fluctuation ( (Qi - Q)2
is calculated in thesame way and is also found to be
small,
so thatQi
is almost
equal
to one of the maximaQ
of3(k)
atalmost all sites i.
This statement is also
expected
to be true at low temperature for an infinite magnet, since an infinite magnet can be considered as anassembly
of interact-ing
cells of sizeL 2,
and the interactions canjust
increase the
tendency
thatQi
lies close to one of the maxima of3(k).
There are,
however,
a few sites at whichQi
is notclose to one of the maxima of
3(k) ;
theproblem
is :can these few sites
destroy L.R.O., i.e.,
canthey
makeequation (13) wrong ? Clearly, they
canonly destroy
L.R.O. if
they
are concentrated in frontiersseparating regions having
different values ofQ.
For a discreteorder parameter, frontiers must be narrow at low temperatures since their entropy is irrelevant and their energy is
proportional
to their thickness. The energy of a narrow frontier per unitlength
is oforder J 1,
and its entropy is of orderKB.
Thereforethe ordered state
(when Qi
is close to the same elementQ
of the star at almost eachsite)
is stable with respectto the formation of one
frontier, provided KB T 1 J I
.A
long
rangeordering
of the type described in sec-tion 4.2 is therefore
expected
below a transitiontemperature of order :
4.4 DIsCUSSION. - We have
given
a ratherrough
argument, for the sake of
brevity
andsimplicity,
but it can
easily
beimproved ;
also note that it caneasily
be extended to quantum systems. It can also be extended to 3-dimensionalspins
with an XYanisotropy,
and in this case, thesemi-polar
repre- sentation[9]
can beused,
with theadvantage
that thespin-wave
calculation can beperformed
in an infinitesample
at once. In theHeisenberg
case,however,
the method of section 4. 3 is theonly
one which canbe
applied.
Also for the sake of
simplicity,
we havegiven
theorder of
magnitude (16)
as ageneral result;
this isnot
strictly
true; moregenerally, KB T,,
isexpected
tobe of order
L1,
where L1 is the energy barrier between the various maxima of3(k).
It canhappen
that L1is
appreciably
lowerthan J 1,
for instance near a Lifshitzpoint [10, 11, 12, 13]
and thisprobably explains why
the Neel temperature of the(expectedly quasi-two-dimensional) helimagnet BaCo2(As04)2
ismuch lower than the average
exchange interaction I J 1 (see
ref.[14]
and[15]).
A
possible approach
tohelimagnets
has beendescribed
by
Mukamel et al.[4]
andby
Garel[5],
who show that within the Landau-Wilson formulation
a
helimagnet
isequivalent
to ananisotropic
ferroma- gnet with(2 nm)-dimensional spins,
if 2 m is thenumber of elements of the star of
Q ;
in thisapproach,
the
possibility
of L.R.O. at low temperature wouldprobably
result from theanisotropy.
A final remark is that a transition is
expected
at a temperature oforder J 1/ KB
for two-dimensionalferromagnets
of the XY type[16],
but this transition involves nolong
rangeordering,
in contrast withproperties (13)
to(15).
Theordering
described in thissection in even more
spectacular
forHeisenberg
systems, since
conventional, Heisenberg ferromagnets
are
presently
believed to have no transition[17].
APPENDIX
Discrete
degeneracy
of theground
state of theHeisenberg
model on a Bravaislattice,
when the struc-ture vector
Q
is not in aspecial position.
- It iseasily
shown
[2]
that the classicalground
state of(1)
on aBravais lattice is the solution of the system :
The summation in
(A .1)
is over the elements of thestar of the maxima of
(3). Eq. (A. 2) yields :
Eq. (A. 4) yields,
after insertion of(A. 1) :
or :
where the sum is over the elements
Q, Q’
of the starand over the
reciprocal
lattice vectors T.When
Q’ = - Q,
there isgenerally only
one termin the sum at the L. H. S. of
(A. 5).
Theexception
iswhen there are 3 elements
Q, Q’, Q"
of the star, whichsatisfy :
This
equation
canonly
be satisfied ifQ
has aparticular position;
forinstance,
if4 Q
= i(A.6)
is satisfied for
Q’
=Q" = - Q ;
it will beassumed,
from now on, that(A. 6)
is not satisfiedby
any 3 vectors of the star. In this case, insertion ofQ’ - - Q
into
(A. 5) yields :
or :
390
where yQ is a real
number,
uQ and vQ areorthogonal
unit vectors.
According
to(A. 4),
one can choose :When
Q’ =1= - Q,
the sum at the L.H.S. of(A. 5) generally
consists of several terms, and the discussion should be continuedseparately
for each type of symmetry; theforthcoming
discussion assumes thatQ
is in
general position;
in this case, the summation over i in(A. 5)
can be limited tor = 0.a)
Monoclinic lattice. - The star consists of 4 vectors ±Ql,
±Qo,
which form arectangle.
Oneof the two
quantities
yo, yl can be assumed to be different from zero, for instance yo :0 0.If, in addition,
y, 00,
uo, ul, vo, vi mustsatisfy
thefollowing
set ofequations,
which iseasily
deduced from(A. 5) :
In
addition,
Uo.Vo = UI.VI = 0. This system has solutions for n >4,
andthey correspond
to thestructure
(4).
For n = 2 or3, (A. 9)
has nosolution,
therefore Y, = 0 : this
corresponds
toequation (2).
b)
Orthorhombic lattices. - The star consists of 8 elements : ±Qo,
+Ql,
±Q2,
±Q3 ;
letQi, Q2, Q3
be deduced fromQo by symmetries through
3orthogonal planes; assuming
yo, yl, Y2 =1=0,
one deduces from(A. 5)
thefollowing homogeneous equations :
In view of the
orthogonality
of u and v, it iseasily
seen that these relations cannot be satisfied
by
unitvectors;
therefore,
one of theyk’s
is zero, for instance Y2 =0 ; but,
in this case, the argument(b)
can beapplied
toQo, - Qo, Q,
and -Ql,
andyields
yo = 0 or y, =
0;
the sameargument
as above holds ifQ2
isreplaced by Q3.
The final result is thatonly
one of the 4 numbers yo, Yl, Y2, Y3 can be different from zero, so that all lowest energy states have the form
(2),
when n = 2 or 3.c) Tetragonal
lattices. - The star consists of 16 ele- ments ; letQl, Q2, Q3
be deduced fromQo by
rota-tions of
7r/2, 7r
and 37r/2
around thetetragonal axis,
and let ±
Q’O,
±Q’,
±Q2
and ±Q’
be the other elements.According
to the same argument as inb),
yl and Y3 should vanish if
yo 0 0 ; yl
andy3
alsovanish if one assumes
Y’ 0 0 0,
forinstance,
ifthe argument
b)
can beapplied
to show that not morethan one of the 3
quantities
yo,y’,
Y2,Y2
can be diffe-rent from zero
if n 3,
so that all lowest energy states have the form(2).
d) Hexagonal
lattice. - In thegeneral
case, thestar of
Q
consists of 24 vectors : letQi ..., Q11
be deduced from
Qo by
rotations around the hexa-gonal
axis c, andby symmetries through planes parallel
to c ; let
Qp+6
besymmetric
ofQp
with respect tothe c
axis ;
assume yo =A0 ;
then the argumentb)
can be
applied
to the 8 vectors ±Qo,
±Q6,
±Qp,
±
Qp+ 6
to show that yp = 0 forall p 1=
0 or6 ; finally
the argumenta)
can beapplied
to the 4 vectors+
Qo
and ±Q6
to show that Y6 also vanishes. Thereason
why
Y6 should be considered at the end of the argument is that there are many vectors(Q, Q’) satisfying 6Q’,Q + Q. + Q6 :0 0,
so that the term contain-ing
Yo y6 is associated to many other terms in(A. 5).
Again,
all lowest energy states are describedby (2)
ifn = 2 or 3.
e)
Rhombohedral lattice. - The starof Q
is reducedto 12 elements and can be deduced from the star
corresponding
to thehexagonal lattice, by
the sup-pression
of 12 vectors, or,alternatively, by
therequire-
ment that the
corresponding yp’s vanish; therefore,
the conclusion thatonly
one of theyp’s
can be diffe-rent from zero, remains correct.
f)
Cubic lattices. - Theproof
isanalogous
buttedious.
g) Exceptions.
- Theargument applies
wheneverit is
impossible
to find 4 vectorsQ1, Q2, Q3, Q4
of the star, which
satisfy :
where r is a
reciprocal
lattice vector.(A 10)
is ageneralization
of(A. 6),
since(A. 5) corresponds
toQs = Q4
The fact that
Q eventually
lies on asymmetry plane
or axis
apparently
does notchange
thegeneral result,
since this does not
imply
that(A. 10)
is satisfied.References
[1] EDWARDS, S. F., ANDERSON, P. W., J. Phys. F 6 (1976) 1927.
[2] YOSHIMORI, A., J. Phys. Soc. Japan 14 (1959) 307.
VILLAIN, J., J. Phys. Chem. Solids 11 (1959) 303.
[3] KOSTER, G. F., in Solid State Physics, Vol. 5, edited by F. Seitz
and D. Turnbull (Acad. Press, New York) 1956.
[4] MUKAMEL, D., KRINSKY, S., Phys. Rev. B 13 (1976) 5065.
[5] GAREL, T., Thesis, Paris (1976).
[6] MERMIN, N. D., WAGNER, H., Phys. Rev. Lett. 17 (1966) 1133.
[7] WEGNER, F., Z. Phys. 206 (1967) 465.
[8] BEREZINSKII, V. L., BLANK, A. Ya., Sov. Phys. J.E.T.P. 37 (1973) 369.
[9] VILLAIN, J., J. Physique 35 (1974) 27.
[10] HORNREICH, R. M., LUBAN, M., SHTRIKMAN, S., Phys. Rev.
Lett. 35 (1975) 1678.
[11] HORNREICH, R. M., LUBAN, M., SHTRIKMAN, S., Phys. Lett.
55A (1975) 269.
[12] HORNREICH, R. M., LUBAN, M., SHTRIKMAN, S., To be publish-
ed in Physica (1976).
[13] VILLAIN, J., To be published in the proceedings of I.C.M.
76, Physica (1977).
[14] REGNAULT, L. P., Thèse de Troisième Cycle, Grenoble (1976).
[15] REGNAULT, L. P., BURLET, P., ROSSAT-MIGNOD, J., To be published in the proceedings of I.C.M. 76, Physica (1977).
[16] KOSTERLITZ, J. M., J. Phys. C 7 (1974) 1046 and references therein.
[17] BREZIN, E., ZINN-JUSTIN, J., Phys. Rev. Lett. 36 (1976) 691.
[18] DZYALOSHINSKII, I. E., Sov. Phys. J.E.T.P. 19 (1964) 960.