HAL Id: jpa-00207339
https://hal.archives-ouvertes.fr/jpa-00207339
Submitted on 1 Jan 1972
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.
Dynamics of KH2PO4 type ferroelectric phase transitions
C. Tsallis
To cite this version:
C. Tsallis. Dynamics of KH2PO4 type ferroelectric phase transitions. Journal de Physique, 1972, 33
(11-12), pp.1121-1127. �10.1051/jphys:019720033011-120112100�. �jpa-00207339�
DYNAMICS OF KH2PO4 TYPE FERROELECTRIC PHASE TRANSITIONS
C. TSALLIS
Service de
Physique
du Solide et de RésonanceMagnétique
Centre d’Etudes Nucléaires de
Saclay,
BP n°2, 91, Gif-sur-Yvette,
France(Reçu
le 14 avril1972,
révisé le 28juillet 1972)
Résumé. 2014
L’énergie
deconfiguration
estajoutée
à la théorie deKobayashi
pour les cristauxferroélectriques
du type KDP. La théorie est ainsi rendue cohérente dupoint
de vuethermodyna- mique, puisque
le paramètre d’ordre et lafréquence
du mode mou associé s’annulent à la même température. Ilapparaît
lapossibilité
d’existence d’un autre type deferroélectriques
où lephéno-
mène
displacif
declenche la transition et induit unphénomène
d’ordre-désordre des ionslégers (protons, deutérons).
Unegrande
sensibilité à lapression,
de diversesgrandeurs physiques statiques
et
dynamiques, suggère
des travauxexpérimentaux
intéressants sur des substancesferroélectriques
mixtes.
Abstract. 2014 We include the
configuration
energy inKobayashi’s theory
forKDP-type
ferroelec-tric
crystals.
Thetheory
is then shown to be coherent from thethermodynamical point
ofview,
asthe order parameter and the
frequency
of the associated soft mode vanish at the same temperature.The results show that it is
possible
to have another kind of mixedferroelectric,
where thedisplacive phenomenon triggers
the transition and induces an order-disorderphenomenon
for thelight
ions(protons, deuterons).
Greatsensitivity
to pressure of various static anddynamic
relevantphysical quantities
suggestsinteresting experimental
work on mixed ferroelectrics.Classification
Physics Abstracts
17.29
1. Introduction. - A
great
amount ofexperimental
and theoretical work has been done on the ferroelec- tric transition of
KH2PO4 (KDP)
andisomorphous crystals (refer
to[1], [2]
forgeneral characteristics)
and the convergence of the results leaves no doubt about the
microscopic
mechanism of the transition(refer
to the excellent paperKobayashi published
in 1968
[3]).
Thecrystal instability
istriggered by
an order-disorder
phenomenon
in theproton assembly (each proton
moves in a double-wellpotential
essen-tially
createdby
theheavy
ionsK+, p+5
and0-2).
Simultaneously, displacements
of theheavy
ionsare induced
by
thestrong coupling
betweenprotons
and the[K-P04] system ([3]
to[13]).
We may thusconsider this transition to be of a mixed
type,
as order-disorder anddisplacive
characteristics appeartogether.
To take account
theoretically
of the collective motion of theprotons,
de Gennes introduced in 1963[14]
a formalism(widely
used sincethen)
wherethe «
right-left » positions
of aproton
in its double well arerepresented by
the «up-down » projections
of a
-1-pseudo-spin.
From this standpoint
there aremany formal
analogies
between a mixed-ferroelectric and amagnetic displacive-ferroelectric crystals (parti- cularly
the collective vibration modes of theprotons
in KDP may beregarded
aspseudo-magnons).
Recently
some work has been done[15]
to[17]
onthe
dynamics
of themagnetic ferroelectrics,
and theconclusions may
be,
to a certain extent,applied
tothe mixed ferroelectrics. In these latter
crystals
thetransition is
characterized,
as is shown inKobayashi’s
paper,
by
thesoftening
of aparticular
mode ofvibration
(which
is shown to be amixing
ofpseudo-
magnon and
phonon)
of the wholesystem,
in a similar way to whathappens
indisplacive
ferroelec-trics
([18]
to[21]).
From the
thermodynamical point
of view it isclear
that,
in a second-order transition due to adynamical softening,
the orderparameter
and thefrequency
of the moderesponsible
for the transition must vanish at the sametemperature (which
ispreci- sely
the definition of the transitiontemperature),
as
they
are « two faces of asingle
coin ».Kobayashi’s theory
fails on thispoint,
and we show in this paper that the consideration of theconfiguration
energy of theheavy
ions raises up this incoherence. Theapproximations
made in this kind oftheory
and thepossibility
of anothertype
of mixed ferroelectric(where
theheavy
ionstrigger
the transition and inducean order-disorder
phenomenon
in theprotons)
appear
clearly
in the boson creation-annihilation formalism used in section 4. Greatsensitivity
onpressure of various static and
dynamic
relevantphysical quantities suggests interesting experimental
work on
KDP-type
ferroelectrics.Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:019720033011-120112100
1122
2. Model. - Let us consider a
crystal
which may beregarded
as thesuperposition
of several sublattices :one of them made
by light
ions(like protons
ordeuterons)
and the othersby heavy
ions(like p+5, K+, 0-2).
Let us suppose there is onelight
ion per unit cell and that it is allowed to move in a double wellpotential (created by
all the other ions andalmost
temperature independent) along
the z-axis.Strictly speaking
thelight
ions do not build up a sublattice unless all of them are on the same side of the double well. KDP hasactually eight protons
per unit
cell ;
nevertheless the meanordering
isalong
the z-axis. Oursimplified
model loosesnothing
essential in
considering only
onelight
ion per unitcell,
and thegeneralization
of thispoint
demandsnothing
butpatience.
Let us call m(where - f m f)
the reduced mean
position
ofprotons
withrespect
to the center of the double
well, and il
a reducedcharacteristic
displacement
between theheavy
ionsublattices ;
both mand il
are assumedparallel
tothe z-axis. The
crystal
is assumed toundergo
aphase
transition in such a way as to
have,
in a certaintemperature domain,
aparaelectric (disordered) phase
where m = il =0, and,
out of thistemperature domain,
a ferroelectric(ordered) phase
where m i= 0and q
= 0.The Hamiltonian may then be written as follows
(our
notation is close to that of Blinc[7]) :
where the
light
ion Hamiltonian may be written[12]
as follows
where T
( > 0)
is thetunnelling frequency
ofprotons through
the barrier of the doublewell,
andJjj, depends only on 1 Ri - Rj’ 1 (Jjj
=0). By Ri
wedenote the
position
of thejth
unit cell withrespect
to an
arbitrary origin, hence j
runs from 1 toN,
the number of unit cells of thecrystal.
Thespin- operators obey
the usual commutation relations.Only
the essential terms have been considered in this Hamiltonian. However more refined interactions may beeasily
included in the formalism. Forexample,
the
influence,
on thetunnelling frequency,
due tothe other
light
ions may be introducedby
a terman
asymmetrical
double wellpotential
may be account- ed forby
aterm - A L Sj [22], [23] ;
and even inter-j
actions between more than two
spins
could be included with no
great
formal difficulties[9], [12].
In the lattice Hamiltonian we shall
only
considerthe harmonic contribution of the
optical
vibrationstrongly coupled
to thelight-ion system (1),
so wehave
where
Zq
are the Fourier transformed coordinates associated to the vibrations(parallel
to thez-axis) strongly coupled
to thelight ions,
andPq
are theircanonical momenta
(we
consider the massequal
tounity). They obey
of course the usual relation[Zq, PQ-,] - ibqq’, (with h
=1).
Let us add thatWq
= Co_q, where q runs over all allowed modes in the first Brillouin zone.Finally
we shallconsider,
in the interaction Hamil-tonian, only
the first order(essentially dipolar)
inter-action between
light
andheavy
ions(particularly
nophonon-assisted tunnelling
of thelight
ions is consi-dered) ;
we may then writewhere
Vq
is real and even in q, andLet us
point
out that in the total Hamiltonian seno
damping
effects are considered. Let us now decom- poseZq
into its thermal average(noted >)
andfluctuating parts,
i. e.Zq = Qq
+ llq where Ilq =Zq
> andQ9 - Zq - llq (hence Qq
> =0).
As we are
considering
a ferroelectricinstability
wehave
Let us now introduce the
phonon
creation and annihi- lationoperators
with the usual transformationwhere
We may now rewrite the total Hamiltonian as fol- lows :
1
(1) It is clear that, in a particular substance where an acous-
tical mode is strongly coupled to the light-ion system, additio- nal terms must be included in KL as well as in J~pL
(see [28]).
Except
for the three last terms, thisexpression
isessentially
the same asKobayashi’s [3], [7].
Tostudy
the thermal statistics of this Hamiltonian we need to know the exact
partition
function Z = Tre - fla (where fl
=1/kB T),
which is verycomplicate
toobtain. So we have to use an
approximate method, namely
a Hartree-Focktype
variational method. This is to say we propose a class of trial Hamiltonians£0 (which depend
on some variationalparameters)
andthe
corresponding
trialdensity
matrix :and we minimize with
respect
to the variational para- meters theapproximate
free energy :where
Fo
= -1 /fi Log
Tre-pJeo
and>o
denotes the thermal average over the states ofXo -
We shall use, in the
following section,
this method tostudy
the statics of oursystem.
3. Statics. - Let us choose as the trial Hamilto- nian the
following
static one(2)
where the molecular field
(h.,, hz) and il
are the varia-tional
parameters.
First of all we
diagonalize Jeo expressing
it in arotated reference
system (ç, Q
such as to haveh,
= 0.Next we
minimize,
withrespect to r¡, hx
andhz,
the free energy
F previously
defined and we obtain :or
(ferroelectric phase),
where J* =Jo
+Vo 0
As we see, the static role
played by
thecoupling
between
light
andheavy
ions isessentially
toreplace Jo by
J*.The transition
temperature
isgiven by
2
(2) The omission of the term N
mo
2 112 in Jeo does not change the results, as it gives no contribution to the entropy.where
In
figure 1,
where we haverepresented
thisfunction,
we may see the
great sensitivity
ofTo
on i, whichis itself very sensitive on pressure. This fact
suggests interesting experimental
work[24].
Existence of the ferroelectric solution atpositive temperatures implies
J* > 0 and r 1.
Stability
of thecrystal
in theparaelectric phase
demands that the orderedphase
be on the low
temperature
side ofTo.
Discussion ofm(T)
shows that the transition is a second order onefor all allowed values of
T, Jo, Vo
and coo. As usual in Hartree-Fock methods weverify
that the free energy isanalytic,
whichimplies
aLandau-type
behaviour for m, i. e.
m2
oc(To - T)
near andbelow
To.
FIG. 1. - Dependence of the transition temperature on the microscopic parameters rand J* .
In
figures
2 and 3 the behaviour of some relevantquantities
is shown.We should like to call attention to an
important point :
the orderparameter
is not m but the electricpolarization
P(linear
combination of m and’1).
In some
respects (the
determination ofTo,
forexample)
the differentiation is irrelevant becauseFIG. 2. - Temperature dependence of the light-ion (proton) contribution (in reduced units) to the polarization.
1124
FIG. 3. - Temperature dependence of total
( si, >)
andtransverse
« sj > )
mean values of the fictitious spin polari-zation (in reduced units).
17
(V 0
m, hence il, m and P vanish simulta-neously at To.
However thefrequency
which becomessingular
atTo is,
as it is shown in section4,
the oneassociated to
P,
and not to m. So it isimportant
toavoid confusion between P and m. Similar conside- rations must be made in situations
((NH4)HZP04-type
antiferroelectrics or more
complicated structures)
where the order
parameter
is not P.Let us
finally point
out anotherpeculiarity
of themixed-ferroelectric
we
arestudying
here. For this let us first consider F : we see thatUsing
theva_rious
relations we have established wemay
rewrite
F asF(q,
m,T),
where it iseasy
toverify
that
oF/or
=ayam
= 0. If wedevelop
F in(q, m)
and retain
only
the second-order terms we obtainThe
eigenvectors
of this matrix are linear combinations of mand 17
withtemperature-dependent
coefficients.Because of this
peculiarity,
thepolarization
P(which is,
in ourmodel,
a linear combination of mand il
with constant
coefficients)
will not be one of theeigenvectors,
as it isquite
often the case.4.
Dynamics.
- Let us introduce thepseudo-
magnon
operators by
Holstein-Primakoff’s transfor- mationwhere
sf = S§ ± iS;
and(a;, aj) satisfy
usual bosonrelations, particularly [a j, a; ]
=b jj’.
Anexpansion
of these
expressions
leads toLet us now
perform
a Hartree-Focktype
linearization(discussed
in section5),
i. e.where the factor 2 comes from the fact that there are
two ways of
constructing
the mean valueaj aj
>.We may now rewrite
(4. l’)
as followswhere we have used
Let us now introduce the Fourier transform boson
operators by
So if we
perform consecutively
transformations(3.1), (4.2)
and(4.3)
in the Hamiltonian(2.1),
we
neglect high-order
terms and we omit zero- andfirst-order terms
(because they
have no influence onthe boson
frequencies)
we obtainwhere
(J,,
is real and even inq).
Let us now
diagonalize
thepart
of Jeconcerning only
thepseudo-magnons.
For this weperform
theBogolyubov
transformationwhere
80 Je becomes
where
A final
diagonalization (see appendix)
leads towhere
and
In this way the
system
may now beregarded
as thecoexistence of two
types
of noninteracting
bosons(quasi-phonons
andquasi-pseudo-magnons
for q -0).
In
figure
4 we havepresented
an illustration[7]
ofrealistic
dispersion
curves for the modesappearing
here.
The
KDP-type phase
transitions are due to aninstability
related to thequasi-pseudo-magnons.
Itsi easy to
verify
that atTo (given by (3.1))
andq = 0 we obtain
We see then
that,
in the caseYq
= 0(no light
ion-lattice
interaction)
J* =Jo,
andcv((To)
=wg(To) =
o.FIG. 4. - Dispersion curves of the relevant vibration modes.
These results are very
important
from the thermo-dynamical standpoint,
for thecrystal instability
andthe
phase
transition are one and the samepheno-
menon, hence the order
parameter
and thefrequency
of the mode
responsible
for theinstability
mustvanish at the same
temperature.
In this way we see that in the caseVq
= 0 the transition isexclusively
due to a
pseudo-magnon
modesoftening
and theorder
parameter
is m. But whenVq :0
0 the transitiontemperature
ishigher (the light
ion-lattice interaction stabilizes the orderedphase),
thepseudo-magnon frequencies
are still different from zero, the transitionis due to a
quasi-pseudo-magnon instability
and theorder
parameter
is a combination of m and il.Let us
finally
addthat,
as is usual in Hartree- Focktype theories,
we obtain nearTo
and on bothsides of it
rog2
=KIT - Tao 1.
If we callKp(Kf)
the
proportionality
coefficient in theparaelectric (ferroelectric) phase
we find thatKflKp >,
2 for allallowed values of the
microscopic parameters
of thetheory (the equality corresponds
to the limitVo -
0 and 2r/J* - 1) ;
furthermore the ratioKf/Kp
isextremely
sensitive to the values of the para- meters 2r/Y*
andiFo/a)o.
Thissuggests (see
also[3])
interesting experiments
with variable pressure.5. Conclusion. - In the
previous
section wehave,
without
justification,
linearized theequations by using
and
.. , ...
However we believe that a more
rigorous procedure
would not
essentially change
thisapproximation.
1126
If we
perform,
on Hamiltonian(2.1), consecutively
transformations
(3.1), (4.1) (or (4. l’», (4.3), (4.4)
and
finally (A. 2),
we obtain(if
we omit lower thanharmonic
terms,
asthey
do notmodify
thedynamics)
Je =
co"(0) Aq Aq
+cv((0) B; Bq
+q
+
(higher-order
terms inAq, AQ , Bq, Bq ) ,
where
cv£(0)
andCOB 0)
denoterespectively
the 0 cK-values of
quasi-pseudo-magnon
andquasi-phonon frequencies.
A treatment of this Hamiltonian with the variational formalismproposed
in Section2, using
the trial Hamiltonian
ico { OA (T) Aq Aq
+Q((T’) B$ Bq} ,
q
(where QA
andQ:
are the variationalparameters)
seems to prove that the linearization we have made in the
previous
section consists of a renormalization ofcvÉ(0)
and00;(0) (to
make themtemperature dependent) by considering only
the low-order inter- actions betweenquasi-pseudo-magnons
andquasi- phonons.
A more
important objection
to the way we have introduced and renormalizedpseudo-magnons
inthe
previous
section is theinjustified
overextension(from
T -- 0 to TTo)
of thetemperature
domain ofvalidity
of theapproximation.
So ourprocedure
does not avoid the usual criticism of most methods
introducing finite-temperature
renormalized magnons.If however a
rigorous procedure
was found for per-forming
thisrenormalization,
thethermodynamic
coherence we want to
point
out(this
is to say thesimultaneity
of a second-orderphase
transition and thedynamical instability inducing it)
should remain.The case we have examined in this paper is a transition
essentially triggered by
an order-disorderinstability (protons
inKDP)
andaccompanied by
an induceddisplacive phenomenon (heavy
ions inKDP).
Theopposite
situation(where
adisplace instability triggers
the transition and induces the order-disorderphenomenon ;
or in anequivalent
way, when the transition is due to aquasi-phonon instability)
isnot
theoretically
forbidden(anharmonic
lattice termsshould of course be included in the
model)
and theexperimental
observation of such mixed cases should be veryinteresting.
Let us
finally
add that neutronscattering experiments performed
onKD2PO, [25], [26]
and(NH4)D2P04 [27]
show aquasi-elastic scattering intensity strongly dependent
ontemperature. However,
if the soft-mode
picture
isright,
theimportant damping
effects should be included in any best
theory
feborevaluable
comparison
withexperience
could be done.The author would like to express his sincere thanks to Dr. N. Boccara for various
suggestions
and criticalreading
of themanuscript,
to Dr. G. Sarma for very fruitful advice and to Dr. R. Bidaux for useful dis- cussions.Appendix
Let us consider
(from
the Hamiltonian(4. 5)
whenwe do not include constant
terms)
where
and
where
(+)
denotes the hermitianadjoint
matrix.Let us now consider the
following diagonalization problem
The existence of non trivial solutions
implies
hence
(we
notecvt
for( - )
andwq
for( + )).
The four
eigenvalues
are notedand the
eigenvectors
are notedLet us define the scalar
product
in thefollowing
wayand let us
denote 1 e’ 1
_+ ale% eq the norm of the
vector
eq.
Replacement
of the foureigenvalues
in(A .1)
with the
condition 1 e’ q 1
= 1(for
all values ofi )
determine
completely
the four realeigenvectors eq,
which constitute an orthonormal basis of the vecto- rial space.
If we note
we may
verify that 1 Tq 1
= 1 and(Tq +)-1 = (Tq 1)+.
.Hence we may rewrite
where
and
(Aq Bq A _ q B_ q)
its hermitianadjoint.
So if we omit constant terms we have
where all the boson relations are
satisfied, particu- larly
References
[1]
KANZIG(W.),
Solid StatePhysics,
ed.Seitz-Turnbull, Pergamon
Press,1957,
Vol. 4.[2]
JONA(F.),
SHIRANE(G.),
FerroelectricCrystals,
Pergamon Press, 1962.[3]
KOBAYASHI(K.), Phys.
Lett.,1967,
26A(1),
55 ; J.Phys.
Soc.Japan,
1968, 24(3),
497.[4]
BLINC(R.),
RIBARIC(M.), Phys.
Rev.,1963,
130(5),
1816.
[5]
TOKUNAGA(M.),
MATSUBARA(T.),
Prog. Th.Phys.
1966,
35(4),
581 ;ibid, 1966,
36(5),
857.[6]
BLINC(R.),
PINTAR(M.),
ZUPANCIC(I.),
J.Phys.
Chem.
Solids,
1967, 28, 405.[7]
BLINC(R.), Theory
of Condensed Matter, Internatio- nal Atomic EnergyAgency, Vienna,
1968.[8]
BLINC(R.),
CEVC(P.),
SHARA(M.), Phys.
Rev., 1967, 159, 411.[9]
VILLAIN(J.),
STAMENKOVIC(S.), Phys.
Stat.Sol., 1966,
15, 585.[10]
BROUT(R.),
MULLER(K. A.),
THOMAS(H.),
SolidState Comm., 1966,
4,
507.[11]
BLINC(R.),
SVETINA(S.), Phys.
Rev.,1966, 147, 423,
430.
[12]
NOVAKOVIC(L.),
J.Phys.
Chem.Solids, 1966, 27, 1469.
[13]
SILVERMAN(B. D.), Phys.
Rev. Lett., 1970, 25(2),
107.
[14]
DE GENNES(P. G.),
Solid State Comm., 1963,1,
132.[15]
MATSUDAIRA(N.),
J.Phys.
Soc. Japan, 1968, 25(5),
1225.
[16]
BAR’YAKHTAR(V. G.),
CHUPIS(I. E.),
Sov.Phys.
Sol.State,
1969, 10(12),
2818.[17]
BOCCARA(N.), Phys.
Stat. Sol.(b),
1971,43,
K 11.[18]
COCHRAN(W.), Phys.
Rev. Lett., 1959, 3(9),
412.Advanc.
Phys., 1960,
9, 387 ;ibid.,
1961,10,
401.[19]
ANDERSON(P. W.),
Izd. An SSSR Moscow, 1960.[20] BOCCARA
(N.),
SARMA(G.), Physics,
1965, 1(4),
219.[21]
VAKS(V. G.),
GALITSKY(V. M.),
LARKIN(A. I.),
Soviet
Phys. JETP, 1966,
51, 1592.[22]
IMRY(Y.),
PELAH(I.),
WIENER(E.),
ZAFRIR(H.),
Solid State Comm., 1967, 5, 41.
[23]
REIN(R.),
HARRIS(F. E.),
J. Chem.Phys., 1964,
41, 3393 and 1965, 42, 2177.[24]
SAMARA(G. A.), Phys.
Rev. Lett., 1971, 27(2),
103.[25]
SKALYO(J.),
FRAZER(B. C.)
and SHIRANE(G.), Phys.
Rev.,1970,
B1, 278.[26]
PAUL(G. L.),
COCHRAN(W.),
BUYERS(W.
J.L.)
andCOWLEY
(R. A.), Phys. Rev.,
1970, B 2, 4603.[27]
MEISTER(H.),
SKALYO(J.),
FRAZER(B. C.)
and SHI-RANE