• Aucun résultat trouvé

Elastic anomalies at structural phase transitions : a consistent perturbation theory. I. One component order parameter

N/A
N/A
Protected

Academic year: 2021

Partager "Elastic anomalies at structural phase transitions : a consistent perturbation theory. I. One component order parameter"

Copied!
16
0
0

Texte intégral

(1)

HAL Id: jpa-00246675

https://hal.archives-ouvertes.fr/jpa-00246675

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.

Elastic anomalies at structural phase transitions : a consistent perturbation theory. I. One component order

parameter

A. Levanyuk, S. Minyukov, M. Vallade

To cite this version:

A. Levanyuk, S. Minyukov, M. Vallade. Elastic anomalies at structural phase transitions : a consistent

perturbation theory. I. One component order parameter. Journal de Physique I, EDP Sciences, 1992,

2 (10), pp.1949-1963. �10.1051/jp1:1992258�. �jpa-00246675�

(2)

Classification

Physics

Abstracts

64.70K 63.75 62.65

Elastic anomalies at structural phase transitions

: a

consistent

perturbation theory. I. One component order parameter

A. P.

Levanyuk ('.~),

S. A.

Minyukov (')

and M. Vallade

(2)

(')

Institute of

Crystallography,

Russian

Academy

of Sciences, 117333 Moscow, Russia (2) Universit6

Joseph

Fourier, Laboratoire de

Spectromdtrie Physique,

B-P. 87, 38402 Saint-

Martin-d'Hbres Cedex, France

(Received 28

February

1992, accepted in

final form

12 June 1992)

Rksumd. La th60rie

perturbative

des anomalies

dlastiques prbs

des transitions de

phase

structurales est rdexaminde. On montre que les

expressions

utilisdes le

plus

souvent pour

interpr6ter

les anomalies ultrasonores ne sont pas correctes, en

particulier

en dessous de la temp6rature de transition de

phase

T~. Une th60rie de

perturbation

auto-coh6rente est dtablie et on

montre que l'attdnuation du son est donnde par une

expression beaucoup

moins

simple

que celle

utilisde d'habitude. Los rdsultats

explicites

quant h la

d£pendance

de l'att£nuation ultrasonore en fonction de la

tempdrature

et de la

frdquence,

sont donnds pour les deux cas extrEmes suivants :

systbmes

ordre-d£sordre et

systbmes displacifs.

Pour les transitions ordre-ddsordre, qui ne se

produisent

pas trop loin du

point tricritique,

l'attdnuation est ddcrite par la forrnule de Landau- Khalatnikov (LK) oh l'expression du

param~tre

d'ordre n'est pas dorm£e par la thdorie du champ moyen. Pour les transitions

displacives,

la partie de type LK et celle due aux fluctuations ont la mEme

d£pendance

en

temp£rature

(dans la r6gion oh le calcul de

perturbations

est valable). En

cons£quence

l'att6nuation basse

frdquence

a le mEme indice

critique

pour (es deux

phases,

mais AT<r~

des

amplitudes critiques

diff£rentes. Le rapport de ces

amplitudes

tend vers z£ro au

point Ar~r~

tricritique.

En ce point (es indices critiques ne sont plus dgaux entre eux.

Abstract. The

perturbation theory

of elastic anomalies near structural

phase

transitions is revisited. It is shown that

expressions

more often used to interpret ultrasonic attenuation anomalies

are not correct,

particularly

below the

phase

transition temperature T~. A consistent

perturbation theory

is worked out and it is shown that the sound wave attenuation coefficient takes a form less simple in the general case than that

usually

assumed. The

explicit

results for the temperature and

frequency dependence

of the sound attenuation coefficient are

given

for two extreme cases : order-

disorder systems and displacive systems. It is found that for the order-disorder transitions, which are not too far from the tricritical

point

the main part of the sound attenuation

anomaly

can be

described by the Landau-Khalatnikov (LK) formula with the order parameter exhibiting a non- mean field behaviour. For

displacive

transitions for both LK and fluctuation, contributions have the same temperature dependence and the same order of magnitude within the perturbative

region.

As a result the low

frequency

sound attenuation coefficient has the same

« critical index » for the two

phases

but different « critical

amplitudes

», the ratio of the

amplitudes Ar~r~/Ar~r~ going

to

zero when the tricritical point is approached. At tricritical

phase

transition the

« critical index » in the

low-temperature phase

is different from that in the

high-temperature

one.

(3)

1950 JOURNAL DE PHYSIQUE I N° 10

1. Introduction.

Ultrasonic anomalies at structural

phase

transitions have been studied for several decades and a great amount of

experimental

data has been obtained

(see

I

]).

To

interpret

these data one often

uses the mean-field

theory

or a

perturbative

method which treats the contribution of the order parameter fluctuations as a small

parameter.

Such theories are

surely

invalid in the close

vicinity

of a second order

phase

transition but

they

may

provide

useful

approximations

outside this « critical

region

».

Although

the basic ideas of

perturbation theory

of ultrasonic anomalies

are

quite

old

[2, 6],

an exhaustive and consistent treatment has never been worked out. This has sometimes led

experimentalists

to make erroneous

interpretations.

The purpose of the present paper is to

give

a consistent

perturbation theory

for the case of a one-component order parameter. As we shall see the

expressions

which describe the ultrasonic

properties

are far from

being simple

in the

general

case. That is

why

we shall discuss in detail two extreme

situations :

displacive

and order-disorder

phase transitions, although

most

phase

transitions fall between these two limits.

The paper is

organized

in the

following

way in section 2 we make some comments on the

present theoretical situation relative to the kinetic coefficients near

displacive

and order- disorder

phase

transitions.

Special

attention is

given

to the temperature

dependence

of the soft mode

damping

coefficient which

plays

an

important

role in ultrasonic anomalies. The

incompleteness (or inconsistency)

of the

theory

is

pointed

out. In section 3 a

perturbation theory,

which takes into account first order fluctuation

effects,

in a consistent way, is

presented.

A summary of the results and a discussion of

possible experimental

verifications of the

theory

are

presented

in section 4.

2. Comments on the

theory

of ultrasonic anomalies.

The ultrasonic

properties

of

isotropic

media are described

by

a

complex frequency-dependent

elastic modulus

>(n)

=

>'(n)+

it

"(n).

The real part

>'(n)

is related to the sound

velocity

and the

imaginary

part >

"(n

to the sound attenuation. In the zero

frequency

limit,

>

(0)

is a

thermodynamic

real

quantity

and its temperature

dependence

can be

correctly

described in the frame of

thermodynamic perturbation theory.

In the ultrasonic

frequency

range, the sound

velocity

can thus be well described

by

such a

theory,

but it is not the case for ultrasonic attenuation which is a

purely

kinetic effect.

For

higher frequencies (typically

those

probed

in Brillouin or inelastic neutron

scattering)

both the real part and the

imaginary part

of >

(n

are not obtainable within the

thermodynamic theory.

Let us first recall the condition of

applicability

of the

thermodynamic theory.

For a one component

order-parameter

~, the Landau

thermodynamic potential

is written :

i~~~ ~~ ~~~~ ~~~~ ~~~~~j

~~

(l)

where A

=

A'(T- T~)

and B and D are assumed to be temperature

independent.

Both the Laudau

(mean-field) theory

and the first order

perturbation theory

are valid

only

when the

following inequality

holds

[7]

:

T T~

B~

T~

(r(

= ~

(2)

Tc A'

D~

The coefficients in

(2)

can be estimated for two

widely

used models. For the

Ising

model with

(4)

nearest

neighbour interactions,

which is the

prototype

oforder-disorder

system,

one finds

[8]

:

A'~l; B~T~d~; D~T~d~

where d is the lattice constant. The

inequality (2)

then reduces to r

~ l. As the

expansion (I )

is not

usually

valid for

large

r, one cannot define a range of r where the mean field

(or perturbation theory)

is

certainly

valid. It does not mean, of course, that this

theory

is never valid for order-disorder

phase

transitions : the real

systems

are

usually

much more

complicated

than the

Ising

model.

For a

displacive phase

transition

(see

e-g-

[9])

one finds :

A'~l; B~T~~d~; D~T~~d~

where T~~ is a

typical

« atomic temperature » (T~~

~10~ -10~ K).

Thus

(2)

reduces to

:

T~

r ~

(3)

T~~

and in this case one can define a range of Tin which the Landau

theory

is

valid,

since

T~«T~~. Strictly speaking

one has to confine oneself to the consideration of

displacive

transitions when

using

the

perturbation theory.

The criterion defined

by (2) is,

however, very

approximate

and a conclusion about the

applicability

of the

perturbation theory

can

only

be

reached after a

comparison

with an

appropriate

set of

experimental

data.

Let us consider now the

perturbation theory

for kinetic coefficients. For an order disorder

phase transition,

the order parameter

equation

of motion is of the relaxation type :

d~F~

yi

+

j

(~ (~)

D A~ = o

(4)

~~

eq

where

(~ )

is the

equilibrium

value of ~ and

F~

is the

homogeneous

part of the free energy

density

in

(I).

The coefficient y is

usually

considered as

temperature independent. Strictly speaking equation (4)

is

only

valid at low

frequencies.

The

general equation

contains

higher

order time derivatives or, in other words, yis

complex

and

frequency dependent. However,

we

are not aware of any reliable calculation of this

frequency dispersion

of y for order-disorder

systems

(this

is unfortunate because, as we shall see, the

high frequency dynamics

of ~ may be

important

for the low

frequency

sound

attenuation).

In the

following

we shall consider y as constant for order-disorder

systems.

For

displacive phase transitions,

the

equation

of motion of the order parameter is of the

oscillatory

type :

d~F~

m4 +

Y4

+ j

(~ (~))

-D A~ = o.

(5)

d~

eq

For small

anharmonicity

the

damping

coefficient y can be evaluated in the frame of

conventional lattice

dynamics.

Several authors

[9-16]

have contributed to the relevant treatment and

having

summarised their results let us

emphasize

first of

all,

that the

frequency dependence

of the

damping

coefficient y proves to be very

important

in the

displacive

case. For the

frequencies

which are close to the soft mode

eigenfrequency

w~

=

~/(d~F/d~

~)~~/m the

damping

coefficient can be

roughly

estimated as r

=

~~~°~~m ~~( ) ~

where

w~~~

m

~~

(5)

1952 JOURNAL DE PHYSIQUE I N° 10

10~~ s~ ' and n

m ÷ 2. This estimate is valid both for low and

high

temperature

phases.

Thus the width of the soft mode neutron

scattering

line is not

expected

to be different in the two

phases.

But as to the

damping

coefficient in the low

frequency region

the difference between the two

phases

becomes

significant.

For

T~T~

there exists a

specific

contribution to

i

y

(0)

which can be estimated as mw~~

~

~

/.

Thus it

acquires

an order of

magnitude

of T~,

T~

w~~ at the

boundary

of

applicability

of the

perturbation theory already,

and within the

~at

perturbation region

it is

expected

to be

dominating.

As it is the soft mode

anharmonicity

which is the

origin

of this contribution it can be

expressed through

the soft mode line width

(r)

and the coefficients of the Landau

thermodynamic potential [6].

In section 3 we shall recalculate this contribution to the order parameter

damping taking

into account term nonlinear in ~

(

~

)

in

equation (5) (in

the nonlinear version of

equation (5)

the coefficient y does not

incorporate

the contribution of the order parameter

fluctuations).

The

aim of the above remark is to

emphasize

that in the

displacive

case at

T~T~

the

« renormalization » due to

q-fluctuations

is more than the

« non-renormalized » parts of the

order parameter

damping

constant unlike to that of the order-disorder case when the

renormalization due to

~-fluctuations

is small as

long

as condition

(2)

is fulfilled.

The mean-field

theory

of sound attenuation

anomaly

was worked out

by

Landau and

Khalatnikov

(LK) [2].

This

anomaly

arises from a

coupling

between the strains

u~~ and the order parameter ~. For

isotropic solids,

the

thermodynamic potential density

F takes the form :

where K and p are the bulk modulus and the

rigidity

modulus

respectively. K,

and the

coupling

constant r are assumed to be temperature

independent.

At zero

applied

stress the

equilibrium

value of the strain is

given by

:

K(uii)

+

r(~ ~)

= 0

(7)

when an ultrasonic

longitudinal

stress with Fourier component tr~,

~

is

applied,

there is an additional

longitudinal

strain e~,

~

= ik uk, w, such that :

> e~~

~

+

r(~

~)k,

w "

"k,

w ~~~~

where >

=

K + ~

p is the

longitudinal

elastic constant. In the mean field

approximation

3

equation (8a)

is linearised

18k,w +~~~~)

~k,w ~k,w' ~~~~

Therefore, the ultrasonic response is

coupled

to the order parameter

only

for T ~ T~ in this

approximation. Taking

into account the kinetic

equations (4)

or

(5)

the

complex

elastic modulus

I (n

can be

easily

calculated.

For n « " one has :

m

i~~(n)

= >

4

r~j~l~

= >

~~~Lj

~~~

28(~) -iny

-i t

(6)

where

A>~~

=

~/

is the difference between the static elastic moduli

corresponding

to

clamped

and free order parameter, t = "

~

is the relaxation time. For nt « : 2 B

(~)

ij~=A>~~ntcc jT-T~j-~ (io)

Going beyond

this mean-field

theory approximation requires

the consideration of the non- linear part of the

coupling

in

(8a).

For

T~T~

the fluctuation

(~~~,~ plays

a role similar to ~~,~ in

equation (8b).

Its

contribution AA

~ to the static elastic constant is well-known

(see [7]) r~ Tc

~m ~ -1'2

(11)

~~F

~

2

~D3'2A~'~

~ ~

At low

frequency,

the fluctuation contribution to

I "(n

takes the form

>J(n

m

~ AAF Aw where Aw is the width of the central maximum in the

spectral density

of thermal fluctuations

~(

n

~)

and it

plays

a role similar to the

reciprocal

of the relaxation time t in

(9). (Let

us

note that this is the width of the

spectral density

of second order

light scattering).

Aw

depends

of course on the

dynamics

of the order parameter. For order-disorder

systems

one

has :

(Aw)~' ~Zcc [T-T~[~~ (12)

which leads to a temperature

dependence

of the ultrasonic attenuation coefficient

[3-5]

:

ij(n)cc jT- T~j-~'~ (13)

For a

displacive

system, the

equation

of motion is

given by equation (5).

The soft mode is

underdamped

in the domain of

validity

of the

perturbation theory

since,

according

to

(3)

y(W T

(

T T 1/2

~

m~ °~~~ j~

~ ~°~

~m °~~j

T~~

~ ~~~~

The

spectral density

of second order

light scattering

exhibits in this case a central

peak

with a width Aw

m r which is

approximately

temperature

independent.

Therefore :

ij(n)

cc

jT- T~j-"~ (15)

as first shown

by

Dvorak

[6].

For T

~

T~,

it is

usually

assumed that the fluctuation contribution

i~(n

is

nearly

the

same

as for T

~ T~ and it can be

simply

added to the mean-field contribution

i~~(n (Eq. (9)).

As

we shall see below this is not correct in the

general

case. Let us compare the

imaginary

parts of these two contributions. For an order-disorder system one finds :

it

AA

~ i

B~

T~ IQ

) >LK " /P @'

~~~~~

(7)

1954 JOURNAL DE PHYSIQUE I N° 10

This ratio is small when

inequality (2)

holds. For the

displacive

case :

it

TB

[A

~'~

_

(16b)

)

~ 2

«rY (°)

D~~~

(we

have used t

=

" ~~~

~

and

y(0)

m y

(w~) j~~ /,

within the

perturbation region).

2 B

(~ ) ~c

Thus the LK contribution has the same

dependence

and the same order of

magnitude

as the fluctuation one. To our

knowledge

this conclusion has never been taken into account in the

interpretation

of

experimental

data on

displacive

systems. It has been overlooked

that,

both in

the order-disorder and in the

displacive

cases, the LK contribution

depends

on material

coefficients which are temperature

independent only

in the mean-field

approximation.

To be

consistent,

the fluctuation induced correction to these coefficients has to be taken into account since

they

occur at the same order of

perturbation

as the « pure » fluctuation terms discussed above.

Zeyher actually

considered all these corrections in a calculation of sound attenuation in

incommensurate

phases [18].

As he focussed his attention

only

on the

phason contribution,

his results will be discussed in a

subsequent

paper.

3. Perturbation

theory.

For the sake of

simplicity

we consider an

elastically isotropic

solid and we assume that the Landau

thermodynamic potential density

has the form

given

in

equation (6).

A)

T

~ T~

The

dynamical equations

can be written :

X~/(k>

)

~k,

w

+ ~

~j ~kj,

wj

~k2,

w2

~k-kj

-k2,

w wj w2 + ~ ~

~j

~kj,

wj

~k-kj,

w wt ~

~

kj, wj kj. wj

k2. »2

(17a)

X

F/ (k,

)

Sk,

w

+ ~

i ~kj,

wj ~k kt. w ml "

~q,

fl

~k,

q

~ fl, w

(~~~)

ki WI

where the

x,~(k, w)

are the

uncoupled susceptibilities

defined

by

X~~ (k, w =

iA

+

Dk~

mw ~ i w y

(w )i~ (18a)

X~~(k,

w

=

> f

(18b)

k

(The damping

coefficient of the acoustic mode has been

neglected

in

(18b).

In first order

perturbation theory,

the second term of

(17a)

can be

neglected. Equations (17a, 17b)

can then be solved

(see appendix)

and one gets :

I (n

= > 4

r2 n(n (19)

with

J7(n)= ~jX~~(k,w)((~)n ~(~) =2T~jX~~(k,w)~~~~~'~ ~°~ (20)

k,w k,w

~ ~

The behaviour of

J7(n ) depends

on the

dynamics

of the order parameter.

(8)

For an order disorder

system (m

=

0,

y

independent

of

w)

the

integration

over k and w can be

performed.

One finds

[4, 17]

:

~~~

~

~~2 ~l/2'j~t l~ ~~~

~~~~

In the low

frequency limit,

one thus obtains :

I (n

> ~~

~c

in y

2

arD~'2

A~'2 ~

fi (22)

For the

displacive

limit ~

» r the main contribution to

J7(n )

comes from the

high

~m

frequency region

w

w~(k)

=

/

~

~~~

As a result one obtains :

m

~~

~

2

ar~~A"~

~

2(i~~r~)

~~~~

We see that in this case the

dispersion frequency (~ r)

does not

depend

on temperature and for

high frequencies (n

~ r

)

the attenuation coefficient has the same

temperature dependence

as

for the low

frequencies.

It is worth

mentioning

that in the low

frequency

limit Im

I (n

cc r~ '

and as r cc T one sees that Im

I (n

does

not contain the small factor T~/T~,. This

specific

feature of acoustic

damping

in a

weakly

anharrnonic

crystal

which was first

pointed

out

by

Sham

[19]

is

quite important

to bear in mind : it means that the sound attenuation anomalies are not small in the

displacive

limit while the

thermodynamic

anomalies are. Let us

emphasize

that in

spite

of the fact that Im

I (n )

does

not contain the small factor Tj/T~~, the corrections, due to the

higher

orders of the

hydrodynamic perturbation theory

used

here,

do contain it and the condition of

applicability

of the

perturbation theory

for the sound attenuation coefficient proves to be

practically

the same as that for

thermodynamic quantities.

B) T~T~

The static

equilibrium

value

(~)

and

(uii)

have now to be taken into account. To be

consistent,

they

have to be calculated up to the first order of

perturbation theory (see

e.g.

[9]).

Keeping

in mind the

possibility

of

applying

our calculations to tricritical transitions we add the term

C~~/6

to the

thermodynamic potential (6).

As a result one has :

A +

h(~)~

+ C

(~)~

+

(h

+ 2

h

+10 C

(~)~)((A~ )~)

=

0

(24a)

luff)

=

~

l~)~

=

~

(l~~)

+

lA~~) (24b)

with

h=B-~~~

and

B=B-~~

(24c)

and

((A~ )~) (which

is a function of

(~))

is to be calculated within the Landau

theory

but with the renormalized

phase

transition

temperature

which is identified with the

experimental

(9)

1956 JOURNAL DE PHYSIQUE I N° lo

one :

~

h

+ 2 B

(24d)

T~ =

Tc

~, a

where a is defined

by

the formula

j (A~ )~j

= a +

f (T

T~*

(25)

with

f(x)

=

0 at x

= 0. For a

phase transition,

which is far from the tricritical

point

one can put

C

= 0 and obtain :

f (T

T~

)

= b T

Tf (26a)

~

A'[T-Tf[ 12jj

(

~

)

= + l + ~ b T Tm

(26b)

h

B

In what follows we shall consider such transitions if not

specially

indicated.

The

dynamical equations

can then be written

X~/(k>

~k,

w

+ ~

~~~)

~k,

w

+

+

~j [3~~~) ~kj,wj ~k-ki,w-wi ~~~~kj,wt ~k-kiw-wjl

k,wt

+ ~

~j ~kj,

wj ~k~, w~ ~k kt k~, w wj w~ ~

°

(27a)

kb ml k~, w~

XEE~(k,

W Sk,

w

+ 2 r

(~ )

~k,

w

+ ~

~j ~kj,

wj>

~k-ki,

w wi

~q,

fl

~k,

q

~fl,

w

(~~~)

kj, wj

~~

~

is the Fourier component of ~

(~), X~~(k, w)

is the

low-temperature uncoupled

suiceptibility

:

X~~

(k,

w

=

iA

+

Dk~

+ 3 B

j

~

~j

+ 2 r

puny

mw ~ i

w y

(w )j~ (28)

As in the

previous

case, the last term in

(27a)

can be

neglected (except

the term 3 B

(A~ ~)

~~,

~

which can be

incorporated

into the first term of

(27a)).

The

equations (27a), (27b)

involve a linear

coupling

between ~~,

~

and e~,

~,

and

they

can be put under the

following

matrix form :

~~~~)

#

f)~

f)~lX

~k,w Xe~ XEE

~j [3

B

(~) ~kj,

wj

~k-kj,

w wi ~ ~

~~ki

wi

~k-ki,

w

l~

x

~"~'

(29)

~

~j ~kj,

wj> ~k kj,

w w1 ~q, fl

~k,

q

~fl,

w

kj, wj

where the

susceptibilities f((k,

w

correspond

to the

linearly coupled

system :

j0j X~/

2 r

(~ )

l~

~

~2rj~j

XiE~

l 4

r~(/)~

Xee X~~ 2 r

(~(~ee

X~~

~ ~

~~~~~

~~~

~~~~

(10)

Both components

~~

n and

s~

n involve a

part

driven

by

the extemal stress

tr~,

n and another

part

which

corresponds

to

th~

indirect influence of

tr~,

n via thermal fluctuations.

The latter contribution can be calculated with

equation (29) by keeping

in the non linear

terms of the

fight

hand side those which are linear in the driven components.

By incorporating

these results back into

(29),

one can then calculate the driven components

(see appendix).

At the end one gets :

~~' "

=

f~~ f~

~ ~

(31)

~q, fl X~e X

ee

"q, fl

where the renormalised

susceptibility

is

given by

:

lf~']~~

=

([28(~)~+Dq~-mn~-iny(n)] -4(~)~

x

~~2

~4

x

9 82 ni (q,

n 12

~ n~ (q,

n

)

+ 4 ~

n~ (q,

n A

)j

lf~

]~~

=

lf~

]~~

=

2 r

(

~

)

l 6

BJ7i (q,

n

)

+ 4

~

J7~

(q,

n

) (32)

A

lf~

]~~ = >

@

4

r~ Hi (q,

n

)

q with

°i

(q> lJ

)

"

~( i(~ (k,

n

) ~(

k, n w

(33a)

H~(q,

n

)

= >

~j I(~ (k,

w

~(_

~, n ~

~) lx~~(k,

w +

xee(q

k, n w

)j (33b)

H3 (q,

n

= >

~') i(~ (k,

w

~(_

~, n ~

~) lx ~~(k,

w

)

+ Keg

(q k,

n w

)j2 (33c) According

to the fluctuation

dissipation theorem,

one has :

~ ~ n

~

~)

= 2 T Im

~~

~ ~~~ ~' ~

(34)

]~ ]~~

is the «

clamped

»

reciprocal susceptibility.

Several authors

[13, 14]

have calculated this

quantity

for different cases, but

equation (32) gives

the

fully

consistent solution.

lj/~

]~~ can be identified with a

piezoelectric

coefficient in the

particular

case of a proper ferroelectric transition.

lj/~

]~~ has the same form as in the

symmetrical phase but,

of course,

J7i(q, n)

now involves the order parameter fluctuations of the low

symmetry phase.

The renormalized elastic constant

I (n

can be deduced from

equation (31), by taking

the limit q - 0 in

lee(q,

n

).

It is convenient to discuss the order-disorder and the

displacive

cases

separately.

Let us

begin

with the former one. As it has been mentioned in section 2 in this case the fluctuation corrections to the order parameter

damping

constant are small within the range of

applicability

of the

perturbation theory.

Thus one can

expand

the

expression

for

I (q,

n

)

in terms of

J7;(q,

n

).

One has :

I(n)=

>

~~l~l~

~

~~~

~

jj4B~]+my(n)j2n~(n)

2Bj~) -my(nj j2B~o-my(njj

~i~ ~i(4 B~ i

+ in Y

(n

)1

l/2(n

+

~l l~ ~' l/3(n )

(35)

where:

~/~A'(T-T/(/B.

JOURNAL DE PHYSIQUEi -T 2, N' IO, OCTOBER 1992 TO

(11)

1958 JOURNAL DE PHYSIQUE I N° 10

This

expression

is very

complicated

in the

general

case, and it can be

easily analysed only

in

some

limiting

cases. Let us assume that p

=

0. Then

lee(k,

w

)

= so that

>

n~(n) n~(n) nj(n)=

=

(see Eq. (33))

and

I(n)

takes then the

simpler

form

:

I(n)=> ~~~(~~~ 4r~J7i(n)l~~~~~~~~ ~

28(~) -iny 2B~o-iny (36)

Hi (n

has been calculated in

[16].

In the low

frequency

limit

°i~~l

»

~

(i +inyi

fl~° " 8D

(37)

with

r)

=

l~ (38)

2

h ~(

so that

I(n)

>

-2r21~

+

~~~ ~ j~j

n~o arD

n

~~2 II

~

~~~

h

~

l

~

~ ~~ ~

B~ ~/

4

arD~

B

Bh (39)

We see that the second term in the

imaginary

part of

I (n )

may be

negative

as well as

positive.

This term,

however,

remains small in

comparison

to the LK correction as

long

as condition

(2)

is fulfilled and the system is far from a tricritical

point.

Close to the tricritical

point

it becomes

quite important

since

h

goes to zero. One can note,

however,

that the contribution to

>

"(n arising

from the last term in

equation (36)

is

always finite,

even when

h

goes to zero.

Therefore the most

important

contribution comes also in this case, from the LK term but this includes

important

fluctuation corrections

through

the non mean field behaviour of the order parameter.

(This

result has been obtained with the

assumption

p

=

0. It is easy to see however that in the

opposite

case, p = co, one has

only

to

change h

into B in

equations (36)-(39),

so that the above conclusions are still valid one can

conjecture

that the above result is also true in the

general case.)

For the

displacive

case, one

has,

from

equation (31)

:

I (n

= >

~

~~~~~

4

r2 n~ (n (40)

li l~

~

]~~]~~

and

lf~~]~~

can then be

approximately

calculated in the two extreme cases p =

0 and p

= co. Let us first consider the case p

= co and

neglect

in

lf~ ]~~

both the term

-mn~ (because

acoustic

frequencies

are much smaller than

optical ones)

and the term

(12)

in y

(n (see

Sect.

2).

One has :

~

l 6

BJ7i (n

)~

>(n)=>

-2r -4r ~

ni(n). (41)

B(1-18BJ7j(n))

For small

frequencies

and within the

perturbation region

one can

expand equation (41)

as

function of

Hi (n ).

To a first

approximation

one

gets

I (n

=

> ~ ~~

16

r2 nj (n ). (42)

In the case p

= 0 one obtains instead of

(42)

:

I(n)=

>

~~~-16r2(~ )~nj(n). (43)

B B

This result has the same form as the y

- 0 limit of

equation (36), (but Hi (n)

has to be calculated for the

displacive case).

Let us

emphasize

once more that this result arises from the fact that the main contribution to the order parameter

damping

constant at small

frequencies

comes from the soft mode

anharmonicity (see

Sect.

2).

An

explicit

calculation of

Hj(n)

can

only

be obtained in the two cases p =0 or p = co. In the latter case, one can put

f(~

= X~~ and use

equation (23), substituting

A

by

28~/.

Then, for

phase

transitions far from

tricriticality,

the « critical indices

» for sound attenuation are the same for both

phases

and the

« critical

amplitude

» ratio is

CT

>T~

/ ~B

~~~~

It is remarkable that this ratio decreases when the tricritical

point

is

approached.

Close to the tricritical

point,

the mean-field order

parameter ~/

=

(A/C )"~

has to substituted into

equation (23).

The « critical indices » are then 1/2 for T ~ T~ and IN for T

~ T~, I.e. sound attenuation is

quite

different for the two

phases

near a tricritical

point.

The same conclusions also hold for p

=

0,

the

only

difference appears in the « critical

amplitude

ratio » which is in this case :

CT>T~ / B~

~~~~

One can

reasonably

expect that these results can be extended to the

general

case,

Let us also mention that, as for T

~

T~,

the

high frequency

and the low

frequency

sound attenuation show the same temperature

dependence.

4, Conclusion.

The main conclusion of this paper is that acoustic anomalies observed near structural

phase

transitions are less

simple

to

analyse

than

usually

assumed. The low

frequency

sound

velocity

is a

thermodynamic quantity

which exhibits an

anomaly

similar to that of the

specific

heat. The difficulties met in

analysing

these

quantities

have been discussed in detail in

[20]. Conceming

the ultrasonic

attenuation,

the temperature

dependence

of the

anomaly

is

expected

to

depend

upon the material constants which characterize the

dynamics

of the system, For T~

T~,

it

depends

on the

dynamical

behaviour of the order parameter: for low

frequencies

>"~

(13)

1960 JOURNAL DE

PHYSIQUE

I N° 10

IT

T~ "~ in the

displacive

limit and > "

IT

T~ ~'~ in the order-disorder limit. For

high

frequencies,

I.e, for

frequencies

which are more than the characteristic

dispersion frequency,

the temperature

dependence

of the attenuation coefficient in the

displacive

limit is the same as at low

frequencies

and is

practically

absent for what we called the order-disorder case

(purely

relaxational

dynamics

of the order

parameter).

Then the

frequency dispersion

is of

simple Debye-type

for

displacive

transitions and the characteristic

frequency

of

dispersion

does not

depend

on temperature while for order-disorder case the

dispersion

is

non-Debye

one, and the characteristic

frequency

is

proportional

to T T~.

For T

~ T~

things

are even more

complicated

since the ultrasonic attenuation also

depends

on

the

dynamics

of the elastic

subsystem

and on the

coupling

between the

order-parameter

and the strains. As a

result,

no

general

closed form

expression

can be found for the attenuation

coefficient,

even in the first order

perturbation theory. Relatively simple expressions

can

only

be obtained at the

price

of additional

assumptions

p

= 0 and p

= co and

again

for the two

extreme cases of the order

parameter dynamics

: pure relaxation and

underdamped

oscillations.

Qualitatively,

the difference between results for the two types of

dynamics

proves to be similar

to that for T

~ T~ with the reservation that there is an additional

Debye-type

contribution.

One has to pay attention to the fact that the conventional

representation

of the attenuation

anomaly

at T

~ T~ as a sum of the mean-field LK term and a

positive

fluctuation contribution is, in most cases, not

justified. Indeed,

for a

displacive

system the LK temperature

dependence

of the attenuation coefficient

(as

T T~ ~) is not

expected

at any temperature.

Moreover,

for

phase

transitions which are not far from the tricritical

point

the attenuation

anomaly

proves to be much more extended in the

high-temperature phase

than in the

low-temperature

one, I.e. it has the assymmetry which is

just opposite

to the « naive »

expectation.

For the order-disorder

system one can, to a first

approximation, si§nply neglect

the fluctuation contribution in the

low-temperature phase

as

long

as

h ~16~

(B fi) (see Eq. (39))

and derive the sound

B

attenuation

anomaly

at T~ T~ from

temperature dependence

of the order parameter. This

anomaly, however,

may be not a mean-field one because the order parameter may exhibit non- classical behaviour.

It would be

quite interesting

to compare the above

predictions

with the

experimental

data.

But this

comparison

is

already beyond

the framework of the present paper as well as discussion of the case when p is finite and non-zero and account for the elastic

anisotropy

which may be

quite important

for a

quantitative interpretation

of

experimental

data.

To

end,

let us not that in this paper we started with a set of

hydrodynamic equations

in which the temperature was not considered as a

hydrodynamic variable,

I-e- we made no distinction

between adiabatic and isothermal elastic moduli in the low

frequency

limit. This

difference,

which is

proportional

to the

specific heat,

is

expected

to be

particularly important

close to a

tricritical

point.

The introduction of the heat transfer

phenomena

into our

perturbation theory

would be

straightforward.

But it is

expected

to introduce

only

additional

quantitative changes

without any

qualitative

new features.

Acknowledgments.

The authors are

grateful

to Professor

Joseph Lajzerowicz

for fruitful discussions.

Appendix.

The non-linear

equations

which

couple

the

order-parameter

components ~~,~ and the

longitudinal

strain components e~_

~

can be solved in the

following

way :

(14)

A)

T

~ T~

Equation (17a) (without

the term

proportional

to

B)

leads to :

~~,~ =

~),~-2rX~~(k, w) ~j

e~,~

~)_~,~_~ (Al)

ki.wj

where

~),~

refers to the thermal fluctuation of the order parameter without the non-linear

coupling

to the strains.

When

inserting (Al

into

(17b)

and

keeping only

terms

quadratic

in the thermal

fluctuations,

one gets

8q, fl " X

ee(~>

~

) ("q,

fl ~

~j [~~,

w

~~

k, fl

w k.w

-4r~)~X~~(q-k, n-w) ~j

e~

~

~(_~_~~

n-w-wj

.

(A2)

kiwj

~'

Taking

statistical average over thermal

fluctuations,

one obtains

e~ n =

x~~(q,

n

tr~

n + 4

r~ ~j X~~(q

k, n w

)( ~)

~

(~)

e~ n =

lee(q,

n tr~ n

k,w

(A3)

with

iie~ (q,

n

= x

ie'(q,

n 4

r2 n(q,

n

) (A4)

and

n(q,

n

=

z

x

~~

(q

k, n w

)

~

j,

~

2j (A5)

~, »

Taking

the limit q - 0

~n2 n2

ii/(q,~l)

m

A(~l)-j=A(~l)-4r~1I(~l)-Pj. (A6)

q~o q q

B) T~T~

Equation (29)

is used to calculate the thermal fluctuation parts of ~~,

~

and e~,

~.

Keeping

only

terms linear in the driven components,

~~

~

and e~

~

one has :

l~~'

~ "

~l'

~

(k~(k,

W

))

X

~~~~ ~~~

6B(~)~~-q,w-fl

~q,fl

+~~[~~-q,w-fl

~q,fl +

~)-q,w-fl q,fllj

X

~~~)

2r~k-q,w-fl

~

~q,fl

where

~i,

w

and

ei,

w

are the thermal fluctuations in the absence of non-linear

coupling.

These fluctuations components are put back into

equation (29)

to calculate the driven components :

~~~~~ (i~(q, l))l~ ~~~~~~

~~'~

~~

~'~~~

~~~~k,w 8q_k ~_~]

q>fl

~

~

~~>

~

~q-k,

fl

w

+ ~T~ ~

~~~~

~'~

(15)

1962 JOURNAL DE

PHYSIQUE

I N° 10

Taking

the statistical average over thermal

fluctuations,

the second term of the

right-hand

side of

(A8)

is written.

[368~(~)~Pj(q, n)

+

24Br(~) P~(q, n)+4r~P~(q, n)] ~~

n

[12

Br

(

~

) Pi (q,

n + 4

r~

P

~

(q,

n

)]

e~,

n

Ii

2 Br

j

~

) Pi (q,

n + 4

r2

P

~

(q,

n ~

~, n

4

r2 Pi (q,

n e~, n ~r~, n

(A9)

with :

~l(~, ~)~ ~j i~~(k, t°)((~~-q,w-fl(~)

k,w

~2(~> ~)

"

~j li~~(k> t°)(8~-q,w-fl ~~-k,fl-w)

+

i~~(k> t°)((~~-q,fl-w (~)l

k,w

~3(~> ~)

~

~j li~~(k> t°)((8~-q,w-fl(~) +~i~~(k> t°)(8~-q,fl-w ~~-k,w-fl)

+

k,w

+I(e(k,w)([~)-q,w-n(~)l. (Aio)

Using

the

fluctuation-dissipation

theorem to evaluate the thermal averages, and

using equation (30),

one gets :

Pi

(q,

n

)

=

Hi (q,

n

(A

I

la)

P~(q,

n

=

~ ~

j~l n~(q,

n

) (Al16)

P3

(q,

n

m

~ ~~

~/

~

H3 (q,

n

)

+ ~~

~

(Al

I

c)

A A

with the definitions of the

J7,(q,

n

given

in

equation (33).

(A

term

~j

Keg

(q

k, w n

) ~)

~

~)

has been

neglected

in

(Al lc).

This term is

~ ~

>

exactly

zero when p

=

0 and it

gives

a

negligible

contribution when

Eq. (2) holds.)

When

equations (A8), (A9)

and

(Al I)

are combined one finds

equation (32).

References

II REHWALD W., Adv. Phys. 22 (1973) 721 (Ferroelectrics 12 (1976) 105).

[2] LANDAU L. D. and LIFSHITz E. M.,

Physical

Kinetics (Nauka, Moscow, 1979)

English

translation

Pergarnon

Press, Oxford.

[3] FIXMAN M., J. Chem.

Phys.

36 (1962) 310.

[4] LEVANYUK A. P., ZhETF 49 (1965) 1304 (Sov.

Phys.

JETP 22 (1966) 901).

[5] KAWASAKI K. and TANAKA M., Proc.

Phys.

Sac. London 90 (1967) 791.

[6] DVORAK V., Czech. J.

Phys.

B

21(1971)

836.

[7] LANDAU L. D. and LIFSHITz E. M., Statistical

Physics

(Nauka, Moscow, 1976)

English

translation Pergamon Press, Oxford.

[8] WHITE R. M., Quantum

Theory

of

Magnetism

(McGraw-Hill Book

Company,

New York, 1970).

[9] VAKS V. G. ZhETF 54 (1968) 910 (Sov. Phys. JETP 27 (1968) 486).

[10] STOLEN R. and DRANSFELD K., Phys. Rev. 139 (1965) 1295.

[I Ii GUREVICH V. L., Transport in Phonon

System

(North-Holland, Amsterdam-New York-Oxford) GUREVICH V. L, and TAGANTSEV A. K., Adv. Phys. 40 (1991) 719.

(16)

[12] TAGANTSEV A. K., ZhETF 80 (1981) 1087 (Sov.

Phys.

JETP 53, 555).

[13] BALAGUROV B. Ya., VAKS V. G, and SHKLOVSKII B. I., Fiz. Tverd. Tela12 (1970) 89 (Sov.

Phys.

Solid State 12

(1970)

70).

[14] COOMBS G. J, and COWLEY R. A., J.

Phys.

C 6 (1973) 121, 143.

[15] TAGANTSEV A. K., ZhETF 86 (1984) 2215 (Sov. Phys. JETP).

[16] LEVANYUK A. P. and SCHEDRINA N. V., Fiz. Tverd. Tela 16 (1974) 1439 (Sov. Phys. Solid State16 (1974), 923).

[17] GINzBURG V. L., SOBYANIN A. A, and LEVANYUK A. P.,

Light scattering

near

phase

transitions,

H. Z. Cummins and A. P.

Levanyuk

Eds. (North-Holland, Amsterdam-New York-Oxford,

1983)

GINzBURG V. L., LEVANYUK A. P. and SOBYANIN A. A., Phys. Rep. 57 (1980) 153.

[18] ZEYHER R., Ferroelectrics (1986) 66 217.

[19] SHAM L. J.,

Phys.

Rev. 156 (1967) 494.

[20] IVANOV N. R., LEVANYUK A. P., MINYUKOV S. A., KROUPA J., FOUSEK J., J.

Phys..

Condens.

Matter. 2 (1990) 5777.

Références

Documents relatifs

The constant tensor E ij is the average strain applied to the unit cell and v is a periodic fluctuation, meaning that it takes the same value at homologous points ( x − , x + ) of

Abstract 2014 The temperature variations of the order parameter invariants are measured in the uniaxial and biaxial nematic phases of different mixtures of potassium

amplitude modes in the biaxial phase mix the components of biaxiality and birefringence of the order parameter, the critical (quasi-biaxial) mode presenting

2014 We present results concerning sample dependence of not only the modulation wave vector, but also of the amplitude of the order parameter describing the N

When the propagators are self-consistently clothed, there is no divergence at 7k, and pertur- bation theory is valid all the way down to just above zero temperature.. Our resistance

Blue Phases I and II, occurring ina small temperature range between an isotropic and a cholesteric phase, have cubic structures with lattice constants of the order of wavelength

In this Appendix, we briefly review the theory of the de SQUID in so far as it concems the effects of trie intrinsic phase shift in YBCO-Pb SQUIDS. We wiii assume that in zero

2: Second proposed set-up: atoms are extracted from the cloud at x A,B and create a matter-wave grating when the wave-packets overlap at X; one detects this grating by shin- ing a