• Aucun résultat trouvé

On considering the elastic constant table as the matrix of an operator ; consequences in ferroelasticity

N/A
N/A
Protected

Academic year: 2021

Partager "On considering the elastic constant table as the matrix of an operator ; consequences in ferroelasticity"

Copied!
17
0
0

Texte intégral

(1)

HAL Id: jpa-00246632

https://hal.archives-ouvertes.fr/jpa-00246632

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.

On considering the elastic constant table as the matrix of an operator ; consequences in ferroelasticity

A. Bulou

To cite this version:

A. Bulou. On considering the elastic constant table as the matrix of an operator ; consequences in ferroelasticity. Journal de Physique I, EDP Sciences, 1992, 2 (7), pp.1445-1460. �10.1051/jp1:1992220�.

�jpa-00246632�

(2)

Classification

Physics

Abstracts

64.70K 62.200 62.65

On considering the elastic constant table

as

the matrix of

an

operator ; consequences in ferroelasticity

A. Bulou

Laboratoire de

Physique

de l'Etat Condens£

(*),

Facult£ des Sciences, Universit£ du Maine, 72017 Le Mans Cedex, France

(Received 24 December 1991,

accepted

in

final form

17 March

1992)

R4sum4.-Les

consdquences

de la convention associde h la contraction des indices des composantes du tenseur des constantes

61astiques

C,~~i sont discutdes. On montre que, pour Etre

manipul£

comme la matrice d'un

op£rateur,

le tableau des constantes

dlastiques

doit Eke dcrit

avec de nouvelles conventions ddfinies par :

C$~

= C,~~t si a, fl

= 1, 2, 3,

C$~

=

/C~~~t

si

(a

= 1, 2, 3 et fl = 4, 5, 6), ou (a = 4, 5, 6 et fl

= 1, 2, 3) et

C$~

= 2C,~~t si a, fl = 4, 5, 6.

Los

C$

~ sont les composantes d'un tableau 6 x 6

[C"].

En

adoptant

pour les d£formations e,~ la convention u~ = e,~ si I

= j et u~ =

/

e;~ si I *

j

on conserve pour

I'£nergie

la for~ne

habituelle E

=

C$~

u~ u~. On montre que les valeurs

©~

per~nettant d'£ctire

I'£nergie £lastique

~ sous une for~ne

diagonale

E

=

£ ©~

[E ]~ (oh les E~ sont des ddfor~nations

sym£trisdes)

sort 2

~ _,

"

les valeurs propres du tableau [C"]

qui

peut ainsi Eke

qualifi£

de matrice des constantes

61astiques

les « valeurs propres » du tableau des constantes

61astiques

d6fini avec les conventions usuelles ne permettent pas d'£crire la forrne

diagonale

correcte dans le cas des groupes

quadratiques

4,

4,

4/m et dans le cas des groupes

rhombo£driques.

Ce rdsultat a des

consdquences

dans l'dtude des transitions

ferrodlastiques

propres ou

pseudo-propres,

ces transitions dtant associ£es au ramollissement de l'un des

C~. Quelques

arguments montrant l'int£rEt de l'utilisation, pour d'autres types de tenseurs, de la convention de contraction

proposde

sont

pr£sent£s.

Abstract The consequences of the convention associated with the contraction of the indices of the elastic constant tensor components C,~~i are discussed. It is shown that, in order to be handled like the matrix of an operator, the elastic constant table must be written with new conventions :

C$~

= C,~~t if a, fl = 1, 2, 3,

C$~

=

/

C,~~i if (a

= 1, 2, 3 and fl = 4, 5, 6) or (a = 4, 5, 6 and fl =1, 2, 3),

C$~

=

2C,~~t

if a, fl = 4, 5, 6. The

C$~

are the components of a 6 x 6 table

[C "].

Using

for the strains

e,~ the convention u~ = e,~ if I

= j and u~ =

/

e,~ if I * j, the usual

for~n

E=jC$~u~u~

for the elastic energy is

preserved.

It is shown that the ter~ns

C~

of the elastic energy written in a

diagonal

for~n E ~

=

£ C~[E~]~

(where the E~ are the 2

« =1

(*) CNRS U-R-A- n 807.

JOURNAL DE PHYSIQUE I T 2, N'7, JULY 1992 52

(3)

symmetry-adapted

strains) are the

eigenvalues

of the table [C "] that can be called matrix of the elastic constants; the

«eigenvalues»

of the elastic constant table defined with the usual conventions do not lead to the

right diagonal

for~n in the cases of the

tetragonal ~,

m

4,

4 groups and in the cases of the

trigonal

groups. This result has consequences in the studies of the proper and

pseudo-proper

ferroelastic

phase

transitions since

they

are associated with the

softening

of one of the

C~.

Some arguments

showing

the

advantage

of an extension of the conventions

proposed

in this paper to other kinds of tensors are

given.

I. Introduction.

The conditions for

stability

of a

crystal

with

respect

to elastic deformations have been

described

by

Born and

Huang [I]. They

deduced

stability

criteria in terms of

specific

conditions between the elastic constants. These conditions became of

special

interest in the field of structural

phase

transitions characterized

by

a deformation of the unit

cell,

the so- called proper

(and pseudo~proper)

ferroelastic

phase

transitions

[2]

sometimes called elastic

phase

transitions. Boccara

[3]

showed that such transitions result from the violation of the

stability

criteria

given by

Bom

Huang

and deduced that

they

are associated with the

vanishing

of some «

eigenvalues

of the elastic constants matrix ».

Aubry

and Pick

[4]

further deduced that such transitions must be associated with the

softening

of sound

velocity propagating

in a

particular

direction with a

peculiar polarization. Hence,

these transitions are of

special

interest since the critical behavior is

strongly

related to the

dimensionality

of the soft acoustic

subspace [5-9]. Fairly

few materials exhibit proper or

pseudo-proper

ferroelastic

phase

transitions

[10].

One of the best known

examples

is encountered in

BiV04

that has been studied in

great

detail

by

David

[11],

from both the

experimental

and theoretical

points

of views.

Although

the nice

approach given by

Boccara to determine the elastic constant combination that softens at a ferroelastic

phase

transition works in a number of cases, it has

already

been

noted that in some circumstances the result differs from the exact combination deduced from the calculation of the extrema of the sound

velocity [12, 13].

In this article it is shown that in

order to be handled like a matrix

(and

to lead to

meaningful eigenvalues),

the elastic constant table

requires

new conventions for the definition of its components. These conventions have been

previously

introduced in reference

[10].

2. Strain energy

density.

A

homogeneous

deformation of a

crystal

is described

by

the

components

of the strain tensor e,~. The strain energy

density

of the

crystal

can then be

expressed

in terms of these strain

components and the isothermal elastic constants

C,~~i

as

~

~

i I I I ~ljkf

E>j ~kf

(~)

~lj~li~lt~l

Since the strain tensor is

symmetrical

e,~

= e~~

)

and since the elastic constants are

symmetric

in the

ij

and

kf indices,

it is

possible

to use the

Voigt

notation where two indices

(ij)

are

replaced by

one

(a

with the

following

convention :

=

(i

+

j)

~ *j.

~~~

(4)

Using

this

notation,

the strain elastic energy

density

can be written with two summation indices

(instead

of

four) running

from one to six.

However,

to do

this,

additional conventions must be used to take into account the fact that, in summation

(I),

some terms appear twice or

four times.

Actually

two conventions are encountered :

in the more common one

[14],

the strain components

e~~ are

replaced by

e~ such that

~"

~ij

~~ ~j

~~~

e~ = 2

e~~ if *

j

while the elastic constants are

simply given by C]p

=

C~~;i

(~) ; then

(I)

reads :

~

~

j I I ~~P

~a ~P

(~)

~lfl~l

The

C(~

are the elements of a 6 x 6 table

[C

~].

some authors

[15, 16]

define the elastic

compliance (or

elastic modulil

by s[

p = si~~i and

so the elastic constants are

given

as follows :

C[~

= C~~~i if a,

p

=

1, 2,

3

if

(a

=

1, 2,

3 and

p

=

4, 5, 6)

C[~

= 2 C,~~y or ~~~

if

(a

=

4, 5,

6 and

fl

= 1, 2,

3)

C[~

=

4 C~~~i if a,

fl

=

4, 5,

6.

Equation (I)

then reads :

E

"

l I ~SP

eP

16)

«=,p=~

provided

the strain

components

are such that :

e~ = e~~ for I

=

j

and for I #

j. (7)

The

C[p

are the elements of a 6 x 6 table

[C~].

As shown in section

3,

in the

general

cases, neither the

[C

~] table nor the

[C

~] table can be

manipulated

like an

ordinary

matrix.

However,

if the

following

conventions are used :

u~=e~~

for

I=j

~ (8)

u~ =

2 ej~ for I #

j

(I)

The elastic compliances are then defined

by

:

s[~

= S,pit if a, fl = 1, 2, 3 if (a

= 1, 2, 3 and fl = 4, 5,

6)

S[~

= 2S,~it or

if (a

= 4, 5, 6 and fl = 1, 2, 3)

S[~

= 4S,~~i if a, fl

= 4, 5, 6.

(5)

C$~

=

C,j~i

if a,

p =1, 2,

3 if

(a

=

1, 2,

3 and

p

=

4, 5, 6)

C$p

=

/ C;y~i

or

(9)

if

(a

=

4, 5,

6 and

p

=

1, 2, 3)

C$~

= 2

C;~ki

if a,

p

=

4, 5,

6

equation (I)

reads :

6 6

E

=

j £ £ C$~

u~ u~

(10)

«=ip=i

and

then,

the elastic constant table

[C "]

exhibits new

properties

and it can be

manipulated

as an

ordinary

matrix. For reasons that will appear

below,

this convention can be called the

«

unitary

» convention or the «

symmetrical

» convention.

In the three above

conventions,

symmetry can further be used to reduce the number of

independent

elastic constants as described

by Nye [14].

VECTORIAL NOTATION. As

pointed

out

by

Boccara

[3],

formulations

(4) (as

well as

(6)

and

(10))

can be written in a shorter way if one considers the e~

(or

e~ or

u~)

as the

components

of

a six-dimension vector, say

e) (or [e)

or

[u)) using

the Dirac

notation,

and the elements

C(p (or C$p

or

C$p)

of the table as the

components

of the matrix of an

operator

C~

(or

C~ or

C~)

in a six dimension space. The strain energy

density

then reads :

E

=

(e[C~[ e) (II)

E

=

(e[C~[ e) (12)

E

=

(u[C~[ u) (13)

Although

this formal notation is

allowed,

the tables

[C

~] and

[C~]

cannot be handled like

ordinary

matrices since the « basis » where

they

are written is not normalized.

3. Action of the

symmetry operators

of the

point

group.

Let us now consider that the symmetry of the

crystal

is described

by

the

point

group

G

=

(R)

with elements R.

3,I MATRiCES OF TRANSFORMATiON OF THE COMPONENTS OF A vEcToR. To each

element R of G

corresponds

in the three dimensional vectorial space an

operator

R that acts on the vectors

(active sense).

This

operator

is

fully

defined

by

its action on the vectors of an

orthonormalized basis vi

)

,

u~)

,

u~)

;

R[u,)

3

= =

£ V(R)j; [u~) (14)

j=1

Then,

under a

symmetry operator R,

a vector

Iv)

such that :

3

[V)

=

£

U,

[Vi) (15)

<=1

(6)

is transforJned into a vector

iv')

:

Iv')

=

Rlv) (16)

with

components

v,'

given by

:

3

vi

=

£

v

(R

);~

vj (17)

j

The V

(R

);~ are the

components

of a 3 x 3 matrix

[V (R)]

and the set of matrices for the different elements of G forms a

representafiion

r~ of the group. This

representation

is

unitary (V (R

),~

= V

(R~ ~)j;)

since the matrices are

representations

of

(symmetry) operators

written

on an orthonormalized basis.

3.2 MATRICES OF TRANSFORMATION OF THE COMPONENTS OF THE

(e)

«VECTORS».

Under a

symmetry

operator

R,

the

component

e;~ of the strain tensor becomes

e(

such that :

3 3

e/j

=

£ £

V

(R

);~ e~i V

(R

~)i~

(18)

f i k i

I-e- the second rank

symmetric

tensor transforms like a matrix. Since the

[V (R )]

matrices are

unitary, (18)

can be written :

3 3

e[

=

£ £

V

(R

)~~ V

(R

)~i e~i

(19)

f I k i

which is the form the most

commonly

used.

A

symmetrical

tensor

(e~i

=

ei~)

can be characterized

by only

six components. In this case

equation (19)

also reads :

~ ~ ~

ei

=

z

v

(R

)~~ v

(R )j,

e~~ +

z z (v (R ),

~ v

(R )ji

+ v

(R ),i

v

(R

)j~

)

e~i

(20)

This relation can

altematively

be written :

3 3

e/j

=

£ £ T(R),~

~i e,I

(21)

k= i f»k

where the terms

T(R)~j

~i are

given by

:

~ ))' ~

=

~ )~ ~ )~

+ V

(R );I

V

(R

)j~

~

#

~~~~

The

T(R);~,~i

are elements of 6 x 6 matrices

[T(R)].

It can be shown that these matrices form a

representation r~

of the group G I-e- for any elements

R~

and

R~

of G :

~T(Ra)iiT(Rb)1~ ~T(RaRb)1. (23)

The characters

X(R)~

of this

representation

are those of the

symmetrized

square of the

r~

representation (characters

X

(R ))

:

~

(x (R )2

+ x

(R

2) x

(R ) (24)

= ~

(7)

The

[T(R)]

are the fix 6 matrices

required

to calculate the transformation of the six

components

of a

symmetrical

second rank tensor like the strain tensor under the action of an

operator

R.

Using

the contracted notation defined

by (7), equation (21)

reads :

6

ej

=

z

T

(R )j

p ep

(25)

«

where the

components T(R )[p

of the two index matrices are

easily

deduced from the four

component

ones

T(R)~j

~i

according

to

(22).

Vectorial notation. As mentioned in section

2,

the e~ can be considered as the

components of a vector

[e)

of a six-dimension space

spanned by

a basis

([ei),

[e~),

..,

[e~)).

Within this

framework, comparing (25)

with

(17),

it appears that the

[T (R )~]

matrices can be identified with the matrices of an

operator T(R

)~ and

by analogy

with

(16)

:

[e')

=

T(R)~ [e) (26)

3.3 MATRiCES OF TRANSFORMATiON OF THE

e)

« vEcToRs ». -The elastic strain energy

density

can be also written in terms of the strains e~ defined

by (3).

The e~ can be connected with the e~

using

a 6 x 6

diagonal

matrix

[A

with components A~~ defined

by

:

A,~

= &~~ if I,

j

=

1,

2 or 3

A,~

=

/

&,~ if I,

j

=

4,

5 or 6

~~~~

(where

&~~ =

l if I

=

j

and

&~~ =

0 if I #

j). Using

the vectorial notation as in

2,

the

components

of

e)

are related with the

components

of

[e) by

:

18> =

A~ ie>

It can be deduced that under a symmetry operator the e~ are transformed into

e[ according

to :

6

e[

=

£ T(R)[p

BP

(28)

a

or in short

e')

=

T

(R

)~ e

) (29)

where the

T(R )[p

are the

components

of the matrices

[T(R )~] given by

:

iT (R

)~

i

= iA

21iT (R )ei iA

21

(30)

The

[T(R)~]

are 6 x 6 matrices that forJn a

representation r~~

of the group G,

equivalent

to

r~ (or r~~, they

have the same characters.

3.4 UNITARY REPRESENTATION. In the

general

case, neither l~~~ nor l~~~ are

unitary since,

as seen in table

I,

the matrices of transformation of

e)

and

[e)

associated with element R~ are not

equal

to the

transposed

matrices of element R. On the other hand the matrices

corresponding

to the transformation of the components of the strain « vector »

[u)

form a

unitary representation r~~ They

are related to those relative to

is )

and

[e) by

:

iT(R)~i

=

iA-hi(R)~iiAi

=

iAiiT(R)eiiA-11. (31)

(8)

Table I. Form

of

the matrices

of transformation of

the components

of

the « vectors »

[e), e)

and

[u)

under the action

of

a

point

symetry operator R.

~~

2 V

(R

)ik V

(R

)~(

jj~

j~~ej

~

~~

~~~ ~ ~~~

(f

# k

)

V

(R )ik

V

(R

)~k V

(R

)~k V

(R

)~( + V

(R

)~( V

(R

)~k

g

*

g

# I and

(I

# k

~ ~~~ ~ ~

~~~

V

(R

)~

k V

(R

)~(

~~ ~~

~~

2 V

(R ),

k V

(R

)~k V

(R ),k

V

(R )( fi~R );(

V

(R

)~ k

g

#

I) g

#

I)

and

(I

# k

(R

)~~ v

(R

~~~

/

V

(R

);~ V

(R

)~i

~~ ~~

~~

/

V

(R

),~ V

(R

)~~ V

(R

),~ V

(R

~ ~(R

),

I V

(R

)~~

Q

# I

) Q

# I and

(f

# k

)

From

(31)

and

(22),

it is

easily

deduced that : if a, p =1,

2,

3

T

(R )j

p = T

(R )j,,

,~ = v

(R ),

~ v

(R ),,

if a

= 1,

2,

3 and p

= 4,

5,

6

T(R )1~

=

T

(R )i.

~i =

/

v

(R ),~

v

(R

)~i

(32)

if a

=

4, 5,

6 and

p =1,

2, 3 T

(R )$

p = T

(R )(,

» =

/

V

(R

);~ V

(R

)y~

j +,

if a =

4, 5,

6 and

p =1, 2,

3

T

(R )$

p = T

(R )(,

w = V

(R

);~ V

(R )ji

+ V

(R );I

V

(R )jk

j *,. t +i

The

unitary

of the

[T(R)~]

matrices follows from the

unitary

of the

[V (R)]

matrices.

3.5 INVARIANCE OF ENERGY UNDER THE ACTION OF THE SYMME~RY OPERATORS. if a

convenient « basis » is

chosen,

the elastic energy

density

can be written in a

diagonal

form as :

6

E

=

£ C[E~]~ (33)

~

a =1

(9)

The

C (having

the dimension of an elastic

constant) only depend

on the elastic constants

C;~~i

and are thus invariant under any symmetry

operator

of the

point

group

G;

the

E~ depend

on the strains e~~. The elastic energy

density being

invariant under all the

symmetry operators

of

G,

the

E~

must transform

according

to its irreducible

representations, they

are

symmetry-adapted

strains. In

particular,

if some

E~ corresponds

to a 2 or 3 order irreducible

representation,

the associated

C

value is

degenerate

2 or 3 times

respectively.

Let us consider the usual

expression

for the energy

given by (4)

based on the

e)

strain.

[C

~]

(like [C

~] and

[C ~])

is invariant under any

symmetry

operator of G while the

e)

are transformed into

e') according

to :

je'j

=

T(R)~ jej

and

je'j

=

jej ni~ (34)

where stands for the

adjoint

operator I-e- with the

transposed

matrix. The invariance of the energy then

requires

that :

E

=

(e[

C~

[e)

=

(e[ T(R)~C~T(R)~ e) (35)

so that :

C~

=

T(R

)~ C~

T(R

)~

(36)

Since,

in the

general

case,

T(R

)~ is not

equal

to

T(R

~)~, the above relation does not ensure

that the

«

operators

» C~ and

T(R

)~ commute I-e- it does not ensure that

they

have common

«eigenvectors

» : the

«eigenvectors

» of C~ may not

belong

to the «

symmetry adapted

strains », invariant

subspaces

of

T(R

)~. On the other hand

according

to

(31)

it can be deduced that :

T

(R

)~ = AT

(R

)~ A and T

(R

)~

= A T

(R

)~ A

(37)

since

[A]

is

diagonal. Then,

since

[T(R )~]

are

unitary, (36)

and

(37)

lead to :

AC ~ A

=

(T(R ~)~)

AC ~ A

(T(R

)~

) (38)

The above relation ensures that AC~A and

T(R)~

commute, and so a common basis of

eigenvectors

for these

«operators»

exists: the

eigenvectors

of the «operator» C~

=

AC ~ A transform like the symmetry

adapted

strains I-e- like the irreducible

representations

of G into which

r~~ (equivalent

to

r~~

and

r~~

reduces.

The

components

of the elastic constant table

[C"]

are deduced from the relation C~

= AC ~ A which leads to the values

given by (9).

The

[C

~] table can be

manipulated

like the matrix of an operator that commutes with the symmetry operators ; in the

following,

it will be called « the matrix of the elastic constants ».

It must be noted that the conventions

(9)

are

required

to

easily

calculate the terms

C

of the

diagonal

form

(31)

as the «

eigenvalues

» of a matrix. This convention is not necessary to calculate the conditions for elastic

stability

as described

by

Bom and

Huang [I]

and

given by Cowley [5].

3.6 EXAMPLE. The fact that the

«

eigenvectors

» of the usual elastic constant table may not be invariant under the symmetry

operators

of the

point

group is

readily

illustrated in the case of the

point

group C~~. The

decomposition

of the

symmetrized

square shows that the

symmetry-adapted

components of the strain tensor transform like :

2Aje2E.

(10)

Then the «

right

» elastic constant matrix must exhibit two

eigenvalues

with

degeneracy

two

associated with the two E symmetry strain components. A

quick

look at the character tables

[17],

or calculation

using

a

projection operator (2)

shows that the E

symmetry

strains

correspond

to :

a

b c

b -a d

c d

~

I-e- to the « vector » :

18E) =a(181) 182)) +(184) +j185) +(186)

The

(conventional)

elastic constant table

[C

~] for this

trigonal

symmetry reads

[18]

:

Cii C12 C13 Cm

° °

C12 Cii C13 -C14

° °

~

ci~ ci~ c~~

0 0 0

~~

~

c14 c14

0 c<J4 0 0

0 0 0 0 c<J4

c56

0 0 0 0

c5~ c~~

with C

~~ =

(C

ii

C12)/2

and

C5~

=

C

j~. Its «

eigenvalues

» C ~ can be obtained

according

to :

C~18E)

=

al EE)

and

they

appear as the solutions of two second-order

equations

:

C~~

(C

ii C

j2 +

C44)

C ~ +

(C

ii

C12) C44 2(C14)~

= 0

@~- (C66

+

C44)

C ~ +

C~~ C44 (C5~)~

= 0.

These two

equations

are different and lead to four different «

eigenvalues

» while two, twice

degenerated,

are

expected. If,

on the other

hand,

one considers the

[C~]

elastic constant

m_atrix,

one has to

replace C44, Cm, Ci~

and

C5~ by C~j

=

2C44, C~j

=

2C~~, C(~

/ C14

and C ]~ =

=

/

C (4

" 2

C14 respectively.

In this case the above two

equations

lead to a

single

one :

C

~~- (C

ii C

i~ + 2

C44)

C ~ +

2(C

ii C

j~) C44 4(C14)~

= 0

so that the

expected degeneracy

conditions for the

eigenvalues

C~ are satisfied.

4. Discussion.

4.I SOFT ELASTIC CONSTANTS ASSOCIATED WITH FERROELASTIC PHASE TRANSITIONS.

Boccara

[3]

showed that those

phase

transitions such that the order

parameter

is a

deformation of the unit cell

(the

so-called proper or

pseudo-proper

ferroelastic

phase

(2) The method is described in detail in [10] within the framework of the convention introduced in the present axticle.

(11)

transitions)

are associated with the

vanishing

of some elastic constant or of some combination of elastic constants

C

associated with

strain(s) E~

that does

(do)

not

belong

to the

totally symmetric

irreducible

representation

of the

point

group,

However,

as shown

above,

the

C

are

eigenvalues

of the

[C

~]

(given

in Tab.

II)

but not of the usual elastic constant table

[C~].

These

eigenvalues

of the

[C~]

are shown in table III.

It appears that the «

eigenvalues

» of the elastic constant table defined with the usual conventions are even

qualitatively

different from the

C

in the

trigonal point

groups and in the

tetragonal 4, 4,

4/m

point

groups. In the latter case, a ferroelastic

phase

transition can be

second order and so the

complete vanishing

of a

C

is

expected;

such a situation is

encountered in

BiV04,

a system

widely

studied. The

present

paper shows

why discrepancies

were observed between the

expressions

of the soft elastic constant as deduced from the

Table II. Elastic constant tables and elastic

compliance

tables as

defined according

to relation

(9) for

the

dijfierent crystallographic point

groups

(with

respect to the system

of

axes

defined

in

[18]).

For the omitted

cubic,

orthorhombic and

tetragonal

4 mm,

422, 12

m, 4/mmm

classes,

these tables have the same

form

as with the usual convention.

Hexagonal

classes

cl cj~ cj~

0 0 0

cj~ cjj cj~

0 o 0

cj~ cj~ cj~

0 0 0

0 0 0

c~4

0 0

0 0 0 0

c~4

0

0 0 0 0 0

cl -cj~

Tetragonal

classes 4.

I, 4/m

c

j

c j~ c j~ 0 0 c j~

C12

C

li

C

13 ° ° C

16

C 13 C13

C13

° ° °

0 0 0

c~j

0 0

0 0 0 0 c~4 0

C 16 C16 ° ° ° C

t6

Monoclinic

point

groups

C~l ~(2 ~~3

° °

~~6

~ (2 ~(2 ~ (3 ° ° ~ (6

~~3 ~(3 ~13

° °

~~6

0 0 0

C~4 C(5

0

0 0 0

C(5 C]5

0

C[6 C(6 C(6

0 0

Cj6

(12)

Table II

(continued).

Trigonal

classes

3, 3

~~l ~~2 ~(3 ~(4 C(5

°

~~2 ~~l ~(3 ~(4 C(5

°

C (3 ~ (3

~(3

~ ~ ~

~(4 ~~4

°

~~

~

~~~~

cj~ -cj~

o 0

C~j /C(~

o o o

/ca /~a

~a ~a

15 14 ll~ 12

Trigonal

classes 3 m,

32, 3m

C~l ~(2 ~~3 C~4

° ~

C~2 ~~l ~(3 ~C~4

° ~

~a13 ~a ~a o o o

13 33

~j~ ~j~

~

C~j

0 0

o o o 0

C~j /C(~

o o o

o

/cj~ cj~ _cj~

«

eigenvalues

» of the elastic constant table

[I Ii

and the values deduced from the extrema of the sound velocities

[12].

4.2 ELASTIC CONSTANT MATRIX AND ELASTIC COMPLIANCE MATRiX. In addition to the

fact that the conventions

presented

in

(8)

and

(9)

are necessary to handle the elastic constant table like an

ordinary matrix,

it has the

advantage

of

being

more «

symmetric

». The relation

between the strains and the stresses :

3 3

«,~

=

£ £ C,j~i

e,I

(39)

t i f i

reads with the usual conventions :

6

"

a "

£

c

j

p 8p

(40)

fl I

where the BP are defined

by (3)

and where the «~ are defined

by

a relation

equivalent

to

(7).

Using

the conventions defined

by (8), (40)

can be written with contracted indices as :

6

r~ =

£ C$p

Up

(41)

fl=i

where the conventions for the contraction of indices for the stress tensor

components

are

defined

exactly

like those for the strains I-e-

~~

"~J

~°~ ~~

(42)

r~ =

/

«~~ for I #

j.

(13)

Table III.

Components C ofthe

elastic energy associated with the

non-totally symmetrical symmetry-adapted

strains. The « vectors »

defining

the

symmetry-adapted

strains are

given

in the

first

column

(upper line) together

with

(lower lines)

the

corresponding

irreducible

representations (in parentheses) for

the

dijfierent point

groups ; in the monoclinic

point

groups,

the

[u~)

axis is

parallel

to the

binary

axis and/or

perpendicular

to the mirror

plane.

The

Care given

in column 2 with the usual notation

for

the elastic constants

C~p

m

C[p (upper line)

and with the convention

given by (9) (lower line).

The

degeneracy ofthe $

is

given by

the

order

of

the associated irreducible

representations.

Cubic classes

Strain

C~

Ul)

~

'~2)

~

'~31' '~I) '~2)

Cjj C12 T(E), Td(E), o(E)

~ ~ ~ ~

Ii 12

Th(Eg), Oh(Eg) l~~~~d~~~)~ ~~~),

~$~

~h(Tg), Oh (T2g)

~

Hexagonal

classes

Strain

C~

('~l) '~21' '~6) )

C6(E2), C6v(E2), D6(E2)

~

II

~12

~°3h(E'), 1~3h(E'), C~l

C(2

~°6h(E2g), 1~6h(E2g) ( ~4) ~5) )

C6(Ei), C6~(Ej), D6(Ej)

2

C~

~3h(E" ), D3h(E" ), C~j

~6h(Elg), 1~6h(Elg)

Trigonal

classes

3,

3

Strain

f

I

j ~c

~~ c

~~ + 2

c~)

±

[(c

ii c

i~ 2

c~)~

+

2

+

16(C

)~ +

C/5)

]~'~)

~~~ iii/i~lli/~'i~~

~ ~~~

i (C ii

C <~ +

C~i)

=

i(C ii

C <~

C~i)~

+

~

~(~l~

~

(14)

Table III

(continued).

Trigonal

classes 3 m,

32, 3

m

Strain

C~

j (C

ii C

i~ + 2

C~~)

±

i(C

ii C

i~ 2

C44)~

+

('~l) '~21' '~6)1'('~41' '~5))

16

C3v(E ), D3(E), D3d(Eg)

~ ~

(C ii

C [2 +

C$)

±

[(C ii

C [2

C~)

+ 2

2 1/2

~ ~ ~41

Tetragonal

classes 4

1,

4/m

Strain

f

(C

ii C

j~ + 2

C~~)

±

[(C

ii C

j~ 2

Cm)~

+

2

j~ j j~ j j~ j

+16

C)~]~'~)

l 2 , 6

~4~~

~'

~4~~

~'

~~~~~~~ (~

~l ~

~2 ~

~$6)

t

[(~

~l ~ ~2

~$6)~

+

+ 8 £~a16

jl/2j

( ~4), ~5) )

~

CM

C4(E), 54(E), C4h(Eg) C~j

Tetragonal

classes 4 mm,

422,

42 m,

4/mmm

Strain

~

'~l) '~2)

~4v(~l), ~4(Bl), D2d(Bl), ~~~ ~~~

D

4h

(B1 )

~

(l

~ (2

C~~(B~),

~~~~),

D~~(B~),

~

~~~

D~~(B~~) ~~~

( ~4) ~5) )

~ £~

C4~(E), D4(E), D2d(E),

~

C~j

D4~ (Eg)

(15)

Table III

(continued).

Orthorhombic

point

groups

Strain

~

U6)

~2v(1~2), D2(Bl),

~

~66

~2h(~ lg) ~$6

(U5)

~2v(l~l ), ~2(1~2),

~

j~~

~

(~ ~55

2h 2g

~~~~ 2

C44

C~~(B~), D~(B~),

~

~2h(1~3g) ~~

Monoclinic

point

groups

Strain

f

1u4>,

1u5>

1(c44

+

c55)

=

[(CM C55)~

+ 4

Cisi~'~i C2(B), C~(A"),

C~~(B ((Cj4+ C]5)

±

[(Cj4- C]5)~

+

4C()l~'~)

~ 2

It follows that the

components

of the elastic

compliance

table

[s~]

must be defined

exactly

like the elastic constant table components

[C "]

the form of the

[C

~]

given

in table II is also the

form of the

[s~].

Nye [14] already pointed

out that the existence of two conventions in

defining

the contracted form of the elastic constants

(C[~

and

C(p

in this

paper)

can make it difficult to compare data

published by

different authors. The choice of one notation rather than the other is

arbitrary.

More

recently

Sirotine and Chaskolskiia

[19] emphasized

that a

«

symmetric

»

convention would be

really

better but

they

did not use it since the data

given

in the literature would have to be corrected. The

present

article

brings

an additional reason for

choosing

such

a convention. The

advantage

of

using

a

symmetric

notation is not limited to the field of

elasticity.

For the different tensors

describing physical properties

it is

required

to know the

specific

conventions for the contraction

(3). Using

the

«

symmetric

»

convention,

a

single

rule would have to be defined I-e- any tensorial

quantity

Q~~__ = Q~~,_ with two indices I and

j running

from I to 3 would be transformed into a

quantity

with a

single

index Q~__

running

(3) The

necessity

of

introducing

a

multipling

factor 2 in the contracted form of the

piezoelectric

stress coefficients is not always mentioned [18].

(16)

from I to

6, provided

Q«,,

=

Q,j,,

if a =

1, 2,

3

Q~

=

/ Q,j

if

a = 4,

5,

6 ;

the process would have to be

applied

as many times as there are

pairs

of indices

(running

from I to

3)

to be contracted.

5. Conclusion.

It has been shown

that,

in the

general

case, the standard elastic constant table cannot be treated like an

ordinary

matrix. On the other

hand,

a «

symmetric

» convention associated to the process of contraction of the indices leads to a form such that the «

eigenvalues

of the

elastic constant table» are

meaningful.

The

«

symmetric»

convention makes it easy to transform the

expression

of the elastic energy into a

diagonal

form. This is useful in the

study

of the so-called ferroelastic

phase transitions,

a

variety

of structural

phase

transitions characterized

by

the

vanishing

of one of the components of the

diagonal

form for the elastic

energy. The results

presented

in the article

explain

the

origin

of the

discrepancies

encountered in the literature

conceming

the soft elastic constant in the ferroelastic

BiV04.

Taking

into account the above

considerations,

the

approach given by

Boccara

[3]

remains the

quickest

and best way to

predict

the soft elastic constants of a ferroelastic

phase

transition.

The above

problem conceming

the elastic

properties pointed

out a more

general problem

related to the use of «

non-symmetric

» conventions when

defining

a contracted form of a tensor. In this paper, it has been shown that the use of a «

symmetric

» notation may

give

the contracted form of a tensor additional useful

properties

; there is no doubt that a « non-

symmetric

» notation cannot do this. This

gives

an additional reason to

apply

a «

symmetric

» convention to any kind of

(symmetric

or

antisymmetric)

tensor, a notation that lies on a

single

rule and so, that would avoid the

knowledge

of the various conventions characteristic of each of them.

Acknowledgments.

Thanks are due to Prof. J. Nouet and to Prof. M. Rousseau for fruitful discussions.

References

[Ii BORN M. and HUANG K., Dynamical

theory

of

crystal

lanices (Oxford U. P., London, 1954).

[2] AIzU K., J. Fhys. Soc. Jpn 27 (1969) 387.

[3] BOCCARA N., Ann. Fhys. 47 (1968) 40.

[4] AUBRY S. and PICK R., J.

Fhys.

France 32 (1971) 657.

[5] COWLEY R. A., Fhys. Rev. B13 (1976) 4877.

[6] DAVID W. I. F., J.

Fhys.

C : Solid State

Fhys.

17

(1984)

385.

[7] FOLK R., IRO H. and SCHWABL F., z.

Fhys.

B 27

(1977)

169.

[8] FOLK R., IRO H, and SCHWABL F.,

Fhys.

Rev. B 20 (1979) 1229.

[9] FOLK R., IRO H. and SCHWABL F., Fhys. Lett. 57A (1976) 112.

[10] BULOU A., RoussEAu M. and NOUET J., Diffusionless Phase Transition and Related Structures in Oxides (Trans. Tech. Publications, 1992).

(17)

[iii DAVID W. I. F., GLAZER A. M. and HEWAT A. W., Phase Transitions1(1979) 155 DAVID W. I. F., J.

Fhys.

C Solid State

Fhys.

16 (1983) 2455 ;

DAVID W. I. F., J.

Fhys.

C : Solid State Fhys. 16 (1983) 5093 DAVID W. I. F., J.

Fhys.

C : Solid State

Fhys.

16 (1983) 5119

DAVID W. I. F. and WOOD I. G., J.

Fhys.

C Solid State Fhys. 16 (1983) 5127 DAVID W. I. F. and WOOD I. G., J.

Fhys.

C Solid State

Fhys.

16

(1983)

5149

[12] Gu BENYUAN G., COPIC M. and CUMMINS H. Z.,

Fhys.

Rev. B 24

(1981)

4098.

[13] CUMMINS H. Z.,

Light Scattering

Near Phase Transitions, H. Z. Cummins and A. P.

Levanyuk

Eds. (North Holland Publishing

Company,

1983).

[14] NYE J. F.,

Physical Properties

of

crystals

(Oxford U. P., London, 1957).

[15] WOOSTER W. A., A text-book on crystal

physics (Cambridge,

1949).

[16] WOOSTER W. A. and BRETON A.,

Experimental Crystal Physics

(Clarendon Press, Oxford, 1970).

[17] DRAGO R. S.,

Physical

Methods in

Chemistry

(Saunders

College Publishing, 1977).

[18] BHAGAVANTAM S.,

Crystal Symmetry

and

Physical Properties (Academic

Press, London and New York, 1966).

[19] SIROTINE Y. and CHASKOLSKAIA M., Fondements de la

Physique

des Cristaux (Edition MIR,

Moscou, 1984).

Références

Documents relatifs

Besides in surjectivity problems, Phragmen-Lindelof conditions also arise in the investigations of Meise, Taylor, and Vogt [14], [15], concerning the existence of continuous

Key words: Mulholland inequality, Hilbert-type inequality, weight function, equiv- alent form, the best possible constant factor.. The above inequality includes the best

In particular, we will prove that, under mild assumption on the base, the logarithmic Sobolev constant of the wreath product can be expressed in terms of the spectral gap of the

Just like the speed of light stops to be a fundamental constant when we express durations in meters, the constant b stops to be a fundamental constant if we use it to define a

(Monograph No. The ICIDH: a review of research and prospects. Report by the Committee on restrictions against disabled people. OFFICE DES PERSONNES HANDICAPEES DU

- In the gel phase, we adjust the value of the magnetic field H(G) in order to keep the birefringence level constant (An is here equal to 0.11 x 10-2) for

Conclusions – Our theoretical results show that variations in the value of the elementary electric charge, e, can occur under specific conditions, consequently producing

This study examines the short- and long-term post-merger performance of American Equity Real Estate Investment Trusts through three different methods: a market model, the Buy-