• Aucun résultat trouvé

Acoustic susceptibility of an insulating spin-glass in an applied magnetic field

N/A
N/A
Protected

Academic year: 2021

Partager "Acoustic susceptibility of an insulating spin-glass in an applied magnetic field"

Copied!
27
0
0

Texte intégral

(1)

HAL Id: jpa-00246341

https://hal.archives-ouvertes.fr/jpa-00246341

Submitted on 1 Jan 1991

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.

Acoustic susceptibility of an insulating spin-glass in an applied magnetic field

P. Doussineau, A. Levelut, W. Schön

To cite this version:

P. Doussineau, A. Levelut, W. Schön. Acoustic susceptibility of an insulating spin-glass in an applied magnetic field. Journal de Physique I, EDP Sciences, 1991, 1 (3), pp.415-440. �10.1051/jp1:1991143�.

�jpa-00246341�

(2)

J.

Phys.

11 (1991) 415-440 MARS 1991, PAGE 415

Classification

Physics

Abstracts

75.50Lk 62.25 + k

Acoustic susceptibility of

an

insulating spin-glass in

an

applied magnetic field

P.

Doussineau,

A. Levelut and,W. Sch6n

(*)

Laboratoire

d'Acoustique

et

Optique

de la Matidre Condens6e

(**),

Universitb Pierre et Marie Curie, Tour 13, 4

place

Jussieu, 75252 Paris Cedex 05, France

(Received

8 November 1990,

accepted16

November

1990)

Abswact. The

propagation

of

longitudinal

acoustic waves of

frequency

between 30 MHz and 800 MHz has been studied in the

insulating spin-glass (CoF2)05(BaF2)02(NaP03)03.

This was achieved in the temperature range 1.2 to 4.2K which includes the critical temperature T~ = 1.8 K, with an

applied

magnetic field up to 9 Teslas. The results are the

following.

i) The

velocity

shows an

anisotropic

behaviour. It

depends

on the

angle

between the field and the acoustic wavevector. ii) The initial

slope

of the

velocity

versus the square of the

magnetic

field which measures a non-linear

magneto-elastic

coefficient presents a very

rapid

variation

(a jump)

at 2~. ii@ Below 2~ the velocity presents a minimum, whereas the attenuation has a maximum for H around I T. iv) Above T~ the

velocity

shows a

complicated

behaviour

: a maximum followed

by

a minimum and

finally

an increase with the field. v) For

high

fields the

velocity

increases with the field, the lower the temperature the steeper the

slope,

and a trend towards saturation is observed at the lowest temperatures for the

highest

fields. In the same field range the attenuation decreases with the field. All these data are

interpreted

within the framework of the static

Sherrington- Kirkpatrick

model

including

two

spin-strain coupling

mechanisms a Waller mechanism and a field induced mechanism. It is shown that, besides terms

coming directly

from the two

coupling

mechanisms, crossed terms not

previously

taken into account are

important.

The value of the

jump

of the non-linear

magneto-elastic

coefficient, the minimum of the

velocity

below 2~, and the succession of a maximum and a minimum above T~ for the velocity are well

explained.

The

dynamical

effects are

briefly

considered for the

paramagnetic phase

in low field.

I. Introduction.

Because of time-reversal invariance the

coupling

of

spins

with an elastic strain is

generally

either

quadratic ~van

Vleck

mechanism)

or bilinear

ovaller mechanism)

in

spin

operators.

As a consequence, acoustic

experiments provide specific

information on

spins

in solids ; for instance

they

allow the measurement of

four-spin

correlation functions. This

interesting

(*)

New address : GEC-ALSTHOM, Division Transports, Tour

Neptune,

Cedex 20, 92086 Paris La Ddfense.

(**)

Associated with the Centre National de la Recherche

Scientifique (URA 800).

(3)

property

applies

to any

magnetic materials, including spin-glasses. However, probably owing

to technical

difficulties,

the acoustic studies of

spin-glasses

are rather rare

[1-3], especially

if their number is

compared

to the multitude of

experiments using

other

techniques [4].

The

situation has been

slowly changing

for the last few years and several measurements of sound

velocity

and attenuation have been

published [5-9],

in

particular

on

insulating spin-glasses.

Recently

we have

reported

a detailed acoustic

study

of a cobalt

fluoro-phosphate spin-glass (COF~)o_~(BaF~)o_~(NaPO~)o~

from both

experimental

and theoretical

points

of

view,

but limited to the case where no

magnetic

field is

applied [7].

In what follows this

study

is referred to as I. In the

present

article we

give

a

report

of an extension of the work described in I.

Experiments

on the same

compound

have been

performed

in a

magnetic

field up to 9 Teslas

(T).

Several new features have been observed :

I)

a minimum of

velocity

as a function of the field below the

spin-freezing

temperature T~ and a maximum followed

by

a minimum above T~

it)

at

higher

fields a

strong

increase of

velocity

which saturates at the

highest

fields

iii)

a

rapid

variation near T~ of the initial

slope

of the elastic constant as a function of the square of the

magnetic

field

iv)

an attenuation decrease when the field increases. The

velocity

data

are

interpreted

in the framework of the

Sherrington-Kirkpatrick

model

[10].

Tliis model considers

only

the static case.

Unfortunately

the

spin dynamics theory

used in I cannot be extended when a finite

magnetic

field is

applied.

As a consequence the

interpretation

of our

attenuation data is

only

schematic. Nevertheless we can

explain

most of our data

qualitatively

and even some of them

quantitatively.

Several

spin-strain coupling

mechanisms have been

considered, including

a

coupling

induced

by

the field which was

evidently

absent in I. Short

reports

of this work have been

previously published [I1, 12],

without

taking

into account the field induced

spin-strain coupling.

The

organization

of the article is as follows. We first present our

experimental

results which include a detailed

study

in very low fields and studies in medium and

high

fields

(Sect. 2).

After a short sketch of the

Sherrington-Kirkpatrick

model

(Sect. 3)

we

apply

the method to the calculation of the static elastic constant for any

spins

S m

1/2 (Sect. 4)

and

then,

with more

details,

for

spins

S

=

1/2 (Sect. 5).

A

dynamical model

valid

only

for S

=

1/2

in low fields is

presented

in section 6.

Finally

we discuss our data in the frame of these theories

(Sect. 7)

and

we conclude

(Sect. 8).

As

already

mentioned the present article is the continuation of another refered to as I. As far as

possible

we have used the same

symbols

in both. However this article is intended to be

self-contained.

2.

Experiment.

2.I METHOD. Ultrasonic waves were

propagated

in a

sample

of the

magnetic insulating

amorphous compound (COF~)o_~ (BaF~)o_~ (NaPO~

)o_~. In this material the

freezing

of the

spins

occurs at T~ =1.8 K

[6],

a critical temperature which is within the

explored temperature

range 1.2 K

< T < 4.2 K.

Longitudinal

ultrasonic waves with

frequency varying

from 30 MHz

up to 800 MHz were used. The

sample

was immersed in a

liquid

helium bath whose

temperature was

changed by pumping.

Two different set-ups were used. In the first one,

magnetic

fields up to 9 T were

produced by

a

superconducting

coil. In this case, the

magnetic

field was

applied parallel

to the acoustic wavevector. In the second case,

magnetic

fields up to 2.I T were delivered

by

an

electromagnet.

Then the

magnetic

field can be rotated in a

plane containing

the wavevector. In both cases the data were collected at fixed

temperature

as a

function of the

magnetic

field

applied

after the

temperature

was reached. A very

good

accuracy was obtained for the relative sound

velocity change

measurements. It was about 10~~ at 100

MHz, decreasing

at

higher frequencies

due to

stronger absorption

in the material.

Changes

in the ultrasonic

absorption

down to 0.02 dB cm~ can be detected.

(4)

bt 3 ACOUSTICS OF A SPIN-GLASS IN A MAGNETIC FIELD 417

2.2 RESULTS. The elastic constant is an even function of the

magnetic

field H.

Consequently

we

present

our data as functions of the square of H. The full

magnetic

field range studied in the

experiments

will be divided into three

parts

for

clarity.

2.2. I

High jieliig,

This domain concerns the field

higher

than 2 T

(H~

~ 4

T~).

The

velocity

results are shown in

figure

I for several fixed

temperatures.

The measurements were done

only

at the lowest

frequency (30 MHz).

The

velocity change

is first

nearly

linear with

H~

up _to

an upper limit which increases with temperature

(it

is

equal

to about

30

T~

at I.4 K and it is

larger

than 80

T~

at 4.2

K),

the lower the

temperature

the

larger

the

slope

of

AV/Vo

versus

H~.

At the

highest

fields the curves bend

around, showing

a trend

towards saturation of the

velocity change.

This behaviour is

clearly

observed

only

for the two lowest temperatures

(I.4

K and 1.8

K).

It is worth

noting

that the

asymptotic

value at 1.8 K is

larger

than it is at I.4K. In the same field range the attenuation decreases

roughly

as

H~,

as shown in

figure

2 for 86 MHz ultrasonic waves.

2.2.2 Intermediate

fieliig.

This domain stretches out from about 0.3 T up to 2 T. It was studied in detail

by varying

the

frequency

and the

angle

q~ between the

magnetic

field and the

ultrasonic wavevector.

v.30 MHz

~o ~~~ ~~~

~o°~

,3 K

oo ++ ~"~

~o# ~

#

DP I

°

"

~o° z °

+~ < °

4

++~ m fl

a

~

~ x

>

~

°

x <

~

a x m

~

_

+g XX~

~

fl

I 4 +~ i B x ~'

o xX

$

°

E

+ x~

«

~ uo

+ °

x a-B

,

+ ° x~ ~/

a $

+ ° x~ ~

° i

+ x~ "

°

°

x ~

~

x ~* M

+~ x *

x

_/

°>~

~~

w*** 4 2

~

°

+o x w*"

~~ ~~

~ ~

*"

o 20 40 60 0 2000 4000 6000

H~ (T~1 H~ lT~l

Fig.

I.

Fig.

2.

Fig.

I. Relative

velocity change

for

longitudinal

elastic waves of

frequency

30 MHz

propagating

in the

spin-glass (CoF2)o

5

(BaF2)o.2(NaP03)o.3

as a function of the square of the

magnetic

field for the full field range up to 9T. The wavevector is parallel to the field. A set of data was taken at the critical

temperature 2~ = 1.8 K

(o).

The other sets

correspond

to a temperature lower than T~

(+)

and two temperatures

larger

than T~

(x)

and

(*).

All the data are

arbitrarily

put at zero for zero field.

Fig.

2. Attenuation of

longitudinal

elastic waves of

frequency

86 MHz propagating in the

spin-glass (CoF2)05(BaF2)02(NaP03)03

as a function of the fourth power of the magnetic field for the full field range at T

=

2.3 K.

(5)

The effect of the

angle

is shown in

figure 3,

where the

velocity change

of a 148 MHz ultrasonic wave at IA K at various fixed values of the

magnetic

field is

plotted

as a function of the

angle.

The main results are the

following.

The

velocity change

can be

analysed

as

bV/Vo

=A

+Bsin~

q~ where coefficients A and B

depend

on the

magnetic

field

(their

variations are shown in

Fig. 4).

The effect is more

anisotropic

for

high

fields. It is

stronger

when the field is

applied perpendicular

and lower when the field is

parallel

to the ultrasonic

wavevector. Because the behaviours are

qualitatively

similar whatever the direction of the

field,

in the

following

the results will be

presented only

for a

magnetic

field

applied

perpendicular

to the ultrasonic wavevector.

o

T.I.4 w

w

v.14B MHz

o

~

x

+ + fl

w a

~ ~ > o

+ + /+ + ~ m

~

~/~

~

$ o

~

l

$ + + X X X X ~ ~

< a b4

x w fl

~

g

x x ,- o ° ~

x x x rJ W

x o b4 °

~ o a >

"

a w

o o w ~

x w m

, J >

, J z Do

, J m

°, Jo m ~ M

" ~,

_~~

J' x T-I 4 K ~

~ «

o

~ v.14B MHz

~ l'

~~

90 0 I 2

ANGLE (degreesl H~ lT~)

Fig.

3.

Fig.

4.

Fig. 3. Relative velocity

change

of

longitudinal

elastic waves of

frequency

148 MHz

propagating

in the

spin-glass (COF~)o~(BaF2)o.2(NaPO~)o~

as a function of the angle q~ between the

applied

magnetic field and the ultrasonic wavevector at the fixed temperature T= IA K. The

origin

of the velocity change is taken at its value without

magnetic

field. The different lines

correspond

to

3V/Vo

=

A + B sin~ ~, where A and B are free parameters

depending

on the

magnetic

field. The various

symbols correspond

to different values of the

magnetic

field

(+)

H= 0.4I T;

(x)

H=0.56T;

(o)

H

=

0.83 T ;

(*)

H

=

1.18 T.

Fig.

4. Parameters A

(o)

and B (*) describing the

anisotropy

of the effect of a

magnetic

field on the

velocity

of elastic waves in the

spin-glass (COF~)05(BaF2)02(NaPO~)03

as a function of the square of the field.

Figure

5 shows several

examples

of

velocity changes

at 148 MHz

plotted

as a function of the square of the

magnetic

field. For T

< T~ a remarkable feature is observed the

velocity

first

decreases,

then passes

through

a minimum and

finally

increases as

expected

from the

high-

field results shown in

figure

I. At T~ T~ the behaviour is

completely

different and more

complicated

: the

velocity

first

increases,

then goes

through

a maximum which is followed

by

a

minimum and

finally

increases

again.

The minimum is not observed below

H~

=

4

T~

for the

(6)

bt 3 ACOUSTICS OF A SPIN-GLASS IN A MAGNETIC FIELD 419

1

1

v .14B MHz

0 1 gi

~ x

~

)

X

X $

x

x ~

x b4

~

~

I

1 x14B

I

x ~

i.3 ~

~

j

~ i X

fi

x

~ ~ x ~

#

~ ~ x

x

j

~

~ <

x

_

~

fl

-I

x x x

o ~

a

o

»

o «

° fl

~

(

o o o 0

°269

c~

~ o

°2.3

~

jf

Do o a a

«

~ w

w z + ~

m

>

*

~ 506 Q

~ ~

~ ~

Z

+ *~

~

~ w

w w ~

~ w* * fl

* j

I

G

+

+

~

+ +

+ + +

4.2 +

+

o

~2 j~2j ~2 j~2j

Fig.

5.

Fig.

6.

Fig.

5. Relative

velocity change

of

longitudinal

elastic waves of

frequency

148 MHz

propagating

in the

spin-glass (COF~)o~(BaF~)o~(NaPO~)o~

as a function of the square of the

magnetic

field,

applied perpendicular

to the ultrasonic wavevector for different temperatures. The vertical shift for the various sets of data is

arbitrary.

The number near each set of data indicates the temperature in Kelvins. The critical temperature is T~ = 1.8 K.

Fig. 6. Relative

velocity change oflongitudinal

elastic waves ofvarious

frequencies propagating

in the spin-glass

(COF~)o~(BaF~)o~(NaPO~)o~

as a function ofthe square ofthe magnetic field measured at the

same temperature T = 1.3 K. The

velocity

is measured with respect to its value in zero field which is in fact

frequency dependent.

The number near each set of data

gives

the

frequency

in MHz.

highest temperatures.

The

frequency dependence

of the

velocity change

is shown in

figure

6 at 1.3 K. The minimum is shifted to

higher

temperatures when the

frequency

increases. The

variation at low fields is

independent

of the

frequency.

In the same field range the ultrasonic attenuation a shows also some

interesting

features.

Figure

7 presents some results transformed into the

imaginary part

of the elastic constant C

(3m~~ Co

=

~a~~~~

as a function of the square of the

magnetic

field. For

w 10

T< T~ the

imaginary part

of the elastic constant has a maximum before to decrease. For T ~ T~ it increases with the field up to

H~

= 4

T~.

The

frequency dependence

of the

imaginary

part of the elastic constant is shown at T

= 1.3 K in

figure

8.

2.2.3 Low

fieliig.

For low

applied fields,

the

velocity changes

have been studied in detail.

Some curves at a

frequency

of 269 MHz are shown in

figure

9. In this field range the

behaviour is

frequency independent

between 90 MHz and 800

MHz,

as can be seen in

figure

6. This is in contrast with what we have observed in zero field

experiments

in the same

frequency

range. In this

figure

all the curves at different

temperatures

have

arbitrarily

the

same value for zero field. This means that the zero field temperature

dependence

which has

(7)

I x

A~

(

(

i i$

g i x

~a~

~i

o + + ~

i i~

a a >

a

~ b4

x a ° o 3.6 f~ ~

x ~

~

o

ji

~9i

fl

Do fl

# x

1 ~

~

nD >

~ i z

x >

+

~ ~

~

o

o + 1 ~

x

q

x14B

+ ~x

~ ~

o m

4

m

x r

p

+x >

x

~_

a fl

i'~ fl

Bo2+x 3

b4 rJ

+ rJ

x x o

a x z

+ fl

x 269 ~

° ~

T.I.3K °

v.14B MHz x x

x Q

i~

"

o 5 ° 5

H~ lT~l H~ lT~l

Fig.

7.

Fig.

8.

Fig.

7.

Imaginary

part of the elastic constant

(it

is related to the attenuation

by

3m

~ ~

=

~'

a

~~ ~~

C w lo

of the

spin-glass (CoF2)o

5

(BaF2)o.2(NaP03

)o.3 as a function of the square of the

magnetic

field. The data

were obtained with

longitudinal

waves of

frequency

148 MHz.

They

are shown for two temperatures

given

in Kelvins

by

the number near each set of data. The zero of the vertical scale is

arbitrary

and

corresponds

to the value in zero field which is in fact temperature dependent.

Fig.

8.

Imaginary

part of the elastic constant for the

spin-glass (COF~)o_~(BaF~)o_~(NaPO~)o~

as a

function of the square of the

magnetic

field. The data were taken at 1.3 K for various

frequencies

of the

longitudinal

elastic waves. The number near each set of data

gives

the

frequency

in MHz. The zero of the vertical scale is

arbitrary.

It

corresponds

to the value without magnetic field which is in fact

frequency dependent.

been studied

previously

is discarded. This initial

slope

of

AV/Vo

as a function of

H~

is

completely

different below and above T~. It is

negative

for T< T~,

positive

for T ~ T~ and zero

just

at 7~.

3. The

Sherrington-llirkpahick

model and its

generalization.

Many

features observed in

spin-glasses

are accounted for

(at

least as a first

approximation) by

a model

proposed by Sherrington

and

ICirkpatrick (SK) [10].

These authors

represent

a SG as

an

assembly

of N

Ising spins coupled together by

random Gaussian

exchange

constants

/,

scaled

by

a standard deviation

$ $

and

a mean value

(IN.

The model may be extended to the case of

spins S~1/2

and the effect of a

magnetic

field H and a

ligand

field D may be

incorporated [13].

Then the Hamiltonian of the

spin

system is

3C=-~z/~S;)-g~pBH£S;-DjjS) (I,j=1,2,..,N ). (I)

~,,J

, ;

(8)

bt 3 ACOUSTICS OF A SPIN-GLASS IN A MAGNETIC FIELD 421

w

*

* * 3.0

*

*

*

*

~ 4.2

*I

.

~ ~

+ + +

a o

fi

a r

~ o o o

x rJ

~

i-B ° $

x

x

x I-a

i~

x o

x

v .269 MHz

o 04 OS

H~ lT~l

Fig.

9. Relative velocity

change

of

longitudinal

elastic waves of

frequency

269 MHz

propagating

in the

spin~glass (COF~)o

~

(BaF~)o ~(NaPO~)o

~ as a function of the square of the

magnetic

field in a reduced field range

showing

the temperature

dependence

of the initial

slope.

The number near each set of data

gives

the temperature in Kelvins. The critical temperature is 2~ = 1.8 K.

As

usual,

g~ is the Land6 factor and pB is the Bohr magneton.

Using

the famous «

replica

trick » the

averaged

free energy is calculated and it is found to be :

fl

=

NkT( p$m~+ (pjj~ ~p~- q~) E(In try e~~~ ~~~)) (2)

with

0~

=

p $

m +

p lv~

t +

pgL

MB H q~~

=

(pjj~ ~p

q +

pD.

2 The Gaussian

expectation

value is defined as :

~"~

"

fi

~~

~~

~~~~~~

The values of the three parameters m, p and q have to be

self~consistently

determined for every value of the

temperature T,

the

magnetic

field H and the

ligand

field D

by

:

m =

E( (S~~)

=

£ (S;)

; q

=

E

((S)()

=

£ ($)~

P#E(lRj~)=(z(().

The

symbol ( )~

denotes a thermal average over a

one~spin

space while the upper bar indicates an average over the disorder. The effective operator

p3C~j

=

0~

S + q~~

S~

which

occurs in the definition of the average

( )

is used for any other

higher

order

parameters.

(9)

For the calculation of an elastic constant in a

magnetic

field we need

three~spin

correlation functions

(in

this case

they

are not

equal

to

0)

and

four~spin

correlation functions :

~

"E(ls~lj)

=

z@

h

=

E( ls~l

i

lsl j)

=

( z (S~) IS,)

~ ~

~( j ~ 3) ~ ~3

fi~

,

r =

E( ls~l i)

=

( z (Sf)

s =

E(iS~>1)

=

z@~

,

~

j~( j~2j j~2) ~ j~2j j~ j2

j fi~ I ,

,

~ ~

E(lsll)

"

i lsl~

,

w

=E(ls~lj lslj) =~zfi)

It is useful to put :

f

=

3 v + s 4 u. A

higher~order

correlation functions is of interest

x =

E( (S~()

=

jj ($)~

We recall that the

freezing temperature

T~ of the

spins (spin~glass transition)

and their

ordering temperature

To are related to the

spin-spin

interaction constants

land $

respectively.

For S

=

1/2

the relations

simply

read :

1=4kT~ $=4kTo.

When we write some

quantities (such

as

/ $,

g~,

D,

T~ and To) without upper

index,

this

means that an elastic strain may be present. On the contrary the upper index 0

specifies

that

no strain is

applied.

4. The static elastic constant in the SK model for any

spins

Sm

I/2.

The

magnetic

contribution to the elastic constant C is :

3C

=

~~~ l(3)

il dE

e =o

~lJ is the volume of the

sample

while E is a static elastic strain.

It is assumed that the two

parameters

which characterize the

spin~spin

interactions become

I= j°(I

+

yE)

and

I= $(I

+

yE)

when the strain

E is

applied.

In the same manner the

Landb factor and the

ligand

field tum out to be g~ =

g((I

+

fE)

and D

=

D°+

GE

(10)

bt 3 ACOUSTICS OF A SPIN~GLASS IN A MAGNETIC FIELD 423

respectively.

y and

f

are dimensionless coefficients G has the dimensions of an energy.

Then the

strain-dependent

free energy reads :

~~~~ ~~~~ ~~~~~~

~

~'~~ ~~~ ~~~~~

~~ ~

~'~~~ ~~ ~~~

E

(In trj

e~~~~~~ ~ ~~~~~

~))

(4)

with

0H(E)

=

P4(1+ y~)m+ pJ°(i

+

y~) It

+

pp~g[(i

+

f~)H

wD(E)

=

(PJ°)~ (i

+

y~)2 ~p-q)

+

p(D°+ G~).

Now the three averages m, p and q

depend

on H and E.

It is easy to calculate the derivatives of the free energy with respect to the strain :

$

=

NkT( ) jp $m2+ (pJ°)2 (1+ y~)~p2- q2)j

+

fpg[

p~ Hm +

pGp (5)

((

=

NkTj y2(pJ0)2 ~p2 q2)

+

pG )

+

+

IYPJ°Yn +fPg[

MB

HI ~

+

Y(PJ°)~ (P )

q

) (6)

The

following

calculations can be put into a compact form with the

help

of a

(3

x

3) symmetric

matrix

£,

whose elements are functions of the

spin

averages

(see Appendix A)

calculated for E

=

0. The relative

change 3C/Co

of the elastic constant C

=

Co

+ 3C turns out to be :

+

2(fl~)~ (fl~)(L12

~IIP

L13 ~ll~)

+

(fl~)~ (L22P~

2

~23P~

+

L33 ~~)l

+ 2

Yf(fl~~@B lf)[(fl~)Lll

~ll +

(fl~)~ (L12P L13~)1

+ 2

Yfl

G

((fl~) L12

~ll +

(fl~)~ (L22P L23 ~)l

+

f~(fl ~~

@ B

H)~ Ll1

+ 2

f (fl~i

@

B

H) fl GL12

+

(fl

~)~

L22)

~~~

JV

=

N/~lJ

is the number of

spins

per unit volume and

Co

is the elastic constant in the same

material without

spins.

The matrix elements L,~ are

given

in

Appendix

A.

A

physical meaning

can be ascribed to the different terms of

3C/Co.

The first term

(proportional

to y is the contribution of the Waller mechanism

(modulation

of the

spin~spin

interaction

by

the

strain)

its Hamiltonian reads

[14]

Jcw=-~yszfs,j.

~

,.j

The last term

(proportional

to

G~

is due to the Van Vleck mechanism

(modulation

of the

ligand field)

; the

corresponding

Hamiltonian is

[15]

:

3Cvv

=

GE

jj $I.

(11)

This

operator

becomes a scalar

(equal

to

-NGS/4)

for

spins

S

=1/2.

The fourth term

~proportional

to

f~

comes from the modulation of the Landd factor

by

the strain. It vanishes when no

magnetic

field is

applied

this is a field induced mechanism. The

corresponding

Hamiltonian reads

[16]

3CFI ~

fg[

MB HE

I $

,

The other terms

originate

in cross~effects. Some of them

disappear

in absence of

magnetic

field.

We recall that the

magnetic susceptibility

x is defined as :

x =

$ (8)

and it reads

X

~~(~~@B)~flLll (9)

This

explains

the

similarity

of the

magnetic susceptibility

x and the field~induced contribution to the elastic constant

(3C/Co)~i.

We

emphasize

that the

spin~averages appearing

in the matrix

£ depend

on the

magnetic

field H.

Finally

we underline that the

expression

of

3C/Co

is valid

only

if the

replica symmetric

solution of the SK model

is,

I-e- above the

generalized

de Almeida~Thouless line

[17, 18].

5. The static elastic constant in the SK model for

spins

S

=1/2.

5,I GENERAL RESULTS. For

spins

S

=

1/2 (with eigenvalues

±

1/2)

the

equations

obtained in the

previous

section become

simpler

because several elements

L;j

are then

equal

to zero.

However the

equations

remain awkward. This is due to numerous coefficients

(1/2)~.

These coefficients are eliminated if

spin operators

s, with

eigenvalues

± I are used. From here we

adopt

this rule as we did in I for the

spin dynamics.

Therefore all the averages have their value between 0 and I instead of 0 and

(1/2)~

for a

n-spin

average. We

distinguish

the new averages

by

a circumflex and we have :

Jii=2m,@=4q, fl=I,d=8a=Jii,$=8b=Jii,d=8c,f=I,§=I,

Q=16u=

@,fit=16w= @,b=16v,I=3@-4f+1,I=64xandj=256y.

Similarly

the new matrix elements

I;y

are related to the old ones

by

:

~°11 "

4

Lll>

~°13

~ 81~13> ~°33 "

161~33

>

while the other three elements are

equal

to zero. More

explicitly

the new set of elements reads :

L12

=

°1 L23

"

°1 L~

= 0

iii

"

((1 4)(1 t/0~) 2(Jii d)~/0~) IA

i~~

=

2

jii g(i q)/oi I

2

g(d e)2/oj IA

£i~

=

2(~ii e)/A

; A

=

(1 fle2) ii

g

(1 4)/oi

+ 2

g(~ii e)2/03

(12)

bt 3 ACOUSTICS OF A SPIN~GLASS IN A MAGNETIC FIELD 425

where

=

T/7$

and g

=

T~/l~° (-

co < g <

).

With these new definitions the

change

of the elastic constant reads :

~~

=

~~~~~

(y~[(1 #~)

+ 2

g~ iii

~ii~ 4

gij~ di#/@

+ 2

i~~ f~/0~]

+

Co

2

Co

°

+ 4 y

fh [g0ijj

Jii

ij~ @]/0

+ 2

f~ h~ijj) (10)

h is a reduced

(dimensionless) magnetic field,

defined

by

h=g[MBH/2k7~.

As

expected

the Van Vleck

mechanism,

as well as the cross-effects where G

appeared,

have vanished. To our

knowledge

the cross-effect

proportional

to

yf

has never been taken into account

previously.

It is

important

in our acoustic

experiments

because

they

were

performed

on a system which combines two conditions : a

high spin density sample

and an

applied magnetic

field.

It has been shown in I that 3

C/Co

must be

negative

for any temperature. This property still holds if a

magnetic

field is

applied.

Indeed the

magnetic

contribution for an

arbitrary

value of the

magnetic

field can be obtained

only by

the means of a

computer. However,

for very low and for very

high fields, analytic expressions

can be calculated.

They

are

reported

in the

following

two sections.

5.2 LOW FIELD EXPANSIONS. Here the case

S=1/2

is

considered,

with

$#0

and

p

g~ p

B H « I. Since the elastic constant is an even function of the

magnetic field, incre~sing-

power

expansions

of 3C

only

contain terms H~ where n is an even

integer.

In fact we limit

ourselves to terms

proportional

to

H°, H~

and

H~.

Indeed the term

independent

of

H has

already

been calculated in I. We have calculated the terms

proportional

to

Hi.

The results

are so much

complicated

that we do not

give

them in the present article. We

only

mention that some of them

diverge

when

approaching

the critical temperature both from

above and from below.

Our calculation is

performed

in two steps : the

paramagnetic phase

first and then the

spin~

glass phase.

We

give

them in terms of the dimensionless reduced parameters 0 and h.

5.2.I

Paramagnetic phase.

The

expansions

of the averages ~ii,

#,

d and are as follows :

~_

0~ ~2

'~~(@-g)

' ~

(@-g)~(@~-l)

~2

~~

~

~4~

~

~2

~

~

~

(@

-g)~ (@~- l)~

'

~

(@

-g)~(@~- l)~~

The matrix elements

I;y

limited to the lowest order are :

~

0

~ 20~

~_

~ 20~

~~

(° gl'

~~ (@

g)~ (@~- l)

' ~~

(@~-

l

The sum of the three contributions to the relative

change

of the elastic constant is :

~ ~ ~~~ ~~'~~

(o

~

g~31°'~+f(~ ~g)l~h~+. (ll)

(13)

We notice that the contribution

proportional

to

h~

is

always negative

since 0 g ~ 0 and the other factor is a

squared quantity.

5.2.2

Spin~glass phase.

The averages Jii, @, d and have the

following expansions

~*

~

lo

g

Ii 40)1~ ~

~

~

io

g

(i <o)12 (

o2

to)

~

d

=

~° ~°

h

=

to

+

~~~ ~° ~°~

~~

h~

1@

g(1 40)1

~

je

g

(1 q0)l~ (@~~ 10)

with

ko=2-17@o+30@o-153~.

The lower index 0 means that the averages are calculated for h

= 0. The matrix elements

I;~

are :

(1 qo)

e

to e2(e2

+ 2

to)

~~~ ~

l~

g

(1 40)1 lo

g

(1 40)l~ (0~- to)

~

2

lo

0

Lj~

= h

lo

g

(1 40)l~ (02 lo)

2

i~

o2

o210

g

(1 40)1 to

+ g

(o

2

i~)

11

~ ~

L33 = + 4 0 h

(°~ ~o) lo

g

(I

@o)l~

(0~- io)~

The total contribution due to the three processes is :

3c sG

JekT) i 0~

+ 3

lo ~

(

f~0 ° ~ ~'

o 2-

i~

~~

~

+

y2(g2(1 q~)3

o 4 g

~°~~ ~°~ ~°

o2

02- to

~ ~

4i kol° g(1 40)1

~4

~

40 li(° g)

~~

(02- to)~ (02- to)2

4010

g

(1 40)1to ~~li

02 i~ io g(i q~)13

g(1- 40)2 (02- i~)

2

i~q~

o

~ ~

(i q~)

o

~~

~~

~ ~

~'~

g

(1 40)l~ (°~ lo)

~

~

° ~

(~ 4°)

~'

5.2.3 Behaviour around T~. In order to

study

the behaviour of

(3C/Co)

near

T~

(I.e.

for 0

m

I)

we put 0 = 1+ 3

(then (3 «1).

We get in the

paramagnetic phase

(3~0):

(14)

bt 3 ACOUSTICS OF A SPIN-GLASS IN A MAGNETIC FIELD 427

~j)~~-~)y~(1-3+.

+

~~~ 6g~

~~

j~~

° ~ o

(I-g)~ (i-g)4

+.

(13a)

~~ ~

=

'~~~~

f2

3 ~

FI

Co

l g I g)2

~

(13b)

~~ ~

=

~~~2

y

fg

2 3

~~

C

Co (i

g

)2 (1

g)3

(13c)

and in the

spin-glass phase (3

<

0)

:

~ )~~=-~~~y2~l+3+.

+

(~(2-3g)

20 ~ ~

°'~

~~° 3(1-g)~ 3(1-g)~

~' ~~' ~~~~~

~ ~ ~~~~~ ~

~

~~~~~) ~~+' (14b)

) ~~

=

~~2 yf

+

~

~2

0 C 0 g

3(j

g

)2

(14c)

If T tends towards T~ the sums of the three contributions take the

following

values in the two

phases

II ~

-

llll~ Y~

+

~~°' ii~~ Ii ~~~~ ~l (15)

~

~~

~~ °'~

~

~

~~~'

g

~~

~'~

~

~

~~

These results deserve several comments.

I) The term

proportional

to

h~

exhibits a finite

discontinuity

at T~:

~ )~- ~ ~~

=

l~~~ "~~~~ ~~

" ~

~)~ ~~~~ h~ (17)

o o o

3(1- g)

it)

The finiteness of the

discontinuity

is not a trivial result.

Indeed,

in the calculation several terms have a

divergence

but

they fortunately

cancel.

iii)

The

discontinuity

vanishes if the Waller coefficient y is

equal

to zero. On the contrary if the coefficient

f

of the field-induced mechanism is

equal

to zero, a

discontinuity

remains.

iv)

The term

proportional

to

h~

in the elastic constant

comes from a contribution

proportional

to

s~h~

in the free energy

expansion.

Therefore the coefficient may be

interpreted

as a non-linear

magnetoelastic

coefficient.

v) This coefficient is

always negative

in the

paramagnetic phase

since g

< I. On the

contrary

it may have any

sign

in the

spin-glass phase.

Exactly

at T~ the three contributions read

~~

=

~~~ 2[~ (h(

W ~ f~0

/(1

g

)

3C

A~k7~

~

~2

~

PI

Co ~

fi

3C

A~k7~ 2g_j

~

~

~° ~~(l-g)~~~

(15)

The field-induced contribution can be obtained

by continuity

from

(13b)

or

(14h).

On the contrary the other two contributions are

singular

and

they

must be calculated

directly.

5.2A

Comparison

between X and

3C/Co.

From the definitions

given by (3)

and

(8),

a

relationship

between 3C and x is

easily

deduced

d~(3

C I

d~fl d~x

dH~

~lT

dH~ ds~ ds~

It is valid for any H and any s, but we use it in the case H

= 0 and s

=

0.

Starting

from

(9)

we get :

d2x A'(gl

pB)~

d2£jj d£jj

~

ds~

4 kT

W

+

~f t

~ ~

~

~~~ ~~~~

For H

=

0, iii simply

reads

:

40

~~~

l

g(1 40)/°

When an elastic strain s is

applied,

the two characteristic temperatures become

7~= 7~(1+ ys)

and

To

=

7j~(1+ ys).

In the same manner the reduced temperature

becomes

0/(1+ ys)

while the ratio g remains

unchanged. Consequently

the parameter

@o which is a function of the reduced

temperature only

is transfornled into

@o(0/(1

+ ye

)).

Then we start from :

~

i

-40(o/(i+ ye))

~~ ~ ~

i

gii 40(o/(i

+

ys))i(i

+

ys)/o

and we find out:

dill ye d40

~

l

#o j

~

e =o

II g(1 40)/°l~

d° ~

°

~~~l

=

Y~°

~

g2 40

3

~ ~ ~

i

40 d#o

~

d40 dE~

e o i g

(1 40)/oi~

° o

w w

+

d@o 2 l

@o d~@o

+ 2 g d° 0 g ° doj

We first remark that each of the three terms in

(18) corresponds

to a

specific

interaction mechanism since

they

are

proportional

to

y~,

y

f

and

f~ respectively.

Moreover we have now

some

physical

argument for the

discontinuity

of

d~(

3 C

)/dH~

at

7~.

Indeed in the

paramagnetic phase

@o =

0 for any T

~ T~ and

consequently d@o/d0

=

0, d~@o/d0~

= 0. On the contrary, in the

spin-glass phase,

in the

vicinity

of T~ we have :

~

dqo d2qo

2

40" (1-°)+j(1-°)

+.

~"-l ~"j.

Références

Documents relatifs

Ces expressions se sont transformées, dans le langage actuel, en « avoir ou tenir quelqu’un à l’œil » ; « ne dormir que d’un œil », traduisant le fait de ne pas

Le potentiel de corrosion de l’acier XC38 sans et avec l'inhibiteur IV se déplace vers des valeurs plus nobles avec l’accroissement de la température (25 à 55°C), par contre pour

We have measured the spin-glass critical field as a function of temperature at different experi- mental time scales in the system Euo.4Sro.6S using the Faraday rotation

In conclusion, our estimate of the small fraction of degrees of freedom associated with the blocking of superparamagnetic clouds is based on the observation of two

The result obtained by this method for the rc(T = 0) is found to be lower than static approximation result which is known from the Ising case to over-estimate the value of

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des

Figure 12: Relative difference between the computed solution and the exact drift-fluid limit of the resolved AP scheme at time t = 0.1 for unprepared initial and boundary conditions ε

Abstract: From the study of a functional equation relating the Gibbs measures at two different tempratures we prove that the specific entropy of the Gibbs measure of