• Aucun résultat trouvé

Dimension effects on spin resonance in graphite intercalation compounds

N/A
N/A
Protected

Academic year: 2021

Partager "Dimension effects on spin resonance in graphite intercalation compounds"

Copied!
20
0
0

Texte intégral

(1)

HAL Id: jpa-00212439

https://hal.archives-ouvertes.fr/jpa-00212439

Submitted on 1 Jan 1990

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.

Dimension effects on spin resonance in graphite

intercalation compounds

M. Saint Jean, C. Rigaux, J. Blinowski

To cite this version:

(2)

Dimension effects

on

spin

resonance

in

graphite

intercalation

compounds

M. Saint

Jean,

C.

Rigaux

and J. Blinowski

(*)

Groupe

de

Physique

des Solides, Ecole Normale

Supérieure,

24 rue Lhomond, 75231 Paris Cedex

05, France

(*)

Institute of Theoretical

Physics,

Warsaw

University,

U1. Hoza 69, 00-681 Warsaw, Poland

(Reçu

le 12 octobre 1989, révisé le 19 janvier et le 1 S

février

1990,

accepté

le 19

février 1990)

Résumé. 2014 Pour

interpréter

les spectres de résonance de

spin électronique

dans les

composés

intercalaires du

graphite (CIG),

nous avons étendu la théorie de

Dyson-Kaplan

au cas des conducteurs

anisotropes

de dimensions finies. Pour les

composés

accepteurs, ce modèle

prédit

que ce sont

principalement

les faces

parallèles

à l’axe c

qui

absorbent l’onde

hyperfréquence

incidente,

et que les spectres ESR

enregistrés

peuvent être

interprétés

dans le cadre du modèle de

Dyson correspondant.

Sur la base de ce modèle nous proposons une

procédure expérimentale

permettant de déterminer les

caractéristiques dynamiques

des porteurs dans les

composés

accepteurs : la conductivité

électrique

suivant l’axe

perpendiculaire

aux

plans graphitiques,

03C3c, la constante de diffusion

planaire Da

et le temps de relaxation

T2.

Pour tester la validité du modèle

théorique,

nous avons étudié

systématiquement

les effets des dimensions réduites de

l’échantillon

(dimension

géométrique

sur

épaisseur

de

peau)

sur la forme de raie. Plusieurs échantillons de

C8AsF5

de différentes tailles ont été étudiées entre 10 et 300 K. Pour modifier

indépendamment l’épaisseur

de peau, nos mesures ont été effectuées à v = 4 et 9 GHz. Les

variations

importantes

de la forme de raie en fonction des dimensions de l’échantillon observées aux deux

fréquences

et leur

complète interprétation

dans le cadre du modèle

théorique

confirment sa validité et celle de la

procédure proposée.

Abstract. 2014 To

interpret

the

experimental

spectra of Conduction Electron

Spin

Resonance

(CESR)

in

Graphite

Intercalation

Compounds (GIC),

we have extended the

Dyson-Kaplan

theory

to the case of

anisotropic

conductors of finite dimensions. For acceptor

compounds,

this model

predicts

that the CESR spectra can be

interpreted

within the framework of the

Dyson

theory considering

the microwave

absorption through

the

sample

faces

parallel

to the c-axis.

Using

this model, we propose an

experimental procedure

to determine the

dynamic

parameters of carriers from CESR spectra in acceptor

compounds.

To check the

validity

of this theoretical

model, we have

systematically

investigated

the effects of the reduced

sample

dimensions

(geometrical

dimension over skin

depth)

on the CESR

lineshape. Many samples

of

C8AsF5

with different sizes have been studied in the temperature range 10-300 K. To

modify

the skin

depth

independently,

CESR measurements have been

performed

at different

frequencies, 03BD

= 4 and

9 GHz. The pronounced effects of

sample

dimensions observed at both

frequencies

and their

complete interpretation

within the framework of the theoretical model confirm its

validity

and the

applicability

of the

proposed procedure.

Classification

Physics

Abstracts 73.60 - 76.30P - 72.15

(3)

1. Introduction.

Conduction Electron

Spin

Resonance

(CESR)

is one of the most

powerful techniques

for

studying

the

dynamic properties

of carriers in conductors. The

analysis

of CESR

lineshape

and linewidth may

yield

information on

physical

characteristics of the résonance : the electrical

conductivity

u, the

spin

relaxation

T2,

the

gyromagnetic

factor as well as the carrier

diffusion constant.

In

Graphite

Intercalation

Compounds (GIC),

as in metallic

systems,

the resonant

absorption

is

accompanied by

another

absorption

process due to the

eddy

current induced

by

the

hyperfrequency electromagnetic

field. Both

absorptions

are

coupled

and determine the field and current distributions in the medium. The direct

magnetic

resonance

absorption

with

the small variations of the

strong

conduction losses leads to the characteristic

asymmetric

Dysonian shape

of the resonance line in the derivative

absorption

spectra.

The

lineshape

is

characterized

by

three

experimental

parameters :

the line

asymmetry

ratio

A /B ;

the

peak-to-peak

linewidth OH and the resonant field

Hres.

These

parameters

are indicated in

figure

1 for a theoretical

Dysonian lineshape plotted against

the reduced variable X =

gJL B(H -

HO) lh.

The

precise

determination of

physical

characteristics of the resonance from these

experimental

parameters

requires

a proper

lineshape analysis procedure.

In

previous

papers

[1, 2],

we have shown that the

Dyson

theory [3]

and the

currently

used

Feher-Kip

procedure

[4],

which have been

developed

to

interpret

the CESR

spectra

in

isotropic

metals with an

infinite flat

plate

geometry,

are not

appropriate

for the case of GIC. To obtain the correct

lineshape analysis

in the case of

anisotropic

conductors of finite

dimensions,

we have

extended the

Dyson-Kaplan theory by taking

into account the

anisotropy

of

conductivity

and

Fig.

1. - ESR line

shape

observed in the derivative spectra

(with

non uniform microwave field inside

(4)

diffusion,

and the finite

sample

dimensions.

Using

this

model,

we propose an

experimental

procedure

to determine the c-axis

conductivity

of

acceptor

compounds [5, 6].

In this paper, we

present

an extensive

study

of CESR

experiments performed

on first

stage

CgAsFs

to check the

validity

of the theoretical model. In these

highly anisotropic

conductors,

we have

systematically investigated

the effects of the reduced

sample

dimensions

(geometrical

dimension over skin

depth)

on the CESR

lineshape.

Many

samples

of different sizes have

been studied in the

temperature

range 10-300 K for both

configurations

of the static

magnetic

field

(parallel

and

perpendicular

to the

c-axis).

To

modify

the skin

depth independently,

CESR measurements have been

performed

at different

frequencies, v

= 4 and 9 GHz.

The

pronounced

effects of

sample

dimensions observed at both

frequencies

and their

complete interpretation

within the framework of our

specific procedure

confirm the

validity

of our theoretical model.

Using

this

procedure,

the

dynamic

characteristics of carriers in

C8AsF5

(c-axis conductivity, spin

relaxation

times,

and

in-plane

diffusion

constant)

can be

determined. The determination of these

physical

parameters

requires

CESR measurements

on a set of identical

samples

of dimensions

comparable

and

larger

than the skin

depth.

This paper is

organized

as follows : the main conclusions of the

lineshape

model in

anisotropic

conductors of finite dimensions are

presented

in section 2.

Experimental

details are described in section 3. In section 4 we

report

the

experimental

results

confirming

the

validity

of the theoretical model obtained at v = 4 and 9 GHz. The

procedure

used for the

Fig.

2. -

(5)

quantitative analysis

of these CESR

spectra

and the

complementary experiments

which confirm its

applicability

are

presented

in section 5.

2. Theoretical model of

lineshape.

The

theory

of the resonance

absorption

P has been

developed

for

anisotropic

metallic

systems

with finite dimensions

by

Blinowski et al.

[2, 3,

5].

In this

model,

the

sample

has a

rectangular

parallelepipedic

form with

length

L,

width f

and thickness d with d

parallel

to the c-axis of

graphite.

The microwave field

Hr

is

perpendicular

to the c-axis and

parallel

to the

length

L of the

sample.

The

sample

dimensions

(L

= 10 mm considered as

infinite, f

= 0.5-2 mm,

d = 0.1-0.4

mm) being

small

compared

to the

wavelength

of the microwave field

(A

= 3

cm),

the field

Hr

is

practically

uniform in the

vicinity

of the

sample.

The

particular geometries

of

sample

and field are shown in

figure

2. The derivations

presented

in this section are limited to

Hol/c

because the extension to the case

Ho

1 c does not introduce any

difficulty

and similar final formulas for the absorbed power are obtained in both cases.

The absorbed power per unit surface is :

where n is a

unitary

vector

perpendicular

to the surface and S =

1/2 (E

A

H * ),

the

Poynting

vector. In these

expressions,

E and H are

respectively

the electric and the

magnetic

fields

measured on the

sample

surface. To determine the absorbed power at resonance, the derivation of S may be limited to the contribution of the microwave

magnetic

field

11,..

In the

particular

geometry

presented

in

figure

2,

the

magnetic

field

Hr

is

parallel

to the

x-axis ;

only the y

and z

components

of the electric field contribute to the energy flux :

Using

the Maxwell

equations,

it is easy to show that in an infinite rod

parallel

to the uniform

oscillating magnetic

field,

the power P absorbed per unit of

length

is

simply

related to the total flux of the

magnetic

induction inside the rod

Hx

and

Mx

are the

complex amplitudes

of the

x-components

of the

oscillating

magnetic

field

H e- iwt

and the

magnetization

M e- iwt

in the

sample.

To determine the field and

magnetization

distributions

inside

the

sample, following

the

procedure proposed by Kaplan [7],

we have solved the Maxwell

equations

and the

generalized

Bloch

equations simultaneously.

The Maxwell

equations,

in which the

displacement

current and the

magnetic

field

dependence

of the

conductivity

tensor have been

neglected,

are :

where the

conductivity

tensor 0=’ is

diagonal,

u xx =

uyy

= O»a, U zz =

U c where U a and Uc are the

planar

and the c-axis conductivities

respectively.

The

magnetization

M is related to

the

magnetic

field H

by

the Bloch

equations.

In metallic

systems,

the

magnetization

is no

(6)

motion can carry the

magnetization

at distances over which the

magnetic

field

changes

considerably.

To account for this

effect,

it is necessary to

complete

the Bloch

equations by

a

diffusion term. For each

i-component

of the

magnetization Ml ( i = x, y, z),

Torrey

has

introduced a

corresponding

current of

magnetization

JMI

[8].

In

anisotropic

systems,

Jm

is

related to

Mi by

the tensor

expression :

JM.

=

D

grad Ml

where

D

is the diffusion tensor. In our case, this tensor is

diagonal, Dxx

=

D yy

=

D a, D zz

=

Dc, Da

and

Dc

are the

planar

and the c-axis diffusion constant

respectively.

The Bloch

equation describing

the evolution of

M

then have to be

completed by -

div

Jmi»

The modified

equations generalized by

the diffusion terms take the

following

form :

where

M =

Ho (Ho.

M) / HJ

and

M -

M -

Mil

are the

longitudinal

and transverse compo-nents of the

magnetization M. Tl

and

T2

are the

longitudinal

and transverse

spin-relaxation

times,

respectively. Mo

= X o

Ho

is the

steady magnetization

induced

by

the static

magnetic

field

Ho,

X o is the static Pauli

susceptibility.

After

eliminating

the electric field we obtain the

followmg

set of

equations :

where

Since it is convenient to limit these derivations

only

to the linear terms in

Hr,

we may take

Mz

= 0. The Maxwell

equations

then involve constant

H,,

and

Ey,

in our

expressions

we have

chosen

Hz = Ey = O.

To solve these

equations

we use the fact that the value of the static Pauli

susceptibility

in

paramagnetic

materials is small and

expand

all fields in a power series in X o

neglecting

quadratic

and

higher

order terms. These

equations

show that

only Hx

has a

nonvanishing

zero

term

HO,;

since

Hy,

Mx

and

My

start with the first order terms

Hÿ,

Mx

and

Mÿ

respectively.

The finite dimensions are introduced

by

the standard

boundary

conditions

imposed

to the solutions at the

sample

surface. The

tangential

component

of the

magnetic

field must be

continuous,

=

Hr,

Hx

= 0 and the normal

component

of the

magnetization gradient

must

vanish

Moreover,

it is convenient to

expand

the fields in double Fourier series so as to reduce the

(7)

Fourier coefficients

(the

choice of sine or cosine series

depends

on the

boundary

conditions

imposed

to the field

[9]).

Since we are

only

interested in the absorbed power at the resonance, it is

justified

to restrict the calculation of P to the linear term in Xo,

AP,

the zero order term

being

constant in the

vicinity

of the resonance. In this

formalism,

we obtain the

following expression

for the absorbed power at the resonance :

where

H"(0,

0)

and

M,’,’(0,

0)

are the zero cosine Fourier coefficients of the first order of

Hx

and

Mx.

These two coefficients can be found

by

the

following

three

step

procédure :

-

First,

the sine Fourier coefficients of

Hf

is determined

by solving

the

corresponding

Maxwell

equations

in which we

put

Hx

=

Hf,

Hy

=

H0y

= 0 and

=

= 0 since we are

limited to zeroth order.

Fig.

3. -

A/B

for the derivative spectra as a function of the reduced

sample

width for different reduced

sample

thickness

1)

very slow diffusion

along the y

and z axes

(T2 ta, T2 «

tc) ;

2)

very

rapid

diffusion

along the y

and z axes

(T2 » Ta, T2 > Tc) ;

3a)

very

rapid

diffusion

along the y

axis and very slow diffusion

along

the z axis

(Ta T2 tc) ;

(8)

- In the second

step,

the cosine Fourier coefficients of

Ml

are determined from the Bloch

equations

in which we

replace

Hx,

Hy,

M,,,

My

by

Hf

0,

M1x

and

M1y

respectively.

- In the third

step,

the sine Fourier coefficients of

Hx

are derived from the Maxwell

equations

in which

Hx

=

Hx

and

Mx

=

Mx.

This

procedure requires

to go over to another Fourier

representation

of a

given

field ;

this

introduces an infinite summation over two indices. The final formula for the absorbed power

then involves infinite summation over six

indices ;

this is unsuitable for

practical

calculations.

However,

this

expression

can be

simplified

in some

particular

cases

corresponding

to different

diffusion

regimes.

The extensive

presentation

of these calculations and the

simplified

final

expression

of the absorbed power can be found in references

[2,

3 and

5].

We recall here the

main conclusions of these derivations.

For each direction

(y

or

z),

the diffusion

regime

can be characterized

by

three different

times. For

instance,

the diffusion

regime along

the

y-axis

is characterized

by

the diffusion time

across the skin

depth ta

=

6 2/2

Da

where

l) a

=

c

is the skin

depth

for microwave

2 w

-B/Vcrc

penetration through

the

planes parallel

to the

c-axis ;

the diffusion time across the

sample

Ta

=

f2 /2D,,

and the

spin

relaxation time

T2.

Corresponding

diffusion

times,

with index

c, can be defined to describe the diffusion

along

the z-axis.

In the

approximation

of slow

(T2

«

ta,

T2 « tc),

fast

(T2 > Ta,

T2 »

rj,

and mixed

(Ta « T2 « t,

and

Tc « T2 « ta)

diffusion

regimes,

the line

asymmetry parameter

A/B

for the field derivative of the power absorbed dP

/dH

was calculated

analytically

as a function of the

two reduced

sample

dimensions A =

f /8 a

and u

=

d / 8 c.

These results are

displayed

in

figure

3.

They

can be

compared

to those obtained

by

Kodera for an infinite

plate

conductor of

arbitrary

thickness

d,

displayed

in

figure

4. A detailed

analysis

of the line

asymmetry

parameter

shows

that,

for very

large

values of

IL / À,

the

asymmetry

ratio

A /B

obtained for

anisotropic

conductors with finite dimensions coincide with those obtained for an infinite flat

Fig.

4. -

(9)

plate of isotropic

conductor : cases 1 and 3b in

figure

3 with the lowest curve in

figure

4 ;

cases

2 and 3a with the

highest

curve. Thus the infinite

plate

model for

isotropic

conductors is

applicable

to finite

samples

of

anisotropic

conductors in all four

limiting

cases when one of the

conditions »

/A

>

100, À

1 J.L

> 100 is satisfied

[2, 3].

For IL

/A

> 100 the

Dyson

formula with

Uc and

Da

is

appropriate.

For À

1 J.L

> 100 one should use ua and

D,.

It is natural to

expect

that these conclusions also

apply

for the intermediate diffusion case.

Owing

to the

strong

conductivity anisotropy

in

acceptor

compounds,

the skin

depth

8 a

exceeds 8 c

by

several orders of

magnitude.

From static

conductivity

measurements, one can

estimate 8c ~

0.5 u

and 8a ~

0. 5 mm at v = 9 GHz at T = 10 K. This

implies

that the

reduces

dimension u =

dl 8 c

is about

103

while À

= f / & ,

is around 1-10 for the

typical sample

dimension

investigated

in CESR

experiments.

Under these

conditions,

the theoretical

lineshape

model

predicts

that the line

asymmetry parameter

is

independent of » (for

» >

100)

but varies

strongly

with A.

3.

Experiments.

To check the

validity

of the theoretical

model,

we have

systematically

studied the effects of

the reduced

sample

dimensions

(geometrical

dimension over skin

depth)

on the CESR

lineshape.

For each

type

of

graphite

host,

a series of

rectangular parallelepipedic samples

with

different

geometrical

dimensions : L = 10 mm, 0.11 mm « d 0.4 mm and

0.5

mm l

2 mm were

investigated (Fig.

2).

To

modify

the skin

depth

independently,

two

different

frequencies

were

used,

4 and 9 GHz.

All CESR

spectra

have been obtained for first

stage

Arsenic Pentafluoride

compounds

of stoechiometric formula

CgAsFs.

The

compounds

are

synthetized following

the method of

Falardeau et al.

[ 11

from

different

graphite

hosts

(Union

Carbide HOPG

type

ZYA, ZYB,

ZYC).

After

intercalation,

the

stage

and the

homogeneity

of the

compounds

are controlled

by

X-Ray

diffraction

analysis [12]

and the

samples

are

encapsulated

in

quartz

tube filled with

helium gas to avoid

sample degradation.

CESR

experiments

were

performed

with a standard reflection method

using

BRUKER

(4 GHz)

and VARIAN E 115

(9

GHz)

spectrometers.

The

sample

geometry

and the field

configurations

are shown in

figure

2. Measurements were made for the static

magnetic

field

Ho

either

parallel

or

perpendicular

to the c-axis. In GIC the

lineshape

is very sensitive to

experimental

measurement conditions

[2].

In order to eliminate the passage and saturation

effects over the whole

temperature

range, measurements were made at a low microwave power of 10

JLW

at both

frequencies.

The field modulation

frequencies

were 1.57 kHz and

35 kHz at 4 and 9 GHz

respectively.

The CESR

experiments

were carried out between 10 and 300 K at 9 GHz and between 90 and 300 K at 4 GHz. The

sample

temperature

was controlled with a accuracy of

T = ± 1 K. All the measurements were

performed during

the

heating

process, the initial

cooling

was achieved after one hour.

4.

Experimental

results.

All CESR

spectra

obtained in

CgAsFs

exhibit an

asymmetric Dysonian lineshape

charac-terized

by

the

asymmetry

ratio

A /B

and the

peak-to-peak

linewidth AH.

Figure

5 shows a

typical experimental

spectrum.

The studies

performed

on several different series of

samples

show that the

Dysonian lineshape

is

extremely

sensitive to the finite

sample

dimensions and to

the

crystallographic

nature of the

graphite

host.

However,

although

the absolute values of

A/B and

AH can

change

from one series to

another,

some

general

features may be observed

(10)

Fig.

5. -

Experimental Dysonian

spectrum observed for a

sample

of d = 0.4 mm ; ZYA f = 2 mm ; at v = 9 GHz ; T = 200 K,

Ho

1 c.

A : theoretical spectrum calculated for À = 6.4 and R = 0.75.

of the

experimental

results is restricted to these characteristic behaviors. The best

experimental

illustrations of each have been selected from our

experimental

results and are

given

below. The

experimental

results chosen do not all

necessarily

to the same

sample

serie.

Fig.

6. - Line asymmetry parameter

A /B

vs. temperature measured at v = 9 GHz on

samples

of

constant width f = 2 mm and of different thicknesses.

(11)

4.1 LINESHAPE INDEPENDENCE OF d. -

To

study

the influence of the

sample

thickness d on the line

shape,

we first recorded the CESR

spectra

on identical

samples

of constant width

(f

= 2

mm)

but with different thicknesses d =

0.11,

0.27 and 0.42 mm. These measurements

carried out at v = 9 GHz do not show a

significant dependence

of A

/B

or OH

parameters

on

sample

thickness. These features are illustrated in

figures

6 and 7

respectively.

A similar invariance of

A IB and

AH was observed at v = 4 GHz.

Fig.

7. -

Temperature dependence

of OH measured on

samples

of constant width f = 2 mm and of

différents thicknesses.

Open symbols :

Hollc,

black

symbols : Ho

1 c.

4.2 INFLUENCE

OF f

ON THE LINESHAPE. In contrast, at both

frequencies,

at any

given

temperature

and

regardless

of the

sample

series,

the

lineshape depends drastically

on the

sample

width f.

a)

Asymmetry

parameter

A/B.

Measurements

performed

at v = 9 GHz for the ZYB series

(Fig. 8)

show a non monotonic

variation of

A IB

with

l :

for

instance,

at T =

100 K,

A /B

varies from about 3 for

f

= 0.5 mm to 8 for

f

= 1 mm and decreases to about 3.5

for f

= 2 mm. In this

temperature

range, the

asymmetry

ratio

corresponding

to

f=1.5mm

is constant and

equal

to

7 ± 0.5. Measurements

performed

at 4 GHz

(Fig. 9)

on the same

samples

show an

opposite

dependence

on

sample

size,

A /B

decreasing

from 6 to 3 as the width is reduced from 2 to

1 mm. This

f-dimension

dependence

was observed for both series

investigated.

Moreover,

the

temperature

variations of

A /B

depend

on the

sample

width

f. At 9 GHz,

A/B

is

independent

of

temperature

between 10-200 K. Above 200

K,

A/B

varies

strongly

with

temperature,

the direction and

amplitude

of these variations

depending

on the

sample

width

f.

For

instance,

the

asymmetry

ratio

A /B

corresponding

to

samples

with f

= 0.5 and

(12)

Fig.

8. - Influence of the f-dimension effects on the line

asymmetry ratio

A /B

observed at 9 GHz on

ZYB series.

o £ = 0.5 mm; .. £ = 1 mm; . £ = 2 mm; d = 0.4 mm.

Fig.

9. - Influence of the f-dimension effects

on the line asymmetry ratio

A IB

observed at 4 GHz on

ZYB series.

(13)

whereas it increases for wide

samples

(f

= 2

mm);

above 260

K,

A IB

decreases for all

samples.

b)

Linewidth : AH.

Whatever the

sample

dimensions,

for each

configuration

of the external field

(Hollc

and

Ho

1

c)

and at each

frequency,

the linewidth exhibits the same

temperature

behavior :

âH1

and

AH_L

decrease with

increasing

temperature,

varying

typically

from a few tens of gauss at low

temperatures

to 0.5 G at room

temperatures.

An

example

of this

temperature

dependence

of the linewidth is

given

in

figure

7.

At low

temperatures,

the linewidth is sensitive to the

sample

width

f.

For

instance,

at

T = 10 K and v = 9

GHz,

the linewidth measured on ZYB

samples

presents

a maximum

around f

= 1 mm

(Fig. 10).

1

Fig.

10. - Influence of

f-dimension

effects on the linewidth OH observed at 9 GHz on ZYB series,

T = 10 K,

HIlc.

Moreover,

the linewidths

depend

on the static

magnetic

field orientation :

AH_L

are

systematically larger

than

AHII (Fig. 7).

Tbis

linewidth

anisotropy AH-L

/ J1HI

is however

independent

of the

sample

dimensions and have a

comparable

value for all

investigated

compounds.

At 4

GHz,

the linewidth exhibits the same three

qualitative

features.

These observations confirm

experimentally

the theoretical

prediction concerning

the

lineshape

in the

acceptor

compounds.

In

particular,

the invariance of the

lineshape

(14)

parameters

with the

sample

width

f

demonstrate

experimentally

that all the microwave power is

practically

absorbed

by

the

sample

faces

(zx) parallel

to the c-axis. We can conclude that

the

commonly

used Feher and

Kip analysis

based on the

Dyson

formula with a-a and

Dc

is

completely unjustified

in

acceptor

compounds.

5. Procédure for a

quantitative interprétation

of CESR

spectra.

5.1 DESCRIPTION OF THE PROCEDURE. - The theoretical model

being

confirmed

by

the

experimental

studies,

we have

developed,

within the framework of the

Dyson

theory

involving

uc and

Da,

a

procedure adapted

for the

quantitative analysis

of the CESR

spectra

in

acceptor

compounds.

This

analysis

leads to the determination of the

dynamic

parameters

of

carriers: u c’ T2

and

D a.

The theoretical

spectrum

is calculated

numerically

as a function of the reduced field X = g¡..t B

T2(H - Ho)/It

using

the theoretical

expression

obtained

by

Kodera

[10]

for the

absorption

derivative. The line

asymmetry parameter

A /B

of the

absorption

derivative

depends

on two

parameters À

and R =

(Tn/T2)1/2

where is the time

required

for

the carriers to diffuse across the skin

depth 8 a.

The theoretical curve

A/B( À ,

R )

is

reported

in

figure

11,

for several values of R. For

sample

widths

comparable

to the skin

depth

Fig.

11. - Theoretical

(15)

(A 2.5 ),

the

magnetic

field inside the

sample

is almost

uniform,

the

asymmetry

parameter

A/B

does not

depend

on R. For

samples

with intermediate

thicknesses,

the distribution of r.f. field and

consequently

A /B

depend

in a

complex

way on both A and R.

A /B (A )

exhibits a

sharp

maximum which occurs in the

region

of A = 3-4. For

larger

values of À

>

6),

the

spins

which leave the skin

depth

relax before

reaching

the middle of the

sample

so that the

microwave field distribution

depends

on the diffusion. Then

A/B( )

is

nearly

constant and exhibits a

regular

increase with the decrease of R.

Moreover,

for a

subsequent analysis

of the

linewidth,

we have calculated the theoretical

line width Ax for different values of the

parameters A,

R

(Fig. 12).

Fig.

12. - Theoretical

dependence

of Ox vs. A. The values of R are indicated on each curve.

For a

quantitative interpretation

of the

spectra,

it is necessary to consider

samples

of different widths

satisfying

the condition either À 2.5 or À > 2.5.

For narrow

samples,

À

2.5,

A /B

is

independent

of the diffusion. Thus measurements of

A/B

on

samples

of dimension

comparable

to the skin

depth provide

a

good

means of

determining k

and

consequently

the c-axis

conductivity

at the

frequencies

used in the

measurement.

The c-axis

conductivity

is deduced from the value of À with the use of the formula :

This

procedure

does not

permit

the value of R to be determined.

However,

the

spin

relaxation time may be evaluated

using the

theoretical linewidth Ax which is

independent

of R for this class of

samples,

Notice that for thèse narrow

samples,

the diffusion

constant cannot be evaluated.

(16)

constant and identical for a

given

set

of samples

of the same series

[6].

Thus we can to deduce the values of À for the

samples

with various widths

f by

using

the

proportionality

between

À and l.

Therefore the

parameter R

can be estimated from the best

agreement

between the

experimental

and theoretical values

of A /B

measured on

large samples.

The

spin

relaxation

time

T2

and the diffusion

Da

can be determined. The

spin

relaxation time

T2 is

estimated

by

comparing

the

experimental

linewidth AH to the theoretical linewidth Ax calculated for the

parameters

À and R obtained from the

lineshape analysis,

using

the relation :

where IL B is the Bohr

magneton.

Notice that it is essential to introduce Ax

(1.2

Ax « 2.1 for

our

experiments)

in order to discuss the relaxation process

correctly.

Indeed,

some

apparent

effects can be observed on OH which

disappear

when the

study

is based on

T2.

After determination of R and

T2,

the in

plane

diffusion constant

Da is

extracted

by

using

the

relation :

5.2 EXPERIMENTAL VALIDATION OF THE PROCEDURE. -

Using

the values of A and R obtained from the

quantitative analysis

of the

experimental

spectra,

the

absorption

Fig.

13. - Values of À determined from theoretical curve.

A : f

= 0.6

mm ; 0 : f = 0.8 mm.

(17)

derivative

dP /dH

was calculated from the theoretical

expression

derived

by

Kodera

[10].

The

comparison

between an

experimental

and the

corresponding

theoretical

spectra

is

presented

in

figure

5. The

good

agreement

between

experimental

and theoretical

lineshapes

indicates that our model

correctly

describes the CESR

lineshapes

in

acceptor

compounds.

In order to select in

experiment

the desired

region

of A 2.5 and to check the

validity

of the

procedure

of

analysis

we have

performed

measurements of

A /B

on a series of narrow

samples

with different

f-dimensions.

The values of À obtained from the

experimental

lineshape

asymmetry

ratio

using

the theoretical

dependence

A/B (l )

are

reported

in

figure

13. The reduced dimensions A are, with excellent accuracy,

proportional

to the

sample

size

f.

This

provides

a

convincing

proof

of the invariance of u c for a

given

set of

samples

of the

same series.

The

validity

of the model is further

supported by

the self

consistency

of the results obtained. In

particular,

this model

provides

a

quantitative interpretation

of the influence of

the

sample

width

e

on the

asymmetry

ratio

A IB

and the variations

of A /B

with

temperature.

The non-monotonic variations of

A /B

with f

reflect the existence of a maximum in the curve

AI B(À).

For

AsFs-graphite compounds,

at v = 9

GHz,

the

corresponding

values of A for

l =

0.5 and 2 mm fall on different sides of the

maximum ; for f

= 1 mm,

A occurs in the

region

of the maximum. At v = 4

GHz,

the k values

corresponding

to

e

= 1 and 2 mm both fall on the

ascending

part

of the curve

A/B (,k ).

Thus for

instance,

the

temperature

dependence

of the line

shape reported

in

figures

7 and 8 for

samples

with different widths is

consistently explained :

as the

temperature

is

raised,

A decreases

implying

an increase

of A /B

for f

= 2 mm and a decrease

of A /B

for

samples

off

= 0.5 and 1 mm at

9 GHz. At 4

GHz,

A IB

decreases with

T,

in

good

agreement

with

experimental

features.

Fig.

14. - C-axis

conductivity

obtained from

A IB

measurements on narrow

samples.

(18)

5.3 EXAMPLES OF QUANTITATIVE RESULTS. - This

general procedure permit

a

complete

interpretation

of CESR

spectra

measured on

acceptor

graphite compounds

and a

precise

determination of the different

physical

characteristics of the

compounds

studied :

T2,

Da.

In this

section,

we

present

some of the main results obtained for first

stage

AsFs-graphite

compounds.

In

particular,

the CESR measurements

provide

direct evidence of the

frequency

depen-dence of the c-axis

conductivity

which was never observed

previously [6]. Figure

14 presents

this

frequency dependence

of Uc determined on narrow

samples

ZYB

f

= 0.5 mm at 9 GHz

and ZYB

f

= 1 mm at 4

GHz,

the c-axis d.c.

conductivity

is measured on a reference

sample

synthetized

in the same reaction as the

sample prepared

for CESR

experiments.

This

observed

dependence

of the c-axis

conductivity

indicates the non standard character of the conduction mecanism

along

the c-axis. Most

probably,

there is a contribution to the a.c.

conductivity

from carriers whose

displacements

in an electric field are limited

[13]

or which

obey

a

hopping

process

[14,

15].

The

spin

relaxation time determined from CESR measurements reveals two

general

features,

whatever the

sample

dimensions or the

frequency :

T2

increases with the

temperature

and is

systematically larger

for

Hollc

than for

Ho

1 c.

Fig.

15.

Temperature dependence

of the

spin

relaxation time obtained from the

lineshape

and line width

analysis ( v

= 4 GHz,

Hollc).

(19)

An

example

of such behaviors is

presented

in

figure

15 which

gives

the variations of

T2( T)

for ZYB

samples

at 4 GHz :

T2

is

equal

to 3 x

10- 9

s at 90 K and increases

by

two

orders of

magnitude

as the

temperature

rises from 90 K to 300 K. The

large uncertainty

observed on

T2

at room

temperature

is due to the

imprecise

determination of the

experimental

linewidth AH and is not to be attributed to the

analysis procedure.

Moreover,

the

comparison

of the

spin

relaxation time determined for

samples

with

different widths

f

indicates

that,

whatever the resonance

frequency, T2

seems to decrease with the

sample

width

f

(Fig. 15). Although

more

complete experiments

will be necessary to

confirm this

sample

dimension

dependence,

we can

already

note that the non monotonic

variation

of J1H ( f)

shown in

figure

10 is not observed in

T2.

This confirms that the linewidth is not the

right

parameter

to discuss the relaxation processes and that a determination of

T2 taking

into account the corrective term introduced

by

the theoretical linewidth X is essential to

interpret

the relaxation processes

correctly.

These variations of the

spin

relaxation time with

temperature

can be

explained qualitatively

by inhomogeneous broadening

mecanisms

[2, 5].

In a

previous

paper, we have

suggested

that

this

narrowing

is due to

exchange

interactions between delocalized carriers and other

magnetic

moments in GIC. Several recent

experimental

results confirm the existence of these localized moments, but their nature and their localization are not

clearly

established. The variation of the linewidth with the atomic number of the intercalant

species

[16, 17]

suggests

that these localized moments are associated with the intercalant

layers

whereas recent

spin

echo results

[18]

and the

anisotropy

of the linewidth with the orientation of

Ho [15]

suggest

that

they

are localized in the

graphite layers.

A

complete

discussion of the

spin

relaxation time

is

presented

in references

[2,

5].

The

planar

diffusion constant

Da

obtained from À and R at 9 GHz

presents

a

systematic

decrease as the

temperature

increases. For

instance,

Da corresponding

to the

sample

ZYB

f = 2 mm,

varies from

6 x 104 cm2 s-1

1

at 10 K to

2 x 103 cm2 s-1

1 at 300 K. This

high

temperature

value seems to be identical for all

samples,

whatever their

f-dimensions.

In

contrast at low

temperature,

Da

increases when the

sample

width decreases. For

instance,

at

10

K,

the diffusion constant obtained for the

sample

ZYB

f

= 1 mm is

equal

to

2.5 x

105

cm2

s-1 whereas

that

corresponding

to ZYB

f

= 2 mm is around 0.6 x

105

cm2

s-1.

Although

very

surprising,

this effect has been observed on different

sample

series and is not

an artefact of the

analysis procedure.

More

experimental

and theoretical effort will be

necessary to confirm and

interpret

this

effect, however,

we can

already

conclude that these

compounds

are not usual two dimensional metallic

systems.

References

[1]

BLINOWSKI J., KACMAN P., RIGAUX C., SAINT JEAN M.,

Synth.

Met. 12

(1985)

419.

[2]

SAINT JEAN M., RIGAUX C., BLINOWSKI J., CLERJAUD B., KACMAN P., FURDIN G., Ann.

Phys.

Fr. 11

(1986)

215.

[3]

DYSON F. J.,

Phys.

Rev. 98

(1955)

349.

[4]

FEHER G., KIP A. F.,

Phys.

Rev. 98

(1955)

337.

[5]

SAINT JEAN M., Thesis Paris

(1989)

unpublished.

[6]

SAINT JEAN M., RIGAUX C., BLINOWSKI J., FURDIN G., MC RAE E., MARÊCHÉ J. F., Solid State

Commun. 69

(1989)

435.

[7]

KAPLAN J. I.,

Phys.

Rev. 115

(1959)

575.

[8]

TORREY H. C.,

Phys.

Rev. 104

(1956)

563.

(20)

[10]

KODERA H., J.

Phys.

Soc.

Jpn

28

(1970)

89.

[11]

FALARDEAU E. R., HANLON H. R., THOMSON T. E., Inorg. Chem. 17

(1978)

301.

[12]

LELAURAIN M., MARÊCHÉ J. F., FURDIN G., MC RAE E., HÉROLD A.,

Synth.

Met. 23

(1988)

365.

[13]

MARKIEWICZ H. S., Solid State Commun. 57

(1986)

237.

[14]

SUGIHARA K.,

Phys.

Rev. B 29

(1984)

497.

[15]

SHIMAMURA S.,

Synth.

Met. 12

(1985)

365.

[16]

DAVIDOV D., MILO O., PALCHAN I., SELIG H.,

Synth.

Met. 8

(1983)

83.

[17]

MURATA M., SUEMATSU H., J.

Phys.

Soc.

Jpn

51

(1982)

1337.

Références

Documents relatifs

modèle de Salamon ne peut expliquer nos résultats expérimentaux. When lowering the temperature, a broadening of the line and a shift of its centre towards low fields

region of the maximum corresponding to the bond length of 5.08 Â is analogous to the behavior of the imaginary part of the FT of metallic Ni for the

For each temperature and for each peak, we calculate the inverse Fourier tranform and obtain the quantities xj(T) relative to the peak j at the temperature T. A very

Résumé.- On a étudié par spectrométrie Mossbauer des composés interfolliaires de graphite G-FeCl (0 ^ x-v ! ) • L'emploi de feuille de Grafoil entraîne un effet de surface

It is shown that characteristic phonon dispersions modulated by a coupling between intercalant atoms and carbon atoms in the adjacent layers play a crucial role in determining

Indeed, this interaction gives rise to a resonance signal which is relatively narrow at low temperature (< 100 gauss residual width) and intense, and whose

Abstract.- The Fermi surfaces of graphite-potassium intercalation compounds, C,,K and C,,K, were composed in such a way that the observed de Haas-van Alphen periods were

The temperature dependence of the recoil-free The decrease in the OM for the graphite-FeC13 fraction of the absorber is given from the Debye mo- compound compared with that