• Aucun résultat trouvé

Structural and magnetic order in CoCl2 intercalated graphite

N/A
N/A
Protected

Academic year: 2021

Partager "Structural and magnetic order in CoCl2 intercalated graphite"

Copied!
16
0
0

Texte intégral

(1)

HAL Id: jpa-00246484

https://hal.archives-ouvertes.fr/jpa-00246484

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.

Structural and magnetic order in CoCl2 intercalated graphite

J. Rogerie, Ch. Simon, I. Rosenman, J. Schweizer, R. Vangelisti, P. Pernot, A.

Perignon

To cite this version:

J. Rogerie, Ch. Simon, I. Rosenman, J. Schweizer, R. Vangelisti, et al.. Structural and magnetic order in CoCl2 intercalated graphite. Journal de Physique I, EDP Sciences, 1992, 2 (3), pp.291-305.

�10.1051/jp1:1992135�. �jpa-00246484�

(2)

Classification

Physics

Abstracts 75.25

Structural and magnetic order in Cocl~ intercalated graphite

J.

Rogerie

(1.

2),

Ch. Simon

(I),

I. Rosenman

(I),

J. Schweizer (2>

3),

R.

Vangelisti (4),

P. Pemot

(4)

and A.

Perignon (4)

ii) Groupe

de

Physique

des Solides (*), Universitds Paris 7 et 6, 75251 Paris Cedex 05, France (2) DRF-DN, Centre d'Et1~des Nuc16aire de Grenoble, France

(3) Institut Laue

Langevin,

Grenoble, France

(4) Laboratoire de Chimie du Solide Min6ral

(**),

Universit6 de Nancy I, Vandceuvre [es Nancy, France

(Received J2

July

J99J, revised 25 November J99J,

accepted

28 November J99J)

Rdsumd. Nous

pr£sentons

des r£sultats de diffraction de rayons X et de neutrons obtenus sur

un monocristal de CoC12 ins£rd dons le

graphite

ok la transition de Kosterlitz-Thouless peut dtre testde car c'est un

systdme

2D XY. La

longueur

de corrdlation du

magndtisme

dons le

plan

des couches augmente

lorsque

la

temp£rature

est abaissde, elle

diverge

h 4,5 K. L'ordre antiferroma-

gn6tique

suivant l'axe c

apparflt

d'abord, h 6,5 K, mais la mise en ordre 2D n'est pas modifide.

C'est pour cette raison que

l'analyse

2D est

justifi£e.

Nous avons ensuite

analys£

[es intensit£s

magn6tiques

et [es

longueurs

de corr61ation en fonction de la

temp6rature

et montr6 la

prdsence

de d6fauts (vortex)

qui

abaissent l'aimantation mesurde dons [es

plans ferromagn6tiques

comme it est

pr£vu

dans [es

syst~mes

2DXY.

Abstract. We present the results of

X-ray

and

magnetic

neutron diffraction studies on a

single crystal

of a rust stage

Cocl~

intercalated

graphite

which is a

typical

2D XY system in which the Kosterlitz-Thouless theory can be tested. The correlation length of the

in-plane ferromagnetism

increases as the temperature is lowered and

diverges

at 4.5 K. The c-axis

antiferromagnetic

order appears first at 6.5 K, but the mechanism of the 2D

ordering

is not modified. For this reason the 2D

analysis

is

justified.

We have then

analysed

the

magnetic

intensities and the correlation

lengths

as functions of temperature and shown the presence of defects (vortices) which lower the measured

magnetization

in the

ferromagnetic planes

as predicted in the 2D XY systems.

1. Introduction.

The search of a

good

2D XY system to test the Kosterlitz-Thouless

(KT) theory [I]

is still a

challenge [2]

and

Cocl~

intercalated in

graphite

was

proposed

a

long

time ago as a

possible

candidate

[3]. Many

measurements of

magnetic susceptibility [4-6]

were

performed

on

samples

with various

stages

of intercalation

[7].

The

shape

of the

peak

observed in these

(*) UA CNRS 17.

(**) UA CNRS 158.

(3)

292 JOURNAL DE PHYSIQUE I N° 3

experiments

was

interpreted

as a 2DXY

transition,

modified

by

the

in-plane

six-fold

symmetry,

the 3D

coupling

and the

in-plane

finite size effects

[6,81. Specific

heat

measurements have shown that

strong

2D fluctuations exist above the critical

temperature [91.

A

quite

extensive neutron

scattering study

has been

performed

on the second stage

compound [10-131, showing

a 3D

antiferromagnetic coupling along

the c-axis and an

in-plane ferromagnetic ordering. However,

there is a

large

c-axis disorder after intercalation

(a

mosaic of more than 10

degrees),

and the

samples

which were

previously

used

present

a

polycrystal-

line nature

(HOPG

I-e-

Highly

Oriented

Pyrolytic Graphite

or stacked Kish

graphite).

Moreover the almost

complete

absence of correlations

along

c-axis in second

stage (there

is

no modulation of the

(hkf

lines

along

f in the

crystal)

makes fine

magnetic crystallography

very difficult to achieve. The determination of the

magnetic

order needs a

good knowledge

of the

crystallographic

structure as any structural defect has

consequencies

on the

magnetic

diffraction. For this reason, we have used for our

study

a

single crystalline

first stage

compound

where the

crystalline

order is much

bener, though

the presence of a 3D

coupling.

Indeed,

the

magnetic specific

heat mesurements

[9]

have shown that 2D strong fluctuations above the transition temperature still exist in the first stage

compound

and it is then

possible

to

study

the

sening

up of a

magnetic in-plane

2D order in it. The

magnetic susceptibility

presents in this stage a double

peak

structure, the

interpretation

of which is still

controversial: it is

interpreted

as the successive

ordering

in the

layer plane (upper temperature)

and 3D

antiferromagnetism (lower temperature) [4,12].

An altemative

interpretation

was

superparamagnetism

due to the islandic structure of the intercalated

layers

and their

coupling by dipolar

effect

[14, 15].

The

spin

arrangement itself was studied in a

powder [16]

and then confirmed in a HOPG

[17].

The structure is similar to that of the second stage

compound.

In this paper, we present a

complete study

of the structural and

magnetic

order. In a first part, we present the

X-ray

diffraction on a

single crystal

of first stage

Cocl~

intercalated in

graphite. Then,

in a second

part,

we

give

a

complete study

of the low

temperature (2 K) magnetic

structure. In a third part, we

present

the

temperature

variation of the correlation

lengths

which shows that the

interpretation

of the two

peak

structure is

quite

different from that

previously proposed

: the upper temperature is the 3D

ordering

and the

lower one

corresponds

to the

divergence

of the

in-plane

correlation

length.

Part four is devoted to the

study

of

higher

stage

compounds

where the 3D

ordering

does not occur. A last part is a discussion of our

results,

in

particular

of our observed transition

temperatures

which

are lower than the usual ones and a discussion of the effect of the 3D

ordering

on the

in-plane

correlation

length.

2.

Samples

and

crystalline

characterization.

The

specimens

were

prepared

in a two-zone fumace

using

natural

single crystals

as host material. First stage

compounds

have been obtained

by

direct action of the gaseous dihalide

with an excess of chlorine gaz

(580

°C 0.7 Atm of

Chlorine,

30

days).

The chemical

analysis,

made on natural

graphite powder (Ceylon

80~&§ ~200 ~L) intercalated under the same conditions

[18]

has

given

an average formula :

C5_23CoC12.12.

When

compared

to the ideal

ratio of

Cocl~

obtained from the

inplane

cell dimensions

assuming

2 carbon atoms per

graphite cell,

I,e. 2 x

(3.56/2.46)2

=

4,19. Such a

stoichiometry

shows that

only

about 85 fb of the available volume is filled with the intercalant. The fact that there is some chlorine excess, here 6

fb,

is now

accepted [19].

This

stoichiometry corresponds

to a Co

filling

factor of 80 fb.

Moreover it should be mentioned that the intercalation in

single crystalline samples

is more

difficult than in

powdered samples,

so the actual

filling

factor in our

sample

is

probably

even

lower than 80 fb.

We

give

both

X-ray

and neutron diffraction results. The

X-ray crystallography

at room

(4)

temperature was

performed

at

Nancy University.

The

crystalline

and

magnetic

structure has been

investigated by

neutron diffraction at the Institut

Laue-Langevin

on a two axis cold

neutron source 2D multidetector

spectrometer

D16

[20].

For the neutron

scattering

experiments,

the

sample

was sealed in an aluminium container and mounted in a variable temperature cryostat of the « orange » type

[20].

As we are here

mainly

interested in the

magnetic

structure, we have used

X-rays

to check the

quality

of the

intercalation,

and we have used the neutron

scattering

data to

analyse

the

magnetic scattering.

For this reason the neutron diffraction

pattems

are corrected for the different effects which broaden them : the mosaic of the

sample,

the resolution of

spectrometer,

and the correlation

length

of the order.

In most of the cases we will discuss now, the mosaic of the c-axis will dominate the

broadening

: 9.6° at the full width at half maximum

(FWHM)

as measured

by

a

rocking

curve

of the

(001) peak,

to be

compared

to the

in-plane

one of

2°,

as measured on the

(hk0) peaks (see below).

2,I STAGE HOMOGENEITY.- The

stage

index

(s

=

I)

and the intercalation

homogeneity

were checked

by (00f) X-ray diffractograms using

o 2 o scans in reflection mode. From the FWHM of the

peak,

we deduce the correlation

length along

the c-axis

(if

=

500

A).

The

comparison

between the calculated and observed intensities

(without taking

into account the thermal coefficient and the

absorption)

leads to the

following crystallo-chemical description

:

the intercalate

layer

consists of a

three-layer

sandwich of Cl-Co-Cl

along

the c axis

(Tab. I).

The c-axis electronic

density gives

a direct

representation

of the

layer stacking [2 II- Figure

I shows the electron

charge density

obtained from the Fourier transform of the diffraction pattem

compared

to the

computed

one

according

to the parameters of table I. This

figure

is

consistent with the

assignment

of a Co

layer

at the center of the intercalate sandwich and Cl- Co-Cl

layers

are

slightly compressed

between the

graphite bounding layers compared

with

pristine

material

[22].

Small satellites in the curves of

figure

I are series determination effects which arise from the finite number

(11)

of terns available in the Fourier transform.

Neutron diffraction was

performed

at 20 K and the results are also listed in table I. The measured correlation

length if

is

larger

than 250

A

which is the

experimental

resolution.

Table I. The parameters

ofthe first

stage

ofcocl~

intercalated in

graphite.

Data at 300 K are obtained

from X-ray study

while those at 20 K come

jfom

neutrons.

la)

The altitude

of

the

different

atoms

ofthe

cell.

lb)

The cell and structure parameters : ac,

ac~ci~,

are

respectively

the cell parameters

of

the

graphite lattice,

and the intercalate

lattice, dc_c

is the

in-plane

distance

between carbon atoms,

fj~, flare

the

in-plane

and c-axis correlation

lengths.

a)

Atom Co Cl C

T(K)

Altitude

(A)

0 1.38 4.695 300

b)

ac

(A) acoci~ (A) dc-c (i) Ic (A) ii iA) fib IA) T(K)

2.456 3.56 1.417 9.42 500 500 300

2.452 3.54 1.415 9.32 ~250 500 20

(5)

294 JOURNAL DE

PHYSIQiJE

I N° 3

c c

4

cl

fl

cl

I

o 3

Fig.

I.

Charge density

p (z)

along

c-axis from a Fourier

synthesis

of the

(00f) integrated

intensities.

The continuous line is

experimental

density obtained from X-ray diffraction and the dashed line is the calculated

profile.

2.2 CRYSTAL STRUCTURE. From

(hk0) X-ray diffraction,

the

hexagonal

lattice constants

are determined

by using

diamond

powder

as a reference. The

in-plane

correlation

length

f~,

~ is determined

by

the FVfHM method and

given

in table I. The

in-plane

structure was

shown

by

monochromatic Laue

photographs [221.

The Laue

diagram

agrees with the

existence of a

CoC12 hexagonal

lattice close to that of the

pristine Cocl~ [23]

and

quasi- parallel

to the

graphite

lattice

(Fig. 2).

However the diffuse spots associated with the intercalate form an arc which shows an orientation of the

crystallographic in-plane

axes not well defined with respect to the carbon lanice. The

angular dispersion

betwen the two a-axes can reach 2° as shown

by

the double rotation

technique [24].

We also note the existence of

diffuse lines at 30° to the a-axis of

graphite

which cannot be attributed to another intercalate lattice at

30°,

but to a disordered state.

2.3 LAYERS STACKING. The

layer stacking

refers to the arrangement of a

given layer

relative to its

neighbouring

ones. The two

in-plane

sublattices do not interfere and diffract

independently

in this

compound (Fig. 2).

The

Cocl~ layers stacking

can be determined from the

profile

of the intensities measured

along (10f)

rods of

graphite

and intercalate lattices.

Figure

3 presents the

(10f) Cocl~

rods obtained

by X-ray (a)

and neutron

(b)

diffraction. The

neutron diffraction pattems are corrected

by

the

geometrical

effects : we have summed the

diffracted intensities of different detector cells and different

sample positions taking

into

account the effect of the mosaic as

long

as it is

responsible

of the measured width

(larger

than

the instrumental resolution and than the width of the diffuse

ridge).

We have to take into account the very

anisotropic

mosaics of the

sample

: the c-axis mosaic

(measured by

a

rocking

curve of the

(001) peak)

is 9.6° at the full width at half maximum

(FWHM),

and the

in-plane

orientational one

(measured

on the

in-plane rocking

curve on

(hk0 ) peaks)

is 2°

(see

below in

«

crystal

structure

»).

The

(10f)c~~i~ pattem

obtained

by X-ray

or

by

neutron diffraction agrees with an

identity period

of 3

Ic

where

I~

is the lattice constant

(9.32 A

at 20 K

by

neutron diffraction and

9.39

A

at 300 K

by X-ray analysis).

This result entails an

a

-p

-y

stacking

similar to that of the

pristine Cocl~.

Numerous faults

perturb

the ideal

stacking

and we have

computed

the diffraction pattem

(appendix I)

with a model which takes into account these

faults, following

Marti et al.

[25].

The diffraction

pattems

of

figures

3a and 3b are different since both

techniques

are not sensitive to the same featl~res :

X-rays

are

mainly

sensitive to the cobalt atoms while neutron diffraction is sensitive to all the defects as the chlorine atoms are the main neutron scanerers. Calculations and

parameters

used in this calculation are defined in

(6)

*

,

,'

,

/

. / / / / /

°i

~d

* «

Q~

(7)

296 JOURNAL DE PHYSIQUE I N° 3

la)

/~ '

w e

" E

g

~

d

#

S "

,

W /

e a /

C ~

~ O

O o

o j~~

0 5 lo 15 0 2

1 1

Fig.

3. The (

lot)

diffraction pattem of the intercalant. a) By

X-ray

at 300 K. b) By neutron at 20 K.

The full line is the fit

by

the model

explained

in

appendix

I and the dashed one represents the

scattering by

the disordered part of the

compound.

The intensities are corrected

by

the

geometrical

factor.

[27].

We worked at low temperature

(2 K)

under a

magnetic

field of 0.3 T

applied

in the

layers plane.

To be sl~re to see

only

the cobalt atoms, we have

performed

the difference between this

pattem

and the

pattem

obtained without

magnetic

field

(the

system is hence

antiferromagnetic

and does not diffuse at small

angles)

or at

higher

temperature. We did not observe any small

angle scattering

due to the cobalt atoms. It is of course

always

difficult to be

sure of the result of a

negative experiment, especially

if the

predicted magnetic signal

is small

(about

I fb of the nuclear small

angle signal). However,

we believe that

previous

small

angle

experiments

on

NiC12 samples [15]

were not relevant in

Cocl~

and that the cobalt vacancies

randomly spread

over the

layer.

This is also the result of a recent

X-ray

and TEM

study

on

CoC12-GIC

of

stages

I and 2

by Speck

and Dresselhaus

[28]

which conclude that the intercalate

layers present

a porous continuous structure rather than an islandic discontinuous

one.

In conclusion, the

(00f)

line characterizes the

staging

and its faults while the

(10f)

neutron diffraction line is sensitive to the

stacking

and to the

chirality

of the Cl-Co-Cl

layer.

The intercalated

layer stacking

remains

nearly

identical to that of the

pristine Cocl~

in

spite

of the presence of a

graphite layer.

This effect characterizes the first

stage compound

and is not observed in the

higher

stages. Similar results are observed in

Cdcl~

and

Cucl~ [26].

For this reason, the

complete magnetic study

can be achieved.

3. The

magnetic

structure.

All the

magnetic

pattems

presented

have been obtained as a difference between two

equivalent pattems

at low and

high temperatures (20 K)

in order to remove the

non-magnetic

part of the

scattering.

The transformation of the intensities from

laboratory

coordinates (@, q~

)

into

reciprocal

space coordinates

(Q)

is

performed taking

into account the

anisotropic

mosaics of the

sample [29].

The

magnetic

calculation takes into account the correction due to the geometry of the

neutron-spin

interaction

((S

x

Q)

x

Q/Q~

where S is the

spin

director

vector and

Q

the neutron transfer

one).

Figure

4 presents the

(00f) magnetic

diffraction pattem at 1.6 K. It indicates a clear c-axis

antiferromagnetism

with cobalt

spins laying

in the

plane.

This structure induces a

doubling

of the cell

along

the c-axis. This result is in agreement with all

previous

studies on

polycrystals.

(8)

'Ii 3/z g~

0 20 40 60 80

26 (degree

Fig.

4. The

magnetic (00i)~~

line at 1.6 K. The intensities are corrected by the

geometrical

factors.

A

=

4.52

h.

From a

comparison

between nuclear and

magnetic intensifies,

we deduce that

only

60 fb of the

spins (1.44 ~B/Co atom)

are ordered

ferromagnetically

at low

temperature (1.6 K) (64

fb

if we take into account the

stoichiometry C~_~~coo,~4Cl~

obtained from the chemical

analysis

(cf.

part

2)

and the

staging

faults

although

the average

magnetization

saturates as will be seen in the next section. The

magnetic

correlation

length along

c-axis is 125

I (about

13

magnetic layers),

smaller than the

crystallographic

one

(f)= 500i).

This is

probably

due to the

importance

of the

stacking

faults. The corrected

(10f)~~ magnetic

diffraction line is flat in

f (Fig. 5). Usually,

in nuclear

diffraction,

such a flat

(10f)

line is characteristic of 2D systems, I-e- of a

complete

absence of correlations in the

stacking.

This effect does not exist in

magnetic

diffraction since the

spins

are attached to the atoms, and we have

computed

the

intensity given by

the 3D

antiferromagnetism (taking

into account the

crystalline

defects and

300

250

3 200

f

lS0

o

° 100

50

0

0 DA 0.8 1.2 1.6 2

~

Fig.

5. The

(10i)~~ magnetic

diffraction pattem at 1.6K, corrected

by

the

geometrical

factors A

= 4.52

h.

The continuous line is the

computed

diffraction

by

the

perfect antiferromagnetic

structure,

taking

into account the

crystalline stacking

correlations, and the horizontal line

corresponds

to a

calculation where

they

are

neglected.

(9)

298 JOURNAL DE PHYSIQUE I N° 3

the resolution of the

spectrometer).

The

computation predicts

indeed the presence of a small modulation of the

magnetic

line

(Fig. 5)

which is too small to be observed

here, given

the statistical error bars.

However the

antiferromagnetic intensity

is non zero and can be obtained from the average value of the

scattering intensity

of the

(10f) magnetic

pattem.

Compared

with the structural

intensity,

it confirms the value of about 64fb for the ordered cobalt

spins, previously

determined on the

00i magnetic

line.

Figure

6

displays

the

(100) magnetic peak

as it is observed on the multidetector : the

peak presents

an

asymmetric

line

shape

which is a convolution of the

in-plane

correlation

length,

the

experimental

resolution and the Lorentz factor which is the convolution of the Warren tail

[30]

and the finite mosaic. We have fitted the

experimental lineshape

on this basis. We have chosen a Gaussian

in-plane

structure factor as it is a

good approximation

of a

quasi-long

range

order as

pointed

out

by

Dutta and Sinha

[31]

and it is

impossible

to

distinguish

from other

forms in such an

experiment.

The width of the Gaussian function is assumed to be

proportional

to the inverse of the

in-plane

correlation

length f~,

~. This correlation

length

is

larger

than the resolution at 1.6 K and can be estimated to be

larger

than

5001.

The

shape

of

the curve at 1.6 K is defined

by

the resolution and the mosaic. The

right

width of this

peak

increases with the temperature. This increase shows the decrease of the

in-plane

correlation

length.

T=10 K

1,",,

,,'

'.

'; "'.,'

.' .' .'" '. .'.

m

..'

_~:

,_T=6 0

I ', ..

g

(

E 6 ~ ~.

)

''

T=4 5 II

I

4

~ fi

d

T=1 6 K

," .'

0

S 4 3 2 0 2 3 4 5

4 (degree)

Fig.

6. The temperature variation of the (100)

magnetic

peak. The fit is a Lorenzian line

shape,

convoluted

by

the

experimental

resolution, the mosaic of the

sample,

and the flat

(10i)

line (Warren

shape).

~P is the 2 0

angle

with the center of the detector at a fixed

position

for Q

= (100) for

~P

=

0.

In

conclusion,

the low temperature

magnetic

structure consists in a

long

range antifer-

romagnetic

order

along

the c-axis

(125 I)

and

a

long

range

ferromagnetic

arrangement in the

plane (more

than 250

I).

No modulation of the

(10i) magnetic

line

was observed because of the

crystallographic

defects. The

complete integrated

intensities of both

(00f)

and

(10i)

lines

give

a number of ordered

spins

of 64

fb, indicating

that 1/3 of the

spins

are

disordered even on a very short

length

scale. This

point

will be discussed in the last part.

(10)

4.

Temperature dependence.

Figure

6

gives

the

temperature

variation of the

(100) magnetic

diffraction line where a dear

broadening

can be seen when the

temperature

increases. This is related to a decrease of

f~,~

and we have fitted the curves

according

to the model described

previously

for 1.6 K.

Figures

7a and 7b show the

temperature dependence

of the

integrated intensity

of

in-plane ((100)

and

out-of-plane ((001/2)

lines and of the inverse of the correlation

lengths I/f~

~ and

I/f~.

(b)

t120 30

loo

u iii

i~i 25

£1,,~i,[

~ ~

i

~~

~f

" ~~ '%

~~

(

~~ loo ©4

~~

~'

~~ ~~

f

~~

p

00112

~~ '

~

i

~

0 0

0 2 4 6 8 lo 12 0 2 4 6 8 lo 12

T(K) T(K)

Fig.

7. a) The temperature variation of the

integrated

intensities of the (001/2)

magnetic peak

and of the (100)

magnetic

part. b) The temperature variation of the inverse of the correlation

lengths

obtained from the half width at half maximum of the diffraction pattems.

ij

are the cristalline correlation

lengths.

The continuous line is a fit by a KT model as described in text ; T~~ is the KT 2D

ordering

temperature, and

T~

is the 3D AF

ordering

temperature.

Two critical temperatures appear on these

figures.

One

corresponds

to the appearance of the 3D

antiferromagnetism:

the

extrapolation

of the

(001/2) peak intensity gives TN

= 6.5 K ± 0.3 K

(Fig. 7a).

The

intensity

of the maximum of the

(100) peak

varies

initially

in the same way as the

(001/2) peak

but does not go to zero above

T~.

The other critical

temperature

is

given by

the

extrapolation

of the variation of the inverse of

in-plane magnetic

correlation

length (Fig. 7b).

This temperature,

T~D,

is about 4.5 K±0.5 K. Above

T~,

a strong 2D

ferromagnetic

order remains with a correlation

length

which is much smaller than at 1.6 K. It is a short range 2D

quasi

order. At

T~,

no

anomaly

was observed in the

variation of the

in-plane magnetic

correlation

length.

This result demonstrates that

previous interpretations

of the upper and lower temperatures, deduced from the

magnetic susceptibility studies,

as T~D and

T~ respectively

were wrong. The upper

temperature corresponds

to the

disappearance

of the

long

range c-axis

antiferromagnetism.

The lower one

corresponds

to the

divergence

of the

in-plane

correlation

length,

and we have called it T~D for this reason.

However at this

temperature

the out-of

plane f~

correlation

length

is

already quite large,

so

the actual order is 3D.

Along

the

c-axis,

a critical

broadening

was

observed,

above 5.5

K,

and

close to the

T~ temperature (6.5K).

This is similar to the results of the second stage

compound [10, Ill

where the 3D order has a smaller correlation

length.

5. Correlation between neutron

scattering

and

magnetic susceptibility.

We have observed two critical

temperatures

T~D and

T~ by

neutron

scattering.

In the first stage

compound,

two

peaks

appear on the

magnetic susceptibility [5].

We have decided to

study

the neutron

scattering

and the

magnetic susceptibility

in the same

sample.

This has been

(11)

300 JOURNAL DE PHYSIQUE I N° 3

done on a first stage HOPG intercalated

sample. Figure

8

presents

the obtained results. It is clear on this

figure

that

T~ corresponds

to the

highest temperature peak

of the

susceptibility.

The

peak

of

susceptibility

at the lowest temperature

corresponds

to the 2D

ordering temperature

T~D. These

temperatures

are

higher

than in the

single crystalline sample presented previously.

This

point

will be discussed below. The

computation

of the actual

behavior of the AC

magnetic susceptibility

in such a

system,

a first

stage Co-GIC,

with

strong

c-axis

coupling

is out of the KT model

[5, 6].

~

. .

. .

"

.

l.6 (~l .'

T ~'~ lb) .

(

( ~ ~ .

(

l.2

j

$

j

~'~ .

~ $ ; % w

~ ~ 'T ~

~ DA .

~ J 2°Qj fi

~

= , ; ~ N

~ DA

d

°'~

~

~ 0

~ ~ ~ ~ ~ ~~ 0 2 4 6 8 lo

T(K) T(K)

Fig. 8.

-Comparison

in a HOPG intercalated sample, between the temperature variation of the

magnetic susceptibility

at 1.0 Oe excitation field (a), and the intensity of the (001/2) magnetic

Bragg

peak (b).

6. The

higher

stages.

The neutron

scattering

in second stage was

extensively

studied

by

Suzuki et al.

[10-13].

The correlation

length along

the c-axis is much smaller

(30 I)

and there is a coexistence between

long

range

(3D)

and very short range

(2D)

order.

However,

we believe that it is

possible

to

give

an

interpretation

of their data similar to our

interpretation

of the results on the first

stage compound.

In the second

stage compound,

the

fitting

of

(100) magnetic peak

is more difficult and

they

did not

study

the

temperature

variation of this

peak.

We have

investigated

a third stage

compound (in

a

powdered sample

on the two-axis

spectrometer DIB)

and found no

antiferromagnetic peaks although

a

peak

was observed in

magnetic susceptibility

measurements at around T~D = 8 K. We have also

investigated

a bi- intercalated

compound GBIC-Cocl~-GaC13.

The

sample

was

prepared

from a

single crystal,

but we could not observe

by

neutron

scattering

the

(100)

nuclear

peak

of the intercalate because the

in-plane

order is very weak. The

(00f) magnetic

line after the correction

by

the Lorentz factor is flat and decreases

slowly

with temperature between 1.6 K and 10

K,

in a way which is very similar to the

integrated intensity

of the

(100) peak

of the first stage

compound.

Unfortunately,

it was not

possible

to determine the

in-plane

correlation

length

as a function of

temperature

because of the absence of the

(100) peak.

In this

sample,

a

peak

was also observed in the

susceptibility

measurements near T~D = 6 K

[32].

In

conclusion,

we believe that the

higher

stages present a

divergence

of the

in-plane

correlation

length

which is very similar to that of the first stage

compound

and

corresponds

to the

peak

of

magnetic susceptibility.

The 3D order decreases as the stage increases. In the second

stage compound,

a short range 3D

antiferromagnetic

order was observed. In

higher stages

and in the bi-intercalated

compound,

no 3D

ordering

was observed. The

comparison

between the

magnetic properties

of the first and the

higher

stages shows that the 2D

ordering

(12)

is

only weakly

influenced

by

the 3D

coupling.

This allows us to

study

this 2D

ordering

even in the first stage

compound.

7. Discussion.

In this

part,

we will discuss three

points

: the

validity

of the 2D XY KT model which is

usually

used to fit the

magnetic susceptibility

in the

higher

stages the critical

temperatures

that we

have

found,

which are very low

compared

to other

studies,

and the

partial ordering

at low

temperatures.

7,I THE VALIDITY oF THE 2D XY MODELS. We have shown in the

previous

sections that

the

temperature

variation of the

in-plane magnetic

correlation

length

is

mainly

driven

by

the 2D

properties

of the

layer

and the 3D

antiferromagnetic

order does not affect it. For this

reason, we will now discuss the 2D XY model and

interprete

our results within this

framework.

The Kosterlitz-Thouless model

[Ii

of the 2DXY transition assumes that the transition is driven

by

the

separation

of vortex-antivortex

pairs

which are

responsible

for the ther-

modynamical stability

of the low

temperature phase.

The correlation

length f

above T~D is

given by

(1/t)

=

(I/to) exp(-

b/t~~~) with t

=

(T T~D)/T~D (I)

This behavior assumes that the system is

purely 2D, ferromagnetic

XY and infinite. In the present case, many deviations exist.

First,

there is a 3D

coupling

which appears at

T~,

and the

system presents strong

3D

correlations,

which however do not

strongly modify

the 2D

ordering.

Moreover in the KT

theory

the correlation

length

does not

really depend

on the dimension of the space

[33].

Such a

system

with 3D

coupling presents

a KT like transition

(which

is in fact

3D)

at a transition temperature

[33]

:

TIT

"

TKT(1

+

9/1r/ (J'/J~) (2)

(TKT

is the 2D

temperature arJs~/2,

s

=

1/2,

J'the

out-of-plane coupling).

This is the case in intercalated

compounds

where the vortices should be in fact 2D. Friedel has

recently

shown that the

opposite assumption

I-e- 3D defects

(here

3D vortices with the axis in any direction of the 3D

space) gives

a

completely

different result, even with a strong

anisotropy [34].

In this

model T~ is very small as J' is not a

perturbative

term and the 2D transition

might

be

destroyed.

We are

obviously

not in such a case.

There is a finite structural

in-plane

correlation

length f~,

~ which is the upper bound for the

magnetic

correlation

length f.

The effect of this finite size on the KT transition has been studied

theoretically by

Szeto et al.

[35].

We have observed that until

f

reaches the value

f~,

~ at 4 K, no effect can be observed on the temperature

dependence

of

f.

For these reasons, we fit the temperature

dependence

of

f by

a KT model as is shown in

figure

7b

(fo

=

8.8

I,

b

=

1.6 and T~D = 4.4

K).

In this

model,

b

parametrizes

the inflexion

changing

of

I/f(T)

as

predicted by

the Kosterlitz-Thouless

model, to,

the

high

temperature

limit of the correlation

length,

is a few times the unit cell parameter

(3.54i)

and

T~D should be the Kosterlitz-Thouless transition

temperature.

With s = 1/2 and

E~ being

the

energy of a vortex

creation,

for

E~/kT WI,

we find

T~~ =NJS~/2

=8.4K. The value

J

= 21.5 K of the XY

ferromagnetic exchange

interaction J was measured

independently by

inelastic

magnetic

neutron

scattering [36].

We will discuss in the next section the

discrepancy

between the measured T~D and

T~~.

The inflexion

point

:

T~

=

5.6 K

(Fig. 7b) corresponds

in

the model to a correlation

length approaching

the size of the core of the vortices:

(13)

302 JOURNAL DE PHYSIQUE I N° 3

f(T~ )

=

20, I

to. Beyond

this

point,

the interaction energy between vortices is

different,

since these can no

longer

be considered as free vortices and the Kosterlitz-Thouless

theory

is not

valid,

but this inversion of the curvature remains as it characterizes the smoothness of the

transition.

7.2 THE VALUES OF THE CRITICAL TEMPERATURES. The critical temperatures

T~~

and

T~~

that we have found in the

single crystal (4

K and 6.5

K)

are lower than those found

by

others

(8

K and 9

K).

The

interpretation

of this

discrepancy

is still not

completely

clear. The presence of a

high

rate of cobalt vacancies in the cobalt

layer

can be at the

origin

of this difference. Nicholls

[19]

has

recently proposed

a model where the vacancies are

completely disordered,

as was

suggested by

our small

angle

neutron

scattering

measurements which were unable to observe any

long

range order in the vacancies. In this

model,

the presence of the

vacancies is treated in a mean field

theory

where the average number of first

neighbors

decreases

(in

the present case 5 instead of

6).

Hence, the effective J

in-plane coupling

value decreases and then T~D decreases. The decrease of

T~

is more difficult to describe since we have to know the

origin

of the 3D

coupling, dipole-dipole

interaction or

super-exchange

across the chlorine and the

graphite layers.

Both models

give

reasonable values of

T~

but the

dipole-dipole

interaction model

[29] predicts

a limitation of

f~

at

f~,~

and a

destruction of the

antiferromagnetic ordering

temperature

TN

for a value of

f~

~ of a few

l~'s

which is indeed observed in

figure

7b. The

discrepancies

between the transition

temperatures among different

samples

then arise from the different

filling

factors of the cobalt atoms in the GICS. It should be mentioned here that it is very difficult to intercalate

graphite, particularly single crystals,

with

CoC12

to the first stage with a

high filling

factor

[17].

7.3 THE PARTIAL ORDERING AT Low TEMPERATURE. As we have

already

mentioned,

only

64 fb of the

spins

are ordered at low temperatures

although

the

peak intensity

saturates. The

integrated

intensities of the

Bragg peaks

measured

by magnetic

neutron

scattering

do not

correspond

to the

magnetization

of one of the two sublattices but

correspond

to the local

magnetization averaged

on a size

equal

to the correlation

length.

For

example,

the presence of vortices in the

layer

decreases the observed

integrated peak intensity

not because the

average

magnetization

should be zero, but because the vortex cores

give

a

scattering

which is very flat in the

reciprocal

lattice and hence is not taken into account in our

integration

of the

peak intensity.

We believe

(but

this remains somehow

speculative)

that the presence of such defects in the low temperature

phase

is due to the presence of the cobalt vacancies which

pin

the vortices.

More measurements will be necessary to confirm or infirm this

model,

and in

particular

measurements under

magnetic

field.

8. Conclusion.

We have measured the

magnetic ordering

of

CoC12

intercalated in

graphite,

and we have

shown that the 2D

ordering

seems to be very

weakly

influenced

by

the 3D

ordering

which appears at

T~.

We have then

analysed

the 2D

ordering

in the Kosterlitz-Thouless framework.

In

particular

the

magnetic in-plane

correlation

length

varies more

smoothly

with temperature than the c-axis one, as is

predicted

in the 2D KT

theory. Only

part of the

spins (64fb)

contributes to the

magnetic Bragg peaks

in the low

temperature phase showing

the presence of local disorder. We attribute these zones to the vortex cores

predicted

in the KT

theory.

The

presence of cobalt vacancies which

explains

the differences in the transition temperatures between different

samples

does not

perturb

too

strongly

the mechanism of appearance of the 2D

ordering.

(14)

Appendix

1.

Stacking

faults of

Coclz

first stage.

The calculation was

performed according

to the

following procedure

:

The

in-plane

lattices of

graphite

and intercalate

layers

are incommensurate, so the two orders do not interfere in the

(10f)

line. Since the effect of the

staging

faults is

negligible compared

to the effect of the

stacking faults,

we have assumed a pure first

stage compound.

We use a model

describing

the N

equidistant layers

as

homogeneous

Markov chains of first order

[25, 26, 36],

which represent a first

neighbour

interactions.

The

probability

to be in the state e~ in the

plane

n is

p(e,)

= p,. The

probability

to be in

the state e~ in the

plane

n if the

plane (n

I

)

is

e~ is

Q(.

Let A be the matrix of coordinates of the

change

of basis which

diagonalized Q.

The diffracted

intensity

is

I(I)

=

Real

((PWI

AAA ~'

W)

where A is a

diagonal

matrix

Aj

=

(I

+ A~

exp(-

2 I

grit)/(I

A~

exp(-

2

igri))

+

+

(I

+ A~

exp(2 igri))/(I

A~

exp(2

I

gri))

The

peaks

occur for A~

exp(±

2

art)

= I.

We also introduce an

in-plane

translation

3

of the

(n

+

I)th plane

with respect to the nth one

by replacing

A~

by A~exp(2iar3(ha*+kb*)).

The

physical meaning

of this translation is not

simple.

It is

actually

an indirect way to introduce random

stacking

faults

[24]

which would have been very difficult to calculate in the presence of commensurate

stacking

faults. The altemative calculation

(only

random

faults)

was

performed

elsewhere and

supply

a

poorer fit

[36].

A

layer

of

Cocl~

is made of three

hexagonal

2D lattices stacked in rhomboedral

positions.

Let u and d be the

symbols

for the upper and lower chlorine

planes,

ABC the three

possible positions

of the 2D

hexagonal

lattice. The

following

table

gives

the different states

(C

is the

permutation

of ABC and 8

exchange

of the two chlorine

positions).

W~

=

b~~

+ 2

b~j

cos 2

ar(df

±

1/3).

State Positions W Structure factor Name

ei

A~ BC~

W+ II

li(ei

+ e~ + e~ + e4 + e~ +

e~)

ej

C

(ej B~ CA~ j

W+ II

li(ei

+ e~ + e~ e4 e~

e~)

e~

C~

~(ei) C~ AB~ j2

4~+ II

/(ei

+

je~

+

j~ e~)

e~

8C~

~(ej B~ AC~ j~

W~ II

/(e4

+

j~

e~ +

je~)

e4

8C(ej) A~ CB~ j4~-

II

~/3(ei

+

j~e~

+

je~)

e~

8~

~(ei) C~ BA~

4~- 1/

/(e4

+

je~

+

j~ e~)

e~

(15)

304 JOURNAL DE PHYSIQUE I N° 3

Since all

positions

are

equivalent,

p

=

1/6. The conditional

probabilities

are listed now

State n, state n + I

Probability

e~,

C(e~)

I

=

1, 2,

3 2 p

e~, C~ ~(e~) I =

4, 5,

6 2 p

e~,

C(ei)

I

=

4, 5,

6 2

a

e~, C~

'(e~)

I

=

1, 2,

3 2 a

e~,

8(e~)

I

= I to 6

fl

e~, e~ I

=

I to 6 e

e~,

8C(e~)

I

=

I to 6 v

e~, SC ~(e~

)

I

=

I to 6 v

with the condition s + 2 a + 2 p + fl + 2 v

=

I.

The

diagonalization gives

the

following eigenvalues

:

i

s+2a+2v-2p-fl

and the

eigenvalues

of the matrix

Is +2ja

+

2/p (fl v)J

(p

+

v)j~

s

+2j~a

+

2jp

The best fit is shown in

figure

3b with the parameters :

a p =

0.13, fl

v

=

0.004,

s a p

=

0.025,

no

background,

755

planes

and 3

=

0.07.

References

[ii KOSTBRLITz J. M., THouLBss D. J., J. Phys. France

Colloq.

34 (1973) C6-l181.

[2] REGNAULT L. P., RossAT-MIGNOT J., Magnetic

properties

of layered transition metals com- pounds, de

Jongh

and Willet Eds.

(Reidel,

Dortricht, 1987).

[31 KARIMOV Yu., Sov.

Phys.

JETP 41

(1976)

772.

[4] SHAYEGAN M., DRESSELHAUS M. S., SALAMANCA-RIBA L., DRBSSBLHAUS G., HERBMANS J., IssI J. P.,

Phys.

Rev. B 28 (1983) 4799.

[5] ELAHY M., DRESSELHAUS G.,

Phys.

Rev. B 30 (1984) 7225.

[6] CHEN S. T., SzETO K. Y., ELAHY M., DRESSELHAUS G., J. Chim.

Phys. 81(1984)

863.

[7] The stage of the intercalation is the number of

graphite planes

which separate two nearest intercalated layers.

[8] SZETO K. Y., CHEN S. T., DRESSELHAUS G., Phys. Rev. B 32 (1985) 4628.

[9] SIMON Ch., BATALLAN F., ROSENMAN I., AYACHE C., BONJOUR E.,

Phys.

Rev. B 35 (1987).

[10] SUZUKI M., IKEDA H., ENDOH Y.,

Synth.

Met. 8 (1983) 8.

[I Ii WIESLER D. G., ZABEL H., SUZUKI M.,

Synth.

Met. 23 (1988) 237.

[12] WIESLER D. G., ZABEL H., SUZUKI M., Phys. Rev. B 36 (1987) 7051.

[13] WIESLER D. G., ZABEL H., Phys. Rev. B 36 (1987) 7303.

[14] RANCOURT D. G., J. Magn. Magn. Mater. 51(1985) 133.

jls] FLANOROIS S., HBWAT A. W., HAUW C., BRAGG R. W.,

Synth.

Met. 7 (1983) 305.

j16] SIMON Ch., BATALLAN F., ROSENMAN I., SCHWBIzBR J., LAUTER H., VANGBLISTIR.,

Synth.

Met.

28 (1983) 53.

(16)

j17] CHOUTEAU G., SCHWEIzER J., TASSET F., YAzAMI R.,

Synth.

Met. 23

(1988)

249.

[18] HACHtM L., Th~se de docteur

ingdnieur

de l'Universitd de

Nancy1 (1984, unpublished).

[19] NicHoLLs J. T., DRESSELHAUS G.,

Phys.

Rev. B

41(1990)

9744.

[20] Neutron research facilities at the ILL flux reactor, H. Blanck Ed. (ILL publications Grenoble, France, 1988).

[21] LEUNG S. Y., DRBSSBLHAUS M. S., UNDBRHILL C., KRAPCHB V., DRESSELHAUS G., WUENSCH B. J.,

Phys.

Rev. B 24 (198I) 3505.

[22] PERNOT P., Wise de I'Universitd de

Nancy1 (1989, unpublished).

[23] WYCKOFF R. W. G.,

Crystal

structures, 1(J.

Wiley

and Son, New York, 1963).

[24] GUtRARD D., LELAURAIN M., AUBRY A., Bull. Sac. Fr. Min. 98 (1975) 43.

[25] MARTI C., THOREL P., CROSET B., Acta

Cryst.

A 37 (1981) 609.

[26] PtRIGNON A., Th~se de docteur de troisidme

cycle

de l'Universitd de

Nancy

(1987,

unpublished).

[27] ROGERIE J., SIMON Ch., ROSENMAN1., SCHWEIzER J., PERNOT P., VANGELISTIR., Intemal report of the ILL

(1988, unpublished).

[28] SPECK J. S., DRESSELHAUS M. S.,

Synth.

Met. 34 (1989) 211.

[29] ROGERIE J., Th~se

unpublished.

[30] WARREN B. E.,

Phys.

Rev. 59 (1941) 693.

[3 Ii DUTTA P., SINHA S. K.,

Phys.

Rev. Lett. 47 (I981) 50.

[32] ROSENMAN I., BATALLAN F., SIMON Ch., HACHIM L., J. Phys. France 47 (1986) 1221.

[33] KOSTERLITz J. M., THouLEss D. J.,

Progress

in Low

Temperature Physics,

D. F. Brewer Ed. 78 (Amsterdam North-Holland 1978) p. 372.

[34] FRIEDEL J., J. Phys. France 49 (1988) 1769.

[35] SzETO K., DRBSSBLHAUS G.,

Phys.

Rev. B 32 (1985) 3142.

[36] ROGERIE J., SIMON Ch., ROSENMAN I., SCHWEIzER J., PERNOT P., VANGELISTI R., J.

Phys.

France

Colloq.

49

(1988)

C8-1437.

[37] HBNDRICKS S. B., TELLER E., J. Chem.

Phys.

10

(1942)

147.

[38] PLANqON A., TCHOUBAR C., J.

Appl. Cryst.

9 (1976) 279.

Références

Documents relatifs

shows the effect of an applied magnetic field H, parallel to the layers plane, on the magnetic suscep-. tibility x of the biintercalated compound : G-CoC12-

In other cases the Curie temperature is smaller than the crystallisation temperature and the annealing up to the Curie temperature induces a more rapidly decrease of p, with

The doubling of the number on nuclear internal fields suggests an antiferromagnetic <( cross structure )) with four magnetic sublattices.. The projections of the

We prove that if the initial magnetic field decays sufficiently fast, then the plasma flow behaves as a solution of the free nonstationnary Navier–Stokes equations when |x| → +∞,

The main result of this work is that the entropy as- sociated with the specific heat indicates that the ~ d spins become substantially ordered as the tem- ~ + perature is

influence of the symmetry of the magnetic interaction and its connection with the Mermin-Wagner theorem, and (ii) the singularity which the surface magneti- zation

We have reported experimental evidence of antiferromagnetic spin 1/2 finite chain behaviour in the new CuC12.2AlC13 graphite intercalated complex, using susceptibility and high

We also note that their comparison with data /3/ obtained above 1.5 K (an extrapolation to low temperatures of a plot of C vs.T) leads to an underestimate of the low