• Aucun résultat trouvé

Model calculation of the static magnetic susceptibility in light rare earth metallic systems

N/A
N/A
Protected

Academic year: 2021

Partager "Model calculation of the static magnetic susceptibility in light rare earth metallic systems"

Copied!
14
0
0

Texte intégral

(1)

HAL Id: jpa-00246369

https://hal.archives-ouvertes.fr/jpa-00246369

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.

Model calculation of the static magnetic susceptibility in light rare earth metallic systems

Y. Hammoud, J. Parlebas

To cite this version:

Y. Hammoud, J. Parlebas. Model calculation of the static magnetic susceptibility in light rare earth metallic systems. Journal de Physique I, EDP Sciences, 1991, 1 (5), pp.765-777. �10.1051/jp1:1991168�.

�jpa-00246369�

(2)

Classification

Physics

Abstracts

71.28 75.30M

Model calculation of the static magnetic susceptibility in fight

rare

earth metallic systems

Y. Hanunoud

(')

and J. C. Parlebas

f)

(') Physics

Department,

Syrian

A-E-C-, Damascus-S-A-R-,

Syria

(2) IPCMS

(*),

Universitk Louis Pasteur, 4 rue Blaise Pascal, 67070

Strasbourg,

France

(Received

7

May

J990, reviYed

23January

J99J,

accepted

30January

J99J)

Rksumk. Nous utilisons le moddle d'Anderson I une

impuretd

dans

l'approximation

des

grands Nr

off

Nr

est la

ddgdndrescence

d'orbitale et de

spin

du niveau f et nous calculons alors la

susceptibilitd paramagndtique statique (I tempdrature nulle)

dans les

systdmes mdtalliques

de terres rares

ldgdres.

Nous effectuons notre calcul pour des valeurs de l'interaction de Coulomb Urr

grandes

par rapport I

l'hybridation

V des

configurations

f et f

avec les dtats de conduction

(c.-i~d.

la

configuration I°~i

nous ne retenons que [es tenures [es

plus importants

dans un

ddveloppement

en

puissances

successives de

lIUrr

et V. Ensuite nous discutons nos rdsultats

numdriques

I

partir

d'une forrne

analytique simple

obtenue pour la

susceptibilitd magndtique

en

fonction de la

position

du niveau f, de

l'hybridation

V, de la forrne et du

remplissage

de la bande de conduction et enfin des effets de Urr fiat. Finalement nous

prdsentons

des courbes de

susceptibilitk

que nous avons calculkes en fonction de V et en liaison avec la transition

a y du cbrium, en utilisant les mdmes

paramdtres

que ceux

qui

interviennent dans un

prkcbdent

calcul de spectres

Lnj d'absorption

de cwur. Ainsi,

kgalement

dans le cas de la

susceptibilitk, l'hybridation

V

apparait-elle

ccmme un

paramdtre important

pour dkcrire le

changement

de

phase

ay du cfirium.

Abstlract.

Using

the

impurity

Anderson model in the

large

Nr

approximation,

where

Nr

is the orbital and

spin degeneracy

of the f level, we calculate the zero temperature static

paramagnetic susceptibility

of

light

rare earth metallic systems. The calculation is

perforrned

for

large

values of the Coulomb Urr electron-electron interactions with respect of the V

hybridization

off' and f~

configurations

vith the conduction states

(I.e.

configuration)

: we

only keep

the

leading

terms in a

development

in successive powers of

I/Urr

and V. Our numerical results on the

magnetic susceptibility

start from a

simple analytic expression

and are discussed in terms of the f level

position,

the

hybridization

V, the

shape

and

filling

of the conduction band and also the finite Urr effects.

Finally

we present calculated curves for the

susceptibility

versus Vin connection with the a y transition of cerium and

utilizing

the same parameters as those used

previously

to obtain

core level

l~ii absorption

spectra : also in the case of the

susceptibility,

the

hybridization

appears

to be an

important

parameter to describe the

phase change

from y to a cerium.

(*)

UMR 46 CNRS.

(3)

1. Inboduction.

The valence

fluctuating

rare-earth systems show a Fermi

liquid

behaviour at low

tempera-

tures their

paramagnetic susceptibility

is of Pauli

type

and

approaches

a finite

(although quite large)

value at zero

temperature.

The

physics

of the finite

susceptibility

at low

temperatures

is the renormalization of the local rare earth moments

by

the conduction electrons

leading

to a

~paramagnetic) singlet ground

state of the system. Therefore no

magnetic ordering

has been observed

alihoujh

one of the valences of the considered

systems

has a

magnetically

active

ground

state

(for example Ce~+

as

compared

to

Ce~+).

From a theoretical

point

of

view,

it has been demonstrated that the low temperature

properties

of the

impurity

Anderson model can be described in an

empirical

local Fermi

liquid theory ([1~4]

and Refs.

therein)

where the

impurity

is described in terms of what it does to the conduction electron states. In the

present

paper we

adopt

an alternative way of

describing

a

single

mixed-valent

impurity,

I.e. we focus on the

ground

state of the

impurity

Anderson

Hamiltonian where various atomic

configurations hybridize

with the conduction states so that

one has a renormalized

configuration picture.

This

point

of view based on a

I/Nr expansion (where Nr represents

both orbital and

spin degeneracy

of the 4f

level)

has been

adopted

to calculate various spectra

~XPS, XAS, BIS...)

of rare-earth

compounds ([5-8]

and Refs.

therein)

: the lowest-order version of this method

(in

terms of successive powers of

I/Nr) provides

an exact solution to the so-called filled band

(insulating)

model

[8]

where the

density

of states is zero above the Fermi level ; in the case of metallic systems

(where

the conduction states extend well above the fermi

energy)

it

provides

an

asymptotically

correct solution in the limit

Nr-

co either for

U~-

co

[5]

or for finite

Urr [6, 7].

The

present

short paper is based on an extension of a

theory

of core level spectra

by

Hammoud et al.

[9, 10]

in order to obtain an

analytic

formulation of the zero

temperature

static

magnetic susceptibility

for finite

(but large)

or infinite Coulomb

U~

electron~electron

interactions. The outline of the paper is as follows : first

(Sect. 2)

we

briefly

recall how to solve the

impurity

Anderson Hamiltonian

using

an extended space of states within which the

ground

state wave function is calculated

[6~8].

Also a small external static

magnetic

field is switched on in order to define

analytically

the static

susceptibility [5] (Sect. 3.I).

Our

analytic

results are then discussed

numerically

in the case of infinite

(Sect. 3.2)

and finite

(Sect. 3.3).

U~. Especially

it is

quite interesting

to calculate a

thermodynamic quantity

such as the zero-

temperature susceptibility utilizing

the same parameters as those used to describe

spectro- scopic properties [9, 10].

This is done for the ay transition of cerium

(Sect. 3.3).

2. Formulation and resolution of the Anderson model.

We consider a metallic

light

rare earth system

consisting

of a conduction band

(CB)

treated as

non-degenerate

and

hybridized

with an

N~fold

orbital and

spin degenerate

f level of energy

El.

The

paramagnetic instability

of the f level is

directly

reflected in the f static

susceptibility.

In order to calculate that

quantity,

it is necessary to switch on a

(small)

extemal static

magnetic

field h which removes the

degeneracy

of the f level and

gives

rise to the Zeemann terms

corresponding

to the ionic

multiplet

of the rare earth element.

Here,

for

simplicity,

we

disregard

the field

dependence

of the CB levels so that there will be no CB electron

polarization

and we write the

h-dependent split

f energy sub~levels s~ as follows

s~=s/-gjp~mh; M=-J, -J+I;

+J

(2,1)

where gj is the Land6

gyromagnetic

ratio of the considered rare-earth ion. The

magnetic

quantum

number, M,

represents both orbital number m and

spin

direction « of an f state with

(4)

Nr

=

2J+1. From the h

dependence

of

equation (2.I)

we can then write an

implicit

h~

dependent

Hamiltonian as :

H =

£

s~

a£~

a~~ +

£

s~

ajj

a~ +

U~ jj ajj

a~

a[,

a~, +

£ V~~~ a[

a~~ + h.c.

,

(2.2)

km M M»M' Mk«

where

a£~

is an

operator creating

a conduction electron with wave number

k, spin

« and

energy s~ whereas

ajj

is that

creating

an f electron in the M state with energy

s~. The third term of

equation (2.2)

describes the Coulomb

repulsion Urr

between two f electrons and the fourth term exhibits the

hybridization V~~~

between f and conduction states. It is well known that a variational calculation of the

ground

state leads to the same results as first order

Brillouin-Wigner perturbation theory

when double

occupation

of the f level

(I.e. 4F configuration)

is

neglected [5, 9, 10].

That

corresponds

to the

asymptotic

limit of

infinite Urr. More

generally,

the

(h-dependent) singlet

mixed valent

ground

state

#) of,£f, given by equation (2.2),

in which

hybridization

admixes

magnetic configurations,

can be written as a linear combination of basic states

corresponding

to 1°,

~

and f~

configurations

and

preserving

the

singlet

character of the

ground

state [6~8]

~'~ ~~ ~

~

~~~" ~~

~~" ~

l, ~ij ~~ ~~'~''~'

"' ~2"2

~~ ~ji, ak,

«iak~

«~l

tbb

) (2.3)

where

4~)

is the

singlet ground

state of the

41° configuration

with energy

E~,

I-e- the filled Fermi sea of the CB electrons up to the Fermi energy s~.

< «

'ffib)

"

fl ~~« '°)

i

l~b

"

I

Sk

(~.4)

km km

In

equations (2.3)

and

(2.4),

the

inequality

< means that

only occupied

states are concemed

both in the summations and in the

product o)

is the vacuum state. The second and third terms of

equation (2.3)

are obtained

by transferring

one and two electrons

(respectively)

from the CB to the f level which

corresponds

to the

4f'

and

4f~ configurations. Now, using

the

(h-

dependent) Schr6dinger equation

Hip)

=

E(h) (~) (2.5)

in the limit of

large degeneracy

and at the lowest orders in

V~~~

and

I/U~ (for example, (V~~~)~/U~

and

higher

power terms are

neglected),

the energy variation

3E(h)

m

E(h) E~

of the considered

system

due to

hybridization

and extemal

magnetic

field

obeys

a

characteristic self-consistent

equation

shown in

Appendix

A.

Then, using

the normalisation condition that

(~ ~

=

l,

we obtain the

following

relation between the various

h-dependent

coefficients of the

ground

state

(2.3)

:

1co(h)

i~

=

ii

+

ci(h)

+

c~(h)1-

'

(2.6)

Ci(h)

=

it lCMk«l~ (2.7)

~

«

and

=

" "

~~~~~ ~j I I '~MM'kj

« k ~ (~

~' "' ~''

' ~ ~

(2.8)

(5)

Explicit equations

for

C~~~

and

C~~,

~~~~~~~~ are

given

in

Appendix

A. Also the

h-dependent

f electron number in the

ground

state is defined as usual

by

:

~f"

I (~ (~~ ~M( ~) (~.~)

M

and

keeping

the

leading

terms in successive powers of

V~~~

and

I/U~

we obtain a

simple analytic expression

of nr, I.e.

n~=

~'(h)+2C2(h)

I +

Cl (h)

+

C2(h) (2.10)

Of course we recover the limit nr -

(I Co ~)

when

U~

- co

(see [9]

for

instance,

and also

Figs. la,

b and

cl.

~°(£> (d> (e>

~

n

o-s

.1.0 0.0 1-O

E/D

~

(i>

~

/

f-level

~

-2.O -I.O O-O I-O a-O

t-lmel

Fig.

I.-Behaviour of nr and e, in the infinite Urr limit, with respect to the f level

position sllDj,

the

origin

of the

energies being

the Fermi level s~:

(a), (b)

and

(c)

represent nr for

(V/Dj)

=

0, 0.04 and 0.08;

(d)

model DOS

represented by

two

semi-elliptic

curves vith

ej/Dj

=

0.50;

s~D~

=

0.25 and

DjDj

=

0.15

(e)

coefficient e vithin the same model DOS as

(d)

and

V/Dj

=

0.12 ; (o coefficient e within a rectangular DOS of energy width 2Di (which allows an analytic

investigation).

3. Calculation of the static

magnetic susceptibility.

In this

section,

first we formulate a new

analytic expression

of the static

paramagnetic

susceptibility

for finite but

large

values of U~r

(Sect. 3.I),

then we look at the infinite

(6)

U~

limit and

present

some numerical results within this limit

(Sect. 3.2). Finally

we discuss our results for finite

Urr (Sect. 3.3)

with a

special emphasis

on the

a y transition of cedum and

by using exactly

the same

parameters

as in the

Lm

XAS

spectra

calculations

[9, 10].

3.I ANALWIC EXPRESSION. In the

preceding

section we calculated how the total energy of

a

polarizable singlet ground

state,

E(h) =E~+ 3E(h), depends

on an extemal static

magnetic

field h. The

paramagnetic susceptibility

can then be deduced

by differentiating

twice

3E(h)

with

respect

to the field and then

by taking

the limit of the

h~dependent

result as h - 0

[5]

:

X = lim

°~~~)~~ (3.I)

h-o ah

Within the

preceding development

in successive powers of

V~~~

and

I/U~

we can cast X in the

following simple analytic

form

(see

also

Appendix B)1

~~ 2

~

§ ~~J

~

Bl~ ~~~

~

l

r

(

n

rj ~~'~~

with

ici(li@11(o)1216 ~~~~liil~~~~~ ~~'~~

In

equations (3.2)

and

(3.3),

r has the dimension of an energy and is defined as : r

= p

(s~) Nr V~

p

(s)

= ~

£ V~~~

~ 3

(s s~) (3.4)

fi~f ~

Mk«

In

equations (3.4), p(s~)

represents the

hybridized density

of states

(D.O.S.)

at the Fermi energy e~. Before

hybridization

the

unperturbed

C.B.D.O.S. is

just

noted

p°(s)

per

spin

direction :

p

°(E)

=

z

8

(E Sk) (3.5)

where N is the number of considered sites in the lattice. Let us also remark that what appears in

equations (3.4) is,

as

usual,

a constant effective

hybridization [4, 8]

:

(V~~)~

=

Nr V~ (3.6)

where

V~~

is

independent

of

Nr

so that V is of the order of

I/ @,

and for

example

the

leading

term in successive powers of

(V/(Urr/

~~ with

j

< 3 is of the order

I/(Ni U~).

We notice

finally

from

equation (3.2)

that X

depends

very

sensitively

on ni,

especially

it increases

(towards

« a

magnetic

state

»)

when ni

approaches unity

as is the case for

(almost~)

trivalent Ce systems like

yce. However,

in addition to nr, there appears also a characteristic

coefficient e in our

analytic

result

(3.2)

and we will see next what is the influence of e and more

generally

what is the behaviour of X first in the infinite

U~

limit

(Sect. 3.2)

and

then in the finite one

(Sect. 3,3).

Our calculations are all

performed

with

Nr

=

14,

I.e. we

neglect

the

spin~orbit splitting,

(7)

3.2 CASE oF INFINITE

U~.

In order to

study

the dimensionless coefficient e of

equations (3.2)

and

(3.3)

in more

details,

we take first the infinite

Urr limit,

in which case

C~

- 0 as well as its derivative versus

3E(h),

and we

adopt

the

following simple

models for

the DOS p

(s)

of the C.B.

(I)

a

rectangular

DOS of width

Dj

centred at sj ;

then, everything

is

analytic and,

after

some

calculations,

the coefficient e reads :

2[8E(0) s/J

+ sj

Dj

e

=

(3.7)

El

Di

(it)

Two

semi-elliptic

DOS as in

YPd~ compounds

for

example ([9]

and Ref.

therein)

for which

p(s)

is

expressed

as

(see Fig. ld)

:

p

(E)

=

£ p;(E) p,(E)

=

p;I(D,)~ (E E;)~l'°/D;, (3.8)

where s;,

D,

and p, are the centre of mass, the half-width and the maximum of the

i~th DOS

(I

=

1, 2).

The behaviour of e is shown in

figure

I with

respect

to the

position e)

of the

level,

the

origin

of the energy axis

being

s~, curves

(e)

and (f~

correspond

to the considered DOS models

(I)

and

(iii respectively. First,

the

particular shape

of the DOS has little influence on the overall

function e of

El only

in the

limiting

case of e-1. This

happens

for

deep

values of

El

as

compared

to the

C.B.,

I.e.

typical

Kondo systems for which

Elm

s~ or more

precisely

:

Nr V~/(s~ El)

«

Dj

nr < I

(Ce) (3.9)

Then it is

possible

to recover a well-known result

[5,

6]

(although

less

general

than

equation (3.2)

which is valid for the whole range of

El values)

x =

j (gJp~)~ J(J

+

1)

~

~) (3. lo)

where

kB TK

= nr r

(3.

I

I)

nr

defines a Kondo temperature which represents the energy

gain

due to the stabilization of the

singlet (see

for ex.

Eq. (4.12)

of Ref.

[6]).

In

equation (3. II)

the Kondo energy is linear in

(I nr)

when V is

kept

constant,

leading

to very low

T~

and very

high (although finite)

X

when nr tends to I like in

(almost)

trivalent Ce systems. Gunnarsson and Sch6nhammer

already pointed

out

numerically

the rather broad range of

validity

of

equation (3.10) (see Eq. (3.19)

and

Fig,

of Ref.

[7]). However,

let us

point

out

that,

for finite but

large

values of

U~

or for infinite

U~,

our

analytical

result

(3.2)

is more valid than

equation (3.10) especially

for

El

> s~ where e increases

linearly

with

El

and cannot be

replaced by

I. In the

vicinity

of s~, e is a non-linear function of

El

via

8E(0).

Since we will be

finally

interested in the behaviour of X with respect to the

hybridization

V,

we also

plot

the e function versus V

(Fig. 2)

for two values of the

El

level below and above the Fermi level s~. the e function increases

sensitively

with

increasing

V in the two cases because

(8E(0) s/(,

which expresses the e

dependence

upon V, also increases with V whatever

El.

For the curve of

figure

2

corresponding

to

El

above s~, e can never reach

I,

even for

(8)

i~

~

o

CID #

5 f i

o.3

o.3

o-i ' Q.z Q-a

M;.bfldization

Fig.

2. Coefficient e

(within

the model DOS of

figure

ld and Urr -

oJ)

with respect to

hybridization V/Dj

for two values of

e)/Dj

= 0.3 and 0.3.

vanishing

V.

Consequently

the

simple

form of x

given by equation (3.10)

should differ

appreciably

from the more

general equation (3.2), containing

e, unless the

typical

case of Kondo systems is considered

according

to

equation (3.9).

When the

position El

is lowered from about the Fermi energy to below the CB

(Fig. 3a)

the

susceptibility

increases more and more whereas the situation is reversed for

SIN

s~

(Fig. 3b).

From the various

hybridizations corresponding

to

figure

3a it is clear that

increasing hybridization

stabilizes the «

non-magnetic

» state

(low susceptibility)

while smaller

hybridi-

zation favours the

tendency

towards

magnetism,

I.e.

large (although finite)

values of the

susceptibility.

A similar

tendency

has been

already

encountered for

example

in SmS

insulating compounds [12].

Instead of

varying hybridization

V, like in

figure 3a,

it is

possible

to obtain the same

qualitative

behaviour of x for a

given

value of V

by varying

the

filling

of the CB

(Fig. 4a).

In the insert of

figure

4 we checked the influence of

Urr

on x as is

explained

below.

3.3 CASE oF FINITE

U~. Analytic equations (3.2)

and

(3.3)

allow us to test the influence of

Urr

on the

susceptibility,

at least at the lowest order of our

development

versus

I/Urr,

X

= X

(Urr-

co

) (I

+ O

(I/Uj)),

I,e, the test will be more correct for

large

values of Urr. From the results of

figure

4b

( Urr

- co

)

and

figure

4c

( Urr

= 2 D

j

),

it appears that the

asymptotic

behaviour of X is almost reached for

El

values in the

vicinity

of the bottom of the CB.

According

to the numerical results of Gunnarsson and Sch6nhammer

(see Fig.

I of Ref.

[7]),

the X

dependence

upon

Urr

which has been

magnified by

the use of a Urr

logarithmic

scale is strong

only

for

relatively

small values of

Urr (typically Urr

< bandwidth

(9)

la

V/D~« V/D~.

0.08

0.08 o, i o

o,16

d ~'~~

p I

I

d

(

(b)

~1

~

~n on m

0,14 M"e'

o, 16 2

(a>

o

-i,o -o~ o-o o~ i-o

f-level

Fig.

3.

-Susceptibility x' (in (Dj)~~ units)

versus f level

position s)/Dj (within

the two

semi-elliptic

model DOS and in the infinite Urr

limit)

for various values of

hybridization

vith the

following

definition x'

= x

(gJHB)~

~ (J +

1) (a)

the range of x' is [0, 12]

(b) magnified

view of x' in the range

[0,

3 0.5].

in the units of Ref.

[7]),

the details of the

dependence changing

with other

parameters.

We also obtain a

significant Urr

variation of X in the low

Urr

range

(see Fig.

5a for

V = 0.96 eV and

El

= 2.5

eV) although

we are not allowed to

explore

this

Urr regime

too

deeply

because of our

asymptotic development

versus

1/Urr,

therefore

V~/Urr

should still be

negligible.

Moreover the

Urr dependence

of x is also found to be very sensitive upon the other parameters,

especially

the V

hybridization

and the

El position (Fig. 5).

Small effects of finite

Urr

values on

susceptibility

versus

hybridization

are also shown in

figure

6a where the parameters chosen

correspond exactly

to those

adopted

in references

[9, 10]

to describe the ay transition of cerium : in order to mimic the 5d-6s band of

Ce,

we use

here

only

one

semi~elliptic

DOS of half-width

Di=4eV

and

(1/3)rd-filled

with

El

= 1.2 eV with

respect

to the Fermi level at 0. Then

again

the

susceptibility

obtained for

Urr=8eV (which

is a reasonable value for

cerium)

is

quite

close to the result for

U~-

co. In

figure

6b the

corresponding

core

absorption spectra (observed experimentally and)

calculated in reference

[9] (and

Ref.

therein)

exhibit the

phase

transition from

yce

with V

< 0.20 eV to ace with V

> 0.20 eV. These orders of

magnitudes

for

hybridi-

zation

yields

a reduction factor of at least

II10

when

going

from X

( YCe

to X

( ace) (Fig. 6a).

Since the

present

model is

quite crude,

it is hard to make a more detailed contact with

experimental susceptibilities;

in

particular

as our calculations

only

include the f-level

(10)

2

1-S

~l~~l"

~

)

,~

(c) (b)

lli j

)

o.50

o~

# I

~Q

(

o-w

-I.o -o.5 0.o 0.5

(

f-level

c9

0l5

'

o.25

",,

(a)

0,oo

o

-i.o -o~ o-o o~ i-o

t-let,el

Fig.

4.

Susceptibility x' (in (Di)~

' units same definition of x' as in

Fig.

3) versus the Llevel

position sllDi (within

the two

semi-elliptic

model

DOS)

:

(a)

for various

fillings

of the band model, I.e. for

various

positions

of

ei/Di

(with s2/D~

=

ej/Dj

0.25,

V/Dj

=

0.12 and Urr - oJ); (b) for

ej/Di

~ 0.50 ;

e2/D2

=

0.25

V/Di

~

0.08 and Urr - oJ

(c)

for

ej/Dj

=

0.50 ; e~/D~ = 0.25 ;

V/Dj

=

0.08 and Urr/Dj = 2.

contribution to X, the CB contribution has to be subtracted from any

comparison

to

experimental

data.

Moreover,

whereas the zero temperature static

susceptibility

of ace is well

defined,

the

corresponding susceptibility

for

yce

is not, since we are then

dealing

with a

high temperature phase. Nevertheless, already

at room temperature,

X(ace)

measured under pressure ~p a 7.6 kbar

[13])

is

approximately

reduced

by

a factor of

I/4

with respect to

X(YCe)

at lower

(or normal)

pressure.

Finally

let us also remark the

analogy

between

increasing

pressure

(experiment)

and

increasing hybridization (Fig. 6a)

in order to make the transition y

- a in cerium without

invoking

a

large magnitude

of valence v

change (see

for

ex.[14]):

e.g. for V=0.16eV we have v= 3.06

(y phase),

whereas for

V

= 0.28eV we obtain v

= 3.38

(a phase).

Moreover our

analysis

is very similar to the

« Kondo Volume

Collapse

» model

[15]

where a ratio

J~/J~

m 2 has been used to describe the y - a

phase

transition in cerium. This

corresponds

also to a ratio

V~/V~

=

/

I.e. if

V~

= 0.16 eV then

V~

= 0.23

eV,

for which value the a

phase begins

to appear in

figure

6b.

4~ Conclusion~

In this paper our calculation of the

paramagnetic susceptibility

x in

light

rare earth metallic

systems

has been based on a

I/Nr expansion.

We obtained a rather

general

and new

analytic

(11)

I'

(a)

= 2.5 eV

f 4QQ

j

bQQ

]

~ =

~

j

n

~

a ?©Q

o.96

ev

Q~

I

,tt

eV

~ 5 IQ 16 2Q 25 JO

V = (b)

= o eV

4

o.96

eV

? 1.92 eV

5 IQ 16

2i

25 JQ

Uyy (eV)

Fig.

5.-

Susceptibility

x'

(in (eV)~'

units same definition of x' as in

Fig.

3) versus Coulomb interaction

Urr(eV

)

(within

one

semi-elliptic

band of half-width

Dj

=

12 eV)

(a)

for

El

= 2.5 eV with V

=

0.96 eV and 1.44 eV (b) for

El

= e~ =

0 eV with V 0.96 eV and 1.92 eV.

(12)

loD

<a> (b)

ao

~ff 3

~'~ ~

JO

< 3.06

~ &o k.o ev

3 16

~

[

z ~Q YCe

0.16

~ ~

u o;

i~ 40 8 ev

o-lo

30 o-m

ace

~~

0.25

-8

lo Wievi

YCe ace

0

o.16 o.2k 0.32 (eV)

Hybddizaticn

Fig.

6.

-(a) Susceptibility

x'

(in (eV)-'

units ; same definition of x' as in

Fig.

3) with respect to

hybridization

for various values of Urr = 4, 8 and

oJ(eV). (b)

Core

absorption

spectra of Ce

(see

Refs. [9, 10]). The

following

model DOS for the CB has been used in

Fig.

6a : one

semi-elliptic

curve of half-width

Di

= 4 eV,

(1/3)rd

filled and with

El

=

-1.2 eV. The valence varies from 3.06

(yce

with V

= 0.16

ev~

to 3.38 (ace with V

=

0.28

ev~,

all parameters except

hybridization being

fixed.

expression

of X, valid for the whole range of

possible

values of the f level

position El,

and for finite

(but large)

or infinite values of the Coulomb

repulsion

Urr. In the case of

typical

Kondo systems, I-e- ni= I for Ce systems, we recovered the standard result of X at zero

temperature.

Within our

development

at the lowest order in V and

I/Urr,

the finite

Urr

result of X for realistic Urr values has been tested

numerically

and shown to be close to the

infinite

U~

result. For the determination of X in the y

- a transition of

cerium, exactly

the

same

parameters

have been used as those

entering

the calculations of the

Lm-XAS

spectra,

demonstrating

once more the

important

role of the

hybridization

effect

(in analogy

to

pressure

effect)

for

describing

the y

- a

phase

transition of Ce. More

precisely, starting

from

yce

with a small

hybridization

and

increasing

the

latter,

you

become,

as it should

finally

in

ace,

less and less

magnetic ~x

smaller and

smaller)

the

important point

here is that our

simple model,

in overall agreement with

experiments,

induces a

change

in valence from y to a cerium which is much smaller than

expected

in older models

(e.g. promotional model).

Let us

finally

mention that the concentration

dependent susceptibility

X in

Ce(Pdj_~Rh~)~

com-

pounds,

measured in reference

[16],

could be

successfully

calculated within a similar theoretical

analysis [17]

as in the

present

paper. In that

example, CePd~

resembles

yce

and

by increasing «alloying

pressure», I.e.

decreasing

average lattice constant and

consequently increasing hybridization,

the

magnitude

of x towards the case of

CeRh~

is reduced

(to

be

compared

to

ace).

(13)

Acknowledglnents.

We would like to thank Prof. F. Gautier for

stimulating

discussions at a first

stage

of this work

as well as Prof. A.

Kotani,

Drs. A.

Meyer,

J.

Besnus,

E.

Beaurepaire

and J. P.

Kappler.

One of the authors

~Y.H.)

is also

grateful

for

obtaining

the «Salam Prize for Scientific

Advancement in

Syria

» as well as a

fellowship

from UNESCO which enabled him to

spend

one month at ICTP in Trieste

during

the summer1990.

Appendix

A.

The characteristic self consistent

equation

for

8E(h)

is written :

~~~~~

~~(h)

+

~~~~~

l~Mk«Mk«

~~~~

with

TMk«M,k,

«, =

Z Z Z ( ("a)

Ml M2

~«j ~"2

ffib

~~«

~M

~~iij

~~i2

~k,

«j ~k~«~ ffi

b) (ffi

b ~~2«2

~~

«j ~M2 ~Mj

~~ji'

~k'

«' ffi

b) 3E(h)

+

s~~ + s~~

s~~ s~~ U~

(A2b) Equation (Al)

is an extension to finite

(but large) U~

of the result obtained in reference

[9]

with

Urr

- co. One of the effects of

Urr

and T

given by equation (A2)

is to renormalize the f level

positions s~)

Similarly

the

k-dependent

coefficients of

equation (2.3)

can be

expressed respectively

as

follows :

~'~~"

3E(h)

+

~~~~~ T~~~ ~~~~

I I

<

~M,

km ffib ~k~«

~~

«,

~~i'

~M

~~M,

~k«

~ b)

~""'~

"'~~

"~

~ ~~

2(8E(h)

+ s~~ + s~~ s~ s~,

Urr)

~~~~

Appendix

B~

Here we use another well~known method to derive our

analytic expression

of the static

magnetic susceptibility

for finite but

large U~.

We start

similarly

from the

appendix

B of reference

[6].

To the usual

impurity

Anderson Hamiltonian

Ho (with

finite

Urr)

such that

Ho 1~)

~

Eol~) (Bl)

we add :

+J

H'

=

hsz

;

Sz

=

£ Majj

a~.

(82)

M= -J

(14)

Equation (Bl)

describes the

coupling

to an extemal

magnetic

field h ;

Sz

is the Z component of the

spin

and

(Eo+AE)

is the

eigenvalue corresponding

to the total Hamiltonian

(Ho

+

H').

In the present method we obtain the static

susceptibility

X as the z

= 0 limit of the

dynamical susceptibility x(z)

defined

by

:

x

(z)

m

Gs(z)

+

Gs(- z) (83)

with

G~(z)

=

~ Sz(z

+

Ho Eo)~ Sz ~ (84)

and then

x = rim x

(z) (85)

Using

the same basis states as in

equation(2.3),

I.e.

[4b), aiiak«(tbb)

and

ajjajj,a~~a~,~, [4~)

for winch the

corresponding eigenvalues

of

Sz

are

0,

M and

(M+ M') respectively,

we are able to cast

G~(z)

as follows :

~ ~

"

ld~(CMk«'~

~~~~~

~

l~

~

Z + hE

El

+ Sk

TMk«Mk«(Z)

<

(M

+

M')~

CMM~

k«k~

«'(

~

~

~~~j

MM'~k'«'Z

+ ~~ ~ ~

~~

~ ~~ ~ ~~'

where

T~~,~~,

has

already

been defined in

equation (A2). Finally using equations (83)

to

(85)

we recover

equation (3.2)

for x. See also reference

[17]

for a similar

analytic

derivation

of x.

References

[Ii

NozitREs P., J. Low-Tenlp. Phys. 17

(1974)

31.

[2] NozitREs P. and BLANDIN A., J.

Phys.

France 41

(1980)

193.

[3] DE CHATEL P. F., Solid State Conlnlun. 41

(1982)

853.

[4] Li T. C., GUNNARSSON O., SCHbNHAMMER K. and ZWICKNAGL G., J.

Phys.

C : Solid State

Phys.

20

(1987)

405.

[5] GUNNARSSON O. and SCHbNHAMMER K.,

Phys.

Rev. 28

(1983)

4315.

[6] GUNNARSSON O. and SCHbNHAMMER K.,

Phys.

Rev. 31

(1985)

4815.

[~ GUNNARSSON O. and SCHbNHAMMER K., in

Theory

of Heavy Fermions and Valence Fluctuations, T. Kasuya and T. Saso Eds.

(Solid-State Sciences)

62

(1985)

l10.

[8] KOTANI A., Jo T. and PARLEBAS J. C., Adv.

Phys.

37

(1988)

37.

[9] HAMMOUD Y., PARLEBAS J. C. and GAUTIER F., J.

Phys.

F: Metal

Phys.

17

(1987)

503.

[10]

HAMMOUD Y., PARLEBAS J. C. and GAUTIER F., J. Magn. Magn. Mater. 63 and 64

(1987)

490.

[I Ii

FULDE P., KELLER J. and ZWICKNAGL G.,

Theory

of

Heavy

Fermion

Systems,

Solid State

Phys.

41

(Academic

Press, New

York)

1988.

[12] HAMMOUD Y., PARLEBAS J. C. and GAUTIER F., J.

Phys.

C Solid State

Phys.

IS

(1985)

6603.

[13]

MACPHERSON M. R., EVERETT G. E., WOHLLEBEN D. and MAPLE M. B.,

Phys.

Rev. Let. 26

(1971)

20.

[14] PARKS R. D., MARTENSSON N. and REIHL B., in Valence Instabilities, P. Wacher Ed.

~North-

Holland,

AInsterdam)

1982, p. 239.

[15] ALLEN J. W. and MARTIN R. M.,

Phys.

Rev. Let. 49

(1982)

l106.

[16] BEAUREPAIRE E., KRILL G., KAPPLER J. P. and ROHLER J., Solid State Conlnlun. 49

(1984)

65.

[17] OKADA M., Master Thesis (1987) Osaka

University.

Références

Documents relatifs

- We present a method for calculating f-electron photoemission spectra of light rare-earth insulating systems within the filled band Anderson impurity model and

and Department of Physical Metallurgy and Science of Materials, University of Birmingham, Edgbaston, Birmingham, England. - Des mesures de I'aimantation de toutes les

The magneto- elastic term AM is the contribution from the higher order terms in (3) and gives the only qualitative diffe- rence in the spinwave expression between

- Paramagnetic spin resonance has been observed in GdZnz, supplemented by static susceptibility mea- surements, over a range of temperatures.. GdZnz has an

opens the possibility for detailed measurements of the crystal field levels and exchange interactions. - Experimental studies of the magne- tic properties of Ce are

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

The aim of this paper is to investigate the unknown magnetic properties of the heavy R E metals in the high temperature solid and liquid state [l] in order to see

Such calculations are not yet available, however recent band structure work on rare earth metallic systems is finally adressing the problem of how to properly treat