• Aucun résultat trouvé

An intrinsic mechanism breaking universality in the conductance fluctuations of realistic mesoscopic disordered metals

N/A
N/A
Protected

Academic year: 2021

Partager "An intrinsic mechanism breaking universality in the conductance fluctuations of realistic mesoscopic disordered metals"

Copied!
9
0
0

Texte intégral

(1)

HAL Id: jpa-00211095

https://hal.archives-ouvertes.fr/jpa-00211095

Submitted on 1 Jan 1989

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.

An intrinsic mechanism breaking universality in the

conductance fluctuations of realistic mesoscopic

disordered metals

M. T. Béal-Monod, G. Forgacs

To cite this version:

(2)

An

intrinsic mechanism

breaking universality

in

the

conductance fluctuations of realistic

mesoscopic

disordered

metals

M. T. Béal-Monod

(1)

and G.

Forgacs

(2)

(1)

Physique

des Solides

(*),

Université de Paris-Sud, 91400

Orsay,

France

(2)

Physics,

Clarkson

University,

Potsdam N.Y.

13376,

U.S.A.

(Reçu

le 6 mars 1989,

accepté

sous

forme définitive

le 2 mai

1989)

Résumé. 2014 Des fluctuations universelles de conductance

(UCF)

de l’ordre de

e2/h,

ont été récemment observées à basse

température

dans des métaux désordonnés

mésoscopiques,

indépendamment

de la taille de l’échantillon et du

degré

de désordre.

L’explication

théorique

qui

suivit

supposait

que les

impuretés,

source du désordre, étaient

indépendantes

les unes des autres.

Nous réexaminons ce

problème

avec

l’hypothèse plus

réaliste que les

impuretés

sont en fait corrélées et d’autant

plus

que leur concentration est

plus

élevée. Nous montrons ainsi que les

UCF sont modifiées par ces interactions : le résultat

e2/h

est

multiplié

par un facteur

numérique

plus grand

ou

plus

petit

que un, suivant que les

impuretés

se repoussent ou s’attirent. Les modifications

dépendent,

en

particulier,

de la concentration en

impuretés

et donc du

degré

de désordre. Par

conséquent

le caractère universal des fluctuations de conductance n’est

plus

vrai dans les

systèmes

réalistes. De nouvelles

expériences

seraient utiles pour tester cette théorie en

comparant des cas où les interactions sont

répulsives

ou attractives. Abstract. 2014 Low

temperature universal conductance fluctuations

(UCF)

of the order of

e2/h

have been

recently

observed in

mesoscopic

disordered metals,

independent

of

sample

size and

degree

of disorder. The theoretical

explanation

which followed assumed that the

impurities,

at the source of the disorder, were

independent

of each other. We reexamine this

problem using

the more realistic

assumption

that the

impurities

are, in fact, correlated and more and more so for

increasing

concentration. We show that the UCF are modified

by

these interactions : the

e2/h

result is

multiplied by

a numerical factor

larger

or smaller than one,

depending

on whether

the

impurities repel

or attract each other. The modifications do

depend,

in

particular,

on the

impurity

concentration and thus on the

degree

of disorder. Therefore the universal character of the conductance fluctuations breaks down in realistic systems. Further

experiments

would be useful to test the present

theory by comparing

cases where the interactions are

repulsive

or

attractive. Classification

Physics

Abstracts 71.55J -

72.15R 72.70

(3)

2710

Introduction.

The recent progress in research on disordered electronic

systems

is marked

by

several

important

milestones

[1] :

the

prediction

and

experimental

observation of deviations from the standard Boltzmann

theory

in metals and the

concept

of Anderson localization of the electrons

yielding

metal-insulator transitions in disordered electronic

systems...

The

latest

discovery

was the existence of universal conductance fluctuations

(UCF)

in

systems

of

«

mesoscopic

» sizes

[2, 3].

Such sizes are intermediate between

microscopic

and

macroscopic

and are relevant to

technological

applications.

It was thus shown that in the

mesoscopic

systems,

new

quantum

mechanical

fluctuations,

as a function of the chemical

potential

or of an

applied

magnetic

field,

show up in the form of time

independent, reproducible, sample

specific, aperiodic

and noiselike structures. The

fluctuations,

of the order of

e2/h,

are the

consequence of

quantum

interference in the

transport

properties

of disordered metals at low

temperature.

They

are

independent

of both the

degree

of disorder and

sample

size

(as

long

as

the inelastic diffusion

length

exceeds the

sample

dimensions)

and,

in this sense, are universal. It was

emphasized

that such fluctuations thus reveal a new and fundamental

aspect

of

quantum transport,

rather than

just

a

simple

finite size effect.

So far in the theories of

mesoscopic

[2, 3]

as well as infinite

[1],

systems

the

randomly

spread impurities

have been considered as

independent.

This may be so

only

in the case of

extreme dilution. But at finite concentration the

impurities

are, in

fact,

correlated and more

and more so as their concentration increases. We indeed

recently

showed

[4]

that even in the

weakly

localized

regime

of infinite disordered

metals,

interactions between

impurities

are

expected

to

play

an

important

role on the Anderson localization of the electrons. Such

interactions

produce,

locally,

an atomic

ordering

which is

accompanied by

oscillating

functions of

(kF d ), d

being

the nearest

neighbor

distance in the lattice and

kF

the Fermi

momentum. This local

order,

in turn, affects the

conductivity

of the infinite

system.

It may induce a metal-insulator transition for a value of the

impurity

concentration for which the

independent impurity

theories

[1]

would lead to a metallic behavior. We showed

[4]

that our

interacting impurity

model can

explain satisfactorily

two different kinds of

experiments.

The purpose of our

present

paper is to reexamine the UCF in

mesoscopic

sysems

taking

into account the interactions between the

impurities.

We use the same

simple

mean field

model to describe the atomic local order induced

by

the

impurity

interactions. Such a model was introduced

long

ago

[5]

to describe the local order between

spins

and

adapted

later on

[6]

to the case of metallic

alloys

with non

magnetic

components ;

in this last case the interactions were of chemical

origin.

Reference

[4]

extensively

recalled the model so that we will not

discuss it further here. Let us

just point

out

that,

already

at the

early

stages

of

references

[5, 6],

the Drude formula for

transport

in disordered metals was shown

[7]

to be modified

by

the

impurity

interactions in such a way that the

(Drude)

conductivity always

decreased when the interactions assumed the screened Coulomb form between excess

charges

on the

impurity

sites. This was recovered in reference

[4]

where,

in

addition,

the

multiply

crossed

conductivity diagram

(the

localization

term)

was also shown to be affected

by

these interactions.

In the next

section,

we

compute

the main contribution to the conductance variance which takes the form of the correlation function of two static conductances related

by

two diffusion

propagators,

shown in

figure

1 ;

reference

[3]

indeed remarked

that,

although

many other

diagrams

are

involved,

the

qualitative

behavior of the final result is the same as that obtained

by only evaluating

the above one. That is

why

we restrict our

study

here to that

diagram only.

(4)

Fig.

1. - A two diffuson

diagram

(the

two shaded

areas)

contributing

to the conductance fluctuations and identical to

figure

5a of reference

[3],

but where interactions between

impurities

are taken into

account as in reference

[4] :

the individual Green’s functions are renormalized

by interacting impurity

scattering

as well as the two diffusons as indicated in

figure

2.

did in reference

[4].

We show that the result of reference

[3]

is

multiplied by

a function of

(kF d )

involving

also the

impurity

concentration and the interaction

strength,

as was the case

in our

conductivity

calculation in reference

[4].

The net result is thus that the universal character of the conductance fluctuations breaks down. We

give

the details of the calculation in three dimensions

(3D)

but the

procedure

is the same in 2D for which case we

only

present

the result. We then end with a discussion and

implications

for other

properties

of

mesoscopic

systems.

Calculation of the two-diffuson conductance fluctuation

diagram.

In this

section,

we evaluate the

diagram

drawn in

figure

1,

that we call K and which

represents,

as recalled above from reference

[3],

a

typical

contribution to the correlation function of two static conductances and thus to the variance of the

conductance,

var

(g ),

where g

is the conductance in units of

e2/h, g

=

G /

(e2/h).

To do so we will use

extensively

the main

ingredients describing

the local order between the

impurities

as derived in reference

[4]

(hereafter

referred to as

BMF),

where the

system

is

considered as an

alloy

of two

types

of atoms A and

B,

the A atoms

being

the

impurities

and the B

being

the host atoms. On the other

hand,

we also follow the

diagrammatic procedure

of

reference

[3] (hereafter

referred to as

LSF).

We confine our discussion to the metallic

regime,

where

kF e >

1,

where

f

is the elastic mean free

path

for

independent

impurities.

As shown in

(5)

2712

VF is the Fermi

velocity ;

T’ is

given

in formula

(22)

of BMF and is recalled here for convenience :

Jo

is the Bessel function of the first kind of index 0

[8].

JI

of index 1 will appear also later in formula

(12) ;

z is the number of nearest

neighbors

in the

lattice,

W =

VAA

+

VBB - 2 VAB,

and the V’s are interaction

energies

between

pairs

of nearest

neighbors

AA,

BB and AB

respectively,

c is the

impurity

concentration.

To

is the

temperature

where the

sample

has been

quenched

from,

and it thus characterizes the

degree

of local order. ro is the elastic

scattering

time in the absence of interaction

(W = 0 )

and

e

=

VF TO.

To evaluate the

diagram

of

figure

1,

we first need to calculate the diffuson T exhibited in

figure

2 and

representing

an infinite ladder of

scattering

of the electron on

interacting

Fig.

2. -

The

diagrammatic Bethe-Salpeter equation

for the diffuson r in the

particle-hole

channel

(formula (5)) :

here too, as in

figure

1, the electron lines should be understood as renormalized with the lifetime T’of formula

(2) ;

the cross is the

scattering potential given by

formula

(3)

involving

the

(6)

impurities.

Each cross denotes a

scattering potential

modified

by

the interactions which has

been calculated in formula

(15)

of BMF as

being :

VA

and

VB

are the

potentials

created

bys

the atoms A and B

respectively, 0

is the

scattering

angle

between momenta pl and p2.

As in

BMF,

we rewrite

(3)

as :

Note that our

present

diffuson vertex T is the

particle-hole analogue

of the

Cooperon

appearing

in the

particle-particle

channel in the

multiply

crossed

conductivity diagram

in BMF. The

Bethe-Salpeter equation

for our r reads :

Then,

as in

BMF,

we linearize

(5)

to first order in

WIT,

and we

get

a structure for the diffuson

analogous

to the one of the

Cooperon

formula

(39)

of BMF :

with the

quantities

A’ and

B’,

different from the

corresponding

A and B ones in BMF and

given by :

Recall that : Z =

[2 7rN (EF)

TO]-l.

After a

lengthy

but

straightforward algebra quite

similar to the one

displayed

in

BMF,

we

(7)

2714

Do

is the bare diffusion coefficient

both

frequency

cases are additive. We may rewrite r as :

One can see that the modifications introduced in r for finite W in the

present

particle-hole

channel

(the

la 3D

factor),

is

exactly

the same as that

occurring

in formula

(49)

of BMF for the

Cooperon

in the

particle-particle

channel. This is not

surprising

since the

frequency

and

momentum

dependences

are

unchanged.

In both cases, the numerical

prefactor

which is involved

depends

on the

sample

characteristics

(c, z, W, kF d, TO).

This last

point

will be

important

for our conclusion.

If,

as was done in the 2nd of reference

[2],

we add to

figure

1 a

similar

diagram

but with 2

Cooperons

instead of 2

diffusons,

the result will thus be the same as

in the

present

case : each

Cooperon

is

multiplied by

(1 + /33D)-

Similarly

in

2D,

,a 2D may be

extracted from BMF and reads :

The

preceding

result will enter into the calculation of K in

figure

1

(identical

to

Fig.

5a in

LSF,

which was said in LSF to be the essential

result).

Now if we go back to the various other

diagrams

considered in

LSF,

with more than 2

diffusons,

it is clear that the

recipe

to calculate them reduces to the

following :

(i)

multiply

each diffuson

(or Cooperon)

by

(1 + 8 ) ;

(ii)

change

the lifetime of the electron from TO to T’ with T’

given

above

by

(2) ;

(iii)

change

the limits in the

final q integral

from with

Actually,

as

long

as L >

1 ’ ,

the

upper’limit (1/f’)

in

the q integral

does not matter in the result as well as

(1/f)

did not matter in LSF when L »

1. Therefore,

provided

that :

only

(i)

and

(ii)

above will affect the LSF result.

In

figure

1

(ii)

will

modify

the

integration

of the various Green’s functions

yielding

a factor

(8)

Similarly,

in

2D,

we

get :

Discussion.

We first

remark,

in connection with

(14)

and

(15),

that K will be enhanced or decreased

by

the

impurity

interactions

depending

on whether W is

positive

or

negative,

i.e. whether the

impurities

repel

each other while

they

are attracted

by

the host atoms, or if

they

attract each other while

being repelled by

the host atoms.

If,

as in reference

[7],

we express

W in terms of the screened Coulomb interactions between excess

charges

on the

impurities

sites,

proportional

to cos

(2 kF

d)l (2

kF

d )3

in 3D and to - sin

(2 kF

d)l (2

kF

d )2

in

2D,

the

change

in K will be an

oscillating

function of

(2 kF d).

Now in order to

complete

the calculation to

get

the variance of the conductance

var

(g ),

it would remain to

perform

the same kind of calculation with

diagrams

with more

than 2 diffusons

(or Cooperons)

with the above

recipe.

We will not do it here but it is

already

clear that var

(g )

will assume the form :

where y will be a function of c, z,

W,

To,

(kF d),

both in 3D and 2D similar to the corrections in formulas

(14)

and

(15).

Therefore,

and this is the crucial

point

of our

present

paper, since

y does

depend

on the various above

parameters,

the conductance fluctuations are not

universal

[9]

in realistic

mesoscopic

disordered

systems,

when the interactions between the

impurities

are taken into account. We believe this could be

usefully

tested

by varying sensibly

the above

parameters,

for instance

by comparing

the results obtained when

dealing

with

impurities

repelling

or, on the

contrary,

attracting

each other.

Aside from the

complete

calculation of var

(g )

sketched

above,

another

interesting

aspect

of this

study

would be to examine how the

impurity

interaction would

modify

the result of reference

[10].

It was shown in reference

[10]

that,

for a

given degree

of weak

disorder,

the conductance was

extremely

sensitive even to the motion of one

single

impurity.

If one

impurity

moved a distance

6r:::. kF 1,

the conductance

changed,

in

3D,

by

5G-(f2 lh) (k2ÎL )- "2

x osc.

(kF 6 r)

where osc.

(kF â r )

denotes an

oscillating

function of

(kF,6r).

In

2D,

the

corresponding expression

is 6G -

(f2/h )(kp P)-nz

x osc.

Thus,

for

2D,

the motion of a

single impurity

leads to

changes

as

big

(if

kF e -

1)

as if the entire

sample

were

changed.

This has been

understood,

on

physical grounds,

since conductance is known

[11]

to

be

proportional

to the

quantum

mechanical transmission

probability

through

the

system,

which itself can be understood in terms of interference between classical

Feynmann paths

through

the

sample

[3].

The

question

which arises in the

present

context is how much ÔG may be

expected

to

change

if several

interacting impurities

are moved ?

Likewise,

how will the

oscillating dependence

in

(kF d )

due to

impurity

interactions combine with the

oscillating dependence

in

(kF,6r)

found in reference

[10] ?

Does the combination of these two

(9)

2716

Acknowledgments.

One of us

(M.

T.

B-M)

enjoyed interesting

comments from B. L. Al’tshuler and S.

Feng.

References

[1]

See the review

by

LEE P. A. and RAMAKRISHNAN T. V., Rev. Mod.

Phys.

57

(1985)

287.

[2]

Theories :

AL’TSHULER B. L., J. E. T. P. Lett. 41

(1985)

648 ;

AL’TSHULER B. L. and KHMEL’NITSKII D. E., J.E.T.P. Lett. 42

(1985)

359 ;

STONE A. D.,

Phys.

Rev. Lett. 54

(1985)

2692 ;

LEE P. A. and STONE A. D.,

Phys.

Rev. Lett. 55

(1985)

1622 ;

IMRY Y.,

Europhys.

Lett. 1

(1986)

249.

Experiments

in Au and Pd

rings :

UMBACH C. P. et al.,

Phys.

Rev. B 30

(1984)

4048 ;

WEBB R. A. et

al. , Phys.

Rev. Lett. 54

(1985)

2696 ;

WASHBURN S. et al.,

Phys.

Rev. B 32

(1985)

4789.

Experiments

in narrow Si inversion

layers :

LICINI J. C. et al.,

Phys.

Rev. Lett. 55

(1985)

2987 ;

SKOCPOL W. J. et al.,

Phys.

Rev. Lett. 56

(1986)

2865 ; 58

(1987)

2347 ;

KAPLAN S. B. and HARSTEIN A.,

Phys.

Rev. Lett. 56

(1986)

2403.

[3]

LEE P. A., STONE A. D. and FUKUYAMA H.,

Phys

Rev B 35

(1987)

1039.

[4]

BÉAL-MONOD M. T. and FORGACS

G., Phys.

Rev. B 37

(1988)

6646.

[5]

DE GENNES P. G. and FRIEDEL J., J.

Phys.

Chem. Sol. 4

(1958)

71.

[6]

BÉAL M. T., J.

Phys.

Chem. Sol. 15

(1960)

72 ; Ph. D. Thesis, Paris

(1963)

unpublished.

[7]

BÉAL M. T. and FRIEDEL J.,

Phys.

Rev. 135

(1964)

A466.

[8]

See for instance in Table of

Integrals,

Series and Products,

by

I. S.

Gradshteyn

and I. W.

Ryzhik

(Academic Press)

1965.

[9]

Actually

the « universal » label should not be taken stricto sensu even in the non

interacting

impurity

case

(we

acknowledge

a remark from one of our referees

reminding

that

point

to

us).

Indeed, even in that case,

only

the central part of the conductance distribution may be described

by

a Gaussian of universal variance but not the tails, as shown in AL’TSHULER B. L.

et al. , J.E. T.P. Lett. 43

(1986)

441 ; J.E. T.P. 64

(1986)

1352. Now, if one took into account the interaction between the

impurities,

as we suggest here, it would

modify

the entire conductance

distribution. In

particular,

the universal character would

disappear

even in the central part.

[10] AL’TSHULER

B. L. and SPIVAK B. Z., J.E.T.P. Lett. 42

(1985)

447 ;

FENG S., LEE P. A. and STONE A. D.,

Phys.

Rev. Lett. 56

(1986),1960

and erratum 56

(1986)

2772.

[11] LANDAUER

R., I.B.M.J. Res. Dev. 1

(1957)

223 ; Localization, Interaction and

Transport

Phenomena in

Impure

Metals, Eds. B. Kramer, G.

Bergman

and Y.

Bruynseraede

(Springer

Références

Documents relatifs

Effect of a bias field on disordered waveguides: Universal scaling of conductance and application to ultracold atoms... application to

Central limit theorem for mesoscopic eigenvalue statistics of deformed Wigner matrices and sample covariance matrices, preprint, arXiv:1909.12821.

Until recently, it was mostly believed (I) that this interesting feature can not be seen in strictly one-dimensional systems since in ID, the localization length, f, is

Judging from the quartic approxi- mation, the characteristic fluctuation frequency may remain of order PI' for intermediate coupling. By comparing the results of the

conductance fluctuation (UCF) of the two-probe conductance in the regime 2.5t 6t in 1D disordered quantum systems in the absence of any inelastic scattering, 1-e-, at zero or a

i) Optical and photoemission studies allow us to assess the positions of electron energy levels away from the Fermi surface and in both pure metals and

Based on these evidentiary support and motivated by the fact that the eigenmatrix of Wishart matrix is Haar (uniformly) distributed, we believe that the eigenmatrix of a

One shows that the features associated with this transition, notably the various scaling behaviors of the height-height correlation functions that have been observed, are