• Aucun résultat trouvé

Oscillator strengths of the optical transitions in a semiconductor superlattice under an electric field

N/A
N/A
Protected

Academic year: 2021

Partager "Oscillator strengths of the optical transitions in a semiconductor superlattice under an electric field"

Copied!
14
0
0

Texte intégral

(1)

HAL Id: jpa-00246501

https://hal.archives-ouvertes.fr/jpa-00246501

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.

Oscillator strengths of the optical transitions in a semiconductor superlattice under an electric field

P. Tronc

To cite this version:

P. Tronc. Oscillator strengths of the optical transitions in a semiconductor superlattice under an electric field. Journal de Physique I, EDP Sciences, 1992, 2 (4), pp.487-499. �10.1051/jp1:1992158�.

�jpa-00246501�

(2)

Classific3tion

Physics

Abstracts

73.00 72.40 78.55

Oscillator strengths of the optical transitions in

a

semiconductor superlattice under

an

electric field

P. Tronc

Laboratoire

d'optique Physique,

Ecole

Supdrieure

de

Physique

et Chimie Industrielles, 10 rue

Vauquelin,

75231 Pads Cedex 05, France

(Received 7

May

1991, revised 21 November 1991, accepted 6December1991)

R4sum4.-Les forces d'oscillateur des transitions

optiques

dans un

superr6seau

semiconducteur soumis h un

champ £lectrique parallble

h la direction de croissance, peuvent %tre calcu16es h l'aide d'un modme de

perturbation

avec des fonctions

enveloppes

de Bloch. Le

champ 61ectrique appliqu£

ainsi que l'interaction dlectron-trou,

qui

induit la formation d'excitons indirects,

entrdment une

asym£trie

entre les forces d'oscillateur des transitions + p et p dons l'dchelle de Wannier-Stark. Certaines

caract6ristiques

des spectres de

photocourant enregistrds

h basse

temp6rature

peuvent %tre

pr6vues

d'une manikre trks

simple.

Abstract. The oscillator strengths of the

optical

transitions in a semiconductor

superlattice

under an electric field

parallel

to the

growth

axis can be calculated

using

a

perturbative

model with Bloch

envelope

functions. The

applied

electric field and the electron-hole interaction

inducing

formation of indirect excitons both induce strength asymmetry between the

oblique

+p and -p transitions of the Wannier-Stark ladder. Features of the photocurrent spectra recorded at low temperature can be accounted for

by

the present model in a very

simple

manner.

1. Introduction.

Due to the lack of resonant

coupling

between

adjacent

quantum Wells when an electric field is

applied along

the

growth

axis of a

superlattice (SL),

the miniband

spectrum

converges towards the

evenly spaced

« Stark ladder »

iii.

In a

type

I

SL,

the

peak corresponding

to the

transition between a hole and an electron with a wavefunction whose maximum is

p

periods

away from the hole state is labelled + p or p

(Fig. I),

the

sign indicating

that its energy is

respectively greater

or smaller than the energy of the

peak corresponding

to an

electron and a hole localized in the same well

(peak

labelled

0).

Photoluminescence

[2, 3]

and

photoconduction [2,4] experiments

show that the +p

transitions are

generally

weaker than their -p

counterparts.

The ratio of the +p to

p transition intensities

depends

on the value of p and on the value of the

applied

electric field F. The

photocurrent

is

proportional

to the

optical absorption

in the structure while the

photoluminescence

spectra are

strongly dependent

on the radiative and non-radiative lifetimes of the carriers. It is

widely

known

[5]

that the matrix element for an

optical

transition is

proportional

to the

overlap

of the electron and hole

envelope

functions.

(3)

/ /

/ /

/ /

' /

,'

,

',

' i /

-2 -1 O+1+2

, , >,

~ , II,

' 'j

,, ,

Fig.

I. Sketch of the conduction and valence band

potential profiles

in a type I SL under an

electric field.

Dignam

et al.

[6]

have fitted the measurements of

photocurrent

made

by Agullo-Rueda

et al.

[4] using

Wannier functions and variational methods but

imposing

at the

beginning

of their calculations an asymmetry of the wavefunction

along

the

growth

direction z in every well.

2. The model.

Before

developing

our model it can

readily

be

shown,

in the

special

case of the functions

proposed by

Bleuse et al.

[I],

that the oscillator

strengths

of the + p and p transitions in a type I SL are

equal

if the electron-hole interaction which induces the formation of indirect

excitons is not considered.

After Bleuse's

model,

the SL

eigenfunction

of a carrier centered on the

q-th

well is :

nq

~

z cn,q

9~(Z

nd)

=

z cn,

q

9~n(Z) (1)

where n is

expanded

over all the N wells of the

SL,

d

being

the

period

and ~

(z )

the

envelope eigenfunction

of the

ground

state of an isolated

quantum

well centered at the

origin.

Vfhen the Nefd total

potential drop

in the SL is

larger

than the miniband A,

C~,~

is

given by

:

C~

~ =

J~

~

(A/2 eFd~ (2)

where

J~(x)

is the Bessel function of

integer

index m, with :

J-

m

(x

=

(-

i )m

Jm (x )

and

~

(

~

Jm (x

)2 =

(3)

In this

theory,

the

approximation

is made that the

~~(z)

functions are

orthogonal.

The

same

approximation

is made if two functions

corresponding respectively

to an electron and a hole located in two different wells are considered.

(4)

The

overlap

of the electron and hole wavefunctions centered at the

origin

is then :

Trill nl)

~

z Cl,

0

Cl, 0(~S ~l) (4)

m

m runs over the wells of the SL

(~$ ~(

does not

depend

on m and is taken to be

equal

to

S.

By translating

the hole

envelope

function

by

± d

along

the z

axis,

one

gets

the

overlap corresponding

to the ± I transition

(mini) =SZCS,oC$,1

~

(5)

(nlln~i) =SZCS,oC$,-i

m

It can

readily

be shown

(Fig. 2)

that C~

~, i =

(- 1)~

+ C

(,

i.

Neglecting

the effect of the limited extension of the SL since the wavefunctions are localized

by

the electric

field,

one

gets

:

(nil n~ j)

=

(trim)) (6)

In the same manner it can be shown that :

(all n~ p)

=

(- IT (all n)) (7)

,'

J i

I i

I j

i

/ ,-, I

f

, i

/

I I

',) electron hole

,

+ d hole

,

d

Fig. 2.- Sketches of the electron wavefunction centered at the

origin

and of the

heavy-hole

wavefunctions translated by ± d

along

the z axis.

In

conclusion,

in the Bleuse's

model,

the + p and p

optical

transitions between free carriers have the same oscillator

strengths

when a

type

I SL is considered. This

result,

which could also be deduced from the

optical absorption

coefficient calculated in reference

[II,

has

to be considered

only

as an

approximation

because the

~~(z)

functions are not

actually

strictly orthogonal. Nevertheless,

it can be noticed that the localization of the carriers induced

by

the electric

field,

which is an

important

feature of the Bleuse's

theory,

has been

unambiguously

verified in all

reported experiments [2-4, 7].

Let us now tum to our more

general theory.

We shall use a

perturbative

model

operating

with the

superlattice envelope

functions of the carriers introduced

by

Bastard

[8].

The

(5)

unperturbed superlattice envelope

functions are Bloch functions

$r~(z),

the allowed values of k

being

determined

by

use of

cyclic boundery

conditions.

Let us consider a SL with an added

perturbative potential Q (z ). Moreover,

if one considers the electron-hole interaction which induces formation of

excitons,

one notices that in a type I SL the system built from an electron centered on the

0~

well and

a hole centered on the

p~

well is the mirror

image

of the system built from the same electron and a hole centered on the

(-p)~

well. The mirror is

parallel

to the

layers

and located at the center of the

ll~

well. The interaction

potential

induced

by

the electron

(hole)

on the hole

(electron) fP~l~~

is

changed

into

f~P~l~~

when one goes from the first

system

to the second.

An

important

step is now to prove the

validity

of

using

a

perturbative

model to take into account the

applied

electric

potential Q(z)

=

efz and the excitons formation.

We have seen that the

applied

electric field F localizes the carriers in the SL. In III-V SLS the

heavy

hole is

strongly

localized due to its

large

effective mass. If N is the number of wells

over which the electron wavefunction is

spread,

N is limited because the

potential drop

over

the whole

spread

of the electron wavefunction cannot exceed the electron first minibandwidth A~

(if

not the

probability

for an electron to tunnel over N wells would be

zero)

:

Nefd = A~

(8)

N is odd from symmetry

arguments.

This value of N can be checked on the Bleuse's functions which

provide

a

good approximation

to the

spread

of the electron

[3]

; table I

displays

the number of wells for these

functions,

with the value of x

=

A~/2

efd which has been used to truncate the

expansion

over the

J~(x)

and the

corresponding

values of

Nefd/A~,

which are all

close to I. The squares of the moduli of these functions are shown in

figure

3 for

N

=

3 and N

=

5. The

voltage drop along

N

wells, being

of the order of A~,

justifies

the use of a

perturbative

model since A~ is

generally widely

smaller than the

heights

of the SL barriers for the electrons and the holes. Moreover the

Rydbergs

of the indirect excitons are small when

compared

to A~

[9] showing

that the

perturbation arising

from the electron hole

interaction is

widely

weaker than the

perturbation

induced

by

the electric field.

The

spread

of the electron wavefunction and hence of the electron-hole system

being

limited to N wells we have to use, to calculate the

perturbed eigenenergies

and the

expansions

of the

perturbed eigenfunctions,

the

corresponding integration

interval over z

(of

extension

Nd~,

the allowed values of k

being

those of a

periodic

array of N wells.

Indeed,

after the electric field has been switched on, the electron wavefunction is the same whether the SL has

only

N wells or more. This

drastically

reduces the number of allowed values of

k to N

(including

the 0

value)

and

drastically

increases the energy difference between the

Table I. Values, versus

N, ofx

used to truncate the

expansion ofthe

Bleuse's

fiznctions

over

the

J~(x)

and

corresponding

values

of Nefd/A~.

N x

=

A~/2

efd

Nefd/A~

3 1.4 1.07

5 2.4 1.04

7 3 1-1?

9 4 1.12

Ii 5.5

13 6 1.08

(6)

(a)

>

« ~

cc

i~

z

(b)

>

j ~

~

~

z

z

Fig.

3.

Square

of the M modulus of the electron wavefunction after Bleuse's model [II a) with N

= 3, b) with N

=

5.

states

corresponding

to two consecutive allowed values of k

(note that,

in the Bleuse's

model,

the

unperturbed

energy

E(k)

is a cosine curve

(Fig. 4)).

The electric field

being

constant, the

envelope

function of a carrier centered on the

n~

well is deduced from the

envelope

function of the same carrier when centered on the

(n p)~

well

by

the

pd

translation.

Obviously

the electric field

splits

the

energies

under the

pd

and +

pd

translations

(Stark ladder).

It is also clear that the

fP~l~~ potential

is invariant under a md translation of the exciton

along

the z

axis,

m

being

an

integer.

The

voltage drop along

N wells

being

of the order of the first electron miniband A~, makes it necessary in a

perturbative model,

to

diagonalize

the electric field

perturbation

over, at

least,

the

subspace corresponding

to the first miniband both for the electrons and the holes.

Higher

minibands will not be considered because their difference in energy is

large

when

compared

to A~, even for the holes. Moreover the contributions to the

overlap

of the electron and hole

envelope

functions is zero for the

envelope

functions from minibands with indexes of

opposite parities

and very weak for indexes of same

parities

but different from

[5].

Let us consider a SL with an

applied

electric field but without

taking

into account the electron-hole interaction. Before

switching

on the electric

field,

the bottom of the first

(7)

'

nld -4nlsd 2nlsd o 2nlsd 4n/5d Rid

k

Fig. 4. Allowed values of k and

corresponding

values of the electron energy at N

=

5 after Bleuse's model.

electron miniband and the top of the

heavy

hole first miniband

correspond

to a

k wavevector

equal

to zero; both

eigenenergies

are not

degenerate.

On the contrary

E

(k )

with k # 0 is twice

degenerate

since E

(k )

= E

(-

k for the SL has even symmetry. The

perturbation arising

from the electric field modifies the

eigenenergies

of the carriers. The allowed values of k are very few and the energy difference between

unperturbed eigenstates

is

large

due to the localization

(see above)

; nevertheless it is not

possible

to assume in every

case that the lowest

perturbed eigenenergy corresponds

to the

perturbed

state

originating

from the

unperturbed

one with k=0

(specially

for the holes since the

heavy-hole

minibandwidth is very small when

compared

to the

potential drop,

which is of the order of the first electron

miniband).

We shall therefore consider in

i)

the situation where the

previous assumption

is valid and in

2)

the situation where the lowest

perturbed eigenenergy corresponds

to a state

originating

from an

unperturbed

one with a wavevector

ko

different from zero at least for the electrons or the holes. In both situations we shall assume that the

radiative recombination takes

place

between the electron and the hole with the lowest

energies.

The effect of the electron-hole interaction

(exciton formation)

which is weak when

compared

to the electric field will be

finally

introduced in both situations.

1)

It is assumed that the

envelope

functions of the electron and the hole which

radiatively

recombine are the Bloch

envelope

functions

xi (z

=

$r(

o

(z )

and

[X((z))

=

$r)~o(z)) perturbed by

the electric

potential Q(z).

These

perturbed

functions are

respectively

written as

(A)~~[c(X(+x[)

and

(~ )~ cl X(

+ X

)) Xi

~~ is

expanded

over the Bloch

envelope

functions

$r((( (z ),

A and ~ are normalization coefficients and

c(l~~

is the

weight

of

x(l~~

in the

perturbed envelope

function.

For

example,

the

overlap

of the

envelope

functions of the electron and the hole centered on the

0~

well is

:

~~~~~~~~

=

~~

)~~-~~~~~~~j~~~~~llxl)1

~~~

with :

lAol~

=

(Clxl+ xllclxl

+

xl)

=

(Cl(~

+

(xllxl)

(~ol~

=

(c(x(+ x)lc(x(+ x))

=

(c((~

+

(x)lx)) (lo)

(xllxl)

=

(xllxl)

=1.

(8)

The matrix element of the electric

potential

is :

~l'$I"

~

(#i(~~~(Z)(Q~~~~(Z)( $i(~~~(Z))

with :

$r(l~~(z)

=

exp(ikz) u(l~~(z)

~~~~

where

u(l~~(z)

is the

periodic

part of the electron

(hole)

Bloch

envelope

function. The

potential

of the barriers

being

even with

respect

to z, one gets :

uif~(z)

=

Ui~~~(- z) (12)

We have seen that

E(k)

is

equal

to

E(- k)

and it can

easily

be shown that :

~((i"

~

(~l'(I' )

* a~d ~- k', k" ~

~k',

k"

therefore

qjl()~

=

qjj>

ar~d

q~jh)

= o

(13)

q_~, ~ = q~, _~ is

purely imaginary

and q~,~ is real.

The coefficient c~ of the

expansion

of

xi(z)

over

$r~,o(z)

is a function of the

q~,~ matrix elements.

At the first

order,

c~ is

[E(0) E(k)]~

qo

~. The second order term is :

iE (o)

E

(k )i-

i

z iE (o)

E

(k') i-

i q~,

~, q~,,o

(14)

~,, o

At the first order c_~ is

equal

to -c~. The second order term is even when

k is

changed

into k. c~ has no defined

parity

and can therefore be written as :

c~ = a~ + b~ with a_~ = a~ k # 0

b~ =

b_

~

a~ is of first

order,

b~ of second order.

Moreover one has :

pj(h)j ~/j,(h))

=

§~,

($~(( $~")

~ ~k

~kk' (Ao(~

=

[c([~+ z (a(+b((~

~'°

~

[~o(~= [c([~+ z (a)+b)[ (15)

k*0

The

overlap

for the ± p transition is calculated

by operating

the ±

pd

translation

along

the

growth

direction on the

$r)= o(z)

function

perturbed by

the

Q(z) potential

:

(x~(x~)

=

(A?~ ~+~)~~ [(c()* c(fro+ (x((x))

*P *P

(16)

(x

x

II

=

Z la(

+

b(

*

[exp (

±

ikpd ) (at

+

b) ) ]

tr~

~~

k#0

j2 ej2~ ~

~e ~

bej2~ h~j2

with:

(A±P

~,~

~ ~

(17) j2 hj2~ ~ ~h

~

bhj2- j~

ll±p

~0 k k °(~

k+0

~k~~-k.

(9)

The numerator of the

overlap

is then :

N

+ p #

(C~)

* C~ ~0 + 2

I

CDS

kpd [(a()

*

a'

+

(b()

*

b'l

~k

k»°

(18)

± 2

z

I sin

kpd (b()

*

at

+

(al)

*

b)I

trk

k>0

One goes from

(x~(X~)

to

(x~(x~) by changing

a, into a~.

P -P

In the above

formulae,

the main contribution comes from the smallest allowed values of

k for the term with k has a contribution

proportional

to

lE~(o) E~(k )j~ jE~(o) E~(k) j~

One

readily

sees that

iApi

=

ii-pi

and

i~pi

=

i~-pi

and it can be

predicted

that the rank of the transitions with the

strongest asymmetry

between

the oscillator

strengths

of the +p and -p transitions are those which are such that

sin

kpd

m I and therefore cos

kpd

« where k has its lowest finite allowed value which is

equal

to 2 ar/Nd. This leads to the criterion :

p = NM

(19)

which will be called «

asymptotic

criterion » for reasons

appearing

when the exciton formation

is considered. This relation

provides

an

integer

value for p on each side of NM

(N

is

odd).

If the formation of excitons is now taken into account, the coefficient of the

expansion

of

xi

(z )

over $r~

, o

(z )

becomes

d~

~k

~ Ck + ~k

e~ is a function of the matrix elements of the electron hole interaction :

g~ejh)_ j~ejh)(~~j

',k" k'

~rpejh)~~~j ~ejh)~~~j

k"

(~Q)

The symmetry of the SL

provides

:

~-i'~~~/"

~

~f'))~~ (21)

We have seen that the electron-hole interaction

perturbation

is weak when

compared

to the electric field one. It is the reason

why

we

keep only

the first order term.

ef

~l~) is then

equal

to

[E(0) E(k)]~ ef[lh)

The formulae are

changed

into :

(X(( X))

~

i [El

~~ +

al

+

b(]

*

[e/P

~ +

exp(± ikpd ) (at

+

b))]

tr~

(22)

~~

k*0

with :

[A±~(~

=

[d([~+ z [e/P~+a(+ b([~

~ ~

~'°

~

(23)

(~±~(

=

[d([

+

z [e/Ph+ exp(±ikpd)(a)+ b))[

k#0

(10)

~±p

~

(d~)* d~

~0 +

I [(~f~)* Ef~

+

(b()* Ef~

~

(a()* Ef~l

"k + k+0

+

z b)[cos kpd(sf~

+

sP[)*

+ I sin

kpd(sf~ eP[)*]

tr~

~"°

(24)

±

z a)[coskpd(sf~- sP[)*

+ I sin

kpd(sf~+ eP[)*]

tr~

k~0

+ 2

~j (cos kpd [(a()

*

at

+

(b()

*

b)]

± I sin

kpd [(b()

*

at

+

(a()

*

b)])

tr~

k>0

and :

'~±p'~

~

(~~(~+ i ((E~~ +~(+ ~((~+ (Ei~~~(+ ~((~)

k~0

[~±~(~= [d([~+ ~j ([eP( +exp(- ikpd)(Ta)+ b))[~ (25)

k>0

+

ei

h + exp

(ikpd ) (± at

+

b)) ~)

Again

the main contribution comes from the smallest allowed value of

(k(.

The criterion for the maximal contribution to the asymmetry is more

complicated

because of the

at

cos

kpd

term and because A and ~ are different for the + p and p transitions.

Nevertheless it can be seen that

e(

is real and that

ei

is a continuous function of

k, making (sf~)*- (sP[)*

to be zero at k= 0. For the smallest allowed values of

(k [,

which

provide

the main contribution to the

overlap, (sf~)* (eP[)*

is therefore small when

compared

to

(ef~)*

+

(eP[)*.

Moreover

fP(z),

which is small when

compared

to

Q(z)

even at p= I, decreases with

increasing

values of p,

making

it clear that

ef

is

always

small when

compared

to a~ which is of first order with respect to the electric

potential

and even very small

except perhaps

for the first values of p

(the

values for which

pd

is

equal

to or smaller than the Bohr radius of the three-dimensional

exciton).

A and ~ are therefore

approximately equal

at the first values of p and very close for

transitions at upper values of p. These considerations show that the

asymptotic

criterion remains still valid with

perhaps

some

discrepancies

for the first values of p.

In a

type

II

SL,

similar conclusions to those drawn in

type

I still hold but

kpd

in the above formulae has to be

changed

into

kp'd,

with

p'= 1/2, 3/2,

2)

It is now assumed that the radiative recombination involves at least one carrier with a wavefunction

originating

from an

unjerturbed

one with a wavevector

ko

different fkom zero.

The

corresponding unperturbed eigenenergy

is twice

degenerate

since

E(ko)

=

E(- ko).

At the lowest order in

perturbation,

the

eigenenergies

are :

l~

(k0)

t

[(~k~

ko)~ +

~k~

ko ~l~~~

(26)

and the ratio c~~/c_~~ is :

1* [(~k~

ko)~ +

~k~

ko ~l~~~

~k~

ko) (~ko, ko)

(27)

The first miniband

originating

fkom an even

single quantum

well

wavefunction,

q~~,~~ is zero

[10].

The lower

perturbed eigenenergy

and the

corresponding

ratio c~~/c_~~ are

respectively equal

to E

(ko) q~~

~~[ and I

(if

one assumes

iq~~

~~ to be

negative).

Even at the lowest

order,

c~ has therefore no defined

parity

versus k

(at

the next order it can be also

(11)

readily

checked that no symmetry property exists versus k between

q~~

~ and

q~~ _~).

As a consequence

asymmetry generally

does exist between the + p and p transitions oscillator

strengths.

In the

special

situation where

only

one

perturbed eigenfunction (electron

or

hole) originates

from an

unperturbed

one with a

ko

wavevector different from zero, the numerator

N±~

of the electron-hole wavefunctions

overlap depends

on p at the lowest order

by

:

2~

~'~[E~(0) E~(ko)]~ ~(q(

~

)* [cos ko pd(I I)

± I sin

ko pd(I

+

I)]

tr~

(28)

This leads to the

following

criterion for the maximal asymmetry

sin

ko pd

=

(29)

which coincides with the

asymptotic

criterion when

ko

is the lowest finite allowed value of k

(which

is necessary the case when N

=

3).

Another

special

situation can be

imagined

where the electron and hole wavefunctions

originate

from

unperturbed

wavefunctions with the same wavevector

ko

different from zero.

At the lowest order N

±~

depends

on p

by

± tr~~ sin

ko pd

if

iq(~,

~~ and

iq(~

~~ have the same

sign

and

by itr~~

cos

ko pd

in the

opposite

case. The first case leads to the criterion

(29).

The second does not present

asymmetry.

This result is inverted if the

unperturbed

wavefunctions have

opposite

wavevectors

ko

and

ko.

If the excitonic interaction is taken into account, it can be seen that

ef~,~ has no defined

parity

versus p and k. The excitonic interaction induces

asymmetry

in any situation. In the

same

special

situations as mentioned

above,

the

(29)

criterion remains

probably approxi- mately

valid since the excitonic interaction is weak when

compared

to the electric field.

3. Discussion.

Table II

displays

the values of sin

kpd

and cos

kpd

versus N for the smallest allowed finite values of

(k(,

in a

type

I

SL,

and the square of the ratio

Ei/E~

of the electron

(hole) eigenenergies

measured from the bottom

(top)

of the miniband and

corresponding

respec-

tively

to the first and second finite allowed values of

(k[

in the Bleuse's

theory.

Are also

displayed

in table II features of the

photocurrent spectra

recorded at 5 K in the

experiments by Agullo-Rueda

et al.

[4]

with the

corresponding

values of the electric field and the

R_~ Rydberg

of the p exciton calculated

[9]

in a SL with

technological

parameters close to those of reference

[4].

The electron minibandwidth is assumed to be

equal

to 65 mev. The values of sin

kpd

which may induce a strong

asymmetry

after the

asymptotic

criterion are underlined. The fit between the

experimental

results and the

asymptotic

criterion is

good, except

for the strong measured

asymmetry

between the +2 and -2 transitions when N

=

5,

which does not

correspond

to a value of sin 2 kd close to I. It may be

imagined (see (2.I))

that the

asymptotic

criterion does not work well at N

=

5 because the difference between the two allowed values of k with the smallest finite modulus

(I.e.

k = 2 ar/5 d and k

= 2 ar/5

d)

is maximal at N

= 5 and

equal

to 4 ar/5 d.

Indeed,

at N

=

3,

the difference between the two

corresponding

allowed values of k

(I,e.

k

=

2 ar/3 d and k

=

2 ar/3

d~

is

only equal

to 2 ar/3 d because of the

cyclic boundary

conditions. Moreover it has to be noticed that

e(j~

=

eP

~j~. On the other hand

(Ei/E~)~,

which,

according

to the Bleuse's

theory,

tends to

1/2i

= 0.0625 for

large

values of N, is maximal and

equal

to 0.15 at N

= 5

(Tab. II).

This,

perhaps, gives

a role to the second finite allowed value of

[k (I.e.

4 ar/5

d~,

a value which

corresponds

to sin 2

~pd

= 0.97

(Tab. II).

It is not

possible,

at this

stage,

to separate the roles

(12)

Table II. Values

of

sin

kpd

and cos

kpd,

versus

N, for

the smallest

finite

allowed values

of

k in a ~ype I SL

and, after

the Bleuse's

model, of

the square

of

the ratio

Ei/E~ of

the electron

eigenenergies corresponding respectively

to the

first

and second

finite

allowed values

of [k features of

the

photocurrent

spectra in the

experiments by Agullo-Rueda

et al. with the

values

of

the F electric

field

and

of

the

R_~ Rydberg of

the p exciton.

kd sin kd sin 2kd sin 3kd sin 4kd sin sin 6kd Features F R_p

(cos kd~ 2kd~ (mev)

2 w/3 0.87 asymmetry 36.I

j =

(0.5) 1/+

2 w/5 0.95 -1/+ 21.7 4

(0.31) asymmetry 2 "

2/+ 2

4 w/5 0.59

2 w/7 0.78 asymmetry 15.5 =

(0.62) 1/+

asymmetry

2 "

2/+ 2

w/9 0.87 asymmetry 12 =

(- 0.5) 1/+

asymmetry 2/+ 2 & 3/+ 3

VIII @ Strong asymmetry 9.85

=

42) (- 0.14) 2/+ 2 & 3/+ 3

w/13 Q69 Strong asymmetry 8.83

3/+ 3 & -4/+ 4

played by

indirect excitons and the electric

potential. Nevertheless,

at N

=

5,

the deviation from the

asymptotic

criterion can

probably

be understood as the

signature

of indirect

excitons.

Another

important

feature appears in the

reported

measurements

[2-4, 7]

: the transitions with the

largest

oscillator

strengths

are the 0 and the -p

(when compared

to the

corresponding

+

p).

No

argument

seems to be drawn from the thermalization of the carriers to

explain

the

photocurrent experiments

results

[2, 4]

since the current is

just proportional

to the

joint density

of states between the electron and

heavy-hole

minibands. On the other

hand,

we can consider that an exciton

originates

from the electric field induced

by

the electron

(hole)

on the hole

(electron).

The mean value of this electric field F~~~

is,

in the case of an

indirect

exciton, parallel

to the

growth

axis. For the p

exciton,

F~~~ and the

applied

electric field F have

opposite

directions

(Fig. 5),

therefore

decreasing

the effect of

F,

whereas

they

have the same direction for the + p exciton. In this

simplified

model

fPhl~)

is taken as

equal

to zero, but the a~ and b~ coefficients are then no

longer equal

for the + p and p excitons since the total F + F~~~ electric field has not the same value ;

they

will be labelled

respectively

a(

and

al'

and

b(

and

bl'.

(13)

F F

Fexc

Fexc

(a)

(b)

Fig. 5. Sketches of the conductions and valence band

potential profiles

for a type I SL under an

electric field F with the electric field F~~~ induced on the hole (electron) by the electron (hole) a) in

a I exciton b) in

a + I exciton.

We have seen that the difference between

a(

and

al'

and

b(

and

bl' originates

from the F~~~ field. One may suppose that

jai al'(

and

b( bl'(

increase with

R~, R~ being

the value

of the

Rydbergs,

that we assume to be

approximately equal [9],

of the +p and p excitons. In the

experiments by Agullo-Rueda

et al.

[4], R~,

which

corresponds

to a

strong

2 versus + 2

asymmetry

when N =

5, 7,

9 or II is

certainly

smaller than the

Rydberg

of the

GaAs bulk exciton

(4.6 mev) [9].

Moreover one would expect that the

asymmetry

decreases when

R~

decreases

(I.e.

for

increasing

values of

N),

which is not true from the measured

photocurrent

spectra

[4].

It is not therefore

possible

to conclude at

large

value of p

~pm3)

that

mainly

the indirect excitons are

responsible

for the oscillator

strength

asymmetry. On the contrary, the strong asymmetry which does exist at

large

values of N can be considered as the

signature

of the electric

potential applied

to the SL. A

complete

conclusion could be drawn

only

from the

computation

of the a~,

b~,

e~ and tr~ coefficients.

Conclusion.

We have calculated the oscillator

strengths

of the various

optical

transitions in a semiconduc- tor

superlattice

under an electric field

using

a

perturbative theory

with the Bloch

envelope

functions. The asymmetry between the p and + p transitions arises fkom both the

applied

electric field and the electron-hole interaction

(indirect excitons).

The rank of the transitions with the strongest asymmetry can be

predicted

versus the

applied

field in a very

simple

manner.

Acknowledgments.

thank G. Bastard, P. Voisin and A. Sibille for very

helpful

discussions.

(14)

References

[II BLEUSE J., BASTARD G. and VOISIN P., Phys. Rev. Lett. 60 (1988) 220.

[2] MENDEz E. E., AGULLO-RUEDA F. and HONG J. M.,

Phys.

Rev. Lett. 60 (1988) 2426.

[3] TRONC P., CABANEL C., PALMIER J. F. and ETIENNE B., Solid State Comman. 75 (1990) 825.

[4] AGULLO-RUEDA F., MENDEz E. E. and HONG J. M.,

Phys.

Rev. B 40 (1989) 1357.

[5] BASTARD G., Wave Mechanics

Applied

to Semiconductor Heterostructures (Les Editions de

Physique,

Les Ulis, France, 1988) p. 246.

[6] DIGNAM M. M, and SIPE J. E.,

Phys.

Rev. Lett. 64 (1990) 1797.

[7] BLEUSE J., VOISIN P., ALLOVON M. and QUILLEC M.,

Appl.

Phys. Lett. 53 (1988) 2632.

[8] BASTARD G., in reference [5] p. 63.

[9] BLUM J. A. and AGULLO-RUEDA F.,

Surf

Sci. 229 (1990) 472.

[10] BASTARD G., in reference [5] p. 18.

Références

Documents relatifs

Two main collision processes will be studied in the present paper : acoustic phonons (deformation poten- tial), and polar optical phonons.. A general solution of these

The virtual photons are absorbed by line transitions of bound electrons of the target plasma, as determined by the bound-bound absorption opacity, which we can write

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

Calculated transitions to conduction band levels n=l,O,and 2 lie at magnetic fields above our experimental range and are therefore not observed.. The data and

Here we find a case where the wavelength is of the order of the thickness a, so that in the Rayleigh-Taylor case we can have a wavelength of the order of lc, iv, or a depending on..

Let us suppose that a sample of n-type nondegenerate semiconductor is subject to an extemal uniform electric field. The general form of the transport equation for electrons will be:..

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

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