• Aucun résultat trouvé

Discussion of the band structure effects on thermoemission as deduced from a simple model

N/A
N/A
Protected

Academic year: 2021

Partager "Discussion of the band structure effects on thermoemission as deduced from a simple model"

Copied!
11
0
0

Texte intégral

(1)

HAL Id: jpa-00208397

https://hal.archives-ouvertes.fr/jpa-00208397

Submitted on 1 Jan 1976

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.

Discussion of the band structure effects on thermoemission as deduced from a simple model

A.K. Bhattacharjee, B. Caroli, D. Saint-James

To cite this version:

A.K. Bhattacharjee, B. Caroli, D. Saint-James. Discussion of the band structure effects on ther- moemission as deduced from a simple model. Journal de Physique, 1976, 37 (2), pp.149-158.

�10.1051/jphys:01976003702014900�. �jpa-00208397�

(2)

149

DISCUSSION OF THE BAND STRUCTURE EFFECTS

ON THERMOEMISSION AS DEDUCED FROM A SIMPLE MODEL

A. K.

BHATTACHARJEE,

B. CAROLI

(*)

and D. SAINT-JAMES

Groupe

de

Physique

des Solides de

l’E.N.S.,

associé au

C.N.R.S.,

Université Paris

VII, 2, place Jussieu,

75221 Paris Cedex

05,

France

(*) Also, Department of Physics, UER Sciences Exactes et

Naturelles, Université de Picardie, 80 Amiens, France.

(Reçu

le 15 mai 1975, révisé le 19

septembre

1975,

accepte

le 10 octobre

1975)

Résumé. 2014 La formulation exacte de l’effet tunnel de Caroli et al. est utilisée pour étudier l’émis- sion thermoionique. Le courant est calculé en présence de potentiel

image

en incluant des correc-

tions quantiques. On montre que l’effet de ces corrections est négligeable. On rend compte des effets de structure de bande à l’aide d’un modèle de Krônig-Penney à une dimension. On montre alors

qu’il n’est pas suffisant de

remplacer m

par m* dans l’expression du courant déduite d’un modèle d’électrons libres. De

plus,

le courant

dépend

fortement de la

position

de la limite entre le métal et le vide. Enfin, on montre que suivant la

position

du niveau du vide par rapport aux bandes inter- dites la loi de Richardson-Laue-Dushman est ou n’est pas vérifiée.

Abstract. 2014 The exact treatment of tunnelling of Caroli et al. is used to study thermoemission.

The current is

computed

in the presence of the image-force

potential

and quantum mechanical correction. The effect of this correction is shown to be

negligible.

Band structure effects are taken

into account

through

a one-dimensional Krônig-Penney model. The current, in this situation, is

not obtained

simply by replacing

m

by

m* in the expression for the current obtained when the metal is described by a free electron gas, even close to band

edges.

Moreover, the current

depends strongly

on the location of the

boundary

between the metal and the vacuum. It is also shown that,

depending

on the location of the vacuum level with respect to the band gaps, the

well-accepted

Richardson- Laue-Dushman law may or may not be valid.

LE JOURNAL DE PHYSIQUE TOME 37, FEVRIER 1976,

Classification

Physics Abstracts

8.340 - 8.940

1. Introduction. - Since the

original

paper

by

Richardson

[1],

the thermoionic emission current has been

expressed

in terms of the Richardson-Laue- Dushman

(R.L.D.) equation [2, 3, 4] :

where T is the

temperature,

T the work function of the

metal,

r the average reflection coefficient of the surface and

Ao

a universal constant which has the value

[3] :

In

using (1),

it is

commonly accepted

that the

reflection coefficient r is

temperature independent

and much less than

unity

if the barrier is

fairly

well

described

by

the

image-force potential.

This assump- tion based on quantum mechanical calculations

by

Nordheim

[5]

and Mac Call

[6],

leads

experimentalists

to use the rather

simple

formula :

where

AR

and QR are constants that are determined

experimentally.

It is

implicitly expected

that

AR

should not be very different from

Ao. However,

the observed

[3, 4]

values of

AR

are

systematically

smaller than the theoretical one,

and,

in some cases,

large

deviations are observed. To try to reconcile

theory

and

experiments,

considerable work has been

done, mainly

in two directions :

i)

The

periodic

nature of the surface

potential

and the value of the

potential

inside and outside the metal

[3]

has been studied in connection with the fact that eq.

(1)

assumes that the electron mass is the same in the metal and in the vacuum,

assumption

which is

only

valid if the

potential gradient parallel

to the surface is zero.

ii)

The effects of the

non-uniformity

of the surface have been

studied,

and

promising

results were obtained

using

the so-called

patch-theory [3]

: the actual

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:01976003702014900

(3)

surface of the metal is viewed as a collection of small

areas

Si,

each of them

contributing

to the thermo- current an

amount ji given by (3);

the total current

is then looked at as the average of the contributions of the different areas. Refinements of this method lead to the more

sophisticated micro-patch theory [7, 8].

In contradistinction with what is

usually

done

when

using

the R.L.D.

law,

we

focus,

in this paper,

our attention on the reflection coefficient

r,

or more

precisely

on the transmission coefficient r defined

as :

The common

assumption, r

= 0 or r =

1,

amounts

to

neglecting

all information about the band structure of the metal.

In section 2 we

give

the

expression

of the thermo-

current j

as a function of the transmission coefficient

T(CO),

where ro is the energy of the

particle,

as mea-

sured from the vacuum level.

Assuming

the metal

to be

represented by

a free electron gas, we show

that j

follows the R.L.D. law

only

if

r(0+) #

0.

In order to

study

the band-structure effects on the

thermo-current,

we consider in section 3 an

exactly

soluble model for the metal : the

Kronig-Penney

model. We show that

taking

into account the energy

dependence

of r

leads,

not

only

to a lower value of the effective Richardson constant

A,

but

also, depend- ing

on the band structure of the

metal,

to a tempe-

rature

dependence

of the thermoionic

current j

which

might

differ from the one

given by

eq.

(1)

or

(3).

We also show

that,

even when the bulk pro-

perties

of the metal are

properly

described

by

the

effective-mass

approximation,

the thermo-current is

not obtained

by simply replacing m by

m* in the free electron metal formula

(1) :

more

complicated

band

effects occur even in that case.

2. The thermoionic current and its

temperature dependence.

- From the

microscopic theory

of

tunnelling developed by

Caroli et al.

[9, 10],

one can

easily

find the

expression

for the

tunnelling

current

(along

the

x-direction)

for a three-dimensional sys- tem,

provided

that one assumes translational inva- riance in the y,

z-directions,

i.e. that the one-electron

potential

is

given by : V(x,

y,

z)

=

V(x).

The

resulting expression

for the current is :

where :

T(CO)

turns out to be the conventional transmission

coefficient, given by :

and

when one assumes that interactions are

negligible

in

the metal. xB defines the

separation plane

between the

left (x xB)

and

right (x

>

xB)

submedia.

is the Fermi

function,

YL

and pR

the chemical

poten-

tials of the isolated

left

and

right submedia, separately

at

equilibrium, WL

and

WR

are the solutions of the

Schrodinger equation :

which behave as

outgoing

waves at x = - oo and

x = + oo

respectively.

Since the Wronskians which appear in eq.

(6)

are constant, xB can be located any- where on the x-axis. It is however most convenient to choose for xB the

boundary

between the two

physical

parts of the

system (e.g.,

for

thermoemission,

the

vacuum-metal

interface).

Eq. (4)

is an exact

expression

for the current in an

out-of-equilibrium

situation

corresponding

to two

electrodes of chemical

potentials

ilL and ,uR. The formulation is

readily

extended to thermoionic emis- sion

by simply remarking

that it may be vizualized as a current

flowing

from a metal

(left submedium)

into

vacuum

(right submedium). Clearly

PL = ,u,

where p

is the Fermi energy of the metal measured from the

vacuum level

(i.e. ,u

= - cp, T

being

the metal work

function), while,uR = -

oo. It is seen

that,

as

expected, only

the electrons

thermally

excited above the vacuum

level contribute to the current.

Indeed,

the transmis- sion coefficient

-r(ro)

is zero for ro

0,

since no

running

wave can propagate, in the vacuum

region,

below the vacuum level.

Note that eq.

(4)

is

completely general

insofar as

it is valid for

arbitrary V(x),

and

that,

when interac- tions are

neglected (i.e. when T(co)

is

given by eq. (6)),

it

gives

the well-known kinetic model of the one

electron

theory [11].

Taking

into account that (p >

kB T,

the thermo- ionic current is

readily

obtained from eq.

(4)

and

(5)

as :

For x --+ oo, the

x-component

of the

outgoing

wave

(4)

151

behaves as

e:tikx u t k(x),

where k is related to the

argument

co of

T(co) by :

Thus it is a function of k ~

w1/2.

This shows that

T(CO) (which

may be

computed

at any

point

x, and

particularly

at x -+

oo)

is a

single

function of

w1/2.

Then :

Eq. (10) clearly

demonstrates that one can

expect

the R. L. D. law

(1)

to be valid

only

if

r(0+ ) # 0,

i. e.

if

T(a))

is discontinuous at the vacuum level.

However,

this necessary condition is

by

no means

sufficient,

since the existence of the R. L. D. law is linked to the

validity

of the power series

expansion

for

t(ro). If,

for

ro >

0, -r(ro)

varies

sharply

in a range of order

kB T,

the

expansion

will be no

longer valid,

and the

tempe-

rature

dependence

of the current will be more involved.

This is for

example

the case when the vacuum level

m = 0 lies near the top of an allowed band of the

periodic potential V(x)

of the

metal,

or in a forbidden

band.

Let us leave for the moment this last case out of our

consideration and concentrate on the behaviour of

T(oi)

close to co = 0.

The

computation

of

T(co) requires

the

knowledge

of

the wave functions

Y’L(X)

and

YR(x) which,

in turn,

are determined as soon as the

potential V(x)

is known.

Since the aim of our calculation is to compare the

commonly

used R. L. D. law

(eq. (3))

with a more

accurate one which includes band

effects,

we will

assume for the

potential V(x)

the

following

form :

where W is the

height

of the conventional step poten- tial

representing

the metal

surface, VL(x)

and

VR(x)

the

potentials

seen

by

electrons

respectively

inside

and outside the metal. xo is the mathematical surface defined

by writing

the

continuity equation

for

V(x).

Note

that,

if we

neglect

the metal and vacuum

potentials ( VL

=

VR

=

0) :

so that

’t(0+)

= 0

[11]

and the thermocurrent reads :

Consequently

there is no reason to

expect

any kind of R.L.D. law in this situation. Moreover it

can be shown from multichannel

scattering theory [ 12J

that

z(0+)

is indeed zero

(i.e. ’t(w)

continuous at

w =

0)

when the

potential V(x)

decreases at

infinity (for x -+ + oo)

faster than

x-lor

is

periodic.

There-

fore the R.L.D. law for

evaporation

may occur

only

in the presence of a Coulomb

potential,

i.e.

. cannot hold for

uncharged particles.

Let us

emphasize

that in most

physical

situations

involving uncharged particles,

a

one-body approxi-

mation cannot be

used,

so that no

simple quantitative prediction

can be made.

In the conventional treatment of

thermoemission,

one retains for

VR(x)

the

following expression [5, 6,13] :

which represents the usual

image-force potential

sup-

plemented by

the quantum mechanical corrections due to

exchange

and correlations

[14].

The

phenome- nological parameter il

is

supposed

to account for the surface

properties

of the metal. The location xo of the mathematical surface is now related

to ’1 by (ao

is the Bohr radius and R the

Rydberg constant) :

Although

it has been shown

[15, 16, 17]

that such a

form for

YR(x)

is incorrect for

light charged particles,

we believe that the detailed

shape

of

V(x)

in the

vacuum

region

will not affect the

qualitative

features

of our

results,

i.e. the

dependence

of the transmission coefficient upon the band structure of the metal.

Moreover,

such a choice for

YR(x)

will allow direct

comparison

with eq.

(3).

Using

the results of references

[5]

and

[6],

one

finds

(1) :

.where :

(1) Wz,,,(z) is the Whittaker function and Hy(z) the Hankel function of order v. The prime denotes derivation with respect

to the argument.

(5)

Therefore,

if we

neglect

the

crystal potentiel (i.e.

VL

=

0), T(O +)

’# 0. This is indeed a consequence of the

x -1

behaviour of

VR(x)

for x -> + oo, i.e.

of the existence of the

image potential.

If

VL(x)

is

such that

t’(w)

is

expandable

in power series of a)’/2 for m >

0,

we then recover the usual R.L.D.

law,

i.e. the

T2 exp(- QlkB T)

temperature

dependence

for the thermocurrent.

Figure

1 shows the value of

r(0+) plotted

as a

function of

R/ W

for the two

limiting

values

of 11

:

11 = 0

and 11

=

x 1 /4,

and for

YL(x)

= 0. We observe

that the quantum mechanical correction

(n = 0)

to the

image

force

potential

introduces no

qualita- tively

new effects. And

since,

for

VL

= 0 and for

actual values of

W(-

10

eV), r(0+)

is

already

close

to

1,

the

commonly accepted assumption, r(0 + )

=

1,

is

justified

when band effects may be

neglected.

FIG. 1. - Plot of 1(0 +) as a function of R/W (eq. (16)) for the two limiting cases : q = 0,ju

= 2 1

and q = x,/4, p

1 (

1 +

4 R W ji12

.

limiting cases: 11 = 0, J.l.

2

and 11 = x J.l.

= 2 1 +

4 W °

However,

we shall see in the

following

section that the band structure of the

crystal

may have

important

consequences on

z(0+),

so

that,

even for

charged particles,

the R.L.D. law may not be valid.

3. Band structure effects : the

Krdnig-Penney

model.

- As is well

known,

band effects in metals are linked to the

periodic

nature of the

.potential V(r).

In order

to

investigate

the influence of this

periodicity

on the

thermocurrent,

we shall solve out a

simple

case,

namely

the

familiar

one dimensional 6-function

Kronig-Penney

model

[18] (K.P. model).

For the sake of

simplicity,

we assume that :

In this

expression

P is a dimensionless

parameter representing

the

strength

of the 6

potential, a

is the

lattice

spacing along

the x

direction,

and b is a para- meter which we allow to vary from - a to zero.

This means that we describe the surface

by

a step

potential

the location of

which,

with

respect

to the

last 6

potential,

may vary within a unit cell.

To be sure, eq.

(18)

does not represent satisfac-

torily

an actual metal in the presence of its

surface,

but it has the

advantage

of

leading

to a

completely

soluble situation and to show that the thermocurrent is not

only

modified

by

the effect of

periodicity,

but also

depends

rather

strongly

on the

position

of

the surface

(i.e.

on

b).

In the metal

region (x xo)

the relevant electron

wave function

VFL(X) compatible

with the

boundary

condition at x = - oo is

readily

calculated and one

finds from eq.

(6) :

where K is defined

by

eq.

( 12b),

and use has been made of the well-known

dispersion -

relation :

Note that the allowed energy bands

correspond

to

those values of K for which the left-hand side of eq.

(21)

lies between - 1 and + 1. It is also seen

that surface states

[19] (Tamm’s states)

which appear in the gaps

only

for co 0

(corresponding

to ima-

ginary

values of

K)

do not contribute to the thermo- current.

The

complete computation

of

T(w) and j requires

also the

knowledge

of yR. In order to

disentangle

somewhat the effects of surface

location, periodicity

and Coulomb

potential,

we shall

investigate

the

two situations of no

image

force

(i.e. VR

= 0 for

x >

xo)

and of a vacuum

potential

as

given by eq. (14).

3.1 TRANSMISSION COEFFICIENT IN THE CASE OF NO IMAGE POTENTIAL. - We set

VR(x)

= 0 for x > 0

(i.e.

xo =

0).

The transmission coefficient is

readily

obtained,

and reads :

(6)

153

while the

density

of states

[20],

for one

spin direction,

is

given by :

Note that there is no

simple

relation between the

density

of states and the transmission

coefficient, and, consequently,

between the

density

of states and

the

thermocurrent,

in contradistinction to what is sometimes assumed.

3.1.1

Effect of surface

location. - It is

interesting

to remark that

T(w) depends strongly

on the surface

position

parameter b. As an

example

we have

computed T((o)

as a function of co for P = 3

n/2, a

= 4

A

and W =

13,5

eV. The results are

displayed

on

figure

2. The two curves

correspond

to the two

different values of b : -

a/2,

0. The effect of the

surface

parameter b is seen to be

quite

drastic.

(Our

choice of the parameters a and W is such that the bottom of the third band lies 0.4 eV below the vacuum

level.)

FIG. 2. - Plot of T(co) as a function of (o (eq. (22)) for different

values of b : W = 13.5 eV (P = 3 Tc/2, a = 4

A).

3.1.2 The

effect of

band structure : the

effective

mass

approximation.

- The effect of the

periodicity

appears in eq.

(22) through periodic

functions of Ka and pa. As well

known,

close to a band

edge,

the

dispersion

relation is

quadratic,

so that one can

introduce an effective mass m*. In the K.P. model the band

edges

are

given by :

the

corresponding energies

cv°

being given by

eq.

(21).

In the

vicinity

of such a band

edge

we have :

where

is the effective mass.

At first

sight

one can be

tempted

to take

advantage

of the

quadratic

relation

(25)

in

computing T((D), by describing,

close to a band

edge,

an electron

inside the metal as a free

particle of

mass

m*, moving

in a constant

potential wo,

the outside

potential being

zero. The

resulting Le.m.(ro)

reads :

This

commonly

used

approach

is however incorrect.

Indeed,

close to the band

edge (dp - 0) 1(cv),

as

given by

eq.

(22),

reads :

It is seen that

T(W)

vanishes at the band

edge.

This

is to be

expected since,

for this energy, the group

velocity

is zero, so that no

particle

can

propagate

in the metal. However,

i(cv)

differs from its

effective

mass value

Te.m.(W) by

a

multiplicative

factor which

depends explicitly

on the characteristics of the

periodic potential. Figure

3 shows the variation of

T((,O)/Te..,.(co)

as a function of b. For

instance,

with

(7)

FIG. 3. - Plot of T(60)/T’...((O) (eq. (27) and (28)) as a function of the

parameter b for W = 13 eV and coo = 0.09 eV (P = 3 7r/2, a = 4

A).

W = 13

eV, P=37r/2, a=4A

and

b = 0,

this

factor is 0.14 for

roO

= 0.09 eV. Note that

depends strongly

on

b,

the

surface

parameter, in

agreement

with what was

already

remarked when

discussing

eq.

(22).

However,

it must be

kept

in mind that the

descrip-

tion of the surface

by

a

step potential

becomes ques- tionable close to the first atomic

layer.

In other

words,

it may be somewhat

misleading

to draw

quantitative predictions

from the above

calculation,

since the variation of b takes

place

in a

region

where

the

potential description

is

questionable.

The reader

should conclude from our results that the transmis- sion coefficient may be

quite

sensitive to the surface

description

and that it is rather

involved

to obtain

information on the surface

by simply studying T((O).

3.2 TRANSMISSION COEFFICIENT IN THE PRESENCE OF AN IMAGE POTENTIAL AND VALIDITY OF THE R.L.D. LAW.

- We now assume that the

potential

outside the

metal is

given by

eq.

(14),

so that yR is

given by

eq.

(16a)

and

(16b). r(co)

is

easily computed

from eq.

(16), (19)

and

(6a).

It is clear

that,

as in section

3.1, T(ro)

vanishes at

a band

edge

and remains zero inside a band gap.

This

immediately

shows that one cannot expect

i(cv)

to be discontinuous at co = 0 if the vacuum

level lies in a forbidden band. We shall therefore

investigate

two

possibilities :

a)

the vacuum level falls into an allowed band.

b)

the vacuum level falls into a forbidden band.

a)

The vacuum level

falls

into a allowed band. - The situation is

pictured

on

figure

4.

i(cv)

is now

discontinuous at co = 0.

However,

as discussed in section

2,

the R.L.D. law is obtained

only

if

i(cv)

varies

slowly

on a range of order

kB

T. Since

Aco)

will vanish at the

top

of the band ro1, we can distin-

guish

between two cases

(2) :

i) co, >> kB

T

(Fig. 4a).

The variation of i takes

place

on an energy range much

larger

than

kB T,

we can then

replace

in eq.

(8) 1(cv) by T(O ’),

and

obtain :

FIG. 4. - The metal band structure is such that the vacuum level lies within an allowed energy band : a) kB T w, ; b) kB T > col.

Therefore,

as

expected,

the thermocurrent follows the R.L.D. law with a Richardson’s constant which

depends

on the band structure and on the

position

of the surface of the metal

through

the transmission

(’) We assume from now on that the width Jf of the forbidden band above wl is much larger than kB T.

(8)

155

FIG. 5. - Plot of i(0+) (K.P. model) as a function of b for

W = 13.5 eV, a = 4 A, P = 3 7r/2,?l = 0.

coefficient

1(0+).

We have

plotted

on

figure

5 the

variation of

1(0 +)

as a function of b for W = 13.5 eV

(P

= 3

n/2, a

= 4

A).

Note

that,

here

again,

the

effect of the surface

parameter

b on the thermocurrent is very

significant.

ii) kB

T >> w1

(Fig.

4.

b).

The variation of z takes

place

on an energy range much smaller than

kB T,

so that a power

expansion

is no

longer

valid. Therefore

one does not

expect

an R.L.D. law for the current.

Moreover,

since the thermal energy

kB

T is small

compared

with the work

function,

the current will

be small.

Anyhow

it is

interesting

to

study

this

situation,

insofar as it shows how band effects can lead to drastic

changes

in the thermocurrent temperature

dependence.

Since w1 is close to the vacuum

level,

we can assume

that it is a

good approximation

to use for yR its

ro = 0+ value

(eq. {16b)).

Near the top of a band eq.

(25)

reads

[20] :

Using

eq.

(19)

and

developing

eq.

(6)

to first order in

bp,

we obtain :

where :

(n

=

1, 2,

... is the band

index).

Since

we can, to first

order, replace exp( - wlkb T) by

1 in the

integrand

of eq.

(8).

The thermocurrent takes the form :

We have written eq.

(33)

under the conventional form. It is seen that the R.L.D. law is not

obeyed,

since the Richardson constant

C(.)(T) depends

on

temperature

(in

this

case j

is even

proportional

to T

instead of

T2.!)

Note

that, similarly

to the case

VR

=

0,

the effect of the

periodic potential

does not appear

only through

the conventional effective mass correction

(mlm*)’I’

but,

in a rather involved way, via T(.). Note also that the value of ’t’(n)’ hence

C(n.)(T), depends

on the

index n of the allowed band considered.

Figure

6

shows the variation of ’t’(n) with b for

FIG. 6. - Plot of t(n) (eq. (32)) for n = 2 as a function of the surface

parameter b whence 1 kB T. Here : W = 9.39 eV, pa = Kl a = 27r

(9)

FIG. 7. - The metal band structure is such that the vacuum level lies within an energy gap. The thermal energy is such that coo kB T wo + A. Hatched areas correspond to allowed

energy bands.

b)

The vacuum level lies within an energy gap. - The situation is

displayed

on

figure

7. Let us call

again

roo the energy of the bottom of the allowed band

lying just

above the vacuum

level,

and A its

width. Then from eq.

(8) :

Eq. (34) pictures

the fact that

r((o)

is zero for a) coo and co > coo + A.

For kB

T coo the current prac-

tically

vanishes as the

exponential

in the

integral

will be very small in the

region

of

integration.

For

kB

T > coo +

A,

one would have to take into account the

explicit dependence

of

-r( OJ)

on co. Let us how-

ever note that this situation is not realistic

since, usually,

the bandwidth is much

larger

than

kB

T.

Therefore,

the

physically interesting

case corres-

ponds

to coo

kB T

d . In this case the thermo- current will be

governed by

the behaviour of

i(cv)

near the bottom of the band. In this energy range yR is

given by eq. (16.a)

and the

co-dependence

of

T(W)

is

governed by

the behaviour

of yL (eq. (19))

near the

bottom of the band. To first order in

bp (eq. (25))

the transmission coefficient reads :

where coo is the energy of the band

bottom, n

=

1, 2, ...

the band index and :

Thus the thermocurrent is

given by :

This

expression

for the thermocurrent is valid

only

when the band structure can be described

by

the effective mass

approximation (3),

i.e. for cv z wo.

The value D of the upper limit of

integration

is

roughly

the energy for which the effective mass

approximation

fails. In thermoionic

emission,

this

energy can be assumed to be

larger

than the thermal energy

kB T,

hence the value of the thermocurrent

j

is

roughly independent

of the exact value of D and

one can write :

(3)

Let us once again emphasize the fact that s(m) is not computed

in an effective mass approximation.

Thus,

one obtains

(cvo « T) :

This result is not

surprising for,

as

long

as we are

interested

only

in the

temperature dependence of j,

we are in a situation very similar to the one described

at the end of section

2, namely

free electron model and a constant

potential

outside the metal.

Indeed,

as we

pointed

out, the

temperature dependence

of

the thermocurrent is

strongly governed by

the beha-

viour of

i(cv)

in the

low-energy region :

this

explains why

the T

dependence of j

is the same in eq.

(39)

as in eq.

(13);

in the first case

I(cvo)

= 0 while in the second we have

-r(0)

= 0.

Moreover,

the

exponential

factor in eq.

(39)

is of the same nature as in eq.

(13)

since

only

electrons with energy w > a)o + (p ~ Q

can escape from the metal.

The effective mass and

periodic potential

effects

are described

through (mlm*)1/2. ten)’

where

ten)

is

given by

eq.

(37).

We have

plotted ten)

as a function

of b in

figure 8,

for W = 13 eV

(P

= 3

a/2, a

= 4

A).

4. Conclusion. - We find that the Richardson law is not

universally

valid for thermoemission. The

(10)

157

FIG. 8. - Plot of t(., (eq. (36)) for n = 3 as a function of the surface parameter b. Here W = 13 eV, po a = 2, Ko a = 2.36 7r so that

mo = 0.09 eV and (m*lm) 1/2 = 0.31 ; tj = 0 (P = 3 n/2, a = 4

A).

thermocurrent

depends quite strongly

on the surface barrier as well as the

periodic potential

inside the emitter.

While the presence of the Coulombian

image

force

outside the emitter is necessary

(but

not

sufficient)

for the Richardson law to be

valid,

the first-order quantum mechanical correction

[14]

to the

image

force does not introduce any

qualitatively

new

effect on the thermocurrent. When the vacuum

level at

infinity (cv

=

0)

falls inside an allowed energy

band,

the top of which lies much

higher

than

kB T,

we find the

following temperature dependence

of

the current :

The latter remains valid if the vacuum

potential

vanishes faster than

x-1

at

infinity.

This result for the thermoemission of neutral

particles should,

how-

ever, be

interpreted

with care. For

instance,

in the

case of

evaporation

of excitons from electron-hole

drops

in

semiconductors,

the

many-body

effects

should

presumably

be non

negligible.

In the presence of the

image force,

the effects of location of the

surface,

as well as of the

periodic potential

inside the

crystal

have been

investigated

in the frame of the

Kronig-Penney

model. The main results of this

study

are enumerated below :

1)

The thermocurrent is

extremely

sensitive to the

location of the surface with respect to the last

spike

of the K.P. array, and

depends explicitly

on the lattice

spacing

for a

given potential strength.

2)

Even near a band

edge,

the current cannot be obtained

by simply replacing m by

m* in the free

electron

expression,

and the temperature

dependence

of the current is sensitive to the

position

of the vacuum

level

(w

=

0)

with respect to the band structure of the emitter. For instance :

a)

When ro = 0 lies inside an allowed

band,

the top of which is at w 1

b)

When co = 0 lies in the gap at a distance -

kB

T

below the bottom of an allowed band

Even

though

the

limiting

cases

(a-ii)

and

(b)

corres-

pond

to small

thermocurrent, they

illustrate

signi-

ficant

departures

from the Richardson law.

It should be remarked that our formulation of the emission

problem

is

effectively one-dimensional,

since

we have made the. conventional

assumption

of

translational invariance in

planes parallel

to the

surface. The fact that the thermocurrent

depends explicitly

on the lattice

spacing

and the band structure

along

the

perpendicular

to the surface

implies

a

very

interesting

result : the

anisotropy

of thermo-

emission from

crystals.

The

problem

of

description

of the surface is illus- trated

by

our K.P. model result that the thermocurrent is very sensitive to the location of the surface

step.

The

representation

of the surface

by

a step

potential (plus

the

image

force

outside)

is

certainly questio-

nable when one is interested in effects on distances of the order of the lattice

spacing. Moreover,

we

cannot rule out the

possibility

that the K.P. model

be

pathological

in relation to the location of the

step.

In any case, whatever be the

model,

it should be

noted that the transmission coefficient

depends

on

the

matching

condition at the

boundary,

and there-

fore the effects of the location

of

the

surface

are

expected

to be

significant.

The

physical

state of the

surface is

expected

to be of consequence, even

though

the localized surface states do not contribute to the thermocurrent.

(11)

References

[1] RICHARDSON, O. W., Phil. Mag. 23 (1912) 594.

[2] DUSHMAN, S., Rev. Mod. Phys. 2 (1930) 381.

[3] HERRING C. and NICHOLS, M. H., Rev. Mod. Phys. 21 (1949)

185.

[4] NOTTINGHAM, W. B., Hand Phys. Band 21 (1956) 1.

[5] NORDHEIM, L. W., Proc. R. Soc. 121 (1928) 626.

[6] MAC CALL, L. A., Phys. Rev. 56 (1939) 699.

[7] GARDNER, F. M., GIROUARD, F. E., BOECK, W. L. and COOMES,

E. A., Surf. Sci. 26 (1971) 605.

[8] NEEHAM, P. L. and COOMES, E. A., Surf. Sci. 44 (1974) 401.

[9] CAROLI, C., COMBESCOT, R., NOZIÈRES, P. and SAINT-JAMES, D., J. Phys. C 4 (1971) 916.

[10] CAROLI, C., COMBESCOT, R., LEDERER, D., NOZIÈRES, P. and SAINT-JAMES, D., J. Phys. C 4 (1971) 2598.

[11] DUKE, C. B., Tunnelling in Solids, Solid State Phys. Suppl. 10,

H. Ehrenreich, F. Seitz & D. Turnbull eds. (Acad. Press, N. Y.) 1969.

[12] NEWTON, R. G., Scattering Theory of Waves and Particles, Chap. 16, 17, (Mc Graw Hill ed., N.Y.) 1966.

[13] CUTLER, P. H. and GIBBONS, J. J., Phys. Rev. 111 (1958) 894.

[14] SACHS, R. G. and DEXTER, D. L., J. Appl. Phys. 21 (1950) 1304.

[15] APPELBAUM, J. A. and HAMAN, D. R., Phys. Rev. B 6 (1972)

1122.

[16] FEIBELMANN, P. J., DUKE, C. B. and BAGCHI, A., Phys. Rev. B 5 (1972) 2436.

[17] HEINRICHS, J., Phys. Rev. B 8 (1973) 1346.

[18] KRÖNIG, R. de L. and PENNEY, W. G., Proc. R. Soc. 130 (1931)

499.

[19] TAMM, I., Phys. Z. Samj. 1 (1932) 733.

[20] DUKE, C. B. and FAUCHIER, J., Surf. Sci. 32 (1972) 175. The

same K. P. model is used there to study the band structure

effects on the field emitted current at T = 0 K. In the range of energy concerned in field emission, the shape of the potential in the vacuum is mainly governed by its field-

dependent part 2014 eEx, so that one can neglect the contri- bution of the image force potential (See also B. Caroli : Surf. Sci., 51 (1975) 237). Thus 03C4(03C9) as given by eq. (22)

is the E ~ 0 limit of Duke and Fauchier’s result.

[21] We recall that, at the top of a band (03C90 = 03C91) one has : K1 a = pa = n03C0, n = 0, 1, 2, ... is the band index and K is given by eq. (12b).

Références

Documents relatifs

Higher order data correspond to a mixture of terms involving higher order correlation functions, higher order potentials, and may involve quantum corrections before they are

En este sentido, el artículo plantea que a partir de estas prácticas en la escena posdramática, el texto teatral pierde jerarquía, pero a la vez se constituye como un

In this study, the marine biota dose rate assessment showed that the predicted dose rates to marine biota attributable to radioactive sea discharges from the La Hague facility

"201, The radiation exposure of transport workers and the general public is subject to the requirements specified in the "Basic Safety Standards for Radiation Protection:

Degeneration of effective diffusion in the presence of periodic potential.. Annales

After incubation, each, well in the microtitre plate (with 25 μl o f crystal violet and C. The experiment was repeated. Controls were parasites incubated without crystal violet. i

A second feature that deserves attention is that as the interest rate rises, the level of technology that is adopted tends to converge to the same value under both competition

l’utilisation d’un remède autre que le médicament, le mélange de miel et citron était le remède le plus utilisé, ce remède était efficace dans 75% des cas, le