• Aucun résultat trouvé

Inelastic scattering of neutral particles by a solid surface

N/A
N/A
Protected

Academic year: 2021

Partager "Inelastic scattering of neutral particles by a solid surface"

Copied!
29
0
0

Texte intégral

(1)

HAL Id: jpa-00211011

https://hal.archives-ouvertes.fr/jpa-00211011

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.

Inelastic scattering of neutral particles by a solid surface

G. Armand

To cite this version:

G. Armand. Inelastic scattering of neutral particles by a solid surface. Journal de Physique, 1989, 50

(12), pp.1493-1520. �10.1051/jphys:0198900500120149300�. �jpa-00211011�

(2)

Inelastic scattering of neutral particles by a solid surface

G. Armand

Service de Physique des Atomes et des Surfaces, Centre d’Etudes Nucléaires de Saclay,

91191 Gif-sur-Yvette Cedex, France

(Reçu le 12 janvier 1989, accepté le 13 février 1989)

Résumé.

2014

La diffusion inélastique de particules couplées à

un

champ de phonon par

un

potentiel d’interaction est analysée

en

détails. Le développement

en

série de perturbation de l’équation de matrice T permet de décomposer la probabilité de transition entre état initial et état final

en

composants élémentaires. Ces termes décrivent les processus d’échange de phonons réels

et virtuels. Ils sont

en

étroite relation

avec

les éléments de matrice du processus élastique :

connaissant leurs expressions analytiques

ou

les diagrammes les représentant, l’on peut déduire les expressions

ou

diagrammes des composants inélastiques correspondant. Les relations d’unitarité valables à chaque ordre de perturbation sont établies. La probabilité de transition

d’échange d’un phonon réel est donnée par

une somme

de termes élémentaires décrivant

l’échange d’un phonon réel et l’échange de 0, 1,

..., n ...

phonons virtuels. Ces termes sont

proportionnels respectivement à T, T2,

...,

Tn + 1 ... à

une

température T du cristal suffisamment élevée. La probabilité d’échange de deux phonons réels peut être écrite de la même façon, les

termes étant alors proportionnels à T2, T3, ..., Tn+2. L’état final peut être soit

un

état du continu

ou un

état lié du potentiel. Dans

ce

dernier

cas

la probabilité de transition donne le coefficient de capture. Les probabilités d’échange d’un phonon réel comprenant tous les termes proportionnels

à T et T2 et celles d’échange de deux phonons réels comprenant tous les termes

en

T2 ont été calculées

en

représentant le potentiel d’interaction par

un

potentiel unidimensionnel.

La probabilite de transition d’un phonon réel présente

un

maximum lorsque la température du

cristal varie. Sa position dépend de la nature de la particule incidente et des conditions d’incidence. Le coefficient de capture croît

avec

la température du cristal. Le domaine de validité de l’approximation dite DWBA est déterminée.

Abstract. 2014 The inelastic scattering of particles coupled to

a

phonon field by

an

interaction

potential is fully analysed. The expansion of the T matrix equation allows

one

to decompose the

transition probability between

an

initial and

a

final state into elementary components. These components exhibit real and virtual phonon processes. They

are

closely linked to the elastic matrix elements : from their analytical expression

or

diagrammatic representation

we can

deduce

the corresponding expression

or

diagram of the inelastic transition probability components.

Unitarity relations to each order of expansion

are

demonstrated. The probability for the single

real phonon exchange is composed of

a sum

of components describing with the exchange of the

real phonon the exchange of 0, 1, ..., n

...

virtual phonons. These terms

are

proportional to T,

T2,

...,

Tn

+

1 ... respectively at sufficiently high crystal temperature T. The probability for two

real phonon exchange

can

be decomposed in the

same

way in terms proportional to T2, T3, ..., Tn+2

...

The final state

can

be either

a

continuum

or a

bound state. In this last

case

the transition probability, gives the sticking coefficient. The transition probabilities for the

one

real

Classification

Physics Abstracts

03.80

-

61.14D - 68.30

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:0198900500120149300

(3)

phonon exchange including all the terms proportional to T and T2 and for the two real phonons

process including all the T2 terms have been calculated using

a one

dimensional potential model.

The single phonon transition probability exhibit

a

maximum

as

the crystal temperature varies. Its

position depends

on

the nature of the scattered particles and

on

the incident conditions. The

sticking coefficient increases with the crystal temperature. The domain of validity of the DWBA is determined.

1. Introduction.

Among the different methods which allow calculation of the inelastic scattering of an incident particle coupled to a phonon field by an interaction potential, the T matrix formalism is

certainly one of the most fruitful.

To the author’s knowledge the first work in the field using this formalism is that of Manson

ans Celli [1]. The authors neglect the processes off of the energy shell and calculate the one-

phonon cross section. Since this work, calculations have in general been limited to the

simplest term given by considering the first order in the expansion of the T matrix equation,

which is called distorted wave Born approximation. The papers published are numerous and the list given here is certainly not exhaustive [2-8].

The state of art is most advanced as far as elastic scattering is concerned. From the work of Levi and Suhl [9], who demonstrate that the elastic intensity is given by the square of the

thermally averaged elastic matrix element, an exact theory of this process has been developed

which includes one [10], two [11] and multiphonon virtual processes [12].

Using the technique elaborated for this study we present in this paper a detailed analysis of

the inelastic processes where the particle exchanges energy via the transfer of real phonons.

However we show that this exchange can be achieved with the simultaneous exchange of

virtual phonons.

Considering the possibility of simultaneous real and virtual phonon processes one feels that the elastic and inelastic processes are closely linked to each other as it is the case on a physical

intuition. Effectively we show that, knowing the elastic matrix element, one can deduce

inelastic transition probabilities and one can demonstrate some useful

«

unitarity

»

relations.

The formalism being well established we write the transition probability component proportional to T and T2 (T is the crystal temperature) for scattering by a simple potential

model. Then the reflection coefficient to a continuum state and the sticking coefficient are

calculated for typical cases.

2. Général theory.

Let us consider a particle scattered by a surface where the interaction potential is thermally displaced due to the thermal agitation of the crystal atoms. This thermal agitation is taken into account through the displacement operator u and the potential can be written as

V (R, z, u ) where R and

z

are the particle coordinate parallel and normal to the surface, respectively. The static potential obtained by performing the thermal average of V over the

crystal phonon field and denoted by V, is, generally speaking, periodic. Hence the particle

is scattered elastically in diffraction channels or inelastically in numerous open channels

corresponding to the exchange of real phonons.

The T matrix equation which describes the scattering process completely can be written as :

with

(4)

where Ef and Eic are the initial energy of the particle and the crystal, respectively, He is the crystal Hamiltonian and

In Ho m is the mass of the incident particle and U(z) the distorted potential chosen for

instance equal to the zero order Fourier component of «V ) .

In order to get a solution of this equation it is convenient to decompose the quantity V (R, z, u ) - U(z ) into two parts :

with

Then following a procedure presented in many textbooks [13] and used in a previous paper to

study the elastic scattering by a surface exhibiting adsorbed defects [14Î, one gets :

with

and :

h is the matrix which describes the elastic scattering by the periodic surface. The h matrix

elements have non zero value on the energy shell, only for the open diffraction channels

corresponding to the initial conditions.

t is the matrix which describes the inelastic scattering as a perturbation of the elastic diffraction process or the scattering by the potential v (R, z, u ) + U(z ).

The transition rate fw; between incident (i) and final (/) particle states is given by :

where Ey and E,P are the final and initial particle energy, respectively, and fT, is the T-matrix element calculated between the final and initial particle states.

Replacing the T-matrix elements by their expression (3) the thermal average factor in (8)

can be written as :

The term «fLi(,k» does not contain correlation functions of the u operator for two different k values and the integration over À with the exponential factor yields a 8 function.

One gets :

(5)

The first and second terms in this expression correspond obviously to elastic and inelastic

scattering respectively.

In order to simplify the subsequent analysis we consider the simpler case of a flat surface or

quasi-flat surface as the dense faces of CFC crystal for which the diffracted peak intensities

are very small. In this case the diffraction process is unefficient and one can takes :

The reflection coefficient or the transition probability i.e. the transition rate divided by the

incident flux hx 2

pi is given by :

incident flux ( 2 7rm ) is given by :

where 2 X is the range parameter of the short range repulsive part of the potential U(z) and p;

=

kf/x with kiz the normal component of the incident particle wavevector.

3. Cohérent and incohérent scattering

-

sticking coefficient.

Let us consider the first term in expression (11). It has been demonstrated previously [10, 11]

that the «fti> values are solutions of the integral equation :

with C

=

EP - Ho and v (R, z, u ( a » is now the potential v in the interaction picture. The

solution of this equation by an iteration procedure shows that the matrix element t is equal to

the sum of elementary contributions, each of them involving only virtual phonon processes.

Each elementary contribution is a matrix element tt/m, n) which can be represented univocally by a diagram and labelled by two numbers (m, n ) : m is the order in perturbation expansion and n the number of virtual phonons. Such a matrix element or a diagram (m, n ) contain the product of n Bose-Einstein factors and consequently are proportional to

Tn at a sufficiently high crystal temperature T. Furthermore one has shown [12] that the

minimum number nm, of phonons involved in a m-order term in the interaction is

m/2 and (m + 1 )/2 for m even or odd respectively. Then we define :

For illustration and further reference figure 1 gives all the diagrams corresponding to one and

two virtual phonon processes.

The detailed form of the different matrix elements depends obviously of the chosen potential V (R, z, u ) which represents the particle crystal interaction but the general analysis

in terms of diagrams remains independent of this particular choice.

As the «fti > matrix element contains only elementary processes of phonon virtual

exchange it can have nonzero value for final states on the energy shell, equal to open

(6)

Fig. 1. - Elastic diagram for

one

and two virtual phonon exchange.

diffracted channels. Then it is easy to show that the coherent reflection coefficient in a

diffracted beam G is given by :

in which now the matrix elements are dimensionless, that is to say are those given by equations (6) or (7) multiplied by the factor 2 m/h2 x2.

Let us consider now the second term in expression (11). The thermally averaged factor

contains the correlation functions «u (0) - u (À » and consequently the integration over Àwill yield a variation of the particle energy by exchange of real phonons. This variation, as well as that of its parallel momentum to the surface, will be equal, respectively, to the algebraic sum

of energy and momentum of the phonons created and annihilated during the scattering event.

It can happen that these two sums are equal to zero and the particle is scattered incoherently

in the specular direction. It can also happen that the particle conserves its energy without

conservation of its parallel momentum to the surface. This particular process supposes the

(7)

exchange of at least two real phonons and can be called elastic incoherent process. It

corresponds to the elastic cross section measured in an experiment where the scattered

particles are resolved in energy for a scattered angle different from the specular direction. In such an experiment one also measures the cross section in the vicinity of this elastic event.

One can expect that this quantity does not exhibit any singularity at the elastic position.

Therefore if a peak is observed for this position it could be due to another effect and can be ascribed to the scattering by surface defects.

Another important point should be outlined. The set of final states are the eigenstates of Ho defined with the potential U(z). Within the assumption of the quasi-flat surface they are approximately the same as those of a Hamiltonian defined with the potential « V (R, z, u » .

Hence if we take a final state such as exp (iKf ’ R) Ob where Ob is a bound state of

U(z ), the reflection coefficient Kf,,Ri gives the probability that the incident particle makes a

transition to a state bound to the surface. As discussed previously [15] this probability is equal

to the sticking coefficient for the bound state considered.

The coherent component of the scattering has been fully analysed in previous studies and the specular intensity as a function of the crystal temperature has been calculated when the

particle is scattered by a one-dimensional potential [11, 12]. It remains to perform the same analysis and calculation for the inelastic component of the scattering including the sticking

coefficient. This will be done in the following sections.

4. Inelastic scattering.

The iteration of equation (7) gives :

where the number under each term labels the expansion or perturbation order.

Replacing t by this expression in equation (11) the bracket under the integration over the À

variable becomes :

or in short hand form, which used the perturbation order notation :

as « 1 > = «v» = 0 (Eq. (2)).

The general term is «m’ m’ » + «m’+ m> for m’= m and «m+ m> with m’

=

m. Hence

the notation m denotes a matrix element taken between the particle states f and i of an operator which contains m potential operators v and m - 1 Green operators. We have :

where we assume the potential v to be real, and 0 is an eingenstate of Ho namely exp (i K - R) 4> p or b where the subscript p or b refers to a continum or bound state respectively.

m + can be written as

(8)

and the sum

4.1 Let us consider the first term in expansion (14). From expression (11) its contribution to the reflection coefficient is given by :

It is interesting to compare this expression to the second order expansion term of the elastic matrix element «;t; » . Equation (12) gives :

In each of these expressions the thermal average bracket contains two potential operators.

Therefore the thermal average operation will yield the same function of the correlated

displacement, namely f( «u (0 ) u (A )» ) and f ( « u (,k ) u (0 ) » ) for (16) and (17), respectively.

To go further one expands f with respect to the correlation function and one gets a general

term proportional to («u(0) u(A )» y or «u(,k)u(O»)" which are proportional to

exp - iA Mhwq and exp + iA M hwq respectively. Then the subsequent integration

over k gives :

- for the inelastic reflection coefficient (16) 2 zr ô Ey - EP - hWq which corresponds

to the exchange of n real phonons.

-

for «it.(2» (17) a dressed propagator G+ located in between two potential operators

now independent of u. G’ writes :

and corresponds in the elastic process to the exchange of n virtual phonons. This matrix element is denoted «itf2, n» .

Due to the form of the correlation function there remain n integrations over w such as

with p a spectral density and « N ( w ) > the Bose-Einstein factor. Therefore these two terms are proportional to T" at sufficiently high crystal temperatures T.

The calculation is achieved by introducing in //2, n) the projector of 03A6) states. This gives the matrix element for an elastic event involving n virtual phonons. In the inelastic reflection coefficient, equation (16) shows that the projection is performed only on the final

state f. Therefore if we know the dimensionless matrix elements itl’ n> we can write the

corresponding inelastic reflection coefficient by using the following rules :

-

replace the dressed propagator by a 6 function which insures the conservation of energy with the phonons involved in the virtual process described by this propagator. This is

equivalent to taking the part of the propagator on the energy shell ;

(9)

-

replace the projection on all states by the projection on the final state ;

-

multiply by 2 7T2/Pi.

This prescription can be applied to the diagrams which represent the il/2, n) matrix

elements. This is equivalent to cutting the dressed propagator between two vertices as shown in figure 2.

Note that the inelastic diagram thus obtained represents a reflection coefficient. They are

denoted (1+ 1 n ) with n the number of real phonons exchanged. At this order n can take on

all integer values.

The above analysis suggests that there exists a relation between the specular intensity

calculated at second order and the total inelastic process calculated at first order. This is demonstrated in appendix I. One finds :

where the sum of fR/l, 1) is extended over all final states and can be separated into two terms,

one for continuum and the other for bound states, as indicated by the subscript c and b respectively. As in the calculation of it/2> we use just one projector this matrix element is also equal to the sum of components due to continuum and bound states. One deduces easily

that :

Fig. 2.

-

Procedure for obtaining inelastic reflection coefficient diagrams (on the right) from elastic

diagrams (on the left). The vertical dashed line is the cutting line. The ¡Ri diagram labelled

(1+ 1, 1 ) is the DWBA term. The last column gives the proportionality to the power of crystal

temperature T (for T sufficiently high such that «N (n ) > f-- T).

(10)

This means that from the calculation of the specular inensity to second order including or disregarding the influence of bound states one can obtain the total sticking coefficient

corresponding to the 1 + 1> inelastic processes.

4.2 The contribution to the inelastic reflection coefficient of the term 1 + 2> is given by expression (16) in which the thermally averaged factor is replaced by :

The corresponding matrix element of the coherent elastic process is « itf3) 1 » . Its thermally

averaged factor is :

For the «2’ 1 > term the expressions corresponding to (20) and (21) are respectively :

and

The latter is equal to (21) in which the time has been reversed.

Because these four expressions contain the same number of potential operators the results demonstrated in the preceding paragraph are valid. The same rules allowing one to obtain the fR; values fR fl, 2) + fR f2, 1)

=

2 Re (fRi1’ 2) and to deduce the diagrammatic representation can

be applied. The Green operator to be suppressed in expressions (21) and (23), or the cutting position in a diagram, is obviously that which corresponds to the position of the projection on

the final state in (20) and (22) respectively.

As indicated in paragraph 3 the «iti(3)> matrix element contains components involving n

virtual phonon processes with n::- 2 (expression (13)). Therefore the matrix elements

« it1(3, n) > and the corresponding reflection coefficient R/I, 2, n) are proportional to Tn and at

least proportional to T 2 at sufficiently high T value. Consequently the only component of

fR; proportional to T is given by the term denoted (1+ 1, 1 ) (Fig. 2) which is the well known distorted wave Born approximation (DWBA).

Figure 3 gives the different diagrams derived from the coherent elastic (3, 2) processes. The

¡Ri components exhibit one real and one virtual phonon processes, or the exchange of two

real phonons and are proportional to T2. Diagrams involving three phonons are obtained by cutting the seven (3, 3) elastic diagrams [12] and are proportional to T3 : there are 2, 3 and 2

¡Ri components with 1, 2 and 3 real phonons exchange respectively.

A unitarity relation analogous to relation (18) for the 1+ 1> term, can be established

(appendix I). One gets :

However to get the matrix element « it1(3 » we should introduce two projectors of the final states f. Therefore it is not a linear combination of continuum and bound states and the

decomposition leading to expression (19) cannot be done.

4.3 The extension of the above procedure to the general term «m"+ m’> + «m’+ m"> is

straightforward and the same rules allowing one to deduce the inelastic reflection coefficient

(11)

Fig. 3. - Inelastic reflection coefficient diagrams obtained with the three elastic diagrams (3, 2) with

the procedure used in figure 2. Cutting

a

propagator located

on

the right of the middle vertex yields

a

diagram with time reversed.

from the elastic matrix element hold. Hence, knowing the latter, one can write the former. In this procedure the diagramatic representation is very helpful because it gives in a simpler way the number of terms in the inelastic reflection coefficient and the number of real and virtual

phonons involved in each inelastic component.

To go into detail, let us consider the elastic matrix element itfm» - It contains

m - 1 Green operators and therefore should be associated with the inelastic terms deduced by suppressing one Green operator, that is to say with all the terms such that

As pointed out in 4.2 when m " >. mi we should associate the «it i(m) > matrix element in which the time has been reversed. However, this fact is included in relation (15) where the

two terms m"+ m> and «m’+ m"> are added and give twice the real part of the former.

Hence it is sufficient to consider the terms for which m" m’. Then for m odd there are

(m - 1 )/2 terms. For m even there are (m - 2 )/2 terms such that m" = m’ to which one

should add the term mi = m" : one the whole there are m/2 terms.

Expression (13) shows that itfm» is a sum of matrix elements «itm,’» where

n

is the number of virtual phonons involved in the process. Since there are different ways to

exchange the n virtual phonons this matrix element is in fact the sum of elementary matrix

(12)

elements : each of them depicts a given process and their number N (m, n) has been calculated

previously [12]. Therefore each «itm, "> matrix element corresponds to a number of

components in the inelastic reflection coefficient equal to :

and

Each of them gives an exchange of p real and q virtual phonons with p + q = n ; p may vary from 1 to n. All of them are always proportional to Tn at sufficiently high T values.

However, the numbers given by (25) are overestimated. Among the elastic matrix elements

or diagrams some of them exhibit an undressed propagator like :

Cutting this line is equivalent, according to our rules, to replacing the Green operator by

the function 2 7T5 (EF - Epf). This leads to an elastic process which does not remain in the inelastic reflection coefficient expression (Eq. (14)) as it cancels with the same term contained in :

This fact is illustrated in figure 4a with a (6, 3) diagram. Furthermore, when m is even, another fact reduces the above calculated number. Then a propagator is a « center of

Fig. 4. - (a) The cutting operation of

an

undressed propagator leads to elastic intensity. (b) Case where

a

cutting operation leads to

one

reflection coefficient ; the propagator cut is in

a

symmetric position. (c)

Case where two cut elastic diagrams yield

one

reflection coefficient component.

(13)

symmetry

»

namely the Green operator located in the middle of the elastic t matrix element :

Cutting the corresponding line leads to an inelastic reflection coefficient of the form

«m + m » and this has been assumed to give N (m, n) inelastic components. In fact, if the elastic

diagram is symmetric with respect to this propagator the cutting operation will lead to one

reflection coefficient (Fig. 4b). On the contrary it is necessary to cut the two elastic diagrams symmetric to each other to get one reflection coefficient (Fig. 4c). Hence cutting N (m, n) diagrams in the middle gives a number of reflection coefficients less than

N (m, n).

Table 1 gives the number Ni of reflection coefficient components deduced for different

elastic (m, n ) matrix elements and the number of these components depicting the exchange of 1, 2 or 3 real phonons. The table contains all the terms proportional to T, T2 or T3. One can remark that there is only one inelastic component proportional to T and 9 proportional to T2 among which 5 depict one real phonon exchange. The exchange of n real phonons starts with terms proportional to Tn.

Table I. m order in t matrix expansion and n number o f virtual phonon processes in the elementary elastic matrix element (t (m, n»), nm is the minimum number o f virtual phonon at

order m. N (m, n) number o f elementary t (m, n) matrix element or diagrams. Ni is the number o f

inelastic reflection coefficient components or inelastic diagrams. The last four columns on the right give the proportionality to the power o f crystal temperature (for sufficiently high T value)

and the number o f inelastic components for 1, 2 and 3 real phonons processes.

Figure 5 gives the inelastic diagrams deduced from the elastic (4, 2) matrix element. With those shown in figures 2 and 3 one has all the single real phonon exchanges proportional to T

and T 2. This will be calculated in the following paragraph. It is interesting to give the unitarity

relation analogous to relations (18) and (24) for this order. One has (appendix 1) :

The last sum comes from the presence of undressed propagators in some elastic diagrams as

mentioned above.

(14)

Fig. 5. - Inelastic reflection coefficient diagrams yielded by cutting the three (4, 2) elastic diagrams

with the procedure used in figure 2. Cutting

a

propagator located

on

the right of the middle propagator yields

a

diagram with time reversed.

5. Flat surface.

As in preceding papers [10-12] we consider the scattering of an incident particle by a flat

surface. The interaction potential is modelled by a simple soft potential :

and the distorted potential U(z) = «V> is of the Morse type. The displacement operator u

acts only on the repulsive part of this potential in the direction normal to the surface. Hence the potential is one dimensional and the particle can gain or lose energy without any change of

its momentum parallel to the surface.

A scattered particle in a continuum state remains in the incident plane and the corresponding calculated reflection coefficient will be greater than that measured in an

experiment. For such experiments where the exchange of parallel momentum is small, one

can expect that the calculation gives the respective strength of the one and two phonon

reflection coefficients and reproduces their evolution with the crystal temperature.

(15)

For particles scattered into bound states one defines the sticking or capture coefficient as

equal to the sum of transition probability into all bound states. This calculated quantity will certainly be dependent on the normal component of the incident particle energy which is for this potential a characteristic quantity. However, one can also expect to obtain a good

evolution of the sticking coefficient as a function of the crystal temperature and also to study

the influence of the different potential parameters on this quantity.

The subsequent different expressions for the reflection coefficient can be calculated directly (expression (11)) or deduced from the elastic matrix element «iti(’» [11] using the rules

exposed in the preceding paragraph. They are expressed with the dimensionless quantities :

-

phonon energy R = A 2 hw

-

eingenvalues of Ho j p 2 f2b = A 2 2 ep = A 1 eb 1 continuum bound state . state

The exchange of energy for the incident particle will be denoted

with

Op and 4>b being the eingenstates of Ho, for continuum and bound states respectively. The potential matrix elements are defined as follows :

The mass of a crystal atom and the spectral density of the u operator are denoted by M and p (d2 ) respectively.

In each fRi expression the final state can be either a continuum (pf) or a bound state

(b ).

5.1 FIRST ORDER. - The two components corresponding to the first and second diagrams of figure 2 are :

The

«

unitary relation » (18) which is valid for each order of phonon exchange is

The imaginary parts of the iti matrix elements are always negative [11] and yield a decrease of the specular intensity when the crystal temperature increases. To a depletion of the specular intensity corresponds an increase of the inelastic scattering. Effectively, the product

p ( n ) « N ( f2 ) > is always positive and the total inelastic reflection coefficient is also positive.

03A9

However one should notice that a calculation of the latter quantity does not allow a

calculation of the specular intensity as one does not know the real part of the elastic matrix

elements.

(16)

5.2 SECOND ORDER. - There are two components for a one phonon real exchange (Fig. 3).

Their reflection coefficients are given respectively by :

with

and

where G (2, 1 ) is the dressed propagator which accounts for the virtual phonon exchange (Q), the real phonon process occurring after (f2’

=

0, Eq. (30a)) or before (f2’ = ,Qo, Eq. (30b)) the virtual one.

The two real phonon exchange gives a component which is written as :

with

and

The particle exchanges the real phonon nt on the first vertex (third term in Fig. 3), then

propagates with an energy pi + f2l 1 and exchanges the second real phonon on the second vertex.

These three terms are proportional to the square of the crystal temperature and to the ratio

m/M.

For all the particle-crystal pairs considered [11], exact numerical calculations indicate that

lm itf3, 2) values are positive leading to an increase of the specular intensity with respect to the value given by the matrix elements « itf2, 1 1) + it.(2, 1 2) >. In agreement with these results the

« unitarity relation » (24) indicates that the total contribution to the inelastic reflection coefficient given by these three terms is negative so that they tend to decrease the inelastic

scattering. However, as we should take the real part of the fA1 expressions the contribution of

(17)

5 function generated in G(2,1) or G(2, 1) does not hold. The only contributions come from the

principal values of these Green functions which depict intermediate processes in which the

.

particle energy is not conserved. One knows that this contribution is not negligible but is in general small compared to the 6 function contribution which depicts the « on energy shell processes

».

Consequently these three components of the inelastic scattering will certainly be

small and probably negligible in many cases. This statement is confirmed by the fact that the elastic matrix element « itf3, 1 2) in both its real and imaginary parts has been found to be small compared to «iti (2,1) > or even «itf2, 2»

5.3 THIRD ORDER. - There are three components for the exchange of one real phonon. The

reflection coefficient is given by :

with fA 1 f4)(4) fA 1 (’)and f A(6) i corresponding to the first, second and third diagrams of figure 5 respectively.

With :

where the g functions are defined in the same way as the four preceding ones, for instance

In expression (34) corresponding to the first diagram of figure 5, the g functions depict all

the virtual phonon processes which end on a continuum state (qgi) or a bound state

(bgi). After these processes the particle propagates freely and exchanges one real phonon at

the last vertex.

Expression (35) corresponding to the second diagram of figure 5 describes the reverse

process. The particle exchanges one real phonon at the first vertex, propagates freely with an

(18)

energy jpi + Q0 and then passes through all elastic virtual phonon processes leading to the

same final state of energy pz + f2o. It is important to notice that, when a phonon of energy

pz + n b (Ei + 1 eb 1 ) is created, the denominator of the second term in (35) becomes equal to

zero. This is a kind of resonance on the final state f

=

b of energy - l1b (- 1 eb ). Then it

appears only in the calculation of the sticking coefficient on the bound state considered. In

particular there is no singularity for a final state in the continuum.

with

which depicts effectively the virtual phonon process (n) and the real phonon exchange (no) on the intermediate vertex (see third diagram of Fig. 5). For purposes of numerical calculation we can write :

which for

In general these three components of the single real phonon exchange contain the product

of two factors of the form :

The product of the two-5 function terms will yield a real negative number and their contribution to the reflection coefficient is effectively contrary to the second order terms

which, as outlined in 5.2, contained just one 6 function. Consequently the third order contribution to the reflection coefficient will be stronger than the second order contribution.

Furthermore one can expect that the product of the two d functions will yield a contribution greater in absolute value than those yielded by the product of the principal parts. In such a

case the fRi(13,1 + 31,1 > value will be negative and proportional to T2. Added to the

D (il, i) proportional to T will yield a parabola shaped variation of the one phonon reflection

coefficient as a function of the crystal temperature always at large T.

For the two real phonon exchange there are two terms and one gets :

A d and A e corresponding to the 4rd and Srd diagrams of figure 5 respectively.

(19)

and

This expression contains a term equal to the product of two matrix elements ¡e q qe i associated with the factor (Cp + f2 + i E )- 1, and another one equal to the preceding but

with time reversed. One phonon is exchanged on the incident vertex and the second phonon

on the final vertex.

fAi is equal to fA d but the C(3,2) ( Q ) functions are replaced by :

The order in which the two phonons are exchanged is inverted in the time reversed part of this expression.

One can demonstrate that the imaginary part of these two term vanishs.

For the purpose of numerical calculation it is advantageous to perform first the integration

over p and/or q variables. Then the product of the two 6 functions yielded by the G (3,2) functions gives a positive contribution. One cannot say anything about the principal part values but one can expect that, on the whole, these two terms give a positive reflection

coefficient component. As all the terms of third order, they are proportional to T 2and (m/M)2.

Now let us look at the unitarity relation (26). For this potential and the order considered it appears as :

For many systems and incident conditions the calculation leads to positive values of

Im {it/4, 2). This confirms our analysis showing that the single real phonon term is negative

and means that the decrease of the single real phonon scattering yielded by this term is large :

it should compensate the increase of the two real phonon processes and of the elastic specular

intensity yielded by the positive value of Im {it/4, 2) .

(20)

6. Results.

The numerical calculations are performed under the same conditions as those for specular intensity calculations reported and discussed in previous papers [11, 12]. We recall here that the u operator is equal to the average of the normal displacements to the surface of the four atoms belonging the unit cell of the (100) face of Copper. This face can be considered as a flat surface. The calculation is purely harmonic ; we do not introduce any anharmonic effects as in

previous papers.

The ten inelastic components written above are calculated and therefore the results include the contribution of all the terms proportional to T and T2. When the components of a given

order have been calculated one can verify the

«

unitarity relation » by calculating indepen- dently the corresponding elastic matrix element. For order one and two (Sects. 5.1 and 5.2)

the error on the

«

unitarity

»

is less than 10- 5. For third order we have pointed out that the expression (35) corresponding to the second diagram of figure 5, exhibits a singularity when

the final state is a bound state. For this final state the singular term is omitted in the calculation and the calculated value of bR fl3,l + 31, 1)i.e. the sticking coefficient for one phonon

events is not correct. The

«

unitarity relation » at third order cannot be satisfied. In order to recover the « unitarity » the

«

unitarity » difference is added to the calculated sticking

coefficient value (6 ) which is called now corrected sticking coefficient (C 6 ). Otherwise stated, this last quantity is given by the difference Ilnel

-

1 (1 + 2 + 3 )c, where I inel is given by

the sum of the left hand side of unitarity relations for order 1, 2 and 3 and I (1 + 2 + 3)c is the total reflection coefficient for continuum states at the three orders. With this procedure, one gets in

a simple way a reliable value of the sticking coefficient. With a three-dimensional potential

like that given by a pairwise potential summation we can show that the above mentioned

singularity disappears. The sticking coefficients is then exactly given by the above procedure.

In the figures presented below the curves labelled (1 )t, (1 + 2 )t and (1 + 2 + 3 )t represent

the values of the total inelastic reflection coefficient calculated with the left hand side of the

unitarity relation for order 1, (1 + 2 ) and (1 + 2 + 3 ), respectively. A label (1 + 2 + 3 )c 1 or (1 + 2 + 3 )c corresponds to the total reflection coefficient to continuum states calculated from the fRi expressions with one and one plus two real phonons processes, respectively. The

calculated sticking coefficient and its corrected value are labelled

6

and C o- respectively.

Finally, the curve labelled 1 - Ioo gives the total inelastic reflection coefficient as

I00 is the calculated specular intensity which includes multiphonon processes [12]. Table II gives the potential parameter values and the bound state energy.

Table II. - Potential parameters and bound state energy for the systems He-Cu (100) [16] and

Ne-Cu (100) [17].

Figures 6 to 8 exhibit the results for an incident Helium atom of kinetic energy 21 mev and incident angles of 73.5, 55.5 and 31.8 degrees respectively. The C6 values are given by the

difference between the values (1 + 2 + 3 )t and (1 + 2 + 3 )c. Figure 9 gives the result for a

Neon atom, the incident angle being 75 degrees.

(21)

Fig. 6.

-

Scattering of He by Cu (100 ) - Ei

=

21 meV, 0i,

=

73.5 degrees, normal particle energy

Ei, = 1 .69 meV. The

curves

labelled (1 )t, (1 + 2 )t, (1 + 2 + 3 )t, 1 - 1 00 gives the total inelastic transition probability calculated with unitarity relations for first, first + second, first + second + third order and to all orders, respectively. The

curve

labelled (1 + 2 + 3)c gives the total inelastic transition

probability up to third order for the continuum states.

In each case (Figs. 6 to 9) the contribution of the two-phonon diagram of figure 2 is completely negligible so that the (1 )t curve corresponds to the values given by the DWBA

term proportional to T. One sees that as the crystal temperature increases, the results yielded by this approximation are no longer good and are greater than the true value 1 - Ioo. Adding

the terms of second order (Fig. 3) does not modify this conclusion as they yield a negligible

contribution compared to the DWBA term. The addition of the third order contribution modifies considerably the preceding values. The total inelastic scattering becomes equal to or

lower than the 1 - Ioo values. The domain of validity of (1 + 2 + 3 ) calculation extends to

higher temperatures and the limit of validity T3 could be appreciated as the temperature where the difference between (1 + 2 + 3 )t and 1 - Ioo curves becomes appreciable (Tab. III).

In particular the C6 values calculated are reliable in this temperature range. Even if this temperature range is reduced as in the Neon case we can see that the Ccr value increases with the temperature and decreases as the normal component of incident energy increases. This last effect can be seen by comparing the results of figures 6, 7 and 8 : for the last case of

relatively high normal energy component the Ca value practically vanishes.

For values of T greater than T3 the contribution of reflection coefficient components proportional to T3 which contain one, two and three real phonon processes becomes non

negligible. They will probably act in such a way that the difference between the

(22)

Fig. 7. - Same

as

figure 6 : 8i

=

55.5 Ei.L = 6.74 meV.

Table III. - Crystal temperature below which an inelastic calculation gives acceptable results.

TD is for the DWBA term (first order) and T3 is for the first plus second and third order with

terms proportional to T and T2 only.

(1 + 2 + 3)t and 1 - 100 curves will be reduced. Also the maximum exhibited by the (1 + 2 + 3 )t curve can be shifted to higher temperatures.

Figures 10 to 13 show the variation of the reflection coefficient with the crystal temperature for particles scattered at a given final angle Of

f

and for the same particles and incident conditions as above. In each case the particle exchanges a small amont of energy (negative or positive) corresponding in the case of single phonon exchange to small phonon momentum.

The curves labelled 1 and 2 give the quoted quantity for one and two real phonon exchange, respectively. The dashed line is the result yielded by the DWBA term.

As has been predicted in the preceding paragraph the two-phonon reflection coefficient

increases as T2. Also in agreement with the above conclusions, the DWBA contribution gives

(23)

Fig. 8. Fig. 9.

Fig. 8. - Same

as

figure 6 : 6i

=

31.8 EiJ. = 15.43 meV.

Fig. 9. - Same

as

figure 6 but for neon. Ei = 63 meV, 6i

=

75 degrees-particle normal energy 4.22 meV. The sticking coefficient is given by the

curve

labelled Co,.

acceptable values in a small domain of temperature, the temperature above which this approximation fails is of the order of TD (see Tab. III). The addition to the DWBA term of

the five contributions proportional to T2, and particularly the three coming from third order, yields a parabola shaped curve for the one-real phonon exchange, as expected. As shown

above these results are valid below the T3 temperature which is close to the maximum exhibited by the one-phonon curves. Beyond this temperature one can expect that the contribution from terms proportional to r3 yields an increase of the one-phonon reflection

coefficient value and consequently also yields a more or less great shift of the maximum

position. For the two-phonon processes, this contribution will certainly be negative yielding

the appearance of a maximum, whereas the three-phonon reflection coefficient will increase

as r3.

7. Discussion.

Generally speaking the reflection coefficient is equal to the probability that the incident

particle undergoes a transition to the given final state. During this transition some real

(24)

Fig. 10. - Transition probability for He scattered by Cu (100) : Ei

=

21 meV, Oi

=

73.5 degrees ; Of

=

80.2 (a) and 65.9 (b) degrees. Ef -- Ep is the particle exchange of energy (meV). Dashed line DWBA - curves labelled 1 and 2 : one plus second and third order calculation with terms proportional

to T and T2 only,

one

and two real phonon exchange respectively.

Fig. 11. - Same

as

figure 10 but 8i = 55 degrees ; 8 f

=

62.5 (a) and 48.6 (b) degrees.

Références

Documents relatifs

The production cross sections of the γ rays observed from the minor isotopes of nat Ti in comparison with TALYS theoretical calculations performed using the default input

However, a significant decrease of the low-frequency Raman intensity at 300 °C can be observed, i.e., at sintering temperatures below the temperature range where the average

We report the Bragg spectroscopy of interacting one-dimensional Bose gases loaded in optical lattices across the superfluid to Mott-insulator phase transition.. Elementary

Inelastic scattering of neutrons has revealed soft optic modes at the T point (b* + c* ) of the Brillouin zone both in Rb~ZnC14 and K2ZnC14 which are responsible for the 22.

Light scattering from the free surface near a second order nematic to smectic a phase

Calcula- for each of the energy spectra, while figure 3 shows tions of the excitation spectrum of 3 ~ e at high pres- the distribution of scattered intensity at the three

Inelastic neutron scattering of hydrogen trapped in solid

Résumé. — Les mesures de la diffusion des neutrons ont donné en détail la forme de la fonction de diffusion pour les fluctuations des spins transverses dans la région critique