• Aucun résultat trouvé

The Compton profile of neon : comparison between experiment and the impulse approximation

N/A
N/A
Protected

Academic year: 2021

Partager "The Compton profile of neon : comparison between experiment and the impulse approximation"

Copied!
6
0
0

Texte intégral

(1)

HAL Id: jpa-00209512

https://hal.archives-ouvertes.fr/jpa-00209512

Submitted on 1 Jan 1982

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.

The Compton profile of neon : comparison between experiment and the impulse approximation

A. Lahmam-Bennani, A. Duguet

To cite this version:

A. Lahmam-Bennani, A. Duguet. The Compton profile of neon : comparison between ex- periment and the impulse approximation. Journal de Physique, 1982, 43 (9), pp.1333-1337.

�10.1051/jphys:019820043090133300�. �jpa-00209512�

(2)

The Compton profile of neon : comparison between experiment and the impulse approximation

A. Lahmam-Bennani and A. Duguet

Laboratoire des Collisions Atomiques et Moléculaires, Bât. 351, Université de Paris-Sud, F 91405 Orsay, France (Rep le 16 decembre 1981, révisé le 30 avril 1982, accepté le 14 mai 1982)

Résumé.

2014

Les profils Compton total et de valence du néon sont mesurés par impact électronique à haute énergie (25 keV). Leur partie symétrique est comparée à des expériences antérieures et à des calculs dans le cadre de l’appro-

ximation de l’impulsion.

Abstract.

2014

The valence and total Compton profiles of

neon are

measured using high-energy (25 keV) electron scattering. Their symmetrical parts

are

compared to previously published experiments and to impulse approxi-

mation calculations.

Classification

Physics Abstracts

34.50H - 34.80D - 34.80G

1. Introduction.

-

Compton profile (CP) determi-

nation is an important tool for studying electronic properties of atoms and solids. These profiles are simply related, within the framework of the impulse approximation (IA1 to the electron momentum den-

sity p(p) of the system under study, and hence they

constitute a very sensitive test of the wavefunction used

to describe this system.

The neon atom was chosen for the present investiga-

tion because of its relative simplicity for theoreticians

as well as for experimentalists. Indeed, a great many theoretical calculations are available for the total and valence CP’s of this atom, mostly within the framework of the IA, using different wavefunctions. However

they differ at the centre of the profile by more than 3 %

from each other, and the best ones are usually 2 %

lower than Eisenberger’s [1] experimental data. These

are valence CP’s, measured in the early seventies using photon impact at two different energies, but they were shown by Mendelsohn et al. [2] to present

some internal inconsistency near q

=

0. Later elec- tron impact data by Wong et al. [3] did not shed more light due to the lack of precision in the scattering angle measurement, and a doubt was left by the

authors as to the definition of an

«

effective

»

valence CP. Such conflicting results warranted a new investi- gation of the Ne valence and total CP’s (the later being here directly measured for the first time), taking advantage of the high accuracy and precision

achieved with high energy electron impact spectro- scopy.

2. Experiment.

-

2.1 METHOD. - The apparatus and data acquisition method have been previously

described [4]. Briefly, a 25 keV, 0.1 to 200 pA 0.25 mm

FWHM electron beam crosses a gaseous target beam of 1 mm FWHM. The scattering intensity is

observed in the angular range 0.70 to 200, With an acceptance angle of 0.010 and an overall uncertainty

in the scattering angle of ± 0.0030. The scattered electrons are energy analysed with a resolution of

2-15 eV over an energy-loss range up to 7.5 keV, depending on the scattering angle under investigation.

At least 20 000 counts are accumulated at the maxi-

mum of the inelastic profile for each scattering angle, except for the measurements taken at - 180 and - 200 where only 10 000 counts are reached because of the weakness of the scattered intensity. Extensive tests

have been carried out in order to eliminate the effects of energy resolution, multiple scattering and other

extraneous scattering.

The relative cross sections measured at fixed

scattering angle 0 are corrected for detector noise and dead time, and for analyser transmi’ssion, then

transformed (see Ref. [4]) to generalized oscillator strengths (GOS), df(K, E)/dE, where K is the momen-

tum transfer and E the energy loss. These GOS’s corrected for the missing portion of the high energy-

loss tail are made absolute by use of the Bethe sum rule.

Following Bonham and Tavard [5], an electron CP, J(q, K), is then defined from the spectra using

whose limit for K -+

oo

is simply the impulse CP, J,A(q). At large but non infinite K values, Gasser

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:019820043090133300

(3)

1334

and Tavard [6] have shown through

a

series develop-

ment that J(q, K) consists of a leading term symme- trical with respect to the q variable and corresponding

to the IA, plus a first order corrective term which is

antisymmetrical with respect to q and whose magni-

tude decreases with increasing K. Hence, if an asymp- totic limit, J(q), is approached for the experimental J(q, K) profiles as a function of K, its symmetrical part, Js(q)

=

(J(q) + J ( - q))12, should be compared

to the calculated JIA(q).

The profile dependence upon K was thus investi-

gated using the following characteristic quantities :

the maximum intensity of the profile, Jmax(q, K ), its full

width at half maximum, Q, and the Compton defect parameters, Aq and 6q, respectively characterizing the

deviation of the maximum of the profile from the

q

=

0 position and the asymmetry of the profile. All

these quantities were found [7] to reach constant values, within experimental uncertainties, in the K-

ranges 4.5 K 7 a.u. and 13 K 15 a.u.,

respectively for the valence and total profiles. (The

upper limit, K

=

15 a.u. is the largest K value consi- dered in this work.) Moreover the area under the CP which is in the IA equal to the number of electrons

involved, N, is not more than 1.5 % lower than N. One may thus consider the asymptotic limit to be suffi- ciently approached in these K-ranges so that Gasser and Tavard’s development may be truncated to the first corrective term (antisymmetrical). That is, the symmetrical part of the experimental CP can be directly compared to the IA CP. Note that our experi- mentally observed plateau for Jmax (q, K) leads to the

conclusion that the K-dependence of the sum of the higher order corrective terms is not strong enough to

be observed within the statistical accuracy of the

experiment.

«

Asymptotic » valence and total CP’s, J (q), were

then determined in the above defined K-ranges from

the average of respectively 8 and 14 experimental runs.

The symmetrical profiles, JS(q), were calculated from these average ones and are shown in tables I and II

for q ranging from 0 to 3 a.u., together with the sym- metrical CP of the 1 s inner shell orbital obtained from the difference of the total and valence ones. The numbers shown in parenthesis are the maximum

deviations from the average values, obtained from a graphical fit to the superposition of all the J(q, K)

CP’s. The standard deviations are about 2-3 times lower. Although the uncertainty of the Is profile is

rather large (about 20 % maximum deviation near

q

=

0) because this profile is obtained from the diffe-

rence of two much larger quantities, the present

measurements constitute the first experimental deter-

mination of a core CP.

2.2 COMPARISON TO THE IA AND DISCUSSION.

-

A

large number of theoretical calculations have been

published for the neon CP, using wavefunctions of different quality. In order to avoid a lengthy presen- tation of the following discussion, we have chosen to

Table I.

-

The measured symmetrical total and

core

profiles compared to IA calculations : the VHS-CI calculations by Smith Jr. and Brown, Ref. [14] the full . correlated calculations by Tong and Lam (TL), Ref [20] ;

the HF-C calculations by Weiss et al., Ref [9].

Table II.

-

Same

as

table I, but for the valence profile, : Eisenberger’s x-ray experiments, Ref. [1]; Wong et al.’s

electron experiment, Rçf. [3] ; L-Corr. IA calculations

by Naon and Cornille, Ref. [21].

first compare all the IA CP’s to the actual results at the maximum of the profile, i.e. in the key region of low

momentum density contributions. Comparison at all q

values will then be made for only the best calculations.

(4)

Table III.

-

Comparison between the experimental and theoretical values of J(0) for the total and valence pro-

files: GTO : Naon et al., Ref [11]; C-CI : Naon and Cornille, Ref [12] ; MCHF : Smith Jr. and Brown, Ref [14]

for the total profile and Mendelsohn et al., Ref. [2] for the valence one ; LDM : Sabin and Trickey, Ref. [ 17], except

for TL : Tong and Lam, Ref. [20]. Other symbols

same as

in tables I and II.

Table III gives for the total and valence profiles the experimentally determined average values of J (0) with

their maximum deviation, as well as the other existing experiments and the most representative calculations.

The different calculations have been roughly classified

into 3 categories : using Hartree-Fock (HF) wave- functions, correlated wavefunctions or local density

models (LDM).

2. 2.1 Total profile.

-

The only other direct experi-

mental investigation of the total CP of Ne is due to

Wellenstein’s group at Brandeis University [8], using

35 keV electron scattering. But their measurements were limited to a single value of the momentum

transfer, K

=

10.7 a.u., where the asymptotic IA limit

is not reached according to the above discussion. Also their profile was not Bethe sum rule normalized but matched to the IA HF one at q

=

0. Hence no compa- rison can be made here with our results. On the other

hand, Eisenberger’s [1] values quoted in table III are

obtained from his experimental valence J(O) values

shown in table II, to which we have added the Is HF contribution calculated by Weiss et al. [9]. Hence, these results will be discussed below, with the valence profile

ones.

The calculations by Weiss et at [9], who used the HF wavefunctions tabulated by Clementi [10] (HF-C),

lead to almost 2 % underestimate of the J(O) value.

Naon et al. [11] have also performed calculations using approximate HF gaussian type orbitals (GTO) wave-

functions. The effect of increasing the size of the

gaussian basis set was found to improve J (0). However, the agreement with our experiment is poorer than the HF-C calculation even for the largest basis set used (shown in table III).

A 1.5 % underestimate of the J (0) value is also observed (5th column of table III) with the value of Naon and Cornille [12] who used the three term confi-

guration interaction wavefunction of Clementi et a1.

[13] (C-CI). This wavefunction takes into account only

radial correlation and gives about 20 % of the correla-

tion energy. Two other calculations by Smith and

Brown [14, 15], who used more correlated wave-

functions, do not bring better agreement with experi-

ment : the multiconfiguration Hartree-Fock (MCHF)

wavefunction of Ahlrichs and Hinze where only the

L-shell correlations are included, and the second order CI wavefunctions of Viers et al. [16] (VHS-CI). The discrepancy seems to become even larger with the

inclusion of more correlations.

In the sixth column of table III are the values obtain- ed by Sabin and Trickey [17] using numerical wave-

functions generated by five LDM for exchange corre-

lation (called Xa models). The value of the local

exchange parameter

a

(usually 2/3

a

1) has

been chosen to be

a =

2/3 in the X(X2/3 model, and

a =

(XvT

=

0.729 97 (the so-called

«

virial theorem »

value) in the XXyT model, while the Xa.

,

one has a stepfunction radial dependence for

a.

The other two models, Ng and HL, due to Ng [18] and to Hedin and Lundqvist [19], contain terms beyond the X a approxi-

mation which are intended to represent correlation corrections.

Except for the first model which leads to about 1 %

overestimate of the experiment, the agreement is very

satisfactory for the others, specially the XaVT, Ng,

and HL ones where the deviation from the experiment

is lower than 0.4 %. Finally, Tong and Lam [20] have

used wavefunctions with full correlation contributions included, in the local density approximation. The J(q) values (labelled TL in table III) are then found as

a

sum of a first term, J ’(q), and a correction term,

J2(q). However, as stated by the authors themselves,

the second term overcorrects at small q values, making J(0) too low. It is obvious, comparing the different slopes dJ (q)ldq from their table II, that this over- correction is quite large, and likely affects not only the J (0) value but also J(0.1).

Our total profile is compared in table I, for all q values, to the most sophisticated IA calculations [14, 20] VHS-CI and TL. Figure 1 gives the relative devia- tion (Js(q) -JIA(q))/Js(q). As discussed above, all the

IA CP’s are below the measured one near q

=

0,

(5)

1336

Fig. 1.

-

Total CP : relative deviations of the IA calculations from

our

symmetrical CP, Js(q). The experiment is repre- sented by the horizontal axis, and the solid lines

are

the

one

standard deviation

error.

The IA calculations

are :

(- -) by Smith Jr. and Brown, Ref. [14], and (-) by Tong

and Lam, Ref. [20].

except in reference [ 17] (not shown in table I as only a

few points are given by the authors). Otherwise, they

more or less satisfactorily reproduce Js (q), within,our

experimental uncertainties, up to q - 2 a.u., but are

again significantly lower at larger q values.

2.2.2 Valence profi°le.

--

Our valence J(O) value is compared in table III to the other existing experi-

mental or theoretical ones. Eisenberger [1] has observ- ed, in two separate experiments, the scattering of

MoKa (17 374 eV) and AgKa (22 163 eV) x-rays

through 1700 from neon. He deduced an

«

experi-

mental » valence CP from the measured total profile by subtracting theoretical core profiles corrected for

the

«

violation » of the IA. The MoKa result agrees with our’s, within the experimental uncertainties. The

AgKa case will be discussed below. Wong et al. [3]

measured the valence CP using 25 keV electron scattering, with an uncertainty of (2-3 %) near the

maximum. Their measurements at three different

scattering angles indicate a definite dependence of J(0) upon the momentum transfer K, the 9.5° value

being almost 4 % lower than the 7° value. Our corres-

ponding K-range is characterized by a plateau, where

the measurements taken at 11 different scattering angles are randomly scattered around the average value with a maximum spread of 1 %. The deviation

of Wong et al.’s results from our’s is respectively about 5, 3 and I% at 7°, and 9.5°. Only the last two values

fall within the combined experimental error bars.

Some inconsistencies are to be noted in reference [3] :

the statement (page 1857) that up to 80 the profile changes to

a «

higher profile

as

the scattering angle

increased » is in contradiction with the data shown in their table V. Also, in the same table, the experimental profiles do not peak at q

=

0, J(o.1) being larger than J(0). We thus suspect that some errors might be present in this table.

Columns 4 and 5 of our table III give the theoretical IA determinations ofJ(O). A 2 % underestimate of the

J(O) value is again observed with the above cited

Weiss et al. [9] HF-C calculations, as well as with the

calculations by both Naon and Cornille [21] (L-Corr.)

who used a variational formalism which reproduces

about 98 % of the L-shell correlation energy, and by

Mendelsohn et al. [2] who used, in their MCHF calcu-

lation, the same correlated Ahlrichs-Hinze wave-

function previously utilized in reference [14]. This

wavefunction is said to account for 67 % of the valence electrons correlation energy.

>

Column 7 of table III shows the only calculation which is not within the framework of the IA. Mendel- sohn et al. [2, 22] have calculated L-shell CP’s, within

the Bom approximation, with an effective nuclear

charge determined from the matching of the impulse hydrogenic CP to the impulse HF profile at q

=

0. The method is referred to

as

the exact hydrogenic (EH)

method. While any IA will be inherently independent of

incident particle energy, the EH method exhibits the

profile dependence on energy. Hence, Mendelsohn

et al. [2] have performed EH calculations for the

Eisenberger [1] x-ray experiments. The MoKa results

agreed well with the experiment, while the AgKa

results were in slightly poorer agreement. Consi-

dering energy dependence arguments, they have

shown that there is,

«

at the very least,... some incon-

sistency between the two sets of experimental data ».

On the other hand, Eisenberger has only reported the

1 % statistical uncertainty as determined by his count

rates. We believe that the systematic error in experi-

mental CP studies is usually not very small compared

to the statistical uncertainty, either in photon or elec-

tron impact work. Hence the 1.3 % difference between the Ag and the Mo experimental J(O) values likely

falls within the error bars. If one considers also i) the

very good agreement of the EH method with our

J(0) determination, the two EH calculations being

within the experimental uncertainty, and ii) the fact

that our result is experimentally observed to be K- independent, one may suspect the silver x-ray value to be slightly too low.

Our valence profile is compared in table II, for all q values, to the previously existing experiments and to

the most sophisticated IA calculations [21] L-corr.

Figure 2 gives the corresponding percentage deviations from our J.(q) values. Eisenberger’s MoKa experiment

agrees reasonably well with our valence one within

the combined experimental uncertainties. Wong et

al.’s one differs from the actual values, because these

(6)

Fig. 2.

-

Valence CP : relative deviations of the previously existing experiments and the IA calculations from

our

symmetrical CP, J5(q). The actual results

are

represented by

the horizontal axis, and the solid lines

are

the

one

standard deviation

error.

(o o o o o) : Eisenberger’s MoKa experi- ment, Ref. [1]; (-A-) : Wong et al.’s electron experiment,

Ref. [3]; (-.-) : the IA calculation by Naon and Comille,

Ref. [21].

authors have only used the

«

high energy-loss side »

of their measured CP. Indeed, their data are found to be in very good agreement with our J (q > 0) ones. As

observed for the total profile, the IA calculations are

below the experiment near q

=

0, while they satis- factorily reproduce Js(q) up to q - 1.5 a.u. within our experimental uncertainties. Above this value, they are again several percents below JS(q).

3. Conclusion

-

We have shown evidence for the existence of a deviation between JIA(q) and the observ-

ed symmetrical CP, Js(q). More confidence in the present experiment is given by the very good agreement between our J(? > 0) values and the « high energy- loss side » electron impact data of Wong et al. obtained using the same incident energy. On the other hand,

the maximum height, Jmax(q, K ), and the FWHM of the CP have reached constant values for the K-ranges investigated in this work, i.e. the asymptotic limit predicted by Gasser and Tavard’s development is reached, at least for the top 50 % part of the profile.

Therefore one may conclude that the symmetrical profile, J,(q), is not fully represented by the IA model and that the inclusion of the next symmetrical term(s)

of the Born’s propagator development is necessary to account for the difference. This clearly shows to what

extent the IA can be used as a comparison model for

the experiment, and points out the need for extending

Gasser and Tavard’s quantitative calculations on H and He to the Ne atom.

The authors would like to acknowledge Pr. M.

Rouault for his constant interest in this work.

References

[1] EISENBERGER, P., Phys. Rev. A 5 (1972) 628 ;

EISENBERGER, P., HENNEKER, W. H., CADE, P. E., J.

Chem. Phys. 56 (1972) 1207.

[2] MENDELSOHN, L. B., BLOCH, B. J., SMITH Jr., V. H., Phys. Rev. Lett. 31 (1973) 266.

[3] WONG, T. C., LEE, J. S., WELLENSTEIN, H. F., BONHAM,

R. A., Phys. Rev. A 12 (1975) 1846.

[4] LAHMAM-BENNANI, A., DUGUET, A., WELLENSTEIN,

H. F., ROUAULT, M., J. Chem. Phys. 72 (1980) 6398.

[5] BONHAM, R. A., TAVARD, C., J. Chem. Phys. 59 (1973)

4691.

[6] GASSER, F., TAVARD, C., C.R. Hebd. Séan. Acad. Sci.

286 (1978) B13 ; J. Chim. Phys. 78 (1981) 341 and

78 (1981) 487.

[7] LAHMAM-BENNANI, A., DUGUET, A., Chem. Phys.

Lett. 74 (1980) 85 ;

DUGUET, A., LAHMAM-BENNANI, A., ROUAULT, M., J.

Physique-Lett. 41 (1980) L-325 ;

LAHMAM-BENNANI, A., DUGUET, A., ROUAULT, M., J.

Chem. Phys., to be published, (1982).

[8] RUECKNER, W. H. E., BARLAS, A. D., WELLENSTEIN,

H. F., Phys. Rev. A 18 (1978) 895;

WONG, T. C., MENDELSOHN, L. B., GROSSMAN, H., WELLENSTEIN, H. F., Phys. Rev. Lett., to be published (1982).

[9] WEISS, R. J., HARVEY, A., PHILLIPS, W. C., Philos, Mag. 17 (1968) 241.

[10] CLEMENTI, E., IBM J. Res. Dev. 9 (1965) 2.

[11] NAON, M., CORNILLE, M., ROUX, M., J. Phys. B 4 (1971) 1593.

[12] NAON, M., CORNILLE, M., J. Phys. B 4 (1971) 1210.

[13] CLEMENTI, E., KRAEMER, W., SALEX, C., J. Chem. Phys.

53 (1970) 125.

[14] SMITH Jr., V. H., BROWN, R. E., Chem. Phys. Lett.

20 (1973) 424.

[15] BROWN, R. E., SMITH Jr., V. H., Mol. Phys. 34 (1977)

713.

[16] VIERS, W., HARRIS, F. E., SCHAFFER III, H. F., Phys.

Rev. A 1 (1970) 24.

[17] SABIN, J. R., TRICKEY, S. B., J. Phys. B8 (1975)

2593.

[18] NG, K. C., Thesis, University of Florida (1974).

[19] HEDIN, L., LUNDQVIST I, B., J. Phys. C 4 (1971) 2064.

[20] TONG, B. Y., LAM, L., Phys. Rev. A 18 (1978) 552.

[21] NAON, M., CORNILLE, M., J. Phys. B 6 (1973) 954.

[22] MENDELSOHN, L. B., BLOCH, B. J., Phys. Rev. A 12

(1975) 551.

Références

Documents relatifs

Indeed, the masses of all elementary fermions follow the formula 2 n .p 2 /2 where n is an integer [3, 4] calculated for the electron, electron neutrino and quark up and p is

Moreover, I reviewed epidemic models that account for host movement to investigate how different types of host movement affect epidemic spread, I find that the main aspect that

In fact, if we are able to quickly compute an upper bound sim UB (qp, p i ) on the similarity between qp and a pattern p i , then it is possible to avoid computing sim(qp, p i )

As a corollary of Theorem 2, we shall deduce now sharp upper bounds for the forward scattering amplitude and for the total cross-section. 25) ensures that. Annales

The Temporary Administration Measures for the Verifi cation and Approval of Overseas Investment Project provides that “overseas invest- ment” refers to investment

As a first step toward determining the full potential energy surfaces of this species for subsequent simulations of its dissociation processes, we investigate the performance

In section 4, we point out the sensibility of the controllability problem with respect to the numerical approximation, and then present an efficient and robust semi-discrete scheme

physical, mathematical, rigid and grid models of the system “gas – rotor – stator” have been worked out; 2 FSI (two-way Fluid-Structure Interaction)