• Aucun résultat trouvé

On the electronic structure of Hgn and Hg+ n aggregates

N/A
N/A
Protected

Academic year: 2021

Partager "On the electronic structure of Hgn and Hg+ n aggregates"

Copied!
10
0
0

Texte intégral

(1)

HAL Id: jpa-00211092

https://hal.archives-ouvertes.fr/jpa-00211092

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.

On the electronic structure of Hgn and Hg+ n aggregates

P. Joyes, R.J. Tarento

To cite this version:

P. Joyes, R.J. Tarento. On the electronic structure of Hgn and Hg+ n aggregates. Journal de Physique,

1989, 50 (18), pp.2673-2681. �10.1051/jphys:0198900500180267300�. �jpa-00211092�

(2)

On the electronic structure of Hgn and Hg+n aggregates

P. Joyes (1) and R. J. Tarento (1,2)

(1) Laboratoire de Physique des Solides, Bât. 510, Université Paris-Sud, 91405 Orsay, France

(2) Laboratoire de Physique des Matériaux, 1 place Aristide Briand, 92295 Meudon, France (Reçu le 19 mai 1989, révisé le 14 juin 1989, accepté le 16 juin 1989)

Résumé.

2014

Nous étudions les agrégats de mercure un changement dans la nature de la liaison

chimique apparaît aux faibles nombres d’atomes. Dans une première partie, nous considérons les

agrégats neutres où la transition métal ~ Van der Waals apparaît pour nc ~ 15, résultat en bon accord avec un autre travail théorique. Dans une deuxième partie, nous étudions l’énergie

d’ionisation I. Pour n ~ 50, un modèle de type métallique donne le bon ordre de grandeur. Pour

n ~ 10, ce modèle ne convient plus et nous présentons une autre méthode où les résultats

expérimentaux, pour n ~ 4, sont utilisés pour fixer les paramètres. Pour 4 n ~ 7, l’accord avec

l’expérience est bon. Nous examinons aussi le problème de la localisation de la charge positive en

excès dans Hg+n.

Abstract.

2014

We study small mercury aggregates in which the nature of the chemical bonding changes with respect of the bulk. In a first part, we consider neutral aggregates for which we find that the transition from the metallic to the Van der Waals bonding occurs for nc ~ 15 in close agreement with a previous theoretical work. In a second part we study the ionization energy I. For n ~ 50, a simple metallic model gives the experimental order of magnitude. This model is

no longer valid in the n ~ 10 range for which we present another calculation. In this study we fix

our parameters by using the experimental values for n ~ 4. For 4 n ~ 7, our results are in agreement with experiment. We also examine the localization of the excess positive charge.

Classification

Physics Abstracts

36.40 - 31.20

1. Introduction.

Mercury is one of the divalent elements where a change in the nature of the chemical bonding

can be expected when the average connectivity varies (here, connectivity means the number

of nearest neighbours per atom). This has first been observed in the bulk phase, when the liquid is expanded, at a density of about 8-9 g/cm3, a metal-insulator transition is observed [1].

This conclusion was obtained by electrical conduction, thermopower, Hall coefficient and

Knight shift studies (Refs. given in [1]). Several theoretical works have been devoted to this

question and the general interpretation is that, as the density decreases, a gap appears at the Fermi level, between the s and the p bands, which causes the insulator behaviour [2, 3].

Another situation where the connectivity is lowered with respect to the bulk appears for

Hgn aggregates. There again, when n decreases, numerous interesting physical effects have

been observed : we will mention three of them.

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:0198900500180267300

(3)

2674

a) One knows that in metallic aggregates (alkalis, noble or transition metals) the ionization

potential I(R) as well as the electronic affinity AE(R) exhibits R-1 dependence.

R is the radius of the aggregate [4-7]

(W is the bulk work-function). For mercury it has been shown [8] that 7(/?) followed relation

(1) only for large sizes (n -- 70). In this study R is related to n, the number of atoms, by

R

=

ro n 1/3 where ro is an atomic radius. For smaller sizes, the experimental values are larger

than the predictions of relation (1) by more than one eV in the range n -- 10. Let us notice that in the analysis of reference [8] the factor 3/8 of relation (1) is replaced by 1/2 but this does not alter the general conclusion.

The previous observation has been interpreted as due to a change in the chemical bonding

which would occur in a progressive way for ne S 70. For n ::-.> nc the nature of the bonding is

the same as in the bulk and for n s nc the metallic character is progressively lost in a transition

towards a Van der Waals bonding (as in rare gases). The physical explanation of this

transition is the same as for the bulk metal-insulator transition mentioned before. As the

connectivity decreases, the s and p bands separate and there is no longer any covalent bonding

because in the s band the bonding and the antibonding levels are complete.

b) Another experimental observation tends to confirm this conclusion. In metallic aggregates it is usually observed that the average interatomic distance d(R) decreases with size [9, 10]. Simple models have been proposed to interpret this behaviour [11, 12]. They lead

to a relation between d and Z (the average connectivity) :

where the subscript oo refers to the bulk phase and where p and q are coefficients which give

the dependence on the interatomic distance r of the repulsion energy between atomic cores

(a exp (- pr ) ) and of the electronic energy (a - exp (- qr ) ). As for the metallic elements

p

>

q, the application of (3) in the whole size range gives the expected decrease of

d with Z. Mercury does not follow this general behaviour since d (R ) increases from the bulk to the dimer (d. = 3.03 Â, d(Hg2)

=

3.36 Â [13]). This effect can be interpreted as a

transition to Van der Waals bonding as n decreases since it is known that in this type of bonding an increase of d(R) is expected as R decreases [14].

An experimental study of d (R ) for mercury has not yet been made. Such a study might

reveal the two regimes discussed before : decrease with R before the Van der Waals transition and increase after it.

c) Another interesting experimental result has been given by the study of the inner shell autoionization spectrum [15, 16]. The atomic transition which is studied is

where the d state is either d5/2 or d3/2. When this kind of excitation occurs in Hgn it first keeps

its atomic character but, since the excitation energy is larger than the ionization energy of the aggregate, ionization occurs after a time of about 10-13 s. Let us notice that ionization occurs

for both d5/2 and d3/2 excitations for n :::. 2 but it occurs only for the d5/2 excitation in the atom.

Two regimes are observed. For Az 12, the excitation energy is inversely proportional to

n in a way which is characteristic of an excitonic behaviour. As an exciton is created one may

(4)

conclude that in this range of n values the bonding is of Van der Waals type. For 12 rx n s 40 the previous 1/n dependence is no longer followed which shows that a deviation from a Van der Waals bonding in the whole aggregate begins to appear. It has also been observed that the shape of the experimental lines is markedly asymmetric for aggregates larger than n - 20 which is an illustration of a gradual evolution towards the bulk character.

The three experimental results mentioned above lead us to the conclusion that a change in

the nature of the chemical bonding occurs in Hgn at a value nc of about some tens of atoms.

The fact that all the experiments do not give exactly the same value for nc seems to be less

important, the main point being the existence of a transition.

Let us notice that, once the covalent bonding begins to appear, the electronic properties are

still far from the bulk ones. The study of the evolution with size of each of them will be done

separately and is not in the scope of this paper.

We will limit ourselves to band structure considerations in Hgn and will study (Sect. 2) the

decrease of the mixing of the s and p bands as the size decreases. We use a tight binding

method whose results agree for small sizes with previous calculations [16] and has the

advantage of being easily extendable to aggregates with n

>

100.

In section 3, we will examine Hg’ ions and particularly the localization of the excess singly positive charge in these aggregates. Let us discuss briefly this question.

According to the general laws of electrostatics when a macroscopic conducting medium is charged, the charges tend to localize at the surface of the medium where they adopt the spatial distribution which minimizes their electrostatic repulsion energy. This kind of model has been extended to metallic aggregates. For example, it has been used for determining the

critical number of atoms, nP (X) above which Xn + aggregates (X is a given element) begin to

be stable. For copper the agreement between theoretical results and experiments is satisfactory : n;ri (Cu )theo

=

14 [17] ; n;ri (Cu )exp

=

20 [18]. Let us notice that the foremen- tioned experimental value has been recently confirmed by results on silver and gold :

n 3j(Ag)

=

22 ; n 3 (Au)

=

17 [19] which are of the same order of magnitude.

As a change in the nature of bonding seems to occur in neutral aggregates, one may wonder whether a similar change could occur in ionized aggregates. In this spirit, a discussion of a

possible « Van der Waals » structure of Hgn + is given in [20] where the

«

2+

»

charge is

localised on one central site. This model allows us to understand why particles such as

Hg 5 2+ do exist whereas a calculation of their stability by a metallic model leads to the opposite

conclusion. Let us notice that the problem of the charge localization in ionized divalent aggregates is a general question also studied in [21] for Mgn + and [22] for Ben + .

2. Neutral Hgn.

For determining the binding energy per atom Ec and also the ionization energy of

Hgn aggregates we use the following Hamiltonian which describes the coupling between an s

and a p bands

where the notation is the same as in reference [3] and where e, and Ep are the energies of the

atomic 6s and 6p levels ; ts, tp and y2 are the intersite hopping integrals s

--+

s,

p -+ p and s

-

p and y 1 is the intrasite hopping integral s

-

p.

(5)

2676

The method of resolution of (4) is detailed in [12]. In this work, it is shown that the energies

of the three bands Ep(i ) (doubly degenerate), ESP(i ) and EPS(i ) (singly degenerate) are given by :

where e (i ) are the n (non-dimensioned) eigenvalues (1 * 1 * n ) of the matrix obtained from 4 by letting Ep

=

-S=tp= yl = y2 = 0 and ts=l.

Our calculations have been performed on three cuboctahedral : shapes with n = 13, 55 and 147, with ep - e,

=

6.1 eV, ts

= -

0.6 eV, tp

= -

0.5 eV, Y2

=

0.2 eV, YI

=

0. These shapes

have already been used for the study of Hgn aggregates [16, 23]. Our parameters are of the

same order of magnitude as those given in [23]. They are justified in particular by the fact they give the correct bulk limit for Ec as we will see below. Let us also note that in our calculation there is one parameter less than in reference [23] since we have only one p --+ p integral

instead of two.

One of the advantages of equations (5) and (6) is to allow the study of large aggregates since

we only need to know the n 6 (i ) values. The method can also be extended to the bulk by using in (5) and (6) the e (i ) values deduced from periodic fcc boundary conditions. We then obtain E,

=

0.52 eV, a value which is close to the measured cohesive energy per atom

-

0.5 eV [20]. Our other Ec values are given in table 1 they are very similar to those obtained in [16], 0.20 and 0.28 eV instead of 0.20 and 0.27 eV for n = 13 and 55, respectively. We also give the values of the energy difference between the one-electron Fermi level and the 6s atomic level. We observe that this quantity increases with n.

An evaluation of the binding energy in a Van der Waals bonding is :

This evaluation, made in [16], gives 0.14 eV and 0.20 eV for n

=

13 and 55 respectively.

A comparison with the Ec values of table 1 would lead to the conclusion that the covalent

bonding is more stable than the Van der Waals bonding, even for n = 13. However we have not taken into account the reduction of the covalent energy due to the electronic correlations.

One knows that the larger the reduction, the smaller n, as it has been shown by using the

Gutzwiller technique for an s band [24, 25]. If we adopt the estimate made in [16] of the

electronic correlation effect (reduction of the cohesive energy of 0.07 eV and 0.03 eV for

n

=

13 and 55 respectively) we are led to corrected covalent energies of 0.13 eV and 0.25 eV for n = 13 and 55 respectively. Now we see that the covalent energy is smaller than

Ev,dw for n

=

13. This means that for n smaller than about 15 there appears a transition from the metallic to the Van der Waals bonding.

3. Ionized Hg§ .

3.1 LARGE SIZES, n > 50.

-

For relatively large sizes (n > 50) in all the models the electronic structure is assumed to be of the same nature as in the bulk (hybridization between

the s and p bands). Then the metallic ionization energy is given by a formula like equation (1)

where the 3 e2/8 R term expresses the potential energy of the Hg’ aggregate which is globally

charged. However equation (1) which is valid for alkalis [5, 6] must be modified before being

applied to our problem.

(6)

Indeed, in the case of alkalis, the one-electron Fermi energy of the s band is, for any n, close to the es atomic energy. In particular when n --+ oo, the one-electron Fermi energy (of

a cc lattice) tends towards es (symmetric band). This is not the case for Hgn since the one-

electron Fermi energy, EF, increases with n (see Tab. I). We have to take this into account by writing, instead of (1)

In table 1 we give the values obtained for I(R) with R

=

ro n1/3,

,

ro

=

1.51 Â (average

distance in the bulk) and the experimental values when available. We see that the agreement between experiment and theory is rather good for n - 55. On the contrary there is a difference between experiment and the theoretical result for n - 13. This discrepancy was expected since

in this range of n values the metal-like structure is no longer valid.

Table I . - Size dependence of the cohesive energy (Ee), the Fermi level (El,) (origin of the energies : the 6s atomic level) and the ionization potentiel (I) in Hgn. The values obtained in

reference [16] are given in parentheses.

3.2 SMALL SIZES, n 10. The fact that neutral Hgn exhibits a changing in the chemical bond prompts an intriguing question concerning the charge localization in ionized

H ’ . Indeed, for n 10, as the structure of the neutral aggregate is Van der Waals-like, one

may think that as in Van der Waals ionized aggregates (Xe’ [26, 27]) the positive hole is

localized on one (or on a few) central site.

In this range of n values the s band is completely separated from the p band. For example,

our calculation of section 2 gives a gap of about 2 eV for n

=

13. Therefore we think that it is

justified to describe mercury ionized aggregates by only considering the external 6s levels.

This assumption, which can also be applied to aggregates of other divalent elements such as

Mg [21], is well justified in the case of Hg where the energy difference between 6s and 6p

levels is large (- 6, 1 eV) in comparison to 2.7 eV for Be or Mg [28].

The main features of our model for Hg’ aggregates with n , 8 are the following :

the wave function is an expansion on a set of atomic determinants. Each determinant describes n - 1 electrons, see figure 1 for the case n

=

3 ;

-

for each n value, we study two kinds of shapes. Firstly, geometries where all the distances are equal and short di = dl. In these shapes the positive charge is shared between all the sites which are almost equivalent : the hole is delocalized. Secondly, geometries with two possible distances dl and d2 > dl where, as we will see, the hole is mainly localized on a

limited number of sites (2 or 3) ;

(7)

2678

-

the parameters used in this calculation are given in table II. They have the following physical meanings.

Fig. 1.

-

The three determinant wave functions for Hg3+ with equal interatomic distances.

Table II.

-

Values o f the parameters (in eV) used for small ionized cluster, their meaning is given in the text.

. Ev and U are atomic energies.

They are determined from the atomic single and double ionization energies.

2022 3 (d) is the transfer energy between two adjacent sites. We will use /3 (d2) = 1 2 (dl).

2022 ER is the repulsion energy between two neutral atoms with distance dl. We assume that ER is negligible for d

=

d2.

. Epol is the attractive polarization energy between an ionized and a neutral site with distance dl. We also assume that Epol is negligible for d

=

d2.

Our assumptions on the variations with d of /3, ER and Epol are justified by the fact that

/3 decreases more slowly than the other two parameters when d increases. The variations with distance of /3 and ER have already been mentioned in section 1.

By using these parameters, we obtain the analytical energies given table III for

n

=

2, 3 and 4.

The ionization energy I is deduced from these expressions by using :

(8)

Table III. - Energies of the localized and delocalized states versus the size.

where E(Hgn) is obtained from equation (7) by considering compact shapes and EYdW(2)

=

0.07 eV. By fitting 9 to the experimental results and by keeping the stablest shapes

we have determined 0, ER and Epol (Tab. II) then we have used these parameters for calculating the other ionization potential given in table IV. For each n value we give in figure 2 the stablest shape of Hg’ in the application of (9). We also give in table IV the weight

of the localized configurations, i.e. the probability of localization of the hole on the central sites drawn as circles in figure 2.

Fig. 2.

-

The stablest shapes of Hgn . We have drawn as circles the sites which are distant from each

other by dl distances. The hole is more localized on these sites than on others (solid circles).

(9)

2680

Let us not that the value obtained for the parameter {3 is of the same order of magnitude as

the value used in the one-electron calculation of the previous sections. There is not a strict

equality because in this part of the study we give a more precise description of the quantuum

structure by considering energies ER and Epol (repulsive and attractive respectively) which

were not included before.

In table IV we can see that for n

>

4 the agreement with experiment is good and that the

charge tends to localize on some central sites. This characteristic is similar to the

Xen behaviour [26].

Table IV. Size dependence o f the experimental and theoretical ionization potential, and the weight o f the localized configuration for Hg£ . (The labels o f the shape are defined in Fig. 2).

Conclusion.

In this paper we have studied neutral Hgn aggregates for which a transition from a metal-like

bonding to a Van der Waals bonding is obtained at about nc - 15. We have applied our model

to Hg+ aggregates with a relatively large number of atoms (n 50 ) and have obtained a

satisfactory agreement with experiment.

For smaller Hg’ (n:5 10 ) a method is developed where only 6s orbitals are taken into account. This method allows one to study the possible localization of the excess positive elementary charge for n . 4. In a way similar to Xe’ aggregates the excess charge tends to

localise on a part of the sites. Our study is limited to n

=

7 where the previous effect still appears.

The model can be extended to Hgn + aggregates where a precise theoretical description has already been published for Hg2 + [30]. Our work on this question is in progress.

References

[1] HENSEL F. and FRANK E. V., Rev. Mod. Phys. 40 (1968) 697.

[2] IONEZAWA F. and MARTINO F., Solid State Commun. 18 (1976) 1471.

[3] LINKE R., MORAN-LOPEZ J. L. and BENNEMANN K. H., Phys. Rev. B 27 (1983) 7348.

[4] WOODS D. M., Phys. Rev. Lett. 46 (1981) 749.

(10)

[5] LINDSAY D. M., WANG T., GEORGE T. F., J. Chem. Phys. 86 (1987) 3500.

[6] KAPPES M. M., SCHÄR M., RADI P., SCHUMACHER, E., J. Chem. Phys. 84 (1986) 1863.

[7] LEOPOLD D. G., Ho J., LINEBERGER W. C., J. Chem. Phys. 86 (1987) 1715.

[8] RADEMANN K., KAISER B., EVEN U. and HENSEL F., Phys. Rev. Lett. 59 (1987) 2319.

[9] MONTANO P. A., SHENOY G. K., ALP E. E., SCHULZE W., URBAN J., Phys. Rev. Lett. 19 (1986)

2076.

[10] JIANG P., JONA F., MARCUS P. M., Phys. Rev. B 36 (1987) 6336.

[11] TOMANEK D., MUKHERJEE S., BENNEMANN K. H., Phys. Rev. B 28 (1983) 665.

[12] JOYES P., Les agrégats inorganiques élémentaires (Editions de Physique) to be published.

[13] LINN S. H., LIAO C. L., BROM, J. M., NG C. Y., Chem. Phys. Lett. 105 (1984) 645.

[14] BRIANT C. L., BURTON J. J., Surf. Sci. 51 (1975) 345.

[15] BRÉCHIGNAC C., BROMER M., CAHUZAC Ph., DELACRETAZ G. , LABASTIE P., WÖSTE L., Chem.

Phys. Lett. 120 (1985) 559 ; Phys. Rev. Lett. 60 (1988) 275.

[16] PASTOR G. M., STAMPFLI P. and BENNEMANN K. H., Europhys. Lett. 7 (1988) 419.

[17] JOYES P. and VAN DE WALLE J., J. Phys. B 18 (1985) 3805.

[18] JOYES P., VAN DE WALLE J. and COLLIEX C., Ultramicroscopy 20 (1986) 65.

[19] EKARDT W., GOLDENFELD I. , PENZAR Z. , SCHULZE W. and WINTER B., to be published.

[20] BRÉCHIGNAC C., BROYER M. , CAHUZAC Ph., DELACRETAZ G. , LABASTIE P. , WÖSTE L., Chem.

Phys. Lett. 118 (1985) 174.

[21] DURANT G., DAUDEY J. P. and MALRIEU J. P., J. Phys. France 47 (1986) 1335.

[22] KHANNA S. N. , REUSE F. and BUTTET J., Phys. Rev. Lett. 61 (1988) 535.

[23] BENNEMANN K. H. and PASTOR G., Microclusters, Springer Ser. Mater. Sci., Ed. S. Suyano, Y.

Nishina, S. Ohnishi, 4 (1986) 54.

[24] JOYES P., Phys. Rev. B 26 (1982) 6307.

[25] JOYES P., Phys. Rev. B 28 (1984) 4006.

[26] AMAROUCHE M., DURAND G. and MALRIEU J. P., J. Chem. Phys. 88 (1988) 1010.

[27] AMAROUCHE M., DURAND G. and MALRIEU J. P., Z. Phys. D. 8 (1988) 289.

[28] JONES R. O., J. Chem. Phys. 71 (1979) 1300.

[29] CABAUD B., HOAREAU A. and MELINON P., J. Phys. D 13 (1980) 1831.

[30] NEISLER R. P. and PITZER K. S. , J. Phys. Chem. 91 (1987) 1084.

Références

Documents relatifs

In this work, the objective is to develop a two-scale kinetic theory description of concentrated suspensions including the modelling of nanotube aggregates and their

Important changes in legislation relating to health called for the role of the nurses and of secondary medical staff to be defined unambiguous( y.. We took advice from

Due to the discrete nature of the charge carriers, the current presents some fluctuations (noise) known as shot noise , which we aim to characterize here (not to be confused with

The irreducible representations can be indexed by the elements of the free monoid N s , and their tensor products are given by formulae which remind the Clebsch–Gordan rules

In this case, the laser pump excites the titanium film and, through a thermoelastic process, leads to the generation of picosecond acoustic pulses that propagate within the copper

represents an irrational number.. Hence, large partial quotients yield good rational approximations by truncating the continued fraction expansion just before the given

Based on DFT methodologies this is quantified for the cohesive energies, relative charge transfers, bulk modules, electronic structures and bonding properties for

· 2CBrClH 2 co-crystal in which both molecular species are ori- entationally ordered undergoes a phase transition around room temperature to a high-temperature hexagonal phase without