• Aucun résultat trouvé

Electron correlations and chemical bonds in aggregates of monovalent elements. A poor man's description

N/A
N/A
Protected

Academic year: 2021

Partager "Electron correlations and chemical bonds in aggregates of monovalent elements. A poor man's description"

Copied!
15
0
0

Texte intégral

(1)

HAL Id: jpa-00210427

https://hal.archives-ouvertes.fr/jpa-00210427

Submitted on 1 Jan 1987

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.

Electron correlations and chemical bonds in aggregates of monovalent elements. A poor man’s description

J. Friedel

To cite this version:

J. Friedel. Electron correlations and chemical bonds in aggregates of monovalent elements. A poor man’s description. Journal de Physique, 1987, 48 (1), pp.93-106. �10.1051/jphys:0198700480109300�.

�jpa-00210427�

(2)

Electron correlations and chemical bonds

in aggregates of monovalent elements. A poor man’s description

J. Friedel

Université Paris-Sud, Laboratoire de Physique des Solides Bât 510, 91405 Orsay, Cedex France

(Requ le 16 juin 1986, accepté le 26 septembre 1986)

Résumé.

2014

A l’aide d’un simple modèle de Hubbard dans la limite des faibles corrélations, nous analysons le rôle des interactions et des corrélations électroniques sur la nature des liaisons dans les agrégats monovalents. Nous discutons la validité de cette approximation simple.

Abstract.

2014

Using a simple Hubbard model in the weak correlation limit, we analyse the role of electron interactions and correlations on the nature of bonding in monovalent aggregates. We discuss the validity of this simple approximation.

Classification

Physics Abstracts

31.20P

-

71.00

Introduction.

The Hubbard model [1, 2] has been developed to analyse cases where electron correlations are too strong

to be treated only as a small correction in the effective

potential of otherwise delocalised independent elec-

trons, and for aggregates too large to be treated by

more exact configurations interaction techniques.

This treatment of electron correlations is compen- sated by a more approximate description of wave functions, based on atomic averages. Thus the wave functions are based on a LCAO approximation:

where a ) are a reduced set of atomic orbitals on sites

a. The kinetic energies are expressed in terms of

intersite transfer integrals

and electron interactions by atomic averages :

etc.

In the simplest Hubbard model, one neglects non orthogonality of atomic functions on different sites

one restricts t to nearest neighbours a, b and only

consider intraatomic Coulomb interactions U.

In the limit of I U /t I small [3], a development will give for the total energy of an aggregate

E=A+Bt+CU+DU2/t+0(U3/t2). (6)

The term A is essentially of atomic origin [4]. Bt is the bonding due to electron delocalisation. CU is the Coulomb repulsive term computed in the Hartree Fock limit for uncorrelated delocalised electrons [5] : their

random distribution over lattice sites, with only the

Pauli exclusion principle on atomic orbitals, leads to charge fluctuations responsible for this term (Fig. 1).

Finally the term DUZ/t corresponds to the first devia-

tion from perfect Hartree Fock disorder [3] : the

interaction U tends to decrease the charge fluctuations

by a correlation of interactomic electronic jumps. In this perturbation limit, it is the first term due to electron

Fig. 1.

-

Charge fluctuations in the Hartree Fock limit.

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:0198700480109300

(3)

correlations ; it is clearly of a less local nature than the Kohn Sham correction [1, 6]. One can finally remark

that the terms in Bt and DU2/t are necessarily negative

while CU is positive.

The purpose of this paper is to test this limit for

problems of small aggregates where the physical proper- ties are generally known and more refined theoretical

treatments at hand. We shall see that this « poor man’s treatment » gives simple and reasonable results, but

that the exact values of the parameters involved, and notably U, are not easy to extract from first principles.

1. Dimers versus close packed aggregates.

We first recall and slightly correct earlier results on the relative stability of a close packed aggregate and of a

collection of dimers [7], thus setting the model and its

approximations.

1.1 DIMERS.

-

For a dimer of hydrogen atoms with

valence electrons in Is orbitals I a ) and l b ) , we can

write for the singlet and triplet states respectively

where

or

Because the Hamiltonian is preserved when the two nuclei a, b are permuted, 141) must be symmetrical or antisymmetrical in this permutation. There are therefore two types of singlet states :

while

With the Hamiltonian written with obvious notations

the energies of these various states are given by

where S, t, U and u are given by equations (2) to (5)

and

If one neglects S, j, v, the preceeding equations give

where

is the repulsive Coulomb energy between the two

neutral atoms on a and b.

In the simplest Hubbard approximation, one neglects

(4)

u with respect to U. The most stable (singlet) state has

then the energy

In the limit I U It I 4, a development in U/t is valid.

It gives, with t 0 for s orbitals, an energy per atom

The two first terms in this development are of atomic origin ; the following ones are due to electron de- localisation ; the terms linear in t and U are obtained in the delocalized Hartree Fock limit, and the term in

U2/t is the first deviation from delocalized Hartree

Fock, due to electron correlations. They could have been computed directly.

More precisely, in the delocalized Hartree Fock

limit, the electrons are assumed to move independently

from each atom (A = 03BC ) , with opposite spins in the singlet state.

The term in t gives the increase in kinetic energy per atom, and thus per electron, due to this delocalization ;

due to the virial theorem, this is also, for the Coulomb

forces involved, the corresponsing lowering to total

energy.

The term in 1 U takes into account the intraatomic 4

interactions between electrons of opposite spins in the

same limit, where the probability that a charge fluctu-

ation of two electrons on the same site is, per site,

1 1 1 2X2 4’

The term in U2/32 t can be considered as due to a

virtual excitation of bonding states into antibonding

ones (Fig. 2), owing to the V12 interactions. Single

excitations correspond to matrix elements such as

with

they vanish by symmetry. The matrix element for double excitation is

the excitation energy - 4 t leads then, in the limit of small U’s, to a contribution per alom

The equilibrium distance r2 between the two protons

can be deduced from the variation with this distance of R and t. We can assume

JOURNAL DE PHYSIQUE. -T. 48, N. 1, JANVIER 1987

Fig. 2.

-

Singlet and triplet states of the H2 molecule.

where direct estimates show that [7]

Then writing

one finds, by minimizing E2 :

with

One sees that the correlation term in U2 both stabilizes the dimer and increases the bond length.

1.2 CLOSE PACKED AGGREGATES. - A development

similar to (24) can be given for a (macroscopic)

aggregate where each of the N atoms is surrounded by

n equidistant nearest neighbours [3, 7].

In the Hubbard limit, the energy pir atom reads

where the effective Is band width is related to its second moment and can be written (n, number of neighbours)

These expressions have numerical factors deduced

(5)

from a rectangular band, with constant density of

states ; but they can be shown to remain good approxi-

mations for more realistic band shapes.’

More precisely, the band term -1 w 4 comes from the

fact that the broadening of the 1s band is essentially symmetrical in energy, because its first moment is zero ; with one electron per atom, the lower half of the band is occupied, with an average energy per electron

4 w. The increase of w with BIn is equivalent to computing the kinetic energy assuming a Brownian

motion of the electron from site to site.

The term in 1 U is the same as for the dimer because 4

the charge fluctuations are the same per atom for dimers and for close packed aggregates.

The term in U2/16 w takes into account virtual collisions of electrons from the occupied into the unoccupied part of the 1s band, with an average energy of excitation w and a probability of charge fluctuations in the initial and final states 1 B2 x 1 - 1. 2 16

Using the variations (25), (26) of t and R with the interatomic distance r, one finally obtains

where w is given by (32) and

1.3 DIMERS VERSUS CLOSE PACKED AGGREGATES. - The preceding analysis can be directly extended to the

valence s electrons of all monovalent elements, with similar values of q/p.

Equations (30) and (35) show that, for small values

of I U/t2 , close packed aggregates have interatomic distances somewhat larger than dimers ; but the corre-

lation correction expands dimers more than close

packed aggregates.

A similar balance is observed for the energy, when

comparing equations (24) to (31) or (29) to (33).

Qualitatively, the Hartree Fock term due to delocalisa- tion stabilises the close packed aggregates, while the correlation term favours dimers. As the repulsive term

also favours the dimers, the critical value of I U/t2 I

above which dimers are preferred and below which

close packed aggregates are more stable somewhat depends on the value of q/p. Table I gives this critical

Table I.

-

Critical value of I Uc/t2 versus q/p.

value I Uc/ t2 l. It is seen that, for the reasonable value

q/p =.--!z 2.5, UC/t2 =-= 2. This is much smaller than the

critical value 4 above which the development in I U /t I used here is no longer valid.

This simple minded analysis is coherent with observa- tion for monovalent elements if U Uc for all alkali

metals, while U:> Uc for H, where the stable state is

made of dimers weakly coupled by dispersion forces neglected here.

Indeed dimers seem to be prefered for all elements

with strong electronegativity (H, 0, N, halogens) while

close packed aggregates are prefered in all other

elements. It is obvious that a similar analysis can be

extended in principle to other types of valence electrons

( p, d ) with similar qualitative conclusions [7]. This

remark therefore suggests that U increases with elec-

tronegativity, a point we will come back to at the end of

this paper.

2. Dimers versus trimers in monovalent elements.

The preceding discussion can be extended to comparing

dimers with small close packed aggregates, and more

specifically with trimers.

The situation for monovalent trimers is indeed well studied theoretically by interaction configurations techniques [8, 9] ; a number of experimental results fit with their conclusions, notably from astrophysical data [10], and from studies in vapours [11] and jets [12].

These studies conclude again to a difference of be- haviour between H and alkali metals.

Close packed H3 trimers are unstable ; only the

situation of a dimer H2 weakly coupled at long range with an H atom is stable. Contrarywise H+ are stable as

close packed equilateral triangles.

For alkali metals, neutral trimers exist as isocele

triangles, with a summit angle somewhat larger than 7T 1i°

The question to be discussed here is how much of this behaviour can be described using the same simple

Hubbard model.

2.1 H TRIMERS.

-

We first compare, for H trimers,

three simple geometries (Fig. 3).

In an independent electron molecular orbitals Hartree

approximation, we can use wave functions of the type

For configurations of figures 3a, a, b, c, the one

electron energy states satisfy respectively

(6)

Hence

E = Eo + 2 t and E = Eo - t (doubly degenerate) for the equilateral triangle.

E = Eo ± t J2 and E = Eo for the linear chain.

E = Eo ± t and E = Eo for the dimer + atom.

These spectra are pictured figure 4.

In the corresponding Hartree Fock approximation, the ground state of H3 can be written as (cf. Appendix A)

Fig. 3.

-

Three simple configurations for trimers : a) equila-

teral triangles ; b) linear chains ; c) dimer + atom.

Fig. 4.

-

One electron energy spectrum in the limit U = 0 for the configurations of figure 3.

where a refers to the molecular orbital with lowest

energy, f3 to the next molecular orbital state and

I ± ) refer to the sign of the spin. With that approxi- mation, the total energies of H3 are :

to be compared with (Eq. (23))

For H3’ , one obtains in a similar way

to be compared with

It has been assumed here that each atom remains neutral and thus that each pair of atoms repels with the

same energy R (r) as for a dimer. This point, which is obviously not exact for H3’ , is discussed in § 2.4 below.

The computation, detailed in Appendix A, is cohe-

rent with the fact that, in such a linear combination of atomic orbitals approximation, the exchange hole of

the Hartree Fock approximation (38) is of atomic dimension. As a result the term in U in the energy of reduces to the sum of Coulomb interactions between « + ) and I a - )

states, and between I (3 + ) and I a - ) states, of antiparallel spins, while the term in U in the energy of

H3 is that between I a ) states of opposite spins.

Because of the asymmetry of the trimers considered,

the charge distribution of each one-electron state is not

uniform, as for a dimer or a close packed aggregate. As

a result, the linear term in U now depends on the

geometry of the trimer.

The correlation terms in U2/t can be computed treating in perturbation the configuration interaction.

For H3, this can be written formally

(7)

where the summation extends over excited states

I .p f ) , of energy Ef 3 A somewhat long but straightfor-

ward computation (Appendix B) gives Y = U2/xt with

x = 18 for the equilateral triangle and x = 17.3 for the chain.

Inclusion of this term gives finally the following developments for H3 :

with

Development (40), valid for H3 , shows that the

bonding term in t due to delocalization strongly stabil-

ises the equilateral triangle. This is due to the large bonding energy of the most bonding state in this configuration (Fig. 4a). The Coulomb term in U also

stabilises the equilateral triangle. As the terms in

U2/t, which are not given here, have similar coef- ficients, it is seen that the equilateral triangle configura-

tion is preferred for any value of I Ult I small enough

for development (40) to be valid.

Development (42), valid for H3, shows that the

delocalization term in t tends to stabilize the equilateral triangle, but the terms in U and in U2/t tend to stabilize

the dimer. There should therefore be a critical ratio of

I U / t I below which the equilateral triangle is preferred

and above which the dimer is preferred. Because the

repulsive term in R favours the dimer, this critical value

should increase with the ratio q/p. More precisely, minimizing the energy with respect to r gives

with

and

Comparison with (29), (30) shows that for

q/p = 2.5, the critical value I U/t2 I is again very near

to 2. Thus the Hubbard model leads to the same

conclusions for the relative stability of a closepacked

trimer versus a dimer as for the relative stability of closepacked macroscopic aggregates versus dimers, i.e.

large values of I U/t2 I lead to dimers, with the same

critical value of I V / t2 I. It is therefore coherent with this conclusion to think again that U > Uc for H and

U Uc for alkali metals, with l Ve/ t2 1== 2.

2.2 NEUTRAL ALKALI TRIMERS.

-

For alkali trimers where I U/t2 I is expected to be small and the equila-

teral triangle to be more stable than the dimer or the linear chain, inspection of figure 4a shows that, in the

limit U -+ 0, one electron occupies a degenerate state

( EJ3 = Ey ). The trimer is then expected to lower its

energy still further by a spontaneous distortion.

In the simplest such Jahn Teller distortion, the equilateral triangle becomes an isocele triangle with a

summit angle 0 # I. 3 Treating 0 - -7jr- 3 as a perturbation

to second order, one can show that the equilibrium

state of lowest energy corresponds to 03B8 > 3’ 3 as

schematically shown in figure 5. The critical angle depends on the ratio q/p and on the value of I U/t2 I ’

However for q/p = 2.5 and vanishingly small values of U, 60° Oc 90°; and it is not very sensitive to U when, as assumed here, I U/t2 2. This is in agree-

ment with more exact computations and the few experimental evidence gathered so far.

More precisely, let t’ and t" be the transfer integrals

involved in the isocele triangle pictured in figure 6. If

we write

the summit angle 8 is larger than 03C0 if T" > T’.

Fig. 5.

-

Jahn Teller distortions from the equilateral triangle.

(8)

Fig. 6.

-

Isocele triangle.

The corresponding one-electron wave functions satisfy, for U = 0, the equation

with roots

One note that

the lifting of the degeneracy is qualitatively that

of figure 5.

Thus the total Hartree energy is, in the limit U - 0,

To simplify the discussion, we shall consider two

types of distortion from the initial equilateral triangle.

We shall consider them as independent, and thus neglect their weak coupling :

a) a pure dilatation of the equilateral triangle, with

T’ = T ". The equilibrium condition for z’ = T" = 0 leads to

D) a snear like moae With

A development of (47) then gives (cf. Appendix C)

If we write

we have

with

With (45) and (49), this gives to second order term in T’:

Putting this expression in (50), we see that, as in any Jahn Teller effect, the linear term :t tT’ lifts the

degeneracy, decreasing the energy in a symmetrical

way whatever the sign of T’, thus of T" - T’. The terms in T’2, which are altogether positive ( t 0 ) , stabilise

the values of T’, thus the total distortion to a finite

r’2 t

value. Finally the term in ± 12 12 stabilizes preferentially

the a’ branch where the third electron is on the E J3

state, thus the distortion with T’ 0 or, according to

The equilibrium distortion is given by minimizing

Ea3’. This gives

With q/p == 2.5, this gives

(9)

Hence

As pr =-= 1.4 for hydrogen [13] and is larger for other elements, it is seen that 0 deviates from 60° by a rather

small amount.

Finally the correction U in the Hartree Fock approxi-

mation can be deduced from computing (Appendix A)

where ’lLaa and ’lLal3 are given by (A.6), (A.13).

From (46), one deduces

and

Hence

with, for the shear mode

one obtains

The Hartree Fock energy then becomes, for the

a’ branch

Minimizing with respect to T’ shows that the correction in U slightly increase the Jahn Teller distortion, which becomes

The effect is not large in the limit I Ult 2 con-

sidered here.

2.3 INTERATOMIC COULOMB TERMS.

-

We have im-

plicitly assumed so far that, in all configurations, atoms keep neutral. This is generally the case, and when not

so, does not change our conclusions.

First, in neutral aggregates with alternate structures

i.e. where closed circuits with only even number of

interatomic jumps can be drawn, a half filled band leads to all atoms being neutral. Particular cases in our

discussion are the dimer and the linear chain. For such cases, if

are the one electron wave functions, they satisfy

To each solution 10 (E) ) with coefficients an corres-

ponds a solution I qi’ ( - E ) ) with coefficients

( -1 ) n an where n is the parity of the site n. The sum

of the contributions an 2 of the lower and upper hand halves of the band on the electronic density on each site

n are thus equal. As for a full (non degenerate s) band,

the electronic density for a full band gives 2 electrons

on each site, a half filled band as here has necessarily

one electron on each atomic site. This can be checked

directly.

Still in neutral aggregates, the use of Bloch states for the equilateral triangle shows that again each atom has

the same electronic charge, that is neutral. This is not so for the choice of stationary wave functions (A3)

made in Appendix A. However because of the degener-

acy of the partially occupied states /3, y with energy

Ep, y = Eo + t, one should make an average between the contributions of the 8 state (as assumed in the computation) and of the y state ; these add to a total of

one electron per site, while the terms in U and

U2 computed in appendices A and B are independent

of the choice of basis.

Now the Jahn Teller distortion lifts the degeneracy

between the /3 and y states, and creates ionic charges on

the atoms of the neutral trimer. These are different for

0 > 3 but not very large. Thus in the limit of weak distortions where the contributions of T’ and T " are

1 1 1

negligible, the ionic charges are ± 3’ 3 + 1, =+= 6 6 6 on atoms

a, b, c for e ’ "3. The effect of these charges has been

3

taken into account in computing the intra-atomic term in U, but not the interatomic terms in R. This correction

changes the overall stability of the (distorted) trimer by

a small term

r =1.5 atomic units, this is of the order of 1.5 eV.

Because of its symmetry for 0 / §, its dependence on r

will not alter the symmetry of the energy splitting for

small distortions nor the fundamental asymmetry for

0 a § 3 shown in figure 5. But this dependence on r will

somewhat shift the equilibrium distortion.

(10)

Finally in H’ , the ionic charge is equally distributed in the equilateral triangle. The ensuing interatomic Coulomb repulsion is largest for that configuration,

where it introduces a positive correction e = 6 eV. e2 3r

From the estimates below for H, this is about

U. Thus the equilateral triangle is still preferred for H

in all the range I U /t 14 where the perturbation

scheme used here is valid.

3. Value of U ; validity of the Hubbard model.

If we take the Hubbard model at its face value, U can

be computed from (3). For 1s wave functions of H, this gives U = 17 eV. Applying (24) to compute the cohe- sive energy - ( E2 - 2 Eo) of H2, which is 4.4 eV, we find I U / t2 1== 2. This is nearly of the right order of magnitude, as the preceding discussion suggests

UI t2 > 2.

It also justifies the use of a perturbation scheme valid

for I UI t2 4.

There are however a number of difficulties.

a) First, as pointed out in (21), if the interatomic interaction u is kept in computing R, it should be kept everywhere ; this amount to replace U by U - u for

dimers. This is physically obvious, as the energy involved in electron correlations is there the difference in energy between two electrons on the same or on

different atoms. In more extended aggregates, the role of u would be more complex ; however, for a half-filled band and weak correlations, one expects the charge

fluctuations to be mostly with very short wavelength,

thus U - u to replace also U in the most effective

charge fluctuations. However a direct computation of u

gives e 2 e2 with r = 1.4 atomic units for H2, thus u == 19

r

eV, even larger than U !

b) The qualitative solution of this difficulty can be

taken as due to the fact that we have used too restricted

a basis of wave functions.

Indeed virtual intra-atomic excitations are known to

reduce both u and U.

a. For two electrons sitting on different atoms, the

possibility of the electrons to be virtually excited by

their interaction from their ground 1s states to excited (mostly 2p) states creates for H2 a van der Waals

attraction which is very effective because of the small-

ness of the distance between protons [141. For two

alkali atoms, where the van der Waals interaction is

smaller, one can take into account a reduction in u due

to the polarization of their inner shells. Finally for aggregates larger than dimers, the correlation between valence electrons screens out their long range interac-

tions ; for large enough aggregates, this effect can be

taken into account by using, in computing u, a dielectric

constant which takes into account this screening by

valence electrons.

In the same way, intratomic correlations, which can

be described in terms of virtual excitations (mostly

Is --+ 2s for H) lower the value of U for two electrons

sitting on the same atom [15]. The ensuing effective

value Uc can be directly measured on free atoms M by

the reaction

which involves the energy

where I is the first ionization potential and A the

electron affinity. For H, this is Ue = 12.8 eV ( U = 17 eV).

b. In an effective Hubbard model, one is then led to

use U, =1- A instead of U and still neglect u (or use a

small value ue Ue). Figure 7 shows that Uc varies in a

rather regular way in the periodic table, in much the

same way as Pauling’s electronegativity. This is not surprising, as this is defined as proportional to

I + A and I >> A. It is thus clear from this table that

electronegative elements (rare gases excluded) have large values of UC. This is in qualitative agreement with the observation that dimers are more stable than

closepacked aggregates for the most electronegative elements, namely H, 0, N+ and halogens [13] ; more precisely, H has a value of Ue more than 2 to 3 times

those of alkali metals, while the corresponding transfer integrals t are not expected to vary so strongly. One can

also note in figure 7 that of all elements forming closepacked aggregates, the divalent metals have the

largest values of Ue’ suggesting that for these metals

electron correlations play a stronger role. This is also in

qualitative agreement with recent theoretical analysis [16].

c) This effective Hubbard model with Ue =- I - A

and ue - 0 however does not fit quantitatively any

Fig. 7.

-

Values of Ue in the periodic table. In parenthesis

values where A is inknown. T transitional elements ; R rare

earth elements. Underlined: elements where dimers are

preferred.

(11)

experiment. This is well known for transitional metals

[3, 17], where the use of such an effective Hubbard model requires the use of U = 2 to 2.5 eV, and not Ue = 6-8 eV as pictured figure 7. In the same way the

use of Ue from figure 7 and of the experimental

cohesive energies [18] would lead to I U/t2 2 for H

and > 2 for alkali metals, in complete contradiction with the conclusion of this paper.

In fact, the argument which led to the replacement of

U by Ue =1- A is incomplete [19]. This is sketched in Appendix D :

-

Because of the electron transfers produced by t,

the inter and intraatomic correlations which lower u

and U are less effective than assumed.

-

On the contrary, the virtual excitations produce supplementary channels of interatomic transfers, thus increase the effective transfer integrals.

The various corrections, together with an account of

the nonorthogonality of the atomic functions (i.e.

S = (a I b) ::F 0) are taken into account in more refined

computations for very small aggregates.

If one wants to keep a simple Hubbard model which

can be extended to larger aggregates, one is therefore bound to use an effective value to U which, if probably varying qualitatively like Ue = I - A, is not quantita- tively given by it.

A related study for 4-atomic Hubbard clusters has been done independently by Y. Ishii and S. Sugano [20]. It leads to similar conclusions.

Appendix A

TRIMERS IN THE HARTREE FOCK LIMIT.

a) One electron states.

-

Let I a ) , l f3 ), l y ) be

the three electron states linear, combinations (36) of

atomic states la), (6), )c) :

satisfying equations such as

Figure 4 gives the energy spectra Ea, E J3’ Ey for the

three simple geometry of the equilateral triangle, the

linear chain and the dimer + atom. The corresponding

values of the coefficients can be taken as

for the equilateral triangle,

for the linear chain.

In fact, because of their energy degenerescence,

states /3 and y are equally populated in the equilateral triangle, leading to the neutrality of each atom here as

for the linear chain. The choice (A.3) is the most

natural for studying the Jahn Teller effect discussed

below.

b) Ground state.

-

For H3’, the corresponding Hartree

Fock ground state is the singlet state built with the one

electron state « a of lowest energy Ea.

For H3, The Hartree Fock ground state l «/Ii) is

built with the one electron states I a ) and l 03B2 ) of

lower energies Ea and EB.

*

It has the form (38) : it is

thus antisymmetrical and is easily checked to be an eigenfunction of Sz = 03A3siz z and of

with eigenvalues respectively equal

c) Coulomb and exchange interactions.

-

For H3’, the quantity to compute is

Using (A.1), this is

Thus X = 1 3 U for the equilateral triangle and X = 3 for 8

the linear chain.

For H3, the quantity to compute is

Developing (38) as

(12)

we can write

where

It is easy to show that

with U.. given by (A.5) and

Using (A.1) one obtains (A.6) and

Finally

with (A.3), (A.4), this gives X = _C/ for the equilateral triangle and X = 5 U for the linear chain.

8

Appendix B

CORRELATION TERM IN H3.

In computing the second order perturbation term Y given by (41), the excited states can be taken as

follows :

One easily deduces, with for instance

(13)

Using (A.1), (A.3), (A.4), the expressions of table

B.I can be computed for the two configurations consid-

ered

Tables B.II and B.III then give the coefficients of Y for the two configurations considered.

Table B. I.

-

Values of integrals for two configurations.

Table B. II.

-

Coefficient of Y for equilateral triangle.

Table B. III.

-

Coefficients of Y for the linear chain.

From these, one deduces that Y = U2/xt, where

x = 18 for the equilateral triangle, and - = 21 =

17.3 for the linear chain

Appendix C

JAHN TELLER EFFECT FOR NEUTRAL TRIMERS.

-

Let

t’, t ", t"’ be the transfer integrals for the most general triangle (Fig. 8). The one electron energies satisfy, for

U = 0, the equation

Fig. 8.

-

Most general triangular shape for a trimer.

If we write

a development in successive powers of the T’s gives for

the fundamental state Ea and the initially excited states

E /3, ’Y the new energies

(14)

where

The developments of the text are obtained for

T fft = T It.

Appendix D

ROLE OF ELECTRON DELOCALIZATION ON THE INTRA- ATOMIC CORRELATIONS.

-

We show in this appendix that, in the delocalised limit, the decrease of U due to

virtual intra-atomic correlations is less than for an

isolated ion, but that this is compensated by an increase

in transfer integrals.

Taking hydrogen as an example, we can write the

lower singlet state in terms of one electron wave function I a(i) ), I b(j)) where now a ) con-

tains not only the ground state a’ ) (ls) but an

excited state a" ) (say 2s) :

The total energy for a H- ion then reads

with

where Eo + 4 is the energy of the excited 2s state.

Also

where

Also

A development in small a " gives

Minimizing E with respect to a " gives

and

For a dimer in the limit U t, we can write in the Hubbard approximation

where

where

A development in small a " gives Ed (a " ) which can

be similarly minimized with respect to a ". The complex

result simplifies somewhat if we neglect differences

between t, t’, t ". Then at equilibrium.

(15)

This formula shows clearly that if we put

as deduced from the energy of the ion, the energy of the dimer is not simply obtained by replacing U by

Ue in the formula

Indeed, in the limit of I Ult l small, and if one neglectec t with respect to V, one would obtain for the correction to the term U

In 2" a term equal to

1 (U- Ue) , thus half that deduced from the ion. This reduction is due to the delocalization of the electrons

produced by t :0 0, which lowers the effect of intra- atomic correlations.

However a larger correction, working in opposite direction, comes from the replacement of V by

V + t in the corrective term. This comes from the increase in the average transfer integral due to the

virtual excitations, indeed from the possibility of cross

Is # 2s interatomic transfers. This correction of t (and

not of U) complicates the development in U and t made

in the text. The approximate formulae developed here

assume anyway I tld 1 but I U/t 4. As the promotion energy 4 is less than the ionization potential I, which as pointed out in the text is of order U, the discussion of this Appendix is only valid for a limited

range of values of I U/t I .

Note added in proof : the discussion in 2.2 can be

simplified and extended by minimizing directly E3 given by (50) with respect to r’ and r". The same

qualitative conclusions hold, with an explicit coupling

between dilation and shear modes.

References

[1] HUBBARD, J., Proc. Roy. Soc. A 276 (1963) 238.

[2] GUTZWILLER, M. C., Phys. Rev. 134 (1964) 4923.

[3] FRIEDEL, J. and SAYERS, C. M., J. Physique 38 (1977)

697.

[4] FRIEDEL, J., Physics of Metals I 2014 Electrons Ed. J. M.

Ziman, (Cambridge University Press) 1969.

[5] FRIEDEL, J., J. Phys. Rad. 16 (1955) 829.

[6] KOHN, W. and SHAM, L. J., Phys. Rev. A 140 (1965)

1133.

[7] FRIEDEL, J., Physics and Chemistry of Electrons and Ions

in Condensed Matter, J. V. Acrivos, N. F. Mott and A. D. Yoffe ed. NATO ASI Series, Serie C

Mathematical and Physical Sciences, 130 (1984) (D.

Deidel Bordrecht Hall).

[8] HIRSCHFELDER, J. O., EYRING, H. and ROSEN, N., J.

Chem. Phys. 4 (1936) 121, cf. also TENNYSON, J.

Chem. Phys. Lett. 86 (1982) 181.

[9] FLAD, J., IGEL-MANN, G., PREUN, H. and STOLL, H., Surf. Sci. 156 (1985) 399 ;

MARTINS, J. L., BUTTET, J. and CAR, R., Bul. Buns.

Phys. Chem. 88 (1984) 239 ;

ANDREONI, W. and MARTINS, J. L., Surf. Sci. 156 (1985) 635 ;

RAO, B. X. and JENA, P., Phys. Rev. B 32 (1985) 2058.

[10] CARNEY, G. D., Com. J. Phys. 62 (1984) 187.

[11] OKA, T., Phys. Rev. Lett. 45 (1980) 531.

[12] KNIGHT, W. D., Helv. Phys. Acta 56 (1983) 521.

[13] KITTEL, C., Introduction to Solid State Physics 3d Ed.

(John Wiley, New York) 1966.

[14] SCHIFF, L. I., Quantum Mechanics (McGraw Hill New York) 1955.

[15] FRIEDEL, J., Physica 109, and 110B (1982) 1421.

[16] DURAND, G., DAUDEY, J. P. and MALRIEU, J. P., J.

Physique (under press).

[17] HERRING, C., Magnetism, G. T. Rado and H. Suhl ed.

(Academic Press New York) 1966 4.

[18] HERZBERG, G., Molecular Spectra and Molecular Struc- ture. I. Spectra ofdiatomic molecules. 2d Ed. (Van Nostrans-Reinhold, Princeton New Jersey) 1950.

[19] FRIEDEL, J. and NOGUERA, C., Intern. J. Quantum

Chem. 23 (1983) 1209.

[20] ISHII, Y. and SUGANO, S., J. Phys. Soc. Japan, 53 (1984)

3895.

Références

Documents relatifs

This indicates that the spin polarization of the electron states in the solid has the dominating influence as opposed to the effects of electron-magnon scattering in the manner

of d-wave superconductivity in trie presence of a van Hove Singularity in trie density of states, scanning trie parameter space with respect ta doping.. Now we want to look in

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

Length and velocity curves were calculated, and ground state correlations were included by the diagram of figure lb which brought the length and velocity

- We study the importance of electron correlations and magnetism in the stability of Re dimers on W(110) surface.. It is found that the magnetism of monomers is

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

of the phonon dispersion of the highest optical branch at the zone boundary and the square of its EPC by almost 80%; 共iii兲 GW reproduces the experimental phonon dispersion near K,

To first order in U and J, the energy difference between a ferromagnetic state of atomic moment and the corresponding paramagnetic state (y = 0). with the same