• Aucun résultat trouvé

Sum rules for inhomogeneous Coulomb fluids, and ideal conductor boundary conditions

N/A
N/A
Protected

Academic year: 2021

Partager "Sum rules for inhomogeneous Coulomb fluids, and ideal conductor boundary conditions"

Copied!
5
0
0

Texte intégral

(1)

HAL Id: jpa-00210218

https://hal.archives-ouvertes.fr/jpa-00210218

Submitted on 1 Jan 1986

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.

Sum rules for inhomogeneous Coulomb fluids, and ideal conductor boundary conditions

B. Jancovici

To cite this version:

B. Jancovici. Sum rules for inhomogeneous Coulomb fluids, and ideal conductor boundary conditions.

Journal de Physique, 1986, 47 (3), pp.389-392. �10.1051/jphys:01986004703038900�. �jpa-00210218�

(2)

Sum rules for inhomogeneous Coulomb fluids,

and ideal conductor boundary conditions

B. Jancovici

Laboratoire de Physique Théorique et Hautes Energies (*),

Université de Paris-Sud, Bâtiment 211, 91405 Orsay Cedex, France

(Reçu le 30 août 1985, accepti le 12 novembre 1985)

Résumé.

2014

On montre que la fonction de corrélation de charge d’un fluide coulombien inhomogène satisfait

une règle de somme qui met en jeu les moments électriques multipolaires d’ordre quelconque; cette règle de somme

est une généralisation de plus de la règle du second moment de Stillinger et Lovett. On donne une application

à un fluide limité par une paroi parfaitement conductrice; on montre comment les potentiels chimiques contrôlent

à la fois les propriétés de volume et de surface.

Abstract

2014

The charge correlation function of an inhomogeneous Coulomb fluid is shown to obey a sum rule involving electrical multipole moments of arbitrary order; this sum rule is a further generalization of the Stillinger-

Lovett second-moment condition. An application is given for a fluid bounded by an ideal conductor wall; it is

shown how the chemical potentials control both the bulk and surface properties.

Classification Physics Abstracts

05.20

-

50.00

-

82.45

1. Introduction.

The present paper is about classical equilibrium

statistical mechanics of Coulomb fluids (plasmas, electrolytes...). The fluid is modelled as a system of

particles interacting through Coulomb’s law plus

some short-range interaction (a hard core repulsion

for instance).

Coulomb fluids exhibit the fundamental property of screening, with the consequence that their charge

correlation functions obey a variety of sum rules.

If p(r) is the microscopic charge density at a point r,

the static charge structure function is defined as

where > denotes an equilibrium average. In the

simplest case of a classical homogeneous fluid (for

a homogeneous fluid ( p(r) > = 0, the truncation T in (1.1) may be omitted, and also S depends only on

the distance r’

-

r 1), S is known to obey the Stil- linger-Lovett [1] rules which give its zeroth and second moments :

and

(where fl is the inverse temperature). In the more complicated case of a classical inhomogeneous fluid,

i.e. a fluid which is not invariant by translations and rotations (for instance because of the presence of walls or of external charges), (1.2) is unchanged and (1. 3) is to be replaced, as shown by Carnie and Chan [2, 3], by

(with an arbitrary choice of the origin); one readily

shows [4] that (1.4) reduces to (1.3) in the homoge-

neous case.

The first Stillinger-Lovett rule (1.2) is a simple

consequence of the following statement : any charge q

introduced into the fluid is screened by a polariza-

tion cloud of charge - q, and therefore the total

excess charge (i.e. the charge q plus the charge of its polarization cloud) vanishes. When applied to the particles of the fluid themselves, this statement leads

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:01986004703038900

(3)

390

to (1.2). However, a more general property of the

excess charge distribution is that all its electrical

moments vanish [5, 6], provided the correlations in the fluid have good decay properties (faster than algebraic); these decay properties will be assumed to

hold for the systems to be considered here. There-

fore, (1.2) can be generalized into

where t’ stands for the polar angles of r’ and Y,m(i’)

is any spherical harmonic. Of course, (1.5) is trivial

for I =1= 0 by rotational invariance in the homogeneous

case, but it becomes a non-trivial statement for an

inhomogeneous fluid.

In the present paper, it will be shown, in section 2, that the generalization of the first Stillinger-Lovett

rule (1.2) into (1. 5) has an analog for the second

Stillinger-Lovett rule, the Camie and Chan form (1. 4)

of which can be generalized into

with spherical harmonics of arbitrary order.

It should be emphasized that (1. 5) and (1. ) do not

hold in general for a Coulomb fluid bounded by an insulating wall (1), because the charge correlation functions near the wall do not have the necessary

good decay properties (in the directions parallel to

a plane wall, the charge structure function (1.1) decays only as the inverse cube of the distance [7, 8]).

However, good decay properties for the charge corre-

lations are expected in a Coulomb fluid bounded

by a conducting wall (a more general conjecture [9]

is that the charge correlations have a faster-than-

algebraic decay whenever every region of the fluid

is completely surrounded by conducting media).

In section 3, equation (1.6) will be used, in its dipolar

form (I = I’ = 1), for a Coulomb fluid bounded by a plane ideal conductor wall, and it will lead to the

simple sum rule

where the origin is on the wall and z is the component of r normal to the wall.

Finally, in *section 4, the thermodynamics of a

Coulomb fluid bounded by ideal conductor walls will be discussed, in the grand-canonical ensemble.

It will be shown how the chemical potentials control

both the bulk and surface properties of the fluid, and

how (1. 7) can be retrieved as a consequence of this

more general study.

(’) They however hold for I

=

l’

=

0.

2. General sum rule.

We derive the general sum rule (1.6) by a simple

extension of the linear response argument which can be used [3] for proving (1.4).

At the (arbitrarily chosen) origin, we put a small point test multipole ql’m which creates an interaction Hamiltonian

The fluid responds by a change of the charge density

at r

The general screening rule [5, 6] discussed in the

Introduction states that the multipole moment (1’, m’)

of bp(r) cancels ql,m, while the other multipole moments

of bp(r) vanish. Therefore

Using (2 . 2) and (2 .1 ) in (2. 3), we obtain at once (1.6).

3. Plane ideal conductor wall

We now apply (1.6) to a semi-infinite Coulomb fluid,

bounded by a plane wall z = 0; the fluid occupies the region z > 0. The wall is assumed to be an ideal con-

ductor. This might be a model for an electrode in an electrolyte. A well-defined model is obtained by assum- ing that the interface between the fluid and the wall is impermeable to the particles of the fluid.

We want to write a sum rule involving the charge

structure function of the fluid alone, i.e. S(r, r’) is

defined by (1.1) where p(r) is the charge density of

the fluid particles. There are also charges induced at

the surface of the ideal conductor : they will be dealt with by the method of images. To be more specific,

we put a small test point dipole in the fluid, at the origin, on the wall (more precisely, infinitely close

to the origin, on the fluid side); the dipole is directed

along the z axis, normal to the wall. We can carry the argument of section 2, for I’ = 1, m’ = 0, with how-

ever the modification that the dipole now has an image and the total interaction Hamiltonian with the fluid is easily found to be twice as large as (2. 1).

Bringing this modification into (1.6), we find

Since S depends on the components of r and r’ parallel

to the wall only through their difference, after a sim- ple manipulation we can rewrite (3.1) in the form (1.7).

It should be noted that the integral (1. 7) is not

absolutely convergent, and the integrations upon z

and z’ cannot be permuted. The first integration

(4)

dr’ S(r, r’) gives in general a non-vanishing result, notwithstanding (1.2), because here there are induced

charges on the conducting wall, which are not includ-

ed in the definition of S(r, r’).

The sum rule (1. 7), which holds for an ideal conduc- tor wall, has some similarity with a rule which holds for an insulating wall [10, 3] :

where also the integrations cannot be permuted Although (1.7) and (3.2) apply to different cases, both can be rewritten in the same form

Indeed, for an insulating wall, z - z’ can be replaced by - z’ because of (1.2), and for an ideal conductor

wall, z - z’ can be replaced by z because of (1. 5) for

l = 1, m = 0.

4. Thermodynamics of a Coulomb fluid with ideal conductor boundary conditions.

We now proceed to study the thermodynamics of a

Coulomb fluid bounded by conducting walls, from

a more general point of view, in the grand-canonical

formalism. Inside a box of volume S2, with conducting

walls kept at potential zero, we consider a fluid made

of M different species of particles. A particle of species i

has a charge ei, there are Ni such particles, and their

chemical potential is pi. It is convenient to introduce the M-dimensional vectors e = (el, ..., em), N = (N t, ..., N M)’ J1 = (Pt, ..., gm).

Coulomb fluids in equilibrium are neutral in the bulk. This has been shown, for the case of systems bounded by insulating walls, in the elaborate mathe- matical investigation of Lieb and Lebowitz [11].

Since we do not expect the bulk properties to depend

on the nature of the walls, we shall admit without proof that bulk neutrality also holds for conducting

walls. Thus, whatever the chemical potentials may be,

in the thermodynamic limit the bulk densities ni = lim Ni >/Q must be such that n e = 0.

U-+ 00

There are only M - 1 independent densities, and

M - 1 chemical potentials should be sufficient for

controlling these densities. The relevant part of the M-component vector J1 can be exhibited by splitting J1

as

where M" is the component of J1 along e :

If N is split, in the same way, as

where

the factor exp(pp - N) in the grand-canonical for-

malism becomes exp[P(u’ . N’ + M" N")]. In the thermodynamic limit N" >/Q always vanishes (this

is the bulk neutrality requirement), the grand-cano-

nical distribution function has a sharp peak at N" = 0,

and therefore the chemical potential p" is that part of p

which is irrelevant for controlling the bulk densities which depend only on the (M - I)-component vec-

tor p’.

.

However, in the case of conducting walls, p" does

have a physical role : it controls the surface structure of the fluid, as it will now be shown. Again, we assume

that the interface between the fluid and the ideal conductor walls is impermeable to the particles of the

fluid The region of the fluid close to the interface is known as the electrical double layer. Although the

bulk fluid must be neutral, the electrical double

layer may carry some surface charge density. An important related quantity is the potential difference

across the electrical double layer : while the conduct-

ing wall is assumed to be at potential zero, in general

the interior of the fluid is at some different electro- static potential 0, and, given the bulk composition of

the fluid and the detail of the interparticle and particle-

wall interactions, the surface structure is determined by 0, or alternatively by p", because there is a very

simple relation between 0 and Jl". Indeed, if 0 is

varied by do, at fixed bulk composition (i.e. fixed p’),

the reversible work for bringing a particle of spe-

cies i from infinity into the fluid varies by ei do (this

is the variation of the work necessary for crossing the

electrical double layer). Thus

and, from the definition (4.2) of p,",

This result (4.6) sustains our claim that p" controls

the surface structure.

Let us now derive a sum rule relating the density

and a correlation function, following a method of

Blum et al. [12]. Let hi(r) be the microscopic number density of species i at r. The grand-canonical for-

malism gives at once

(5)

392

On the other hand,

Using (4. 5) and (4. 7) in (4. 8), we obtain the sum rule

This sum rule has been previously conjectured by

Forrester [13]. Note that it can be considered as the

analog for a conducting wall of the dipole sum rule

of Blum et al. [10] which holds at a plane charged insulating wall creating an electrical field E :

Finally, in the limiting case of a semi-infinite fluid bounded by a plane ideal conductor wall, we can

retrieve the sum rule (1.7) as a consequence of (4. 9).

The potential difference across the electrical double

layer is related by simple electrostatics to the dipole

moment of the charge density :

Since

combining (4. 9) and (4 .11 ), we obtain (1.7).

5. Concluding remarks.

Although we have pointed out analogies between conducting and insulating walls, there is an impor-

tant difference : for a Coulomb fluid bounded by insulating walls, Il" is a totally irrelevant parameter

which controls neither the bulk nor the surface pro-

perties. This can be easily seen in the simple case of a

fluid in an insulating uncharged spherical box of

radius R. If the fluid had a surface charge density a,

the potential at the surface would be 4 n u R, becom- ing infinite in the thermodynamic limit R - oo, cr fixed. The reversible work pic for bringing a particle

of species i from infinity into the fluid would have an

infinite part ei 4 naR. In other words, in the grand-

canonical formalism, with finite chemical potentials,

it is not possible to control the surface charge density,

which will remain zero, like the bulk charge density.

It should also be noted, from the derivation in section 4 of the conducting wall sum rules (4.9) and (1. 7), that they remain valid in quantum mechanics without any change and that they also hold for the time-displaced correlation functions, i.e. when the

densities at r and r’ are taken at different times 0 and t.

This is to be contrasted with the insulating wall sum

rule (3.2), which can be extended to quantum- mechanical and time-displaced correlations only for

a one-component plasma, and in a modified form [14]

where wp is the plasma frequency.

6. Acknowledgments.

I am indebted to L. Blum, J. L. Lebowitz, and M. L.

Rosinberg for stimulating discussions. Part of this work was done at Rutgers University and at the University of Puerto Rico; I thank J. L. Lebowitz and L. Blum respectively for their kind hospitality

at these institutions.

References

[1] STILLINGER, F. H. and LOVETT, R., J. Chem. Phys. 49 (1968) 1991.

[2] CARNIE, S. L. and CHAN, D. Y. C., Chem. Phys. Lett.

77 (1981) 437.

[3] CARNIE, S. L., J. Chem. Phys. 78 (1983) 2742.

[4] OUTHWAITE, C. W., Chem. Phys. Lett. 24 (1974) 73.

[5] GRUBER, Ch., LEBOWITZ, J. L. and MARTIN, Ph. A.,

J. Chem. Phys. 75 (1981) 944.

[6] BLUM, L., GRUBER, C., LEBOWITZ, J. L. and MARTIN, P., Phys. Rev. Lett. 48 (1982) 1769.

[7] USENKO, A. S. and YAKIMENKO, I. P., Soviet Tech.

Phys. Lett. 5 (1979) 549.

[8] JANCOVICI, B., J. Stat. Phys. 28 (1982) 43.

[9] JANCOVICI, B., J. Stat. Phys. 34 (1984) 803.

[10] BLUM, L., HENDERSON, D., LEBOWITZ, J. L., GRUBER,

Ch. and MARTIN, Ph. A., J. Chem. Phys. 75 (1981)

5974.

[11] LIEB, E. H. and LEBOWITZ, J. L., Adv. Math. 9 (1972)

316.

[12] BLUM, L., LEBOWITZ, J. L. and HENDERSON, D., unpu- blished.

[13] FORRESTER, P. J., J. Phys. A 18 (1985) 1419.

[14] JANCOVICI, B., LEBOWITZ, J. L. and MARTIN, Ph. A.,

J. Stat. Phys. 41 (1985) 941.

Références

Documents relatifs

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

(2006) have demonstrated that when a cylindrical scatterer is illuminated by random plane P and S waves at equipartition in 2-D, the cross-correlation of the wavefield allows

Then, denoting by l(a) the number of occurrences of the pattern “01” in the binary expansion of a, we give the asymptotic behavior of this probability distribution as l(a) goes

On radial Schr¨ odinger operators with a Coulomb potential: General boundary conditions.. Jan

As usual, denote by ϕ(n) the Euler totient function and by [t] the integral part of real t.. Our result is

These numbers were introduced in 1944 (cf. [1]) by Alaoglu and Erdős who did not know that, earlier, in a manuscript not yet published, Ramanujan already defined these numbers

As usual, denote by ϕ(n) the Euler totient function and by [t] the integral part of real t.. Our result is

Up to normalization, Macdonald polynomials can alternatively be defined as the unique eigenfunctions of certain linear difference operators acting on the space of all