PHYSICA

ELSEVIER PhysicaA 241 (1997) 216~222

On the Clausius-Mossotti-Lorenz-Lorentz
formula
A.P. Vinogradov*
Scientific Center for Applied Problems in Electrodynamics, Russian Academy of Sciences,
Izhorskaja 13/19, Moscow, 127412, Russian Federation

Abstract
The difference in application of the Clausius-Mossotti-Lorenz-Lorentz formula for bulk and
thin systems is discussed, including size-effects.

PACS." 72.15.Gd; 05.70.Jk

Keywords: Local field; Permittivity; Permeability

1. Introduction

A commonly accepted way to describe the interaction of electromagnetic field with

material is to employ the macroscopic Maxwell equations together with constitutive
relations and boundary conditions. The most popular form of all these relations is
based on the introduction of the permittivity t~ and permeability /~, for we obtain a
common language for both static and alternating fields. That is why the problem of
permittivity and permeability averaging in composite materials challenged scientists for
a long time [1]. Though this problem has a long history [2,3], there are still unclear
moments. First of all, it concerns a local field which acts on a separate inclusion. The
local field is not identical with both the external field Eext and the mean field (E).
The field Eloca I is the sum of the external field and the fields produced by all other
inclusions or atoms. The most widespread approximation for Elocaj is the Clausius-
Mossotti-Lorenz-Lorentz (CMLL) formula [4-6]:

where P is the mean polarization of the material. It is the local field that is responsible
for the difference in properties of bounded systems near the surfaces from the bulk

* Fax: (095)4852633; e-mail: vin@eldyn.msk.ru.

0378-4371/97/$17.00 Copyright @ 1997 Elsevier Science B.V. All rights reserved

PII S0378-4371 (97)00085-X
 A.P. VinooradovlPhysica A 241 (1997) 216-222 217

properties of the same material. This difference is due to the fact that each surface
atom or inclusion are predominately illuminated from one side only.
In the following we consider as how the local field varies with geometry of the
problem.

2. The CMLL formula for bulk material

The well-known Lorentz derivation of the CMLL formula begets many questions
(see discussion in [7]). Following [5], consider Lorentz's speculation:
"Let our reference molecule A be surrounded by an imaginary sphere of such an
extent that beyond it, the dielectric can be treated as a continuum. If the molecules
inside this sphere were removed while the polarization outside remains frozen, the field
acting on A would stem from two sources: from the free charges at the electrodes of
the plate capacitor (Eext), and from the free ends of the dipole chains that line the
cavity walls (E2). Actually there are molecules inside the sphere and they are so near
to A that their individual positions and shapes have to be considered. This adds an
additional contribution E3 to the local field Elocal ":

Elocal = Eext + E2 + E3 • (2)

Naturally, the next step is a calculation of the fields E2, E3. These fields seem to be
expressed through the sums of the same dipole fields:

Edipole = 3 ( p " r ) r p
r5 r3 (3)

(see [6]). The field E3 comes about from summation of Edipole over the Lorentz sphere,
whereas the field E2 is due to the summation over the rest volume of the insulator
[8]. The results of the summation seem to depend on the radius of the Lorentz sphere.
Traditionally, for uniform distribution of dipoles the results are taken as independent of
the Lorentz sphere radius (see [5,8]). What is more puzzling is that they are different:

E2 = 4 n P E3 = 0. (4)

As a consequence we obtain an expression resembling the CMLL formula:

Elocal = Eext + 47tP (5)

which links Elocal with Eext (see [5,8]). Expression (5) differs from (1). To prove the
equivalence of (1) and (5) some authors assume that 1

Eext = ( E ) . (6)

1R.E. Collin [8] even tries to prove (6) for a regular array of dipoles. Actually, he implicitly introduces (6)
as a starting point of his proof when applying periodic conditions on electrical potential inside the lattice in
the form: ~(r) = q~(r + a) - Eext • a.
218 A.P. Vinogradov/Physica A 241 (1997) 216 222

It was shown in [18], instead of (6), at least for non-magnetic media, the following
expression should be used: Eext q,,o = (E)q,,o +41zPq, ~ +4rt~o[q × [q x Pq,,,,]]/(q2 _ 0 ) 2 / c 2 ).
Thus, (5) cannot be directly brought to (1). Actually, (4) does not come from the
summation of dipole fields. It results from consideration of the dipoles placed outside
the Lorentz sphere as a uniform media. Such a consideration inevitably brings into the
problem the mean field (E). This field is equal neither to Eext nor to the sum of dipole
fields (3). To average the field of a dipole we should consider not only its "external"
part (3) but also its "internal" part, assuming the dipole to be a dielectric sphere 2
Simple calculation shows that the averaged "extemal" part equals zero, whereas the
"internal" part produces the CMLL result (1)3
The averaging procedure may be conceived as blowing up each dipole until the
dipoles occupy the whole volume. The interior of such a blown-up dipole should be
filled by medium with the mean polarization. Thus, the field inside this blown-up filled
dipole should be equal to the averaged electric field (E), whereas the field inside the
empty blown-up dipole is equal to Elocal. The value of Elocal depends on the shape of
the blown-up dipole, and the factor ~ in (1) is an approximation, corresponding to a
spherical shape of the blown-up dipole [2,3].
It remains to be seen as what is wrong with (5). The key moment here is that the
integral or sum one tries to evaluate, is a non-absolutely convergent integral

I = / Edipole dr. (7)

Integral (7) does not exist in rigorous mathematical sense because the integral of the
absolute value IEdipoll over the whole volume diverges logarithmically (see [9]). Thus,
depending on the method of summation one can obtain different results. Indeed, instead
of treating integral (7) the corresponding sum

(2m 2 -- n2 _ S 2 )
S= Z (m 2 + n 2 + s 2)5/2
n, m,s=-- oc~

where the term with m = n = s = 0 is omitted, and changing the order of summation
we arrive at different results:

o~ (m 2 _ n2) ~ (m 2 _ s2)
S= Z (m2 + n 2 + s 2)5/2 + (m 2 + n 2 + s 2)5/2 = 0 (8)
n, re,s= -- oo n, m, s=-- o<z

2The "internal" part corresponds to the delta function term introduced in (3) by Jackson [6] to arrive at
results (4).
3 It is clear as why in plasma one should not distinguish local and mean fields. Indeed, in plasma the charges
o f opposite sign do not form bonded, point-like systems, hence, there is no reason to distinguish " e x t e m a | "
and "internal" parts of the fields (see also [7]).
 A.P. Vinogradov/PhysicaA 241 (1997) 216-222 219

and using the Poisson summation formula we obtain (see [8])

~1 ~ (m2-n2-s2) ~ (2m2 - s 2 )
S=4 ~ + (m-~7~ n g 7~ 72)5/2 + 2 E (m 2 + n 2)5/2
m=--oo n=-- 1,s=l,m=--oo n=l,m=--oc
oo
+2 E (2m2- s2)/(m2 + s2)5/2 m 4.808 - 64rt2K0(27t) ~ 4.228,
s=l,m=--c~
where Ko(x) is the modified Bessel function.
When deducing (1) one avoids the direct summation of non-existing sum since, the
same, absolutely divergent at the infinity portion, is a constituent of both the average
and local fields [3]. Indeed, as r'~r

[ 1 / ((p'rt)r t p ) d3r/]
(E)+Eext = E * + ~ ~ 3 r, 5 r, 3
r L Vcell(r)
[ 3 ( P : -r ) r P]
= > E . -1- ~ L r5 r3 '
r
where the summation is going over the dipoles located outside the cell, E . is the field
induced by charges and currents in the cell. Thus, the long-length terms cancel out
in (1). As a result, Eq. (1) has a "local" origin. It appears as a result of averaging
a dipole field over a volume of blown-up dipole reducing the Lorentz sphere to the
Onsager one [19]. Obviously, the relation between the external and local fields as well
as between the external and average fields depends on the shape of the sample, keeping
(1) to be valid.
To illustrate this abstract speculation let us consider a sphere placed in a uniform
external field Eext. Inside the sphere there are dipoles producing the average polar-
ization P. We purposely choose the sphere to obtain the uniform polarization. The
mean macroscopic field {E) inside the sphere is equal to E e x t - ( ~ ) P . The field
Elocal inside a small hollow sphere, placed for simplicity in the center of the primary
sphere, is equal to the sum of the external field Eext and the fields of the dipoles
contained between the spheres. Due to symmetry of the problem the contribution (E2)
of all the other dipoles is equal to zero; and we come to the usual relation (1):
Elocal--=Eext -1-E2 = Eext = <E) q- ('~)P.
For alternating in time fields we should take into account the radiative self-field
(2q/3c2)d3r/dt3 (see [3,10]): Elocal= (E) + (~- )P - (2/3)d3p/d3t.

3. The local field in bounded system

As we deal with bounded, systems the inclusions near the surface are illuminated by
a different local field from those in the bulk, since they have neighborhood inclusions
on one side. If we deal with a boundary between media, the inclusions near the interface
will be affected by both of the materials. Hence, the values of the effective parameters
220 A.P. Vinogradov/PhysicaA 241 (1997) 216-222

ascribed to the interface layer have to depend upon the properties of both the media.
This is the reason why we cannot correctly introduce e and /~ for thin systems [11].
Instead of the standard Maxwell boundary conditions (continuity of the tangential field
components outside and inside the material) we should use transition conditions relating
the fields on both sample sides with mean polarization of the sample (see for review
[12]). To find the mean polarization we should know a local field. Direct calculation
of the local field reduces to the summation of "external" fields of the dipoles. The
result is a non-zero value (2).
Sivukhin [13] was the first to apply (2) to description of optical properties of thin
films. Recently his result was reproduced in [8,15-17]:

Ex2 q- Ex3 = ~i ( px 2X2 - y2i 2xi Yi ~ (9)

Z + = =

Ey2 q- Ey3 = ~i. (py2y2~x2 +3px .~) .(y) 1 (v)D

= Yll ~ P y = 7fi "y, (10)

Ez2 q- Ez3 = - ~i" P~ = 7~I)P~, 11)

where (xi, yi,gi) are the coordinates of the ith dipole on the plane (z = 0). The Poisson
summation formula yields the following results (see [8], p. 760):

?{]) = -27(,~ ) 7~]) = 7{y) = 4.8 - 32~z2K0(27t) ~ 4.51. (12)

The planar value of the 7{11) is near the CMLL result 4n/3 .~ 4.188.
In spite of the apparent proximity of the results for bulk and thin system they have
different origin. For bulk system the term E2 = ( ~ ) P appears due to the averaging of
the "internal" dipole field, whereas for thin system the term E2 = ?P comes from the
summation of the "external" dipole fields.
Mahan and Obermair [14] have generalized the Sivukhin result, considering n layers
of dipoles. The authors have shown [14] that, only inclusions in one or two layers near
the surface see local field which differs from that of the bulk. These authors suggested
(see also [13, 15-17]) the following expression for the local field in jth layer:
//

Elocalj = Eext -+- Z 7jiei' (13)

i~l
where Pi is a polarization of ith layer and 7ji are real-valued factors. Mahan and
OG
Obermair [14] (see also [13]) shown that 7 = ~i=_oc~)l,i =4n/3. Formally, in the
limit n ~ oo expression (13) gives the wrong result (5). To clear up the question, we
observe the transition from (13) to (5), focusing on a bounded system with a finite
number of dipoles. As sum (11) includes a finite number of terms, it exists.
Direct calculation shows that it is enough to consider the system of size L > 20 in
order that the value of 7 ceases to depend on L. In Fig. 1 we can see the distribution
of 70x) over a finite plate. Only a small number of dipoles near the boundaries has
 A.P. Vinooradov/PhysicaA 241 (1997) 216-222 221

5.0 .~m

4.5

4.0

3.fi

3.(

2.!
2.
1,
30

1.u
-60
-40 o

-20 0 20 -60

X 40 60

Fig. 1. The distribution of the 7-factor over the 400 × 400-system.

~(lx) differing from (9). What will happen if we add one more layer? We should take
into account more terms in the expression (13). In the general case of n layers the
field in kth layer depends upon the difference m = ( k - l): 7kl = f ( l k - l],n).
In Fig. 2 we can see, that the more the ratio l/n differs from zero, the more 7
distinguishes from a constant. For the center point of the cube (l/n= 1) we obtain
7 = 0 in accordance with the previous speculations.

4. Conclusions

We have seen that the formula (13) for the local field, can only be approximately em-
ployed for bounded thick systems, being valid for a planar infinite system with 7 = h i .
The cooperative action of dipoles producing the local field is not so short-range as
was considered previously [15,16]: to obtain a rapid attenuation of the interaction, one
should consider infinitely large planes. Moreover, in continuous limit these planes cease
interact. Indeed, each layer being regarded as a continuous medium can be imagined
as an oblate ellipsoid with two semi-axes tending to infinity, whereas the third one
tends to zero. The field in such an ellipsoid tends to the value of the external field:
(E) = Eext. Since the uniform layer does not disturb the external field, all other layers
are situated in the same external field. Hence, the consideration of a bulk system as a
sum of thin uniform layers implies that ( E ) = Eext and leads to unphysical result (5).
Deviation of 7 from 711 in thin system is due to discrete nature of the material.
The disturbance of the electric field Eext is observed in the vicinity of the layer at a
distance of the order of lattice period. As a consequence the value of the y-factor for
thin system depends on the type of the lattice too [16].
222 A.P. Vinogradov/Physica A 241 (1997) 216~22

3.0 T N=50

2.5

2.0 J
N=100

1.5

1.0

0.5
N=M=2~ diN
0.0 i I , I , I L i ,
0.0 0.2 0.4 0.6 0.8 1.0

Fig. 2. The dependence of the 9-factor for the dipole placed on the normal to the layers. The normal passes
through the center of the sample. The sample is system, d is the distance between the point of observation
and the center of the sample.

References

[1] R. Landauer, in: Proc. of AlP Conf. No. 40, eds. J.C. Garland and D.B. Tanner (American Institute of
Physics, New York, 1978) p. 2.
[2] G. Russakoff, Am. J. Phys. 38 (1970) 1188.
[3] J.E. Sipe and J. Van Kranendonk, Can. J. Phys. 53 (1975) 2095.
[4] M. Born and E. Wolf, Principles of Optics (Pergamon Press, Oxford, 1993).
[5] A.R. von Hippel, Dielectrics and Waves (Artech House Books, London, 1995).
[6] J.D. Jackson, Classical Electrodynamics, 2nd Ed. (Wiley, Singapore, 1990).
[7] V.L. Ginzburg, Propagation of Electromagnetic Waves in Plasma (Pergamon Press, New York, 1964).
[8] R.E. Collin, Field Theory of Guided Waves, 2nd Ed. (IEEE Press Series on Electromagnetic Waves,
New York, 1993).
[9] L. Sehwarz, Methodes Mathematiques pour les Scieces Physiques (VI: Hermann, Paris 1961 ).
[10] L. Mandelstam, Phys. Zs. 8 (1907) 608; Ann d. Phys. 23 (4) (1907) 626.
[11] A.P. Vinogradov, I.G. Busarov, O.P. Posudnevsky and V.E. Romanenko, J. Phys.: Condens. Matter
6 (1974) 4351.
[12] T.B.A Senior and J.L. Volakis, Approximate boundary conditions in electromagnetics, IEE
Electromagnetic Waves Series 41 (The Institution of Electrical Engineers, London, 1995).
[13] D.V. Sivukhin, Zh. Eksp. Teor. Fiz. 18 (1948) 976; 21 (1951) 367; 30 (1956) 374 (in Russia).
[14] G.D. Mahan and G. Obermair, Phys. Rev. 183 (1969) 834.
[15] M.R. Philpott, J. Chem. Phys. 60 (1974) 1410, 2520.
[16] M.R. Philpott and G.D. Mahan, J. Chem. Phys. 59 (1973) 445.
[17] C.A. Grimes and D.M. Grimes, IEEE Transaetions of Magnetics 29 (1993) 4092.
[18] I.M. Ryasanov, Electrodynamics of Condensed Matter (Nauka, Moscow, 1984) (in Russian).
[19] L. Onzager, J. Am. Chem. Soc. 58 (1936) 1486.