PH Y SICA L

VOLUME 22, NUMBER

REVIEW B

10

15

NO

VEMBER 1980

Absorption spectrum of clusters of spheres from the general solution of Maxwell's equations.
The long-wavelength limit
J, M. Gerardy and M. Ausloos
Institut de Physique, B5, Universite de Liege, B-4000 Sart Tilman/Liege
(Received 17 December 1979)

1, Belgium

A theory is presented to calculate the infrared absorption spectrum of N dielectric spheres of arbitrary sizes,
embedded in a dielectric matrix. The Laplace equation is solved exactly for this system, and general solutions take
into account interactions at all multipolar orders. Phase factors are introduced when retardation effects become
important. Absorption 'and absorbed power spectra are calculated for two (different or identical) spheres and for an
infinite linear chain. of identical spheres for different field directions. The theory is applied to MgO spheres. It is
found that the Frohlich mode for a single sphere splits into several resonant modes, and that quadrupolar-order
interactions and incident field direction have drastic effects on the predicted infrared spectrum. Experimental
consequences regarding damping-factor effects and the position of the modes are discussed.

A large variety of particle sizes, compositions,

and environments are met in nature. Among experimental techniques chosen for particle-size
measurement and distribution, light scattering
allows particle monitoring in the 0.08 to 100 pm
range. This encompasses an important part of
all "industrial particles.
Much work has already been presented on the light scattered by
powders, with emphasis on metallic powders. '
Investigative techniques have made less use of
infrared and Raman absorption spectrum analysis, although several approximate theories have
been recently presented in particular about infrared absorption in dielectric powders'
(as
well as in metals' ). We will report elsewhere
on the general solution of Maxwell's equation for
a powder, and consider here the particular "longwavelength limit.
We follow the usually ac-

"

"

"

ceptable statistical assumption,

ly symmetric particles. '
Without loss of generality, let
case of an ionic powder in an as
matrix. The infrared absorption

viz. , of spherical-

us consider the
yet unspecified
amplitude A(&u)
(which is proportional to the absorption coefficient
in the case of a dilute layer of particles) is expressed in a continuum theory" in terms of the
imaginary part of the directionally averaged particle susceptibility, (y(&u)), by
A

(&u)

= (2Mv/vNq') ((u/(u r) Im(y ((u)),

where M is the reduced mass, v the particle volume (=4mB'/3), q the ionic charge, N the number
of ion pairs in the microcrystal, (d~ the transverse-optical phonon frequency of the (bulk) ma-

terial, and
3 Mv

(u'

(o((u+iy) '

(2)
22

is the dipolar strength,

'"

the pocu
larization mode frequency, and y an appropriate
(small) damping factor. Although these formulas
are typical of "rigid-ion diatomic lattice"
substitution of (2) into (I) indicates that
theory,
the relevant unknown quantities are the u 's.
Clearly, most theories have attempted to obtain
such values by "brutal" lattice-dynamical calculations. It is of interest to ask whether such
values, and hence the absorption spectrum, can
be obtained in a continuum approximation, given
measured macroscopic quantities characterizing
the powder and its matrix, i.e. , the respective
dielectric constant e; (ur) and
frequency-dependent
e, (&u). In particular, one can examine the effect
of particle interaction in a continuum theory as
a function of interparticle distance and radius.
In so doing one can check how the single dipolar-active infrared mode for an isolated sphere,
the Frohlich mode, which obeys

where C~

I. INTRODUCTION

'"

"

e, (&u, ) + 2e, (u), ) = 0,

is modified.
We obtain here general formulas giving a/l inor not, modes (in principle for
any 2l polar order) for an arbitrary distribution
of spherical particles in the long-wavelength
limit, while also taking into account retardation
effects. The N-sphere case without retardation
is treated in Sec. II. Retardation effects are included in Sec. III. One could perform calculations
for arbitrary 2I, polar order, but for comparison
with other work, illustrations of the results are
given only for a linear chain of identical spheres
and for two spheres of different sizes (Sec. IV).
Quadrupolar effects are included in the former

frared-active,

case.
In Sec. IV,

we present the absorption spectrum

for different cases (spheres of equal or unequal
4950

1980 The American Physical Society

ABSORPTION SPECTRUM OF CLUSTERS OF SPHERES. . .

22

4951

radii, two spheres and the linear chain, with or

without damping factor and with or without inclusion of quadrupolar effects). In Sec. V, a
brief comparison is made with other work, and
some emphasis is placed on new results presented

here.
II. INTERACTION BETWEEN SPHERICAL

IONIC

PARTICLES
A. Single-sphere case

Consider an isolated sphere of radius

8 and

characteristic dielectric constant ~; in a uniform

isotropic matrix of dielectric constant e, . The
electric field E; in the sphere can be written in
terms of the potential Q,. as
E( = Vfg

p,

(4)

-z, g a.H.(r, e, q),

(5)

n, m

with

FIG. 1. Reference frame S and B with position vectors

and r~ centered, respectively, on sphere S and B.
The center-to-center vector r~-rs is characterized by
the angle e s~ and the angle P sz with respect to the
original fixed reference frame.

The functions X,~(S) described in a frame centered on sphere S can be expressed in terms of
functions H(B) centered on sphere 8 as

H(r, e, p) =r"Y(e,y),
where the Y(8,p) are the usual spherical harmonics.
Q,

x(s) = g

The potential P, outside the sphere is

with

" 'Y.-(e-, q).

+ [(n+ 1)jn]~,
(~B ~ )H2n+1

e,

Also, let the incident radiation field be defined

by an electric field

Q, =

s.p, e

=-vQ,

and a corresponding

cp)

nm

d,

b, (S)(X, (S)~H(B)) =

where the notation

with S IB, and

electrical potential

- E, Q dH (r,8, .

(14)

Rewriting the boundary conditions in terms of the

above expressions (5), (7), and (10), using (14),
and eliminating a(B) between these equations
leads to a matrix equation for the b(B)

=-E, g bx(r,e, y),

x(r, e, y) =r

. .

&x(s)lH. (B)&H. (B)

n, m

Z, ~, means Z~, Z, =, Zi,

(15)
='

(10)

n, m

B. N-sphere case
Owing to the presence of neighboring spheres,
the field inside and outside a sphere (for example,
B) is modified, , and its interior potential is hereby called Q;(B) with coefficients a(B). The potential outside B due to the "incident" potential
$0 and that due to the other (for example, S)
spheres is given by
y

Q, (B) = Q, (B)+

S&8

g, (s)+ go.

The usual boundary conditions on the sphere of

radius A~ and dielectric constant ~; are

4';(B) ls, = 4'~(B) ls,

e; VP, (B) nlrb =e,~&~(B) nlrb

where n is the normal at the surface of sphere

(12)

B.

FIG. 2. A point r in the space is defined in both

rotated frames rs and r~ by the polar angles (respectively, 8&, ys and 0&, yz). The common z axis gives
here ps = qy~. Notice that spheres S and 8 are not
necessarily touching each other.

(&(5)III

J.

..

M. GKRARDY AND M. AUSLOOS

axis by the same angle n, in order to obtain

two new frames with a common a axis (see Fig.

(B)&

f.a& p(rs

8s Ws)~* (ra 8a qa)della

1).
The relation between the spherical harmonics
of the original system and the final system can be
written using Jeffrey's theorem"

f&aI!&. (ra 8a &a)l'della

is, in fact, the hardest part to calculate.

In order to do so, it is necessary to express
the Y, (8s, gs) spherical harmonics in the B reference frame. All frames are chosen with each
axis parallel to a general reference frame. First,
both reference frames S and B are rotated around
the z axis by the same angle P~, then around the

which

Y.(8q,)=e" l=-n

O(n, m,

L,

with

'n)~'
cosao,
sin
g (-1"

-m
-r)!(n
r!(n
r)!(m+L+

"' ~ '

zr+m+'

0(n, m,

L,

n) = (-1)""[(n+m)!(n -m)!(n+L)!(n L)!]'~'

~)Y,(8q,),

)r!

-m L) and g = min(n L, n -m). Taking into account the distance between reference
where
frames, it is now easy to relate the spherical harmonics in different frames,
(n+&)|Y (8a~ga)
(2n+ I)'~'
Ir ral
Y(8s,gs)
),
'
'

Psa""
-m)!]'
m)!(0+m)!]'
[(2@+1)(k
[(n+m)!(n
(r

= max(0,

r(""

(see Fig. 2 for definition of parameters. )

Using (16), (17), (19), and standard orthogonality

where

6 is

l,

the Kronecker delta function,

2n+

substituted into (15), allows one to describe the field outside the spheres in the absence of retardation effects. Equation (15) consists of a generalization to arbitrary clusters of
the method proposed by McKenzie et al. and by
McPhedran el' al. for calculating the static properties of lattices of polarizable spheres.

which,

"

"

III. RETARDATION EFFECTS

effects is valid as

The neglect of retardation

long as the spheres are close to each other, i.e. ,

where k, is the wave number in the
~k, asa~
matrix. Nevertheless, situations arise in which
the intersphere distance is very large (inter-

1,

Et

6e

),B",]', 'b.{B)+

relations for the spherical harmonics,

(20)

we obtain

(-1)"(q+n)! 2q+1 '';&, !s

a~sa""

(19)

(-1)'O(q, P, !, )O(n, m, !,ns

~
(q+
'~'
~s f(n+
o's

L)!(n

L)!

L)!(q

L)!J

stellar medium) or the incident radiation contains

small wavelengths and the field oscillates between
distant spheres In suc. h a case, ~k, asa~ is not
negligible with respect to unity. The phase of the
incident field and of the diffracted field must be
taken into account, and phase factors must be
introduced in each term of Eq. (15). The new
equation, being much more complex, is not
written here but will be discussed in a later work.
However, assuming that (1) the phase of the incident field on the sphere is equal to that at the
sphere center, and (2) the phase difference between diffracted fields by different spheres is
that between the sphere centers, Eq. (15) is easily
modified to read

g 5(S)(X(B)ja.(B))e*""a=-d.e'" 'a.

S,p, q

The above assumptions are strictly valid for

"point spheres" and, by extension, for slowly
varying fields inside and outside each sphere:

(21)

and ~A+a( 1, where k; is the wave

(k; Ra(
number in the B sphere. Typically, we have
Rq & 1000 A.

(22)

ABSORPTION SPECTRUM OF CLUSTERS OF SPHERES. . .

22

IV. EXAMPLES
A. Linear chain, identical spheres, regularly spaced

As shown elsewhere,

chains of identical spheres

are seen in pictures of powders" and are representative of the structure of a real powder.
Their study can give a good idea of the bounds

"

1.0x

and shape of the infrared spectrum.

We suppose the chain to be parallel to the z
axis. We consider only the long-wavelengthlimit case (i.e. , Ik, =0), but we will look for

0.9-

chain

+1

2 spheres

cases corresponding

to a different polarization

mode of the incident field. We can write, using

0.8

(21),
(q+n) ~[(-I)'+ (-1)"]5.,
~( ""'I"-"" [("~).
(' p). ("p).(. -p) ] "
=

(2q+ 1)'~'(-1)~& (q+ n + 1)

(2n + I) &/ aga

+ n+

&

(23)

where a is the distance between the centers of

spheres and f(n) is the Riemann
zeta function.
By symmetry, all the spheres oscillate in
phase: b(S)= b(B)= b in such a limit. In
order to solve Eq (15),. the sum in the left-hand
side (i.e., the development of the diffracted field)
is limited to a given polar order (here to the
quadrupolar order, n & 2). The matrix equation
is separable into two parts, one involving the
dipolar modes only, the other the quadrupolar
modes only. The resulting resonant frequencies
are a superposition of two independent equations.
The dipolar modes u, are found from the poles
of the expression giving b, , i.e. , from

"

two neighboring

j,
" (.(a/8)'+
/~)'-2~
where r (3) = 1.2021 and

j,

g(3)
'
~(3)

(24a)

takes the values

for m=+1, and+2 for m=0. Similarly,

quencies of the quadrupolar modes,
tained from the poles of b, or from
$(

2m)

8(

2m)

+,

(a/Z) 5--;j, g (5)

(+/~)5+ g(5)

-1

the

fre-

, are ob-

(24b)

j,

where g (5) = 1.0369 and

takes the values 2 for
m =+ 2, -8 for m =a 1, and + 12 for m = 0.
Fuchs" has shown that the use of the bulkmaterial dielectric constant e;(&u) is a good approximation for describing the true dielectric
constant of cubic dielectric particles. For testing
our method, we have thus taken a well-known
model dielectric constant

f (((a&) = f ~ + (65 6'~)(dr/((dr 4P

extending

Fuchs results to spherical shapes.

(25)

FIG. 3. Ratio of the dipolar resonant frequency

wand the usual noninteracting single-particle Frohlich
frequency co& as a function of the interparticle distance
measured in sphere radius R units for Mgo particles.
Crosses and
correspond to equal-size spheres linearSolid and dashed lines correscluster (R =R&R&). The inner
modes (dashed line) are doubly degenerate.

chain resonant modes.

pond to the two-sphere

In Fig. 3, the crosse s and show the variation o f the

dipolar mode frequencies &u(= ~, ) as a function
of the ratio a/8 for the case of a linear chain of
MgO spheres in vacuum; ~~ is the Frohlich fre-

quency and ~~(~r) is the longitudinal (transverse)

plasmon bulk resonance given by e;(co~) = 0
[e;(~r) =~]. For MgO we have taken c, =9.8,
2

95

~z=7. 5x10" s ' ~~=13.67x10" s '

'.

and u~ = 11.58 x 10" s

When the incident uniform field is parallel to
the axis of the chain, the m = 0 mode (longitudinal
polarization of the field) is the only active one.
When the field is normal to the chain axis (transversal polarization) only the degenerate mode
m =+ 1 is active. In Fig. 4, we have given the
variation of the quadrupolar mode frequencies
&u(= &u, ) as a function of the ratio g/B in the
same cases as in Fig. 3. The mode multiplicity
is indicated. All these modes are nonactive in a

uniform incident field. The characteristic frequency +@ of the so-called quadrupolar mode of
a single sphere is defined by

~;(~o) = , ~8(~g)

(26)

[viz. , taking the limit a-~ in Eq. (24b)] and

serves as a normalizing frequency.
B. Two

spheres with different radius

Limiting Eq. (15) to the dipolar order (n & 1)

and neglecting retardation effects again, we obtain the resonant modes by solving
2

,',

'j

A3 .

[&'~'(~)+2~' (~)]= R,',

[~"(~) -~, (~)],

(27)

J.

4954

M. GERARDY AND M. AUSLOOS

where the integration is taken over the volume
]d'or parallel
v~ of the sphere. Average normal and
susceptibilities are defined by

Mg0

1.04

PB((u) = VB[X(B,(u)1)I+X, (B, Cu) 1,

are unit vectors, respectively,

1~~ and 1
parallel and perpendicular to the incident field
E,. Only X(B,Iv) is discussed here F. rom (28),
(29), and (30), we have
where

1.02

0.9 8

1 1 e B
'
XII(B, co) =
4mEO,
v~ 4

0.9 6
0.9 4-

FIG. 4. Quadrupolar resonance frequency ~ vs a/R

ratio in the case of a linear chain of equal-size MgO
spheres (radius R, distance

a).

The mode multiplicity

is given in parentheses; the number

m characterizes
part of the spherical harmonics.

the azimuthal

V. OPTICAL ABSORPTION

xR '"

(28)

where r~ is the radial vector described in the

frame centered on the B sphere. We define the
average electrical moment for the B particle as

(r)=

f p)r,
Vg

~)dr

(29)

1,)drB.

(31)

'b

e, )]

(B)

(32)

Gauss theorem, using expression (5) for

the potential inside the sphere, and orthogonality
relations (20) for the spherical harmonics, we
obtain from (31) and (32)

Applying

.
3/2

X,

(B, ) =I

.[", (

-1]/[".(-) - ]R.'

x (2 '~2sinoI[b(B)e'B

b, ,(B)e '
]

+ (coso.')b(B)],

(33)

where n and )8 are spherical angles defining the

direction of the incident field in the Cartesian
frame centered on sphere B (see, e.g. , Fig.

1). Because

XB(o) = (1/4~)

where
A

In order to describe the absorption spectrum

of a powder we can consider either the absorption
coefficient, as defined in (1), or the total absorbed power. For the former quantity, we calculate an "averaged susceptibility" using the
average polarization of each sphere. This form
takes into account the interaction of neighboring
spheres through the results of Sec. III.
The polarization density can be written as

Notice that when E;(rB, Iv) is uniform over the

volume vB, (31) corresponds to the Fuchs's
macroscopic definition of the susceptibility.
From the calculation made in Sec. II, one can
easily obtain

)B~B

[(~/~)' 1],

(a ) is

phonon frequency,

p(rB, ~) = [(1 c'; )/4v]E, (rB, &u),

E;(rB, Iv)

a(B)= [(2n+ 1)/n] [e,/(e;

where e' ) (v) is the dielectric constant of the sth

= 2 for m = 0.
sphere and = -1 for m =z 1, and
The solutions of (27) have been calculated in the
case of two MgO spheres and are shown in Fig. 3.
Notice that the downward shifts of the Frohlich
mode are much smaller than those of the linear
chain, but the upward shifts have similar magnitudes. As in the case of the chain, the modes
m = 0 are active only for a longitudinal polarization of the field, while the modes m =+ 1 are active for a transverse polarization. Notice that
when R, =R the upper frequency mode in the
longitudinal polarization geometry and the lower
frequency mode in the transverse polarization
case are always nonactive. Only one mode is thus
active in such a case (R, =R,) for a given polarization. The quadrupolar modes are more complicated to obtain and are left for further work.

rr

(30)

(34)

the longitudinal (transverse)

we can rewrite (1) as

(B,(o) = 8(uu)rB/(IvqB Ia)rB) ImXII(Br

Id)

(35)

The absorption can be deduced immediately from

(33) and (35). For N (different or not) spheres,
the total absorption is

+ II(IV)

B= 1

~BA II(

+) )

(36)

where $~ is a normalization factor taking into

account the particle volume (hence the polarization moment density),
$B

VB/V

V=+ Vs.
S=x

(37)
(38)

We show in Fig. 5 the absorptionAII(&d) for the

case of two uneq'ual-size MgO spheres for different field incidences and arbitrary damping fac-

ABSORPTION SPECTRUM OF CLUSTERS OF SPHERES. . .

22
40

RR1-

R2= 100A

30-

A ) )(tilI

Mg0

50k

MgO

10-

150A

~l

Y = 0.04 0dT

I 1
I
I

4955

R1 =R2=75A

a =150k

20

10

0100

120

110

130
ft)(10 sec )

1.3

FIG. 5. Absorption A{u) for two touching different

MgO spheres (R2=2Rq=100 A) as a function of frequency.
The damping factor y is chosen equal to 0.04 ~
Curve
is given for a field direction such that the
angle n between the center-to-center axis and the inn= v/5; curve
cident field Eo is n=0; curve
~
~
a=v/2. gs is the
n=v/2; and curve

:
.

. :

17

FIG. 6. Absorption A, {co) for two touching equal-size

MgO spheres {radii R&=R2= 75 A). The damping factor
y is 0.04 co z. The incidence angle 0. of the field Eo is
x/3. Curve ~ is obtained by limiting equations up to
takes into account dipolar
dipolar terms, and curve
and quadrupolar interaction as well.

z.

1.4

Frohlich resonant mode.

tor (y=0.04&or), with the

by the spheres can be -obtained from the

h's

field distribution

calculated taking
into account the dipolar terms only. In Fig. 6,
we show the absorption A. (((~) for two equal-size
MgO spheres, but with the 5;, 's calculated taking
into account the quadrupolar (n= 2) terms
On the other hand, the total power ~' absorbed

w(

)=- Zw(8

) =

'

cleB,

i.e.,

electric
for parti-

we have

(('((), ~)=~;(~)

(E;(&, ~)l'B l(~.&l)

()

0gy

in reduced units, which for N spheres sums up to

)E

'

in the particles,

MgO
3

'" z-'"-

I(

(8)I')

(40)

Is

)l

1
2

&=225--a
d. 45 --

c(-900 .

100

110

120

130
1

~(10 sec )

FIG. 7. Reactive power W as a function of frequency u for two values of the damping factor {p = 0: infinite-peak
curves; y= 0.04 cur: finite-maximum curves)'for two touching MgO spheres with equal radius R = 75 A. Different angjes
e are considered and indicated on the figure.

J.

4956

M. GERARD Y AND M. AUS LOOS

Log W

fs
90

~~

~ ~
~ ~
~ ~

-1

100

120

110

130

~(10 sec")

FIG. 8. Reactive power W' (~) for two unequal-size

fieM directions e are indicated.

MgO spheres

The active and reactive power are, respecIn

tively, defined by Im~ = W" and Be' =
absence of damping, the active power &" vanishes, while the absorption spectrum consists
of delta functions. The reactive power W' as
seen from Eq. (40) is a linear function of the real
part of the dielectric constant e'(e) times a complicated function of the frequency. It can be as
usual related to the imaginary part of the dielec-

~'.

(R2= 2A1= 75 A, a = 150 A) with y= 0. The incident

tric constant, through Kramers-Kronig

relations,

and hence to experimental data.

The reactive power W' has been calculated in
the dipolar approximation (n & 1) for two MgO
spheres of equal sizes for four different field
incidences and different damping factors (y = 0,

or y = 0.04&or) (Fig. 7). On Figs. 8 and 9, the

reactive power of two unequal-size Mgo spheres
is shown as a function of , for different damping

Log W

2.0

1.0

0.0
100

120

110

co (10" sec" )
FIG. 9. Reactive power W'(co) in the same case as that of Fig. 6 for different angles
dipolar terms. The damping factor y is chosen to be equal to 0.04~&.

n.

Equations are limited up to

ABSORPTION SPECTRUM OF CLUSTERS OF SPHERES. . .

22

factors, and for different field incidences.

VI. DISCUSSION

The most important improvement of our work

over previous theories is the inclusion of the effect of all 2l polar modes on the dipolar resonant
modes and on the dipolar absorption spectrum.
Furthermore, the method is general enough to
permit the calculation of absorption modes and
the absorption spectrum of any distribution of
spheres in a medium as long as their dielectric
constant and radius are known. The full calculation needs much computer work and has not yet
been attempted. In fact, it is necessary to know
powder statistics before analyzing a given spectrum. The inverse problem (obtaining the statistics from the spectrum} seems, after this
stage, partially feasible when the spherical shape
approximation is removed.
We will discuss successively:
(a) the effect of the incident field direction on the
absorption,
(b) the position of dipolar modes with and without quadrupolar effects,
(c) the possible generalization of implicit
formulas giving the mode positions when higher
multipolar orders are included,
(d) the damping-factor influence on the spectrum,
(e) the influence of the shape or size of the par-

ticles.
For a cluster of

two identical spheres, both absorbing modes were already found by Clippe
et al. However, the position of the four resonant modes predicted by (27), when the spheres
have unequal size, is a new result. Notice the
drastic effect of the incident field direction (with
respect to the center-to-center axis) on the absorption amplitude and absorbed power spectrum.
Longitudinal absorption modes only appear when
the "excitation" is parallel to the common axis
of both spheres, while transverse modes appear
when E is perpendicular to this axis. As this
observation also holds true for the spectral bounds
of the linear chain, one can easily imagine how
useful the observation of a powderlike system
would be in polarized light. This would lead to
some knowledge of the configurations of spheres
in clusters.
However, before one can be fully certain about
this point, it is necessary to have a precise idea
of the position of the resonant dipolar modes.
Our work in Secs. III and IV indeed indicates that
an appreciable shift in the theoretical mode position occurs according to the level of accuracy
imposed on the solution of Maxwell's equations.

'

We have indeed indicated in Fig. 6 the important

displacement of the mode when quadrupolar terms
and
are taken into account in calculating the
the ~ 's. It would be of interest to examine
whether the "motion" of the modes toward lower
and upper values is asymptotic or not. The answer is beyond the scope of this paper and is left

b's

for future work.

Furthermore, the appearance of low shoulders
in the upper part of the spectrum is a new feature
which requires extension of this work. The variation of dipolar modes as a function of a/R has already been examined, e.g. , by Ronveaux et al.
Their results agree with ours, although some
approximation (at the final stage) leads them into
completely missing the m =0 mode. These authors
indeed take into account only the multipolar fields
for which n= )m), with n~ 1. The power-law
variation of &u vs a/R is identical to ours, however. In the Appendix, we correct the a pro~i
truncation of the field equations Iby Ronveaux
et al. and recover our results.
An interesting point concerns the implicit equations allowing the extraction of the dipolar and
quadrupolar (~k, =0) resonant modes [i.e., Eqs.
(24) and (25), and their generalization to higherorder modes of the linear chain]. Preliminary
results indicate that the form of the equations is
more complicated and cannot be written in general

"

"

as
&(

r)

f+1
1

e(

urn)

(a/R)"" a,j,f(2l + 1}
( /R)2l+1+P

'

g(2)+ 1)

(41)
Inclusion of the octupolar term in the field equations already breaks the "symmetry" of the equations. For the two-sphere cluster, inclusion of
quadrupolar terms also leads to a more elaborate
form of an implicit equation like (27). Eq. (41) is
only exact in the limit of large sphere separation.
Nevertheless, it would be interesting to observe
on experimental data whether Eq. (41) corresponds
to and may serve as some realistic approximation
to characteristic mode frequencies.
Another interesting result concerns the influence
of the damping factor on the spectrum. Of course,
the greater the damping, the broader the spectrum. However, it is remarkable that for any incident field direction, and for a given damping y,
the absorption amplitude (or the absorbed power)
has the same magnitude at a particular frequency
*
(see Figs.
&u, (~,*) close to the Frohlich frequency
5 and 9). Therefore, a change in incident field
direction and observation of u,* (&u,*) would allow
a determination of the experimental parameter y.
We have obtained, in the dipolar approximation,
an analytic expression for y as a function of ~,*

J.
1 &2(4+s)(l

-f')'

M. GERARDY AD M. AUSLOOS

=&@2/vz, fz =+~/uyr, 5 =y/&oz, and s

=(R/a)' for two identical spheres, or s =2/(S)
(R/a)2 for the linear chain (s CQ).
The variation of ~*, with y is rather weak, however, and does not seem to be observable. This

where

feature of the spectrum remains when higherorder terms are included in the calculation for
a single-strand chain of different or identical
particles.
The damping factor y has been an a Priori
parameter introduced into Eg. (25), for example,
in order to smooth the spectrum and make it
"more realistic. " For y=O, a complete solution
of the field equations would lead to an infinite
number of modes [or at least to NI(I +2) modes
for an N-sphere cluster when the sum in (15) is
limited to the 2l polar order]. The largest amplitude of the absorption modes would be that for
the dipolar modes, followed by that for the quadrupolar modes, etc. The total set of modes would
seem to form a continuum. In such a sense, y
can be considered as reproducing (in the dipolar
approximation) the effect of higher-order terms
A self-consistent treatment of an effective medium
dielectric constant would in fact introduce a real
and an imaginary part of the resonant frequencies;
the latter part as y introduces a broadening of
the spectrum. '
In usual experimental situations, the particles
dispersed in a medium are not truly spherical
but have irregular shapes. This causes further
substantial broadening of the absorption peaks
and spectrum; Fuchs's and I angbein's work'"
on independent cubes has indeed shown that the
inclusion of higher-order multipolar terms leads
to an additional structure in the absorption spectrum.
According to their results, the strongest absorption is a measure of the equivalent "Frohlich mode" of the cube. It remains to be discussed how the broadening due to the irregular
shape of the particles and that due to the distribution of (spherical) particles in a cluster are
affected and in which proportion. Both effects
might play a role in reducing the observed "strongest absorption peak" and lead to an erroneous
attribution of the mode. Much more work is obviously needed in such a direction.
Finally, even in the framework of our model
with "all spherical particles,
we can now cover
a larger part of the absorption spectrum of a
powder by including higher-order multipolar

"

terms.

-f ')'

2g, g(1+s) (f2~ -f')(1 f ') e'(2-s)(f',

(2 s)&2 + 2e, &(1+s) e2(4 +s)

(42)

ACKNOWLEDGMENT

One of us (J. M. G. ) would like to thank the Institut. pour la Recherche Scientifique dans 1'Industrie et 1'Agriculture (IRSIA), Brussels, Belgium, for financial support.
APPENDIX

For a linear strand chain, Ecl. (18) becomes

O(n, m, l, o)

=5, ,

(Al)

I, ~) =(- I)""5. ,

O(n, m,

Consider identical particles of radius R and center-to-center distance a. Inserting (Al) into ex-

pression (21) and limiting Eg. (15) for

~1 to

"

~m~

n=q=m, i.e. , following the Ronveaux et al. approximations,

we obtain

+(n+I)c, (&o)

n&, ((o)

n[E;((d) t

(&d)]

b Il&

p (R/a)2n+1

Olj

where
n = m
~

)'2

~22+1

d ~2n+1
n

(A2)

j and
~

k are the sphere indices. Modes

are twice degenerate (n = +m [) and are
~

given by Ronveaux's

equation

det(MXE)=0,

(AS)

where M is the matrix

(N -1~
~

(A-I)

"

(N 2)

"

(N

3)

"

~ ~

"

-~
)~

(N 2)

'

~ ~ ~

(A4)
and where

n2:;(v)

+)&I, (+) a
&2(4J)] R

(+n

n[fq(4P)

"
(A5)

p, = 2n +1. This sort of truncation neglects

the m =0 (nondegenerate) modes bees, use n ~ 1.
With the same approximations [Ecl. (15) limited
up to dipolar terms], the m =0 modes are obtained
from (AS) and (A4) by adding the condition M

with

=Mand

1 &;((u) +2&.((a)
2 e;((o) &, ((u) R

(A6)

ABSORPTIO SPECTRUM OF CLUSTERS OF SPHERES. . .

'.

which differs from X by a factor ,

Modes m =0 are quite important because they
lead to a greater splitting than others (see, e. g. ,

Meeting on Small Particles and Inorganic

Clusters 1976 [J. Phys. (Paris) Colloq. 38, C-2, S7
(1977) and references thereinJ.
R. Fuchs, Phys. Bev. B 11, 1732 (1975).
3R. Frech, Phys. Rev. B 13, 2342 (1976).
P. Clippe, R. Evrard, and A. A. Lucas, Phys. Rev.
B 14, 1715 (1976).
R. Buppin, Surf. Sci. 58, 550 (1976).
6R. Ruppin, Surf. Sci. 51, 149 (1975).
YR. Ruppin, Surf. Sci. 62, 206 (1977).
T. P. Martin, Phys. Rev. B 7, 3906 (1973).
R. Huppin, Phys. Status Solidi. 8 87, 819 (1978).
M. Ausloos, P. Clippe, and A. A. Lucas, Phys Rev.
B 18, 7176 (1978).
J. H. Weaver, R. W. Alexander, L. Terg, R. A. Mann,
and R. J. Bell, Phys. Status Solidi. A 20, 321 (1973).
'2R. Fuchs, Phys. Rev. B 18, 7160 (1978).
T. P. Martin and H. Schaber, Phys. Status Solidi B 81,

International

K41 (1977}.

C. G. Granqvist

and N. Calander,

Solid State Commun.

31, 249 (1979).

F. F. Y. Wang, C. H.

Lin, K. C. Yoo,

P. J. Herley,

Fig. 3). Note that all modes (except n =1) given

are nonactive in a uniform field because

by (A3)

the

d's differ

from 0 only when n

=1.

and Y, H. Kao, Philos. Mag. B39, 277

(1979).
A. J. Hunt, T. R. Steyer and D. A. Huffman, Surf. Sci.
36, 454 (1973).
~~A. A. Lucas, J. Chem. Phys. 48, 3156 (1968).
H. Frohlich, Theory of Dielectrics (Clarendon, Oxford, 1949).
B. Jeffreys, Geophys. J. R. Astron. Soc. 10, 141 (1965).
D. R. McKenzie, R. C. McPhedran, and G. H. Derrick,
Proc. R, Soc. London Ser. A 362, 211 (1978).
R. C. Mc Phedran and D. R. McKenzie, Proc. R. Soc.
London Ser. A 359, 1I5 (1978).
M. Abramowitz and I. A. Stegun, Handbook of ~athematica/ Fhnctions (Dover, New York, 1972).
A. Ronveaux, A. Moussiaux, and A. A. Lucas, Can. J.
Phys. 55, 16 (1977).
24J. E. Gubernatis, E/ectrical Transport and Optical
Properties of Inhomogeneous Media, edited by J. C.
Garland and D. B. Tanner (AIP, New York, 1977),
p. 84.
D. Langbein, J. Phys. A 9, 627 (1976).

H. Herman,