PHYSICAL REVIEW B 70, 235103 (2004)

Surface plasma resonance and magneto-optical enhancement in composites containing

multicore-shell structured nanoparticles
Masanori Abe and Takeshi Suwa
Department of Physical Electronics, Tokyo Institute of Technology, Ookayama, Meguro-ku, Tokyo 152-8552, Japan
(Received 5 April 2004; published 1 December 2004)
We have formulated the conditions in which dipolar surface plasma resonance is excited by light waves in
composites containing nanometer-sized, n-fold multicore-shell structured particles (referred henceforth as
nano-onions) dispersed in matrices. The nano-onions have ellipsoidal shape and an arbitrary layer number n,
which contain shell(s) (or core) of a metal having Drude type free electrons responsible for the surface plasma
oscillation. By solving a quasistatic potential boundary problem in a nano-onion that is exposed to an external
static electric field, we derived the effective dielectric permittivity tensor, including off-diagonal elements, for
the composites, based on the Maxwell Garnett theory. The results were utilized not only to derive the resonance
conditions but also to formulate the surface charge densities on the metal surfaces, from which we determined
the symmetry of the dipolar surface plasmon polaritons excited in the metal shells. Calculations made on the
composites containing model nano-onions of spherical shape having n-fold core-shell structure of sodium and
a dielectric revealed the following results: (1) The surface plasmon resonance occurs at n eigenfrequencies,
similar to the mechanical oscillation in n-fold coupled oscillators; (2) at these eigenfrequencies, the composite
causes resonant peaks of light extinction coefficient, and (3) the magneto-optical Kerr effect induced by a static
external magnetic field is remarkably enhanced at the resonance frequencies. The magneto-optical enhancement is augmented by hypothetically reducing the dielectric loss in Na, thus increasing the quality factor Q of
the surface plasmon resonance. The validity limit in our calculations based on the effective medium approximation by the Maxwell Garnett theory is discussed, comparing with the calculations made by Sinzig and
Quinten [Appl. Phys. A 58, 157 (1994)] based on a rigorous Mie scattering theory treatment.
DOI: 10.1103/PhysRevB.70.235103

PACS number(s): 78.67.n, 45.70.n, 78.40.q, 78.20.e

I. INTRODUCTION

In this study we analyzed surface plasma resonance in

nanoparticles with multicore-shell structures, and found that
the magneto-optical effect is remarkably enhanced near the
resonance. Such multifold core-shell structured nanoparticles, referred to here as nano-onions,1 are attracting extensive attention at present, since the multilayered structures
are expected to give particular magnetic properties and exhibit unique optical and magneto-optical effects similarly observed in metallic multilayers of nanometer-scale thickness.2
Surface plasma (or plasmon) resonance is an excitation of
charge density surface waves, which propagate along a metallic surface or on metal films.35 The free electrons, following the Drude theory, play an essential role in surface plasmon resonance, which is most prominently observed in noble
and alkaline metals. The surface plasmons can be excited by
light waves, making a contrast with volume plasmons that
can be excited not by light waves but by electron beams.
Therefore, the surface plasmon resonance has attracted special interest among researchers studying optically induced
phenomena in matter. They found that the surface plasmon
resonance enhances magneto-optical effects,612 nonlinear
optical processes,1113 surface Raman scatterings,1416 and
photon-induced catalysis reactions17 to a remarkable extent.
This created much interest not only in solid-state physics but
also in sensor technology, because the physical effect enhancements by the surface plasmons facilitated a variety of
highly sensitive surface analysis devices, including chemical
sensors18 and biomedical sensors.19
1098-0121/2004/70(23)/235103(15)/$22.50

Optical excitation of surface plasmons on planar metal

surfaces can be realized only by evanescent waves, because
surface plasmons have retarded dispersion relations with respect to the excitation light waves.36,911,13 However, even
using ordinary (nonevanescent) light waves, surface plasmons can be excited on fine metal particles,7,8,12,14,15,17,2023
which may be embedded in matrices (thus forming
granular composites)7,12,17,2023 or located on planar
surfaces.8,1417,24,25 The particles on the planar surfaces offer
a model for rough surfaces on which surface plasmons are
excited to enhance the surface Raman scattering.1316 Surface
plasmon resonance excited by light waves in fine metal particles have therefore been studied extensively.
Granqvist and Hunderi21 analyzed the optical absorption
by ultrafine (3 4 nm in diameter) Ag spheres in terms of
surface plasmon resonance. Applying the Maxwell Garnett
(MG) theory26 and the Bruggeman theory,27 they calculated
effective dielectric constants for composites containing the
Ag nanoparticles dispersed in matrices based on effective
medium approximation. The MG theory presupposes that the
particles are sparsely dispersed, or the filling factor of the
particles is small, while the Bruggeman theory is free from
such a premise on the filling factor, as will be described in
Sec. V, Hui and Stroud7 showed using the MG theory that
magneto-optical Faraday rotation in dilute suspension of
small particles of a Drude metal is enhanced by the surface
plasma resonance on the metal surfaces. The analyses were
extended to higher concentrations of the suspensions by applying the Bruggeman theory.23 The analyses of surface plasmon based on the MG theory were further extended to such

235103-1

2004 The American Physical Society

PHYSICAL REVIEW B 70, 235103 (2004)

M. ABE AND T. SUWA

core/shell structured nanoparticles as Ag spheres having inner dielectric cores28 or outer dielectric shells,29 with which
they explained the optical absorption and the surfaceenhanced Raman scattering by the plasma resonance in the
Ag shells or the cores.
Almost a century ago, Mie30 showed with his scattering
theory that spherical metal particles exhibit resonant extinction of light, which was later interpreted by Kreibig et
al.20,31,32 by the concept of electric and magnetic multipolar
excitations due to surface plasmon polaritons and eddy currents, respectively. They revealed, based on the Mie theory,
that the resonant light absorption observed for spherical
nanoparticles of Ag and Au can be interpreted by invoking
the surface plasma resonance of electric multipolar modes,
mainly from the lowest electric excitation mode called the
dipolar surface plasmon polariton mode. From the Mie scattering theory, Kerker et al.33 and Messinger et al.34 explained
the surface enhanced Raman scattering in spherical metal
particles in terms of surface plasmon resonance. Wokaun35
extended the Mie formalism for the surface plasmon resonance to spheroidal metal particles, with which he investigated the marked surface enhancements of Raman scattering
and the catalytic activity of the Ag nanoparticles. The Mie
theory analyses on the surface plasmon resonance were further extended to the double layered nanoparticles having
spherical dielectric (core)/Ag (shell),36 spheroidal latex
(core)/Ag (shell),37 and spherical Au (core)/Pt (shell)38 structures.
Sinzig and Quinten39 investigated the surface plasmon
resonance in n-fold multi-core-shell structured particles with
arbitrary number n. Calculating the Mie scattering intensities, they showed that the multicore-shell structures with alternate stratification of Na and a dielectric exhibit resonant
peaks of light extinction, which they guessed to be ascribed
to the surface plasmon polaritons excited in the Na shells and
cores. They also proposed that in the core(dielectric)/
shell(metal) structured n = 2 particles embedded in a dielectric matrix, two surface plasmon eigenmodes are excited,
having symmetric and anti-symmetric charge distributions,
respectively, on the inner and outer surfaces of the metal
shell. Their calculations were, however, not derived analytically, but from analogy with the surface plasmon eigenmodes
excited in a metal planar layer. By the Mie scattering theory
they calculated the fields induced in and scattered outside the
particles, using complicated recurrent formulas; the Mie scattering theory cannot give analytical expressions for the fields
or, therefore, the conditions in which surface plasmon resonance occurs.
In this study we derived, based on the MG theory, analytical formulas for the fields induced in the n-fold nanoonions, and formulated the conditions in which the surface
plasmon resonance takes place in the composites containing
the nano-onions. We also formulated the charge distributions
induced on the surfaces of the metal shells, with which we
determined the symmetry of the surface plasmon polariton
eigenmodes induced in the metal shells. We revealed that the
surface plasmon resonance enhances not only light extinction
but also magneto-optical effects. In our calculations we assumed that the nano-onions much smaller than the light
wavelength are dispersed sparsely in a dielectric matrix to

form a composite. The light waves propagate in such a composite as if it were a continuous medium, called an effective
medium, having an effective dielectric permittivity. We
derived the effective permittivity based on the MG theory in
tensor form, including the off-diagonal terms that are responsible for the magneto-optical effects.
The effective dielectric permittivity tensor was derived by
Lissberger and Saunders40 based on the MG theory in order
to explain the magneto-optical Kerr effect for cermets, or
composites in which magnetized spherical particles are embedded in dielectric materials. Carey et al.41 extended the
effective dielectric tensor to spherical particles with a core/
shell structure, thus explaining the magneto-optical effects in
granular films in which surface-oxidized Co particles are dispersed in a dielectric matrix. Abe,42 one of the authors of this
paper, generalized the effective dielectric permittivity tensor
to composites containing magnetized, oriented ellipsoid particles embedded in matrices which may be magnetic or nonmagnetic. The effective dielectric tensor will be further extended in this study to the nano-onions having an arbitrary
number of core/shell layers.
In Sec. II, we will solve a quasistatic potential boundary
problem for the electric fields induced in the core and shells
of a magnetized n-fold nano-onion in order to derive the
electric polarizability of the nano-onion. The result will be
used to formulate, based on the MG theory, the effective
dielectric permittivity tensor for composites containing the
magnetized nano-onions. In Sec. III, we will derive the conditions upon which the surface plasmon resonance occurs in
the composites, and derive an equation that we can use to
determine the symmetry of the charge distributions of surface plasmon polariton eigenmodes. In Sec. IV, we will investigate the resonant conditions for the surface plasmons in
a composite containing the spherical, n-fold nano-onions
made of Na and a dielectric. It will be shown that the
magneto-optical Kerr effect, as well as the light extinction, is
remarkably enhanced by the surface plasmon oscillation. In
Sec. V, we will discuss the applicability limits in the MG
theory and then conclude.
II. DERIVATION OF EFFECTIVE DIELECTRIC TENSOR
A. Potential boundary problem

In order to derive the effective dielectric permittivity tensor based on the MG theory for the composites containing
ellipsoidal nano-onions (Fig. 1) we first solve the potential
boundary problem for the electric fields induced in the core
and shells of a magnetized nano-onion that is embedded in a
matrix. Using the quasistatic approximation, we neglect the
spatial dependence of the electric field of the light, but the
time dependence is introduced through the wavelength dependent complex dielectric functions for the core and the
shells.
Consider that a uniform, isotropic medium has in it a
uniform, or quasistatic electric field

Fx0

F0 = F0y .

2.1

Fz0

Let a nano-onion, ellipsoidal in shape (which is generalized

from spheroidal in our previous study42), be embedded in the

235103-2

PHYSICAL REVIEW B 70, 235103 (2004)

SURFACE PLASMA RESONANCE AND MAGNETO-

x = + a2 + a2 + a21/2/b2 a2c2 a21/2 ,

2.4a
y = + b2 + b2 + b21/2/c2 b2a2 b21/2 ,
2.4b
z = + c2 + c2 + c21/2/a2 c2b2 c21/2 ,
2.4c
which are valid in the ranges, satisfy the following relations:
b 2 c 2,

c 2 ,

a 2 b 2 .
2.5a

FIG. 1. A composite containing an oriented, spatially random

array of ellipsoidal nano-onions embedded in a host material in
which a uniform magnetic field F0 exists.

matrix (numbered as m = 0), as shown in Fig. 2. The outermost surface of each nano-onion has principal radii of a, b,
and c, along which we define the x, y, and z axes, respectively. The nano-onion has a co-centric, n-fold core/shell
structure whose boundaries are expressed by the following
quadratic equations:
x2 y 2 z2
2 + 2 + 2 =1
am
bm cm

m = 1,2, . . . ,n.

2.2

In this equation, a, b, and c represent the maximum, midpoint, and minimum values of a, b, and c, respectively, that
satisfy the relationship
c b a .

Using the radial coordinate , the outer boundary of the mth

medium is expressed by

= m

b1 = b,

c1 = c.

m = 1,2,3, . . . ,n,

2.6a

where

1 = 0,

2.6b

and ms satisfy the following equations:

Here, am, bm, and cm are the principal radii of the outer
surface of the mth shell or the core m = n, and thus
a1 = a,

2.5b

2.3

Now, we introduce ellipsoidal coordinates , , and as

follows:43

am = m + a21/2 ,

2.7a

bm = m + b21/2 ,

2.7b

cm = m + c21/2 ,

2.7c

m = 1,2, . . . ,n.
It should be noted that am : bm : cm depends on m, and, therefore, the boundary surfaces are not of similar shape, that is,
the further out the more spherical in shape is the shell.
Let the nano-onions and the matrix be magnetized along
an arbitrary direction, and express the dielectric permittivity
tensor m for the mth medium to the first order of magnetization as follows:

xy
xz
m m
m

m = myx m myz
zx
m

zy
m

m = 0,1,2, . . . ,n,

xy
zy
zx
xz
+ myx = myz + m
= m
+ m
= 0.
m

2.8a

2.8b

Our task is to obtain the electric field

x
Em

Em = Emy

m = 0,1,2, . . . ,n,

2.9

z
Em

FIG. 2. Cross section in xy plane for an n-fold nano-onion.

which is induced in the mth medium. The electric flux density in the mth medium
235103-3

PHYSICAL REVIEW B 70, 235103 (2004)

M. ABE AND T. SUWA

x
Dm

Dm =

m = 0,1,2, . . . ,n

Dmy
z
Dm

2.10

is given by the product of m and Em as

Dm = mEm .

x
Fm
Fmy
z
Fm

because the core has no inner structure.

In the ellipsoidal coordinate system, Em is expressed in
terms of m as

Em

2.11

The external field F0 is no longer equal to E0 in the matrix or

Em m 1 in the shells or core, because F0 induces electric
dipole moments inside the nano-onion. Now we assume that
in the shells and core, uniform electric fields

Fm =

m = 1,2, . . . ,n

2.12

Em

m = 1,2, . . . ,n

Pm = Pmy

2.13

z
Pm

Dm

m =

m = 0,1,2, . . . ,n,

x x
Cm
A x

Fmy

z
z z
Fm
+ Cm
A z

i Emiui

m = 0,1,2, . . . ,n,

uy = y,

s + a

2 3/2

A y =

2.20a

Em
=m = Em1
=m

m = 1,2, . . . ,n,

2.20b

Dm
=m = Dm1
=m

m = 1,2, . . . ,n.

2.20c

+ 1F + N

m1Fm1 = N
m m
m1
m1
m
m m
m
m1

2.15

uz = z,

s + b

2 1/2

s + c

2 1/2

2.21

mm1Cm1 = m m1Fm
1C .
+ mm m1N
m
m
m
2.22
Here we rewrote

i
Cm

i = x , y , z in a vector form, as

x
Cm

Cm = Cmy ,

2.23

z
Cm

ds,
and put

s + b23/2s + c21/2s + a21/2ds,

=
N
m

s + c23/2s + a21/2s + b21/2ds.

2.16c

Note that the dipole field is not induced in the core; namely,
235103-4

1 0 0
1 = 0 1 0 ,
0 0 1

2.16b
A z =

2.19

2.14c

2.16a

m = 0,1,2, . . . ,n.

Em

m = 1,2, . . . ,n,

and expressed the ith component of the dipole field in the

i i
A , with Ai given by43,44
mth medium as Cm

Em

As shown in the Appendix, we derive from Eq. (2.20) the

following recurrent vector formula:

i = x,y,z; m = 0,1,2, . . . ,n.

ux = x,

Em

Em
=m = Em1
=m

Here, we rewrote x, y, and z as ui i = x , y , z, i.e.,

A =

2.18

2.14b

2.14d

m = 0,1,2, . . . ,n,

N
m
m
m1 Cm

i
i
i i
Em
= Fm
+ Cm
A

On the mth ellipsoidal surface, Em and Dm must satisfy the

following boundary conditions:43,44

2.14a

CmyAyy

m
m
m

m
m
Dm = Dm
= m

are induced in the nano-onion with respect to the matrix. We

express Em in terms of electric field potential m to the form
Em = m

h1
3 m/

where h1, h2, and h3 are metrical coefficients,44 and Dm is

related to Em by

Dm

x
Pm

h1
1 m/

= h1
Em = Em
2 m/

exist and uniform polarizations

x
Fm

2.17

Cxn = Cny = Czn = 0,

x
Nm

Nmy

i
= Ai0ambmcm/2
Nm

2.24

m = 1,2, . . . ,n,

2.25

z
Nm

i = x,y,z; m = 1,2, . . . ,n,

2.26

PHYSICAL REVIEW B 70, 235103 (2004)

SURFACE PLASMA RESONANCE AND MAGNETO-

m = 2abc/ambmcm

m = 1,2, . . . ,n,

2.27

= 1 N
+ 1,
A
m
m m
m1
m1
m1

2.30a

i
Nm

where
is the depolarization factor along the ith axis for
the mth shell (core) and m is a constant inversely proportional to the volume inside the mth boundary surface.
The recurrent relations expressed by Eqs. (2.21) and
(2.22) can be rewritten to the form

Fm1
Fm
= T
m
Cm1
Cm

m = 1,2, . . . ,n,

= 1 N
N
1

B
m
m m1 m
m
m1
m
m
m1 ,
2.30b

2.28

is a super matrix whose components are given

where T
m
by the following matrices:

= Am Bm
T
m

C
m Dm

m = 1,2, . . . ,n,

2.30c

= 1 N
+ 1.
D
m
m
m1
m
m
m1

2.30d

They are expressed as

x
Nm
m m1 + m1

x
xy
xy
Nm
m
m1

x
xz
xz
Nm
m
m1

yx

Nmymyx m1

Nmym m1 + m1

yz
Nmymyz m1

= 1
A
m
m1

= 1
B
m
m m1

2.29

= 11 ,
C
m
m
m1
m m1

z
zx
m
Nm

zx
m1

z
zy
Nm
m

zy
m1

z
Nm
m

m1 + m1

x
x
Nm
Nm
1m m1

x y
xy
xy
Nm
Nmm
m1

x z
xz
xz
Nm
Nmm
m1

x
yx
NmyNm
myx m1

NmyNmy 1m m1

z
yz
NmyNm
myz m1

z
z
NmNm 1m m1

z x
xz
Nmm
Nm

xz
m1

z y
zy
Nm
Nmm

zy
m1

xy
xy
xz
xz
m m1 m
m1
m
m1

2.31a

= 11 yx yx m m1 yz yz ,
C
m
m
m1
m
m1
m m1

= 1
D
m
m1

zx
m

zx
m1

zy
m

zy
m1

m m1

x
Nm
m m1 + m

xy
xy
Nmym
m1

z
xz
xz
Nm
m
m1

x
yx
Nm
myx m1

Nmym m1 + m

z
yz
Nm
myz m1

x
zx
m
Nm

zx
m1

zy
Nmym

zy
m1

B. Polarizability of a nano-onion

In order to derive the polarizability for the nano-onion, let

us express Fm in terms of F0. Substituting Eq. (2.17) into Eq.
(2.28), we obtain

Fm

Cm

T
T

= T
m+1
m+2
n

Fn
0

F0

C0

m1 + m

2.31c

T
T
F .
F0 = T
1
2
n 1,1 n

2.31d

2.33b

Here 1,1 designates the first row first column components

of the super matrix. Combination of Eqs. (2.33a) and (2.33b)
results in

m = 1,2, . . . ,n 1,
T
T
T
T
T
1 F .
Fm = T
m+1
m+2
n 1,1
1
2
n 1,1 0

2.32a

z
Nm
m

2.31b

T
Fn ,
T
= T
1
2
n
0

2.32b

2.34

Thus the matrix Sm, which connects F0 to Fm by

where 0 expresses the matrix having only zero components,

and thus we obtain
T
T
F ,
Fm = T
m+2
n 1,1 n
m+1

2.33a

Fm = SmF0
is expressed as

235103-5

m = 1,2, . . . ,n,

2.35

PHYSICAL REVIEW B 70, 235103 (2004)

M. ABE AND T. SUWA

S = T
T
T
T
T
T
1 .
m
m+1
m+2
n 1,1
1
2
n 1,1

2.36

tmm 0S m.

2.44

m=1

For a spherical, core/shell structured particle n = 2, we confirmed that the polarizability given by Eq. (2.44) reduces to
that reported in literature.41

Some of Sms are given as follows:

n = 1 (nonstructured):
1 = A
1;
S1 = T
1 1,1
1

2.37
C. Effective dielectric permittivity tensor

n = 2 (core/shell structured):
S = T
T
T
1 = A
A
A

1
1
2 1,1
1
2 1,1
2
1 2 + B 1C 2 ,
2.38a
S = T
T
1 = A
A

1
2
1
2 1,1
1 2 + B 1C 2 ;

2.38b

n = 3 (core/inner-shell/outer-shell structured):
S = T
T
T
T
T
1
1
2
3 1,1
1
2
3 1,1

Now, we derive the effective dielectric permittivity tensor

for the composite containing the ellipsoidal nano-onions dispersed in a host medium at a volume fraction f. The nanoonions are, as already shown in Fig. 1, the same in shape and
orientation (principal axes parallel to x, y, and z directions)
but not necessarily in size. The exciting light field E is not
equal to F0 but is given by averaging the uniform, quasistatic
fields Fms over the matrix and the core and shells of the
nano-onions as
n

= A
2 3 + B2C3A1A2A3 + B2C3
C
A

1
+B
1
2 3 + D2C3 ,

E = 1 fF0 + f
2.39a

t mF m .

2.45

m=1

Substituting Eq. (2.35) into Eq. (2.45), we obtain

S = T
T
T
T
1
2
3 1,1
1
2
3 1,1

E =

A
A
A

1
=A
3
1
2 3 + B2C3 + B1C2A3 + D2C3
2.39b
S = T
T
T
1
3
1
2
3 1,1

The polarization Pm [cf. Eq. (2.13)] that is induced in the

mth medium with respect to the matrix having 0, satisfies
the relation
2.40

Summing up all the Pms given by Eq. (2.40), the total polarization induced in the nano-onion amounts to

P =

V mP m

m=1

2.41

Here we introduced tm as the fraction of the volume Vm of

the mth shell (core) to the volume Vo of the nano-onion,
expressed as

m = n.

= 0 + f

tmm 0S
m

m=1

+f
1 f1

= 0 + f
by defining

tmSm

2.49

tmm 0S m

m=1

= S 1 f1
+f
S
m
m

m=1

We can simplify Eq. (2.49) to

2.42a

2.50

m=1

tmSm

2.51

2.42b

Combining Eq. (2.35) and Eq. (2.46), we observe that S
m
connects Fm to the light field E by

2.43

E.
Fm = S
m

, for the nano-onion defined by

The polarizability
F0 ,
P =

2.48

is finally obtained by substituting Eqs. (2.43), (2.44), and

(2.46) into Eq. (2.47) to the form

tm = Vm/V0 = ambmcm am+1bm+1cm+1/abc

tn = Vn/V0 = anbncn/abc

2.47

D = E

tmm 0Fm .

m=1

m = 1,2, . . . ,n 1,

2.46

F0 .

for the comTherefore, the effective permittivity tensor

posite defined by

Vo =

D = 0E + fP.

2.39c

m = 1,2, . . . ,n.

tmSm

m=1

The total electric flux density D for the whole composite

relative to the matrix with the permittivity tensor 0 is expressed as

A
A

1
= A
1
2 3 + B2C3 + B1C2A3 + D2C3 .

mFm = 0Fm + Pm

+f
1 f1

is expressed from Eqs. (2.41) and (2.35) as

2.52

We can further rewrite Eq. (2.50) to a brief form as

235103-6

PHYSICAL REVIEW B 70, 235103 (2004)

SURFACE PLASMA RESONANCE AND MAGNETO-

n

mm ,

= + i = 1

2.53

m=0

where ms m = 1 n are the virtual volume fractions for

shells and core that are defined as

m = ftmS
m

m = 1,2, . . . ,n,

2.54a

and 0 is the virtual volume fraction for the matrix given

by
n

m .

0 = 1

2.54b

m=1

Now let us extend to the composites containing N

types of nano-onions that differ in shape and/or dielectric
constants. Describing the filling factor of the jth type j
= 1 , 2 , . . . , N of nano-onion ensemble as f j, the polarizability
is extended from Eq. (2.44) to the following:

j=1

nj

fj

j
tmjmj 0S

m
m=1

Here tmj and Smj are the volume fraction and field connecting
tensor, respectively, for the mth shell (or core) belonging to
the jth type nano-onions that have n j-fold core/shell structure. Equation (2.49) is then extended to

nj

= 0 +

m=1

j=1

j
0S
m

tmjmj

j=1

fj

j=1

nj

f 1 +
j

tmjmj

j
0S
m

m=1

= 0 +
as we set

j=1

nj

m=1

j=1

fj

j=1

nj

f 1 +
j

j ,
0S
m

j = S j
S
m
m

tmjmj

Here p is the volume plasmon angular frequency and the

relaxation constant that is responsible for the dielectric loss
or the imaginary dielectric function, , of the metal. The
real dielectric function becomes zero at = 2p 2,
where in bulk metal samples volume charge density waves
are resonantly excited. In the range , becomes
negative in sign, which facilitates the excitation of the surface charge density waves or the surface plasmon resonance.
Let us consider an unstructured n = 1 metal particle ellipsoidal in shape. When a quasistatic electric field Fx0 is
applied to the particle along the x direction by the deriving
light wave, a uniform electric field Fx1 is induced parallel to
the light field inside the particle. The field is obtained by
substituting Eqs. (2.31a) and (2.37) into Eq. (2.35) as
Fx1 =

m=1

tmjmj

j
0S
m

Fx0
1
=
Fx0 ,

1/S1xx A1xx

3.2a

x
x
= N11 + 1 N10 .
A
1 xx
0

3.2b

= 0,
Re1/S
1 xx

3.3a

= 0,
ReA
1 xx

3.3b

Therefore, when

or
1

and, thus

2.56

Equation (2.56) is simplified to the form

N

3.1

2.55

2p
2p
.
+
i
2 + 2 2 + 2

2.57

Re1 =

2.58

III. RESONANCE CONDITIONS AND SURFACE

PLASMON EIGENMODES

In this section, we derive the conditions in which dipolar

surface plasmon resonance occurs in the composites containing the n-fold nano-onions, and investigate the symmetry of
the plasmon eigenmodes from the polarity of the charges
induced on the metal surfaces. The metal in the nano-onions
is assumed to have dielectric function (: light angular
frequency) of Drude type due to free electron gas as
follows:31

Nx1

3.4

holds, the induced field Fx1 takes the resonant maximum

value given by35
Fx1 = i

1 Nx1

1
1
Fx0 = i
Fx0 .

Im1/S1xx
ImA1xx

3.5

Because this field is 90 out of phase with the driving optical

field Fx0, the particles absorb maximum light power at the
resonance frequency, similar to a resonantly driven cavity.46
The quality factor Q of the resonance is inversely propor .35 If the metal has no dielectric loss
tional to Im1 / S
1 xx
= 0, Fx and Q be = 0, or Im1 = 0 and thus Im1 / S
1 xx
1
come infinitely large. Therefore, even if the external field
Fx0 is absent, the inner field Fx1 oscillating at the resonance
frequency can exist, which induces inside the particle the
polarization and excites the charge density waves on the particle surface. This is the surface plasma resonance in an ellipsoidal metal particle.
For the composite containing the n-fold nano-onions, the
condition in which surface plasmon resonance occurs in the
mth shell (or core) is given by extending Eq. (3.3a) to

235103-7

PHYSICAL REVIEW B 70, 235103 (2004)

M. ABE AND T. SUWA

=0
Re1/S
m xx

m = 1,2, . . . ,n.

3.6

connects the light field E to the internal field F

Here S
m
m
as in Eq. (2.52). When resonated, the mth shell absorbs
maximum light power, which is analogous to the dissipative
m-fold coupled oscillators with dampers.45 By substituting
Eq. (2.36) into Eq. (3.6), we can rewrite the resonance condition as follows:

FIG. 3. Spherical nano-onions of alternate Na (shaded area) and

dielectric shells and core.

T
T
T
T
T
1 = 0.
ReT
1
2
n 1,1
m+1
m+2
n 1,1 xx

x
z
Nm
= Nmy = Nm
= 1/3,

3.7

If the nano-onions and the surrounding medium are nonmagxy

zx
= myz = m
= 0; m = 0 , 1 , . . . , n) having no dielecnetic (i.e., m
tric losses (i.e., Imm = 0; m = 0 , 1 , . . . , n), all the compo given in Eq. (2.31) become
nents of the super matrices T
m
diagonal tensors with real number components. Equation
(3.7) then reduces to the form
T
T
T
T
T
1 = 0,
T
1
2
n 1,1 xx
m+1
m+2
n 1,1 xx

3.8

T
T


because
both
T
and
T
1
2
n 1,1
m+1
1

Tm+2 Tn1,1 are also diagonal tensors. From Eq. (3.8)

we have n resonance angular frequencies, 1 , 2 , . . . , n,
which are the solutions for the nth order equation of :
T
T
= 0.
T
1
2
n 1,1 xx

3.9

We also have n-m anti-resonance frequencies, where zero

field is induced in the mth shell or core,46 obtained by solving the (n minus m)-th order equation of
T
T
= 0.
T
m+1
m+2
n 1,1 xx

1 / S
m xx

3.10

T
m+1

When
Eq.
(3.10)
holds,
1

Tm+2 Tn1,1xx diverges to infinities, and thus Eq.

(2.52) yields Fm = 0.
The charge density m induced on the mth boundary surface by the light field Ex is formulated from Eqs. (2.40) and
(2.52) as follows:

am = bm = cm

m = 1,2,3, . . . ,n.

4.1a
4.1b

Starting from a Na sphere n = 1, an inner core of a dielectric

and that of Na are added by turns, as shown in Fig. 3. Here
we assume
a1 a2 = a2 a3 = = an1 an = an ,

4.2

or the core radius is equal to each shell thickness. We also

assume that Na has p = 5.95 eV and = 0.31 eV as reported in literature,3 and the dielectric has = 10 independent
of photon energy, similar as assumed by Sinzig and
Quinten.39 Such a fictitious, high permittivity was chosen for
better recognition of the effect of adding the dielectric to the
metal. The nano-onions are dispersed in a matrix with 0
= 1 (as in vacuum, similar as Sinzig and Quinten39) at a volume fraction of f = 0.01. We assumed such a small filling
factor to conform to MG theory and to reduce the imaginary
, thus facilitating the resonance even
(or loss) term of S
m xx
at low photon energy range, where approaches .
For the composites containing the nano-onions with
stacking number n = 1 7, we calculated the light extinction
coefficient

= Imxx1/2

4.3

as a function of in the range 0.4 6 eV. Figure 4 shows

semi-logarithmic plots of , in which the spectra are shifted
along the ordinate by arbitrary factors for better presentation.
The n-fold nano-onions exhibit n resonant absorption peaks
at = 1 , 2 , . . . , n, except n = 7, where the lowest
resonant energy 7 runs off the lower limit of the calcula-

mx = Pmx Pm1
= m 0S
m xx
m1 0Sm1xx

Ex

m = 1,2,3, . . . ,n.

3.11

If no dielectric loss exists in the shells (or core), and thus no

phase retardance exists between the outer and inner surfaces
x
x
/ m1
whether the
of a metal shell, we can determine by m
surface plasmon polariton eigenmodes excited in the mth
shell are symmetric or anti-symmeteric.

IV. SURFACE PLASMONS IN COMPOSITE CONTAINING

Na/DIELECTRIC NANO-ONIONS
A. Resonant light extinction and symmetry of plasmon
eigenmodes

First, we consider nonmagnetic nano-onions of spherical

shape for simplicity, having

FIG. 4. Extinction coefficient spectra (shifted along the ordinate

by arbitrary factors) calculated for composites containing spherical,
n-fold Na/dielectric nano-onions at a volume fraction of f = 0.01.

235103-8

PHYSICAL REVIEW B 70, 235103 (2004)

SURFACE PLASMA RESONANCE AND MAGNETO-

calculated as a function of photon energy

FIG. 5. Re 1 / S
1 xx
for a composite which contains spherical, nonstructured nanoparticles of Na having as reported in the literature (shown by solid
line) and = 0 (dotted line).

tion range. The peak-position frequencies are very close to

the surface plasmon resonance frequencies, where
= 0 holds, as shown in Figs. 57 for n = 1, 2, and
Re1 / S
m xx
4, respectively. This is similar as in the forced harmonic
motion of the dissipative, coupled oscillators.45
with
In Figs. 57 we compare the spectra of Re1 / S
m xx
those calculated assuming = 0 in order to investigate the
effect of the dielectric loss. For the simple Na spheres n
= 1 the resonance occurs at 1 = 3.4 eV (Fig. 5), which
does not change appreciably by introducing the dielectric
loss. Figure 6 shows that as the dielectric core is inserted at
the center of the Na sphere, thus making n = 2, the resonance
energy 1 shifts to higher values, and another resonance
appears at a lower energy 2. When = 0 [Fig. 6a],
diverges to infinities at an anti-resonance enRe1 / S
1 xx
ergy a between 1 and 2. When 0 [Fig. 6(a)], the
is deformed to a continuous
diverging curve of 1 / S
1 xx
curve crossing the horizontal axis at a.
Figures 57 indicate that when = 0, each n-fold nanoonion has n eigenfrequencies 1 , 2 , . . . , n, and each mth
shell (or core) has n-m anti-resonance frequencies (a, b,
a, etc.), as is expected from Eqs. (3.9) and (3.10). Introducing the dielectric loss smears out some of the resonance and
the anti-resonance [e.g., 4 and c in Figs. 7(a) and 7a,
respectively].

m = 1 4 calculated for a composite

FIG. 7. Re 1 / S
m xx
which contains spherical, fourfold n = 4 nano-onions, using as
reported in the literature [(a)(d)] and = 0 [ad].

Assuming = 0, we calculated from Eq. (3.11) surface

charge densities induced on the mth surface, with which we
determined their polarity and thus the symmetry of the surface plasmon eigenmodes induced in the Na shells, as given
in Table I. Letters S and A in the table indicate symmetric
and anti-symmetric eigenmodes (cf. Fig. 8),25,39 respectively.
For example, when n = 4 (i.e., fourfold nano-onions) at 1,
charge densities , , , and in sign are induced on the
surfaces of numbers m = 1, 2, 3, and 4, respectively. Thus, the
Na shell has symmetric and anti-symmetric surface plasmon
modes on its outer (m = 1 and 2) and inner (m = 3 and 4)
shells, respectively. One will notice that in all nano-onions
with n = 2 5, the outermost (m = 1 and 2) Na shell has only
symmetric eigenmodes at 1, while it has only antisymmetric eigenmodes at other resonance energies, 2,
3, 4, and 5.
B. Magneto-optical enhancement

Applying an external magnetic field of B = 1 T along the z

axis, we calculate the magneto-optical Kerr effect for the
composite containing the spherical nano-onions. The offxy
xy
= m
diagonal dielectric function m
+ imxy for Na is expressed in terms of the cyclotron angular frequency c, as
well as p and for the free electrons as follows:46
xy
m
= myx = ic

2p
+ i2 2c

c = eB/m * .
and Re 1 / S
calculated for a comFIG. 6. Re 1 / S
1 xx
2 xx
posite which contains spherical, twofold n = 2 core (dielectric)/
shell (Na) structured nano-onions, using as reported in the literature [(a) and (b)] and = 0 [a and b].

4.4a
4.4b

Here, e and m* are charge and effective mass of the electrons, respectively. The polar Kerr rotation angle K and ellipticity angle K for the composite containing the nanoonions were calculated from the complex equation

235103-9

PHYSICAL REVIEW B 70, 235103 (2004)

M. ABE AND T. SUWA

TABLE I. Polarity of surface charge densities induced on the mth surfaces and symmetry (S: symmetric, A: anti-symmetric) of surface
plasmon eigenmodes generated in the Na shells in spherical, n-fold Na/dielectric nano-onions, calculated at respective resonance energies.
Parentheses indicate that the resonance is smeared as is changed from 0 to 0.31 eV, the literature value.

Resonance
energy

Surface
number
m

1
2
3
4
5

1
2
3
4
5

1
2
3
4
5

1
2
3
4
5

1
2
3
4
5

n=5
Polarity

n=4

Symmetric

+
+

+
+

Polarity

n=3
Polarity

Symmetric

Polarity

Symmetric

Polarity

+
+

+
+

+
+

+
+

+
+

+
+

+
+

n=1

Symmetric

n=2

K + iK =

ixy

xxxx 1

4.5

As an example, the case for n = 2 is shown in Fig. 9. The

diagonal and off-diagonal elements of the complex effective
dielectric tensor, xx =xx + ixx and xy =xy

FIG. 8. Surface charges induced in the mth shell by (a) symmetric and (b) antisymmetric surface plasmon eigenmodes.

+ ixy, exhibit resonant dispersion relations in close vicinity of the surface plasmon resonance frequencies as shown in
Fig. 9. However, the Kerr rotation K and ellipticity K exhibit a resonant dispersion relation at frequencies slightly
different from the surface plasmon resonance frequencies.
This is because at these frequencies xx 1, which appears
in the denominator on the right side of Eq. (4.5), approaches
zero.
Figure 10 shows K and K for n = 3. Because Na fills the
composite by factor ft1 + t3, we defined Kerr effect enhancement factors by
MK =

kPcomp
,
ft1 + t3KNa

4.6a

MK =

KPcomp
.
ft1 + t3KNa

4.6b

Here, KPcomp [or KPcomp] expresses the resonant peak

height of the Kerr rotation (ellipticity) for the composite and

235103-10

PHYSICAL REVIEW B 70, 235103 (2004)

SURFACE PLASMA RESONANCE AND MAGNETO-

TABLE II. Enhancement factors for Kerr rotation and ellipticity

for a composite containing spherical, threefold n = 3 nano-onions
calculated at the respective resonance energies for two values of .
MK

MK

= 0.31 eVa = 0.06 eVb = 0.31 eVa = 0.06 eVb

1
2
3
aReported
bReduced

7.7
56
6.1

82
1166
159

10
88
6.4

82
1698
85

in the literature.
by a factor 5 from the literature value.

hancement factor is augmented as the Q factor of the resonance becomes high.

C. Effect of the shape of nano-onions

FIG. 9. Spectra for effective dielectric permittivity tensor elements and Kerr effect calculated for a composite containing spherical, twofold Na/dielectric nano-onions.

KNa KNa expresses the rotation (ellipticity) calculated for Na bulk sample at resonance frequencies. As shown
in Table II, we obtained for n = 3, MK and MK
6 90, which increased about one order of magnitude, to
80 1700, when was hypothetically reduced by a factor
5, from 0.31 to 0.06 eV. Thus, the magneto-optical en-

FIG. 10. Kerr effect spectra calculated for a composite containing spherical, threefold Na/dielectric nano-onions.

Next, we changed the shape of the nano-onions to spheroidal prolate Nx = Ny = 0.35 and oblate Nx = Ny = 0.28 to
determine how the depolarization factor affects the magnetooptical response. For the composites containing the spheroidal, twofold n = 2 nano-onions we calculated xy and Kerr
effect. The results are shown in Figs. 11 and 12. Comparing
Fig. 11 (prolate) with Fig. 9 (spherical), one notices that
increasing the depolarization factor Nx along the light field
direction augments the resonant peak heights of xy, K,
and K at 1 slightly, but does not change appreciably the
peak heights at 2. On the other hand, Fig. 12 (oblate) shows
that decreasing Nx deforms the resonant curves and decreases
their peak heights of the Kerr effect at both 1 and 2.

FIG. 11. Off-diagonal element for effective dielectric permittivity tensor and Kerr effect calculated for a composite containing
prolate, twofold Na/dielectric nano-onions.

235103-11

PHYSICAL REVIEW B 70, 235103 (2004)

M. ABE AND T. SUWA

FIG. 12. Off-diagonal element for dielectric permittivity tensor

and Kerr effect calculated for a composite containing oblate, twofold Na/dielectric nano-onions.
V. DISCUSSION AND CONCLUSION

By applying the MG theory, which is based on the quasistatic approximation, we formulated the effective dielectric
permeability tensors for the composites containing ellipsoidal, n-fold nano-onions as given by Eq. (2.49), or in more
simplified forms of Eqs. (2.50) and (2.53). When there are
two or more different types of nano-onions dispersed in the
matrix, the formula is extended as given by Eq. (2.57). Because the polarizations induced in the nano-onions are
treated as point dipole moments in our quasistatic approximation, the size of the nano-onions does not appear in the
effective dielectric tensor explicitly, but enters as the filling
factors f and tm.
Using the effective dielectric tensor, we analytically formulated the resonance conditions for the dipolar surface
plasmons as Eq. (3.6). We also derived Eq. (3.11), a formula
giving the surface charge density induced on the shell surfaces, with which we can determine the symmetry of the
surface plasmon polariton eigenmodes excited in the metal
shells.
On the composites containing model nano-onions spherical in shape having n-fold Na/dielectric alternate stratification, our findings are summarized as follows.
(1) If the dielectric loss of Na is neglected, there are n
surface plasmon eigenmodes, in each of which an infinitely
large electric field is induced in every shell or core. In addition in the mth shell or core, anti-resonance (where zero field
is induced) occurs at n-m frequencies. The occurrence of the
resonance and anti-resonance is similar as found in the n-fold
coupled oscillators.45
(2) As the dielectric loss is introduced, the light extinction coefficient exhibits resonant peaks at the surface plasma
frequencies, although some of the resonance and anti-

resonance (which occurred in case of no dielectric loss) are

smeared out.
(3) The magneto-optical Kerr effect, as well as the offdiagonal dielectric permittivity tensor elements, is prominently enhanced by the surface plasmon resonance, especially when the dielectric loss is weak and the resonance has
a high quality factor Q. For the Na/dielectric double layered
nano-onions, we found that changing the particles shape to
prolate spheroidal (and thus increasing the depolarization
factor) increases the resonant peak heights of xy, K, and
K at 1 only slightly.
We47 have given a preliminary report on the effective permittivity tensor [Eq. (2.49)], with only a brief outline of its
derivation, and have calculated the magneto-optical Kerr effect in composites containing nano-onions of Fe/ Au
multicore-shell structures; they did not show definite features
of plasma resonance in Au shells. We48 later found that inserting a dielectric layer between the Fe core and Au shell,
and thus increasing the Q factor of the resonance, enhances
Kerr effect. This is similar as increasing Q by decreasing
in Na enhances Kerr effect in the composite containing Na/
dielectric nano-onions.
The composite containing threefold Na/dielectric nanoonions have large Kerr effect enhancement factors [MK
and MK 6 90], which are, however, augmented by a
factor 1 / ft1 + t2 =1 / 135; actual K and K are reduced by
factors of 0.050.7 from those for bulk Na. The figures of
merit for the Kerr effect, defined by R1/2K and R1/2K (R:
reflectivity), are reduced more, by factors of 0.010.14, because the composite has a much lower reflectivity R
0.04 than Na bulk R 0.9. This, as well as the combustible nature of Na, makes the composites containing the Na/
dielectric nano-onions unfeasible for practical applications,
although they provide a good model to study optical surface
plasmon resonance.
In our calculation, the filling factor of the nano-onions
was fixed to f = 0.01, small enough to conform to the MG
theory that was derived assuming sparse dispersion of the
particles. Let us reexamine the low filling factor limit in the
calculations based on the MG theory. The MG theory26 was
derived originally utilizing the Rayleigh scattering theory, a
kind of quasistatic approximation, which assumes that homogeneous fields are induced within and outside the particles by
the light field. In other words, the MG theory presupposes
that the particles are much smaller than the light wavelength
and the particles are sparsely filling the matrix. However, the
MG theory has been applied beyond the low filling factor
limit, and is justified by experimental results. Kreibig et al.49
revealed that the plasmon resonance absorption spectra for
the composites containing Ag spherical nanoparticles
(8 75 nm in diameter) are fitted by the MG theory rather
well even for high filling factor, up to f = 0.4, as long as the
particles were not coagulated. When the composites had partial coagulation, additional peaks appeared, which was explained in term the surface plasma resonance in various types
of clusters of the Ag nanoparticles. Therefore, the MG theory
is shown to be applicable even for high volume fraction of
the particles when no coagulation aggregates exist.
One can explain this as follows. The MG theory is considered as a generalization of the Clausius-Mossotti theory50

235103-12

PHYSICAL REVIEW B 70, 235103 (2004)

SURFACE PLASMA RESONANCE AND MAGNETO-

that gives the dielectric constant for crystals, i.e., the ensembles of spherical atoms in vacuum to the ensembles of
the small particles in the matrix. In the MG theory, the local
field acting on the particles is approximated by the Lorentz
local field, similar as in Clausius-Mossotti theory.31,42,51 The
Clausius-Mossotti theory is applicable to cubic crystals or
completely disordered amorphous crystals, in which vacuum
is embedded with atoms at a fairly large filling factor, even
up to 0.74, the maximum value obtained for bcc or hexagonal close packing of spherical atoms. Therefore, the MG
theory will be applicable even up to large value of f, when
the matrix surrounding the particles are composed of grains
much smaller than the particles (i.e., the host material can be
considered as continuous, similar as vacuum in the ClausiusMossotti theory) and the particles are uniformly dispersed
without being coagulated. If the grains of the surrounding
matrix are not sufficiently smaller than the particles, we must
use the Bruggeman theory27 to symmetrize the roles of inclusion (particles) and the host (matrix) when the filling factor is not small.51 If the particles are coagulated, the effects
by clustering of the particles must be taken into account.
Therefore, the MG theory will be applicable even when f is
large for an ideal composite in which the nano-onions are
dispersed without coagulation in a matrix composed of very
small grains.
Now let us discuss the limit inherent in the calculation by
the MG theory, comparing with that by the rigorous Mie
scattering theory. For Ag spheres with radius a, the limit
allowing 10% difference between the calculations by the MG
theory and the Mie theory is estimated as a 0.03 (:
light wavelength).33,35,37 This is about the size perimeter usually taken to delimit the small particle Rayleigh approximation (based on which the MG theory is derived) from the
accurate Mie theory, which amount to a = 6.9 and 105 nm at
our measurement boundaries = 6 and 0.4 eV, respectively.
As we already described, Sinzig and Quinten39 revealed
using the Mie scattering calculation that the spherical nanoonions (having diameter 2a = 2 12 nm), which have similar
Na/dielectric stratified structure n 7 dispersed in vacuum
as ours, exhibit similar plasmon resonant absorption peaks.
The background of the extinction spectra = 0.5 6 eV
rises prominently at higher photon energies, while our extinction spectra (Fig. 4) do not show such a particular increase. This is because the light scattering intensity, calculated by the Mie theory (which involves space dependence
of, as well as multipole scattering by, the polarization and
field induced in particles), increases as a / increases at
higher photon energies.33 However, it should be noted that
by the Mie theory the light extinction is calculated from the
scattering cross section by a single particle, without considering any mutual interactions between the particles. The Mie
calculation is hence limited to the very sparse dispersion of
the nano-onions. On the other hand, our calculation is applicable to more concentrated dispersion of the particles, because the MG theory incorporates the interparticle interaction through the local Lorentz field acting on the particles.
It should be also noted that the dielectric function reported for bulk metal samples may not be directly applicable
to small particles. If the dimension d (core radius and shell
thickness) for core/shell structured particles becomes compa-

rable to the mean free path of the free electrons, the relaxation constant for the metal core or shell(s) is modified
from the bulk value to21,22,29

= 1 + /d.

5.1

Here, the term / d expresses the effects of free electron collision at the surface. For Na is reported to be 34 nm.31
Furthermore, for small particles we must consider the quantum size effect and defects, which will also change the dielectric function of the particles.
In conclusion, we have clarified the physical picture of the
surface plasma resonance occurring in the n-fold stratified
Na/dielectric nano-onions, and revealed that magneto-optical
effect, as well as light extinction, is prominently enhanced by
the surface plasmon resonance in the Na shells and cores.
Our calculation, based on the MG theory, should prove useful for the composites containing nano-onions even at a high
volume fraction as far as the particles much smaller than the
light wavelength are dispersed without coagulation and the
dielectric functions of the cores and shells of the nano-onions
are known.
ACKNOWLEDGMENTS

The authors are indebted to Mr. J. Kuroda for his help in

developing the computer program and performing the calculations. They are thankful to Prof. Vince Harris of Northeastern University for the critical reading of the manuscript.
APPENDIX: DERIVATION OF EQS. (2.21) AND
(2.22)

Substituting Eq. (2.14) into Eq. (2.18), we obtain the electric field on the mth boundary surface as follows:

m =
Em

i h11

i
Em

ui Emi i
+
u

=
m

i Emi ni
m

A1a

Em
m =

m =
Em

i
h1
2 Em

i
h1
3 Em

ui

ui

i Emi ni

i Emi ni

, A1b

. A1c

Here m indicates = m, and we defined

Ei = Fi + Ni 1Ci
m
m
m
m
m

i = x,y,z,

A2

, 1 given by Eqs. (2.24)(2.27). We also introwith m, N

m
duced unitary vectors along the , , and coordinates as
follows:44

235103-13

PHYSICAL REVIEW B 70, 235103 (2004)

M. ABE AND T. SUWA

n = nx,ny,nz =

1 x 1 y 1 z
,
,
,
h1 h1 h1

1 x 1 y 1 z
,
,
,
n = nx,ny,nz =
h2 h2 h2
n = nx,ny,nz =

m
m =

A3a

1 x 1 y 1 z
,
,
.
h3 h3 h3

= m
A3b
+
A3c

n n = 1,

A4a

n n = 0,

A4b

n n = n ,

A4c

n n = n ,

A4d

n n = n .

A4e

m =
m

=
m

A y

,
3
am
b mc m

xm = 2am

x
,

=
m

=
m

1
3
a mb mc m

ym = 2bm

y
,

1
3
a mb m
cm

= m

m = 1,2, . . . ,n,

z
,

A6

m
Em +

m
Em

m
E m m .

A10

x i, j = y,z

A11

ij
m
nki,j

ij

A12

Substituting Eqs. (A1a), (A9), (A10), and (A12) into Eq.

(A8), we get

Dm
m = m

i
ij
Em
m
nki,jni
i Emi ni +
ij
m

+ nki,jnim
= m

i
ij
Em
m
nki,ki,j
i Emi ni +
ij
m

A13
which is further transformed to

x + xyEy + xzEz n
m = mE
Dm
x m
m m
m
m m
m
m m
x
y + yzEz n
+ myxEm
m + mE
y m
m m
m
m m
zx x
zy y
z n .
+ m
Emm + m
Emm + mE
z m
m m

A7

in order for Eq. (A6) to hold for arbitrary values of and .

Substituting Eq. (A1c) into Eq. (2.20b) as well as gives Eq.
(A7).
The left side of the remaining boundary condition [Eq.
(2.20c)] is expanded as

Dm
m =

A9

ki,z = y i, j = z,x .
z i, j = x,y

i
i
m and Em1
m are constants but nim is a funcSince Em
i
i
m = Em1
m i = x , y , z, or
tion of and , we must have Em
in vector form,
i
i
Em
m = Em1
m

= m ,

in which we used Eq. (A4b) and introduced the suffix k i , j

as

which are derived from Eqs. (2.4), (2.16), and (2.24)(2.27).

Substituting Eq. (A1b) into the boundary condition of Eq.
(2.20a), we obtain
m

ij
ni ni + m
nin j n jni

i,j
ij

ij
m
nki,j

ij

A5b

i
ni
i Emi ni = i Em1

ij
m
+ mjinin j

ij

m
m =

A5a

zm = 2cm

Similarly, we obtain

i ni ni

ij
m
ni n j

i,j

In deriving Eq. (A1), we used, in addition to Eq. (A4), the

relations

Ax

by using Eqs. (2.8b) and (A4a). In a similar way we obtain

m as
m

They satisfy the orthonormal relations

ij
m
nin j

i,j

A14
Replacing the suffix m with m 1 in Eq. (A14) gives

x + xy Ey + xz Ez
m = m1E
Dm1
m1 m
m1 m1 m
m1 m1 m

A8

m by the unitary transformation as

Here, we obtain m
follows:

235103-14

yx
x
y
Em1
m + m1E
nxm + m1
m1 m
yz
z
zx
x
+ m1
Em1
mnym + m1
Em1
m
zy
y
z n . A15
+ m1
Em1
m + m1E
z m
m1 m

PHYSICAL REVIEW B 70, 235103 (2004)

SURFACE PLASMA RESONANCE AND MAGNETO-

Substituting Eqs. (A14) and (A15) into Eq. (2.20c), we obtain the following vector formulas:

Fm1 + mN
m m1 = Fm + mNmCm ,

m1E
m1 m
m1 m11Em1m

1C
m1Fm1 + mm1N
m
m1
m1

+ 1E
= mE
m m
m
m
m m
m = 1,2, . . . ,n.

1C .
= mFm + mmN
m
m
m

A16

By substituting Eqs. (2.14d) and (A2) into Eqs. (A7) and

(A16), we get

1 J.

A17

Wiggins, E. E. Carpenter, and C. J. OConnor, J. Appl. Phys.

87, 5651 (2000).
2
T. Katayama, H. Awano, and Y. Nishihara, J. Phys. Soc. Jpn. 55,
2539 (1986).
3
H. Raether, Excitation of Plasmons and Interband Transitions by
Electrons (Springer-Verlag, Berlin, 1980), p. 116.
4
H. Raether, Surface Plasmons on Smooth and Rough Surfaces
and on Gratings (Springer-Verlag, Berlin, 1988).
5 W. Steinmann, Phys. Status Solidi 28, 437 (1968).
6 P. E. Ferguson, O. M. Stafsudd, and R. F. Wallis, Physica B & C
89, 91 (1977).
7 P. M. Hui and D. Stroud, Appl. Phys. Lett. 50, 950 (1987).
8 V. A. Kosobukin, Surf. Sci. 406, 32 (1998).
9 V. A. Kosobukin, Solid State Commun. 101, 497 (1997).
10 V. A. Kosobukin, J. Magn. Magn. Mater. 153, 397 (1996).
11 G. Tessier and P. Beauvillain, Appl. Surf. Sci. 164, 175 (2000).
12 T. V. Murzina, T. V. Misuryaev, A. F. Kravets, J. Gdde, D.
Schuhmacher, G. Marowsky, A. A. Nikulin, and O. A. Aktsipetrov, Surf. Sci. 482-485, 1101 (2001).
13 S. Kawata, Near-Field Optics and Surface Plasmon Polaritons
(Springer-Verlag, Berlin, 2001).
14 O. A. Aktsipetrov, E. M. Dubinina, S. S. Elovikov, E. D.
Mishina, A. A. Nikulin, N. N. Novikova, and M. S. Strebkov,
Solid State Commun. 70, 1021 (1989).
15 S. L. McCall, P. M. Platzman, and P. A. Wolff, Phys. Lett. 77A,
381 (1980).
16 J. Gersten and A. Nitzan, J. Chem. Phys. 73, 3023 (1980).
17 P. V. Kamat, M. Flumiani, and A. Dawson, Colloids Surf., A 202,
269 (2002).
18 J. Homola, S. S. Yee, and G. Gauglitz, Sens. Actuators B 54, 3
(1999).
19 W. Lukosz, Biosens. Bioelectron. 12, 175 (1997).
20 U. Kreibig and P. Zacharias, Z. Phys. 231, 128 (1970).
21 C. G. Granqvist and O. Hunderi, Phys. Rev. B 16, 3513 (1977).
22 G. A. Niklasson and C. G. Granqvist, J. Appl. Phys. 55, 3382
(1984).
23 T. K. Xia, P. M. Hui, and D. Stroud, J. Appl. Phys. 67, 2736
(1990).
24 M. Inoue, Phys. Rev. B 36, 2852 (1987).
25 P. Royer, J. P. Goudonnet, R. J. Warmack, and T. L. Ferrell, Phys.
Rev. B 35, 3753 (1987).
26 J. C. Maxwell Garnett, Philos. Trans. R. Soc. London 203, 385
(1904).

A18

Eliminating Cm1 or Fm1 from Eqs. (A17) and (A18) gives

Eq. (2.21) or Eq. (2.22), respectively.

A. G. Bruggeman, Ann. Phys. (Leipzig) 24, 636 (1935).

G. Granqvist and O. Hunderi, Z. Phys. B 30, 47 (1978).
29
C. A. Murray, J. Opt. Soc. Am. B 2, 1330 (1985).
30 G. Mie, Ann. Phys. (Paris) 25, 377 (1908).
31
U. Kreibig and M. Vollmer, Optical Properties of Metal Clusters
(Springer, Berlin, 1995), pp. 1531, 134, and 141.
32
U. Kreibig, B. Schmitz, and H. D. Breuer, Phys. Rev. B 36, 5027
(1987).
33 M. Kerker, D. S. Wang, and H. Chew, Appl. Opt. 19, 4159
(1980).
34 B. J. Messinger, K. Ulrich von Raben, R. K. Chang, and P. W.
Barber, Phys. Rev. B 24, 649 (1981).
35 A. Wokaun, Mol. Phys. 56, 1 (1985).
36 M. Kerker and C. G. Blatchford, Phys. Rev. B 26, 4052 (1982).
37 P. Barnickel and A. Wokaun, Mol. Phys. 67, 1355 (1989).
38 S. Remita, G. Picq, J. Khatouri, and M. Mostafavi, Radiat. Phys.
Chem. 54, 463 (1999).
39 J. Sinzig and M. Quinten, Appl. Phys. A: Solids Surf. 58, 157
(1994).
40 P. H. Lissberger and P. W. Saunders, Thin Solid Films 34, 323
(1976).
41 R. Carey, D. M. Newman, and P. Bouvier, IEEE Trans. Magn.
33, 3256 (1997).
42 M. Abe, Phys. Rev. B 53, 7065 (1996).
43 L. D. Landau, E. M. Lifshitz, and L. P. Oitaevskii, Electrodynamics of Continuous Media, 2nd ed., translated by J. B. Sykes, J. S.
Bell, and M. J. Kearsley (Pergamon, Oxford, 1981), p. 19.
44 J. A. Stratton, Electromagnetic Theory (McGraw-Hill, New York,
1941), pp. 34, 48, and 207.
45 K. U. Ingard, Fundamentals of Waves & Oscillations (Cambridge
University Press, New York, 1988), p. 113.
46 J. G. Mavroides, in Optical Properties of Solids, edited by F.
Abeles (North-Holland, Amsterdam, 1972), p. 357.
47 M. Abe and J. Kuroda, J. Appl. Phys. 91, 7305 (2002).
48 M. Abe and J. Kuroda, to be presented at the MMM Conf., Jacksonville, 711 Nov. 2004.
49 U. Kreibig, A. Althoff, and H. Pressmann, Surf. Sci. 106, 308
(1981).
50 H. Frhlich, Theory of Dielectrics: Dielectric Constant and Dielectric Loss, 2nd ed. (Oxford University Press, London, 1986).
51 R. Landauer, Electrical Transport and Optical Properties of Inhomogeneous Media, edited by J. C. Garland and D. B. Tanner,
AIP Conf. Proc. No. 40 (AIP, New York, 1978), p. 2.
27 D.
28 C.

235103-15