Chapter 1

Optical Properties of Plasmonic

Materials
Maxwell’s equations state that a dielectric–metallic interface can support
surface plasmon polaritons (SPPs), which are coherent electron oscillation
waves that propagate along the interface with an electromagnetic wave. The
unique properties of the interface waves result from the frequency-dependent
dispersion characteristics of metallic and dielectric materials. This chapter
provides an introduction to alternative plasmonic materials, as well as the
rationale for each material choice. The comprehensive optical properties of
various materials, including noble metals and semiconductors, are presented.
The optical properties are evaluated based on the permittivity and permeability
defined by either the Drude or Lorentz model. Furthermore, the noble metals
are described from the generally approved data in a general handbook of solid
materials, such as the Handbook of Optical Constants of Solids, edited by Palik.
This chapter outlines the effective medium approaches for describing the
effective dielectric functions of composite nanostructures. It also provides a
reference for finding better plasmonic materials at specific frequencies.

1.1 Electromagnetic Waves Propagating through Materials

1.1.1 Fundamental equations of electromagnetic waves
The modern electromagnetic wave theory has been developed from the group
work in theory by James Clerk Maxwell, Oliver Heaviside, and Josiah Willard
Gibbs from 1861 to 1865, and verified experimentally by Heinrich Hertz in
1887. During the establishment processes of the electromagnetic wave theory,
Maxwell’s contribution was settled in his two famous papers: “On Physical
Lines of Force” in 1861 and “A Dynamical Theory of the Electromagnetic
Field” in 1865. In his theory, four equations composed the fundamentals of the
electromagnetic wave theory and are now universally known as Maxwell’s
equations. They demonstrate the unifying connection between electromagnetic
waves and light, from the extremely long wavelengths of radio, television,

1
2 Chapter 1

radar, and microwaves, to the shorter wavelengths of visible light and very
short ultraviolet light.
The electromagnetic wave theory describes the fundamental concepts of
the interaction processes between matter and electromagnetic energy in
numerous scientific disciplines. Since the 20th century, it has been understood
that the theory of quantum electrodynamics can give a more accurate and
fundamental explanation of some phenomena related to photons, photon
scattering, and quantum optics. Regardless, the sophisticated electromagnetic
wave theory has been furthering our comprehension of the flourishing
scientific disciplines of plasmonic optics.
In electromagnetic wave theory, the conventional Maxwell’s equations are
formulated via vector notation in two forms of integral equations and
differential forms. The two forms are mathematically equivalent and useful in
same physical meaning. The integral equations are often used to directly
calculate fields in the conditions of symmetric distributions of electric charges
and electric currents. At the same time, the different equations are convenient
formulations in more complicated situations, such as using the finite element
analysis method.
There are also microscopic and macroscopic variants of Maxwell’s
equations. The differential forms of macroscopic Maxwell equations are as
follows:
­ ­
∇E¼ B, ∇  H ¼ D þ Jext , (1.1a)
­t ­t

∇ · B ¼ 0, ∇ · D ¼ r, (1.1b)

where E (volts/m) and H (amperes/m) describe the electric and magnetic

field, respectively. D (coulombs/m2) and B (webers/m2) correspond to the
electric displacement and the magnetic flux density vectors, respectively.
These four variables are generally time-dependent and position-dependent
parameters, which are created by electric charges or electric currents and
thus expressed by the local charge density per unit volume r (coulombs/m3)
and the external current density Jext (amperes/m2).
From the macroscopic point of view, Eqs. (1.1a) describe Faraday’s law of
induction and Ampère’s circuital law. The former states that a time harmonic
magnetic field induces an electric field, whereas the latter states that both an
electric current and a varying electric field can generate a magnetic field.
Equations (1.1b) give Gauss’ laws for magnetic and electric fields. Gauss’ laws
for electric fields formulate the conversion between a static electric field and
the electric charges; Gauss’ law for magnetic field states that the sum total
magnetic flux through any finitude volume surface is zero, which means that
there are no magnetic charges. In a 3D vector system, the parameters B, H, D,
r, and Jext are considered to be functions of both position and time (r, t).
Optical Properties of Plasmonic Materials 3

In the macroscopic Maxwell’s equations, the total electric charge density r

is factorized into a bound component rb and a free counterpart rf, just as the
total current density J is separated into a free component Jf and a bound
component Jb:

r ¼ rb þ rf , J ¼ J b þ Jf : (1.2)

The bound charge density rb and current density Jb are defined by introducing
the terms of polarization P and magnetization M in the form of
­P
rb ¼ ∇ · P, Jb ¼ ∇  M þ : (1.3)
­t
The conservation law for the electric charge density and the external current
density is described as
­
∇  Jext ¼  r: (1.4)
­t
Equation (1.4) indicates that the divergence of the current density leaving a
volume equals the decrease rate of the charge density in the same volume. This
equation can be obtained from Eqs. (1.1) by taking the divergence of
Ampère’s circuital law and introducing Gauss’ laws for electric fields.
In the case of a nonmagnetic medium, where only the free current density
remains, the magnetization and the subsequent bound current density are
zero. The continuity equation [Eq. (1.4)] becomes the partial component
corresponding to the free charges
­
∇  Jf ¼  r: (1.5)
­t f
In a bulk matter without electric charges or currents, the macroscopic
Maxwell’s equations are presented in the form of the microscopic variant:
­B
∇ · E ¼ 0, ∇  E ¼ m0 , (1.6)
­t

­E
∇ · H ¼ 0, ∇  H ¼ ε0 : (1.7)
­t
Equations (1.6) and (1.7) describe an electric and magnetic field in a vacuum,
where m0 and ε0 are the permittivity and permeability in the free space of
vacuum (m0  4p  10–7 henry/meter and ε0  8.85  10–12 farad/meter).
They refer only to the cases that do not account for the medium without
charges and currents; they do not mean that the space of the medium is empty
of charge or current. For an oscillating wave exp(jvt), Eqs. (1.1a) can be
rewritten in the simple form of multiplication by the time derivatives:
4 Chapter 1

∇  E ¼ jvB, ∇  H ¼ jvD þ Jext , (1.8)

where v ¼ 2pf represents the angular frequency of an oscillating wave. A
time-oscillatory plane wave propagating in 3D notation is

uðr,tÞ ¼ A expðjvt þ jkrÞ, (1.9)

where the complex amplitude A refers to any one component of the
aforementioned field vectors. The wavenumber k is defined by the number of
wavelengths per 2p units

k ¼ 2p=l ¼ 2pf =yp ¼ v=yp , (1.10)

where f refers to the frequency of the propagating wave, l is the wavelength in

a media, and yp is the phase velocity of the wave field. A series of sources
generate a linear combination of time-dependent wave fields:

X
N
Uðr,tÞ ¼ Ai expðjvi t þ jk i rÞ: (1.11)
i¼1

In general cases, the complex amplitude Ai depends on the initial and

boundary conditions. The combined fields are called the spatial spectrum. The
time-oscillating propagating waves can also be described in any specific axis.
The notation r in Eqs. (1.9) and (1.11) is replaced by another displacement
vector, such as in the x, y, and z axes.

1.1.2 Constitutive equations of inhomogeneous media

The constitutive equations of a medium specify the physical kinetic response
to external stimuli combined with other physical laws. In electromagnetic
wave theory, they are applied to describe the electrical and magnetic response
of various media. In Maxwell’s macroscopic equations, the constitutive
equations describe the relations between the displacement field D and the
electric field E, and the magnetizing field H and the magnetic field B. In other
words, they describe the dynamic response of bound charges and current to
external applied fields.
Within the vast majority of isotropic materials, the electric and magnetic
fields are often investigated separately. The constitutive relations of general
materials without polarization and magnetization are commonly written as:

D ¼ εE, H ¼ B=m, (1.12)

where ε and m are the absolute permittivity and permeability of a general
medium, respectively. Even for simple linear media, the constitutive relations
Eq. (1.12) have various complications. For example, in homogeneous materials,
ε and m are constant values throughout the media, while in inhomogeneous
Optical Properties of Plasmonic Materials 5

materials, they are position dependence within the materials. For anisotropic
materials, ε and m are in the form of vectors, while for isotropic materials, they
are denoted as scalars. Because general materials are dispersive, both ε and m
are dependent on the frequency of electromagnetic wave.
The vast majority of natural materials are electrically neutral at the
macroscopic level. This macroscopic electrical neutrality is the consequence
of the internal equilibrium of collective charge interactions. When an electro-
magnetic wave impinges on electrical neutrally media, the time-dependent
electric and magnetic field will induce separate oscillatory charge displace-
ments. Such local separation of positive and negative charges from their
original positions in opposite directions manifests the media to present in the
form of induced electric dipole momentum. These phenomena are called
polarizations. Similar effects induced by magnetic field are magnetization or
magnetic polarization. For an inhomogeneous medium with polarization and
magnetization, the continuous constitutive relations are described to be the
functions of space coordinates and time variables:2

Dðr,tÞ ¼ ε0 Eðr,tÞ þ Pðr,tÞ, (1.13a)

Bðr,tÞ ¼ m0 Hðr,tÞ þ Mðr,tÞ: (1.13b)

Equation (1.13a) indicates that a dielectric medium is characterized by a free-

space part ε0E(r, t) and a polarization vector P(r, t). The latter represents the
electric dipole moment. The electric flux density D represents the organization
of electric charges induced by an external electric field E. In the presence of
the external field, the polarization vector is caused by induced dipole
moments, alignment of the permanent dipole moments, and the migration of
electric charges. Equation (1.13b) states that a magnetic medium can also be
described by a free-space part m0H(r, t) and a magnetization vector M(r, t).
A medium is called diamagnetic if m , m0, whereas it is paramagnetic if
m . m0. For a diamagnetic medium, the induced magnetic moments tend
toward the opposite direction of the external magnetic field, whereas a
paramagnetic medium features an alignment of magnetic moments within the
medium.
The instantaneous polarizations of a medium to an electromagnetic field
are the convolution integration over the previous history response:2
ðt
Pðr,tÞ ¼ ε0 ke ðr,t  tÞEðr,tÞdt, (1.14a)
`

ðt
Mðr,tÞ ¼ m0 km ðr,t  tÞHðr,tÞdt, (1.14b)
`
6 Chapter 1

where the scalar functions of ke and km are the electric and magnetic
susceptibilities, both of which are time-dependent scalar. In the frequency
domain, the aforementioned induced polarizations are represented by a
multiplication operation:

Pðr,vÞ ¼ ε0 ke ðr,vÞEðr,vÞ, (1.15a)

Mðr,vÞ ¼ m0 k m ðr,vÞHðr,vÞ, (1.15b)

where ke(r, v) and km(r, v) are the Fourier-transformed kernels correspond-
ing to the electric and magnetic susceptibilities ke and km, respectively.
The constitutive relations of an inhomogeneous medium in the frequency
domain are:2

D ¼ ε0 ½1 þ ke ðr,vÞE ¼ ε0 εr ðr,vÞE ¼ εðr,vÞE, (1.16a)

B ¼ m0 ½1 þ km ðr,vÞH ¼ m0 mr ðr,vÞH ¼ mðr,vÞH, (1.16b)

where εr(r, v) ¼ 1 þ ke(r, v) and mr(r, v) ¼ 1 þ km(r, v) are the relative

permittivity and relative permeability in frequency region, respectively.
ε(r, v) ¼ ε0εr(r, v) and m(r, v) ¼ m0mr(r, v) are the corresponding absolute
permittivity and absolute permeability. Here, both the electric field E and the
magnetic field H are complex amplitudes. From the macroscopic point of
view, these vector quantities are the averaging of the microscopic electric and
magnetic field vectors over the unit cell of a medium.
In plasmonic optics, the complex artificial nanostructures, such as cubic
lattice, may be constructed by kinds of nanoparticles, even with different
natural substrates. In such inhomogeneous conditions, the parameters εr(r,v)
and mr(r,v) in Eq. (1.16) are spatially variant.
The dependency characteristics of the frequencies of time-harmonic fields
are called the materials’ dispersion. In general, all physical materials have
some dispersion because no material can respond instantaneously to external
acting fields. Additionally, in a nonlinear medium, both the permittivity and
the permeability depend on the strength of the electric field and the external
magnetic field, respectively.

1.1.3 Isotropic and anisotropic media

The responses of an isotropic medium to an external electromagnetic field are
invariant in notational directions. Thus, the magnetization, polarization, and
other parameters of an isotropic medium can be expressed simply by the
scalar coefficients of permittivity and permeability. However, numerous
composite materials, especially artificial nanostructures emerging in the field
of plasmonic optics, are anisotropic media because their internal micro-
structures are an asymmetric configuration of lattice patterns.
Optical Properties of Plasmonic Materials 7

Although they remain spatially homogeneous, these anisotropic materials

have distinct axis directions. The responses of anisotropic materials to an
external field depend on the orientation of the external field with respect to its
internal alignment. In such cases, the constitutive equations should be
generalized by introducing medium parameters in the form of the permittivity
and permeability tensors.
Another complication of anisotropic materials related to their complex
geometry is the magneto-electric coupling effect, i.e., an external electric field
induces not only electric dipole momentum but also magnetic dipole
momentum inside an anisotropic material. An applied magnetic field
generally generates both magnetic and electric polarizations. For a simply
anisotropic material, the magneto-electric coupling effect occurs at perpen-
dicular polarization directions, consisting of the orthogonal components of
electric and magnetic components.
In such bi-anisotropic media with magneto-electric coupling effects, the
general constitutive equations are formulated by introducing coupling
tensors as
pffiffiffiffiffiffiffiffiffiffi
D ¼ εE þ P0 þ ε0 m0 kem H, (1.17a)

pffiffiffiffiffiffiffiffiffiffi
B ¼ mH þ M0 þ ε0 m0 kme E, (1.17b)

where ε ¼ ε0(1 þ ke) and m ¼ m0(1 þ km) are the absolute permittivity and
permeability, respectively. P0 and M0 are called the static electric polarization
and static magnetization in the absence of external alternating fields. The
dimensionless tensors of kem and kme are introduced to describe the coupling
effect between the electric polarization induced by a magnetic field and the
magnetization induced by an electric field. The magnetic susceptibility
describes the linear magneto-electric effect, which is often observed in
artificial bi-anisotropic materials, such as split-ring resonators and wires.

1.1.4 Constitutive equations of dielectric media

The discussed constitutive equations covering polarization and magnetization
expansions address the major phenomena encountered for the vast majority of
artificial nanostructures and natural materials. In the cases of a symmetric
lattice, the number of tensor components can be reduced. For example, the
magneto-electric coupling effect within isotropic media is orthorhombic
independent, i.e., four scalar parameters characterize such bi-isotropic
materials instead of four tensors.
Dielectric media are some of the most dominant materials used for
traditional optical components due to their effective manipulation of light
waves. The physical understanding for propagating waves within dielectric
materials can also be analyzed using Maxwell’s equations and constitutive
8 Chapter 1

relations. When the instantaneous responses of homogeneous dielectric media

are not considered, the constitutive relations are expressed as:5
D ¼ ε0 E þ P ¼ ε0 ð1 þ xe ÞE, (1.18a)

B ¼ m0 H þ M ¼ m0 ð1 þ xm ÞH, (1.18b)
where P ¼ ε0xeE and M ¼ m0xmH are the polarization density and the
induced magnetization within dielectric media, respectively. The introduced
terms of xe and xm are called the electric and magnetic susceptibility,
respectively. The relative permittivity and permeability are denoted as
εr ¼ 1 þ xe and mr ¼ 1 þ xm. Both the electric displacement D and magnetic
field B can be considered to be spatially dependent in inhomogeneous media
(and likely time-harmonic dependent, too).
Likewise, the instantaneous electromagnetic responses of a homogeneous
dielectric medium at a certain time depend on the fields at that time and the
evolutionary progress over a period of past time. Thus, the constitutive
relations involve the time evolutionary variation as:
ðt
DðtÞ ¼ ε0 EðtÞ þ ε0 xe ðt  tÞEðtÞdt, (1.19a)
`
ðt
BðtÞ ¼ m0 H þ m0 xm ðt  tÞHðtÞdt: (1.19b)
`

The constitutive relations are still valid as long as the susceptibilities are
independent of the strength of the electric and magnetic fields. The
evolutionary constitutive relations in the frequency domain are written as
DðvÞ ¼ ε0 ½1 þ xe ðvÞEðvÞ, (1.20a)

BðvÞ ¼ m0 ½1 þ xm ðvÞHðvÞ: (1.20b)

In plasmonic optics, both dielectric and metallic materials are employed to
construct the unit cells in plasmonic devices, such as metamaterials and
metasurfaces. Metals are much more dispersive than dielectric materials in the
visible and near-infrared region. Nevertheless, the constitutive relations of
dielectric materials enable us to understand the origin of the frequency
dispersion in plasmonic components. In particular, when the range of interest
is extended to the mid-infrared or longer wavelengths, the properties of the
frequency dispersion for most dielectric materials begin to change. For
example, silicon dioxide, which is transparent in wavelengths less than
4.0 mm, becomes opaque when the wavelengths extend to the mid-infrared
and longer than 5.0 mm. Silicon dioxide shows evident reflectivity in the
wavelength range of 8.0–10.0 mm. The development of plasmonic optical
devices should account for the constitutive relations of both dielectric and
Optical Properties of Plasmonic Materials 9

metallic constituents. The material constituents and their patterns determine

the performance of the plasmonic nanostructures in most cases.

1.2 Electromagnetic Properties of Materials

1.2.1 Permittivity and permeability
In the theory of electromagnetic wave propagation presented here, the terms
“permittivity” and “permeability” require clarification. For an optical medium,
they define the specifications for how electromagnetic waves propagate through
a given medium.
Regarding electromagnetic fields with very low frequencies, the auxiliary
magnetic field is proportional to the magnetic field through a scalar
permeability. However, at a very high frequency of oscillatory fields, a medium
behaves like a dynamic system. The quantities will respond to each other with a
phase delay described in the form of:

HðvÞ ¼ H0 ejvt , BðvÞ ¼ B0 ejðvtdÞ , (1.21)

where d denotes the phase delay between the magnetic fields B(v) and the
corresponding auxiliary magnetic fields H(v). In electromagnetic wave
theory, permeability describes the magnetization that a medium undergoes
in response to an acting magnetic field. The permeability is defined as the ratio
of the magnetic flux density to the magnetic field. In the conditions of phase
delay existing, the permeability becomes a complex value:

B0 ejðvtdÞ B0 jd
mðvÞ ¼ ¼ e : (1.22)
H0 ejvt H0
The complex permeability is a scalar for an isotropic medium and a tensor for
an anisotropic medium. The complex permeability is translated from a polar
coordinate form to a rectangular coordinate one as
mðvÞ ¼ B0 =H0 ðcos d  j sin dÞ ¼ m1 ðvÞ  jm2 ðvÞ, (1.23)
where m1(v) and m2(v) refer to the real and imaginary parts of the complex
permeability, respectively. Likewise, a phase delay emerges between a
displacement field and an electric field when the frequency of an
electromagnetic wave increases a certain amount. A complex permittivity is
used to describe the polarization within a medium; it is expressed in the form
of real and imaginary parts as
εðvÞ ¼ D0 =E0 ðcos d  j sin dÞ ¼ ε1 ðvÞ  jε2 ðvÞ: (1.24)
In Eqs. (1.23) and (1.24), the real parts of ε1(v) and m1(v) are connected with
the stored energy within a medium, while their imaginary parts are related to
the energy dissipation. The complex permittivity and permeability are usually
10 Chapter 1

a complicated description of dispersion phenomena (which will be discussed

later in the optical properties of media). In standard international (SI) units,
the permittivity and electric susceptibility are related to each other by

ε ¼ ε0 ð1 þ xe Þ ¼ ε0 εr , (1.25a)

m ¼ m0 ð1 þ xm Þ ¼ m0 mr : (1.25b)

The actual permittivity is calculated by multiplying the relative permittivity

and the electric permittivity of free space, just as the actual permeability is
calculated.
For conventional materials working at visible and near-infrared
wavelengths, the relative permeability is often considered to be unity because
the magnetic susceptibility becomes negligible. Such conditions will simplify
the description of the optical refractive index. The particular example is that
the refraction and reflection properties of dielectric media, including both the
magnitude and phase differences, are characterized by Fresnel equations,
where neither permeability nor magnetic susceptibility is considered.
For the media with a unit value of magnetic permeability at optical
frequency, the complex permittivity ε(v) is determined experimentally through
the reflectivity measurement. The complex refractive index is linked with the
complex permittivity in the relations of

ε1 ðvÞ ¼ n21  n22 , ε2 ðvÞ ¼ 2n1 k, (1.26a)

where n1 and n2 are the real and imaginary value of the refractive index. The
complex refractive index can then be determined by
 qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 ε
n1 ¼
2
ε1 þ ε21 þ ε22 , n2 ¼ 2 : (1.26b)
2 2n1

The imaginary part n2 implies the energy absorption within a medium. Taking
into account the permeability, the refractive index is expressed as

ðn1 þ jn2 Þ2 ¼ ðε1 m1  ε2 m2 Þ þ jðε1 m2 þ ε2 m1 Þ: (1.27)

In the emerging fields of metamaterials and metasurfaces, nanostructure

materials with a negative or near-zero refractive index are now being
achieved. To achieve a negative real part of the refractive index, the imaginary
part in Eq. (1.26) must be negative. It indicates that the negative refractive
index can be achieved by a sufficiently large imaginary part of permittivity
and/or permeability. It is not necessary that the real parts of both ε1 and m1
are negative. In the visible light region, most noble metals have a large
imaginary part of permittivity |ε1| .. |ε2|. This property enables them to be
frequently used as materials to construct plasmonic devices.
Optical Properties of Plasmonic Materials 11

1.2.2 Loss tangent

The curl equation of a magnetic field in Maxwell’s equations is expressed as
∇  H ¼ ðvε2 þ sÞE þ jvε1 E: (1.28)
The real part ε1 relates to the lossless component. The imaginary part ε2
represents the energy loss component. The quantity s(v) is defined by a
constitutive relation between the internal current density J and the electric
field E as4
JðvÞ ¼ sðvÞ  EðvÞ: (1.29)
The quantity s(v) refers to the conductivity, which is an intrinsic
characteristic of a medium determined by the relation

N  q2e s0
sðvÞ ¼ ¼ , (1.30)
me ð1=t  jvÞ 1  jvt
where me is the effective mass of an electron, N refers to the electron number
in unit volume, and g is the damping ratio of the oscillating electrons. The
term s0 is defined as

N  q2e g
s0 ¼ : (1.31)
me
In the presence of an external electric field, the loss tangent is defined by the
ratio of the loss component to the lossless component in Eq. (1.28):
vε2 þ s
tan h ¼ : (1.32)
vε1
For dielectric media with a small amount of energy loss, Eq. (1.32)
approaches a simple form by the approximation tan h  h. Likewise, the
electric and magnetic loss tangent of a transparent dielectric medium are
quantified similarly in the subsequent definitions of
m2 ε2
tan hm ¼ , tan he ¼ : (1.33)
m1 ε1
The loss tangent quantifies the inherent dissipation of electromagnetic energy
due to the imaginary parts of its dielectric properties.

1.2.3 Penetration depth and skin depth

When an electromagnetic wave impinges on the surface of a medium, part of
the wave reflects from the surface, and a small fraction of the wave will
penetrate into the medium. The penetrated wave interacts with the atoms and
electrons of the medium, which will contribute significantly to the SPPs. The
12 Chapter 1

Beer–Lambert law indicates that the intensity of the electromagnetic wave

penetrating into a medium falls off exponentially from the air–medium
interface.5,6 The penetrating intensity is described by the exponential function of

I ðzÞ ¼ I 0 ezaðvÞ , (1.34)

where z denotes the penetrating distance from the air–medium interface. The
attenuation factor a(v) depends on the frequency of the impinging
electromagnetic wave. I(z) and I0 are the penetrating intensity at depth z
and the initial intensity at the interface, respectively.
A penetration depth is defined as the penetrating distance at which the wave
intensity decays to the value 1/e (0.37) of its initial intensity. Conventionally,
the penetration depth is defined in terms of the attenuation factor

dp ðvÞ ¼ a1 ðvÞ: (1.35)

In the condition of a static electric field, the current density within a conductor
wire distributes uniformly throughout the cross-section of the wire. On the other
hand, when the frequencies of an electromagnetic wave increase, the current
flow tends to approach the surface of a conductor medium. The penetration
depth decreases sharply with the frequency increasing. A similar concept of skin
depth is used to describe the current flow in a conductor medium.
The energy of a wave propagating through a particular medium is
proportional to the square of its intensity. From the energy point of view,
when the amplitude of the energy attenuates to 1/e of its initial energy, the
depth below the air–medium interface is defined to be the skin depth of a
medium, which relates with the penetration depth as

1 1
ds ðvÞ ¼ ¼ dp ðvÞ: (1.36)
2aðvÞ 2

The skin depth is one characteristic of an electromagnetic medium whose

value also depends on the frequencies of electromagnetic waves.
For a solid wire, as shown in Fig. 1.1(a), the current concentrates on the
outer surface. In the case of a micro-strip layout, as shown in Fig. 1.1(b), the
current concentrates nearest to the substrate dielectric material. As suggested
in Eqs. (1.34)–(1.36), there is an exponential decay of the electromagnetic
amplitude or energy. A more gradual penetration of the electromagnetic
energy into the metal occurs, although the high-amplitude regions are
depicted as having a sharp interface between the “bulk” metal and the near-
surface regions.
In the case of a poor conductor, such as a dielectric medium, or in the
presence of an electromagnetic wave with a high frequency, a general
calculation with a more exact solution is expressed as:
Optical Properties of Plasmonic Materials 13

Figure 1.1 Schematic diagram of current distribution within conductor media. (a) Current
flows through a cylindrical conductor wire and (b) in metal strips of a composite structure.
The dark color indicates regions with a higher electromagnetic amplitude. The regions of
metal and dielectric materials are labeled by their permittivity.

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dS ðvÞ ¼ 2rc =vm 1 þ ðrc εvÞ2 þ rc εv, (1.37)

where rc is the resistivity of conductors (ohm  meter), ε and m are the

absolute permittivity and permeability, and v is the angular frequency of an
impinging wave.
When the frequency is much lower than the value 1/rcε, the calculated
value of the square root in Eq. (1.37) approaches unity. In such a case, the
skin depth of a medium is estimated by
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dS ðvÞ ¼ 2rc =vm: (1.38)

Equation (1.38) applies to the materials with a larger skin depth at a much
lower frequency. For instance, it would hold true for metal copper at
frequencies below 1018 Hz. Nevertheless, for materials with a rather low
conductivity corresponding to a shallow skin depth or materials that work at
frequencies much higher than the value of 1/rcε, the skin depth approaches an
asymptotic value rather than monotonously decreasing:
pffiffiffiffiffiffiffiffi
dS ðvÞ  2rc ε=m: (1.39)

Equation (1.39) applies to materials with low conductivity working at high

frequencies. For instance, the skin depth of undoped silicon is 40 m
irradiated by waves with a frequency of 100 kHz. When the frequency of the
wave increases to the microwave region (1 GHz to 30 GHz), with a
corresponding wavelength of 30 cm to 10 mm, the skin depth of the same
undoped silicon closes to the asymptotic value of 11 m.
For good conductors working at a sufficiently high frequency, the skin
depth becomes tiny, which implies that the electric charges approach the air–
medium interface. This characteristic deserves attention with respect to
nanostructures and plasmonic devices. In such conditions, the dimensional
14 Chapter 1

Figure 1.2 Spectral dependence of skin depth for frequently used metals when a plane
wave impinges on the air–metal interface at normal incidence. The parameter values for the
estimation of skin depth are adopted from Rodrigo et al.7

sizes approach the scale of skin depth at the visible or near-infrared

wavelength.
Figure 1.2 displays the skin depth spectra of a few frequently used noble
metals. For those metals with good conductivity working at the visible and
near-infrared wavelength, a majority of the electric current flows within a thin
region, which is less than tens of nanometers beneath the air–medium
interface. For example, the skin depth of gold at a wavelength of 500 nm
reaches a peak value of 40 nm and then decays to a skin depth of 25 nm at
a near-infrared wavelength of 900 nm.7 Table 1.1 shows the skin depths of a
few frequently used noble metals at certain frequencies.
In the condition of normal incidence of an electromagnetic wave, the
attenuation factor is also proportional to the imaginary part of the refractive
index. The following relationship holds true for the attenuation factor defined
by Eq. (1.34):

Table 1.1 Bulk resistivity and skin depths of frequently used noble metals.
Skin Depth (mm)
Bulk Resistivity
Material (Ω  10 8 m) 1 MHz 10 MHz 100 MHz 1 GHz 10 GHz 100 GHz
Aluminum 2.65 81.9 25.9 8.19 2.59 0.819 0.259
Copper 1.69 65.4 20.7 6.54 2.07 0.654 0.207
Gold 2.2 74.7 23.6 7.47 2.36 0.747 0.236
Graphite 783.7 1409 446 141 44.6 14.1 4.46
Lead 20.6 228 72.2 22.8 7.22 2.28 0.722
Nickel 6.9 9.3 3.0 .093 0.30 0.093 0.030
Silver 1.63 64.3 20.3 6.43 2.03 0.643 0.203
Titanium 54 370 117 37.0 11.7 3.70 1.17
Optical Properties of Plasmonic Materials 15

aðvÞ ¼ 2n2 ðvÞv=c, (1.40)

where n2(v) denotes the imaginary part of the complex refractive index. So
both the penetration depth and the skin depth can be estimated directly from
the imaginary part of the refractive index.

1.3 Optical Properties of Metals

The obvious difference between the optical responses of metals and dielectrics
enables them to be fundamental elements for designing plasmonic devices. In
plasmonic optics, metals are incorporated into nanostructures for the
emerging characteristics of the SPPs and other light–matter interactions. This
section emphasizes the optical performance of metals following the theory
model for free electrons and bound electrons, pursing the approach for
realizing the optimal plasmonic materials at the nanoscale.

1.3.1 Free electrons and interband transitions

To investigate the optical properties of solid materials, we start from
the conduction electrons within metals and semiconductors. There are
conduction electrons and interband transition electrons within media.8
Conduction electrons oscillate freely within bulk metals, whereas interband
transition electrons can only be induced by the incident photons with energy
greater than the bandgap energy. Both of them determine the complex
dielectric functions, such as permittivity and permeability, that describe the
optical properties of metals.
Conduction electrons are associated principally with free carriers within
metal media. The mechanism of this intraband electronic conduction is
explained by the Drude theory. For the electrons bounded into occupied
states, the electrons can jump up from the energy band below the Fermi level
to the conduction bands only when excited by photons with enough energy.
This process of electron transition is called an interband transition, as shown
in Fig. 1.3, where no energy states are allowed between the filled valence band
and an empty conduction band. Higher-energy photons can promote the

E Empty

hw

Filled
k

Figure 1.3 Schematic diagram of an interband transition of bound electrons.

16 Chapter 1

(a) Conduction (b) (c)

band Conduction
band Conduction
band
Electron energy

Band Band
gap gap

Valence
Valence band
Valence band
band
insulator semiconductor metal
Momentum k
Figure 1.4 Typical band structures of insulator, semiconductor, and metal materials.

bound electrons from lower energy levels into the empty conduction band.
Such interband transitions often occur in the light–matter interaction at the
visible or infrared wavelength. The process of interband transitions produces a
pair carrying an electron and a hole. In practice, only a portion of interband
transitions contribute to the conduction mechanism, which corresponds to a
few specific energy bands.
Solid materials have their own energy band structures, as shown in Fig. 1.4.
The various characteristics of energy band structures determine the multitudi-
nous electrical properties of solid materials. For example, the bandgap of the
undoped silicon is 1.1 eV, whereas diamond has a 5-eV bandgap.8
The electron transitions in a medium between the energy levels determine its
dielectric function and then the optical properties. For an insulator, a large gap
exists between the valence and conduction band, as shown in Fig. 1.4(a). No
electron jumps into the conduction band without enough energy. The electrons
in insulators are rare compared with semiconductors. As shown in Fig. 1.4(b),
semiconductor materials have basically the similar band structure as insulators
have; however, the bandgap Eg of semiconductors is much smaller than that of
insulators. The number of electrons within semiconductors would be increased
greatly after the photons are excited to achieve electron conduction. A deep
doping will greatly reduce their semiconductor band gaps, which enables more
available electrons to contribute to the interband transition.
In the case of metal materials, however, the energy bands overlap each
other or, in certain conditions, are partially filled, as shown in Fig. 1.4(c). The
overlap of valence bands and free conduction bands enables electrons to move
freely across the bands. Interband transitions within metals can easily occur
under the excitation by photons with low frequencies. The free carriers
contribute largely to the conduction. Although the densities of free carriers are
low, the high mobility of free carriers contributes to the conductivity of metals.
For semiconductors, the energy gaps are often temperature dependent.
For example, the energy gap of germanium depends on the temperature as
Optical Properties of Plasmonic Materials 17

expressed by an empirical formula Eg ¼ 0.742 – 4.8  10–4 T2/(T þ 235)(eV),12

where T is the temperature in kelvin. Also, the energy gap bands of
semiconductors are determined by the doping density of either donors or
acceptors. For example, deep doping will increase the band gap of germanium
by the following approximation formula:12
pffiffiffiffiffiffiffi pffiffiffiffiffiffiffi
• For pn-type Ge,
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi DE g ¼ 4.31  109 N d þ 8.67  106 3 N d þ 8.14 
105 4 100  N d ðeVÞ, pffiffiffiffiffiffi pffiffiffiffiffiffi
• For pp-type
ffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi Ge: DE g ¼ 5.77  109 N a þ 8.21  106 3 N a þ 9.18 
105 4 100N a ðeVÞ,
where Nd and Na are the doping density of the donor and the acceptor,
respectively. Increases to the doping concentration enlarge the energy
bandgap, slowly at first and then sharply.

1.3.2 Harmonic oscillator model

The Drude and Lorentz models were developed in the electronic kinetic
theory of microscopic electrons to explain the optical properties of materials.
The models were further extended into the Drude–Lorentz model, which
describes the dielectric properties of solid materials.
As shown in Fig. 1.5, a charge within a medium is treated as a harmonic
oscillator, which is bounded with a nucleus. Under the excitation of an incident
electromagnetic wave, the oscillator will oscillate in the oppositional phase
relative to the electric field. Here, only the conditions of a nonpolar molecule
are discussed, in which no dipole moment exists in the electron–nucleus system.
From the dynamic point of view, the charge oscillation will lead to the
charge redistribution, which will create an additional induced electric field.
The induced field will restore the charge to its equilibrium position.
Newton’s second law states that the production of mass multiplying the
acceleration equals the sum of the forces applied on the system. The governing
equation for the system in Fig. 1.5 is
­2 rðtÞ X
me ¼ Fi ¼ FE ðtÞ þ FD ðtÞ þ FS ðtÞ, (1.41)
­t2 i

where me denotes the effective mass of a charge, r(t) is the instantaneous

distance deviation from its equilibrium position. FE(t) ¼ qeE0e–jvt is the local
→
EL

C, γ e¯ m
·

Nucleus r

P = -er
Figure 1.5 Schematic of electron motion for a harmonic oscillator model in an external
electric field.
18 Chapter 1

force attributed to an external alternating electric field, where qe is the electric

quantity of an charge, and E0 is the electric field amplitude. FS(t) ¼ –q  r(t) is
the restoration force proportional to the distance deviation from the
equilibrium position q ¼ mev02, where v0 is the natural frequency of the
bound oscillator. The restoration forces are zero for free electrons within
conductor materials because they are not bound to a particular nucleus.
FD ¼ –g[­r(t)/­t] refers to the damping force due to the collision energy loss,
where g is the damping coefficient in hertz.
Note that the restoration and damping forces are negative because they
are opposite the direction of the motion. Thus, the differential form in
Eq. (1.41) becomes

­2 r ­r
me þ me g þ me v20 r ¼ qe Eo ejvt : (1.42)
­t 2 ­t
In general, the condition g . 0 leads to the damping attenuation.
The instaneous distance solution of Eq. (1.42) for a monochromatic
electric field is solved to be
qe
rðvÞ ¼ EðvÞ: (1.43)
me ðv20  v2  jgvÞ
The polarization of a dipole moment causes the electric-charge timing
distance to deviate from its balanced position. The gross polarization of the
whole electrons in unit volume is

N e q2e
P ¼ N e qe rðvÞ ¼ EðvÞ, (1.44)
me ðv20  v2  jgvÞ
where Ne is the electron density per unit volume. The induced dielectric
polarization density is proportional to an electric field by the constant electric
susceptibility as
P ¼ ε0 xe EðvÞ ¼ ε0 ðεr  1ÞEðvÞ: (1.45)
The susceptibility is xe ¼ εr – 1. The permittivity function is thus obtained:11

v2p
εðvÞ ¼ 1 þ , (1.46)
v20  v2  jgv
where vp is the plasma frequency of the bulk metal, which is calculated by5

N e q2e
v2P ¼ : (1.47)
me ε0
The dielectric function [Eq. (1.46)] presents explicitly the real and imaginary
parts in the form of
Optical Properties of Plasmonic Materials 19

v2p ðv20  v2 Þ
ε1 ðvÞ ¼ 1 þ , (1.48a)
ðv20  v2 Þ2 þ g2 v2

gv2p v
ε2 ðvÞ ¼ : (1.48b)
ðv20  v2 Þ2 þ g2 v2

1.3.3 Drude model and Lorentz model

The interactions between metals and electromagnetic waves are firstly
determined by the collective movement of free electrons. In this case, the
electrons are not bound to any particular nucleus, which are considered to
move about freely around the metal lattice in the absence of a restoration
force. The motion equation of a free electron in an alternating electric field is
described by

d 2r dr
me þ me g ¼ qe Eo ejvt , (1.49)
dt2 dt
where qe is the electric charge of a free electron. the damping effect g is
proportional to the Fermi velocity g ¼ n/l ¼ 1/t, where n denotes the Fermi
velocity, and l is the mean free path of an electron between successive collision
events. The relaxation time t is the averaged interval time between subsequent
collisions of an electron. In general, the relaxation time is about 10–14 s. The
solution of instaneous distances in Eq. (1.49) for a monochromatic electric
field is solved to be
qe
rðvÞ ¼ EðvÞ: (1.50)
me ðv þ jvgÞ
2

The polarization density is the total dipole moment per unit volume. The
gross polarization of all of the electrons in the unit volume is

N  q2e
P ¼ N  qe  rðvÞ ¼ EðvÞ, (1.51)
me ðv20 þ jgvÞ

where N is the electron density per unit volume. A comparison of Eqs. (1.45)
and (1.51) leads to the frequency-dependent permittivity of metals as

v2P
εðvÞ ¼ 1  , (1.52)
v2 þ jvg
where vp is the plasma frequency of the bulk media, which has a similar
physical mechanism as the definition in Eq. (1.47). At the plasma frequency,
the electromagnetic response of a material changes from the metallic
behaviors to those of dielectric materials. At the frequency below the plasma
20 Chapter 1

frequency, the optical properties of a medium exhibit a metal-like behavior.

However, at a frequency greater than the plasma frequency, their optical
properties look more like those of dielectric media. Equation (1.52) describes
the contribution of free electrons to the permittivity of metals. The
permittivity in Eq. (1.52) is rewritten into the real and imaginary parts
 
v2 gv2P
εðvÞ ¼ ε1 þ jε2 ¼ 1  2 P 2 þ j : (1.53)
v þg vðv2 þ g2 Þ
In the conditions of v .. g, Eq. (1.53) is simplified to be

v2P v2P
εðvÞ  1  þ j g: (1.54)
v2 v3
The permittivity of metal gold is calculated by Eq. (1.53), ranging from
visible to infrared wavelengths. As shown in Fig. 1.6, the real part of the
permittivity for metal gold remains negative throughout the wavelengths in
which the frequencies are less than the plasma frequency. The negative
permittivity results in an imaginary part of the refractive index. For metals,
the imaginary part of the permittivity implies an energy dissipation that is
relevant with the thermal motion of electrons.
Figure 1.6 indicates that, in the infrared wavelengths longer than
650 nm, the Drude model gives an accurate permittivity of gold, and the
experimental data clearly agree well with the Drude theory. However, there is
an obvious deviation of the imaginary part in the visible wavelengths shorter
than 650 nm. The measured imaginary part increases much more sharply than
that predicated by the Drude theory. This difference is attributed to the fact
that the interband transitions of the bound electrons excited by the photons

Figure 1.6 Permittivity of metal gold estimated by the Drude model (dotted line) and the
measurement (solid line). The measured data come from Ref. [10]. The parameter values in
the Drude model are adopted from Table 1.2.
Optical Properties of Plasmonic Materials 21

with higher energy have not been taken into account in the Drude model,
where, however, the interband transitions become significant.
The contribution from the interband transitions of bound electrons to the
dielectric permittivity looks like the corresponding resonance in dielectric
media. These transitions are obtained from the governing equations without
consideration of the free electrons as

v21
εðvÞ ¼ 1 þ , (1.55)
v20  v2  jgv
where v0 is the oscillation frequency of a bound electron under an electric
field, and v1 and g are related to the corresponding frequency and damping of
a bound electron, respectively.
Figure 1.7 shows the dielectric permittivity of metal gold considering
only the interband contribution of bound electrons. The real part shows a
dispersion behavior while the imaginary part displays a clear resonant behavior.
The vertical dotted line represents the resonant frequency v0 that is consistent
with the zero value of the second part on the right side of Eq. (1.52). The
resonant frequency v0 also agrees well with the peak value of the imaginary
part, so it is sometimes called the absorption frequency. For the real parts, there
is a peak value in the lower frequency and a valley value in the higher frequency
that is greater than the resonant frequency. The frequency region between the
two extreme values is called the abnormal dispersion region, where the
permittivity values decrease with the increasing frequency.
For a Lorentz oscillator, the real part of permittivity is a positive value in
the range less than the resonance frequency, while there is a negative valley at
the slightly higher frequency than the resonance frequency. Another feature in
Fig. 1.7 is that the dielectric permittivity has a nonzero asymptotic value when

Figure 1.7 Contribution of the bound electrons in metal gold to the permittivity. The values
of the parameters used are v1 ¼ 4.55  1015 s1, g ¼ 9.10  1016 s1. The central
resonance wavelength is 450 nm (v0 ¼ 4.19  1017 s1).
22 Chapter 1

the wavelength increases much longer than the resonance wavelength. After
all, if the interband electron transitions within media have been taken into
account, such an asymptotic constant should approach one summary value.11
If the higher order of interband electrons transition is taken into account,
the dielectric permittivity is expressed as the model superposition of the
Lorentz model as

X
K
f i v2i
εðvÞ ¼ 1 þ , (1.56)
i v2i  v2  jgi v

where i refers to the resonant modes, vi corresponds to the resonance

frequencies, fi indicates the weighting coefficients, and gi is the damping effect.
Each electron is assumed to contribute to the dipole polarization. The
atomic nucleus also makes a slight contribution, which is ignored because
the mass of the nucleus is much greater than that of the electron. The
aforementioned formulae of dielectric permittivity in the Drude theory are
derived under the approximation of infinite nucleus mass. Another assumption
is that only one electron works in one electronic-dynamics system. In such
conditions, the electric properties of commonly used noble metals and their
corresponding bulk plasma frequencies (as well as the damping constant) are
listed in Table 1.2.
Figure 1.8 compares the characteristics of metals using the Drude model
and the Lorentz model. At the plasma frequency in the Drude model, the
dielectric function that passes through zero corresponds to the collective

Table 1.2 Electric properties of noble metals and bulk plasma frequencies.
Plasma Frequency Plasma Frequency Damping Constant
Metal (eV) (1015 s–1) (1015 s–1)
Au 9.1 13.8 0.011
Ag 9.2 14.0 0.032
Al 15.1 22.9 0.92
Cu 8.8 13.4 0.14

(a) (b)
Dielectric function

ε2 ε2

0 ε1

ε1 0

ωp ωp ωr

Figure 1.8 Components of the complex dielectric functions for (a) metals described using
the Drude model and (b) dielectrics using the Lorentz equations.
Optical Properties of Plasmonic Materials 23

electronic oscillation within metals. Figure 1.8(b), described by the Lorentz

equations, shows a resonant frequency that corresponds to the bound
electrons, which is consistent with the absorption peak.

1.3.4 Drude–Lorentz model

Both bound electrons and free ones contribute to the optical properties of a
general metallic medium. Therefore, the resembling complex dielectric
permittivity contains both the Drude component for the intraband effect
and the Lorentz term for the interband transition in the form of the Drude–
Lorentz model

v2p X
K
f i v2i
εðvÞ ¼ 1  þ , (1.57)
v2 þ jvgd i¼1 v2i  v2  jvgi

where K is the total number of higher-energy oscillators with the corresponding

resonant frequency vj, weighting coefficient fj, and lifetime 1/gj. In the second
term in Eq. (1.57), vp refers to the plasma frequency relevant to the intraband
transitions with a damping constant gd.
As shown in Figs. 1.6 and 1.7, the interband transition has a nonzero
asymptotic value of the dielectric permittivity at a wavelength much further
away from the resonance wavelength. It is noteworthy that, within noble
metals, there are usually multiple interband transitions due to the various band
structures of bound electrons. Therefore, after taking all pertinent transitions
into account, we use a constant offset value to denote the constant offset when
working at much longer wavelengths than the resonance wavelength. Thus, the
modified Drude–Lorentz model considering the fundamental oscillator is

v2p v2p g
εðvÞ ¼ ε`  þ j : (1.58)
v2 þ g2 vðv2 þ g2 Þ

For example, the interband offset for gold is about ε` ¼ 1.2 from Eq. (1.58) by
setting v ! 0.
Optimal values of typical metals for the Drude–Lorentz model are listed
in Table 1.3. These include noble metal materials (Au, Ag, and Cu),
aluminum, and transition metals (Cr, Ni, Pd, Pt, and Ti). These metals have
been frequently used in optoelectronic devices. The measurements of these
optical data are strongly dependent on the experimental conditions (in terms
of geometry morphology and the fabrication process), so the relevant optical
data must be carefully chosen.12
A correction may be made to the model data shown in Fig. 1.5 obtained
by the Drude–Lorentz model. Taking into account the contribution from both
free electrons and a single interband transition, the dielectric function of gold
progresses toward the measured results. Only one interband transition has
24 Chapter 1

Table 1.3 Parameter values for the Drude–Lorentz model. Adapted from Ref. 12 with
permission from the American Physical Society, © 1998.
Parameters Ag Au Cu Al Be Cr Ni Pt Ti W
f0 0.845 0.760 0.575 0.523 0.084 0.168 0.096 0.333 0.148 0.206
gd 0.048 0.053 0.030 0.047 0.035 0.047 0.048 0.080 0.082 0.064
f1 0.065 0.024 0.061 0.227 0.031 0.151 0.100 0.191 0.899 0.054
g1 (eV) 3.886 0.241 0.378 0.333 1.664 3.175 4.511 0.517 2.276 0.530
v1 (eV) 0.816 0.415 0.291 0.162 0.100 0.121 0.174 0.780 0.777 1.004
f2 0.124 0.010 0.104 0.050 0.140 0.150 0.135 0.659 0.393 0.166
g2 0.452 0.345 1.056 0.312 3.395 1.305 1.334 1.838 2.518 1.281
v2 4.481 0.830 2.957 1.544 1.032 0.543 0.582 1.314 1.545 1.917
f3 0.011 0.071 0.723 0.166 0.530 1.149 0.106 0.547 0.187 0.706
g3 0.065 0.870 3.213 1.351 4.454 2.676 2.178 3.668 1.663 3.332
v3 8.185 2.969 5.300 1.808 3.183 1.970 1.597 3.141 2.509 2.580
f4 0.840 0.601 0.638 0.030 0.130 0.825 0.729 3.576 0.001 2.590
g4 0.916 2.494 4.305 3.382 1.802 1.335 6.292 8.517 1.762 5.836
v4 9.083 4.304 11.18 3.473 4.604 8.775 6.089 9.249 19.43 7.498
f5 5.646 4.384 — — — — — — — —
Г5 2.419 2.214 — — — — — — — —
v5 20.29 13.32 — — — — — — — —

been considered in Fig. 1.9, so the mode curves still fail to reproduce the
experimental results at wavelengths below 500 nm. For a practical metal,
there are usually multiple interband transitions in the short-wavelength
spectrum because the higher energy of photons promotes interband transitions
of bound electrons.

Figure 1.9 Permittivity of gold estimated by the Drude–Lorentz model (dotted line)
considering the contribution from both the free electron and a single interband transition. The
experimental values (solid line) from Ref. [10] are compared. The model parameters for Au
are taken from Table 1.3. Note the different scales for the real part in the lower half and the
imaginary part in the upper half.
Optical Properties of Plasmonic Materials 25

Figure 1.10 (a) Real and (b) imaginary parts of the permittivity for metal aluminum at
wavelengths ranging from 200 nm to 1200 nm. The first, second, and third oscillators of
interband transitions are integrated into the effect. The experimental values are taken from
Ref. [10], and the parameter values for the model are from Table 1.3.

When more interband transitions are integrated into Eq. (1.57), the
calculated date would progress closely to the experimental data. For
example, the dielectric functions of aluminum are shown in Fig. 1.10. Both
the real part and the imaginary part of the permittivity approach the
experimental results when the higher-energy interband transitions are
integrated into the effect. Some rapid variations occur around 800 nm,
which corresponds to the lowest-lying interband transition of the bound
electrons in the outer shells.
The dielectric permittivity of metal silver and the complex refractive index
are plotted in Fig. 1.11. The contribution of both free and bound electrons is
considered. The measured data are consistent with the results estimated by the
Drude–Lorentz model. At optical frequencies, the imaginary part of the
dielectric permittivity is positive, whereas the real part has negative values at
wavelengths longer than the resonant wavelength. The dielectric function of
Im(ε) ,, Re(|ε|) holds true. However, the imaginary part of the refractive
index is incrementally greater than the corresponding real part, which is
slightly greater than zero.
26 Chapter 1

Figure 1.11 (a) Permittivity and (b) corresponding refractive index of metal silver. The
Drude–Lorentz model in (a) considers the contribution of both free and bound electrons. The
experimental data come from Ref. [10].

The reflection coefficient in Fresnel equations is presented by (n1  n2)/

(n1 þ n2), where n1 is the real part and n2 is the imaginary part of the refractive
index. In the conditions of n2 .. n1 at the wavelengths longer than the
corresponding plasma frequency of the interband transition, the reflection
coefficient approaches an absolute value of unity, which indicates that most of
the parts of light are reflected at the interface between a dielectric medium and
silver metal.
The dielectric functions for the metals of Au, Cu, Ni, and Cr are
compared in Fig. 1.12. The imaginary parts of the dielectric permittivity of
these metals remain positive at optical wavelengths ranging from 200 nm to
900 nm, whereas the real parts of dielectric functions are distinctly negative.
An increasing absolute value of the real part indicates that most of the light
reflects at the dielectric–metallic interface. The imaginary part of metals Ni
and Cr increase prominently after a wavelength of 500 nm while their real
parts have a faint value. The reducing reflection coefficient means that the
light will be absorbed in those wavelengths.
Following a similar methodology as the deducing of the Lorentz model,
the effective permeability can be obtained as
Optical Properties of Plasmonic Materials 27

Figure 1.12 (a) Real part and (b) imaginary part of the dielectric functions for metals Au,
Cu, Ni, and Cr at wavelengths ranging from 200 nm to 900 nm. Fitting lines are shown for
clarity. The experimental data come from Ref. [10].

v2pm
mr ðvÞ ¼ 1 þ , (1.59)
v2om  v2  igm v
where vpm, vom, and gm are the plasma frequency, resonant frequency, and the
damping coefficient in the magnetization of a magnetic dipole. For multiple
orders of magnetic resonance, Eq. (1.59) is extended into a multiple form as

X
M v2pmk
mr ðvÞ ¼ 1 þ : (1.60)
k¼1
v2omk  v2  igmk v

The resonant components vpmk correspond with k orders of resonances

of magnetic dipoles. The relative permeability of metals and dielectric media
is often assumed to be unity. However, the effective permeability for
composite structures made of dielectric or metal materials can be tuned to
values not accessible in natural materials.13 A large imaginary component of
permeability helps one obtain composite structures with a negative refractive
index.14 The extension of relative permeability allows for arbitrary spectral
dispersion.
28 Chapter 1

1.4 Optical Properties of Dielectric Materials

Both dielectric and metallic materials constitute materials used to design
nanostructures in plasmonic optics. In conventional geometrical optics, the
vast majority of functional parts are made from dielectric materials, such as
crystalline and glassy materials. The characteristics of dielectric materials
enable them to be useful for various band energies.

1.4.1 Dielectric function of dielectric media

Figure 1.13 shows the dielectric functions of silicon and germanium. Likewise,
two resonances for these semiconductors appear indistinctly at the visible
wavelength region, with positive imaginary values, and with two near
resonances crossing zero points at a wavelength of 300 nm. In the conditions
of wavelengths approaching 900 nm, the real parts of the dielectric functions
approach asymptotic values, and the imaginary values approach zero limits.9
Photonic bandgap structures are made of semiconductor materials with
periodic structures to modulate the refractive index, where the dimensional
sizes are comparable to the wavelength of light. Such electromagnetic
bandgap structures that work in the microwave or visible range are called

Figure 1.13 Dielectric functions for semiconductor materials of silicon and germanium.
The experimental data comes from Ref. [10].
Optical Properties of Plasmonic Materials 29

Table 1.4 Bandgap and optical properties of frequently used dielectric materials.
Material InAs Ge Si InP GaAs TiO2 ZnO SiO2 Al2O3
Bandgap (eV) 0.35 0.66 1.12 1.34 1.424 3.0 3.44 8.5 8.9
l (mm) 3.50 1.87 1.10 0.92 0.87 0.41 0.36 0.15 0.14
Carrier density (cm–3) 1015 2  1013 1010 1.3  107 2.1  106 — — — —

semiconductor crystals. The photonic bandgap structures exhibit a periodic

potential by designing allowed and forbidden energy bands, which influence
the motion equations of electrons.
In order to obtain periodic potentials, these bandgap semiconductor
crystals are constructed by materials with a sharp contrast of dielectric function
in the constituted cell. For instance, a nanostructure combines a dielectric
material with a metal. The prominent superiority of such artificial structures has
wide applications, for example, to enable light propagation in specified
directions, to localize light in particular zones, or to completely prohibit light
propagation within a certain bandgap. Table 1.4 lists the bandgap and optical
properties of a few frequently used dielectric materials.10,16
There are no absolute conductors or dielectrics: the behavior of metals is
primarily conductive due to the domination of the intraband transitions at a
low frequency. However, metals behave like dielectrics at frequencies greater
than the plasma frequency of the ultraviolet wavelength.

1.4.2 Kramers–Kronig relation

The dielectric functions and complex refractive index are the inherent
definitions of optical properties of media. Unfortunately, they are not measured
directly in experiments. What we can obtain physically are the transmissivity
and reflectivity of a medium. The Kramers–Kronig formula relates the complex
optical properties with the physical observables, which are useful tools to obtain
the dielectric functions.17
When the real part ε1 (or the imaginary part ε2) of a permittivity function
has been obtained over a frequency region (vl, vh), the complex counterpart
will be determined from the Kramers–Kronig formula18
ðv
2 h ṽ  ε2 ðṽÞ
ε1 ðvÞ ¼ 1 þ Pc d ṽ, (1.61a)
p vl ṽ2  v2

ðv
2v h ε1 ðṽÞ
ε2 ðvÞ ¼  P d ṽ, (1.61b)
p c vl ṽ2  v2

where Pc is the Cauchy principal value. For the refractive index, the real part
and its imaginary part are related with the Kramers–Kronig relations18
30 Chapter 1

ðv
2 h ṽ  n2 ðṽÞ  vn2 ðvÞ
n1 ðvÞ ¼ 1 þ Pc d ṽ, (1.62a)
p vl ṽ2  v2

ðv
2v h n1 ðṽÞ  n1 ðvÞ
n2 ðvÞ ¼  P d ṽ: (1.62b)
p c vl ṽ2  v2

The relation between the reflectivity R(v) and its phase difference Dw(v) are
also defined by the Kramers–Kronig form
ðf
2 h ṽ  DðṽÞ  v  DwðvÞ
ln RðvÞ ¼ Pc d ṽ, (1.63a)
p fl ṽ2  v2

ðv
2v h ln RðṽÞ  ln RðvÞ
DwðvÞ ¼  P d ṽ: (1.63b)
p c vl ṽ2  v2

The Kramers–Kronig theory is widely useful to quantify the optical

characteristics of natural and artificial materials, which enable us to measure
optical properties.

1.4.3 Obtaining optical functions from physical observables

The complex dielectric functions (permittivity and permeability) are linked to
each other with the refractive index, as expressed in Eqs. (1.26) and (1.27), given
the permeability is zero. At a semiconductor-metal interface, the reflectivity and
the associated phase difference in normal incidences is calculated by12

ðn0  n1 Þ2 þ n22
R¼ , (1.64a)
ðn0 þ n1 Þ2 þ n22

2n0 n2
Dw ¼ arctan , (1.64b)
ðn20  n21  n22 Þ

where n0 refers to the refractive index of the incident medium, and n1 þ jn2 is
the complex refractive index of the substrate material. In general, at the
boundary of free space and a nonmagnetic substrate material, the reflection
coefficient is presented through Snell’s law:3

n1  1 þ jn2
R  ejvu ¼ , (1.65)
n1 þ 1 þ jn2

where u refers to a series of incident angles that depend on the wave frequency.
The optical functions are then related to the physical observables by
Optical Properties of Plasmonic Materials 31

1  R2
n1 ¼ , (1.66a)
1 þ R2  2R  cos u

2R  sin u
n2 ¼ : (1.66b)
1 þ R2  2R  cos u
In Eqs. (1.64)–(1.66), optical parameters are the function of optical frequency.
The optical refractive index can be determined in Eq. (1.66) by the reflectivity
over a frequency region.
The reflection coefficient and the corresponding phase difference can be
calculated from the frequency-dependent reflectivity. In order to calculate
the complex refractive index n1 þ jn2, two independent measurements could
be made to relate the complex dielectric function to physical observables.
One convenient method measures the reflectivity in different conditions, for
example, different incident angles of light with the same wavelength or
various wavelengths with the same incident angles. Both the real and
imaginary parts of a complex dielectric function can be related with the
Kramers–Kronig relations.

1.5 Effective Medium Approach for Composite Nanostructures

1.5.1 Effective medium theory
In plasmonic optics, composite nanostructures are often made from two or
multiple materials in either an arbitrary fashion or ordered patterns. Among
them, the host materials are often electromagnetically continuous media,
while a small number of inclusions or particles are incorporated in host
materials. For the composite nanostructures made from metallic and dielectric
components, the overall optical properties may be significantly different from
those of constituent host materials, and also differ greatly from its inclusions.
The electromagnetic properties are determined by the electrical dimensional
sizes of the inclusions and the interparticle distance. It is hard to develop a
universal method to analyze the optical properties of such arbitrary
nanostructure materials.11
In general, composite nanostructures are designed based on one of two
strategies: (1) random metal–dielectric composites or (2) well-structured
building blocks. In both of these cases, the dimensional sizes of the constituted
structures are designed intentionally smaller than the wavelength. In such a
condition, the composite nanostructure is assumed to be an effective
continuum medium. The overall effective optical effects can be described by
the previously known dielectric functions of individual components and their
volume fractions. There are several analytical approaches to derive the
effective electromagnetic response of the composited materials, which are
often known as the effective medium approach.19,20
32 Chapter 1

In the effective medium theory, the electromagnetic responses of the

composites are assumed to be electric dipoles, and the collective responses of
the electric dipoles take on an overall dielectric response of the matter. In this
case, the diffuse field is neglected because it is much smaller compared to the
coherent field. Several models of the effective medium theory have been
proposed, e.g., the Maxwell–Garnett theory, the Bruggerman model, and the
asymptotic multiscale theory.21

1.5.2 Topologies of metal–dielectric composites

The effective medium approach attempts to describe the heterogeneous
medium using effective dielectric functions (effective permittivity and
permeability) or optical functions (effective refractive index). Without losing
generality in the following sections, the media made of metal–dielectric
components are assumed to be composed of isotropic inclusions.
Two typical topologies are presented as shown in Fig. 1.14. When the
relative concentration of inclusion particles embedded in the host material is
small and the inclusions have well-defined shapes, this type of composite
topology is called Maxwell–Garnett geometry.21 When the percentage
compositions of two constituent materials are almost identical to each other,
as shown in Fig. 1.14(b), the two compositions play comparable roles in its
optical parameters. This type of heterogeneous topology is called Brugge-
man geometry.22 Corresponding to these topology geometries are two
effective medium approaches that are widely used to obtain the effective
parameters, i.e., Maxwell–Garnett theory (MGT) and Bruggeman effective
medium theory (EMT).

1.5.3 Lorentz cavity model

As illustrated in Fig. 1.15, a Lorentz cavity model is used to evaluate the local
field surrounding the particle. In this model, an individual particle immersed
in a uniform medium is excited by an applied field. The external electric field

Figure 1.14 Typical topology geometries of metal–dielectric composites: (a) Maxwell–

Garnett geometry and (b) Bruggeman geometry. The bright and dark areas represent the
host dielectric components and the embedded metal inclusions, respectively.
Optical Properties of Plasmonic Materials 33

EO
Ed

EI
A

P B

Figure 1.15 Lorentz cavity model to calculate the local field of the central particle.

E induces the redistribution of surface charges of the cavity model, as marked

in the two-sided ordinate. The charge redistribution results in an additional
depolarized field Ed, which results in the polarized surface charges of the
central particle as labeled in bilateral symmetry. The surface charges lead to
the polarization field EI inside the particle.
The local field EL at the site of the particle refers to a microscopic field,
which acts on the particle and fluctuates sharply in the Lorentz cavity. Thus,
the local field EL is decomposed into four components11

EL ¼ E þ Ed þ EI þ Ep : (1.67)

It is notable that the term Ep denotes the field induced by other particles
dispersed throughout the whole system. When the particles are randomly
distributed, such as the conditions of particles dispersing in a liquid or gas
environment, the term Ep would vanish due to the counterbalance of the
symmetric lattice. In the familiar cubic crystal lattice, the Ep term is also zero
due to the lattice symmetry.
The depolarized local field Ed relates the macroscopic polarization P by the
relation Ep ¼ –P/ε0, where ε0 is the permittivity of the free space in the
aforementioned cavity. It is assumed that the particles are immersed in the host
medium of vacuum. Both E and Ed are field values averaged within the entire
composite medium. For the electric fields inside the particle, the total charge is
integrated over its normal surface. Because of the structural lattice symmetry,
the total field EI, ascribed to all of surface charges accumulated on the particles,
is parallel with the external field. The total polarization field is estimated with a
magnitude of11,24
EI ¼ P=3ε0 : (1.68)
Therefore, the local field upon the particle is

EL ¼ E þ P=3ε0  P=ε0 : (1.69)

Following the treatment derived from the Drude model, the polarization P is
related to the electric dipole moment. The macroscopic polarization is
expressed by
34 Chapter 1

P ¼ f aðE þ P=3ε0 Þ, (1.70)

where f is the volume density of the particles, and a is the polarizability of the
individual particle. A combination of the constitutive equation P ¼ ε0(εr – 1)E
and Eq. (1.70) leads to the particle polarizability formulated with the relative
dielectric permittivity
εr  1 a
¼f , (1.71a)
εr þ 2 3ε0

3ε0 εr  1
a¼ , (1.71b)
f εr þ 2
where ε ¼ ε0εr is the permittivity of the individual particle. This notion is called
the Clausius–Mossotti theory, which calculates the microscopic polarizability of
an individual particle from the observable dielectric function of bulk materials.
It provides a distinct connection between the electric response of a microscopic
particle and the macroscopic behavior of a bulk material.
If the particle polarizability volume a0 ¼ 4pε0a(m3) is used in Eq. (1.71),
the Clausius–Mossotti formula is rewritten in the form of
ε  ε0 4p 0
¼ Va : (1.72)
ε þ 2ε0 3
The following section adopts the particle polarizability a, as in Eqs. (1.71).

1.5.4 Maxwell–Garnett theory

In a metal–dielectric composite medium with the Maxwell–Garnett geometry,
in which the volume fraction follows that shown in Fig. 1.16(a), the metal
particles and the dielectric constituent are viewed as the inclusions and the
host medium, respectively. When the metallic particles are spherical particles
and the host medium is an isotropic dielectric substance, the Clausius–
Mossotti relation of Eq. (1.71) becomes
fa ε  εh
¼ : (1.73a)
3ε0 εh ε þ 2εh
When considering a composite inclusion with permittivity εi, the parameter
ε in Eq. (1.71a) represents the effective permittivity of the composite inclusion.
In these conditions, the polarizability in Eq. (1.69b) becomes
εi  εh
a ¼ 3ε0 εh f i , (1.73b)
εi þ 2εh
where fi is the volume-filling fraction of particles in the composite, and εh and
εi are the relative permittivity of the host medium and the composite particle,
Optical Properties of Plasmonic Materials 35

(a) (b)
2.6 80

2.2 60
Re (ε)/εh

Im (ε)/εh
1.8 40

1.4 20

1.0 0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
f f

Figure 1.16 Effective dielectric functions as a function of the volume fraction for composite
materials as predicted by the Maxwell–Garnett and Bruggeman formula: (a) dielectric
function of a composite for permittivity ratio εi/εh ¼ 3, and (b) imaginary part of the permittivity
for εi/εh ¼ 5  j100. Both feature predictions using the Maxwell–Garnett (solid) and the
Bruggeman (dashed) formula.

respectively. Equation (1.73b) is substituted into Eq. (1.73a), and the

Clausius–Mossotti relation is rewritten as
ε  εh ε  εh
¼ fi i : (1.74)
ε þ 2εh εi þ 2εh
The equivalent dielectric function of a metal–dielectric composite medium is
obtained as

1 þ 2Gi εi  εh
ε ¼ εh , Gi ¼ f i : (1.75)
1  Gi εi þ 2εh
Equation (1.75) formulates the equivalent dielectric function of a composite
medium by the individual dielectric functions of the inclusions and the host
material, i.e., the Maxwell–Garnett theory. In the MGT, the embedded
particles and the host solid material are treated in an unsymmetrical manner.
When the difference of the dielectric functions between the two composite
materials are obvious, this asymmetry becomes particularly sharp. For a
dilute composite material with the volume-filling factor less than a unit away
fi ,, 1, the equivalent dielectric function of the composite materials appears
in the function expansion of the Taylor series as

ε ¼ εh þ 3Gi εh þ Oð3f 2i Þ: (1.76)

It indicates that the equivalent dielectric function of a dilute composite

material is a linear function of the volume density and the dielectric function
of their individual inclusions. Equation (1.75) is still valid in the limiting
conditions: when the constituted phase of inclusions vanishes fi ! 0, the
estimated dielectric function approaches that of the host material by ε ! εh.
36 Chapter 1

On the other hand, when the host medium phase vanishes fi ! 1, it is

predicted to be the dielectric function of inclusions by ε ! εi.
For a single metallic, spherical particle embedded in a host medium, the
equivalent dielectric function in Eq. (1.75) satisfies the resonance requisition
εi ¼ 2εh for a surface plasmon resonance, which means that the dielectric
function of the inclusion materials equals the negative double value of the
surrounding medium.

1.5.5 Bruggeman medium theory

The Maxwell–Garnett theory only predicts the approximation of equivalent
dielectric function in the dilute two-phase composite. In fact, with the
increasing of the volume fraction of inclusions, the deviation between the
effective properties predicted by the Maxwell–Garnett formula and by Lord
Rayleigh become obvious.24 Bruggeman proposed a widely known mean-field
theory to evaluate the effective dielectric function of composite media. In the
Bruggeman theory, the inclusions and the surrounding medium are weighted
symmetrically using the volume fraction of the individual components. For a
composite medium made of two inclusions, the formula is expressed as22
ε  εh ε  εh ε  εh
¼ f1 1 þf2 2 , (1.77)
ε þ 2εh ε1 þ 2εh ε2 þ 2εh
where εj (j ¼ 1, 2) and εh are the dielectric functions of the dispersed spherical
particles and the environment medium, and fj (j ¼ 1, 2) is the volume-filling
fraction of i-order phase inclusions. Because the two-phase inclusions appear
in a symmetric manner, the total sum of the volume-filling fractions equals
one unit. It can obviously be generalized to systems that contain more than
two phases by adding more terms:
.X X
f i ¼ ni nm , f i ¼ 1, (1.78)
m i

where m represents the total phase of inclusions. Furthermore, when ε = εh,

there is always a valid expression
ε1  ε ε ε
f1 þf2 2 ¼ 0: (1.79)
ε1 þ 2ε ε2 þ 2ε
The equivalent dielectric function to this quadratic equation is
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
εε
ε ¼ G2  ½G2 2 þ 1 2 , (1.80a)
2

3ðε1 f 1 þ ε2 f 2 Þ  ðε1 þ ε2 Þ
G2 ¼ : (1.80b)
4
Optical Properties of Plasmonic Materials 37

The þ/ sign in Eq. (1.80a) is determined to ensure that the imaginary part of
the obtained dielectric function is positive.11
Based on Eq. (1.79) for an effective medium expression, it is reasonable to
get the general theory for a composite material with multiple components in a
symmetrical manner
X εi  ε X
fi ¼ 0, f i ¼ 1: (1.81)
i
εi þ 2ε i

Equation (1.81) is the popular Bruggeman form of the EMT. This formula
can be expanded to any number of components in a variety of composite
materials because each component is treated equally in the mixture. For two-
phase composite materials, the Taylor series expansion of the Bruggeman
approach in the case of low-volume fractions (fi ,, 1) becomes

ε ¼ εh þ 3Gi εh þ Oð9f 2i Þ: (1.82)

The second order of volume fraction O(9Vi2) differs from the expansion in the
MGT in Eq. (1.76).
Figure 1.16 illustrates the effective dielectric function of a composite
material made of two components predicted by the Maxwell Garnett and the
Bruggeman formula. For a low-volume fraction of fi , 0.33, both the Maxwell–
Garnett and the Bruggeman formula present nearly the same curves. However,
when the volume fraction increases, the predictions become strongly different.
Especially for the imaginary part, the deviations increase sharply. For the
whole region, the Maxwell–Garnett prediction remains lower than that of the
Bruggeman estimate. The deviation of the imaginary part of the relative
dielectric functions suddenly becomes obvious after the volume fraction
increases over the threshold value fi  0.33.
When the shape effect of the inclusions are taken into account, the term
εi þ 2ε is replaced by εi þ (d1)ε in the Maxwell–Garnett and the Bruggeman
formula for a more-general d-dimensional inclusion. For a two-phase
d-dimensional particle, the Maxwell–Garnett formula becomes the general-
ized formula
ε  εh εi  εh
¼ fi , (1.83)
ε þ ðd  1Þεh εi þ ðd  1Þεh
and the Bruggeman expression becomes
ε1  ε ε2  ε
f1 þf2 ¼ 0, (1.84)
ε1 þ ðd  1Þε ε2 þ ðd  1Þε
where d refers to the dimensions of the inclusions determined by the shape and
the orientation of the nanoparticles with respect to the external electric field.
The value 2 is usually taken corresponding to spherical inclusions. For 2D
38 Chapter 1

metal–dielectric composite films, the resonance peak in the MGT will be

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
vp = 1 þ ðn  1Þεh , and the inflexion of the effective dielectric function in the
Bruggeman prediction will be 0.5.
The aforementioned volume–weight approaches for the mean value of
permittivity are just analytical approximations for the effective dielectric
function of composite nanostructures. The term “effective” indicates that they
are equivalent in physical concept. It is a simple implemention for low-
dimensional composite materials. A more-general approach for effective
medium analysis is the Bergman theory or the Bergman–Milton theory, which
is often called the spectral representation method. Further details about this
approach can be found in textbooks23 and the original papers.25

References
1. J. Aukong, Electromagnetic Wave Theory, John Wiley & Sons, New York
(1986).
2. P. Šolín, Partial Differential Equations and the Finite Element Method,
John Wiley & Sons, New York (2006).
3. N. Johnson et al, Electromagnetic Characterization of Complex Nano-
structured Materials: An Overview, ECONAM CSA project, http://
econam.metamorphose-vi.org/index.php?option=com_docman&task=
doc_download&gid=263&Itemid=24 (2011).
4. J. M. Lourtioz, Photonic Crystals: Towards Nanoscale Photonic Devices,
p. 122, Springer, Berlin (2005).
5. J. E. Conrad, Electromagnetic Waves and Radiating Systems, 2nd ed.,
Prentice Hall, Englewood Cliffs, NJ (1968).
6. A. vander Vorst, A. Rosen, and Y. Kotsuka, RF/Microwave Interaction
with Biological Tissues, John Wiley & Sons, New York (2006)
7. S. G. Rodrigo, F. J. García-Vidal, and L. Martín-Moreno, “Influence of
material properties on extraordinary optical transmission through holes
arrays,” Physical Review B 77(7), 075401 (2008).
8. M. S. Dresselhaus, “Solid State Physics Part II: Optical Properties of
Solids,” lecture notes, Massachusetts Institute of Technology, Cambridge,
MA (1999).
9. R. J. Van Overstraeten and R. P. Mertens, “Heavy doping effects in
silicon,” Solid State Electron. 30(11), 1077–1087 (1987).
10. E. D. Palik, Handbook of Optical Constants of Solids, Academic Press
Limited, New York (1998).
11. W. Cai and V. M. Shalaev, Optical Metamaterials: Fundamentals and
Applications, pp. 26–37, Springer, New York (2010).
12. D. Rakić, A. B. Djurišić, J. M. Elazar, and M. L. Majewski, “Optical
properties of metallic films for vertical-cavity optoelectronic devices,”
Appl. Optics 37(22), 5271–5283 (1998).
Optical Properties of Plasmonic Materials 39

13. J. B. Pendry, A. J. Holden, and D. J. Robbins, “Magnetism from

Conductors and Enhanced Nonlinear Phenomena,” IEEE Trans. Micro-
wave Theory Techniques 47(11), 2075–2084 (1999).
14. M. Dalarsson, Z. Jakšić, and P. Tassin, “Structures Containing Left-
Handed Metamaterials with Refractive Index Gradient: Exact Analytical
Versus Numerical Treatment,” Microwave Review, (2009).
15. S. C. Jain and D. J. Roulston, “A simple expression for band gap
narrowing (BGN) in heavily doped Si, Ge, GaAs and GexSi1–x strained
layers,” Solid-State Electronics 34(5), 453–465 (1991).
16. C. Kittel, Introduction to Solid State Physics, Wiley, New York (2004).
17. V. Lucarini, “Kramers-Kronig relations and sum rules in optical
harmonic generation processes: theory and applications,” Ph.D. disserta-
tion, University of Joensuu, Finland (2003).
18. V. Lucarini, J. J. Saarinen, K.-E. Peiponen, and E. M. Vartiainen, Kramer-
Kronig Relations in Optical Materials Research, Springer, Berlin (2005).
19. Wikipedia, “Effective medium approximations: for the Maxwell Garnett
equation and Bruggeman’s model,” http://en.wikipedia.org/wiki/
Effective_medium_approximations.
20. A. A. Maradudin, Structure Surfaces as Optical Metamaterials,
Cambridge University Press, Cambridge, UK (2011).
21. C. C. Tuck, Effective Medium Theory, Oxford University Press, Oxford,
UK (1999).
22. D. E. Aspnes, “Local-field effects and effective medium theory: A
microscopic perspective,.” Am. J. Phys. 50(8) (1982).
23. J. C. M. Garnett, “Colours in metal glasses and in metallic films,” Phil.
Trans. R. Soc. Lond. 203, 237–288 (1905).
24. G. W. Milton, The Theory of Composites, Cambridge University Press,
Cambridge, UK (2002).
25. D. J. Bergman, “The dielectric constant of a composite material—A
problem in classical physics,” Phys. Rep. 43(9), 377–407 (1978).