Mackay 2008

E.
Wolf, Progress in Optics 51

© 2008 Elsevier B.V.
All rights reserved
Chapter 3
Electromagnetic fields in linear bianisotropic

mediums
by
Tom G. Mackay
School of Mathematics, The University of Edinburgh, Edinburgh EH9 3JZ, Scotland, UK
e-mail: T.Mackay@ed.ac.uk
Akhlesh Lakhtakia
Department of Engineering Science and Mechanics, The Pennsylvania State University,
University Park, Pennsylvania 16802-6812, USA
e-mail: akhlesh@psu.edu
ISSN: 0079-6638 DOI: 10.1016/S0079-6638(07)51003-6

121
Contents
Page
§ 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
§ 2. The Maxwell postulates and constitutive relations . . . . . . . . . . . 124
§ 3. Linear mediums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
§ 4. Plane-wave propagation . . . . . . . . . . . . . . . . . . . . . . . . . 156
§ 5. Dyadic Green functions . . . . . . . . . . . . . . . . . . . . . . . . . 175
§ 6. Homogenization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
§ 7. Closing remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
122
§ 1. Introduction
An isotropic medium has electromagnetic properties which are the same in all di-
rections. However, the notion of isotropy, as encountered in elementary treatments
of electromagnetics, is an abstraction which requires qualification when applied
to real materials. For example, liquids and random composite mediums may be
isotropic on a statistical basis, while cubic crystals are isotropic when viewed
at macroscopic length-scales. Electromagnetically isotropic mediums are charac-
terized simply by scalar constitutive parameters which relate the induction field
phasors D and H to the primitive field phasors E and B. Often, naturally occur-
ring materials and artificially constructed mediums are more accurately described
as anisotropic rather than isotropic. Anisotropic mediums exhibit directionally
dependent electromagnetic properties, such that D and E are not aligned or H
and B are not aligned. Dyadics (i.e., second-rank Cartesian tensors) are needed to
relate the primitive and the induction field phasors in anisotropic mediums.
Bianisotropy is the natural generalization of anisotropy. In the electromagnetic
description of a bianisotropic medium, both D and H are anisotropically coupled
to both E and B. Hence, in general, a linear bianisotropic medium is character-
ized by four 3×3 constitutive dyadics. Though seldom described in standard text-
books, bianisotropy is commonplace. Suppose a certain medium is characterized
as an isotropic dielectric medium by an observer in an inertial reference frame Σ.
The same medium generally exhibits bianisotropic properties when viewed by an
observer in another reference frame that translates at uniform velocity with re-
spect to Σ. Aside from relativistic scenarios, bianisotropic effects are observed
at low frequencies and temperatures in a host of naturally occurring minerals
(O’Dell [1970], Schmid [2003]). Furthermore, the phenomenon of bianisotropy
looks set to play an increasingly important role in the rapidly burgeoning fields re-
lating to complex composite mediums. In particular, bianisotropic mediums may
be readily conceptualized through the process of homogenization of a composite
of two or more constituent mediums. Thereby, metamaterials may be realized that
exhibit novel electromagnetic properties which are not exhibited (or at least not
exhibited to the same degree) by the constituent mediums (Walser [2003]).
123
124 Electromagnetic fields in linear bianisotropic mediums [3, § 2
A broad overview of electromagnetic bianisotropy is provided in the follow-

ing sections. Our survey includes anisotropic mediums as an important subcat-
egory of bianisotropic mediums. Where appropriate, the presentation of bian-
isotropic and anisotropic properties is preceded by a brief description of the
corresponding properties for isotropic mediums. We are confined to passive
mediums throughout this chapter. In § 2 the constitutive relations for bian-
isotropic mediums, and their general properties, are discussed. Commonly en-
countered classifications of bianisotropic mediums are presented in § 3. The
propagation of plane waves in bianisotropic mediums is considered in § 4.
The Green-function method is widely used in source-field problems. Repre-
sentations of dyadic Green functions for bianisotropic mediums are described
in § 5. The conceptualization of bianisotropic mediums as homogenized com-
posite mediums is the subject of § 6. A few closing remarks are provided
in § 7.
The following notation is adopted: 6-vectors (3-vectors) are in bold (normal)
face and underlined, whereas 6 × 6 (3 × 3) dyadics are in bold (normal) face and
twice underlined. The caret symbol ˆ placed over a 3-vector signifies a unit vector,
and the unit Cartesian vectors are denoted by x̂, ŷ and ẑ. The adjoint, determinant,
inverse, transpose and trace of a dyadic Q are represented by adj Q, det Q, Q−1 ,
QT and tr Q, respectively. Furthermore,
⎫
diag(Qx , Qy , Qz ) = Qx x̂ x̂ + Qy ŷ ŷ + Qz ẑẑ⎪
⎬

Q 0 . (1.1)
Diag(Q , Q ) = 1 ⎪
⎭
1 2 0 Q
2
The 6×6 (3×3) identity dyadic is represented by I (I ); the 4×4 identity matrix is
written as I. The tilde symbol ˜ placed over a field dyadic or vector indicates that
the quantity is in the time domain; the same quantity in the frequency domain is
denoted without a tilde. Contraction of indexes for dyadics and vectors is denoted
by the dot product. The permittivity and permeability of free space (i.e., vacuum
in the absence of any gravitational field) are written as 0 and μ0 , respectively,
and c0 = (0 μ0 )−1/2 represents the speed of light in free space. The angular
frequency is ω. The real and imaginary parts of z ∈ C are specified as Re{z} and
Im{z}, respectively, whereas z∗ denotes the complex conjugate of z.
§ 2. The Maxwell postulates and constitutive relations
The action of complex assemblies of matter on electromagnetic waves lies at the

heart of this chapter. The atomic (and sub-atomic) nature of matter does not di-
3, § 2] The Maxwell postulates and constitutive relations 125
rectly concern us here, since we adopt a macroscopic viewpoint wherein the elec-
tromagnetic wavelengths are large compared with interatomic distances. Thus, we
refer in this chapter to ‘mediums’ and ‘materials’ rather than ‘assemblies of atoms
and molecules’. The essential theoretical basis is provided by the Maxwell postu-
lates for macroscopic fields combined with constitutive relations, but we keep in
mind that electromagnetism is fundamentally a microscopic science. In this sec-
tion, we present the salient features of macroscopic electromagnetic theory which
underpin the remainder of the chapter.
2.1. Maxwell postulates
The basic framework for our description of electromagnetic anisotropy and bian-
isotropy is constructed in terms of the four macroscopic electromagnetic fields
Ẽ(r, t), D̃(r, t), B̃(r, t) and H̃ (r, t). These are piecewise differentiable vector
functions of position r and time t which arise as spatial averages of microscopic
fields and bound sources (Jackson [1999]). The fields Ẽ(r, t) and B̃(r, t) are
directly measurable quantities which produce the Lorentz force. Accordingly,
Ẽ(r, t) and B̃(r, t) are viewed as the primitive fields. The fields D̃(r, t) and
H̃ (r, t) develop within a medium in response to the primitive fields; hence, they
are considered as induction fields. Conventionally, Ẽ(r, t) and D̃(r, t) are called
the electric field and the dielectric displacement, respectively. The conventional
terms for B̃(r, t) and H̃ (r, t), namely the magnetic induction and magnetic field,
respectively, are confusing and are avoided in this chapter.
The physical principles governing the behaviour of Ẽ(r, t), D̃(r, t), B̃(r, t)
and H̃ (r, t) are encapsulated by the Maxwell curl postulates
⎫
∂
∇ × H̃ (r, t) − D̃(r, t) = J̃ e (r, t) ⎪ ⎪
⎬
∂t
(2.1)
∂ ⎪
⎪
∇ × Ẽ(r, t) + B̃(r, t) = −J̃ m (r, t) ⎭
∂t
and divergence postulates

∇ • D̃(r, t) = ρ̃e (r, t)
. (2.2)
∇ • B̃(r, t) = ρ̃m (r, t)
The terms on the right sides of (2.1) and (2.2) represent sources of fields. Whereas
J̃ e (r, t) and ρ̃e (r, t) are the externally impressed electric current and electric
charge densities, respectively, the magnetic current and magnetic charge densi-
ties – denoted by J̃ m (r, t) and ρ̃m (r, t) – do not represent physical quantities
but are incorporated for mathematical convenience (Lakhtakia [1995]). In con-

sonance with our macroscopic viewpoint, the source terms are also piecewise
differentiable and satisfy the continuity relations
⎫
∇ • J̃ e (r, t) + ρ̃e (r, t) = 0 ⎪
∂
⎪
⎬
∂t
. (2.3)
⎪
∇ • J̃ m (r, t) + ρ̃m (r, t) = 0⎪
∂
⎭
∂t
A redundancy is implicit in eqs. (2.1)–(2.3), from the macroscopic viewpoint.
The continuity relations (2.3), when combined with the Maxwell curl postulates
(2.1), yield the Maxwell divergence postulates (2.2). Therefore, under the pre-
sumption of source continuity, there is no need for us to consider explicitly the
divergence postulates (2.2).
2.2. Constitutive relations
The Maxwell curl postulates (2.1) provide us with a system of two linear vec-
tor differential equations in terms of the two primitive vector fields Ẽ(r, t) and
B̃(r, t) and the two induction vector fields D̃(r, t) and H̃ (r, t). In order to solve
these differential equations, further information – in the form of constitutive rela-
tions relating the induction fields to the primitive fields – is needed. It is these con-
stitutive relations which characterize the electromagnetic response of a medium.
The constitutive relations may be naturally expressed in the general form

D̃(r, t) = F Ẽ(r, t), B̃(r, t)

, (2.4)
H̃ (r, t) = G Ẽ(r, t), B̃(r, t)
wherein F and G are linear/nonlinear functions of Ẽ(r, t) and B̃(r, t) for lin-
ear/nonlinear mediums.
In general, the electromagnetic response of a medium is nonlocal with respect
to both space and time. Thus, the constitutive relations of a linear medium should
be stated as (Weiglhofer [2003])
⎫

D̃(r, t) = ˜ EB (r , t ) • Ẽ(r − r , t − t ) ⎪
⎪
⎪
⎪
⎪
⎪

t r ⎪
⎪
3 ⎪
+ ξ̃ (r , t ) B̃(r − r , t − t ) d r dt ⎪
• ⎬
EB
⎪
, (2.5)
⎪
H̃ (r, t) = ζ̃ (r , t ) • Ẽ(r − r , t − t ) ⎪
⎪
⎪
⎪
EB ⎪
⎪
t r
⎪
⎪
• 3 ⎭
+ ν̃ EB (r , t ) B̃(r − r , t − t ) d r dt
where ˜ EB (r, t), ξ̃ (r, t), ζ̃ (r, t) and ν̃ EB (r, t) are constitutive dyadics
EB EB
(i.e., second-rank Cartesian tensors) that can be interpreted as 3 × 3 matrixes.
While spatial nonlocality can play a significant role when the wavelength is com-
parable to some characteristic length-scale in the medium (Ponti, Oldano and Bec-
chi [2001]), it is commonly neglected and lies outside the scope of this chapter.
On the other hand, temporal nonlocality is almost always a matter of central im-
portance, because of the high speeds of electromagnetic signals. We therefore
concentrate here upon linear, spatially local, constitutive relations of the form

⎫
D̃(r, t) = ˜ EB (r, t ) • Ẽ(r, t − t ) + ξ̃ (r, t ) • B̃(r, t − t ) dt ⎪
⎪
⎪
EB ⎪
⎬
t
.
• •
⎪ ⎪
H̃ (r, t) = ζ̃ (r, t ) Ẽ(r, t − t ) + ν̃ EB (r, t ) B̃(r, t − t ) dt ⎪ ⎪
EB ⎭
t
(2.6)
2.3. The frequency domain
The convolution integrals appearing in the constitutive relations (2.6) result in

mathematical complexities when those relations are substituted in the Maxwell
postulates. A widely used technique to circumvent those difficulties (often without
loss of essential physics) is to introduce the temporal Fourier transforms via
∞
Z(r, ω) = Z̃(r, t) exp(iωt) dt, (2.7)
−∞
with Z standing in for EB , ξ , ζ , ν EB , E, D, B and H , while ω is called the
√ EB EB
angular frequency and i = −1. After taking the temporal Fourier transforms of
(2.6) and implementing the convolution theorem (Arfken and Weber [1995]), the
frequency-domain constitutive relations emerge as

D(r, ω) = EB (r, ω) • E(r, ω) + ξ (r, ω) • B(r, ω)
EB
. (2.8)
H (r, ω) = ζ (r, ω) • E(r, ω) + ν EB (r, ω) • B(r, ω)
EB
Often in electromagnetic theory, Ẽ(r, ω) is partnered with H̃ (r, ω) rather than
B̃(r, ω); for example, in the formulation of boundary conditions and the defin-
ition of the Poynting vector (Chen [1983]). Consequently, it is convenient and
widespread practice to express frequency-domain constitutive relations as

D(r, ω) = EH (r, ω) • E(r, ω) + ξ (r, ω) • H (r, ω)
EH
. (2.9)
B(r, ω) = ζ (r, ω) • E(r, ω) + μ (r, ω) • H (r, ω)
EH EH
The names Boys–Post and Tellegen are often associated with the constitutive rela-
tions (2.8) and (2.9), respectively (Weiglhofer [1998a]). A one-to-one correspon-
dence between the Boys–Post representation and the Tellegen representation is
straightforwardly established via (Weiglhofer [2003])
⎫
EB (r, ω) = EH (r, ω) − ξ (r, ω) • μ−1 (r, ω) • ζ (r, ω)⎪
⎪
EH EH EH ⎪
⎪
ξ (r, ω) = ξ (r, ω) • μ−1 (r, ω) ⎪
⎬
EB EH EH
(2.10)
ζ (r, ω) = −μ−1 (r, ω) • ζ (r, ω) ⎪
⎪
⎪
EB EH EH ⎪
⎪
⎭
ν EB (r, ω) = μ−1 (r, ω)
EH
and
⎫
EH (r, ω) = EB (r, ω) − ξ (r, ω) • ν −1 (r, ω) • ζ (r, ω)⎪
⎪
EB EB EB ⎪
⎪
ξ (r, ω) = ξ (r, ω) • ν −1 (r, ω) ⎪
⎬
EH EB EB
, (2.11)
ζ (r, ω) = −ν −1 (r, ω) • ζ (r, ω) ⎪
⎪
⎪
EH EB EB ⎪
⎪
⎭
μ (r, ω) = ν −1
EB
(r, ω)
EH
wherein the invertibility of ν EB (r, ω) and μ (r, ω) has been assumed. We

EH
largely make use of the Tellegen representation in this chapter, but with occa-
sional recourse to the Boys–Post representation where necessary.
The mathematical simplicity of the frequency-domain formulation as com-
pared to the time-domain formulation is gained at a cost in terms of physi-
cal interpretation. The frequency-dependent constitutive dyadics EB,EH (r, ω),
ξ (r, ω), ζ (r, ω), ν EB (r, ω) and μ (r, ω) are complex-valued quan-
EB,EH EB,EH EH
tities; so also are the frequency-dependent field phasors E(r, ω), D(r, ω),
B(r, ω) and H (r, ω). The real-valued physical entities they represent surface
only indirectly upon subjecting them to the inverse temporal Fourier transform.
In this chapter, phasors are also called fields – in keeping with widespread usage.
The corresponding frequency-domain Maxwell curl postulates develop as

∇ × H (r, ω) + iωD(r, ω) = J e (r, ω)
, (2.12)
∇ × E(r, ω) − iωB(r, ω) = −J m (r, ω)
where the source terms J e,m (r, ω) are the Fourier transforms of J̃ e,m (r, t), de-
fined as in (2.7) with Z = J e,m . The constitutive relations (2.9) – or equally
(2.8) – together with the Maxwell curl postulates (2.12) form a self-consistent
system into which anisotropy and bianisotropy are incorporated.
2.4. 6-vector/6 × 6 dyadic notation
The use of a 6-vector/6 × 6 dyadic notation allows the Tellegen constitutive rela-
tions (2.9) to be expressed very compactly as
C(r, ω) = K EH (r, ω) • F(r, ω), (2.13)
with the 6-vectors
T
C(r, ω) = D(r, ω), B(r, ω) (2.14)
and
T
F(r, ω) = E(r, ω), H (r, ω) (2.15)
containing components of the electric and magnetic fields, while the 6 × 6 consti-
tutive dyadic

EH (r, ω) ξ (r, ω)
K EH (r, ω) = EH . (2.16)
ζ (r, ω) μ (r, ω)
EH EH
The result of combining the constitutive relations (2.9) with the Maxwell curl
postulates (2.12) is thereby succinctly expressed as

L(∇) + iωK EH (r, ω) • F(r, ω) = Q(r, ω), (2.17)
with the linear differential operator

0 ∇ ×I
L(∇) = , (2.18)
−∇ × I 0
and the source 6-vector
T
Q(r, ω) = J e (r, ω), J m (r, ω) . (2.19)
In a similar fashion, the four 3 × 3 dyadics EB (r, ω), ξ (r, ω), ζ (r, ω)
EB EB
and ν EB (r, ω), which specify the constitutive properties in the Boys–Post repre-
sentation (2.8), may be represented by the 6 × 6 constitutive dyadic

EB (r, ω) ξ (r, ω)
K EB (r, ω) = EB . (2.20)
ζ (r, ω) ν EB (r, ω)
EB
The transformations (2.11) and (2.10) may then be expressed in terms of the in-
vertible 6 × 6 dyadic function τ which we define through the following relation-
ships:

K EB (r, ω) ≡ τ K EH (r, ω)

EH (r, ω) − ξ (r, ω) • μ−1 (r, ω) • ζ (r, ω) ξ (r, ω) • μ−1 (r, ω)
= EH
−1
EH EH EH
−1
EH
−μ (r, ω) • ζ (r, ω) μ (r, ω)
EH EH EH
(2.21)
and

K EH (r, ω) ≡ τ −1 K EB (r, ω)

EB (r, ω) − ξ (r, ω) • ν −1
EB
(r, ω) • ζ (r, ω) ξ (r, ω) • ν −1
EB
(r, ω)
= EB
−1
EB EB
.
−ν EB (r, ω) • ζ (r, ω) ν −1
EB
(r, ω)
EB
(2.22)
2.5. Form invariances
The Maxwell postulates retain their form under certain linear coordinate-and-field
transformations, thereby leading to the concepts of spatial and temporal invari-
ances as well as spatiotemporal covariance. While spatiotemporal covariance is
of immense theoretical importance, invariances with respect to spatial and tem-
poral transformations are commonly applied in many practical situations. Chiral
invariance, which captures the non-uniqueness of the Maxwell postulates under
linear field transformations, is also significant. An invariance of the Maxwell pos-
tulates in the frequency domain to a certain transformation involving complex
conjugates has recently been reported. Finally in this subsection, implications
of various transformations on electromagnetic energy and momentum are out-
lined.
2.5.1. Time reversal

We denote the operation of time reversal by T , i.e., T {t} = −t. Under the pre-
sumption that electric and magnetic source densities transform as (Kong [1972])

T ρ̃e (r, t) = ρ̃e (r, −t)

, (2.23)
T ρ̃m (r, t) = −ρ̃m (r, −t)
the continuity relations (2.3) yield

T J̃ e (r, t) = −J̃ e (r, −t)

, (2.24)
T J̃ m (r, t) = J̃ m (r, −t)
and the electromagnetic fields must therefore transform as

T Ẽ(r, t) = Ẽ(r, −t), T {D̃(r, t)} = D̃(r, −t)

(2.25)
T B̃(r, t) = −B̃(r, −t), T H̃ (r, t) = −H̃ (r, −t)
in order to preserve the form of the Maxwell postulates. From the definition of the
temporal Fourier transform (2.7), we see that the frequency-domain counterparts
of (2.23), (2.24) and (2.25) are

⎫
T ρe (r, ω) = ρe∗ (r, ω), T J e (r, ω) = −J ∗e (r, ω)⎪
⎪

⎪
⎪
T ρm (r, ω) = −ρm ∗
(r, ω), T J m (r, ω) = J ∗m (r, ω) ⎬

∗

, (2.26)
T E(r, ω) = E (r, ω), T D(r, ω) = D ∗ (r, ω) ⎪ ⎪
⎪

⎪
⎭
T B(r, ω) = −B ∗ (r, ω), T H (r, ω) = −H ∗ (r, ω)
where the superscript ∗ indicates the complex conjugate. Therefore, under time
reversal, the constitutive dyadics transform as

⎫
T EB,EH (r, ω) = ∗EB,EH (r, ω) ⎪ ⎪

⎪
⎪
T ξ (r, ω) = −ξ ∗ (r, ω) ⎪
⎪
⎪
EB,EH EB,EH ⎪
⎬

∗
T ζ (r, ω) = −ζ (r, ω) , (2.27)

EB,EH

EB,EH ⎪
⎪
⎪
⎪
T ν EB (r, ω) = ν ∗EB (r, ω) ⎪
⎪

⎪
⎪
T μ (r, ω) = μ (r, ω) ∗ ⎭
EH EH
by virtue of (2.8) and (2.9). The time-reversal asymmetry which is exhibited by
the magnetoelectric constitutive dyadics ξ (r, ω) and ζ (r, ω) origi-
EB,EH EB,EH
nates from irreversible physical processes, such as can develop through the appli-
cation of quasistatic biasing fields or by means of relative motion (Post [1997]).
We enlarge upon these matters in § 3 in the context of Faraday chiral mediums
and Lorentz-transformed constitutive dyadics.
2.5.2. Spatial inversion

Let us turn now to the inversion of space, denoted by the operator P as P{r} =
−r. Similarly to the time-reversal scenario, if we assume that the electric and
magnetic charge densities transform as (Kong [1972])

P ρ̃e (r, t) = ρ̃e (−r, t)

, (2.28)
P ρ̃m (r, t) = −ρ̃m (−r, t)
then

P J̃ e (r, t) = −J̃ e (−r, t)

(2.29)
P J̃ m (r, t) = J̃ m (−r, t)
by virtue of the continuity relations (2.3); furthermore, the form invariance of the
Maxwell postulates enjoins the relationships

P Ẽ(r, t) = −Ẽ(−r, t), P D̃(r, t) = −D̃(−r, t)

. (2.30)
P B̃(r, t) = B̃(−r, t), P H̃ (r, t) = H̃ (−r, t)
The action of the spatial-inversion operator P on the field quantities is not altered
upon switching from the time domain to the frequency domain. Thus, from the
constitutive relations (2.8) and (2.9) we find

⎫
P EB,EH (r, ω) = EB,EH (−r, ω) ⎪ ⎪

⎪
⎪
P ξ (r, ω) = −ξ (−r, ω) ⎪
⎪
⎪
EB,EH EB,EH ⎪
⎬

P ζ (r, ω) = −ζ (−r, ω) . (2.31)

EB,EH

EB,EH ⎪
⎪
⎪
⎪
P ν EB (r, ω) = ν EB (−r, ω) ⎪
⎪

⎪
⎪
P μ (r, ω) = μ (−r, ω) ⎭
EH EH
2.5.3. Lorentz covariance
Suppose an inertial reference frame Σ moves with constant velocity v = v v̂

with respect to an inertial reference frame Σ. The Lorentz transformation (Chen
[1983])
⎫
r = Y • r − γ vt ⎪
⎬
r •v , (2.32)
t =γ t− 2 ⎪ ⎭
c0
relates the space–time coordinates (r , t ) in Σ to the space–time coordinates

(r, t) in Σ, wherein
⎫
Y = I + (γ − 1)v̂ v̂ ⎪
⎪
⎬
1
2 −2
γ = 1−β ⎪
, (2.33)
⎪
⎭
β = c0−1 v
The Maxwell postulates are Lorentz covariant, which means that they retain
their form under the spatiotemporal transformation (2.32). The Lorentz covari-
ance of the Maxwell postulates has far-reaching implications for the constitutive
relations that develop in uniformly moving reference frames, as mentioned in
§ 3.3.1.
2.5.4. Chiral invariance
As well as being form-invariant under spatial, temporal and spatiotemporal trans-

formations described in the previous subsections, the Maxwell postulates do not
change under the following transformation of fields (Lakhtakia [1995])

⎫
Rψ Ẽ(r, t) = Ẽ(r, t) cos ψ − Z H̃ (r, t) sin ψ ⎪
⎪

⎪
⎪
Rψ H̃ (r, t) = Z Ẽ(r, t) sin ψ + H̃ (r, t) cos ψ ⎬
−1

(2.34)
Rψ B̃(r, t) = B̃(r, t) cos ψ + Z D̃(r, t) sin ψ ⎪
⎪

⎪
⎪
−1 ⎭
Rψ D̃(r, t) = −Z B̃(r, t) sin ψ + D̃(r, t) cos ψ
and source densities

⎫
Rψ ρ̃e (r, t) = ρ̃e (r, t) cos ψ − Z −1 ρ̃m (r, t) sin ψ ⎪ ⎪

⎪
⎪
Rψ ρ̃m (r, t) = Z ρ̃e (r, t) sin ψ + ρ̃m (r, t) cos ψ ⎬

. (2.35)
Rψ J̃ e (r, t) = J̃ e (r, t) cos ψ − Z −1 J̃ m (r, t) sin ψ ⎪
⎪
⎪

⎪
⎭
Rψ J̃ m (r, t) = Z J̃ e (r, t) sin ψ + J̃ m (r, t) cos ψ
Here ψ is a complex-valued angle and the scalar Z is an impedance re-

quired to maintain dimensional integrity. When ψ ∈ R (the set of all real
numbers), the transformation operator Rψ represents a rotation of the fields.
Hence, the Maxwell postulates are said to possess the property of chiral invari-
ance.
An especially interesting transformation arises for ψ = π/2: The electric and
magnetic fields, and similarly the electric and magnetic charge densities, inter-
change under Rπ/2 , which is often called the duality transformation (Jackson
[1999]). Owing to the duality of the electric charge and the magnetic charge, it
is merely a matter of convention whether a particular charged particle is said to
possess an electric charge or a magnetic charge.
Furthermore, the question of the existence of magnetic monopoles is more fun-
damentally the question of whether all charged carriers possess the same pro-
portion of electric charge and magnetic charge. If the answer to this question
is in the affirmative, then – by applying a Rψ transformation with the appro-
priate choice of ψ – either the magnetic monopole or the electric monopole
could be said to not exist. Let us also note that duality is best considered glob-
ally (i.e., for all mediums, at all times, and everywhere) so that the appropriate
choice of ψ is made for physical certainty; however, that choice does not pre-
clude the later application of duality in a local context for mathematical conve-
nience.
The constitutive relations (2.6) retain their form under the transformation of
fields (2.34) provided that the constitutive dyadics transform as

⎫
Rψ ˜ EB (r, t) = cos2 ψ ˜ EB (r, t) + Z −2 sin2 ψ μ̃ (r, t) ⎪⎪
EB ⎪
⎪
− Z −1 sin ψ cos ψ ξ̃ (r, t) + ζ̃ (r, t) ⎪ ⎪
⎪
⎪
EB EB ⎪
⎪

⎪
Rψ ξ̃ (r, t) = sin ψ cos ψ Z ˜ EB (r, t) − Z μ̃ (r, t) ⎪
−1 ⎪
⎪
⎪
EB EB ⎪
⎪
+ cos ψ ξ̃ (r, t) − sin ψ ζ̃ (r, t)
2 2 ⎪
⎬
EB EB

. (2.36)
Rψ ζ̃ (r, t) = sin ψ cos ψ Z ˜ EB (r, t) − Z −1 μ̃ (r, t) ⎪ ⎪
⎪
EB EB ⎪
⎪
+ cos2 ψ ζ̃ (r, t) − sin2 ψ ξ̃ (r, t) ⎪
⎪
⎪
⎪
EB EB ⎪
⎪

⎪
Rψ μ̃ (r, t) = Z sin ψ ˜ EB (r, t) + cos ψ μ̃ (r, t) ⎪
2 2 2 ⎪
⎪
EB EB ⎪⎪
⎪
+ Z sin ψ cos ψ ξ̃ (r, t) + ζ̃ (r, t) ⎭
EB EB
2.5.5. Conjugate invariance

The frequency-domain Maxwell curl postulates (2.12) are invariant under a fur-
ther transformation, namely the conjugate transformation denoted by C. The
conjugate-transformed fields are defined as (Lakhtakia [2004a])

C E(r, ω) = E ∗ (r, ω), C H (r, ω) = H ∗ (r, ω)

, (2.37)
C D(r, ω) = −D ∗ (r, ω), C B(r, ω) = −B ∗ (r, ω)
whereas the source densities transform as

C ρe (r, ω) = −ρe∗ (r, ω), C ρm (r, ω) = −ρm ∗
(r, ω)

. (2.38)
C J e (r, ω) = J ∗e (r, ω), C J m (r, ω) = J ∗m (r, ω)
When applied to linear materials, the Maxwell postulates remain invariant, pro-
vided that the constitutive dyadics undergo the following transformations:

C EB (r, ω) = − ∗EB (r, ω), C ξ (r, ω) = ξ ∗ (r, ω)

EB

EB
(2.39)
C ζ (r, ω) = ζ ∗ (r, ω), C ν EB (r, ω) = −ν ∗EB (r, ω)
EB EB
and

C EH (r, ω) = − ∗EH (r, ω), C ξ (r, ω) = −ξ ∗ (r, ω)

EH
EH
. (2.40)
C ζ (r, ω) = −ζ ∗ (r, ω), C μ (r, ω) = −μ (r, ω)
EH EH EH EH
The conjugation symmetry represented in eqs. (2.37)–(2.40) arises as a gener-
alization of the transformation which reverses the sign of the real permittivity
and permeability scalars for isotropic dielectric–magnetic mediums (Lakhtakia
[2004a]). The significance of this transformation is discussed further in § 4, in the
context of negative-phase-velocity mediums. The effect of the conjugate trans-
formation (2.39) would be observable in, for example, plane-wave propagation
through a material slab.
2.5.6. Energy and momentum

It is of practical interest to consider how the spatial, temporal and field transfor-
mations influence measurable quantities. To this end, let us draw attention to the
energy flow density, as quantified via the instantaneous Poynting vector
S̃(r, t) = Ẽ(r, t) × H̃ (r, t); (2.41)
the total energy density
1
W̃ (r, t) = D̃(r, t) • Ẽ(r, t) + B̃(r, t) • H̃ (r, t) ; (2.42)
2
and the Maxwell stress dyadic
1
T̃ (r, t) = − D̃(r, t) • Ẽ(r, t) + B̃(r, t) • H̃ (r, t) I
2
+ D̃(r, t)Ẽ(r, t) + B̃(r, t)H̃ (r, t). (2.43)
By the straightforward application of the field transformations (2.25) and (2.30),
respectively, we see that

⎫
T S̃(r, t) = −S̃(r, −t)⎪ ⎬

T W̃ (r, t) = W̃ (r, −t) (2.44)

⎪
⎭
T T̃ (r, t) = T̃ (r, −t)
and

⎫
P S̃(r, t) = −S̃(−r, t)⎪ ⎬

P W̃ (r, t) = W̃ (−r, t) . (2.45)

⎪
⎭
P T̃ (r, t) = T̃ (−r, t)
The chiral invariance of the Maxwell postulates carries over to measurable
quantities. From (2.34), it follows immediately that

⎫
Rψ S̃(r, t) = S̃(r, t) ⎪ ⎬

Rψ W̃ (r, t) = W̃ (r, t) . (2.46)

⎪
⎭
Rψ T̃ (r, t) = T̃ (r, t)
Therefore, the electromagnetic fields cannot be uniquely determined from mea-
surements of electromagnetic energy and/or momentum.
2.6. Constitutive dyadics
Let us now look more closely at the constitutive dyadics which characterize the
electromagnetic response of a medium. In the most general linear scenario, the
6 × 6 constitutive dyadic K EH (r, ω) comprises 36 complex-valued parameters.

This vast parameter space may be reduced through the imposition of physical
constraints which require the constitutive dyadics to satisfy certain symmetries.
Also, we often have occasion to limit our attention to special cases and idealiza-
tions which manifest as symmetries of the constitutive dyadics.
2.6.1. Constraints
2.6.1.1. Causality and Kramers–Kronig relations The formulations of consti-
tutive relations must conform to the principle of causality; i.e., an ‘effect’ must
appear after its ‘cause’. So, neither can a cause and its effect be simultaneous, nor
can an effect precede its cause. The principle of causality is most transparently
implemented in the time domain for constitutive relations of the form (2.4).
The induced fields D̃(r, t) and H̃ (r, t) develop in response to the primitive
fields Ẽ(r, t) and B̃(r, t), such that

D̃(r, t) = 0 Ẽ(r, t) + P̃ (r, t)
, (2.47)
H̃ (r, t) = μ−1
0 B̃(r, t) − M̃(r, t)
where 0 = 8.854 × 10−12 F m−1 and μ0 = 4π × 10−7 H m−1 are the permittiv-
ity and permeability of free space, respectively. The polarization P̃ (r, t) and the
magnetization M̃(r, t) indicate the electromagnetic response of a medium, and
must therefore be causally connected to the primitive fields.
Therefore, with regard to the time-domain linear constitutive relations (2.5)–
(2.6), causality dictates that
⎫
˜ EB (r, t) − 0 δ(r)I ≡ 0 ⎪ ⎪
⎪
⎪
ξ̃ (r, t) ≡ 0 ⎪
⎬
EB
for t 0, (2.48)
ζ̃ (r, t) ≡ 0 ⎪
⎪
EB ⎪
⎪
⎪
⎭
μ−10 δ(r)I − ν̃ EB (r, t) ≡ 0
where δ(·) is the Dirac delta function.1 When translated into the frequency do-
main, the causality requirement (2.48) gives rise to integral relations between the
real and imaginary parts of the frequency-dependent constitutive parameters, as
follows.
Let the scalar function f˜(r, t) represent an arbitrary component of a Boys–Post
constitutive dyadic; i.e., f˜(r, t) is a component of ˜ EB (r, t) − 0 δ(r)I , ξ̃ (r, t),
EB
1 The Dirac delta function is defined in (5.5).

ζ̃ (r, t) or μ−1 ˜
EB 0 δ(r)I − ν̃ EB (r, t). The temporal Fourier transform of f (r, t)
may be expressed as
∞
f (r, ω) = f˜(r, t) exp(iωt) dt, (2.49)
0
wherein the causality constraint (2.48) has been applied to set the lower limit
of integration equal to zero. The analytic continuation of f (r, ω) in the upper
complex ω plane is provided by the Cauchy integral formula

1 f (r, s)
f (r, ω) = ds, (2.50)
2πi s−ω
where the integration contour extends around the upper half plane. The integrand
in (2.50) vanishes as |s| → ∞ for Im{s} > 0 due to the exp(iωt) factor occurring
in the integral representation (2.49). Hence, the contour integral (2.50) reduces to
an integral along the real axis. Counting the single pole on the real axis at ω = s
as a half-residue, we have
∞
1 f (r, s)
f (r, ω) = P ds, (2.51)
πi s−ω
−∞
where P indicates the Cauchy principal value. Therefore, we see that the real and
imaginary parts of f (r, ω) are related as the Hilbert transforms
⎫
∞ ⎪

1 Im{f (r, s)} ⎪
⎪
Re f (r, ω) = P ds ⎪
⎪
π s−ω ⎪
⎪
−∞
⎬
. (2.52)
∞ ⎪

1 Re{f (r, s)} ⎪ ⎪
⎪
Im f (r, ω) = − P ds ⎪
⎪
⎪
π s−ω ⎭
−∞
Finally, since f˜(r, t) is real-valued, the symmetry condition
f (r, −ω) = f ∗ (r, ω) (2.53)

relates f to its complex conjugate f ∗ . Thus, (2.52) yield the Kramers–Kronig

relations (Post [1997])2
⎫
∞ ⎪
⎪

2 s Im{f (r, s)} ⎪
Re f (r, ω) = P ds ⎪
⎪
s −ω
2 2 ⎪
⎪
π ⎬
0
. (2.54)
∞ ⎪

2 ω Re{f (r, s)} ⎪ ⎪
Im f (r, ω) = − P ds ⎪
⎪
⎪
π s 2 − ω2 ⎪
⎭
0
Although the relations (2.54) are presented here for components of the Boys–
Post constitutive dyadics, analogous relations hold for components of the Tellegen
constitutive dyadics by virtue of eqs. (2.11).
The Kramers–Kronig relations are a particular example of dispersion relations
that apply generally to frequency-dependent, causal, linear systems (Hilgevoord
[1962]). These may be usefully employed in experimental determinations of con-
stitutive parameters (Bohren and Huffman [1983]).
2.6.1.2. Post constraint A structural constraint, known as the Post constraint, is

available for those linear mediums which exhibit magnetoelectric coupling (Post
[1997]). The Post constraint may be expressed as

tr ζ (r, ω) − ξ (r, ω) = 0, (2.55)
EB EB
or equivalently as

tr μ−1 (r, ω) • ζ (r, ω) + ξ (r, ω) = 0. (2.56)

EH EH EH
As a result, only 35 independent complex-valued parameters are needed to char-
acterize the most general linear medium.
The origins of the Post constraint lie in the microscopic nature of the prim-
itive electromagnetic fields and the Lorentz covariance of the Maxwell equa-
tions (Lakhtakia [2004b]). Post established his eponymous constraint more than
40 years ago (Post [1997]). Two additional independent proofs, one based on a
uniqueness requirement (Lakhtakia and Weiglhofer [1996a, 1996b]) and another
based on multipole considerations (de Lange and Raab [2001]), further secured
the standing of the Post constraint. However, there is recent experimental evidence
that the Post constraint is violated at low frequencies (Hehl, Obukhov, Rivera and
Schmid [in press]), but there is no microscopic understanding as yet of this ev-
idence. The incorporation of the hitherto-undiscovered axion will lead to a re-
2 An alternative approach to the derivation of the Kramers–Kronig relations, exploiting the proper-
ties of Herglotz functions, has recently been reported by King [2006].
evaluation of the Post constraint even for free space (Hehl and Obukhov [2005],
Lakhtakia [2006]).
2.6.1.3. Onsager relations The Onsager relations are a set of reciprocity re-
lations which are applicable generally to coupled linear phenomena at macro-
scopic length-scales (Onsager [1931a, 1931b], Casimir [1945]). They were orig-
inally established for instantaneous phenomenons, but their scope was extended
by means of the fluctuation–dissipation theorem (Callen and Greene [1952]) to in-
clude time-harmonic phenomenons as well (Callen, Barasch and Jackson [1952]).
Central to the Onsager relations is the assumption of microscopic reversibility.
Consequently, in order to apply the Onsager relations to electromagnetic consti-
tutive relations, the contribution of free space must be excluded because micro-
scopic processes cannot occur in free space. The frequency-dependent quantities
P (r, ω) and M(r, ω), which are the temporal Fourier transforms of the polar-
ization P̃ (r, t) and the magnetization M̃(r, t), represent the electromagnetic re-
sponse of a medium in relation to the electromagnetic response of free space. For
linear homogeneous mediums, the constitutive relations (2.8) reduce to
⎫
P (r, ω) = EB (ω) − 0 I • E(r, ω) + ξ (ω) • B(r, ω) ⎬
EB

.
M(r, ω) = − ζ (ω) • E(r, ω) + ν (ω) − μ−1 I • B(r, ω) ⎭
EB 0
EB
(2.57)
In an external magnetostatic field B dc , the application of the Onsager relations
to the constitutive relations (2.57) yields the constraints (Lakhtakia and Depine
[2005])
⎫
η (ω)B = ηT (ω)−B (η = , ν)⎬
EB EB
dc dc , (2.58)
ξ (ω) B dc
= ζ T (ω) −B dc
⎭
EB EB
where the superscript T denotes the transpose operation. The equivalent con-
straints for the Tellegen constitutive dyadics follow immediately from (2.10) as
⎫
η (ω)B = ηT (ω)−B (η = , μ)⎬
EH EH
dc dc . (2.59)
ξ (ω) B dc
= −ζ T (ω) −B dc
⎭
EH EH
2.6.2. Specializations
2.6.2.1. Lorentz reciprocity The issue of Lorentz reciprocity – which is closely
related to the topics of time reversal and the Onsager relations – crops up fre-
quently in theoretical studies involving complex mediums (Altman and Suchy
[1991]). It is often described in terms of the interchangeability of transmitters and

receivers.
Let us consider two frequency-domain electric source current densities, namely
J ae (r, ω) and J be (r, ω), and two frequency-domain magnetic source current den-
sities, namely J am (r, ω) and J bm (r, ω). The sources labelled a generate fields de-
scribed by the electric and magnetic fields E a (r, ω) and H a (r, ω), whereas the
sources labelled b generate fields described by the electric and magnetic fields
E b (r, ω) and H b (r, ω). The interaction of the a sources with the field generated
by the b sources is quantitated as the reaction (Kong [1986])

a

a, b = J e (r, ω) • E b (r, ω) − J am (r, ω) • H b (r, ω) d3 r, (2.60)

Va
where the integration region Va contains the a sources. Similarly, the interac-
tion of the b sources with the field generated by the a sources is represented by
the reaction
b, a. If the medium which supports J a,b

e,m (r, ω), E
a,b (r, ω) and
a,b
H (r, ω) is such that

a, b =
b, a, (2.61)
then it is called Lorentz-reciprocal.
Combining the Tellegen constitutive relations (2.9) with the Maxwell curl pos-
tulates (2.12) and integrating thereafter, we obtain the reaction difference

a, b −
b, a

b
= −iω E (r, ω) • EH (r, ω) − TEH (r, ω) • E a (r, ω)
Va ∪Vb

+ H a (r, ω) • μ (r, ω) − μT (r, ω) • H b (r, ω)
EH EH

+ E b (r, ω) • ξ (r, ω) + ζ T (r, ω) • H a (r, ω)
EH EH

+ H a (r, ω) • ζ (r, ω) + ξ T (r, ω) • E b (r, ω) d3 r, (2.62)

EH EH
where the integration region Va ∪ Vb contains both the sources a and b. Thus,
Lorentz reciprocity is signalled by (Krowne [1984])
⎫
η (r, ω) = ηT (r, ω) (η = , μ)⎬
EH EH
. (2.63)
ξ (r, ω) = −ζ T (r, ω) ⎭
EH EH
The corresponding symmetries for the Boys–Post representation follow immedi-
ately from (2.11) as
⎫
η (r, ω) = ηT (r, ω) (η = , ν)⎬
EB EB
. (2.64)
ξ (r, ω) = ζ T (r, ω) ⎭
EB EB
The Lorentz-reciprocity conditions (2.63) and (2.64) coincide with the Onsager
relations (2.59) and (2.58), respectively, in the absence of a magnetostatic field.
Many widely studied anisotropic and bianisotropic mediums are required to sat-
isfy the Lorentz-reciprocity conditions. Lorentz-reciprocal mediums arise com-
monly as dielectric and magnetic crystals, whereas plasmas are not Lorentz-
reciprocal, as becomes clear from § 3.
2.6.2.2. Neglect of dissipation By virtue of the principle of causality, no

medium – with the unique exception of free space, which is not a mater-
ial – responds instantaneously to an applied electromagnetic field (Chen [1983],
Weiglhofer and Lakhtakia [1996]). Therefore, dissipation is exhibited by all ma-
terial mediums. However, on occasion it may be expedient to neglect dissipation,
particularly if attention is confined to a narrow ω-range wherein dissipation is tiny
over length scales of interest.
In order to focus upon dissipation in the frequency domain, we introduce the
time-averaged Poynting vector
1
S(r, ω) t = Re E(r, ω) × H ∗ (r, ω) (2.65)

2
of a monochromatic field. Of particular relevance to us here is the divergence
of (2.65). Utilizing the Maxwell curl postulates (2.12) in the absence of sources
(i.e., J e,m (r, ω) ≡ 0), together with the Tellegen constitutive relations (2.9), we
find that eq. (2.65) provides (Kong [1986])
iω ∗
∇ • S(r, ω) t = E (r, ω) • EH (r, ω) − +EH (r, ω) • E(r, ω)
4

+ H ∗ (r, ω) • μ (r, ω) − μ+ (r, ω) • H (r, ω)
EH EH

+ E ∗ (r, ω) • ξ (r, ω) − ζ + (r, ω) • H (r, ω)
EH EH

+ H ∗ (r, ω) • ζ (r, ω) − ξ + (r, ω) • E(r, ω) ,

EH EH
(2.66)
with the superscript + indicating the conjugate transpose. A medium is nondissi-
pative provided that ∇ •
S(r, ω)t = 0. Thus, dissipation is neglected by enforc-
ing the equalities (Chen [1983])
⎫
η (r, ω) = η+ (r, ω) (η = , μ)⎬
EH EH
, (2.67)
ξ (r, ω) = ζ + (r, ω) ⎭
EH EH
Table 1
The conditions imposed upon the Tellegen constitutive parameters by Lorentz reciprocity and by the
neglect of dissipation; three forms of K EH are represented
K EH Lorentz-reciprocal Nondissipative
I ξ I ξ = −ζ = ∗
ξ = ζ∗
ζ I μI
μ = μ∗
⎡ ⎤ ∗
11 0 0 ξ11 0 0 ξ = −ζ =
⎢ 0 22 0 0 ξ22 0 ⎥ ∗
ξ = ζ
⎢ 0 0 33 0 0 ξ33 ⎥
⎢ ⎥
⎢ ζ11 0 0
μ11 0 0 ⎥ μ = μ∗
⎣ ⎦
0 ζ22 0 0 μ22 0
0 0 ζ33 0 0 μ33
⎡ ⎤ ∗
11 12 13 ξ11 ξ12 ξ13 m = m m = m
⎢ 21 22 23 ξ21 ξ22 ξ23 ⎥ ξm = −ζm ∗
ξm = ζm
⎢ 31 32 33 ⎥
⎢ ξ31 ξ32 ξ33 ⎥ μm = μm
⎢ ⎥ μm = μ∗m
⎢ ζ11 ζ12 ζ13 μ
11 μ12

μ13 ⎥
⎣ μ21 μ22 23
μ ⎦
ζ21 ζ22 ζ23
ζ31 ζ32 ζ33 μ31 μ32 μ33
Notice that for the medium represented in the first example, the Lorentz-reciprocity condition ξ = −ζ
must be satisfied in order to comply with the Post constraint.
or equivalently
⎫
η (r, ω) = η+ (r, ω) (η = , ν)⎬
EB EB
. (2.68)
ξ (r, ω) = −ζ + (r, ω) ⎭
EB EB
The distinction between the conditions for the neglect of dissipation and for
Lorentz reciprocity must be noted. These are summarized in table 1 for three
commonly encountered forms of the constitutive dyadic K EH (r, ω).
§ 3. Linear mediums
The notions of anisotropy and bianisotropy may be simply characterized in

terms of what they are not: the adjectives ‘anisotropic’ and ‘bianisotropic’ de-
scribe those mediums which are not isotropic. Therefore, prior to discussing
anisotropic and bianisotropic mediums in this section, we first establish our
terms of reference by briefly considering isotropic mediums. We then present
a survey of the commonly encountered classifications of anisotropy and bian-
isotropy, as expressed in terms of their constitutive relations and constitutive
dyadics.
3, § 3] Linear mediums 143
3.1. Isotropy
The primitive fields and induced fields in an isotropic medium are co-directional
so that its constitutive dyadics reduce to scalars.
3.1.1. Free space

Let us begin with the most fundamental medium in electromagnetics: free space,
otherwise referred to as vacuum. From our viewpoint of classical physics, free
space is devoid of all matter. Quantum-electrodynamical processes, through
which energetic fluctuations may become interwoven with free space, are not
relevant to us here. Thus, free space represents the reference medium, relative
to which the electromagnetic responses of all mediums are gauged (Weiglhofer
[2003]).
Our discussion of the electromagnetic properties of free space in this section
pertains to flat space–time exclusively. The subject of free space in generally
curved space–time is taken up in § 3.4.2.
By definition, free space is clearly both isotropic and homogeneous. Its con-
stitutive properties are specified by the scalar permittivity 0 and scalar perme-
ability μ0 . Free space holds the unique distinction of being the only medium for
which there is an exact, spatiotemporally local relationship between the induc-
tion fields and the primitive fields. The time-domain constitutive relations of free
space are

D̃(r, t) = 0 Ẽ(r, t)
, (3.1)
H̃ (r, t) = μ−1
0 B̃(r, t)
and its frequency-domain constitutive relations have the same form:

D(r, ω) = 0 E(r, ω)
. (3.2)
H (r, ω) = μ−1
0 B(r, ω)
If axions are ever discovered, eqs. (3.1) and (3.2) shall have to be modified by the
incorporation of magnetoelectric terms that violate the Post constraint (Hehl and
Obukhov [2005]).
3.1.2. Dielectric–magnetic mediums

The simplest material mediums are the isotropic, homogeneous, dielectric–mag-
netic mediums. Their electromagnetic properties are characterized in terms of the
scalar permittivity (ω) and scalar permeability μ(ω) by the frequency-domain
Tellegen constitutive relations

D(r, ω) = (ω)E(r, ω)
. (3.3)
B(r, ω) = μ(ω)H (r, ω)
Causality demands that both and μ be ω-dependent, complex-valued parameters
with Im{} > 0 and Im{μ} > 0. The placement of and μ in the upper half of
the complex plane follows from the sign convention for the exponent of Fourier
transform kernel (2.7).
Parenthetically, we remark that though isotropic dielectric–magnetic mediums
have been very widely studied since the earliest days of electromagnetics, their
properties are still the subject of ongoing research. For example, negative refrac-
tion – which follows from the condition
Re{(ω)} Re{μ(ω)}
+ <0 (3.4)
Im{(ω)} Im{μ(ω)}
being satisfied at a particular frequency ω – is a topic of intense current interest
(Lakhtakia, McCall and Weiglhofer [2003], Litchinitser, Gabitov, Maimistov and
Shalaev [2008]); see § 4.7 for further details.
3.1.3. Isotropic chirality and biisotropy

A homogeneous, isotropic chiral medium is described by the frequency-domain

D(r, ω) = (ω)E(r, ω) + χ(ω)H (r, ω)
. (3.5)
B(r, ω) = −χ(ω)E(r, ω) + μ(ω)H (r, ω)
Three complex-valued, ω-dependent scalars – permittivity , permeability μ and
the chirality parameter χ – characterize it (Lakhtakia [1994]).
Notice that, in view of the definition presented in § 2.6.2, isotropic chiral medi-
ums are Lorentz-reciprocal. The corresponding non-reciprocal isotropic medium,
as would be described by the Tellegen constitutive relations

D(r, ω) = (ω)E(r, ω) + χ1 (ω)H (r, ω)
, (3.6)
B(r, ω) = χ2 (ω)E(r, ω) + μ(ω)H (r, ω)
with χ1 = −χ2 , is in violation of the Post constraint (Lakhtakia [2004b]). Such a
medium is called a biisotropic medium.
3.2. Anisotropy
The primitive fields and induced fields in isotropic mediums are co-directional. In
contrast, the defining characteristic of anisotropic mediums is that E and D are
not aligned and/or B and H are not aligned. Hence, whereas scalars provide the
constitutive characterizations for isotropic mediums, the constitutive properties of

anisotropic mediums are described using dyadics.
Numerous examples of dielectric anisotropy crop up within the context of crys-
tal optics (Born and Wolf [1980], Nye [1985]). Nematic and smectic liquid crys-
tals fall into the anisotropic dielectric category (Chandrasekhar [1992], de Gennes
and Prost [1993]). Magnetic anisotropy is an important characteristic of many dia-
magnetic and paramagnetic substances (Nye [1985]).
3.2.1. Uniaxial anisotropy

At macroscopic length scales, a single distinguished axis identifies a uniaxial
medium. The distinguished axis may originate from microscopic attributes, such
as the underlying crystalline structure or particulate geometry. Let the unit vec-
tor û point in the direction of the distinguished axis. The frequency-domain con-
stitutive relations of a uniaxial, homogeneous, dielectric medium are then given
in the Tellegen representation as

D(r, ω) = uni (ω) • E(r, ω)
, (3.7)
B(r, ω) = μ0 H (r, ω)
where the ω-dependent permittivity dyadic uni may be expressed as
uni (ω) = (ω)(I − ûû) + u (ω)ûû. (3.8)
The permittivity scalars and u are complex-valued quantities; causality requires

that Im{} > 0 and Im{u } > 0.
Without any loss of generality, we may choose to orient our coordinate system
in such a way that the distinguished direction coincides with the z axis. Thereby,
uni (ω) acquires the diagonal matrix form

uni (ω) = diag (ω), (ω), u (ω) . (3.9)
Similarly, a uniaxial, homogeneous magnetic medium may be described by the


D(r, ω) = 0 E(r, ω)
, (3.10)
B(r, ω) = μ (ω) • H (r, ω)
uni
with
μ (ω) = μ(ω)(I − ûû) + μu (ω)ûû. (3.11)

uni
The ω-dependent parameters μ and μu of causal mediums lie in the upper half of
the complex plane. The choice û = ẑ leads to the permeability matrix representa-
tion

μ (ω) = diag μ(ω), μ(ω), μu (ω) . (3.12)
uni
A uniaxial dielectric–magnetic medium is described by the Tellegen constitu-
tive relations

D(r, ω) = uni (ω) • E(r, ω)
. (3.13)
B(r, ω) = μ (ω) • H (r, ω)
uni
3.2.2. Biaxial anisotropy

The natural generalization of the uniaxial dielectric medium specified by eqs. (3.7)–
(3.9) is the orthorhombic biaxial dielectric medium described by the Tellegen
constitutive relations

D(r, ω) = ortho
bi
(ω) • E(r, ω)
, (3.14)
B(r, ω) = μ0 H (r, ω)
where the permittivity dyadic ortho
bi
(ω) has the diagonal form

ortho
bi
(ω) = diag x (ω), y (ω), z (ω) . (3.15)
Causality dictates that the imaginary parts of the ω-dependent permittivity scalars
x , y and z are positive-valued.
While the diagonal dyadic representation (3.15) can be mathematically conve-
nient, it is not particularly insightful from a physical perspective. The equivalent
representation (Chen [1983])
ortho
bi
(ω) = a (ω)I + b (ω)(ûm ûn + ûn ûm ) (3.16)
highlights the biaxial symmetry via the two unit vectors ûm and ûn (Weiglhofer
and Lakhtakia [1999]). In the absence of dissipation, the ω-dependent permittivity
scalars a and b are real-valued and the unit vectors ûm and ûn are aligned with
the optic ray axes. In a general direction through a biaxial medium, electromag-
netic radiation may propagate in one of two different modes; the two modes are
distinguished from each other by their different rates of energy flow. However, in
the two privileged directions provided by the optic ray axes, only one energy ve-
locity is permissible (Born and Wolf [1980]).3 For dissipative mediums, both a
3 The optic ray axes should not be confused with the optic axes that identify the two privileged
directions in which electromagnetic waves may propagate through a biaxial medium with only one
phase velocity. See § 4.4.2.
and b are necessarily complex-valued, in which case Re{a , b } and Im{a , b }

are generally associated with different pairs of (ûm , ûn ) axes (Mackay and Wei-
glhofer [2000]).
More general biaxial, homogeneous, dielectric mediums may be characterized
by the Tellegen constitutive relations
⎫
mono (ω) ⎪
⎪
D(r, ω) = tri bi • E(r, ω)⎬
bi (ω) , (3.17)
⎪
⎪
B(r, ω) = μ H (r, ω) ⎭
0
where the 3 × 3 permittivity dyadics have the symmetric forms

⎛ ⎞
x (ω) α (ω) 0
mono
bi
(ω) = ⎝ α (ω) y (ω) 0 ⎠ (3.18)
0 0 z (ω)
and
⎛ ⎞
x (ω) α (ω) β (ω)
tri (ω) = ⎝ α (ω) y (ω) γ (ω) ⎠ . (3.19)
bi
β (ω) γ (ω) z (ω)
The constitutive permittivity dyadics which specify biaxial anisotropy are sum-
marized in table 2.
As in the orthorhombic scenario, the ω-dependent permittivity scalars x , y
and z are complex-valued, with causality dictating that Im{x , y , z } > 0. The
off-diagonal ω-dependent scalars α , β and γ are real-valued. Biaxial medi-
ums belonging to the monoclinic and triclinic crystal systems (Nye [1985]) are
described by (3.18) and (3.19), respectively. The distinction between the three
Table 2
The constitutive permittivity dyadics which describe the three biaxial crystal
systems
Crystal Constitutive Complex-valued and
system dyadic form real-valued scalars
Orthorhombic x , y , z ∈ C
x 0 0
0 y 0
0 0 z
x , y , z ∈ C
Monoclinic x α 0
α ∈ R
α y 0
0 0 z
x , y , z ∈ C
Triclinic x α β
α y γ α , β , γ ∈ R
β γ z
biaxial crystal systems stems from the symmetries of the primitive unit cell be-
longing to the underlying Bravais lattice. If all three basis vectors of the primitive
unit cell are orthogonal then the crystal structure is orthorhombic; only two basis
vectors are orthogonal in the monoclinic system; whereas there are no orthogonal
basis vectors for triclinic crystals (Ashcroft and Mermin [1976]).
Biaxiality is not restricted to dielectric mediums. In a precisely analogous man-
ner, biaxial magnetic mediums are classified. Thus, the Tellegen constitutive rela-
tions
⎫
D(r, ω) = 0 E(r, ω) ⎪
⎪
⎫ ⎪
⎪
μ ortho
⎪
(ω) ⎪ ⎪
⎬
bi ⎬ , (3.20)
B(r, ω) = μ
mono
(ω) • H (r, ω)⎪ ⎪
bi ⎪ ⎪
⎪
⎪ ⎪
μtri (ω) ⎭ ⎭
bi
describe orthorhombic, monoclinic and triclinic biaxial magnetic mediums, re-

spectively, where the corresponding permeability dyadics have the symmetric
forms
⎫
μortho (ω) = diag μx (ω), μy (ω), μz (ω) ⎪ ⎪
bi ⎛ ⎞⎪⎪
⎪
⎪
μx (ω) μα (ω) 0 ⎪
⎪
⎝ ⎠ ⎪
⎪
μ mono
(ω) = μα (ω) μy (ω) 0 ⎬
bi
0 0 μz (ω) ⎪ , (3.21)
⎛ ⎞ ⎪ ⎪
μx (ω) μα (ω) μβ (ω) ⎪
⎪
⎪
⎪
μtri (ω) = ⎝ μα (ω) μy (ω) μγ (ω) ⎠ ⎪ ⎪
bi ⎪
⎭
μβ (ω) μγ (ω) μz (ω)
with μx , μy , μz ∈ C and μα , μβ , μγ ∈ R.
A biaxial dielectric–magnetic medium is described by the Tellegen constitutive
relations

D(r, ω) = bi (ω) • E(r, ω)
, (3.22)
B(r, ω) = μ (ω) • H (r, ω)
bi
wherein the symmetric permittivity and permeability dyadics bi (ω) and μ (ω),
bi
respectively, may be of the orthorhombic, monoclinic or triclinic type.
3.2.3. Gyrotropy
The uniaxial and biaxial manifestations of anisotropy discussed in §§ 3.2.1

and 3.2.2 are represented mathematically in terms of symmetric permittivity and
permeability dyadics. Uniaxial and biaxial mediums are therefore Lorentz-reci-

procal. Let us turn now to a fundamentally different type of anisotropy, namely
gyrotropy, which is characteristic of mediums that are not Lorentz-reciprocal.
Manifestations of gyrotropy are found within the realm of magneto-optic medi-
ums (Mansuripur [1995], Gersten and Smith [2001]). As an illustrative example,
we consider an incompressible plasma of electrons in thermal motion (Felsen and
Marcuvitz [1994]). A quasistatic magnetic field B 0 = B0 û of strength B0 is ap-
plied in the direction of the unit vector û. Adopting a macroscopic viewpoint, we
assume that the plasma with charge density −n0 q and mass density n0 m is homo-
geneously distributed in free space. Additionally, the plasma density is taken to be
sufficiently low that collisions between electrons may be neglected. The average
electron velocity is ṽ(r, t).
The governing equations are provided by the time-domain Maxwell curl postu-
lates
⎫
∂ ⎪
∇ × H̃ (r, t) − 0 Ẽ(r, t) = −n0 q ṽ(r, t)⎪ ⎬
∂t
, (3.23)
∂ ⎪
⎪
∇ × Ẽ(r, t) + μ0 H̃ (r, t) = 0 ⎭
∂t
along with the equation of motion
∂
−n0 q Ẽ(r, t) + ṽ(r, t) × B 0 = n0 m ṽ(r, t). (3.24)
∂t
Using eq. (3.24) to eliminate ṽ(r, t) from eqs. (3.23) and thereupon comparing
with the source-free versions of the Maxwell curl postulates (2.1), we find that
the constitutive relations

D̃(r, t) = ˜ gyro Ẽ(r, t)
(3.25)
H̃ (r, t) = μ0 −1 B̃(r, t)
emerge (Weiglhofer [2003]). The permittivity dyadic operator ˜ gyro {•} is defined
formally by
−1
˜ gyro = 0 I + 0 ωp2 I ∂tt − ωc (û × I )∂t , (3.26)
√
with ωp = q n0 /m0 being the plasma frequency and ωc = qB0 /m being the
gyrofrequency. The shorthand notation ∂t ≡ ∂/∂t and ∂tt ≡ ∂ 2 /∂t 2 has been
introduced in eq. (3.26), and the inverse of the temporal differential operator may
be interpreted as
t
∂t−1 f (t) ≡ f (t ) dt . (3.27)
−∞
By straightforward dyadic manipulations, eq. (3.26) may be expressed in the

more convenient form (Chen [1983])

ωp2 ωp2 ωc
˜ gyro = 0 1 + 2 (I − ûû) + 2 û × I
ωc + ∂tt (ωc + ∂tt )∂t

ωp2
+ 1+ ûû , (3.28)
∂tt
where the interpretation
1
g(t) = f (t) ⇔ (C + ∂tt )g(t) = f (t) (3.29)
C + ∂tt
is to be understood.
The differential operators appearing in the representations (3.26) and (3.28)
signify temporal dispersion. Upon applying the temporal Fourier transform (2.7),
the differential operator ∂t transforms to −iω. Thus, the time-domain consti-
tutive relations (3.25) transform into the frequency-domain constitutive rela-
tions

D(r, ω) = gyro (ω) • E(r, ω)
(3.30)
B(r, ω) = μ0 H (r, ω)
in the Tellegen representation. The ω-dependent permittivity dyadic gyro is spec-
ified as

ωp2 iωp2 ωc
gyro (ω) = 0 1 + 2 (I − ûû) + û × I
ωc − ω 2 (ωc2 − ω2 )ω

ωp2
+ 1 − 2 ûû . (3.31)
ω
Without loss of generality, let us orient our coordinate system such that the
quasistatic biasing magnetic field is directed along the z axis; i.e., we choose
û ≡ ẑ. Thereby, the constitutive dyadic (3.31) may be reformulated as
⎛ ωp2 iωp2 ωc ⎞
1 + ω2 −ω 2 − (ω2 −ω 2 )ω 0
⎜ c c ⎟
⎜ 2 ωp2 ⎟
gyro (ω) = 0 ⎜ iω p ωc
1 + 0 ⎟. (3.32)
⎝ (ωc2 −ω2 )ω ωc2 −ω2 ⎠
ωp2
0 0 1− ω2
The gyrotropic form (3.32) is characterized by a uniaxial anisotropy in the direc-
tion of the quasistatic biasing magnetic field, together with antisymmetric terms
associated with the plane perpendicular to that direction.
With reference to the results of § 2.6.2, notice that the gyrotropic medium de-
scribed jointly by eqs. (3.30) and (3.31) is nondissipative. A more realistic plasma
representation may be developed through taking account of electron collisions.
This may be achieved formally via the substitution ω → ω − iΩ, where Ω ∈ R
is a suitably chosen damping parameter (Weiglhofer [2003]).
Finally, we mention that gyrotropy also arises in magnetic mediums, for exam-
ple in ferrites (Lax and Button [1962]). The Tellegen constitutive relations for a
gyrotropic magnetic medium may be expressed as

D(r, ω) = 0 E(r, ω)
, (3.33)
B(r, ω) = μ (ω) • H (r, ω)
gyro
with the ω-dependent permeability dyadic
μ (ω) = μ(ω)(I − ûû) + iμg (ω)û × I + μu (ω)ûû; (3.34)

gyro
μ, μu , μg ∈ R for nondissipative mediums and μ, μu , μg ∈ C for dissipative

mediums. When û = ẑ, eq. (3.34) may be recast in matrix notation as
⎛ ⎞
μ(ω) −iμg (ω) 0
μ (ω) = ⎝ iμg (ω) μ(ω) 0 ⎠. (3.35)
gyro
0 0 μu (ω)
3.3. Bianisotropy
Bianisotropy arises from the combination of anisotropy with magnetoelectric cou-

pling. Thus, in a bianisotropic medium, D is anisotropically coupled to both
E and B, and H is anisotropically coupled to both E and B as well (Weiglhofer
[2003]). In general, a linear bianisotropic medium is described by four 3 × 3 con-
stitutive dyadics.
3.3.1. Mediums moving at constant velocity

Isotropy and anisotropy are not invariant under the Lorentz space–time transfor-
mation (2.32). A medium which is isotropic or anisotropic in one inertial reference
frame is generally bianisotropic in all other inertial reference frames. However,
a homogeneous medium that is spatially local, but not temporally local, in one
reference frame is generally spatiotemporally nonlocal in another reference frame
(Lakhtakia and Weiglhofer [1996a], Censor [1998]). This fact confines analysis
on simply moving, physically realistic materials to a consideration only of plane
waves.
3.3.2. Uniaxial and biaxial bianisotropy

The unixial and biaxial classifications introduced in §§ 3.2.1 and 3.2.2, respec-
tively, for dielectric and magnetic mediums carry over to bianisotropic mediums
in a straightforward way. Thus, a general biaxial structure for a homogeneous
bianisotropic medium may be specified by the Tellegen constitutive relations
(Mackay and Weiglhofer [2001a])
T T
D(r, ω), B(r, ω) = K EH (ω) • E(r, ω), H (r, ω) , (3.36)
where the 6 × 6 constitutive dyadic K EH is assembled from four symmetric 3 × 3
dyadics as

EH (ω) ξ (ω)
K EH (ω) = EH
ζ (ω) μ (ω)
EH EH
⎡ (ω) (ω) (ω) ξ (ω) ξ (ω) ξ (ω) ⎤
x α β x α β
⎢ α (ω) y (ω) γ (ω) ξα (ω) ξy (ω) ξγ (ω) ⎥
⎢ β (ω) γ (ω) z (ω) ξ β (ω) ξγ (ω) ξz (ω) ⎥
=⎢
⎢ ζx (ω) ζα (ω) ζβ (ω) μx (ω) μα (ω) ⎥,
⎣ μβ (ω) ⎥ ⎦
ζα (ω) ζy (ω) ζγ (ω) μα (ω) μy (ω) μγ (ω)
ζβ (ω) ζγ (ω) ζz (ω) μβ (ω) μγ (ω) μz (ω)
(3.37)
with x,y,z , ξx,y,z , ζx,y,z , μx,y,z ∈ C and α,β,γ , ξα,β,γ , ζα,β,γ , μα,β,γ ∈ R. The
orthorhombic bianisotropic specialization follows from setting
⎫
α,β,γ (ω) = 0 ⎪ ⎪
⎬
ξα,β,γ (ω) = 0
. (3.38)
ζα,β,γ (ω) = 0 ⎪ ⎪
⎭
μα,β,γ (ω) = 0
3.3.3. Faraday chiral mediums

Faraday chiral mediums (FCMs) are conceptualized as homogenized composite
mediums (Engheta, Jaggard and Kowarz [1992]). Their bianisotropic nature arises
through combining
• natural optical activity, as exhibited by isotropic chiral mediums (Lakhtakia
[1994]), with
• Faraday rotation, as exhibited by gyrotropic mediums (Lax and Button [1962],
Chen [1983], Collin [1966]).
Two examples of FCMs are the chiroferrite medium which develops from the
homogenization of an isotropic chiral medium with a magnetically biased ferrite
(Weiglhofer, Lakhtakia and Michel [1998]), and the chiroplasma medium which
develops from the homogenization of an isotropic chiral medium with a magneti-

cally biased plasma (Weiglhofer and Mackay [2000]).
The frequency-domain constitutive relations of a homogeneous FCM are rigor-
ously established as (Weiglhofer and Lakhtakia [1998])
⎫
D(r, ω) = FCM (ω) • E(r, ω) + iξ (ω) • H (r, ω) ⎬
FCM
, (3.39)
B(r, ω) = −iξ (ω) E(r, ω) + μ
• (ω) • H (r, ω)⎭
FCM FCM
in the Tellegen representation, where the 3 × 3 constitutive dyadics

η (ω) = η(ω)I + ηu (ω) − η(ω) ûû + iηg (ω)û × I (η = , ξ, μ),
FCM
(3.40)
all have the same form. The constitutive parameters , u , g , ξ , ξg , ξu , μ, μg and
μu are complex-valued in general, but are real-valued for a nondissipative FCM.
The FCM structure described by eq. (3.40) is associated with the homoge-
nization of particulate composite mediums based on spherical particulate geom-
etry. More general FCM forms develop through the homogenization of compos-
ite mediums containing nonspherical particles (Weiglhofer and Mackay [2000],
Mackay, Lakhtakia and Weiglhofer [2001a]).
3.4. Nonhomogeneous mediums
So far in this section, the descriptions of anisotropy and bianisotropy have been
provided in terms of constitutive dyadics which are independent of the posi-
tion vector r. However, the macroscopic properties of many important classes
of complex mediums cannot be adequately described without accounting for the
dependency of the constitutive properties upon r. Nonhomogeneity increases the
complexity of our mathematical analyses; nevertheless, theoretical descriptions
of electromagnetic properties are well-established for certain nonhomogeneous
anisotropic mediums.
3.4.1. Periodic nonhomogeneity
One of the most extensively studied examples of anisotropy in nonhomoge-

neous mediums is provided by cholesteric liquid crystals (CLCs). The periodic
nonhomogeneity of CLCs stems from their helicoidal arrangement of aciculate
molecules. In essence, CLCs are periodically twisted, uniaxial, dielectric medi-
ums. Their frequency-domain constitutive relations are given by (Chandrasekhar
[1992], de Gennes and Prost [1993])

D(r, ω) = CLC (r, ω) • E(r, ω)
, (3.41)
B(r, ω) = μ0 H (r, ω)
where the permittivity dyadic

CLC (r, ω) = a (ω)I + b (ω) − a (ω) û(r)û(r) (3.42)
is described in terms of the complex-valued, ω-dependent, scalars a and b and
the unit vector û(r) which represents the local optic axis. After choosing the z axis
to coincide with the helicoidal axis, the unit vector û(r) may be expressed as
πz πz
û(r) ≡ û(z) = x̂ cos + ŷ sin , (3.43)
Ω Ω
where Ω is the dielectric periodicity along the z direction.
The helicoidal conformation of CLCs also forms the basis of the recently de-
veloped chiral sculptured thin films (CSTFs) (Venugopal and Lakhtakia [2000]).
These comprise parallel helical nanowires deposited on a substrate. The helical
shape engenders structural chirality, while the nanowires on the average are char-
acterized by local orthorhombic symmetry. CSTFs are of considerable technolog-
ical interest because of their optical properties.
The constitutive relations of a CSTF may expressed as

D(r, ω) = CSTF (r, ω) • E(r, ω)
, (3.44)
B(r, ω) = μ0 H (r, ω)
with the permittivity dyadic

CSTF (z, ω) = S z (z) • S y (ψ) • diag b (ω), c (ω), a (ω)
• S −1
y
(ψ) • S −1
z
(z). (3.45)
Here, a,b,c are three complex-valued, ω-dependent parameters. The tilt dyadic
⎛ ⎞
cos ψ 0 − sin ψ
S y (ψ) = ⎝ 0 1 0 ⎠ (3.46)
sin ψ 0 cos ψ
and rotation dyadic
⎛ ⎞
cos πz −h sin πz 0
Ωπz πzΩ
S z (z) = ⎝ h sin Ω cos Ω 0⎠, (3.47)
0 0 1
specify the nanowire morphology wherein Ω is the half-pitch of the helical nano-
wires, h = ±1 denotes the structural handedness, and ψ ∈ (0, π/2]. For further
details on the optical characteristics of CSTFs, the reader is referred to the spe-
cialist literature (Lakhtakia and Messier [2005]).
The generalization of the CSTF form (3.44)–(3.45) leads to (Lakhtakia and
Weiglhofer [1995, 1997a])

D(r, ω) = HBM (z, ω) • E(r, ω) + ξ (z, ω) • H (r, ω)
HBM
, (3.48)
B(r, ω) = ζ (z, ω) • E(r, ω) + μ (z, ω) • H (r, ω)
HBM HBM
which are the constitutive relations for a helicoidal bianisotropic medium (HBM).
Herein, all four of the 3×3 nonhomogeneous constitutive dyadics have the general
form
⎛ ⎞
η11 (z, ω) η12 (z, ω) η13 (z, ω)
η (z, ω) = ⎝ η21 (z, ω) η22 (z, ω) η23 (z, ω) ⎠ (η = , ξ, ζ, μ),
HBM
η31 (z, ω) η32 (z, ω) η33 (z, ω)
(3.49)
and are factorizable as
η (z, ω) = S z (z) • η (0, ω) • S −1
z
(z) (η = , ξ, ζ, μ). (3.50)
HBM HBM
3.4.2. Gravitationally induced bianisotropy

In the preceding subsections of § 3, the backdrop for our description of electro-
magnetic phenomena has been flat space–time. We now digress in order to show
that the classical vacuum in curved space–time is not isotropic but bianisotropic
instead.
In the general-relativistic literature, the space–time curvature induced by a
gravitational field is conventionally characterized using a symmetric covariant
tensor of rank 2, known as the metric (d’Inverno [1992]). In keeping with the no-
tational practices implemented throughout this chapter, we represent the space–
time metric by the symmetric, real-valued, 4 × 4 matrix g̃(r, t). The components
of g̃(r, t) are the space–time metric components. This matrix is spatiotemporally
nonhomogeneous and local, in general.
Following a standard procedure wherein the covariant Maxwell equations are
expressed in noncovariant form (Landau and Lifshitz [1975], Skrotskii [1957],
Plébanski [1960]), the electromagnetic response of vacuum in curved space–time
may be represented by the time-domain constitutive relations (Schleich and Scully
[1984], Lakhtakia and Mackay [2004a])
⎫
D̃(r, t) = 0 γ̃ (r, t) • Ẽ(r, t) − c0−1 Γ̃ (r, t) × H̃ (r, t) ⎬
(3.51)
B̃(r, t) = μ0 γ̃ (r, t) • H̃ (r, t) + c0−1 Γ̃ (r, t) × Ẽ(r, t)⎭
in Tellegen form. Herein, the 3 × 3 dyadic γ̃ (r, t) has components
[− det g̃(r, t)] 1/2

γ̃ (r, t) m = − g̃−1 (r, t) +1,m+1 , m ∈ {1, 2, 3} ,
[g̃(r, t)]1,1
(3.52)
and the 3-vector Γ̃ (r, t) is given as
−1 T
Γ̃ (r, t) = g̃(r, t) 1,1 g̃(r, t) 1,2 , g̃(r, t) 1,3 , g̃(r, t) 1,4 , (3.53)
with [g̃(r, t)],m denoting the (, m)th element of g̃(r, t). Thus, the mathemat-
ical description of electromagnetic fields in gravitationally affected vacuum is
isomorphic to the description of electromagnetic fields in a fictitious, instanta-
neously responding medium described by the constitutive relations (3.51). Con-
sequently, analytical techniques commonly used to investigate electromagnetic
problems (without considering the effects of gravitational fields) may be applied
to the study of gravitationally affected vacuum – courtesy of the medium de-
scribed by eqs. (3.51).
This fictitious medium is generally bianisotropic, the symmetries of its consti-
tutive dyadics being determined by those of the underlying metric. The constitu-
tive dyadics are both spatially nonhomogeneous and time-varying. The medium is
neither Lorentz-reciprocal in general (cf. § 2.6.2.1) nor dissipative (cf. § 2.6.2.2),
but it does satisfy the Post constraint (2.56). As the matrix g̃(r, t) is real sym-
metric, the permittivity and permeability dyadics 0 γ̃ (r, t) and μ0 γ̃ (r, t), respec-
tively, are orthorhombic and have the same eigenvectors. Finally, let us note that
in flat space–time the matrix g̃(r, t) simplifies to diag(1, −1, −1, −1) and the
constitutive relations (3.51) reduce to the familiar form (3.1).
§ 4. Plane-wave propagation
From a theoretical perspective, the passage of electromagnetic signals through

a complex medium is generally a complicated issue – and perhaps it is best
broached by considering the propagation of plane waves. Although plane waves
themselves are idealizations, being of limitless spatial and temporal extents and
possessing infinite energy, they can provide a reasonable understanding of fields
far away from their sources. Furthermore, realistic signals may be represented as
superpositions of plane waves.
Descriptions of plane-wave propagation in isotropic dielectric mediums, as
well as in certain relatively simple, anisotropic dielectric mediums, are widely
3, § 4] Plane-wave propagation 157
available in the literature (Born and Wolf [1980], Chen [1983], Nye [1985], Kong
[1986]). This is especially true when the effects of dissipation are neglected, the
topic then often coming under the heading of ‘crystal optics’. In this section, a sur-
vey of plane-wave solutions of the Maxwell curl postulates in anisotropic and
bianisotropic mediums, including dissipative and nonhomogeneous mediums, is
presented. We begin with a general description of plane waves. Next, as a pre-
cursor to the later subsections on anisotropic and bianisotropic mediums, a brief
outline of plane-wave propagation in isotropic mediums is provided. Thereafter,
we discuss the modes of uniform plane-wave propagation which are supported
by various anisotropic and bianisotropic mediums, including nonhomogeneous
mediums. Plane-wave propagation in isotropic mediums is independent of di-
rection of propagation, which contrasts sharply with plane-wave propagation in
anisotropic and bianisotropic mediums.
4.1. Uniform and non-uniform plane waves
Let us recall from § 2 that the electric and magnetic field phasors E(r, ω) and
H (r, ω), respectively, can be conveniently combined in the 6-vector field phasor
F(r, ω) defined in eq. (2.15). Electromagnetic plane waves may be represented
mathematically by field phasors of the form

F(r, ω) = F 0 (ω) exp i(k • r − ωt) , (4.1)
where F 0 (ω) ∈ C6 is independent of r. The wavevector k ∈ C3 , in general, and

may be expressed as
k = kR k̂ R + ikI k̂ I , (4.2)
wherein the scalars kR,I ∈ R and the unit vectors k̂ R,I ∈ R3 . Parenthetically, note
that although k varies with ω, for convenience we do not express this dependence
explicitly.
In light of eq. (4.2), the field phasor (4.1) may be written as

F(r, ω) = F 0 (ω) exp −kI k̂ I • r exp i kR k̂ R • r − ωt . (4.3)
On planes in R2 specified by
k̂ R • r = constant, (4.4)
we see from eq. (4.3) that F(r, ω) has constant phase. In other words, a propagat-
ing plane of constant phase is described by the field phasor (4.1). The planes of
constant phase propagate in the direction of k̂ R with velocity

ω
vp = k̂ . (4.5)
kR R
This velocity is called the phase velocity, and its magnitude is the phase speed.
Plane waves may be classified as either uniform or nonuniform as follows:
• If k̂ R = ±k̂ I then the wavevector k may be stated as
k = k k̂, (4.6)
with the complex-valued wavenumber k = kR ± ikI and the real-valued unit

vector k̂ ≡ k̂ R . Thus, on planes of constant k̂ • r, F(r, ω) has both constant
phase and constant amplitude. Accordingly, these plane waves are known as
uniform plane waves. Uniform plane waves also arise in the nondissipative sce-
nario characterized by kI = 0.
• If k̂ R = ±k̂ I and kI = 0, then the amplitude of F(r, ω) is generally not uni-
form on planes of constant k̂ R • r. Accordingly, these plane waves are known
as nonuniform plane waves.
The wavevector k has inherently complex-valued components for non-uniform
plane waves, whereas for uniform plane waves k may have either complex- or
real-valued components. Uniform plane waves with real-valued k propagate with-
out attenuation. For practical applications, uniform plane waves are encountered
much more often than nonuniform plane waves. For this reason, we concentrate
on uniform plane waves in the remainder of this section.
4.2. Eigenanalysis
The starting point for our analysis is provided by the frequency-domain Maxwell
curl postulates in the absence of sources; i.e.,

∇ × H (r, ω) + iωD(r, ω) = 0
. (4.7)
∇ × E(r, ω) − iωB(r, ω) = 0
Let us consider the most general linear homogeneous medium – namely, a ho-
mogeneous bianisotropic medium, whose Tellegen constitutive relations are ex-
pressed as

D(r, ω) = EH (ω) • E(r, ω) + ξ (ω) • H (r, ω)
EH
. (4.8)
B(r, ω) = ζ (ω) • E(r, ω) + μ (ω) • H (r, ω)
EH EH
The combination of (4.7) and (4.8) provides us with

L(∇) + iωK EH (ω) • F(r, ω) = 0, (4.9)
wherein the 6-vector/6 × 6 dyadic notation of § 2.4 has been implemented, with
the constitutive dyadic

EH (ω) ξ (ω)
K EH (ω) = EH . (4.10)
ζ (ω) μ (ω)
EH EH
On restricting our attention to plane-wave solutions, as represented by the field
phasor (4.1), the differential dyadic equation (4.9) simplifies to

L(ik) + iωK EH (ω) • F 0 (ω) = 0. (4.11)
The algebraic dyadic equation (4.11) is generally amenable to eigenanalysis.
The existence of nonzero solutions to (4.11) immediately leads us to the dis-
persion relation

det L(ik) + iωK EH (ω) = 0. (4.12)
For uniform plane waves, i.e., k = k k̂ with k ∈ C and k̂ ∈ R3 , the dispersion

relation (4.12) yields a polynomial of the fourth degree in k. Accordingly, it has
four solutions k = k , ∈ [1, 4].
The dispersion relation (4.12) is quadratic in k 2 for all Lorentz-reciprocal medi-
ums and all Lorentz-nonreciprocal mediums with null-valued magnetoelectric
constitutive dyadics (such as gyrotropic mediums). Hence, for these mediums two
solutions, k = k1 and k = k2 , say, hold for propagation co-parallel with k and two
solutions, k = k3 and k = k4 , say, hold for propagation anti-parallel to k, wherein
k1 = −k3 and k2 = −k4 . If k1 = k2 (or, equivalently, k3 = k4 ), the medium is
supposed to possess birefringence. In contrast, the term unirefringence is used to
describe those instances in which k1 = k2 (or, equivalently, k3 = k4 ).
For bianisotropic Lorentz-nonreciprocal mediums (such as Faraday chiral
mediums), four distinct roots – that do not come in positive/negative pairs – gen-
erally emerge from the dispersion relation (4.12).
Equation (4.11) yields

F 0 (ω) = adj L(ik) + iωK EH (ω) • N, (4.13)
where N is an arbitrary 6-vector. Without the boundary/initial conditions being
specified, only the relative orientations of E(r, ω) and H (r, ω) may be deduced
in general. It follows from eq. (4.1) that

E(r, ω) = E 0 (ω) exp i(k • r − ωt)
, (4.14)
H (r, ω) = H 0 (ω) exp i(k • r − ωt)
wherein E 0 (ω), H 0 (ω) ∈ C3 . Furthermore, it is helpful to decompose these pha-

sors as

E 0 (ω) = E 0R (ω) + iE 0I (ω)
, (4.15)
H 0 (ω) = H 0R (ω) + iH 0I (ω)
with E 0R,0I (ω), H 0R,0I (ω) ∈ R3 . The relative orientations and magnitudes of
E 0R (ω) and E 0I (ω) determine the polarization state of the electric field:
• if E 0R (ω) × E 0I (ω) = 0 then the electric field is linearly polarized;
• if E 0R (ω) • E 0I (ω) = 0 and |E 0R (ω)| = |E 0I (ω)| then the electric field is
circularly polarized;
• otherwise the electric field is elliptically polarized.
The polarization state of the magnetic field is similarly determined by the relative
orientations and magnitudes of H 0R (ω) and H 0I (ω). In optics, the ‘polarization
state’ of a plane wave is that of its electric field.
Although the derivation of the k roots of eq. (4.12) and the associated eigenvec-
tors F 0 (ω) is mathematically straightforward, the process can involve unwieldy
expressions, especially for bianisotropic mediums. Symbolic manipulations pack-
ages – such as Mathematica™ and Maple™ – can be usefully employed in the
analysis.
4.3. Isotropic scenarios
4.3.1. Dielectric–magnetic mediums

Isotropic dielectric–magnetic mediums, as characterized by the Tellegen constitu-
tive relations (3.3), are unirefringent. The corresponding dispersion relation (4.12)
yields two roots:
1/2
k1 = k2 = ω (ω)μ(ω)
1/2 . (4.16)
k3 = k4 = −ω (ω)μ(ω)
As regards the orientations of E(r, ω) and H (r, ω), a straightforward manipula-
tion of the Maxwell curl postulates (4.7) delivers the orthogonality relations
k • E(r, ω) = k • H (r, ω) = 0. (4.17)
Plane waves in isotropic dielectric–magnetic mediums can be arbitrarily polar-
ized: linearly, circularly, or elliptically. The time-averaged Poynting vector, de-
fined in eq. (2.65), simplifies to
exp(−2kI k̂ • r) 2
S(r, ω) t = E 0 (ω) kR k̂ (4.18)
2ωμ(ω)
for isotropic dielectric–magnetic mediums, and therefore is always aligned with

the wavevector k.
4.3.2. Isotropic chiral mediums

Isotropic chiral mediums, characterized by the Tellegen constitutive relations (3.5),
are birefringent. The wavenumbers which emerge from the dispersion relation
(4.12) are
1/2

k1 = −k3 = ω (ω)μ(ω) − iχ(ω)
1/2
. (4.19)
k2 = −k4 = ω (ω)μ(ω) + iχ(ω)
To determine the associated fields, it is mathematically convenient to introduce
the Beltrami fields (Lakhtakia [1994])
& '⎫
1 μ(ω) 1/2 ⎪
Q 1 (r, ω) = E(r, ω) + i H (r, ω) ⎪
⎪
⎬
2 (ω)
& 1/2 '⎪ . (4.20)
1 (ω) ⎪
⎪
Q 2 (r, ω) = H (r, ω) + i E(r, ω) ⎭
2 μ(ω)
Thereby, a simplified representation of the Maxwell curl postulates (4.7) is deliv-
ered in the form of the two uncoupled first-order differential equations

∇ × Q 1 (r, ω) − k1 Q 1 (r, ω) = 0
. (4.21)
∇ × Q 2 (r, ω) + k2 Q 2 (r, ω) = 0
Thus, we find that plane waves in an isotropic chiral medium must be circularly
polarized. The left circular polarization state is represented by Q 1 (r, ω), and the
right circular polarization state by Q 2 (r, ω). Furthermore, since k1 = k2 , the left
and right circularly polarized plane waves propagate with different phase veloc-
ities and with different attenuation rates. Accordingly, isotropic chiral mediums
are said to be circularly birefringent.
A Beltrami description is also possible for plane waves in homogeneous bi-
isotropic mediums characterized by eqs. (3.6), as delineated by Chambers [1956].
4.4. Anisotropic scenarios
4.4.1. Uniaxial mediums

We commence our survey of uniform plane-wave propagation in homogeneous
anisotropic mediums with the uniaxial dielectric medium, characterized by the
Tellegen constitutive relations (3.7) with permittivity dyadic (3.8). The distin-
guished axis û in eq. (3.8) is parallel to the sole crystallographic axis of the
medium.
The wavenumbers are extracted from the dispersion relation (4.12) as
1/2 ⎫
k1 = −k3 = ω (ω)μ0 ⎪
⎪
⎬
1/2
(ω)u (ω)μ0 . (4.22)
k2 = −k4 = ω ⎪
⎪
k̂ • uni (ω) • k̂ ⎭
Two distinct modes of plane-wave propagation are supported: ordinary as spec-

ified by the wavenumber k1 = −k3 , and extraordinary as specified by the
wavenumber k2 = −k4 . The phase speed of an extraordinary plane wave de-
pends on the direction of propagation, whereas all ordinary plane waves have the
same phase speed.
For the special propagation direction specified by k̂ = û, we see from
eqs. (4.22) that k1 = k2 (and, equivalently, k3 = k4 ). Therefore, parallel to ±û
both plane waves propagate with only one phase speed. The unique direction spec-
ified by û is called the optic axis. The optic axis for any uniaxial dielectric medium
coincides with the crystallographic axis.
The relative orientations of k, E(r, ω) and H (r, ω) may be determined from
eq. (4.13). For both ordinary and extraordinary plane waves, H (r, ω) • k = 0,
D(r, ω) • k = 0 and B(r, ω) • k = 0. Whereas E(r, ω) • k = 0 for ordinary plane
waves, E(r, ω) • k = 0 for extraordinary plane waves in general.
The time-averaged Poynting vector for the uniaxial dielectric medium with per-
mittivity dyadic (3.8) is given by (Chen [1983])
exp(−2kI k̂ • r) 2
S(r, ω) t = Re E 0 (ω) k ∗ − E 0 (ω) • k ∗ E ∗0 (ω) .

2ωμ0
(4.23)
Therefore, for ordinary plane waves the time-averaged Poynting vector is directed
along k̂, whereas it lies in the plane formed by k̂ and Re{[E 0 (ω) • k ∗ ]E ∗0 (ω)} for
extraordinary plane waves.
Mathematically, plane-wave propagation in uniaxial magnetic mediums, as de-
scribed by the Tellegen constitutive relations (3.10) with permeability dyadic
(3.11), is isomorphic to plane-wave propagation in uniaxial dielectric medi-
ums.
Significantly different plane-wave properties are exhibited by uniaxial dielect-
ric–magnetic mediums, as specified by the Tellegen constitutive relations (3.13)
with permittivity and permeability dyadics (3.8) and (3.11), respectively. The dis-
persion relation (4.12) yields the four solutions

1/2 ⎫
(ω)u (ω)μ(ω) ⎪
⎪
k1 = −k3 = ω ⎪
⎪
k̂ • uni (ω) • k̂ ⎬
1/2 . (4.24)
(ω)μ(ω)μu (ω) ⎪
⎪
k2 = −k4 = ω ⎪
⎪
• •
⎭
k̂ μ (ω) k̂
uni
All wavenumbers are dependent on the direction of propagation – i.e., unlike uni-
axial dielectric and uniaxial magnetic mediums, there is no ‘ordinary’ plane-wave
mode which propagates with the same phase speed in all directions. When k̂ = û,
we see from eqs. (4.24) that k1 = k2 (and, equivalently, k3 = k4 ). Thus, as is true
for uniaxial dielectric and uniaxial magnetic mediums, incidental unirefringence
occurs for propagation parallel to the crystallographic axis û. In addition, we ob-
serve that pathological unirefringence, characterized by k1 = k2 (and, equiva-
lently, k3 = k4 ) for all k̂, arises in the special case, identified by (Lakhtakia,
Varadan and Varadan [1991])
(ω) μ(ω)
= . (4.25)
u (ω) μu (ω)
4.4.2. Biaxial mediums
Next, let us consider a biaxial dielectric medium characterized by the Tellegen


D(r, ω) = bi (ω) • E(r, ω)
, (4.26)
B(r, ω) = μ0 H (r, ω)
where the symmetric permittivity dyadic bi (ω) may be of the orthorhombic,

monoclinic or triclinic type, as listed in table 2. The corresponding dispersion
relation (4.12) delivers the biquadratic polynomial (Michel [1997])
abi k 4 + bbi k 2 + cbi = 0, (4.27)
with
⎫
abi = k̂ • bi (ω) • k̂ ⎪
⎪

⎬
bbi = ω μ0 k̂ • bi (ω) • bi (ω) • k̂ − k̂ • bi (ω) • k̂ tr bi (ω) . (4.28)
2
⎪
⎪
c = ω4 μ2 det (ω) ⎭
bi 0 bi
Two, generally independent, wavenumbers emerge from eq. (4.27):

2 − 4a c )1/2 1/2
⎫
−bbi + (bbi bi bi ⎪
⎪
k1 = −k3 = ⎪
⎬
2abi

2 − 4a c )1/2 1/2 ⎪
. (4.29)
−bbi − (bbi bi bi ⎪
⎪
k2 = −k4 = ⎭
2abi
For further understanding of plane-wave propagation in biaxial dielectric medi-
ums, it is illuminating to confine ourselves to the realm of crystal optics, wherein
the components of the permittivity dyadic bi (ω) are real-valued. By manipulat-
ing eq. (4.11) and its complex conjugate, we find that (Chen [1983])

k4
I− (ω) k̂ k̂ • E 0 (ω) × E ∗0 (ω) = 0
• (4.30)
det bi (ω) bi
and

k 4 k̂ • bi (ω) • k̂
I− k̂ k̂ • H 0 (ω) × H ∗0 (ω) = 0. (4.31)
det bi (ω)
Hence, E 0 (ω) × E ∗0 (ω) = 0 and H 0 (ω) × H ∗0 (ω) = 0 unless the determinants
of the coefficient dyadics on the left sides of eqs. (4.30) and (4.31), respectively,
are null-valued. The following two central attributes of plane-wave propagation
in nondissipative biaxial dielectric mediums may thus be deduced:
• Plane waves associated with the distinct wavenumbers k1 and k2 (or, equiv-
alently, k3 and k4 ) are linearly polarized with mutually orthogonal planes of
polarization.
• For two directions of propagation, specified implicitly by
2
bbi − 4abi cbi = 0, (4.32)
the wavenumbers k1 and k2 (and, equivalently, k3 and k4 ) coincide. Along these
two directions – known as the optic axes – plane waves can have any polariza-
tion state. Unlike uniaxial dielectric mediums and uniaxial magnetic mediums,
the optic axes of a biaxial medium do not coincide with its crystallographic
axes.
The time-averaged Poynting vector (Chen [1983])
ω(k • E 0 )2
S(r, ω) t = k • bi (ω) • k k + ω2 μ0 bi (ω)
2 det[ω μ0 bi (ω) − k I ]
2 2

• (ω) − tr (ω) I + k 2 (ω) • k (4.33)
bi bi bi
is not generally aligned with the wavevector k.
The mathematical description of plane-wave propagation in biaxial magnetic

mediums, as described by the Tellegen constitutive relations (3.20) with perme-
ability dyadic being of the orthorhombic, monoclinic or triclinic type, as specified
in eqs. (3.21), is equivalent to that for biaxial dielectric mediums.
The Tellegen constitutive relations (3.22) specify a biaxial dielectric–magnetic
medium. The increased parameter space associated with such mediums, as com-
pared with biaxial dielectric mediums or biaxial magnetic mediums, gives rise
to plane-wave propagation with a richer palette of attributes. For example, under
certain pathological conditions, biaxial dielectric–magnetic mediums may exhibit
only one optic axis (Shen, Win, Chen, Fan, Ding, Wang, Tian and Ming [2005]).
4.4.3. Gyrotropic mediums

Turning now to an example of a Lorentz-nonreciprocal anisotropic medium, we
focus on the gyrotropic dielectric medium characterized by the Tellegen consti-
tutive relations (3.30) wherein the components of the antisymmetric permittivity
dyadic
gyro (ω) = (ω)(I − ûû) + ig (ω)û × I + u (ω)ûû (4.34)
are complex-valued, in general. Thus, the dispersion relation (4.12) simplifies to
the biquadratic polynomial (Chen [1983])
agyro k 4 + bgyro k 2 + cgyro = 0, (4.35)
with
⎫
agyro = k̂ • gyro (ω) • k̂ ⎪
⎪
⎪
⎬

bgyro = ω μ0 k̂ • adj gyro (ω) − tr adj gyro (ω) I • k̂ .
2
(4.36)
⎪
⎪
⎪
⎭
cgyro = ω4 μ20 det gyro (ω)
Two, generally independent, wavenumbers emerge therefrom:
⎫
−bgyro + (bgyro
2 − 4agyro cgyro )1/2 1/2 ⎪
k1 = −k3 = ⎪
⎪
2agyro ⎬
. (4.37)
−bgyro − (bgyro
2 − 4agyro cgyro )1/2 1/2 ⎪
⎪
⎪
k2 = −k4 = ⎭
2agyro
The following two special scenarios are worthy of particular mention:
• For propagation either co-parallel or antiparallel to the biasing magnetic field
(cf. § 3.2.3), k̂ • û = ±1; hence,
1/2
k1 = −k3 = ω2 μ0 (ω) − g (ω)
1/2 . (4.38)
k2 = −k4 = ω2 μ0 (ω) + g (ω)
When dissipation is negligibly small, we may set , g , u ∈ R; and the upper

and lower wavenumbers specified by eqs. (4.38) then correspond to left and
right circularly polarized plane waves, respectively. The difference in phase
speeds associated with the wavenumbers k1 and k2 (or, equivalently, k3 and k4 )
results in a rotation of the plane of polarization known as Faraday rotation.
• For propagation perpendicular to the biasing magnetic field, the wavenumbers
2 ⎫
(ω) − g2 (ω) 1/2 ⎪
⎬
k1 = −k3 = ω μ0 2
(ω) (4.39)
1/2 ⎪
⎭
k2 = −k4 = ω2 μ0 u (μ)
emerge by using k̂ • û = 0 in eqs. (4.37). Whereas k1 and k3 depend upon the
strength of the biasing magnetic field, k2 and k4 do not. On ignoring dissipation
(i.e., setting , g , u ∈ R), it transpires that the upper and lower wavenumbers
specified by eqs. (4.39) correspond to elliptically and linearly polarized waves,
respectively.
The mathematical description of plane-wave propagation for gyrotropic mag-
netic mediums is equivalent to that for gyrotropic dielectric mediums.
4.4.4. Voigt waves

As described in §§ 4.4.1–4.4.3, two distinct plane waves generally propagate in
each direction in anisotropic mediums. We close our consideration of plane-wave
propagation in anisotropic mediums with an anomalous example wherein the two
plane waves coalesce to form a Voigt wave. Experimental observations of these
waves were reported by Woldemar Voigt [1902] more than 100 years ago. While
the theoretical basis for Voigt waves was developed in the 1950s (Pancharatnam
[1958]), this topic has attracted renewed attention lately as a consequence of ad-
vances in metamaterial technologies. Recent work has demonstrated the possibil-
ity of constructing complex composite materials which support the propagation of
Voigt waves, using component materials which do not themselves support Voigt-
wave propagation (Mackay and Lakhtakia [2003]).
Let us take a general anisotropic dielectric medium, specified by the Tellegen

D(r, ω) = aniso (ω) • E(r, ω)
, (4.40)
B(r, ω) = μ0 H (r, ω)
with complex-valued permittivity dyadic
⎛ ⎞
11 (ω) 12 (ω) 13 (ω)
aniso (ω) = ⎝ 21 (ω) 22 (ω) 23 (ω) ⎠ , (4.41)
31 (ω) 32 (ω) 33 (ω)
as the setting for our analysis. For definiteness – and without loss of generality –
we examine uniform plane-wave propagation along the direction of the z axis;
i.e., k̂ = ẑ. Thus, focussing on the electric field phasor E(r, ω), we consider
plane waves of the form
⎡ ⎤ ⎡ ⎤
Ex (z, ω) E0x (ω)
E(r, ω) ≡ ⎣ Ey (z, ω) ⎦ = ⎣ E0y (ω) ⎦ exp i(kz − ωt) . (4.42)
Ez (z, ω) E0z (ω)
Combining eqs. (4.40) and (4.11), we eliminate E0z (ω) to obtain

ω2 μ0 δ11 δ12 E0x (ω) 2 E0x (ω)
• =k , (4.43)
33 (ω) δ12 δ22 E0y (ω) E0y (ω)
where
⎫
δ11 = 11 (ω)33 (ω) − 13 (ω)31 (ω)⎪
⎪
⎪
⎪
δ = (ω) (ω) − (ω) (ω)⎬
12 12 33 13 32
. (4.44)
δ21 = 21 (ω)33 (ω) − 23 (ω)31 (ω)⎪
⎪
⎪
⎪
⎭
δ22 = 22 (ω)33 (ω) − 23 (ω)32 (ω)
The matrix on the left side of eq. (4.43) has only one eigenvalue but two linearly
independent eigenvectors for isotropic dielectric mediums. The normal situation
for anisotropic mediums is that of birefringence, wherein the matrix in eq. (4.43)
has two distinct eigenvalues and two independent eigenvectors. The general solu-
tion for the x and y components of E(r, ω) may then be expressed as
&
Ex (z, ω) Ex1 (ω)
= C1 exp(ik1 z)
Ey (z, ω) Ey1 (ω)
'
Ex2 (ω)
+ C2 exp(ik2 z) exp(−iωt). (4.45)
Ey2 (ω)
Herein, the eigenvectors [Ex1 , (ω), Ey1 (ω)]T and [Ex2 , (ω), Ey2 (ω)]T corre-
spond to the two wavenumbers
⎫
μ0 1/2
1/2 ⎪
⎪
k1 = ω (δ11 + δ22 ) + (δ11 − δ22 ) + 4δ12 δ21
2 ⎪
⎬
233 (ω)
1/2 ⎪ ,
μ0 1/2
⎪
⎪
k2 = ω (δ11 + δ22 ) − (δ11 − δ22 )2 + 4δ12 δ21 ⎭
233 (ω)
(4.46)
respectively, while C1,2 are amplitude coefficients which may be determined from
boundary/initial conditions.
Anomalously, the matrix in eq. (4.43) can have only one independent eigen-
vector – and therefore only one eigenvalue k. Accordingly, the matrix cannot be
a scalar matrix. In this case, the general solution for the x and y components of
E(r, ω) may be expressed as

Ex (z, ω) Ex1 (ω) Ex2 (ω)
= C1 + ikzC2 exp i(kz − ωt) ,
Ey (z, ω) Ey1 (ω) Ey2 (ω)
(4.47)
which represents a Voigt wave. A prominent distinguishing feature of the solu-
tion (4.47) is that the plane-wave amplitude has a linear dependence on propaga-
tion distance.
Sufficient conditions for Voigt-wave propagation are (Gerardin and Lakhtakia
[2001])
• (δ11 − δ22 )2 + 4δ12 δ21 = 0, and
• |δ12 | + |δ12 | = 0.
These conditions cannot be satisfied by uniaxial dielectric mediums, but can be
satisfied by certain non-orthorhombic biaxial dielectric mediums and gyrotropic
mediums (Agranovich and Ginzburg [1984]). Furthermore, in connection with
the remaining subsections in this section, we note that Voigt-wave propagation
has also been investigated in certain bianisotropic (Berry [2005]) and periodically
nonhomogeneous mediums (Lakhtakia [1998a]).
4.5. Bianisotropic scenarios
As the parameter space associated with bianisotropic mediums is much larger

than that associated with anisotropic mediums, an extremely rich and diverse ar-
ray of plane-wave propagation characteristics can be exhibited by bianisotropic
mediums. We consider two examples of plane-wave propagation in bianisotropic
scenarios. The first is plane-wave propagation in a bianisotropic medium which is
isotropic dielectric–magnetic when viewed by an observer in a co-moving refer-
ence frame. As indicated in § 4.5.1, this bianisotropic medium is in fact unirefrin-
gent. The second is the Faraday chiral medium, which can support plane waves
with four distinct wavenumbers for propagation along an arbitrary axis. Both ex-
amples are of Lorentz-nonreciprocal mediums.
The dispersion relation (4.12) reduces to a polynomial that is quadratic in k 2
for all Lorentz-reciprocal mediums, whether bianisotropic or not. Consequently,
Lorentz-reciprocal bianisotropic mediums support plane waves with at most two
independent wavenumbers for propagation along any axis.
4.5.1. Mediums moving at constant velocity

Suppose an isotropic dielectric–magnetic medium is characterized by the Tellegen

D (r , ω ) = Σ

(ω )E (r , ω )

(4.48)
B (r , ω ) = μΣ (ω )H (r , ω )
in an inertial reference frame Σ , which moves at a constant velocity v = v v̂
relative to the inertial reference frame Σ. A plane wave, described by the phasor

F (r , ω ) = F0 (ω ) exp i(k • r − ω t ) (4.49)
with respect to the reference frame Σ , is described by eq. (4.1) with respect
to the reference frame Σ. Herein {t, r} and {t , r } are related via the Lorentz
transformations (2.32), and (Chen [1983])

ωv
k = γ k • v̂ − 2 v̂ + (I − v̂ v̂) • k, (4.50)
c0
ω = γ (ω − k • v). (4.51)
Since the medium is unirefringent with respect to Σ ,it is clear from eq. (4.50)
that the medium is also unirefringent with respect to Σ. Furthermore, even though

the plane wave may be taken to be uniform with respect to Σ (i.e., k = k k̂ with

k̂ ∈ R3 ), we see from eq. (4.50) that the plane wave is generally nonuniform with
respect to Σ.
4.5.2. Faraday chiral mediums

The concept of a Faraday chiral medium (FCM) was presented in § 3.3.3. In the
Tellegen representation, the frequency-domain constitutive relations of a FCM are
given by eqs. (3.39) with the 3 × 3 constitutive dyadics (3.40).
The corresponding dispersion relation (4.12) yields a quartic polynomial in k
with four distinct roots. These may be extracted by standard algebraic or numeri-
cal methods (Abramowitz and Stegun [1965]).
The mathematical description of plane-wave propagation in FCMs simplifies
considerably in the particular case that k̂ = û; i.e., propagation along the direction
of the biasing magnetic field. The four wavenumbers
1/2 1/2
⎫
k1 = ω (ω) + g (ω) μ(ω) + μg (ω) − ξ(ω) − ξg (μ) ⎪ ⎪
1/2 1/2
⎪⎪
⎪
k2 = ω − (ω) + g (ω) μ(ω) + μg (ω) − ξ(ω) − ξg (μ) ⎬
1/2 1/2
⎪,
k3 = ω (ω) − g (ω) μ(ω) − μg (ω) + ξ(ω) − ξg (μ) ⎪ ⎪
1/2 1/2
⎪⎪
⎭
k4 = ω − (ω) − g (ω) μ(ω) − μg (ω) + ξ(ω) − ξg (μ)
(4.52)
then emerge from the dispersion relation (4.12). Lorentz-nonreciprocity is evident

from the fact that k1 + k3 = 0 and k2 + k4 = 0. The orientations of the electro-
magnetic field phasors are determined by analyzing eq. (4.13): thereby, it is found
that there is no component of the electric field in the direction of û, whereas the
time-averaged Poynting vector is aligned parallel to û (Mackay and Lakhtakia
[2004]).
4.6. Nonhomogeneous mediums
Let us now discuss propagation in two types of nonhomogeneous mediums. The

first is the general class of helicoidal bianisotropic mediums (HBMs). These rep-
resent a wide range of technologically important materials such as CLCs and
CSTFs. The second is gravitationally affected vacuum, which is of particular in-
terest in view of the recent discovery that certain space–time metrics can support
propagation with negative phase velocity – a topic that will be discussed further
in § 4.7.
The mediums discussed in §§ 4.6.1 and 4.6.2 are atypical of nonhomogeneous
mediums insofar as analytical (or semi-analytical) methods may be fruitfully de-
ployed to explore the characteristics of plane-wave propagation therein. Gener-
ally, wave-propagation studies for nonhomogeneous mediums pose formidable
challenges to theoreticians, particularly if closed-form solutions are sought.
4.6.1. Periodic nonhomogeneity
Let us recall from § 3.4.1 that the constitutive relations for a HBM with its heli-
coidal axis aligned with the Cartesian z axis may stated in the Tellegen represen-
tation as eqs. (3.48), with constitutive dyadics (3.49) and (3.50). On combining
the constitutive relations (3.48) and the Maxwell curl postulates (4.7) we find

∇ × E(r, ω) = iω ζ (z, ω) • E(r, ω) + μ (z, ω) • H (r, ω)
HBM HBM
.
∇ × H (r, ω) = −iω HBM (z, ω) • E(r, ω) + ξ (z, ω) • H (r, ω)
HBM
(4.53)
As a step towards eliminating the z-dependence on the right side of (4.53) for
propagation along the helicoidal axis, the Oseen transformation (Oseen [1933])

E (r, ω) = S −1
z
(z) • E(r, ω)
, (4.54)
H (r, ω) = S −1
z
(z) • H (r, ω)
is implemented. Thereby, eqs. (4.53) are recast as

⎫
S −1 (z) • (∇ × I ) • S z (z) • E (r, ω)
z ⎪
⎪
⎪
= iω ζ (z, ω) • E (r, ω) + μ (z, ω) • H (r, ω) ⎪⎬
HBM HBM
. (4.55)
−1
S z (z) • (∇ × I ) • S z (z) • H (r, ω) ⎪
⎪
⎪
⎪
⎭
= −iω HBM (z, ω) • E (r, ω) + ξ (z, ω) • H (r, ω)
HBM
Upon introducing the Fourier representations

E (r, ω) = ex (z, ω)x̂ + ey (z, ω)ŷ + ez (z, ω)ẑ exp(iκx)
, (4.56)
H (r, ω) = hx (z, ω)x̂ + hy (z, ω)ŷ + hz (z, ω)ẑ exp(iκx)
wherein κ may be interpreted as a wavenumber in the xy plane, eqs. (4.55) gen-

erally reduce to the 4 × 4 matrix differential equation
∂
F (z, ω) = M (z, ω)F (z, ω). (4.57)
∂z
Herein,
T
F (z, ω) = ex (z, ω), ey (z, ω), hx (z, ω), hy (z, ω) (4.58)
is a column 4-vector, and the 4 × 4 matrix function M (z, ω) – which is too cum-
bersome to reproduce explicitly here – may be conveniently expressed as

iπz iπz
M (z) = A + κ C exp + C1,−1 exp −
1,1 Ω Ω

i2πz i2πz
+ κ 2 C2,2 exp + C2,0 + C2,−2 exp − , (4.59)
Ω Ω
where the 4 × 4 matrixes A , C1,±1 , C2,0 , and C2,±2 are independent of z and κ
but not of Ω and ω.
For axial propagation, κ = 0 and the closed-form solution to eq. (4.57) arises
as (Lakhtakia and Weiglhofer [1995])
F (z, ω) = exp(iA z)F (0, ω). (4.60)
For nonaxial propagation, κ = 0 and the solution to eq. (4.57) may be ex-
pressed in terms of a power series in z through exploiting the representation (4.59)
(Lakhtakia and Weiglhofer [1997a]). Alternatively, a piecewise uniform approxi-
mation may be implemented (Lakhtakia and Messier [2005]).
4.6.2. Gravitationally affected vacuum

As described in § 3.4.2, the electromagnetic properties of vacuum in generally
curved space–time are equivalent to those of a fictitious, instantaneously respond-
ing, bianisotropic medium. The time-domain constitutive relations for the ficti-
tious medium (3.51) are expressed in terms of the 3×3 dyadic γ̃ (r, t) and 3-vector
Γ̃ (r, t), which are derived from the space–time metric and specified in eqs. (3.52)
and (3.53), respectively.
The nonhomogeneous nature of the equivalent medium represented by eq. (3.51)
may be effectively dealt with by implementing a piecewise uniform approxima-
tion (Lakhtakia, Mackay and Setiawan [2005]). Attention is focussed on a space–
time neighbourhood which is sufficiently small so that the non-uniform quantities
γ̃ (r, t) and Γ̃ (r, t) may be replaced by their uniform equivalents γ̄ and Γ̄ , respec-
tively. Thereby, the propagation of plane waves within the neighbourhood may be
analysed using the methods described in § 4.2. Global solutions may then be de-
veloped by stitching together the plane-wave solutions from adjacent space–time
neighbourhoods. This piecewise uniform approximation technique is widely used
in solving differential equations with nonhomogeneous coefficients (Hoffman
[1992]) and also for nonaxial propagation in CLCs and CSTFs (Lakhtakia and
Messier [2005]).
Let us now proceed to plane-wave solutions

Ẽ(r, t) = Re E 0 (ω) exp i(k • r − ωt)

, (4.61)
H̃ (r, t) = Re H 0 (ω) exp i(k • r − ωt)
within a neighbourhood which is sufficiently small that the non-uniform consti-
tutive relations (3.51) may be replaced by their uniform counterparts (Lakhtakia,
Mackay and Setiawan [2005])
⎫
D̃(r, t) = 0 γ̄ • Ẽ(r, t) − c0−1 Γ̄ × H̃ (r, t) ⎬
. (4.62)
B̃(r, t) = μ0 γ̄ • H̃ (r, t) + c0−1 Γ̄ × Ẽ(r, t)⎭
The phasor amplitudes E 0 (ω) and H 0 (ω) in eq. (4.61) are determined by the
boundary/initial conditions. Combining the source-free Maxwell curl postulates
⎫
∇ × H̃ (r, t) − D̃(r, t) = 0⎪
∂
⎪
⎬
∂t
, (4.63)
⎪
∇ × Ẽ(r, t) + B̃(r, t) = 0 ⎪
∂
⎭
∂t
with the constitutive relations (4.62), we obtain the eigenvector equation
& 2 '
ω
det(γ̄ ) − p • γ̄ • p I + pp • γ̄ · E 0 (ω) = 0, (4.64)
c0
where
ω
p=k− Γ¯ . (4.65)
c0
The dispersion relation
2 2
ω
p γ̄ p −
• • det(γ̄ ) = 0 (4.66)
c0
thereby arises, from which four roots emerge:
ω ⎫
k1 = −k3 = k̂ • γ̄ • Γ¯ + (k̂ • γ̄ • Γ¯ )2 ⎪
⎪
⎪
c0 k̂ • γ̄ • k̂ ⎪
⎪
1/2
⎪ ⎪
¯ ¯ ⎪
⎪
− k̂ γ̄ k̂(Γ γ̄ Γ − det γ̄ )
• • • • ⎬
. (4.67)
ω ⎪
k2 = −k4 = k̂ • γ̄ • Γ¯ − (k̂ • γ̄ • Γ¯ )2 ⎪
⎪
⎪
c0 k̂ • γ̄ • k̂ ⎪
⎪
1/2
⎪ ⎪
¯ ¯ ⎪
⎭
− k̂ γ̄ k̂(Γ γ̄ Γ − det γ̄ )
• • • •
Just as described in § 4.5.1, the wavenumbers k1 and k2 (or, equivalently,

k3 and k4 ) given by eqs. (4.67) are not independent, and the medium represented
by eqs. (3.51) is unirefringent accordingly.
The general solution to eq. (4.64) may be expressed as the sum
E 0 (ω) = Aa (ω)ê a (ω) + Ab (ω)ê b (ω), (4.68)
wherein the complex-valued amplitude scalars Aa,b (ω) are determined from
boundary/initial conditions, and the unit vectors êa,b (ω) are taken as
⎫
γ̄ −1 • ŵ ⎪
⎪
ê a (ω) = −1 ⎪
⎪
|γ̄ • ŵ| ⎪
⎬
. (4.69)
γ̄ −1 • [p × ê a (ω)] ⎪
⎪
⎪
ê b (ω) = −1 ⎪
⎪
|γ̄ • [p × ê a (ω)]| ⎭
The unit vector ŵ in eqs. (4.69) is orthogonal to p, i.e., ŵ • p = 0, but is otherwise

arbitrary. In a similar vein,
H 0 (ω) = Aa (ω)ĥ a (ω) + Ab (ω)ĥ b (ω), (4.70)
with the unit vectors

1 −1
ĥ a,b (ω) = γ̄ • p × ê a,b (ω) . (4.71)
ωμ0
From eqs. (4.68) and (4.70), we find that the corresponding time-averaged
Poynting vector
1
S(r, ω) t = Aa (ω)2 ê (ω) • γ̄ • ê (ω)
a a
2ωμ0 det γ̄
2
+ Ab (ω) ê b (ω) • γ̄ • ê b (ω) γ̄ • p (4.72)
is aligned with γ̄ • p.
Note that a local observer is cognizant only of a flat space–time, and will
therefore choose a space–time coordinate system such that γ̄ = I and Γ¯ = 0
(Lakhtakia, Mackay and Setiawan [2005]).
4.7. Plane waves with negative phase velocity
We bring our survey of plane-wave propagation to a close by considering the

relationship between the phase velocity v p , as defined in eq. (4.5), and the time-
averaged Poynting vector
S(r, ω)t , as defined in eq. (2.65). The phase velocity
may be classified as being (Mackay and Lakhtakia [in press-a])
• positive if v p •
S(r, ω)t > 0, and
• negative if v p •
S(r, ω)t < 0.
Conventional materials support plane-wave propagation with positive phase ve-
locity (PPV). However, due to recent progress in the development of novel materi-
als and metamaterials, attention is turning increasingly towards parameter regimes
giving rise to negative phase velocity (NPV). Furthermore, the concept of infinite
phase velocity – which arises at the boundary between NPV and PPV propagation
in isotropic dielectric–magnetic mediums – has come up (Lakhtakia and Mackay
[2004b]).
For isotropic dielectric–magnetic mediums, the phenomenon of NPV propaga-
tion is closely associated with negative refraction (Lakhtakia, McCall and Wei-
glhofer [2003], Ramakrishna [2005]). An explosion of interest in this topic took
place at the beginning of this century, following an experimental report of mi-
crowave negative refraction in an effectively homogeneous metamaterial which
comprised a periodic array of metallic wire and ring inclusions embossed on
circuit boards (Shelby, Smith and Schultz [2001]). Subsequent efforts by exper-
imentalists and theorists have been directed towards higher-frequency regimes,
with optical negative refraction being achieved most recently, albeit with consid-
erable associated dissipation (Dolling, Wegener, Soukoulis and Linden [2007],
Shalaev [2007]). The prospect of fabricating lenses with extremely high resolving
power from negatively refracting materials has fuelled much of this work (Pendry
[2004]).
3, § 5] Dyadic Green functions 175
A simple construction method for isotropic dielectric–magnetic metamaterials

which support NPV propagation has recently been proposed. The method is based
on the homogenization of a random assembly of two different types of spherical
particle. The two types of constituent particles, type a and type b, can each be
made of an isotropic, homogeneous, dielectric–magnetic material, with permittiv-
ities a and b , and permeabilities μa and μb , respectively. Provided that a,b and
μa,b lie within certain parameter ranges, with the real parts of a,b being negative-
valued and the real parts of μa,b being positive-valued (or vice versa), the bulk
constituent materials do not support NPV propagation whereas the correspond-
ing homogenized composite medium does. Whether or not NPV propagation is
supported by the homogenized composite medium depends upon the relative pro-
portions of the constituents, as well as the size (Lakhtakia and Mackay [2006])
and spatial distribution (Mackay and Lakhtakia [2006]) of the their particles.
As compared with isotropic dielectric–magnetic mediums, anisotropic and
bianisotropic mediums offer greater scope for achieving NPV propagation on ac-
count of their larger parameter spaces (Hu and Lin [2003], Mackay and Lakhtakia
[2004]). In particular, bianisotropic homogenized composite mediums which sup-
port NPV propagation may be conceptualized – as Faraday chiral mediums
(Mackay and Lakhtakia [2004]), for example – which arise from constituent medi-
ums which do not themselves support NPV propagation.
There are also noteworthy manifestations of NPV in astrophysical scenarios,
with potentially important consequences for observational astronomy. The prop-
erty of supporting NPV propagation is not Lorentz covariant (cf. § 2.5.3). Thus,
a medium may support PPV propagation from the perspective of one inertial ob-
server, but support NPV propagation from the perspective of another inertial ob-
server (Mackay and Lakhtakia [in press-a]). In addition, vacuum itself – which
does not support NPV propagation in any inertial reference frame in flat space–
time – can support NPV propagation for certain curved space–times, such as the
Kerr space–time (Mackay, Lakhtakia and Setiawan [2005a]) and Schwarzschild–
anti-de Sitter space–time (Mackay, Lakhtakia and Setiawan [2005b]). This dis-
tinction essentially arises from the differences between covariant and noncovari-
ant formalisms (Mackay and Lakhtakia [in press-b], McCall [2007]).
§ 5. Dyadic Green functions
What is the (frequency-domain) electromagnetic field generated by a given dis-

tribution of sources immersed a specific medium? Due to the linearity of the
Maxwell postulates, this fundamental question may be tackled by means of dyadic
Green functions (DGFs) (Tai [1994], Weiglhofer [1995]). The explicit delineation
of DGFs for complex mediums poses formidable challenges to theoreticians.

Closed-form representations of DGFs are available for isotropic mediums and
for some classes of relatively simple anisotropic and bianisotropic medium, but
not for general anisotropic and bianisotropic mediums. However, integral repre-
sentations of DGFs can always be found. For certain applications – such as in
the determination of the scattering response of electrically small particles, uti-
lized in homogenization studies – approximative solutions may be constructed
using only the singular part of the DGF, which gives rise to the depolarization
dyadic. In this section, we survey DGFs and depolarization dyadics for unbounded
anisotropic/bianisotropic mediums. The mediums described are homogeneous,
with the exception of HBMs in § 5.3.2.
5.1. Definition and properties
Let Q(r, ω) defined in eq. (2.19) be the source 6-vector immersed in the homo-
geneous bianisotropic medium characterized by the Tellegen 6 × 6 constitutive
dyadic K EH (ω). The relationship between Q(r, ω) and F(r, ω) is dictated by the
frequency-domain Maxwell curl postulates

L(∇) + iωK EH (ω) • F(r, ω) = Q(r, ω). (5.1)
Since (5.1) is a linear differential equation, its solution may be expressed in terms
of a 6 × 6 DGF G(r − r , ω). Thus, the field at position r is given as

F(r, ω) = Fcf (r, ω) + G(r − r , ω) • Q(r , ω) d3 r , (5.2)
V
wherein the source points r are confined to the region of integration V . The
6-vector Fcf (r, ω) denotes the corresponding complementary function as per

L(∇) + iωK EH (ω) • Fcf (r, ω) = 0. (5.3)
By construction, the DGF is the solution of the differential equation

L(∇) + iωK EH (ω) • G(r − r , ω) = δ(r − r )I, (5.4)
where
∞
1
δ(r) = exp(iq • r) d3 q (5.5)
2π 3
−∞
is the Dirac delta function. Hence, the DGF may be viewed informally as repre-
senting the ‘response’ of the medium to a point ‘source’. Furthermore, the DGF
is required to satisfy the Sommerfeld radiation condition (Felsen and Marcuvitz

[1994]).
The 6 × 6 DGF G(r − r , ω) may be expressed in terms of its four component
3 × 3 DGFs as
ee
G (r − r , ω) Gem (r − r , ω)
G(r − r , ω) = . (5.6)
Gme (r − r , ω) Gmm (r − r , ω)
These 3 × 3 DGFs satisfy the differential equations (Weiglhofer [1995])
⎫
L e (∇, ω) • Gee (r − r , ω) = iωI δ(r − r )
⎪
⎪
⎪
⎪
⎪
⎪
L e (∇, ω) • G (r − r , ω)
em
⎪
⎪
−1 ⎪
= − ∇ × I + iωξ (ω) μ (ω)δ(r − r )⎪
• ⎬
EH EH
, (5.7)
L m (∇, ω) G (r − r , ω)
• me ⎪
⎪
⎪
⎪
= ∇ × I − iωζ (ω) • −1 (ω)δ(r − r ) ⎪
⎪
⎪
⎪
EH
EH ⎪
⎪
⎭
L m (∇, ω) • Gmm (r − r , ω) = iωI δ(r − r )
wherein the linear differential operators

⎫
L e (∇, ω) = ∇ × I + iωξ (ω) • μ−1 (ω) • ∇ × I − iωζ
(ω) ⎪
⎪
EH EH ⎪
⎪ EH
− ω2 EH (ω) ⎬
−1 .
L m (∇, ω) = ∇ × I − iωζ (ω) • EH (ω) • ∇ × I + iωξ (ω) ⎪ ⎪
⎪
EH EH ⎪
⎭
− ω2 μ (ω)
EH
(5.8)
Thus, if Gee (r − r , ω) is known, then Gme (r − r , ω) may be calculated directly
as
1 −1
Gme (r − r , ω) = μ (ω) • ∇ × I − iωζ (ω) • Gee (r − r , ω),
iω EH EH
(5.9)
in lieu of solving the differential equation for Gme (r−r , ω) in eq. (5.7). Similarly,
Gem (r − r , ω) may be determined as
1 −1
Gem (r − r , ω) = − EH (ω) • ∇ × I + iωξ (ω) • Gmm (r − r , ω),
iω EH
(5.10)
provided that Gmm (r − r , ω) is known. Also, once one of the four 3 × 3 DGFs in
eq. (5.6) known, the others can often be deduced from symmetry considerations.
5.2. Closed-form representations
A variety of ingenious methods, commonly based on dyadic operator manipula-

tions and/or field transformations, have been devised in the pursuit of closed-form
representations for DGFs (Weiglhofer [1993], Olyslager and Lindell [2002]). Of-
ten such methods have been developed on an ad hoc basis. In contrast, the fol-
lowing dyadic scalarization and factorization technique is generally applicable,
at least in principle: Formally, the 6 × 6 DGF may be expressed as (Weiglhofer
[1993])
†
G(r − r , ω) = L(∇) + iωK(ω) σ (r − r , ω). (5.11)
Herein, the adjoint operation indicated by the superscript † is defined implicitly

via
†
L(∇) + iωK(ω) • L(∇) + iωK(ω) = H(∇, r, ω)I, (5.12)
whereas the scalar Green function σ (r − r , ω) is the solution of
H(∇, r, ω)σ (r − r , ω) = δ(r − r ), (5.13)

with H(∇, r, ω) being a scalar fourth-order differential operator. Straightforward,
albeit lengthy, matrix-algebraic operations result in the construction of the adjoint
operator [L(∇) + iωK(ω)]† for anisotropic and bianisotropic mediums. However,
solutions to the fourth-order partial differential equation represented by eq. (5.13)
are generally elusive.
The existence of closed-form expressions for DGFs is closely linked to the
factorization properties of H(∇, r, ω). In many instances where H(∇, r, ω) is
expressible as a product of two second-order differential operators, closed-form
representations for DGFs have been established (Weiglhofer [2000], Olyslager
and Lindell [2002]).
A different approach for the determination of DGFs is provided by spatial
Fourier transformations. Upon transforming the spatial coordinates in eq. (5.4),
the differential equation is converted into a soluble algebraic equation. However,
the process of inverse Fourier transformation, which is required to extract an ex-
plicit DGF representation, is generally problematic.
5.2.1. Isotropic mediums

5.2.1.1. Dielectric–magnetic mediums The 3 × 3 DGF Gee iso
(r − r , ω) for the
isotropic dielectric–magnetic medium described by the Tellegen constitutive rela-
tions D(r, ω) = (ω)E(r, ω) and B(r, ω) = μ(ω)H (r, ω) emerges as a result of
applying ∇ • from the left to both sides of eq. (5.7). Thereby, we find that

1
Gee (r − r , ω) = iωμ(ω) I + ∇∇ giso (r − r , ω), (5.14)
iso ω2 (ω)μ(ω)
with the scalar Green function giso (r − r , ω) satisfying the scalar differential
equation
2
∇ + ω2 (ω)μ(ω) giso (r − r , ω) = −δ(r − r ). (5.15)
The solution of eq. (5.15), namely
1 1/2
giso (r − r , ω) =
exp iω (ω)μ(ω) |r − r | , (5.16)
4π|r − r |
is known very well (Chen [1983]).
At locations outside the source region (i.e., r = r ), the explicit representation
(Chen [1983], Van Bladel [1991])

G iso (R, ω) = iωμ(ω) (I − R̂ R̂)giso (R, ω)
ee
(5.17)
i
+ (I − 3R̂ R̂)giso (R, ω)
ω[(ω)μ(ω)]1/2 R

1
− 2 (I − 3 R̂ R̂)g iso (R, ω) (5.18)
ω (ω)μ(ω)R 2
follows from combining eqs. (5.14) and (5.16), with R = R R̂ = r − r . At
locations inside the source region the DGF is singular, as discussed in § 5.4.
In view of the expression (5.14) for Gee
iso
(r − r , ω), the dual DGF
(ω) ee
Gmm
iso
(r − r , ω) =
G (r − r , ω) (5.19)
μ(ω)
emerges immediately, whereas the expressions
Gem
iso
(r − r , ω) = −Gme
iso
(r − r , ω) = ∇ × g iso (r − r , ω)I (5.20)
follow from eqs. (5.9) and (5.10).
5.2.1.2. Isotropic chiral mediums The study of fields and sources in isotropic
chiral mediums is efficiently carried out via the introduction of the Beltrami fields
Q1 (r, ω) and Q2 (r, ω) defined in eqs. (4.20) and the corresponding Beltrami
source current densities (Lakhtakia [1994])
& '⎫
1 μ(ω) 1/2 ⎪
W 1 (r, ω) = i J e (r, ω) − J m (r, ω) ⎪
⎪
⎬
2 (ω)
& '⎪ . (5.21)
1 (ω) 1/2 ⎪
⎪
W 2 (r, ω) = J e (r, ω) − i J m (r, ω) ⎭
2 μ(ω)
Then the Maxwell curl postulates can be set down as
∇ × Q (r, ω) + (−1) k Q (r, ω) = W (r, ω) ( = 1, 2), (5.22)
with wavenumbers
1/2

k1 = ω (ω)μ(ω) − iχ(ω)
1/2
. (5.23)
k2 = ω (ω)μ(ω) + iχ(ω)
The solution of eq. (5.22) is

Q (r, ω) = Qcf (r, ω) + G (r − r , ω) • W (r , ω) d3 r ( = 1, 2),
V
(5.24)
where Qcf (r, ω) is the complementary function satisfying the equation

∇ × I + (−1) k I • Qcf (r, ω) = 0 ( = 1, 2). (5.25)
The dyadic Beltrami–Green functions

G (r − r , ω) = ∇ × I − (−1) k I

1
• I + ∇∇ g (r − r , ω) ( = 1, 2), (5.26)
k2
employ the scalar Green functions
exp(ik |r − r |)
g (r − r , ω) = ( = 1, 2), (5.27)
4π|r − r |
that are isomorphic to giso (R, ω).
A similar procedure is useful for biisotropic mediums characterized by
eqs. (3.6), as shown by Monzon [1990].
5.2.2. Uniaxial dielectric–magnetic mediums
Let us now turn to the anisotropic dielectric–magnetic mediums characterized by

a single distinguished axis, oriented arbitrarily along the direction of the unit vec-
tor û. These mediums are characterized by the constitutive relations (3.13) with
permittivity and permeability dyadics (3.8) and (3.11), respectively.
By a process involving the diagonalization of the linear differential operator L e
(or, equivalently, L m ) combined with dyadic factorizations, the 3 × 3 DGFs for
the uniaxial dielectric–magnetic medium are determined as (Weiglhofer [1990])

& ⎫
Gee −
= −T − ⎪
⎪
(r r , ω) iωμ(ω) (r r , ω) ⎪
⎪
uni ⎪
' ⎪
⎪
⎪
∇∇ ⎪
−1
+ u (ω) uni (ω) + 2 guni (r − r , ω) ⎪
⎪
⎪
⎪
ω (ω)μ(ω) ⎪
⎪
⎪
⎪
−1
G uni (r − r , ω) = −(ω) uni (ω) • (∇ × I )
em ⎪
⎪
⎪
⎪
−1
⎪
⎪
• μu (ω)g
μ
uni (r − r , ω)μ (ω) + T (r − r , ω) ⎬
uni
.
Gme (r − r , ω) = −μ(ω)μ−1 (ω) • (∇ × I ) ⎪
⎪
⎪
uni
uni
⎪
⎪
• −u (ω)g (r − r , ω) −1 (ω) + T (r − r , ω)
⎪
⎪
⎪
⎪
&
uni uni ⎪
⎪
⎪
⎪

G uni (r − r , ω) = iω(ω) T (r − r , ω)
mm ⎪
⎪
⎪
⎪
'⎪
⎪
⎪
⎪
∇∇ ⎪
−1
guni (r − r , ω) ⎪

μ
+ μu (ω)μ (ω) + 2 ⎭
uni ω (ω)μ(ω)
(5.28)
− r , ω)
,μ
The scalar Green functions guni (r in eqs. (5.28) are defined by
exp{iω[(ω)μ(ω)]1/2 [ηu (ω)R • η−1 (ω) • R]1/2 }

η uni
guni (R, ω) =
4π[ηu (ω)R • η−1 (ω) • R]1/2
uni
(η = , μ), (5.29)
while the dyadic T (r − r , ω) is specified as4

(R × û)(R × û) u (ω) μu (ω) μ
T (R, ω) = g (R, ω) − g (R, ω)
(R × û)2 (ω) uni μ(ω) uni

1 2(R × û)(R × û)
+ I − ûû −
iω[(ω)μ(ω)]1/2 (R × û)2 (R × û)2

× guni (R, ω) u (ω)R • −1
1/2
uni
(ω) • R
1/2
− guni (R, ω) μu (ω)R • μ−1 (ω) • R

μ
. (5.30)
uni
The specializations appropriate to uniaxial dielectric mediums and uniaxial
magnetic mediums follow straightforwardly from eqs. (5.28), by implementing
the substitutions μ (ω) = μ0 I and uni (ω) = 0 I , respectively.
uni
4 The T (r, ω) defined by eq. (5.30) is not related to the Maxwell stress dyadic T̃ (r, t) of eq. (2.43).
5.2.3. More complex mediums
For gyrotropic dielectric mediums and gyrotropic magnetic mediums, i.e., those
with permittivity and permeability dyadics of the form
η (ω) = η(ω)(I − ûû) + iηg (ω)û × I + ηu (ω)ûû (η = , μ), (5.31)

gyro
respectively, DGFs in terms of one-dimensional integrals involving cylindri-

cal functions are available, but closed-form representations are not (Weiglhofer
[1994]). Similarly, some analytical progress towards closed-form DGFs for
general uniaxial bianisotropic mediums described by the Tellegen constitutive
dyadics
η (ω) = η(ω)(I − ûû) + ηu (ω)ûû (η = , ξ, ζ, μ), (5.32)

uni
has been reported, but has not yet reached a satisfactory degree of conclusion
(Weiglhofer and Lindell [1994], Olyslager and Lindell [2002]).
A noteworthy example of a bianisotropic scenario for which a closed-form
DGF is available arises within the context of gravitationally affected vacuum.
As discussed in §§ 3.4.2 and 4.6.2, the electromagnetic properties of vacuum in
generally curved space–time are equivalent to those of a fictitious, instantaneously
responding, nonhomogeneous, bianisotropic medium. Within a space–time neigh-
bourhood which is small compared to the curvature of space–time, in a noncovari-
ant formalism the nonhomogeneous bianisotropic medium may be approximated
by the homogeneous bianisotropic medium with time-domain constitutive rela-
tions (4.62).
On using the frequency-domain equivalent of eqs. (4.62), the corresponding
DGFs may be derived by the following two-step procedure. First, by applying the
field-source transformations

F (r, ω) = F(r, ω) exp −iω(0 μ0 )1/2 Γ̄ • r
, (5.33)
Q (r, ω) = Q(r, ω) exp −iω(0 μ0 )1/2 Γ̄ • r
the problem simplifies to that for an orthorhombic biaxial dielectric–magnetic
medium. Second, under the affine transformations

F (r, ω) = Diag γ̄ 1/2 , γ̄ 1/2 • F γ̄ 1/2 • r, ω
, (5.34)
Q (r, ω) = Diag adj γ̄ 1/2 , adj γ̄ 1/2 • Q γ̄ 1/2 • r, ω
wherein γ̄ 1/2 • γ̄ 1/2 = γ̄ , the problem further simplifies to that for an isotropic
dielectric–magnetic medium. Thus, the 3 × 3 DGFs emerge as (Lakhtakia and
Mackay [2005])
⎫
1
Gee (r − r
, ω) = iωμ adj γ̄ 1/2 •
I+ γ̄ • ∇∇ ⎪
⎪
GAV 0
ω2 0 μ0 det γ̄ ⎪
⎪
⎪
⎪
⎪
⎪

× gGAV (r − r , ω) ⎪
⎪
⎪
⎪
⎪
⎪
−1 • ⎪
⎬
G GAV (r − r , ω) = −γ̄
em
∇ × I − ω(0 μ0 ) Γ̄ × I
1/2
,
• Gee (r − r , ω) ⎪
⎪
⎪
⎪
GAV
⎪
⎪
⎪
⎪
Gme (r − r , ω) = −G em
(r − r , ω) ⎪
⎪
GAV GAV ⎪
⎪
⎪
⎪
0 ee ⎪
⎭
G GAV (r − r , ω) =
mm
G GAV (r − r , ω)
μ0
(5.35)
with the scalar Green function gGAV (r − r , ω) being given by
exp{iω(0 μ0 )1/2 [Γ̄ • R + (det γ̄ )1/2 |γ̄ −1/2 • R|]}

gGAV (R, ω) = . (5.36)
4π|γ̄ −1/2 • R|
5.2.4. Huygens principle
The Huygens principle represents a keystone of electromagnetic scattering the-

ory (Brau [2004]). By means of this principle, the electromagnetic fields in a
source-free region may be related to the DGF and the tangential field compo-
nents on a closed surface enclosing the source region. The Huygens principle
may be applied, for example, to diffraction from an aperture, wherein the aper-
ture is formally represented as an equivalent source (Chen [1983]); likewise, it
can be used to formulate the Ewald–Oseen extinction theorem and the T matrix
method (Lakhtakia [1994]). Due to the general scarcity of DGFs in closed form,
exact formulations of the Huygens principle are available only for isotropic medi-
ums (Lakhtakia [1994]) and the simplest of anisotropic mediums (Ogg [1971],
Lakhtakia, Varadan and Varadan [1989], Tai [1994]).
5.3. Eigenfunction representations
5.3.1. Homogeneous mediums
As an alternative to the closed-form representations considered in § 5.2, DGFs

may be expressed in terms of expansions of eigenfunction solutions of eq. (5.4).
Thereby, the spectral representation (Kong [1972])

3
1
G(R, ω) = Ǧ(q, ω) exp(iq • R) d3 q (5.37)
2π
q
emerges. Although closed-form representations of DGFs for anisotropic and bian-

isotropic mediums are relatively scarce, the spatial Fourier transform (Kong
[1972])

Ǧ(q, ω) = G(R, ω) exp(−iq • R) d3 R (5.38)
R
can always be found. Thus, for the homogeneous bianisotropic medium charac-
terized by the Tellegen 6 × 6 constitutive dyadic K EH (ω), the spatial Fourier
transform of eq. (5.4) yields
1 adj Ǎ(q, ω)
Ǧ(q, ω) = , (5.39)
iω det Ǎ(q, ω)
where

0 (q/ω) × I
Ǎ(q, ω) = + K EH (ω). (5.40)
−(q/ω) × I 0
Whereas the analytic properties of the inverse Fourier transform
3
1 1 adj[Ã(q, ω)]
G(R, ω) = exp(iq • R) d3 q (5.41)
2π iω det[Ã(q, ω)]
q
have been determined (Cottis and Kondylis [1995]), numerical techniques are
generally needed to explicitly evaluate the right side of eq. (5.41).
Equation (5.37) is a spectral representation in terms of plane waves. Simi-
lar representations of DGFs have been established for other sets of eigenfunc-
tions. Most notably, expansions in terms of cylindrical vector wavefunctions
(Chew [1999], Li, Kang and Leong [2001]) have been developed for certain uni-
axial bianisotropic mediums (Cheng [1997], Cheng, Ren and Jin [1997], Wu
and Yasumoto [1997], Tan and Tan [1998], Li, Leong, Kooi and Yeo [1999],
Ren [1999]). Also, formulations are available in terms of spherical harmonics
(Monzon [1989]). However, these DGF representations are exceedingly cumber-
some to handle and restricted in scope, and generally require numerical imple-
mentation.
5.3.2. Nonhomogeneous mediums

DGFs for nonhomogeneous mediums are quite painful to even contemplate.
A conspicuous exception is the matrix Green function of a HBM, which may be
developed as follows. Let us recall from § 4.6.1 that for a HBM with its helicoidal
axis aligned with the Cartesian z axis, as characterized by the Tellegen constitu-
tive relations (3.48), with constitutive dyadics (3.49) and (3.50), propagation in
source-free regions is governed by the 4 × 4 matrix differential equation (4.57).
In the presence of an Oseen-transformed source term J (z, ω), corresponding to
the Oseen-transformed field term F (z, ω), the governing equation (4.57) may be
re-expressed as (Lakhtakia and Weiglhofer [1997b])
∂
F (z, ω) = M (z, ω)F (z, ω) + J (z, ω). (5.42)
∂z
The solution of eq. (5.42) is compactly stated as
z

F (z, ω) = Ǧ
HBM
(z, ω)F (0, ω) + Ǧ
HBM
(z − zs , ω)J (zs , ω) dzs ,
0
(5.43)
in terms of the spectral matrix Green function
−1
Ǧ (z − zs , ω) = Z (z, ω) Z (zs , ω) . (5.44)
HBM
The matrizant Z (z, ω) in eq. (5.44) is the 4×4 matrix that satisfies the differential
equation (Yakubovich and Starzhinskii [1975])
∂
Z (z, ω) = M (z, ω)Z (z, ω), (5.45)
∂z
with boundary condition
Z (0, ω) = I. (5.46)
5.4. Depolarization dyadics
Often, it is sufficient to construct approximate solutions of restricted validity. An

important example – explored in detail in § 6 – is provided by the homogenization
of particulate composite materials (Lakhtakia [1996, 2000a]). Here, the scattering
response of an electrically small and homogeneous particle embedded within a
homogeneous ambient medium is required.
Let K p (ω) and K amb (ω) denote the Tellegen constitutive dyadics for the parti-
cle and the ambient mediums, respectively. Then the corresponding Maxwell curl
postulates (5.1) may be written in the form

L(∇) + iωK amb (ω) • F(r, ω) = Q equiv (r, ω). (5.47)
Herein, the particle is represented by an equivalent source current density
&
iω[K amb (ω) − K p (ω)] • F(r, ω), r ∈ V ,
Q equiv (r, ω) = (5.48)
0, r∈/ V ,
which is assumed to be uniform within the region V occupied by the particle.
As the particle is sufficiently small relative to all relevant electromagnetic wave-
lengths, the Rayleigh approximation may be implemented to estimate the uniform
field F(r, ω) in V as
F(r, ω) ≈ Fcf (r, ω) + D(r, ω) • Q equiv (r, ω) (r ∈ V ). (5.49)

The 6 × 6 dyadic

D(r, ω) = G(r − r , ω) d3 r (5.50)
V
is the depolarization dyadic of a region of the same shape, orientation and size as
the particle (Lakhtakia [2000a]). The depolarization dyadic owes its existence to
the singularity of the DGF of the ambient medium.
5.4.1. Ellipsoidal shape

Consider an ellipsoidal particle with surface parameterized as
r e (θ, φ) = ρU • r̂(θ, φ), (5.51)

where r̂(θ, φ) is the radial unit vector specified by the spherical polar coordinates
θ and φ, and ρ > 0 is a measure of the linear dimensions. The shape dyadic
U is a real-valued 3 × 3 dyadic with positive eigenvalues (Lakhtakia [2000b]),
the normalized lengths of the ellipsoid semi-axes being specified by the eigen-
values of U . Suppose that the particle is embedded within a homogeneous ambi-
ent medium described by 3 × 3 Tellegen constitutive dyadics amb (ω), ξ (ω),
amb
ζ (ω) and μ (ω); i.e., the ambient medium is generally bianisotropic. In the
amb amb
limit ρ → 0, the 6 × 6 depolarization dyadic for the ellipsoidal particle (5.50)
may be expressed as (Michel and Weiglhofer [1997])
2π π
∞
D(r, ω) = Ǧ U −1 • q̂, ω sin θ dθ dφ, (5.52)
φ=0 θ=0
with q̂ = (sin θ cos φ, sin θ sin φ, cos θ ), and wherein the spectral representa-
tion (5.41) is exploited to provide
∞
Ǧ (q̂, ω) = lim Ǧ(q, ω). (5.53)
q→∞
Utilizing eq. (5.39) to expand the integrands in eq. (5.52), we have

U/amb
D ee (ω) D U/ambem
(ω)
D U/amb
(ω) = , (5.54)
D U/amb
me
(ω) D U/amb (ω)
mm
where
(ω) = U −1 • d λλ (ω) • U −1 (λλ = ee, em, me, mm), (5.55)

U/amb U/amb
D λλ
and
⎫
2π π q̂ • U −1 • μ (ω) • U −1 • q̂ ⎪
⎪
⎪
⎪
d U/amb (ω) = amb
q̂ q̂ sin θ dθ dφ ⎪
⎪
ee 4πiωb(θ, φ) ⎪
⎪
⎪
⎪
φ=0 θ=0 ⎪
⎪
⎪
⎪
2π π q̂ • U −1 • ξ −1
(ω) • U • q̂ ⎪
⎪
⎪
d em (ω) =
U/amb
− amb
⎪
⎪
⎪
4πiωb(θ, φ) ⎪
⎬
φ=0 θ=0
.
2π π q̂ • U −1 • ζ (ω) • U −1 • q̂ ⎪
⎪
⎪
⎪
d U/amb (ω) = − amb
⎪
me ⎪
⎪
4πiωb(θ, φ) ⎪
⎪
φ=0 θ=0 ⎪
⎪
⎪
⎪
q̂ • U −1 •
2π π −1 ⎪
⎪
(ω) • U • q̂ ⎪
⎪
d mm (ω) =
U/amb amb
q̂ q̂ sin θ dθ dφ ⎪ ⎪
⎪
4πiωb(θ, φ) ⎪
⎭
φ=0 θ=0
(5.56)
Herein,

b(θ, φ) = q̂ • U −1 • amb (ω) • U −1 • q̂ q̂ • U −1 • μ (ω) • U −1 • q̂
amb

− q̂ • U −1 • ξ (ω) • U −1 • q̂ q̂ • U −1 • ζ (ω) • U −1 • q̂ ,
amb amb
(5.57)
and
q̂ = x̂ sin θ cos φ + ŷ sin θ sin φ + ẑ cos θ. (5.58)

Thus, eqs. (5.56) provide the components of the integrated singularity of the DGF.
The depolarization dyadic does not involve the constitutive properties of the par-
ticle that occupies V . Depolarization dyadics associated with ellipsoidal particles
which are small compared to all relevant wavelengths, but not vanishingly small,
may also be derived from the spectral representation (5.41) (Mackay [2004], Cui
and Mackay [2007]).
Explicit expressions for the integrals in eqs. (5.56) are available for isotropic
ambient mediums as well as several classes of anisotropic ambient mediums.
Parenthetically, expressions for depolarization dyadics associated with certain
cylindrical shapes have also been derived (Cottis, Vazouras and Spyrou [1999],
Weiglhofer and Mackay [2002]).
5.4.2. Spherical shape

In view of their use in the homogenization formalisms discussed in § 6, let us
move on to depolarization dyadics for spherical shapes (i.e., U = I ).
5.4.2.1. Isotropic ambient mediums When the ambient medium is isotropic

dielectric–magnetic, i.e.,
⎫
η (ω) = η(ω)I (η = , μ)⎬
amb
, (5.59)
ξ (ω) = ζ (ω) = 0 ⎭
amb amb
the corresponding 3 × 3 depolarization dyadics reduce to the well-known forms

(Van Bladel [1991])
⎫
D ee (ω) =
I /iso 1
I , D em (ω) = 0
I /iso ⎪
⎪
⎬
3iω(ω)
. (5.60)
1 ⎪
D Ime/iso
(ω) = 0, D Imm
/iso
(ω) = I⎪
⎭
3iωμ(ω)
When the ambient medium is isotropic chiral, i.e.,
⎫
η (ω) = η(ω)I (η = , μ) ⎬
amb
, (5.61)
ξ (ω) = −ζ (ω) = χ(ω)I ⎭
amb amb
the 3 × 3 depolarization dyadics have the scalar matrix forms

⎫
μ(ω) χ(ω) ⎪
D Iee/iso (ω) = I , D Iem
/iso
(ω) = − I⎪
⎬
3iωΥ (ω) 3iωΥ (ω)
, (5.62)
χ(ω) (ω) ⎪
D Ime
/iso
(ω) = I , D Imm
/iso
(ω) = I ⎪⎭
3iωΥ (ω) 3iωΥ (ω)
where
Υ (ω) = (ω)μ(ω) + χ 2 (ω). (5.63)

5.4.2.2. Anisotropic ambient mediums Let us now turn to the uniaxial dielectric–
magnetic scenario wherein the constitutive dyadics of the ambient medium are
⎫
η (ω) = η (ω) (η = , μ)⎬
amb uni
, (5.64)
ξ (ω) = ζ (ω) = 0 ⎭
amb amb
with uni (ω) and μ (ω) as specified in eqs. (3.8) and (3.11), respectively. The
uni
3 × 3 depolarization dyadics are now given by (Michel [1997])
⎫
D Iee/uni (ω) = D (ω)(I − ûû) + Du (ω)ûû ⎪
⎪
⎬
D Iem/uni
(ω) = D Ime
/uni
(ω) = 0 , (5.65)
⎪
⎪
D I /uni
(ω) = D (ω)(I − ûû) + D (ω)ûû
μ μ ⎭
mm u
with
⎫
D (ω) =
η 1 ⎪
⎪
⎪
⎪
2iω[η(ω) − ηu (ω)] ⎪
⎪
⎪
⎪
& γη (ω) sinh −1 1−γη (ω) 1/2 ' ⎪
⎬
γη (ω)
× 1− (η = , μ),
[1 − γη (ω)]1/2 ⎪
⎪
⎪
& sinh−1 1−γη (ω) 1/2 '⎪
⎪
⎪
1 γη (ω) ⎪
⎪
Duη (ω) = − 1 ⎪
⎭
iω[η(ω) − ηu (ω)] [1 − γη (ω)] 1/2
(5.66)
and
ηu (ω)
γη (ω) = (η = , μ). (5.67)
η(ω)
The depolarization dyadics relevant to uniaxial dielectric mediums and uni-
axial magnetic mediums follow immediately from eqs. (5.65), upon substitut-
ing μ (ω) = μ0 I and uni (ω) = 0 I , respectively. Depolarization dyadics of
uni
the form (5.65) also arise for a spheroidal shape, with its rotational axis aligned
with û, immersed in an isotropic dielectric–magnetic ambient medium (Mackay
and Lakhtakia [2005]).
If the ambient medium is either gyrotropic dielectric, i.e.,
⎫
amb (ω) = gyro (ω) ⎪
⎪
⎬
ξ (ω) = ζ (ω) = 0 , (5.68)
amb amb ⎪
⎪
μ (ω) = μ I ⎭
0
amb
or gyrotropic magnetic, i.e.,
⎫
amb (ω) = 0 I ⎪
⎪
⎪
⎬
ξ (ω) = ζ (ω) = 0 , (5.69)
amb amb ⎪
⎪
⎪
μ (ω) = μ (ω) ⎭
amb gyro
with gyro (ω) and μ (ω) specified in eq. (5.31), the skew-symmetric dyadic
gyro
components make no contribution to the depolarization dyadics.5 Hence, the de-
polarization dyadics for gyrotropic dielectric mediums and gyrotropic magnetic
mediums are the same as those for the corresponding uniaxial dielectric mediums
and uniaxial magnetic mediums, respectively.
Finally, suppose that the ambient medium is orthorhombic dielectric–magnetic;
i.e.,
⎫
η (ω) = ηortho (ω) (η = , μ)⎬
amb bi
, (5.70)
ξ (ω) = ζ (ω) = 0 ⎭
amb amb
with
ηortho (ω) = ηx (ω)x̂ x̂ + ηy (ω)ŷ ŷ + ηz (ω)ẑẑ

bi
(η = , μ). (5.71)
The corresponding 3 × 3 DGFs for spherical shapes are given as (Weiglhofer

[1998b])
1 ⎫
D Iee/ortho (ω) = Dx (ω)x̂ x̂ + Dy (ω)ŷ ŷ + Dz (ω)ẑẑ ⎪
⎪
⎪
iω0 ⎪
⎪
⎬
D Iem
/ortho
= D Ime
/ortho
=0 . (5.72)
⎪
⎪
1 μ ⎪
⎪
D Imm
/ortho
= Dx (ω)x̂ x̂ + Dy (ω)ŷ ŷ + Dz (ω)ẑẑ ⎪
μ μ ⎭
iωμ0
5 In fact, it follows from the expressions (5.56) that the skew-symmetric components of any consti-
tutive dyadic vanish from the depolarization dyadics.
Here,
1/2 ⎫
ηy (ω)[F (λ1 , λ2 ) − E(λ1 , λ2 )] ⎪
⎪
Dxη = ⎪
⎪
[ηy (ω) − ηx (ω)][ηz (ω) − ηx (ω)] 1/2 ⎪
⎪
& 1/2 ⎪
⎪
⎪
⎪
1 ηx (ω) − ηy (ω) ηz (ω) − ηx (ω) ⎪
⎪
Dyη = − ⎪
⎪
ηy (ω) − ηx (ω) ηz (ω) − ηy (ω) ηy (ω) ⎪
⎪
⎪
⎬
ηx (ω)
× F (λ1 , λ2 ) ⎪
,
ηz (ω) − ηx (ω) ⎪
⎪
' ⎪
⎪
ηy (ω) ⎪
⎪
− E(λ1 , λ2 ) ⎪
⎪
ηz (ω) − ηy (ω) ⎪
⎪
& 1/2 ' ⎪ ⎪
⎪
⎪
1 ηy (ω) ⎪
⎪
Dz =
η
1− E(λ1 , λ2 ) ⎭
ηz (ω) − ηy (ω) ηz (ω) − ηx (ω)
(η = , μ), (5.73)
involve F (λ1 , λ2 ) and E(λ1 , λ2 ) as elliptic integrals of the first and second kinds
(Gradshteyn and Ryzhik [1980]), respectively, with arguments
⎫
−1 ηz (ω) − ηx (ω) 1/2 ⎪ ⎪
λ1 = tan ⎪
⎬
ηx (ω)
& '1/2 (η = , μ). (5.74)
ηz (ω)[ηy (ω) − ηx (ω)] ⎪
⎪
λ2 = ⎪
⎭
ηy (ω)[ηz (ω) − ηx (ω)]
Depolarization dyadics for the corresponding orthorhombic dielectric mediums

and orthorhombic magnetic mediums follow immediately from eqs. (5.72), upon
substituting μortho (ω) = μ0 I and ortho
bi
(ω) = 0 I , respectively. The form of the
bi
depolarization dyadics (5.72) also arises for an ellipsoidal shape, with rotational
axes aligned with x̂, ŷ and ẑ, in an isotropic dielectric–magnetic ambient medium
(Fricke [1953], Weiglhofer [1998b]).
5.4.2.3. Bianisotropic ambient mediums Closed-form representations of depo-

larization dyadics are not available for a general bianisotropic medium, but an-
alytical progress has been reported for certain uniaxial bianisotropic mediums
(Michel and Weiglhofer [1997]). When the ambient medium is characterized by
the constitutive relations (4.62), the skew-symmetric components of the magneto-
electric constitutive dyadics do not contribute to the associated depolarization
dyadics; therefore, the depolarization dyadics are simply those of the correspond-
ing anisotropic dielectric–magnetic mediums.
§ 6. Homogenization
Bianisotropic mediums may be readily conceptualized as homogenized compos-

ite mediums (HCMs), arising from constituent mediums which are themselves
not bianisotropic (or even anisotropic). The constitutive parameters of HCMs are
estimated by homogenization formalisms.
A mixture of two or more different mediums may be viewed as being effectively
homogeneous provided that wavelengths are much longer than the length scales
of the nonhomogeneities (Garland and Tanner [1978], Priou [1992], Lakhtakia
[2000a]). In fact, the process of homogenization implicitly underpins our de-
scriptions of anisotropy and bianisotropy, since the constitutive relations relate
macroscopic electromagnetic fields which represent the volume-averages of their
microscopic counterparts (Van Kranendonk and Sipe [1977], Buchwald [1985]).
On a more practical level, homogenization is important in the interpretation of
experimental measurements and in material design. This latter topic continues to
motivate research.
HCMs, and anisotropic and bianisotropic HCMs in particular, furnish prime
examples of metamaterials (Walser [2003]). These are artificial composite mate-
rials which exhibit properties that are either not exhibited at all or not exhibited
to the same extent by their constituent materials. The concept that complex medi-
ums may be realized through the homogenization of relatively simple constituent
mediums was promoted initially for isotropic chiral HCMs (Sihvola and Lindell
[1990], Lakhtakia [1993]), and subsequently extended to anisotropic and bian-
isotropic HCMs (Michel [2000]).
The origins of formalisms used to estimate the constitutive parameters of
HCMs may be traced back to the earliest years of electromagnetic theory. For
a historical perspective, see the introduction to an anthology of milestone papers
on HCMs (Lakhtakia [1996]). Two of the most widely used formalisms today –
the Maxwell Garnett formalism (Maxwell Garnett [1904]) and the Bruggeman
formalism (Bruggeman [1935]) – were first developed for isotropic dielectric–
magnetic HCMs. However, generalizations of these formalisms appropriate to
anisotropic and bianisotropic HCMs emerged much more recently (Weiglhofer,
Lakhtakia and Michel [1997]), following the development of expressions for the
corresponding depolarization dyadics (Michel and Weiglhofer [1997]).
In the conventional approaches to homogenization, as exemplified by the
Maxwell Garnett and Bruggeman formalisms, the distributional statistics of the
HCM constituents are described solely in terms of their respective volume frac-
tions. A more sophisticated approach is taken in the strong-property-fluctuation
theory (SPFT), wherein spatial correlation functions of arbitrarily high order may
3, § 6] Homogenization 193
be accommodated (Tsang and Kong [1981]). The development of the SPFT for
bianisotropic HCMs began in the 1990s (Zhuck [1994], Michel and Lakhtakia
[1995]) and was achieved at a practical level at the beginning of the twenty-first
century (Mackay, Lakhtakia and Weiglhofer [2000]).
The estimation of constitutive dyadics of linear bianisotropic HCMs is de-
scribed in the following subsections, using the Maxwell-Garnett, Bruggeman and
SPFT formalisms. These approaches are rigorously established, in contrast to ex-
trapolations of isotropic formalisms (Sihvola and Pekonen [1996, 1997], Shanker
[1997]).
6.1. Constituent mediums
Let us focus upon HCMs arising from two particulate constituent mediums, la-
belled a and b. Each constituent medium is itself homogeneous. The Tellegen con-
stitutive dyadics of the constituent mediums are denoted by K a (ω) and K b (ω).
The constituent particles are taken to be generally ellipsoidal in shape. The el-
lipsoids of each constituent medium are conformal and have the same orientation,
but are randomly distributed. The shapes of the ellipsoids are characterized by the
real-symmetric 3 × 3 dyadics U a and U b , as introduced in eq. (5.51). The sizes of
the ellipsoids are not specified, but their linear dimensions must be much smaller
than the electromagnetic wavelength(s).6
Let V denote the unbounded space occupied by the composite medium. This
space is partitioned into the disjoint regions Va and Vb which contain the con-
stituent mediums a and b, respectively. In order to completely fill V with ellip-
soids, a fractal-like distribution of the constituent particles is implicit. The distri-
butions of the two constituent mediums throughout V are specified in terms of the
characteristic functions
&
1, r ∈ V
Φ (r) = ( = a, b). (6.1)
0, r ∈ / V
The nth moment of Φ (r) is the ensemble average
Φ (r 1 ) · · · Φ (r n )e , which
represents the probability of r 1 , . . . , r n ∈ V ( = a, b). The volume fraction of
constituent medium is given by f =
Φ (r)e ( = a, b); clearly, fa + fb = 1.
6 We refrain from discussing homogenization formalisms which take into account finite sizes of
constituent particles, as these are largely restricted to isotropic dielectric–magnetic (Doyle [1989],
Dungey and Bohren [1991]) and isotropic chiral (Shanker and Lakhtakia [1993], Shanker [1996])
HCMs. Exceptions are provided by certain extensions to the Maxwell Garnett formalism (Lakhtakia
and Shanker [1993]) and the SPFT (Mackay [2004], Cui and Mackay [2007]).
Only the first moments of Φa,b (r) are used in the Maxwell Garnett and Brugge-
man formalisms, whereas arbitrarily high-order moments can be accommodated
in the SPFT (Tsang and Kong [1981], Mackay, Lakhtakia and Weiglhofer [2000]).
The polarizability density dyadic of an ellipsoid of medium ( = a, b) em-
bedded in a homogeneous ambient medium described by the Tellegen constitutive
dyadic K amb (ω), is defined as (Michel [2000])

α /amb (ω) = K (ω) − K amb (ω) • I + iωDU /amb (ω)

−1
• K (ω) − K (ω) ( = a, b), (6.2)
amb

with DU /amb (ω) being the depolarization dyadic defined in eq. (5.54). The dyadic
α /amb (ω) is central to the Maxwell Garnett and Bruggeman formalisms.
6.2. Maxwell Garnett formalism
The Maxwell Garnett homogenization formalism has been used extensively, de-
spite its applicability being limited to dilute composites (Lakhtakia [1996]). A rig-
orous basis for this formalism was established by Faxén [1920] for isotropic
dielectric-in-dielectric composite mediums, and its standing was further bolstered
by the establishment of the closely related Hashin–Shtrikman bounds (Hashin and
Shtrikman [1962]).
In the Maxwell Garnett formalism, the mixture of the two constituent medi-
ums may be envisaged as a collection of well-separated particles (of medium a,
say) randomly dispersed in a simply connected host medium (medium b, say).
The Maxwell Garnett estimate of the Tellegen constitutive dyadic of the HCM is
(Weiglhofer, Lakhtakia and Michel [1997])
−1
K MG (ω) = K b (ω) + fa α a/b (ω) • I − iωfa DI /b (ω) • α a/b (ω) . (6.3)
Herein, the depolarization dyadic DI /b (ω) for a spherical region in constituent
medium b indicates the incorporation of a spherical Lorentzian cavity in the
Maxwell Garnett formalism. Notice that if DI /b (ω) in eq. (6.3) were to be re-
a
placed by DU /b (ω), then the Bragg–Pippard formalism would arise (Sherwin
and Lakhtakia [2002]).
The Maxwell Garnett estimate (6.3) is valid for fa 0.3 only. In order to
overcome this restriction, the following two refinements of the Maxwell Garnett
formalism have been established.
In the incremental Maxwell Garnett homogenization formalism, the estimate
of the HCM constitutive dyadic is constructed incrementally, by adding the con-
stituent medium a to the constituent medium b not all at once but in a fixed
number N of stages (Lakhtakia [1998b]). After each increment, the composite

is homogenized using the Maxwell Garnett formalism. Thus, the iteration scheme
1/N a/n
K[n + 1](ω) = K[n](ω) + 1 − fb α (ω)
1/N I /n −1
• I − iω 1 − f D (ω) • α a/n (ω) (6.4)
b
emerges for n = 0, 1, . . . , N − 1, where K[0](ω) = K b (ω). In eq. (6.4), α a/n (ω)

is the polarizability density dyadic of a constituent material phase a particle, of
shape specified by U a , relative to the homogeneous medium characterized by the
Tellegen constitutive dyadic K[n](ω). The incremental Maxwell Garnett estimate
of the Tellegen constitutive dyadic of the HCM is delivered as
K IMG (ω) = K[N ](ω). (6.5)

In the limit N → ∞, the incremental Maxwell Garnett formalism gives rise to
the differential Maxwell Garnett homogenization formalism (Michel, Lakhtakia,
Weiglhofer and Mackay [2001]). That is, the differential Maxwell Garnett esti-
mate K DMG (ω) of the Tellegen constitutive dyadic of the HCM is the solution of
the ordinary differential equation
∂ 1
K(ω, s) = α a/s (ω), (6.6)
∂s 1−υ
with initial value K(ω, 0) = K b (ω). The continuous variable s represents the
volume fraction of constituent medium a added to medium b, and α a/s (ω) is the
polarizability density dyadic of a particle made of constituent medium a, and
of shape given by U a , immersed in a medium with constitutive dyadic K(ω, s).
The particular solution of eq. (6.6) satisfying s = fa is the differential Maxwell
Garnett estimate; i.e.,
K DMG (ω) = K(ω, fa ). (6.7)

We note that the recursive approach that underpins the incremental and dif-
ferential Maxwell Garnett formalisms has been applied quite widely within
the realm of isotropic dielectric HCMs (Ghosh and Fuchs [1991], Barrera and
Fuchs [1995], Fuchs, Barrera and Carrillo [1996], Sosa, Mendoza and Barrera
[2001]).
6.3. Bruggeman formalism
A central characteristic of the Bruggeman homogenization formalism is that the

constituent mediums a and b are treated symmetrically (Milton [1985], Ward
[2000]). The Bruggeman formalism is thereby applicable for all volume fractions
fa ∈ (0, 1), unlike the conventional Maxwell Garnett formalism. The symmet-
rical treatment of the constituent mediums enables the phenomenon of percola-
tion threshold to be predicted by the Bruggeman formalism, albeit not always
correctly (Brouers [1986], Berthier and Peiro [1998], Goncharenko and Venger
[2004]). Several variations on the Bruggeman formalism have been developed for
isotropic dielectric–magnetic mediums (Hanai [1960], Niklasson and Granqvist
[1984], Goncharenko [2003]). Indeed, by a microscopic treatment of the local
field effects, the Bruggeman formalism may be shown to arise naturally from
the Maxwell Garnett formalism (Aspnes [1982]). The rigorous establishment of
the Bruggeman formalism for the most general linear scenarios follows from the
SPFT (Mackay, Lakhtakia and Weiglhofer [2000]).
The ellipsoidal particles made of the two constituent mediums can have dif-
ferent shapes. The polarizability density dyadics implemented in the Bruggeman
formalism, for each constituent material phase, are calculated relative to the HCM
itself; i.e., the polarizability density dyadic α /Br (ω) ( = a, b) is defined in
terms of the Bruggeman estimate K Br (ω) of the Tellegen constitutive dyadic of
the HCM.
The assertion that the net polarizability density is zero throughout the HCM
underlies the Bruggeman homogenization formalism. Thus, the Bruggeman es-
timate of the HCM constitutive dyadic is provided implicitly by the nonlinear
equation (Weiglhofer, Lakhtakia and Michel [1997])
fa α a/Br (ω) + fb α b/Br (ω) = 0, (6.8)
from which K Br (ω) can be extracted by applying the simple Jacobi technique
(Michel, Lakhtakia and Weiglhofer [1998]). Hence, the iterative solution

K Br [n](ω) = S K Br [n − 1](ω) (n = 1, 2, . . .) (6.9)
is developed, with the initial value K Br [0](ω) = K MG (ω). The action of the
operator S is defined by

a
−1
S K Br (ω) = fa K a (ω) • I + iωDU /Br (ω) • K a (ω) − K Br (ω)
b
−1
+ fb K b (ω) • I + iωDU /Br (ω) • K b (ω) − K Br (ω)
a
−1
• fa I + iωDU /Br (ω) • K (ω) − K (ω)
a Br
b
−1 −1
+ fb I + iωDU /Br (ω) • K b (ω) − K Br (ω) .
(6.10)
6.4. Strong-property-fluctuation theory
6.4.1. Background
The provenance of the strong-property-fluctuation theory7 (SPFT) lies in wave-
propagation studies for continuous random mediums (Ryzhov and Tamoikin
[1970], Tatarskii and Zavorotnyi [1980]). However, in recent years it has been
adapted to estimate the constitutive parameters of HCMs (Tsang and Kong
[1981]). In contrast to the conventional Maxwell Garnett and Bruggeman for-
malisms, the SPFT accommodates a comprehensive description of the distribu-
tional statistics of the constituent material phases. Thereby, coherent scattering
losses may be accounted for.
The SPFT approach to homogenization is based upon the iterative refinement
of a comparison medium; i.e., the homogeneous medium specified by the Telle-
gen constitutive dyadic K comp (ω). The refinement process involves applying a
Feynman-diagrammatic technique to ensemble-average a Born series representa-
tion of the electromagnetic fields. While the straightforward approach is limited
to weak spatial fluctuations in the infinity norm of
K a (ω)Φa (r) + K b (ω)Φb (r) − K comp (ω), (6.11)
a careful consideration of the singularity of the DGF G comp (r −r , ω) of the com-

parison medium allows a reformulation that is applicable for strong fluctuations.
When electromagnetic wavelengths are much larger than the length scales of
inhomogeneities, the mixture of material phases may viewed as being effectively
homogeneous. In this long-wavelength regime, the nth-order estimate K[n] SPFT
(ω)
of the Tellegen constitutive dyadic of the HCM is delivered by the SPFT as
(Mackay, Lakhtakia and Weiglhofer [2000, 2001a])
1 −1
K[n]
SPFT
(ω) = K comp (ω) − I + Σ [n] (ω) • DU/comp (ω) • Σ [n] (ω),
iω
(6.12)
wherein the mass operator term Σ [n] (ω) – which has an infinite series represen-
tation – is defined in terms of the DGF for the comparison medium, together with
the generalized polarizability density dyadic
α U/comp (r, ω) = α a/comp (ω)Φa (r) + α b/comp (ω)Φb (r). (6.13)
7 It is otherwise known as the strong-permittivity-fluctuation theory in the context of dielectric medi-

ums.
The two populations of ellipsoids are required to have the same shape dyadics;
i.e., U a = U b = U . The condition
U/comp
α (r, ω) e = 0 (6.14)
is imposed in order to eliminate secular terms from the Born series representation
(Frisch [1970]).
6.4.2. SPFT estimates

6.4.2.1. Zeroth and first orders Both the zeroth-order and the first-order
mass operator are null-valued; i.e., Σ [0] (ω) = Σ [1] (ω) = 0. The estimates
K[0]
SPFT
(ω) = K[1] SPFT
(ω) = K comp (ω) may be extracted from the condition (6.14).
In fact, as the condition (6.14) is equivalent to the Bruggeman equation (6.8), the
lowest-order SPFT estimates of the Tellegen constitutive dyadic of the HCM are
identical to its Bruggeman estimate.
6.4.2.2. Second order The SPFT is most commonly implemented at the second-
order level of approximation – otherwise known as the bilocal approximation
(Tsang and Kong [1981], Stogryn [1983], Genchev [1992], Zhuck [1994], Michel
and Lakhtakia [1995], Mackay, Lakhtakia and Weiglhofer [2000]). Therein, the
distributional statistics of the constituent mediums a and b are characterized by
the two-point covariance function

Λ(r − r ) = Φa (r)Φa (r ) e − Φa (r) e Φa (r ) e

= Φb (r)Φb (r ) e − Φb (r) e Φb (r ) e , (6.15)
along with its associated correlation length L. Within a region of linear dimen-
sions given by L, and of shape dictated by the covariance function, the correlated
responses of scattering centres give rise to an attenuation of the macroscopic co-
herent field. On the other hand, the responses of scattering centres separated by
distances much greater than L are statistically independent.
The second-order mass operator is given by (Mackay, Lakhtakia and Weigl-
hofer [2000])

Σ [2] (ω) = −ω2 α a/comp (ω) − α b/comp (ω)

• PΛ (ω) • α a/comp (ω) − α b/comp (ω) , (6.16)
where

P (ω) = P
Λ
Λ(R)G comp (R, ω) d3 R, (6.17)
R
(
with P . . . d3 R denoting principal value integration.
The principal value integral (6.17) has been investigated theoretically and nu-
merically for various physically motivated choices of covariance function (Tsang,
Kong and Newton [1982], Mackay, Lakhtakia and Weiglhofer [2001b]), and most
notably for the step function Λ(r − r ) = Λstep (r − r ) wherein
)

f , |U −1 • (r − r )| L
Φ (r)Φ (r ) e = ( = a, b). (6.18)
f2 , |U −1 • (r − r )| > L
These studies have revealed that the form of the covariance function has only a
weak influence upon the SPFT estimate K[2]SPFT
(ω) (Mackay, Lakhtakia and Wei-
glhofer [2001b]).
After choosing the covariance function Λstep (r − r ), the principal value inte-
gral (6.17) may be expressed in the more tractable form
2π π ∞
f a fb
P Λstep
(ω) = Q(q, ω) sin θ dq dθ dφ, (6.19)
2π 2
φ=0 θ=0 q=0
with

0 −1
sin qL
Q(q, ω) = Ǧ comp U • q, ω − L cos qL , (6.20)
q
where q = q q̂ and q̂ = (sin θ cos φ, sin θ sin φ, cos θ ). Furthermore,
0 ∞
Ǧ comp (q, ω) = Ǧ comp (q, ω) − Ǧ comp (q̂, ω), (6.21)
with Ǧ comp (q, ω) being the spatial Fourier transform of the DGF for the compar-
∞
ison medium, as defined in eq. (5.38), which tends to Ǧ comp (q̂, ω) in the limit
q → ∞, as defined in eq. (5.53).
Some analytical progress has been reported towards the simplification of the in-
tegral (6.19) on an unbounded domain: for certain anisotropic (Genchev [1992],
Zhuck [1994]) and bianisotropic (Mackay, Lakhtakia and Weiglhofer [2000])
comparison mediums, a two-dimensional integral emerges from eq. (6.19), but
numerical methods are generally required to evaluate even that integral.
6.4.2.3. Third order To calculate the third-order (i.e., trilocally approximated)

mass operator term Σ [3] , a three-point covariance function is required (Miller
[1969], Milton and Phan-Thien [1982], Mackay, Lakhtakia and Weiglhofer
[2001c]). The two-point covariance function (6.18) incorporated into Λstep (r −r )
generalizes to yield
⎧ 3
⎪
⎪ fa , min{L12 , L13 , L23 } > L,
⎨
f , max{L12 , L13 , L23 } L,
θa (r)θa (r )θa (r ) = 1 a (6.22)
⎪
⎪ (f + 2fa3 ), one of L12 , L13 , L23 L,
⎩ 31 a
3 (2fa + fa ), two of L12 , L13 , L23 L,
3
wherein
⎫
L12 = U −1 • (r − r ) ⎪⎪
−1 ⎬

L13 = U •
(r − r ) . (6.23)
−1 ⎪
⎪
⎭

L23 = U •
(r − r )
Upon using eq. (6.22), the third-order mass operator term emerges as (Mackay,
Lakhtakia and Weiglhofer [2001c])
iω3 fa (1 − 2fa ) a/comp
Σ [3] (ω) = Σ [2] (ω) + α (ω)
3(1 − fa )2

• V(ω) • α a/comp (ω) • PΛstep (ω)
+ PΛstep (ω) • α a/comp (ω) • V(ω)

+ PΛstep (ω) • α a/comp (ω) • PΛstep (ω) • α a/comp (ω), (6.24)
in which
1 −1
V(ω) = K (ω) − DU/comp (ω). (6.25)
iω comp
Strictly, eq. (6.24) holds for isotropic chiral comparison mediums combined
with spherical constituent particles, but it is highly probable that it also holds
for weakly bianisotropic mediums with diagonally dominant 3 × 3 constitu-
tive dyadics. By implementing eq. (6.24), SPFT convergence at the level of
the bilocal approximation has been demonstrated for certain isotropic chiral
HCMs, as well as bianisotropic HCMs of the Faraday chiral type which are
both weakly uniaxial and weakly gyrotropic (Mackay, Lakhtakia and Weiglhofer
[2001c]).
6.5. Anisotropy and bianisotropy via homogenization
Manifestations of anisotropy and bianisotropy may be readily conceptualized

through the process of homogenization. HCMs ‘inherit’ the symmetries of their
constituent mediums. Therefore, complex HCMs can arise from relatively simple
constituent mediums.
3, § 7] Closing remarks 201
Biaxial HCMs have been comprehensively studied in this regard, in both

anisotropic (Mackay and Weiglhofer [2001b, 2000]) and bianisotropic (Mackay
and Weiglhofer [2001a]) scenarios. A biaxial HCM generally arises when the
constituent mediums present two, noncollinear, distinguished axes of symmetry.
These distinguished axes may be either of electromagnetic origin, as in uniaxial
constituent mediums for example, or they may derive from the particulate shapes
of the constituent mediums, as in spheroidal particles for example. Thus, a biax-
ial HCM – which is anisotropic or bianisotropic – may develop from constituent
mediums a and b wherein:
• both constituent mediums are uniaxial and distributed as spherical particles;
• both constituent mediums are isotropic and distributed as spheroidal particles;
• both constituent mediums are isotropic, with one medium distributed as spher-
ical particles and the other as ellipsoidal particles; or
• one of the constituent mediums is uniaxial while the other is isotropic, and one
of the constituent mediums is distributed as spheroidal particles while the other
as spherical particles.
If the distinguished axes of the constituent mediums are orthogonal, then or-
thorhombic biaxial HCMs generally develop; otherwise monoclinic or triclinic
biaxial HCMs arise.
A second prime example of bianisotropic HCMs arising from relatively sim-
ple constituents is provided by Faraday chiral mediums, as discussed in § 3.3.3.
These have been comprehensively investigated using the Maxwell Garnett and
Bruggeman formalisms (Weiglhofer, Lakhtakia and Michel [1998], Weiglhofer
and Mackay [2000]), as well as the SPFT (Mackay, Lakhtakia and Weiglhofer
[2000]).
§ 7. Closing remarks
The most general description of a linear medium in electromagnetics is bian-

isotropic, which is Lorentz-covariant. But bianisotropy is rarely discussed in
even graduate-level treatments of electromagnetics. Bianisotropy arises very nat-
urally when considering descriptions of a linear medium by uniformly moving
observers. Furthermore, bianisotropy may be observed in numerous naturally oc-
curring minerals when viewed from a co-moving reference frame. In relation to
technological applications, it may well transpire that the most significant mani-
festation of bianisotropy lies in homogenized composite mediums. Through the
process of homogenization, the vast parameter space of bianisotropy may be read-
ily accessed, thereby making possible the realization and exploitation of highly
complex electromagnetic behaviour.
202 Electromagnetic fields in linear bianisotropic mediums [3
References
Abramowitz, M., Stegun, I.A. (Eds.), 1965, Handbook of Mathematical Functions, Dover, New York,
NY, USA.
Agranovich, V.M., Ginzburg, V.L., 1984, Crystal Optics with Spatial Dispersion, and Excitons,
Springer, Berlin, Germany.
Altman, C., Suchy, K., 1991, Reciprocity, Spatial Mapping and Time Reversal in Electromagnetics,
Kluwer, Dordrecht, The Netherlands.
Arfken, G.B., Weber, H.J., 1995, Mathematical Methods for Physicists, fourth ed., Academic Press,
London, UK.
Ashcroft, N.W., Mermin, N.D., 1976, Solid State Physics, Saunders College, Philadelphia, PA, USA.
Aspnes, D.E., 1982, Local-field effects and effective-medium theory: A microscopic perspective, Am.
J. Phys. 50, 704–709.
Barrera, R.G., Fuchs, R., 1995, Theory of electron energy loss in a random system of spheres, Phys.
Rev. B 52, 3256–3273.
Berry, M.V., 2005, The optical singularities of bianisotropic crystals, Proc. R. Soc. Lond. A 461,
2071–2098.
Berthier, S., Peiro, J., 1998, Anomalous infra-red absorption of nanocermets in the percolation range,
J. Phys.: Condens. Matter 10, 3679–3694.
Bohren, C.F., Huffman, D.R., 1983, Absorption and Scattering of Light by Small Particles, Wiley,
New York, NY, USA.
Born, M., Wolf, E., 1980, Principles of Optics, sixth ed., Pergamon, Oxford, UK.
Brau, C.A., 2004, Modern Problems in Classical Electrodynamics, Oxford University Press, New
York, NY, USA.
Brouers, F., 1986, Percolation threshold and conductivity in metal–insulator composite mean-field
theories, J. Phys. C: Solid State Phys. 19, 7183–7193.
Bruggeman, D.A.G., 1935, Berechnung verschiedener physikalischer Konstanten von heterogenen
Substanzen, I. Dielektrizitätskonstanten und Leitfähigkeiten der Mischkörper aus isotropen Sub-
stanzen, Ann. Phys. Lpz. 24, 636–679.
Buchwald, J.Z., 1985, From Maxwell to Microphysics, University of Chicago Press, Chicago, IL,
USA.
Callen, H.B., Barasch, M.L., Jackson, J.L., 1952, Statistical mechanics of irreversibility, Phys. Rev. 88,
1382–1386.
Callen, H.B., Greene, R.F., 1952, On a theorem of irreversible thermodynamics, Phys. Rev. 86, 702–
710.
Casimir, H.B.G., 1945, On Onsager’s principle of microscopic reversibility, Rev. Mod. Phys. 17, 343–
350.
Censor, D., 1998, Electrodynamics, topsy–turvy special relativity, and generalized Minkowski consti-
tutive relations for linear and nonlinear systems, Prog. Electromagn. Res. PIER 18, 261–284.
Chambers, Ll.G., 1956, Propagation in a gyrational medium, Quart. J. Mech. Appl. Math. 9, 360–370.
Chandrasekhar, S., 1992, Liquid Crystals, second ed., Cambridge University Press, Cambridge, UK.
Chen, H.C., 1983, Theory of Electromagnetic Waves, McGraw–Hill, New York, NY, USA.
Cheng, D., 1997, Eigenfunction expansion of the dyadic Green’s function in a gyroelectric chiral
medium by cylindrical vector wave functions, Phys. Rev. E 55, 1950–1958.
Cheng, D., Ren, W., Jin, Y.-Q., 1997, Green dyadics in uniaxial bianisotropic-ferrite medium by cylin-
drical vector wavefunctions, J. Phys. A: Math. Gen. 30, 573–585.
Chew, W.C., 1999, Waves and Fields in Inhomogenous Media, IEEE Press, Piscataway, NJ, USA.
Collin, R.E., 1966, Foundations for Microwave Engineering, McGraw–Hill, New York, NY, USA.
Cottis, P.G., Kondylis, G.D., 1995, Properties of the dyadic Green’s function for an unbounded
anisotropic medium, IEEE Trans. Antennas Propagat. 43, 154–161.
3] References 203
Cottis, P.G., Vazouras, C.N., Spyrou, C., 1999, Green’s function for an unbounded biaxial medium in
cylindrical coordinates, IEEE Trans. Antennas Propagat. 47, 195–199.
Cui, J., Mackay, T.G., 2007, Depolarization regions of nonzero volume in bianisotropic homogenized
composites, Waves Random Complex Media 17, 269–281.
de Gennes, P.G., Prost, J.A., 1993, The Physics of Liquid Crystals, second ed., Clarendon Press,
Oxford, UK.
de Lange, O.L., Raab, R.E., 2001, Post’s constraint for electromagnetic constitutive relations, J. Opt.
A: Pure Appl. Opt. 3, L23–L26.
d’Inverno, R., 1992, Introducing Einstein’s Relativity, Clarendon Press, Oxford, UK.
Dolling, G., Wegener, M., Soukoulis, C.M., Linden, S., 2007, Negative-index metamaterial at 780 nm
wavelength, Opt. Lett. 32, 53–55.
Doyle, W.T., 1989, Optical properties of a suspension of metal spheres, Phys. Rev. B 39, 9852–9858.
Dungey, C.E., Bohren, C.F., 1991, Light scattering by nonspherical particles: A refinement to the
coupled-dipole method, J. Opt. Soc. Am. A 8, 81–87.
Engheta, N., Jaggard, D.L., Kowarz, M.W., 1992, Electromagnetic waves in Faraday chiral media,
IEEE Trans. Antennas Propagat. 40, 367–374.
Faxén, H., 1920, Der Zusammenhang zwischen den Maxwellschen Gleichungen für Dielektrika und
den atomistischen Ansätzen von H.A. Lorentz u.a., Z. Phys. 2, 218–229.
Felsen, L.B., Marcuvitz, N., 1994, Radiation and Scattering of Waves, IEEE Press, Piscataway, NJ,
USA.
Fricke, H., 1953, The Maxwell–Wagner dispersion in a suspension of ellipsoids, J. Phys. Chem. 57,
934–937.
Frisch, U., 1970, Wave propagation in random media, in: Bharucha-Reid, A.T. (Ed.), Probabilistic
Methods in Applied Mathematics, vol. 1, Academic Press, London, UK, pp. 75–198.
Fuchs, R., Barrera, R.G., Carrillo, J.L., 1996, Spectral representations of the electron energy loss in
composite media, Phys. Rev. B 54, 12824–12834.
Garland, J.C., Tanner, D.B., 1978, Electrical Transport and Optical Properties of Inhomogeneous Me-
dia, American Institute of Physics, New York, NY, USA.
Genchev, Z.D., 1992, Anisotropic and gyrotropic version of Polder and van Santen’s mixing formula,
Waves Random Media 2, 99–110.
Gerardin, J., Lakhtakia, A., 2001, Conditions for Voigt wave propagation in linear, homogeneous,
dielectric mediums, Optik 112, 493–495.
Gersten, J.I., Smith, F.W., 2001, The Physics and Chemistry of Materials, Wiley, New York, NY, USA.
Ghosh, K., Fuchs, R., 1991, Critical behaviour in the dielectric properties of random self-similar com-
posites, Phys. Rev. B 44, 7330–7343.
Goncharenko, A.V., 2003, Generalizations of the Bruggeman equation and a concept of shape-
distributed particle composites, Phys. Rev. E 68, 041108.
Goncharenko, A.V., Venger, E.F., 2004, Percolation threshold for Bruggeman composites, Phys. Rev.
E 70, 057102.
Gradshteyn, I.S., Ryzhik, I.M., 1980, Table of Integrals, Series, and Products, fourth ed., Academic
Press, London, UK.
Hanai, T., 1960, Theory of the dielectric dispersion due to the interfacial polarization and its applica-
tions to emulsions, Kolloid Z. 171, 23–31.
Hashin, Z., Shtrikman, S., 1962, A variational approach to the theory of the effective magnetic perme-
ability of multiphase materials, J. Appl. Phys. 33, 3125–3131.
Hehl, F.W., Obukhov, Yu.N., 2005, Linear media in classical electrodynamics and the Post constraint,
Phys. Lett. A 334, 249–259.
Hehl, F.W., Obukhov, Y.N., Rivera, J.-P., Schmid, H., in press. Relativistic analysis of magnetoelectric
crystals: Extracting a new 4-dimensional P odd and T odd pseudoscalar from Cr2 O3 data. Phys.
Lett. A, doi:10.1016/j.physleta.2007.08.069.
Hilgevoord, J., 1962, Dispersion Relations and Causal Descriptions, North-Holland, Amsterdam, The
Netherlands.
Hoffman, J.D., 1992, Numerical Methods for Engineers and Scientists, McGraw–Hill, New York,
USA.
Hu, L., Lin, Z., 2003, Imaging properties of uniaxially anisotropic negative refractive index materials,
Phys. Lett. A 313, 316–324.
Jackson, J.D., 1999, Classical Electrodynamics, third ed., Wiley, New York, NY, USA.
King, F.W., 2006, Alternative approach to the derivation of dispersion relations for optical constants,
J. Phys. A: Math. Gen. 39, 10427–10435.
Kong, J.A., 1972, Theorems of bianisotropic media, Proc. IEEE 60, 1036–1046.
Kong, J.A., 1986, Electromagnetic Wave Theory, Wiley, New York, NY, USA.
Krowne, C.M., 1984, Electromagnetic theorems for complex anisotropic media, IEEE Trans. Antennas
Propagat. 32, 1224–1230.
Lakhtakia, A., 1993, Frequency-dependent continuum electromagnetic properties of a gas of scattering
centers, Adv. Chem. Phys. 85 (2), 312–359.
Lakhtakia, A., 1994, Beltrami Fields in Chiral Media, World Scientific, Singapore.
Lakhtakia, A., 1995, Covariances and invariances of the Maxwell postulates, in: Barrett, T.W., Grimes,
D.M. (Eds.), Advanced Electromagnetism: Foundations, Theory and Applications, World Scien-
tific, Singapore, pp. 390–410.
Lakhtakia, A. (Ed.), 1996, Selected Papers on Linear Optical Composite Materials, SPIE Optical
Engineering Press, Bellingham, WA, USA.
Lakhtakia, A., 1998a, Anomalous axial propagation in helicoidal bianisotropic media, Opt. Com-
mun. 157, 193–201.
Lakhtakia, A., 1998b, Incremental Maxwell Garnett formalism for homogenizing particulate compos-
ite media, Microw. Opt. Technol. Lett. 17, 276–279.
Lakhtakia, A., 2000a, On direct and indirect scattering approaches for homogenization of particulate
composites, Microw. Opt. Technol. Lett. 25, 53–56.
Lakhtakia, A., 2000b, Orthogonal symmetries of polarizability dyadics of bianisotropic ellipsoids,
Microw. Opt. Technol. Lett. 27, 175–177.
Lakhtakia, A., 2004a, A conjugation symmetry in linear electromagnetism in extension of materials
with negative real permittivity and permeability scalars, Microw. Opt. Technol. Lett. 40, 160–161.
Lakhtakia, A., 2004b, On the genesis of Post constraint in modern electromagnetism, Optik 115, 151–
158.
Lakhtakia, A., 2006, Boundary-value problems and the validity of the Post constraint in modern elec-
tromagnetism, Optik 117, 188–192.
Lakhtakia, A., Depine, R.A., 2005, On Onsager relations and linear electromagnetic materials, Int. J.
Electron. Commun. (AEÜ) 59, 101–104.
Lakhtakia, A., Mackay, T.G., 2004a, Towards gravitationally assisted negative refraction of light by
vacuum, J. Phys. A: Math. Gen. 37, L505–L510; Corrigendum: J. Phys. A: Math. Gen. 37 (2004)
12093.
Lakhtakia, A., Mackay, T.G., 2004b, Infinite phase velocity as the boundary between positive and
negative phase velocities, Microw. Opt. Technol. Lett. 41, 165–166.
Lakhtakia, A., Mackay, T.G., 2005, Dyadic Green function for an elecromagnetic medium inspired by
general relativity, Chin. Phys. Lett. 23, 832–833.
Lakhtakia, A., Mackay, T.G., 2006, Negative phase velocity in isotropic dielectric–magnetic media
via homogenization: Part II, Microw. Opt. Technol. Lett. 48, 709–712.
Lakhtakia, A., Mackay, T.G., Setiawan, S., 2005, Global and local perspectives of gravitationally as-
sisted negative-phase-velocity propagation of electromagnetic waves in vacuum, Phys. Lett. A 336,
89–96.
3] References 205
Lakhtakia, A., McCall, M.W., Weiglhofer, W.S., 2003, Negative phase-velocity mediums, in: Weigl-
hofer, W.S., Lakhtakia, A. (Eds.), Introduction to Complex Mediums for Optics and Electromag-
netics, SPIE Press, Bellingham, WA, USA, pp. 347–363.
Lakhtakia, A., Messier, R., 2005, Sculptured Thin Films: Nanoengineered Morphology and Optics,
SPIE Press, Bellingham, WA, USA.
Lakhtakia, A., Shanker, B., 1993, Beltrami fields within continuous source regions, volume integral
equations, scattering algorithms, and the extended Maxwell-Garnett model, Int. J. Appl. Electro-
magn. Mater. 4, 65–82.
Lakhtakia, A., Varadan, V.K., Varadan, V.V., 1989, A note on Huygens’s principle for uniaxial dielec-
tric media, J. Wave–Mater. Interact. 4, 339–343.
Lakhtakia, A., Varadan, V.K., Varadan, V.V., 1991, Plane waves and canonical sources in a gyroelec-
tromagnetic uniaxial medium, Int. J. Electron. 71, 853–861.
Lakhtakia, A., Weiglhofer, W.S., 1995, On light propagation in helicoidal bianisotropic mediums,
Proc. R. Soc. Lond. A 448, 419–437; Corrections: Proc. R. Soc. Lond. A 454 (1998) 3275.
Lakhtakia, A., Weiglhofer, W.S., 1996a, Lorentz covariance, Occam’s razor, and a constraint on linear
constitutive relations, Phys. Lett. A 213, 107–111; Corrections: Phys. Lett. A 222 (1996) 459.
Lakhtakia, A., Weiglhofer, W.S., 1996b, Constraint on linear, spatiotemporally nonlocal, spatiotem-
porally nonhomogeneous constitutive relations, Int. J. Infrared Millim. Waves 17, 1867–1878.
Lakhtakia, A., Weiglhofer, W.S., 1997a, Further results on light propagation in helicoidal bianisotropic
mediums: Oblique propagation, Proc. R. Soc. Lond. A 453, 93–105; Corrections: Proc. R. Soc.
Lond. A 454 (1998) 3275.
Lakhtakia, A., Weiglhofer, W.S., 1997b, Green function for radiation and propagation in helicoidal
bianisotropic mediums, IEE Proc. Microw. Antennas Propag. 144, 57–59.
Landau, L.D., Lifshitz, E.M., 1975, The Classical Theory of Fields, Clarendon Press, Oxford, UK.
Lax, B., Button, K.J., 1962, Microwave Ferrites and Ferrimagnetics, McGraw–Hill, New York, NY,
USA.
Li, L.-W., Kang, X.-K., Leong, M.-S., 2001, Spheroidal Wave Functions in Electromagnetic Theory,
Wiley, Hoboken, NJ, USA.
Li, L.-W., Leong, M.-S., Kooi, P.-S., Yeo, T.-S., 1999, Comment on “Eigenfunction expansion of the
dyadic Green’s function in a gyroelectric chiral medium by cylindrical vector wave functions”,
Phys. Rev. E 59, 3767–3771.
Litchinitser, N.M., Gabitov, I.R., Maimistov, A.I., Shalaev, V.M., 2008, Negative refractive index
metamaterials in optics, in: Prog. Optics, vol. 51, pp. 1–67 (Chapter 1 in this volume).
Mackay, T.G., 2004, Depolarization volume and correlation length in the homogenization of
anisotropic dielectric composites, Waves Random Media 14, 485–498; Erratum: Waves Random
Complex Media 16 (2006) 85.
Mackay, T.G., Lakhtakia, A., 2003, Voigt wave propagation in biaxial composite materials, J. Opt. A:
Pure Appl. Opt. 5, 91–95.
Mackay, T.G., Lakhtakia, A., 2004, Plane waves with negative phase velocity in Faraday chiral medi-
ums, Phys. Rev. E 69, 026602.
Mackay, T.G., Lakhtakia, A., 2005, Anisotropic enhancement of group velocity in a homogenized
dielectric composite medium, J. Opt. A: Pure Appl. Opt. 7, 669–674.
Mackay, T.G., Lakhtakia, A., 2006, Correlation length and negative phase velocity in isotropic
dielectric–magnetic materials, J. Appl. Phys. 100, 063533.
Mackay, T.G., Lakhtakia, A., in press-a. Positive-, negative-, and orthogonal-phase-velocity propaga-
tion of electromagnetic plane waves in a simply moving medium: Reformulation and reappraisal.
Optik, doi:10.1016/j.ijleo.2007.07.004.
Mackay, T.G., Lakhtakia, A., in press-b. Response to ‘On negative refraction in classical vacuum’.
J. Mod. Optics.
Mackay, T.G., Lakhtakia, A., Setiawan, S., 2005a, Electromagnetic negative-phase-velocity propaga-
tion in the ergosphere of a rotating black hole, New J. Phys. 7, 171.
Mackay, T.G., Lakhtakia, A., Setiawan, S., 2005b, Electromagnetic waves with negative phase veloc-
ity in Schwarzschild–de Sitter spacetime, Europhys. Lett. 71, 925–931.
Mackay, T.G., Lakhtakia, A., Weiglhofer, W.S., 2000, Strong-property-fluctuation theory for homog-
enization of bianisotropic composites: Formulation, Phys. Rev. E 62, 6052–6064; Corrections:
Phys. Rev. E 63 (2001) 049901.
Mackay, T.G., Lakhtakia, A., Weiglhofer, W.S., 2001a, Ellipsoidal topology, orientation diversity and
correlation length in bianisotropic composite mediums, Int. J. Electron. Commun. (AEÜ) 55, 243–
251.
Mackay, T.G., Lakhtakia, A., Weiglhofer, W.S., 2001b, Homogenisation of similarly oriented, metal-
lic, ellipsoidal inclusions using the bilocal-approximated strong-property-fluctuation theory, Opt.
Commun. 197, 89–95.
Mackay, T.G., Lakhtakia, A., Weiglhofer, W.S., 2001c, Third-order implementation and convergence
of the strong-property-fluctuation theory in electromagnetic homogenisation, Phys. Rev. E 64,
066616.
Mackay, T.G., Weiglhofer, W.S., 2000, Homogenization of biaxial composite materials: Dissipative
anisotropic properties, J. Opt. A: Pure Appl. Opt. 2, 426–432.
Mackay, T.G., Weiglhofer, W.S., 2001a, Homogenization of biaxial composite materials: Bian-
isotropic properties, J. Opt. A: Pure Appl. Opt. 3, 45–52.
Mackay, T.G., Weiglhofer, W.S., 2001b, Homogenization of biaxial composite materials: Nondissipa-
tive dielectric properties, Electromagnetics 21, 15–26.
Mansuripur, M., 1995, The Physical Principles of Magneto–Optical Recording, Cambridge University
Press, Cambridge, UK.
Maxwell Garnett, J.C., 1904, Colours in metal glasses and in metallic films, Phil. Trans. R. Soc. Lond.
A 203, 385–420.
McCall, M.W., 2007, Classical gravity does not refract negatively, Phys. Rev. Lett. 98, 091102.
Michel, B., 1997, A Fourier space approach to the pointwise singularity of an anisotropic dielectric
medium, Int. J. Appl. Electromagn. Mech. 8, 219–227.
Michel, B., 2000, Recent developments in the homogenization of linear bianisotropic composite ma-
terials, in: Singh, O.N., Lakhtakia, A. (Eds.), Electromagnetic Fields in Unconventional Materials
and Structures, Wiley, New York, NY, USA, pp. 39–82.
Michel, B., Lakhtakia, A., 1995, Strong-property-fluctuation theory for homogenizing chiral particu-
late composites, Phys. Rev. E 51, 5701–5707.
Michel, B., Lakhtakia, A., Weiglhofer, W.S., 1998, Homogenization of linear bianisotropic particulate
composite media – Numerical studies, Int. J. Appl. Electromag. Mech. 9, 167–178; Corrections:
Int. J. Appl. Electromag. Mech. 10 (1999) 537–538.
Michel, B., Lakhtakia, A., Weiglhofer, W.S., Mackay, T.G., 2001, Incremental and differential
Maxwell Garnett formalisms for bi-anisotropic composites, Compos. Sci. Technol. 61, 13–18.
Michel, B., Weiglhofer, W.S., 1997, Pointwise singularity of dyadic Green function in a general bian-
isotropic medium, Int, J. Electron. Commun. (AEÜ) 51, 219–223; Corrections: J. Electron. Com-
mun. (AEÜ) 52 (1998) 310.
Miller, M.N., 1969, Bounds for effective electrical, thermal, and magnetic properties of heterogeneous
materials, J. Math. Phys. 10, 1988–2004.
Milton, G.W., 1985, The coherent potential approximation is a realizable effective medium scheme,
Commun. Math. Phys. 99, 463–500.
Milton, G.W., Phan-Thien, N., 1982, New bounds on effective elastic moduli of two-component ma-
terials, Proc. R. Soc. Lond. A 380, 305–331.
Monzon, J.C., 1989, Three-dimensional field expansion in the most general rotationally symmetric
anisotropic material: Application to scattering by a sphere, IEEE Trans. Antennas Propagat. 37,
728–735.
Monzon, J.C., 1990, Radiation and scattering in homogeneous general biisotropic regions, IEEE
Trans. Antennas Propagat. 38, 227–235.
3] References 207
Niklasson, G.A., Granqvist, C.G., 1984, Optical properties and solar selectivity of coevaporated Co–
Al2 O3 composite films, J. Appl. Phys. 55, 3382–3410.
Nye, J.F., 1985, Physical Properties of Crystals, Oxford University Press, Oxford, UK.
O’Dell, T.H., 1970, The Electrodynamics of Magneto–Electric Media, North-Holland, Amsterdam,
The Netherlands.
Ogg, N.R., 1971, A Huygen’s principle for anisotropic media, J. Phys. A: Math. Gen. 4, 382–388.
Olyslager, F., Lindell, I.V., 2002, Electromagnetics and exotic media: A quest for the holy grail, IEEE
Antennas Propagat. Mag. 44, 48–58.
Onsager, L., 1931a, Reciprocal relations in irreversible processes. I, Phys. Rev. 37, 405–426.
Onsager, L., 1931b, Reciprocal relations in irreversible processes. II, Phys. Rev. 38, 2265–2279.
Oseen, C.W., 1933, The theory of liquid crystals, J. Chem. Soc. Faraday Trans. II 29, 883–899.
Pancharatnam, S., 1958, Light propagation in absorbing crystals possessing optical activity – Electro-
magnetic theory, Proc. Ind. Acad. Sci. A 48, 227–244.
Pendry, J.B., 2004, Negative refraction, Contemp. Phys. 45, 191–202.
Plébanski, J., 1960, Electromagnetic waves in gravitational fields, Phys. Rev. 118, 1396–1408.
Ponti, S., Oldano, C., Becchi, M., 2001, Bloch wave approach to the optics of crystals, Phys. Rev.
E 64, 021704.
Post, E.J., 1997, Formal Structure of Electromagnetics, Dover, New York, NY, USA.
Priou, A. (Ed.), 1992, Dielectric Properties of Heterogeneous Materials, Elsevier, New York, NY,
USA.
Ramakrishna, S.A., 2005, Physics of negative refractive index materials, Rep. Progr. Phys. 68, 449–
521.
Ren, W., 1999, Comment on “Eigenfunction expansion of the dyadic Green’s function in a gyroelectric
chiral medium by cylindrical vector wave functions”, Phys. Rev. E 59, 3772–3773.
Ryzhov, Yu.A., Tamoikin, V.V., 1970, Radiation and propagation of electromagnetic waves in ran-
domly inhomogeneous media, Radiophys. Quantum Electron. 14, 228–233.
Schleich, W., Scully, M.O., 1984, General relativity and modern optics, in: Grynberg, G., Stora, R.
(Eds.), New Trends in Atomic Physics, Elsevier, Amsterdam, Holland, pp. 995–1124.
Schmid, H., 2003, Magnetoelectric effects in insulating magnetic materials, in: Weiglhofer, W.S.,
Lakhtakia, A. (Eds.), Introduction to Complex Mediums for Optics and Electromagnetics, SPIE
Press, Bellingham, WA, USA, pp. 167–195.
Shalaev, V.M., 2007, Optical negative-index metamaterials, Nature Photonics 1, 41–48.
Shanker, B., 1996, The extended Bruggeman approach for chiral-in-chiral mixtures, J. Phys. D: Appl.
Phys. 29, 281–288.
Shanker, B., 1997, A comment on ‘Effective medium formulae for bi-anisotropic mixures’, J. Phys.
D: Appl. Phys. 30, 289–290.
Shanker, B., Lakhtakia, A., 1993, Extended Maxwell Garnett model for chiral-in-chiral composites,
J. Phys. D: Appl. Phys. 26, 1746–1758.
Shelby, R.A., Smith, D.R., Schultz, S., 2001, Experimental verification of a negative index of refrac-
tion, Science 292, 77–79.
Shen, N.-H., Win, Q., Chen, J., Fan, Y.-X., Ding, J., Wang, H.-T., Tian, Y., Ming, N.-B., 2005, Opti-
cally uniaxial left-handed materials, Phys. Rev. B 72, 153104.
Sherwin, J.A., Lakhtakia, A., 2002, Bragg–Pippard formalism for bianisotropic particulate compos-
ites, Microw. Opt. Technol. Lett. 33, 40–44.
Sihvola, A.H., Lindell, I.V., 1990, Chiral Maxwell-Garnett mixing formula, Electronics Lett. 26, 118–
119.
Sihvola, A.H., Pekonen, O.P.M., 1996, Effective medium formulae for bi-isotropic mixtures, J. Phys.
D: Appl. Phys. 29, 514–521.
Sihvola, A.H., Pekonen, O.P.M., 1997, Six-vector mixing formulae defended. Reply to comment on
‘Effective medium formulae for bi-isotropic mixtures’, J. Phys. D: Appl. Phys. 30, 291–292.
Skrotskii, G.V., 1957, The influence of gravitation on the propagation of light, Soviet Phys. Dokl. 2,
226–229.
Sosa, I.O., Mendoza, C.I., Barrera, R.G., 2001, Calculation of electron-energy-loss spectra of com-
posites and self-similar structures, Phys. Rev. E 63, 144201.
Stogryn, A., 1983, The bilocal approximation for the electric field in strong fluctuation theory, IEEE
Trans. Antennas Propagat. 31, 985–986.
Tai, C.T., 1994, Dyadic Green Functions in Electromagnetic Theory, second ed., IEEE Press, Piscat-
away, NJ, USA.
Tan, E.L., Tan, S.Y., 1998, On the eigenfunction expansions of dyadic Green’s functions for bian-
isotropic media, Prog. Electromagn. Res. 20, 227–247.
Tatarskii, V.I., Zavorotnyi, V.U., 1980, Strong fluctuations in light propagation in a randomly inhomo-
geneous medium, Prog. Optics 18, 204–256.
Tsang, L., Kong, J.A., 1981, Scattering of electromagnetic waves from random media with strong
permittivity fluctuations, Radio Sci. 16, 303–320.
Tsang, L., Kong, J.A., Newton, R.W., 1982, Application of strong fluctuation random medium the-
ory to scattering of electromagnetic waves from a half-space of dielectric mixture, IEEE Trans.
Antennas Propagat. 30, 292–302.
Van Bladel, J., 1991, Singular Electromagnetic Fields and Sources, Oxford University Press, Oxford,
UK.
Van Kranendonk, J., Sipe, J.E., 1977, Foundations of the macroscopic electromagnetic theory of di-
electric media, Prog. Optics 15, 245–350.
Venugopal, V.C., Lakhtakia, A., 2000, Sculptured thin films: Conception, optical properties and appli-
cations, in: Singh, O.N., Lakhtakia, A. (Eds.), Electromagnetic Fields in Unconventional Materials
and Structures, Wiley, New York, NY, USA, pp. 151–216.
Voigt, W., 1902, On the behaviour of pleochroitic crystals along directions in the neighbourhood of an
optic axis, Phil. Mag. 4, 90–97.
Walser, R.M., 2003, Metamaterials: An introduction, in: Weiglhofer, W.S., Lakhtakia, A. (Eds.), In-
troduction to Complex Mediums for Optics and Electromagnetics, SPIE Press, Bellingham, WA,
USA, pp. 295–316.
Ward, L., 2000, The Optical Constants of Bulk Materials and Films, second ed., Institute of Physics,
Bristol, UK.
Weiglhofer, W.S., 1990, Dyadic Green’s functions for general uniaxial media, IEE Proc., Part H 137,
5–10.
Weiglhofer, W.S., 1993, Analytic methods and free-space dyadic Green’s functions, Radio Sci. 28,
847–857.
Weiglhofer, W.S., 1994, A dyadic Green’s functions representation in electrically gyrotropic media,
Int. J. Electron. Commun. (AEÜ) 48, 125–130.
Weiglhofer, W.S., 1995, Frequency-dependent dyadic Green functions for bianisotropic media, in:
Barrett, T.W., Grimes, D.M. (Eds.), Advanced Electromagnetism: Foundations, Theory and Ap-
plications, World Scientific, Singapore, pp. 376–389.
Weiglhofer, W.S., 1998a, A perspective on bianisotropy and Bianisotropics ’97, Int. J. Appl. Electro-
magn. Mech. 9, 93–101.
Weiglhofer, W.S., 1998b, Electromagnetic depolarization dyadics and elliptic integrals, J. Phys. A:
Math. Gen. 31, 7191–7196.
Weiglhofer, W.S., 2000, The connection between factorization properties and closed-form solutions
of certain linear dyadic differential operators, J. Phys. A: Math. Gen. 33, 6253–6261.
Weiglhofer, W.S., 2003, Constitutive characterization of simple and complex mediums, in: Weiglhofer,
W.S., Lakhtakia, A. (Eds.), Introduction to Complex Mediums for Optics and Electromagnetics,
SPIE Press, Bellingham, WA, USA, pp. 27–61.
Weiglhofer, W.S., Lakhtakia, A., 1996, On causality requirements for material media, Int. J. Electron.
Commun. (AEÜ) 50, 389–391.
3] References 209
Weiglhofer, W.S., Lakhtakia, A., 1998, The correct constitutive relations of chiroplasmas and chiro-
ferrites, Microw. Opt. Technol. Lett. 17, 405–408.
Weiglhofer, W.S., Lakhtakia, A., 1999, On electromagnetic waves in biaxial bianisotropic media, Elec-
tromagnetics 19, 351–362.
Weiglhofer, W.S., Lakhtakia, A., Michel, B., 1997, Maxwell Garnett and Bruggeman formalisms for a
particulate composite with bianisotropic host medium, Microw. Opt. Technol. Lett. 15, 263–266;
Corrections: Microw. Opt. Technol. Lett. 22 (1999) 221.
Weiglhofer, W.S., Lakhtakia, A., Michel, B., 1998, On the constitutive parameters of a chiroferrite
composite medium, Microw. Opt. Technol. Lett. 18, 342–345.
Weiglhofer, W.S., Lindell, I.V., 1994, Analytic solution for the dyadic Green function of a nonrecip-
rocal uniaxial bianisotropic medium, Int. J. Electron. Commun. (AEÜ) 48, 116–119.
Weiglhofer, W.S., Mackay, T.G., 2000, Numerical studies of the constitutive parameters of a chiro-
plasma composite medium, Int. J. Electron. Commun. (AEÜ) 54, 259–265.
Weiglhofer, W.S., Mackay, T.G., 2002, Needles and pillboxes in anisotropic mediums, IEEE Trans.
Antennas Propagat. 50, 85–86.
Wu, X.B., Yasumoto, K., 1997, Cylindrical vector-wave-function representations of fields in a biaxial
Ω-medium, J. Electromagn. Waves Applics. 11, 1407–1423.
Yakubovich, V.A., Starzhinskii, V.M., 1975, Linear Differential Equations with Periodic Coefficients,
Wiley, New York, NY, USA.
Zhuck, N.P., 1994, Strong-fluctuation theory for a mean electromagnetic field in a statistically homo-
geneous random medium with arbitrary anisotropy of electrical and statistical properties, Phys.
Rev. B 50, 15636–15645.

Mackay 2008

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Mackay 2008

Uploaded by

Copyright:

Available Formats

E.

Wolf, Progress in Optics 51

Electromagnetic fields in linear bianisotropic

ISSN: 0079-6638 DOI: 10.1016/S0079-6638(07)51003-6

A broad overview of electromagnetic bianisotropy is provided in the follow-

§ 2. The Maxwell postulates and constitutive relations

The action of complex assemblies of matter on electromagnetic waves lies at the

2.1. Maxwell postulates

but are incorporated for mathematical convenience (Lakhtakia [1995]). In con-

2.2. Constitutive relations

2.3. The frequency domain

The convolution integrals appearing in the constitutive relations (2.6) result in

wherein the invertibility of ν EB (r, ω) and μ (r, ω) has been assumed. We

2.4. 6-vector/6 × 6 dyadic notation

2.5. Form invariances

2.5.1. Time reversal

of (2.23), (2.24) and (2.25) are

2.5.2. Spatial inversion

P ζ (r, ω) = −ζ (−r, ω) . (2.31)

2.5.3. Lorentz covariance

Suppose an inertial reference frame Σ  moves with constant velocity v = v v̂

relates the space–time coordinates (r  , t  ) in Σ  to the space–time coordinates

2.5.4. Chiral invariance

As well as being form-invariant under spatial, temporal and spatiotemporal trans-

change under the following transformation of fields (Lakhtakia [1995])

and source densities

Here ψ is a complex-valued angle and the scalar Z is an impedance re-

2.5.5. Conjugate invariance

2.5.6. Energy and momentum

T W̃ (r, t) = W̃ (r, −t) (2.44)

P W̃ (r, t) = W̃ (−r, t) . (2.45)

Rψ W̃ (r, t) = W̃ (r, t) . (2.46)

2.6. Constitutive dyadics

6 × 6 constitutive dyadic K EH (r, ω) comprises 36 complex-valued parameters.

1 The Dirac delta function is defined in (5.5).

Finally, since f˜(r, t) is real-valued, the symmetry condition

f (r, −ω) = f ∗ (r, ω) (2.53)

relates f to its complex conjugate f ∗ . Thus, (2.52) yield the Kramers–Kronig

2.6.1.2. Post constraint A structural constraint, known as the Post constraint, is

tr μ−1 (r, ω) • ζ (r, ω) + ξ (r, ω) = 0. (2.56)

[1991]). It is often described in terms of the interchangeability of transmitters and

a, b = J e (r, ω) • E b (r, ω) − J am (r, ω) • H b (r, ω) d3 r, (2.60)

b, a . If the medium which supports J a,b

+ H a (r, ω) • ζ (r, ω) + ξ T (r, ω) • E b (r, ω) d3 r, (2.62)

2.6.2.2. Neglect of dissipation By virtue of the principle of causality, no

S(r, ω) t = Re E(r, ω) × H ∗ (r, ω) (2.65)

+ H ∗ (r, ω) • ζ (r, ω) − ξ + (r, ω) • E(r, ω) ,

The notions of anisotropy and bianisotropy may be simply characterized in

3.1.1. Free space

3.1.2. Dielectric–magnetic mediums

3.1.3. Isotropic chirality and biisotropy

constitutive characterizations for isotropic mediums, the constitutive properties of

3.2.1. Uniaxial anisotropy

uni (ω) = (ω)(I − ûû) + u (ω)ûû. (3.8)

The permittivity scalars and u are complex-valued quantities; causality requires

Similarly, a uniaxial, homogeneous magnetic medium may be described by the

μ (ω) = μ(ω)(I − ûû) + μu (ω)ûû. (3.11)

3.2.2. Biaxial anisotropy

and b are necessarily complex-valued, in which case Re{a , b } and Im{a , b }

where the 3 × 3 permittivity dyadics have the symmetric forms

describe orthorhombic, monoclinic and triclinic biaxial magnetic mediums, re-

The uniaxial and biaxial manifestations of anisotropy discussed in §§ 3.2.1

permeability dyadics. Uniaxial and biaxial mediums are therefore Lorentz-reci-

By straightforward dyadic manipulations, eq. (3.26) may be expressed in the

Suppose an inertial reference frame Σ moves with constant velocity v = v v̂

relates the space–time coordinates (r , t ) in Σ to the space–time coordinates

b, a. If the medium which supports J a,b

S(r, ω) t = Re E 0 (ω) k ∗ − E 0 (ω) • k ∗ E ∗0 (ω) .

F (z, ω) = exp(iA z)F (0, ω). (4.60)