Progress in Electromagnetics Research B, Vol. 18, 165-183, 2009

Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

Progress In Electromagnetics Research B, Vol.

18, 165–183, 2009

ELECTROMAGNETIC SOURCE TRANSFORMATIONS


AND SCALARIZATION IN STRATIFIED GYROTROPIC
MEDIA

P. De Visschere
Ghent University
ELIS, Sint-Pietersnieuwstraat 41, Gent B-9000, Belgium

Abstract—It is known that with restrictions on the type of the


constitutive equations, Maxwell’s equations in non-uniform media can
sometimes be reduced to two 2nd order differential equations for 2
scalar quantities only. These results have previously been obtained
in two quite different ways, either by a “scalarization of the sources”,
where the relevant scalar quantities are essentially vector potential
components, and the derivation was limited to isotropic media, or
alternatively by using the “scalar Hertz potentials”, and this method
has been applied to more general media. In this paper, it is shown
that both methods are equivalent for gyrotropic media. We show that
the scalarization can be obtained by a combination of transformations
between electric and magnetic sources and gauge transformations.
It is shown that the method based on the vector potential, which
previously used a non-traditional definition of the vector potentials,
can also be obtained using the traditional definition provided a proper
gauge condition is applied, and this method is then extended from
isotropic to gyrotropic media. It is shown that the 2 basic scalar Hertz
potentials occurring in the second method are invariant under the
source scalarization transformations of the first method and therefore
are the natural potentials for obtaining scalarization. Finally, it is
shown that both methods are also equivalent with a much older third
method based on Hertz vectors.

1. INTRODUCTION

Usually problems in electromagnetics are reduced to solving a 2nd


order vectorial equation for either electric field or magnetic field [1, 2].
However for some problems, the radiation of a dipole in a stratified
Corresponding author: P. De Visschere (pdv@elis.ugent.be).
166 De Visschere

medium being a good example, it is still useful to use a representation


for the fields in terms of auxiliary functions (vector and scalar
potentials or Hertz vector potentials) and then solving equations
for these auxiliary functions instead of for the fields directly. It is
well-known that for simple uniform electromagnetic media, Maxwell’s
equations can in this way be reduced to 2 scalar uncoupled 2nd
order differential equations, corresponding to the TE/TM modes.
For uniform media this has been generalized to more complex
(decomposable) media [3]. Also for non-uniform, usually stratified,
media such a scalar decomposition has been obtained.
One of the earliest systematic treatments of this “reduction
problem” was given by Nisbet [4, 5]. First, a general formulation in
terms of 2 Hertz vector potentials was given validly for non-uniform
anisotropic media, and then it was noticed that considerable freedom
exists for choosing these potentials since they can be subjected to a
gauge transformation. Using this freedom the Hertz vectors could
be reduced to single component vectors, and conditions were derived
under which the resulting differential equations for these components
were of 2nd order, at least for isotropic media. The same idea was
later extended to anisotropic media [6, 7]. The reduction to two scalar
potentials was also extended to gyrotropic media [8, 9] and to even
more complicated media [10]. Only rather recently the case of a
uniaxial medium was given explicitly [11]. Whereas in these earlier
publications [5, 7] a general coordinate system was considered, the
later extensions to more complex media usually considered a cartesian
coordinate system only, and the non-uniformity was limited to a
stratification along e.g., the symmetry axis of the uniaxial medium [11].
Although the 2 scalar potentials are referred to as “scalar Hertz
potentials” the link with the Hertz vector potentials is not obvious
anymore, and for the stratified uniaxial medium the (initially 4) scalar
functions are instead defined by applying a Helmholtz-decomposition
to the electric and magnetic field components perpendicular to the
symmetry axis [11].
Subsequently Weiglhofer and Georgieva [12, 13] arrived at the
same scalar equations, at least for isotropic media, following a
completely different method, which rests mainly on the so-called
scalarization of sources. It was shown that arbitrary current density
distributions can be replaced by equivalent distributions but oriented
along a fixed direction. With a proper (unconventional) choice of
the vector potentials the latter could then also be scalarized, and the
resulting equations are exactly the same as those found using the scalar
Hertz potentials. This remarkable correspondence was noticed, but no
explanation was given [13].
Progress In Electromagnetics Research B, Vol. 18, 2009 167

The main purpose of this paper is to shed some light on this


finding, which can not be a coincidence. At the same time, we
will extend the source scalarization method explained in [12, 13]
to more general (gyrotropic) anisotropic media. We believe that
“source scalarization” can be understood best as an application of
the well-known equivalence between electric and magnetic charges and
currents. This is first presented in § 2. In the following sections the
scalarization problem is solved using different potentials. As in most
of the referenced papers we consider only cartesian coordinates. A
stratified initially uniaxial medium is considered where the symmetry
axis is perpendicular to the layers everywhere. The latter condition
is necessary to avoid mixing between the longitudinal and transverse
field components by applying the constitutive equations. In § 6 the
theory is extended to a stratified gyrotropic medium. We will use c
as a unit vector along the symmetry axis and c as the corresponding
coordinate whereas transversal vector components will be labeled by
⊥; in particular, the transversal nabla operator will be written as ∇⊥ .

2. SOURCE TRANSFORMATIONS

We use Maxwell’s equations in the standard form including electric and


magnetic charge and current densities, where the latter are labeled by
a superscript star
∂B ∗
∇×E = − −J (1)
∂t
∂D
∇×H = +J (2)
∂t
∇·D = ρ (3)
∇ · B = ρ∗ (4)
The possibly position dependent constitutive properties of the medium
are given by D = ² · E and B = µ · H. We will write the electric charge
and current densities in general as
∂p
J = +∇×m (5)
∂t
ρ = −∇ · p (6)
where p and m either are given polarization and magnetization
densities or must be considered as stream potentials for given ρ and
J [5]. In either case p and m can be subjected to a gauge transformation
168 De Visschere

which leaves ρ and J invariant [5]


p0 = p + ∇ × G (7)
∂G
m0 = m − + ∇g (8)
∂t
where G, g are arbitrary functions. Since our aim is to transform
transversal sources into longitudinal ones, 2 possibilities arise. With
∇⊥ g = −m⊥ a transversal magnetization is turned into a longitudinal
one ∂g∂c c. And with a longitudinal G = Gc c a transversal polarization
p⊥ = −∇⊥ Gc × c is turned into a longitudinal magnetization − ∂G ∂t c.
c

It is well-known that the polarization p and the magnetization m can


equally well be represented by magnetic charge and current densities,
which is most easily seen by rearranging the terms in the Maxwell
equations as follows
¡ ¢ ∂ ¡ ¢ ∂µ · m
∇ × E + ²−1 · p = − B−µ·m − + ∇ × ²−1 · p (9)
∂t ∂t
¡ ¢ ∂ ¡ ¢
∇× H −m = D+p (10)
¡ ¢ ∂t
∇· D+p = 0 (11)
¡ ¢
∇ · B − µ · m = −∇ · (µ · m) (12)
The equivalent magnetic sources are thus given by
∗ ∂µ · m
J = − ∇ × ²−1 · p (13)
∂t
ρ∗ = −∇ · (µ · m) (14)
Contrary to the gauge transformations (7) and (8), in this case the
source transformation (from electrical charges to magnetic charges) is
accompanied by the following field transformations
0
E = E + ²−1 · p (15)
0
B = B−µ·m (16)
In what follows we will label a polarization/magnetization density
which is represented by magnetic charges by a superscript star (p∗ , m∗ ).
We can thus freely exchange p (or m) for p∗ (or m∗ ) and vice versa as
long as p + p∗ (or m + m∗ ) remains invariant, and we take the field
transformations (15) and (16) into account. These “magnetic” stream
potentials can also be subjected to a gauge transformation [5]
∂L
²−1 · p0∗ = ²−1 · p∗ − + ∇l (17)
∂t
µ · m0∗ = µ · m∗ − ∇ × L (18)
Progress In Electromagnetics Research B, Vol. 18, 2009 169

Table 1. The equivalent source contributions due to (external)


polarization and magnetization. The starred quantities allow to make
a distinction between sources modeled with electric charges/currents
and those modeled with magnetic ones.
polarization magnetization
∂p
J ∂t ∇×m
ρ −∇ · p
∗ ∂µ·m∗
J −∇ × ²−1 · p∗ ∂t
ρ∗ −∇ · µ · m∗

for arbitrary L, l. With l we can turn a transversal polarization


into a longitudinal one, and with L = Lc c we can turn a transversal
magnetization into a longitudinal polarization. For further reference
the different representations are tabulated in Table 1.
Using the electric/magnetic charge transformations and gauge
transformations if needed we can now scalarize an arbitrary current
density† along a fixed direction defined by the unit vector c. The
goal of the scalarization process is to replace the current density by
equivalent electric and magnetic current densities parallel to c. For a
stratified medium, c is perpendicular to the layers, and as mentioned
in the “Introduction” this is also the direction of the symmetry axis of
the uniaxial medium. The current density can always be written as
J = Jc c + ∇⊥ v × c + ∇⊥ u (19)
In the context of scalar Hertz potentials the functions u, v are known
as auxiliary functions [11], and they can be found by solving the 2-
dimensional potential problems
∇2⊥ u = ∇⊥ · J ⊥ (20)
2
¡ ¢
∇⊥ v = −c · ∇⊥ × J ⊥ (21)
Scalarization of the 2nd term in (19) is straightforward, since according
to Table 1 it can be attributed to a magnetization m = vc which can
also be represented by a magnetic current density along c
∗ ∂µ · m∗ ∂v
J = = µ// c (22)
∂t ∂t
where µ// is the permeability along c. This transformation is
accompanied by a field transformation according to (16)
0
B = B − µ// vc (23)
†We will consider an electric current density, but the same method can be applied to a
magnetic current density.
170 De Visschere

Scalarization of the last contribution in (19) cannot be obtained


simply by transforming electric into magnetic sources. However, this
transformation can always be combined with a gauge transformation.
The last contribution in (19) can then be scalarized by attributing
R the
current density ∇⊥ u to a transversal polarization density p = ∇⊥ udt
which, as we have seen, can be turned into a longitudinal one in the
“magnetic” domain by a proper choice of l in (17) (with L = 0) namely
Z
u
l=− dt (24)
²⊥
where ∇⊥ ²⊥ = 0 has been used. We then end up with a scalarized
current density µ ¶
0 ∂ u
Jc = −²// (25)
∂c ²⊥
Due to the 2 electric/magnetic transformations preceding and
following the gauge transformation we must take into account of a
transformation of the electric field according to (15)
µZ ¶
0 u
E − E = ∇l = ∇ dt (26)
²⊥
These results as well as similar ones for the magnetic current density
are tabulated in Tables 2 and 3. To conclude this section we make 2
remarks
(i) It can be proved that the scalarizations summarized in Table 2
and Table 3 are unique;
(ii) Under these source transformations the total longitudinal current
densities (Jc + ∂D ∗ ∂Bc
∂t and Jc + ∂t ) are invariant, since
c

∂Dc ¡ ¢
Jc + = ∇×H ·c (27)
∂t
∂Bc ¡ ¢
Jc∗ + = − ∇×E ·c (28)
∂t
0 0
and H − H and E − E are either parallel to c or equal to a
gradient.

3. VECTOR POTENTIALS

We introduce conventional vector and scalar potentials A, φ and also



comparable potentials A , φ∗ for handling the magnetic sources

∂A
B = ∇×A−µ· − µ · ∇φ∗ (29)
∂t
∂A ∗
E = − − ∇φ − ²−1 · ∇ × A (30)
∂t
Progress In Electromagnetics Research B, Vol. 18, 2009 171

Table 2. Using proper transformations between electric and


magnetic charge representations from Table 1 and possibly gauge
transformations of the stream potentials, arbitrary transverse electric
current densities can be replaced by equivalent electric or magnetic
current densities along the symmetry direction c. Each transformation
is also accompanied by a transformation of the fields, shown in the last
2 rows.
initial current density J = ∇⊥ v × c J = ∇⊥ u
R
initial stream potential m = vc p = ∇⊥ udt

R
final stream potential m∗ = vc p = −²// ∂c ( ²u⊥ dt)c

final current density J = µ// ∂v
∂t c J = −²// ∂c∂ u
( )c
0 R u ²⊥
E −E ∇( ²⊥ dt)
0
B −B −µ// vc

Table 3. Using proper transformations between electric and


magnetic charge representations from Table 1 and possibly gauge
transformations of the stream potentials, arbitrary transverse magnetic
current densities can be replaced by equivalent electric or magnetic
current densities along the symmetry direction c. Each transformation
is also accompanied by a transformation of the fields, shown in the last
2 rows.
∗ ∗
initial current density J = ∇⊥ v ∗ × c J = ∇⊥ u∗
R ∗
initial stream potential p∗ = −²// v ∗ c m = ∇⊥ µu⊥ dt


R ∗
final stream potential p = −²// v ∗ c m∗ = − ∂c ( µu⊥ dt)c
∗ ∗ ∂ u∗
final current density J = −²// ∂v
∂t c J = −µ// ∂c ( µ⊥ )c
0
E −E v∗ c
0 R u∗
B −B µ · ∇( µ⊥ dt)

Substitution into Maxwell’s curl-equations gives initially


∂φ
L(², µ)A + ² · ∇ = J (31)
∂t
∗ ∂φ∗ ∗
L(µ, ²)A + µ · ∇ = J (32)
∂t
where the operator L(², µ) is defined by
∂2
L(², µ) = ² · + ∇ × µ−1 · ∇× (33)
∂t2
172 De Visschere

The scalar potentials are eliminated using gauge conditions. If these


gauge conditions do not mix-up the electric and magnetic quantities
then the resulting equations will also remain uncoupled. We first
mention the gauge conditions used by Nisbet [5]
∂φ
∇·²·A+α = 0 (34)
∂t
∗ ∂φ∗
∇ · µ · A + α∗ = 0 (35)
∂t
where α, α∗ are scalars which can still be chosen. As will become
clear, these gauge conditions allow scalarization only with additional
restrictions on the position dependence of the material parameters.
This restriction is eliminated in [12, 13] but only for an isotropic
medium by using a different definition for the potentials and using
different gauge conditions. At first, this leads to equations for A and

A which are coupled but under the restrictions ∇⊥ ² = 0 and ∇⊥ µ = 0
these equations become uncoupled and scalarizable. Although this
method can be extended to uniaxial media we prefer to stick to the
conventional decompositions (29) and (30), and we will now derive
gauge conditions which allow to scalarize (31) and (32) for a non-
uniform uniaxial medium.
We assume that the sources have already been scalarized so that
in (31) and (32) only longitudinal current densities Jc , Jc∗ occur.
Splitting these equations into longitudinal and transversal components
the latter equations will only allow the null solution for the transversal
components of the vector potentials if their longitudinal components
do not occur in these transversal equations. These conditions are easily
found by assuming A = Ac c and equating the transversal component
of the LHS of (31) to zero
· µ ¶ ¸
∂ Ac ∂φ
∇⊥ + ²⊥ =0 (36)
∂c µ⊥ ∂t
and a similar “magnetic” equation, which is obtained by replacing
unstarred quantities by starred ones and by switching the roles of ²
and µ. Using the gauge condition (34), this condition can only be met
if ²⊥ , µ⊥ do not depend on the longitudinal coordinate c or if ε// µ⊥ is
independent of c, and then only by choosing α = ²// ²⊥ µ⊥ . For a more
general result we must instead choose the following gauge conditions
∂φ
∇ · µ−1
⊥ A + ²⊥ = 0 (37)
∂t
∗ ∂φ∗
∇ · ²−1
⊥ A + µ⊥ = 0 (38)
∂t
Progress In Electromagnetics Research B, Vol. 18, 2009 173

From the longitudinal components of (31), (32) and these gauge


conditions, we then find the final scalarized equations
µ 0 ¶
Ac
Ls (², µ) = −J 0 c (39)
µ⊥
µ 0∗ ¶
Ac ∗
Ls (µ, ²) = −J 0 c (40)
²⊥
where scalarized quantities are now explicitly marked by an accent
with the (scalar) wave operator defined by
∂ ∂ ∂2
Ls (², µ) = ∇2⊥ + ²// ²−1 − ² // µ⊥ (41)
∂c ⊥ ∂c ∂t2
Taking into account of (29), (30) and the field transformations due to
the prior source scalarization, the fields are given by
µ ¶ µ Z · µ ¶¸ ¶
∂A0 ∗c −1 0 −1 ∗ ∂ A0 ∗c
H = v− c+µ⊥ ∇⊥ A c ×c−∇ µ⊥ u − dt (42)
∂t ∂c ²⊥
µ 0
¶ µ Z · µ ¶¸ ¶
∗ ∂A c −1 0∗ −1 ∂ A0 c
E =− v + c−²⊥ ∇⊥ A c ×c−∇ ²⊥ u− dt (43)
∂t ∂c µ⊥

4. HERTZ VECTORS

The main idea behind the use of Hertz vectors is to introduce


an additional differentiation in such a way that the new poten-
tials (= Hertz vectors) are governed by equations with the polariza-
tion/magnetization (or the stream potentials) as sources instead of the
current densities. Since for such Hertz vectors the starred/unstarred
stream potentials are equivalent we can eliminate e.g., the “magnetic
charge” sources from the start and use instead of (29), (30) the simpler
decompositions
B + µ · m∗ = ∇ × A (44)
∂A
E − ²−1 · p∗ = − − ∇φ (45)
∂t
still complying with (1) and (4). The Hertz vector equations are
obtained most easily in the temporal gauge [14] (φ = 0), and we then
define the Hertz vectors Πe , Πm following [5]
∂Πe
A= + ²−1 · ∇ × Πm (46)
∂t
After substituting these equations in (2), (3) and following a standard
procedure [15], we find the equations
L(², µ)Πe = p + p∗ + ² · ∇ψe (47)

L(µ, ²)Πm = µ · (m + m ) + µ · ∇ψm (48)
174 De Visschere

where the operator L(², µ) has been defined in (33). The functions
ψe , ψm can be chosen arbitrarily and are in fact redundant due to
the gauge transformations (7), (8), (17), and (18). However, an extra
term ∇ψe should now be added to the RHS of (45). Substituting (46)
in (44) and (45) and replacing the 2nd order time derivative using (47)
and (33) we find the following symmetric expressions for the fields
∂Πe
B + µ · m∗ = ∇ × + ∇ × ²−1 · ∇ × Πm (49)
∂t
∂Πm
² · E + p = ∇ × µ−1 · ∇ × Πe − ∇ × (50)
∂t
Using gauge transformations (7), (8), (17), and (18), one can try to
simplify equations (47) and (48). One possibility is to eliminate the
magnetizations (m + m∗ → 0) so that also Πm → 0. This is the
essence of the 1-Hertz-vector method followed by Sein [16]. In this
case it is limited to a uniform isotropic medium and therefore using
the conventional Lorentz gauge. Another possibility is to scalarize
the stream functions so that p + p∗ → p0 c c and m + m∗ → m0 ∗c c [7].
For these scalarized sources it is now possible to choose the (gauge)
functions ψe , ψm in (47) and (48) in such a way that Πe , Πm are
also scalarized and thus have only components along c. To that end
we assume Πe/m = Π(e/m)c c and collect the transversal components of
e.g., (47)
∂ £ −1 ¤
µ⊥ (∇⊥ Πec ) = ²⊥ ∇⊥ ψe (51)
∂c
The transversal components of the vector potentials will then vanish
if we choose
1 ∂ Πec
ψe = (52)
²⊥ ∂c µ⊥
This is a particular form of the condition found by Mohsen [6] for a
coordinate system more general than the cartesian system considered
here. The longitudinal part of the same equation is given by
∂ 2 Πec Πec ∂ψe
²// − ∇2⊥ = p0 c + ²// (53)
∂t2 µ⊥ ∂c
and using (52) we obtain
µ 0 ¶
Π ec
Ls (², µ) = −p0 c (54)
µ⊥
where the scalar operator Ls has been defined in (41). A similar
equation can be found starting from (48)
µ 0 ¶
Π mc ∗
Ls (µ, ²) = −µ// m0 c (55)
²⊥
Progress In Electromagnetics Research B, Vol. 18, 2009 175

Unlike the scalarized current densities, which are unique, the scalarized
stream potentials and corresponding Hertz vector components are not
unique, since they can always be subjected to a gauge transformation.
However, we can in particular choose the scalarized stream potentials
as follows
∂p0 c ∗ ∂m0 ∗c
J 0c = J 0 c = µ// (56)
∂t ∂t
where J 0 c and J 0 ∗c are the (unique) scalarized current densities.
Comparing (54)and (55) with (39) and (40) we conclude that in that
case
∂Π0 ec ∗ ∂Π0 mc
A0 c = A0 c = (57)
∂t ∂t
This correspondence can also be checked by comparing the field
expressions (49) and (50) with (42) and (43) where in (49) and (50)
one should also take into account of the field transformations due to
the scalarization of the current densities (see Tables 2 and 3).

5. SCALAR HERTZ POTENTIALS

The “scalar Hertz potential” formulation introduced by Weiglhofer [11]


starts by decomposing the fields and equations into transversal
and longitudinal components and parts. From these equations the
longitudinal components Ec , Hc can be eliminated, leaving 4 equations
for the 4 unknown transversal components. Up to this point, the
method is identical to the 4 × 4 matrix method used for solving
Maxwell’s equations in stratified media [17]. However, for dealing with
the source terms and unlike the 4 × 4 matrix method, the transversal
field components are then expressed using scalar (Hertz) potential
functions
E ⊥ = ∇⊥ Φ + ∇⊥ × Θc (58)
H ⊥ = ∇⊥ Π + ∇⊥ × Ψc (59)
In what follows we will give a compact derivation of the scalar Hertz
potential equations, following [11]. Splitting Maxwell’s curl-equations
parallel and perpendicular to c we obtain
¡ ¢ ∂Dc
c · ∇⊥ × H ⊥ = Jc + (60)
µ ∂t ¶
¡ ¢ ∂Bc
c · ∇⊥ × E ⊥ = − Jc∗ + (61)
∂t
176 De Visschere

and
µ ¶
∂H ⊥ ∂D⊥
∇⊥ Hc − = c × J⊥ + (62)
∂c ∂t
µ ¶
∂E ⊥ ∗ ∂B ⊥
∇⊥ Ec − = −c × J ⊥ + (63)
∂c ∂t
Taking the cross-product with c, the latter 2 equations become
¡ ¢
∂ c × H⊥ ∂D⊥
c × ∇ ⊥ Hc − = −J ⊥ − (64)
¡ ∂c ¢ ∂t
∂ c × E⊥ ∗ ∂B ⊥
c × ∇⊥ Ec − = J⊥ + (65)
∂c ∂t
The 4 × 4 matrix method is based on (63) and (64), without the source
terms, where (60) and (61) are used for eliminating Ec and Hc , after
inserting the constitutive equations
Dc = ²// Ec D ⊥ = ²⊥ E ⊥ (66)
Bc = µ// Hc B ⊥ = µ⊥ H ⊥ (67)
Weiglhofer [11] deals with the source terms by operating with ∇⊥ · on
the 4 equations (62)–(65) and by introducing the auxiliary functions
already defined in (20) and (21). Using constitutive equations (66),
(67) and decompositions (58), (59), all terms then contain the
Laplacian ∇2⊥ which can be dropped, yielding
∂Π ∂Θ
Hc − − ²⊥ = v (68)
∂c ∂t
∂Φ ∂Ψ
Ec − + µ⊥ = −v ∗ (69)
∂c ∂t
∂Φ ∂Ψ
²⊥ − = −u (70)
∂t ∂c
∂Π ∂Θ
µ⊥ + = −u∗ (71)
∂t ∂c
On the other hand the longitudinal equations (60) and (61) become
∂Ec
−∇2⊥ Ψ = Jc + ²// (72)
∂t
∂Hc
∇2⊥ Θ = Jc∗ + µ// (73)
∂t
Finally ∂Ec /∂t and ∂Hc /∂t can be calculated from (68)–(71) as a
function of Ψ, Θ only, and when substituted in (72) and (73) one
obtains two uncoupled 2nd order equations in Ψ, Θ. We remind
Progress In Electromagnetics Research B, Vol. 18, 2009 177

the reader that the RHSs of (60), (61) and thus also of (72), (73)
are invariant under the source scalarization transformations shown
in Tables 2 and 3, meaning that ∇⊥ × E ⊥ and ∇⊥ × H ⊥ , and also
Ψ, Θ are invariant under these transformations. Therefore, the scalar
potentials Ψ, Θ can only depend on the unique scalarized current
densities J 0 c and J 0 ∗c . Whereas using the vector potential method
in § 3 or the Hertz vector method in § 4, scalarization could only
be obtained with some effort by applying appropriate source and
gauge transformations; the scalar Hertz potentials Ψ and Θ are the
natural potentials for obtaining scalarization since they are invariant
under the required transformations. The final equations can then also
be obtained immediately by making the RHSs of (68)–(71) zero and
replacing Jc , Jc∗ in (72) and (73) by the scalarized versions J 0 c , J 0 ∗c
leading to
Ls (², µ) (Ψ) = −J 0 c (74)
0∗
Ls (µ, ²) (−Θ) = −J c (75)
However note that Ec , Hc and Φ, Π are not invariant. From (70) and
(71), we find
Z µ ¶
−1 ∂Ψ
Φ = ²⊥ −u + dt (76)
∂c
Z µ ¶
−1 ∗ ∂Θ
Π = −µ⊥ u + dt (77)
∂c
and subsequently from (68) and (69)
∂Φ ∂Ψ
Ec = −v ∗ + − µ⊥ (78)
∂c ∂t
∂Π ∂Θ
Hc = v + + ²⊥ (79)
∂c ∂t
The total fields can then µ be written¶as
∂Ψ
E = − v ∗ + µ⊥ c + ∇⊥ Θ × c + ∇Φ (80)
∂t
µ ¶
∂Θ
H = v + ²⊥ c + ∇⊥ Ψ × c + ∇Π (81)
∂t
These expressions confirm the field transformations in Tables 2 and 3,
and they correspond term for term with the expressions in (42) and
(43) with the correspondence
A0 c ∂ Π0 ec A0 ∗c ∂ Π0 mc
Ψ= = −Θ= = (82)
µ⊥ ∂t µ⊥ ²⊥ ∂t ²
Z Z ∗⊥
u u
Φ = −φ − dt Π = −φ∗ − dt (83)
²⊥ µ⊥
178 De Visschere

which was already (partially) apparent from (74), (75) and (39), (40).

6. GYROTROPIC MEDIA

For a gyrotropic medium the transversal dielectric tensor is given by [8]


· ¸
²⊥ j²0⊥
²⊥ = = ²⊥ I + j²0⊥ c × I (84)
−j²0⊥ ²⊥
with a similar expression for the permittivity tensor.
0
Since the scalarization of the solenoidal parts of J ⊥ and J ⊥ does
not involve these transversal constitutive tensors no changes are needed
here. However, for the irrotational parts the reasoning leading to (24),
(25) and (26) must be extended with additional terms. Introducing the
gauge function Gc in the “electric” domain (7) and (8) and as before
l in the “magnetic” domain (17) we obtain the transformed stream
potentials
Z
p = ∇⊥ udt + ² · ∇l + ∇⊥ Gc × c (85)
∂Gc
m∗ = − c (86)
∂t
If ²0⊥ 6= 0 then the 2nd term on the RHS of (85) contains an extra term
which can be compensated by the last term if
Gc = j²0⊥ l (87)
In this way the extra transversal term in the polarization is transformed
into a longitudinal magnetization, which can be represented by a
longitudinal magnetic current density
∗ ∂ 2 Gc
J 0 c = −µ// (88)
∂t2
Since (24), (25) and (26) remain valid it suffices thus to add a magnetic
current density
∗ ²0 ∂u
J 0 c = j ⊥ µ// (89)
²⊥ ∂t
and due to the electric/magnetic switch (m → m∗ ) in (86) we must
also add a matching field transformation
0 ∂Gc ²0
B − B = µ// c = −j ⊥ µ// uc (90)
∂t ²⊥
Again Jc + ∂B∂t remains invariant for this additional transformation,
c

and therefore the formulation using the scalar Hertz potentials should
still automatically lead to scalarized equations as shown in [9].
Progress In Electromagnetics Research B, Vol. 18, 2009 179

Considering the formulation using Hertz vectors, extra (transver-


sal) terms in j²0⊥ will also occur in (51). However, the functions ψe/m
in (47) and (48) are only part of the gauge transformation, and more
in general we can also add terms in G respectively in L as in (7), (8)
and (17), (18) to the RHS of (47), (48). With in particular G = Gc c
(and L = Lc c) we then obtain instead of (51) condition
µ ¶ µ ¶
∂ µ⊥ ∂ jµ0⊥
∇⊥ Πec − c × ∇⊥ Πec
∂c |µ⊥ | ∂c |µ⊥ |
= ²⊥ ∇⊥ ψe + j²0⊥ c × ∇⊥ ψe + ∇Gc × c (91)
where |µ⊥ | = det µ⊥ = µ2⊥ −µ0 2⊥ . This condition is fulfilled by choosing
µ ¶
−1 ∂ µ⊥
ψe = ²⊥ Πec (92)
∂c |µ⊥ |
· ¸
∂ µ0 ⊥ ²0 ⊥ ∂ µ⊥
Gc = j + Πec (93)
∂c µ⊥ ²⊥ ∂c |µ⊥ |
The longitudinal Equation (53) is replaced by
µ ¶ µ ¶
∂ 2 Πec 2 µ⊥ 0 ∂ψe ∂Lc
²// − ∇⊥ Πec = p c + ²// − (94)
∂t2 |µ⊥ | ∂c ∂t
where −Lc is given by a similar expression as in (93). We notice that
except for the replacement of µ−1 ⊥ by µ⊥ / |µ⊥ | the main change is
the occurrence of a cross-coupling term due to the additional gauge
functions Gc , Lc which indeed mix between electric and magnetic
stream functions. If we replace (41) by the more general expression
∂ −1 ∂ |µ⊥ | ∂ 2
Ls (², µ) = ∇2⊥ + ²// ²⊥ − ²// (95)
∂c ∂c µ⊥ ∂t2
then the scalarized equations for a gyrotropic medium (omitting the
accents) are given by
· ¸
∂ ²0 ⊥ µ0 ⊥ ∂ ∂Σm
Ls (², µ)Σe + j²// + = −pc (96)
∂c ²⊥ µ⊥ ∂c ∂t
· ¸
∂ µ0 ⊥ ²0 ⊥ ∂ ∂Σe
Ls (µ, ²)Σm − jµ// + = −µ// m∗c (97)
∂c µ⊥ ²⊥ ∂c ∂t
where Σe = |µµ⊥ ⊥|
Πec and Σm = |²²⊥ ⊥|
Πmc and as for the uniaxial
case these Hertz vector components are equivalent to the scalar Hertz
potentials with Ψ = ∂Σe /∂t and −Θ = ∂Σm /∂t.
Finally, we consider the formulation using ordinary but standard
vector potentials. Comparing (31), (32) with (47), (48) the close
180 De Visschere

resemblance between both formulations is apparent where in particular


−∂φ(∗) /∂t ⇔ ψe/m and therefore scalarization of (31), (32) might also
be possible. However, whereas for a uniaxial medium the functions φ(∗)
or ψe/m are sufficient for obtaining scalar equations, for a gyrotropic
medium the additional freedom offered by the gauge function Gc in (91)
(and Lc ) is needed, and these functions do not occur in (31) and (32).
With some hindsight we realize that transformations of the “electric”
and “magnetic” stream potentials according to
²−1 · p0 = ²−1 · p − E (98)
²−1 · p0∗ = ²−1 · p∗ + E (99)
m0 = m+H (100)
m0∗ = m∗ − H (101)
correspond with the following transformations of the current densities
0 ∂E
J = J +∇×H−²· (102)
∂t
0∗ ∂H∗
J = J −µ·
−∇×E (103)
∂t
These transformations are thus legitimate, provided the fields are
transformed according to
0 0
E =E+E H =H +H (104)
With additional transformations (102) and (103) of the current
densities the formulations based on the standard vector potentials
on one hand and on the Hertz vectors on the other hand become
fully equivalent. The former can thus also be scalarized with a
proper choice of the derivatives of the potentials ∂φ(∗) /∂t and of the
longitudinal fields Ec and Hc. As a result, the scalar equations (96) and
(97) also hold for the vector potentials with the proper substitutions
(∗)
Ac ⇔ Π(e/m)c , pc ⇔ Jc and µ// m∗c ⇔ Jc∗ (omitting the accents).
Also for this transformation ∇⊥ × E ⊥ and ∇⊥ × H ⊥ are invariant,
explaining why it is also automatically included using the scalar Hertz
potentials.

7. CONCLUSIONS

We have compared three “potential” methods for solving Maxwell’s


equations with arbitrary sources in a lineair stratified gyrotropic
medium where the longitudinal symmetry axis (sometimes referred to
as the distinguished axis) is perpendicular to the strata. In particular,
Progress In Electromagnetics Research B, Vol. 18, 2009 181

we studied the reduction of Maxwell’s equations to two scalar equations


(scalarization). A prerequisite for scalarization to be possible and
which we have accepted without proof, is that the constitutive tensors
should not introduce mixing between longitudinal and transversal
field components. A second condition which was often required in
the derivations is that the transversal constitutive tensors should not
depend on the transverse coordinates. It is perhaps interesting to
note that there are no restrictions on the position dependence of the
longitudinal properties ²// and µ// .
Introducing the conventional vector/scalar potentials or the Hertz
vectors two uncoupled vector equations are obtained. We have shown
that with the limitations already mentioned these equations can
always be scalarized. First the sources must be scalarized: Using
the equivalence between electric and magnetic sources and gauge
transformations for the polarizations (stream potentials), the current
densities can always be replaced by unique current densities along the
distinguished axis. The sources for the Hertz vector equations (the
stream potentials) can be scalarized with the gauge transformations
only, but it is easier to use (56) and the scalarized current densities.
At this stage the problem cannot yet be reduced to two scalar
differential equations because there is still cross-coupling between
the longitudinal and transversal components of the vector potentials
or Hertz vectors. However, using gauge transformations (see (36),
(51) and (91)) these cross-coupling terms can always be eliminated,
and two scalar equations are obtained. When using the vector
potentials, for a gyrotropic medium, an additional transformation
between electric/magnetic sources must be performed (see (98)–(104)),
and for such a gyrotropic medium the final scalar equations are also
coupled. The two methods are found to be fully equivalent, and the
“vector potential” quantities are merely the time derivatives of the
corresponding “Hertz vector” quantities.
We noticed that the appropriate electric/magnetic source
transformations leave the longitudinal “total” current densities ∇⊥ ×
E ⊥ and ∇⊥ × H ⊥ invariant, and this obviously holds also for the
gauge transformations. In a third method, 4 scalar Hertz potentials
are defined based on Helmholtz decompositions (58) and (59) in the
transversal plane, and it immediately follows that 2 of those scalar
potentials, Ψ and Θ, are also invariant under all transformations
needed to obtain scalarization. These are the natural potentials for
obtaining scalarization, and this invariance explains why, when using
these scalar Hertz potentials, the scalarized current densities emerge
effortlessly. These scalar Hertz potentials are also equivalent with the
longitudinal components of the scalarized traditional Hertz vectors
182 De Visschere

used in the 2nd method, and eventually we conclude that the three
methods are fully equivalent. It remains to be investigated whether this
conclusion still holds for more complex media for which scalarization
has been obtained using the scalar Hertz potentials [10].

REFERENCES

1. Joannopoulos, J. D., S. G. Johnson, J. N. Winn, and R. D. Meade,


Photonic Crystals, 3nd edition, Princeton University Press, 2008.
2. Monk, P., Finite Element Methods for Maxwell’s Equations,
Numerical Mathematics and Scientific Computation, Oxford
University Press, 2003.
3. Olyslager, F. and I. Lindell, “Field decomposition and factoriza-
tion of the Helmholtz determinant operator for bianisotropic me-
dia,” IEEE Transactions on Antennas and Propagation, Vol. 49,
No. 4, 660–665, Apr. 2001.
4. Nisbet, A., “Hertzian electromagnetic potentials and associated
gauge transformations,” Proceedings of the Royal Society of
London. Series A, Mathematical and Physical Sciences (1934–
1990), Vol. 231, No. 1185, 250–263, Aug. 1955.
5. Nisbet, A., “Electromagnetic potentials in a heterogeneous non-
conducting medium,” Proceedings of the Royal Society of London.
Series A, Mathematical and Physical Sciences (1934–1990),
Vol. 240, No. 1222, 375–381, Jun. 1957.
6. Mohsen, A., “Electromagnetic field representation in inhomoge-
neous anisotropic media,” Applied Physics A: Materials Science
& Processing, Vol. 2, No. 3, 123–128, Sep. 1973.
7. Mohsen, A., “Electromagnetic-field representation in source
region,” Applied Physics, Vol. 10, No. 1, 53–55, Jan. 1976.
8. Przeźiecki, S. and R. A. Hurd, “A note on scalar Hertz potentials
for gyrotropic media,” Applied Physics, Vol. 20, No. 4, 313–317,
Dec. 1979.
9. Weiglhofer, W. S. and W. Papousek, “Scalar Hertz potentials for
anisotropic media,” Aeu-Arch Elektron. Ub, Vol. 39, No. 6, 343–
346, Jan. 1985.
10. Weiglhofer, W. and S. Hansen, “Faraday chiral media revisited.
I. Fields and sources,” IEEE Transactions on Antennas and
Propagation, Vol. 47, No. 5, 807–814, May 1999.
11. Weiglhofer, W. S., “Scalar Hertz potentials for nonhomogeneous
uniaxial dielectric-magnetic mediums,” Int. J. Appl. Electrom.,
Vol. 11, No. 3, 131–140, Jan. 2000.
Progress In Electromagnetics Research B, Vol. 18, 2009 183

12. Georgieva, N. and W. S. Weiglhofer, “Electromagnetic vector


potentials and the scalarization of sources in a nonhomogeneous
medium,” Physical Review E, Vol. 66, No. 4, Oct. 8, 2002.
13. Weiglhofer, W. S. and N. Georgieva, “Vector potentials
and scalarization for nonhomogeneous isotropic mediums,”
Electromagnetics, Vol. 23, No. 5, 387–398, Jul. 2003.
14. Jackson, J. D., “From Lorenz to Coulomb and other explicit gauge
transformations,” Am. J. Phys., Vol. 70, No. 9, 917–928, 2002.
15. Jackson, J. D., Classical Electrodynamics, 3rd edition, 280–282,
John Wiley & Sons, 1999.
16. Sein, J. J., “Solutions to time-harmonic Maxwell equations with
a hertz vector,” Am. J. Phys., Vol. 57, No. 9, 834–839, Jan. 1989.
17. Berreman, D. W., “Optics in stratified and anisotropic media —
4 × 4 — matrix formulation,” J. Opt. Soc. Am., Vol. 62, No. 4,
502–510, Jan. 1972.

You might also like