ELE 2612 – Draft of Lecture 1.

On the Electromagnetic Fields in Planarly

Layered Uniaxially-Anisotropic Media
Part I
Guilherme Simon da Rosa

(Rev. 2020.10.09)

Contents
1 Electromagnetic Fields in Homogeneous Anisotropic Media 2
1.1 Axial Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.1 Solving the Wave Equation via the Method of Separa-
tion of Variables . . . . . . . . . . . . . . . . . . . . . 4
1.2 Transversal Fields . . . . . . . . . . . . . . . . . . . . . . . . . 6

2 Electromagnetic Fields Along Layered Anisotropic Media 8

2.1 Reflection and Transmission from a Two-Layer Medium . . . . 8
2.1.1 TMz Fields . . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.2 TEz Fields . . . . . . . . . . . . . . . . . . . . . . . . . 11

References 12

1
1 Electromagnetic Fields in Homogeneous
Anisotropic Media
The Maxwell’s equations in a linear homogeneous and anisotropic medium
are
∇ × E = iω µ̄ ¯ · H, (1)
∇ × H = −iω ¯ · E + J, (2)
∇ · (¯ · E) = %, (3)
¯ · H) = 0,
∇ · (µ̄ (4)
where E and H are the electric and the magnetic fields, respectively, due to
the impressed volumetric electric current and charge densities J and %. Our
media is characterized by the complex permeability
¯ = diag(µs , µs , µz ), with µ{s,z} = µ0 µ{s,z}
µ̄ (5)
and permittivity
¯ = diag(s , s , z ), with {s,z} = 0 r{s,z} + iσ{s,z} /ω (6)
tensors, represented in Cartesian coordinates. We have decomposed the ten-
sors into axial (along z) and transverse components, with subscripts z and
s, respectively. We have also expressed the constitutive parameters in terms
of the usual vacuum constants 0 and µ0 .
We next consider the fields in a source-free region and decompose the
problem into axial and transversal components as
F = Fs + ẑFz , (7)
in which F = E or H, Fz = Ez or Hz , and
∂
∇ = ∇s + ẑ
, (8)
∂z
where ∇s denotes the transverse two-dimensional nabla operator given by
∂ ∂
∇s = x̂ + ŷ . (9)
∂x ∂y

1.1 Axial Fields

Taking the curl of the equation (1) and projecting the resulting vector onto
ẑ, we can obtain
¯ · H)].
ẑ · [∇ × (∇ × E)] = iωẑ · [∇ × (µ̄ (10)

2
In what follows, we will show that the above scalar wave equation can be
represented in a more familiar form.
As the problem at hand deal with homogeneous media, the right-hand
side (RHS) of (10) can be simplified by using
¯ · H)] = µs ẑ · (∇ × H).
ẑ · [∇ × (µ̄ (11)
Noting that ẑ · (∇ × H) = −iωz Ez , we can derive
¯ · H)] = −iωµs z Ez .
ẑ · [∇ × (µ̄ (12)
Consequently, (10) becomes
ẑ · [∇ × (∇ × E)] = ω 2 µs z Ez . (13)
The left-hand side (LHS) of (13) can be simplified after using the vector
identity
∇ × ∇ × E = ∇(∇ · E) − ∇2 E. (14)
The divergence of the electric and magnetic fields do not vanish anymore
as in the isotropic media [1, 2]. The Gauss’s law for magnetism (4) can be
decomposed as
 
∂
¯ · H) = ∇ + ẑ
∇ · (µ̄ · (µs Hs + ẑHz )
∂z
∂Hz
= µs ∇ s · Hs + µz
 ∂z 
µz ∂Hz
= µs ∇ · H − µs 1 − = 0, (15)
µs ∂z
and a dual equation can be easily established for the electric field, such we
can derive
 
z ∂Ez
∇·E= 1− , and (16)
s ∂z
 
µz ∂Hz
∇·H= 1− . (17)
µs ∂z
As expected, ∇ · E and ∇ · H vanish for the isotropic limits s = z and
µs = µz .
Combining the above results when we take the projection of equation (14)
onto ẑ allow us to find the LHS of (13):
ẑ · [∇ × (∇ × E)] = ẑ · ∇(∇ · E) − ẑ · (∇ · ∇)E (18)
z ∂ 2 Ez
 
= 1− − ∇2 Ez , (19)
s ∂z 2

3
where ∇2 is the scalar Laplacian in Cartesian coordinates given by

∂2 ∂2 ∂2
∇2 = + + . (20)
∂x2 ∂y 2 ∂z 2
Finally, after substituting (19) into (13), one obtains

z ∂ 2 Ez
∇2s Ez + + ω 2 µs z Ez = 0, (21)
s ∂z 2
in which the two-dimensional Laplacian operator is defined by

∂2
∇2s = ∇2 − . (22)
∂z 2
A wave equation dual to (21) can be derived for the axial magnetic field
using a procedure analogous to that shown above. It follows that

µz ∂ 2 Hz
∇2s Hz + + ω 2 µz s Hz = 0. (23)
µs ∂z 2

1.1.1 Solving the Wave Equation via the Method of Separation of

Variables
It is clear that (21) and (23) are homogeneous Helmholtz wave equations for
Ez and Hz whose solutions can be constructed by the method of separation
of variables [3, Ch. 5]. The Helmholtz equation in Cartesian coordinates

pz ∂ 2
 
∂ ∂ 2
+ + + ω pz p̃s Fz = 0, (24)
∂x2 ∂y 2 ps ∂z 2

where Fz = {Ez , Hz }, p = {, µ} and p̃ = {µ, }, can be solved by the

elementary wave functions. Following the method of separation of variables,
we seek to find solutions of the form

Fz = X(x)Y (y)Z(z). (25)

Substitution of (25) into (24), division by Fz , and noting that the partial
derivatives now becomes total derivatives, allow us to derive

1 d2 X 1 d2 Y pz 1 d2 Z
2
+ 2
+ 2
+ ω 2 pz p̃s = 0. (26)
X dx Y dy ps Z dz
the above equation should be satisfied every (x, y, z). Notice the last term
is explicitly independent of x, y, z, i.e., a constant, and this imply that the

4
remaining terms should be constants too because the equation is to sum to
zero for all (x, y, z). Hence, we can define

1 d2 X
2
= −kx2 , (27)
X dx
1 d2 Y
= −ky2 , (28)
Y dy 2
1 d2 Z
= −kz2 , (29)
Z dz 2
where kα , α = {x, y, z}, are constant parameters. Substitution of the above
into (26) gives the separation equation
pz
− kx2 − ky2 − kz + ω 2 pz p̃s = 0, (30)
ps
or, the simplified form
ps 2
ks2 = kρ + kz2 , (31)
pz
where we have introduced the transversal and radial wavenumbers ks2 =
ω 2 ps p̃s = ω 2 µs s and kρ2 = kx2 + ky2 , respectively.
In view of (27), the Helmholtz equation (24) was separated into the fol-
lowing equations:

d2 A
2
+ kα2 A = 0, for A = {X, Y, Z} and α = {x, y, z}. (32)
dα
The above are one-dimensional non-homogeneous Laplace’s equations whose
solutions are given by the linear combinations of any pair of harmonic func-
tions such as

A(kα α) → e+ikα α , e−ikα α , sin(kα α), cos(kα α). (33)

Finally, the elementary solution for (24) can be written as

Cx+ e+ikx x + Cx− e−ikx x

 
Fz,kx ky kz = (34)
× Cy+ e+iky y + Cy− e−iky y
 
(35)
× Cz+ e+ikz z + Cz− e−ikz z ,
 
(36)

where Cα± are constants to be determined by enforcing the boundary condi-

tions.

5
 Linear combination of the elementary wave functions are also a solution
to (24) [3, p. 200], such that the general solution is in the shape of
Z Z
Fz = dky dkx C(kx , ky ) Fz,kx ky kz , or (37)
kx ky
XX
Fz = Ckx ,ky Fz,kx ky kz , (38)
kx ky

where the continuous function C(kx , ky ) or the discrete constants Ckx ,ky are
given by a source boundary condition.
It is important to emphasize that, despite the similarities that we can
infer for Ez and Hz , their separation equations are different:
s 2
ks2 = kρ + kz2 for Ez , and (39)
z
µs
ks = kρ2 + kz2 for Hz .
2
(40)
µz

Alternative Choice In some cases, instead of defining the separation vari-

able kz as in (29), in view of (26), the choice

pz 1 d2 Z 2 d2 Z ps   2
= −k̃z ⇒ + k̃z Z = 0 (41)
ps Z dz 2 dz 2 pz
can also confer proper results. In this case, the separation equation becomes

ω 2 pz p̃s = kρ2 + k̃z , (42)

and the solution for Z now assumes the form

r
ps
 √ √ r
ps
 r
ps

+i pps k̃z z −i pps k̃z z
Z k̃z z → e z , e z , sin k̃z z , cos k̃z z .
pz pz pz
(43)

The above render different values of ∂ 2 /∂z 2 for Ez and Hz . This is in contrast
with the former choice in (29).

1.2 Transversal Fields

The transversal field components of the fields can be expressed as a combi-
nation of the axial ones, as will be shown in the following. The Maxwell’s

6
curl equations (1) and (2) in a source-free region can now be rewritten as

∇s × (ẑEz ) ± ikz ẑ × Es = iωµs Hs (44)

∇s × (ẑHz ) ± ikz ẑ × Hs = −iωs Es (45)
∇s × Es = iωµz ẑHz (46)
∇ s × Hs = −iωz ẑEz , (47)

where we have assumed that the axial dependence of all vector fields match
with exp(ikz z). In the above derivation we have replaced ∂/∂z by ikz , in
accordance with (29).
Initially, by the vector multiplication of ẑ on both sides of the equations
in (44) and (45), and using the identities [1, p. 165]

ẑ × (∇s × ẑGz ) = ∇s Gz , and (48)

ẑ × (ẑ × Gs ) = −Gs , (49)

and after a few manipulations, we can express the transversal fields as a

combination of the axial ones via
1
ikz αe2 ∇s Ez + iωµs αh2 ∇s × (ẑHz ) ,
 
Es = (50)
kρ2
1
ikz αh2 ∇s Hz − iωs αe2 ∇s × (ẑEz ) ,
 
Hs = 2 (51)
kρ

where we have introduced the anisotropic coefficients

r r
z µz
αe = , αh = . (52)
s µs

In case of an axially-backward propagating wave with the dependence

exp(−ikz z), the above holds but with kz → −kz because of ∂/∂z → −ikz .
By assuming the axial fields behaves as the plane wave

eikx x eiky y eikz z , (53)

we can derive

∇s = ikx x̂ + iky ŷ (54)

∇s × ẑ = −ikx ŷ + iky x̂, (55)

7
and then the transversal field components reduce to
1
−kz kx αe2 Ez − ωµs ky αh2 Hz


Ex = (56)
kρ2
1
−kz ky αe2 Ez + ωµs kx αh2 Hz


Ey = 2 (57)
kρ
1
−kz kx αh2 Hz + ωs ky αe2 Ez


Hx = 2 (58)
kρ
1
−kz ky αh2 Hz − ωs kx αh2 Ez .


Hy = 2 (59)
kρ

In a compact matrix-notation, we can write them as

    
Ex 1 −kz kx αe2 −ωµs ky αh2 Ez
= 2 (60)
Hx kρ ωs ky αe2 −kz kx αh2 Hz
  2  
1 −kz kx −ωµs ky αe 0 Ez
= 2 2 (61)
kρ ωs ky −kz kx 0 αh Hz
  −1  
1 −kz kx −ωµs ky s 0 Dz
= 2 , (62)
kρ ωs ky −kz kx 0 µs Bz

    
Ey 1 −kz ky αe2 ωµs kx αh2 Ez
= 2 (63)
Hy kρ −ωs kx αe2 −kz ky αh2 Hz
  −1  
1 −kz ky ωµs kx s 0 Dz
= 2 . (64)
kρ −ωs kx −kz ky 0 µs Bz

2 Electromagnetic Fields Along Layered

Anisotropic Media
2.1 Reflection and Transmission from a Two-Layer
Medium
Consider a plane wave propagating in region 1, that is incident on region
2 at z = 0. At the interface, the following boundary conditions should be
observed:

ẑ × E1 = ẑ × E2 (65)
ẑ × H1 = ẑ × H2 . (66)

8
Due to the translational invariance of the medium in the traversal direction
(s, i.e, along x and y), field solutions for all layers must have the the eiks ·rs
variation, i.e., kx = kjx and ky = kjy for each j = {1, 2, 3, . . . } layer. As a
result, we can write the component α = {x, y, z} of a field at the jth region
as the superposition of longitudinally forwardly- and backwardly-propagating
fields, namely,

Fjα = Cj+ e+ikjz z + Cj− e−ikjz z g(kx x, ky y)

 
(67)
= fjα (z)g(kx x, ky y). (68)

In view of that, notice that ∂/∂z → ±ikjz for the field propagating to z ≷ 0.
The boundary conditions require the continuity of field components Ex ,
Ey , Hx and Hy at z = 0. By using (60)–(62), under the invariance of kx ,
ky , kρ , and ω along the regions, we can simplify the boundary-condition
requirements to only depend on Ez or Hz .
For Ez 6= 0 and Hz = 0 (TMz fields), we can obtain
2 2
k1z α1e e1z (z = 0) = k2z α2e e2z (z = 0), for the continuity of Ex and Ey
(69)
2 2
1s α1e e1z (z = 0) = 2s α2e e2z (z = 0), for the continuity of Hx and Hy .
(70)

Notice that in the above we have assumed ∂/∂z → ikz . In case of a axially-
backward propagatin wave with the exp(−ikz z) factor, the above holds but
with the substitution kz → −kz .
The conditions in (69) and (70) are equivalent to imposing the continuity
of
∂ d
αe2 ez ⇒ −1
s dz , and (71)
∂z dz
z ez ⇒ dz . (72)

Notice we have used s αe2 = z . The above reveals that the boundary con-
dition enforcement can done via the axial electric field ez as well as via the
electric flux density dz . The latter was employed in the formulation presented
in [2, Ch. 2] for modeling layered isotropic media.
The boundary conditions for Hz 6= 0 and Ez = 0 (TEz fields) can be
obtained by duality, resulting in
∂ d
αh2 hz ⇒ µ−1
s bz , and (73)
∂z dz
µz hz ⇒ bz . (74)

9
2.1.1 TMz Fields
Following the compact notation introduced in [2], the z-dependent parcel of
the axial electric field in region 1 can be written as the superposition

e1z (z) = e0 e−ik1z z + RT M (z) e0 e+ik1z z , (75)

while in region 2, we have

e2z (z) = T T M (z) e0 e−ik2z z . (76)

By enforcing the boundary conditions at z = 0, we obtain

   
1z T M (z) 2z
k1z (1 − R )= k2z T T M (z) (77)
1s 2s
1z 1 + RT M (z) = 2z T T M (z) .


(78)

By solving the above, we obtain

2s k1z − 1s k2z
RT M (z) = , (79)
2s k1z + 1s k2z
 
1z 22s k1z
T T M (z) = , (80)
2z 2s k1z + 1s k2z

that are Fresnel reflection and transmission coefficients for uniaxially anisotropic
media for describing the Ez field along a two-layer interface.

Alternative Choice If instead of describing the wave reflection and trans-

mission phenomena by using the Ez -based equations (75) and (76) we prefer
to employ a Dz -based formalism, we should start by writing the fields in
regions 1 and 2 in the form

d1z (z) = d0 e−ik1z z + RT M d0 e+ik1z z , (81)

d2z (z) = T T M d0 e−ik2z z . (82)

By enforcing the boundary conditions, we can obtain

2s k1z − 1s k2z
RT M = , (83)
2s k1z + 1s k2z
22s k1z
T TM = . (84)
2s k1z + 1s k2z

10
Observe that the above R and T Fresnel coefficients are related to the ones
in (79) and (80) presented in via:

RT M (z) = RT M (85)
T M (z) 2 TM
T = α12e T , (86)
2 2 2
where α12e = α1e /α2e = 1z /2z . For the special case of isotropic media, the
coefficients (83) and (84) recover the solution presented in [2, p. 49].

2.1.2 TEz Fields

The z-dependent parcel of the axial magnetic field in regions 1 and 2 can be
written as

h1z (z) = h0 e−ik1z z + RT E(z) h0 e+ik1z z , (87)

h2z (z) = T T E(z) h0 e−ik2z z . (88)

By duality with the presented in previous section, we can derive:

µ2s k1z − µ1s k2z
RT E(z) = , (89)
µ2s k1z + µ1s k2z
 
µ1z 2µ2s k1z
T T E(z) = , (90)
µ2z µ2s k1z + µ1s k2z

Alternative Choice By employing a Bz -based formalism, we have

b1z (z) = b0 e−ik1z z + RT E b0 e+ik1z z , (91)

b2z (z) = T T E b0 e−ik2z z . (92)

Invoking the duality principle in view of (83) and (84), we obtain

µ2s k1z − µ1s k2z
RT E = , (93)
µ2s k1z + µ1s k2z
2µ2s k1z
T TE = . (94)
µ2s k1z + µ1s k2z
Notice that

RT E(z) = RT E (95)
T T E(z) = α12h
2
T TE, (96)
2 2 2
where α12h = α1h /α2h = µ1z /µ2z . Again, for the special case of isotropic
media, the above recover the well-known solution presented in [2, p. 49].

11
References
[1] J. Kong, Electromagnetic Wave Theory. New York, NY, USA: Wiley,
1986.

[2] W. C. Chew, Waves and Fields in Inhomogeneous Media. New York,

NY, USA: John Wiley & Sons, 1995.

[3] R. F. Harrington, Time-harmonic electromagnetic fields. New York, NY,

USA: McGraw-Hill, 1961.

12