Waveguide notes 2020

Electromagnetic waves in free space

We start with Maxwell’s equations for an LIH medum in the case that the
source terms ρf and jf are both zero.

∇ · D = ∇ · εE = 0 ∇·B =0
∂B
∇×E =−
∂t
∂D
∇×H =
∂t
Take the curl of Faraday’s law, and then use Ampere’s law:
  ∂∇ × B ∂∇ × H
∇× ∇×E = − = −µ
∂t ∂t
  ∂ ∂ D
∇ ∇ · E − ∇2 E = −µ
∂t ∂t
Use the first Maxwell equation (the "H" in LIH assures us that spatial derivatives
√
of ε are zero1 ), and we obtain the wave equation with wave speed vφ = 1/ µε

∂2E
∇2 E = µε
∂t2
∂2E
= vφ2 ∇2 E
∂t2
A similar derivation gives the same equation for B. Now let’s look at a plane
wave solution:
 
E = E0 exp ik · x − iωt
 
B = B0 exp ik · x − iωt + φ

where ω/k = vφ , the wave phase speed. By including the phase constant φ
in the expression for B we allow for a possible phase shift bewteen E and B.
Inserting these expressions into Maxwell’s equations, we have

∇·E = 0 ⇒ k · E0 = 0 (1)
∇·B = 0 ⇒ k · B0 = 0

Thus both E and B are perpendicular to the direction of propagation. From

Faraday’s law
   
k × E0 exp ik · x − iωt = ω B0 exp ik · x − iωt + φ
1 Here we also assume that ε is independent of t.

1
Since this relation must be true for all x and t, and k is real2 , we have φ = 0
(E and B oscillate in phase) and

k 1
B0 = × E0 = k̂ × E0 (2)
ω vφ

Thus B is also perpendicular to E, and its magnitude is E/vφ .

If the waves propagate in a vacuum, the derivation goes through in the same
√
way and the only difference is that the wave speed is c = 1/ µ0 ε0 . In an LIH
s s
medium, vφ = c/n, where the refractive index n = εµ/ε0 µ0 ε/ε0 .
Electromagnetic fields in a wave guide
A wave guide is a region with a conducting boundary inside which EM waves
are caused to propagate. In this confined region, the boundary conditions create
constraints on the wave fields. We shall idealize, and assume that the walls are
perfect conductors. If they are not, currents flowing in the walls lead to energy
loss. See Jackson Ch 8 for a discussion of this case, especially sections 1 and 5.

The boundary conditions at the walls of our perfectly conducting guide are
(see notes 1 eqns 10,12,13 and 15):

n̂ · D = Σ (3)

where Σ is the free surface charge density on the wall,

n̂ × E = 0 (4)
2 k is not always real. We will see later some situations where k is complex. In these cases

E and B may be out of phase.

2
 n̂ · B = 0 (5)
and
n̂ × H = K (6)
where K is the free surface current density. (Note that we have taken H = 0
inside the conducting material. This is true here because all our fields are
time-dependent, and then non-zero ∂ B/∂t implies non-zero E, by Faraday’s
law. Non-zero E is not allowed inside a perfect conductor, and so H must be
zero too. ) Since we do not know Σ or K, equations (4) and (5) will be most
useful.
Now we use cylindrical coordinates with ẑ along the guide in the direction
of wave propagation. The transverse coordinates will be chosen to match the
cross-sectional shape of the guide — Cartesian for a rectangular guide and polar
for a circular guide. Next we assume that all fields may be written in the form

E = E0 (x) e−iωt

We are not making any special assumptions about the time variation, because
we can always Fourier transform the fields to get combinations of terms of this
form. Then Maxwell’s equations in the guide take the form:

∇ × E = iω B

∇·D = 0 ∇·B =0
and
∇ × H = −iω D
Taking the curl of Faraday’s law, and inserting ∇ × B from Ampere’s law, we
get:
     
∇× ∇×E = ∇ ∇ · E − ∇2 E = iω∇ × B = iω −iωµεE
∇2 E = −ω2 µεE (7)

This equation is the same as we obtained for free space. Note that ε is usually
a function of ω.
Next we look for solutions that take the form of waves propagating in the
z−direction, that is:
E0 (x) = Ea (x, y) eikz
The wave equation (7) then becomes:

∇2t E + ω 2 µεE − k2 E = 0 (8)

where ∇2t is the Laplacian operator in the two transverse coordinates (x and y,
or ρ and φ, for example.) Thus equation (8) is an equation for the function Ea
of the two transverse coordinates.

3
 Since we were able to simplify the equations by separating the function E
into its dependence on the coordinates along and transverse to the guide, we
now try to do the same thing with the components. At first glance, and based
on equation (1), you might want to jump to the conclusion that there is no
z−component of a wave propagating in the z−direction, but in general there
is. The waves are propagating between conducting boundaries, and we have to
allow for the possibility that waves travel at an angle to the guide center-line,
and bounce back and forth off the walls as they travel. Since E is perpendicular
to the wave vector, E in such a bouncing wave has a z−component. The total
electromagnetic disturbance in the guide is a sum of such waves. The sum is a
combination of waves that interfere constructively. Thus we take

Ea = Ez ẑ + Et

and similarly
Ba = Bz ẑ + Bt
This decomposition simplifies the boundary conditions, since the normal n̂ on
the boundary has no z−component. Then eqn (5) becomes

n̂ · Bt = 0 on S (9)

However equation (4) has two components. The transverse component gives

Ez = 0 on S (10)

while the z− component gives

n̂ × Et = 0 on S (11)

Next we put these components into Maxwell’s equations. The “divergence”

equations are scalar equations, so let’s start with them:
 
∂Ez
∇·D = 0= ε + ∇t · Et
∂z

and evaluating the z−derivative, we get

ikEz + ∇t · Et = 0 (12)

Similarly:

ikBz + ∇t · Bt = 0 (13)
We separate the curl equations into transverse and z−components. Take the
dot product of Faraday’s law with ẑ:
   
ẑ · ∇ × E = iωBz = ẑ · ∇t × Et (14)

4
and also the cross product:
 
ẑ × ∇ × E = ẑ × iω B

Let’s investigate the triple cross product on the left. Since ẑ is a constant, we
may move it through the ∇ operator in the BAC-CAB rule:
     
ẑ × ∇ × E = ∇ ẑ · E − ẑ · ∇ E = ẑ × iω Bt

The derivative ∂Ez /∂z times ẑ appears in both terms in the middle, and so
cancels, leaving:
∂
∇t Ez − Et = iωẑ × Bt
∂z
and evaluating the z−derivative, we get

∇t Ez − ikEt = iωẑ × Bt (15)

Similarly, from Ampere’s law, we have the transverse component:

∇t Bz − ikBt = −iωεµẑ × Et (16)

and the z−component

 
−iωεµEz = ẑ · ∇t × Bt (17)

Equations (12), (13), (17) and (14) show that the longitudinal components
Ez and Bz act as sources of the transverse fields Bt and Et .
Now we can simplify by looking at the normal modes of the system.
Transverse Electric (TE) (or magnetic) modes.
In these modes there is no longitudinal component of E:

Ez ≡ 0 everywhere

Thus boundary condition (10) is automatically satisfied. The remaining bound-

ary conditions are (9) and (11), and we can find the version that we need by
taking the dot product of n̂ with equation (16):
 
n̂ · ∇t Bz − ikn̂ · Bt = −iωεµn̂ · ẑ × Et

The second term is zero on S (eqn 9), and we rearrange the triple scalar product
on the right, leaving:
∂Bz  
= iωεµẑ · n̂ × Et = 0 on S (18)
∂n

where we used the other boundary condition (11) for E.

5
 Transverse Magnetic (TM) (or electric) modes.
In these modes there is no longitudinal component of B:
Bz ≡ 0 everywhere
Thus the boundary condition (18) is trivially satisfied, and we must impose the
remaining condition (10)
Ez = 0 on S.
Since the Maxwell equations are linear, we can form superpositions of these
two sets of modes to obtain fields in the guide with non-zero longitudinal com-
ponents of both E and B. These modes are the result of the constructive inter-
ference mentioned above.
Transverse electromagnetic (TEM) modes
In these modes both Ez and Bz are zero everywhere. Then from (12) and
(14), ∇t × Et and ∇t · Et are zero everywhere. This means we can express Et
as the gradient of a scalar function Φ that satisfies Laplace’s equation in two
dimensions. The boundary condition (11) becomes
 
∂Φ
n̂ × ∇Φ = n̂ × ŝ = 0 on S
∂s
where s is a coordinate parallel to the surface S. Since n̂ and ŝ are perpendicular,
n̂× ŝ is not zero, and so Φ = constant on S and therefore is constant everywhere
inside the volume V, making Et = 0. Thus these modes cannot exist inside a
hollow guide. They may exist, and in fact become the dominant modes, inside a
guide with a separate inner boundary, like a coaxial cable. We will not consider
them further here.
Now let’s see how the equations simplify for the TE and TM modes..
TM modes
We start by finding an equation for Ez . Since Bz ≡ 0, equation (16) simplifies
to:
−ikBt = −iωεµẑ × Et
ω
Bt = εµẑ × Et (19)
k
and we substitute this result back into equation (15).
ω 
−ikEt + ∇t Ez = iωẑ × εµẑ × Et
k
ω2
∇t Ez = ikEt − i εµEt
 k 
ω2
= ik 1 − 2 εµ Et (20)
k
and finally we substitute this result for Et back into equation (12):
∇t Ez
ikEz + ∇t ·  2  = 0
ik 1 − ωk2 εµ
 
∇2t Ez + ω2 εµ − k2 Ez = 0

6
or
∇2t Ez + γ 2 Ez = 0 (21)
with
γ 2 ≡ ω 2 εµ − k2 (22)
3
Equation (21) is the defining differential equation for Ez . Once we have solved
for Ez , we can find Et from equation (20) and then Bt from equation (19).
TE modes
The argument proceeds similarly. We start with equation (15) with Ez = 0,
to get:
ω
Et = − ẑ × Bt (23)
k
and substitute into equation (16)
 ω 
−ikBt + ∇t Bz = −iωεµẑ × − ẑ × Bt
 k
ω 2 εµ
∇t Bz = i k − Bt (24)
k
Then finally from equation (13) we have:
∇t Bz
ikBz + ∇t ·  2
 = 0
i k − ω kεµ
∇2t Bz + γ 2 Bz = 0
which is the same differential equation that we found for Ez in the TM modes.
The solutions are different because the boundary conditions are different. Thus
the solution for the two modes proceeds as follows:

TM modes TE modes
assumed Bz ≡ 0 Ez ≡ 0

differential equation ∇2t Ez + γ 2 Ez = 0 ∇2t Bz + γ 2 Bz = 0 This is an eigenvalue/

∂Bz
boundary condition Ez = 0 on S ∂n = 0 on S eigenfunction problem.

ik ik
next find Et = γ 2 ∇t Ez Bt = γ 2 ∇t Bz

ω
then find Bt = k εµẑ × Et Et = − ωk ẑ × Bt

The differential equation plus boundary condition is an eigenvalue problem

that produces a set of eigenfunctions Ez,n (or Bz ,n ) and a set of eigenvalues
γ n . The wave number kn is then determined from equation (22):
kn2 = ω 2 εµ − γ 2n (25)
3 This equation is the Helmholtz equation.

7
 √
Clearly if γ n is greater than ω εµ = ω/v, where v is the wave phase speed
in unbounded space, kn becomes imaginary and the wave does not propagate.
There is a cut-off frequency for each mode, given by

ωn = γ n v

If γ c is the lowest eigenvalue for any mode, the corresponding frequency ωc is the
cutoff frequency for the guide, and waves at lower frequencies cannot propagate
in the guide.
A few things to note: the wave number kn is always less than the free-
space value ω/v, and thus the wavelength is always greater than the free-space
wavelength. The phase speed in the waveguide is
ω 1 1 1
vφ = =√ s > √ =v
k µε 1 − ω 2n /ω 2 µε

and we can differentiate eqn (25) to get

dω
2ω εµ = 2k
dk
Then the group speed in the guide is
u
dω 2k 1 1 ω 2n 1
vg = = = =√ 1− < √ =v
dk 2ωεµ µεvφ µε ω2 µε
Thus information travels more slowly than if the wave were to propagate in free
space.
TM modes in a rectangular wave guide
We use Cartesian coordinates with origin at one corner of the guide. Let
the guide have dimensions a in the x-direction by b in the y−direction. Let the
interior be full of air so ε/ε0 = µ/µ0 1. Then the differential equation for Ez
is (21)  2 
∂ ∂2 2
+ + γ Ez = 0
∂x2 ∂y 2
As usual we look for a separated solution, choosing Ez = X (x) Y (y) to obtain:
X Y
+ + γ2 = 0
X Y
Each term must separately be constant, so we have:
X
= −α2
X
Y
= −β 2
Y
and
−α2 − β 2 + γ 2 = 0

8
 The boundary condition is Ez = 0 on S, so:

X = 0 at x = 0 and x = a

and
Y = 0 at y = 0 and y = b
Thus the appropriate solutions are X = sin αx and Y = sin βy with eigenvalues
chosen to fit the second boundary condition in each coordinate:
nπ mπ
α= and β =
a b
Thus, putting back the dependence on z and t, we have
nπx mπy ikz−iωt
Ez = Enm sin sin e (26)
a b
and  nπ 2  mπ 2
γ 2nm = + (27)
a b
Notice that the lowest possible values of n and m are 1 in each case, since taking
m or n = 0 would render Ez identically zero. Thus the lowest eigenvalue is
u
1 1
γ 11 = π +
a2 b2
and the cutoff frequency for the TM modes is:
u u
1 1 cπ a2
ω c,TM = cγ 11 = cπ + = 1 +
a2 b2 a b2
Jackson solves for the TE modes (pg 361). The eigenvalues are the same,
but in this case it is possible for one (but not both) of m and n to be zero,
leading to a lower cutoff frequency for the TE modes:
cπ
ωc,TE =
max(a, b)
This would be the cutoff frequency for the guide.
In the TM mode, the remaining fields are (eqns 20 and 26):
ik nπx mπy ikz−iωt
Et = ∇t Enm sin sin e
γ2 a b
ikπ n nπx mπy m nπx mπy  ikz−iωt
= E nm x̂ cos sin + ŷ sin cos e (28)
γ2 a a b b a b
and (eqn 19)
ω ik  nπ nπx mπy mπ nπx mπy  ikz−iωt
Bt = ẑ × E nm x̂ cos sin + ŷ sin cos e
kc2 γ2 a a b b a b
ω  nπ nπx mπy mπ nπx 
mπy ikz−iωt
= i 2 2 Enm ŷ cos sin − x̂ sin cos e (29)
c γ a a b b a b

9
where (eqns 25 and 27)
u  nπ 2  mπ 2
k= ω 2 εµ − −
a b
As usual, the physical fields are given by the real part of each mathematical
expression, so that, for example, Et ∝ sin (kz − ωt). You should verify that
these fields satisfy the boundary conditions at x = 0, a and at y = 0, b.
Power
The power transmitted by the waves in the guide is:
1
S (t) = E×B
µ0

where here we must take the real, physical fields. Usually we are interested in
the time-averaged Poynting flux, which is given by
1
< S > = Re E × B∗ (30)
2µ0

where the fields on the right are the complex functions we have just found.
Proof of this result:
If E = E0 e−iωt = êE0 eiφE e−iωt and similarly for B, where E0 and B0 are
real, then
1    
S= Re E × Re B
µ0
and the time average is
E0 B0
< S > = <ê × b̂ cos (φE − ωt) cos (φB − ωt) >
µ0
E0 B0
= <ê × b̂ (cos φE cos ωt + sin φE sin ωt) (cos φB cos ωt + sin φB sin ωt) >
µ0
E0 B0 
= ê × b̂ cos φE cos φB cos2 ωt+
µ0

(cos φE sin φB + cos φB sin φE ) cos ωt sin ωt + sin φE sin φB sin2 ωt
E0 B0 E0 B0
= ê × b̂ (cos φE cos φB + sin φE sin φB ) = ê × b̂ cos (φE − φB )
2µ0 2µ0
1
= Re E × B∗
2µ0

so the two results are the same. (See also Notes 2 page 11.)
Using the solution (28, 29), the components of S in the rectangular guide

10
are:
1  
< Sz > = Re Ex By∗ − Ey Bx∗
2µ0
 
1 ik ω nπ nπx mπy 2  mπ nπx mπy 2
= Re Enm (−i) 2 2 Enm cos sin + sin cos
2µ0 γ 2 c γ a a b b a b
2
 2  2

Enm k ω nπ nπx mπy mπ nπx mπy
= cos sin + sin cos
2µ0 γ 4 c2 a a b b a b
s     
2
Enm ω2 /c2 − γ 2 ω nπ nπx mπy 2 mπ nπx mπy 2
= cos sin + sin cos
2µ0 γ4 c2 a a b b a b
t    
2 2  
mπy 2  mπ mπy 2
ω 2 nπ
2
ε0 Enm ω c2 − a − mπ b nπ nπx nπx
= k   2 2 l cos sin + sin cos
2 nπ 2 a a b b a b
a + mπ b

Check the dimensions! < Sz > is positive for all values of x and y, showing that
power is propagating continuously along the guide in the positive z-direction.
The transverse component Sx is:
1   1  
< Sx > = Re Ey Bz∗ − Ez By∗ = Re −Ez By∗
2µ0 2µ0
 
1 nπx mπy ω nπ nπx mπy
= Re −E0 sin sin (−i) 2 2 E0 cos sin
2µ0 a b c γ a a b
= 0
Because there is no real part, the time averaged power flowing across the guide
is zero. Power sloshes back and forth, but there is no net energy transfer.
Fields in a parallel plate wave guide
By simplifying the shape of the guide even more, we can demonstrate how
the wave modes are formed by reflection of waves at the guide walls. Let this
guide exist in the region 0 ≤ y ≤ a, with infinite extent in x. For the TM modes,
the equation to be satisfied is:
 2 
∇t + γ 2 Ez = 0
with
Ez = 0 at y = 0 and y = a
Because the region is infinite in the x−direction, the appropriate solution has
no x−dependence:
nπy ikz−iωt
Ez = En sin e (31)
a
with
nπ
γn =
a
Thus the cutoff frequency for these modes is
πc
ω c,TM = γ 1 c = (32)
a

11
Then the other components of the fields are:
ik ik nπ nπy ikz−iωt
Et = ∇t Ez = 2 En cos e ŷ
γ2 γ a a
ika nπy ikz−iωt
= En cos e ŷ (33)
nπ a
and
ω ω ika nπy ikz−iωt
Bt = 2
ẑ × Et = − 2 En cos e x̂
kc kc nπ a
ω ia nπy ikz−iωt
= − 2 En cos e x̂ (34)
c nπ a

Let’s look at the electric field first. We write the sine and cosine as combi-
nations of complex exponentials. From (31),
 iγy   iγy 
e − e−iγy ikz−iωt γ −e + e−iγy
Ez = En e = i En eikz−iωt
2i γ 2
and from (33)  
ik eiγy + e−iγy
Ey = En eikz−iωt
γ 2
Thus we can write the electric field as a superposition
1 
En,total = En1 + En2
2
where the two superposed fields are
En
En1 = i (kŷ − γ ẑ) exp (ikz + iγy) e−iωt
γ
and
En
En2 = i (kŷ + γ ẑ) exp (ikz − iγy) e−iωt
γ
Similarly:
 iγy 
ω e + e−iγy 1 
Bt = −i 2
E n x̂ eikz−iωt = Bn1 + Bn2
c γ 2 2
with
ω
Bn1,2 = −i En x̂ exp [i (kz ± γy)] e−iωt
γc2
Now define the four vectors

u1 = kŷ − γ ẑ; u2 = kŷ + γ ẑ

and
v1 = γ ŷ + kẑ; v2 = −γ ŷ + kẑ

12
Then for i = 1, 2, the vectors are perpendicular:

ui · vi = 0

and
  ω2
v1 × u1 = (kẑ + γ ŷ) × (kŷ − γ ẑ) = −x̂ γ 2 + k2 = −x̂ 2 = v2 × u2
c
where we used equation (25). The vectors are shown in the diagram below.

The two electric field components are then:

En En
En1 = iu1 exp (iv1 · x) e−iωt and En2 = iu2 exp (iv2 · x) e−iωt
γ γ
while
v1 × En1 ω
= −i 2 x̂En exp [i (kz + γy)] = Bn1
ω γc
consistent with (2) and (34).
Each of these sets of fields (En1 and Bn1 , En2 and Bn2 ) has the form of
a free-space wave propagating in the direction given by the vectors v1 and v2
respectively and with wave number
s ω
|v1 | = |v2 | = γ 2 + k2 =
c
. These waves are moving across the guide at an angle given by
vy γ γ
tan θ = = ± = ±t
vz k ω2 2
c2 − γ

that is, the waves are reflecting off the plates at y = 0, a, as shown in the
diagram. When the angle θ becomes π/2, the wave ceases to propagate along
the guide, but just bounces back and forth, perpendicular to the walls. This
happens when tan θ → ∞, or
ω
γ=
c
This gives the cut-off frequency (32) we found before.
See also Lea and Burke pages 1058-1060.

13