Microwaves Locked PDF

Department of Electrical Engineering
Technion – Israel Institute of Technology
MICROWAVES – #46216
LECTURE NOTES
based upon lectures delivered by
Prof. L. Schächter
M
Maarrcchh 22000099
This work is subject to copyright. All rights reserved, whether the whole part or part of the material is concerned,
specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on
microfilm, or in any other way, and storage in data banks and electronic storage. Duplication of this publication or
parts thereof is permitted only with the explicit written permission of the authors. Violations are liable for
prosecution.
C ONTENTS
1 Transmission Lines 1
1.1 Simple Model ………………………………………………………… 2
1.2 Coaxial Transmission Line …………………………………………… 7
1.3 Low Loss System ……………………………………………………… 11
1.4 Generalization of the Transmission Line Equations ………………… 13
1.5 Non-Homogeneous Transmission Line ……………………………… 21
1.6 Coupled Transmission Lines ………………………………………… 24
1.7 Microstrip ……………………………………………………………… 26
1.8 Stripline ……………………………………………………………… 38
1.9 Resonator Based on Transmission Line ……………………………… 46
1.9.1 Short Recapitulation ………………………………………… 46
1.9.2 Short-Circuited Line ………………………………………… 50
1.9.3 Open-Circuited Line ………………………………………… 52
1.10 Pulse Propagation ……………………………………………………… 53
1.10.1 Semi-Infinite Structure ……………………………………… 53
1.10.2 Propagation and Reflection ………………………………… 59
i
1.11 Appendix ……………………………………………………………… 64
1.11.1 Solution to Exercise 1.10 ……………………………………… 64
2 Waveguides – Fundamentals 71
2.1 General Formulation …………………………………………………… 71
2.2 Transverse Magnetic (TM) Mode [Hz = 0] …………………………… 79
2.3 Transverse Electric (TE) Mode [Ez = 0] ……………………………… 82
2.4 Power Considerations ………………………………………………… 85
2.4.1 Power Flow …………………………………………………… 85
2.4.2 Ohm Loss ……………………………………………………… 88
2.4.3 Dielectric Loss ………………………………………………… 90
2.5 Mode Comparison …………………………………………………… 93
2.6 Cylindrical Waveguide ………………………………………………… 95
2.6.1 Transverse Magnetic (TM) Mode [Hz = 0] …………………… 95
2.6.2 Transverse Electric (TE) Mode [Ez = 0] ……………………… 97
2.6.3 Power Considerations ………………………………………… 100
ii
2.6.4 Ohm Loss ……………………………………………………… 104
2.7 Pulse Propagation ……………………………………………………… 107
2.8 Waveguide Modes in Coaxial Line …………………………………… 113
3 Waveguides – Advanced Topics 114

3.1 Hybrid Modes ………………………………………………………… 115
3.2 Dielectric Loading – TM01 …………………………………………… 125
3.3 Cross-Section Variation and Mode Coupling ………………………… 130
3.3.1 Step Transition – TE mode …………………………………… 130
3.3.2 Step Transition – TM0n Mode in Cylindrical Waveguide …… 139
3.4 Reactive Elements …………………………………………………… 146
3.4.1 Inductive Post ………………………………………………… 146
3.4.2 Diaphragm …………………………………………………… 153
3.5 Excitation of Waveguides ……………………………………………… 157
3.6 Coupling between Waveguides by Small Apertures ………………… 164
3.7 Surface Waveguides …………………………………………………… 165
3.7.1 Dielectric Layer above a Metallic Surface …………………… 166
3.7.2 Surface Waves along a Dielectric Fiber ……………………… 176
3.8 Transients in Waveguides ……………………………………………… 181
iii
3.9 Waveguide Based Cavities …………………………………………… 185
3.9.1 Power and Energy Considerations …………………………… 185
3.9.2 Rectangular Waveguide Cavity ……………………………… 187
3.9.3 Cylindrical Resonator ……………………………………… 191
3.9.4 Open Resonator – Rectangular Geometry (TM - Mode) ……… 193
3.9.5 Exciting a Cavity ……………………………………………… 194
3.10 Wedge in a Waveguide ……………………………………………… 197
3.10.1 Metallic Wedge ……………………………………………… 197
3.10.2 Dielectric Wedge ……………………………………………… 204
3.11 Appendix ……………………………………………………………… 207
4 Matrix Formulations 211

4.1 Impedance Matrix …………………………………………………… 211
4.2 Scattering Matrix ……………………………………………………… 215
4.3 Transmission Matrix …………………………………………………… 220
4.4 Wave-Amplitude Transmission Matrix ……………………………… 221
4.5 Loss Matrix …………………………………………………………… 223
4.6 Directional Coupler …………………………………………………… 224
iv
4.7 Coupling Concepts …………………………………………………… 225
4.7.1 Double Strip Line: Parameters of the Line …………………… 225
4.7.2 Equations of the System ……………………………………… 230
4.7.3 Simple Coupling Process ……………………………………… 233
5 Nonlinear Components 237

5.1 Transmission Line with Resistive Nonlinear Load …………………… 237
6 Periodic Structures 241

6.1 The Floquet Theorem ………………………………………………… 242
6.2 Closed Periodic Structure ……………………………………………… 252
6.2.1 Dispersion Relation …………………………………………… 254
6.2.2 Modes in the Groove …………………………………………… 259
6.2.3 Spatial Harmonics Coupling …………………………………… 261
6.3 Open Periodic Structure ……………………………………………… 264
6.3.1 Dispersion Relation …………………………………………… 266
6.4 Transients ……………………………………………………………… 271
v
7 Generation of Radiation 276
7.1 Single-Particle Interaction …………………………………………… 277
7.1.1 Infinite Length of Interaction ………………………………… 277
7.1.2 Finite Length of Interaction …………………………………… 280
7.1.3 Finite Length Pulse …………………………………………… 281
7.1.4 Cerenkov Interaction ………………………………………… 282
7.1.5 Compton Scattering: Static Fields …………………………… 284
7.1.6 Compton Scattering: Dynamic Fields ………………………… 285
7.1.7 Uniform Magnetic Field ……………………………………… 286
7.1.8 Synchronism Condition ……………………………………… 287
7.2 Radiation Sources: Brief Overview …………………………………… 290
7.2.1 The Klystron ………………………………………………… 290
7.2.2 The Traveling Wave Tube ……………………………………… 291
7.2.3 The Gyrotron ………………………………………………… 293
7.2.4 The Free Electron Laser ……………………………………… 295
7.2.5 The Magnetron ………………………………………………… 297
7.3 Generation of radiation in a waveguide ……………………………… 299
7.4 Generation of radiation in a cavity …………………………………… 308
vi
Chapter 1: Transmission Lines
In this chapter we shall first recapitulate some of the topics learned in the
framework of the course ``Waves and Distributed Systems'' and then we shall
extend the analysis to topics that are of
importance to microwave devices. But first a
few examples:
1
1.1 Simple Model
First we shall examine the propagation of an w x z
electromagnetic wave between two parallel plates located
at a distance a one of the other as illustrated in the y
figure. The principal assumptions of this simple model
are as follows: a
1. No variation in the x direction i.e. ∂ x = 0 .
2. Steady state e.g. exp ( jωt ) .
3. The distance between the two plates ( a ) is very small so that even if there is any
(field) variation in the y direction, it is negligible on the scale of the wavelength ( a λ )
∂ ∂ ∂
⇒ ∼ 0. (1.1.1)
∂y ∂z ∂y
4. The
constitutive
relations
of the vacuum:
D = ε 0 E , B = µ0 H where µ0 = 4π ×10 −7 [Henry/meter] and
ε 0 = 8.85 ×10 −12 [Farad/Meter].
w x z
Based on the assumptions above, ME may be simplified.
y
a
2

(a) Gauss' law ∇ ⋅ E = 0 ∂ z Ez = 0 ⇒ Ez = const. we conclude that Ez is uniform
between the two plates. Imposing next the boundary conditions on the two plates
Ez ( y = 0) = 0 Ez ( y = a ) = 0 (1.1.2)
which means that the longitudinal electric field vanishes ( Ez ≡ 0) .

(b) In a similar way the magnetic induction satisfies ∇ ⋅ B = 0 and it may be shown
that the longitudinal component of the magnetic induction vanishes ( Bz = 0) .

(c) Faraday's equation reads ∇ × E = − jω B thus explicitly
1x 1y 1z 1x : −∂ z E y = − jω Bx

0 0 ∂ z = − jω B ⇒ 1y : ∂ z Ex = − jω By (1.1.3)
Ex E y 0 1z : 0 = 0
There is no variation in the y direction therefore since E x = 0 for both y = 0
and y = a , as in the case of Ez , we have E x ≡ 0 therefore By = 0 thus
∂
E y = jωµo H x . (1.1.4)
∂z
3
w x z

(d) Ampere's law reads ∇ × H = jω D , or explicitly taking
y
advantage of the vanishing components we get
1x 1y 1z 1x : 0 = 0 a

0 0 ∂ z = jω D ⇒ 1y : ∂ z H x = jω Dy (1.1.5)

Hx 0 0 1z : 0 = 0
hence
∂
H x = jωε o E y . (1.1.6)
∂z
From these two equations [(1.1.4) and (1.1.6)] it can be readily seen that we obtain the
wave-equation for each one of the components:
∂ 
H x = jωε o E y 
∂z   ∂2 ω 2 
 ⇒  2 + 2  Ey = 0 (1.1.7)
∂  ∂ z c 
E y = jωµo H x 
∂z 
which has a solution of the form
 ω   ω 
E y = A exp  − j z  + B exp  j z  (1.1.8)
 c   c 
4
w x z
It is convenient at this point to introduce the notation in terms of
y
voltage and current. The voltage can be defined since ∫ ⋅ d = 0 ; it
E
reads a
V ( z ) = − E y ( z )a. (1.1.9)
In order to define
the current
we recall that based on the boundary conditions we have

n × ( H1 − H 2 ) = K where K is the surface current. Consequently, denoting by w the
height of the metallic plates, the local current is I = K z w or
I ( z ) = H x ( z )w. (1.1.10)
Based on these two equations [(1.1.9)–(1.1.10)] it is possible to write
∂ ∂  V I ∂  a
E y = jωµo H x ⇒  −  = jωµ ⇒ V = − jω  µo  I
∂z ∂z  a  ∂z
o
w  w
(1.1.11)
∂ ∂ I   V ∂  w
H x = jωε o E y ⇒   = jωε o −  ⇒ I = − jω ε o V
∂z ∂z  w   a ∂z  a
The right hand side in both lines of (1.1.11) represent the so-called transmission line
equation also known as telegraph equations.
5
w x z
d
V ( z ) = − jω L I ( z ) y
dz
(1.1.12)
d a
I ( z ) = − jωC V ( z )
dz
C being the capacitance per unit length whereas L is the inductance per unit length;
a w
L = µo and C = ε o . As expected, these two equations lead also to the wave equation
w a
dV   d2 2
= − jω LI ( z )  + β
dz   dz 2 V ( z ) = 0
   (1.1.13)
dI
= − jωCV ( z )  β = ω LC
dz 
− jβ z jβ z
The general solution is V ( z ) = Ae + Be and correspondingly, the expression for
the current is given by
−1 dV β  − jβ z jβ z 
I ( z) = = Ae − Be (1.1.14)
jω L dz ω L 
 

β ω LC L
defining the characteristic impedance Z c−1 = = or Z c ≡ , we get
ωL ωL C
6
1 − jβ z + jβ z
I ( z) = [ Ae − Be ]. (1.1.15)
Zc
In the specific case under consideration
a
µo
Zc = w = η a , β = ω LC = ω . (1.1.16)
o
w w c
εo
a
1.2 Coaxial Transmission Line
As indicated in the previous case, two parameters are

to be determined: the capacitance per unit length (C )
and the inductance per unit length ( L) . According to
(1.1.11) these two parameters can be determined in
static conditions. We determine next the capacitance 2R int
per unit length of a coaxial structure. For this
purpose it is assumed that on the inner wire a voltage 2R ext
Vo is applied, whereas the outer cylinder is grounded.
7
Consequently, the potential is given by
ln( r / Rext )
ϕ ( r ) = Vo . (1.2.1)
ln( Rint / Rext )
and the corresponding electric field associated with this potential is
∂ϕ 1 1
Er = − = −Vo . (1.2.2)
∂r r ln( Rint / Rext )

The charge per unit surface at r = Rext is calculated based on n ⋅ ( D1 − D2 ) = ρ s and it is
given by
V 1
ρs = ε 0 0 . (1.2.3)
Rext ln( Rext / Rint )
Based on this result, the charge per unit length ( ∆ z ) may be expressed as
Q V 1 ε oVo 2π
= ρ s 2π Rext = ε 0 o 2π Rext = . (1.2.4)
∆z R 
Rext ln( Rext /Rint )
ln ext 
 Rint 
Consequently, the capacitance per unit length is given by
8
Q/∆ z 2πε o
C≡ = . (1.2.5)
Vo ln( Rext /Rint )
In a similar way, we shall calculate the inductance per unit length. Assuming that
the inner wire carries a current I , based on Ampere law the azimuthal magnetic field is
I
Hφ ( r ) = o . (1.2.6)
2π r
With this expression for the magnetic field, we can calculate the magnetic flux. It is given
by
Rext I R
Φ = µo ∆ z ∫ drH φ (r ) = µo ∆ z ln ext . (1.2.7)
Rint 2π Rint
The inductance per unit length ∆ z is
Φ/∆ z µo  Rext 
L≡ = ln  . (1.2.8)
I 2π  Rint 
Io
∆z
To summarize the parameters of a coaxial transmission line
L 1  Rext 
Zc = = ηo ln 
C 2π  Rint 
(1.2.9)
ω
β=
c
Exercise 1.1: Determine Z c and β for a coaxial line filled with a material ( ε r , µ r )?
10
w x z
1.3 Low Loss System
y
Based on Ampere's law we obtained a
d
∇ × H = jωε oε E I ( z ) = − jωCV ( z ),
dz
(1.3.1)
where we assumed a line without dielectric ( ε r ) and Ohm (σ ) loss. In the case of
dielectric loss we have
ε r = ε ′ − jε ′′ (1.3.2)
or in our case
jω C → jω C + G ≡ Y . (1.3.3)
In a similar way based on Faraday's law
d
∇ × E = − jωµo µ r H ⇒ V ( z ) = − jω LI ( z ) (1.3.4)
dz
and the magnetic losses
µ r = µ r′ − j µ r′′
allows us to extend the definition according to
jω L jω L + R ≡ Z (1.3.5)
hence the equations
11
w x z
d
I ( z ) = −YV ( z ) y
dz
(1.3.6)
d a
V ( z ) = − ZI ( z )
dz
may be conceived as a generalization of the transmission line equations in the presence of
loss. The characteristic impedance for small loss line is
Z L  G R 
Zc = 1 + j  −  (1.3.7)
Y C  2ωC 2ω L  
and the wave number
γ = α + jβ
R GZ o
α + (1.3.8)
2Z o 2
 RG G2 R2 
β ω LC 1 − 2 + 2 2 + 2 2
 4ω LC 8ω C 8ω L 
Exercise 1.2: Prove the relations in Eq. (1.3.8).
12
1.4 Generalization of the Transmission Line Equations
The fundamental assumptions of the analysis are:

(i) TEM, (ii) the wave propagates in the z direction, (iii) we distinguish between
∂
longitudinal ( z ) and transverse ( ⊥ ) components ∇ = ∇ ⊥ + 1z .
∂z
From Faraday law, ∇ × E = − jωµo µ r H , we obtain

1x 1y 1z 1x ( − ) ∂ z E y

∂ x ∂ y ∂ z = 1y ∂ z Ex (1.4.1)

Ex E y 0 1z ∂ x E y − ∂ y Ex
thus

1x : −∂ z E y = − jωµo µ r H x  ∂ E⊥
 1 : 1z × = − jωµ o µ r H ⊥
1y : +∂ z E x = − jωµo µ r H y  ⊥ ∂z (1.4.2)

1z : ∂ x E y − ∂ y E x = 0  1z : ∇ ⊥ × E⊥ = 0.
13
In a similar way, from Ampere's law we have

 ∂H ⊥
1 : 1 × = jωε oε r E⊥
∇ × H = + jωε oε r E ⇒  ⊥ z ∂z (1.4.3)

1 : ∇ × H
 z ⊥ ⊥ = 0.
Fromthe two curl equations
∇ ⊥ × E⊥ = 0 ⇒ E⊥ = g ( z )∇ ⊥ϕ ( x, y ) ⇒ ∇ 2⊥ϕ = 0
(1.4.4)
∇ ⊥ × H ⊥ = 0 ⇒ H ⊥ = h( z )∇ ⊥ψ ( x, y ) ⇒ ∇ ⊥ψ = 0,
2
we conclude that the transverse variations of the transverse field components are
determined by 2D Laplace equation justifying the use of DC quantities adopted above
(capacitance and inductance per unit length). From the other two equations we get the
wave equation

∂  ∂E⊥  ∂H ⊥
1
 z ×  = − jωµ µ
∂z  ∂z  ∂z
o r
(1.4.5)
 ∂ 2 E⊥   ∂H ⊥ 
1z × 1z × 2 
= − jωµ µ 1
o r  z ×  = − jωµo µ r [ jωε oε r E⊥ ],
 ∂z   ∂z 
or explicitly
14
 ∂2 ω 2 
 ∂z 2 + c 2 µ rε r  E⊥ = 0. (1.4.6)
 
The last equation determines the dynamics of g ( z ) [see (1.4.4)] and
− jβ z
the solution has the form g ( z ) = e where β = (ω / c ) ε r µ r . Note that

∂E⊥ µo µ r
1z × = − jωµo µr H ⊥ ⇒ 1z × E⊥ = H
∂z εo εr ⊥
(1.4.7)
∂H ⊥ ε ε
1z × = jωε oε r E⊥ ⇒ 1z × H ⊥ = − o r E⊥
∂z µo µr
15
V = V0
s1
x
As in the previous two cases we shall see next how y V= 0
the electric parameters can be calculated in the s2
general case and for the sake of simplicity we shall z
assume that the medium has uniform transverse and
longitudinal properties. The electric field in the entire space is given by
− jβ z
E⊥ = −(∇ ⊥ϕ )e whereas the magnetic field is
εo εr − jβ z
H⊥ = ( − 1z ×∇ ⊥ϕ )e . (1.4.8)
µo µ r
Note that associated to this electric field, one can define the voltage
s2 s2 s2
Vo = − ∫ E⊥ ⋅d = ∫ ∇ ⊥ϕ ⋅ d = ∫ dϕ (1.4.9)
s1 s1 s1
− jβ z
such that V ( z ) = Vo e . On the two (ideal) conductors the electric field generates a
surface charge given by

ρ s = ε o ε r n ⋅ E⊥ , (1.4.10)
therefore the charge per unit length is
Q
∆z
= ∫ dl ρ s = ∫ dl ε oε r ( n ⋅ E⊥ ). (1.4.11)
Since by virtue of linearity of Maxwell's equations
16
V = V0
s1
x
y V=0
s2
the charge per unit length is proportional to the applied z
Q
voltage = CV0 we get
∆z
1
V0 ∫
C = ε oε r d ( n ⋅ E⊥ ). (1.4.12)
In a similar way, the magnetic field generates on the metallic electrode (wire) a surface
current given by
J s = n × H⊥. (1.4.13)
ε oε r
Since it was shown that H ⊥ = 1z × E⊥ we conclude that
µo µ r
ε oε r ε oε r
J s = n × H ⊥ = n × [1z × E⊥ ] = (n ⋅ E⊥ )1z (1.4.14)
µo µ r µo µr
hence the total current is
ε oε r
I 0 = ∫ J s ⋅ 1z dl =
µo µ r
(
∫ dl n ⋅ E⊥ ) (1.4.15)
At this point rather than calculating the inductance per unit length we combine the
previous result for the charge per unit length and (1.4.15) the result being
17
V = V0
s1
x
y V=0
s2
ε oε r z
Io µ µ
o r
∫
dl n ⋅ E⊥
c
= = . (1.4.16)
Q/∆ z ε oε r ∫ dl n ⋅ E⊥ µr ε r
However, having established this relation between the current and the charge per unit
length we may use again the linearity of Maxwell's equations and express Q / ∆ z = CV0 .
Substituting in Eq. (1.4.16) we get
I0 c
= (1.4.17)
CVo µrε r
but by definition
V0
= Zc , (1.4.18)
I0
which finally implies that
1 c
= .
CZ c µrε r
This result leads to a very important conclusion namely, in a transmission line of
uniform electromagnetic properties it is sufficient to calculate the capacitance per
unit length. Bearing in mind that Z c = L / C we find that once C is established,
18
V = V0
s1
x
y V=0
s2
µrε r z
L= . (1.4.19)
Cc 2
It is important to re-emphasize that this relation is valid only if the electromagnetic
properties (ε r , µ r ) are uniform over the cross-section.
Exercise 1.3: Calculate the capacitance per unit length of two wires of radius R which
are at a distance d > 2 R apart.
Another quantity that warrants consideration is the average power
{ ε oε r
}
* *
1
2
{ ( 1
2
) } 
P = Re ∫ dxdy E⊥ × H ⊥ ⋅ 1z = Re ∫ dxdy E⊥ × 1z × E⊥ ⋅ 1z


 µo µ r
( )
* ε oε r ε oε r * 
1
= Re
2
{ ∫ dxdy E⊥ ⋅ E⊥ } =
2
µ o µ r ε oε r µo µ r
1
 4 ∫ dxdy ε o ε r ⊥ ⋅ E⊥ 
E

(1.4.20)
2 c
= We = [We + Wm ].
ε o µ oε r µ r ε r µr
19
Exercise 1.4: In the last expression we used the fact that We = Wm -- prove it.
1 1
Exercise 1.5: Show that the power can be expressed as P = Z c | I o |2 = Vo I o* .
2 2
Finally, we may define the energy velocity as the average power propagating along the
transmission line over the total average energy per unit length
P c
Ven = = .
We + Wm µrε r
Exercise 1.6: Show that the material is not frequency dependent, this quantity equals
exactly the group velocity. What if not? Namely ε r (ω ) .
20
1.5 Non-Homogeneous Transmission Line
There are cases when either the electromagnetic properties or the

geometry vary along the structure. In these cases the impedance per unit
length ( Z ) and admittance (Y ) per unit length are z -dependent i.e.
dV ( z )
= −Z ( z) I ( z)
dz
(1.5.1)
dI ( z )
= −Y ( z )V ( z ).
dz
As a result, the voltage or current satisfy an equation that to some extent differs from the
regular wave equation
d 2V dZ ( z ) dI ( z )
= − I ( z ) − Z ( z )
dz 2 dz dz
(1.5.2)
d  dV
=  ln [ Z ( z ) ] + Z ( z )Y ( z )V ( z )
 dz  dz
A solution of a general character is possible only using
numerical methods. However, an analytic solution is
possible if we assume an `èxponential'' behavior of the
21
form
Z ( z ) = jω Lo exp ( qz )
(1.5.3)
Y ( z ) = jωCo exp ( − qz )
Substituting these expressions in Eq. (1.5.2) we get
d 2V ( z ) dV
2
− q + ω 2
LoCoV = 0 (1.5.4)
dz dz
−γ z
therefore assuming a solution of the form V ( z ) = Vo e 1 we conclude that
1
γ 1 = −  q ± q 2 − 4ω 2 LoCo . (1.5.5)
2 
In a similar way, the equation for the current is given by
d 2 I dI d
− ln[Y ( z )] − YZI ( z ) = 0. (1.5.6)
dz 2 dz dz
−γ z
Assuming a solution of the form I = I o e 2 we obtain
1
γ 2 =  q ± q 2 − 4ω 2 LoCo . (1.5.7)
2 
It is convenient to define
q2 / 4
ωc ≡ 2 2 ,
2
(1.5.8)
LoCo
22
which sets a ``cut-off'' in the sense that for ω < ωc both γ 1 and γ 1 are real.
The second important result is that the impedance along the transmission
line
−γ z
V ( z ) Vo e 1 qz
= = Z (0) e
I ( z ) I o e−γ 2 z
c
(1.5.9)
is frequency independent.
Exercise 1.7: Plot the average power along such a transmission line as well as the
average electric and magnetic energies. What is the energy velocity?
23
1.6 Coupled Transmission Lines
Microwave or high frequency circuits consist

typically of many elements connected usually with wires that
may be conceived as transmission lines. The proximity of
one line to another may lead to coupling phenomena. Our purpose in this section is to
formulate the telegraph equations in the presence of coupling. With this purpose in mind
let us assume N transmission lines each one of which is denoted by an index
n(= 1,2 N ) -- as illustrated in the figure above. Ignoring loss in the system we may
conclude that the relation between the charge per unit length of each ``wire'' is related to
the voltages by
N
Qν
= ∑ Cν nVn (1.6.1)
∆ z n=1
Cν ,n being the capacitance matrix per unit length. In a similar way, it is possible to
establish the inductance matrix per unit length relating the voltage on wire ν with all the
currents
Φν N
= ∑ Lν n I n . (1.6.2)
∆ z n =1
24
Having these two equations [(1.6.1)–(1.6.2)] in mind, we may naturally extend the
telegraph equations to read
M M
d d
Vν ( z ) = − jω ∑ Lν n I n ( z ) Iν ( z ) = − jω ∑Cν nVn ( z ). (1.6.3)
dz n =1 dz n =1
Subsequently we shall discuss in more detail phenomena linked to this coupling process
however, at this point we wish to emphasize that the number of wave-numbers ( β 2 )
corresponds to the number of ports. This is evident since
− jβ z − jβ z
V ( z ) = V0 e I ( z ) = I0 e (1.6.4)
enabling to simplify
(1.6.3) to read

βV0 = ω LI 0 β I 0 = ωCV0 (1.6.5)
= =
thus the wavenumber is the non-trivial solution of
[ω 2 LC − β 2 δ ]V0 = 0 or [ω 2 C L − β 2 δ ]I 0 = 0. (1.6.6)
= = = = = =
wherein δ is the unity matrix. Clearly the normalized wave number β 2 ≡ ( β c / ω ) are
2
=
the eigen-values of the matrix LC ( = C L since both matrices are symmetric) and if the
= = = =
dimension of C and L is N then, the number of the eigen wavenumber is also N .
= =
25
1.7 Microstrip
In this section we shall discuss in some detail

some of the properties of a microstrip which is an
essential component in any micro-electronic as well
as microwave circuit. The microstrip consists of a ∆
thin and narrow metallic strip located on a thicker y
dielectric layer. On the other side of the latter, there h
εr
is a ground metal; the side walls have been w x
introduced in order to simplify the analysis and the
width w is large enough such that it does not affect the physical processes in the vicinity
of the strip. We shall examine a simplified model of this system and for this purpose we
make the following assumptions: (i) The width of the device is much larger than the
height ( w h ) and the width ( w ∆ ) of the strip.
(ii) The charge on the strip is distributed uniformly.
Our goal is to calculate the two parameters of the transmission line: capacitance
and inductance per unit length. With this purpose in mind we shall start with
evaluation of the capacitance therefore let us assume a general charge distribution on the
26
∆
y
h εr
w x
strip
 w ∆
η ( x ) | x − |<
ρ s ( x) =  2 2 (1.7.1)
 0 w ∆
| x − |>
 2 2
With the exception of y = h the potential is given by
 ∞  π nx   π ny 
∑ n  w 
A sin sinh   0≤ y≤h
 n =0    w 
ϕ ( x, y ) =  π n ( y −h ) (1.7.2)
 −
∞
 π nx 
∑ n  B sin e
w y ≥ h.
 n =0  w 
The continuity of the potential at y = h implies
 x  π nh   π nx 
φ ( x, y = h) = ∑ An sin  π n  sinh   = ∑ Bn sin  
n =0  w  w  n  w 
(1.7.3)
 π nh 
⇒ An sinh   = Bn .
 w 
27
∆
y
h εr
w x
The electric induction D y is discontinuous at this plane. In each
one of the two regions the field is given by
 π n   π nx   π ny 
Dy ( x,0 < h) = −ε oε r ∑ An  sin
   cosh  
n  w   w   w 
(1.7.4)
 π n   π nx   π n 
Dy ( x, y > h) = ε o ∑ Bn   sin   exp  − ( y − h) .
n  w   w   w 

With these expressions, we can write the boundary conditions i.e., n ⋅ ( D1 − D2 ) = ρ s in
the following form
 w ∆
 η ( x ) | x − |<
 π n   π nx    π nh    2 2
ε o ∑   sin    Bn + ε r An cosh  = (1.7.5)
n  w   w   w   w ∆
0 | x − |> .
 2 2
Using the orthogonality of the sin function we obtain [for this reason the two side walls
were introduced]
w+∆
 π nh  1 2 2 dxη ( x)sin  π nx .
Bn + ε r An cosh 
 w


=
ε o π n ∫ w − ∆ 
 w


(1.7.6)
2
28
∆
y
h εr
w x
The next step is to substitute (1.7.3) into the last expression. The
result is
w+∆
  π nh   π nh   1 2 2 dxη ( x )sin  π nx .
An sinh 
  w 
+
 r ε cosh   =
 w   εo π n
∫ w − ∆ 
 w 
 (1.7.7)
2
Consequently, subject to the assumption that η ( x ) is known, the potential is known in the
entire space and specifically at y = h is given by
w+∆
 π nx  1 1 2  π nx′ 
φ ( x, y = h) = ∑ sin   ∫ 2 ′ ′
dx η ( x )sin  . (1.7.8)
n  w  1 + ε ctanh  π nh  ε o π n w − ∆  w 
r   2
 w 
In principle, this is an integral equation which can be solved numerically since the
potential on the strip is constant and it equals V0 ; many source solution.
At this point we shall employ our second assumption namely that the charge is
uniform across the strip and determine an approximate solution. The first step is to
average over the strip region, | x − w / 2 |≤ ∆ / 2 . The left hand side is by definition constant
thus
29
∆
y
h εr
w x
w+∆
1
Vo = ∫ 2 dx φ ( x, y = h)
∆ w−∆
2
 w+∆  w+∆ 
=∑
(1/ε o )(2/π n)  1 2 dxsin  π nx    2 dx′η ( x′)sin  π nx′  .
 π   ∆ ∫w − ∆     ∫w − ∆  
1 + ε r ctanh 
nh  w   w 
  
n
 w  2  2 
(1.7.9)
Explicitly our assumption that the charge is uniformly distributed, implies η Q / ∆ z ∆ ,
therefore
2
 w+∆ 
Q 1 2 1 1 2 dxsin  π nx  
Vo = ∑
∆z ε0 n π n  π nh   ∆ ∫w − ∆

 w 

1 + ε r ctanh    
 w  2 
and finally the capacitance per unit length is
30
∆
y
h εr
w x
Q/∆ z ε oπ /2
C= (1.7.10)
Vo 1/n 2 π n  2  π n∆ 
∑  π nh 
sin 
 2
 sinc 


 2w 
n
1 + ε r ctanh  
 w 
With this expression we can, in principle calculate all the parameters of the
microstrip. For evaluation of the inductance per unit length we use the fact that the
dielectric material cannot have any impact on the DC inductance. Moreover, we know
that in the absence of the dielectric ( ε r = 1), the propagation number is ω / c and the
characteristic impedance satisfies
1 1 L(ε r = 1)
Zc = 1= . (1.7.11)
C (ε r = 1) c C (ε r = 1)
Since the DC the magnetic field is totally independent of the dielectric coefficient of the
medium (electric property), we deduce from the expression of above that
1
⇒ L(ε r = 1) = 2 , (1.7.12)
c C (ε r = 1)
or explicitly
31
∆
y
h εr
w x
1  h 2 π ∆
∞
2  h  
L = µo ∑  exp
π ν =0 2ν + 1    −π (2ν + 1)
w sinh 

π (2ν + 1)  
w 
sinc 
2 w
(2ν + 1) 

. (1.7.13)
With the last expression and (1.7.10) we can calculate the characteristic impedance of the
microstrip
1/2
 − hν  1/2
L 2 e sinh(hν )sinc (∆ν )
2
 ∞
sinc (∆ν )
2

Zc = = ηo  ∑  ∑  , (1.7.14)
C π ν =0 2ν + 1  ν =0 (2ν + 1)[1 + ε r ctanh(hν )] 
 
π ∆
where hν ≡ π (2ν + 1)h / w and ∆ν ≡ (2ν + 1) . The next parameter that remains to be
2w
determined is the phase velocity. Since L and C are known, we know that β = ω LC
implying that
∞
sinc 2 (∆ν ) 1
1
∑ 2ν + 1 1 + ε ctanh(h )
ν =0 r ν
Vph = =c ∞ . (1.7.15)
LC sinc (∆ν )
2
∑
ν =0 2ν + 1
exp ( − hν ) sinh(hν )
32
∆
y
h εr
w x
Contrary to cases encountered so far the dielectric material fills only part of the entire
volume. As a result, only part of the electromagnetic field experiences the dielectric. It is
therefore natural to determine the effective dielectric coefficient experienced by the
field. This quantity may be defined in several ways. One possibility is to use the fact that
c
when the dielectric fills the entire space we have the phase velocity in Vph = it
εr
c2
becomes natural to define the effective dielectric coefficient as ε eff ≡ 2 thus
Vph
∞
sinc 2 (∆ν )
∑ 2ν + 1
exp ( − hν ) sinh(hν )
ε eff = ν =0∞ . (1.7.16)
sinc (∆ν )
2
1
∑
ν =0 2ν + 1 1 + ε r ctanh(hν )
The following figures illustrate the dependence of the various parameters on the
geometric parameters.
33
70 0.40 7.2
7.0
60
0.39
Zc [ohm]
6.8
Vph / c
ε eff⋅
50
6.6
0.38
40
6.4
30 0.35 6.2
1.0 2.0 3.0 4.0 1.0 2.0 3.0 4.0 1.0 2.0 3.0 4.0
∆[mm ] ∆[mm ] ∆[mm ]
(a) (b) (c)
(a) Characteristic impedance vs. ∆ ; h = 2 mm, w = 20 mm, ε r = 10 and ν < 100 .

(b) Phase velocity vs. ∆ ; h = 2 mm , w = 20 mm, ε r = 10 and ν < 100 .
(c) Effective dielectric coefficient vs. ∆ ; h = 2 mm , w = 20 mm, ε r = 10 and ν < 100 .
34
70 0.41 7.2
7.0
60 0.40
6.8
Zc [ohm]
Vph / c
ε eff⋅
50 0.39
6.6
40 0.38
6.4
30 0.37 6.2
1.0 2.0 3.0 4.0 1.0 2.0 3.0 4.0 1.0 2.0 3.0 4.0
h [mm ] h [mm ] h [mm ]
(a) (b) (c)
(a) Characteristic impedance vs. the height h ; ∆ = 2 mm, w = 20 mm, ε r = 10 , ν < 100 .
(b) Phase velocity vs. the height h ; ∆ = 2 mm , w = 20 mm, ε r = 10 , ν < 100 .
(c) Effective dielectric coefficient vs. the height h ; ∆ = 2 mm, w = 20 mm, ε r = 10 ,
ν < 100 .
35
Finally the figure below shows several alternative configurations
36
Exercise 1.8: Determine the effective dielectric coefficient relying on energy
confinement.
Exercise 1.9: What fraction of the energy is confined in the dielectric and how the
various parameters affect this fraction?
Exercise 1.10: Examine the effect of the dielectric coefficient on Z c , ε eff and V ph .
Compare with the case where the dielectric fills the entire space. (For solution see
Appendix 11.1)
Exercise 1.11: Show that if w h, ∆ the various quantities are independent of w .
Explain!! (For solution see Appendix 11.2)
Exercise 1.12: Analyze the effect of dielectric and permeability loss on a micro-strip.
Exercise 1.13: Calculate the ohmic loss. Analyze the effect of the strip and ground
separately.
Exercise 1.14: Determine the effect of the edges on the electric parameters ( L, C ) .
37
1.8 Stripline
Being open on the top side, the microstrip has limited ability to confine the
electromagnetic field. For this reason we shall examine now the stripline which has a
metallic surface on its top. The basic configuration of a stripline is illustrated below
ε
y
x h d
The model we shall utilize first replaces the central strip with a wire as illustrated below
and as in Section 1.7 our goal is to calculate the parameters of the line.
ε
y
x h d
38
ε
y
x h d
For evaluation of the capacitance per unit length it is first assumed that the charge density
is given by
Q
ρ ( x, y ) = δ ( x )δ ( y − h ), (1.8.1)
∆z
and we need to solve the Poisson equation subject to trivial boundary conditions on the
two electrodes. Thus
ρ 
∇t2φ = −
ε oε r 
  π ny 
φ ( y = 0) = 0  ⇒ φ ( x, y ) = ∑φn ( x)sin  . (1.8.2)
 d 
φ( y = d) = 0
n


Substituting the expression in the right hand side in the Poisson equation we have
 d2 πn 
2
  π ny  −Q
∑n  dx 2 n  d  n   d  ε ε ∆ δ ( x )δ ( y − h)
φ ( x ) − φ ( x ) sin = (1.8.3)
  o r z
and the orthogonality of the trigonometric function we obtain
39
ε
y
x h d
 d2 πn   2
−Q 2  π nh 
 2 − 
 n φ ( x ) = δ ( x ) sin   = −Qnδ ( x ), (1.8.4)
 dx  d   ε ε ∆
o r z d  d 
where
2Q  π nh 
Qn ≡ sin  .
ε oε r d ∆ z  d 
The solution of (1.8.4) is given by
  x
A
 n exp  −π n  x>0
  d
φn ( x) =  (1.8.5)
 B exp  π n x  x<0
 n  
 d
and since the potential has to be continuous at x = 0 then
An = Bn , (1.8.6)
integration of (1.8.4) determines the discontinuity:
dϕ n dϕ πn 
− n = −Qn ⇒ −   [ An + Bn ] = −Qn . (1.8.7)
dx x =0+ dx x =0−  d 
From (1.8.6) and (1.8.7) we find
40
ε
y
x h d
Qn Q h  h
An = Bn = = sinc  π n . (1.8.8)
 π n  ε oε r ∆ z d  d
2 
 d 
This result permits us to write the solution of the potential in the entire space as
 −
π nx
 Q h ∞
 π nh   π ny 
ε ε ∆ d

∑
n =1
sinc   sin 
 d   d 
e
d x≥0
φ ( x, y ) =  o r z (1.8.9)
 +
π nx
 Q h sinc  π nh  sin  π ny  e d
∞
 ε oε r ∆ z d ∑     x ≤ 0.
n =1  d   d 
At this stage we can return to the initial configuration and assume that the central
strip is a superposition of charges Qi located at xi and since the system is linear, we
apply the superposition principle thus
h/d  π nh   π ny   | x − xi | 
φ ( x, y ) = ∑  d   d  ∑i i 
ε oε r ∆ z n
sinc sin Q exp −π n
d 
. (1.8.10)
In the case of a continuous distribution we should replace
41
∆
ε
y
−
π n|x− x | −
π n|x− x ′| x h d
i 1
∑ Qi e d
i
=
∆ ∫ dx ′Q ( x ′)e d (1.8.11)
and consequently
h/d  π nh   π ny  1  πn 
φ ( x, y ) = ∑
ε oε r ∆ z n
sinc  sin
 
 d   d ∆
 ∫ dx ′Q ( x ′) exp −
 d
| x − x′ | , (1.8.12)

which again leads us to an integral equation; note that the surface changes density as
ρ s ( x) = Q( x) / ∆∆ z . As in the microstrip case, we shall assume uniform distribution
therefore
h/d  π nh   π ny  Qo ∆/2 ′  πn 
φ ( x, y ) = ∑  d   d  ∆ ∫− ∆/2
ε oε r ∆ z n
sinc sin dx exp 

−
d
| x − x′ | . (1.8.13)

The potential is constant on the strip
42
∆
ε
y
x h d
1 ∆/2
∆ ∫− ∆/2
Vo = dx φ ( x, y = h)
−1 (1.8.14)
( h/d )Qo ∞  π nh  2 π nh  1 ∆ /2 1 ∆/2  π n 
= ∑ 
ε oε r ∆ z n =1  d   sin 
 d ∆
 ∫− ∆/2 dx ∫− ∆/2 dx′ exp  −
∆  d
| x − x′ | 

and the two integrals may be simplified to read
−1
1 ∆/2 1 ∆/2  πn   πn ∆    πn ∆   π n∆  
∆ ∫− ∆/2 ∆ ∫− ∆/2
dx dx ′ exp  − | x − x′ =
    1 − exp  −  sinhc  
 d   2 d   2 d  2d  
(1.8.15)
such that
 2h 
(h/d )Qo   ∞
 ∆  sinc2  π nh  1 − exp  − π n∆  sinhc  π n∆  .
Vo =
ε oε r ∆ z
∑
n =1

 d



 2 d



 2 d


(1.8.16)
The last result enables us to write the following expression for the capacitance per unit
length
−1
Qo /∆ z 1  ∆d    π nh  
C= = ε oε r  2  ∑ sinc2   1 − exp ( −ξ n ) sinhc(ξ n )   (1.8.17)
Vo 2  h  n  d  
43
whereas ξn = π n∆ / 2d , thus with it the characteristic impedance reads
1 2 h2  π nh 
Zc =
CV ph
= ηo ∑
ε r ∆d n
sinc2 
 d 
 1 − exp ( −ξ n ) sinhc(ξ n ) . (1.8.18)
The two frames show the impedance dependence on the width and height of the strip
100 200
80 150
Z c [ohm]
Z c [ohm]
60 100
40 50
20 0
1.0 2.0 3.0 4.0 1.0 2.0 3.0 4.0
∆[mm ] h[mm ]
Left: Characteristic impedance vs. the width ∆ [ h = 2 mm, ε r = 10 and n = 0,1,...200 ]

Right: Characteristic impedance vs. the height h [ ∆ = 2 mm, ε r = 10 and n = 0,1,...200 ]
44
Exercise 1.15: Determine the inductivity per unit length and analyse the dependence of
the various characteristics on the geometric parameters. (For solution see Appendix 11.3)
Exercise 1.16: Compare the dependence of the various characteristics of the stripline
and microstrip as a function of the geometric parameters.
Exercise 1.17: Compare micro-strip and strip-line from the perspective of sensitivity to
the dielectric coefficient. (For solution see Appendix 11.4)
Exercise 1.18: Determine the error associated with the assumption that the charge is
uniform across the strip.
Exercise 1.19: Analyze the effect of a strip of finite thickness. Remember that
throughout this calculation the strip was assumed to have a negligible thickness.
45
1.9 Resonator Based on Transmission Line
1.9.1 Short Recapitulation

Resonant circuits are of great importance for oscillator circuits, tuned amplifiers,
frequency filter networks, wavemeters for measuring frequency. Electric resonant circuits
have many features in common, and it will be worthwhile to review some of these by
using a conventional lumped-parameter RLC parallel network as an example Figure 8
illustrates a typical low-frequency resonant circuit. The resistance R is usually only an
equivalent resistance that accounts for the power loss in the inductor L and capacitor C
as well as the power extracted from the resonant system by some external load coupled to
the resonant circuit. One possible definition of
resonance relies on the fact that at resonance the I
input impedance is pure real and equal to R
implying Z in V R L C
P + 2 jω (Wm − We )
Z in = l . (1.9.1)
II * / 2
Although this equation is valid for a one-port
circuit, resonance always occurs when Wm = We , if we define resonance to be that
condition which corresponds to a pure resistive input impedance or explicitly
ω0 = 1/ LC ; note that these are the lumped capacitance (C ) and inductance ( L) .
46
An important parameter specifying the frequency selectivity, and performance in general,
of a resonant circuit is the quality factor, or Q . A very general definition of Q that is
applicable to all resonant (We = Wm ) systems is
ω (time − average energy stored in the system)
Q= 0 (1.9.2)
energy loss per second in the system.
hence,
Q = ω0 RC = R / ω0 L. (1.9.3)
In the vicinity of resonance, say ω = ω0 + ∆ω , the input impedance can be expressed in a
relatively simple form. We have
ω02 RL R
Zin = 2 = . (1.9.4)
ω0 L + j 2 R∆ω 1 + j 2Q( ∆ω / ω0 )
Z in
A plot of Zin as a function of ∆ω / ω0 is given below.
When | Zin | has fallen to 1/ 2 (half the power) of its R
maximum value, its phase is 45 if ω < ω0 and -45 if
0 .707 R
ω > ω0 thus
+ 90
∆ω ω0 Z in
2Q = 1⇒ ∆ω = (1.9.5)
ω0 2Q ∆ω
ω0

− 90
BW ∠ Z in
47
The fractional bandwidth BW between the 0.707 R points is twice this value, hence
ω 1
Q= 0 = . (1.9.6)
2∆ω BW
If the resistor R in Fig. 8 represents the loss in the resonant circuit only, the Q give by
(1.9.3) is called the unloaded Q . If the resonant circuit is coupled to an external load that
absorbs a certain amount of power, this loading effect can be represented by an additional
resistor RL in parallel with R . The total resistance is now less, and consequently the new
Q is also smaller. The Q , called the loaded Q and denoted QL , is
RRL / ( R + RL )
QL = . (1.9.7)
ω0 L
The external Q , denoted Qe , is defined to be the Q that would result if the resonant
circuit were loss-free and only the loading by the external load was present. Thus
R
Qe = L (1.9.8)
ω0 L
leading to
1 1 1
= + . (1.9.9)
QL Qe Q
Another parameter of importance in connection with a resonant circuit is the decay factor
τ . This parameter measures the rate at which the oscillations would decay if the driving
48
source were removed. Significantly, with losses present, the energy stored in the resonant
circuit will decay at a rate proportional to the average energy present at any time (since
Pl ∝ VV * and W ∝ VV * , we have Pl ∝ W ), so that
dW 2  t
= − W ⇒ W = W0 exp  −2  (1.9.10)
dt τ  τ 
where W0 is the average energy present at t = 0 . But the rate of decrease of W must
equal the power loss, so that
dW 2
− = W = Pl
dt τ
and consequently,
1 P ω P ω
= l = 0 l = 0. (1.9.11)
τ 2W 2 ω0W 2Q
Thus, the decay factor is proportional to the Q . In place of (1.9.10) we now have
 ω 
W = W0 exp  − 0 t . (1.9.12)
 Q 
49
l
Z in Zc , β , α
1.9.2 Short-Circuited Line
By analogy to the previous section, consider a short-circuited line of length l ,
parameters R, L, C per unit length, as in Fig. 10. Let l = λ0 / 2 at f = f 0 , that is, at
ω = ω0 . For f near f 0 , say f = f 0 + ∆f , β l = 2π f l / c = πω / ω0 = π + π∆ω / ω0 , since at
ω0 , β l = π . The input impedance is given by
tanh α l + j tan β l
Zin = Z c tanh( j β l + α l ) = Z c . (1.9.13)
1 + j tan β l tanh α l
But tanh α l α l since we are assuming small losses, so that
α l 1. Also tan β l = tan(π + π∆ω / ω0 ) = tan π∆ω / ω0 π∆ω / ω0
since ∆ω / ω0 is small. Hence
α l + jπ∆ω / ω0  ∆ω 
Zin = Z c ≈ Z c  α l + jπ  (1.9.14)
1 + jα lπ∆ω / ω0  ω0 
since the second term in the denominator is very small. Now

1
Z c = L / C , α = RYc = ( R / 2) C / L , and β l = ω0 LCl = π ; so
2
π / ω0 = l LC , and the expression for Zin becomes
50
R0
Z in L0
L l C  1
Zin =  R + j ∆ ωl LC  = Rl + jlL∆ω. C0
C 2 L  2
(1.9.15)
It is of interest to compare (1.9.15) with a series R0 L0 C0 circuit illustrated above. For this
circuit
Z in = R0 + jω L0 (1 − 1/ ω 2 L0C0 ).
If we let ω02 = 1 / L0C0 , then Z in = R0 + jω L0 (ω 2 − ω02 ) / ω 2 . Now if ω − ω0 = ∆ω is small
then
Z in R0 + 2 jL0 ∆ω. (1.9.16)
By comparison with (1.9.15), we see that in the vicinity of the frequency for which
l = λ0 / 2 , the short-circuited line behaves as a series resonant circuit with resistance
R0 = Rl / 2 and inductance L0 = Ll / 2 . We note that Rl , Ll are the total resistance and
inductance of the line; so we might wonder why the factors 1/ 2 arise: recall that the
current on the short-circuited line is half sinusoid, and hence the effective circuit
parameters R0 , L0 are only one-half of the total line quantities.The Q of the short-
circuited line may be defined as for the circuit
ωL ωL β
Q= 0 0 = 0 = . (1.9.17)
R0 R 2α
51
l
R0
Z in Zc ≡ L0
C0
1.9.3 Open-Circuited Line
By means of an analysis similar to that used earlier, it is readily verified that an open-
circuited transmission line is equivalent to a series resonant circuit in the vicinity of the
frequency for which it is an odd multiple of a quarter wavelength long. The equivalent
relations are
Z in α l + j (π / 2 ) ( ∆ω / ω0 )  Z c = Rl / 2 + j∆ω Ll
(1.9.18)
l = λ0 / 4, R0 = Rl / 2, L0 = Ll / 2, ω0 = 1/ L0C0
2
Comment: Note that formally from (1.9.13) in the lossless case i.e., Zin = jZ c tan β l we
conclude that there are many (infinite) resonances
since tan( β l ) vanishes for β l = π but also for
β l = π n , n = 1,2,3… corresponding to all the
``series'' resonances. In case of ``parallel'' resonances
the condition tan β l ±∞ is satisfied for
π
βl = + π n , n = 1,2,… . In practice, only the first
2
resonance is used since beyond that the validity of
the approximations leading to the equations are
questionable.
52
1.10 Pulse Propagation
1.10.1 Semi-Infinite Structure

So far the discussion has focused on solution of problems in the frequency domain.
In this section we shall discuss some time-domain features. Let us assume that at the
input of a semi-infinite and lossless transmission line we know the voltage pulse
V ( z = 0, t ) = V0 (t ) . In general in the absence of reflections
jωt − jβ (ω ) z
V ( z , t ) = ∫d ωV (ω ) e (1.10.1)
and specifically
jωt
V ( z = 0, t ) = ∫d ωV (ω ) e (1.10.2)
or explicitly, the voltage spectrum V (ω ) is the Fourier transform of the input voltage
1 ∞ − jωt 1 ∞ − jωt
2π ∫−∞ 2π ∫−∞
V (ω ) = dtV ( z = 0, t ) e = dtV 0 (t ) e . (1.10.3)
Consequently, substituting in Eq.(1.10.1) we get
∞ jωt − j β (ω ) z 1 ∞ ′ − jωt ′
V ( z, t ) = ∫ dω e
2π ∫−∞
dt V ( z = 0, t ′) e
−∞
(1.10.4)
∞ 1 ∞ ′
jω (t −t )− j β (ω ) z
= ∫ dt ′V ( z = 0, t ′)
2π ∫−∞
d ω e
−∞
53
ω
and in the case of a dispersionless line we have β (ω ) = ε r which leads to
c
∞  z   z 
V ( z , t ) = ∫ dt ′V ( z = 0, t ′)δ  t ′ − t + ε r  = V  z = 0, t ′ = t − ε r  (1.10.5)
−∞
 c   c 
implying that the pulse shape is preserved as it propagates in the z -direction. If the
phase velocity is frequency-dependent, then different frequencies propagate at different
velocities and the shape of the pulse is not preserved. As a simple example let us assume
that the transmission line is filled with gas
ω 2p
ε (ω ) = 1 − 2 . (1.10.6)
ω
Since
∞ 1  z 2
V ( z, t ) = ∫ dt ′V ( z = 0, t ′ )
2π ∫
d ω exp  jω (t − t ′) − ω 2
p − ω  (1.10.7)
−∞ c
it is evident that sufficiently far away from the input, the low frequencies (ω < ω p ) have
no contribution and the system acts as a high-pass filter.
The dispersion process may be used to determine the frequency content of a signal.
In order to envision the process let us assume that the spectrum of the signal at the input
is given by
54
M   ω − ω 2 
V (ω ) = ∑aν exp  −  ν
  (1.10.8)
v =1   ∆ων  
where all the parameters are known and that ∆ων / ων 1 ; note that at the limit that
∆ων / ων → 0 the Gaussian function behaves very similar to Dirac delta function. In such
a case it is convenient to expand
 1 ∂ε (ω ) 
ε (ω ) ε (ων ) +   (ω − ων ) (1.10.9)
 2 ε (ω ) ∂ ω ω =ω
ν
and also to use the fact that
1 ∂β ∂ ω  ε (ω ) ω 1 ∂ε (ω )
= = ε (ω ) = +
Vgr ∂ω ∂ω  c  c c 2 ε (ω ) ∂ω
(1.10.10)
1 ω 1 1 ∂ε (ω )
= + .
V ph c 2 ε (ω ) ∂ω
With these two observations, we now aim to develop an analytic expression for the
voltage at any location z and at any time t . Specifically, to demonstrate that the peak of
the Gaussian pulses depends on t = z / Vgr (ωr ) and reveal the way this signal varies in
space.
55
The voltage variation in time is given by
∞ jωt − j β (ω ) z
V ( z , t ) = ∫ dω V (ω ) e
−∞
∞ jωt − j β (ω ) z   ω − ων 

2 (1.10.11)
= ∑aν ∫ dω e exp  −   
ν
−∞
  ∆ων  
ω
wherein the wave number is β (ω ) = ε (ω ) therefore,
c
∞   ω − ω 2 
V ( z, t ) = ∑aν ∫ dω exp  −  ν
 
ν
−∞
  ∆ων  
(1.10.12)
 
 ω   1 ∂ε  
× exp  jωt − j z  ε (ων ) + (ω − ων )   
 c  2 ε (ω ) ∂ω ω =ων 
 
Rearranging the terms we have
56
  z 
V ( z , t ) = ∑aν exp  jων t
 c− ε (ων 
)
ν  
  z z ων  d ε   
×∫dω exp  j (ω − ων ) t − ε (ων ) −     (1.10.13)
  c c 2 ε (ων )  d ω  
ων 
  1 z 1  d ε   
× exp −(ω − ων )  2
+j    .
 ∆ω
 ν
2
c 2 ε (ων 
) d ω ων  
The first term in the integrand can be written as
  z 
exp  j (ω − ων )  t −
 V  
(1.10.14)
  gr  

whereas for the second term it is convenient to write
1 1  dε  1  1 1 
 =  −  (1.10.15)
c 2 ε (ων )  dω ω =ω ω V
ν  gr V ph 
ν
and define
1 1 1  ων   c c 
≡ + j 2  z  − . (1.10.16)
∆Ων2 ( z ) ∆ων2 ων  c   Vgr ,ν V ph ,ν 
57
Consequently,
  z ∞   z   ω − ων  
2
V ( z , t ) = ∑aν exp  jων t − ε (ων )   ∫ dω exp  j (ω − ων )  t −  −   


 c   Vgr   ∆Ων  
−∞
ν  
  z   ∞  2  z 
= ∑aν exp  jων t −   ∆Ων ∫−∞dξ exp  −ξ + jξ∆Ων  t −   .
ν   V ph (ων )     Vgr  
(1.10.17)
d α e−α = π such that finally we
∞ 2
The integral may be evaluated analytically since ∫−∞
get
  
2

   ∆Ων ( z ) 
 
M   z z  
V ( z , t ) = ∑aν ∆Ων ( z ) π exp  jων  t − exp −  t−   . (1.10.18)
 
 V  
ν =1      2
gr ,ν 

Vgr ,ν   

According to (1.10.16) the width of the pulse varies with the distance from the input.
Comment: Discuss possibilities of spectrum measurement by its decomposition in space.
58
1.10.2 Propagation and Reflection
Other phenomena which may significantly affect the propagation of a pulse along a
transmission line are discontinuities. For a glimpse into those phenomena let us consider
three transmission lines connected in series.
Z1 Z2 Z3
ρ τ
z =0 z =d
The reflection and transmission coefficients in the frequency-domain are
59
cosψ ( Z1 − Z 3 ) Z 2 + j sinψ ( Z1Z 3 − Z 22 )
ρ=
cosψ ( Z1 + Z 3 ) Z 2 + j sinψ ( Z1Z 3 + Z 22 )
2Z1Z 2
τ= (1.10.19)
cosψ ( Z1 + Z 3 ) Z 2 + j sinψ ( Z1Z 3 + Z 22 )
ψ = β 2d.
With these coefficients we aim to determine the transmitted signal in the time domain.
Assuming that the incoming voltage is
∞
V ( in ) ( z , t ) = ∫ dω V (ω ) exp  jωt − j β1 (ω ) z  (1.10.20)
−∞
wherein
1 ∞ − jωt
∫
V (ω ) =( in )
dtV ( z = 0, t ) e . (1.10.21)
2π −∞
Based on this expression, the transmitted signal is

∞
V ( tr )
( z , t ) = ∫ dω V (ω ) exp [ jωt − j β 3 (ω )( z − d ) ]τ (ω ) (1.10.22)
−∞
which assuming dispersionless transmission lines entails
60
∞  1 ∞ − jωt ′ 
V ( tr ) ( t , z ) = ∫ dω  ∫ dt ′V ( in ) ( z = 0, t ′) e 
−∞
 2π −∞
  z − d  (1.10.23)
4 Z1Z 2 exp ( − jψ ) exp  jω  t − 
  V 3 
× .
−2 jψ −2 jψ
(1 + e )( Z1 + Z 3 ) Z 2 + (1 − e )( Z1Z 3 + Z 2 )
2
We now define
4 Z1Z 2 −( Z1 + Z 3 ) Z 2 + Z1Z 3 + Z 22
ξ= ,χ=
Z 2 ( Z1 + Z 3 ) + Z1Z 3 + Z 22 ( Z1 + Z 3 ) Z 2 + Z1Z 3 + Z 22
enabling us to write
  z − d d  
′
exp  jω t − t − − 
∞ 1 ∞   V V2  
V ( z , t ) = ξ ∫ dt ′V ( z = 0, t ′)
2π ∫−∞
dω
( tr ) ( in ) 3
. (1.10.24)
−∞  ω 
1 − χ exp  −2 j d 
 V2 
Since χ < 1 we may expand
61
∞
1
= ∑uν (1.10.25)
1 − u ν =0
and further simplify
∞ 1 ∞   z − d d  
V ( tr )
( z, t ) = ξ ∫ dt ′V ( in ) ( z = 0, t ′) ∫−∞dω exp  jω t − t − V3 − V2  
′
−∞ 2π   (1.10.26)
∞
 ω 
×∑χ exp  −2 jν d .ν
ν =0  V2 
The integration over ω is straight forward resulting in a Dirac delta function therefore
∞
 z−d 
V ( z , t ) = ξ ∑χ νV ( in )  z = 0, t −
( tr )
(2ν + 1) . (1.10.27)
ν =0  V3 
A delay in the occurrence of the pulse reflects in the third term. The term ν = 0
represents the first pulse which reaches the output end. After a full round-trip, the second
contribution occurs. Effects of round-trip time and pulse duration as well as the sign of χ
are revealed by the following frames.
62
2.0 1.0
Z 1 = 200 Z 1 = 200
1.5 β 2 = 0.75 0.8 β 2 = 0.75
) χ = −0.143 χ = 0.286
)
Z 2 = 100 Z 2 = 40
V z / d , 9T p
V z / d , 9T p
1.0 0.6
β 3 = 0.9 β 3 = 0.9
Z 3 = 0.4 Z 3 = 100
0.5 d = 1.0 0.4 d = 1.0
(
(
T p = 2.0 T p = 2.0
0.0 0.2
-0.5 0
0 5 10 15 20 25 0 5 10 15 20 25
z/d z/d
2.0 1.4
Z 1 = 200 Z 1 = 200
1.2
1.5 β 3 = 0.85 β 2 = 0.85
χ = −0.143 1.0 χ = 0.286
)
)
Z 2 = 100 Z 2 = 40
V z / d , 4T p
V z / d , 4T p
1.0
β 3 = 0.9 0.8 β 3 = 0.9
Z 3 = 0.4 Z 3 = 100
0.6
0.5 d = 0 .5 d = 0 .5
(
(
T p = 2.0 0.4 T p = 2.0
0.0
0.2
-0.5 -0.5
0 2 4 6 8 10 12 0 2 4 6 8 10 12
z/d z/d
63
1.11 Appendix
1.11.1 Solution to Exercise 1.10
70 0.40 7.2
60
0.39
V ph / c
Z c [Ω]
6.8
ε eff
50
0.38
40 6.4
30 0.37
Microstrip: 1 2 3 4 1 2 3 4 1 2 3 4
dependence of ∆[mm] ∆ [mm] ∆[mm]
various
parameters on 70 0.42 7.2
the geometric 60
0.40
V ph / c
Z c [Ω]
parameters and 6.8
ε eff
50
the dielectric 0.38
coefficient. For 40 6.4
all the graphs the 30 0.36
1 2 3 4 1 2 3 4 1 2 3 4
paramters are
h [mm] h [mm] h [mm]
(when they are
not the variable): 140 1.00 7.0
∆ = 2 mm,
0.82
w = 20 mm,
Z c [Ω]
V ph / c
104 4.6
ε eff
h = 2 mm, 0.65
ε r = 10 and 68 0.47 2.2
ν = 0,1100 . 0.30
1 4 7 10 1 4 7 10 1 4 7 10
εr εr εr
64
Show that if w h,∆ the various quantities are independent of w .
60
n=100, 150, 200
52
44 0.415
Z c [Ω]
n=50
36
28 0.408
20
0 100 200 300
V ph / c
w[mm] 0.400
6.75
n=50 n=50, 100, 150, 200
Dependence of various parameters in 6.70 0.393
w . [ ∆ = 2 mm, ε r = 10 , h = 2 mm and 6.65
ν = 0,1100 ].
ε eff
n=100
6.60 n=150, 200 0.385
5 36.67 68.33 100
6.55 w[mm]
6.50
5 70 135 200
65 w[mm]
Determine the inductivity per unit length and determine its dependence on the
various parameters.
The inductivity per unit length of the stripline can be determined according to Eq.
(1.7.12) L = 1/ c 2C (ε r = 1) substituting the expression for the capacitance per unit length
h2 ∞ 2 h  sinh(ξ n ) 
given by Eq. (1.8.17) L = 2 µ0 ∑ sinc  π n d  1 − exp ( −ξ n ) ξ 
∆d n =1  n 
1 ∆
where ξn = π n .
2 d 0.36
0.34
L[µH / m]
0.32
The figure shows the inductance per unit length of a
stripline as a function of h . Clearly it is symmetrical around
the maxima point h = 2 mm, corresponding to the 0.30
symmetrical stripline configuration. [ d = 4 mm, ∆ = 2 mm,
n = 1,2, 200 ].
0.28
1.0 1.5 2.0 2.5 3.0
66
h[mm]
0.8
0.6
L[µH / m]
The figure shows the dependence of L on ∆ , the strip 0.4
width. The inductance per unit length of the stripline decreases
as the strip width increases, and saturates as ∆ becomes very
large compared to h . From this result one can conclude that 0.2
the inductance per unit length of a metallic wire is larger than
that of a strip [ h = 2 mm, d = 4 mm n = 1,2, 200 ]. 0
1 2 3 4
∆[mm]
60
40
Z c [Ω]
The characteristic impedance of a stripline as a function
of d (Eq. (1.8.18)). The impedance decreases as d decreases,
due to increase in the capacitance per unit length of the 20
stripline as shown in Figure 20. The symmetrical case where
d = 2h , is obtained at the knee of the graph corresponding to
d = 4 mm. An asymptotic behavior is observed as d becomes
large compared to h,∆ . [ h = ∆ = 2 mm, ε r = 10 ,
0
0 5 10 15 20
n = 1,2, 200 ].
d [mm]
67
Compare micro-strip and strip-line from the perspective of sensitivity to the

dielectric coefficient.
The characteristic impedance of a

microstrip (the red solid line) and a stripline
(the blue dotted line) as a function of ε r . The 150
characteristic impedances of both the microstrip
and the stripline decrease as ε r increases. This
result can be explained by noting that increasing
100
Z c [Ω]
the dielectric constant increases the capacitance Microstrip
per unit length of the line (C ) and thus
decreases the characteristic impedance, which is
inverse proportional to the capacitance
according to Z c = 1 / CV ph , where V ph is the 50
phase velocity, in turn ε r decreases V ph but in Stripline
lower rate than the increase of C . [ h = ∆ = 2
mm, d = 4 mm (stripline), ω = 20 mm, 0
n = 1,2, 200 , and ν = 0,1,… 200 ]. 0 5 10 15
εr
68
The derivative of the characteristic impedance of a
microstrip (solid line) and a stripline (dotted line) as a 0
function of ε r . At low values of ε r , the absolute value
of the dZ c / d ε r is larger in stripline (dotted line) than Stripline
in microstrip (solid line), implying a higher sensitivity
-20 Microstrip
dε r
dZ c
to variations in ε r in the stripline geometry at that low
range of ε r . As ε r is increased, the sensitivity of the
-40
microstrip characteristic impedance ( Z c ) becomes
slightly larger than that of the stripline. At high values
of ε r , the sensitivity of both configurations approach -60
zero asymptotically as expected. [ h = ∆ = 2 mm, 0 5 10 15
d = 4 mm, ω = 20 mm, n = 1,2, 200 , and
ν = 0,1,… 200 ]. εr
35
Sensitivity to dielectric coefficient of a [pF / m ]
microstrip (solid line) and a stripline (dotted 30 Stripline
line) as a function of ε r . The figure shows
clearly that the stripline geometry is more
sensitive with respect to ε r than the microstrip
25
configuration [ h = ∆ = 2 mm, d = 4 mm
dε r
dC
(stripline), ω = 20 mm, n = 1,2, 200 , and 20 Microstrip

15
0 5 10 15
69
εr
ν = 0,1,… 200 ].
The normalized phase velocity of a microstrip (solid line)
and a stripline (dotted line) as a function of ε r ,V ph of stripline is
simply c / ε r (for a TEM mode), V ph of microstrip (for the 0.8
Microstrip
V ph / c
quasi-TEM mode) is c / ε eff , where ε eff is given by Eq. (7.17).
As ε r is increased, V ph / c decreases as expected, according to 0.6
Stripline
1 / ε r and 1 / ε eff relations, for stripline and microstrip cases
0.4
respectively, where the following inequality holds 1 < ε eff < ε r . In
the special case ε r = 1 where only vacuum is experienced by the 0.2
electromagnetic field, V ph is simply c for both geometries 0 5 10 15
[ h = ∆ = 2 mm, d = 4 mm (stripline), ω = 20 mm,
n = 1,2, 200 , and ν = 0,1,… 200 ]. εr
The derivative of the normalized phase velocity of 0

a microstrip (solid line) and a stripline (dotted line) as a
function of ε r . The absolute value of the derivative is
-0.2 Stripline
1 dV ph
higher for the stripline case compared to the microstrip
c dε r
case, implying higher sensitivity in the stripline with Microstrip
respect to changes in ε r . For relatively high values of ε r
the sensitivity of both configurations approaches zero -0.4
asymptotically [ h = ∆ = 2 mm, d = 4 mm (stripline),
-0.6
0 5 10 15
70 εr
Chapter 2: Waveguides – Fundamentals
2.1 General Formulation
So far we have examined the propagation of electromagnetic waves in a structure

consisting of two or more metallic surfaces. This type of structure supports a transverse
electromagnetic (TEM) mode. However, if the electromagnetic characteristics of the
structure are not uniform across the structure, the mode is not a pure TEM mode but it
has a longitudinal field component.
In this chapter we consider the propagation of an electromagnetic wave in a closed
metallic structure which is infinite in one direction ( z ) and it has a rectangular (or
cylindrical) cross-section as illustrated in Fig. 1. While the use of this type of waveguide
is relatively sparse these days, we shall adopt it since it provides a very convenient
mathematical foundation in the form of a set of trigonometric functions. This is an
orthogonal set of functions which may be easily manipulated. The approach is valid
whenever the transverse dimensions of the structure are comparable with the wavelength.
71
72
µ0 µr
b ε0εr y
z
a x
Rectangular waveguide; a and b are the dimensions of the rectangular cross section. a corresponding to
the x coordinate, b to the y coordinate.
The first step in our analysis is to establish the basic assumptions of our approach:
a) The electromagnetic characteristics of the medium: µ = µ0 µ r and ε = ε 0ε r .

b) Steady state operation of the type exp ( jωt ) .
c) No sources in the pipe.
d) Propagation in the z direction -- exp ( − jkz z ) ; kz can be either real or imaginary or
complex number.
e) The conductivity (σ ) of the metal is assumed to be arbitrary large (σ → ∞ ).
73
µ0 µr
b ε0εr y
z
Subject to these assumptions Maxwell's Equations may be a x
written in the following
form

∇ × E = − jωµ H ∇ × H = jωε E (2.1.1)
Substituting
one equationinto the other we
obtain
the wave equation

∇ × (∇ × E ) = − jωµ (∇ × H ) ∇ × (∇ × H ) = jωε (∇ × E )

∇(∇ ⋅ E ) − ∇ E = − jωµ ( jωε E )
2
∇(∇ ⋅ H ) − ∇ H = jωε ( − jωµ H )
2
 ω 2  ω 2
∇ (∇ ⋅ E ) − ∇ E =   E
2
∇ (∇ ⋅ H ) − ∇ H =   H
2
v v (2.1.2)

∇ ⋅ε E = 0 ∇ ⋅ µH = 0
 2  ω 2   2  ω 2 
∇ +    E = 0 ∇ +    H = 0
  v     v  
where v = 1/ µε is the phase-velocity of a plane wave in the medium. Specifically, we
conclude that the z components of the electromagnetic field satisfies
 2 ω2   2 ω2 
∇ + v 2  Ez = 0, ∇ + v 2  H z = 0 (2.1.3)
   
74
µ0 µr
b ε0εr y
and subject to assumption (d) we have z
 2 x
ω2   2 ω2  a
 ∇ ⊥ − k z + 2  Ez = 0  ∇ ⊥ − k z + 2  H z = 0.
2 2
 v   v 
As a second step, it will be shown that assuming the longitudinal components of the
electromagentic field are known, the transverse components are readily established. For
this purpose we observe that Faraday's Law reads
1x 1y 1z

∇ × E = − jωµ H ⇒ ∂ x ∂ y − jkz = − jωµ H . (2.1.4)
Ex Ey Ez
(i) 1x : ∂ y Ez + jkz E y =− jωµ H x jkz E y + jωµ H x = − ∂ y Ez
(ii) 1y : − (∂ x Ez + jkz Ex ) = − jωµ H y − jkz Ex + jωµ H y = ∂ x Ez
(iii) 1z : ∂ x E y − ∂ y Ex = − jωµ H z ∂ x E y − ∂ y Ex =− jωµ H z .
In a similar way, Ampere's law reads
1x 1y 1z

∇ × H = jωε E ⇒ ∂ x ∂ y − jkz = jωε E, (2.1.5)
Hx H y Hz
75
µ0 µr
b ε0εr y
z
a x
or explicitly
(iv) 1x : ∂ y H z + jkz H y = jωε Ex jωε Ex − jk z H y = ∂yHz
(v) 1y : − (∂ x H z + jkz H x ) = jωε E y jωε E y + jkz H x = − ∂xHz
(vi) 1z : ∂xH y − ∂ yHx = jωε Ez ∂xH y − ∂ yHx = jωε Ez .
From equations (ii) and (iv) we obtain
jkzωε 1 1 
Hy =  ∂ E + ∂ H z
− jkz Ex + jωµ H y = ∂ x Ez  kz 2 − (ω /v) 2 k
 z
x z
ωε y

→ (2.1.6)
jωε Ex − jkz H y = ∂ y H z  E = j
 k ∂ E + ωµ∂ y H z 
2  z x z
kz − (ω /v)
x 2
It is convenient at this point to define the transverse wavenumber

ω2
k ≡
2
⊥ 2
− kz 2 (2.1.7)
v
That as we shall shortly see, has a special physical meaning.
76
µ0 µr
b ε0εr y
z
This allows us to write the last two expressions in the following x
a
form
− jk jωµ
Ex = 2 z ∂ x Ez − 2 ∂ y H z , (2.1.8)
k⊥ k⊥
− jωε jkz
Hy = 2
∂ E
x z − 2
∂ y Hz. (2.1.9)
k⊥ k⊥
In a similar way, we use equations (i) and (v) and obtain
− jk jωµ
E y = 2 z ∂ y Ez + 2 ∂ x H z (2.1.10)
k⊥ k⊥
jωε jk
H x = 2 ∂ y Ez − 2z ∂ x H z . (2.1.11)
k⊥ k⊥
Equations (2.1.8),(2.1.10) and (2.1.9),(2.1.11) can be written in a vector form
jkz jωµ
E⊥ = − 2 ∇ ⊥ Ez + 2 1z ×∇ ⊥ H z (2.1.12)
k⊥ k⊥
jkz jωε
H ⊥ = − 2 ∇ ⊥ H z − 2 1z ×∇ ⊥ Ez . (2.1.13)
k⊥ k⊥
77
Comments:
1. The wave equations for E z and H z with the corresponding boundary conditions and
the relations in ((2.1.12)--(2.1.13)) determine the electromagnetic field in the entire
space (at any time).
2. Note that the only assumption made so far was that in the z direction the
propagation is according to exp ( − jkz z ) . No boundary conditions have been
imposed so far.
3. Therefore it is important to note within the framework of the present notation that
TEM mode ( H z = 0, Ez = 0) is possible provided that k⊥ ≡ 0 or substituting in the
wave equations

∇ ⊥ E⊥ = 0;
2
∇ ⊥ H ⊥ = 0.
2
(2.1.14)
4. By the superposition principle and the structure of ((2.1.12)--(2.1.13)), the ransverse
field components may be derived from the longitudinal ones.
Complete Solution = ( Ez = 0) and( H z =/ 0) + ( Ez =/ 0) and ( H z = 0) .

Transverse Electric(TE) Transverse Magnetic(TM)
78
µ0 µr
b ε0εr y
z
2.2 Transverse Magnetic (TM) Mode [ H z = 0 ] x
a
In this section our attention will be focused on a specific case where H z = 0 . This
step is justified by the fact that equations ((2.1.12)--(2.1.13)) are linear, therefore by
virtue of the superposition principle (e.g. circuit theory) and regarding H z and Ez as
sources of the transverse field, we may turn off one and solve for the other and vice
versa. As indicated in the last comment of the previous section, the overall solution is,
obviously the superposition of the two. The boundary conditions impose that the
longitudinal electric field Ez vanishes on the metallic wall therefore
 π mx   π ny  − jkzn ,m z
Ez = ∑ Anm sin   sin  e . (2.2.1)
n ,m  a   b 
This further implies that the transverse wave vector, k⊥ , is entirely determined by the
geometry of the waveguide (substitute in (2.1.3))
2 2
 π m   π n  ω2
 +  = 2 − kz .
2 2
k⊥ =  (2.2.2)
 a   b  v
From these two equations we obtain
79
µ0 µr
b ε0εr y
2 2 z
ω 2
ω 2
πm  πn  x
kz, 2n ,m = − k⊥2 = −   − 
a
v2 2
v  a   b 
2 2
ω2  π m 
πn 
kz = ± 2
−  −
   . (2.2.3)
v  a   b 
This expression represents the dispersion equation of the electromagnetic wave in the
waveguide.
Exercise 2.1: Analyze the effect of the material characteristics on the cut-off frequency.
Exercise 2.2: What is the impact of the geometry?
Exercise 2.3: Can two different modes have the same cut-off frequency?!
What is the general condition for such a degeneracy to occur?
80
µ0 µr
b ε0εr y
z
Comments: x
a) Asymptotically (ω k⊥ v ) this dispersion relation behaves as if a
no walls were present i.e. ω kz v .
b) There is an angular frequency ωc,n ,m for which the wavenumber kz vanishes. This is
called the cutoff frequency.
2 2 2 2
πm πn
(ωc )m,n ≡ vk⊥ = v   +   ⇒ ( f c )m,n = v   +   ,
1 m n
(2.2.4)
 a   b  2  a  b
where v = c/ ε r µr .
c) Below this frequency the wavenumber kz is imaginary and the wave decays or grows
exponentially in space.
ω
d) The indices n and m define the mode TM m,n ; m
represents the wide transverse dimension ( x ) whereas n
represents the narrow transverse dimension ( y ).
ω = k zv
ω c12
ω c11
81
kz
µ0 µr
b ε0εr y
z
a x
2.3 Transverse Electric (TE) Mode [ Ez = 0 ]
The second possible solution according to (2.1.12)--(2.1.13) is when Ez = 0 and

since the derivative of the longitudinal magnetic field H z vanishes on the walls (see
(2.1.12)) we conclude that
 π mx   π ny  − jkzm,n z
H z = ∑ Am,n cos   cos  e . (2.3.1)
m ,n  a   b 
The expression for the transverse wavenumber k⊥ is identical to the TM case and so is
the dispersion relation. However, note that contrary to the TM mode where if n or m
were zero the field component vanishes, in this case we may allow n = 0 or m = 0
without forcing a trivial solution.
82
µ0 µr
b ε0εr y
z
For convenience, we present next a comparison table of the x
a
various field components of the two modes.
TM mode TE mode
 π mx   π ny  − jkzmn z  π mx   π ny  − jkmn z
Ez = ∑ mn Amn sin   sin  e H z = ∑ mn Bmn cos   cos  e
 a   b   a   b 
jωε  π n   π mx   π ny  jωµ  π n   π mx   π ny 
H x = ∑ mn Amn 2   sin   cos  … E x = ∑ mn Bmn 2   cos   sin  …
k ⊥ n ,m  b   a   b  k⊥  b   a   b 
− jωε  π m   π mx   π ny  jωµ  π m   π mx   π my 
H y = ∑ mn An   cos   sin  … E y = ∑ mn Bmn −  sin   cos  …
k⊥2  a   a   b  k⊥2  a   a   b 
Hz = 0 Ez = 0
− jkz  π m   π mx   π ny  + jkz  π m   π mx   π my 
E x = ∑ mn Amn   cos   sin  … H x = ∑ mn Bmn   sin   cos  …
k⊥2  a   a   b  k⊥2  a   a   b 
− jk  π n   π mx   π ny  + jk  π n   π mx   π my 
E y = ∑ mn Amn 2 z   sin   cos  … H y = ∑ mn Bmn 2 z   cos   sin  …
k⊥  b   a   b  k⊥  b   a   b 
1/2 −1/ 2
µ   f c ,m ,n  2  E y ,mn µ   f c ,m ,n  2 
(TM )
Ex ,mn = Z mn H y ,mn (TM )
; Z mn = 1 −    H x ,mn = − (TE )
; Z mn = 1 −   
ε   f  
(TE )
Z mn ε   f  
E y ,mn = − Z mn
( TM )
H x ,mn E x ,mn
H y ,mn = ( TE )
Z mn
Exercise 2.4: Check all the expressions presented in the table above.
83
µ0 µr
b ε0εr y
z
a x
Comments:
1. The phase velocity of the wave is the velocity an imaginary observer has to move
in order to measure a constant phase i.e. ωt − kz z = const.. This implies,
−1/2
ω  f 
2
v ph ≡ = v 1 −  .
c
2
(2.3.2)
kz  f 
2. The phase velocity is always faster than v !! Specifically, in vacuum the phase
velocity of a wave is larger than c . In fact close to cutoff this velocity becomes
`ìnfinite"!!
3. The group velocity is defined from the requirement that an observer sees a constant
envelope in the case of a relatively narrow wave packet. At the continuous limit this
is determined by
1/2
∂ω  f c2 
vgr = = v 1 − 2  . (2.3.3)
∂kz  f 
4. The group velocity is alway smaller than v . Specifically, in vacuum it is always
smaller than c . It is the group velocity is responsible to information transfer.
5. When the waveguide is uniform
v ph v gr = v 2 . (2.3.4)
84
µ0 µr
b ε0εr y
z
2.4 Power Considerations a x
2.4.1 Power Flow

Let us now consider the power associated with a specific TM mode; say
 π mx   π ny  − jkz ,mn z
Ez = Amn sin   sin  e .
 a   b 
At this stage, for simplicity sake, we assume that this is the only mode in the
waveguide. Based on Poynting's theorem, the power carried by this mode is given by
Pmn = Re{∫ 0
a b
}
dx ∫ dy Sz,mn .
0
(2.4.1)
Explicitly the longitudinal component of the Poynting vector is
1 * 1
S z ,mn = ( E × H ) ⋅ 1z = [ E x H *y − E y H x* ]
2 2
1  (TM )  1 (TM ) 
H y + H x .
2 2 2 2
= Z Hy + Z ( TM )
Hx = Z (2.4.2)
2  2  
Above cutoff the characteristic impedance is a real number therefore the next step is to
substitute the explicit expressions for the magnetic field components and perform the
spatial integration:
85
µ0 µr
b ε0εr y
z
1 (TM ) a b
 H 2 + H 2 x
Pmn = Z
2 ∫0
dx ∫0
dy

 y x 

a
 
2 ω ε a b πm 
2 2  2
πn  
2
1 (TM ) 
= Z Amn 4
⋅   +  
2 k⊥ 2 2 
a  b  
 2 
 k⊥ 
2 ω ε ab
2 2
1 (TM )
Pmn = Z Amn . (2.4.3)
2 k⊥2 4
The last expression represents the average power carried by the specific mode.
Exercise 2.5: What is the power at any particular point in time?

Exercise 2.6: Since the two sets sin(π mx / a ) and sin(π ny / b ) are two orthogonal sets of
functions, the total average power carried by the wave in the forward direction is a
superposition of the average power carried by each individual mode separately. In other
words show that P = ∑ n ,mPmn .
Exercise 2.7: Show that below cutoff, the power is identically zero although the field is
not zero.
86
µ0 µr
b ε0εr y
z
a x
Exercise 2.8: Note that the average power is proportional to the
average magnetic energy per unit length. Calculate this quantity. Compare it with the
average electric energy per unit length.
Exercise 2.9: Calculate the energy velocity of a specific mode v EM = 〈 P〉 /WEM . Compare
to the group velocity. What happens if the frequency is below cutoff?
Exercise 2.10: Repeat the last exercise for a superposition of modes An ,m .
87
µ0 µr
b ε0εr y
z
a x
2.4.2 Ohm Loss

So far it was assumed that the walls are made of an ideal metal (σ → ∞ ) . If this is
not the case (σ ) a finite amount of power is absorbed by the wall. In order to calculate
this absorbed power we firstly realize that the magnetic field is ``discontinuous" which is
compensated by a surface current

J s = n × H. (2.4.4)
This current flows in a very thin layer which is assumed to be on the scale of the skin-
depth [ J = J s / δ ] therefore, the dissipated average power per unit length is given by
11 2
2 σ ∫δ
PD = dxdy J s , (2.4.5)
or explicitly
2
1 Js 1 2 Rs 2
PD =
2σ ∫ dlδ δ = 2δσ ∫ dl J s = 2 ∫ dl J s , (2.4.6)
where δ ≡ 2 / ωµoσ , Rs ≡ ωµo / 2σ and the integration is over the circumference of

the waveguide
88
µ0 µr
b ε0εr y
z
1 2 x
PD = Rs ∫ dl J s . (2.4.7) a
2
This is the average electromagnetic power per unit length which is converted into heat
(dissipation) due to Ohm loss. Based on Poynting's theorem we may deduce that the
spatial change in the electromagnetic power is given by
d
P = − PD (2.4.8)
dz
and since in case of a single mode both P and PD are proportional to | A |2 ,
2 2
P∝ A and PD ∝ A (2.4.9)
we conclude that the change in the amplitude of the mode is given by
2
= −2α A ⇒ A( z ) = A( z = 0) e−2α z .
d A 2 2 2
(2.4.10)
dz
The coefficient α represents the exponential decay of the amplitude and based on the
arguments of above is given by
P
α ≡ D. (2.4.11)
2P
Let us denote by kz (0) the wavenumber in a lossless waveguide. Subject to the
89
µ0 µr
b ε0εr y
z
assumption of small losses ( kz (0) α ) we can generalize the a x
solution in a waveguide with lossy walls by kz = kz (0) − jα .
Exercise 2.11: Based on the previous calculation of the power show that this parameter is
given by
(TM ) 2 Rs 1 m 2b 3 + n 2 a 3
α m,n = (2.4.12)
η 1 − ( f c /f ) 2 b m b a + n a
2 2 2 3
Rs is the surface resistance. Note that α is very large close to cutoff. Explain the
difficulty/contradiction.
2.4.3 Dielectric Loss

If the dielectric coefficient of the material is not ideal, in other words, it has an
imaginary component ε r = ε ' − jε '' , then the wavenumber is given by
1ω ω
kz kz (0) − jε '' , (2.4.13)
2 c ckz (0)
where we assumed that (i) the dielectric loss is small and (ii) the system operates remote
from cutoff conditions [i.e. kz (0) ε '' ω / c ].
90
µ0 µr
b ε0εr y
z
We can now repeat the entire procedure described in the x
a
previous subsection for a TE mode. Here are the main steps and
results
 π mx 
H z = Amn cos 
 a 
 π ny  − jkz ,mn z
 cos 
 b
e
 0
a
0
b
{
, Pmn = Re ∫ dx ∫ dyS z ,mn , }
S z ,mn = ( Ex H y − E y H x ) =
1 1  2
Ex + E y  .
* * 2
(2.4.14)
2 mn 2 Z mn 
( TE )  mn
2 ω µ  π n  1 1 ω 2µ 2  π m  1 1 
2 2 2 2
1 1
Pmn = (TE )
Amn  4   a b+ 4   a b
2 Z mn  ⊥ 
k b  2 2 k ⊥  a  2 2 
(2.4.15)
2ω µ
2 2
1 ab
= (TE )
Amn
8 Z mn k⊥2
2 Rs 1 m 2 (b/a )3 + n 2
α (TE )
= . (2.4.16)
b m 2 (b/a ) 2 + n 2
mn
2 1/2
η 1 − ( f c /f ) 
91
Exercise 2.12: Check equation (2.4.16). In particular check the cases n = 0 or m = 0 .
Repeat all the exercises from the above (TM mode) for the TE mode. Make a comparison
table where relevant.
Exercise 2.13: Note that both α and α (TM ) are large close to cutoff and increase as
( TE )
ω for large frequencies. In between there is a minimum loss for an optimal

frequency. Calculate it.
Exercise 2.14: Calculate the loss very close to cutoff.
92
2.5 Mode Comparison
Mode comparison for a rectangular waveguide.
93
TE 10 TE 11 TE 21
3 3 3
2 1 1
2
1 2
y z y z y z
x x x
1 1 1
3 2 3 2 3 2
TE 20 TM 11 TM 21
3 2 3 3
2 2
1 1 1
y y
z z
x x
1 1 1
3 2 3 2 3 2
94
2.6 Cylindrical Waveguide
2.6.1 Transverse Magnetic (TM) Mode [ H z = 0]
In this section we shall investigate the propagation of a wave in a cylindrical

waveguide. The longitudinal component of the electric field satisfies
[∇ 2⊥ + k⊥2 ] Ez = 0, (2.6.1)
ω2
where k =2
⊥ − k z2 or explicitly
v2
 ∂2 1 ∂ 1 ∂2 2
+ +
 ∂r 2 r ∂r r 2 ∂ϕ 2 + k ⊥  E z = 0. (2.6.2)
 
The solution of this equation subject to the boundary conditions Ez ( r = R ) = 0 reads
 r  − jkz s ,n z
E z = ∑ J n  ps , n  e [ Ans cos( nϕ ) + Bns sin( nϕ )], (2.6.3)
n ,s  R 
where J n (u ) if the n 'th order Bessel function of the first kind. This function behaves
similar to a trigonometric function ( sin or cos ). It has zeros, denoted by ps ,n i.e.,
ps ,n : J n ( ps ,n ) ≡ 0.
95
The first few zeros of the Bessel function are tabulated next.
s=1 s=2 s=3

n= 0 2.405 5.52 8.654
n=1 3.832 7.016 10.174
n=2 5.135 8.417 11.62
Substituting (2.6.3) in (2.6.2) we obtain

 ∂ 2 1 ∂ 1 ∂ 2  ps , n  2 
 2+ + 2 +   Ez = 0 (2.6.4)
 ∂ r r ∂ r r ∂ ϕ 2
 R  
thus
2
ω  s ,n  ω 2 ps , n
2 2
p
k⊥ ≡ 2 − k z = 
2 2
 ⇒ k z s ,n = 2 − 2 .
2
(2.6.5)
v  R  v R
Based on this expression the characteristic impedance of the TM mode is given by
kz,s ,n v
Z s ,n = η
(TM )
. (2.6.6)
ω
96
2.6.2 Transverse Electric (TE) Mode [ E z = 0]
In this case the wave equation reads

[∇ 2⊥ + k⊥2 ] H z = 0 (2.6.7)
and its solution has the form
 ′ r  − jkz s ,n z
H z = ∑ J n  ps , n  e  As ,n cos( nϕ ) + Bs ,n sin( nϕ ) , (2.6.8)
s ,n  R 
where
∂H z
= 0 ⇒ p′s ,n : J n′ ( p′s ,n ) = 0. (2.6.9)
∂r r = R
The first few zeros of the derivatives of the Bessel function are
s=1 s=2 s=3
N=0 3.832 7.016 10.174
N=1 1.841 5.331 8.536
N=2 3.054 6.706 9.970
thus
2
 ω   p′s ,n 
2
ω
n =η
Z s(,TE =   −
)
, kz,s ,n  (2.6.10)
vk z , s ,n v  R 
97
Summary
TE modes TM modes
 r  − jksn z cos nϕ
Hz = J n  p′sn  e  0
 R sin nϕ
 r  − jksn z cos nϕ
Ez = 0 J n  psn  e 
 R sin nϕ
jk sn p′sn  r  − jksn z cos nϕ
Hr = − J ′  p′ e  − Eϕ /Z sn(TM )
sin nϕ
n sn
Rk⊥2 ,sn  R
jnksn  r  − jksn z − sin nϕ
Hϕ = − J p ′
n  sn e  Er /Z sn(TM )
rk⊥2 ,sn  R  cos nϕ
jksn psn  r  − jksn z cos nϕ
Er = Z sn(TE ) H ϕ − J p
n′  sn e 
Rk⊥2,sn  R sin nϕ
jnksn  r  − jksn z − sin nϕ
Eϕ = − Z sn(TE ) H r − J n  psn  e 
 cos nϕ
2
rk⊥ ,sn  R
1/2 1/2
 ω 2  p′sn 2   ω 2  psn 2 
ksn =  2 −    2 −  
 v  R    v  R  
ω vk
Z sn(TE ) = η Z sn(TM ) = sn η
vksn ω
k⊥ ,sn = p′sn /R psn /R
98
TE modes TM modes
η k0 ksnπ η k0 ksnπ
Power= 4
( psn'2 − n 2 ) J n2 ( p′sn ) 4
psn2 [ J n′ (k⊥ ,sn R)]2
2k ⊥ , sn g 0,n 2k ⊥ , sn g0,n
−1/2 −1/2
Rs  k⊥ ,sn v   k⊥2,snc 2 n2  Rs  k⊥ ,sn v 
2 2 2 2
α= 1 −   +  1 − 
Rη  ω2   ω
2
psn'2 − n 2  Rη  ω2 
1 n = 0
1 µ0 µr 
v= , η= g 0,n =
µ0 µr ε 0ε r ε 0ε r 2 n =/ 0

99
2.6.3 Power Considerations
According to Maxwell's Equations for a single mode we have
jk E
E⊥ = − 2z ∇ ⊥ Ez , Hϕ = r , H r = − Eϕ / ZTM .
k⊥ ZTM
Consequently, the average Poynting vector is
1 *
( )
S z = E⊥ × H ⊥ ⋅ 1z
2
and the average power flowing in the waveguide
{ }
1 *  1
P = Re ∫ da  E⊥ × H ⊥   = Re ∫ da  Er H ϕ* − Eϕ H r* 
 2 c. s . z
 2 c. s.
  2

1  E
  = 1 Re { }
2
| E | ϕ  E 2 + E 2
= Re ∫ da  r + ∫c.s.  r
da ϕ 
2  c.s.  ZTM ZTM   2 ZTM 
  
  
2

=
1
2 ZTM


c.s.
kz
Re ∫ da 2  ∇ ⊥ Ez  =
k
 ⊥
2
 2
1 k z2
Z TM k 4
⊥
Re ∫ {
c. s .
da ∇ E
⊥ z }
2
. (2.6.11)
In order to further simplify the last expression let us examine the wave equation:
( ∇2⊥ + k⊥2 ) Ez = 0 (2.6.12)
100
we multiply by the complex conjugate of Ez
Ez* ( ∇ 2⊥ + k⊥2 ) Ez = 0 (2.6.13)
and integrate over the entire cross section
∆ z ∫ da  Ez*∇ 2⊥ Ez + k⊥2 Ez  = 0
2
 
∆ z ∫ da ∇ ⊥ ( Ez*∇ ⊥ Ez ) − ∇ ⊥ Ez + k⊥2 Ez  = 0.
2 2
(2.6.14)
 
The first term in the integrand is zero since

∫ ∇  ∇  ∫
 ⋅ 
 ∇ ⊥ Ez  = 0
* *
da ⊥  E z ⊥ E z = d
⊥ E z
hence
∫ da ∇⊥ Ez = k⊥ ∫ da Ez .
2 2 2
(2.6.15)
Now back to the propagating power for a superposition of modes starting from the
expression for a single mode we get
101
Ps ,n =
1 k2
(TM ) 4
2 Z s ,n k⊥
{
Re ∫ da ∇ ⊥ Ez
2
} =
2Z
1 k z2
(TM ) 2 ∫
k⊥
da Ez
2
 kz  r 
 ∑ k s , n n  s ,n R 
A J p cos( n ϕ ) 
1 π  
dφ ∫ drr  
2 R s ,n ⊥
P=
2η 0∫ 0  ω / v  r 
 ∑ As ',n ' J n '  ps ',n '  cos(n 'ϕ )  
  s ',n ' ⊥
k  R  
The integration over the angle it is sraight forward
1 n=0
2π 
∫0 d φ cos( n ϕ ) cos( n ' ϕ ) = 2 π g δ
n n,n ' , g n = 1
n =/ 0
 2
and after integration over r we get
2
1 k z ,s ,n 2 R 2
P =2π ∑ g n
2
(TM ) 2
A s ,n 
 J ′
n ( p s , n  δ ss′
) 
n , s , s′ 2Z s ,n ⊥,s ,nk 2
where we used
R  r  r  R2 2
∫0 dr r J

p
n  n,s J
 n  n,s ' 
R 
p
R 2
= 
 J n
′ ( p n,s  δ s,s '
)  (2.6.16)
102
In the case of a single mode we have
 1 2  ω   ω 2
p 2 
 J ′ ( p ) 
2
Ps ,n =  RAs ,n   R  Re  R 2 − s 2,n  π g n  n s ,n  . (2.6.17)

 2η   v   v R 
  ps , n 
Note that there is power flow only if the wave is above cutoff and as in the rectangular
case, the the total power is the superposition of the power in each mode separately
Exercise 2.15: Calculate the average energy per unit length stored in the
electromagnetic field.
103
2.6.4 Ohm Loss
Now to the general expression for the losses. Starting from the dissipated power
2
Rsu 2 1 2 1 Er , sn
∫ ∫ ∫
( su )
PD , sn = J z1 d
= Rs H φ d
= Rs (TM )
d
2 bound 2 2 Z sn
2 2  J n ( pns ) 
2
1 Rs ′ 2
= k R   2π g n Asn R
2 RZ (TM ) 2  p ns 
sn
1 Rs  ω   J n′ ( pns ) 
2 2
2
=  R    2π g n Asn R (2.6.18)
2 η R 2  v   pns 
which finally entails for a single mode
1 Rs  v ph ,sn 
α ns(TM ) =  . (2.6.19)
R η  v 
Exercise 2.16: Check Eq. (2.6.19).
(TE )
Exercise 2.17: Calculate α ns .
Exercise 2.18: Calculate the exponential decay due to dielectric loss.
104
TE11 TM 01 TE 01
0.010
Attenuation, decibels/meter
0.008
0.006
0.004
0.002
0
0 0.4 0.8 1.2 1.6 20 24 28 32 36 40
Diameter
TE 11 TE 21 TE 01 TE 31 TE 41 TE 12 Wavelength
fel ( fc )TE 11 Attenuation due to copper losses in circular waveguides at 3000 Mc/sec.
0 1 2 3
TM 01 TM 11 TM 21 TM 02
0.07
Relative cutoff frequencies of waves in a circular guide.
0.06 TM 01
Attenuation, decibels/meter
0.05
0.04
0.03
0.02 TE 01
TE11
0.01
1000 5000 9000 13000

Frequency, megacycles/second
Attenuation due to copper losses in circular
waveguides; diameter = 2 in.
Mode comparison for cylindrical waveguide.
105
Wave Type TM 01 TM 02 TM 11 TE 01 TE 11
Field distribution
in cross-sectional
plane, at plane of
maximum
Distributions Distributions
transverse fields Below Along
This Plane
Below Along
This Plane
Field distribution
along guide
Field components E z , Er , H ϕ E z , Er , H ϕ Ez , Er , Eϕ , H r , Hϕ H z , H r , Eϕ H z , H r , Hϕ , Er , Eϕ
present
p or p ′ 2.405 5.52 3.83 3.83 1.84
( kc ) 2.405 5.52 3.83 3.83 1.84
α α α α α
( λc ) 2.61α 1.14α 1.64α 1.64α 3.41α
( fc ) 0.383 0.877 0.609 0.609 0.293
α µε α µε α µε α µε α µε
Attenuation due to Rs 1 Rs 1 Rs 1 Rs ( f c / f )2 Rs 1
imperfect α n 1 − ( f c / f )2 α n 1 − ( f c / f )2 α n 1 − ( f c / f )2 α n 1 − ( f c / f )2 α n 1 − ( f c / f )2
conductors
106
2.7 Pulse Propagation
Let us consider an azimuthally symmetric TM mode described by
∞
r ∞
Ez (r , z , t ) = ∑J 0 ( ps ) ∫ dω exp [ jωt − Γ s z ] Es (ω ) (2.7.1)
R −∞
s =1
wherein Γ s = ps / R 2 − ω 2 / c 2 and Es (ω ) is the Fourier transform of this field component

2 2
at z = 0
1 R  r 1 ∞ ′
Es (ω ) = 2 ∫ dr ′r ′J 0  ps  ∫ dt exp ( − jωt ′ ) Ez (r ′, z = 0, t ′). (2.7.2)
R 2
J 1 ( ps )
0
 R  2π −∞
2
Let us now calculate the energy associated with the radiation field as it propagates in
an empty waveguide. The transverse field components are
 rp ∞ EΓ
Er (r , z , t ) = ∑J1  ps  s ∫ dω exp ( jωt − Γ s z ) s s 2
 R  R −∞ ω
s =1
Γ 2s + 2
c
∞
 r  ps ∞ 1 jω Es
H ϕ (r , z , t ) = ∑J1  ps  ∫ dω exp ( jωt − Γ s z ) (2.7.3)
 R  R −∞ µ c 2
ω 2
s =1 0
Γ 2s + 2
c
therefore, the z -component of the Poynting vector is
107
 ∞  r  ∞  R  
S z (r , z , t ) = ∑J1  ps  ∫ dω exp ( jωt − Γ s z ) Es (ω )  Γ s  
 s =1  R 
−∞
 ps  
 ∞  r ∞ 1 R  
× ∑J1  pσ  ∫ d Ω exp ( jΩt − Γσ (Ω) z ) Eσ (Ω)  jΩ  . (2.7.4)
σ =1  R  −∞ η0  cpσ  
Using the orthogonality of Bessel functions
R  r  r  R2 2
∫0 drrJ1  ps R  J1  pσ R  = 2 J1 ( ps )δ sσ (2.7.5)
the power propagating is

∞
R2 2  ∞  R  
P( z, t ) = 2π ∑ J1 ( ps )  ∫ dω exp ( jωt − Γ s z ) Es (ω )  Γ s  
s =1 2  −∞  ps  
 ∞ 1 R  
×  ∫ d Ω exp ( jΩt − Γ s z ) Es (Ω)  jΩ  (2.7.6)

−∞ η 0  cp 
s 
and the energy associated with this power
108
∞
WR ( z ) = ∫ dtP ( z, t )
−∞
∞
R2 2
= (2π )∑ J1 ( ps ) ∫d ω Es (ω ) exp  −Γ s (ω ) z  ∫d ΩEs (Ω)exp  −Γ s ( Ω ) z 
s =1 2
(2.7.7)
 R1 Ω 1  ∞
×  Γ s (ω )   j R  ∫ dt exp [ j (ω + Ω)t ].
 ps  η0  c ps  −∞
1 j (ω +Ω)t
2π ∫
With the definition of the Dirac delta function δ (Ω + ω ) ≡ dt e we have
∞
R2 2
{ }
∞
WR ( z ) = (2π ) ∑ 2
J1 ( ps ) ∫ dω Es (ω )Es (−ω )exp − z Γ s (ω ) + Γ s ( −ω ) 
2 −∞
s =1
(2.7.8)
 Γ s (ω ) R  1  ω 1 
×   − j R .
 p s  ηo  c ps 
For proceeding it is important to emphasize two features: since the field components are
real functions, it is evident that the integrand ought to satisfy Fs ( −ω ) = Fs* (ω ) .
Consequently
Es ( −ω ) = Es* (ω ) (2.7.9)
and
Γ s ( −ω ) = Γ*s (ω ). (2.7.10)
109
This last conclusion implies that if the frequency is below cut-off (| ω |< ps c / R ) then
Γ s ( −ω ) =| Γ s (ω ) | (2.7.11)
whereas if | ω |> cps / R
Γ s ( −ω ) = − j | Γ s (ω ) |. (2.7.12)
With these observations we conclude that
(2π ) 2 ∞ R 2 2  cps /R −2 Γs (ω ) z
WR ( z ) = ∑ J ( p )2Re ∫0 dω | Es (ω ) | e
η0 s =1 2 1 s
2

(2.7.13)
 − jω R | Γ s (ω ) | R  ∞ 2  ω R | Γ s (ω ) | R  

×  ∫cp /R
+ d ω | Es (ω ) |   .
 cps ps  s  cps ps  
Clearly the first integrand is pure imaginary therefore, its contribution is identically
zero and as a result, in the lossless case considered here, the energy associated with the
propagating signal does not change as a function of the location
(2π )2 ∞ R 2 2 2  ω R | Γ s (ω ) | R 
∞
W= ∑
η0 s =1 2 1 s ∫cps / R
J ( p ) 2 d ω | Es ( ω ) |  cp p . (2.7.14)
 s s 
110
Comments:
1. Only the propagating components contribute to the radiated energy.
2. The non-propagating components are confined to the close vicinity of the input
(where initial conditions were defined).
3. If all the spectrum is confined to the region 0 < ω < p1 c / R , no energy will
propagate.
4. Since the waveguide is lossless, the propagating energy does not change as a
function of z .
Exercise 2.19: Calculate the electromagnetic energy per unit length. Compare with
(2.7.14).
Let us now simplify the discussion and focus on a source which excites only the first
mode ( s = 1) i.e., Ez (r , z = 0, t ) = J 0 ( p1 r/R) E0 (t ) therefore according to Eq.(2.7.2) we get
1  R  r   r   1 ∞ ′ − jωt ′
Es (ω ) = 2  ∫ dr ′r ′J 0  ps  J 0  p1  
 R   R   2π
∫−∞
dt e E0 (t ')
J 1 ( ps ) 
R 2 0
2 (2.7.15)
δ ∞ − jωt ′
= s ,1 ∫ dt ′ e E0 (t ')
2π −∞
111
As an example, consider a signal starting from t=0 ramping up and oscillating
and decaying
 t  δ s ,1 ΩT 2
E0 (t ') = E0 sin ( Ωt ) exp  −  h ( t ) ⇒ Es (ω ) = E0 (2.7.16)
 T  2 π 1 + 2 jωT + T ( Ω − ω )
2 2 2
Substituting in Eq.(2.7.14)
1 2 4  J1 ( p1 ) 
2
∞ ( Ω T ) (ω ω − ω )
2 4 2 2
1
W=
η0 c 2
E0R 
 p 1


∫ω dω
1 = cp1/R
1 + T ( Ω − ω )  + ( 2ωT )
2 2 2
2 2
(2.7.17)
 
Normalizing to the cutoff frequency, u = ω / ω1 as well as Ω = Ω / ω1 , T = T ω1 the last
integral simplifies to read
J12 ( p1 ) 2 ∞ u u2 −1
W = ε0E R 2
0
3
Ω ∫ du 2
(2.7.18)
− −
T + Ω − u  + 4u T
p1 1 2 2 2 2 2
Its numerical analysis reveals that

2 3 J1 ( p1 ) π
2
1 Ω <1
W ∼ ε 0 E0 R Ω2  (2.7.19)
p1 4 T Ω > 1
Exercise 2.20: Compare with the dependence of the propagating energy in the frequency
Ω when the signal prescribed by Eq.(2.7.16) propagates in free space.
112
2.8 Waveguide Modes in Coaxial Line
113
Chapter 3: Waveguides – Advanced Topics
After paving the foundations of electromagnetic phenomena in waveguides, in this

chapter we shall consider some advanced phenomena including:
1. Hybrid modes, whereby the TM and TE are coupled
2. Dielectric loaded waveguide
3. Mode coupling due to geometric discontinuity
4. Reactive elements
5. Excitation of Waveguides
6. Coupling between two waveguides
7. Surface waveguides
8. Transients
9. Cavities
10. Wedge in a Waveguide
114
3.1 Hybrid Modes
In Section 2.1 it was demonstrated that if the waveguide's electromagnetic

characteristics are uniform on the cross-section, the transverse components may be
derived from the two longitudinal components that in turn are independent. For example
in a rectangular waveguide we had
⎛ x⎞ ⎛ y ⎞ −γ nx ,n y z
Ez ( x, y, z;ω ) = ∑ An ,n sin ⎜ π nx ⎟ sin ⎜ π n y ⎟ e
nx , n y
x y
⎝ ax ⎠ ⎜⎝ a y ⎟⎠
⎛ x⎞ ⎛ y ⎞ −γ mx ,m y z
H z ( x, y, z;ω ) = ∑ Bm ,m cos ⎜ π mx ⎟ cos ⎜ π m y ⎟ e (3.1.1)
⎜ a y ⎟⎠
⎝ ax ⎠
x y
mx , m y ⎝
wherein γ = k − ε r μ r (ω / c ) and k = (π nx / ax ) + (π n y / a y )
2 2 2 2 2 2
⊥ ⊥ for the transverse
magnetic (TM) mode or k⊥2 = (π mx / ax ) + (π m y / a y ) for the transverse electric (TE)
2 2
mode. The transverse components being given by

G γ jωμ G G γ jωε G
E⊥ = − 2 ∇ ⊥ Ez + 2 1z × ∇ ⊥ H z , H ⊥ = − 2 ∇ ⊥ H z − 2 1z × ∇ ⊥ Ez (3.1.2)
k⊥ k⊥ k⊥ k⊥
115
Let us now assume that the relative dielectric coefficient (ε r ) depends on the
transverse coordinates (e.g. x and y in a rectangular geometry) and so is the relative
permeability coefficient ( μ r ) , the goal being to determine the character of the modes
supported by theG structure. OurG starting pointG again G is Maxwell's equations
∇ × E = − jωμ0 μ r H ∇ ⋅ ( μ0 μ r H ) = 0
G G G G (3.1.3)
∇ × H = jωε o ε r E ∇ ⋅ (ε 0 ε r E ) = 0
which lead to the Gfollowing
G G wave equations G G G G
∇ × (∇ × E ) = − jωμ0 ∇ × ( μ r H ) = − jωμ0 [∇μ r × H + μ r ∇ × H ]
G G G G G
−∇ E + ∇(∇ ⋅ E ) = − jωμ0 [∇μ r × H + μ r ( jωε 0 ε r E )]
2
⎡ 2 ω2 ⎤ G G G G
⎢∇ + c 2 ε r μ r ⎥ Ez = ⎡⎣∇(∇ ⋅ E ) + jωμ0 ∇μ r × H ⎤⎦ 1z (3.1.4)
⎣ ⎦ G G G G G
Bearing in mind Gauss' law ∇ ⋅ (ε r E ) = 0 ⇒ ∇ ⋅ E = − E ⋅ ∇ lnε r we get
⎡ 2 ω2 ⎤ G G G
⎢∇ ⊥ + γ +ε r μr c 2 ⎥ Ez = γ ( E⊥ ⋅ ∇ ⊥ lnε r ) + jωμ0 μr (∇ ⊥ ln μr × H ⊥ ) ⋅ 1z
2
(3.1.5)
⎣ ⎦
116
and in a similar way
⎡ 2 ω2 ⎤ G G G
⎢∇ ⊥ + γ +ε r μ r c 2 ⎥ H z = −γ ( H ⊥ ⋅ ∇ ⊥ ln μ r ) − jωε 0 ε r (∇ ⊥ lnε r × E⊥ ) ⋅ 1z
2
(3.1.6)
⎣ ⎦
It is evident from these two equations (3.1.5)--(3.1.6) that Ez and H z are no longer
independent if transverse variations in the cross-section occur. Modes which are linear
combinations of TE and TM modes are called hybrid modes and they frequently
occur in microwave components.
In order to illustrate the coupling of TE and y

TM modes let us consider a dielectric loaded
waveguide as illustrated in the figure. A fraction of b
the waveguide (0 < x ≤ d ) is filled with dielectric
εr
material (ε r ) whereas the remainder ( d < x ≤ a ) has
the same characteristics as the vacuum. In each
region the material is uniform which means that the x
wave equation is d a
117
y
b
εr
⎡ 2 ⎛ ε r ⎞ ω2 ⎤
∇ +
⎢ ⊥ ⎜ ⎟ 2 + γ 2
⎥ Ez = 0 (3.1.7) d a
x
⎣ 1
⎝ ⎠ c ⎦
and its solution is
⎧ ⎡ ⎛ π m ⎞2 ω 2 ⎤
⎪ A sinh ⎢ x ⎜ ⎟ − γ −ε r 2 ⎥
2
0≤ x≤d
⎪ ⎢⎣ ⎝ b ⎠ c ⎥⎦
−γ z ⎛ y ⎞⎪
Ez = e sin ⎜ π m ⎟ ⎨ (3.1.8)
⎝ b ⎠⎪ ⎡ πm ⎞
2
ω2 ⎤
⎛
⎪ B sinh ⎢( x − a ) ⎜ ⎟ −γ − 2 ⎥ d ≤ x ≤ a
2
⎪⎩ ⎢⎣ ⎝ b ⎠ c ⎥
⎦
wherein we impose the boundary condition at x = 0 and x = a .
Similarly, for H z we have

⎡ 2 ⎛ ε r ⎞ ω2 ⎤
∇ +
⎢ ⊥ ⎜ ⎟ 2 + γ 2
⎥ Hz = 0 (3.1.9)
⎣ ⎝1⎠c ⎦
and
118
y
b
εr
⎧ ⎡ ⎛ π m ⎞2 ω 2 ⎤
⎪ C cosh ⎢ x ⎜ x
⎟ − γ −ε r 2 ⎥ 0≤ x≤d
2
d a
⎪
⎣⎢ ⎝ ⎠ ⎦⎥
b c
−γ z ⎛ πm ⎞⎪
Hz = e cos ⎜ y⎟⎨ (3.1.10)
⎝ b ⎠⎪ ⎡ πm ⎞
2
ω2 ⎤
⎛
⎪ D cosh ⎢( x − a ) ⎜ ⎟ −γ − 2 ⎥ d ≤ x ≤ a
2
⎣⎢
⎪⎩ ⎝ b ⎠ c ⎥
⎦
Imposing the boundary condition at x=d leads us to the dispersion relation. In order to
emphasize the coupling between TE and TM modes we shall develop the dispersion
relation in two stages.
Stage I: we impose the continuity of E z and H z facilitating to redefine the amplitudes

thus A and D according to
⎧ sinh(Λ ε x)
⎪ sinh(Λ d ) 0≤ x≤d
−γ z ⎛ y ⎞⎪ ε
Ez = Ae sin ⎜ π m ⎟ ⎨ (3.1.11)
⎝ b ⎠ ⎪ sinh [ Λ v ( a − x) ]
d ≤x≤a
⎪⎩ sinh [ Λ v ( a − d ) ]
and
119
y
b
εr
⎧ cosh(Λ ε x)
⎪ cosh(Λ d ) 0≤ x≤d x
−γ z ⎛ y ⎞⎪ ε d a
H z = De cos ⎜ π m ⎟ ⎨ (3.1.12)
⎝ b ⎠⎪ cosh [ Λ v ( a − x ) ] d ≤x≤a
⎪⎩ cosh [ Λ v ( a − d ) ]
wherein Λ ε2 ≡ (π m / b ) − γ 2 − ε (ω / c ) is the horizonthal wave number in the dielectric
2 2
(subscript ε ) and Λ 2v ≡ (π m / b ) − γ 2 − (ω / c ) represents the horizonthal wave number in

2 2
the vacuum (subscript v). It is evident that if d = 0 or d = a (i.e. the waveguide is filled
uniformly), the regular solutions may be readily retrieved Λ v a = jπ n or Λ ε a = jπ n .
Stage II: In our second step if 0 < d < a we need to impose continuity of E y and H y at
x = d . According to (3.1.2) the transverse components are
γ jωμ0
Ey = − ∂ E + ∂x Hz
ω 2
ω 2
y z
γ 2 +ε r 2 γ 2 +ε r 2
c c
γ jωε 0 ε r
Hy = − ∂ H − ∂ x Ez (3.1.13)
ω 2
ω 2
y z
γ 2 +ε r 2 γ 2 +ε r 2
c c
120
y
b
εr
Based on (3.1.13) and (3.1.11)-(3.1.12) we get
γ ⎛πm⎞ jωμ0 d
x
E y ( x = d − 0) = A ⎜ ⎟ + D Λ ε tanh( Λ ε d ) a
ω 2
⎝ b ⎠ ω 2
γ 2 +ε r 2 γ 2 +ε r 2
c c
γ πm jωμ0
E y ( x = d + 0) = A − D Λ v tanh[ Λ v ( a − d )] (3.1.14)
ω 2
b ω 2
γ2 + 2 γ2 + 2
c c
thus their continuity implies
⎛ πm πm ⎞ ⎡ ⎤
⎜ γ γ ⎟ ⎢ Λ v tanh [ Λ v (a − d ) ] Λε tanh(Λε d ) ⎥
A⎜ b − b = − jωμ0 D ⎢ +
2 ⎟ ⎥ (3.1.15)
⎜⎜ γ 2 +ε ω 2
ω ω 2
ω 2
γ 2 + 2 ⎟⎟ ⎢ γ2+ 2 γ 2 +ε r 2 ⎥
⎣⎢ ⎦⎥
r 2
⎝ c c ⎠ c c
121
y
b
εr
Similarly
x
−γ ⎛ πm⎞ jωε 0 ε r d a
H y ( x = d − 0) = − − Λ coth( Λ ) (3.1.16)
ω 2 ⎜⎝ b ⎟⎠ 2
D A ε ε d
ω2
γ +ε r 2
2
γ +ε r 2
c c
−γ ⎛ πm⎞ jωε 0
H y ( x = d + 0) = D ⎜ − ⎟ − A ( −Λ v )coth[ Λ v ( a − d )] (3.1.17)
ω 2
ω 2
γ2 + 2 ⎝ ⎠
b
γ2 + 2
c c
thus
⎧ ⎫ ⎛ πm πm ⎞
⎪⎪ jωε 0 jωε 0 ε r ⎪⎪ ⎜ − γ γ ⎟
A⎨ Λ v coth[Λ v (a − d )] + Λ ε coth(Λε d ) ⎬ + D ⎜ b + b 2 ⎟ = 0.
⎪ γ 2 +ω ω ⎜⎜ γ 2 +ε ω 2 ω
2 2 2
γ 2
+ε ⎪ γ + 2 ⎟⎟
⎩⎪ ⎭⎪
r 2 r 2
c 2
c ⎝ c c ⎠
(3.1.18)
The dispersion relation, in a matrix form, may by now be expressed using (3.1.15) and
(3.1.18)
122
y
b
εr
⎡ TE ⎛ ⎞ ⎤
⎢ Λ tanh(Λ d ) Λ v tanh[Λ v (a − d )] πm ⎜ 1 1 ⎟ ⎥ x
⎢ ε ε
+ γ⎜ − ⎟ ⎥d a
⎢ ⎛ γ 2 +ε ω ⎞ 1
2
⎛ 2 ω2 ⎞ 1 b ⎜ 2 ω2 ω2 ⎟⎟ ⎥
⎜ γ + c 2 ⎟ jωμ ⎜ γ +ε r 2 γ + 2
2
⎢ ⎜⎝ r
c 2 ⎟⎠ jωμ0 ⎝ ⎠ ⎝ c c ⎠ ⎥⎛ D⎞
⎢ 0
⎥⎜ ⎟ = 0
⎢ ⎛ ⎞ TM ⎥ ⎜⎝ A ⎟⎠
⎢ πm ⎜ −1 1 ⎟ jωε 0 Λ v jωε 0 ε r Λε ⎥
⎢ γ⎜ + 2 ⎟
coth[ Λ ( a − d )] + coth( Λ ε d ) ⎥
⎢ ⎜⎜ γ 2 +ε ω 2
ω ω 2 v
ω 2
⎥
γ 2 + 2 ⎟⎟
b
γ2 + 2 γ 2 +ε r 2
⎢⎣ ⎝
r
c 2
c ⎠ c c ⎥⎦
(3.1.19)
or formally
⎛ D11 D12 ⎞ ⎛ D ⎞
⎜ ⎟⎜ ⎟ = 0 (3.1.20)
⎜ ⎟⎜ ⎟
⎜D ⎟
⎝ 21 D22 ⎠ ⎜⎝ A ⎟⎠
implying that for a non trivial solution the determinant of the matrix must be zero
⎛ D11 D12 ⎞
det ⎜⎜ ⎟ = 0 ⇒ D D + D 2 = 0.
⎟ 11 22 ( 12 ) (3.1.21)
⎜D ⎟
⎝ 21 D22 ⎠
123
y
b
εr
Comments:
1. The diagonal terms represent the dispersion relation of TE and x
d a
TM modes.
2. If there are no variations in one transverse dimension (e.g. m = 0 ), the TE and
TM modes are decoupled. Such an example will be considered in the next
section.
Exercise 3.1: Analyze the solution of the dispersion relation as a function of d .

Specifically, compare to the case d = 0 and d = a .
Exercise 3.2: Calculate the group velocity and the energy velocity.
Exercise 3.3: For a given propagating power, determine the ratio of electric and magnetic
energy (per unit length) as a function of d .
124
3.2 Dielectric Loading – TM 01
In the previous section it was concluded that partial

loading in one direction and the field variations in the other
direction leads to hybrid modes. Moreover, we claimed
that if there are no field variations in the second
transverse direction the modes are not coupled. We
shall examine this kind of configuration next, namely, a Rd
R
partially filled with dielectric ε r configuration as
illustrated in Figure 2. The radius of the waveguide is
R and the inner radius of the dielectric is denoted with
Rd . We shall assume a solution of the form
Ez ( r , z ) = E (r )exp ( − jkz ) in which case the amplitude E ( r )
satisfies
⎡1 d d ω2 2⎤
⎢ r dr dr c 2 r ( )
r + ε r − k ⎥ E (r ) = 0 (3.2.1)
⎣ ⎦
125
Rd
R
in the dielectric region or explicitly for Rd < r < R :
⎡ ⎤
⎢1 d d ω 2 ⎥
⎢ r + 2 ε − k ⎥ E1 (r ) = 0
2
(3.2.2)
⎢ r dr dr
c ⎥
⎢⎣ κε 2
⎥⎦
whereas in the vacuum 0 < r < Rd :
⎡ ⎤
⎢1 d d ω 2 ⎥
⎢ r + 2 − k ⎥ E2 (r ) = 0.
2
(3.2.3)
⎢ r dr dr c
⎥
⎢⎣ κ v2 ⎥⎦
It is therefore convenient to define the radial wave number κ v2 ≡(ω / c ) − k 2 in vacuum
2
(subscript v) whereas the corresponding quantity in the dielectric is defined as

κ ε2 ≡ ε (ω / c ) − k 2 based on which the solution reads
2
126
Satisfies the boundary
condition on the
external wall
Rd
Satisfies the boundary ⎧ AS0 ( r ) 0 < r < Rd R
condition in the center Ez ( r ) = ⎨
⎩ BT0 (r ) Rd < r < R
⎧ J 0 (κ ε r )Yo (κ ε R ) − Yo (κ ε r ) J 0 (κ ε R ) κ ε2 > 0
T0 (r ) ≡ ⎨ (3.2.4)
⎩ I 0 (| κ ε | r ) K 0 (| κ ε | R) − K 0 (| κ ε | r ) I 0 (| κ ε | R ) κ ε < 0
2
⎧ J 0 (κ v r ) κ v2 > 0
S0 (r ) = ⎨
⎩ I 0 (| κ v | r ) κ v < 0
2
As in the previous section the dispersion relation will be now determined by imposing
the boundary conditions at the interface between the two regions. Firstly, we impose the
continuity of Ez at r = Rd :
Ez ⇒ AS0 ( Rd ) = BT0 ( Rd ) (3.2.5)
and secondly, the continuity of Hϕ at r = Rd
jωε 0ε r ∂Ez | κε | D | κv | D
Hϕ = ⇒ ε r BT 0 ( Rd ) = 2 A S 0 ( Rd ) (3.2.6)
k − ε r (ω / c ) ∂r κ ε2 κv
2 2
wherein
127
Rd
D ⎧− J (κ r )Y (κ R ) + Y1 (κ ε r ) J 0 (κ ε R ) 2
κε > 0 R
T 0 (r ) ≡ ⎨ 1 ε o ε
⎩ I1 (| κ ε | r ) K 0 (| κ ε | R ) + K1 (| κ ε | r ) I 0 (| κ ε | R) κ ε < 0
2
(3.2.7)
D ⎧ − J (κ r ) κ v > 0 2
S 0 (r ) = ⎨ 1 v
⎩ I1 (| κ v | r ) κ v < 0
2
From the previous two equations[(3.2.5), (3.2.6)] we obtain the dispersion equation
| κε | D | κv | D
ε r S0 T 0 = 2 S 0 T0 (3.2.8)
κε
2
κv
which can be solved numerically as
illustrated below. 20
According to the frequency, the mode
varies from the characteristics of empty
space (close to cut-off) of dielectric filled 15
f [GH z ]
case. empty waveguide
10
5 dielectric filled
waveguide
0
0 1 2 3 4 5 6
128
kR
Exercise 3.4: Analyse the effect of Rd on the phase velocity, group velocity and ohm
loss of the metal. What about dielectric loss?
Exercise 3.5: Analyze the ratio of the electric energy stored per unit length vs. magnetic
energy per unit length.
Exercise 3.6: Compare the energy stored in the dielectric with that in air.
Exercise 3.7: Determine the higher modes of a coaxial line illustrated below. The radii
are Rin , Rd , Rext .
dielectric
metal
air
129
3.3 Mode Coupling
3.3.1 Step Transition – TE mode
A longitudinal change in the cross-section of the waveguide leads to a coupling

process between virtually all possible modes. Excitation of modes is an inherent process
associated with the necessity to satisfy the boundary conditions. Whether these modes
propagate or not, it is only a question of what is the frequency (or frequencies) of the
incoming signal. In general, the coupling occurs between TE and TM or between various
hybrid modes however, in what follows we shall consider a configuration whereby the
coupling is between the various eigen-modes of a TE mode.
130
d
z
y z=0 d
a
b z
x z=0
a x
A TE 01 wave propagates from z → −∞ towards the discontinuity at z = 0

⎛ x⎞
H z(inc) = H 0 cos ⎜ π ⎟ exp ( jkz z ) (3.3.1)
⎝ a⎠
in a waveguide of width a and height b; note that k z2 = (ω / c ) − (π / a ) . At the
2 2
discontinuity the width of the waveguide varies abruptly from a to d < a . A direct result
of the discontinuity is a reflected field described by
∞
⎛ π nx ⎞
H z = H 0 ∑ ρ n cos ⎜ ( ) ( ) ( )
2 2
(ref )
⎟ exp Γ n z ; Γ m = π m / a − ω / c (3.3.2)
n =0 ⎝ a ⎠
131
d
a
ρ m are the reflection coefficients of the various modes. z

x z=0
Similarly, the transmitted field is given by
∞
⎛ π mx ⎞
H z = H 0 ∑τ m cos ⎜ ( ) ( ) ( )
2 2
(tr )
⎟ exp −Λ m z ; Λ m = π m / d − ω / c (3.3.3)
m =0 ⎝ d ⎠
τ m representing the transmitted coefficient of each mode. Three observations at this
stage:
1. The amplitude of both reflected ( ρ m ) and transmitted (τ m ) will be determined by the
boundary conditions to be imposed shortly.
2. The character of the mode (propagating or evanescent) is determined by the angular
frequency (ω ) of the incident wave.
3. The choice of TE 01 is dictated by the need for a simple analysis. If a higher mode is
launched (e.g. TE 11 ) there will be coupling to TM modes.
132
d
a
Step 1: Continuity of the y -component of the electric field
in the range 0 < x < a implies
⎧ E y( tr ) |z =0 0< x<d z
(E (inc)
y +E | =⎨
(sc)
y z =0 x (3.3.4)
z=0
⎩0 d < x < a.
Faraday's law, in our case ∂ x E y = − jωμ0 H z , thus
⎛ a ⎞ ⎛πx⎞
E y( inc ) |z =0 = − jωμ0 H o ⎜ ⎟ sin ⎜ ⎟
⎝π ⎠ ⎝ a ⎠
⎛ a ⎞ ⎛ π mx ⎞
E y(sc) |z =0 = − jωμ0 H o ∑ρ m ⎜ ⎟ sin ⎜ ⎟
m ⎝ π m ⎠ ⎝ a ⎠
⎛ d ⎞ ⎛ π m′x ⎞
E y( tr ) |z =0 = − jωμ0 H o ∑τ m′ ⎜ ⎟ sin ⎜ ⎟. (3.3.5)
m′ ′
⎝πm ⎠ ⎝ d ⎠
Substituting in (3.3.4) we get
⎧ d ⎛ π m′x ⎞
a ⎛ π mx ⎞ ⎪∑ τ sin ⎜ ⎟ 0< x<d
∑m π m m,1 m ⎜⎝ a ⎟⎠ ⎨ m′
[δ + ρ ]sin = π m ′ m′
⎝ d ⎠ (3.3.6)
⎪0 d < x < a.
⎩
Using the orthogonality of the trigonometric function we obtain
133
d
a
z
x z=0
a 1 d 1 d ⎛ π m′x ⎞ ⎛ π mx ⎞
[δ m,1 + ρ m ] = ∑ τ m′ ∫ dxsin ⎜ ⎟ sin ⎜ ⎟, (3.3.7)
πm 2 m ′ π m′ a 0
⎝ d ⎠ ⎝ a ⎠
or
δ m,1 + ρ m = ∑ Amm′τ m′ , (3.3.8)
m′
wherein the matrix Amm′ is proportional to the overlap integral of the modes from both
sides of the discontinuity
2
⎛d ⎞ ⎛ m ⎞1 d ⎛ π m′ ⎞ ⎛ π m ⎞
Amm′ ≡ 2 ⎜ ⎟ ⎜ ⎟ ∫ dx sin ⎜ x ⎟ sin ⎜ x ⎟. (3.3.9)
⎝ a ⎠ ⎝ m′ ⎠ d 0
⎝ d ⎠ ⎝ a ⎠
Step 2: Continuity of the x -component of the magnetic field in the region 0 < x < d
implies
H x( inc ) + H x( sc ) |z =0 = H x( tr ) |z =0 (3.3.10)
G G
and since ∇ ⋅ H = 0 or explicitly ∂ z H z + ∂ x H x = 0 we get at z = 0
134
d
a
z
a ⎛πx ⎞
= − H 0 sin ⎜ ⎟ ( − jkz01 )
(inc)
H x z=0
π
x
⎝ a ⎠
a ⎛ π mx ⎞
H x(sc) = − H 0 ∑ρ m sin ⎜ ⎟ Γm
m π m ⎝ a ⎠
d ⎛ π m′x ⎞
H x( tr ) = − H 0 ∑τ m′ sin ⎜ ⎟ (−Λ m ). (3.3.11)
m′ πm ′ ⎝ d ⎠
Thus, in the region z = 0 and 0 < x < d
a ⎛ π mx ⎞ d ⎛ π m′x ⎞
∑ π m 0m m,1 m ⎜⎝ a ⎟⎠ ∑
Γ [ −δ + ρ ]sin = −
m ′ π m′
Λ τ
om ′ m ′ sin ⎜
⎝ d ⎠
⎟; (3.3.12)
this time we use the orthogonality of sin (π m′x / d ) in the range 0 ≤ x ≤ d we get
1 ⎡ d ⎤ ⎡ a ⎤ 1 d ⎛ π m′ ⎞ ⎛ π m ⎞
τ m′ ⎢ Λ m′ ⎥ = ∑ ⎢ Γ m ⎥ ⎡⎣δ m ,1 − ρ m ⎤⎦ ∫ dx sin ⎜ x ⎟ sin ⎜ x ⎟, (3.3.13)
2 ⎣ π m′ ⎦ m ⎣π m ⎦ d 0
⎝ d ⎠ ⎝ a ⎠
or
τ m′ = ∑ Bm′m [δ m ,1 − ρ m ], (3.3.14)
m
where the matrix Bm′m is also proportional to the overlap integral
135
d
a
a m′ Γ m 1 d z
⎛ π m′ ⎞ ⎛ π m ⎞
Bm′m ≡ 2
d m Λ m′ d ∫0
dx sin ⎜ x ⎟ sin ⎜
⎝ d ⎠ ⎝ a ⎠
x ⎟. x (3.3.15)
z=0
We can now write G(3.3.8) and (3.3.14) in aGvector form

G G G G
δ + ρ = Aτ and τ = B(δ − ρ ) (3.3.16)
G GG G
respectively. Substituting the expression for τ we obtain δ + ρ = AB(δ − ρ ) , or
G G
[ I + AB ]ρ = [ AB − I ]δ . Finally the amplitudes of the reflected and transmitted
amplitudes are given by G
G
ρ = [ I + AB] [ AB − I ]δ
−1
G G (3.3.17)
τ = 2[ I + B A]−1δ .
For evaluation of the matrices A and B we need to evaluate the integral
1 d ⎛ π m′x ⎞ ⎛ π mx ⎞ ⎡ ⎛ d ⎞⎤ ⎡ ⎛ d ⎞⎤
d ∫0
dx sin ⎜ sin
⎟ ⎜
⎝ d ⎠ ⎝ a ⎠
⎟ = sinc ⎢π
⎣ ⎝
⎜ m ′ − m ⎟⎥
a ⎠⎦
− sinc ⎢π
⎣ ⎝
⎜ m ′ + m ⎟ ⎥. (3.3.18)
a ⎠⎦
136
d
a
With the amplitudes established we may proceed and evaluate z

the power in the system: x z=0
2
1 2 a ∞ 2 ⎛ a ⎞
Pinc = H 0 b ∑ | δ m ,1 | ⎜ ⎟ Re(− jωμ0Γ m )
2 2 m =1 ⎝πm ⎠
2
1 2 a ∞ ⎛ a ⎞
Pref = H 0 b ∑ | ρ m |2 ⎜ ⎟ Re( jωμ0Γ m )
2 2 m =1 ⎝ π m ⎠
2
1 2 d ∞ ⎛ d ⎞
Ptr = H 0 b ∑ | τ m′ |2 ⎜ ⎟ Re(− jωμ0 Λ m ) (3.3.19)
2 2 m =1 ⎝ π m ′ ⎠
Assuming that at the operating frequency there is only a single propagating mode in
each region, then the average power may be written as in a transmission line
1 ω k a ab
Pinc = | I1 |2 Z1 ⇒ Z1 = η0 , I1 = H 0 1
2 ck1 π 2
1
Pref = | I1 |2 | ρ1 |2 Z1
2
1 ω kd db
Ptr = | I 2 |2 | τ 1 |2 Z 2 ⇒ Z 2 = η0 , I2 = H0 2 (3.3.20)
2 ck2 π 2
137
(ω / c ) − (π / a ) and k2 = (ω / c ) − (π / d ) .
2 2 2 2
wherein k1 =
1.2
Reflection coefficient
Let us examine closely the normalized reflected power
1.0
⎡ P1( ref ) / P1(inc ) ⎤ in the first mode as a function of the
⎣ ⎦
0.8
frequency:
1. The power can be injected at frequencies 0.6
above 6GHz (a=2.5cm).
2. To the point whereby the first mode in the 0.4
4 6 8 10 12 14 16 18
second part is above cutoff
f c 8.485 GHz ( d = a / 2 ) f [GH z ]
the reflection coefficient of the first mode is unity.
3. Above 8.485GHz power can be transferred to the second part therefore the
reflected power drops yet P1( ) = P1( ) + P1( ) .
inc ref tr
4. Above 12 GHz for which the second mode in the left-hand side may carry
power, the power reflected in the first mode decreases further yet beyond
this point P1( ) ≠ P1( ) + P1( ) but rather P1( ) = P1( ) + P2( ) + P1( ) .
inc ref tr inc ref ref tr
138
3.3.2 Step Transition – TM 0 n Mode in Cylindrical Waveguide
An additional example illustrating mode-coupling due

to discontinuity in the transverse geometry is brought next. In
fact, in this case there are two discontinuities, and the first
and last sections have the same cross section. It will be
reiterated that even if the incident wave is composed of a Rint
single (lowest) mode, the coupling causes other modes that in d
turn may carry part of the electromagnetic energy.
In the left hand side ( z < 0 ) of a cylindrical waveguide Rext
of radius Rext , there is an incident and a reflected TM wave
⎛ r ⎞
⎟ exp ( −Γ1 z )
Ez(inc) (r , z ) = E0 J 0 ⎜ p1 z
z=0
⎝ Rext ⎠
∞
⎛ r ⎞
Ez (r , z ) = E0 ∑ ρ s J 0 ⎜ ps
(ref )
⎟ exp ( Γ1 z ), (3.3.21)
s =1 ⎝ Rext ⎠
where Γ s ≡ ( ps /Rext ) 2 − (ω /c) 2 and ρ s represents the normalized amplitudes of the
139
Rint
d
reflected modes. In the region 0 ≤ z ≤ d the radius of the waveguide is Rext
smaller Rint < Rext however for z ≥ d the radius returns to its initial value.
z
Consequently, in the other regions we have z=0
∞
⎛ r ⎞
Ez (r , z < d ) = E0 ∑ J 0 ⎜ pn
(d )
⎟ ⎡⎣ An exp ( −Λ n z ) + Bn exp ( Λ n z ) ⎤⎦
n =1 ⎝ Rint ⎠
(3.3.22)
∞
⎛ r ⎞
Ez( tr ) (r , z > d ) = E0 ∑ τ s J 0 ⎜ ps ⎟exp [ −Γ s ( z − d ]
s =1 ⎝ Rext ⎠
where Λ n = ( pn / Rint ) 2 − (ω / c) 2 . As previously, the goal is to determine the

amplitudes of the transmitted and reflected waves (τ s and ρ s ). For this purpose let us
concentrate on the field components at z = 0 .
⎛ H ϕ(left ) ⎞ Rext ⎛ jωε 0 ⎞ ⎛ r ⎞ ⎛ δ s ,1 exp ( −Γ1 z ) + ρ s exp ( Γ s z ) ⎞
⎜ (left ) ⎟ = E0∑ ⎜ Γ ⎟ 1⎜ s
J p ⎟⎜ ⎟
⎝ r ⎠
E s p 1 ⎝ s ⎠ ⎝ R δ
ext ⎠ ⎝ s ,1 exp ( −Γ 1 z ) − ρ s exp ( Γ s z ) ⎠
(3.3.23)
⎛ Hϕ ⎞ Rint ⎛ jωε 0 ⎞ ⎛ r ⎞ ⎛ An exp ( −Λ n z ) + Bn exp ( Λ n z ) ⎞
(d)
⎜ (d) ⎟ = E0∑ ⎜ Λ ⎟ 1⎜ n
J p ⎟⎜ ⎟
⎝ r ⎠
E n p n ⎝ n ⎠ ⎝ Rint ⎠ ⎝ n
A exp ( −Λ n z ) − B n exp ( Λ n z ) ⎠
With these field components we impose the boundary conditions at z = 0 . Continuity of
140
Rint
d
the radial electric field implies Rext
− ⎧ Er ( r , z = 0 + ) 0 < r < Rint

Er (r , z = 0 ) = ⎨ (3.3.24) z
⎩0 Rint < r < Rext z=0
Using the orthogonality of the Bessel functions

Rext ⎛ r ⎞ ⎛ r ⎞ Rext 2
∫0 ( ps )δ ss ′
2
drr J 1 ⎜ p s ⎟ J 1 ⎜ p s ′ ⎟ = J 1 (3.3.25)
⎝ Rext ⎠ ⎝ Rext ⎠ 2
we obtain from (3.3.24)
1 ∞
Rint ⎛ r ⎞ ⎛ r ⎞
∑
Rint
δ s ,1 − ρ s = Λ [ A − B n ∫0
] drr J 1⎜ s
p ⎟ 1⎜ n
J p ⎟ (3.3.26)
Γ s Rext Rext2
2 n =1 pn
n n
⎝ Rext ⎠ ⎝ Rint ⎠
J 1 ( ps )
ps 2
or
∞
δ s ,1 − ρ s = ∑ Lsn[ An − Bn ], (3.3.27)
n =1
where the matrix Lsn is proportional to the overlap integral

ps Λ n 1 Rint 3
2 Rint ⎛ r ⎞ ⎛ r ⎞
Ri2nt ∫0
Lsn ≡ U s ,n , U s ,n ≡ dr r J 1⎜ s
p ⎟ 1⎜ n
J p ⎟ (3.3.28)
pn Γ s J12 ( ps ) Rext
3
⎝ Rext ⎠ ⎝ Rint ⎠
141
Rint
d
Rext
The continuity of the azimuthal magnetic field in the region 0 < r < Rint is
Hφ( d ) ( r , z = 0− ) = Hφ( d ) ( r , z = 0+ ) . Again we use the orthogonality of the
z
z=0
Bessel functions in the inner region

Rint 2
Rint ∞
R Rint ⎛ r ⎞ ⎛ r ⎞
[ An + Bn ] J1 ( pn ) = ∑[δ s ,1 + ρ s ] ext ∫ drr J1 ⎜ ps
2
⎟ 1⎜ n
J p ⎟ (3.3.29)
pn 2 s =1 ps 0 ⎝ Rext ⎠ ⎝ Rint ⎠
hence
An + Bn = ∑ Mns (δ s ,1 + ρ s ), (3.3.30)
s
where using the definition (3.3.28) we have further defined

R p 1
Msn ≡ ext n 2 U s ,n . (3.3.31)
Rint ps J1 ( pn )
From these expressions we can easily deduce the two equations representing the
boundary conditions at z = d :
∞
An exp ( −Λ n d ) + Bn exp ( Λ n d ) = ∑ Mnsτ s (3.3.32)
s =1
142
Rint
d
∞ Rext
τ s = ∑ Ls ,n[ An exp ( −Λ n d ) − Bn exp ( Λ n d )]. (3.3.33)
n =1 z
z=0
All four equations (3.3.27), (3.3.30), (3.3.32), (3.3.33) can be formulated
in a vector (matrix) notation
( 3.3.27 ) → δ − ρ = L ( A − B ) ⎫ ⎛ A ⎞ ⎛ I − I ⎞ ⎛ L−1 −L−1 ⎞ ⎛ δ ⎞

−1
⎬⇒⎜ ⎟=⎜
( 3.3.30 ) → A + B = M (δ + ρ )⎭ ⎝ B ⎠ ⎝ I I ⎟⎠ ⎜⎝ M M ⎟⎠ ⎜⎝ ρ ⎟⎠
( 3.3.32 ) → Ae−Λd + BeΛd = Mτ ⎫⎪ ⎛ A ⎞ ⎛ e−Λd 0 ⎞−1 ⎛ I − I ⎞−1 ⎛ L−1 −L−1 ⎞ ⎛ τ ⎞
⎬⇒⎜ ⎟=⎜ ⎟ ⎜ ⎟⎜ ⎟
( 3.3.33) → L ( Ae − Be ) = τ ⎪⎭ ⎝ B ⎠ ⎝ 0 eΛd ⎠ ⎜⎝ I I ⎟⎠ ⎝ M
−Λd Λd
M ⎠⎝ 0⎠
(3.3.34)
Substituing we get
143
Rint
d
Rext
z
z=0
−1 −1 −1
⎛ δ ⎞ ⎛ L−1 −L−1 ⎞ ⎛ I − I ⎞ ⎛ e −Λd 0 ⎞ ⎛I − I ⎞ ⎛ L−1 −L−1 ⎞ ⎛ τ ⎞
⎜ρ⎟=⎜ ⎟ ⎜ ⎜
I ⎟⎠ ⎝ 0
⎟ ⎜ ⎜ ⎟
I ⎟⎠ ⎝ M M ⎠ ⎜⎝ 0 ⎟⎠
⎝ ⎠ ⎝ M M ⎠ ⎝I e Λd ⎠ ⎝ I
or (3.3.35)
−1 −1
⎛τ ⎞ ⎛ L−1 −L−1 ⎞ ⎛ I − I ⎞ ⎛ e −Λd 0 ⎞⎛ I − I ⎞ ⎛ L−1 −L−1 ⎞ ⎛ δ ⎞
⎜0⎟ = ⎜ ⎟ ⎜ ⎜
I ⎠⎟ ⎝ 0
⎟⎜ ⎜ ⎟
I ⎟⎠ ⎝ M M ⎠ ⎜⎝ ρ ⎟⎠
⎝ ⎠ ⎝M M ⎠ ⎝I e Λd ⎠ ⎝ I
or using a short notation
⎛ Q11 Q12 ⎞⎛ δ ⎞ ⎛ τ ⎞
⎜Q ⎟⎜ ρ ⎟ = ⎜ 0 ⎟ (3.3.36)
⎝ 21 Q 22 ⎠⎝ ⎠ ⎝ ⎠
thus Q11δ + Q12 ρ = τ and Q21δ + Q22 ρ = 0 . Consequently,
ρ = −Q22−1Q21δ , τ = [Q11 − Q12Q22−1Q21 ]δ . (3.3.37)
The Figure below illustrates the reflection coefficient and power conservation as a
function of the frequency. We observe that below the cut-off of the second mode all the
propagating power is in the first mode. As the frequencies exceed the cut-off of the
second mode, part of the power leaves the system through the second mode.
144
0.9 1.1
0.8
2
1.0
ρ11 + τ 11
ρ11
0.7
2
0.9
0.6
0.5 0.8
5 10 15 20 5 10 15 20
f [GH z ] f [GH z ]
The reflection coefficient as a function of the frequency and energy conservation of the first mode. Note that as
long as the higher mode is below cutoff, | ρ 1,1 |2 + | τ 1,1 |2 = 1 while above cutoff, energy is transferred from the
first mode to the second and therefore | ρ 1,1 |2 + | τ 1,1 |2 ≠ 1 .
145
3.4 Reactive Elements
Obstacles in waveguides or transmission lines may cause reflections which in turn
may be detrimental. Power reflected implies obviously reduction in the transmitted power
but it entails that some energy is stored in the evanescent waves which develop in the
vicinity of the obstacle.
3.4.1 Inductive Post
As an example we shall examine a cylindrical post and an incident TE 01 mode

determined by
⎛ x⎞
E y( ) ( x, z ) = E0sin ⎜ π ⎟ exp ( −Γ1 z ) ; Γ1 = (π / a ) 2 − (ω / c) 2 = j β
p
(3.4.1)
⎝ a⎠
In zero order this primary field
excites in the metallic cylinder a y
current density
J y ( x, z ) = Iδ ( z )δ ( x − d ) (3.4.2)
which may be interpreted as surface
current density
b 2R
d x
146
a
∞
⎛ x⎞
J y , s ( x ) = I δ ( x − d ) = ∑U m sin ⎜ π m ⎟ (3.4.3)
m =1 ⎝ a⎠
hence
2 ⎛ d⎞
Um = I sin ⎜ π m ⎟. (3.4.4)
a ⎝ a ⎠
This current generates a secondary field
∞
⎛ x⎞
E y ( x, z ) = ∑Em sin ⎜ π m ⎟ exp ( −Γ m z )
(sec)
(3.4.5)
m =1 ⎝ a⎠
where Γ 2m = (π m / a )2 − ω 2 / c 2 . In order to establish the relation between the current's
``harmonics'' and the secondary field we observe that based on Maxwell's equations the
magnetic field
−1 ∞ ⎛ x⎞
(sec)
H x ( x, z ) = ∑ m m ⎜⎝ a ⎟⎠ exp ( −Γ m z ) sgn( z )
jωμ0 m =1
Γ E sin π m (3.4.6)
enables via its discontinuity condition

H x ( x, z = 0 + ) − H x ( x, z = 0 − ) = J y ,s (3.4.7)
to determine the amplitudes of the modes describing the secondary field in terms of the
current on the filament
147
y
b 2R
I ⎛ d ⎞ jωμ0 x
Em = − sin ⎜ π m ⎟
d
. (3.4.8) a
a ⎝ a ⎠ Γm
We have yet to determine the relation between the current ( I ) in the filament and the
amplitude of the incident wave. This is done by relaxing the condition of our infinitely
small filament and imposing that the tangential electric field vanishes on the wall of the
post. Specifically, averaging the electric field over the azimuthal coordinate at r = R we
must get zero total field, namely
2π
∫0
dφ[ E y( inc ) + E y(sec) ]x = d + R cos φ = 0. (3.4.9)
z = Rsinφ
explicitly
∞
⎡π m ⎤
( )
2π
∑
m =1
∫0
d φ ⎡ E δ
⎣ 0 m ,1 exp ( −Γ m R sin φ ) + E m exp −Γ m R sinφ ⎤ sin
⎦ ⎢⎣ a ( d + R cos φ ) ⎥⎦ (3.4.10)
which formally may be rewritten by defining
1 2π ⎡π m ⎤
( ) ( )
2π ∫0
L(1)
m ≡ d φ exp −Γ m R sin φ sin ⎢⎣ a d + R cos φ ⎥⎦ (3.4.11)
and
1 2π ⎡π m ⎤
L(2)
m ≡
2π ∫ d φ exp ( −Γ m R sin φ ) sin ⎢ ( d + R cos φ ) ⎥ (3.4.12)
0
⎣ a ⎦
148
y
b 2R
hence d
a
x
∞
1 ⎛ d ⎞ jωμ0 (2) L1(1) E0 a
E L − I ∑ sin ⎜ π m ⎟
(1)
Lm = 0 ⇒ I = .
0 1
⎝ ⎠ mΓ ∞
η0
∑L ( jω / cΓ ) sin (π md / a )
m =1 a a (2)
m m
m =1
(3.4.13)
At this stage we may define the reflection coefficient
Em −L1 ( jω / cΓ m ) sin (π md / a )
(1)
ρm ≡ = ∞ (3.4.14)
∑L(2)n ( jω / cΓn ) sin (π nd / a )
E0
n =1
or for the fundamental
−L1(1) (ω / cβ ) sin (π d / a )
ρ1 = ∞
. (3.4.15)
L1(2) (ω / cβ ) sin (π d / a ) + ∑L(2)
n ( jω / cΓ n ) sin ( π nd / a )
n =2
At this point it can be shown that since m = 1 is a propagating mode L1(1) = Re( L1(2) ) and
the second term in the denominator is pure imaginary
−1
ρ1 = . (3.4.16)
(2)
Im(L1 ) ∞
Ln ⎛ β ⎞ sin (π nd / a )
(2)
1+ j
L1(1)
+ j ∑ (1) ⎜ ⎟
n =2 L1 ⎝ Γ n ⎠ sin ( π d / a )
149
y
b 2R
d x
a
Based on this last result, we conclude that the post may be represented by a reactance
since according to the equivalent scheme below
−1 −1
ρ1 = =
1 + 2Z L 1 + 2 j X L
and Z0 ZL Z0
1 Im[L1(2) ] ∞ L(2) β sin (π nd / a )

XL =
2 L1 (1)
+ ∑ n
(1)
n =2 L1 Γ n sin (π d / a )
.
Further simplification, is achieved assuming that R d we get

⎛ d⎞ 1 π ⎛ d⎞
Lm sin ⎜ π m ⎟
(1)
∫ d φ exp ( −Γ R sin φ ) sin ⎜ π m ⎟ I0 ( Γm R ) (3.4.17)
⎝ a ⎠ 2π −π m
⎝ a ⎠
⎛ d⎞ 2 π
Lm sin ⎜ π m ⎟
(2)
∫ dφ exp ( −Γ m R sin φ )
⎝ a ⎠ 2π 0
(3.4.18)
⎛ d⎞⎡ ∞
⎛1 ⎞ ⎤
sin ⎜ π m ⎟ ⎢ I 0 ( Γ m R ) + 2∑ ( −1) sinc ⎜ π k ⎟ I k ( Γ m R ) ⎥
k
⎝ a⎠⎣ k =1 ⎝2 ⎠ ⎦
thus
150
⎡ I 0 ( Γ n R ) 4 ∞ ( −1)k I 2 k +1 ( Γ n R ) ⎤
∞
2⎛ d⎞
∑ ⎢
n =2 ⎢ Γ
−
π
∑k = 0 2k + 1 Γ
⎥ sin ⎜
⎝
π n ⎟
⎠
⎣ n R n R ⎥
⎦ a
XL = (3.4.19)
J0 ( β R) 2 ⎛ d ⎞
sin ⎜ π ⎟
βR ⎝ a⎠
At the limit R → 0 or more precisely assuming β R 1 and Γ n R 1 thus
Iν ( x 1) ( x / 2 ) / ν ! we get
ν
⎛a 1⎞
Int ⎜
4 ∞ ( −1) ( Γ n R ) ⎛ 1 ⎞ ⎤
⎟
⎝ Rπ ⎠ ⎡ 1 k 2k 2 k +1
2⎛ d⎞
∑ ⎢ − ∑ ⎜ ⎟ ⎥ sin ⎜ π n ⎟
n =2 ⎢⎣ Γ n R π k =0 2k + 1 ( 2k + 1)! ⎝ 2 ⎠ ⎥⎦ ⎝ a⎠
XL = (3.4.20)
1 ⎛ d⎞
sin 2 ⎜ π ⎟
βR ⎝ a⎠
151
Exercise 3.8: Analyze (numerically) the dependence of the normalized reactance on three
parameters; the angular frequency (ω ) , its location ( d ) and its radius ( R ) . It is possible
to simplify the two quantities defined in (3.4.11) and (3.4.12) by using
⎡ z ⎛ 1 ⎞⎤ ∞
exp ⎢ ⎜ t − ⎟ ⎥ = ∑ t k J k ( z ) ; J − k ( z ) = (−1) k J k ( z ).
⎣ 2 ⎝ t ⎠ ⎦ k = −∞
(For solution see Appendix 10.1)
Exercise 3.9: Calculate the energy stored in the vicinity of the post and compare it with
energy per unit length carried by the incident wave.
Exercise 3.10: Analyze the ratio of electric to magnetic energy stored around the post.
152
3.4.2 Diaphragm
Another element (obstacle) y

that may deflect radiation but not
absorb it is the diaphragm. This is a
thin metallic layer occupying part of
the cross-section. Consider a
diaphragm occupying the region b
d <x<a its thickness being
negligible compared to the
wavelength and the skin depth. The d x
electric field is given by a
⎧ ⎛ x⎞ ∞ ⎛ x⎞
⎪⎣⎡ exp ( − j β z ) 1 ( )⎦ ⎜ ⎟ ∑ m ( m ) ⎜
+ ρ exp j β z ⎤ sin π + ρ exp Γ z sin π m ⎟ z<0
⎪ ⎝ a ⎠ m =2 ⎝ a⎠
E y = E0 ⎨ ∞
⎪τ e− j β z sin ⎛ π x ⎞ + τ exp ( −Γ z ) sin ⎛ π m x ⎞
⎪⎩ 1 ⎜ ⎟ ∑ m m ⎜ ⎟ z > 0.
⎝ a ⎠ m =2 ⎝ a ⎠
(3.4.21)
Let us denote by E y ( x, z = 0 ) = E0E ( x) the transverse electric field in the aperture, thus
153
y
d x
2 d ⎛ x⎞
a
m = 1:
a ∫
1 + ρ1 = τ 1 =
0
dx E ( x ) sin ⎜π ⎟
⎝ a⎠
(3.4.22)
2 d ⎛ x⎞
m≥ 2: ρm τ m
= =
a ∫0
dx E ( x ) sin ⎜
⎝
π m ⎟.
a⎠
(3.4.23)
E ( x) represents the horizontal variation of the vertical electric field ( E y ) . Continuity of

H x in the aperture (0 < x < d ) implies
⎛ x⎞ ∞ ⎛ x⎞ ⎛ x⎞ ∞ ⎛ x⎞
− j β (1 − ρ1 ) sin ⎜ π ⎟ + ∑ρ m Γ m sin ⎜ π m ⎟ = − j βτ 1 sin ⎜ π ⎟ − ∑τ m Γ m sin ⎜ π m ⎟.
⎝ a ⎠ m =2 ⎝ a⎠ ⎝ a ⎠ m =2 ⎝ a⎠
(3.4.24)
Substituting (3.4.23) we get
⎛ x⎞ ∞
⎛ Γm ⎞ ⎛ x⎞4 d ⎛ x′ ⎞
−2 ρ1 sin ⎜ π ⎟ = − ∑ ⎜ ⎟ sin ⎜ π m a ⎟ a ∫0 dx E ( x ) sin ⎜ π m a ⎟
′ ′ (3.4.25)
⎝ a⎠ m =2 ⎝ − j β ⎠ ⎝ ⎠ ⎝ ⎠
and further multiplying and deviding by 1 + ρ1 and using (3.4.22) we finally obtain
2 ρ1 ⎛ x⎞ d ⎛ x' ⎞ ∞
⎛ Γm ⎞ ⎛ x⎞ d ⎛ x′ ⎞
sin ⎜ π ⎟ ∫ dx E ( x )sin ⎜ π ⎟ = 2 j ∑ ⎜
′ ′ sin
⎟ ⎜ π m ⎟ ∫0 dx′E ( x′ )sin ⎜ π m ⎟.
1 + ρ1 ⎝ a ⎠ 0
⎝ a ⎠ m=2 ⎝ β ⎠ ⎝ a ⎠ ⎝ a ⎠
(3.4.26)
154
y
2 ρ1
Since the admittance is Y L = − in the left hand side may be rewritten
b
1 + ρ1 d x
a
⎛ x⎞ d ⎛ x' ⎞ ∞
⎛ Γm ⎞ ⎛ x⎞ d ⎛ x′ ⎞
YL sin ⎜ π ⎟ ∫ dx E ( x )sin ⎜ π ⎟ = −2 j ∑ ⎜
′ ′ sin
⎟ ⎜ π m ⎟ ∫0 dx ′E ( x ′)sin ⎜ π m ⎟. (3.4.27)
⎝ a ⎠ 0
⎝ a ⎠ m=2 ⎝ β ⎠ ⎝ a ⎠ ⎝ a ⎠
So far the formulation is exact (of course subject to the specified assumptions). At this
stage we multiply both sides by E ( x) and integrate over the cross section of the aperture
and get
2
⎡ d ⎛ x′ ⎞ ⎤
∞
⎛Γ ⎞ ⎢ ∫ dx′E ( x′)sin ⎜ π m ⎟ ⎥
⎝ a ⎠⎦
YL = −2 j ∑ ⎜ m ⎟ ⎣
0
. (3.4.28)
m=2 ⎝ β ⎠ ⎡ d
2
⎛ x ' ⎞⎤
⎢ ∫0 dx′E ( x′)sin ⎜ π ⎟⎥
⎣ ⎝ a ⎠⎦
which evidently entails that the distribution of the electric field in the aperture
determines the admittance of the diaphragm as experienced by the wave. For a rough
approximation we may consider the simplest form of E ( x ) that satisfies the boundary
conditions at x = 0 and x = d but it ignores edge effect namely, E ( x) ∝ x ( d − x ) . An
estimate of the edge effect may be considered by E ( x) ∝ x ( d − x ) with 0 < ν ≤ 1.
ν
155
Exercise 3.11: Compare the admittace using the approximation E ( x) = x ( d − x ) with that
evaluated by Lewin (Theory of Waveguides)
2π ⎡ −4 ⎛ π d ⎞ ⎤
YL = j ⎢1 − sin ⎜ ⎟ ⎥.
βa ⎣ ⎝ 2 a ⎠⎦
Exercise 3.12: What is the effect of the edge of the diaphragm?
156
3.5 Excitation of Waveguides – Probe Antenna
Coupling of fields from one guide to another may be accomplished by means of

small antennas by means of small radiating apertures located at appropriate positions in
the common wall separating the two guides. A completely rigorous solution of the
antenna boundary-value problem in a waveguide is beyond the scope of this course.
Nevertheless, by making suitable approximations, solutions which shed considerable
light on the behavior of antennas in waveguides, as well as providing useful engineering
results, can be obtained for many practical cases. In this section attention will be confined
to the small coaxial-line probe.
x
Input impedance. The type y
of coaxial-line probe antenna h

L
to be analyzed is illustrated a
next. It consists of a small
coaxial line, terminated in b
a
d
the center of the broad face
x
of a rectangular guide, with
its inner conductor extending z
a distance d into the guide. t
157
In order to have the antenna radiate in one direction only, a short-circuiting plunger is
placed at a distance L to the left of the probe. The basic concept is to generate a current
( I ) at a location where the parallel electric field is maximum -- quarter wavelength
from a short circuiting plane.
y
L
z
2Rext
2Rint
The total field in the waveguide could be found if Ea and H a , the total tangential
fields in the aperture were known. Unfortunately, these aperture fields are unknown,
but, if the coaxial-line opening is small, we may, to a first approximation, assume that
the higher-order coaxial-line modes which being excited, are negligible. The field in the
aperture will then be that associated with the incident and reflected TEM modes in the
coaxial line.
158
y
L
Relying on Poynting's complex theorem we find by denoting by P the 2Rext
2Rint
power flow far away from the probe that
1
− ∫dVJ *y E y = P + 2 jω [〈WM 〉 − 〈WE 〉 ] (3.5.1)
2
where the current density may be approximated, assuming a thin probe, by
I ( y) ⎡
( ) + ( z − L) 2 ⎤.
2
J y ( x, y , z ) = u R − x − h (3.5.2)
π Rint ⎢⎣
2 int ⎥⎦
This current density excites a TE nm mode satisfying
⎛ 2 ω2 ⎞ 1 ∂2
⎜ ∇ + c 2 ⎟ E y = jωμ0 J y − jωε ∂y 2 J y (3.5.3)
⎝ ⎠ 0
which in terms of Green's function reads
⎡ c2 ∂2 ⎤
E y ( x, y, z ) = − jωμ0 ∫dx′dy′dz′G y ( x, y, z | x′, y′, z′) ⎢ J y ( x ', y ', z ' ) + 2 J ( x ', y ', z ' ) ⎥.
⎣ ω ∂y ' 2 y
⎦
(3.5.4)
In turn Green's function for our configuration is given by
159
y
L
G G ∞
1 ⎛ x⎞ ⎛ x′ ⎞
2Rext
G y (r | r ') = ∑
n =1, m =0 ( 2 Γ ) ab
(1 + δ m ,0 )sin ⎜
⎝
π n
a ⎠
sin
⎟ ⎜
⎝
π n
a
⎟
⎠
2Rint
nm
⎛ y⎞ ⎛ y′ ⎞
⎝ b⎠ ⎝ b⎠
{
× cos ⎜ π m ⎟ cos ⎜ π m ⎟ exp ⎡⎣ −Γ n ,m z − z′ ⎤⎦ − exp ⎡⎣ −Γ n ,m ( z + z′) ⎤⎦ }
(3.5.5)
Exercise 3.13: Prove that this is the solution of
⎛ 2 ω2 ⎞
⎜∇ + 2 ⎟ G y = −δ ( x − x′)δ ( y − y′)δ ( z − z′)
⎝ c ⎠
∂G y
G y ( x = 0, a ) = 0, = 0, G y ( x, y, z = 0) = 0.
∂y y =0,b
Substituting (3.5.4) and (3.5.2) in (3.5.1) as well as taking advantage of the probe being
thin we have
jωμ0 ⎡ c2 ∂2 ⎤
∫ ∫ ′ ′ ′ + ′ = P + 2 jω[〈WM 〉 − 〈WE 〉 ] (3.5.6)
*
dyI ( y ) dy g ( y | y ) ⎢ I ( y ) I ( y ) ⎥
2 ⎣ ω ∂y
2 '2
⎦
wherein
160
y
L
1 2Rext
g ( y | y′) ≡ G ( x = h, y, z = L | x′ = h, y′, z′ = L ) = ∑ (1 + δ m ,0 )
( 2Γ ) ab
2Rint
n =1 n ,m
m =0
(3.5.7)
⎛ h⎞ ⎛ y⎞ ⎛ y′ ⎞
× sin 2 ⎜ π n ⎟ ⎡⎣1 − exp ( −2Γ n ,m L ) ⎤⎦ cos ⎜ π m ⎟ cos ⎜ π m ⎟
⎝ a⎠ ⎝ b⎠ ⎝ b⎠
In terms of the current at the aperture of the coax (interface coax-waveguide) the current
is I (0) therefore, we may define the input impedance
P + 2 jω[〈WM 〉 − 〈WE 〉 ]
Z in ≡ (3.5.8)
1 2
| I (0) |
2
hence
d * d ⎡ c2 ∂2 ⎤
Z in jωμ0 ∫ dyI ( y ) ∫ dy′ g ( y | y′) ⎢ I ( y ') + 2 '2 I ( y′) ⎥. (3.5.9)
0 0
⎣ ω ∂y ⎦
At the end of the probe the current is zero, therefore, for simplicity sake we assume
2
I ( y) ⎛ y − d ⎞
I ( y) = =⎜ ⎟ (3.5.10)
I (0) ⎝ d ⎠
or
161
d
2
⎛ y−d ⎞ d ′ ⎡ ⎛ y ′ − d ⎞ 2 2c 2 ⎤
Z in = jωμ0 ∫ dy ⎜ ⎟ ∫0 dy g ( y | y′) ⎢⎜ ⎟ + 2 2⎥ (3.5.11)
0
⎝ d ⎠ ⎢⎣⎝ d ⎠ ω d ⎥⎦
Let us assume now that except the first mode, all the modes are evanescent. Clearly
their contribution to the impedance is pure imaginary (inductive). Only the propagating
mode has both imaginary and real contribution to the impedance
η0 d 2 ω ⎛ h⎞ ⎡ ⎛ω ⎞ ⎤
−2
Z in = sin 2 ⎜ π ⎟ sin ( β L ) ⎡⎣sin ( β L ) + j cos ( β L ) ⎤⎦ ⎢1 + 6 ⎜ d ⎟ ⎥ (3.5.12)

18 ab cβ ⎝ a⎠ ⎢⎣ ⎝ c ⎠ ⎥⎦
y
L
Comment: The required values of L and d b
to make X = 0 and R = Z c , so that all the k0 d z
incident power is coupled into the guide,

2Rext
2Rint
may be found graphically from a plot of the 3

X = 0 and R = constant contours in the
k0 d − β L plane. Such a plot is given in the R = 100 Ω
R = 70 Ω
figure above for a guide of dimensions R = 50 Ω
2.3 × 1cm and a probe diameter equal to 2
0.23cm at a wavelength of 3.14cm. The
X =0
1
162 π /4 π /2 3π / 4 π β 10 L
intersection of the two curves determines the required parameters. From the figure it is
seen that the self-resonant length of the antenna, corresponding to β L = 0 or π / 2 , is
0.28λ0 . Also it is seen that, for k0 d less than 1.22 ( d < 0.61 centimeter), a value of β L
which will make X = 0 does not exist. The reason is that the TE 10 mode standing wave
between the antenna and the short-circuiting plunger cannot provide enough inductive
reactance to counterbalance the high capacitive reactance of a short probe.
163
3.6 Coupling between Waveguides by Small Apertures
Electromagnetic energy may be coupled from one waveguide into another guide or
into a cavity resonator by a small aperture located at a suitable position in the common
wall. For apertures whose linear dimensions are small compared with the wavelength, an
approximate theory is available which states that the aperture is equivalent to a
combination of radiating electric and magnetic dipoles, whose dipole moments are
respectively proportional to the normal electric field and tangential magnetic field of the
incident wave. This theory was originally developed by H.A. Bethe in ``Theory of
Diffraction by Small Holes", Physical Review, vol. 66, pp. 163--182, 1944.
164
3.7 Surface Waveguides
In addition to the closed cylindrical conducting tube and the conventional TEM
wave-transmission line, there exists a class of open boundary structures which are
capable of guiding an electromagnetic wave. The field is characterized by an exponential
decay away from the surface and having the usual propagation function exp ( − j β z )
along the axis of the structure. Moreover it has a discrete spectrum of eigenmodes. This
type of wave is called a surface wave, and the structure which guides this wave may be
appropriately referred to as a surface waveguide. Some typical structures that are capable
of supporting a surface wave are illustrated in Figure 10. These consist of dielectric-slabs,
fibers, dielectric-coated planes and wires and corrugated planes and wires.
z
z
(a) (b)
z z
(c) (d)
165
Although surface waveguides have several features which are similar to those of the
cylindrical conducting-tube guide, they have many characteristics which are quite
different. Some of the outstanding differences are:
1. Propagation with no low-frequency cutoff.
2. Finite number of discrete modes of propagation at a given frequency.
3. Phase velocity smaller than that of plane wave in vacuum.
3.7.1 Dielectric Layer above a Metallic Surface
Consider a TM mode and for the sake of simplicity, yet without significant loss of
generality, we shall examine the case when ∂ y = 0 , or more specifically TM 01 . This mode
may be derived from the longitudinal electric field
x
d εr
z
Basic configuration of the open dielectric waveguide.
166
x
d εr
z
⎧ ⎛ ω2 ⎞
⎪ A sin ⎜ x 2 ε r − k ⎟ 2
0< x<d
⎪⎪ ⎜ c ⎟
⎝ ⎠
Ez ( x, z ) = exp ( − jkz ) ⎨ (3.7.1)
⎪ ⎡ ω2 ⎤
⎪ Bexp ⎢ −( x − d ) k − c 2 ⎥
2
x>d
⎪⎩ ⎢⎣ ⎥⎦
G G
The x -component of the electric field may be derived using Gauss' law (∇ ⋅ E = 0)
⎧ − jkA ⎛ ω2 ⎞
⎪ cos ⎜ x 2 ε r − k ⎟ 0 < x < d
2
⎪ ω 2 ⎜ c ⎟
ε − k 2 ⎝ ⎠
⎪⎪ c 2 r
Ex ( x, z ) = exp ( − jkz ) ⎨ (3.7.2)
⎪ − jkB ⎡ ω ⎤2
⎪ exp ⎢ −( x − d ) k − 2 ⎥
2
x>d
⎪ k 2 −ω
2
⎢⎣ c ⎥⎦
⎪⎩ c2
G G G
for establishing the magnetic field we use Ampere's law ( ∇ × H = jωε 0 ε r E )
167
x
d εr
z
⎧ − jωε ε A ⎡ ω2 2 ⎤
⎪ 0 r
cos ⎢ x ε r 2 − k ⎥ 0< x<d
⎪ ω 2
⎢⎣ c ⎥⎦
⎪⎪ ε r − k 2
c2
H y ( x, z ) = exp ( − jkz ) ⎨
⎪ − jωε 0 B ⎡ ω2 ⎤
⎪ exp ⎢ −( x − d ) k − 2 ⎥ d < x < ∞.
2
⎪ k 2 −ω
2
⎢⎣ c ⎥⎦
⎪⎩ c2
(3.7.3)
Continuity of the tangential electric field implies
⎡ ω2 2 ⎤
Ez : A sin ⎢ d ε r 2 − k ⎥ = B (3.7.4)
⎣⎢ c ⎥⎦
whereas the continuity of the tangential magnetic field
εr ⎡ ω2 2 ⎤ 1
Hy : A cos ⎢ d ε r 2 − k ⎥ = B. (3.7.5)
ω 2
⎢⎣ c ⎥⎦ ω 2
εr 2 − k2 k2 − 2
c c
These two equations lead to the following dispersion relation
168
x
d εr
z
ω2
ξ tan(ξ ) = ε r d k −
2
, (3.7.6)
c2
ω2 ω
where ξ ≡ d ε r − k 2 . It is also convenient to define ξ0 = d ε r − 1 therefore, we
c c
may write
⎡ ω2 ω2 2 ⎤ 2 ω
2
ξ = d ⎢(ε r − 1) 2 + 2 − k ⎥ ⇒ d k − 2 = ξ 02 − ξ 2
2 2
(3.7.7)
⎣ c c ⎦ c
Consequently, the dispersion relation reads
ξ tan(ξ ) =ε r ξ02 − ξ 2 . x
d εr
(3.7.8) z
The two sides of the equation are plotted in the figure below
169
x
d εr
z
Clearly for a single mode operation, we require

ω π 1
that ξ0 < π ⇒ < . ξ tan (ξ )
c d εr −1
ε r ξ 03
ξ tan (ξ )
ε r ξ 02
We shall examine what happens at low
frequencies i.e. ξ → 0, ξ0 → 0 . At this limit the ε r ξ 01
dispersion relation reads
ξ 2 =ε r ξ02 − ξ 2 (3.7.9)
ξ 01 π ξ 02 π ξ 03 3π ξ
The solution of the fourth order polynomial which
corresponds to a low-frquency propagating wave 2 2
is
1⎡ 2 ξ02 ⎤
ξ = ⎢ −ε r + ε r 1 + 4 2 ⎥
2 2
(3.7.10)
2 ⎢⎣ ε r ⎥⎦
and now substituting the definitions of ξ and ξ0 we get
2
⎛ω ⎞
k = ε eff ⎜ ⎟
2
(3.7.11)
⎝c⎠
which is only weakly dependent on the thickness (provided d λ ) but is dependent on
170
x
d εr
z
the dielectric coefficient (but ε r > 1). To be more accurate
ε r2 ⎡ ⎤
2
ω2 4 ⎛ω ⎞
ε r 2 − k = 2 ⎢ −1 + 1 + 2 ⎜ d ⎟ ( ε r − 1) ⎥
2
c 2d ⎢ εr ⎝ c ⎠ ⎥⎦
⎣ (3.7.12)
2 ω2
k =
c2
where the effective dielectric coefficient is unity.
171
x
The figure below illustrates the dispersion relation of the system.
d εr
εr = 3 ω z
d εr
kd d = 0.5cm c
ω
d
c
ω
d
c
172
x
d εr
z
Our next step is to examine the energy distribution in the system. Denoting
ξ ≡ d ε r (ω / c ) − k 2 and χ ≡ d k 2 −(ω / c ) we calculate the average electric energy
2 2
per unit surface ( Δ y Δ z ) in the dielectric and the air

⎡ ( ) ⎤
2
ε ε d
2⎛ x ⎞ kd 2⎛ x ⎞
∫ dx ⎢ A sin ⎜ ξ ⎟ + A ξ⎟⎥
(d ) 2 2
WE = 0 r
2 cos ⎜
4 0
⎢⎣ ⎝d ⎠ ξ ⎝ d ⎠ ⎥⎦
(3.7.13)
A ⎧⎪ sin(2ξ ) ⎛ kd ⎞ ⎡ sin(2ξ ) ⎤ ⎫⎪
2 2
= ε 0 ε r d ⎨1 − + ⎜ ⎟ ⎢1 + ⎬
8 ⎪⎩ 2ξ ⎝ξ ⎠ ⎣ 2ξ ⎦⎥ ⎪⎭
ε0 ⎡ k 2d 2 ⎤
∞ ⎛ x−d ⎞
W = A sin (ξ )∫ dx ⎢1 + 2 ⎥ exp ⎜ −2
( air ) 2 2
χ⎟
E
4 d
⎣ χ ⎦ ⎝ d ⎠
2 ε0 ⎡ k 2d 2 ⎤ d
= A sin (ξ )⎢1 + 2 ⎥ .
2
(3.7.14)
4 ⎣ χ ⎦ 2χ
Similarly the average magnetic energy
173
μ0 d ω 2ε 02 ε r 2 d 2 2 2 ⎛ x ⎞ 1 2 ω ε0 εr d 1 ⎡
2 2 2 2
sin ( 2ξ ) ⎤
∫ ξ μ +
(d )
W = dx A cos ⎜ ⎟ = A d ⎢1 ⎥
M
4 0 ξ2 ⎝d ⎠ 4
0
ξ2 2 ⎣ 2ξ ⎦
2 ω ε0 d
2 2 2
1 d
W (a)
= μ0 A sin ξ
2
M
4 χ 2 2χ
(3.7.15)
1.0
Wd
The ratio of energy stored in the dielectric relative to 0.8 WT
the total energy is illustrated below as a function of εr = 3
d = 0.5 cm W m ,d
the frequency 0.6
WT
0.4
We ,d
0.2 WT
0
0 5 10 15 20
ΩK i , 0
174
Exercise 3.14: Show that for a dielectric layer (ε r ) of thickness 2d there are even TE
modes that satisfy
ξ tan ξ = ξ02 − ξ 2 (3.7.16)
and odd TE modes satisfying
ξ ctan ξ = − ξ02 − ξ 2 . (3.7.17)
What happens at low frequencies in each case? Calculate the energy ratio in each case.
Exercise 3.15: A metallic ``wire'' of radius Rint is coated with a dielectric layer (ε r ) the
external radius being Rext . Show that the TM on mode is described by
⎧ A K 0 ( Γr ) r > Rint
Ez = ⎨
⎩ B[ J 0 (Λr )Y0 (ΛRint ) − Y0 (Λr ) J 0 (ΛRint ) Rint < r < Rext
wherein Γ = k 2 − ω 2 / c 2 and Λ = εω 2 / c 2 − k 2 . Demonstrate that the dispersion
relation is
K1 ( ΓRext ) ε J1 ( ΛRext )Y0 ( ΛRint ) − Y1 ( ΛRext ) J 0 ( ΛRint )
=− .
ΓK 0 ( ΓRext ) Λ J 0 ( ΛRext )Y0 ( ΛRint ) − Y0 ( ΛRext ) J 0 ( ΛRint )
Analyze the mode.
175
3.7.2 Surface Waves along a Dielectric Fiber
R θ
k z
As a second example of a structure capable of supporting a surface wave, we

consider a dielectric rod of radius a , as in the figure above. Dielectric rods are similar in
behavior to dielectric sheets in that a number of surface-wave modes exist. Pure TM or
TE modes are possible only if the field is independent of the azimuthal coordinate φ . As
the radius of the rod increases, the number of TM and TE modes also increases. These
modes do, however, have a cutoff point such that, below some minimum value of R/λ0 ,
the mode cannot exist any longer. All modes with angular dependence are a combination
of a TM and a TE mode, and are classified as hybrid EH or HE modes, depending on
whether the TM or TE mode predominates, respectively. All these modes, with the
exception of the HE 11 mode, exhibit cutoff phenomena similar to that of the axially
symmetric modes. Since the HE 11 mode has no low-frequency cutoff, it is the dominant
mode. For small-diameter rods, the field extends for a considerable distance beyond the
176
surface, and the axial propagation constant β is only slightly larger than k0 . As the radius
increases, the field is confined closer and closer to the rod, and β approaches β = k0 ε r
in the limit of infinite radius. Since β > k0 , the phase velocity is less than that of plane
waves in free space. In all of the above respects, the dielectric rod does not differ from
the plane-dielectric sheet. Omitting the term exp ( − jnφ − j β z ) , the field expansion
components are
r<R r>R
Ez = An J n (Λr ) E z = Cn K n ( Γ r )
jβ nωμ jβ nωμ
Er = − An J n′ (Λr ) − 2 0 Bn J n (Λr ) Er = Cn K n′ (Γr ) + 2 0 Dn K n (Γr )
Λ Λr Γ Γr
nβ jωμ0 nβ jωμ0
Eφ = − 2 An J n (Λr ) + Bn J n′ (Λr ) Eφ = 2 Cn K n (Γr ) − Dn K n′ (Γr )
Λr Λ Γr Γ
H z = Bn J n (Λr ) H z = Dn K n (Γr )
nωε jβ nωε jβ
H r = 2 An J n (Λr ) − Bn J n′ (Λr ) H r = − 2 0 C n K n ( Γr ) + Dn K n′ (Γr )
Λr Λ Γr Γ
jωε nβ jωε 0 nβ
Hφ = − An J n′ (Λr ) − 2 Bn J n (Λr ) Hφ = Cn K n′ (Γr ) + 2 Dn K n (Γr )
Λ Λr Γ Γr
177
ω2 ω2
Λ =ε
2
− β and Γ = β − 2 ; An , Bn , Cn , Dn are normalized amplitude; and the
2 2 2
c2 c
prime indicates differentiation with respect to the arguments Λr or Γr .
Imposition of the boundary conditions at r = R leads to equations for determining
the relative amplitudes of the coefficients, and also the eigenvalue equation. The
eigenvalue equation is
2
⎡ ε J n′ (u1 ) K n′ (u2 ) ⎤ ⎡ J n′ (u1 ) K n′ (u2 ) ⎤ ⎡ nβ (u22 + u12 ) ⎤
⎢ − ⎥⎢ + ⎥=⎢ 2 2 ⎥ (3.7.18)
⎣ u1 J n (u1 ) u2 K n (u2 ) ⎦ ⎣ u1 J n (u1 ) u2 K n (u2 ) ⎦ ⎣ k0 u1 u2 ⎦
where u1 = ΛR , u2 = ΓR . When n = 0 , the right-hand side vanishes, and each factor on
the left-hand side must equal zero. These two factors give the eigenvalue equations for
the axially symmetric TM and TE modes:
ε J 0′ (u1 ) K ′ (u )
= 0 2 TM modes (3.7.19)
u1 J 0 (u1 ) u2 K 0 (u2 )
J 0′ (u1 ) K 0′ (u2 )
=− TE modes. (3.7.20)
u1 J 0 (u1 ) u2 K 0 (u2 )
In addition, u1 and u2 are related by the equation
2
⎛ω ⎞
u12 + u22 =( ε − 1)⎜ R ⎟ . (3.7.21)
⎝c ⎠
178
The ratio of β to k0 as a function of 2 R / λ0
for a polystyrene rod is given below for ε = 2.56
1.5
axially symmetric TM 1 and TE 1 modes and
the HE 11 dipole mode. Both the TM 1 and 1.4
TE 1 modes are cut off for 2a < 0.163λ0 ,
β / k0
1.3
while the HE 11 mode has no low-frequency HE11 TE 1 TM 1
cutoff. 1.2
1.1
0.613
Comments: 1.0
1. The problem of exciting surface waves 0.2 0.4 0.6 0.8 1.0 1.2
differs somewhat from that of exciting a 2 a / λ0
propagating mode in an ordinary closed-
boundary guide, in that some of the power radiated by the source goes into the
propagating field – not necessarily bounded to the dielectric.
2. The conclusion reached in (3.7.12) that there is smooth ``transition'' to ω → 0 is not
general for surface waves. In order to demonstrate this fact let us return to (3.7.19). It is
possible to write the dispersion relation in the following form
179
(
J1 ψ 02 − ψ 2 ) =
1 1 K1 (ψ )
. (3.7.22)
ψ −ψ J 0 ψ −ψ
2
0
2
( 2
0
2
1 ) ε r ψ K 0 (ψ )
ω
wherein ψ 0 = R ε − 1 and ψ = R k 2 − ω 2 / c 2 as a function of the frequency for
c
several phase velocities reads
1 1 1 K1 (ψ )
= . (3.7.23)
2 ε r ψ K 0 (ψ )
The solution denoted by ψ c (ε r ) implies
2⎛ 2 ω ⎞
2
ω 2 ψ c2 (ε r )
ψc = R ⎜k − 2 ⎟ k = 2 +
2 2
2
(3.7.24)
⎝ c ⎠ c R
and since for example for ε r = 3, the solution is ψ c = 0.963 , we conclude that there is no
solution which is consistent with the assumption ψ → 0 . Consequently, there is no
solution at low frequencies.
180
3.8 Transients in Waveguides
In any waveguide, the phase velocity v ph , is a function of frequency. As a

consequence, all signals having a finite frequency spectrum will undergo dispersion when
transmitted through a length of guide. The phase relationship between the frequency
components of the original signal at the feeding point continually changes as the signal
progresses along the guide. Any realistic analysis should take into account the frequency
characteristics of the antenna or aperture that couples the signal into the guide as well as
the characteristics of the circuit elements used to extract the signal at the receiving end. In
this section we shall consider only the properties of the guide itself. The effect of losses
and their variation will also be neglected. In practice, this does not lead to significant
errors, because in most cases the frequency bandwidth of the signal is relatively narrow,
and the mid-band frequency of operation is usually chosen far enough above the cutoff
frequency so that the attenuation curve is approximately constant throughout the band.
Before some time, which we choose as the time origin t = 0 , the disturbance in the
guide is zero. When a current element is introduced into the guide, a disturbance or signal
is generated. This signal is a solution of the time-dependent field equations. Let
E (r , t ), H(r , t ) be the time-dependent field vectors. For a unit impulse current element
δ (t − t ′) applied at time t = t ′ and located at the point ( x ′, y ′, z ′) , the field vectors are a
solution of
181
∂H
∇ × E = − μ0
∂t
∂E
∇ × H = ε0 + sδ (t − t ′)δ (r − r') (3.8.1)
∂t
where t is a unit vector giving the direction of the current element, and r designates the
field point ( x, y , z ) , while r' designates the source point or location. If the Laplace
transform of these equations is taken, we get
∇ × E(r, s ) = − μ0 sH (r, s )
′
∇ × H (r , s ) = ε 0 sE(r, s ) + se− st δ (r − r') (3.8.2)
where E(r ,s)= E (r ,t)e−st dt
∞
∫0
with a similar definition for H(r , s ) . These equations are formally the same as those
obtained by assuming a time dependence e jωt , and the solution may be obtained by the
methods we have previously discussed. All our previous solutions may be converted into
jωt ′
solutions of (3.8.2) by replacing jω by s and e by e− st . The Laplace transform has
the effect of suppressing the time variable. The solution to (3.8.2) constitutes the Laplace
transform of the time-dependent Green's function. Inverting the transform yields the
time-dependent Green's function. For an arbitrary spatial and time variation, the solution
may be obtained by a super position integral.
182
If we restrict ourselves to a line current extending across the narrow dimension of a
G
rectangular guide, the appropriate Green's function corresponding to E y ( r )
− μ0 ∞ s exp {− st ′ − [(nπ /a ) 2 + s 2 / c 2 ]1/2 | z − z′ |}
G (r | r', s ) = ∑ φn ( x)φn ( x′) 1/2
(3.8.3)
a n =1 ⎡⎣(nπ /a ) + s / c ⎤⎦
2 2 2
where φn ( x ) = sin( nπ x / a ) . The inversion of this expression gives us the time-dependent

Green's function corresponding to the disturbance set up in the waveguide by an impulse
line current located at x ′, z ′ . A typical term from (3.8.3) gives
1 { 1/2
exp s (t − t ′) − ⎡⎣(nπ /a ) 2 + s 2 / c 2 ⎤⎦ | z − z′ | } ds
2π j ∫
C
⎡⎣ (nπ /a ) 2 + s 2 / c 2 ⎤⎦
1/2
⎧ ⎧ nπ 2 2 1/2 ⎫ | z − z′ |
⎪ ⎨
cJ [ c (t − t ′) 2
− ( z − z ′) ] ⎬ 0 < t − t ′ <
=⎨ 0⎩ a ⎭ c (3.8.4)
⎪ 0 otherwise
⎩
where J 0 is the Bessel function of the first kind. At any given distance | z − z ′ | from the
source, the disturbance is zero until a time t = t ′+ | z − z′ | /c is reached when the presence
of the signal first becomes known to the observer at this position. No information reaches
the observer in a time interval less than the time required to propagate a disturbance with
183
the velocity of light. The velocity of light is, therefore, the wavefront velocity.
The time derivative of the above function is
−c 3 (nπ /a )(t − t ′) ⎧ nπ 2 2 1/2 ⎫
J 1⎨ [ c (t − t ′) 2
− ( z − z ′) ] ⎬ (3.8.5)
[c 2 (t − t ′) 2 − ( z − z′) 2 ]1/2 ⎩ a ⎭
where J1 is the Bessel function of the first kind and order 1. The solution for the time-
dependent Green's function which is equal to E y (r , t ) becomes
1 ∞ nπ J1 {(nπ /a )[c 2 (t − t ′) 2 − ( z − z′) 2 ]1/2 }
G (r | r', t − t ′) = ∑ φ ( x)φn ( x′) [c(t − t′)]
ε 0 a n =1 a n [c 2 (t − t ′) 2 − ( z − z′) 2 ]1/2
(3.8.6)
for t ′ < t < t ′+ | z − z′ | /c and zero otherwise.
184
3.9 Waveguide Based Cavities
3.9.1 Power and Energy Considerations

The goal of this sub-section is to determine a general relation between power, energy and
impedance at the input of an electromagnetic device. We shall consider the power in a
volume V and an envelope S ; through this envelope an average power P crosses
G G
1
2
{ G
P = Re v∫ E × H * ⋅ ds .} (3.9.1)
Thus based on the complex Poynting theorem
1 G G* G ⎡ 1 G G * 1 G G * ⎤ 1 G G*
2 ∫
v ( E × H ) ⋅ ds = −2 jω ∫ dV ⎢ B ⋅ H − E ⋅ D ⎥ − ∫ E ⋅ J dV .
V
⎣4 4 ⎦ 2 V
(3.9.2)
Assuming linear medium characterized by μ = μ ′ − j μ ′′ , ε = ε ′ − jε ′′ and σ we can

separate the real and imaginary part of this theorem
⎧1 G G * G⎫ ⎡ 1 ′′ 1 2⎤ 1
Re⎨ vv∫∫ × ⋅ − ω ∫V ⎣⎢ 4 μ + ε ′′ ⎥⎦ 2 ∫
− σ
2 2
E H ds ⎬ = 2 dV | H | | E | dV | E |
⎩2 ⎭ 4
⎧1 G G * G⎫ ⎡μ′ G 2 ε ′ G 2 ⎤
⎩2
∫ ∫ E × H ⋅ ds ⎬⎭ = −2ω∫V dV ⎢⎣ 4 | H | − 4 | E | ⎦⎥.
Im ⎨ vv (3.9.3)
185
Several comments are in place
1. The terms proportional to ωμ ′′ or ωε ′′ can be interpreted as loss.
2. The reactive energy flow into the volume V equals exactly 2ω the reactive energy
difference (WM − WE ) stored in the volume.
3. In terms of a lumped parameters ( RLC ): if the current at the entrance is denoted by I
and the voltage V , we may write the complex power at the input as
1 * 1 * 1 *⎡ 1 ⎤
VI = ZII = II ⎢ R + jω L + ⎥, (3.9.4)
2 2 2 ⎣ jωC ⎦
where PL = R | I |2 / 2 represents all losses in the system, WM = L | I |2 / 4 represents the
average magnetic energy stored in the system whereas WE = C | V |2 / 4 represents the
average electric energy stored in the system hence
1 * 1
VI = Z | I |2 = PL + 2 jω(WM − WE ) (3.9.5)
2 2
or
P + 2 jω[WM − WE ]
Z= L . (3.9.6)
1 2
|I|
2
This can be conceived as a generalization of the impedance concept.
186
3.9.2 Rectangular Waveguide Cavity
b
d
a x
z
Consider a rectangular waveguide shortened at z = 0 and z = d . Without shortening

plates the wavenumber is given by
2 1/2
⎡ ω2 ⎛ π n ⎞ ⎛ π m ⎞ ⎤
2
β nm = ⎢ε r μ r 2 − ⎜ ⎟ − ⎜ ⎟ ⎥ . (3.9.7)
⎣ c ⎝ a ⎠ ⎝ b ⎠ ⎦
Because of the shortening planes
πl
β nm = (3.9.8)
d
187
thus the resonance frequencies are given by
2 1/2
c 1 ⎡ ⎛πn ⎞ ⎛πm ⎞ ⎛πl ⎞ ⎤
2 2
f n ,m,l = ⎢⎜ ⎟ +⎜ ⎟ +⎜ ⎟ ⎥ . (3.9.9)
2π ε r μ r ⎣⎝ a ⎠ ⎝ b ⎠ ⎝ d ⎠ ⎦
We shall examine next electromagnetic field characteristics in a cavity. In particular we
shall calculate the quality factor of a TE101 . Choosing d > a > b , this is the lowest mode.
It has the following field (non-zero) components
− jβ z jβ z ⎤ ⎛πx⎞
H z = ⎡ Ae + Be cos ⎜ ⎟
⎣⎢ ⎦⎥ ⎝ a ⎠
jβ a ⎡ − jβ z jβ z ⎤ ⎛ π x ⎞
Hx = Ae − Be sin ⎜ ⎟ (3.9.10)
π ⎣ ⎢ ⎥
⎦ ⎝ a ⎠
jω a ⎛πx⎞
Ey = − ε r μ rη0 ⎡ Ae− j β z + Be j β z ⎤ sin ⎜ ⎟
cπ ⎢⎣ ⎥⎦ ⎝ a ⎠
Bearing in mind that E y vanishes at z = 0 we obtain A + B = 0 and applying the
− jβ d jβ d
same condition at z = d we get Ae − Be = 0 ⇒ − 2 jA sin( β d ) = 0 therefore
for a non trivial solution β has to satisfy β d = π l which is just (3.9.8).
188
Consequently, the mode TE101 is described by
⎛πx ⎞ ⎛πz ⎞
H z ( x, z;ω ) = −2 jA cos ⎜ ⎟ sin ⎜ ⎟
⎝ a ⎠ ⎝ d ⎠
a ⎛πx ⎞ ⎛πz ⎞
H x ( x, z; ω ) = 2 jA sin ⎜ ⎟ cos ⎜ ⎟ (3.9.11)
d ⎝ a ⎠ ⎝ d ⎠
1/2
a ⎡⎛ π a ⎞ π 2 ⎤
2
⎛πx⎞ ⎛πz ⎞
E y ( x, z;ω ) = −2 A η ⎢⎜ + ⎥ sin ⎟ sin ⎜ ⎟
π ⎢⎣⎝ d ⎟⎠ d 2 ⎥⎦ ⎜
⎝ a ⎠ ⎝ d ⎠
The electric and the magnetic energy stored in the system
2 ⎡π π 2 ⎤a d
2 2
1 a b d 1 2 a
WE = ε o ∫ dx ∫ dy ∫ dz| E y | = ε o | 2 A | 2 η ⎢ 2 + 2 ⎥
2
b = WM
4 0 0 0 4 π ⎣a d ⎦2 2
Clearly at resonance the electric energy equals the magnetic energy stored in the system
therefore the total average energy stored in the cavity is
1 ⎡ a2 ⎤ 2
WT = 2WE = μo (bad ) ⎢1 + 2 ⎥| A | . (3.9.12)
2 ⎣ d ⎦ G G G
Next we shall calculate the loss. The surface current is given by J s = n × H therefore the
dissipated power is given by
189
2
1 J2 1 ⎛ Js ⎞ 1 1 | H x |2 Rs
Ploss = ∫ daδ = ∫ daδ ⎜ ⎟ = ∫ da = ∫ da | H x |2
2 σ 2 ⎝δ ⎠ σ 2 δσ 2 (3.9.13)
=| A |2 Rs ( 2a 3b + 2d 3b + ad 3 + da 3 ) / d 2
Consequently, the quality factor
3/2
⎡(π / a )2 + (π / d )2 ⎤ (ad )3
ωWT 2ωWE η ⎣ ⎦
Q= = = . (3.9.14)
Ploss Ploss Rs 2π (2a b + 2d b + a d + d a )
2 3 3 3 3
As an example consider the following parameters

σ Cu ∼ 5.8 × 107 [ Ohm/m ] , a = b = d = 3[ cm ] → f = 7.0 [ GHz ] , T = 1/ f 0.14[ nsec ]
Rs ∼ 0.022Ω, Q ∼ 12700, τ = 2Q / ω0 = Q / π f 0.58[ msec].
In the case of dielectric losses
1 2ωWE ε ′ 1 1 1
Ploss,d = ωε ∫ dV | E | ⇒ Qd =
′′ 2
= ⇒ = + (3.9.15)
2 V Ploss ,d ε ′′ Qeff Qd Q
since the total power loss is the sum of the two mechanisms.
Exercise 3.18: Check whether (3.9.13) is correct. In particular check if H z does not
contribute to the relevant surface current. Calculate the impedance in (3.9.6).
190
3.9.3 Cylindrical Resonator 2R
In this subsection we briefly repeat the

previous exercise in cylindrical geometry.
Consider a cylindrical waveguide of radius R .
with two shorting planes at z = 0 and z = d . d
We shall examine in this case the TE11 mode
⎛ ′ r⎞
H z = J1 ⎜ p11 cos φ ⎡ Ae− j β z + Be j β z ⎤ (3.9.16)
⎟ ⎢
⎣ ⎥
⎦
⎝ R ⎠
wherein p11 ′ = 1.841. Determine all the other field components!! As in the previous
subsection
Er ( z = 0, d ) = 0 ⎫
⎬ ⇒ −2 jA sin( β d ) ⇒ β d = Aπ ⇒ β = π A/d . (3.9.17)
Eφ ( z = 0, d ) = 0 ⎭
thus
2 1/2
ω ⎛ π ⎞ ⎛ p11′ ⎞ ⎤
⎡ 2
= ⎢⎜ ⎟ + ⎜ ⎟ ⎥ . (3.9.18)
c ⎣⎝ d ⎠ ⎝ R ⎠ ⎦
191
Exercise 3.19: Prove that the quality factor for a TE mode is given by
3/2
⎡ ⎛ n ⎞2 ⎤ ⎡ ⎛ πl ⎞ ⎤
2
⎥ ⎢( pnm′ ) +⎜ R⎟ ⎥
2
⎢1 − ⎜ ⎟
1 ⎢⎣ ⎝ pnm ′ ⎠ ⎥ ⎢⎣ ⎝ d ⎠ ⎥⎦
(TE )
Qnml = ⎦ (3.9.19)
ω π
2
⎛ π ⎞
2
δ s ′ ) + 2 ⎛ R ⎞ + ⎛1 − 2 ⎞
R l R nl R
c ( pnm
2
⎜ ⎟ ⎜ ⎟⎜ ′ ⎟
d⎝d ⎠ ⎝ d ⎠ ⎝ pnm d⎠
For a TM mode the quality factor is given by
⎧⎡ 2 1/2
⎪ ⎢ pnm ⎛ πl ⎞ ⎤
2
+⎜ R⎟ ⎥
⎪ ⎢⎣ ⎝ d ⎠ ⎥⎦
⎪ l>0
(TM ) 1 ⎪ 1+ 2
R
Qnml = ⎨ (3.9.20)
ω0 ⎪ d
δs
c ⎪ pnm
l=0
⎪ R
⎪⎩ 1+
d
Exercise 3.20: Retrieve the result in (3.9.19)--(3.9.20). Compare to a rectangular cavity

of the same volume or same surface.
192
3.9.4 Open Resonator – Circular Geometry
In many cases the cavity relies on evanescent waves that decay in one direction and they
do not carry any power in that direction. One example is illustrated below: A cylindrical
waveguide of radius R with a short-circuiting plane at z=0, a dielectric ( ε r ) of thickness
d.
Exercise 3.21: Determine the resonant frequency and the

quality factor for a TMnm mode. Repeat the exercise for
TEnm mode. Consider both metallic and dielectric loss.
2R
εr
193
3.9.5 Exciting a Cavity
Coaxial Coaxial
line line
(a) (b)
There are two main methods to excite cavities: with coaxial
line (electric or magnetic coupling) or with a waveguide
Cavity
through an aperture.
In what follows, we shall consider a simple example Waveguide Aperture

of excitation of a cavity by a waveguide through a small
(circular) aperture of radius r0 as illustrated in lower frame.
(c)
Ignoring loss processes the input impedance is given by the
parallel impedance of the aperture
( j X L )( j tan β d )
Z in = . (3.9.21)
j X L + j tan( β d )
To include propagation loss (still assuming ideal aperture) the input impedance is
j X L tanh(α d + j β d )
Z in = .
j X L + tanh(α d + j β d )
y
(3.9.22)
2r0 b Zc =1 jX L
d z d
z=0 z=d
194
(a) (b)
8 ⎛ π r02 ⎞
In case of a TE 10 the normalized impedance of the aperture is X L = ⎜ ⎟ β r0
3π ⎝ ab ⎠
revealing that this is proportional to the areas ratio. For low loss
( j X L ) j tan( β L)
Z in (3.9.23)
j X L + j tan β A + tanh(α d )
and the condition for resonance
D (ω ) ≡ X L + tan β A = 0 (3.9.24)
implies that at resonance ( ω = ω0 )
2
XL
Z in (ω = ω0 ) . (3.9.25)
αd
In the vicinity of the resonance
2
X Z in (ω = ω0 )
Z in (ω )
L
(3.9.26)
∂D ⎛ j ⎞ ∂D
αd + j (ω − ω0 ) 1+ ⎜ ⎟ (ω − ω0 )
∂ω ω =ω ⎝ α d ⎠ ∂ω ω =ω
0 0
Since for an RLC circuit the impedance in the vicinity of the resonance is
Z in (ω ) Z in (ω = ω0 ) [1 + j (ω − ω0 ) / Δω ] we conclude by analogy that the bandwidth is
−1
Δω = α d / D′ (ω = ω0 ) implying that the quality factor Q = ω0 / Δω = ω0 D′ (ω = ω0 ) / α d .
195
Comment: It is interesting to note that D′ (ω = ω0 ) depends on the group velocity since
X L =Uβd
∂ ∂β ⎡ 1 ⎤
D ′ ( ω = ω0 ) = [U β d + tan( β d )] = d ⎢U + ⎥
∂ω ∂ω ⎣ cos β d ⎦ω =ω0
2
(3.9.27)
d d
= ⎡⎣U + 1 + tan 2 ( β d ) ⎤⎦ = [U + 1 + (U β d ) 2 ]
Vgr ω =ω0 Vgr
8 ⎛ π r02 ⎞ r0
where U = ⎜ ⎟ implying that the quality factor is inversely proportional to the
3π ⎝ ab ⎠ d
group velocity in the waveguide consisting the cavity Q ∝ D′ (ω = ω0 ) ∝ 1/ Vgr .
196
3.10 Wedge in a Waveguide
3.10.1 Metallic Wedge

The goal of this section is to analyze a wave guide of a circular cross-section but which
lacks a slice as illustrated below
2R
α
x
In order to keep the problem relatively simple, we shall limit the discussion to the
evaluation of the cut-off frequency. For this purpose we recall that in Section 2.1 we
197
2R
found that
G jkz jωμ G
E⊥ = − 2 ∇ ⊥ Ez + 2 1z × ∇ ⊥ H z
k⊥ k⊥
G jkz jωε G
H ⊥ = − 2 ∇ ⊥ H z − 2 1z × ∇ ⊥ Ez (3.10.1)
k⊥ k⊥
k⊥2 = ω 2εμ − k 2
And both longitudinal components satisfy the wave equation. Limitint the analysis to the
cut-off frequencies k z = 0 these equations simplify
⎧ 1 1 ∂
⎪ E = Hz
G G jωε r ∂φ
r
1 ⎪
E⊥ = − 1z × ∇ ⊥ H z ⇒ ⎨
jωε ⎪E = − 1 ∂ H
⎪⎩ φ jωε ∂r
z
(3.10.2)
⎧ −1 1 ∂
⎪ Hr = Ez
G 1 G ⎪ jωμ r ∂φ
H⊥ = + 1z × ∇ ⊥ Ez ⇒ ⎨
jωμ ⎪ H = 1 ∂ E
⎪⎩ φ jωμ ∂r z
and the wave equation simplifies
198
2R
⎛ 1 ∂ ∂ 1 ∂2 ⎞⎛ Ez ⎞
⎜ r + 2 2 + ω με ⎟⎜ ⎟ = 0
2
(3.10.3)
⎝ r ∂r ∂r r ∂ φ ⎠⎝ z ⎠
H
Obviously, the solution of this set of equations describe a resonator
whereby the field does not vary in the z-direction.
For the TM modes the solution is of the form

Ez ( r ,φ ) ∝ Jν ( qr ) ⎡⎣ A cos (νφ ) + B sin (νφ ) ⎤⎦ (3.10.4)
subject to the boundary conditions
Ez ( r = R,α / 2 < φ < 2π − α / 2 ) = 0
Ez ( r ,φ = α / 2 ) = 0 (3.10.5)
Ez ( r ,φ = 2π − α / 2 ) = 0
The first condition implies that
Jν ( qR ) = 0 (3.10.6)
wherein q is yet to be determined and from the other two conditions we will specify ν :
Ez ( r ,φ = α / 2 ) = 0 ⇒ A cos (να / 2 ) + B sin (να / 2 ) = 0
(3.10.7)
Ez ( r ,φ = 2π − α / 2 ) = 0 ⇒ A cos ⎡⎣ν ( 2π − α / 2 ) ⎤⎦ + B sin ⎡⎣ν ( 2π − α / 2 ) ⎤⎦ = 0
or explicitly
199
2R
⎛ cos (να / 2 ) sin (να / 2 ) ⎞⎛ A⎞

⎜⎜ ⎟⎟ ⎜ ⎟ = 0 (3.10.8)
⎝ cos ⎣ (
⎡ν 2π − α / 2 )⎦
⎤ sin ⎣ (
⎡ν 2 π − α / 2 )⎦ ⎠ ⎝ B ⎠
⎤
For a non-trivial solution the determinant of the matrix is zero
πn 1
ν= =n (3.10.9)
2π − α 2 −α /π
wherein n = 1, 2,3....∞ . Now that we have determined ν we may proceed and establish
the zeros of the Bessel function:
J 1 ( p) = 0 (3.10.10)
n
2 −α / π
As an example let us consider three Bessel
functions in one case ν = 1 and for 1
comparison we illustrate below α = π / 5
thus ν ( n = 1) = 5 / 9 and ν ( n = 2 ) = 10 / 9 .
Obviously the Bessel function of a non- Jn( 1 , x) 0.5
integer order has also zeros: Jn( ν , x)
pn , s s =1 s = 2 s = 3 Jn( 2⋅ ν , x)
ν =1 3.832 7.016 10.173 0
ν (n = 1) = 5 / 9 3.220 6.366 9.509

ν (n = 2) = 10 / 9 3.981 7.175 10.337
0.5
0 5 10
x
200
2R
Explicitly
∞
⎛ r ⎞ ⎛ φ −α / 2 ⎞
Ez ( r ,φ ) = ∑U
n , s =1
n,s J
n
1 ⎜ n , s ⎟ sin ⎜ π n
⎝
p
R⎠ ⎝
⎟
2π − α ⎠
2 −α / π
pn , s : J
n
1 ( p) = 0
2 −α / π
(3.10.11) Ez
and the cut-off frequencies are determined by

c
ωco = pn , s (3.10.12)
R
The top frame in the right reveals the contours of constant Ez for
s = 1and n = 1. The central frame illustrates the case s = 1and n = 2
whereas the lower frame corresponds to s = 2and n = 1.
Ez2
201
Ez3
A similar approach may be followed for the TE modes
H z ( r ,φ ) ∝ Jν ( qr ) ⎡⎣ A cos (νφ ) + B sin (νφ ) ⎤⎦ (3.10.13)
subject to the boundary conditions
Eφ ( r = R,α / 2 < φ < 2π − α / 2 ) = 0
Er ( r ,φ = α / 2 ) = 0 (3.10.14)
Er ( r ,φ = 2π − α / 2 ) = 0
The first condition [Eq. (3.10.2)] implies that
⎡d ⎤
⎢⎣ dr J ν ( qr ) ⎥⎦ = 0 (3.10.15)
r=R
wherein q is yet to be determined and from the other two conditions we specify ν :
Er ( r ,φ = α / 2 ) = 0 ⇒ − A sin (να / 2 ) + B cos (να / 2 ) = 0
(3.10.16)
Er ( r ,φ = 2π − α / 2 ) = 0 ⇒ − A sin ⎡⎣ν ( 2π − α / 2 ) ⎤⎦ + B cos ⎡⎣ν ( 2π − α / 2 ) ⎤⎦ = 0
or explicitly
⎛ − sin (να / 2 ) cos (να / 2 ) ⎞⎛ A⎞
⎜⎜ ⎟⎟ ⎜ ⎟ = 0 (3.10.17)
⎝ − sin ⎡
⎣ν ( 2π − α / 2 ) ⎤
⎦ cos ⎡
⎣ ν ( 2π − α / 2 ) ⎤
⎦ ⎠⎝ ⎠B
For a non-trivial solution the determinant of the matrix is zero
202
πn 1
=n ν= (3.10.18)
2π − α 2 −α /π
wherein n = 1, 2,3....∞ . Now that we have determined ν we may proceed and establish
the zeros of the Bessel function:
J′
n
1 ( p) = 0 (3.10.19)
2 −α / π
From here the approach is virtually identical
∞
⎛ ′ r⎞ ⎛ φ −α / 2 ⎞
H z ( r ,φ ) = ∑ U n,s J
n , s =1
n
1 ⎜ n,s ⎟
⎝
p
R ⎠
cos ⎜π n
⎝
⎟
2π − α ⎠
2 −α / π (3.10.20)
pn′ , s : J ′
n
1 ( p) = 0
2 −α / π
Exercise 3.22: Draw the contours of constant H z . Compare the field distribution of the
TE and TM modes. For both modes compare the magnetic and electric energy per-unit
length.
203
3.10.2 Dielectric Wedge y
The goal of this section is to analyze a waveguide of a circular
cross-section but which has a material ( μ r , ε r ) slice as illustrated
2R
below μr = 1
εr = 1
As in the metallic case we keep the problem relatively simple, and α
limit the discussion to the evaluation of the cut-off frequency. As a x
result, we can separate the analysis of the TE and TM modes. So ( μr , ε r )
let us start with the TM mode.
Vacuum Slice
⎛ 1 ∂ ∂ 1 ∂2 ω 2 ⎞ ⎛ 1 ∂ ∂ 1 ∂2 ω 2 ⎞
⎜ r + 2 2 + 2 ⎟ Ez = 0 ⎜ r + + μ ε
r r ⎟ Ez = 0
⎝ r ∂r ∂r r ∂φ c ⎠ ⎝ r ∂r ∂r r 2
∂φ 2
c 2
⎠
−1 1 ∂ −1 1 ∂
Hr = Ez Hr = Ez
jωμ0 r ∂φ jωμ0 μr r ∂φ
1 ∂ 1 ∂
Hφ = Ez Hφ = Ez
jωμ0 ∂r jωμ0 μr ∂r
A differential approach (separation of variables) is not applicable since the radial
204
variation needs to be the same in both regions but if this is the case the function does not
satisfy the wave equation. We must seek an integral approach. For determining the cut-
off frequencies associated with the TM-like modes we need to solve
⎛ 1 ∂ ∂ 1 ∂2 ω 2 ⎞ ω2
⎜ r + 2 2 + 2 ⎟ E z = − 2 f (φ ) E z (3.10.21)
⎝ r ∂r ∂r r ∂φ c ⎠ c
wherein f (φ ) is a function which equals μ r ε r − 1 in the region(s) with material and zero
otherwise. A solution which satisfies the boundary conditions
∞ ∞
⎛ r⎞
Ez ( r ,φ ) = ∑ ∑ En , s J n ⎜ pn , s ⎟ exp ( jnφ ) (3.10.22)
n =−∞ s =1 ⎝ R ⎠
substituting in (3.10.21) we get
⎛ pn2, s ω 2 ⎞ ω2 ∞ ∞
En , s χ n ,n ⎜ − 2 + 2 ⎟ = − 2 ∑ ∑ Em ,σ χ mσ ,,ns Fn ',m
s,s
⎝ R c ⎠ c m =−∞ σ =1
1 1
χ s,s '
n,n ' ≡ ∫ dx xJ n ' ( pn ', s ' x ) J n ( pn , s x ) , χ s,s '
n,n = δ s , s ' ∫ dx xJ n ( pn , s x ) J n ( pn , s x ) (3.10.23)
0 0
π
1
Fn ',m =
2π ∫π dφ f (φ ) exp ⎡⎣− j ( n '− m )φ ⎤⎦
−
205
Rewriting
2
⎛ω ⎞ ∞ ∞
⎜ R ⎟ ∑ ∑
⎝ c ⎠ m =−∞ σ =1
E χ σ ,s
⎡ F
m ,σ m , n ⎣ n , m + δ δ ⎤
n ,m σ , s ⎦ = E χ s,s 2
n , s n , n pn , s (3.10.24)
and defining the matrices

χ mσ ,,ns
C{n , s},{m ,σ } = ⎡⎣ Fn ,m + δ n ,mδσ , s ⎤⎦ (3.10.25)
χ p
s ,s
n ,n
2
n,s
thus defining ω = ω R / c .
∑σ C{ E
n , s},{m ,σ } {m ,σ }
= ω −2
∑σ I{ E{m ,σ }
n , s},{m ,σ }
(3.10.26)
{m , } {m , }
implying that the (square of the ) cut-off frequencies are the inverse of the eigen-values
of the matrix C
C − ω −2I = 0 . (3.10.27)
Exercise 3.23: Calculate the first cut-off frequency of a waveguide with dielectric slice
identical with the metallic slice. Compare the two results. Examine the convergence of
the solution.
Exercise 3.24: Repeat Exercise 3.23 for the TE mode including the formulation.
206
y
b 2R
3.11 Appendix
d x
a
Analyze (numerically) the dependence of the normalized reactance on three
parameters; the angular frequency (ω ) , its location ( d ) and its radius ( R ) .
0.6
The normalized reactance of a cylindrical post located
in a rectangular waveguide as a function of f . The
normalized reactance given by Eq. (3.4.20) is shown to
0.5 increase with frequency, which implies an inductive behavior.
[ a = 2.5 cm, b = 1 cm, ε r = 1 , d = a / 2 = 1.25 mm,
XL
R = a / 20 , n = 2,3," 200 ].
0.4
0.3
8 9 10
y
f [GHz ]
b 2R
d x
a
207
6
4 10 [GHz] The normalized reactance of a cylindrical post

located in a rectangular waveguide as a function of
XL
d . The solid line and the dotted line correspond to

f = 8 GHz and f = 10 GHz, respectively. A
2 8[GHz ] minima in the normallized reactance is observed at

the middle of the rectangular waveguide, whereas
close to the edges the normalized reactance
increases. [ a = 2.5 cm, b = 1 cm, ε r = 1 , R = a / 20 ,
n = 2,3," 200 ].
0
5 10 15 20
d [mm]
208
y
b 2R
d x
a
The normalized reactance of a cylindrical post

located in a rectangular waveguide as a function of
R . The solid line and the dotted line correspond to
f = 8 GHz and f = 10 GHz. In this figure the post
is assumed to be located at the middle of the
waveguide. [ a = 2.5 cm, b = 1 cm, ε r = 1 ,
d = a / 2 = 125 mm, n = 2,3," 200 ].
209
y
b 2R
d x
a
1.5
1.0 10[GHz ]
XL
The normalized reactance of a cylindrical

post located in a rectangular waveguide as a
0.5 function of R . The solid line and the dotted line
correspond to f = 8 GHz and f = 10 GHz, In
8[GHz ] this figure the post is assumed to be located close
to one of the edges of the waveguide. [ a = 2.5
0.0 cm, b = 1 cm, ε r = 1 , d = 5 mm, n = 2,3," 200 ].
0.5 1.0 1.5
R[mm]
210
Chapter 4: Matrix Formulations
4.1 Impedance Matrix

Consider an N port systems that is characterized at a given frequency by set of complex
numbers Zij ∼ i.e. 2 N 2 parameters
 V1   Z11 Z1N   I1 
V     
  = 
2    (4.1.1)
     
     
V
 N Z
 N1 Z NN   IN 
Vi
or, explicitly, Z ij = .
Ij
Ik ≠ j =0
(4.1.2)
V1
I1
V2 I2
IN
211 VN
In a similar way we can define the admittance matrix
 I1   Y11 Y1N   V1 
I  Y Y  V 
 2  21 2N   2
  =    . (4.1.3)
     
     
I  Y   
 N  N 1 YNN   VN 
Example: For the sake of simplicity, let us determine the Z matrix of a section of
transmission line.
This two port system is characterized by
 V1   Z11 Z12   I1 
V  = Z  I  (4.1.4)
 2  21 Z 22   2
and for example the first term is
V
Z11 = 1 (4.1.5)
I1 I =0
2
wherein I 2 = − I ( z = ) . Without loss of
I1 I2
generality, the voltage and current wave are
given by
V1 Z0 V2

212
Z =0 Z =
− jβ z jβ z
V ( z ) = V+ e + V− e
− jβ z jβ z  1
I ( z ) = V+ e − V− e (4.1.6)
  Z
o
respectively. At the `ìnput'', i.e. at z = 0
1
V1 = V+ + V− I1 = (V+ − V− ). (4.1.7)
Zo
Based on (4.1.5) we explicitly have
V +V
Z11 = Z o + − (4.1.8)
V+ − V− I ( )=0
wherein the condition I ( ) = 0 implies
1  − jβ jβ  −2 jβ
I () = V+ e − V− e = 0 V− = V+ e (4.1.9)
Z o  
hence
−2 jβ
1+ e cos( β )
Z11 = Z o = Zo = − jZ o ctan( β ) (4.1.10)
−2 jβ j sin β
1− e
Z11 = − jZ o ctan( β ). (4.1.11)
213
In a similar way we may establish the other parameters of the matrix. For example,
V V+ + V−
Z12 = 1 = Z0 (4.1.12)
I 2 I =0 ( −) V+ e
− j β
− V− e
j β 
1  
1
and since I1 = [V+ − V− ] V+ = V− we get
Zo
2 − jZ 0
Z12 = − Z 0 = (4.1.13)
− jβ j β sin( β )
e −e
In a similar way Z 21 = Z12 and Z 22 = Z11 implying
 − jZ o 
 − jZ ctan(θ ) 
sin θ
o
Z= . (4.1.14)
 − jZ o
− jZo ctan(θ ) 
 sin θ 
214
V1+
V1−
4.2 Scattering Matrix
In many cases it is more convenient to measure the V N+

amplitudes of the reflected and transmitted waves V2+
V N−
V2−
 v1−   v1+ 
   S11 S1N   
 v 2−     v 2+ 
     
  =     (4.2.1)
     
 
 − S   +
 vN   N 1 S NN   vN 
   
+ v1+ − v1−
1v ≡ v1 ≡ (4.2.2)
Zo Zo
the reason for this normalization becomes evident since
1 + 2 V + (V + )* 1 + + *
P= V = = v (v ) . (4.2.3)
2Zo 2Zo 2
215
The S matrix elements are defined by
−
vi
Sij = +
. (4.2.4)
+
v v
j
k ≠ j =0
+
In order to achieve v = 0 the port is loaded with a
k
matched impedance, namely
the diagonal term Sii represents the reflection v k+ = 0
coefficient at the i th port when all the other ports are
matched.
Z0 Z0
Describing linear systems, both S and Z matrices
are interrelated. Firstly it is convenient to define
Zij v v k−
Z ij = , v= and I = I Z o consequently
Zo Zo
+ − + −
v = v + v and I = v − v .
Using now the definition of the Z matrix
+ − + −
v = v + v = Z v − v  (4.2.5)
 
216
+ −
[ I − Z ]v = −[ I + Z ]v (4.2.6)
− +
thus since v = Sv
S = ( Z + I )−1 ( Z − I ). (4.2.7)
Properties of the S matrix:
Property 1: If the system is reciprocal S is symmetrical

Sij = S ji or S = St. (4.2.8)
1
This is true provided that at all ports the power is given by | v n |2 i.e. characterization
2
impedance at all ports is the same. In order to prove this statement we recall that based on
Lorentz reciprocity theorem (in absence of sources)

∫  1 2 2 1 

s
E × H − E × H  ⋅ nds = 0 ⇒ ∫ i j ∫ j i
E × H
i
da = E × H
j
da (4.2.9)
and since the electric field is associated with the voltage whereas the local magnetic
field is linked to the current E1 ∼ V1 , H 2 ∼ I 2 thus
vi I j = v j I j . (4.2.10)
217
Bearing in mind that Ii = Σ jYij v j and I j = ΣiY ji vi we conclude that
vi v jYij = vi v jY ji ⇒ Yij = Y ji ⇒ Z ij = Z ji (4.2.11)
Namely both the admitance and impedance matrices are symmetric aobviously since
S = ( Z + I ) −1 ( Z − 1) if Z is symmetrical so is S
Sij = S ji . (4.2.12)
Property 2. For a lossless system, S is unitary namely

S −1 = ( S * )T ( S * )+ S = I . (4.2.13)
For prooving this statement we bear in mind that power conservation implies
T
+
*
       
−
T

v  v  − v  v = 0 ⇒ v
+ − *
  + T
  + *

  I v  − Sv { }{ }
+ T + *
Sv = 0
(4.2.14)
{ }
+ T + * + T + * + T + *
v  I v  − v  S t S * v  = 0 ⇒ v  I − S S v  = 0
t *
           
Example: Consider a two ports system with a load Z L at the output
218
v1−   S S12  v1 
+
  =  11    (4.2.15)
v −2   S 21 S 22  v +2 
 
whereas the load
+
Z − Zo v 2
SL = ρ L = L = −. (4.2.16)
Z L + Zo v 2
Combining the two entails
− + + + −
v1 = S11 v1 + S12 v 2 = S11 v1 + S12 S L v 2
− + − − +
v 2 = S21 v1 + S22 S L v 2 → v 2 (1 − S22 S L ) = S21 v1 (4.2.17)
Thus the reflection coefficient
v1− S S S
ρin = Sin = + = S11 + 12 21 L . (4.2.18)
v1 1 − S L S22
+ +
v1 v2
ZL
− −
v1 219 v2
4.3 Transmission Matrix
 v1   A B   v2 
 I  =  C D  I  (4.3.1)
 1   2 
v1 = Z11I1 − Z12 I 2
(4.3.2)
v2 = Z 21I1 − Z 22 I 2
Thus
 Z11 Z11Z 22 − Z122 
  v
 v1   Z12 Z12  1 
I  =   I1 
(4.3.3)
 1 1 Z 22
 
 Z12 Z12 
Det{ ABCD} ≡ 1 (4.3.4)
I1 I2
v1 v2
220
4.4 Wave-Amplitude Transmission Matrix
 v1−   T T   v 2+ 
  =  11 12    (4.4.1)
 v1+   T21 T22   v 2− 
   
Relation to S matrix
S S S S 1
T11 = S12 − 11 22 , T12 = 11 , T21 = − 22 , T22 = (4.4.2)
S 21 S 21 S 21 S 21
In a reciprocal system S is symmetrical and S12 = S21 . Therefore
S
Det(T ) = 12 = 1 (4.4.3)
S21
+ +
v1 v2
− 221
−
v1 v2
Example: Scattering and transmission matrices of a section of transmission line
 v1−   S11 S12   v1+ 

 −=   (4.4.4)
 v2   S 21 S22   v2+ 
v1− v2− v1− − jβ v2− − jβ

S11 = + = 0, S22 = + = 0, S12 = + =e , S21 = + =e
v1 v+ =0 v2 v+ =0 v2 v+ =0 v1 v+ =0
2 1 1 2
(4.4.5)
 0 − jβ 

e
  e− jβ 0 
S =  T =  (4.4.6)
 − jβ   0 − jβ 
 e 
e 0 
+ +
v1 v2
Z0
− 222 −
v1 v2
4.5 Loss Matrix
Based on the result in Eq.(4.2.14)

Q = I − ( S * )T S (4.5.1)
and since the dissipated power is
1   + T +
Pdis =  v
2  
{ I − ( S * )T S } v 

(4.5.2)
the matrix defined above determines this dissipated power

1   + T +
Pdis =  v Qv  . (4.5.3)
2   
223
4.6 Directional Coupler
• 4 ports, lossless and reciprocal input direct

• At least one symmetry plane
• Each port is matched when all
1 3
the
other are loaded with Z o
2 4
• The power entering (1) splits
equally between (3) and (4). insulated coupled
No power reaches (2).
Performance parameters:
Coupling : C = 10log ( P1 / P4 )
Directivity : D = 10log ( P4 / P2 )
S matrix:
Isolation condition : S12 = S 21 = S34 = S43 = 0
Matching condition : S11 = S 22 = S33 = S 44 = 0
Symmetrical condition : S13 = S31 , S23 = S32 , S14 = S 41 , S24 = S42
 0 0 S13 S14 
 0 0 S 23 S 24 
S =
 S13 S 23 0 0 224
 
 S14 S 24 0 0 
4.7 Coupling Concepts
We shall discuss several aspects of the coupling effect. Here we shall investigate
coupling in transmission line configuration.
4.7.1 Double Strip Line: Parameters of the Line
x=a
(1) (2) x
δ
a/2 z
x=0
z=0
The basic configuration under consideration is illustrated above. Each one of the wires is
225
charged thus the charge distribution in the system is assumed to be
 a  ∆  a  ∆
ρ1 ( x, z ) = λ1δ  x −  δ  z −  , ρ 2 ( x, z ) = λ2δ  x −  δ  z + 
 2  2  2  2
For the electrostatic potential we may write the following expression
∞
 π nx 
Φ( x, z ) = ∑ϕ n ( z )sin 
n =1  a 
Which satifies the boundary conditions on the two plates. This potential satisfies the
Poisson equation i.e.,
 ∂2 ∂2  1
 2 + 2 
ϕ ( x , z ) = − ρ ( x, z ) ⇒
 ∂x ∂z  ε oε
 d 2  π n 2  1 2  π nx 
 2  −

  n

ϕ ( z ) =
ε ε ∫ dx ρ ( x , z )sin 



(4.7.1)
 dz a  o a a
1 2  ∆ν   π n 
=− ∑ λ
ε oε a ν ν 
δ z − sin
2   2 

where ν = 1,2 and ∆1 = ∆, ∆ 2 = −∆ .
1 ∞  π nx   π n   n ∆  n ∆ 
ϕ ( x, z ) = ∑ sin sinc λ
 1 exp  −π z + + λ exp −π z −   (4.7.2)
2  
2
2 n=1  a   2   a  a 2 
226
For a strip of width δ we have
 − ∆2 + δ2 
1  πn  
δ ∫ d ξ λ1 (ξ )exp −
 a z − ξ  
∆ δ
1 ∞  π nx   π   − 2 − 2 
ϕ ( x, z ) = ∑ sin sinc n   ∆ δ  (4.7.3)
2 n=1  a   2  +

+ 1 2 2
 π n 
∫ dξ λ2 (ξ )exp  − z − ξ 
 δ ∆ δ  a 
 −
2 2 
Assuming that λ1 (ξ ) and λ2 (ξ ) are uniform
1 ∞  π nx  π 
ϕ ( x, z ) = ∑ sin sinc  n
2 n=1  a  2 
 − ∆2 + δ2 ∆ δ
+  (4.7.4)
λ λ 2 2
  πn 
×  1 ∫ dξ + 2 ∫ dξ  exp  − z −ξ 
 δ − ∆ −δ δ ∆ δ 
−
 a 
 2 2 2 2 
Moreover, the potential is uniform on the strip
227
∆ δ
− +
2 2
1 a
V1 = ∫
δ ∆ δ
dzϕ ( x =
2
, z)
− −
2 2
∆ δ ∆ δ
− + − +
λ ∞
πn  π 
2 2
1 2 2
 πn 
= 1
2δ
∑
n =1
sin  
 2 
sinc n  ∫ dz ∫
 2  − ∆ −δ δ − ∆ −δ
d ξ exp −
 a z − ξ  (4.7.5)
2 2 2 2
∆ δ ∆ δ
− + +
λ ∞
πn  π 
2 2
1 2 2
 πn 
+ 2
2δ
∑ sin  
 2 
sinc n  ∫ dz
 2  − ∆ −δ δ
∫ d ξ exp −
 a z − ξ 
n =1 ∆ δ
−
2 2 2 2
The integrals may be evaluated analytically therefore
2
1a  ∞
πn     π nδ   π nδ 
U11 = ∑
2 δ n=1 
sinc   
 2  
1 − exp  −
 2a 
 sinhc 
 2a


2
(4.7.6)
1a ∞   π n    π nδ  2  π nδ    n 
U12 = ∑ 
2 δ n =1 
sinc    
 2    2a 
 sinhc  
 2a  
exp 

−π ∆
a 
Similar considerations can be applied to the second strip hence V2 = U 21λ1 + U 22λ2
and detailed calculation shows that U 21 = U12 and U 22 = U11 . Note that the off
228
diagonal term is proportional to
 ∆
∑ exp  −π n a  ( )n (4.7.7)
which indicates that the coupling drops rapidly. Furthermore, U11 are ∆ independent!
If the strips have different geometries U11 =/ U 22 but still U12 = U 21 . In a matrix form:
 V1   U11 U12  λ1   λ1   C11 C12  V1 
 =   ⇒   =    (4.7.8)
         
V  U       
 2   21 U 22  λ2   λ2   C12 C22  V2 
A similar approach can be applied for evaluation of the inductance
 Φ1   L11 L12  I1 
 =   (4.7.9)
    
Φ   L  
 2   12 L22  I 2 
Exercise 4.1: Analyze the coupling term.
Exercise 4.2: Calculate the inductance matrix and analyze the coupling term.
Exercise 4.3: Analyze the case when the two strips are not equally wide.
Exercise 4.4: Analyze the case when the two strips are not equally wide and are not in
the same plane.
229
4.7.2 Equations of the System
The transmission line equations in this case read

 dV1 
 dz   L11 L12   I1 
d     
V = − jω LI   = − j ω
dz   
 dV2  L  
 12 L22   I 2 
 
 dz 
(4.7.10)
 dI1 
 dz   C11 C12   V1 
d     
I = − jωCV   = − jω
dz   
 dI 2  C  
 12 C22  V2 
 
 dz 
It is convenient in this case to investigate the waves which propagate in this system
d2 dI
2
V = − jω L = − jω L ( − jω )CV = −ω 2
LCV (4.7.11)
dz dz
Let us define the wavenumber matrix
230
 K112 K122 
 
K ≡ ω 2 LC = 
2
 (4.7.12)
 K12 K 22 
2 2
 
− jβ z
Assume a solution of the form e
 − β 2 0   K112 K122    V1 
     
(− β 2 I + K )V = 0 ⇒  +
2
   = 0 (4.7.13)
 0
 − β 2  K122
  K 222    V2 
Explicitly this reads
 K112 − β 2 K122   V1 
  
   = 0 ⇒ ( K 2
11 − β 2
)( K 2
22 − β 2
) − K 4
12 =0 (4.7.14)
 K122
 K 222 − β 2   V2 
Let us assume for simplicity sake that the two lines are identical i.e. K11 = K 22 hence
( K112 − β 2 ) 2 − K124 = 0 β ±2 = K121 ± K122 (4.7.15)
The eigen-vectors corresponding to β +2 may be derived from one of the two equations
231
K112 − β +2
eg., ( K − β )V1 + K V = 0 threfore choosing V1 = 1 we get V2 = −
2
11
2
+
2
12 2 2
= 1 implying
K12
that the normalized egen-vector is
1 / 2 
 
V+ =  (4.7.16)
1 / 2 
 
And in a similar way, the normalized eigen-vector corresponding to β −2 is
 1/ 2 
 
V− =  (4.7.17)
 −1/ 2 
 
The general solution has the form
− jβ + z + j β+ z  − j β− z + j β− z 
V ( z ) = V +  Ae + Be + V − Ce + De (4.7.18)
   
232
1
4.7.3 Simple Coupling Process 2

L
In this subsection we ignore reflections. The schematics of the system under

consideration is illustrated in the right. Consider a solution of the form
− jβ+ z − j β− z
V ( z ) = V + Ae + V − Be (4.7.19)
At z = 0 only the mode in the first line exists. The question is how the power varies with
the length L .
1/ 2   1/ 2 
  − j β+ z   − j β− z
V ( z) =   Ae +   Be
1/ 2   −1/ 2 
   
 1/ 2 Ae− j β+ z + 1/ 2 Be− j β− z 
( ) ( )
(4.7.20)
 
= 
 
 1/ 2 Ae− j β+ z − 1/ 2 Be− j β− z 
(  ) ( ) 
In the absence of coupling β + = β − = K11
233
1
2
L
1 0
− jK11z − jK11z
V ( z) = C   e + D e (4.7.21)
   
   
0 1
Now since V ( z = 0) = V0 for line 1 and V ( z = 0) = 0 for line 2 we get C = V0 and D = 0
implying that the voltage on line 2 is always zero.
Let us reinstate the coupling process: the voltage along the transmission lines is
given by
 V0   A/ 2 + B/ 2 
 
V ( z = 0) =   = 
V0
 ⇒ A = B = (4.7.22)
0   2
   A / 2 − B / 2 
hence
  1  1 
e − j β z − j β − 
z V
 o exp − j
 2 +( β + β − ) z cos
  2 +( β − β − ) z  
+
+e
Vo    
V ( z) =  = 
2   
 e− j β+ z − e− j β− z    1  1 
   Vo exp  − j ( β + + β − ) z  j sin  ( β + − β − ) z  
  2  2 
In order to illustrate the coupling process, we examine the voltage on each one of the
234
1
2
lines L
 1  1 
V1 = Vo exp  − j z ( β + + β − ) z  cos( ) Vo cos  ( β + − β − ) z  (4.7.23)
 2  2 
 1  1 
V2 = jVo exp  − j z ( β + + β − ) z  sin( ) Vo sin  ( β + − β − ) z  (4.7.24)
 2  2 
Schematically these voltages are illustrated in the figure in the right. If we terminate
the ``coupler'' at a length L such that
1 π
( + −)
β − β L = (4.7.25)
2 2
the voltage is maximum on line 2 and zero on line
1. Recall β ± = K112 ± K122 . For simplicity we
shall assume K112 K122 hence
1 K122
β ± = K11 ± (4.7.26)
2 K11
and
235
1
2
L
1 1 1 K122  1 K122   1 K122
( β + − β − ) L  K11 + −  K11 −  L = L
2 2 2 K11  2 K11   2 K11
(4.7.27)
π πK
= ⇒ L = 211
2 K12
This clearly indicates that the coupling is directly related to the off diagonal term in the
2
matrix K .
Exercise 4.5: Analyze L as a function of the geometry (separation ∆ , width δ and planes
separation a ) in the previous section. Make sure that the result is independent of the
number of harmonics.
236
Chapter 5: Nonlinear Components
5.1 Transmission Line with Resistive Nonlinear Load
Consider a transmission line with a non-linear load

VL (t ) = V0 tanh( I L /I 0 ). (5.1.1)
The voltage-wave has two components, V+ propagating in the forward direction and V−
representing the reflected wave
 z  z
V ( z, t ) = V+  t −  + V−  t +  (5.1.2)
 v  v
wherein v is the phase-velocity.
Load
Z0
237 z=0
∞1 2
In a similar way the current W− = ∫ dt V− (t ).
−∞ Z0
1   z  z 
I ( z, t ) = V+  t −  −  t +  .
− V (5.1.3)
Z 0   v   v 
At the location of the load
VL (t ) = V ( z = 0, t ) = V+ (t ) + V− (t )
1 (5.1.4)
I L (t ) = I ( z = 0, t ) =
[V+ (t ) − V− (t )]
Z0
thus after substituting in the constitutive relation
 1 
V+ (t ) + V− (t ) = V0 tanh  [V+ (t ) − V− (t )]. (5.1.5)
 Z0 I 0 
It is convenient to define
V V
V± = ± and Z L = 0 (5.1.6)
Z0 I 0 Z0 I0
hence
V + + V − = Z L tanh(V + − V − ). (5.1.7)
238
10
Normalized Reflected Voltage

Assuming that the incident wave is known (V + ) this V+
Z L = 15
expression determines the reflected wave. The figure 5 Z L = 10
above shows this reflected wave for several ZL =8
normalized impedance parameters.
0
Several comments about the plot are evident:
ZL =4
(i) at the limit Z L → ∞ for a finite Z L = 0. 8
-5
solution, the reflected wave equals the
incident one. −V +
-10
(ii) At the other extreme, as Z L → 0 , the -3 -2 -1 0 1 2 3
τ
reflected wave is V − = −V + therefore
any solution is in between these two
values.
Another aspect of interest is the energy balance. The incident and reflected energy may
be evaluated by
239
∞ 1 2 ∞ 1
W+ = ∫ dt V+ (t ), W− = ∫ dt V−2 (t ). (5.1.8)
−∞ Z0 −∞ Z0
In a linear system the ratio of the two is
2
W−  Z L − 1 
ρW ≡ =  (5.1.9)
W+  Z L + 1 
wherein Z L = Z L / Z0 . The figure in the right illutrates ρW 1.0
when the maximum value of the incident (Gaussian) wave 0.8 Linear load
is V+( max ) = 2,5,10 . This figure clearly reveals that there is
Energy Ratio
0.6
full reflection when Z L is zero or tends to infinity. Zero V max = 2
reflection occurs when 0.4
(max ) V0 V+(max ) V max = 5
ZL =V+ ⇒ = ⇒ V+(max ) = V0. 0.2
V max = 10
Z0 I0 Z0 I0
0
0 5 10 15 20
ZL
240
Chapter 6: Periodic Structures
In a variety of applications it is necessary to control the phase velocity of the wave

or the propagation time. Periodic metallic or dielectric structures play an important role in
implementation of these devices. The periodic geometry can be conceived as a set of
obstacles delaying the propagation of the wave due to the multi-reflection process and, as
will be shown during our discussion on Floquet's theorem, an infinite spectrum of spatial
harmonics develops. A few of which cases these harmonics may propagate with a phase
velocity larger or equal to c but the absolute majority have a smaller phase velocity.
In the first section we present the basic theorem of periodic structures namely,
Floquet's theorem. This is followed by an investigation of closed periodic structures in
Section 6.2 and open structures in the third. Smith-Purcell effect is considered as a
particular case of a Green's function calculation for an open structure and a simple
scattering problem is also considered. The chapter concludes by presenting a simple
transient solution in a periodic structure.
241
6.1 The Floquet Theorem
A periodic function, f ( z ) , is a function whose value at a given point z is equal to

its value at a point z + L i.e.,
f ( z ) = f ( z + L), (6.1.1)
where L is the periodicity of the function. Any periodic function can be represented as a
series of trigonometric functions exp( − j 2π nz / L) and since this is an orthogonal and
complete set of functions, it implies
∞
− j 2π nz / L
f ( z ) = ∑ fn e . (6.1.2)
n = −∞
The amplitudes f n are determined by the value of the function f ( z ) in a single cell.
j 2π mz / L
Specifically, we multiply (6.1.2) by e and integrate over one cell i.e.,
∞
L j 2π mz / L L j 2π mz / L − j 2π nz / L
∫ 0
dzf ( z ) e =
0 ∫ dze ∑ n
n = −∞
f e . (6.1.3)
Using the orthogonality of the trigonometric function we have

1 L j 2π mz / L
f m = ∫ dzf ( z )e . (6.1.4)
L 0
This presentation is called the Fourier series representation and it is valid for a static
phenomenon in the sense that the value of f ( z ) at the same relative location in two
242
different cells is identical. It can not describe a propagation phenomenon, thus it can not
represent a dynamic system. In the latter case the function f ( z ) has to satisfy
f ( z ) = ξ f ( z + L), (6.1.5)
which means that the value of the function is proportional to the value of the function in
the adjacent cell up to a constant, ξ , whose absolute value has to be unity otherwise at
z → ±∞ the function diverges or is zero as can be concluded from
f ( z ) = ξ n f ( z + nL), (6.1.6)
where n is an arbitrary integer. Consequently, the coefficient ξ can be represented as a
phase term of the form ξ = exp( jψ ) hence
jψ
f ( z ) = e f ( z + L); (6.1.7)
ψ is also referred to as the phase advance per cell. Without loss of generality one can
redefine this phase to read ψ = kL . Since a-priori we do not know ψ , this definition does
not change the information available. Nonetheless based on the Fourier series in (6.1.2)
we can generalize the representation of a dynamic function in a periodic structure to
∞
− j 2π nz / L − jkz
f ( z) = ∑ fn e e , (6.1.8)
n = −∞
and realize that it satisfies
jkL
f ( z) = e f ( z + L), (6.1.9)
243
which is identical with the expression in (6.1.7). The last two expressions are different
representations of the so-called Floquet's Theorem. Later we shall mainly use the form
presented in (6.1.8), however in order to illustrate the use of Floquet's theorem in its latter
representation, we investigate next the propagation of a TM wave in a periodically loaded
waveguide.
g εr L
2R
Let us consider a waveguide of radius R which is loaded with dielectric layers: a

representative cell ( 0 ≤ z ≤ L ) consists of a region, 0 ≤ z ≤ g , filled with a dielectric, ε r ,
and the remainder is vacuum -- see Figure 1. We shall determine the dispersion relation
of this structure and for this purpose we write the solution of the magnetic vector
244
potential and electromagnetic field (steady state) in the dielectric ( 0 ≤ z ≤ g ):
∞
 r −Γ z +Γ z
Az ( r , z ) = ∑J 0  ps   As e d ,s + Bs e d ,s ,
s =1  R   
c2 ∞
 r −Γ z +Γ z
Γ d ,s J1  ps   As e d ,s − Bs e d ,s ,
ps
Er ( r , z ) =
jωε r
∑
s =1 R  R   
c2 ∞
 r   −Γd ,s z +Γd ,s z 
Ez ( r, z ) = − ∑
jωε r s =1
Γ 2
J
d ,s 0  s
p
 R 
 s
A e + Bs e


,
1 ∞ ps  r   −Γd ,s z +Γd ,s z 
Hϕ ( r , z ) = ∑   J p
µ0 s =1 R 1  s R   s
A e + Bs e

, (6.1.10)
where Γ2d , s = ( ps / R )2 − ε r (ω / c )2 . In a similar way, we have in the vacuum ( g < z < L ):
245
∞
 r −Γ ( z − g ) +Γ ( z − g ) 
Az ( r , z ) = ∑J 0  ps  Cs e s + Ds e s ,
s =1  R 
 

c 2 ∞ ps  r   −Γs ( z − g ) +Γs ( z − g ) 
Er ( r , z ) = ∑ s 1  s R   s
jω s =1 R
Γ J p C e − D s e

,
c 2 ∞ 2  r   −Γs ( z − g ) +Γ ( z − g ) 
Ez ( r , z ) = − ∑
jω s =1
Γ s J 0  ps  C s e
 R 

+ Ds e s


,
1 ∞ ps  r   −Γs ( z − g ) +Γ ( z − g ) 
Hϕ ( r , z ) = ∑ J 1  ps  C s e
µ0 s =1 R  R   
+ Ds e s


, (6.1.11)
with Γ2s = ( ps / R )2 − (ω / c )2 . At this point we shall consider only the TM 01 mode ( s = 1 )

thus the continuity of the radial electric field at z = g implies
−Γ g +Γ g
Γ d ,1  A1e d ,1 − B1e d ,1  = Γ1 [C1 − D1 ],
1
(6.1.12)
εr  
and in a similar way the continuity of the azimuthal magnetic field reads
−Γ g +Γ g
A1e d ,1 + B1e d ,1 = C1 + D1. (6.1.13)
Last two equations express the relation between the amplitudes of the field in the
dielectric and vacuum.
In the dielectric filled region of next cell ( L ≤ z ≤ L + g ) the field has a similar form
246
as in (6.1.10) i.e.,
∞
 r −Γ ( z − L ) +Γ ( z − L ) 
Az ( r , z ) = ∑J 0  ps   As' e d ,s + Bs' e d ,s ,
s =1  R   

c2 ∞  r   ' −Γd ,s ( z − L ) ' +Γd ,s ( z − L ) 
Er ( r , z ) = ∑ s d ,s 1  s R   s
jωε r s=1
p R Γ J p A e − B se 
,
c2 ∞
 r   ' −Γd ,s ( z − L ) ' +Γd ,s ( z − L ) 
Ez ( r , z ) = − ∑
jωε r s=1
Γ 2
J
d ,s 0  s
p
 R 
 s
A e − Bse 

,
1 ∞  r   ' −Γd ,s ( z − L ) ' +Γd ,s ( z − L ) 

Hϕ ( r , z ) = ∑ R J
µ0 s=1 s 1  s R   s
p p A e + B se 
(6.1.14)
Accordingly, the boundary conditions at z = L read

1 −Γ ( L − g ) Γ ( L− g ) 
Γ d ,1  A1' − B1'  = Γ1 C1e 1 − D1e 1 , (6.1.15)
εr 
 

and
−Γ1 ( L − g ) Γ1 ( L − g )
A1' + B1' = C1e + D1e . (6.1.16)
The relation between the amplitudes of the wave in the second cell ( L < z < 2 L ) and
the first cell can be represented in a matrix form
247
a' = Ta, (6.1.17)
where the components of a' are A1' and B1' and similarly the components of a are A1 and
B1 . According to Floquet's theorem (6.1.9) the two vectors are expected to be related by
− jkL
a' = e a, (6.1.18)
− jkL
thus e represents the eigen-values of the single cell transmission matrix T :
− jkL
|T−e I |= 0. (6.1.19)
Explicitly this reads
−2 jkL − jkL
e −e (T11 + T22 ) + T11T22 − T12T21 = 0. (6.1.20)
For a passive system the determinant of the matrix T is unity (prove this statement!!),
thus
−2 jkL − jkL
e −e (T11 + T22 ) + 1 = 0. (6.1.21)
The fact that the last term in this equation is unity indicates that if k is a solution of
(6.1.21) −k is also a solution. Consequently, we can write
1
cos( kL) = (T11 + T22 ). (6.1.22)
2
Note that this is an explicit expression for k as a function of the frequency and the other
geometric parameters. In principle there are ranges of parameters where the right-hand
248
side is larger than unity and there is no real k which satisfies this relation. If only the
frequency is varied then this result indicates that there are frequencies for which the
solution of the dispersion relation is real thus a wave can propagate, or the solution is
imaginary and the amplitude of the wave is zero. The frequency range for which the wave
is allowed to propagate is called the passband. Explicitly, the right-hand side of (6.1.22)
reads
1 ( Z1 + Z 2 )2 ( Z1 − Z 2 )2
(T11 + T22 ) = cosh(ψ + χ ) − cosh(ψ − χ ), (6.1.23)
2 4 Z1 Z2 4 Z1 Z2
where ψ = Γ1 ( L − g ) , χ = Γ d ,1 g , the characteristic impedances are
η0 cΓ d ,1 η cΓ
2 Z1 = , Z2 = 0 1 , (6.1.24)
jωε r jω
20
and η0 = 377Ω is the impedance of the
1 vacuum.
15
f (GHz )
0
10
-1
5
-2
0 5 10 15 20 25 0
249 0.0 0.4 0.8
f (GH z ) kL / π
Figure 2 illustrates the right-hand side of (6.1.22) as a function of the frequency
( ε r = 10,R = 2 cm, L = 1 cm and g = L / 2 ). The blocks at the bottom, illustrate the
forbidden frequencies, namely at these frequencies TM waves can not propagate. In
Figure 3 the dispersion relation of the first three passbands is presented; these branches
correspond only to the TM 01 mode. Higher symmetric or asymmetric modes have
additional contributions in this range of frequencies.
Comment 1. The expression in (6.1.22) is the dispersion relation of a TM 01 mode in the
periodic structure illustrated in Figure 1. From this simple example however we observe
that the dispersion relation of a periodic structure is itself periodic in k with a
periodicity 2π / L . This is a general feature which can be deduced from (6.1.9). If the
latter is satisfied for k = k0 then (6.1.9) is satisfied also for k = k0 + 2π / L as shown next
j ( k + 2π / L ) L jk L j 2π
f ( z + L )e 0 = f ( z + L )e 0 e ,
jk L
= f ( z + L)e 0 = f ( z ). (6.1.25)
Consequently, since the dispersion relation is periodic in k , it is sufficient to represent
its variation with k in the range −π / L ≤ k ≤ π / L ; this k domain is also called the first
Brillouin zone.
Comment 2. Bearing in mind the last comment, we can re-examine the expression in
250
(6.1.8) and realize that f ( z ) is represented by a superposition of spatial harmonics
exp( − jkn z ) where
2π
kn = k + n, (6.1.26)
L
which all correspond to the solution of the dispersion relation of the system. According
to this definition the phase velocity of each harmonic is
ω
vp h , n = , (6.1.27)
ckn
and for a high harmonic index, n , this velocity decreases as n −1 . Furthermore, all
harmonics with negative index correspond to backward propagating waves. In addition,
note that the zero harmonic ( n = 0 ) has a positive group velocity for π / L > k > 0 and
negative in the range −π / L < k < 0 . This is a characteristic of all spatial harmonics.
Since the group velocity is related to the energy velocity, one can conclude that although
the the wave number of a particular space harmonic is positive, the power it carries may
flow in the negative direction (if the group velocity is negative).
251
d
6.2 Closed Periodic Structure

2R ext 2R int
Based on what was shown in the previous section
one can determine the dispersion relation of a TM 01
L
mode which propagates in a corrugated waveguide
[Brillouin (1948)]. Its periodicity is L , the inner
radius is denoted by Rint and the external by Rext ; the
distance between two cavities (the drift region) is d . Using Floquet's Theorem (6.1.8)
we can write for the magnetic potential in the inner cylinder ( 0 < r < Rint ) the following
expression
∞
− jkn z
Az ( r , z ) = ∑ An e I0 ( Γ n r ), (6.2.1)
n = −∞
and accordingly, the electromagnetic field components read
252
c2 ∞ − jkn z
Er ( r , z ) = ∑
jω n= −∞
( − jkn Γ n ) An e I1 ( Γ n r ),
c2 ∞ − jkn z
Ez ( r , z ) = ∑ n n
jω n= −∞
( −Γ 2
) A e I0 ( Γ n r ),
∞
1 − jkn z
Hϕ ( r , z ) = −
µ0
∑Γ
n = −∞
n An e I1 ( Γ n r ) (6.2.2)
In these expressions,
ω2
Γ =k −
2
n
2
n 2
, (6.2.3)
c
and I0 ( x ), I1 ( x ) are the zero and first order modified Bessel functions of the first kind
respectively. This choice of the radial functional variation is dictated by the condition of
convergence of the electromagnetic field on axis.
In each individual groove (superscript σ ) the electromagnetic field can be derived
from the following magnetic vector potential:
∞
A(σ )
z ( r , z ) = ∑Bν(σ ) cos [ qν ( z − zσ − d )] t0,ν ( r ), (6.2.4)
ν =0
where qν = πν / ( L − d ) ,
t0,ν ( r ) = I0 ( Λν r )K 0 ( Λν Rext ) − K 0 ( Λν r )I0 ( Λν Rext ), (6.2.5)
253
and Λν2 = qν2 − (ω / c )2 . The electromagnetic field reads
c2 ∞
(σ )
Er ( r , z ) = ∑
jω ν =0
( − qν )Λν Bν(σ ) sin[ qν ( z − zσ − d )]t1,ν ( r ),
c2 ∞
E (σ )
z (r, z ) = ∑
jω ν =0
( −Λν2 ) Bν(σ ) cos[ qν ( z − zσ − d )]t0,ν ( r ),
∞
1
(σ )
Hϕ ( r , z ) = −
µ0
∑
ν
Λν Bν σ
=0
( )
cos[ qν ( z − zσ − d )]t1,ν ( r ), (6.2.6)
In these expressions t1,ν ( r ) is the derivative of t0,ν ( r ) defined by

t1,ν ( r ) = I1 ( Λν r )K 0 ( Λν Rext ) + K1 ( Λν r )I0 ( Λν Rext ), (6.2.7)
and except at r = Rint all the boundary conditions are satisfied.
6.2.1 Dispersion Relation
Our next step is to impose the continuity of the boundary conditions at the interface
( r = Rint ). The continuity of the longitudinal component of the electric field
[ Ez ( r = Rint , −∞ < z < ∞) ] reads
254
c2 ∞ − jkn z
∑ n n
jω n = −∞
( −Γ 2
) A e I0 ( Γ n Rint )

0 for zσ < z < zσ + d ,

= (6.2.8)
 c2 ∞
− ∑
 jω ν =0
Λν2 Bν(σ ) cos[ qν ( z − zσ − d )]t0,ν ( Rint ) for zσ + d < z < zσ + L,
and the azimuthal magnetic field [ Hϕ ( r = Rint , zσ + d < z < zσ + L) ] reads

1 ∞ − jkn z 1 ∞
−
µ0
∑ Γn An e I1 ( Γn Rint ) = − ∑Λν Bν(σ ) cos[ qν ( z − zσ − d )]t1,ν ( Rint ). (6.2.9)
n = −∞ µ0 ν =0
From these boundary conditions the dispersion relation of the structure can be
developed. For this purpose we analyze the solution in the grooves having Floquet's
theorem in mind. The latter implies that the longitudinal electric field in the σ 's groove
has to satisfy the following relation:
∞ ∞
2 (σ +1) jkL
∑ ν ν
Λ
ν =0
2 (σ )
B cos[ qν ( z − zσ − d )]t 0,ν ( r ) = ∑ ν Bν e cos[qν ( z + L − zσ +1 − d )]t0,ν (r ).
Λ
ν =0
255
(6.2.10)
But by definition zσ +1 − zσ = L therefore, the last expression implies that
− jkzσ
Bν(σ ) Bν e . (6.2.11)
This result permits us to restrict the investigation to a single cell and without loss of
generality we chose zσ =0 = 0 since if we know Bν in one cell, the relation in (6.2.11)
determines the value of this amplitude in all other cells. With this result in mind we
jk z
multiply (6.2.8) by e m and integrate over one cell; the result is
∞ ∞ L jk z
∑ n n n,m 0 n int ∑ ν ν 0,ν int ∫ dze m cos[ qν ( z − d )],
Γ
n = −∞
2
A δ LI ( Γ R ) Λ 2
B t
ν =0
( R )
d
(6.2.12)
here δ n ,m is the Kroniker delta function which equals 1 if n = m and zero otherwise. We
also used the orthogonality of the Fourier spatial harmonics. We follow a similar
procedure when imposing the continuity of the magnetic field with one difference, (6.2.9)
is defined only in the groove aperture thus we shall utilize the orthogonality of the
trigonometric function cos[ qν ( z − d )] . Accordingly, (6.2.9) is multiplied by
cos[ qµ ( z − d )] and we integrate over d < z < L ; the result is
∞ ∞
L − jkn z
∑Γ
n = −∞
n An I1 ( Γ n Rint ) ∫ dz cos  qµ ( z − d )  e
d
= ∑Λν Bν t1,ν ( Rint )( L − d ) g µ δν , µ ,
ν =0
(6.2.13)
256
where g0 = 1 and g µ ≠ 0 = 0.5 otherwise. It is convenient to define the quantity
1 jkn z
[ ]
L
Ln ,ν ( k ) =
L−d ∫d
dz cos qν ( z − d ) e , (6.2.14)
which allows us to write (6.2.12) as
1 L−d ∞ 2
An = 2
Γ n I0 ( Γ n Rint ) L ν =0
∑ Λν t0,ν ( Rint )Ln ,ν ( k ) Bν , (6.2.15)
and (6.2.13) as
∞
1
Bν = ∑
Λν t1,ν ( Rint ) gν n = −∞
An Γ n I1 ( Γ n Rint )L*n ,ν ( k ). (6.2.16)
These are two equations for two unknown sets of amplitudes ( An , Bν ) and the
dispersion relation can be represented in two equivalent ways: One possibility is to
substitute (6.2.16) in (6.2.15) and get
∞  
L − d Γ m I1 ( Γ m Rint ) ∞ t0,ν ( Rint )Λν
∑ δ
 n,m
m = −∞ 
− ∑
L Γ n I0 ( Γ n Rint ) ν =0 t1,ν ( Rint ) gν
2
Ln ,ν m ,ν  Am = 0,
L*
(6.2.17)

whereas the other possibility is to substitute (6.2.15) in (6.2.16) and obtain
257
∞  L − d Λ µ t0, µ ( Rint ) ∞ I1 ( Γ n Rint ) *
2

∑
µ =0 
δ
 ν ,µ − ∑
L Λν t1,ν ( Rint ) gν n = −∞ Γ n I0 ( Γ n Rint )
Ln ,ν n , µ  Bµ 0.
L (6.2.18)

In both cases the dispersion relation is calculated from the requirement that the
determinant of the matrix which multiplies the vector of amplitudes, is zero.
Although the two methods are equivalent, we found that the latter expression is by
far more efficient for practical calculation because of the number of modes required to
represent adequately the field in the groove compared to the number of spatial harmonics
required to represent the field in the inner section. In the case of single mode operation
we found that 1 to 3 modes are sufficient for description of the field in the grooves and
about 40 spatial harmonics are generally used in the inner section. As indicated by these
numbers it will be much easier to calculate the determinant of a 3 × 3 matrix rather than
40 × 40 one; we shall quantify this statement later. At this point we shall discuss the
design of a disk-loaded structure assuming that the number modes in the grooves and
harmonics in the inner space are sufficient.
Let assume that we want to determine the geometry of a disk-loaded structure which
enables a wave at 10 GHz to be in resonance with electrons with β = 0.9 and the phase
advance per cell is assumed to be kL = 2π / 3. These two conditions determine the period
of the structure. In our case L = 9 mm. There are three additional geometric parameters
to be determined: Rext ,Rint and d . The last two have
258
11
R ext = 14mm
Rint = 8mm
L = 9mm
d = 2mm
10 kL = 2π / 3
a dominant effect on the width of the passband and for the
f (GHz )
lowest mode, the passband increases with increasing Rint and
decreases with increasing d . The passband, ∆ω , of a mode sets
a limit on the maximum group velocity as can be seen bearing in 9
mind that the half width of the first Brillouin zone is ∆k = π / L .
Consequently, vgr = ∆ω / ∆k < ∆ω L / π . A solution of the Beam Line
dispersion relation in (6.2.18) is illustrated the geometry chosen β = 0.9
is: L = 9 mm, Rint = 8 mm and d = 2 mm and from the condition 8
of phase advance per cell of 120o at 10 GHz, we determined, 0.0 0.4 0.8
using the dispersion relation, the value of the external radius to kL / π
be Rext = 13.96 mm.
6.2.2 Modes in the groove

Therefore, before we conclude this subsection, we shall quantify the effect of higher
modes. The first mode in the groove (ν = 0 ) represents a TEM mode which propagates in
the radial direction. Other modes (TM 0,ν >0 ) are either propagating or evanescent. The
amplitudes of the magnetic and electric field ( Ez ) of the TEM mode are constant at the
groove aperture thus the choice of using a single mode in the groove is equivalent to the
259
average process at the boundary
-- approach usually adopted in 11.0 8.75
Modes=1 Modes=1
the literature. The Figure =2 =2
illustrates the dependence of 10.8 =3 8.73 =3
upper and lower cut-off
10.6 8.71
f (GHz)
f (GHz)
frequency on the number of
harmonics used in the
10.4 8.69
calculation; the number of
modes in the grooves is a 10.2 8.67
parameter. For the geometry
presented above, the number of 10.0 8.65
harmonics required is 20 or 0 20 40 60 80 100 0 20 40 60 80 100
larger; typically about 40 Total Number of Total Number of
Harmonics Harmonics
harmonics are being used. The
effect of the ν = 1 mode is
negligible in this case as seen for both upper and lower cut-off frequencies. The effect of
the higher mode introduces a correction on the order of 1% which for most practical
purposes is sufficient.
260
6.2.3 Spatial Harmonics Coupling
Contrary to uniform dielectric structures, here each mode consists of a superposition

of an infinite number of spatial harmonics. These harmonics are all coupled by the
conditions imposed on the electromagnetic field by the geometry at r = Rint . We shall
limit the investigation to the accuracy associated with a single mode taken in the groove,
therefore according to (6.2.15), we have
1 ω2
An = − 2 t ( Rint )Ln ,0 ( k ) B0 , (6.2.19)
Γ n I0 ( Γ n Rint ) c 2 0,0
and in this particular case

L−d 1   1 
Ln ,0 ( k ) = sinc  kn ( L − d )  exp  j kn ( L + d ) . (6.2.20)
L  2   2 
Let us compare the first few spatial harmonics relative to the zero harmonic. For this
purpose we take f = 10 GHz, v0 = 0.9c,Rint = 8 mm, L = 9 mm and d = 2 mm. The ratio
of the first two amplitudes is
261
A−1 A A A
= 8 × 10−3 , 1 = 3 × 10−6 , −2 = 2 × 10−6 , 2 = 1 × 10−8 (6.2.21)
A0 A0 A0 A0
This result indicates that on axis, the amplitude of the interacting harmonic is dominant.
At the interface with the grooves ( r = Rint ) the ratio between the contribution of the zero
and n th harmonic is much closer to unity and it can be checked that it reads
| Ez ,n ( r = Rint ) | | sinc( kn ( L − d ) / 2) |
= , (6.2.22)
| Ez ,0 ( r = Rint ) | | sinc( k0 ( L − d ) / 2) |
which is a virtually unity.
A more instructive picture is obtained by examining the average power flowing in
one time and spatial period of the system:
Rint 1 L 1 
P = 2π ∫ drr ∫ dz  Er ( r , z ) Hϕ* ( r , z ) . (6.2.23)
0 L 0
2 
According to the definition in (6.2.2) we have
π ∞ 2 ck n
Γ n Rint
P=
η0
∑ n
n = −∞
| cA
ω
|
0 ∫ dxxI 2
1 ( x ); (6.2.24)
the integral can be calculated analytically [Abramowitz and Stegun (1968) p.484] and it
reads
262
ξ 1
U (ξ ) ≡ ∫ dxxI12 ( x )ξ I0 (ξ )I1 (ξ ) + ξ 2 [I12 (ξ ) − I02 (ξ )]. (6.2.25)
0 2
Based on these definitions we can calculate the average power carried by each harmonic
as
π ck
Pn = | cAn |2 n U ( Γ n Rint ), (6.2.26)
η0 ω
and the result is listed below
P−2 P P P
= −3 × 10−3 , −1 = −0.16, 1 = 1 × 10−4 , 2 = 3 × 10−3 (6.2.27)
P0 P0 P0 P0
Although there is a total flow of power along (the positive) direction of the z axis, a
substantial amount of power is actually flowing backwards. In this numerical example for
all practical purposes we can consider only the lowest two harmonics and write the total
power which flows, normalized to the power in the zero harmonic. Thus if the latter is
unity, then the power in the forward is 1 − 0.16 = 0.84 . This result indicates that if we
have a finite length structure with finite reflections from the input end, then in this
periodic structure we have an inherent feedback even if the output end is perfectly
matched.
263
6.3 Open Periodic Structure
2 R ext 2Rint
In this section an analysis similar to L
that in Section 6.2 is applied to an open
periodic structure. As we shall see the
number of modes which may develop in d
such a structure is small and therefore
mode competition is minimized.
We shall consider a system in which the wave propagates along the periodic
structure which consists of a disk-loaded wire, as illustrated in Figure 5 whose periodicity
is L , the inner radius is denoted by Rint , the external by Rext and the distance between two
cavities is d . Floquet's theorem as formulated in (6.1.8) allows us to write for the
magnetic potential in the external region ( ∞ > r ≥ Rext ) the following expression
∞
− jkn z
Az ( r , z ) = ∑ An e K 0 ( Γ n r ), (6.3.1)
n = −∞
and accordingly, the electromagnetic field components read
264
c2 ∞ − jkn z
Er ( r , z ) = ∑ n n n
jω n = −∞
( jk Γ ) A e K1 (Γ n r ),
c2 ∞ − jkn z
Ez (r , z ) = ∑
jω n = −∞
( −Γ n2 ) Ane K 0 (Γ n r ), (6.3.2)
∞
1 − jkn z
Hϕ (r , z ) = − ∑ (−Γn ) Ane
µ0 n = −∞
K1 ( Γ n r )
In these expressions K0 ( x ), K1 ( x ) are the zero and first order modified Bessel functions
of the second kind respectively and Γ2n = kn2 − (ω / c )2 . This choice of the radial functional
variation is dictated by the condition of convergence of the electromagnetic field far
away from the structure.
In each individual groove the electromagnetic field can be derived from the
following magnetic vector potential:
∞
A (σ )
z ( r , z ) = ∑Bν(σ ) cos[ qν ( z − zσ − d )]t0,ν ( r ), (6.3.3)
ν =0
where qν = πν / ( L − d ) ,
t0,ν ( r ) = I0 ( Λν r )K 0 ( Λν Rint ) − K 0 ( Λν r )I0 ( Λν Rint ), (6.3.4)
and Λν2 = qν2 − (ω / c )2 . The electromagnetic field reads
265
c2 ∞
E (σ )
r (r, z ) = ∑
jω ν =0
( − qν )Λν Bν(σ ) sin[ qν ( z − zσ − d )]t1,ν ( r ),
c2 ∞
E (σ )
z (r, z ) = ∑
jω ν =0
( −Λν2 ) Bν(σ ) cos[ qν ( z − zσ − d )]t0,ν ( r ),
∞
1
(σ )
Hϕ ( r , z ) = −
µ0
∑
ν
Λν Bν σ
=0
( )
cos[ qν ( z − zσ − d )]t1,ν ( r ). (6.3.5)
The index σ labels the ``cavity" and in these expressions we used

t1,ν ( r ) = I1 ( Λν r )K 0 ( Λν Rint ) + K1 ( Λν r )I0 ( Λν Rint ). (6.3.6)
The solution above satisfies all boundary conditions with the exception of r = Rext .
6.3.1 Dispersion Relation
Our next step is to impose the continuity of the boundary conditions at the interface
( r = Rext ). Continuity of the longitudinal component of the electric field implies
Ez ( r = Rext , −∞ < z < ∞ ) , reads
266
c2 ∞ − jkn z
∑ n n
jω n = −∞
( −Γ 2
) A e K0 ( Γ n Rext ) =

0 for zσ < z < zσ + d ,

 (6.3.7)
 c2 ∞
− ∑
 jω ν =0
Λν2 Bν(σ ) cos[ qν ( z − zσ − d )]t0,ν ( Rext ) for zσ + d < z < zσ + L,
and the azimuthal magnetic field, Hϕ ( r = Rext , zσ + d < z < zσ + L) , reads

1 ∞ − jkn z 1 ∞
µ0
∑ Γn An e K1 ( Γn Rext ) − ∑Λν Bν(σ ) cos[ qν ( z − zσ − d )]t1,ν ( Rext ). (6.3.8)
n = −∞ µ0 ν =0
Following the same arguments as in Section 6.2.1 it can be shown that

− jkzσ
Bν(σ ) = Bν e and consequently we can limit the discussion to a single cell. We
jk z
multiply (6.3.7) by e m and integrate over one cell; the result being
∞ ∞ L jkm z
∑ n n n,m 0 n ext ∑ ν ν 0,ν ext ∫
Γ
n = −∞
2
A δ LK ( Γ R ) Λ 2
B t ( R
ν =0
) dze cos[ qν ( z − d )].
d
(6.3.9)
267
We follow a similar procedure when imposing the continuity of the magnetic field; the
difference in this case is that (6.3.8) is defined only in the groove's aperture thus we shall
utilize the orthogonality of the trigonometric function cos[ qν ( z − d )] . Accordingly,
(6.3.8) is multiplied by cos[ qµ ( z − d )] and we integrate over d < z < L ; the result is
∞ ∞
L − jkn z
∑Γ
n = −∞
n An K1 ( Γ n Rext ) ∫ dz cos[ qµ ( z − d )]e
d
− ∑Λν Bν t1,ν ( Rext )( L − d ) g µ δν , µ .
ν =0
(6.3.10)
In this expression g0 = 1 and g n ≠ 0 = 0.5 . It is convenient to define the quantity
1 L jkn z
Ln ,ν ( k ) =
L−d ∫d
dz cos[ qν ( z − d )]e , (6.3.11)
by whose means, (6.3.9) reads
1 L−d ∞ 2
An = 2
Γ n K 0 ( Γ n Rext ) L ν =0
∑Λν t0,ν ( Rext )Ln,ν ( k ) Bν , (6.3.12)
whereas (6.3.10)
∞
1
Bν − ∑
Λν t1,ν ( Rext ) gν n = −∞
An Γ n K1 ( Γ n Rext )L*n ,ν ( k ). (6.3.13)
These are two equations for two unknown sets of amplitudes ( An , Bν ) . The
268
dispersion relation can be represented in two equivalent ways: One possibility is to
substitute (6.3.13) in (6.3.12) and obtain one equation for the amplitudes of the various
harmonics
∞  
L − d Γ m K1 ( Γ m Rext ) ∞ t0,ν ( Rext )Λν
∑ δ n ,m +
m = −∞  L 2 ∑
Γ n K 0 ( Γ n Rext ) ν =0 1,ν ext ν
t ( R ) g
*
Ln ,ν Lm,ν  Am = 0. (6.3.14)

The other possibility is to substitute (6.3.12) in (6.3.13) and obtain one equation for the
amplitudes of the various modes in the groove
∞  
L − d Λ µ t0, µ ( Rext ) ∞ K1 ( Γ n Rext ) *
2
∑ δν , µ +
µ =0  L Λ t ( R ) g
∑ Γ K ( Γ R )
Ln ,ν Ln , µ  Bµ = 0. (6.3.15)
ν 1,ν ext ν n = −∞ n 0 n ext 
In both cases the dispersion relation is calculated from the requirement that the
determinant of the matrix which multiplies the vector of amplitudes, is zero. As in the
closed structure the two methods are equivalent, but the last expression is by far more
efficient for practical calculation.
There is one substantial difference between open and closed periodic structures. In the
latter case, the radiation is guided by the waveguide and there is an infinite discrete
spectrum of frequencies which can propagate along the system. In open structures, modes
can propagate provided that the projection of the wavenumbers of all harmonics in the
first Brillouin zone corresponds to waves whose phase velocity is smaller than c ; in other
269
ω
β ph = 1 β ph = 1
words, no radiation propagates outwards
(radially). The figure illustrates the two
regions of interest: in the shadowed region no
solutions are permissible and in the remainder
the solution is possible with an adequate
choice of the geometric parameters. It is
evident from this picture that waves at
frequencies higher than -6 -4 -2 0 2 4 kL / π
1c
f ≥ , (6.3.16)
2L 50
Rint = 15 mm
R ext = 21mm
40
can not be supported by a disk-loaded wire, regardless of the L = 3mm
geometrical details of the cavity. With this regard, an open d = 1mm
structure forms a low pass filter. Figure 3.1 illustrates the 30
f (GHz )
dispersion relation of such a system for L = 3 mm, d = 1 mm,
Rint = 15 mm and Rext = 21 mm. For comparison, in the same 20
frequency range (0--50 GHz) there are 6 symmetric TM modes β ph = 1
which can propagate in a closed system of the same geometry;
obviously there are many others at higher frequencies. 10
0
270 0.0 0.2 0.4 0.6 0.8 1.0
kL / π
6.4 Transients
In order to illustrate the effect of the periodicity on the propagation of a wave packet
we shall consider at t = 0 the same wavepacket a ( z ) , in vacuum and in a periodic
structure. The propagation in vacuum will be represented by a dispersion relation
k 2 = ω 2 / c 2 , therefore a scalar wave function Ψ ( z , t ) is given by
∞ − jkz 1  jkct − jkct 
Ψ ( z , t ) = ∫ dkψ ( k )e e +e . (6.4.1)
−∞ 2  

Since at t = 0 this function equals a ( z ) , the amplitudes ψ ( k ) can be readily determined
using the inverse Fourier transform hence
1 ∞ jkz
ψ (k ) =
2π ∫−∞
dza ( z ) e . (6.4.2)
Substituting back into (4.1) we find that
1
Ψ ( z , t ) = [ a ( z − ct ) + a ( z + ct )], (6.4.3)
2
which basically indicates that the pulse moves at the speed of light in both directions and
asymptotically, it preserves its shape. In a periodic structure the description of the
wavepacket is complicated by the dispersion relation which in its lowest order
approximation (e.g., first TM symmetric mode in a waveguide) can be expressed as
271
ω( k ) = ω − δω cos( kL), (6.4.4)
where ω = (ω0 + ωπ ) / 2 is the average frequency between the low ( kL = 0 ) cut-off
denoted by ω0 and the high ( kL = π ) cut-off denoted by ωπ . The quantity
δω = (ωπ − ω0 ) / 2 is half the passband width and L is the period of the structure.
Contrary to the previous case k here denotes the wavenumber in the first Brillouin zone.
In the framework of this approximation we can use Floquet's representation to write
 ∞ π /L jω( k )t − jkn z 
Ψ ( z , t ) = Re  ∑ ∫ dkψ n ( k )e , (6.4.5)
−π / L
 n = −∞ 
where kn = k + 2π n / L . The amplitudes ψ n ( k ) are determined by the value of the
function at t = 0 hence
∞ jk z
ψ n ( k ) = 12π ∫ dza( z )e n . (6.4.6)
−∞
Substituting back into (6.4.5) we have
 1 ∞ ∞ π /L jt[ω −δω cos( kL)]− jkn ( z −ζ ) 
Ψ ( z , t ) = Re  ∫ d ζ a (ζ ) ∑ ∫ dke . (6.4.7)
 2π −∞
n = −∞
− π / L

At this point we can take advantage of
272
1
ζ (τ −1/τ ) ∞
e2 ≡
ν
∑ τ ν Jν (ζ ),
= −∞
(6.4.8)
and simplify the last equation to read

∞ ∞
 1 ∞ π /L jωt j( kL−π /2)ν − jkn ( z −ζ ) 
Ψ ( z , t ) = Re  ∫ d ζ a (ζ ) ∑ ∫ dke × ∑ Jν (δωt )e e  ,
 2π −∞
n = −∞
− π / L
ν = −∞ 
(6.4.9)
which after the evaluation of the integrals and summation (over n ) reads
∞
Ψ( z, t ) =
ν
∑ a ( z −ν L)Jν (δωt ) cos(ωt − πν / 2).
= −∞
(6.4.10)
273
0.5 0.5
The figure illustrates the propagation of two t = 3L / c t = 6L / c
wavepackets in vacuum (dashed line) and in a
periodic structure. The latter is characterized by
Ψ (z, t )
Ψ (z, t )
ω = 2π ×10 GHz, δω = ω / 30 and a spatial 0.0 0.0
periodicity of L = 1 cm. At t = 0 the distribution is a
Gaussian, a ( z ) = exp[ −( z / L)2 ] . In each one of the
frames Ψ ( z , t ) was plotted at a different time as a -0.5 -0.5
0 5 10 15 0 5 10 15
function of z . Characteristic to all the frames is the z/L z/L
relatively large peak following the front of the pulse. 0.5 0.5
t = 12 L / c t = 9L / c
It is evident that although the front of the pulse
propagates at the speed of light (as in vacuum) the
Ψ (z, t )
Ψ (z, t )
main pulse propagates slower. In fact, a substantial 0.0 0.0
fraction of the energy remains at the origin even a
long time after t = 0 . For the parameters used, the
amplitude of the signal at the origin ( z = 0 ) is
-0.5 -0.5
dominated by the zero order Bessel function i.e., 0 5 10 15 0 5 10 15
J 0 (δωt ) therefore the energy is drained on a time z/L z/L
scale which is determined by the asymptotic behavior
of the Bessel function namely ∝ 1 / δωt . Clearly the wider the passband the faster the
energy is drained from the origin.
274
Exercise 6.1: Based on the solution for Az ( r , z ) in Section 6.1 determine the Floquet
representation of the magnetic vector potential (TM 01 ). In other words write
 r − jkn z
Az ( r , z ) = J 0  p1  ∑an ( k )e
 R n
and determine an ( k ) .
Exercise 6.2: Find all the waves which can propagate between f = 0 to 20 GHz,
including asymmetric modes for the system described in Section 6.1. Repeat this exercise
for the branches of the TE modes.
Exercise 6.3: Analyze the coupling of spatial harmonics for the system in Section 6.3 in
a similar way as in Section 6.2.2.
Exercise 6.4: Repeat the calculation of the propagation of a transient in a periodic
structure (Section 6.4) but this time for a TEM-like mode. [Hint: take ω = ωπ ∼ ( kh ) .]
275
Chapter 7: Generation of radiation
Throughout these notes it was always assumed that somewhere there is a radiation source
providing us with the necessary energy. In what follows we shall skim through the
fundamentals of generation of radiation. Our discussion will be limited to ``vacuum''
devices, this is to say that the energy is extracted from free electrons in vacuum.
Employed on medium ( < 1kW) and high ( > 1kW) power devices, the concepts to be
discussed rely solely on energy and momentum conservation. Another very important
family of devices relying on electrons moving in ``solid state'', will not be discussed here.
276
7.1 Single-Particle Interaction
On its own, an electron cannot transfer energy via a linear process to a

monochromatic electromagnetic wave in vacuum if the interaction extends over a very
long region. Non-linear processes may facilitate energy exchange in vacuum, but this
kind of mechanism is rarely used since most systems require a linear response at the
output. Therefore throughout this text we shall consider primarily linear processes and in
this introductory chapter we shall limit the discussion to single-particle schemes.
Collective effects, where the current is sufficiently high to affect the electromagnetic
field, are also important but cannot be covered at this stage and they are the essence of a
different course.
7.1.1 Infinite Length of Interaction
Far away from its source, in vacuum, an electromagnetic wave forms a plane wave
which is characterized by a wavenumber whose magnitude equals the angular frequency,
ω , of the source divided by c = 299,792,458m s −1 , the phase velocity of the plane wave
in vacuum, and its direction of propagation is perpendicular to both the electric and
magnetic field. For the sake of simplicity let us assume that such a wave propagates in
the z direction and the component of the electric field is parallel to the x axis i.e.,
277
  z 
E x ( z , t ) = E0 cos ω  t −  . (7.1.1)
  c 
If a charged particle moves at v parallel to z axis, then the electric field this charge
experiences (neglecting the effect of the charge on the wave) is given by
  z (t )  
E x ( z (t ), t ) = E0 cos ω  t −  . (7.1.2)
  c 
A crude estimate for the particle's trajectory is
z (t ) vt, (7.1.3)
therefore if the charge moves in the presence of this wave from t → −∞ to t → ∞ then
the average electric field it experiences is zero,
∞   v 
∫−∞ dt cos ωt 1 − c   = 0,
   
(7.1.4)
even if the particle is highly relativistic. The lack of interaction can be illustrated in a
clearer way by superimposing the dispersion relation of the wave and the particle on the
same diagram. Firstly, the relation between energy and momentum for an electron is
given by
E = c p 2 + ( mc )2 , (7.1.5)
where m = 9.1094 ×10−31 Kg is the rest mass of the electron. Secondly, the corresponding
278
relation for a photon in free space is
E = cp. (7.1.6)
For the interaction to take place the electron has to change its initial state, subscript i,
denoted by ( Ei , pi ) along the dispersion relation to the final, subscript f, denoted by
( Ef , pf ) in such a way that the resulting photon in case of emission or absorbed photon
for absorption, has exactly the same difference of energy and momentum i.e.,
Ei = Ef + Eph , (7.1.7)
and
pi = pf + pph . (7.1.8)
In vacuum this is impossible, as can be shown by
substituting (7.1.5)--(7.1.6)in (7.1.7)--(7.1.8). We E
can also reach the same conclusion by examining
Fig. 1.1 The expression, E = cp , which describes E i
the photon's dispersion relation, is parallel to the 2 2 2
asymptote of the electron's dispersion relation. Thus, E = c p + m c
if we start from one point on the latter, a line parallel
to E = cp will never intersect (7.1.5) again. In other Ef
words, energy and momentum can not be conserved E = cp
simultaneously in vacuum.
279
pf pi p
7.1.2 Finite Length of Interaction
If we go back to (7.1.4) we observe that if the electron spends only a finite time in the
interaction region then it can experience a net electric field. Let us denote by −T the time
the electron enters the interaction region and by T the exit time. The average electric
field experienced by the electron (subject to the same assumptions indicated above) is
1 T
dt cos ωt (1 − vc ) ,
2T ∫−T
〈 E 〉 = E0
  v 
= E0 sinc ωT 1 −  ; (7.1.9)
  c 
here sinc( x ) = sin( x ) / x . That is to say that if the time the electron spends in the
interaction region, is small on the scale of the radiation period T0 = 2π / ω then the net
electric field it experiences, is not zero. From the perspective of the conservation laws,
the interaction is possible since although the energy conservation remains unchanged i.e.,
Ei = Ef + ω, (7.1.10)
the constraint on momentum conservation was released somewhat and it reads
280
E
Ei
E = c p 2 + m 2c 2
ω
| pi − pf − |< cT , (7.1.11) E
c f
which clearly is less stringent than in (7.1.8) as also illustrated
E = cp
in the figure; = 1.05457 ×10−34 Jsec is the Planck constant. The
operation of the klystron to be discussed subsequently relies on
the interaction of an electron with a wave in a region which is pi p
shorter than the radiation wavelength. pf
7.1.3 Finite Length Pulse

Another case where energy transfer is possible in vacuum is when the pulse duration
is short. In order to examine this case we consider, instead of a periodic wave whose
duration is infinite, a short pulse of a typical duration τ . In order to visualize the
configuration, consider a field given by
2 2
E x ( z , t ) = E0 e− ( t − z / c ) /τ . (7.1.12)
A particle following the same trajectory as in (7.1.3) will clearly E
experience an average electric field which is non-zero even when E i
the interaction duration is infinite. This is possible since the E = c p 2 + m 2c2
spectrum of the radiation field is broad -- in contrast to Sect. 7.1.1
where it was peaked -- therefore again the constraint of the Ef
conservation laws is less stringent:
E = cp
281
pf pi p

| Ei − Ef |< , (7.1.13)
τ
and

| pi − pf |< . (7.1.14)
cτ
Schematics of this mechanism is illustrated in the figure. It should be also pointed out
that in this section we consider primarily the kinematics of the interaction and we pay no
attention to the dynamics. In other words, we examined whether the conservation laws
can be satisfied without details regarding the field configuration.
7.1.4 Cerenkov Interaction
It was previously indicated that since the dispersion

relation of the photon is parallel to the asymptote of the E
electron's dispersion relation, the interaction is not possible
Ei
in an infinite domain. However, it is possible to change the
``slope'' of the photon, namely to change its phase velocity E = c p 2 + m2c2
-- see Figure. The easiest way to do so is by ``loading" the
medium where the wave propagates with a material whose Ef
dielectric coefficient is larger than one. Denoting the
E = cp
E = (c / n )p
282
pf pi p
refraction coefficient by n , the dispersion relation of the photon is given by
c
Eph = pph , (7.1.15)
n
while the dispersion relation of the electron remains unchanged. Substituting in the
expressions for the energy and the momentum conservation laws we find that the
condition for the interaction to occur is
c
= v, (7.1.16)
n
where it was assumed that the electron's recoil is relatively small i.e., ω / mc 2 1. The
result in (7.1.16) indicates that for the interaction to occur, the phase velocity of a plane
wave in the medium has to equal the velocity of the particle. This is the so-called
Cerenkov condition. Although dielectric loading is conceptually simple, it is not always
practical because of electric charges which accumulate on the surface and of a relatively
low breakdown threshold which is critical in high-power devices. For these reasons the
phase velocity is typically slowed down using metallic structures with periodic
boundaries. The operation of traveling wave tubes (or backward wave oscillators) relies
on this concept.
283
E
Ei
E = c p 2 + m 2c 2
7.1.5 Compton Scattering: Static Fields
Ef
E = cp
It is not only a structure with periodic boundaries
nh / L
which facilitates the interaction between electrons and
electromagnetic waves but also periodic fields. For example,
if a magneto-static field of periodicity L is applied on the pf pi p
electron in the interaction region, then this field serves as a
momentum ``reservoir" which can supply momentum quanta of n(2π / L) where
n = 0, ±1, ±2,... ; see Figure. The energy conservation law remains unchanged i.e.,
Ei = Ef + Eph , (7.1.17)
but the momentum is balanced by the applied static field
2π
pi = pf + pph + n. (7.1.18)
L
For a relativistic particle ( β 1) and when the electron's recoil is assumed to be small,
these two expressions determine the so-called resonance condition which reads
 2π c 
ω 2γ 2  n , (7.1.19)
 L 
where γ ≡ [1 − ( v / c )2 ]−1/ 2 . Note that the frequency of the emitted photon depends on the
284
velocity of the electron which means that by varying the velocity we can change the
operating frequency. A radiation source which possesses this feature is a tunable source.
Identical result is achieved if we assume a periodic electrostatic field and both field
configurations are employed in the so-called ``free electron lasers''.
7.1.6 Compton Scattering: Dynamic Fields
Static electric or magnetic field can be conceived as limiting cases of a dynamic

field of zero or vanishingly small frequency and we indicated above that they facilitate
the interaction between an electron and a wave. Consequently we may expect that the
interaction of an electron with a wave will occur in the presence of another wave. Indeed,
if we have an initial wave of frequency ω1 and the emitted wave is at a frequency ω2 the
conservation laws read
Ei + ω1 = Ef + ω2 , (7.1.20)
and
ω1 ω2
pi = pf + . + (7.1.21)
c c
Following the same procedure as above we find that the ratio between the frequencies of
the two waves is
285
E
E1
Ei
E2
ω2
4γ 2 , (7.1.22)
ω1 Ef
E = cp
which is by a factor of 2 larger than in the static case. The figure E = c p 2 + m 2c 2

illustrates this process.
pf pi p
7.1.7 Uniform Magnetic Field
A periodic magnetic field can provide quanta of momentum necessary to satisfy the
conservation law. It does not affect the average energy of the particle. An opposite
situation occurs when the electron moves in a uniform magnetic field ( B ): there is no
change in the momentum of the particle whereas its energy is given by
E = c p 2 + ( mc )2 − 2neB , (7.1.23)
where e = 1.6022 ×10 −19 C is the charge of the electron and n = 0, ±1, ±2... .
For most practical purposes the energy associated with the magnetic field is much
smaller than the energy of the electron therefore we can approximate
286
E = c p 2 + m 2 c 2 − 2 neBh / 2π
E
Ei
n −1
ec 2 B ec 2 B n
Ei − n1 = Ef − n2 + Eph , (7.1.24)
Ei Ef Ef
and the momentum conservation remains unchanged i.e., n +1
pi = pf + pph . (7.1.25)
E = cp
From these two equations we find that the frequency of the
emitted photon is
pf pi p
eB 2  eB 
ω = 2γ = 2γ  . (7.1.26)
m  mγ 
The last term is known as the relativistic cyclotron angular frequency, ωc,rel ≡ eB / mγ .
The figure illustrates schematically this type of interaction. It indicates that the dispersion
line of the electron is split by the magnetic field in many lines (index n ) and the
interaction is possible since the electron can move from one line to another. Gyrotron's
operation relies on this mechanism and it will be discussed briefly in the next section.
7.1.8 Synchronism Condition
All the processes in which the interaction of an electron with a monochromatic wave
extends to large regions, have one thing in common: the velocity of the electron has to
287
equal the effective phase velocity of the pondermotive wave along the electron's main
trajectory, namely
v = vph,eff . (7.1.27)
Here by pondermotive wave we mean the effective wave along the longitudinal trajectory
of the particle which accounts for transverse or longitudinal oscillation. In the Cerenkov
case we indicated that the phase velocity is c / n and there is no transverse motion
therefore the condition for interaction implies n = c / v where n is the refraction
coefficient of the medium. In the presence of a periodic static field the wavenumber of
the pondermotive wave is ω / c + 2π / L therefore
ω
vph,eff = , (7.1.28)
ω / c + 2π / L
and for a dynamic field
ω2 − ω1
vph,eff = , (7.1.29)
k2 + k1
where k1,2 = ω1,2 / c are the wavenumbers of the two waves involved. Finally, in a
uniform magnetic field only the effective frequency varies
ω − ωc,rel
vph,eff = . (7.1.30)
k
The reader can check now that within the framework of this formulation we obtain
288
(7.1.19) from (7.1.28), in the case of the dynamic field we have from (7.1.29) the 4γ 2
term as in (7.1.22) and finally (7.1.30) leads to the gyrotron's operation frequency
presented in (7.1.26).
289
7.2 Radiation Sources: Brief Overview
There are numerous types of radiation sources driven by electron beams. Our
purpose in this section is to continue the general discussion from the previous section and
briefly describe the operation principles of one ``member" of each class of what we
consider the main classes of radiation sources. A few comments on experimental work
will be made and for further details the reader is referred to recent review studies. The
discussion continues with the classification of the major radiation sources according to
several criteria which we found to be instructive.
7.2.1 The Klystron

The klystron was one of the first radiation sources to be developed. It is a device in
which the interaction between the particle and the wave is localized to the close vicinity
of a gap of a cavity, as illustrated in the Figure.
cathode
collector
input
output
290
Electrons move along a drift tube and its geometry is chosen in such a way that at
the frequency of interest it does not allow the electromagnetic wave to propagate. The
latter is confined to cavities attached to the drift tube. The wave which feeds the first
cavity modulates the velocity of the otherwise uniform beam. This means that after the
cavity, half of the electrons have a velocity larger than the average beam velocity
whereas the second half has a smaller velocity. According to the change in the (non-
relativistic) velocity of the electrons the beam becomes bunched down the stream since
accelerated electrons from one period of the electromagnetic wave catch up with the slow
electrons from the previous period. When this bunch enters the gap of another cavity it
may generate radiation very efficiently. In practice, several intermediary cavities are
necessary to achieve good modulation.
7.2.2 The Traveling Wave Tube
The traveling wave tube (TWT) is a Cerenkov device, namely the phase velocity of
the interacting wave is smaller than c and the interaction is distributed along many
wavelengths. Generally speaking, as the beam and the wave advance, the beam gets
modulated by the electric field of the wave and in turn, the modulated beam increases the
amplitude of the electric field. In this process both the beam modulation and the radiation
field grow exponentially in space.
291
cathode
collector
(a) input output
cathode output
output
(b)
input
cathode output
(c)
input
cathode
(d)
cathode
(e) 292
In the interaction process the electron oscillates primarily along the major axis ( z
direction ) and the interaction is with the parallel component of the electric field.
Correspondingly, the interaction occurs here with the transverse magnetic (TM) mode.
7.2.3 The Gyrotron
The gyrotron relies on the interaction between an annular beam, gyrating around the
axis of symmetry due to an applied uniform magnetic field, and a transverse electric (TE)
mode. The concept of generating coherent radiation from electrons gyrating in a magnetic
field was proposed independently by three different researchers in the late fifties, and it
has attracted substantial attention due to its potential to generate millimeter and
submillimeter radiation.
In this device electrons move in the azimuthal direction and they get bunched by the
corresponding azimuthal electric field. As in the case of the TWT the bunches act back
on the field and amplify it. In contrast to traveling wave tubes or klystrons in which the
beam typically interacts with the lowest mode, in the gyrotron the interaction is with high
modes therefore various suppression techniques are employed in order to obtain coherent
operation with a single mode.
The operation frequency is determined by the applied magnetic field, the energy of
the electrons and, in cases of high mode operation, also by the radius of the waveguide:
293
ω = ωcγ + γβ ωc2 + ωc2o , (7.2.1)
where β = v / c , ωc = eB / m and ωco is the cutoff frequency of the mode. The operating
frequency in this case can reach very high values: for a magnetic field of 1T and γ 2.5
the operation frequency is of the order of 150 GHz or higher according to the mode with
which the electrons interact.
Circuit Coil
First Anode
-350kV
Second Anode Open Cavity
Cathode
500kV
Gun Coil
Since the interaction of the electrons is with an azimuthal electric field, it is

necessary to provide the electrons with maximum momentum in this direction. The
parameter which is used as a measure of the injected momentum is the ratio of the
294
transverse to longitudinal momentum α ≡ v⊥ / vz . This transverse motion is acquired by
the electrons in the gun region as can be deduced from the schematics illustrated in the
Figure. In relativistic devices this ratio is typically smaller than unity whereas in non-
relativistic devices it can be somewhat larger than one.
Beam location is also very important. In the TWT case the interaction is with the
lowest symmetric TM mode. Specifically the electrons usually form a pencil beam and
they interact with the longitudinal electric field which has a maximum on axis. We
indicated that gyrotrons operate with high TE modes and the higher the mode, the higher
the number of nulls the azimuthal electric field has along the radial direction. Between
each two nulls there is a peak value of this field. It is crucial to have the annular beam on
one of these peaks for an efficient interaction to take place.
7.2.4 The Free Electron Laser

As the gyrotron, it is a fast-wave device in the sense that the interacting
electromagnetic wave has a phase velocity larger or equal to c but instead of a uniform
magnetic field it has a periodic magnetic field. The ``conventional" free electron laser
(FEL) has a magnetic field perpendicular to the main component of the beam velocity. As
a result, the electrons undergo a helical motion which is suitable for interaction with
either a TE or a TEM mode. The oscillation of electrons is in the transverse direction but
the bunching is longitudinal and in this last regard the process is similar to the one in the
295
traveling wave tube. However, its major advantage is the fact that it does not require a
metallic (or other type of) structure for the interaction to take place. Consequently, it has
the potential to either generate very high power at which the contact of radiation with
metallic walls would create very serious problems, or produce radiation at UV, XUV or
X-ray where there are no other coherent radiation sources. The Figure illustrates the basic
configuration.
cathode S N S N S N S N
N S N S N S N S
296
7.2.5 The Magnetron
v
The magnetron was invented at the beginning
of the 20th century but because of its complexity B
there is no analytical model, as yet, which can
describe its operation adequately as a whole. In
recent years great progress has been made in the
understanding of the various processes with the aid
of particle in cell (PIC) codes. Its operation combines potential and kinetic energy
conversion. The Figure illustrates the basic configuration. Electrons are generated on the
cathode (inner surface) and since a perpendicular magnetic field is applied they form a
flow which rotates azimuthally. The magnetic field and the voltage applied on the anode
are chosen in such a way that, in equilibrium, the average velocity of the electrons equals
the phase velocity of the wave supported by the periodic structure at the frequency of
interest.
A simplistic picture of the interaction can be conceived in the following way:
electrons which lose energy to the wave via the Cerenkov type interaction, move in
upward trajectories -- closer to the anode. Consequently, two processes occur. Firstly, the
closer the electron is to the periodic surface the stronger the radiation field and therefore
297
the deceleration is larger, causing a further motion upwards. Secondly, as it moves
upwards its (dc) potential energy varies. Again, this is converted into electromagnetic
energy.
Two major differences between the magnetron and other radiation sources
mentioned above, are evident: (i) in the magnetron the beam generation, acceleration and
collection occur all in the same region where the interaction takes place. (ii) The
potential energy associated with the presence of the charge in the gap plays an important
role in the interaction; the other device where this is important is the vircator which will
be briefly discussed next.
298
7.3 Generation of radiation in a waveguide
In this subsection we consider the electromagnetic field associated with the

symmetric transverse magnetic (TM) mode in a dielectric filled waveguide. As in the
previous subsection, the source of thiss field is a particle moving at a velocity v0 ,
however, the main difference is that the solution has a constraint since on the waveguide's
wall ( r = R ) the tangential electric field vanishes. Therefore, we shall calculate the Green
function in the frequency domain subject to the condition G ( r = R, z | r' , z' ) = 0 . We
assume a solution of the form
∞
 r
G ( r , z | r , z ) = ∑Gs ( z | r' , z' )J 0  ps ,
' '
(7.3.1)
s =1  R
substitute in (2.4.11) and use the orthogonality of the Bessel functions we find that
 r'  1
Gs ( z | r , z ) = J 0  p s 
' '
g s ( z | z' ), (7.3.2)
1
 R  R2 J2 ( p )
1 s
2
where g s ( z | z' ) satisfies
 d2 2 1
 dz 2 − Γ s s g ( z | z '
) = − δ ( z − z' ), (7.3.3)
  2π
and
299
ps2 ω2
Γ = 2 − εr 2 .
2
s (7.3.4)
R c
'
For z > z the solution of (7.3.3) is
−Γ s ( z − z' )
g s ( z | z ) = A+ e
'
, (7.3.5)
and for z < z' the solution is
Γ s ( z − z' )
g s ( z | z ) = A− e
'
. (7.3.6)
Green's function is continuous at z = z' i.e.,
A+ = A− , (7.3.7)
and its first derivative is discontinuous. The discontinuity is determined by integrating
(7.3.3)from z = z' − 0 to z = z' + 0 i.e.,
d '  d '  1
g
 dz s ( z | z )  ' − g
 dz s ( z | z )  ' = − . (7.3.8)
z = z +0 z = z −0 2π
Substituting the two solutions introduced above, and using (7.3.7)we obtain
1 −Γs | z − z' |
gs ( z | z ) =
'
e . (7.3.9)
4πΓ s
Finally, the explicit expression for the Green's function corresponding to azimuthally
symmetric TM modes in a circular waveguide is given by
300
∞
J 0 ( ps r / R )J 0 ( ps r' / R ) 1 −Γs | z − z' |
G(r, z | r , z ) = ∑
' '
e . (7.3.10)
s =1
1 2 2
R J 1 ( ps )
4πΓ s
2
In this expression it was tacitly assumed that ω > 0 and Γ s [defined in (7.3.4)] is non-
zero.
With Green's function established, we can calculate the magnetic vector potential as
generated by the current distribution described in (2.4.10); the result is
R ∞
Az ( r , z , ω ) = 2πµ0 ∫ drr ∫ dz' G ( r , z | r' , z' ) J z ( r' , z' )
0 −∞
eµ ∞
J 0 ( ps r / R ) 2
∑
− j ( ω / v0 ) z
= − 02 e . (7.3.11)
8π s =1
1
R 2 J12 ( ps ) Γ s + ω / v0
2 2 2
2
It will be instructive to examine this expression in the time domain; the Fourier transform
is
e β2
Az ( r , z , t ) = − 2
2π ε 0 R 2 1 − n 2 β 2
∞ jω ( t − z / v0 )
J ( p r / R) ∞ e
×∑ 0 2 s ∫ dω , (7.3.12)
s =1 J 1 ( ps ) −∞ ω 2 + Ω 2s
301
where
2
 p c
2
β
2
Ω = s  . (7.3.13)
 R  1− n β
s 2 2
The problem has been now simplified to the evaluation of the integral
∞ e jωτ
Fs (τ = t − z / v0 ) ≡ ∫ d ω 2 , (7.3.14)
−∞ ω + Ωs 2
which in turn is equivalent to the solution of the following differential equation

 d2 2
 dτ 2 −Ω s  Fs (τ ) = −2πδ (τ ). (2.4.39)
 
If the particle's velocity is smaller than the phase velocity of a plane wave in the medium
( nβ < 1 ) then Ω 2s > 0 and the solution for τ > 0 is
−Ω sτ
Fs (τ > 0) = A+ e , (7.3.15)
or
Ω sτ
Fs (τ < 0) = A− e . (7.3.16)
As previously, in the case of Green's function, Fs (τ ) has to be continuous at τ = 0

and its derivative is discontinuous:
302
 d   d 
 F (τ )  − Fs (τ )  = −2π . (7.3.17)
 dτ τ =0+  dτ
s
τ =0−
When the velocity of the particle is smaller than c / n (i.e., nβ < 1 ) the characteristic
frequency Ω s is real, therefore
π −Ω s |τ |
Fs (τ ) = e , (7.3.18)
Ωs
and
e β2 ∞
J 0 ( ps r / R ) −Ω s |t − z / v0 |
Az ( r , z , t ) = − ∑
2πε 0 R 2 1 − β 2 n 2
s =1 J1 ( ps )Ω s
2
e . (7.3.19)
This expression represents a superposition of evanescent modes attached to the particle. It

is important to emphasize that since the phase velocity of a plane wave in the medium is
larger than the velocity of the particle there is an electromagnetic field in front ( τ < 0 ) of
the particle. The situation is different in the opposite case, β > 1 / n , since Ω 2s < 0 . In this
case the waves are slower than the particle and there is no electromagnetic field in front
of the particle i.e.,
Fs (τ < 0) = 0. (7.3.20)
By virtue of the continuity at τ = 0 we have for τ > 0
Fs (τ > 0) = A+ sin (| Ω s | τ ). (7.3.21)
303
Substituting these two expressions in (2.4.42) we obtain
2π
Fs (τ ) = − sin (| Ω s | τ )h(τ ), (7.3.22)
| Ωs |
and the magnetic vector potential reads
e β2
Az ( r, z, t ) = −
πε 0 R 2 n 2 β 2 − 1
J 0 ( ps r / R )   z   z 
×∑ 2 sin | Ω s |  t −   h  t − , (7.3.23)
J
s =1 1 ( p s ) | Ω s |   v0   v0 
where h(ξ ) is the Heaviside step function. This expression indicates that when the
velocity of the particle is larger than c / n , there is an entire superposition of propagating
waves traveling behind the particle. Furthermore, all the waves have the same phase
velocity which is identical with the velocity of the particle, v0 . It is important to bear in
mind that this result was obtained after tacitly assuming that ε r is frequency independent
which generally is not the case, therefore the summation is limited to a finite number of
modes. The modes which contribute are determined by the Cerenkov condition
n(ω = Ω s ) β > 1.
After we established the magnetic vector potential, let us now calculate the average
power which trails behind the particle. Firstly, the azimuthal magnetic field is given by
304
1 ∂
Hϕ ( r , z , t ) = − Az ( r , z , t )
µ 0 ∂r
1 ps  r 
=
µ0
∑As
s =1
J 1  ps 
R  R
  z   z 
× sin | Ω s |  t −  h  t − , (7.3.24)
  v0    v0 
where
e β2 1
As = − . (7.3.25)
πε 0 R 2 n 2 β 2 − 1 J12 ( ps ) | Ω s |
Secondly, the radial electric field is determined by the electric scalar potential which in
turn is calculated using the Lorentz gauge and it reads
305
∂
Er ( r , z , t ) = − Φ( r , z , t )
∂r
c2 ps  r 
= ∑ A J p
ε r v0 s =1 s R 1  s R 
  z   z 
× sin | Ω s |  t −   h  t − . (7.3.26)
  v0    v0 
With these expressions we can calculate the average electromagnetic power trailing the
particle. It is given by
e2 β c 1 1
P= ∑
2πε 0ε r R 2 ε r β 2 − 1 s =1 J12 ( ps )
. (7.3.27)
Note that for ultra relativistic particle ( β → 1) the power is independent of the particle's
energy. In order to have a measure of the radiation emitted consider a very narrow bunch
of N ∼ 1011 electrons injected in a waveguide whose radius is 9.2 mm. The waveguide is
filled with a material whose dielectric coefficient is ε r 2.6 and all electrons have the same
energy 450 keV. If we were able to keep their velocity constant, then 23 MW of power at
11.4 GHz (first mode, s = 1 ) will trail the bunch. Further examining this expression we
note that the average power is quadratic with the frequency i.e.,
306
( Ne)2 | Ω s |2
P ≡ ∑Ps = ∑ . (7.3.28)
s =1 2πε 0ε r β c s =1 [ ps J1 ( ps )] 2
In addition, based on the definition of the Fourier transform of the current density in
(2.4.10), we conclude that the current which this macro-particle excites in the s th mode
is I s eN Ω s / 2π . With this expression, the radiation impedance of the first mode ( s = 1 ) is
P1 4π
RC,1 = = η0 . (7.3.29)
ε r β [ p1J1 ( p1 )]
2
1
| I1 |2
2
For a relativistic particle, β 1 , a dielectric medium ε r = 2.6 the radiation impedance
corresponding to the first mode is 1200Ω which is one order of magnitude larger than
that of a dipole in free space or between two plates. Note that this impedance is
independent of the geometry of the waveguide and for an ultra-relativistic particle it is
independent of the particle's energy.
307
7.4 Generation of radiation in a cavity
In order to examine transient phenomena associated with reflected waves we shall

calculate the electromagnetic energy in a cavity as a single (point) charge traverses the
structure. Consider a lossless cylindrical cavity of radius R and length d . A charged
particle ( e) moves along the axis at a constant velocity v0 . Consequently, the longitudinal
component of the current density is the only non-zero term, thus
1
J z (r , t ) = −ev0 δ ( r )δ ( z − v0 t ). (7.4.1)
2π r
It excites the longitudinal magnetic vector potential Az (r , t ) which for an azimuthally
symmetric system satisfies
1 ∂ 1 ∂2 1 ∂2 
 r ∂r r ∂r + ∂z 2 − c 2 ∂t 2  Az ( r , z, t ) = − µ0 J z ( r , z , t ). (7.4.2)
 
In this section we shall consider only the internal problem, ignoring the electromagnetic
phenomena outside the cavity. The boundary conditions on the internal walls of the
cavity impose that Ez ( r = R, z , t ) = 0 , Er ( r , z = 0, t ) = 0 and Er ( r , z = d , t ) = 0 therefore
the magnetic vector potential reads
∞
 r πn 
Az ( r , z , t ) = ∑ As ,n (t )J 0  ps  cos  z . (7.4.3)
s =1,n =0  R   d 
308
Using the orthogonality of the trigonometric and Bessel functions we find that the
amplitude As ,n (t ) satisfies
 d2 2  ev0 1 1
 dt 2 + Ω A
s , n  s ,n ( t ) − −
  2πε 0 1 R 2 J 2 ( p ) g n d
1 s
2
πn   v 
× cos  v0 t   h(t ) − h  t − 0  , (7.4.4)
 d   d 
where
 1 for n = 0,
gn  (7.4.5)
0.5 otherwise,
and
2 2
 p  πn 
Ω s ,n = c  s  +   , (7.4.6)
  
R d 
are the eigen-frequencies of the cavity. Before the particle enters the cavity ( t < 0 ), no
field exists, therefore
As ,n (t < 0) = 0. (7.4.7)
For the time the particle is in the cavity namely, 0 < t < d / v0 , the solution of (7.4.4)
309
consists of the homogeneous and the excitation term:
 d 
As ,n  0 < t <  = B1 cos(Ω s ,n t ) B2 sin(Ω s ,n t )
 v0 
+α s ,n cos(ωn t ), (7.4.8)
where
ev0 1 1 1
α s ,n = − , (7.4.9)
2πε 0 R 2 J 2 ( p ) g n d Ω s ,n − ωn
1 2 2
1 s
2
and
πn
ωn = v0 . (7.4.10)
d
Since both the magnetic and the electric field are zero at t = 0 , the function As ,n (t ) and its
first derivative are zero at t = 0 hence
B1α s ,n = 0, (7.4.11)
and
B2 = 0. (7.4.12)
Consequently, the amplitude of the magnetic vector potential [ As ,n (t ) ] reads
As ,n (t ) = α s ,n cos(ωn t ) − cos(Ω s ,n t ) . (7.4.13)
310
Beyond t = d / v0 , the particle is out of the structure thus the source term in (7.4.4) is zero
and the solution reads
 d    d 
As ,n  t >  = C1 cos Ω s ,n  t −  
 v0    v0  
  d 
+ C2 sin Ω s ,n  t −  . (7.4.14)
  v0 
As in the previous case, at t = d / v0 both As ,n (t > d / v0 ) and its derivative, have to be

continuous:
  d 
α s ,n ( −1)n − cos  Ω s ,n   = C1 , (7.4.15)
  v0  
 d 
α s ,n Ω s ,n sin  Ω s ,n  = C2 Ω s ,n . (7.4.16)
 v0 
For this time period, the explicit expression for the magnetic vector potential is
311
 d    d    d 
As ,n  t >  = α s ,n ( −1)n − cos  Ω s ,n   cos Ω s ,n  t −  
 v0    v0     v0  
  d 
+α s ,n sin ( Ω s ,n dv0 ) sin Ω s ,n  t −  , (7.4.17)
  v0  
The expressions in (7.4.7), (7.4.13), (7.4.17) describe the magnetic vector potential in the
cavity at all times. The Figure illustrates schematically this solution.
During the period the electron spends in the d

cavity, there are two frequencies which are v0 R
excited: the eigen-frequency of the cavity Ω s ,n and r
the ``resonances" associated with the motion of the z
particle, ωn . The latter set corresponds to the case
when the phase velocity, vphω / k , equals the d
velocity v0 . Since the boundary conditions impose R v0
kπ n / d and the resonance implies
r
 ω2 d 
v0 = vph = c  2 , (7.4.18) z
 c πn 
d
R v0
312
r
z
thus we can immediately deduce the resonance frequencies ωn as given in (7.4.10).
Now that the magnetic vector potential has been determined, we consider the effect
of the field generated in the cavity on the moving particle. The relevant component is
 d  r π 
Az  r , z ,0 < t <  = ∑ α s ,n J 0  ps  × cos  z  cos (ωn t ) − cos(Ω s ,n t ) . (7.4.19)
 v0  s =1,n=0  R  nd 
Note that the upper limit in the double summation was omitted since in practice this limit
is determined by the actual dimensions of the particle, which so far was considered
infinitesimally small. In order to quantify this statement we realize that the summation is
over all eigenmodes which have a wavenumber much longer than the particle's dimension
i.e., Ω s , n Rb / c < 1 .
According to Maxwell's equations, the longitudinal electric field is
∂ 1 ∂
ε 0 E z (r , t ) = − J z (r , t ) + rHϕ (r , t ). (7.4.20)
∂t r ∂r
Furthermore, the field which acts on the particle does not include the self field, therefore
we omit the current density term. Using the expression for the magnetic vector potential
[(2.1.36)], we have
1 ∂ ∂
Ez (r , t ) = −c 2 ∫dt r Az (r , t ), (7.4.21)
r ∂r ∂r
or explicitly,
313
π n  sin(ωn t ) sin(Ω s ,n t ) 
2
 d   cps   r 
Ez  r , z ,0 < t <  = ∑ α s ,n  J
 0 s  p × cos( z)  − .
 v0  s =1,n =0  R   R d  ωn Ω s ,n 
(7.4.22)
In a lossless and closed cavity the total power flow is zero, therefore Poynting's theorem
in its integral form reads
dW R d
= −2π ∫ drr ∫ dzEz ( r , z , t )J z ( r, z, t ). (7.4.23)
dt 0 0
Thus substituting the current density [(7.4.1)] we obtain

d / v0
W = ev0 ∫ dtEz ( r , z = v0 t , t ), (7.4.24)
0
which has the following explicit form
2
 cps  d / v0  sin(ωn t ) sin(Ω s ,n t ) 
W = ev0 ∑ α s ,n   ∫0 dt cos(ωn t ) ×  − . (7.4.25)
s =1,n =0  R   ω n Ω s ,n 
The time integral in this expression can be evaluated analytically. As can be readily
deduced, the first term represents the non-homogeneous part of the solution and its
contribution is identically zero whereas the second's reads
 cps  1 − ( −1) cos(Ω s ,n d / v0 )
2 n
W = −ev0 ∑ α s ,n   . (7.4.26)
s =1,n =0  R  Ω s ,n − ωn
2 2
314
Substituting the explicit expression for α s , n we have
−1
 e2 
W ≡W  
 4πε d
0 
2
 2 ps  1 1   Ω s ,n  
= ∑   × 1 − ( −1) cos 
n
d   .(7.4.27)
2 2
 ps + (π nR / d γ )  
J
s =1,n =0  1 ( p )
s  g n
2 v
 0 
In the Figure we illustrate two typical terms from the expression above as a function of
the particle's momentum. The (normalized) energy stored at 10.7 GHz (corresponding to
s = 1, n = 1) is shown in the left frame and we observe that for γβ = 2.5 the energy
reaches its asymptotic value [W ( s = 1, n = 1) 4 ]. This is in contrast to the energy stored
in the 35.5 GHz ( s = 3,n = 3 ) wave
4.0 R = 1.5 cm 4.0 R = 1 .5 cm
which at the same momentum
reaches virtually zero level; the d = 2.0 cm d = 2.0 cm
asymptotic value 3.0 3.0
W (3,3)
[W ( s = 3, n = 3) 0.5] is reached
W (1,1)
for a much higher momentum 2.0 2.0

(γβ = 15 ).
1.0 1.0
0.0 0.0
0.0 1.0 2.0 0.0 1.0 2.0
315 γβ γβ

Microwaves Locked PDF

Uploaded by

Document Information

Original Title

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

Microwaves Locked PDF

Uploaded by

Copyright:

Available Formats

Department of Electrical Engineering

Technion – Israel Institute of Technology

based upon lectures delivered by

3 Waveguides – Advanced Topics 114

4 Matrix Formulations 211

5 Nonlinear Components 237

6 Periodic Structures 241

1.2 Coaxial Transmission Line

As indicated in the previous case, two parameters are

Exercise 1.2: Prove the relations in Eq. (1.3.8).

The fundamental assumptions of the analysis are:

Since by virtue of linearity of Maxwell's equations

Another quantity that warrants consideration is the average power

There are cases when either the electromagnetic properties or the

Microwave or high frequency circuits consist

In this section we shall discuss in some detail

(a) (b) (c)

(a) Characteristic impedance vs. ∆ ; h = 2 mm, w = 20 mm, ε r = 10 and ν < 100 .

(a) (b) (c)

and the orthogonality of the trigonometric function we obtain

In the case of a continuous distribution we should replace

Left: Characteristic impedance vs. the width ∆ [ h = 2 mm, ε r = 10 and n = 0,1,...200 ]

1.9.1 Short Recapitulation

since the second term in the denominator is very small. Now

1.10.1 Semi-Infinite Structure

V ( z , t ) = ∑aν exp  jων t − ε (ων )   ∫ dω exp  j (ω − ων )  t −  −   

Comment: Discuss possibilities of spectrum measurement by its decomposition in space.

The reflection and transmission coefficients in the frequency-domain are

Based on this expression, the transmitted signal is

parameters and 6.8

Compare micro-strip and strip-line from the perspective of sensitivity to the

The characteristic impedance of a

(stripline), ω = 20 mm, n = 1,2, 200 , and 20 Microstrip

The derivative of the normalized phase velocity of 0

2.1 General Formulation

So far we have examined the propagation of electromagnetic waves in a structure

a) The electromagnetic characteristics of the medium: µ = µ0 µ r and ε = ε 0ε r .

v v (2.1.2)

It is convenient at this point to define the transverse wavenumber

Complete Solution = ( Ez = 0) and( H z =/ 0) + ( Ez =/ 0) and ( H z = 0) .

Exercise 2.2: What is the impact of the geometry?

no walls were present i.e. ω  kz v .

The second possible solution according to (2.1.12)--(2.1.13) is when Ez = 0 and

2.4.1 Power Flow

Exercise 2.5: What is the power at any particular point in time?

2.4.2 Ohm Loss

where δ ≡ 2 / ωµoσ , Rs ≡ ωµo / 2σ and the integration is over the circumference of

Let us denote by kz (0) the wavenumber in a lossless waveguide. Subject to the

2.4.3 Dielectric Loss

ω for large frequencies. In between there is a minimum loss for an optimal

Mode comparison for a rectangular waveguide.

2.6.1 Transverse Magnetic (TM) Mode [ H z = 0]

In this section we shall investigate the propagation of a wave in a cylindrical

s=1 s=2 s=3

Substituting (2.6.3) in (2.6.2) we obtain

In this case the wave equation reads

Ps ,n =  RAs ,n   R  Re  R 2 − s 2,n  π g n  n s ,n  . (2.6.17)

1000 5000 9000 13000

Mode comparison for cylindrical waveguide.

wherein Γ s = ps / R 2 − ω 2 / c 2 and Es (ω ) is the Fourier transform of this field component

the power propagating is

and the energy associated with this power

Its numerical analysis reveals that

(stripline), ω = 20 mm, n = 1,2, 200 , and 20 Microstrip

no walls were present i.e. ω kz v .

Further simplification, is achieved assuming that R d we get