Progress In Electromagnetics Research, PIER 59, 267297, 2006

A TIME-DOMAIN THEORY OF WAVEGUIDE

W. Geyi
Research In Motion
295 Phillip Street, Waterloo, Ontario, Canada N2L 3W8
AbstractA new time-domain theory for waveguides has been
presented in the paper. The electromagnetic elds are rst expanded
by using the complete sets of vector modal functions derived from
the transverse electric eld. The expansion coecients are then
determined by solving inhomogeneous Klein-Gordon equation in
terms of retarded Greens function. The theory has been validated
by considering propagation problems excited by various excitation
waveforms, which indicates that the higher order modes play a
signicant role in the eld distributions excited by a wideband signal.
1. INTRODUCTION
In recent years the major advances made in ultrawideband techniques
have made time-domain analysis of electromagnetic phenomena an
important research eld. A short pulse can be used to obtain
high resolution and accuracy in radar and to increase information
transmission rate in communication systems. Another important
feature of a short pulse is that its rate of energy decay can be slowed
down by decreasing the risetime of the pulse [1].
According to linear system theory and Fourier analysis, the
response of the system to an arbitrary pulse can be obtained by
superimposing its responses to all the real frequencies. In other words,
the solution to a time domain problem can be expressed in terms of a
time harmonic solution through the use of the Fourier transform. This
process can be assisted by the fast Fourier transforms and has been
used extensively in studying the transient response of linear systems.
This procedure is, however, not always most eective and is not a
trivial exercise since the harmonic problem must be solved for a large
range of frequencies and only an approximate time harmonic solution
valid over a nite frequency band can be obtained. Another reason is
that the original excitation wave may not be Fourier transformable.
268 Geyi
Thus we are forced to seek a solution in the time domain in some
situations.
In high-speed circuits the signal frequency spectrum of a
picosecond pulse may extend to terahertz regime and signal integrity
problems may occur, which requires a deep understanding of the
propagation characteristics of the transients in a waveguide. The
waveguide transients have been studied by many authors [224]. One
of the research topics is to determine the response of the waveguide
to an arbitrary input signal. If the input signal of a waveguide in a
single-mode operation is x(t), then after traveling a distance z, the
output from the waveguide is given by the Fourier integral
y(t) =
1
2

X()e
j[tz]
d (1)
where X() is the Fourier transform of x(t); is the propagation
constant =
_

2
c
; and
c
is the cut-o frequency of the
propagating mode and c = ()
1/2
with and being the
permeability and permitivity of the medium lling the waveguide
respectively.
Several methods have been proposed to evaluate the Fourier
integral in (1), such as the method of saddle point integration [2],
method of stationary phase [3, 4], contour integration technique [5],
and the quadratic approximation of the propagation constant around
the carrier frequency [69]. A serious drawback to stationary phase
is that it contradicts the physical realizability [10] as well as causality
[13] (i.e., the response appears before the input signal is launched). A
more rigorous approach is based on impulse response function for a
lossless waveguide, which is dened as the inverse Fourier transform
of the transfer function e
jz
. The impulse response function can
be expressed as an exact closed form and has been applied to study
transient response of waveguide to various input signals [1216].
The response given by (1) is, however, hardly realistic for
describing the propagation of a very short pulse or an ultrawideband
signal since it is based on an assumption that the waveguide is in a
single-mode operation. This assumption is reasonable only for a narrow
band signal but not valid for a short pulse, which covers a very wide
range of frequency spectrum, and will excite a number of higher order
modes in the waveguide. It should be mentioned that most authors
have discussed transient responses for various input signals, such as a
step function, a rectangular pulse or even a impulse on the basis of
an assumption that the waveguide is in a single-mode operation. This
kind of treatment has oversimplied the problem and the theoretical
Progress In Electromagnetics Research, PIER 59, 2006 269
results obtained cannot accurately describe the real transient process
in the waveguide.
In order to get the real picture of the transient process in an
arbitrary waveguide, we must solve the time-domain Maxwell equations
subject to initial conditions, boundary conditions and excitation
conditions, and we should include the higher order mode eects.
One such approach is based on the Greens function of Maxwell
equations and has been discussed in [20]. Another approach is
based on the eld expansions in terms of the eigenfunctions in the
waveguide [2124], which are usually derived from the eigensolutions
of the longitudinal elds. When these eld expansions are introduced
into homogeneous Maxwell equations one nds that the expansion
coecients satisfy the homogeneous Klein-Gordon equations. A wave
splitting technique has been used to solve the homogeneous Klein-
Gordon equations by introducing two Greens functions G
+
(z, t) and
G

(z, t), which satisfy an integro-dierential equation [21], and the

excitation problem in a waveguide has been studied by introducing the
time-domain vector mode functions, each of which is time dependent
and satises the homogeneous Maxwell equations. The total elds
generated by the source can then be expressed as an innite sum
of convolutions of expansion coecients and the time-domain vector
mode functions. The approach based on wave splitting seems rigorous
but very complicated. In addition it is based on the eld expansions
in terms of the longitudinal eld components. Consequently the
theory is only suitable for a hollow waveguide and cannot be applied
to transverse electromagnetic (TEM) transmission lines whose cross
section is multiple connected.
However a hollow waveguide is not suitable for carrying a
wideband signal since it blocks all the low frequency components
below the cut-o frequency of dominant mode, which may cause severe
distortion. For this reason, a hollow waveguide is only appropriate
for narrow band applications where the frequency spectrum of the
transmitted signal falls in between the cut-o frequency of the
dominant mode and the cut-o frequency of the next higher order
mode. On the other hand the cut-o frequency of the dominant mode
of a TEM transmission line is zero. Therefore it is often used for
wideband applications.
Hence it is necessary to develop a more general time-domain
theory for the waveguide, which is suitable for hollow waveguides as
well as TEM transmission lines. It is the purpose of present paper to
achieve this goal. A more straightforward and concise time-domain
approach for the analysis of waveguides has been presented in this
paper. Instead of starting from eigenvalue theory of the longitudinal
270 Geyi
eld components, we use the eigenvalue theory of the transverse
electric eld. The eigenfunctions of this transverse eigenvalue problem
constitute a complete set and can be used to expand all eld
components in an arbitrary waveguide (hollow or multiple connected).
A set of inhomogeneous Klein-Gordon equations for the eld expansion
coecients can be obtained when these eld expansions are substituted
into Maxwell equations. As a result, the excitation problems may
be reduced to the solution of these inhomogeneous scalar equations.
On the contrary, all the previous publications deal with homogeneous
Klein-Gordon equations and leave the excitation problem in the
waveguide very complicated [21]. To solve the inhomogeneous Klein-
Gordon equation, the Greens function method has been used. The new
time-domain theory has been applied to study the transient process in
both hollow waveguide and TEM transmission line with the eects of
higher order mode being taken into account. It should be mentioned
that, as far as the author can determine, there is no a single numerical
example in previous publications that has really considered the higher
order mode eects when dealing with the transients in a waveguide.
For the purpose of validating the new time-domain theory for the
waveguide, a typical excitation problem by a sinusoidal wave, which is
turned on at t = 0, has been investigated. Our theoretical prediction
shows that the steady state response of the eld distribution tends to
the well-known result derived from time-harmonic theory as t .
Other excitation waveforms are also discussed, which indicates that
the higher order modes play a signicant role in the eld distributions
excited by a wideband signal. To further validate the theory three
appendices have been attached. Appendix A gives a simple proof of the
completeness of the transverse eigenvalue problem for the waveguide.
Appendices B and C investigate the numerical examples studied in
Section 4 by directly solving Maxwell equations subject to the initial
and boundary conditions in the waveguide. It is shown that both
approaches give the exactly same results.
2. FIELD EXPANSIONS BY EIGENFUNCTIONS
Consider a perfect conducting waveguide, which is uniform along z-axis
and lled with homogeneous and isotropic medium. The cross-section
of the waveguide is denoted by and the boundary of by . Note
that the cross-section can be multiple connected. The transient
electrical eld in a source free region of the waveguide will satisfy the
Progress In Electromagnetics Research, PIER 59, 2006 271
following equations
_

2
E(r, t)
2
E(r, t)/c
2
t
2
= 0, r
E(r, t) = 0, r
u
n
E(r, t) = 0, r
(2)
where u
n
is the outward normal to the boundary . The solution
can then be expressed as the sum of a transverse component and a
longitudinal component, both of which are the separable functions of
transverse coordinates and the longitudinal coordinate with time, i.e.,
E(r, t) = (e
t
+u
z
e
z
)u(z, t)
where e
t
and e
z
are functions of transverse coordinates only. For
a hollow waveguide (i.e., the cross-section is simple connected),
the transverse elds may be expressed in terms of the longitudinal
elds. Therefore the analysis for the hollow waveguide can be based
on the longitudinal elds and this is the usual way to study the
transient process in the hollow waveguide [2124]. If the cross-section
of the waveguide is multiple connected, such as multiple conductor
transmission line for which the dominant mode has no longitudinal
eld components at all, the above analysis is no longer valid. To ensure
the theory to be applicable to the general situations we may use the
transverse elds. Inserting the above equation into (2) and taking the
boundary conditions into account, we obtain
_
e
t
e
t
k
2
c
e
t
= 0, r
u
n
e
t
= e
t
= 0, r
where k
2
c
is the separation constant. The above system of equations
constitutes an eigenvalue problem and has been studied by Kurokawa
[25]. It has been be shown that the system of eigenfunctions {e
tn
|n =
1, 2, . . .} or vector mode functions is complete in the product space
L
2
() L
2
(), where L
2
() stands for the Hilbert space consisting
of square-integrable real functions. Kurokawas approach is rigorous
enough for engineering purposes although he does not provide the proof
of the existence of eigenfunctions of the above eigenvalue problem.
In Appendix A of present paper, a more rigorous proof on the
completeness of the eigenfunctions of the above eigenvalue problem
has been presented.
The corresponding eigenvalues {k
2
cn
0, k
cn+1
k
cn
|n =
1, 2, . . .} are the cut-o wavenumbers. These mode functions can
be arranged to fall into three dierent categories (1) TEM mode,
e
tn
= 0, e
tn
= 0; (2) TE mode, e
tn
= 0, e
tn
= 0;
272 Geyi
and (3) TM mode, e
tn
= 0, e
tn
= 0. It should be noted that
only TEM modes correspond to a zero cut-o wavenumber. The cut-
o wavenumbers for TE and TM modes are always great than zero.
From the complete set {e
tn
} other three complete systems may be
constructed
{u
z
e
tn
|u
n
u
z
e
tn
= 0, r }
{ e
tn
/k
cn
| e
tn
/k
cn
= 0, r }
{u
z
(e
tn
/k
cn
), c|u
n
[u
z
(e
tn
/k
cn
)] = 0, r }
where c is a constant. Hereafter we assume that the vector mode
functions are orthonormal, i.e.,
_

e
tm
e
tn
d =
mn
. If the set
{e
tn
} is orthonormal, then all the above three complete sets are also
orthonormal. According to the boundary conditions that the vector
mode functions must satisfy, {e
tn
} are electric eld-like and most
appropriate for the expansion of the transverse electric eld; {u
z
e
tn
}
are magnetic eld-like and most appropriate for the expansion of
the transverse magnetic eld; { e
tn
/k
cn
} are electric eld-like
and most appropriate for the expansion of longitudinal electric eld;
{ e
tn
/k
cn
} are magnetic eld-like and most appropriate for the
expansion of longitudinal magnetic eld. Notice that E is magnetic
eld-like while H is electric eld-like. It should be notied that
all the modes and cut-o wavenumbers are independent of time and
frequency and they only depend on the geometry of the waveguide.
Therefore these modes can be used to expand the elds in both
frequency and time domain. Following a similar procedure described
in [25] the transient electromagnetic elds in the waveguide can be
represented by
E =

n=1
v
n
e
tn
+u
z

n=1
_
e
tn
k
cn
_
e
zn
(3)
H =

n=1
i
n
u
z
e
tn
+
u
z

u
z
H

d +

n=1
_
e
tn
k
cn
_
h
zn
(4)
E =

n=1
_
v
n
z
+ k
cn
e
zn
_
u
z
e
tn
+

n=1
k
cn
v
n
_
e
tn
k
cn
_
(5)
H =

n=1
_

i
n
z
+ k
cn
h
zn
_
e
tn
+u
z

n=1
k
cn
i
n
_
e
tn
k
cn
_
(6)
where
v
n
(z, t) =
_

E e
tn
d, i
n
(z, t) =
_

H u
z
e
tn
d
Progress In Electromagnetics Research, PIER 59, 2006 273
h
zn
(z, t) =
_

H
_
e
tn
k
cn
_
d, e
zn
(z, t) =
_

u
z
E
_
e
tn
k
cn
_
d
We will call v
n
and i
n
the time-domain modal voltage and time-domain
modal current respectively. Substituting the above expansions into the
time-domain Maxwell equations
_

_
E(r, t) =
H(r, t)
t
J
m
(r, t)
H(r, t) =
E(r, t)
t
+J(r, t)
and comparing the transverse and longitudinal components, we obtain

i
n
z
+ k
cn
h
zn
=
v
n
t
+
_

J e
tn
d (7)
k
cn
i
n
=
e
zn
t
+
_

u
z
J
_
e
tn
k
cn
_
d, for TM modes (8)
v
n
z
+ e
zn
k
cn
=
i
n
t

_

J
m
u
z
e
tn
d (9)
k
cn
v
n
=
h
zn
t

_

(u
z
J
m
)
_
u
z
e
tn
k
cn
_
d, for TE modes
(10)

t
_

H u
z

d =
_

u
z
J
m

d, for TE modes (11)

It is easy to show that the modal voltage and modal current for TEM
mode satisfy the one-dimensional wave equation
_

2
v
TEM
n
z
2

1
c
2

2
v
TEM
n
t
2
=

t
_

J e
tn
d

z
_

J
m
u
z
e
tn
d

2
i
TEM
n
z
2

1
c
2

2
i
TEM
n
t
2
=

z
_

J e
tn
d +

t
_

J
m
u
z
e
tn
d
(12)
from (7) and (8). For TE modes the modal voltage v
TE
n
satises the
following one-dimensional Klein-Gordon equation, i.e.,

2
v
TE
n

2
z

1
c
2

2
v
TE
n
t
2
k
2
cn
v
TE
n
=

t
_

J e
tn
d

z
_

J
m
u
z
e
tn
d
274 Geyi
+k
cn
_

(u
z
J
m
)
_
u
z
e
tn
k
cn
_
d (13)
from (7), (9) and (10). The modal current i
TE
n
can be determined by
a time integration of v
TE
n
/z. For TM modes, it can be shown that
the modal current i
TM
n
also satises the Klein-Gordon equation

2
i
TM
n

2
z

1
c
2

2
i
TM
n
t
2
k
2
cn
v
TM
n
=

z
_

J e
tn
d+

t
_

J
m
u
z
e
tn
dk
cn
_

u
z
J
_
e
tn
k
cn
_
d (14)
from (7), (8) and (9). The modal voltage v
TM
n
can be determined
by a time integration of i
TM
n
/z. Now we can see that the
excitation problem in a waveguide is reduced to the solution of a
series of inhomogeneous Klein-Gordon equations. Compared to the
treatment in previous publications, where only homogeneous Klein-
Gordon equations are involved but very complicated time-domain
vector mode functions have to be introduced to solve an excitation
problem [e.g., 21], our approach is much simpler, and at the same time
more general since it can be applied to a hollow waveguide as well as
a multiple conductor transmission line.
3. SOLUTIONS OF INHOMOGENEOUS
KLEIN-GORDON EQUATION
To get the complete solution of the transient elds in the waveguide,
we need to solve the inhomogeneous Klein-Gordon equation. This can
be done by means of the retarded Greens function.
3.1. Retarded Greens Function of Klein-Gordon Equation
The retarded Greens function for Klein-Gordon equation is dened by
_

_
_

2
z
2

1
c
2

2
t
2
k
2
cn
_
G
n
(z, t; z

, t

) = (z z

)(t t

)
G
n
(z, t; z

, t

t<t

= 0
(15)
where the second equation represents the causality condition. Taking
the Fourier transform, the rst equation reduces to

G
n
(p, ; z

, t

) = c
2
e
jpz

jt

/
_

2
p
2
c
2
k
2
cn
c
2
_
Progress In Electromagnetics Research, PIER 59, 2006 275
The inverse Fourier transform may then be written as
G
n
(z, t; z

, t

) =
c
2
(2)
2

e
jp(zz

)
dp

e
j(tt

)
(
2
p
2
c
2
k
2
cn
c
2
)
d
To calculate the second integral we may extend to the complex
plane and use the residue theorem in complex variable analysis. There
are two simple poles
1,2
=
_
p
2
c
2
+ k
2
cn
c
2
in the integrand. To
satisfy the causality condition, we only need to consider the integral
of e
j(tt

)
/(
2
p
2
c
2
k
2
cn
c
2
) along a closed contour consisting of the
real axis from to and an innite semicircle in the upper half
plane. Note that the poles on the real axis have been pushed up a little
bit by changing them into
1,2
=
_
p
2
c
2
+ k
2
cn
c
2
+ j to satisfy the
causality condition, where will be eventually made to approach zero.
The contour integral along the large semicircle will be zero for t > t

.
Using the residue theorem and [27]

_
0
_
sinq

x
2
+ a
2

x
2
+ a
2
_
cos bxdx =

2
J
0
_
a
_
q
2
b
2
_
H(q b),
a > 0, q > 0, b > 0
we obtain
G
n
(z, t; z

, t

) =
c

_
0
_
_
_
sin
_
c(t t

)
_
p
2
+ k
2
cn
_
_
p
2
+ k
2
cn
_
_
_
cos p(z z

)dp
=
c
2
J
0
_
k
cn
c
_
(t t

)
2
|z z

|
2
/c
2
_
H(t t

|z z

|/c) (16)
where J
0
(x) is the Bessel function of rst kind and H(x) the unit step
function. Note that when k
cn
= 0 this reduces to Greens function of
wave equation.
3.2. Solution of Inhomogeneous Klein-Gordon Equation
Consider the inhomogeneous Klein-Gordon equation
_

2
z
2

1
c
2

2
t
2
k
2
cn
_
u
n
(z, t) = f(z, t) (17)
and (15). Both equations can be transformed into the frequency
domain by Fourier transform
_

2
z
2
+
2
cn
_
u
n
(z, ) =

f(z, )
276 Geyi
left-traveling wave
J
right-traveling wave
z=a z=b
Figure 1. Left-traveling wave and right-traveling wave in waveguide.
_

2
z
2
+
2
cn
_

G
n
(z, ; z

, t

) = (z z

)e
jt

where
2
cn
=
2
/c
2
k
2
cn
. Multiplying the rst equation and the second
equation by

G
n
and u respectively and then subtracting the resultant
equations yield
u
n
(z, )

2

G
n
(z, ; z

, t

)
z
2


G
n
(z, ; z

, t

2
u
n
(z, )
z
2
= (z z

) u
n
(z, )e
jt

f(z, )

G
n
(z, ; z

, t

) (18)
We assume that the source distribution f(z, t) is limited in a nite
interval (a, b) as shown in Figure 1. Taking the integration of (18) over
the interval [a, b] and then taking the inverse Fourier transform with
respect to time, we obtain
u
n
(z, t) =
_
_

G
n
(z, t; z

, t

)
u
n
(z

, t

)
z

dt

u
n
(z

, t

)
G
n
(z, t; z

, t

)
z

dt

_
_
b
z

=a

b
_
a
dz

f(z

, t

)G
n
(z, t; z

, t

)dt

, z (a, b) (19)
where we have used the symmetry of Greens function about z and z

.
Note that if a , b , the above expression becomes
u
n
(z, t) =

dz

f(z

, t

)G
n
(z, t; z

, t

)dt

, z (, ) (20)
Next we will show that the solution in the region (z
+
, +)(z
+
b)
and (, z

)(z

a) may be expressed in terms of its boundary

Progress In Electromagnetics Research, PIER 59, 2006 277
values at b and a. Without loss of generality, we assume that a < 0
and b > 0. Taking the integration of (18) over [z
+
, +] with z
+
b
and using integration by parts, we obtain

G
n
(z
+
, ; z

, t

)
u
n
(z
+
, )
z
u
n
(z
+
, )

G
n
(z
+
, ; z

, t

)
z
= u
n
(z

, )e
jt

where we have used the radiation condition at z = +. Taking the

inverse Fourier transform and letting z

= z
+
lead to
u
n
(z, t) + cJ
0
(k
cn
ct)H(t)
u
n
(z, t)
z
= 0, z b > 0 (21)
where we have replaced z
+
by z and t t

by t since z
+
and t

are
arbitrary, and denotes the convolution with respect to time. This
equation is called right-traveling condition of the wave [21]. Similarly
taking the integration of (18) over (, z

] with z

a, we may
obtain left-traveling condition
u
n
(z, t) cJ
0
(k
cn
ct)H(t)
u
n
(z, t)
z
= 0, z a < 0 (22)
Both (21) and (22) are integral-dierential equations. If the source is
turned on at t = 0, all the elds must be zero when t < 0. In this
case, (21) and (22) can be solved by single-sided Laplace transform.
Denoting the Laplace transform of u
n
by u
n
we have
u
n
(z, s) +
_
(s/c)
2
+ k
2
cn
_
1/2
u
n
(z, s)
z
= 0, z b > 0
u
n
(z, t)
_
(s/c)
2
+ k
2
cn
_
1/2
u
n
(z, s)
z
= 0, z a < 0
The solutions of the above equations can be easily found. Making use
of the inverse Laplace transforms listed in [26], the solutions of (21)
and (22) can be expressed as
u
n
(z, t) = u
n
_
b, t
z b
c
_
ck
cn
(z b)

t
zb
c
_
0
J
1
_
k
cn
c
_
(t )
2
(z b)
2
/c
2
_
_
(t )
2
(z b)
2
/c
2
u
n
(b, )d
(z b > 0)
278 Geyi
u
n
(z, t) = u
n
_
a, t +
z a
c
_
+ ck
cn
(z a)

t+
za
c
_
0
J
1
_
k
cn
c
_
(t )
2
(z a)
2
/c
2
_
_
(t )
2
(z a)
2
/c
2
u
n
(a, )d
(z a < 0)
The above two equations have also been derived by the method
of impulse response function [16] and wave splitting technique [21].
Once the input signal is known the output signal after traveling a
certain distance in the waveguide can be determined by a convolution
operation. Some numerical results can be found in references [16, 21].
However there is a common mistake in studying the propagation of
transients in a waveguide, which assumes a wideband input signal (such
as a step function or a rectangular pulse) and a single-mode operation
at the same time. The electric eld at the output is then determined
by the product of the response determined by one of the above two
equations and the transverse mode function. This process totally
ignores the contributions from higher order modes and is not accurate
enough to describe the actual response in the waveguide. Indeed if the
excitation pulse is wideband, the waveguide will be overmoded and the
eects of higher order modes cannot be neglected in most situations.
This will be demonstrated below.
4. TRANSIENT PROCESS IN WAVEGUIDES
Now the above general theory will be applied to study the propagation
characteristics of electromagnetic pulse in waveguides. In general a
wideband pulse in the waveguide will generate a number of higher
order modes, and the eld distributions inside the waveguide should
be determined by (3) and (4), which will be approximated by a nite
summation of m terms.
4.1. Transient Process in a Rectangular Waveguide
Let us rst consider a rectangular waveguide of width a and height b
as depicted in Figure 2. If the waveguide is excited by a line current
extending across the waveguide located at x
0
= a/2, then current
density is given by
J(r, t) = u
y
(x x
0
)(z z
0
)f(t)
Since the line current density is uniform in y direction, the eld excited
by the current will be independent of y. As a consequence, only TE
n0
Progress In Electromagnetics Research, PIER 59, 2006 279
o
x
y
0 a
b
J
x
Figure 2. Cross-section of a rectangular waveguide.
mode will be excited. In this case we have
k
cn
=
n
a
, e
tn
(x, y) = e
TE
n0
(x, y) = u
y
_
2
ab
_
1/2
sin
nx
a
, n = 1, 2, 3 . . .
From (13), (20) and the above equation, we obtain
v
TE
n
(z, t) =
b
2
_
2
ab
_
1/2
sin
n
a
x
0

t|zz
0
|/c
_

df(t

)
dt

J
0
_
k
cn
c
_
(tt

)
2
|zz
0
|
2
/c
2
_
dt

(23)
where =
_
/. Thus the time-domain voltages v
TE
n
for TE
n0
(n =
2, 4, 6, . . .) vanish. From (3) the total electric eld in the waveguide
may be approximated by a nite summation of m terms
E = u
y
E
y
= u
y
_
2
ab
_
1/2
m

n=1
v
TE
n
sin
nx
a
= u
y
m

n=1

a
sin
n
a
x
0
sin
nx
a

t|zz
0
|/c
_

df(t

)
dt

J
0
_
k
cn
c
_
(t t

)
2
|z z
0
|
2
/c
2
_
dt

(24)
The above expression has been validated in Appendix B by using
a totally dierent approach. The theory can also be validated by
considering the time-domain response to a continuous sinusoidal wave
turned on at t = 0. We should expect that the time-domain response
tends to the well-known steady state response as time goes to innity.
Hereafter we assume that x
0
= a/2, and z
0
= 0 for all numerical
280 Geyi
examples. To validate the theory let f(t) = H(t) sint. Then (23)
may be expressed as
v
TE
n
(z, t) = v
TE
n
(z, t)

steady
+ v
TE
n
(z, t)

transient
, t > |z|/c
where k = /c and v
TE
n
(z, t)

steady
represents the steady-state part of
the response and v
TE
n
(z, t)

transient
the transient part of the response
v
TE
n
(z, t)

steady
=
b
2
_
2
ab
_
1/2
ka sin
n
2

_
|z|/a
cos ka(ct/a u)J
0
_
k
cn
a
_
u
2
|z|
2
/a
2
_
du
v
TE
n
(z, t)

transient
=
b
2
_
2
ab
_
1/2
ka sin
n
2

_
ct/a
cos ka(ct/a u)J
0
_
k
cn
a
_
u
2
|z|
2
/a
2
_
du
Note that the transient part of the response approaches to zero as
t . To investigate the steady-state part of the response the
following calculations are needed [27]

_
a
J
0
_
b
_
x
2
a
2
_
sincxdx =
_
0, 0 < c < b
cos
_
a

c
2
b
2
_
/

c
2
b
2
, 0 < b < c

_
a
J
0
_
b
_
x
2
a
2
_
cos cxdx =
_
_
_
exp
_
a

b
2
c
2
_
/

b
2
c
2
, 0 < c < b
sin
_
a

c
2
b
2
_
/

c
2
b
2
, 0 < b < c
Thus we have
v
TE
n
(z, t)

steady
=
b
2
_
2
ab
_
1/2 ka sin
n
2
_
(ka)
2
(k
cn
a)
2

_
sin
_
ka
ct
a

|z|
a
_
(ka)
2
(k
cn
a)
2
_
, k>k
cn
cos
_
ka
ct
a
_
exp
_
|z/a|
_
(k
cn
a)
2
(ka)
2
_
, k<k
cn
Therefore only those modes satisfying k > k
cn
will propagate in the
steady state. When v
TE
n

steady
are inserted into (24) it will be found
Progress In Electromagnetics Research, PIER 59, 2006 281
0 2 4 6 8 10
4
2
0
2
4
v
TE
1
ct/a{z/a = 2, ka = 1.2}
v
TE
3
Figure 3. Time-domain modal voltages excited by a sinusoidal wave.
0 5 10 15 20
0
3
4.5
ct/a{x = a/2, z = 2a, ka = 1.2}
E
y
{m = 1} E
y
{m = 3}
Figure 4. Electric elds excited by a sinusoidal wave.
that the steady state response of the electric eld agrees with the
traditional time harmonic theory of waveguides (see Eqn. (74), Chapter
5 of [20]). The variation of v
TE
n
=

2v
TE
n
/
_
b/a and E
y
= aE
y
/
with the time at z = 2a for dierent modes have been shown in Figure 3
and Figure 4 respectively, where m stands for the number of terms
chosen in (24). The operating frequency has been chosen in between
the rst cut-o frequency and the second one. It can be seen that the
contribution from the higher order modes are negligible. However this
is not true for other wideband excitation pulse.
282 Geyi
0 2 4 6 8 10
1
0
1
ct/a{z = 2a}
v
TE
1
v
TE
5
Figure 5. Time-domain modal voltages excited by a unit step
waveform.





0 2 4 6 8 10
2
0
2
{ }
/ /2 , 2 ct a x a z a = =
{ }
1
y
E m=
{ }
39
y
E m=
Figure 6. Electric elds excited by a unit step waveform.
Let us consider a unit step pulse, i.e., f(t) = H(t). The time-
domain voltages v
TE
n
for the rst and fth mode are shown in Figure 5,
which clearly indicates that the voltage for the higher order modes
cannot be ignored in this case. As indicated by Figure 6, the electric
elds at z = 2a, obtained by assuming m = 1 (only the dominant mode
is used) and m = 39 (the rst 39 modes are used), are quite dierent
due to the signicant contributions from the higher order modes. It
is seen that time response of the eld is totally dierent from the
original excitation pulse (i.e., a unit step function) due to the fact that
a hollow waveguide is essentially a high pass lter, which blocks all
Progress In Electromagnetics Research, PIER 59, 2006 283
the low frequency components below the rst cut-o frequency. In
addition the waveguide exhibits severe dispersion. It should be noted
that the singularities in the electric eld distributions come from the
time derivatives of the excitation waveform in (23), and the periodic-
like performance of the eld distribution results from the behavior of
Bessel functions.
Thus it is clear that a hollow metal waveguide is not an ideal
medium to carry a wideband signal that contains signicant low
frequency components below the rst cut-o frequency. In this case
one should use a multi-conductor transmission line supporting a TEM
mode whose cut-o frequency is zero.
0

a
y
x
m
J
b
Figure 7. Cross-section of a coaxial waveguide.
4.2. Transient Process in a Coaxial Cable
To see how a pulse propagates in a TEM transmission line as well as
the eects of the higher order modes we may consider a coaxial line
consisting of an inner conductor of radius a and an outer conductor
of radius b, as shown in Figure 7. Let the coaxial line be excited by a
magnetic ring current located at z = 0, i.e.,
J
m
(r, t) = u

f(t)(z)(
0
), a <
0
< b
where (, , z) are the polar coordinates and u

is the unit vector

in direction. According to the symmetry, only TEM mode and
those TM
0q
modes that are independent of will be excited. The
orthonormal vector mode functions for these modes are given by [28]
k
c1
= 0, e
t1
(, ) = u
r
1

2 lnc
1
k
cn
=

n
a
,
284 Geyi
e
tn
(, ) = u

n
a
J
1
(
n
/a)N
0
(
n
) N
1
(
n
/a)J
0
(
n
)
_
J
2
0
(
n
)/J
2
0
(c
1

n
) 1
, n 2
where c
1
= b/a, u

is the unit vector in direction, and

n
is the nth nonvanishing root of the eqauation J
0
(
n
c
1
)N
0
(
n
)
N
0
(
n
c
1
)J
0
(
n
) = 0. It follows from (12), (14) and (20) that
i
TEM
1
=

2 lnc
1
f(t |z z
0
|/c)
i
TM
n
=

2a
J
1
(
n

0
/a)N
0
(
n
) N
1
(
n

0
/a)J
0
(
n
)
_
J
2
0
(
n
)/J
2
0
(c
1

n
) 1

t|zz
0
|/c
_

df(t

)
dt

J
0
_
k
cn
c
_
(t t

)
2
|z z
0
|
2
/c
2
_
dt

Therefore if the highest frequency component of the excitation

waveform is below the cut-o frequency of the rst higher mode, the
coaxial line will be an ideal medium for a distortion-free transmission
of signals. From (4) the magnetic eld in the coaxial cable is given by
H = u

= u

i
TEM
n
1

2 lnc
1
+u

n=2
i
TM
n

n
2a
J
1
(
n
/a)N
0
(
n
) N
1
(
n
/a)J
0
(
n
)
_
J
2
0
(
n
)/J
2
0
(c
1

n
) 1
=
u

2 lnc
1
f(t |z z
0
|/c)
u

n=2

2
n

0
4a
2

J
1
(
n

0
/a)N
0
(
n
) N
1
(
n

0
/a)J
0
(
n
)
_
J
2
0
(
n
)/J
2
0
(c
1

n
) 1

J
1
(
n
/a)N
0
(
n
) N
1
(
n
/a)J
0
(
n
)
_
J
2
0
(
n
)/J
2
0
(c
1

n
) 1

t|zz
0
|/c
_

df(t

)
dt

J
0
_
k
cn
c
_
(t t

)
2
|z z
0
|
2
/c
2
_
dt

(25)
In the following all numerical examples are based on the assumptions
that
0
= (a+b)/2, and z
0
= 0. The time-domain currents for the rst
three modes excited by a unit step waveform f(t) = H(t) are depicted
in Figure 8, where i
TEM
1
= i
TEM
1
and i
TM
n
= i
TM
n
. Figure 9 gives
Progress In Electromagnetics Research, PIER 59, 2006 285
0 2 4 6 8 10
0
1
2
i
TEM
1
ct/a{z = 2a}
i
TM
3
i
TM
2
Figure 8. Time-domain modal currents excited by unit step
waveform.
0 2 4 6 8 10
0
0.5
1
1.5
ct/a{r = 1.1a, z = 2a}
H

{m = 5}
H

{m = 1}
Figure 9. Magnetic elds excited by unit step waveform.
the magnetic elds H

= H

at r = 1.1a, z = 2a, where m denotes

the number of terms chosen in (25). If we neglect the contribution of
the higher modes by letting m = 1, the magnetic eld is a perfect step
function. When higher modes are included (m = 5, the rst ve modes
are used), some ripples will occur in the response of the magnetic eld,
which stands for the contribution of higher order modes excited by the
high frequency components of the unit step waveform. In this situation
the signal has been distorted when it propagates in the coaxial line.
Fig. 10 and Fig. 11 give the time-domain currents and the
286 Geyi
0 2 4 6 8 10
0
1
2
i
TEM
1
i
TM
2
i
TM
3
ct/a{z = 2a, cT/a = 3}
Figure 10. Time-domain modal currents excited by a rectangular
pulse.
0 2 4 6 8 10
0.5
0
0.5
1
H

{m = 5}
H

{m = 1}
ct/a{r = 1.1a, z = 2a, cT/a = 3}
Figure 11. Magnetic elds excited by rectangular pulse.
magnetic elds in the coaxial line excited by a rectangular pulse f(t) =
H(t) H(t T) respectively. It can be seen that the ripples occur
not only inside the pulse but also outside the pulse. Therefore when a
rectangular pulse train passes through a transmission line the pulse will
spread in time, which causes the pulse to smear into the time intervals
of succeeding pulses. This phenomenon is very similar to the situation
of a pulse train passing through a bandlimited channel, where the
pulses will also spread in time, introducing intersymbol interference.
Note that the time spread in the transmission line is caused by the
Progress In Electromagnetics Research, PIER 59, 2006 287
higher order modes while the time spread in a bandlimited channel is
due to the shortage of bandwidth. To reduce the eects of time spread
in both cases, the pulse shaping techniques can be used to restrain the
high frequency components.
5. CONCLUSIONS
In this paper a new and concise approach for the time-domain theory
of waveguides has been presented. The theory has been applied to
the time-domain analysis of excitation problems in typical waveguides.
Numerical results for various typical excitation pulses have been
expounded, which give a real physical picture of the transient process in
a waveguide. In a special case where the excitation pulse is a sinusoidal
wave turned on at t = 0, the steady state response is shown to approach
to the well-known result from time-harmonic theory, which validates
our theory. More validations can be found in the Appendices.
Our analysis results have also indicated that the contributions
from the higher order modes excited by a wideband waveform are
signicant. Therefore the input signal at a given reference plane cannot
simply be written as a single term of a separable function of space
and time when studying the propagation of wideband signals in a
waveguide or a feeding line of an antenna. Instead the expression
of the input signal should consist of a number of such terms given
by (3) and (4), each term having a dierent time variation than the
others. The number of terms to be selected depends on the accuracy
required as well as the bandwidth of excitation waveform. The wider
the bandwidth of the excitation waveform, the more terms of the higher
order modes must be included.
APPENDIX A. PROOF OF THE COMPLETENESS OF
THE TRANSVERSE EIGENFUNCTIONS
To prove the completeness of the transverse eigenfunctions we need the
following theorem [29, pp. 284].
Theorem: Let H be a real separable Hilbert space with dimH = .
Let B : D(B) H H be a linear, symmetric operator. We further
assume that B is strongly monotone, i.e., there exists a constant c
1
such that (Bu, u) > c
1
u
2
for all u D(B). Let B
F
be the Friedrichs
extension of B and H
B
be the energy space of operator B. We further
assume that the embedding H
B
H is compact. Then the following
eigenvalue problem
B
F
u = u, u D(B
F
)
288 Geyi
has a countable eigenfunctions {u
n
}, which form a complete
orthonormal system in the Hilbert space H, with u
n
H
B
. Each
eigenvalue corresponding to u
n
has nite multiplicity. Furthermore we
have
1

2
and
n

n
.
Now consider the following eigenvalue problem for the transverse
electric eld in the waveguide
_
B(e) = e ( e) = k
2
c
e, r
u
n
e = e = 0, r
(A1)
where k
2
c
=
2

2
, and B = . The domain of
denition of operator B is dened as follows
D(B) =
_
e

e (C

())
2
, u
n
e = e = 0 on
_
where C

() stands for the set of functions that have continuous

partial derivatives of any order. Let L
2
() stand for the space
of square-integrable functions dened in and H = (L
2
())
2
=
L
2
() L
2
(). For two transverse vector elds e
1
and e
2
in (L
2
())
2
,
the usual inner product is dened by
(e
1
, e
2
) =
_

e
1
e
2
d.
and the corresponding norm is denoted by = (, )
1/2
. In the above
a bar over a letter is used to represent the complex conjugate. Now we
modify (A1) as equivalent form by adding a term e on both sides
_
A(e) = e ( e) + e =
_
k
2
c
+
_
e, r =
u
n
e = e = 0, r
where is an arbitrary positive constant. In order to apply the previous
theorem we need to prove that the operator A is symmetric, strongly
monotone and the embedding H
A
H is compact. The proof will be
carried out in three steps.
Step 1: We rst show that A is symmetric, strongly monotone. For all
e
1
, e
2
D(A) = D(B), we have
(e
1
, e
2
)
A
= (A(e
1
), e
2
) =
_

[e
1
( e
1
) + e
1
] e
2
d
=
_

[e
1
e
2
+ ( e
1
)( e
2
) + e
1
e
2
]d (A2)
Progress In Electromagnetics Research, PIER 59, 2006 289
Therefore the new operator A is symmetric. Thus we can assume that
e is real. A is also strongly monotone since
(A(e), e) =
_

[e e + ( e)( e) + e e]d e
2
Step 2: We then demonstrate some important properties of energy
space H
A
. The energy space H
A
is the completion of D(A) with
respect to the norm
A
= (, )
1/2
A
. Now let e H
A
, and by
denition, there exists admissible sequence {e
n
D(A)} for e such
that
e
n
e

n
0
and {e
n
} is a Cauchy sequence in H
A
. From (A2) we obtain
e
n
e
m

2
A
= e
n
e
m

2
+ e
n
e
m

2
+e
n
e
m

2
Consequently {e
n
} and { e
n
} are Cauchy sequences in H. As
a result, there exist h H, and L
2
() such that
e
n

n
h, e
n

n

From integration by parts
_

e
n
d =
_

e
n
d, (C

0
())
2
_

( e
n
)d =
_

e
n
d, C

0
()
we obtain
_

h d =
_

e d, (C

0
())
2
_

d =
_

e d, C

0
().
In the above C

0
() is the set of all functions in C

() that vanish
outside a compact set of . Therefore e = h and e =
in the generalized sense. For arbitrary e
1
, e
2
H
A
, there are two
admissible functions {e
1n
} and {e
2n
} such that e
1n
e
1

n
0 and
e
2n
e
2

n
0. We dene
(e
1
, e
2
)
A
= lim
n
(e
1n
, e
2n
)
A
=
_

[e
1
e
2
+ ( e
1
)( e
2
) + e
1
e
2
] d
290 Geyi
where the derivatives must be understood in the generalized sense.
Step 3: Finally we prove that the embedding H
A
H is compact. Let
J(e) = e, e H
A
Then the linear operator J : H
A
H is continuous since
J(e)
2
= e
2

1
_
e
2
+e
2
+ e
2
_
=
1
e
H
A
.
A bounded sequence {e
n
} H
A
implies
e
n

2
H
A
= e
n

2
+e
n

2
+ e
n

2
=
_

_
(e
nx
)
2
+ (e
ny
)
2
+ (e
nx
)
2
+ (e
ny
)
2
_
d c

where c

is a constant. So the compactness of the operator J follows

from the above inequality and the following Rellichs theorem.
Rellichs theorem [25]: Any sequence {f
n
}, which satises f
n

2
=
_

f
2
n
d c
1
and f
n

2
=
_

(f
n
)
2
d c
2
, where c
1
and c
2
are constants, has a subsequence still denoted by {f
n
} such that
lim
m,n
_

w(f
m
f
n
)
2
d = 0, where w is the weight function.
From the previous theorem, the set of eigenfunctions {e
n
}
constitutes a complete system in (L
2
())
2
. The proof is completed.
APPENDIX B. A DIFFERENT APPROACH TO THE
TRANSIENT RESPONSE OF RECTANGULAR
WAVEGUIDE
To validate (24), let us take a totally dierent approach to the
excitation problem considered in Section 4.1. The electromagnetic
elds in the waveguide satisfy the wave equations
E +

2
E
t
2
=
J
t
, H +

2
H
t
2
= J
From the wave equations and the property of exciting source only the
y component of the electric eld E is excited. The y component of the
electric eld E
y
satises
_

2
x
2
+

2
z
2


2
t
2
_
E
y
= f

(t)(x x
0
)(z z
0
)
Progress In Electromagnetics Research, PIER 59, 2006 291
and the boundary condition E
y
(x, z, t)

x=0,a
= 0. Making use of the
Fourier transform pair with respect to z and t,

E
y
(x, p, ) = F(E
y
) =

E
y
(x, z, t)e
jpzjt
dzdt
E
y
(x, z, t) = F
1
(

E
y
) =
1
(2)
2

E
y
(x, p, )e
jpz+jt
dpd
we obtain
_

_
_

2
x
2
p
2
+

2
c
2
_

E
y
= j

f()e
jpz
0
(x x
0
)

E
y
(x, p, )

x=0,a
= 0
(B1)
where c = 1/

and

f() is the Fourier transform of f(t). According
to the boundary condition we may use the well-known complete
orthonormal set {
_
2
a
sin
n
a
x|n = 1, 2, . . .} to expand the electric eld
[30]. So we have

E
y
=

n=1
e
n
(z, t)
_
2
a
sin
n
a
x (B2)
Inserting this into (B1) we get

n=1
e
n
(z, t)
_
2
a
sin
n
a
x
_

_
n
a
_
2
p
2
+

2
c
2
_
= j

f()e
jpz
0
(xx
0
)
Multiplying both sides by
_
2/a sin(nx/a) and taking the integration
over [0, a] gives
e
n
(z, t) =
jc
2

f()
_
2/a sin(nx
0
/a)e
jpz
0

2
c
2
(n/a)
2
p
2
c
2
The electric eld can be obtained by taking the inverse Fourier
transform
E
y
=

n=1
F
1
[ e
n
(z, t)]
_
2
a
sin
n
a
x
292 Geyi
where
F
1
[ e
n
(z, t)] = c
2
_
2
a
sin
nx
0
a
F
1
_
j

f()e
jpz
0

2
c
2
(n/a)
2
p
2
c
2
_
=
c
2
(2)
2
_
2
a
sin
nx
0
a
f

(t)

e
jpzjpz
0
dp

e
jt

2
p
2
c
2
c
2
(n/a)
2
d
The integral with respect to can be evaluated by residue theorem
and result is

e
jt
d

2
p
2
c
2
c
2
(n/a)
2
=
2 sint
_
p
2
c
2
+ c
2
(n/a)
2
_
p
2
c
2
+ c
2
(n/a)
2
Thus we have
F
1
[ e
n
(z, t)] =
c
2

_
2
a
sin
nx
0
a
f

(t)

_
0
cos p(z z
0
)
sintc
_
p
2
+ (n/a)
2
_
p
2
+ (n/a)
2
dp
=

2
_
2
a
sin
nx
0
a
f

(t)
J
0
_

n
c
_
t
2
|z z
0
|/c
2
_
H(t |z z
0
|/c)
Finally the electric eld is given by
E = u
y
E
y
= u
y

n=1

a
sin
nx
0
a
sin
n
a
x

t|zz
0
|/c
_

df(t

)
dt

J
0
_
k
cn
c
_
(t t

)
2
|z z
0
|
2
/c
2
_
dt

(B3)
which is exactly the same as (24).
APPENDIX C. A DIFFERENT APPROACH TO THE
TRANSIENT RESPONSE OF COAXIAL WAVEGUIDE
To validate (25), the excitation problem discussed in Section 4.2 will
be studied by directly solving Maxwell equations. The electromagnetic
Progress In Electromagnetics Research, PIER 59, 2006 293
elds in the coaxial waveguide satisfy
H+

2
H
t
2
=
J
m
t
, E+

2
E
t
2
= J
m
(C1)
From the above equations and the symmetry of the source only the
eld components E

, E
z
, and H

are excited. Thus we have

H =
H

z
u

+
1

(H

)u
z
The boundary condition u
n
E = 0 on the conductor requires

(H

=a,b
= 0. Equation (C1) reduces to

2
H

2
+
1

2
+

2
H

z
2

2
H

t
2
= f

(t)(z z
0
)(
0
)
Making use of the Fourier transform pair with respect to z and t,

(, , p, ) = F(H

) =

(, , z, t)e
jpzjt
dzdt
H

(, , z, t) = F
1
(

H

) =
1
(2)
2

(, , p, )e
jpz+jt
dpd
we get
_

2

H

2
+
1

+
_

2
_

H

= j

f()e
jpz
0
(
0
)
(

H

=a,b
=

+

H

=a,b
= 0
(C2)
where
2
=
2
p
2
. To solve this equation, we may use the method of
eigenfunction expansion [30]. We rst consider the following eigenvalue
problem
_

2
h

2
+
1

+
_

2
_
h

= 0

+ h

=a,b
= 0
(C3)
This is a typical eigenvalue problem of Sturm-Liouville type. The
general solution of the above equation is
h

= a
1
J
1
() + a
2
N
1
()
294 Geyi
Applying the boundary conditions yields
a
1
/a
2
= N
0
(a)/J
0
(a) = N
0
(b)/J
0
(b)
The above equation determines the eigenvalues
n
, n = 1, 2, . . . , which
satisfy
J
0
(c

n
)N
0
(
n
) J
0
(
n
)N
0
(c

n
) = 0
where
n
=
n
a is the nth root of the above equation, and c

= b/a.
The eigenfunctions corresponding to
n
are given by
h
n
() = b
n
[J
1
(
n
/a)N
0
(
n
) N
1
(
n
/a)J
0
(
n
)]
The constants b
n
can be determined by using the normalized condition
b
_
a
h
2
n
d = 1, which gives b
n
=

n

2a
_
J
2
0
(
n
)
J
2
0
(c

n
)
1
_
1/2
. The normalized
eigenfunctions may be expressed as
h
n
() =

n

2a
[J
1
(
n
/a)N
0
(
n
) N
1
(
n
/a)J
0
(
n
)]
_
J
2
0
(
n
)/J
2
0
(c

n
) 1

1/2
, n = 1, 2, . . .
In solving (C3) we have assumed that = 0. If = 0, (C3) reduces to
_

2
h

2
+
1

2
= 0
h

+ h

=a,b
= 0
The general solution of the above equation is h

= c
1
+ c
2
/, where
c
1
and c
2
are two constants. The boundary condition requires that
c
1
= 0. Therefore h

= c
2
/. Applying the normalized condition we
obtain eigenfunction corresponding to = 0
h
0
=
1

lnc

Thus the set of eigenfunctions {h

n
, n = 0, 1, 2, . . .} is complete by
Sturm-Liouville theory [30], and can be used to expand the solution of
(C2)

n=0
g
n
h
n
(C4)
where g
n
are the expansion coecients to be determined. Substituting
this into (C2) we obtain

n=0
g
n
_

2
h
n

2
+
1

h
n

+
_

2
_
h
n
_
= j

f()e
jpz
0
(
0
)
Progress In Electromagnetics Research, PIER 59, 2006 295
i.e.,

n=0
g
n
_

2
n
_
h
n
= j

f()e
jpz
0
(
0
)
Multiplying both sides by h
n
and taking the integration over [a, b],
we get
g
n
=
jc
2

0

f()e
jpz
0
h
n
(
0
)

2
p
2
c
2

2
n
c
2
The magnetic eld can then be obtained by taking the inverse Fourier
transform of (C4)
H

n=0
F
1
( g
n
)h
n
where
F
1
( g
n
) = c
2

0
h
n
(
0
)F
1
_
j

f()e
jpz
0

2
p
2
c
2

2
n
c
2
_
Similar to the discussion in Appendix B, we may obtain
F
1
( g
n
) =

0
h
n
(
0
)
2

t|zz
0
|/c
_

df(t

)
dt

J
0
_
k
cn
c
_
(t t

)
2
|z z
0
|
2
/c
2
_
dt

The magnetic eld is given by

H = u

=
u

2 lnc

f(t |z z
0
|/c)
u

n=1

2
n

0
4a
2

[J
1
(
n

0
/a)N
0
(
n
) N
1
(
n

0
/a)J
0
(
n
)]
_
J
2
0
(
n
)/J
2
0
(c

n
) 1

1/2

[J
1
(
n
/a)N
0
(
n
) N
1
(
n
/a)J
0
(
n
)]
_
J
2
0
(
n
)/J
2
0
(c

n
) 1

1/2

t|zz
0
|/c
_

df(t

)
dt

J
0
_
k
cn
c
_
(t t

)
2
|z z
0
|
2
/c
2
_
dt

which is exactly the same as (25).

REFERENCES
1. Geyi, W., Further research on the behavior of energy density
of electromagnetic pulse, Microwave and Optical Technology
Letters, Vol. 9, 331335, Aug. 20, 1996.
296 Geyi
2. Namiki, M. and K. Horiuchi, On the transient phenomena in the
wave guide, J. Phys. Soc. Japan, Vol. 7, 190193, 1952.
3. Ito, M., Dispersion of very short microwave pulses in waveguide,
IEEE Trans. Microwave Theory and Techniques, Vol. 13, 357364,
May 1965.
4. Stenius, P. and B. York, On the propagation of transients in
waveguides, IEEE Antennas and Propagation Magazine, Vol. 37,
No. 2, 3944, 1995.
5. Dvorak, S. L., Exact, closed-form expressions for transient
elds in homogeneous lled waveguides, IEEE Trans. Microwave
Theory and Techniques, Vol. 42, 21642170, Nov. 1994.
6. Elliott, R. S., Pulse waveform degradation due to dispersion
in waveguide, IRE Trans. Microwave Theory and Techniques,
Vol. 5, 254257, Oct. 1957.
7. Karbowiak, A. E., Propagation of transients in waveguides,
Proc. IEE (London), Vol. 104C, 339349, 1957.
8. Gajewski, R., Inuence of wall losses on pulse propagation in
waveguides, J. Appl. Phys., Vol. 29, 2224, 1958.
9. Forrer, M. P., Analysis of millimicrosecond RF pulse transmis-
sion, Proc. IRE, Vol. 46, 18301835, Nov. 1958.
10. Knop, M. and G. I. Cohn, Comments on pulse waveform
degradation due to dispersion in waveguides, IEEE Trans.
Microwave Theory and Techniques, Vol. MTT-11, 445447, Sept.
1963.
11. Knop, M., Further comments on pulse waveform degradation due
to dispersion in waveguides, IEEE Trans. Microwave Theory and
Techniques, Vol. MTT-18, 663665, Sept. 1970.
12. Cotte, M., Propagation of a pulse in a waveguide, Onde
Electron., Vol. 34, 143146, 1954.
13. Poincelot, P., Propagation of a signal along a waveguide, Ann.
Telecommun., Vol. 9, 315317, Nov. 1954.
14. Saxon, W. A. and H. J. Schimidt, Transients in a large
waveguide, Proc. IEEE, Vol. 51, 405406, Feb. 1963.
15. Berger and J. W. E. Griemsmann, Transient electromagnetic
guided wave propagation in moving media, IEEE Trans.
Microwave Theory and Techniques, Vol. MTT-16, 842849, Oct.
1968.
16. Schulz-DuBois, E. O., Sommerfeld pre- and postcursors in the
context of waveguide transients, IEEE Trans. Microwave Theory
and Techniques, Vol. MTT-18, 455460, Aug. 1970.
17. Nikitskiy, S. B., O. A. Tretyakov, and K. M. Yemelyanov, Ar-
Progress In Electromagnetics Research, PIER 59, 2006 297
bitrary signal propagation in waveguides, Mathematical Methods
in Electromagnetic Theory, MMET, Conference Proceedings, 260
263, 1996.
18. Nikitskiy, S. B., O. A. Tretyakov, and K. M. Yemelyanov, Waveg-
uide propagation of electromagnetic step signal, Proceedings
of the Mediterranean Electrotechnical Conference - MELECON,
Vol. 1, 26326, 1998.
19. Nikitskiy, S. B., O. A. Tretyakov, and K. M. Yemelyanov,
Electromagnetic step signal propagation in lossy waveguide,
Telecommunications and Radio Engineering (English translation
of Elektrosvyaz and Radiotekhnika), Vol. 51, No. 2, 149159, 1997.
20. Collin, R. E., Field Theory of Guided Waves, 2nd edition, IEEE
Press, 1991.
21. Kristensson, G., Transient electromagnetic wave propagation in
waveguides, Journal of Electromagnetic Waves and Applications,
Vol. 9, 645671, 1995.
22. Harmuth, H. F., Antennas and Waveguides for Nonsinusoidal
Waves, Academic Press, 1984.
23. Tretyakov, O. A., Evolutionary equations for the theory of
waveguides, IEEE Antennas and Propagation Society, AP-S
International Symposium (Digest), Vol. 3, 19731976, 1994.
24. Tretyakov, O. A., Evolutionary waveguide equations, Soviet
Journal of Communications Technology & Electronics (English
translation of Elektrosvyaz and Radiotekhnika), Vol. 35, No. 2,
717, 1990.
25. Kurokawa, K., An Introduction to the Theory of Microwave
Circuits, Academic Press, New York, 1969.
26. Erdelyi, A. (ed.), Tables of Integral Transform, Bateman
Manuscript Project, Vol. 1, McGraw-Hill, 1954.
27. Grandshteyn, I. S. and I. M. Ryzhik, Table of Integrals, Series,
and Products, Academic Press, New York, 1994.
28. Marcuvitz, N., Waveguide Handbook, McGraw-Hill, 1951.
29. Zeidler, E., Applied Functional Analysis-Applications to Mathe-
matical Physics, Springer-Verlag, 1995.
30. Courant, R. and D. Hilbert, Methods of Mathematical Physics,
Interscience Publishers, New York, 1953.