Wave Propagation in Cylindrical Waveguides

EL:313Electromagnetic Theory
May 19, 2009

1 Problem Formulation
Cylindrical waveguides are hollow tubes made of good conductors. Phe-
nomenon of wave propagation takes place inside the tube which may contain
a number of cylindrical structures within the outer shell which also acts as
a shield. It is desired find out all the possible solutions of electromagnetic
fields which satisfy Maxwells equations and the boundary conditions in ab-
sence of sources. Therefore it is considered that there is no excitation due to
charges or currents, i.e., = 0 and J = 0. It is assumed that all the fields
have time harmonic dependence exp(jt) which will be suppressed and will
not be stated explicitly. Maxwells equations in this situation dictate that
Helmholtzs equation should be satisfied, i.e.,

2 E + k 2 E = 0,
2 H + k 2 H = 0,

where k 2 = 2 . The axis of the cylinder is called the longitudal direction

and is taken as the zaxis. The plane perpendicular to the longitudal direc-
tion is called the transverse plane. It is desired to find out all the nontrivial,
i.e., nonzero solutions of electric and magnetic fields that propagate along the
waveguide axis, i.e., along zdirection. Therefore it is assumed that the fields
have the form E = E(x, y) exp(z) and H = H(x, y) exp(z). Propaga-
tion constant is as much an unknown of the problem as are the fields. It
is expedient to define transverse gradient operator

t = z
=x .
+y
z x y
With the assumed exponential dependance on longitudal coordinate the
Helmholtzs equation for electromagnetic fields, which now depend only on

1
the transverse coordinates, is transformed as follows,

2 E(x, y) + kc2 E(x, y) = 0, (1)

2 H(x, y) + kc2 H(x, y) = 0, (2)

where kc2 = k 2 + 2 . Since is an unknown it implies that kc is also an

unknown and has to be solved for. The total fields will now be decomposed
into transverse and longitudal components as follows

E(x, y) = Et (x, y) + Ez (x, y)z and H(x, y) = Ht (x, y) + Hz (x, y)z

where Et and Ht lie completely in the transverse plane. It will now be shown
that the transverse field components can be derived from the longitudal field
components. Consider Maxwells curl equation for electric field


E = t + z (Et + Ez z)
z
= t Et + t (Ez z) z Et
= t Et + t Ez z z Et = j (Ht + Hz z)

Noting that t Et is completely in the longitudal direction, transverse and

londitudal components in the above equation can be separated. as follows

jHt = t Ez z z Et (3)
jHz z = t Et (4)

Similarly using Maxwells curl equation for magnetic field the following rela-
tions can be obtained,

jEt = t Hz z z Ht (5)
jEz z = t Ht (6)

From (3) and (5) Ht can be eliminated to obtain a relationship for Et in the
following manner,

1
jEt = t Hz z z (t Ez z z Et )
j

This can be simplified to obtain the desired result for Et

1
Et = [t Ez + j z t Hz ] (7)
kc2

2
In a similar manner we obtain
1
Ht = 2 [t Hz j z t Ez ] (8)
kc
Equations (7) and (8) show that the transverse components of electric and
magnetic fields can be derived from longitudal field components Ez and Hz .
Therefore only the longitudal component of equations (1) and (2) have to
be really solved. Since Maxwells equations are linear, superposition holds
and each of the equations can be solved separately. There are two possible
solutions. Let Hz = 0 and Ez 6= 0. This is called transverse magnetic (TM)
solution or mode which is obtained by solving
2 Ez + kc2 Ez = 0. (9)
Conversely we can set Ez = 0 and Hz 6= 0. This is called transverse electric
(TE) solution or mode which is obtained by solving
2 Hz + kc2 Hz = 0 (10)
It is also possible to set both Ez = 0 and Hz = 0. In this case transverse
equations (3) and (5) imply that
(k 2 + 2 )Et = (k 2 + 2 )Ht = 0
and since we are looking for nontrivial solutions we must set (k 2 + 2 ) =
kc2 = 0. This solution is called transverse electromagnetic (TEM) mode. The
propagation constant for this mode is = jk.Therefore the phase velocity

of TEM waves is 1/ same as the phase velocity of plane waves in free
space. Transverse fields for this mode cannot be derived from (7) and (8)
because kc = 0. The equation for transverse fields can be readily obtained
by observing that the curl of Et in (4) is zero and hence may be derived from
the gradient of some function, i.e.,
E = Et exp(jkz) = t (x, y) exp(jkz) (11)
where the negative sign is for propagation in +z direction and positive sign is
for propagation in z direction. Transverse magnetic field may be obtained
from (3) as
1
H = Ht exp(jkz) = z t (x, y) exp(jkz) (12)

p
where = /. Since divergence of E is zero in charge free region therefore
(11) implies that
2t = 0 (13)

3
 Table 1: Waveguide Boundary Value Problems

Mode Differential Equation Boundary Condition

TEM 2t = 0 =Constant/s p jk
TM 2 Ez + kc2 Ez = 0 Ez = 0 j pk 2 kc2
TE 2 Hz + kc2 Hz = 0 =0
Hz n j k 2 kc2

It has been already determined that the propagation constant for the
TEM mode is = jk. For TE and TM modes mathematicians have deter-
mined that the eigenvalues kc2 that satisfy the boundary value problem must
be discrete, infinitely many and real positive values. Solutions for which
eigenvalues kc < k propagate in the zdirection and are called propagating
modes with = j. Conversely solutions for which eigenvalues kc > k decay
in the zdirection and are called evanescent modes. A cut off frequency may

be defined as c = kc / . A mode whose frequency is greater than the cut
off frequency of that mode will propagate otherwise it will decay.

2 Modal Solutions
From the above discussion it can been concluded that waveguide modes are
of three types. These modal solutions are obtained by solving appropriate
boundary value problem. For TEM modes the function is obtained by
solving Laplaces equation with the boundary condition is constant on the
cylindrical surfaces of the waveguide. This boundary condition ensures that
tangential electric field and normal magnetic field are zero on the perfectly
conducting surface. This is because the electric field is proportional the t
which in turn is normal to surface of constant . The magnetic field is
proportional to z t and thus it is parallel to the conducting surface. It
may be noted that a tube with a single closed surface cannot support a TEM
wave because (x, y) = constsnt is the solution to the Laplaces equation
and satisfies the boundary condition. The fields due to this solution are
identically zero.
TM modes satisfy Helmholtzs equation (9) for Ez with the boundary
condition Ez = 0 on the perfectly conducting boundary. This condition is
enough to ensure that tangential electric field and normal magnetic field are
zero on the perfectly conducting surface. In the first place Ez is tangential to
the surface and it is zero on the surface. Transverse component of the electric
field is proportional to t Ez and transverse component of the electric field is

4
 Table 2: Properties of Waveguide Modal Solutions

Mode Ez Hz Et Ht Zt
TEM 0 0 t (x, y) (1/)z t (x, y)
TM Ez 0 (/kc2 )t Ez (j/kc2 )z t Ez /j
TE 0 Hz (j/kc2 )z t Hz (/kc2 )t Hz j/

proportional to z t Ez . This situation is the same as was in TEM case, i.e.,

the conductor is a surface of constant Ez . Thus all the boundary conditions
for a perfectly conducting surface are satisfied.
Similarly TE modes satisfy Helmholtzs equation (10) for Hz with the
boundary condition Hz n = Hz /n = 0 where n is a unit vector normal
to the perfectly conducting boundary. Therefore t Hz is totally tangential
to the waveguide surface which in turn implies that normal magnetic field is
zero on the boundary. The total electric field is transverse and is proportional
to z t Hz and thus it is normal to the conducting surface.
It may be noted that transverse electric and magnetic fields are not only
orthogonal to the direction of propagation but for each mode they are also
orthogonal to each other. It is possible to define a mode impedance Zt for
each mode such that Et = Zt Ht z
The power transferred by a propagating mode may be calculated as
Z Z
1 1
WT = <(E H ) z dS = <(Et Ht ) z dS
2 2
Z
1
= <(Zt (Ht z) Ht ) z dS
2
Z
Zt
= <((Ht Ht )z) (Ht z)Ht ) z dS
2
Z Z
Zt 2 Yt
= |Ht | dS = |Et |2 dS (14)
2 2
Power is also lost in the waveguides. There are two sources of power loss,
losses in dielectric and losses due to resistivity of the walls. Although the
boundary value problem is solved with perfectly conducting wall and perfect
insulator as dielectric, it is assumed that losses are small, i.e., resistivity of
of the walls is low and conductivity of the dielectric is also small. It is also
assumed that due to small losses the solution to the boundary value problem
remains unchanged.

5
 Let us assume that the dielectric constant is lossy, i.e., = 0 j00 , where
00
= d /. When >> c then the propagation constant is modified with
the help of binomial theorem as follows,
p p
= kc2 k 2 = kc2 2 (0 j00 )
p 2 00
j 2 0 kc2 + p = j + d (15)
2 2 0 kc2
Therefore all the field components propagate as exp(jz) and their ampli-
tude decays as exp(d z). The attenuation constant due to lossy dielectric
is denoted as d = d /2. It depends on the mode only through its cut
off frequency.
The attenuation due to finite conductivity of the walls is more signifi-
cant than the attenuation due to conductivity of the dielectric. Assuming
small losses, the mode functions and propagation factor do not change
but only a decay constant c results due to wall losses. The fields will de-
cay as exp(c z) and power transferred WT being quadratic will decay as
exp(2c z). Therefore,
WT = P0 exp(2c z) (16)
where P0 is the power crossing z = 0 plane. The rate of decay of power per
unit length is given as
dWT
= 2c WT (17)
dz
Hence the decay constant may be calculated as
WT0 (z) Decrease of power per unit length
c = = (18)
2WT Twice average power transferred
Therefore to calculate c we have to calculate the power lost per unit length
in the conducting walls of the waveguide. Surface current induced on the
walls Js = n H causes power loss per unit area given as (1/2)Rs |Js |2 ,
where Rs = 1/. Thus power lost per unit length wL is given as
I I I
0 1 2 Rs 2 Rs
wL = WT = Rs |Js | dl = |Js | dl = |H|2 dl (19)
2 2 2
where the path of integration is along the circumference of the tube. The
attenuation constant in terms of mode functions may be written as
H
Rs {|Ht |2 + |Hz |2 } dl
c = R (20)
2Zt |Ht |2 dS
It may be noted that c is mode dependent.