You are on page 1of 9

0.

1 Dielectric Slab Waveguide

At high frequencies (especially optical frequencies) the loss associated with the induced current in the metal
walls is too high. A transmission line lled with dielectric material but without conducting walls is another
structure that may be used to guide electromagnetic waves. This dielectric slab waveguide eliminates the
metallic absorption loss.
Consider a dielectric slab that is surrounded by another dielectric material that has a lower permittivity. A
representative slab is shown in Fig. 1. The waveguide thickness is 2d and the center region (core) has a higher
permittivity than the two outer regions (cladding) (
1
>
2
). We assume propagation in the z direction and
no variation of elds in the y direction (W 2d). This makes the problem simpler, because it reduces to a
two-dimensional analysis.
z
x
y
>
c
x=-d
x=d
propagating
wave
evanescent
wave
evanescent
wave
region 1
(core)
region 2
region 2

o
,
2

o
,
2

o
,
1
>
2
E
H
}
Figure 1: Dielectric slab waveguide of width 2d.
Since the dielectric waveguide is intended to guide the light, the elds in the cladding region should be
evanescent or decay in amplitude away from the slab. This guiding property requires the ray angle to be past
the critical angle >
c
= sin
1
__

1
_
= sin
1
_
n
2
n
1
_
. This requires the propagation constant to be in the
range
n
1
k
o
sin (
c
) < < n
1
k
o
sin (90
o
) (1)
n
2
k
o
< < n
1
k
o
. (2)
We could use the concept of computing E
z
or H
z
and then applying Maxwells equations to obtain the
remaining eld components just like we did in the metallic waveguides. However, the fact that we are now
in two dimensions makes things a little easier. We will still use TE and TM to indicate that the eld has no
longitudinal electric or magnetic eld, respectively. In the case of TE modes, the electric eld will only have
a y component (why?). We can then easily nd the magnetic eld using Faradays law. A similar argument
holds for TM modes, as we will see shortly.
Two other key concept concerning dielectric waveguides deserve attention. The rst is that, due to the
symmetry of the geometry, the elds will either be symmetric or anti-symmetric about the y-z plane. The
second is that in order for the eld to be guided by the high-permittivity dielectric slab, the elds outside the
ECEn 462 7 September 6, 2005
slab must be evanescent, i.e. they decay in the x direction. We will use these observations in the formulations
that follow.
0.1.1 TE Modes
The electric eld for the TE modes must satisfy our wave equation as given by
d
2
E
y
dx
2
+
_

2

2
_
E
y
= 0 (3)
This wave equation is valid in all regions. However, remember that the permittivity is different in the two
regions.
Our experience with rectangular waveguides tells us what the solutions must be (before application of the
boundary conditions). However, our argument about symmetry makes it so that within the slab, the variation
will either be
E
y
= Acos k
1x
x (4)
or
E
y
= Bsin k
1x
x. (5)
Note that we used the sin and cos form rather than the exp form. This is because we know that the elds
form standing waves within the waveguide region. By substituting these into the wave equation (Eq. 3), it
can be shown that
k
2
1x
=
2

2
= k
2
1

2
. (6)
The elds outside of the slab (in the cladding regions) are also of the same basic form.
E
y
= Ce
jk
2x
x
x d
E
y
= De
jk
2x
x
x d
(7)
These eld solutions use the exponential form because we know that they will not form standing waves. By
substituting these in Eq. (3), it can be shown that
k
2
2x
=
2

2
= k
2
2

2
. (8)
In order to maintain guiding, the elds in the cladding must be evanescent or decay in amplitude with
distance away from the slab. This requirement caused the propagation constant to be in the range of n
2
k
o
<
. Therefore, the propagation constant in the cladding regions is complex as given by
k
2x
= j
2
, (9)
where

2
=
_

2
k
2
2
. (10)
The sign of k
2x
is chosen such that the elds decay with distance away from the waveguide. The resulting
elds in the cladding regions are then given by
E
y
= Ce

2
x
x d
E
y
= De

2
x
x d
(11)
ECEn 462 8 September 6, 2005
Therefore, the electric eld in the various regions is given by
E = y
_

_
E
2
e
xjz
x d
E
1
_
sink
x
x
cos k
x
x
_
e
jz
|x| d
_

+
_
E
2
e
+xjz
x d
(12)
where the top and bottom lines in the braces refer to the antisymmetric and symmetric modes, respectively.
The solution that uses the cos is called the symmetric solution and the solution that uses the sin is called the
anti-symmetric solution.
Using Faradays law, we can now compute the magnetic elds,
H =
1
j
E =
_

_
E
2

0
( x zj) e
xjz
x d
E
1

_
x
_
sink
x
x
cos k
x
x
_
zjk
x
_
cos k
x
x
sin k
x
x
__
e
jz
|x| d
_

+
_
E
2

0
( x + zj) e
+xjz
x d
(13)
Note that in these eld expressions, we have used the fact that the z variation is of the form e
jz
both
inside and outside the slab. Why do we know that this propagation constant is the same in both regions?
Note also that we have 4 unknowns: E
2
/E
1
, k
x
, , and .
Since these elds must obey the wave equation (with
2
/y
2
= 0), we know that
k
2
x
+
2
= k
2
1
=
2

1
(14)

2
+
2
= k
2
2
=
2

2
(15)
which gives us two constraints for determining our unknowns. We need two additional constraints in order
to nd all 4 unknowns.
Lets start by enforcing continuity of tangential electric elds at the core-cladding interface. We will rst
consider the symmetric modes. Therefore, at x = d
E
1
cos(k
x
d)e
jz
= E
2
e
d
e
jz
cos(k
x
d)E
1
e
d
E
2
= 0 (16)
Note that applying continuity at x = d results in an identical equation, so this does not help us. This stems
from the symmetry of the problem, and in reality we have already used this symmetry to break the problem
into symmetric and antisymmetric modes.
Since we need one more equation, we will apply continuity of tangential (i.e. z component) magnetic eld
at the boundary. At x = d we have
jk
x
E
1

sin(k
x
d)e
jz
= j
E
2

e
d
e
jz
k
x
sin(k
x
d)E
1
e
d
E
2
= 0 (17)
and again, we get the exact same equation at x = d. The easiest thing to do is to divide these two equations
to simplify them.
k
x
sin (k
x
d) E
1
cos (k
x
d) E
1
=
e
d
E
2
e
d
E
2
(18)
ECEn 462 9 September 6, 2005
= k
x
tan(k
x
d) (19)
which can be re-written as
(d) = (k
x
d) tan(k
x
d) symmetric TE modes (20)
Subtracting (14) and (15) yields
k
2
x
+
2
=
2

0
(
1

2
) (21)
or as
(k
x
d)
2
+ (d)
2
=
2

0
(
1

2
)d
2
(22)
We can actually solve this graphically, as will be shown later. Alternately, we can combine these two
equations to obtain

2
+k
2
x
=
2

o
_
n
2
1
n
2
2
_
k
2
x
tan
2
(k
x
d) +k
2
x
=
2

o
_
n
2
1
n
2
2
_
tan
2
(k
x
d) + 1 =

2

o(n
2
1
n
2
2
)
k
2
x
tan(k
x
d) =
_

0
(n
1
n
2
)
k
2
x
1
(23)
and solve this with a nonlinear solver on a calculator or computer.
1. Solutions in the range (m1)/2 k
x
d m/2 m = 1, 3, 5, . . . we will call TE
m
modes. These
correspond to the symmetric TE modes.
2. Cutoff occurs when the mode is no longer guided, which occurs as soon as becomes negative. So,
we dene cutoff as the frequency at which = 0. Using (20), this implies that tan(k
x
d) = 0, so that
k
x
d = (m1)/2, m = 1, 3, 5, . . .. Using (22) with = 0 leads to
f
c,m
=
c(m1)
4d

r
1
(24)
for the cutoff frequencies of the TE modes. Note that f
c,1
= 0, so the lowest order mode propagates
at any frequency. Furthermore, since at cutoff = k
2
and
2
+ k
2
x
= k
2
1
, the angle of incidence of
the wave on the dielectric boundary can be expressed as

i
= sin
1

_

2
+k
2
x
= sin
1
k
2
k
1
= sin
1
_

1
=
c
(25)
which you may recognize as the critical angle. So, cutoff occurs when the angle of incidence on the
boundary is smaller than the critical angle. Makes sense, doesnt it?
Observe also that the cutoff condition of = k
2
means that the propagation constant becomes that of
the surrounding medium.
3. Note that k
x
is frequency dependent, unlike in the rectangular waveguide. From the two dispersion
relations, we can see that

1
, which should be intuitive.
4. As the frequency gets larger, which means that the eld decays very rapidly outside the
dielectric. The behavior of the mode becomes like that of a parallel plate waveguide lled with a
dielectric.
ECEn 462 10 September 6, 2005
Note that we could repeat the entire procedure for the antisymmetric TE modes. The dispersion relation (22)
remains the same. The guidance condition becomes
(d) = (k
x
d) cot(k
x
d) antisymmetric TE modes (26)
Again, cutoff occurs for k
x
d = (m1)/2, m = 2, 4, 6, . . .. These are therefore the even order TE modes.
0.1.2 TM Modes
We can repeat the whole process for TM modes. In this case, we have
H = y
_

_
H
2
e
xjz
x d
H
1
_
sin k
x
x
cos k
x
x
_
e
jz
|x| d
_

+
_
H
2
e
+xjz
x d
(27)
where the top and bottom lines in the braces refer to the antisymmetric and symmetric modes, respectively.
Using Amperes law, we can now compute the electric elds
E =
1
j
H =
_

_
H
2

0
( x + zj) e
xjz
x d
H
1

_
x
_
sin k
x
x
cos k
x
x
_
+ zjk
x
_
cos k
x
x
sin k
x
x
__
e
jz
|x| d
_

+
_
H
2

0
( x zj) e
+xjz
x d
(28)
We go through the exact same sequence of steps for this case. The dispersion relations remain the same.
The guidance conditions become
(d) =

0

(k
x
d) tan(k
x
d) symmetric TM modes (29)
(d) =

(k
x
d) cot(k
x
d) antisymmetric TM modes (30)
0.1.3 Graphical Mode Solution
Solving the simultaneous equations (22) and (20) (symmetric modes) or (26) (antisymmetric modes) for the
values of k
x
and is difcult, because the equations are transcendental. But it is easy to nd an approximate
solution using a graphical technique. This also provides valuable physical insight.
Each solution involves a guidance equation and a dispersion equation. These equations are:
ECEn 462 11 September 6, 2005
(d) = (k
x
d) tan(k
x
d) (31)
(d) = (k
x
d) cot(k
x
d) (32)
(k
x
d)
2
+ (d)
2
=
2

0
(
1

2
)d
2
(33)
Notice that we have multiplied each of the equations by d to make them dimensionless, which allows us to
use one plot for any value of the frequency or slab thickness. The idea is to plot these equations on the same
axes. Think of k
x
d as x and d as y. These equations then become
(d) = (k
x
d) tan(k
x
d) y = xtan(x) (34)
(d) = (k
x
d) cot(k
x
d) y = xcot(x) (35)
(k
x
d)
2
+ (d)
2
=
2

0
(
1

2
)d
2
x
2
+y
2
=
2

0
(
1

2
)d
2
(36)
The rst two equations are periodic functions. These functions are independent of both the premittiv-
ities and the the thickness of the waveguide. Figure 2 shows these plots.
The second equation is a circle with a radius of R =
_

0
(
1

2
)d. Thus, the radius of this
circle changes with waveguide parameters.
0 2 4 6 8 10 12
0
2
4
6
8
10
12 m=1 2 3 4 5 6 7 8
kx d (x)

d

(
y
)
Figure 2: Plots of the guidance condition for a TE dielectric slab waveguide. The solid lines are for the
symmetric modes and the dashed lines are for teh antisymmetric modes.
The second equation can be easily ploted as a quarter circle onto the lines shown in Fig. 2. Each intersection
of the curves corresponds to a mode. The coordinates of the point give us the value of k
x
d and d, which
can be easily divided by d to give k
x
and .
ECEn 462 12 September 6, 2005
Dielectric Waveguide Example
How many modes exist in a dielectric waveguide that has the following parameters? index of refraction of
the core n
1
= 1.6, index of refraction of the cladding n
2
= 1.5, wavelength = 1.0m, waveguide core
thickness 2d = 4m.
The equations are
d = k
y
d tan (k
y
d) (37)
d = k
y
d cot (k
y
d) (38)
(k
y
d)
2
+ (d)
2
= (k
o
d)
_
n
2
1
n
2
2
_
(39)
Using k
y
d = x and d = y these equations become
y = xtan x (40)
y = xcot x (41)
x
2
+y
2
= (k
o
d)
2
_
n
2
1
n
2
2
_
(42)
For this example the radius of the circle is given by
r =
2
1.0
4
2
_
1.6
2
1.5
2
(43)
r = 2.23 = 7.0 (44)
The equation xtan x is equal to zero when x = 0, 2, 3, ...m and is equal to when x =

2
,
3
2
,
5
2
, ...

2
+
m.
The equation xcot x is equal to zero when x =

2
,
3
2
,
5
2
, ...

2
+ m and is equal to when x =
, 2, 3, ...m. And when x = 0 xcot x = 1.
The radius of the circle for this problem is r = 7.0 = 2.23. There are 3 even modes (0, , 2) and 2 odd
modes (0.5, 1.5).
What is the waveguide thickness for single mode operation? We need
r < 0.5 (45)
2
1.0
d
_
1.6
2
1.5
2
<

2
(46)
d < 0.449 (47)
or a slab thickness of 2d = 0.9m
ECEn 462 13 September 6, 2005
0 2 4 6 8 10 12
0
2
4
6
8
10
12 m=1 2 3 4 5 6 7 8
kx d (x)

d

(
y
)
Figure 3: Graphical mode solution with n
1
= 1.6, n
2
= 1.5, 2d = 4m, and = 1.0m.
ECEn 462 14 September 6, 2005
0.1.4 Example - TM Modes
For the TM modes, the graphical solution method is complicated slightly by the fact that Eqs. (29) and (30)
depend on the permittivity of the slab. This means that we need to plot curves for different values of
r
. The
required plot is shown below.
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
k
x
d
d

r
= 1
2
4
10
Figure 4: Graphical solution curves for TM modes of a dielectric slab waveguide for various values of
r
.
(Note that the
r
= 1 curve can also be used for TE modes.)
Consider an example with slab thickness 2 cm, relative permittivity 2, and operating frequency 20 GHz. The
rst few cutoff frequencies are
f
c,1
= 0
f
c,2
= 7.5 GHz
f
c,3
= 15 GHz
f
c,4
= 22.5 GHz
From these values, we know that there are three propagating TM modes (and three TE modes as well). In
order to nd the values of k
x
and , we must nd the intersections of a circle of radius d

r
1/c 4.2
with the curves in Fig. 4.
For the dominant mode, this intersection point lies at d 4, so that = 4 Np/cm. This tells us that after
1/ .25 cm, the elds outside the slab have decayed by a factor of 1/e. The values of as well as other
constants can be found similarly for the higher order propagating modes. Will these modes be more or less
tightly bound to the slab?
ECEn 462 15 September 6, 2005