You are on page 1of 25


Problems 361

improving liquid crystal displays, and other products, such as various optoelectronic
components, cosmetics, and hot and cold mirrors for architectural and automotive
windows. 9
8.14 Problems Waveguides
8.1 Prove the reectance and transmittance formulas (8.4.6) in FTIR.
8.2 Computer ExperimentFTIR. Reproduce the results and graphs of Figures
8.3 Computer ExperimentSurface Plasmon Resonance. Reproduce the results and graphs of
8.4 Working with the electric and magnetic elds across an negative-index slab given by Eqs. (8.6.1)
and (8.6.2), derive the reection and transmission responses of the slab given in (8.6.8).
8.5 Computer ExperimentPerfect Lens. Study the sensitivity of the perfect lens property to the
Waveguides are used to transfer electromagnetic power efciently from one point in
deviations from the ideal values of  = 0 and = 0 , and to the presence of losses by
reproducing the results and graphs of Figures 8.6.3 and 8.6.4. You will need to implement
space to another. Some common guiding structures are shown in the gure below.
the computational algorithm listed on page 329. These include the typical coaxial cable, the two-wire and mictrostrip transmission lines,
hollow conducting waveguides, and optical bers.
8.6 Computer ExperimentAntireection Coatings. Reproduce the results and graphs of Figures
In practice, the choice of structure is dictated by: (a) the desired operating frequency
band, (b) the amount of power to be transferred, and (c) the amount of transmission
8.7 Computer ExperimentOmnidirectional Dielectric Mirrors. Reproduce the results and graphs
losses that can be tolerated.
of Figures
8.8 Derive the generalized Snels laws given in Eq. (8.10.10). Moreover, derive the Brewster angle
expressions given in Eqs. (8.11.4) and (8.11.5).
8.9 Computer ExperimentBrewster angles. Study the variety of possible Brewster angles and
reproduce the results and graphs of Example 8.11.1.
8.10 Computer ExperimentMultilayer Birefringent Structures. Reproduce the results and graphs
of Figures

Fig. 9.0.1 Typical waveguiding structures.

Coaxial cables are widely used to connect RF components. Their operation is practi-
cal for frequencies below 3 GHz. Above that the losses are too excessive. For example,
the attenuation might be 3 dB per 100 m at 100 MHz, but 10 dB/100 m at 1 GHz, and
50 dB/100 m at 10 GHz. Their power rating is typically of the order of one kilowatt at
100 MHz, but only 200 W at 2 GHz, being limited primarily because of the heating of
the coaxial conductors and of the dielectric between the conductors (dielectric voltage
breakdown is usually a secondary factor.) However, special short-length coaxial cables
do exist that operate in the 40 GHz range.
Another issue is the single-mode operation of the line. At higher frequencies, in order
to prevent higher modes from being launched, the diameters of the coaxial conductors
must be reduced, diminishing the amount of power that can be transmitted.
Two-wire lines are not used at microwave frequencies because they are not shielded
and can radiate. One typical use is for connecting indoor antennas to TV sets. Microstrip
lines are used widely in microwave integrated circuits.
9.1. Longitudinal-Transverse Decompositions 363 364 9. Waveguides

Rectangular waveguides are used routinely to transfer large amounts of microwave where , denote the permittivities of the medium in which the elds propagate, for
power at frequencies greater than 3 GHz. For example at 5 GHz, the transmitted power example, the medium between the coaxial conductors in a coaxial cable, or the medium
might be one megawatt and the attenuation only 4 dB/100 m. within the hollow rectangular waveguide. This medium is assumed to be lossless for
Optical bers operate at optical and infrared frequencies, allowing a very wide band- now.
width. Their losses are very low, typically, 0.2 dB/km. The transmitted power is of the We note that z z = 1,
z z = 0,
z ET = 0, z T Ez = 0 and that
z ET and
order of milliwatts. z T Ez are transverse while T ET is longitudinal. Indeed, we have:

z ET =
z (x
Ex + y Ex x
Ey )= y Ey
9.1 Longitudinal-Transverse Decompositions T ET = (x
x + y
y )(x Ex + y z(x Ey y Ex )
Ey )=

In a waveguiding system, we are looking for solutions of Maxwells equations that are Using these properties and equating longitudinal and transverse parts in the two
propagating along the guiding direction (the z direction) and are conned in the near sides of Eq. (9.1.4), we obtain the equivalent set of Maxwell equations:
vicinity of the guiding structure. Thus, the electric and magnetic elds are assumed to
have the form:
T Ez
z j
z ET = jHT
T Hz
z j
z HT = jET

E(x, y, z, t)= E(x, y)ejtjz T ET + j

z Hz = 0
(9.1.1) (9.1.5)
H(x, y, z, t)= H(x, y)e jtjz T HT j
z Ez = 0
T ET jEz = 0
where is the propagation wavenumber along the guide direction. The corresponding T HT jHz = 0
wavelength, called the guide wavelength, is denoted by g = 2/.
Depending on whether both, one, or none of the longitudinal components are zero,
The precise relationship between and depends on the type of waveguiding struc-
we may classify the solutions as transverse electric and magnetic (TEM), transverse elec-
ture and the particular propagating mode. Because the elds are conned in the trans-
tric (TE), transverse magnetic (TM), or hybrid:
verse directions (the x, y directions,) they cannot be uniform (except in very simple
structures) and will have a non-trivial dependence on the transverse coordinates x and
y. Next, we derive the equations for the phasor amplitudes E(x, y) and H(x, y). Ez = 0, Hz = 0, TEM modes
Because of the preferential role played by the guiding direction z, it proves con- Ez = 0, Hz = 0, TE or H modes
venient to decompose Maxwells equations into components that are longitudinal, that Ez = 0, Hz = 0, TM or E modes
is, along the z-direction, and components that are transverse, along the x, y directions. Ez = 0, Hz = 0, hybrid or HE or EH modes
Thus, we decompose:
In the case of TEM modes, which are the dominant modes in two-conductor trans-
mission lines such as the coaxial cable, the elds are purely transverse and the solution
E(x, y)= x Ey (x, y) +
Ex (x, y)+y z Ez (x, y) ET (x, y)+
z Ez (x, y) (9.1.2) of Eq. (9.1.5) reduces to an equivalent two-dimensional electrostatic problem. We will
transverse longitudinal discuss this case later on.
In all other cases, at least one of the longitudinal elds Ez , Hz is non-zero. It is then
In a similar fashion we may decompose the gradient operator:
possible to express the transverse eld components ET , HT in terms of the longitudinal
ones, Ez , Hz .
x + y y +
z z = T +
z z = T j
z (9.1.3)
   Forming the cross-product of the second of equations (9.1.5) with z and using the
transverse BAC-CAB vector identity, z (z HT )= z HT )HT (
z( z z)= HT , and similarly,
where we made the replacement z j because of the assumed z-dependence. In- z (
T Hz z)= T Hz , we obtain:
troducing these decompositions into the source-free Maxwells equations we have:
T Hz + jHT = j
z ET

E = jH T j
( z)(ET +
z Ez )= j(HT +
z Hz ) Thus, the rst two of (9.1.5) may be thought of as a linear system of two equations
H = jE T j
( z)(HT +
z Hz )= j(ET +
z Ez ) in the two unknowns z ET and HT , that is,
E=0 T j
( z)(ET +
z Ez )= 0 z ET HT = j
z T Ez
H=0 T j
( z)(HT +
z Hz )= 0 T Hz
z ET HT = j

9.1. Longitudinal-Transverse Decompositions 365 366 9. Waveguides

The solution of this system is: where the medium impedance is = /, so that /c = and c = 1/. We note the
j j
z ET =
z T Ez
T Hz
k2c k2c TE 2
(9.1.7) TE TM = 2 , = 2 2 (9.1.13)
j j TM c
HT = z T Ez 2 T H z

k2c kc 
where we dened the so-called cutoff wavenumber kc by: Because c/ = 1 2c /2 , we can write also:

2 2c
k2c = 2  2 = 2 = k2 2 (cutoff wavenumber) (9.1.8) TE =  , TM = 1 (9.1.14)
c2 2 2
1 c2
The quantity k = /c =  is the wavenumber a uniform plane wave would
have in the propagation medium , .
With these denitions, we may rewrite Eq. (9.1.7) as follows:
Although k2c stands for the difference 2  2 , it turns out that the boundary
conditions for each waveguide type force k2c to take on certain values, which can be j

positive, negative, or zero, and characterize the propagating modes. For example, in a z ET =
z T Ez + TE T Hz

dielectric waveguide k2c is positive inside the guide and negative outside it; in a hollow (9.1.15)
j 1

conducting waveguide k2c takes on certain quantized positive values; in a TEM line, k2c HT = z T Ez + T Hz

is zero. Some related denitions are the cutoff frequency and the cutoff wavelength k2c TM
dened as follows: z (
Using the result z ET )= ET , we solve for ET and HT :

2 j

c = ckc , c = (cutoff frequency and wavelength) (9.1.9) ET = T Ez TE

z T Hz
kc k2c
(transverse elds) (9.1.16)
We can then express in terms of and c , or in terms of and c . Taking j 1

HT = T Hz + z T Ez

the positive square roots of Eq. (9.1.8), we have: k2c TM

 An alternative and useful way of writing these equations is to form the following
1 2 linear combinations, which are equivalent to Eq. (9.1.6):
= 2 2
c = 1 c2 and = 2c + 2 c2 (9.1.10)
c c
1 j
Often, Eq. (9.1.10) is expressed in terms of the wavelengths = 2/k = 2c/, HT z ET =
T Hz
c = 2/kc , and g = 2/. It follows from k2 = k2c + 2 that (9.1.17)
z= T Ez
1 1 1
= + g =  (9.1.11)
2 2c 2g 2
1 2 So far we only used the rst two of Maxwells equations (9.1.5) and expressed ET , HT
in terms of Ez , Hz . Using (9.1.16), it is easily shown that the left-hand sides of the
Note that is related to the free-space wavelength 0 = 2c0 / = c0 /f by the remaining four of Eqs. (9.1.5) take the forms:
refractive index of the dielectric material = 0 /n.
It is convenient at this point to introduce the transverse impedances for the TE and j 2

T ET + j
z Hz = z T Hz + k2c Hz

TM modes by the denitions: k2c
j 2

T HT j z Ez = 2 z T Ez + k2c Ez
c j

TE = = , TM = = (TE and TM impedances) (9.1.12) T ET jEz = 2 2T Ez + k2c Ez

c  kc

T HT jHz = 2 2T Hz + k2c Hz
9.1. Longitudinal-Transverse Decompositions 367 368 9. Waveguides

where 2T is the two-dimensional Laplacian operator:

2T = T T = 2x + 2y (9.1.18)
and we used the vectorial identities T T Ez = 0, T (
z T Hz )=
z T Hz , and
T (
z T Hz )= 0.
It follows that in order to satisfy all of the last four of Maxwells equations (9.1.5), it
is necessary that the longitudinal elds Ez (x, y), Hz (x, y) satisfy the two-dimensional
Helmholtz equations: Fig. 9.1.1 Cylindrical coordinates.

2T Ez + k2c Ez = 0 The Helmholtz equations (9.1.19) now read:

(Helmholtz equations) (9.1.19)
T Hz + kc Hz = 0

1 Ez 1 2 Ez
These equations are to be solved subject to the appropriate boundary conditions for
+ + k2c Ez = 0
2 2
each waveguide type. Once, the elds Ez , Hz are known, the transverse elds ET , HT are (9.1.23)
computed from Eq. (9.1.16), resulting in a complete solution of Maxwells equations for 1 Hz 1 2 Hz 2
+ + k c Hz = 0
the guiding structure. To get the full x, y, z, t dependence of the propagating elds, the 2 2
above solutions must be multiplied by the factor ejtjz .
Noting that and
= =
z , we obtain:
The cross-sections of practical waveguiding systems have either cartesian or cylin-
drical symmetry, such as the rectangular waveguide or the coaxial cable. Below, we 1
( Hz )
z T Hz = ( Hz )
summarize the form of the above solutions in the two types of coordinate systems.

The decomposition of a transverse vector is ET = E . The cylindrical

E +
Cartesian Coordinates
coordinates version of (9.1.16) are:
The cartesian component version of Eqs. (9.1.16) and (9.1.19) is straightforward. Using
z T Hz = y
the identity x Hz x
y Hz , we obtain for the longitudinal components:
j 1
j 1

E = Ez TE Hz H = Hz + Ez
(x + y )Ez + kc Ez = 0
2 2 2 k2c k2c TM
(9.1.20) , (9.1.24)
(2x + 2y )Hz + k2c Hz = 0 j 1
j 1 1

E = Ez + TE Hz H = Hz Ez
k2c k2c TM
Eq. (9.1.16) becomes for the transverse components:
For either coordinate system, the equations for HT may be obtained from those of
ET by a so-called duality transformation, that is, making the substitutions:
j 1

Ex = x Ez + TE y Hz Hx = x Hz y Ez
k2c k2c TM
, (9.1.21) E H, H E ,  ,  (duality transformation) (9.1.25)
j 1

Ey = 2 y Ez TE x Hz Hy = 2 y Hz + x Ez
kc kc TM These imply that 1 and TE 1
TM . Duality is discussed in greater detail in
Sec. 18.2.
Cylindrical Coordinates
The relationship between cartesian and cylindrical coordinates is shown in Fig. 9.1.1. 9.2 Power Transfer and Attenuation
From the triangle in the gure, we have x = cos and y = sin . The transverse
gradient and Laplace operator are in cylindrical coordinates: With the eld solutions at hand, one can determine the amount of power transmitted
along the guide, as well as the transmission losses. The total power carried by the elds
1 , 1 1 2 along the guide direction is obtained by integrating the z-component of the Poynting
T =
+ 2T = + (9.1.22)
2 2 vector over the cross-sectional area of the guide:
9.2. Power Transfer and Attenuation 369 370 9. Waveguides

Second, the magnetic elds on the conductor surfaces are determined and the corre-

1 sponding induced surface currents are calculated by Js = n H, where n is the outward
PT = Pz dS , where Pz = Re(E H )
z (9.2.1)
S 2 normal to the conductor.
Third, the ohmic losses per unit conductor area are calculated by Eq. (2.8.7). Figure
It is easily veried that only the transverse components of the elds contribute to
9.2.1 shows such an innitesimal conductor area dA = dl dz, where dl is along the
the power ow, that is, Pz can be written in the form:
cross-sectional periphery of the conductor. Applying Eq. (2.8.7) to this area, we have:
Pz = Re(ET H z
T ) (9.2.2) dPloss dPloss 1
2 = = Rs |Js |2 (9.2.6)
dA dldz 2
For waveguides with conducting walls, the transmission losses are due primarily to
where Rs is the surface resistance of the conductor given by Eq. (2.8.4),
ohmic losses in (a) the conductors and (b) the dielectric medium lling the space between
the conductors and in which the elds propagate. In dielectric waveguides, the losses  1 2
are due to absorption and scattering by imperfections. Rs = = = , = = skin depth (9.2.7)
2 2
The transmission losses can be quantied by replacing the propagation wavenumber
by its complex-valued version c = j, where is the attenuation constant. The Integrating Eq. (9.2.6) around the periphery of the conductor gives the power loss per
z-dependence of all the eld components is replaced by: unit z-length due to that conductor. Adding similar terms for all the other conductors
gives the total power loss per unit z-length:
ejz ejc z = e(+j)z = ez ejz (9.2.3)  
dPloss 1 1
The quantity is the sum of the attenuation constants arising from the various loss Ploss = = Rs |Js |2 dl + Rs |Js |2 dl (9.2.8)
dz Ca 2 Cb 2
mechanisms. For example, if d and c are the attenuations due to the ohmic losses in
the dielectric and in the conducting walls, then

= d + c (9.2.4)

The ohmic losses in the dielectric can be characterized either by its loss tangent,
say tan , or by its conductivity d the two being related by d =  tan . More
generally, the effective dielectric constant of the medium may have a negative imaginary
part I that includes both conductive and polarization losses, ()=  jI , with
I =  tan . Then, the corresponding complex-valued wavenumber c is obtained by
the replacement:
= 2  k2c c = 2 ()k2c Fig. 9.2.1 Conductor surface absorbs power from the propagating elds.

For weakly lossy dielectrics (I ), we may make the approximation: where Ca and Cb indicate the peripheries of the conductors. Finally, the corresponding
 attenuation coefcient is calculated from Eq. (2.6.22):
  2 I 2 I
c = 2 ( jI )k2c = 2 j2 I = 1j
2 2
 c = (conductor losses) (9.2.9)
Resulting in the attenuation constant, after setting  = 1/c2 and c/ = 1 2c /2 ,
Equations (9.2.1)(9.2.9) provide a systematic methodology by which to calculate the
2 I 1 2  tan transmitted power and attenuation losses in waveguides. We will apply it to several
d = = tan =  (dielectric losses) (9.2.5)
2 2 2c 1 2c /2 examples later on. Eq. (9.2.9) applies also to the dielectric losses so that in general Ploss
arises from two parts, one due to the dielectric and one due to the conducting walls,
The conductor losses are more complicated to calculate. In practice, the following
Ploss P + Pcond

approximate procedure is adequate. First, the elds are determined on the assumption = = diel = d + c (attenuation constant) (9.2.10)
that the conductors are perfect.
9.3. TEM, TE, and TM modes 371 372 9. Waveguides

Eq. (9.2.5) for d can also be derived directly from Eq. (9.2.10) by applying it sepa-
rately to the TE and TM modes. We recall from Eq. (1.9.6) that the losses per unit vol- 1
HT = z ET
ume in a dielectric medium, arising from both a conduction and polarization current, T
Jtot = J + jD, are given by,
where T is the transverse impedance of the particular mode type, that is, , TE , TM
dPloss 1   1   in the TEM, TE, and TM cases.
= Re Jtot E = I E E 
dV 2 2 Because of Eq. (9.3.1), the power ow per unit cross-sectional area described by the
Poynting vector Pz of Eq. (9.2.2) takes the simple form in all three cases:
Integrating over the cross-sectional area of the guide gives the dielectric loss per unit
waveguide length (i.e., z-length),
1 1 1
Pz = Re(ET H z=
T ) |ET |2 = T |HT |2 (9.3.2)
1 2 2T 2
Pdiel = I |E|2 dS
2 S

Applying this to the TE case, we nd, TEM modes

In TEM modes, both Ez and Hz vanish, and the elds are fully transverse. One can set
1 1
Pdiel = I |E|2 dS = I |ET |2 dS Ez = Hz = 0 in Maxwell equations (9.1.5), or equivalently in (9.1.16), or in (9.1.17).
2 S 2 S
From any point view, one obtains the condition k2c = 0, or = c. For example, if
1 1
PT = Re(ET H z dS =
T ) |ET |2 dS = |ET |2 dS the right-hand sides of Eq. (9.1.17) vanish, the consistency of the system requires that
S 2 2TE S 2 S TE = TM , which by virtue of Eq. (9.1.13) implies = c. It also implies that TE , TM
must both be equal to the medium impedance . Thus, the electric and magnetic elds
Pdiel 2 I
d = = satisfy:
2PT 2

The TM case is a bit more involved. Using Eq. (9.13.1) from Problem 9.11, we nd, 1
after using the result, 2 + k2c = 2 , HT = z ET

1 1   These are the same as in the case of a uniform plane wave, except here the elds
Pdiel = I |E|2 dS = I |Ez |2 + |ET |2 dS
2 S 2 S are not uniform and may have a non-trivial x, y dependence. The electric eld ET is
determined from the rest of Maxwells equations (9.1.5), which read:
1 2 1 2
= I
|Ez |2 +
| E
T z | 2
dS =  I 1 + |Ez |2 dS
2 S k4c 2 kc2 S
2 T ET = 0
1   (9.3.4)
PT = |ET |2 dS = 2
4 | T Ez | dS = 2 |Ez |2 dS T ET = 0
2TM S 2 S kc 2kc S

1 2 These are recognized as the eld equations of an equivalent two-dimensional elec-
 I 1 +
P 2 k2c 2 I trostatic problem. Once this electrostatic solution is found, ET (x, y), the magnetic eld
d = diel = = is constructed from Eq. (9.3.3). The time-varying propagating elds will be given by
2k2c Eq. (9.1.1), with = c. (For backward moving elds, replace by .)
We explore this electrostatic point of view further in Sec. 11.1 and discuss the cases
of the coaxial, two-wire, and strip lines. Because of the relationship between ET and HT ,
9.3 TEM, TE, and TM modes
the Poynting vector Pz of Eq. (9.2.2) will be:
The general solution described by Eqs. (9.1.16) and (9.1.19) is a hybrid solution with non- 1 1 1
zero Ez and Hz components. Here, we look at the specialized forms of these equations Pz = Re(ET H z=
T ) |ET |2 = |HT |2 (9.3.5)
2 2 2
in the cases of TEM, TE, and TM modes.
One common property of all three types of modes is that the transverse elds ET , HT
are related to each other in the same way as in the case of uniform plane waves propagat-
ing in the z-direction, that is, they are perpendicular to each other, their cross-product
points in the z-direction, and they satisfy:
9.3. TEM, TE, and TM modes 373 374 9. Waveguides

TE modes
2T Ez + k2c Ez = 0
TE modes are characterized by the conditions Ez = 0 and Hz = 0. It follows from the
second of Eqs. (9.1.17) that ET is completely determined from HT , that is, ET = TE HT
z. j
The eld HT is determined from the second of (9.1.16). Thus, all eld components ET = T Ez
k2c (TM modes) (9.3.10)
for TE modes are obtained from the equations:
HT = z ET

2T Hz + k2c Hz = 0
Again, the relationship of ET and HT is identical to that of uniform plane waves
HT = T Hz (TE modes) (9.3.6) propagating in the z-direction, but the wave impedance is now TM . The Poynting vector
takes the form:
1 1 1 2
Pz = Re(ET H z=
T ) |ET |2 = T Ez |2
| (9.3.11)
The relationship of ET and HT is identical to that of uniform plane waves propagating 2 2TM 2TM k4c
in the z-direction, except the wave impedance is replaced by TE . The Poynting vector
of Eq. (9.2.2) then takes the form: 9.4 Rectangular Waveguides
Next, we discuss in detail the case of a rectangular hollow waveguide with conducting
1 1 1 1 2
Pz = Re(ET H z=
T ) T H z |2
|ET |2 = TE |HT |2 = TE 4 | (9.3.7) walls, as shown in Fig. 9.4.1. Without loss of generality, we may assume that the lengths
2 2TE 2 2 kc
a, b of the inner sides satisfy b a. The guide is typically lled with air, but any other
The cartesian coordinate version of Eq. (9.3.6) is: dielectric material , may be assumed.

(2x + 2y )Hz + k2c Hz = 0

j j
Hx = x Hz , Hy = 2 y Hz (9.3.8)
k2c kc
Ex = TE Hy , Ey = TE Hx

And, the cylindrical coordinate version:

Fig. 9.4.1 Rectangular waveguide.

1 Hz 1 2 Hz 2
+ + kc Hz = 0 The simplest and dominant propagation mode is the so-called TE10 mode and de-
2 2
pends only on the x-coordinate (of the longest side.) Therefore, we begin by looking
j Hz j 1 Hz (9.3.9)
H = , H = 2 for solutions of Eq. (9.3.8) that depend only on x. In this case, the Helmholtz equation
k2c kc
reduces to:
E = TE H , E = TE H
2x Hz (x)+k2c Hz (x)= 0
where we used HT
z = ( H )
H + H +
z = H .
The most general solution is a linear combination of cos kc x and sin kc x. However,
TM modes only the former will satisfy the boundary conditions. Therefore, the solution is:

TM modes have Hz = 0 and Ez = 0. It follows from the rst of Eqs. (9.1.17) that HT is Hz (x)= H0 cos kc x (9.4.1)
completely determined from ET , that is, HT = 1
z ET . The eld ET is determined
where H0 is a (complex-valued) constant. Because there is no y-dependence, it follows
from the rst of (9.1.16), so that all eld components for TM modes are obtained from
from Eq. (9.3.8) that y Hz = 0, and hence Hy = 0 and Ex = 0. It also follows that:
the following equations, which are dual to the TE equations (9.3.6):
j j j
Hx (x)= x Hz = 2 (kc )H0 sin kc x = H0 sin kc x H1 sin kc x
k2c kc kc
9.4. Rectangular Waveguides 375 376 9. Waveguides

Then, the corresponding electric eld will be:

Ey (x)= TE Hx (x)= TE H0 sin kc x E0 sin kc x

where we dened the constants:

j Fig. 9.4.2 Electric eld inside a rectangular waveguide.

H1 = H0
j 9.5 Higher TE and TM modes
E0 = TE H1 = TE H0 = j H0
kc c
where we used TE = /c. In summary, the non-zero eld components are: To construct higher modes, we look for solutions of the Helmholtz equation that are
factorable in their x and y dependence:

Hz (x, y)= F(x)G(y)

Hz (x)= H0 cos kc x Hz (x, y, z, t)= H0 cos kc x ejtjz
Hx (x)= H1 sin kc x Hx (x, y, z, t)= H1 sin kc x ejtjz (9.4.3) Then, Eq. (9.3.8) becomes:

Ey (x)= E0 sin kc x jtjz

Ey (x, y, z, t)= E0 sin kc x e
F (x) G (y)
F (x)G(y)+F(x)G (y)+k2c F(x)G(y)= 0 + + k2c = 0 (9.5.1)
Assuming perfectly conducting walls, the boundary conditions require that there be F(x) G(y)
no tangential electric eld at any of the wall sides. Because the electric eld is in the Because these must be valid for all x, y (inside the guide), the F- and G-terms must
y-direction, it is normal to the top and bottom sides. But, it is parallel to the left and be constants, independent of x and y. Thus, we write:
right sides. On the left side, x = 0, Ey (x) vanishes because sin kc x does. On the right
F (x) G (y)
side, x = a, the boundary condition requires: = k2x , = k2y or
F(x) G(y)
Ey (a)= E0 sin kc a = 0 sin kc a = 0
F (x)+k2x F(x)= 0 , G (y)+k2y G(y)= 0 (9.5.2)
This requires that kc a be an integral multiple of : where the constants k2x and k2y are constrained from Eq. (9.5.1) to satisfy:
kc a = n kc = (9.4.4) k2c = k2x + k2y (9.5.3)
These are the so-called TEn0 modes. The corresponding cutoff frequency c = ckc , The most general solutions of (9.5.2) that will satisfy the TE boundary conditions are
fc = c /2, and wavelength c = 2/kc = c/fc are: cos kx x and cos ky y. Thus, the longitudinal magnetic eld will be:

cn cn 2a Hz (x, y)= H0 cos kx x cos ky y (TEnm modes) (9.5.4)

c = , fc = , c = (TEn0 modes) (9.4.5)
a 2a n
It then follows from the rest of the equations (9.3.8) that:
The dominant mode is the one with the lowest cutoff frequency or the longest cutoff
wavelength, that is, the mode TE10 having n = 1. It has:
Hx (x, y) = H1 sin kx x cos ky y Ex (x, y) = E1 cos kx x sin ky y
c c (9.5.5)
kc = , c = , fc = , c = 2a (TE10 mode) (9.4.6) Hy (x, y) = H2 cos kx x sin ky y Ey (x, y) = E2 sin kx x cos ky y
a a 2a

Fig. 9.4.2 depicts the electric eld Ey (x)= E0 sin kc x = E0 sin(x/a) of this mode where we dened the constants:
as a function of x. jkx jky
H1 = H0 , H2 = H0
k2c k2c
ky kx
E1 = TE H2 = j H0 , E2 = TE H1 = j H0
c kc c kc
9.5. Higher TE and TM modes 377 378 9. Waveguides

The boundary conditions are that Ey vanish on the right wall, x = a, and that Ex 9.6 Operating Bandwidth
vanish on the top wall, y = b, that is,
All waveguiding systems are operated in a frequency range that ensures that only the
Ey (a, y)= E0y sin kx a cos ky y = 0 , Ex (x, b)= E0x cos kx x sin ky b = 0 lowest mode can propagate. If several modes can propagate simultaneously, one has
no control over which modes will actually be carrying the transmitted signal. This may
The conditions require that kx a and ky b be integral multiples of : cause undue amounts of dispersion, distortion, and erratic operation.
A mode with cutoff frequency c will propagate only if its frequency is c ,
n m or < c . If < c , the wave will attenuate exponentially along the guide direction.
kx a = n , ky b = m kx = , ky = (9.5.6)
a b This follows from the , relationship (9.1.10):

 correspond to the TEnm modes. Thus, the cutoff wavenumbers of these modes
These 2 2c
kc = k2x + k2y take on the quantized values: 2 = 2c + 2 c2 2 =
 2  2 If c , the wavenumber is real-valued and the wave will propagate. But if
n m
kc = + (TEnm modes) (9.5.7) < c , becomes imaginary, say, = j, and the wave will attenuate in the z-
a b direction, with a penetration depth = 1/:
The cutoff frequencies fnm = c /2 = ckc /2 and wavelengths nm = c/fnm are:
ejz = ez
 2  2
n m 1
fnm = c + , nm =  2  2 (9.5.8) If the frequency is greater than the cutoff frequencies of several modes, then all
2a 2b n m
+ of these modes can propagate. Conversely, if is less than all cutoff frequencies, then
2a 2b none of the modes can propagate.
The TE0m modes are similar to the TEn0 modes, but with x and a replaced by y and If we arrange the cutoff frequencies in increasing order, c1 < c2 < c3 < ,
b. The family of TM modes can also be constructed in a similar fashion from Eq. (9.3.10). then, to ensure single-mode operation, the frequency must be restricted to the interval
Assuming Ez (x, y)= F(x)G(y), we obtain the same equations (9.5.2). Because Ez c1 < < c2 , so that only the lowest mode will propagate. This interval denes the
is parallel to all walls, we must now choose the solutions sin kx and sin ky y. Thus, the operating bandwidth of the guide.
longitudinal electric elds is: These remarks apply to all waveguiding systems, not just hollow conducting wave-
guides. For example, in coaxial cables the lowest mode is the TEM mode having no cutoff
Ez (x, y)= E0 sin kx x sin ky y (TMnm modes) (9.5.9) frequency, c1 = 0. However, TE and TM modes with non-zero cutoff frequencies do
exist and place an upper limit on the usable bandwidth of the TEM mode. Similarly, in
The rest of the eld components can be worked out from Eq. (9.3.10) and one nds optical bers, the lowest mode has no cutoff, and the single-mode bandwidth is deter-
that they are given by the same expressions as (9.5.5), except now the constants are mined by the next cutoff frequency.
determined in terms of E0 : In rectangular waveguides, the smallest cutoff frequencies are f10 = c/2a, f20 =
c/a = 2f10 , and f01 = c/2b. Because we assumed that b a, it follows that always
jkx jky
E1 = E0 , E2 = E0 f10 f01 . If b a/2, then 1/a 1/2b and therefore, f20 f01 , so that the two lowest
k2c k2c
cutoff frequencies are f10 and f20 .
1 jky 1 1 jkx 1 On the other hand, if a/2 b a, then f01 f20 and the two smallest frequencies
H1 = E2 = E0 , H2 = E1 = H0
TM c kc TM c kc are f10 and f01 (except when b = a, in which case f01 = f10 and the smallest frequencies
are f10 and f20 .) The two cases b a/2 and b a/2 are depicted in Fig. 9.6.1.
where we used TM = c/. The boundary conditions on Ex , Ey are the same as It is evident from this gure that in order to achieve the widest possible usable
before, and in addition, we must require that Ez vanish on all walls. bandwidth for the TE10 mode, the guide dimensions must satisfy b a/2 so that the
These conditions imply that kx , ky will be given by Eq. (9.5.6), except both n and m bandwidth is the interval [fc , 2fc ], where fc = f10 = c/2a. In terms of the wavelength
must be non-zero (otherwise Ez would vanish identically.) Thus, the cutoff frequencies = c/f , the operating bandwidth becomes: 0.5 a/ 1, or, a 2a.
and wavelengths are the same as in Eq. (9.5.8). We will see later that the total amount of transmitted power in this mode is propor-
Waveguide modes can be excited by inserting small probes at the beginning of the tional to the cross-sectional area of the guide, ab. Thus, if in addition to having the
waveguide. The probes are chosen to generate an electric eld that resembles the eld
Murphys law for waveguides states that if a mode can propagate, it will.
of the desired mode.
9.7. Power Transfer, Energy Density, and Group Velocity 379 380 9. Waveguides


H1 = H0 , E0 = TE H1 = j H0 (9.7.2)
kc c
The Poynting vector is obtained from the general result of Eq. (9.3.7):

1 1 1
Pz = |ET |2 = |Ey (x)|2 = |E0 |2 sin2 kc x

The transmitted power is obtained by integrating Pz over the cross-sectional area

Fig. 9.6.1 Operating bandwidth in rectangular waveguides.
of the guide:
a b
widest bandwidth, we also require to have the maximum power transmitted, the dimen- 1
PT = |E0 |2 sin2 kc x dxdy
sion b must be chosen to be as large as possible, that is, b = a/2. Most practical guides 0 0 2TE
follow these side proportions.
Noting the denite integral,
If there is a canonical guide, it will have b = a/2 and be operated at a frequency
a a
that lies in the middle of the operating band [fc , 2fc ], that is, x
sin2 kc x dx = sin2 dx = (9.7.3)
c 0 0 a 2
f = 1.5fc = 0.75 (9.6.1) 
a and using TE = /c = / 1 2c /2 , we obtain:
Table 9.6.1 lists some standard air-lled rectangular waveguides with their naming
designations, inner side dimensions a, b in inches, cutoff frequencies in GHz, minimum

and maximum recommended operating frequencies in GHz, power ratings, and attenua- 1 1 2c
tions in dB/m (the power ratings and attenuations are representative over each operating PT = |E0 | ab = |E0 |2 ab 1 (transmitted power) (9.7.4)
4TE 4 2
band.) We have chosen one example from each microwave band.
We may also calculate the distribution of electromagnetic energy along the guide, as
name a b fc fmin fmax band P
measured by the time-averaged energy density. The energy densities of the electric and
WR-510 5.10 2.55 1.16 1.45 2.20 L 9 MW 0.007 magnetic elds are:
WR-284 2.84 1.34 2.08 2.60 3.95 S 2.7 MW 0.019
WR-159 1.59 0.795 3.71 4.64 7.05 C 0.9 MW 0.043 1 1
WR-90 0.90 0.40 6.56 8.20 12.50 X 250 kW 0.110 we = Re E E = |Ey |2
2 2 4
WR-62 0.622 0.311 9.49 11.90 18.00 Ku 140 kW 0.176
WR-42 0.42 0.17 14.05 17.60 26.70 K 50 kW 0.370 1 1

wm = Re H H = |Hx |2 + |Hz |2
WR-28 0.28 0.14 21.08 26.40 40.00 Ka 27 kW 0.583 2 2 4
WR-15 0.148 0.074 39.87 49.80 75.80 V 7.5 kW 1.52
WR-10 0.10 0.05 59.01 73.80 112.00 W 3.5 kW 2.74 Inserting the expressions for the elds, we nd:

1 1

Table 9.6.1 Characteristics of some standard air-lled rectangular waveguides. we = |E0 |2 sin2 kc x , wm = |H1 |2 sin2 kc x + |H0 |2 cos2 kc x
4 4

Because these quantities represent the energy per unit volume, if we integrate them
over the cross-sectional area of the guide, we will obtain the energy distributions per
9.7 Power Transfer, Energy Density, and Group Velocity unit z-length. Using the integral (9.7.3) and an identical one for the cosine case, we nd:

Next, we calculate the time-averaged power transmitted in the TE10 mode. We also calcu- a b a b
1 1
late the energy density of the elds and determine the velocity by which electromagnetic We = we (x, y) dxdy = |E0 |2 sin2 kc x dxdy = |E0 |2 ab
0 0 0 0 4 8
energy ows down the guide and show that it is equal to the group velocity. We recall a b
that the non-zero eld components are: 1

Wm = |H1 |2 sin2 kc x + |H0 |2 cos2 kc x dxdy = |H1 |2 + |H0 |2 ab

0 0 4 8
Hz (x)= H0 cos kc x , Hx (x)= H1 sin kc x , Ey (x)= E0 sin kc x (9.7.1)
9.8. Power Attenuation 381 382 9. Waveguides

Although these expressions look different, they are actually equal, We = Wm

. In- The eld expressions (9.4.3) were derived assuming the boundary conditions for
deed, using the property 2 /k2c + 1 = (2 + k2c )/k2c = k2 /k2c = 2 /2c and the relation- perfectly conducting wall surfaces. The induced surface currents on the inner walls of
ships between the constants in (9.7.1), we nd: the waveguide are given by Js = n H, where the unit vector n
is x
and y on the
left/right and bottom/top walls, respectively.

|H1 |2 + |H0 |2 = |H0 |2 2 + |H0 |2 = |H0 |2 2 = 2 |E0 |2 = |E0 |2 The surface currents and tangential magnetic elds are shown in Fig. 9.8.1. In par-
kc c ticular, on the bottom and top walls, we have:
The equality of the electric and magnetic energies is a general property of wavegui-
ding systems. We also encountered it in Sec. 2.3 for uniform plane waves. The total
energy density per unit length will be:

W = We + Wm

= 2We = |E0 |2 ab (9.7.5)

According to the general relationship between ux, density, and transport velocity
given in Eq. (1.6.2), the energy transport velocity will be the ratio ven = PT /W . Using

Eqs. (9.7.4) and (9.7.5) and noting that 1/ = 1/  = c, we nd:

 Fig. 9.8.1 Currents on waveguide walls.

PT 2
ven = =c 1 c2 (energy transport velocity) (9.7.6)
Js = y
H = y
Hx +
zHz )= (
z Hx + x
Hz )= (
z H1 sin kc x + x
H0 cos kc x)
This is equal to the group velocity of the propagating mode. For any dispersion
relationship between and , the group and phase velocities are dened by Similarly, on the left and right walls:

d Js = x
H = x
Hx +
zHz )= y
Hz = y
H0 cos kc x
vgr = , vph = (group and phase velocities) (9.7.7)
At x = 0 and x = a, this gives Js = y
(H0 )= y
H0 . Thus, the magnitudes of the
For uniform plane waves and TEM transmission lines, we have = c, so that vgr = surface currents are on the four walls:
vph = c. For a rectangular waveguide, we have 2 = 2c + 2 c2 . Taking differentials of 
|H0 |2 , (left and right walls)
both sides, we nd 2d = 2c2 d, which gives: 2
|Js | =
|H0 |2 cos2 kc x + |H1 |2 sin2 kc x , (top and bottom walls)

d c2 2c The power loss per unit z-length is obtained from Eq. (9.2.8) by integrating |Js |2
vgr = = =c 1 (9.7.8)
d 2 around the four walls, that is,
a b
where we used Eq. (9.1.10). Thus, the energy transport velocity is equal to the group 1 1
velocity, ven = vgr . We note that vgr = c2 / = c2 /vph , or
Ploss = 2 Rs |Js |2 dx + 2 Rs |Js |2 dy
2 0 2 0
a b

vgr vph = c2 (9.7.9) = Rs |H0 |2 cos2 kc x + |H1 |2 sin2 kc x dx + Rs |H0 |2 dy

0 0
The energy or group velocity satises vgr c, whereas vph c. Information trans-
Rs a 2b

mission down the guide is by the group velocity and, consistent with the theory of = Rs |H0 |2 + |H1 |2 + Rs b|H0 |2 = |H0 |2 + |H1 |2 + |H0 |2
relativity, it is less than c. 2 2 a

Using |H0 |2 +|H1 |2 = |E0 |2 /2 from Sec. 9.7, and |H0 |2 = (|E0 |2 /2 )c2 /2 , which
follows from Eq. (9.4.2), we obtain:
9.8 Power Attenuation
Rs a|E0 |2 2b 2c
In this section, we calculate the attenuation coefcient due to the ohmic losses of the Ploss = 1+
22 a 2
conducting walls following the procedure outlined in Sec. 9.2. The losses due to the
lling dielectric can be determined from Eq. (9.2.5). The attenuation constant is computed from Eqs. (9.2.9) and (9.7.4):
9.8. Power Attenuation 383 384 9. Waveguides

The cutoff frequency of the TE10 mode is fc = c/2a = 3.71 GHz. The maximum operating
bandwidth is the interval [fc , 2fc ]= [3.71, 7.42] GHz, and the recommended interval is
Rs a|E0 |2 2b 2c [4.64, 7.05] GHz.
Ploss 22 a 2
c = =  Assuming copper walls with conductivity = 5.8107 S/m, the calculated attenuation
2PT 1 2 constant c from Eq. (9.8.1) is plotted in dB/m versus frequency in Fig. 9.8.2.
2 |E0 |2 ab 1 c2
Attenuation Coefficient Power Transmitted
0.1 1.5
which gives:
2b 2c
1+ 1
Rs a 2


c =  (attenuation of TE10 mode) (9.8.1) 0.06
b 2c
2 0.04

This is in units of nepers/m. Its value in dB/m is obtained by dB = 8.686c . For a 0.02

given ratio a/b, c increases with decreasing b, thus the smaller the guide dimensions,
the larger the attenuation. This trend is noted in Table 9.6.1. 0 0
0 1 2 3 4 5 6 7 8 9 10 11 12 0 1 2 3 4 5 6 7 8 9 10 11 12
The main tradeoffs in a waveguiding system are that as the operating frequency f f (GHz) f (GHz)
increases, the dimensions of the guide must decrease in order to maintain the operat-
ing band fc f 2fc , but then the attenuation increases and the transmitted power Fig. 9.8.2 Attenuation constant and transmitted power in a WR-159 waveguide.
decreases as it is proportional to the guides area.
The power transmitted PT is calculated from Eq. (9.7.4) assuming a maximum breakdown
Example 9.8.1: Design a rectangular air-lled waveguide to be operated at 5 GHz, then, re- voltage of E0 = 1.5 MV/m, which gives a safety factor of two over the dielectric breakdown
design it to be operated at 10 GHz. The operating frequency must lie in the middle of the of air of 3 MV/m. The power in megawatt scales is plotted in Fig. 9.8.2.
operating band. Calculate the guide dimensions, the attenuation constant in dB/m, and 
the maximum transmitted power assuming the maximum electric eld is one-half of the Because of the factor 1 2c /2 in the denominator of c and the numerator of PT ,
dielectric strength of air. Assume copper walls with conductivity = 5.8107 S/m. the attenuation constant becomes very large near the cutoff frequency, while the power is
almost zero. A physical explanation of this behavior is given in the next section. 

Solution: If f is in the middle of the operating band, fc f 2fc , where fc = c/2a, then
f = 1.5fc = 0.75c/a. Solving for a, we nd
9.9 Reection Model of Waveguide Propagation
0.75c 0.7530 GHz cm
a= = = 4.5 cm
f 5
An intuitive model for the TE10 mode can be derived by considering a TE-polarized
uniform plane wave propagating in the z-direction by obliquely bouncing back and forth
For maximum power transfer, we require b = a/2 = 2.25 cm. Because = 1.5c , we
have c / = 2/3. Then, Eq. (9.8.1) gives c = 0.037 dB/m. The dielectric strength of air between the left and right walls of the waveguide, as shown in Fig. 9.9.1.
is 3 MV/m. Thus, the maximum allowed electric eld in the guide is E0 = 1.5 MV/m. Then, If is the angle of incidence, then the incident and reected (from the right wall)
Eq. (9.7.4) gives PT = 1.12 MW. wavevectors will be:
At 10 GHz, because f is doubled, the guide dimensions are halved, a = 2.25 and b = 1.125 k=x
kx +
z kz = x
k cos +
z k sin

cm. Because Rs depends on f like f 1/2 , it will increase by a factor of 2. Then, the factor

Rs /b will increase by a factor of 2 2. Thus, the attenuation will increase to the value k = x
kx +
z kz = x
k cos +
z k sin

c = 0.037 2 2 = 0.104 dB/m. Because the area ab is reduced by a factor of four, so
will the power, PT = 1.12/4 = 0.28 MW = 280 kW. The electric and magnetic elds will be the sum of an incident and a reected com-
The results of these two cases are consistent with the values quoted in Table 9.6.1 for the ponent of the form:
C-band and X-band waveguides, WR-159 and WR-90. 

E1 ejk r = y
E1 ejkr + y
E=y E1 ejkx x ejkz z = E1 + E 1
E1 ejkx x ejkz z + y
Example 9.8.2: WR-159 Waveguide. Consider the C-band WR-159 air-lled waveguide whose
1 1
characteristics were listed in Table 9.6.1. Its inner dimensions are a = 1.59 and b = a/2 = H= k E1 + k E 1
0.795 inches, or, equivalently, a = 4.0386 and b = 2.0193 cm.

9.9. Reection Model of Waveguide Propagation 385 386 9. Waveguides

The boundary condition on the right wall requires sin kx a = 0, which gives rise to
the same condition as (9.4.4), that is, kc a = n.
This model claries also the meaning of the group velocity. The plane wave is bounc-
ing left and right with the speed of light c. However, the component of this velocity in
the z-direction will be vz = c sin . This is equal to the group velocity. Indeed, it follows
from Eq. (9.9.3) that: 
vz = c sin = c 1 = vgr (9.9.5)
Eq. (9.9.3) implies also that at = c , we have sin = 0, or = 0, that is, the wave
is bouncing left and right at normal incidence, creating a standing wave, and does not
propagate towards the z-direction. Thus, the transmitted power is zero and this also
Fig. 9.9.1 Reection model of TE10 mode. implies, through Eq. (9.2.9), that c will be innite.
On the other hand, for very large frequencies,  c , the angle will tend to 90o ,
causing the wave to zoom through guide almost at the speed of light.
where the electric eld was taken to be polarized in the y direction. These eld expres- The phase velocity can also be understood geometrically. Indeed, we observe in the
sions become component-wise: rightmost illustration of the above gure that the planes of constant phase are moving

obliquely with the speed of light c. From the indicated triangle at points 1,2,3, we see that
Ey = E1 ejkx x + E1 ejkx x ejkz z
the effective speed in the z-direction of the common-phase points will be vph = c/ sin
so that vph vgr = c2 .
Hx = sin E1 ejkx x + E1 ejkx x ejkz z
(9.9.1) Higher TE and TM modes can also be given similar geometric interpretations in terms
of plane waves propagating by bouncing off the waveguide walls [890].
Hz = cos E1 ejkx x E1 ejkx x ejkz z

The boundary condition on the left wall, x = 0, requires that E1 + E1 = 0. We may write 9.10 Resonant Cavities
therefore, E1 = E1 = jE0 /2. Then, the above expressions simplify into:
Cavity resonators are metallic enclosures that can trap electromagnetic elds. The
Ey = E0 sin kx x ejkz z boundary conditions on the cavity walls force the elds to exist only at certain quantized
resonant frequencies. For highly conducting walls, the resonances are extremely sharp,
Hx = sin E0 sin kx x ejkz z having a very high Q of the order of 10,000.
Because of their high Q , cavities can be used not only to efciently store electro-
j magnetic energy at microwave frequencies, but also to act as precise oscillators and to
Hz = cos E0 cos kx x ejkz z
perform precise frequency measurements.
Fig. 9.10.1 shows a rectangular cavity with z-length equal to l formed by replacing
These are identical to Eq. (9.4.3) provided we identify with kz and kc with kx , as the sending and receiving ends of a waveguide by metallic walls. A forward-moving wave
shown in Fig. 9.9.1. It follows from the wavevector triangle in the gure that the angle will bounce back and forth from these walls, resulting in a standing-wave pattern along
of incidence will be given by cos = kx /k = kc /k, or, the z-direction.

c 2c
cos = , sin = 1 (9.9.3)

The ratio of the transverse components, Ey /Hx , is the transverse impedance, which
is recognized to be TE . Indeed, we have:

TE = = =  (9.9.4)
Hx sin 2 Fig. 9.10.1 Rectangular cavity resonator (and induced wall currents for the TEn0p mode.)
1 c2

9.10. Resonant Cavities 387 388 9. Waveguides

Because the tangential components of the electric eld must vanish at the end-walls, where we used the following denite integrals (valid because kc = n/a, = p/l) :
these walls must coincide with zero crossings of the standing wave, or put differently, an a a l l
integral multiple of half-wavelengths must t along the z-direction, that is, l = pg /2 =
a l
sin2 kc x dx = cos2 kc x dx = , sin2 z dz = cos2 z dz = (9.10.5)
p/, or = p/l, where p is a non-zero integer. For the same reason, the standing- 0 0 2 0 0 2
wave patterns along the transverse directions require a = nx /2 and b = my /2, or The ohmic losses are calculated from Eq. (9.2.6), integrated over all six cavity sides.
kx = n/a and ky = m/b. Thus, all three cartesian components of the wave vector
 The surface currents induced on the walls are related to the tangential magnetic elds
are quantized, and therefore, so is the frequency of the wave = c kx2 + k2y + 2 : by J s = n
Htan . The directions of these currents are shown in Fig. 9.10.1. Specically,
we nd for the currents on the six sides:
 2  2  2 2
n m p 2
nmp = c + + (resonant frequencies) (9.10.1) H0 sin z
(left & right)
a b l 2
|J s | = H0 cos2 kc x sin2 z + H12 sin2 kc x cos2 z
(top & bottom)

Such modes are designated as TEnmp or TMnmp . For simplicity, we consider the case H1 sin2 kc x (front & back)
TEn0p . Eqs. (9.3.6) also describe backward-moving waves if one replaces by , which
also changes the sign of TE = /c. Starting with a linear combination of forward The power loss can be computed by integrating the loss per unit conductor area,
and backward waves in the TEn0 mode, we obtain the eld components: Eq. (9.2.6), over the six wall sides, or doubling the answer for the left, top, and front

sides. Using the integrals (9.10.5), we nd:
Hz (x, z) = H0 cos kc x Aejz + Bejz ,  
1 bl al ab

Ploss = Rs |J s |2 dA = Rs H02 + (H02 + H12 ) + H12
Hx (x, z) = jH1 sin kc x Aejz Be jz
, H1 = H0 (9.10.2)
2 walls 2 4 2
kc   (9.10.6)

1 2
= Rs H02 l(2b + a)+ 2 a(2b + l)
Ey (x, z) = jE0 sin kc x Aejz + Bejz , E0 = H0 4 kc
where c = ckc . By requiring that Ey (x, z) have z-dependence of the form sin z, the where we substituted H12 = H02 2 /k2c . It follows that the Q -factor will be:
coefcients A, B must be chosen as A = B = j/2. Then, Eq. (9.10.2) specializes into:
W (k2c + 2 )(abl)
Hz (x, z) = H0 cos kc x sin z , Q= = 2
Ploss 2Rs kc l(2b + a)+2 a(2b + l)
For the TEn0p mode we have = p/l and kc = n/a. Using Eq. (9.2.7) to replace
Hx (x, z) = H1 sin kc x cos z , H1 = H0 (9.10.3)
kc Rs in terms of the skin depth , we nd:

Ey (x, z) = jE0 sin kc x sin z , E0 = H0 n2 p2
Q= a2 l2
As expected, the vanishing of Ey (x, z) on the front/back walls, z = 0 and z = l, and     (9.10.7)
n2 2 1 p2 2 1
on the left/right walls, x = 0 and x = a, requires the quantization conditions: = p/l + + +
a2 a b l2 l b
and kc = n/a. The Q of the resonator can be calculated from its denition:
The lowest resonant frequency corresponds to n = p = 1. For a cubic cavity, a =
Q= (9.10.4) b = l, the Q and the lowest resonant frequency are:
where W is the total time-averaged energy stored within the cavity volume and Ploss is a c 2 c
Q= , 101 = , f101 = = (9.10.8)
the total power loss due to the wall ohmic losses (plus other losses, such as dielectric 3 a 2 a 2
losses, if present.) The ratio = Ploss /W is usually identied as the 3-dB width of the For an air-lled cubic cavity with a = 3 cm, we nd f101 = 7.07 GHz, = 7.86105
resonance centered at frequency . Therefore, we may write Q = /. cm, and Q = 12724. As in waveguides, cavities can be excited by inserting small probes
It is easily veried that the electric and magnetic energies are equal, therefore, W that generate elds resembling a particular mode.
may be calculated by integrating the electric energy density over the cavity volume:
a b l
1 1
W = 2We = 2 |Ey (x, z)|2 dx dy dz = |E0 |2 sin2 kc x cos2 z dx dy dz 9.11 Dielectric Slab Waveguides
4 vol 2 0 0 0
1 1 2 1 k2c + 2 A dielectric slab waveguide is a planar dielectric sheet or thin lm of some thickness,
= |E0 |2 (abl)= |H0 |2 2 (abl)= |H0 |2 (abl)
8 8 c 8 k2c say 2a, as shown in Fig. 9.11.1. Wave propagation in the z-direction is by total internal
9.11. Dielectric Slab Waveguides 389 390 9. Waveguides

x, and exist effectively within a skin depth distance 1/c from the slab. Setting kc1 = kc
and kc2 = jc , Eqs. (9.11.1) become in this new notation:

k2c = k20 n21 2 k2c = k20 n21 2

2c = k20 n22 2 2c = 2 k20 n22

Similarly, Eqs. (9.11.2) read:

2x Hz (x)+k2c Hz (x)= 0 for |x| a

2x Hz (x)2c Hz (x)= 0 for |x| a
The two solutions sin kc x and cos kc x inside the guide give rise to the so-called even
Fig. 9.11.1 Dielectric slab waveguide.
and odd TE modes (referring to the evenness or oddness of the resulting electric eld.)
For the even modes, the solutions of Eqs. (9.11.4) have the form:
reection from the left and right walls of the slab. Such waveguides provide simple

H sin kc x , if a x a
models for the conning mechanism of waves propagating in optical bers. 1
The propagating elds are conned primarily inside the slab, however, they also Hz (x)= H2 ec x , if xa (9.11.5)

exist as evanescent waves outside it, decaying exponentially with distance from the slab.
H3 ec x , if x a
Fig. 9.11.1 shows a typical electric eld pattern as a function of x.
For simplicity, we assume that the media to the left and right of the slab are the The corresponding x-components Hx are obtained by applying Eq. (9.3.8) using the
same. To guarantee total internal reection, the dielectric constants inside and outside appropriate value for k2c , that is, k2c2 = 2c outside and k2c1 = k2c inside:
the slab must satisfy 1 > 2 , and similarly for the refractive indices, n1 > n2 .

j j
We only consider TE modes and look for solutions that depend only on the x co-
2 x Hz (x)= H1 cos kc x , if a x a

kc kc

ordinate. The cutoff wavenumber kc appearing in the Helmholtz equation for Hz (x) j j
depends on the dielectric constant of the propagation medium, k2c = 2  2 . There- Hx (x)= 2 x Hz (x)= H2 ec x , if xa (9.11.6)

fore, k2c takes different values inside and outside the guide:
c c

j j

x Hz (x)= H3 ec x , if x a
k2c1 = 2 1 0 2 = 2 0 0 n21 2 = k20 n21 2 (inside) 2c c
(9.11.1) The electric elds are Ey (x)= TE Hx (x), where TE = 0 / is the same inside
k2c2 = 2 2 0 2 = 2 0 0 n22 2 = k20 n22 2 (outside)
and outside the slab. Thus, the electric eld has the form:
where k0 = /c0 is the free-space wavenumber. We note that , are the same inside

E cos kc x , if a x a
and outside the guide. This follows from matching the tangential elds at all times t 1
and all points z along the slab walls. The corresponding Helmholtz equations in the Ey (x)= E2 ec x , if xa (even TE modes) (9.11.7)

regions inside and outside the guide are:
E3 ec x , if x a

x Hz (x)+k2c1 Hz (x)=
0 for |x| a where we dened the constants:
2x Hz (x)+k2c2 Hz (x)= 0 for |x| a j j j
E1 = TE H1 , E2 = TE H2 , E3 = TE H3 (9.11.8)
kc c c
Inside the slab, the solutions are sin kc1 x and cos kc1 x, and outside, sin kc2 x and
cos kc2 x, or equivalently, ejkc2 x . In order for the waves to remain conned in the near The boundary conditions state that the tangential components of the magnetic and
vicinity of the slab, the quantity kc2 must be imaginary, for if it is real, the elds would electric elds, that is, Hz , Ey , are continuous across the dielectric interfaces at x = a
propagate at large x distances from the slab (they would correspond to the rays refracted and x = a. Similarly, the normal components of the magnetic eld Bx = 0 Hx and
from the inside into the outside.) therefore also Hx must be continuous. Because Ey = TE Hx and TE is the same in
If we set kc2 = jc , the solutions outside will be ec x . If c is positive, then only both media, the continuity of Ey follows from the continuity of Hx . The continuity of
the solution ec x is physically acceptable to the right of the slab, x a, and only ec x Hz at x = a and x = a implies that:
to the left, x a. Thus, the elds attenuate exponentially with the transverse distance
9.11. Dielectric Slab Waveguides 391 392 9. Waveguides

We note that the electric elds Ey (x) given by Eqs. (9.11.7) and (9.11.15) are even or
H1 sin kc a = H2 ec a and H1 sin kc a = H3 ec a (9.11.9) odd functions of x for the two families of modes. Expressing E2 and E3 in terms of E1 ,
we summarize the forms of the electric elds in the two cases:
Similarly, the continuity of Hx implies (after canceling a factor of j):

E cos kc x , if a x a
H1 cos kc a =
H2 ec a and
H1 cos kc a =
H3 ec a (9.11.10)
kc c kc c Ey (x)= E1 cos kc a ec (xa) , if xa (even TE modes) (9.11.19)

E1 cos kc a ec (x+a) , if x a
Eqs. (9.11.9) and (9.11.10) imply:

H2 = H3 = H1 ec a sin kc a = H1 ec a cos kc a (9.11.11)
E sin kc x , if a x a
kc 1
Ey (x)= E1 sin kc a ec (xa) , if xa (odd TE modes) (9.11.20)

Similarly, we nd for the electric eld constants:
E1 sin kc a ec (x+a) , if x a

kc Given the operating frequency , Eqs. (9.11.3) and (9.11.13) or (9.11.18) provide three
E2 = E3 = E1 ec a cos kc a = E1 ec a sin kc a (9.11.12)
c equations in the three unknowns kc , c , . To solve them, we add the two equations
(9.11.3) to eliminate :
The consistency of the last equations in (9.11.11) or (9.11.12) requires that:

kc 2 2
cos kc a = sin kc a c = kc tan kc a (9.11.13) 2c + k2c = k20 (n21 n22 )= (n1 n22 ) (9.11.21)
c c20
For the odd TE modes, we have for the solutions of Eq. (9.11.4):
Next, we discuss the numerical solutions of these equations. Dening the dimen-
sionless quantities u = kc a and v = c a, we may rewrite Eqs. (9.11.13), (9.11.18), and

H cos kc x , if a x a
1 (9.11.21) in the equivalent forms:
Hz (x)= H2 ec x , if xa (9.11.14)

H3 ec x , if x a
The resulting electric eld is: v = u tan u v = u cot u
(even modes) , (odd modes) (9.11.22)
v2 + u2 = R2 v2 + u2 = R2

E sin kc x , if a x a
Ey (x)= E2 ec x , if xa (odd TE modes) (9.11.15) where R is the normalized frequency variable:

E3 ec x , if x a
a 2f a 2a
The boundary conditions imply in this case: R = k0 aNA = NA = NA = NA (9.11.23)
c0 c0
H2 = H3 = H1 ec a cos kc a = H1 ec a sin kc a (9.11.16) where NA = n21 n22 is the numerical aperture of the slab and = c0 /f , the free-space
and, for the electric eld constants: Because the functions tan u and cot u have many branches, there may be several
possible solution pairs u, v for each value of R. These solutions are obtained at the
kc intersections of the curves v = u tan u and v = u cot u with the circle of radius R,
E2 = E3 = E1 ec a sin kc a = E1 ec a cos kc a (9.11.17) that is, v2 + u2 = R2 . Fig. 9.11.2 shows the solutions for various values of the radius R
corresponding to various values of .
The consistency of the last equation requires: It is evident from the gure that for small enough R, that is, 0 R < /2, there is
only one solution and it is even. For /2 R < , there are two solutions, one even
c = kc cot kc a (9.11.18) and one odd. For R < 3/2, there are three solutions, two even and one odd, and
for an optical ber, the single-mode condition reads 2aNA / < 2.405, where a is the core radius.
9.11. Dielectric Slab Waveguides 393 394 9. Waveguides

sin u cos(m/2) cos u sin(m/2)

tan(u m/2)=
cos u cos(m/2)+ sin u sin(m/2)

Therefore, to nd the mth mode, whether even or odd, we must nd the unique
solution of the following system in the u-range Rm u < Rm+1 :

v = u tan(u Rm )
(mth mode) (9.11.29)
v2 + u2 = R2

If one had an approximate solution u, v for the mth mode, one could rene it by using
Newtons method, which converges very fast provided it is close to the true solution. Just
such an approximate solution, accurate to within one percent of the true solution, was
given by Lotspeich [930]. Without going into the detailed justication of this method,
Fig. 9.11.2 Even and odd TE modes at different frequencies. the approximation is as follows:

u = Rm + w1 (m)u1 (m)+w2 (m)u2 (m) , m = 0, 1, . . . , M (9.11.30)

so on. In general, there will be M + 1 solutions, alternating between even and odd, if R
where u1 (m), u2 (m) are approximate solutions near and far from the cutoff Rm , and
falls in the interval:
w1 (m), w2 (m) are weighting factors:
M (M + 1) 
R< (9.11.24) 1 + 2R(R Rm ) 1 Rm
2 2 u1 (m)= , u2 (m)=
R 2 R+1
Given a value of R, we determine M as that integer satisfying Eq. (9.11.24), or, M

2R/ < M + 1, that is, the largest integer less than 2R/: w1 (m)= exp (R Rm )2 /Vm2
, w2 (m)= 1 w1 (m) (9.11.31)
  1 /4 + Rm
2R Vm = Rm
M = oor (maximum mode number) (9.11.25) ln 1.25 cos(/4)

This solution serves as the starting point to Newtons iteration for solving the equa-
Then, there will be M+ 1 solutions indexed by m = 0, 1, . . . , M, which will correspond tion F(u)= 0, where F(u) is dened by
to even modes if m is even and to odd modes if m is odd. The M + 1 branches of tan u 
and cot u being intersected by the R-circle are those contained in the u-ranges: F(u)= u tan(u Rm )v = u tan(u Rm ) R2 u2 (9.11.32)

Newtons iteration is:

Rm u < Rm+1 , m = 0, 1, . . . , M (9.11.26)

where for i = 1, 2 . . . , Nit do:

F(u) (9.11.33)
m u=u
Rm = , m = 0, 1, . . . , M (9.11.27) G(u)
where G(u) is the derivative F (u), correct to order O(F):
If m is even, the u-range (9.11.26) denes a branch of tan u, and if m is odd, a branch
of cot u. We can combine the even and odd cases of Eq. (9.11.22) into a single case by v u R2
noting the identity: G(u)= + + (9.11.34)
u v u

tan u , if m is even The solution steps dened in Eqs. (9.11.29)(9.11.34) have been implemented in the
tan(u Rm )= (9.11.28) MATLAB function dslab.m, with usage:
cot u , if m is odd
[u,v,err] = dslab(R,Nit); % TE-mode cutoff wavenumbers in a dielectric slab
This follows from the trigonometric identity:
9.11. Dielectric Slab Waveguides 395 396 9. Waveguides

where Nit is the desired number of Newton iterations (9.11.33), err is the value of F(u)
at the end of the iterations, and u, v are the (M + 1)-dimensional vectors of solutions. The cutoff frequencies fm are in GHz. We note that as the mode number m increases,
The number of iterations is typically very small, Nit = 23. the quantity c decreases and the effective skin depth 1/c increases, causing the elds
The related MATLAB function dguide.m uses dslab to calculate the solution param- outside the slab to be less conned. The electric eld patterns are also shown in the gure
eters , kc , c , given the frequency f , the half-length a, and the refractive indices n1 , n2 as functions of x.
of the slab. It has usage: The approximation error, err, is found to be 4.8851015 using only three Newton itera-
[be,kc,ac,fc,err] = dguide(f,a,n1,n2,Nit); % dielectric slab guide tions. Using two, one, and no (the Lotspeich approximation) iterations would result in the
errors 2.381108 , 4.029104 , and 0.058.
where f is in GHz, a in cm, and , kc , c in cm . The quantity fc is the vector of
The lowest non-zero cutoff frequency is f1 = 8.6603 GHz, implying that there will be a
the M + 1 cutoff frequencies dened by the branch edges Rm = m/2, that is, Rm =
single solution if f is in the interval 0 f < f1 . For example, if f = 5 GHz, the solution is
m aNA /c0 = 2fm aNA /c0 = m/2, or,
= 1.5649 rad/cm, kc = 1.3920 rad/cm, and c = 1.1629 nepers/cm.

mc0 The frequency range over which there are only four solutions is [25.9808, 34.6410] GHz,
fm = , m = 0, 1, . . . , M (9.11.35) where the upper limit is 4f1 .
We note that the function dguide assumes internally that c0 = 30 GHz cm, and therefore,
The meaning of fm is that there are m + 1 propagating modes for each f in the
the calculated values for kc , c would be slightly different if a more precise value of c0
interval fm f < fm+1 . is used, such as 29.9792458 of Appendix A. Problem 9.13 studies the sensitivity of the
Example 9.11.1: Dielectric Slab Waveguide. Determine the propagating TE modes of a dielectric solutions to small changes of the parameters f , a, c0 , n1 , n2 . 

slab of half-length a = 0.5 cm at frequency f = 30 GHz. The refractive indices of the slab
and the surrounding dielectric are n1 = 2 and n2 = 1. In terms of the ray picture of the propagating wave, the angles of total internal
reection are quantized according to the values of the propagation wavenumber for
Solution: The solution is obtained by the MATLAB call:
the various modes.
f = 30; a = 0.5; n1 = 2; n2 = 1; Nit = 3; If we denote by k1 = k0 n1 the wavenumber within the slab, then the wavenumbers
[be,kc,ac,fc,err] = dguide(f,a,n1,n2,Nit) , kc are the z- and x-components kz , kx of k1 with an angle of incidence . (The vectorial
relationships are the same as those in Fig. 9.9.1.) Thus, we have:
The frequency radius is R = 5.4414, which gives 2R/ = 3.4641, and therefore, M = 3.
The resulting solutions, depicted in Fig. 9.11.3, are as follows:
= k1 sin = k0 n1 sin
TE Modes for R = 5.44 Electric Fields kc = k1 cos = k0 n1 cos

The value of for each mode will generate a corresponding value for . The at-
1 0 1
0 tenuation wavenumber c outside the slab can also be expressed in terms of the total
5 1
2 internal reection angles:
Ey (x) / E1

4 2 3
0 c = 2 k20 n22 = k0 n21 sin2 n22

2 3 Since the critical angle is sin c = n2 /n1 , we may also express c as:

1 1 
c = k0 n1 sin2 sin 2c (9.11.37)
0 1 2 3 4 5 6 7 3 2 1 0 1 2 3
u x/a Example 9.11.2: For the Example 9.11.1, we calculate k0 = 6.2832 and k1 = 12.5664 rad/cm.
The critical and total internal reection angles of the four modes are found to be:
Fig. 9.11.3 TE modes and corresponding E-eld patterns.  
c = asin = 30o
m u v kc c fm  

0 1.3248 5.2777 12.2838 2.6497 10.5553 0.0000 = asin = {77.8275o , 65.1960o , 51.5100o , 36.0609o }
1 2.6359 4.7603 11.4071 5.2718 9.5207 8.6603
2 3.9105 3.7837 9.8359 7.8210 7.5675 17.3205
As required, all s are greater than c . 

3 5.0793 1.9519 7.3971 10.1585 3.9037 25.9808
9.12. Asymmetric Dielectric Slab 397 398 9. Waveguides

9.12 Asymmetric Dielectric Slab which combine to dene the allowed range of for the guided modes:

The three-layer asymmetric dielectric slab waveguide shown in Fig. 9.12.1 is a typical
component in integrated optics applications [912933]. nc ns nf (9.12.3)
A thin dielectric lm nf of thickness 2a is deposited on a dielectric substrate ns .
Above the lm is a dielectric cover or cladding nc , such as air. To achieve propagation where the lower limit = k0 ns denes the cutoff frequencies, see Eq. (9.12.13).
by total internal reection within the lm, we assume that the refractive indices satisfy
nf > ns nc . The case of the symmetric dielectric slab of the previous section is
obtained when nc = ns .
TE modes
We consider the TE modes rst. Assuming only x-dependence for the Hz component, it
must satisfy the Helmholtz equations in the three regions:

(2x + k2f )Hz (x)= 0, |x| a

(x c )Hz (x)= 0,
2 2
(x s )Hz (x)= 0,
2 2
x a

The solutions, decaying exponentially in the substrate and cover, can be written in
the following form, which automatically satises the continuity conditions at the two
boundaries x = a:

H1 sin(kf x + ) ,
|x| a
Hz (x)= H1 sin(kf a + )ec (xa) , xa (9.12.4)

Fig. 9.12.1 Three-layer asymmetric dielectric slab waveguide. H1 sin(kf a )es (x+a) , x a

In this section, we briey discuss the properties of the TE and TM propagation modes. where is a parameter to be determined. The remaining two components, Hx and Ey ,

Let k0 = 0 c0 = /c0 = 2f /c0 = 2/0 be the free-space wavenumber at the are obtained by applying Eq. (9.3.8), that is,
operating frequency or f in Hz. The t, z dependence of the elds is assumed to be
the usual ejtjt . If we orient the coordinate axes as shown in the above gure, then Hx = x Hz , Ey = TE Hx TE =
the decay constants s and c within the substrate and cladding must be positive so
that the elds attenuate exponentially with x within both the substrate and cladding,
This gives in the three regions:
hence, the corresponding transverse wavenumbers will be js and jc . On the other

hand, the transverse wavenumber kf within the lm will be real-valued. These quantities

j H1 cos(kf x + ) , |x| a
satisfy the relations (we assume = 0 in all three media):


k2f = k20 n2f 2 k2f + 2s = k20 (n2f n2s ) Hx (x)= j H1 sin(kf a + )ec (xa) , xa (9.12.5)


2s = 2 k20 n2s k2f + 2c = k20 (n2f n2s )(1 + )= k20 (n2f n2c ) (9.12.1) j H1 sin(kf a )es (x+a) , x a
2c = 2 k20 n2c 2c 2s = k20 (n2f n2s ) = k20 (n2s n2c )
Since we assumed that = 0 in all three regions, the continuity of Ey across the
boundaries x = a implies the same for the Hx components, resulting in the two con-
where we dened the asymmetry parameter :
n2s n2c 1 1 c
= (9.12.2) cos(kf a + ) = sin(kf a + ) tan(kf a + ) =
n2f n2s kf c kf
1 1 s
Note that 0 since we assumed nf > ns nc . Because kf , s , c are assumed to cos(kf a ) = sin(kf a ) tan(kf a ) =
be real, it follows that must satisfy the inequalities, k0 nf , k0 ns , and k0 nc , kf s kf
9.12. Asymmetric Dielectric Slab 399 400 9. Waveguides

Since the argument of the tangent is unique up to an integral multiple of , we may For a given operating frequency f , the value of R is xed. All allowed propagating
invert the two tangents as follows without loss of generality: modes must satisfy Rm R, or,

kf a + = arctan
+ m 1 1 
2R arctan
kf m + arctan R m
2 2

s This xes the maximum mode index M to be:
kf a = arctan

2R arctan
which result in the characteristic equation of the slab and the solution for : M = oor (maximum TE mode index) (9.12.14)

1 1 s 1 c
kf a = m + arctan + arctan (9.12.7)
2 2 kf 2 kf Thus, there are (M + 1) modes labeled by m = 0, 1, . . . , M. In the symmetric case,
= 0, and (9.12.14) reduces to Eq. (9.11.25) of the previous section. The corresponding

1 1 c 1 s cutoff frequencies are obtained by setting:
= m + arctan arctan (9.12.8)
2 2 kf 2 kf 1 1 

2fm a  m + arctan
where the integer m = 0, 1, 2, . . . , corresponds to the mth mode. Eq. (9.12.7) and the
Rm = n2f n2s = 2 2
2a  2
fm (9.12.15)
three equations (9.12.1) provide four equations in the four unknowns {, kf , s , c }. c0 2
nf ns
Using the trig identities tan(1 +2 )= (tan 1 +tan 2 )/(1tan 1 tan 2 ) and tan()= c0
tan( + m), Eqs. (9.12.7) and (9.12.8) may also be written in the following forms:
which can be written more simply as fm = f Rm /R, m = 0, 1, . . . , M, where f = c0 /0 .
kf (c + s ) kf (c s ) For each of the M+1 propagating modes one can calculate the corresponding angle of
tan(2kf a)= , tan(2)= (9.12.9)
k2f c s k2f + c s total internal reection of the equivalent ray model by identifying kf with the transverse
propagation wavenumber, that is, kf = k0 nf cos , as shown in Fig. 9.12.2.
The form of Eq. (9.12.7) is preferred for the numerical solution. To this end, we introduce
the dimensionless variables:
 2f a  2 a 2
R = k0 a n2f n2s = nf n2s = 2 nf n2s
c0 0 (9.12.10)
u = kf a , v = s a , w = c a
Then, Eqs. (9.12.7) and (9.12.1) can be written in the normalized forms:
1 1 v 1 w
u= m + arctan + arctan
2 2 u 2 u
u2 + v2 = R2 (9.12.11)
Fig. 9.12.2 Ray propagation model.
w2 v2 = R2
The characteristic equation (9.12.7) can be given a nice interpretation in terms of the
Once these are solved for the three unknowns u, v, w, or kf , s , c , the propagation ray model [925]. The eld of the upgoing ray at a point A at (x, z) is proportional, up
constant , or equivalently, the effective index n = /k0 can be obtained from: to a constant amplitude, to
  ejkf x ejz
 k2f  u2
= k20 n2f k2f n = = nf 2 = n2f 2 2
(9.12.12) Similarly, the eld of the upgoing ray at the point B at (x, z + l) should be
k0 k0 k0 a
ejkf x ej(z+l) (9.12.16)
To determine the number of propagating modes and the range of the mode index
m, we set v = 0 in the characteristic
equation (9.12.11) to nd the radius Rm of the mth But if we follow the ray starting at A along the zig-zag path AC CS SB, the ray
mode. Then, u = Rm and w = Rm , and we obtain:
will have traveled a total vertical roundtrip distance of 4a and will have suffered two
1 1 
total internal reection phase shifts at points C and S, denoted by 2c and 2s . We
Rm = m + arctan , m = 0, 1, 2, . . . (9.12.13)
2 2
9.12. Asymmetric Dielectric Slab 401 402 9. Waveguides

recall that the reection coefcients have the form = e2j for total internal reection, Universal mode curves for TE modes
as given for example by Eq. (7.8.3). Thus, the eld at point B would be 1
ejkf (x+4a) e2js e2jc ej(z+l)
0.8 m=0
This must match (9.12.16) and therefore the extra accumulated phase 4kf a 2s 2c
must be equal to a multiple of 2, that is, 0.7 1

1 1 1 0.6 2
4kf a 2s 2c = 2m kf a = m + s + c
2 2 2 0.5

As seen from Eq. (7.8.3), the phase terms are exactly those appearing in Eq. (9.12.7): 0.4 4

c s c s 0.3
tan c = , tan s = c = arctan , s = arctan
kf kf kf kf
0.2 =0
A similar interpretation can be given for the TM modes. 0.1 =1
It is common in the literature to represent the characteristic equation (9.12.11) by = 10
means of a universal mode curve [927] dened in terms of the following scaled variable: 0
0 1 2 3 4 5 6 7 8 9 10 11
k20 n2s
b= = 2 2 2 (9.12.17)
R2 k0 (nf ns ) Fig. 9.12.3 Universal mode curves.
which ranges over the standardized interval 0 b 1, so that
u = R 1 b, v = R b, w=R b+ (9.12.18) where is a parameter to be determined. Then, the Ex component is:

Then, Eq. (9.12.11) takes the universal form in terms of the variables b, R:

j E1 cos(kf x + ) , |x| a

b b +
2R 1 b = m + arctan + arctan (9.12.19) Ex (x)= j

E1 sin(kf a + )ec (xa) , xa (9.12.21)
1b 1b

It is depicted in Fig. 9.12.3 with one branch for each value of m = 0, 1, 2, . . . , and for

j E1 sin(kf a )es (x+a) , x a
the three asymmetry parameter values = 0, 1, 10.
A vertical line drawn at each value of R determines the values of b for the propagating The boundary conditions require the continuity of the normal component of dis-
modes. Similar curves can be developed for TM modes. See Example 9.12.1 for a concrete placement eld Dx = Ex across the interfaces at x = a, which is equivalent to the
example that includes both TE and TM modes. continuity of the tangential eld Hy because Hy = Ex /TM = Ex / = Dx /. Thus,
the boundary conditions at x = a require:
TM modes f c c
cos(kf a + ) = sin(kf a + ) tan(kf a + ) = pc
The TM modes are obtained by solving Eqs. (9.3.10) in each region and applying the kf c kf
boundary conditions. Assuming x-dependence only, we must solve in each region:
f s s
cos(kf a ) = sin(kf a ) tan(kf a ) = ps
j 1 kf s kf
(2x + k2f )Ez = 0 , Ex = x Ez , Hy = Ex , TM =
k2f TM 
where we dened the ratios:
The solution for Ez (x) is given by a similar expression as Eq. (9.12.4):
f n2f f n2f
pc = = 2, ps = = 2 (9.12.23)

E1 sin(kf x + ) ,
|x| a c nc s ns
Ez (x)= E1 sin(kf a + )ec (xa) , xa (9.12.20)

E1 sin(kf a )es (x+a) , x a
R is usually denoted by the variable V.
9.12. Asymmetric Dielectric Slab 403 404 9. Waveguides

Inverting the tangents we obtain: which can be written more simply as fm = f Rm /R, m = 0, 1, . . . , M, where f = c0 /0 .
The corresponding angles of total internal reection in the equivalent ray model are
c obtained by solving kf = k0 nf cos .
kf a + = arctan pc + m
kf Because pc > 1, we observe that the maximum mode index M and the cutoff fre-
quencies fm will satisfy the following inequalities for the TE and TM cases:
kf a = arctan ps
MTM MTE , fm,TE fm,TM (9.12.30)
These give the characteristic equation of the slab and :
Numerical Solutions
1 1 s 1 c
kf a = m + arctan ps + arctan pc (9.12.24) Next we look at the numerical solutions of Eqs. (9.12.27). The TE case is also included
2 2 kf 2 kf
by setting ps = pc = 1. A simple and effective iterative method for solving such char-
acteristic equations was given in Ref. [963]. By replacing v, w in terms of u, let F(u)
1 1 c 1 s
= m + arctan pc arctan ps (9.12.25) denote the right-hand side of Eq. (9.12.27):
2 2 kf 2 kf
and, as in Eq. (9.12.9), we can write: 1 1 v 1 w
F(u)= m + arctan ps + arctan pc
2 2 u 2 u
kf (ps s + pc s ) kf (pc c ps s )
tan(2kf a)= , tan(2)= (9.12.26) The problem then becomes that of nding the xed-point solutions u = F(u). The
k2f ps pc s c k2f + ps pc s c
method suggested in Ref. [963] is to use the iteration:
In terms of the normalized variables u, v, w, R, we have:
    un+1 = F(un ) , n = 0, 1, 2, . . .
1 1 v 1 w
u = m + arctan ps + arctan pc
2 2 u 2 u initialized at u0 = R. This simple iteration does converges in many cases, but fails in
2 2 2 (9.12.27) others. We have found that a simple modication that involves the introduction of a
u +v =R
relaxation parameter r such that 0 r 1, enables the convergence of even the most
w2 v2 = R2 difcult cases. The modied iteration has the form:

un+1 = rF(un )+(1 r)un

The number of propagating modes and the range of the mode index m, are again
determined by setting v = 0, u = Rm , and w = Rm :
Explicitly, the iteration starts with the initial values:
1 1 

Rm = m + arctan pc , m = 0, 1, 2, . . . 
2 2 u0 = R , v0 = 0 , w0 = R (9.12.31)
The allowed propagating modes must satisfy Rm R, or,
and proceeds iteratively, for n = 0, 1, 2, . . . , until two successive un values become closer

1 1 
2R arctan pc to each other than some specied error tolerance, say tol, such as tol = 1010 :
m + arctan pc R m
2 2     
1 1 vn 1 wn
which xes the maximum mode index M to be: un+1 = r m + arctan ps + arctan pc + (1 r)un
2 2 un 2 un

2R arctan pc if |un+1 un | < tol, then exit, else continue
M = oor (maximum TM mode index) (9.12.28)  (9.12.32)

vn+1 = R2 u2n+1
The (M + 1) modes are again labeled by m = 0, 1, . . . , M. The corresponding cutoff 
wn+1 = R2 vn+ 1
frequencies are obtained by setting:

1 1 
The MATLAB function dguide3.m implements the method and has usage:
2fm a  m + arctan pc
Rm = 2
nf ns 2
= 2 2
2a  2
fm (9.12.29) [be,kf,as,ac,fm,Nit] = dguide3(a,ns,nf,nc,mode,r,tol);
c0 2
nf ns
9.12. Asymmetric Dielectric Slab 405 406 9. Waveguides

where the inputs and outputs have the following meanings: Example 9.12.1: For comparison purposes, we consider the same benchmark example dis-
cussed in [963] consisting of a silicon lm of thickness of 1 m with nf = 3.5, an oxide
a = half-width of slab in units of the free-space wavelength 0 substrate with ns = 1.45, and air cover, with operating wavelength 0 = 1.55 m. The
ns , nf , nc = refractive indices of substrate, lm, and cladding (nf > ns >= nc ) following MATLAB code generates both the TE and TM modes, with the numerical outputs
mode = TE or TM listed in the tables below.
r = relaxation parameter (default r = 0.5)
tol = error tolerance (default tol = 1010 ) nf=3.5; ns=1.45; nc=1; % oxide substrate | silicon film | air cover
la0 = 1.55; a = 0.5; % units of microns
= propagation wavenumbers in units of k0 = 2/0 a = a/la0; % half-thickness in units of la0
kf = transverse wavenumbers inside slab in units of k0
r=0.3; % default value r=0.5 fails to converge for the TE modes
s , c = decay wavenumbers in substrate and cladding in units of k0 tol=1e-10;
fm = cutoff frequencies in units of f = c0 /0
Nit = number of iterations it takes to converge to within tol [be,kf,as,ac,fm,Nit] = dguide3(a,ns,nf,nc,te,r,tol); % TE modes
[be,kf,as,ac,fm,Nit] = dguide3(a,ns,nf,nc,tm,r,tol); % TM modes

Internally, the function determines M from Eq. (9.12.14) or (9.12.28) and calculates
, kf , s , c , fm as (M + 1)-dimensional column vectors representing the M + 1 modes.
To clarify the computations, the essential part of the code is listed below: m /k0 kf /k0 s /k0 c /k0 fm /f m
0 3.434746 0.6727 3.1137 3.2860 0.0247 78.92
k0 = 2*pi; % la0 = 2*pi/k0 = 1 in the assumed units 1 3.232789 1.3413 2.8894 3.0742 0.2679 67.47 (TE modes)
R = k0*a * sqrt(nf^2-ns^2); % (u,v) circle radius, note k0*a = 2*pi*(a/la0) 2 2.872310 2.0000 2.4794 2.6926 0.5112 55.15
d = (ns^2-nc^2)/(nf^2-ns^2); % asymmetry parameter 3 2.302025 2.6364 1.7880 2.0735 0.7545 41.13
4 1.451972 3.1846 0.0756 1.0527 0.9978 24.51
if strcmpi(mode,TE) % mode can also be entered in lower case
ps = 1; pc = 1;
else m /k0 kf /k0 s /k0 c /k0 fm /f m
ps = nf^2/ns^2; pc = nf^2/nc^2;
0 3.416507 0.7599 3.0935 3.2669 0.1028 77.46
1 3.154191 1.5169 2.8011 2.9915 0.3461 64.32 (TM modes)
M = floor((2*R - atan(pc*sqrt(d)))/pi); % highest mode index
2 2.668932 2.2642 2.2407 2.4745 0.5894 49.69
m = (0:M); % vector of mode indices 3 1.865244 2.9616 1.1733 1.5745 0.8327 32.20
4 1.0760
u = R*ones(M+1,1); % initialize iteration variables u,v,w
v = zeros(M+1,1); % u,v,w are (M+1)x1 vectors The /k0 column is the effective phase index of the modes. The default value of the
w = R*sqrt(d)*ones(M+1,1); relaxation parameter r = 0.5 did not work in this case and caused the TE iteration to
diverge and the smaller value r = 0.3 was chosen. The number of iterations were Nit = 57
Nit = 1; % number of iterations for TE and Nit = 66 for TM. The TIR angles were computed by the following command:
% while loop repeats till convergence
while 1
thm = acos(kf/nf)*180/pi; % degrees
unew = r*(m*pi/2 + atan(ps*v./u)/2 + atan(pc*w./u)/2) + (1-r)*u;
if norm(unew-u) <= tol, break; end
Nit = Nit + 1; We note that all TIR angles are greater than the critical angles computed by:
u = unew;
v = sqrt(R^2 - u.^2); ns nc
s = arcsin = 24.47o , c = arcsin = 16.60o
w = sqrt(R^2*d + v.^2); nf nf
if Nit>1000, break; end % safeguard against possible divergence
There are ve TE modes and four TM ones. The fth TE mode is very weakly bound to the
substrate side because its decay parameter s is very small, its cutoff frequency is very
kf = u/(k0*a); % kf in units of k0, i.e., kf/k0 = u/(k0*a)
as = v/(k0*a); near the operating frequency f = c0 /0 , and its TIR angle, very close to the critical angle.
ac = w/(k0*a); With reference to the inequality (9.12.30), it so happened that in this example f falls in the
be = sqrt(nf^2 - kf.*kf); % beta in units of k0, i.e., beta/k0
range f4,TE < f < f4,TM , and therefore, the fth TM mode f4,TM is not excited, but f4,TE is.
Rm = m*pi/2 + atan(pc*sqrt(d))/2; % cutoff radius for m-th mode The convergence can be veried for all modes at once by computing the vector error norm
fm = Rm/R; % cutoff frequencies in units of f = c0/la0 of the characteristic equations, that is,
thm = acos(kf/nf); % angles of total internal reflection
9.12. Asymmetric Dielectric Slab 407 408 9. Waveguides

M = length(be)-1; m = (0:M); 9.13 Problems

Err = norm(kf*2*pi*a - m*pi/2 - atan(ps*as./kf)/2 - atan(pc*ac./kf)/2);
9.1 An air-lled 1.5 cm3 cm waveguide is operated at a frequency that lies in the middle of its
This error is of the order of the assumed tolerance, indeed, we have Err = 2.94 1010 for TE10 mode band. Determine this operating frequency in GHz and calculate the maximum
TE, and Err = 2.53 1010 for TM. We note that the quantity kf*2*pi*a represents the power in Watts that can be transmitted without causing dielectric breakdown of air. The
variable u in our units, indeed, u = kf a = (kf /k0 )k0 a = (kf /k0 )2(a/0 ). dielectric strength of air is 3 MV/m.

Finally, Fig. 9.12.4 displays the TE and TM solutions on the universal mode curves, see e.g. 9.2 It is desired to design an air-lled rectangular waveguide such that (a) it operates only in the
Eq. (9.12.19). 
 TE10 mode with the widest possible bandwidth, (b) it can transmit the maximum possible
power, and (c) the operating frequency is 12 GHz and it lies in the middle of the operating
band. What are the dimensions of the guide in cm?
1 9.3 An air-lled rectangular waveguide is used to transfer power to a radar antenna. The guide
must meet the following specications: The two lowest modes are TE10 and TE20 . The op-
erating frequency is 3 GHz and must lie exactly halfway between the cutoff frequencies of
0.8 these two modes. The maximum electric eld within the guide may not exceed, by a safety
m=0 margin of 3, the breakdown eld of air 3 MV/m.
a. Determine the smallest dimensions a, b for such a waveguide, if the transmitted power
1 is required to be 1 MW.

b. What are the dimensions a, b if the transmitted power is required to be maximum?

0.4 What is that maximum power in MW?
0.3 4 9.4 It is desired to design an air-lled rectangular waveguide operating at 5 GHz, whose group
velocity is 0.8c. What are the dimensions a, b of the guide (in cm) if it is also required to carry
0.2 maximum power and have the widest possible bandwidth? What is the cutoff frequency of
the guide in GHz and the operating bandwidth?
9.5 Show the following relationship between guide wavelength and group velocity in an arbitrary
0 air-lled waveguide: vg g = c , where g = 2/ and is the free-space wavelength.
0 1 2 3 4 5 6 7 8 9 10
R Moreover, show that the and g are related to the cutoff wavelength c by:

1 1 1
Fig. 9.12.4 Universal mode curves. TE (solid lines/lled circles), TM (dashed lines/open circles). = +
2 2g 2c

9.6 Determine the four lowest modes that can propagate in a WR-159 and a WR-90 waveguide.
Example 9.12.2: A second, more difcult, example from [963] has the parameters 0 = 1.55 m, Calculate the cutoff frequencies (in GHz) and cutoff wavelengths (in cm) of these modes.
a = 0.5 m, nf = 3.3, ns = 3.256, nc = 1.
9.7 An air-lled WR-90 waveguide is operated at 9 GHz. Calculate the maximum power that
The same MATLAB code applies here, but we used the default value r = 0.5, which con- can be transmitted without causing dielectric breakdown of air. Calculate the attenuation
verges in 8 and 10 iterations respectively for the TE and TM modes. Only one (M = 0) TE constant in dB/m due to wall ohmic losses. Assume copper walls.
and one TM mode are supported with parameters given in the table below. The critical TIR
9.8 A rectangular waveguide has sides a, b such that b a/2. Determine the cutoff wavelength
angles are in this example:
c of this guide. Show that the operating wavelength band of the lowest mode is 0.5c

ns nc c . Moreover, show that the allowed range of the guide wavelength is g c / 3.
s = arcsin = 80.63o , c = arcsin = 17.64o
nf nf 9.9 The TE10 mode operating bandwidth of an air-lled waveguide is required to be 47 GHz.
What are the dimensions of the guide?
9.10 Computer Experiment: WR-159 Waveguide. Reproduce the two graphs of Fig. 9.8.2.
mode /k0 kf /k0 s /k0 c /k0 fm /f m
9.11 A TM mode is propagated along a hollow metallic waveguide of arbitrary but uniform cross
TE 3.265996 0.4725 0.2553 3.1091 0.6427 81.77 section. Assume perfectly conducting walls.
TM 3.263384 0.4902 0.2194 3.1064 0.7142 81.46
a. Show that the Ez (x, y) component satises:

The computational errors in the characteristic equations were Err = 1.63 10 for TE,
T Ez |2 dS = k2c
| |Ez |2 dS , (kc = cutoff wavenumber) (9.13.1)
and Err = 1.52 1011 for TM. 
 S S
9.13. Problems 409 410 9. Waveguides

b. Using the above result, show that the energy velocity is equal to the group velocity. a. Show that PT is given as the sum of the following two terms, where the rst one
represents the power owing within the slab, and the second, the power owing outside
T B)= T A T B + A 2T B, for scalar A, B.
Hint: Use the identity: T (A the slab:
9.12 Computer Experiment: Dielectric Slab Waveguide. Using the MATLAB functions dslab and  
0 akc + sin(akc )cos(akc ) 0 sin2 (akc )
dguide, write a program that reproduces all the results and graphs of Examples 9.11.1 and PT = |H1 |2 3 + |H1 |2
2k c 23c
9.13 Show that if the speed of light c0 is slightly changed to c = c0 + c0 (e.g. representing a where H1 is the amplitude dened in Eq. (9.11.5). Without loss of generality, from now
more exact value), then the solutions of Eq. (9.11.29) for kc , c change into: on set, H1 = 1.
   b. Show that the electric and magnetic energy densities are given as follows, where again,
kc c0
kc + kc = kc the rst terms represent the energy contained within the slab, and the second, the
1 + c a c0
  energy outside the slab:
k2c a c0
c + c = c c +  
1 + c a c0 0 (2 + k2c ) akc + sin(akc )cos(akc ) 0 (2 2c )sin2 (akc )
We = 3 +
4kc 43c
For Example 9.11.1, calculate the corrected values when c0 = 30 and c = 29.9792458
GHz cm. Compare with the values obtained if c0 is replaced by c inside the function dguide. 0 (2 + k2c )akc + 0 (2 k2c )sin(akc )cos(akc ) 0 (2 + 2 )sin2 (akc )
Wm = +
4k3c 43c
More generally, consider the sensitivity of the solutions of Eq. (9.11.29) to any of the
eters 0 , a, c0 , n1 , n2 , which affect the solution through the value of R = a0 c01 n21 n22 . c. Using the above expressions and Eq. (9.11.13), show the equality
A small change in one or all of the parameters will induce a small change R R + R. Show
that the solutions are changed to We = Wm

u R Thus, the total energy density is W = We + Wm

= 2We .
u + u = u +
1+v R
  d. From parts (a,b,c), show that
u2 R
v + v = v + v + 0 (1 + ac )
1+v R PT =
2c k2c
In particular, for simultaneous changes in all of the parameters, show that  2 
0 ( + k2c )ac + 2
W =
R a 0 c0 2n1 n1 2n2 n2 2c k2c
= + +
R a 0 c0 n21 n22
e. By differentiating Eqs. (9.11.3) and (9.11.13) with respect to , show that
From these results, show that the changes due to a change a a + a of the slab thickness
are given by, d (2 + kc2 )ac + 2
kc c d 1 + ac
kc + kc = kc a
1 + c a
f. Combining the results of parts (e,f), show nally that
c + c = c + a
1 + c a (1 + ac ) (1 + ac )
ven = vgr = = 2
9.14 For the dielectric slab waveguide shown in Fig. 9.11.1, demonstrate that the energy transport (2 + k2c )ac + 2 0 0 n21 ac + 2
velocity is equal to the group velocity. Specically, consider the case of even TE modes
9.15 Computer Experiment. Asymmetric Slab Waveguide. Reproduce all the results and Fig. 9.12.4
dened by Eqs. (9.11.3)(9.11.13), and show that ven = vgr , where
of Example 9.12.1. Moreover, make a separate graph of Fig. 9.12.4 that zooms into the
PT d neighborhood of the fth TE mode to make sure that it is indeed below cutoff.
ven = , vgr = (9.13.2)
W d

where PT is the time-averaged power transmitted in the z-direction through the cross-
sectional area dened by 0 y 1 and < x < , and W is the energy contained
in the volume dened by the above area and unit-z-length, i.e.,

1 1  
PT = |Ey (x)|2 dx , W = |Ey (x)|2 + |Hx (x)|2 + |Hz (x)|2 dx
2TE 4

Because of the substantial amount of algebra involved, break the calculation as follows: