800

18. Radiation from Apertures

18
Radiation from Apertures

18.1 Field Equivalence Principle

The radiation elds from aperture antennas, such as slots, open-ended waveguides,
horns, reector and lens antennas, are determined from the knowledge of the elds
over the aperture of the antenna.
The aperture elds become the sources of the radiated elds at large distances. This
is a variation of the Huygens-Fresnel principle, which states that the points on each
wavefront become the sources of secondary spherical waves propagating outwards and
whose superposition generates the next wavefront.
Let E a , H a be the tangential elds over an aperture A, as shown in Fig. 18.1.1. These
elds are assumed to be known and are produced by the sources to the left of the screen.
The problem is to determine the radiated elds E(r), H(r) at some far observation point.
The radiated elds can be computed with the help of the eld equivalence principle
[12881295,1685], which states that the aperture elds may be replaced by equivalent
electric and magnetic surface currents, whose radiated elds can then be calculated using
the techniques of Sec. 15.10. The equivalent surface currents are:
Ha
Js = n
Ea
J ms = n

(electric surface current)

(18.1.1)

Fig. 18.1.1 Radiated elds from an aperture.

The screen can also be a perfectly conducting surface, such as a ground plane, on
which the aperture opening has been cut. In reector antennas, the aperture itself is
not an opening, but rather a reecting surface. Fig. 18.1.2 depicts some examples of
screens and apertures: (a) an open-ended waveguide over an innite ground plane, (b)
an open-ended waveguide radiating into free space, and (c) a reector antenna.

Fig. 18.1.2 Examples of aperture planes.

(magnetic surface current)

is a unit vector normal to the surface and on the side of the radiated elds.
where n
Thus, it becomes necessary to consider Maxwells equations in the presence of magnetic currents and derive the radiation elds from such currents.
The screen in Fig. 18.1.1 is an arbitrary innite surface over which the tangential
elds are assumed to be zero. This assumption is not necessarily consistent with the
radiated eld solutions, that is, Eqs. (18.4.9). A consistent calculation of the elds to
the right of the aperture plane requires knowledge of the elds over the entire aperture
plane (screen plus aperture.)
However, for large apertures (with typical dimension much greater than a wavelength), the approximation of using the elds E a , H a only over the aperture to calculate
the radiation patterns is fairly adequate, especially in predicting the main-lobe behavior
of the patterns.

There are two alternative forms of the eld equivalence principle, which may be used
when only one of the aperture elds E a or H a is available. They are:
Js = 0
E a)
J ms = 2(n
H a)
J s = 2(n
J ms = 0

(perfect electric conductor)

(18.1.2)

(perfect magnetic conductor)

(18.1.3)

They are appropriate when the screen is a perfect electric conductor (PEC) on which
E a = 0, or when it is a perfect magnetic conductor (PMC) on which H a = 0. On the
aperture, both E a and H a are non-zero.

18.2. Magnetic Currents and Duality

801

Using image theory, the perfect electric (magnetic) conducting screen can be eliminated and replaced by an image magnetic (electric) surface current, doubling its value
over the aperture. The image eld causes the total tangential electric (magnetic) eld to
vanish over the screen.
If the tangential elds E a , H a were known over the entire aperture plane (screen plus
aperture), the three versions of the equivalence principle would generate the same radiated elds. But because we consider E a , H a only over the aperture, the three versions
give slightly different results.
In the case of a perfectly conducting screen, the calculated radiation elds (18.4.10)
using the equivalent currents (18.1.2) are consistent with the boundary conditions on
the screen.
We provide a justication of the eld equivalence principle (18.1.1) in Sec. 18.10
using vector diffraction theory and the Stratton-Chu and Kottler formulas. The modied
forms (18.1.2) and (18.1.3) are justied in Sec. 19.2 where we derive them in two ways:
one, using the plane-wave-spectrum representation, and two, using the Franz formulas
in conjunction with the extinction theorem discussed in Sec. 18.11, and discuss also
their relationship to Rayleigh-Sommerfeld diffraction theory of Sec. 19.1.

Next, we consider the solution of Maxwells equations driven by the ordinary electric
charge and current densities , J, and in addition, by the magnetic charge and current
densities m , J m .
Although m , J m are ctitious, the solution of this problem will allow us to identify
the equivalent magnetic currents to be used in aperture problems, and thus, establish
the eld equivalence principle. The generalized form of Maxwells equations is:




J

J m
m

Jm

m
J

(18.2.1)

A

A m
m

J 
J m


m


=

cos
sin

(duality)

(18.2.2)

where , A and m , A m are the corresponding scalar and vector potentials introduced
below. These transformations can be recognized as a special case (for = /2) of the
following duality rotations, which also leave Maxwells equations invariant:



E
H

J
Jm


(18.2.3)

Am
1

m jA m +
H =

(18.2.4)

The expression for H can be derived from that of E by a duality transformation of

the form (18.2.2). The scalar and vector potentials satisfy the Lorenz conditions and
Helmholtz wave equations:

2 + k2 =

A m + j m = 0

and

2 A + k2 A = J

2 m + k2 m =

(18.2.5)

2 A m + k2 A m =  J m


The solutions of the Helmholtz equations are given in terms of G(r r )=


V

(r )G(r r ) dV ,

J(r )G(r r ) dV ,


m (r) =


A m (r) =

ejk|rr |
:
4|r r |

m (r )G(r r ) dV

(18.2.6)

 J m (r )G(r r ) dV

where V is the volume over which the charge and current densities are nonzero. The
observation point r is taken to be outside this volume. Using the Lorenz conditions, the
scalar potentials may be eliminated in favor of the vector potentials, resulting in the
alternative expressions for Eq. (18.2.4):
E=
H=

Am

m
A

sin
cos

jA
E =

There is now complete symmetry, or duality, between the electric and the magnetic
quantities. In fact, it can be veried easily that the following duality transformation
leaves the set of four equations invariant :
E H
H E

E
H 

Under the duality transformations (18.2.2), the rst two of Eqs. (18.2.1) transform
into the last two, and conversely, the last two transform into the rst two.
A useful consequence of duality is that if one has obtained expressions for the electric eld E, then by applying a duality transformation one can generate expressions for
the magnetic eld H. We will see examples of this property shortly.
The solution of Eq. (18.2.1) is obtained in terms of the usual scalar and vector potentials , A, as well as two new potentials m , A m of the magnetic type:

A(r) =

E = J m jH
H=

(r) =

H = J + jE
E=

18. Radiation from Apertures

A + j = 0

18.2 Magnetic Currents and Duality

802

j

 1

(
A)+k2 A A m


 1

A m )+k2 A m + A
(
j

(18.2.7)

These may also be written in the form of Eq. (15.3.9):

E=
H=

j


1
A) J] A m
(



1
A m ) J m ]+ A
(
j

(18.2.8)

18.3. Radiation Fields from Magnetic Currents

803

Replacing A, A m in terms of Eq. (18.2.6), we may express the solutions (18.2.7) directly in terms of the current densities:
E=
H=

j
1

804

18. Radiation from Apertures

F +
F , and similarly for Fm ,
F and
F
r)=
rF=
Noting that
r (F
we nd for the polar components of Eq. (18.3.4):


 G j J m  G dV
k2 J G + (J  )

(18.2.9)


 G + j J  G dV
k2 J m G + (J m  )

E = jk


ejkr 
(F Fm )
(F + Fm )+
4r


jk ejkr 
(F + Fm )
H=
(F Fm )+
4r

(18.3.6)

Alternatively, if we also use the charge densities, we obtain from (18.2.4):


E=


H=

j J G +  G J m  G dV


The Poynting vector is given by the generalization of Eq. (16.1.1):


m 
G + J  G dV
j J m G +

(18.2.10)

and  jk

ejkr
F(, ) ,
4r

A m (r)= 

ejkr
Fm (, )
4r

(18.3.7)

(18.3.1)

(18.3.8)

18.4 Radiation Fields from Apertures

For an aperture antenna with effective surface currents given by Eq. (18.1.1), the volume
integrations in Eq. (18.2.9) reduce to surface integrations over the aperture A:

r. Then, the vector potentials of Eq. (18.2.6) take the simplied form:
where k = k
A(r)=


dP
k2 
= r 2 Pr =
|F + Fm |2 + |F Fm |2
d
322

U(, )=

The radiation elds of the solutions (18.2.7) can be obtained by making the far-eld
approximation, which consists of the replacements:

ejk|rr |
ejkr jkr

e

4|r r |
4r



1
k2
|F + Fm |2 + |F Fm |2 =
Re(E H )=
r
r Pr
322 r 2
2

and the radiation intensity:

18.3 Radiation Fields from Magnetic Currents

G(r r )=

P=

E=

(18.3.2)

H=

j


 G + k2 J s G j J ms  G dS
(J s  )


A


 G + k2 J ms G + j J s  G dS
(J ms  )

(18.4.1)

where the radiation vectors are the Fourier transforms of the current densities:


F(, ) =

jkr

J(r )e

dV


Fm (, ) =

and, explicitly in terms of the aperture elds shown in Fig. 18.1.1:



(radiation vectors)
jkr

J m (r )e

(18.3.3)

dV
H=

Setting J = J m = 0 in Eq. (18.2.8) because we are evaluating the elds far from the
current sources, and using the approximation = jk = jk
r, and the relationship
k/ = , we nd the radiated E and H elds:

E = j
r (A
r)
r A m = jk
H=

E=



ejkr

r F
r Fm
4r




jk ejkr
j 

r (A m
r)+
rA =
r F + Fm
r

4r

(18.3.4)

H=

rE

j
1


 (
 G)+k2 (n
 G dS
H a )G + j(n
E a )
H a )
(n


A


 (
 G)k2 (n
 G dS
E a )
E a )G + j(n
H a )
(n

(18.4.2)
These are known as Kottlers formulas [12931298,1287,12991302,1324]. We derive
them in Sec. 18.12. The equation for H can also be obtained from that of E by the
application of a duality transformation, that is, E a H a , H a E a and  , .
In the far-eld limit, the radiation elds are still given by Eq. (18.3.6), but now the
radiation vectors are given by the two-dimensional Fourier transform-like integrals over
the aperture:

These generalize Eq. (15.10.2) to magnetic currents. As in Eq. (15.10.3), we have:

1

(18.3.5)

F(, ) =


Fm (, ) =

J s (r )ejkr dS =




A

H a (r )ejkr dS

n

J ms (r )ejkr dS =

(18.4.3)


E a (r )ejkr dS

n

18.4. Radiation Fields from Apertures

805

806

18. Radiation from Apertures

gx + y
gy )= y
gx x
gy
zg=
z (x
F=
fx + y
fy )= x
fy y
fx
Fm =
z f =
z (x

(18.4.7)

The polar components of the radiation vectors are determined as follows:

F=
(y
gx x
gy )= gx sin cos gy cos cos
F =
x
y
) and (
) from Eq. (15.8.3). The remaining
where we read off the dot products (
polar components are found similarly, and we summarize them below:

F = cos (gy cos gx sin )

F = gx cos + gy sin
Fm = cos (fy cos fx sin )
Fig. 18.4.1 Radiation elds from an aperture.

(18.4.8)

Fm = (fx cos + fy sin )

It follows from Eq. (18.3.6) that the radiated E-eld will be:

Fig. 18.4.1 shows the polar angle conventions, where we took the origin to be somewhere in the middle of the aperture A.
The aperture surface A and the screen in Fig. 18.1.1 can be arbitrarily curved. However, a common case is to assume that they are both at. Then, Eqs. (18.4.3) become
ordinary 2-d Fourier transform integrals. Taking the aperture plane to be the xy-plane
=
as in Fig. 18.1.1, the aperture normal becomes n
z, and thus, it can be taken out of
the integrands. Setting dS = dx dy , we rewrite Eq. (18.4.3) in the form:


F(, ) =


Fm (, ) =
jkr

J s (r )ejkr dx dy =

z



A

J ms (r )ejkr dx dy =

z

H a (r )ejkr dx dy


A

E = jk


ejkr 
(fx cos + fy sin )+ cos (gy cos gx sin )
4r


ejkr 
cos (fy cos fx sin )(gx cos + gy sin )
E = jk
4r

The radiation elds resulting from the alternative forms of the eld equivalence
principle, Eqs. (18.1.2) and (18.1.3), are obtained from Eq. (18.4.9) by removing the g- or
the f -terms and doubling the remaining term. We have for the PEC case:

(18.4.4)

E = 2jk

E a (r )ejkr dx dy

=e
and kx = k cos sin , ky = k sin sin . It proves convewhere e
nient then to introduce the two-dimensional Fourier transforms of the aperture elds:


f(, )=

g(, )=
A

jkr

E a (r )e

dx dy =

H a (r )ejkr dx dy =


ejkr 
fx cos + fy sin
4r


ejkr 
cos (fy cos fx sin )
E = 2jk
4r

jkx x +jky y

(18.4.9)

(18.4.10)

and for the PMC case:



jkx x +jky y

E a (x , y )e

dx dy

(18.4.5)


H a (x , y )ejkx x +jky y dx dy

Fm (, ) =
z f(, )


ejkr 
cos (gy cos gx sin )
4r


ejkr 
(gx cos + gy sin )
E = 2jk
4r

Then, the radiation vectors become:

F(, ) =
z g(, )

E = 2jk

(18.4.11)

In all three cases, the radiated magnetic elds are obtained from:
(18.4.6)

Because E a , H a are tangential to the aperture plane, they can be resolved into their
Eax + y
Eay . Then, the quantities f, g can be
cartesian components, for example, E a = x
fx + y
fy . Thus, we have:
resolved in the same way, for example, f = x

H =

E ,

H =

(18.4.12)

We note that Eq. (18.4.9) is the average of Eqs. (18.4.10) and (18.4.11). Also, Eq. (18.4.11)
is the dual of Eq. (18.4.10). Indeed, using Eq. (18.4.12), we obtain the following Hcomponents for Eq. (18.4.11), which can be derived from Eq. (18.4.10) by the duality
transformation E a H a or f g , that is,

18.5. Huygens Source

807


ejkr 
gx cos + gy sin
H = 2jk
4r

ejkr 
cos (gy cos gx sin )
H = 2jk
4r

808

18. Radiation from Apertures

The Huygens source condition implies the same relationship for the Fourier trans=
z)
forms of the aperture elds, that is, (with n
(18.4.13)
g=

At = 90o , the components E , H become tangential to the aperture screen. We

note that because of the cos factors, E (resp. H ) will vanish in the PEC (resp. PMC)
case, in accordance with the boundary conditions.

f
n

18.5 Huygens Source

Ha =

Ea
n

(Huygens source)

(18.5.1)

where is the characteristic impedance of vacuum.

For example, this is the case if a uniform plane wave is incident normally on the
aperture plane from the left, as shown in Fig. 18.5.1. The aperture elds are assumed to
be equal to the incident elds, E a = Einc and H a = Hinc , and the incident elds satisfy
Hinc =
z Einc /.

gx =

gy =

fy ,

fx

(18.5.2)

Inserting these into Eq. (18.4.9) we may express the radiated electric eld in terms
of f only. We nd:

E = jk

The aperture elds E a , H a are referred to as Huygens source if at all points on the
aperture they are related by the uniform plane-wave relationship:


ejkr 1 + cos 
fx cos + fy sin
2r
2

(18.5.3)


ejkr 1 + cos 
fy cos fx sin
E = jk
2r
2

The factor (1 + cos )/2 is known as an obliquity factor. The PEC case of Eq. (18.4.10)
remains unchanged for a Huygens source, but the PMC case becomes:

E = jk



ejkr
cos fx cos + fy sin
2r

E = jk


ejkr 
fy cos fx sin
2r

(18.5.4)

We may summarize all three cases by the single formula:

E = jk


ejkr 
c fx cos + fy sin
2r



ejkr
E = jk
c fy cos fx sin
2r

(elds from Huygens source)

(18.5.5)

where the obliquity factors are dened in the three cases:

Fig. 18.5.1 Uniform plane wave incident on an aperture.

The Huygens source condition is not always satised. For example, if the uniform
plane wave is incident obliquely on the aperture, then must be replaced by the transverse impedance T , which depends on the angle of incidence and the polarization of
the incident wave as discussed in Sec. 7.2.
Similarly, if the aperture is the open end of a waveguide, then must be replaced by
the waveguides transverse impedance, such as TE or TM , depending on the assumed
waveguide mode. On the other hand, if the waveguide ends are ared out into a horn
with a large aperture, then Eq. (18.5.1) is approximately valid.

c
c


=

1
2

1 + cos
1 + cos

 
,

1
cos

 
,

cos
1


(obliquity factors)

(18.5.6)

We note that the rst is the average of the last two. The obliquity factors are equal to
unity in the forward direction = 0o and vary little for near-forward angles. Therefore,
the radiation patterns predicted by the three methods are very similar in their mainlobe
behavior.
In the case of a modied Huygens source that replaces by T , Eqs. (18.5.5) retain
their form. The aperture elds and their Fourier transforms are now assumed to be
related by:
1
1

Ha =

g=
(18.5.7)
z Ea
zf

Inserting these into Eq. (18.4.9), we obtain the modied obliquity factors :

c =

1
[1 + K cos ] ,
2

c =

1
[K + cos ] ,
2

K=

(18.5.8)

18.6. Directivity and Effective Area of Apertures

809

18.6 Directivity and Effective Area of Apertures

810

18. Radiation from Apertures

conservation and compute the total power that ows into the right half-space through
the aperture. Assuming a Huygens source, we have:

For any aperture, given the radiation elds E , E of Eqs. (18.4.9)(18.4.11), the corresponding radiation intensity is:


1 
1
dP
U(, )=
= r 2 Pr = r 2
|E |2 + |E |2 = r 2
|E(, )|2
d
2
2

(18.6.1)


Prad =

U(, )d ,
0

Prad
UI =
4

(18.6.2)

The directive gain is computed by D(, )= U(, )/UI , and the normalized gain
by g(, )= U(, )/Umax . For a typical aperture, the maximum intensity Umax is
towards the forward direction = 0o . In the case of a Huygens source, we have:

U(, )=

k2
82

c2 |fx cos + fy sin |2 + c2 |fy cos fx sin |2

where |f|2max

k2
82

|fx cos + fy sin |2 + |fy cos fx sin |2

Umax =

Umax =

1
22

Dmax = 4

1 + cos
2

1 + cos
2

|E a (r )|2 dS

(18.6.8)

kx =ky =0




E a (r ) dS

1
22

|f |2max =

E a (r ) dS

22
A
1

(18.6.9)

E a (r ) dS

4
A
4Aeff
= 2 
=

2
 2

|E a (r )| dS

(directivity)

(18.6.10)

E a (r ) dS

A
= 
A
|E a (r )|2 dS

(effective area)

(18.6.11)

(18.6.5)

2

|f(, )|2
|f |2max

(18.6.6)

The square root of the gain is the (normalized) eld strength:

Umax
Prad

Aeff

In the case of Eq. (18.4.9) with c = c = (1 + cos )/2, this simplies further into:

|E(, )|
= g(, ) =
|E |max

It follows that the maximum effective area of the aperture is:

(18.6.4)

|f |2max

and the aperture efciency:

|f |2max

c2 |fx cos + fy sin |2 + c2 |fy cos fx sin |2

|fx |2 + |fy |2
=
|f |2max

1
2

Dividing (18.6.9) by (18.6.8), we nd the directivity:

=0

It follows that the normalized gain will be:

g(, )= c2

Therefore, the maximum intensity is given by:


k2 
k2
=
|fx |2 + |fy |2 =0 =
|f |2max
2
8
82


= |fx |2 + |fy |2 =0 . Setting k = 2/, we have:

g(, )=

z Re E a Ha dS =


jkr
|f|2max =

dS

E a (r )e

(18.6.3)

Assuming that the maximum is towards = 0o , then c = c = 1, and we nd for

the maximum intensity:

Umax =

 /2  2
0

1
2

Because = 0 corresponds to kx = ky = 0, it follows from the Fourier transform

denition (18.4.5) that:

Because the aperture radiates only into the right half-space 0 /2, the total
radiated power and the effective isotropic radiation intensity will be:

Prad =

Pz dS =

|f(, )|
|f |max

E a (r ) dS

Aeff
= A
ea =
1
A
A
|E a (r )|2 dS

(18.6.12)

The inequalities in Eqs. (18.6.11) and (18.6.12) can be thought of as special cases of
the Cauchy-Schwarz inequality. It follows that equality is reached whenever E a (r ) is
uniform over the aperture, that is, independent of r .
Thus, uniform apertures achieve the highest directivity and have effective areas equal
to their geometrical areas.
Because the integrand in the numerator of ea depends both on the magnitude and the
phase of E a , it proves convenient to separate out these effects by dening the aperture
taper efciency or loss, eatl , and the phase error efciency or loss, epel , as follows:

(18.6.7)

The power computed by Eq. (18.6.2) is the total power that is radiated outwards from
a half-sphere of large radius r . An alternative way to compute Prad is to invoke energy

(aperture efciency)

eatl

|E a (r )| dS

A

=
,
 2
A
|E a (r )| dS
A

epel

E a (r ) dS

=
 A

|E a (r )| dS

(18.6.13)

18.7. Uniform Apertures

811

812

18. Radiation from Apertures

so that ea becomes the product:

ea = eatl epel

(18.6.14)

We note that Eq. (18.6.10) was derived under the assumption that the aperture elds
were Huygens sources, that is, they were related by, Ha =
z Ea . This assumption
is not necessarily always true, but may be justied for electrically large apertures, i.e.,
with dimensions that are much larger than the wavelength .
The correct expression for the directivity is given by Eq. (19.7.10) of Sec. 19.7 and
compared with (18.6.10). Directivities that are much larger (even innite) than those of
the uniform case are theoretically possible, but almost impossible to realize in practice.
The issues associated with such superdirective apertures are discussed in Sec. 20.22.

ratio in the gain (18.6.7) and that f (0, )= (1/A)

gain and eld strengths:

|E(, )|
= g(, ) =
|E |max

In uniform apertures, the elds E a , H a are assumed to be constant over the aperture
area. Fig. 18.7.1 shows the examples of a rectangular and a circular aperture. For convenience, we will assume a Huygens source.

dS = 1, we nd for the normalized

1 + cos
2


|f (, )|

(18.7.3)

18.8 Rectangular Apertures

For a rectangular aperture of sides a, b, the area integral (18.7.2) is separable in the xand y-directions:

f (, )=

18.7 Uniform Apertures

ab

 a/2  b/2
a/2

b/2

ejkx x +jky y dx dy =

 a/2

a/2

ejkx x dx

 b/2
b/2

ejky y dy

where we placed the origin of the r integration in the middle of the aperture. The above
integrals result in the sinc-function patterns:

f (, )=

sin(kx a/2) sin(ky b/2)

sin(vx ) sin(vy )
=
kx a/2
ky b/2
vx
vy

(18.8.1)

where we dened the quantities vx , vy :

vx =

1
1
a
sin cos
kx a =
ka sin cos =
2
2

vy =

1
1
b
ky b =
kb sin sin =
sin sin
2
2

(18.8.2)

The pattern simplies along the two principal planes, the xz- and yz-planes, corresponding to = 0o and = 90o . We have:

f (, 0o ) =
Fig. 18.7.1 Uniform rectangular and circular apertures.

f (, 90o ) =

The eld E a can have an arbitrary direction, with constant x- and y-components,
E0x + y
E0y . Because E a is constant, its Fourier transform f(, ) becomes:
Ea = x


f(, )=
A

E a (r )ejkr dS = E a

ejkr dS A f (, ) E a

(18.7.1)

ejkr dS

(uniform-aperture pattern)

vx
sin(vy )

vy

sin (a/)sin
(a/)sin

sin (b/)sin
(b/)sin

(18.7.2)

The quantity f (, ) depends on the assumed geometry of the aperture and it, alone,
determines the radiation pattern. Noting that the quantity |E a | cancels out from the

(18.8.3)

Fig. 18.8.1 shows the three-dimensional pattern of Eq. (18.7.3) as a function of the
independent variables vx , vy , for aperture dimensions a = 8 and b = 4. The x, y
separability of the pattern is evident. The essential MATLAB code for generating this
gure was (note MATLABs denition of sinc(x)= sin(x)/(x)):
a = 8; b = 4;
[theta,phi] = meshgrid(0:1:90, 0:9:360);
theta = theta*pi/180; phi = phi*pi/180;

where we introduced the normalized scalar quantity:

f (, )=

sin(vx )

vx = a*sin(theta).*cos(phi);
vy = b*sin(theta).*sin(phi);
E = abs((1 + cos(theta))/2 .* sinc(vx) .* sinc(vy));
surfl(vx,vy,E);
shading interp;

colormap(gray(16));

18.8. Rectangular Apertures

813

814

18. Radiation from Apertures

Radiation Pattern for = 0o

Radiation Pattern for = 90o

1

0.5

3 dB

field strength

3 dB

field strength

field strength

13.26 dB

0.5

13.26 dB

0.5

0
8
8

4
vy

0
0

10

20

30

4
8 8

40

50

(degrees)

60

70

80

0
0

90

10

20

30

40

50

(degrees)

60

70

80

90

vx

Fig. 18.8.2 Radiation patterns along the two principal planes (a = 8, b = 4).

Fig. 18.8.1 Radiation pattern of rectangular aperture (a = 8, b = 4).

As the polar angles vary over 0 90o and 0 360o , the quantities vx and
vy vary over the limits a/ vx a/ and b/ vy b/. In fact, the physically
realizable values of vx , vy are those that lie in the ellipse in the vx vy -plane:

vy2
vx2
1
+
2
a2
b2

(visible region)

= 50.76o ,
a
a

y = 0.886

= 50.76o
b
b

G=

(18.8.4)

The realizable values of vx , vy are referred to as the visible region. The graph in
Fig. 18.8.1 restricts the values of vx , vy within that region.
The radiation pattern consists of a narrow mainlobe directed towards the forward
direction = 0o and several sidelobes.
We note the three characteristic properties of the sinc-function patterns: (a) the 3dB width in v-space is vx = 0.886 (the 3-dB wavenumber is vx = 0.443); (b) the rst
sidelobe is down by about 13.26 dB from the mainlobe and occurs at vx = 1.4303; and
(c) the rst null occurs at vx = 1. See Sec. 22.7 for the proof of these results.
The 3-dB width in angle space can be obtained by linearizing
the relationship vx =

(a/)sin about = 0o , that is, vx = (a/) cos

=0 = a/. Thus, =
vx /a. This ignores also the effect of the obliquity factor. It follows that the 3-dB
widths in the two principal planes are (in radians and in degrees):

x = 0.886

For aperture antennas, the gain is approximately equal to the directivity because the
losses tend to be very small. The gain of the uniform rectangular aperture is, therefore,
G  D = 4(ab)/2 . Multiplying G by Eqs. (18.8.5), we obtain the gain-beamwidth
product p = G x y = 4(0.886)2 = 9.8646 rad2 = 32 383 deg2 . Thus, we have an
example of the general formula (16.3.14) (with the angles in radians and in degrees):

(18.8.5)

The 3-dB angles are x = x /2 = 25.4o /a and y = y /2 = 25.4o /b.

Fig. 18.8.2 shows the two principal radiation patterns of Eq. (18.7.3) as functions of
, for the case a = 8, b = 4. The obliquity factor was included, but it makes essentially no difference near the mainlobe and rst sidelobe region, ultimately suppressing
the response at = 90o by a factor of 0.5.
The 3-dB widths are shown on the graphs. The rst sidelobes occur at the angles
a = asin(1.4303/a)= 10.30o and b = asin(1.4303/b)= 20.95o .

9.8646
32 383
=
x y
ox oy

(18.8.6)

18.9 Circular Apertures

For a circular aperture of radius a, the pattern integral (18.7.2) can be done conveniently
using cylindrical coordinates. The cylindrical symmetry implies that f (, ) will be
independent of .
Therefore, for the purpose of computing the integral (18.7.2), we may set = 0. We
have then k r = kx x = k sin cos  . Writing dS =  d d , we have:

f ()=

 a  2

a2

ejk

sin cos 

 d d

(18.9.1)

The  - and  -integrations can be done using the following integral representations
for the Bessel functions J0 (x) and J1 (x) [1790]:

J0 (x)=

1
2

 2

ejx cos d

1
and

J0 (xr)r dr =

J1 (x)
x

(18.9.2)

Then Eq. (18.9.1) gives:

f ()= 2

J1 (ka sin )
J1 (2u)
,
=2
ka sin
2u

u=

1
a
sin
ka sin =
2

(18.9.3)

This is the well-known Airy pattern [638] for a circular aperture. The function f ()
is normalized to unity at = 0o , because J1 (x) behaves like J1 (x) x/2 for small x.

815

Fig. 18.9.1 shows the three-dimensional eld pattern (18.7.3) as a function of the independent variables vx = (a/)sin cos and vy = (a/)sin sin , for an aperture
radius of a = 3. The obliquity factor was not included as it makes little difference
near the main lobe. The MATLAB code for this graph was implemented with the built-in
function besselj:

816

18. Radiation from Apertures

Radiation Pattern of Circular Aperture
1
3 dB

field strength

18.9. Circular Apertures

17.56 dB

0.5

field strength

0
0

10

0.5

30

40

50

(degrees)

60

70

80

90

Fig. 18.9.2 Radiation pattern of circular aperture (a = 3).

0
3

The 3-dB angle is 3dB = 3dB /2 = 0.2572/a = 14.74o /a and the rst-null
angle null = 0.6098/a. Fig. 18.9.2 shows the radiation pattern of Eq. (18.7.3) as a
function of , for the case a = 3. The obliquity factor was included.
The graph shows the 3-dB width and the rst sidelobe, which occurs at the angle a =
asin(0.817/a)= 15.8o . The rst null occurs at null = asin(0.6098/a)= 11.73o ,
whereas the approximation null = 0.6098/a gives 11.65o .


The gain-beamwidth product is p = G(3dB )2 = 4(a2 )/2 (0.514/a)2 =
2
2
2
2
4 (0.5144) = 10.4463 rad = 34 293 deg . Thus, in radians and degrees:

3
0
vy

0
vx

3 3

Fig. 18.9.1 Radiation pattern of circular aperture (a = 3).

a = 3;
[theta,phi] = meshgrid(0:1:90, 0:9:360);
theta = theta*pi/180; phi = phi*pi/180;

G=

vx = a*sin(theta).*cos(phi);
vy = a*sin(theta).*sin(phi);
u = a*sin(theta);

10.4463
34 293
=
(3dB )2
(o3dB )2

(18.9.6)

The rst-null angle null = 0.6098/a is the so-called Rayleigh diffraction limit for
the nominal angular resolution of optical instruments, such as microscopes and telescopes. It is usually stated in terms of the diameter D = 2a of the optical aperture:

E = ones(size(u));
i = find(u);
E(i) = abs(2*besselj(1,2*pi*u(i))./(2*pi*u(i)));
surfl(vx,vy,E);
shading interp;

20

= 1.22

= 70o
D
D

(Rayleigh limit)

(18.9.7)

colormap(gray(16));

18.10 Vector Diffraction Theory

The visible region is the circle on the vx vy -plane:

vx2 + vy2

a
2

(18.9.4)

The mainlobe/sidelobe characteristics of f () are as follows. The 3-dB wavenumber

is u = 0.2572 and the 3-dB width in u-space is u = 20.2572 = 0.5144. The rst null
occurs at u = 0.6098 so that the rst-null width is u = 20.6098 = 1.22. The rst
sidelobe occurs at u = 0.8174 and its height is |f (u)| = 0.1323 or 17.56 dB below the
mainlobe. The beamwidths in angle space can be obtained from u = a()/, which
gives for the 3-dB and rst-null widths in radians and degrees:

3dB = 0.5144

= 29.47o ,
a
a

null = 1.22

= 70o
a
a

(18.9.5)

In this section, we provide a justication of the eld equivalence principle (18.1.1) and
Kottlers formulas (18.4.2) from the point of view of vector diffraction theory. We also
discuss the Stratton-Chu and Franz formulas. A historical overview of this subject is
given in [1302,1324].
In Sec. 18.2, we worked with the vector potentials and derived the elds due to
electric and magnetic currents radiating in an unbounded region. Here, we consider the
problem of nding the elds in a volume V bounded by a closed surface S and an innite
spherical surface S , as shown in Fig. 18.10.1.
The solution of this problem requires that we know the current sources within V
and the electric and magnetic elds tangential to the surface S. The elds E1 , H1 and

18.10. Vector Diffraction Theory

817

818

18. Radiation from Apertures

Using Eq. (18.10.2), the second term on the left may be integrated to give E(r):

E(r ) (2 G + k2 G) dV =

E(r ) (3) (r r ) dV = E(r)

where we assumed that r lies in V. This integral is zero if r lies in V1 because then r
can never be equal to r. For arbitrary r, we may write:

E(r )

(3)

E(r), if r V
(r r ) dV = uV (r) E(r)=
0,
if r
 V


(18.10.4)

where uV (r) is the characteristic, or indicator, function of the volume region V:

1, if r V
uV (r)=
0, if r
 V

Fig. 18.10.1 Fields outside a closed surface S.

current sources inside the volume V1 enclosed by S have an effect on the outside only
through the tangential elds on the surface.
We start with Maxwells equations (18.2.1), which include both electric and magnetic
currents. This will help us identify the effective surface currents and derive the eld
equivalence principle.
Taking the curls of both sides of Amp`
eres and Faradays laws and using the vector
E)= (
E)2 E, we obtain the following inhomogeneous Helmholtz
identity (
equations (which are duals of each other):

2 E + k2 E = j J +

2 H + k2 H = j J m +

+ J m
1

(18.10.1)

m J

We may now solve Eq. (18.10.3) for E(r). In a similar fashion, or, performing a duality
transformation on the expression for E(r), we also obtain the corresponding magnetic
eld H(r). Using (18.10.1), we have:

 

 


G
E
dS
G  G  J m dV +
E
G



n
n
V
S

 
 

1
G
H
H(r) =
H
G
j G J m G  m + G  J dV +
dS



n
n
V
S
j G J

E(r) =

G(r r )=

 


G2 E E 2 G dV =

(18.10.2)

4|r r |

G
S+S

E
G
dS ,
E
n
n


+

V
The

 

E
G
G
dS
E
G( E + k E)E ( G + k G) dV =


n
n
S

2

2

precise conditions are: r|E| const. and r|E H

r| 0 as r .

(18.10.3)



G J m  G dV



 G dS
H)+(n
E)  G + (n
E)
j G(n

 
H(r)=


=n
n

where G and E stand for G(r r ) and E(r ) and the integration is over r . The quantity
. The negative sign in the right-hand side
/n is the directional derivative along n
that is pointing into the volume V.
arises from using a unit vector n
The integral over the innite surface is taken to be zero. This may be justied more
rigorously [1295] by assuming that E and H behave like radiation elds with asymptotic
form E const. ejkr /r and H
r E/. Thus, dropping the S term, and adding
2
and subtracting k G E in the left-hand side, we obtain:

j G J +

E(r)=

jk|rr |

where  is the gradient with respect to r . Applying Greens second identity given by
Eq. (C.27) of Appendix C, we obtain:

(18.10.6)
Because of the presence of the particular surface term, we will refer to these as
the Kirchhoff diffraction formulas. Eqs. (18.10.6) can be transformed into the so-called
Stratton-Chu formulas [12931298,1287,12991302,1324]:

We recall that the Greens function for the Helmholtz equation is:

2 G + k2 G = (3) (r r ) ,

(18.10.5)


+

j G J m +


m 
G + J  G dV

(18.10.7)


 G dS
E)+(n
H)  G + (n
H)
j G(n

The proof of the equivalence of (18.10.6) and (18.10.7) is rather involved. Problem
18.4 breaks down the proof into its essential steps.
Term by term comparison of the volume and surface integrals in (18.10.7) yields the
effective surface currents of the eld equivalence principle:
H,
Js = n
Technically

E
J ms = n

(18.10.8)

[1301], one must set uV (r)= 1/2, if r lies on the boundary of V, that is, on S.
[1289,1296,1302,1324] for earlier work by Larmor, Tedone, Ignatowski, and others.
Initially derived by Larmor and Love [1302,1324], and later developed fully by Schelkunoff [1288,1290].
See

18.10. Vector Diffraction Theory

819

Similarly, the effective surface charge densities are:

E,
s =  n

H
ms = n

(18.10.9)

Eqs. (18.10.7) may be transformed into the Kottler formulas [12931298,1287,1299

1302,1324], which eliminate the charge densities , m in favor of the currents J, J m :
E(r)=

+
H(r)=

 

j

j


 G j J m  G dV
k2 J G + (J  )

Next, we specialize the above formulas to the case where the volume V contains
no current sources (J = J m = 0), so that the E, H elds are given only in terms of the
surface integral terms.
This happens if we choose S in Fig. 18.10.1 such that all the current sources are
inside it, or, if in Fig. 18.10.2 we choose S such that all the current sources are outside
it, then, the Kirchhoff, Stratton-Chu, Kottler, and Franz formulas simplify into:

 
E(r) =




  G + j(n
 G dS
H)+ (n
H)
E)
k2 G(n

 

18. Radiation from Apertures

820


 G + j J  G dV
k2 J m G + (J m  )


S

(18.10.10)
The steps of the proof are outlined in Problem 18.5.
A related problem is to consider a volume V bounded by the surface S, as shown in
pointing again into
Fig. 18.10.2. The elds inside V are still given by (18.10.7), with n
the volume V. If the surface S recedes to innity, then (18.10.10) reduce to (18.2.9).




  G + j(n
 G dS
E) (n
E)
H)
k2 G(n

E(r) =
H(r) =

j

j
1

j

H(r) =




  G + j(n
 G dS
H )+ (n
H )
E )
k2 G(n

H )dS +
G(n


S

E )dS
G(n
(18.10.13)


G
H
dS
H
G


n
n


 G dS
H )
E )+(n
H )  G + (n
j G(n


Finally, the Kottler formulas may be transformed into the Franz formulas [1298,1287,1299
1301], which are essentially equivalent to Eq. (18.2.8) amended by the vector potentials
due to the equivalent surface currents:


 G dS
H )+(n
E )  G + (n
E )
j G(n

 

Fig. 18.10.2 Fields inside a closed surface S.


G
E

G
dS
n
n




  G + j(n
 G dS
E ) (n
E )
H )
k2 G(n


S


S

E )dS +
G(n


S

H )dS
G(n

(18.10.14)
where the last equations are the Franz formulas with A = A m = 0.
Fig. 18.10.3 illustrates the geometry of the two cases. Eqs. (18.10.13) and (18.10.14)
represent the vectorial formulation of the Huygens-Fresnel principle, according to which
the tangential elds on the surface can be considered to be the sources of the elds away
from the surface.

 1


(A + A s ) J (A m + A ms )




 1

(A m + A ms )  J m + (A + A s )
j

(18.10.11)

where A and A m were dened in Eq. (18.2.6). The new potentials are dened by:

A s (r) =


A ms (r) =

J s (r )G(r r )dS =

Fig. 18.10.3 Current sources are outside the eld region.



H(r ) G(r r )dS
n

 J ms (r )G(r r )dS =


S



E(r ) G(r r )dS
 n

(18.10.12)

18.11. Extinction Theorem

821

18.11 Extinction Theorem

In all of the equivalent formulas for E(r), H(r), we assumed that r lies within the volume
V. The origin of the left-hand sides in these formulas can be traced to Eq. (18.10.4), and
therefore, if r is not in V but is within the complementary volume V1 , then the left-hand
sides of all the formulas are zero. This does not mean that the elds inside V1 are
zeroit only means that the sum of the terms on the right-hand sides are zero.
To clarify these remarks, we consider an imaginary closed surface S dividing all
space in two volumes V1 and V, as shown in Fig. 18.11.1. We assume that there are
current sources in both regions V and V1 . The surface S1 is the same as S but its unit
1 points into V1 , so that n
1 = n
. Applying (18.10.10) to the volume V, we have:
vector n

Fig. 18.11.1 Current sources may exist in both V and V1 .

j




  G + j(n
 G dS
H)+ (n
H)
E)
k2 G(n


S

 
E(r), if r V

1
2




G j J m G dV =
k J G + (J )
+
0,
j V
if r V1
The vanishing of the right-hand side when r is in V1 is referred to as an extinction
theorem. Applying (18.10.10) to V1 , and denoting by E1 , H1 the elds in V1 , we have:

j
+




  G + j(n
 G dS
1 H1 )+ (n
1 H1 )
1 E1 )
k G(n

S1

0,

if r V
2




G j J m G dV =
k J G + (J )
E1 (r), if r V1

j

V1

1 = n
, and on the surface E1 = E and H1 = H, we may rewrite:
Because n

j




  G + j(n
 G dS
H)+ (n
H)
E)
k2 G(n

j

V1


 G j J m  G dV =
k2 J G + (J  )

0,

if r V
E1 (r), if r V1

Adding up the two cases and combining the volume integrals into a single one, we obtain:
1

j
In

18. Radiation from Apertures

This is equivalent to Eq. (18.2.9) in which the currents are radiating into unbounded
space. We can also see how the sources within V1 make themselves felt on the outside
only through the tangential elds at the surface S, that is, for r V :


V+V1

E(r), if r V

 G + k2 GJ j J m  G dV =
(J  )
E1 (r), if r V1

fact, it can be used to prove the Ewald-Oseen extinction theorem that we considered in Sec. 15.6.

j
1

V1


 G j J m  G dV
k2 J G + (J  )

j




  G + j(n
 G dS
H)+ (n
H)
E)
k2 G(n

18.12 Vector Diffraction for Apertures

The Kirchhoff diffraction integral, Stratton-Chu, Kottler, and Franz formulas are equivalent only for a closed surface S.
If the surface is open, as in the case of an aperture, the four expressions in (18.10.13)
and in (18.10.14) are no longer equivalent. In this case, the Kottler and Franz formulas
remain equal to each other and give the correct expressions for the elds, in the sense
that the resulting E(r) and H(r) satisfy Maxwells equations [1289,1287,1302,1324].
For an open surface S bounded by a contour C, shown in Fig. 18.12.1, the Kottler
and Franz formulas are related to the Stratton-Chu and the Kirchhoff diffraction integral
formulas by the addition of some line-integral correction terms [1296]:

E(r)=

j




  G + j(n
 G dS
H )+ (n
H )
E )
k2 G(n

H )dS +
G(n

E )dS
G(n
j
S
S



 G dS
H )+(n
E )  G + (n
E )
=
j G(n

 
=

=

=

j
1

H
S

j

 G)H dl
(




  G + j(n
 G dS
E ) (n
E )
H )
k G(n
2


S

E )dS +
G(n


S

H )dS
G(n


 G dS +
E )+(n
H )  G + (n
H )
j G(n

 
=

(18.12.1)




G
1
E
 G)H dl
G
G E dl
(
dS


n
n
j C
C

H(r)=

822




G
1
H
 G)E dl
G
G H dl +
(
dS


n
n
j C
C


C

 G)E dl
(

(18.12.2)
The proof of the equivalence of these expressions is outlined in Problems 18.7 and
18.8. The Kottler-Franz formulas (18.12.1) and (18.12.2) are valid for points off the
aperture surface S. The formulas are not consistent for points on the aperture. However,
they have been used very successfully in practice to predict the radiation patterns of
aperture antennas.

18.13. Fresnel Diffraction

823

824

18. Radiation from Apertures

Fig. 18.12.1 Aperture surface S bounded by contour C.

The line-integral correction terms have a minor effect on the mainlobe and near
sidelobes of the radiation pattern. Therefore, they can be ignored and the diffracted
eld can be calculated by any of the four alternative formulas, Kottler, Franz, StrattonChu, or Kirchhoff integralall applied to the open surface S.

18.13 Fresnel Diffraction

 
S

E1


G
E1
dS

G
n
n

(18.13.1)

where E1 is the spherical wave from the source point P1 evaluated at the aperture point
P , and G is the Greens function from P to P2 :

E1 = A1

where A1 is a constant. If r1 and r2 are the vectors pointing from the origin to the source
and observation points, then we have for the distance vectors R1 and R2 :
R1 = r1 r ,

In Sec. 18.4, we looked at the radiation elds arising from the Kottler-Franz formulas,
where we applied the Fraunhofer approximation in which only linear phase variations
over the aperture were kept in the propagation phase factor ejkR . Here, we consider
the intermediate case of Fresnel approximation in which both linear and quadratic phase
variations are retained.
We discuss the classical problem of diffraction of a spherical wave by a rectangular
aperture, a slit, and a straight-edge using the Kirchhoff integral formula. The case of a
plane wave incident on a conducting edge is discussed in Problem 18.11 using the eldequivalence principle and Kottlers formula and more accurately, in Sec. 18.15, using
Sommerfelds exact solution of the geometrical theory of diffraction. These examples
are meant to be an introduction to the vast subject of diffraction.
In Fig. 18.13.1, we consider a rectangular aperture illuminated from the left by a point
source radiating a spherical wave. We take the origin to be somewhere on the aperture
plane, but eventually we will take it to be the point of intersection of the aperture plane
and the line between the source and observation points P1 and P2 .
The diffracted eld at point P2 may be calculated from the Kirchhoff formula applied
to any of the cartesian components of the eld:

E=

Fig. 18.13.1 Fresnel diffraction through rectangular aperture.

jkR1

R1

G=

R2 = r2 r ,

1
 = R

4R2

,
R1

2
 = R

R2

1 and R
2 are the unit vectors in the directions of R1 and R2 . Thus, we have:
where R



E1
E1
1
ejkR1

1
1 ) jk +

n
n
n
=

E
=

=
(

A1
R
R
1
n
R1
R1
R1
 jkR2

e
G
G
1
2
2 ) jk +
 G = n
R
R
=n
= (n
n
R2
R2 4R2

(18.13.4)

Dropping the 1/R2 terms, we nd for the integrand of Eq. (18.13.1):

E1


G
E1
jkA1 
2 )(n
1 ) ejk(R1 +R2 )
R
R
G
=
(n
n
n
4R1 R2

Except in the phase factor ejk(R1 +R2 ) , we may replace R1  r1 and R2  r2 , that is,

E1


G
E1
jkA1 


G
=
(n
r2 )(n
r1 ) ejk(R1 +R2 )
n
n
4r1 r2

(18.13.5)

Thus, we have for the diffracted eld at point P2 :

(18.13.2)

(18.13.3)

Therefore, the gradient operator  can be written as follows when it acts on a function
of R1 = |r1 r | or a function of R2 = |r2 r |:

jkR2

R1 = |r1 r | = r12 2r1 r + r r

R2 = |r2 r | = r22 2r2 r + r r

E=


jkA1 


(n
r2 )(n
r1 )
4r1 r2

ejk(R1 +R2 ) dS

S

(18.13.6)

18.13. Fresnel Diffraction

825

r2 )(n
r1 ) is an obliquity factor. Next, we set r = r1 + r2 and
The quantity (n
dene the free-space eld at the point P2 :

ejk(r1 +r2 )
ejkr
= A1
r1 + r 2
r

E0 = A1

If the origin were the point of intersection between the aperture plane and the line
when the aperture and screen are absent.
The ratio D = E/E0 may be called the diffraction coefcient and depends on the
aperture and the relative geometry of the points P1 , P2 :


E
jk 


(n
=
r2 )(n
r1 )
E0
4F

18. Radiation from Apertures

It follows that the Fresnel approximation of the diffraction coefcient for an arbitrary
aperture will be given by:

(18.13.7)

P1 P2 , then E0 would represent the eld received at point P2 in the unobstructed case

D=

826

ejk(R1 +R2 r1 r2 ) dS

D=

E
jk(n
r2 )
=
E0
2F

ejk(b b )/(2F) dS

(18.13.11)

A further simplication is obtained by assuming that the aperture plane is the xyplane and that the line P1 P2 lies on the yz plane at an angle with the z-axis, as shown
in Fig. 18.13.2.

(18.13.8)

where we dened the focal length between r1 and r2 :

1

r1

r2

F=

r1 r2
r1 + r2

(18.13.9)

The Fresnel approximation is obtained by expanding R1 and R2 in powers of r  and

keeping only terms up to second order. We rewrite Eq. (18.13.3) in the form:


R1 = r1

2
r1 r

r1

r r

r12


R2 = r2

2
r2 r

r2

r r

r22

Next, we apply the Taylor series expansion up to second order:

1+x=1+

1
1
x x2
2
8

Fig. 18.13.2 Fresnel diffraction by rectangular aperture.

=
sin . It follows that
+ y y
, n
Then, we have r = x x
z cos + y
z, and
r2 =

n
r2 = cos , and the perpendicular distance b b becomes:

This gives the approximations of R1 , R2 , and R1 + R2 r1 r2 :

R1 = r1
r1 r +


1   
r r (
r1 r )2
2r1

r2 r +
R2 = r2


1   
r r (
r2 r )2
2r2

1
R1 + R2 r1 r2 = (
r1 +
r2 )r +
2




b b = r r (
r2 )2 = x2 + y2 (y sin )2 = x2 + y2 cos2
r
Then, the diffraction coefcient (18.13.11) becomes:
1

r1

r2

r1 r )2
r2 r )2
(
(
r r

r1
r2


To simplify this expression, we now assume that the origin is the point of intersection
of the line of sight P1 P2 and the aperture plane. Then, the vectors r1 and r2 are antiparallel and so are their unit vectors
r1 =
r2 . The linear terms cancel and the quadratic
ones combine to give:

2

1   
1

1  

r 
R1 +R2 r1 r2 =
r r (
b b (18.13.10)
r2 r )2 =
r2 (r
r2 )
=
2F
2F
2F
where we dened b = r
r2 (r
r2 ), which is the perpendicular vector from the point
P to the line-of-sight P1 P2 , as shown in Fig. 18.13.1.

D=

jk cos
2F

 x2  y2
x1

y1

2 +y2

ejk(x

cos2 )/2F

dx dy

(18.13.12)

where we assumed that the aperture limits are (with respect to the new origin):

x1 x x2 ,

y1 y y2

The end-points y1 , y2 are shown in Fig. 18.13.2. The integrals may be expressed
in terms of the Fresnel functions C(x), S(x), and F(x)= C(x)jS(x) discussed in
Appendix F. There, the complex function F(x) is dened by:

x
F(x)= C(x)jS(x)=

ej(/2)u du
2

(Fresnel function)

(18.13.13)

18.14. Knife-Edge Diffraction

827

828

18. Radiation from Apertures

We change integration variables to the normalized Fresnel variables:


u=

k 
x ,
F

v=

k 
y cos
F

(18.13.14)

where b = y cos is the perpendicular distance from P to the line P1 P2 , as shown in

Fig. 18.13.2. The corresponding end-points are:


ui =

k
xi ,
F


vi =

k
yi cos =
F

k
bi ,
F

i = 1, 2

(18.13.15)

Note that the quantities b1 = y1 cos and b2 = y2 cos are the perpendicular
distances from the edges to the line P1 P2 . Since du dv = (k cos /F)dx dy , we
obtain for the diffraction coefcient:

D=

 u2

u1

eju

/2

 v2
du

v1

ejv

/2

dv =

j



F(v2 )F(v1 )

F(u2 )F(u1 )

Noting that F(x) is an odd function and that j/2 = 1/(1 j)2 , we obtain:

D=

E
F(u1 )+F(u2 ) F(v1 )+F(v2 )
=
E0
1j
1j

(rectangular aperture)

Fig. 18.14.1 Illuminated and shadow regions in straight-edge diffraction.

Positive values of v correspond to positive values of the clearance distance b1 , placing the point P2 in the illuminated region, as shown in Fig. 18.14.1. Negative values of
v correspond to b1 < 0, placing P2 in the geometrical shadow region behind the edge.
The magnitude-square |D|2 represents the intensity of the diffracted eld relative
to the intensity of the unobstructed eld. Since |1 j|2 = 2, we nd:

(18.13.16)

|D(v)|2 =

The normalization factors (1 j) correspond to the innite aperture limit u1 , u2 , v1 ,

v2 , that is, no aperture at all. Indeed, since the asymptotic value of F(x) is
F()= (1 j)/2, we have:

D=

E
F(v1 )+F(v2 )
=
E0
1j

(diffraction by long slit)

(18.13.17)

18.14 Knife-Edge Diffraction

The case of straight-edge or knife-edge diffraction is obtained by taking the limit y2
, or v2 , which corresponds to keeping the lower edge of the slit. In this limit
F(v2 ) F()= (1 j)/2. Denoting v1 by v, we have:
1

D(v)=

1j

F(v)+

1j
2


v=

k
b1
F

(18.14.1)

(18.14.3)

or, in terms of the real and imaginary parts of F(v):

|D(v)|2 =

F()+F() F()+F()
F(u1 )+F(u2 ) F(v1 )+F(v2 )

=1
1j
1j
1j
1j
In the case of a long slit along the x-direction, we only take the limit u1 , u2 :

|E|2

F(v)+ 1 j

2
|E0 |
2
2

1
2


C(v)+

1
2

2

 

1 2
+ S(v)+
2

(18.14.4)

The quantity |D(v)|2 is plotted versus v in Fig. 18.14.2. At v = 0, corresponding to

the line P1 P2 grazing the top of the edge, we have F(0)= 0, D(0)= 1/2, and |D(0)|2 =
1/4 or a 6 dB loss. The rst maximum in the illuminated region occurs at v = 1.2172
and has the value |D(v)|2 = 1.3704, or a gain of 1.37 dB.
The asymptotic behavior of D(v) for v is obtained from Eq. (F.4). We have
for large positive x:


1j
j jx2 /2
e
F(x)
+
2
x
This implies that:

1 j jv2 /2

, for v +

1 2v e
D(v)=

1 j jv2 /2

,
for v
e
2v

(18.14.5)

We may combine the two expressions into one with the help of the unit-step function

where,

v
F(v)=

eju

/2

du ,

D(v)=

1
1j

v

eju

/2

du

(18.14.2)

u(v) by writing D(v) in the following form, which denes the asymptotic diffraction
coefcient d(v):
D(v)= u(v)+d(v)ejv

/2

(18.14.6)

18.14. Knife-Edge Diffraction

829

830

18. Radiation from Apertures

Diffraction Coefficient in dB

Diffraction Coefficient
1.5
1.25

20 log10|D()|

|D()|2

1
0.75

12

0.5

18

0.25
0
3

24
3

Fig. 18.14.3 Communicating antennas over an obstacle.

Fig. 18.14.2 Diffraction coefcient in absolute and dB units.

be expressed in terms of the heights:

where u(v)= 1 for v 0 and u(v)= 0 for v < 0.

With u(0)= 1, this denition requires d(0)= D(0)v(0)= 0.5 1 = 0.5. But if
we dene u(0)= 0.5, as is sometimes done, then, d(0)= 0. The asymptotic behavior of
D(v) can now be expressed in terms of the asymptotic behavior of d(v):
1j
,
2v

for

2


L = 10 log10
d(v)
= 10 log10 22 v2 ,

as v

For values v 0.7, the diffraction loss can be approximated very well by the following function [1309]:

b1 = l1 l2 1 2

(18.14.10)

and in terms of , we have:

1 =

l2
,
l1 + l2

2 =

l1
l1 + l2

b1 = F

v=

2F

(18.14.11)

The case of multiple obstacles has been studied using appropriate modications of the
knife-edge diffraction problem and the geometrical theory of diffraction [13981413]. 


the corresponding Fresnel zone ellipsoids help answer the question of what the minimum
value of the clearance b should be for efcient communication between the antennas.
Diffraction Coefficient in dB
3

(18.14.9)

Example 18.14.1: Diffraction Loss over Obstacles. The propagation path loss over obstacles and
irregular terrain is usually determined using knife-edge diffraction. Fig. 18.14.3 illustrates
the case of two antennas communicating over an obstacle. For small angles , the focal
length F is often approximated in several forms:

F=

Example 18.14.2: Fresnel Zones. Consider two antennas separated by a distance d and an obstacle at distance z from the midpoint with clearance b, as shown below. Fresnel zones and

% knife-edge diffraction coefcient D(v)




2
L = 10 log10
D(v)
= 6.9 + 20 log10
(v + 0.1)2 +1 v 0.1

b1  l1 1  l2 2

(18.14.8)

The MATLAB function diffr.m, mentioned in Appendix F, calculates the diffraction

coefcient (18.14.1) at any vector of (real) values of v. It has usage:
D = diffr(v);

The distance b1 can also be expressed approximately in terms of the subtended angles 1 ,
2 , and , shown in Fig. 18.14.3:

(18.14.7)

In the illuminated region D(v) tends to unity, whereas in the shadow region it decreases to zero with asymptotic dB attenuation or loss:


h1 d2 + h2 d1
h cos
d1 + d2

r1 r2
d1 d2
l1 l2


r1 + r 2
d1 + d2
l1 + l2

These approximations are valid typically when d1 , d2 are much greater than and the
height h of the obstacle, typically, at least ten times greater. The clearance distance can

exact
asymptotic
extrema
fresnel zone

20 log10|D()|

d(v)=


b1 = y1 cos =

1
0

1
2
3
0

18.14. Knife-Edge Diffraction

831

The diffraction coefcient D(v) and its asymptotic form were given in Eqs. (18.14.1) and
(18.14.5), that is,

D(v)=

1
1j

F(v)+

1j
2


v=

k
b=
F

b,

F=

d1 d2
d1 + d2

1 j jv2 /2
1
2
e
ej(v /2+1/4)
=1
2v
2v

18. Radiation from Apertures

If the obstacle is at midpoint (z = 0), the minimum clearance becomes:

b=

(18.14.13)

A common interpretation and derivation of Fresnel zones is to consider the path difference
between the rays following the straight path connecting the two antennas and the path
getting scattered from the obstacle, that is, l = l1 + l2 d. From the indicated triangles,
and assuming that b  d1 and b  d2 , we nd:

b2
l1 = d21 + b2  d1 +
,
2d1

ejn = 1

2vn

(1)n

(18.14.15)

An alternative set of vs, also corresponding to alternating almost extremum values, are
those that dene the conventional Fresnel zones, that is,

un =

n = 1, 2, . . .

2n ,

(18.14.16)

These are indicated by open circles on the graph. The corresponding D(v) values are:

ej/4
Das (un )= 1
(1)n
2un

(18.14.17)

For clearances b that correspond to vs that are too small, i.e., v < 0.5, the diffraction
coefcient D(v) becomes too small, impeding efcient communication. The smallest acceptable clearance b is taken to correspond to the rst maximum of D(v), that is, v = v1

or more simply v = u1 = 2.
The locus of points (b, z) corresponding to a xed value of v, and hence to a xed value
of the diffraction coefcient D(v), form an ellipsoid. This can be derived from (18.14.12)
by setting d1 = d/2 + z and d2 = d/2 z, that is,


v=

b2 =

F
2

v2 =

(d2 /4 z2 ) 2
v ,
2d

because F =

d1 d2
d2 /4 z2
=
d1 + d2
d

l = l1 + l2 d =

v2 d

b2 +

d2


b2 +

d2


z2 = 1

d1

d2


=

b2
2F

v2

The corresponding phase difference between the two paths, ejkl , will be then:

ejkl = ejv

2 /2

(18.14.20)

which has the same form as in the diffraction coefcient Das (v). The values v = un =

2n will make the path difference a multiple of /2, that is, l = n/2, resulting in the
alternating phase ejkl = (1)n .
The discrepancy between the choices vn and un arises from using D(v) to nd the alternating maxima, versus using the plain phase (18.14.20).



The Fresnel approximation is not invariant under shifting the origin. Our choice of
origin above is not convenient because it depends on the observation point P2 . If we
choose a xed origin, such as the point O in Fig. 18.14.4, then, we must determine the
corresponding Fresnel coefcient.
We assume that the points P1 , P2 lie on the yz plane and take P2 to lie in the shadow
region. The angles 1 , 2 may be chosen to be positive or negative to obtain all possible
locations of P1 , P2 relative to the screen.
The diffraction coefcient is still given by Eq. (18.13.8) but with r1 , r2 replaced by
the distances l1 , l2 . The unit vectors towards P1 and P2 are:
l1 =
sin 1 ,
z cos 1 y

l1 r = y sin 1 ,

z2 = 1

For v = u1 = 2, this denes the rst Fresnel zone ellipse, which gives the minimum
acceptable clearance for a given distance z:

b2

l2 =
sin 2
z cos 2 y

(18.14.21)

=
+ y y
and n
Since r = x x
z, we nd:

which can be rearranged into the equation of an ellipse:

b2
l2 = d22 + b2  d2 +
2d2

which leads to the following path length l, expressed in terms of v:

The corresponding values of D(v), shown on the gure with black dots, are given by

2vn

(18.14.19)

For example, for a distance of d = 1 km, using a cell phone frequency of f = 1 GHz,

corresponding to wavelength = 30 cm, we nd b = d/2 = 8.66 meters.

As can be seen in the above gure on the right, the diffraction coefcients D(v) and
Das (v) agree closely even for small values of v. Therefore, the extrema can be obtained
from the asymptotic form. They correspond to the values of v that cause the exponential
in (18.14.13) to take on its extremal values of 1, that is, the vs that satisfy v2 /2 + 1/4 = n,
with integer n, or:

vn = 2n 0.5 , n = 1, 2, . . .
(18.14.14)

Das (vn )= 1

1
d
2

(18.14.12)

and for positive and large clearance b, or equivalently, for large positive v,

Das (v)= 1

832

(18.14.18)

l2 r = y sin 2 ,

l1 = cos 1 ,
n

The quadratic approximation for the lengths R1 , R2 gives, then:

l2 = cos 2
n

18.14. Knife-Edge Diffraction

833

834

18. Radiation from Apertures

where the function D(v) was dened in Eq. (18.14.1). Putting all the factors together,
we may write the diffracted eld at the point P2 in the form:

ejkl2
E = Eedge
Dedge
l2

(straight-edge diffraction)

(18.14.23)

where we set ky02 /2F = v2 /2 and dened the incident eld Eedge at the edge and the
overall edge-diffraction coefcient Dedge by:

Eedge = A1

ejkl1
,
l1

Dedge =

FF
l2

cos 1 + cos 2
2


ejv

/2

D(v)

(18.14.24)

The second factor (ejkl2 / l2 ) in (18.14.23) may be interpreted as a cylindrical wave

emanating from the edge as a result of the incident eld Eedge . The third factor Dedge is
the angular gain of the cylindrical wave. The quantity v may be written as:

Fig. 18.14.4 Fresnel diffraction by straight edge.




(l1 r )2
(l2 r )2
(r r )
+

l1
l2
l1
l2


 2 
1
1 x
cos2 1
cos2 2 y2
+
+
+
= y (sin 1 + sin 2 )+
l1
l2
2
l1
l2
2

1
R1 + R2 l1 l2 = (l1 + l2 )r +
2


1 2
1  2
x +
y + 2F y (sin 1 + sin 2 )
2F
2F

1 2
1
1
x +
(y + y0 )2  y02
2F
2F
2F

where we dened the focal lengths F, F and the shift y0 :

1

l1

l2

F

cos2 1

l1

cos2 2


jkA1 ejk(l1 +l2 ) 
l2 )(n
l1 )
(n
4l1 l2

y0 = F (sin 1 + sin 2 )

(18.14.22)

Dedge =

2 /2Fjk(y +y

ejkx

2

0 ) /2F

dx dy

The x -integral is over the range <x < and can be converted to a Fresnel
integral with the change of variables u = x k/(F):

jkx2 /2F

dx =

F
k

ju2 /2

du =

e
0

jk(y +y0 )2 /2F

dy =

F
k


e
v

ju2 /2

du =

l1
(1 j)(cos 1 + cos 2 )

l1 + l2 4 k(sin 1 + sin 2 )

(18.14.26)

= 1 + 2 , as shown in Fig. 18.14.4 and using the trigonometric identity:



1 + 2
cos 1 + cos 2
= cot
sin 1 + sin 2
2
Then, Eq. (18.14.26) may be written in the form:

F
(1 j)
k

The y -integral is over the upper-half

of the xy-plane,that is, 0 y < . Dening


the Fresnel variables u = (y + y0 ) k/(F ) and v = y0 k/(F ), we nd:

(18.14.25)

This expression may be simplied further by dening the overall diffraction angle

ejk(R1 +R2 l1 l2 ) dS

2
jkA1 ek(l1 +l2 )

(cos 1 + cos 2 )ejky0 /2F
4l1 l2

kF
(sin 1 + sin 2 )

Depending on the sign and relative sizes of the angles 1 and 2 , it follows that
v > 0 when P2 lies in the shadow region, and v < 0 when it lies in the illuminated
region. For large positive v, we may use Eq. (18.14.5) to obtain the asymptotic form of
the edge-diffraction coefcient Dedge :


FF cos 1 + cos 2 jv2 /2 1 j jv2 /2
FF cos 1 + cos 2 1 j
e
Dedge =
e
=
l2
2
2v
l2
2
2v

Writing F/l2 = l1 /(l1 + l2 ) and replacing v from Eq. (18.14.25), the F factor

Using these approximations in Eq. (18.13.6) and replacing r1 , r2 by l1 , l2 , we nd:

E=

cancels and we obtain:

l2

v=

k
y0 =
F

F
(1 j)D(v)
k

Dedge =

l1
(1 j)

cot
2
l1 + l2 4 k

(18.14.27)

The asymptotic diffraction coefcient

 is obtained from Eqs. (18.14.26) or (18.14.27)
by taking the limit l1 , which gives l1 /(l1 + l2 ) 1. Thus,

Dedge =

(1 j)(cos 1 + cos 2 )
(1 j)

cot
=
2
4 k(sin 1 + sin 2 )
4 k

(18.14.28)

18.15. Geometrical Theory of Diffraction

835

Eqs. (18.14.27) and (18.14.28) are equivalent to those given in [1302].

The two choices for the origin lead to two different expressions for the diffracted
elds. However, the expressions agree near the forward direction,  0. It is easily
veried that both Eq. (18.14.1) and (18.14.27) lead to the same approximation for the
diffracted eld:

ejkl2
l1
1j

E = Eedge
(18.14.29)
l1 + l2 2 k
l

836

18. Radiation from Apertures

z Ez , and
Two polarizations may be considered: TE, in which the electric eld is E =
TM, which has H =
z Hz . Using cylindrical coordinates dened in Eq. (E.2) of Appendix
E, and setting /z = 0, Maxwells equations reduce in the two cases into:

(TE)

2 Ez + k2 Ez = 0,

H =

(TM)

2 Hz + k2 Hz = 0,

E =

18.15 Geometrical Theory of Diffraction

Geometrical theory of diffraction is an extension of geometrical optics [13981413]. It
views diffraction as a local edge effect. In addition to the ordinary rays of geometrical
optics, it postulates the existence of diffracted rays from edges. The diffracted rays
can reach into shadow regions, where geometrical optics fails.
An incident ray at an edge generates an innity of diffracted rays emanating from the
edge having different angular gains given by a diffraction coefcient Dedge . An example
of such a diffracted ray is given by Eq. (18.14.23).
The edge-diffraction coefcient Dedge depends on (a) the type of the incident wave,
such as plane wave, or spherical, (b) the type and local geometry of the edge, such as a
knife-edge or a wedge, and (c) the directions of the incident and diffracted rays.
The diffracted eld and coefcient are usually taken to be in their asymptotic forms,
like those of Eq. (18.15.25). The asymptotic forms are derived from certain exactly
solvable canonical problems, such as a conducting edge, a wedge, and so on.
The rst and most inuential of all such problems was Sommerfelds solution of a
plane wave incident on a conducting half-plane [1287], and we discuss it below.
Fig. 18.15.1 shows a plane wave incident at an angle on the conducting plane
occupying half of the xz-plane for x 0. The plane of incidence is taken to be the xyplane. Because of the cylindrical symmetry of the problem, we may assume that there
is no z-dependence and that the elds depend only on the cylindrical coordinates , .

1 1 Ez
,
j

1 Hz

j

H =

1 Ez
j

E =

Hz
j
1

(18.15.1)

where k2 = 2 , and the two-dimensional 2 is in cylindrical coordinates:

2 =


+

1 2
2 2

(18.15.2)

The boundary conditions require that the tangential electric eld be zero on both
sides of the conducting plane, that is, for = 0 and = 2. In the TE case, the
tangential electric eld is Ez , and in the TM case, Ex = E cos E sin = E =
(1/j)(Hz /), for = 0, 2. Thus, the boundary conditions are:
(TE)

Ez = 0,

for = 0 and = 2

(TM)

Hz
= 0,

for = 0 and = 2

(18.15.3)

In Fig. 18.15.1, we assume that 0 90o and distinguish three wedge regions
dened by the half-plane and the directions along the reected and transmitted rays:
reection region (AOB):
transmission region (BOC):
shadow region (COA):

0
+
+ 2

(18.15.4)

The case when 90o 180o is shown in Fig. 18.15.2, in which has been
redened to still be in the range 0 90o . The three wedge regions are now:
reection region (AOB):
transmission region (BOC):
shadow region (COA):

0
2
2 2

(18.15.5)

We construct the Sommerfeld solution in stages. We start by looking for solutions

of the Helmholtz equation 2 U + k2 U = 0 that have the factored form: U = ED, where
E is also a solution, but a simple one, such as that of the incident plane wave. Using the
differential identities of Appendix C, we have:



2 U + k2 U = D 2 E + k2 E + E2 D + 2 E D
Fig. 18.15.1 Plane wave incident on conducting half-plane.

Thus, the conditions 2 U + k2 U = 0 and 2 E + k2 E = 0 require:

E2 D + 2 E D = 0

ln E)
D = 0
2 D + 2(

(18.15.6)

18.15. Geometrical Theory of Diffraction

837

838

18. Radiation from Apertures

where f and v must be chosen to satisfy the four conditions:

2 f = 0,

f f = k2

2 v = 0,

2 f v

v v

= g (v)= v

(18.15.12)

It can be veried easily that the functions u = a cos a and u = a sin a are solutions of the two-dimensional Laplace equation 2 u = 0, for any value of the parameter
a. Taking f to be of the form f = Aa cos a, we have the condition:



sin a
cos a
f = Aaa1

This immediately implies that a = 1 and A2 = k2 , so that A = k. Thus, f =

A cos = k cos . Next, we choose v = Ba cos a. Then:

Fig. 18.15.2 Plane wave incident on conducting half-plane.

If we assume that E is of the form E = ejf , where f is a real-valued function, then,

equating to zero the real and imaginary parts of 2 E + k2 E = 0, we nd for f :



2 E + k2 E = E k2 f f + j2 f = 0

2 f = 0 ,

f f = k2 (18.15.7)

sin )
cos
f = A(


sin a
cos a
v = Baa1


f v = ABaa1 cos cos a + sin sin a = ABaa1 cos( a)

v v = B2 a2 2(a1)

Next, we assume that D is of the form:

v
D = D0

ejg(u) du

(18.15.8)

where D0 is a constant, v is a function of , , and g(u) is a real-valued function to be

determined. Noting that D = D0 ejg v and g = g (v) v, we nd:

D = D0 ejg v ,


2 D = D0 ejg 2 v jg (v) v v)

ln E)
D = 2 D + j
f D and:
Then, it follows from Eq. (18.15.6) that 2 D + 2(


f D = D0 ejg 2 v + j(2 f v g v v) = 0
2 D + j

Then, the last of the conditions (18.15.12) requires that:

1 2 f v
2A12a cos( a)
=
=1
v v v
aB2 cos a
which implies that a = 1/2 and B2 = 2A/a = 4A/. But since A = k, only the
case A = k is compatible with a real coefcient B. Thus, we have B2 = 4k/, or,

B = 2 k/.
In a similar fashion, we nd that if we take v = Ba sin a, then a = 1/2, but now

2
B = 4A/, requiring that A = k, and B = 2 k/. In summary, we have the
following solutions of the conditions (18.15.12):

Equating the real and imaginary parts to zero, we obtain the two conditions:
2

v = 0,

2 f v

v v

f = +k cos ,


= g (v)

D(v)=

1
1j

1j
1
+ F(v) =
2
1j

v

eju

/2

du

(18.15.10)

Therefore, we are led to choose g(u)= u2 /2 and D0 = 1/(1 j). To summarize,

we may construct a solution of the Helmholtz equation in the form:

2 U + k2 U = 0 ,

v = 2


(18.15.9)

Sommerfelds solution involves the Fresnel diffraction coefcient of Eq. (18.14.1),

which can be written as follows:

f f = A2 a2 2(a1) = k2

U = ED = ejf D(v)

(18.15.11)

f = k cos ,

v = 2

k 1/2

cos

(18.15.13)

k 1/2

sin

The corresponding solutions (18.15.11) of the Helmholtz equation are:


jk cos

U(, )= e

D(v) ,

v = 2


jk cos

U(, )= e

D(v) ,

v = 2

k 1/2

cos

2
k 1/2

sin

(18.15.14)

18.15. Geometrical Theory of Diffraction

839

The function D(v) may be replaced by the equivalent form of Eq. (18.14.6) in order
to bring out its asymptotic behavior for large v:



2
U(, )= ejk cos u(v)+d(v)ejv /2 ,


2
U(, )= ejk cos u(v)+d(v)ejv /2 ,

v = 2

v = 2

k 1/2

cos

2
k 1/2

sin

Using the trigonometric identities cos = 2 cos2 (/2)1 = 1 2 sin2 (/2), we

nd for the two choices of v:

k cos v2 = k cos 2 cos2

2
2

k cos

v = 2

U(, )= ejk cos u(v)+ejk d(v) ,

v = 2

For the TM case, we form the sum instead of the difference:



Hz = H0 ejk cos i D(vi )+ejk cos r D(vr )

= k

k 1/2

cos



Ez = E0 ejk cos i u(vi )E0 ejk cos r u(vr )+E0 ejk d(vi )d(vr )

The image of this electric eld with respect to the perfect conducting plane will
cos + y
sin ), resulting in
be the reected eld Er = E0 ejkr r , where kr = k(x
Er = E0 ejk cos(+) . The sum Ei + Er does vanish for = 0 and = 2, but it also
vanishes for = . Therefore, it is an appropriate solution for a full conducting plane
(the entire xz-plane), not for the half-plane.
Sommerfelds solution, which satises the correct boundary conditions, is obtained
by forming the linear combinations of the solutions of the type of Eq. (18.15.14):

Ez = E0 e

D(vi )e

D(vr )

(TE)

(18.15.17)

where

i = ,
r = + ,

k 1/2
i
cos

2

k 1/2
r
vr = 2
cos

vi > 0,
vi > 0,
vi < 0,

vr > 0
vr < 0
vr < 0

Ez = Ei + Er + Ed
Ez = Ei + Ed
Ez = Ed

reection region:
transmission region:
shadow region:

(18.15.21)

(18.15.22)

where we dened the incident, reected, and diffracted elds:

Ei = E0 ejk cos i
Er = E0 ejk cos r


Ed = E0 ejk d(vi )d(vr )

(18.15.16)

Ei = E0 ejkr = E0 ejk cos()

0 < ,
< < + ,
+ < 2,

The unit-step functions will be accordingly present or absent resulting in the following elds in these three regions:

k r = k(cos cos + sin sin )= k cos( )

jk cos r

(18.15.20)

(18.15.15)

k 1/2

sin

from which the Sommerfeld solution is built.

Considering the TE case rst, the incident plane wave in Fig. 18.15.1 is E =
z Ei ,
cos + y
sin and k = k(x
cos + y
sin ). It
where Ei = E0 ejkr , with r = x
follows that:

jk cos i

(18.15.19)

The boundary conditions (18.15.3) are satised by both the TE and TM solutions.
As we see below, the choice of the positive sign in the denitions of vi and vr was
required in order to produce the proper diffracted eld in the shadow region. Using the
alternative forms (18.15.15), we separate the terms of the solution as follows:

reection region:
transmission region:
shadow region:

Shifting the origin of the angle still leads to a solution. Indeed, dening  =
, we note the property / = /, which implies the invariance of the Laplace
operator under this change. The functions U(, ) are the elementary solutions

(TM)

The rst two terms correspond to the incident and reected elds. The third term is
the diffracted eld. The algebraic signs of vi and vr are as follows within the reection,
transmission, and shadow regions of Eq. (18.15.4):

Thus, an alternative form of Eq. (18.15.14) is:

U(, )= ejk cos u(v)+ejk d(v) ,

18. Radiation from Apertures

= k

v2 = k cos + 2 sin2
2
2

840

(18.15.23)

The diffracted eld is present in all three regions, and in particular it is the only one
in the shadow region. For large vi and vr (positive or negative), we may replace d(v) by
its asymptotic form d(v)= (1 j)/(2v) of Eq. (18.14.7), resulting in the asymptotic
diffracted eld:

Ed = E0 ejk

1j
2

vi

vr

1j

jk

= E0 e

22 k/1/2




1
cos(i /2)

cos(r /2)

which can be written in the form:

vi = 2

(18.15.18)

Ed = E0

ejk
Dedge
1/2

(18.15.24)

18.16. Problems

841

with an edge-diffraction coefcient:

1j

Dedge =

4 k

cos

i
2

cos

18. Radiation from Apertures

18.2 Prove the rst pair of equations for E, H of the previous problem by working exclusively
with the Kottler formulas (18.4.2) and taking their far-eld limits.

842

(18.15.25)

18.3 Explain in detail how the inequality (18.6.12) for the aperture efciency ea may be thought
of as an example of the Schwarz inequality. Then, using standard properties of Schwarz
inequalities, prove that the maximum of ea is unity and is achieved for uniform apertures.
As a reminder, the Schwarz inequality for single-variable complex-valued functions is:

2 


b

b
b

f (x)g(x) dx

|f (x)|2 dx
|g(x)|2 dx

a
a
a

Using a trigonometric identity, we may write Dedge as follows:

1j

Dedge =

4 k

1
cos

cos

1
1 j sin 2 sin 2
=
+
k cos + cos

(18.15.26)

Eqs. (18.15.22) and (18.15.24) capture the essence of the geometrical theory of diffraction: In addition to the ordinary incident and reected geometric optics rays, one also
has diffracted rays in all directions corresponding to a cylindrical wave emanating from
the edge with a directional gain of Dedge .
For the case of Fig. 18.15.2, the incident and reected plane waves have propagation
sin ) and kr = k(
sin ). These correspond to
vectors k = k(
z cos + y
z cos y
the incident and reected elds:

Ei = E0 ejkr = E0 ejk cos(+) ,

18.4 To prove the equivalence of the Kirchhoff diffraction and Stratton-Chu formulas, (18.10.6)
and (18.10.7), use the identities (C.29) and (C.32) of Appendix C, to obtain:

 
 



1
j G J + G  + G  J m dV =
j G J  G + J m  G dV


V
V
 


G+n
J m G dS

n

S
Then, using the identity (C.33), show that Eq. (18.10.6) can be rewritten in the form:

 


j G J  G + J m  G dV

V
 


G+n
J m G dS
n
+

S



 G G n
 E)(n
E)
(
E)  G dS
G  E (n
n

E(r)=

Er = E0 ejkr r = E0 ejk cos()

In this case, the Sommerfeld TE and TM solutions take the form:



Ez = E0 ejk cos i D(vi )ejk cos r D(vr )


Hz = H0 ejk cos i D(vi )+ejk cos r D(vr )
where, now:

(18.15.27)

Finally, use / =  E and  E + J m = jH to obtain (18.10.7).

i = + ,
r = ,

k 1/2
i
sin
vi = 2

2

k 1/2
r
sin
vr = 2

18.5 Prove the equivalence of the Stratton-Chu and Kottler formulas, (18.10.7) and (18.10.10), by
rst proving and then using the following dual relationships:

18.16 Problems
18.1 Show that Eq. (18.4.9) can be written in the compact vectorial form:



ejkr

r
r (
z f
z g) ,
4r

H=



jk ejkr

r
r (
z f)+
zg
4r

Similarly, show that Eqs. (18.4.10) and (18.4.11) can be written as:



ejkr

r
zf ,
4r


ejkr

r
r (
E = 2jk
z g) ,
4r

E = 2jk

H=



2jk ejkr

r
r (
z f)
4r

H = 2jk



ejkr

r
zg
4r

(18.15.28)

The choice of signs in vi and vr are such that they are both negative within the
shadow region dened by Eq. (18.15.5). The same solution can also be obtained from
Fig. 18.15.1 and Eq. (18.15.17) by replacing by .

E = jk


V


 G dV =
 G (J  )
j



  G j(n
 G
H)
E)
(n


 G dV =
jm  G (J m  )



  G + j(n
 G
E)
H)
(n


S

To prove these, work component-wise, use Maxwells equations (18.2.1), and apply the divergence theorem on the volume V of Fig. 18.10.1.
18.6 Prove the equivalence of the Kottler and Franz formulas, (18.10.10) and (18.10.11), by using
A)= (
A)2 A, and by replacing the quantity k2 G(r r ) by
the identity (
(3) (r r )2 G. Argue that the term (3) (r r ) makes a difference only for the volume
integrals, but not for the surface integrals.
18.7 Prove the equivalence of the modied Stratton-Chu and Kirchhoff diffraction integral formulas of Eq. (18.12.1) and (18.12.2) by using the identity (C.42) of Appendix C and replacing
 E = 0 and  E = jH in the source-less region under consideration.
18.8 Prove the equivalence of the Kottler and modied Stratton-Chu formulas of Eq. (18.12.1) and
(18.12.2) by subtracting the two expressions, replacing jE =  H , and using the Stokes
identity (C.38) of Appendix C.

18.16. Problems

843

18.9 Consider a reector antenna fed by a horn, as shown

on the right. A closed surface S = Sr + Sa is such
that the portion Sr caps the reector and the portion
Sa is an aperture in front of the reector. The feed
lies outside the closed surface, so that the volume V
enclosed by S is free of current sources.

19
Diffraction Plane-Wave Spectrum

Applying the Kottler version of the extinction theorem of Sec. 18.10 on the volume V, show
that for points r outside V, the eld radiated by the induced surface currents on the reector
Sr is equal to the eld radiated by the aperture elds on Sa , that is,
E rad (r) =

j
1

j


Sr

Sa




k2 G J s + J s   G dS



 2

  G + j(n
 G dS
H )+ (n
H )
E )
k G(n

r H and J ms = n
r E, and
where the induced surface currents on the reector are J s = n
on the perfectly conducting reector surface, we must have J ms = 0.
This result establishes the equivalence of the so-called aperture-eld and current-distribution
methods for reector antennas [1686]. See also Sec. 21.9.
18.10 Consider an x-polarized uniform plane wave incident obliquely on the straight-edge aperture
1 =
sin 1 . First show that the
of Fig. 18.14.4, with a wave vector direction k
z cos 1 + y
+ y y
on the aperture above the straight-edge
tangential elds at an aperture point r = x x
are given by:
E0 ejky
Ea = x

 sin

Ha = y

E0

cos 1 ejky sin 1
0

Then, using Kottlers formula (18.12.1), and applying the usual Fresnel approximations in
the integrand, as was done for the point source in Fig. 18.14.4, show that the diffracted
wave below the edge is given by Eqs. (18.14.23)(18.14.25), except that the eld at the edge
is Eedge = E0 , and the focal lengths are in this case F = l2 and F = l2 / cos2 2
Finally, show that the asymptotic diffracted eld (when l2 ), is given near the forward
direction  0 by:
ejkl2 1 j

E = Eedge

l2 2 k

18.11 Assume that the edge in the previous problem is a perfectly conducting screen. Using the
eld-equivalence principle with effective current densities on the aperture above the edge
J s = 0 and J ms = 2
n E a , and applying the usual Fresnel approximations, show that the
diffracted eld calculated by Eq. (18.4.1) is still given by Eqs. (18.14.23)(18.14.25), except
that the factor cos 1 + cos 2 is replaced now by 2 cos 2 , and that the asymptotic eld and
edge-diffraction coefcient are:

ejkl2
E = E0
Dedge ,
l2

Dedge =

(1 j)2 cos 2

4 k(sin 1 + sin 2 )

This chapter continues the previous one on radiation from apertures. The emphasis
is on Rayleigh-Sommerfeld diffraction theory, plane-wave spectrum representation for
scalar and vector elds, radiated and reactive power of apertures, integral equations
for apertures in conducting screens, revisiting the Sommerfeld half-plane problem using Wiener-Hopf factorization techniques, the Bethe-Bouwkamp model of diffraction by
small holes, and the Babinet principle.

19.1 Rayleigh-Sommerfeld Diffraction Theory

In this section, we recast Kirchhoffs diffraction formula in a form that uses a Dirichlet Greens function (i.e., one that vanishes on the boundary surface) and obtain the
Rayleigh-Sommerfeld diffraction formula. In the next section, we show that this reformulation is equivalent to the plane-wave spectrum approach to diffraction, and use
it to justify the modied forms (18.1.2) and (18.1.3) of the eld equivalence principle.
In Chap. 20, we use it to obtain the usual Fresnel and Fraunhofer approximations and
discuss a few applications from Fourier optics.
We will work with the scalar case, but the same method can be used for the vector
case. With reference to Fig. 19.1.1, we consider a scalar eld E(r) that satises the
source-free Helmholtz equation, (2 + k2 )E(r)= 0, over the right half-space z 0.
We consider a closed surface consisting of the surface S of a sphere of very large
radius centered at the observation point r and bounded on the left by its intersection S
with the xy plane, as shown in the Fig. 19.1.1. Clearly, in the limit of innite radius, the
volume V bounded by S + S is the right half-space z 0, and S becomes the entire xy
plane. Applying Eq. (18.10.3) to volume V, we have:


V

Show that this expression agrees with the exact Sommerfeld solution (18.15.26) at normal
incidence and near the forward diffracted direction.


G(2 E + k2 E)E (2 G + k2 G) dV =



G
E

E
G
dS (19.1.1)
n
n
S+S

is the negative of the

The surface integral over S can be ignored by noting that n
radial unit vector and therefore, we have after adding and subtracting the term jkEG:


G

  




G
E
E
G
E
dS =
G
+ jkE E
+ jkG dS


n
n
r
r
S